Mechanisms of thrombin receptor agonist specificity. Chimeric receptors and complementary mutations identify an agonist recognition site.

Identification of the docking interactions by which peptide agonists activate their receptors is critical for understanding signal transduction at the molecular level. The human and Xenopus thrombin receptors respond selectively to their respective hexapeptide agonists, SFLLRN and TFRIFD. A systematic analysis of human/Xenopus thrombin receptor chimeras revealed that just two human-for-Xenopus amino acid substitutions, Phe for Asn in the Xenopus receptor's amino-terminal exodomain and Glu for Leu in the second extracellular loop, conferred human receptor-like specificity to the Xenopus receptor. This observation prompted complementation studies to test the possibility that Arg in the human agonist peptide might normally interact with Glu in the human receptor. The mutant agonist peptide SFLLEN was a poor agonist at the wild type human receptor but an effective agonist at a mutant human receptor in which Glu was converted to Arg. An “arginine scan” of the receptor's extracellular surface revealed additional complementary mutations in the vicinity of position 260 and weak complementation at position 87 but not elsewhere in the receptor. Strikingly, a double alanine substitution that removed negative charge from the Glu region of the human receptor also effectively complemented the SFLLEN agonist. The functional complementation achieved with single Arg substitutions was thus due at least in part to neutralization of a negatively charged surface on the receptor and not necessarily to introduction of a new salt bridge. By contrast, charge neutralization did not account for the gain of responsiveness to SFLLRN seen in the human/Xenopus receptor chimeras. Thus two independent approaches, chimeric receptors and arginine scanning for complementary mutations, identified the Glu region and to a lesser degree Phe as important determinants of agonist specificity. These extracellular sites promote receptor responsiveness to the “correct” agonist and inhibit responsiveness to an “incorrect” agonist. They may participate directly in agonist binding or regulate agonist access to a nearby docking site.

Identification of the docking interactions by which peptide agonists activate their receptors is critical for understanding signal transduction at the molecular level. The human and Xenopus thrombin receptors respond selectively to their respective hexapeptide agonists, SFLLRN and TFRIFD. A systematic analysis of human/Xenopus thrombin receptor chimeras revealed that just two human-for-Xenopus amino acid substitutions, Phe for Asn 87 in the Xenopus receptor's aminoterminal exodomain and Glu for Leu 260 in the second extracellular loop, conferred human receptor-like specificity to the Xenopus receptor. This observation prompted complementation studies to test the possibility that Arg 5a in the human agonist peptide might normally interact with Glu 260 in the human receptor. The mutant agonist peptide SFLLEN was a poor agonist at the wild type human receptor but an effective agonist at a mutant human receptor in which Glu 260 was converted to Arg. An "arginine scan" of the receptor's extracellular surface revealed additional complementary mutations in the vicinity of position 260 and weak complementation at position 87 but not elsewhere in the receptor. Strikingly, a double alanine substitution that removed negative charge from the Glu 260 region of the human receptor also effectively complemented the SFLLEN agonist. The functional complementation achieved with single Arg substitutions was thus due at least in part to neutralization of a negatively charged surface on the receptor and not necessarily to introduction of a new salt bridge. By contrast, charge neutralization did not account for the gain of responsiveness to SFLLRN seen in the human/Xenopus receptor chimeras. Thus two independent approaches, chimeric receptors and arginine scanning for complementary mutations, identified the Glu 260 region and to a lesser degree Phe 87 as important determinants of agonist specificity. These extracellular sites promote receptor responsiveness to the "correct" agonist and inhibit responsiveness to an "incorrect" agonist. They may participate directly in agonist binding or regulate agonist access to a nearby docking site.
