Identification of Functionally Important Residues of the Silkmoth Pheromone Biosynthesis-activating Neuropeptide Receptor, an Insect Ortholog of the Vertebrate Neuromedin U Receptor*

Background: The moth pheromone biosynthesis-activating neuropeptide (PBAN) and vertebrate neuromedin U (NMU) have a similar biologically essential C-terminal motif (FX1PRX2-NH2). Results: Mutation data revealed important residues in the silkmoth PBAN receptor for ligand binding and signaling. Conclusion: Two glutamate residues conserved in the PBAN/NMU receptor family of GPCRs are responsible for ligand recognition. Significance: A novel ligand-receptor interaction is proposed for the PBAN/NMU family of neuropeptides and receptors. The biosynthesis of sex pheromone components in many lepidopteran insects is regulated by the interaction between pheromone biosynthesis-activating neuropeptide (PBAN) and the PBAN receptor (PBANR), a class A G-protein-coupled receptor. To identify functionally important amino acid residues in the silkmoth PBANR, a series of 27 alanine substitutions was generated using a PBANR chimera C-terminally fused with enhanced GFP. The PBANR mutants were expressed in Sf9 insect cells, and their ability to bind and be activated by a core PBAN fragment (C10PBANR2K) was monitored. Among the 27 mutants, 23 localized to the cell surface of transfected Sf9 cells, whereas the other four remained intracellular. Reduced binding relative to wild type was observed with 17 mutants, and decreased Ca2+ mobilization responses were observed with 12 mutants. Ala substitution of Glu-95, Glu-120, Asn-124, Val-195, Phe-276, Trp-280, Phe-283, Arg-287, Tyr-307, Thr-311, and Phe-319 affected both binding and Ca2+ mobilization. The most pronounced effects were observed with the E120A mutation. A molecular model of PBANR indicated that the functionally important PBANR residues map to the 2nd, 3rd, 6th, and 7th transmembrane helices, implying that the same general region of class A G-protein-coupled receptors recognizes both peptidic and nonpeptidic ligands. Docking simulations suggest similar ligand-receptor recognition interactions for PBAN-PBANR and the orthologous vertebrate pair, neuromedin U (NMU) and NMU receptor (NMUR). The simulations highlight the importance of two glutamate residues, Glu-95 and Glu-120, in silkmoth PBANR and Glu-117 and Glu-142 in human NMUR1, in the recognition of the most functionally critical region of the ligands, the C-terminal residue and amide.

The successful propagation of many moth species is dependent on the female's ability to attract conspecific males via species-specific sex pheromones (1,2). These sex pheromone components are predominantly unsaturated, acyclic, aliphatic C 10 -C 18 compounds containing an oxygenated functional group (e.g. aldehyde, alcohol, or acetate ester). They are synthesized de novo in the pheromone gland (specialized tissue located between the 8th and 9th abdominal segments) from acetyl-CoA through fatty acid synthesis with varied desaturation and limited chain-shortening reactions followed by reductive modification of the acyl group (2,3). In the silkmoth, Bombyx mori, production of the principal sex pheromone component bombykol, (E,Z)-10,12-hexadecadien-1-ol, is initiated during photophase starting from the day of adult eclosion (4,5). Like most moth sex pheromones, bombykol is synthesized from acetyl-CoA through palmitate (16:0) but is then stepwise converted to the bioactive compound by ⌬11 desaturation, ⌬10,12 desaturation, and fatty acyl reduction to the alcohol (2, 6 -8).
In most Lepidoptera, pheromone biosynthesis is regulated by a C-terminally amidated 33-amino acid neuropeptide termed pheromone biosynthesis-activating neuropeptide (PBAN) 4 that originates from the subesophageal ganglion. PBAN was initially isolated in 1989 from the corn earworm Helicoverpa zea (9) and B. mori (10) and has since been identified in a variety of species (11). Structure-function studies have determined that the minimal sequence necessary for pheromonotropic activity resides in the C-terminal five residues and amide group, FXPRL-NH 2 (X ϭ S, T, G, or V) (12). Several neuropeptides containing the C-terminal FXPRL-NH 2 motif have been identified from a number of insect orders, including Lepidoptera, Diptera, and Orthoptera (13). These peptides regulate diverse biological activities, including initiation of B. mori embryonic diapause (14), lepidopteran larval melanization (15), ecdysteroidogenesis in the prothoracic gland of B. mori (16), hindgut/oviduct contraction in cockroach (17) and locust (18), and acceleration of pupariation in flies (19).
Structure-function studies of PBANRs have revealed a number of intracellular domains and sites crucial for receptor activation and regulation (27,36,37). The roles of the extracellular loops (ECLs) and transmembrane (TM) domains in PBAN binding and signaling, however, have not been as well defined. To begin to elucidate the structural determinants that govern ligand discrimination, Choi et al. (38) used a series of domain swaps to implicate ECL3 and the N terminus in PBAN-induced activity. Point mutations of a limited number of residues (five in total) implicated Ser-300 and Phe-303 in ECL3 and two potential N-glycosylation sites in the N terminus (38,39). The applicability of those findings to other PBANRs is questionable as a series of truncations that removed the first 27 residues, includ-ing the homologous N-glycosylation sites, from the B. mori PBANR had no effect on cell surface targeting or ligand-induced internalization (37). Molecular modeling and evolutionary trace approaches designed to facilitate identification of the PBAN-binding pocket identified a number of potential ligand interaction sites (40,41), many of which map to the ECLs. Those sites, however, have yet to be experimentally verified.
