Microbiology (2014), 160, 2780–2793
DOI 10.1099/mic.0.082180-0
Metabolic impact of an NADH-producing
glucose-6-phosphate dehydrogenase in
Escherichia coli
K. Olavarria,1 J. De Ingeniis,2 D. C. Zielinski,3 M. Fuentealba,1 R. Muñoz,1
D. McCloskey,3 A. M. Feist3,4 and R. Cabrera1
Correspondence
1
Ricardo Cabrera
2
[email protected]
Adam M. Feist
Facultad de Ciencias, Universidad de Chile, Santiago, Chile
Sanford-Burnham Medical Research Institute, La Jolla, CA, USA
3
Department of Bioengineering, University of California San Diego, San Diego, CA, USA
[email protected]
4
Received 3 July 2014
Accepted 19 September 2014
In Escherichia coli, the oxidative branch of the pentose phosphate pathway (oxPPP) is one of the
major sources of NADPH when glucose is the sole carbon nutrient. However, unbalanced
NADPH production causes growth impairment as observed in a strain lacking
phosphoglucoisomerase (Dpgi). In this work, we studied the metabolic response of this bacterium
to the replacement of its glucose-6-phosphate dehydrogenase (G6PDH) by an NADH-producing
variant. The homologous enzyme from Leuconostoc mesenteroides was studied by molecular
dynamics and site-directed mutagenesis to obtain the NAD-preferring LmG6PDHR46E,Q47E.
Through homologous recombination, the zwf loci (encoding G6PDH) in the chromosomes of WT
and Dpgi E. coli strains were replaced by DNA encoding LmG6PDHR46E,Q47E. Contrary to some
predictions performed with flux balance analysis, the replacements caused a substantial effect on
the growth rates, increasing 59 % in the Dpgi strain, while falling 44 % in the WT. Quantitative
PCR (qPCR) analysis of the zwf locus showed that the expression level of the mutant enzyme was
similar to the native enzyme and the expression of genes encoding key enzymes of the central
pathways also showed moderate changes among the studied strains. The phenotypic and qPCR
data were integrated into in silico modelling, showing an operative G6PDH flux contributing to the
NADH pool. Our results indicated that, in vivo, the generation of NADH by G6PDH is beneficial or
disadvantageous for growth depending on the operation of the upper Embden–Meyerhof
pathway. Interestingly, a genomic database search suggested that in bacteria lacking
phosphofructokinase, the G6PDHs tend to have similar preferences for NAD and NADP. The
importance of the generation of NADPH in a pathway such as the oxPPP is discussed.
Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby,
Denmark
INTRODUCTION
Escherichia coli has different metabolic pathways to oxidize
available carbon sources, but each of them has a different
yield of reduced cofactors NADH and NADPH. The strong
segregation in the cofactor specificity among the dehydrogenases of the central pathways suggests that the coupling of
the generation of either NADH or NADPH with certain
metabolites could be a key property for network function.
Abbreviations: DW, dry weight; FRT, FLP recognition target; G6PDH,
glucose-6-phosphate dehydrogenase; MD, molecular dynamics; oxPPP,
oxidative branch of the pentose phosphate pathway; qPCR, quantitative
PCR.
Supplementary material is available with the online Supplementary
Material.
2780
For example, generation of pentose phosphate is coupled
with NADPH production, whereas pyruvate synthesis is
coupled with NADH generation through the Embden–
Meyerhof pathway.
In E. coli, the dehydrogenases involved in the oxidative
branch of the pentose phosphate pathway (oxPPP) have
been regarded as NADP-specific enzymes. Accordingly, this
pathway is considered the source of ~30–40 % of total
NADPH production when glucose is the sole carbon source
(Sauer et al., 2004). The generation of NADPH must be
balanced with its consumption to maintain optimal growth
rates. For example, phosphoglucoisomerase (Dpgi) mutants
grown on glucose as the sole carbon source exhibit reduced
growth rates (Canonaco et al., 2001). In this case, all the
glucose is channelled through oxPPP, leading to an
Downloaded from www.microbiologyresearch.org by
082180 G 2014 The Authors
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Printed in Great Britain
Effect of an NADH-producing G6PDH in E. coli
unbalanced production of NADPH, and growth rates fall
~75 % with respect to WT. An additional factor in the
redox-balancing capacity of E. coli is the presence of two
kinds of transhydrogenases (Fuhrer & Sauer, 2009; Sauer
et al., 2004). For example, it was observed that the
overexpression of the transhydrogenase UdhA helps to
dissipate the NADPH imbalance in the Dpgi mutant,
increasing the growth rate of this mutant strain (Canonaco
et al., 2001). Given the balancing capacity provided by the
transhydrogenases, it is not clear why the dehydrogenases
in the oxPPP from E. coli have evolved to be specific for
NADP. It is also not clear why NADPH production is
coupled with carbon flux through the oxPPP. We believe
that the high cofactor specificity of the dehydrogenases
from oxPPP in E. coli is related to the location of these
enzymes after the branching points between the major
glycolytic pathways, enabling the regulation of how much
carbon flux is coupled to the generation of NADPH.
Consequently, the cofactor specificity of these dehydrogenases should affect the magnitude of the carbon flux
through these pathways. To test this hypothesis, we depend
on the utilization of a dehydrogenase from the oxPPP
showing a strong preference for NAD instead of NADP. As
we are introducing a single and well-defined stoichiometric
change in the metabolic network, we believed flux balance
analysis could be a suitable tool to predict the physiological
impact of this kind of change.
The preference of the glucose-6-phosphate dehydrogenase
(EC 1.1.1.49) from E. coli (EcG6PDH) for NADP over NAD
has been quantified, showing a specificity constant (kcat/
Km) for NADP 410 times higher than for NAD (Olavarrı́a
et al., 2012). The homologous enzyme from Leuconostoc
mesenteroides (LmG6PDH) has also been thoroughly studied
in terms of structural and functional aspects, with both NAD
and NADP (Naylor et al., 2001; Vought et al., 2000). In the
present research, we aimed to obtain a G6PDH variant with
a higher discrimination power against NADP, use it to replace
the native EcG6PDH, and evaluate the metabolic impact of
this change. For this reason, we took the well-studied
LmG6PDH as a starting point to perform molecular dynamics
(MD) simulations of the interactions between this enzyme
and the cofactors NAD and NADP, envisioning the design of
the point mutations needed to obtain an NAD-preferring
G6PDH.
There are previous observations with glyceraldehyde 3-phosphate dehydrogenase, glutamate dehydrogenase, and isocitrate dehydrogenase indicating that cofactor specificity is a
key factor in determining the distribution of metabolic fluxes
and/or growth in bacteria (Martı́nez et al., 2008; Marx et al.,
1999; Zhu et al., 2005). In spite of its clear contribution to
redox balance and growth, in E. coli, the oxPPP has not been
studied in this regard.
In this work, we generated an NAD-preferring G6PDH and
replaced the native EcG6PDH with the engineered protein.
The effects of this change in WT and Dpgi genetic backgrounds were compared. We examined the changes in the
http://mic.sgmjournals.org
expression of various genes encoding key enzymes of the
central pathways and the growth rates during aerobic growth
on glucose. We compared predictions performed by flux
balance analysis with the experimental results and found
that in spite of the prediction, the flux through the oxPPP
was active after the replacement. A search in genome
databases suggested that the cofactor specificity in G6PDH is
related to the bifurcation of the carbon flux between the
Embden–Meyerhof pathway and the oxPPP.
METHODS
MD analysis. MD simulations were performed using a model of the
3D structure of LmG6PDH in complex with NADP (Protein Data
Bank ID: 1H9A) (Naylor et al., 2001). The in silico mutations R46E
and Q47E were generated using the Mutagenesis plugin implemented
in PyMOL (http://www.pymol.org/) over the 1H9A structure and were
submitted to 150 000 steps of conjugate gradient energy minimization prior to the MD simulations. Starting from these structures we
generated the NAD complexes by removing the 29-phosphate group
of NADP in each case, leaving a hydroxyl group. The complexes were
simulated under explicit solvent (i.e. simulating the presence of water
molecules in the systems) using NAMD 2.9 software (Phillips et al.,
2005) and the Amber99SB-ILDN force field (Lindorff-Larsen et al.,
2010). The systems were neutralized using Na+ counter-ions. The
parameters and topology for the NAD and NADP were taken from
simulations performed over alcohol dehydrogenase (Ryde, 1995) and
lactate dehydrogenase (Holmberg et al., 1999), respectively. The
simulations were performed for 10 ns using a 1 fs timestep measuring
the non-bonded interactions in a radius of 12 Å every 2 fs. The
simulated temperature was 300 K without additional user-defined
restraints. The calculations were performed on a 64-core AMD
Opteron server.
