Minimum Energy Conformations of Proline-Containing Helices ALEXANDER POLINSKY,’ MURRAY G O O D M A N , ’ KAREN A. WILLIAMS,2 and CHARLES M. DEBER2,*
’ Department of Chemistry,
University of California-San Diego, La Jolla, California 92093, USA; *Research Institute, Hospital for Sick Children, Toronto M5G 1X8, and Department of Biochemistry, University of Toronto, Toronto M5S 1A8, Ontario, Canada
SYNOPSIS
Proline occurs frequently in transmembrane a-helices of transport and receptor proteins even though statistical surveys demonstrate the overwhelming preference of this residue for a non-a-helical, hydrophilic environment. As a result, membrane-buried proline has been proposed to be functionally important, with function arising from structural discontinuity or destabilization of the helix. Destabilization may occur by Pro-mediated conformational transitions between discrete states, and may be manifested in membrane protein systems through reversible processes such as channel opening and closing or signal transduction. In this study, computer modeling of a model transmembrane a-helix, ( Ala)8-LeuPro-Phe-(Ala)8,in a medium of low polarity (dielectric = 2 ) , is used to examine the occurrence and energetic accessibility of Pro-mediated conformational interconversions. Leu I)and xl,Pro I), and Phe 4 and X1 torsion angles were assigned random values so that a data base of 200 conformations for each of the cis and trans states was generated. The conformations were minimized and low-energy structures organized into families. This analysis demonstrated that the most populated lowest energy family is the Trans-I conformation, corresponding to proline in a kinked a-helix. Two additional trans structures, Trans-I1 and Trans-111, as well as a cis conformation, Cis-I, are also energetically competitive. Interconversions between the trans states could thus be mediated by changes at a single torsion angle, accompanied by minor local hydrogen-bonding rearrangements. This work substantiates that membrane-buried proline can provide the basis for conformational transitions between discrete a-helix-based structures in a nonpolar environment.
INTRO DUCT10 N Proline residues are widely observed in the putatively a-helical transmembrane ( T M ) segments of many integral membrane proteins that function as receptor subunits or transporters, even though this residue favors a hydrophilic environment’ and tends not to occur within the a-helices of globular proteins?,* Pro may cause structural discontinuity in an a-helix by virtue of the pyrrolidine ring, which sterically hinders adjacent residues, and also limits the dihedral angle 4 to an a-helical value (-60°) * To whom correspondence should be sent at the Research Institute. Biopolymers, Vol. 32, 399-406 (1992) 0 1992 John Wiley & Sons, Inc.
CCC 0006-3525/92/040399-08$04.00
and leaves the 4 angle flexible but hindered. Further contributing to destabilization is the absence of a proton on the tertiary nitrogen and the increased backbone polarity a t the X-Pro secondary amide.6,7 These characteristics are manifested as structural and/or dynamic roles where the function of a particular TM Pro depends upon the local context and the requirements of the parent protein.’ Destabilization facilitates conformational transitions that may constitute the basis for dynamic processes such as channel “opening and closing” or signal transduction. Conformational transitions between discrete states may occur via cis-trans isomerization of the X-Pro bond (Figure 1) or through more subtle structural modifications. The cis conformation of the X-Pro bond, unlike that of all other peptide bonds 399
400
POLINSKY ET AL. R
0
I
II
CH-C -NAu
H
'/
'N-C
I
/
H
6CH2
R
I
C-
\a $CH2
- N I
H
Y
cis
No
6
'N-CH
/
\ 2
H-Ca -C
'CH(
trans
4" -
YCH2
'CH(
II
B
0
Figure 1. Isomeric trans and cis states of the X-Pro peptide bond. The structures are related by 180° rotation about the C-N bond.
in mammalian proteins, is energetically accessible t o c y ~ l e . ' ~The - ~ ~ facility for conformational interdue to relative destabilization of the trans conforconversion is also suggested by 13C-and 'H-nmr exmation? Cis X-Pro bonds have been observed in periments which study the transfer of the synthetic globular proteins, where they tend to occur in peptide, Boc- ( Ala)3-Leu-Pro-Phe-OH, from water several patterns of local sequence.z,lzIsomerization to a lipid micelle en~ironment.'~The consensus has been shown to be involved in the slow phase of triad, Leu-Pro-Phe, typifies those that frequently surround membrane-buried Pro residues in transprotein foldingl3 and in giving rise to conformational port proteins.' This transfer elicits chemical shift heter~geneity.'~.'~ Interest in cis-trans X-Pro isomchanges which were interpreted as arising from a erism has heightened since it was demonstrated that membrane-induced y-turn involving the Leu-Procyclophilin, the protein to which the immunosupressant drug cyclosporin A binds, is a P P I ~ S ~ . " , ~ Phe ~ triad.23 If the observed selective inclusion of Pro residues Isomerization of an X-Pro bond, within or adjacent in transmembrane segments of many multispanning to a membrane, could alter the alignment of a segmembrane proteins has a dynamic-rather than a ment, and thereby provide the conformational basis purely static-structural role, then local membranefor a reversible regulatory mechanism for channel buried Pro-containing segments should have the inopening and closing, although rates of isomerization herent propensity to exist in multiple conformamay be limiting." The occurrence of isomerization tional states of competitive minimum energy. In the at a membrane-buried Pro has not yet been dempresent work, we show that this is indeed the case, onstrated; experiments employing site-directed muthrough analysis of the conformations accessible to tagenesis offer indirect evidence that it does not oca model transmembrane segment, (Ala)*-Leu-Procur on the basis of the ease with which Pro could Phe-(Ala)*, in an environment of low dielectric be m ~ t a t e d . ' ~ - ' ~ constant. A preliminary account of this work has Pro-mediated conformational transitions need appeared?