Proc. Natl. Acad. Sci. USA Vol. 74, No. 1, pp. 13-17, January 1977
Chemistry
Stable conformations of aliphatic disulfides: Influence of 1,4 interactions involving sulfur atoms (1,4 carbon-sulfur interaction/1,4 nitrogen-sulfur interaction/rotamers)
HAROLD E. VAN WART* AND HAROLD A.
SCHERAGAtf
* Biophysics Research Laboratory, Department of Biological Chemistry, Harvard Medical School, Boston, Massachusetts 02115; and t Department of Chemistry, Cornell University, Ithaca, New York 14853
Contributed by Harold A. Scheraga, October 29, 1976
The present state of knowledge of the stable ABSTRACT conformations of the XCCSSCCX fragment (in which X is any saturated group) of aliphatic disulfides is examined. In particular, the evidence for the existence of attractive 1,4 interactions between CH groups and S atoms across C-S bonds and their influence on the potential function for rotation about C-S bonds is discussed. The effects of similar attractive interactions between NH groups and S atoms on the potential function for rotation about C-C bonds also are considered. It is concluded that weak attractive interactions between CH groups and S atoms are capable of stabilizing rotamers with unusually low (about 30°) values of the SS-CC dihedral angle.
In a recent series of papers (1-8) from this laboratory, various experimental and theoretical techniques were used to study the conformations of aliphatic disulfides structurally related to cystine. On the basis of these studies, it was suggested that there exist weak attractive interactions between CH groups and S atoms that are separated by three intervening single bonds. These were referred to as 1,4 carbon-sulfur interactions, and were thought to be responsible for the presence of rotamers about the C-S bonds of aliphatic disulfides with unusually low values (about 300) of the SS-CC dihedral angle. Such rotamers were referred to as A conformations, and were thought to comprise about 20% of all rotamers about C-S bonds. The existence of A conformations is unexpected on the basis of "classical" conformational analyses of saturated molecules in which heavy groups/atoms generally would be considered to be only gauche or trans (SS-CC dihedral angles of +600 and 1800, respectively) to each other. While the available experimental and theoretical evidence strongly supports the existence of A conformations, a direct proof of this hypothesis has been difficult to establish because of inherent experimental and theoretical limitations. We, therefore, thought it useful to examine briefly what is known about the conformation of aliphatic disulfides in general, devoting special attention to the evidence regarding the existence of 1,4 carbon-sulfur interactions and A conformations. The possible existence of attractive 1,4 interactions between S atoms and NH groups also will be considered.
S-S bond is quite hindered and that the equilibrium value of the CS-SC dihedral angle [x(CS-SC)] for open-chain disulfides is approximately +85'. This is confirmed by the results of x-ray crystallographic studies (summarized in refs. 4, 6, and 7), and by gas phase electron diffraction (9, 10) and microwave (11) studies on various disulfides in which values of x(CS-SC) within 200 of ±850 are invariably observed. It should be pointed out that, because the CSSC unit has C2 symmetry and its S-S bond is a chiral center, conformations with values of X(CS-SC) of +0 and -0 are mirror images of one another and, hence, isoenergetic. Therefore, symmetric disulfides with values of X(CS-SC) of +85° and -85° correspond to different screw-senses of the same conformation (i.e., they are enantiomers). Ring closure or other structural constraints can cause the S-S bond to adopt values of Ix(CS-SC)I in the range of 0-60° or 110-180°. As Ix(CS-SC)I is reduced from near 850 to near 00, there is a lengthening of the S-S bond (2, 7) from about 2.02 to about 2.10 A, a reduction in the S-S stretching frequency (1, 2, 7, 8) [v(S-S)] from about 510 to about 480 cm-1, and a shift in the longest wavelength ultraviolet absorption band from about 250 to about 380 nm (1). The value of v(S-S) observed for strained disulfides can be used to estimate the value of
lx(CS-SC)I (7, 8). Rotation about C-S bonds Before discussing the stable conformations about the C-S bonds of the XCCSSCCX unit, we first define the nomenclature to be used to describe the possible conformations. The C (for cis), A, G (for gauche), B, and T (for trans) conformations are those with SS-CC dihedral angles [x(SS-CC)] of roughly 0, +30, +60, ±90, and 180°, respectively. It should be noted that, for a given screw sense of the disulfide bond, conformations of the CCSSCC unit with values of +0 and -0 for x(SS-CC) are nonequivalent and, hence, not necessarily isoenergetic. For example, when x(CS-SC) is +85°, the G rotamer with the value of X(SS-CC) of -600 is thought (ref. 12 and references therein) to be of higher energy than the G rotamer with the value of X(SS-CC) of +600 because of steric repulsions between CH groups across the S-S bond. To distinguish between these two rotamers, the higher energy one will always be designated with a prime (e.g., A', G', or B'). Using the G rotamer as an example, it should also be noted that x(SS-CC) = +600 when '(CS-SC) = +85°, but X(SS-CC) = -600 when X(CS-SC) = -85°. Similarly, for the G' rotamer, X(SS-CC) = -60° when X(CS-SC) = +850, but x(SS-CC) = +60° when X(CS-SC) = -850. Some time ago Scott et al. (12) pointed out that the Raman and infrared spectra of diethyl disulfide contained bands arising from the coexistence of rotational isomers about its C-S bonds. More recently, Sugeta et al. (13, 14) showed that the Raman spectra of a series of primary disulfides larger than dimethyl disulfide all exhibit bands arising from such rotamers [a primary
RESULTS AND DISCUSSION The conformation of the XCCSSCCX fragment will be considered, where X is any saturated group such as the amino group of cystine. The stable conformations of this unit, arising from rotations about the S-S, C-S, and C-C bonds, will be considered separately. Rotation about S-S bonds The results of several different types of molecular orbital calculations (summarized in ref. 2) indicate that rotation about the t To whom requests for reprints should be addressed.