In this study, we used multiple sequence alignments and PBANR homology modeling to identify 27 amino acid residues in the B. mori PBANR sequence as potential ligand interaction sites. To assess their functional roles, the identified residues were Ala-substituted and expressed as PBANR-EGFP fusion proteins in cultured insect cells with their subcellular localization, ligand binding, and ligand-induced mobilization of extracellular Ca 2ϩ assessed.
Docking simulations using a homology-based molecular model of PBANR with the solution structure of the minimal active fragment of PBAN (i.e. FSPRL-NH 2 , referred to as C5PBAN) incorporating a type I ␤-turn (42)(43)(44) were consistent with both mutational data and the degree and nature of sequence conservation across PBANRs and related receptors. Independent docking simulations using human neuromedin U (NMU) with human NMU receptor 1 (NMUR1) suggested that the ligand-receptor pairs of PBAN-PBANR and NMU-NMUR1 share similar intermolecular interactions. The docking models also suggest that the functionally essential amide and C-terminal residue of both PBAN and NMU interact with two glutamate residues in TM2 and TM3, both of which are highly conserved across the NMU/PBANR family of GPCRs.
Construction of Ala-substituted PBANR-EGFP Expression Plasmids-The expression plasmid encoding a chimeric B. mori PBANR-EGFP protein was prepared as described (27). Individual Ala substitutions were generated via point mutations using the parental PBANR-EGFP plasmid and the QuikChange (Agilent Technologies, Santa Clara, CA) method. The coding regions of the respective PBANR-EGFP mutants were sequence-verified.
Synthesis and Purification of PBAN Analogs-Reagents for peptide synthesis were purchased from Watanabe Chemical Industries (Hiroshima, Japan). The 10-amino acid peptide SKTRYFSPRL-NH 2 (termed C10PBAN R2K ), which corresponds to the C-terminal 10 amino acids of B. mori PBAN, was synthesized on an automated Apex 369 peptide synthesizer (AAPPTec, Louisville, KY) using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase protocols. To facilitate efficient labeling with Rhodamine Red-X succinimidyl ester (Invitrogen), an R2K mutation was introduced into the peptide during synthesis. After de-protection and cleavage from the resin, the synthetic C10PBAN R2K peptide was purified by reverse-phase HPLC on a PEGASIL ODS column (10 mm inner diameter ϫ 150 mm, Senshu Kagaku, Tokyo, Japan). The peptide was eluted at a flow rate of 4.0 ml/min using a binary solvent system consisting of 0.05% trifluoroacetic acid (TFA) (solvent A) and 90% CH 3 CN, 0.05% TFA (solvent B). The elution profile began with an isocratic flow of 0% B for 2.5 min, followed by a linear gradient from 0 to 75% B in 15 min, and ended with a 100% B isocratic flow for 30 min. Tritium-labeled C10PBAN R2K was synthesized by S- 3  Confocal Scanning Laser Microscopy-Co-localization of a red fluorescent analog of B. mori PBAN, Rhodamine Red-Xlabeled C10PBAN R2K (RR-C10PBAN R2K ), and the PBANR-EGFP mutants was assessed using a confocal scanning laser microscope. Sf9 insect cells, derived from the fall armyworm Spodoptera frugiperda (57,58), were adherently cultured on 35-mm glass bottom dishes (Asahi Glass, Tokyo, Japan) at 28°C for 2 days with 1000 l of IPL-41 (Invitrogen) supplemented with 10% FBS (Nichirei Bioscience, Tokyo, Japan). Cells were transfected with 5 g of each PBANR-EGFP expression plasmid using 10 l of Cellfectin II (Invitrogen) and incubated at 28°C for 16 h. Transfected cells were washed with 10% FBS/ IPL-41 containing 200 g/ml kanamycin and incubated at 28°C for 24 h. To examine co-localization of RR-C10PBAN R2K with the PBANR-EGFP mutants, cells were incubated in 500 l of IPL-41 containing 100 nM RR-C10PBAN R2K in the dark at 4°C for 60 min and then washed three times with PBS and fixed with 4% formaldehyde in PBS. Fluorescence images of PBANR-EGFP and RR-C10PBAN R2K were obtained using a FV-1000D confocal scanning laser microscope (Olympus, Tokyo, Japan) following excitation at 488 nm (EGFP) and 569 nm (Rhodamine Red). Images obtained correspond to a minimum of 20 individual cells expressing the wild type and mutant constructs. The fluorescent data were analyzed using Fluoview (Olympus), and statistical analyses were performed using a Tukey-Kramer test. These experiments were performed in four replicates such that comparative analyses were performed with data from 80 individual cells.
