Simulations on dimeric peptides: evidence for ... - Semantic Scholar

1 downloads 0 Views 645KB Size Report
Mar 10, 1997 - 36 Monga, M., Ku, C.-Y., Dodge, K. and Sanborn,. 37 Bussolati, G., Asboth, G., Negro, F., Stella, A. and. 521 -528. Reprod. Immunol. 26, 19-22.
Biochemical Society Transactions

I066

24 Tsukaguchi, H., Matsubara, €I., Mori, Y., Yoshimasa, Y., Yoshimasa, T., Nakao, K. and Inada, M. (1995) Biochem. Biophys. Res. Commun. 211,967-977 25 Hausmann, H., Richters, A., Kreienkamp, H.-J., Meyerhof, W., Mattes, H., Lederis, K., Zwiers, H. and Richter, D. (1996) Proc. Natl. Acad. Sci. U.S.A. 93,6907-6912 26 Mouillac, B., Chini, B., Balestre, M.-N., Elands, J., Trumpp-Kallmeyer, S., Hoflack, J., Hibert, M., Jard, S. and Barberis, C. (1995) J. Biol. Chem. 270, 25771-25777 27 Chini, B., Mouillac, B., Ala, Y., Balestre, M.-N., Trumpp-Kallmeyer, S., Hiflack, J., Elands, J., Hibert, M., Manning, M., Jard, S. and Barberis, C. (1995) EMBO J. 14,2176-2182 28 Fuchs, A. R., Fields, M. J., Freidman, S., Shemesh, M. and hell, R. (1995) Adv. Exp. Med. Biol. 395, 405-420 29 Zingg, 11. H., Rozen, F., Breton, C., Larcher, A., Neculcea, J., Chu, K., Russo, C. and Arslan, A. (1995) Adv. Exp. Med. Biol. 395, 395-404 30 Bale, T. L., Pedersen, C. A. and Dorsa, D. M. (1995) Adv. Exp. Med. Biol. 395, 269-280 31 Ostrowski, N. L. and Lolait, S. J. (1995) Adv. Exp. Med. Biol. 395, 329-340 32 Flint, A. P. F., Lamming, G. E., Stewart, H. J. and Abayasekara, D. R. E. (1994) Philos. Trans. R. SOC. London Ser. B 344. 291-304

33 Sheldrick, E. L., Flick-Smith, H. C. and Dos Santos Cruz, G. J. (1993) J. Reprod. Fertil. 98, 521-528 34 Bazer, F. W. and Johnson, H. M. (1991) Am. J. Reprod. Immunol. 26, 19-22 35 Flint, A. P. F. (1995) Reprod. Fertil. Dev. 7, 3 13-3 18 36 Monga, M., Ku, C.-Y., Dodge, K. and Sanborn, B. M. (1996) Biol. Reprod. 55, 427-432 37 Bussolati, G., Asboth, G., Negro, F., Stella, A. and Sapino, A. (1995) Adv. Exp. Med. Biol. 395, 553-554 38 Phaneuf, S., Asboth, G., Carrasco, M. P., EuropeFinner, G. N., Saji, F., Kimura, T., Harris, A. and Lopez-Bernal, A. (1997) J. Endocrinol., in the press 39 Asselin, E., Bazer, F. W. and Fortier, M. A. (1997) Biol. Reprod. 56, 402-408 40 Maggi, M., Peri, A., Baldi, E., Mancina, R., Granchi, S., Fantoni, G., Finetti, G., Forti, G., Casini Raggi, C. and Serio, M. (1996) Am. J. Physiol. 271 (Endocrinol. Metab. 34), E840-E846 41 Rozen, F., Russo, C. and Zingg, H. H. (1995) Endocrine Society Annual Meeting, abstract P1-155

Received 10 March 1997

Simulations on dimeric peptides: evidence for domain swapping in G-protein-coupled receptors? P. R. Gouldson and C. A. Reynolds Department of Biological and Chemical Sciences, Central Campus, University of Essex, Wivenhoe Park, Colchester, Essex C04 3SQ, U.K.

