A second visual rhodopsin gene, rh1-2, is expressed in zebrafish photoreceptors and found in other ray-finned fishes

Vertebrate photoreception is mediated by opsins, which are members of the G protein-coupled receptor (GPCR) superfamily of proteins (Terakita, 2005). In the dark, opsins are covalently bound to a light-sensitive chromophore, 11-cis retinal, which acts as an inverse agonist to suppress dark state activation (Menon et al., 2001). When activated by light, the chromophore isomerizes to its all-trans conformation, which initiates a signaling cascade within the cell (Baylor, 1996). Visual opsins are responsible for initiating the visual transduction cascade, while non-visual opsins are involved in processes such as circadian entrainment (Doyle et al., 2008) and the metabolism of retinal (Bellingham et al., 2003a), with some possibly contributing indirectly to image formation (Cheng et al., 2009). Rhodopsin is the visual opsin expressed in rod photoreceptors responsible for mediating dim-light vision in vertebrates (Nathans, 1992), and was the first GPCR to have its crystal structure resolved at high resolution (Palczewski et al., 2000).

While gene duplications have occurred multiple times in invertebrate opsins (Rivera et al., 2010; Serb et al., 2013) and vertebrate cone opsins (Hunt et al., 1998; Matsumoto et al., 2006), visual rhodopsin is generally considered to be a single copy gene, with only a few exceptions. Several eel species have two rhodopsins, one freshwater (rh1fwo) and one marine (rh1dso), and expression shifts from the former to the latter following migration during maturation (Beatty, 1975; Hope et al., 1998; Zhang et al., 2000; Zhang et al., 2002). The short-fin pearleye (Scopelarchus analis), a deep-sea teleost, also expresses an additional rh1 gene, rh1b, in the accessory retina of adult fish after descending to greater ocean depths (Pointer et al., 2007). Other examples of multiple rh1 genes are usually the result of species-specific duplication events (Lim et al., 1997). It is also interesting to note that the middle wavelength-sensitive cone opsin groups, which absorb similar wavelengths of light to rhodopsins, have been suggested to make greater contributions to adaptive vision through gene duplication (Gojobori and Innan, 2009).

Despite the scarcity of rh1 gene duplications, zebrafish (Danio rerio) is a logical candidate for opsin gene duplication, considering most of its tissues are directly photoentrainable (Whitmore et al., 2000) and its array of nine visual opsins is a large complement even among teleost fish (Chinen et al., 2003). Many non-visual opsins have been discovered in zebrafish through traditional sequencing studies, including five melanopsins (Bellingham et al., 2002; Davies et al., 2011), teleost multiple tissue (tmt) opsin (Moutsaki et al., 2003), two vertebrate ancient long (VAL) opsins (Kojima et al., 2008) and exo-rhodopsin (Mano et al., 1999). Additionally, a recent functional genomics screen identified 10 novel non-visual opsins (Davies et al., 2015). The non-visual exo-rhodopsin (exorh), expressed in the pineal gland of the brain and not in retinal photoreceptors, is not involved in vision, but is also orthologous to rh1 of non-teleost vertebrates, with rh1 in teleosts being the product of an ancient retrotransposition event that contains no introns (Fitzgibbon et al., 1995). This duplication is thought to have occurred no later than 284 million years ago, marking the onset of the radiation of ray-finned fish (Hurley et al., 2007), as basal Actinopterygians such as the sturgeon and gar also have intronless rh1 genes (Bellingham et al., 2003b).

We previously identified another rhodopsin-like gene, rh1-2, in juvenile and adult zebrafish, but limited functional characterization prevented its classification as either a visual or non-visual opsin (Morrow et al., 2011). This novel gene was found to be expressed in the retina of adult zebrafish, but not in the brain. When regenerated with 11-cis retinal, Rh1-2 produced an absorption spectrum with a λ_max value of approximately 500 nm (Morrow et al., 2011), similar to both rhodopsin (Chinen et al., 2003) and exo-rhodopsin (Tarttelin et al., 2011). Orthologous sequences were also found in three other cyprinid fish, suggesting that rh1-2 is not the result of a zebrafish-specific duplication event, and initial phylogenetic analyses hinted at a divergence from rh1 during the earlier stages of teleost evolution (Morrow et al., 2011).

