The Mammalian Crx Genes Are Highly Divergent Representatives of the Otx5 Gene Family, a Gnathostome Orthology Class of Orthodenticle-Related Homeogenes Involved in the Differentiation of Retinal Photoreceptors and Circadian Entrainment Jean-Louis Plouhinec,* Tatjana Sauka-Spengler,* Agne`s Germot,*  Chantal Le Mentec,* The´re`se Cabana,à Gavan Harrison,§ Claude Pieau,k Jean-Yves Sire,{ Ge´raldine Ve´ron,# and Sylvie Mazan* *CNRS-Universite´ Paris, Orsay, France;  Faculte´ des Sciences et Techniques, Universite´ de Limoges, Limoges, France; àDe´partement de Sciences Biologiques, Universite´ de Montre´al, Montre´al, Canada; §School of Science Food and Horticulture, University of Western Sidney, Parramata, Australia; kInstitut Jacques Monod, CNRS-Universite´s Paris, Paris, France; {Evolution et De´velopement du Squelette dermique, Universite´ Paris, Paris, France; and #Mammife`res et Oiseaux, Muse´um National d’Histoire Naturelle, Paris, France The mammalian Crx genes are highly divergent orthodenticle (otd)-related homeogenes that play important roles in the differentiation of retinal photoreceptors and the circadian entrainment. However, their evolutionary origin and orthological relationships with other otd-related genes remain unclear. An orthology relationship of these genes with the highly conserved Otx5 genes identified in fish and amphibians, and also expressed in the eye and epiphysis, has been proposed previously but remains controversial. To test this hypothesis, we have identified Crx genes in a wide range of mammals, including three marsupials, and Otx5-related genes in a lizard, a turtle, and two archosaurs (crocodile and chick), as well as in the pufferfish. Phylogenetic analyses of the coding sequences show that the mammalian Crx genes are orthologous to the Otx5-related genes isolated in other gnathostomes. They also indicate that a duplication event has taken place in actinopterygians, after the splitting of the Cladistia, and that a relaxation of the structural constraints acting on the gene coding region has occurred early in the mammalian lineage. This process may be linked not only to the loss of ancestral Otx5/Crx functions during gastrulation or in the retinal pigmented epithelium, but also to the evolution of photic entrainment mechanisms in mammals.

Introduction In vertebrates, orthodenticle (otd)-related homeodomain genes (Otx) are involved in multiple aspects of head development, including the early specification of anterior neuroectoderm, the regionalization of the forebrain, and the morphogenesis of sense organs (Acampora, Gulisano, and Simeone 2000). Among jawed vertebrates, three orthology classes, Otx1, Otx2, and Otx5, which have emerged after the splitting of the cephalochordates and before the gnathostome radiation, have been identified (Williams and Holland 1998; Germot et al. 2001). Whereas Otx1-related and Otx2-related genes have been identified in all major gnathostome taxa, the Otx5 class remains more poorly characterized. Otx5 genes showing highly conserved coding regions have been characterized in cartilaginous and ray-finned fish (dogfish, reedfish, and zebrafish), as well as in amphibians. These genes share highly conserved expression domains in the pineal gland and the eye, in addition to more variable expression sites such as the olfactory placodes in the dogfish, the cement gland in X. laevis, or the balancers in P. waltl (Kuroda et al. 2000; Vignali et al. 2000; Germot et al. 2001; SaukaSpengler et al. 2001; Gamse et al. 2002; Sauka-Spengler et al. 2002). In line with these expression features, the zebrafish DrOtx5 gene is essential for the transactivation of pineal genes showing a circadian gene expression (Gamse et al. 2002). Despite this important role, unKey words: Otx5, Crx, circadian entrainment, specification of photoreceptors. E-mail: [email protected]. Mol. Biol. Evol. 20(4):513–521. 2003 DOI: 10.1093/molbev/msg085 Ó 2003 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038

