made possible by decoupling and deuterium exchange experiments as outlined

Transcription

1 A CONFORMATIONAL ANALYSIS OF GRAMICIDIN S-A BY NUCLEAR MAGNETIC RESONANCE* BY ARNOLD STERN, t WILLIAM A. GIBBONS, AND LYMAN C. CRAIG THE ROCKEFELLER UNIVERSITY Communicated July 8, 1968 Currently there is great interest in the precise conformation of complicated natural products in solution. Nuclear magnetic resonance (NMR) has great promise in this connection and has made possible a high degree of sophistication in the conformational analysis of the steroids. However, less progress has been made with peptides. In this paper, we wish to report our analysis of 100- and 220-Mc NMIR spectra of the cyclic antibiotic decapeptide gramicidin S-A in solution. Methods.-Proton magnetic resonance spectra were obtained on Varian Associates high-resolution spectrometers operating at 100 and 220 Mc. The internal reference was tetramethylsilane (TMS), which appears at 0 ppm in all spectra. Sample concentrations were 3-6% (gm/100 ml). The normal operating temperature was 32. Gramicidin S-A was isolated from a crude preparation by countercurrent distributions and lyophilized in glacial acetic acid. Deuterated dimethylsulfoxide (DMSO) and deuterated methanol (CD30D) were purchased from NMR specialties. DMSO-d6 was dried when necessary over calcium hydride. Results and Discussion. A twofold axis of molecular symmetry in the crystal has been demonstrated for gramicidin-s by X-ray crystallography.2 From the known amino acid sequence,3' 4 it seemed likely that this conformation would hold in solution; this is confirmed by the NMR results reported here. Ten amino acid residues, two of each kind, are present. The strong single resonance band at 7.3 ppm in CD30D (and in DMSO if the partially interfering amide signal is subtracted, see later) must result from the ten aromatic protons of the two phenylalanines. With this as a basic inference, integration under the peak areas assigned to amide bands then corresponds to two protons for each of four regions, although these areas are not all clearly separated by a single solvent and temperature condition. Their assignment to a particular amino acid residue is made possible by decoupling and deuterium exchange experiments as outlined below. The assignments of the proton chemical shifts in the gramicidin S-A spectra are based on the following: (a) comparison with NMR spectra of constituent amino acids and known peptides; (b) spectra taken on instruments of varying field strengths (60, 100, and 220 Miec); (c) spectra taken in solvents that promote and solvents that prohibit proton exchange; (d) double resonance experiments; and (e) variable temperature spectra. Decoupling experiments were necessary in order (a) to show that the splitting of the amide proton peaks was due to spin-spin interaction and not to chemical shifts (a fact also proved by the invariability of the amide coupling constants in both the 100- and 220-Mc spectra) and (b) to determine which Ca, Ca, and NH protons are coupled together. The coupling constants between these types of protons will be designated J<,,,B and JNC. A knowledge of these should define the stereochemistry around the Cc-C" and N-Ca bonds. 734

2 VOL. 61, 1968 BIOCHEMISTRY: STERN ET AL. 735 The 100- and 220-Mc NMR spectra of gramicidin S-A acetate in DMSO-d6 and CD30D are shown in Figures 1 and 2. The assignments of the chemical shifts and integrated areas are shown in Table 1. The relationship between the Ca, Cc, and amide proton areas, as detected by decoupling experiments, and a summary of the observations are shown in Table 2 and indicated by the symbols in Figure 1. Irradiation of each of the three amide doublets did not always produce a noticeable effect on the Ca proton region, but irradiation throughout the Ca proton region did collapse the amide doublets to singlets and had observable effects on the multiplet structures of certain side-chain regions. This permitted identification of the Cc, Ca, and amide proton regions of a single amino acid in the molecule. In CD3OD, the water signal at 4.8 ppm makes integration of the Ca region difficult, but if the temperature is raised or lowered, the water band can be shifted sufficiently to again reveal five areas, which are from the five pairs of Ca proton resonances. Four of these can be related to the four downfield amide resonances by appropriate spin decoupling experiments. The fifth area, by elimination, must be the proline Ca resonance, which has no amide proton with which to couple. The phenylalanine residue protons appear at 8.9, 7.3, 4.5, and 3.0 ppm. Leucine protons appear at 8.7, 4.6, and 1.5 ppm when it is assumed that the complex multiplet centered at 1.7 ppm (in CD3OD) is produced by the Ca protons of leucine, proline, valine, and ornithine and that the relative chemical shifts of their C protons do not differ greatly from those in the four free amino acids. The valine residue chemical shifts are identified on the assumption that the amino acid residue in the Ca proton region at 4.2 ppm that could give a triplet before decoupling (actually two superimposed doublets with JNC = Jap) is most probably valine. Conversion of this triplet to a doublet (Ja3s = 9.0 cps) by irradiation at 7.7 ppm in CD30D is further evidence for the valine assignments. The 220-Mc spectrum in DMSO-d6 shows a Ca proton region at a = 4-5 ppm consisting of two areas of two protons each and a complex area of six protons. The amide region, which is similar to that in methanol, has three doublets, a singlet, and a broad six-proton peak, although variable temperature studies were necessary to prove this. The assignments of Ca and amide protons is somewhat hampered because three pairs of Ca protons overlap at 4.3 ppm. Thus, irradiation at this field value sharpened the broad 9.15-ppm singlet, converted the amide doublet at 7.2 ppm to a singlet, and produced noticeable decoupling effects in the area at 2.9 ppm. This indicates that the 7.2 ppm- or 9.15-ppm area belongs to phenylalanine. To be consistent with the methanol assignments, the 9.15-ppm peak is assigned to Phe and the 7.2 ppm to Val. The 8.7-ppm amide doublet is assigned to Om. That at 8.35 ppm was assigned with certainty to Leuby the same reasoning, with methanol as solvent. The Om NH3+ groups are assigned a chemical shift at 8.1 ppm because of their behavior with change of temperature and their rate of exchange and because the area is equivalent to six protons. In addition, they did not exhibit coupling to any Ca proton. Acceptance of the amide assignments and other data can lead to the assignments of the C proton region in Table 1.

