Development and Application of
13
C-Labeling Techniques
Analyzing the Pentose Phosphate Pathway of Penicillium chrysogenum
Roelco Kleijn
Development and Application of
13
C-Labeling Techniques
Analyzing the Pentose Phosphate Pathway of Penicillium chrysogenum
Proefschrift
ter verkrijging van de graad van doctor
aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof. dr. ir J.T. Fokkema,
voorzitter van het College voor Promoties
in het openbaar te verdedigen op maandag 2 april 2007 te 15:00 uur
door
Roelof Jacobus KLEIJN
ingenieur in de Bioprocestechnologie
geboren te Vlaardingen
Dit proefschrift is goedgekeurd door de promotor:
Prof. dr. J.J. Heijnen
Samenstelling promotiecommissie:
Rector Magnificus
Voorzitter
Prof. dr. J.J. Heijnen
Technische Universiteit Delft, promotor
Prof. dr. J.T. Pronk
Technische Universiteit Delft
Prof. dr. J. Nielsen
Technical University of Denmark, Lyngby, Denmark
Prof. dr. E. Heinzle
Saarland University, Saarbrücken, Germany
Dr. W.A. van Winden
Technische Universiteit Delft
Dr. D. Schipper
DSM
Prof. dr. J.H. de Winde
Technische Universiteit Delft, reservelid
The studies performed in this thesis were performed at the Bioprocess Technology section,
Department of Biotechnology, Delft University of Technology. The research was financially
supported by the Dutch Ministry of Economic Affairs via the EET program (Project No.
EETK20002) and DSM.
If facts are the seeds that later produce knowledge and wisdom, then the emotions and the impressions of the senses are the fertile soil in which the seeds
must grow.
- Rachel Carson
Table of Contents
List of abbreviations
ix
Chapter 1
General Introduction
1
Chapter 2
Revisiting the 13C-label distribution of the non-oxidative
branch of the pentose phosphate pathway based upon
kinetic and genetic evidence
29
Chapter 3
Metabolic flux analysis of a glycerol-overproducing
Saccharomyces cerevisiae strain based on GC-MS, LCMS and NMR derived 13C-labeling data
53
Chapter 4
13
C-labeling based metabolic network and flux analysis
of penicillin-G producing and non-producing chemostat
cultures of Penicillium chrysogenum
81
Chapter 5
13
C-labeled gluconate tracing as a direct and accurate
method for determining the pentose phosphate pathway
split-ratio in Penicillium chrysogenum
129
Chapter 6
Cytosolic NADPH metabolism in penicillin-G producing
and non producing chemostat cultures of Penicillium
chrysogenum
157
Chapter 7
Discussion and future directions
179
Summary
187
Samenvatting
191
List of Publications
195
Curriculum Vitae
197
Dankwoord
199
vii
Abbreviations
A list of the most common abbreviations used throughout this thesis:
1,3PG
2/3PG
2PG
3PG
6PG
8-HPA
Aald
AcCoA
ACV
ADP
AKG
AMP
ASP
AT
ATP
AVCS
C1
CID
CITR
CoA
COSY
DHAP
DSH
E4P
E-C2
E-C3
ETOH
F6P
FAD(H)
FBP
FE
FUM
G1P
G3P
G6P
GAP
GC-MS
GLC
GLN
GLY
1,3-Bisphosphoglycerate
Combined pool of 2PG and 3PG
2-Phosphoglycerate
3-Phosphoglycerate
6-Phosphogluconate
8-Hydroxypenillic acid
Acetaldehyde
Acetyl-Coenzyme A
δ‑(L‑α-Aminoadipyl)-L-cysteinyl-D-valine
Adenosine diphosphate
α-Ketogluterate
Adenosine monophosphate
Aspertate
Acyl-coA:IPN acyl transferase
Adenosine triphosphate
Non-ribosomal ACV synthetase
Methylenetetrahydrofolate
Collisionally induced dissociation
Citrate
Coenzyme A
Correlation spectroscopy
Dihydroxyacetone-phosphate
Direct sulfhydrylation
Erythrose-4-phosphate
Glycolaldehyde moiety covalently bound to the thiamine
pyrophosphate/transketolase complex
Dihydroxyacetone moiety covalently bound to the enzyme
transaldolase
Ethanol
Fructose-6-phosphate
Flavin adenine dinucleotide
Fructose-1,6-bisphosphate
Fractional enrichment
Fumerate
Glucose-1-phosphate
Glycerol-3-phosphate
Glucose-6-phosphate
Glyceraldehyde-3-phosphate
Gas chromatography mass spectrometry
Glucose
Gluconate
Glycine
ix
Abbreviations
GLYOX
GOH
IPN
IPNS
LC-MS
M6P
MAL
MDV
MFA
MG
NAD(H)
NADP(H)
NMR
O8P
OAA
o-OH-PAA
OPC
P5P
PAA
PEP
PIO
POA
PPP
PYR
R5P
RQ
RU5P
S7P
s2res
SER
SFL
SS
SUC
T6P
TA
TCA
THR
TK
TPP
TS
X5P
Glyoxylate
Glycerol
Isopenicillin-N
IPN synthetase
Liquid chromatography mass spectrometry
Mannose-6-phosphate
Malate
Mass isotopomer distribution vector
Metabolic flux analysis
Methylgyoxal
Nicotinamide dinucleotide
Nicotinamide dinucleotide phosphate
Nuclear magnetic resonance
Octulose-8-phosphate
Oxaloacetate
Ortho-hydroxyphenylacetic acid
6-Oxopiperidine-2-carboxylic acid
Combined pool of R5P, X5P and RU5P
Phenylacetic acid
Phosphoenolpyruvate
Penicilloic acid
Phenocyacetic acid
Pentose phosphate pathway
Pyruvate
Ribose-5-phosphate
Respiratory quotient
Ribulose-5-phosphate
Sedoheptulose-7-phosphate
Estimated error variance
Serine
Summed fractional labeling
Summed squared residuals
Succinate
Trehalose-6-phosphate
Transaldolase
Tri carboxylic acid
Threonine
Transketolase
Thiamine pyrophosphate
Transsulfuration
Xylulose-5-phosphate
CHAPTER
General introduction
1
Chapter 1
1.1 History of Penicillin Production
1.1.1From discovery to production
Many lives have been saved by the development of antibiotics for the treatment of
infectious diseases. In that sense, the discovery of antibiotics is perhaps the most important
breakthrough in the history of therapeutic medicine. This discovery is generally attributed to
Alexander Fleming, who worked at the St. Mary's Medical School at London University and
in 1929 published an article in which he showed that a species of Penicillium (later identified
as Penicillium chrysogenum) exhibited strong antibacterial activity towards gram-positive
bacteria [21]. Although Fleming was not the first to report on the antibacterial properties of
moulds and fungi, he was the first to recognize the importance of his findings [29]. For this
reason he was one of the recipients of the Nobel price for medicine in 1945.
In the years following the 1929 publication of Fleming, his results did not evoke much interest
outside scientific circles and was merely seen as a scientific curiosity. In fact, at that time it
was not considered important that basic research should lead to a practical application. By
1932, Fleming had abandoned his work on the antibacterial agent (which he called penicillin),
also because his efforts to stabilize and purify penicillin were to no avail. Almost a decade
later Howard Florey and Ernst Chain at the Sir William Dunn School of Pathology at Oxford
University were the first to recognize the importance of Fleming’s 1929 publication. Thanks to
the pioneering work of Norman Heatley, an extraction procedure based upon a solvent/water
mixture was developed which enabled the production and extraction of enough penicillin for
clinical trials [38]. After successfully showing the chemotherapeutic action of penicillin in mice
injected with a lethal dose of bacteria [5], the first dose of penicillin was administered to a
human on January 27, 1941. Ensuing studies demonstrated that injections of penicillin caused
rapid recoveries in patients suffering from a variety of infections [1].
From 1941 onwards it became clear that if available in sufficient quantity, penicillin had an
enormous potential as antibacterial agent. Studies demonstrated that penicillin inhibited the
growth of pathogens such as gas gangrene and syphilis, which were widespread due to the
outbreak of World War II. However, due to war conditions the British pharmaceutical industry
was incapable of producing sufficient penicillin as materials needed for the production of
penicillin were limited. For this reason, Florey and Heatley headed for the United States on
June 27, 1941 to urge pharmaceutical companies to enter into mass production.
Initially, the US pharmaceutical industry was hesitant to invest in the development of expensive
fermentation plants, as they feared that penicillin would soon be chemically synthesized. To
this end, studies on the structure of penicillin were initiated in 1941. However, resolving the
chemical structure proved to be difficult and it soon became clear that the production of penicillin
could perhaps be accomplished more easily by fermentation. In 1943 four companies took the
lead in developing pilot plants and new methods for the mass production of penicillin: Merck &
Co., E.R. Squibb & Sons, Chas. Pfizer & Co and Abbott Laboratories. On March 1, 1944 Pfizer
opened the first commercial plant for the large-scale production of penicillin by submerged
fermentation. Until today, penicillin has always been produced in a fermentative process by
P. chrysogenum, despite numerous attempts to synthesize it chemically. In fact it would take
until 1957 before a procedure was developed for the complete chemical synthesis of penicillin.
Penicillin V was the only useful derivative made by this procedure, and even this was easier
Introduction
produced by fermentation [29]. Aside from being easier, the fermentative approach remains
cheaper and more environmental friendly, making the industrial production of penicillin by
fermentation one of the outstanding examples of biotechnology.
1.1.2
Strain improvement techniques
In the 60 years following the initial commercial production of penicillin, its biosynthesis
developed from a specialty process yielding gram scale quantities, to a bulk process with
a world-wide annual production exceeding 60.000 tons [2, 53]. Prices have dropped from a
staggering $83.000/kg in 1943 to less than $15/kg nowadays [2, 16, 29, 46]. The penicillin titers
of P. chrysogenum have increased accordingly from less than 0.003 g/l for the strain isolated
by Fleming in 1929, to 40-50 g/L nowadays [14, 16, 29] (See Figure 1). This increase in titer is
partly the result of more optimal cultivation conditions and improved fermentation techniques,
but above all due to an ongoing quest for superior strains. Numerous strain improvement
programs were initiated to increase the productivity and fitness of P. chrysogenum via classical
mutagenesis techniques (e.g. via irradiation and chemical agents) and subsequent screening
for better producing strains.
120
50
Estimate
40
80
30
60
Estimate
20
40
10
20
NRLL1951
1940
Penicillin-G titers (g/L)
Penicillin-G price ($/kg)
100
Lilly-E15.1
WisQ-176
Panlabs-8
1960
1980
2000
0
year
Figure 1 Penicillin prices [2, 29, 46] and titers [16, 23, 29, 46] from the start of commercial production
until today. Titers are listed together with the P. chrysogenum strain in which the titer was attained.
From 1980 onwards only estimates for titers are available. Arrows denote the corresponding y-axis
for the two curves.
An inherent disadvantage of classical mutagenesis is its random nature. Besides beneficial
mutations resulting in increased product titers, the improved strain will also accumulate many
undefined and undesirable mutations. In addition, the process is labour and resource intensive
due to the many mutants that have to be created and screened before a robust and higher
Chapter 1
producing strain is isolated. Conversely, the development of recombinant DNA technology
over the past decades has allowed researchers to introduce precise modifications in the
genetic makeup of a cell, thereby redesigning only selected parts of the metabolism. This
more rational approach towards strain improvement is referred to as metabolic engineering
and increases the chance of successfully directing the metabolic fluxes towards the desired
end product. In short, metabolic engineering consists of a first metabolic characterization step
in which more insight in the cellular metabolism of the studied strain is obtained, followed by
a genetic modification step using recombinant DNA technology, in which this insight is used
to improve the properties of the strain. For lack of a complete understanding of the cell, this
is a cyclic process; multiple steps of characterization and modification are needed to further
improve the strain [60].
BOX 1: Taxonomy and morphology of P. chrysogenum
Penicillium is a genus of filamentous fungi belonging to the division of ascomycota (sac fungi).
Characteristic of ascomycota is that they produce spores in a distinctive compartment called the
ascus. Species of Penicillium are recognized by their dense brush-like spore-bearing structures.
This is also where the name Penicillium comes from (brush is penicillus in Latin). The Penicillium
species studied in this thesis, Penicillium chrysogenum (also known as Penicillium notatum), is the
main production host of penicillin (see main text). Penicillin prevents the growth of gram-positive
bacteria by binding to the peptidoglycan crosslinking enzyme transpeptidase, thereby disrupting
bacterial cell wall synthesis. As a result, the formation of crosslinks between peptidoglycan polymers
is inhibited, making the affected cells more susceptible to lysis during cell division.
Like with all fungi, spores form the start (germination)
and endpoint (sporulation) of the developmental phase of
Apical
P. chrysogenum. Spore germination can be divided into
three stages: spore swelling, germtube emergence and
Subapical
germtube elongation [49]. During spore swelling, the spore
grows spherically and new cell material is produced. After
Hyphal
establishing growth polarity, germtubes emerge from the
spore at one or two areas. During the elongation phase
the germtube grows and eventually develops into an
elongated filament consisting of thin, needle-shaped, multicellular structures called a hypha. Growth of the hypha is
initially supported by storage compounds in the spore, but
as the hypha extends it grows on nutrients taken up from
the medium [51]. Hyphal growth is highly polarized and
occurs only at the hyphal tip (apex). Periodically branches
are formed at or near the hyphal tip, allowing the fungus to
form densely branched hyphal elements [25, 48].
Spore
The morphologic differentiation of a spore to a hypha leads
to different cell-types. In general, a hypha can be divided
into three distinct compartments each with its own function:
Figure 2 Structure of a hyphal
the apical, the subapical and the hyphal compartment
element emerging from a spore.
(Figure 2) [46]. The apical compartment is situated at the
Shown are the apical, subapical
tip of the hypha, contains no cross-walls (septa) and is
and the hyphal compartment.
directly involved in the supply of cellular material necessary
Septa in the subapical and hyphal
for tip extension (e.g. cytoplasmic material and building
compartments are also shown.
blocks for wall synthesis). The subapical compartment is
Introduction
located further down the hypha. The cells in this compartment are divided by septa, but a (free) flow
of nutrients and organelles from one compartment to another is still possible. Most probably, this
compartment also plays a role in supplying the hyphal tip with cellular building blocks. The hyphal
compartment is found at the basis of the hypha, containing cells with large vacuoles [49, 50]. These
cells are believed to be important for creating a sufficient intracellular pressure to ensure transport
of protoplasm towards the tip section.
In contrast to unicellular eukaryotes the two processes of cell division (genome duplication and
cell division) do not necessarily occur consecutively in fungal cell. As a result, fungal cells contain
multiple nuclei. Formation of a new apical compartment consists of three stages: (i) the length of
the apical compartment is increased until the volume of cytoplasm per nucleus reaches a critical
ratio (ii) genome duplication (mitosis) is initiated and continued until the number of nuclei has
doubled (iii) the compartment is divided by a septum, forming a new apical compartment [46].
Aside from the morphological differentiation of a single hypha (microscopic morphology) the
interaction of multiple hyphae with each other also causes morphological differentiation at a higher
level (macroscopic morphology). Within submerged cultures the interaction of hyphae can lead
to distinct particles (pellets), connected networks of hyphae (aggregates), or separate hyphae
(dispersed mycelia) [11]. Factors which affect the macroscopic morphology include the level and
type of inoculum, medium shear, medium constituents and the pH [51].
1.2 Metabolic Engineering and P. chrysogenum
1.2.1
Production pathway
Metabolic engineering of P. chrysogenum requires fundamental knowledge on the biosynthesis
pathway leading to the formation of penicillin (see Figure 3). The first step in the penicillin
biosynthesis pathway is the condensation of the three amino acid precursors (L-valine, Lα-aminoadipic acid and L-cysteine) into the tripeptide δ‑(L‑α-aminoadipyl)-L-cysteinyl-Dvaline (LLD-ACV) by the non-ribosomal peptide synthetase ACV-synthetase. In the second
step isopenicillin N synthetase (IPNS) catalyzes the oxidative ring closure of the tripeptide,
resulting in a four-membered β-lactam ring fused to a five-membered thiazolidine which is
characteristic for all penicillins. In the third and final step the L-α-aminoadipyl side chain of
IPN is exchanged for a CoA-activated acyl group by acyl-CoA:isopenicillin-N-acyl transferase
(AT). Strictly speaking L-α-aminoadipic acid is not a precursor as it is split-off and (partially)
recycled after the formation of the β-lactam nucleus. The type of penicillin produced by P.
chrysogenum depends on the acyl side-chain precursor added to the medium. Commonly
used side-chain precursors are phenylacetic acid (PAA) and phenoxyacetic acid (POA),
leading to the formation of penicillin‑G and penicillin‑V, respectively. Aside from the above
described main reaction pathway, several spontaneous chemical conversions lead to the
(irreversible) formation of penicillin side-products such as ortho-hydroxyphenylacetic acid (oOH-PAA), 6-oxopiperidine-2-carboxylic acid (OPC), 8-hydroxypenillic acid (8-HPA), penicilloic
acid (PIO) and bis-ACV [17, 30].
In P. chrysogenum the genes encoding for ACVS, IPNS and AT (pcbAB, pcbC, penDE) are
arranged in a single gene cluster [42]. As a result, the most straightforward way to increase
the productivity of β-lactams would be to over-express these gene clusters. Indeed, several
studies have shown that over-expression of these three clustered genes in the Wis54-1255
strain (the ancestor of most present day industrial penicillin producing strains) leads to an
Chapter 1
H 2N
H 2N
COOH
HOOC
HOOC
L-Aad
H 2O
HOOC
L-Cys
3ATP
COOH
CH3
CH3
L-Val
bis-ACV
ACVS
3AMP
NH
HOOC
O
SH CH
3
1/
2 O2
H2O
O
IPNS
2H 2O
H
N
H2 N
O
O
CH3
S
CH3
N
O H
CH3
COOH
IPN
PAA
H2 N
HOOC
COOH
COOH
Aad
ATP + CoA
PCL
AT
H 2O + AMP
CoA
PAA-CoA
H2 N
S
6-APA
O
N
H
N
CH3
CH3
COOH
O
AT
H 2O
CoA
N
HO
HOOC
O
CO 2
COOH
S
N
8-HPA
H2 N
AT
CH3
CH3
COOH
O
S
N
H
N
OO C N
OH H
S
COOH
CH3
S
CH3
S
N
O H
N
N
H
H
N
H2N
CH3
H2 N
+
COOH NADP NADPH
COOH
TR
N
O H
O2
HOOC
O
COOH
H
N
H 2N
LLD-ACV
O
OPC
H2N
SH
COOH
Aad + CoA
CH3
Penicillin-G
CH3
COOH
CH3
CH3
COOH
PIO
Figure 3 Biosynthesis pathway of penicillin-G and the related by-products. For other abbreviations see
text.
CH3
COOH
Introduction
increased penicillin production rate [59, 70]. Furthermore, Fierro et al. [18] showed that
classical strain improvement has led to a 6-14 fold increase in the copy number of the gene
cluster in the nowadays available high penicillin producing strains compared to the Wis541255 strain. For obvious reasons there are no public reports on the number of gene clusters
in the currently used industrial strains.
1.2.2
Primary metabolism
There is a limit to increasing penicillin production via the increased amplification of the
penicillin gene cluster. At some point the supply of carbon precursors, cofactors and energy
by the primary metabolism will limit penicillin production, causing a shift in the metabolic
bottleneck from the production pathway to the primary metabolism [27, 63]. For example,
Jorgensen et al. [31] observed that supplementing fed-batch cultivations with cysteine, valine
and α-aminoadipic acid increased the penicillin-V production by about 20%, indicating that
in the studied strain amino acid availability may limit penicillin production. In general, three
possible limitations in the primary metabolism of P. chrysogenum can be identified when
synthesizing large amounts of penicillins. These are the supply of the carbon precursors; the
supply of energy in the form of ATP; and the availability of electrons in the form of NADPH.
All these three constituents of primary metabolism are needed to construct the three amino
acid precursors for penicillin synthesis. In addition, ATP and NADPH are essential cofactors
for the conversion of the precursors to the product penicillin Several studies have addressed
one or more of these potential bottlenecks in high-yielding former production strains of P.
chrysogenum [9, 10, 28, 31, 62, 63].
Supply of carbon precursor The supply of sufficient carbon precursors for the synthesis
of penicillin is an obvious target when optimizing penicillin production. Van Gulik et al. [63]
showed in a chemostat cultivation that flux distributions around the four principal metabolite
nodes of penicillin production in P. chrysogenum (glucose-6-phosphate, glyceraldehyde-3phosphate, mitochondrial pyruvate and mitochondrial isocitrate) are highly flexible. It was
therefore considered unlikely that in the studied industrial strain the primary carbon metabolism
forms a potential bottleneck in penicillin-G synthesis.
Supply of energy The synthesis of penicillin is an energy demanding process. The ATP
consumption associated with penicillin production from glucose was experimentally determined
by van Gulik et al. [62] to be 73 mol ATP/mol penicillin-G, which corresponded with a theoretical
maximum yield of penicillin-G on glucose of 0.18 mol/mol glucose. This value is much lower
than the calculated values (0.43-0.50 mol/mol glucose) based on the known ATP stoichiometry
of the penicillin biosynthesis pathway [31]. If the value reported by van Gulik et al. [62] is
correct a considerable amount of ATP is consumed by as yet unknown and unaccounted
processes which are related to penicillin synthesis. Possible explanations for this increased
ATP demand are high turnover rates for the enzymes involved in penicillin biosynthesis, ATPdriven secretion systems [43] and/or the hydrolysis of penicillin intermediates. Kallow et al.
[32], for example, showed that hydrolysis of the tripeptide LLD-ACV may result in an actual
requirement of 20 mol ATP per mole of tripeptide formed, while the stoichiometric energy
requirement is only three mol ATP.
Chapter 1
Supply of reducing power Reducing power in the form of NADPH is important for the
biosynthesis of the two amino acid precursors of the β-lactam nucleus, cysteine and valine.
The total stoichiometric NADPH demand for penicillin synthesis ranges from 7-10 mol/mol
penicillin [63]. However, similar to the ATP demand for penicillin synthesis, the actual value
can be several times higher as a result of unaccounted processes (e.g. the spontaneous
oxidation of ACV into bis-ACV, which is reduced back to ACV by thioredoxin reductase at the
cost of one NADPH equivalent).
The pentose phosphate pathway (PPP) is the main producer of NADPH in P. chrysogenum.
Therefore, an important metabolic network parameter is the fraction of g6p entering the
oxidative branch of the PPP in relation to the total uptake of glucose by the cell (from hereon
referred to as the PPP split-ratio). In general, an increased demand for NADPH is associated
with an increase in the PPP split-ratio [74, 79]. In the past, several studies quantified the relation
between the PPP split-ratio and the overproduction of (partly) PPP-originating products [12, 78]
or amino acids requiring considerable amount of NADPH for their biosynthesis [54]. Several
theoretical studies were also performed on P. chrysogenum in which it was speculated that the
PPP split-ratio was strongly correlated to penicillin production [28, 31]. This hypothesis was
supported by the results of van Gulik et al. [63], who showed that a stepwise increase in the
total metabolic demand for NADPH (by altering between glucose, acetate, ethanol or xylose
as carbon source and ammonia or nitrate as nitrogen source) resulted in a stepwise decrease
in penicillin-G production. On the other hand, experimental 13C-based flux determinations by
Christensen et al. [10] in a slightly different P. chrysogenum strain showed no relation between
penicillin production and the flux through the oxidative PPP.
1.3 Metabolic flux analysis
1.3.1
Principle
At the basis of investigating the central metabolism of a micro-organism lies a tool called
metabolic flux analysis (MFA). MFA, also referred to as metabolite balancing, is a generally
applicable characterization method within the field of metabolic engineering, aimed at quantifying
intracellular steady state fluxes within a predefined metabolic network model. Via this metabolic
snapshot researchers are able to better understand and predict the phenotypic behaviour
of a micro-organism as a result of genetic alterations [15, 57] and different environmental
conditions [13, 34, 63] Moreover, an accurate description of the metabolic network and the
corresponding metabolic fluxes form the basis of kinetic models for in silico network analysis.
With these models researchers can predict how changes in the underlying metabolic network
(e.g. changed enzyme levels or enzyme properties as a result of genetic modification) affect
the fluxes in the cell, making these a powerful tool for metabolic engineering.
1.3.2
Experimental framework
A prequisite for determining the flux distribution in a cell is that the intracellular metabolite
concentrations are in pseudo steady state. Studied micro-organisms should thus be cultivated
in controlled bioreactors operated in batch, fed batch or continuous mode at a constant
growth-rate, dissolved oxygen tension, pH and temperature. In batch cultures the steady-
Introduction
BOX 2: The pentose phosphate pathway
The pentose phosphate pathway (PPP) plays a crucial role in supplying the cell with precursors
for amino acid and nucleotide biosynthesis and maintaining the NADP+/NADPH balance. Note
that in order to maintain the NADPH balance the flux through the PPP is generally much larger
than the drain on PPP metabolites for the biosynthesis of building blocks. Figure 4 shows the two
parts of the PPP: i) the oxidative branch for the production of NADPH and ii) the non-oxidative
branch for the synthesis of biomass precursors such as erythrose-4-phosphate (e4p) and ribose5-phosphate (r5p), and for the recycling of surplus carbon back to glycolysis. Other metabolites
within the PPP are glucose-6-phosphate (g6p), 6-phosphogluconate (6pg), ribulose-5-phosphate
(ru5p), xylulose-5-phosphate (x5p), sedoheptulose-7-phosphate (s7p), fructose-6-phosphate (f6p)
and glyceraldehyde-3-phosphate (gap). Neglecting the withdrawal of precursors for amino acid and
nucleotide synthesis, the overall reactions of the oxidative and the non-oxidative branch are:
3 g6p + 3H2O + 6NADP+ → 3ru5p + 3CO2 + 6NADPH + 6H+
(1.1)
3ru5p ↔ 2r5p + x5p ↔ 2 f6p + gap
(1.2)
In the oxidative branch g6p is irreversibly oxidized and hydrolyzed to 6pg by the enzyme g6p
dehydrogenase and gluconolactonase, respectively. The formed 6pg is oxidatively decarboxylated
by the enzyme 6pg dehydrogenase yielding ribulose-5-phosphate (ru5p) and CO2. In each
oxidation step one molecule of NADP+ is reduced to NADPH, resulting in the net formation of
2 mol NADPH per mol of g6p converted. In the non-oxidative branch of the PPP ru5p can be
isomerized or epimerized into, respectively, r5p or x5p by the enzymes phosphopentose isomerase
and epimerase. These metabolites form the starting point for the ensuing carbon transfer reactions
catalyzed by transketolase and transaldolase. In these reactions two- (transketolase) and three(transaldolase) carbon fragments are transferred from a ketose substrate (x5p, s7p, f6p) to an
aldose acceptor (r5p, gap, e4p). In most textbooks the transketolase and transaldolase catalyzed
reactions are depicted as three highly reversible reactions as shown in Figure 4. It should be
stressed that this is an oversimplified representation as argued by van Winden et al. [65] and in
chapter 2 of this thesis.
glc
Oxidative branch
NADP NADPH+CO2
NADP NADPH
g6p
g6p dehydrogenase &
gluconolactonase
6pg
6pg
dehydrogenase
epimerase
ru5p
isomerase
x5p
f6p
r5p
transketolase
transketolase
e4p
transaldolase
g3p
towards lower
glycolysis /TCA
s7p
Non-oxidative branch
Figure 4 Simplified representation of the oxidative and non-oxidative branch of the
pentose phosphate pathway. glc: glucose. For other abbreviations see text.
Chapter 1
10
state assumption only holds during the exponential phase of growth. As a result, reported
fluxes for batch cultures automatically correspond with those at maximal growth rate. In
contrast, continuous and fed batch cultivations allow flux analysis experiments for a whole
range of growth rates by simply altering the imposed feed rate and establishing (pseudo)
steady states.
1.3.3
Mathematical framework
MFA is based on the principle of mass conservation of the intracellular metabolites within a
defined stoichiometric network. By measuring the net conversion rates of the extracellular
metabolites, assuming (pseudo) steady state for the intracellular metabolite concentrations
and neglecting the dilution effects of growth, the mass balances of the metabolites are written
as:
0
S
 ⋅v =  
R
 
 rm 
(1.3)
In Eq. 1.3, S is the N-by-V stoichiometry matrix, where N is the number of intracellular
metabolites and V the total number of fluxes in the flux vector (v). R is the M-by-V measurement
matrix, where M is the number of measured net conversion rates in the rate vector (rm). R
contains a unity entry for each measured net conversion rate.
S
In case the rank of   equals the number of fluxes (V), the system of equations is determined
R 
and all fluxes can be calculated by inverting the matrix. Often, the system of equations is
S
undetermined (rank   < V) in which case only a linear combination of the fluxes can be
R 
calculated [67]:
S  0 
S
v = pinv    + null  β
 R   rm 
R 
(1.4)
In Eq. 1.4 ‘pinv’ denotes the pseudo-inverse and ‘null’ denotes the nullspace. Vector β contains
the linear coefficients of the columns that span the null space and thus represent the degrees
of freedom that remain after combining the balances and the extracellular measurements.
Figure 5 shows the application of MFA for resolving the flux patterns in 4 simplified metabolic
networks. Fluxes through linear and diverging nodes can be readily calculated as illustrated
by Figure 5-A. A shortcoming of metabolic balancing is, however, that it fails to identify parallel
metabolic pathways (Figure 5-B), bidirectional reaction steps and metabolic cycles (Figure 5C). For all these three networks the set of linear equations is undetermined, and only a subset
of fluxes can be quantified. The single column of the nullspace indicates that one additional
measurement is needed to determine all fluxes.
1.3.4
Parallel
primary
operate
Solving underdetermined systems
pathways, bidirectional reactions steps and/or metabolic cycles also occur in the
metabolism of a cell. For example, the upper part of the glycolysis and the PPP
in parallel, as both pathways convert g6p into gap. A possible way to overcome this
Introduction
problem is the inclusion of conserved moiety balances such as ATP, NADH and NADPH. A
difference between the upper glycolysis and the PPP is the production of NADPH in the latter
pathway. Based upon the total NADPH demand of the cell the amount of g6p metabolized
via the oxidative branch of the PPP is fixed and consequently also the flux through the upper
glycolysis. The result of including conserved moiety balances is depicted in Figure 5-D, where
the fluxes through a parallel pathway can be calculated as a result of the conserved moiety
balance.
Inclusion of conserved moiety balances has its own shortcomings, since these balances do
not resolve the flux through bidirectional reactions and are generally based on incomplete
stoichiometric information [4, 72]. Uncertainties about the ATP yield of the respiratory system
and NADPH cofactor specificities of (iso)enzymes, the presence of transhydrogenation cycles
and ATP-wasting futile cycles and the occurrence of as yet unknown sinks of NAD(P)H (e.g.
caused by oxidative stress), make flux predictions based upon assumed cofactor balances
prone to error.
A
S
v1
r1=100
1

1
 S  
R = 0
  
0
0

I1
I2
v2
r2=30
0
1
0
0
1
0
0
S
I1
I2
r2
r2
P
P
v1
S
r1 =100
I1 N
v2
 v1   50  
 1 −1 −1 0    
 
S 
  v 2   −50  
 R  =  1 −1 0 −1 ⇒  r  =  100  + 
  
1

 0 0 1 0   r   100  
 
 2 
where, β ≤ 100
I2
 v1   50   0.50 
 1 1 −1 0    
 

S 
  v 2   50   −0.50 
=
1
1
0
−
1
⇒
=
+
R 
  r   100   0 iβ
  
  1 
0
0
1
0

  r   100   0 
 

 2 
where, − 100 ≤ β ≤ 100
D
v1
r1=100
I1
v2
0   v1   30 
   

−1 0   v 2   70 




0 −1 ⇒ r1 = 100 
   

0 0   r2   30 
1 0   r3   70 
C
r11 =100
P2
1 −1 0
0
v1
S
r3
I3
B
P1
N‘
I2
r2
P
v2
0.5 

0.5 
iβ
0 

0 
 1 1 −1

 S  1 1 0
 R  =  1 −1 0
 

0 0 1
0   v1   50 
   

−1  v 2   50 
⇒
=
0   r1   100 
   

0   r2   100 
Figure 5 Application of MFA for resolving flux patterns in a diverging pathway (A), a parallel pathway
(B), a bidirectional pathway or metabolic cycle (C) and a parallel pathway with a conserved moiety
(D). Pathways A and D produce a determined system of equations and therefore yield a unique flux
solution. In contrast, pathways B and C form an underdetermined system. Some of the fluxes can
be calculated, while the remaining fluxes are expressed as a linear combination of the minimumnorm solution and the nullspace.
11
Chapter 1
12
1.4 13C-labeling technique
1.4.1
Principle
An alternative method for resolving the fluxes through parallel pathway, bidirectional reactions
and metabolic cycles is by complementing the MFA method with 13C-labeling experiments.
13
C-labeling experiments are based on feeding 13C-labeled substrate to a biological system,
allowing the labeled carbon atoms to distribute over the metabolic network, and subsequently
measuring the 13C-label distributions of intracellular compounds. Based upon the different
rearrangement of the 13C-labeled positions in the intracellular metabolites by different
reactions, this method enables the flux distribution through pathways to be discriminated even
if the pathways have the same overall stoichiometry.
An example of this is shown in the metabolic network model depicted in Figure 6, where the
product can be synthesized via two parallel pathways. As shown before (Figure 5), the system
of equations formed by the mass balances is underdetermined and thus does not lead to a
unique flux solution. The flux distribution between the two pathways can be determined by
feeding a mixture of unlabeled and uniformly labeled [u-13C3] substrate and measuring the
labeling patterns of the product. The product produced via the top pathway will have the same
labeling as the substrate, while the bottom pathway yields [1,2-13C2] and [2,3-13C2] product.
The ratio between partially labeled product and fully labeled or unlabeled product is thus a
direct measure for the flux distribution between the two parallel pathways. Note that in this
S
r 1 = 100
I2
v1
I1
v3
I3
v2
v4
I5
r2
P
I4
1 0

1 −1
S 
 RS =  0 0
  
0 1
0 0

1
0
1
0
0
 v1   50   0.5 
−1 0    
 

v2
50   0.5 

0 0 0   
 v 3   50   −0.5 
−1 0 0  ⇒   = 
+
iβ
  v 4   50   −0.5 
1 0 −1  
 100   0 
r1
0 1 0    
 

