System-wide mapping of peptide-GPCR interactions in C. elegans

Graphical abstract.

Summary

Neuropeptides and peptide hormones are ancient, widespread signaling molecules that underpin almost all brain functions. They constitute a broad ligand-receptor network, mainly by binding to G protein-coupled receptors (GPCRs). However, the organization of the peptidergic network and roles of many peptides remain elusive, as our insight into peptide-receptor interactions is limited and many peptide GPCRs are still orphan receptors. Here we report a genome-wide peptide-GPCR interaction map in Caenorhabditis elegans. By reverse pharmacology screening of over 55,384 possible interactions, we identify 461 cognate peptide-GPCR couples that uncover a broad signaling network with specific and complex combinatorial interactions encoded across and within single peptidergic genes. These interactions provide insights into peptide functions and evolution. Combining our dataset with phylogenetic analysis supports peptide-receptor co-evolution and conservation of at least 14 bilaterian peptidergic systems in C. elegans. This resource lays a foundation for system-wide analysis of the peptidergic network.

Introduction

Peptide hormones and neuropeptides are one of the largest and most widespread classes of signaling molecules that regulate physiology and behavior in all animals. Although bioactive peptides are also secreted from non-neuronal tissues, most are released by the nervous system and signal via G protein-coupled receptors (GPCRs). Peptide messengers are among the most ancient signaling molecules, likely predating the evolution of neurons,^1,2 and many peptidergic systems have broadly conserved functions across bilaterian animals.^3–7 Because of their crucial roles, peptide-GPCR pairs are increasingly gaining traction as potential therapeutic targets for human diseases.^8–11 GPCRs represent over 34% of the targets for prescribed drugs, and abnormalities in peptidergic transmission have been implicated in multiple diseases.^8,10,12 Indeed, these signaling systems are involved in a broad range of biological processes, including development, reproduction, feeding, nociception, learning, and memory, as well as the regulation of behavioral states, such as sleep and arousal.^{2,3,7,13–19}

While several efforts have been made to map the brain-wide expression of peptide and GPCR genes,^20–23 our insight into the structure of the peptidergic signaling network is limited, as huge gaps remain in our understanding of peptide-receptor interactions in animals. Neuropeptides mainly bind to GPCRs that activate second-messenger pathways and, because the number of second-messenger pathways is limited, the diversity of peptide-GPCR couples is thought to provide the complexity required for brain function.²⁴ Most animal genomes harbor around 100 to 150 putative peptide-GPCR genes and encode hundreds of bioactive peptides that are cleaved from prepropeptides, containing one or multiple peptide sequences.^25–31 Some receptor genes also code for multiple GPCR isoforms, which may have different ligand interactions.^25,26 In addition, peptide ligands can bind to multiple receptors and vice versa.^27–29 These complex interactions further diversify the peptidergic network.

Peptide-GPCR couples are challenging to predict from prior knowledge, such as neuronal connectivity or sequence information, as peptide ligands have only short sequences and activate GPCRs that are often distantly located from their release site.^30–32 Reverse pharmacology has proved to be a successful approach for the deorphanization of GPCRs by expressing receptors in heterologous cells and identifying their ligand(s) in compound libraries.³³ This way, peptide ligands have been found for a broad range of GPCRs, including at least 138 receptors in humans, 36 in Drosophila melanogaster, and 29 in Caeno-rhabditis elegans.^29,34–50 Nevertheless, many predicted peptide GPCRs remain orphan receptors, hampering investigations into their functions.⁵¹ Although some large-scale efforts have been undertaken, such as for Platynereis and human peptide GPCRs,^29,52,53 most deorphanization studies have focused on single receptors or a subset of GPCR candidates, which may overlook complex interactions of peptides with multiple receptors.^29,54 To gain a deeper understanding of the peptidergic network, system-wide approaches are required that examine as many interactions as possible between peptides and GPCR isoforms encoded in animal genomes.

Due to its compact and well-defined nervous system, the nematode C. elegans represents an attractive model to map and functionally characterize the peptidergic network at an organismal level. The complete anatomy and wiring diagram of its nervous system has been defined, facilitating studies of peptidergic interactions with the wired circuitry.^55–58 The C. elegans genome encodes over 300 peptides that are classified as FMRFamide-like (FLP), insulin-like (INS), and other neuropeptide-like (NLP) peptides.^59,60 In addition, its genome is thought to encode around 150 putative peptide GPCRs.^36,61 Many of these peptidergic systems are shared between nematodes and other animals, including humans, and have conserved functions.^6,31,62,63 Transcriptional profiles are available for individual neuron classes among the 302 neurons of the C. elegans hermaphrodite nervous system.^21,64 As in other animals, distinct neuron types in C. elegans were found to express unique codes of neuropeptide genes and receptors, suggesting different roles for neuron classes in transmitting peptidergic signals.^20,21,23 However, the relationships between sender and receiver cells in the extensive peptide network have not been determined, as over 80% of the peptide-GPCR candidates in C. elegans are still orphan, and target receptors for most C. elegans peptides remain unknown.^36,61

Here, we describe a large-scale reverse pharmacology pipeline for pairing C. elegans peptides to GPCRs and identify 461 cognate peptide-receptor couples. A system-wide screening approach allows us to discover novel ligands of known and orphan peptide GPCRs and reveals several types of complex ligand-receptor interactions in the peptidergic signaling network, including promiscuous receptors and different interactions encoded within single peptide genes. We combine our dataset with phylogenetic analysis to delineate the evolutionary history of C. elegans peptidergic systems. This resource of ligand-receptor pairs, combined with single-cell transcriptome data,²¹ also provides a basis for constructing a connectome of neuropeptidergic signaling in C. elegans.⁶⁵

Results

Large-scale screening for C. elegans peptide-GPCR pairs

The C. elegans genome has at least 149 genes for putative peptide GPCRs (Table S2), most of which encode orphan receptors.^36,61,66,67 To map the peptide-GPCR interaction network, we set up a reverse pharmacology platform in which we systematically screened for peptide ligands of these receptors in cultured cells (Figure 1A). We first constructed a GPCR library by cloning receptor cDNA sequences in a pcDNA3.1 vector. Multiple GPCR isoforms were included, as receptor variants may be differentially activated by peptide ligands.^25,26,68 In total, we cloned 161 receptors, covering 87% of peptide-GPCR genes that are known and predicted in C. elegans (Table S2; Data S1).

Figure 1. Systematic ligand screening of peptide-GPCR candidates in C. elegans.

(A) Platform for identifying peptide ligands of C. elegans GPCRs. Candidate receptors are cloned under control of the cytomegalovirus immediate-early promoter (pCMV) for heterologous expression in CHO cells. Each GPCR is expressed together with the calcium-activated photoprotein aequorin and the promiscuous human Gα₁₆ subunit and is screened with a synthetic library of C. elegans FLP and NLP peptides. Upon GPCR activation, phospholipase Cβ (PLCβ) is activated and hydrolyzes phosphatidylinositol bisphosphate (PIP₂) into diacylglycerol (DAG) and inositol trisphosphate (IP₃), which activates IP₃-dependent calcium channels, resulting in calcium release from intracellular storage sites and increased luminescence. Screening of 55,384 individual peptide-GPCR interactions identified putative ligands for 114 out of 161 GPCRs. See also Figure S1, Tables S1–S4, and Data S1.

(B) For each GPCR, putative hits are iteratively selected and each interaction is validated by measuring concentration-response curves, from which half-maximal effective concentrations (EC₅₀) values are calculated. Out of 776 putative hits, 459 pairs show concentration-dependent GPCR activation. See Figures S2–S7 and Tables S4–S6 for tested and validated couples.

To identify peptide-receptor couples, we screened each GPCR with a comprehensive library of 344 synthetic C. elegans peptides, belonging to the FMRFamide-like (FLP) and neuropeptide-like protein (NLP) families, that we compiled from biochemical identifications and bioinformatic predictions (Table S3). Insulin-like peptides were not included in the library, as they mainly activate receptor tyrosine kinases and are structurally less well characterized in C. elegans.^69,70 We expressed C. elegans GPCRs in Chinese hamster ovary (CHO) cells, which are widely used for heterologous GPCR expression, and screened for ligand-receptor interactions in an established calcium mobilization assay (Figure 1A).^{5,6,38,39,44,52,71–75} Each receptor was transiently expressed in CHO cells that stably express the calcium-sensitive bioluminescent protein aequorin together with a promiscuous Gα₁₆ protein. The Gα₁₆ subunit allows coupling of many GPCRs to phospholipase Cβ (PLCβ) activity, resulting in calcium release from intracellular storage sites and an increase in luminescence (Figure 1A).^75–77 Using a high-throughput fluorometric imaging plate reader (FLIPR) screening system, we challenged each GPCR with peptides from the synthetic library (at 10 μM), testing 55,384 pairs, and monitored GPCR activation by measuring luminescent responses.

