Use of model cell membranes to demonstrate templated binding of vancomycin group antibiotics

摘要

J. Chem. Soc. Perkin Trans. 1 1997 2911 Use of model cell membranes to demonstrate templated binding of vancomycin group antibiotics Andrew C. Try Gary J. Sharman Robert J. Dancer Ben Bardsley Richard M. H. Entress and Dudley H. Williams *,dagger; Cambridge Centre for Molecular Recognition University Chemical Laboratory Lensfield Road Cambridge UK CB2 1EW In this paper we demonstrate the importance of binding geometry and dimerisation at the surface of model cell membranes in the mode of action of the clinically important glycopeptide antibiotics. This has been achieved through the use of model cell membranes (micelles and vesicles) to which cell wall analogues are anchored via a hydrophobic decanoyl chain. A number of ndash;D-Ala-terminating cell wall analogues ranging from two to six residues in length have been used.Dipeptide pentapeptide and hexapeptide display enhanced binding to the antibiotic at the model cell surface but tripeptide and tetrapeptide do not. The possible implications of the observed binding geometries for bacterial systems are discussed. Introduction The vancomycin group of antibiotics kill Gram-positive bacteria by binding to cell wall precursors terminating in ndash;Lys-DAla- D-Ala,1 preventing peptidoglycan polymerisation and subsequent cross-linking and in doing so weakening the cell wall ultimately causing cell lysis.2 Two members of the family vancomycin and teicoplanin are clinically important in the treatment of methicillin-resistant Staphylococcus aureus (MRSA) and are currently the last line of defence against such infections. The binding of the vancomycin group of antibiotics to model cell wall precursor peptides such as di-N-acetyl-Lys-D-Ala-DAla (Ac-tri-Ala Dagger;) in free solution has been studied extensively and these studies provide valuable insights into their mode of action (Fig.1).3ndash;5 We have also shown that almost all glycopeptide antibiotics dimerise and that dimerisation of the antibiotic is in all but one case cooperative with ligand binding i.e. the antibiotic dimerisation constant is higher in the presence of ligand than when free and the ligand binds with a higher affinity to antibiotic dimer than to antibiotic monomer.6 We therefore hypothesised that dimerisation might play an important role in the mode of action of these antibiotics in that the second binding event between a dimer and the surface of a growing bacterium would be effectively intramolecular thus allowing a chelate-like enhancement of binding.7 We have subsequently provided evidence for this hypothesis through the use of in vitro bacterial assays which reveal a correlation between dimerisation constant and the ability of the antibiotic to kill bacteria in the presence of competing peptides.8 To further demonstrate the importance of dimerisation in the mode of action of glycopeptide antibiotics we have devised model membrane systems designed to mimic the growing bacterial cell wall thus allowing for the expression of any binding enhancement due to dimerisation at a surface.In these models the bacterial cell membrane is represented by sodium dodecyl sulfate (SDS) micelles or phosphatidylcholine (PC) vesicles and the cell wall precursors by cell wall peptide analogues with an N-terminal decanoyl chain designed to insert into the model membranes.This N-terminal membrane anchor is similar to the dagger; E-Mail dhw1@cam.ac.uk Dagger; All peptides used in this study are abbreviated in this style in order to indicate the nature of their a-acyl chain (acetyl = Ac decanoyl = dec docosanoyl = docos) the number of residues (di tri tetra penta or hexa based on the sequence Gly-L-Ala-D-g-Glu-Lys-D-Ala-D-Ala counting from the C-terminus) and the identity of the C-terminal residue (ndash;D-Alanine = Ala or ndash;D-Lactate = Lac). C11 chain present in the antibiotic teicoplanin which we have previously shown to associate with model membranes.9 The whole arrangement of micelle/vesicle and anchored cell wall analogue is therefore similar to that at the bacterial cell surface where cell wall precursors are anchored to the cell membrane via a C55 hydrocarbon chain.10 Using such a system we have demonstrated that for a dimerndash; ligand complex dimerisation and membrane anchoring result in enhanced binding affinity.11 This was achieved by comparing the binding constants of ristocetin A (another member of the vancomycin group of antibiotics) to the cell wall analogues Na- decanoyl-D-Ala-D-Ala (dec-di-Ala) and N-a-acetyl-D-Ala-DFig.1 Exploded view of the complex formed between chloroeremomycin (CE) or biphenylchloroeremomycin (BCE) (the two members of the vancomycin group of antibiotics studied in this paper) and di-Nacetyl- Lys-D-Ala-D-Ala. Protons referred to in the text are labelled. Aromatic rings are numbered according to residue number.N O N H N O H O N H O OH H O H NH2Me H H H O O N O H H H H O N O ndash; O2C H OH OH HO HO CH3 O H3C O OH O HO HO O ndash; O H N N H3C H O CH3 H H O NH2 O H3N HO CH3 O H3C RH2N H Cl N H3C O HN H CH3 O H Cl H 7 5 4 2 + + 7d 7f w2 b d 6 e e-Ac BCE R = CE R = H + g 2912 J. Chem. Soc. Perkin Trans. 1 1997 Ala (Ac-di-Ala) in the presence of SDS micelles. The complex formed between the antibiotic and Ac-di-Ala results from an intermolecular association Fig. 2(a) whereas the complex formed on the surface of a micelle with dec-di-Ala is essentially intramolecular Fig. 2(b). However inspection of Coreyndash; Paulingndash;Kulton (CPK) models reveals that this peptide might be rather short and therefore limited in its ability to anchor to the model cell membrane when bound to the dimeric antibiotic.We have therefore decided to investigate the behaviour of a number of longer anchored peptides (ranging from two to six residues) in their complexes with antibiotic dimers in the model system predicting increased anchoring from the longer peptides which should act as better models of the natural peptides. The sequence followed for the peptides was that of the natural cell wall precursor namely ndash;L-Ala-D-g-Glu-Lys-D-Ala-D-Ala with an N-terminal glycine added as a spacer in the hexapeptide ligand (Fig. 3). In this paper we have therefore attempted to investigate the effect of the length of the peptide portions of ndash;D-Alaterminating cell wall precursor analogues on the enhancement of binding observed with glycopeptides at model cell surfaces. The influence that peptide length has on the geometry of binding to the antibiotic is also considered.Results and discussion Binding on a micelle surface Initially the binding of the ligands was assayed by a simple NMR method which involved measurement of the limiting chemical shift of the amide residue 2-proton (w2) of the antibiotic (labelled in Fig. 1) under conditions where the antibiotic was 95 bound (dw2 lim). In the complex between antibiotic and ligand w2 is involved in a crucial hydrogen bond to the ligand carboxylate such that its chemical shift moves dramatically downfield upon binding. We have previously observed a strong correlation between dw2 lim and the free energy of ligand binding.11ndash;14 It is therefore possible to use dw2 lim as a qualitative measure of binding affinity. The glycopeptide antibiotic chloroeremomycin (CE LY264826 Fig.1) was chosen for these studies as it binds ligand strongly and has a high dimerisation constant.6 It was therefore anticipated that it might exhibit a high degree of cooperativity when binding at a surface. The results obtained for dw2 lim with this antibiotic upon binding to the decanoyl ligands in the presence of SDS micelles are shown in Table 1. It should be noted that additional N-terminal residues beyond lysine add a negligible amount to the intrinsic binding energy but removal of the lysine does significantly reduce this intrinsic binding energy.15 The dw2 lim for CE bound to Ac-tri-Ala in the presence of SDS micelles was observed to be 11.11 ppm. The value when bound Fig. 2 Schematic illustration of a glycopeptide dimer binding cell wall analogues at the surface of a micelle.