The visual pigment xenopsin is widespread in protostome eyes and impacts the view on eye evolution

Photoreceptor cells in the eyes of Bilateria are often classified into microvillar cells with rhabdomeric opsin and ciliary cells with ciliary opsin, each type having specialized molecular components and physiology. First data on the recently discovered xenopsin point towards a more complex situation in protostomes. In this study, we provide clear evidence that xenopsin enters cilia in the eye of the larval bryozoan Tricellaria inopinata and triggers phototaxis. As reported from a mollusc, we find xenopsin coexpressed with rhabdomeric-opsin in eye photoreceptor cells bearing both microvilli and cilia in larva of the annelid Malacoceros fuliginosus. This is the first organism known to have both xenopsin and ciliary opsin, showing that these opsins are not necessarily mutually exclusive. Compiling existing data, we propose that xenopsin may play an important role in many protostome eyes and provides new insights into the function, evolution, and possible plasticity of animal eye photoreceptor cells.


Introduction
The photoreceptor cells (PRCs) in animal eyes are often classified according to their structure, that is depending on whether the sensory surface is enlarged by microvilli or by cilia (Eakin, 1979;Eakin, 1963;Eakin, 1968). The first type of PRCs in many protostomes was shown to depolarize in response to light and to employ rhabdomeric opsin (r-opsin) as a visual pigment, which signals via the Ga q mediated IP 3 cascade opening TRP ion channels in the PRC membrane (Fain et al., 2010;Shichida and Matsuyama, 2009). In contrast, ciliary PRCs of vertebrate eyes are known to signal via the Ga i/t mediated cGMP cascade closing CNG channels and leading to a hyperpolarization. Since both are found in protostome and deuterostome animals and due to their distinct molecular signatures, it is assumed that these two kinds of PRCs were already present in the last common ancestor of bilaterian animals (Arendt, 2008;Arendt et al., 2004;Arendt et al., 2002;Gomez et al., 2009;Gehring, 2014;Nasi and Gomez, 2009;Panda et al., 2002). This classification of PRCs became a sound basis for comparative eye research from sensory biology to molecular physiological, developmental, and evolutionary biology. We present data suggesting that in protostomes, an additional second kind of ciliary PRCs is widespread and that this may be evolutionarily closer to microvillar PRCs than to vertebrate ciliary eye PRCs.
Recently, a new type of visual opsins, xenopsin, has been characterized. It shares important functional sequence motifs with ciliary opsins (c-opsins) and has been shown to signal most likely also via Ga I in a flatworm (Rawlinson et al., 2019). Nonetheless, xenopsins and c-opsins do not group in phylogenetic analyses (Ramirez et al., 2016;Rawlinson et al., 2019;Vö cking et al., 2017) indicating a distinct evolutionary origin. Surprisingly, xenopsins and c-opsins are mutually exclusively distributed across the animal kingdom, which is difficult to explain from a genomic perspective and seemingly doubts the phylogenetic analyses. In this study, we report the first organism having both xenopsin and c-opsin. In congruence with thorough phylogenetic and gene structure analyses, this provides further support for a distinct evolutionary origin of these visual pigments.
Despite increasing knowledge on the presence of xenopsin in many animal groups, only very few data on cellular expression and function of xenopsin exist. So far it turned out to be this new opsin type and not c-opsin that is present in ciliary eye PRCs of larval brachiopods (Passamaneck et al., 2011;Vö cking et al., 2017) and in larval ciliary eye PRCs and adult extraocular ciliary PRCs in a flatworm (Rawlinson et al., 2019). Furthermore, xenopsin has been found coexpressed with r-opsin in eye PRCs exhibiting both microvilli and cilia in the larva of a mollusc (Vö cking et al., 2017), thereby raising the question, whether protostome eye PRCs had the potential to change between microvillar and ciliary organization during evolution.
To obtain a broader overview of the role of xenopsin in animal eyes, we investigated larva of the annelid Malacoceros fuliginosus (Claparède, 1868), and the bryozoan Tricellaria inopinata d 'Hondt & Occhipinti Ambrogi, 1985 in which RNA-seq data pointed to the presence of xenopsin. We find it expressed in ciliary eye PRCs of the bryozoan larva, and we present unambiguous evidence that xenopsin enters the cilia and likely triggers the phototactic response of the larva. Further, we find xenopsin coexpressed with r-opsin in eye PRCs of the annelid larva similar to the earlier finding in a larval chiton (Vö cking et al., 2017). We propose that (1) Xenopsin is an important visual pigment in protostomes, (2) ciliary eye PRCs may not be of the same evolutionary origin in protostomes and deuterostomes, and (3) ciliary and microvillar eye PRCs may be evolutionarily linked in protostomes. The findings impact the current understanding of how animal eyes evolved and diversified and provides insights on the plasticity that cell types can exhibit in the course of evolution.

