Introduction
Because of their participation in numerous cell processes, GPCRs are the targets of approximately half of marketed drugs, and new GPCR ligands continue to be pursued and developed6. Naïve approaches toward GPCR ligand identification typically involve high-throughput screening of a molecular library (102 to 105 unique members) in functional, cell-based assays6, 7, 8. A powerful alternative approach for rapidly isolating new ligands is in vitro peptide selection using mRNA display, which allows access to very high library complexities (>1013) in a robust format4. High-affinity peptide ligands for RNA, small-molecule, and protein targets have been identified by mRNA display selection5.
The crystal structure of the hexahistidine-tagged Mth ectodomain shows that the mature, N-terminal extracellular domain of Mth is a stably folded, glycosylated protein of 195 residues9. Given that the ectodomains of other class B GPCRs maintain recognition of their cognate ligands independently of their transmembrane cores10, 11, we targeted the Mth ectodomain for in vitro selection to isolate putative modulators of Mth signaling. We expressed and purified a specifically biotinylated construct of the Mth ectodomain12 to avoid using the weak hexahistidine epitope as an immobilization tag for selection13. Hence, the Mth ectodomain was homogeneously presented, perhaps mimicking the juxtaposition of the ectodomain and transmembrane domain in the context of the full-length receptor.
We constructed a random 27-mer peptide mRNA display library. After eight rounds of selection, with the final four rounds including preclearing steps on matrix without target and specific elution with free, nonbiotinylated Mth, we obtained a final eighth-round pool that showed high activity for Mth and negligible nonspecific binding (Fig. 1). DNA sequencing of individual clones from the final selection round revealed a highly conserved consensus—[R/P]XXWXXR—which we term the RWR motif (Table 1). This motif was not found in the recently identified Mth peptide agonist Stunted14, and the shortness of the consensus precluded any statistically significant homology to the D. melanogaster proteome.
Figure 1: Selection of a 27-mer peptide library against the Mth ectodomain.
RNase-treated, 35S-methionine–labeled mRNA displayed peptides from each round of selection were assayed for binding to immobilized Mth (black) or to matrix alone (white). Preclearing and competitive elutions were performed in the fifth through eighth rounds to eliminate nonspecific binding peptides.
Full size image (20 KB)
We synthesized several selected peptides for binding analysis by surface plasmon resonance (SPR). Peptides containing the RWR motif demonstrated high affinity (Kd < 30 nM) to the Mth ectodomain (Table 1). A scrambled version of the R8-12 synthetic peptide, as well as W7A and R10A mutants of R8-14, showed no measurable binding by SPR at concentrations up to 5 
M. Additionally, a fluorescently labeled analog of R8-12 was shown to bind full-length Mth expressed in cell culture (Supplementary Fig. 1 online). Hence, despite targeting the Mth ectodomain, selected peptides recognize the full-length GPCR.
It is not clear why arginine and proline should be interchangeable at the first consensus position of the RWR motif. Proline could provide a conformational anchor, whereas arginine makes favorable electrostatic contacts that ultimately result in two peptides with similar binding affinities to Mth. Analysis of the amino acid types at each position in the aligned peptides also reveals several trends that are dependent on whether arginine or proline is in the first position (Supplementary Fig. 2 and Supplementary Table 1 online).
In vitro competition binding studies suggest that the selected peptides share the same binding site. Synthetic, unlabeled peptide R8-01 (10 M) competed with radiolabeled, full-length R8-01 and R8-04 for binding to immobilized Mth ectodomain, which resulted in 96% and 94% reductions in binding, respectively, compared with the amount bound without competitor. N-Stunted, a 30-mer synthetic peptide previously shown to activate the Mth receptor14, also competed for binding to the Mth ectodomain: at 30 M, N-Stunted reduced binding of radiolabeled R8-01 by 79% compared with binding without competitor. These results suggest that the natural ligand binding site is an interaction "hot spot"15 and is at least partially reconstituted by the Mth ectodomain. Alternatively, allosteric competition may occur through Mth conformational changes upon ligand binding.
