Discovery of Human Signaling Systems: Pairing Peptides to G Protein-Coupled Receptors

Abstract

The peptidergic system is the most abundant network of ligand-receptor-mediated signaling in humans. However, the physiological roles remain elusive for numerous peptides and more than 100 G protein-coupled receptors (GPCRs). Here we report the pairing of cognate peptides and receptors. Integrating comparative genomics across 313 species and bioinformatics on all protein sequences and structures of human class A GPCRs, we identify universal characteristics that uncover additional potential peptidergic signaling systems. Using three orthogonal biochemical assays, we pair 17 proposed endogenous ligands with five orphan GPCRs that are associated with diseases, including genetic, neoplastic, nervous and reproductive system disorders. We also identify additional peptides for nine receptors with recognized ligands and pathophysiological roles. This integrated computational and multifaceted experimental approach expands the peptide-GPCR network and opens the way for studies to elucidate the roles of these signaling systems in human physiology and disease. VIDEO ABSTRACT.

Keywords: GPCR; deorphanization; endogenous ligand; evolution; genomics; machine learning; orphan receptor; peptide ligand; pharmacological screening; receptor internalization.

Graphical abstract

Figure S1

Peptide-Receptor Pairing Approach, Related to Figures 2, 3, 4, and 5 21 orphan receptor targets were selected based on shared characteristics of known peptide-activated GPCRs. A library of 218 peptides was generated using a proteome-wide machine-learning approach. Peptides were screened using three complementary functional assays. Putative peptide-oGPCRs pairings were validated using additional assays. Predicted cleavage variants of discovered peptide agonists were tested to gain insights into determinants of peptide potency.

Figure 1

The Human G Protein-Coupled Receptor-Ligand System (A) GPCRs represent the predominant targets for endogenous ligands. Peptides are more numerous, larger and bind with higher affinity than non-peptide ligands. From the top: (1) distinct endogenous ligands by target family; (2) endogenous GPCR ligands, of which “principal” ligands are considered most physiologically relevant; (3) peptide and small-molecule binding receptors, of which “paired” ones have a known principal endogenous ligand; and (4) ligands per receptor and vice versa (averages). (B and C) Ligand molecular weight distribution (B) and cognate receptor affinity (C) (boxplots show a median and interquartile range of 1.5; Wilcoxon rank-sum test, p < 1 × 10⁻⁵). Data are from the Guide to Pharmacology database (Harding et al., 2018). (D) GPCR-ligand systems vary in complexity from 1:1 to many:many (gray circles show numbers of each system; data are shown in Table S1). See also Figure S2.

Figure S2

The Peptidergic Signaling System Has Been Shaped by Co-evolution of Ligands and Their Receptor Targets, Related to Figure 1 (A) Receptors and their ligands are ubiquitously expressed in human organs and tissues. The data is ordered by mean level of peptide receptor expression. Peptide receptors consistently had lower median expression levels than non-peptide GPCRs. Similar peptide receptor-ligand expression levels were observed for liver and smooth muscle, whereas granulocytes and lung tissue had higher expression of receptors than ligand precursors (Wilcoxon rank sum test; P values: < 1x10⁻⁵ (granulocytes) and 0.02 (lung)). (B) Evolutionary fingerprints indicated conservation (gray) or absence (white) of receptor and peptide precursor gene orthologs in 313 species (representatives shown). The fingerprint identity (%) reflects the evolutionary relationship of peptide-receptor pairs. Photos from Ensembl genome database project. (C) The average percentage identity of evolutionary receptor-ligand pairs for all endogenous receptor-ligand pairs is increased when fingerprints of ligands for the same receptor are merged and is greater than for a random protein pair (permutation tests by performing 10,000 randomizations, Wilcoxon rank sum test P value < 1x10⁻⁵). (D) Jaccard index similarity of human peptide receptor and ligand precursor repertoires (n = 131 and 130, respectively) to selected species ordered by evolutionary distance. Data in Tables S1 and S2.

