It

Ra

b

a

ARRAA

KRDIPELARBC

1

nipoca(opp[

NR

(

0d

Chemico-Biological Interactions 178 (2009) 188–196

Contents lists available at ScienceDirect

Chemico-Biological Interactions

journa l homepage: www.e lsev ier .com/ locate /chembio int

dentification of Aldh1a, Cyp26 and RAR orthologs in protostomes pushes backhe retinoic acid genetic machinery in evolutionary time to the bilaterian ancestor

icard Albalata,∗, Cristian Canestrob,∗

Departament de Genètica, Facultat de Biologia, Universitat de Barcelona, Av. Diagonal, 645, E-08028 Barcelona, SpainInstitute of Neuroscience, University of Oregon, Eugene, OR 97403, USA

r t i c l e i n f o

rticle history:eceived 8 August 2008eceived in revised form 5 September 2008ccepted 9 September 2008vailable online 24 September 2008

eywords:etinoic acid signalingeuterostomes

nvertebrates

a b s t r a c t

In vertebrates, retinoic acid (RA) is an important morphogenetic signal that controls embryonic devel-opment, as well as organ homeostasis in adults. RA action depends on the function of the RA-geneticmachinery, which includes a metabolic module and a signaling module. The metabolic module regulatesthe spatiotemporal distribution of RA by the combined action of the RA-synthesizing Aldh1a enzymes,and the RA-degrading Cyp26 enzymes. The signaling module includes members of the nuclear hormonereceptors family RAR and RXR, and controls the transcriptional state of RA-target genes. RA-signaling hasbeen described primarily in chordates, but the recent finding of elements of the RA-genetic machineryin non-chordate deuterostomes has changed our perspective on the evolutionary origin of this morpho-genetic signal, challenging previous assumptions that related the invention of the RA-genetic machinery

rotostomescdysozoansophotrocozoansldhetinaldehyde dehydrogenasesilaterian evolution

with the origin of the chordate body plan. To illuminate the evolutionary origin of the RA machinery wehave conducted an extensive survey of Aldh1a, Cyp26 and RAR orthologs in genomic databases of 13 non-deuterostome metazoans. Our results show for the first time the presence of Aldh1a, Cyp26 and RAR inprotostomes, which implies that the components of the RA machinery may be ancient elements of animalgenomes, already present in the last common ancestor of bilaterians. Interestingly, our data also revealthat independent losses of the RA toolkit have occurred multiple times during animal evolution, which

r the

pshCcoaias

omparative genomicsmay have been relevant fo

. Introduction

Retinoic acid (RA), a bioactive derivative of vitamin A, is a sig-aling molecule that regulates the expression of genes involved

n the development of vertebrate embryos, as well as in numeroushysiological processes in adults (reviewed in [1]). In humans, as inther vertebrates, the alteration of the RA-signaling system causesongenital malformations, fertility problems and vision defects,nd can lead to tumorigenesis and neurodegenerative disordersreviewed in [2]). For instance, an excess of RA during embry-

nic development induces homeotic transformations in the axialatterning and changes the fate of anterior structures into moreosterior ones, a phenotype known as RA-induced posteriorization3], which has been hitherto reported only in species of the chordate

Abbreviations: Aldh, aldehyde dehydrogenase; Cyp, cytochrome P450 enzyme;R, nuclear hormone receptor; RA, all-trans retinoic acid; RAR, retinoic acid receptor;XR, retinoid X receptor; RARE, retinoic acid response element.∗ Corresponding author.

E-mail addresses: ralbalat@ub.edu (R. Albalat), cristian@uoregon.eduC. Canestro).

gpbd

mtpraet

009-2797/$ – see front matter © 2008 Elsevier Ireland Ltd. All rights reserved.oi:10.1016/j.cbi.2008.09.017

evolution and developmental diversity of the current metazoan lineages.© 2008 Elsevier Ireland Ltd. All rights reserved.

hylum. Because the genetic machinery that mediates the RA-ignaling had previously been described only in chordates, it wasypothesized that the invention of the RA machinery (i.e., Aldh1a1,yp26 and RAR) was an evolutionary milestone in the origin of thehordate body plan (reviewed in [4,5]). Recent extensive surveysf genomic databases, however, have challenged this hypothesisfter revealing the presence of the key players of the RA machineryn non-chordate deuterostomes [6–8]. Interestingly, phylogeneticnalysis showed that the RA-genetic machinery diversified sub-tantially in different animal lineages due to recurrent extensiveene duplications and losses [6]. Such diversification raised theossibility that interspecific differences in the RA machinery coulde related to morphological and developmental changes in extanteuterostomes [6].