We determined the crystal structure of the Mth ectodomain in complex with an RWR motif peptide to identify the binding site. Electron density putatively corresponding to the R8-01 15-mer peptide (Table 1) places the binding site near the C terminus of the ectodomain (Fig. 2a). In the context of the full-length receptor, this suggests that the peptides bind at an interface between the Mth ectodomain and extracellular loops (Fig. 2b). These results contradict the hypothesis that the single exposed tryptophan residue in the Mth ectodomain is the binding site for the natural ligand9. Further fluorescence studies with Mth, R8-01 and R8-04 confirmed that Trp120 is not required for peptide binding to Mth (Supplementary Fig. 3 online).
Figure 2: Structure of the Mth ectodomain in complex with the R8-01 15-mer peptide.
(a) Electron density reveals the putative peptide binding site on the Mth ectodomain (shown as a ribbon diagram) from an averaged 3.5-Å Fo – Fc map contoured at 9 
. Trp120 (a previously proposed natural ligand binding site9), Tyr130 (at the R8-01 binding site), and Asp46 and Phe153 (suggested to interact with the extracellular face of the transmembrane domain9) are shown as stick models. (b) Scaled model depicting the full-length structure of Mth (adapted from ref. 9). The transmembrane domain is depicted by the structure of rhodopsin30 as a representative GPCR. Data collection and refinement statistics are available in Supplementary Table 2 online.
The second extracellular loop (EL2) of Mth contains a WXXR peptide sequence (a partial RWR motif) that may interact with the ectodomain in a similar fashion as the RWR motif peptides and form a distinct surface for agonist binding (Fig. 2b). However, at concentrations of up to 150 M, a synthetic EL2 peptide did not compete with radiolabeled R8-01. This may not preclude an in vivo interaction, because the affinity may be enhanced in the full-length receptor, in which the EL2 sequence and the ectodomain are colocalized. Additionally, a high-affinity, optimized interaction may be undesirable for the modulation of natural signaling. Indeed, the bias of in vitro selection for high-affinity ligands may favor the recovery of antagonists rather than agonists, which are evolved for function rather than binding.
The Stunted peptide was previously identified as an agonist for Mth using a cell-based fluorescence reporter system for calcium mobilization, which is a common consequence of GPCR activation14. We used the same technique here to determine whether the selected peptides antagonize Mth signaling. R8-12 and R8-14 were strong antagonists of N-Stunted–induced Mth signaling, with 50% inhibitory concentration (IC50) values of 45 
10 and 115 25 nM, respectively (Fig. 3). The W7A and R10A mutants of R8-14, as well as the R8-12 scrambled peptide, failed to antagonize Mth at concentrations up to 10 M. No signaling was induced by any of the peptides in nontransfected control human embryonic kidney (HEK) cells. RWR motif peptides likely antagonize Mth by blocking interaction with agonist. However, it is also possible that the peptides desensitize the receptor (for example, by triggering internalization) or block GPCR oligomerization, which may be required for proper signaling.
Figure 3: RWR motif peptides are antagonists of Mth signaling.
(a) Mth activation by Stunted results in intracellular calcium mobilization and enhanced fluorescence. The N-Stunted agonist peptide (20 M final) was added to HEK-Mth cells preincubated with and without the indicated concentration of R8-12 peptide. The dashed line indicates a control in which only buffer (without Stunted agonist) was added. (b) Concentration dependence of the inhibition of Mth signaling by the R8-12 and R8-14 peptides. The fluorescence values at a time point 
13 s after the addition of N-Stunted agonist (10 M final) are expressed as a fraction of the fluorescence observed in the absence of antagonists ( s.d. when more than one trial was performed).