Ligand Binding Assay-A radioligand binding assay was performed using the filter plate method (59). Membrane suspensions from Sf9 cells expressing wild type or a mutant PBANR-EGFP were incubated with different concentrations of Cys([ 3 H]Me)-C10PBAN (100 mCi/mmol) in a buffer containing 500 mM NaCl, 25 mM HEPES-NaOH (pH 7.5), and 1% (w/v) BSA in a total volume of 200 l for 60 min at room temperature. Unbound radioligand was removed by rapid filtration through a GF/F filter (GE Healthcare) presoaked in 0.3% (v/v) polyethyleneimine, and the filter was rinsed twice with 1 ml of ice-cold buffer. Each filter was placed in a scintillation vial, and 5 ml of scintillation mixture were added. The filter-bound radioactivity was measured using the liquid scintillation counter (LSC-6100, Aloka, Tokyo, Japan). Nonspecific binding was determined in parallel reaction in the presence of 200 M unlabeled Cys(Me)-C10PBAN. Equilibrium dissociation constants (K d ) for Cys([ 3 H]Me)-C10PBAN were calculated by nonlinear curve fitting to the saturation curves using IGOR Pro (WaveMetrics, Lake Oswego, OR).
Confocal Laser Scanning Microscopy-based Ca 2ϩ Imaging Assay-Sf9 cells were harvested and incubated in a 24-well glass bottom plate (Asahi Glass) at 28°C for 2 days in 500 l of IPL-41 medium supplemented with 10% FBS. Cells were transfected overnight with 1 g of plasmid DNA in 250 l of IPL-41 and 2.5 l of Cellfectin II (Invitrogen). On the day of the experiment, cells were washed three times with 500 l of IPL-41 and then incubated in the dark at 28°C for 30 min with 250 l of IPL-41 supplemented with 0.75 l each of Pluronic F-127 (Invitrogen) and Fura Red AM (1 mM stock; Invitrogen). After incubation, the cells were washed three times with 500 l of IPL-41 and then maintained in the dark at 28°C for 20 min with 300 l of IPL-41 to allow hydrolysis of the Fura Red AM ester bond. The fluorescence intensities of EGFP and Fura Red were measured using an FV-1000D confocal laser microscope following excitation at 488 and 548 nm, respectively. To measure the Ca 2ϩ mobilization responses of PBANR-EGFP and the Ala substitution mutants, fluorescence at 509 and 610 nm was monitored over 40 scans (1.08 s/scan) following addition of 100 l of C10PBAN R2K to each well after 10 scans. Fura Red fluorescence was analyzed using Fluoview software. The data obtained from five replicates were statistically analyzed using a Tukey-Kramer test.
Molecular Modeling and Docking Simulation of PBAN-PBANR and NMU-NMUR-Homology-based molecular models of PBANR were constructed using the Molecular Operating Environment software suite (Chemical Computing Group, Montreal, Canada) with the crystal structures of agonist-bound turkey ␤1AR (PDB code 2Y03 (60)) and human A 2A AR (PDB code 3QAK (61)), used as templates. Docking simulations of PBAN with PBANR were performed using AutoDock Vina (62) with PBANR conformations in which the N-and C-terminal ends of TM1-5 and TM6 -7, respectively, were opened and separated by 0 -5 Å. In the docking simulation, a subset of the functionally important PBANR residues were defined as flexible residues: Glu-95, Glu-120, Asn-124, Ser-207, Phe-211, Phe-12, Phe-276, Trp-280, Phe-283, His-284, and Arg-287. The PBAN molecular model was restricted to the minimal C-terminal active core (FSPRL-NH 2 , C5PBAN) and was derived from the NMR solution structure of C10PBAN dissolved in 30% (v/v) 2,2,2-trifluoroethanol-d 3 in 50 mM sodium phosphate buffer (pH 6.0) and 0.02% (w/v) sodium azide. Under these conditions, the C-terminal four residues form a type I ␤-turn, whereas the N-terminal five residues are highly disordered (44). These latter residues are not essential for PBAN activity and were not used in the docking simulation. An acetyl group was added to the N terminus of C5PBAN in silico to avoid unfavorable interactions by the N-terminal ␣-amino group that does not exist in the full-length PBAN.
Homology-based molecular models of human NMUR1 were constructed based on the crystal structures of agonist-bound turkey ␤1AR and A 2A AR as described above. The molecular model of the C-terminal active core of NMU, FRPRN-NH 2 , termed C5NMU, was constructed based on the C5PBAN molecular model. The docking simulation of C5NMU to NMUR1 was performed with AutoDock Vina as described. Intermolecular interactions in the docking models were analyzed with PISA (63) and visualized with PyMOL (64).