Seven transmembrane domain G protein-coupled receptors respond to a structurally diverse set of ligands to regulate a host of biological processes. Catecholamines and certain other small ligands elicit responses by binding to their receptors' transmembrane regions (1). The docking interactions by which peptide agonists activate their receptors are less well-characterized (2)(3)(4)(5)(6)(7). The thrombin receptor can be viewed as a specialized peptide receptor that contains its own "tethered ligand." Thrombin activates its receptor by cleaving the receptor's amino-terminal exodomain. This limited proteolysis unmasks a new amino terminus which then functions as a tethered peptide agonist, binding to the body of the receptor to cause signaling (8 -10). Synthetic peptides which mimic the tethered ligand domain behave as peptide agonists, activating the receptor independent of thrombin. Identification of the interactions by which the thrombin receptor's tethered ligand domain triggers transmembrane signaling is central to understanding signaling by this and perhaps other peptide receptors. Such information may also aid the development of novel pharmaceuticals for inhibiting thrombotic, inflammatory, and proliferative actions of thrombin (11,12).
Three-dimensional structures for seven transmembrane domain G protein-coupled receptors are currently of low resolution, especially in the extramembranous regions (13). Thus structures that reveal the details of agonist docking, particularly for agonists that interact with the receptor's extracellular loops, are unlikely to be available in the near future. Mutagenesis studies which examine functional end points remain a valuable approach that provides constraints for model building and analysis of future structural data.
We exploited the specificity of the human and Xenopus thrombin receptor homologues for their respective agonist peptides to identify the receptor domains which distinguish between these agonists (14). A chimeric receptor in which the Xenopus receptor's extracellular surface (its amino-terminal exodomain and three extracellular loops) was replaced with that of the human receptor showed a remarkable gain of responsiveness to the human agonist and loss of responsiveness to the Xenopus agonist, resulting in human receptor-like agonist specificity (14). More limited replacement of Xenopus with human receptor sequence suggested that two regions accounted for this change in specificity: residues 244 -268 in the second extracellular loop and residues 76 -93 located in the amino-terminal exodomain near the start of transmembrane domain 1 (14). An antiserum which recognized the latter region blocked receptor activation by agonist peptide, consistent with a role for this region in agonist function (15,16).
With the long term goal of defining the agonist-receptor interactions which mediate receptor activation, we analyzed progressively finer chimeras to identify the specific amino acids in the human and Xenopus thrombin receptors which distinguish their cognate agonists. Ultimately, two single human for Xenopus amino acid substitutions in the Xenopus receptor, Phe for Asn 87 in the amino-terminal exodomain and Glu for Leu 260 in the second extracellular loop, proved sufficient to confer robust responses to human agonist. The gain of function conferred by the Leu 260 Glu substitution prompted complementation studies to test the hypothesis that Arg 5a in the human agonist peptide might normally interact with Glu 260 in the human receptor (Fig. 1). These studies again pointed to the receptor domains surrounding Glu 260 and to a lesser degree Phe 87 but revealed that these domains can contribute to receptor specificity by excluding activation by unwanted agonists instead of, or in addition to, providing direct docking interactions. Overall, our results strongly suggest that these extracellular receptor regions of the thrombin receptor participate in receptor activation by binding agonist directly or by regulating agonist access to a nearby docking site.

MATERIALS AND METHODS
Responses of wild type and mutant thrombin receptors to peptide agonists were determined in the Xenopus oocyte expression system using previously described procedures (8,14). Briefly, cDNAs encoding mutant human and Xenopus thrombin receptors were generated by standard techniques (17,18) and confirmed by dideoxy sequencing (19). All wild type and mutant receptors used in this study were epitope tagged with the FLAG sequence present at the amino terminus of the mature receptor protein (14,20). To describe the various mutants, we use the amino acid number of the human thrombin receptor sequence with the start methionine designated as one to refer to both human and their cognate Xenopus receptor residues (8,14); the relevant alignment is shown in Fig. 1 and Table I. cRNAs for the wild type and mutant receptors were transcribed from cDNAs subcloned into pFROG (8). 12.5 ng of wild type cRNA and 12.5-25 ng of mutant receptor cRNAs were injected/oocyte (8,14). After culture for 24 h, receptor expression on the oocyte surface was measured as specific binding of antibody to the FLAG epitope (14). The quantity of cRNA injected was adjusted to maintain surface expression of mutant receptors between 50 and 150% of wild type. Responses to wild type and mutant human and Xenopus agonist peptides were measured as agonist-induced 45 Ca efflux which reflects phosphoinositide hydrolysis in these cells (8,14). For determining EC 50 s, four to five different concentrations of each agonist (from sub EC 50 to maximally activating) were employed. All determinations of EC 50 values for mutant receptor activation included concentration response curves for the appropriate wild type receptor(s) in the same experiment. All experiments were replicated at least twice.