r
 2   100   0 
0
where, − 100 ≤ β ≤ 100
Figure 6 Use of 13C-labeling technique for resolving flux patterns in a parallel pathway. Open
circles denote a 12C atom and black circles denote a 13C atom. By feeding a mixture of uniformly
and unlabeled substrate and measuring the labeling distribution of product P, the flux distribution
between the upper and lower pathway can be derived.
Introduction
case the feeding of [1-13C] substrate does not resolve the flux through the upper and lower
pathway, showing the importance of a priori analysing the identifiability of the metabolic fluxes
[67] and choosing the appropriate 13C-substrate labeling [45].
A major advantage of 13C-labeling experiments is the richness of the acquired 13C-labeling data.
The total number of possible combinations of a 12C and a 13C atom in a molecule containing ncarbon atoms is equal to 2n (from hereon these combinations are referred to as isotopomers,
short for isotopic isomers). By setting up isotopomer balances (similar to the mass balances
for MFA) additional constraints are imposed on the set of flux balances. The maximum number
of constraints imposed by an isotopomer balance equals 2n-1: 2n isotopomers minus one
dependent fraction. The number of constraints imposed by 13C-labeling data on the set of
flux balances is thus much bigger then the constraints imposed by the conserved moiety
balances.
1.4.2
Experimental framework
The three most often used methods for quantifying the isotopic enrichment of the intracellular
metabolites are 2D [13C,1H] COSY NMR [40], GC-MS [20, 24] and LC-MS [69]. An overview of
the advantages and disadvantages of the three analytical platforms is given in Table 1.
Table 1 Advantages and disadvantages of the three analytical platforms commonly used for measuring
13
C-labeling distributions.
Analysis
platform
What is measured?
Advantages
Disadvantages
LC-MS
Mass isotopomers
of intracellular
metabolites
1. Sensitive
2. Direct 13C-labeling
information on
metabolites
3. Fast isotopic steady
state
1. Rapid sampling
and quenching of
metabolism in sample
2. Cell averaged 13Clabeling information
GC-MS
Mass isotopomers of
proteinogenic intact
and fragmented amino
acids
1. Sensitive
2. Compartment specific
13
C-labeling information
in eukaryotic cells
3. Easy sampling
1. Derivatization of
sample prior to analysis
2. Indirect 13C-labeling
information on
metabolites
3. Slow isotopic steady
state
NMR
Fine structures of
proteinogenic amino
acids
1. Non-destructive
analysis
2. Compartment specific
13
C-labeling information
in eukaryotic cells
3. Easy sampling
4. No separation
technique needed
1. Insensitive
2. Indirect 13C-labeling
information on
metabolites
3. Slow isotopic steady
state
13
Chapter 1
14
2D [13C,1H] COSY NMR NMR measurements of 13C-distributions are based on the splitting
of the 1H-NMR signals of protons that are bonded to 13C atoms and on the splitting of the
13
C-NMR signals of 13C atoms bonded to 13C atoms. Detected fine structures represent the
interaction of the measured 13C atom with its neighbouring 13C atoms. Depending on the
number of 13C-13C scalar couplings the following fine structures can be detected: singlet (s, no
13
C-13C scalar coupling), doublet (d, one 13C-13C scalar coupling), triplet (t, two identical 13C13
C scalar couplings) and double doublet (dd, two different 13C-13C scalar couplings). Recently,
van Winden et al. [64] developed software specifically aimed at fitting and analyzing NMR
spectra from 13C-labeling experiments.
Inherent to NMR analysis is its low sensitivity. As a result, the 13C-labeling of the intracellular
metabolites is inferred from accumulating biomass components such as proteinogenic amino
acids and storage carbohydrates (Figure 7). Knowing the biosynthetic pathway from precursor
to amino acid, the labeling patterns of the primary metabolites can be easily deduced. Of all
biomass constituents the proteinogenic amino acids contain most information on the labeling
of the primary metabolites as
they are synthesized from a large variety of precursors originating from glycolysis, pentose
phosphate pathway and TCA cycle. Note that since these precursors are localized in predefined
compartments of the cell, the measured isotopic enrichment is compartment specific.
Turnover times for biomass constituents are slow (in the order of hours). As a consequence,
long 13C-substrate feeding times are needed to reach constant 13C-labeling patterns for the
measured biomass constituents (isotopic steady state). Replacement of 12C by 13C takes place
mostly by formation of new and washout of old biomass from the chemostat. In a batch and
fed-batch cultivation this means that all added substrate has to be 13C-labeled. In a chemostat
system three residence times of 13C-feeding are needed to replace 95% (1-e-3) of the original
biomass, as calculated by first order washout kinetics. These long 13C-feeding times are a
disadvantage of NMR analysis, as they render the method unsuitable for observing metabolic
transients and require large amounts of expensive 13C-labeled substrate. Despite the fact that
prices for 13C-labeled substrate have dropped drastically over the past years, they remain
considerable (e.g. the current price of [u-13C6]glucose is $175/g). Due to the slow turnover
time of biomass, sample harvesting does not have to be rapid. After harvesting and filtering
the biomass it is hydrolyzed to free the proteinogenic amino acids, lyophilized and dissolved
in D2O prior to analysis.
GC-MS A second analytical technique that is often used in 13C-tracer studies is GC-MS.
GC-MS analysis consists of two steps; first the injected sample is heated and separated
into individual components via gas chromatography (GC), after which the components
are identified based upon their mass (MS). GC-MS is used in 13C-labeling experiments to
distinguish between isotopomers with different atomic masses (Figure 7). These so-called
mass isotopomers describe the number of 13C atoms in the molecule, but not the positions of
these 13C atoms. Similar to NMR, GC-MS determines the 13C-labeling of primary metabolites
indirectly, by measuring the 13C-labeling of proteinogenic amino acids. As a result, equally
long 13C- feeding times are needed to reach isotopic steady state for the proteinogenic amino
acids. An advantage of GC-MS is the fact that proteinogenic amino acids can be fragmented
Introduction
to obtain additional labeling information. As a result, GC-MS does not only measure the mass
isotopomers of the intact amino acid molecule, but also of specific amino acids fragments.
Sample harvesting and preparation is similar to NMR analysis, the only added step being the
derivatization of the lyophilized proteinogenic amino acids to make them volatile for the GC
step [8]. Compared to NMR, GC-MS is much more sensitive and thus requires less biomass
for analysis.
LC-MS A relatively new technique for directly determining the 13C-labeling distribution of
primary metabolites is LC-MS, consisting of a high-performance liquid chromatography
step (LC) and a MS identification step. Unlike with NMR and GC-MS this method directly
assesses the mass isotopomer distribution of primary metabolites, thereby preventing errors
due to assumptions on the amino acids biosynthesis pathways (Figure 7). Turn-over times for
primary metabolites are in the order of seconds, as a result only a short 13C-substrate feeding
time is needed to reach isotopic steady state (compared to GC-MS and NMR analysis).
Since anabolic pathways are generally unidirectional, the inflow of unlabeled carbon from the
polymeric biomass constituents can be safely neglected. Possible exceptions to this are the
inflow of unlabeled storage carbohydrates [44, 69] and the transamination reactions involved
in amino acid synthesis [26]. Due to these exceptions the attainment of isotopic steady state
can take several hours.
Because turnover times for primary metabolites are small, rapid sampling and quenching
techniques are required to correctly determine the 13C-labeling patterns of the metabolites. The
sample preparation protocol for LC-MS analysis consists of the rapid withdrawal of biomass,
the immediate quenching of metabolism, the separation of cells from the extracellular liquid,
and the extraction of primary metabolites [37]. Note that in compartmented micro-organisms
some metabolites are localized in multiple compartments (e.g. pyruvate). LC-MS analysis
yields the cell averaged 13C-labeling patterns for these metabolites, since all compartments
in the cell are lysed during the metabolite extraction step. It should be noted that the LC-MS
platform can give both the mass isotopomer distribution as well as the absolute levels of the
primary metabolites.
13
C Flux analysis methods
1.4.3
At present two main methods exist for deriving metabolic flux patterns from measured 13Clabel distributions; the local flux analysis approach and the whole isotopomer modeling
approach [55]. The local flux analysis approach determines the intracellular fluxes around a
selected metabolite node in the central metabolism by analytically interpreting the measured
13
C-labeling patterns of surrounding metabolites. Generally, algebraic equations relate the
measured 13C pattern to the flux distribution around the studied node. In contrast, the whole
isotopomer modeling approach aims at estimating all fluxes throughout a predefined metabolic
network model. Normally, there exists no algebraic flux solution for this approach. Since all
measured 13C-label distributions are fitted simultaneously, the resulting flux set is statistically
optimal for the complete network, but not necessarily optimal for each separate metabolite
node.
15
Figure 7 Overview of three
methods used for quantifying
the isotopic enrichment of in- M+0
tracellular metabolites: NMR,
GC-MS and LC-MS. Shown are M+1
the measured fine structures or
mass isotopomer fractions and
their corresponding isotopom- M+2
ers. NMR analysis yields the
relative intensities of proteino- M+3
genic amino acid fragments
(e.g. Ser-Cαβ and Ser-αβ). GCMS analysis yields the mass isotopomer fractions of (fragmented) proteinogenic amino acids
(e.g. Ser-(all) and Ser-(1+2)).
Based upon the known biosynthetic pathway from precursor
to amino acid, 13C-labeling patterns for intracellular metabolites are deduced from the NMR
and/or GC-MS data. In contrast,
LC-MS analysis directly yields
the mass isotopomer fractions
of intracellular metabolites
(e.g. 3pg: 3-phosphoglycerate)
1
3
3
2
M+2
1
m/z ->
M+1
OH
M+0
3-phoshoglycerate
O3PO
LC-MS
2
M+3
OH
O
d*
dd
s
d
1
1
2
Ser-Cαβ
M+3
M+2
M+1
M+0
3
2
dd
3
Ser-(all)
d* d
s
dd
HO
δ
M+2
2
M+3
1
d
dd
d*
NH 2
M+2
GC-MS
m/z ->
M+1
M+0
d
δ
s
Serine
OH
NMR
β α C
3
m/z ->
M+1
dd
M+0
d
1
2
3
Ser-αβ
2
s
d
M+2
M+1
M+0
Ser-(1+2)
16
O
3pg
Chapter 1
Introduction
Local node flux analysis The local node flux analysis approach has long been used to
quantify individual flux partitioning ratios [35]. In one of the earliest report on the use of 13Clabeling experiments Walsh and Koshland [71] applied NMR spectroscopy to characterize
the branch point of the tricarboxylic acid (TCA) cycle and glyoxylate shunt in Escherichia coli.
Later on both Malloy et al. [41] and Katz et al. [33] used 13C as a tracer to study specific parts
of anaplerosis and the TCA cycle in mammalian organs using NMR and GC-MS, respectively.
Other examples of flux partitioning ratios that have been quantified using a local flux analysis
approach are the PPP split-ratio [6, 39] and the degree of anaplerosis [52].
A particularly informative methodology based upon the local node flux analysis approach
is metabolic flux ratio analysis (METAFOR) [40]. With this method 13C-labeling patterns of
proteinogenic amino acids from a single experiment are analytically interpreted to quantify flux
partitioning ratios for several converging nodes in the central metabolism of a studied microorganism. The individually quantified flux partitioning ratios are largely independent of each
other and require no input of physiological information (e.g. extracellular conversion rates).
Furthermore, the determined flux ratios can be used as additional constraints for metabolic
flux analysis, thereby producing a global net flux solution for the complete metabolic network
[56, 73].
The METAFOR method was originally developed for NMR-analysis by Szyperski [58]. By
growing cells on a mixture of 10% [u-13C6]glucose and 90% naturally labeled glucose and
by measuring the 13C-labeling of the proteinogenic amino acids, non-random 13C-labeling
patterns can be identified arising from the incorporation of intact two-carbon and three-carbon
fragments originating from a single source molecule of glucose. Syperski [58] developed
probabilistic equations to relate the observed multiplet intensities of the 13C fine structures to
the relative abundance of the intact carbon fragments. These probabilistic equations enable the
derivation of fragment balancing equations, which provide accurate flux information. Fischer et
al. [20] modified the METAFOR method for calculating flux partitioning ratios based on GC-MS
derived mass isotopomers of proteinogenic amino acids from either [1-13C]glucose or [u-13C6]
glucose experiments. In total 14 ratios of fluxes through converging pathways of the central
metabolism of E. coli were identified. As the flux ratios based on GC-MS measurements are
intuitively better understandable these will be explained in more detail here. For more detailed
information on how to calculate the flux ratios from 13C-NMR multiplet data see Syperski [58]
and Maaheimo et al. [40].
Flux partitioning ratios for GC-MS measurements are calculated by setting up mass isotopomer
balances around converging metabolite nodes. Consider the case of a converging node
consisting of two inflowing fluxes:
A
v
1
v
C
B
v
2
3
17
Chapter 1
18
The mass balance and the mass isotopomer balance for this node are:
v1 + v 2 = v 3
(1.5)
v1 ⋅ MDVA + v 2 ⋅ MDVB = v 3 ⋅ MDVC
(1.6)
where MDVX is the mass isotopomer distribution vector of metabolite X (individual mass
isotopomers should sum up to one). Substituting Eq. 1.5 into Eq. 1.6 and solving for the flux
partitioning ratio (v1/v2) yields:
v1
v2
=
(MDVC − MDVB )
(MDVA − MDVC )
(1.7)
An example of a flux ratio calculated via METAFOR is the contribution of the non-oxidative
branch of the PPP via transketolase to the synthesis of the triose pool [3, 20]. This specific
flux ratio is discussed here since the PPP is one of the focal points of this thesis and because
the PPP plays an important role in penicillin synthesis. When cells are grown on a mixture
of [u-13C6]glucose and naturally labeled glucose, glucose metabolized through glycolysis
yields trioses with all carbon-carbon bonds left intact, while glucose metabolized via the PPP
produces trioses with partially cleaved bonds. The fraction of trioses in which the C1-C2 bond
has been formed in a metabolic reaction (e.g. both carbon atoms originate from a different
glucose molecule) is calculated from the mass isotopomers of the C1-C2 carbon fragment
of phosphoenol-pyruvate (PEP12). Note that mass isotopomers of PEP12 are indirectly
obtained via GC-MS analysis of the C1-C2 carbon fragments of the proteinogenic amino acids
phenylalanine and/or tyrosine. PEP12 can either be synthesized from an intact two-carbon unit
of glucose (GLU2U) or from two one-carbon units of glucose (GLU1UxGLU1U). As a result
the fraction of phosphoenol-pyruvate (PEP) formed by at least one action of transketolase
(TK) equals:
f(PEP → TK) =
MDVPEP12 − MDVGLU2U
MDVGLU1U ⋅ MDVGLU1U − MDVGLU2U
(1.8)
An upper bound for the contribution of the PPP to triose formation can be calculated by
assuming that five trioses are produced from three pentoses and that at least two of these
trioses are rearranged by a transketolase. The amount of PEP synthesized via the PPP then
becomes:
f(PEP → PPP) = 5 ⋅ f(PEP → TK)
2
(1.9)
A disadvantage of the METAFOR method (and more generally the local flux node analysis
approach) is that the 13C-labeling patterns for the required metabolites are not always
measurable. In these cases the missing labeling patterns are derived from metabolites situated
Introduction
up- or downstream, thereby assuming identical 13C-labeling patterns for these metabolites. If
this assumption is invalid, it can easily lead to erroneous outcomes for the local node flux
analysis. In the above example, the 13C-labeling of gap produced by transketolase is inferred
from the labeling of PEP. This inference is based on the assumption that the 13C-labeling of
PEP is not influenced by other reactions, such as the gluconeogenic reaction catalyzed by
PEP carboxykinase.
Another disadvantage of the local node flux analysis approach is that flux partitioning ratios in
the studied metabolic network can only be determined for convergent nodes. In the example
discussed above the PPP split-ratio can be deduced from the convergent node created by
triose molecules originating from either the PPP or glycolysis. The derived flux ratio is an
upper bound, as It does not take into account the formation of additional cleaved C1-C2 triose
fragments through reversibility of the transketolase. Direct determination of the PPP split-ratio
is not possible, since this is a divergent metabolite node.
Whole isotopomer modeling Fundamental to the whole isotopomer modeling approach is
the formulation of a whole isotopomer model as described by Schmidt et al. (1997). This
whole isotopomer model defines the complete set of balances of each isotopomer fraction
of each individual metabolite pool in the studied network. Combined with the measured
uptake and secretion rates the model relates the isotopic enrichment of metabolites to all
intracellular fluxes. However, due to the non-linearity of the isotopomer balances, in all but
very simple metabolic network models, fluxes can not be expressed as an explicit function
of the measured 13C-labeling data. Consequently, flux-patterns are calculated by iteratively
fitting simulated 13C-label distributions for a chosen set of metabolic fluxes to the measured
13
C-label distributions. The flux-set that gives the best correspondence between the measured
and simulated 13C-label distribution is determined by non-linear optimization and denoted as
the optimal flux-fit.
The whole isotopomer modeling approach has been applied in many studies to derive metabolic
flux patterns from GC-MS [9, 19, 24, 36, 79], NMR [68, 75] and LC-MS [69] measurements.
In most of these studies fluxes are estimated by simulating mass isotopomers or relative
intensities and fitting these to their measured counterpart. Christensen et al. [9] introduced a
slightly different approach by, prior to fitting, transforming the mass isotopomer distributions
of the measured proteinogenic amino acids (GC-MS) into summed fractional labelings (SFL).
The SFL of an amino acid or fragment thereof is equal the sum of its positional enrichments:
SFL =
0 ⋅ Im + 0 + 1⋅ Im +1 + 2 ⋅ Im + 2 .. + n ⋅ Im +n
Im + 0 + Im +1 + Im + 2 .. + Im +n
(1.10)
where I is the peak intensity, m is the monoisotopic mass and n is the total number of carbon
atoms in the studied amino acid (fragment). Even though SFL’s are intuitively understandable
and speed up the flux fitting procedure, their condensed form means that the isotopomer
information is not used to its full potential. For example: two metabolite pools, one that is
100% [1-13C] labeled and one that is 50% unlabeled and 50% [1,2-13C]-labeled have identical
SFL’s. Nevertheless, Christensen et al. [7] showed that the fluxes obtained with this method
19
Chapter 1
20
are very reproducible and sensitive to small metabolic variations caused by changes in growth
conditions or genetic make-up of the micro-organism.
A relative new application of whole isotopomer modeling is the respirometric 13C flux analysis
method introduced by Yang et al. [76]. In this method the only 13C-labeling input comes from
online CO2 labeling measurements. The information content of CO2 is extremely low, as it
contains only one independent isotopomer. Additional 13C-labeling constrains are imposed on
the fluxes in the studied network by performing multiple parallel 13C-labeling experiments using
different 13C-labeled substrates. However, in general not all (exchange) fluxes in the studied
metabolic network can be estimated. Despite this drawback, Yang et al. [77] showed that
for a lysine-producing Corynebacterium glutamicum mutant the net fluxes in the glycolysis,
TCA cycle, anaplerosis and the PPP could be accurately quantified. Since CO2 does not
accumulate in the cell and its labeling can be quantified online, the respirometric method is
suitable for instationary flux analysis [47]. In addition the method is non-invasive allowing the
study of e.g. biofilms, sediments or immobilized cell cultures, without disturbing or destroying
the system.
An obvious advantage of the whole isotopomer modeling approach is insight in all metabolic
fluxes of the studied network. Due to the input of quantitative physiological data such as
extracellular net conversion rates and biomass composition the determined flux values are
absolute. Furthermore, the high information content of the 13C-labeling data constrains the flux
fit and results in an over-determined system of equations, allowing researchers to statistically
verify the topology of the studied metabolic network model. By investigating the statistical
acceptability of the flux fit, shortcomings in the stoichiometry of the metabolic network model
can be localized and alterations to the model can be hypothesized and validated [9, 52, 68].
A major disadvantage of the whole isotopomer modeling approach is the high interconnectivity
of cellular metabolism and the non-linear fit problem. Incorrectly simulated 13C-labeling patterns
in one part of the network as a result of a missing or erroneous reaction may very well propagate
and cause incorrectly estimated fluxes throughout the studied network. Furthermore, changes
in the metabolite-labeling can be counteracted by changed flux-patterns in other parts of the
metabolism. As a result, the fitting-procedure can produce multiple ‘optimal’ flux sets each with
their own flux distribution. A good understanding of the studied network and a critical statistical
assessment of the flux fitting results is thus essential for accurately determining fluxes.
1.5 Aim and outline of the thesis
13
C-labeling experiments have been a focus of research at the bioprocess technology group
of the Department of Biotechnology at Delft University of Technology since 1997. Innovations
were made to the 13C-labeling technique with respect to the experimental, mathematical
and biochemical aspects [66]. Initial research focused on further developing the 13C-labeling
methodology in order to maximize the information content of 13C-labeling data [65, 67]. The first
metabolic flux analysis for P. chrysogenum was performed using a whole isotopomer modeling
approach based upon 2D [13C,1H] COSY NMR derived relative intensities for the proteinogenic
amino acids [68]. Meanwhile, the concurrent development of experimental tools for stimulus
response experiments, such as the rapid sampling and analysis of intracellular metabolites
Introduction
by LC-MS [37, 61], provided a novel method for quantifying the 13C-label distribution in the
cell. With this method the mass isotopomer fractions of 13C-labeled primary metabolites were
directly measured and used for analyzing the metabolic fluxes of Saccharomyces cerevisiae
[69]. The above results formed the basis of the research described in this thesis. This thesis
aims to apply and further develop the 13C-labeling techniques and the available analytical
platforms for the analysis of the metabolic fluxes in P. chrysogenum, thereby specifically
focussing on the flux through the non-oxidative branch of the PPP which is of prime importance
for penicillin synthesis.
Chapter 2 describes the development of a new metabolic network model for the PPP based
upon the established ping-pong kinetic mechanism of the enzymes transketolase and
transaldolase. The first 13C-labeling based metabolic flux analysis of the PPP was performed
by Follstad et al. [22], who constructed a general metabolic network model in order to
determine the degree of reversibility of the individual reactions in the non-oxidative branch of
the PPP. Van Winden et al. [65] further investigated the PPP and showed that six additional
reactions could take place in the non-oxidative branch of the PPP. All six additional reactions
were included in the new stoichiometric model for the PPP, but restructured into metabolite
specific, reversible, C2 and C3 fragments producing and consuming half-reactions. It is shown
that a stoichiometric model based upon these half-reactions is fundamentally different from
the currently applied stoichiometric models and can lead to different label distributions for 13Ctracer experiments.
Chapter 3 and 4 focus on the application of whole isotopomer models for estimating the
metabolic fluxes in, respectively, S. cerevisiae and the closely related micro-organism
P. chrysogenum. In these two studies different analytical platforms were used to quantify
the 13C-labeling distribution in the cell. Comparisons were made between the estimated
flux patterns for the different techniques and a sensitivity analysis of the flux estimates for
important metabolite nodes was performed. In Chapter 3 we investigate a glycerol hyperproducing tpi1∆nde1,2∆gut2∆ S. cerevisiae strain by measuring the isotopic enrichment of the
intracellular primary metabolites both directly (LC-MS) and indirectly from proteinogenic amino
acids (NMR and GC-MS). New insight in the carbon and redox metabolism of the studied
strain is presented. In addition, it is shown that the three 13C-quantification techniques lead to
similar flux-patterns but have different flux sensitivities for important metabolite nodes such
as the PPP split-ratio. In Chapter 4 the fluxes of P. chrysogenum are determined under both
penicillin producing and non-producing conditions to examine the effect of penicillin synthesis
on primary metabolism. Two analytical platforms are compared: NMR and LC-MS. In addition,
we show how labeling redundancies can be used to reconstruct and validate a metabolic
network model. In contrast to S. cerevisiae, highly different flux patterns are presented for the
two applied 13C-quantification techniques, making it difficult to draw clear conclusions on the
effect of penicillin-G production on the primary metabolism of P. chrysogenum.
In Chapter 5 a local node flux analysis approach is used to directly determine the flux
through the oxidative branch of the PPP. Since the PPP split-ratio is a divergent node the flux
partitioning around this node is normaly not directly accessible. In this chapter the oxidative
branch of the PPP is artificially turned into a convergent node by the simultaneous feeding of
unlabeled glucose and trace amounts of [u-13C]gluconate. A sensitivity analysis shows that this
21
Chapter 1
22
method is more accurate than the whole isotopomer modeling approach applied in Chapter
3 and 4. In Chapter 6 the gluconate-tracer method of Chapter 5 is applied to a penicillin-G
producing and non-producing chemostat culture. In accordance with earlier claims, this study
conclusively shows for the first time that the flux through the oxidative branch of the PPP is
strongly correlated to β-lactam antibiotic production and that the oxidative branch of the PPP
produces the majority of the cytosolic NADPH needed for penicillin synthesis.
Finally, Chapter 7 discusses the outcome of the studies presented in this thesis and provides
the reader with possible leads for future research.
Introduction
23
References
[1] Abraham, E.P., Chain, E., Fletcher, C.M., Gardner,
A.D., Heatley, N.G., Jennings, M.A. and Florey,
H.W. (1941) Further observations on penicillin. The
Lancet 238, 177-189.
[2] Biwer, A., Griffith, S. and Cooney, C. (2005)
Uncertainty analysis of penicillin V production using
Monte Carlo simulation. Biotechnol. Bioeng. 90,
167-179.
[3] Blank, L.M. and Sauer, U. (2004) TCA cycle activity
in Saccharomyces cerevisiae is a function of the
environmentally determined specific growth and
glucose uptake rates. Microbiol. 150, 1085-1093.
[4] Bonarius, H.P.J., Schmid, G. and Tramper, J.
(1997) Flux analysis of underdetermined metabolic
networks: the quest for the missing constraints.
Trends Biotechnol. 15, 308-314.
[5] Chain, E., Florey, H.W., Adelaide, M.B., Gardner,
A.D., Oxfd, D.M., Heatley, N.G., Jennings, M.A., OrrEwing, J. and Sanders, A.G. (1940) Penicillin as a
chemotherapeutic agent. The Lancet 236, 226-228.
[6] Christensen, B., Christiansen, T., Gombert, A.K.,
Thykaer, J. and Nielsen, J. (2001) Simple and
robust method for estimation of the split between
the oxidative pentose phosphate pathway and
the
Embden-Meyerhof-Parnas
pathway
in
microorganisms. Biotechnol. Bioeng. 74, 517-523.
[7] Christensen, B., Gombert, A.K. and Nielsen, J.
(2002) Analysis of flux estimates based on 13Clabeling experiments. Eur. J. Biochem. 269, 27952800.
[8] Christensen, B. and Nielsen, J. (1999) Isotopomer
Analysis Using GC-MS. Metab. Eng. 1, 282-290.
[9] Christensen, B. and Nielsen, J. (2000) Metabolic
network analysis of Penicillium chrysogenum using
13
C-labeled glucose. Biotechnol. Bioeng. 68, 652659.
[10] Christensen, B., Thykaer, J. and Nielsen, J. (2000)
Metabolic characterization of high- and low-yielding
strains of Penicillium chrysogenum. Appl. Microb.
Biot. 54, 212-217.
[11] Cox, P.W., Paul, G.C. and Thomas, C.R. (1998)
Image analysis of the morphology of filamentous
micro-organisms. Microbiol. 144, 817-827.
[12] Dauner,
M.,
Sonderegger,
M.,
Hochuli,
M.,
Szyperski, T., Wuthrich, K., Hohmann, H.P., Sauer,
U. and Bailey, J.E. (2002) Intracellular carbon fluxes
in riboflavin-producing Bacillus subtilis during growth
on two-carbon substrate mixtures. Appl. Environ.
Microb. 68, 1760-1771.
[13] David, H., Akesson, M. and Nielsen, J. (2003)
Reconstruction of the central carbon metabolism of
Aspergillus niger. Eur. J. Biochem. 270, 4243-4253.
[14] Demain, A.L. and Elander, R.P. (1999) The betalactam antibiotics: past, present, and future. Anton.
Leeuw. Int. J. G. 75, 5-19.
[15] Edwards, J.S., Ibarra, R.U. and Palsson, B.O. (2001)
In silico predictions of Escherichia coli metabolic
capabilities are consistent with experimental data.
Nat. Biotechnol. 19, 125-130.
[16] Elander, R.P. (2003) Industrial production of betalactam antibiotics. Appl. Microb. Biot. 61, 385-392.
[17] Evers, M.E., Trip, H., van den Berg, M.A.,
Bovenberg, R.A.L. and Driessen, A.J.M. (2004)
Compartmentalization and transport in beta-lactam
antibiotics biosynthesis. In: Molecular biotechnology
of fungal beta-lactam antibiotics and related peptide
synthetases (Brakhage, A.A., Ed.), Vol. 88, pp. 111135. Springer Berlin, Heidelberg.
[18] Fierro, F., Garcia-Estrada, C., Castillo, N.I.,
Rodriguez, R., Velasco-Conde, T. and Martin, J.F.
(2006) Transcriptional and bioinformatic analysis of
the 56.8 kb DNA region amplified in tandem repeats
containing the penicillin gene cluster in Penicillium
chrysogenum. Fungal Genet. Biol. 43, 618-629.
[19] Fischer, E. and Sauer, U. (2003) A novel metabolic
cycle catalyzes glucose oxidation and anaplerosis in
hungry Escherichia coli. J. Biol. Chem. 278, 4644646451.
[20] Fischer, E. and Sauer, U. (2003) Metabolic flux
profiling of Escherichia coli mutants in central
carbon metabolism using GC-MS. Eur. J. Biochem.
270, 880-891.
[21] Fleming, A. (1929) On the antibacterial action of
cultures of a Penicillium, with special reference to
their use in the isolation of Bacillus influenzae. Brit.
J. Exp. Path. 10, 226-236.
[22] Follstad, B.D. and Stephanopoulos, G. (1998) Effect
of reversible reactions on isotope label redistribution
- Analysis of the pentose phosphate pathway. Eur. J.
Chapter 1
24
Biochem. 252, 360-371.
[23] Gailey, F.B., Stefaniak, J.J., Olson, B.H. and Johnson,
M.J. (1946) A Comparison of Penicillin-Producing
Strains of Penicillium notatum-chrysogenum. J.
Bacteriol. 52, 129-140.
[24] Gombert, A.K., dos Santos, M.M., Christensen,
B. and Nielsen, J. (2001) Network identification
and flux quantification in the central metabolism of
Saccharomyces cerevisiae under different conditions
of glucose repression. J. Bacteriol. 183, 1441-1451.
[25] Grimm, L.H., Kelly, S., Krull, R. and Hempel, D.C.
(2005) Morphology and productivity of filamentous
fungi. Appl. Microb. Biot. 69, 375-384.
[26] Grotkjaer, T., Akesson, M., Christensen, B., Gombert,
A.K. and Nielsen, J. (2004) Impact of transamination
reactions and protein turnover on labeling dynamics
in 13C-labeling experiments. Biotechnol. Bioeng. 86,
209-216.
[27] Gunnarsson, N., Eliasson, A. and Nielsen, J. (2004)
Control of fluxes towards antibiotics and the role
of primary metabolism in production of antibiotics.
In: Molecular biotechnology of fungal beta-lactam
antibiotics and related peptide synthetases
(Brakhage, A.A., Ed.), Vol. 88, pp. 137-178. Springer
Berlin, Heidelberg.
[28] Henriksen, C.M., Christensen, L.H., Nielsen, J.
and Villadsen, J. (1996) Growth energetics and
metabolic fluxes in continuous cultures of Penicillium
chrysogenum. J. Biotechnol. 45, 149-164.
[29] Hobby, G.L. (1985) Pencillin. Meeting the Challenge.
Yale University Press, New Haven.
[30] Jorgensen, H., Nielsen, J., Villadsen, J. and
Mollgaard, H. (1995) Analysis of Penicillin-V
biosynthesis during fed-batch cultivations with A
high-yielding strain of Penicillium chrysogenum.
Appl. Microb. Biot. 43, 123-130.
[31] Jorgensen, H., Nielsen, J., Villadsen, J. and
Mollgaard, H. (1995) Metabolic flux distributions
in Penicillium chrysogenum during fed-batch
cultivations. Biotechnol. Bioeng. 46, 117-131.
[32] Kallow, W., von Dohren, H. and Kleinkauf, H. (1998)
Penicillin biosynthesis: Energy requirement for
tripeptide precursor formation by delta-(L-alphaaminoadipyl)-L-cysteinyl-D-valine synthetase from
Acremonium chrysogenum. Biochemistry 37, 59475952.
[33] Katz, J., Lee, W.N.P., Wals, P.A. and Bergner,
E.A. (1989) Studies of glycogen-synthesis and
the Krebs cycle by mass isotopomer analysis with
[U-13C]Glucose in rats. J. Biol. Chem. 264, 1299413004.
[34] Kayser, A., Weber, J., Hecht, V. and Rinas, U. (2005)
Metabolic flux analysis of Escherichia coli in glucoselimited continuous culture. I. Growth-ratedependent
metabolic efficiency at steady state. Microbiol. 151,
693-706.
[35] Kelleher, J.K. (2001) Flux estimation using isotopic
tracers: Common ground for metabolic physiology
and metabolic engineering. Metab. Eng. 3, 100110.
[36] Klapa, M.I., Aon, J.C. and Stephanopoulos, G.
(2003) Systematic quantification of complex
metabolic flux networks using stable isotopes and
mass spectrometry. Eur. J. Biochem. 270, 35253542.
[37] Lange, H.C., Eman, M., van Zuijlen, G., Visser,
D., van Dam, J.C., Frank, J., de Mattos, M.J. and
Heijnen, J.J. (2001) Improved rapid sampling
for in vivo kinetics of intracellular metabolites in
Saccharomyces cerevisiae. Biotechnol. Bioeng. 75,
406-415.
[38] Lax, E. (2004) The Mould in Dr Florey's Coat. Little,
Brown, London.
[39] Lee, W.N.P., Boros, L.G., Puigjaner, J., Bassilian, S.,
Lim, S. and Cascante, M. (1998) Mass isotopomer
study of the nonoxidative pathways of the pentose
cycle with [1,2-13C2]glucose. Am. J. Physiol-Endoc.
M. 37, E843-E851.
[40] Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E.,
Sauer, U. and Szyperski, T. (2001) Central carbon
metabolism of Saccharomyces cerevisiae explored
by biosynthetic fractional 13C-labeling of common
amino acids. Eur. J. Biochem. 268, 2464-2479.
[41] Malloy, C.R., Sherry, A.D. and Jeffrey, F.M.H. (1988)
Evaluation of carbon flux and substrate selection
through alternate pathways involving the citric-acid
cycle of the heart by 13C NMR spectroscopy. J. Biol.
Chem. 263, 6964-6971.
[42] Martin, J.F. (1998) New aspects of genes and
enzymes for beta-lactam antibiotic biosynthesis.
Appl. Microb. Biot. 50, 1-15.
[43] Martin, J.F., Casqueiro, J. and Liras, P. (2005)
Secretion systems for secondary metabolites: how
producer cells send out messages of intercellular
communication. Curr. Opin. Microbiol. 8, 282-293.
Introduction
[44] Mashego, M.R., Wu, L., van Dam, J.C., Ras, C.,
Vinke, J.L., van Winden, W.A., van Gulik, W.M. and
Heijnen, J.J. (2004) MIRACLE: mass isotopomer
ratio analysis of [U-13C]-labeled extracts. A new
method for accurate quantification of changes
in concentrations of intracellular metabolites.
Biotechnol. Bioeng. 85, 620-628.
[45] Mollney, M., Wiechert, W., Kownatzki, D. and de
Graaf, A.A. (1999) Bidirectional reaction steps in
metabolic networks: IV. Optimal design of isotopomer
labeling experiments. Biotechnol. Bioeng. 66, 86103.
[46] Nielsen, J. (1997) Physiological engineeering
aspects of Penicillium chrysongenum. World
Scientific Publising Co., Singapore.
[47] Noh, K. and Wiechert, W. (2006) Experimental
design principles for isotopically instationary 13Clabeling experiments. Biotechnol. Bioeng. 94, 234251.
[48] Papagianni, M. (2004) Fungal morphology and
metabolite production in submerged mycelial
processes. Biotechnol. Adv. 22, 189-259.
[49] Paul, G.C., Kent, C.A. and Thomas, C.R. (1993)
Viability testing and characterization of germination
of gungal spores by automatic image-analysis.
Biotechnol. Bioeng. 42, 11-23.
[50] Paul, G.C., Kent, C.A. and Thomas, C.R. (1994)
Hyphal vacuolation and fragmentation in Penicillium
chrysogenum. Biotechnol. Bioeng. 44, 655-660.
[51] Pazouki, M. and Panda, T. (2000) Understanding the
morphology of fungi. Bioprocess Eng. 22, 127-143.
[52] Petersen, S., de Graaf, A.A., Eggeling, L., Mollney,
M., Wiechert, W. and Sahm, H. (2000) In vivo
quantification of parallel and bidirectional fluxes in
the anaplerosis of Corynebacterium glutamicum. J.
Biol. Chem. 275, 35932-35941.
[53] Robin, J., Bruheim, P., Nielsen, M.L., Noorman,
H. and Nielsen, J. (2003) Continuous cultivations
of a Penicillium chrysogenum strain expressing
the
expandase
gene
from
Streptomyces
clavuligerus:
Kinetics
of
adipoyl-7-aminodeacetoxycephalosporanic acid and byproduct
formations. Biotechnol. Bioeng. 83, 353-360.
[54] Sahm, H., Eggeling, L. and de Graaf, A.A. (2000)
Pathway analysis and metabolic engineering in
Corynebacterium glutamicum. Biol. Chem. 381,
899-910.
[55] Sauer, U. (2004) High-throughput phenomics:
experimental methods for mapping fluxomes. Curr.
Opin. Biotech. 15, 58-63.
[56] Sauer, U., Hatzimanikatis, V., Bailey, J.E., Hochuli,
M., Szyperski, T. and Wuthrich, K. (1997) Metabolic
fluxes in riboflavin-producing Bacillus subtilis. Nat.
Biotechnol. 15, 448-452.
[57] Stuckrath, I., Lange, H.C., Kotter, P., van Gulik,
W.M., Entian, K.D. and Heijnen, J.J. (2002)
Characterization of null mutants of the glyoxylate
cycle and gluconeogenic enzymes in Saccharomyces
cerevisiae through metabolic network modeling
verified by chemostat cultivation. Biotechnol. Bioeng.
77, 61-72.
[58] Szyperski, T. (1995) Biosynthetically directed
fractional 13C-Labeling of proteinogenic amino acids An efficient analytical tool to investigate intermediary
metabolism. Eur. J. Biochem. 232, 433-448.
[59] Theilgaard, H.A., van den Berg, M.A., Mulder,
C.A., Bovenberg, R.A.L. and Nielsen, J. (2001)
Quantitative analysis of Penicillium chrysogenum
Wis54-1255 transformants overexpressing the
penicillin biosynthetic genes. Biotechnol. Bioeng.
72, 379-388.
[60] Thykaer, J. and Nielsen, J. (2003) Metabolic
engineering of beta-lactam production. Metab. Eng.
5, 56-69.
[61] van Dam, J.C., Eman, M.R., Frank, J., Lange,
H.C., van Dedem, G.W.K. and Heijnen, S.J.
(2002) Analysis of glycolytic intermediates in
Saccharomyces cerevisiae using anion exchange
chromatography and electrospray ionization with
tandem mass spectrometric detection. Anal. Chim.
Acta 460, 209-218.
[62] van Gulik, W.M., Antoniewicz, M.R., deLaat, W.T.,
Vinke, J.L. and Heijnen, J.J. (2001) Energetics of
growth and penicillin production in a high-producing
strain of Penicillium chrysogenum. Biotechnol.
Bioeng. 72, 185-93.
[63] van Gulik, W.M., de Laat, W.T., Vinke, J.L. and
Heijnen, J.J. (2000) Application of metabolic
flux analysis for the identification of metabolic
bottlenecks in the biosynthesis of penicillin-G.
Biotechnol. Bioeng. 68, 602-618.
[64] van Winden, W., Schipper, D., Verheijen, P. and
Heijnen, J. (2001) Innovations in generation and
analysis of 2D [(13C,1H] COSY NMR spectra for
metabolic flux analysis purposes. Metab. Eng. 3,
322-343.
25
Chapter 1
26
[65] van Winden, W., Verheijen, P. and Heijnen, S.
(2001) Possible pitfalls of flux calculations based on
13
C-labeling. Metab. Eng. 3, 151-162.
[66] van Winden, W.A. (2003) 13C-Labeling Technique for
Metabolic Network and Flux Analysis: Theory and
Application. Thesis. Delft University of Technology.
[67] van Winden, W.A., Heijnen, J.J., Verheijen, P.J. and
Grievink, J. (2001) A priori analysis of metabolic
flux identifiability from 13C-labeling data. Biotechnol.
Bioeng. 74, 505-516.
[68] van Winden, W.A., van Gulik, W.M., Schipper,
D., Verheijen, P.J.T., Krabben, P., Vinke, J.L. and
Heijnen, J.J. (2003) Metabolic flux and metabolic
network analysis of Penicillium chrysogenum
using 2D [13C, 1H] COSYNMR measurements
and cumulative Bondomer simulation. Biotechnol.
Bioeng. 83, 75-92.
[69] van Winden, W.A., van Dam, J.C., Ras, C., Kleijn,
R.J., Vinke, J.L., van Gulik, W.M. and Heijnen, J.J.
(2005) Metabolic flux analysis of Saccharomyces
cerevisiae CEN.PK113-7D based on mass
isotopomer measurements of 13C-labeled primary
metabolites. FEMS Yeast Res. 5, 559-568.
[70] Veenstra, A.E., Vansolingen, P., Bovenberg, R.A.L.
and Vandervoort, L.H.M. (1991) Strain Improvement
of Penicillium-Chrysogenum by Recombinant-Dna
Techniques. J. Biotechnol. 17, 81-90.
[71] Walsh, K. and Koshland, D.E. (1984) Determination
of flux through the branch point of 2 metabolic cycles
- the tricarboxylic acid cycle and the glyoxylate
shunt. J. Biol. Chem. 259, 9646-9654.
[72] Wiechert, W. (2001) 13C metabolic flux analysis.
Metab. Eng. 3, 195-206.
[73] Wittmann, C. and Heinzle, E. (2001) Application
of MALDI-TOF MS to lysine-producing Corynebacterium glutamicum - A novel approach for
metabolic flux analysis. Eur. J. Biochem. 268, 24412455.
[74] Wittmann, C. and Heinzle, E. (2002) Genealogy
profiling through strain improvement by using
metabolic network analysis: Metabolic flux
genealogy of several generations of lysine-producing
corynebacteria. Appl. Environ. Microb. 68, 58435859.
[75] Yang, C., Hua, Q. and Shimizu, K. (2002) Metabolic
flux analysis in Synechocystis using isotope
distribution from 13C-labeled glucose. Metab. Eng. 4,
202-216.
[76] Yang, T.H., Wittmann, C. and Heinzle, E. (2006)
Respirometric 13C flux analysis, Part I: Design,
construction and validation of a novel multiple
reactor system using on-line membrane inlet mass
spectrometry. Metab. Eng. 8, 417-431.
[77] Yang, T.H., Wittmann, C. and Heinzle, E. (2006)
Respirometric 13C flux analysis, Part II: In vivo flux
estimation of lysine-producing Corynebacterium
glutamicum. Metab. Eng. 8, 432-446.
[78] Zamboni, N., Fischer, E., Laudert, D., Aymerich, S.,
Hohmann, H.P. and Sauer, U. (2004) The Bacillus
subtilis yqiI gene encodes the NADP(+)-dependent
6-P-gluconate dehydrogenase in the pentose
phosphate pathway. J. Bacteriol. 186, 4528-4534.
[79] Zhao, J. and Shimizu, K. (2003) Metabolic flux
analysis of Escherichia coli K12 grown on C-13labeled acetate and glucose using GG-MS and
powerful flux calculation method. J. Biotechnol. 101,
101-117.
Introduction
27
CHAPTER
2
Revisiting the 13C-label distribution
of the non-oxidative branch of
the pentose phosphate pathway
based upon kinetic and
genetic evidence
Roelco J. Kleijn, Wouter A. van Winden,
Walter M. van Gulik and Joseph J. Heijnen
Based upon:
FEBS Journal
(2005) 272 (19): 4970-4982
Chapter 2
30
ABSTRACT
The currently applied reaction structure in stoichiometric flux balance models for the
non‑oxidative branch of the pentose phosphate pathway (PPP) is not in accordance with the
established ping‑pong kinetic mechanism of the enzymes transketolase and transaldolase.
Based upon the ping‑pong mechanism, the traditional reactions of the non‑oxidative branch
of the PPP are replaced by metabolite specific, reversible, C2 and C3 fragments producing
and consuming half‑reactions. It is shown that a stoichiometric model based upon these
half‑reactions is fundamentally different from the currently applied stoichiometric models with
respect to the number of independent C2 and C3 fragment pools in the PPP, and can lead to
different label distributions for 13C‑tracer experiments. To investigate the actual impact of the new
reaction structure on the estimated flux‑patterns within a cell, mass isotopomer measurements
from a previously published 13C‑based metabolic flux analysis of Saccharomyces cerevisiae
were used. Different flux‑patterns were found.
From a genetic point of view, it is well known that several micro-organisms, including Escherichia
coli and S. cerevisiae, contain multiple genes encoding for isoenzymes of transketolase and
transaldolase. However, it remains the question to which extent these gene products are also
actively expressed. It is shown that the newly proposed stoichiometric model allows studying
of the effect of isoenzymes on the 13C‑label distribution in the non‑oxidative branch of the
PPP by extending the half‑reaction based stoichiometric model with two distinct transketolase
enzymes instead of one. Results show that inclusion of isoenzymes affects the ensuing
flux‑estimates.
Revisiting the Pentose Phosphate Pathway
2.1
INTRODUCTION
During the past decade 13C‑labeling based metabolic flux analysis (MFA) has increasingly been
used to understand the effect of genetic alterations [40, 48], changes in external conditions
[8, 20] and different nutritional regimes [13, 42] on the metabolism of micro-organisms.
13
C‑labeling based MFA relies on the feeding of 13C‑labeled substrate to a biological system,
allowing the labeled carbon atoms to distribute over the metabolic network, and subsequently
measuring the 13C‑label distributions of intracellular and/or secreted compounds by means
of NMR spectroscopy or MS. The flux‑patterns within a metabolic network model can be
calculated by iteratively fitting simulated 13C‑label distributions for a chosen set of metabolic
fluxes to the measured 13C‑label distributions [44]. Apart from MFA, the information richness
of 13C‑labeling data also permits verification of the topology of metabolic network models.
Furthermore, shortcomings in the stoichiometry of the metabolic network can be localized and
alterations to the model can be hypothesized and validated [4, 26, 42].
A part of the metabolic network that has received relatively little attention from the MFA
community with respect to model validation is the pentose phosphate pathway (PPP). This
is rather surprising since the PPP plays several key roles in the cell metabolism. Apart from
supplying the cell with precursors for amino acid and nucleotide biosynthesis, it also plays
a crucial role in maintaining the cytosolic NADP+/NADPH balance. In order to maintain this
balance, the flux through the oxidative branch of the PPP is usually much bigger than the
drain on PPP metabolites for the biosynthesis of building blocks, resulting in a significant
recycling and redistribution of the carbon atoms via the non‑oxidative branch. Incorrectly
mapped carbon atom distributions, due to for example an incomplete or incorrect metabolic
model, can lead to erroneously predicted label distributions (and consequently flux estimates)
for 13C‑tracer experiments.
Practically all stoichiometric flux balance models of the non‑oxidative branch of the PPP
consist of three reversible reactions; two transketolase (TK) catalyzed reactions (r.1&2) and
one transaldolase (TA) catalyzed reaction (r.3) [7, 13, 21, 32, 47, 50].
TK
x5p + r5p ↔ s7p + gap
TK
x5p + e4p ↔ f6p + gap
TA
s7p + gap ↔ f6p + e4p
(r.1)
(r.2)
(r.3)
Van Winden et al. [41] argued that the non‑oxidative branch of the PPP consists of more
reactions than the three conventional reactions shown above. Supporting evidence from
literature was presented indicating that six additional reactions could take place [5, 10,
19, 46]. Furthermore, van Winden et al. [42] demonstrated that the incorporation of these
reactions in the metabolic network model of Penicillium chrysogenum significantly increased
the goodness‑of‑fit to measured 13C‑label distribution data and also resulted in a changed
flux distribution. The six additional reactions consist of five stoichiometric neutral reactions,
two of which are catalyzed by transaldolase (r.8&9) and three of which by transketolase
(r.5‑7), and one additional reversible transketolase‑catalyzed reaction (r.4). Although the
31
Chapter 2
32
stoichiometric neutral reactions have no effect on the mass balances set up over the system,
they do influence the labeling pattern of the metabolite pools and thus need to be incorporated
into the metabolic network for 13C‑based flux estimations [1, 10]. The structure of reactions
r.1‑9 is such that a carbon‑fragment is transferred from one substrate to another, yielding two
products. From hereon any non‑oxidative PPP reactions abiding by this structure are denoted
as traditional reactions.
TK
f6p + r5p ↔ e4p + s7p
TK
g3p + x5p → x5p + gap
TK
f6p + e4p → e4p + f6p
TK
r5p + s7p → s7p + r5p
TA
f6p + g3p → gap + f6p
TA
e4p + s7p → s7p + e4p
(r.4)
(r.5)
(r.6)
(r.7)
(r.8)
(r.9)
In this article results of genetic and kinetic studies into the non‑oxidative branch of the PPP are
analyzed and used to obtain a more realistic stoichiometric flux balance model. Based upon
the kinetic mechanism of transaldolase and transketolase, an alternative reaction structure
for tracing the distribution of 13C through the non‑oxidative branch of the PPP is proposed.
It is shown that a stoichiometric flux balance model based upon this new reaction structure
is fundamentally different from the current models with respect to 13C‑label distribution and,
consequently, can yield different flux‑patterns. Moreover, the new reaction structure facilitates
the estimation of the metabolic fluxes from the 13C‑labeling data due to a smaller number of
parameters. Following genetic evidence, the presence of isoenzymes for transketolase and
transaldolase is incorporated to further refine the stoichiometric model. The effect of these
model alterations on the estimated 13C‑based flux patterns is examined using a recently
published MFA for Saccharomyces cerevisiae based upon mass isotopomer measurements
of 13C‑labeled primary metabolites [43].
2.2 THEORY
2.2.1
Kinetic mechanism of the non‑oxidative branch of the PPP
The enzymes transketolase and transaldolase catalyze the transfer of two‑ and three‑carbon
fragments from a ketose donor to an aldose acceptor. Transketolase performs this C2‑transfer
using a tightly bound thiamine pyrophosphate (TPP) as cofactor. The second carbon atom of
the thiazole ring of TPP readily ionizes to give a carbanion, which can react with the carbonyl
group of the ketose substrates; xylulose‑5‑phosphate (X5P), fructose‑6‑phopshate (F6P) or
sedoheptulose‑7‑phosphate (S7P). The phosphorylated part of the ketose substrate splits off,
leaving a negatively charged glycolaldehyde (C2) attached to TPP. Resonance forms keep
the glycolaldehyde unit attached to TPP until a suitable acceptor has been found in the form
Revisiting the Pentose Phosphate Pathway
of ribose‑5‑phosphate (R5P), erythrose‑4‑phosphate (E4P) or glyceraldehyde‑3‑phosphate
(GAP) [37]. In contrast to transketolase, transaldolase does not contain a prosthetic group.
Instead, a Schiff base is formed between the carbonyl group of the ketose substrate (F6P,
S7P) and the ε‑amino group of a lysine residue of the active site of the enzyme, leading to
the formation of either GAP or E4P while leaving behind the bound dihydroxyacetone (C3).
The nitrogen atom of the Schiff base (similar to the nitrogen atom in the thiazole ring of
transketolase) stabilizes the dihydroxyacetone unit using resonance forms until a suitable
aldose (GAP, E4P) acceptor is bound [37].
The kinetic mechanism employed by both enzymes has been characterized as a reversible
ping‑pong mechanism [14, 18, 49]. Bi‑bi reactions use this mechanism to shuttle molecule
fragments from one compound to another, epitomized by the fact that the first substrate
is released from the holoenzyme before the second substrate binds. For the enzymes
transketolase and transaldolase this implies that the cleaved phosphorylated fragment of the
ketose substrate is first detached from the enzyme, before the stabilized carbon fragment
(glycoaldehyde for transketolase and dihydroxyacetone for transaldolase) is donated to a
suitable aldose acceptor. This mechanism is in conflicts with the traditional reactions. The
structure of the traditional reactions is such that a C2 or C3 fragment is transferred from one
specific donor to one specific acceptor molecule. This reaction structure is in agreement
with a so‑called ordered sequential kinetic mechanism. The difference between a sequential
and a ping‑pong kinetic mechanism is illustrated in Figure 1. Whereas the correct ping‑pong
mechanism for transketolase and transaldolase was adopted by several researchers in the
1990’s to construct detailed kinetic models [1, 22, 23, 27], this has been largely overlooked by
the metabolic engineering community.
In accordance with the ping‑pong mechanism employed by transaldolase and transketolase,
the traditional reactions of the non‑oxidative branch of the PPP can be represented as
metabolite specific, reversible C2 and C3 fragments producing and consuming half‑reactions
for each of the metabolites S7P, F6P, X5P, R5P, E4P and GAP (r.10‑14). Note that the C2 and
C3 fragments remain bound to the holoenzyme (E) until they are transferred to an acceptor.
TK
x5p ↔ gap + E −C2
TK
f6p ↔ e4p + E −C2
TK
s7p ↔ r5p + E −C2
TA
f6p ↔ gap + E −C3
TA
s7p ↔ e4p + E −C3
(r.10)
(r.11)
(r.12)
(r.13)
(r.14)
Using the above half‑reactions, a C2 fragment producing reaction (e.g. x5p → gap + E −C2 )
can be coupled to a C2 fragment consuming reaction (e.g. e4p + E −C2 → f6p ), leading to one
of the traditional reactions (in this case r.2: x5p + e4p → f6p + gap ). In total thirteen different
combinations of half‑reactions are possible; the three C2 fragment donating half‑reactions
can be combined with three C2 fragment accepting half‑reactions and the two C3 fragment
33
Chapter 2
34
donating half‑reactions can be combined with the two C3 fragment accepting half‑reactions,
leading to the three conventional reactions (r.1‑3) and the six additional reactions (r.4‑9).
Interestingly, half‑reactions r.10‑14 can be used to show that a stoichiometric model for
the non‑oxidative branch of the PPP based upon traditional reactions r.1‑3 is in essence
incomplete. In order to perform these three reactions in forward and backward direction, all
five proposed half‑reactions (r.10‑14) are needed. The reversibility of the traditional reactions
was argued by Follstad et al. [11]; a claim supported by most textbooks [25, 37]. However,
recombination of the half‑reactions into their traditional counterparts leads to nine reversible
reactions (r.1‑9), as shown in the previous paragraph. Therefore, given the reversibility of
the transketolase‑ and transaldolase‑catalyzed reactions and their demonstrated ping‑pong
mechanism, one has to conclude that in addition to traditional reactions r.1‑3, one should also
incorporate the other six traditional reactions (r.4‑9) when constructing a stoichiometric model
for the non‑oxidative branch of the PPP.
A
A
K
A
C
K
K
C
A
E
K
K
C
A
E
B
C
C
E
E
A
K
C
A
E
E
A
K
C
K
E
C
A
E
K
K
A
E
C
C
A
E
K
C
A
E
Figure 1 Schematic representation of the two kinetic mechanisms used for modeling the
transketolase- and transaldolase-catalyzed reactions of the PPP: ping-pong mechanism (A) and
(ordered) sequential mechanism (B). Depicted are the ketose substrate (diamond K), the aldose
acceptor (triangle A), the transferred carbon-fragment (circle C) and the enzyme/cofactor complex
(form E).
Revisiting the Pentose Phosphate Pathway
2.2.2
Traditional vs. half‑reactions: implications for 13C‑labeling
From a labeling point of view, the main difference between modeling the stoichiometry of the
non‑oxidative branch of the PPP using either traditional reactions or half‑reactions, is the
number of independent C2 and C3 fragment pools each approach generates. The traditional
reactions will lead to separate C2 and C3 fragment pools for each of the nine possible reactions
(r.1‑9), while the half‑reactions by definition lead to only one C2 and one C3 fragment pool,
from which carbon fragments are retrieved and attached to any suitable acceptor (Figure
2). As the number of non‑oxidative PPP reactions increases, application of the traditional
reactions leads to an increase in the number of distinct C2 and C3 fragment pools. Due to
these segregated pools, the 13C‑labeling of the C2 and C3 fragments (and consequently, the
labeling of the metabolites formed from these) can differ from the 13C‑labeling of the single C2
and C3 fragment pools generated by the half‑reactions.
A
1
2
n
E -C 2 E - C 2 ....E - C 2
Transketolase
B
E -C 2
1
2
m
E - C 3 E -C 3 ....E - C 3
Transaldolase
Figure 2 Number of C2 and
C3 fragment pools in the nonoxidative branch of the PPP
based upon a stoichiometric
model
constructed
from
traditional reactions (A) and
half-reactions (B). The number
of C2 and C3 fragment producing
reactions when applying the
traditional reactions is denoted
by n and m, respectively.
E -C 3
2.2.3
Genetic organization of the non‑oxidative branch of the PPP
In recent years, the genes encoding the enzymes of the non‑oxidative branch of the PPP
have been sequenced and cloned for many micro-organisms. It was found that several
micro-organisms, including Escherichia coli and S. cerevisiae, contain two transketolase
genes, named tkl1 and tkl2 in S. cerevisiae [30, 38] and tktA and tktB in E. coli [35]. The
combined fact that several micro-organisms possess two transketolase‑genes and that most
stoichiometric flux balance models of the non‑oxidative branch of the PPP contain only two
transketolase‑catalyzed reactions (r.1‑2), has led to the common misunderstanding that each
reaction is catalyzed by a separate transketolase (either tkl1 or tkl2). In several publications
it is assumed that the transketolase encoded by tkl1/tktA specifically catalyzes the reversible
reaction r.1, while the transketolase encoded by tkl2/tktB catalyzes the reversible reaction r.2
[9, 28, 32, 34, 50, 52]. In reality the transketolase gene products in S. cerevisiae and E. coli
are isoenzymes, each of which is capable of non‑specifically catalyzing both reactions r.1 and
r.2 in the non‑oxidative branch of the PPP [3, 6, 12, 16, 17]. As expected the two isoenzymes
of S. cerevisiae and E. coli show a strong resemblance with respect to amino acid residues;
35
Chapter 2
36
homology was measured to be 71% [30] and 74% [35], respectively.
The presence of isoenzymes for transaldolase has been studied to a lesser extent. Microorganisms containing multiple genes with transaldolase activity do exist, an example being
E. coli, which contains two isoenzymes for transaldolase (talA/talB) [16, 17]. The talB gene of
E. coli has been shown to encode a functional transaldolase [36], while the functionality of the
talA gene has not been shown up to date. S. cerevisiae contains one verified transaldolase
gene named tal1 [29]. Recently a hypothetical ORF for a possible second transaldolase was
found [3, 16].
Using this genetic information the stoichiometric model for the non‑oxidative branch of the
PPP can be further refined. Although homology between isoenzymes is normally quite high,
differences in substrate affinity are common [24]. If evidence for isoenzymes of transketolase
and/or transaldolase exists, one can opt for a model with two sets of half‑reactions, in which
each set of half‑reactions models the transfer of the C2 or C3 fragments for one isoenzyme.
As a result of this modification, a second set of C2 and C3 fragment pools is created in the
non‑oxidative branch of the PPP. Note that genetic evidence alone is not sufficient proof
for the actual expression of isoenzymes, this expression should be verified under relevant
cultivation conditions. Literature shows that in S. cerevisiae the activity of the tkl2‑encoded
transketolase appears to be very low when growing cells on a synthetic batch mineral salts
medium with glucose as carbon‑source [30]. Furthermore, deletion mutants of tkl2 showed no
changed phenotype, while deletion mutants of tkl1 displayed a declined growth‑rate [31]. A
similar trend was found for the isoenzymes of E. coli, where the tktA‑encoded transketolase
and talB‑encoded transaldolase accounted for the majority of the cellular activity [15, 33, 51].
2.2.4
Model construction and analysis
Using the five half‑reactions (r.10‑14) a new stoichiometric model for the combined glycolysis
and PPP was constructed as shown in Figure 3‑II (From hereon referred to as the half‑reaction
model). Note that this model does not yet take into account the presence of isoenzymes for
transketolase and transaldolase, as it only contains a single C2 and C3 fragment pool. As a
comparison, Figure 3‑I shows the equivalent stoichiometric model based upon the traditional
reactions (henceforth called the traditional model). Note that this model contains both the
conventional non‑oxidative PPP reactions (Figure 3‑IA) as well as the six additional reactions
proposed by van Winden et al. [41] (Figure 3‑IB). The traditional model has previously been
used to fit the metabolic fluxes of P. chrysogenum and S. cerevisiae [42, 43].
The half‑reaction model of Figure 3‑II covers the complete range of possible reactions, yet it
significantly reduces the number of free fluxes that have to be estimated from the 13C‑labeling
data during the flux fitting procedure. The model contains twelve reactions (n1‑n12) and eight
reversibilities, which are constrained by ten mass balances over the intracellular metabolites in
a (pseudo) steady state. When normalizing the rates relative to the uptake rate of glucose, nine
free fluxes remain to be estimated from the 13C‑labeling data. The corresponding traditional
model (Figure 3‑I) contains sixteen reactions (v1‑v16) and seven reversibilities. Under (pseudo)
steady state conditions eight reaction rates are fixed by mass balances over the intracellular
metabolites. Normalization of the fluxes to the glucose uptake rate thus leaves fourteen free
fluxes.
Revisiting the Pentose Phosphate Pathway
The half‑reaction model can be extended with a second set of half‑reactions to account for the
possible presence of isoenzymes for transketolase (r.10‑13) and/or transaldolase (r.14‑15).
This extension will increase the number of free fluxes that have to be estimated from the
13
C‑labeling data. In the case of two actively expressed genes for transketolase this will result
in five additional free fluxes, since the six additional half‑reactions are constrained by one
extra mass balance over the second C2 fragment pool. As a result, the total number of free
fluxes (14) equals the number of free fluxes in the traditional model.
I
A
g1p
glc
v 2 f,b
I
v1
v7
g6p
B
p5p
p5p
v 3 f,b
v 11
f6p
v 9 f,b
f6p
v4
v 15
fbp
v 10 f,b
v 12
v 8 f,b
v 13
e4p
e4p
v 16
v 14 f,b
v 5 f,b
gap
gap
s7p
s7p
v6
towards lower
glycolysis /TCA
II
g1p
glc
n1
n 2 f,b
n7
g6p
n 3 f,b
f6p
n4
n 9 f,b
p5p
n 10 f,b
E -C 2
n 8 f,b
s7p
n 11 f,b
fbp
n 5 f,b
gap
n6
towards lower
glycolysis /TCA
E -C 3
n 12 f,b
e4p
Figure 3 Traditional (I) and half-reaction
(II) stoichiometric flux balance models
for the upper glycolysis and PPP. The
non-oxidative PPP reactions of the
traditional model are split-up into the three
conventional reactions (r.1-3) (IA) and the
six additional reactions (r.4-9) (IB). Closed
arrows denote the direction of the forward
flux in the case of reversible reactions.
P5P: the pentose pool consisting of
xylulose-5-phosphate, ribose-5-phosphate
and ribulose-5-phosphate, G1P: glucose1-phosphate, TCA-cycle: tricarboxylic acid
cycle. See text for other abbreviations.
37
Chapter 2
38
2.3 MATERIALS AND METHODS
2.3.1
Transketolase experiments
Transketolase A (kindly donated by Dr. R. Schoevaart, Department of Organic Chemistry, Delft
University of Technology) was purified from recombinant Escherichia coli K12 cells carrying
the homologous cloned tktA gene on a PUC19-derived plasmid, analogous to the method used
by Sprenger et al. (1995). The reaction mixture for the transketolase experiments consisted
of 5 units/mL of transketolase A solution, 250 μM of each substrate (GAP, R5P, X5P, F6P
and/or G6P), 500 μM MgCl2 and 100 μM thiamine pyrophosphate (TPP). The total volume of
the reaction mixture equaled 1 mL and was buffered using 50 mM triethanolamine (TEA) (pH
8.3). The samples were incubated for 20 minutes at 30oC, after which they were immediately
stored on ice.
Unconsumed substrate(s) and formed product(s) were first separated by high-performance
anion exchange chromatography, using an Alliance pump system (Waters, Milford, USA)
followed by an IonPac AS11 (250 x 4 mm) column equipped with an AG11 (50 x 4 mm) guardcolumn (both from Dionex, Sunnyvale, CA., USA). The flow rate was 1mL/min.
Subsequent MS analyses were performed with a Quatro-LC triple quadrupole mass
spectrometer (Micromass Ltd., Manchester, UK) equipped with an electrospray ionization
interface with a mass range up to 1600 m/z. All samples were analyzed in the negative mode
giving [M-H]- ions, which were either monitored in the multiple reaction monitoring (MRM)
mode or in the single ion recording (SIR) mode. When monitoring in the MRM-mode the [M-H]ions were fragmented by collisionally induced dissociation (CID) with argon at a pressure of
5.10-4 mbar. In all cases 10 μl of reaction mixture was injected. Further details on the applied
method are given in van Dam et al. [39].
2.3.2
Metabolic network model
Apart from the variations in the stoichiometric model of the PPP discussed in this work,
the other parts of the stoichiometric model used for fitting the fluxes of S. cerevisiae were
identical to that presented by van Winden et al. [43]. For simplicity reasons the consumption
of metabolites for the synthesis of biomass precursors and the reversible flux towards storage
carbohydrates are not shown in the metabolic network model depicted in Figure 3, but these
were accounted for when fitting 13C‑labeling data. The reversible reactions in Figure 3 were
modeled as separate forward and backward reactions and are referred to as net and exchange
fluxes, where:
v net = v forward − v backward
(2.1)
v exchange = min(v forward ,v backward )
(2.2)
2.3.3
Flux fitting procedure
The employed flux fitting procedure is described in detail by van Winden et al. [43]. In short,
the procedure uses the cumomer balances and cumomer to isotopomer mapping matrices
introduced by Wiechert et al. [45] to calculate the isotopomer distributions of metabolites in
a pre‑defined metabolic network model for a given flux‑set. The flux‑set that gives the best
Revisiting the Pentose Phosphate Pathway
correspondence between the measured and simulated 13C‑label distribution is determined by
non‑linear optimization and denoted as the optimal flux‑fit. All calculations were performed in
Matlab 6.1 (The Mathworks Inc, Natick, USA).
2.4 RESULTS AND DISCUSSION
2.4.1
Experimental evidence for additional transketolase reactions
In order to further justify the incorporation of reaction r.4 in the traditional model of the non
oxidative PPP as proposed by van Winden et al. [41], the substrate-specificity of transketolase
was tested in six in vitro enzymatic conversion experiments (Table 1). The substrate-specificity
of transketolase was tested in experiments I and IV, while the other four experiments served
as control experiments.
Table 1 Results of the six in vitro enzymatic conversion experiments catalyzed by transketolase.
The second column indicates whether the reaction (R) was allowed to run to completion or stopped
immediately after the addition of the substrate(s). The third column indicates whether transketolase (TK)
was added to the reaction mixture. For each measured metabolite is shown if its presence is expected
(E) and observed (O).
Expected
product(s)
PPP intermediates with LC-MS/MSa
Exp#
R
TK
Added
substrate(s)
E
O
E
O
E
O
I
+
+
R5P + F6P
GAP + S7P
+
+
+
+
+
+
II
+
-
R5P + F6P
none
+
+
+
+
-
-
P5Pc
F6P
S7P
III
-
+
R5P + F6P
none
+
+
+
+
-
-
IV
+
+
GAP + F6P
X5P + E4P
+
+
+
+
-
±b
V
+
-
GAP + F6P
none
-
-
+
+
-
-
VI
+
+
X5P
none
+
+
-
-
-
±b
GAP and E4P could not be measured with this method.
Trace amounts of this compound were detected.
c
X5P and R5P could not be distinguished with this method.
a
b
In experiment I we clearly observed the formation of S7P after the addition of F6P and R5P
as substrate, indicating that transketolase is indeed capable of catalyzing this reaction. This
supports the inclusion of this ‘missing’ stoichiometric non-neutral reaction in the metabolic
network model of the non-oxidative branch of the PPP. Experiments II and III served as
negative controls; in experiment II transketolase was absent from the reaction mixture, while
in experiment III the reaction was immediately halted after the addition of the substrates
by storing the reaction mixture on ice. As expected, no products were formed in these two
experiments.
To confirm that the enzyme used was indeed transketolase, experiment IV was performed
with F6P and GAP as substrates. The expected conversion would be one of the traditional
PPP, the conversion of F6P and GAP to the ketose X5P and the aldose E4P (r.2). As shown
in Table 1, the enzyme was indeed capable of catalyzing this conversion. However, besides
39
Chapter 2
40
the expected products, small amounts of S7P were formed. A possible explanation for this
phenomenon might be contamination of the transketolase with the enzymes phosphopentose
epimerase and isomerase. These enzymes catalyze the epimerization and isomerization of
X5P into R5P, which can be converted by the conventional transketolase-catalyzed reaction
(r.1) into S7P and GAP. As the used LC-MS method could not distinguish between R5P and
X5P, a transketolase experiment with solely X5P as a substrate was performed (experiment
VI). Even in the presence of one substrate, S7P was formed, suggesting indeed the presence
of some phosphopentose epimerase and isomerase activity. It must be noted that the S7Pconcentrations measured in experiments IV and VI were small (5-10%) compared to the
amount formed in experiment I.
2.4.2
Presence of octulose-8-phophate
Similar to the above experiments, it was tested if the eight-carbon sugar-phosphate, octulose8-phosphate (O8P), could be produced in a transketolase-catalyzed reaction. Addition of F6P
and G6P to a reaction mixture with transketolase and subsequent metabolite analysis did
indeed show the formation of a compound with the same molecular mass as O8P (Mw = 318
g/mol). Furthermore, fragmentation of the compound indicated the formation of a product-ion
with the same molecular mass as phosphate (H2PO4-, Mw = 97 g/mol).
Extension of the traditional metabolic network model with the transketolase and transaldolase
reactions for O8P results in an additional 7 reactions, four for transketolase (r.15-17 and r.20)
and three for transaldolase (r.18-19 and r.21). Under (pseudo) steady state the fluxes of these
7 reactions and 5 reversibilities are only constrained by one additional mass balance for O8P,
resulting in 11 extra parameters to be fixed by the 13C-labeling data.
TK
f6p + g6p  o8p + e4p
TK
x5p + g6p  gap + o8p
TA
r5p + f6p  o8p + gap
TK
g6p + s7p  o8p + r5p
TA
s7p + r5p  e4p + o8p
TK
g6p + o8p → o8p + g6p
TA
r5p + o8p → o8p + r5p
(r.15)
(r.16)
(r.17)
(r.18)
(r.19)
(r.20)
(r.21)
Modeling of these new transketolase and transaldolase reactions according to a ping-pong
kinetic mechanism yields only two new reactions, one reversible reaction for cleavage of a C2
fragment from O8P (r.22) and another reversible reaction for the splitting-off of a C3 fragment
(r.23). Thus, when calculating the flux-patterns in the half-reaction model only three additional
parameters have to be estimated from the 13C-labeling data.
Revisiting the Pentose Phosphate Pathway
TK
o8p  g6p + E −C2
TA
o8p  r5p + E −C3
(r.22)
(r.23)
Clearly, the extension of the half-reaction model with transketolase- and transaldolasecatalyzed reactions for O8P results in a smaller number of additional parameters that have to
be fitted by 13C-labeling data. However, since the presence of O8P could not be conclusively
tested due to the absence of a commercially available standard and since the intensities of
the O8P mass peaks were rather low, its formation was not included in the metabolic network
models discussed in the remainder of this study.
2.4.3
Traditional vs. half‑reaction model: three theoretical cases
To illustrate the difference in 13C‑labeling distribution when using either traditional reactions
or half‑reactions to model the non‑oxidative branch of the PPP, three simplified metabolic
networks were formulated (cases 1‑3). Note that the three networks are oversimplified and
are solely used to clarify the difference in 13C‑label distribution that can occur between the two
different modeling approaches.
For case 1 consider the traditional model of Figure 3, but now containing only the conventional
non‑oxidative PPP reactions (Figure 3‑IA). The reversibilities of the three bidirectional
non‑oxidative PPP reactions and the three bidireactional glycolytic reactions are set at zero,
such that the PPP overall converts three P5P molecules into two F6P molecules and one
GAP molecule. Consequently, only the forward reactions of the PPP (v8f, v9f, v14f) and the
glycolysis (v2f, v3f, v5f) are active. Analogous to the traditional model only the forward glycolytic
reactions are included in the half‑reaction model (n5f, n3f and n2f). Using the relations in
Appendix A, the active non‑oxidative PPP reaction rates in the traditional model are converted
to the corresponding rates in the half‑reaction model, resulting in substantial throughput for
half‑reactions n8f, n9b, n10b, n11b, n12f. Investigation of the acceptor and the donor of the C2
fragment in both models, shows that the traditional model contains two C2 fragment pools
created by reactions v8f and v9f, while the half‑reaction model by definition contains one single
C2 fragment pool that is solely formed by reaction n8f (Figure 4). However, both C2 fragment
pools in the traditional model are formed by the cleavage of P5P and can thus be lumped
into a single pool, resulting in identical C2 fragment pools for both modeling approaches.
Examination of the origin of the C3 fragment pools shows that both models contain only one
C3 fragment‑producing reaction, both with S7P as donor (v14f, n12f). So in essence both models
described in this case contain one C2 and one C3 fragment pool. As a result the redistribution
of 13C‑atoms in the PPP is identical for both models.
For case 2 consider the same traditional model as used in case 1, supplemented with the
stoichiometric neutral exchange reaction for E4P and F6P (v12 in Figure 3‑IB). In the half‑reaction
model this means an increase in n9f and n9b (see Appendix A). Due to this additional reaction,
C2 fragments are now also produced from F6P, thus increasing the number of C2 fragment
pools in the traditional model to three (Figure 4). The absence of bidirectional reactions makes
it impossible for the three C2 fragment pools, originating from either P5P or F6P, to efface
their labeling differences. A different labeling of F6P (in comparison to P5P) therefore by
41
Chapter 2
42
necessity leads to two unique C2 fragment pools in the traditional model. The half‑reaction
model inherently contains one single C2 fragment pool that comprises all distinct C2 fragment
pools of the traditional model, as shown in Figure 4. From this single pool a C2 fragment is
randomly retrieved and attached to any suitable acceptor. Consequently, the top two carbon
atoms of S7P synthesized in the half‑reaction model can originate from either F6P or P5P,
while in the traditional model they can only originate from P5P. In a 13C‑labeling experiment
with 100% 13C1‑glucose this will result in the synthesis of unlabeled and 13C1‑labeled S7P for
the half‑reaction model, contrary to only unlabeled S7P for the traditional model.
For case 3 consider the same traditional and half‑reaction model as used in case 2, but now
with all bidirectional reactions set at maximum reversibility (99.9%). Due to this reversibility
assumption the number of C2 fragment‑producing reactions in the traditional model increases
from two to four (v8f, v8b, v9f, v9b). However, the high reversibility of the bidirectional reactions
also ensures that the label distributions of the C2 fragment pools (and also the C3 fragment
pools) are fully exchanged, effacing the differences in labeling pattern amongst the separate
p5p
p5p
v 9f
=
C2
v 9f
p5p
v 8f
n 8f
Case 1
C2
C2
v 8f
C2
v9f
f6p
f6p
v8f
C2
v8f
s7p
f6p
p5p
p5p
=
s7p
Half-reaction Model
Traditional Model
v9f
n 10b
f6p
s7p
f6p
p5p
n 9b
=
v12
C2
n 8f
n 9f
Case 2
v12
n 9b
C2
n 10b
f6p
f6p
s7p
Figure 4 Route traversed by the C2 fragments of the transketolase-catalyzed reactions present
in the simplified traditional model and half-reaction model of cases 1 and 2 (see main text). The
colored spheres represent the carbon atoms of which the C2 fragment is constructed. A different
13
C-labeling of the C2 fragment is denoted by a different color. Consequently, the 13C-labeling of the
top two-carbon fragments of the P5P and F6P depicted in this figure is different.
Revisiting the Pentose Phosphate Pathway
pools. As a result, no difference in isotopomer distribution is observed between the two models
under conditions of high reversibility.
The three cases discussed above show that the difference in 13C‑label distribution amongst the
two modeling approaches becomes more pronounced as the number of C2 and C3 fragment
producing reactions increases, while high reaction reversibilities diminish this difference. In
reality the non‑oxidative branch of the PPP contains multiple C2 and C3 fragment producing
reactions, thereby in essence creating different 13C‑label distributions. As shown in case 3,
these differences can be alleviated by high reversibilities for the non‑oxidative PPP reactions.
Even though the reversibility of these reactions was argued by Follstad et al. [11], it remains
questionable whether this reversibility is high enough to efface the difference in 13C‑label
distribution created by the multiple C2 and C3 fragment producing reactions.
2.4.4
Application of the half‑reaction model: Flux‑patterns in S. cerevisiae
To investigate the actual difference in estimated flux‑patterns when applying either the
traditional model or the half‑reaction model shown in Figure 3, measured mass isotopomers
of 13C‑labeled primary metabolites [43] were used to re‑fit the fluxes in the glycolysis and the
PPP of S. cerevisiae CEN.PK113‑7D. Similarly to the previously published fit, only measured
mass isotopomer fractions larger than 0.03 were included. Figure 5‑I&II and Table 2 show the
previously estimated flux‑patterns for the traditional model, as well as the newly estimated
flux‑patterns using the half‑reaction model. In order to facilitate the comparison of the two
flux‑sets in Table 2, the flux‑estimates for the traditional model have been converted into their
corresponding half‑reaction rates using the equations given in Appendix A. The difference in
flux‑pattern is evident, though on average not very large. As expected, the largest differences
are found for the PPP split‑ratio and the fluxes of the non‑oxidative branch of the PPP.
The minimized covariance‑weighted sum of squared residuals (SSres) in these fits was
calculated to be 20.9 and 6.5 for the half‑reaction and traditional model, respectively. The SSres
is distributed according to a χ2(n‑p) distribution, with n‑p being the degrees of freedom, equal
to the number of independent data points (n=26) minus the number of free parameters (p=14
and 9 for the traditional and the half‑reaction model, respectively). Given the probabilities P(
χ2 (12) > 6.5 ) =0.89 and P( χ2(17) > 20.9 ) = 0.23, it follows that within the 95% confidence
interval both models give statistically acceptable flux‑estimates. Even though both models are
statistically acceptable, it must be noted that the discrepancy between the measured and the
fitted mass isotopomers (SSres) is higher for the half‑reaction model. A possible explanation for
the bigger SSres in the half‑reaction model is an overparameterization of the traditional model.
In an overparameterized model some parameters are actually used to fit measurement errors,
thereby underestimating the true SSres [2]. To determine the extent of this overparameterization
the estimated error variance (s2res) criterion can be used:
s2res =
SSres
n−p
(2.3)
This criterion minimizes the variance of the sum of squared residuals by dividing the SSres of a
model by its degrees of freedom. Since the traditional model contains more parameters than
43
Chapter 2
the half‑reaction model, this will result in a smaller denominator for s2res, thus compensating
for any possible overparameterization. Nevertheless, the traditional model gives a s2res of 0.54
compared to 1.23 for the half‑reaction model, implying that the traditional model performs
better from a statistical point of view.
44
IA
IB
glc
p5p
100
g1p
26(105)
24
g6p
p5p
8
f6p
50(>1000)
6(1)
f6p
7(2)
63
e4p
0
8(0)
63(194)
>1000
e4p
24
fbp
1(5)
148
gap
gap
s7p
s7p
119
towards lower
glycolysis /TCA
II
III
glc
100
g1p
100
26(134)
18
g6p
p5p
g1p
56(>1000)
f6p
glc
-3(10)
-6(124)
E -C 2
26(103)
50(>1000)
f6p
9(4)
65
s7p
-6(25)
fbp
65(221)
gap
121
towards lower
glycolysis /TCA
E -C 3
-6(98)
0(1)
63
p5p
E1 - C 2
E 2 -C 2
6(0)
8(10)
0(6)
-8(5)
s7p
-8(24)
6(0)
e4p
24
g6p
fbp
63(199)
8(0)
E -C 3
e4p
gap
118
towards lower
glycolysis /TCA
Figure 5 Fitted fluxes for the traditional model (I), half-reaction model (II) and ‘double TK’ halfreaction model (III) based upon the mass isotopomer measurements of 13C-labeled primary
metabolites as presented in van Winden et al. [43]. Fluxes are normalized for the glucose-uptake
rate. Values outside brackets denote the net fluxes, while values between brackets represent the
exchange fluxes. Solid arrowheads denote the direction of the net flux. See text for abbreviations.
Revisiting the Pentose Phosphate Pathway
Table 2 Comparison of the flux-estimates for the traditional and the half-reaction model presented in
Figure 5-I&II. The PPP fluxes in the traditional model have been converted to their corresponding fluxes
in the half-reaction model using the equations in Appendix A.
Reaction #
n1
Fluxes in half-reaction
model
Converted fluxes in
traditional model
Relative Change (%)
100
100
0
n2 net
26
26
0
n2 exch
134
105
21
n3 net
56
50
11
>1000
>1000
-
n4
65
63
3
n5 net
65
63
3
n5 exch
221
194
13
n6
121
119
2
n7
n3 exch
18
24
36
n8 net
9
13
48
n8 exch
4
10
>100
n9 net
-3
-5
67
n9 exch
10
155
>100
n10 net
-6
-8
38
124
4898
>100
n11 net
-6
-8
38
n11 exch
25
24
4
n12 net
6
8
38
n12 exch
0
0
0
n10 exch
A second explanation for the higher SSres found for the half‑reaction model might be the
presence of isoenzymes for transketolase. As stated before, the genome of S. cerevisiae
contains two genes encoding for a transketolase, which adds a second C2 fragment pool
to the metabolic network model. To test whether the introduction of an isoenzyme for
transketolase in the metabolic network model results in a better fit, the half‑reaction model
in Figure 3 was expanded with a second set of transketolase half‑reactions (r.10‑12) and
subsequently used to fit the measured mass isotopomer fractions of S. cerevisiae. Figure
5‑III shows the estimated reaction rates for the so‑called ‘double TK’ half‑reaction model. The
SSres for this model was 6.5, meaning that this model also adequately fitted the measured
mass isotopomer fractions (P( χ2 (12) > 6.5 ) =0.89). Interestingly, exactly the same values
for the minimized SSres and the number of free parameters (14) were found for both the
‘double TK’ half‑reaction and the traditional model, making it impossible to distinguish the two
models using the s2res criterion. Table 3 shows that the flux‑estimates for both models were
also very similar. The resemblance between the two models can be understood when one
realizes that both models, unlike the half‑reaction model, have the ability to create separate
C2 fragment pools. Considering the reported finding that tkl1 encodes for the majority of the
transketolase activity in S. cerevisiae cells grown in synthetic mineral medium on glucose, it
45
Chapter 2
was not anticipated that the addition of a transketolase isoenzyme to the metabolic network
model would result in an increased goodness‑of‑fit. It must be noted that the prevalence of
the tkl1‑encoded transketolase was measured under excess glucose conditions, while the
13
C‑labeling experiment was performed in a chemostat under glucose‑limiting conditions.
46
Table 3 Comparison of the flux-estimates for the traditional and the ‘double TK’ half-reaction model
presented in Figure 5-II&III. The two separate fluxes for the transketolase-catalyzed half reactions in the
‘double TK’ half-reaction model have been summed to allow for comparison with the converted fluxes of
the traditional model shown in Table 2.
Reaction #
n1
Converted fluxes
in 'double TK’ halfreaction model
Converted fluxes in
traditional model
100
Relative Change (%)
100
0
n2 net
26
26
0
n2 exch
103
105
2
n3 net
50
50
1
>1000
>1000
-
n4
63
63
0
n5 net
63
63
0
n5 exch
199
194
3
n6
118
119
0
n7
25
24
2
n8 net
13
13
3
n8 exch
10
10
0
n9 net
-5
-5
4
55
n3 exch
n9 exch
100
155
n10 net
-8
-8
3
n10 exch
11
4898
>100
n11 net
-8
-8
3
n11 exch
24
24
2
n12 net
8
8
3
n12 exch
0
0
0
2.5 Conclusion
This study shows that a good understanding of enzyme genetics and kinetics is crucial
for a correct 13C‑label distribution prediction in stoichiometric flux balance models. When
comparing two models of the non‑oxidative branch of the PPP based, respectively, on the
traditional reactions and the kinetically derived half‑reactions, it was demonstrated that the
main difference between the two reaction structures is the number of independent C2 and C3
fragment pools present in the stoichiometric model. Whereas the traditional reactions lead
to multiple independent pools, the half‑reactions result in only one C2 and one C3 fragment
Revisiting the Pentose Phosphate Pathway
pool. This difference in C2 and C3 fragment pools influences the ensuing label distribution
when conducting 13C‑tracer experiments. An additional advantage of using half‑reactions is
the decreased number of free parameters that have to be estimated by fitting 13C‑labeling data
to the stoichiometric model.
Mass isotopomer measurements from a previously published study on S. cerevisiae were
used to compare the traditional and half‑reaction model depicted in Figure 3, resulting in
statistically acceptable fits for both models. Different flux‑patterns were found for the two
models, but no major rerouting of metabolic fluxes was observed. The incorporation of
genetic knowledge into the metabolic network model for the non‑oxidative branch of the PPP
introduced the possibility of modeling the presence of isoenzymes for transketolase and
transaldolase. Extending the half‑reaction model from one to two autonomously functioning
transketolase enzymes, resulted in a doubling of the number of C2 fragment pools. The fitting
of measurement data to a ‘double TK’ half‑reaction model yielded flux‑estimates and a SSres
similar to the traditional model. The similarity of the flux estimate indicates that the presence
of isoenzymes reduces the difference in 13C‑label distribution between the two models and
impedes their discrimination. This shows that for S. cerevisiae more accurate measurement
techniques are needed to discriminate between the different stoichiometric models for the
non‑oxidative branch of the PPP, in combination with genetic and biochemical evidence on
the number of active transketolase and transaldolase isoenzymes under the experimental
conditions used. In spite of their practical similarity, clear differences between the traditional
and half‑reaction model were illustrated by means of three theoretical cases. Therefore,
considering the established ping‑pong mechanism of transketolase and transaldolase, we
recommend the use of the half‑reaction model when modeling the label distribution in the
non‑oxidative PPP, keeping in mind that isoenzymes for transketolase and transaldolase may
exist.
ACKNOWLEDGEMENTS
This work was financially supported by the Dutch EET program (Project No. EETK20002) and
DSM.
47
Chapter 2
Appendix A: Relations between the non‑oxidative PPP fluxes of the traditional model
and the half‑reaction model
48
From the traditional and the half‑reaction model of the non‑oxidative PPP depicted in Figure 3,
linear dependencies can be derived relating the non‑oxidative PPP fluxes of the two models.
These non‑redudant linear dependencies are given in Eq. A-1‑10.
n8f = v 8f + v 9f + v11
(A-1)
n8b = v 8b + v 9b + v11
(A-2)
n9f = v 9b + v10f + v12
(A-3)
n9b = v 9f + v10b + v12
(A-4)
n10f = v 8b + v10b + v13
(A-5)
n10b = v 8f + v10f + v13
(A-6)
n11f = v14b + v15
(A-7)
n11b = v14f + v15
(A-8)
n12f = v14f + v16
(A-9)
n12b = v14b + v16
(A-10)
Revisiting the Pentose Phosphate Pathway
REFERENCES
[1] Berthon, H.A., Bubb, W.A. and Kuchel, P.W. (1993)
13
C NMR isotopomer and computer simulation studies
of the nonoxidative pentose phosphate pathway of
human erythrocytes. Biochem. J. 296, 379-387.
[2] Burnham, K.P. and Anderson, D.R. (2002) Model
selection and multimodel inference : a practical
information-theoretic approach. Springer, New York.