The RA machinery consists of two basic modules: a metabolicodule and a signaling module. The metabolic module controls

he spatiotemporal levels of RA by the coordinated action of RA-

roducing and RA-degrading enzymes. This module includes theetinaldehyde dehydrogenase enzymes (Aldh1a, formerly Raldh)nd the cytochrome P450 enzymes Cyp26 [9]. Vertebrate Aldh1anzymes are cytosolic proteins that catalyze the irreversible oxida-ion of retinal to its carboxylic acid, retinoic acid (reviewed in [10]).

ologica

VataAedohmmbwtie

gewltt(drbascttmbts

nRlo1dntsarta(if

2

2

Rsai(m

pp1m(easTgioT

rdtwmomsdtEi(c

2

OaISrfnpsulatcs

3

3

riiol4e

R. Albalat, C. Canestro / Chemico-Bi

ertebrates have three Aldh1a paralogs, named Aldh1a1, Aldh1a2nd Aldh1a3, with the exception of rodents that have an addi-ional Adh1a enzyme (Aldh1a4 in rat, and Aldh1a7 in mouse) [11]nd teleosts, which appear to have only two forms (Aldh1a2 andldh1a3) [6,12]. Outside vertebrates, an extensive catalog of Aldh1anzymes has recently been identified in cephalochordates, urochor-ates and hemichordates, showing for the first time the presencef the Aldh1a family in non-chordate organisms [6]. The Cyp26ydroxylases of the cytochrome P450 superfamily are endoplas-ic reticulum enzymes that transform RA to biologically inactiveetabolites such as 4-hydroxy-RA and 4-oxo-RA. Typically, verte-

rates have three Cyp26 enzymes, Cyp26A1, Cyp26B1 and Cyp26C1,hich are responsible for catabolism of RA in embryonic and adult

issues [13–16]. Outside vertebrates, Cyp26 enzymes have beendentified in cephalochordates, urochordates, hemichordates andchinoderms [6,17].

The RA-signaling module is responsible for the regulation ofene expression in a ligand-dependent mode. RA activates het-rodimers of a retinoic acid receptor/retinoid X receptor (RAR/RXR),hich bind to retinoic acid response elements (RARE) in the regu-

atory regions of direct target genes. RAR and RXR are members ofhe nuclear hormone receptor (NR) superfamily of ligand-activatedranscription factors [18]. Vertebrates possess three different RAR�, � and �) and three RXR (�, � and �) whereas cephalochor-ates and urochordates have a single representative of each type ofeceptor [6–8,17–20]. The only non-chordate RAR so far identifiedelongs to the sea urchin Strongylocentrotus purpuratus [6–8,21],nd no RAR orthologs have been reported in any protostomepecies. Conversely, orthologs of the RXR, a promiscuous receptorapable of forming heterodimers with many other NR such as thehyroid hormone receptor (THR), the vitamin D receptor (VDR) orhe ecdysone receptor (EcR), have been identified in nearly all ani-

als analyzed. Therefore, in contrast to the RAR that specificallyinds RA to mediate gene transcription regulation, the identifica-ion of RXR is not diagnostic of the presence of an RA-signalingystem in a given organism.

