Journal of Antimicrobial Chemotherapy (2003) 51, 821–831
DOI: 10.1093/jac/dkg170
Advance Access publication 13 March 2003
Structure–activity relationships for a series of peptidomimetic
antimicrobial prodrugs containing glutamine analogues
Neil J. Marshall1, Ryszard Andruszkiewicz2, Sona Gupta1, Sßawomir Milewski2 and John W. Payne1*
1School
of Biological Sciences, University of Wales Bangor, Gwynedd LL57 2UW, UK; 2Department of
Pharmaceutical Technology and Biochemistry, Technical University of Gdaásk, Gdaásk, 80-952, Poland
Received 3 December 2002; returned 18 January 2003; revised 21 January 2003; accepted 21 January 2003
Synthetic glutamine analogues such as N 3 -(4-methoxyfumaroyl)-L-2,3-diaminopropanoic acid
(FMDP) inhibit purified glucosamine-6-phosphate synthase, an intracellular enzyme that is
essential for microbial cell wall synthesis, but they are inactive against intact organisms
because they cannot enter the cell. However, when the analogues are linked to a peptide they can
be actively transported, and FMDP peptidomimetics show broad-spectrum antimicrobial activity. To characterize this process in more detail, the antibacterial activities of various synthetic
peptidomimetics containing glutamine analogues have been determined against isogenic
strains of Escherichia coli in which one or more of its three peptide transporters Dpp, Opp and
Tpp have been mutated. In addition, their affinities for DppA and OppA, the binding-protein components of the transporters, have been measured. In general, antibacterial activities against the
various transport mutants correlated with binding to DppA and OppA. Xaa-FMDP compounds
have greater activities than FMDP-Xaa analogues. To explore structure–activity relationships
for the peptidomimetics, molecular modelling was used to determine the conformational forms
they adopt in solution. The relative bioactivities of the peptidomimetics correlated with the
percentage of conformers that had backbone torsions matching those previously defined for
the molecular recognition templates of the peptide transporters. However, the large size of the
N-terminal residue in the FMDP-Xaa analogues appears to interfere with transport and thus to
limit antibacterial activity. Overall, the results provide the structural rationale for the identification in silico of analogues with optimal bioactivities, which decreases the need for extensive
chemical syntheses and testing.
Keywords: cell wall biosynthesis, peptide transport, molecular modelling, molecular recognition
Introduction
When designing novel antimicrobial agents, enzymes involved
in the biosynthesis of microbial cell walls are generally good
targets. In this regard, glucosamine-6-phosphate synthase
[L-glutamine: D-fructose-6-phosphate amidotransferase (hexose isomerising)], which is responsible for the synthesis of
D-glucosamine-6-phosphate, is particularly attractive because
it is involved in the first step in the formation of the core
amino-sugar N-acetyl glucosamine that is an essential part of
the unique peptidoglycan and chitin components of the cell
walls of bacteria and fungi, respectively. Although the analogous enzyme is present in man, its inhibition is not a problem
because of the slow turnover of its relevant polysaccharide
products. Thus, compounds that are able to inhibit this
enzyme can potentially be broad-spectrum antimicrobial
agents, having both antibacterial and antifungal activities.
Several well-studied glutamine analogues, e.g. 6-diazo-5-oxoL-norleucine, azaserine and anticapsin, do inhibit the enzyme
but lack specificity and inhibit other enzymes that use glutamine as a substrate.1,2 A rationally designed glutamine analogue, N 3 -(4-methoxyfumaroyl)-L-2,3-diaminopropanoic
acid (FMDP), was shown to be both a strong and selective
inhibitor of the enzyme in vitro, inhibiting the enzyme by
forming a covalent linkage with the Cys residue located in the
glutamine-binding domain.3–5 Unfortunately, FMDP proved
..................................................................................................................................................................................................................................................................
*Corresponding author. Tel: +44-1248-382349; Fax: +44-1248-370731; E-mail: [email protected]
...................................................................................................................................................................................................................................................................
821
© 2003 The British Society for Antimicrobial Chemotherapy
N. J. Marshall et al.
ineffective against whole cells because it could not gain access
to the intracellular location of the enzyme.3,6 Subsequently,
FMDP was converted into an active antimicrobial agent by
incorporating it as part of a peptide, which allowed its active
accumulation by peptide transporters in the form of a
peptidomimetic prodrug susceptible to intracellular peptidase
action.3,6,7
FMDP-peptides have been shown to inhibit various
bacterial species, including Escherichia coli (ATCC 9637),
Shigella sonnei 433, Staphylococcus aureus 209P, Bacillus
pumilus 1697, Bacillus subtilis3,6,8,9 and also fungi, e.g. Candida albicans.6,9 However, the inhibition assays allowed only
a rough ranking of relative antibacterial activities, e.g. for
E. coli MICs of 100 or >200 mg/L,3 or the inhibitory effects of
a single amount of each inhibitor using disc diffusion assays.8
The rate-limiting step for antimicrobial activity is generally
uptake, although for B. subtilis uptake rates were generally
faster than the intracellular hydrolysis rates that liberated free
FMDP.8 Furthermore, certain of these compounds have
demonstrated chemotherapeutic activity in a mouse model of
generalized candidiasis,10,11 showing them to be potentially
valuable, broad-spectrum antimicrobial compounds.
Molecular modelling of small peptides (two to five
residues) derived from protein hydrolysis has allowed identification of the conformational forms they adopt in solution,
and revealed how various peptide transporters have evolved
complementary specificities to recognize different conformational types as ligands.12,13 From evaluation of the structural
and electronic features needed for this ligand recognition,
individual molecular recognition templates (MRTs) for the
microbial peptide transporters have been defined.14–17 The
following features are involved in defining MRTs: (i) charged
N-terminal α-amino and C-terminal α-carboxylate groups,
permitting charge and hydrogen-bond donor and acceptor
properties; (ii) backbone torsion angles psi, phi and omega;
(iii) chirality at α-carbons; (iv) N–C distance between terminal amino and carboxylate groups; (v) chi space torsions
and size of side chains; (vi) H-bond donor and acceptor properties of peptide bond atoms; and (vii) charge fields around
N-terminal α-amino group and C-terminal α-carboxylate. To
define an MRT, each of these seven features must be of a
certain type, e.g. positively-charged N-terminal α-amino
group or L-chirality, or value, e.g. psi torsion angle of 150°.
Knowing these MRTs, compounds can be modelled to assess
their match to these parameters and thus their potential as
putative transport ligands.