Molecular phylogeny of animal xenopsins and c-opsins
We screened Tricellaria inopinata assembly one for opsins by blasting with a broad set of metazoan opsin sequences as query and successive reciprocal blast against Genbank. The sequences were further checked for the presence of the PFAM 7tm_1 domain and the residue Lys296, which is predictive for chromophore binding in opsins and for the NPXXY motif at positions 302-306 ( Figure 1) contributing to signal transduction in G protein-coupled receptors. We blasted the hits against assembly two and elongated the sequences if longer hits were retrieved. We screened the transcriptomic resources of Malacoceros fuliginosus in the same manner, but only for the presence of xenopsins and c-opsins. We retrieved five hits from the T. inopinata assembly, which all gave xenopsins as first hits by reciprocal blast. Since we had evidence for contamination of the T. inopinata assembly (see Materials and methods), we cloned all sequences and tested them by ISH for expression in T. inopinata larva. Only one sequence gave positive signals and was further used in this study, while the others were no longer considered as they might be from other bryozoan species. Three sequences were retrieved from the M. fuliginosus assembly. After reciprocal blast against Genbank, one sequence gave c-opsins as first hits and the other two xenopsins. For further analyses, we kept the potential c-opsin and that potential xenopsin, for which we obtained positive results after in situ hybridization in larvae.
We added the sequences and few recently described xenopsin sequences from the molluscs Sepia officinalis and Ambigolimax valentianus, the bryozoan Bugula neritina, the flatworm Maritigrella crozieri, and the chaetognath Pterosagitta draco to the opsin sequence set (https://doi.org/ 10.7554/eLife.23435.009) analyzed by Vö cking et al., 2017 and ran maximum likelihood and Bayesian phylogenetic analyses to study opsin molecular evolution with a focus on the relationships of xenopsins and c-opsins. All major opsin groups described by Ramirez et al., 2016, Vö cking et al., 2017, and Rawlinson et al., 2019 xenopsins were recovered with high support values (Figure 2, Figure 2-figure supplements 1 and 2). One sequence from M. fuliginosus falls into c-opsins, while another one falls into xenopsins. The opsin of T. inopinata likewise falls into xenopsin and groups with the sequence of the bryozoan Bugula neritina. The topology within the xenopsin clade suggests an early divergence of xenopsin in two clades xenopsin A and xenopsin B containing opsin from several animal groups similar as described by Vö cking et al., 2017 and Rawlinson et al., 2019. Yet, the support values for the two subclades are not as high as for xenopsisn as a whole and other large opsin groups. We tested robustness of the split into xenopsin A and B against changes in the outgroup by calculating trees of xenopsins only (unrooted) and trees with few cililary opsins, few cnidops and few c-opsins and cnidpos as outgroup. The split is retained in all cases with the exception of an outgroup composed of cnidops and c-opsins, where xenopsin B is a paraphyletic assemblage (Figure 2-figure supplements 5-8). The position of one brachiopod sequence (Lingula anatina melanopsin like XP 013397676.1) is not stable, in some cases it falls into xenopsin A, in others it groups with xenopsin B sequences or has a basal position. Accordingly, our data suggest an early diversification of xenopsins, but with moderate support only. Since M. fuliginosus xenopsin groups in all trees with xenopsin B representatives, we regard it as likely being the first known annelid xenopsin B. Several flatworm xenopsin B sequences stand out by strong modifications in the NPXXY and tripeptide motif (Figure 1-figure supplement 1) questioning the capability of the opsins to induce G-protein based light transduction. In difference, these motifs are conserved in the xenopsin of M. fuliginosus.
Gene structure analysis corroborates molecular phylogeny Several sequences (for example from Idiosepius paradoxus and Terebratalia transversa), which in Ramirez et al., 2016, Vö cking et al., 2017, Rawlinson et al., 2019 and this study group within xenopsins were earlier classified as c-opsins (Passamaneck et al., 2011;Yoshida et al., 2015). This view  was either based on automated gene annotation, similarity searches, or phylogenetic analyses with only low taxon sampling. Nonetheless, it is in congruence with the presence of the NKQ tripeptide pattern (Figure 1) in the fourth cytoplasmic loop, which is in c-opsins crucial for specific binding to Ga i/t (Marin et al., 2000). To test if the grouping of the new opsin sequences found in T. inopinata and M. fuliginosus may result from tree inference artifacts -we cloned the respective genes from genomic DNA, analyzed gene structure and mapped it together with gene structure data generated by Notably, the closest related clade is neither a deuterostome specific nor a protostome specific opsin group, but cnidops. Yet, validation of this sister group relationship by gene structure data is not possible, since cnidops are lacking introns.