VMD 1.9.1 software (Humphrey et al., 1996) was used to analyse the
trajectories. The interactions described in Fig. 1 were quantified from
the distances between the following pairs of atoms (protein and
ligand): (I) the main-chain nitrogen hydrogen of position 46 and the
N3A nitrogen in the adenine moiety; (II) any nitrogen hydrogen in
the side-chain of R46 and the 29-oxygen of NAD or any oxygen of the
29-phosphate group of NADP (WT systems); any oxygen in the sidechain of E46 and the 29-oxygen of NAD or any oxygen of the 29phosphate group of NADP (mutant systems); (III) the main-chain
nitrogen hydrogen of T14 and 39-oxygen in the ribose moiety; (IV)
the c-oxygen of T14 and 39-hydrogen of the ligand (NAD systems);
the c-hydrogen of T14 and either 39-oxygen or any oxygen in the
29-phosphate (NADP systems); and (V) the e-oxygen of residue 47
and the c-hydrogen of T14. These distances were evaluated every
50 ps; one interaction was counted when the distance between the
involved atoms was ,2.5 Å.
Enzyme mutagenesis and purification. The double-mutant
R46E,Q47E of LmG6PDH was obtained by site-directed mutagenesis
using the plasmid pET25-LmG6PDH (lacO, lacI, ColE1, AmpR) as the
template. Dr Michael Cosgrove (Syracuse University, New York, NY,
USA) kindly donated this plasmid and it contained the amino acid
codifying sequence for the WT LmG6PDH (GenBank accession
number M64446.1). The primers employed were: 59-CATTGTTGGAACGGCCCAGGAAGCCCTCAATGATG-39 and 59-GGCCGTTCCAACAATGGCAAAATGCTTTTGCAAG-39. The bold bases in the first
primer denote the intended mutations. The mutant sequence was also
cloned in plasmid pTEV (Rocco et al., 2008) to facilitate the
overexpression and His-tag-based purification of the mutant enzyme.
E. coli BL21-DE3 cells transformed with the derived plasmid were
aerobically grown in Luria–Bertani (LB) medium at 37 uC to OD600
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2781
K. Olavarria and others
Time (ns)
10
II
8
I
I
II
6
4
IVa IVb
IV
III
III
V
2
V
0
LmG6PDH
LmG6PDHR46E,Q47E
Interactions with NAD I
II III IV V
I
II III IV V Interactions with NADP
0
LmG6PDH
LmG6PDHR46E,Q47E
2
I
I
4
II
II
III
6
IV
IVa IVb
8
V
III
V
10
Time (ns)
Fig. 1. Network of interactions with the adenosine moiety of the cofactor at the active site of WT and double-mutant
LmG6PDH. The graphs show the occurrence, along 10 ns molecular simulation, of the five specific interactions labelled in the
molecular schemes adjacent to each graph. Interactions I and II involve main- and side-chain atoms of residue 46, respectively.
Interactions III and IV involve main- and side-chain atoms of residue 14, respectively. Interaction V corresponds to a hydrogen
bond between the side-chains of residues 14 and 47. In the case of NADP (right half of the panel), the hydrogen bond
described by interaction IV in the WT system is lost in the double mutant and the side-chain of T14 rearranges to bind an
oxygen atom in the 29-phosphate in the cofactor ligand.
0.5. After 4 h of induction with 1 mM IPTG (US Biologicals), the
cells were harvested by centrifugation (10 min, 2000 g, 4 uC),
resuspended in 50 mM cold Tris (Merck), 10 mM MgCl2 (Merck)
and 1 mM PMSF (Calbiochem), pH 8.2, and sonicated. The cellular
extract was centrifuged for 30 min at 80 000 g, 4 uC and the
supernatant was inoculated in a HiTrap Q HP (GE Healthcare)
ion-exchange column previously equilibrated with a buffer of 50 mM
Tris, 10 mM MgCl2 and 40 mM NaCl (Merck), pH 8.2. The bound
proteins were fractionally eluted by increasing the concentrations of
NaCl inside the column from 40 mM to 1 M. The fractions with the
highest G6PDH activities were supplemented with 40 mM imidazol
(Merck) and inoculated in a HisTrap (GE Healthcare) column
previously charged with NiSO4 (Merck) according to the manufacturer’s
instructions. The bound proteins were eluted using a buffer 50 mM Tris
and 10 mM MgCl2, pH 8.2, supplemented with increasing concentrations of imidazol from 40 to 600 mM. Before setting the proteolytic cut
to remove the His6 tail from the molecules of LmG6PDHR46E,Q47E, the
fractions with the highest activities were pooled and dialysed against the
buffer 25 mM Tris/HCl, pH 8.0, 50 mM NaCl and 1 mM EDTA to
chelate the soluble nickel that interfered with the activity of the protease
2782
TEV. A home-made preparation of the protease TEV, expressed from
plasmid pRK793 (Addgene), which had its own poly-His tail, was
utilized. The proteolysis was performed for 1 h at 25 uC with 100 : 1
(w/w) substrate/protease and tris(2-carboxyethyl)phosphine (SigmaAldrich) was added up to a final concentration of 1 mM (bmercaptoethanol provoked the precipitation of the protease TEV).
After the digestion, the protein mixture was inoculated again in a
HisTrap column, and the unbound protein was collected whilst the Histagged TEV and the His6 tails remained bound to the column. The
purity of the protein obtained was checked by running a sample with
10–20 mg total protein in SDS-PAGE and staining with Coomasie blue
(Merck). The effectiveness of the proteolytic cut was checked by
Western blotting using an Ni-NTA HRP Conjugate kit (Qiagen). The
purified LmG6PDHR46E,Q47E was concentrated to a final concentration
of 1 mg ml21 after the addition of 50 % (v/v) pure glycerol (Merck) and
stored at 220 uC until the kinetic assays were performed.
Enzyme kinetics. Just before the assays, a sample of pure
LmG6PDHR46E,Q47E was dialysed against a buffer of 100 mM Tris,
10 mM MgCl2 and b-mercaptoethanol, pH 8.0, at 25 uC. The assay
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
buffer had similar characteristics: it was supplemented with the
substrates and did not contain b-mercaptoethanol. The substrates
NAD, NADP and glucose-6-phosphate were prepared for the kinetic
assays as described previously (Olavarrı́a et al., 2012). The minimal
enzymatic concentration that could be used safely during the assays
without appreciable inactivation was determined using the test
described previously (Selwyn, 1965). The reactions were triggered by
the addition of glucose-6-phosphate. The velocities were assessed by
monitoring the formation of NAD(P)H following the temporal changes
in A340 at 25 uC in a Synergy 2 (Biotek) UV/visible spectrophotometer.
The protein concentration was determined using a Bio-Rad protein
assay taking the dialysis buffer as the blank. The kinetic parameters of
the mutant LmG6PDHR46E,Q47E were determined by an initial velocities
approach. We set up the conditions to minimize the accumulation of
products during the mixing before beginning the measurements. All the
kinetic assays were performed using low-protein-absorption 96-well
plates (Nunc).
Generation of mutant strains. The FRT-Kanr-FRT cassette [contain-
ing the FLP recognition target (FRT)] from the plasmid pKD13
(Datsenko & Wanner, 2000) was amplified using ad hoc designed
primers that included restriction sites for the endonucleases XhoI and
SalI. Both the amplified FRT-Kanr-FRT cassette and the plasmid pETLmG6PDHR46E,Q47E were cut with the endonucleases XhoI and SalI.
After gel purification, the cassette and the cleaved vector were joined
using a Quick Ligation kit (New England Biolabs). The resultant
plasmid had the FRT-Kanr-FRT cassette just downstream of the aminoacid-encoding sequence for LmG6PDHR46E,Q47. The ligation product
was used to transform E. coli TOP10 cells (Invitrogen) to propagate the
resultant plasmid. The cells transformed with the resultant plasmid
gained resistance to kanamycin conferred by the Kanr gene. The
correctness of the ligation was checked by DNA sequencing. Following
a routinely used recombination protocol (Datsenko & Wanner, 2000),
the second step was to generate, by PCR, a linear DNA molecule with
the sequence LmG6PDHR46E,Q47E-FRT-Kanr-FRT flanked with nucleotide sequences identical to the chromosomal 36 bp upstream and
the 36 bp downstream of the 1476 bp native amino-acid-encoding
sequence for the enzyme EcG6PDH.
The parental strains for the recombination procedure were: WT
(MG1655 K12) and Dpgi (MG1655 K12 Dpgi; Charusanti et al., 2010).
The linear DNA molecule obtained was purified and introduced into
electrocompetent cells from the parental strains transformed previously with plasmid pKD46 (Datsenko & Wanner, 2000). To
accomplish the recombination, the electrocompetent cells were grown
to OD600 0.4–0.5 at 30 uC and the expression of the Red recombinase
system was induced with the addition of 10 mM arabinose to the
culture for 1 h before the treatment to transform the bacteria into
electrocompetent cells. It is highly recommended to introduce the
linear DNA into freshly prepared electrocompetent cells. The
recombinant bacteria were detected by their ability to grow in LBagar-kanamycin. The Kanr gene was further eliminated according
with a procedure described previously (Datsenko & Wanner, 2000),
and the sensitivity of the resultant strain to ampicillin, kanamycin and
chloramphenicol was checked by negative selection in plates with LBagar supplemented with the corresponding antibiotics. Colony PCR,
chromosomal DNA sequencing, quantitative PCR (qPCR) and measurement of G6PDH specific activity in the cellular extracts were useful to
confirm the success of the recombination procedure. A detailed
description of the method for homologous recombination can be
found elsewhere (Datsenko & Wanner, 2000). In this way, the
following recombinant strains were obtained: NAD-G6PDH (MG1655
K12 Dzwf : : zwfLmG6PDHR46E,Q47E), inact-G6PDH (MG1655 K12
Dzwf : : zwfLmG6PDHD16GR46E,Q47E) and Dpgi-NAD (MG1655 K12 Dpgi
Dzwf : : zwfLmG6PDHR46E,Q47E), where inact-G6PDH refers to a strain
bearing an inactive form of G6PDH.
http://mic.sgmjournals.org
All the resultant strains were whole-genome sequenced and no other
unintended mutations were detected.
