Eye evolution: common use and independent recruitment of genetic components

Abstract

Animal eyes can vary in complexity ranging from a single photoreceptor cell shaded by a pigment cell to elaborate arrays of these basic units, which allow image formation in compound eyes of insects or camera-type eyes of vertebrates. The evolution of the eye requires involvement of several distinct components—photoreceptors, screening pigment and genes orchestrating their proper temporal and spatial organization. Analysis of particular genetic and biochemical components shows that many evolutionary processes have participated in eye evolution. Multiple examples of co-option of crystallins, Gα protein subunits and screening pigments contrast with the conserved role of opsins and a set of transcription factors governing eye development in distantly related animal phyla. The direct regulation of essential photoreceptor genes by these factors suggests that this regulatory relationship might have been already established in the ancestral photoreceptor cell.

1. Introduction

Eyes of some sort occur in many animal phyla, but their anatomy, ontogenetic origin and degree of sophistication vary enormously (Land & Nilsson 2002). For the purpose of this text, we use the minimal definition of an eye as a photoreceptor cell in the close vicinity of a screening pigment (Arendt & Wittbrodt 2001; Land & Nilsson 2002)—a situation found, for example, in Hesse eye cups of amphioxus (Lacalli 2004). As these two components are formed by different genes and genetic pathways, which have different evolutionary histories, the evolution of an eye then becomes a question of evolutionary history of these separate components and their continuous or repeated integration to an organ called ‘eye’. The aim of this review is to gather available genetic and biochemical data regarding evolution of particular eye components, i.e. phototransduction (which is extensively reviewed elsewhere in this issue), pigmentation and regulatory genes involved, and unite these processes.

2. Genetic components of photoreceptors

Despite a remarkable variation in size and complexity, the common indispensable basis of all animal eyes is the photoreceptor cell containing photosensitive molecules, which are connected to a downstream phototransduction cascade. The photoreceptor cells are classified according to the morphology of membrane protrusions bearing visual pigments as ‘rhabdomeric’, which form microvilli, and ‘ciliary’, where the membrane surface is increased by folding the membrane of the cilium (Eakin 1979; Arendt 2003). Primary observations revealed the rhabdomeric photoreceptors as being predominantly present in the eyes of invertebrates, whereas the vertebrate eyes employ the ciliary type; however, several exceptions from this rule do exist (Arendt & Wittbrodt 2001; Land & Nilsson 2002). Both photoreceptor cell types have always co-existed in bilaterians, as suggested by both types found in amphioxus (Lacalli 2004) and confirmed by recent morphological and molecular studies (Arendt et al. 2004; Velarde et al. 2005). Still, it is not clear why the two types of photoreceptors were employed in the eyes of invertebrates versus vertebrates in a mutually exclusive way.

(a). Opsins: variation on ancestral theme

The first step of photoreception is mediated by light-sensitive transmembrane proteins containing retinal chromophore, generally termed rhodopsins. They have been found in most groups of organisms including archeal prokaryotes (Blanck & Oesterhelt 1987), unicellular eukaryotes (Nagel et al. 2002), fungi (Bieszke et al. 1999) and metazoa. The functions of rhodopsins in these organisms vary from photon-driven ionic pumps in prokaryotes (Blanck & Oesterhelt 1987), sensory molecules in fungi, to a light-gated ion channel in the eyespot of green algae (Spudich et al. 2000; Nagel et al. 2002). Metazoan opsins are seven transmembrane proteins belonging to the superfamily of G-protein coupled receptors (GPCRs) and often are coupled to a Gα protein to mediate a phototransduction cascade. Despite these distinct functions of rhodopsins, several common features, such as a transmembrane structure, conserved retinal group covalently bound to a lysine residue and similarities in exon–intron structure, suggested the possibility of common evolutionary origin of all opsins. However, a recent bioinformatic study (Larusso et al. 2008) provided strong evidence that at least prokaryotic and metazoan opsins are not homologous, thus revealing their common features as a remarkable example of convergent evolution.

With increasing number of sequenced genomes available, the opsin gene repertoire has been described in several animal species (Raible et al. 2006; Plachetzki et al. 2007; Holland et al. 2008). This information then enables the understanding of metazoan opsin evolution and origin (figure 1), which is inseparably connected with the origin of eyes itself. The metazoan opsins seem to originate early in metazoan evolution from a single GPCR gene by acquisition of light sensitivity. As no opsin genes have been found in available genomic data for the choanoflagellate Monosiga and the poriferan Amphimedon (Plachetzki et al. 2007; Suga et al. 2008), it is probable that this event pre-dated the cnidarian–bilaterian ancestor. Before the split of cnidarians and bilaterians, the newly formed ur-opsin gene underwent duplication producing two ancestral classes of opsins: ciliary opsin class (c-opsin) conserved in cnidarians and bilaterians and a second opsin gene eventually giving rise to the rest of bilaterian opsins (see below) and persisting in cnidarians as ‘cnidopsin’ class (figure 1) (Plachetzki et al. 2007).