(a) The complex formed with Acdi- Ala is intermolecular whereas (b) that with dec-di-Ala is essentially intramolecular. to Ac-di-Ala in the presence of SDS was not observed due to exchange-broadening of the spectrum but is expected to be lower than 11.11 ppm as Ac-di-Ala binds more weakly to CE than Ac-tri-Ala. However for CE bound to dec-di-Ala in the presence of SDS dw2 lim is 11.50 ppmmdash;much further downfield than dw2 lim when bound even to the longer peptide Ac-tri-Ala. This significant increase in dw2 lim when bound to the decanoyl ligand is consistent with observations made previously with the antibiotic ristocetin A i.e. an enhancement in binding affinity for the intramolecular system. However the similarity of dw2 lim for CE bound to N-a-decanoyl-N-e-acetyl-Lys-D-Ala-D-Ala (dec-tri-Ala) and N-a-decanoyl-D-g-Glu-N-e-acetyl-Lys-D-Ala- D-Ala (dec-tetra-Ala) in the presence of SDS to that when bound to Ac-tri-Ala suggested a negligible enhancement of binding in these systems (11.13 and 11.15 ppm for dec-tri-Ala and dec-tetra-Ala respectively vs.11.11 ppm for Ac-tri-Ala). The shifts when bound to N-a-decanoyl-L-Ala-D-g-Glu-N-eacetyl- Lys-D-Ala-D-Ala (dec-penta-Ala) and N-a-decanoyl- Gly-L-Ala-D-g-Glu-N-e-acetyl-Lys-D-Ala-D-Ala (dec-hexa- Ala) (11.18 and 11.25 ppm respectively) suggest that as the peptides become longer there is some enhancement to the binding of the anchored ligand relative to the binding of the unanchored Ac-tri-Ala but substantially less enhancement than for binding to dec-di-Ala. The physical reasons for this enhancement to the binding of the anchored dipeptide ligand but not to the binding of anchored tripeptide are discussed later in this paper.Binding on a vesicle surface To further investigate these complexes we used a model system in which the SDS micelles were replaced by PC vesicles which are of a closer size surface charge and curvature to the natural cell membrane. These features make them a better mimic of the bacterial membrane surface which is composed of similar lipids. However they are unsuitable for the w2 binding assay described above since their large size results in a very long correlation time and consequently very broad NMR resonances. An antibiotic associated with the vesicle surface via binding to anchored ligands thus takes on the tumbling properties of the vesicle itself with the effect that its resonances also appear as broad lines.Whilst this renders any high-resolution structural analysis almost impossible we used this broadening to determine whether a complex was indeed anchored to the surface of the vesicle. The results of these experiments are shown in Fig. 4. This shows the aromatic region of the 1H NMR spectra of the complexes of CE with various length peptides in the presence of PC vesicles. For the complexes of CE with dec-di-Ala and dechexa- Ala no signals were observed. This was interpreted as being due to tight binding of CE to the peptides on the surface of the vesicle resulting in fully broadened NMR resonances i.e. all of the antibiotic was associated via the bound peptide to the surface of the vesicle.In these cases the broadening was probably exacerbated by some aggregation of the vesicles; the sample became visibly turbid. The complex with dec-penta-Ala also shows some degree of broadening but those with dec-tri- Ala and dec-tetra-Ala both result in well-resolved spectra. In these three cases (dec-tri- -tetra- and -penta-Ala) the ligand Table 1 Limiting chemical shifts of the residue 2 amide proton (w2) dw2 lim of CE with various decanoyl ligands in the presence of SDS at pH 4.5 Ligand dec-di-Ala dec-tri-Ala dec-tetra-Ala dec-penta-Ala dec-hexa-Ala dw2 lim 11.50 11.13 11.15 11.18 11.25 J. Chem. Soc. Perkin Trans. 1 1997 2913 Fig. 3 Structures of the Ala-terminating decanoyl peptides used in these studies CO2H HN NH HN NH O O O HN O CO2H HN NH O O CO2H HN NH HN O O O HN O O CO2H CO2H HN NH HN NH HN O O O HN O O CO2H O CO2H HN NH HN NH HN NH O O O HN O O CO2H O O N-a-dec-D-Ala-D-Ala (dec-di-Ala) N-a-dec-N-e-Ac-Lys-D-Ala-D-Ala (dec-tri-Ala) N-a-dec-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-tetra-Ala) N-a-dec-L-Ala-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-penta-Ala) N-a-dec-Gly-L-Ala-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-hexa-Ala) was confirmed to be bound to the antibiotic by the downfield shift of the proton w2 indicative of ligand binding.However the lack of broadening of the resonances for the complexes with dec-tri-Ala and dec-tetra-Ala suggested that even though the ligand was bound to the antibiotic it was not simultaneously associated with the vesicles. In the case of dec-penta-Ala the partial broadening observed indicates that the antibioticndash;ligand complexes are in slow exchange on the NMR timescale between being vesicle-associated and free in solution.Competition experiments We have also used a competition strategy illustrated schematically in Fig. 5 to estimate a binding constant of dec-hexa-Ala to CE on the surface of vesicles. In these experiments a nonanchored cell wall analogue Ac-tri-Ala (1.0 mM) was added to a solution of CE (0.5 mM) bound to dec-hexa-Ala (1.0 mM) on the surface of vesicles. The results of this experiment and a control experiment in the absence of dec-hexa-Ala showed (Fig. 6) that even in the presence of an excess of the antagonist most of the antibiotic appears to remain in an anchored complex on the surface of the vesicle. Based on the relative peak integrals in the presence and absence of dec-hexa-Ala Figs.6(d ) and (b) and concentrations of added Ac-tri-Ala in these two solutions a binding enhancement of approximately one hundred-fold is estimated for the anchored hexapeptide assembly over the non-anchored Ac-tri-Ala/CE complex. This is similar to the four hundred-fold binding enhancement due to templating measured for ristocetin A binding to dec-di-Ala in the presence of SDS micelles.11 Therefore the above observations can be rationalised by the hypothesis that dec-tri-Ala and dec-tetra-Ala when bound to the antibiotic are simply not long enough for the alkyl chain to insert into the model membrane Fig. 7(a) whereas dec-penta- Ala and more effectively dec-hexa-Ala are of sufficient length Fig. 7(b). If this is the case the seemingly anomalous cooperativity exhibited by the dec-di-Ala/CE complex when anchoring to micelles or vesicles might arise as a result of some alternative binding geometry.It was anticipated that when binding in vivo the antibiotics would normally interact with cell wall peptides at the bacterial 2914 J. Chem. Soc. Perkin Trans. 