To identify hits in the peptide library, we manually inspected peptide-evoked luminescent responses and determined an activation value for each peptide-GPCR pair based on Z scores (see STAR Methods). We calculated Z scores for each pair, standardized the scores for each GPCR by dividing by its maximal activation, and log2 transformed this fraction to rank individual peptide-GPCR couples (Table S4). Known peptide-GPCR couples with sub-micromolar half-maximal effective concentrations (EC₅₀ < 1 μM) nearly all showed Z scores above 20 and log2-standardized values above -1.2 in our assays (Figure S1A; Table S4). Therefore, we considered all peptide-GPCR couples with a strong calcium response over background (Z score >20), reaching at least 40% of that of the receptor’s maximum Z score (log2-standardized Z > -1.2), as potential hits in the library. Based on these criteria, we found 416 putative hits for 114 GPCRs derived from 81 peptide- and 95 receptor-encoding genes (Figure S1; Table S4).

Validation of peptide-GPCR interactions

We first screened for peptide-GPCR pairs by testing high peptide concentrations (10 μM). However, peptide GPCRs typically show concentration-dependent activation with EC₅₀ values in the nanomolar range.³² To determine whether the potential peptide hits are cognate GPCR ligands, we quantified GPCR activation for decreasing peptide concentrations (Figure 1B). Prior to concentration-response measurements, synthetic peptides were purified by reversed phase high-performance liquid chromatography (HPLC) and verified by mass spectrometry (see STAR Methods). We then performed concentration-response tests for all putative hits identified in library screens. Because false-positive pairs could bias hit selection, we assigned hits by iterative ranking of peptide-GPCR couples. When the top interaction was not confirmed by concentration-response tests, we restandardized the Z scores for that receptor relative to the next top hit in the screen. Concentration-response assays were then repeated until all pairs matching the hit criteria were tested (Z score >20, log2-standardized Z > -1.2). For each hit, we screened all peptides derived from the same peptide-encoding gene, which may be co-released from peptidergic cells (except for promiscuous receptors; see STAR Methods). In total, we tested 776 peptide-GPCR couples, of which 459 pairs showed concentration-dependent receptor activation (Figures 1B and S2–S7; Table S4). None of the peptide ligands evoked a calcium response in CHO cells transfected with an empty pcDNA3.1 plasmid (Table S5).

Among the 459 validated pairs, we identified ligands for 66 GPCRs (including isoforms) and receptors for 151 peptides (Table S4). We characterized GPCRs for peptides of all 31 flp genes and 22 nlp encoding genes (Table S1; Figures S2–S7). These include 39 receptors that are activated only by RFamide peptides, 24 GPCRs that interact with peptides of the NLP family, and three receptors that are activated by peptides from both families. EC₅₀ values, as calculated from concentration-response curves, range between 0.1 pM and 22 μM but are in the sub-micromolar range for 73% of the couples (Figure 2A; Table S6), suggesting that the majority of the identified interactions are likely cognate pairs.

Figure 2. System-wide characterization of peptide GPCRs identifies a complex network of ligand-receptor interactions.

(A) Cumulative frequency plot for log(EC₅₀) values of 459 validated peptide-GPCR couples.

(B) Venn diagram of peptides in the synthetic library that have been biochemically isolated⁵⁹ and/ or for which a receptor has been identified.

(C) Heatmap of log10(EC₅₀) values for validated peptide-GPCR pairs. When a GPCR interacts with multiple peptides from a single peptide-encoding gene, only the log(EC₅₀) value of the most potent ligand is included. Histograms indicate the number of interactions for each GPCR (top) or peptide precursor (right). See also Figures S2–S8 and Tables S1, S6, and S7.

To probe whether peptide-GPCR couples are cognate pairs in vivo, we set up a collaborative network through which researchers could prioritize GPCRs for ligand screening and functionally test peptide-receptor interactions in C. elegans during our study (via https://worm.peptide-gpcr.org). This way, over 10 novel peptide-GPCR couples in our dataset have already been validated in in vivo studies.^{62,63,78–84} These revealed roles of peptide systems in the regulation of foraging, nociception, locomotory arousal, aversive learning, and other behaviors,^{62,63,78–80} and illustrate that our screening approach allows identifying physiologically relevant interactions. Here, we present the complete resource of peptide-GPCR pairs (Figures S2–S7); in vivo studies that have validated and reported functional ligand-receptor couples in this dataset are listed in Table S1.

Discovery of novel peptide receptors and additional ligands for peptide GPCRs

Deorphanization efforts previously identified peptide-receptor couples for 16 nlp, 19 flp, and 29 peptide-GPCR genes in C. elegans (Table S1). We compared these reported interactions to the 459 peptide-GPCR couples discovered in our system-wide screen and confirmed known agonists for 23 receptors (Table S1). Our dataset recapitulates the majority of high-potency interactions (EC₅₀ < 500 nM) reported previously (Table S1). For eight peptide GPCRs, we discovered novel peptide ligands in addition to known agonists. By systematically screening for peptide-GPCR interactions, we identified 79 additional ligands for DMSR-1, NPR-6, NPR-11, NPR-12, NTR-1, CKR-1, and two EGL-6 receptors (Table S1). This underscores the importance of system-wide screening approaches to map peptide-GPCR interactomes.

In addition, our dataset reveals peptide ligands for 35 orphan GPCRs (including isoforms; Table S1). These include receptors for peptides from 58 peptide-encoding genes (31 flp and 27 nlp genes). We also asked whether our screen might uncover receptors for predicted peptides that have as yet not been biochemically isolated. To probe this, we compared our dataset of peptide-GPCR couples to the C. elegans neuropeptidome, a resource of peptide sequences identified by mass spectrometry.⁵⁹ Of the 344 peptide sequences in our synthetic library, 206 peptides were detected in whole-mount peptide extracts using mass spectrometry. Among these, 115 peptides were identified as agonists in our screen (Figure 2B). In addition, we found receptors for 36 peptides that so far have not been biochemically isolated (Figure 2B; Table S7), which suggests bioactive roles for these peptides.

A system-wide screening strategy reveals promiscuous peptide GPCRs and ligands

Peptides of the FLP family share a C-terminal RFamide motif, whereas NLPs form a structurally diverse class of peptides that lack a common consensus sequence. In agreement with these structural differences, most peptide GPCRs in our dataset are activated by either FLP or NLP ligands. Only three receptors (NTR-1, FRPR-16, and NPR-22) have ligands from both peptide families. For NPR-22, NLP and FLP ligands also show clear sequence similarity: NLP-72 peptides have a C-terminal RYamide motif that resembles the RFamide sequence of FLP ligands (Figure S2; Table S3).

To further probe the diversity of ligands for individual GPCRs, we generated a heatmap of EC₅₀ values for the identified peptide-receptor pairs (Figure 2C). The majority of receptors (48 out of 66 GPCRs) were activated by ligands from a single peptide-encoding gene, showing specific interactions. However, we also identified six GPCRs with many ligands, from at least eight peptide-encoding genes (Figure 2C). These promiscuous receptors are encoded by dmsr-1, dmsr-7, frpr-8, and egl-6 genes and are all activated by diverse RFamide (FLP) peptides. Most FLPs also activate multiple GPCRs encoded by different receptor genes (Figure 2C). FLP-1, FLP-8, and FLP-14 are among the most versatile peptides and interact with 8 to 12 GPCRs. By contrast, peptides of the NLP family activate only a few (one to three) receptors (Figure 2C).

We aimed at further validating the promiscuity of peptide GPCRs and asked which second-messenger pathways might be activated in response to different ligands. We first focused on DMSR-7 that shows the highest promiscuity. Previous studies suggested that DMSR-7 has inhibitory effects in vivo.^82,85 To validate its ligand interactions and determine its cellular response, we measured DMSR-7 activation in different heterologous expression systems in the absence of the promiscuous Gα₁₆ subunit. First, we expressed DMSR-7 in Xenopus laevis oocytes and measured responses to multiple FLP ligands, with EC₅₀ values in the nanomolar to micromolar range, by means of two-electrode voltage-clamp (TEVC) recording. DMSR-7 strongly responded to FLP-1, its most potent ligand, causing robust stimulation of K⁺ current through G protein inward rectifying potassium (GIRK) channels (Figures 3A and 3B). Treatment with pertussis toxin (PTX) inhibited the activation of DMSR-7 to the same extent as the known G_i/o-coupled muscarinic M2 receptor (Figures 3A and 3B), suggesting that DMSR-7 is also a G_i/o-coupled GPCR. In the oocyte system, DMSR-7 was also activated by multiple RFamide peptides, while none of them elicited a response in oocytes that lacked the receptor (Figures 3C and 3D). Using a luminescence-based cyclic AMP (cAMP) reporter in HEK cells, we found that DMSR-7 inhibits cAMP signaling in response to 10 different FLP ligands (Figures 3E and 3F). These peptides did not evoke a calcium response in cells expressing DMSR-7 in the absence of Gα₁₆ (Figures 3G and 3H). Our results thus suggest that DMSR-7 responds to diverse RFamide peptides by activating the G_i/o pathway, not the calcium pathway, which is in agreement with functional studies.^82,85 We also investigated the second-messenger responses of two other promiscuous FLP receptors, FRPR-8 and EGL-6 (Figure 2C). Activation of EGL-6 inhibited cAMP signaling in response to all RFamide ligands (Figures S8A, S8B, S8E, and S8G). By contrast, activation of FRPR-8 increased cAMP levels, although these results were less conclusive, because three ligands evoked a significant cAMP response in control cells and three other peptides did not activate the receptor in the absence of Gα₁₆ (Figures S8C–S8G). Together, these results suggest that promiscuous peptide GPCRs stimulate or inhibit cAMP signaling and that the same second-messenger pathway is induced by different FLP ligands. We conclude that promiscuous FLP receptors are activated by multiple diverse RFamide peptides, which may result in similar cellular responses.