Figure 1.

A schematic diagram of opsin distribution among eyes in different animal phyla. Particular opsins subfamilies are distinguished by different colouring. The lines leading to different phyla depict the hypothetical evolutionary fate of given opsin lineage based on available genomic and other data. The question mark denotes such animals where the presence of a given opsin lineage has not been confirmed yet. The colour of the eye-like pictogram corresponds to the type of photoreceptor cell employed in the eye. The class of opsin employed in the eye is represented by the colour of the rectangle next to the eye-like pictogram. If known, the Gα subunit interacting with the opsin is shown. Note that a small subset of vertebrate retinal ganglion cells expresses melanopsin coupled to a Gq signalling cascade (Panda et al. 2005). Although these cells fulfil the definition of a minimal eye, they are not the major photoreceptors of the eye and are not considered in this figure.

In cnidarians, the c-opsins are expressed in the ciliary photoreceptors of adult eyes of hydrozoan Cladonema, cubozoans Tripedalia and Carybdea (Koyanagi et al. 2008; Kozmik et al. 2008a; Suga et al. 2008).

The bilaterian opsin repertoire comprises the c-opsin class and the second ancestral class which has diversified into several subclasses termed as rhabdomeric opsins (r-opsins), Go-coupled opsins, neuropsins and RGR (Terakita 2005). From these, only three groups (namely c-, r- and Go-coupled opsins) seem to have been recruited for visual purposes, whereas the function of other subclasses is probably supportive, as, for example, photoisomerases (RGR-opsin) involved in the retinal visual cycle (Radu et al. 2008). With one exception of Go-coupled opsin mediating the phototransduction in the ciliary part of the retina in scallop Patinopecten yessoensis (Kojima et al. 1997), the rhabdomeric and ciliary photoreceptors of bilaterians consistently employ r- and c-opsins, respectively.

(b). Phototransduction in ciliary photoreceptors: promiscuity in Gα coupling

The different morphology of rhabdomeric and ciliary photoreceptors is further reflected in the level of phototransduction cascades operating in these cells, although some common elements, such as opsins or arrestins, participate in both. The rhabdomeric phototransduction cascade, which is mediated by Gα_q and phospholipase C (Suzuki et al. 1995), seems to be evolutionarily conserved from protostomes to vertebrate retinal ganglion cells (Koyanagi et al. 2005; Panda et al. 2005; Contin et al. 2006; Graham et al. 2008). In contrast, the ciliary photoreceptors may employ both G_o-opsin (Kojima et al. 1997) and c-opsins, whose Gα protein partners may be rather variable. In vertebrate rods and cones, the c-opsin couples via transducin Gα_t to downstream hyperpolarizing phototransduction cascade. Because the Gα_t subfamily originated together with other rod and cone specific phototransduction genes during vertebrate-specific whole-genome duplications (Nordstrom et al. 2004; Milligan & Kostenis 2006), the Gα_t subunits are not present in invertebrates. A question arises—what then are the Gα proteins involved in ciliary phototransduction cascades in invertebrates and cnidarians? Apparently, the members of the Gα_i/o protein subfamily, from which Gα_t proteins evolved, could be plausible candidates for this function. Consistent with this assumption, a Gα_i1 protein subunit is expressed in the ciliary photoreceptor cells of Ciona intestinalis (Yoshida et al. 2002). In a reptile Uta stansburiana, a vertebrate c-opsin (parietopsin), expressed in the parietal eye retina, signals via a Gα_o protein (Su et al. 2006), and in the same cell, another vertebrate visual/non-visual opsin, pinopsin, couples with gustducin—a third vertebrate paralogue of transducin not used in rods and cones. In contrast to its reptile counterpart, chicken pinopsin has been shown to interact with Gα₁₁ subunit (Kasahara et al. 2002), which is closely related to Gα_q protein. An exciting surprise came from a recent study (Koyanagi et al. 2008), which revealed that phototransduction in cubomedusan Carybdea rastonii is mediated by Gα_s cascade.

