Structural mechanisms of constitutive activation in the C5a receptors with mutations in the extracellular loops: molecular modeling study

Abstract

Previously we demonstrated by random saturation mutagenesis a set of mutations in the extracellular (EC) loops that constitutively activate the C5a receptor (C5aR) ^1,2. In the current study, molecular modeling revealed possible conformations for the extracellular loops of the C5a receptors with mutations in the EC2 loop or in the EC3 loop. Comparison of low-energy conformations of the EC loops defined two distinct clusters of conformations typical either for strongly constitutively active mutants of C5aR (the CAM cluster) or for non-constitutively active mutants (the non-CAM cluster). In the CAM cluster, the EC3 loop was turned towards the transmembrane (TM) helical bundle and more closely interacted with EC2 than in the non-CAM cluster. This suggested a structural mechanism of constitutive activity where EC3 contacts EC2 leading to EC2 interactions with helix TM3, thus triggering movement of TM7 towards TM2 and TM3. The movement initiates rearrangement of the system of hydrogen bonds between TM2, TM3 and TM7 including formation of the hydrogen bond between the side chains of D82^2.50 in TM2 and N296^7.49 in TM7, which is crucial for formation of the activated states of the C5a receptors ³. Since the relative large length of EC3 in C5aR (13 residues) is comparable with those in many other members of rhodopsin family of GPCRs (13-19 residues), our findings might reflect general mechanisms of receptor constitutive activation. The very recent X-ray structure of the agonist-induced constitutively active mutant of rhodopsin ⁴ is discussed in view of our modeling results.

Introduction

G-protein coupled receptors (GPCRs) constitute a vast protein superfamily of membrane proteins (almost 1000 members in humans). Their structures feature seven-helical transmembrane fragments (TM helices, TMs), the N- and C-terminal tails as well as the extra- and intracellular loops (EC and IC loops) connecting the TM helices. Binding of extracellular agonists triggers GPCR activation that involves conformational movements facilitating interaction of GPCRs with corresponding G-proteins inside the cell. Detailed knowledge of molecular mechanism of GPCR's transition from their ground conformational states to the activated states is of paramount pharmacological importance, since it has been estimated that almost 50% of the therapeutic compounds in use act on GPCRs ⁵.

Mutations in many types of GPCRs can result in constitutive activity, i.e., spontaneous GPCR activity in the absence of an agonist. Constitutively active mutants (CAMs) have been described for many GPCRs and are widely studied; recently, two volumes of Methods in Enzymology (vol. 484 and 485, 2010) extensively presented the current view on constitutive activity of GPCRs. It is commonly believed that the ground conformational states of CAMs mimic the activated conformations of GPCRs. For instance, constitutive activity-inducing mutations occurring in the TM helices may result in different packing of TM fragments within membrane; if they occur in the IC loops, they may change conformational possibilities of the IC loops directly facilitating interactions between CAMs and G-proteins. However, if CAMs feature mutations in the EC loops, their structural interpretation may not be straightforward.

Examples of CAMs with mutations in the EC loops are rare. For instance, of the almost 50 naturally occurring CAMs of the TSH receptor, only 3 possessed mutations in EC1 connecting TM2 and TM3 (all at position 486), 2 in EC2 connecting TM4 and TM5 (both in position 568), and 2 in EC3 connecting TM6 and TM7 (at positions 650 and 656) ⁶. In addition, mutation of residue K191 in EC2 of the GnRH receptor resulted in constitutive activity ⁷. In contrast, using the technique of random saturation mutagenesis of the EC2 loop in the complement factor 5a receptor (C5aR) in yeast, we found 24 constitutively active mutants (ranging from strong to weak CAMs) out of total number of 29 functionally active receptors ¹, and one strong CAM with mutations in EC3 ². The goal of the present study was to suggest possible structural mechanisms for triggering constitutive activity in the CAMs of C5aR with mutations in EC2 or EC3 by modeling conformational possibilities of their EC loops.

