Spatial Approximation between Two Residues in the Mid-region of Secretin and the Amino Terminus of Its Receptor

Photoaffinity labeling of receptors by bound agonists can provide important spatial constraints for molecular modeling of activated receptor complexes. Secretin is a 27-residue peptide hormone with a diffuse pharmacophoric domain that binds to the secretin receptor, a prototypic member of the Class B family of G protein-coupled receptors. In this work, we have developed, characterized, and applied two new photolabile probes for this receptor, with sites for covalent attachment in peptide positions 12 and 14, surrounding the previously most informative site of affinity labeling of this receptor. The [Tyr10,(BzBz)Lys12]rat secretin-27 probe covalently labeled receptor residue Val6, whereas the [Tyr10,(BzBz)Lys14]rat secretin-27 probe labeled receptor residue Pro38. When combined with previous photoaffinity labeling data, there are now seven independent sets of constraints distributed throughout the peptide and receptor amino-terminal domain that can be used together to generate a new molecular model of the ligand-occupied secretin receptor. The aminoterminal domain of this receptor presented a stable platform for peptide ligand interaction, with the amino terminus of the peptide hormone extended toward the transmembrane helix domain of the receptor. This provides clear insights into the molecular basis of natural ligand binding and supplies testable hypotheses regarding the molecular basis of activation of this receptor.

A detailed understanding of the molecular basis for agonist binding to receptors and their activation can provide key insights for the rational design and refinement of receptor-active drugs. The Class B family of guanine nucleotide-binding protein (G protein) 1 -coupled receptors includes several very impor-tant potential drug targets, including receptors for parathyroid hormone, calcitonin, vasoactive intestinal polypeptide, and glucagon (1). The secretin receptor was the first member of this family to be cloned and is prototypic of this group of receptors (2). All of the natural ligands for the Class B G protein-coupled receptors, like secretin, represent moderately long linear peptide hormones that, by first principles, are quite flexible and provide substantial challenges for determination of the appropriate conformation for receptor docking. Although several of these ligands have had solution structures established by NMR (3), it is unclear that these are relevant to the conformations of these peptides when bound to the receptor.
Additionally, the amino-terminal domain of receptors in this family are large (greater than 120 residues), with complex topology, including three conserved disulfide bonds that are critical for ligand binding (1,4). Attempts to gain insight into these structures have come from studies in which the nonglycosylated amino-terminal regions of the parathyroid hormone receptor and the corticotropin-releasing factor receptor have been produced in Escherichia coli, denatured, refolded, and had their disulfide bonds determined directly (5,6). Of interest, the pattern of disulfide bonding for the amino terminus of the corticotropin-releasing factor receptor determined under these conditions was different from that evaluated previously by mutagenesis of the entire intact receptor (7). The recent analysis by Taylor et al. (8) using ab initio modeling and careful evaluation of all permutations of disulfide bond connectivity identified a number of plausible patterns, with no unique solution. Thus, even the constraints for molecular modeling of disulfide bonding are not as certain as would be optimal.
Using the direct approach of intrinsic photoaffinity labeling, we have previously established five distinct spatial approximations between residues in various positions throughout the pharmacophoric domain of secretin and its receptor (9 -13). Of note, four of these five constraints, experimentally established for secretin peptide residues 6, 18, 22, and 26 (of 27 residues), covalently labeled residues within the first 36 residues at the distal end of the amino terminus of the receptor (9 -12). Therefore, this extensive set of data was only able to support a very limited and focused molecular model of this most distal region of the secretin receptor, and even that conformation was not firmly established (12). Only the most recent study, labeling through peptide residue 13, established a covalent bond with a distinct region of the amino terminus of the receptor, at residue 103 (13). This provided adequate information to propose a preliminary model of the entire amino-terminal domain of this receptor (13).
As part of the current project, we have extensively refined and expanded that model, and have experimentally tested it by performing two additional series of photoaffinity labeling experiments with new probes on either side of this most informative site. These probes incorporated photolabile residues into positions 12 or 14. Both bound specifically and saturably to the secretin receptor where they stimulated full biological responses. Each of these probes also covalently labeled this receptor in single, unique, and distinct residues. These residues were quite close to the sites predicted by the working model, and were subsequently incorporated into this model to further refine it.
The seven photoaffinity labeling cross-links, together with the three disulfide bonds in the receptor amino-terminal domain, constitute a significant set of topological constraints for the ligand-receptor complex. Although there is still considerable conformational flexibility in both peptide and receptor, this set of ten constraints reduces dramatically the number of viable three-dimensional models for the peptide-receptor complex that must be considered. The resulting models provide additional insight into the nature of the ligand-receptor interaction for Class B G protein-coupled receptors and suggest new experiments to further test the conformation and the molecular basis of agonist-induced activation of this receptor.
