Previous studies have shown that VKRs constitute a distinct RTK family. These receptors were described in 15 different protostomes, including insects and the platyhelminth S. mansoni. A vkr gene was also found in the deuterostome Strongylocentrotus purpuratus 12 . The aim of this study was to extend the VKR family to a large number of phyla in order to evaluate the place of these receptors in the animal kingdom and to contribute to a better understanding of their evolution.

Using BLAST approaches, we searched for vkr genes in multiple genomic databases (see Methods). Results showed that vkr genes were present in at least 36 species from six distinct phyla (Cnidaria, Echinodermata, Platyhelminthes, Mollusca, Annelida and Arthropoda, see Table 1). The protein architecture of each putatively encoded VKR was analysed using the SMART online software http://smart.embl-heidelberg.de/ in order to verify that newly discovered genes effectively encoded structurally conserved proteins, composed of an extracellular VFT module and an intracellular TK domain linked together by a single transmembrane α-helix. In Platyhelminthes, two putative vkr genes were detected in both Clonorchis sinensis and S. mansoni 14 parasitic trematodes. By contrast, only one vkr was found in the genome of the three parasitic cestodes, Echinococcus multilocularis, Echinococcus granulosus and Hymenolepis microstoma. No unequivocal VKR sequence could be found in the genome of the planarian Schmidtea mediterranea, but a sequence encoding a polypeptide composed of a TK domain and a truncated VFT module (GenBank: AAWT01078636.1) was identified, suggesting that a vkr gene might still exist in Turbellaria. However, deeper investigations are needed to extend reliably the existence of VKR to planarian species and to the whole of the Platyhelminth phylum. Several new vkr genes were also found in diverse insect genomes. We identified 15 new vkr genes in 13 species belonging to the Hymenoptera and Lepidoptera orders. In Hymenoptera, on top of the two genes previously characterized in the honeybee Apis mellifera and in the parasitic wasp Nasonia vitripennis 12 , we were able to identify 11 new vkr genes in the genomes of eight Formicidae and three Apidae species. In Lepidoptera, we could show that both the silkworm Bombyx mori and the monarch butterfly Danaus plexippus possess two vkr genes located in tandem on a single scaffold (GenBank accession numbers DF090406.1and JH386161.1, respectively). Additionally, we could find truncated vkr sequences in other arthropod genomes, such as those of Rhodnius prolixus (Hemiptera), Glossina morsitans morsitans (Diptera) and Daphnia pulex (Cladocera). However, the bad scores of similarity registered from sequence alignments and the lack of protein structure conservation led us to exclude these sequences from further phylogenetic analyses. These were performed to assess that the predicted proteins were VKR orthologs belonging to the same family. A maximum likelihood phylogenetic tree was generated under the JTT+I+G model with the support of three outgroups composed respectively of insulin (outgroup 1), RTK-like orphan (ROR) (outgroup 2) and epidermal growth factor (EGF) (outgroup 3) receptors of invertebrates (Figure 1). Results showed that all VKR sequences consistently form a monophyletic group distinct from the three other RTK families. Inside of the VKR family, distinct and robust groups were formed by hymenopteran, coleopteran or dipteran sequences. Platyhelminth VKRs form a cluster made of two distinct cestode and trematode branches. In the trematode branch, VKR1 proteins grouped distinctly from the VKR2 ones. Similarly, lepidopteran VKR1 and VKR2 sequences (BmVKR1/2 and DpVKR1/2) were split into two different branches, but surprisingly lepidopteran VKR1 proteins were rejected out of the arthropod group and placed at the root of the VKR tree. Finally, the cnidarian (NveVKR), annelid (CtVKR), mollusc (LgVKR) and echinoderm (SpVKR) proteins were correctly bound to the VKR tree, but no conclusion can be drawn about their phylogenetic proximity since only one vkr is available for each phylum. Thus, in these studies, we have confirmed the distribution of VKR in arthropods (in insects particularly), platyhelminths and echinoderms and found new VKR orthologs in three additional phyla, Annelida (Capitella teleta), Mollusca (Lottia gigantea) and Cnidaria (N. vectensis) (Figure 2). Additionally, putative VKR sequences were found in EST databases of different orders of Arthropoda (Cochliomyia hominivorax (gb|FG295125.1), Rhipicephalus microplus (gb|FG302900.1), Coptotermes formosanus (GI:345171826) and Crassostrea gigas (GI:313329111)), in the mollusc Biomphalaria glabrata (Contig1132.1, Biomphalaria glabrata Genome Initiative, biology.unm.edu/biomphalaria-genome/index.html), in the annelid Helobdella robusta (gb|EY370614.1) and in the echinoderm Paracentrotus lividus ( emb|AM524433.1). The identification of a putative VKR in this second echinoderm indicates that vkr genes could be present in multiple deuterostomes, and excludes a recent horizontal gene transfer or a genomic material contamination. The presence of VKR in Cnidaria, an animal lineage early diverging from Bilateria, suggests that the VKR family emerged prior to the expansion of Bilateria, the clade that comprises almost all extant animals. However, though we identified NveVKR in the anthozoan Nematostella, we could not detect any VKR sequence in the genome of the hydrozoan Hydra magnipapillata, a result in agreement with the recent genome-wide RTK screening performed for this species 15 . Such a difference could be related to those already described for gene diversity and content between anthozoan and hydrozoan genomes 16 . At this time, cnidarians are the first branch of Metazoa in which vkr genes have been found and the question of their presence or not throughout all phyla of Bilateria is still open. Extensive research in vertebrate and roundworm genomes allowed us to conclude that VKR is absent from Chordata and Nematoda but the lack of genomic data for species of the Acoela, Rotifera, Ectoprocta and Brachiopoda phyla still stands in the way of a better understanding of the place of VKR throughout evolution (Figure 2). Moreover, the existence of vkr in Premetazoa remains to be considered because we have recently noticed in choanoflagellates that the “unclassified” TK annotated as UTK12 of Monosiga brevicollis 5 as well as the RTKS kinase of Salpingoeca rosetta 17 have both a protein architecture similar to that of VKR proteins.