Figure 3. The promiscuous receptor DMSR-7 inhibits cAMP signaling in response to diverse RFamide peptides.

(A) Continuous TEVC recordings from untreated (left) and PTX-injected (right) Xenopus laevis oocytes expressing DMSR-7, mGIRK1, and mGIRK2 and treated with 100 nM FLP-1-6 peptide. FLP-1-6 activates DMSR-7, resulting in a robust stimulation of K⁺ current, which is blocked in PTX-injected oocytes.

(B) The percentage GIRK activation for at least seven oocytes expressing DMSR-7 or the G_i/o-coupled muscarinic M2 receptor, mGIRK1, and mGIRK2 with or without PTX treatment. M2 was exposed to acetylcholine and DMSR-7 to FLP-1-6 (100 nM). Boxplots indicate 25th (lower boundary), 50th (central line), and 75th (upper boundary) percentiles. Whiskers show minimum and maximum values.

(C and D) Mean percentage GIRK activation for oocytes injected with DMSR-7, mGIRK1, and mGIRK2 (C), or with mGIRK1/2 alone (D), in response to different peptides at 1 μM(n ≥ 5). NLP-3-1 does not activate DMSR-7 (Table S6) and was included as a negative control.

(E and F) Activation of DMSR-7 by diverse FLP ligands (10 μM) significantly decreases cAMP levels in HEK cells (E). None of the peptides affect cAMP signaling in cells transfected with a control vector (F). Data are shown as relative luminescence after normalization to the ligand-free control (n ≥ 6).

(G and H) Activation of DMSR-7 does not affect calcium levels in CHO cells in the absence of the Gα₁₆ protein. Values are reported as the ratio of the total calcium response in cells transfected with DMSR-7 (G) or empty vector (H), and challenged with peptides (10 μM), BSA (negative control), or ATP (positive control) (n ≥ 6). Error bars indicate SEM. Significance was assessed by Mann-Whitney test (A), Kruskal-Wallis with Dunn’s test (C, D, G, and H), and one-way ANOVA with Dunnett’s test (E and F). ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05; ns, not significant. See also Figure S8.

The RFamide peptide-receptor network is characterized by peptidergic crosstalk

Our results suggested that the RFamide signaling network shows a higher level of ligand-receptor crosstalk in comparison to non-RFamide (NLP) peptide systems. RFamide peptides also interact with a receptor network that is largely distinct from that of other peptides. To further investigate the structure of FLP and NLP ligand-receptor networks, we used a bipartite network representation and generated its projections. We focused our analysis on interactions with sub-micromolar EC₅₀ values (<1 μM), comprising most pairs in our dataset. Bipartite graphs of both networks show different topologies and more crosstalk for ligands and receptors of the RFamide family than for other peptidergic systems (Figure 4A). To investigate the relationships among FLP and NLP peptides, we projected the bipartite networks into simple, monopartite ligand networks in which the nodes (peptides) are connected only if they have at least one common receptor. Nearly all clusters in the NLP network include peptides from a single precursor, which often have similar sequences and activate the same receptor(s) (Figure 4B). The only two exceptions are the C. elegans RPamide (NLP-2, NLP-22, and NLP-23) and PDF-like peptides (PDF-1 and NLP-37). Both are families of homologous peptide genes that likely arose through gene duplication and share significant sequence similarity.^83,86 By contrast, the monopartite projection of the FLP network shows a higher number of edges, consistent with the receptor promiscuity that we observed for RFamide peptides (Figure 4C).

Figure 4. Specific and combinatorial interactions in the peptidergic network.

(A) Bipartite graphs of NLP and FLP peptide-GPCR networks. Nodes represent peptides (purple) and receptors (orange). Edges between two nodes depict peptide-GPCR couples with sub-micromolar EC₅₀ values (EC₅₀ < 1 μM) identified in this resource. Asterisks indicate receptors in both NLP and FLP networks.

(B and C) Monopartite peptide projections of the NLP (B) and FLP (C) bipartite peptide-GPCR networks. Nodes are sized by the total number of connections (degree) and colored by module.

(D) Monopartite receptor projection of the FLP bipartite network. Nodes are sized by degree and colored by module.

(E) Different types of peptide-receptor interaction motifs in the peptidergic network. See also Figures S9–S12.

Many complex networks have a modular structure, whereby they contain subsets of nodes—called modules—that are more densely interconnected with each other than with the rest of the network. We further analyzed the modularity of the FLP ligand network and identified several modules, two of which contained the majority of RFamide peptides (red and dark blue in Figure 4C). Peptides in these two modules interact with one or multiple receptors. However, they all activate the promiscuous receptors DMSR-1 and/or DMSR-7 (Figure 2C). One module comprises all peptides that activate DMSR-7 but not DMSR-1 (dark blue in Figure 4C). The second module groups all ligands of DMSR-1 (red in Figure 4C), of which some—closest to the DMSR-7 module—also activate DMSR-7. The peptide agonists of DMSR-1 and DMSR-7 thus only partially overlap, although many of them also activate other receptors (Table S6). Besides the two DMSR modules, we found eight smaller groups of FLPs that constitute separate modules in the ligand network, such as FLP-2, FLP-3, FLP-15, FLP-20, FLP-21, and FLP-34 peptides (Figure 4C). These include several peptides of the evolutionarily conserved short neuropeptide F (sNPF) and neuropeptide Y/F (NPY/F) families (e.g., FLP-15, FLP-21, and FLP-34),⁸⁷ which arose in a common ancestor of bilaterian animals and, through ligand-receptor co-evolution, may have evolved interactions with different receptor targets than those of other FLPs.^31,88

We found similar modules in a receptor-focused projection of the bipartite FLP network, in which nodes (receptors) are connected when they share at least one peptide ligand. Two modules are centered on DMSR-1 and DMSR-7 (red and dark blue in Figure 4D). Other receptors that share multiple ligands with these promiscuous GPCRs, such as DMSR-5 and DMSR-7, cluster in the same modules, while seven modules are distinct from the DMSR-1 and DMSR-7 clusters (Figure 4D). Consistent with the ligand projection, receptors of the sNPF receptor family (NPR-1, NPR-2, NPR-3, NPR-4, NPR-5, NPR-6, and NPR-10)⁸⁷ cluster in separate modules, as they share none or only a few ligands with other GPCRs. In addition, the FLP-receptor projection revealed a novel module in the FLP signaling network, consisting of seven GPCRs that are uniquely activated by FLP-8 and FLP-14 peptides (light blue in Figure 4D).

Taken together, our results indicate three main types of ligand-receptor interactions in the peptide-GPCR network (Figure 4E): (1) complex “many-to-many” interactions mediated by promiscuous receptors (e.g., DMSR-1 and DMSR-7), which have diverse RFamide peptide ligands that often activate other GPCRs as well; (2) divergent “one-to-many” interactions by RFamide peptides (e.g., FLP-1, FLP-8 and FLP-14) that activate multiple specific receptors; and (3) specific “one-to-one” interactions between peptides and receptors from single-peptide- and GPCR-encoding genes. Most non-RFamide systems belong to this category, whereas the majority of RFamide peptides cross-interact with many GPCRs.

Different interactions encoded within single peptidergic genes

Peptide-encoding genes often code for multiple peptide sequences that may be co-released.^32,89 To probe how this multi-peptide signaling affects the peptide-receptor network, we compared sequences and receptor interactions for peptides derived from the same precursor. Most peptides from a single gene show clear sequence similarity and activate the same receptor(s) (Figures S9–S11). However, we also found several genes, such as flp-1, flp-11, nlp-2, and nlp-23, of which individual peptides differ in their potency to activate specific GPCRs (Figures S10 and S11). For example, FLP-1 peptides with the PNFLRF-NH₂ motif (FLP-1-1 to -7 and FLP-1-10) activate DMSR-5, DMSR-6, and FRPR-7, while these receptors are not activated by the two PNFMRY-NH₂ FLP-1 peptides (FLP-1-8 and FLP-1-9) that interact with NPR-22 (Figure S10; Table S3).