Xenopsin is expressed in cilia of the eye photoreceptor cells in larval T. inopinata
Larvae of T. inopinata possess one median eye apical of the anterior vibratile plume and one pair of lateral eyes halfway down from the apical to the abapical pole ( Figure 3A). All eyes can be easily identified in live animals due to their red pigmentation. EM sections show that all three eyes form epidermal invaginations ( Figure 4A, Figure 4-figure supplements 1A and 2), and whole-mount in situ hybridization revealed that Tin-xenopsin is strongly expressed in the region of all three eyes ( Figure 3B,C). Besides, we found weak expression of Tin-xenopsin in few other cells, which are not associated with shielding pigments. One pair of cells lies on the rim of the anterior ciliary groove. Another pair lies lateral to the axial nerve running down from the apical organ ( Figure 3B  Source data 1. Accession numbers of the genes used for gene tree inference.   Gene structures of all sequences, which were used for gene tree calculation and for which genomic information was available or generated in this study, mapped on the un-curated protein sequence alignment. Figure supplement 4. Intron phase and position of all sequences, which were used for gene tree calculation and for which genomic information was available or generated in this study, mapped on the un-curated protein sequence.    hybridization signal was much stronger in the eye regions than in the extraocular cells. Custom made antibodies against Tin-xenopsin specifically stain the eye regions ( Figure 3D), but no significant staining appeared in the extraocular Tin-xenopsin expressing or any other cells.
To get insights into the fine structure of the eyes, we performed serial section electron microscopy. The invagination of the lateral eyes is 5 mm deep, and it is formed by two neighboring coronal epidermal cells (PCC1 and PCC2 in Figure 4 and extending into the eye invagination. In the right eye of the cryo-fixed specimen, we counted 170 cilia. The axonemal microtubules of the cilia are arranged in 9 Â 2+two pattern ( Figure 4B). The cilia of the PRC penetrate the cuticle (cu in Figure 4A       . It sends a slender dendrite running upwards on the abapical side of the eye PRC and forms an anteriorly projecting pillar-like elevation emerging from the abapical wall of the eye invagination ( Figure 4-figure supplement 2). On top of the elevation, 15 cilia with a 9 Â 2 +two axoneme emerge from the tip of the dendrite and penetrate the cuticle.
The invagination of the median eye is formed by two coronal cells with shielding pigment granules and two PRCs (PRC1, PRC2 in Figure 4-figure supplement 1A). Subcellular characteristics of the coronal cells and PRCs are similar to those of the lateral eyes, but the cellular arrangement is different. The coronal cells line the bottom as well as the apical and abapical walls of the invagination, while the two photoreceptor cells line the left and the right wall. The perikarya of the PRCs lie distant to each other on the left and the right from the invagination and the ciliary bundles of the PRCs project from both sides into the eye invagination.
Knowing the ultrastructure of the eyes makes it possible to localize Tin-xenopsin mRNA and protein on the subcellular level by combining fluorescence in situ hybridization (FISH) with immunohistochemistry. In the lateral eye, the Tin-xenopsin FISH signal surrounds a nucleus next to the base of the eye invagination ( Figure 3E). It matches well the position of the eye PRC nucleus in the EM dataset. The anti-Tin-xenopsin antibody signal is directly adjacent to the FISH signal and co-localizes with the cilia labeled by anti-acetylated a-tubulin in the eye invagination ( Figure 4C-F). The median eye shows a similar pattern. While the FISH signal stains one cell on each side of the invagination, the opsin antibody stains the cilia inside the eye invagination ( Figure 3F; Figure 4-figure supplement 1B-E). Accordingly, Tin-xenopsin mRNA is located throughout the soma of the eye photoreceptor cells of T. inopinata, whereas the opsin protein resides in the ciliary bundles emerging from these cells.
Tin-xenopsin likely is most sensitive in blue light and triggers the phototactic response of the larva Since we could not detect the expression of any other opsin than Tin-xenopsin in the eyes of T. inopinata, we were interested in behavioral responses, which depend on directional detection of light by the eyes for the first functional characterization of this new opsin. We assayed the phototactic displacement of freshly hatched larvae under different wavelengths. The animals showed the biggest displacement towards blue light (454 nm) but still showed displacement towards green (513 nm To assess the presence of ciliary structures in the ventral and dorsal eye and to achieve quantitative data on the surface extension of microvillar and ciliary structures, we analyzed a 3D electron microscopic data set of ventral and dorsal eyes in a 72 hpf stage larva in detail. In the lateral and the medial cells of the ventral eyes, only a basal body with an accessory centriole underneath the apical cell membrane (in the medial cell we could see an accessory centriole only on the right body side), but no cilia are present, which gives rise to the microvillar brushes ( Figure 6-figure supplement 1). In contrast, the middle cell bears a long cilium projecting together with the microvillar brushes into the eye cavity ( Figure 6G-K). This cilium is also visible in light microscopic stainings ( Figure 6L). We estimated the microvillar surface of the middle PRC as 296 mm 2 based on the average diameter of the microvilli, the number of microvilli per area, and the total volume of the space filled by the microvilli assessed from the 3D image stack. The ciliary surface is 10.7 mm 2 based on the length and the diameter of the cilium. Accordingly, the ratio of ciliary to microvillar membrane surface is 1:27.7. The PRC of the dorsal eye likewise possesses a long cilium projecting together with the PRC microvilli into the eye cavity ( Figure 6M-N).

