The putative GPCR-encoding transcripts of eleven decapod species (including crab, shrimp, crayfish and lobster from both freshwater and seawater) were compiled into a single catalogue using the contemporary methods described below. Previously published GPCRomes (i.e., all proteins identified as putative GPCRs of an organism) that have been incorporated into this study include: Procambarus clarkii (21), Sagmariasus verreauxi (10), Nephrops norvegicus (22), Carcinus maenas (23), Gecarcinus lateralis (24) and Penaeus monodon (25). Publicly available transcriptomic reads accessed from NCBI Transcriptome Sequence Assembly (TSA) database include: Homarus americanus, Cherax quadricarinatus and Eriochier sinensis. Novel GPCRomes were generated for Panulirus ornatus, C. quadricarinatus and E. sinensis and are presented as part of this study. A summary of the species, tissues, assembly technology, and origin of transcriptomic reads is summarized in Table 1 .

The sampling of tissue, RNA extraction, RNA quality assurance and RNA read sequencing for P. ornatus have been outlined previously (26–28). The resulting read sequences were subjected to the de novo assembly algorithm of CLC Genomics Workbench v9.5 (https://www.qiagenbioinformatics.com/) and cDNA was translated into the longest open reading frames (ORFs) using CLC. All ORFs were then screened against the Pfam database for predicted GPCR structural domains using the CLC platform with the Pfam module (v29). Search results displaying maximum E-values of 10^-3 were extracted and iteratively used as a reference list. Additionally, a search was conducted on UniprotKB database using the terms “family: G protein coupled receptor 1 family and taxonomy: Arthropoda [6656]” for Rhodopsin receptors, and “family: G protein coupled receptor 2 family and taxonomy: Arthropoda [6656]” for Secretin receptors. Rhodopsin and Secretin reference BLAST databases were generated from these lists and a local BLASTp search was conducted to discriminate the class A (Rhodopsin-like) and class B (Secretin-like) decapod GPCRs. The N-terminus and C-terminus of all predicted GPCR-encoding transcripts were trimmed to also yield lists of 7TM domains (including ECLs and ICLs).

Predicted class A (Rhodopsin-like) decapod GPCR-encoding transcripts were then subjected to a bioinformatic workflow which was designed to assist in the classification and functional annotation of GPCRs by reducing the relevant search spaces. Key steps in the workflow included clustering analysis (incorporating deorphanized and predicted orthologues), and the determination of the topological GPCR domains and conserved motifs (that is DRY, CWxP, NPxxY). The relationships between the sequences of the function-related DRY motif locus and the phylogeny of the 7TM domains was then analyzed for predicted receptors of the GnRH-like superfamily of neuropeptides. The congruence and compartmentalization of species, conserved motifs and phylogenetic clades was then qualified collectively, allowing the annotation of receptors into clades.

Firstly, all available class A decapod GPCR ORFs were downloaded from the Interpro database using the search terms ‘Rhodopsin-like’ and ‘7TM’. A combined list of class A decapod GPCRs was compiled that incorporated the transcripts from Interpro, our eleven decapod species and selectively ‘seeded’ decapod GnRH-like receptor transcripts that have previously been functionally or bioinformatically annotated. Clustering analysis was then performed on the combined list using the CLANS2 algorithm (29) with a P-value of 1e-50 for 10,000 rounds in three dimensions. The resulting clusters were then collapsed to two dimensions for visualization. In brief, CLANS2 performs all-against-all BLAST searches in the three-dimensional space where clustering is performed using attractive forces proportional to the negative logarithm of the BLAST P-values, and a uniform repulsive force.