* not involve isomerization, however, since interconversions between discrete all-trans states may be mediated by transitions between favored states a t MATERIALS A N D METHODS the hindered Pro $ angle. This "flexibility" stems from Pro-induced destabilization of an a-helix and The calculations were carried out on an IRIS 4Dthe increased basicity of the N-adjacent carbonyl 25TG workstation using the QUANTA and oxygen, which increases its propensity for hydrogen CHARMm 2.1 programs of Polygen." All simulabonding (e.g., participation in the 10-membered 0tions were carried out using a dielectric constant of turnzz and 7-membered y-turn structure^^^). Evi2 to mimic a membrane environment. The consensus dence for Pro-mediated conformational flexibility is triad, Leu-Pro-Phe, links two a-helical regions with obtained from the well-characterized integral memPro as the major helix-disturbing residue. Throughbrane protein, bacteriorhodopsin (bR), the lightout the search, Leu r$ and Phe $ torsions were kept driven proton pump of the purple membrane of H. a t a-helical values, and only Leu $ and X I , Pro $ halobium, 24 which has three highly conserved TM and Phe r$ and x1 were varied. Pro residues ( Proso,Progl,Pro1%).Site-directed TM Pro mutants and 15N-labeled Pro residues of bR Conformational Search studied with Fourier transform ir (FTIR) spectrosTwo hundred conformations were generated by ascopy demonstrated that one or more of the TM Pros signing random values to Leu $ and x l , Pro $, and undergoes a structural change during the bR pho-
'
MINIMUM ENERGY CONFORMATIONS
Phe and XI torsions. After an initial minimization of each structure ( 5 0 steps of steepest descent) designed to remove obvious steric overlaps, structures with energies up to 100 kcal higher than the lowest energy structure (95 structures with trans and 107 with cis) were selected for full minimization with the adopted basis Newton-Rafson algorithm. Lowenergy structures with energies up to 15 kcallmol higher than the lowest energy structure were then selected for cluster analysis (36 with trans and 41 with cis). Cluster Analysis of Backbone Conformations
The conformations were divided into families (clusters) based on the similarity of the backbone conformations for the Leu-Pro-Phe triad. The cluster analysis was carried out by first designating the lowest energy conformation from the entire set of structures as the nucleus for the first cluster. The remaining structures were then compared to the nucleus; if none of the backbone dihedral angles differed from the corresponding angles in the nucleus by more than a specified threshold value, that structure was added to the cluster. After all structures were compared, the lowest energy conformation remaining was extracted and used as the nucleus of the second cluster. This process was continued until all structures were assigned to clusters. The algorithm used differed from that supplied with the standard QUANTA package in that instead of calculating the overall rms difference in torsions between two conformations, each torsion was compared individually. The threshold value was determined empirically such that the standard deviation of any of the backbone dihedral angles of the structures comprising each cluster did not exceed 15”. A value of this magnitude is typically observed in molecular dynamics simulations for molecules fluctuating around stationary states between conformational transitions.
401
RESULTS AND DISCUSSION Strategies for the Search
A systematic search with even coarse 60” increments of the backbone torsional angles and 120’ increments of the XI angles would require generation and minimization of 1944 structures for each cis and trans conformation of the Leu-Pro peptide bond. To decrease the extent of calculations involved, we used a random search to sample the conformational space during which 200 structures were generated with the Leu-Pro-Phe triad in random conformations for each cis and trans conformation of the Leu-Pro amide bond. After the minimization we found that a number of conformations, though being slightly different, possessed similar energies. Thus a question arises as to which of these conformations is most probable. We believe for a molecule undergoing thermal motion, the term “preferred conformation” can refer to a set, or family, of closely related structures differing from each other only by minor changes in torsional angles. To define these families among the low-energy conformations obtained, we used cluster analysis (Tables I and 11). The variation of the torsional angles allowed within one family was set at 15”.The families in Tables I and I1 thus represent an average dynamic structure rather than one static “preferred” conformation. The populations of various families differ significantly. Before any suggestion can be made explaining these differences, the statistical validity of the results has to be checked, i.e., are a sufficient number of random conformations being generated and minimized to provide a convergence in the family populations? Figure 2 shows the dependence of family populations ( in % ) on the sample size. As expected, the populations fluctuate sharply when the sample size is small, but converge once the sample size reaches 75-100 structures, as may be anticipated because variations of backbone torsional angles of
Table I Mean Values and RMS Deviations (in Parentheses) for the Stable Conformations with the trans Leu-Pro Peptide Bond
Family
Number of Structures
min E (kcal/mol)
Leu 6
Leu J/
Pro J,
Trans-I Trans-I1 Trans-I11
20 9 7
0 6.9 11.0
-47 (4) -59 (4) -51 (1)
-57 (7) 123 (15) -46 (2)
-45 (7) -41 (4) 70 (11)
Phe -62 (10) -56 (2) -46 (4)
Phe J, -43 (8) -50 (4) -40 (3)
402
POLINSKY ET AL.
Table I1 Mean Values and RMS Deviations (in Parentheses) for the Stable Conformations with the cis Leu-Pro Peptide Bond
Family
Number of Structures
min E (kcal/mol)
1 (Cis-I) 2 3 4 5 6 7 8 9
10 5 4 3 3 2 2 2 2
12.6 13.3 19.0 17.4 17.9 11.7 14.5 18.2 20.9
Leu I+?