The peptide agonists used in these studies were synthesized as the carboxy amide forms and were purified by high performance liquid chromatography before use (14,21).

RESULTS AND DISCUSSION
The tethered ligand domains of the human and Xenopus thrombin receptors are strikingly different: SFLLRN for human and TFRIFD for Xenopus (14) (Fig. 1). Synthetic peptides mimicking these domains show specificity for their respective receptors. Indeed, the human agonist peptide is approximately 1000-fold more potent at the human receptor than at the Xenopus receptor. Our previous studies showed that replacement of the Xenopus thrombin receptor's amino-terminal exodomain and second extracellular loop with the corresponding human sequences resulted in a chimeric receptor with human receptorlike agonist specificity. Individually, these two substitutions resulted in substantial but less dramatic gains of responsiveness to the human agonist (14).
To identify the specific amino acids in the human and Xenopus thrombin receptors responsible for distinguishing between their cognate agonists, we constructed progressively finer chimeras and determined their responsiveness to human versus Xenopus agonist peptides ( Fig. 1 and Table I). We first focused on the putative binding region in the receptor's amino-terminal exodomain (14). Substitution of human receptor amino-terminal exodomain (HAT) residues 76 -93 for the cognate sequence in the Xenopus receptor yielded some gain of responsiveness to the human agonist but an exaggerated loss of responsiveness to the Xenopus agonist. More informative results were obtained by substituting smaller overlapping segments (76 -86, 82-90, and 87-94). The HAT82-90 chimera exhibited a substantial gain of responsiveness to the human agonist. HAT76 -86 had agonist specificity that differed little from the Xenopus wild type receptor and HAT87-94 was uninformative, showing loss of responsiveness to both human and Xenopus agonists ( Table  I). The contrasting phenotypes of the HAT76 -86 and HAT82-90 chimeras pointed to residues 87-90 as mediating gain of responsiveness to human agonist. Indeed, HAT86 -90 showed a more than 10-fold lower EC 50 for the human agonist compared to the wild type Xenopus receptor. Within this region, HAT87 also yielded a more than 10-fold gain of responsiveness to the human agonist and HAT89,90 a 5-fold gain. HAT86 showed no gain, and Ile 88 is conserved in the two receptors.
The HAT87 human-for-Xenopus amino acid exchange disrupted a consensus sequence for N-linked glycosylation by substituting a Phe for Asn 87 (NIT, Fig. 1). Substitution of Ala for Asn 87 in the Xenopus receptor produced no gain of responsiveness to human agonist peptide (HAT87A , Table I), thus the gain of responsiveness to human agonist seen with HAT87 was due to introduction of the phenyl side chain and not to ablation of an N-linked glycosylation site. Taken together, the data presented above show that human thrombin receptor aminoterminal exodomain residues 87-90 play an important role in defining agonist specificity with Phe 87 being particularly important.
We used a similar approach to identify the amino acids in the receptor's second extracellular loop that distinguish between the human and Xenopus agonists. This loop can be divided into two halves separated by a cysteine thought to participate in a disulfide bridge between the first two extracellular loops (8). HECL2, 254 -268, a chimera in which the "second half" of the human receptor's second extracellular loop was substituted for the cognate Xenopus residues, showed a remarkable gain of responsiveness toward the human peptide ( Fig. 1, Table I). By contrast, a chimera containing the "first half" of this loop (HECL2, 244 -254) had Xenopus receptor-like agonist specificity (Fig. 1, Table I). Because residues 254 -258 are conserved in the Xenopus and human receptors, we next analyzed a series of chimeras in which overlapping segments covering human residues 259 -268 were substituted for the corresponding Xenopus sequences. HECL2, 259 -262 showed a greater than 30-fold decrease in EC 50 to the human agonist; little change was seen with the HECL2, 261-265 or 264 -268 substitutions. To determine whether the consensus N-linked glycosylation site introduced by the HECL2, 259 -262 substitution played a role in the gain of responsiveness to human agonist, HECL2, 259 -262 AETL was constructed. This construct, which substituted Ala-Glu-Thr-Leu (AETL) rather than Asn-Glu-Thr-Leu (NETL) for the cognate Xenopus residues DLKD, still yielded a remarkable gain of responsiveness to the human peptide, but less impressive than that seen with HECL2,259 -262. Whether glycosylation per se or the Asn side chain itself account for Asn's contribution in this context is not known. Taken together, these data pointed to residues 259 and 260, Asn and Glu in the human receptor versus Asp and Leu in the Xenopus, as important for agonist specificity (Table I). Notably, changing Leu 260 in the Xenopus receptor to Glu (HECL2, 260) yielded a gain of responsiveness to the human peptide equivalent to that achieved with substitution of the entire second extracellular loop (Table I).