Introduction It is known that peptides derived from intracellular loop 3 of G-protein-coupled receptors are able to activate G-proteins [l], and these studies have shed some light on the mechanism of activation. Recently, however, it has been shown that dimers of these peptides show enhanced potency, suggesting that multiple interaction sites on both the receptor and the G-protein are important in activation. Here we use molecular-dynamics simulations on a peptide dimer to investigate whether these observations are consistent with an activation process involving receptor dimers. The simulations were performed on Qdimer, where peptide Q was taken from the C-terminus of intracellular loop 3 of the aW-adrenergic receptor. (Intracellular loop 3 is known to play a

Volume 25

key role in receptor activation [2].) This peptide dimer has been shown to be 100-fold more effective than the monomer at inhibiting highaffinity a2 agonist binding and substantially more potent in inhibiting a2 agonist-stimulated GTPase activity. In CHO cell membranes, the was found to be about two potency of Qdlmer orders of magnitude greater than that of Q for GTPase activity with GJG, [3]. Since there have been a number of recent reports suggesting that G-protein-coupled receptor activation may involve dimerization [4-71, these observations on the Qdimer would appear to be consistent with this. However, closer examination of the structure of Qdlmer with the electron-cryomicroscopy studies of rhodopsin [8,8a], which also show the presence of 1,Z-dimers, suggest that the associa-

G-Protein-Coupled Receptors for Peptide Hormones

tion of Qdimer with the receptor dimers is not straightforward because of steric considerations. Thus, in order to facilitate this comparison, we have modified our helical adrenergic receptor model (P. R. Gouldson, C. R. Snell and C. A. Reynolds, unpublished work) to generate a fullloop model. We have also carried out simulations on Qdimer and monitored the distribution of key distances between the charged (arginine) residues believed to be important in activation. These distances will be compared with those in the various receptor dimers, namely the l,Z-dimer, the 5,6-domain-swapped dimer, which we believe to be important in activation (P. R. Gouldson, C. R. Snell, P. R. Bywater and C. A. Reynolds, unpublished work), and the 4,s-domain-swapped dimer, for which limited circumstantial evidence exists [ 111. (Although we are not aware of any reference to domain swapping in the activation of G-protein-coupled receptors by other authors, domain swapping is a known mechanism for forming dimers in other proteins and is described fully elsewhere [ 121.) T h e methods for constructing the receptor model and the protocol for running the simulations are described in the following section.

Methods

role in the a-P switch ([15,16]; P. R. Gouldson, P. R. Bywater and C. A. Reynolds, unpublished work). Consequently, Asn3I2faces inwards in our model. (The alternative is to use the sitedirected mutagenesis information of Zhou et al. [18], which suggests that Asp79on helix 2 makes a contact with Am3” on helix 7.) The helical model, which is supported by additional experimental data, will be fully described elsewhere (P. R. Gouldson, C. R. Snell and C. A. Reynolds, unpublished work), and the co-ordinates have been deposited at the GPCRDB website [19]. The loop model and the peptide dimer

T h e majority of the loops were added to the helical model using a search algorithm to find similar loops from the Brookhaven protein database. Extracellular loop 2 was added in an extended conformation and constrained molecular dynamics used to form the disulphide bond and to generate the loop conformation. T h e final loop structure was obtained through simulated annealing by decreasing the simulation temperature from 500 K to 298 K over 5ps while keeping the rest of the receptor structure fixed. T h e Nand the C-termini of intracellular loop 3 were predicted to be a-helical (see Table 1) and so were added as helix extensions, and the remain-

Receptor model

The Bz-adrenergic receptor model is based on the bacteriorhodopsin template [ 131 modified to conform to the electron cryomicroscopy projection structure of rhodopsin [8,8a]. A careful check was made to ensure that 22 of the key functional rhodopsin residues on all seven helices were facing inwards. A multiple sequence alignment involving bacteriorhodopsin, rhodopsin and PI-, Pz- and P3-adrenergic receptors was then used to define the position of the Pz-adrenergic residues. Again, a careful check was made to ensure that an additional nine key functional residues, including Se?04 and SerZo7,on helices 2, 3, 5, 6 and 7, were facing inwards. Additional refinements to the length of certain helices were then made on the basis of the results of the helix-prediction algorithm of Rost ~141. The main ambiguity in the model, which does not affect the results presented here, probably concerns the orientation of helix 7 and here we have followed the site-directed mutagenesis data of Suryanarayana et al. [15,15a], which suggest that Am3’‘ is important for binding the ether oxygen of b-antagonists and plays a key

Secondary structure prediction on intracellular loop 2 and intracellular loop 3 of the &adrenergic receptor and peptide P and peptide Q taken from the crzA-adrenergicreceptor The prediction is given directly below the sequence; H denotes helix, L denotes a loop region.