13
14
Chemistry: Van Wart and Scheraga
disulfide is one in which the carbon atom adjacent to the disulfide bond is a primary (-CH2-S) carbon atom]. In particular, a prominent v(S-S) band near 510 cm-' and a satellite v(S-S) band near 525 cm-1 were observed for methyl ethyl disulfide and all larger primary alkyl disulfides studied. Both Scott et al. (12) and Sugeta et al. (13, 14) made the assumption (adapted from disubstituted ethanes) that rotation about C-S bonds resuited in only either G or T rotamers. The v(S-S) band near 510 cm I was attributed (13-15) to a conformation of the CCSSCC unit with G rotamers about both C-S bonds, and the v(S-S) band near 525 cm-1 to a conformation with a G rotamer about one C-S bond and a T rotamer about the other. The conformation with T rotamers about both C-S bonds would have been expected (13, 14) to have a value of v(S-S) of about 540 cm-1. The differences in the values of v(S-S) of the G and T rotamers was attributed (13, 14) to a greater degree of vibrational coupling between the CCS deformation ['y(CCS)J and v(S-S) motions in the T relative to the G form. Calculations of normal mode frequencies, claiming to support these proposed correlations, have been carried out (15, 16). From an examination of the temperature dependence of the low frequency Raman spectrum of methyl ethyl disulfide [the simplest disulfide for which Raman spectral evidence of rotational isomerism about its C-S bond had been reported (13, 14)], it subsequently was established (5) that at least three rotamers about the CH3CH2-SS bond coexist. In particular, two of the rotamers were thought to be responsible for the prominent S-S stretching band near 510 cm-1 and a third (higher energy) rotamer for the weaker 525 cm-' band. In order to establish frequency-conformation correlations that could be used to determine the conformations of cystine residues in proteins from their Raman spectra, it was necessary to ascertain the conformations of the rotamers present about the C-S bonds. Rather than assume the existence of any particular rotamer a priori, it was decided (6) to survey the conformations about the C-S bonds of small open-chain cystine-related disulfides whose structures were known from x-ray crystallography to determine which rotamers about the C-S bond commonly occur. About a dozen such structures are known and a histogram, shown in Fig. 1A, illustrating the frequency of occurrence of various ranges of Ix(SS-CC)I § in 100 increments, was constructed from these data. The data plotted are those of the 11 primary aliphatic disulfides of Table IV in ref. 6 and also the n-propyl half of thiamine n-propyl disulfide (17). Most of the occurrences are in the region of the B conformation (with a few spreading into the G region), and the remaining are in the T region. This implies that there are minima in the potential function for rotation about the C-S bonds of these disulfides in the B, B', and T regions. The Raman spectra of several of these crystalline disulfides of known structure (from Fig. 1A) were examined to correlate the conformation about their C-S bonds with their values of v(S-S). It was found (6) that compounds with the G (bis[2-
(N,N'-dimethylamino)ethyl]disulfide), G' (cyclo-L-cystine), B (N,N'-diglycyl-L-cystine and dibenzyl disulfide), and T (dithioglycolic acid) conformations about both C-S bonds all have values of v(S-S) close to 510 cm-'. This finding contradicts §
There is a roughly equal distribution of disulfide screw senses in proteins. It should be recalled that the sign of X(SS-CC) for a given rotamer (e.g., G, G', etc.) depends upon the screw sense of the CSSC fragment. In order to plot the frequencies of occurrence of the values of X(SS-CC) conveniently on one graph, we have plotted x(SS-CC)i and, hence, have not distinguished between, for example, G and G' conformations. However, as expected, the majority of occurrences lie in the unprimed domain of the values of x(SS-CCQ.
Proc. Natl. Acad. Sci. USA 74 (1977) A. Model Disulfides
x2
6H
6
1
OL_
B. Cys- X2
0
04y
20
,0
O
10 _
D. Leu-X1 0
30_
O _ 25 0
60
Ix'
I
1
120
180
FIG. 1. Frequency of occurrence of various ranges of dihedral angles determined by x-ray crystallography for (A) the values of X(SS-CC)I of model disulfides, (B) the values of Ix2(CaCC-SS)I of cystine residues in eight proteins, (C) the values of xl I of cystine and cysteine residues in 26 proteins, and (D) and (E), respectively, the values of x I and x21 for the leucine residues in 26 proteins.