Taken together, these findings indicate that the coupling specificity of ciliary opsins could be rather promiscuous in comparison to the rhabdomeric opsins retaining more strictly the Gα_q specificity. Since the origin of G proteins pre-dated the origin of opsins, the opsin–Gα protein interaction evolved by co-option and subsequent coevolution (Plachetzki & Oakley 2007). Multiple co-option events during c-opsin evolution might explain their ‘promiscuity’ and reconcile an apparent discrepancy between the data pointing to G_i/o phototransduction cascade in Tripedalia (Kozmik et al. 2008a) and G_s-mediated transduction in Carybdea (Koyanagi et al. 2008). Although Carybdea and Tripedalia are closely related, the opsin sequences identified in the studies are rather diversified and suggest that two different phototransduction cascades in structurally the same eyes might be possible. Moreover, all the opsins found to be expressed in the eye of hydrozoan Cladonema radiatum and assigned to the ciliary class (Suga et al. 2008) show even more sequence diversification (figure S1, electronic supplementary material). Together with low bootstrap support in phylogenetic trees and discrepancies in total number of Hydra and Nematostella opsins identified in two independent studies (Plachetzki et al. 2007; Suga et al. 2008), further analyses are required to fully resolve the relationships among cnidarian opsins and address their Gα coupling specificity and role in phototransduction.

3. Dark pigments: redeployment without logic

The second essential component of a postulated minimal eye is a dark shielding pigment, which brings the organism additional information of light direction. Generally, three types of compounds—melanins, ommochromes and pterins—serve as shielding pigments in most animal eyes (see electronic supplementary material for detailed information). The distribution of screening pigments in the animal kingdom (figure 2) does not follow any rule and is not correlated with a certain eye or photoreceptor cell. This leads to an assumption that the union of a dark pigment and photoreceptor cannot be traced to any ancestral condition, but is rather an outcome of independent assembly of these two components in different animal phyla. Another possibility is that the pigments coexisted in ancient pigmented photoreceptor cells and were lost in several extant pigment cells.

Figure 2.

A schematic diagram of distribution of screening pigments in different animal phyla. The photoreceptor cell type screened by particular pigments is depicted on the left side of the box of each phylum. There is no apparent correlation between the photoreceptor cell type and shielding pigment used. Note the unknown type of photoreceptor and unique deployment of haemoglobin as a shielding pigment of nematode Mermis, caused probably by independent origin of the eye in this organism. See electronic supplementary material for further details and references.

4. Lens-containing eyes: a story of endless crystallin diversity

Lens-containing eyes are a feature of a surprisingly broad spectrum of organisms across the animal kingdom, which represent a significant improvement of the simple eye composed of just photoreceptor cells and pigment cells. Highly abundant, water-soluble proteins that contribute to the main optical properties of lens, transparency and refractive power, are collectively called crystallins (reviewed in Piatigorsky 2007). The molecular mechanism of their optical function is based on their exceptional solubility and stability at high concentration. In some cases, crystallins may account for up to 90 per cent of the dry weight of the lens. Remarkably and in striking contrast to the universal conservation of opsins as the visual pigments in the photoreceptors, the lens crystallins are enormously diverse proteins that are often taxon-specific, i.e. different proteins function as crystallins in different species (Wistow & Piatigorsky 1988; de Jong et al. 1989; Piatigorsky & Wistow 1989). Crystallin diversity might be a consequence, at least in some cases, of multiple independent origins of lenses during evolution. A surprising number of proteins has biophysical properties that enable them to fulfil an optical role as lens crystallins in different taxonomic groups. Nevertheless, the selection of a particular polypeptide as a crystallin appears not to be entirely random since certain proteins are preferentially recruited as lens crystallins. Proteins used as lens crystallins are often related or identical to ubiquitously expressed metabolic enzymes or physiological stress proteins. For example, all vertebrate lenses contain the α-crystallins that belong to the family of small heat shock proteins (Ingolia & Craig 1982; de Jong et al. 1993) and the β/γ-crystallins that are related to microbial stress proteins (Wistow 1990; D'Alessio 2002). Vertebrate α-crystallins are indeed effective chaperones that protect partially denatured proteins from aggregating in the lens (Horwitz 1992). The co-option of a protein with chaperone-like activity as a lens crystallin makes sense, in light of the fact that proteins in vertebrate lenses must remain functional and soluble for the entire lifespan of an organism. The selective advantage of recruiting enzyme-crystallins is much less clear. First, they do not fall within a common metabolic category. The catalytic activities include, but are not limited to, lactate dehydrogenase (ε-crystallin in crocodiles and birds), argininosuccinate lyase (δ-crystallin in reptiles and birds), α-enolase (τ-crystallin in turtles), glutathione S-transferase (S-crystallin in cephalopods), ADP-ribosylglycohydrolase (J1-crystallin in cubozoan jellyfish) or glyceraldehyde 3-phosphate dehydrogenase (π-crystallin in diurnal geckos) (Piatigorsky 2007). Furthermore, some enzyme-crystallins have no or diminished catalytic activity, e.g. aldehyde dehydrogenase/Ω-crystallin in scallops (Piatigorsky et al. 2000) or argininosuccinate lyase/δ1-crystallin in chickens (Piatigorsky et al. 1988), suggesting that their function in the lens is limited to a refractive role.