Figure 2

Universal Precursor Processing and Peptide Ligand Gene Conservation Hotspots (A) Potential precursors can be mined from the human proteome based on the presence of secretion signal peptides and an unknown or ligand-precursor-like function (Table S4). (B) The vast majority of GPCR peptide ligands are cleaved from precursors at specific dibasic sites. (C) GPCR peptide ligands are more evolutionarily conserved than random sequences of similar length (up to 45 residues) (Wilcoxon rank-sum test, p < 1 × 10⁻⁵). (D) Human peptide ligands can be deduced from precursor cleavage sites and conservation hotspots. The example depicts the pro-opiomelanocortin precursor containing endogenous ligands for melanocortin (α-MSH, β-MSH, γ-MSH, and ACTH) and opioid (β-endorphin) receptors.

Figure 3

Peptide Receptors Share Distinct Sequence and Structural Characteristics (A) The majority of class A GPCRs cluster by endogenous ligand type based on ligand-interacting residue analysis with multi-dimensional scaling. (B) Peptide receptors with a structure (n = 21, left) share a characteristic β sheet (green) substructure (left) and sequence (right) in extracellular loop 2 (ECL2), which includes a conserved cysteine, Cys^45×50 (red, center). (C) A long ECL2 segment (>20 residues) after Cys^45×50 is an overrepresented feature of peptide/protein receptors (Wilcoxon rank-sum test, p < 1 × 10⁻⁵). (D) Principal-component analysis of receptor structures in a 2D plot (top left) and dendrogram (bottom) demonstrate separation of peptide (green), non-peptide receptors (beige), and outliers (gray). Differences are predominantly found in the extracellularly facing ligand-binding domain, as shown by residue displacements from the mean (right). (E) Ligand-binding pocket volumes are larger in peptide than non-peptide class A receptors. See also Table S3 for related 3D PCA and ECL2 motif data.

Figure S3

Receptor Expression and G Protein Coupling in Orphan GPCR Cell Lines, Related to Figure 4 (A) Cell-surface expression of induced oGPCRs measured by ELISA. Data shown as mean ± SEM from n = 3-7 independent experiments performed in triplicate. (B-D) Constitutive cAMP and IP₁ production (i.e., in the absence of ligand) upon orphan receptor induction provide insights into G protein-coupling. Data shown as mean ± SEM from n = 3-5 independent experiments, except for BRS3/BB₃ (n = 2 in (B) and (C)), GPR32 (n = 2 in (B)) and GPR3 (n = 2 in (C)). (E) For three receptors, we discovered constitutive G protein-coupling unreported in Guide to Pharmacology or literature (Doi et al., 2016, Harding et al., 2018, Inoue et al., 2012, Martin et al., 2015, Muppidi et al., 2014, Pera et al., 2018, Suply et al., 2017), and for GPR15 robust G_i/o signaling in cAMP accumulation assays.

Figure 4

General versus Assay-Specific Responses and Novel Peptide-Receptor Pairings (A) The multifaceted screen of 218 peptides identified a variety of multiple and single-assay responses, including hits for all 21 predicted peptide receptors. Mass redistribution data revealed repeat hitters (denoted with asterisks) that reflect peptide-dependent responses from endogenous targets. Screening results for additional class A orphan and peptide GPCRs are provided in Table S6. (B) Pairing of 17 peptides with five orphan receptors. Colored circles show pEC₅₀ values and concentration-response curves the most potent ligand for each receptor. Other assays used were G_q/11 (IP₁), G_s and G_i/o (cAMP), and β-arrestin recruitment (PathHunter). An asterisk indicates a new cleavage variant of a known GPCR peptide ligand with the amino acid range of the cleaved peptide shown in subscript; empty circles indicate inactivity. All data represent mean ± SEM for 3–4 independent experiments, each performed in triplicate. Table S7 provides all related data and data for indicative pairings (tested in a single assay) for GPR17, GPR161, GPR176, GPR183, and MAS1.