Since the EC loops, especially the large EC2 loop (the average number of residues of EC2 in GPCRs being 20-35 residues ⁸), are highly flexible, direct interpretation of experimental structural data on GPCRs in terms of involvement of the EC loops in molecular mechanisms of activation is limited. For instance, the X-ray spectroscopy obtains only a single “frozen” snapshot of the set of possible conformations of the EC loops, and spectroscopic techniques in solution, such as NMR, produce loop structures “averaged” over the set of possible conformations. Besides, the available X-ray data on GPCRs (seventeen structures of six GPCRs referred in the minireview ⁹, and seven structures of GPCRs released very recently ^4,10-15) clearly demonstrated that conformations of the EC loops are very different from one GPCR to another. This observation, in our view, precludes using the X-ray data for structural homology modeling of the EC loops.

Earlier, we have developed a de novo modeling procedure that restores with reasonable accuracy the conformational possibilities of the EC loops in GPCRs ¹⁶ (see also the very recent discussion on the issue ^17-19). In the present study, we employed this procedure for modeling conformational possibilities of the EC loops in the series of C5aR CAMs with mutations in EC2 and EC3. Comparing the conformational space of the EC loops available for constitutively active and non-constitutively active mutants, we have determined geometrically similar conformations of the EC3 loop characteristic for CAMs that differ from conformations for non-CAMs. The major differences occur in spatial positions of the EC3 loop, as well as in subtle conformational changes of inter-residue interactions within the TM regions. Notably, hydrogen bonding between the side chains of D82^2.50 in TM2 and N296^7.49 in TM7 was facilitated in CAMs when compared to non-CAMs. (Residues in TM helices are numbered both sequentially and according to the Ballesteros-Weinstein notation ²⁰ throughout the paper.)

Methods

All modeling procedures used in the present study were developed earlier and were described in details elsewhere ^3,16,21. Only significant differences in calculation protocols related to the present study are outlined below. Sequential steps of the general calculation procedure are described in the Results and Discussion section.

Energy calculations were performed employing the ECEPP/2 force field with rigid valence geometry ^22,23 and trans-conformations of Pro residues; residues Arg, Lys, Glu and Asp were regarded as charged species. Energy calculations were routinely performed with the value of the macroscopic dielectric constant ε of 80 mimicking, to some extent, hydrophilic environment of the extracellular loops.

The possible ground states of the TM region of C5aR found previously ³ were used as templates for mounting the extracellular loops on the rigid four-residue stems of the TM helices further represented as the 20-residue fragments (see Results and Discussion for more details). The stems were kept in place by the system of the harmonic restraint potentials between Cα-atoms (U₀ = 10 kcal/mol*Ų). The same type of harmonic restraint potentials was employed also for keeping together the ends of the 20-residue TM helical fragments at some steps of the calculation procedure (the first run of energy calculations for the EC1+EC2+EC3+TM constructs, see below). The highly conserved disulfide bridge between the C109^3.25 in the TM3 helical stem and C188 in EC2 was represented in the templates by the standard ECEPP harmonic potentials for the disulfide bond. The initial sets of low-energy conformations of the EC2 and EC3 loops used in our calculations were the same as found earlier for C5aR ²¹.

Each separate run of energy calculations involved energy minimization over all dihedral angles of the loops and, at the second run for the EC1+EC2+EC3+TM constructs, also over all dihedral angles of TM helices. Repacking of spatial arrangements for each side chain in the TM region and in the EC loops was performed at each convergence step of energy minimization by an algorithm developed previously ²⁴. Finally, all rms values throughout this paper were calculated between the heavy atoms of the loop backbones with either TM stems or TM helices overlapped.

Results and Discussion

Mutants of C5aR selected for modeling

Table I lists sequences and functional response of 26 C5a receptors with mutations in EC2 or EC3 obtained by our random saturation mutagenesis ^1,2 selected for modeling. We have divided them into four groups. The first group consists of strong constitutively active receptors R25, R17, R115, R111, R127, R125 and R128 (notation according to Ref. ¹). The second group is the collection of non-CAMs R12, R6, R27 and R20; the WT receptor also belongs to this group. Modeling results obtained for these two groups were used for determining structural features characteristic for CAMs and non-CAMs and for deducing the loop conformations common either for CAMs or for non-CAMs (the CAM and non-CAM clusters, see below). The third group (R18, R10, R8, R53, R102, R105, R109, R104 and R118) comprised receptors possessing constitutive activity that was weaker than for receptors in the first group. The previous three groups are mutated in EC2, whereas the fourth group consists of receptors that contain mutations in EC3, namely K13, K97, K129, K132, K17 and K127 ². Loop conformations found by modeling for the third and fourth groups of mutants were further compared to the CAM and non-CAM clusters to determine whether is it possible to predict constitutive activity for mutants in these groups by modeling results.