Here, we further characterized the expression and function of Rh1-2 by comparing it with both rhodopsin and exo-rhodopsin in order to gain a better understanding of its role in photoreception and its potential as a visual opsin. We localized rh1-2 expression in the retina to a subset of peripheral photoreceptors, a pattern that partially overlaps rh1 expression but that is distinct from exorh expression in the pineal gland of the brain. When monitored using fluorescence spectroscopy following photoactivation, Rh1-2 was shown to release retinal at a rate comparable to rhodopsin and approximately 5 times slower than exo-rhodopsin. Finally, rh1-2 was identified in three additional species, including one outside of the family Cyprinidae, Misgurnus anguillicaudatus, and phylogenetic analyses show the rh1-2 gene family is likely a sister group to ostariophysian rh1 genes, having been subjected to purifying selection following duplication and divergence. This study adds new insights into the visual system of zebrafish, and explores the implications of a second visual rhodopsin gene expressed in some teleost fish.

RNA was extracted from adult eyes of various teleost fishes using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and cDNA libraries were generated using the SMART cDNA Library Construction Kit (BD Biosciences, San Jose, CA, USA). Genomic DNA was extracted from various tissues of teleost fish using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany). All specimens were killed prior to tissue extraction using an overdose of tricaine methane sulfonate (MS222, 300 mg l⁻¹; Sigma-Aldrich, St Louis, MO, USA) buffered to neutral pH with sodium bicarbonate prior to fish immersion. Gene fragments of rh1 and rh1-2 were amplified from cDNA libraries and genomic DNA, respectively, using either previously designed rh1-2 (Morrow et al., 2011) or acanthomorph rh1 primers (Chen et al., 2003), then cloned into the pJET1.2 cloning vector (Fermentas, Waltham, MA, USA). Full-length sequences of rh1, rh1-2 and exorh were amplified from zebrafish eye (rh1, rh1-2) or brain (exorh) cDNA libraries. PCR was performed using PfuTurbo (Agilent, Santa Clara, CA, USA); the resulting fragments were cloned into the p1D4-hrGFP II expression vector (Morrow and Chang, 2010). All vectors were sequenced using a 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA).

In total, 19 new nucleotide sequences are introduced in this study, including 16 new rh1 sequences (GenBank accession numbers: KY026025–KY026040) and three new rh1-2 sequences (GenBank accession numbers: KY026041–KY026043) (Table S1). These sequences were combined with 117 rhodopsin gene sequences (rh1, rh1-2, exorh) obtained from GenBank in order to maintain even sampling across vertebrates within the limits of available data. Among these were four additional rh1-2 genes not previously studied, three from species of the Sinocycloheilus genus that were predicted as rhodopsin-like sequences from a recent cavefish genome project (Yang et al., 2016), and one that was extracted from the carp (Cyprinus carpio) genome. A second putative rh1-2 sequence fragment was also identified within the carp genome, which is not unexpected given a recent tetraploidization (Larhammar and Risinger, 1994), but was not included because of gaps in the first transmembrane domain. Four vertebrate rh2 sequences were also obtained from GenBank and used as an outgroup. Sequences were aligned using the webPRANK (Löytynoja and Goldman, 2010) implementation of PRANK (Löytynoja and Goldman, 2005). Species and accession numbers for all sequences used in the study are provided in Table S1.

Eyes were dissected from 21 and 175 days post-fertilization (dpf) zebrafish, Danio rerio (Hamilton 1822), that had been anesthetized with 160 mg l⁻¹ tricaine (ethyl 3-aminobenzoate methanesulfonate salt; Sigma-Aldrich); eyes were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C overnight, and rinsed in PBS with 0.1% Tween-20 (PBT), then in methanol, before being stored in fresh methanol at −20°C. In situ hybridizations were performed on 3 dpf (control) and 5 dpf zebrafish embryos (whole mount), and eyes from 21 and 175 dpf zebrafish, as previously described (Wong et al., 2010). DIG-labeled RNA probes 700 bp in length, as well as unlabeled, full-length blocking RNA for rh1-2 (control), were amplified from rh1 and rh1-2 sequences inserted into the pBluescript cloning vector using T3 RNA Polymerase (Fermentas, Waltham, MA, USA). For our control using unlabeled rh1-2 blocking RNA, the labeled rh1 probe was pre-adsorbed with blocking RNA prior to in situ hybridization. A probe concentration of 1 ng µl⁻¹ was using during the 70°C hybridization. Semi-thin plastic sections were made from 5 dpf embryos, where whole mounts were rinsed with PBT, dehydrated using increasing concentrations of ethanol (from 30% to 90% in PBT), followed by 100% ethanol, and then embedded with increasing concentrations of Spurr's resin in ethanol (3:1, 1:1, 1:3). Samples were then left to polymerize at 65°C in 100% Spurr's resin. Semi-thin coronal sections were cut with a glass knife using an ultramicrotome and dried onto glass slides. Sections were 1.5 µm thick without counterstaining to maximize visualization. Cryosections were performed on 21 and 175 dpf zebrafish eyes, which were washed 3 times in PBS, then put through a sucrose gradient at room temperature, 30 min per step: 5% sucrose in PBS, 2:1 5%:30%, 1:1 5%:30%, 1:2 5%:30%, with a final step in 30% sucrose in PBS at 4°C overnight. Eyes were incubated in 2:1 30% sucrose in PBS:Tissue-Tek OCT compound (VWR, Radnor, PA, USA) for 4 h at room temperature, then 4°C overnight. Cryosections were performed at 20 μm on a Leica CM3050S cryostat, and collected on Superfrost Plus slides (VWR) mounted in 90% glycerol/10% PBS. All images were taken on a Leica DM4500B compound microscope with a Qimaging digital camera and OpenLab 4.0.2 software (Improvision, Coventry, UK).