The discovery of Aldh1a, Cyp26 and RAR homologues inon-chordate deuterostomes [6] led us to wonder whether theA-genetic machinery was an innovation of the deuterostome

ineage or, alternatively, whether it had a deeper evolutionaryrigin. To answer this question, in this work we have screened3 metazoan genomes for Aldh1a, Cyp26 and RAR orthologs as aiagnostic feature of the presence of a RA-genetic machinery inon-deuterostome animals. Our results show for the first timehe presence of the three main components of RA machinery inome protostome lineages. This finding pushes back in evolution-ry time our understanding of the origins of the RA machinery andeveals that it was already present in the last common ancestor ofhe bilaterians. Interestingly, independent losses of the RA toolkitppear to have happened in the two main branches of protostomesi.e., lophotrochozoans and ecdysozoans), supporting the possibil-ty that the diversification of the RA-genetic machinery could haveavored the evolution and the developmental diversity of animals.

. Materials and methods

.1. Sequence analyses and gene identification

Human ALDH1A, ALDH2, ALDH1L, CYP26A, CYP51, CYP4V2,ARA, RXRA and THRA, and D. melanogaster Eip75B and EcR protein

equences were used as starting queries for TBLASTN searches [22]gainst public genomic and EST databases of 13 animal species,ncluding lophotrochozoans such as the annelids Capitella sp.coverage 7.9-fold) and Helobdella robusta (coverage 7.9-fold), the

ollusc Lottia gigantea (coverage 8.9-fold); ecdysozoans such as the

Attcg

l Interactions 178 (2009) 188–196 189

latyhelminth Schistosoma mansoni (coverage 7.0-fold), the arthro-ods Anopheles gambiae (African malaria mosquito) (coverage0.2-fold), Apis mellifera (honeybee) (coverage 7.5-fold), Drosophilaelanogaster (fruit fly), Tribolium castaneum (red flour beetle)

coverage 7.3-fold) and Daphnia pulex (common water flea) (cov-rage 8.7-fold), the nematodes Brugia malayi (coverage 9.0-fold)nd Caenorhabditis elegans; and the cnidarian Nematostella vecten-is (starlet sea anemone) (coverage 7.8-fold) and the placozoanrichoplax adhaerens (coverage 8.1-fold). Automatically annotatedenes were revised and edited manually to maximize the similar-ty with available ESTs or with human proteins. Accession numbersf the genes and database URLs used in this work are provided inable 1.

The orthology of the proteins was inferred initially by recip-ocal best BLASTP searches against human and Drosophila proteinatabases [23], and confirmed by phylogenetic reconstructions. Forhe Aldh family, in addition to the reciprocal BLASTP approach,e combined phylogenetic analysis with other types of infor-ation with low homoplasy, such as differences in exon–intron

rganization and the presence or absence of specific sequenceotifs (e.g., N-terminal signaling sequences that determine the

ubcellular localization of the Aldh enzymes), to unambiguouslyistinguish between Aldh1a and Aldh2 families [6]. Gene struc-ures were deduced by merging the genomic sequences withSTs when available, or by comparison with well character-zed Aldh genes described in other species. The iPSORT programhttp://hc.ims.u-tokyo.ac.jp/iPSORT/) was used to predict the sub-ellular localization of the deduced enzymes [24].

.2. Phylogenetic analysis

Protein sequence alignments were generated with clustalX [25].nly the regions rendering unambiguous alignments among par-logs were considered for the phylogenetic analysis: from codon40 to I513 of human ALDH1A2 for Aldh alignment; from codon74 to E156 and from I236 and S417 of human RARA for nucleareceptor alignment; and from codon P45 to F490 of human CYP26Aor cytochrome P450 alignment. A neighbor-joining (NJ) phyloge-etic tree was constructed and visualized with NJPlot and Unrootedrograms [26]. Confidence in each node was assessed by 1000 boot-trap replicates. Maximum-likelihood (ML) analysis was performedsing the PHYML 2.4.5 program [27] from the MacGDE package, fol-

owing the JTT model of amino acid substitution [28], estimating themino acid frequencies from the data set and taking into accounthe among-site rate heterogeneity with four gamma-distributedategories. The confidence of each node was assessed by 500 boot-trap replicates.

