Bioorganic & Medicinal Chemistry 21 (2013) 5806–5810

Contents lists available at ScienceDirect

Bioorganic & Medicinal Chemistry

journal homepage: www.elsevier .com/locate /bmc

Structural studies of b-hairpin peptidomimetic antibiotics that targetLptD in Pseudomonas sp.

0968-0896/$ - see front matter � 2013 Published by Elsevier Ltd.http://dx.doi.org/10.1016/j.bmc.2013.07.013

⇑ Corresponding author. Tel.: +41 44 6354242.E-mail address: robinson@oci.uzh.ch (J.A. Robinson).

Jasmin Schmidt a, Krystyna Patora-Komisarska a, Kerstin Moehle a, Daniel Obrecht b, John A. Robinson a,⇑a Chemistry Department, University of Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerlandb Polyphor AG, Hegenheimermattweg 124, 5011 Allschwil, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history:Received 2 May 2013Revised 5 July 2013Accepted 8 July 2013Available online 17 July 2013

Keywords:AntibacterialConformationPeptidePeptidomimeticPseudomonas aeruginosaLipopolysaccharide transportb-Barrel membrane protein

We report structural studies in aqueous solution on backbone cyclic peptides that possess potent antimi-crobial activity specifically against Pseudomonas sp. The peptides target the b-barrel outer membraneprotein LptD, which plays an essential role in lipopolysaccharide transport to the outer membrane. Thepeptide L27-11 contains a 12-residue loop (T1W2L3K4K5R6R7W8K9K10A11K12) linked to a DPro–LPro tem-plate. Two related peptides were also studied, one with various Lys to ornithine or diaminobutyric acidsubstitutions as well as a DLys6 (called LB-01), and another containing the same loop sequence, butlinked to an LPro–DPro template (called LB-02). NMR studies and MD simulations show that L27-11and LB-01 adopt b-hairpin structures in solution. In contrast, LB-02 is more flexible and importantly,adopts a wide variety of different backbone conformations, but not b-hairpin conformations. L27-11and LB-01 show antimicrobial activity in the nanomolar range against Pseudomonas aeruginosa, whereasLB-02 is essentially inactive. Thus the b-hairpin structure of the peptide is important for antimicrobialactivity. An alanine scan of L27-11 revealed that tryptophan side chains (W2/W8) displayed on oppositefaces of the b-hairpin represent key groups contributing to antimicrobial activity.

� 2013 Published by Elsevier Ltd.

1. Introduction

The cyclic peptidomimetic L27-11 (Fig. 1A) shows antimicro-bial activity in the nanomolar range against Gram-negative Pseu-domonas sp.1 The amino acid sequence of L27-11 is unrelated tothat of any known naturally occurring cationic antimicrobialpeptide, although it does contain a mix of hydrophobic (includ-ing aromatic) and cationic residues.2,3 Related molecules withoptimized drug-like properties include POL7001, which has amuch-improved stability towards proteolysis in human plasmadue to the replacement of multiple Lys/Arg residues by diamin-obutyric acid (Dab), although it retains the same potent andselective antimicrobial activity for Pseudomonas sp.1 A relatedpeptide is now in clinical development as a potential new anti-biotic targeting the important human pathogen Pseudomonasaeruginosa.4

These peptidomimetic antibiotics have a novel mechanism ofaction, involving an interaction with the outer membrane (OM)protein LptD.1,5 The function of LptD in Gram-negative bacteriahas been studied most thoroughly in Escherichia coli, where to-gether with the lipoprotein LptE, it plays an essential role in lipo-

polysaccharide (LPS) transport to the cell surface.6–10 Sevenessential Lpt (lipopolysaccharide transport) proteins (LptA-G) areknown to mediate LPS transport from the inner membrane (IM)to the outer leaflet of the asymmetric OM (Fig. 1B).11,12 LptD is a�100 kD OM protein, comprising a C-terminal b-barrel domainembedded within the OM and an N-terminal domain located inthe periplasm (Fig. 1B). The lumen of the LptD b-barrel domain isoccupied at least partly by the lipoprotein LptE.13 The LptD/E com-plex accepts LPS molecules from LptA in the periplasm and trans-locates them by an unknown mechanism into the outer leaflet. Theantibiotic L27-11 binds to LptD and inhibits transport of LPS to thecell surface.1,5