1 1997 Fig. 4 Aromatic region of the 500 MHz 1H NMR spectra (D2O pD 6.2) of CE (0.5 mM) with the various decanoyl peptides (1 mM) in the presence of vesicles (10 mM). In the case of dec-di-Ala and dec-hexa-Ala the signals from the complex are broadened due to anchoring to the vesicle. cell surface with their ring-4 sugars the disaccharide glucose:4- epi-vancosamine in CE (Fig. 1) in contact or near to the cell surface. This supposition was based on the fact that teicoplanin 9,16 and the semi-synthetic glycopeptide biphenylchloroeremomycin (BCE LY307599 Fig.1) 17 have putative membrane anchors attached to their ring-4 sugars. Additionally the Fig. 5 Schematic illustration of the competition experiment in which antibiotic dimer bound to anchored ligands is displaced from the surface of a vesicle Fig. 6 Aromatic region of the 500 MHz 1H NMR spectra (D2O pD 6.2) of (a) CE (0.5 mM) in the presence of vesicles (10 mM) and (b) following the addition of Ac-tri-Ala (1 mM). (c) The same region of the spectrum of CE (0.5 mM) in the presence of vesicles (10 mM) and dechexa- Ala (1 mM) and (d) following the addition of Ac-tri-Ala (4 mM) (* these resonances result from residual amide protons of the added ligand). natural precursor peptide analogue penta-Ala has been shown to bind CE with its N-terminal L-alanine residue adjacent to the ring-4 sugars.18 Inspection of CPK models suggested that if the decanoyl chain of dec-di-Ala were to pass over ring-7 rather than adjacent to ring-6 (which is the orientation necessary if the ring-4 sugars are to interact with the micelle surface) there would be more of the hydrocarbon chain available to insert into the micelle.We therefore hypothesised that in binding to the anchored dec-di-Ala ligand CE might be oriented lsquo;upside downrsquo; i.e. with the sugars projecting away from the micelle Fig. 7(c) thus allowing the decanoyl chain to pass over ring-7 of the antibiotic and insert into the membrane. Such a conformation is not available to longer peptides because the lysine side chain occupies this position over ring-7.5,18ndash;21 With the lysine side-chain positioned over ring-7 the L-stereochemistry of the lysine forces the additional residues of the longer peptides and indeed the alkyl chain into an orientation in which they are directed toward the ring-4 sugars.In the case of binding by dec-tri-Ala and dec-tetra-Ala the alkyl chain is then not long enough to be able to insert into the membrane Fig. 7(a); it is only for dec-penta-Ala and dec-hexa-Ala that this is possible Fig. 7(b) as observed experimentally. Fig. 7 Schematic representation of the hypothesis which explains lack of templating for dec-tri-Ala and dec-tetra-Ala in the presence of vesicles. (a) Dec-tri-Ala is not long enough to reach the surface of the model membrane when bound to antibiotic whereas (b) dec-hexa-Ala is able to both anchor and bind antibiotic.(c) Dec-di-Ala can both anchor to the model membrane and bind antibiotic but only when oriented lsquo;upside downrsquo; (a geometry not accessible to longer peptides because of the presence of the lysine side chain). J. Chem. Soc. Perkin Trans. 1 1997 2915 To test the validity of this hypothesis experiments were performed with two new ligands N-a-decanoyl-Gly-D-Ala-D-Ala and N-a-docosanoyl-N-e-acetyl-Lys-D-Ala-D-Ala (docos-tri- Ala docosanoyl = C22) binding to CE in the presence of vesicles. It was anticipated that for the former ligand the absence of the lysine side chain would allow it to take up a conformation in the binding pocket similar to that of dec-di- Ala with the decanoyl chain passing over ring-7 so that binding to the antibiotic on the vesicle surface would be facilitated.For docos-tri-Ala it was predicted that the binding geometry would be the same as for dec-tri-Ala but that the much longer hydrocarbon chain would now be able to reach the membrane again allowing binding of the antibiotic to the vesicle surface. The results were exactly as anticipated with the complexes of CE bound to each ligand in the presence of vesicles resulting in substantially broadened antibiotic 1H NMR signals. Also the addition of Ac-tri-Ala did not result in the antibiotic becoming displaced from its complexes with the anchored ligands on the surface of vesicles (Fig. 8 shows the aromatic region of the spectra obtained with docos-tri-Ala). One possible alternative explanation for the difference in binding to dec-di-Ala dec-penta-Ala and dec-hexa-Ala compared to dec-tri-Ala and dec-tetra-Ala could lie in the relative abilities of the peptides to form self-micelles or self-vesicles.This self-association could thus be responsible for the observed 1H NMR line broadening and enhancement of binding of the first three named peptides to CE instead of the templated binding with antibiotics on PC vesicle surfaces as described above. However 1H NMR spectra of these ligands in solution with vesicles at the same concentration to that used in the binding experiments show that their signals all broaden to a similar degree (although predictably the C-terminal signals of the longer peptides are sharper than their N-terminal signals) indicating that all of the ligands associate with the PC vesicles to a similar extent. Additionally given the relatively low concentration of ligands used in these experiments (typically twice that of antibiotic) aggregates composed entirely of such ligands would possess only sufficient surface area to bind a small fraction of the antibiotic present and therefore could not lead to the complete broadening of signals observed particularly for dec-di-Ala and dec-hexa-Ala.Confirmation of ligand orientation To further support the hypothesis outlined in the previous section two-dimensional NOESY spectra were acquired in order to identify the orientation of the decanoylated peptides with respect to the antibiotic in the presence of SDS. For the complex of CE with dec-di-Ala where the decanoyl chain was anticipated to pass over ring-7 in the lsquo;upside downrsquo; arrange- Fig. 8 Aromatic region of the 1H NMR spectra (D2O pD 6.2) of (a) CE (0.5 mM) in the presence of vesicles (10 mM) (b) following the addition of docos-tri-Ala (1 mM) cf.Fig. 6(b) and (c) after the addition of Ac-tri-Ala (2 mM) ment NOESY cross peaks were observed between the aromatic protons of ring-7 of the antibiotic and the methylene protons of the decanoyl chain Fig. 9(a). From this data we conclude only that the decanoyl chain of dec-di-Ala must lie over ring-7. A more precise interpretation of the data does not seem to be warranted because of (i) the anticipated dynamic behaviour of this portion of the ligand and (ii) spin diffusion. Thus the decanoyl chain projects from the antibiotic toward the micelle (or vesicle) in an orientation which places the ring-4 sugars away from the surface of the model membrane.For the complex with dec-tri-Ala NOESY cross peaks were observed from the lysine side chain to ring-7 Fig. 9(b) and the methylene groups of the decanoyl chain gave cross peaks to residue 6 of the antibiotic. This positions the decanoyl chain such that it could anchor the complex to a membrane in the case of either the longer peptides (dec-penta-Ala and dec-hexa-Ala) or the tripeptide with a longer acyl chain (docos-tri-Ala) while simultaneously positioning the ring-4 sugars adjacent to the membrane surface. Thus the orientation of the ligands that was predicted by the results of the dw2 lim measurements and vesicle binding experiments was shown to exist by two-dimensional NMR spectroscopy. Three-dimensional representations of the complexes formed between CE and dec-di-Ala and dec-tri-Ala are illustrated in Fig.10(a) and 10(b) respectively. Conclusions We have employed two model membrane systems in an attempt to establish the optimum conditions for the expression of cooperativity due to binding of ligands to glycopeptides on a surface. Each system has its advantages and disadvantages. SDS is available in deuterated form and the micelles it forms are small in diameter. These features enable complexes formed on the surface of micelles to be studied by high-resolution 1H NMR spectroscopy providing a wealth of structural information. However the small diameter of the SDS micelles (approximately 25ndash;30 Aring;) results in a high degree of curvature at the surface leading perhaps to a non-ideal binding geometry with certain ligands. PC vesicles are much larger in diameter (1000ndash; 10 000 Aring;) and are thus expected to overcome any problems related to surface curvature.The size of PC vesicles precluded a detailed study of vesicle-bound complexes by NMR spectroscopy but allowed a qualitative determination of the extent to which templated binding was achieved for CE binding to a series of decanoylated ligands. We have thus been able to demonstrate in a direct manner the importance of binding geometry and dimerisation in the mode of action of these antibiotics. We believe that the origin of this enhanced binding lies in the chelate-like enhancement conveyed Fig. 9 Portions of the NOESY spectra of the complexes formed in the presence of micelles between CE and (a) dec-di-Ala and (b) dec-tri- Ala illustrating the cross peaks from ring-7 of the antibiotic to ligand protons.In (a) lsquo;dec CH2rsquo; is used to signify those methylene groups of the decanoyl chain which have unresolved chemical shifts. 