M. fuliginosus is the first organism known to have both xenopsin and c-opsin
Based on phylogenies with broad taxon sampling across the animal kingdom, Ramirez et al., 2016 and Vö cking et al., 2017 reported a secondary loss of xenopsins, as well as c-opsins in several major animal groups. Notably, xenopsins and c-opsins were not known to occur together. Annelids are the only group in which both opsin types were found, while other spiralians have only xenopsin and arthropods and deuterostomes have only c-opsins (Ramirez et al., 2016;Rawlinson et al., 2019;Vö cking et al., 2017). But even within annelids, mutually exclusive distribution of these opsins was reported. Xenopsin was only found in the basally branching oweniids (Vö cking et al., 2017), whereas c-opsins were only found in Platynereis dumerilii (Arendt et al., 2004) and sabellids (Bok et al., 2017) and genomic loss of both opsins are evident for Capitella teleta and Helobdella   robusta (Vö cking et al., 2017). To our knowledge, not a single organism has been hitherto reported to have both a c-opsin and a xenopsin. No evolutionary or functional explanation has been given so far as to why xenopsins and c-opsins do not co-occur in any of the animals screened. Accordingly, some uncertainty remained, whether the distinction between xenopsins and c-opsins is a mere tree inference artefact. We now provide evidence for an independent origin of these opsins based on thorough molecular phylogeny, the exon-intron structure of the genes and a clear case of co-occurrence in a single species, M. fuliginosus.

Xenopsin is employed in the eyes of several protostomes
In vertebrates, c-opsins constitute the visual pigments of the retinal rods and cones in the eyes and serve additional functions when expressed in the pineal or deep brain PRCs (Blackshaw and Snyder, 1999;Kawano-Yamashita et al., 2014;Hankins et al., 2014). In protostomes, c-opsins were only reported from annelids and arthropods (Bok et al., 2017;Cronin and Porter, 2014;Hering and Mayer, 2014;Ramirez et al., 2016;Vö cking et al., 2017). Moreover, the expression of c-opsin has not been reported from PRCs in cerebral eyes, but from extraocular brain photoreceptors and in the case of the annelid subgroup of Sabellida in tentacular crown eyes (Arendt et al., 2004;Beckmann et al., 2015;Bok et al., 2017;Velarde et al., 2005;Verasztó et al., 2018). Instead, in many protostomes, r-opsins sense light in microvillar eye photoreceptor cells (Fain et al., 2010;Ramirez et al., 2016;Terakita, 2005). Besides, the evidence is increasing that xenopsins play important roles in protostome eyes. So far, xenopsin expression has been reported in the eye PRCs and serially homologous extraocular photoreceptor cells in chiton larva (Vö cking et al., 2017), in the eyes of larval brachiopods (Passamaneck et al., 2011) and recently in flatworm brain PRCs (Rawlinson et al., 2019). In this study, we report expression in the eyes of a larval annelid and eyes of a larval bryozoan. Seemingly, xenopsin is more common in the eyes of those protostomesthan hithero anticipated.

Xenopsin enters cilia
Subcellular targeting of opsins is an important prerequisite for visual perception in PRCs. Being transmembrane proteins, opsins travel integrated into vesicle membranes from the Golgi to the plasma membrane. Once there, they can enter plane plasma membrane areas similar to melanopsin in intrinsic light-sensitive retinal ganglion cells in the vertebrate retina (Belenky et al., 2003) and like many of the vertebrate non-visual c-opsins expressed in deep brain receptors, inner layers of the retina and in several other tissues (Foster and Bellingham, 2004;Hunt et al., 2014). Access to specialized membrane extensions like cilia or microvilli depends on specific active transport mechanisms (Schopf and Huber, 2017;Wang and Deretic, 2014;Wingfield et al., 2018). Though the evolution of sequence motifs relevant for protein binding to the respective transport machinery is not well understood, the capability of certain opsin types to enter either cilia or microvilli is seemingly very well conserved. To our knowledge, no c-opsin entering microvilli and no r-opsin entering cilia are known. We provide unambiguous evidence that xenopsin enters cilia in T. inopinata. Likely, this is also the case in the eye photoreceptor cells of the larval brachiopod Terebratalia, where xenopsin expressing cells show a ciliary organization (Passamaneck et al., 2011) and in brain PRCs in the flatworm Maritigrella (Rawlinson et al., 2019). The subcellular localization of xenopsin in PRCs of larval Leptochiton asellus (Vö cking et al., 2017) and larval M. fuliginosus is not clear since custom-made antibodies against the opsin did not reveal positive staining. However, in Leptochiton asellus, the presence of prominent ciliary structures beside microvilli expressing r-opsin provides the structural prerequisites for a similar opsin targeting. Similarly, in M. fuliginosus the dorsal eye PRC and the second ventral eye PRC bear a prominent cilium in addition to microvilli. Whether the xenopsin expressed enters these cilia remains speculative. Only for xenopsin A exist data on the subcellular localization of the protein (always in cilia), but so far not for xenopsin B. The first and third PRC of the ventral eye bear in addition to microvilli only basal bodies and accessory centriols close to the apical surface, which may be remnants of cilia. The xenopsin expressed in these cells may not enter any surface extensions, enter microvilli or may even not be translated into protein. If it would enter microvilli in certain PRCs of M. fuliginosus, xenopsin would be the first opsin group known to have the capability to enter both cilia and microvilli.