the correlations proposed earlier (13-15), whereby aliphatic disulfides with T conformations about one or both C-S bonds were thought to have values of v(S-S) of about 525 and 540 cm-, respectively. The finding that all these conformations have nearly the same value of v(S-S), despite their widely different values of x(SS-CC), rules out vibrational coupling between y(CCS) and v(S-S) modes of the kind proposed (13, 14, 16) as an explanation for the existence bf rotational isomers with v(S-S) bands of different frequency. If such vibrational coupling existed, the values of v(S-S) of the five compounds mentioned above would not have been the same. Furthermore, conclusions regarding the frequencies of the various rotamers, drawn from normal mode analyses (15, 16) that were based upon the existence of such coupling interactions, are equally invalid; this is indicated by their erroneous prediction that T rotamers should have higher values of v(S-S) than G rotamers. The above results (combining Raman and x-ray data on crystalline dithioglycolic acid) clearly indicate that the T rotamer is not responsible for the 525 cm-1 band in methyl ethyl disulfide. Unfortunately, the above studies were unable to establish the identity of the rotamer responsible for this band. Yokozeki and Bauer (10) attempted to elucidate the struc-
Chemistry:
Van Wart and Scheraga
tures of the rotamers present about the C-S bond of methyl ethyl disulfide in the gas phase using the electron diffraction technique. Because of the coexistence of at least three rotamers, it was not possible to determine the structures of the rotamers present uniquely, and an extremely large number of possible solutions was possible. The authors were able to conclude, however, that there exists a substantial fraction of the T form. The potential function for rotation about the C-S bonds of several alkyl disulfides was investigated using the CNDO/2 (complete neglect of differential overlap) approximate molecular orbital method (3, 4). It was hoped that the results of these calculations would provide insight into the conformations of the rotamers present. The potential functions obtained were characterized by minima at the C and T positions. The appearance of the C minima was unexpected, and an analysis (4) of the nonbonded contact distances in C conformations between the S atoms and CH groups in the 1,4 positions across the C-S bonds revealed that the calculations resulted in contacts shorter than those observed experimentally. This result, taken together with the known tendency of the CNDO/2 method to underestimate overlap repulsions, led to the suggestion (4) that the C minima were probably artifacts, and that the true minima were probably somewhere between the C and G conformations (at about 300) where more normal CH.. S contact distances occur. The CNDO/2 calculations did not reveal the expected minimum in the region of the B conformation (Fig. IA), presumably because it was overshadowed by the overestimation of the stability of the C conformation. While the presence of C minima in the calculated potential functions was considered to be an artifact, it seemed that the calculations might be qualitatively correct in that they pointed to real, though weak, attractive nonbonded interactions between S atoms and CH groups in the 1,4 positions. This led to a literature search in which an unusually large number of short nonbonded contacts between CH groups and S atoms were revealed (4), supporting the existence of a 1,4 carbon-sulfur interaction and strengthening the idea that a rotamer with a value of X(SS-CC) between the C and G regions-the A rotamer-existed. The conformations about the C-S bonds of model disulfides in crystals are all found to be in the lower energy B and T conformations. It was thought that there might be a chance of observing higher energy rotamers about the C-S bonds of cystine residues in protein crystals where long-range intramolecular interactions might provide sufficient energy to populate higher energy minima. The conformations about the C-S bonds of all the cystine residues in eight proteins (6) whose coordinates were available at the time were examined (6) and the histogram§ shown in Fig. lB was constructed. As expected from the crystal conformations of smaller disulfides (Fig. 1A), there are peaks in the B and T regions. In addition, however, a third peak near 30'-due to the A conformation-is apparent. On the basis of these data, it was suggested (6) that this was the "missing" rotamer that was responsible for the v(S-S) band near 525 cm-' in the Raman spectra. Accordingly (6), conformations of the CCSSCC moiety with an A conformation about one C-S bond and either a B or T conformation about the other have values of v(S-S) of about 525 cm-I, while the presence of A conformations about both C-S bonds results in a value of v(S-S) of about 540 cm-1 It should be pointed out that A conformations are observed in crystals (4) of small aromatic disulfides (namely, those in which the S atoms are bonded directly to phenyl rings). Furthermore, the v(S-S) band of crystalline diphenyl disulfide (in which there are A conformations about both C-S bonds) lies at
Proc. Natl. Acad. Sci. USA 74 (1977)
15
about 540 cm-1, in agreement with the value predicted for aliphatic disulfides (6). In solution, diphenyl disulfide has v(S-S) bands at about 510, 525, and 540 cm-l corresponding to molecules with A conformations about neither, one, and both C-S bonds, respectively (6). Although there is no reason to expect a prior that the dependence of the values of v(S-S) on conformation about C-S bonds in aliphatic and aromatic disulfides should be the same, the agreement found above for diphenyl disulfide (using the proposed correlation) suggests that this may be the case. In principle, one could examine the Raman spectra of all proteins having cystine residues with A conformations about their C-S bonds to see if their presence correlates with Raman bands at about 525 and 540 cm-1. In practice, such a comparison requires an extremely high quality protein Raman spectrum. For lysozyme, a protein for which a high resolution spectrum in the v(S-S) region was available, this comparison (6) was made, and the frequencies and intensities of the bands observed in the v(S-S) region were found to be consistent with the proposed correlation. The reliability of the histogram shown in Fig. IB, in pointing to the existence of A conformations, is difficult to establish unambiguously. While the errors in the dihedral angles obtained from x-ray data on proteins are generally larger than for small molecules, the S atoms are usually among the best resolved atoms and, hence, the values of X(SS-CC) are among the most accurately known. Even assuming an error for each value of X(SS-CC) as large as 200, it seems unlikely that random error would result in the presence of the well-defined peak in the A region. To gain some insight into how such a histogram, derived from protein x-ray data, would appear for a side-chain rotational angle whose energy minima are better understood, similar histograms for the values of xI and x2 of leucine (in which there are heavy atoms, but not sulfur atoms, in the 1,4 positions across C-C bonds) are shown in Fig. ID and E. The data set included all the leucine residues in the twenty proteins used in ref. 18 plus those of the A and B chains of horse deoxy (19) and methemoglobin (A. D. McLachlan, personal communication to the Brookhaven Protein Data Bank), rubredoxin (20), a-chymotrypsinogen (21), dogfish lactate dehydrogenase (22), and trypsin (R. Stroud, personal communication). Leucine was chosen because it has a large hydrocarbon side chain whose conformations are well understood (i.e., its values of xI and x2 are expected to cluster in only the gauche and trans regions). Furthermore, since it is a residue generally found in the protein interior, its atomic coordinates are less likely to be influenced by thermal motion than those of exterior residues, such as lysine or glutamic acid. As can be seen from Fig. ID and E, the expectation that peaks in only the G and T regions occur is borne out, and no new or unexpected conformations are indicated. Hence, these histograms reliably reflect the identity of the rotamers present. Unfortunately, since few of the proteins whose structures are known from x-ray studies contain cystine residues, the number of occurrences in Fig. 1B is smaller than in Fig. iD and E and, hence, must be considered somewhat less reliable. Recently, Allinger and coworkers (23) calculated the potential function for rotation about the C-S bond of methyl ethyl disulfide using a semiempirical molecular mechanics method. They found energy minima in only the B, B', and T regions and not in the A region. The B and T conformations were of nearly equal energy, while the B' conformation was higher in energy by about 0.6 kcal/mol (1 cal = 4.184 J). The potential was found to be very soft, with energies less than about 1 kcal/mol for all