Peptide P RIYQIAKRRTRV HHHHHHHHHH-Peptide Q RWRGRQN REKRFT - - - - HHHHHHHHH lntracellular loop 3 SRVFQVAKRQLQKIDKSEGRFHAQNLSQVE L-HHHHHHHHHHHH----L---LLLLLLLL QDGRSGHGLRSSSKFCLKEHKALKTLGIIM LLLLLLLLL----- HH- HHHHHHH- - - L lntracellular loop 2 IAVDRYVAITSPFKYQSLLTKNKARWIL LL ......LLL __....__ HHHHHH--L

I997

I067

Biochemical Society Transactions

I068

ing 25 residues were predicted to be loop residues and so were added using the same protocol as for extracellular loop 2. The peptide dimer (Qdimer) was predicted to be largely a-helical (see Table 1). Consequently it was constructed in three initial starting geometries: a fully extended chain, an a-helical structure with a cis disulphide bond and an a-helical structure with a trans disulphide bond. All of these configurations converged to similar L-shaped configurations during the molecular dynamics, and so the results presented below are compiled from the simulations on both the cis and trans structures. The interactive modelling was performed using the W A T I F proteinmodelling software [ZO].

Distribution of distances between the key arginine cr-carbon atoms of the Qdlmer

0.3

1

n

R, - R,.

Molecular-dynamics protocol

The same protocol was observed for the minimization and dynamics of both the receptor and the peptide dimer. The initial stage involved ten steps of steepest descent and 9990 steps of conjugate gradient-energy minimization, followed by 1 ps of molecular dynamics at 10 K, 5 ps of heating from 10 to 100 K, 20 ps of heating from 100 to 298 K, followed by 40 ps of molecular dynamics at 298 K. The simulations employed the AMBER all-atom force field [21,21a], as implemented in the AMBER 4.1 suite of programs [22].A distance-dependent dielectric constant was used, as described elsewhere [23].A time step of 0.0005 ps and a non-bonded cut-off of lOA was used for the molecular-dynamics simulations. During the simulations on the peptide dimer, the co-ordinates of the key C, atoms were dumped every 0.5 ps over a period of 400 ps for subsequent analysis.

Results The distance-distribution functions between the key arginines of Qdimer are shown in Figure 1; the secondary-structure predictions on peptides P and Q and intracellular loops 2 and 3 are given in Table 1.

Discussion The primary conclusion to emerge from the is that if Qdimer functions by simulations on Qdimer mimicking a receptor dimer, it does so by mimicking a 5,6-dimer. The l,Z-dimer, as observed in the rhodopsin projection map, places the two copies of intracellular loop 3 about 70w

Volume 25

v 23.0

C(a1pha) - C(a1pha) distance

apart, whereas Figure 1 shows that the range of distances observed between the key arginines in Qdimer is about 11-24 A. The 5,6-domainswapped dimer is shown on the right hand side of Figure 2 and in Figure 3 ( a ) . Here the A domain (helices 1-5) is shown in black while the B domain (helices 6 and 7) is shown in grey on the far left hand side of Figure 2, so that the domain-swapping rearrangement can be visualized. Figure 2 shows how domain swapping brings the two copies of intracellular loop 3 in close proximity. The 5,bcontact dimer (Figure 3 b ) would also bring together two copies of intracellular loop 3, although the loops would be in a different orientation with respect to each other. Consequently, Figure 3 shows how the relative orientation of intracellular loop 3 varies for selected modes of dimerization. In the apo5,6-domain-swapped receptor cimer, the helix 6-helix 6' distance is about 30A, so it is possible to imagine that this distance is comparable with the 24A observed in Qdimer. However, in the 5,6-domain-swapped dimer containing one agonist, computer simulations predict that there is approximately an 8 A shift in the intracellular end of helix 6 (see below), reducing this distance from about 30w to about 22w. Thus the results in Figure 1 are consistent with activation involving receptor dimerization, and, indeed, it may be that the short length of Qdimer plays an important part in bridging key regions on the G-protein. It is also clear from Figure 1 that Qdimer cannot be