In the present study, structure–activity relationships have
been explored for FMDP peptides using E. coli. The substrate
specificities of the generic peptide transporters Opp, Dpp and
Tpp have well-characterized features that allow them to be
distinguished from each other, but they also share certain substrate recognition features, so that some peptides can be taken
up by more than one transporter.14,17 Dpp and Opp are typical
ABC transporters energized by ATP; each comprises four
membrane proteins plus a periplasmic peptide binding protein DppA or OppA, respectively.17 In contrast, Tpp lacks a
binding protein and comprises a single membrane protein that
is energized by a proton-motive force.17 To be transported by
Dpp or Opp, ligands must be recognized and bound by their
respective binding proteins. A strong correlation exists
between the affinity of ligands for these binding proteins and
uptake rates through the respective transporters.12,14,16,18 Thus,
compounds can be evaluated as potential transport ligands by
measuring their abilities to compete for binding of radioactively labelled ligands to DppA and OppA, which is not
only more convenient than directly assaying transport but in
requiring less sample is particularly convenient when little
material is available, as with many of the peptidomimetics
here. For antibacterial compounds, such binding studies
can complement inhibition assays. Here, structure–activity
results have been obtained using three approaches: first, antibacterial activities against isogenic strains of E. coli that
possess different complements of the peptide transporters;
secondly, peptide binding to purified peptide binding proteins; and thirdly, molecular modelling to derive structural
information on the conformations adopted by the peptidomimetics in solution. The results should aid the design of antimicrobial peptidomimetic prodrugs of glutamine analogues.
Materials and methods
Peptide compounds
Peptidomimetics containing FMDP, N 3 -(fumaramoyl)-L2,3-diaminopropanoic acid (FCDP) or N4-(4-methoxyfumaramoyl)-L-2,4-diaminobutanoic acid (FMDB) were synthesized as described previously.3,8 The iodinated peptide
ligands, Gly[125I]Tyr and [125I2]TyrGlyGly were prepared as
described previously.18 The synthetic inhibitor, alafosfalin
[Ala-L-1-aminoethylphosphonic acid; AlaAla(P)] was a gift
from Roche Products, Welwyn Garden City, UK.14,18
Determination of the antibacterial activities of
peptidomimetics
Peptidomimetics were tested using the following strains, the
isolation and characterization of which were described previously:18 a parental strain, E. coli K-12 Morse 2034 (trp, leu)
(CGSC 5071), which is a wild-type with respect to Opp, Dpp
and Tpp; and three isogenic mutants: (i) strain PA0183
(∆opp), derived from strain Morse 2034, in which the deletion
extends from tonB to tdk; (ii) strain PA0333 (∆opp,dpp)
derived from PA0183; and (iii) strain PA0410 (∆opp,tpp) also
derived from PA0183. Inhibition was determined using the
following disc diffusion assay, adapted from recommended
procedures.19 Bacteria were grown in minimal ‘A’ medium18
822
Structure–activity relationship of glutamine analogues
supplemented with 0.5 mM L-Leu, 0.2 mM L-Trp, 0.05 mM
FeCl3 and 0.4 % (w/v) D-glucose at 37°C with rotary shaking.
Soft agar overlays (0.6 % w/v agar), similarly supplemented
with Leu, Trp and glucose, were inoculated with exponentialphase bacteria to give ∼1 × 107 cells/mL, to allow semi-confluent growth, and used to overlay a plate (1.5 % w/v agar) of
the same composition. Typically, six filter paper discs were
placed on the surface, 10 µL amounts of different concentrations of the peptidomimetics were added, giving a range of
30–300 nmol, and the plates were incubated at 37°C. After 17
h incubation, the appearance of any inhibition zones was
noted and their diameters measured. Antibacterial activity
was expressed as the amount of inhibitor (nmol) required to
produce an inhibition zone of 25 mm and was determined
from semi-logarithmic plots of log10 concentrations of inhibitor versus inhibition zone diameters. These isogenic strains
all have the same rates of growth in the above liquid medium
and on these agar plates, so zone sizes are not susceptible to
different growth rates18 (data not shown).
Assays for peptide binding to DppA and OppA
DppA and OppA were isolated from strain Morse 2034 by an
osmotic shock procedure, purified to homogeneity by sequential use of ion-exchange chromatography, freed from any
endogenous ligands using reverse-phase chromatography,
lyophilized and stored at –20°C, as described.14,18 Competitive filter-binding assays were carried out using Gly[125I]Tyr
as the radioactive ligand for DppA, and [125I2]TyrGlyGly for
OppA, as described previously.18 All competitor peptidomimetics were tested at a molar ratio of 10:1 relative to the
ligands, and some were tested at other molar concentrations.
As control competitors, AlaAla was used with DppA and
AlaAlaAla with OppA. Controls were performed in which
either the binding protein or a competing peptide was omitted.
with relevant side-chain torsions, chi (χ) and the distance
(N–C) between the nitrogen of the N-terminal α-amino group
and α-carbon of the C-terminal α-carboxylate group. For
tripeptide conformers, analogous measurements were made,
together with those for (ψi) Tor6, (ω2) Tor7 and (φi+1) Tor8
(Figure 1). For each dipeptide compound, ψ and φ torsional
space for the peptide unit was divided up into 36 10° sectors
and the percentage contributions of conformers having
particular ψ–φ combinations were aggregated. Molecular
modelling of peptides has previously indicated that conformers adopt a limited set of torsional combinations, which
have been classified by dividing ψ and φ conformational
space into 12 30° sectors designated A1 to A12 (0° to ±180°)
and B1 to B12 (0° to ±180°), respectively.12,14–16 In this way,
particular conformational forms can be defined by reference
to their location within the conformational grid, e.g. A7B9.
Analogous procedures were adopted for tripeptides.16 A
complete profile of the conformers of any compound can be
displayed using 3D pseudo-Ramachandran plots (3DPR) that
relate cumulative percentage of conformer to backbone ψ and
φ torsions.12,14–16,20
The maximum lengths and volumes of the side chains of
FCDP, FMDP, FMDB and of natural amino acids were computed by placing a volume mesh around the residues modelled
in fully extended conformations. Lengths were measured
between the Cα atom and the outermost heavy atom of a
chain. Volumes were determined for the side chain only, i.e.
excluding the α-amino and α-carboxylate, and were related to
the volume of a water molecule (17.7 Å3).