3 736 BIOCHEMISTRY: STERN ET AL. PROc. N. A. S. FIG. 1.-(Lower) 100-Mc spectra of gramicidin S-A in CD30D, and (upper) in DMSO-d6. Resonances: 0, 0, X, and + indicate regions connected by decoupling FIG. 2.-(Lower) 220-Mc spectra of gramicidin S-A in CDgOD, and (upper) in DMSO-d,. An additional peak appears at 8.9 in CHsOH.

4 VOL. 61, 1968 BIOCHEMISTRY: STERN ET AL. 737 TABLE 1. Assignment of resonances. DMSO-d6 CD30D No. of (ppm) (ppm) protons Type Methyl (Val and Leu) C-CH2-C (Pro, Orn, Leu), f, and CH (Leu, Val) Methyl (acetate ion) 2.9, , CH2 heteroatom or benzene ring H20 contaminant (DMSO); solvent impurity a CH (2 Phe (4.3), 2 Pro (4.3), 2 Val (4.3), 2 Leu (4.6), 2 Orn (4.8)) a CH (2 Val (4.2), 2 Pro (4.35), 2 Phe (4.5), 2 Leu (4.7), 2 Om (4.95)) Amide N-H (Val) move upfield at higher temp Aromatic rings (Phe) Amineprotons (Omn) Amide N-H (Val) Amide N-H (Leu) Amide N-H (Orn) Amide N-H (Phe) The variation in the NMR spectra of gramicidin S-A in both CH30H and DMS0 with change of temperature is shown in Figures 3 and 4. These spectra show quite clearly that the chemical shifts of all amide protons are to higher field with increasing temperature. An additional feature not shown was the uncovering of the ornithine and leucine Cc proton resonances in methanol due to shifts of the water proton signal relative to the Ca proton resonance. Stereochemistry of gramicidin S-A: A brief NMR study in DMS0 at 100 Me has been published.' However, certain of their spectral assignments and exchange data to not agree with the present work, e.g., (a) incorrect assignment of the Ca and amide protons to specific areas by analogy with the values found in helical polyamino acids; (b) assignment of the phenylalanine amide peak at TABLE 2. Area irradiated (ppml) Uncoupling experiments. Change observed in CD30D Triplet at 4.2 to doublet None 8.8 None 8.9 None 4.2 Doublet at 7.7 to singlet 4.5 Broad multiplet at 3.0 to narrower and structured multiplet; sharpening at Doublet at 8.8 to a singlet (downfield portion of doublet of doublets at 8.75); upfield portion (1.5) of the complex multiplet (1.7) becomes sharper 4.95 Doublet at 8.7 to singlet (upfield portion of doublet of doublets at 8.75) Area irradiated (ppm) Change Observed in DMSO None Broad quartet at 4.6 to triplet 8.7 Broad peak at 4.8 sharpens 9.15 None 4.3 Doublet at 7.2 to singlet; singlet at 9.15 sharpens; broad multiplet at 2.9 to differently structured multiplet 4.6 Doublet at 8.35 to singlet (upfield portion (1.3) of the complex multiplet (1.5) becomes sharper) 4.8 Doublet at 8.7 to singlet