Next, we asked whether ligand interactions differ between GPCR isoforms encoded by the same gene. Our dataset includes isoforms of 11 GPCR genes that differ at N-terminal, C-terminal, or intertransmembrane regions (Figure S12A). Most receptor variants display similar ligand interactions (Figure S12B). However, two N-terminal variants of the NLP-17 receptor NPR-37 showed a 1,000-fold difference in the respective EC₅₀ values for their ligands (Figure S12B). Two variants of the Drosophila myosuppressin receptor ortholog DMSR-1, which differ at their intracellular C terminus, also displayed different interactions (Figures S12A and S12B). FLP-25 peptides potently activate DMSR-1B (DMSR-1-2), with picomolar EC₅₀ values, whereas these peptides activate DMSR-1A (DMSR-1-1) only at high micromolar concentrations (Figure S12B). Because EC₅₀ values can differ depending on the heterologous system,^90–92 we compared the concentration-dependent activation of the two DMSR-1 variants in Xenopus oocytes. In this system, FLP-25 potently activated both DMSR-1 isoforms with similar EC₅₀ values (Figure S12C). EC₅₀ values were also comparable for most other DMSR-1 agonists in our calcium mobilization assay (Figure S12B), suggesting that both isoforms show similar ligand interactions.

Characterization of ancestral bilaterian peptidergic systems in C. elegans

Many peptidergic systems show long-range evolutionary conservation. Comparative genomic analyses of these systems across widely divergent animal phyla revealed at least 31 peptidergic systems that are ancestral to bilaterian animals, such as oxytocin and vasopressin, cholecystokinin, neuropeptide Y (NPY), and gonadotropin-releasing hormone (GnRH) systems.^31,93–95 Several bilaterian systems are also thought to be conserved in C. elegans, but not all predicted peptide-receptor couples have been experimentally demonstrated. To further clarify the conservation of peptidergic systems in C. elegans, we performed a phylogenetic analysis of nematode GPCRs in our screen, identified receptor orthologs for bilaterian peptide systems, and compared their ligands to peptides that activate known representatives of the same receptor family.

To reconstruct phylogenetic relationships, we first used the sequences of all rhodopsin and secretin-type peptide GPCRs from species in eight bilaterian phyla as bait to identify potential homologs in transcriptomic and expressed sequence tag (EST) datasets of 34 phylogenetically dispersed nematode species (Table S8). We then constructed a maximum-likelihood tree, which revealed 31 bilaterian receptor clusters that contain a diversified set of protostome and deuterostome GPCRs (Figures 5A and 5B; Data S1). Representative C. elegans receptors were present in 17 of the 31 ancestral receptor families (Figure 5C; Table S9). These GPCR families were well conserved in nematode species across different clades (Figure 5C).

Figure 5. Peptide-GPCR pairs support conservation of bilaterian peptidergic systems and expansions of peptide-GPCR families in nematodes.

(A) Maximum-likelihood tree of bilaterian rhodopsin peptide GPCRs. Subtrees are numbered according to the receptor families in (C). Subtrees comprising bilaterian families with nematode representatives are indicated in blue; those without nematode representatives are in red. Subtrees containing only protostomian sequences are colored yellow and nematode-specific subtrees are green. Node support values at the root of all delineated subtrees are above 0.95 unless depicted otherwise. Atypical peptide receptors (angiotensin, bradykinin, and chemokine receptors) were used as outgroup (gray).

(B) Maximum-likelihood tree of bilaterian secretin receptors. Color-coding, numbering, and node support values as in (A). Adhesion and cadherin receptors were used as outgroup.

(C) Inferred evolutionary relationships for nematode peptide-receptor systems. Names of peptidergic systems are assigned according to the classification by Mirabeau and Joly³¹ and Elphick et al.⁹⁴ and color-coded as in (A). Gray squares mark the presence of a receptor ortholog. A receptor is considered present in a nematode clade when it is positioned inside a well-supported subtree (branch support value >0.95). The number of C. elegans receptors in each family is depicted in the left column and squares are colored green if at least one receptor is paired with a peptide ligand.

(D) Elevenin-like SNET-1 (LDCRKFSFAPACRGIML; C3–C12) and calcitonin/DH31-like NLP-73 (NRQCLLNAGLSQGCDFSDLLHAQTQARKFMSFAGPamide; C4–C14) peptides activate the elevenin and calcitonin/DH31 receptor orthologs NPR-34 and SEB-2. Calcium responses of CHO cells expressing NPR-34 or SEB-2 are shown relative (%) to the highest value (100% activation) after normalization to the total calcium response. Error bars represent SEM (n ≥ 6). See also Data S1 and Tables S6, S8, and S9.

We identified ligand-receptor pairs for 12 of the 17 bilaterian families of peptide GPCRs that are conserved in C. elegans (Figure 5C; Table S9). In most cases, the C. elegans receptor was activated by a peptide that was a previously recognized ortholog of the peptide ligands of related GPCRs.^31,93,94 Consistent with previous reports, our data support the conservation of vasopressin/oxytocin (NTC-1/NTR-1), tachykinin (NLP-58/TKR-1), GnRH (NLP-47/GNRR-1), cholecystokinin (NLP-12/CKR-1/ CKR-2), neuromedin U (CAPA-1/NMUR-1/NMUR-2), NPY (FLP-34/NPR-11), thyrotropin-releasing hormone (TRH-1/ TRHR-1), luqin (LURY-1/NPR-22), pigment-dispersing factor (PDF-1/PDF-2/PDFR-1), and myoinhibitory peptide (MIP-1/ SPRR-2) systems in C. elegans.^{5,6,41,46,49,50,62,63,74,79,86,96–98} We also identified peptide ligands for four C. elegans GPCRs related to neuropeptide FF/SIFamide (NPR-35), tachykinin (TKR-2), NPY (NPR-12), and myoinhibitory peptide (SPRR-1) receptors (Table S9). Our results provide biochemical evidence for the predicted interactions of these receptors with orthologs of neuropeptide FF/SIFamide (NLP-10), tachykinin (NLP-58), NPY (FLP-33), and myoinhibitory peptides (NLP-42).^31,46,49,99 In addition, we identified a ligand for SEB-3, a GPCR that belongs to the same bilaterian secretin superfamily of receptors as bilaterian CRH/DH44 and calcitonin/DH31 receptors, vertebrate VIP/ PACAP receptors, and invertebrate PDF receptors.^31,78,93,94 SEB-3 is most closely related to PDF receptors (Figure 5B) and orthologous to orphan GPCRs in many other species.^31,94 The identification of a peptide ligand for SEB-3 may help to characterize ligands of other receptors within this family. A functional study of this peptide-receptor pair validated its interaction in vivo and characterized its role in the regulation of locomotion, arousal, and reproduction.⁷⁸

C. elegans receptors in five bilaterian GPCR families remain orphan. For the C. elegans galanin/allatostatin A (NPR-9) and leucokinin (TKR-3) receptor orthologs,³¹ peptides related to allatostatin A (NLP-6) and leucokinin (NLP-43) were included in our screen but did not activate their predicted GPCR. Reasons for this lack of interaction include (1) the functional expression of GPCRs in CHO cells may have failed, or (2) in the case of NPR-9, the predicted receptor cDNA sequence could not be cloned (Data S1). For C. elegans orexin/allatotropin (NPR-14), elevenin (NPR-34), and calcitonin/DH31 (SEB-2) receptor orthologs, the predicted ligands were initially not included in our peptide library. C. elegans homologs of orexin/allatotropin (NLP-59) and elevenin (SNET-1) were identified after constructing the library,^94,100 whereas a calcitonin-like peptide (NLP-73) was included but lacked a disulfide bridge that may be critical for receptor activation (Table S3). We therefore retested these receptors with their predicted ligands and found that SNET-1 and NLP-73 activate NPR-34 and SEB-2, respectively (Figure 5D; Table S5). We did not observe activation of NPR-14 with NLP-59, which may be because this receptor did not express well or is activated by another, unidentified orexin/allatotropin-like peptide. Taken together, our results support the conservation of at least 14 ancestral bilaterian peptide systems in C. elegans (Figure 5C). Based on these findings, we suggest naming these peptide and receptor-encoding genes in accordance with their evolutionary relationship to bilaterian peptidergic systems (Table S9).