Evolution of bilaterian eye PRCs
For a long time, hypotheses on the evolution of eye PRCs focused mainly on PRCs expressing either r-opsin or c-opsin. R-opsin expressing cells employing the same kind of phototransduction cascade and with similar electrophysiological responses are found in the eyes of protostomes and deuterostomes (Arendt et al., 2002;Gomez et al., 2009;Fain et al., 2010;Koyanagi et al., 2005;Panda et al., 2002;Shichida and Matsuyama, 2009). Accordingly and due to conserved patterns in development, the presence of these PRCs already in the eyes of the bilaterian ancestor has been suggested (Arendt, 2003;Arendt, 2008;Arendt et al., 2004;Fernald, 2006;Gehring, 2014;Lamb, 2013;Shubin et al., 2009). Though c-opsins detect light in rods and cones of the vertebrate retina, its ancestral expression is assumed in brain extraocular photoreceptors (Arendt, 2008;Arendt et al., 2004;Shubin et al., 2009). Only in such cells were c-opsins found in arthropods (Beckmann et al., 2015;Velarde et al., 2005) and annelids (Arendt et al., 2004). Similarly, many kinds of non-visual c-opsins of vertebrates like encephalopsins, TMT-opsins, or VA-opsins are found in the brain in addition to possible functions in the inner layers of the retina (Hunt et al., 2014;Pérez et al., 2019). Accordingly, the employment of c-opsin cells in the visual cells of cerebral eyes evolved later, most probably in the lineage leading to chordates (Vopalensky et al., 2012).
Morphological data on the presence of ciliary PRCs in the eyes of several less studied protostome animals like bryozoans (Reed et al., 1988;Woollacott and Eakin, 1973;Woollacott and Zimmer, 1972), gastrotrichs (Woollacott and Eakin, 1973) and nemerteans (von Dö hren and Bartolomaeus, 2018) are, however, in conflict with this scenario. Vö cking et al., 2017 proposed that in addition to rhabdomeric and c-opsin, xenopsin is a third important player in PRC and eye evolution questioning a common origin of ciliary eye PRCs in protostomes and deuterostomes. Our study is providing further support for this view.
Though in protostomes, c-opsins only exist in annelids and arthropods, ancestral employment in extraocular brain PRCs is still likely. But seemingly many kinds of protostome ciliary PRCs do not employ c-opsin. Instead, cellular expression of xenopsin is reported from ciliary PRCs in the eyes of larval brachiopods (Passamaneck et al., 2011) and bryozoans (this study) and meanwhile also from extraocular and eye ciliary PRCs in flatworms (Rawlinson et al., 2019). Further, the presence of xenopsin is also known from molluscs, annelids, chaetognaths, and rotifers (Ramirez et al., 2016;Rawlinson et al., 2019;Vö cking et al., 2017). Coexpression with r-opsin is evident in a larval chiton (Vö cking et al., 2017) and an annelid (this study) in microvillar eye PRCs, which partly also exhibit ciliary structures. Xenopsin may have been co-opted by these mainly microvillar cells (Figure 7 scenario A), but it is also conceivable that the observed cellular coexpression with r-opsin in two subgroups of lophotrochozoans points towards an evolutionary link between ciliary and microvillar PRCs. During the evolution of protostomes eyes, formerly microvillar PRCs may have changed into mixed microvillar/ciliary cells coexpressing r-opsin and xenopsin (Figure 7, scenario B). Since co-expression is evident from xenopsin A and B, this happened likely before the diversification of xenopsins in protostomes. Clear hypotheses on the ancestral targeting of xenopsin (cilia and/or microvilli) may need further investigation, but existing data so far point towards cilia as targets. Even the presence of mixed cells in the eyes of the last common ancestor of bilaterians is conceivable (Figure 7, scenario C), since opsin tree topology suggests a genomic loss of xenopsin in the deuterostome stem lineage. Within protostomes, the mixed organization was retained in some extant organisms (several molluscs, certain annelids) and transformed in other organisms into a purely ciliary or microvillar organization going along with loss or downregulation of r-opsins (bryozoans, brachiopods) or xenopsins (arthropods, certain annelids), respectively. Such a hypothesis also raises the question, whether the co-occurrence of ciliary and microvillar PRCs within the same eye as known, for example from several molluscs (Bartolomaeus, 1992;Blumer, 1998;Salvini-Plawen, 2008) or the larva of the polyclad flatworms (Eakin and Brandenburger, 1981;Rawlinson et al., 2019) may be the result of integrating or co-opting ciliary cells into microvillar eyes or caused by duplication and diversification of formerly mixed microvillar/ciliary PRCs. Interestingly, the polyclad flatworm Maritigrella crozieri has several kinds of ciliary PRCs, which are expressing xenopsin in adults outside of pigmented eyes, in larval epidermal eyes, and in larval cerebral eyes, which also contain r-opsin expressing microvillar PRCs (Rawlinson et al., 2019). The evolutionary origin of the extraocular PRCs and the epidermal eyes is unclear. Nonetheless, developmental data from Schmidtea mediterranea, indicate homology of flatworm cerebral eyes to those of other protostomes (Lapan and Reddien, 2011;Lapan and Reddien, 2012). Therefore, origination from eyes with mixed r-opsin/xenopsin+ microvillar/ciliary PRCs is conceivable.