Specific activities in cell extracts. The cells obtained by centri-
fugation (2500 g, 10 min) during the exponential growth phase in LB
were washed twice with ice-cold 100 mM Tris, 10 mM MgCl2 and
5 mM NaCl, pH 8.0, and centrifuged again (2500 g, 10 min). The
resultant pellets were dissolved in the same buffer, but this time
supplemented with a protease inhibitor cocktail (Roche). The cells
were disrupted by sonication and the micelles were centrifuged at
20 000 g for 20 min at 4 uC. The supernatants were rescued and
placed on ice, and the pellets were discarded. The specific activities
were checked by adding aliquots of the supernatants to defined
volumes of a buffer of 50 mM Tris, 10 mM MgCl2, and 1 mM glucose6-phosphate, pH 8.0, supplemented with NAD or NADP at 0.4, 1 or
3 mM, and monitoring the initial rate of formation of NAD(P)H
following the changes in the absorbance at 340 nm at 25 uC. A coupled
assay was used to determine the specific activity of phosphoglucoisomerase (Salas et al., 1965). The concentration of protein in the
cellular extracts was estimated using a commercial protein assay (BioRad). The enzymic activities were normalized by the concentration of
total proteins in the respective cellular supernatants.
Growth conditions and sampling. During the genetic manipula-
tions and the assessment of the enzymic specific activities in the
cellular extracts, the strains were grown up in LB rich medium with or
without agar, supplemented with ad hoc antibiotics. All physiological
experiments were done by growing the cells in full aerobic conditions,
with vigorous agitation (300 r.p.m.), at 37 uC in M9 minimal medium
supplemented with 0.2 mm-filtered glucose as the sole carbon
source (4.5 g l21) (M9-glucose). M9 minimal medium was prepared
as described previously (Fuhrer & Sauer, 2009). All experiments were
performed in biological triplicates and every measurement in a
biological replicate had technical triplicates.
In all the experiments, the starting biological material was bacteria
sampled from isolated colonies previously grown up overnight in
50 ml M9-glucose before being inoculated into the flasks for the
experiments. To measure the growth rate, the glucose uptake rate
and the acetate excretion rate, samples were taken every 30 min to
measure the OD600, and broth samples were filtered using a 0.2 mm
filter. These samples were immediately chilled at 280 uC until
analysis. When the cultures reached OD600 1.0, samples of biomass
were treated and saved for measurement of mRNA. Samples of
extracellular media and measurements of optical density were taken
until the cultures approached the stationary phase.
qPCR. An aliquot of 1 ml broth from each culture was added to 2 vol.
RNAprotect Bacterial Reagent (Qiagen), and the total RNA was isolated
by using RNeasy columns (Qiagen) and treated with DNase I (Promega).
Total RNA yields were measured by quantifying A260, and quality was
checked by visualization on agarose gels and by measuring the A260/280.
cDNAs were obtained by using SuperScript II reverse transcriptase
(Invitrogen). The transcript levels of the genes ackA, gapA, gnd, icd,
maeA, maeB, mdh, nadk, pgi, pntA, udhA and zwf were determined by
qPCR (SYBR GreenER qPCR SuperMix Universal; Invitrogen).
The reaction mixtures were monitored on an MX3000P (Stratagene)
qPCR analyser with the following protocol: denaturation at 95 uC for
10 min, followed by 40 cycles of 95 uC for 30 s, annealing at 56 uC for
60 s and extension at 72 uC for 30 s. The specificity of the products
was verified by melting curves analysis. The levels of the target
mRNAs were normalized to the level of mRNA encoding the
housekeeping protein DnaA. The sequences of the primers used can
be found in the Supplementary Material.
Computational systems analysis of physiological and qPCR
data. The latest comprehensive genome-scale model of E. coli,
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2783
K. Olavarria and others
iJO1366, was used as the WT default Systems Biology Markup
Language model (Orth et al., 2011).
Modelling was performed by solving flux balance analysis problems
with targeted constraints. Generally, the flux balance analysis problem
is a linear optimization problem formulated as:
T
min c v
s.t.
Sv50
vlb,v,vub
A number of existing expression integration methods exist, as recently
reviewed (Machado & Herrgård, 2014). These methods have been
shown to provide little increased accuracy in flux prediction from
incorporation of gene expression data. For this reason, we did not
attempt to quantitatively predict flux levels, a deviation from the goal of
previous methods, but instead only sought to determine the possible
systems-level impact of observed gene expression changes, taking into
account the limitations of expression as a proxy for flux levels.
Phenotypic measurements. Glucose and acetate concentrations in
where c is the objective vector containing the reaction to be maximized
or minimized (for negative and positive elements, respectively), v is the
vector of fluxes, S is the stoichiometric matrix, and vlb and vub are the
lower and upper flux bounds, respectively. Additional constraints to
model the effects of gene deletions can be added by constraining flux
bounds to zero for reactions catalysed by deleted genes. The formalism
was previously described in detail (Orth et al., 2010). All the
computational simulations were performed with COBRA Toolbox 2.0.3
(Schellenberger et al., 2011) in MATLAB.
For qPCR and robustness analysis, a mutant G6PDH model with the
cofactor specificity of G6PDH changed from NADP to NAD was
created. Robustness analyses were performed using the methodology
developed previously (Varma et al., 1993) as implemented in the
COBRA Toolbox (Schellenberger et al., 2011). Additional constraints
were added to specific reactions (transhydrogenase and phosphoglucoisomerase) as indicated in the Results and Discussion during the
analysis. A G6PDH knockout was simulated by constraining the
G6PDH reaction to have zero flux. For the in silico robustness
analysis, a glucose uptake rate of 8 mmol [g dry weight (DW)]21 h21
and oxygen uptake rate of 10 mmol gDW21 h21 was assumed. qPCR
measurements were then incorporated as follows.
The WT flux state was first calculated by using flux balance analysis
with the WT model, by maximizing ATP production, and
subsequently minimizing the length of the flux vector necessary to
achieve the measured growth rate given measured glucose uptake and
acetate excretion. To enforce known flux splits, previously measured
fractions of NADPH production from the membrane-bound transhydrogenase and flux through oxPPP and the NADP-specific malic
enzyme were constrained in the WT model (Perrenoud & Sauer, 2005;
Sauer et al., 2004). Then, to estimate the mutant flux state in each
case, for each one of the genes with its level of expression measured
using qPCR, the mutant flux was constrained to be equal to the WT
flux multiplied by the ratio of mutant/WT glucose uptake and the
ratio of mutant/WT gene expression. This procedure was performed
based on the assumption that expression changes correlate directly
with flux changes, which has obvious exceptions but is simply
intended to provide a systems analysis of the possible integrated
effects of gene expression changes (Fig. S1). The reaction catalysed by
6-phosphogluconate dehydrogenase was excluded from analysis in the
G6PDH ‘knockout’ model as the oxPPP was assumed to not carry
flux once G6PDH was inactive. Glyceraldehyde 3-phosphate dehydrogenase was left unconstrained in all models despite the availability
of qPCR measurements because constraining flux based on gene
expression was found to lead to an unfeasible flux solution. We found
this to be due to an inflexibility of glycolysis due to the lack of alternate
pathways that could sustain energy production and metabolite
secretion production given a downregulation of the glycolytic enzyme
glyceraldehyde 3-phosphate dehydrogenase in the mutant strain. Flux
solutions for mutant strains were then calculated as with the WT model
by maximizing ATP production and subsequently minimizing the
length of the flux vector necessary to satisfy all constraints. This choice
2784
of objective is most consistent with both physiological data and reports
that bacterial flux states tend to fall near a Pareto surface determined by
the competing objectives of growth, energy production and minimal
use of enzymes (Schuetz et al., 2012).
the extracellular filtrates were analysed by HPLC using an Aminex
HPX-87H (Bio-Rad) ion-exchange column at 65 uC and an IR HP
1047A (Hewlett Packard) detector. No other organic compound was
detected consistently. The mobile phase was degassed with 5 mM
H2SO4 at a flow rate of 0.5 ml min21. The concentration was assessed
by comparison with standards of known concentrations run previously
under the same conditions. For the conversion of the optical density
data to gDW, we previously studied the relationship between these two
variables, observing a linear relationship for values of OD600,0.6.
Therefore, the broth samples with OD600.0.5 were diluted and measured
again, considering the dilution factor (Neidhardt et al., 1990). From this
study we obtained the conversion factor 0.45 gDW l21 OD21.