16
Chemistry: Van Wart and Scheraga
values of x(SS-CC), except those close to the C form. Clearly, the results derived from any semiempirical calculation reflect solely those interactions that are "built-in" in the parameterization. Since the authors admittedly (23) made no effort to take into account the existence of the 1,4 carbon-sulfur interaction, thought (4,6) to stabilize the A conformation (which could have been done, e.g., by computing the crystal structures of those organosulfur molecules that show short CH... S contacts), they pointed out that their failure to find the A conformation did not necessarily mean that it did not exist. Indeed, the fact that their calculated potential in the region of the A conformation is so soft indicates that, had a small stabilization energy been included in that region, such as that from the proposed (4) 1,4 carbon-sulfur interaction, a local energy minimum capable of accounting for an A rotamer population of 20% could easily result. Allinger et al. (23) also calculated the relative energies of the minimum- and maximum-energy conformations (a total of six values) in the potential function obtained from the molecular mechanics calculated using a minimum basis set ab initio molecular orbital method. It was hoped that this method would reveal any unexpected electronic effects in a 1,4 carbon-sulfur interaction without having to be parameterized for it. The authors (23) found that these ab initio calculations did not provide any evidence for the existence of such an interaction. Unfortunately, their ab initio calculations completely omitted the very region in which the A conformations are proposed (6) to exist (about 300); the closest conformations examined had values of x(SS-CC) of 00 and 720. For this reason, these calculations cannot decide anything about the existence of A conformations involving 1,4 carbon-sulfur interactions. Because of the inherently weak nature of any attractive 1,4 carbonsulfur interaction, only a thorough examination of the A region of the C-S potential function, using a large basis set ab initio technique in which geometries are optimized at each conformation, could be trusted to reflect the true situation. For economic reasons, such calculations have proven to be beyond present means. The molecular mechanics calculations discussed above are valuable in that they provide an idea of the relative stability of the B and T rotamers and in that they predict the presence of the B' rotamer. However, in spite of the claims by the authors (23), these calculations are not in agreement with the experimental data. Since only B, B', and T minima are found in these calculations, the presence of the 525 cm-' band in the Raman spectrum of methyl ethyl disulfide cannot be accounted for; it was shown earlier (6) that this band cannot be due to any of these three conformers. Furthermore, the calculations are not compatible with either the existence of A conformations about the C-S bonds of cystine residues in proteins (shown in Fig. 1B) or the multitude of experimentally observed (4) short nonbonded CH ...S contacts. Finally, while the calculations are consistent with the electron diffraction data (10), so are an infinite number of other mixtures of rotamers that include the A conformation. In summary, the available experimental and theoretical evidence appears to be consistent with a rotamer population about the C-S bonds of aliphatic disulfides of about 20% A, 40% B+B', and 40% T. This distribution is consistent with the electron diffraction and Raman data, the histogram in Fig. 1B, and the combined results of the CNDO/2 and molecular mechanics calculations. While it would be highly desirable to find examples of A conformations in crystals of smaller aliphatic disulfides (instead of only aromatic disulfides and cystine residues in protein crystals) whose Raman spectra could be studied directly,
Proc. Natl. Acad. Sci. USA 74 (1977)
this has not proved possible so far, presumably because these are higher-energy rotamers. Hence, allowing for the possibility that the histogram in Fig. IB, for one reason or another, is misleading, all of the evidence so far supporting the existence of A conformations is somewhat indirect. On the other hand, if one is to argue that A conformations do not exist, then the presence of the 525 cm-' band in the Raman spectra of aliphatic disulfides must still be accounted for by other than the G, G', B, B', or T rotamers. Hopefully, other techniques, such as microwave or nuclear magnetic resonance spectroscopy, may be able to provide more detailed information about the rotamers present. Rotation about C-C bonds We now consider the possible existence of 1,4 X,S interactions in the XCCSSCCX unit and the effect that such interactions might have on the potential function for rotation about C-C bonds. While attractive X... S interactions are possible for a variety of groups, discussion will be limited specifically to the case where X is NH, since attractive interactions between these groups in the form of NH .. S hydrogen bonds are known (24) to exist. However, the results for this group can be generalized to other groups (such as OH, etc.) capable of such attractive interactions with S atoms. The case where X is NH arises for cystine residues in proteins. One might expect attractive 1,4 NH-.. S interactions to influence the potential function for rotation about the Ca-C0 bonds of cystine and cysteine. To see if this is the case, the distribution of conformations observed about the Ca-C, bonds of all of the cystine and cysteine residues in 26 proteins (refs. 18-22; A. D. McLachlan, personal communication to Brookhaven Protein Data Bank and R. Stroud, personal communication) has been examined. The results are shown in the histogram in Fig. iC, in which the frequency of occurrence of various ranges of values (in 10° increments) of Xl(NCa-COS) is plotted. The peaks in this histogram fall in the G and T regions only, with the broad peak in the G region [X(NCa-C0S) = 60° ± 60° ] accounting for 78% of all occurrences, and the broad peak in the T region [X(NCa-C0S) = 180° ± 60°] accounting for the remaining 22%. The histogram in Fig. IC differs from that in Fig. 1B in that there is no peak in the A region. Hence, whereas 1,4 CH... S interactions across the C8fS bonds of cystine residues result in A conformations, 1,4 NH ...S interactions across the Ca-Cbonds of cystine and cysteine residues do not result in A conformations. Instead, the distribution of conformations observed about the Ca-C/3 bonds of cystine and cysteine residues qualitatively resembles that found for leucine (Fig. iD). Quantitatively, however, the distribution for cystine and cysteine favors the G region over the T (78:22) to a somewhat greater extent than for leucine (67:33). The differences between the effects of 1,4 CaH .. S interactions on the C0-S potential function and of 1,4 NH-.. S interactions on the Ca-Co potential function can be understood by considering the differences in the inherent structural features of the CaCCSS and NCaC'CS units. For example, an essential structural feature that will influence the potential function for rotation about a given bond is the bond length. Considering Ca-S bonds (1.83 A)
Chemistry
Stable conformations of aliphatic disulfides: Influence of 1,4 interactions involving sulfur atoms (1,4 carbon-sulfur interaction/1,4 nitrogen-sulfur interaction/rotamers)
HAROLD E. VAN WART* AND HAROLD A.