Table I.

C5a receptors selected for modeling (the data adopted from Ref. ¹ and Ref. ²)

Receptors were co-expressed in BY1142 yeast with and without the C5a ligand. Receptor signaling strength was measured by the ability of yeast to grow in the absence of histidine and at the various concentrations of 3-aminotriazole (AT) ²⁷: ++++++, growth on 50 mM AT; +++++, on 20 mM AT; ++++, on 10 mM AT; +++, on 5 mM AT.

Selection of the template for the TM region

At initial steps of modeling (see below), the EC loops were restored on the rigid four-residue helical stems (TM2 95-98 and TM3 107-109 for EC1; TM4 169-172 and TM5 199-202 for EC2; and TM6 264-267 and TM7 281-284 for EC3). The stems were parts of the seven-helical template representing the TM region of C5aR in the ground state.

However, as we showed earlier, several configurations of TM helices (configurations A_1-4) may be considered as possibilities for the ground state models of C5aR ³. The configurations are very close to each other (with the rms values from 0.54 to 1.29 Å) and differ from the recently determined X-ray structures of the TM bundles of GPCRs by the rms of ca. 3 Å ³. Spatial positions of the TM2-TM3 stems were geometrically similar (by rms cut-off of 1 Å) for all four configurations. At the same time, in regard of spatial positions of the TM4-TM5 stems, configuration A₁ was similar to A₃ (A₁ ≈ A₃) and configuration A₂ was similar to A₄ (A₂ ≈ A₄), whereas in regard of spatial positions of the TM6-TM7 stems configuration A₁ was similar to A₂ and A₃ but not to A₄ (A₁ ≈ A₂ ≈ A₃). Accordingly, the template based on configuration A₁ may be used instead of A₃ for restoration of EC2, and instead of A₂ and A₃ for restoration of EC3. Also, the template based on configuration A₄ may be used instead of A₂ for restoration of EC2. Therefore, two most different templates may be used for restoring both EC2 and EC3, namely A₁ (or A₃) and A₄. One can expect that sampling of the conformational space for the EC loops mounted onto the two most different templates (specifically, configurations A₁ and A₄ were selected) will cover possible conformations of the loops mounted onto the two other templates, A₂ and A₃, as well.

All four configurations were used as templates in calculations performed for WT and R25 yielding different numbers of low-energy conformations of the EC loops selected at the sequential steps of modeling. In both cases, however, the number of low-energy conformations modeled employing template A₁ was significantly higher than that for A₄ (e.g., 144 vs. 3 for WT and 80 vs. 3 for R25 upon modeling of the EC1+EC2+EC3 package; see below). Therefore, only configuration A₁ was used for modeling of all other receptors as the template yielding more comprehensive sampling of the EC loop conformational space.

Sequential steps of modeling

For all receptors, the first step of modeling was to determine the low-energy conformations of the mutated EC2 loops (or the EC3 loops for the receptors in the fourth group of Table I) starting from the sets of low-energy conformations found previously for WT C5aR, namely from 339 conformations for EC2 and from 44 conformations for EC3 ²¹. The next step involved clustering the selected EC2 or EC3 conformations by their geometrical similarity (spatial positions of fragments 172-199 or 264-284) within the rms cut-off of 4 Å (3 Å for EC3). Clusters obtained for EC2 or EC3 in mutant receptors were combined with the clusters of the low-energy conformations of the other loops in WT, namely with 3 clusters for EC1 (the energy cut-off of 10 kcal/mol and the rms cut-off of 3 Å), 27 clusters for EC2 and 21 cluster for EC3 into the EC1+EC2+EC3 “packages” consisting of the EC loops and the corresponding helical stems, including the TM1 38-40 stem. At the next step, fragments of the TM helices of the equal length of 20 residues (TM1 38-57; TM2 79-98; TM3 107-126; TM4 153-172; TM5 199-218; TM6 248-267; and TM7 281-300) were added to the low-energy structures of the EC1+EC2+EC3 packages. The resulting constructs (EC1+EC2+EC3+TM) were subjected to energy minimization by two sequential runs.