We report here NMR studies of L27-11, which show that thepeptide adopts a folded b-hairpin conformation in solution(Fig. 1A). This structural information is important for a more de-tailed understanding of how the peptide interacts and binds toLptD. L27-11 contains a 12-residue loop (TWLKKRRWKKAK) cou-pled to a DPro–LPro template.14 The structures of two additionalcyclic peptides have been studied (LB-01 and LB-02, Fig. 1A), withaltered conformational properties in aqueous solution, which to-gether provide evidence that the b-hairpin conformation is impor-tant for antimicrobial activity. Finally, an alanine scan of L27-11 isreported, which highlight side chains that are important for anti-microbial activity.

Figure 1. (A) Structures of the cyclic peptidomimetic antibiotics. Orn = L-ornithine, Dab = L-diaminobutyric acid, DK = DLys. (B) The peptidomimetic antibiotics target LptD, anOM protein essential for LPS transport to the cell surface (see text). LPS = lipopolysaccharide, PE/PG = phosphatidyl-ethanolamine/glycerol, IM = inner membrane, OM = outermembrane.

Figure 2. Coupling constants (3JHNa), temperature coefficients (Dd/DT) of amide NH resonances, and relative peptide NH exchange rates (open circle = fast, halffilled = slower, filled = slowest) for residues 1–12 in each peptide.

J. Schmidt et al. / Bioorg. Med. Chem. 21 (2013) 5806–5810 5807

2. Results

2.1. Solution structure of L27-11

In water at pH 5.0, L27-11 exists as a mixture of two slowlyinterconverting rotational isomers (rotamers) in 1:6 ratio arisingfrom cis to trans isomerization at the Lys12–DPro13 peptide bond.The two rotamers show characteristic 1H–1H NOEs between Lys12

Ha and DPro13 Ha (cis), or Lys12 Ha and DPro13 Hd (trans), withthe major form being trans. In a d3-TFE/water (1/1 v/v) solutiononly the trans form was observed.

A complete assignment of 1H and 13C NMR spectra was onlypossible for the predominant trans rotamer of L27-11. The assign-ments and an analysis of chemical shift deviations (CSDs) from sta-

tistical random coil values ðDd ¼ dobserved � drandomÞ are given in theSupplementary data (SI). The measured 3JHNa coupling constants(Fig. 2) all lie in the range 6–8 Hz, likely due to fluctuations inthe backbone torsion angle /. Amide proton (HN) resonance tem-perature coefficients (Dd/DT)15,16 were determined, and valuessmaller than �5 ppb/K were observed for Thr1, Leu3, and Trp8, aswell as Arg6 in the turn region (Fig. 2). 1H NMR spectra measuredin D2O revealed rapid exchange (within minutes at 300 K) of allpeptide amide protons.

Analysis of NOESY plots for L27-11 revealed medium and longrange NOEs characteristic of b-hairpin conformations, mostly forresidues close to the DPro–LPro template (Fig. 3). However, reso-nance overlap for backbone and side-chain protons in lysine resi-dues 4, 5 and 10 hindered the identification of specific NOE

Figure 3. Long range NOEs and average solution structures. Dotted arrows indicate long range NOEs observed for L27-11, LB-01 and LB-02 (red arrow on L27-11 indicatesNOE that cannot be unambiguously assigned due to signal overlap (see text); red arrow on LB-01 indicates an NOE that should be observable, but was not detected due toinsufficient separation of resonances); the resulting bundle of 20 average solution structures for each peptide; and one representative structure for LB-01 (template in lightbrown/blue).

Table 1Summary of statistics for the final 20 NMR structures calculated using DYANA forL27-11, LB-01 and LB-02

L27-11 LB-01 LB-02

Number of NOE upper-distancelimitsIntraresidueSequentialMedium- and long-range

74273314

181494448

8638399

Residual target function value (Å2) 0.57 ± 0.08 0.62 ± 0.05 0.48 ± 0.05Mean rmsd values (Å)All backbone atoms