Prediction of PBANR Ligand Recognition Sites-Knowledge
of GPCR three-dimensional structures can provide mechanistic insights into ligand binding and aid in the determination of ligand interaction sites. To facilitate identification of these sites in PBANR, we aligned the B. mori PBANR sequence with related GPCRs (H. zea PBANR, B. mori DHR, D. melanogaster pyrokinin 1 receptor and pyrokinin 2 receptor, and human NMUR-1 and -2), and two class A GPCRs, human ␤2AR and human A 2A AR, whose ligand recognition residues have been identified. The potential three-dimensional spatial coordinates of the conserved residues were then approximated using a homology-based molecular model of B. mori PBANR generated from the human ␤2AR (PDB codes 2RH1 and 3D4S) and human A 2A AR (PDB code 3EML) crystal structures. These sites were then compared with known ligand recognition sites in the two human receptors to identify 27 amino acids residues that potentially include the B. mori PBANR ligand-binding pocket (Fig. 1). These amino acids are predicted to reside in the follow-  ined in cultured insect Sf9 cells transiently expressing the respective mutants.
Cell Surface Localization-Most of the mutants localized to the plasma membrane as evidenced by a ring of EGFP-associated fluorescence at the cell surface. In contrast, the fluorescence profile of the S207A, F211A, F212A, and H284A mutants was largely intracellular (Fig. 2). Ser-207, Phe-211, and Phe-212 are predicted to reside within TM5, although the fourth residue, H284A, is predicted to be part of TM6. The impaired localization observed with Ala substitution of these residues suggests that their side chains may play a role in stabilizing the PBANR conformation for transport to the plasma membrane.
Binding of a Red Fluorescent PBAN Analog-To examine the ligand-binding ability of each Ala-substituted PBANR-EGFP, Sf9 cells transiently expressing the individual mutants were incubated with RR-C10PBAN R2K , a red fluorescent PBAN analog. A total of 80 cells for each construct (wild type or the individual mutant PBANR-EGFP) were observed using a confocal scanning laser microscope using excitation wavelengths at 488 and 560 nm for EGFP and Rhodamine Red, respectively. Co-localization of the green and red fluorescence signals was considered an indication of ligand binding (Fig. 2).
Variations in receptor expression can affect the degree of ligand binding such that low receptor expression would result in fewer sites for ligand interaction and thus yield a phenotype indistinguishable from loss of a ligand-receptor binding site. To address this issue, we used total cellular EGFP fluorescence as a means of quantifying PBANR expression (Fig. 3A). Based on this metric, we found that expression of the mutants, irrespective of their subcellular localization, was largely comparable with that of wild type PBANR-EGFP with deviations that did not exceed more than 2-fold. We next assessed the affinity of each mutant for the RR-C10PBAN R2K ligand by determining the fluorescence ratio of Rhodamine Red to EGFP at the edge of the cell (Fig. 3B). To ensure sufficient fluorescence intensity, a high RR-C10PBAN R2K concentration (100 nM) was used. As expected, the four mutants (S207A, F211A, F212A, and H284A) with impaired cell surface localization had minimal RR-C10PBAN R2K binding. Markedly diminished (statistical significance at p Ͻ 0.01) RR-C10PBAN R2K binding was observed for 14 mutants (E95A in TM2; E120A and N124A in TM3; V195A in ECL2; F209A in TM5; F276A, W280A, F283A, and R287A in TM6; F303A and Y307A in ECL3; and T311A, G315A, and F319A in TM7). This dramatic reduction in binding suggests that these residues may be critical for ligand recognition. An approximate 50% reduction in ligand binding was observed for Y105A (ECL1), K196A (ECL2), and F312A (TM7), suggesting that they may help coordinate ligand recognition. The W100A, I113A, S119A, V200A, H201A, and Y318A mutations had negligible effects on ligand binding. Furthermore, the poor RR-C10PBAN R2K binding was not correlated with PBANR-EGFP expression (see F209A in Fig. 3, A and B), indicating that the impaired binding was specific to the nature of the Ala substitution and not to the degree of expression.
Binding of a Radiolabeled PBAN Analog-To validate the fluorescence-based binding data, quantitative ligand-binding analyses using tritium-labeled C10PBAN R2K were performed with wild type PBANR-EGFP and three mutants (E95A, I113A, and E120A). The criteria for selecting the mutants used in the assay were based on their observed binding effects and spatial distribution. The TM-residing E95A and E120A mutations both had significant effects on ligand binding (Fig. 3B), whereas the I113A mutation, which is predicted to include a portion of ECL1, had no effect. The binding isotherms (Fig. 3C) for both wild type PBANR-EGFP and the I113A mutation were saturating with comparable density of binding site values (B max ) (wild type, 1.4 nM; I113A, 1.2 nM) and K d values (wild type, 240 nM; I113A, 110 nM), indicating that the cell surface expression levels of the receptors and affinities for the radiolabeled ligand were similar. In contrast, we were unable to generate binding isotherms for the E95A and E120A mutants (Fig. 3C) despite receptor expression levels comparable with that of wild type, indicating that the E95A and E120A mutations effectively inhibited specific binding of the radiolabeled ligand. These findings are consistent with the fluorescence-based binding data shown in Fig. 3B.