In the first run, the dihedral angles in the TM helices were frozen at the same values as in the TM template and the harmonic potentials between the ends of the TM helices were added to keep the TM bundle together. In the second run, the selected EC1+EC2+EC3+TM structures were re-minimized without any additional restraints and with all dihedral angles allowed to vary. Finally, the low-energy structures of EC1+EC2+EC3+TM constructs were selected for further geometrical comparisons.

Table II lists the numbers of conformations selected at the each step of modeling by the uniformly selected energy cut-offs, which are also shown in Table II. In some cases, different energy cut-offs were applied to prevent yielding the too large/small numbers of the resulting conformations; these cases are also noted in Table II.

Table II. Numbers of conformations selected at each calculation step by specified energy cut-offs (ΔE, kcal/mol).

Receptor	EC2 (ΔE < 30) or EC3 (ΔE < 15), numbers of clusters	EC1+EC2+EC3 (ΔE < 30)	EC1+EC2+EC3+TM
			First run (ΔE < 70)	Second run (ΔE < 50)
R25	29	80	27	15
R17	17	76	12a	6
R115	25	280	47	32
R111	20	60	15	12
R127	31	33	11a	6b
R125	18	227	41	24
R128	28	146	38	10
WT	27	144	27	17
R12	36	161	33	17
R6	25	93	19	13
R27	29	268	61	19
R20	33	77	25	10
R18	30c	111	31	24
R10	25c	204	45	31
R8	29c	259	80	19
R53	21c	118	41	12
R102	49c	187	57	36
R105	18c	80	30	14
R109	33c	308	71	22
R104	10b	146	44	28
R118	17c	67d	13	6
K13	14	110	24	11
K97	11	182	56	20
K129	11	298	81	18
K132	13	169	48	20
K17	10	69	16	14
K127	9	53	15c	9

^aΔE < 100 kcal/mol;

^bΔE < 70 kcal/mol;

^cΔE < 25 kcal/mol;

^dΔE < 40 kcal/mol.

Loop conformations characteristic for CAMs and non-CAMs

Visual inspection of the low-energy structures of the EC1+EC2+EC3+TM constructs showed that main geometrical differences between them were in spatial positions of the EC3 loops. For each of the modeled C5aR mutants, the EC structures may be roughly divided into two broad groups, namely those with either “closed” or “opened” EC3 conformations, i.e., turned either towards or away from the TM helical bundle. Accordingly, we compared all obtained structures to each other by the rms values calculated for EC3 (fragment 267–281). For comparison, all EC1+EC2+EC3+TM structures were superimposed on the common template selected earlier for the 20-residue TM fragments (see above). The goal of comparison was to determine possible conformations of EC3 geometrically similar for all strong CAMs (or for all non-CAMs), i.e., determining whether it would be possible to find clusters of conformations of EC3 (consisting of one conformation for every CAM or non-CAM), with geometrical pair-wise differences between all conformations in the cluster not exceeding an rms cut-off of 5.0 Å.

For CAMs belonging to the first group (the strong CAMs), only one such cluster was found; the pair-wise rms values between corresponding common conformations of the seven different mutants (R25, R17, R115, R111, R127, R125 and R128) varied from 2.1 Å (between R25 and R128) to 4.6 Å (between R125 and R128). In this cluster, the EC3 loop was “closed”, i.e., turned towards the TM bundle and EC2. For non-CAMs (WT, R12, R6, R27 and R20), comparison also found only one cluster of common conformations. This non-CAM cluster featured the “opened” EC3 loop turned outwards TM bundle with the rms values between conformations in the cluster varying from 3.1 Å (between R12 and R20) to 4.9 Å (between R12 and R6).

The two clusters are depicted in Figs 1A (the CAM cluster) and 1B (the non-CAM cluster), respectively. The rms values averaged over all conformations within each cluster were 3.1 Å for the CAM cluster and 3.5 Å for the non-CAM cluster. For comparison, the average rms value calculated between the clusters was 8.4 Å, which demonstrates clear distinction between the CAM cluster and the non-CAM cluster.

Figure 1.

