cnidarian opsins are evolving by linage-specific duplications
In order to investigate patterns of molecular evolution of opsins in H. vulgaris we first curated opsin sequences in the recently released and improved genome, Hydra 2.0 Genome Project (formerly H. magnipapillata) . By searching a ab initio transcriptome, phylogenetically-informed annotation (PIA) database , and an improved reference genome, we identified 45 opsin genes in H. vulgaris (Supplemental Table S1-2). This result differed from that of 63 opsin genes found by Suga et al.  using the first genome release. Given the highly fragmented nature of the original assembly, we believe that the difference in opsin gene number between our studies is due to misalignments or haplotypes in the original assembly.
Next, we generated a cnidarian opsin phylogeny and included outgroups placozoa, humans, and Drosophila (Fig. 2). We made placozoa the root of the tree as determined by Feuda et al. [3, 42]. Based on previous studies, we expected to see lineage-specific duplications of opsins in Cnidaria with Hydra opsins forming two groups [25, 26] or we expected to see the opsin tree recapitulate the evolutionary history of the species (Fig. 1B-C). Our phylogenetic tree supported claims that opsins are evolving by lineage-specific duplications as Hydra, Cladonema, Tripedalia, and Nematostella opsins group together by species rather than opsin type (Fig. 2). Generally, the opsin phylogeny reflects the cnidarian cladogram with Hydra, Cladonema and Podocoryna closer together, next Tripedalia, and Nematostella a little further away (Fig. 2). Our opsin phylogeny provides support for previous suggested cnidarian opsin phylogenetic relationships. Similar to previous studies, we found ctenophore opsins Mnemiopsis opsin1 and opsin2 grouping together while Mnemiopsis opsin3 branches separately (Fig. 2) [24, 42]. We also found that Podocoryna opsins do not group together  and that both Cladonema and Tripedalia opsins form 2 groups [25, 26].
Interestingly, we discovered some differences from previous studies as to the placing of a N. vectensis opsin group and two H. vulgaris opsins. Suga et al. and Liegertová et al. found that N. vectensis opsins cluster into 3 and 4 groups respectively [25, 26]. In Suga et al. one of the opsin groups clusters together with the other cnidarian opsins while one of the groups is sister to rhabdomeric opsin (R-opsin) and another is an outgroup . Conversely, Liegertová et al. found that two opsin groups fell within the cnidopsins, one group was potentially sister to ciliary or rhabdomeric opsin, and they were not able to place the fourth group . Here, we found that Nematostella opsins formed three groups; group 3 clusters with the cnidopsins, group 2 is outside of ciliary opsins (C-opsin) and cnidopsins, and group 1 is sister to rhabdomeric opsins (Fig. 2). Regarding Hydra, Suga et al. and Liegertová et al. found that Hydra opsins cluster into 2 and 3 groups respectively [25, 26]. We found that H. vulgaris opsins clustered into 2 main groups, but we also uncovered that 2 genes fall outside of these two large groups, so we refer to each of these its own group. HvOpB1 (group B Hydra opsin) falls within Mnemiopsis opsin3 and outside of a group of cnidopsins and HvOpA1 (group A) is sister to a group of Placozoan opsins (Fig. 2).
As a majority of the cnidarian opsin genes form clusters, this suggests that opsin genes are expanding by linage-specific duplications rather than a large expansion in their common ancestor. In addition, we named our opsin genes based on location on the genome and found that many H. vulgaris opsin genes that are in close proximity in the genome are also next to or very close to each other on the phylogeny. As an example, opsin genes in group C HvOpC1-5 are all on the same scaffold (Table S1) and next to each other on the phylogeny (Fig. 2). HvOpD1-4 are also on the same scaffold but only HvOpD2-3 group together. HvOpD5-6 are on the same scaffold and branch together on the phylogeny. Other examples include HvOpD9-10, HvOpD12-15, HvOpD16-19, and HvOpD22-24. These groupings of genes on same scaffolds in the opsin phylogenetic tree suggest that H. vulgaris opsins could be expanding by tandem duplications (Fig. 2).
Expression patterns of H. vulgaris opsins in the Hydra body, during budding, and during regeneration
Investigating the expression patterns of genes, especially when comparing tissues, can give some insight into their potential functions. We quantified the expression of the H. vulgaris opsins in the H. vulgaris body, during budding, and during regeneration . Opsin genes that were expressed more highly (>2 fold change) in the foot compared to other tissues were HvOpD21, HvOpD27, HvOpD33, HvOpD36, and HvOpD38 (Fig. 3A; Fig. S1A). Interestingly, all of these genes are near each other on the opsin phylogeny and belong to an opsin gene cluster for which a Podocoryna opsin is an outgroup (Fig. 2). In the hypostome, the genes that were more highly expressed (>2 fold change) relative to other tissues were HvOpB1, HvOpD2, HvOpD11, HvOpD12, HvOpD14, HvOpD15, HvOpD19, HvOpD29, HvOpD32, and HvOpD37 (Fig. 3A; Fig. S1A). These genes are not all near each other on the phylogeny, however HvOp12, HvOp14 and HvOp15 belong to a branch that includes genes located on the same scaffold and they have similar expression patterns across tissues (Fig. 4A). In the tentacle, opsin genes HvOpC1, HvOpC2, HvOpC4, HvOpD4, HvOpD8, HvOpD9, HvOpD13, HvOpD22, HvOpD23, and HvOpD24 were expressed more highly (2x) relative to other tissues (Fig. 3A; Fig. 4A). HvOpC1-2 and HvOpC4, and HvOp22-24 are next to each other in the genome, have similar sequences based on the opsin phylogeny, and have similar expression patterns across tissues. This suggests that these genes may have shared functions (Fig. 2, Fig. S1A).