Molecular modelling of peptides and peptidomimetics
Conformational analysis of zwitterionic peptides and peptidomimetics was carried out using SYBYL software (Tripos Inc.,
St Louis, MO, USA), in a similar manner to that described in
detail previously.12–16,20,21 Briefly, starting structures were
assigned appropriate atom types and charges, and were energyminimized before being subjected to random searches (2–5k
search cycles) with an absolute energy cut-off of 40 kcal/mol,
using a distance-based dielectric constant of 80 to simulate a
water environment. Unique conformers were distinguished
by setting a root mean square threshold of 0.2 Å, together with
chirality checking. The computed collection of conformers
for each compound was first ordered according to energy,
before the percentage contribution of each conformer was
determined using a Boltzmann distribution. For each dipeptide conformer, backbone torsion angles, psi (ψ) (Tor2),
omega (ω) (Tor3) and phi (φ) (Tor4) were measured, together
Figure 1. Schematic of tripeptide mimetic LysNva-FMDP showing the
amino acid residues i–1, i and i+1, with their associated charges. Backbone torsion angles: Tor2 is ψi–1, Tor3 is ω, Tor4 is φi, Tor6 is ψi, Tor7
is ω2 and Tor8 is φi+1. Side chain torsion angle χ1 is shown for Lys. N–C
distance is measured from the N-terminal α-amino nitrogen to the
C-terminal α-carboxylate carbon.
823
N. J. Marshall et al.
Results
Antibacterial activities of peptidomimetics
The antibacterial activities of the peptidomimetics were
tested against the parent strain M2034 and peptide-transport
mutants of E. coli (Table 1). Because only limited amounts of
the synthetic peptidomimetics were available, all compounds
were initially tested against the parent strain and those that did
not give a measurable level of inhibition, e.g. Gly-FMDP
(Table 1), were not tested with the mutants. In addition, tests
with the mutants were performed on the basis of their substrate specificities, e.g. Opp favours peptides with three to
five residues and transports dipeptides poorly, whereas Dpp
and Tpp transport di- and tripeptides only but with different
specificities.17,18,22 Alafosfalin, a well-characterized, synthetic,
antimicrobial peptide prodrug (smugglin),7,22 was used
(0.5 nmol) as a control to check reproducibility of assay procedures: for example, an inhibition zone of 18.6 ± 0.9 mm was
found for all assays with strain M2034.
Table 1. Antibacterial activities of glutamine analogue
peptidomimetics against strains of E. coli having different
peptide transporters
Peptidomimetics
M2034a
PA0183a
PA0410a
PA0333a
Leu-FMDP
Lys-FMDP
Met-FMDP
Nva-FMDP
Phe-FMDP
Tyr-FMDP
Val-FMDP
Nva-FMDB
FMDP-Ala
Gly-FMDP
AcNva-FMDP
FMDP-Leu
FMDP-Met
FMDP-FMDP
FCDP-Ala
FCDP-Met
Nva-FMDP-Nva
SarNva-FMDP
LysNva-FMDP
80
180
90
60
80
200
60
150
700
>3000b
>5000b
>1000b
>3500b
>5000b
>5000b
>5000b
230
290
150
200
250
ND
ND
250
550
80
190
NI
450
240
750
60
600
900
100
NI
NI
180
NI
500
500
1000
>1000
550
NI
NI
NIc
NIc
NIc
Activities are expressed as the amounts in nmol producing inhibition zones of
25 mm, obtained by extrapolation from plots of inhibition zone diameters
versus amounts of peptidomimetics.
aBacterial strains have the following phenotypes for peptide transporters:
wild-type strain M2034 (Opp+,Dpp+,Tpp+); PA0183 (Opp–,Dpp+,Tpp+);
PA0410 (Opp–,Dpp+,Tpp–); PA0333 (Opp–,Dpp–,Tpp+).
bBecause these activities were so low only the wild-type strain was tested.
cBecause no inhibition was detected with this Opp– strain the other strains
containing the same opp deletion were not tested.
Nva, norvalyl (2-amino pentanoic acid); Sar, sarcosyl (N-methyl glycyl); ND,
not determined because of limiting material; NI, no inhibition was detected at
any amount up to the maximum 300 nmol tested.
From Table 1, possible trends can be inferred on the
inhibitory activities of the various analogues towards strain
M2034. First, Xaa-FMDP dipeptides are markedly more
active than FMDP-Xaa dipeptides. This can be seen specifically when comparing Leu-FMDP and Met-FMDP with
FMDP-Leu and FMDP-Met. Secondly, Xaa-FMDP dipeptides appear in general to have greater activity than analogous
oligopeptides. Thirdly, the antibacterial activities of analogues
of FMDP are higher than those of FMDB (compare NvaFMDP with Nva-FMDB) and FCDP analogues (compare
FMDP-Ala with FCDP-Ala). These observations agree with
the relative inhibitory activities of the free glutamine analogues against the purified target enzyme: their IC50 values
being FMDP 15 µM, FMDB 56 µM and FCDP 100 µM.1
However, from these inhibition data with M2034, it is unclear
whether variations in activities arise primarily from differences in uptake or in intracellular cleavage, making results
with transport mutants helpful.