Nematode expansions of evolutionarily conserved peptidergic systems

Besides the conservation of bilaterian peptide systems, phylogenetic analysis revealed expansions of peptide-GPCR families in nematodes, including GnRH, somatostatin, sNPF, myosuppressin, and FMRFamide-related receptors (Figure 5A; Table S9; Data S1). We find that several C. elegans receptors of these expanded families are activated by peptides of the corresponding related peptide family, which supports the conservation of these peptidergic systems in nematodes. For example, one nematode-expanded receptor family clustered near the bilaterian family of somatostatin (SST), allatostatin C (AstC), and opioid (Op) receptors (cluster 37; Figure 5A; Data S1). This group of 12 C. elegans GPCRs (Table S9) includes NPR-17, which was previously shown to interact with the opioid-like peptide NLP-24.³⁷ We found two other receptors in this family, NPR-24 and NPR-32, to interact with orthologs of the SST/AstC family, NLP-62 and NLP-64 (Figure S2), supporting the conservation of SST/AstC-like receptors in C. elegans.^31,101 A second expansion occurred for GnRH-related receptors (cluster 38; Figure 5A; Data S1). C. elegans GNRR-1 clusters most closely to GnRH-like receptors in other animals and is specifically activated by NLP-47 (Figure S3; Data S1), which is orthologous to GnRH and the related insect adipokinetic hormone (AKH).⁹⁷ Seven additional GnRH-like receptors (GNRR-2 to GNRR-8) form a paralogous group of which two are activated by RPamide peptides (Figure S3), a nematode family of peptides that shows sequence similarity to GnRH/AKH peptides.⁸³

In addition, three protostomian families of peptide GPCRs diversified in nematodes. The largest expansion occurred in the FMRFamide-like receptor family, related to the Drosophila FMRFamide receptor (FMRFaR), including 37 GPCRs in C. elegans (cluster 33; Figure 5A; Table S9; Data S1). As expected, nearly all GPCRs in this cluster interact with FLP peptides (Table S9). In addition, the protostomian sNPF and myosuppressin receptor families expanded in nematodes (clusters 32 and 34; Figure 5A; Data S1). Protostomian sNPF receptors are orthologous to deuterostomian prolactin-releasing peptide (PrRP) receptors, although protostomian sNPF peptides and deuterostomian PrRP-type peptides share low levels of sequence similarity.^88,93 In C. elegans, the sNPF receptor family includes nine GPCRs, of which eight have sNPF-like peptide ligands with the canonical motif XLRFamide (FLP-15, FLP-18, FLP-21, and FLP-26)⁸⁷ (Table S9; Data S1). The myosuppressin-like receptor family includes 16 C. elegans receptors (Table S9; Data S1), three of which are activated by FLP-1 peptides that show sequence similarity to insect myosuppressins (Figure S6).¹⁰² Interestingly, these FLP-1 receptors—DMSR-5 to– DMSR-7—all cluster together in the phylogenetic tree along with several orphan GPCRs (DMSR-4, DMSR-9, and DMSR-11 to 16), whereas other DMSRs form separate sub-groups and are activated by different, more divergent peptide ligands (Table S9; Data S1). In addition, we found that some RFamide peptides, such as FLP-3 and FLP-14, activate multiple receptors that belong to different expanded peptide-GPCR families (Table S9). Because the FLP system has largely diversified in nematode evolution, it is conceivable that some receptors from phylogenetically distinct clusters evolved the ability to interact with similar peptides, possibly at different evolutionary times. Promiscuous FLP receptors also belong to these expanded GPCR families (Table S9), which may explain ligand promiscuity. Taken together, the biochemical interactions identified for all these receptors provide general insights into the evolutionary history and diversification of peptidergic systems in nematodes.

Discussion

Bioactive peptides and peptide GPCRs constitute the most diverse ligand-receptor signaling network in animals. However, many GPCRs are still orphan receptors, precluding the construction of peptidergic signaling maps at organismal scale. By system-wide screening of peptide-receptor interactions, we have generated a resource of 461 peptide-GPCR couples in C. elegans that, along with the synaptic connectome, single-cell transcriptome data and other resources in this model,^21,55–57 lays a foundation for functional studies of the peptidergic signaling network at system level.

A peptide-GPCR interaction resource for mapping peptidergic circuits and functions in C. elegans

Reverse pharmacology strategies based on Gα₁₆-mediated calcium signaling have been widely used for characterizing GPCR ligands.^6,52,103,104 Using this strategy, we identify 68 peptide GPCRs (including isoforms) that interact with 153 peptides of 60 peptide-encoding genes (31 flp and 29 nlp genes; Table S1). This dataset expands the number of known peptide GPCRs in C. elegans from 29 to 60 receptor genes. Many of the interactions that we uncover have been validated in other in vitro systems and in in vivo studies (Table S1), illustrating that this resource can guide targeted research into peptidergic pathways and functions. Furthermore, we identify additional ligands for peptide GPCRs that were previously deorphanized.

This peptide-GPCR interaction dataset combined with single-cell transcriptome data²¹ provides a basis for mapping the peptide signaling network in C. elegans. Peptide-receptor interactions are one of the most ancient types of signaling pathways that link up cells into complex chemical networks.^1,105 While synaptically wired circuits can be anatomically mapped, the brain’s chemo-connectome constituted by numerous neuromodulators, such as neuropeptides, monoamines, and their receptors, cannot be resolved solely from anatomical expression, as this extensive “wireless” network relies on chemical messengers that often act remotely from their release site.^30,106,107 Knowledge of ligand-receptor interactions is therefore crucial to delineate which nodes or cells in the network can communicate with each other. In a complementary study, we have generated a draft connectome of neuropeptidergic signaling in the C. elegans nervous system by integrating this interaction resource with gene expression and anatomical datasets.⁶⁵ The development of novel tools to delineate peptidergic circuits, such as genetically encoded sensors for peptide-GPCR signaling, provides a promising avenue to further understand the functional organization of this broad signaling network and to validate peptide-receptor interactions in vivo.^108–110 In addition, the ligands identified in this resource allow further investigation of the cellular responses of peptidergic signaling by studying the second-messenger pathways and G protein coupling of peptide GPCRs.

System-wide screening of peptide-GPCR pairs reveals specific and combinatorial signaling motifs in the peptide network

Our dataset reveals three main types of peptide-GPCR interactions that show different levels of complexity: specific, one-to-many, and many-to-many interactions. Many peptide GPCRs interact with ligands derived from only a single peptide-encoding gene, generating a large diversity of specific ligand-receptor pairs for information coding. By contrast, we also identify peptides that interact with a broad range of specific GPCRs as well as promiscuous peptide receptors. This suggests that the peptidergic system forms a complex interaction network in which information is coded not only through a binary mechanism of specific ligand-receptor couples but also through complex combinatorial peptide-receptor signaling.

We find that peptidergic crosstalk is a feature characteristic of the C. elegans FLP-receptor network, as all promiscuous receptors interact with diverse FLP peptides, typified by a carboxyterminal RFamide sequence motif. A few FLPs (e.g., FLP-1, FLP-14) activate multiple closely related as well as phylogenetically more distant receptors. Since the number of RFamide peptide and receptor genes has diversified extensively in nematode evolution, the complex cross-interactions in the RFamide network may be explained by the large expansion of the FLP system in C. elegans.^31,94,111 However, ligand and receptor promiscuity has been demonstrated for peptidergic systems that did not undergo large expansions in other species as well.^29,54,112 For example, human neuropeptide FF receptors are activated by all five groups of mammalian RFamide peptides.^113–116 Peptides that activate multiple unrelated GPCRs have also been described in other animals.^117–120 Receptor promiscuity and peptidergic crosstalk thus appear to be general features of the peptidergic system, especially of RFamide peptides and their receptors.

Our results indicate that promiscuous peptide GPCRs (DMSR-7, FRPR-8, and EGL-6) can stimulate or inhibit cAMP signaling in response to diverse ligands. This finding is in agreement with functional genetic studies showing that DMSR-7 and EGL-6 function as inhibitory receptors in vivo.^82,85,121 The promiscuous receptor DMSR-1 also inhibits foraging and sleep-like behavior by interacting with multiple RFamide peptides. Stress-induced sleep is regulated by FLP-13/DMSR-1 signaling,³⁹ whereas FLP-5 peptides—which activate the same receptor with a higher EC₅₀ value than FLP-13 (Table S6)—regu-late temperature- and food-dependent behavioral states through DMSR-1.⁸² FLP-6 and its receptor EGL-6 were also shown to regulate these behavioral states, while activation of the same EGL-6 receptor by other FLP peptides (FLP-10 and FLP-17) inhibits egg-laying behavior.^82,121 These findings suggest that promiscuous peptide GPCRs can be involved in the regulation of different behaviors in vivo by interacting with different ligands. RFamide peptides that activate many specific GPCRs, such as FLP-1 and several of its receptors, are also involved in multiple biological processes.^{45,85,122,123} In nematodes, these are among the most highly conserved RFamide peptide genes (e.g., flp-1 and flp-14),¹¹¹ which during evolution may have acquired multiple functions mediated by different receptors. Although more functional studies of peptide-GPCR couples are needed, distinct types of interactions in the peptidergic network may serve specialized functions. We speculate that peptide-GPCR couples with complex or promiscuous interactions may be important for orchestrating global states and cellular activity by integrating or broadcasting peptidergic signals. Unique peptide-receptor couples, on the other hand, may have important roles in mediating specific local or long-range signaling.

Finally, multiple bioactive peptides are often encoded within the same peptide precursor. These peptides can activate different receptors, as has been described for the mammalian pro-opiomelanocortin (POMC) precursor.¹²⁴ In C. elegans, we find that different peptides from a single peptide-encoding gene often interact with the same receptor(s). As peptidergic communication is limited by diffusion, encoding multiple ligands of a single receptor may amplify signaling by increasing the concentration of secreted peptide ligands.¹⁰⁵ However, we also identify peptide genes encoding ligands for different receptors, which may diversify signaling.