The radioligand used for binding (i.e. [Tyr 10 ]rat secretin-27) and the (BzBz)Lys 12 and the (BzBz)Lys 14 probes were radioiodinated oxidatively with Na 125 I (PerkinElmer Life Sciences, Boston, MA). Upon exposure to the solid-phase oxidant, IODO-BEADs, for 15 s, the resulting product was purified by reversed-phase high-performance liquid chromatography to yield specific radioactivities of 2000 Ci/mmol (15).
Development of a new secretin receptor mutant that incorporated an additional site for CNBr cleavage in a key position was necessary for the current work. This represented mutation of Pro 8 to Met (P8M), prepared using an oligonucleotide-directed approach with the Quik-Change TM site-directed mutagenesis kit from Stratagene. The P8M secretin receptor construct was subcloned into the eukaryotic expression vector, pcDNA3 (Invitrogen, Carlsbad, CA), and its sequence was confirmed by direct DNA sequencing (17). The P8M secretin receptor construct was expressed transiently in COS cells (American Type Cell Collection, Manassas, VA) after transfection using a modification of the DEAE-dextran method (18). These cells were harvested mechanically 72 h after transfection. Plasma membranes from receptor-bearing cells were prepared using a procedure that has been well described, which included cell disruption by Dounce homogenization and sucrose gradient ultracentrifugation (19).
Ligand Binding-This assay was used for characterization of the newly synthesized photolabile probes that were used in the current study. Briefly, increasing concentrations of the (BzBz)Lys 12 or (BzBz)Lys 14 probe (from 0 to 1 M) were incubated with a constant amount of radioligand 125 I-[Tyr 10 ]rat secretin-27, and 5 g enriched CHO-SecR plasma membranes for 1 h at room temperature in Krebs-Ringer-HEPES (KRH) medium (25 mM HEPES, pH 7.4, 104 mM NaCl, 5 mM KCl, 1 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 2 mM CaCl 2 ) containing 0.01% soybean trypsin inhibitor, and 0.2% bovine serum albumin. After incubation, bound and free radioligand were separated using a Skatron cell harvester (Molecular Devices, Sunnyvale, CA) with receptor-binding glass-fiber filter mats that had been soaked in 0.3% Polybrene for 1 h, with the bound radioactivity being quantified in a ␥-spectrometer. Nonspecific binding was determined in the presence of 1 M secretin and represented less than 20% of total radioligand in the incubation.
The same assay was also utilized to characterize the binding activity of the COS cells transiently expressing the P8M secretin receptor construct. Binding curves were analyzed and plotted using the non-linear regression analysis program in the Prism software package (GraphPad Software, San Diego, CA). Binding kinetics were determined by analysis with the LIGAND program of Munson and Rodbard (20). Data are reported as the means Ϯ S.E. of duplicate determinations from a minimum of three independent experiments.
Biological Activity Assay-This was done by measuring the intracellular cAMP accumulation in CHO-SecR cells in response to stimulation by secretin or the photolabile analogues, representing either the (BzBz)Lys 12 or the (BzBz)Lys 14 probes, using reagents provided by Diagnostic Products Corp. (Los Angeles, CA). Briefly, Cells grown in 24-well plates were washed in ice-cold phosphate-buffered saline and stimulated by increasing concentrations (0 -1 M) of peptide for 30 min at 37°C in KRH medium containing 1 mM 3-isobutyl-1-methylxanthine, 0.01% soybean trypsin inhibitor, 0.1% bacitracin, and 0.2% bovine serum albumin. After incubation, cells were lysed by 6% perchloric acid, and cell lysates were then adjusted to pH 6.0 by adding 30% KHCO 3 and cleared by centrifugation before being introduced into the assay tubes, as described previously (21). Radioactivity was quantified by scintillation counting in a liquid scintillation Beckman LS6000 counter. Assays were performed in duplicate and repeated in at least three independent experiments.
Receptor Photoaffinity Labeling-Covalent labeling of the secretin receptor was achieved as described previously (9). In brief, 50 g of enriched receptor-bearing plasma membranes were incubated with 0.1 nM Peptide Mapping of the Site of Covalent Labeling-After gel electrophoresis, labeled bands prepared in large scale were cut out, eluted, lyophilized, and ethanol-precipitated, before being used for deglycosylation and peptide mapping by chemical and enzymatic cleavage.