. Results

.1. Identification of Aldh1a and Aldh2 in protostomes

To unambiguously distinguish Aldh1a enzymes from the closelyelated mitochondrial Aldh2 enzymes, we used a recently proposedntegrative approach that combines phylogenetic reconstructions,dentification of Aldh-family-specific signatures (the exon–intronrganization of the Aldh genes) and the prediction of the subcellu-ar localization of the deduced proteins [6]. Aldh1a genes lack intronbut include an extra intron 12b, whereas the opposite – the pres-

nce of intron 4 and absence of 12b – is characteristic of Aldh2 genes.

ldh1a proteins are cytosolic enzymes, whereas Aldh2 localize in

he mitochondria. During our in silico screening, identification ofhe Aldh1L genes, the next most closely related Aldh family, wasonsidered evidence that the complete catalog of Aldh1a and Aldh2enes present in the databases had been retrieved.

190R

.Albalat,C.Canestro

/Chemico-BiologicalInteractions

178(2009)

188–196

Table 1Phylogenetic distribution of RA-genetic machinery and related sequences.

Species Aldh1a Aldh2 Aldh1L Cyp26 Cyp51 Cyp4 RAR THR Eip75B-Ppar EcR-Lxr RXR

BILATERIANSDeuterostomiaChordataa (Verte-

brata + Urochordata + Cephalochordata)+ + + + + + + + + + +

Ambulacrariaa (Hemichor-data + Echinodermata)

+ + + + + + + + + + +

LophotrochozoaAnnelida + + + + + + + + + + +

Capitella sp.b 151890 183731 228199 212322 173561 + 168520 167148 62897 125155 164614145966 150007159056 95922

129163Helobdella robustac 186284 194011 108094 nf 157030 + nf nf 67463 108893 186353

Mollusca + + + + + + + + + + +Lottia gigantead 207312 157947 198550 232238 205907 + 142757 224598 136477 170342 206562

164897 111018

Platyhelminthes + + − − − + − + + − +Schistosoma mansonie Smp 050390 Smp 022960 nf nf nf + nf AAR29358 AAR30507 nf AAD16119

AAR29359 AAD45325

EcdysozoaArthropoda + + + − − + − + + + +

Anopheles gambiaef XP 313331 XP 313425 XM 318614 nf nf + nf nf XP 320316 XP 320323 XP 320944XP 319075

Apis melliferaf XP 392104 + XP 623252 XP 623084 XP 623798 nf nf + nf nf NP 001073579 NP 001091685 NP 001011634Drosophila.melanogasterf AAF56646 NP 609285 NP 610107 nf nf + nf nf NP 730321 NP 724456 NP 476781Tribolium castaneumf XP 970835 XP 967960 XP 969916 nf nf + nf nf XP 971362 NP 001107650 NP 001107766Daphnia pulexg 215225 318586 326816 nf nf + nf 316465 205205 319648 219609

305826 301472

Nematoda − + + − − + − +i + + +Brugia malayif nfh nfh nfh nf nf + nf nf XP 001899570 ABQ28713 ABQ28715Caenorhabditis elegansf nf NP 503467 NP 498081 NP 502054 nf nf nf nf nf Q9XUK7 nf nf

Abbreviations: Aldh: Aldehyde dehydrogenase; Cyp: Cytochrome P450 enzyme; RAR: Retinoic acid receptor; THR: Thyroid hormone receptor; Eip75B: Ecdysone-induced protein 75B; Ppar: Peroxisome proliferator activatedreceptor; EcR: Ecdysone receptor; Lxr: Liver X receptor; RXR: Retinoid X receptor.

a Chordate and Ambulacraria sequences are those reported in [6]. Ambulacraria LXR and PPAR correspond to the predicted sea urchin XP 779997 and XP 781750 proteins, respectively.b JGI Capitella sp. I v1.0: http://genome.jgi-psf.org/Capca1/Capca1.home.html.c JGI H. robusta v1.0: http://genome.jgi-psf.org/Helro1/.d JGI L. gigantea v1.0: http://genome.jgi-psf.org/Lotgi1/Lotgi1.home.html.e GeneDB S. manosoni: http://www.genedb.org/genedb/smansoni/blast.jsp.f NCBI: www.ncbi.nlm.nih,gov.g JGI D. pulex v1.0: http://genome.jgi-psf.org/Dappu1/Dappu1.home.html.h Three partial sequences similar to Aldh1a, Aldh2 and Aldh1L enzymes can be found in the incomplete B. malayi genome project (TIGR database: http://www.tigr.org/tdb/e2k1/bma1/); accession numbers: BRBFS81TR,

BRQGH23TF and BRNJU76TF.i Schmidtea mediterranea THRa and THRb sequences reported in [56].