Figure S4

Peptide-GPCR Pairings across Multiple Assay Formats, Related to Figures 4 and 5 (A) Concentration-response measurements of BB₃, GPR1, GPR15, GPR55 and GPR68 peptide ligands with activity in at least two assays. Assays used were ligand-dependent dynamic mass redistribution, receptor internalization (TR-FRET), G_q/11 (IP₁), G_s (GloSensor cAMP) and β-arrestin recruitment (Tango and PathHunter). (B) Indicative pairings from a single assay (Tango) for additional receptor targets. All data represent mean ± SEM for n = 3-4 independent experiments performed in triplicate, except for BB₃/peptide 158 in IP₁ accumulation (n = 2) and GPR55/peptide 156 (n = 1). Pharmacological parameters are provided in Table S7.

Figure 5

Potential Peptide Cleavage Variants Elicit Increased GPR15 Signaling Responses (A) Evolutionary trace and cleavage site (gray bars) analysis of the GPR15L gene-encoded precursor presents potential alternative peptide cleavage variants. (B–D) GPR15-mediated responses for GPR15L cleavage variants in (B) cAMP inhibition, (C) mass redistribution, and (D) receptor internalization assays. The most potent peptide is the longest, 57-residue form the recently named “GPR15L” (Suply et al., 2017) (excluded in C because of assay interference). Data represent mean ± SEM for 3–4 independent experiments performed in triplicate. Related pharmacological data for GPR15 as well as for cleavage variants of peptides activating BB₃, GPR55, and GPR1 are shown in Table S7 and Figures S4 and S5.

Figure S5

Peptide Ligand Variants Elicit Differential oGPCR Signaling Responses, Related to Figure 5 (A) BB₃ responses for neuromedin B (NMB) and gastrin-releasing peptide (GRP) cleavage variants. (B) GPR1 responses for gastrin-releasing peptide, osteocrin (OSTN) and cholecystokinin (CCK-33) cleavage variants. (C) GPR55 responses for PACAP cleavage variants. Mass redistribution data show an apparently biphasic response for PACAP peptides, which could indicate an additional intracellular signaling pathway mediated via GPR55. Numbers in parenthesis indicate positions in the full-length protein.All data represent mean ± SEM for n = 3-4 independent experiments performed in triplicate.

Figure S6

mRNA Expression Profiles of Paired Peptide Ligand Precursors and Receptors, Related to Figure 4 Human tissue expression profile of peptide-receptor pairs. Some receptors or precursors have low abundance (e.g., FMRF amide-related peptide) or restricted expression patterns (e.g., BB₃, gastrin-releasing peptide), whereas the majority are ubiquitously expressed (source data from massive mining of publicly available RNA-seq data for 52 tissues provided by ARCHS4). Peptides are cleaved and secreted from their tissue of origin and may act at distant tissues.

Figure 6

Disease Associations for Novel Peptide-Receptor Pairs Diseases associated with paired receptors and peptide precursors from https://www.opentargets.org. Open Targets presents associations with therapeutic areas by agglomerating data; e.g., genome-wide associations, genetic variants, expression and animal models. Disease association scores between 0 and 1 (color intensity) summarize the strength of evidence. For precursors with disease correlation similar to the associated receptor target, the Pearson correlation value is indicated. UniProt names for precursors are shown, with peptide library designation in parentheses (Table S4).

Figure S7

Confirmation of Proposed Pairings and Secondary Ligands for Known Peptide Receptors, Related to Figure 7 (A) Of the 14 pairings proposed in literature, half were reproduced in mass redistribution and/or internalization assays (Table S5). Note: P2RY10–lysophosphatidylserine (LPS) activity was marginal. ^∗We tested cleavage variants of cortistatin (CST-14) and somatostatin (SRIF-28). (B-C), Nine known peptide receptors were activated by their cognate agonists (internal controls) and, unexpectedly, 22 additional peptides. These indicate as yet unappreciated cross-pharmacology (Table S7). All data represent mean ± SEM for n = 3 independent experiments performed in triplicate.

Figure 7

Expansion of the Human Peptidergic Receptor Signaling System The new pairings (colored lines) increase the number of known ligand-receptor connections (edges) from 348 to 407 (putative peptide ligands from 185 to 214 and putative peptide receptors from 120 to 130). Ligand-receptor systems are shown with increasing ligand-to-receptor ratios (top to bottom). There are more ligands per receptor in both the established and novel peptidergic receptor systems.