The p1D4-hrGFP II expression vector constructs containing full coding sequences of zebrafish rh1, rh1-2 and exorh were used to transiently transfect cultured HEK293T cells (ATCC CRL-11268) using Lipofectamine 2000 (Invitrogen). Cells were harvested 48 h post-transfection and opsins were regenerated using 11-cis retinal, generously provided by Dr Rosalie Crouch (Medical University of South Carolina). Visual pigments were solubilized in 1% N-dodecyl-β-d-maltoside (DM) and immunoaffinity purified with 1D4 monoclonal antibody (University of British Columbia no. 95-062, lot no. 1017; Molday and MacKenzie, 1983), as previously described (Morrow and Chang, 2010). Purified visual pigment samples were eluted in sodium phosphate buffer (50 mmol l⁻¹, 0.1% DM, pH 7). The ultraviolet–visible absorption spectra of purified opsin were recorded at 25°C using a Cary4000 double-beam spectrophotometer (Agilent) and quartz absorption cuvettes (Helma, Paris, France). All λ_max values were calculated after fitting absorbance spectra to a standard template for A1 visual pigments (Govardovskii et al., 2000).

The protocol used to determine retinal release rates of visual pigments was modified from that of Farrens and Khorana (1995). Briefly, 0.05–0.20 μmol l⁻¹ visual pigment samples were incubated in sodium phosphate buffer (50 mmol l⁻¹, 0.1% DM, pH 7) at 20°C using submicro fluorometer cell cuvettes (Agilent) and bleached for 30 s using a Fiber-Lite MI-152 Illuminator external light source (Dolan-Jenner, Boxborough, MA, USA), using a filter to restrict wavelengths of light below 475 nm. Fluorescence measurements were integrated for 2 s at 30 s intervals using a Cary Eclipse fluorescence spectrophotometer, with temperature maintained by a Cary Temperature Controller employing a Peltier Multicell Holder (Agilent) and monitored by a temperature probe. The excitation wavelength was 295 nm (1.5 nm slit width) and the emission wavelength was 330 nm (10 nm slit width); no noticeable pigment bleaching by the excitation beam was detected. Retinal release was demonstrated through a sharp initial rise in intrinsic tryptophan fluorescence, representing a decrease in fluorescence quenching of W265 by the retinal chromophore. Data from the initial rise were fitted to a three-variable, first-order exponential equation y=y_o+a(1−e^−bx), with half-life values calculated based on the rate constant b (t_1/2=ln2/b). All curve fitting resulted in r² values greater than 0.9.

A maximum-likelihood rhodopsin gene tree was estimated in PhyML 3.0 (Guindon et al., 2010) under the GTR+I+G model using a BioNJ starting tree, the best of a NNI and SPR tree improvement, and 100 bootstraps. A Bayesian rh1 gene tree was also constructed in MrBayes 3 (Ronquist and Huelsenbeck, 2003) using reversible jump MCMC with a gamma rate parameter (nst=mixed, rates=gamma), which explores the parameter space for the nucleotide model and the phylogenetic tree simultaneously. The analysis was run for 5 million generations with a 25% burn-in. Convergence was confirmed by checking that the standard deviations of split frequencies approached zero and that there was no obvious trend in the log likelihood plot.