ambiguous orthologs have not been reported in amniotes so far. The evolutionary origin of the highly divergent otdrelated Crx genes, which have been isolated in the rat, mouse, human, and ox, is another unresolved question. These genes have aroused much interest in the past few years both because of their essential roles in the differentiation of retinal photoreceptors and circadian entrainment (Furukawa et al. 1999) and because of their involvement in various hereditary retinal degenerations in humans (reviewed in Rivolta, Berson, and Dryja 2001). Phylogenetic constructs and site-by-site comparisons of protein sequences have led to the hypothesis that Crx genes may be the mammalian representatives of the Otx5 class (Germot et al. 2001; Sauka-Spengler et al. 2001). However, the hypothesis of a single Otx5/Crx class is contradicted by the isolation in zebrafish of both DrOtx5 and a putative ortholog of the mammalian Crx genes, DrCrx (Liu et al. 2001; Gamse et al. 2002). Whereas the zebrafish Otx5 gene shows a strongly supported sequence relationship with its amphibian counterparts (Gamse et al. 2002), the grouping of DrCrx with the mammalian Crx sequences is less clear. No monophyletic group containing these two sequences is retrieved in one study (Gamse et al. 2002), and another study shows only a poorly supported relationship (Sauka-Spengler et al. 2001). In the latter study, the suggestion has been raised that the grouping of the zebrafish and the mammalian Crx sequences in a monophyletic group may be artifactual, resulting from the high divergence of their protein sequences rather than from a true orthology relationship. Such an artifact could be caused by a phenomenon known as ‘‘long-branch 513

514 Plouhinec et al.

attraction,’’ commonly encountered when dealing with divergent sequences. Its resolution usually relies on the addition in the reconstructions of ‘‘intermediate’’ sequences, which emerge at the transition between slow-evolving and fast-evolving sequences (Philippe 2000). The origin of the mammalian Crx genes and their relationship with the Otx5 class are important issues in the understanding of their structural and functional evolution. They are also crucial for the use of relevant nonmammalian animal models in this field. To resolve these questions, we have performed an exhaustive search for both Otx5 and Crx genes in a wide range of species representative of the main amniote lineages (the lizard Gallotia stehlini; the turtle Emys orbicularis; the crocodile Crocodylus niloticus; the chick Gallus gallus; three marsupials, Monodelphis domestica, Macropus eugenii, and Isoodon macrourus; and four eutherian mammals, the sloth Bradypus tridactylus, the wild-cat Felis sylvestris, the pangolin Manis javanica, and the squirrel Sciurus vulgaris) and in the pufferfish Fugu rubripes genome database. Phylogenetic analysis of the coding sequences resolves the relationships among Otx5 and Crx genes, thus providing new insights into their structural and functional evolution. Materials and Methods Isolation of Otx5 and Crx Coding Sequences in Amniotes Fragment I (fig. 1A) was amplified and characterized as described in Germot et al. (2001). The 59 part of the coding region was amplified using a seminested semidegenerate RT-PCR approach. The 59 primer was 59CCNCAYTAYGCNGTNAA-39, which codes for PHYAVN (positions 9 to 14 [fig. 1B]), and specific 39 primers were chosen in the sequence previously determined for fragment I. Amplified fragments were subcloned and sequenced. Sequences were deduced from the consensus of at least three independent clones. As a control, both fragment I and the 59 part of the coding region were amplified in a single RT-PCR step and sequenced in each case. Database Searches Amphioxus and gnathostome Otx sequences were retrieved from GenBank as described in Germot et al. (2001), as well as the zebrafish sequences DrOtx5 (AAK62029) and DrCrx (AAM27445). The pufferfish F. rubripes FrOtx5a and FrOtx5b coding sequences were retrieved using the BLAST algorithm from http:// www.ncbi.nlm.nih.gov/BLAST/fugu.html, respectively, from genomic scaffolds 5499 (positions 3042 to 2943, 2161 to 2007, and 1917 to 1278) and 2954 (positions 2030 to 2126, 2882 to 3033, and 3209 to 3847). Blast searches in the mouse and human genomes were performed starting from the NCBI genome blast page at http://www. ncbi.nlm.nih.gov/BLAST. Phylogenetic Analysis The alignment and phylogenetic analyses were performed as described in Germot et al. (2001), except