Pro b,t
Phe dJ
150 (12) -47 (2) 152 (5) -39 (8) -48 (4) -33 (5) 115 (4) 132 (30) 172 (1)
-42 (11) 133 (8) -46 (4) -48 (10) 129 (14) 178 (2) 15 (1) 54 (3) 168 (1)
-50 (11) -57 (3) -170 (4) 52 (15) 33 (7) -47 (3) -70 (10) 67 (3) -45 (1)
Leu dJ -67 -54 -62 -55 -49 -53 -78 -82 -66
(4) (8) (3) (6) (3) (4) (4) (14) (1)
only three residues (with one of them being Pro) were analyzed in the cluster analysis procedure. We used a sample size of 200 conformations that is far beyond these values, and accordingly, the family populations found in our search prove to be statistically valid. In this analysis, the population differences may be indicative of the shape of the potential energy surface. The greater number of randomly selected starting conformations that converge to a certain family of local minima, the wider is the “valley” on the energy surface. Given that the minimum energies of two valleys are close, the wider valley may correspond to the conformational state of higher entropy. This conformational state, therefore, has higher statistical weight since it has lower free en-
0
-
so
-64 -48 -54 -80 -88 -45 -65 -78 -49
(4)
(6) (5) (5) (8) (2) (7) (6) (1)
ergy. Since the shape of the potential energy surface is not known, no quantitative relationship between family populations and relative entropy can be derived. However, the population comparison may prove useful as a qualitative criterion for choosing the most probable conformational states of the molecule. Minimum Energy Structures of (Ala),-Leu-ProPhe-( Ala)8
The conformational search for low-energy structures containing a trans Leu-Pro bond suggests the presence of three families (Table I ) . The preferred structure for the peptide corresponds to an a-helix with a kink about the Pro residue (Trans-I, Figure
Trans-Ill
Y l
Phe I+?
100
150
200
Sample size Figure 2. The dependence of family populations (in % ) on the sample size for the families with a trans Leu-Pro peptide bond.
MINIMUM ENERGY CONFORMATIONS
Tms-1
403
Trans-II
Figure 3. Backbone conformations corresponding to the most stable families.
3 ) . Families I1 and I11 represent alternative structures that the molecule can assume upon conformational transition ( Trans-I1 and Trans-111,Figure 3 ) . They are less populated and possess higher energies than the Trans-I family. It is interesting to note that all three trans conformations differ essentially in only one torsion angle. This further suggests the possibility of interconversion between the structures. The increased stability of the Trans-I family over the other two tram families can be attributed to the retention of the a-helical hydrogen-bond pattern even though some of the hydrogen bonds are bifurcated due to distortion of the helix. The Trans-I1 structure has a slightly distorted type I11 0-turn around Pro and Phe, where again the hydrogen bond is bifurcated because of &-helical distortion. The Trans-I11 conformation has a y-turn involving Pro, with a hydrogen bond between the Phe NH and Leu CO. The conformational search for the cis Leu-Pro peptide bond (Table 11) reveals multiple families of conformations. Although an unambiguous choice can not be made regarding the most probable conformation, we selected the family Cis-I (Figure 3 ), which is comparable to Trans-I1 and Trans-I11 in both population and energy.
The four families of structures (Trans-I, -11, -111; Cis-I) selected on the basis of low energy and high population were subjected to optimization of the side chains in the Leu and Phe residues. The final structures (torsion angles and hydrogen bonds) are presented in Table 111. The Leu-Pro-Phe region of the peptide for each of these conformations is shown in Figure 4.
Conformational Transitions Among low-Energy Structures
Conformational interconversion between discrete states requires that each of the alternative conformations be of low energy, that the energy of activation of the transition not be limiting, and that the magnitude of structural change be appropriate for the requisite function. A comparison of relative energies for the similarly populated Trans-11,Trans111, and Cis-I conformations suggests that Trans-I1 is the structure most likely to be assumed upon conformational transition from the a-helical Trans-I structure since it has the lowest energy. It should be noted, however, that relative energies calculated in vacuo (even with the dielectric constant corrected to 2 to reflect the hydrocarbon interior of the mem-
404
POLINSKY ET AL.
Table I11 Backbone Torsion Angles and Hydrogen Bonds for the Lowest Energy Structures with Optimized Conformations of Side Chains
E Structure Trans-I Trans-I1 Trans-I11 Cis-I
(kcal/mol)
Leu 4
Leu$
Pro$
Phe 4
Pheq
0
-50.6
-53.1
-39.4
-72.9
-38.3
6.9 11.0 12.6
-58.0 -53.3 -65.6
116.0 -47.7 155.7
-42.9 59.9 -48.1
-56.2 -43.1 -52.4
-53.4 -42.6 -71.8
brane) may differ from energies in the real environment. As a result, we prefer not to exclude any of these structures on the basis of difference in energy only. Our calculations provide no direct information concerning the energy of activation ( E a ) ,which would characterize any of the possible transitions, nor the mechanism through which substrate or ligand activation might promote the transition by reducing this value during the functioning of a membrane protein. It seems reasonable to assume that the change in shape of a transmembrane segment during a con-
H Bonds CO (Leu) - NH (Ala12, Ala13); CO (Ala8) - NH (Phe, Ala12) CO (Leu) - NH (Ala12, Ala 13) CO (Leu) - NH (Phe) CO (Ala8) - NH (Ala12, Ala 13)
formational transition associated with membrane protein function should not be so dramatic as to disturb adjacent lipids. Although the overall shape of the structures appears similar (Figure 3 ) , they differ in the direction and magnitude of the a-helical bend. Figure 5 schematically shows all four structures with the Ala ( 1)-Ala (8)regions superimposed. Trans-I and Cis-I are bent in the same direction, with Cis-I having a larger bend angle. Trans-I1 and Trans-I11 bend in a direction essentially opposite that of Trans-I. Dynamic conformational interconversions of precisely this sort could be envisioned
Figure 4. The Leu-Pro-Phe region of the lowest energy conformations in various families.
MINIMUM ENERGY CONFORMATIONS Trans-II
405
This work was supported, in part, by grants to MG from the National Institutes of Health GM18694, and to CMD from the Natural Sciences and Engineering Research Council of Canada (NSERC). KAW holds an NSERC studentship.