It turned out that most crystallins are not lens-restricted proteins being expressed at a low level in other tissues as well. This means that apart from their optical role in the lens, crystallins serve an apparently different function elsewhere. A concept of ‘gene sharing’ has emerged from crystallin studies (Piatigorsky et al. 1988; Piatigorsky & Wistow 1989; Piatigorsky 2007). Gene sharing proposes that the protein encoded by a single gene may perform two entirely different functions: a structural (refractive) function in the lens as a lens crystallin and a catalytic or stress function elsewhere in the organism. An important implication of gene sharing is that a protein can evolve a new role, without losing its original function, simply by a change in gene expression (Piatigorsky & Wistow 1991). It has become apparent over time that gene sharing and repeated use of proteins for new tasks are not limited to crystallins and lenses, but they rather represent a common evolutionary strategy (Piatigorsky 2007).

The overall diversity of crystallins throughout the animal kingdom is indicative of convergent evolutionary solution: lens cells simply choose to synthesize a suitable protein to high levels to make their lenses optically useful. Such opportunistic molecular strategy emphasizes the key role of transcriptional regulation in crystallin gene recruitment. The high level lens-preferred expression of crystallin genes is often regulated by transcription factors implicated in eye development as well (Duncan et al. 2004). Perhaps the best example of such a dual role of transcription factors as regulators of both eye development and crystallin gene expression is the case of Pax proteins (reviewed in Cvekl & Piatigorsky 1996; Gehring & Ikeo 1999; Cvekl et al. 2004; Kondoh et al. 2004; Kozmik 2005). The presence of Pax regulatory elements in non-homologous crystallin genes in phylogenetically distant animal species is due to convergent evolution (Carosa et al. 2002; Cvekl et al. 2004; Kozmik et al. 2008b). It reflects the situation that the diverse crystallin genes have been recruited independently during evolution and hence their regulatory elements to achieve lens-preferred expression must have been acquired by independent events as well (figure 3).

Figure 3.

Dual role of transcription factors in regulation of both eye development and differentiation genes. The box on the left-hand side represents the sum of largely unknown developmental genes regulated by corresponding transcription factors based on functional data. The letters represent different animals (V, vertebrates; A, ascidians; D, Drosophila; C, cnidarians; M, molluscs; P, planarians). The arrows on the right-hand side represent a direct influence of a given factor on differentiation set of genes proved by biochemical methods (DNA-binding assay, ChIP, transgenesis, luciferase assays, etc.) The green arrows indicate the ancestral interaction proposed by the ‘bipartite’ model. The red arrow highlights the proposed role of Otx in the regulation of ancestral phototransduction genes. Co-option of a certain transcription factor to a new role is indicated by dashed line. We propose that the transcription factors were independently co-opted for regulation of genes governing eye development in different species and these downstream genes may vary among species. Please note that cross-regulatory interactions of transcription factors are not considered in this scheme for simplicity.

5. Redeployment of a selected set of transcription factors for animal eye development

Structural components of simple or complex (lens-containing) eyes described above are encoded by genes expressed during terminal cell differentiation at the end of developmental processes. Any developmental process to be completed properly requires tight regulation by a dedicated set of transcription factors. Given the enormous diversity of animal eyes, it came as a surprise that certain transcription factors are redeployed for visual system development far more often than others. In addition to governing eye morphogenesis, some of these transcription factors are directly involved in the regulation of differentiation genes encoding structural eye components.