To estimate the strength and form of selection acting on rhodopsin, the alignment, along with the maximum-likelihood gene tree, was analyzed with the codeml package of PAML 4 (Yang, 2007) using random sites models (M0, M1a, M2a, M3, M7, M8a and M8), branch model, branch-site model and clade model C (CmC). Analyses were run on the complete rh1 alignment and tree as well as on two subsets, one pruned to only include ray-finned fish rhodopsin genes (including exorh as the outgroup) and the other pruned to contain only rh1-2 (no outgroup).

Comparisons between the PAML random sites models were used to test for variation in ω (M3 versus M0) and for the presence of a positively selected class of sites (M2a versus M1a and M8 versus M7 and M8a). All analyses were run starting with the branch lengths estimated by PhyML repeated at least three times with varying initial starting points of κ (transition to transversion ratio) and ω to avoid potential local optima. The model pairs were compared using a likelihood ratio test with a χ² distribution.

The branch, branch-site (Zhang et al., 2005) and CmC models (Bielawski and Yang, 2004) were used to test for changes in selective constraint and positive selection on the branch leading to the rh1-2 clade and between the rh1-2 clade and other rhodopsins. The branch model estimates a single ω value for each branch and/or clade type specified a priori. This model is useful for testing for overall changes in selective constraint between branches/clades. The branch-site and clade models allow ω to vary both among sites and between branches/clades. The branch-site model has four site classes: (0) 0<ω₀<1 for all branches; (1) ω₁=1 for all branches, (2a) ω_2a=ω_2b≥1 in the foreground and 0<ω_2a=ω₀<1 in the background, and (2b) ω_2b=ω_2a≥1 in the foreground and ω_2b=ω₁=1 in the background. This model provides a test for positive selection on specified branches/clades and incorporates a Bayes’ Empirical Bayes (BEB) analysis to identify codon sites under positive selection (Yang et al., 2005). CmC assumes that some sites evolve conservatively across the phylogeny (two classes of sites where 0<ω₀<1 and ω₁=1), while a class of sites is free to evolve differently among two or more partitions (e.g. ω_D1>0 and ω_D1≠ω_D2>0), which can be branches, clades or a mix of the two. Rather than a test for positive selection, this provides a test for divergent selective pressure, although a test for positive selection can be performed if desired (see Chang et al., 2012). These models were applied only to the dataset pruned to contain only ray-finned fish rhodopsins.

We performed a series of in situ hybridizations to investigate the onset of rh1-2 expression and to localize its expression in the retina. In situ hybridizations were performed using 700 bp coding sequence probes for both rh1 and rh1-2 in order to localize cellular expression in whole-mount embryos 5 dpf, as well as both juvenile (21 dpf) and adult (175 dpf) eyes. At 5 dpf, expression of rh1 was strongest in the peripheral retina, although some limited expression was also seen in the central retina (Fig. 1A), while rh1-2 was only detected in a limited portion of the ventral peripheral retina (Fig. 1B). At 21 dpf, both rh1 and rh1-2 expression in the peripheral retina was more prominent relative to expression at 5 dpf (Fig. 1C,D). While the central retina contained more widespread rh1 expression (Fig. 1E), there was no rh1-2 expression (Fig. 1F). Expression at 175 dpf was similar to that at 21 dpf, with a strong rh1 signal throughout the photoreceptor layer (Fig. 1G). However, rh1-2 was still localized to the ventral peripheral retina, with no staining shown in the central retina (Fig. 1H). Patterns of rh1 expression were similar to those presented in previous studies (Raymond et al., 1995; Robinson et al., 1995; Takechi and Kawamura, 2005).

Expression of rh1-2 was consistent with previous RT-PCR results that showed expression in 21 dpf juvenile fish and the adult retina but at significantly lower levels than rh1 (Morrow et al., 2011). Furthermore, all expression of both rh1 and rh1-2 was confined to the outer nuclear layer (ONL), consisting of the cell bodies of rod and cone photoreceptors, suggesting that rh1-2 protein expression occurs in photoreceptors as opposed to other retinal cells. Another interesting feature of rh1-2 expression is that it often overlaps rh1 expression, which suggests the possibility of co-expression of the genes in the same photoreceptor. Because of this overlap and as the nucleotide sequences of rh1 and rh1-2 share approximately 75% similarity, a sense-strand probe control experiment was run to exclude the possibility of cross-hybridization. The same 700 bp rh1 probes were used to stain 3 dpf embryos both with and without the addition of full-length rh1-2 blocking RNA, present at double the concentration of the rh1 probe. The presence of the rh1-2 blocking RNA did not have a significant effect on rh1 staining, which suggests that there is likely no cross-hybridization between rh1/rh1-2 probes and the opposite transcripts (Fig. S1).