G-Protein-Coupled Receptors for Peptide Hormones

Proposed domain-swapping rearrangement For clarity, the dimerization is illustrated using the chirnaeric adrenergic (grey)-muscarinic (black) receptors of Maggio et al. [5,5a].

I069

-

Muscarinic M Adrenergic a 2

mimicking the 4,5-domain-swapped dimer, as the corresponding distances would be too long. T h e chimaeric receptor studies of Maggio et al. [5,5a] enable us to distinguish between the 5,bcontact dimer and the 5,6-domain-swapped dimer. Maggio constructed chimaeric receptors where the A domain was taken from the a2-adrenergic receptor and the B domain was taken from the muscarinic M3 receptor. These receptors and the alternative muscarinic-adrenergic chimaeras were unable to bind ligand or to

stimulate phosphatidylinositol hydrolysis. However, binding and activity were restored when the two chimaeras were co-expressed, and indeed our proposed mechanism of how domain swapping can regenerate intact receptors from two chimaeras is shown in Figure 2. Similar evidence for domain swapping can be derived from the studies of Monnot et al. [ l l ] , except here we conclude that domain swapping resulted in the formation of a 4,5-domain-swapped dimer leading to the restoration of binding, as an intact receDtor was regenerated. but not activitv. " . .probably because the loops were in the wrong relative orientation (see Figure 3 4 . Naturally, it is possible that G-protein activation normally involves receptor monomers and that dimers only become involved during- experi. ments on defective receDtors when there is an opportunity for &by a second receptor. However, if defective receptors are able to domain-swap to regain activity, it is highly likely that fully functional G-protein-coupled receptors also domain-swap. Indeed, there is much additional evidence that G-protein-coupled receptor dimers play a role in activation (P. R. Gouldson, C. R. Snell, P. R. Bywater and C. A. Reynolds, unpublished work). For example, the results of Hebert et al. [4] on the ability of a peptide derived from helix 6 to inhibit both dimerization and activation are particularly convincing, since this provides additional evidence that helix 6 may be involved in the formation of the dimer interface. Other evidence for a 5,6-domain-swapped dimer may be derived from the immunological studies of Ciruela et al. [6]. Thus the simulations on Qdlmer provide evidence consistent with the domain-swapping hypothesis, but the results are not conclusive because other interpretations are possible, for example Qdlmer could in some way mimic P-Q even though the percentage identity is very low. ~

Various possible receptor dimers, showing the different relative orientations of the two copies of intracellular loop 3 which connects helices 5 and 6 (a) The 5,6-domain-swapped dimer which we propose is involved in receptor activation, ( b ) the 5,6-contact dimer, (c) the I ,2-contact dirner, as observed in the rhodopsin projection maps, (d) the 4.5-domain-swapped dimer. which may be inferred from the studies of Monnot Despite carrying loss-of-function mutations on helices 3 and 5, this receptor will bind ligand although it will not activate G-proteins.

(a)

(C)

(b)

(d)