Cartoons of the loop structures in the CAM (A) and non-CAM (B) clusters. TM stems are included. EC1 is shown in grey, EC2 in white and EC3 in dark grey.

Structural features characteristic for CAM and non-CAM clusters

Due to the large number of amino acid substitutions in the EC2-mutated receptors, the structures within the CAM and non-CAM clusters displayed residue-residue interactions varying from receptor to receptor. Nevertheless, some specific types of interaction motifs for EC3 were characteristic either for the CAMs or for the non-CAMs.

Table III lists possible contacts between the side chains of EC3 and all other side chains in the structures of the CAM and non-CAM clusters for the first and second groups of mutants, respectively. (A contact was defined as occurrence of any inter-atomic distance between residues less than 4.5 Å.) The data in Table III suggest several differences in the systems of inter-residue interactions in the CAM and non-CAM clusters. First, the central fragment of EC3 (S272-L277) in the structures of the CAM cluster more likely interacted with the residues from the central part of EC2, fragment 178-194. Second, the systems of hydrogen bonding involving the side chains of EC3 were different in the CAM and non-CAM clusters (see contacts shown in bold in Table III; a possible hydrogen bond was defined as the O…H distance less than 3.5 Å). Specifically, within the CAM cluster, the hydrogen bonding of E269 with R200^5.36 in TM5 was characteristic for R127, R25, R17 and R128, while K185 in EC2 was involved in hydrogen bonds with S272 and T274 in R17, and the N279 side chain formed hydrogen bonds with S283^7.36 in TM7 in R125. The latter interaction was also characteristic for the non-CAM cluster (in WT and R12), as well as interactions of N279 with K185 and S189 (both in EC2) in R20. For illustration, see Fig. 2 depicting most essential residue-residue interactions between EC2 and EC3 (and within EC2) for R17.

Table III.

Contacts between the side chains of EC3 and other residues of the EC1+EC2+EC3+TM constructs in the structures of the CAM (top half) and non-CAM clusters (bottom half). Occurrences of hydrogen bonding are shown in bold.

Receptor	Residues of EC3
	L268	E269	P270	S271	S272	P273	T274	F275	L276	L277	L278	N279	K280
R25		R200	M265			Y181			Y258			F190
			R200			Y192			M265
						F193
R17		F181	R200		K185	K185	K185				M264	L187
		R200	M265			L187						S283
						V190
						Y192
R115	R200	M264									D191	S283
		F267									M265
											L281
R111	F267	E195	L194			R178	Y192		L194		M264	V190
			E195			P183			R200		M265	S283
						L194			M265		L281
R127		R200	R197				K191		Y192		T261	S283
			R200						R200		M265
											L281
R125	F267	M265	M265				H194	E191	L187			V190
								R200	E191			D282
								M265				S283
R128		T193	D191			A189				M264	V190
		R200	T193			H194				L281	D191
			R200								T261
			M265								M265
WT		M264	F267					K185		L281	T261	L187
		F267						L187			M265	S283
R12	Y192	M264	F267					M264		V189	M264	V189
		F267						L281		D191	M265	S283
											L281
R6	R200	M264	M264					K185		M264	M264	Q187
			F267									S283
R27	H194	F267					M264	M265			M264	V186	V186
											L281		V187
													S283
R20		M264	F267					M264			T261	K185	S283
		F267						L281			M264	L187
											M265	S189
											L281

Figure 2.

Essential side chain – side chain interactions within EC2 and between EC2 and EC3 in the R17 mutant. All hydrogens are omitted. Hydrogen bonds are sketched as thick dashed lines. Labels for the EC3 residues are underlined.

Our modeling demonstrated more close interactions between the residues of EC3 and EC2 in the structures of the CAM clusters rather than in structures of the non-CAM cluster (compare top and bottom halves of Table III). These interactions are of special significance, since the spatial position of EC2 within the transmembrane segments makes it likely to be involved in a conserved mechanism of receptor activation. The segment of EC2 (from 182 to 192) containing the highly conserved C188, which is involved in the disulfide bond C109^3.25-C188, interacted with fragment 110-117 in TM3 both in the CAM and non-CAM clusters (the data not shown). In the CAM cluster, interactions between EC2 and EC3 led to re-orientation of TM helices; namely, the slight kink in TM2 at P90^2.58 was straightened and TM7 slightly rotated and moved closer to TM2 and TM3 compared to structures of the non-CAM cluster (as illustrated by Fig. 3). For instance, the average distances between Cα atoms of N296^7.49 (TM7) and D82^2.50 (TM2) were 7.6 Å for the CAM cluster and 9.5 Å for the non-CAM cluster, and the distances between Cα atoms of N296^7.49 (TM7) and N119^3.35 (TM3) were 11.2 Å and 13.6 Å for the CAM cluster and for the non-CAM cluster, respectively. At the same time, the average Cα – Cα distances between D82^2.50 and N119^3.35 for both clusters were very close, 6.7 Å and 6.6 Å, respectively.