Active Xaa-FMDP analogues can generally be transported
by both Dpp and Tpp, i.e. strains PA0410 and PA0333 were
inhibited, with activity against the former usually being
greater (Table 1). Their activities against strains M2034 and
PA0183 indicated that uptake by Opp also contributed to their
overall antibacterial activity (Table 1). For the three oligopeptides, inhibition of wild-type M2034 and no antibacterial
activity against PA0183 (Opp–,Dpp+,Tpp+) indicate that their
uptake occurs exclusively by Opp (Table 1). Thus, although
both Dpp and Tpp can transport ‘folded’ conformers of many
natural tripeptides, these particular tripeptide mimetics are
effectively recognized only by Opp, which recognizes ‘elongated’ conformers.14,17,22
Binding of peptidomimetics to DppA and OppA
Several trends can be observed from the relative abilities of
the dipeptide analogues to compete for binding with
Gly[125I]Tyr to DppA (Table 2). Xaa-FMDP dipeptides are
generally better competitors than FMDP-Xaa dipeptides; this
was exemplified specifically when comparing Leu-FMDP
and Met-FMDP with FMDP-Leu and FMDP-Met. Most XaaFMDP dipeptides showed broadly similar levels of competitive ability, although no activity was detectable for GlyFMDP and FMDP-FMDP. These results are all in broad
agreement with the relative inhibitory activities of these
compounds (Table 1), indicating that transport rates appear to
be the main determinant of antibacterial activities. AcNvaFMDP also showed no competitive activity (Table 2), in
accordance with the requirement for a protonated N-terminal
α-amino group for ligand binding to DppA and transport by
Dpp.17,18,22 Consequently, lack of binding and presumed
transport is sufficient explanation for its failure to inhibit
(Table 1). In addition to the results for a ligand:competitor
ratio of 1:10 (Table 2), Nva-FMDP and Nva-FMDB were also
assayed at 1:1 and 1:5 ratios and found to inhibit 73% and
824
Structure–activity relationship of glutamine analogues
Table 2. Competitive binding of peptidomimetics to DppA
and OppA
Percentage
inhibition
of binding
Peptidomimetics to DppA
Percentage
inhibition
of binding
Peptidomimetics to OppA
Gly-FMDP
Leu-FMDP
Met-FMDP
Nva-FMDP
Phe-FMDP
Tyr-FMDP
Val-FMDP
AcNva-FMDP
Nva-FMDB
FMDP-Ala
FMDP-Leu
FMDP-Met
FMDP-FMDP
FCDP-Ala
FCDP-Met
SarNva-FMDP
LysNva-FMDP
Nva-FMDP-Nva
Lys-FMDP
0
80
88
99
94
73
75
0
93
14
36
5
0
49
47
83
89
91
24
The percentage inhibition of binding to DppA was determined using
Gly[125I]Tyr as ligand, and for OppA [125I2]TyrGlyGly was used as ligand. The
molar ratio of peptidomimetic to ligand was 10:1 in all cases.
93%, and 89% and 100%, respectively. Similarly, FCDP-Ala
and FCDP-Met are better competitors than FMDP-Ala and
FMDP-Met (Table 2), compatible with them being well transported and implying that the very poor antibacterial activity of
the FCDP-containing dipeptides (Table 1) relates to the lower
enzyme inhibition of the FCDP warhead compared with
FMDP.1
As with Dpp, substrates for Opp can be ranked by their
abilities to bind to OppA,23 and these can be measured in an
analogous competitive assay using [125I2]TyrGlyGly as
ligand. The three tripeptide analogues showed comparable
competitive binding to OppA (Table 2), which accords with
their similar antibacterial activities and supports their transport by Opp (Table 1). Lys-FMDP showed a lower level of
binding, as is normal for dipeptides, but in accordance with
uptake by Opp contributing to the inhibition seen with these
compounds in M2034 (Table 1).
Molecular modelling of peptides and peptidomimetics
Molecular modelling of various classes of natural and synthetic peptides has allowed identification of the bioactive
conformational forms recognized by different peptide transporters and peptidases, and provided an understanding of
structure–activity relationships for substrates of these proteins.12–14,16,20 Backbone torsions are a very important feature
of MRTs for peptide transporters.12,14–17 For the ω peptide
bond only the trans form is recognizable.24,25 A variety of
values can be adopted by ψ and φ torsions. Dipeptides mostly
adopt Tor2 of A4 (–50° to –85°), A7 (+140° to –175°) and
A10 (+50° to +85°), and Tor4 of B2 (+40° to +85°), B9 (–50°
to –95°) and B12 (–130° to +175°) (Figure 1), and these
torsions help to determine the ligand specificities of peptide
transporters.12,14–17 Thus, in the present case, conformational
analysis of the peptidomimetics can provide an estimate of the
extent to which the conformer repertoire of each matches the
MRTs for substrate(s) of each peptide transporter. This evaluation of the peptidomimetics as putative ligands of Dpp, Tpp
and Opp can be related to the ligand binding data, and, in turn,
to the antibacterial results, so as to explore the basis for
structure–inhibition relationships. Random search conformational analysis for each peptidomimetic and related Xaa-Gln
dipeptides showed they occurred as several hundred distinct
conformers (results not shown), comparable to the finding for
most other natural peptides.15,16 The various conformers for
any one compound, e.g. 768 for Val-FMDP, are distinguishable either by having different combinations of backbone torsions (ψ, ω and φ), or having comparable backbone torsions
but different orientations of side chains (χ torsions). Using a
random search procedure, the minimum energy conformation
identified may or may not be the actual global minimum conformer and it may or may not be a bioactive form. However,
this is not of great importance here, for what is critically
important when considering molecular recognition of flexible
molecules is to consider the whole population of conformational forms and to be able to estimate with good precision the
percentage of conformers that match MRTs.13,21 Thus, there is
a direct correlation between the proportion of conformers that
match a particular MRT and the bioactivity of the compound,
e.g. affinity to the ligand, transport rate, etc.12–14,20,24
For these compounds, the minimum energy form typically
accounts for ∼8–30% of the total conformer pool, and there
are broad similarities between the energies of Xaa-Gln
peptides and their respective peptidomimetics. Because all
the peptidomimetics comprise L-amino acids with charged
N- and C-terminal groups, when the minimum energy conformers have backbone torsions that match those for peptide
transporter MRTs they can be considered as putative ligands.
For each peptidomimetic, data for the percentage of its conformers having backbone torsions and lengths matching the
MRTs for Dpp, Tpp and Opp are shown in Table 3. Threedimensional pseudo-Ramachandran plots showing the complete conformer distribution for representative inhibitors are
shown in Figure 2. From these results, several general trends
can be seen regarding structure–activity relationships for the
inhibitors. For the Xaa-FMDP compounds, there is broad
similarity in their conformer profiles with those of the related
Xaa-Gln dipeptides, and they have comparable proportions of
conformers matching the different MRTs, e.g. Tyr-FMDP
(65%, 18%) and Tyr-Gln (64%, 17%); Val-FMDP (55%,
825
N. J. Marshall et al.
Table 3. Percentage of conformers for glutamine dipeptides
and peptidomimetics with backbone torsions matching those
for the MRTs of Dpp, Tpp and Opp
Compounds
Dpp
Tpp
Opp
Gly-FMDP
Val-FMDP
Leu-FMDP
Met-FMDP
Phe-FMDP
Tyr-FMDP
Lys-FMDP
Nva-FMDP
AcNva-FMDP
Nva-FMDB
FMDP-Ala
FMDP-Leu
FMDP-Met
FMDP-FMDP
FCDP-Ala
FCDP-Met
Nva-FMDP-Nva
SarNva-FMDP
LysNva-FMDP
AlaGln
PheGln
LysGln
LeuGln
SerGln
ThrGln
ValGln
TyrGln
26a
55
15
41
57
65
17
42
11
50
71
37
50
27
67
49
–
–
–
39
70
44
21
34
45
54
64
53a
19
55
14
13
18
59
40
46
16
11
16
11
9
14
25
–
–
–
30
11
28
50
38
31
23
17
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
–
70
60a
89
–
–
–
–
–
–
–
–
–, not determined.