R. Albalat, C. Canestro / Chemico-Biological Interactions 178 (2009) 188–196 191

Table 2Invertebrate Aldh1a and Aldh2. Accession numbers and family signatures.

Species Gene Accession number pMTP I4/I12b

Capitella sp. CspAldh1a1/2/3a 151890 − −/+CspAldh1a1/2/3b 145966 − −/+CspAldh1a1/2/3c 159056 − −/+CspAldh2 183731 + +/−

Helobdella robusta HrAldh1a 1/2/3 186284 − −/+HrAldh2 194011 + +/−

Lottia gigantea LgAldh1a1/2/3a 207312 − −/+LgAldh1a1/2/3b 164897 − −/+LgAldh2 157947 + +/−

Placopecten magellanicus Pm˝-crystallin AAF73122 − −/+Schistosoma mansoni SmAldh1a1/2/3 Smp 050390 − −/+

SmAldh2 Smp 022960 − +/−Anopheles gambiae AgAldh1a1/2/3a XP 313331 − na

AgAldh1a1/2/3b XP 319075 − naAgAldh2 XP 313425 + na

Apis mellifera AmAldh1a1/2/3 XP 392104+XP 623252 − −/+AmAldh2 XP 623084 + −/−

Drosophila melanogaster DmAldh1a1/2/3 AAF56646 − naDmAldh2 NP 609285 + na

Tribolium castaneum TcAldh1a1/2/3 XP 970835 − −/+TcAldh2 XP 967960 + −/−

Daphnia pulex DpAldh1a1/2/3a 215225 − −/+DpAldh1a1/2/3b 305826 − −/+DpAldh2 318586 + −/−

Caenorhabditis elegans CeAldh2a NP 503467 − naCeAldh2b NP 498081 + na

Nematostella vectensis NvAldh1a-2 245626 − −/+NvAldh1a-2 181421 − −/+NvAldh1a-2 179476 + −/+

Trichoplax adhaerens TaAldh1a-2 37388 − +/+TaAldh1a-2 35686 − +/+TaAldh1a-2 63774 − +/+

p SORTI genes

tlftyArtjdtfcwtfot4iislhetOmel

c�ecol

agfDfiitrettbTbtdon

MTP: presence (+) or absence (−) of a mitochondrial-targeting peptide based on iP4/I12b: presence (+) or absence (−) of the family-specific introns 4 and 12b in Aldh

To recognize the protostome Aldh1a, we identified all the puta-ive Aldh1a and Aldh2 genes in the genomic databases of elevenophotrochozoan and ecdysozoan species, which covered five dif-erent phyla (Tables 1 and 2). Our genomic survey yielded aotal of twenty-five Aldh sequences (Table 1). Phylogenetic anal-ses revealed that each protostome Aldh grouped either into theldh1a or Aldh2 families. Significantly, we found that at least oneepresentative of both Aldh families was identified in most ofhe species analyzed (Fig. 1). Maximum-likelihood and neighbor-oining methods for tree construction yielded the same topologyefining the two families of enzymes. Phylogenetic assignment ofhe protostome enzymes was consistent with the Aldh1a and Aldh2amily-specific signatures (Fig. 2) and with the prediction of sub-ellular localization for the new protostome enzymes (Table 2),hich provided further support for the orthology relationships of

he enzymes. In lophotrochozoans, seven Aldh1a proteins (threerom Capitella sp., one from H. robusta, two from L. gigantea andne from S. mansoni) were predicted to be cytosolic (Table 2), andheir genes showed the Aldh1a signature (i.e., absence of intron