5 738 BIOCHEMISTRY: STERN ET AL. PROC. N. A. S FIG. 3.-(Lower) Spectra of gram- FIG. 4.-(Lower) Spectra of gramicidin icidin S-A in CH30H at -35, and S-A in DMS0-H6 at 400, and (upper) at (upper) at 200 and and ppm to an impurity; (c) assignment of the peak at 8.1 ppm, which we assign to the amino groups of ornithine, to four amide protons; and (d) failure to detect the two valine amide protons at 7.2 ppm. Furthermore, the disappearance of the amide peaks during exchange as reported by them did not occur in the same order observed here. They found that the amide protons at 8.7 ppm exchanged more slowly than those at 8.3 ppm, in contrast to our results in Table 3. The data in Table 1 indicate four amide and five Ca proton chemical shifts, with the integrated area of each being equivalent to two protons. This means that the Cc and NH protons of each of the pair of corresponding amino acids are magnetically equivalent and consequently have the same electronic environment. This can mean that the pairs of protons are related by a C2v symmetry axis. The magnitudes of the coupling constants JNC also give information about the backbone stereochemistry, i.e., rotation around the N-C bonds. The amide proton signals in methanol may be divided into two groups on the basis of coupling constants, the broad singlet in the most downfield position and the three doublets further upfield having JNC = cps. This suggests that the average (not necessarily the maximum) N-C bond dihedral angle of the three amino acid residues represented by these doublets is similar but different from that of Phe. Examination of the literature reveals a reasonable relationship between JNC and the dihedral angle 4. Thus the coupling constants of N-alkyl-substituted formamides have been evaluated.7 The average vicinal coupling constant JNC for an HNCH proton pair is 5 cps, assuming free rotation about the N-CH3 axis in N-methylformamide. Free rotation about the N- TABLE 3. Approximate rates of proton-deuterium exchange. -Val-----~ -0Orn--- - Leu Phe OGm NH3 + ppm T/2 ppm T/2 ppm T/2 ppm T/2 ppm T/2 CD30D 7.7 >2 weeks hr 8.8 abouti 8.9 <0.5 hr week DMSO(D.) 7.2 >2 weeks hr 8.35 about <0. 5 hr hr + 5% week D20

6 VOL. 61, 1968 BIOCHEMISTRY: SPERN VT AL. 739 alkyl bond should be relatively more restricted when R = ethyl and even more when R = isopropyl. The latter two compounds have coupling constants of 6.0 and 7.7 cps, respectively. Also, the NMR spectrum of poly-l-alanine in deuterochloroform exhibits an amide proton peak split into a symmetrical doublet with JNC = 6 cps. In the alpha helical structure of poly-l-alanine, the configuration around the NH-CH bond is trans.6 The diketopiperazines have either a cis or slightly gauche conformation in solution. The JNC values of some diketopiperazines measured in several solvents range from 0 to less than 3 CpS.8 Therefore, it seems reasonable to assume that small values of JNC correspond to large NC dihedral angles, whereas large values of JNC correspond to small dihedral angles. The fact that JNC in six of the eight possible amino acid residues is cps suggests strongly that their dihedral angles are small (q6 about 30, nearly trans), whereas the remaining two Phe residues have dihedral angles which are large (about 1500, nearly cis).t Knowledge of the complete stereochemistry of the side chains requires a knowledge of the relative populations of all the isomers formed by rotation around the Ca l& bonds and the J., of each. However, J#'s for the two valines (9.0 cps in DMSO) and leucines (identical to Jaj in the free amino acid) implies the same C2 symmetry axis as the backbone. The data suggest that this is true for the phenylalanines and ornithines, but with these and the prolines further measurements are needed. Conformational model of gramicidin S-A: Any proposed conformational model of gramicidin S-A must satisfy the following minimum requirements: (1) It must have (a) a C2v axis of symmetry, (b) a small (trans) dihedral angle (4) that is similar for Orn, Leu, and Val, (c) a large (cis) dihedral angle for Phe. (2) It must account for four slowly and four less slowly exchangeable protons (see Table 3). (3) It must explain the coupling constants and chemical shifts of the Ca, Ca, and amide protons. (4) The postulated hydrogen bonds must involve the amide protons of Val and Leu. A model for gramicidin S-A that satisfies these restrictions was easily constructed from Corey-Pauling-Koltun models in the following way: two tripeptides L-Val-L-Orn-L-Leu were arranged with 4 = 300 and co = 00, and were placed antiparallel side by side. Each tripeptide was then joined by two D-Phe-L-Pro arranged with OPhe = 1500 and 4tPro about 600 and with the proline ring tipped in a nearly vertical position (see drawing). The four required bonds were joined easily and adjusted without strain. This model, Figure 5, permits hydrogen bonding of the valine amide protons to the carbonyl oxygens of leucine, and of the leucine amide protons to the carbonyl oxygen of valine. The valine amide protons are located near the bend formed by the Phe-Pro sequence and have a shorter bond distance than the leucine amide protons. This could explain the relative rates of exchange of the valine and leucine amide protons. The model also. can explain the rapid exchange of the ornithine and phenylalanine amide protons because they point outward from the ring and cannot be internally hydrogen-bonded. The rates are supported by tritium exchange data.1' The model also predicts the relatively high-field chemical shift of the valine amide protons on the basis of a