Full-length gene sequences coding for zebrafish rhodopsin, Rh1-2 and exo-rhodopsin were inserted into the p1D4-hrGFP II expression vector (Morrow and Chang, 2010) and transiently transfected into HEK293T cells. All three pigments successfully bound 11-cis retinal, producing dark spectra with λ_max values of 500.6±0.4, 495.7±0.3 and 496.8±0.5 nm for rhodopsin (N=3), Rh1-2 (N=3) and exo-rhodopsin (N=3), respectively (Fig. 2). These values were consistent over three separate expressions and with previous in vitro expression studies (Chinen et al., 2003; Morrow et al., 2011; Tarttelin et al., 2011). This shows that the λ_max of Rh1-2 is more similar to that of exo-rhodopsin than to rhodopsin, but all three are around 500 nm, characteristic of most rhodopsins and many non-visual opsins (Kojima et al., 2000; Bowmaker, 2008).

We also measured the rate of release of all-trans retinal that occurs after photoactivation, requiring both hydrolysis of the Schiff base linkage between opsin and retinal and dissociation of retinal from opsin. Using fluorescence spectroscopy, we measured the retinal release half-life of zebrafish rhodopsin as 6.5±0.3 min (N=6; Fig. 3), which is comparable to our previous results (Morrow and Chang, 2015). Despite much lower expression levels, Rh1-2 had a very similar retinal release half-life of 7.6±0.8 min (N=3; Fig. 3). Conversely, the non-visual exo-rhodopsin released retinal approximately 5 times faster than rhodopsin, with a half-life of 1.6±0.3 min (N=5; Fig. 3). This is the first time that retinal release has been measured in a non-visual opsin. These results show that the kinetics of photoactivation in Rh1-2 are more similar to those of rhodopsin than to exo-rhodopsin, despite the fact that in vitro expression levels suggest that Rh1-2 may be less stable than both rhodopsin and exo-rhodopsin.

Prior analyses lacked the taxonomic sampling to resolve the origins of rh1-2, although there were hints that it might be an ancient gene duplication in teleost fish (Morrow et al., 2011). Here, we amplified additional rh1 and rh1-2 sequences in order to better resolve the evolutionary history of rh1-2, and its relationship to the rh1 and exorh genes. Both maximum-likelihood and Bayesian phylogenetic analyses recovered with high bootstrap and posterior probability support, respectively, a single clade of rh1-2 that was most closely related to anchovies and herring (Clupeomorpha) and ostariophysian fishes (Fig. 4; Fig. S2). Interestingly, the rh1-2 clade was not most closely related to other duplicated ray-finned fish rh1 genes, such as exorh, eel deep-sea rhodopsin (dso) and freshwater rhodopsin (fwo), or the pearleye rh1a and rh1b. The resulting topology largely recovered expected species relationships, particularly for major lineages, including the placement of lampreys and ray-finned fish (Hurley et al., 2007; Nakatani et al., 2011; Near et al., 2012). Interestingly, the two clades of eel rh1 paralogs did not resolve into a monophyletic group, although this may have been due to unusual sequence evolution in nearby clades, which may be resolved with additional sequence data. Together, this suggests that the duplication that led to rh1-2 occurred in the ancestor of a major group of bony fishes including anchovies, herrings and ostariophysian fishes (Ostarioclupeomorpha). Previous phylogenetic analysis of rh1-2 found, with weak support, that rh1-2 was sister to ostariophysians plus acanthomorphs (Morrow et al., 2011). This discrepancy is likely due to the increased taxon sampling in the current study. These results further suggest that additional copies of rh1-2 have yet to be identified from several groups of ray-finned fish. Additional sequences that become available as emerging genome projects of ostariophysian and clupeomorph fishes are annotated may help to further resolve the placement of this clade.

Molecular evolutionary analyses were used to determine what changes in selective constraint occurred during and after the duplication that led to rh1-2. Random sites models as implemented in PAML revealed that vertebrate rhodopsins as a whole were under strong selective constraint (average ω=0.07, M0) (Table S2) with no evidence of positive selection (M2a versus M1a; M8 versus M8a, P>>0.5 in all cases) (Table S2). Significant among-site rate variation was found, as would be expected for functional protein coding genes (M3 versus M0, P<0.00) (Table S2). This was also true when only ray-finned fish were considered (ω=0.08, M0; M2a versus M1a; M8 versus M8a, P>>0.5 in all cases; M3 versus M0, P<0.00) (Table S2) and when only rh1-2 was considered (ω=0.09, M0; M2a versus M1a; M8 versus M8a, P>>0.5 in all cases; M3 versus M0, P<0.00) (Table S2).

