EVOLUTION & DEVELOPMENT 8:5, 394–406 (2006) Is retinoic acid genetic machinery a chordate innovation? Cristian Can˜estro,a John H. Postlethwait,a Roser Gonza`lez-Duarte,b and Ricard Albalatb,Ã aInstitute of Neuroscience, University of Oregon, Eugene, OR 97403, USA bDepartament de Gene`tica, Universitat de Barcelona, Av. Diagonal 645, 08028 Barcelona, Spain ÃAuthor for correspondence (email: [email protected]) SUMMARY Development of many chordate features and showed for the first time that RA genetic machineryF depends on retinoic acid (RA). Because the action of RA that is Aldh1a, Cyp26, and Rar orthologsFis present in during development seems to be restricted to chordates, it had nonchordate deuterostomes. This finding implies that RA been previously proposed that the ‘‘invention’’ of RA genetic genetic machinery was already present during early machinery, including RA-binding nuclear hormone receptors deuterostome evolution, and therefore, is not a chordate (Rars), and the RA-synthesizing and RA-degrading enzymes innovation. This new evolutionary viewpoint argues against Aldh1a (Raldh) and Cyp26, respectively, was an important the hypothesis that the acquisition of gene families under- step for the origin of developmental mechanisms leading lying RA metabolism and signaling was a key event for to the chordate body plan. We tested this hypothesis the origin of chordates. We propose a new hypothesis in by conducting an exhaustive survey of the RA machinery which lineage-specific duplication and loss of RA machinery in genomic databases for twelve deuterostomes. We genes could be related to the morphological radiation of reconstructed the evolution of these genes in deuterostomes deuterostomes. INTRODUCTION which appears to be the sister group of chordates (Cameron et al. 2000, but see Delsuc et al. 2006). To differentiate orthologs The origin of chordates and their innovative body plan re- from paralogs, we investigated the phylogenetic relationships mains controversial (Holland 2005b; Delsuc et al. 2006). Be- of deuterostome genes implicated in the evolution of the RA cause the action of retinoic acid (RA) in patterning embryonic machinery and closely related families. Understanding the axes seems to be restricted to chordates, it had been proposed evolution of Aldh1a had been obscure, due to poorly sup- that the morphogenetic role of RA was a chordate novelty ported gene phylogenies, confusion with the closely related linked to the origin of chordate-specific features (Shimeld Aldh2 (Fujiwara and Kawamura 2003), and the small number 1996; Manzanares et al. 2000; Schilling and Knight 2001; of taxa in which Aldh1a hadbeenreported(onlyafew Wada 2001; Holland 2005a). Vertebrates regulate RA action vertebrates and ascidians). To illuminate Aldh1a evolution, at two levels: metabolism and signaling (Fig. 1). Machinery we identified 73 Aldh1a-related genes in publicly available governing RA metabolism includes the RA-synthesizing en- genome/EST databases of seven vertebrates, one cephalo- zymes (retinaldehyde dehydrogenases Aldh1a, formerly chordate, three urochordates, and two nonchordate deuteros- Raldh) and the RA-degrading enzymes (Cyp26), which to- tomes. In this work we reveal the presence of the three main gether regulate the spatio-temporal distribution of RA during components of RA machinery, Aldh1a, Cyp26,andRar,in embryogenesis (Niederreither et al. 2002; Reijntjes et al. 2005). nonchordate animals, revealing for the first time that these Machinery for RA-signaling includes the RA-binding nuclear genes were not a chordate innovation. hormone receptors (Rars), which mediate RA action on target genes (e.g., Hox genes) (Marshall et al. 1994). Because Aldh1a, Cyp26 and Rar had been described only in chordates, it had been proposed that the acquisition of these gene families was MATERIALS AND METHODS a key step for the innovation of the chordate body plan (re- Sequence analysis and identification of new genes viewed in Fujiwara and Kawamura 2003). This hypothesis Sequences used in this work were assembled from data obtained by predicts that Aldh1a, Cyp26,andRar genes should not be in silico screening of public databases (accession numbers and da- found outside the Chordata. tabase URLs used in this work are provided in Table A1 in the To test this hypothesis, we searched for RA genetic ma- appendix). Human reference proteins from each analyzed gene chinery in Ambulacraria (echinoderms plus hemichordates), family were used as starting queries for BLAST searches (Altschul 394 & 2006 The Author(s) Journal compilation & 2006 Blackwell Publishing Ltd. Canìestro et al. Retinoic acid and chordate origins 395 A B CD Fig. 1. (A) Retinoic acid (RA) genetic machinery regulates RA action at two levels: metabolism and signaling. Aldh1a (red) and Cyp26 (blue) regulate the spatio-temporal distribution of RA. Heterodimers of Rar and Rxr (green and gray, respectively) mediate RA signaling to RA-target genes (e.g., Hox). In contrast to Rar, Rxr can heterodimerize with other nuclear receptors, and its presence in protostomes suggests a more ancient origin (reviewed in Escriva et al. 2000). (B) Gene phylogenies corroborate the orthology of nonchordate Aldh1a, Cyp26, and Rar proteins (colored). Consistent with the presence of Rar in sea-urchin, we also found a sea-urchin Rxr (in bold). Tree branch lengths correspond to neighbor-joining distances, and numbers are the bootstrap values supporting each node (n 5 1000; poorly supported nodes o50% were collapsed). The same tree topologies were supported by maximum-likelihood and maximum-parsimony methods. (C) The finding of Aldh1a, Cyp26,andRar orthologs in Ambulacraria suggests a new evolutionary scenario, in which the RA genetic machinery was already present before the divergence of extant deuterostomes, and consequently, it is not a chordate innovation. (D) The Aldh1a phylogeny illustrates taxon-specific variation of RA-related gene families caused by independent gene duplication and loss during de- uterostome evolution. To understand the evolution of the Aldh1a subfamily, it was necessary to include in the analysis the next two most related Aldh families (Aldh2 and Aldh1l), and to consider exon–intron organization. The putative Aldh1a2 described in ascidians (Fujiwara and Kawamura 2003) has been renamed here as Aldh1a1/2/3a to reflect its phylogenetic affinities. The position of the urochordate Aldh1a1/ 2/3 cluster close to the Aldh1l family is probably distorted by an artifact due to long branch attraction. Vertebrates: Hs, Homo sapiens;Mm, Mus musculus;Rn,Rattus novergicus;Gg,Gallus gallus;Xt,Xenopus tropicalis;Dr,Danio rerio; and Tr, Takifugu rubripes. Cephalo- chordates: Bf, Branchiostoma floridae. Urochordates: (larvaceans) Od, Oikopleura dioica; (ascidians) Ci, Ciona intestinalis and Cs, Ciona savignyi; Hemichordates: (acorn-worm) Sk, Saccoglossus kowalevskii; Echinoderms: (sea-urchin) Sp, Strongylocentrotus purpuratus. et al. 1997) against EST and genomic databases. The orthology of From nonassembled genomes, Aldh genes were deduced by assem- proteins was deduced initially by reciprocal best BLAST searches bling cDNA and genomic contigs from 1009 trace sequences and against human genome GenBank database (Wall et al. 2003). 108 ESTs from NCBI. Gene structures and protein sequences were 396 EVOLUTION&DEVELOPMENT Vol. 8, No. 5, September^October 2006 deduced after merging the genomic sequences with ESTs when plete genome sequence for this species, we cannot discard the available, or by comparison with well characterized Aldh, Cyp26, presence of Rar in this hemichordate. Orthologies of the and Rar genes described in other species. Predicted genes from newly identified genes were strongly supported by sequence automatically annotated genomes were verified by eye, and errors similarity (Tables A2 and A3) and by reciprocal BLAST in automatic annotations were corrected to maximize the similarity against the human databases. In the case of the sea-urchin with ESTs when available, and other known enzymes. The partially Rar, the most significant BLAST hits were against the RARs predicted Strongylocentrotus purpuratus Rar protein sequence was (E-value of 9e-117) and the second next hit was against the completed by in silico genomic walking over 70 sequence traces. Zebrafish Aldh1a3 protein was initially predicted from putative THRB (E-value substantially lower, at 4e-69). In the case of exons inappropriately assembled by the Ensembl Zv5 database into the acorn-worm and sea-urchin Cyp26 proteins, the most three nonoverlapping genomic contigs (Table A1). To verify our significant BLAST hits in the human genome were the human zebrafish Aldh1A3 prediction, we amplified embryonic cDNA CYP26s (E-values were 2e-98 and 9e-72, respectively) and the by PCR and cloned a cDNA containing the complete coding next hits were against the human CYP51 and CYP3A, sequence. respectively (E-values were much lower, at 2e-26 and 9e-21, respectively). Gene phylogenies inferred by maximum Phylogenetic analysis likelihood and maximum parsimony (data not shown) Protein sequence alignments were generated with clustalX (Thomp- showed the same tree topologies as the neighbor-joining tree son et al. 1997) and corrected by eye. Only conserved parts of the (Fig. 1B) and thus corroborated the Rar and Cyp26 proteins, whose alignments were unambiguous among paralogs, orthologies inferred by the reciprocal best hit method were considered for the phylogenetic analysis: from codon I40 to (Wall et al. 2003). The next closest related families accord-