All heavy atoms1.85 ± 0.513.65 ± 0.92

0.47 ± 0.211.27 ± 0.21

2.32 ± 0.634.44 ± 0.22

Residual NOE violationsNumber >0.2 Å

Maximum (Å)80.33

50.25

40.0

5808 J. Schmidt et al. / Bioorg. Med. Chem. 21 (2013) 5806–5810

contacts to these residues. Some medium and long range NOEscharacteristic of b-hairpin structures are clearly seen, includingHa-Ha NOEs between Trp2 and Ala11, and Lys4/5/10 and Lys9, whichwould be expected between cross-strand residue pairs whose pep-tide bonds do not hydrogen bond (NHB) across the b-hairpin. Inaddition, cross-strand side-chain NOEs are seen between Trp2-Ala11 and Trp2–Lys9. However, the absence of a more extensivenetwork of cross-strand NOEs, in particular, any cross-strandNH–NH NOEs, likely reflects partial melting of a regular hairpinstructure. The available NOEs were used to derive distance re-straints for structure calculations with DYANA (Table 1). Theresulting average NMR structures (Fig. 3) could be superimposedwith a backbone rmsd of 1.85 Å. However, nascent b-hairpin-likestructures are readily discernible in all 20 NMR structures, mostlywith distorted type-II b-turns in Arg6–Arg7 at the tip.

To characterize the dynamic properties of the molecule, onetypical NMR structure was used for a 100 ns MD simulation withtime-averaged distance restraints in explicit solvent water at300 K. An analysis of the resulting MD trajectory shows that thecyclic peptide remains predominantly in a b-hairpin-like geometry,but shows increasing //w-angle fluctuations upon proceedingalong both nascent b-strands from the template (DPro–LPro) tothe tip of the loop (Arg6–Arg7) (see SI). In summary, the structureand MD calculations suggest that L27-11 populates b-hairpin-likestructures, with residues attached to the template being moststrongly constrained into regular b-structure, but with increasingbackbone flexibility of the backbone hairpin geometry towardsthe tip of the loop.

2.2. Solution structure of LB-01

Cyclic peptide LB-01 is identical to L27-11 except selected Lysresidues have been substituted by Orn or Dab to facilitate full 1Hbackbone and side-chain resonance assignments, and DLys is intro-duced at position-6 at the tip to stabilize a regular b-hairpin con-formation (Fig. 1A). LB-01 shows the same potent antimicrobialactivity (MIC = 0.01 lg/ml) against Pseudomonas aeruginosa as the

lead L27-11. LB-01 appears as a single rotamer (>98%) in aqueoussolution, with all peptide bonds trans. A complete assignment ofthe 1H NMR spectrum is given in the SI.

Analysis of coupling constants reveals 3JHNa values P8.5 Hz forseveral residues (Fig. 2), consistent with extended backbone con-formations. Spectra measured in D2O showed several slowlyexchanging amide protons, mostly at predicted cross-strandhydrogen bonding (HB) positions (Fig. 2). The slower NH exchangeseen for Trp2 might be due to shielding from the solvent by the aro-matic side chain. Finally, the NH temperature coefficients of Thr1,Orn5, Orn9, Dab10 and Lys12 show significantly lowered values(Fig. 2), consistent with HB and/or shielding from solvent.

In 2D NOESY plots for LB-01, numerous long-range cross-strandbackbone and side chain NOEs are seen (Fig. 3). This network ofcross-strand NOEs is a clear indication of regular b-hairpin struc-tures. Average structures calculated using NOE-derived distancerestraints now converge to a narrow range of backbone torsion an-gles characteristic of a b-hairpin, with an rmsd of 0.47 Å (Fig. 3 andTable 1). In the tip region (DLys6–Arg7) a type II’ b-turn is found inall structures, consistent with observed Ha-HN and HN–HN NOEsbetween DLys6–Trp8 and Arg7–Trp8, respectively. One typical

Figure 4. UV–CD spectra recorded for L27-11, LB-01 and LB-02 in water pH 5.0 at298 K.

Table 2Structure–activity studies

Peptide/mutant MIC (lg/ml)

L27-11Thr1AlaTrp2AlaLeu3AlaLys4AlaLys5AlaArg6AlaArg7AlaTrp8AlaLys9AlaLys10AlaLys12AlaTrp2PheTrp2TyrTrp2LeuTrp8PheTrp8TyrTrp8Leu

0.010.01>160.10.050.10.030.06>160.060.060.030.060.061.00.10.30.1

The residues mutated in L27-11 are shown. The antimicro-bial activity against P. aeruginosa ATCC 27853 is the minimalinhibitory concentration (MIC) measured in Mueller–Hintonbroth containing 0.05% (w/v) bovine serum albumin. Theaccuracy of MIC determinations is typically ±100%.20

J. Schmidt et al. / Bioorg. Med. Chem. 21 (2013) 5806–5810 5809

NMR structure of LB-01, in a MD simulation in a solvent water boxwith time-averaged distance restraints, maintained //w backbonetorsion angles in the b-region with only modest fluctuations overthe entire simulation, indicating a stable b-hairpin structure (seeSI).