Vkr gene name is specified for each species, as well as protein or scaffold accession number and website link of the database in which the gene was found.

Previous studies have already demonstrated that vkr genes were substantially heterogeneous in length as well as in intron-exon composition throughout species, while their organization within a given order was rather well conserved, like in Diptera 12 . In this work, we have made an extensive study of the exon/intron structure of all the discovered vkr genes using the GenScan 18 and Augustus 19 gene prediction Web servers. Data in Table 2 illustrate the gene composition of vkr genes from all the species grouped into families. They indicate for each vkr gene the total number of exons and specify the exons that encode VFT, TM or TK domains. The results confirmed a wide heterogeneity for vkr genes across the diverse phyla. In arthropods, vkr genes have highly variable size (estimated from 4 to 65 kb) and exon numbers (from 5 in Drosophilidae and Culicidae to 13 in the ant Acromyrmex echiniator). As it was already observed in the case of dipterans 12 , sizes and/or exon-intron structures were homogeneous inside of the hymenopteran order, particularly for Formicidae and Apidae. The low number of exons found in vkr genes of Drosophilidae and Culicidae could be due to the high degree of intron loss that has occurred during the evolution of the protostome lineage leading to flies and mosquitoes 20 .

Data illustrate the composition of vkr genes from all the animal species (see Table 1) which have been classified into their respective phylum, order and family. They indicate for each vkr gene the number of exons and specify those that encode VFT, TM or TK domains. Stars indicate genes for which the complete cDNA sequence was obtained.

Vkr genes found in lophotrochozoan organisms (annelids, molluscs, platyhelminths), are overall more complex (15 to 18 exons) than the insect ones, except in the cestode H. microstoma (11 exons). Lophotrochozoan vkr genes are according to this more similar to that one detected in the phylogenetically basal animal N. vectensis. Indeed, Nvevkr is also intron-rich (15 exons), respecting therefore the known high complexity of the genes present in early animal genomes 20 . Spvkr found in the echinoderm S. purpuratus remains the most complex vkr gene found with a size of 60kb and a total of 21 exons. In trematodes, the organisation of Smvkr1 and Smvkr2 genes was shown to be quite identical (see Table 2, and 14 ), arguing for a duplication event in S. mansoni.

As a preliminary approach to understand vkr gene evolution, we have analysed the conservation of intron positions throughout the diverse animal groups (Figure 3). First results overall indicated that vkr genes were all sharing at least one conserved intron position, arguing that they might belong to the same family.

We also noticed that 12 out of 14 intron positions of Nvevkr were found in at least one phylum and that most of these conserved exon-intron boundaries were located in regions that encode VFT and TK domains. Most of these positions were shown to be conserved in the annelid (Ctvkr), mollusc (Lgvkr) and echinoderm (Spvkr) genes. Inversely, only a very limited number of intron positions seem to be conserved in arthropods as well as in platyhelminthes, indicating a profound reorganization of vkr genes along evolution. Finally, specific or non-conserved intron positions were also found in various vkr genes (Figure 3). Taken together, these studies demonstrate that vkr genes are highly variable in size and in complexity but that, in spite of their heterogeneity, all of them possess common features, which are conserved from Cnidaria to the other phyla.