[3] Cherry, J.M., Adler, C., Ball, C., Chervitz, S.A.,
Dwight, S.S., Hester, E.T., Jia, Y.K., Juvik, G.,
Roe, T., Schroeder, M., Weng, S.A. and Botstein,
D. (1998) SGD: Saccharomyces Genome
Database. Nucleic Acids Res. 26, 73-79.
[4] Christensen, B. and Nielsen, J. (2000) Metabolic
network analysis of Penicillium chrysogenum using
13
C-labeled glucose. Biotechnol. Bioeng. 68, 652-659.
[5] Clark, M.G., Williams, J.F. and Blackmore,
P.F.
(1971)
Transketolase
exchange
reaction in vitro. Biochem. J. 125, 381-384.
[6] [7] [8] [9] Daran-Lapujade, P., Jansen, M.L.A., Daran,
J.M., van Gulik, W., de Winde, J.H. and Pronk,
J.T. (2004) Role of transcriptional regulation in
controlling fluxes in central carbon metabolism
of Saccharomyces cerevisiae - A chemostat
culture study. J. Biol. Chem. 279, 9125-9138.
dos Santos, M.M., Gombert, A.K., Christensen, B.,
Olsson, L. and Nielsen, J. (2003) Identification of in
vivo enzyme activities in the cometabolism of glucose
and acetate by Saccharomyces cerevisiae by using
13
C-labeled substrates. Eukaryot. Cell 2, 599-608.
Fiaux, J., Andersson, C.I.J., Holmberg, N., Bulow, L.,
Kallio, P.T., Szyperski, T., Bailey, J.E. and Wuthrich,
K. (1999) 13C NMR flux ratio analysis of Escherichia
coli central carbon metabolism in microaerobic
bioprocesses. J. Am. Chem. Soc. 121, 1407-1408.
Fiaux, J., Cakar, Z.P., Sonderegger, M., Wuthrich,
K., Szyperski, T. and Sauer, U. (2003) Metabolic-flux
profiling of the yeasts Saccharomyces cerevisiae
and Pichia stipitis. Eukaryot. Cell 2, 170-180.
[10] Flanigan, I., Collins, J.G., Arora, K.K., Macleod,
J.K. and Williams, J.F. (1993) Exchange reactions
catalyzed by group transferring enzymes oppose the
quantitation and the unraveling of the identity of the
pentose pathway. Eur. J. Biochem. 213, 477-485.
[11] Follstad, B.D. and Stephanopoulos, G. (1998)
Effect of reversible reactions on isotope label
redistribution - Analysis of the pentose phosphate
pathway. Eur. J. Biochem. 252, 360-371.
[12] Forster, J., Famili, I., Fu, P., Palsson, B.O.
and
Nielsen,
J.
(2003)
Genome-scale
reconstruction of the Saccharomyces cerevisiae
metabolic network. Genome Res. 13, 244-253.
[13] Gombert, A.K., dos Santos, M.M., Christensen, B.
and Nielsen, J. (2001) Network identification and
flux quantification in the central metabolism of
Saccharomyces cerevisiae under different conditions
of glucose repression. J. Bacteriol. 183, 1441-1451.
[14] Horecker, B.L., Cheng, T. and Pontremoli, S. (1963)
Coupled reaction catalyzed by enzymes transketolase
and transaldolase .2. reaction of erythrose 4phosphate and transaldolase-dihydroxyacetone
complex. J. Biol. Chem. 238, 3428-3431.
[15] Iida, A., Teshiba, S. and Mizobuchi, K. (1993)
Identification and characterization of the tktB gene
encoding a 2Nd transketolase in Escherichia
coli K-12. J. Bacteriol. 175, 5375-5383.
[16] Kanehisa, M. and Goto, S. (2000) KEGG:
Kyoto
Encyclopedia
of
Genes
and
Genomes. Nucleic Acids Res. 28, 27-30.
[17] Karp, P.D., Riley, M., Paley, S.M. and PelligriniToole,
A. (1996) EcoCyc: An encyclopedia of Escherichia coli
genes and metabolism. Nucleic Acids Res. 24, 32-39.
[18] Kato, N., Higuchi, T., Sakazawa, C., Nishizawa,
T., Tani, Y. and Yamada, H. (1982) Purification
and properties of a transketolase responsible
for formaldehyde fixation in a methanol utilizing
yeast, Candida boidinii (Kloeckera Sp) No
2201. Biochim. Bioph. Acta 715, 143-150.
[19] Ljungdahl, L., Wood, H.G., Couri, D. and Racker,
E. (1961) Formation of unequally labeled fructose
6-phosphate by an exchange reaction catalyzed
by transaldolase. J. Biol. Chem. 236, 1622-1625.
[20] Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E.,
Sauer, U. and Szyperski, T. (2001) Central carbon
metabolism of Saccharomyces cerevisiae explored
by biosynthetic fractional 13C labeling of common
amino acids. Eur. J. Biochem. 268, 2464-2479.
[21] Marx, A., deGraaf, A.A., Wiechert, W., Eggeling, L.
and Sahm, H. (1996) Determination of the fluxes
in the central metabolism of Corynebacterium
glutamicum by nuclear magnetic resonance
49
Chapter 2
spectroscopy
combined
with
metabolite
balancing. Biotechnol. Bioeng. 49, 111-129.
50
[22] Mcintyre, L.M., Thorburn, D.R., Bubb, W.A. and
Kuchel, P.W. (1989) Comparison of computer
simulations of the F-type and L-type nonoxidative
hexose-monophosphate
shunts
with 31P-NMR experimental data from human
erythrocytes. Eur. J. Biochem. 180, 399-420.
[23] Melendezhevia,
E.,
Waddell,
T.G.
and
Montero, F. (1994) Optimization of metabolism
- the evolution of metabolic pathways toward
simplicity through the game of the pentosephosphate cycle. J. Theor. Biol. 166, 201-219.
[24] Moss,
Chapman
D.W.
and
(1982)
Hall
Ltd,
Isoenzymes.
London.
[25] Nelson,
D.L.
and
Cox,
M.M.
(2000)
Lehninger
Principles
of
Biochemistry.
Worth
Publishers,
New
York.
[26] Petersen, S., de Graaf, A.A., Eggeling, L.,
Mollney, M., Wiechert, W. and Sahm, H. (2000)
In vivo quantification of parallel and bidirectional
fluxes in the anaplerosis of Corynebacterium
glutamicum. J. Biol. Chem. 275, 35932-35941.
[27] Sabate, L., Franco, R., Canela, E.I., Centelles,
J.J. and Cascante, M. (1995) A Model of
the pentose phosphate pathway in rat
liver cells. Mol. Cell. Biochem. 142, 9-17.
[28] Sauer, U., Lasko, D.R., Fiaux, J., Hochuli, M., Glaser,
R., Szyperski, T., Wuthrich, K. and Bailey, J.E.
(1999) Metabolic flux ratio analysis of genetic and
environmental modulations of Escherichia coli central
carbon metabolism. J. Bacteriol. 181, 6679-6688.
[29] Schaaff, I., Hohmann, S. and Zimmermann, F.K.
(1990) Molecular analysis of the structural gene for
yeast transaldolase. Eur. J. Biochem. 188, 597-603.
[30] Schaaff-Gerstenschlager, I., Mannhaupt, G., Vetter,
I., Zimmermann, F.K. and Feldmann, H. (1993) tkl2,
A second transketolase gene of Saccharomyces
cerevisiae - Cloning, sequence and deletion
analysis of the gene. Eur. J. Biochem. 217, 487-492.
[31] Schaaff-Gerstenschlager, I. and Zimmermann,
F.K.
(1993)
Pentose
phosphate
pathway
in
Saccharomyces
cerevisiae
Analysis
of
deletion
mutants
for
transketolase,
transaldolase,
and
glucose-6-phosphate
dehydrogenase. Curr. Genet. 24, 373-376.
[32] Schmidt,
K.,
Nielsen,
J.
and
Villadsen,
J.
(1999) Quantitative analysis of metabolic
fluxes
in
Escherichia
coli,
using
twodimensional NMR spectroscopy and complete
isotopomer models. J. Biotechnol. 71, 175-189.
[33] Schorken, U., Thorell, S., Schurmann, M.,
Jia, J., Sprenger, G.A. and Schneider, G.
(2001) Identification of catalytically important
residues in the active site of Escherichia coli
transaldolase. Eur. J. Biochem. 268, 2408-2415.
[34] Siddiquee, K.A., Arauzo-Bravo, M.J. and Shimizu,
K. (2004) Metabolic flux analysis of pykF gene
knockout Escherichia coli based on 13C-labeling
experiments together with measurements of
enzyme activities and intracellular metabolite
concentrations. Appl. Microb. Biot. 63, 407-417.
[35] Sprenger, G.A. (1995) Genetics of pentosephosphate pathway enzymes of Escherichia
coli K-12. Arch. Microbiol. 164, 324-330.
[36] Sprenger, G.A., Schorken, U., Sprenger, G.
and Sahm, H. (1995) Transaldolase-B of
Escherichia coli K-12 - Cloning of its gene,
talB, and characterization of the enzyme from
recombinant strains. J. Bacteriol. 177, 5930-5936.
[37] Stryer,
L.
(1995)
Biochemistry.
Freeman
and
Company,
New
W.H.
York.
[38] Sundstrom, M., Lindqvist, Y., Schneider, G.,
Hellman, U. and Ronne, H. (1993) Yeast tkl1
gene encodes a transketolase that is required for
efficient glycolysis and biosynthesis of aromatic
amino acids. J. Biol. Chem. 268, 24346-24352.
[39] van Dam, J.C., Eman, M.R., Frank, J., Lange,
H.C., van Dedem, G.W.K. and Heijnen, S.J.
(2002) Analysis of glycolytic intermediates
in Saccharomyces cerevisiae using anion
exchange chromatography and electrospray
ionization with tandem mass spectrometric
detection. Anal. Chim. Acta 460, 209-218.
[40] Van Dien, S.J., Strovas, T. and Lidstrom, M.E.
(2003) Quantification of central metabolic fluxes
in the facultative methylotroph Methylobacterium
extorquens AM1 using 13C-label tracing and mass
spectrometry. Biotechnol. Bioeng. 84, 45-55.
[41] van Winden, W., Verheijen, P. and Heijnen,
S. (2001) Possible pitfalls of flux calculations
based on 13C-labeling. Metab. Eng. 3, 151-162.
[42] van Winden, W.A., van Gulik, W.M., Schipper, D.,
Verheijen, P.J.T., Krabben, P., Vinke, J.L. and Heijnen,
J.J. (2003) Metabolic flux and metabolic network
Revisiting the Pentose Phosphate Pathway
analysis of Penicillium chrysogenum using 2D [13C,
1
H] COSYNMR measurements and cumulative
Bondomer simulation. Biotechnol. Bioeng. 83, 75-92.
[43] van Winden, W.A., van Dam, J.C., Ras, C.,
Kleijn, R.J., Vinke, J.L., van Gulik, W.M. and
Heijnen, J.J. (2005) Metabolic flux analysis of
Saccharomyces cerevisiae CEN.PK113-7D based
on mass isotopomer measurements of 13C-labeled
primary metabolites. FEMS Yeast Res. 5, 559-568.
13
[44] Wiechert,
W.
(2001)
C
flux analysis. Metab. Eng. 3,
metabolic
195-206.
[45] Wiechert, W., Mollney, M., Isermann, N., Wurzel,
M. and de Graaf, A.A. (1999) Bidirectional
reaction steps in metabolic networks: III.
Explicit solution and analysis of isotopomer
labeling systems. Biotechnol. Bioeng. 66, 69-85.
[46] Williams, J.F., Blackmore, P.F. and Clark, M.G. (1978)
New reaction sequences for the non-oxidative pentose
phosphate pathway. Biochem. J. 176, 257-282.
[47] Wittmann, C. and Heinzle, E. (2001) Application of
MALDI-TOF MS to lysine-producing Corynebacterium
glutamicum - A novel approach for metabolic
flux analysis. Eur. J. Biochem. 268, 2441-2455.
[48] Wittmann, C. and Heinzle, E. (2002) Genealogy
profiling through strain improvement by using
metabolic network analysis: Metabolic flux
genealogy of several generations of lysine-producing
corynebacteria.Appl. Environ. Microb. 68, 5843-5859.
[49] Wood T. (1985) The Pentose Phosphate
Pathway.
Academic
Press
Inc.,
London.
[50] Zhang, H.M., Shimizu, K. and Yao, S.J. (2003)
Metabolic flux analysis of Saccharomyces cerevisiae
grown on glucose, glycerol or acetate by 13Clabeling experiments. Biochem. Eng. J. 16, 211-220.
[51] Zhao, G.S. and Winkler, M.E. (1994) An
Escherichia coli K-12 tktA tktB mutant deficient in
transketolase activity requires pyridoxine (VitaminB-6) as well as the aromatic amino acids and
vitamins for growth. J. Bacteriol. 176, 6134-6138.
[52] Zhao, J. and Shimizu, K. (2003) Metabolic flux
analysis of Escherichia coli K12 grown on 13C-labeled
acetate and glucose using GG-MS and powerful flux
calculation method. J. Biotechnol. 101, 101-117.
51
CHAPTER
3
Metabolic flux analysis of a
glycerol-overproducing
Saccharomyces cerevisiae
strain based on GC-MS, LC-MS
and NMR derived 13C-labeling data
Roelco J. Kleijn, Jan-Maarten A. Geertman, Beckley K. Nfor,
Cor Ras, Dick Schipper, Jack T. Pronk, Joseph J. Heijnen,
Antonius J.A. van Maris and Wouter A. van Winden
Based upon:
FEMS Yeast Research
(2007) In Press
Chapter 3
54
Abstract
This study focuses on unraveling the carbon and redox metabolism of a previously developed
glycerol over-producing Saccharomyces cerevisiae strain with deletions in the structural genes
encoding triosephosphate isomerase (tpi1), the external mitochondrial NADH dehydrogenases
(nde1 and nde2) and the respiratory chain-linked glycerol-3-phosphate dehydrogenase (gut2).
For the analysis of the metabolic fluxes two methods were used: metabolite balancing and
13
C-labeling based metabolic flux analysis. The isotopic enrichment of intracellular primary
metabolites was measured both directly (LC-MS) and indirectly through proteinogenic amino
acids (NMR and GC-MS). Because flux sensitivity around several important metabolic nodes
proved to be dependent on the applied technique, the combination of the three 13C-quantification
techniques generated the most accurate overall flux pattern. When combined, the measured
conversion rates and 13C-labeling data provided evidence that a combination of assimilatory
metabolism and pentose phosphate pathway activity diverted some of the carbon away from
glycerol formation. Metabolite balancing indicated that this results in excess cytosolic NADH,
suggesting the presence of a cytosolic NADH sink in addition to the ones that were deleted.
The exchange flux of four-carbon dicarboxylic acids across the mitochondrial membrane, as
measured by the 13C-labeling data, supports a possible role of a malate/aspartate or malate/
oxaloacetate redox shuttle in the transfer of these redox equivalents from the cytosol to the
mitochondrial matrix.
Metabolic flux analysis of S. cerevisiae
3.1
Introduction
Methods to enhance the production of glycerol from sugar using Saccharomyces cerevisiae
have been developed since the start of the 20th century. Traditionally, glycerol was produced
in oxygen-limited fermentations of S. cerevisiae using the Neuberg process. In this process,
bisulphite was used to trap acetaldehyde, the electron acceptor of alcoholic fermentation, thus
yielding excess reduction power for the reduction of dihydroxyacetone phosphate (DHAP) to
glycerol [2, 27, 28]. A more recent attempt, using metabolic engineering to improve glycerol
production in S. cerevisiae, successfully redirected the carbon flow by deletion of the structural
gene (tpi1) encoding triosephosphate isomerase [6, 7]. The strategy behind this deletion was
to force equimolar formation of glyceraldehyde-3-phosphate (GAP) and DHAP. Although
the obtained yield (0.90 mol glycerol per mol glucose) was indeed close to the maximum
theoretical yield of 1.0 for a non-growing tpi1∆ S. cerevisiae strain, the mutant did not grow
on glucose as the sole carbon source. This was attributed to reactions that compete with
glycerol-3-phosphate dehydrogenase (G3PDH) for the NADH that is generated in the lower
branch of glycolysis, causing toxic accumulation of DHAP [29].
Indeed, additional deletion of the structural genes for the external mitochondrial NADH
dehydrogenases (nde1 and nde2) and the respiratory chain-linked glycerol-3-phosphate
dehydrogenase (gut2) eliminated this growth defect [29]. Quantitative analysis of the
tpi1∆nde1,2∆gut2∆ S. cerevisiae strain displayed a difference in the glycerol yield on glucose
between aerobic batch cultures (0.99 mol glycerol per mol glucose) and aerobic glucoselimited chemostat cultures (0.83 mol·mol-1), which was attributed to the increased assimilatory
metabolism in the latter cultivations [29]. Not only the biosynthesis of cell wall constituents,
storage carbohydrates, amino acids and fatty acids, but also the generation of NADPH via
the pentose phosphate pathway (PPP) result in a decreased flux from glucose to DHAP,
thereby decreasing the glycerol production rate. Alternatively, the lower glycerol yield can also
be explained by the breakdown of DHAP via the methylglyoxal (MG) pathway. This bypass
converts DHAP to pyruvate and involves a direct transfer of reduction equivalents to the
electron transport chain via FAD-linked, mitochondrial d-lactate dehydrogenase [15, 22, 32].
Both the PPP and the MG pathway prevent equimolar fluxes through G3PDH and glyceraldehyde3-phosphate dehydrogenase (GAPDH), causing a glycerol yield on glucose lower than 1.0.
The PPP results in a higher flux through GAPDH, while the MG pathway lowers the flux through
G3PDH. Consequently, activity of either pathway would result in excess cytosolic NADH,
requiring an alternative reoxidation mechanism, such as for instance redox shuttles [2]. NADH
shuttles enable the transfer of NADH equivalents from the cytosol to the mitochondrial matrix,
where they are oxidized by the mitochondrial internal NADH dehydrogenase [2]. Notably, the
difference in glycerol yield between batch and chemostat cultivation suggests the involvement
of mechanisms sensitive to high glucose concentrations.
Metabolite balancing and 13C-labeling based metabolic flux analysis (MFA) are established
tools to study the distribution of metabolic fluxes. Metabolite balancing is based on the
principle of mass conservation of the intracellular metabolites within a defined stoichiometric
network operating under steady-state conditions [42]. Intracellular fluxes can be calculated
straightforwardly by combining the stoichiometric network with measured extracellular
conversion rates. A downside of this method is that it fails to identify parallel metabolic
55
Chapter 3
56
pathways, metabolic cycles and bidirectional reaction steps. Furthermore, it often requires
the inclusion of conserved moiety balances (i.e. ATP, NADH, NADPH), which are generally
based on incomplete stoichiometric information [3, 41]. These problems can be (partly)
overcome by using 13C as a tracer. The 13C-labeling based MFA method relies on feeding
13
C-labeled substrate to a culture of the studied micro-organism and measuring the 13C-label
distribution in various intracellular compounds under isotopic (pseudo) steady state. The
isotopic enrichment of the intracellular metabolites is measured directly using LC-MS [37] or
indirectly by measuring the 13C-labeling of proteinogenic amino acids using NMR [21] or GCMS [10, 14]. The flux patterns within a metabolic network model are subsequently calculated
by iteratively fitting simulated 13C-label distributions for a chosen set of metabolic fluxes to the
measured 13C-label distributions [41]. Up to now, all published 13C-labeling based MFA studies
relied on one [10, 14, 21, 37] or at the utmost two [9, 11] independent measurement methods
for determining the 13C-label distribution within the cell.
This study aims at mapping the distribution of carbon fluxes and reduction equivalents within
a glycerol overproducing tpi1∆nde1,2∆gut2∆ S. cerevisiae strain (CEN.PK 546-AHG) using
both metabolite balancing and 13C-labeling based MFA. 13C-labeling patterns are measured
using three independent measurement techniques: NMR, GC-MS and LC-MS. In addition,
the sensitivity of each of the three 13C-measurement techniques is evaluated for important
metabolic nodes.
3.2 Materials and methods
3.2.1
Strain use and maintenance
All experiments described in this study were performed with an evolutionary engineered
glycerol overproducing tpi1Δnde1,2Δgut2Δ S. cerevisiae strain (CEN.PK 546-AHG) [29].
Stock cultures were grown to stationary phase in shake flask cultures on mineral medium
supplemented with vitamins and trace elements [40], which was set to pH 6.0 using potassium
hydroxide and contained 2 % (w·v-1) glucose. After addition of sterile glycerol (20 % v·v-1),
2 mL aliquots were stored at -80 oC and subsequently used for inoculating precultures for
chemostat experiments.
3.2.2
Chemostat cultivation and 13C-labeling
Two duplicate aerobic, glucose-limited chemostat cultivations were carried out in a 1 L
laboratory fermentor (Applikon, Schiedam, The Netherlands) that was temperature controlled
at 30 oC. The working volume was kept at 0.6 L by means of an overflow system. The pH
was kept constant at 5.0 by an Applikon ADI 1030 biocontroller, by addition of 2 M potassium
hydroxide. Defined aerobic, continuous-culture media were prepared as described previously
[40], containing 7.5 g·L-1 glucose as the sole carbon source. The fermentor was sparged with
air (0.33 vvm) and stirred at 600 rpm. An overpressure of 0.1 bar was applied to the system to
facilitate the rapid sampling of broth. The dissolved oxygen tension was continuously monitored
with an oxygen electrode (Ingold, model 34 100 3002, Mettler, Utrecht, The Netherlands) and
never dropped below 50 %. Offgas was cooled in a condenser (2 oC) and analyzed for oxygen
and carbon dioxide concentrations with a VC Prima600 mass spectrometer (Thermo Electron
Metabolic flux analysis of S. cerevisiae
Corporation, Waltham MA, USA). The exact gas flow rate was determined using a Saga digital
flow meter (Ion Science, Cambridge, UK). The dilution rate was set to 0.05 h-1. Steady-state
cultures did not exhibit detectable metabolic oscillations. Chemostat cultures were routinely
checked for contaminations by phase-contrast microscopy.
A metabolic steady-state was defined as the situation in which at least four volume changes had
passed since the last change in culture parameters and in which the biomass concentration,
as well as all other specific production or consumption rates, had remained constant (< 2 %
variation) for 24 hours. Upon reaching a metabolic steady-state (at 85 h), the original medium
containing 7.5 g·L-1 naturally labeled glucose was replaced by chemically identical medium,
but with 10 % of the glucose replaced by [U‑13C6]d‑glucose (Isotec OH, USA) and 90%
replaced by [1-13C1]d-glucose (Campro Scientific, Veenendaal, The Netherlands). Grotkjaer
et al. [16], developed a detailed dynamic model describing carbon atoms transitions in the
central metabolism of S. cerevisiae to study the rate at which 13C is incorporated into biomass.
The model showed that for the first three residence times the labeling of proteinogenic amino
acids deviated significantly from the commonly assumed first-order washout kinetics, as a
result of transamination reactions and protein turnover. To ignore the impact of these reactions
and thus ensure full isotopic steady state for all biomass components, the chemostat was run
4 residence times on the 13C-labeled medium before sampling.
3.2.3
Rate determination
Biomass dry-weight determination and protein determination were performed in samples that
were taken from the effluent and kept on ice [13]. Extracellular samples were acquired by
rapidly sampling 2 mL of broth into a syringe containing cold steel beads (‑18oC) [23]. Culture
supernatants and culture medium were analyzed for glucose, glycerol and ethanol by HPLC
mounted with a dual-wavelength absorbance detector (Waters 2487, Milford MA, USA) and
a refractive index detector (Waters 2410, Milford MA, USA). An Aminex HPX-87H (Biorad,
Hercules CA, USA) column was used and eluted with dilute sulfuric acid (5 mM; 0.6 mL·min-1)
at 60 oC.
3.2.4
LC-MS: metabolite sampling, extraction and analysis
Samples (1 mL, ≈ 1.7 mg biomass dry-weight) for LC-MS were taken using the rapid sampling
setup described by Lange et al. [20]. Immediate quenching of the metabolism, separation of
the cells from the extracellular liquid, and metabolite extraction were performed as described
by Kleijn et al. [17]. The obtained supernatant was stored at -80 oC prior to LC-MS analysis.
The mass isotopomer distributions of the intracellular metabolites were measured as described
by van Winden et al. [37]. In short, metabolites were first separated by high-performance
anion exchange chromatography (Waters, Milford MA, USA) followed by MS analysis with a
Quattro-LC triple quadrupole mass spectrometer (Mircomass Ltd., Manchester, UK) equipped
with an electrospray ionization interface. LC-MS was used to analyse glucose-6-phosphate
(G6P), fructose-6-phosphate (F6P), 6-phosphogluconate (6PG), manose-6-phosphate (M6P),
fructose-1,6-bisphosphate (FBP), phosphoenol pyruvate (PEP), the combined pool of 2- and
3-phosphoglycerate (2/3PG), the combined pool of xylulose-5-phosphate, ribose-5-phosphate
and ribulose-5-phosphate (P5P) and sedoheptulose-7-phosphate (S7P).
57
Chapter 3
Mass fractions of CO2 in the offgas were determined using a VC Prima600 mass spectrometer.
The mass spectrometer was calibrated using two 13CO2 calibration gases containing 2% CO2
and 98% N2 with a 13CO2:12CO2 ratio of 1:2 and 1:4. The calibration gases were prepared by,
respectively, dissolving 1.0 g and 1.2 g of unlabeled K2CO3 with 0.5 g and 0.3 g of 13C-labeled
K2CO3 (Cambridge Isotope Laboratories, Andover MA, USA) in 30 mL of water. Through
addition of 10 mL of 4M HCl, the solution was brought to pH 1 and subsequently heated to
50oC. 0.24L of CO2 gas was captured in a 10L gas bag (Alltech, Deerfield IL, USA) and diluted
with N2 gas.
58
3.2.5
GC-MS: biomass sampling and analysis
Biomass samples (115 mL ≈ 200 mg biomass dry-weight) for the GC-MS analysis were taken
from the chemostat, spun down directly for 5 min at 2000 g, washed with 0.9 % NaCl solution
and lyophilized. Aliquots (5 mg) of the lyophilized material were hydrolyzed in 1 mL 6 M HCl,
dried and derivatized as described by Perrenoud & Sauer [30]. GC-MS analysis and raw data
analysis were performed as described by Fischer & Sauer [10]. Mass isotopomer distributions
were measured for the following proteinogenic amino acid (fragments): alanine(all), (‑1),
aspartate(all), (‑1), (1+2), glutamate(all), (‑1), glycine(all), (‑1), isoleucine(all), (‑1), leucine(all),
(‑1), lysine(all), (‑1), phenylalanine(all), (‑1), (1+2), proline(all), (‑1), serine(all), (‑1), (1+2),
threonine(all), (‑1), tyrosine(all), (‑1), (1+2), valine(all), (‑1), (1+2), where (all) denotes the
complete amino acid, (‑1) is the amino acid minus the carboxyl group and (1+2) is the amino
acid fragment consisting of only the first two carbon atoms (position 1 being the carboxyl
group). The obtained mass isotopomer fractions were corrected for the occurrence of natural
isotopes of N, H, O, S and Si in both the amino acid and the derivatizing agent and for the
occurrence of natural isotopes of C in the derivatizing agent only. The applied correction was
the inverse of the procedure proposed by van Winden et al. [39].
3.2.6
2D [13C,1H] COSY NMR: biomass sampling and analysis
Together with the sample taken for GC-MS analysis 385 mL of culture broth (≈ 665 mg
biomass) was spun down directly for 5 min at 2000 g, washed with 0.9 % NaCl solution
and demineralized water and stored at -80 oC. Prior to 2D [13C,1H] COSY NMR analysis the
biomass was lyophilized and hydrolyzed. Subsequently, the amino acids were purified and
prepared for measurement as described by van Winden et al. [38]. The NMR spectra were
recorded and analyzed by means of spectral fitting software as described by van Winden et
al. [36]. The resulting data are relative intensities of fine structures observed in the multiplets
of the proteinogenic amino acids phenylalanine‑α, ‑β, glycine‑α, histidine‑α, ‑β, serine-α, ‑β,
tyrosine‑α, ‑β, alanine‑α, ‑β, aspartate‑α, ‑β, glutamate‑α, ‑β, ‑γ, isoleucine‑β, ‑γ1, ‑δ, lysine‑β,
‑γ, leucine‑α, ‑β, ‑γ, ‑δ1, ‑δ2, methionine‑α, proline‑α, ‑δ, arginine‑β, ‑δ, threonine‑α, ‑β, ‑γ,
valine‑α, ‑γ1 and the storage carbohydrate trehalose‑C1. Relative intensities for the symmetrical
component glycerol‑C1/C3 were measured in filtrate samples. Note that the nomenclature for
numbering the carbon atoms of amino acids used is as follows; C is the C‑terminus of the
amino acid (carboxyl group); α is the carbon atom next to C; β is the carbon atom next to α,
etc. Standard three letter abbreviations were used for the amino acids.
Metabolic flux analysis of S. cerevisiae
3.2.7
Metabolite balancing
A previously developed compartmentalized stoichiometric model of S. cerevisiae [19] was
used for metabolite balancing. Calculations were carried out with the software package
MNAv3.0 (SpadIT, Nijmegen, The Netherlands). To describe the genetically engineered
tpi1∆nde1,2∆gut2∆ strain, the reactions carried out by triosephosphate isomerase, the
external NADH dehydrogenases and FAD-dependent glycerol-3-phosphate dehydrogenase
were removed. Subsequently, eight versions of the stoichiometric model were constructed
using different assumptions with respect to the existence of a MG bypass, the existence of
a mitochondrial NADH shuttle and the cofactor specificity of acetaldehyde dehydrogenase
(see results section). The measured biomass specific conversion rates of glucose, glycerol,
O2 and CO2 were used as input for the metabolite balancing. These rates were calculated
from the average steady state values of the measured flows and concentrations. The flux
balancing procedure was weighted using the full variance matrix associated with the calculated
conversion rates.
13
C-labeling based MFA
3.2.8
The metabolic network models described above formed the basis of the model used for
the 13C-labeling based MFA. The following adaptations were made: (i) mass balances for
reduction equivalents were removed, (ii) reaction reversibilities were included, (iii) the
conventional reactions of the non-oxidative branch of the PPP were replaced by metabolite
specific, reversible, C2 and C3 fragment-producing and consuming half reactions as proposed
by Kleijn et al. [18], (iv) glycine synthesis via the enzyme threonine aldolase was included,
(v) the glycine decarboxylase complex (GDC) catalyzing the reversible cleavage of glycine
into CO2 and 5,10-CH2-H4folate (E-C1) was included [31], (v) the transport of acetyl-CoA
from the cytosol to the mitochondrion was included, (vii) the gluconeogenic enzyme PEP
carboxykinase catalyzing the conversion of oxaloacetate into PEP was included, (viii) malic
enzyme catalyzing the mitochondrial decarboxylation of malate into pyruvate was included
and (ix) scrambling reactions for the symmetrical molecules glycerol and succinate (which
forms part of the oxaloacetate pool) were included. Metabolite pools in the network that had
only one influx were removed, while all metabolite pools in isotopic equilibrium due to fast
exchange reactions were lumped into one single pool [35]. Reversible reactions were modeled
as separate forward and backward reactions and are referred to as net and exchange fluxes,
where:
v net = v forward − v backward
(3.1)
v exchange = min(v forward ,v backward )
(3.2)
All major CO2 producing and consuming reactions were incorporated in the metabolic network
model, making it possible to fit the measured mass fractions of CO2. To correctly estimate the
isotopic enrichment of the CO2 pool, the inflow of naturally labeled CO2 as a result of aeration
was also included in the metabolic network model. No distinction was made between CO2
produced and consumed in different compartments.
The employed flux fitting procedure is described in detail by van Winden et al. [37]. Standard
59
Chapter 3
deviations for the measured 13C-labeling data were fixed at 1%. The variance-weighted sum
of squared residuals (SSres) between the simulated and measured data was used as the target
function in a minimization procedure based on sequential quadratic programming that was
implemented in Matlab (The MathWorks Inc., Natick MA, USA).
60
3.3 Results and Discussion
3.3.1
Macroscopic data
The biomass-specific consumption and production rates of the tpi1Δnde1,2Δgut2Δ strain are
given in Table 1. As expected for a tpi1Δ strain the biomass yield (0.24 ± 0.01 g biomass·(g
glucose)-1) on glucose was significantly lower compared to a wild-type under the same
conditions. The glycerol yield on glucose during these cultivations was calculated to be 0.80 ±
0.01 mol·mol-1. In line with this high glycerol production, the respiratory quotient (RQ) was 1.24
mol CO2·(mol O2)-1. The carbon and the degree of reduction (γ) balances closed to 99.9±0.1
% and 101.4 ± 0.7 % respectively, indicating that apart from biomass, CO2 and glycerol, no
significant amounts of other compounds were produced. In addition, black box balancing of
the conversion rates (Table 1) and gross error detection of the measured conversion rates
according to van der Heijden [33], showed that the measurement set was statistically acceptable
(p-value = 0.21). The p-value represents the probability that the discrepancy between the
balanced and measured conversion rates can be explained by measurement error. Typically,
p-values of 0.05 or lower signify the presence of other errors (e.g. systematic errors, erroneous
model assumptions). Measured biomass specific consumption and production rates were,
subsequently, used for metabolite balancing and 13C-labeling based MFA.
Table 1 Measured and reconciled conversion rates for an evolutionary engineered glycerol overproducing
tpi1Δnde1,2Δgut2Δ S. cerevisiae strain grown in duplicate aerobic, glucose-limited chemostat cultures at
a dilution rate of 0.05 h-1.
Conversion rates
Unit
Glucose consumption
Measured flux
Reconciled flux
mmol·(CmolX·h)
-1
30.0±1.0
30.3
Oxygen consumption
mmol·(CmolX·h)
-1
47.7±2.4
48.5
Carbon dioxide
production
mmol·(CmolX·h)-1
60.3±3.0
62.4
Glycerol production
mmol·(CmolX·h)-1
23.9±0.8
23.8
Ethanol production
mmol·(CmolX·h)-1
< 0.03
0.0
3.3.2
Metabolite balancing
To gain more insight in the primary metabolism of the adapted tpi1∆nde1,2∆gut2∆ strain
eight different variants of the stoichiometric model were constructed and used for metabolite
balancing (see Materials and Methods, and Table 2). The basic stoichiometric model (model
1) contained no MG bypass, no mitochondrial NADH-shuttle and a NADP+-dependent
acetaldehyde (Aald) dehydrogenase. It should be stressed here that metabolite balancing
Metabolic flux analysis of S. cerevisiae
requires the inclusion of reduction-equivalent balances (NAD(H) and NADP(H)) in order to
determine all fluxes. Statistical testing of the balanced conversion rates for this model gave
a high χ2-test value and, subsequently a low p-value (p<0.001, Table 2), indicating that the
redundant measured conversion rates could not be reconciled by the stoichiometric model.
The statistical rejection of model 1 is in part caused by an overestimation of the glycerol
production rate (0.84 mol·(mol glucose)-1, Figure 1) compared to the measured rate of 0.80
mol·(mol glucose)-1. In stoichiometric model 1 the glucose entering the cell can be metabolized
towards GAP via two pathways; the glycolysis and the PPP. Since the PPP is the main pathway
for cytosolic NADPH formation, the flux through the PPP is constrained by the mass balance
for NADPH and thus depends directly on the biosynthetic demand for NADPH. In model 1,
this maximum biosynthetic NADPH demand fixed the oxidative PPP flux at 0.15 mol·(mol
glucose)-1, which resulted in a glycerol production of 0.84 mol·(mol glucose)‑1 (see Figure
1). A second reason for the statistical rejection of model 1 was the estimation of a ethanol
production rate of 0.11 mol·(mol glucose)-1, whereas the measured ethanol production rates
were below detection level (<0.001 mol·(mol glucose)‑1). Ethanol was formed in model 1 to
remove the surplus of cytosolic NADH formed in anabolism (0.07 mol·(mol glucose)-1) and
from the GAP synthesized in the PPP (0.04 mol·(mol glucose)‑1). Due to the exclusion from
the model of the external mitochondrial NADH dehydrogenases and the glycerol-3-phosphate
dehydrogenase, ethanol formation was the only remaining pathway to reoxidize the excess
cytosolic NADH.
Table 2 Metabolic fluxes for several important metabolic nodes in a tpi1∆nde1,2∆gut2∆ S. cerevisiae strain
as calculated via metabolite balancing for eight different stoichiometric models. The eight stoichiometric
models differed with respect to the existence of a MG bypass, the existence of a putative NADH shuttle
and the cofactor specificity of the enzyme acetaldehyde dehydrogenase. Fluxes are normalized for the
glucose uptake rate, which was set to 100.
Model
NADH
shuttle
MG
bypass
Specificity
acetaldehyde
dehydrogenase
Flux through
PPP
Flux through
methylglyoxal χ2-value p-valuea
split ratio NADH shuttle
bypass
1
-
-
NADP+
15.2
0
0
1900 <0.001
2
-
+
NADP
+
17.1
0.0
-12.3
257 <0.001
3
+
-
NADP
+
16.4
11.9
0
24.8 <0.001
4
+
+
NADP
+
16.1
16.1
4.5
5
-
-
NAD+
17.6
0
0
2230 <0.001
6
-
+
NAD+
22.9
0
-24.7
558 <0.001
7
+
-
NAD
+
21.0
22.9
0
12.6
0.002
8
+
+
NAD
+
20.8
25.5
2.9
3.58
0.167
3.58
0.167
The p-value represents the probability that the discrepancy between the balanced and measured
conversion rates can be explained by measurement error alone. Typically, p-values < 0.05 signify an
erroneous stoichiometric model.
a
61
Chapter 3
glc
100
100
biomass
biomass
+
NAD
7
7
f6p
84
83
NADH
biomass
X
dhap
NAD
84
79
62
g6p
75
74
5
biomass
15
16
PPP
9
9
4
5
gap
biomass
+
88
88
NADH
etoh
goh
biomass
pep
11
0
84
83
pyr
oaa
biomass
X
nd
i1
AcCoA
pyr
biomass
NAD
50
63
oaa
50
63
45
58
9
9
biomass
4
5
NAD+
NADH 187
260
NADH
aald
20
9
55
68
gut2
XUQ
nde1,2
8
8
+
NAD
+
NADH
AcCoA
2 NADH
citr
+
2 NAD
NAD+
NADH
NADH
biomass
+
NAD
?
NADH
+
NAD
16
50
63
akg
Figure 1 Intracellular metabolic fluxes in a tpi1Δnde1,2Δgut2Δ S. cerevisiae strain quantified via
metabolite balancing. The strain was grown in an aerobic, glucose-limited chemostat culture at
D=0.05 h-1. Fluxes were determined by combining the measured conversion rates with a standard
stoichiometric model (model 1, bold) and an extended stoichiometric model containing a putative
NADH-shuttle and the MG bypass (model 8, italic). Apart from the carbon flow through the primary
metabolism, the NADH production or consumption is specified for each reaction. Fluxes are
normalized for the glucose uptake rate. Abbreviations: glc: glucose, G6P: glucose-6-phosphate,
F6P: fructose-6-phosphate, PPP: pentose phosphate pathway, dhap: dihydroxyacetone phosphate,
GAP: glyceraldehyde-3-phosphate, goh: glycerol, PEP: phosphoenol pyruvate, PYR: pyruvate,
aald: acetaldehyde, etoh: ethanol, AcCoA, acetyl-CoA, OAA: oxaloacetate, CITR: citrate, AKG:
α‑ketogluterate.
Metabolic flux analysis of S. cerevisiae
To remedy the overestimation of the glycerol and ethanol production rates in model 1,
additional reactions were included into the stoichiometric model. The first discrepancy, the
overestimation of the glycerol production, was addressed by incorporation of a MG bypass
(model 2). This reaction created an alternative mode of DHAP dissimilation. Addition of the
MG bypass increased the statistical acceptability of the model (lower χ2-test value), but still led
to model rejection (p<0.001). In fact, a negative flux was fitted for the MG bypass, signifying
the conversion of pyruvate into DHAP. The reverse flux actually mimics the role of the enzyme
triosephosphate isomerase by converting GAP into DHAP and thus indirectly removes part of
the surplus NADH in the cytosol at the cost of an increased glycerol production rate. Note that
in vivo the MG bypass is an irreversible pathway, thereby rendering the proposed carbon-flow
infeasible. These results indicate that the cytosolic redox balance has a larger impact on the
statistical acceptability of the model than the overestimation of the glycerol production rate.
The apparent cytosolic redox stress was addressed in model 3 by the introduction of an
alternative pathway to reoxidize cytosolic reduction equivalents, in the form of a putative NADH
shuttle between the cytosol and the mitochondrion. The putative NADH shuttle allowed for the
net oxidation of cytosolic NADH and reduction of mitochondrial NAD+. It significantly increased
the statistical acceptability of the model (χ2-test value=24.8) and abolished ethanol formation.
The equivalent of 0.12 mol NADH·(mol glucose)-1 was transported into the mitochondrial
matrix. Nevertheless, the model was still statistically rejected (p=0.002).
Since the separate inclusion of the MG bypass and the putative redox shuttle had a
positive but insufficient effect on the acceptability of the model, both were incorporated in
stoichiometric model 4. This combination yielded statistically acceptable flux-patterns when
balancing the variance-weighted conversion rates (Figure 1). The flux through the MG
bypass and the putative NADH shuttle were estimated at 0.045 and 0.161 mol·(mol glucose)-1
respectively. These results suggest that the adapted tpi1∆nde1,2∆gut2∆ strain possesses a
significant cytosolic NADH sink, most likely translocating excess NADH from the cytosol to the
mitochondria. The estimated flux through the MG bypass was 15-fold higher than the value
reported by Martins et al. [22] for wild-type S. cerevisiae, who measured the MG bypass flux
to be 0.3% of the total glycolytic flux independent of its magnitude. The PPP split ratio was
estimated at 0.16 mol·(mol glucose)-1. This value corresponds well with the value reported by
Gombert et al. [14] (0.42 mol·(mol glucose)-1) for wild-type S. cerevisiae, which gives a PPP
split-ratio of 0.20 mol·(mol glucose)-1 for the tpi1∆nde1,2∆gut2∆ strain when corrected for its
higher biomass yield (Yxs=0.52 g· (g glucose)-1).
A higher flux through the PPP directs part of the carbon around FBP, resulting in reduced
DHAP formation and thus less glycerol production. However, as described before the flux
through the PPP is constrained by the NADPH balance. Therefore, in models 5-8 the cofactor
specificity of cytosolic Aald dehydrogenase, required for the formation of cytosolic acetyl-CoA,
was changed from NADP+ to NAD+. Even though most literature sources point to an NADP+
dependency of the cytosolic Aald dehydrogenase (ald6) [8, 24], S. cerevisiae also contains
two genes (ald2 and ald3) encoding for stress-induced, cytosolic NAD+ dependent Aald
dehydrogenases [26]. As a result of the change in cofactor specificity, estimated PPP splitratios for models 5-8 were on average 0.05 mol·(mol glucose)-1 higher than those of models
1-4. The increased PPP split-ratio also lowered the MG bypass flux (model 8), as less surplus
63
Chapter 3
64
carbon had to be withdrawn from the glycerol pathway. In contrast to the lower MG bypass
flux, the flux through the putative NADH shuttle increased as a result of the additional NADH
produced by Aald dehydrogenase. Models 4 and 8 both had the same p-value (p<0.167),
indicating that the cofactor specificity of the Aald dehydrogenase did not affect the statistical
acceptability.
The above stoichiometric models indicate the presence of pathways that divert carbon away
from glycerol formation (e.g. the pentose phosphate pathway and the methylglyoxal bypass)
and an additional sink for cytosolic NADH in the investigated tpi1∆nde1,2∆gut2∆ S. cerevisiae
strain. In a next step, 13C-labeling experiments were performed to validate the presence of
these pathways and to gain further insight in the actual mechanism of the cytsolic NADH
sink.
3.3.3
Metabolic flux analysis using 13C-labeling
LC-MS, GC-MS and NMR datasets The tpi1∆nde1,2∆gut2∆ strain was grown on a mixture of
10% [U-13C]glucose and 90% [1-13C]glucose. The inflow of labeled material was followed by
online measurements of the 13CO2/12CO2 ratio in the offgas of the chemostat (Figure 2). CO2 is
primarily produced by the catabolic pathways of the cell (e.g. TCA cycle). The sharp increase
in the 13CO2 concentration seen directly after switching to 13C-labeled glucose (at t = 85 h),
therefore indicates a rapid distribution of 13C-label throughout the primary metabolism. As CO2
is also produced during the synthesis and turnover of macromolecular biomass components
(e.g. storage carbohydrates, proteins and lipids) it takes several residence times before a
constant ratio is reached (>3 residence times). Figure 2 shows that the 4 residence times
of 13C-labeling applied in this study were sufficient to ensure full isotopic steady state for all
biomass components (see also materials and methods).
The 13C-labeling patterns measured with LC-MS, GC-MS and NMR are given in the Appendix
(Table A-1). The LC-MS derived mass isotopomer fractions of the hexose mono-phosphates
are in accordance with the applied substrate labeling (10% [U-13C6] and 90% [1-13C]glucose).
The observed decrease in the m+1, m+2 and m+6 mass fractions in favor of the unlabeled
m+0 fraction is a result of the oxidative branch of the PPP, in which the first carbon atom
(carboxyl group) of 6-phosphogluconate is split off to form CO2. No notable discrepancies in
labeling are observed between the hexose mono-phosphates. The identical labeling patterns
for G6P/F6P and F6P/M6P indicate the reversibility of the enzymes G6P isomerase and M6P
isomerase, respectively.
Analysis of the carbon transitions in the PPP and the upper glycolysis shows that the fragment
consisting of the last three carbon atoms of the upper glycolytic and PPP metabolites remains
intact during conversion and thus always ends up as GAP by the action of either FBP aldolase or
transaldolase. In this specific strain the carbon transition is further constrained by the absence
of the enzyme triosephosphate isomerase, preventing the exchange of label between DHAP
and GAP. Consequently, the 13C-labeling of GAP is identical to the 13C-labeling of the bottom
three carbon atoms of the glucose fed to the chemostat. This is confirmed by the LC-MS
derived mass isotopomer fractions of both PEP and 2/3PG (Table A-1) which were measured
to be 90% naturally labeled (m+0/m+1/m+2) and 10% uniformly labeled (m+3). Similar mass
isotopomer distributions were measured using GC-MS for the amino acid fragments Phe/
Metabolic flux analysis of S. cerevisiae
Thr(1+2) and Ser(1+2) (See Table A-1), which correspond with the first two carbon atoms of
PEP and 3PG, respectively. The top three carbon atoms of the glucose fed to the chemostat
end up in DHAP and are thus primarily converted into glycerol. The exact labeling of glycerol
depends on the extent of the carbon redistribution in the non-oxidative branch of the PPP, but
it will be either 13C-labeled at the first carbon position (<90%) or uniformly labeled (<10%).
In accordance with this, NMR-derived relative intensities of the first carbon atom of glycerol
contained a singlet peak of 85% and a doublet peak of 15%, indicating a ratio of 85:15 for the
labeling patterns ‘10X:11X’ of glycerol, where ‘0’ indicates a 12C-atom, ‘1’ indicates a 13C-atom
and ‘x’ can denote either.
0.30
65
0.25
0.15
13
CO2 / 12CO2
0.20
0.10
0.05
0.00
0
4
5
6
7
8
9
Residence times (-)
Figure 2 Online measurement of the 13CO2/12CO2 ratio in the offgas of an aerobic glucose-limited
chemostat culture of a tpi1Δnde1,2Δgut2Δ S. cerevisiae strain. After 4.25 residence times the
naturally-labeled feed was replaced by a chemically identical feed, but enriched in 13C. Samples
for measurement of the isotopic enrichment of the intracellular components were harvested 4
residence times later.
Metabolic flux analysis for the individual datasets 13C-labeling based MFA was performed
for each of three datasets presented in Table A-1. GC-MS and NMR data reflect the 13Clabeling pattern of proteinogenic amino acids and storage carbohydrates. These components
are synthesized from metabolic precursors that are localized in a specific cellular compartment.
In contrast, the LC-MS derived mass isotopomers contain 13C-labeling information about cellaveraged metabolite pools, irrespective of the compartments in which they are localized. Due
to insufficient knowledge on the distribution of metabolites localized in multiple compartments
(such as pyruvate and the TCA-cycle metabolites) only the mass isotopomers of the glycolytic
and PPP metabolites were measured and the metabolic flux-estimates for the LC-MS dataset
Chapter 3
66
were limited to glycolysis and the PPP. Estimated flux-patterns for the three datasets are
displayed in Figure 3 (first three values).
The most prominent feature of Figure 3 is the similarity in flux estimates for the three
individually fitted datasets. Differences in flux patterns are seen for the reversibilities of the
transketolase and transaldolase catalyzed reactions in the non-oxidative branch of the PPP.
The reversibilities in the non-oxidative branch of the PPP have always proven difficult to
estimate accurately [12]. However, this has no influence on the flux estimates in the other
parts of the network. A similar observation was made by Christensen et al. [5] when estimating
fluxes in the central metabolism of S. cerevisiae using samples taken at different time points
during the cultivation. A second part of the metabolism that shows different flux-estimates for
the GC-MS and NMR datasets in Figure 3 is the transport of pyruvate and acetyl-CoA across
the mitochondrial membrane. With the help of a sensitivity analysis, it can be shown that
the employed 13C-subtrate labeling can not distinguish the two alternative routes (see next
section).
Interestingly, several researchers have speculated that the LC-MS derived flux patterns,
based upon a direct measurement of the 13C-labeling of primary metabolites, might differ
from the flux patterns derived from proteinogenic amino acids measured using NMR and GCMS [5, 34]. They argued that the labeling patterns of proteinogenic amino acids specifically
reflect the flux distribution in a cell during the cell cycle phase when proteins are synthesized
(G1-phase), while the labeling patterns of the primary metabolites reflect the average flux
distribution during a cell cycle. The data presented here do not support this hypothesis.
Metabolic flux analysis for the combined dataset To further investigate the comparability
among the three measurement sets, 13C-labeling based MFA was performed on pairwise
combinations of the three datasets as well as on the combination of the three datasets. Figure
4 compares the build-up of the SSres for the combined datasets with those of the individual
datasets, where the SSres is a measure for the discrepancy between the measured and
simulated 13C-labeling distribution. In general, combining the datasets did not lead to a major
increase in the SSres. The total SSres when combining all three datasets was only 20% higher
than the summed SSres for each separately fitted dataset, thereby confirming the coherence
between the three datasets. Estimated fluxes for the combination of all three datasets are
represented by the bottom values in Figure 3. In line with the minimal increase in SSres, similar
fluxes were fitted for the combined datasets compared to those fitted for the three individual
datasets (Figure 3).
As an additional check of the accuracy of the estimated 13C-labeling based flux patterns the
combined dataset was refitted, but now with the triosephosphate isomerase catalyzed reaction
included in the stoichiometric network model of the tpi1Δnde1,2Δgut2Δ strain. It was checked
whether the metabolic flux through the isomerase would be fitted to be negligible. Indeed, a
negligible flux of 0.0028 mol·(mol glucose)-1 was fitted for the forward reaction from GAP to
DHAP, while the flux through the backward reaction was completely absent.
Metabolic flux analysis of S. cerevisiae
glc
25
25
23
24
100
11
biomass
g6p
65 (>1000)
65 (>1000)
66 (709)
66 (>1000)
f6p
80
80
80
80
fbp
80 (425)
80 (64)
80 (131)
80 (>1000)
0
biomass
dhap
80
80
80
80
0
0
1
1
-7 (2)
-7 (0)
-7 (157)
-7 (5)
-8 (21)
-8 (0)
-8 (29)
-8 (12)
X
gap
nde1,2
X
UQ
XUQ
oaa
5 (10)
8 (18)
8 (19)
oaa
oaa
s7p
8 (6)
8 (7)
8 (375)
8 (4)
E-C3
pep
2
82
85
84
9
13
13
67
1
biomass
1
C1
ser
pyr
biomass
-8 (>1000)
-8 (68)
-8 (326)
-8 (790)
15 (14)
15 (9)
14 (23)
15 (11)
E-C2
3
2
2
85 (150)
84(193)
85 (238)
85 (188)
0
0
0
gut2
p5p
e4p
goh
4
biomass
1
2 (2)
2 (1)
1 (2)
1
biomass
67
51
64
6
20
6
1
7
2mass
pyr
3
pyr
biomass
3
0
58
16
59 AcCoA
3
58
0 (0)
0 (0)
0 (0)
gly
2
biomass
thr
0
1
1
aald
67
51
64
AcCoA
9
oaa
biomass
58
42
56
AcCoA
53
54
54
citr
citr
biomass
5
akg
biomass
58
59
58
Figure 3 Intracellular metabolic fluxes in a tpi1Δnde1,2Δgut2Δ S. cerevisiae strain determined by
independently fitting the mass isotopomer fractions of the intracellular metabolites measured by LCMS (top values, bold), by independently fitting the mass isotopomer fractions of the proteinogenic
amino acids measured by GC-MS (top-middle values, italic), by independently fitting the relative
intensities of the proteinogenic amino acids measured by NMR (bottom-middle values, bold italic)
and by fitting the combined 13C-labeling dataset (NMR, LC-MS and GC-MS) (bottom values, normal).
Fluxes are normalized for the glucose uptake rate. Values outside parentheses denote the net
fluxes, while values inside parentheses represent the exchange fluxes. Solid arrowheads denote the
direction of the net flux. Abbreviations: FBP: fructose-1,6-bisphospate, P5P: pentose-5-phopshate,
S7P: sedoheptulose-7-phosphate, E4P: erythrose-4-phosphate, C1: methylenetetrahydrofolate, EC2: glycolaldehyde moiety covalently bound to the thiamine pyrophosphate/transketolase complex,
E-C3: dihydroxyacetone moiety covalently bound to the enzyme transaldolase, ser: serine, gly:
glycine, thr: threonine, other abbreviations are listed in Figure 1.
Chapter 3
700
600
LCMS
NMR
GCMS
500
SSres
400
300
200
100
LCMS
NMR
LCMS
GCMS
GCMS
NMR
in
dv
co .
m
b.
dv
co .
m
b.
in
in
dv
co .
m
b.
0
in
dv
co .
m
b.
68
LCMS
NMR
GCMS
Figure 4 Contribution of the LC-MS, GC-MS and NMR datasets to the minimized sum of squared
residuals (SSres) as calculated via 13C-labeling based MFA. Minimized SSres values were first derived
for each individual dataset (indv.), after which combinations (comb.) of the 13C-datasets were used
for flux-estimation.
Flux-sensitivity analysis The sensitivity of the calculated 13C-based fluxes to measurement
error was determined for several important metabolic nodes in the metabolism of the quadruple
deletion mutant as described by Kleijn et al. [17]. The measured 13C-labeling distributions were
independently re-fitted for a range of fixed flux-values for the studied metabolic node and the
fold-change in the SSres was used as a measure for the flux sensitivity. Flux sensitivities were
determined for the following nodes: i) the PPP split-ratio ii) the pyruvate carboxylase catalyzed
reaction iii) the reversibility of the oxaloacetate transport and iv) the pyruvate decarboxylase
catalyzed reaction. Figure 5 shows the fold-changes in the SSres for the different datasets.
The PPP split-ratio can be estimated accurately with all three 13C-labeling measurement
techniques (Figure 5-A). Nonetheless, the GC-MS measurements are most sensitive to
changes in the oxidative PPP flux; denoted by the large fold-change in the SSres when the
PPP split-ratio was fixed at suboptimal values. The minimal flux needed for the synthesis of
biosynthetic precursors (e.g. amino acids) determines the lower boundary for the PPP splitratios at 0.04 mol·(mol glucose)-1, while the fixed glycerol production rate and biosynthetic flux
from the G6P-pool determine the maximum PPP split-ratio at 0.25 mol·(mol glucose)-1.
The joint action of the enzymes pyruvate kinase, pyruvate carboxylase and PEP carboxykinase
creates a futile cycle within the metabolic network of Figure 3. Due to uncertainties concerning
energy consuming processes this cycle could not by examined via metabolite balancing.
Omission of the ATP balance in the 13C-labeling based MFA enabled the flux through the futile
cycle to be quantified. Figure 5-B shows that the anaplerotic flux from pyruvate to oxaloacetate
is best estimated via NMR. Note that the same sensitivity holds for the fluxes catalyzed by
Metabolic flux analysis of S. cerevisiae
pyruvate kinase and PEP carboxykinase that take part in the same metabolic cycle. The
anaplerotic flux has a lower limit of 0.09 mol·(mol glucose)-1 due to the fixed biomass synthesis
fluxes for cytosolic oxaloacetate and mitochondrial α‑ketoglutarate.
The reversibility of the oxaloacetate transporter was determined for both the NMR and the
GC-MS dataset. The largest fold-change in the SSres when fixing the reversibility of the
oxaloacetate transporter was observed for the NMR dataset (Figure 5-C). Assuming a twofold increase in the SSres as significantly different, the reversibility of the transporter will be
within the interval 0.60 to 0.80. In terms of relative flux size this means an exchange flux
ranging from 0.12 to 0.40 mol·(mol glucose)‑1.
The distribution of the flux from pyruvate towards acetyl-CoA via the two possible routes (see
Figure 3) is insensitive to both the GC-MS and NMR dataset (Figure 5-D). This insensitivity
is caused by the similar labeling of cytosolic and mitochondrial pyruvate and hence cytosolic
and mitochondrial acetyl-CoA for both datasets [5]. The similar labeling of cytosolic and
mitochondrial pyruvate in the GC-MS dataset follows from the mass isotopomer fractions of
Phe(1+2) and Tyr(1+2) (corresponding with the first two carbon atoms of cytosolic pyruvate)
and those of Val(1+2) (corresponding with the first two carbon atoms of mitochondrial pyruvate).
The similar labeling of cytosolic and mitochondrial pyruvate in the NMR dataset follows from
the relative intensities of Tyr-α and Phe-α (corresponding with cytosolic pyruvate) and those
of Ala-α (corresponding with mitochondrial pyruvate). Note that the flux for the malic enzyme
catalyzed reaction may cause a difference between cytosolic and mitochondrial pyruvate
through an extra inflow of label into the mitochondrial pyruvate pool. The bigger difference in
13
C-labeling between cytosolic and mitochondrial pyruvate for the NMR-data in comparison
to the GC-MS data therefore results in a bigger flux-estimate for the malic enzyme catalyzed
reaction from the NMR dataset (Figure 3).
In the flux sensitivity analysis of the PPP split-ratio (Figure 5-A), the fixed glycerol production
rate and fixed biosynthetic efflux from the G6P pool, determined the maximum split-ratios at
0.25 mol·(mol glucose)-1. However, these fluxes derived from biomass synthesis also have a
certain degree of error. Figure 6‑A shows the effect of varying the biosynthetic G6P efflux on
the SSres of the flux fit. Note that the performed flux fits were based upon the combined dataset
(LC-MS, GC-MS and NMR) and not on the individual datasets used for the flux sensitivity
analyses of Figure 5. The combined dataset gave the most accurate flux estimates for the
PPP split-ratio, for reasons explained below. Figure 6-A shows that the optimal biosynthetic
G6P efflux (0.10 mol NADH·(mol glucose)-1) closely resembled the calculated G6P efflux (0.11
mol NADH·(mol glucose)-1 in Figure 3). More importantly, Figure 6-A shows that the optimal
PPP split-ratio remained unchanged at 0.25 mol·(mol glucose)‑1, despite the possibility of
also fitting higher PPP split-ratio values. In Figure 6-B the contribution of the three different
datasets to the optimal SSres are plotted, showing that the LC-MS dataset is most sensitive to
changes in the G6P efflux and hence the PPP split-ratio. In retrospect, the earlier observed
sensitivity of the GC-MS and NMR dataset to changes in the PPP split-ratio (based on Figure
5-A) denotes in actuality the sensitivity of these datasets to the MG bypass flux. By varying the
biosynthetic G6P efflux, the PPP split-ratio was uncoupled from the MG bypass flux, resulting
in a lower fold-change in the SSres for the NMR and the GC-MS datasets. In accordance with
this a small MG bypass flux was fitted, independent of the chosen G6P efflux (Figure 6-A).
69
Chapter 3
12
Fold change in SSres
NMR
GC-MS
LC-MS
A
10
6
5
4
8
3
6
2
4
1
2
0
70
5
0
6
10
15
PPP split-ratio
20
0
25
5
4
3
2
1
0
0.0
10
6
C
Fold change in SSres
Fold change in SSres
B
Fold change in SSres
14
0.2
0.4
0.6
0.8
oxaloacetate transport reversibility
40
D
5
4
3
2
1
0
1.0
20
30
pyruvate carboxylase flux
(mol.[100 mol glucose]-1)
0
10
20
30
40
pyruvate decarboxylase flux
(mol.[100 mol glucose]-1)
50
Figure 5 Observed fold increase in the SSres when independently refitting the LC-MS, GC-MS
and NMR 13C-labeling data for fixed values of (A) the PPP split-ratio, (B) the pyruvate carboxylase
catalyzed flux, (C) the reversibility of the oxaloacetate transporter and (D) the pyruvate decarboxylase
catalyzed flux. Metabolic flux-estimates for the LC-MS dataset were limited to glycolysis and the
PPP. As a result, the sensitivities of nodes B, C and D were not analyzed for the LC-MS dataset.
MG bypass flux
PPP split-ratio
Fold change in SSres
A
2
50
40
30
6
B
Flux (mol.[100 mol glucose]-1)
Fold change in SSres
3
5
NMR
LC-MS
GC-MS
4
3
20
1
10
0
0
4
8
12
biosynthetic g6p efflux
(mol.[100 mol glucose]-1)
16
0
.
2
1
0
0
4
8
12
16
biosynthetic g6p efflux
(mol.[100 mol glucose]-1)
Figure 6 Observed fold increase in the total SSres (A) and the SSres of the individual datasets (B)
when refitting the combined LC-MS, GC-MS and NMR 13C-labeling data for fixed values of the
biosynthetic efflux from the g6p pool. Estimated values for the PPP split-ratio and the MG bypass
flux for the different biosynthetic g6p effluxes are given in Figure 6-A.
Metabolic flux analysis of S. cerevisiae
3.3.4
Metabolite balancing versus 13C-based MFA
NAD(P)H balances The NADH and NADPH balance are used to constrain the fluxes derived
via metabolite balancing, but not those derived via 13C-labeling based MFA. The consistency
of the NADH and NADPH balance for the 13C-labeling derived fluxes (combined datasets,
Figure 3) was checked by calculating the total production of NADH and NADPH in the cytosol
and comparing these values to those derived via metabolite balancing. In these calculations
the enzyme Aald dehydrogenase was considered to be NADP+-dependent. In addition, the
pyruvate decarboxylase bypass flux was set at 0.08 mol·(mol glucose)-1, which corresponded
with the minimal flux needed to cover the biosynthetic need for cytosolic acetyl-CoA. A more
precise flux estimate could not be made due to the insensitivity of this node (See Figure 5-D).
Note that a minimal flux was chosen, since a high flux through the pyruvate decarboxylase
bypass is energetically highly unfavorable to the cell. The conversion of acetate into acetyl‑CoA
by the pyrophosphate-generating enzyme acetyl‑CoA synthetase costs two ATP equivalents.
The total NADPH production in the cytosol was calculated to be 0.56 mol NADPH·(mol glucose)1
for the 13C-derived flux-pattern. This is substantially higher than the 0.41 mol NADPH·(mol
glucose)-1 needed for biomass synthesis as calculated from the stoichiometric model used for
metabolite balancing. The total NADH production in the cytosol was calculated to be 0.94 mol
NADH·(mol glucose)-1. In total 0.80 mol NADH·(mol glucose)-1 is reduced in the cytosol via
glycerol synthesis, resulting in an excess of 0.14 mol NADH·(mol glucose)-1 for transportation
via a redox-shuttle. This value agrees with the 0.16 mol NADH·(mol glucose)-1 calculated via
metabolite balancing.
Upper glycolysis The absence of the triosephosphate isomerase in the tpi1Δnde1,2Δgut2Δ
strain assured an equal FBP aldolase catalyzed flux towards DHAP and GAP. Given the
measured glycerol production rate (0.80 mol·(mol glucose)‑1), this meant that 0.20 moles of
glucose had to be metabolized via a route other than the upper glycolysis, for instance via
assimilation or via the PPP. Alternatively, the surplus of carbon could be metabolized by an
alternative DHAP dissimilation pathway, such as the MG bypass. Comparison of the 13Cbased flux pattern with the flux pattern derived from metabolite balancing shows a similar
assimilation flux from the G6P pool. However, in the 13C-labeling based MFA a higher flux
through the oxidative branch of the PPP and, consequently, a lower flux through the MG
bypass is fitted compared to the metabolite balancing method. This discrepancy can to a
certain extent be explained by the sensitivity of the PPP split-ratio to small deviations in
glycerol formation. Every 0.01 mol·(mol glucose)-1 lowering of the glycerol yield will result in a
0.03 mol·(mol glucose)-1 increase in the PPP split-ratio. However, a second explanation for the
discrepancy in PPP split-ratio between the metabolite balancing and the 13C-labeling method
is the difference in the constraints imposed by the two MFA methods.
Metabolite balancing constrains the flux through the PPP based on assumptions on the
cytosolic NADPH demand of the cell, which leads to a diversion of excess carbon through
the MG bypass. Uncertainty about cofactor specificities of enzymes, especially in case of
multiple isoenzymes, the presence of transhydrogenation cycles and the occurrence of as yet
unknown sinks of NAD(P)H (e.g. caused by oxidative stress), makes predictions based upon
balances of reduction equivalents prone to error.
71
Chapter 3
72
In contrast, the 13C-labeling based MFA primarily constrains the flux through the MG bypass
as a result of the labeling of alanine. The measured m+1 mass isotopomer fraction and singlet
peak measured for ala-all and ala-α, respectively, are enough to account for the presence of
naturally labeled 13C, but leave little room for the inflow of 1‑[13C1] DHAP via the MG bypass.
The excess carbon is therefore shuttled to the other remaining option: the PPP. Generally, 13Clabeling based MFA will lead to more reliable estimates for the MG bypass flux and coupled
to that the PPP split-ratio, as these calculations are based upon actual (13C)-measurements.
In accordance with this the 13C-labeling based estimate of the MG bypass flux (0.5% of the
total glucose uptake flux) resembles the value reported by Martins et al. [22] for wild-type
S. cerevisiae much closer than the metabolite balancing based estimate of 5%. Seemingly,
the MG bypass only plays a minor role in diverting carbon away from glycerol formation.
Redox-shuttle mechanisms Metabolite balancing showed the necessity of incorporating a
putative NADH shuttle in the stoichiometric network model to allow for the reoxidation of the
surplus of NADH formed in the cytosol. In the absence of the external mitochondrial NADH
dehydrogenases and the glycerol-3-phosphate shuttle, S. cerevisiae apparently possesses
other mechanisms to oxidise cytosolic NADH. Several redox-shuttle mechanisms that
enable the transfer of reduction equivalents from the cytosol to the mitochondrion have been
proposed, such as the ethanol-acetaldehyde shuttle; the malate-oxaloacetate (MAL/OAA)
shuttle; the malate-aspartate (MAL/ASP) shuttle and the malate-pyruvate (MAL/PYR) shuttle
[2]. All key enzymes needed to operate these shuttles are present in S. cerevisiae. However,
no conclusive evidence for in vivo activity of these shuttles has been reported to date, except
maybe for the activity of the ethanol-acetaldehyde shuttle in an adh3Δ S. cerevisiae strain
[1]. In this context, it is relevant to note that the malate-aspartate shuttle has recently been
proposed to play an important role during growth of S. cerevisiae on acetate and fatty acids
[4].
The 13C-labeling MFA of this study provides additional insight into the redox-shuttle mechanism
used by S. cerevisiae. 13C-based flux-patterns displayed in Figure 3 show a considerable
reversibility for the transport of oxaloacetate. In the stoichiometric model for the 13C-labeling
based MFA both the cytosolic and mitochondrial pools of malate and oxaloacetate are lumped,
as no separate labeling information was available for malate. In principle, the transport of
oxaloacetate across the mitochondrial membrane can form part of the MAL/OAA shuttle or
the MAL/ASP shuttle. In the MAL/OAA shuttle cytosolic NADH is used to reduce oxaloacetate
to malate, which in turn is reoxidized to oxaloacetate in the mitochondrion producing NADH.
The MAL/ASP shuttle works similarly; apart from the fact that oxaloacetate is first converted
into aspartate which is then transported in symport with glutamate. Interestingly, cytosolic
malate dehydrogenase, a crucial enzyme in both shuttles, is deactivated in the presence
of high glucose concentrations [25]. This might explain why Overkamp et al. [29] observed
higher glycerol yields in batch cultivations of the tpi1Δnde1,2Δgut2Δ strain. Under glucose
excess conditions the MAL/ASP or MAL/OAA shuttles are inactive, leaving more NADH for
the formation of glycerol.
Direct transport of oxaloacetate (MAL/OAA shuttle) or transport in the form of aspartate (MAL/
ASP shuttle) could not be distinguished due to the fact that GC-MS and NMR analysis only
Metabolic flux analysis of S. cerevisiae
provide the labeling of aspartate and not that of its precursor oxaloacetate. If the observed
transport of oxaloacetate is indeed part of the MAL/ASP or the MAL/OAA shuttle, the net
NADH production in the mitochondrion (0.19 mol·(mol glucose)‑1 for the combined dataset,
see Figure 3) is enough to remove the surplus of NADH created in the cytosol as calculated
via metabolic balancing (0.16 mol·(mol glucose)‑1, see Figure 1). As shown in Figure 5, NMRderived relative intensities allowed for the most accurate estimate of the reversibility of the
oxaloacetate transporter. Based upon a twofold increase in the SSres, the net flux through the
oxaloacetate transport was estimated to be between 0.12 and 0.40 mol·(mol glucose)‑1.
It was shown for the metabolite balancing method that altering the cofactor specificity of the
cytosolic Aald dehydrogenase from NADP+ to NAD+ increased the PPP split-ratio from 16
to 21%, making the oxidative branch of the PPP the sole producer of cytosolic NADPH. As
a result of the changed cofactor specificity, additional NADH was produced in the cytosol,
thereby increasing the flux for the putative NADH shuttle from 0.16 mol·(mol glucose)‑1 to
0.26 mol·(mol glucose)‑1. This value is slightly higher than the exchange flux estimated for the
combined 13C-labeling dataset, but still falls within the range identified via the flux sensitivity
analysis (0.12 to 0.40 mol·(mol glucose)‑1).
Activity of the MAL/PYR shuttle, in which pyruvate is reduced to malate in the cytosol via
pyruvate carboxylase and malate dehydrogenase and again oxidized to pyruvate in the
mitochondrion via malic enzyme, is improbable given the estimated low flux through malic
enzyme. Activity of the ethanol-acetaldehyde shuttle could not be verified using the employed
13
C-labeling experiment, since no information on the 13C-label distribution of ethanol or
acetaldehyde was available.
3.4 Conclusions
This study provides further insight into the carbon and redox metabolism within a glycerol
overproducing tpi1∆nde1,2∆gut2∆ S. cerevisiae strain. For the first time the isotopic enrichment
of the intracellular compounds in a single 13C-labeling experiment was measured using GCMS, NMR and LC-MS. The combination of the three datasets resulted in similar flux-patterns
compared to the individual data sets and only a minor increase (<20%) in the summed SSres,
which shows that the three 13C measurement techniques yield consistent results. In addition,
combining the different techniques led to a more accurate final flux-pattern as the sensitivity
of the fluxes around several important metabolic nodes in the primary metabolism of the
tpi1Δnde1,2Δgut2Δ strain proved to be dependent on the method of analysis. For example,
NMR-derived relative intensities enabled an accurate estimation of the reversibility of the
oxaloacetate transporter, the PPP split-ratio was most accurately estimated by the LC-MS
derived mass fractions, while the MG bypass flux was best estimated via the GC-MS derived
mass fractions. This observation calls for an approach with multiple 13C-measurement tools in
future 13C-labeling based MFA experiments.
The combination of metabolite balancing and 13C-labeling based MFA provided physiological
evidence that three pathways were used to divert carbon away from glycerol formation (in
order of flux size): (i) the PPP, (ii) the assimilatory pathway towards storage carbohydrates
and (iii) the MG bypass. The 13C-derived PPP split-ratio and MG flux are believed to be more
73
Chapter 3
accurate than the fluxes derived via metabolite balancing, as the latter method relies heavily
on an uncertain NADPH balance. Based on the 13C-labeling data an almost negligible flux
through the MG bypass was fitted, accounting for 0.5% of the total glucose uptake flux.
Most carbon was diverted away from glycerol formation through the PPP (24%). Metabolite
balancing showed that an alternative mechanism for the oxidation of cytosolic NADH was
required to maintain redox balance. The 13C-derived exchange flux of oxaloacetate (lumped
together with malate) across the mitochondrial membrane provided evidence for the MAL/
OAA or alternatively the MAL/ASP shuttle as alternative redox mechanism.
74
Acknowledgements
The Ph.D. research of R.J.K. was financially supported by the Dutch EET program (Project
No. EETK20002) and DSM. The Ph.D. research of J.M.A.G. was financed by Tate & Lyle
Ingredients Americas, Inc. The research groups of J.T.P. and J.J.H. are part of the Kluyver
Centre for Genomics of Industrial Fermentation, which is supported by the Netherlands
Genomics Initiative. GC-MS measurements and data processing were made possible by
financial support (travel grant R81-743) of the Netherlands Organization for Scientific Research
(NWO). W.v.W acknowledges the hospitality and support of the Sauer Group of the Institute
of Biotechnology, ETH Zurich, Switzerland.
Metabolic flux analysis of S. cerevisiae
Appendix a: 13C-Labeling data
Table A-I Measured steady-state GC-MS mass isotopomer fractions, NMR fine structures and LC-MS
mass isotopomer fractions for an aerobic glucose-limited chemostat culture of a tpi1Δnde1,2Δgut2Δ
S. cerevisiae strain grown on a mixture of 10% [U-13C]glucose and 90% [1-13C]glucose. Standard three
letter abbreviations were used for the proteinogenic amino acids. The following NMR fine structures
were measured: singlets (s), doublets (d), triplets (t) and double doublets (dd), where d* indicates the
doublet with the larger one-bond scalar coupling constant. For the GC-MS mass fractions, (all) denotes
the complete amino acid, (‑1) is the amino acid minus the carboxyl group and (1+2) is the amino acid
fragment consisting of only the first two carbon atoms (position 1 being the carboxyl group).
GC-MS
AA fragment
Ala (all)
mass
fraction
m+0
m+1
m+2
m+3
Ala (-1)
m+0
m+1
m+2
Asp (all)
m+0
m+1
m+2
m+3
m+4
Asp (-1)
m+0
m+1
m+2
m+3
Asp (1+2) m+0
m+1
m+2
Glu (all)
m+0
m+1
m+2
m+3
m+4
m+5
Glu (-1)
m+0
m+1
m+2
m+3
m+4
Gly (all)
m+0
m+1
m+2
Gly (-1)
m+0
m+1
Ile (all)
m+0
m+1
m+2
m+3
m+4
m+5
m+6
IIe (-1)
m+0
m+1
m+2
m+3
NMR
Measured
0.8550
0.0410
0.0120
0.0920
0.8591
0.0470
0.0939
0.7310
0.1630
0.0540
0.0445
0.0075
0.7449
0.1648
0.0714
0.0190
0.8711
0.0480
0.0809
0.6757
0.1652
0.1316
0.0220
0.0035
0.0020
0.7126
0.1654
0.1065
0.0145
0.0010
0.8826
0.0338
0.0836
0.8935
0.1065
0.6576
0.1617
0.1115
0.0552
0.0110
0.0020
0.0010
0.6394
0.1758
0.1414
0.0340
AA fragment
Phe-α
Phe-β
Gly-α
His-α
His-β
Ser-α
Ser-β
Tyr-α
Tyr-β
Ala-α
Ala-β
Asp-α
Asp-β
fine
structure
s
d*
d
dd
s
d*
d
dd
s
d
s
d*
d
dd
s
d*
d
dd
s
d*
d
dd
s
d
s
d*
d
dd
s
d*
d
dd
s
d*
d
dd
s
d
s
d*
d
dd
s
d*
d
dd
LC-MS
Measured
0.1031
0.0066
0.0017
0.8886
0.0862
0.0192
0.7958
0.0988
0.1857
0.8143
0.1154
0.0001
-0.0002
0.8847
0.0891
0.0004
0.3477
0.5628
0.1284
0.0184
0.3877
0.4655
0.4823
0.5177
0.1115
-0.0002
0.0089
0.8798
0.0817
0.0222
0.7777
0.1184
0.1197
0.0280
0.0449
0.8074
0.1533
0.8467
0.2968
0.0693
0.2325
0.4013
0.2498
0.2521
0.3714
0.1266
Metabolite
G6P
F6P
FBP
M6P
2/3PG
PEP
6PG
P5P
massfraction
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+0
m+1
m+2
m+3
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
Measured
0.0768
0.7298
0.0615
0.0335
0.0186
0.0062
0.0736
0.0876
0.7065
0.0650
0.0391
0.0230
0.0071
0.0717
0.0920
0.6529
0.0680
0.0865
0.0646
0.0069
0.0292
0.0838
0.7157
0.0644
0.0357
0.0202
0.0063
0.0739
0.8372
0.0414
0.0158
0.1057
0.7934
0.0776
0.0211
0.1079
0.1103
0.6880
0.0672
0.0348
0.0178
0.0088
0.0730
0.6969
0.1413
0.0518
75
Chapter 3
76
m+4
m+5
Leu (all)
m+0
m+1
m+2
m+3
m+4
m+5
m+6
Leu (-1)
m+5
m+1
m+2
m+3
m+4
m+5
Lys (all)
m+0
m+1
m+2
m+3
m+4
m+5
m+6
Lys (-1)
m+0
m+1
m+2
m+3
m+4
m+5
Phe (all)
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+7
m+8
m+9
Phe (-1)
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+7
m+8
Phe (1+2) m+0
m+1
m+2
Pro (all)
m+0
m+1
m+2
m+3
m+4
m+5
Pro (-1)
m+0
m+1
m+2
m+3
m+4
Ser (all)
m+0
m+1
m+2
m+3
0.0080
0.0015
0.6977
0.0641
0.2082
0.0100
0.0180
0.0000
0.0020
0.6648
0.1379
0.1608
0.0250
0.0100
0.0015
0.6384
0.1668
0.1588
0.0280
0.0070
-0.0010
0.0020
0.6399
0.2227
0.1113
0.0241
0.0020
0.0000
0.6433
0.0882
0.0772
0.1062
0.0581
0.0130
0.0090
0.0050
0.0000
0.0000
0.6384
0.0913
0.1479
0.0426
0.0592
0.0090
0.0115
0.0000
0.0000
0.9005
-0.0010
0.1005
0.6777
0.1639
0.1248
0.0281
0.0035
0.0020
0.6996
0.1718
0.1087
0.0165
0.0035
0.8492
0.0642
0.0348
0.0518
Glu-α
s
d*
d
dd
Glu-β
s
d
t
Glu-γ
s
d*
d
dd
Ile-β
s
d
t
Ile-δ
s
d
Lys-β
s
d
t
Lys-γ
s
d
t
Leu-α
s
d*
d
dd
Leu-β
s
d
t
Leu-δ1
s
d
Leu-δ2
s
d
Met-α
s
d*
d
dd
Pro-α
s
d*
d
dd
Pro-δ
s
d
Arg-β
s
d
t
Arg-δ
s
d
Thr-α
s
d*
d
dd
Thr-β
s
d
t
Thr-γ
s
d
Val-α
s
d*
d
dd
Val-γ1
s
d
Glycerol c1 s
d
0.3054
0.1215
0.4413
0.1318
0.6854
0.2520
0.0626
0.1397
0.7552
0.0166
0.0885
0.1030
0.7666
0.1104
0.7044
0.2956
0.6747
0.3077
0.0176
0.6641
0.2933
0.0426
0.1363
0.0189
0.7541
0.0907
0.8036
0.1964
0.0000
0.1541
0.8459
0.8881
0.1119
0.2978
0.0886
0.2042
0.4094
0.3912
0.1614
0.3494
0.0980
0.1325
0.8675
0.6680
0.2881
0.0439
0.1457
0.8543
0.3029
0.0661
0.2346
0.3964
0.2629
0.6284
0.1087
0.7021
0.2979
0.1085
0.0225
0.7757
0.0933
0.1651
0.8349
0.8588
0.1412
S7P
CO2
m+3
m+4
m+5
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+7
m+0
m+1
0.0300
0.0276
0.0524
0.4812
0.2752
0.1047
0.0412
0.0338
0.0468
0.0113
0.0058
0.7900
0.2100
Metabolic flux analysis of S. cerevisiae
Ser (-1)
Ser (1+2)
Thr (all)
Thr (-1)
Tyr (all)
Tyr (-1)
Tyr (1+2)
Val (all)
Val (-1)
Val (1+2)
m+0
m+1
m+2
m+0
m+1
m+2
m+0
m+1
m+2
m+3
m+4
m+0
m+1
m+2
m+3
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+7
m+8
m+9
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+7
m+8
m+0
m+1
m+2
m+0
m+1
m+2
m+3
m+4
m+5
m+0
m+1
m+2
m+3
m+4
m+0
m+1
m+2
0.8342
0.1091
0.0568
0.8944
0.0010
0.1046
0.7329
0.1678
0.0464
0.0469
0.0060
0.7488
0.1665
0.0718
0.0130
0.6707
0.0815
0.0655
0.0980
0.0495
0.0150
0.0080
0.0040
0.0065
0.0015
0.6500
0.0866
0.1497
0.0401
0.0561
0.0055
0.0090
0.0010
0.0020
0.9016
-0.0020
0.1004
0.7560
0.0570
0.0880
0.0870
0.0030
0.0090
0.7490
0.0670
0.1650
0.0100
0.0090
0.8831
0.0070
0.1099
Threhalose c1
s
d
0.9009
0.0991
77
Chapter 3
References
[1] Bakker, B.M., Bro, C., Kotter, P., Luttik, M.A.H., van
Dijken, J.P. and Pronk, J.T. (2000) The mitochondrial
alcohol dehydrogenase adh3p is involved in a redox
shuttle in Saccharomyces cerevisiae. J. Bacteriol.
182, 4730-4737.
[2] Bakker, B.M., Overkamp, K.M., van Maris, A.J.A.,
Kötter, P., Luttik, M.A.H., van Dijken, J.P. and Pronk,
J.T. (2001) Stoichiometry and compartmentation of
NADH metabolism in Saccharomyces cerevisiae.
FEMS Microbiol. Rev. 25, 15-37.
78
[3] Bonarius, H.P.J., Schmid, G. and Tramper, J.
(1997) Flux analysis of underdetermined metabolic
networks: the quest for the missing constraints.
Trends in Biotechnology 15, 308-314.
[4] Cavero, S., Vozza, A., del Arco, A., Palmieri, L.,
Villa, A., Blanco, E., Runswick, M.J., Walker, J.E.,
Cerdan, S., Palmieri, F. and Satrustegui, J. (2003)
Identification and metabolic role of the mitochondrial
aspartate-glutamate transporter in Saccharomyces
cerevisiae. Mol Microbiol 50, 1257-1269.
[5] Christensen, B., Gombert, A.K. and Nielsen, J. (2002)
Analysis of flux estimates based on 13C-labeling
experiments. Eur. J. Biochem. 269, 2795-2800.
[6] Compagno, C., Boschi, F. and Ranzi, B.M. (1996)
Glycerol production in a triose phosphate isomerase
deficient mutant of Saccharomyces cerevisiae.
Biotechnol. Prog. 12, 591-595.
[7] Compagno, C., Brambilla, L., Capitanio,
F., Ranzi, B.M. and Porro, D. (2001)
of the glucose metabolism in a triose
isomerase-negative Saccharomyces
mutant. Yeast 18, 663-670.
D., Boschi,
Alterations
phosphate
cerevisiae
[8] Eglinton, J.M., Heinrich, A.J., Pollnitz, A.P.,
Langridge, P., Henschke, P.A. and Lopes, M.D. (2002)
Decreasing acetic acid accumulation by a glycerol
overproducing strain of Saccharomyces cerevisiae
by deleting the ALD6 aldehyde dehydrogenase
gene. Yeast 19, 295-301.
[9] Fischer, E. and Sauer, U. (2003) A novel metabolic
cycle catalyzes glucose oxidation and anaplerosis in
hungry Escherichia coli. J. Biol. Chem. 278, 4644646451.
[10] Fischer, E. and Sauer, U. (2003) Metabolic flux
profiling of Escherichia coli mutants in central carbon
metabolism using GC-MS. Eur. J. Biochem. 270,
880-891.
[11] Fischer, E., Zamboni, N. and Sauer, U. (2004) High-
throughput metabolic flux analysis based on gas
chromatography-mass spectrometry derived 13C
constraints. Anal. Biochem. 325, 308-316.
[12] Follstad, B.D. and Stephanopoulos, G. (1998) Effect
of reversible reactions on isotope label redistribution
- Analysis of the pentose phosphate pathway. Eur. J.
Biochem. 252, 360-371.
[13] Geertman, J.M.A., van Dijken, J.P. and Pronk,
J.T. (2006) Engineering NADH metabolism in
Saccharomyces cerevisiae: formate as an electron
donor for glycerol production by anaerobic, glucoselimited chemostat cultures. FEMS Yeast Res. In
press.
[14] Gombert, A.K., Moreira dos Santos M., Christensen,
B. and Nielsen, J. (2001) Network identification
and flux quantification in the central metabolism of
Saccharomyces cerevisiae under different conditions
of glucose repression. J. Bacteriol. 183, 1441-1451.
[15] Grandier-Vazeille, X., Bathany, K., Chaignepain,
S., Camougrand, N., Manon, S. and Schmitter,
J.M. (2001) Yeast mitochondrial dehydrogenases
are associated in a supramolecular complex.
Biochemistry 40, 9758-9769.
[16] Grotkjaer, T., Akesson, M., Christensen, B., Gombert,
A.K. and Nielsen, J. (2004) Impact of transamination
reactions and protein turnover on labeling dynamics
in 13C-labeling experiments. Biotechnol. Bioeng. 86,
209-216.
[17] Kleijn, R.J., van Winden, W.A., Ras, C., van Gulik,
W.M., Schipper, D. and Heijnen, J.J. (2006) 13CLabeled gluconate tracing as a direct and accurate
method for determining the pentose phosphate
pathway split ratio in Penicillium chrysogenum. Appl.
Environ. Microbiol. 72, 4743-4754.
[18] Kleijn, R.J., van Winden, W.A., van Gulik, W.M.
and Heijnen, J.J. (2005) Revisiting the 13C-label
distribution of the non-oxidative branch of the
pentose phosphate pathway based upon kinetic and
genetic evidence. FEBS J 272, 4970-4982.
[19] Lange, H.C. (2002) Quantitative physiology of
Saccharomyces cerevisiae using metabolic network
analysis. Thesis, Delft University of Technology, the
Netherlands.
[20] Lange, H.C., Eman, M., van Zuijlen, G., Visser, D.,
van Dam, J.C., Frank, J., Teixeira de Mattos, M.J.
and Heijnen, J.J. (2001) Improved rapid sampling
for in vivo kinetics of intracellular metabolites in
Saccharomyces cerevisiae. Biotechnol. Bioeng. 75,
406-415.
Metabolic flux analysis of S. cerevisiae
[21] Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E.,
Sauer, U. and Szyperski, T. (2001) Central carbon
metabolism of Saccharomyces cerevisiae explored
by biosynthetic fractional 13C-labeling of common
amino acids. Eur. J. Biochem. 268, 2464-2479.
[32] Rojo, E.E., Guiard, B., Neupert, W. and Stuart,
R.A. (1998) Sorting of D-lactate dehydrogenase to
the inner membrane of mitochondria. Analysis of
topogenic signal and energetic requirements. J. Biol.
Chem. 273, 8040-8047.
[22] Martins, A.M., Cordeiro, C.A. and Ponces, F. (2001)
In situ analysis of methylglyoxal metabolism in
Saccharomyces cerevisiae. FEBS lett. 499, 41-44.
[33] van der Heijden, R.T.J.M. (1991) State estimation
and error diagnosis for biotechnological processes.
TU Delft, Delft, the Netherlands.
[23] Mashego, M.R., van Gulik, W.M., Vinke, J.L. and
Heijnen, J.J. (2003) Critical evaluation of sampling
techniques for residual glucose determination in
carbon-limited chemostat culture of Saccharomyces
cerevisiae. Biotechnol. Bioeng. 83, 395-399.
[34] van Winden, W.A. (2002) 13C-labeling technique for
metabolic network and flux analysis - theory and
applications. Thesis, Delft University of Technology,
the Netherlands.
[24] Meaden, P.G., Dickinson, F.M., Mifsud, A., Tessier,
W., Westwater, J., Bussey, H. and Midgley, M.
(1997) The ALD6 gene of Saccharomyces cerevisiae
encodes a cytosolic, Mg2+-activated acetaldehyde
dehydrogenase. Yeast 13, 1319-1327.
[25] Minard, K.I. and Mcalisterhenn, L. (1992) Glucose
induced degradation of the Mdh2 isozyme of malate
dehydrogenase in yeast. J. Biol. Chem. 267, 1745817464.
[26] Navarro-Avino, J.P., Prasad, R., Miralles, V.J.,
Benito, R.M. and Serrano, R. (1999) A proposal
for nomenclature of aldehyde dehydrogenases in
Saccharomyces cerevisiae and characterization of
the stress-inducible ALD2 and ALD3 genes. Yeast
15, 829-842.
[27] Neuberg, C. and Reinfurth, E. (1919) Die Festlegung
der Aldehydstufe bei der alkoholischen Gärung;
ein experimenteller Beweis der AcetaldehydBrenztraubensäuretheorie. Biochem. Z. 89, 365415.
[28] Neuberg, C. and Reinfurth, E. (1919) Natürliche und
erzwungene Glycerinbildung bei der alkoholische
Gärung. Biochem. Z. 92, 234-266.
[29] Overkamp, K.M., Bakker, B.M., Kötter, P., Luttik,
M.A.H., van Dijken, J.P. and Pronk, J.T. (2002)
Metabolic engineering of glycerol production in
Saccharomyces cerevisiae. Appl. Environ. Microbiol.
68, 2814-2821.
[30] Perrenoud, A. and Sauer, U. (2005) Impact of
global transcriptional regulation by ArcA, ArcB, Cra,
Crp, Cya, Fnr, and Mlc on glucose catabolism in
Escherichia coli. J. Bacteriol. 187, 3171-3179.
[31] Piper, M.D., Hong, S.P., Ball, G.E. and Dawes, I.W.
(2000) Regulation of the balance of one-carbon
metabolism in Saccharomyces cerevisiae. J. Biol.
Chem. 275, 30987-30995.
[35] van Winden, W.A., Heijnen, J.J., Verheijen, P.J. and
Grievink, J. (2001) A priori analysis of metabolic
flux identifiability from 13C-labeling data. Biotechnol.
Bioeng. 74, 505-516.
[36] van Winden, W.A., Schipper, D., Verheijen, P.J.
and Heijnen, J.J. (2001) Innovations in generation
and analysis of 2D [13C,1H] COSY NMR spectra for
metabolic flux analysis purposes. Metab Eng. 3, 322343.
[37] van Winden, W.A., van Dam, J.C., Ras, C., Kleijn,
R.J., Vinke, J.L., van Gulik, W.M. and Heijnen, J.J.
(2005) Metabolic-flux analysis of Saccharomyces
cerevisiae CEN.PK113-7D based on mass
isotopomer measurements of 13C-labeled primary
metabolites. FEMS Yeast Res. 5, 559-568.
[38] van Winden, W.A., van Gulik, W.M., Schipper, D.,
Verheijen, P.J., Krabben, P., Vinke, J.L. and Heijnen,
J.J. (2003) Metabolic flux and metabolic network
analysis of Penicillium chrysogenum using 2D [13C,
1
H] COSY NMR measurements and cumulative
bondomer simulation. Biotechnol. Bioeng. 83, 7592.
[39] van Winden, W.A., Wittmann, C., Heinzle, E. and
Heijnen, J.J. (2002) Correcting mass isotopomer
distributions for naturally occurring isotopes.
Biotechnol. Bioeng. 80, 477-479.
[40] Verduyn, C., Postma, E., Scheffers, W.A. and van
Dijken, J.P. (1992) Effect of benzoic acid on metabolic
fluxes in yeasts: a continuous culture study on the
regulation of respiration and alcoholic fermentation.
Yeast 8, 501-517.
[41] Wiechert, W. (2001) 13C metabolic flux analysis.
Metab Eng. 3, 195-206.
[42] Wiechert, W. (2002) Modeling and simulation: tools
for metabolic engineering. J. Biotechnol. 94, 37-63.
79
CHAPTER
13
4
C-labeling based metabolic network
and flux analysis of penicillin G
producing and non-producing
chemostat cultures of
Penicillium chrysogenum
Roelco J. Kleijn, Wouter A. van Winden,
Walter M. van Gulik, Cor Ras,
Dick Schipper and Joseph J. Heijnen
Manuscript in preparation
Chapter 4
82
ABSTRACT
13
C-labeling experiments were conducted to characterize the metabolic fluxes of a highyielding Penicillium chrysogenum strain grown under penicillin-G producing and nonproducing conditions. Two 13C-quantification techniques were used; liquid chromatographymass spectrometry (LC-MS) providing direct information on the labeling of the intracellular
primary metabolites and 2D [13C,1H] correlation nuclear magnetic resonance spectroscopy
(NMR) yielding the 13C-labeling of proteinogenic amino acids fragments.
Prior to performing a 13C‑labeling based metabolic flux analysis (MFA), a node by node
analysis of the proposed primary metabolic network model was performed using labeling
redundancies in the measured 13C-labeling data and enzymatic analysis. This resulted in: (i) a
modified compartmental origin of several amino acid precursors, (ii) the identification of rapidly
equilibrated metabolic pools and, (iii) evidence for the presence of a mitochondrial acetyl-CoA
transporter and of the enzymes threonine aldolase, phosphoenol-pyruvate carboxykinase and
malic enzyme.
Using the obtained network model, metabolic fluxes were derived for P. chrysogenum by first
independently fitting the LC-MS and NMR dataset, followed by a flux fit for the combined
datasets. Despite the updated reaction network model all flux fits had to be rejected on statistical
grounds. Furthermore, the two 13C-analysis techniques yielded very different flux estimates
for the major part of the metabolism. Further research is needed to determine whether the
complexity of fungal cultivations, due to cell differentiation and complex formation leads to
different metabolic flux pattern when applying the nowadays available 13C-quantification
techniques.
Metabolic flux analysis of P. chrysogenum
4.1 INTRODUCTION
The family of β-lactam antibiotics includes penicillins, cephalosporins, cephamycins,
oxacephems, carbapenems and monobactams. Over the past 60 years the biosynthesis of βlactam antibiotics has developed from a specialty process yielding gram scale quantities, to a
bulk process with annual production rates exceeding 60.000 tons. This increase in production
rate is primarily the result of an ongoing quest for superior strains with respect to productivity
and fitness, first via the classical approach of random mutagenesis and screening, and
increasingly in the last decade by the more rational approach of metabolic engineering [31].
In both bacteria and filamentous fungi the genes involved in the biosynthesis of β-lactams
antibiotics are arranged in clusters [23]. Therefore, the most straightforward way to increase
the productivity of β-lactams would be to over-express these gene clusters. In accordance
with this, Theilgaard et al. [30] showed that over-expression of the three clustered structural
biosynthetic genes for penicillin (pcbAB, pcbC and penDE) in the wild-type derived Penicillium
chrysogenum Wis54-1255 strain led to an increased penicillin production rate. Furthermore,
comparison of this strain with the nowadays available high penicillin-producing strains of
P. chrysogenum, shows that classical strain improvement has led to a 6-14-fold increase in
the copy number of the gene cluster coding for enzymes of the penicillin biosynthesis pathway
[9]. However, there is a limit to increasing penicillin production via further amplification of the
penicillin gene cluster. At some point the supply of carbon precursors, cofactors and energy by
the primary metabolism will become more important for penicillin production, causing a shift in
the metabolic bottleneck from the secondary metabolism to the primary metabolism [13, 34].
Metabolic engineering plays a crucial role in identifying these shifts in metabolic bottlenecks
as a consequence of a further increase of the productivity, as it allows for a more rational
approach towards strain improvement by redesigning only specific parts of the metabolism.