To further test for differences in selective constraint between rh1-2 and other rh1 genes, we used the branch-site, branch and clade models on the dataset pruned to contain only ray-finned fish rhodopsin sequences. Using the branch-site model, we observed increased positive selection on the branch leading to the rh1-2 clade (Table S3), which approached significance when compared with the null model (P=0.071), and identified a number of BEB sites with posterior probabilities above 0.8. Conversely, branch-site and CmC analyses with the rh1-2 clade placed in the foreground found no evidence of either positive or divergent selection, supporting earlier random sites model results, which found that the rh1-2 clade is under similar selective pressure to other rh1 genes. Finally, the branch model was used to test for overall changes in selective constraint both on the branch leading to rh1-2 and on the entire clade. We found that the ω along the branch and for the whole clade did not significantly differ from background ω (P>>0.5) (Tables S3 and S4). This supports the hypothesis that rh1-2 is a functional gene as it has been maintained under high levels of negative selection, corroborated by the M3 results, which show significant rate variation. Together, these results suggest that rh1-2 may have experienced a burst of positive selection following duplication and divergence from rh1, and was later subject to purifying selection, which highlights an evolutionary path typical of genes that survive duplication and divergence events (Lynch and Conery, 2000).

Using in situ hybridization, along with both absorbance and fluorescence spectroscopy, we have shown that zebrafish rh1-2 is expressed in peripheral photoreceptors of the retina and codes for a functional opsin protein that releases retinal at a similar rate to rhodopsin following photoactivation. However, rh1-2 expression only starts around 5 dpf and is weak in vitro, both traits that are uncharacteristic of traditional rh1 genes. Despite low expression levels, Rh1-2 has a similar half-life of retinal release to rhodopsin, almost 5 times greater than that of the non-visual exo-rhodopsin, suggesting the potential for a role in vision as a result of functional similarities to rhodopsin. Phylogenetic analyses place the rh1-2 clade as a sister group to ostariophysian and clupeomorph rh1, suggesting it originated following a fairly ancient duplication event, independent of other rh1 duplications previously characterized in teleost fish. Below, we discuss potential functional roles for rh1-2 considering our findings, and the implications of this opsin gene for the zebrafish visual system.

The retinas of teleosts experience persistent neurogenesis in postembryonic fish, with new neurons arising from two different populations of stem cells. The first are multipotent stem cells residing in the ciliary, or circumferential, marginal zone (CMZ), where progenitor cells give rise to Müller glia and all retinal neurons, except for rod photoreceptors, proliferate outward from the peripheral retina (Johns, 1977; Hitchcock et al., 2004; Stenkamp, 2007). The second are Müller glia cells, which give rise to rod precursor cells in the inner nuclear layer; once these rod precursors reach the ONL, they divide and differentiate as rod photoreceptors (Bernardos et al., 2007; Nelson et al., 2008). Our in situ results show that expression of rh1-2 is consistently located in the peripheral ONL of the retina, near the CMZ, but does not persist as cells are repositioned to a more central location in juvenile and adult fish. This suggests that rh1-2 expression is limited to either rod photoreceptors derived from peripheral Müller glia cells that have recently emerged from the CMZ, or to cone photoreceptors recently derived from retinal progenitors of the CMZ that are still located in the periphery. Either possibility is consistent with the idea that rh1-2 is more similar to visual opsins and expressed in retinal photoreceptors, as opposed to a non-visual opsin that can be expressed in other retinal cells, including horizontal cells (Cheng et al., 2009) and retinal ganglion cells (Dacey et al., 2005; Panda et al., 2005), as well as in neural tissues outside of the retina, including the pineal gland (Mano et al., 1999), cerebellum (Blackshaw and Snyder, 1999) and spinal cord (Tartellin et al., 2003).