Deglycosylation of labeled secretin receptor was performed with endoglycosidase F under the conditions we previously reported (19). CNBr and endoproteinase Lys-C were used separately and in sequence to cleave the labeled secretin receptor and its fragments, following procedures described previously (9). Products of cleavage were resolved on 10% NuPAGE gels using MES buffer system (Invitrogen), with labeled products being visualized by autoradiography. The apparent molecular masses of labeled receptor fragments were determined by interpolation on a plot of the mobility of Multimark TM protein standards (Invitrogen) versus the log values of their apparent masses.
Immunoprecipitation of labeled intact and digested HA-tagged receptor constructs using anti-HA monoclonal antibody was performed to determine the identities of some of the labeled fragments using procedures that we described previously (13).
Identification of the Covalently Labeled Residues-Purified CNBr and endoproteinase Lys-C fragments from the radiolabeled secretin receptor were utilized for identification of the residues labeled with the (BzBz)Lys 12 and the (BzBz)Lys 14 probes by radiochemical Edman degradation sequencing. Receptor fragments containing cysteine residues were immobilized through their sulfhydryl groups utilizing the bifunctional cross-linker, m-maleimidobenzoyl-N-hydroxysulfosuccinimide ester, and N-(2-aminoethyl-1)-3-aminopropyl glass beads. Edman degradation was manually repeated up to eight cycles, in a manner that has been previously reported in detail (23,24), and the radioactivity released in each cycle was quantified in a ␥-spectrometer.
Molecular Modeling of the Receptor Amino-terminal Region-Multiple sequence alignment was performed for Class B G protein-coupled receptors to ascertain those family members with significant sequence identity (Ͼ30%) and sequence similarity (Ͼ50%) to the amino-terminal domain of the secretin receptor. The NCBI, SwissProt, and EMBL databases were then searched for homologues of each of the closely related Class B family amino-terminal domains. Homologues with experimentally determined three-dimensional structures were assessed for suitability as templates in a homology modeling exercise using structure-based sequence alignment of the secretin receptor amino terminus. Three-dimensional homology models for the rat secretin receptor amino terminus were then built.
Multiple sequence alignments were generated using the AMPS package (25), and FASTA3 was used for all sequence data base searches (26). Structure-based sequence alignments were generated using the program MOE with the Blossum62 scoring matrix (27). Homology models were created using Modeler (28). Final models were assessed for stereochemical quality and side-chain packing profiles using the PRO-CHECK (29) and QPACK programs (30).
Ligand Docking to the Receptor Amino-terminal Domain-The NMRderived solution structure of secretin (3) was manually docked close to its putative binding site in the three-dimensional model, guided by our previous photoaffinity labeling data (9 -13). The docked complex model was then subjected to 100 steps of in vacuo energy minimization with no constraints, followed by brief (5 ps), low temperature (40 K) molecular dynamics simulations using a generalized Born model. Five constraints (Table I) were applied to impose peptide-receptor contacts determined from previous photoaffinity labeling experiments during the MD simulation, using a harmonic restraining potential with a 20 kcal/mol/Å force constant. The last configuration from the simulation was then energy minimized to generate a final structure for the complex. This same protocol was used to refine additional models that included the three putative disulfide bonds as extra topological constraints (5,6).
Finally, distance geometry calculations were used to search for alternate three-dimensional structures for the ligand-receptor complex that might also satisfy the photoaffinity labeling cross-links and disulfide bond constraints equally well. Both metric matrix and random embedding methods were used to generate three-dimensional structures. For the random embedding procedure, 75,000 iterations were employed for the embedding process, and 500 independent structures were generated by varying the initial seed value. To test the dependence of final structures on starting model conformations, fully extended and various random, compact initial conformations were used for both the receptor amino-terminal and peptide hormone fragments. After embedding, resultant structures were refined with 200 steps of energy minimization. In addition to the ten experimental constraints from cross-linking and disulfide binding patterns, empirical distance constraints were included to impose trans peptide bond conformations for the entire structure. Control calculations were performed without the ten experimental constraints included. All distance geometry calcula-tions were performed using the NAB package (31). The cluster analysis package OC was used to group similar final structures into conformational subfamilies (32).