7 740 BIOCHEMISTRY: STERN ET AL. PROC. N. A S. FIG. 5.-Schematic drawing of model with w angles all 0 except Leu 10. vc angles of Val, Orn, and Leu = 300, Phe = , angles of Val, Orn, and Leu = 0, Phe = 330, and Pro = Dashed lines H indicate hydrogen bonds. shielding effect from one or both carbonyl groups of the phenylalanine and valine residues. At least six of the peptide carbonyl groups and the corresponding amide hydrogen bond are in the plane perpendicular to the C2 axis, while two phenylalanine peptide carbonyl groups and amide hydrogen bonds are almost parallel to this axis; therefore, this molecule could be expected to exhibit infrared dichroism9 in the carbonyl and amide regions and possibly the per cent of cis found. 13 The over-all symmetry and dimensions of this model are consistent with those found in the crystal2 and with surface film measurements made by A. Rothen.10 The latter indicated a flat molecule 9 A in thickness with a hydro-

8 VOL. 61, 1968 BIOCHEMISTRY: STERN ET AL. 741 phobic and a hydrophilic side. The rigidity of the molecule is confirmed by thin-film dialysis and possibly by the optical rotatory dispersion data.12 Many different conformations have been proposed for gramicidin S.2, 9, 5, None of these is entirely consistent with our data, except possibly that of Hodgkin -and Oughton2 and that of Schwyzer.16 However, the latter fixes only part of the conformation now indicated by our data. Further work with computersimulated conformations is in progress. We wish to thank the Bell Telephone Laboratories for use of their 100- and 220-Mc NMR spectrometers and especially Dr. Frank Bovey and associates for their advice and helpful discussions. * This work was supported in part by grants AM and CA from the National Institutes of Health. t Supported by postdoctoral fellowship G. M. 29,426 from the National Institutes of Health. I For nomenclature see J. Biol. Chem., 241, 1004 (1966). la Synge, R. L. M., Biochem. J., 39, 363 (1945); bcraig, L. C., J. D. Gregory, and J. T. Barry, in Cold Spring Harbor Symposia on Quantitative Biology, vol. 14 (1950), p Hogkin, D. C., and B. M. Oughton, Biochem. J., 65, 752 (1957). Consden, R. J., A. H. Gordon, A. J. P. Martin, and R. L. M. Synge, Biochem. J., 41, 596 (1947). 4Battersby, A. R., and L. C. Craig, J. Am. Chem. Soc., 73, 1887 (1951). 5 Liquori, A. M., and F. Conti, Nature, 217, 635 (1968). 6Schellman, J. A., and C. Schellman, in The Proteins, ed. H. Neurath (New York: Academic Press, 1964), vol. 2, 2nd ed., p. 1. 7La Planche, L., and M. T. Rogers, J. Am. Chem. Soc., 86, 337 (1964). 8 Bovey, F., personal communication. 9 Abbott, N. B., and E. J. Ambrose, Proc. Roy. Soc. (London), Ser. A, 219, 17 (1953). 10 Rothen, A., unpublished results. 11 Laiken, S., M. Printz, and L. C. Craig, in preparation. 12 Craig, L. C., these PROCEEDINGS, 61, 152 (1968). 13 Balasubramanian, D., J. Am. Chem. Soc., 89, 5445 (1967). 14Vanderkooi, G., S. J. Leach, G. Nemethy, R. A. Scott, and H. A. Scheraga, Biochemistry, 5, 2991 (1966). 15 Warner, D. T., Nature, 190, 120 (1961). 16Schwyzer, R., in Amino Acids and Peptides with Antimetabolic Activity, ed. G. E. W. Wolstenholme (London: J. and A. Churchill, Ltd., 1958), p Gavrilov, N. J., N. A. Poddubnaja, L. N. Akimova, and E. M. Grigor'eva, Zh. Obshch. Khim., 26, 2029 (1956).