for the following modifications. Neighbor-Joining (NJ) distances were computed by the method of Rzhetsky and Nei (1994), using the a value (c shape parameter) of the best maximum-likelihood (ML) tree. The likelihood, branch lengths, and RELL proportions of the 2,000 trees constructed with PROTML were recomputed using PAML (Yang 2000) with the JTT-f1ÿ model to obtain the best ML tree and the bootstrap proportions. For the tests of topologies, exhaustive sets of strictly dichotomous trees were generated from the multifurcating topologies used as hypotheses (see online Supplementary Material). These sets were compared using the Shimodaira-Hasegawa test implemented in PAML with the JTT-f1ÿ model (Shimodaira and Hasegawa 1999). Results and Discussion Characterization of Otx5 and Crx Sequences in Amniotes and in the Pufferfish To identify Otx5-related and Crx-related genes in the species studied, we first used the degenerate PCR strategy previously described, starting from genomic DNA (Germot et al. 2001). This strategy relies on the use of degenerate primers, which encode protein motifs conserved among all chordate Otx proteins, respectively, located in the homeodomain (VWFKNR and AKCRQQ) and at the C-terminal end of the molecule (WKFQVL). As expected, this led to the amplification of fragments ranging in size from 550 pb to 600 pb (fragment I [fig. 1A]). In all species studied, cloning and sequencing of 10 to 20 independent clones led to the characterization of either Otx5-related or Crx-related coding sequences, as well as Otx2 sequences (which were also obtained with the strategy used [data not shown]). In some of the species studied (G. stehlini, E. orbicularis, C. niloticus, G. gallus, and M. domestica), the characterization of full-length coding sequences could be almost completed in a second RT-PCR step, starting from eye cDNA. In these species, the sequenced fragment (fragment II [fig. 1A]) spans the full length of the coding sequence, with the exception of 13 to 14 N-terminal amino acids and six C-terminal amino acids. The deduced amino acid sequences obtained in the lizard, turtle, crocodile, and chick appear highly similar to the dogfish, reedfish, and zebrafish Otx5 sequences (respectively, 84%, 83%, 80%, and 73% identity with the dogfish ScOtx5 sequence versus 54%, 54%, 55%, and 51% with the human HsCrx sequence). These sequences were therefore termed GsOtx5, EoOtx5, CnOtx5, and GgOtx5, respectively. Among mammals, the sequences determined in eutherians (B. tridactylus, F. sylvestris, M. javanica, and S. vulgaris) show a close relationship to the rat, mouse, human, and ox Crx sequences previously characterized (respectively, 90%, 93%, 85%, and 99% identity with HsCrx but less than 43% in all cases with ScOtx5) and were termed accordingly (BtrCrx, FsCrx, MjCrx, and SvCrx). The sequences characterized in the marsupials M. domestica, M. eugenii, and I. macrourus are equally related to Otx5 and mammalian Crx protein sequences (respectively, 57%, 51%, and 49% with ScOtx5 and 56%, 51%, and 50% with HsCrx). They were

The Otx5/Crx Orthology Class in Gnathostomes 515

FIG. 1.—Comparison of gnathostome Otx5 and Crx amino acid sequences. (A) Amplification of gnathostome Otx5 and Crx sequences. The coding region is shown by an open bar, with the homeodomain (HD) in black and the location of introns indicated by vertical empty arrowheads. Thick black lines delineate the fragments characterized in this study and amplified either from cDNA (fragment II) or from genomic DNA (fragment I). Black arrowheads symbolize the degenerate amplification primers that were used. (B) Alignment of gnathostome Otx5 and Crx protein sequences. Dashes indicate residue identities with the dogfish Otx5 sequence (ScOtx5, first line), stars indicate missing residues, and dots indicate undetermined sequence data. The homeodomain is boxed, black arrows overline each copy of a 10 to 19 amino acid domain repeated at the C-terminal end of the molecule (domains A and B), and the Otx tail is underlined. Residues shaded in light gray are conserved among all Otx5/Crx proteins; residues shaded in dark gray correspond to conservations selectively observed in the Otx5/Crx class (but not in the Otx1 and Otx2 classes [also see fig. 5 in the online Supplementary Material]). Residues shown below the alignment correspond to sequence polymorphisms detected in human Crx sequences and showing a cosegregation with a retinopathy (white characters on a black background); are associated to a retinopathy without demonstrated cosegregation (white characters on a gray background); or are observed without reported mutant phenotype (black characters). Sequences determined in this study are underlined. Abbreviations: Bt: Bos taurus; Btr: Bradypus tridactylus; Cn: Crocodylus niloticus; Dr: Danio rerio; Ec: Erpetoichthys calabaricus; Eo: Emys orbicularis; Fr: Fugu rubripes; Fs: Felis sylvestris; Gg: Gallus gallus; Gs: Gallotia stehlini; Hs: Homo sapiens; Im: Isoodon macrourus; Md:

516 Plouhinec et al.

FIG. 2.—Phylogenetic trees showing the relationships among gnathostome Otx5 and Crx protein sequences over fragment I. The phylogenetic trees shown in (A), (B), and (C) were calculated using NJ, MP, and ML algorithms, respectively. Boxed numbers indicate the bootstrap values supporting the corresponding nodes. Only bootstrap values greater than 50% are shown. The scale bar represents in (A) the number of estimated differences, in (B) the number of steps, and in (C) the number of substitutions per position for a unit branch length. The NJ tree was constructed with a 5 1.7. The MP tree is the strict consensus of the eight most-parsimonious trees (length 5 1625; RI 5 0.703; CI 5 0.559). The likelihood of the best ML tree was lnL 5 28009.0. Same abbreviations as in figure 1. Bf: Branchiostoma floridae; Ci: Ciona intestinalis; Cs: Ciona savignyi; Hc: Herdmania curvata; Hr: Halocynthia roretzi; Lj: Lampetra japonica; Mg: Myxine glutinosa; Pm: Petromyzon marinus.