Ligand-induced Ca 2ϩ Mobilization-We next sought to examine the effects of the Ala substitutions on PBANR activation. Previous reports have demonstrated that PBAN-induced activation of heterologously expressed PBANR triggers an influx of extracellular Ca 2ϩ (20,27). We consequently used intracellular Ca 2ϩ imaging to assess the effects of the Ala substitutions in response to varying concentrations (1, 10, and 100 nM) of C10PBAN R2K (Fig. 4). A comparison of the wild type PBANR-EGFP K d (240 nM, Fig. 3C) and EC 50 (Ͻ 0.1 nM, Fig. 4B) indicates that the receptor was sufficiently overexpressed such that only a small fraction of cell surface receptor was required for maximal signaling. The comparable expression levels with wild type PBANR-EGFP (Fig. 3A) suggests similar conclusions could be drawn for the mutants. This explains why the cells

Ligand-Receptor Interactions of PBAN-PBANR and NMU-NMUR
expressing the F212A mutant, which exhibited impaired cell surface localization (Fig. 2), had a moderate Ca 2ϩ response (Fig.  4C); some portion of the receptor translocated to the cell surface but was below our threshold of detection (i.e. fluorescent visualization). Consistent with our previous findings, the fluorescence changes associated with Ca 2ϩ mobilization (Fig. 4A) in cells expressing the E120A, S207A, and F211A mutations were essentially undetectable at all three ligand concentrations tested. The reduced Ca 2ϩ mobilization profiles of cells expressing the E95A, E120A, N124A, F276A, W280A, F283A, R287A, Y307A, and F319A mutants (Fig. 4C) suggests a critical role for these residues in PBANR activation. In contrast, Ca 2ϩ mobili- zation in cells expressing the V195A, T311A, and Y318A mutants was only moderately reduced compared with control cells (Fig. 4C), suggesting a secondary role for these residues in transmitting the PBAN signal. The intracellular fluorescence of cells expressing the remaining 11 mutants (i.e. W100A, Y105A, I113A, S119A, K196A, V200A, H201A, F209A, F303A, F312A, and G315A) was indistinguishable from control cells, indicating that these residues likely do not contribute to receptor activation.
Homology-based Molecular Modeling of PBANR and Location of Functionally Important Residues-To examine the intramolecular location of the functionally important residues in PBANR in relation to a bound ligand, a second set of homology-based models were constructed using the crystal structures of thermostabilized turkey ␤1AR with bound isoprenaline (PDB code 2Y03) and human A 2A AR with bound UK-432097 (PDB code 3QAK). The resulting models (supplemental Data S1) diverged significantly with a root mean square deviation of 3.2 Å for C ␣ atoms, and ligand-docking simulations (see below) were only successful with the A 2A AR-derived model. Consequently, we used the A 2A AR-derived model to map the functionally important PBANR residues (Fig. 5). The four residues essential for the efficient localization of PBANR to the plasma membrane (i.e. Ser-207, Phe-211, Phe-212, and His-284) clustered within a spatially limited region of TM5 and TM6 (Fig.  5A). Similarly, the residues critical for ligand binding and signaling clustered together with the putative ligand-binding pocket in the three-dimensional model (Fig. 5, B and C). Intriguingly, most of the PBANR residues critical for C10PBAN R2K binding and signal transduction share the same general spatial coordinates as the ligand-binding residues in A 2A AR, ␤2AR, ␤1AR, and D 3 -dopamine receptor (65). The ligand-binding residues of neu-FIGURE 5. Location of functionally important residues in PBANR. The important residues identified by Ala substitution for plasma membrane localization (A), C10PBAN R2K binding (B), and Ca 2ϩ mobilization are colored and labeled in the homology-based model of PBANR (C). Residues shown in darker colors are more important than those depicted in lighter colors. D, functionally important residues in PBANR and their corresponding residues in PBANR-like class A GPCRs and class A GPCRs whose agonist-bound crystal structures have been solved. Biological functions affected by Ala substitution of the PBANR residues are indicated by the colored circles (localization, green; ligand binding, blue; signaling, red). Amino acid residues completely conserved with PBANR are highlighted using the same color scheme but with purple to indicate residues involved in both ligand binding and signaling. Residues shown to form contact points with bound ligand in the solved crystal structures are indicated by solid boxes, and those predicted in the docking simulations are indicated by dotted lines. JULY (66), however, only partially overlap with PBANR and the above-mentioned class A GPCRs ( Fig. 5D and supplemental Fig. S1).