aPercentages for peptides containing Gly or Sar residues include conformers
with Tor2 values outside the torsion angles adopted by other amino acids;
this approach is adopted because of the extended backbone torsional space
accessible to these residues although the conformers are not equally good
ligands.13,15,16
19%) and Val-Gln (54%, 23%); Leu-FMDP (15%, 55%) and
Leu-Gln (21%, 50%) for Dpp and Tpp, respectively. These
percentage values for the Xaa-FMDP peptidomimetics are
generally in broad agreement with their relative antibacterial
activities (Table 1). The antibacterial activity of these
compounds appears mainly to arise from uptake by Dpp, in
accordance with them having a higher proportion of conformers that match the MRT for Dpp and, also, as a result
of the greater inherent transport capacity of Dpp.17,22 For
AcNva-FMDP, its inactivity can be attributed to substitution
of its N-terminal α-amino group, which must possess a positive charge to achieve ligand binding to Dpp, Tpp and Opp.
For both DppA and OppA, this involves interaction of the
positive charge on the α-amino group with the side-chain
carboxylate of an aspartic acid residue.12,17,22 The antibacterial activity of Gly-FMDP is extremely poor, which is in
accordance with the general finding that any dipeptide
containing Gly shows low transport activity by Dpp, Tpp and
Opp.17 This arises from several causes. First, a high proportion of their conformers have cis ω bonds, which are not
recognized.14,24,25 Secondly, the small side chain allows
greater flexibility of their ψ and φ torsions so they can adopt a
wider array of conformational forms, very few of which
match well the backbone torsions favoured for recognition by
the transporters. The need to define weightings for the relative
contributions of such conformers with torsions outside the
optimum range (Table 3) still needs to be adequately
addressed.15,24 Significantly, the most effective inhibitor of
mutant PA0333 (Tpp+) is Leu-FMDP (Table 1) and this
accords well with the finding that (but for Lys-FMDP) it has
the highest proportion of conformers (55%) that match precisely the backbone torsions of the MRT for Tpp (Table 3,
Figure 2). Thus, Lys-FMDP alone appears anomalous in the
relationship between its antibacterial activity and conformer
distribution based on backbone torsions, and an explanation
for this is considered below.
The FMDP-Xaa analogues have conformer profiles
broadly similar to those of Xaa-FMDP and they also possess
high proportions of conformers matching the backbone torsions of the MRTs for Dpp and Tpp (Figure 2, Table 3), and yet
they have significantly lower antibacterial activities (Table 1)
and comparably poorer affinities for DppA (Table 2).
Therefore, a further structural/electronic feature(s) must be
responsible for this difference in their bioactivities. To
explore the possibility that this might arise from the size of
FMDP, the dimensions of the side chains for the analogues
were computed and compared with those for examples of the
largest side chains of protein amino acids. Results for lengths,
molecular volumes and number of water equivalents were,
respectively: FMDB (12.3 Å; 145 Å3; 9); FMDP (9.8 Å; 130
Å3; 8); FCDP (8.6 Å; 117 Å3; 7); Gln (5.1 Å; 86 Å3; 5); Lys
(6.4 Å; 107 Å3; 6) and Trp (5.4 Å; 128 Å3; 8). Thus, the
glutamine analogues can adopt more extended conformations
than any protein amino acid and some have larger volumes.
Results with Lys-FMDP provide a further test of the present approach of using conformational analysis to probe
structure–activity relationships. Lys-FMDP inhibited the
parent strain and mutants PA0183 (Dpp+,Tpp+) and PA0410
(Dpp+) but showed no activity against mutant PA0333 (Tpp+)
(Table 1), indicating that its uptake is predominantly by Dpp,
a little by Opp, but not at all by Tpp. These results appear at
odds with the finding that the majority of its conformers have
backbone torsions that match the MRT for Tpp (Table 3,
Figure 2). However, particular properties of a side chain can
also markedly influence the molecular recognition of peptides, as illustrated above regarding size, and described earlier
for certain side chains that are charged and have particular
χ torsions.12–14 Examination of the molecular structures of
A4B9, A4B12 conformational forms of Lys-FMDP, which
826
Structure–activity relationship of glutamine analogues
Figure 2. Three-dimensional pseudo-Ramachandran plots for the dipeptides (a) LysGln and (c) TyrGln, and the dipeptide mimetics (b) Lys-FMDP
and (d) Tyr-FMDP. The aggregated percentages of conformers with particular combinations of backbone torsions ψ (Tor2) and φ (Tor4) are plotted
against those torsions. The axes are labelled with the torsional ranges that are best recognized by Dpp: A7 (+140° to –175°) with B9 (–50° to –95°)
and B12 (–130° to +175°); and by Tpp: A4 (–50° to –85°) and A10 (+50° to +85°) with B9 (–50° to –95°) and B12 (–130° to +175°). Conformers with
B2 torsions (+40° to +85°) are not recognized by either transporter.
have the lowest energies and are most abundant, indicates that
both side chains are contiguous and fully extended in which
trans χ torsions predominate, allowing stabilizing interactions
that are unique to these conformational forms (Figure 3).
Thus, Lys-FMDP is quite atypical, and in its A4(B9,B12)
conformers the side chains appear as a ‘fused unit’. In contrast, natural dipeptides having charged or polar N-terminal
residues, e.g. LysGln has mainly A7 conformers in which the
side chains are separated.13,15
The results showing inhibition of the parent strain but
not of mutant PA0183 (Opp–,Dpp+,Tpp+) by the tripeptide
analogues indicate that Opp is exclusively responsible for
their transport (Table 1). In support of this, each showed good
binding to OppA (Table 2). The steric problems of the long
side chain of FMDP described above would be exacerbated
in ‘folded’ conformations of the tripeptides that are putative
ligands of Dpp and Tpp, precluding them from being recognized by these transporters.12,17
Discussion
All microorganisms have peptide transporters to absorb the
products of protein breakdown. Several transporters usually
coexist; these have overlapping and complementary specificities that have evolved to optimize uptake of the nutritionally
important components of the peptide pool. However, competition between microorganisms has also led to them producing
a wide variety of natural antimicrobial peptidomimetics
designed to exploit peptide transporters for their delivery.