Figure 3.

The system of hydrogen bonding between selected side chains in structures characteristic for non-CAMs (WT, left) and CAMs (R127, right). TM2, TM3 and TM7 are shown as cylindrical cartoons. All hydrogens are omitted. Hydrogen bonds are sketched as thick dashed lines.

Movement of TM7 suggested by modeling also triggered significant changes in the system of hydrogen bonds within the TM region. Possible hydrogen bonds between the side chains in the structures of both clusters are listed in Table IV. Notably, most conformations in the CAM cluster displayed hydrogen bonding between the side chains of D82^2.50 in TM2 and N296^7.49 in TM7 (those belonging to R25, R115, R111, R127, R125 and R128), while none of the conformations in the non-CAM cluster featured this interaction. We believe this finding is noteworthy, because the D82^2.50-N296^7.49 interaction was shown to be crucial for stabilization of the activated states of C5aR: the mutation N296^7.49 A resulted in loss of constitutive activity of the strong CAM, I124^3.40N/L127^3.43Q ²⁵.

Table IV. Hydrogen bonds between selected side chains in the structures of the CAM and non-CAM clusters.

H-bonds	CAM cluster							Non-CAM cluster
	R25	R17	R115	R111	R127	R125	R128	WT	R12	R6	R27	R20
N48-N296									*
D82-C86			*									*
D82-N119					*			*		*	*
D82-S123			*
D82-N292									*			*
D82-C293											*	*
D82-N296	*		*	*	*	*	*
S85-N119		*			*			*		*
N119-S123	*					*		*	*			*
N119-Y258									*
N119-C293	*			*		*
S123-N292					*
S250-N292		*		*
S250-N296		*

The possible structural mechanism leading to formation of the D82^2.50-N296^7.49 hydrogen bond in the CAM cluster is shown in Fig. 3, exemplified by WT (non-CAM) and R127 (CAM). In R127, TM7 moves by ca. 2 Å closer towards TM2 and TM3. As a result, the side chain of S123^3.39 in R127 re-orients towards N292^7.45 thus decreasing the hydrogen bonding with the side chain of N119^3.35 in WT. In turn, movement of the N292^7.45 side chain in R127 allows re-orientation of the N296^7.49 side chain from the hydrogen bonding with the backbone carbonyl of N292^7.45 (not shown in Fig. 3) towards the D82^2.50 – N296^7.49 hydrogen bonding.

Predicting constitutive activation in C5a receptors with mutations in EC loops

Our modeling found varieties of the “closed” and “opened” conformations of the EC3 loops in the C5aR mutants. Among them was the structure type characteristic for the strong CAMs (conformations in the CAM cluster), and another type of structure that was characteristic for non-CAMs (conformations in the non-CAM cluster). Comparing these two types of structures with low-energy conformations found by modeling for any other C5aR receptor with mutations in the EC loops might be, therefore, used as a tool for predicting the constitutive activation of a given mutant.

However, low-energy conformations similar to the structures of the non-CAM cluster were found even for some receptors from the first modeled group of the highly strong CAMs (R25, R17, R115, R111, R127, R125 and R128). Specifically, geometrical comparison determined at least one conformation with the structure of EC3 that was similar to all structures of the non-CAM cluster by the rms cut-off of 4.9 Å (the maximal value for the pair-wise rms between the structures in the non-CAM cluster) in R25, R115 and R125. Also, at least one conformation similar to all structures of the CAM cluster was found for most of non-CAMs from the second modeled group (WT, R12, R6 and R27). In other words, if presence or absence of structures similar to the CAM or non-CAM clusters in the sets of low-energy conformations of the C5aR mutants with mutations in EC2 is used for prediction of whether the mutants from the first and second groups may display constitutive activity, four of seven mutants would be correctly predicted as CAMs (R17, R111, R127 and R128) and one out of five receptors would be correctly predicted as a non-CAM (R20). All other mutants did not fit into a clear classification of CAMs (or non-CAMs), since their sets of low-energy conformations contained structures similar to both CAM and non-CAM clusters.