I997

Biochemical Society Transactions

I070

(The peptide P-Q itself could mimic a receptor dimer but would be more likely to mimic the N-terminal and C-terminal portions of a single copy of intracellular loop 3; since P and Q are both highly charged and because charge is important in activation, this may be the origin of the activity of Qdimer.) In order to relate the distance-distribution to the receptor structure, it functions for Qdimer was also desirable to perform simulations on the receptor. Because the a=-adrenergic receptor has a very long intracellular loop 3 (158 residues), the loop modelling and hence the simulations were performed on the homologous D2-adrenergic receptor where our loop model contains only 60 residues. The structural comparisons were actually carried out using the static receptor model, but the simulations help in determining the uncertainty in the distances. The following conformational changes are worthy of note. First, the conformational changes in the loop model occur mainly in helices 5 and 6 and are essentially the same as those that occur in the helix-only model, as observed by ourselves (P. R. Gouldson, C. R. Snell and C. A. Reynolds, unpublished work) and others [24,24a], except that there is less 'fraying' at the ends of the loops in the loop model. There is a lateral movement in the intracellular ends of helices 5 and 6 of about 8 A in the presence of agonist, and this adds some uncertainty in the comparison of the helix 6-helix 6' distances with the corresponding distance-distribution functions (Figure l ) , but fortunately the effect is to decrease the distances within the receptor dimer. Secondly, there is a 15-20" rotation in helix 5 in the loop model, which makes SerZw and SerZo7slightly more accessible to the agonist than in the helix-only model. The origin of this twist probably lies within the additional interactions between thf helix extensions. Thirdly, helix 5 lies 4-6 A higher in the full-loop model compared with the helix-only model (such a change, however, does not have a significant effect on the distance between the key Asp"3 (helix 3) and the serine on helix 5 (SerZo4and Ser"'). As far as the transmembrane helices are concerned, however, the full-loop model gives structural changes that are very similar to those in the helix-only model (P. R. Gouldson, C. R. Snell and C. A. Reynolds, unpublished work). This agreement is very encouraging, since loop modelling is very much in its infancy and so any major model-dependent discrepancies would be cause for great concern.

Volume 25

During the 200 ps of molecular dynamics carried out on the apo receptor, it was clear that the extracellular loops formed a tight canopy, raising questions as to how the ligand may enter the binding site. There are a number of possible solutions to this problem. First, the apparent rigidity may be an artifact of the short simulation time and the loops may possibly open up to permit ligand binding during much longer simulations on the receptor in the presence of lipid and water. Secondly, the ligand may enter from the lipid. If this is the preferred route, the most likely entry point is between helices 1 and 7, since these have parallel helix dipoles and should therefore have the weakest helix-helix interaction. Some movement similar to that in the initial stages of the domain-swapping pathway shown in Figure 2 would assist this process. Thirdly, if the receptor binds directly from the solvent, it may bind as the receptor opens up during domain swapping, as shown in the centre of Figure 2. This particular ligand-binding problem requires further investigation.

Conclusions Here we have presented simulations on Qdimer. The peptide Q, which is based on the C-terminus of intracellular loop 3 of the a=-adrenergic receptor, has been shown to stimulate GJG, but Qdimer iS even more potent. Similar observations have been made for the peptide P and Pdimer. Although there may be a number of reasons for these experimental observations, one explanation is that Qdimer is more effective than peptide Q in G-protein activation because the normal activation process involves G-protein-coupled receptors dimers and that Qdimer is mimicking a receptor dimer. The results of our moleculardynamics simulations on Qdimer are consistent with the hypothesis that G-protein-coupled receptors dimerize, but only if Qdimer is mimicking the 5,6-domain-swapped dimer, since this is the only form of the dimer that brings two copies of intracellular loop 3 into close proximity with each other and which is consistent with the chimaeric receptor studies of Maggio et al. [ S ] . Further evidence for the 5,6-dimer interface comes from an analysis of the external correlated mutations in G-protein-coupled receptors and from simulations on receptor dimers ( [ 2 5 ] ; P. R. Gouldson, C. R. Snell, P. R. Bywater and C. A. Reynolds, unpublished work). The simulations on the loop model gave very similar structural changes in the presence of

G-Protein-Coupled Receptors for Peptide Hormones

agonist or antagonist to those observed in the helix-only model, suggesting that the common practice of modelling only the transmembrane regions of G-protein-coupled receptors is justified provided that the results are evaluated critically. These initial simulations on a loop model of the P2-adrenergic receptor suggest that the pathway for ligand binding is not immediately apparent, as the loops form a tightly binding canopy over the binding site. T h e inherent fluidity of the G-protein-coupled receptors may enable the ligand to bind either from the lipid, entering from between helices 1 and 7, or from the solvent, entering as the helices move apart, as for example during domain swapping. We acknowledge the EPSCR (94309861) and the BBSCR (B/06081) for support; we are grateful to Dr. Christopher R. Snell for his role in developing the ideas of domain swapping in the activation of G-protein-coupled receptors.