but presence of intron 12b) (Fig. 2). The opposite situation –.e., prediction of mitochondrial localization, and the presence ofntron 4 and absence of 12b – was found in four putative Aldh2equences in the same species (Table 2 and Fig. 2). Additionalophotrocozoan Aldh sequences from other species whose genomesave not been sequenced were included in the analysis. One Aldhnzyme from Enchytraeus buchholzi (earthworm) (CAA64680) and

he �-crystallins of Octopus dofleini (giant octopus) (AAA29392),mmastrephes sloani (arrow squid) (AAA29406) and Placopectenagellanicus (sea scallop) (AAF73122) had been reported to be

volutionarily related to the Aldh1a-Aldh2 enzymes [29]. Our phy-ogenetic analysis suggested that these proteins belong to the

gwgte

predictions.; na, not applicable because gene structure is completely reorganized.

ytoplasmic Aldh1a family (Fig. 2). Consistent with this result,-crystallins lack the mitochondrial-targeting signal ([29] and ref-

rences therein), and the structure of the sea scallop gene [30]oncurs with the Aldh1a-signature (Fig. 2 and Table 2). Overall,ur analysis revealed the presence of Aldh1a and Aldh2 forms inophotrochozoan species.

In ecdysozoans, Aldh exon–intron organizations were more vari-ble than in lophotrocozoans, but some sequences still showed aene structure informative for family assignment. One Aldh generom A. mellifera, one gene from T. castaneum and two genes from. pulex showed the Aldh1a signature (Fig. 2). These genes encode

or proteins predicted to localize in the cytosol (Table 2), andn phylogenetic reconstructions, grouped together with the othernvertebrate Aldh1a enzymes (Fig. 1). In addition, A. mellifera, T. cas-aneum and D. pulex genomes contained one Aldh gene that partiallyetained the Aldh2 signature (absence of intron 12b), encoded fornzymes predicted to have a mitochondrial localization, and clus-ered within the Aldh2 group in the phylogenetic tree. Therefore,he overall analysis allowed us to convincingly assign each Apis, Tri-olium and Daphnia Aldh to either the Aldh1a or the Aldh2 families.he Aldh nature of the genes from D. melanogaster (2 genes), A. gam-iae (3 genes) and C. elegans (2 genes) (Table 1) was inferred fromheir position in the evolutionary tree (Fig. 1) and from their pre-icted subcellular localization (Table 2). The exon–intron structuref the Aldh genes in such species was poorly conserved and it couldot be used for family assignment. One D. melanogaster and two A.

ambiae proteins clustered with the ecdysozoa Aldh1a forms andere predicted to be cytoplasmic, while one D. melanogaster, one A.

ambiae and the two C. elegans sequences grouped within the inver-ebrate Aldh2 and were predicted to be mitochondrial (except C.legans Aldh2a; Table 2). Thus, based on the phylogenetic relation-

192 R. Albalat, C. Canestro / Chemico-Biological Interactions 178 (2009) 188–196

Fig. 1. Phylogenetic analysis of Aldh1a and Aldh2 families reveals the presence of the Aldh1a in protostomes. The unrooted phylogenetic tree was generated by the neighbor-joining (NJ) method based on the clustalX alignment from I40 to I513 of human ALDH1A2. The same tree topology was obtained by the maximum-likelihood (ML) method.Figures at the nodes are the percentage of bootstrap values supporting the Aldh1a and Aldh2 families, and vertebrate Aldh1a and Aldh2 groups (n = 1000 for NJ, and n = 500f oridaeE culus;S

salptzt

3

doisammMofispoCeh

3r

RaidCb

iddwsmtBa(if

or ML (underlined)). Species abbreviations: Ag, A. gambiae; Am, A. mellifera; Bf, B. fl. buchholzi; Gg, G. gallus; Hr, H. robusta; Hs, H. sapiens; Lg, L. gigantea; Mm, M. musp. S. purpuratus; Tc, T. castaneum; Tr, T. rubripes; Xt, X. tropicalis.

hips and the predictions of subcellular localization, fly, mosquitond worm enzymes were classified into Aldh1a and Aldh2 fami-ies. In conclusion, our analysis provided robust evidence for theresence of orthologs to the deuterostome Aldh1a genes in thewo main branches of protostomes, lophotrochozoans and ecdyso-oans, and therefore indicated that the origin of Aldh1a predatedhe deuterostome–protostome split.