In short, metabolic engineering consists of a metabolic characterization step in which more
insight in the cellular metabolism of the production strain is obtained, followed by a genetic
modification step, in which this insight is used to improve the properties of the strain using
recombinant DNA technology. Note that this is an iterative process, as multiple steps of
characterization and modification are, most probably, needed to further improve the strain
[31].
Metabolic flux analysis (MFA) is one of the major characterization tools in metabolic engineering,
enabling the quantification of intracellular fluxes by combining uptake and secretion rates with
a predefined metabolic network model. Nowadays, 13C-labeling experiments are increasingly
being used to complement the conventional MFA method in order to: (i) remove uncertain
cofactor balances from the metabolic network model, (ii) validate the metabolic network
and (iii) estimate fluxes through parallel pathways and metabolic cycles [42]. Several MFA
studies have investigated the relation between penicillin production and the supply of carbon
precursors (valine and cysteine), cofactors (NADPH) and energy (ATP) by the primary
metabolism of P. chrysogenum [3, 4, 15, 16, 34, 39]. However, different flux outcomes make it
difficult to unambiguously draw conclusions on potential metabolic bottlenecks.
Jorgensen et al. [16] were the first to set up a metabolic network model to study the flux
distribution during three phases of a fed batch cultivation of a high-yielding former production
strain of P. chrysogenum. A similar version of the model was used by Henriksen et al. [15] to
83
Chapter 4
84
study the metabolic fluxes of the same strain during chemostat cultivations. Both these studies
presented model based evidence that β-lactam production is correlated with the flux through
the PPP; the main source of cytosolic NADPH in P. chrysogenum [14]. A similar observation,
albeit with different absolute values, was made by van Gulik et al. [34] when analyzing
metabolic fluxes in a different high-yielding former production strain of P. chrysogenum grown
in chemostat cultures at different growth-rates on glucose, acetate, ethanol or xylose as carbon
source and ammonia or nitrate as nitrogen source. Christensen et al. [3, 4] were the first to
perform a 13C-labeling based MFA for P. chrysogenum, and compared the metabolism of a
high- and a low-yielding strain of P. chrysogenum at a growth-rate of 0.06-0.07 h-1. In contrast
to earlier observations, they found that penicillin production had no influence on the estimated
flux through the oxidative part of the PPP. Van Winden et al. [39] compared the conventional
MFA and the 13C-labeling based MFA method by growing the same strain used by van Gulik et
al. [34] in a continuous culture at a growth rate of 0.03 h-1 with either ammonia or nitrate as the
nitrogen source. They showed that the outcomes of the 13C-labeling based MFA substantially
differed from those of the conventional MFA. In addition, fluxes determined using 13C-labeling
data were shown to be highly dependent on the chosen metabolic network.
In this study, 13C-labeling experiments were used to characterize the metabolic fluxes of a
high-yielding P. chrysogenum strain grown under penicillin-G producing and non-producing
conditions. As described above, a validated metabolic network model is fundamental to
accurately estimating metabolic fluxes. Therefore, prior to performing a 13C‑labeling based
MFA, a node by node analysis of the proposed primary metabolic network was performed using
labeling redundancies in the measured 13C-labeling data. Two 13C-quantification techniques
were used; LC-MS providing direct information on the labeling of the intracellular primary
metabolites [40] and 2D [13C,1H] COSY NMR yielding the 13C-labeling of proteinogenic amino
acid fragments [22, 39].
4.2 THEORY
Fundamental to analyzing metabolic fluxes in a micro-organism is a valid metabolic network
model. When performing a 13C‑labeling experiment, a first validation of the metabolic network
model can be performed by analyzing the isotopomer distributions of metabolites directly
surrounding the structural elements constituting the model. Based upon the number of fluxes
that enter and leave a metabolite pool, three types of structural elements (metabolic nodes) can
be distinguished within a metabolic network model: linear, diverging and converging nodes.
As has been described by van Winden et al. [38], a property of linear or divergent node is that
they contain only one influx and therefore lead to labeling redundancies for the metabolites
directly surrounding the node. Consequently, a correct designation of the linear and divergent
nodes in the metabolic network model can be tested by comparing the isotopomer labeling of
the metabolites directly produced from the same node. Different isotopomer labeling patterns
point to an unaccounted influx in the examined node and thus require a revision of the initial
metabolic network model.
The linearity and divergence of several metabolic nodes in primary metabolism can be
examined by directly measuring the mass isotopomer fractions of primary metabolites via
Metabolic flux analysis of P. chrysogenum
LC-MS. Similar mass fractions for the primary metabolites directly surrounding a divergent or
linear node are indications for valid model assumptions. In addition, the 13C-labeling of primary
metabolites can be used to identify metabolic pools connected by a highly reversible reaction.
Well known examples of highly equilibrated pools are the hexose monophosphate pools in
the upper glycolysis, the triose monophosphate pools in the lower glycolysis, the pentose
monophosphate pools in the PPP and the four‑carbon organic acids pools in the TCA‑cycle
(tri carboxylic acid).
The above described characteristics of metabolic nodes can also be used to study the amino
acid biosynthesis pathways, which consist of linear and divergent nodes [29]. Due to the
divergence of these pathways several amino acids share common precursors; allowing for
assumptions with respect to the common origin of amino acids to be tested by comparing the
labeling of these amino acids [22, 37]. In this study the labeling of amino acids was quantified
via NMR. NMR derived fine structures correspond with the 13C labeling of the central carbon
atom, the two adjacent carbon atoms and, in some cases, the carbon atoms next to these. If all
the carbon atoms in the analyzed carbon fragment of the compared amino acids are obtained
from the same precursor, one can simply compare the fine structures of the two supposedly
identical amino acid fragments. However, if only parts of the molecular fragments measured in
the multiplets are derived from the same precursor, only these two parts should be compared
with each other. This is done by summing the relative intensities of the isotopomers differing
only in the carbon atom that is not derived from the same precursor molecule (see Figure 1).
The identified linear, divergent and highly equilibrated pools in the proposed metabolic
network model form the basis of the model reduction proposed by van Winden et al. [38].
A
α
s
d
d
d*
dd
B
s
d*
d
dd
γ
β
α
β
d
s‘
d‘
s
d*
d
dd
γ
d
s
dd
dd
d‘
α
d‘
dd d*
β
dd
dd
dd
d*
α
d* dd
s‘
β
d‘
Figure 1 13C-labeling comparison
s‘ = s+d
d‘ = d*+dd of two amino acids in which only
part of the carbon backbone is
derived from the same precursor.
The following NMR fine structures
are depicted: singlets (s),
doublets (d and d*), and double
doublets (dd), where d* indicates
the doublet with the larger onebond scalar coupling constant.
s‘ = s+d
In this example only the first two
d‘ = d*+dd carbon atoms of the amino acids
stem from the same precursor
molecule. By summing the fine
structures corresponding to the
labeling patterns 010+011 (s+d)
and 110+111 (d*+dd) for both
amino acid fragments (where ‘0’
denotes 12C and ‘1’ denotes 13C),
one compares only the labeling
pattern of the first two carbon
atoms. The third carbon atom
can be either 12C or 13C-labeled.
85
Chapter 4
Different 13C-labeling
Model reduction takes place by removing the metabolite pools that have only one influx (linear
and divergent nodes) and lumping the metabolite pools that are in isotopic equilibrium due
to fast exchange reactions. In Figure 2 the applied procedure for constructing, validating and
reducing a metabolic network model prior to performing a 13C‑labeling based MFA is presented
in a flow diagram.
86
1) Define reaction network model based upon:
• Textbook knowledge
• Scientific publications
• Thermodynamic principles
Revise reaction
network model
2) Use isotopomer measurements (LC-MS, GC-MS and/or NMR)
to check for:
• Similar 13C-labeling for the metabolites forming a linear or
a divergent node
• Similar 13C-labeling for the metabolite pools connected
by a high exchange flux
Similar 13C-labeling
3) Simplify the reaction network by removing metabolite pools with a
single influx (linear and divergent nodes) and lumping metabolite pools
that are connected by a high exchange flux
4) Use simplified reaction network model for
based metabolic flux analysis
13C-labeling
-
Figure 2 Flow diagram for constructing, validating and reducing a metabolic network model prior to
performing a 13C‑labeling based MFA.
4.3
MATERIALS AND METHODS
4.3.1
Strain, media and cultivation conditions
A high-producing industrial P. chrysogenum strain (code name DS17690, kindly donated by
DSM Anti‑Infectives, Delft, The Netherlands) was cultivated in a carbon-limited chemostat
system at a dilution rate of 0.02 h‑1, both in the absence and presence of phenylacetic acid
(PAA), the side-chain precursor for penicillin‑G biosynthesis. Cultivations were carried out as
previously described [19]. Cells were grown on a scaled down minimal medium described by
van Gulik et al. [34]. The medium supported a biomass dry weight concentration of about 1 to
1.5 g/L. The minimal medium contained: 3.3 g/L glucose·H2O, 0.68 g/L sodium-acetate·3H2O,
0.35 g/L KH2PO4, 1.54 g/L (NH4)2SO4, 0.22 g/L MgSO4·7H2O and 0.90 mL/L trace element
solution. The trace element solution contained 75 g/L Na2-EDTA·2H2O, 2.5 g/L CuSO4·5H2O,
Metabolic flux analysis of P. chrysogenum
10 g/L ZnSO4·7H2O, 10 g/L MnSO4·H2O, 20 g/L FeSO4·7H2O, and 2.5 g/L CaCl2·2H2O. Sodiumacetate was added to the medium to introduce an additional inflow of labeled carbon into the
metabolism for a better estimation of the fluxes in the lower part of the metabolism.
At the end of the batch phase (typically after 50 h) the reactor was switched to continuous
mode using the minimal medium described above. After 120 h this medium was replaced by
a chemically identical medium, but with 60 Cmol% of the naturally labeled glucose replaced
by specifically labeled 1‑13C1 glucose (Sigma-Aldrich, St. Louis, MO, USA), 20 Cmol% of the
naturally labeled glucose replaced by [U‑13C]glucose (Sigma-Aldrich) and 100 Cmol% of the
naturally labeled acetate replaced by [U‑13C]acetate (Sigma-Aldrich).
The penicillin‑G production medium was prepared by the addition of 0.533 g/L PAA to the
previously described batch and the continuous medium. The PAA-level in the medium was
chosen such that the steady state residual concentration in the chemostat was approximately 3
mM. At this concentration PAA was neither limiting for penicillin‑G production nor inhibiting the
growth of P. chrysogenum. Medium was prepared according to Kleijn et al. [19]. The CO2 and
O2 concentrations in the offgas from the chemostat culture were measured online with a nondispersive/paramagnetic infrared Rosemount NGA 2000 gas analyzer (Fisher-Rosemount,
Hanau, Germany).
4.3.2
Dry weight determination, filtrate sampling and extracellular metabolite
analysis
Duplicate 10 mL samples were withdrawn from the bioreactor for determining the biomass
dry weight concentrations. Samples were filtered over preweighted glass fiber filters (PALL,
East-Hills, NY, USA) and dried at 70oC for at least 24 h. The collected filtrate was immediately
frozen in liquid nitrogen and stored at ‑80oC prior to analysis. The extracellular concentrations
of penicillin‑G and PAA in the filtrate samples were measured by isocratic HPLC analysis
using a Platinum EPS C18 column (Alltech, Deerfield, IL, USA) at 30°C. The mobile phase
consisted of 28% acetonitrile (v/v) with 5 mM KH2PO4 and 6 mM H3PO4. The extracellular
concentrations of the byproducts of penicillin‑G synthesis; ortho-hydroxyphenylacetic acid
(o‑OH-PAA), 6-oxopiperidine-2-carboxylic acid (OPC), isopenicillin‑N (IPN), 8‑hydroxypenillic
acid (8‑HPA) and penicilloic acid (PIO), were determined with the HPLC method described by
Christensen et al. [5]. The extracellular 6‑amino‑penicillanic acid (6‑APA) concentration was
determined using the HPLC method described by van Can et al. [32].
Samples for residual substrate determination were acquired by rapidly sampling 2 mL of broth
into a syringe containing precooled stainless steel beads (‑18oC), immediately followed by
separation of cells and medium by filtration as described by Mashego et al. [24]. The glucose
and acetate concentrations in the samples were determined enzymatically (Enzytec, Scil
Diagnostics, Viernheim, Germany).
4.3.3
Metabolite sampling and LC-MS analysis
After approximately 185 h of chemostat cultivation on 13C‑labeled medium, samples for
intracellular metabolite isotopomer determinations were obtained. Hereby 1 mL of broth was
rapidly withdrawn from the bioreactor, followed by direct injection of the sample in 5 mL of a
60% (v/v) methanol/water mixture (‑40oC) for instantaneous quenching of the cell metabolism
87
Chapter 4
as described by Lange et al. [21]. In total 16 intracellular metabolite samples were acquired.
Samples were combined (pair wise) to compensate for the low biomass concentration [19] and
the intracellular metabolites were extracted from the samples as described previously [26]. All
samples were stored at ‑80oC, prior to LC‑MS analysis. The mass isotopomer distributions of
the intracellular metabolites were measured as described by van Winden et al. [40]. Metabolites
were first separated by high-performance anion exchange chromatography (Waters, Milford,
MA, USA) followed by MS analysis with a Quatro‑LC triple quadrupole mass spectrometer
(Micromass Ltd., Manchester, UK) equipped with an electrospray ionization interface.
Quantitative analysis of the following metabolites was carried out: glucose‑6‑phosphate (G6P),
fructose‑6‑phosphate (F6P), 6‑phosphogluconate (6PG), mannose‑6‑phosphate (M6P),
1,6‑fructose-bisphosphate (FBP), phosphoenol‑pyruvate (PEP), the combined pool of 2- and
3‑phosphoglycerate (2/3PG), the combined pool of xylulose‑5‑phosphate, ribose‑5‑phosphate
and ribulose‑5‑phosphate (P5P), sedoheptulose‑7‑phosphate (S7P), α‑ketoglutarate (AKG),
succinate (SUC), fumarate (FUM) and malate (MAL).
88
13
C‑Label distribution of glucose and acetate
4.3.4
The isotopic enrichment of the glucose and acetate in the feed of the 13C‑based labeling
experiments were determined via LC-MS analysis. Since glucose could not be directly measured
on the LC-MS, the mixture of [1‑13C]glucose, [U‑13C]glucose and naturally labeled glucose was
first phosphorylated to G6P by incubating 200 µL of medium with 15 µL of 0.25M ATP (pH 7.0)
and 3 µL of hexokinase (1500 U/mL, Roche Diagnostics, Almere, The Netherlands) for 30
minutes at 30oC. Enzyme and metabolites were separated by centrifugation at 6000 g for 30
min on an Ultrafree‑MC 10.000 NMWL spinfilter (Millipore, Billerica, MA, USA). Filtrates were
stored at ‑80oC prior to LC-MS analysis.
4.3.5Broth sampling and 2D [13C,1H] COSY NMR analysis
After withdrawal of the samples for intracellular metabolite analysis, biomass samples for
NMR analysis were obtained by filtrating 200 mL of the broth over a glass fiber filter (PALL).
The filter cake was washed with 0.9% NaCl-solution and demineralized water. Prior to NMR
analysis the biomass was hydrolyzed, lyophilized and dissolved in D2O as described by van
Winden et al. [39]. NMR measurements were performed at 600 Mhz at 37oC on a Bruker
Avance 600 spectrometer. The recorded 2D[13C,1H] COSY spectra were processed by means
of the spectral fitting software described by van Winden et al. [36]. The resulting data are
relative intensities of fine structures observed in the multiplets of the following proteinogenic
amino acids: phenylalanine‑α, ‑β, glycine‑α, histidine‑α, ‑β, ‑δ, serine‑α, ‑β, tyrosine‑α, ‑β, ‑δ,
‑ε, alanine‑α, ‑β, aspartate‑α, ‑β, glutamate‑α, ‑β, isoleucine‑α, ‑β, ‑γ1, ‑γ2, ‑δ, lysine‑β, ‑γ,‑δ,
‑ε, leucine‑α, ‑β, ‑γ, ‑δ1, ‑δ2, proline‑α, ‑β, ‑γ, ‑δ, arginine‑β, ‑γ, ‑δ, threonine ‑β, ‑γ, and valine‑α,
‑γ1, ‑γ2. In the penicillin‑G producing culture 2D [13C,1H] COSY spectra were also recorded
for the penicillin‑G in the filtrate, resulting in additional relative intensities for valine‑α, ‑γ1, ‑γ2
and cysteine‑α, ‑β. Standard three letter abbreviations are used for the amino acids in the
remainder of the text. Note that the nomenclature for numbering the carbon atoms of amino
acids used from here on is as follows; C is the carboxyl group of the amino acid: α is the
carbon atom next to C; β is the carbon atom next to α, etc.
Metabolic flux analysis of P. chrysogenum
4.3.6
Enzymatic analysis
Cells (approximately 80 mg dry weight) were harvested from the glucose-limited chemostat
cultures for the preparation of cell extracts as described by Harris et al. [14]. Cell extracts were
analyzed for G6P dehydrogenase (EC 1.1.1.49) and glucose oxidase (EC 1.1.3.4) activity
using the assay described by Harris et al. [14]. Threonine aldolase (EC 4.1.2.5) activity was
tested using the assay described by van Maris et al. [35]. PEP carboxykinase (EC 4.1.1.49)
activity was measured using the assay described by de Jong-Gubbels [7].
4.3.7
Metabolic flux analysis
Metabolic network Model Starting point for the metabolic network model used in this study
was the stoichiometric model of P. chrysogenum developed by van Gulik et al. [34]. Based
upon textbook and thermodynamic knowledge, reversibilities were added for the appropriate
reactions. Furthermore, the input of isotopic enrichment data enabled the removal of the
conserved moiety balances (i.e. NADH, NADPH) which are generally based on incomplete
stoichiometric information [1]. Further refinements to the metabolic network model were made
based upon the procedure described in the theory section. In short, the following changes
were made: (i) several amino acid biosynthesis routes were adapted as outlined in the result
section (ii) the conventional reactions of the non-oxidative branch of the PPP were replaced
by metabolite specific, reversible, C2 and C3 fragment producing and consuming half reactions
as proposed by Kleijn et al. [20] (iii) the synthesis of glycine via the reversible enzyme
threonine aldolase was included (iv) the glycine decarboxylase complex (GDC) catalyzing the
reversible cleavage of glycine into CO2 and 5,10‑CH2‑H4folate (E‑C1) was included [28] (v) the
reversible transport of acetyl‑CoA from the cytosol to the mitochondrion was introduced (vi)
the gluconeogenetic enzyme PEP carboxykinase catalyzing the conversion of oxaloacetate
(OAA) into PEP was included (vii) malic enzyme catalyzing the mitochondrial decarboxylation
of malate into pyruvate (PYR) was included and (viii) a scrambling reaction was included for
the symmetrical molecule succinate (which formed part of the four‑carbon pool (succinate,
fumerate, malate and oxaloacetate).
To correctly estimate the isotopic enrichment of the CO2 pool, the inflow of naturally labeled
CO2 as a result of aeration was also included in the metabolic network model. No distinction
was made between CO2 produced and consumed in different compartments. A schematic
representation of the metabolic network model is shown in Figure 5. Note that reversible
reactions were modeled as separate forward and backward reactions and are referred to as
net and exchange fluxes, where:
v net = v forward − v backward
(4.1)
v exchange = min(v forward ,v backward )
(4.2)
Flux fitting procedure The employed flux fitting procedure has been described in detail by
van Winden et al. [40]. In short, the procedure uses the cumomer balances and cumomer to
isotopomer mapping matrices introduced by Wiechert et al. [43] to calculate the isotopomer
distributions of metabolites in a pre-defined metabolic network model for a given flux set. Using
a sequential quadratic programming algorithm the flux set that gave the best correspondence
89
Chapter 4
between the measured and simulated 13C‑label distribution was selected and denoted as the
optimal flux fit. All calculations were performed in Matlab 7.0 (The Mathworks Inc, Natick,
USA).
90
4.4 RESULTS AND DISCUSSION
4.4.1
Steady state characterization
P. chrysogenum was cultivated aerobically in a carbon-limited chemostat system at a dilution
rate of 0.02 h‑1, both in the absence and presence of PAA. Table 1‑A shows the steadystate biomass dry weight concentrations and the calculated recoveries of carbon and PAA.
Measured biomass specific conversion rates in steady-state are shown in Table 1‑B. The
measured biomass dry weight concentrations were in accordance with values predicted
from the yield and maintenance parameters published by van Gulik et al. [33]; namely 1.08
g/L and 1.41 g/L for the producing and non-producing chemostat cultivation, respectively.
Furthermore, the biomass specific production rates of the β‑lactam compounds were in
accordance with those reported by van Gulik et al. [33] and Kleijn et al. [18]. Carbon recoveries
were between 90-95%, most probably due to the formation of polymeric byproducts (peptides
and polysaccharides) in the filtrate of the culture [26, 34] which were not measured. The total
amount of carbon containing (by) products in the filtrate was determined via TOC analysis and
was assumed to consist of extracellular peptides and polysaccharides at a ratio of 1:2. The
recovery of consumed PAA as the sum of penicillin-G, PIO and o-OH-PAA was 102.9±2.6%,
indicating that PAA was not metabolized by the cells.
Table 1-A Steady-state biomass concentrations and recoveries for carbon and PAA in a penicillin-G
producing and non-producing chemostat cultivation operated at a growth-rate of 0.02 h-1.
Biomass concentration and
recovery balances
Unit
Biomass concentration
g/L
Carbon recovery
(%)
97.6±5.6
92.9±1.7
Degree of reduction
(%)
103.7±4.9
102.5±3.2
PAA recovery
(%)
102.9±2.6
-
Producing
Non-producing
1.07±0.06
1.34±0.05
Data reconciliation based upon the conservation relations for the elements and gross error
detection according to van der Heijden et al. [41] indicated an error in the measured biomass
specific O2 consumption rate. This error was most probably caused by improper drying of the
air before it entered the gas analyzer. Apart from the O2 consumption rate, no other significant
discrepancies were observed between the measured and the reconciled conversion rates (see
Table 1‑B). For both chemostat cultivations the p‑values associated with the reconciliation
were higher than 0.05, indicating that the discrepancy between the measured and reconciled
conversion rates could be accounted for by measurement error. The CO2 concentrations
in the offgas of the chemostat were measured via a gas analyzer equipped with a nondispersive infrared absorption cell. Since the absorption for 13CO2 is lower from that of 12CO2
and since the offgas contained both 12CO2 and 13CO2 during the 13C‑labeling phase, the total
Metabolic flux analysis of P. chrysogenum
Table 1-B Measured and reconciled biomass specific conversion rates in steady state for a penicillin-G
producing and non-producing chemostat cultivation operated at a growth-rate of 0.02 h-1.
Biomass specific
conversion rates
(mmol·CmolX-1·h-1)
Producing (p ≥ 0.68)a
Non-producing (p ≥ 0.48)a
‘Measured’
‘Measured’
Reconciled
Reconciled
Glucose consumption
-8.36±0.75
-7.85
-6.96±0.56
-6.35
Acetate consumption
-2.52±0.23
-2.44
-2.05±0.17
-1.95
Oxygen consumption
-29.71±2.45
-24.99
-22.94±1.70
-15.46
PAA consumption
-0.57±0.21
-0.56
<0.01
0.00
Carbondioxide production
27.09±2.22
26.49
16.98±1.24
16.47
Biomass production
19.49±0.28
19.50
20.70±0.29
20.72
0.47±0.05
0.46
<0.01
0.00
IPN production
<0.01
0.00
<0.01
0.00
6-APA production
<0.01
0.00
0.05±0.00
0.05
b
Penicillin-G & PIO production
8-HPA production
<0.01
0.00
<0.01
0.00
o-OH-PAA production
0.10±0.02
0.10
<0.01
0.00
OPC production
0.08±0.01
0.08
<0.01
0.00
0.37±0.57
0.60
1.43±0.28
1.47
Extracellular peptidesc
Extracellular polysaccharides
0.75±1.14
1.20
2.85±0.56
2.92
P‑values were determined via a χ2 ‑distribution and denote the probability that the discrepancy between
the measured and reconciled conversion rates is a result of measurement error. P-values ≤ 0.05 were
considered as statistically significant, thus indicating a clear deviation between the measured and
reconciled conversion rates.
b
The specific O2 consumption rate was found to be erroneous, and was thus not used in the final
reconciliation of the specific conversion rates.
c
Based upon TOC measurements of the filtrate. The missing carbon in the filtrate carbon balance was
assigned to extracellular peptides and polysaccharides at a ratio of 1:2 [34].
c
a
CO2 concentration in the offgas was underestimated. As a result only the CO2 concentrations
before the 13C‑labeling phase were used for calculating the CO2 production rate.
After reaching isotopic steady state, the two 13C‑labeled chemostat cultivations were sampled
and the mass isotopomer distributions of the primary metabolites and relative intensities of the
proteinogenic amino acid fragments were analyzed via LC-MS and NMR, respectively (Table
A-I and A-II). On first sight little difference in 13C‑labeling is observed between the penicillin‑G
producing and non-producing chemostat. This is confirmed by the parity plots shown in Figure
3. The relative difference in labeling between the two chemostats is the largest for the NMR
dataset and lies, generally, between 0.01-0.05. An exception to this are the relative intensities
of glycine, which show a difference in the order of 0.15. The biggest difference in 13C‑labeling
between the two LC-MS datasets is seen for the mass isotopomer distribution of phosphoenol
pyruvate (PEP).
The measured mass isotopomer distributions of the 13C‑labeled glucose and the acetate
added to the feed are shown in Table A-I and were found to be in close agreement with
the applied 13C‑labeling of 20% [1‑13C1]glucose, 60% [u‑13C6]glucose, 20% naturally labeled
91
Chapter 4
glucose and 100% [u‑13C2]acetate. The mass isotopomer distribution of glucose was analyzed
separately for each cultivation, while the mass isotopomer distribution of acetate was only
determined once. Reason for this was that the applied labeled glucose consisted of a mixture
of naturally, C1 and uniformly labeled glucose, while the acetate added to the feed was solely
uniformly labeled. Measured mass distributions were used to calculate the isotopic purity of
the uniformly labeled acetate and glucose:
isotopic purity = 100 * n
(4.3)
1.0
Non-producing chemostat
Where, m is the monoisotopic mass, n is the total number of carbon atoms and F(x) is the
fraction of mass isotopomer x. The isotopic purity of glucose was calculated to be 98.1%
and 98.5% in the producing and non-producing chemostat cultivation, respectively. Note that
this means that a considerable amount of glucose molecules (approximately 10%), are not
uniformly labeled. The isotopic purity of the uniformly labeled acetate was calculated to be
98.7%. All isotopic purities appeared to be within the specification of the manufacturer (>98%
isotopically pure).
Non-producing chemostat
92
F(m + n)
F(m + n − 2) + F(m + n − 1) + F(m + n)
A
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
Producing chemostat
1.0
1.0
B
0.8
0.6
gly-α
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Producing chemostat
Figure 3 Parity plots depicting the measured mass isotopomer fractions (A) and the relative intensities
(B) in the penicillin-G producing and non-producing chemostat cultivation.
4.4.2
Metabolic network analysis
Prior to performing a 13C‑labeling based MFA, a node by node analysis of the proposed primary
metabolic network was performed (see Theory section). Using the labeling redundancies in the
LC‑MS derived mass isotopomer fractions (See Appendix A, Table A-I) and the NMR derived
relative intensities (See Appendix A, Table A-II), the consistency of the metabolic nodes in the
proposed metabolic network model was verified.
LC-MS derived mass isotopomer fractions Similar mass isotopomer fractions were found
for G6P and 6PG (see Table A-I) confirming that the 6PG node has one single influx and
thus leads to a labeling redundancy. Importantly, the similar labeling of G6P/6PG suggests
Metabolic flux analysis of P. chrysogenum
the absence of activity of the enzyme glucose oxidase. The presence of glucose oxidase
would create an additional inflow into the 6PG pool by converting extracellular glucose into
extracellular gluconate, which is subsequently taken up by the cell and converted into 6PG
by the enzyme gluconate kinase. In that case the mass isotopomer distribution of 6PG would
have lied in-between those of G6P and glucose. This is not observed in Table A-I.
To identify the metabolite pools connected by large exchange fluxes, 13C-labeling patterns
were compared for: the hexose pools in the upper glycolysis (G6P & F6P and F6P & M6P),
the pools connecting the upper and the lower glycolysis (FBP & 2/3PG), the triose pools
in the lower glycolysis (2/3PG & PEP) and the C4 carboxylic acid pools in the TCA‑cycle
(succinate & fumerate & malate). Similar mass isotopomer distributions were measured for
the G6P & F6P pools and the carboxylic acids (succinate to a somewhat lesser extent, see
Table A-I), suggesting that the corresponding enzymes indeed operate near equilibrium in
P. chrysogenum. The high reversibility of these enzymes was also confirmed by Nasution
et al. [26], based upon constant mass action ratios obtained from intracellular metabolite
measurements in P. chrysogenum during short term highly dynamic conditions in a glucose
pulse experiment. On the other hand, the difference in the labeling between FBP & 2/3PG (by
combining the labeling of two 2/3PG molecules, not shown), and 2/3PG & PEP indicated that
their interconversion was not fast enough to ensure equilibrium. Furthermore, the difference
in labeling indicates that there was an additional inflow of carbon into either the 2/3PG or
the PEP pool. This additional inflow most likely resulted from the gluconeogenic reaction
catalyzed by PEP carboxykinase, converting one molecule of oxaloacetate into PEP and CO2
at the expense of one ATP. A possible explanation for the presence of this enzyme was its
induction by the co-substrate acetate and/or the low residual glucose concentration at a growth
rate of 0.02 h‑1. Note that in combination with the anaplerotic enzyme pyruvate carboxylase
(converting pyruvate into oxaloacetate) and the glycolytic enzyme pyruvate kinase (converting
PEP into pyruvate) a futile cycle is created at the cost of 1 ATP/cycle.
NMR derived relative intensities NMR derived relative intensities were used to test the
assumptions made with respect to the common origin of amino acids made in the proposed
metabolic network model as described in the theory section. Note that in most cases only
the common origin of the amino acid (fragments) could be validated and not the actual
precursor used for synthesizing the amino acid (fragments). Reason for this was that LC‑MS
derived mass isotopomer fractions of the primary metabolites can not be transformed to their
corresponding relative intensities for comparison with the measured relative intensities of the
amino acid fragments.
Assumption 1: Tyr‑Cαβ and Phe‑Cαβ are both synthesized from cytosolic PEP
Table 2‑A clearly demonstrates that the 13C‑labeling of the 3 carbon fragment around Phe‑α
was identical to that of Tyr‑α. Note that the underlined carbon atom refers to the atom for which
the NMR spectrum is recorded. The only isotopomers that showed some deviation were the
011 and 111 fragments for the penicillin‑G producing chemostat cultivation. This assumption
was further checked by comparing the labeling of the fragments Phe‑β and Tyr‑β. To determine
the labeling of these fragments the relative intensities of the isotopomers differing only in the
93
Chapter 4
γ carbon atom were summed (see Figure 1). Table 2‑B shows that identical labeling patterns
were also observed for Phe‑αβ and Tyr‑αβ, confirming that the first three carbon atoms of Tyr
and Phe are synthesized from the same precursor, which is cytosolic PEP.
Table 2-A Relative intensities of Phe-Cαβ and Tyr-Cαβ under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
Fragment
Phe-Cαβ
Tyr-Cαβ
Phe-Cαβ
Tyr-Cαβ
010
0.126
0.120
0.098
0.093
110
0.155
0.190
0.179
0.185
011
0.055
0.055
0.046
0.045
111
0.665
0.635
0.677
0.677
Table 2-B Summed relative intensities of Phe-αβ and Tyr-αβ under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
94
Producing
Non-producing
Fragment
Phe-αβ
Tyr-αβ
Phe-αβ
Tyr-αβa
01
0.566
0.568
0.580
nd
11
0.434
a
nd = not determined
0.432
0.420
nd
Assumption 2: Asp, Thr and Ile‑Cα and Ile‑γ1δ are synthesized from cytosolic oxaloacetate
Tables 3‑A to 3‑D show the measured relative intensities for Asp, Thr and Ile. The similarity
in 13C‑labeling indicates that these three amino acids are indeed synthesized from the same
precursor, oxaloacetate. The amino acid Met is also assumed to be synthesized from cytosolic
oxaloacetate, but was not taken into account in this comparison as for both chemostats
undefined peaks were found in the corresponding NMR spectra. Similar coupling constants
of the carbon atoms in the Thr‑β spectrum made it impossible to distinguish between the 110
and the 011 fragment in Table 3‑B. To allow for comparison, the fine structures of the 110 (d)
and 011 (d*) fragment for Asp‑αβγ were summed. Even though the origin of the oxaloacetate
used for the biosynthesis of Asp, Thr, Ile and Thr can not be deduced from this data, all
previously described metabolic network models [3, 15, 34] indicate a cytosolic origin. Based
Table 3-A Relative intensities of Asp-Cαβ and Thr-Cαβ under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
fragment
Asp-Cαβ
Thr- Cαβa
Asp- Cαβa
Thr- Cαβ
010
0.312
nd
nd
0.305
110
0.217
nd
nd
0.219
011
0.251
nd
nd
0.243
111
0.220
a
nd = not determined
nd
nd
0.234
Metabolic flux analysis of P. chrysogenum
upon the presented results there is no reason to assume the precursor has a different cellular
localization.
Table 3-B (Summed) relative intensities of Asp-αβγ and Thr-αβγ under both penicillin-G producing and
non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Asp-αβγ
Thr-αβγ
Asp-αβγ
Thr-αβγ
010
0.334
0.292
0.322
0.315
110+011
0.479
0.483
0.463
0.477
111
0.187
0.226
0.214
0.208
Table 3-C Summed relative intensities of Ile-Cα, Asp-Cα and Thr-Cα under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Ile-Cα
Asp-Cα
Thr-Cα
Ile-Cα
Asp-Cαa
Thr-Cα
01
0.533
0.529
nd
0.552
nd
0.523
0.471
nd
0.448
nd
0.477
11
0.468
a
nd = not determined
a
Table 3-D Relative intensities of Ile-γ1δ and Thr-βγ under both penicillin-G producing and non-producing
conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is recorded.
Producing
Non-producing
Fragment
Ile-γ1δ
Thr-βγ
Ile-γ1δ
Thr-βγ
01
0.430
0.417
0.439
0.421
11
0.571
0.583
0.561
0.580
Assumption 3: Gly is synthesized by cleavage of the αβ carbon‑bond of Ser
Several Gly synthesis pathways are described in the literature, of which the most common
route is via the reversible cleavage of the αβ carbon‑bond of Ser. This is also the route
included in the metabolic network model of van Gulik et al. [34]. An alternative Gly synthesis
pathway is via the cleavage of the αβ carbon‑bond of Thr by threonine aldolase, producing
both Gly and acetealdehyde. In Table 4 the summed relative intensities of Gly are compared
with the Cα‑fragments of Ser‑α and Thr‑α and Asp‑α. The labeling of Asp‑Cα is shown instead,
because the labeling of Thr‑Cα was not measurable for the producing chemostat culture.
In assumption 2 it was demonstrated that both amino acids are synthesized from the same
precursor and thus have an identical labeling.
The results in Table 4 show a big difference between the two chemostats. The relative intensities
of Gly‑Cα in the penicillin‑G producing chemostat correspond better with those of Thr‑Cα/
Asp‑Cα, while under non‑producing conditions the relative intensities of Gly correspond much
better with Ser‑Cα. These results suggest that under penicillin‑G producing conditions the Gly
biosynthesis route via Thr is active, while under non-producing conditions Gly is synthesized
from Ser. The difference in biosynthetic pathways might be explained by a higher demand for
cytosolic cysteine under producing conditions. Cytosolic cysteine is synthesized from serine
95
Chapter 4
and is one of the carbon precursors for synthesizing β‑lactams. If under penicillin‑G producing
conditions the majority of serine is used for synthesizing cysteine, the cleavage of threonine
can provide the cell with the needed Gly.
Another alternative route for glycine synthesis is via the glycine decarboxylation complex
which catalyzes the reversible conversion of CO2 and 5,10‑CH2‑H4folate into glycine. The
presence of this reaction could not be tested, but it was incorporated in the metabolic network
model used for the 13C‑labeling based MFA.
Table 4 Summed relative intensities of Ser-Cα, Gly-Cα, Thr-Cα and Asp-Cα under both penicillin-G
producing and non-producing conditions. The underlined carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
96
Non-producing
Fragment Ser-Cα
Gly-Cα
Thr-Cαa
Asp-Cα
Ser-Cα
Gly-Cα
Thr-Cα
Asp-Cαa
01
0.458
nd
0.529
0.281
0.309
0.523
nd
11
0.693
0.542
a
nd = not determined.
nd
0.471
0.719
0.691
0.477
nd
0.307
Assumption 4: Ser is synthesized from 3PG
LC‑MS results revealed that no (isotopic) equilibrium exists between the metabolites 2/3PG
and PEP. In view of this it was examined whether the 13C‑labeling of Ser‑α (synthesized from
3PG) differs from that of Phe‑α and Tyr‑α, (synthesized from PEP, see Assumption 1). Table
5‑A shows that the relative intensities of Ser‑Cαβ indeed differed from those of Tyr‑Cαβ and
Phe‑Cαβ.
In assumption 3 it was explained that Gly could be produced via the reversible cleavage
of the α‑β carbon‑bond of Ser. If this is the case, the labeling of Phe‑Cαb and of Tyr‑Cαβ
should be different from that of Ser‑Cαβ, no matter the size of exchange reaction between
2/3PG and PEP. Consequently, only the Cα‑fragment of Tyr, Phe and Ser were compared with
each other in Table 5‑B. Surprisingly, the relative intensities of the Cα‑fragments are rather
similar. This contradicts with the observation that the mass isotopomer fractions of PEP and
2/3PG are different (Table A-I). However, the NMR results presented in Table 5‑B are relative
intensities of fine structures, while the LC‑MS yields a distribution of the mass isotopomers.
Comparison of the constraints imposed on the isotopomers of 2/3PG and PEP showed that
these metabolites could display differing mass isotopomer distributions, but at the same time
similar relative intensities (results not shown). In other words, similarity of NMR data does not
necessarily imply that the isotopomer labeling of two compounds are identical.
Table 5-A Relative intensities of Tyr-Cαβ, Phe-Cαβ and Ser-Cαβ under both penicillin-G producing and
non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Tyr-Cαβ
Phe-Cαβ
Ser-Cαβ
Tyr-Cαβ
Phe-Cαβ
Ser-Cαβ
010
0.120
0.126
0.165
0.093
0.098
0.149
110
0.190
0.155
0.142
0.185
0.179
0.132
011
0.055
0.055
0.271
0.045
0.046
0.301
111
0.635
0.665
0.422
0.677
0.677
0.419
Metabolic flux analysis of P. chrysogenum
Table 5-B Summed relative intensities of Tyr-Cα, Phe-Cα and Ser-Cα under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Tyr-Cα
Phe-Cα
Ser-Cα
Tyr-Cα
Phe-Cα
Ser-Cα
01
0.310
0.281
0.307
0.278
0.277
0.281
11
0.690
0.719
0.693
0.722
0.723
0.719
Assumption 5: Ala is synthesized from cytosolic pyruvate, while Val‑Cα, Val‑βc1, Ile‑βγ2,
Leu‑γδ1 are synthesized from mitochondrial pyruvate
In the model of van Gulik et al. [34] the pyruvate used for Ala synthesis has a cytosolic origin,
while Val and Ile‑βγ2 and Leu‑γδ1 are derived from mitochondrial pyruvate. The measured
labeling distributions in this study seem to refute this assumption, as comparable summed
relative intensities are found for Val- Cα and Ala‑Cα (Table 6-A) and Ala‑αβ, Val‑βγ1, Ile‑βγ2
and Leu‑γδ1 (Table 6-B).Note that, by itself, this does not necessarily imply that the origin of
the precursor is the same for all amino acids, since the labeling of cytosolic and mitochondrial
pyruvate can also be identical. Further proof for the mitochondrial origin of the pyruvate used
for Ala synthesis is given in assumption 6, where it is shown that the labeling of cytosolic PEP
(Tyr‑Cαβ and Phe‑Cαβ) does not correspond with the labeling of the pyruvate used for Ala
synthesis.
Table 6-A Summed relative intensities of Ala-Cα and Val-Cα under both penicillin‑G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
Fragment
Ala-Cα
Val-Cαa
Ala-Cα
Val-Cα
01
0.323
0.349
0.332
0.296
11
0.677
0.652
0.668
0.704
a
Based upon the Val- spectrum measured in penicillin-G. The spectrum of the free amino acid fragment
of Val-α was not measurable as it overlapped with Thr-α.
Table 6-B Summed relative intensities of Ala-αβ, Val-βγ1, Ile-βγ2 and Leu-γδ1 under both penicillin-G
producing and non-producing conditions. The underlined carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
Non-producing
Fragment Ala-αβ
Val-βγ1
Ile-βγ2
Leu-γδ1
Ala-αβ
Val-βγ1
Ile-βγ2
Leu-γδ1
01
0.560
0.600
0.558
0.588
0.601
0.595
0.555
0.569
11
0.440
0.400
0.442
0.412
0.399
0.405
0.446
0.432
Assumption 6: The carbon atoms of Leu‑Cα originate from mitochondrial acetyl‑CoA while
those of Lys‑Cα originate from cytosolic acetyl‑CoA
Table 7‑A shows the relative intensities for the Cα fragment of Leu and Lys. Unfortunately
the spectrum of Lys‑α in the penicillin‑G producing cultivation overlapped with Glu‑α, making
a comparison between the two amino acid fragments impossible. However, based upon the
97
Chapter 4
98
similar relative intensities found for the non-producing cultivations the acetyl‑CoA used for
synthesizing Leu and Lys seems to have the same origin. This is in contradiction with the
assumption made by van Gulik et al. [34] that the AcCoA used for synthesizing the two amino
acids differed in its cellular localization.
Since AcCoA is synthesized from the second and third carbon atom of pyruvate, information
about the origin of the AcCoA was obtained by comparing the summed relative intensities in
Table 7‑A with those of Table 6‑B. The 13C‑labeling of the AcCoA used for Leu‑Cα and Lys‑Cα
differed from the labeling of the pyruvate used for the synthesis of Ala‑αβ, Val‑βγ1, Leu‑γδ1
and Ile‑βγ2. Considering the above it is highly unlikely that the AcCoA used for Leu‑Cα
and Lys‑Cα originated from pyruvate. More likely is that the AcCoA in Leu‑Cα and Lys‑Cα
predominantly originated from the fully labeled acetate added to the medium and thus has a
cytosolic localization. This idea is strengthened by the high relative intensity of the ‘11’ fragment
(corresponding with uniformly labeled AcCoA) in Table 7‑A. The somewhat lower intensity
of the ‘11’ fragment under non-producing chemostat indicates a slightly bigger contribution
of other reactions (e.g. pyruvate decarboxylase, threonine aldolase). The cytosolic origin of
Leu‑Cα was also observed by Christensen et al. [3] when conducting a 13C‑labeling based
MFA on a penicillin‑V producing P. chrysogenum chemostat culture. Note that the values in
Table 7-A are relative values of the ‘11’ fragment with respect to the ‘01’ fragment and thus
give no information on the total fraction of Leu‑Cα and Lys‑Cα fragments that is uniformly
labeled.
Based upon these observations the following metabolic network around AcCoA is
hypothesized:
i.
The labeled acetate added to the medium is the main source of cytosolic AcCoA. AcCoA
is produced from acetate via the enzyme AcCoA synthetase.
ii. Most pyruvate produced in the cytosol is transported into the mitochondrion via a
pyruvate transporter, where it is converted to AcCoA, needed to sustain the flux through
the TCA‑cycle. As a result, the cytosolic pyruvate decarboxylase complex, which
decarboxylates pyruvate into acetaldehyde, has a very low activity.
iii. Given the high relative intensity measured for uniformly labeled AcCoA, there seems to
be little transport of AcCoA from the mitochondrion to the cytosol.
In order to verify hypothesis 2, the relative intensities of Tyr‑Cαβ and Phe‑Cαβ were compared
to those of Ala‑Cαβ (Tables 7‑B&C). It was assumed that the majority of the pyruvate produced
in the cytosol stemmed from PEP. The results are not very conclusive; the relative intensities
of Table 7‑C match very well, while the relative intensities of Table 7‑B lie further apart. The
differences observed between Tyr/Phe‑Cαβ and Ala‑Cαβ in Table 7‑B can be explained by an
additional influx into the mitochondrial pyruvate pool. A possible candidate for such a reaction
is the conversion of malate into pyruvate catalyzed by malic enzyme. In Saccharomyces
cerevisiae this enzyme has been extensively measured and found to be active in the
mitochondrion. In line with this, Harris et al. [14] found trace activities of a NADP+-dependent
malic enzyme in a glucose-limited chemostat cultivation of P. chrysogenum. Altogether the
above results give no reason to reject hypothesis 2.
Metabolic flux analysis of P. chrysogenum
Table 7-A Summed relative intensities of Leu-Cα and Lys-Cα under both penicillin-G producing and
non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Leu-Cα
Lys-Cαa
Leu-Cα
Lys-Cα
01
0.038
nd
0.122
0.153
11
0.962
a
nd = not determined
nd
0.878
0.847
Table 7-B Summed relative intensities of Tyr-Cαβ, Phe-Cαβ and Ala-Cαβ under both penicillin-G
producing and non-producing conditions. The underlined carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
Non-producing
Fragment
Tyr-Cαβ
Phe-Cαβ
Ala-Cαβ
Tyr-Cαβ
Phe-Cαβ
Ala-Cαβ
010
0.120
0.126
0.142
0.093
0.098
0.141
110
0.190
0.155
0.181
0.185
0.179
0.191
011
0.055
0.055
0.100
0.045
0.046
0.082
111
0.635
0.665
0.577
0.677
0.677
0.586
Table 7-C (Summed) relative intensities of Tyr-αβ, Phe-αβ and Ala-αβ under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Tyr-αβ
Phe-αβ
Ala-αβ
Tyr-αβa
Phe-αβ
Ala-αβ
01
0.568
0.566
0.560
nd
0.580
0.601
0.434
0.440
nd
0.420
0.399
11
0.432
a
nd = not determined
Assumption 7: The carbon skeletons of Glu, Pro, Arg and Lys‑βγδe are synthesized from
cytosolic α‑ketoglutarate
In the model of van Gulik et al. [34] the amino acids Glu, Pro, Arg and Lys‑βγδε were all
assumed to be synthesized in the cytosol with α-ketogluterate as precursor. In Tables 8‑A
to 8‑D the relative intensities of the amino acid fragments are presented. The results clearly
show that the amino acids were synthesized from the same precursor. However, based upon
these results no statement could be made about the localization of the α-ketogluterate.
In the first part of the TCA‑cycle AcCoA and oxaloacetate are condensed into citrate by citrate
synthase, citrate is isomerized to iso‑citrate by aconitase and isocitrate is subsequently
oxidatively decarboxylated to α-ketogluterate by isocitrate dehydrogenase. In the
decarboxylation step of isocitrate the third carbon atom is split off in the form of CO2, meaning
that the fourth and fifth carbon atom of α-ketogluterate can be traced back to AcCoA. Based
on this, the 13C‑labeling of the last two carbon atoms of α-ketogluterate were compared with
the AcCoA used for the synthesis of Leu‑Cα and Lys‑Cα (Table 8‑E) and the AcCoA produced
from the pyruvate used for Ala (Table 8‑F&G). Note that in this comparison carbon atoms C
and α of Leu and Lys should be compared with, respectively, carbon atoms δ and γ of Glu,
Pro and Arg. The relative intensities of the last two carbon atoms of Glu, Arg, Pro and Lys
99
Chapter 4
corresponded much better with the relative intensities of Ala than with those of Leu‑Cα and
Lys‑Cα. Considering the hypothesis postulated for the distribution of AcCoA (assumption 6)
it is likely that the α-ketogluterate used for the synthesis of Glu, Arg, Pro and Lys originated
from the mitochondrion.
Table 8-A Relative intensities of Glu-Cαβ and Pro-Cαβ under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
Fragment Glu-Cαβ
Pro-Cαβ
Glu-Cαβ
Pro-Cαβ
010
0.298
0.304
0.316
0.317
110
0.217
0.224
0.245
0.244
011
0.268
0.259
0.229
0.232
111
0.217
0.213
0.211
0.207
Table 8-B Relative intensities of Glu-αβγ, Pro-αβγ, Arg-αβγ and Lys-βγδ under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
100
Producing
Non-producing
Fragment Glu-αβγ
Pro-αβγ
Arg-αβγ
Lys-βγδ
Glu-αβγ
Pro-αβγ
Arg-αβγ
Lys-βγδ
010
0.380
0.374
0.367
0.350
0.334
0.295
0.299
0.333
110+011
0.477
0.466
0.483
0.492
0.479
0.502
0.488
0.487
111
0.144
0.160
0.150
0.158
0.188
0.203
0.213
0.181
Table 8-C Relative intensities of Glu-βγδ, Pro-βγδ, Arg-βγδ and Lys-γδε under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment Glu-βγδa
Pro- βγδ
Arg-βγδ
Lys-γδε
Glu-βγδ
Pro- βγδ
Arg-βγδ
Lys-γδε
010
nd
0.310
0.304
0.326
0.301
0.287
0.318
0.311
110+011
nd
0.510
0.514
0.512
0.330
0.501
0.508
0.520
111
nd
0.179
a
nd = not determined
0.183
0.162
0.164
0.211
0.174
0.169
Table 8-D Relative intensities of Pro-γδ, Arg-γδ and Lys-δε under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
Fragment
Pro-γδ
Arg-γδ
Lys-δε
Pro-γδ
Arg-γδ
Lys-δε
01
0.204
0.208
0.192
0.185
0.187
0.190
11
0.796
0.792
0.809
0.815
0.813
0.810
The difference in relative intensities observed in Tables 8‑F&G indicate that the labeling of the
mitochondrial AcCoA produced from pyruvate was different from the AcCoA incorporated into
citrate via citrate synthase. Furthermore, the tables show that the relative intensity of the fully
labeled carbon fragment (11) is higher for Glu‑δγ, Pro‑δγ, Arg‑δγ and Lys‑εδ than for Ala‑αβ.
Metabolic flux analysis of P. chrysogenum
The higher relative intensity of the fully labeled carbon fragment suggests the transport of
uniformly labeled AcCoA (originating from the uniformly labeled acetate added to the medium)
into the mitochondrion, thereby causing an additional influx in the mitochondrial AcCoA pool.
AcCoA can be transported into the mitochondrion either directly via an AcCoA transporter
(e.g. a carnithine shuttle), or indirectly via the transport of cytosolic citrate synthesized from
cytosolic AcCoA and oxaloacetate via citrate synthase. The presence of a cytosolic citrate
synthase seems rather unlikely, given the earlier finding that Glu, Pro, Arg and Lys are most
probably synthesized from mitochondrial α-ketogluterate. Furthermore, localization studies
performed by Harris et al. [14] showed that citrate synthase was exclusively located in the
mitochondrion of P. chrysogenum cells. The direct transport of AcCoA from the cytosol to the
mitochondrion is thus the most plausible pathway.
Table 8-E Summed relative intensities of Leu-Cα, Lys-Cα and Glu-δγ under both penicillin-G producing
and non-producing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum
is recorded.
Producing
Non-producing
Fragment
Leu-Cα
Lys-Cα
Glu-δγ
Leu-Cα
Lys-Cα
Glu-δγ
01
0.038
nd
nd
0.122
0.153
0.465
nd
nd
0.878
0.847
0.535
a
11
0.962
a
nd = not determined.
a
Table 8-F Summed relative intensities of Ala-αβ and Glu-δγ under both penicillin-G producing and nonproducing conditions. The underlined carbon atom refers to the atom for which the NMR spectrum is
recorded.
Producing
Non-producing
Fragment
Ala-αβ
Glu-δγa
Ala-αβ
Glu-δγ
01
0.560
nd
0.601
0.465
11
0.440
a
nd = not determined.
nd
0.399
0.535
Table 8-G (Summed) relative intensities of Arg-δγ, Pro-δγ, Lys-εδ and Ala-αβ under both penicillin-G
producing and non-producing conditions. The underlined carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
Non-producing
Fragment Arg-δγ
Pro-δγ
Lys-εδ
Ala-αβ
Arg-δγ
Pro-δγ
Lys-εδ
Ala-αβ
10
0.208
0.204
0.192
0.243
0.187
0.185
0.190
0.223
11
0.792
0.796
0.809
0.758
0.813
0.815
0.810
0.777
Assumption 8: The mitochondrial and cytosolic oxaloacetate pool are connected by a high
exchange flux
To determine the extent of label exchange between the cytosolic and mitochondrial
oxaloacetate pool, the amino acids originating from cytosolic oxaloacetate (Asp and Thr)
were compared with the amino acids originating from mitochondrial α-ketogluterate (Lys, Pro,
Arg and Glu). The carbon transitions within the TCA‑cycle are such that the second, third
and fourth carbon atom of oxaloacetate map to the third, fourth and fifth carbon atom of α-
101
Chapter 4
102
ketogluterate, respectively. Consequently, the relative intensities of Asp‑αβγ were compared
with those of Glu‑βαC and Pro‑βαC in Table 9‑A. Similar relative intensities were observed,
thereby providing the first evidence that the cytosolic and mitochondrial oxaloacetate pool
were connected via a highly reversible transporter. The large exchange flux for the transporter
also followed from the identical labeling of Asp‑Cαβ and Asp‑αβγ in the penicillin‑G producing
chemostat culture and Thr‑Cαβ and Thr‑αβγ in the non-producing chemostat culture (Table
9‑B). The identical labeling of these amino acid fragments indicates symmetry in the labeling
of the cytosolic oxaloacetate pool. This symmetry is a result of the symmetric precursor
molecules succinate and fumerate. Pines et al. [27] showed that in S. cerevisiae only the
labeling pattern of mitochondrial oxaloacetate is scrambled as a result of symmetric succinate
and fumerate molecules (low reversibility of the cytosolic fumerase). Therefore, only in the
case of a large exchange flux across the mitochondrial membrane, will this scrambling of label
be observable in the cytosolic oxaloacetate pool.
The apparent high equilibration of the cytosolic and mitochondrial oxaloacetate pool also
provided further proof for the activity of malic enzyme. In assumption 6 it was proposed that
the difference in labeling between Tyr/Phe‑Cαβ and Ala‑Cαβ (Table 7‑B) was caused by
malic enzyme, where malic enzyme splits off the fourth carbon atom of mitochondrial malate
to produce pyruvate. Assuming the labeling of mitochondrial malate to be similar to that of
cytosolic oxaloacetate (Asp‑ Cαβ and Thr‑ Cαβ); an influx of malate could effectively explain
the difference in labeling between Ala‑Cαβ and Tyr/Phe‑Cαβ observed in Table 7‑B. The
measured relative intensities of Ala‑Cαβ were best explained by a 1:4 ratio of the relative
intensities of Asp‑Cαβ/Thr‑Cαβ and Tyr/Phe‑Cαβ respectively, indicating that approximately
20% of mitochondrial pyruvate originated from malate.
Table 9-A Summed relative intensities of Asp-αβγ, Glu-βαC and Pro-βαC under both penicillin-G
producing and non-producing conditions. The underlined carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
Non-producing
Fragment
Asp-αβγ
Glu-βαC
Pro-βαC
Asp-αβγ
Glu-βαC
Pro-βαC
010
0.334
0.298
0.304
0.322
0.316
0.317
011
0.260
0.268
0.259
0.217
0.229
0.232
110
0.219
0.217
0.224
0.246
0.245
0.244
111
0.187
0.217
0.213
0.214
0.211
0.207
Table 9-B Relative intensities of Asp-Cαβ and Asp-αβγ for the penicillin-G producing cultivation and ThrCαβ and Thr-αβγ for the non-producing cultivation. The underlined carbon atom refers to the atom for
which the NMR spectrum is recorded.
Producing
Non-producing
Fragment
Asp-Cαβ
Asp-αβγ
Thr-Cαβ
Thr-αβγ
010
0.312
0.334
0.305
0.315
110+011
0.468
0.479
0.462
0.477
111
0.220
0.219
0.234
0.208
Metabolic flux analysis of P. chrysogenum
Assumption 9: The cysteine incorporated into penicillin‑G is synthesized from serine
The labeling of Cys‑Cαβ and Cys‑αβ in the β‑lactam nucleus of penicillin‑G was measured for
the producing chemostat culture and compared to the labeling of Ser‑Cαβ and Ser‑αβ (Table
10‑A&B). It was assumed that cysteine was synthesized from cytosolic serine. Surprisingly,
clear differences were observed between the labeling of the two amino acid fragments,
suggesting a (partial) different origin of cysteine. A plausible explanation was not found. Aside
from cysteine, the β‑lactam nucleus of penicillin‑G is also constructed from valine. In contrast
to the labeling of cysteine, the labeling of Val‑βγ1 and Val‑βγ2 corresponded well with the
labeling in the corresponding proteinogenic amino acid fragments (Table 10‑C&D).
Table 10-A Relative intensities of Cys-Cαβ
and Ser-Cαβ under both penicillin-G producing
and non-producing conditions. The underlined
carbon atom refers to the atom for which the
NMR spectrum is recorded.
Table 10-B Relative intensities of Cys-αβ and
Ser-Cαβ under both penicillin-G producing
and non-producing conditions. The underlined
carbon atom refers to the atom for which the
NMR spectrum is recorded.
Producing
Fragment
Cys-Cαβ
Ser-Cαβ
010
0.205
0.165
011
0.147
0.142
110
0.309
0.271
111
0.339
0.422
Table 10-C Relative intensities of Val-βγ1
measured in penicillin-G (Pen) and as a
proteinogenic amino acid (AA) under penicillinG producing and non-producing conditions. The
underlined carbon atom refers to the atom for
which the NMR spectrum is recorded.
Producing
Fragment
Cys-αβ
Ser-αβ
01
0.747
0.657
11
0.253
0.343
Table 10-D Relative intensities of Val-βγ2
measured in penicillin-G (Pen) and as a
proteinogenic amino acid (AA) under penicillinG producing and non-producing conditions. The
underlined carbon atom refers to the atom for
which the NMR spectrum is recorded.
Producing
Producing
AA: Val-βγ2
Pen: Val-βγ2
Fragment
AA: Val-βγ1
Pen: Val-βγ1
Fragment
01
0.600
0.549
01
0.754
0.765
11
0.400
0.451
11
0.246
0.235
Overview
Based upon the above comparisons an overview was made to illustrate the origin of the
amino acids and penicillin-G synthesized by P. chrysogenum. Figure 4 shows how the carbonatoms of the primary metabolites map into the corresponding amino acids and peniciilin-G.
Furthermore the overview shows which carbon bonds in the precursor are preserved in
synthesis of the amino acid and which are broken. The illustration was adapted from the
overview given by Maaheimo et al. [22] for S. cerevisiae.
Figure 4 shows that several amino acids (e.g. Tyr and Phe) are constructed from multiple
precursors. When joining two precursors, the simultaneous 13C‑labeling of two carbon atoms
between which a new bond is formed, is the result of a random process. Consider, for example,
the biosynthesis of the amino acid Cαβ in which the Cα‑fragment is joined to the β‑fragment:
the chance that both the α‑carbon atom and the β‑carbon atom are 13C‑labeled determines the
103
Chapter 4
Primary metabolites
Cytosol
3-phoshoglycerate phosphoenol-pyruvate erythrose-4-phosphate
2
OH
O
O
3
2
1
O
H H
2
4
3
2
O
1
1
3
H
O
2
1
2
3
acetyl-CoA
2
H
O
4
O3PO 5
H
pyruvate
O
4
3
oxaloacetate
α-ketoglutarate
5
O3PO 4
1
1
1
OH
CO2
O
3
O
OH
O
1
O
O
2
OH OH
O
H
3
OH OH
O
O3PO
OPO3
O
OH
ribose-5-phosphate
1
S
Mitochondrion
Amino Acids
β α
3
104
2
β α
3
1
Ala
γ2
β α
3
γ
1
2
δ2
3
1
δ1
Val
3
δ'
γ
4
2
2
1
ε
3
δ
α
1
α
2
1
γ
2
3
2
1
3
1
Gly
β
α
2
2
δ γ
1
1
2
1
δ γ
β α
1
3
2
4
5
β α
4
3
2
1
β α
3
4
δ
γ1
4
3
5
α
2
γ2
3
β
ε
δ γ
β
α
1
2
4
2
3
2
1
Ile
1
Lys
His
Tyr, Phe
γ
Asp, Met, Thr
Glu, Pro, Arg
Leu
β α
3
α
2
Ser, Cys
3
ε'
ξ
2
2
Penicillin-G
PAA
N
Cys
2
3
1
N
S
3
2
3
2
1
Val
Figure 4 Overview of the precursors used for amino acids and penicillin-G synthesis in P. chrysogenum as
described by van Gulik et al. [34] updated with the additional insight obtained during this study (overview
is adapted from Maaheimo et al. [22]). Shown are the carbon skeletons of the primary metabolites
(precursors) and the resulting amino acids and penicillin-G. The box around the precursor and amino acid
indicates the localization, a dashed line being mitochondrial and a solid line being cytosolic. The carbon
bonds of the amino acids can be both thick or thin lines. A thick carbon bond indicates that the bond is
preserved in the synthesis of the amino acid, while a thin carbon bond indicates that the two connected
carbon atoms are not from the same precursor molecule. The origin of cysteine was uncertain (see main
text), but was assumed to be synthesized from 3-phosphoglycerate in this overview. PAA: phenylacetic
acid.
Metabolic flux analysis of P. chrysogenum
fractional enrichment of the β‑carbon atom. In this case, van Winden et al. [37] showed that
the fractional enrichment (FE) equals:
FE =
f110
f111
=
f110 + f010 f111 + f011
(4.4)
where fx is the relative intensity of the specified fragment x, in which ‘1’ denotes a 13C‑labeled
carbon atoms and ‘0’ denotes an 12C‑labeled carbon atom.
In Table 11 the fractional enrichments are given for the following carbon atoms: Val‑β, Phe‑γ,
Tyr‑γ, Ile‑β, Glu‑β, Leu‑β and Lys‑β. Note that fractional enrichments could only be calculated
for spectra in which the relative intensities of the two doublet fine structures (d and d*) could
be separately determined. Tables 11-A and B show higher fractional enrichments for the
amino acids that were synthesized further downwards in the metabolism. This increase is
a result of the uniformly labeled acetic acid added to the medium. Mitochondrial import and
subsequent incorporation of cytosolic uniformly labeled AcCoA in the TCA‑cycle intermediate
citrate increased the abundance of 13C in the TCA‑cycle intermediates. As a result, the highest
fractional enrichments were calculated for Glu‑β and Lys‑β. Fractional enrichments for Phe‑γ
and Tyr‑γ were the lowest as the labeling of these amino acids was not influenced by the
uniformly labeled acetate. Compared to Phe‑γ and Tyr‑γ, slightly higher fractional enrichments
were observed for the carbon atoms originating from mitochondrial pyruvate (Val‑β, Ile‑β
and Leu‑β). This increase was most probably a result of malic enzyme activity, resulting in
an inflow of 13C‑enriched malate (from the uniformly labeled acetate) in the mitochondrial
pyruvate pool.
Table 11-A Fractional enrichments of the proteinogenic amino acid fragments Phe‑αβγ, Tyr‑αβγ, val-Cαβ,
Ile-Cαβ, Leu-Cαβ, Glu-δγβ and Lys-Cαβ in the penicillin-G producing chemostat cultivation. The underlined
carbon atom refers to the atom for which the NMR spectrum is recorded. Fractional enrichments were
calculated for the most right carbon atom.
Carbon atom
Phe-γ
Tyr-γ
Val-β
Ile-β
Leu-βb
Glu-βa
Lys-βa
011/(011+010)
0.190
0.209
0.241
0.223
na
nd
nd
111/(111+110) 0.177
0.191
0.195
0.210
0.227
nd
nd
a
nd = not determined.
b
na= not accurate. These values were based upon small relative intensities resulting in relative large
contributions of measurement error to the calculated fractional enrichment.
Table 11-B Fractional enrichments of the proteinogenic amino acid fragments Phe-αβγ, Tyr‑αβγ, val-Cαβ,
Ile-Cαβ, Leu-Cαβ, Glu-δγβ and Lys-Cαβ in the non‑producing chemostat cultivation. The underlined
carbon atom refers to the atom for which the NMR spectrum is recorded. Fractional enrichments were
calculated for the most right carbon atom.
Carbon atom
Phe-γ
Tyr-γa
Val-β
Ile-β
Leu-βb
Glu-β
Lys-βb
011/(011+010)
0.198
nd
0.252
0.244
na
0.383
na
111/(111+110) 0.189
nd
0.229
0.242
0.233
0.353
0.381
a
nd = not determined.
b
na = not accurate. These values were based upon small relative intensities resulting in relative large
contributions of measurement error to the calculated fractional enrichment.
105
Chapter 4
106
Enzymatic analysis Analysis of the measured 13C‑labeling data indicated the presence of
both the enzymes threonine aldolase and PEP carboxykinase. To verify these observations,
activities of these enzymes were determined in cell extracts of the producing and nonproducing chemostat cultivations. In addition, the activity of glucose oxidase was measured
in both cell extracts and in culture supernatants. G6P dehydrogenase activity served as a
positive control and was found to be present in all tested cell extracts.
As expected, no measurable enzyme activities were found for glucose oxidase, thereby
supporting the notion that under the applied growth conditions there exists no metabolic
bypass from glucose to 6PG via gluconate. This is in accordance with the findings of Harris et
al. [14], who also found no glucose oxidase activity in P. chrysogenum cell extracts and culture
supernatant grown under glucose‑limiting conditions.
No measurable enzyme activities were observed for threonine aldolase in both the penicillin‑G
producing and non-producing chemostat cultivations, indicating that threonine aldolase
activities were below the detection limit of the applied assay (<0.005 U/mg protein). Cell
extracts from a chemostat cultivation of a Gly1 over expressing mutant of S. cerevisiae served
as a positive control for the assay. Analogous to our observations, threonine aldolase activity
could also not be measured in the empty vector strain of S. cerevisiae (without Gly1) [35].
Nevertheless, several flux analysis papers have illustrated the importance of incorporating
the enzyme in the metabolic network of S. cerevisiae [8, 11]. Furthermore, Christensen et
al. [3] did measure threonine aldolase activity in crude biomass extracts of P. chrysogenum.
Therefore, despite the inability to measure the presence of threonine aldolase it was decided
to include the enzyme in the metabolic network model.
Trace amounts of enzyme activity were measured for PEP carboxykinase in both the
penicillin‑G producing and non-producing chemostat cultures. Measured enzyme activities
were close to the detection limit of the assay (±0.005 U/min/mg protein). The reaction catalyzed
by PEP carboxykinase was included in the metabolic network model used in this study. The
importance of including a reaction converting four‑carbon metabolites from TCA cycle to threecarbon metabolites from the glycolysis in the metabolic network model of P. chrysogenum was
recognized by Christensen et al. [3], but in this study they could not distinguish between the
enzymes oxaloacetate decarboxylase, malic enzyme and PEP carboxykinase. More recently,
van Winden et al. [39] showed that inclusion of the PEP carboxykinase catalyzed reaction
dramatically improved the fit between measured and simulated 13C-labeling data when
performing a MFA of P. chrysogenum.
4.4.3
Accuracy of the measurement error
LC-MS derived mass isotopomer fractions The measurement errors of the LC-MS derived
mass isotopomer fractions were based upon 5 independent sample injections (Table A-I).
Standard deviations for these measurements were on average ±0.01. To test the reproducibility
of the measured mass isotopomer fractions, the variability between multiple samples was
compared with the variability of the 5 independent measurements within one sample using a
one-way Analysis of Variance test (ANOVA). The ANOVA test evaluates the hypothesis that
the samples all have the same mean. Typically, a p-value smaller than 0.05 casts doubt on the
null hypothesis and suggests that at least one sample mean is significantly different from the
Metabolic flux analysis of P. chrysogenum
other sample means. For nearly all mass fractions of the measured compounds it was found
that the variability between samples was not significantly bigger than the variability within a
sample (i.e. between multiple injections). As an example, Table 12 shows the ANOVA’s for
the mass isotopomers of malate. Similar probabilities were found for the other measured
metabolites (results not shown).
As an additional check, metabolite samples were also withdrawn during cultivation on unlabeled
substrate, to determine the natural 13C-labeling of the primary metabolites. Differences
Table 12 One-way ANOVA for the measured mass isotopomer fractions of malate in the penicillin-G
producing and non-producing chemostat cultivation. The variability of the measured mass isotopomer
fractions in multiple samples is compared with the variability of the measured mass isotopomer fractions
in one sample (multiple sample injection). P-values represent the probability that all analyzed metabolite
samples have the same mean. Typically, a p-value smaller than 0.05 casts doubt on the null hypothesis.
Chemostat
cultivation
Mass
fraction
Producing
m+0
m+1
m+2
m+3
m+4
Nonproducing
m+0
m+1
m+2
m+3
m+4
Sum of
Squares
DF
Mean
Square
F-value
P-value
Between samples 1.44E-04
2
7.20E-05
1.689
0.226
4.26E-05
1.864
0.197
0.117
0.891
0.212
0.812
4.214
0.041
3.442
0.101
4.142
0.076
0.135
0.723
0.363
0.564
0.178
0.684
Within samples
5.12E-04
12
Total
6.56E-04
14
Between samples 2.08E-04
2
1.04E-04
5.57E-05
Within samples
6.68E-04
12
Total
8.76E-04
14
Between samples 1.18E-05
2
5.91E-06
Within samples
6.06E-04
12
5.05E-05
Total
6.18E-04
14
Between samples 5.67E-06
2
2.83E-06
Within samples
1.60E-04
12
1.34E-05
Total
1.66E-04
14
Between samples 1.75E-05
2
8.74E-06
Within samples
2.49E-05
12
2.07E-06
Total
4.24E-05
14
Between samples 3.09E-04
1
3.09E-04
Within samples
7.18E-04
8
8.98E-05
Total
1.03E-03
9
Between samples 1.61E-04
1
1.61E-04
Within samples
3.11E-04
8
3.88E-05
Total
4.71E-04
9
Between samples 2.30E-06
1
2.30E-06
Within samples
1.37E-04
8
1.71E-05
Total
1.39E-04
9
Between samples 6.72E-06
1
6.72E-06
Within samples
1.48E-04
8
1.85E-05
Total
1.55E-04
9
Between samples 6.25E-07
1
6.25E-07
Within samples
2.81E-05
8
3.51E-06
Total
2.87E-05
9
107
Chapter 4
between the measured natural labeling and the calculated natural labeling based upon a
natural abundancy of 13C of 1.07%, were on average ±0.01 (results not shown). Based upon
the above results the standard deviations for the measured mass isotopomer fractions were
fixed at 0.01 when weighting the sum of squared residuals ( SSres) between the simulated and
measured mass isotopomer fractions in the metabolic flux estimation. The weighted SSres was
subsequently used as minimization function when iteratively determining the optimal flux set.
108
NMR derived relative intensities The errors in the measurements of the labeling of the
amino acid fragments (Table A-II) were directly computed by the peak-fitting software from
the residual spectrum of the measured and optimally fitted spectra [36]. The advantage of
this method is that the NMR noise in only one spectrum is sufficient to derive the standard
deviations of the relative intensities. The validity of the estimated measurement errors in
this study was examined by calculating the covariance-weighted SSres between the relative
intensities of (i) amino acid fragments analyzed in duplo and (ii) two amino acid fragments
with an assumed common origin. P-values corresponding with the calculated SSres were
determined based upon a chi-squared distribution.
Table 13 shows the probabilities that the relative intensities of the amino acid fragments
analyzed in duplo are identical. Apart from Tyr-δεξ for the non-producing culture and Val-Cαβ
and Tyr-γδε for the producing culture clear differences were observed between the two sets
of relative intensities. In addition, measurement error could seldom explain the differences
in relative intensities of two amino acid fragments with an assumed common origin. As an
example, the similarity of the relative intensities of Ala-αβ, Val-βγ1, Ile-βγ2 and Leu-γδ1 are
presented in Table 14, showing that only the relative intensities of Ala-αβ and Ile-βγ2 were
statistically identical. The inability to explain the observed difference in relative intensities
indicates an underestimation of the measurement error. A similar observation was made
by van Winden et al. [37] when comparing amino acid fragments with identical biosynthetic
pathways in yeast. It was proposed that an additional source of errors might be introduced
when deriving the one-dimensional 13C-spectra used for peak fitting from the two-dimensional
[13C,1H] COSY spectra. By making a cross section at a single 1H-frequency, the relative peak
areas may not be representative for the relative peak volumes in the two-dimensional spectra.
Standard deviations for the measured relative intensities in Table A-II were on average ±0.007.
To compensate for the more than two-fold underestimation of the measurement error, standard
deviations were fixed at 0.02 when weighting the SSres in the metabolic flux estimation.
Table 13 Statistically testing the equality between duplo measurements for the amino acid fragments
Tyr-γδε, Tyr-δεξ, Val-Cαβ, Val-βγ1, Val-βγ2 and His-γδ. Statistical testing was based upon a chi-squared
distribution of the sum of squared residuals (SS) with DF degrees of freedom. P-values represent the
probability that the relative intensities of the amino acid fragment analyzed in duplo are identical. Typically,
a p-value smaller than 0.05 casts doubt on the null hypothesis.
Producing
Tyr-γδε
Non-producing
Tyr-δεξ
Val-Cαβ
Val-βγ1
Val-βγ2
Tyr-γδε
Tyr-δεξ
His-γδ
SS
0.875
7.7936
70.748
3.6081
4.8766
12.877
3.4813
58.761
DF
2
2
3
1
1
2
2
1
0.646
0.020
<0.001
0.057
0.027
0.002
0.175
<0.001
P-value
Metabolic flux analysis of P. chrysogenum
Table 14 Statistically testing the equality between the amino acid fragments originating from the second
and third carbon atoms of pyruvate (Ala-α, Val-βγ1, Ile-βγ2 and Leu-γδ1) assuming a chi-squared
distribution of the sum of squared residuals (SS) with DF degrees of freedom. P-values represent the
probability that the relative intensities of the compared amino acid fragments are identical. Typically, a
p-value smaller than 0.05 casts doubt on the null hypothesis.
Producing
Ala-αβ
Val-βγ1
Ile-βγ2
Val-βγ1
Ile-βγ2
SS
19.5
0.1
DF
1
<0.001
P-value
Non-producing
Leu-γδ1
Val-βγ1
Ile-βγ2
14.3
0.5
27.1
Leu-γδ1
13.4
1
1
1
1
1
0.467
0.719
<0.001
<0.001
<0.001
SS
33.5
2.0
91.4
21.3
DF
1
1
1
1
<0.001
0.154
<0.001
<0.001
P-value
SS
27.8
7.9
DF
1
1
P-value
<0.001
0.005
13
4.4.4
C-labeling based metabolic flux analysis
Estimated Fluxes Figures 5-A and 5-B show the estimated fluxes for the penicillin-G
producing and non-producing chemostat cultivation, respectively. Flux patterns were first
fitted independently for the LC-MS and the NMR dataset, followed by a combined fit for
both datasets. Table A-I and A-II show the simulated mass isotopomer fractions and relative
intensities for the independently fitted datasets, respectively. Note that the flux-estimates for
the LC-MS dataset were limited to the glycolysis and the PPP. Reason for this was that the LCMS derived mass isotopomers provided 13C-labeling information on cell-averaged metabolite
pools, requiring the incorporation of additional parameters in the model to fit the distribution
of metabolites located in multiple compartments (such as the TCA-cycle metabolites). As a
result of these additional parameters, TCA-cycle fluxes could not be accurately estimated and
were thus omitted in Figure 5.
Very different estimates were derived for the primary metabolic fluxes of P. chrysogenum
when fitting the LC-MS, NMR and combined datasets. Most noteworthy differences were the
flux through the oxidative branch of the PPP, the flux through the anaplerotic/gluconeogenic
metabolic cycle catalyzed by pyruvate kinase, pyruvate carboxylase and PEP carboxykinase
and the reversibility of the G6P-isomerase catalyzed reaction.
A possible explanation for the difference in flux estimates might be the used 13C-quantification
technique. While the NMR-derived labeling patterns of proteinogenic amino acids specifically
reflect the flux distribution during the G1-phase of the cell-cycle (protein synthesis), the labeling
patterns of the LC-MS derived primary metabolites reflect the average flux distribution during
a complete cell cycle [2, 37]. However, a recent comparison of the different 13C-quantification
techniques for a glycerol-overproducing S. cerevisiae strain showed that the application of
labeling data of proteinogenic amino acids (NMR and GC-MS) and primary metabolites (LCMS) for 13C-based MFA yielded similar flux patterns [17].
A notable difference between cultivating S. cerevisiae and P. chrysogenum is the much
109
Chapter 4
glc
51
29
45
100
biomass
15
biomass
g6p
34 (>1000)
-15 (50)
56 (0)
-7 (17)
40 (>1000)
-13 (53)
3
biomass
f6p
63
-17 (75)
70
-10 (>1000)
65
-15 (76)
fbp
63 (38)
70 (236)
65 (44)
biomass
1
tp
16
15
110
154
29
78
biomass
5
oaa
23 (>1000)
8 (115)
15
0
oaa
97
93
4
pep
299
138
195
105
101
citr
ser
0
0
106 (>1000)
115 (3)
pyr
s7p
17 (9)
10 (236)
15 (13)
2
biomass
3
biomass
penicillin-G
pyr
-17 (5)
-10 (50)
-15 (>1000)
e4p
9
3
32 (12)
17 (21)
28 (8)
E-C2
E-C3
129 (159)
132 (0)
128 (195)
173
10
71
2
p5p
E-C1
8 (20)
6 (17)
-5 (0)
-4 (0)
gly
4
biomass
biomass
thr
18
-10 (13)
-7 (15)
AcCoA
4 (0)
7 (0)
oaa
31
Ac
20
biomass
penicillin-G
101
95
AcCoA
8
8
citr
8
biomass
Figure 5-A MFA of a high-yielding P. chrysogenum strain grown under penicillin-G producing conditions
at a growth-rate of 0.02 h-1, based upon LC-MS derived mass isotopomer fractions (top values, regular
font), NMR derived relative intensities (middle values, italic) and a combination of the LC-MS and NMR
dataset (bottom values, bold). Values outside parentheses denote the net fluxes, while values inside
parentheses represent the exchange fluxes. Solid arrowheads denote the direction of the net flux. Fluxes
are normalized for the glucose uptake rate. Abbreviations: glc: glucose, G6P: glucose-6-phosphate, F6P:
fructose-6-phosphate, FBP: fructose-1,6-bisphospate, P5P: pentose-5-phopshate, S7P: sedoheptulose7-phosphate, E4P: erythrose-4-phosphate, E-C2: glycolaldehyde moiety covalently bound to the thiamine
pyrophosphate/transketolase complex, E-C3: dihydroxyacetone moiety covalently bound to the enzyme
transaldolase, ser: serine, gly: glycine, thr: threonine, tp: combined pool of triose-phosphates containing
glyceraldehyde-3-phosphate, 2- and 3-phosphoglycerate and 1,3-bisphosphoglycerate, PEP: phosphoenol
pyruvate, PYR: pyruvate, Ac: acetate, AcCoA, acetyl-CoA, OAA: oxaloacetate, CITR: citrate.
Metabolic flux analysis of P. chrysogenum
glc
40
19
23
100
biomass
24
g6p
37 (>1000)
58 (122)
53 (>1000)
4
biomass
f6p
56
63
62
7
28
0
80
78
3
biomass
4
5
biomass
4
0
0
pyr
66 (>1000)
92 (32)
38 (210)
10 (128)
oaa
333
136
182
pyr
90
89
citr
14
80
78
s7p
e4p
ser
5
-13 (0)
-6 (41)
-8 (>1000)
13 (6)
6 (35)
8 (2)
8
8
pep
oaa
E-C2
E-C3
113 (86)
121 (122)
119 (126)
biomass
23 (13)
9 (2)
13 (8)
-13 (91)
-6 (226)
-8 (82)
56 (46)
63 (62)
62 (66)
1 tp
biomass
192
66
85
3
p5p
-10 (33)
-3 (35)
-5 (46)
fbp
224
20
67
biomass
E-C1
4 (15)
4 (13)
0 (0)
0 (0)
gly
111
5
biomass
biomass
thr
22
1 (6)
1 (14)
AcCoA
oaa
31
10 (2)
10 (0)
Ac
biomass
AcCoA
10
10
citr
10
biomass
Figure 5-B MFA of a high-yielding P. chrysogenum strain grown under penicillin-G non-producing
conditions at a growth-rate of 0.02 h-1, based upon LC-MS derived mass isotopomer fractions (top values,
regular font), NMR derived relative intensities (middle values, italic) and a combination of the LC-MS and
NMR dataset (bottom values, bold). Values outside parentheses denote the net fluxes, while values inside
parentheses represent the exchange fluxes. Solid arrowheads denote the direction of the net flux. Fluxes
are normalized for the glucose uptake rate. For abbreviations see Figure 5-A.
Chapter 4
bigger heterogeneity in cell-types for P. chrysogenum. Whereas the unicellular organism
S. cerevisiae reproduces via cell-budding, a complex morphological differentiation takes place
in P. chrysogenum leading from conidia to elongated multi-cellular complexes called hyphae
[12]. The elongation process is highly polarised occurring only on the hyphal tip, resulting in
different cell types from the extending apices to the degenerated distal regions. Apart from
the morphology of a single hyphae, the heterogeneity during cultivation is further increased
by the different structures formed by interacting hyphae. Within submerged cultures hyphae
can form distinct particles (pellets), connected networks of hyphae (aggregates), or occur as
single hyphae (dispersed mycelia) [6]. Due to transport limitations, considerable differences in
metabolic state can occur between hyphae positioned at the centre and the edge of a pellet
or aggregate. The effect of cell differentiation and transport limitation on the metabolic state
of a P. chrysogenum culture is largely unexplored, but a large variation in metabolic state may
well lead to different flux estimates for 13C-quantification techniques that over-represent the
metabolic state of a certain cell type (for example protein-synthesizing cells).
112
Statistical acceptability Table 15 shows the minimized SSres for the LC-MS, NMR and the
combined dataset under penicillin-G producing and non-producing conditions. The SSres is
a measure for the goodness-of-fit and is distributed according to a chi-squared distribution
with n-p degrees of freedom, where n is the number of independent data points (see Table
A-I and A-II) and p is the number of free parameters in the model. Given the probability
P(χ2 (n − p) > SS) it was tested whether the estimated fluxes are statistically acceptable
within the 95% confidence interval. Table 15 clearly shows that all flux fits are rejected on
statistical grounds, indicating that either (i) the assigned measurement error is too small or (ii)
the chosen metabolic network model is erroneous and/or incomplete, resulting in incorrectly
calculated 13C-label distribution within the cell.
Table 15 Minimized variance-weighted sum of squared residuals (SSres) for the independently fitted
LC-MS and NMR datasets and the combined dataset. The SSres is distributed according to a χ2(n-p)
distribution, with n being the number of independent datapoints, p the number of model parameters and
n-p the degrees of freedom. P(χ2(n-p)>SSres) represents the probability that the measured and simulated
13
C-lableing data are statically identical. Typically, a p-value smaller than 0.05 casts doubt on the null
hypothesis.
Dataset
LC-MS
Parameter
SSres
Independent
datapoints (n)
model parameters
(p)
Degrees of
freedom (n-p)
P χ2 (n − p) > SS
NMR
Combined
nonnonproducing
producing
producing
producing
producing
nonproducing
95.9
75.7
251.0
289.8
510.4
608.5
75
75
83
88
158
163
24
24
19
19
24
24
51
51
64
69
134
139
1.4E-04
1.4E-02
5.8E-24
7.6E-29
2.6E-45
5.8E-60
Metabolic flux analysis of P. chrysogenum
Comparison of the calculated p-values for the LC-MS and the NMR dataset shows that
especially the NMR-derived relative intensities are significantly different from the measured
ones. Table A-II shows that the biggest differences between the measured and simulated
relative intensities are observed for the amino acids originating from the lower part of the
metabolism (anaplerotic pathway, TCA-cycle), such as Asp and Thr (originating from cytosolic
oxaloacetate) and Glu, Lys, Pro and Arg (originating from mitochondrial α-ketogluterate).
Interestingly, the simulated singlet fractions of Asp, Thr, Glu, Lys, Pro and Arg are all
underestimated with respect to the other fine structures, thus indicating a too high inflow of
13
C-label in the lower part of the primary metabolism. In accordance herewith, the fit improved
by lowering the influx of [u‑13C2]acetate by 20-40%. However, despite the volatility of acetate, a
significant lower acetate influx seemed implausible given the measured acetate concentration
in the medium and the chemostat. Combining the NMR and LC-MS datasets into one flux fit
resulted in a >50% increase in the SSres with respect to the SSres for the independently fitted
datasets. This shows that the large differences between the separately fitted flux-sets are not
only due to insensitivity of the fluxes towards the 13C-data.
PPP split-ratio The PPP split-ratio represents the flux through the oxidative branch of the
PPP with respect to the uptake rate of glucose. Significantly higher PPP split ratios were
consistently estimated for the penicillin-G producing cultivation in comparison to the nonproducing cultivation: 51% versus 40% for the LC-MS dataset and 29% versus 19% for
the NMR dataset. The only difference between the two chemostats is the addition of PAA.
Since the uncoupling effect of PAA (by dissipating the proton-motive force across the plasma
membrane) was shown to be negligible under the applied PAA concentrations and pH [33]
and since there was no noteworthy catabolism of PAA by the mycelia (see PAA balance in
Table 1-A), the observed difference in flux distribution can be completely attributed to the
increased demand for cytosolic NADPH as a result of the formation of penicillin-G. This is
the first experimental evidence that penicillin-G production is correlated with flux through the
oxidative branch of the PPP as hypothesized earlier by Jorgensen et al., Henriksen et al. and
van Gulik et al. [15, 16, 34].
Note that a quantitative determination of the PPP split-ratio was hampered by the fact that the
biosynthetic efflux from the G6P pool (primarily towards storage carbohydrates) is dependent
on the assumed ratio between extracellular polysaccharides and peptide formation (Table 1B). Based upon the results of van Gulik et al. [34] this ratio was set at 2:1. Variation of the G6P
efflux showed that a lower G6P biosynthetic efflux resulted in an increased PPP split-ratio and
vice-versa (results not shown).
ATP-dissipation Under both producing and non-producing conditions the anaplerotic flux
through PEP carboxykinase in Figure 5 is far in excess of the anabolic requirement for TCA
cycle intermediates, thereby creating a futile cycle catalyzed by the enzyme pyruvate kinase,
pyruvate decarboxylase and PEP carboxykinase. The presence of this futile cycle is a direct
result of the enzyme PEP carboxykinase, which is normally not expressed during growth on
glucose as the sole carbon source, but may be derepressed/activated by the co-substrate
acetate and/or the low residual glucose concentration in the glucose limited chemostat at a
113
Chapter 4
114
growth rate of 0.02 h‑1. The recycling of one PEP molecule via this cycle costs 1 ATP, thereby
resulting in the dissipation of a significant amount of ATP. In both cases the surplus of ATP
consumption corresponds well with the ATP dissipated by the action of the earlier described
futile cycle (PEP carboxykinase, oxaloacetate carboxylase and pyruvate kinase). The action
of this futile cycle was not taken into account by van Gulik et al. [33] when estimating the
cost of ATP for mycelium growth, penicillin production and maintenance and the overall
stoichiometry of oxidative phosphorylation in P. chrysogenum. Van Gulik et al. estimated the
growth-associated maintenance parameter (Kx) to be 0.38 mol ATP/Cmol biomass. Based
upon the flux distribution for the combined dataset (Figure 5) the futile cycle under producing
and non-producing conditions accounts for, respectively, 76% and 54% of this value (0.29
and 0.21 mol ATP/Cmol biomass, respectively). ATP-dissipation via PEP carboxykinase
was recently also found to be active at dilution rates up to 0.30 h-1 in Escherichia coli [25]
and was linked to the so-called catabolic PEP-glyoxylate cycle as identified by Fischer et
al. [10] in slow-growing E. coli chemostat cultures. Fischer et al. hypothesized that the PEPglyoxylate cycle ran parallel with the TCA cycle and provided E. coli with an alternative route
for catabolism without the formation of excess NADPH (via the NADP+ dependent isocitrate
dehydrogenase). In a compartmented organism such as P. chrysogenum this would provide
the cell with a way to prevent the build-up of NADPH in the mitochondria. Consequently, it
was examined whether extension of the metabolic network model with the glyoxylate cycle
improved the fit. No improvement was observed.
Conserved moieties 13C-labeling based MFA does not need to be constrained by the mass
balances for NADPH, NADH and ATP to obtain flux estimates. By using the fluxes estimated
from 13C-based MFA, shown in Figure 5, the consistency of the NADPH and the ATP balance
was checked for both the producing and non-producing cultivation conditions (see Table 16
and 17). These calculations were based on the stoichiometric model for P. chrysogenum
developed by van Gulik et al. [34] and the ATP stoichiometry parameters estimated by van
Gulik et al. [33]. From Table 16 it can be seen that the total rate of NADPH supply calculated
from the fluxes estimated from the combined dataset is lower than the NADPH demand
under both penicillin-G producing and non-producing conditions. The difference between
NADPH supply and NADPH demand was even bigger for the NMR-derived fluxes, while it
was practically negligible for the LC-MS derived fluxes (results not shown). The difference in
NADPH recovery for the different datasets was primarily caused by the estimated flux through
the oxidative branch of the PPP; the main producer of NADPH in the cell. Conservation of
the NADPH moiety in Table 16 required an oxidative PPP flux of 57% and 45% of the total
glucose-uptake rate under penicillin-G producing and non-producing conditions, respectively.
As expected, these values correspond best with the LC-MS derived PPP split-ratios.
Furthermore, the 13C-derived flux estimates for the combined dataset were used to calculate
the total ATP production and consumption in the penicillin-G producing and non-producing
chemostat cultures (Table 17). ATP synthesis was divided into two parts: direct synthesis of
ATP via substrate-level phosphorylation and indirect formation via the oxidation of NADH
and FADH by the electron transfer chain. In both 13C-labeling experiments the amount of ATP
consumed was slightly higher than the amount produced. ATP recovery rates were 102% and
Metabolic flux analysis of P. chrysogenum
106% in the penicillin-G producing and non-producing chemostat cultures, respectively. As
described in the previous paragraph, the ATP consumed by the PEP carboxykinase catalyzed
reaction is included in the growth associated maintenance parameter (Kx) and was thus
omitted from the ATP balance shown in Table 17.
Table 16 Overview of the NADPH producing and non-producing reactions under penicillin-G producing and
non-producing conditions, based upon the 13C-labeling derived flux patterns for the combined dataset.
Supply of NADPH
NADPH
stoichiometry
 mol NADPH 