(a). Pax family of transcription factors

Pax transcription factors are defined by the presence of a highly conserved DNA binding domain, the paired domain (Burri et al. 1989; Treisman et al. 1991). In addition to a paired domain, some Pax proteins (such as Pax6) contain a second DNA binding domain, a homeodomain. Because a PaxB-like gene was identified in porifera (Hoshiyama et al. 1998) and placozoa (Hadrys et al. 2005), the origin of Pax genes predates the origin of eyes and the nervous system. The widespread use of Pax genes in the genetic programme underlying eye formation throughout the animal kingdom is remarkable. Mutations in the Pax6 gene disrupt eye development in both mammals (Hill et al. 1991) and insects (Quiring et al. 1994). The ability to induce ectopic eyes through Pax6 misexpression has furthermore been demonstrated in Drosophila and vertebrates (Halder et al. 1995; Chow et al. 1999). The key role of the Pax6 gene for eye morphogenesis in such diverse species led to the proposal of Pax6 being a ‘master control gene’ in animal visual system development (Gehring & Ikeo 1999). Such a simplified scenario, however, has to be dismissed. The term ‘master control gene’ implies that the Pax6 transcription factor is located at the top of a gene cascade and initiates eye development in almost any tissue where it is ectopically expressed. However, neither seems to be the case. For instance, in the absence of Pax6, presumptive retina can develop up to the optic cup stage, albeit abnormally (Grindley et al. 1995). Moreover, within the lens placode, Pax6 expression is under the control of other transcription factors, Meis1 and Meis2 (Zhang et al. 2002). In Drosophila, the ability of the two Pax6 paralogues, ey (Halder et al. 1995) and toy (Czerny et al. 1999), to induce ectopic eyes is restricted both spatially and temporally. These limitations suggest that ey/toy modify an existing programme of sensory organ development rather than initiate the entire eye morphogenesis. Furthermore, toy controls more than just eye morphogenesis because a loss-of-function mutation produces flies missing an entire head (Kronhamn et al. 2002). Likewise, in mice, Pax6 is a pleiotropic regulator of development (Simpson & Price 2002).

Despite these limitations, there is no doubt that Pax6 and other Pax genes have been frequently redeployed for visual system development. Apart from Pax6/ey/toy, three other Pax genes (Pax2, Eyg, toe) might have a role in the genetic programme underlying Drosophila eye morphogenesis (Fu & Noll 1997; Jang et al. 2003; Dominguez et al. 2004). Likewise, in mice, Pax2 cooperates with Pax6 in the development of the retinal pigment epithelium (Baumer et al. 2003) and the mutual repression of Pax6 and Pax2 is responsible for morphogenesis of the entire mouse optic primordium (Schwarz et al. 2000). Even in species where genetic studies have not been done, there has been a generally good correlation between the presence of eyes and Pax gene expression (Loosli et al. 1996; Glardon et al. 1997, 1998; Tomarev et al. 1997; Callaerts et al. 1999; Arendt et al. 2002; Hartmann et al. 2003; Kozmik et al. 2003; Quigley et al. 2007), although the expression is never eye-restricted. There are few notable examples, however, known so far among bilaterians of eyes developing in the absence of Pax6. Pax6 is apparently not expressed in developing Limulus eyes (Blackburn et al. 2008), developing Platynereis adult eyes (Arendt et al. 2002), Hesse eye cups of amphioxus (Glardon et al. 1998) and its function is not required for the eye regeneration in planarians (Pineda et al. 2002) as well as the Bolwig organ in Drosophila (Suzuki & Satoh 2000). Nonetheless, Pax genes arguably have an ancient and fundamental role in visual system development. The bipartite model (Kozmik 2005) proposes that the two independent DNA binding domains within a single Pax transcription factor have been co-opted for two essential features of the prototypical eye, production of a dark shielding pigment and production of a photopigment. Frequent deployment of Pax genes in eye development in phylogenetically diverse species may reflect their ancestral role in the regulation of key differentiation genes (figure 3).

(b). Sine oculis (Six) and eyes absent (Eya)

Eye specification in Drosophila is governed by the members of the retinal determination gene network that includes, apart from Pax6 paralogues (ey, toy), eyes absent (eya), sine oculis (so/six) and dachshund (dac) (Pappu & Mardon 2004; Silver & Rebay 2005; Friedrich 2006). This highly interactive network of genes is sometimes referred to as Pax–Six–Eya–Dach network (PSEDN; Kawakami et al. 2000) to reflect the situation that it has been co-opted for non-retinal roles in other species and developmental contexts (Heanue et al. 1999; Xu et al. 1999; Ozaki et al. 2004; Kozmik et al. 2007). Six and Eya are evolutionarily old gene families, whose origin like in the case of Pax predates the origin of eyes and that are characterized by the conserved biochemical roles of the encoded proteins (Bebenek et al. 2004; Silver & Rebay 2005). This is perhaps best documented by Eya gene that encodes a protein phosphatase (Li et al. 2003; Rayapureddi et al. 2003; Tootle et al. 2003). Eya phosphatase activity is required for eye development in Drosophila, yet the same phosphatase activity is already found in the plant orthologue (Silver & Rebay 2005). Eya functions in the transcription factor complex with members of Six gene family (Ohto et al. 1999), which might explain consistent co-expression in many different developmental settings (Silver & Rebay 2005).