We hypothesized that some of the genes that were expressed more highly in the hypostome and tentacles relative to other tissues would have expression that increased during budding and regeneration. For the hypostome, HvOpB1 increases in expression during both budding and regeneration (Fig. S1A-C). HvOpD2 and HvOpD37 increase in expression during regeneration but do not show a temporal trend during budding (Fig. S1B-C). Conversely, HvOpD14 and HvOpD32 increase in expression during budding but do not have a directional change during regeneration (Fig. S1B-C). For the tentacle, HvOpD4 increases during both regeneration and budding. HvOpD13 only increases during budding while HvOpD24 and HvOpC2 increase during regeneration. These findings are interesting because HvOpB1 is one of the most highly expressed genes in the hypostome and HvOpC2, HvOpD4, and HvOpD24 are some of the most highly expressed genes in the tentacle and these four genes all show trend of increasing either in budding, regeneration, or both. These genes likely play an important function in the Hydra head. It is important to note that HvOpB1 falls outside of the two H. vulgaris opsin gene groups C and D. Instead, HvOpB1 serves as an outgroup to all Hydrazoan opsins and one group of the Tripadelia opsins.
While HvOpB1, HvOpC2, HvOpD4, and HvOpD24 are expressed highly in the H. vulgaris head region and have dynamic expression during budding and regeneration, we found another interesting gene due to its very high expression in H. vulgaris. HvOpA1 is expressed almost 200-fold more than the other opsin genes (Fig. 4). We did not detect a significant difference in expression between body parts nor during different stages and times of budding and regeneration. The high expression of this gene throughout the H. vulgaris body suggests that it is a gene of importance with a general function. Similar to HvOpB1, HvOpA1 does not fall within the H. vulgaris opsin gene clusters. Instead, HvOpA1 groups with Placozoan opsins (Fig. 2).
To increase our power, we also looked at opsin expression across all samples used together (Fig. 5A; Fig. S2). From this analysis we notice three sets of genes that are upregulated in the hypostome, tentacle or foot. According to gene expression z-scores across all samples HvOpB1, HvOpD3, HvOpD11, HvOpD15, HvOpD19, HvOpD29, and HvOpD37 have higher expression in the hypostome compared to other tissue types and also increased during budding. HvOpC1, HvOpC2, HvOpC3, HvOpC4, HvOpC5, HvOpD1, HvOpD4, HvOpD7, HvOpD8, HvOpD9, HvOpD10, HvOpD16, HvOpD18, HvOpD22, HvOpD23, HvOpD24, and HvOpD26 group together as having similar expression patterns and were more highly expressed in the tentacles compared to other tissue types and time points in budding and regeneration (Fig. 5A; Fig. S2). HvOpD21, HvOpD27, HvOpD33, HvOpD36, and HvOpD38 are more highly expressed in the foot compared to other tissue types and time points in budding and regeneration (Fig. 5A; Fig. S2). For the most part, an analysis among all samples reflects similar patterns of gene expression as an analysis between tissue types.
Phototransduction cascade genes in H. vulgaris
In order to detect whether any of these opsins might function similar to vertebrate ciliary or invertebrate rhabdomeric opsins, we searched the Hydra for phototransduction genes using M. leidyi sequences following the example of Schnitzler et al. . As mentioned above, cnidarians are of interest because they are basal invertebrates with ciliary opsins similar to vertebrates, thus we expected to find evidence of ciliary phototransduction cascade components. Ciliary and rhabdomeric photoreceptors are similar in that the general transduction pathway is the same beginning with activation by rhodopsin, transduction via G-protein coupled receptor and ion channels, and finally termination. However, some of the messenger genes that they employ vary. In Drosophila melanogaster (a model for invertebrate phototransduction), activation of rhodopsin by light causes the release of Gαq which activates phospholipase C (PLC) . The transduction of the signal is carried out by Ca2+-permeable transient receptor potential (TRP) channels that cause depolarization of the cell [44, 45]. Finally, phototransduction is terminated when the activated rhodopsin (metarhodopsin) binds arrestin or is phosphorylated by rhodopsin kinase [46–48]. In vertebrates, activated rhodopsin works through GTP-binding transducin which releases Gtα and binds guanosine monophosphate phosphodiesterase (GMP-PDE) . Instead of TRP, opening of cyclic nucleotide gated ion channels (CNG) cause the photoreceptor cell to hyperpolarize . Similar to ciliary cells, rhodopsin kinase and arrestin terminate the cascade by deactivating rhodopsin . In addition in vertebrates, G Protein-coupled receptor kinase 1 (GRK1) and regulator of G protein signaling 9 (RGS9) regulate G protein signaling while recovering inhibits phosphorylation of light-activated rhodopsin .