Using the multiple alignment ClustalW algorithm, we have compared the TK domain sequences of the 40 VKRs and generated an identity matrix. As it could be expected for catalytic structures, the sequences are relatively well conserved across all species, with the best scores of identity observed between the species belonging to a same order. For example, in Hymenoptera the TK domains of Bombus and Apis species are more than 92% identical, in Diptera, those of Drosophila species are more than 75% identical and those of the mosquitoes A. gambiae and A. aegypti are 93%. For platyhelminth VKRs, identities between TK domains scored between 57 to 96%, with the best score registered between the two cestode parasites E. multilocularis and E. granulosus. In Lepidoptera, the TK domains of Bombyx and Danaus VKRs were less conserved, except for BmVKR2 and DplVKR2 that share 81% of identity.

In the aligned sequences, we could identify most of the residues shown to be essential for TK activity 21 . As indicated in Figure 4, the glycine-rich motif G₈xGxxGxV₁₅ required for the correct positioning of ATP is found in all VKRs except in that of D. ananassae, which lacks the first G residue. The V₃₂AxK(16x)E₅₂ motif essential for ATP stabilization is also tightly conserved, except for the hymenopteran N. vitripennis VKR. The phosphotransfer site H₁₄₇RD (L/V/I)xxRNxL₁₅₆ is also present in the catalytic loop of all VKRs but in the insect VKRs of P. humanus corporis and H. saltator, this motif is interrupted by an insertion. Without exception, all VKRs possess the D₁₉₄FG₁₉₆ site essential for the binding to the catalytic magnesium ion. In the activation loop, the two juxtaposed Y₂₁₁Y₂₁₂ autophosphorylation site that allows an open access to ATP and substrates in many activated RTKs (like insulin receptors) 22 , is found in all VKRs, except in VKR1 isoforms of B. mori and D. plexippus in which a single Y is found at this position. The M₂₆₂(A/S)PE₂₆₅ motif implicated in the stabilization of the active kinase core is present in most VKRs, but the A/S residue is replaced by a P residue in all the Drosophila species as well as in cestodes. We can note also that this motif is totally absent from the VKR of the ant L. humile. Other hydrophobic residues M₅₆, L₆₇, L_91, (L/V/I)_284, (L/V/I)₂₈₈ , that compose in every active kinase spatially conserved motifs, termed spines, and that play a major role in kinase dynamic assembly and activity 23 , are present in all VKRs. Overall, these data showed that a large part of VKR sequences contained all the motifs essential for TK activity. Furthermore, our previous demonstrations 12 14 that recombinant receptors from A. mellifera and S. mansoni were catalytically active and able to autophosphorylate, strongly suggest that most of the VKRs identified in this work might exhibit similarly kinase activity. Concerning the ones (DaVKR, NvVKR, BmVKR2, DplVKR2, PhcVKR, HsVKR and LhVKR), that lack one or several motifs essential for catalytic activity, we could suggest that they constitute dead or pseudokinases, but this conclusion should be taken with much caution because of possible artefacts in gene prediction or sequences. Further analyses of their kinase potential would be needed to conclude.

Multiple alignment of the VFT module sequences from the 40 VKRs shows that protein sequences are relatively well conserved, particularly in insect species (Figure 5). Excluding the two lepidopteran VKRs, all insect VFT sequences share from 29% (between PhcVKR and DpseVKR) to 98% (between the two Bombus species) of identity. Interestingly, VFT sequences are highly conserved within a given order. As an example, VKRs from the 13 hymenopteran species all share at least 70% identity. Likewise, there is at least 50% identity between the VKRs identified in all dipteran species. Taken together, these results agree with an evolutionary conservation of insect VFT sequences. However, an exception concerns the lepidopteran VKRs (BmVKR1/2 and DplVKR1/2) for which VFT module sequences are highly divergent from each other and from all insect VFT domains (less than 16% of identity). Still, the BmVKR2 VFT module seems closer to that of DplVKR2 (40% of identity). Concerning the VFT sequences of platyhelminths, intermediate levels of conservation were observed between trematode and cestode sequences (24% to 44% identity) while we could note, as previously observed for the TK domain, a marked identity (93%) within the genus Echinococcus between EgVKR and EmVKR VFT domains. Finally, other VKRs (from Annelida, Mollusca and Cnidaria species) were found to be highly divergent from platyhelminth and lepidopteran VKRs in their VFT domains, but as discussed above concerning their phylogeny (Figure 1), available sequences are not sufficient to state about receptor evolution in these phyla.