 mol flux 
Producing
Flux
 mol flux 


 100 mol S 
Non-producing
NADPH
 mol NADPH 


 100 mol S 
Flux
 mol flux 


 100 mol S 
NADPH
 mol flux 


 100 mol S 
Oxidative branch of
the PPP
2
45
90
23
46
Acetaldehyde dha
1
8
8
10
10
Malic enzyme
1
0
0
0
0
Cytosolic isocitrate
dh
1
0
0
0
0
98
Total
56
NADPH consumption
Penicillin-G synthesis
Biomass synthesis
10b
5.9
59
0.8
8
0.25c
248
63
326
83
122
91
Total
dh = dehydrogenase
b
As shown by Kleijn et al. [18]
c
Based on the stoichiometric model developed by van Gulik et al. [34], the cytosolic NADPH demand for
biomass synthesis was calculated to be 0.25 mol NADPH/CmolX.
a
Flux sensitivity analysis The sensitivity of the estimated PEP carboxykinase flux and PPP
split-ratio to measurement error was determined with the method proposed by van Winden et
al. (2005), in which the measured 13C-labeling data was refitted for a range of fixed flux values
for the studied metabolic node (Figure 6). The subsequent fold-change in the SSres was used
as a measure for the flux sensitivity.
Similar sensitivity plots were observed for the producing and non-producing chemostat
cultivation. In general, the fold-changes in the SSres for the two metabolic nodes presented
in Figure 6 are much smaller than those previously observed for a sensitivity analysis of
S. cerevisiae [17]. This lack of sensitivity clearly indicates that given the applied substrate
labeling and metabolic network model the metabolic fluxes can be more accurately determined
in S. cerevisiae than in P. chrysogenum. In part, the lower sensitivity of the P. chrysogenum
fluxes can be ascribed to the relatively high SSres of the P. chrysogenum flux fits. If, for
example, a certain mass fraction or relative intensity has a big contribution to the SSres, but is
highly insensitive to changes in the studied metabolic flux, this will lead to a relatively smaller
increase in the SSres and thus downscale the overall fold change.
115
Chapter 4
Table 17 Overview of ATP consuming and producing reactions under penicillin-G producing and nonproducing conditions, based upon the 13C-labeling derived flux patterns for the combined dataset (Figure
5).
116
Supply of ATP
Substrate-level
phosphorylation
Pyruvate kinase
Phosphoglycerate kinase
Succinyl-CoA synthase
Oxidative phosphorylation
NADHmitb Pyruvate dha
Isocitrate dh
α-ketoglutarate dh
Malate dh
NADHcytc Glyceraldehyde3p dh
Biomass synthesis
Penicillin-G
synthesis
FADHc
Succinate dh
Total
Producing
ATP
stoichiometry
Flux
ATP
 mol ATP  

  mol flux   mol ATP 
 mol flux   100 mol S   100 mol S 


Non-producing
Flux
ATP
 mol flux   mol ATP 

 