1 Dalman, H. M. and Neubig, R. R. (1991) J. Biol. Chem. 266, 11025-1 1029 2 Strader, C. D., Fong, T. M., Tota, M. R. and Underwood, D. (1994) Annu. Rev. Biochem. 63, 101-132 3 Wade, S. M., Dalrnan, H. M., Yang, S. Z. and Neubig, R. R. (1994) Mol. Pharmacol. 45, 1191-1197 4 Hebert, T. E., Moffett, S., Morello, J.-P., Loisel, T. P., Bichet, D. G. and Bouvier, M. (1996) J.Biol. Chern. 271, 16384-16392 5 Maggio, R., Vogel, Z. and Wess, J. (1993) Proc. Natl. Acad. Sci. U.S.A. 90, 3103-3107 5a Maggio, R., Barbier, P., Fornai, F. and Corsini, G. U. (1996) J. Biol. Chem. 271, 31055-31060 6 Ciruela, F., Casado, V., Mallol, J., Canela, E. I., Lluis, C. and Franco, R. (1995) J. Neurosci. Res. 2, 818-828 7 Romano, C., Yang, W.-L. and O’Malley, K. L. (1996) J. Biol. Chern. 271,28612-28616 8 Schertler, G. F. X., Villa, C. and Henderson, R. (1993) Nature (London) 362, 770-772 8a Unger, V. M. and Schertler, G. F. X. (1995) Biophys. J. 68, 1776-1786

9 Reference deleted 10 Reference deleted 1 Monnot, C., Bihoreau, C., Conchon, S., Curnow, K. M., Corvol, P. and Clauser, E. (1996) J. Biol. Chern. 271, 1507-1513 2 Bennett, M. J., Schlunegger, M. P. and Eisenberg, D. (1995) Protein Sci. 4, 2455-2468 3 Henderson, R., Baldwin, J. M., Ceska, T. A., Zernlin, F., Beckmann, E. and Downing, K. H. (1990) J. Mol. Biol. 213, 899-929 14 Rost, B. (1996) Methods Enzyrnol. 266, 525-539 15 Suryanarayana, S. and Kobilka, B. K. (1993) Mol. Pharrnacol. 44, 111-1 14 15a Suryanarayana, S., Daunt, D. A., von Zastrow, M. and Kobilka, B. K. (1991) J. Biol. Chem. 266, 15488-15492 16 Gouldson, P. R., Bywater, R. P. and Reynolds, C. A. (1997) Biochern. Soc. Trans. 25,434s 17 Reference deleted 18 Zhou, W., Flanagan, C., Ballesteros, J. A., Konvicka, K., Davidson, J.S., Weinstein, H., Millar, R. P. and Sealfon, S. C. (1994) Mol. Pharrnacol. 45, 165-170 19 GPCRDB: Information system for G-proteincoupled receptors (GPCRs), European Molecular Biology Laboratory, Heidelberg, Germany (http://www.sander.embl-heidelberg.de/7trn) 20 Vriend, G. (1990) J. Mol. Graph. 8, 52-56 21 Weiner, S. J., Kollrnan, P. A., Case, D. A., Singh, U. C., Ghio, C., Alagona, G., Profeta, Jr., S. and Weiner, P. (1984) J. Am. Chem. SOC.106, 765-784 21aWeiner, S. J., Kollman, P. A., Nguyen, D. T. and Case, D. A. (1986) J. Comput. Chem. 7,230-252 22 Singh, U. C., Weiner, P. K., Caldwell, J. W. and Kollrnan, P. A. (1988) AMBER (Version 4.0), Department of Pharmaceutical Chemistry, University of California, San Francisco 23 Gouldson, P. R., Winn, P. J. and Reynolds, C. A. (1995) J. Med. Chem. 38,4080-4086 24 Luo, X., Zhang, D. and Weinstein, H. (1994) Protein Eng. 7, 1441-1448 24a Zhang, D. and Weinstein, H. (1993) J. Med. Chern. 36,934-938 25 Gouldson, P. R., Bywater, R. P. and Reynolds, C. A. (1997) Biochern. Soc. Trans. 25, 5298 Received 15 April 1997

I997

1071