By definition of the CAM and non-CAM clusters, each of the strong CAMs from the first group contained one conformation belonging to the CAM cluster, and, vice versa, each of the non-CAMs from the second group contained one conformation belonging to the non-CAM cluster. Therefore, the above comparison was inherently biased and could not produce a wrong prediction, i.e., prediction of CAM as non-CAM and vice versa. However, when low-energy conformations found by modeling C5a receptors from two independent groups (receptors R18, R10, R8, R53, R102, R105, R109, R104 and R118 possessing weaker constitutive activity, the third group in Table I, and receptors K13, K97, K129, K132, K17 and K127 with mutations in EC3, the fourth group in Table I) were compared with the CAM and non-CAM clusters, the predictions of CAMs were still more accurate than predictions of non-CAMs. Four receptors, R18, R53, R104 and K127, were correctly predicted as CAMs, whereas only one CAM, R109, was incorrectly predicted to be a non-CAM. All other CAMs (R10, R8, R102, R105, R109 and R118) were not identified as CAMs or non-CAMs. At the same time, non-CAMs K13, K97, K129 and K132 were not discerned as either CAMs or non-CAMs, and the non-CAM K17 was incorrectly predicted as CAM.

Functional role of EC2 and EC3 in displaying constitutive activity

Our modeling suggested that the EC3 loop adopts the “closed” spatial position for triggering constitutive activity in the C5a receptors with mutations in EC2 or EC3. This stabilizes spatial position of the EC2 fragment 188-199 (“downstream” of the highly conserved residue C188). The downstream fragment of EC2 is located between the extracellular ends of TM3, TM4, TM5 and TM6, which form the pocket for the second binding site of the natural agonist of C5aR, anaphylotoxin C5a ^26-28. The site consists of several residues of C5aR that interact with C5a. Based on previous docking studies, E199^5.35, R200^5.36 and R206^5.42 in TM5 interact with Lys68, Asp69 and the C-terminal carboxyl of Arg74 in C5a, respectively ²¹. Our results showed that the side chains of residues E199^5.35 and R200^5.36 were involved in various contacts with the side chains of EC2 (mostly with fragment 190-194) in all structures from the CAM and non-CAM clusters, whereas contacts of EC2 with the side chain of R206^5.42 were rare. The side chain of E199^5.35 was more likely involved in hydrogen bonding with the EC2 side chains in the structures of the CAM cluster (with Y192 in R127, K185 and S193 in R111, R193 in R115, H191 in R17 (see Fig. 2), R197 in R128, and H193 in R125), than in the structures of the non-CAM cluster (with S193 in R20, and R193 and H194 in R12). On the other hand, the side chain of R200^5.36 was involved in hydrogen bonding with E269 in EC3 in the structures from the CAM cluster (see Fig. 2), but not the non-CAM cluster (see Table III).

Previously, we concluded that EC2 might serve as a negative regulator of C5aR function, most likely by forming interactions that stabilize the ground state of the receptor ¹. Structural features found by our modeling now allow further refinement of this concept. Mutations in EC2 initiate subtle conformational shifts in the ground state of C5aR, which facilitate the activated states of C5aR that are stabilized by interactions such as the hydrogen bond between the side chains of D82^2.50 and N296^7.49. These shifts are triggered by interactions between EC3 and EC2 as well as between EC2 and residues in the TM region. Importantly, in the structures characteristic for constitutively active mutants, the EC3 loop in the “closed’ conformation not only stabilizes EC2, but is directly involved in interactions with the TM region including tight interactions with R200^5.36 that are important for binding the natural ligand of C5aR, C5a.