In the exponential growth phase, the temporal evolution of the dry
weight DW(t) is described by the equation:
lnDW(t)5kt+lnDW0
where DW0 is the initial dry weight, k is the specific growth rate of the
strain and t is time (the independent variable). Given this linear form,
it was possible to statistically compare the specific growth rates
(slopes) of the different strains by an analysis of covariance, with the
null hypothesis of no difference among the specific growth rates.
For the concentration of any metabolite detected in the extracellular
media, the instantaneous concentration X(t) is described by:
X(t)5r6[DW(t)/k]+X0
where r is the specific uptake/excretion rate and X0 is the initial
concentration of the metabolite X. The above equation also represents
a line where the slope is the relevant specific uptake/excretion rate, so
it was also possible to statistically compare the glucose uptake rates
and the acetate excretion rates of the different strains.
RESULTS AND DISCUSSION
In this work, we used a multidisciplinary approach to
understand the importance of the reduced cofactor produced
in the reaction catalysed by EcG6PDH. To the best of our
knowledge, the present work is the first study in E. coli where
a dehydrogenase from the oxPPP was replaced by a
homologue with a different specificity. Furthermore, the
inserted gene was not an intact, heterologous gene, but rather
an engineered gene, which we name a heteromutant gene.
Design and kinetic study of an NAD-preferring
G6PDH
According to previous work (Vought et al., 2000), the
presence of a carboxylate group at the side-chain of
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
position 46 or 47 antagonizes the negative charge of the 29phosphate of NADP. We wanted to evaluate to what extent
two negative charges would reinforce the discrimination
against NADP without disturbing the interactions with
NAD. With this aim, we simulated the effect of the doublemutant R46E,Q47E over NAD(P) discrimination by MD.
Trajectories of 10 ns were run for the different enzyme–
cofactor complexes in explicit solvent, with either NAD or
NADP for both LmG6PDH and LmG6PDHR46E,Q47E (Fig.
1). For LmG6PDH, the side-chain atoms of the residues
T14 and R46 establish the network of interactions with the
adenosine moiety of NADP. Conversely, when the 29phosphate is absent from the ligand (NAD molecule), the
interactions are mostly provided by the main-chain atoms
of those residues, together with a stabilizing hydrogen
bond formed between the side-chains of Q47 and T14. In
the case of the mutant, the hydrogen-bond network around
NADP is weak because of the absence of the strong ionic
interaction provided by the guanidinium group of R46 to
the 29-phosphate. Indeed, the simulation shows that the
negative charge of the carboxylate in E46 lies very close
to the phosphate group (interaction II), increasing the
instability of the system. The scheme is not so complicated in the double-mutant active site when it comes to the
interactions network around the adenosine moiety of
NAD. In fact, it seems that the carboxylate of E46 could
establish relatively weak interactions with the 29-hydroxyl
of ribose. These observations suggest that the doublemutant R46E,Q47E will be able to use NAD as the substrate
more effectively than NADP, providing a discrimination
against the 29-phosphate stronger than the single-mutant
R46E (Vought et al., 2000), due to an increased coulombic
repulsion. Thus, we evaluated in vitro the effect of the
proposed double mutation.
Using site-directed mutagenesis, we obtained the doublemutant enzyme LmG6PDHR46E,Q47E, which was studied
kinetically (Table 1, Fig. S2) using either NAD or NADP as
cofactor in the presence of glucose-6-phosphate at 800 mM
(a typical concentration for this metabolite in the cell;
Lowry et al., 1971) or 10 mM. For the coenzyme NADP, a
change was seen from a Km of 8 mM with EcG6PDH to almost
4.8 mM with LmG6PDHR46E,Q47E (Table 1). The change in
Km for NAD was not as drastic as that for NADP; however, it
was more than three times lower than the Km for NAD
of EcG6PDH. The MD simulations predicted that in
LmG6PDHR46E,Q47E the interactions with NAD around the
adenosine moiety would not be severely compromised after
the amino acid replacements. Therefore, the increase in the
Km for NAD in comparison with the unmodified LmG6PDH
could be related to the loss of interactions in other regions of
the molecule or the effect of the presence of the glucose-6phosphate at the active site, which was not considered in the
simulations. The most noteworthy effect of the double
mutation is that the kcat for NADP was almost 13 times lower
than the kcat for NAD (Table 1). With a specificity constant
(kcat/Km) for NAD 37 times higher than for NADP, we almost
doubled the discrimination power against NADP observed in
the mutant with the highest NAD preference obtained
previously (Vought et al., 2000). When we increased the
glucose-6-phosphate concentration to 10 mM to obtain
more saturating conditions for this substrate, we observed
an increase in the specificity constant for NAD up to a value
50 times higher than the specificity constant for NADP (Fig.
S2). Moreover, considering the Km for NADP and the
physiological concentration of this nucleotide, the event of
oxidation of a glucose-6-phosphate by an NADP catalysed by
LmG6PDHR46E,Q47E must be very infrequent under physiological conditions. Therefore, the kinetic properties of the
double-mutant G6PDH allow its use in vivo as an almost
exclusive NADH-generating enzyme.
An inactive form of the double-mutant enzyme was
obtained by the additional mutation D16G that nullifies
the activity of the enzyme with both NAD and NADP. The
D16 side-chain was observed to contact the nicotinamide
amide in the 3D structures of LmG6PDH with both NAD
and NADP (Naylor et al., 2001), as well as in the course of
our simulations (Fig. S3). In the absence of this sidechain, the nicotinamide moiety of the cofactor must be
unable to acquire the correct orientation for hydride
transfer.
Table 1. Apparent kinetic parameters for the double-mutant LmG6PDHR46E,Q47E at 800 mM glucose-6-phosphate, using either NAD
or NADP as the cofactor
For comparison, the corresponding parameters of LmG6PDH and EcG6PDH are shown as reported by Lee & Levy (1992) and Olavarrı́a et al.
(2012), respectively.
NAD
Km (mM)
kcat (s21)
NADP
Km (mM)
kcat (s21)
(kcat/Km)NADP : (kcat/Km)NAD
http://mic.sgmjournals.org
LmG6PDHR46E,Q47E
LmG6PDH
EcG6PDH
1626±243
736±32
162±18
727
5090±400
288±5
4757±724
58±3
0.027
8.0±0.7
337
9.4
7.5±0.8
174±2
410
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2785
K. Olavarria and others
Active LmG6PDHR46E,Q47E is expressed from the
chromosomal locus of zwf
Taking advantage of the homologous recombination mediated
by the Red recombinase from phage lambda (Datsenko &
Wanner, 2000), the amino-acid-encoding sequence from the
chromosomal copy of zwf was replaced by the amino-acidencoding sequence for LmG6PDHR46E,Q47E, preserving the
endogenous cis-regulatory elements (Fig. S4). The replacement was done both in the WT strain and in an isogenic
Dpgi strain. In this way, two recombinant strains (NADG6PDH and Dpgi-NAD-G6PDH) were obtained. A third
recombinant strain (inact-G6PDH) was generated where the
mutant gene (encoding LmG6PDHD16G,R46E,Q47E) should be
transcribed and translated, but the protein was inactive.
Thus, it was expected that the strain harbouring the gene
encoding LmG6PDHD16G,R46E,Q47E would resemble the
metabolic phenotype of a Dzwf strain (Fig. 2c). Cellular
extracts from batch cultures of all the strains studied were
used to test the specific G6PDH activities using either NADP
or NAD as the cofactor. The results indicated that
LmG6PDHR46E,Q47E was translated in an active form and
suggested that it could sustain the flux through the oxPPP in
the recombinant strains while producing NADH instead of
NADPH (Fig. 2f). The specific activities observed with both
cofactors were consistent with the kinetic parameters
obtained for the EcG6PDH (Olavarrı́a et al., 2012) and the
engineered LmG6PDHR46E,Q47E (Table 1, Fig. S2). No
phosphoglucoisomerase activity was detected in the cellular
extracts from Dpgi and Dpgi-NAD-G6PDH strains (data not
shown), confirming that the upper part of the Embden–
Meyerhof pathway is blocked in these strains.
Modelling optimum growth predicts a low impact
for the shift of cofactor specificity of G6PDH over
different genetic backgrounds
Constraint-based modelling, as used here, is particularly
useful for examining the limits of functionality for a
particular organism (Price et al., 2004). Fig. 3 shows a
robustness analysis (Schellenberger et al., 2011) where the
maximum possible growth rate is predicted given various
amounts of flux passing through the G6PDH reaction (either
through the WT G6PDH or through the NAD-specific
G6PDH). Simulations were run in the WT, Dpgi and DpntAB
(the membrane-bound transhydrogenase) backgrounds.
Overall, the impact that the NAD-specific G6PDH had on
the maximal possible growth rate was rather modest: in the
most extreme case (DpntAB background), the growth of the
strain bearing the NAD-specific G6PDH was ~2 % lower
than the value for the strain bearing the NADP-specific
G6PDH form. For the WT background, nearly half of the
glucose-6-phosphate must be oxidized through the oxPPP to
accomplish the maximum growth rate and this maximal
growth rate is only ~1 % higher than the growth rate
attainable when there is no flux through G6PDH. The
simulations also predicted that after the replacement of
the NADP-specific EcG6PDH by the NAD-specific
2786
LmG6PDHR46E,Q47E, any flux through the oxPPP is detrimental to the theoretical maximum growth rate, albeit a
small change. In the case where PntAB supplied 40 % of the
NADPH production, no appreciable changes were expected.