Light-detecting rhodopsin is comprised of an opsin protein bound to a retinal molecule known as a chromophore, 11-cis-3-hydroxyretinal in D. melanogaster and 11-cis-retinal in mammals . The chromophore is transported to the photoreceptor cell by a retinal binding protein, cellular retinaldehyde-binding protein (CRALBP) in mammals and prolonged depolarization afterpotential is not apparent (PINTA) in D. melanogaster [50, 51]. The chromophore binds opsins at a conserved retinal-binding lysine in the seventh transmembrane helix. In order to identify which of the H. vulgaris opsins may function in phototransduction, we investigated which had the conserved lysine necessary for chromophore binding. We found that all opsins excluding five have the lysine amino acid necessary for phototransduction. The five opsins missing the lysine were: HvOpA1, HvOpB1, HvOpD4, HvOpD9, and HvOpD26 (Table S1).
In terms of ciliary components, H. vulgaris differed from M. leidyi in that the top hit to G-alpha-i subunit is a G-alpha-o subunit (Gαo) (Table 1). Although we did not have an exact predicted protein match, Gαo and Gαi belong to the same Gα protein subfamily and are expected to have similar functions in signal transduction . H. vulgaris also had two genes similar to Transducin G-gamma-t1 which we refer to as Gtγ1 and Gtγ2, two GMP-PDE alpha rod genes (GMP-PDEα1 and GMP-PDEα2), only one cyclic nucleotide gated ion channel (CNG – as opposed to two in M. leidyi), and the top hit to Recoverin is a Neurocalcin-like gene (Table 1). Neurocalcin is in the same gene family as recoverin and also expressed in the retina but not in the rods and cones . In addition, the top hit for GRK1 was G protein-coupled receptor kinase 5-like (GRK5-like), the top hit for RGS9-1 was regulator of G-protein signaling 12-like (RGS12-like), and the top hit for GC1 guanylyl cyclase were two atrial natriuretic peptide receptor 1-like (ANPR1-like and ANPR1-like2). For the rhabdomeric components, the top hit for TRP-C was an Ankyrin-3-like gene (Table 1). A reason for this might be that ankyrin repeats are part of TRP channels but H. vulgaris is likely missing a TRP ortholog . Lastly, for shared components, H. vulgaris differed from M. leidyi in that we found three Visual G beta genes (Gβ1, Gβ2, and Gβ3) (Table 1).
We next looked at the expression patterns of phototransduction genes to see whether they have similar expressions to the opsins. Interestingly, we identified a group of genes that contained most of the necessary components of the phototransduction cascade and two opsins (Fig. S3). This group had genes with high expression in the tentacle and hypostome and increasing expression during budding. This group included ANPR1−like2, GMP−PDEα2, Gαq, Gtαi, Gβ1, SEC14−like, Arrestin, Neurocalcin−like, Gαs, GRK5−like, GMP−PDEδ, PLC, GMP−PDEα1, ANPR1−like, and Rh kinase (Fig. 5B). When visualized together with the opsins, two opsins HvOpC5 and HvOpD1 have similar expression patterns to these genes (Fig. S3). If similar expression patterns in these genes means that they are expressed together then these results imply that H. vulgaris is using components from both ciliary and rhabdomeric receptors to transduce a signal (see Discussion).
Differentiation trajectories clustering
To further detect in which cells opsins and phototransduction genes are likely expressed in, we looked at which gene clusters they belong to in a stem cell differentiation trajectories clustering by Siebert et al. . We expected to see phototransduction genes and one or more opsins expressed in similar cell clusters. We did not find unique matches for all of our opsin genes, but we were able to determine in which clusters 19 of them are expressed (Table 3). We only listed the top clusters, which we selected based on higher expression and expression in more cells in a cluster. We found that most opsins clustered as cells of the neuronal cells of the interstitial lineage in the endoderm or ectoderm (Table 3). HvOpA1 and HvOpB1 again were expressed in many more cells and cell clusters. HvOpA1 had the densest expression in clusters of the nematocyte and nematoblast of the interstitial cell lineage (Table 3). HvOpB1 was expressed more heavily in granular mucous gland cells and spumous mucous gland cells of the interstitial lineage (Table 3). Unlike the opsins, most phototransduction genes were expressed in all cell clusters (Table S3). Two of the genes that were not expressed in all clusters were CNG and GMP-PDEα1 which were expressed in neuronal ectoderm and endoderm cells of the interstitial cell lineage similar to HvOpC5 and HvOpD1 which we predict might be functioning together in phototransduction (Fig. 6; Table 3; Table S3).