Given that downregulation of mth increases life span3, we speculated that the in vivo application of a Mth antagonist would affect D. melanogaster longevity. We generated transgenic flies that constitutively express the R8-12 peptide using the UAS/GAL4 system16. Expression of the R8-12 peptide extended the mean and maximal adult life spans at 29 °C by 38% and 26%, respectively (Fig. 4a). Two other independent UAS–R8-12 insertion lines showed similar life span increases (Supplementary Fig. 4a online). F1 heterozygous lines, each carrying one of the two P elements (da-GAL4 or UAS–R8-12), showed life spans similar to those of parental controls, which suggests that the longevity seen with peptide expression is not due to heterosis (Supplementary Fig. 4b). Additionally, expression of a mutant R8-12 peptide in which critical residues of the RWR motif were mutated to alanine resulted in no extension of life span (Fig. 4b). Male survival curves at 25 °C also show life span extension with R8-12 expression (Supplementary Fig. 4c). These results suggest that the expressed R8-12 peptide antagonizes Mth in vivo. However, because the role of other Mth-like receptors in longevity is unknown, and because other Mth-like extracellular domains have significant sequence similarity to the Mth ectodomain (up to 60% identity), the action of the R8-12 peptide may be mediated by other proteins. Studying the effects of R8-12 expression on other phenotypes observed in the mth mutant may reveal the relationships between aging, synaptic transmission and fertility3, 17, 18, 19.
Figure 4: Expression of the Mth antagonist R8-12 peptide extends fly life span.
(a) Male life span for the UAS–R8-12 transgene insertion line (41.6). 
, heterozygous control for the daughterless-GAL4 driver (da-GAL4/+). 
, heterozygous control for the UAS–R8-12 peptide construct in the absence of the driver (UAS–R8-12/+). 
, the combination in which the da-GAL4 driver, by activating UAS promoter elements, drives expression of the R8-12 peptide (UAS–R8-12/da-GAL4). Data are mean s.d. (b) Male life span for a mutant UAS–R8-12 transgene (Ala2). Flies were maintained at 29 °C, and approximately 90 to 120 flies were used for each trial.
In vitro selection has emerged as a powerful alternative to screening approaches for the identification of modulators of G protein signaling20, 21. Our work provides a clear strategy for designing functional ligands for class B GPCRs, a protein family associated with a number of human diseases2. New expression and presentation platforms for transmembrane proteins are needed to broaden the utility of mRNA display selection for developing novel ligands. Selections targeting full-length GPCRs expressed in cells remain an unattractive option owing to the large background of cell-surface proteins. The recent successes in assembling pure GPCRs in paramagnetic proteoliposomes22 and in a reconstituted bilayer on a biosensor surface23, 24 may be favorable alternatives for presenting a transmembrane protein for selection. The method we describe for the identification of GPCR effectors may provide new tools that are applicable to aging and other fields.
Methods
mRNA display library preparation.
The initial DNA pool was generated by PCR (0.1 M initial template, 5 total cycles of PCR amplification using Herculase DNA polymerase (Stratagene)) of the 142.1 template(5'-TTAAATAGCGGATGCACGCAGACCGCCACTAGT(SNN)27CATTGTAATTGTAAATAGTAATTGTCCC-3'; where N is A, C, G, or T, and S is C or G) with the primers 47T7FP (5'-GGATTCTAATACGACTCACTATAGGGACAATTACTATTTACAATTAC-3') and 21.2 (5'-TTAAATAGCGGATGCACGCAG-3'). This library encoded a T7 promoter for transcription, a 5' untranslated region sequence, and an open reading frame for the peptide MX27TSGGLRASAI. The template DNA was transcribed and purified as described previously25. Cross-linking of a puromycin-psoralen linker to the mRNA was performed as described using oligo 28A.1 (5'-(Ps)UAGCGGAUGC(dA)16(S9)2(dCdC)Pu-3'; where unlabeled bases are 2'-OMe RNA, Ps is psoralen C6, S9 is spacer phosphoramidite 9, and Pu is puromycin-CPG; Glen Research)26.
In vitro translation with the mRNA-28A.1 library was performed in Red Nova Lysate (Novagen) according to the manufacturer's instructions with optimized conditions (100 mM KOAc, 0.5 mM MgOAc, 0.4 M mRNA-28A.1, and 25 M overall L-methionine; 10 ml total reaction volume) and 35S-methionine (0.5 mCi ml-
1 final). RNA-peptide fusions were salt-treated and purified as described previously25. Fusions were reverse-transcribed (oligo 21.2) and desalted into Mth buffer (50 mM HEPES-KOH at pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.1% (w/v) BSA, 1 g ml-
1 yeast tRNA and 0.05% Tween 20) by gel filtration (NAP-25, GE Healthcare). Based on the 35S-methionine incorporated into the RNA-peptide fusions and an average of 1.3 methionine residues per peptide, the initial complexity of the library was approximately 1.5 
1013.