Evolution of peptidergic signaling systems

Because biologically active peptides typically have short sequences and co-evolve with their receptors, identifying orthology between peptidergic systems from distant animals remains a challenge. Annotations of the first invertebrate genome sequences revealed a large number of GPCRs that resembled vertebrate peptide receptors but relatively few vertebrate-like peptides.^125–127 Initially, this led to the assumption that vertebrate peptidergic systems are not well conserved in C. elegans and other invertebrates. The increasing number of available genome sequences, however, has enhanced our knowledge on the long-range evolution of peptidergic systems,^{1,5,31,88,93,94} revealing deep conservation of these systems across bilaterian animals, which in some cases has been confirmed by biochemical interactions.^31,52,88 Our dataset provides further evidence for orthologs of at least 14 bilaterian peptidergic systems in C. elegans (Table S9). In addition, we clarify the diversification of specific peptide-receptor families in nematodes. Indeed, many peptide receptors that we characterized belong to expanded GPCR families, for which specific ligands could not have been predicted based on available data, as the corresponding peptide families (e.g., RFamide, sNPF, and somatostatin-like peptides) also diversified in nematodes.^31,101

Our phylogenetic analysis suggests that additional peptide ligands remain to be discovered in C. elegans. For example, we identified a large family of orphan nematode GPCRs that may be activated by potentially unknown peptides (cluster 39, Table S9). Likewise, we discovered two orphan receptor families related to the Drosophila trissin and CNMamide receptors (clusters 35 and 36, Table S9) but orthologous peptides have not been identified in nematodes so far.^128,129 Peptidomics and comparative genomic studies also identified novel peptides in C. elegans.^59,130 Peptide discovery by these and other approaches, such as machine learning methods,^29,131 will facilitate future deorphanizations.

In addition, this resource can facilitate predictions of peptide-GPCR couples in other species, such as parasitic nematodes. Conservation of peptide-receptor interactions has been shown between C. elegans and nematode parasites,^132,133 and peptidergic signaling regulates key aspects of nematode biology, such as reproduction, development, and locomotion.^{36,130,134,135} The discovery of peptide-GPCR pairs in parasitic species may thus provide promising leads for the development of novel anthelminthics. Comparisons of peptide-GPCR pairs between nematode species may also further clarify the diversification and ligand promiscuity of GPCRs.

Limitations of the study

Although our system-wide screen more than doubles the number of characterized peptide GPCRs in C. elegans, this peptide-receptor interaction map is incomplete for several reasons. First, we did not identify ligands for all known and predicted peptide receptors, likely because not all GPCRs are functionally expressed in the heterologous system that we used. There are several possible explanations for this lack of functionality, including improper folding or trafficking, insufficient membrane expression, or the requirement of specific co-factors, such as receptor-activity modifying proteins (RAMPs).^136,137 We tested untagged receptors, as tagging might interfere with ligand-GPCR signaling. However, GPCR tagging strategies could be used in future studies to verify membrane expression in heterologous cells. Alternatively, the remaining orphan receptors may be activated by other ligands that were not included in our library or remain to be identified. Likewise, some peptides may interact with other GPCRs than those in our screen or with other types of receptors, such as ion channels.¹²⁰ Second, the promiscuous Gα₁₆ subunit couples many but not all GPCRs to the PLCβ pathway and is less effective in linking G_i-coupled receptors to the release of intracellular calcium than, for example, chimeric Gα proteins.^76,138 Co-expressed Ga proteins may also influence the EC₅₀ values of ligand-GPCR interactions,¹³⁸ which potentially differ from those in vivo. Finally, GPCRs can show constitutive, ligand-independent activity or signal via G protein-independent pathways,^136,139 which are not captured by our readout. Different heterologous systems and screening strategies could be combined in future studies, for example, using chimeric G proteins and other readouts of key signaling events in the receptor activation cascade,^{46,104,138,140,141} to further complete the peptide-receptor network.

Star⋆Methods

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
E. coli: Strain OP50	Caenorhabditis Genetics Center	WB OP50; RRID:WB-STRAIN:OP50
Chemicals, peptides, and recombinant proteins
See Table S3 for a list of peptides	GL Biochem Ltd. and ThermoFisher Scientific	N/A
1% Penicilin/Streptomycin Mixture	Gibco	Cat#15140-122
Zeocin selection reagent	Invitrogen	Cat#R25001
Puromycin	Sigma-Aldrich	Cat#P8833
Fetal Bovine Serum	Sigma-Aldrich	Cat#F7524
Lipofectamine LTX, Plus Reagent	Invitrogen	Cat#15338–500, Cat#11514015
Coelenterazin H	Invitrogen	Cat#C6780
Critical commercial assays
MycoAlert Mycoplasma Detection Kit	Lonza	Cat#LT07-418
Steady Plus Reporter Gene Assay System	PerkinElmer	Cat#6066751
T3 mMessage mMachine transcription kit	Thermo Fisher Scientific	Cat#AM1348
GeneJET RNA purification kit	Thermo Fisher Scientific	Cat#K0731
Experimental models: Cell lines
CHO/mtAEQ/Gα16	PerkinElmer	ES-000-A24
CHO/mtAEQ	PerkinElmer	ES-000-A12
HEK 293T	ATCC	Cat#CRL-11268
Experimental models: Organisms/strains
C. elegans: N2 Bristol wild type	Caenorhabditis Genetics Center	N2 Bristol
Oligonucleotides
See Table S2 for a list of primers	Integrated DNA Technologies	N/A
Software and algorithms
Prism	Graphpad	https://www.graphpad.com/scientific-software/prism/
ScreenWorks System control software	Molecular Devices	FLIPR Tetra control program
MicroBeta control software	PerkinElmer	Microbeta control program
RoboCyte2 control software	Multichannel systems	Roboocyte2 control program
RoboCyte2+ analysis software	Multichannel systems	Roboocyte2+ analysis program
Other
Xenopus laevis oocytes	EcoCyte Bioscience	https://ecocyte-us.com/products/xenopus-oocyte-delivery-service/

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Isabel Beets (isabel.beets@kuleuven.be).

Materials availability

The HEK 293T and CHO-K1 cell lines stably expressing mitochondrial-targeted apo-aequorin are under MTA and cannot be freely distributed.

Experimental Model and Subject Details

C. elegans maintenance

The wild-type C. elegans (N2 Bristol) strain was obtained from the Caenorhabditis Genetics Center (University of Minnesota). Worms were cultivated at 20°C on nematode growth medium (NGM) plates seeded with E. coli OP50 bacteria and used for mRNA extraction and subsequent cDNA synthesis. cDNA prepared from mixed-stage culture plates was used as template for PCR of peptide GPCR coding sequences.

Microbe strains

The Escherichia coli OP50 strain was used as a food source for C. elegans.

Cell lines

CHO-K1 cells stably expressing mitochondrial-targeted apo-aequorin and a promiscuous human Gα₁₆ protein were used for reverse pharmacology screening (CHO/mtAEQ/Gα₁₆, ES-000-A24, PerkinElmer). Calcium measurements in the absence of Gα₁₆ were done in CHO cells that stably express only apo-aequorin (CHO/mtAEQ, ES-000-A12, PerkinElmer). Cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM)/Nutrient Mixture F-12 Ham (Sigma-Aldrich), supplemented with 10% fetal bovine serum (FBS, Sigma-Aldrich) and 1% penicillin/streptomycin mixture (10.000 units/ml penicillin and 10 mg/mL streptomycin, Gibco). Growth medium was supplemented with 250 μg/mL zeocin or 5 μg/mL puromycin, which serves as a selection reagent for CHO cells with or without the Gα₁₆ subunit, respectively. For cAMP-based receptor activation assays, GPCRs were transiently expressed in HEK 293T cells cultured in DMEM/high glucose supplemented with 10% FBS and 1% penicillin/streptomycin. All cells were grown in stable conditions of 37°C, 5% CO₂ and high relative humidity, unless otherwise indicated. Mycoplasma tests were performed using the MycoAlert Mycoplasma Detection Kit (Lonza) to verify that cells were free of mycoplasma contamination.

Xenopus laevis oocytes

Defolliculated Xenopus Laevis oocytes were sourced from EcoCyte Bioscience and maintained in ND96 solution (in mM: 96 NaCl, 1 MgCl2, 5 HEPES, 1.8 CaCl2, 2 KCl) at 16°C until RNA injection.