In order to function as a visual rhodopsin, rh1-2 would likely have to be expressed in rods; however, expression in cones is also possible as there are rare cases where rhodopsins and cone opsins are expressed in the opposing photoreceptor type in reptiles and amphibians (Kojima et al., 1992; McDevitt et al., 1993; Schott et al., 2016). While a ∼496 nm peak corresponding to Rh1-2 was not detected in previous microspectrophotometry studies, which could have helped to localize expression at the cellular level (Nawrocki et al., 1985; Robinson et al., 1993; Cameron, 2002), this is likely the result of either a lack of sensitivity to detect the poorly expressing opsin or confounding signals with other opsins that have similar λ_max values, such as rhodopsin, Rh2-3 or Rh2-4 (Chinen et al., 2003). Interestingly, the expression pattern of rh1-2 most resembles that of rh2-3 and rh2-4, which is confined to peripheral photoreceptors near the CMZ in embryos and juveniles (Takechi and Kawamura, 2005). In adult fish, however, rh2-3 and rh2-4 expression expands to additional portions of the peripheral retina (Takechi and Kawamura, 2005), while the pattern of rh1-2 stays mostly the same. It is possible that even limited rh1-2 expression in the ventral peripheral retina, which produces a slightly blue-shifted pigment compared with rhodopsin, may be advantageous for detecting the spectrum of down-welling light, a phenomenon that has been noted in cone opsin duplicates (Temple, 2011). Overall, the fact that rh1-2 has both a different and more restrictive expression pattern than rh1 in the retina is not unusual, as a diversity of expression patterns seem to be a common feature of duplicated opsin genes in teleost fish (Hofmann and Carleton, 2009).

Along with expression in the ONL, there are several functional features of the Rh1-2 protein that suggest it is likely to be a visual opsin gene. We previously reported that Rh1-2 successfully bound 11-cis retinal to form a functional pigment that activates in response to light, with a λ_max of approximately 500 nm (Morrow et al., 2011). Here, we present a more precise λ_max estimation of 495.7±0.3 nm, calculated via curve fitting to an A1 visual pigment template (Govardovskii et al., 2000). While this is ∼5 nm blue-shifted compared with rhodopsin, and more similar to exo-rhodopsin, all three pigments have λ_max values within the typical range of most vertebrate rhodopsins (Bowmaker, 2008). The ∼4 nm blue shift of exo-rhodopsin relative to rhodopsin was suggested to be due to A124 (Tarttelin et al., 2011), as A124G substitutions in some deep-sea fish rhodopsins were associated with red-shifts of up to 3 nm (Hunt et al., 2001). Rh1-2 also has A124, suggesting the potential for a similar spectral tuning mechanism to that of exo-rhodopsin. However, the G124A substitution in zebrafish rhodopsin was recently shown to have no significant effect on spectral tuning (Morrow and Chang, 2015). While identical substitutions in different rhodopsin sequences can lead to variable shifts in spectral sensitivity, it is also possible that the blue shift of Rh1-2 relative to rhodopsin is due to an accumulation of minor substitutions throughout its sequence, as no other differences were identified at sites known to be involved in rhodopsin spectral tuning (Yokoyama, 2000; Hunt et al., 2001).

Following photoactivation, Rh1-2 releases retinal at a rate similar to rhodopsin, suggesting that key structural aspects of rhodopsins, such as Schiff base stability and the hydrogen bond network of the chromophore binding pocket, are likely maintained in Rh1-2 (Janz and Farrens, 2004). This point is reflected in the significantly faster retinal release of cone visual pigments (Chen et al., 2012) and in our measurements of the non-visual exo-rhodopsin, which releases retinal almost 5 times faster following photoactivation. Because retinal release is a step in the retinoid cycle (Kiser et al., 2012), the metabolic cycle responsible for providing new 11-cis retinal molecules to regenerate free opsin, the slower retinal release of Rh1-2 also suggests that it may only have access to the retinal pigment epithelium-mediated pathway of retinal regeneration (Lamb and Pugh, 2004), and not the Müller cell-mediated pathway upon which cone opsins rely to help maintain rapid response kinetics (Wang et al., 2009; Wang and Kefalov, 2011). These results support the hypothesis that Rh1-2 is a visual opsin with rhodopsin-like functional properties.

An alternative hypothesis concerning the role of rh1-2 is that it is a gene duplicate experiencing low levels of expression that has no unique functional role in photoreception. Where traditional rhodopsin genes maintain a high level of expression in the retina, the duplication event that gave rise to rh1-2 failed to transfer the same regulatory elements that drive rh1 expression (Kennedy et al., 2001), resulting in much lower expression levels. Additionally, in vitro expression of Rh1-2 suggests it is considerably less stable than rhodopsin. Despite having some characteristics of a redundant gene duplicate, this classification is usually attributed to genes resulting from relatively recent gene duplication events, with the vast majority of gene duplicates being silenced within a few million years (Lynch et al., 2001). However, unlike a more recent rhodopsin gene duplication, which may generate species-specific duplicates (Lim et al., 1997), our analyses suggest a much more ancient origin for the rh1-2 gene, within the rh1 gene family of a major group of bony fishes, which would place the duplication leading to the birth of rh1-2 somewhere between 153 and 248 million years ago (Nakatani et al., 2011; Chen et al., 2013). PAML analyses also suggest the potential for increased evolutionary rates at a variety of sites along the branch leading to the rh1-2 clade, followed by strong selective constraint characteristic of rh1 genes within the rh1-2 clade. This pattern of evolutionary rates is typical of genes that survive duplication and divergence events (Lynch and Conery, 2000), including retrogenes (Gayral et al., 2007). This pattern will likely become more significant when additional rh1-2 sequences are discovered and included in future analyses.