A list of the available functionally annotated receptors of the decapod superfamily of GnRH-like neuropeptides (that is, GnRH, corazonin, RPCH, ACP, CCAP and vasopressin) was compiled as seeds. The seeded transcripts included the functionally annotated CCAPR of Scylla paramamosain (Sp_CCAPR) (30), and corazonin receptor (Cm_CrzR: accession number MF974386) and RPCH receptor (Cm_RPCHR: accession number MF974387) of Cm (31), as well as the bioinformatically annotated vasopressin receptors (VRs) of Portunus pelagicus (Pp_VR: accession number MZ147830.1) and of Litopenaeus vannamei (Lv_VR: accession number ROT66533.1), and the ACP receptor of Macrobrachium rosenbergii (Mr_ACPR) (32). Transcripts that clustered with these seeds were extracted and considered putative GnRHR-like transcripts.

The candidate GnRHR-like transcripts were then subjected to the online machine-learning DeepTMHMM algorithm in the prediction of GPCR topological domains (including N-terminus, C-terminus, ECLs, ICLs, and 7TM) (33). Identification of these domains enabled the subsequent prediction of the highly conserved DRY, CWxP, and NPxxY motifs. The aa sequences of both the 7TM (including ECLs and ICLs) and the transcript ORFs of the GnRHR-like receptors were aligned using the MUSCLE algorithm (34) implemented in MEGA-X (35). Then neighbor joining phylogenetic trees were constructed (bootstrap values of 1,000) using Mega-X. The resulting phylogenetic clades that were constructed using both the 7TMs (including ECLs and ICLs) and the ORFs were compared and contrasted to assess the propriety of using either method in this study.

Snake plots were generated using Protter (wlab.ethz.ch/protter) (36) for the functionally annotated Cm_CrzR, Cm_RPCHR and Sp_CCAPR, and three transcripts from this catalogue (Gl_GPCR_A7, Po_GPCR_A146 and Cm_GPCR_A117) that respectively shared a high degree of sequence identity. Snake plots of the ACPR1 and ACPR2 clades were generated to compare the intra-clade and inter-clade similarities and differences, and especially the ECLs. All snake plots were augmented with the species, conserved GPCR motifs, and a color-coded annotation.

The aa sequences of the 7TM (including ECLs and ICLs) of the GnRHR-like transcripts were aligned and Neighbor Joining phylogenetic trees were constructed, as described above. The phylogenetic trees were also labeled with species and associated DRY motif sequence. The compartmentalization of labels and the strength of phylogenetic clades were analyzed and visually represented. In considering the two ACPR clades (grey and orange), that lacked a functionally annotated constituent, we identified the sequences of the conserved motifs of the two functionally annotated ACPR-like receptor variants, BNGR-A28 (accession number: NP_001127726.1) and BNGR-A29 (accession number: NP_001127745.1) of silkworm moth Bombyx mori (37). The sequences of the conserved DRY, CWxP, NPxxY motifs of all receptor transcripts were tabulated and sorted. The congruence and compartmentalization of species and conserved motifs was analyzed, and color-coded annotations were then ascribed.

A representative of each of the color-coded GnRHR-like clades was subjected to CrustyBase expression pattern analysis. That is, a tBLASTN search of the aa was conducted for the following four molt-related transcriptomes: multiple tissues, 11 stages of embryo development, 12 stages of larval development of the ornate spiny lobster (P. ornatus), and Y-organ across five molt stages of the blackback land crab (G. lateralis). The contigs with E-values less than approximately 1e-50 were all downloaded. The GPCR topological domains (predicted using DeepTMHMM) and conserved GPCR motifs (DRY, CWxP and NPxxY) were sequenced for these contigs to determine which clade they actually represented. This is, as opposed to just returning the lowest E-value in the tBLASTN search. The expression patterns of the legitimized contigs were then compiled for each of the transcriptome contexts and are presented in the results.

The class A and class B GPCR-encoding transcripts of P. clarkii, H. americanus, S. verreauxi, Cancer borealis, N. norvegicus, C. maenas, G. lateralis, and P. monodon, whose aa sequences have been predicted previously by the authors and others, were subjected to contemporary workflows and compiled in one convenient spreadsheet (retaining original names, as applicable). We also present the novel GPCR-encoding transcripts of P. ornatus, C. quadricarinatus and E. sinensis (see Supplementary Material S1 ).