Nonetheless, the current evidence strongly suggests that Eya and Six have an ancient role in eye development. Orthologues of both genes are involved in visual system development in both invertebrates and vertebrates. In Drosophila, deficiency mutations of so and eya are characterized by loss of all visual sense organs (Bonini et al. 1993; Cheyette et al. 1994; Zimmerman et al. 2000; Friedrich 2006). Likewise, eya and so are expressed in the embryonic visual system of directly developing insects (Dong & Friedrich 2005) and their knockdown induces a long-term arrest of eye development (Dong & Friedrich in press). In addition, a divergent Six3/6-like gene, optix, does not synergize with eya and contributes to compound eye morphogenesis in Drosophila by a mechanism that is apparently Pax6/ey independent (Seimiya & Gehring 2000). In the annelid Platynereis, the orthologue of Six1/2 is expressed in all components of the larval and adult visual system (Arendt et al. 2002). Orthologues of eya and so are functionally required for planarian eye regeneration (Pineda et al. 2002; Mannini et al. 2004). In the hydrozoan jellyfish Cladonema, a species with well-developed eyes, orthologues of Six1/2 and Six3/6 are expressed, both during normal development and during the process of regeneration (Stierwald et al. 2004). Eya and Six4/5 genes in the chordate amphioxus are transiently expressed in the two rhabdomeric photoreceptive neurons that flank a biconcave pigment cell to comprise the first Hesse eyecup (Kozmik et al. 2007). Among the Six genes that are expressed in vertebrate eyes only Six3/6-like genes appear to have a critical role during development (Zuber et al. 1999; Kobayashi et al. 2001; Carl et al. 2002; Li et al. 2002; Zhu et al. 2002; Lopez-Rios et al. 2003; Liu et al. 2006). In addition, Six3 has been implicated in the regulation of mouse rhodopsin gene (Manavathi et al. 2007). The analysis of a functional role of Eya in vertebrate eye development is complicated by the fact that three mouse Eya genes (Eya1–3) are expressed in the developing eye (Xu & Saunders 1997) and so their combined loss-of-function phenotype has to be generated in order to unmask possible redundancy. Unlike the other trascriptional regulators discussed in this review, Eya has so far not been associated with the expression of key differentiation genes such as opsins or genes regulating dark pigment formation (figure 3).

(c). Orthodenticle-related homeobox (Otx)

The first member of the Otx gene family, orthodenticle (Otd), has been isolated from Drosophila and shown to be necessary for development of photoreceptors in the compound eye, Bolwig organ and the ocelli (Finkelstein et al. 1990; Royet & Finkelstein 1995; Vandendries et al. 1996). Otd also participates in terminal photoreceptor differentiation. It has been shown to directly regulate opsins (Tahayato et al. 2003) and influence the expression of genes involved in rhabdomeric phototransduction cascade (Ranade et al. 2008). In other invertebrates, the expression of Otx genes in photoreceptors has been reported in planarians (Umesono et al. 1999), putative eye-field precursor of the annelid Hydroides elegans (Arenas-Mena & Wong 2007), ciliary and rhabdomeric photoreceptor cells of Platynereis dumerilii (D. Arendt 2008, personal communication), sensory pigment cells of ascidians (Wada et al. 1996) and the frontal eye region of amphioxus (Williams & Holland 1996). Reciprocal rescue experiments with Drosophila and mammalian Otx orthologues demonstrated that at least part of ancestral genetic and biochemical interactions is still conserved between vertebrates and invertebrates (Acampora et al. 1998; Nagao et al. 1998).