In summary, the results provide clear evidence that LB-01adopts a stable b-hairpin structure in solution, characterized by aconsistent network of cross-strand NOEs and cross-strand HBinteractions.

2.3. Solution structure of LB-02

In cyclic peptide LB-02 (Fig. 1A) the order of the heterochiralprolines in the template is changed to LPro13–DPro14, which wasexpected to cause a major change in the entire backbone confor-mation compared to LB-01. LB-02 shows no antimicrobial activityagainst Pseudomonas sp. (MIC P32 lg/ml). The 1H NMR spectrumnow reveals two rotamers in slow exchange due to cis–trans isom-erism (�1:2 ratio) at the LPro13–DPro14 peptide bond. The majorform is the trans rotamer, with a characteristic Ha–Hd NOE, andthe minor form is the cis rotamer, with a Ha–Ha NOE betweenPro13 and DPro14. Due to resonance overlap, unambiguous NMRassignments were only possible for the trans rotamer (see SI).

Except for Thr1, the majority of 3JHNa values (Fig. 2) lie in therange 6.1–7.7 Hz, consistent with fast conformational averaging.All amide NH protons exchanged rapidly with solvent in D2O,although several amide NH temperature coefficients (�Dd/DT)were smaller than 4.0, suggesting possible involvement in hydro-gen bonding or shielding from solvent. In 2D NOESY plots, mostof the observed NOEs are intra-residue or sequential. Only a fewweak medium range i, i + 2 NOEs were seen, and none of thelong-range cross-strand NOEs expected for regular b-hairpin struc-tures were observed, except for residues directly attached to theLPro13–DPro14 template (Fig. 3).

With the available distance restraints, structure calculationswere performed as for the previous examples, however, the result-ing average structures revealed no regular secondary structure.Upon superimposition of the 20 selected structures, meanbackbone rmsd values were large (2.3 Å) (Table 1 and Fig. 3). Char-acteristic NOEs (Thr1 HN–DPro14 Hd, Thr1 HN–Pro13 Ha and Thr1

Hb–Lys12 HN) around the Pro13–DPro14 template suggest the for-mation of a type II b-turn in the major trans isomer. However, thereis no evidence for b-hairpin structures from residues 2 to 11. Thisconclusion was supported by an MD simulation in water solventwith time-averaged restraints starting from one typical NMR struc-ture of LB-02 (see SI).

The results reveal that exchanging the sense of chirality of thePro residues in the template, from LB-01 to LB-02, has a profoundeffect on the backbone peptide conformation. LB-02 exhibits con-siderable backbone flexibility but does not adopt hairpin-likeconformations.

2.4. Circular Dichroism (CD)

The far-UV CD spectra of L27-11, LB-01 and LB-02 in aqueoussolution are shown in Figure 4. Both L27-11 and LB-01 show abroad negative peak at 215 nm, although the mean residue elliptic-ity is higher for LB-01. In contrast, the shape of the CD curve for LB-02 is quite different, showing that this peptide adopts backboneconformations that are different from those populated in L27-11and LB-01.

2.5. Structure–activity studies

To evaluate the importance of each side chain in L27-11 forantimicrobial activity, an alanine scan was performed, by replacing

each residue in the b-hairpin loop in turn by L-alanine. The synthe-sis and characterization of the library followed established meth-ods (see SI). The results (Table 2) show that most of thesubstitutions lead at most to a 3–10 fold drop in antimicrobialactivity. Notably, however, substitution of either Trp2 or Trp8 ledto complete loss of antimicrobial activity. Analysis of 2D NMRspectra of the Trp8Ala mutant suggests that the backbone confor-mation is not changed significantly compared to L27-11, however,for the Trp2Ala mutant fewer long range NOEs are apparent, sug-gesting a more flexible structure. To explore the importance ofthese side chains further, each Trp was substituted for either an-other aromatic residue (Phe or Tyr) or an aliphatic non-b-branchedresidue (Leu). The results show that Phe and Tyr are tolerated atposition-2 with about sixfold drop in activity, however, Phe andTyr are tolerated less well at position 8. Replacement of each Trpby Leu also led to a significant drop in activity. These studies high-light the importance of the Trp2 and Trp8 side chains, displayed onopposite sides of the b-hairpin structure, for potent antimicrobialactivity.