nevertheless termed MdCrx, MeCrx, and ImCrx because they share several unique features (A in position 15, QQP in positions 108 to110, AP in positions 225 to 226, F in position 231, P in position 311, as well as several insertion sites in positions 132 to 134, 208 to 218, and 235 to 237 [fig. 1B]) with eutherian Crx proteins. In addition, searches for similarities with Otx5 and Crx coding sequences in the pufferfish F. rubripes genome database revealed the presence of two Otx5-related coding sequences, showing, respectively, 75% and 65% identity with ScOtx5 but only 52% and 48% identity with HsCrx. These two sequences were termed FrOtx5a and FrOtx5b. Searches for Otx5related coding sequences were also performed in the human and mouse genome databases but only led to the identification of the previously characterized Otx1, Otx2, and Crx sequences and, in the human genome, identification of an Otx2-related pseudogene located on the chromosome 9 (contig NT 008580.10, positions 7636086 to 7636539) showing features of a processed retrogene (39 truncation, presence of three stop codons in the coding

sequence and of a poly (T) tail, and absence of introns [data not shown]). No canonical Otx5 sequence was thus retrieved from the mammalian genomes. Phylogenetic Analyses Support the Presence of a Single Otx5/Crx Orthology Class in Gnathostomes To study the phylogenetic relationships among Otx5 and Crx genes, we included all gnathostome Otx5 and Crx protein sequences in an alignment also containing the Otx1-related and Otx2-related sequences previously characterized in gnathostomes, the lamprey and hagfish sequences, the Amphioxus BfOtx sequence, and the ascidian HcOtx, HrOtx, CsOtx, and CiOtx (fig. 5 in the online Supplementary Material). The ascidian sequences were always found clustered in a monophyletic group, which was used as an outgroup. Phylogenetic analyses of these sequences were successively performed over fragment I (257 sites, 234 of which informative for parsimony [fig. 2]) and fragment II (345 sites, 257 informative for

The Otx5/Crx Orthology Class in Gnathostomes 517

FIG. 3.—Phylogenetic trees showing the relationships among gnathostome Otx5 and Crx protein sequences over fragment II. Same legend as in figure 2, except for the tree parameters. The NJ tree was constructed with a 5 0.7. The MP tree is the strict consensus of the five most parsimonious trees (length 5 1458; RI 5 0.693; CI 5 0.602). The likelihood of the best ML tree was lnL 5 27656.0.

parsimony [fig. 3]), using NJ, maximum-parsimony (MP), and ML algorithms. As in previous analyses, the relative branching orders of the genes identified in cyclostomes remain largely unresolved, but the Otx1 and Otx2 classes are retrieved with good confidence levels (Otx1: 50%, 71%, and 91% [fig. 2A–C] and 68%, 72%, and 89% [fig. 3A–C]; Otx2: 62%, 88%, and 100% [fig. 2A–C] and 73%, 91%, and 100% [fig. 3A–C]). In all the reconstructions, a monophyletic group containing all gnathostome Otx5 and Crx sequences is also recovered with strong statistical supports (bootstrap proportions [BP] 5 75%, 62%, and 98% [fig. 2A–C] and 91%, 76%, and 100% [fig. 3A–C] in the NJ, MP, and ML analyses, respectively), which fully confirms the previously reported relationship between Otx5 and Crx genes (Germot et al. 2001; Sauka-Spengler et al. 2001). Inside this class, the branching orders are largely similar in the two data sets used (fragments I and II) but can vary depending on the reconstruction method. However, three monophyletic groups are retrieved with good statistical supports in all six reconstructions. First, the marsupial and eutherian Crx sequences are always found clustered as two strongly supported sister-groups. This mammalian Crx monophyletic group is supported by very high bootstrap values (>90%) in all cases, confirming the assignment of all the mammalian sequences obtained in this study to the