Multiple vertebrate orthologoues of otd termed Otx1, Otx2 and Crx/Otx5 (Germot et al. 2001; Plouhinec et al. 2003) probably arose during whole genome duplication, since a single Otx gene is present in the genome of C. intestinalis (Wada et al. 2003) and amphioxus (Williams & Holland 1998). Besides the role of Otx genes in early vertebrate development of anterior neural structures and the brain (Simeone et al. 2002; Acampora et al. 2005), these genes are necessary for the proper development of the pineal gland and the eye (Martinez-Morales et al. 2001; Nishida et al. 2003; Plouhinec et al. 2005). Later in development, Otx genes play a crucial role in the terminal differentiation of photoreceptors and their maintenance during postnatal development (Nishida et al. 2003; Koike et al. 2007). The expression of Otx genes has been also detected in immature retinal ganglion cells (Bovolenta et al. 1997; Martinez-Morales et al. 2001; Rath et al. 2007)—putative descendants of the rhabdomeric photoreceptor line in vertebrates (Arendt 2003).

The role of vertebrate Otx genes in the regulation of eye-specific genes has been extensively studied and led to the discovery of many direct target genes. Crx, strongly expressed in differentiated photoreceptor cells, directly regulates the phototransduction genes, rhodopsin, β-PDE, arrestin and guanylate cyclase, via binding the PCE element in the promoters (Chen et al. 1997; Furukawa et al. 1997; Qian et al. 2005) (for a review, see Hennig et al. 2008). The expression of ciliary-phototransduction cascade genes in the vertebrate pineal gland is mediated by the action of Otx genes as well (Appelbaum & Gothilf 2006; Takechi et al. 2008). Besides the direct regulation of photoreceptor-specific genes, Otx genes are involved in the regulation of pigmentation. In ascidians, the Tyrp gene is a direct target of Otx (Wada et al. 2002). The vertebrate homologue Otx2 has been shown to bind to the promoters of Mitf, tyrosinase and Tyrp1 (Martinez-Morales et al. 2003) as well as Tyrp2 in the retinal pigmented epithelium (Takeda et al. 2003).

The fact that Otx genes are expressed in both rhabdomeric and ciliary photoreceptors across animal phyla points to their ancient role in photoreceptor cell differentiation (Ranade et al. 2008). The differentiation processes and regulatory subcircuits composed of differentiation genes regulated by certain transcription factors seem to be well conserved during evolution (Arendt 2008). The direct regulation of opsins by Otx could be an example of such a differentiation subcircuit dating back before the split of r- and c-opsins. Based on the scenario of opsin evolution proposed in Plachetzki et al. (2007), the regulatory relationship must have been already established between an ancestral Otx and ur-opsin before the split of cnidarians and bilaterians. Then, after the diversification of c-opsins and the second opsin class, this regulatory unit has been preserved in both rhabdomeric and ciliary photoreceptors (figure 4).

Figure 4.

A hypothetical scenario suggesting the ancient regulatory relationship between Otx and ur-opsin predating the cnidarian–bilaterian split. The ur-opsin has been already regulated by the Otx gene and probably by Pax (not excluding other transcription factors involved). After the duplication and diversification of r- and c-opsin, the regulation by Otx has been preserved in both lineages, whereas Pax-dependent regulation has been lost in c-opsin lineage. Retinal homeobox Rx might have been recruited for regulation of c-opsin. Alternatively, Rx has regulated the ur-opsin and this role has been lost in r-opsin lineage. With increasing complexity of animal body plans, all the transcription factors have consequently acquired additional roles in eye development.

(d). Retinal homeobox (Rx)

During vertebrate development, Rx genes are expressed in the anterior forebrain, retinal primordia and pineal gland (reviewed in Bailey et al. 2004). The over-expression of Rx in Xenopus leads to ectopic formation of retinal tissue (Mathers et al. 1997). In zebrafish chokh mutants, a non-sense mutation in Rx3 paralogue leads to the loss of eyes (Loosli et al. 2003) and Rx knock-out mice lack the eye and the anterior brain structures (Mathers et al. 1997). Similar to vertebrates, the ascidian homologue of Rx is expressed in the anterior brain and the knock-down resulted in the loss of photoreceptor cells (D'Aniello et al. 2006). In contrast, Rx is not expressed in planarian eyes (Salo et al. 2002) and genetic studies in Drosophila have shown a clear dispensability of Rx for compound eye development (Davis et al. 2003). The explanation of this result came from the emerging concept of sister cell types (Arendt 2003) and the fundamental discovery of Rx expression in the ciliary photoreceptors of Platynereis brain (Arendt et al. 2004). These findings led to the identification of Rx as a ciliary photoreceptor-specific marker. One may speculate, what is the reason for keeping the expression of Rx in differentiated ciliary photoreceptors in distantly related species. Analogous to the Otx scenario, the direct regulation of vertebrate c-opsins by Rx (Kimura et al. 2000; Wang et al. 2004; Pan et al. 2006) may point to an ancestral condition of ciliary photoreceptor cell type. In these cells, the regulation of c-opsins by Rx might have been already established and, being a differentiation subcircuit, it was conserved throughout evolution. Rx might be later co-opted for new roles in developmental regulatory networks operating in the anterior body part. Since this hypothesis is based on an Rx/c-opsin regulatory relationship so far confirmed solely in vertebrates, one has to keep in mind a possible co-option of Rx for this function.