In addition, recently we identified a previously unappreciated role for S272 in EC3 of C5aR and its interaction with the N terminus of C5aR ²⁹. These studies demonstrated a new microswitch region within the C5aR for receptor activation. It is unclear how the N terminus might affect the conformations of the EC3 in the mutant receptors in the current study since the highly flexible N-termini were excluded from the modeling to avoid combinatorial explosion of sampling the possible conformations of the “N-termini + loops” packages. However, if the N-terminus helps stabilize the ground state of the receptor, we would predict that this interaction could stabilize the “open” conformation of EC3.

Now we may conclude that mutations in EC2 that destabilize the ground state of C5aR (i.e., shift equilibrium of receptor conformations to the activated state) do so in part by interacting with EC3. This conclusion is in line with the important functional role of the EC3 loop in various GPCRs (see, e.g., Ref. ³⁰). The most frequent lengths of the EC3 loops in GPCRs belonging to the rhodopsin family were estimated as being in the regions of either 5-13 or 13-19 residues ⁸. The size of EC3 adopted in the C5aR modeling was 13 residues; as we showed, this length allows both the “opened” and the “closed” conformations of EC3. Longer EC3 loops would also be predicted to be able to adopt these two conformations. The average lengths of the upstream and downstream fragments of EC2 in C5aR (14 and 10 residues, respectively) are also conserved in many GPCRs ³¹. Therefore, the typical “closed” conformations of the EC3 loop that caps the downstream fragment of EC2 found by our modeling could be readily available in many GPCRs. Also, D82^2.50 and N296^7.49 are highly evolutionary conservative residues (94% and 75% of conservation in the rhodopsin family ⁸). Taken together, these observations may indicate a possible general structural mechanism involving EC2 and EC3 in displaying constitutive activity within GPCRs belonging to the rhodopsin family.

Concluding Remarks

In the present study, de novo molecular modeling procedures developed earlier ^3,16 were applied to extensive sampling of possible conformational space available for the EC loops of the C5a receptors with mutations in EC2 or EC3. The obtained low-energy conformations of the EC loops defined two distinct clusters of conformations typical either for the strong constitutively active mutants (the CAM cluster) or for non-constitutively active mutants (the non-CAM cluster). The CAM cluster featured only the “closed” structure of the EC3 loop turned towards the TM bundle, while the non-CAM cluster, on the contrary, featured only the “opened” EC3 structure turned outwards the TM bundle (see Fig. 1). This modeling predicts a role for the EC3 loop in which its contacts with EC2 and with TM3 are necessary for constitutive activity. These interactions trigger movement of TM7 towards TM2 and TM3, which may be the first step in destabilizing the ground state of the receptor. This movement initiates rearrangement of the system of hydrogen bonds between TM2, TM3 and TM7 including formation of the hydrogen bond between the side chains of D82^2.50 in TM2 and N296^7.49 in TM7. As shown elsewhere ³, this hydrogen bond was crucial for formation of the activated states of the C5a receptors, thus providing strong support for the mechanism of constitutive activity suggested by the current de novo modeling. Given the relative large length of EC3 in C5aR (13 residues) comparable with those in many other rhodopsin family members of GPCRs (13-19 residues, which allows the “closed” conformations of EC3), the findings of this modeling might reflect general mechanisms of receptor constitutive activation. However, flexibility of EC3 may be limited by interactions with the N-terminus that was not present in our calculations.

Interestingly, the EC3 loop in the very recently released X-ray structure of the agonist-induced constitutively active mutant of rhodopsin, E113^3.28N, did not show the “closed” conformation, though some EC3 residues contacted residues of EC2 (e.g., Q279 - P194) ⁴. A major consideration when interpreting this result is that an engineered disulfide bond C2-C282 between the N-terminus and EC3 likely limited the conformational flexibility of EC3. Also, the EC3 loop of rhodopsin is only 8 residues long, which might preclude its ability of assume a “closed” conformation. Of note, the constitutive activity of the disulfide bonded rhodopsin mutant N2C/E113^3.28N/D282C was not reported in functional assays. Nevertheless, the X-ray structure featured the hydrogen bonding D^2.50 – N^7.49 discussed above, which occurred through an intermediate water molecule. Importantly, the mutant was co-crystallized with the C-terminal peptide of the α-transducin, which stabilized the active state of rhodopsin (see also Ref. ³², where the same conformational shift was demonstrated for opsin bound to C-terminal peptide of α-transducin).