REFERENCES
Figure 5. Relative spatial arrangement of the N-terminal and C-terminal helices in the most stable families.
as the basis for the opening and closing of a transmembrane channel to allow access to substrates or provide the molecular basis for propagation of signaling events. The transition from Trans-I to CisI will not necessarily be easier than that to TransI1 or Trans-111,however, because a-helical fragments can rotate around the helical axis within a membrane. Simultaneous variation of the appropriate torsion angle ( s ) and rotation around the helical axis can conserve the overall shape of the molecule during the transition from Trans-I to any of the three alternatives.
CONCLUSIONS The present work substantiates that energetically accessible structures are available that provide the molecular basis for local conformational transitions about Pro sites in an environment of low dielectric constant. These transitions may produce conformational fluctuations through rearrangement of local H-bonding patterns (e.g., from a-helical to yturn-like structures) while all-trans peptide bonds are preserved, or through trans-cis interconversion of X-Pro peptide bonds. Both cis and trans X-Pro peptide bonds have been observed experimentally in solvents of low p ~ l a r i t y ,while ~' y-turn structures have been deduced for model Pro-containing peptides in membrane-mimetic environment^.^^ Further exploration of these phenomena awaits experimental determination of the actual conformation ( s ) of XPro peptide bonds within membrane proteins in both resting (e.g., bacteriorh~dopsin~~ ) and activated states. Note added in proof. In computer modelling studies on a Pro-containing helix, Yun et al. (ref. 32) similarly observed a variety of hydrogen-bonding schemes.
1. Brandl, C. J. & Deber, C. M. (1986) Proc. Natl. Acad. Sci. USA 83,917-921. 2. Richardson, J. S. & Richardson, D. C. (1989) in Pre-
diction of Protein Structure and Principles of Protein Conformation, Fasman, G. D., Ed., Plenum Press, New York, pp. 1-99. 3. O'Neil, K. T. & DeGrado, W. F. (1990) Science 2 5 0 , 646-65 1. 4. Chou, P. Y. & Fasman, G. D. (1978) Adu. Enzymol. 4 7 , 45-148. 5. Carver, J. P. & Blout, E. R. (1967) in Treatise on Collagen, Vol. 1, Academic Press, New York, pp. 441526. 6. Rose, G. D., Geselowitz, A. R., Lesser, G., Lee, R. & Zehfus, M. H. (1985) Science 229,9834-9838. 7. Veis, A. & Nawrot, C. F. (1970) J. Am. Chem. SOC. 9 2 , 3910-3914. 8. Williams, K. A. & Deber, C. M. (1991) Biochemistry 30,8919-8923. 9. Zimmerman, S. S. & Scheraga, H. A. ( 1976) Mucromolecules 9 , 408-416. 10. Schmid, F. X., Grafl, R., Wrba, A. & Beintema, J. J. (1986) Proc. Natl. Acad. Sci. USA 83,872-876. 11. Stewart, D. E., Sarkar, A. & Wampler, J. E. (1990) J. Mol. Biol. 2 1 4 , 253-260. 12. Frommel, C. & Preissner, R. (1990) FEBS Lett. 277, 159-163. 13. Brandts, J. F., Halvorson, H. R. & Brennan, M. (1975) Biochemistry 14,4953-4963. 14 Evans, P. A., Dobson, C. M., Kautz, R. A., Hatfield, G. & Fox, R. 0. (1987) Nature 3 2 9 , 266-268. 15. Kikuchi, M., Yamamoto, Y., Taniyama, Y., Ishimau, K., Yoshikawa, W., Kaisho, Y. & Ikehara, M. (1988) Proc. Natl. Acad. Sci. USA 8 5 , 9411-9415. 16. Takahashi, N., Hagano, T. & Suzuki, M. (1989) Nature 337,473-475. 17. Fischer, G., Wittmann-Liebold, B., Lang, K., Kiefhaber, T. & Schmidt, F. X. (1989) Nature 3 4 1 , 755757. 18. Galardy, R. E., Alger, J. R. & Liakopoulou-kyriakides, M. (1982) Znt. J . Peptide Protein Res. 1 9 , 123-132. 19. Ahl, P. L., Stern, L. J., During, D., Mogi, T. & Rothschild, K. J. (1988) J . Biol. Chem. 263,13594-13601. 20. Mogi, T., Stern, L. J., Chao, B. H. & Khorana, H. G. (1989) J. Biol. Chem. 264, 14192-14196. 21. Consler, T. G., Tsolas, 0. & Kaback, H. R. (1991) Biochemistry 3 0 , 1291-1298. 22. Smith, J. A. & Pease, L. G. (1980) CRC Crit. Rev. Biochem. 8 , 315-399.
406
POLINSKY ET AL.
23. Deber, C. M., Glibowicka, M. & Woolley, G. A. (1990) Biopolymers 29, 149-157. 24. Khorana, H. G. (1988) J. Biol. Chem. 263, 74397442. 25. Rothschild, K. J., He, Y.-W., Gray, D., Roepe, P. D., Pelletier, S. L., Brown, R. S. & Herzfeld, J. ( 1989) Proc. Natl. Acad. Sci. USA 86,9832-9835. 26. Rothschild, K. J., He, Y.-W., Mogi, T., Marti, T., Stern, L. J. & Khorana, H. G. (1990) Biochemistry 29,5954-5960. 27. Gerwert, K., Hess, B. & Engelhard, M. (1990) FEBS Lett. 277,159-163. 28. Deber, C. M., Polinsky, A. & Goodman, M. (1991) in Peptides 1990,Giralt, E. & Andreu, D., Eds., ESCOM
29.
30. 31. 32.
Science Publishers, Leiden, The Netherlands, pp. 485487. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S. & Karplus, M. (1983) J. Comp. Chem. 4,187. Deber, C. M., Madison, V. & Blout, E. R. (1976) Acct. Chem. Res. 9,106-113. Deber, C. M., Sorrell, B. J. & Xu, G.-Y. (1990) Biochem. Biophys. Res. Commun. 172,862-869. Yun, R. H., Anderson, A. & Hermans, J. (1991) Proteins: Struct. Func. & Gen. 10, 219-228.