Xenopsin function and physiology
Strong phototactic responses, as we observed in the larva of T. inopinata, depend on directional detection of light by pigmented eyes. Since we could not find any opsin other than xenopsin expressed in the eyes of T. inopinata, we suggest that the xenopsin here is responsible for light reception and triggers the phototactic behavior of the larva. Accordingly, it has a similar visual function as r-opsins have in microvillar eye PRCs of several other protostomes (Fain et al., 2010;Jékely et al., 2008;Neal et al., 2019;Randel et al., 2013). The ciliary surface of the PRC in T. inopinata is even three times larger than the microvillar surface of the middle eye PRC in M. fuliginosus. Notably, heterologous expression of Maritigrella crozieri xenopsin suggests that it acts mainly via Ga i and possibly to a lower extent also via Ga s , but not via Ga q signaling (Rawlinson et al., 2019) as r-opsins do (Fain et al., 2010;Shichida and Matsuyama, 2009). Hence, similar to the case of r-opsins and c-opsins, downstream signaling and even the cellular electrophysiological response may also be different upon the excitation of r-opsins and xenopsins.
Most PRCs express only one type of visual opsin. If PRCs employ more than one opsin, they are usually from the same subgroup of visual pigments and activate the same transduction cascade (Applebury et al., 2000;Arikawa et al., 2003;Dalton et al., 2015;Isayama et al., 2014;Figure 7. Scenarios on eye PRC evolution in Bilateria. The bilaterian ancestor had extraocular c-opsin+ ciliary PRCs. These became integrated into the eyes in the lineage leading to vertebrates and were lost in many protostomes along with secondary loss of c-opsin. Scenario A: Cerebral eyes contained microvillar r-opsin+ PRCs in the bilaterian ancestor. Xenopsin was co-opted several times independently by microvillar PRCs, and eye PRCs were several times independently transformed into or replaced by ciliary xenopsin+ PRCs. Scenario B: Ancestral r-opsin+ microvillar eye PRCs were transformed once into mixed microvillar/ciliary PRCs coexpressing r-opsin and xenopsin. In extant organisms, those were retained or changed into purely microvillar r-opsin+ or ciliary xenopsin+ PRCs going along with genomic loss or downregulation of xenopsin or r-opsin, respectively. Scenario C: Mixed ciliary/microvillar PRCs were already present in the bilaterian ancestor. Katti et al., 2010;Parry and Bowmaker, 2016;Rajkumar et al., 2010), which is suggested, ultimately, to expand the visual spectrum. The same function is assumed in the annelid P. dumerilii, where Go-opsin and r-opsin co-occur in eye PRCs (Gühmann et al., 2015). If Ga i signaling of xenopsin is conserved in protostomes, PRCs coexpressing xenopsin and r-opsin, as we observed in M. fuliginosus and are known in Leptochiton asellus (Vö cking et al., 2017), might potentially be polymodal sensory cells with complex physiology capable of integrating different stimuli by activation of different signaling cascades. The ciliary surface of the middle eye PRC of M. fuliginosus is nearly 30 times less than the microvillar surface. This may hint to a minor role of xenopsin signaling in this eye, but other parameters like the efficiency of the specific sensory transduction pathway and the protein content in the membrane certainly impact the sensitivity as well. The contribution of cilia in light detection may be higher in eyes where microvilli are accompanied by higher numbers of cilia as described in several molluscs (Blumer, 1995;Blumer, 1996;Hughes, 1970;Zhukov et al., 2006).

Conclusion
Xenopsin seems to be an important visual pigment in protostome eyes. This opsin type was overlooked for a long time, probably because most molecular data on protostome light perception are from arthropods, which secondarily lost xenopsin. M. fuliginosus is the first organism known to have xenopsin and c-opsin corroborating the distinct evolutionary origin of these opsin types inferred from phylogenetic and gene structure analysis. All other organisms studied thus far, for reasons unknown, have either c-opsin or xenopsin. Xenopsin, like c-opsin, enters cilia. In protostomes, it is employed in purely ciliary PRCs and found coexpressed with r-opsin in microvillar PRCs that also have ciliary structures. Xenopsin or xenopsin expressing cells might have been recruited several times independently in the eyes of protostomes. Alternatively, these eyes already early in evolution employed a possibly polymodal r-opsin+ and xenopsin+ microvillar/ciliary PRCs. Further studies on the employment of xenopsin in protostomes will be of high interest for a better understanding of evolution, function, and plasticity of animal photoreceptor cells and eyes. Counterparts of vertebrate c-opsin employing ciliary PRCs in protostomes probably exist only in certain annelids and arthropods, since c-opsin according to available sequence resources has been lost in all other protostome animals.