VFT modules constitute the binding pocket of various receptors activated by small molecules 13 . They are composed of two lobes connected via flexible tethers that close around the bound ligand. In most class C GPCRs, these modules contain the binding site for natural amino acids or derivatives and ligand recognition is dependent on a consensus motif of 8 residues that participate to the binding of the α-amino acid functions (i.e. primary amine and carboxylic acid) 24 . Among these residues, the serine that binds the COOH group of glutamate in mGluR1 (Ser₁₆₅), is the most conserved residue in class C GPCRs. In most VKRs (except in cestode and lepidopteran receptors), this residue is strictly conserved (at position 66 in the VFT sequence, Figure 5). The R_107, another important residue composing the consensus site for AA binding in VFT modules 24 is also present in all VKRs (except in Lepidoptera). These observations are totally in agreement with our demonstration that amino-acids, and particularly L-arginine, are able to bind and activate both honey bee and schistosome VKRs 12 14 and we have recently confirmed the requirement of S₆₆ and R₁₀₇ for the amino-acid recognition by VKR receptors (unpublished). Interestingly, if the six other elements composing the consensus motif in VFT modules for amino acid binding 24 are different in the ligand-binding domains of VKRs, we could note that at their exact positions, many VFT domains present identical residues (highlighted in green in Figure 5), and this is particularly obvious inside of the Hymenoptera order. Among these residues, Y₄₀₆ was found to be present in all VKRs (except in lepidopteran sequences). Y₄₀₆ is exactly at the position K₅₀₉ in mGluR1, the residue involved in the binding of the terminal carboxylic group of glutamate 12 , and this strengthens our observation that glutamate is not a ligand for VKR. Additionally, we can see in the VFT alignment several conserved residues and motifs (highlighted in red, Figure 5), such as C₃₅ and W₁₂₈ which are present in all domains and the motif (G₂₄₇Y(V/I)WFLPxWL₂₅₆) which is present in most VKRs. Finally to investigate the potential importance of these conserved residue positions for the VFT properties, we performed a comparative modeling of the VFT domain of VKR using the ModWeb server of the ModBase databases. A VFT model of AmVKR was generated with the human Glutamate Receptor 5 (mGluR5) as homologous template (PDB: 3lmkA) (Figure 6). Results confirmed the position inside of the ligand pocket formed between the two lobes of the VFT, of the residues potentially involved in amino-acid recognition. The model also indicates that many of the other conserved residues are constituents of alpha-helices or beta-sheets, and thus take part very likely in the tertiary structure of the domain. Interestingly, two residues T₁₀₈ and I₁₀₉, highly conserved in most VKRs, are located in the putative ligand binding pocket, suggesting that they could contribute with the conserved residues of the amino-acid binding motif to ligand specificity and/or affinity. Taken together, these results let suppose that in spite of a partial conservation of primary structures, the VFT domains of many VKRs can bind a common ligand.

Putative vkr sequences were searched using tBLASTn on genomic sequences available in the following databases: GenBank (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi), FlyBase (http://flybase.org/), VectorBase (http://www.vectorbase.org), Wellcome Trust Sanger institute databases (http://www.sanger.ac.uk/resources/databases/) and the JGI genome portal (http://genome.jgi-psf.org/). Additionally, we also searched for vkr sequences in Hydra magnipapillata (hydrazome), Schmidtea mediterranea (SmedDB) and Schistosoma japonicum(schistoDB) genome databases.

Selected genomic sequences were analysed by GenScan (http://genes.mit.edu/GENSCAN.html) and Augustus (http://bioinf.uni-greifswald.de/augustus/) gene prediction servers, and putative VKR proteins were then determined. The presence of VFT, TM and TK domains was verified, and their delimitation defined using the SMART software (http://smart.embl-heidelberg.de/). Finally, TM regions were confirmed with the TMHMM server (http://www.cbs.dtu.dk/services/TMHMM).

Protein sequences (listed in Additional file 1) were aligned using ClustalW algorithm in the BioEdit v7.1 software, and manually corrected. Maximum likelihood trees were built using MEGA5 25 under the JTT+I+G model, with 100 bootstrap repetitions.

Comparative modeling was performed on the VFT sequence domain of VKRs using the comparative modeling web-server Modweb (https://modbase.compbio.ucsf.edu/scgi/modweb.cgi) of the ModBase databases. Calculation and evaluation of models were performed with ModPipe software using sequence-sequence, sequence-profile and profile-sequence methods for fold assignment and target-template alignment.