Receiued J u n e 10, 1991 Accepted July 24, 1991
’ Department of Chemistry,
University of California-San Diego, La Jolla, California 92093, USA; *Research Institute, Hospital for Sick Children, Toronto M5G 1X8, and Department of Biochemistry, University of Toronto, Toronto M5S 1A8, Ontario, Canada
SYNOPSIS
Proline occurs frequently in transmembrane a-helices of transport and receptor proteins even though statistical surveys demonstrate the overwhelming preference of this residue for a non-a-helical, hydrophilic environment. As a result, membrane-buried proline has been proposed to be functionally important, with function arising from structural discontinuity or destabilization of the helix. Destabilization may occur by Pro-mediated conformational transitions between discrete states, and may be manifested in membrane protein systems through reversible processes such as channel opening and closing or signal transduction. In this study, computer modeling of a model transmembrane a-helix, ( Ala)8-LeuPro-Phe-(Ala)8,in a medium of low polarity (dielectric = 2 ) , is used to examine the occurrence and energetic accessibility of Pro-mediated conformational interconversions. Leu I)and xl,Pro I), and Phe 4 and X1 torsion angles were assigned random values so that a data base of 200 conformations for each of the cis and trans states was generated. The conformations were minimized and low-energy structures organized into families. This analysis demonstrated that the most populated lowest energy family is the Trans-I conformation, corresponding to proline in a kinked a-helix. Two additional trans structures, Trans-I1 and Trans-111, as well as a cis conformation, Cis-I, are also energetically competitive. Interconversions between the trans states could thus be mediated by changes at a single torsion angle, accompanied by minor local hydrogen-bonding rearrangements. This work substantiates that membrane-buried proline can provide the basis for conformational transitions between discrete a-helix-based structures in a nonpolar environment.
INTRO DUCT10 N Proline residues are widely observed in the putatively a-helical transmembrane ( T M ) segments of many integral membrane proteins that function as receptor subunits or transporters, even though this residue favors a hydrophilic environment’ and tends not to occur within the a-helices of globular proteins?,* Pro may cause structural discontinuity in an a-helix by virtue of the pyrrolidine ring, which sterically hinders adjacent residues, and also limits the dihedral angle 4 to an a-helical value (-60°) * To whom correspondence should be sent at the Research Institute. Biopolymers, Vol. 32, 399-406 (1992) 0 1992 John Wiley & Sons, Inc.
CCC 0006-3525/92/040399-08$04.00
and leaves the 4 angle flexible but hindered. Further contributing to destabilization is the absence of a proton on the tertiary nitrogen and the increased backbone polarity a t the X-Pro secondary amide.6,7 These characteristics are manifested as structural and/or dynamic roles where the function of a particular TM Pro depends upon the local context and the requirements of the parent protein.’ Destabilization facilitates conformational transitions that may constitute the basis for dynamic processes such as channel “opening and closing” or signal transduction. Conformational transitions between discrete states may occur via cis-trans isomerization of the X-Pro bond (Figure 1) or through more subtle structural modifications. The cis conformation of the X-Pro bond, unlike that of all other peptide bonds 399
400
POLINSKY ET AL. R
0
I
II
CH-C -NAu
H
'/
'N-C
I
/
H
6CH2
R
I
C-
\a $CH2
- N I
H
Y
cis
No
6
'N-CH
/
\ 2
H-Ca -C
'CH(
trans
4" -
YCH2
'CH(
II
B
0
Figure 1. Isomeric trans and cis states of the X-Pro peptide bond. The structures are related by 180° rotation about the C-N bond.
in mammalian proteins, is energetically accessible t o c y ~ l e . ' ~The - ~ ~ facility for conformational interdue to relative destabilization of the trans conforconversion is also suggested by 13C-and 'H-nmr exmation? Cis X-Pro bonds have been observed in periments which study the transfer of the synthetic globular proteins, where they tend to occur in peptide, Boc- ( Ala)3-Leu-Pro-Phe-OH, from water several patterns of local sequence.z,lzIsomerization to a lipid micelle en~ironment.'~The consensus has been shown to be involved in the slow phase of triad, Leu-Pro-Phe, typifies those that frequently surround membrane-buried Pro residues in transprotein foldingl3 and in giving rise to conformational port proteins.' This transfer elicits chemical shift heter~geneity.'~.'~ Interest in cis-trans X-Pro isomchanges which were interpreted as arising from a erism has heightened since it was demonstrated that membrane-induced y-turn involving the Leu-Procyclophilin, the protein to which the immunosupressant drug cyclosporin A binds, is a P P I ~ S ~ . " , ~ Phe ~ triad.23 If the observed selective inclusion of Pro residues Isomerization of an X-Pro bond, within or adjacent in transmembrane segments of many multispanning to a membrane, could alter the alignment of a segmembrane proteins has a dynamic-rather than a ment, and thereby provide the conformational basis purely static-structural role, then local membranefor a reversible regulatory mechanism for channel buried Pro-containing segments should have the inopening and closing, although rates of isomerization herent propensity to exist in multiple conformamay be limiting." The occurrence of isomerization tional states of competitive minimum energy. In the at a membrane-buried Pro has not yet been dempresent work, we show that this is indeed the case, onstrated; experiments employing site-directed muthrough analysis of the conformations accessible to tagenesis offer indirect evidence that it does not oca model transmembrane segment, (Ala)*-Leu-Procur on the basis of the ease with which Pro could Phe-(Ala)*, in an environment of low dielectric be m ~ t a t e d . ' ~ - ' ~ constant. A preliminary account of this work has Pro-mediated conformational transitions need appeared?