These peptidomimetics usually comprise an impermeable
compound, which is inhibitory to an intracellular target,
linked into a peptide so as to smuggle the antimicrobial agent
into the competing microorganisms in the form of a prodrug.7,17,22,23,26 Amongst the many examples of such natural
antibacterial and antifungal agents, e.g. bacilysin, valclavam,
lindenbein, bialaphos, polyoxins and nikkomycins, some
have been studied for their therapeutic potential.7,23,26 This
827
N. J. Marshall et al.
Figure 3. Ball-and-stick representations of specific low energy conformers derived from random searches: (a) Lys-FMDP (A4B12); (b) TyrGln
(A7B12); (c) Leu-FMDP (A4B9); and (d) FMDP-FMDP (A7B2). The structures were created in CPK colours, but atoms retain recognizable
shading when displayed in black and white. To aid identification, for Lys-FMDP the structure is orientated with the protonated N-terminal α-amino
group of Lys at the bottom left, the side chain protonated ε-amino group of Lys at the top left, the C-terminal α-carboxylate of FMDP at the bottom
right, and with the carbonyl oxygen of the peptide bond directed behind and the nitrogen of the peptide bond directed forwards. The peptide bond is
orientated analogously for the other structures.
approach has also been explored for targeting a variety of
synthetic antimicrobial agents.7,26,27 This peptide prodrug
concept has wider potential benefits, since peptide transporters in the gut and kidney with similar specificities can
endow such therapeutic agents with oral activity. Recent
studies have sought to define the structural basis for ligand
recognition by peptide transporters, so as to allow a
rational approach to the design and synthesis of peptidomimetics with either antimicrobial or other therapeutic activities.12–14,17,24,28,29
Examples of the glutamine analogue peptidomimetics
studied here have been shown previously to have broadspectrum antimicrobial activities and chemotherapeutic
activity in a mouse model of candidiasis. In addition, extensive data have been obtained previously on their transport and
cleavage by microorganisms, but these have not provided
an understanding of the structural basis for their individual
activities.8–11 In the present study, the relative antibacterial
activities of a series of these peptidomimetics against isogenic
strains of E. coli differing only in their complement of peptide
transporters, has shown the extent to which the different transporters recognize and absorb the various compounds. Each
strain has an identical complement of peptidases, so intracellular cleavage of peptidomimetics is comparable in each.
Given the limited availability of compounds, the disc diffusion assay was chosen as a suitable test to obtain quantitative
antibacterial data that was dependent upon the uptake of the
various compounds through distinguishable peptide transporters. In this assay, a query can always be raised regarding
possible variations in diffusion rates of compounds and
whether hypothetically this may influence zone sizes, but
only in very rare cases would it be possible to measure diffu-
828
Structure–activity relationship of glutamine analogues
sion with any precision in these assays so as to check this
speculation. In general, diffusion may be influenced by mass,
shape, pKa, hydrophobicity, etc., but for compounds as
closely related to each other as these it seems improbable that
any such variations could have any significant effect on diffusion rates. Certainly, no correlation exists between the sizes of
the compounds and their zone sizes. Thus, as is usual with this
assay, it appears sensible to accept the inhibition zones as
directly reflecting antimicrobial activities. This approach is
supported by the correlations between the antibacterial assays
and the results obtained from measurements of the relative
affinities of the compounds for the peptide-binding proteins
DppA and OppA, which are the components responsible for
ligand specificities in the Dpp and Opp transporters. In
general, there was good agreement between ligand binding
and antibacterial activity; and this same conclusion has
been reached previously for other antimicrobial peptidomimetics.17,18
Explanation of the structural basis for these activities was
provided by comparison with results of molecular modelling
of the various compounds. These peptidomimetics are the
first example of a series of antimicrobial compounds for
which such an analysis has been attempted. They are flexible
molecules that exist as a variety of conformers in solution, and
to understand how this determines their molecular recognition, transport, antibacterial activities, etc. it is necessary to
analyse the complete repertoire of conformers for each
compound. Previous analysis of natural peptide substrates has
identified the important structural and electronic features, e.g.
charged termini, particular backbone torsions, L-chirality,
etc., which must be possessed by a compound for it to be wellrecognized by peptide transporters in microorganisms and
other species. Thus, Dpp recognizes dipeptide conformers
with torsions of A7 paired with B9 and B12, whereas Tpp recognizes a different set of dipeptide conformers with torsions
of A4 and A10 paired with B9 and B12; Dpp and Tpp also
recognize ‘folded’ tripeptide conformers with the above torsional values and appropriate N–C distances. Opp recognizes
‘elongated’ conformers of tri- and higher oligopeptides (and
to a lesser extent dipeptides) with A7 combined with B9 and
B12 and N–C distances greater than ∼5.5 Å.12,14–16 These conclusions have been corroborated by studies of bound peptide
ligands in crystal structures of DppA and OppA.16,30–33 These
structural parameters have been incorporated into definitions
of MRTs for each microbial transporter, which also appear to
relate to mammalian and plant transporters.12,13,16,24 Thus, in
principle, comparison of the extent to which the conformers
of the peptidomimetics match the MRTs allows estimation of
their transport rates and antibacterial activities.