The best single amino acid substitutions identified in each of the two "specificity regions" defined above were combined to generate "HAT87 ϩ HECL2, 260." This chimera had human receptor-like agonist specificity despite bearing only two human-for-Xenopus amino acid substitutions ( Fig. 2 and Table I). The corresponding double alanine substitution mutant had only a small relative gain of responsiveness to human agonist compared to the wild type Xenopus receptor. Taken together with our earlier studies (14), this completes a systematic analysis of human/Xenopus thrombin receptor chimeras to identify the specific receptor residues which distinguish the human from Xenopus agonists. The data raised the possibility that the agonist peptide might interact directly with the receptor's extracellular surface at or near positions 87 and 260 and suggested a candidate docking interaction. In the agonist peptide, position 5 changes from the aromatic Phe 5a in Xenopus to the basic Arg 5a in human. In the receptor, position 260 changes from the hydrophobic Leu 260 in Xenopus to the acidic Glu 260 in human. The remarkable gain of responsiveness to human agonist conferred by replacing Xenopus receptor's Leu 260 with Glu thus suggested that Arg 5a in the human agonist might dock with Glu 260 in the human receptor ( Fig. 1 and Table I).
This hypothesis was first tested by introducing potentially complementary amino acid substitutions at these positions in the human agonist and receptor. Replacement of Arg 5a in the human agonist peptide with Glu caused a remarkable loss of function at the wild type human receptor. Strikingly, this was remedied by replacing Glu 260 in the human receptor with Arg but not Ala (Fig. 3).
To determine the specificity of this complementation phenomenon, we first tested agonist peptides in which Glu or Asp replaced the native agonist residues at positions 2-5 for their ability to activate wild type versus Glu 260 Arg and Phe 87 Arg mutant human thrombin receptors. Only the agonists with substitutions at position five were effectively complemented by the Glu 260 Arg mutation (Fig. 4A). Asp and Glu substitutions behaved similarly, consistent with a role for the negatively charged side chain. It was possible that the side chains of the   2). This experiment was replicated once. Note the loss of function of the SFLLENPNDK and SFLLDNPNDK peptides at the wild type receptor which was remedied by the Glu 260 Arg mutation and to a lesser extent by the Phe 87 Arg mutation. B, arginine scan of receptor. Xenopus oocytes expressing wild type human thrombin receptor (HWT) or receptors bearing the indicated single arginine substitutions were assayed for responsiveness to SFLLENPNDK (100 M) versus SFLLRNPNDK (10 M). All mutant receptors were expressed at comparable levels on the oocyte surface. (*) indicates "loss of function" mutations; these mutant receptors yielded responses to both wild type and mutant agonists that were less than 15% of those seen with the wild type receptor. All other mutant receptors had responses to wild type agonist that were indistinguishable from wild type receptor. Accordingly, responses to the mutant peptide SFLLENPNDK were expressed as percent of the wild type receptor's response to wild type agonist determined in parallel in each experiment. The data shown are the means of duplicate determinations. Similar results were obtained in two or more experiments with each mutant receptor. Receptors showing "positive" complementation were arbitrarily defined as those that conferred responses to maximal concentrations of SFLLENPNDK that were Ͼ50% of the maximal response of wild type receptor to wild type agonist. C, arginine scan lacked Arg 5a . In this series, SFLLFNPNDK was an effective agonist at the wild type receptor, all aspartate substitutions caused loss of agonist function at the wild type receptor, and only SFLLDNPNDK was effectively complemented by the Glu 260 Arg receptor mutation (data not shown). We cannot exclude the possibility that other intra-agonist interactions specific to positions 2-4 make negatively charged side chains introduced at these positions unavailable to the receptor. However, at face value, these data suggest that receptor position 260 looks primarily at agonist position 5.