In fact, the membrane-bound transhydrogenase has been
reported to be responsible for between 35 and 45 % of
NADPH production during aerobic batch growth of E. coli
(Fuhrer & Sauer, 2009; Sauer et al., 2004). In other words, the
use of an NAD-specific G6PDH would not be optimal for the
formation of biomass at any flux through oxPPP. Therefore,
we should expect that if the internal fluxes are distributed
such that the formation of biomass is maximal, the flux
through oxPPP will be zero and the metabolic phenotype
of the NAD-G6PDH strain must resemble the metabolic
phenotype of the inact-G6PDH strain.
A shift in the cofactor preference of G6PDH
significantly affects the growth rate and depends
on the genetic background
In order to test the modelling predictions, the growth rates
and the glucose uptake rates were observed first on the Dpgi
and Dpgi-NAD-G6PDH strains, where all the catabolic
carbon flows through the oxPPP when glucose is the sole
carbon source. This was also an in vivo test of functionality
for LmG6PDHR46E,Q47E, as a Dpgi strain will not grow
under these conditions in the absence of flux through the
oxPPP (Sauer et al., 2004). When these strains were grown
aerobically with glucose as the sole carbon source, the
mutant Dpgi-NAD-G6PDH showed a growth rate 59 %
faster than that observed for the parental Dpgi (Table 2).
Canonaco et al. (2001) demonstrated that the impaired
growth rate of the Dpgi strain is caused by the relative
excess of NADPH produced when almost the entire
assimilated glucose flows through the oxPPP. The results
shown here are consistent with the results of Canonaco et al.
(2001) because the substitution of an NADPH-producing
enzyme by an NADH-producing enzyme would partially
relieve the relative excess of NADPH observed in Dpgi,
improving its growth rate. However, the glucose uptake rate
and the growth rate of the Dpgi-NAD-G6PDH strain
remained far from those observed in the WT strain, giving
additional support for the hypothesis suggested by Canonaco
et al. (2001) about the inability of the pentose phosphate
pathway alone to carry a glycolytic flux of the same magnitude
observed in the WT strain. Taken together, the specific
G6PDH activities measured in the cellular extracts and the
growth rate observed in the Dpgi-NAD-G6PDH strain show
that the expressed molecules of LmG6PDHR46E,Q47E were
catalytically capable of sustaining flux through the oxPPP.
The impact on the growth rate of the replacement of
the NADP-specific EcG6PDH by the NAD-specific
LmG6PDHR46E,Q47E was then studied in the WT background. This mutation provoked a fall of 44 % in the
growth rate in comparison with the WT strain.
Surprisingly, this effect was even more negative than that
observed in the inact-G6PDH strain, where the growth rate
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
(a)
G6P
F6P
DHAP
6PG
RL5P
E4P
X5P
S7P
R5P
(b)
G6P
F6P
DHAP
GA3P
G6P
F6P
DHAP
6PG
(d)
RL5P
E4P
X5P
S7P
R5P
GA3P
DHAP
DHAP
S7P
R5P
RL5P
E4P
X5P
S7P
R5P
GA3P
1,3BPG
6PG
RL5P
E4P
X5P
S7P
R5P
GA3P
1.4
Specific activity (U mg–1)
F6P
X5P
6PG
F6P
(f)
G6P
E4P
GA3P
G6P
1,3BPG
(e)
RL5P
1,3BPG
1,3BPG
(c)
6PG
NADP NAD
0.4 mM 0.4 mM
1 mM 1 mM
3 mM 3 mM
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1,3BPG
WT
NAD-G6PDH
Dpgi
Dpgi-NAD-G6PDH
Fig. 2. Metabolic pathways around the node of glucose-6-phosphate in the five strains involved in this study. Blue circles
represent NADP-specific dehydrogenases; red circles represent NAD-specific dehydrogenases. (a) WT, (b) NAD-G6PDH, (c)
inact-G6PDH, (d) Dpgi and (e) Dpgi-NAD-G6PDH. G6P, glucose-6-phosphate; 6PG, 6-phosphogluconate; RL5P, ribulose 5phosphate; F6P, fructose 6-phosphate; E4P, erythrose 4-phosphate; X5P, xylulose 5-phosphate; S7P, sedoheptulose 7phosphate; R5P, ribose 5-phosphate; DHAP, dihydeoxyacetone phosphate; GA3P, glyceraldehyde 3-phosphate; 1,3BPG,
1,3-bisphosphoglyceric acid. (f) Mean specific G6PDH activities, detected in the cellular extracts from the represented strains,
using NAD or NADP as the cofactor at different concentrations. The concentration of glucose-6-phosphate used in these
assays was 1 mM. No activities were detected with NAD or NADP in the inact-G6PDH strain (data not shown).
fell just 17 % (Table 2). As no other change beyond the
intended gene replacement was observed, the differences in
the growth rates could be explained by the generation in
the oxPPP of some NADH instead of only NADPH.
Interestingly, despite the differences in the growth rates, no
changes in the yield of biomass per consumed substrate
were observed between WT and NAD-G6PDH strains
(Table 2). Therefore, flux balance analysis correctly predicted
no changes in the growth yields, but failed to predict the
existence of some flux through the modified oxPPP.
http://mic.sgmjournals.org
These results suggest that the effect on the growth rate of
the generation of NADH by G6PDH depends on the
presence of a branching point for the carbon flux: this
effect was positive when the upper Embden–Meyerhof
pathway was blocked and was negative when both the
upper Embden–Meyerhof pathway and the oxPPP were
active. To obtain further evidence for this statement, we
searched for the presence of phosphofructokinases (which
catalyse the irreversible committed step of the upper
Embden–Meyerhof pathway) in the genomes of the
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2787
K. Olavarria and others
(b)
(a)
0.78
Growth rate (h–1)
Growth rate (h–1)
0.78
0.77
0.76
WT
NAD-G6PDH
WT – THD 40% NADPH.
NAD-G6PDH – THD 40% NADPH.
0.75
0
25
50
75
0.77
0.76
100
0.75
0
Incoming glucose through G6PDH (%)
Dpgi
Dpgi NAD-G6PDH
DpntAB
DpntAB NAD-G6PDH
25
50
75
100
Incoming glucose through G6PDH (%)
Fig. 3. Robustness analysis simulating the impact of a variable flux through an NAD- or NADP-preferring G6PDH on the growth
rate for different genetic backgrounds. The changing flux through the G6PDH reaction is represented on the x-axis and the
predicted optimal growth rate achievable is represented on the y-axis. Overall, the impact of changing the flux through G6PDH
was modest. However, there was a small advantage to using the native NADP-specific over the NAD-specific G6PDH. (a) THD
40 % represents a constraint forcing 40 % of the generated NADPH to be made by the membrane-bound transhydrogenase
(pntAB). (b) Simulations performed using a pgi gene knockout as a background to force flux through the oxPPP.
prokaryotic organisms whose G6PDHs have been characterized using NAD and NADP as cofactors. We considered the
four kinds of phosphofructokinases (EC 2.7.1.11, EC
2.7.1.56, EC 2.7.1.105 and EC 2.7.1.146) described so far
(Table 3). It is noteworthy that, for the 16 cases analysed, the
presence of a dual G6PDH was associated with the absence,
non-confirmed functionality or very weak phosphofructokinase activity. Phosphofructokinase activity has been
reported in the three organisms where the G6PDHs were
.100 times more specific for NADP than for NAD.
Integrating qPCR and phenotypic data with in
silico modelling
One possible explanation for the differences in the growth
rates could come from changes in the expression levels of
different enzymes involved in the NAD(P)(H) balance. We
addressed this issue by performing qPCRs. The results with
the primers designed for hybridization with the cDNAs
derived from the mRNA transcribed from the zwf loci in
WT, inact-G6PDH and NAD-G6PDH strains were consistent with the genetic manipulations performed, and their
expression levels were similar under the conditions studied
here: the normalized values were 0.041±0.070 in WT,
0.035±0.080 in inact-G6PDH and 0.034±0.090 in NADG6PDH. For the other loci explored, the levels of
expression did not change more than 1.7-fold (Table 4).
Therefore, despite the elimination of at least one source of
NADPH, the transcriptional regulatory network did not
provoke important changes in the loci observed. These
results indicate that the mechanisms that regulate the
transcription of the enzymes associated with NAD(P)(H)
Table 2. Physiological parameters during the exponential growth phase for the different strains
Strain
Dpgi
Dpgi-NAD-G6PDH
WT
inact-G6PDH
NAD-G6PDH
ND,
Growth rate (h”1)
Glucose uptake rate
(mmol gDW”1 h”1)
Acetate excretion rate
(mmol gDW”1 h”1)
Yield (gDW mmol”1)
0.145±0.003
0.231±0.003
0.64±0.02
0.53±0.02
0.36±0.02
1.63±0.04
3.04±0.04
8.08±0.20
7.19±0.18
4.53±0.17
ND
0.089
0.076
0.079
0.074
0.079
ND
1.91±0.17
2.93±0.15
ND
Not detected.