Transcript ORFs that clustered strongly with the seeded transcripts of predicted receptors of the GnRH superfamily of neuropeptides (that is, displaying E-values less than 1e-50) were identified and extracted (see Supplementary Material S1 ). Figure 1 depicts the clustering of the class A receptor transcripts ( Figure 1A ) and the constituent members of the GnRHR-like superfamily ( Figure 1B ). The topological GPCR domains of the GnRHR-like predicted ORFs, including the N-terminus and C-terminus, ICLs and ECLs, and the canonical 7TM domains have been predicted and are presented in Supplementary Material S1 . DRY, CWxP, and NPxxY motif loci were sequenced for the GnRHR-like transcripts and have also been reported (see Supplementary Material S1 ).

Table 2 summarizes the diversity observed in the aa sequences of the function-related DRY motif loci of this dataset of decapod GPCRs and, for comparison, a dataset of human GPCRs analyzed by others (8). Noting the high degree of conservation of the DRY motif, and the similarity between datasets.

Supplementary Material S2 : Table 1 depicts the diversity of aa residues observed at the DRY motif locus for the 1289 decapod GPCR-encoding transcripts of this study. The conserved DRY motif was observed in 40.8% of the transcripts. Whereas ERY (10.6%), DRF (8.8%), ERF (6.8%), GRF (1.8%), DKY (1.7%), DSY (0.9%), DRC (0.7%), ERC (0.6%), GRY (0.5%), and VRY (0.3%) collectively were observed in 32.7% of the transcripts. Meaning that the variation observed in the DRY motif loci of 73.5% of the transcripts is explained by a point mutation in a single nucleotide of an aa codon. DRY locus motifs containing R as the second aa residue were observed in 93.8% of the transcripts (refer to Table 2 and the left-hand column of Supplementary Material S2 : Table 1 ). Interestingly, in the third aa or ‘Y’ locus of the DRY motif, an aromatic aa (that is, W = tryptophan, Y = tyrosine, F= phenylalanine and H= histidine) was observed in 84.5% of the transcripts. Moreover, DRM and GRF motifs were only observed in 0.9% and 1.8% of the GPCR-encoding transcripts, respectively. The clusters of motifs that deviated from the DRY sequence represented in Supplementary Material S2 : Table 1 (that is, away from the upper left corner) are noted as being biologically less probable, as is detailed in the accompanying notes ( Supplementary Material S2 ). These deviations raise the possibility of function-related consequence if they are not bioinformatic anomalies (for example GRF sequence in the DRY locus).

The snake plots in Figure 2 depict the topological domains of three pairs of GnRHR-like proteins. The receptors on the left-hand side are the functionally annotated receptors of corazonin, RPCH and CCAP. The three receptors on the right-hand side (Gl_GPCR_A7, Po_GPCR_A146 and Cm_GPCR_A117) are transcripts from this dataset that share aa sequence identity with the annotated receptors. Note the high degree of similarity of the N-termini, ECLs, ICLs, C-termini, conserved N-glycosylation sites and conserved motifs (that is, DRY, CWxP and NPxxY). Observed differences in the lengths of N-terminus and C-terminus may be a sequencing artefact.

The relationship between the phylogeny of the 7TMs (including ICLs and ECLs) of the GnRHR-like protein transcripts, and the associated sequence of the DRYx motif locus and species is summarized in Figure 3 . Noting the high degree of compartmentalization of the labels (that is, species and DRY motif sequence), and very high bootstrap values within color-coded clades. Moreover, the receptor pairs depicted in the snake plots of Figure 2 cluster together.