In vitro selection.
Fusions of peptides with RNA bound to complementary DNA were incubated with 0.1 ml of Mth-agarose at 4 °C for 1 h, then filtered and washed with 4 1 ml Mth buffer followed by 2 1 ml Mth buffer without BSA or tRNA. Bound fusions were eluted with 2 100 l of 0.15% SDS. After removal of the SDS (SDS-OUT, Pierce), fusions were isopropanol-precipitated (50 g ml-
1 linear acrylamide, 1/40 volume of 3 M NaOAc at pH 5.2, and 1 volume of isopropanol). The reduced salt used for isopropanol precipitation was necessary to prevent inhibition of subsequent PCR, owing to the high salt introduced by the SDS-OUT reagent. Precipitated cDNA was PCR-amplified to produce a new double-stranded DNA pool.
Further rounds of selection were performed as described for the initial round except that in vitro translation reactions were smaller (0.3 ml), less immobilized Mth was used for the selective step (20 l), and in rounds 5 through 8, bound fusions were eluted by competition with nonbiotinylated Mth (0.5 mg ml-
1) in Mth buffer without BSA or tRNA. Additionally, rounds 5 through 8 included a preclearing step in which the precipitated RNA-cDNA–peptide fusions were passed through columns containing NeutrAvidin-agarose and/or protein G-sepharose to remove peptides with high nonspecific binding for the immobilization matrix.
Kinetics determination by SPR.
SPR measurements were performed at 25 °C on a BIAcore 2000 instrument equipped with research-grade streptavidin sensor chips. Biotinylated Mth was immobilized to a surface density of 450 to 700 response units. HBS-EP (10 mM HEPES at pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.005% polysorbate 20 (Tween 20)) was used as the running buffer for all experiments. To collect kinetics data, a concentration series of each peptide was injected for at least 60 s at a flow rate of >45 l min-
1. Raw data was processed with Scrubber (http://www.cores.utah.edu/interaction/scrubber.html) and globally fit with CLAMP using a 1:1 bimolecular interaction model27. Sensorgrams for the R8-04 peptide are shown in Supplementary Figure 5 online.
Cell-based GPCR signaling assay.
Calcium response assays were performed with HEK 293 cells stably expressing Mth essentially as described14. We used the Mth-B splice variant, which has a longer and different C-terminal cytoplasmic tail than Mth-A. Fluorescence spectra were divided by a baseline average, calculated from the region of data before the addition of N-Stunted. Further details are provided in the Supplementary Methods online.
Life span assays.
P-element transformants with the full-length R8-12 peptide transgene were generated in a w1118 background by standard techniques28. R8-12Ala2 is a W5A and R8A mutant of the full-length R8-12 sequence. UAS-controlled R8-12 lines (UAS–R8-12) were crossed with da-GAL4, which had previously been outcrossed with w1118. F1 heterozygous adults (0–4 d old) were transferred to new bottles and aged for 2 d, and males were separated under CO2 anesthesia (30 males per vial). After allowing the flies to recover overnight at 25 °C, flies (3–7 d old) were incubated in a light- and humidity-controlled environment at the indicated temperature. The UAS–R8-12 and da-GAL4 lines were each crossed with w1118 to produce F1 heterozygous animals for controls. Flies were transferred to fresh food vials ((0.45% agar, 5% dextrose, 2.5% sucrose, 8.3% corn meal and 1.5% dried yeast, all (w/v)) with phosphoric and propionic acids supplemented to prevent mold, as previously described29) every 3–4 d and scored for survival.
Other methods.
Additional methods are provided in the Supplementary Methods online.
Accession codes.
Protein Data Bank: The atomic coordinates for the structure of the Mth ectodomain with a peptide antagonist have been deposited under PDB ID code 2PZX.
Note: Supplementary information is available on the Nature Chemical Biology website.
Author Contributions
W.W.J., A.P.W., Jr. and S.L.D. performed research; all authors designed research and discussed results; W.W.J., A.P.W., Jr., S.B. and R.W.R. wrote the manuscript.