Our second test of specificity was an "arginine scan" of the receptor's extracellular surface, a search for mutant receptors that would respond to the SFLLEN mutant agonist (Fig. 4, B  and C). Remarkably, complementary mutations were found only in the regions previously identified as important for receptor specificity by the chimera studies. Arg substitutions at Glu 260 and Glu 264 provided strong complementation and Arg substitutions at nearby residues 261, 263, 265, 268, and 269 caused lesser but reproducible gains in responsiveness to SFLLEN relative to SFLLRN (Fig. 4). Arg substitution at position 87, the other site identified in the chimera studies, provided weak complementation. By contrast, Arg substitutions in extracellular loops 1 and 3 and in the first half of loop 2, all sites found to be unimportant for agonist specificity in the chimera studies, failed to yield complementation. These findings again suggest that the Glu 260 region and possibly Phe 87 are positioned to interact with agonist position five.
What is the physical basis for the functional complementation seen in these studies? The most dramatic complementation was observed when Arg replaced Glu at positions 260 or 264. By replacing or neutralizing negative charges, Arg substitutions at these and nearby positions might be acting to eliminate repulsive electrostatic interactions with position five of the SFLLEN peptide rather than (or in addition to) providing a salt bridge for binding Glu 5a in the mutant agonist (Fig. 5A). Arginine scanning of the Glu 260 region did reveal an apparent helical periodicity, with positive complementation occurring every three to four residues (Fig. 4B), and it was possible that the local structure of this loop placed the Glu 260 and Glu 264 side chains in proximity. We therefore tested the double alanine substitution Glu 260 Ala ϩ Glu 264 Ala for gain of responsiveness to the SFLLEN peptide. In contrast to single alanine substitutions which had little effect ( Fig. 3 and data not shown), the double alanine substitution caused a gain of responsiveness to the SFLLEN peptide comparable to that seen with the Glu 260 Arg substitutions (Fig. 5). This suggests that the Arg substitutions at positions 260 or 264 had complemented the SFLLEN peptide's agonist function at least in part by neutralizing nearby negative charge at positions 264 or 260, respectively (Fig. 5).
Does neutralization of repulsive electrostatic interactions also account for the gain of responsiveness to the human agonist peptide achieved in the receptor chimera studies? The Leu 260 Glu replaced a neutral side chain with an acidic one. Might this acidic side chain be acting by neutralizing nearby basic lysine side chains to decrease repulsive interactions with Arg 5a in the human agonist (Fig. 5B)? Substitution of alanines for the two lysines near position 260 in the Xenopus receptor removed the positively charged side chains from this region but caused neither gain of responsiveness to the human agonist nor loss of responsiveness to the Xenopus agonist (Table I) Residues that were altered in one or more of the mutants are shown in bold. The predicted net charge of this region at physiological pH is indicated at right. Note that the Glu 260 region in the wild type human thrombin receptor (HWT) has a net charge of Ϫ2 and might repel the SFLLEN mutant agonist. Arginine substitutions in this region reduced this net charge, perhaps accounting for the observed gain of responsiveness to SFLLEN. A double alanine substitution mutant that would also neutralize the Glu 260 region (E260A, E264A) was therefore tested for possible gain of responsiveness to the SFLLEN agonist (C). B, charged residues in the region of the Xenopus thrombin receptor corresponding to the human receptor's Glu 260 region. Two lysines that might interfere electrostatically with the action of the SFLLRN agonist are present at positions 261 and 264. The HECL2,260 substitution, which caused a gain of responsiveness to SFLLRN, did add a negatively charged side chain to this region. A double alanine substitution mutant that removed the potentially inhibitory lysines was tested for possible gain of responsiveness to SFLLRN (see Table I appear to be due to removal of electrostatic interactions that interfere with the action of the human agonist on the Xenopus receptor. Alanine scanning of the Phe 87 and Glu 260 regions in the human receptor was undertaken to define the importance of individual side chains for responsiveness to agonist. Phe 87 Ala caused a small loss of responsiveness to human agonist, a 13-fold increase in EC 50 . Ala substitutions for the adjacent Ile 88 and Ser 89 caused 40-and 23-fold increases while substitutions at Glu 90 and Asp 91 had little effect (Table II). In the second extracellular loop, little or no loss of function occurred with alanine substitutions at residues 259 -265. Thus in general, individual alanine substitutions at the residues that the chimera and complementation studies pointed to as important for maintaining receptor specificity caused little loss of responsiveness to agonist (Table II). However, larger deletions and substitutions in either region did cause substantial loss of receptor function (Table II and data not shown). The lack of dramatic loss of function from single alanine substitutions in the receptor's specificity regions is perhaps not surprising. In the agonist itself, positions 3-6 presumably mediate receptor specificity ( Fig. 1 and Ref. 14) Single alanine substitutions at these positions caused only partial loss of activity (14,21), and each amino acid in the agonist probably interacts with several in the receptor.