* not involve isomerization, however, since interconversions between discrete all-trans states may be mediated by transitions between favored states a t MATERIALS A N D METHODS the hindered Pro $ angle. This "flexibility" stems from Pro-induced destabilization of an a-helix and The calculations were carried out on an IRIS 4Dthe increased basicity of the N-adjacent carbonyl 25TG workstation using the QUANTA and oxygen, which increases its propensity for hydrogen CHARMm 2.1 programs of Polygen." All simulabonding (e.g., participation in the 10-membered 0tions were carried out using a dielectric constant of turnzz and 7-membered y-turn structure^^^). Evi2 to mimic a membrane environment. The consensus dence for Pro-mediated conformational flexibility is triad, Leu-Pro-Phe, links two a-helical regions with obtained from the well-characterized integral memPro as the major helix-disturbing residue. Throughbrane protein, bacteriorhodopsin (bR), the lightout the search, Leu r$ and Phe $ torsions were kept driven proton pump of the purple membrane of H. a t a-helical values, and only Leu $ and X I , Pro $ halobium, 24 which has three highly conserved TM and Phe r$ and x1 were varied. Pro residues ( Proso,Progl,Pro1%).Site-directed TM Pro mutants and 15N-labeled Pro residues of bR Conformational Search studied with Fourier transform ir (FTIR) spectrosTwo hundred conformations were generated by ascopy demonstrated that one or more of the TM Pros signing random values to Leu $ and x l , Pro $, and undergoes a structural change during the bR pho-
'
MINIMUM ENERGY CONFORMATIONS
Phe and XI torsions. After an initial minimization of each structure ( 5 0 steps of steepest descent) designed to remove obvious steric overlaps, structures with energies up to 100 kcal higher than the lowest energy structure (95 structures with trans and 107 with cis) were selected for full minimization with the adopted basis Newton-Rafson algorithm. Lowenergy structures with energies up to 15 kcallmol higher than the lowest energy structure were then selected for cluster analysis (36 with trans and 41 with cis). Cluster Analysis of Backbone Conformations
The conformations were divided into families (clusters) based on the similarity of the backbone conformations for the Leu-Pro-Phe triad. The cluster analysis was carried out by first designating the lowest energy conformation from the entire set of structures as the nucleus for the first cluster. The remaining structures were then compared to the nucleus; if none of the backbone dihedral angles differed from the corresponding angles in the nucleus by more than a specified threshold value, that structure was added to the cluster. After all structures were compared, the lowest energy conformation remaining was extracted and used as the nucleus of the second cluster. This process was continued until all structures were assigned to clusters. The algorithm used differed from that supplied with the standard QUANTA package in that instead of calculating the overall rms difference in torsions between two conformations, each torsion was compared individually. The threshold value was determined empirically such that the standard deviation of any of the backbone dihedral angles of the structures comprising each cluster did not exceed 15”. A value of this magnitude is typically observed in molecular dynamics simulations for molecules fluctuating around stationary states between conformational transitions.
401
RESULTS AND DISCUSSION Strategies for the Search
A systematic search with even coarse 60” increments of the backbone torsional angles and 120’ increments of the XI angles would require generation and minimization of 1944 structures for each cis and trans conformation of the Leu-Pro peptide bond. To decrease the extent of calculations involved, we used a random search to sample the conformational space during which 200 structures were generated with the Leu-Pro-Phe triad in random conformations for each cis and trans conformation of the Leu-Pro amide bond. After the minimization we found that a number of conformations, though being slightly different, possessed similar energies. Thus a question arises as to which of these conformations is most probable. We believe for a molecule undergoing thermal motion, the term “preferred conformation” can refer to a set, or family, of closely related structures differing from each other only by minor changes in torsional angles. To define these families among the low-energy conformations obtained, we used cluster analysis (Tables I and 11). The variation of the torsional angles allowed within one family was set at 15”.The families in Tables I and I1 thus represent an average dynamic structure rather than one static “preferred” conformation. The populations of various families differ significantly. Before any suggestion can be made explaining these differences, the statistical validity of the results has to be checked, i.e., are a sufficient number of random conformations being generated and minimized to provide a convergence in the family populations? Figure 2 shows the dependence of family populations ( in % ) on the sample size. As expected, the populations fluctuate sharply when the sample size is small, but converge once the sample size reaches 75-100 structures, as may be anticipated because variations of backbone torsional angles of
Table I Mean Values and RMS Deviations (in Parentheses) for the Stable Conformations with the trans Leu-Pro Peptide Bond
Family
Number of Structures
min E (kcal/mol)
Leu 6
Leu J/
Pro J,
Trans-I Trans-I1 Trans-I11
20 9 7
0 6.9 11.0
-47 (4) -59 (4) -51 (1)
-57 (7) 123 (15) -46 (2)
-45 (7) -41 (4) 70 (11)
Phe -62 (10) -56 (2) -46 (4)
Phe J, -43 (8) -50 (4) -40 (3)
402
POLINSKY ET AL.
Table I1 Mean Values and RMS Deviations (in Parentheses) for the Stable Conformations with the cis Leu-Pro Peptide Bond
Family
Number of Structures
min E (kcal/mol)
1 (Cis-I) 2 3 4 5 6 7 8 9
10 5 4 3 3 2 2 2 2
12.6 13.3 19.0 17.4 17.9 11.7 14.5 18.2 20.9
Leu I+?