Method Details

Cloning of peptide GPCR candidates

A list of peptide GPCR candidates (Table S2) was compiled from in silico searches for putative peptide GPCR genes in the C. elegans genome.^36,61,66,67 Oligonucleotide primers for amplifying receptor cDNAs (Table S2) were designed based on gene models in Wormbase (http://wormbase.org, version WS240). Only receptors with a seven transmembrane topology, as predicted from the translated cDNA sequence (http://www.cbs.dtu.dk/services/TMHMM), were considered as peptide GPCR candidates. Forward primers included a ‘CACC’ sequence at the 5′ end that introduces a partial Kozak sequence for increased translation efficiency in mammalian cells. Receptor cDNAs were amplified by PCR with Q5 High-Fidelity DNA Polymerase (New England Biolabs) from cDNA of mixed-stage populations of wild-type C. elegans, and were cloned in a pcDNA3.1 vector (ThermoFisher Scientific). Plasmids for C16D6.2, T14B1.2,⁴² F59D12.2, and C54A12.2⁴⁰ were kind gifts from Dr. Derek Sieburth (University of Southern California, Los Angeles, USA), Dr. Rachel McMullan (The Open University, Milton Keynes, UK), and Dr. David Raizen (University of Pennsylvania, Philadelphia, USA). All plasmids were verified by sequencing of the full receptor cDNA and purified with an EndoFree Plasmid Maxi Kit (Qiagen). cDNA sequences of all GPCR variants cloned in pcDNA3.1 are summarized in Data S1.

Peptide library composition and synthesis

A library of 344 synthetic peptides was generated for reverse pharmacology screening of C. elegans peptide GPCRs. For this, we compiled a list of known C. elegans peptide precursors of the FLP and NLP families by assembling peptide sequences from pepti-domics studies and in silico searches in the C. elegans genome.^{5,6,31,60,86,96,97,99,102,127,135,142–155} We assembled 94 precursors and searched each of them for peptide sequences identified by peptidomics or flanked by mono- or dibasic cleavage sites for proprotein convertases (RK, RR, KR, KK or [RK]-X_2/4/6/8-[RK]).⁹⁹ This search yielded 344 mature peptide sequences. Replicates of the identified peptides were synthesized as crude peptides (40%–70% purity) by GL Biochem Ltd. and ThermoFisher Scientific, and were chemically modified to match post-translational modifications (e.g., disulfide bridge formation, C-terminal amidation or N-terminal pyro-glutamation) that are commonly found in bioactive peptides.⁵⁹ Peptides carrying a disulfide bridge were synthesized at >95% purity. Peptide stocks were reconstituted in 80% acetonitrile and vortexed. If needed, peptide solutions were sonicated to enhance solubilization, and hydrophobic peptides were first dissolved in small amounts of solvent and then diluted using water. Sequences of all synthetic peptides in the library are listed in Table S3.

Transfection of CHO and HEK cells

GPCRs were heterologously expressed in CHO or HEK cells by transient transfection. CHO cells, used in aequorin-based GPCR activation assays, were transfected with GPCR/pcDNA3.1 plasmids when cells reached 60–70% confluency, using Lipofectamine LTX with Plus Reagent (Invitrogen) in serum-free medium. Shifting the cultivation temperature from 37°C to 28°C, at 24h after transfection, was shown to increase the response of C. elegans peptide GPCRs in CHO cells.^92,156,157 Transfected cells were therefore supplemented with complete growth medium the day after transfection and transferred to an incubator at 28°C 16 h prior to cell harvesting for the GPCR activation assay.

HEK 293T cells, used for cAMP-based GPCR activation assays, were co-transfected with the cAMP indicator CRE(6x)-luciferase and GPCR/pcDNA3.1 plasmids in a 1:1 ratio when cells reached 60–70% confluency. Transfection medium contained Opti-MEM (Gibco), CRE(6x)-luciferase and receptor plasmid DNA, Plus Reagent (Invitrogen), and Lipofectamine LTX (Invitrogen). One day post-transfection, fresh culture medium was added and cells were kept at 37°C. Two days post-transfection, the cAMP reporter assay was performed.

Aequorin-based GPCR activation assay

We used an aequorin-based calcium mobilization assay to identify and validate peptide-GPCR couples in the synthetic C. elegans peptide library (in CHO/mtAEQ/Gα₁₆ cells). Two days after transfection, CHO cells were detached and collected in phosphate buffered saline (PBS) containing 0.2% EDTA. Viable cells were quantified using a Biorad TC20 Cell counter. Cells were pelleted by 5 min centrifugation at 800 rpm at room temperature and resuspended in DMEM/BSA (DMEM/F-12 without phenol red, with L-glutamine and 10 mM HEPES, 0.1% bovine serum albumin; Gibco, Thermo Fisher Scientific) at a concentration of 5 3 10⁶ cells/ml. Coelenterazine H (Invitrogen) was added at a final concentration of 2.5 μM, and cells were gently shaken at room temperature for 4 h in the dark, allowing the aequorin holoenzyme to be reconstituted. After a 10-fold dilution in DMEM/BSA, cells were incubated for 30 min before starting the assay.

Library screening was performed on an FLIPR Tetra High-Throughput Cellular Screening System with ScreenWorks software (Molecular Devices). Using this system, cells were transferred to 96-well plates (25,000 cells/well) containing 50 μL of synthetic peptide, dissolved in DMEM/BSA, and calcium responses were simultaneously monitored for 36 s. Peptides were individually tested in black 96-well plates with clear bottom (Greiner Bio One) at a final concentration of 10 μM. Wells containing 50 μL DMEM/BSA were used as a negative control. ATP (1.5 μM in DMEM/BSA), which activates an endogenous receptor in CHO cells, was used asa positive control. After measuring peptide-evoked responses, 50 μL of Triton X-100 (0.2% in DMEM/BSA) was added to lyse the cells and obtain a measure of the maximum calcium response.

Selection of putative peptide hits

Peptide-GPCR interactions were initially tested at high peptide concentrations (10 μM) in luminescence-based GPCR activation screens with a synthetic peptide library. From this data, we selected putative receptor ligands by manually inspecting relative light unit (RLU) plots of calcium responses and by ranking peptides based on standardized Z scores. We integrated peptide-evoked calcium responses over time and normalized the signals per assay plate and subsequently per receptor to compute an activation value for each peptide-GPCR pair. Specifically, we performed Z score normalization on the integrated RLU signal using the mean and variance of the on-plate negative controls to obtain an activation value for each peptide-GPCR pair. Next, we standardized the Z scores for each GPCR by dividing by its maximal activation Z score, and log2-transformed this fraction to rank individual peptide-GPCR couples. Based on the scores obtained for known peptide-GPCR interactions, we considered all peptide-GPCR couples with a strong calcium response over background (Z score >20), reaching at least 40% of that receptor’s maximum activation (log2-standardized Z > -1/2), as potential hits in the peptide library. When the top hit (log2-standardized Z =0) could not be validated in dose-response tests, Z scores for that receptor were restandardized to the next top hit in the library screen and dose-response assays were performed for all peptides matching the selection criteria above. The NLP-2-1 synthetic peptide was excluded from the standardization procedure (log2-standardized Z was set to NA), as it produced a signal (Z score ≥5) in 136 out of 176 assays and was therefore considered to be a likely false positive.

When concentration-response measurements validated a peptide-receptor pair, all peptides derived from the same peptide precursor were systematically tested in concentration-response assays. An exception was made for GPCRs with peptide hits originating from more than 10 different FLP precursor proteins (DMSR-1, DMSR-7, and FRPR-8). For these promiscuous receptors, at least one FLP peptide from each FLP precursor was tested for concentration-dependent receptor activation.

Concentration-response measurements

All putative peptide hits identified in library screens were further tested in concentration-response assays (in CHO/mtAEQ/Gα₁₆ cells). Crude peptides were first purified by reversed phase HPLC on a Symmetry-C8 Prep Column (7.83300 mm HPLC cartridge with pore size of 100 Å and particle size of 7 μM), typically yielding peptide stocks of >95% purity. Peptide masses of purified peptides were verified by MALDI-TOF/TOF mass spectrometry on a rapifleX MALDI Tissuetyper (Bruker Daltonic) and peptide concentrations were measured using the bicinchoninic acid (BCA) protein assay.

Purified peptides were tested in concentration series using the aequorin-based GPCR activation assay on a MicroBeta2 LumiJET luminometer with MicroBeta control software (PerkinElmer). Positive and negative controls were similar to those used in library screens. Calcium responses were measured for 30 s after adding cells to the compound plate. After the addition of Triton X-100 (0.2% in DMEM/BSA), a maximum response per well was measured for 30s, which was used for normalization of peptide-evoked calcium responses.

Concentration-response measurements were performed in triplicate on at least two independent days. For each peptide concentration and replicate, calcium responses were first normalized to the maximum calcium response (sum of peptide- and Triton-evoked response) and subtracted with the negative control (BSA) value. Next, a relative calcium response (%) compared to the maximum peptide-evoked response (100% activation) was calculated. Concentration-response data were fitted in function of log[peptide] using GraphPad Prism 5 or 7. EC₅₀ values were calculated from concentration-response curves by fitting a 3- or 4-parameter concentration-response curve.