Referring again to the snake plots ( Figure 2 ) and the DRYx motif ( Figure 3 ), we present this as evidence that Gl_GPCR_A7, Po_GPCR_A146, and Cm_GPCR_A117 are orthologues of Cm_CrzR, Cm_RPCHR and Sp_CCAPR, respectively. Thereby representing candidates for functional annotation as receptors of corazonin, RPCH, and CCAP, respectively.

The Neighbor Joining phylogenetic trees of both the 7TMs (including ECLs and ICLs) and of the ORFs of the GnRHR-like receptors (seen in Figure 3 and Supplementary Material S2 : Figure 1 , respectively) are presented for comparison. The eight discernible and color-coded clades were composed of the same constituent receptors and had similar bootstrap values. The yellow, green and magenta clades contained the functionally annotated receptors of corazonin, RPCH and CCAP, respectively. The maroon and light red clades contained the two bioinformatically annotated variants of the VR. The blue clade contains Ha_GPCR_A67 of which GNRH2R (accession number: XP_042226332.1) (38) is a truncated version. Notably, the GnRHRs (blue) were more closely related to the VR/CCAPR clade than the CrzR/ACPR1/ACPR2/RPCHR clade, respectively (albeit with the low bootstrap values of 54 and 34). Snake plots of the ACPR1 (orange clade) and ACPR2 (grey clade) receptors are included in Supplementary Material S2 : Figures 2 , 3 , respectively. Comparison and contrast of the topological domains depicted in these figures reveals a high degree of intra-clade similarity, and inter-clade similarity but with a discernible degree of inter-clade difference. For example, the length of ECL3, which may have implications for function and/or ligand binding. The orange clade contained the bioinformatically annotated ACPR1 receptor, Mr_ACPR.

Table 3 summarizes the canonical conserved motifs (DRY, NPxxY, and CWxP) observed in GnRHR-like transcripts of our dataset and in the aa sequences of GPCRs annotated by others (functionally and bioinformatically). Noting the high degree of conservation, congruence and compartmentalization of sequence motif variants within clades of the same predicted ligand, and by extension, the same predicted function. Members of the GnRHR clade (blue) were observed to possess a distinctive DRHEAV sequence in the DRYxxx motif, and lack any discernible CWxP motif, which was exceptional in the dataset tested. Further we noted that tyrosine (Y) is aromatic, hydrophobic and non-charged, whereas histidine (H) is aromatic, hydrophobic and positively charged, which likely has implications for functionality. Altogether, suggesting that the GnRHR receptors (blue) represent a putative novel GPCR model. Moreover, as new models they will likely need novel bioassays for functional annotation. Considering the ACPR clades (orange and grey), the silkworm moth B. mori ACPR-like BNGR-A28 and BNGR-A29 had conserved DRYxxx, CWxP and NPxxY motifs of DRFFAI, CWFP and DPLVY, and DRFFAV, CWLP and NPLVY, respectively. Noting a, not unexpected, lack of total congruence between the conserved motifs of the two ACP receptors of B. mori (Taxonomical order: Lepidoptera) and the decapods of this study (Taxonomical order: Crustacea) (see Table 3 ). Considering all of the above, this suggests that the yellow, green, magenta, maroon, light red, blue, orange and grey clades represent CrzR, RPCHR, CCAPR, VR1, VR2, and the newly annotated GnRHR, ACPR1 and ACPR2, respectively.

Figure 4 summarizes the findings of the CrustyBase expression pattern analysis of the GnRHR-like transcripts. In general, across tissues of P. ornatus, gene expression levels were low, however, the highest expression was in neural tissues (eyestalks, brain, thoracic ganglia), antennal gland, and the adult testis. Higher expression of GnRHR-like and CCAPR-like were observed throughout Po embryo and larval developmental stages, respectively. Notably the expression of CrzR-like was elevated over the molt stages of G. lateralis. ACPR2 (grey) and VR2 (light red) were observed to be expressed in the adult testis, suggesting putative reproduction-related function.