Pro b,t
Phe dJ
150 (12) -47 (2) 152 (5) -39 (8) -48 (4) -33 (5) 115 (4) 132 (30) 172 (1)
-42 (11) 133 (8) -46 (4) -48 (10) 129 (14) 178 (2) 15 (1) 54 (3) 168 (1)
-50 (11) -57 (3) -170 (4) 52 (15) 33 (7) -47 (3) -70 (10) 67 (3) -45 (1)
Leu dJ -67 -54 -62 -55 -49 -53 -78 -82 -66
(4) (8) (3) (6) (3) (4) (4) (14) (1)
only three residues (with one of them being Pro) were analyzed in the cluster analysis procedure. We used a sample size of 200 conformations that is far beyond these values, and accordingly, the family populations found in our search prove to be statistically valid. In this analysis, the population differences may be indicative of the shape of the potential energy surface. The greater number of randomly selected starting conformations that converge to a certain family of local minima, the wider is the “valley” on the energy surface. Given that the minimum energies of two valleys are close, the wider valley may correspond to the conformational state of higher entropy. This conformational state, therefore, has higher statistical weight since it has lower free en-
0
-
so
-64 -48 -54 -80 -88 -45 -65 -78 -49
(4)
(6) (5) (5) (8) (2) (7) (6) (1)
ergy. Since the shape of the potential energy surface is not known, no quantitative relationship between family populations and relative entropy can be derived. However, the population comparison may prove useful as a qualitative criterion for choosing the most probable conformational states of the molecule. Minimum Energy Structures of (Ala),-Leu-ProPhe-( Ala)8
The conformational search for low-energy structures containing a trans Leu-Pro bond suggests the presence of three families (Table I ) . The preferred structure for the peptide corresponds to an a-helix with a kink about the Pro residue (Trans-I, Figure
Trans-Ill
Y l
Phe I+?
100
150
200
Sample size Figure 2. The dependence of family populations (in % ) on the sample size for the families with a trans Leu-Pro peptide bond.
MINIMUM ENERGY CONFORMATIONS
Tms-1
403
Trans-II
Figure 3. Backbone conformations corresponding to the most stable families.
3 ) . Families I1 and I11 represent alternative structures that the molecule can assume upon conformational transition ( Trans-I1 and Trans-111,Figure 3 ) . They are less populated and possess higher energies than the Trans-I family. It is interesting to note that all three trans conformations differ essentially in only one torsion angle. This further suggests the possibility of interconversion between the structures. The increased stability of the Trans-I family over the other two tram families can be attributed to the retention of the a-helical hydrogen-bond pattern even though some of the hydrogen bonds are bifurcated due to distortion of the helix. The Trans-I1 structure has a slightly distorted type I11 0-turn around Pro and Phe, where again the hydrogen bond is bifurcated because of &-helical distortion. The Trans-I11 conformation has a y-turn involving Pro, with a hydrogen bond between the Phe NH and Leu CO. The conformational search for the cis Leu-Pro peptide bond (Table 11) reveals multiple families of conformations. Although an unambiguous choice can not be made regarding the most probable conformation, we selected the family Cis-I (Figure 3 ), which is comparable to Trans-I1 and Trans-I11 in both population and energy.
The four families of structures (Trans-I, -11, -111; Cis-I) selected on the basis of low energy and high population were subjected to optimization of the side chains in the Leu and Phe residues. The final structures (torsion angles and hydrogen bonds) are presented in Table 111. The Leu-Pro-Phe region of the peptide for each of these conformations is shown in Figure 4.
Conformational Transitions Among low-Energy Structures
Conformational interconversion between discrete states requires that each of the alternative conformations be of low energy, that the energy of activation of the transition not be limiting, and that the magnitude of structural change be appropriate for the requisite function. A comparison of relative energies for the similarly populated Trans-11,Trans111, and Cis-I conformations suggests that Trans-I1 is the structure most likely to be assumed upon conformational transition from the a-helical Trans-I structure since it has the lowest energy. It should be noted, however, that relative energies calculated in vacuo (even with the dielectric constant corrected to 2 to reflect the hydrocarbon interior of the mem-
404
POLINSKY ET AL.
Table I11 Backbone Torsion Angles and Hydrogen Bonds for the Lowest Energy Structures with Optimized Conformations of Side Chains
E Structure Trans-I Trans-I1 Trans-I11 Cis-I
(kcal/mol)
Leu 4
Leu$
Pro$
Phe 4
Pheq
0
-50.6
-53.1
-39.4
-72.9
-38.3
6.9 11.0 12.6
-58.0 -53.3 -65.6
116.0 -47.7 155.7
-42.9 59.9 -48.1
-56.2 -43.1 -52.4
-53.4 -42.6 -71.8
brane) may differ from energies in the real environment. As a result, we prefer not to exclude any of these structures on the basis of difference in energy only. Our calculations provide no direct information concerning the energy of activation ( E a ) ,which would characterize any of the possible transitions, nor the mechanism through which substrate or ligand activation might promote the transition by reducing this value during the functioning of a membrane protein. It seems reasonable to assume that the change in shape of a transmembrane segment during a con-
H Bonds CO (Leu) - NH (Ala12, Ala13); CO (Ala8) - NH (Phe, Ala12) CO (Leu) - NH (Ala12, Ala 13) CO (Leu) - NH (Phe) CO (Ala8) - NH (Ala12, Ala 13)
formational transition associated with membrane protein function should not be so dramatic as to disturb adjacent lipids. Although the overall shape of the structures appears similar (Figure 3 ) , they differ in the direction and magnitude of the a-helical bend. Figure 5 schematically shows all four structures with the Ala ( 1)-Ala (8)regions superimposed. Trans-I and Cis-I are bent in the same direction, with Cis-I having a larger bend angle. Trans-I1 and Trans-I11 bend in a direction essentially opposite that of Trans-I. Dynamic conformational interconversions of precisely this sort could be envisioned
Figure 4. The Leu-Pro-Phe region of the lowest energy conformations in various families.
MINIMUM ENERGY CONFORMATIONS Trans-II
405
This work was supported, in part, by grants to MG from the National Institutes of Health GM18694, and to CMD from the Natural Sciences and Engineering Research Council of Canada (NSERC). KAW holds an NSERC studentship.
REFERENCES
Figure 5. Relative spatial arrangement of the N-terminal and C-terminal helices in the most stable families.
as the basis for the opening and closing of a transmembrane channel to allow access to substrates or provide the molecular basis for propagation of signaling events. The transition from Trans-I to CisI will not necessarily be easier than that to TransI1 or Trans-111,however, because a-helical fragments can rotate around the helical axis within a membrane. Simultaneous variation of the appropriate torsion angle ( s ) and rotation around the helical axis can conserve the overall shape of the molecule during the transition from Trans-I to any of the three alternatives.