 100 mol S   100 mol S 
1
1
1
195
143
93
195
143
93
182
127
78
182
127
78
1.84
1.84
1.84
1.84
95
93
93
93
175
171
171
171
78
78
78
78
144
144
144
144
1.10
143
158
0.04d
248
9
127
326
140
12
4.42e
5.9
26
1.10
93
103
1414
0.8
78
4
86
1203
ATP consumption
Glucokinase
1
100
100
100
100
Phosphofructokinase
1
65
65
56
62
Pyruvate carboxylase
1
78
78
94
85
Acetyl-CoA synthase
1
31
31
30
31
Penicillin-G synthesis
9e
5.9
53
0.8
7
Stoichiometric biomass
f
0.6
149
synthesis
248
326
124
Additional ATP for growth (Kx)
0.38g
248
94
0.8
57
Additional ATP for penicillin
73g
428
(Kp)
5.9
326
196
Non-growth related
g
395
maintenance (ms)
474g
Total
1393
1136
a
dh = dehydrogenase.
b
The P/O ratio for the oxidation of mitochondrial NADH in P. chrysogenum was calculated to be 1.84 by
van Gulik et al. [33].
c
The P/O ratio for the oxidation of cytosolic NADH and FADH was set at 0.6·P/O-ratio [33].
d
Using the stoichiometric model developed by van Gulik et al. [34] the cytosolic NADH production
during biomass synthesis was calculated to be 0.033 mol NADH/CmolX.
e
As shown by van Gulik et al. [34].
f
Using the stoichiometric model developed by van Gulik et al. [34] the ATP demand for biomass
synthesis was calculated to be 0.60 mol ATP/CmolX.
g
As shown by van Gulik et al. [33].
Metabolic flux analysis of P. chrysogenum
Figure 6-A shows that the PPP split ratio is somewhat more sensitive to changes in the LC-MS
measurements than in the NMR measurements. PPP split-ratios below ±0.07 mol·(mol glucose)1
were not feasible due to the minimal flux needed for the synthesis of biosynthetic precursors
(e.g. amino acids). Sensitivity analyses for the gluconeogenic flux from oxaloacetate to PEP
shows that at flux values below the optimal flux this node is best estimated via LC-MS, while
at values above the optimal flux the NMR measurements are most sensitive (Figure 6-B). In
principle, a combination of the two datasets would thus have resulted in the most accurate flux
estimate, were it not for the fact that very different flux estimates were computed for the two
datasets. Interestingly, Figure 6-B shows a big overlap in the range of fluxes at which both the
LC-MS and NMR measurements are sensitive to the PEP carboxykinase catalyzed flux (0-150
mol∙100 mol glucose-1). Despite this similar sensitivity range for both analytical methods, very
different optimal fluxes were estimated for the PEP carboxykinase catalyzed reaction. This
again shows that the two 13C-datasets represent fundamentally different flux-patterns.
A-I
Fold change in SS
LC-MS
NMR
1.8
1.6
1.4
1.2
1.0
0.0
0
2.0
Fold change in SS
2.0
20
60
40
PPP split-ratio
A-II
1.6
1.4
1.2
1.0
0.0
0
50
100
150
200
PEP carboxykinase flux
(mol.[100 mol glucose]-1)
LC-MS
NMR
250
117
1.6
1.4
1.2
1.0
20
0
60
40
PPP split-ratio
80
B-II
2.0
LC-MS
NMR
1.8
B-I
1.8
0.0
80
Fold change in SS
Fold change in SS
2.0
LC-MS
NMR
1.8
1.6
1.4
1.2
1.0
0.0
0
50
100
150
200
250
PEP carboxykinase flux
(mol.[100 mol glucose]-1)
Figure 6 Observed fold increase in the SSres when independently refitting the LC-MS and NMR 13Clabeling data from the penicillin-G producing (A) and non-producing (B) chemostat for fixed values
of the PPP split-ratio (I) and the PEP carboxykinase catalyzed flux (II).
Chapter 4
118
4.5 CONCLUSIONS
Labeling redundancies in the measured LC-MS derived mass isotopomers and the NMR
derived relative intensities were combined with enzymatic analysis to reconstruct and validate
the metabolic network model proposed by van Gulik et al. [34]. As a result: (i) precursor origins
for the amino acids Leu, Pro, Glu, Lys and Arg were altered (ii) the reactions catalyzed by
threonine aldolase, PEP carboxykinase and malic enzyme were included and the reaction
catalyzed by glucose oxidase was excluded from the metabolic network model (iii) a
mitochondrial acetyl-CoA transporter was included and (iv) the transport of oxaloacetate, the
interconversion of G6P/F6P and the interconversion of the C4 carboxylic acids in the TCAcycle were identified as highly equilibrated reactions.
Metabolic fluxes in the penicillin-G producing and non-producing chemostat cultivation of
P. chrysogenum were derived by first independently fitting the LC-MS and NMR datasets,
followed by a flux fit for the combined datasets. Despite the reconstructed metabolic network
model, all six flux fits presented in this study had to be rejected on statistical grounds.
Furthermore, highly different flux patterns were estimated for the two applied 13C-quantification
techniques. The presented flux patterns and ensuing hypotheses should thus be treated with
appropriate caution. Nevertheless, flux analysis of P. chrysogenum indicated the existence of
a considerable ATP-dissipating flux as result of the simultaneous activity of PEP carboxykinase
and pyruvate carboxylase (accounting for 5-7% of the total ATP consumed by the cell) and a
pronounced increase in the PPP split-ratio under penicillin‑G producing conditions (a 12-19%
increase). Considering the increased demand for ATP and NADPH under penicillin producing
conditions, these two metabolic nodes form interesting starting points for future research on
metabolic engineering.
The question remains whether the statistical rejection and the different flux patterns for the
two 13C-quantification techniques is the result of an incomplete/incorrect metabolic network
model or due to the earlier described heterogeneity of a fungal cultivation. Further research is
needed to determine whether the complexity of fungal cultivations, due to cell differentiation
and complex formation, indeed leads to different metabolic flux pattern when applying the
nowadays available different 13C-quantification techniques.
ACKNOWLEDGEMENTS
This work was financially supported by the Dutch EET program (Project No. EETK20002) and
DSM. We gratefully acknowledge Diana Harris for her help with the enzymatic analysis.
Metabolic flux analysis of P. chrysogenum
Appendix a: 13C-Labeling data
Table A-I Measured and simulated steady-state LC-MS derived mass isotopomers of 13C-labeled primary
metabolites in a penicillin-G producing and non-producing chemostat cultivation of P. chrysogenum.
Cells were grown on a mixture of 13C-labeled glucose (0.1 Cmol/L; 20% [U-13C6]glucose and 60% [113
C1]glucose) and 13C-labeled acetate (0.01 Cmol/L; 100% [U-13C2]acetate).
Metabolite
Glucose
in mediuma
Acetate
in mediumb
G6P
6PG
F6P
M6P
FBP
2/3PG
PEP
Producing
Mass
fraction
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+0
m+1
m+2
measured
0.188
0.570
0.037
0.008
0.002
0.019
0.177
0.000
0.025
0.975
0.200
0.413
0.129
0.084
0.058
0.028
0.088
0.205
0.382
0.146
0.081
0.063
0.036
0.088
0.192
0.379
0.130
0.087
0.059
0.032
0.121
0.205
0.393
0.137
0.090
0.063
0.031
0.081
0.223
0.345
0.142
0.128
0.074
0.028
0.060
0.514
0.253
0.081
0.152
0.446
0.285
0.126
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.000
0.001
0.000
0.000
0.001
0.003
0.003
0.000
0.005
0.005
0.012
0.011
0.005
0.003
0.004
0.002
0.004
0.008
0.016
0.016
0.002
0.005
0.004
0.003
0.014
0.033
0.010
0.009
0.011
0.005
0.017
0.006
0.016
0.005
0.004
0.005
0.003
0.004
0.014
0.008
0.004
0.008
0.003
0.002
0.003
0.010
0.006
0.013
0.007
0.009
0.007
0.006
Non-producing
simulated
0.205
0.409
0.134
0.086
0.059
0.030
0.075
0.204
0.408
0.135
0.087
0.059
0.030
0.075
0.205
0.406
0.136
0.088
0.060
0.030
0.073
0.205
0.406
0.136
0.088
0.060
0.030
0.073
0.225
0.348
0.144
0.128
0.068
0.030
0.054
0.511
0.249
0.094
0.143
0.454
0.276
0.130
measured
0.189
0.569
0.037
0.007
0.001
0.017
0.180
0.000
0.025
0.975
0.194
0.428
0.122
0.089
0.057
0.028
0.083
0.201
0.394
0.133
0.090
0.062
0.031
0.089
0.193
0.424
0.132
0.102
0.065
0.027
0.057
0.200
0.406
0.133
0.088
0.065
0.031
0.077
0.226
0.333
0.132
0.132
0.085
0.034
0.057
0.508
0.258
0.077
0.157
0.409
0.317
0.158
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.002
0.002
0.000
0.000
0.000
0.000
0.002
0.000
0.005
0.005
0.012
0.012
0.005
0.004
0.001
0.003
0.004
0.014
0.008
0.001
0.005
0.003
0.001
0.006
0.033
0.037
0.033
0.009
0.011
0.012
0.017
0.016
0.020
0.017
0.011
0.007
0.004
0.004
0.015
0.018
0.010
0.006
0.003
0.003
0.004
0.013
0.010
0.003
0.006
0.009
0.008
0.007
simulated
0.195
0.419
0.128
0.090
0.059
0.025
0.081
0.195
0.418
0.129
0.091
0.060
0.025
0.081
0.195
0.412
0.133
0.094
0.062
0.025
0.076
0.195
0.412
0.133
0.094
0.062
0.025
0.076
0.226
0.339
0.143
0.140
0.071
0.027
0.052
0.514
0.247
0.090
0.145
0.415
0.294
0.161
119
Chapter 4
P5P
S7Pc
AKG
SUC
MAL
120
FUM
m+3
m+0
m+1
m+2
m+3
m+4
m+5
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+0
m+1
m+2
m+3
m+4
m+0
m+1
m+2
m+3
m+4
m+0
m+1
m+2
m+3
m+4
0.143
0.481
0.213
0.091
0.075
0.062
0.078
0.221
0.286
0.192
0.121
0.084
0.067
0.029
0.153
0.257
0.280
0.184
0.093
0.033
0.245
0.288
0.289
0.129
0.048
0.276
0.302
0.240
0.136
0.046
0.268
0.305
0.243
0.136
0.048
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.005
0.017
0.006
0.016
0.005
0.009
0.014
0.012
0.008
0.008
0.005
0.006
0.004
0.003
0.004
0.016
0.012
0.040
0.011
0.002
0.010
0.010
0.009
0.005
0.009
0.006
0.008
0.007
0.003
0.002
0.010
0.008
0.005
0.003
0.003
0.135
0.485
0.203
0.111
0.084
0.049
0.066
0.223
0.292
0.194
0.120
0.079
0.064
0.029
0.148
0.262
0.287
0.196
0.083
0.023
0.239
0.308
0.254
0.147
0.050
0.263
0.303
0.232
0.147
0.054
0.262
0.302
0.232
0.148
0.054
0.116
0.483
0.206
0.130
0.073
0.032
0.076
0.222
0.280
0.193
0.123
0.094
0.061
0.028
0.140
0.240
0.273
0.208
0.101
0.039
0.244
0.286
0.289
0.135
0.045
0.267
0.298
0.239
0.145
0.052
0.263
0.295
0.243
0.141
0.058
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.007
0.047
0.018
0.040
0.014
0.009
0.020
0.015
0.008
0.014
0.007
0.005
0.004
0.002
0.004
0.005
0.008
0.004
0.007
0.002
0.005
0.007
0.005
0.004
0.005
0.010
0.007
0.004
0.004
0.002
0.012
0.006
0.009
0.009
0.004
0.126
0.479
0.199
0.118
0.091
0.042
0.069
0.223
0.285
0.194
0.127
0.079
0.064
0.027
0.135
0.242
0.290
0.208
0.095
0.029
0.247
0.304
0.249
0.148
0.050
0.248
0.304
0.248
0.148
0.050
0.248
0.303
0.248
0.148
0.051
Mass isotopomer distribution of the 13C-labeled glucose added to the feed medium
Mass isotopomer distribution of the [U-13C2]acetate added to the feed medium
c
The m+7 mass fraction of S7P was not measured, as a result the presented values were normalized for
a
b
mass fractions m+0 to m+6.
Metabolic flux analysis of P. chrysogenum
Table A-II Measured and simulated steady-state NMR derived relative intensities for the 13C-labeled
proteinogenic amino acids in a penicillin-G producing and non-producing chemostat cultivation of
P. chrysogenum. Simulated relative intensities correspond with those for the independently fitted NMR
dataset. The following NMR fine structures were measured: singlets (s), doublets (d), triplets (t) and
double doublets (dd), where d* indicates the doublet with the larger one-bond scalar coupling constant.
Amino acid fragments are ordered as 010, 011, 110 and 111, where ‘0’ denotes 12C-labeled and ‘1’ denotes
13
C-labeled. Standard three letter abbreviations are used to denote the amino acids.
amino
acid
fine
structure
Phe-α
s
d
d*
dd
s
d*
d
dd
s
d
s
d
d*
dd
s
d*
d
dd
s
d
s
d
d*
dd
s
d
s
d
d*
dd
s
d*
d
dd
s
d
t
s
d
t
s
d
d*
dd
s
d
s
d
d*
dd
s
d*
Phe-β
Gly-α
His-α
His-β
His-δ
Ser-α
Ser-β
Tyr-α
Tyr-β
Tyr-δ
Tyr-ε
Ala-α
Ala-β
Asp-α
Asp-β
Producing
measured
0.126
0.155
0.055
0.665
0.459
0.108
0.357
0.077
0.458
0.542
0.059
0.033
0.102
0.806
0.153
0.035
0.320
0.491
0.560
0.440
0.165
0.142
0.271
0.422
0.657
0.343
0.120
0.190
0.055
0.635
0.450
0.119
0.349
0.083
0.448
0.491
0.061
0.167
0.305
0.528
0.142
0.181
0.100
0.577
0.560
0.440
0.312
0.217
0.251
0.220
0.334
0.260
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.004
0.005
0.005
0.007
0.011
0.012
0.013
0.015
0.007
0.007
0.002
0.004
0.003
0.005
0.005
0.007
0.009
0.006
0.009
0.009
0.007
0.007
0.009
0.009
0.018
0.018
0.006
0.009
0.008
0.011
0.007
0.009
0.009
0.011
0.009
0.005
0.012
0.005
0.004
0.007
0.004
0.006
0.006
0.007
0.006
0.006
0.006
0.007
0.006
0.009
0.007
0.008
Non-producing
simulated
0.107
0.168
0.070
0.655
0.435
0.116
0.355
0.095
0.457
0.543
0.064
0.026
0.093
0.817
0.146
0.042
0.309
0.503
0.589
0.411
0.213
0.171
0.232
0.384
0.622
0.378
0.107
0.168
0.070
0.655
0.435
0.116
0.355
0.095
0.392
0.502
0.106
0.201
0.314
0.485
0.134
0.175
0.110
0.581
0.552
0.448
0.260
0.206
0.295
0.238
0.261
0.303
measured
simulated
0.098
0.179
0.046
0.677
0.465
0.115
0.340
0.080
0.309
0.691
0.079
0.039
0.111
0.771
0.139
0.041
0.343
0.477
0.548
0.453
0.149
0.132
0.301
0.419
0.663
0.337
0.093
0.185
0.045
0.677
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.003
0.005
0.004
0.006
0.005
0.005
0.005
0.006
0.003
0.003
0.001
0.002
0.002
0.003
0.002
0.002
0.003
0.003
0.004
0.004
0.002
0.003
0.003
0.003
0.012
0.012
0.005
0.010
0.008
0.012
0.095
0.159
0.065
0.681
0.448
0.122
0.338
0.092
0.384
0.616
0.061
0.029
0.093
0.817
0.119
0.031
0.362
0.487
0.555
0.445
0.181
0.158
0.270
0.390
0.664
0.336
0.095
0.159
0.065
0.681
0.454
0.478
0.069
0.172
0.300
0.528
0.141
0.191
0.082
0.586
0.601
0.399
±
±
±
±
±
±
±
±
±
±
±
±
0.011
0.007
0.014
0.006
0.006
0.009
0.002
0.003
0.002
0.003
0.008
0.008
0.415
0.486
0.099
0.194
0.315
0.491
0.144
0.176
0.118
0.563
0.564
0.436
0.322
0.217
±
±
0.003
0.003
0.299
0.250
121
Chapter 4
Glu-α
Glu-β
Glu-γ
Ile-α
Ile-γ1
Ile-γ2
Ile-δ
Lys-α
122
Lys-β
Lys-γ
Lys-δ
Lys-ε
Leu-α
Leu-β
Leu-δ1
Leu-δ2
Pro-α
Pro-β
Pro-γ
Pro-δ
Arg-β
d
dd
s
d
d*
dd
s
d
t
s
d*
d
dd
s
d
d*
dd
s
d
t
s
d
s
d
s
d
d*
dd
s
d
t
s
d
t
s
d
t
s
d
s
d
d*
dd
s
d
t
s
d
s
d
s
d
d*
dd
s
d
t
s
d
t
s
d
s
d
t
0.219
0.187
0.298
0.217
0.268
0.217
0.380
0.477
0.144
±
±
±
±
±
±
±
±
±
0.008
0.010
0.003
0.004
0.003
0.005
0.008
0.006
0.010
0.203
0.233
0.262
0.203
0.303
0.233
0.331
0.489
0.179
0.414
0.119
0.370
0.098
0.399
0.500
0.101
0.558
0.442
0.430
0.571
±
±
±
±
±
±
±
±
±
±
±
0.004
0.004
0.004
0.006
0.006
0.004
0.008
0.004
0.004
0.003
0.003
0.359
0.108
0.411
0.123
0.434
0.465
0.100
0.552
0.448
0.381
0.619
0.158
0.541
0.302
0.350
0.492
0.158
0.326
0.512
0.162
0.192
0.809
0.026
0.012
0.744
0.218
0.229
0.642
0.129
0.588
0.412
0.789
0.211
0.304
0.224
0.259
0.213
0.374
0.466
0.160
0.310
0.510
0.179
0.204
0.796
0.367
0.483
0.150
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.004
0.004
0.006
0.009
0.006
0.012
0.006
0.005
0.009
0.002
0.002
0.002
0.004
0.004
0.004
0.003
0.003
0.005
0.005
0.005
0.003
0.003
0.004
0.004
0.004
0.006
0.006
0.004
0.007
0.005
0.004
0.007
0.004
0.004
0.007
0.004
0.008
0.110
0.540
0.350
0.331
0.489
0.179
0.306
0.497
0.197
0.225
0.775
0.046
0.014
0.723
0.217
0.150
0.664
0.185
0.552
0.448
0.769
0.231
0.262
0.203
0.303
0.233
0.331
0.489
0.179
0.306
0.497
0.197
0.225
0.775
0.331
0.489
0.179
0.246
0.214
0.316
0.245
0.229
0.211
0.334
0.479
0.188
0.301
0.330
0.164
0.205
0.418
0.135
0.340
0.108
0.424
0.489
0.087
0.555
0.446
0.439
0.561
0.081
0.072
0.524
0.322
0.161
0.529
0.310
0.333
0.487
0.181
0.311
0.520
0.169
0.190
0.810
0.090
0.032
0.674
0.205
0.166
0.660
0.174
0.569
0.432
0.779
0.221
0.317
0.244
0.232
0.207
0.295
0.502
0.203
0.287
0.501
0.211
0.185
0.815
0.299
0.488
0.213
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.003
0.004
0.007
0.010
0.007
0.013
0.004
0.005
0.006
0.002
0.002
0.002
0.003
0.002
0.003
0.002
0.003
0.004
0.002
0.004
0.002
0.002
0.003
0.003
0.006
0.012
0.011
0.012
0.007
0.005
0.008
0.008
0.005
0.011
0.009
0.018
0.019
0.005
0.005
0.001
0.002
0.002
0.003
0.006
0.005
0.008
0.004
0.004
0.003
0.003
0.001
0.002
0.002
0.003
0.009
0.008
0.012
0.006
0.004
0.007
0.004
0.004
0.012
0.009
0.017
0.237
0.214
0.278
0.224
0.256
0.243
0.281
0.498
0.221
0.269
0.338
0.175
0.219
0.348
0.113
0.407
0.132
0.414
0.475
0.110
0.564
0.436
0.410
0.590
0.033
0.022
0.570
0.375
0.088
0.523
0.389
0.281
0.498
0.221
0.269
0.512
0.219
0.193
0.807
0.042
0.014
0.714
0.231
0.125
0.671
0.204
0.564
0.436
0.755
0.245
0.278
0.224
0.256
0.243
0.281
0.498
0.221
0.269
0.512
0.219
0.193
0.807
0.281
0.498
0.221
Metabolic flux analysis of P. chrysogenum
Arg-γ
s
0.304
±
0.010
0.306
0.318
±
0.005
0.269
d
0.514
±
0.006
0.497
0.508
±
0.003
0.512
t
0.183
±
0.013
0.197
0.174
±
0.006
0.219
Arg-δ
s
0.208
±
0.003
0.225
0.187
±
0.002
0.193
d
0.792
±
0.003
0.775
0.813
±
0.002
0.807
Thr-α
s
0.305
±
0.001
0.248
d
0.219
±
0.002
0.214
d*
0.243
±
0.001
0.232
dd
0.234
±
0.002
0.306
Thr-β
s
0.292
±
0.005
0.261
0.315
±
0.004
0.299
d
0.483
±
0.004
0.506
0.477
±
0.002
0.487
t
0.226
±
0.007
0.233
0.208
±
0.005
0.214
Thr-γ
s
0.417
±
0.004
0.381
0.421
±
0.002
0.410
d
0.583
±
0.004
0.619
0.580
±
0.002
0.590
val-α
s
0.221
±
0.002
0.241
d
0.075
±
0.003
0.078
d*
0.543
±
0.002
0.514
dd
0.161
±
0.004
0.166
Val-γ1
s
0.600
±
0.006
0.552
0.595
±
0.004
0.564
d
0.400
±
0.006
0.448
0.405
±
0.004
0.436
Val-γ2
s
0.754
±
0.007
0.769
0.752
±
0.005
0.755
d
0.246
±
0.007
0.231
0.248
±
0.005
0.245
cys-α a
s
0.205
±
0.007
0.213
d
0.147
±
0.014
0.171
d*
0.309
±
0.012
0.232
dd
0.339
±
0.010
0.384
cys-β a
s
0.747
±
0.009
0.622
d
0.253
±
0.009
0.378
val-α a
s
0.265
±
0.002
0.238
d
0.084
±
0.004
0.072
d*
0.525
±
0.003
0.532
dd
0.127
±
0.005
0.159
Val-γ1 a
s
0.549
±
0.006
0.552
d
0.451
±
0.006
0.448
Val-γ2 a
s
0.765
±
0.005
0.769
d
0.235
±
0.005
0.231
a
Relative intensities were derived from the penicillin-G in the filtrate. The β-lactam nucleus of penicillin-G
is constructed from cysteine and valine.
123
Chapter 4
References
[1] Bonarius, H.P.J., Schmid, G. and Tramper, J.
(1997) Flux analysis of underdetermined metabolic
networks: the quest for the missing constraints.
Trends Biotechnol. 15, 308-314.
[2] Christensen, B., Gombert, A.K. and Nielsen, J.
(2002) Analysis of flux estimates based on 13Clabeling experiments. Eur. J. Biochem. 269, 27952800.
[3] Christensen, B. and Nielsen, J. (2000) Metabolic
network analysis of Penicillium chrysogenum using
13
C-labeled glucose. Biotechnol. Bioeng. 68, 652659.
[4] Christensen, B., Thykaer, J. and Nielsen, J. (2000)
Metabolic characterization of high- and low-yielding
strains of Penicillium chrysogenum. Appl. Microb.
Biot. 54, 212-217.
124
[5] Christensen, L.H., Mandrup, G., Nielsen, J. and
Villadsen, J. (1994) A Robust liquid-chromatographic
method for measurement of medium components
during penicillin fermentations. Anal. Chim. Acta
296, 51-62.
[6] Cox, P.W., Paul, G.C. and Thomas, C.R. (1998)
Image analysis of the morphology of filamentous
micro-organisms. Microbiol. 144, 817-827.
[7] de Jong-Gubbels P., Vanrolleghem P., Heijnen, J.J.,
van Dijken, J.P. and Pronk, J.T. (1995) Regulation
of carbon metabolism in chemostat cultures of
Saccharomyces cerevisiae grown on mixtures of
glucose and ethanol. Yeast 11, 407-418.
[8] dos Santos, M.M., Gombert, A.K., Christensen,
B., Olsson, L. and Nielsen, J. (2003) Identification
of in vivo enzyme activities in the cometabolism of
glucose and acetate by Saccharomyces cerevisiae
by using 13C-labeled substrates. Eukaryot. Cell 2,
599-608.
[9] Fierro, F., Garcia-Estrada, C., Castillo, N.I.,
Rodriguez, R., Velasco-Conde, T. and Martin, J.F.
(2006) Transcriptional and bioinformatic analysis of
the 56.8 kb DNA region amplified in tandem repeats
containing the penicillin gene cluster in Penicillium
chrysogenum. Fungal Genet. Biol. 43, 618-629.
[10] Fischer, E. and Sauer, U. (2003) A novel metabolic
cycle catalyzes glucose oxidation and anaplerosis in
hungry Escherichia coli. J. Biol. Chem. 278, 4644646451.
[11] Gombert, A.K., dos Santos, M.M., Christensen,
B. and Nielsen, J. (2001) Network identification
and flux quantification in the central metabolism of
Saccharomyces cerevisiae under different conditions
of glucose repression. J. Bacteriol. 183, 1441-1451.
[12] Grimm, L.H., Kelly, S., Krull, R. and Hempel, D.C.
(2005) Morphology and productivity of filamentous
fungi. Appl. Microb. Biot. 69, 375-384.
[13] Gunnarsson, N., Eliasson, A. and Nielsen, J. (2004)
Control of fluxes towards antibiotics and the role
of primary metabolism in production of antibiotics.
In: Molecular biotechnology of fungal beta-lactam
antibiotics and related peptide synthetases
(Brakhage, A.A., Ed.), Vol. 88, pp. 137-178. Springer
Berlin, Heidelberg.
[14] Harris, D.M., Diderich, J.A., van der Krogt, Z.A.,
Luttik, M.A.H., Raamsdonk, L.M., Bovenberg,
R.A.L., van Gulik, W.M., van Dijken, J.P. and
Pronk, J.T. (2006) Enzymic analysis of NADPH
metabolism in [beta]-lactam-producing Penicillium
chrysogenum: Presence of a mitochondrial NADPH
dehydrogenase. Metab. Eng. 8, 91-101.
[15] Henriksen, C.M., Christensen, L.H., Nielsen, J.
and Villadsen, J. (1996) Growth energetics and
metabolic fluxes in continuous cultures of Penicillium
chrysogenum. J. Biotechnol. 45, 149-164.
[16] Jorgensen, H., Nielsen, J., Villadsen, J. and
Mollgaard, H. (1995) Metabolic flux distributions
in Penicillium chrysogenum during fed batch
cultivations. Biotechnol. Bioeng. 46, 117-131.
[17] Kleijn, R.J., Geertman J.M.A., Nfor B.K., Ras, C.,
Schipper, D., Pronk, J.T., Heijnen, J.J., van Maris,
A.J.A. and van Winden, W.A. (2006) Metabolic flux
analysis of a glycerol-overproducing Saccharomyces
cerevisiae strain based on GC-MS, LC-MS and
NMR derived 13C-labeling data . FEMS Yeast Res.
In press.
[18] Kleijn, R.J., Liu F., van Winden, W.A., van Gulik,
W.M., Ras, C. and Heijnen, J.J. (2006) Cytosolic
NADPH metabolism in penicillin-G producing and
nonproducing chemostat cultures of Penicillium
chrysogenum. Metab. Eng. In press.
[19] Kleijn, R.J., van Winden, W.A., Ras, C., van Gulik,
W.M., Schipper, D. and Heijnen, J.J. (2006) 13CLabeled gluconate tracing as a direct and accurate
method for determining the pentose phosphate
pathway split ratio in Penicillium chrysogenum. Appl.
Metabolic flux analysis of P. chrysogenum
Environ. Microb. 72, 4743-4754.
[20] Kleijn, R.J., van Winden, W.A., van Gulik, W.M.
and Heijnen, J.J. (2005) Revisiting the 13C-label
distribution of the non-oxidative branch of the
pentose phosphate pathway based upon kinetic and
genetic evidence. FEBS J. 272, 4970-4982.
[21] Lange, H.C., Eman, M., van Zuijlen, G., Visser,
D., van Dam, J.C., Frank, J., de Mattos, M.J. and
Heijnen, J.J. (2001) Improved rapid sampling
for in vivo kinetics of intracellular metabolites in
Saccharomyces cerevisiae. Biotechnol. Bioeng. 75,
406-415.
[22] Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E.,
Sauer, U. and Szyperski, T. (2001) Central carbon
metabolism of Saccharomyces cerevisiae explored
by biosynthetic fractional 13C-labeling of common
amino acids. Eur. J. Biochem. 268, 2464-2479.
[23] Martin, J.F. (1998) New aspects of genes and
enzymes for beta-lactam antibiotic biosynthesis.
Appl. Microb. Biot. 50, 1-15.
[24] Mashego, M.R., van Gulik, W.M., Vinke, J.L. and
Heijnen, J.J. (2003) Critical evaluation of sampling
techniques for residual glucose determination in
carbon-limited chemostat culture of Saccharomyces
cerevisiae. Biotechnol. Bioeng. 83, 395-399.
[25] Nanchen, A., Schicker, A. and Sauer, U. (2006)
Nonlinear dependency of intracellular fluxes on
growth rate in miniaturized continuous cultures of
Escherichia coli. Appl. Environ. Microb. 72, 11641172.
[26] Nasution, U., van Gulik, W.M., Kleijn, R.J., van
Winden, W.A., Proell, A. and Heijnen, J.J. (2006)
Measurement of intracellular metabolites of primary
metabolism and adenine nucleotides in chemostat
cultivated Penicillium chrysogenum. Biotechnol.
Bioeng. 94, 159-166.
[27] Pines, O., EvenRam, S., Elnathan, N., Battat, E.,
Aharonov, O., Gibson, D. and Goldberg, I. (1996)
The cytosolic pathway of L-malic acid synthesis in
Saccharomyces cerevisiae: The role of fumarase.
Appl. Microb. Biot. 46, 393-399.
[28] Piper, M.D., Hong, S.P., Ball, G.E. and Dawes, I.W.
(2000) Regulation of the balance of one-carbon
metabolism in Saccharomyces cerevisiae. J. Biol.
Chem. 275, 30987-30995.
[29] Stryer, L. (1995) Biochemistry. W.H. Freeman and
Company, New York.
[30] Theilgaard, H.A., van den Berg, M.A., Mulder,
C.A., Bovenberg, R.A.L. and Nielsen, J. (2001)
Quantitative analysis of Penicillium chrysogenum
Wis54-1255 transformants overexpressing the
penicillin biosynthetic genes. Biotechnol. Bioeng.
72, 379-388.
[31] Thykaer, J. and Nielsen, J. (2003) Metabolic
engineering of beta-lactam production. Metab. Eng.
5, 56-69.
[32] van Can, H.J.L., Braake, H.A.B.T., Bijman, A.,
Hellinga, C., Luyben, K.C.A.M. and Heijnen, J.J.
(1999) An efficient model development strategy
for bioprocesses based on neural networks in
macroscopic balances: Part II. Biotechnol. Bioeng.
62, 666-680.
[33] van Gulik, W.M., Antoniewicz, M.R., deLaat, W.T.,
Vinke, J.L. and Heijnen, J.J. (2001) Energetics of
growth and penicillin production in a high-producing
strain of Penicillium chrysogenum. Biotechnol.
Bioeng. 72, 185-93.
[34] van Gulik, W.M., de Laat, W.T., Vinke, J.L. and
Heijnen, J.J. (2000) Application of metabolic
flux analysis for the identification of metabolic
bottlenecks in the biosynthesis of penicillin-G.
Biotechnol. Bioeng. 68, 602-618.
[35] van Maris, A.J.A., Luttik, M.A.H., Winkler, A.A., van
Dijken, J.P. and Pronk, J.T. (2003) Overproduction
of threonine aldolase circumvents the biosynthetic
role of pyruvate decarboxylase in glucose-limited
chemostat cultures of Saccharomyces cerevisiae.
Appl. Environ. Microb. 69, 2094-2099.
[36] van Winden, W., Schipper, D., Verheijen, P. and
Heijnen, J. (2001) Innovations in generation and
analysis of 2D [(13C,1H] COSY NMR spectra for
metabolic flux analysis purposes. Metab. Eng. 3,
322-343.
[37] van Winden, W.A. (2003) 13C-Labeling technique
for metabolic network and flux analysis: Theory and
application. Thesis. Delft University of Technology,
the Netherlands.
[38] van Winden, W.A., Heijnen, J.J., Verheijen, P.J. and
Grievink, J. (2001) A priori analysis of metabolic
flux identifiability from 13C-labeling data. Biotechnol.
Bioeng. 74, 505-516.
[39] van Winden, W.A., van Gulik, W.M., Schipper,
D., Verheijen, P.J.T., Krabben, P., Vinke, J.L. and
Heijnen, J.J. (2003) Metabolic flux and metabolic
network analysis of Penicillium chrysogenum
using 2D [13C, 1H] COSYNMR measurements
125
Chapter 4
and cumulative Bondomer simulation. Biotechnol.
Bioeng. 83, 75-92.
[40] van Winden, W.A., van Dam, J.C., Ras, C., Kleijn,
R.J., Vinke, J.L., van Gulik, W.M. and Heijnen, J.J.
(2005) Metabolic flux analysis of Saccharomyces
cerevisiae CEN.PK113-7D based on mass
isotopomer measurements of 13C-labeled primary
metabolites. FEMS Yeast Res. 5, 559-568.
[41] Vanderheijden, R.T.J.M., Heijnen, J.J., Hellinga,
C., Romein, B. and Luyben, K.C.A.M. (1994)
Linear constraint relations in biochemical reaction
systems .1. Classification of the calculability and
the balanceability of conversion rates. Biotechnol.
Bioeng. 43, 3-10.
[42] Wiechert, W. (2001) 13C metabolic flux analysis.
Metab. Eng. 3, 195-206.
126
[43] Wiechert, W., Mollney, M., Isermann, N., Wurzel,
M. and de Graaf, A.A. (1999) Bidirectional reaction
steps in metabolic networks: III. Explicit solution and
analysis of isotopomer labeling systems. Biotechnol.
Bioeng. 66, 69-85.
Metabolic flux analysis of P. chrysogenum
127
CHAPTER
13
5
C-labeled gluconate tracing as a
direct and accurate method for
determining the pentose
phosphate pathway split-ratio
in Penicillium chrysogenum
Roelco J. Kleijn, Wouter A. van Winden, Cor Ras,
Walter M. van Gulik, Dick Schipper and
Joseph J. Heijnen
Based upon:
Applied and Environmental Microbiology
(2006) 72 (7): 4743-4754
Chapter 5
ABSTRACT
In this study a new method is introduced for accurately determining the pentose phosphate
pathway (PPP) split-ratio, an important metabolic parameter in the primary metabolism of
a cell. The method is based upon the simultaneous feeding of unlabeled glucose and trace
amounts of [U-13C]gluconate, followed by measurement of the mass isotopomers of the
intracellular metabolites surrounding the 6-phospho-gluconate (6PG) node. The gluconatetracer method was applied to a penicillin-G producing chemostat-culture of the filamentous
fungus Penicillium chrysogenum. As a comparison, 13C-labeling based metabolic flux analysis
(MFA) was performed for the glycolysis and the PPP of P. chrysogenum. For the first time mass
isotopomer measurements of 13C-labeled primary metabolites are reported for P. chrysogenum
and used for 13C-based MFA.
Estimation of the PPP split-ratio of P. chrysogenum at a growth-rate of 0.02 h-1 yielded
comparable values for the gluconate-tracer and the 13C-based MFA method, namely 51.8%
and 51.1%, respectively. A sensitivity analysis of the estimated PPP split-ratios showed that
the 95% confidence interval was almost three times as small for the gluconate-tracer method
compared to the 13C-based MFA method: [40.0-63.5] and [46.0-56.5], respectively. From these
results it was concluded that the gluconate-tracer method permits an accurate determination
of the PPP split-ratio, but yields no information on the remaining cellular metabolism, while the
13
C-based MFA method allows for the estimation of multiple fluxes, but at the same time yields
a less accurate estimation of the PPP split-ratio.
130
13
C-Labeled gluconate tracing: method
5.1 INTRODUCTION
The quantification of primary metabolic fluxes within microorganisms provides researchers with
an important tool for a more rational approach to metabolic engineering. A part of the cellular
metabolism that has received special attention when exploring possible strain improvement
strategies is the flux distribution around branch-points [9, 30, 31]. At these metabolic branch
points an entering flux diverges into two or more different directions, thus forming a potential
target for rerouting fluxes. Under glucose-feeding conditions the first metabolic node
encountered by the glucose entering a cell is situated around glucose-6-phosphate (G6P),
resulting in a carbon flux partitioning towards glycolysis, the pentose phosphate pathway
(PPP), storage carbohydrates and in some prokaryotes the Entner-Doudoroff pathway. Each
of these pathways has its own unique function in the cell. The glycolysis (combined with the
TCA-cycle) is the general route for glucose catabolism and energy formation in the cell, while
the PPP plays a crucial role in the redox-metabolism of the cell. In addition to their catabolic
function these pathways also have an anabolic function as they provide the precursors for the
monomers (amino acids, fatty acids, nucleotides, sugar phosphates, etc.) required for growth
and product formation. Therefore, an accurate determination of the flux distribution around the
G6P-node provides valuable insight into the functioning of a cell.
An important flux-ratio of the G6P-node is the fraction of G6P entering the oxidative branch
of the PPP in relation to the total uptake of glucose by the cell (from hereon referred to as the
PPP split-ratio). From an industrial point of view, split-ratio determinations are interesting in
the quest for strains that overproduce (partly) PPP-originating products such as phenylalanine
and riboflavin. Recent publications have also revealed the importance of the oxidative branch
of the PPP in regulating the cytosolic NADPH-levels when overproducing specific amino acids
[20] and antibiotics [34].
One of the earliest documented quantifications of the PPP split-ratio dates from 1955, when
Katz et al. [18] developed a method to calculate the flux into the oxidative branch of the
PPP based upon the evolution of 14CO2 in rat liver cells that were consecutively fed with
[U‑14C]glucose, [6‑14C]glucose and [1‑14C]glucose. In the ensuing decades the flux of the
oxidative branch of the PPP has usually been estimated by feeding cells with specifically,
13
C- or 14C-labeled glucose, followed by analysis of the 13C or 14C-distribution in the produced
CO2 or derivatives of the produced triose-phosphates (e.g. lactate, glycerol, amino acids) [8,
21, 25]. Some of these methods are rather laborious, requiring parallel experiments with two
differently labeled substrates, while others can be quite costly due to the applied specifically
labeled substrates. The major problem with these methods is, however, that they are based
upon assumptions with respect to the recycling of fructose-6-phosphate produced in the PPP;
the reversibility [12] and structure [22] of the reactions in the non-oxidative branch of the PPP;
the degree of isotopic equilibrium of the hexose-mono-phosphate pool; and the drain on PPP
metabolites as precursors for the synthesis of biomass [19]. These various assumptions lead
to widely differing determinations of the PPP split-ratio.
Recently, a more robust method has been developed by Christensen et al. [3] in which an
overall 13C balance is set up over the upper part of the glycolysis and the PPP. This method
is capable of estimating the PPP flux irrespective of isotope redistributions arising from F6P
recycling and PPP reversibilities. However, this method is also based upon assumptions that
131
Chapter 5
132
limit the validity of the estimated PPP split-ratios. It is assumed that the C-5 of G6P is naturally
labeled when cells are fed with [1-13C]glucose; that the label pattern of glyceraldehyde-3phopshate (G3P) is identical to that of serine; and that the labeling of C-1 and C-2 of serine
are identical, thereby neglecting the possible effect of the enzymes threonine aldolase and
glycine decarboxylase on the label distribution of serine.
Apart from dedicated 13C tracer experiments aimed at determining the metabolism around
one specific branch point, 13C tracer experiments are increasingly used for simultaneously
determining the flux distributions of multiple convergent metabolic nodes. This method, often
referred to as the 13C-based metabolic flux analysis (MFA), combines 13C-labeling data of
primary metabolites with uptake and secretion rates and biomass composition to determine
the fluxes throughout a metabolic network. The labeling of the primary metabolites can be
either directly measured using LC-MS [37] or indirectly by measuring the 13C-labeling of
proteinogenic amino acids using GC-MS [11, 13] and/or NMR [26]. Up to now, however, it has
proven to be difficult to accurately determine the PPP split-ratio using this method. Van Winden
et al., for example, showed that the PPP split-ratio estimations in Penicillium chrysogenum are
highly dependent on the chosen network stoichiometry [35], while a large confidence interval
was found for the PPP split-ratio of Saccharomyces cerevisiae and Bacillus subtilus [6, 37].
In this study a new method for determining the PPP split-ratio is introduced based upon
the simultaneous feeding of unlabeled glucose and trace amounts of [U‑13C]gluconate. By
measuring the mass isotopomers of the intracellular metabolites around the G6P-node
the flux into the oxidative branch of the PPP can be directly calculated. The applicability
of the gluconate-tracer method is studied in a penicillin-G producing chemostat-culture of
the filamentous fungus P. chrysogenum. The sensitivity and accuracy of the estimated
PPP split-ratio is compared to the PPP split-ratio estimated from a 13C-labeling based MFA
of P. chrysogenum grown under identical conditions. For the first time mass isotopomer
measurements of 13C-labeled primary metabolites are reported for P. chrysogenum and used
for 13C-based MFA.
5.2 THEORY
The first step in the PPP is the irreversible dehydrogenation of G6P to 6‑phosphoglucono‑
δ ‑lactone catalyzed by G6P-dehydrogenase, followed by a hydrolysis step to 6-phosphogluconate (6PG). 6-phospho-gluconate is subsequently catabolized via the Entner-Douderoff
pathway or the PPP. Under most growth conditions the irreversibility of these two pathways
ensures that 6PG can only originate from G6P, resulting in an identical labeling pattern for
both metabolites. An additional inflow into the 6PG-pool can be obtained by feeding gluconate
to the cells. After its uptake, gluconate is directly phosphorylated to 6PG by the enzyme
gluconokinase, thereby entering the PPP directly after the G6P branch point. The co-feeding
of [U‑13C]gluconate to a culture grown on naturally labeled glucose will result in a different
labeling pattern for G6P and 6PG. This difference in labeling pattern is directly related to the
amount of G6P that enters the oxidative branch of the PPP.
Figure 1 schematically shows the primary metabolism of a cell grown simultaneously on
glucose (glc) and gluconate (gln). The flux into the oxidative branch of the PPP (v3) can be
13
C-Labeled gluconate tracing: method
calculated by setting up a metabolite balancev and a mass isotopomer balance over the
metabolite 6PG.
v2 + v3 = v 4 + v5
(5.1)
m + 0
m + 0
m + 0






m + 1
m + 1
m + 1
v 2 
+ v 3 
= (v 4 + v 5 )
  
  
  






m
+
6
m
+
6

gln

g6p
 m + 6 6pg
(5.2)
Eq. 5.1 shows the mass balance for 6PG, while Eq. 5.2 shows the labeling balance in which the
fluxes are multiplied by the mass isotopomer distribution vector of the respective metabolite.
Substituting Eq. 5.1 into Eq. 5.2 yields:
 m + 0 
 m + 0 
m + 0 
m + 0 




 


 
m + 1 
 m + 1
 m + 1 
 m + 1 
v 2  
−
=
v

−
3 
  
   
  
   




 


 
 m + 6 gln  m + 6 6pg 
 m + 6 6pg  m + 6 g6p 
(5.3)
By measuring the uptake rate of gluconate (v2) and the mass isotopomer distributions of G6P,
6PG and gluconate an over-determined system is obtained, from which v3 can be estimated.
Since measurement errors occur on both sides of Eq. 5.3, a sequential quadratic programming
algorithm was used for estimating v3. The PPP split-ratio is obtained by dividing v3 by the
uptake rate of glucose (v1). Note that for an accurate estimation of the PPP split-ratio, the
measured mass isotopomer distribution vectors in Eq. 5.3 should be corrected for naturally
occurring isotopes of hydrogen, nitrogen and oxygen [36].
By introducing a labeling difference amongst the metabolites directly surrounding the entry
point of the PPP and by measuring the labeling pattern of these metabolites, the split-ratio can
be accurately determined without any of the assumptions made in the previously developed
methods. Furthermore, a direct assessment of the surrounding metabolites rules out any
uncertainties introduced by deducing the labeling of primary metabolites from measured
labeling patterns of proteinogenic amino acids.
In order to accurately measure the PPP split-ratio using the method described above, several
conditions have to be met. First of all, the uptake-rate of gluconate should be small, not to
disturb the overall metabolism of the cell. Gluconate should thus only be added in tracer
amounts to the medium, hence the name gluconate-tracer method. Secondly, the investigated
microorganism has to simultaneously metabolize glucose and gluconate. In general, mixtures
of carbon-sources are only taken up simultaneously by a cell under carbon-limiting conditions
[10]. Under conditions of excess carbon, the molecular mechanisms of catabolite repression
and inducer exclusion ensure that the cell first depletes the energetically more favorable
carbon-source. The simultaneous growth on both glucose and gluconate has been reported
for various microorganisms, such as Corynebacterium glutamicum [24], Bacillus subtilis
133
Chapter 5
[7], Escherichia coli [17] and Schizosaccharomyces pombe [2]; in this study we focus on
P. chrysogenum. A last prerequisite for accurately applying the gluconate-tracer method is
knowledge about the glucose metabolism of the studied micro-organism. Glucose-oxidizing
reactions catalyzed by glucose oxidase or glucose dehydrogenase, can give rise to an influx
of unlabeled carbon (originating from glucose) into the uniformly 13C-labeled gluconate pool,
thereby complicating the determination of the PPP split-ratio. In general, however, these
enzymes are only expressed under conditions of glucose-excess. Harris et al. [15], for
example, showed that the enzyme glucose oxidase could not be detected in a glucose-limited
chemostat culture of P. chrysogenum.
n- C6
12
v2
v1
Biomass
g6p
u-13C6
gln
glc
v3
6pg
v5
v4
f6p
134
Entner
Doudoroff
pathway
Figure 1 Schematic diagram of the
metabolic network surrounding the G6Pnode when growing cells simultaneously
on glucose and gluconate. The PPP
split-ratio is defined as v3/v1. (G3P:
glyceraldehyde-3-phopshate,
other
abbreviations are listed in the text)
Non-oxidative
branch of the
PPP
g3p
Lower
glycolysis and
TCA-cycle
5.3 MATERIALS AND METHODS
5.3.1
Strain and cultivation conditions
All cultivations described in this study were performed in a carbon-limited chemostat system
operated at a growth rate of 0.02 h-1, using a high-yielding industrial P. chrysogenum strain
(code name DS17690, kindly donated by DSM Anti-Infectives, Delft, The Netherlands). This
strain was a re-isolation of the strain (code name DS12975) previously used in our lab [34].
Both the gluconate-tracer and the 13C-based MFA method for determining the PPP split-ratio
were carried out in a 1 L bioreactor (Applikon, Schiedam, The Netherlands). The working
volume of the reactor was kept constant at 0.6 L by means of an overflow system. Effluent
was removed discontinuously by pumping out liquid at fixed time intervals (1 min every 10
min). Temperature was controlled at 25oC­, while the pH was controlled at 6.50 by automatic
titration with 0.5 M NaOH and 0.25 M H2SO4. To keep the dissolved oxygen tension above
50% the bioreactor was equipped with one Rushton turbine stirrer (600 rpm) and aerated with
13
C-Labeled gluconate tracing: method
pressurized air at 10 L/h (0.28 vvm) at an overpressure of 0.1 bar. Silicone antifoam agent
(BDH, Poole, UK) was diluted 1:10 (v/v) and added to the reactor at a fixed rate of 0.2 mL/h.
Cells were grown on a scaled down version of the minimal medium described by van Gulik
et al. [34] in order to minimize the costs of the labeling experiments. For the 13C-based
MFA experiment the minimal medium contained: 3.3 g/L glucose.H2O, 0.68 g/L sodiumacetate.3H2O, 0.35 g/L KH2PO4, 1.54 g/L (NH4)2SO4, 0.22 g/L MgSO4.7H2O, 0.53 g/L phenylacetic acid (PAA) and 0.90 mL/L trace element solution. The trace element solution contained
75 g/L Na2-EDTA.2H2O, 2.5 g/L CuSO4.5H2O, 10 g/L ZnSO4.7H2O, 10 g/L MnSO4.H2O, 20
g/L FeSO4.7H2O, and 2.5 g/L CaCl2.2H2O. An identical medium composition was used for the
gluconate-tracer experiment, the only modification being the replacement of 5 Cmol% glucose
by an equimolar amount of glucono- δ -lactone (0.005 Cmol/l). The reason for the addition
of a small amount of sodium-acetate to the medium was to introduce an additional inflow of
labeled carbon into the metabolism for a better estimation of the fluxes in the lower part of the
metabolism using the 13C-based MFA method.
PAA, the side chain precursor for penicillin-G, was added to the medium at such a
concentration that the residual concentration in the chemostat remained at around 3 mM.
At this concentration PAA was neither limiting for penicillin-G production nor inhibiting for the
growth of P. chrysogenum. For the preparation of the minimal medium the appropriate amount
of PAA was dissolved in 1 L of demineralized water, set at a pH of 5.40 using 1 M KOH and
autoclaved at 121oC for 40min. The remaining medium components were dissolved in 4 L of
demineralized water, set at a pH of 5.40 using 1 M KOH and added filter-sterilized to the PAA
solution using a Acropak20 filter (PALL, East-Hills, NY, USA). After preparation the minimal
medium was stirred for at least 12 hours on a magnetic stirrer. At a medium pH of 5.40, these
12 hours ensured that all the added glucono- δ -lactone was hydrolyzed to gluconate [32].
Apart from the smaller chemostat volume, the batch phase and the first part of the chemostat
phase were carried out as described by van Gulik et al. [34]. The batch phase of the cultivation
was inoculated with spores from 2.0 g of rice grains. The sole energy source during the batch
phase was 3.3 g/L of glucose.H2O. The end of the batch phase was typically reached after 50 h,
after which the reactor was switched to continuous mode using the minimal medium described
before. After two residence times this medium was replaced by a chemically identical medium,
but with part of the naturally labeled carbon replaced by 13C. For the 13C-based MFA, 60
Cmol% of the glucose was replaced by specifically labeled 1-13C1 glucose (Sigma-Aldrich,
St. Louis, MO, USA), 20 Cmol% of the glucose was replaced by [U‑13C]glucose (SigmaAldrich) and 100 Cmol% of the acetate was replaced by [U‑13C]acetate (Sigma-Aldrich). For
the gluconate-tracer experiment 100 Cmol% of the gluconate (0.005 Cmol/l) was replaced by
[U‑13C]glucono- δ -lactone (Omicron, South Bend, IN, USA).
5.3.2Biomass sampling
Duplicate 10 mL samples were withdrawn from the bioreactor every second day for determining
the biomass dry weight. Samples were filtered over preweighted glass fiber filters (PALL) and
dried at 70oC for at least 24 h. The collected filtrate was immediately frozen in liquid nitrogen
and used for analyzing the extracellular penicillin-G and PAA concentrations.
Recently, Mashego et al. [28] observed that the feeding-time with 13C-labeled substrate
135
Chapter 5
needed to reach isotopic steady-state of the intracellular metabolites is longer than previously
anticipated. A possible explanation for the slow replacement of unlabeled metabolites is the
turnover of unlabeled storage sugars. To ensure that the intracellular primary metabolites
measured in this study were both in chemical and isotopic steady-state, the cultivation was
continued for 3 residence times after switching to the 13C-labeled medium, before harvesting
broth samples for determining the extracellular residual substrate concentrations and the mass
isotopomer distributions of intracellular metabolites. Extracellular samples were acquired by
rapidly sampling 2 mL of broth into a syringe containing precooled stainless steel beads (‑18oC),
immediately followed by separation of cells and medium by filtration [27]. Samples were stored
at –80oC prior to analysis. Samples for intracellular metabolite determinations were obtained
by rapidly withdrawing 1 mL of broth from the bioreactor, followed by direct injection of the
sample in 5 mL of a 60% (v/v) methanol/water mixture (-40oC) for instantaneous quenching
of the cell metabolism [23]. In total 16 x 1 mL of broth was sampled from the bioreactor and
further processed for metabolite extraction.
136
5.3.3
Metabolite extraction
8 out of the 16 samples harvested for metabolite extraction were centrifuged for 5 min at
5000 g in a cooled centrifuge (-20oC) equipped with a precooled (-40oC) swing-out rotor with
four buckets (Heraeus, Hanau, Germany). After decanting the supernatant, each pellet was
resuspended with one of the 8 remaining samples, resulting in a doubling of the amount
of biomass in each tube. This step was performed to compensate for the low biomass
concentration in the bioreactor (+/-1 g/L). The pooled samples were centrifuged for 5 min
at 5000 g. After decanting the supernatant, the pellet was resuspended in fresh 60% (v/v)
methanol/water (-40oC) and centrifuged for 5 min at 5000 g. This step ensured the removal
of any extracellular components (e.g. salts) that would otherwise hamper the metabolite
analysis. After repeating this washing-step for a second time, the samples were decanted
and stored at –40oC in a cryostat. The metabolites were extracted from the pellets using the
boiling ethanol procedure described by Lange et al. [23]. The boiled extracts were evaporated
under controlled vacuum and temperature for 45 min in a Rapid-Vap (Labinco, Kansas City,
MO, USA) to remove the ethanol and traces of methanol. It was shown that within 45 min all
ethanol and methanol had evaporated from the samples, leaving a mixture of metabolites and
cell debris suspended in 200 µl of water. The samples were centrifuged for 10 min at 13000 g
and the supernatant was stored at -80oC prior to LC-MS analysis.
5.3.4
Sample analysis
The mass isotopomer distributions of the intracellular metabolites were measured as
described by van Winden et al. [37]. Metabolites were first separated by high-performance
anion exchange chromatography (Waters, Milford, MA, USA) followed by MS analysis
with a Quatro-LC triple quadrupole mass spectrometer (Micromass Ltd., Manchester, UK)
equipped with an electrospray ionization interface. Samples were analyzed for the following
intermediates of the glycolysis and the PPP: G6P, F6P, 6PG, gln, manose-6-phopshate (M6P),
1,6-fructose-bis-phosphate (FBP), phosphoenol-pyruvate (PEP), the combined pool of 2- and
3-phosphoglycerate (2/3PG), the combined pool of xylulose-5-phosphate, ribose-5-phosphate
13
C-Labeled gluconate tracing: method
and ribulose-5-phosphate (P5P) and sedoheptulose-7-phosphate (S7P).
The concentrations of glucose, gluconate and acetate in the medium and the bioreactor were
enzymatically determined (Scil Diagnostics, Viernheim, Germany).
5.3.5
Enzymatic analysis
Cell extracts for the enzymatic analysis of glucose oxidase, glucose dehydrogenase and 6PG
phosphatase were prepared according to Harris et al. [15]. Cell extracts were analyzed for
glucose dehydrogenase activity using the assay described by van Dijken et al. [33]. Filtrate
and cell extracts were analyzed for glucose oxidase activity using the assay described by
Harris et al. [15]. 6PG phosphatase activity was tested by incubating (25oC, 30 min) 1 mL cell
extract with 100 µmol 6PG and enzymatically measuring the production of gluconate (Scil
Diagnostics, Viernheim, Germany). Background gluconate levels in the cell extract were used
as off-set.
13
C-Label distribution of glucose and gluconate
5.3.6
Mass isotopomer distribution of the labeled glucose mixture used for the 13C-labeling based
MFA and the [U‑13C]gluconate used for the gluconate-tracer method were determined via
LC-MS analysis. Since glucose could not be directly measured on the LC-MS, the mixture of
[1‑13C]glucose, [U‑13C]glucose and naturally labeled glucose was first phosphorylated to G6P
by incubating 200 µL of medium with 15 µL of 0.25M ATP (pH 7.0) and 3 µl of hexokinase
(1500 U/mL, Roche Diagnostics, Almere, The Netherlands) for 30 minutes at 30oC. Enzyme
and metabolites were separated by centrifugation at 6000 g for 30 min on an Ultrafree-MC
10.000 NMWL spinfilter (Millipore, Billerica, MA, USA). Supernatants were stored at ‑80oC
prior to LC-MS analysis.
13
C-based metabolic flux analysis
5.3.7
The metabolic network model of the glycolysis used for the 13C-based MFA of P. chrysogenum
contained all conventional reactions. The reactions of the non-oxidative branch of the PPP were
modeled as metabolite specific, reversible, C2 and C3 fragments producing and consuming
half-reactions as proposed by Kleijn et al [22]. For reasons stated in the results section, the
metabolic network model also included the decarboxylation of oxaloacetate (OAA) into PEP
catalyzed by the enzyme PEP-carboxykinase. The reversible reactions were modeled as
separate forward and backward vreactions and are referred to as net and exchange fluxes,
where:
v net = v forward − v backward
(5.4)
v exchange = min(v forward ,v backward )
(5.5)
The employed flux fitting procedure is described in detail by van Winden et al. [37]. In short,
the procedure uses the cumomer balances and cumomer to isotopomer mapping matrices
introduced by Wiechert et al. [39] to calculate the isotopomer distributions of metabolites in
a pre-defined metabolic network model for a given flux-set. The flux-set that gives the best
correspondence between the measured and simulated 13C-label distribution is determined by
non-linear optimization and denoted as the optimal flux-fit. All calculations were performed in
137
Chapter 5
Matlab 6.1 (The Mathworks Inc, Natick, MA, USA).
5.4 RESULTS
5.4.1
Validity of the gluconate-tracer method
Prior to applying the gluconate-tracer experiment for the direct determination of the PPP splitratio it was checked whether P. chrysogenum could simultaneously consume glucose and
gluconate under carbon-limiting conditions. Furthermore, it was investigated to which extent
the metabolism of P. chrysogenum was influenced by the replacement of 5 Cmol% glucose
by gluconate. A trial cultivation was performed in which P. chrysogenum was grown for 4.0
residence times on the 13C-based MFA medium containing only glucose and acetate, followed
by 4.0 residence times of cultivation on a gluconate-tracer medium with 5 Cmol% of gluconate.
As shown in Figure 2 no significant changes in biomass dry-weights, penicillin‑G production
rates and PAA consumption rates were observed for the two different medium compositions.
Throughout the cultivation the uptake rate of PAA matched the production rate of penicillin-G,
indicating that PAA was not catabolized by the cell.
1.6
1e-3
Medium switch
1.4
Dry weight (g/L)
138
1.0
6e-4
0.8
4e-4
0.6
0.4
Dryweight
-qPAA
qPenicillin-G
0.2
-qPAA and qPenicllin-G
(mol.Cmol-1h-1)
8e-4
1.2
2e-4
0.0
0.0
0
2
4
6
Residence Time (-)
8
10
Figure 2 Penicillin-G production rate, PAA consumption rate and biomass concentration in the
chemostat cultivation of P. chrysogenum performed to validate the gluconate-tracer method. After ~
4 residence times the composition of the feed was slightly altered; 5 Cmol% of the original glucose
concentration was replaced by an equimolar amount of gluconate. At the same time the chemostat
was pulsed with a gluconate solution such that the initial concentration in the chemostat equaled
0.160 g/L.
The effect of the gluconate addition on the intracellular metabolism was investigated by
measuring the steady-state intracellular metabolites levels before and after switching to
the gluconate-containing medium (Figure 3). Metabolite levels were measured by isotope
dilution mass spectrometry using u-13C-labeled metabolite extracts as internal standard as
13
C-Labeled gluconate tracing: method
described by Wu et al. [40]. In this same paper it was found that for most measured metabolite
levels the maximal standard deviation did not exceed 10%. A t-test based upon this 10%
maximal standard deviation showed that the metabolite levels before and after the addition
of gluconate were not significantly different, the only exception being FBP. The observed
discrepancy in FBP-level might be due to the extremely low concentrations of this metabolite
in the sample. The metabolite levels presented in Figure 3 seem to be slightly lower after
the addition of gluconate. However, this lowering is not necessarily an effect of the added
gluconate. Presented metabolite levels are biomass specific, an overestimation of the
biomass concentration at the time of sampling will, therefore, result in an underestimation of
all biomass specific metabolite levels.
4
w/o gluconate
w/ gluconate
Metabolite concentration
(µmol/g dry weight)
3
2
139
1
6PG
FBP
2/3PG
PEP
T6P
M6P
F6P
G6P
G1P
α-KG
MAL
FUM
PYR
*
Glyox
0
Metabolite
Figure 3 Steady state intracellular metabolite concentrations of P. chrysogenum before and after
switching to the gluconate containing feed. Broth samples for metabolite analysis were harvested
after about 4 and 8 residence times. Displayed standard deviations are fixed at 10%. Metabolites
showing a significant difference in concentration before and after the addition of gluconate were
found by using a two-tailed equal variance t-test and are marked with an asterisk. A p-value smaller
than 0.01 was considered significant. (GLYOX: glyoxylate, PYR: pyruvate, FUM: fumerate, MAL:
malate, AKG: α-ketogluterate, G1P: glucose-1-phosphate, T6P: threhalose-6-phosphate, other
abbreviations are listed in the text).
Chapter 5
Directly after switching to the gluconate-containing medium the bioreactor was pulsed with
a gluconate solution to attain the same gluconate concentration in the bioreactor as in the
medium (~0.160 g/L). After administering the pulse, the decrease in the extracellular residual
gluconate concentration in the bioreactor was followed over time (Figure 4). Within half an
hour after the pulse the gluconate concentration in the bioreactor started to decrease. The
residual glucose concentration in the bioreactor remained constant at ~3 mg/L throughout
the experiment. These results clearly demonstrate that P. chrysogenum is capable of
simultaneously metabolizing glucose and gluconate under carbon-limiting conditions. This
simultaneous uptake is most probably triggered by the low glucose concentration during the
experiment, enabling the induction of the genes needed for the uptake of other substrates
(absence of catabolite repression). The gluconate consumption rate was calculated from
the time pattern of the extracellular gluconate concentration. Figure 4, shows that during the
first 12 hours the specific uptake rate of gluconate reaches ~6.10‑3 mol/Cmol/h, followed by
a decrease to the steady-state uptake rate of 0.64.10-3 mol/Cmol/h. P. chrysogenum clearly
needs time to fully adjust to the presence of gluconate, indicating that the required transport
system for gluconate and/or the enzyme gluconokinase are induced rather than constitutively
expressed.
Gluconate chemostat concentration
Gluconate specific uptake-rate
Glucose chemostat concentration
140
Concentration
(mg/L)
140
6e-3
120
5e-3
100
4e-3
80
3e-3
60
2e-3
40
1e-3
20
0
0.0
20.0
40.0
60.0
80.0
Specific uptake-rate of gluconate
(mol/Cmol/h)
7e-3
160
0
100.0
Time after medium switch (h)
Figure 4 Measured gluconate and glucose concentration in the chemostat before and after the
addition of the gluconate-pulse (t=0 h). The specific uptake rate of gluconate was estimated based
upon the measured uptake dynamics of gluconate.
5.4.2
The gluconate-tracer method and P. chrysogenum
The gluconate-tracer method was used to determine the PPP split-ratio in a penicillin-G
producing chemostat cultivation of P. chrysogenum operated at a dilution-rate of 0.02 h-1. The
first data-column of Table 1 gives the ‘measured’ macroscopic parameters of the cultivation.
Apart from an erroneous oxygen consumption-rate, identified via rate-balancing according
13
C-Labeled gluconate tracing: method
to van der Heijden et al. [38], no major discrepancies were found between the balanced and
‘measured’ macroscopic rates. Note that the uptake rate of gluconate was just below 5% of
the glucose uptake rate.
Table 1 Macroscopic parameters of the gluconate-tracer experiment and the 13C-based MFA experiment.
(nd: not determined).
Experiment
Macroscopic parameter
Unit
Glucose consumption
mmol/CmolX/h
8.60±0.40
8.37±0.45
Gluconate consumption
mmol/CmolX/h
0.42±0.02
nd
Acetate consumption
mmol/CmolX/h
2.67±0.12
2.52±0.14
Oxygen consumption
mmol/CmolX/h
48.91±2.10
32.6±1.3
PAA consumption
mmol/CmolX/h
0.50±0.05
0.55±0.06
Carbondioxide production
mmol/CmolX/h
31.34±1.57
26.01±4.53
Biomass production
1/h
0.020±0.001
0.019±0.001
Penicillin-G production
mmol/CmolX/h
0.51±0.05
0.48±0.08
Biomass concentration
g/L
1.04±0.05
1.07±0.06
Carbon recovery
%
104.0±2.2
92.2±5.9
Gluconate-tracer
13
C-based MFA
The measured mass isotopomer distribution of the [U‑13C]gluconate added to the medium is
shown in the upper part of Table 2. Based upon the measured mass fractions an isotopic purity
of 99% (~0.941/6) was calculated. This value is higher than the 95% isotopic purity specified
by the manufacturer, from which a m+6 mass isotopomer fraction of only 0.74 is predicted for
gluconate. Moreover, use of the manufacturer specified isotopic purity would have resulted in
a very different estimate for the PPP split-ratio, illustrating the importance of measuring the
label distribution of the substrate(s) used.
After 3 residence times of feeding on medium containing [U‑13C]gluconate, broth was harvested
for measuring the mass isotopomer fractions of 6PG, G6P, F6P and intracellular gluconate
(Table 2). As expected the addition of [U‑13C]gluconate led to a distinct m+6 mass isotopomer
fraction for 6PG. No discernable m+6 mass fractions were observed for G6P and F6P due to
the decarboxylation of 6PG in the oxidative branch of the PPP and its carbon redistribution in
the non-oxidative branch of the PPP.
Unexpected mass fractions were observed for intracellular gluconate. The unlabeled fraction
(m+0) of the intracellular gluconate was measured to be 22%, while this fraction was
undetectable in the gluconate added to the medium (Table 2). Apparently, an unidentified
reaction caused an inflow of unlabeled carbon into the otherwise uniformly 13C-labeled
gluconate pool. Two possible reactions were hypothesized to explain this phenomenon; the
oxidation of glucose to gluconate by either glucose oxidase or glucose dehydrogenase and the
dephosphorylation of intracellular 6PG to gluconate by a phosphatase (Figure 5). In literature
several aspecific phosphatases have been reported for P. chrysogenum [14, 29]. Given the
competitive inhibition of these phosphatases by inorganic phosphate (ki = 0.42 mM), it seems
unlikely that these enzymes are expressed under the relatively high phosphate concentrations
141
Chapter 5
Table 2 Measured mass isotopomer fractions and standard deviations for the gluconate-tracer experiment.
Presented mass fractions have been corrected for the natural isotopes of the elements hydrogen and
oxygen.
Measured metabolite
Mass isotopomer
Measured
Gluconate in medium
m+0
0.000±0.000
m+1
0.000±0.000
m+2
0.000±0.000
m+3
0.000±0.000
m+4
0.002±0.000
m+5
0.059±0.001
m+6
0.939±0.001
m+0
0.217±0.006
m+1
0.019±0.001
m+2
0.003±0.001
m+3
0.001±0.000
m+4
0.002±0.001
m+5
0.040±0.003
m+6
0.718±0.007
m+0
0.866±0.003
m+1
0.094±0.002
m+2
0.017±0.002
m+3
0.015±0.000
m+4
0.007±0.000
m+5
0.001±0.000
m+6
0.001±0.000
m+0
0.852±0.004
m+1
0.094±0.003
m+2
0.024±0.001
m+3
0.016±0.001
m+4
0.012±0.002
m+5
0.002±0.001
m+6
0.000±0.000
m+0
0.786±0.003
m+1
0.082±0.003
m+2
0.015±0.001
m+3
0.011±0.000
m+4
0.016±0.003
m+5
0.008±0.001
Gluconatea
G6Pa
142
F6Pa
6PGa
m+6
a
Intracellular metabolite surrounding the G6P-node.
0.081±0.003
employed in this study. However, the same restriction holds for the glucose-oxidizing enzymes.
As pointed out in the theoretical section, expression of these enzymes is normally triggered by
excess glucose, while in this study the chemostat cultivations were preformed under carbonlimiting conditions.
13
C-Labeled gluconate tracing: method
Enzymatic analysis on cell extracts from the gluconate-tracer experiment showed no
measurable activity for any of the three proposed enzymes, demonstrating that enzyme
activities were below the detection limit of the applied assays. Based upon the conceived
fluxes for the oxidation of glucose or the dephosphorylation of 6PG (see Appendix A, Table AI), the minimal specific activity for all three enzymes was estimated to be around 2 ⋅ 10-4 µmol/
mg protein/min. This value is indeed much lower than the detection limit of the assays (±0.01
µmol/mg protein/min). In spite of the fact that the in vitro assays for these enzymes showed
no measurable activity, it can therefore not be excluded that the measured m+0 fractions of
intracellular gluconate were caused by the activity of one of these enzymes.
v1
v1-v6
Biomass
glucose oxidase/
v2
dehydrogenase
gln ex
glc
v6
v2
gln in
glc
g6p
v3
v7
6pg
v
v8 phosphatase
Figure 5 Two candidate reactions for
explaining the observed unlabeled
mass
isotopomer
fraction
of
intracellular gluconate (glnin): the
dephosphorylation of 6PG (v8) by a
phosphatase and the oxidation of
glucose (v6) by either glucose-oxidase
or glucose dehydrogenase.
v5
4
In appendix A the effect of the two proposed metabolic scenarios on the estimation of the PPP
split-ratio is examined. The oxidation of unlabeled glucose to gluconate causes a small fraction
of the unlabeled carbon to flow into the 6PG-pool via the uptake of gluconate instead of via the
conventional oxidation of G6P, thereby altering the estimated PPP split-ratio. However, due to
the small size of the glucose-oxidation flux, the change in the PPP split-ratio was practically
negligible. The dephosphorylation of 6PG has no effect on the label-inflow into the 6PG-pool.
Irrespective of whether the phosphatase is actively expressed, all unlabeled carbon in 6PG
originates from G6P, while all 13C-label originates from [U‑13C]gluconate added to the medium.
Hence an unchanged estimate for the PPP split-ratio was obtained. These findings justify
the negligence of the glucose-oxidizing reaction in the PPP split-ratio estimations presented
below.
The mass isotopomers of Table 2 and the measured uptake rates of glucose and gluconate
were combined with Eq. 5.3 to estimate the PPP split-ratio in P. chrysogenum. The 95%
confidence interval of the PPP split-ratio was determined by Monte Carlo simulation in which
the added noise was normally distributed with the measured standard deviations of the mass
isotopomer fractions. The optimization routine yielded a PPP flux of 4.45±0.36 mmol/Cmol/h,
meaning that 51.7±4.2% of the total glucose entering the cell was metabolized via the PPP.
Note that for every mole of glucose catabolized in the oxidative branch of the PPP two moles of
NADPH are produced, while for every mole of gluconate catabolized only one mole of NADPH
is produced. Since the flux through the PPP is related to the cytosolic NADPH demand of
143
Chapter 5
P. chrysogenum [34], the PPP split-ratios calculated via the gluconate-tracer method will be
slightly overestimated. In the most extreme case of direct proportionality between the PPP flux
and the cytosolic NADPH demand, the PPP split-ratio should be slightly corrected to 49.5%.
However, normalization of this value for the total uptake-rate of glucose and gluconate (v1+v2)
increases the PPP split-ratio again to 51.8%.
Mass isotopomer fractions smaller than 0.03 were not included in the fitting procedure as it
was observed that these fractions led to a considerable increase in the variance-weighted
sum of squared residuals (SSres). This increase may be explained by the fact that the standard
deviations shown in Table 2 are only based upon precision limitations of the LC-MS and do
not account for any systematic measurement errors. Especially for the small mass isotopomer
fractions these systematic errors can result in large contributions to the minimized sum of
squared residuals. A minimal SSres of 2.59 was found when fitting the PPP split-ratio. To test
whether the SSres is solely caused by normally distributed measurement errors a χ2-distribution
was used, in which the degrees of freedom (df) equal the number of fitted mass isotopomers
fractions minus the number of fitted parameters. The exclusion of the m+2, m+3 and m+4
mass isotopomer fractions of G6P, 6PG and gluconate leaves 3 degrees of freedom. Given
the probability P(χ2(3)<2.59)=0.5408, it follows that the fitted PPP split-ratio is well within the
statistical acceptable range.
144
13
C-based MFA of P. chrysogenum
5.4.3
By measuring the mass isotopomers of 13C-labeled primary metabolites the metabolic fluxes in
the PPP and glycolysis of P. chrysogenum were estimated in a second chemostat cultivation
operated under conditions identical to that of the gluconate-tracer experiment, but now with
labeled glucose instead of gluconate added to the feed. The macroscopic parameters for the
cultivation are shown in Table 1. As expected, similar biomass concentrations and uptake and
secretion rates were observed for the gluconate-tracer experiment and the 13C-based MFA
experiment.
The first data-column of Table 3 shows the measured mass isotopomer fractions of the
glucose in the medium and the 9 intracellular metabolites after growing P. chrysogenum for
3 residence times on the 13C-labeled medium. The mass fractions of glucose were used to
calculate the isotopic purity of the [U‑13C]glucose (98%) and [1‑13C]glucose (99%) used in the
feed. Both isotopic purities matched the manufacturer-specified values.
Prior to performing any calculations, a qualitative analysis of the obtained mass isotopomer
data provides additional insight in the primary metabolism of P. chrysogenum under the applied
cultivation conditions. For example, the near identical mass isotopomer fractions for G6P and
F6P in Table 3 indicate that the glucose-isomerase catalyzed reaction is close to equilibrium.
Only a high reversibility of the glucose-isomerase catalyzed reaction can efface the labeling
difference in the G6P and F6P pool caused by F6P originating from the non-oxidative branch
of the PPP.
The difference in labeling between the 2/3PG and PEP measurements of Table 3 suggests
that the reaction catalyzed by enolase is not fully reversible. Furthermore, the difference in
labeling also indicates that PEP does not only originate from 3PG. The most likely candidate
for a second PEP-producing reaction is the gluconeogenic reaction catalyzed by PEP-
13
C-Labeled gluconate tracing: method
Table 3 Measured and fitted mass isotopomer fractions for the 13C-based MFA experiment.
Measured Metabolite
Glucose in
medium
Mass isotopomer Measured
Fitted
Difference
m+0
0.196±0.001
m+1
0.569±0.002
m+2
0.037±0.000
m+3
0.012±0.000
m+4
0.002±0.000
m+5
0.019±0.000
m+6
0.166±0.001
a
G6P
m+0
0.200±0.012
0.205
-0.005
m+1
0.413±0.011
0.423
-0.011
m+2
0.129±0.005
0.124
0.005
m+3
0.084±0.003
0.081
0.003
m+4
0.058±0.004
0.053
0.005
m+5
0.028±0.002
0.027
0.002
m+6
0.088±0.004
0.084
0.004
a
6PG
m+0
0.205±0.008
0.205
0.000
m+1
0.382±0.016
0.422
-0.040
m+2
0.146±0.016
0.124
0.021
m+3
0.081±0.002
0.082
-0.001
m+4
0.063±0.005
0.053
0.009
m+5
0.036±0.004
0.027
0.009
m+6
0.088±0.003
0.084
0.004
a
F6P
m+0
0.192±0.006
0.206
-0.014
m+1
0.379±0.016
0.414
-0.035
m+2
0.130±0.005
0.130
0.001
m+3
0.087±0.004
0.086
0.001
m+4
0.059±0.005
0.056
0.003
m+5
0.032±0.003
0.027
0.005
m+6
0.121±0.004
0.079
0.042
a
M6P
m+0
0.205±0.006
0.206
0.000
m+1
0.393±0.016
0.414
-0.021
m+2
0.137±0.005
0.130
0.007
m+3
0.090±0.004
0.086
0.004
m+4
0.063±0.005
0.056
0.006
m+5
0.031±0.003
0.027
0.004
m+6
0.081±0.004
0.079
0.002
FBPa
m+0
0.223±0.014
0.223
0.000
m+1
0.345±0.012
0.349
-0.004
m+2
0.142±0.008
0.143
-0.001
m+3
0.128±0.010
0.130
-0.002
m+4
0.074±0.004
0.068
0.006
m+5
0.028±0.003
0.029
0.000
m+6
0.060±0.006
0.056
0.004
2/3PGa
m+0
0.514±0.010
0.501
0.013
m+1
0.253±0.006
0.254
-0.001
m+2
0.081±0.013
0.097
-0.015
m+3
0.152±0.007
0.144
0.008
a
PEP
m+0
0.446±0.009
0.456
-0.010
m+1
0.285±0.007
0.277
0.008
m+2
0.126±0.006
0.123
0.003
m+3
0.143±0.005
0.139
0.003
P5Pa
m+0
0.481±0.017
0.486
-0.005
m+1
0.213±0.006
0.205
0.009
m+2
0.091±0.016
0.108
-0.017
m+3
0.075±0.005
0.081
-0.006
m+4
0.062±0.009
0.052
0.010
m+5
0.078±0.014
0.066
0.012
a,b
S7P
m+0
0.221±0.012
0.231
-0.011
m+1
0.286±0.008
0.292
-0.006
m+2
0.192±0.008
0.188
0.004
m+3
0.121±0.005
0.119
0.002
m+4
0.084±0.006
0.080
0.004
m+5
0.067±0.004
0.064
0.003
m+6
0.029±0.003
0.027
0.003
a
Intracellular glycolytic or PPP metabolites
b
The m+7 mass fraction of S7P was not measured; presented values are normalized for m+0 to m+6.
145
Chapter 5
146
carboxykinase, converting one molecule of oxaloacetate (OAA) into PEP and CO2 at the
expense of one ATP. A possible explanation for the presence of this enzyme is its induction by
the co-substrate acetate. Nevertheless, it remains intriguing why a microorganism expresses
an energy-consuming gluconeogenic reaction, when it can produce all upper glycolytic
intermediates from glucose. Van Winden et al. [35] have previously shown the importance of
including a PEP-carboxykinase in fitting P. chrysogenum 13C-labeling data.
Based upon the qualitative analysis of the measurements in Table 3 a refined metabolic
network model of the glycolysis and the PPP was constructed for the 13C-based MFA of
P. chrysogenum. Flux estimates for the glycolytic and PPP reactions were obtained by fitting
the mass isotopomer distribution model to the measured mass isotopomer distributions. The
fitted mass isotopomer fractions and the estimated fluxes are shown in the second column of
Table 3 and Figure 6, respectively. All fluxes are normalized (on mole base) to a glucose influx
of 100. Using the 13C-based MFA method the PPP split-ratio was estimated to be 51.1%, which
is very similar to the value determined with the gluconate-tracer method. Figure 6 also shows
that, in accordance with earlier observations, a flux was fitted for the PEP-carboxykinase
catalyzed reaction. However, the extent of this flux could not be quantified due to the fact that
only mass isotopomer fractions of glycolytic and PPP metabolites were measured.
The fit presented in Figure 6 yielded a minimized variance-weighted SSres of 60.2. As
described before, this variable follows a chi-squared distribution with, in this case, 30
degrees of freedom (47 independent data points used for fitting 17 parameters). Given the
probability P(χ2(30)<60.2)=0.999, it follows that within a 95% confidence interval the fit has to
be statistically rejected. Analysis shows that the mass isotopomer fractions of F6P and 6PG
contribute the most to the total SSres (data not shown).
glc
CO2
100
biomass
21
g6p
28 (>1000)
3
biomass
51
f6p
15 (30)
32 (14)
s7p
17 (68)
17 (28)
E-C3
e4p
57 (38)
1
biomass
biomass
g3p
2
129 (248)
4
biomass
17 (29)
E-C2
57
fbp
p5p
2
biomass
pep
> 129
>0
CO2
Lower
Metabolism
oaa
Figure 6 Flux estimates for the 13Cbased MFA of P. chrysogenum. Values in
grey represent the drain on the primary
metabolites
for
monomer-synthesis
(e.g.
amino-acids,
polysaccharides,
fatty-acids, nucleotides etc.). Fluxes
are normalized for the glucose uptake
rate. Values outside parentheses denote
the net fluxes, while values inside
parentheses represent the exchange
fluxes. Solid arrowheads denote the
direction of the net flux. (E4P: erythrose4-phosphate,
E-C2:
glycolaldehyde
moiety covalently bound to the thiamine
pyrophosphate/transketolase complex,
E-C3: dihydroxyacetone moiety covalently
bound to the enzyme transaldolase, other
abbreviations are listed in the text)
13
C-Labeled gluconate tracing: method
5.4.4
Accuracy and reproducibility of the estimated PPP split-ratios
The accuracy of the two methods for determining the PPP split-ratio was examined by studying
the sensitivity of the minimized SSres to changes in the PPP split-ratio. The metabolic network
model used for the 13C-based MFA is non-linear making the estimation of the confidence
interval based on linearization of the model around the optimally fitted fluxes prone to errors.
Therefore, the measured mass isotopomer fractions of both methods were re-fitted at fixed
values for the PPP split-ratio, ranging from 20% to 80% at a 5% interval. The SSres for each fit
is plotted against the fixed PPP split-ratio in Figure 7. By determining the points of intersection
of the curve with a horizontal line at the 95% confidence value for SSres and by subsequently
projecting these intersection points onto the x-axis, the 95% confidence interval for the PPP
split-ratio is obtained. The 95% confidence value for the SSres can be calculated as follows
(see Appendix B):
SS(βˆ )
95%
SSres
= SS(βˆ ) +
⋅ F0.05 (1,n − p)
(5.6)
n−p
where SS(βˆ ) is a constant corresponding to the globally fitted minimum for the SSres at the
optimal PPP split-ratio, n is the number of independent data points and p is the total number
of parameters in the model.
95%
SSres
-values were calculated to be 68.6 and 11.3 for the 13C-based MFA and the gluconatetracer method, respectively, which corresponded with a 95% confidence interval of [40.0-63.5]
and [47.6-56.2] (Figure 7). Note that the confidence interval for the gluconate-tracer method
matches well with the previously estimated confidence interval using Monte Carlo simulations
[47.7-56.1]. Results clearly show that the PPP split-ratio estimation via the gluconate-tracer
method is more precise.
50
100
40
30
80
SSres
SSres
120
20
60
10
40
0
20
30
40
50
60
PPP split-ratio
70
80
0
20
30
40
50
60
70
PPP split-ratio
Figure 7 Sensitivity analysis of the PPP split-ratio determination for the C-based MFA method (A)
and the gluconate-tracer method (B). The horizontal dashed line represents the 95% cut-off value
for the SSres, as can be calculated from Eq.5. Vertical dashed lines represent the upper and lower
boundary for the 95% confidence interval of the PPP split-ratio.
13
80
147
Chapter 5
Reproducibility of the estimated PPP split-ratios depends on the consistency of three factors:
(i) the chemostat cultivations, (ii) the rapid sampling procedure and (iii) the sample analysis.
The reproducibility of the chemostat cultivations follows from Table 1, showing that despite
the presence of 5 Cmol% gluconate, similar macroscopic parameters are obtained for the
gluconate-tracer method and the 13C-labeling based MFA method. Furthermore, repetition of
the gluconate-tracer experiment with naturally labeled gluconate added to the feed medium
yielded similar macroscopic parameters (results not shown). During each experiment multiple
independent samples were taken using the rapid sampling procedure. A one-way analysis
of variance (ANOVA) showed that variability between samples was not significantly larger
than the variability within a sample (i.e. between multiple injections on the LC-MS). Samples
withdrawal from the chemostat is thus also reproducible. The reproducibility of the sample
analysis follows from the small standard deviations of the 5 independent injections on the LCMS (<1%, see Table 2 & 3).
148
5.5 DISCUSSION AND ConClusion
Replacement of 5% Cmol glucose by [U‑13C]gluconate in a glucose-fed chemostat culture of
P. chrysogenum allows for an accurate determination of the PPP split-ratio without disturbing
the metabolism of the cell. Application of the method to a penicillin-G producing culture of
P. chrysogenum yielded a PPP split-ratio of 51.8%. Determination of the PPP split-ratio in
P. chrysogenum via a 13C-based MFA yielded a similar PPP split-ratio of 51.1%. This is the
first report of the use of mass isotopomer measurements of 13C-labeled primary metabolites
for determining the metabolic fluxes in P. chrysogenum.
Sensitivity analysis of the two estimated PPP split-ratios showed that the gluconate-tracer
method is more sensitive, illustrated by a 95% confidence interval for the PPP split-ratio that
was three times as small. The lower sensitivity of the 13C-based MFA method can be primarily
attributed to the nature of the method. While the gluconate-tracer method is dedicated to
the estimation of the PPP split-ratio, the 13C-based MFA method aims at the determination
of multiple fluxes within a defined metabolic network model. The high interconnectivity of
cellular metabolism allows changes in the metabolite-labeling (due to for example an altered
PPP split-ratio) to be counteracted by changed flux-patterns in other parts of the metabolism.
As a result, the fitting-procedure can produce similar minimized SSres for dissimilar metabolic
flux patterns. The preferred method thus depends on the objective of the experiment. The
gluconate-tracer method allows for an accurate determination of solely the PPP split-ratio,
whereas the 13C-based MFA method provides an estimate of the global flux distribution in the
cell, albeit not as accurate as the gluconate-tracer method.
In principle, the gluconate-tracer method can be extended to other parts of the metabolic
network. By using route specific 13C-labeled tracers it is possible to determine flux-patterns in
parts of the metabolic network, which otherwise might be hard to accurately determine using
the ‘holistic’ 13C-labeled MFA method. Essential for such tracer-studies is the simultaneous
uptake of the applied tracer with the main carbon-source. Furthermore, the effect of the tracer
addition on the overall cellular metabolism should be kept to a minimum. Quantification of the
flux distribution in a metabolic node requires an influx of the tracer directly after the branch-
13
C-Labeled gluconate tracing: method
point and measurement of the metabolites directly surrounding the branch-point.
Table 4 gives an overview of the previously published PPP split-ratios in a penicillin-producing
chemostat cultivation of P. chrysogenum, together with the measured penicillin productionrate, the employed growth-rate and the applied estimation method. PPP split-ratios varying
from 33% to 75% have been reported for P. chrysogenum. In part, this large variation can be
explained by different experimental conditions, such as differences in the penicillin productionrate (due to strain diversity) and the applied growth-rate. Nonetheless, these large variations
are also a result of assumptions made on the underlying stoichiometric model. The difference
in PPP split-ratio observed by van Gulik et al. [34] and Henriksen et al. [16], for example, can
be largely explained by the inclusion of a different cysteine biosynthesis pathway. Furthermore,
van Winden et al. [35] showed that different metabolic networks have a large impact on the
estimated PPP split-ratio in a 13C-based MFA.
As described in the previous paragraph, the main advantage of the gluconate-tracer method
is that it does not rely on a metabolic network model, but merely on an accurate determination
of the metabolite labeling directly surrounding the G6P-pool. The improved sensitivity of the
gluconate-tracer method in comparison to the 13C-based MFA method described in this study,
further adds to the credibility of the gluconate-tracer method. In that respect, none of the other
studies listed in Table 4 mention the confidence interval associated with the estimated PPP
split-ratio.
Table 4 Overview of the previously measured PPP split-ratios in a penicillin-producing chemostat
cultivation of P. chrysogenum.
PPP splitratio
(%)
Reference
µ
(h-1)
Penicillin production
Employed method
(mmol/CmolX/h)b
Henriksen et al. [16]
61
0.1
0.66
MFA
van Gulik et al. [34]
37
0.045
~0.55
MFA
Christensen et al. [4]
75
0.08
0.72
13
C-based MFA
C-based MFA
C-based MFA
70
0.06
0.31
13
Christensen et al. [5]
van Winden et al. [35]
33
0.03
0.45
13
Christensen et al. [3]
75
0.06
0.31
algebreic solution
51
0.02
0.45
13
This study
a
a
C-based MFA
This study
51
0.02
0.51
gluconate-tracer
Both these articles are based upon the same 13C-labeling data-set, but differ in the method used for
calculating the PPP split-ratio.
b
The molecular mass of biomass used for calculating the specific penicillin production rates was assumed
to be growth-rate independent and equal to 28.05 g dryweight per Cmol biomass [34].
a
ACKNOWLEDGEMENTS
This work was financially supported by the Dutch EET program (Project No. EETK20002)
and DSM. Diana Harris (Delft University of Technology) is gratefully acknowledged for her
practical help with the enzymatic determinations.
149
Chapter 5
APPENDIX A: Glucose oxidase or phosphatase acitivity
Two candidate reactions were hypothesized for explaining the difference in labeling between
the gluconate added to the medium and the intracellular gluconate in the gluconate-tracer
experiment (Table 2): the oxidation of glucose to gluconate by either glucose oxidase or
glucose dehydrogenase and the dephosphorylation of intracellular 6PG to gluconate by a
phosphatase (Figure 5). Flux-estimates for these reactions were derived by combining the
mass and labeling balance around the intracellular gluconate pool; similar to the derivation
of Eq. 5.3 in the theoretical section. In the case of a glucose-oxidation reaction (v6) the mass
balance around the intracellular gluconate pool becomes:
v7 = v 2 + v 6
(A-1)
Combining Eq. A-1 with the labeling balance yields:
 m + 0 