2788
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
Table 3. Cofactor specificity of G6PDHs and presence/absence of phosphofructokinase in 16 bacterial species
Species
(kcat/Km)NADP : (kcat/Km)NAD
Observations
Acetobacter5Gluconacetobacter hansenii
Pseudomonas fluorescens
0.2 (Ragunathan & Levy, 1994)
0.34* (Lessmann et al., 1975)
Burkholderia multivorans
1 (Cacciapuoti & Lessie, 1977)
Azotobacter5(Pseudomonas?) vinelandii
2 (Anderson & Anderson, 1995)
Pseudomonas aeruginosa
Streptomyces5Kitasatospora aureofaciens
Leuconostoc mesenteroides
Burkholderia cepacia5Pseudomonas C
2 (Lessie & Neidhardt, 1967)
2 (Neuzil et al., 1988)
9 (Vought et al., 2000)
13 (Ben-Bassat & Goldberg, 1980)
Methylomonas sp.
Neisseria gonorrhoeae
Aquifex aeolicus 70 uC
Aquifex aeolicus 40 uC
Gluconobacter oxydans5Acetobacter
suboxydans
Thiobacillus ferrooxydans
13 (Steinbach et al., 1978)
16 (Cacciapuoti & Morse, 1980)
6 (Iyer et al., 2002)
21 (Iyer et al., 2002)
48 (Rauch et al., 2010)
Bacillus5Geobacillus stearothermophilus
Escherichia coli
Thermotoga maritima
60 (Tabita & Lundgren, 1971)
102 (Okuno et al., 1985)
410 (Olavarrı́a et al., 2012)
1268 (Hansen et al., 2002;
Holm et al., 2010)
No phosphofructokinases (Kanehisa, 2002)
EC 2.7.1.56 predicted, not characterized (The UniProt
Consortium, 2012)
EC 2.7.1.11 predicted, not characterized (The UniProt
Consortium, 2012)
EC 2.7.1.11 predicted, not characterized (The UniProt
Consortium, 2012)
No phosphofructokinases (Kanehisa, 2002)
No phosphofructokinases (Novotna & Hostalek, 1985)
No phosphofructokinases (Kanehisa, 2002)
No phosphofructokinases (Kanehisa, 2002),
6-phosphogluconate dehydrogenase is dual
(Benbassat & Goldberg, 1980)
No phosphofructokinases (Kanehisa, 2002)
No phosphofructokinases (Kanehisa, 2002)
EC 2.7.1.11 predicted, not characterized (The UniProt
Consortium, 2012)
No phosphofructokinases (Krajewski et al., 2010), 6phosphogluconate dehydrogenase is dual (Rauch et al., 2010)
EC 2.7.1.11 is 800-fold less active than phosphofructokinase-2
from E. coli (Wang et al., 2012)
EC 2.7.1.11 and EC 2.7.1.56 (Kanehisa, 2002)
EC 2.7.1.11 and EC 2.7.1.56 (Kanehisa, 2002)
EC 2.7.1.11 and EC 2.7.1.56 (Kanehisa, 2002)
*Data refer to G6PDH committed to the Entner–Doudoroff pathway; there is also an NADP-preferring G6PDH in the same organism.
metabolism did not change their patterns in the short term
after a shift in the cofactor specificity in G6PDH.
We then sought to estimate computationally the possible
impact of transcriptional changes that did occur on the
NADH and NADPH production pathways. The growth
rates, glucose uptake rates, acetate excretion rates and
qPCR data were considered as constraints to calculate the
possible internal flux distributions in the WT, inactG6PDH and NAD-G6PDH strains. The production and
consumption patterns of NADH and NADPH were
compared considering the most relevant sources and sinks
in each strain. For the NAD-G6PDH strain, the simulation
shows an operative zwf flux contributing to the NADH
pool and the possible metabolic response to achieve a new
balance in the presence of NADH production by oxPPP
(Fig. 4). However, using this steady-state analysis we could
obtain no inferences regarding changes in the size of the
NADH or NADPH pools.
Taken together, the physiological results suggest that the
drop in the growth rate observed in the NAD-G6PDH
strain could be the consequence of a limited systemic
capacity to overcome, in the short term, the effects of the
production of NADH by G6PDH. The qPCR data suggest
that the regulatory network did not respond in the short
term to compensate for the shortage in NADPH production
http://mic.sgmjournals.org
through the oxPPP. It would be possible, however, that the
growth rate of the NAD-G6PDH strain could be augmented
after selecting for the cells with the fastest growth in the
context of adaptive laboratory evolution. Furthermore,
given the finding that the response was not near the optimal
Table 4. Fold change in the expression levels from different
loci for the inact-G6PDH and NAD-G6PDH strains relative to
the values in WT
Locus
ID in iJO1366
inact-G6PDH
NADG6PDH
zwf
gnd
icd
nadk
pntA
maeA
maeB
ackA
pgi
udhA
mdh
gapA
b1852
b2029
b1136
b2615
b1603
b1479
b2463
b2296
b4025
b3962
b3236
b1779
0.82
1.47
1.31
1.22
1.63
0.76
1.21
1.62
1.54
0.69
1.18
1.33
0.81
1.35
1.20
1.16
1.31
0.70
1.05
1.32
1.29
0.85
1.23
1.61
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2789
K. Olavarria and others
predicted phenotype, additional algorithms, such as MOMA
(Segrè et al., 2002) or ROOM (Shlomi et al., 2005), could be
pertinent for examining the flux state of the cells
immediately post-perturbation.
However, the reducing power of NADPH is used to drive
antioxidant and biosynthetic reactions. The majority of the
reducing equivalents harnessed in the NADPH pool remain
in the cell, unlike the reducing equivalents in the NADH
pool, which leave the cell. Given these differences in their
physiological roles, the redox status of these pools must
be regulated by different means. The redox status of the
internal NAD : NADH pair depends on the redox potential
of the external electron acceptors (de Graef et al., 1999),
but the NADP : NADPH redox status must be regulated
by controlling the flux through the NADPH-producing
branching points. One way to control the generation of
NADPH is by partitioning the carbon flux at branching
points under some kind of regulatory control where only
one of the emerging pathways bears NADPH-producing
enzymes. The amount of carbon going through each branch
varies according to the particular system. It depends on the
kinetic parameters and relative abundances of the enzymes
It seems that metabolic balances are insufficient to explain
why EcG6PDH enzyme is highly specific for NADP. It is
important to note that among the NADPH generation
pathways of E. coli, the oxPPP is dedicated exclusively to
NADPH production and the flux through this pathway
does not affect NADH metabolism directly. The importance of having NADP-specific dehydrogenases is explained
as follows.
Despite the structural similarities, NAD(H) and NADP(H)
have very different metabolic roles. The reoxidation of
NADH is coupled to energy-conserving processes fuelled
by the difference between the redox potential of the electron
sources and the redox potential of the final acceptors.
WT
NAD-specific
zwf
gdhA (0.66)
sthA (0.18)
cysI/J (0.11)
idnD (0.09)
icd (0.51)
pntAB (0.44)
zwf (0.30)
NADPH
gnd (0.30)
folD (0.08)
gapA (1.68)
pntAB (0.58)
ilvC (0.07)
zwf (0)
proC (0.06)
gnd (0.40)
ghrAB (0.06)
folD (0.08)
cysI/J (0.10)
idnD (0.06)
icd (0.67)
asd (0.08)
pntAB (0.72)
ilvC (0.07)
zwf (0)
proC (0.07)
gnd (0)
ghrAB (0.08)
asd (0.08)
ibvC (0.06)
NADPH
proC (0.06)
ghrAB (0)
folD (0.07)
thrA, metL (0.05)
thrA, MetL (0.05)
ydfG (0)
ydfG (0.06)
ydfG (0)
gapA (1.64)
WT
aceEF, lpd (1.13)
mdh (0.57)
mdh (0.71)
lpd, sucAB (0.42)
lpd, sucAB (0.52)
gapA (1.62)
NAD-specific
zwf
nuoA–M (4.28)
sthA (0.16)
NADH
serA (0.13)
pntAB (0.44)
idnD (0.12)
glxR, garR (0.06)
glxR, garR (0.08)
adhR, mhpF (0)
adhR, mhpF (0.06)
Dzwf
aceEF, lpd (0.76)
mdh (0.68)
nuoA–M (3.77)
sthA (0.18)
serA (0.13)
cysI/J (0.11)
thrA, metL (0.05)
aceEF, lpd (1.10)
idnD (0.09)
NADPH
gdhA (0.69)
sthA (0.13)
idnD (0.12)
icd (0.60)
asd (0.08)
Dzwf
gdhA (0.67)
sthA (0.16)
lpd, sucAB (0.58)
nuoA–M (3.46)
sthA (0.13)
NADH
serA (0.12)
pntAB (0.58)
NADH
idnD (0.06)
gbcR, garR (0)
adhR, mhpF (0.21)
puuC (0)
puuC (0)
puuC (0.05)
zwf (0)
zwf (0.40)
zwf (0)
pntAB (0.72)
Legend
High flux (>20%)
Medium flux (>10%)
Low flux (>0%)
Negligible flux (0%)
Fig. 4. In silico modelling of the production and consumption of NADH and NADPH pools in the different strains generated in
this work by integrating phenotypic and qPCR data. The top panel shows the production and consumption of NADPH, whereas
the bottom panel shows the same for NADH. The amount of flux (expressed as mmol gDW”1 h”1) is scaled to the amount of
glucose coming into the cell for each case. As expected, dehydrogenases from oxPPP, isocitrate dehydrogenase and PntAB are
the major contributors of NADPH. In the presence of an NAD-preferring G6PDH, oxPPP is used at an increased relative rate,
but its contribution to the NADPH pool falls by 33 %. However, the simulation indicates that the contribution of the engineered
G6PDH to the NADH pool must be balanced mainly by a higher consumption, particularly by PntAB transhydrogenase.