Ligand-Receptor Interactions of PBAN-PBANR and NMU-NMUR
Docking Simulation of PBAN with PBANR-Finally, to gain insights into PBAN binding itself, we used AutoDock Vina to simulate binding of C5PBAN ( 29 FSPRL 33 -NH 2 ), the smallest active fragment of PBAN (12), with the PBANR homology models generated from the UK-432097-bound A 2A AR structure. We were particularly interested in the molecular interactions between C5PBAN and the residues we determined experimentally to be critical for binding and activation. Simulations were performed using C5PBAN with a type I ␤-turn as NMR studies of PBAN have suggested that the C-terminal active site of PBAN adopts a ␤-turn (42)(43)(44). The ligand-binding pocket of the PBANR molecular models, however, was too small to fully accommodate C5PBAN with the type I ␤-turn conformation. We consequently artificially widened the pocket by splitting the N-and C-terminal parts of PBANR (i.e. TM1-5 and TM6 -7, respectively) and expanded the gap by 1-5 Å. This split GPCR method was selected based on a report demonstrating that the N-and C-terminal regions of ␤2AR corresponding to TM1-5 and TM6 -7, respectively, form a functional ␤2AR when co-expressed (67). The following criteria were used to evaluate reasonable docking models: 1) the ligand-receptor interaction is favorable in terms of binding energy; 2) the bound ligand is accommodated in the ligand-binding pocket of the receptor; 3) the bound ligand interacts with the functionally important residues of the receptor; and 4) the N terminus of the ligand extends beyond the entrance of the ligand-binding pocket. With a receptor separation of Յ2 Å, no docking model met the above criteria. An appropriate docking model with an affinity of Ϫ10.1 kcal/mol was obtained using a 3-Å separated conformation derived from the A 2A AR template. This docking model (supplemental Data S2) was chosen as the most appropriate because it met all the above criteria with a minimal receptor separation. In the docking model, a number of contact points were observed between C5PBAN and PBANR (Fig. 6, A and C, and supplemental Data S3) as follows: Phe-29, six contact points in TMs 6 and 7; Ser-30, five contact points in TMs 6 and 7; Pro-31, seven contact points in TMs 5-7; Arg-32, eight contact points in TMs 3 and 5-7; Leu-33, five contact points in TMs 2, 3, 6, and 7; and the terminal NH 2 , three contact points in TMs 2 and 7. Intriguingly, although no interactions were observed between C5PBAN and the ECLs, all four residues important for the normal transport of PBANR (i.e. Ser-207, Phe-211, Phe-212, and His-284) to the plasma membrane were predicted to interact with C5PBAN, suggesting that these residues may also be crucial for ligand recognition. Docking Simulation of NMU to NMUR-To validate the PBAN-PBANR docking model, we performed a similar simulation using human NMU and human NMUR1, the vertebrate ortholog pair of PBAN and PBAN. NMU has been implicated in the regulation of smooth muscle contraction, blood pressure, and local blood flow, ion transport in the gut, stress responses, cancer, gastric acid secretion, pronociception, and feeding behavior (68). Human NMU is a 25-amino acid peptide with a PBAN-like C-terminal motif, Phe-21-Arg-22-Pro-23-Arg-24 -Asn-25-NH 2 (portions of the sequence identical between PBAN and NMU are underlined), that is highly conserved across a number of vertebrate species and is critical for biological activity (68). Human NMU binds to two receptor types, NMUR1 and NMUR2. For the docking simulation, we modeled the C-terminal five residues and amide of NMU (C5NMU) with a type I ␤-turn similar to C5PBAN, and we generated homology models of NMUR1 using the same structural templates (PDB codes 2Y03 and 3QAK; supplemental Data S4) as those used to construct the PBANR models. The same criteria for evaluating reasonable docking models were used. An appropriate docking model with an affinity of Ϫ10.7 kcal/mol was obtained using a 2-Å separated NMUR1 model derived from the A 2A AR template. This docking model (supplemental Data S5) was chosen as the most appropriate because it met all the above criteria with a minimal receptor separation. Inspection of the predicted C5NMU-NMUR1 contact residues revealed significant conservation with those predicted for C5PBAN and PBANR (Fig. 6, B and C, and supplemental Data S6): Phe-21, four contact residues in TMs 6 and 7; Arg-22, six contact residues in TMs 5-7; Pro-23, eight contact residues in TMs 5-7; Arg-24, eight contact residues in TMs 3 and 5-7; Asn-25, eight contact residues in TMs 2, 3, and 5-7; and the terminal NH 2 , four contact residues in TMs 2, 6, and 7. Similar to C5PBAN and PBANR, no interactions were seen between C5NMU and the ECLs of NMUR1. The observed conservation in the number, location, and physicochemical properties of the predicted interactions comprising the putative PBANR-and NMUR1-binding pockets supports the validity of these docking models.

Comparison of Class A GPCR Agonist Recognition Residues-
To obtain a clearer picture of the ligand-binding pocket across multiple class A GPCRs, the three-dimensional conformations of the ligand-bound PBANR (Fig. 7A) and NMUR1 (Fig. 7B) homology models were compared with agonist-bound crystal structures (Fig. 7, C-E) of ␤1AR (agonist/isoprenaline; PDB code 2Y03), A 2A AR (agonist/UK-432097; PDB code 3QAK), and the neurotensin receptor 1 (NTSR1; agonist/neurotensin 8 -13; PDB code 4GRV). The difference in sizes of the bound ligands in ␤1AR and A 2A AR suggests a structural reason why A 2A AR proved to be the more suitable template for generating the PBANR and NMUR1 models (Fig. 7, A-D). In the NTSR1 structure (Fig. 7E), the C-terminal, biologically active fragment of the peptide hormone neurotensin (RRPYIL-OH, termed C6NTS) is bound. Unlike the molecular models of C5PBAN and C5NMU, this peptide agonist does not form a ␤-turn but rather adopts an extended conformation with its C terminus oriented toward the receptor core. The bound C6NTS, however, does not penetrate as deeply into the receptor as the bound ligands in the ␤1AR and A 2A AR crystal structures or the PBANR and NMUR docking models. Intriguingly, the locations of the agonist recognition sites are well conserved across PBANR, NMUR, A 2A AR, and ␤1AR (Fig. 7, A-D) but less so for NTSR1 (Fig. 7E). The physicochemical properties of the side chains that compose the agonist recognition sites however are exceedingly varied across the GPCRs with only Phe-276 and trp-280 (PBANR numbering) conserved. This diversity of ligand recognition residues in terms of location within the receptor helical bundle and side chain functionalities likely reflects the diverse array of ligands recognized by class A GPCRs.