Animal culture
Adults of the polychaete Malacoceros fuliginosus (Claparède, 1868) were collected from Pointe de Mousterlin, Fouesnant, France. The animals were maintained in the lab facility in sediment containing seawater tanks at 18˚C and fed with ground fish food flakes (TetraMin granules, Tetra). Mature males and females were picked, rinsed several times with filtered seawater, and kept in separate bowls until they spawned. Staging was started from the time gametes were combined in a fresh bowl. Bowls were kept at 18˚C under 12:12 hr light-dark cycle, and water was replaced every day or every second day. Larvae were fed with the microalga Chaetoceros calcitrans from 24 hpf onwards after each water change. Colonies of the bryozoan Tricellaria inopinata (d'Hondt & Occhipinti Ambrogi, 1985) were collected in Brest,France (48˚23'38.3"N 4˚25'57.4"W). The colonies were maintained at 18˚C under 12:12 hr light-dark cycle in the lab animal facility.

RNA-seq and transcriptome assembly
For studies on M. fuliginosus, we used transcriptomic resources prepared in an earlier study (Kumar et al., 2020) from pooled larvae of several stages. For T. inopinata, we performed RNA-seq and de novo transcriptome assembly. The release of larvae was triggered by the onset of light in the tanks. Two hours later, swimming larvae were attracted by a light bulb and cryo-fixed. We extracted total RNA using the Agencourt RNAdvance Tissue Kit (Beckman Coulter, Brea, California). Library preparation and sequencing were performed by EMBL (Heidelberg, Germany) Genomics Core Facility using cation-based chemical fragmentation of RNA, Illumina (San Diego, California) Truseq RNA-Sample Preparation Kit and 1 lane of 100 bp paired-end read sequencing on Illumina HiSeq 2000. We used Cutadapt 1.2.1 (RRID:SCR_011841) for trimming and the ErrorCorrectReads tool implemented in Allpaths-LG (RRID:SCR_010742) for error correction of the raw reads and Trinity (RRID: SCR_013048) for de novo assembly. We performed two rounds of RNA-seq and assembly. For the first data set (assembly 1) the collected colonies were thoroughly cleaned. However, by microscopic inspection, we observed a minor proportion of zooids of other bryozoan species, which were dispersed across the colonies and could not be entirely removed. We assessed contamination of the respective assembly by screening for cytochrome oxidase subunit I (COI), and we found sequences indeed from four different bryozoans, but none from other animal groups. The second data set (assembly 2) showed contamination also with sequences from other taxa and was only used for corroboration and elongation of sequences retrieved from assembly 1.

Opsin gene structure analysis
For the two opsins found in M. fuliginosus and the one found in T. inopinata, the whole genes were cloned from genomic DNA for subsequent analysis of exon-intron boundaries. Genomic DNA was extracted with the Nucleospin Tissue Kit (Machery-Nagel, Dü ren, Germany) and tested for fragment length larger than 20 kb. As a starting point, gene-specific primers were designed based on the transcript sequences. For genome walking, four libraries were prepared with Universal Genome Walker Kit (Takara Bio, Mountain View, California) by enzymatic digestion and used for sequence elongation starting from exonic fragments. In parallel, long amplicons bridging smaller introns were also directly amplified from genomic DNA using LA Taq (Takara Bio), iProof (Biorad, Hercules, California, USA), and HotStarTaq Plus (Qiagen, Hilden, Germany) polymerases. Obtained amplicons up to 8 kb were cloned using pGEM-T easy Vector (Promega, Madison, Wisconsin) TOPO XL PCR cloning kit (Thermo Fisher Scientific), TopTen chemically competent cells (Thermo Fisher Scientific, Waltham, Massachusetts) and Sanger sequenced. Obtained sequences were used to design further primers for ongoing sequence elongation. Read assembly was performed with CLC Main Workstation (RRID:SCR_ 000354) 7.1. Gene structures of the opsins from M. fuliginosus and T. inopinata were determined based on the cloned genomic and the protein sequences retrieved from RNA-seq using WebScipio (Hatje et al., 2011). The obtained gene structures were together with the gene structures prepared by Vö cking et al., 2017 mapped onto the un-curated sequence alignment using Genepainter (Hammesfahr et al., 2013) to identify conservation of intron positions and phases.

Custom antibodies
Custom polyclonal antibodies were prepared and affinity-purified against peptides of Tin-xenopsin by 21 st Century Biochemicals (Marlboro, Massachusetts) and Mfu-r-opsin-3 by Eurogentec (Liè ge, Belgium). For Tin-xenopsin, the peptide sequences VKAAGKKFGGDDAASQ from the 3 rd cytoplasmic loop and ATKPAPSATQAPREKKATAL from the cytosolic tail and for Mfu-ropsin3 RHSE VPSGDGKKDTL and CKNRAIDKGGDEESDN both from the cytoplasmic tail were chosen as antigens.
To assure antigen specificity, all peptide sequences were blasted against the T. inopinata, and M. fuliginosus transcriptome and gave only the respective opsin sequences as hits. The antibodies raised against the peptides ATKPAPSATQAPREKKATAL and RHSEVPSGDGKKDTL gave the best results and were used for the stainings.