2916 J. Chem. Soc. Perkin Trans. 1 1997 by the ability of these antibiotics to dimerise at the cell surface resulting in a tightening of all interactions within the complex thus giving rise to enthalpic as well as entropic gains.22 The greatest degree of templated binding was achieved with the longest and also paradoxically the shortest of the anchored ligands studied. We have shown semi-quantitatively that the enhancement to binding to CE due to templating for dec-hexa- Ala is similar to that for dec-di-Ala measured previously.11 We have put forward a physical model with accompanying evidence which accounts for this anomaly. These results provide a rationale as to why membrane anchors on naturally occurring glycopeptide antibiotics are located on the ring-4 sugars,16 and why membrane anchors on the most active semi-synthetic antibiotics are similarly located.17 The L-stereochemistry of the lysine residue present in Fig.10 Three dimensional representations of half of the dimeric complexes formed between CE and (a) dec-di-Ala and (b) dec-tri-Ala. Note how the orientation of the decanoyl chain attached to the dipeptide results in more exposure of hydrocarbon than the chain attached to the tripeptide. cell wall peptides directs the antibiotic to bind in a fashion that places the residue-4 sugars in close proximity to the bacterial cell membrane. Nature takes advantage of this by placing a locating device (in the case of teicoplanin a C11 acyl chain) at precisely this point.Despite this rationale we do not preclude the possibility of enhancement of antibiotic action through the location of hydrophobic chains at alternative sites. The model systems studied in this work thus present a more detailed picture of how the vancomycin group antibiotics function in biological systems. The results support the hypothesis that these antibiotics bind to nascent bacterial cell walls with their ring-4 saccharides adjacent to the cell membrane and the parallel nature of these saccharides in the antibiotic dimers may reflect a similar parallel arrangement of the peptidoglycan strands of growing cell wall. Experimental sect; Preparation of phosphatidylcholine vesicles Type XV1-E L-a-phosphatidylcholine from fresh egg yolk (Sigma 80 mg) was dissolved in chloroform (2 ml) which had been rendered ethanol-free by passage through a column of activated alumina.The solution was then evaporated under reduced pressure to yield a thin film on the wall of the flask. The flask was flushed with nitrogen followed by addition of D2O (5 ml) or 50 mM NaH2PO4 pH 6.2 buffer (5 ml). The mixture was shaken for 20 min then sonicated for 90 min to yield a slightly turbid suspension of vesicles (20 mM phosphatidylcholine). 1H NMR spectroscopy Sodium 2H25dodecyl sulfate (SDS; 98 atomD) was purchased from Euriso-top. All 1H NMR spectroscopy experiments were performed on 500 MHz Bruker DRX-500 and AM500 spectrometers at 300 K. Suppression of the solvent resonance was achieved using WATERGATE23 or pre-saturation. One-dimensional spectra were recorded using 32k complex data points.In two-dimensional experiments 4k complex points were acquired in f2 with 512 increments in f1. TPPI was used to achieve quadrature detection in the indirect dimension. Data was processed with XWIN-NMR software using a sinesquared window function and zero-filling in f1 up to 1k or 2k points. Two-dimensional NOESY experiments employed mixing times ranging through 50ndash;150 ms and were used to confirm all w2 assignments. In experiments involving micelles or vesicles the ligand was added to the vesicle/micelle solution and the mixture was sonicated to facilitate insertion. Experiments involving SDS employed a concentration of 70 mM SDS (above the SDS critical micelle concentration) 5 mM antibiotic and 10 or 20 mM ligand. These concentrations ensured that a high (90) proportion of antibiotic was bound by ligand but also that the SDS was not lsquo;overloadedrsquo; with ligand.Only when dissolution was complete was the antibiotic added. In the vesicle experiments vesicles were prepared as described above and used as a 10 mM solution; concentrations of antibiotics used were as described in the individual figure legends. In the competition experiments the unanchored ligand was added to the NMR tubes as a concentrated solution (50 mM) so as not to change the concentration of the contents of the tube signifi- cantly and to allow accurate concentrations to be achieved on the addition of the appropriate volume. Acknowledgements Eli Lilly and Co. are thanked for generously providing samples of CE (LY264826) and BCE (LY307599).The EPSRC (G. J. S.) sect; Details of the synthesis of the peptides are available as supplementary material (SUP 57275; 33 pp.) deposited with the British Library. Details are available from the editorial office. J. Chem. Soc. Perkin Trans. 1 1997 2917 Xenova (A. C. T.) The Wellcome Trust (R. J. D.) the EPSRC and Roussel (B. B.) and the BBSRC and Roussel (R. M. H. E.) are thanked for financial support. We also thank the Biomedical NMR Centre NIMR Mill Hill London for access to NMR equipment. References 1 A. N. Chatterjee and H. R. Perkins Biochem. Biophys. Res. Commun. 1966 24 489. 2 D. C. Jordan and P. E. Reynolds in Antibiotics ed. J. W. Corcoran and F. E. Hahn Springer-Verlag Berlin 1974 vol. III. 3 J. R. Kalman and D. H. Williams J. Am. Chem. Soc. 1980 102 906. 4 D.H. Williams M. P. Williamson D. W. Butcher and S. J. Hammond J. Am. Chem. Soc. 1983 105 1332. 5 J. C. J. Barna D. H. Williams and M. P. Williamson J. Chem. Soc. Chem. Commun. 1985 254. 6 J. P. Mackay U. Gerhard D. A. Beauregard R. A. Maplestone and D. H. Williams J. Am. Chem. Soc. 1994 116 4573. 7 J. P. Mackay U. Gerhard D. A. Beauregard M. S. Westwell M. S. Searle and D. H. Williams J. Am. Chem. Soc. 1994 116 4581. 8 D. A. Beauregard D. H. Williams M. N. Gwynn and D. J. C. Knowles Antimicrob. Agents Chemother. 1995 39 781. 9 M. S. Westwell U. Gerhard and D. H. Williams J. Antibiot. 1995 48 1292. 10 J. M. Ghuysen in Topics in Antibiotic Chemistry ed. P. G. Sammes Ellis Horwood Chichester 1980 vol. 5 p. 31. 11 M. S. Westwell B. Bardsley R. J. Dancer A. C. Try and D. H. Williams Chem.Commun. 1996 589. 12 P. Groves M. S. Searle M. S. Westwell and D. H. Williams J. Chem. Soc. Chem. Commun. 1994 1519. 13 G. J. Sharman M. S. Searle B. Benhamu P. Groves and D. H. Williams Angew. Chem. Int. Ed. Engl. 1995 34 1483. 14 M. S. Searle G. J. Sharman P. Groves B. Benhamu D. A. Beauregard M. S. Westwell R. J. Dancer A. J. Maguire A. C. Try and D. H. Williams J. Chem. Soc. Perkin Trans. 1 1996 2781. 15 M. Nieto and H. R. Perkins Biochem. J. 1971 123 780. 16 J. C. J. Barna D. H. Williams D. J. M. Stone T.-W. C. Leung and D. M. Doddrell J. Am. Chem. Soc. 1984 106 4895. 17 R. D. G. Cooper N. J. Snyder M. J. Zweifel. M. A. Staszak S. C. Wilkie T. I. Nicas D. L. Mullen T. F. Butler M. J. Rodriguez B. E. Huff and R. C. Thompson J. Antibiot. 1996 49 575. 18 W. G. Prowse A. D. Kline M.A. Skelton and R. J. Loncharich Biochemistry 1995 34 9632. 19 S. W. Fesik T. J. Orsquo;Donnell R. T. Gampe and E. T. Olejniczak J. Am. Chem. Soc. 1986 106 3165. 20 P. Groves M. S. Searle J. P. Mackay and D. H. Williams Structure 1994 2 747. 21 P. Groves M. S. Searle J. P. Waltho and D. H. Williams J. Am. Chem. Soc. 1995 117 7958. 22 M. S. Searle M. S. Westwell and D. H. Williams J. Chem. Soc. Perkin Trans. 2 1995 141. 