Molecular Modeling of the Ligand-receptor Complex-The transmembrane helix sequences of the rat secretin receptor were aligned with those of bovine rhodopsin based on the previously reported alignment derived using the "cold spot" method (33). The end of each helix was adjusted based on the recent x-ray crystal structure of bovine rhodopsin (34). The final three residues of helix 6 and the first three residues of helix 7 as indicated by the sequence alignment were unwound to create the third extracellular loop. A homology model for the transmembrane domain was built with systematic replacement of rhodopsin side chains, followed by optimization search to remove steric clashes and distortions in geometry. Extracellular and cytosolic loops were built by first defining anchoring residues located at each helix terminus, then searching a database of protein loops for a proper template to model them. The last carboxyl-terminal helix of the aminoterminal domain model described above was treated as the beginning of transmembrane helix 1, these fragments were superimposed, and the two domains connected to create an intact secretin receptor model.
The ligand-intact receptor models were further refined with additional energy minimization and low temperature molecular dynamics, using weak harmonic constraints (2.0 kcal/mol/Å) to maintain transmembrane helix backbone atoms close to their initial positions, and stronger harmonic constraints as described above to impose the seven photoaffinity labeling cross-links.

Characterization of [Tyr 10 ,(BzBz)Lys 12 ]Rat
Photoaffinity Labeling of the Secretin Receptor-The (BzBz)-Lys 12 and (BzBz)Lys 14 probes were further examined for their ability to covalently photoaffinity label the secretin receptor. Fig. 2 shows that they both labeled this receptor, with labeling inhibited by secretin in a concentration-dependent manner ((BzBz)Lys 12 , IC 50 ϭ 15 Ϯ 3 nM; (BzBz)Lys 14 , IC 50 ϭ 14 Ϯ 3 nM). The bands specifically labeled with each of the probes migrated at approximate M r ϭ 70,000 and shifted to approximate M r ϭ 42,000 after deglycosylation with endoglycosidase F, similar to those labeled by each of the other photolabile secretin receptor probes used previously (9 -13). They did not saturably label any protein in membranes prepared from nonreceptor-bearing CHO cells. These data indicated that both probes were able to label the secretin receptor efficiently and specifically. Thus, we proceeded with identification of their regions of labeling by peptide mapping.
Identification of the Regions of Labeling-To gain an initial indication of regions of labeling with the (BzBz)Lys 12 and (BzBz)Lys 14 probes, we first chose cleavage of the labeled receptor utilizing CNBr that cleaves at the carboxyl side of Met residues. As shown in Fig. 3, CNBr cleavage of the secretin receptor labeled with either the (BzBz)Lys 12 or (BzBz)Lys 14 probe resulted in a single band migrating at approximately M r ϭ 19,000 that shifted to approximately M r ϭ 10,000 after Leu 36 11 deglycosylation. Considering the molecular masses of the attached probes (3258 and 3286 Da for the (BzBz)Lys 12 and (BzBz)Lys 14 probes, respectively), and the presence of glycosylation, the two glycosylated fragments at the amino terminus of the receptor (highlighted in black circles in Fig. 3) represent the best candidates for the domain of labeling with each of the probes.
To definitively identify which of the two fragments represented the region of labeling, two previously well-characterized HA-tagged secretin receptor-bearing cell lines (CHO-SecR-HA37 and CHO-SecR-HA79) were used for immunoprecipitation studies (9). Fig. 4 shows that both HA-tagged secretin receptors labeled with the (BzBz)Lys 12 or (BzBz)Lys 14 probe were immunoprecipitated by the anti-HA antibody, which was competed by excess HA peptide. These data indicated that the labeled HA-tagged receptors were well recognized by the anti-HA antibody. However, after CNBr cleavage, only the immunoprecipitated fragment from the deglycosylated SecR-HA37  CNBr cleavage patterns were identical for both probes, with the fragment from the labeled native receptor migrating at approximate M r ϭ 19,000, and that from the labeled deglycosylated receptor shifting to approximate M r ϭ 10,000. This is representative of at least 10 experiments. The two glycosylated fragments (bold circles) at the receptor amino terminus are the best candidates to represent the domain of labeling for each of the probes.
receptor labeled with either the (BzBz)Lys 12 or (BzBz)Lys 14 probe was radioactive. This indicated that both probes labeled residues within the distal amino-terminal CNBr fragment (Ala 1 -Met 51 ) of the secretin receptor.
Endoproteinse Lys-C, which cleaves at the carboxyl side of Lys residues, was used to further narrow the domain of labeling for each of the probes. As shown in Fig. 5, Lys-C cleavage of the CNBr fragment from the SecR-HA37 receptor labeled with the (BzBz)Lys 12 probe resulted in a non-glycosylated fragment migrating at approximate M r ϭ 6,000, representing the receptor segment Ala 1 -Lys 30 at the amino-terminal distal end of the secretin receptor. In contrast, Lys-C cleavage of the CNBr fragment from the SecR-HA37 receptor labeled with the (BzBz)Lys 14 probe resulted in a glycosylated fragment migrating at approximate M r ϭ 15,000, which shifted to M r ϭ 6,500 after deglycosylation. This indicated that the (BzBz)Lys 14 probe labeled a distinct domain within the amino terminus of the secretin receptor that represented the segment Gly 34 -Met 51 (Fig. 5).