 m + 0 
m + 0
m + 0 








 
m + 1
 m + 1

 m + 1 
 m + 1 
v 2  
−
=
v

−
6 

  
  
   
 








 
 m + 6 gln_ ex  m + 6 gln_ in 
 m + 6 gln_ in  m + 6 glc 




150
(A-2)
In which gln_in and gln_ex are the intracellular and extracellular gluconate pool, respectively.
In the case of a dephosphorylation reaction (v8) the mass balance around the intracellular
gluconate pool becomes:
v7 = v 2 + v8
(A-3)
Combining Eq. A-3 with the labeling balance gives:
 m + 0 

 m + 0 
m + 0
m + 0 








 
m + 1
 m + 1

 m + 1 
 m + 1 
v 2  
−
=
v

−
8 

  
  
   
 








 






 m + 6 gln_ ex  m + 6 gln_ in 
 m + 6 gln_ in  m + 6 6pg 




(A-4)
Using a sequential quadratic programming algorithm the fluxes v6 (using Eq. A-2) and v8 (using
Eq. A-4) were estimated to be 0.13 and 0.15 mmol/Cmol/h, respectively. Both fits yielded a
similar SSres (14.85 and 15.86), making it impossible to distinguish between the two proposed
reactions based upon the measured labeling of intracellular gluconate. The flux-estimates for
v6 and v8 were, subsequently, used for estimating the corresponding oxidative PPP flux (v3).
Due to the changed mass and labeling balance around the 6PG-pool (Figure 5), Eq. 5.3 could
not be directly used for estimating v3. A new mass balance around the 6PG-pool was set up:
13
C-Labeled gluconate tracing: method
v7 + v3 = v 4 + v5 + v8
(A-5)
Combining Eq. A-5 with the labeling balance around 6PG yielded a new relation:
 m + 0 
 m + 0 
m + 0 
m + 0 




 


 
m + 1 
m + 1 
 m + 1 
 m + 1 


v7 
−
= v3 
−
  
   
  
   








 
 m + 6 gln_ in  m + 6 6pg 
 m + 6 6pg  m + 6 g6p 




(A-6)
Oxidative PPP fluxes (v3) were calculated for the metabolic pathways containing either the
glucose-oxidation reaction or the 6PG-dephosphorylation reaction, by substituting Eq. A-1 or
A-3 into A-6, respectively. An overview of the estimated fluxes is given in Table A-I.
The dephosphorylation of 6PG has no effect on the estimated PPP split-ratio given the fact
that it does not alter the origin of the label entering the 6PG-pool. Irrespective of an active
phosphatase, all unlabeled carbon entering the 6PG-pool originates from G6P, and all 13Clabel enters via the uptake of gluconate. Mathematical proof hereof is obtained by substituting
Eq. A-3 and A-4 in A-6, resulting in the originally derived equation (Eq. 5.3) for calculating the
PPP split-ratio (not shown).
In contrast, the oxidation of unlabeled glucose to gluconate causes an inflow of unlabeled
carbon into the 6PG-pool via gluconate-uptake instead of via the classical oxidation of G6P.
Consequently, a slightly changed PPP split-ratio was found. However, due to the small size of
the glucose-oxidation flux, the observed effect on the PPP split-ratio was marginal.
Table A-I Estimated oxidative PPP flux (v3) and the corresponding PPP split-ratio for a stoichiometric
model containing either the glucose-oxidation (v6) or the 6PG-dephosphorylation (v8) flux.
Metabolic fluxes (mmol/Cmol/h)
Pathway
Oxidation of glucose
PPP split-ratio
SSres
(%)
v1
v2
v3
v6
v7
8.60
0.42
4.30
0.13
0.55
-
50.8a
3.48
0.57
0.15
51.7
2.59
Dephosphorylation
8.60
0.42
4.45
of 6PG
a
Normalized for the uptake-rate of glucose (v1 minus v6).
v8
151
Chapter 5
APPENDIX B: Confidence interval of SSres
The asymmetric confidence interval of parameter (β) in a non-linear model can be derived by
refitting the measurement data for a fixed range of values for β. By plotting the minimized sum
of squared residuals ( SS(β) ) for each fit as a function of β, an asymmetrical curve is obtained
from which the confidence interval can be calculated by defining a cut-off value for the SS(β)
using the relation [1]:
(SS(β) − SS(βˆ ))
 F(1,n − p)
SS(βˆ ) /(n − p)
(B-1)
where p equals the total number of parameters in the model and SS(βˆ ) is a constant
corresponding to the global minimized sum of squared residuals (determined when β is not
fixed). The cut-off value for SSres at a chosen confidence level (α) is obtained by solving Eq.
B-1 for SS(β) :
SS(β) ≤ SS(βˆ ) +
SS(βˆ )
⋅ Fα (1,n − p)
n−p
(B-2)
ˆ
Note that the term SS(β) corresponds with the estimator for the measurement error ( σˆ 2 ).
n−p
152
13
C-Labeled gluconate tracing: method
REFERENCES
[1] Bates, D.M. and Watts, D.G. (1988) Nonlinear
regression analysis and its applications. John Wiley
and Sons, Inc., New York.
profiling of Escherichia coli mutants in central
carbon metabolism using GC-MS. Eur. J. Biochem.
270, 880-891.
[2] Caspari, T. and Urlinger, S. (1996) The activity of the
gluconate-H+ symporter of Schizosaccharomyces
pombe cells is down-regulated by D-glucose and
exogenous cAMP. FEBS Lett. 395, 272-276.
[12] Follstad, B.D. and Stephanopoulos, G. (1998) Effect
of reversible reactions on isotope label redistribution
- Analysis of the pentose phosphate pathway. Eur. J.
Biochem. 252, 360-371.
[3] Christensen, B., Christiansen, T., Gombert, A.K.,
Thykaer, J. and Nielsen, J. (2001) Simple and
robust method for estimation of the split between
the oxidative pentose phosphate pathway and
the
Embden-Meyerhof-Parnas
pathway
in
microorganisms. Biotechnol. Bioeng. 74, 517-523.
[13] Gombert, A.K., dos Santos, M.M., Christensen,
B. and Nielsen, J. (2001) Network identification
and flux quantification in the central metabolism
of Saccharomyces cerevisiae under different
conditions of glucose repression. J. Bacteriol. 183,
1441-1451.
[4] Christensen, B. and Nielsen, J. (2000) Metabolic
network analysis of Penicillium chrysogenum using
13
C-labeled glucose. Biotechnol. Bioeng. 68, 652659.
[5] Christensen, B., Thykaer, J. and Nielsen, J. (2000)
Metabolic characterization of high- and low-yielding
strains of Penicillium chrysogenum. Appl. Microb.
Biot. 54, 212-217.
[14] Haas, H., Redl, B., Leitner, E. and Stoffler, G.
(1991) Penicillium chrysogenum extracellular
acid phosphatase - Purification and biochemical
characterization. Biochim. Bioph. Acta 1074, 392397.
[6] [7] [8] [9] Dauner, M., Bailey, J.E. and Sauer, U. (2001)
Metabolic flux analysis with a comprehensive
isotopomer model in Bacillus subtilis. Biotechnol.
Bioeng. 76, 144-156.
Dauner, M., Sonderegger, M., Hochuli, M.,
Szyperski, T., Wuthrich, K., Hohmann, H.P., Sauer,
U. and Bailey, J.E. (2002) Intracellular carbon fluxes
in riboflavin-producing Bacillus subtilis during growth
on two-carbon substrate mixtures. Appl. Environ.
Microb. 68, 1760-1771.
Delgado, T.C., Castro, M.M., Geraldes, C.F. and
Jones, J.G. (2004) Quantitation of erythrocyte
pentose pathway flux with [2-13C]Glucose and 1H
NMR analysis of the lactate methyl signal. Magnet.
Reson. Med. 51, 1283-1286.
Drysch, A., El Massaoudi, M., Wiechert, W., de
Graaf, A.A. and Takors, R. (2004) Serial flux
mapping of Corynebacterium glutamicum during
fed-batch L-lysine production using the sensor
reactor approach. Biotechnol. Bioeng. 85, 497-505.
[10] Egli, T., Lendenmann, U. and Snozzi, M. (1993)
Kinetics of microbial growth with mixtures of carbon
sources. Anton. Leeuw. Int. J. G. 63, 289-298.
[11] Fischer, E. and Sauer, U. (2003) Metabolic flux
[15] Harris, D.M., Diderich, J.A., van der Krogt, Z.A.,
Luttik, M.A.H., Raamsdonk, L.M., Bovenberg,
R.A.L., van Gulik, W.M., van Dijken, J.P. and
Pronk, J.T. (2006) Enzymic analysis of NADPH
metabolism in [beta]-lactam-producing Penicillium
chrysogenum: Presence of a mitochondrial NADPH
dehydrogenase. Metab. Eng. 8, 91-101.
[16] Henriksen, C.M., Christensen, L.H., Nielsen, J.
and Villadsen, J. (1996) Growth energetics and
metabolic fluxes in continuous cultures of Penicillium
chrysogenum. J. Biotechnol. 45, 149-164.
[17] Izu, H., Adachi, O. and Yamada, M. (1997) Gene
organization and transcriptional regulation of the
gntRKU operon involved in gluconate uptake and
catabolism of Escherichia coli. J. Mol. Biol. 267,
778-793.
[18] Katz, J., Abraham, S., Hill, R. and Chaikoff, I.L.
(1955) The occurence and mechanism of the
hexose monophosphate dhunt in rat liver slices. J.
Biol. Chem. 214, 853-868.
[19] Katz, J. and Rognstad, R. (1967) Labeling of
pentose phosphate from glucose-14C and estimation
of rates of transaldolase transketolase contribution
of pentose cycle and ribose phosphate synthesis.
Biochemistry 6, 2227-2247.
[20] Kiefer, P., Heinzle, E., Zelder, O. and Wittmann,
C. (2004) Comparative metabolic flux analysis of
lysine-producing Corynebacterium glutamicum
153
Chapter 5
cultured on glucose, or fructose. Appl. Environ.
Microb. 70, 229-239.
[21] Kingsley-Hickman, P.B., Ross, B.D. and Krick, T.
(1990) Hexose monophosphate shunt measurement
in cultured cells with [1-13C]glucose: Correction for
endogenous carbon sources using [6-13C]glucose.
Anal. Biochem. 185, 235-237.
[22] Kleijn, R.J., van Winden, W.A., van Gulik, W.M.
and Heijnen, J.J. (2005) Revisiting the 13C-label
distribution of the non-oxidative branch of the
pentose phosphate pathway based upon kinetic and
genetic evidence. FEBS J. 272, 4970-4982.
[23] Lange, H.C., Eman, M., van Zuijlen, G., Visser,
D., van Dam, J.C., Frank, J., de Mattos, M.J. and
Heijnen, J.J. (2001) Improved rapid sampling
for in vivo kinetics of intracellular metabolites in
Saccharomyces cerevisiae. Biotechnol. Bioeng. 75,
406-415.
[24] Lee, H.W., Pan, J.G. and Lebeault, J.M. (1998)
Enhanced L-lysine production in threonine-limited
continuous culture of Corynebacterium glutamicum
by using gluconate as a secondary carbon source
with glucose. Appl. Microb. Biot. 49, 9-15.
154
[25] Lee, W.N.P., Boros, L.G., Puigjaner, J., Bassilian, S.,
Lim, S. and Cascante, M. (1998) Mass isotopomer
study of the nonoxidative pathways of the pentose
cycle with [1,2-13C2] glucose. Am. J. Physiol-Endoc.
M. 37, E843-E851.
[26] Maaheimo, H., Fiaux, J., Cakar, Z.P., Bailey, J.E.,
Sauer, U. and Szyperski, T. (2001) Central carbon
metabolism of Saccharomyces cerevisiae explored
by biosynthetic fractional 13C labeling of common
amino acids. Eur. J. Biochem. 268, 2464-2479.
[27] Mashego, M.R., van Gulik, W.M., Vinke, J.L. and
Heijnen, J.J. (2003) Critical evaluation of sampling
techniques for residual glucose determination in
carbon-limited chemostat culture of Saccharomyces
cerevisiae. Biotechnol. Bioeng. 83, 395-399.
[28] Mashego, M.R., Wu, L., van Dam, J.C., Ras, C.,
Vinke, J.L., van Winden, W.A., van Gulik, W.M. and
Heijnen, J.J. (2004) MIRACLE: mass isotopomer
ratio analysis of U-13C-labeled extracts. A new
method for accurate quantification of changes
in concentrations of intracellular metabolites.
Biotechnol. Bioeng. 85, 620-628.
[29] Politino, M., Brown, J. and Usher, J.J. (1996)
Purification and characterization of an extracellular
alkaline phosphatase from Penicillium chrysogenum.
Prep. Biochem. Biotech. 26, 171-181.
[30] Pronk, J.T., Steensma, H.Y. and vanDijken, J.P.
(1996) Pyruvate metabolism in Saccharomyces
cerevisiae. Yeast 12, 1607-1633.
[31] Sauer, U., Hatzimanikatis, V., Bailey, J.E., Hochuli,
M., Szyperski, T. and Wuthrich, K. (1997) Metabolic
fluxes in riboflavin-producing Bacillus subtilis. Nat.
Biotechnol. 15, 448-452.
[32] Sawyer, D.T. and Bagger, J.B. (1959) The lactone
acid salt equilibria for D-glucono-delta-lactone and
the hydrolysis kinetics for this lactone. J. Am. Chem.
Soc. 81, 5302-5306.
[33] van Dijken, J.P., van Tuijl, A., Luttik, M.A.H.,
Middelhoven, W.J. and Pronk, J.T. (2002) Novel
pathway for alcoholic fermentation of deltagluconolactone in the yeast Saccharomyces bulderi.
J. Bacteriol. 184, 672-678.
[34] van Gulik, W.M., de Laat, W.T., Vinke, J.L. and
Heijnen, J.J. (2000) Application of metabolic
flux analysis for the identification of metabolic
bottlenecks in the biosynthesis of penicillin-G.
Biotechnol. Bioeng. 68, 602-618.
[35] van Winden, W.A., van Gulik, W.M., Schipper,
D., Verheijen, P.J.T., Krabben, P., Vinke, J.L. and
Heijnen, J.J. (2003) Metabolic flux and metabolic
network analysis of Penicillium chrysogenum
using 2D [13C, 1H] COSYNMR measurements
and cumulative Bondomer simulation. Biotechnol.
Bioeng. 83, 75-92.
[36] van Winden, W.A., Wittmann, C., Heinzle, E. and
Heijnen, J.J. (2002) Correcting mass isotopomer
distributions for naturally occurring isotopes.
Biotechnol. Bioeng. 80, 477-479.
[37] van Winden, W.A., van Dam, J.C., Ras, C., Kleijn,
R.J., Vinke, J.L., van Gulik, W.M. and Heijnen, J.J.
(2005) Metabolic-flux analysis of Saccharomyces
cerevisiae CEN.PK113-7D based on mass
isotopomer measurements of 13C-labeled primary
metabolites. FEMS Yeast Res. 5, 559-568.
[38] Vanderheijden, R.T.J.M., Heijnen, J.J., Hellinga,
C., Romein, B. and Luyben, K.C.A.M. (1994)
Linear constraint relations in biochemical reaction
systems .1. Classification of the calculability and
the balanceability of conversion rates. Biotechnol.
Bioeng. 43, 3-10.
[39] Wiechert, W., Mollney, M., Isermann, N., Wurzel,
M. and de Graaf, A.A. (1999) Bidirectional reaction
steps in metabolic networks: III. Explicit solution and
analysis of isotopomer labeling systems. Biotechnol.
Bioeng. 66, 69-85.
13
C-Labeled gluconate tracing: method
[40] Wu, L., Mashego, M.R., van Dam, J.C., Proell, A.M.,
Vinke, J.L., Ras, C., van Winden, W.A., van Gulik,
W.M. and Heijnen, J.J. (2005) Quantitative analysis
of the microbial metabolome by isotope dilution
mass spectrometry using uniformly 13C-labeled cell
extracts as internal standards. Anal. Biochem. 336,
164-171.
155
CHAPTER
6
Cytosolic NADPH metabolism
in penicillin-G producing and
non producing chemostat cultures
of Penicillium chrysogenum
Roelco J. Kleijn, Feng Liu, Wouter A. van Winden,
Walter M. van Gulik, Cor Ras
and Joseph J. Heijnen
Based upon:
Metabolic Engineering
(2007) 9 (1):112-123
Chapter 6
ABSTRACT
This study addresses the relation between NADPH supply and penicillin synthesis, by
comparing the flux through the oxidative branch of the pentose phosphate pathway (the main
source of cytosolic NADPH) in penicillin-G producing and non-producing chemostat cultures
of Penicillium chrysogenum. The fluxes through the oxidative part of the pentose phosphate
pathway (PPP) were determined using the recently introduced gluconate-tracer method.
Significantly higher oxidative PPP fluxes were observed in penicillin-G producing chemostat
cultures, indicating that penicillin production puts a major burden on the supply of cytosolic
NADPH. To our knowledge this is the first time direct experimental proof is presented for
the causal relationship between penicillin production and NADPH-supply. Additional insight
in the metabolism of P. chrysogenum was obtained by comparing the PPP fluxes from the
gluconate-tracer experiment to oxidative PPP fluxes derived via metabolic flux analysis, using
different assumptions for the stoichiometry of NADPH consumption and production.
158
13
C-Labeled gluconate tracing: application
6.1 INTRODUCTION
In the ongoing quest for strains with higher product titers and yields, metabolic engineering
has allowed for a more rational approach to strain improvement [18]. At the level of cellular
metabolism, techniques such as metabolic network and flux analysis provide metabolic
engineers with a better understanding of the metabolic network with respect to stoichiometry
and flux distribution [1, 5, 10, 21]. This knowledge enables a more directed genetic modification
of industrial micro-organisms through recombinant DNA technology, thereby increasing the
chance of successfully directing the metabolic fluxes towards the desired end product.
Metabolic flux analysis (MFA) is frequently applied to study microorganisms that either
overproduce primary metabolites (e.g. ethanol, glycerol, lactate) [17, 25] or products closely
related to the primary metabolism (e.g. amino acids) [10]. MFA can, however, also be used to
study the metabolism of micro-organisms applied for secondary metabolite production, thereby
paying special attention to the interconnectivity between the primary and secondary metabolic
pathways. In wild-type microorganisms secondary metabolites are normally produced at levels
that are several orders of magnitude lower than the primary metabolites. For this reason,
classical strain improvements strategies have primarily focused on increasing the levels of
enzymes in the product pathway. However, continuous amplification of the enzyme levels in
the secondary metabolism will, at some point, lead to a shift of the metabolic bottleneck from
secondary metabolism towards primary metabolism. Particularly in high-producing strains
the role of primary metabolism in supplying the carbon precursors, cofactors and energy for
product formation should not be underestimated. A well known example is the filamentous
fungus Penicillium chrysogenum, of which the β-lactam antibiotic productivity has been
increased with several orders of magnitude as a result of more than 50 years of classical
strain improvement. A detailed review on the role of primary metabolism in the production of
antibiotics is given by Gunnarsson et al. [6].
In general, three potential limitations in the primary metabolism of P. chrysogenum can be
identified when synthesizing large amounts of β-lactam antibiotics (e.g. penicillin-G and
penicillin-V). These are the supply of the three precursor amino acids (α-aminoadipate, valine
and cysteine); the availability of electrons in the form of NADPH; and the supply of energy in
the form of ATP. Note that strictly speaking α-aminoadipate is not a precursor of penicillin as it
is split-off and (partially) recycled after the formation of the β-lactam nucleus. Several studies
have addressed one or more of these potential bottlenecks by metabolic flux analysis [3, 8, 9,
19, 20]. However, uncertainties in the metabolism of P. chrysogenum have resulted in different
assumptions with respect to the applied stoichiometric models for metabolic flux analysis,
making it difficult to unambiguously draw conclusions on potential metabolic bottlenecks. Van
Gulik et al. [19], for example, estimated from experimental data a much higher ATP consumption
associated with penicillin production than the theoretical value used by Jorgensen et al. [9],
resulting in a much lower theoretical maximum yield of penicillin-G on glucose (0.18 mol/mol
glucose versus 0.50 mol/mol glucose, respectively). Similarly, Jorgensen et al. [9] observed
that the supply of the three precursor amino acids α-aminoadipate, valine and cysteine to
a fed-batch cultivation increased the penicillin-V production by about 20%, indicating that
precursor availability may limit penicillin production at high rates. On the other hand, van Gulik
et al. [20] concluded for a different P. chrysogenum strain that primary carbon metabolism was
159
Chapter 6
unlikely to form a potential bottleneck in penicillin-G synthesis, based upon the flexibility of
four principal metabolic nodes.
The supply of sufficient reducing power in the form of NADPH is especially important for the
biosynthesis of the two amino acid precursors of the β-lactam nucleus, cysteine and valine.
Using stoichiometric models for the primary and secondary metabolism of P. chrysogenum,
several groups have speculated that the flux through the oxidative branch of the pentose
phosphate pathway (PPP), being the main source of cytosolic NADPH in P. chrysogenum, is
strongly correlated to β-lactam production [8, 9]. This hypothesis was supported by the results
of van Gulik et al. [20], who showed that a stepwise increase in the total metabolic demand for
NADPH resulted in a stepwise decrease of the penicillin-G production.
In the past, the flux through the oxidative branch of the PPP has often been estimated via 13Clabeling based MFA. Recently, a new method has been developed for estimating the oxidative
PPP flux and the PPP split-ratio (the fraction of glucose taken up by the cell that enters the
oxidative branch of the PPP) [11]. This method is based on the co-feeding of trace amounts
of 13C labeled gluconate and produces more accurate estimates for the oxidative PPP flux
than the 13C-labeling based MFA method. In this study the gluconate-tracer method is applied
to measure the PPP split-ratio, under penicillin-G producing and non-producing conditions
in carbon-limited chemostat cultures of P. chrysogenum at two different specific growth
rates. By comparing the estimated PPP split-ratios with those predicted from stoichiometric
models based upon the recently found NADPH-specificity of various enzymes in the primary
metabolism of P. chrysogenum [7], the role of NADPH in the biosynthesis of β-lactam antibiotics
is further investigated.
160
6.2 MATERIALS AND METHODS
6.2.1
Strain
A high-yielding industrial P. chrysogenum strain (code name DS17690) was kindly donated by
DSM Anti-Infectives (Delft, The Netherlands).
6.2.2
Cultivation conditions
P. chrysogenum was cultivated in a carbon-limited chemostat system at two different dilution
rates (0.020 h-1 and 0.052 h-1), both in the absence and presence of phenylacetic acid (PAA),
the side-chain precursor for penicillin-G biosynthesis. All four cultivations were carried out in
a 1 L (working volume 0.6 L) bioreactor (Applikon, Schiedam, The Netherlands) equipped
with a 6-bladed Rusthon turbine stirrer (D = 45 mm). The stirrer speed was 600 rpm and the
aeration rate was 11 L/h (0.30 vvm). Under these conditions the dissolved oxygen tension was
always above 50%. Temperature was controlled at 25oC, the pH at 6.5 and the head space
overpressure at 0.1 bar. Effluent was removed discontinuously via an overflow tube by means
of a peristaltic pump which was operated at fixed time intervals (1 min every 10 min). Silicone
antifoam agent (BDH, Poole, UK) was diluted 1:10 (v/v) with demineralized H2O and fed to the
reactor at a fixed rate of 0.2 mL/h.
The batch phase was initiated by inoculating the reactor with spores from 2.0 g of rice grains.
During the germination phase of the spores (first 30 hours), the reactor was run at a stirrer
13
C-Labeled gluconate tracing: application
speed of 100 rpm and an aeration rate of 1 L/h. The sole carbon source during the batch
phase was 3.3 g/L of glucose.H2O. The end of the batch phase was typically reached after 50
h, after which the reactor was switched to continuous mode. The continuous mode consisted
of two phases; during the first phase P. chrysogenum was grown on an unlabeled medium for
2 residence times, followed by at least 3 residence times of feeding on a chemically identical
medium, but containing uniformly labeled [u-13C]glucono-δ-lactone (Omicron, South Bend,
USA) instead of naturally labeled glucono-δ-lactone.
6.2.3
Media
Cells were grown on a medium specifically developed for the gluconate-tracer method [11],
containing: 3.30 g/L glucose.H2O, 0.15 g/L glucono-δ-lactone, 0.68 g/L sodium-acetate.3H2O,
0.35 g/L KH2PO4, 1.54 g/L (NH4)2SO4, 0.22 g/L MgSO4.7H2O and 0.90 mL/L trace element
solution. The trace element solution contained 75 g/L Na2-EDTA.2H2O, 2.5 g/L CuSO4.5H2O,
10 g/L ZnSO4.7H2O, 10 g/L MnSO4.H2O, 20 g/L FeSO4.7H2O, and 2.5 g/L CaCl2.2H2O. The
small amount of sodium-acetate was added to the medium to compare these experiments
with previously performed 13C-labeling experiments in which acetate was added for a better
estimation of the fluxes in the lower part of the metabolism.
In two of the four chemostat experiments phenylacetic acid (PAA) was added to the medium
to induce the production of penicillin-G. Depending on the applied specific growth rate (0.020
h-1 or 0.052 h-1), the medium of these experiments contained 0.533 g/L and 0.461 g/L PAA,
respectively. The PAA-level in the medium was chosen such that the steady state concentration
in the chemostat was approximately 3 mM. At this concentration PAA was neither limiting for
penicillin-G production nor inhibiting for the growth of P. chrysogenum.
For the preparation of the minimal medium the appropriate amount of PAA was dissolved in
demineralized water, adjusted to a pH of 5.40 using 1 M KOH and autoclaved at 121oC for 40
min. The remaining medium components were filter-sterilized using an Acropak20 filter (PALL,
East-Hills, NY, USA) and added to the PAA solution. After preparation the minimal medium
was stirred for at least 12 hours on a magnetic stirrer to ensure that all the added glucono-δlactone was hydrolyzed to gluconate [16].
6.2.4Broth sampling and dry weight determination
Duplicate 10 mL samples were withdrawn from the bioreactor for determining the biomass
dry weight. Samples were filtered over preweighted glass fiber filters (PALL, East-Hills, NY,
USA) and dried at 70oC for at least 24 h. The collected filtrate was immediately frozen in liquid
nitrogen and stored at -80oC prior to analysis.
6.2.5
Rapid sampling, quenching and metabolite extraction
Extracellular samples were acquired by rapidly sampling 2 mL of broth into a syringe containing
precooled stainless steel beads (-18oC), immediately followed by separation of cells and
medium by filtration as described by Mashego et al. [13]. Samples for intracellular metabolite
determinations were obtained by rapidly withdrawing 1 mL of broth from the bioreactor,
followed by direct injection of the sample in 5 mL of a 60% (v/v) methanol/water mixture
(‑40oC) for instantaneous quenching of the cell metabolism as described by Lange et al. [12].
161
Chapter 6
Intracellular metabolites were extracted from these samples as described previously [11, 14].
6.2.6
Analytical procedures
The mass isotopomer distributions glucose-6-phosphate (G6P), 6-phospho-gluconate (6PG)
and gluconic acid (gln) were measured by LC-MS as described by van Winden et al. [23].
The concentrations of glucose, gluconate and acetate in the medium and the bioreactor were
determined enzymatically (Enzytec, Scil Diagnostics, Viernheim, Germany).
The concentration of PAA, penicillin-G, ortho-hydroxyphenylacetic acid (o-OH-PAA), 6oxopiperidine-2-carboxylic acid (OPC), isopenicillin-N (IPN), 8-hydroxypenillic acid (8-HPA),
6-amino-penicillanic acid (6-APA) and penicilloic acid (PIO) in the filtrate samples was
analyzed at 27oC at 600 MHz on a Bruker Avance 600 nuclear magnetic resonance (NMR)
spectrometer equipped with an inverse triple-resonance cryoprobe and a pulse field gradient
system. Samples were prepared for analysis by combining 1.0 mL of the filtrate with 0.5 mL of
standard solution, containing an exactly known concentration of maleic acid for quantification
and EDTA to chelate the paramagnetic metal ions. The combined sample was lyophilized
and redissolved in 600 µl D2O. For each sample the 1H-NMR spectrum was recorded using
a relaxation delay of 30 seconds, ensuring full relaxation of all the hydrogen atoms between
pulses. The integrals of the characteristic resonances for each component and the internal
standard (singlet at 6.1 ppm) were measured, and the contents of the individual components
were calculated.
162
6.2.7
Gluconate-tracer method
The gluconate-tracer method was specifically designed for the accurate determination of the
PPP split-ratio and has been described in detail by Kleijn et al. [11]. In short, the method
is based upon the simultaneous feeding of unlabeled glucose and a trace amount of [u13
C]gluconate to a carbon-limited chemostat cultivation (Figure 1). Once isotopic steady state
is reached, the mass isotopomer distributions of the intracellular metabolites surrounding the
6PG-node and the gluconate-uptake rate are measured. Substitution of these measurements
in the mass isotopomer balances of 6PG (Eq.6.1) produces an over-determined system from
which the oxidative PPP flux (v2) is calculated using a sequential quadratic programming
algorithm. By normalizing v2 with the uptake rate of glucose, the PPP split-ratio is obtained.
 m + 0 
 m + 0 
m + 0 
m + 0 




 


 
m + 1 
 m + 1
 m + 1 
 m + 1 
v1  
−
=
v

−
2 
  
   
  
   




 