Calcium-based assay in the absence of Gα₁₆

Calcium signaling in the absence of Gα₁₆ was measured in CHO/mtAEQ cells using the aequorin-based GPCR activation assay. Peptides were dissolved in DMEM/BSA and 50 μL of the peptide solution was added to the wells of a white flat bottom 96-well plate. Wells containing DMEM/BSA were used as a negative control, while wells containing 1.5 μM ATP were used as a positive control. Incubated cells were added to the wells at a density of 25,000 cells/well and luminescence was monitored for 30 s on a MicroBeta2 LumiJET luminometer (PerkinElmer). After 30 s, 0.2% Triton X-100 dissolved in DMEM/BSA was added to lyse the cells and light emission was recorded for another 30 s. Light emission from each well was calculated relative to the total calcium response (ligand + Triton X-100). Assays were done on at least two independent days in triplicate.

cAMP-based GPCR activation assay

On the day of the assay, cells were detached, counted, pelleted, and resuspended in 200 μM 3-isobutyl-1-methylxanthine (IBMX) medium. Peptides were also dissolved in IBMX medium and 50 μL of the peptide solution was added to the wells of a white flat bottom 96-well plate. Wells containing 50 μL IBMX medium, without peptide, were used as negative control. Each well was supplemented with 50 μL of the cell suspension (50,000 cells/well) and the plate was incubated at 37°C for 3.5 h. Cells were then loaded with 100 μL SteadyLitePlus substrate (PerkinElmer) and incubated on a shaking plate for 15 min under dark conditions at RT. Finally, luminescence was measured twice for 5s (at 0s and 5s) per well at 469 nm on a Mithras LB940 luminometer (Berthold Technologies). Measurements were performed in triplicate on at least two independent days.

Two-electrode voltage clamp (TEVC) recording

Plasmids for expressing C. elegans GPCRs in Xenopus laevis oocytes were generated using the KSM backbone. cDNA sequences of dmsr-1a (DMSR-1-1), dmsr-1b (DMSR-1-2), and dmsr-7 (DMSR-7-1) were inserted into the KSM vector, which contains Xenopus β-globin UTR regions and a T3 promoter, using HiFi assembly. Mouse G protein inward rectifying potassium channels, mGIRK1 (NM_001355118.1) and mGIRK2 (NM_001025584.2), and the mouse muscarinic GPCR M2 (NM_001411688.1) were prepared from mouse cDNA as previously described.¹⁵⁸

Prior to RNA injection, KSM plasmids were linearised using NotI and 5′-capped cRNA was synthesized in vitro using the T3 mMessage mMachine transcription kit (Thermo Fischer Scientific). RNA was then purified using the GeneJET RNA purification kit (Thermo Fischer Scientific). The defolliculated Xenopus oocytes were placed individually into V-bottom 96 well plates and injected with RNA using the Roboinject system (Multi Channel Systems GmbH). For each oocyte a total of 10 ng RNA was injected for the GPCR and 7.5 ng RNA for mGIRK1 and mGIRK2 each. All constructs were injected in one mix and the expression of the GIRKs was used to determine if the injection had been successful. After injection, the oocytes were incubated for 2 days in ND96 at 16°C until recording began.

Two-electrode voltage-clamp recordings were performed using the Robocyte2 recording system (Multi Channel Systems). Recording electrodes were purchased from Multichannel systems and typically had a resistance of 0.7–2 MΩ. The pipettes were filled with 1.5 M KCl and 1 M acetic acid. Oocytes were clamped at -80 mV and continuous recordings were done at 500 Hz. All recordings were collected using the Robocyte2 control software and the data was extracted using the Robocyte2+ analysis software.

DMSR-7 promiscuity analysis was performed by first perfusing the oocytes with ND96 to estimate resting potential of the oocytes. The high K⁺ solution (96 mM KaCl, 1 mM MgCl2, 5 mM HEPES, 1.8 mM CaCl2, 2 mM NaCl) was then applied for 40s to open the GIRK channels and get their baseline current. Peptide solutions were prepared in the High K⁺ solution and then subsequently applied for 20s. The ratio between the baseline High K⁺ current and the Peptide/High K⁺ current was used to determine the ‘% GIRK activation’. The oocytes were then perfused with ND96 again to bring the oocyte back to resting potential. Each application of the peptide was performed on an individual oocyte to ensure desensitisation of the GPCR did not affect the results.

Concentration-response analysis of DMSR-1 isoforms was performed in a similar manner with an initial baseline High K⁺ application for 40s, and then increasing concentrations of the peptide in High K⁺ was applied stepwise for 20s each, without a wash period in between. The protocol ended with a 30s ND96 washout to bring the oocyte back to resting potential. Data was normalised by calculating the I/Imax for each oocyte. Normalized mean, SD and n numbers were imported into GraphPad where data was plotted and EC₅₀ values were calculated by fitting to the Hill equation using three parameter slopes to obtain the highest degree of fit.

To determine the G protein coupling of DMSR-7, on the recording day half the oocytes were injected with 50 pg pertussis toxin (PTX) and incubated at room temperature for 6 h. The TEVC recordings were performed and analyzed as described above for the promiscuity experiments using FLP-1-6. M2 is a known G_i/o-coupled GPCR and was used as a control.

Bipartite network analysis

Bipartite graphs were constructed for all FLP and NLP peptide-GPCR couples with sub-micromolar EC₅₀ values (<1 μM) identified in this resource (Table S6). Promiscuous receptors were assumed to interact with all peptides from the same peptide precursor unless our results indicate otherwise. Monopartite network projections were generated and analyzed using custom scripts in MATLAB; scripts are available upon request, MATLAB version 9.10.0.1602886 (R2021a). Modules were defined using a popular consensus clustering approach,¹⁵⁹ with the Louvain community detection algorithm as implemented in the Brain Connectivity Toolbox.^160,161 The Fruchterman-Reingold force-directed layout algorithm was used to layout graphs in 2D space for visualisation.¹⁶²

Peptide sequence alignments

Pairwise peptide sequence alignment scores in Figure S9 were generated using the stringDist method from the R/Bioconducor package Biostrings. The BLOSUM50 substitution matrix was used with gap opening and gap extension penalties set to 13 and 2, respectively.

Phylogenetic analysis

We first drew a list of 150 C elegans sequences predicted to be peptide GPCRs (Table S8B). We then added to this list the predicted peptide GPCRs from non-nematode bilaterian genomes of Homo sapiens (Vertebrata), Takifugu rubripes (Vertebrata), Branchiostoma floridae (Cephalochordata), Saccoglossus kowalevskii (Hemichordata), Strongylocentrotus purpuratus (Echinodermata), Capitella teleta (Annelida), Lottia gigantea (Mollusca), Dapnia pulex (Arthropoda) and Drosophila melanogaster (Arthropoda) and those from the nematode species Strongyloides ratti, Trichinella spiralis, Brugia malayi, Pristionchus pacificus, Loa loa and Onchocercus volvulus downloaded from the Ensembl metazoan database, using the strategy described in Mirabeau and Joly, 2013.³¹ We complemented these sequences with the predicted peptide GPCRs from the complete set of nematode ESTs available from NCBI (https://www.ncbi.nlm.nih.gov/nucest). All nematode species used in the construction of the phylogenetic tree are listed in Table S8A.

We retrieved 150 secretin-like and 1963 rhodopsin-like GPCRs for this analysis (Tables S8C and S8D) and annotated each bilaterian sequence that was included in Mirabeau and Joly, 2013 for its membership in a conserved bilaterian group (for example, the Capitella sequence jgi|Capca1|219484|estExt_fgenesh1_pg.C_20057 was annotated as jgi|Capca1|219484|estExt_fgenesh1_ pg.C_20057_Calc/DH31). We then constructed two separate phylogenetic trees with the SATé -II maximum-likelihood algorithm,^163,164 with default settings (Aligner:MAFFT,¹⁶⁵ Merger:MUSCLE,¹⁶⁶ Tree Estimator:FASTTREE,¹⁶⁴ substitution model:GTR+G20, Maximum size subproblem:50%, Decomposition:Centroid, Stop after Last Improvement). Branch support values were generated using the implementation of the Shimodaira-Hasegawa test of FASTTREE, displayed as “label” in supplemental tree data.

Quantification and Statistical Analysis

Tests used to determine statistical significance were one-way ANOVA with Dunnett’s multiple comparisons test (cAMP-based reporter assays), the Kruskal-Wallis test with Dunn’s test for multiple comparisons (calcium-based reporter and oocyte assays), the Pearson’s Chi-squared test (GPCR activation probability), two-way ANOVA with Tukey post-hoc test (empty vector experiments), and the Mann-Whitney test (PTX-assays in oocytes). Statistical information for each experiment can be found in the corresponding figure legend.

Highlights.

• System-wide reverse pharmacology deorphanizes 68 C. elegans peptide GPCRs

• Discovery of 461 peptide-GPCR pairs and additional ligands for characterized GPCRs

• Peptide ligands show specific and complex combinatorial receptor interactions

• Peptide-GPCR pairs support long-range conservation and expansion of peptide systems

In brief.

Beets et al. present a system-wide resource of 461 peptide-GPCR couples in C. elegans, revealing binary and complex combinatorial signaling motifs in the peptide-receptor network, which provide insight into the evolution of peptidergic systems and lay a foundation for studying the functional organization of the peptide signaling network at system level.

Data and code availability

• All data generated in this study is reported in this paper and will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.