To further localize the site of labeling with the (BzBz)Lys 12 probe, a series of Met secretin receptor mutants were used. Among these were SecR-HA37-V16M (V16M) and SecR-HA37-V13M (V13M) that were characterized previously (10,11). A new receptor mutant SecR-HA37-P8M (P8M) was prepared for this study. It bound secretin with similar affinity to the wild type secretin receptor (K i ϭ 1.7 Ϯ 0.5 nM) and had a normal cAMP response to secretin (EC 50 ϭ 101 Ϯ 19 pM). Fig. 6 shows that the (BzBz)Lys 12 probe was able to efficiently and specifically label V16M, V13M, and P8M. CNBr cleavage of the labeled V16M, V13M, and P8M secretin receptor constructs resulted in non-glycosylated bands migrating at approximately M r ϭ 5000 (Ala 1 -Met 16 ), 4500 (Ala 1 -Met 13 ), and 4000 (Ala 1 -Met 8 ), respectively. These data clearly indicated that the eightresidue segment Ala 1 -Pro 8 at the distal amino terminus of the secretin receptor contained the site of labeling of the (BzBz)Lys 12 probe.
A previously well characterized Met secretin receptor mutant, A41M (11), was utilized to further localize the site of labeling with the (BzBz)Lys 14 probe. Fig. 6 shows that the (BzBz)Lys 14 probe efficiently and specifically labeled the A41M receptor mutant. CNBr cleavage of the labeled A41M receptor yielded a non-glycosylated band migrating at approximately M r ϭ 7500, representing the segment Ala 1 -Met 41 . Taken together with above identification of the segment Gly 34 -Met 51 as the domain of labeling, the (BzBz)Lys 14 probe labeled a residue within an eight-residue segment, representing Gly 34 -Ala 41 .
Identification of the Sites of Labeling-The radiochemically pure CNBr fragment (Ala 1 -Met 51 ) from the labeled secretin receptor was utilized for manual Edman degradation sequencing to identify the specific residue labeled with the (BzBz)Lys 12 probe. Of note, we previously demonstrated that the postulated signal sequence in the mature secretin receptor is cleaved as was predicted by Ishihara et al. (2,10). As shown in Fig. 7, the eluted radioactive peak was found in cycle 6, corresponding to the labeling of residue Val 6 .
To identify the specific residue labeled with the (BzBz)Lys 14 probe, the purified endoproteinase Lys-C fragment (Gly 34 -Met 51 ) from the labeled secretin receptor was used in radiochemical sequencing experiments. Shown also in Fig. 7 is the profile of eluted radioactivity in which a peak was found in cycle 5. This identified Pro 38 of the secretin receptor as the site of labeling with the (BzBz)Lys 14 probe.
Three-dimensional Model of the Amino Terminus of the Secretin Receptor-The amino-terminal domain of the rat secretin receptor has a predicted length of about 120 residues and is regarded as large enough to be folded as an independent domain. Database searches revealed a few structurally well characterized sequence homologues for the secretin receptor aminoterminal domain having greater than 50% sequence similarity. However, none of these candidates satisfied our standards for the comparative modeling template. For example, the homologous region was either a fragment of a larger protein domain or Hence, we decided to expand our homology template search using amino-terminal domain sequences from other closely related Class B G protein-coupled receptors. This homologue search strategy has been proposed previously and is based on the rationale that if sequence A is homologous to sequence B, and B is homologous to C, then A must be homologous to C, even if the sequence similarity between their sequences is low (39).
Using this database search strategy, the amino-terminal fragment of ribonuclease Mc1 from the seeds of bitter gourd (PDB code: 1BK7) was identified, with ϳ27% sequence identity and 54% sequence similarity to the amino-terminal domain of the glucagon-like peptide 1 receptor. Further analysis of the ribonuclease Mc1 three-dimensional structure confirmed that the amino terminus is an independent folding domain, making it suitable as a homology modeling template. Sequence alignment between the amino termini of ribonuclease Mc1 and rat secretin receptor showed that, although there is only 19% sequence identity between the two sequences, the total sequence similarity is about 50% (Fig. 8, upper panel). Therefore, ribonuclease Mc1 was chosen as a reasonable template for modeling the rat secretin receptor amino-terminal domain.