5810 J. Schmidt et al. / Bioorg. Med. Chem. 21 (2013) 5806–5810

3. Discussion

The results provide strong evidence that the cyclic peptidesL27-11 and LB-01 adopt b-hairpin structures in aqueous solution,whereas the related molecule LB-02 adopts quite different back-bone conformations. Structure calculations and MD simulationsshow that LB-02 is quite flexible and adopts a wide variety of back-bone //w angles but does not adopt b-hairpin conformations. TheUV–CD spectra of all three peptides (Fig. 4), underscores the simi-larity between L27-11 and LB-01, and the different structures pop-ulated by LB-02. The negative peak near 215 nm, with near to zerointensity at 200 nm, is close to that expected for a combination ofantiparallel b-sheets and type-II’ b-turns, whereas the weakerpeaks and maxima observed for LB-02 approximates more closelythat expected for a random coil peptide.17

L27-11 adopts a significant amount (ca. 17%) of the cis-rotamerat the Lys12–D-Pro13 peptide bond. However, only trans-peptidebonds are seen in LB-01, which has the same potent antimicrobialactivity as L27-11. The NMR data for L27-11 and LB-01 show thatboth molecules adopt b-hairpin structures in solution. However,the b-strands and tip of the hairpin loop in L27-11 appear moreflexible, characterized by larger fluctuations in backbone //w an-gles than with LB-01. This increased flexibility correlates with lessstable cross-strand hydrogen bonding between the peptide amidegroups of residue pairs at HB positions (i.e., residue pairs: 1 = 12,3 = 10 and 5 = 8) of the hairpin. LB-01 on the other hand, adoptsstable b-hairpin structures characterized by significant cross-strand hydrogen bonding between the residue pairs at HB posi-tions. This difference in structure and dynamics is due to the DLys6

residue at the tip of the hairpin, which adopts a hairpin-stabilizingtype-II’ b-turn. Moreover, the hairpin structure of L27-11 and LB-01 appear to be very important for antimicrobial activity, since acomplete loss of activity is seen with LB-02, which adopts differentbackbone conformations.

The b-hairpin structure in L27-11 might be especially well sui-ted to bind to the b-barrel domain in the target LptD. For example,the b-strands in the ligand might interact with edges of antiparallelb-sheets in the b-barrel. The results of the alanine scan reveal thecritical importance of both tryptophan side chains (Trp2 andTrp8), which are displayed on opposite faces of the hairpin struc-ture, and likely participate in an extended network of interactionsbetween the ligand and its target. For example, one side chainmight interact with the inner lumen of the b-barrel while the otherinteracts with the outside lipid environment of the OM. In addition,it is interesting that changing charged side chains (Arg/Lys) to Alaleads to only small losses of antimicrobial activity, possibly due tothe reduced overall positive charge. Clearer insights into the struc-ture–activity profile, however, will require a detailed structuralanalysis of the antibiotic-binding site in LptD.

4. Experimental section

Peptide synthesis: The synthesis of L27-11 was described ear-lier.1 The peptides LB-01 and LB-02 and the mutants listed in Ta-ble 2 were synthesized in the same way (see SI).

NMR spectroscopy and structure calculations: 1H NMR measure-ments were performed either in H2O/D2O (9:1), or pure D2O ateither pH 5.0 or 3.0, or in a mixture (1:1) of trifluoroethanol (TFE-d3) and H2O/D2O (9:1), pH 5.0. Spectra were acquired on BrukerAV-600 or AV-700 spectrometers at 300 K, and assigned usingstandard methods (see SI). Water suppression was performed by

presaturation. All spectra were referenced using internal TSPstandard. Distance restraints were obtained from NOESY and ROESYspectra with a mixing time of 250 ms. The structure calculationswere performed by restrained molecular dynamics in torsionangle space by applying the simulated annealing protocol imple-mented in the program DYANA.18 Starting from 100 randomizedconformations a bundle of 20 conformations is selected, which incurthe lowest DYANA target energy function (Table 1). MOLMOL19 wasused for structure analysis and visualization of the molecularmodels.