2790
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
competing for the same metabolite, the rate of generation of
the shared metabolite and other regulatory mechanisms
(LaPorte et al., 1984). The best-studied case so far is the
branching point between isocitrate dehydrogenase and
isocitrate lyase where the behaviour of the emerging fluxes
upon different changes in the concentration of isocitrate and
the maximum velocity of the isocitrate dehydrogenase were
modelled (LaPorte et al., 1984). It is also well known that
isocitrate dehydrogenase can be transiently inactivated by
phosphorylation, enabling an increase in the flux through
the glyoxylate cycle (Garnak & Reeves, 1979). In the case of
the partitioning of the flux around glucose-6-phosphate, it
was observed that it does change after an increase in the
demand (Rui et al., 2010) or the offer of NADPH (Martı́nez
et al., 2008). To the best of our knowledge, there are no
studies on how much control EcG6PDH could exert upon
the flux through the oxPPP, although increasing its
concentration above the typical level did not increase
significantly the flux through this pathway (Nicolas et al.,
2007; Orthner & Pizer, 1974).
The branching points described could function as valves
because NADPH is usually generated in reactions that are
irreversible under physiological conditions. Interestingly,
all the sources of NADPH in the central pathways of E. coli
are irreversible reactions (under physiological conditions)
and are placed at branching points (with the exception of
the membrane-bound transhydrogenase). In the case of the
oxPPP, there are two steps with the ability to generate
NADPH: the irreversible conversion of glucose-6-phosphate into 6-phospho-gluconolactone (at the branching
point between the Embden–Meyerhof pathway and the
oxPPP) and the decarboxylative oxidation of 6-phosphogluconate to form ribulose 5-phosphate (at the branching
point between the Entner–Doudoroff pathway and the
oxPPP). Therefore, partitioning the carbon flux through
the non-essential oxPPP and the Entner–Doudoroff pathways allows an effective control of the amount of NADPH
produced by the cell without direct interference in the
generation of NADH.
Therefore, the results presented here expand our current
interpretation of the importance of generating NADPH by
EcG6PDH as a mechanism to protect the cell against
oxidative stress to a view where the NADP specificity of the
EcG6PDH could have also emerged from systemic
metabolic constraints.
REFERENCES
Anderson, B. M. & Anderson, C. D. (1995). Purification and
characterization of Azotobacter vinelandii glucose-6-phosphate dehydrogenase: dual coenzyme specificity. Arch Biochem Biophys 321, 94–100.
Ben-Bassat, A. & Goldberg, I. (1980). Purification and properties of
glucose-6-phosphate dehydrogenase (NADP+/NAD+) and 6-phosphogluconate dehydrogenase (NADP+/NAD+) from methanolgrown Pseudomonas C. Biochim Biophys Acta 611, 1–10.
Cacciapuoti, A. F. & Lessie, T. G. (1977). Characterization of the fatty
acid-sensitive glucose 6-phosphate dehydrogenase from Pseudomonas
cepacia. J Bacteriol 132, 555–563.
Cacciapuoti, A. F. & Morse, S. A. (1980). Glucose-6-phosphate
dehydrogenase from Neisseria gonorrhoeae: partial characterization of
the enzyme and inhibition by long-chain fatty acid acyl-coenzyme A
derivatives. Can J Microbiol 26, 863–873.
Canonaco, F., Hess, T. A., Heri, S., Wang, T. T., Szyperski, T. & Sauer,
U. (2001). Metabolic flux response to phosphoglucose isomerase
knock-out in Escherichia coli and impact of overexpression of the
soluble transhydrogenase UdhA. FEMS Microbiol Lett 204, 247–252.
Charusanti, P., Conrad, T. M., Knight, E. M., Venkataraman, K., Fong,
N. L., Xie, B., Gao, Y. & Palsson, B. Ø. (2010). Genetic basis of growth
adaptation of Escherichia coli after deletion of pgi, a major metabolic
gene. PLoS Genet 6, e1001186.
Datsenko, K. A. & Wanner, B. L. (2000). One-step inactivation of
chromosomal genes in Escherichia coli K-12 using PCR products. Proc
Natl Acad Sci U S A 97, 6640–6645.
de Graef, M. R., Alexeeva, S., Snoep, J. L. & Teixeira de Mattos, M. J.
(1999). The steady-state internal redox state (NADH/NAD) reflects
the external redox state and is correlated with catabolic adaptation in
Escherichia coli. J Bacteriol 181, 2351–2357.
Fuhrer, T. & Sauer, U. (2009). Different biochemical mechanisms
ensure network-wide balancing of reducing equivalents in microbial
metabolism. J Bacteriol 191, 2112–2121.
Garnak, M. & Reeves, H. C. (1979). Phosphorylation of isocitrate
dehydrogenase of Escherichia coli. Science 203, 1111–1112.
Hansen, T., Schlichting, B. & Schönheit, P. (2002). Glucose-6-
phosphate dehydrogenase from the hyperthermophilic bacterium
Thermotoga maritima: expression of the g6pd gene and characterization of an extremely thermophilic enzyme. FEMS Microbiol Lett 216,
249–253.
Holm, A. K., Blank, L. M., Oldiges, M., Schmid, A., Solem, C., Jensen,
P. R. & Vemuri, G. N. (2010). Metabolic and transcriptional response
to cofactor perturbations in Escherichia coli. J Biol Chem 285, 17498–
17506.
Holmberg, N., Ryde, U. & Bülow, L. (1999). Redesign of the coenzyme
specificity in L-lactate dehydrogenase from Bacillus stearothermophilus
using site-directed mutagenesis and media engineering. Protein Eng
12, 851–856.
Humphrey, W., Dalke, A. & Schulten, K. (1996). VMD: visual molecular
dynamics. J Mol Graph 14, 33–38, 27–28.
ACKNOWLEDGEMENTS
Iyer, R. B., Wang, J. Q. & Bachas, L. G. (2002). Cloning, expression,
This work was supported by MECESUP-Chile (UCH-0713 and UCH0717), The Novo Nordisk Foundation and FONDECYT-Chile
(1121170). K. O. and R. C. are very grateful to Dr Rosa Deves for
decisive institutional support and Dr Juan Carlos Letelier for
enlightening discussions about metabolism. We are indebted to Dr
Bernhard Ø. Palsson and Dr Andrei Osterman, as a significant part of
the experiments presented here were done in their laboratories. K. O.
thanks Xiaoqing Lee, Pep Charusanti, David Scott and Oleg Kurnasov
for important technical advice.
http://mic.sgmjournals.org
and characterization of the gsdA gene encoding thermophilic glucose6-phosphate dehydrogenase from Aquifex aeolicus. Extremophiles 6,
283–289.
Kanehisa, M. (2002). The KEGG database. Novartis Found Symp 247,
91–101, discussion 101–103, 119–128, 244–252.
Krajewski, V., Simic, P., Mouncey, N. J., Bringer, S., Sahm, H. & Bott,
M. (2010). Metabolic engineering of Gluconobacter oxydans for
improved growth rate and growth yield on glucose by elimination
of gluconate formation. Appl Environ Microbiol 76, 4369–4376.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2791
K. Olavarria and others
LaPorte, D. C., Walsh, K. & Koshland, D. E., Jr (1984). The branch
Orthner, C. L. & Pizer, L. I. (1974). An evaluation of regulation of the
point effect. Ultrasensitivity and subsensitivity to metabolic control.
J Biol Chem 259, 14068–14075.
hexose monophosphate shunt in Escherichia coli. J Biol Chem 249,
3750–3755.
Lee, W. T. & Levy, H. R. (1992). Lysine-21 of Leuconostoc mesenteroides
glucose-6-phosphate dehydrogenase participates in substrate binding
through charge–charge interaction. Protein Sci 1, 329–334.
Perrenoud, A. & Sauer, U. (2005). Impact of global transcriptional
Lessie, T. & Neidhardt, F. C. (1967). Adenosine triphosphate-linked
Phillips, J. C., Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E., Villa,
E., Chipot, C., Skeel, R. D., Kalé, L. & Schulten, K. (2005). Scalable
control of Pseudomonas aeruginosa glucose-6-phosphate dehydrogenase. J Bacteriol 93, 1337–1345.
regulation by ArcA, ArcB, Cra, Crp, Cya, Fnr, and Mlc on glucose
catabolism in Escherichia coli. J Bacteriol 187, 3171–3179.
molecular dynamics with NAMD. J Comput Chem 26, 1781–1802.