DISCUSSION
Neuropeptides mediate numerous biological events in both vertebrates and invertebrates. Intriguingly, despite millions of years of evolutionary divergence, subsets of neuropeptides from these two diverse groups are characterized by similar biologically relevant motifs. One such assemblage of peptides

Ligand-Receptor Interactions of PBAN-PBANR and NMU-NMUR
includes the NMU/neuromedin S and the Insecta FXPRL-NH 2 peptides (i.e. PBAN, DH, melanization, and reddish coloration hormone and pyrokinin), both of which share a C-terminal FX 1 PRX 2 -NH 2 motif as the active core (68 -70). Based on these sequence similarities, and those between certain mammalian receptors and putative GPCRs in the Drosophila genome, Hewes and Taghert (71) postulated that peptide ligands with a C-terminal PRX-NH 2 motif co-evolved with their respective receptors. They suggested that PBAN-like peptides could interact with receptors exhibiting sequence similarity to mammalian NMURs. Park et al. (72) extended this hypothesis and demonstrated that H. zea PBAN activated a Drosophila GPCR with high similarity to NMUR. Reciprocal experiments measuring the in vivo effects of NMU in H. zea revealed that the mammalian peptide stimulated pheromone production (20). The fact that NMU can activate PBANR suggests that the PBANR-and NMUR-binding pockets are similar.
An initial effort to model the structural features of the H. zea PBANR on the bovine rhodopsin crystal structure identified 20 residues that potentially compose the inner side of the binding pocket (40). Experimental verification of those sites, many of which map to the extracellular domains, has yet to be demonstrated. Furthermore, the utility of using the rhodopsin structure as a starting point for GPCR homology building can be problematic (73)(74)(75). Mutational studies designed to identify specific ligand-interacting residues in the PBANR pocket have likewise yielded mixed results. Domain swaps of the H. zea PBANR ECLs suggested roles for ECL3 and the N terminus in ligand binding (38). Site-directed mutagenesis of five residues (two in the N terminus and three in the ECL3) within these regions implicated potential N-terminal N-glycosylation sites and ECL3 residues Ser-300 and Phe-303 (38,39). The impaired PBAN-induced Ca 2ϩ response in those mutational studies, however, could have resulted from reduced representation at the cell surface due to compromised trafficking, altered stability of the binding pocket, and/or impaired signaling as well as reduced ligand binding. Moreover, incremental truncations of the B. mori PBANR N terminus, which removed homologous N-glycosylation sites, had no effect on cell surface localization or PBAN-induced receptor internalization, a regulatory process dependent on functional PBANR signaling (37).
To begin to address some of these discrepancies, we sought to combine molecular modeling techniques with mutational analyses to generate a more robust characterization of the PBANR-binding pocket. We mapped the conserved residues from multiple PBANRs and PBANR-like receptors onto homology-based models of the B. mori sequence that were built using spatial coordinates derived from the crystal structures of two GPCRs. This approach facilitated the identification of 27 residues that potentially compose the putative PBAN ligandbinding pocket. The number and location of the residues identified by our mapping approach differ substantially from those previously predicted to include the putative PBAN-binding pocket. Only 3 of the 20 potential residues identified in the rhodopsin-based model of the H. zea PBANR (40) were also identified in our study. In contrast, all of the residues identified based on evolutionary trace analysis of PBANR-related sequences (41) are represented in our prediction. The predic-tive power of that analysis, however, was limited to 11 potential PBANR-interacting sites.
The role the 27 residues have on cell surface trafficking of PBANR, ligand binding, and PBAN-induced Ca 2ϩ mobilization was assessed via sequential Ala substitutions. Four point mutants (S207A, F211A, F212A, and H284A) impeded PBANR trafficking to the plasma membrane (Fig. 2). Based on our homology model, these four residues, which are located in TM5 (S207A, F211A, and F212A) and TM6 (H284A), are predicted to contribute to a number of inter-helix interactions (Fig. 5A) as follows: Ser-207 with Arg-287 (TM6); Phe-211 with three residues in TM3 (Asn-124, Val-127, and Leu-128) and two residues in TM4 (Phe-166 and Thr-170); Phe-212 with one residue in TM3 (Val-127) and four residues in TM6 (Phe-276, Trp-280, Ala-281, and His-284); and His-284 with three residues in TM5 (Ser-208, Phe-212, and Val-213). These interactions strongly suggest that these side chains are important for stabilizing the PBANR conformation and imply that the observed defect in plasma membrane trafficking was attributable to impaired PBANR folding. Indeed, the fluorescence profile of the four mutants was strikingly reminiscent of that seen with endoplasmic reticulum localization (76). Docking simulations suggested that these four residues might also be involved in ligand recognition as C5PBAN contacted all four in the docking model. Consistent with this model, the corresponding residues in A 2A AR, ␤1AR, and ␤2AR are also involved in ligand recognition (Fig. 5D).