For most of the Xaa-FMDP compounds and the tripeptidomimetics, structure–activity relationships were clear, with
preferences for Dpp, Tpp or Opp, and thus antibacterial
activities against the different strains, being related to the pro-
portions of conformers that adopted the particular backbone
torsions recognized by each transporter. However, a particular strength of this study is how it has provided insight into the
influence of side chain properties, e.g. overall size and χ torsions, on molecular recognition. Thus, the lower bioactivities
of FMDP-Xaa compounds compared with the corresponding
Xaa-FMDP analogues, which are not explicable in terms of
their backbone torsions, can be related to the dimensions of
the side chain of FMDP. These values can be related to results
of X-ray crystal studies on DppA and OppA, which show that
they have water-filled pockets to accommodate the variety of
side chains found in natural peptides. These have evolved so
that different numbers of water molecules are displaced
depending on the size of the side chain.30–33 DppA possesses
two pockets: thus, for dipeptide ligands, the N- and C-residue
side chains are readily accommodated.31 For DppA (and Tpp)
to bind tripeptides, a tripeptide conformer must be ‘folded’ so
that its ψi–1 and φi torsions and the locations of its N- and
C-terminal charged groups mimic those in dipeptide ligand
conformers. This requires that the C-terminal region (second
peptide bond and the central and C-terminal side chains) is
recognized as a ‘unit’ that fits into the C-terminal pocket.12,14–16
To achieve this, the C-terminal pocket needs to be larger than
the N-terminal pocket and this is exactly what has been
found.31 Consequently, the greater size of FMDP compared
with the largest protein amino acids makes it likely that it is
not easily accommodated in the N-terminal water-filled
pocket but it can be in the C-terminal one. This observation
provides a plausible explanation for the much poorer affinities
of the FMDP-Xaa compounds compared with the Xaa-FMDP
analogues for DppA (Table 2) (and implicitly for Tpp) and for
their lower antibacterial activities (Table 1). FMDP-FMDP
showed no binding to DppA and lacked measurable antibacterial activity. This can now also be rationalized on the
basis of its large side chains and also the fact that about 60% of
its conformers have unrecognizable B2 torsions (data not
shown). With regard to the dimensions of the other glutamine
analogues, the even larger size of FMDB would be expected
to limit severely its transport as part of a peptidomimetic,
which would further diminish its poor enzyme inhibition. The
small size of FCDP supports the finding that DppA binds
FCDP-Ala and FCDP-Met better than FMDP-Ala and
FMDP-Met (Table 2), which endorses the suggestion that the
poor antibacterial activities of FCDP peptidomimetics are
probably a consequence of the intrinsically lower inhibition
of the target enzyme by this analogue. Nature endorses this
influence of side chain size on bioactivity. Thus, most natural
antimicrobial dipeptidomimetics have a protein amino acid as
their N-terminus, commonly Ala, and the larger antimicrobial
compound forms the C-terminus. This is invariably the case
when the inhibitory moiety is large, as in valclavam, polyoxins and nikkomycins.17,22,23,26,27 Interestingly, the natural
antibiotic lindenbein, FCDP-Ala, has the warhead at the
829
N. J. Marshall et al.
N-terminus. However, as shown above, being the smallest of
these glutamine analogues it can be accommodated in the
N-terminal pocket, allowing it to be well bound by DppA
(Table 2), although FCDP has poor inhibitory activity against
the target enzyme (Table 1). Molecular modelling of AlaFCDP, an alternative theoretical lindenbein analogue, shows
that compared with the natural form it possesses a much lower
proportion of conformers that match the MRT for Dpp
(results not shown). It appears that nature chose the form best
suited to deliver the antimicrobial moiety.
Lys-FMDP provides a further example of the ability of side
chains to affect molecular recognition. Its conformers are
mostly A4(B9,B12), which are putative ligands specifically
for Tpp, but it was completely inactive with this transporter.
The explanation for this became clear from inspection of the
modelled conformers in which their side chains interact with
one another. This may improve stability but would seem to
make the ‘fused side chains’ too large to allow binding by
Tpp. With this insight, it would be relatively easy to evaluate
conformer databases of related compounds to identify any in
which the side chains are comparably close and so may constitute an unrecognizable ‘fused unit’.
In conclusion, conformational analysis of these peptidomimetics has provided detailed insights into their structure–
activity relationships. This approach would speed up identification of compounds with optimal bioactivities whilst
dramatically decreasing the need for extensive chemical
syntheses and testing. In addition, molecular modelling could
be performed on, for example, FMDP-containing pseudopeptides, having chemical features such as peptide-bond isosteres
that could enhance overall pharmacokinetics.34 In a broader
context, the study extends the use of the concept of MRTs, and
provides important structural information that will aid the
design not only of antimicrobial compounds but also of all
therapeutic agents targeted for delivery by peptide transporters.
2. Badet, B., Vermoote, P., Haumont, P. Y., Lederer, F. &
LeGoffic, F. (1987). Glucosamine synthetase from Escherichia coli:
purification, properties and glutamine-utilizing site location. Biochemistry 26, 1940–8.
3. Andruszkiewicz, R., Chmara, H., Milewski, S. & Borowski, E.
(1987). Synthesis and biological properties of N-3-(4-methoxyfumaroyl)-L-2,3-diaminopropanoic acid dipeptides, a novel group of
antimicrobial agents. Journal of Medicinal Chemistry 30, 1715–9.
4. Andruszkiewicz, R., Chmara, H., Milewski, S., Kasprzak, L. &
Borowski, E. (1993). Structural determinants of inhibitory activity of
N-3-(4-methoxyfumaroyl)-L-2,3-diaminopropanoic acid towards
glucosamine-6-phosphate synthase. Polish Journal of Chemistry
67, 673–83.
5. Kucharczyk, N., Denisot, M. A., LeGoffic, F. & Badet, B. (1990).
Glucosamine-6-phosphate synthase from Escherichia coli—
determination of the mechanism of inactivation by N-3-fumaroyl-L2,3-diaminopropionic derivatives. Biochemistry 29, 3668–76.
6. Milewski, S., Andruszkiewicz, R., Kasprzak, L., Mazerski, J.,
Mignini, F. & Borowski, E. (1991). Mechanism of action of anticandidal dipeptides containing inhibitors of glucosamine-6phosphate synthase. Antimicrobial Agents and Chemotherapy 35,
36–43.
7. Payne, J. W. (1995). Bacterial peptide permeases as a drug
delivery target. In Peptide Based Drug Design: Controlling Transport
and Metabolism (Taylor, M. D. & Amidon, G. L., Eds), pp. 341–67.
American Chemical Society, Washington, DC, USA.
8. Chmara, H., Milewski, S., Andruszkiewicz, R., Mignini, F. &
Borowski, E. (1998). Antibacterial action of dipeptides containing an
inhibitor of glucosamine-6-phosphate isomerase. Microbiology 144,
1349–58.
9. Andruszkiewicz, R., Milewski, S., Zieniawa, T. & Borowski, E.
(1990). Anticandidal properties of N-3-(4-methoxyfumaroyl)-L2,3-diaminopropanoic acid oligopeptides. Journal of Medicinal
Chemistry 33, 132–5.
10. Milewski, S., Chmara, H., Andruszkiewicz, R., Borowski, E.,
Zaremba, M. & Borowski, J. (1988). Antifungal peptides with novel
specific inhibitors of glucosamine 6-phosphate synthase. Drugs
under Experimental and Clinical Research 14, 461–5.