SCHERAGAtf
* Biophysics Research Laboratory, Department of Biological Chemistry, Harvard Medical School, Boston, Massachusetts 02115; and t Department of Chemistry, Cornell University, Ithaca, New York 14853
Contributed by Harold A. Scheraga, October 29, 1976
The present state of knowledge of the stable ABSTRACT conformations of the XCCSSCCX fragment (in which X is any saturated group) of aliphatic disulfides is examined. In particular, the evidence for the existence of attractive 1,4 interactions between CH groups and S atoms across C-S bonds and their influence on the potential function for rotation about C-S bonds is discussed. The effects of similar attractive interactions between NH groups and S atoms on the potential function for rotation about C-C bonds also are considered. It is concluded that weak attractive interactions between CH groups and S atoms are capable of stabilizing rotamers with unusually low (about 30°) values of the SS-CC dihedral angle.
In a recent series of papers (1-8) from this laboratory, various experimental and theoretical techniques were used to study the conformations of aliphatic disulfides structurally related to cystine. On the basis of these studies, it was suggested that there exist weak attractive interactions between CH groups and S atoms that are separated by three intervening single bonds. These were referred to as 1,4 carbon-sulfur interactions, and were thought to be responsible for the presence of rotamers about the C-S bonds of aliphatic disulfides with unusually low values (about 300) of the SS-CC dihedral angle. Such rotamers were referred to as A conformations, and were thought to comprise about 20% of all rotamers about C-S bonds. The existence of A conformations is unexpected on the basis of "classical" conformational analyses of saturated molecules in which heavy groups/atoms generally would be considered to be only gauche or trans (SS-CC dihedral angles of +600 and 1800, respectively) to each other. While the available experimental and theoretical evidence strongly supports the existence of A conformations, a direct proof of this hypothesis has been difficult to establish because of inherent experimental and theoretical limitations. We, therefore, thought it useful to examine briefly what is known about the conformation of aliphatic disulfides in general, devoting special attention to the evidence regarding the existence of 1,4 carbon-sulfur interactions and A conformations. The possible existence of attractive 1,4 interactions between S atoms and NH groups also will be considered.
S-S bond is quite hindered and that the equilibrium value of the CS-SC dihedral angle [x(CS-SC)] for open-chain disulfides is approximately +85'. This is confirmed by the results of x-ray crystallographic studies (summarized in refs. 4, 6, and 7), and by gas phase electron diffraction (9, 10) and microwave (11) studies on various disulfides in which values of x(CS-SC) within 200 of ±850 are invariably observed. It should be pointed out that, because the CSSC unit has C2 symmetry and its S-S bond is a chiral center, conformations with values of X(CS-SC) of +0 and -0 are mirror images of one another and, hence, isoenergetic. Therefore, symmetric disulfides with values of X(CS-SC) of +85° and -85° correspond to different screw-senses of the same conformation (i.e., they are enantiomers). Ring closure or other structural constraints can cause the S-S bond to adopt values of Ix(CS-SC)I in the range of 0-60° or 110-180°. As Ix(CS-SC)I is reduced from near 850 to near 00, there is a lengthening of the S-S bond (2, 7) from about 2.02 to about 2.10 A, a reduction in the S-S stretching frequency (1, 2, 7, 8) [v(S-S)] from about 510 to about 480 cm-1, and a shift in the longest wavelength ultraviolet absorption band from about 250 to about 380 nm (1). The value of v(S-S) observed for strained disulfides can be used to estimate the value of
lx(CS-SC)I (7, 8). Rotation about C-S bonds Before discussing the stable conformations about the C-S bonds of the XCCSSCCX unit, we first define the nomenclature to be used to describe the possible conformations. The C (for cis), A, G (for gauche), B, and T (for trans) conformations are those with SS-CC dihedral angles [x(SS-CC)] of roughly 0, +30, +60, ±90, and 180°, respectively. It should be noted that, for a given screw sense of the disulfide bond, conformations of the CCSSCC unit with values of +0 and -0 for x(SS-CC) are nonequivalent and, hence, not necessarily isoenergetic. For example, when x(CS-SC) is +85°, the G rotamer with the value of X(SS-CC) of -600 is thought (ref. 12 and references therein) to be of higher energy than the G rotamer with the value of X(SS-CC) of +600 because of steric repulsions between CH groups across the S-S bond. To distinguish between these two rotamers, the higher energy one will always be designated with a prime (e.g., A', G', or B'). Using the G rotamer as an example, it should also be noted that x(SS-CC) = +600 when '(CS-SC) = +85°, but X(SS-CC) = -600 when X(CS-SC) = -85°. Similarly, for the G' rotamer, X(SS-CC) = -60° when X(CS-SC) = +850, but x(SS-CC) = +60° when X(CS-SC) = -850. Some time ago Scott et al. (12) pointed out that the Raman and infrared spectra of diethyl disulfide contained bands arising from the coexistence of rotational isomers about its C-S bonds. More recently, Sugeta et al. (13, 14) showed that the Raman spectra of a series of primary disulfides larger than dimethyl disulfide all exhibit bands arising from such rotamers [a primary
RESULTS AND DISCUSSION The conformation of the XCCSSCCX fragment will be considered, where X is any saturated group such as the amino group of cystine. The stable conformations of this unit, arising from rotations about the S-S, C-S, and C-C bonds, will be considered separately. Rotation about S-S bonds The results of several different types of molecular orbital calculations (summarized in ref. 2) indicate that rotation about the t To whom requests for reprints should be addressed.