In contrast to the individual alanine or arginine substitutions within the specificity regions of the receptor, single substitutions at aromatic or hydrophobic residues bounding these domains caused substantial loss of function (Table II and Fig.  4). Similarly, in the agonist peptide, in contrast to "specificity" positions 3-6, alanine substitution for the conserved Phe at agonist position 2 ablates agonist function (14,21,22). The finding of critical hydrophobic residues next to specificity residues in both receptor and agonist is tantalizing. Wells and colleagues (23) have recently found that the majority of the binding energy for growth hormone-receptor docking is contributed by hydrophobic interactions, with nearby charged and hydrophilic residues apparently helping to maintain specific-ity. Moreover, the docking interactions of thrombin's anionbinding exosite with the thrombin receptor itself relies on the interactions of aromatic and hydrophobic residues that are surrounded by charged surfaces (9,24). These analogies prompt the speculation that the critical Phe 2a in the thrombin receptor's agonist peptide may dock in a hydrophobic pocket bounded by the Glu 260 and perhaps the Phe 87 specificity regions. Candidates for contributing to this putative hydrophobic pocket include the conserved aromatic cluster (266 -271) at the carboxyl end of the thrombin receptor's second extracellular loop and/or Ile 88 in its amino-terminal exodomain; point mutations at these sites cause dramatic of loss of function and both sequences are continuous with the specificity regions.
In summary, a systematic analysis of human-Xenopus chimeric thrombin receptors identified two small regions, residues 82-90 in the amino-terminal exodomain and 259 -262 in the second extracellular loop, as important for distinguishing the human versus Xenopus agonist peptides. Within these domains, Phe 87 and Glu 260 made major contributions. These gain of function studies suggested the possibility of direct interactions between these regions and the human agonist peptide and raised the hypothesis that Arg 5a in the agonist might interact with Glu 260 . This prompted a systematic arginine scan of the thrombin receptor's extracellular surface in search of mutations which would complement an Arg 5a Glu mutant agonist peptide. Strikingly, this scan again identified the Glu 260 region and to a lesser degree Phe 87 , but not other sites in the receptor. The successful complementation of the Arg 5a Glu mutant agonist peptide by arginine substitutions in the receptor's Glu 260 region appeared to be due to removal of repulsive electrostatic interactions. By contrast, the gain of function to human agonist seen in the chimera studies could not be accounted for by such a mechanism.
Taken together, our data suggest two alternative models. The Phe 87 and Glu 260 regions of the receptor may simply play a gatekeeper role, regulating agonist access to a nearby binding site that is responsible for receptor activation. Alternatively, Arg 5a and perhaps nearby residues in the human agonist peptide may dock directly with the Glu 260 region and/or Phe 87 in the receptor and contribute to the conformational change that causes receptor activation. Our data show that these regions can both inhibit receptor activation by an "incorrect" agonist and promote receptor activation by the "correct" agonist, both features expected for a docking site. Moreover, our recent finding that mutations in the Glu 260 region can cause constitutive activation of the human thrombin receptor support a direct role for this region in receptor activation. 1