Significantly reduced ligand binding and Ca 2ϩ mobilization were seen with 11 point mutants, (E95A, E120A, N124A, V195A, F276A, W280A, F283A, R287A, Y307A, T311A, and F319A). In contrast, three mutants (F209A, F303A, and G315A) exhibited defects in lowered binding, although a single mutant (Y318A) was defective only in the Ca 2ϩ response (Figs. 3 and 4). Surprisingly, despite exhibiting reduced binding, the Ca 2ϩ response profile of cells expressing the F209A, F303A, and G315A mutants was comparable with wild type cells, suggesting that a sufficient amount of ligand had bound to trigger a Ca 2ϩ response. The Y318A mutant is the only mutant with normal ligand binding but impaired signaling abilities, indicating that this tyrosine side chain may be important for the putative conformational change induced in response to PBAN binding. Phe-212, Phe-276, Trp-280, Phe-283, and Phe-319 are highly conserved in class A GPCRs (Fig. 5D) and are involved in formation of the receptor active conformation (77). This likely accounts for the conservation observed in our initial alignments and suggests that they may not necessarily be directly involved in ligand binding. The effects of the mutations could be allosteric rather than impairing ligand recognition. Asn-124 and Phe-212, which are located deep within the core of the receptor, may likewise have allosteric roles that contribute indirectly to ligand binding. Our decision to model the backbone conformation of C5PBAN/C5NMU as a type I ␤-turn was based on NMR structural analyses of PBAN (43,44) and a pheromonotropic eightresidue cyclic peptide containing a highly similar sequence to C5PBAN (underlined), cyclo(NTSFTPRL) (42,44). The docking simulations of C5PBAN-PBANR and C5NMU-NMUR1 highlighted the functional relevance of this conformation in both PBAN and NMU binding. Although artificial widening by 2-3 Å was required (Fig. 6), the C-terminal active core of both peptides was largely accommodated by the ligand-binding pocket of the respective receptors. In contrast, the full-length peptide hormones (33 amino acids in PBAN and 25 amino acids in NMU) would be sterically hindered from entering the sizelimited ligand-binding pockets. As a consequence, the N-terminal flanking regions of C5PBAN and C5NMU, which are not biologically necessary, are proposed to interact with the respective receptor ECLs to strengthen ligand affinity. These varying contact points could potentially function as a selectivity filter for differentiating between ligands with similar active core sequences (e.g. PBAN and DH). In support of this, two residues located in the ECLs, Val-195 (ECL2) and Phe-303 (ECL3), appear to be important for C10PBAN R2K binding (Figs. 3 and 5, B and D) but are not C5PBAN-interacting sites in the docking model (Fig. 6, A and C). Five neuropeptides with the C-terminal FXPRL-NH 2 motif have been identified in B. mori (78), two of which, PBAN and DH, regulate different biological activities and bind two different but homologous receptors, PBANR and DHR. When the functionally important residues of PBANR and the corresponding residues of DHR are compared, all the residues are identical except for Val-195/Gln (ECL2), Phe-303/Pro (ECL3), and Phe-319/Tyr (TM7) (Fig. 5D), which supports the hypothesis that the ECLs of PBANR, DHR, and related GPCRs function as a ligand selection filter (this study and Ref. 38).
The positions of the putative ligand-binding residues in PBANR are highly conserved with the ligand-binding residues in nonpeptidic class A GPCRs (A 2A AR, ␤1AR, ␤2AR, and dopamine D3 receptor), but not with those of NTSR1, a class A peptide ligand GPCR (Figs. 5D and 7 and supplemental Fig. S1). Thus, class A GPCRs can recognize various ligands, both peptidic and nonpeptidic, in distinctive manners using varied residues at different positions.
In summary, the results obtained in this study indicate that the ligand-binding pocket of PBANR/NMUR and related GPCRs recognizes the ␤-turn conformation of their peptidic ligands and that this pocket, similar to nonpeptidic ligand class A GPCRs, is buried relatively deep within the receptor helical bundle. The specific ligand-binding residues, however, are not conserved, which could account for the multitude of ligands recognized by class A GPCRs. In addition, two glutamate residues in PBANR and NMUR1, Glu-95/Glu-117 (TM2) and Glu-120/Glu-142 (TM3), were shown to be critically important for ligand binding and signal transduction and are proposed to be involved in the recognition of the C-terminal amide and the C-terminal residue of both PBAN and NMU. Our results and simulation data have significantly broadened our understanding of the molecular interactions underlying peptidic ligand interactions with class A GPCRs, in particular those between PBAN-PBANR and NMU-NMUR.