The three-dimensional structure of rat secretin receptor residues 9 -108 was modeled by direct homology with the aminoterminal domain of ribonuclease Mc1 (Fig. 9). The first eight residues of the secretin receptor amino-terminal domain were built in extended conformation, and the final fourteen residues were modeled as two ␣-helices approximately perpendicular to each other, as suggested by previous NMR studies on the corresponding fragment of the amino-terminal sequence of the parathyroid hormone receptor (40). The final helical segment in this secretin receptor amino-terminal domain model corresponds to the top of transmembrane helix 1 in the intact receptor structure. Structural analysis of the final, refined amino-terminal domain model with PROCHECK indicates that 97.6% of the , angles are in the energetically allowed regions, and the structure has an overall G-factor of Ϫ0.18, indicating good stereochemistry. The side-chain packing profile computed with QPACK is ϳ101, indicating a well packed structure.
Docking Model of the Receptor Amino-terminal Domain with an Agonist-Using the NMR-derived solution conformation of secretin as a starting structure, the peptide was docked into its binding site in the receptor amino terminus model, based on previous photoaffinity cross-linking data. During refinement, the secretin peptide adopted a more extended conformation, and the final complex satisfied all original cross-linking constraints (Fig. 10, upper left panel). This model also accommodates the new photoaffinity labeling data reported here quite well, even though these most recent two cross-links were not included as constraints in the homology model construction. This is most encouraging, because these cross-links involve receptor residue positions well separated in the primary sequence (peptide position 12 to receptor residue 6; peptide position 14 to receptor residue 38) and thus impose meaningful constraints on the ligand-receptor complex three-dimensional structure.
Disulfide Bond Pattern in the Amino-terminal Domain-A reliable model should agree well with all the available experimental data. One of the striking characteristics of the aminoterminal domains of Class B G protein-coupled receptors is the six conserved Cys residues forming three disulfide bonds (5,6,41). However, the disulfide bond pattern for this domain has not been established definitively. The most compelling experimental data has come from refolding of denatured, non-glycosylated peptides corresponding to the amino-terminal regions of the parathyroid hormone and corticotrophin-releasing factor receptors (5,6). Interestingly, the reported disulfide bond pattern is consistent in both proteins, where the first conserved Cys forms a disulfide bond with the third, the second Cys with the fifth, and the fourth Cys with the sixth. The final refined receptor amino-terminal domain model incorporating all five photoaffinity labeling cross-links and the three disulfide bonds is shown in Fig. 10 (upper right panel).
Assessment of Model Variability-Our three-dimensional docking model for the secretin receptor amino-terminal domain complex is based primarily on the ribonuclease Mc1 structure, with modest influence from the photoaffinity labeling crosslinks and disulfide bond constraints. Although these ten constraints do impose significant topological restrictions on the final three-dimensional structure for the ligand-receptor complex, ten well characterized distance constraints are clearly not sufficient to uniquely define the three-dimensional structure for a complex comprised of 148 residues. We performed distance geometry calculations, using only these ten distance constraints from photoaffinity labeling cross-links and disulfide bonds, to explore other three-dimensional structures for the ligand-receptor complex that might satisfy these constraints equally well. The random embedding procedure generally yielded better results, with smaller error function violations and more compact, globular structures. Alteration of starting structure had little impact on final results. The root mean square deviation (r.m.s.d.) between the backbones of embedded structures starting from unique initial conformations ranged from 8.6 to 16.9 Å, whereas the r.m.s.d. results between embedded structures derived from a single, extended starting conformation, but with varying initial seed numbers, ranged from 4.7 to 19.7 Å.
As expected, embedded structures generated without imposing the ten cross-linking and disulfide bond constraints displayed great variability, with pairwise r.m.s.d. values as large as 22 Å. In these models, the peptide hormone was not maintained in an intact complex with the amino-terminal domain, and there was no obvious correspondence or similarities between any of the structures. In contrast, the 500 structures generated with inclusion of the ten experimental constraints yielded pairwise r.m.s.d. values ranging from 4.7 to 19.7 Å. Cluster analysis results indicated that these 500 structures could be grouped into ten distinct conformational subfamilies. Although these calculations are not exhaustive, they do suggest that there may be only a finite number of three-dimensional models that possess reasonable compact, globular structures and satisfy the ten experimentally determined distance constraints. Several of the conformational subfamilies generated in the distance geometry calculations exhibit folding topology quite similar to our homology-generated model structure, and most of the other conformational subfamilies differ from the homology model topology in anticipated ways, such as placing a helical segment on top of the packing core rather than below it, etc.