.2. Identification of Cyp26 enzymes in protostomes

No Cyp26 proteins had been previously identified outsideeuterostomes. To discover whether protostomes have Cyp26rthologs, we analyzed the same set of protosome genomes usedn the Aldh analysis. In lophotrochozoans, we identified four Cyp26equences in the Capitella sp. genome and two in L. gigantea (Table 1nd Fig. 3), while no Cyp26 genes were found in H. robusta or S.ansoni. Orthologies, initially established by the reciprocal BLASTethod [23], were corroborated by phylogenetic analyses (Fig. 3).embers of the Cyp51 family (the closest family to Cyp26) and

ther cytochrome P450 enzymes such as Cyp4 forms were searchedor and used as outgroups in the phylogenetic analyses. Interest-ngly, our surveys revealed that none of the seven ecdysozoanpecies (A. gambiae, A. melliferea, D. melanogaster, T. castaneum, D.

ulex and C. elegans) had convincing Cyp26 orthologs. Identificationf more distant cytochrome P450 genes in the BLAST searches usingyp26 and Cyp51 proteins as starting queries was considered asvidence that Cyp26 genes have either been lost or their sequencesave diverged beyond recognition in the species analyzed.

gn(tw

; Ce, C. elegans; Csp, Capitella sp.; Dm, D. melanogaster; Dp, D. pulex; Dr, D. rerio; Eb,Od, O. dofleini; Os, O. sloani; Pm, P. magellanicus; Rn, R. novergicus; Sm, S. mansoni;

.3. Identification of RAR, RXR and other nuclear hormoneeceptors

Heterodimers of RAR and RXR mediate RA-signaling. BecauseXR have been identified in the most basal metazoan phyla, suchs Porifera [31,32] and Cnidaria [33], the presence of RXR orthologsn most animal species was expected. In accordance with this pre-iction, RXR orthologs were found in all species analyzed, except in. elegans (Table 1), in which the lack of RXR orthologs had alreadyeen reported [34,35].

In contrast, RAR homologs had hitherto been described onlyn deuterostomes [6–8]. To determine whether or not non-euterostomes possess RAR orthologs, we surveyed the genomicatabases of the eleven protostome species analyzed in thisork. We found, for the first time, putative RAR genes out-

ide deuterostomes: in the annelid Capitella sp. and in theollusc L. gigantea (Table 1). The orthology between deuteros-

ome and protostome RAR was initially based on the reciprocalLAST method [23] and was robustly supported by phylogeneticnalyses (Fig. 3). Members of the thyroid hormone receptorsTHR), the closest related receptors to the RAR family, were alsodentified in Capitella sp. and L. gigantea genomes (Table 1),urther supporting the RAR nature of the newly identified

enes. Additional invertebrate members of the subfamily I ofuclear receptors such as the ecdysone-induced proteins 75BEip75B), which are related to vertebrate peroxisome prolifera-or activated receptors (PPAR), and the ecdysone receptors (EcR),hich are closely related to the vertebrate liver X receptors

R. Albalat, C. Canestro / Chemico-Biological Interactions 178 (2009) 188–196 193

Fig. 2. Comparison of the gene structure of Aldh1a and Aldh2 families across bilaterians reveals the presence of family-specific intron positions, and validates the use of theexon–intron organization as a homoplastic feature for family assignments [6]. Chordate gene structures are represented by H. sapiens ALDH1A1, urochordate C. intestinalisAldh1a1/2/3a and cephalochordate B. floridae Aldh1a1/2/3e genes. For clarity, intron nomenclature (1–12, 12b and 13) is as in [6]. The positions of the introns 4 and 12b, whichdefine the Aldh-family signatures, are indicated by black arrowheads on a gray background; conserved intron positions are indicated with white arrowheads; lineage-specificintron positions are indicated with gray arrowheads (not numbered).