(e). Microphthalmia-associated transcription factor (Mitf)

Mitf is a member of the Mitf/TFE family of bHLH-lecine-zipper transcription factors (Hodgkinson et al. 1993). Homozygous-mutant Mitf mice show severe defects of body pigmentation, have small unpigmented eyes, lack melanocytes in the inner ear and are deaf (reviewed in Steingrimsson et al. 2004). Genetic experiments have shown that multiple isoforms of Mitf are responsible for driving the expression of Tyr and Tyrp in melanocytes and retinal pigmented epithelium in cooperation with Otx and Pax genes (Martinez-Morales et al. 2004; Murisier & Beermann 2006; Bharti et al. 2008). Besides its role in melanogenesis, Mitf has been suggested to be involved in the regulation of the pteridine synthesis pathway in zebrafish (Ziegler 2003). In ascidians, Mitf is expressed in the precursors of pigmented cells in the brain vesicle (Yajima et al. 2003). Although the regulatory mechanism of ascidian Tyr and Tyrp genes is not yet fully understood (Toyoda et al. 2000), ascidian Tyrp gene has been shown to be directly regulated by Otx (Wada et al. 2002) and over-expression experiments suggest that Mitf and Pax are involved as well (Yajima et al. 2003; Toyoda et al. 2004).

The role of Mitf in eye development does not seem to be restricted to deuterostomes. Drosophila homologue of Mitf has been shown to be expressed in the eye-antennal imaginal disc and expression of dominant negative form of Drosophila Mitf resulted in enlarged photoreceptor field (Hallsson et al. 2004). Strikingly, a cnidarian homologue of Mitf has been isolated (Kozmik et al. 2008a) and shown to be expressed in the melanin-pigmented photoreceptor cells of the camera-type eye of Tripedalia cystophora. The expression of Mitf in cnidarians as well as bilaterian eyes and its interaction with Otx and Pax genes in deutorostomes raises speculations about Mitf–Pax–Otx cooperation within an ancestral pigmented photoreceptor cell.

6. Conclusions

The commonalities in the use of structurally similar seven-transmembrane receptors (opsins) as animal eye photopigments stem from their shared evolutionary history. The handfull of dark (shielding) pigments that animals can make through various biosynthetic pathways are all apparently used for screening purposes without any evolutionary pattern or developmental logic. The recruitment of lens crystallins provides an extreme example of an opportunistic use of almost any soluble cytoplasmic protein for a refractive role. In contrast, there is a homology at the level of transcriptional regulators operating in developmental programmes in structurally diverse eyes. Although the expression of genes in all metazoa is generally regulated by a large number of diverse transcription factors often belonging to distinct families, certain transcription factors (i.e. Pax, Six, Eya, Otx, Mitf) have been deployed for the regulation of eye development far more often than others. The reasons for this phenomenon are not entirely clear at the moment, however, might be due to several underlying molecular mechanisms that are not necessarily mutually exclusive. Anteriorly expressed transcription factors are more likely to be co-opted for the regulation of an eye programme due to the fact that eyes are located more often at the front rather than at the back of the animal. Perhaps the most important aspect contributing to the observed homology of transcription factors is due to stochastically chosen ancestral regulatory connections. We can reasonably well argue that certain transcription factors were more or less randomly chosen for the regulation of essential genes in the ancestral (pigmented?) photoreceptor cell such as the gene encoding an ur-opsin. Gene duplication events generated opsin gene duplicates that began the process of divergence both in the coding sequence and in the regulatory regions. One might expect that the diverging genes encoding essential eye components have retained in their promoters binding sites of any transcription factor provided that such regulatory link was useful and the spatio-temporal expression pattern of the regulator was maintained. Such transcriptional regulators were later co-opted into new roles as ‘higher structure’ organizers of increasingly complex but diverse eyes. Frequent deployment in eye development in phylogenetically diverse species, thus, most likely reflects the ancestral role of a particular transcription factor in the regulation of a key differentiation gene.