23 M. Piotto V. Saudek and V. Sklenaacute;r J. Biomol. NMR 1992 2 661. Paper 7/01880K Received 18th March 1997 Accepted 19th June 1997 J. Chem. Soc. Perkin Trans. 1 1997 2911 Use of model cell membranes to demonstrate templated binding of vancomycin group antibiotics Andrew C. Try Gary J. Sharman Robert J. Dancer Ben Bardsley Richard M. H. Entress and Dudley H.Williams *,dagger; Cambridge Centre for Molecular Recognition University Chemical Laboratory Lensfield Road Cambridge UK CB2 1EW In this paper we demonstrate the importance of binding geometry and dimerisation at the surface of model cell membranes in the mode of action of the clinically important glycopeptide antibiotics. This has been achieved through the use of model cell membranes (micelles and vesicles) to which cell wall analogues are anchored via a hydrophobic decanoyl chain. A number of ndash;D-Ala-terminating cell wall analogues ranging from two to six residues in length have been used. Dipeptide pentapeptide and hexapeptide display enhanced binding to the antibiotic at the model cell surface but tripeptide and tetrapeptide do not. The possible implications of the observed binding geometries for bacterial systems are discussed.Introduction The vancomycin group of antibiotics kill Gram-positive bacteria by binding to cell wall precursors terminating in ndash;Lys-DAla- D-Ala,1 preventing peptidoglycan polymerisation and subsequent cross-linking and in doing so weakening the cell wall ultimately causing cell lysis.2 Two members of the family vancomycin and teicoplanin are clinically important in the treatment of methicillin-resistant Staphylococcus aureus (MRSA) and are currently the last line of defence against such infections. The binding of the vancomycin group of antibiotics to model cell wall precursor peptides such as di-N-acetyl-Lys-D-Ala-DAla (Ac-tri-Ala Dagger;) in free solution has been studied extensively and these studies provide valuable insights into their mode of action (Fig.1).3ndash;5 We have also shown that almost all glycopeptide antibiotics dimerise and that dimerisation of the antibiotic is in all but one case cooperative with ligand binding i.e. the antibiotic dimerisation constant is higher in the presence of ligand than when free and the ligand binds with a higher affinity to antibiotic dimer than to antibiotic monomer.6 We therefore hypothesised that dimerisation might play an important role in the mode of action of these antibiotics in that the second binding event between a dimer and the surface of a growing bacterium would be effectively intramolecular thus allowing a chelate-like enhancement of binding.7 We have subsequently provided evidence for this hypothesis through the use of in vitro bacterial assays which reveal a correlation between dimerisation constant and the ability of the antibiotic to kill bacteria in the presence of competing peptides.8 To further demonstrate the importance of dimerisation in the mode of action of glycopeptide antibiotics we have devised model membrane systems designed to mimic the growing bacterial cell wall thus allowing for the expression of any binding enhancement due to dimerisation at a surface.In these models the bacterial cell membrane is represented by sodium dodecyl sulfate (SDS) micelles or phosphatidylcholine (PC) vesicles and the cell wall precursors by cell wall peptide analogues with an N-terminal decanoyl chain designed to insert into the model membranes. This N-terminal membrane anchor is similar to the dagger; E-Mail dhw1@cam.ac.uk Dagger; All peptides used in this study are abbreviated in this style in order to indicate the nature of their a-acyl chain (acetyl = Ac decanoyl = dec docosanoyl = docos) the number of residues (di tri tetra penta or hexa based on the sequence Gly-L-Ala-D-g-Glu-Lys-D-Ala-D-Ala counting from the C-terminus) and the identity of the C-terminal residue (ndash;D-Alanine = Ala or ndash;D-Lactate = Lac).C11 chain present in the antibiotic teicoplanin which we have previously shown to associate with model membranes.9 The whole arrangement of micelle/vesicle and anchored cell wall analogue is therefore similar to that at the bacterial cell surface where cell wall precursors are anchored to the cell membrane via a C55 hydrocarbon chain.10 Using such a system we have demonstrated that for a dimerndash; ligand complex dimerisation and membrane anchoring result in enhanced binding affinity.11 This was achieved by comparing the binding constants of ristocetin A (another member of the vancomycin group of antibiotics) to the cell wall analogues Na- decanoyl-D-Ala-D-Ala (dec-di-Ala) and N-a-acetyl-D-Ala-DFig.1 Exploded view of the complex formed between chloroeremomycin (CE) or biphenylchloroeremomycin (BCE) (the two members of the vancomycin group of antibiotics studied in this paper) and di-Nacetyl- Lys-D-Ala-D-Ala. Protons referred to in the text are labelled. Aromatic rings are numbered according to residue number. N O N H N O H O N H O OH H O H NH2Me H H H O O N O H H H H O N O ndash; O2C H OH OH HO HO CH3 O H3C O OH O HO HO O ndash; O H N N H3C H O CH3 H H O NH2 O H3N HO CH3 O H3C RH2N H Cl N H3C O HN H CH3 O H Cl H 7 5 4 2 + + 7d 7f w2 b d 6 e e-Ac BCE R = CE R = H + g 2912 J.Chem. Soc. Perkin Trans. 1 1997 Ala (Ac-di-Ala) in the presence of SDS micelles. The complex formed between the antibiotic and Ac-di-Ala results from an intermolecular association Fig. 2(a) whereas the complex formed on the surface of a micelle with dec-di-Ala is essentially intramolecular Fig. 2(b). However inspection of Coreyndash; Paulingndash;Kulton (CPK) models reveals that this peptide might be rather short and therefore limited in its ability to anchor to the model cell membrane when bound to the dimeric antibiotic. We have therefore decided to investigate the behaviour of a number of longer anchored peptides (ranging from two to six residues) in their complexes with antibiotic dimers in the model system predicting increased anchoring from the longer peptides which should act as better models of the natural peptides.The sequence followed for the peptides was that of the natural cell wall precursor namely ndash;L-Ala-D-g-Glu-Lys-D-Ala-D-Ala with an N-terminal glycine added as a spacer in the hexapeptide ligand (Fig. 3). In this paper we have therefore attempted to investigate the effect of the length of the peptide portions of ndash;D-Alaterminating cell wall precursor analogues on the enhancement of binding observed with glycopeptides at model cell surfaces. The influence that peptide length has on the geometry of binding to the antibiotic is also considered. Results and discussion Binding on a micelle surface Initially the binding of the ligands was assayed by a simple NMR method which involved measurement of the limiting chemical shift of the amide residue 2-proton (w2) of the antibiotic (labelled in Fig.1) under conditions where the antibiotic was 95 bound (dw2 lim). In the complex between antibiotic and ligand w2 is involved in a crucial hydrogen bond to the ligand carboxylate such that its chemical shift moves dramatically downfield upon binding. We have previously observed a strong correlation between dw2 lim and the free energy of ligand binding.11ndash;14 It is therefore possible to use dw2 lim as a qualitative measure of binding affinity. The glycopeptide antibiotic chloroeremomycin (CE LY264826 Fig. 1) was chosen for these studies as it binds ligand strongly and has a high dimerisation constant.6 It was therefore anticipated that it might exhibit a high degree of cooperativity when binding at a surface.The results obtained for dw2 lim with this antibiotic upon binding to the decanoyl ligands in the presence of SDS micelles are shown in Table 1. It should be noted that additional N-terminal residues beyond lysine add a negligible amount to the intrinsic binding energy but removal of the lysine does significantly reduce this intrinsic binding energy.15 The dw2 lim for CE bound to Ac-tri-Ala in