Aside from rh1-2, only two other rh1 gene duplications are known to have been retained in Actinopterigian fishes, making rh1, along with sws1, the least common opsin gene to experience duplication. The first example is from eels, which express the rh1fwo gene with the 11-cis 3,4-dehydroretinal (A2) chromophore in the early stages of life, where a red-shifted rhodopsin is thought to provide an advantage in the more long wavelength-shifted spectral environment of freshwater (Bridges, 1972; Loew, 1995). During maturation, eels migrate to a marine environment, with a more restricted and blue-shifted light spectrum, coupled with expression of a blue-shifted rh1dso gene, regenerated with 11-cis retinal (A1) chromophore (Hope et al., 1998; Zhang et al., 2000). This switch of both opsin and chromophore is a clear example of an adjustment of the visual system as a result of a change in photic environment. Another example is the deep-sea pearleye, S. analis, which has a more traditional rh1A gene, along with rh1B, expressed alongside rh1A in adult fish living over 900 m below the surface (Pointer et al., 2007). The pearleye has unique cylindrical eye morphology, containing both a main retina, used for image formation, and an accessory retina, likely only capable of gross light perception (Collin et al., 1998), with rh1B expression being localized in this accessory retina (Pointer et al., 2007). Zebrafish do not experience an ontogenetic migration, possess only A1 chromophore-based visual pigments (Allison et al., 2004), do not occupy deep-sea habitats and do not have abnormal eye or retinal morphology, suggesting that it is unlikely that rh1-2 serves a similar function to the duplicated rhodopsin genes in either eels or the pearleye. This is supported by our phylogenetic analyses, which found rh1fwo, rh1dso, rh1B and rh1-2 all form distinct lineages that resulted from separate duplication and divergence events.

Perhaps the most intriguing result highlighted in this study is the partially overlapping expression patterns of rh1 and rh1-2, suggesting the possibility that the two genes may be co-expressed in a subset of photoreceptors, which could influence both the structure and function of these cells. Previous studies have hypothesized that cone opsin co-expression in humans could provide a developmental advantage (Xiao and Hendrickson, 2000). Alternatively, cone opsin co-expression in the cichlid fish Metriaclima zebra is thought to contribute to spectral tuning, although the λ_max differences in these opsins is 35–48 nm (Dalton et al., 2014), far exceeding the 5 nm difference between rhodopsin and Rh1-2. However, rhodopsin also serves an important structural role in rod photoreceptors, where it is packed into the outer segments and forms an array of dimers (Fotiadis et al., 2003). This arrangement could help to maximize the capacity of the rod outer segments, but likely also serves a functional purpose, with higher order rhodopsin oligomers being a more active species than monomers (Fotiadis et al., 2006). Considering the relatively low stability of Rh1-2 compared with rhodopsin, the incorporation of the former into a rhodopsin oligomer array could influence the structure of peripheral rod photoreceptors.

Co-expression of multiple rh1 genes also raises the possibility of the formation of heterodimers, common in some GPCRs where it allows for differential binding between ligand and G protein (Waldhoer et al., 2005; Monnier et al., 2011). In fact, a functional dichotomy where one monomer responds to stimuli and the other binds the G protein was predicted for rhodopsin by molecular dynamics simulations (Neri et al., 2010), while alternative conformations for each monomer following activation could promote distinct functions from otherwise identical subunits (Jastrzebska et al., 2013). The presence of a rhodopsin/Rh1-2 heterodimer would likely influence the properties of a rod photoreceptor; however, further study is required to investigate this possibility. These studies will be challenging because of the low expression levels of rh1-2 and its significant sequence similarity to the highly expressed rh1. Nonetheless, the potential for this interaction as well as the presence of the rh1-2 gene in other teleost fish should promote additional investigation into the influence of a second visual rhodopsin gene on the vertebrate visual system.

We acknowledge Dr Rosalie Crouch for generously providing 11-cis retinal, and the CSB Imaging Facility at the University of Toronto.