Three-dimensional Model of the Ligand-Receptor Complex-Given a three-dimensional homology model for the secretin receptor amino-terminal domain with bound peptide ligand, we next attempted to construct a model for the intact receptorligand complex to explore possible aspects of receptor activation as a function of ligand complex formation.
Despite the fact that the sequence identity between Class A and Class B G protein-coupled receptors is quite low, Ͻ20%, it is generally believed that both families share a similar sevenhelical bundle domain structure. Because the recently published x-ray structure of bovine rhodopsin is the only experimentally determined high resolution structure for the entire superfamily (34), its coordinates were used to construct a comparative model for the transmembrane domain of the rat secretin receptor (42).
Given the generally low sequence identity between Class A and Class B, it is quite difficult to obtain a reliable alignment between the transmembrane domains of bovine rhodopsin and rat secretin receptor using standard sequence alignment tools. Over the years, various methods have been proposed to tackle this challenge, including a novel "cold spot" method (33). This method aligns protein sequences based on residue conservation rather than sequence similarity, because it is believed that structure and function are generally more conserved than sequences, and highly conserved residues are usually strongly associated with important structural and functional details. We therefore used this approach to construct an alignment between bovine rhodopsin and the rat secretin receptor.
The homology model for the secretin receptor transmembrane domain exhibited good stereochemical quality (G-factor ϭ 0.0) and side-chain packing. It should be noted, however, that the helix bundle topology and conformations for the Class B family of G protein-coupled receptors are likely rather different from the Class A family (43,44). Therefore, the helix bundle domain of the secretin receptor model reported here should be regarded as a rough approximation at this stage. The aminoterminal receptor domain was attached to the transmembrane domain model as described above, yielding the intact receptorligand complex (Fig. 10, bottom panel). The resulting model suggests that the bound peptide hormone is positioned with its amino terminus quite close to the top of the transmembrane helix bundle. Although there are no experimental data yet available for the secretin receptor, photoaffinity labels placed at the amino-terminal residue of parathyroid hormone do, in fact, covalently label the top of the transmembrane helix bundle in the parathyroid hormone receptor (45,46), consistent with the model proposed here for the secretin receptor.

DISCUSSION
In this series of studies, we have generated two additional agonist ligand-receptor cross-links that serve as further constraints for modeling secretin docked to its receptor. These are particularly important, because we now have three consecutive residues in the peptide ligand (positions 12, 13, and 14) that label three highly dispersed residues in the receptor amino terminus. These seven photoaffinity labeling constraints, spread throughout the diffuse pharmacophoric domain of this important peptide hormone and the receptor amino terminus, together with the three highly conserved disulfide bonds, impose considerable topological and conformational constrictions on the ligand-receptor complex, as suggested by distance geometry calculations reported here. All of these spatial constraints are well accommodated in the current molecular model of the secretin-receptor complex.
The rationale for developing the first iteration of this model and its subsequent refinement has been carefully developed under "Experimental Procedures." Using a series of homology searches with a set of Class B G protein-coupled receptor sequences that are homologous to the amino-terminal domain of the secretin receptor, the amino-terminal domain from ribonu- clease Mc1 (PDB code: 1BK7) was established as the most logical structural template for the homology modeling exercises.
The remainder of the receptor amino terminus, representing the first eight residues of the receptor and the fourteen residues closest to the plasma membrane were added, based on existing complementary information. The distal tail residues were attached in an extended conformation, and the residues adjacent to the membrane were added in conformation analogous to that determined by NMR for the corresponding region of the parathyroid hormone receptor (40). The latter was felt to be reasonable, based on a series of observations: (i) there is greater than 30% sequence identity and 50% sequence similarity between the amino-terminal domains of the secretin receptor and the parathyroid hormone receptor; (ii) both domains have six conserved Cys residues and the pattern of disulfide bonds within them is likely similar (5); and (iii) mutation of a single amino acid in the second transmembrane segment of the secretin receptor to the corresponding residue in the parathyroid hormone receptor (I234N) results in a receptor that can bind and signal in the presence of parathyroid hormone and the reciprocal mutation on the parathyroid hormone receptor produces the analogous response to secretin (47).
After manual docking of secretin in the receptor aminoterminal domain, guided by the previously published affinity labeling data (9 -13), and subsequent structure refinement by energy minimization and molecular dynamics, the peptide hormone adopted an extended conformation. This extended conformation of the amino-terminal region of a ligand for a Class B G protein-coupled receptor is consistent with previous NMR studies of a pituitary adenylate cyclase-activating peptide agonist, which assumed an extended conformation in this region upon receptor binding (48).