 
 m + 6 gln  m + 6 6pg 
 m + 6 6pg  m + 6 g6p 
(6.1)
Using the gluconate-tracer method the PPP split-ratio of P. chrysogenum was calculated for
four different chemostat-cultivations. For an accurate calculation of the PPP split-ratio the
measured mass isotopomers were corrected for naturally occurring isotopes of hydrogen and
oxygen [22]. To obtain a measure for the PPP split-ratio in a chemostat grown solely on
glucose, the estimated PPP split-ratios were corrected: (i) for the additional influx of gluconate
and (ii) for the fact that per mol of glucose catabolized in the oxidative branch of the PPP,
13
C-Labeled gluconate tracing: application
two mol of NADPH are produced, while per mol of catabolized gluconate only one mole of
NADPH is produced. The latter correction was based on the assumption that the PPP flux
was directly proportional to the cytosolic NADPH demand. Note that the first correction is for
an underestimation of the PPP, while the second correction is for an overestimation of the
PPP, thereby resulting in only minor changes in the final PPP split-ratio. The 95% confidence
intervals of the estimated PPP split-ratios were determined by Monte Carlo simulation in which
the added noise was normally distributed with the measured standard deviations of the mass
isotopomer fractions.
n-12C6
glc
u-13C6
gln
Figure 1 Schematic diagram of the
metabolic network surrounding the G6Pnode when growing cells simultaneously
on glucose (glc) and gluconate (gln).
v1
g6p
glycolysis
v2
6pg
v3
PPP
P. chrysogenum
6.2.8
Metabolic flux analysis
Metabolite balancing in combination with a previously described stoichiometric model for
growth and product formation of P. chrysogenum [20] was used to estimate intracellular
metabolic fluxes. Calculations were carried out with the software package MNAv3.0 (SpadIT,
Nijmegen, the Netherlands). Four versions of the stoichiometric model were constructed using
different assumptions for NADPH consumption and production (see results section). MFA was
carried out using the measured biomass specific rates of glucose, gluconate, acetate, CO2,
PAA, penicillin-G, o-OH-PAA, PIO, 8-HPA, 6-APA and OPC, which were calculated from the
average steady state values of the measured flows and concentrations. For each data set the
full covariance matrix associated with the calculated conversion rates was used in the flux
balancing procedure, which was carried out as described by van Gulik et al. [20]. The number
of available rates was sufficient to result in an overdetermined and thus solvable system.
6.3 RESULTS AND DISCUSSION
6.3.1
Macroscopic data
P. chrysogenum was cultivated in a carbon-limited chemostat system at a dilution rate of
0.020 h-1 and 0.052 h-1, both in the absence and presence of PAA. Steady-state was assumed
after 4 residence times. Steady-state biomass concentrations and calculated recoveries of
carbon, degree of reduction and PAA for the four cultivations are presented in Table 1-A.
Measured biomass specific conversion rates in steady-state are shown in Table 1-B. In all
163
Chapter 6
four experiments the carbon recovery was close to 100%, while the degree of reduction
recovery was significantly higher than 100%. Furthermore, the calculated respiratory quotients
(RQ) where all well below the expected value of 1.07 for all four fermentations, indicating
an error in the measured biomass specific O2 consumption rate. Data reconciliation under
the constraint of the conservation relations for the elements and gross error detection was
carried out according to Vanderheijden et al. [24] and confirmed this error. Apart from the O2
consumption rate, no other significant discrepancies were observed between the measured
and the reconciled conversion rates (see Table 1-B). Note that for all four cultivations the
gluconate uptake-rate was approximately 5% (w/w) of the glucose uptake-rate.
Table 1-A Steady-state biomass concentrations and recovery balances for the four gluconate-tracer
experiments. Chemostat cultivations were performed under penicillin-G producing and non-producing
conditions at two different growth-rates.
Biomass
concentration and
recovery balances
164
Unit
µ = 0.020 h-1
Producing
µ = 0.052 h-1
Non-producing
Producing
Non-producing
Biomass concentration g/L
1.04±0.05
1.25±0.05
1.60±0.09
1.88±0.02
Carbon recovery
(%)
104.0±2.0
94.4±3.0
97.1±4.2
106.9±2.2
Degree of reduction
(%)
120.0±4.4
128.2±7.0
108.3±6.5
114.6±1.7
PAA recovery
(%)
114.7±33.5
-
100.9±5.3
-
Figure 2 displays the measured specific consumption rate of PAA and the specific production
rate of penicillin-G, β-lactam intermediates (8-HPA, 6-APA, IPN) and side-products (OPC and
o-OH-PAA) during the course of the four chemostat cultivations. As expected no penicillin-G
formation was observed in the absence of the side-chain precursor PAA. However, the absence
of PAA did not completely halt the synthesis of β-lactam intermediates as small amounts of IPN,
6-APA and 8‑HPA (the carboxylated form of 6‑APA) were produced. Interestingly, the formation
of penicillin-G intermediates seems to be growth rate dependent. At a specific growth rate of
0.020 h‑1 the amount of β-lactam compounds produced in the absence of PAA accounted for
13% of the total β-lactam produced under producing conditions, while this value increased to
28% at a specific growth rate of 0.052 h-1. The measured specific penicillin-G production rates
at dilution rates of 0.020 h‑1 and 0.052 h-1 corresponded well with the penicillin-G production
rates reported previously by van Gulik et al. [20] for this P. chrysogenum strain.
From Table 1-B it can be seen that the presence of penicillin-G production coincides with a
lower mycelium concentration and increased biomass specific conversion rates of substrates,
oxygen and carbon dioxide. These observations are indicative of the burden imposed by
penicillin-G production on the catabolic activity of the cell (formation of ATP en NADPH).
Conversely, the absence of penicillin-G production decreases the catabolic demand of the
cell in favor of the assimilation of biomass, thereby increasing the steady state mycelium
concentration in the chemostat.
C-Labeled gluconate tracing: application
13
µ = 0.020 h-1
µ = 0.052 h-1
8.60±0.40
Measured
2.66
0.42
8.25
Reconciled
2.30±0.10
0.35±0.02
7.14±0.29
Measured
2.28
0.35
6.72
Reconciled
0.49±0.03
47.47±3.65
4.40±0.26
0.69±0.04
13.91±0.81
0.52
37.37
4.41
0.69
14.23
Measured Reconciled
<0.01
37.68±1.12
3.74±0.04
0.58±0.01
11.81±0.12
Measured
0.00
26.23
3.74
0.58
11.85
Reconciled
Non-producing (p ≥ 0.73)a
0.42±0.02
0.00
23.15
Producing (p ≥ 0.26)a
Glucose consumption
2.67±0.09
<0.01
40.90±2.94
Non-producing (p ≥ 0.08)a
Gluconate consumption
0.57
30.48
Producing (p ≥ 0.66)a
Acetate consumption
0.53±0.12
48.91±2.10
Biomass specific
conversion rates
(mmol·CmolX-1·h-1)
Oxygen consumptionb
Carbondioxide production
20.30±1.00
31.34±1.57
0.00
0.48
20.67
32.26
0.02±0.00
<0.01
20.26±1.00
22.60±0.93
0.02
0.00
21.98
24.31
0.01±0.00
0.49±0.02
52.29±1.00
41.00±1.65
0.01
0.52
52.21
40.78
0.04±0.00
<0.01
52.18±1.00
30.11±0.31
0.04
0.00
51.55
28.86
PAA consumption
Biomass production
<0.01
0.48±0.06
Penicillin-G & PIO production
<0.01
0.00
<0.01
0.00
<0.01
<0.01
0.02±0.01
0.00
0.00
0.02
<0.01
0.10±0.01
0.01±0.00
0.00
0.10
0.01
IPN production
6-APA production
0.04
0.07
0.00
0.07±0.01
<0.01
0.14
0.04±0.01
0.14±0.01
0.00
0.01
0.09
0.01±0.00
<0.01
0.08
0.09±0.01
o-OH-PAA production
0.08±0.01
8-HPA production
OPC production
a
p‑values were determined via a χ 2 ‑distribution and denote the probability that the discrepancy between the measured and the reconciled conversion
rates is a result of measurement error. P-values ≤ 0.05 were considered statistically significant, thus indicating a clear deviation between the measured
and reconciled conversion rates.
The erroneously measured O2 consumption rate was omitted when reconciling the conversion rates.
b
Table 1-B Measured and reconciled biomass specific conversion rates for the four gluconate-tracer experiments in steady-state. Chemostat cultivations
were performed under penicillin-G producing and non-producing conditions at two different growth rates.
165
µ = 0.020 h-1
Conversion rate
(mmol Cmol-1h-1)
µ = 0.052 h-1
Conversion rate
(mmol Cmol-1h-1)
0
2e-4
4e-4
6e-4
0
2e-4
4e-4
6e-4
0
0
C
A
1
1
3
4
2
Residence Time (-)
3
4
2
Residence Time (-)
Penicillin-G production
5
5
6
6
0
2e-5
4e-5
6e-5
8e-5
1e-4
0
3e-5
6e-5
9e-5
12e-5
Conversion rate
(mmol Cmol-1h-1)
Conversion rate
(mmol Cmol-1h-1)
166
Figure 2 Biomass specific
conversion
rates
of
penicillin-G, IPN, 8-HPA,
6-APA, OPC, PAA and oOH-PAA during the course
of the four gluconate-tracer
experiments. Conversion
rates were derived from the
measured concentrations
of the components in
the filtrate. The dashed
line represents the point
at
which
steady-state
was assumed (about 4
residence times).
0
0
D
B
1
1
5
3
4
2
5
Residence Time (-)
3
4
2
Residence Time (-)
6
6
-qPAA
qPenG
q6-APA
q8-HPA
qIPN
qOPC
qo-OH-PAA
No penicillin-G production
Chapter 6
13
C-Labeled gluconate tracing: application
6.3.2
PPP split-ratio determination
Biomass samples for intracellular metabolite determination were harvested from the four
chemostat cultivations after 3 residence times of feeding on a medium with [u‑13C]gluconate.
Table 2 shows the measured mass isotopomer fractions of the gluconate added to the medium
and the intracellular metabolites G6P, 6PG and gln.
Table 2 shows that for all four chemostat experiments the measured mass isotopomer
distributions of intracellular gluconate and the gluconate present in the medium are different,
especially with respect to the m+0 and m+6 fractions, indicating that an unidentified reaction
Table 2 Measured mass isotopomer fractions and standard deviations for the four gluconate-tracer
experiments. Presented mass fractions have been corrected for the natural isotopes of the elements
hydrogen and oxygen as described in the materials and methods section.
µ=0.020
Measured Mass
Metabolite fraction Penicillin-G
production
Gln in
mediuma
Glnb
G6Pb
6PGb
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
m+0
m+1
m+2
m+3
m+4
m+5
m+6
0.000
0.000
0.000
0.000
0.002
0.059
0.939
0.217
0.019
0.003
0.001
0.002
0.040
0.718
0.866
0.094
0.017
0.015
0.007
0.001
0.001
0.786
0.082
0.015
0.011
0.016
0.008
0.081
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.000
0.000
0.000
0.000
0.000
0.001
0.001
0.006
0.001
0.001
0.000
0.001
0.003
0.007
0.003
0.002
0.002
0.000
0.000
0.000
0.000
0.003
0.003
0.001
0.000
0.003
0.001
0.003
µ=0.052
No penicillin-G
production
Penicillin-G
production
No penicillin-G
production
0.000
0.000
0.000
0.000
0.002
0.059
0.939
0.257
0.022
0.002
0.003
0.006
0.043
0.669
0.873
0.084
0.014
0.015
0.009
0.002
0.002
0.772
0.073
0.012
0.011
0.009
0.011
0.111
0.000
0.000
0.000
0.000
0.002
0.059
0.939
0.100
0.010
0.002
0.001
0.002
0.052
0.833
0.864
0.095
0.019
0.014
0.008
0.001
0.001
0.783
0.084
0.017
0.010
0.011
0.010
0.086
0.000
0.000
0.000
0.000
0.002
0.059
0.939
0.096
0.010
0.002
0.001
0.002
0.053
0.836
0.860
0.093
0.020
0.016
0.009
0.001
0.001
0.767
0.082
0.017
0.013
0.008
0.011
0.103
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.000
0.000
0.000
0.000
0.000
0.001
0.001
0.005
0.003
0.001
0.001
0.001
0.004
0.005
0.001
0.001
0.000
0.000
0.000
0.000
0.000
0.003
0.001
0.000
0.000
0.001
0.000
0.002
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.000
0.000
0.000
0.000
0.000
0.001
0.001
0.003
0.001
0.000
0.000
0.000
0.002
0.005
0.002
0.002
0.001
0.000
0.000
0.000
0.000
0.002
0.001
0.001
0.001
0.001
0.001
0.003
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
0.000
0.000
0.000
0.000
0.000
0.001
0.001
0.005
0.001
0.001
0.000
0.000
0.003
0.007
0.003
0.003
0.001
0.000
0.000
0.000
0.000
0.002
0.002
0.001
0.000
0.001
0.001
0.001
The mass isotopomer fractions of pure gluconate were measured once. The same gluconate was used
for all four gluconate-tracer experiments.
b
Mass isotopomer fractions of the intracellular metabolite.
a
167
Chapter 6
causes an inflow of unlabeled carbon into the otherwise uniformly 13C-labeled gluconate
pool. Kleijn et al. [11] also observed this phenomenon when developing the gluconatetracer method and proposed three candidate reactions to explain the presence of unlabeled
intracellular gluconate; the oxidation of glucose to gluconate by either glucose oxidase or
glucose dehydrogenase and the dephosphorylation of intracellular 6PG to gluconate by a
phosphatase. Furthermore, Kleijn et al. [11] found the enzyme activities of the three proposed
enzymes to be below the detection limit of the applied assays, which was not unexpected
considering the low flux estimates for the three proposed reactions. Examination of the effect
of these metabolic scenarios on the actual PPP split-ratio showed that an active phosphatase
had no effect, while the oxidation of unlabeled glucose to gluconate led to a slightly altered
PPP split-ratio. Appendix A shows that for all four chemostat cultivations a small glucoseoxidation flux suffices to explain the changed labeling of intracellular gluconate. From these
calculations it was concluded that the possible presence of a glucose-oxidizing reaction could
be safely neglected in the PPP split-ratio estimations presented below.
Table 3 shows the calculated PPP split-ratios with their 95% confidence intervals. For both
specific growth rates a difference is observed between the PPP split-ratio of mycelia cultivated
in the absence and presence of PAA. It has been reported previously that the uncoupling
effect of PAA (by dissipating the proton-motive force across the plasma membrane) in this
strain is negligible under the applied PAA concentration and pH [20]. Furthermore, PAA is not
catabolized by the cell, as can be inferred from the mass balance for PAA (Table 1). Therefore
the observed difference in flux distribution can be completely attributed to the formation of
penicillin-G.
168
Table 3 Estimated PPP split-ratios for two penicillin-G producing and two non-producing chemostat
cultivations of P. chrysogenum operated at a growth-rate of 0.020 h-1 and 0.052 h-1. The 95% confidence
interval for the PPP split-ratios was determined using Monte Carlo simulations. To obtain the PPP splitratio in a chemostat grown solely on glucose, the estimated PPP split-ratios were corrected for the
additional influx of gluconate and for the slightly lower NADPH production rate in the oxidative branch of
the PPP as a result of gluconate uptake.
Chemostat cultivation
95% lower
boundary
PPP split-ratio
95% upper
boundary
+
0.472
0.517
0.559
0.020
-
0.355
0.381
0.407
0.052
+
0.447
0.494
0.538
0.052
-
0.376
0.411
0.443
µ (h-1)
Penicillin-G
production
0.020
Contrary to the above findings, Christensen et al. [3] did not observe a correlation between the
PPP split-ratio and penicillin production when investigating the metabolism of a high and a lowyielding strain of P. chrysogenum cultivated in a chemostat at a specific growth rate of 0.060.07 h-1. In this study the PPP split-ratio was determined by growing the cells on specifically
labeled [1-13C] glucose, followed by a 13C-based MFA in which metabolite balances were
combined with labeling patterns of proteinogenic amino acids measured with GC-MS. Using
13
C-Labeled gluconate tracing: application
phenoxyacetic acid (POA) as the side-chain precursor, the PPP split-ratio under penicillin‑V
producing conditions in the high- and low-yielding strain was estimated to be 0.70 and 0.66,
respectively. Cultivation of the high-yielding strain in a medium without POA resulted in an
estimated PPP split-ratio of 0.71. Based upon these findings it was speculated that the flux
through the oxidative branch of the PPP might be strain-specific and not a result of metabolic
burden.
The rigidity of the PPP split-ratio observed by Christensen et al. [3] can be partly ascribed
to the use of a different P. chrysogenum strain that had a 30% lower penicillin production
rate compared to this study (0.012 vs 0.017 mmol/g biomass/h) under the applied cultivation
conditions. Due to this lower production rate less NADPH was needed for penicillin synthesis,
resulting in a diminished split-ratio difference between the producing and non-producing
chemostat-culture.
Furthermore, the relatively high specific growth rate at which Christensen et al. [3] performed
their chemostat cultures reduced the chance of observing a measurable difference in splitratio between a producing and a non-producing chemostat-culture. In general, the PPP splitratio of a cell is positively correlated with its specific growth rate because at low specific
growth rates most substrate entering a cell is converted into energy (ATP) for maintenance
requirements, causing its metabolism to be dominated by catabolism. At increased growth
rates the corresponding increase in anabolic activity causes an increase in the cell’s
biosynthetic NADPH-demand, which increases the PPP split-ratio. This effect is visualized
in Figure 3, where the relationship between the specific growth rate and the PPP split-ratio
is plotted for both a penicillin-G producing and non-producing chemostat culture. Split-ratios
were calculated using the relation presented in Appendix B, based upon the stoichiometric
model and the energetic parameters proposed by van Gulik et al. [19, 20]. Cysteine synthesis
was modeled according to the transsulfuration pathway and the maximal biomass specific
penicillin-G production rate (qp) was set at a constant value of 0.017 mmol/g biomass/h
independent of the specific growth rate. Note that Eq. B-5 is based upon assumptions with
respect to the yield parameters of P. chrysogenum, making it difficult to draw conclusions
based upon the absolute values presented in Figure 3. Nonetheless, the two curves clearly
demonstrate that at increasing specific growth rates the difference in PPP split-ratio between
a β-lactam producing and a non-producing chemostat culture decreases.
The relation plotted in Figure 3 is in good accordance with the results of the present study.
Table 3 shows that a smaller difference in PPP split-ratio is observed between the producing
and non-producing culture for the experiments performed at a specific growth rate of 0.052
h-1 compared to those at 0.020 h-1. Since the maximal specific penicillin production rates were
practically identical for the two tested specific growth rates (Table 1), the observed difference
can be fully attributed to the expected higher biosynthetic NADPH-demand at increased
specific growth rates.
Previous studies have shown that both the gluconate tracer method and the 13C-labeling
based MFA produce accurate estimates for the PPP split-ratio. Kleijn et al. [11] and Table 3
show that the 95% confidence interval of the PPP split-ratio for the gluconate tracer method is
±0.04, while Christensen et al. [2] show that in Saccharomyces cerevisiae the 95% confidence
interval of the PPP split-ratio for the 13C-based MFA method is ±0.02. Based upon these
169
Chapter 6
confidence intervals and Figure 3, it was checked whether the two methods are sensitive
enough to distinguish between penicillin producing and non-producing conditions. At the
specific growth rates used by Christensen et al. [3] the difference in PPP split ratio between
penicillin producing and non-producing conditions may be hard to distinguish (0.04 at μ=0.07
h-1), but at the lower specific growth rates used in this study the difference can be reliably
observed by both methods (0.12 at μ=0.02 h-1).
0.6
Figure 3 Theoretical
relation
between
the PPP split-ratio
and specific growth
rate in a penicillin-G
producing and nonproducing chemostat
cultivation (Eq. B-4).
The derivation of the
relation is described
in Appendix B.
PPP split-ratio (-)
0.5
0.4
0.3
0.2
0.1
0.0
0.00
qp= 0.017 mmol/g biomass/h
qp= 0 mmol/g biomass/h
0.02
0.04
0.06
0.08
0.10
Growth rate (h-1)
170
6.3.3
Sources and sinks of NADPH
The increased PPP split-ratio under penicillin-producing conditions indicates a higher demand
for cytosolic NADPH, which is primarily caused by the synthesis of the two amino acid precursors
of the β-lactam nucleus: valine and cysteine. It is known that the biosynthesis of 1 mol of valine
requires 1 mol of cytosolic NADPH and 1 mol of mitochondrial NADPH. The biggest burden on
the cytosolic NADPH pool is, however, imposed by the biosynthesis of cysteine. This is a direct
consequence of the fact that cysteine is a sulfur-containing amino acid, requiring the active
uptake and reduction of sulfate (SO43-) from the medium. The reduction of sulfate to sulfide
(H2S) requires 4 cytosolic NADPH equivalents. Cysteine is formed by coupling sulfide to the
carbon backbone of serine. In P. chrysogenum two different cysteine biosynthesis pathways
exist [15]: (i) the direct sulfhydrylation pathway where sulfide is coupled in a single step to
serine and (ii) the transsulfuration pathway where sulfide is first coupled to acetyl-homoserine
forming homocysteine (an intermediate in methionine biosynthesis), then coupled to serine
to form cystathionine and subsequently, split into α-ketobutyrate and cysteine. Because the
cytosolic NADPH requirement for the synthesis of serine is 1 mol/mol and for homocysteine 3
mol/mol, the total NADPH requirement per mol of synthesized cysteine is 5 mol for the direct
sulfhydrylation pathway and 8 mol for the transsulfuration pathway.
Recently, Harris et al. [7] identified the enzymes involved in the cytosolic NADPH metabolism
of an aerobic glucose-limited P. chrysogenum chemostat culture. Enzyme assays confirmed
13
C-Labeled gluconate tracing: application
the strict NADP+-specificity of the G6P and 6PG dehydrogenases that catalyze the oxidative
branch of the PPP. In addition, a cytosolic NADP+-dependent isocitrate dehydrogenase was
detected. No NAD+-dependent activities were detected for these three enzymes in cell-free
extracts. Other cytosolic redox enzymes, such as glyceraldehyde-3-phosphate dehydrogenase
and acetaldehyde dehydrogenase, were not NADPH-specific. The observations of Harris
et al. [7] confirm to a large extent the assumptions on the cytosolic NADPH metabolism of
P. chrysogenum made in the stoichiometric model of van Gulik et al. [20].
6.3.4
Stoichiometric modeling
The first MFA for P. chrysogenum was performed by Jorgensen et al. [9] based upon a
stoichiometric model containing the transsulfuration pathway for cysteine synthesis and a
NAD+-dependent isocitrate dehydrogenase. Jorgensen et al. [9] showed that the PPP splitratio increases when the metabolism shifts from rapid growth to slow growth and mainly
penicillin production. Furthermore, they showed that the maximum theoretical yield of penicillin
on glucose increased by 20% if cysteine is synthesized by direct sulfhydrylation rather than
transsulfuration.
Additional insight in the NADPH metabolism of P. chrysogenum was obtained by performing
a conventional MFA for different assumptions of the NADPH stoichiometry, using the
stoichiometric model of van Gulik et al. [20] and comparing the estimated PPP split-ratios
to the experimentally obtained results of the gluconate-tracer method (Figure 3). Four
stoichiometric models were constructed, each containing (i) a cytosolic NADP+- or a NAD+dependent isocitrate dehydrogenase, and (ii) a cysteine synthesis route based upon the
transsulfuration or the direct sulfhydrylation pathway. As input for the MFA the measured
extracellular rates obtained from the four chemostat conditions were used, with the exception
of the erroneous oxygen consumption rate (Table 1). All presented MFA-derived PPP splitratios were statistically acceptable within the 95% confidence interval.
The incorporation of a solely NAD+-dependent isocitrate dehydrogenase lead to MFA-derived
PPP split-ratios which were much higher than the results obtained with the gluconate tracer
method, irrespective of the employed pathway for cysteine synthesis (Figure 4-B,C&D). This
is in agreement with the findings of Harris et al.[7] that the cytosolic isocitrate dehydrogenase
is NADP+-specific. Flux analysis shows that the relative contribution of the NADP+-specific
isocitrate dehydrogenase to the total cytosolic NADPH synthesis rate was about 13%. The
majority (~87%) of cytosolic NADPH thus has to be synthesized by the oxidative branch of
the PPP.
The results for the two alternative pathways for cysteine synthesis are less straightforward
to interpret, but tend to support the transsulfuration pathway. At a specific growth rate of
0.020 h-1 similar PPP split-ratios were estimated for the gluconate-tracer method and the
transsulfuration-based model, while at a specific growth rate of 0.052 h-1 a somewhat better
correspondence was observed for the direct sulfhydrylation-based model. For both specific
growth rates the difference in PPP split-ratio between a producing and non-producing
chemostat culture was best explained by the transsulfuration pathway. At a specific growth rate
of 0.020h-1 and 0.052h‑1 the difference in PPP split-ratio for the gluconate-tracer method was
measured to be 0.136 and 0.083, respectively. Slightly smaller differences were calculated for
171
Chapter 6
the transsulfuration pathway (0.116 and 0.040, respectively), while much smaller differences
were observed for the direct sulfhydrylation pathway (0.045 and 0.010, respectively).
Both cysteine biosynthesis pathways have been identified in P. chrysogenum by [15].
In fact, Evers et al. [4] provided proof that each pathway has its own distinctive role and
compartmentation: the mitochondrial direct sulfhydrylation pathway produces growth-related
cysteine, and the cytosolic transsulfuration pathway produces cysteine for penicillin synthesis.
This implies that in the absence of penicillin synthesis only the direct sulfhydrylation pathway
is active. Verification of this hypothesis using the results in this study was impossible, as only
marginal differences were observed between the two cysteine-biosynthesis pathways in the
MFA-derived PPP split-ratios of Figure 4‑B&D. These small differences were caused by the
fact that very little cysteine is needed for biomass synthesis (~0.15.10-3 mol cysteine/Cmol
biomass). An implication of the low flux through the direct sulfhydrylation pathway is that
the cysteine biosynthesis is predominated by the transsulfuration pathway under penicillinproducing conditions. This is in accordance with the findings of this study as explained in the
previous paragraph.
µ = 0.052 h-1
PPP split-ratio (%)
172
Penicillin-G production
70
A
PPP split-ratio (%)
60
50
40
70
50
40
30
Gluconate-tracer
TS & Icdh (NADP+)
DSH & Icdh (NADP+)
TS & Icdh (NAD+)
DSH & Icdh (NAD+)
40
70
C
B
50
30
60
No penicillin-G production
60
30
PPP split-ratio (%)
µ = 0.020 h-1
PPP split-ratio (%)
70
D
60
50
40
30
Figure 4 Comparison of the PPP split-ratio estimated via the gluconate-tracer method with the
MFA-derived PPP split-ratios for four different stoichiometric models of P. chrysogenum. The
four stoichiometric models each contained (i) a cytosolic NAD+-dependent or NADP+-dependent
isocitrate dehydrogenase (Icdh), and (ii) a cysteine biosynthesis pathway based upon either
transsulfuration (TS) or direct sulfhydrylation (DSH). PPP split-ratios were derived via MFA using
the measured conversion rates in Table 1-B.
13
C-Labeled gluconate tracing: application
Apart from identifying the cytosolic NADPH-producing enzymes in P. chrysogenum, Harris et
al. [7] also revealed the presence of a mitochondrial NADPH dehydrogenase that oxidizes
cytosolic NADPH via the mitochondrial respiration chain. Active expression of this enzyme
causes an increased cytosolic NADPH-consumption, which requires an increased PPP splitratio. Analysis of the MFA-derived PPP split-ratios depicted in Figure 4 shows that there is little
room for this increase, indicating that under the applied cultivation conditions the mitochondrial
NADPH dehydrogenase only fulfills a minor catabolic function. A quantitative determination of
the mitochondrial NADPH dehydrogenase flux was not possible, since its inclusion in the
stoichiometric models of Figure 4 led to a parallel route.
6.4 CONCLUSIONS
Quantification of the PPP split-ratio in penicillin-G producing and non-producing chemostat
cultures of P. chrysogenum confirmed that the flux through the oxidative branch of the PPP
is strongly correlated to β-lactam antibiotic production. Furthermore, it was shown through
metabolic flux analysis that the oxidative branch of the PPP produces the majority of the
cytosolic NADPH needed for penicillin synthesis; the only other supplier of cytosolic NADPH
being the recently identified NADP+-dependent isocitrate dehydrogenase in P. chrysogenum.
The observed increase in PPP split-ratio under penicillin producing conditions is best explained
by a stoichiometric model in which cysteine is synthesized via the transsulfuration pathway,
requiring 9 mol of cytosolic NADPH for every mole of penicillin produced from glucose. The
P. chrysogenum strains currently employed by industry have penicillin titers that are several
factors higher than those measured in this study, meaning that the metabolic burden on the
supply of cytosolic NADPH will be even higher. Limiting the cytosolic NADPH demand for
penicillin synthesis, by for example introducing a cytosolic direct sulfhydrylation pathway for
cysteine synthesis as earlier proposed by Jorgensen et al. and Ostergaard et al. [9, 15], can
thus form an interesting target for future metabolic engineering of P. chrysogenum.
173
ACKNOWLEDGEMENTS
This work was financially supported by the Dutch EET program (Project No. EETK20002) and
DSM.
Chapter 6
APPENDIX A: Glucose oxidase and phosphatase activity
Two candidate reactions were hypothesized for explaining the difference in 13C-labeling
between the gluconate added to the medium and the intracellular gluconate (Table 2); the
oxidation of glucose to gluconate by either glucose oxidase or glucose dehydrogenase and the
dephosphorylation of intracellular 6PG to gluconate by a phosphatase. By setting up a labeling
balance around the intracellular gluconate pool, estimates for the 6PG dephosphorylation
flux and the glucose-oxidation flux were determined. Corresponding PPP split-ratios were
determined by means of a second labeling balance around the 6PG-pool (similar to Eq. 6.1).
The applied labeling balances are described in detail by Kleijn et al. [11].
In all four chemostat experiments the estimated glucose oxidation and 6PG dephosphorylation
flux was only a fraction of the uptake rate of glucose (Table A-I). P-values for the flux-estimates
indicate that the unlabeled mass fraction of intracellular gluconate is somewhat better explained
by the dephosphorylation of 6PG to gluconate. Note that the dephosphorylation of 6PG has
no effect on the label-inflow into the 6PG-pool, resulting in unchanged PPP split-ratios. Due
to the small size of the glucose oxidation flux only marginally different PPP split-ratios were
observed for this metabolic scenario. PPP split-ratios were statistically acceptable for both
candidate reactions in all four chemostat experiments (p-value>0.05).
Table A-I Flux estimates and corresponding PPP split-ratios for the two metabolic pathways hypothesized
to explain the unlabeled mass fraction of intracellular gluconate: the oxidation of glucose and the
dephosphorylation of 6PG. To obtain the PPP split-ratio in a chemostat grown solely on glucose, the
estimated PPP split-ratios were corrected for the additional influx of gluconate and for the slightly lower
NADPH production rate in the oxidative branch of the PPP as a result of gluconate uptake.
Chemostat
cultivation
µ (h-1)
174
Glucose oxidation
FluxPenicillinestimate
G
P-valuea
(mmol/
production
Cmol/h)
6PG dephosphorylation
FluxPPP
estimate
split- P-valuea
(mmol/
ratio
Cmol/h)
p-valuea
PPP
Psplitvaluea
ratio
0.020
+
0.13
<0.01
0.511
0.32
0.15
<0.01
0.517
0.46
0.020
-
0.14
0.19
0.371
0.18
0.17
0.49
0.381
0.11
0.052
+
0.09
<0.01
0.490
0.13
0.10
0.79
0.494
0.13
0.052
0.07
0.13
0.408 0.41
0.08
0.92
0.411 0.38
a
p‑values for the flux-fits were determined via a χ 2 ‑distribution and denote the probability that the
discrepancy between the measured and simulated mass isotopomer distributions in the labeling balance
is a result of measurement error. A P-value ≤ 0.05 was considered as statistically significant and thus
indicates a clear deviation between the measured and simulated mass isotopomer distribution.
13
C-Labeled gluconate tracing: application
APPENDIX B: Theoretical calculation of the PPP split-ratio
The biomass specific glucose uptake-rate ( qs ) is derived from the well-known Herbert-Pirt
equation for substrate consumption:
qs =
qp
µ
+ max + ms
Ysxmax Ysp
(B-1)
max
Where Ysxmax is the maximum theoretical yield of biomass on glucose, Ysp
is the maximum
theoretical yield of penicillin-G on glucose and ms is the non-growth related maintenance
coefficient. These parameters have been determined and reported for P. chrysogenum by van
Gulik et al. [19].
A similar relation can be derived for the biomass specific cytosolic NADPH consumption rate
( qnadph ), neglecting the NADPH required for maintenance:
qnadph =
qp
µ
+ max
max
Ynx
Ynp
(B-2)
Ynxmax is the maximum theoretical yield of biomass on cytosolic NADPH (3.93 Cmol biomass/
mol NADPH) which can be derived from the stoichiometric model proposed by van Gulik et
max
al. [20], Ynp
is the maximum theoretical yield of penicillin-G on cytosolic NADPH (0.111 mol
penicillin-G/mol NADPH, see main text) based upon cysteine synthesis via the transsulfuration
pathway.
Taking into account that the oxidative branch of the PPP produces two molecules of NADPH
per cycle, the PPP split-ratio (PPPSR) can be defined as the ratio between qnadph and qs :
PPPSR =
qnadph ⋅ α
2 ⋅ qs
(B-3)
175
where α is a correction factor for the amount of NADPH synthesized via the oxidative branch
of the PPP with respect to the cytosolic NADP+-dependent isocitrate dehydrogenase (±0.87,
see main text).
By combining Eq. B-1, B-2 and B-3 and substituting the appropriate values for the yield
parameters the PPP split-ratio as a function of the specific growth rate and the biomass
specific penicillin-G production rate is obtained:
PPPSR =
0.21⋅ µ + 7.50 ⋅ qp
0.50 ⋅ µ + 11.5 ⋅ qp + 0.0030
(B-4)
Chapter 6
REFERENCES
[1] Blank, L.M. and Sauer, U. (2004) TCA cycle activity
in Saccharomyces cerevisiae is a function of the
environmentally determined specific growth and
glucose uptake rates. Microbiol. 150, 1085-1093.
[2] Christensen, B., Gombert, A.K. and Nielsen, J. (2002)
Analysis of flux estimates based on C-13-labelling
experiments. Eur. J. Biochem. 269, 2795-2800.
[3] Christensen, B., Thykaer, J. and Nielsen,
J.
(2000)
Metabolic
characterization
of
high- and low-yielding strains of Penicillium
chrysogenum. Appl. Microb. Biot. 54, 212-217.
[4] Evers, M.E., Trip, H., van den Berg, M.A.,
Bovenberg, R.A.L. and Driessen, A.J.M. (2004)
Compartmentalization and transport in betalactam antibiotics biosynthesis. In: Molecular
biotechnology of fungal beta-lactam antibiotics and
related peptide synthetases (Brakhage, A.A., Ed.),
Vol. 88, pp. 111-135. Springer Berlin, Heidelberg.
[5] Gombert, A.K., dos Santos, M.M., Christensen,
B. and Nielsen, J. (2001) Network identification
and flux quantification in the central metabolism of
Saccharomyces cerevisiae under different conditions
of glucose repression. J. Bacteriol. 183, 1441-1451.
176
[6] Gunnarsson, N., Eliasson, A. and Nielsen, J.
(2004) Control of fluxes towards antibiotics and
the role of primary metabolism in production
of antibiotics. In: Molecular biotechnology
of fungal beta-lactam antibiotics and related
peptide synthetases (Brakhage, A.A., Ed.), Vol.
88, pp. 137-178. Springer Berlin, Heidelberg.
[7] Harris, D.M., Diderich, J.A., van der Krogt, Z.A.,
Luttik, M.A.H., Raamsdonk, L.M., Bovenberg,
R.A.L., van Gulik, W.M., van Dijken, J.P. and
Pronk, J.T. (2006) Enzymic analysis of NADPH
metabolism in [beta]-lactam-producing Penicillium
chrysogenum: Presence of a mitochondrial
NADPH dehydrogenase. Metab. Eng. 8, 91-101.
[8] Henriksen, C.M., Christensen, L.H., Nielsen, J.
and Villadsen, J. (1996) Growth energetics and
metabolic fluxes in continuous cultures of Penicillium
chrysogenum. J. Biotechnol. 45, 149-164.
[9] Jorgensen, H., Nielsen, J., Villadsen, J. and
Mollgaard, H. (1995) Metabolic Flux Distributions
in Penicillium-Chrysogenum During Fed-Batch
Cultivations. Biotechnol. Bioeng. 46, 117-131.
[10] Kiefer, P., Heinzle, E., Zelder, O. and Wittmann,
C.
(2004)
Comparative
metabolic
flux
analysis of lysine-producing Corynebacterium
glutamicum cultured on glucose, or fructose.
Appl.
Environ.
Microb.
70,
229-239.
[11] Kleijn, R.J., van Winden, W.A., Ras, C., van
Gulik, W.M., Schipper, D. and Heijnen, J.J. (2006)
13C-Labeled Gluconate Tracing as a Direct and
Accurate Method for Determining the Pentose
Phosphate Pathway Split Ratio in Penicillium
chrysogenum. Appl. Environ. Microb. 72, 4743-4754.
[12] Lange, H.C., Eman, M., van Zuijlen, G., Visser, D., van
Dam, J.C., Frank, J., de Mattos, M.J. and Heijnen, J.J.
(2001) Improved rapid sampling for in vivo kinetics
of intracellular metabolites in Saccharomyces
cerevisiae. Biotechnol. Bioeng. 75, 406-415.
[13] Mashego, M.R., van Gulik, W.M., Vinke, J.L. and
Heijnen, J.J. (2003) Critical evaluation of sampling
techniques for residual glucose determination in
carbon-limited chemostat culture of Saccharomyces
cerevisiae. Biotechnol. Bioeng. 83, 395-399.
[14] Nasution, U.,
van Winden,
J.J.
(2006)
metabolites of
nucleotides in
chrysogenum.
van Gulik, W.M., Kleijn, R.J.,
W.A., Proell, A. and Heijnen,
Measurement
of
intracellular
primary metabolism and adenine
chemostat cultivated Penicillium
Biotechnol. Bioeng. 94, 159-166.
[15] Ostergaard, S., Theilgaard, H.B.A. and Nielsen,
J. (1998) Identification and purification of Oacetyl-L-serine sulphhydrylase in Penicillium
chrysogenum. Appl. Microb. Biot. 50, 663-668.
[16] Sawyer, D.T. and Bagger, J.B. (1959) The
Lactone Acid Salt Equilibria for D-Glucono-DeltaLactone and the Hydrolysis Kinetics for This
Lactone. J. Am. Chem. Soc. 81, 5302-5306.
[17] Sonderegger, M., Jeppsson, M., Hahn-Hagerdal, B.
and Sauer, U. (2004) Molecular basis for anaerobic
growth of Saccharomyces cerevisiae on xylose,
investigated by global gene expression and metabolic
flux analysis. Appl. Environ. Microb. 70, 2307-2317.
[18] Thykaer,
J.
and
Nielsen,
Metabolic
engineering
of
production.
Metab.
Eng.
J.
(2003)
beta-lactam
5,
56-69.
[19] van Gulik, W.M., Antoniewicz, M.R., deLaat,
W.T., Vinke, J.L. and Heijnen, J.J. (2001)
Energetics of growth and penicillin production
in a high-producing strain of Penicillium
13
C-Labeled gluconate tracing: application
chrysogenum. Biotechnol. Bioeng. 72, 185-193.
[20] van Gulik, W.M., de Laat, W.T., Vinke, J.L.
and Heijnen, J.J. (2000) Application of
metabolic flux analysis for the identification of
metabolic bottlenecks in the biosynthesis of
penicillin-G. Biotechnol. Bioeng. 68, 602-618.
[21] van Winden, W.A., van Gulik, W.M., Schipper, D.,
Verheijen, P.J.T., Krabben, P., Vinke, J.L. and Heijnen,
J.J. (2003) Metabolic flux and metabolic network
analysis of Penicillium chrysogenum using 2D [C13, H-1] COSYNMR measurements and cumulative
Bondomer simulation. Biotechnol. Bioeng. 83, 75-92.
[22] van Winden, W.A., Wittmann, C., Heinzle,
E. and Heijnen, J.J. (2002) Correcting mass
isotopomer distributions for naturally occurring
isotopes. Biotechnol. Bioeng. 80, 477-479.
[23] van Winden, W.A., van Dam, J.C., Ras, C.,
Kleijn, R.J., Vinke, J.L., van Gulik, W.M. and
Heijnen, J.J. (2005) Metabolic-flux analysis of
Saccharomyces cerevisiae CEN.PK113-7D based
on mass isotopomer measurements of 13C-labeled
primary metabolites. FEMS Yeast Res. 5, 559-568.
[24] Vanderheijden, R.T.J.M., Heijnen, J.J., Hellinga,
C., Romein, B. and Luyben, K.C.A.M. (1994)
Linear Constraint Relations in Biochemical
Reaction
Systems
.1.
Classification
of
the Calculability and the Balanceability of
Conversion Rates. Biotechnol. Bioeng. 43, 3-10.
[25] Zhu, J. and Shimizu, K. (2005) Effect of a singlegene knockout on the metabolic regulation in
Escherichia coli for D-lactate production under
microaerobic condition. Metab. Eng. 7, 104-115.
177
CHAPTER
Discussion and
future directions
7
Chapter 7
Discussion and Future Directions
The aim of metabolic flux and network analysis is to accurately characterize the metabolic
network and flux distribution within a cell. Many new advances have been made within this
field over the past decade, as discussed in the first chapter of this thesis. Most notable is the
introduction of the 13C-labeling technique, which has led to the development of new analytical
platforms for determining the 13C-labeling of intracellular metabolites and the development of
novel flux analysis methods for interpreting the measured 13C-labeling distributions. Despite
these advancements many uncertainties still remain, preventing a complete understanding of
the metabolic network of the cell and causing inaccuracies in the determined flux distributions.
To further improve the quality of the derived metabolic flux patterns several issues need to be
addressed.
180
7.1 Incomplete metabolic network models
Within a cell thousands of different metabolic reactions take place. In general, only a small
subset of these reactions is included in the currently applied metabolic network models. The
omission of reactions influencing the isotopomer distribution in the cell can easily lead to
erroneous flux distributions. Note that this problem is most pronounced for the whole isotopomer
modeling approach. Nonetheless, erroneous network assumptions and simplifications can
also be made when determining flux ratio via the local flux node analysis approach. The effect
of an incomplete metabolic network model was illustrated in Chapter 2 when applying the
whole isotopomer modeling approach to Saccharomyces cerevisiae and the non-oxidative
branch of the pentose phosphate pathway (PPP). Aside from missing reactions, erroneous
flux patterns can also arise as a result of aspecific enzyme reactions in the cell. This issue
was also encountered in Chapter 2, where the broad substrate specificity of transketolase and
transaldolase (octulose-8-phosphate formation) and the possible presence of isoenzymes
were studied.
To prevent erroneous flux patterns, independent verification of the metabolic network model is
important. Labeling redundancies in the measured 13C‑labeling data and enzymatic analysis
data can be used to verify the proposed metabolic network model (see Chapter 3 and 4).
However, a more comprehensive method would be to reconstruct the metabolic network
using currently available genomic, transcriptomic, biochemical and physiological information
to produce more realistic models [1-3]. A limiting factor with these models is the computational
power needed to calculate the isotopomer distributions of all metabolites included in the model;
consequently, no such models exist as of yet. In that sense, the novel framework for modeling
isotopic distributions based on elementary metabolite units as introduced by Antoniewicz et al.
[4] may help to overcome these computational limitations.
7.2 Culture and cell heterogeneity
Metabolic flux analysis (MFA) implicitly assumes the metabolic state of each cell in the studied
culture to be identical. However, this assumption is only valid when the 13C-labeling pattern of
a metabolite pool is homogeneously distributed both amongst the cells in the studied culture
(macroscopic level) and within a single cell (microscopic level). Therefore, additional insight
Discussion and future directions
in the sources of cell and culture heterogeneity is needed. Note that the relation between
fluxes and isotopomer distributions is non-linear. In the case of heterogeneously distributed
metabolite pools, averaged 13C-labeling patterns will not necessarily yield the ‘average’ fluxpattern within the cell culture.
7.2.1
Macroscopic level
Possible sources of heterogeneity at the macroscopic level are [5, 6]:
•
Asynchronical cell growth The various stages in the cell cycle each have their own
physiological function in the overall growth process [7]. As a result, the metabolic fluxes
in one stage of the cell cycle (e.g. the G1/G2-phase in which protein is synthesized) can
differ from other cell cycle stages (e.g. the S-phase).
Cell differentiation This problem is most pronounced when culturing multi-cellular
•
organisms, such as filamentous fungi. These organisms contain different cell types each
with their own specialized function and thus metabolism.
Strain instability If the cultured strain is genetically unstable, sub-cultures of mutant
•
strains can co-exist with the parent strain. Generally, the metabolism of the mutant differs
from that of the parent strain.
Synchronized cell cultures can be used to study cell-cycle dependent variations in metabolic
fluxes during a time period of one to two cell divisions. Currently, experiments are being
performed to study these variations in a culture of S. cerevisiae grown aerobically on 13Clabeled glucose [8]. Interestingly, several researchers have speculated that the metabolic
fluxes determined via the labeling patterns of proteinogenic amino acids (NMR and GC-MS
analysis) specifically reflect the flux distribution during the cell cycle phases when proteins
are synthesized (G1/G2-phase) The LC-MS derived flux patterns, on the other hand, are
based upon a direct measurement of the 13C-labeling of primary metabolites and thus do
not necessarily overrepresent a specific cell cycle stage. It should be pointed out here that
depending on variations in the absolute metabolite concentrations from one cell-cycle stage
to another [9], specific stages can still be overrepresented. Despite these uncertainties, the
similar flux distributions for the three 13C-measurement techniques in S. cerevisiae (Chapter
3) are a first indication that metabolic variations amongst the cell cycle stages do not play a
major role.
Since S. cerevisiae is a unicellular organism, the effect of cell differentiation on culture
heterogeneity is negligible. In contrast, a complex morphological differentiation takes place
in the filamentous fungi Penicillium chrysogenum. The statistical rejection and differing LCMS- and NMR-derived flux patterns found in Chapter 4 for P. chrysogenum might thus well be
the result of different cell types, each with their own metabolic state. Note that apart from cell
differentiation, the heterogeneity of the studied culture was further increased by the different
structures (pellets, aggregates and/or dispersed mycelia) formed by interacting hyphae. As
of yet no research has been conducted on the metabolic flux patterns within the different cell
types of a multi-cellular organism.
Strain instability in combination with selection pressure can result in the co-existence of
(multiple) mutant strains alongside the parental strain. For example, Christensen et al. [10]
181
Chapter 7
showed for an industrial P. chrysogenum strain that a progressive loss in penicillin production
capacity was correlated with the formation of mutants. In the 13C-labeling experiments described
in this thesis little degeneration of penicillin production was observed, indicating that strain
instability played only a minor role in the studied P.chrysogenum strains. Nevertheless, strain
instability can dramatically distort the outcome of flux fits when growing less stable strains for
a prolonged time in a fermentor.
7.2.2
Microscopic level
Possible sources of heterogeneity at the microscopic level are [5, 6]:
Compartmentation Eukaryotic cells contain multiple compartments (e.g. nucleus,
•
mitochondria and peroxisomes) each with their own specific function in the cell. As a
result of the compartmented metabolism in these cells, subsets of metabolite pools can
exist, each with their own specific 13C-labeling distribution.
Metabolite channeling Metabolite channeling describes the process in which a product
•
of one enzymatic reaction is transferred to the next enzyme without mixing with the bulkphase metabolite pool. As described in detail by van Winden et al. [11], channeling may
lead to the ‘micro-compartmentation’ of the reaction intermediates, which is in conflict
with the assumption of homogeneous metabolite pools.
By constructing a compartmented metabolic network model of the studied eukaryotic cell,
one can examine the intracellular flux distributions in the individual compartments. In much
the same way, the effect of metabolite channeling on the flux distribution can be studied by
modeling each micro-compartment as a separate metabolite pool [11]. Both these approaches
require detailed knowledge on the metabolism of the studied organism, which might not always
be available (see previous section). In that respect it is interesting to take a step back and
test the variability in flux estimates amongst the different 13C-quantification techniques (see
Chapter 3 and 4) in a less complex and uncompartmented micro-organism, such as the wellcharacterized, unicellular, gram-negative prokaryote Escherichia coli.
182
7.3 Measurement error
To draw conclusions on the validity and precision of the determined flux patterns it is essential
to accurately determine the measurement error of the determined 13C-labeling distributions.
In general, three sources of measurement error can be identified: (i) the 13C-labeling analysis;
(ii) the sampling and (iii) the fermentation. The first two sources of measurement error
have been addressed in this thesis and were studied before by Christensen et al. [12]. The
measurement errors of the applied 13C-labeling analysis method and sampling procedure can
be straightforwardly determined by measuring the error variability within a single sample and
in between multiple samples, respectively (see chapter 4 and 5). By running multiple identical
13
C-labeling experiments one can also determine the measurement error of the fermentation
process and, consequently, the reproducibility of the determined flux patterns.
Discussion and future directions
7.4 Local flux node analysis
In principle, route specific 13C-labeled tracers can be used to determine flux-patterns in parts
of the metabolic network, which otherwise might be hard to determine accurately. An example
of this was shown in Chapter 6, where the flux through the oxidative branch of the PPP was
accurately determined by co-feeding trace amounts of 13C-labeled gluconate. In such tracerstudies it is essential that: (i) the applied tracer is taken up simultaneously with the main carbonsource; (ii) tracer amounts are such that the effect on the studied metabolism is negligible and
(iii) the 13C-labeling distributions of the metabolites directly surrounding the branch-point are
measurable. If these prequisites hold, this method can, for example, be used to estimate the
gluconeogenic flux catalyzed by phosphoenol pyruvate (PEP) carboxykinase by feeding trace
amounts of 13C-labeled aspartate or lactate, setting up an overall 13C-balance over the PEPpool and measuring the 13C-labeling distribution of oxaloacetate and PEP. Note that this flux
previously proved difficult to estimate in P. chrysogenum cultivations (see Chapter 4).
7.5 Pentose phosphate pathway
Quantification of the PPP split-ratio in penicillin-G producing and non-producing chemostat
cultures of P. chrysogenum showed that the flux through the oxidative branch of the PPP
is strongly correlated to β-lactam antibiotic production. NADPH balances based upon these
estimated PPP split-ratios revealed that NADPH production in the cell was far in excess
of the stoichiometric requirements for NADPH. Further research is needed to identify and
incorporate any missing NADPH consuming reactions in the metabolic network model.
Candidate reactions are those catalyzed by enzymes having multiple cofactor specificities
(e.g. as a result of isoenzymes) and reactions involved in transhydrogenation cycles and in
dealing with oxidative stress.
Restructuring the metabolic network model for the non-oxidative branch of the PPP allowed
for a more accurate description of the 13C-label and, consequently, flux distribution in this
part of the metabolism. Despite this improvement the reversibility of the transketolase and
transaldolase catalyzed reactions remains difficult to estimate using 13C-labeling based MFA.
183
Chapter 7
ReferenceS
184
[1] Borodina, I., Krabben, P. and Nielsen, J. (2005)
Genome-scale analysis of Streptomyces coelicolor
A3(2) metabolism. Genome Res. 15, 820-829.
[2] Forster, J., Famili, I., Fu, P., Palsson, B.O. and
Nielsen, J. (2003) Genome-scale reconstruction of
the Saccharomyces cerevisiae metabolic network.
Genome Res. 13, 244-253.
[3] Reed, J.L., Vo, T.D., Schilling, C.H. and Palsson,
B.O. (2003) An expanded genome-scale model of
Escherichia coli K-12 (iJR904 GSM/GPR). Genome
Biology 4: R54.
[4] Antoniewicz,
M.R.,
Kelleher,
J.K.
and
Stephanopoulos, G. Elementary metabolite units
(EMU): A novel framework for modeling isotopic
distributions. Metab. Eng. In Press.
[5] Christensen, B. (2001) Metabolic network analysis:
Principles, methodologies and applications. Thesis.
Technical University of Denmark, Denmark.
[6] van Winden, W.A. (2003) 13C-Labeling Technique
for Metabolic Network and Flux Analysis: Theory
and Application.
Thesis. Delft University of
Technology, the Netherlands.
[7] Lloyd, D. and Murray, D.B. (2006) The temporal
architecture of eukaryotic growth. FEBS Lett. 580,
2830-2835.
[8] Costenoble, R., Muller, D., Barl, T., van Gulik,
W.M., van Winden, W.A., Reuss, M. and Heijnen,
J.J. (2006) 13C-isotopically labeled metabolic
flux analysis of a synchronised culture of
Saccharomyces cerevisiae. FEMS Yeast Res. In
Press.
[9] Wittmann, C., Hans, M., van Winden, W.A., Ras, C.
and Heijnen, J.J. (2005) Dynamics of intracellular
metabolites of glycolysis and TCA cycle during
cell-cycle-related oscillation in Saccharomyces
cerevisiae. Biotechnol. Bioeng. 89, 839-847.
[10] Christensen, L.H., Henriksen, C.M., Nielsen, J.,
Villadsen, J. and Egel-Mitani, M. (1995) Continuous
cultivation of Penicillium chrysogenum. Growth on
glucose and penicillin production. J. Biotechnol. 42,
95-107.
[11] van Winden, W., Verheijen, P. and Heijnen, S.
(2001) Possible pitfalls of flux calculations based
on 13C-labeling. Metab. Eng. 3, 151-162.
[12] Christensen, B., Gombert, A.K. and Nielsen, J.
(2002) Analysis of flux estimates based on 13Clabeling experiments. Eur. J. Biochem. 269, 27952800.
Discussion and future directions
185
Summary
Summary of the thesis: ‘Development and Application of 13C-Labeling Techniques:
Analyzing the Pentose Phosphate Pathway of Penicillium chrysogenum’ by Roelco
Kleijn
The 13C-labeling technique is a powerful characterization tool within the field of metabolic
engineering aimed at determining intracellular steady state fluxes. The provided metabolic
snapshots enable researchers to better understand and predict the phenotypic behavior of
a micro-organism as a result of genetic alterations and/or different environmental conditions.
In general, two main methods can be distinguished for deriving metabolic flux patterns from
measured 13C-label distributions; (i) the local flux analysis approach, which determines the
intracellular fluxes around a selected metabolite node and (ii) the whole isotopomer modeling
approach which aims at estimating all fluxes throughout a predefined reaction network
model. In this thesis the different 13C-labelling techniques were further developed and the
available analytical platforms were applied for the analysis of the metabolic fluxes in the
well-characterized yeast Saccharomyces cerevisiae and the less-studied filamentous fungi
Penicillium chrysogenum. Throughout this thesis special attention was paid to the flux through
the oxidative branch of the pentose phosphate pathway (PPP), which is of prime importance
for penicillin synthesis in P. chrysogenum.
A prerequisite for accurately determining metabolic flux patterns using the whole isotopomer
modeling approach is a correct metabolic network model. In Chapter 2 it is discussed that
the often applied reaction structure in metabolic network models for the non-oxidative branch
of the PPP is not in accordance with the established ping-pong kinetic mechanism of the
enzymes transketolase and transaldolase. Consequently, the traditional reactions of the
non-oxidative branch of the PPP were replaced by metabolite specific, reversible, C2 and C3
fragments producing and consuming half-reactions. Application of this fundamentally different
metabolic network model to a previously published 13C-based metabolic flux analysis (MFA) of
S. cerevisiae resulted in different flux estimates. In addition to better describing the reaction
structure in the non-oxidative branch of the PPP, the new metabolic network model also allowed
one to study the effect of isoenzymes on the 13C-label distribution in the non-oxidative branch
of the PPP. It is well known that several micro-organisms contain multiple genes encoding for
isoenzymes of transketolase and transaldolase. Results showed that inclusion of isoenzymes
affected the ensuing flux-estimates.
Several analytical platforms exist for measuring the 13C-labeling distribution of intracellular
primary metabolites. In Chapter 3 the three most common platforms, liquid chromatographymass spectrometry (LC-MS), gas chromatography-mass spectrometry (GC-MS) and 2D
[13C,1H] correlation nuclear magnetic resonance spectroscopy (NMR), were compared and
used to unravel the carbon and redox metabolism of a glycerol over-producing S. cerevisiae
strain with deletions in the structural genes encoding triosephosphate isomerase (tpi1), the
external mitochondrial NADH dehydrogenases (nde1 and nde2) and the respiratory chainlinked glycerol-3-phosphate dehydrogenase (gut2). Consistent fluxes were estimated for the
three analytical platforms. In addition, flux sensitivities around several important metabolic
nodes proved to be dependent on the applied technique, indicating that flux patterns were
best resolved by a combination of the three 13C-quantification techniques. From a biological
point of view it was found that a combination of assimilatory metabolism and PPP activity
187
Summary
188
diverted carbon away from glycerol formation. A direct consequence of this was the formation
of excess cytosolic NADH, suggesting the presence of a fourth cytosolic NADH sink in addition
to the three that were knocked-out in the studied mutant. Based upon the estimated exchange
flux of four-carbon dicarboxylic acids across the mitochondrial membrane, it was suggested
that the malate/aspartate or malate/oxaloacetate redox shuttle plays a role in the transfer of
redox equivalents from the cytosol to the mitochondrial matrix.
In Chapter 4 the analytical platform comparison was extended to a high-yielding P.
chrysogenum strain grown under penicillin-G producing and non-producing conditions. Two
13
C-quantification techniques were compared; LC-MS and NMR. As the metabolic network
model of P. chrysogenum was primarily based on that of S. cerevisiae, the 13C-based MFA
was preceded by a node by node analysis of the proposed primary metabolic network. This
resulted in: (i) a modified compartmental origin of several amino acid precursors (ii) the
identification of rapidly equilibrated metabolic pools and (iii) evidence for a mitochondrial acetylCoA transporter and the enzymes threonine aldolase, phosphoenol-pyruvate carboxykinase
and malic enzyme. Despite the altered metabolic network model, the two 13C quantification
techniques yielded very different flux estimates. The heterogeneity of the P. chrysogenum
culture was suggested as a possible explanation for the discrepancy between the two applied
analytical platforms. While S. cerevisiae is an unicellular micro-organism, P. chrysogenum
forms elongated structures that contain multiple cells each with there own specific function.
Further research is needed to determine whether the complexity of fungal cultivations, due
to cell differentiation and complex formation leads to different metabolic flux pattern when
applying the nowadays available 13C-quantification techniques.
The MFA of P. chrysogenum performed in Chapter 4 provided no conclusive evidence on
the flux through the oxidative branch of the PPP. As a result, a novel 13C-tracer method was
developed in Chapter 5. In contrast to the whole isotopomer modeling approach applied in
Chapters 2-4, this method was aimed at determining the flux through one specific metabolic
node, namely the oxidative branch of the PPP (local flux analysis approach). In general, the
local flux analysis approach allows a more accurate flux determination of a specific metabolic
node, but yields no information on the remaining cellular metabolism. The flux through the
oxidative branch of the PPP was calculated by simultaneously feeding unlabeled glucose
and trace amounts of [U-13C]gluconate, followed by measurement of the mass isotopomers
of 6-phospho-gluconate and glucose-6-phosphate. Application of the method to a penicillinG producing culture of P. chrysogenum yielded a PPP split-ratio of 51.8%. Surprisingly, the
obtained value was very comparable to the value determined in Chapter 4 (51.1%). On the
other hand, the sensitivity of the gluconate-tracer method was much higher, resulting in a 95%
confidence interval that was more than three times as small.
In Chapter 6 the gluconate tracer method developed in Chapter 5 was used to study the
flux through the oxidative branch of the PPP in penicillin-G producing and non-producing
chemostat cultures of P. chrysogenum. Significantly higher oxidative PPP fluxes were
observed in penicillin-G producing chemostat cultures, thereby for the first time presenting
experimental proof for the causal relationship between penicillin production and NADPHsupply. By comparing the PPP fluxes from the gluconate-tracer experiment to oxidative
PPP fluxes derived via metabolic flux analysis it was shown that the oxidative branch of the
PPP produces the majority of the cytosolic NADPH needed for penicillin synthesis; the only
Summary
other supplier of cytosolic NADPH being the recently identified NADP+-dependent isocitrate
dehydrogenase in P. chrysogenum. The observed increase in PPP split-ratio under penicillin
producing conditions was best explained by a stoichiometric model in which cysteine is
synthesized via the transsulfuration pathway. With the transsulfuration pathway included, the
synthesis of one mol penicillin from glucose requires 9 mol of cytosolic NADPH.
In conclusion, this thesis presents innovations with respect to the experimental and mathematical
aspects of the 13C-labeling technique. By applying these new aspects additional insight in the
primary carbon metabolism of S. cerevisiae and P. chrysogenum was obtained. Despite these
advancements many uncertainties still remain, preventing a complete understanding of the
metabolic network of the cell and causing inaccuracies in the determined flux distributions.
In Chapter 7 some recommendations are made to further improve the quality of the derived
metabolic flux patterns in future research.
189
Samenvatting
Samenvatting van het proefschrift: ‘Ontwikkeling en Toepassing van 13Clabelingstechnieken: Een analyse van de pentose fosfaat route in Penicillium
chrysogenum’ door Roelco Kleijn
Binnen de industriële biotechnologie bestaat een continue vraag naar micro-organismen die
in staat zijn een gewenst product sneller en efficiënter te synthetiseren. Nu verschillende
processen in een cel steeds beter worden begrepen, zijn wetenschappers steeds vaker in
staat om op basis van rationele beslissingen het celmetabolisme aan te passen, en zodoende
de productvorming in een cel te verhogen. Voor het bestuderen van celmetabolisme wordt
vaak gebruikt gemaakt van 13C labelingstechnieken. 13C is een stabiel, natuurlijk isotoop van
koolstof met een afwijkende massa en nucleaire draaiing. Door een substraat te voorzien
van een 13C label en de distributie van 13C binnen het celmetabolisme te meten, kunnen de
koolstof fluxen door de verschillende stofwisselingsroutes gemeten worden en kunnen de
invloeden van verschillende omgevingsfactoren of genetische aanpassingen beter begrepen
worden. Men kan twee soorten 13C-labelingstechnieken onderscheiden; (1) de lokale flux
analyse methode, waarmee flux patronen rond een aantal specifieke knooppunten in het
metabolisme bepaald worden, en (2) de complete flux analyse methode, waarmee alle fluxen
in een metabool netwerk bepaald worden met behulp van een isotopomeer model dat de
distributie van alle koolstof atomen binnen dit netwerk tracht te beschrijven.
In dit proefschrift worden de verschillende 13C labelingstechnieken eerst verder ontwikkeld,
om vervolgens het metabolisme van de gist Saccharomyces cerevisiae en de schimmel
Penicillium chrysogenum beter in kaart te brengen. Speciale aandacht wordt besteed aan
de flux door de oxidatieve tak van de pentose fosfaat route. De pentose fosfaat route is een
belangrijk onderdeel van de stofwisseling en bestaat uit 2 fases; (1) de oxidatieve fase waarin
NADPH gegenereerd wordt, en (2) de non-oxidatieve fase waarin metabolieten gevormd
worden die essentieel zijn voor de vorming van aminozuren, nucleotiden en nucleïnezuren.
Eerder onderzoek heeft aangetoond dat de oxidatieve fase van de pentose fosfaat route
mogelijk van belang is voor de productie van penicilline in P. chrysogenum.
Het nauwkeurig bepalen van metabole fluxen met behulp van de complete flux analyse
methode vereist een correct model van het metabool netwerk. In Hoofdstuk 2 wordt
aangetoond dat het tot nu toe gebruikte model voor de non-oxidatieve fase van de pentose
fosfaat route niet overeenkomt met het ping-pong reactie mechanisme van de enzymen
transketolase en transaldolase. Naar aanleiding hiervan werden de reacties in het traditionele
model vervangen door metaboliet specifieke, reversibele half reacties. Aan de hand van een
eerder gepubliceerde flux analyse van S. cerevisiae werd aangetoond dat de nieuwe reactie
structuur resulteerde in afwijkende flux patronen. Naast een verbeterde beschrijving van de
13
C-label distributie in de pentose fosfaat route, kan met het nieuwe model ook het effect
van isoenzymen op de 13C-label distributie worden bestudeerd. De inclusie van isoenzymen
voor transketolase en transaldolase in het metabool model leidde ook tot veranderde flux
patronen.
Er bestaan verschillende analytische methoden om de 13C-labeling van intracellulaire primaire
metabolieten te meten. In Hoofdstuk 3 worden de meest gebruikte analysemethoden (LCMS, GC-MS en NMR) vergeleken en gebruikt om het koolstof en redox metabolisme van een
glycerol-overproducerende S. cerevisiae stam te onderzoeken. Er werden consistente flux
191
Samenvatting
192
patronen gevonden voor alle drie de analysemethoden, echter de flux gevoeligheid rondom een
aantal belangrijke knooppunten bleek afhankelijk van de gebruikte techniek. In dit hoofdstuk
werd voor het eerst aangetoond dat een combinatie van de drie analysemethoden de meest
nauwkeurige flux patronen oplevert. Vanuit biologisch oogpunt bleek dat de pentose fosfaat
route en een aantal anabole routes de koolstof flux naar glycerol verminderden. Uitgaande
van het huidige metabool model voor S. cerevisiae, zou dit moeten leiden tot een disbalans
in het cytosolaire redox metabolisme met celdood tot gevolg. Het feit dat deze S. cerevisiae
stam toch in staat was om te groeien, wijst op de aanwezigheid van een additionele NADH
shuttle. Via een metabole flux analyse werden de malaat-aspartaat shuttle en de malaatoxaloacetaat shuttle als kandidaat shuttles aangewezen.
In Hoofdstuk 4 wordt de vergelijking van de analytische methoden voortgezet in een voormalige
productie stam van P. chrysogenum. Met behulp van LC-MS en NMR werden metabole
fluxen bepaald onder zowel penicilline producerende en niet-producerende condities.
Aangezien het metabole model van P. chrysogenum voornamelijk gebaseerd was op dat van
S. cerevisiae, werd eerst een uitvoerige analyse van het metabole netwerk uitgevoerd. Dit
resulteerde in: 1) een afwijkende compartimentalisering van een aantal aminozuur synthese
routes, 2) de identificatie van snel geëquilibreerde metaboliet koppels, en 3) het bewijs dat
P. chrysogenum beschikt over een mitochondriële acetyl-CoA transporter en de enzymen
threonine aldolase, fosfo-enolpyruvaat carboxykinase en malic enzyme. Ondanks de
aangebrachte verbeteringen in het model werden er afwijkende flux patronen gevonden voor
de twee vergeleken analysemethoden. Een mogelijke verklaring hiervoor is de heterogeniteit
van P. chrysogenum; in tegenstelling tot het eencellige micro-organisme S. cerevisiae, zorgt
celdifferentiatie en complexformatie in P. chrysogenum namelijk voor subpopulaties met elk
hun eigen metabolisme. Verder onderzoek is nodig om te bepalen of de complexiteit van
schimmel fermentaties inderdaad leidt tot afwijkende flux patronen voor de hedendaags
gebruikte 13C analysemethoden.
De metabole flux analyse van P. chrysogenum zoals beschreven in Hoofdstuk 4 gaf geen
eenduidig antwoord op de vraag of penicilline productie van invloed is op de flux door de
oxidatieve fase van de pentose fosfaat route. Om deze reden werd een nieuwe 13C-tracer
methode ontwikkeld in Hoofdstuk 5. In tegenstelling tot de eerder gebruikte complete flux
analyse methode (Hoofdstuk 2 t/m 4), is de nieuwe methode specifiek gericht op de flux
door de oxidatieve fase van de pentose fosfaat route (lokale flux analyse). Het onderliggende
principe bij deze methode is de simultane opname van zowel ongelabeld glucose als kleine
hoeveelheden 13C-gelabeld gluconaat. Door de massa isotopomeren van 6 fosfo-gluconaat
en glucose-6-fosfaat te meten, kan vervolgens eenvoudig berekend worden hoeveel van de
opgenomen glucose gemetaboliseerd wordt via de oxidatieve fase van de pentose fosfaat
route (deze verhouding wordt de split ratio genoemd). Toepassing van deze methode in P.
chrysogenum leverde een split ratio op van 51.8%. Verassend genoeg kwam deze split ratio sterk
overeen met de split ratio bepaald in Hoofdstuk 4 (51.1%). Het 95% betrouwbaarheidsinterval
van de split ratio bepaald in Hoofdstuk 5 was echter drie keer kleiner dan van de split ratio
bepaald in Hoofdstuk 4.
In Hoofdstuk 6 wordt de gluconaat tracer methode toegepast om de flux door de oxidatieve
fase van de pentose fosfaat route in een penicilline-G producerende en niet-producerende
chemostaat cultuur van P. chrysogenum te onderzoeken. In de penicilline-G producerende
Samenvatting
chemostaat culturen werd een significant verhoogde oxidatieve flux waargenomen. Aangezien
de oxidatieve fase van de pentose fosfaat route de belangrijkste bron van cytosolair NADPH
is, werd hiermee voor het eerst experimenteel bewijs geleverd voor een causale relatie
tussen penicilline productie en NADPH toevoer. De enige andere leverancier van cytosolair
NADPH in P. chrysogenum is de onlangs geïdentificeerde NADP+-afhankelijke isocitraat
dehydrogenase. Berekeningen lieten zien dat de verhoogde split ratio onder penicilline
producerende condities het beste verklaard kon worden door een stoichiometrisch model
waarin cysteïne gesynthetiseerd wordt via de transsulfuratie route. Dit betekent dat minimaal
9 mol cytosolair NADPH nodig is voor de synthese van 1 mol penicilline-G uit glucose.
Concluderend, in dit proefschrift worden een aantal innovaties voor de 13C-labelingstechniek
beschreven. Het toepassen van deze innovaties heeft geleid tot meer inzicht in het primaire
koolstofmetabolisme van S. cerevisiae and P. chrysogenum. Ondanks de nieuwe inzichten
resten er nog vele vraagtekens in het metabole netwerk van S. cerevisiae and P. chrysogenum,
waardoor vooralsnog onnauwkeurigheden in de flux bepalingen blijven bestaan. In Hoofdstuk
7 worden enkele aanbevelingen gedaan om in toekomstig onderzoek de kwaliteit van metabole
flux patronen te verbeteren.
193
Publications
List of publications
Kleijn, R.J., Liu, F., van Winden, W.A., van Gulik, W.M., Ras, C., Heijnen, J.J. (2007) Cytosolic
NADPH metabolism in penicillin-G producing and non-producing chemostat cultures of
Penicillium chrysogenum. Metab. Eng. 9, 112-123.
Kleijn, R.J., Geertman, J.M., Nfor, B.K., Ras, C., Schipper, D., Pronk, J.T., Heijnen, J.J.,
van Maris, A.J., van Winden, W.A. (2006) Metabolic flux analysis of a glycerol-overproducing
Saccharomyces cerevisiae strain based on GC-MS, LC-MS and NMR-derived 13C-labelling
data. FEMS Yeast Res. (Epub)
Kleijn, R.J., van Winden, W.A., Ras, C., van Gulik, W.M., Schipper, D., Heijnen, J.J. (2006)
13
C-labeled gluconate tracing as a direct and accurate method for determining the pentose
phosphate pathway split ratio in Penicillium chrysogenum. Appl. Environ. Microbiol. 72, 47434754.
Nasution, U., van Gulik, W.M., Kleijn, R.J., van Winden, W.A., Proell, A., Heijnen, J.J. (2006)
Measurement of intracellular metabolites of primary metabolism and adenine nucleotides in
chemostat cultivated Penicillium chrysogenum. Biotechnol. Bioeng. 94, 159-166.
Kleijn, R.J., van Winden, W.A., van Gulik, W.M., Heijnen, J.J. (2005) Revisiting the 13C-label
distribution of the non-oxidative branch of the pentose phosphate pathway based upon kinetic
and genetic evidence. FEBS J. 272, 4970-4982.
van Winden, W.A., van Dam, J.C., Ras, C., Kleijn, R.J., Vinke, J.L., van Gulik, W.M., Heijnen,
J.J. (2005) Metabolic-flux analysis of Saccharomyces cerevisiae CEN.PK113-7D based on
mass isotopomer measurements of 13C-labeled primary metabolites. FEMS Yeast Res. 5,
559-568.
Osinga, R., Kleijn, R., Groenendijk, E., Niesink, P., Tramper, J., Wijffels, R.H. (2001)
Development of in vivo sponge cultures: particle feeding by the tropical sponge Pseudosuberites
aff. andrewsi. Mar. Biotechnol. (NY). 3, 544-554.
195
Curriculum Vitae
Curriculum Vitae
Roelco Kleijn was born on 20th of July 1977 in Vlaardingen, the Netherlands. He attended the
International School of Manila, the Philippines, and after his graduation he started with the
BSc and MSc program of Bioprocestechnology at Wageningen University, the Netherlands.
He performed a 7-month internship at the Food and Bioprocess Engineering Group, where he
modeled and cultured sponges and micro-algae in an airlift-loop reactor, under supervision
of Dr. R. Osinga and Prof. R. Wijffels. This work was followed by a second internship at the
Biochemistry Group, where he performed optimization studies on the production of 3-phenyl1-propanol by Pyrococcus furiosus, under supervision of Dr. E. van de Ban and Dr. H. Haaker.
His final internship was carried out at Solvay Pharmaceuticals, Weesp, where he worked on
adsorption of CTAB with amberlite XAD-4 during the production of a subunit influenza vaccine
under the supervision of Dr. J. Louwerens.
After receiving his MSc degree in February 2002, he started as a PhD student at the Department
of Biotechnology, Delft University of Technology, the Netherlands under supervision of Prof.
J. Heijnen. During his PhD he worked on the development and application of 13C labeling
techniques, the results of which are described in this thesis. From November 1st 2006 he is
employed as a postdoctoral researcher at the Institute of Molecular Systems Biology of ETH
Zürich, Switzerland.
197
Dankwoord
Hoe zou het zijn als op zekere dag of nacht een demon uw eenzaamste eenzaamheid binnendrong en
zei: ‘Dit leven zoals je het nu leeft en tot dusver geleefd hebt, zul je nogmaals en dan nog ontelbare keren
opnieuw moeten leven, en er zal niets nieuws zijn; elke pijn, genieting en zucht, het allerkleinste zowel
als het grootste in je huidige leven zal wederkeren, alles in dezelfde maat en volgorde...’ Zoud u zichzelf
dan tandenknarsend op de knieën werpen en de demon vervloeken om deze vreselijke influistering? Of
hebt U ooit een overweldigend moment gekend waarin zulk een demon geantwoord zou hebben: ‘U bent
waarlijk een god, nooit hoorde ik iets goddelijkers!’ (...) Hoezeer zoud u zichzelf en uw leven toegenegen
zijn, als u niets met grotere hartstocht verlangde te vernemen dan deze ultieme lotsbezegeling.
Friederich Nietzsche, Die fröhliche Wissenschaft (1882)
Waarom ben je AIO geworden? Dat is een vraag die me de afgelopen jaren regelmatig
gesteld is door de mensen om me heen. En ook een vraag waarop ik zo nu en dan het
antwoord schuldig moest blijven. De zogeheten ‘AIO-dip’ is niet voor niets het grootste cliché
in het leven van een AIO. Toch zijn er ook meerdere malen de ‘overweldigende momenten’
geweest, zoals die hierboven beschreven worden door Nietzsche. Op wetenschappelijk vlak,
maar vooral ook op sociaal vlak, zonder welke dit proefschrift nooit tot stand zou zijn gekomen.
Deze laatste pagina’s zijn dan ook gewijd aan alle vrienden, collega’s en familie die er voor
hebben gezorgd dat ik met veel plezier terugkijk op de afgelopen jaren.
In de eerste plaats wil ik graag mijn promotor Sef bedanken voor zijn goede raad, kritische
blik en steun. Je hebt me altijd vrij gelaten om dat te doen wat ik het meest interessant
vond, maar verloor nooit de rode lijn van het onderzoek uit het oog. Zoals het een goede
prof betaamt, eindigde een werkbespreking met jou altijd in een hoop nieuwe vragen en
gedachtespinsels. Bij het zien van fermentatie data was er altijd één vraag die keer op keer
terugkwam: ‘Kloppen de koolstof en reductiegraad balans?’. Naast Sef, wil ik ook Walter en
Wouter bedanken voor de dagelijkse begeleiding. Walter, je hebt me in het begin enorm op
weg geholpen met je kennis van Penicillium. Daarnaast hielpen je praktische inzichten bij het
opzetten van de experimenten in het lab. Wouter, je was een onuitputtelijke bron van ideeën.
Je nam altijd de tijd om over nieuwe experimenten en ideeën na te denken, ook al was die
tijd niet altijd voorhanden. Ik heb me een aantal keren serieus afgevraagd wat er van mijn
proefschrift geworden zou zijn als jij na je promotie niet in Delft was gebleven. Daarnaast kon
ik ook altijd aankloppen voor morele ondersteuning, bedankt!
Naast begeleiding was er ook veel ondersteuning in het lab. Jan en Cor, bedankt voor
alle isotopomeer bepalingen die jullie voor mij hebben uitgevoerd. Jullie hebben aan elk
wetenschappelijk hoofdstuk in dit proefschrift bijgedragen. Ko, met jouw kennis ging het
opzetten van de fermentaties een stuk makkelijker. Zonder jouw hulp zou het percentage
succesvolle fermentaties een stuk lager liggen. Daarnaast wil ik graag Angie bedanken voor
de penicilline analyses en Dick voor de NMR metingen. Dick, het heeft even geduurd voordat
de NMR-metingen tot hun recht kwamen, maar uiteindelijk hebben ze wel tot een mooie
vergelijking tussen de verschillende 13C-meetmethoden geleid. Furthermore, I would like to
thank all (former) BPT-colleagues: Liang, Mlawule, Lodewijk, Hilal, Emrah, Andre, Penia,
Uly, Roeland, Michiel, Zheng, Frederik, Sergio and Reza, for the advices, help and good
atmosphere. Thanks also go out to ‘my’ students Feng and Kungah: your contributions have
each ended up in a chapter. I learned a great deal by supervising both of you, hope the same
holds for you. Special thanks to my pen-buddies in the lab. Uly, I have great respect for the
way you combined your work with being a mom. Diana, je was altijd in voor een praatje,
199
Dankwoord
200
wetenschappelijk of niet. Daarnaast heb je me met raad en daad bijgestaan tijdens de enzym
bepalingen. En nee, Februari in Zürich betekent niet alleen mist, regen en sneeuw!
Behalve het werk aan Penicillium heb ik tijdens mijn promotie ook een klein uitstapje
gemaakt naar gist. Jan-Maarten, Ton en Jack, bedankt voor de prettige samenwerking. Wat
in eerste instantie een incoherente dataset leek te zijn, is uiteindelijk toch verworden tot een
mooi hoofdstuk. En dan te bedenken dat de oorspronkelijke invalshoek van de experimenten
compleet anders was.
Naast de wetenschap was er gelukkig ook genoeg tijd voor ontspanning in het lab. De
vele praatjes op de gang, tijdens de koffiepauzes en lunch hielpen me altijd om de boel
enigszins te relativeren. Alle mensen die zijn aangeschoven tijdens de koffiepauzes op de
2e verdieping, bedankt voor alle smeuïge verhalen en roddels. Ik vrees alleen dat een aantal
van deze verhalen nog terug gaat komen in de toekomst. Het woord papaja heeft in ieder
geval een compleet andere betekenis gekregen..... Ik denk ook met veel plezier terug aan
de organisatie van Biotour 2003. Hoogtepunten uit Canada: inwijding in de Chinese keuken
door Liang en Xiaonan, de fabuleuze routeboekjes van Merle, buffet bij de Nederlandse
ambassadeur gecombineerd met hotdogs bij Hooters..... I also have good memories of the
weekly football games on Monday evening with ‘Op ’t randje’. Due to the ‘sports-card affair’
the ending was rather abrupt…but I really enjoyed playing with you guys.
At certain times the turn-over rate of roommates seemed to mimic the metabolic systems
studied in the lab. Thanks to all my roommates for the good memories. A few people I want
to thank in person. peNia, my smiling room-buddy from the beginning, keep in touch! Inés,
thanks for moving to the window …. I’m not such a quiet and serious Dutch guy after all hè?
Rutger and Marija, enjoy the peace and quietness in the room…. heater on 5?
Helmer, Floris, Jasper, Guido, Marjolein, Sipke, Cindy, Remco, Jeroen en Natalia bedankt
voor alle gezellige etentjes, feestjes, weekendjes weg en wat nog niet meer. Ook al waaieren
we steeds verder uit over Nederland en daarbuiten, laten we proberen een aantal tradities
in stand te houden: watjes weekend, ski vakanties, mannen weekend, bbqs, oud en nieuw
(hmm... dat is er dit jaar niet van gekomen). Ik reken erop dat jullie langskomen in Zürich!
To my Spanish-speaking friends, Inés, María and Carol: Hicisteis del último año en el
laboratorio el más inolvidable de todos. A pesar de que sólo nos hemos conocido durante un
año os considero uno de mis mejores amigos. Cada uno de vosotros ocupa un lugar especial
en mi corazón. Thanks for all the Spanish lessons (Tiempo de mierda or mierda de tiempo?),
bullshitting, botanical garden walks & talks, drinks at Belvédère, diners and moral support.
Mijn paranimfen, Jasper en María, bedankt voor alle hulp nu ik hier in Zürich zit. Ik ben trots
dat jullie aan mijn zijde willen staan tijdens de promotie. En bedenk dat als ik kan promoveren,
jullie dat zeker kunnen! Zijn jullie er al uit wie de stellingen gaat voorlezen?
Paps en mams, bedankt voor jullie onvoorwaardelijke steun. Het is een onbeschrijfelijk
goed gevoel om te weten dat ik altijd bij jullie terecht kan. Gewoon even lekker uitwaaien in
het hoge Noorden en gedachten op nul. Willemijn, Els, Bernd, Sebas, Henk, Diny, Madyjan,
Angelique, Björn, jullie waren altijd geïnteresseerd naar mijn vorderingen.... nu is het eindelijk
af! Ik hoop jullie allemaal snel in Zürich te zien.
Jacq, bedankt voor wie je bent. Zonder jouw steun, geduld en hulp zou dit boekje er nooit
zijn geweest, dat weet ik zeker. De tijd is voor ons tijdloos ...
- Roelco -