Crx class. The phylogenetic relationships among eutherian sequences remain largely unresolved, except for the early emergence of the sloth BtrCrx sequence, which receives a good statistical support (BP 5 84%, 61%, and 100% in NJ, MP, and ML analyses, respectively [fig. 2]), in agreement with recently published phylogenies of mammals (Madsen et al. 2001; Murphy et al. 2001). The second monophyletic group, which is recovered whatever the data set and reconstruction method used, contains the mammalian Crx and the archosaurian (crocodile and chick) Otx5 sequences as sister-groups. This grouping is supported by high bootstrap values in ML (respectively, 96% and 90% [figs. 2C and 3C)], but lower values in NJ and MP (respectively, 56% and 41% [fig. 2A and B] and 71% and 45% [fig. 3A and B]). It is not unexpected in the hypothesis of a single Otx5/Crx class, since several molecular studies have led to a similar clustering of archosaurs and mammals (Hedges, Moberg, and Maxson 1990). However, the phylogeny of amniotes remains largely unresolved (Zardoya and Meyer 1998) and in this study, the statistical support for a group containing archosaurs and mammals remains relatively low, especially in NJ and MP analyses. Finally, a third monophyletic group, containing the zebrafish Otx5 and Crx sequences as well as the two sequences retrieved from the pufferfish genome database

518 Plouhinec et al.

(FrOtx5a and FrOtx5b), emerges in all our tree reconstructions with very good statistical supports in NJ and ML analyses (respectively, 80% and 99% [fig. 2A and C] and 77% and 100% [fig. 3A and C]) but a lower one in MP (respectively, 58% and 65% [figs. 2B and 3B]). Inside this group, the zebrafish Crx and the pufferfish Otx5b sequences appear closely related (BP 5 77% and 87% in MP [figs. 2B and 3B]; BP . 96% in all other reconstructions), which strongly supports an orthology relationship between these genes. The clustering of the four teleostean genes indicates that they have arisen through duplication events, which have occurred among actinopterygians, after the splitting of the Cladistia (the earliest emerging group in actinopterygians, represented in our phylogenetic reconstructions by the reedfish EcOtx5 sequence). In line with this conclusion, an increasing amount of evidence suggests that massive gene duplications have occurred in this taxon, even though their chronology remains elusive (Robinson-Rechavi et al. 2001). In order to directly compare the hypothesis of a single Otx5/Crx class in gnathostomes to the one recently proposed (two distinct gnathostome Otx5 and Crx orthology classes [Gamse et al. 2002]), we generated two exhaustive sets of strictly dichotomous trees, both consistent with widely accepted phylogenetic relationships among gnathostomes, and each representing one of these two scenarios (see fig. 6 in the online Supplementary Material). Since the lamprey and hagfish sequences display highly variable branching orders in the tree reconstructions, only gnathostome Otx sequences were included in this analysis. The Amphioxus Otx sequence was used as an outgroup. These two sets were tested using the Shimodaira-Hasegawa (SH) method (Yang 2000). The best ML tree obtained in the hypothesis of distinct gnathostome Otx5 and Crx classes (lnL 5 24815.8) was rejected with a very high probability (P-value of the SH test 5 20.002) in favor of the best ML tree obtained in the hypothesis of a single Otx5/Crx class (lnL 5 24773.6). The SH test thus strongly rejects the hypothesis of two distinct Otx5 and Crx classes. The Otx Tail Is Repeated in Crx, as in Otx5 Proteins, but One Copy Is Degenerate in Zebrafish and Eutherian Mammals In all deuterostomes, otd-related proteins share a 12 to 13 amino acid motif known as the Otx tail close to their C-terminal end (Furukawa, Morrow, and Cepko 1997) (fig. 1B). This motif plays an important role in Crx proteins, since its deletion results in a substantially decreased (albeit not abolished) transactivation of photoreceptor-specific genes by Crx (Chau et al. 2000). It is part of a longer conserved domain that displays a tandem duplication in all gnathostome Otx1, Otx2, and Otx5 as well as in lamprey and hagfish Otx proteins (domains A and B in fig. 1A). In contrast, a single copy is present in protochordates and echinoderms. The presence of this repeat is thus an important phylogenetic hallmark, which provides strong evidence that craniate Otx genes have arisen through the duplication of a single ancestral gene,