13
14
Chemistry: Van Wart and Scheraga
disulfide is one in which the carbon atom adjacent to the disulfide bond is a primary (-CH2-S) carbon atom]. In particular, a prominent v(S-S) band near 510 cm-' and a satellite v(S-S) band near 525 cm-1 were observed for methyl ethyl disulfide and all larger primary alkyl disulfides studied. Both Scott et al. (12) and Sugeta et al. (13, 14) made the assumption (adapted from disubstituted ethanes) that rotation about C-S bonds resuited in only either G or T rotamers. The v(S-S) band near 510 cm I was attributed (13-15) to a conformation of the CCSSCC unit with G rotamers about both C-S bonds, and the v(S-S) band near 525 cm-1 to a conformation with a G rotamer about one C-S bond and a T rotamer about the other. The conformation with T rotamers about both C-S bonds would have been expected (13, 14) to have a value of v(S-S) of about 540 cm-1. The differences in the values of v(S-S) of the G and T rotamers was attributed (13, 14) to a greater degree of vibrational coupling between the CCS deformation ['y(CCS)J and v(S-S) motions in the T relative to the G form. Calculations of normal mode frequencies, claiming to support these proposed correlations, have been carried out (15, 16). From an examination of the temperature dependence of the low frequency Raman spectrum of methyl ethyl disulfide [the simplest disulfide for which Raman spectral evidence of rotational isomerism about its C-S bond had been reported (13, 14)], it subsequently was established (5) that at least three rotamers about the CH3CH2-SS bond coexist. In particular, two of the rotamers were thought to be responsible for the prominent S-S stretching band near 510 cm-1 and a third (higher energy) rotamer for the weaker 525 cm-' band. In order to establish frequency-conformation correlations that could be used to determine the conformations of cystine residues in proteins from their Raman spectra, it was necessary to ascertain the conformations of the rotamers present about the C-S bonds. Rather than assume the existence of any particular rotamer a priori, it was decided (6) to survey the conformations about the C-S bonds of small open-chain cystine-related disulfides whose structures were known from x-ray crystallography to determine which rotamers about the C-S bond commonly occur. About a dozen such structures are known and a histogram, shown in Fig. 1A, illustrating the frequency of occurrence of various ranges of Ix(SS-CC)I § in 100 increments, was constructed from these data. The data plotted are those of the 11 primary aliphatic disulfides of Table IV in ref. 6 and also the n-propyl half of thiamine n-propyl disulfide (17). Most of the occurrences are in the region of the B conformation (with a few spreading into the G region), and the remaining are in the T region. This implies that there are minima in the potential function for rotation about the C-S bonds of these disulfides in the B, B', and T regions. The Raman spectra of several of these crystalline disulfides of known structure (from Fig. 1A) were examined to correlate the conformation about their C-S bonds with their values of v(S-S). It was found (6) that compounds with the G (bis[2-
(N,N'-dimethylamino)ethyl]disulfide), G' (cyclo-L-cystine), B (N,N'-diglycyl-L-cystine and dibenzyl disulfide), and T (dithioglycolic acid) conformations about both C-S bonds all have values of v(S-S) close to 510 cm-'. This finding contradicts §
There is a roughly equal distribution of disulfide screw senses in proteins. It should be recalled that the sign of X(SS-CC) for a given rotamer (e.g., G, G', etc.) depends upon the screw sense of the CSSC fragment. In order to plot the frequencies of occurrence of the values of X(SS-CC) conveniently on one graph, we have plotted x(SS-CC)i and, hence, have not distinguished between, for example, G and G' conformations. However, as expected, the majority of occurrences lie in the unprimed domain of the values of x(SS-CCQ.
Proc. Natl. Acad. Sci. USA 74 (1977) A. Model Disulfides
x2
6H
6
1
OL_
B. Cys- X2
0
04y
20
,0
O
10 _
D. Leu-X1 0
30_
O _ 25 0
60
Ix'
I
1
120
180
FIG. 1. Frequency of occurrence of various ranges of dihedral angles determined by x-ray crystallography for (A) the values of X(SS-CC)I of model disulfides, (B) the values of Ix2(CaCC-SS)I of cystine residues in eight proteins, (C) the values of xl I of cystine and cysteine residues in 26 proteins, and (D) and (E), respectively, the values of x I and x21 for the leucine residues in 26 proteins.