11. Milewski, S., Mignini, F., Micossi, L. & Borowski, E. (1998). Antihistoplasmal in vitro and in vivo effect of Lys-Nva-FMDP. Medical
Mycology 36, 177–80.
Acknowledgements
We are grateful to Barry M. Grail, Gillian M. Payne and
Craig J. Winstanley for their contributions to these studies.
Financial support for this research was provided by the British
Society for Antimicrobial Chemotherapy (Grant GA313),
and the Biotechnology and Biological Sciences Research
Council (Grant P12847).
References
1. Chmara, H., Andruszkiewicz, R. & Borowski, E. (1986). Inactivation of glucosamine-6-phosphate synthetase from Salmonella
typhimurium Lt2 by fumaroyl diaminopropanoic acid-derivatives, a
novel group of glutamine analogs. Biochimica et Biophysica Acta
870, 357–66.
12. Payne, J. W., Grail, B. M. & Marshall, N. J. (2000). Molecular
recognition templates of peptides: driving force for the molecular
evolution of peptide transporters. Biochemical and Biophysical
Research Communications 267, 283–9.
13. Payne, J. W., Marshall, N. J., Grail, B. M. & Gupta, S. (2002).
Conformer profiles and biological activities of peptides. Current
Organic Chemistry 6, 1221–46.
14. Payne, J. W., Grail, B. M., Gupta, S., Ladbury, J. E., Marshall,
N. J., O’Brien, R. et al. (2000). Structural basis for recognition of
dipeptides by peptide transporters. Archives of Biochemistry and
Biophysics 384, 9–23.
15. Grail, B. M. & Payne, J. W. (2000). Predominant torsional forms
adopted by dipeptide conformers in solution: parameters for
molecular recognition. Journal of Peptide Science 6, 186–99.
830
Structure–activity relationship of glutamine analogues
16. Marshall, N. J., Grail, B. M. & Payne, J. W. (2001). Predominant
torsional forms adopted by oligopeptide conformers in solution:
parameters for molecular recognition. Journal of Peptide Science 7,
175–89.
17. Payne, J. W. & Marshall, N. J. (2001). Peptide Transport. In
Microbial Transport Systems (Winkelman, G., Ed.), pp. 139–64.
Wiley-VCH, Weinheim, Germany.
18. Smith, M. W., Tyreman, D. R., Payne, G. M., Marshall, N. J. &
Payne, J. W. (1999). Substrate specificity of the periplasmic
dipeptide-binding protein from Escherichia coli: experimental basis
for the design of peptide prodrugs. Microbiology 145, 2891–901.
19. Andrews, J. M. (2001). BSAC standardized disc susceptibility
testing method. Journal of Antimicrobial Chemotherapy 48, Suppl.
S1, 43–57.
20. Grail, B. M. & Payne, J. W. (2002). Conformational analysis of
bacterial cell wall peptides indicates how particular conformations
have influenced the evolution of penicillin-binding proteins, β-lactam
antibiotics and antibiotic resistance mechanisms. Journal of
Molecular Recognition 15, 113–25.
21. Marshall, N. J. & Payne, J. W. (2001). The importance of
electrostatic charge and dielectric constant in conformational analysis of biologically active dipeptides. Journal of Molecular Modelling
7, 112–9.
22. Payne, J. W. & Smith, M. W. (1994). Peptide transport by microorganisms. Advances in Microbial Physiology 36, 1–80.
23. Tyreman, D. R., Smith, M. W., Marshall, N. J., Payne, G. M.,
Schuster, C. M., Grail, B. M. et al. (1998). Peptides as prodrugs: the
smugglin concept. In Peptides in Mammalian Protein Metabolism:
Tissue Utilization and Clinical Targeting (Grimble, B. M. & Backwell,
F. R. C., Eds), pp. 141–57. Portland Press, London, UK.
24. Payne, J. W., Payne, G. M., Gupta, S., Marshall, N. J. & Grail,
B. M. (2001). Conformational limitations of glycylsarcosine as a
prototypic substrate for peptide transporters. Biochimica et Biophysica Acta 1514, 65–75.
25. Brandsch, M., Knutter, F., Thunecke, F., Hartrodt, B., Born, I.,
Borner, V. et al. (1999). Decisive structural determinants for the
interaction of proline derivatives with the intestinal H+/peptide symporter. European Journal of Biochemistry 266, 502–8.
26. Ringrose, P. S. (1985). Peptides as antimicrobial agents. In
Scientific Basis of Antimicrobial Chemotherapy (Greenwood, D. &
O’Grady, F., Eds), pp. 219–66. Cambridge University Press, Cambridge, UK.
27. Taylor, M. D. & Amidon, G. L. (Eds) (1995). Peptide Based Drug
Design: Controlling Transport and Metabolism. American Chemical
Society, Washington, DC, USA.
28. Doring, F., Will, J., Amasheh, S., Clauss, W., Ahlbrecht, H. &
Daniel, H. (1998). Minimal molecular determinants of substrates for
recognition by the intestinal peptide transporter. Journal of Biological Chemistry 273, 23211–8.
29. Bailey, P. D., Boyd, C. A. R., Bronk, J. R., Collier, I. D.,
Meredith, D., Morgan, K. M. et al. (2000). How to make drugs
orally active: a substrate template for peptide transporter PepT1.
Angewandte Chemie International Edition 39, 506–8.
30. Davies, T. G., Hubbard, R. E. & Tame, J. R. H. (1999). Relating
structure to thermodynamics: the crystal structures and binding
affinity of eight OppA–peptide complexes. Protein Science 8,
1432–44.
31. Dunten, P. & Mowbray, S. L. (1995). Crystal structure of the
dipeptide binding protein from Escherichia coli involved in active
transport and chemotaxis. Protein Science 4, 2327–34.
32. Sleigh, S. H., Seavers, P. R., Wilkinson, A. J., Ladbury, J. E. &
Tame, J. R. H. (1999). Crystallographic and calorimetric analysis of
peptide binding to OppA protein. Journal of Molecular Biology 291,
393–415.
33. Tame, J. R. H., Sleigh, S. H., Wilkinson, A. J. & Ladbury, J. E.
(1996). The role of water in sequence-independent ligand binding
by an oligopeptide transporter protein. Nature Structural Biology 3,
998–1001.
34. Gupta, S. & Payne, J. W. (2001). Evaluation of the conformational propensities of peptide bond analogues as a basis for
selecting bioactive peptidomimetics. Journal of Peptide Research
58, 546–62.
831