after the splitting of cephalochordates (Williams and Holland 1998; Germot et al. 2001). However, its presence in Crx proteins remains controversial. We have previously proposed that two copies of the Otx tail are present in eutherian Crx proteins, although the first one (positions 304 to 316) is degenerate (Germot et al. 2001). This hypothesis is fully confirmed by the marsupial Crx sequences obtained in this study, since they display almost perfect tandem repeats that can be readily aligned with their eutherian counterparts (domains A and B in fig. 1B). Similarly, despite a broad deletion in its C-terminal region (positions 328 to 335), the N-terminal residues (positions 317 to 327) of domain B, which contain a truncated second copy of the Otx tail (positions 324 to 327), can be recognized in the zebrafish Crx sequence. Protein sequence analysis of its putative pufferfish ortholog FrOtx5b, which contains both domain A and domain B, with full-length copies of the Otx tail (fig. 1B), supports this conclusion. Taken together, these data indicate that the C-terminal repeat characterized in all craniate Otx proteins is also an ancestral feature of Crx proteins, which has been partially lost in eutherians and in some teleosts through independent genetic events. This result provides an additional argument in favor of a recent origin of Otx5/ Crx genes, from a gene duplication event that has occurred after the splitting of cephalochordates but before the gnathostome radiation. Whether the degenerate copy of the Otx tail could contribute to the residual transactivation activity observed in mouse Crx proteins after the deletion of the canonical second copy remains to be assessed (Chau et al. 2000). Structural Evolution of Otx5/Crx Protein Sequences in Gnathostomes The comparative analysis of Otx5 and Crx sequences among a wide range of gnathostomes provides a mean to accurately identify the constraints acting on their primary structure. Residues, which are strictly conserved among all Otx5 and Crx proteins, can be identified inside (49 residues) as well as outside (46 residues) the homeodomain, suggesting an involvement not only in protein/DNA contacts but also in protein/protein contacts (fig. 1B). Most of these residues are also conserved in the other orthology classes, Otx1 and Otx2. In an attempt to identify constraints specifically acting on the Otx5/Crx class, we first searched residues, which are selectively conserved among the canonical Otx5 proteins, characterized in fish (ScOtx5, EcOtx5, and FrOtx5a), amphibians (PwOtx5, XlOtx5a, and XlOtx5b) and amniotes (GgOtx5, CnOtx5, GsOtx5, and EoOtx5). Ten such residues were found (H in position 10, A in position 32, YS in positions 152 to 153, TP in positions 164 to 165, L in position 248, S in position 255, G in position 277, and V in position 303 [fig. 1B]). They correspond either to a conserved residue of a different identity (V and M in Otx1 and Otx2, respectively, instead of L in position 248, P instead of H in position 10, S instead of A in position 32, and FT instead of YS in positions 152 to 153) or to an unconstrained position in Otx1 and Otx2 proteins (H, P, Q, S, and T instead of T in position 164; A, S, T, and V instead of P in position 165;

The Otx5/Crx Orthology Class in Gnathostomes 519

FIG. 4.—Functional evolution of Otx genes in chordates. The phylogeny shown is a consensus, but the basal position of Urochordates and the monophyly of the Cyclostomes remain controversial. The major changes in the structure or the function of Otx genes suggested by this study, Germot et al. 2001, and Sauka-Spengler et al. 2001, are indicated, together with the approximate time period when they have taken place (thick bars).

A, G, S, and T instead of S in position 255; A, P, S, and T instead of G in position 277; and A, S, and T instead of V in position 303 [see fig. 5 in the online Supplementary Material]). This number is low relative to the ones observed in the other classes (27 for the Otx1 class, 33 for the Otx2 class [see fig. 5 in the online Supplementary Material]), possibly due to the large number of species considered and the presence of relatively divergent sequences, such as the chick GgOtx5 sequence. However, the majority of these residues are strictly conserved among the mammalian Crx sequence (nine out of ten [fig. 1]). Most of them are also found in the zebrafish Crx sequence and its ortholog FrOtx5b (six out of ten). This result provides an additional support for the orthology relationship between Otx5 and Crx genes and suggests that the residues thus identified might correspond to constraints selectively acting on Otx5 and Crx proteins. The high divergence observed between the canonical Otx5 proteins characterized in chondrichthyans, amphibians, sauropsids, and their mammalian Crx counterparts raises the possibility that different structural constraints

may act in these groups of species. In line with this possibility, a number of residues, which are highly conserved in Otx5 but also in Otx1 and Otx2 proteins, appear to vary or change identity among mammalian Crx proteins (for instance, positions 15, 22 to 23, 113 to 114, 182, 226, 231, and 233). This suggests that a relaxation in the structural constraints acting on Otx5/Crx proteins may have occurred early during mammalian evolution. Whether novel, group-specific constraints may have been fixed among mammals remains difficult to address formally but is suggested by the relative homogeneity of the sequences observed, for instance, among eutherians. The biological significance of the constraints acting on Otx5/Crx proteins and of their evolution in mammals remains to be addressed by detailed analyses of structurefunction relationships, and their understanding at the molecular level will require an exhaustive characterization of the interactions, either protein-protein or DNA-protein, involving this family of transcription factors. In the absence of extensive mutational analyses, human polymorphisms corresponding to missense mutations could