CONCLUSIONS The present work substantiates that energetically accessible structures are available that provide the molecular basis for local conformational transitions about Pro sites in an environment of low dielectric constant. These transitions may produce conformational fluctuations through rearrangement of local H-bonding patterns (e.g., from a-helical to yturn-like structures) while all-trans peptide bonds are preserved, or through trans-cis interconversion of X-Pro peptide bonds. Both cis and trans X-Pro peptide bonds have been observed experimentally in solvents of low p ~ l a r i t y ,while ~' y-turn structures have been deduced for model Pro-containing peptides in membrane-mimetic environment^.^^ Further exploration of these phenomena awaits experimental determination of the actual conformation ( s ) of XPro peptide bonds within membrane proteins in both resting (e.g., bacteriorh~dopsin~~ ) and activated states. Note added in proof. In computer modelling studies on a Pro-containing helix, Yun et al. (ref. 32) similarly observed a variety of hydrogen-bonding schemes.
1. Brandl, C. J. & Deber, C. M. (1986) Proc. Natl. Acad. Sci. USA 83,917-921. 2. Richardson, J. S. & Richardson, D. C. (1989) in Pre-
diction of Protein Structure and Principles of Protein Conformation, Fasman, G. D., Ed., Plenum Press, New York, pp. 1-99. 3. O'Neil, K. T. & DeGrado, W. F. (1990) Science 2 5 0 , 646-65 1. 4. Chou, P. Y. & Fasman, G. D. (1978) Adu. Enzymol. 4 7 , 45-148. 5. Carver, J. P. & Blout, E. R. (1967) in Treatise on Collagen, Vol. 1, Academic Press, New York, pp. 441526. 6. Rose, G. D., Geselowitz, A. R., Lesser, G., Lee, R. & Zehfus, M. H. (1985) Science 229,9834-9838. 7. Veis, A. & Nawrot, C. F. (1970) J. Am. Chem. SOC. 9 2 , 3910-3914. 8. Williams, K. A. & Deber, C. M. (1991) Biochemistry 30,8919-8923. 9. Zimmerman, S. S. & Scheraga, H. A. ( 1976) Mucromolecules 9 , 408-416. 10. Schmid, F. X., Grafl, R., Wrba, A. & Beintema, J. J. (1986) Proc. Natl. Acad. Sci. USA 83,872-876. 11. Stewart, D. E., Sarkar, A. & Wampler, J. E. (1990) J. Mol. Biol. 2 1 4 , 253-260. 12. Frommel, C. & Preissner, R. (1990) FEBS Lett. 277, 159-163. 13. Brandts, J. F., Halvorson, H. R. & Brennan, M. (1975) Biochemistry 14,4953-4963. 14 Evans, P. A., Dobson, C. M., Kautz, R. A., Hatfield, G. & Fox, R. 0. (1987) Nature 3 2 9 , 266-268. 15. Kikuchi, M., Yamamoto, Y., Taniyama, Y., Ishimau, K., Yoshikawa, W., Kaisho, Y. & Ikehara, M. (1988) Proc. Natl. Acad. Sci. USA 8 5 , 9411-9415. 16. Takahashi, N., Hagano, T. & Suzuki, M. (1989) Nature 337,473-475. 17. Fischer, G., Wittmann-Liebold, B., Lang, K., Kiefhaber, T. & Schmidt, F. X. (1989) Nature 3 4 1 , 755757. 18. Galardy, R. E., Alger, J. R. & Liakopoulou-kyriakides, M. (1982) Znt. J . Peptide Protein Res. 1 9 , 123-132. 19. Ahl, P. L., Stern, L. J., During, D., Mogi, T. & Rothschild, K. J. (1988) J . Biol. Chem. 263,13594-13601. 20. Mogi, T., Stern, L. J., Chao, B. H. & Khorana, H. G. (1989) J. Biol. Chem. 264, 14192-14196. 21. Consler, T. G., Tsolas, 0. & Kaback, H. R. (1991) Biochemistry 3 0 , 1291-1298. 22. Smith, J. A. & Pease, L. G. (1980) CRC Crit. Rev. Biochem. 8 , 315-399.
406
POLINSKY ET AL.
23. Deber, C. M., Glibowicka, M. & Woolley, G. A. (1990) Biopolymers 29, 149-157. 24. Khorana, H. G. (1988) J. Biol. Chem. 263, 74397442. 25. Rothschild, K. J., He, Y.-W., Gray, D., Roepe, P. D., Pelletier, S. L., Brown, R. S. & Herzfeld, J. ( 1989) Proc. Natl. Acad. Sci. USA 86,9832-9835. 26. Rothschild, K. J., He, Y.-W., Mogi, T., Marti, T., Stern, L. J. & Khorana, H. G. (1990) Biochemistry 29,5954-5960. 27. Gerwert, K., Hess, B. & Engelhard, M. (1990) FEBS Lett. 277,159-163. 28. Deber, C. M., Polinsky, A. & Goodman, M. (1991) in Peptides 1990,Giralt, E. & Andreu, D., Eds., ESCOM
29.
30. 31. 32.
Science Publishers, Leiden, The Netherlands, pp. 485487. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J., Swaminathan, S. & Karplus, M. (1983) J. Comp. Chem. 4,187. Deber, C. M., Madison, V. & Blout, E. R. (1976) Acct. Chem. Res. 9,106-113. Deber, C. M., Sorrell, B. J. & Xu, G.-Y. (1990) Biochem. Biophys. Res. Commun. 172,862-869. Yun, R. H., Anderson, A. & Hermans, J. (1991) Proteins: Struct. Func. & Gen. 10, 219-228.
Receiued J u n e 10, 1991 Accepted July 24, 1991