the correlations proposed earlier (13-15), whereby aliphatic disulfides with T conformations about one or both C-S bonds were thought to have values of v(S-S) of about 525 and 540 cm-, respectively. The finding that all these conformations have nearly the same value of v(S-S), despite their widely different values of x(SS-CC), rules out vibrational coupling between y(CCS) and v(S-S) modes of the kind proposed (13, 14, 16) as an explanation for the existence bf rotational isomers with v(S-S) bands of different frequency. If such vibrational coupling existed, the values of v(S-S) of the five compounds mentioned above would not have been the same. Furthermore, conclusions regarding the frequencies of the various rotamers, drawn from normal mode analyses (15, 16) that were based upon the existence of such coupling interactions, are equally invalid; this is indicated by their erroneous prediction that T rotamers should have higher values of v(S-S) than G rotamers. The above results (combining Raman and x-ray data on crystalline dithioglycolic acid) clearly indicate that the T rotamer is not responsible for the 525 cm-1 band in methyl ethyl disulfide. Unfortunately, the above studies were unable to establish the identity of the rotamer responsible for this band. Yokozeki and Bauer (10) attempted to elucidate the struc-
Chemistry:
Van Wart and Scheraga
tures of the rotamers present about the C-S bond of methyl ethyl disulfide in the gas phase using the electron diffraction technique. Because of the coexistence of at least three rotamers, it was not possible to determine the structures of the rotamers present uniquely, and an extremely large number of possible solutions was possible. The authors were able to conclude, however, that there exists a substantial fraction of the T form. The potential function for rotation about the C-S bonds of several alkyl disulfides was investigated using the CNDO/2 (complete neglect of differential overlap) approximate molecular orbital method (3, 4). It was hoped that the results of these calculations would provide insight into the conformations of the rotamers present. The potential functions obtained were characterized by minima at the C and T positions. The appearance of the C minima was unexpected, and an analysis (4) of the nonbonded contact distances in C conformations between the S atoms and CH groups in the 1,4 positions across the C-S bonds revealed that the calculations resulted in contacts shorter than those observed experimentally. This result, taken together with the known tendency of the CNDO/2 method to underestimate overlap repulsions, led to the suggestion (4) that the C minima were probably artifacts, and that the true minima were probably somewhere between the C and G conformations (at about 300) where more normal CH.. S contact distances occur. The CNDO/2 calculations did not reveal the expected minimum in the region of the B conformation (Fig. IA), presumably because it was overshadowed by the overestimation of the stability of the C conformation. While the presence of C minima in the calculated potential functions was considered to be an artifact, it seemed that the calculations might be qualitatively correct in that they pointed to real, though weak, attractive nonbonded interactions between S atoms and CH groups in the 1,4 positions. This led to a literature search in which an unusually large number of short nonbonded contacts between CH groups and S atoms were revealed (4), supporting the existence of a 1,4 carbon-sulfur interaction and strengthening the idea that a rotamer with a value of X(SS-CC) between the C and G regions-the A rotamer-existed. The conformations about the C-S bonds of model disulfides in crystals are all found to be in the lower energy B and T conformations. It was thought that there might be a chance of observing higher energy rotamers about the C-S bonds of cystine residues in protein crystals where long-range intramolecular interactions might provide sufficient energy to populate higher energy minima. The conformations about the C-S bonds of all the cystine residues in eight proteins (6) whose coordinates were available at the time were examined (6) and the histogram§ shown in Fig. lB was constructed. As expected from the crystal conformations of smaller disulfides (Fig. 1A), there are peaks in the B and T regions. In addition, however, a third peak near 30'-due to the A conformation-is apparent. On the basis of these data, it was suggested (6) that this was the "missing" rotamer that was responsible for the v(S-S) band near 525 cm-' in the Raman spectra. Accordingly (6), conformations of the CCSSCC moiety with an A conformation about one C-S bond and either a B or T conformation about the other have values of v(S-S) of about 525 cm-I, while the presence of A conformations about both C-S bonds results in a value of v(S-S) of about 540 cm-1 It should be pointed out that A conformations are observed in crystals (4) of small aromatic disulfides (namely, those in which the S atoms are bonded directly to phenyl rings). Furthermore, the v(S-S) band of crystalline diphenyl disulfide (in which there are A conformations about both C-S bonds) lies at
Proc. Natl. Acad. Sci. USA 74 (1977)
15
about 540 cm-1, in agreement with the value predicted for aliphatic disulfides (6). In solution, diphenyl disulfide has v(S-S) bands at about 510, 525, and 540 cm-l corresponding to molecules with A conformations about neither, one, and both C-S bonds, respectively (6). Although there is no reason to expect a prior that the dependence of the values of v(S-S) on conformation about C-S bonds in aliphatic and aromatic disulfides should be the same, the agreement found above for diphenyl disulfide (using the proposed correlation) suggests that this may be the case. In principle, one could examine the Raman spectra of all proteins having cystine residues with A conformations about their C-S bonds to see if their presence correlates with Raman bands at about 525 and 540 cm-1. In practice, such a comparison requires an extremely high quality protein Raman spectrum. For lysozyme, a protein for which a high resolution spectrum in the v(S-S) region was available, this comparison (6) was made, and the frequencies and intensities of the bands observed in the v(S-S) region were found to be consistent with the proposed correlation. The reliability of the histogram shown in Fig. IB, in pointing to the existence of A conformations, is difficult to establish unambiguously. While the errors in the dihedral angles obtained from x-ray data on proteins are generally larger than for small molecules, the S atoms are usually among the best resolved atoms and, hence, the values of X(SS-CC) are among the most accurately known. Even assuming an error for each value of X(SS-CC) as large as 200, it seems unlikely that random error would result in the presence of the well-defined peak in the A region. To gain some insight into how such a histogram, derived from protein x-ray data, would appear for a side-chain rotational angle whose energy minima are better understood, similar histograms for the values of xI and x2 of leucine (in which there are heavy atoms, but not sulfur atoms, in the 1,4 positions across C-C bonds) are shown in Fig. ID and E. The data set included all the leucine residues in the twenty proteins used in ref. 18 plus those of the A and B chains of horse deoxy (19) and methemoglobin (A. D. McLachlan, personal communication to the Brookhaven Protein Data Bank), rubredoxin (20), a-chymotrypsinogen (21), dogfish lactate dehydrogenase (22), and trypsin (R. Stroud, personal communication). Leucine was chosen because it has a large hydrocarbon side chain whose conformations are well understood (i.e., its values of xI and x2 are expected to cluster in only the gauche and trans regions). Furthermore, since it is a residue generally found in the protein interior, its atomic coordinates are less likely to be influenced by thermal motion than those of exterior residues, such as lysine or glutamic acid. As can be seen from Fig. ID and E, the expectation that peaks in only the G and T regions occur is borne out, and no new or unexpected conformations are indicated. Hence, these histograms reliably reflect the identity of the rotamers present. Unfortunately, since few of the proteins whose structures are known from x-ray studies contain cystine residues, the number of occurrences in Fig. 1B is smaller than in Fig. iD and E and, hence, must be considered somewhat less reliable. Recently, Allinger and coworkers (23) calculated the potential function for rotation about the C-S bond of methyl ethyl disulfide using a semiempirical molecular mechanics method. They found energy minima in only the B, B', and T regions and not in the A region. The B and T conformations were of nearly equal energy, while the B' conformation was higher in energy by about 0.6 kcal/mol (1 cal = 4.184 J). The potential was found to be very soft, with energies less than about 1 kcal/mol for all