520 Plouhinec et al.

provide important insights into the biological significance of the constraints acting on Crx primary structure. However, all of those, which show a clear cosegregation with retinopathies, have been found in the homeodomain (positions 44, 83, and 93 [fig. 1B]), and more exhaustive polymorphism data are clearly needed for a more comprehensive comparison with the conservation map generated by the comparative analysis. Conclusion Our results provide strong evidence that the mammalian Crx genes are orthologous to the Otx5 genes characterized in chondrichthyans, actinopterygians, and amphibians. They also indicate that the presence of two genes in zebrafish and pufferfish results from an independent duplication event in actinopterygians. Finally, the high divergence displayed by the mammalian Crx genes implies that Otx5/Crx genes have undergone a very high evolutionary rate early in the mammalian lineage. These results have important implications as to the functional evolution of Otx5/Crx genes (fig. 4). First, they lead to a reinterpretation of the functional characterization of the zebrafish paralogs DrOtx5 and DrCrx. In this species, DrOtx5 has been shown to regulate circadian gene expression, whereas DrCrx can only transactivate retinal photoreceptor-specific genes (Liu et al. 2001; Gamse et al. 2002). Considering our results, this divergence among paralogs may correspond to a recent partition of functions, which were both fulfilled by the Otx5/Crx ancestral gene before its duplication in the actinopterygian lineage. Such processes are likely to play an important part in the functional evolution of vertebrate multigene families (Force et al. 1999). In line with this possibility, the Otx5 genes characterized in the dogfish and Xenopus, as well as their mammalian counterparts, share prominent retinal and pineal expression domains, which strongly suggests an ancestral role in both organs. Second, they raise new questions as to the biological significance of the high sequence divergence displayed by the mammalian Crx genes. Comparisons of expression patterns suggest that it could be linked to the loss of some ancestral functions. For instance, the dogfish and amphibian Otx5 genes are both expressed in gastrulating embryos as well as in the retinal pigmented epithelium (Vignali et al. 2000; Sauka-Spengler et al. 2001; Sauka-Spengler and Mazan [unpublished data]). In contrast, no Crx expression has been reported in these domains among mammals, suggesting the loss of these ancestral expression territories in this taxon. Another possibility is suggested by expression patterns and functional analyses, which point to an important ancestral role of Otx5/Crx genes in the adult and embryonic pineal gland, possibly in the regulation of circadian expressed genes or in the differentiation of pinealocytes. The loss of a direct sensitivity to light in the epiphysis of adult mammals may have been linked to substantial changes in these roles, as well as in the corresponding molecular interactions, which could also account for the high divergence rate of Crx genes in mammals. In this respect, it is noticeable that relatively high divergence rates are also observed for Otx5 in the chick, a species that has lost

functional photoreceptors in the pineal gland, and for the zebrafish DrCrx gene, which, as indicated above, may have lost some ancestral Otx5 roles (fulfilled by DrOtx5) in this organ. Expression pattern comparisons in a wide range of amniotes will help to test these hypotheses, but a deeper understanding of the structural and functional evolution of Otx5/Crx genes will require an exhaustive characterization of the molecular interactions involved in their different functions in the eye and in the pineal gland. The comparative analysis reported here should provide a strong structural basis for further analyses of structurefunction relationships. Acknowledgments This work was supported by grants from the Centre National de la Recherche Scientifique and the Universite´ Paris-Sud, by Ministe` re de la Recherche et de la Technologie doctoral fellowships to J.L.P. and T.S.S. and a Fe´de´ration des Aveugles et Handicape´s Visuels award to J.L.P., and by a Fondation de la Recherche Me´dicale postdoctoral fellowship to A.G. We thank A. Viczian, L. Coutte, and D. Casane for critical reading of the manuscript and M. Pradels and N. Narradon for their technical support. We are grateful to L. Fougeirol (Ferme aux crocodiles, Pierrelatte, France) for giving us free access to living growth series of C. niloticus. The Fugu data has been provided freely by the Fugu Genome Consortium for use in this publication only. Literature Cited Acampora, D., M. Gulisano, and A. Simeone. 2000. Genetic and molecular roles of Otx homeodomain proteins in head development. Gene 246:23–35. Chau, K. Y., S. Chen, D. J. Zack, and S. J. Ono. 2000. Functional domains of the cone-rod homeobox (CRX) transcription factor. J. Biol. Chem. 275:37264–37270. Force, A., M. Lynch, F. B. Pickett, A. Amores, Y. L. Yan, and J. Postlethwait. 1999. Preservation of duplicate genes by complementary, degenerative mutations. Genetics 151:1531– 1545. Furukawa, T., E. M. Morrow, and C. L. Cepko. 1997. Crx, a novel otx-like homeobox gene, shows photoreceptorspecific expression and regulates photoreceptor differentiation. Cell 91:531–541. Furukawa, T., E. M. Morrow, T. Li, F. C. Davis, and C. L. Cepko. 1999. Retinopathy and attenuated circadian entrainment in Crx-deficient mice. Nat. Genet. 23:466–470. Gamse, J. T., Y. C. Shen, C. Thisse, B. Thisse, P. A. Raymond, M. E. Halpern, and J. O. Liang. 2002. Otx5 regulates genes that show circadian expression in the zebrafish pineal complex. Nat. Genet. 30:117–121. Germot, A., G. Lecointre, J. L. Plouhinec, C. Le Mentec, F. Girardot, and S. Mazan. 2001. Structural evolution of Otx genes in craniates. Mol. Biol. Evol. 18:1668–1678. Hedges, S. B., K. D. Moberg, and L. R. Maxson. 1990. Tetrapod phylogeny inferred from 18S and 28S ribosomal RNA sequences and a review of the evidence for amniote relationships. Mol. Biol. Evol. 7:607–633. Kuroda, H., T. Hayata, A. Eisaki, and M. Asashima. 2000. Cloning a novel developmental regulating gene, Xotx5: its potential role in anterior formation in Xenopus laevis. Develop. Growth. Differ. 42:87–93.