the presence of SDS micelles was observed to be 11.11 ppm. The value when bound Fig. 2 Schematic illustration of a glycopeptide dimer binding cell wall analogues at the surface of a micelle. (a) The complex formed with Acdi- Ala is intermolecular whereas (b) that with dec-di-Ala is essentially intramolecular. to Ac-di-Ala in the presence of SDS was not observed due to exchange-broadening of the spectrum but is expected to be lower than 11.11 ppm as Ac-di-Ala binds more weakly to CE than Ac-tri-Ala.However for CE bound to dec-di-Ala in the presence of SDS dw2 lim is 11.50 ppmmdash;much further downfield than dw2 lim when bound even to the longer peptide Ac-tri-Ala. This significant increase in dw2 lim when bound to the decanoyl ligand is consistent with observations made previously with the antibiotic ristocetin A i.e. an enhancement in binding affinity for the intramolecular system. However the similarity of dw2 lim for CE bound to N-a-decanoyl-N-e-acetyl-Lys-D-Ala-D-Ala (dec-tri-Ala) and N-a-decanoyl-D-g-Glu-N-e-acetyl-Lys-D-Ala- D-Ala (dec-tetra-Ala) in the presence of SDS to that when bound to Ac-tri-Ala suggested a negligible enhancement of binding in these systems (11.13 and 11.15 ppm for dec-tri-Ala and dec-tetra-Ala respectively vs.11.11 ppm for Ac-tri-Ala). The shifts when bound to N-a-decanoyl-L-Ala-D-g-Glu-N-eacetyl- Lys-D-Ala-D-Ala (dec-penta-Ala) and N-a-decanoyl- Gly-L-Ala-D-g-Glu-N-e-acetyl-Lys-D-Ala-D-Ala (dec-hexa- Ala) (11.18 and 11.25 ppm respectively) suggest that as the peptides become longer there is some enhancement to the binding of the anchored ligand relative to the binding of the unanchored Ac-tri-Ala but substantially less enhancement than for binding to dec-di-Ala. The physical reasons for this enhancement to the binding of the anchored dipeptide ligand but not to the binding of anchored tripeptide are discussed later in this paper. Binding on a vesicle surface To further investigate these complexes we used a model system in which the SDS micelles were replaced by PC vesicles which are of a closer size surface charge and curvature to the natural cell membrane.These features make them a better mimic of the bacterial membrane surface which is composed of similar lipids. However they are unsuitable for the w2 binding assay described above since their large size results in a very long correlation time and consequently very broad NMR resonances. An antibiotic associated with the vesicle surface via binding to anchored ligands thus takes on the tumbling properties of the vesicle itself with the effect that its resonances also appear as broad lines. Whilst this renders any high-resolution structural analysis almost impossible we used this broadening to determine whether a complex was indeed anchored to the surface of the vesicle.The results of these experiments are shown in Fig. 4. This shows the aromatic region of the 1H NMR spectra of the complexes of CE with various length peptides in the presence of PC vesicles. For the complexes of CE with dec-di-Ala and dechexa- Ala no signals were observed. This was interpreted as being due to tight binding of CE to the peptides on the surface of the vesicle resulting in fully broadened NMR resonances i.e. all of the antibiotic was associated via the bound peptide to the surface of the vesicle. In these cases the broadening was probably exacerbated by some aggregation of the vesicles; the sample became visibly turbid. The complex with dec-penta-Ala also shows some degree of broadening but those with dec-tri- Ala and dec-tetra-Ala both result in well-resolved spectra.In these three cases (dec-tri- -tetra- and -penta-Ala) the ligand Table 1 Limiting chemical shifts of the residue 2 amide proton (w2) dw2 lim of CE with various decanoyl ligands in the presence of SDS at pH 4.5 Ligand dec-di-Ala dec-tri-Ala dec-tetra-Ala dec-penta-Ala dec-hexa-Ala dw2 lim 11.50 11.13 11.15 11.18 11.25 J. Chem. Soc. Perkin Trans. 1 1997 2913 Fig. 3 Structures of the Ala-terminating decanoyl peptides used in these studies CO2H HN NH HN NH O O O HN O CO2H HN NH O O CO2H HN NH HN O O O HN O O CO2H CO2H HN NH HN NH HN O O O HN O O CO2H O CO2H HN NH HN NH HN NH O O O HN O O CO2H O O N-a-dec-D-Ala-D-Ala (dec-di-Ala) N-a-dec-N-e-Ac-Lys-D-Ala-D-Ala (dec-tri-Ala) N-a-dec-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-tetra-Ala) N-a-dec-L-Ala-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-penta-Ala) N-a-dec-Gly-L-Ala-D-g-Glu-N-e-Ac-Lys-D-Ala-D-Ala (dec-hexa-Ala) was confirmed to be bound to the antibiotic by the downfield shift of the proton w2 indicative of ligand binding.However the lack of broadening of the resonances for the complexes with dec-tri-Ala and dec-tetra-Ala suggested that even though the ligand was bound to the antibiotic it was not simultaneously associated with the vesicles. In the case of dec-penta-Ala the partial broadening observed indicates that the antibioticndash;ligand complexes are in slow exchange on the NMR timescale between being vesicle-associated and free in solution. Competition experiments We have also used a competition strategy illustrated schematically in Fig. 5 to estimate a binding constant of dec-hexa-Ala to CE on the surface of vesicles.In these experiments a nonanchored cell wall analogue Ac-tri-Ala (1.0 mM) was added to a solution of CE (0.5 mM) bound to dec-hexa-Ala (1.0 mM) on the surface of vesicles. The results of this experiment and a control experiment in the absence of dec-hexa-Ala showed (Fig. 6) that even in the presence of an excess of the antagonist most of the antibiotic appears to remain in an anchored complex on the surface of the vesicle. Based on the relative peak integrals in the presence and absence of dec-hexa-Ala Figs. 6(d ) and (b) and concentrations of added Ac-tri-Ala in these two solutions a binding enhancement of approximately one hundred-fold is estimated for the anchored hexapeptide assembly over the non-anchored Ac-tri-Ala/CE complex.This is similar to the four hundred-fold binding enhancement due to templating measured for ristocetin A binding to dec-di-Ala in the presence of SDS micelles.11 Therefore the above observations can be rationalised by the hypothesis that dec-tri-Ala and dec-tetra-Ala when bound to the antibiotic are simply not long enough for the alkyl chain to insert into the model membrane Fig. 7(a) whereas dec-penta- Ala and more effectively dec-hexa-Ala are of sufficient length Fig. 7(b). If this is the case the seemingly anomalous cooperativity exhibited by the dec-di-Ala/CE complex when anchoring to micelles or vesicles might arise as a result of some alternative binding geometry. It was anticipated that when binding in vivo the antibiotics would normally interact with cell wall peptides at the bacterial 2914 J.Chem. Soc. Perkin Trans. 