16
Chemistry: Van Wart and Scheraga
values of x(SS-CC), except those close to the C form. Clearly, the results derived from any semiempirical calculation reflect solely those interactions that are "built-in" in the parameterization. Since the authors admittedly (23) made no effort to take into account the existence of the 1,4 carbon-sulfur interaction, thought (4,6) to stabilize the A conformation (which could have been done, e.g., by computing the crystal structures of those organosulfur molecules that show short CH... S contacts), they pointed out that their failure to find the A conformation did not necessarily mean that it did not exist. Indeed, the fact that their calculated potential in the region of the A conformation is so soft indicates that, had a small stabilization energy been included in that region, such as that from the proposed (4) 1,4 carbon-sulfur interaction, a local energy minimum capable of accounting for an A rotamer population of 20% could easily result. Allinger et al. (23) also calculated the relative energies of the minimum- and maximum-energy conformations (a total of six values) in the potential function obtained from the molecular mechanics calculated using a minimum basis set ab initio molecular orbital method. It was hoped that this method would reveal any unexpected electronic effects in a 1,4 carbon-sulfur interaction without having to be parameterized for it. The authors (23) found that these ab initio calculations did not provide any evidence for the existence of such an interaction. Unfortunately, their ab initio calculations completely omitted the very region in which the A conformations are proposed (6) to exist (about 300); the closest conformations examined had values of x(SS-CC) of 00 and 720. For this reason, these calculations cannot decide anything about the existence of A conformations involving 1,4 carbon-sulfur interactions. Because of the inherently weak nature of any attractive 1,4 carbonsulfur interaction, only a thorough examination of the A region of the C-S potential function, using a large basis set ab initio technique in which geometries are optimized at each conformation, could be trusted to reflect the true situation. For economic reasons, such calculations have proven to be beyond present means. The molecular mechanics calculations discussed above are valuable in that they provide an idea of the relative stability of the B and T rotamers and in that they predict the presence of the B' rotamer. However, in spite of the claims by the authors (23), these calculations are not in agreement with the experimental data. Since only B, B', and T minima are found in these calculations, the presence of the 525 cm-' band in the Raman spectrum of methyl ethyl disulfide cannot be accounted for; it was shown earlier (6) that this band cannot be due to any of these three conformers. Furthermore, the calculations are not compatible with either the existence of A conformations about the C-S bonds of cystine residues in proteins (shown in Fig. 1B) or the multitude of experimentally observed (4) short nonbonded CH ...S contacts. Finally, while the calculations are consistent with the electron diffraction data (10), so are an infinite number of other mixtures of rotamers that include the A conformation. In summary, the available experimental and theoretical evidence appears to be consistent with a rotamer population about the C-S bonds of aliphatic disulfides of about 20% A, 40% B+B', and 40% T. This distribution is consistent with the electron diffraction and Raman data, the histogram in Fig. 1B, and the combined results of the CNDO/2 and molecular mechanics calculations. While it would be highly desirable to find examples of A conformations in crystals of smaller aliphatic disulfides (instead of only aromatic disulfides and cystine residues in protein crystals) whose Raman spectra could be studied directly,
Proc. Natl. Acad. Sci. USA 74 (1977)
this has not proved possible so far, presumably because these are higher-energy rotamers. Hence, allowing for the possibility that the histogram in Fig. IB, for one reason or another, is misleading, all of the evidence so far supporting the existence of A conformations is somewhat indirect. On the other hand, if one is to argue that A conformations do not exist, then the presence of the 525 cm-' band in the Raman spectra of aliphatic disulfides must still be accounted for by other than the G, G', B, B', or T rotamers. Hopefully, other techniques, such as microwave or nuclear magnetic resonance spectroscopy, may be able to provide more detailed information about the rotamers present. Rotation about C-C bonds We now consider the possible existence of 1,4 X,S interactions in the XCCSSCCX unit and the effect that such interactions might have on the potential function for rotation about C-C bonds. While attractive X... S interactions are possible for a variety of groups, discussion will be limited specifically to the case where X is NH, since attractive interactions between these groups in the form of NH .. S hydrogen bonds are known (24) to exist. However, the results for this group can be generalized to other groups (such as OH, etc.) capable of such attractive interactions with S atoms. The case where X is NH arises for cystine residues in proteins. One might expect attractive 1,4 NH-.. S interactions to influence the potential function for rotation about the Ca-C0 bonds of cystine and cysteine. To see if this is the case, the distribution of conformations observed about the Ca-C, bonds of all of the cystine and cysteine residues in 26 proteins (refs. 18-22; A. D. McLachlan, personal communication to Brookhaven Protein Data Bank and R. Stroud, personal communication) has been examined. The results are shown in the histogram in Fig. iC, in which the frequency of occurrence of various ranges of values (in 10° increments) of Xl(NCa-COS) is plotted. The peaks in this histogram fall in the G and T regions only, with the broad peak in the G region [X(NCa-C0S) = 60° ± 60° ] accounting for 78% of all occurrences, and the broad peak in the T region [X(NCa-C0S) = 180° ± 60°] accounting for the remaining 22%. The histogram in Fig. IC differs from that in Fig. 1B in that there is no peak in the A region. Hence, whereas 1,4 CH... S interactions across the C8fS bonds of cystine residues result in A conformations, 1,4 NH ...S interactions across the Ca-Cbonds of cystine and cysteine residues do not result in A conformations. Instead, the distribution of conformations observed about the Ca-C/3 bonds of cystine and cysteine residues qualitatively resembles that found for leucine (Fig. iD). Quantitatively, however, the distribution for cystine and cysteine favors the G region over the T (78:22) to a somewhat greater extent than for leucine (67:33). The differences between the effects of 1,4 CaH .. S interactions on the C0-S potential function and of 1,4 NH-.. S interactions on the Ca-Co potential function can be understood by considering the differences in the inherent structural features of the CaCCSS and NCaC'CS units. For example, an essential structural feature that will influence the potential function for rotation about a given bond is the bond length. Considering Ca-S bonds (1.83 A)