It is noteworthy that this model easily accommodated the two new photoaffinity labeling cross-links generated in the current work. These were ultimately utilized to further refine the molecular model. It was also fully consistent with the recently proposed disulfide bonding pattern established for refolded, non-glycosylated peptides corresponding to the amino-terminal domains of the parathyroid hormone receptor and bond constraints (C 24 -C 53 , C 44 -C 85 , and C 67 -C 101 ). The three disulfide bonds are colored purple. The receptor backbone is shown in blue, with key residues involved in photoaffinity labeling studies highlighted in red. The secretin peptide backbone is shown in cyan, with key residues that have been involved in photoaffinity labeling studies displayed in yellow. Bottom panel, side view of a complex model of secretin bound to the secretin receptor. The receptor amino-terminal domain is shown in purple, with two newly found key residues involved in photoaffinity labeling studies highlighted in red. The peptide hormone is displayed in cyan, with two newly found key residues that have been involved in photoaffinity labeling studies displayed in yellow. The receptor transmembrane domain is shown in blue. The extracellular loops have been removed in this view to facilitate visualization of the peptide contacts with the amino terminus of the receptor. of the corticotrophin releasing factor receptor (5,6). It is reassuring that establishment of those bonds did not disrupt the experimentally derived spatial constraints coming from the photoaffinity labeling experiments, and continued to be consistent with all of the published mutagenesis data for this receptor.
There is always a hierarchy of data to be accommodated into molecular models. One tries to rely most heavily on the data that are most informative and unambiguous, assigning relative value to each type of observation. The best such data would come from a high resolution structure of a biologically active, high affinity receptor or domain of that receptor. Unfortunately, such data are not yet available for the secretin receptor or for any closely related Class B G protein-coupled receptor. The next most direct type of data that is currently available for this receptor comes from photoaffinity labeling studies utilizing probes, like those in the present study, that are fully biologically active and that bind to the receptor with appropriate specificity and affinity. These are capable of providing useful spatial approximation constraints.
It is also important that the molecular model be consistent with all existing structure-activity data and mutagenesis data. Indeed, in the current work, residues such as Asp 98 and Lys 173 are found to be in regions of tight packing or in regions closely associated with the docked peptide. However, building models solely on the basis of such observations is extremely difficult. Specific mechanisms and explanations will have to be developed for each of these natural or engineered receptor mutations or polymorphisms that might be described in the future. For example, previous secretin receptor mutagenesis studies supported the possibility of receptor loop residues interacting with Asp 3 of secretin during ligand binding and receptor activation (49,50). Indeed, this specific contact is quite feasible in the current model.
Of note, all seven constraints derived from photoaffinity labeling studies involve the amino terminus of the secretin receptor, a region known to be important for natural peptide agonist binding, based on deletions, site mutagenesis, chimeric receptor analysis, and photoaffinity labeling studies (9,18,51,52). It is also the site of a deletion that has been recently described in certain neoplasms associated with a non-functional secretin receptor that can actually act as a dominantnegative inhibitor of wild type secretin receptor (53,54). The amino-terminal domain has been shown to be critically important for ligand binding throughout the Class B family of G protein-coupled receptors (1,41). Because many class B receptors also have peptide ligands that are similar to secretin, it is likely that many details of peptide-receptor interactions and receptor activation defined for the secretin receptor will be applicable to other members of the family as well. This is quite important, because many of the Class B G protein-coupled receptors represent potentially important drug targets.
One area of great interest in the current work is the placement of the amino-terminal region of secretin in the docked ligand-receptor model, because it provides insight into the possible molecular basis of receptor activation. Our current model places the amino terminus of the peptide ligand adjacent to the transmembrane domain of the secretin receptor, in the area of the sixth transmembrane segment. Indeed, that is precisely the area that was photoaffinity labeled through the amino-terminal residue of parathyroid hormone, a natural ligand for another important member of this receptor family (45,46). Tethering of the peptide between its major docking site within the receptor amino terminus and the top of the transmembrane domain has been proposed as a molecular mechanism for transduction of the activated state of the parathyroid hormone re-ceptor (55)(56)(57). Of note, the adjacent third exoloop may also be involved in this by a cooperative process involving a ligand sandwiched between the exodomain and a core domain of the receptor (58 -60). This has yet to be directly studied for the secretin receptor, but our current model suggests this hypothesis merits further experimental investigation.