1 1997 Fig. 4 Aromatic region of the 500 MHz 1H NMR spectra (D2O pD 6.2) of CE (0.5 mM) with the various decanoyl peptides (1 mM) in the presence of vesicles (10 mM). In the case of dec-di-Ala and dec-hexa-Ala the signals from the complex are broadened due to anchoring to the vesicle. cell surface with their ring-4 sugars the disaccharide glucose:4- epi-vancosamine in CE (Fig. 1) in contact or near to the cell surface. This supposition was based on the fact that teicoplanin 9,16 and the semi-synthetic glycopeptide biphenylchloroeremomycin (BCE LY307599 Fig. 1) 17 have putative membrane anchors attached to their ring-4 sugars. Additionally the Fig. 5 Schematic illustration of the competition experiment in which antibiotic dimer bound to anchored ligands is displaced from the surface of a vesicle Fig.6 Aromatic region of the 500 MHz 1H NMR spectra (D2O pD 6.2) of (a) CE (0.5 mM) in the presence of vesicles (10 mM) and (b) following the addition of Ac-tri-Ala (1 mM). (c) The same region of the spectrum of CE (0.5 mM) in the presence of vesicles (10 mM) and dechexa- Ala (1 mM) and (d) following the addition of Ac-tri-Ala (4 mM) (* these resonances result from residual amide protons of the added ligand). natural precursor peptide analogue penta-Ala has been shown to bind CE with its N-terminal L-alanine residue adjacent to the ring-4 sugars.18 Inspection of CPK models suggested that if the decanoyl chain of dec-di-Ala were to pass over ring-7 rather than adjacent to ring-6 (which is the orientation necessary if the ring-4 sugars are to interact with the micelle surface) there would be more of the hydrocarbon chain available to insert into the micelle.We therefore hypothesised that in binding to the anchored dec-di-Ala ligand CE might be oriented lsquo;upside downrsquo; i.e. with the sugars projecting away from the micelle Fig. 7(c) thus allowing the decanoyl chain to pass over ring-7 of the antibiotic and insert into the membrane. Such a conformation is not available to longer peptides because the lysine side chain occupies this position over ring-7.5,18ndash;21 With the lysine side-chain positioned over ring-7 the L-stereochemistry of the lysine forces the additional residues of the longer peptides and indeed the alkyl chain into an orientation in which they are directed toward the ring-4 sugars.In the case of binding by dec-tri-Ala and dec-tetra-Ala the alkyl chain is then not long enough to be able to insert into the membrane Fig. 7(a); it is only for dec-penta-Ala and dec-hexa-Ala that this is possible Fig. 7(b) as observed experimentally. Fig. 7 Schematic representation of the hypothesis which explains lack of templating for dec-tri-Ala and dec-tetra-Ala in the presence of vesicles. (a) Dec-tri-Ala is not long enough to reach the surface of the model membrane when bound to antibiotic whereas (b) dec-hexa-Ala is able to both anchor and bind antibiotic. (c) Dec-di-Ala can both anchor to the model membrane and bind antibiotic but only when oriented lsquo;upside downrsquo; (a geometry not accessible to longer peptides because of the presence of the lysine side chain).J. Chem. Soc. Perkin Trans. 1 1997 2915 To test the validity of this hypothesis experiments were performed with two new ligands N-a-decanoyl-Gly-D-Ala-D-Ala and N-a-docosanoyl-N-e-acetyl-Lys-D-Ala-D-Ala (docos-tri- Ala docosanoyl = C22) binding to CE in the presence of vesicles. It was anticipated that for the former ligand the absence of the lysine side chain would allow it to take up a conformation in the binding pocket similar to that of dec-di- Ala with the decanoyl chain passing over ring-7 so that binding to the antibiotic on the vesicle surface would be facilitated. For docos-tri-Ala it was predicted that the binding geometry would be the same as for dec-tri-Ala but that the much longer hydrocarbon chain would now be able to reach the membrane again allowing binding of the antibiotic to the vesicle surface.The results were exactly as anticipated with the complexes of CE bound to each ligand in the presence of vesicles resulting in substantially broadened antibiotic 1H NMR signals. Also the addition of Ac-tri-Ala did not result in the antibiotic becoming displaced from its complexes with the anchored ligands on the surface of vesicles (Fig. 8 shows the aromatic region of the spectra obtained with docos-tri-Ala). One possible alternative explanation for the difference in binding to dec-di-Ala dec-penta-Ala and dec-hexa-Ala compared to dec-tri-Ala and dec-tetra-Ala could lie in the relative abilities of the peptides to form self-micelles or self-vesicles. This self-association could thus be responsible for the observed 1H NMR line broadening and enhancement of binding of the first three named peptides to CE instead of the templated binding with antibiotics on PC vesicle surfaces as described above.However 1H NMR spectra of these ligands in solution with vesicles at the same concentration to that used in the binding experiments show that their signals all broaden to a similar degree (although predictably the C-terminal signals of the longer peptides are sharper than their N-terminal signals) indicating that all of the ligands associate with the PC vesicles to a similar extent. Additionally given the relatively low concentration of ligands used in these experiments (typically twice that of antibiotic) aggregates composed entirely of such ligands would possess only sufficient surface area to bind a small fraction of the antibiotic present and therefore could not lead to the complete broadening of signals observed particularly for dec-di-Ala and dec-hexa-Ala.Confirmation of ligand orientation To further support the hypothesis outlined in the previous section two-dimensional NOESY spectra were acquired in order to identify the orientation of the decanoylated peptides with respect to the antibiotic in the presence of SDS. For the complex of CE with dec-di-Ala where the decanoyl chain was anticipated to pass over ring-7 in the lsquo;upside downrsquo; arrange- Fig. 8 Aromatic region of the 1H NMR spectra (D2O pD 6.2) of (a) CE (0.5 mM) in the presence of vesicles (10 mM) (b) following the addition of docos-tri-Ala (1 mM) cf. Fig. 6(b) and (c) after the addition of Ac-tri-Ala (2 mM) ment NOESY cross peaks were observed between the aromatic protons of ring-7 of the antibiotic and the methylene protons of the decanoyl chain Fig.9(a). From this data we conclude only that the decanoyl chain of dec-di-Ala must lie over ring-7. A more precise interpretation of the data does not seem to be warranted because of (i) the anticipated dynamic behaviour of this portion of the ligand and (ii) spin diffusion. Thus the decanoyl chain projects from the antibiotic toward the micelle (or vesicle) in an orientation which places the ring-4 sugars away from the surface of the model membrane. For the complex with dec-tri-Ala NOESY cross peaks were observed from the lysine side chain to ring-7 Fig. 9(b) and the methylene groups of the decanoyl chain gave cross peaks to residue 6 of the antibiotic. This positions the decanoyl chain such that it could anchor the complex to a membrane in the case of either the longer peptides (dec-penta-Ala and dec-hexa-Ala) or the tripeptide with a longer acyl chain (docos-tri-Ala) while simultaneously positioning the ring-4 sugars adjacent to the membrane surface.Thus the orientation of the ligands that was predicted by the results of the dw2 lim measurements and vesicle binding experiments was shown to exist by two-dimensional NMR spectroscopy. Three-dimensional representations of the complexes formed between CE and dec-di-Ala and dec-tri-Ala are illustrated in Fig. 10(a) and 10(b) respectively. Conclusions We have employed two model membrane systems in an attempt to establish the optimum conditions for the expression of cooperativity due to binding of ligands to glycopeptides on a surface.Each system has its advantages and disadvantages. SDS is available in deuterated form and the micelles it forms are small in diameter. These features enable complexes formed on the surface of micelles to be studied by high-resolution 1H NMR spectroscopy providing a wealth of structural information. However the small diameter of the SDS micelles (approximately 25ndash;30 Aring;) results in a high degree of curvature at the surface leading perhaps to a non-ideal binding geometry with certain ligands. PC vesicles are much larger in diameter (1000ndash; 10 000 Aring;) and are thus expected to overcome any problems related to surface curvature. The size of PC vesicles precluded a detailed study of vesicle-bound complexes by NMR spectroscopy but allowed a qualitative determination of the extent to which templated binding was achieved for CE binding to a series of decanoylated ligands.We have thus been able to demonstrate in a direct manner the importance of binding geometry and dimerisation in the mode of action of these antibiotics. We believe that the origin of this enhanced binding lies in the chelate-like enhancement conveyed Fig. 9 Portions of the NOESY spectra of the complexes formed in the presence of micelles between CE and (a) dec-di-Ala and (b) dec-tri- Ala illustrating the cross peaks from ring-7 of the antibiotic to ligand protons. In (a) lsquo;dec CH2rsquo; is used to signify those methylene groups of the decanoyl chain which have unresolved chemical shifts. 