The Otx5/Crx Orthology Class in Gnathostomes 521

Liu, Y., Y. C. Shen, J. S. Rest, P. A. Raymond, and D. J. Zack. 2001. Isolation and characterization of a zebrafish homologue of the cone rod homeobox gene. Invest. Ophtalmol. Visual. Sci. 42:481–487. Madsen, O., M. Scally, C. J. Douady, D. J. Kao, R. W. DeBry, R. Adkins, H. M. Amrine, M. J. Stanhope, W. W. de Jong, and M. S. Springer. 2001. Parallel adaptive radiations in two major clades of placental mammals. Nature 409:610–614. Murphy, W. J., E. Eizirik, W. E. Johnson, Y. P. Zhang, O. A. Ryder, and S. J. O’Brien. 2001. Molecular phylogenetics and the origins of placental mammals. Nature 409:614–618. Philippe, H. 2000. Opinion: long branch attraction and protist phylogeny. Protist 151:307–316. Rivolta, C., E. L. Berson, and T. P. Dryja. 2001. Dominant Leber congenital amaurosis, cone-rod degeneration, and retinitis pigmentosa caused by mutant versions of the transcription factor Crx. Hum. Mut. 18:488–498. Robinson-Rechavi, M., O. Marchand, H. Escriva, P. L. Bardet, D. Zelus, S., Hughes, and V. Laudet. 2001. Euteleost fish genomes are characterized by expansion of gene families. Genome Res. 11:781–788. Rzhetsky, A., and M. Nei. 1994. Unbiased estimates of the number of nucleotide substitutions when substitution rate varies among different sites. J. Mol. Evol. 38:295–299. Sauka-Spengler, T., B. Baratte, D. L. Shi, and S. Mazan. 2001. Structure and expression of an Otx5-related gene in the dogfish Scyliorhinus canicula: evidence for a conserved role

of Otx5 and Crx genes in the specification of photoreceptors. Dev. Genes Evol. 211:533–544. Sauka-Spengler, T., A. Germot, D. L. Shi, and S. Mazan. 2002. Expression patterns of an Otx2 and an Otx5 orthologue in the pleurodeles P. waltl: implications on the evolutionary relationships between the balancers and cement gland in amphibians. Dev. Genes Evol. 212:380–387. Shimodaira, H., and M. Hasegawa. 1999. Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Mol. Biol. Evol. 16:1114–1116. Vignali, R., S. Colombetti, G. Lupo, W. Zhang, S. Stachel, R. M. Harland, and G. Barsacchi. 2000. Xotx5b, a new member of the Otx gene family, may be involved in anterior and eye development in Xenopus laevis. Mech. Dev. 96:3–13. Williams, N., and P. W. H. Holland. 1998. Gene and domain duplication in the chordate Otx gene family: insights from amphioxus Otx. Mol. Biol. Evol. 15:600–607. Yang, Z. 2000. Phylogenetic analysis by maximum likelihood (PAML), Version 3.0. University College London, London, England. Zardoya, R., and A. Meyer. 1998. Complete mitochondrial genome suggests diapsid affinities of turtles. Proc. Natl. Acad. Sci. USA 95:14226–14231.

Diethard Tautz, Associate Editor Accepted November 5, 2002