The NMR structures of L27-11 and LB-01 have been depositedin the PDB databank under the accession numbers 2M7I and2M7J, respectively.

5. CD spectroscopy

The CD spectra were measured using a Jasco J-175 spectrometerequipped with a temperature controller using a quartz cell of 1 cmpath length. The spectra were recorded at 283 K from 195 to260 nm, at 0.5 nm resolution with a scan rate of 5 nm/min and asensitivity of 50 mdeg. Three scans were accumulated and aver-aged for each spectrum. Raw data were manipulated by smoothingand subtraction of solvent spectra. CD values were expressed as themean residue molar ellipticity.

Acknowledgments

The authors thank the Swiss National Science Foundation andthe European Union Framework 7 project NABATIVI for supportingthis work.

Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.bmc.2013.07.013.

References and notes

1. Srinivas, N.; Jetter, P.; Ueberbacher, B. J.; Werneburg, M.; Zerbe, K.; Steinmann,J.; Van der Meijden, B.; Bernardini, F.; Lederer, A.; Dias, R. L. A.; Misson, P. E.;Henze, H.; Zumbrunn, J.; Gombert, F. O.; Obrecht, D.; Hunziker, P.; Schauer, S.;Ziegler, U.; Kach, A.; Eberl, L.; Riedel, K.; DeMarco, S. J.; Robinson, J. A. Science2010, 327, 1010.

2. Nguyen, L. T.; Haney, E. F.; Vogel, H. J. Trends Biotechnol. 2011, 29, 464.3. Yount, N. Y.; Yeaman, M. R. In Ann. Rev. Pharmacol. Toxicol.; Insel, P. A., Amara,

S. G., Blaschke, T. F., Eds.; Annual Reviews: Palo Alto, 2012; Vol. 52, p 337.4. Moir, D. T.; Opperman, T. J.; Butler, M. M.; Bowlin, T. L. Curr. Opin. Pharmacol.

2012, 12, 535.5. Werneburg, M.; Zerbe, K.; Juhas, M.; Bigler, L.; Stalder, U.; Kaech, A.; Ziegler, U.;

Obrecht, D.; Eberl, L.; Robinson, J. A. ChemBioChem 2012, 13, 1767.6. Bos, M. P.; Tefsen, B.; Geurtsen, J.; Tommassen, J. Proc. Natl. Acad. Sci. U.S.A.

2004, 101, 9417.7. Braun, M.; Silhavy, T. J. Mol. Microbiol. 2002, 45, 1289.8. Chng, S.-S.; Ruiz, N.; Chimalakonda, G.; Silhavy, T. J.; Kahne, D. Proc. Natl. Acad.

Sci. U.S.A. 2010, 107, 5363.9. Sperandeo, P.; Dehó, G.; Polissi, A. Biochim. Biophys. Acta 2009, 1791, 594.

10. Wu, T.; McCandlish, A. C.; Gronenberg, L. S.; Chng, S.-S.; Silhavy, T. J.; Kahne, D.Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 11754.

11. Bos, M. P.; Robert, V.; Tommassen, J. Annu. Rev. Microbiol. 2007, 61, 191.12. Ruiz, N.; Kahne, D.; Silhavy, T. J. Nat. Rev. Microbiol. 2009, 7, 677.13. Freinkman, E.; Chng, S. S.; Kahne, D. Proc. Nat. Acad. Sci. U.S.A. 2011, 108, 2486.14. Robinson, J. A. Acc. Chem. Res. 2008, 41, 1278.15. Baxter, N. J.; Williamson, M. P. J. Biomol. NMR 1997, 9, 359.16. Cierpicki, T.; Otlewski, J. J. Biomol. NMR 2001, 21, 249.17. Greenfield, N. J. Nat. Protoc. 2004, 1, 2876.18. Guntert, P.; Mumenthaler, C.; Wuthrich, K. J. Mol. Biol. 1997, 273, 283.19. Koradi, R.; Billeter, M.; Wuthrich, K. J. Mol. Graph. 1996, 14, 51.20. Andrews, J. M. J. Antimicrob. Chemother. 2001, 48, 5.