2916 J. Chem. Soc. Perkin Trans. 1 1997 by the ability of these antibiotics to dimerise at the cell surface resulting in a tightening of all interactions within the complex thus giving rise to enthalpic as well as entropic gains.22 The greatest degree of templated binding was achieved with the longest and also paradoxically the shortest of the anchored ligands studied.We have shown semi-quantitatively that the enhancement to binding to CE due to templating for dec-hexa- Ala is similar to that for dec-di-Ala measured previously.11 We have put forward a physical model with accompanying evidence which accounts for this anomaly. These results provide a rationale as to why membrane anchors on naturally occurring glycopeptide antibiotics are located on the ring-4 sugars,16 and why membrane anchors on the most active semi-synthetic antibiotics are similarly located.17 The L-stereochemistry of the lysine residue present in Fig.10 Three dimensional representations of half of the dimeric complexes formed between CE and (a) dec-di-Ala and (b) dec-tri-Ala. Note how the orientation of the decanoyl chain attached to the dipeptide results in more exposure of hydrocarbon than the chain attached to the tripeptide. cell wall peptides directs the antibiotic to bind in a fashion that places the residue-4 sugars in close proximity to the bacterial cell membrane. Nature takes advantage of this by placing a locating device (in the case of teicoplanin a C11 acyl chain) at precisely this point. Despite this rationale we do not preclude the possibility of enhancement of antibiotic action through the location of hydrophobic chains at alternative sites. The model systems studied in this work thus present a more detailed picture of how the vancomycin group antibiotics function in biological systems.The results support the hypothesis that these antibiotics bind to nascent bacterial cell walls with their ring-4 saccharides adjacent to the cell membrane and the parallel nature of these saccharides in the antibiotic dimers may reflect a similar parallel arrangement of the peptidoglycan strands of growing cell wall. Experimental sect; Preparation of phosphatidylcholine vesicles Type XV1-E L-a-phosphatidylcholine from fresh egg yolk (Sigma 80 mg) was dissolved in chloroform (2 ml) which had been rendered ethanol-free by passage through a column of activated alumina. The solution was then evaporated under reduced pressure to yield a thin film on the wall of the flask.The flask was flushed with nitrogen followed by addition of D2O (5 ml) or 50 mM NaH2PO4 pH 6.2 buffer (5 ml). The mixture was shaken for 20 min then sonicated for 90 min to yield a slightly turbid suspension of vesicles (20 mM phosphatidylcholine). 1H NMR spectroscopy Sodium 2H25dodecyl sulfate (SDS; 98 atomD) was purchased from Euriso-top. All 1H NMR spectroscopy experiments were performed on 500 MHz Bruker DRX-500 and AM500 spectrometers at 300 K. Suppression of the solvent resonance was achieved using WATERGATE23 or pre-saturation. One-dimensional spectra were recorded using 32k complex data points. In two-dimensional experiments 4k complex points were acquired in f2 with 512 increments in f1. TPPI was used to achieve quadrature detection in the indirect dimension.Data was processed with XWIN-NMR software using a sinesquared window function and zero-filling in f1 up to 1k or 2k points. Two-dimensional NOESY experiments employed mixing times ranging through 50ndash;150 ms and were used to confirm all w2 assignments. In experiments involving micelles or vesicles the ligand was added to the vesicle/micelle solution and the mixture was sonicated to facilitate insertion. Experiments involving SDS employed a concentration of 70 mM SDS (above the SDS critical micelle concentration) 5 mM antibiotic and 10 or 20 mM ligand. These concentrations ensured that a high (90) proportion of antibiotic was bound by ligand but also that the SDS was not lsquo;overloadedrsquo; with ligand. Only when dissolution was complete was the antibiotic added. In the vesicle experiments vesicles were prepared as described above and used as a 10 mM solution; concentrations of antibiotics used were as described in the individual figure legends.In the competition experiments the unanchored ligand was added to the NMR tubes as a concentrated solution (50 mM) so as not to change the concentration of the contents of the tube signifi- cantly and to allow accurate concentrations to be achieved on the addition of the appropriate volume. Acknowledgements Eli Lilly and Co. are thanked for generously providing samples of CE (LY264826) and BCE (LY307599). The EPSRC (G. J. S.) sect; Details of the synthesis of the peptides are available as supplementary material (SUP 57275; 33 pp.) deposited with the British Library. Details are available from the editorial office. J.Chem. Soc. Perkin Trans. 1 1997 2917 Xenova (A. C. T.) The Wellcome Trust (R. J. D.) the EPSRC and Roussel (B. B.) and the BBSRC and Roussel (R. M. H. E.) are thanked for financial support. We also thank the Biomedical NMR Centre NIMR Mill Hill London for access to NMR equipment. References 1 A. N. Chatterjee and H. R. Perkins Biochem. Biophys. Res. Commun. 1966 24 489. 2 D. C. Jordan and P. E. Reynolds in Antibiotics ed. J. W. Corcoran and F. E. Hahn Springer-Verlag Berlin 1974 vol. III. 3 J. R. Kalman and D. H. Williams J. Am. Chem. Soc. 1980 102 906. 4 D. H. Williams M. P. Williamson D. W. Butcher and S. J. Hammond J. Am. Chem. Soc. 1983 105 1332. 5 J. C. J. Barna D. H. Williams and M. P. Williamson J. Chem. Soc. Chem. Commun. 1985 254. 6 J. P. Mackay U. Gerhard D. A. Beauregard R.A. Maplestone and D. H. Williams J. Am. Chem. Soc. 1994 116 4573. 7 J. P. Mackay U. Gerhard D. A. Beauregard M. S. Westwell M. S. Searle and D. H. Williams J. Am. Chem. Soc. 1994 116 4581. 8 D. A. Beauregard D. H. Williams M. N. Gwynn and D. J. C. Knowles Antimicrob. Agents Chemother. 1995 39 781. 9 M. S. Westwell U. Gerhard and D. H. Williams J. Antibiot. 1995 48 1292. 10 J. M. Ghuysen in Topics in Antibiotic Chemistry ed. P. G. Sammes Ellis Horwood Chichester 1980 vol. 5 p. 31. 11 M. S. Westwell B. Bardsley R. J. Dancer A. C. Try and D. H. Williams Chem. Commun. 1996 589. 12 P. Groves M. S. Searle M. S. Westwell and D. H. Williams J. Chem. Soc. Chem. Commun. 1994 1519. 13 G. J. Sharman M. S. Searle B. Benhamu P. Groves and D. H. Williams Angew. Chem. Int. Ed. Engl. 1995 34 1483.14 M. S. Searle G. J. Sharman P. Groves B. Benhamu D. A. Beauregard M. S. Westwell R. J. Dancer A. J. Maguire A. C. Try and D. H. Williams J. Chem. Soc. Perkin Trans. 1 1996 2781. 15 M. Nieto and H. R. Perkins Biochem. J. 1971 123 780. 16 J. C. J. Barna D. H. Williams D. J. M. Stone T.-W. C. Leung and D. M. Doddrell J. Am. Chem. Soc. 1984 106 4895. 17 R. D. G. Cooper N. J. Snyder M. J. Zweifel. M. A. Staszak S. C. Wilkie T. I. Nicas D. L. Mullen T. F. Butler M. J. Rodriguez B. E. Huff and R. C. Thompson J. Antibiot. 1996 49 575. 18 W. G. Prowse A. D. Kline M. A. Skelton and R. J. Loncharich Biochemistry 1995 34 9632. 19 S. W. Fesik T. J. Orsquo;Donnell R. T. Gampe and E. T. Olejniczak J. Am. Chem. Soc. 1986 106 3165. 20 P. Groves M. S. Searle J. P. Mackay and D. H. Williams Structure 1994 2 747. 21 P. Groves M. S. Searle J. P. Waltho and D. H. Williams J. Am. Chem. Soc. 1995 117 7958. 22 M. S. Searle M. S. Westwell and D. H. Williams J. Chem. Soc. Perkin Trans. 2 1995 141. 23 M. Piotto V. Saudek and V. Sklenaacute;r J. Biomol. NMR 1992 2 661. Paper 7/01880K Received 18th March 1997 Accepted 19th June 1997