In situ hybridization
For gene cloning cDNA, was prepared from total RNA with Super Script II (Thermo Fisher Scientific), sequences of interest were PCR amplified with gene-specific primers, and amplicons were subsequently ligated into pgemT-easy vector (Thermo Fisher Scientific) and cloned into Top10 chemically competent E. coli (Thermo Fisher Scientific). Sanger sequencing was used to verify the cloned sequences before DIG-and FITC-labeled sense and antisense probes were generated with T7 and SP6-RNA Polymerases (Roche, Basel, Switzerland) or with Megascript Kit (Thermo Fisher Scientific). If needed, Smarter Race (Takara Bio) was used to elongate ends of transcript sequences. In situ hybridization experiments were performed as described previously (Vö cking et al., 2015) if formamide based hybridization buffers were used. Otherwise, we followed # (Sinigaglia et al., 2018) # for urea-based hybridization buffers. In brief, animals were fixed in 4% PFA in phosphate buffer and with Tween20 (PTW; pH 7.4) and subsequently washed and stored in methanol. For Tricellaria inopinata larvae, a 2 min prefixation with 0.3% Glutaraldehyde in 4% PFA was necessary. After rehydration in PTW, samples were briefly digested with Proteinase K, washed and prehybridized in hybridization with or without 5% dextran. Samples were hybridized with RNA probes for 72 hr. Tricellaria inopinata larvae required that each washing step after hybridization was extended to 30 min. Stainings were done with a combination of FastBlue (Sigma-Aldrich) and Fast Red (Roche). The significance of expression signals was evaluated by using sense probes as control experiments. All in situ hybridization experiments were performed on at least 15 specimens per gene for each sense, and anti-sense probe and the experiments were repeated at least two times.

Electron microscopy
For electron microscopic studies, two kinds of sample preparations were used. For chemical fixation larvae were relaxed for 3 min in 7% MgCl 2 and seawater mixed 1:1 and then fixed in 2.5% glutaraldehyde in PBS, postfixed in 1% Osmium tetroxide in the same buffer, en-bloc stained with reduced Osmium, dehydrated in a graded ethanol series and embedded in Epon/Araldite as described in Vö cking et al., 2015. Cryo-fixation was performed at the EM Core facility at EMBL (Heidelberg, Germany). Larvae were relaxed as described above and then high-pressure-frozen with hexadecene acting as filler in an HPM 010 from Balzers (Balzers, Liechtenstein). Freeze substitution with 2% OsO 4 and 0.1% uranyl acetate in a mixture of 95% acetone and 5% water was performed in a Leica (Wetzlar, Germany) EM AFS2 for 46 hr at À90˚C. Samples were slowly warmed to À30˚C, kept at this temperature for 6 hr, and slowly warmed to 0˚C before they were taken out from the freeze-substitution device. Samples were rinsed several times in acetone at 0˚C and at room temperature, stepwise transferred to Epon, and cured for 48 hr at 60˚C. Serial sections of 70 nm were cut with an ultra 35˚diamond knife (Diatome, Biel, Switzerland) on a UC7 ultramicrotome (Leica) and collected on Beryllium-Copper slot grids (Synaptek, Reston, Virginia, USA) coated with Pioloform (Ted Pella, Redding, California, USA) and counterstained with 2% uranyl acetate and lead citrate. Complete series were imaged with STEM-in-SEM as described by Kuwajima et al., 2013 at a resolution of 4 nm/ pixel with a Supra 55VP (Zeiss, Oberkochen, Germany) equipped with Atlas (Zeiss) for automated large field of view imaging. Acquired images were processed with Adobe Photoshop CC, first registered rigidly followed by affine and elastic alignment  with TrakEM2  implemented in Fiji (RRID:SCR_002285).

Behavioral assays of T. inopinata larva
Freshly hatched larvae were placed in a small clear plastic container that was situated inside of a chamber with two infrared long-pass filters (RG610, Reichmann Feinoptic, Brokdorf, Germany) installed above and below the container. The chamber was placed on a Zeiss Stemi 2000 stereo microscope. Through a hole on one side of the chamber, an LED served as the light stimulus. Eight different LEDs covered wavelengths from UV 375 nm to 630 nm in the red part of the spectrum. For each experiment between 50 to 230 freshly hatched larvae were placed inside the chamber. We recorded the reaction of the animals to the stimulus with an industrial monochrome CMOS camera (DMK 23U445, The Imaging Source, Bremen, Germany). Each recording starts in darkness for 30 s, followed by 15 s of illumination and another 45 s of darkness. The response of the animals to each wavelength was assayed at least three times, always with a new batch of animals and an extra four without any light stimulus as a control. From each recording, we removed the background and enhanced the contrast in Fiji. Subsequently, we tracked the animals' position in enhanced recordings for each frame with the Fiji plugin Trackmate (Tinevez et al., 2017). The tracking information was used to calculate the mean and median position of the animals for each frame for a single axis. To make different recordings comparable, we used the mean and median position of animals during the initial darkness to subtract from each position for each frame. Boxplots Violinplots were inferred from all the tracked positions of all the animals during a time of guaranteed illumination (second 40 to 42, Figure 5A, dashed box).