Lessmann, D., Schimz, K. L. & Kurz, G. (1975). D-Glucose-6-phosphate
Price, N. D., Reed, J. L. & Palsson, B. O. (2004). Genome-scale models
dehydrogenase (Entner–Doudoroff enzyme) from Pseudomonas fluorescens. Purification, properties and regulation. Eur J Biochem 59, 545–559.
of microbial cells: evaluating the consequences of constraints. Nat Rev
Microbiol 2, 886–897.
Lindorff-Larsen, K., Piana, S., Palmo, K., Maragakis, P., Klepeis, J. L.,
Dror, R. O. & Shaw, D. E. (2010). Improved side-chain torsion potentials
Ragunathan, S. & Levy, H. R. (1994). Purification and characteriza-
for the Amber ff99SB protein force field. Proteins 78, 1950–1958.
Lowry, O. H., Carter, J., Ward, J. B. & Glaser, L. (1971). The effect of
carbon and nitrogen sources on the level of metabolic intermediates
in Escherichia coli. J Biol Chem 246, 6511–6521.
Machado, D. & Herrgård, M. (2014). Systematic evaluation of
tion of the NAD-preferring glucose-6-phosphate dehydrogenase from
Acetobacter hansenii (Acetobacter xylinum). Arch Biochem Biophys 310,
360–366.
Rauch, B., Pahlke, J., Schweiger, P. & Deppenmeier, U. (2010).
Characterization of enzymes involved in the central metabolism of
Gluconobacter oxydans. Appl Microbiol Biotechnol 88, 711–718.
methods for integration of transcriptomic data into constraint-based
models of metabolism. PLOS Comput Biol 10, e1003580.
Rocco, C. J., Dennison, K. L., Klenchin, V. A., Rayment, I. &
Escalante-Semerena, J. C. (2008). Construction and use of new
Martı́nez, I., Zhu, J., Lin, H., Bennett, G. N. & San, K.-Y. (2008).
cloning vectors for the rapid isolation of recombinant proteins from
Escherichia coli. Plasmid 59, 231–237.
Replacing Escherichia coli NAD-dependent glyceraldehyde 3-phosphate dehydrogenase (GAPDH) with a NADP-dependent enzyme
from Clostridium acetobutylicum facilitates NADPH dependent pathways. Metab Eng 10, 352–359.
Rui, B., Shen, T., Zhou, H., Liu, J., Chen, J., Pan, X., Liu, H., Wu, J.,
Zheng, H. & Shi, Y. (2010). A systematic investigation of Escherichia
Marx, A., Eikmanns, B. J., Sahm, H., de Graaf, A. A. & Eggeling, L.
(1999). Response of the central metabolism in Corynebacterium
Ryde, U. (1995). On the role of Glu-68 in alcohol dehydrogenase.
coli central carbon metabolism in response to superoxide stress. BMC
Syst Biol 4, 122.
Protein Sci 4, 1124–1132.
glutamicum to the use of an NADH-dependent glutamate dehydrogenase. Metab Eng 1, 35–48.
Salas, M., Vinuela, E. & Sols, A. (1965). Spontaneous and
Naylor, C. E., Gover, S., Basak, A. K., Cosgrove, M. S., Levy, H. R. &
Adams, M. J. (2001). NADP+ and NAD+ binding to the dual coenzyme
enzymatically catalyzed anomerization of glucose-6-phosphate and
anomeric specificity of related enzymes. J Biol Chem 240, 561–568.
specific enzyme Leuconostoc mesenteroides glucose-6-phosphate dehydrogenase: different interdomain hinge angles are seen in different binary
and ternary complexes. Acta Crystallogr D Biol Crystallogr 57, 635–648.
Sauer, U., Canonaco, F., Heri, S., Perrenoud, A. & Fischer, E. (2004).
Neidhardt, F. C., Ingraham, J. L. & Schaechter, M. (1990). Physiology
of the Bacterial Cell: A Molecular Approach. Sunderland, MA: Sinauer.
Neuzil, J., Novotná, J., Erban, V., B?hal, V. & Hostálek, Z. (1988).
Glucose-6-phosphate dehydrogenase from a tetracycline producing
strain of Streptomyces aureofaciens: some properties and regulatory
aspects of the enzyme. Biochem Int 17, 187–196.
Nicolas, C., Kiefer, P., Letisse, F., Kroemer, J., Massou, S., Soucaille, P.,
Wittmann, C., Lindley, N. D. & Portais, J.-C. (2007). Response of the central
metabolism of Escherichia coli to modified expression of the gene encoding
the dehydrogenase glucose-6-phosphate. FEBS Lett 581, 3771–3776.
Novotna, J. & Hostalek, Z. (1985). Phosphorylation of hexoses in
Streptomyces aureofaciens – evidence that the phosphoenolpyruvatesugar phosphotransferase system is not operative. FEMS Microbiol
Lett 28, 347–350.
Okuno, H., Nagata, K. & Nakajima, H. (1985). Purification and
properties of glucose-6-phosphate dehydrogenase from Bacillus
stearothermophilus. J Appl Biochem 7, 192–201.
Olavarrı́a, K., Valdés, D. & Cabrera, R. (2012). The cofactor preference
of glucose-6-phosphate dehydrogenase from Escherichia coli –
modeling the physiological production of reduced cofactors. FEBS J
279, 2296–2309.
Orth, J. D., Thiele, I. & Palsson, B. Ø. (2010). What is flux balance
analysis? Nat Biotechnol 28, 245–248.
Orth, J. D., Conrad, T. M., Na, J., Lerman, J. A., Nam, H., Feist, A. M. &
Palsson, B. Ø. (2011). A comprehensive genome-scale reconstruction
of Escherichia coli metabolism – 2011. Mol Syst Biol 7, 535.
2792
The soluble and membrane-bound transhydrogenases UdhA and
PntAB have divergent functions in NADPH metabolism of Escherichia
coli. J Biol Chem 279, 6613–6619.
Schellenberger, J., Que, R., Fleming, R. M., Thiele, I., Orth, J. D., Feist,
A. M., Zielinski, D. C., Bordbar, A., Lewis, N. E. & other authors (2011).
Quantitative prediction of cellular metabolism with constraint-based
models: the COBRA Toolbox v2.0. Nat Protoc 6, 1290–1307.
Schuetz, R., Zamboni, N., Zampieri, M., Heinemann, M. & Sauer, U.
(2012). Multidimensional optimality of microbial metabolism.
Science 336, 601–604.
Segrè, D., Vitkup, D. & Church, G. M. (2002). Analysis of optimality in
natural and perturbed metabolic networks. Proc Natl Acad Sci U S A
99, 15112–15117.
Selwyn, M. J. (1965). A simple test for inactivation of an enzyme
during assay. Biochim Biophys Acta 105, 193–195.
Shlomi, T., Berkman, O. & Ruppin, E. (2005). Regulatory on/off
minimization of metabolic flux changes after genetic perturbations.
Proc Natl Acad Sci U S A 102, 7695–7700.
Steinbach, R. A., Sahm, H. & Schütte, H. (1978). Purification and
regulation of glucose-6-phosphate dehydrogenase from obligate
methanol-utilizing bacterium Methylomonas M15. Eur J Biochem 87,
409–415.
Tabita, R. & Lundgren, D. G. (1971). Glucose-6-phosphate dehydro-
genase from the chemolithotroph Thiobacillus ferrooxidans. J Bacteriol
108, 343–352.
The UniProt Consortium (2012). Reorganizing the protein space at
the Universal Protein Resource (UniProt). Nucleic Acids Res 40
(Database issue), D71–D75.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
Microbiology 160
Effect of an NADH-producing G6PDH in E. coli
Varma, A., Boesch, B. W. & Palsson, B. O. (1993). Stoichiometric
interpretation of Escherichia coli glucose catabolism under various
oxygenation rates. Appl Environ Microbiol 59, 2465–2473.
Vought, V., Ciccone, T., Davino, M. H., Fairbairn, L., Lin, Y., Cosgrove,
M. S., Adams, M. J. & Levy, H. R. (2000). Delineation of the roles of
amino acids involved in the catalytic functions of Leuconostoc
mesenteroides glucose-6-phosphate dehydrogenase. Biochemistry 39,
15012–15021.
http://mic.sgmjournals.org
Wang, H., Liu, X., Liu, S., Yu, Y., Lin, J., Lin, J., Pang, X. & Zhao, J.
(2012). Development of a markerless gene replacement system for
Acidithiobacillus ferrooxidans and construction of a pfkB mutant. Appl
Environ Microbiol 78, 1826–1835.
Zhu, G. P., Golding, G. B. & Dean, A. M. (2005). The selective cause of
an ancient adaptation. Science 307, 1279–1282.
Edited by: F. Sargent
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Sat, 17 Jun 2017 21:56:54
2793