Fig. 3. Phylogenetic analysis of Cyp26 and RAR. Neighbor-joining (NJ) and maximum-likelihood (ML) methods yielded the same tree topologies. Figures at the nodes arethe percentage of bootstrap values supporting each node (n = 1000 for NJ, and n = 500 for ML, in italics; poorly supported nodes <50% were collapsed). The tree of the Cypenzymes was rooted with the Cyp4v2 sequences, and the tree of the RAR, with the RXR sequences. Capitella sp. and L. gigantea Cyp26 and RAR sequences clearly groupedwith deuterostome Cyp26 and RAR proteins, supporting their orthology to the deuterostome proteins.

194 R. Albalat, C. Canestro / Chemico-Biological Interactions 178 (2009) 188–196

F protoa elatedA

((

4

4

ctsianopCpaa

dopwotwoRceorClaaima

4d

rRlspgpascntaff[

isglAttditdi[

l

ig. 4. The phylogenetic distribution of the RA-genetic machinery (black boxes) inncestor. The lack of RA-genetic machinery (empty box) and the presence of closely rldh proteins was not conclusive (half filled boxes).

LXR) [36], were also identified in most of the phyla analyzedTable 1).

. Discussion

.1. Origin of the RA-genetic machinery

The identification of Aldh1a, Cyp26 and RAR orthologs in non-hordate deuterostomes [6] pushed back in the evolutionary ladderhe origin of the RA-genetic machinery and challenged the relation-hip between the invention of this machinery and chordate-specificnnovations in the body plan. To further investigate the evolution-ry origin of the RA machinery, we analyzed a catalogue of elevenon-deuterostome genomes, searching for Aldh1a, Cyp26 and RARrthologs as diagnostic features of the RA-genetic machinery in therotostome lineage. Our finding of clear orthologs of the Aldh1a,yp26 and RAR families in several protostome species (Table 1)ushes back once again the origin of the RA-genetic machinerynd implies that these genes already existed in the last commonncestor of bilaterians (Fig. 4).

In an attempt to narrow down the origin of the RA machineryuring animal evolution, we explored available genomic databasesf non-bilaterian organisms such as the cnidarian N. vectensis andlacozoan T. adhaerens. Three putative Aldh1a-Aldh2 sequencesere identified in the N. vectensis genome (Table 2), but no Cyp26

r RAR orthologs were recognized in the database. In T. adhaerens,hree Aldh1a-Aldh2 (Table 2) and one Cyp26 (ID 60776) sequencesere found, but no RAR orthologs were identified. Although neither

f the two non-bilaterian animals analyzed in this work have a fullA toolkit, the taxonomic diversity is still too narrow to draw firmonclusions about whether a complete RA-genetic machinery couldxist in stem metazoans. We also analyzed the genomic databasef the choanoflagellate Monosiga brevicollis, the closest unicellularelative of animals [37,38]. The absence of any convincing Aldh1a,yp26 and RAR genes in M. brevicollis suggested that this organism

acks the RA-genetic machinery (Fig. 4), which is consistent with thebsence of members of the NR superfamily in choanoflagellates [39]nd in any other non-metazoan organism [19,40]. Overall, our find-ngs suggest that the RA-genetic machinery might be confined to

etazoans, although a definite answer must await larger databasesnd broader phylogenetic samplings.

vRtmt9

stomes and deuterostomes indicates that it was already present in the bilateriangenes (grey boxes) are also depicted. The family identity of cnidarian and placozoan

.2. Divergent patterns of conservation of the RA machineryuring protostome evolution

Our genomic survey reveals that protostome species haveetained the components of RA machinery (i.e., Aldh1a, Cyp26,AR orthologs) to different extents. Some species, principally from

ophotrochozoa, have preserved most of the elements while otherpecies, especially from ecdysozoa, have lost nearly all the com-onents. The cases of the lophotrochozoans Capitella sp. and L.igantea are particularly interesting, because the finding of com-lete RA-gene toolkits in their genomes provides for the first timemechanistic framework for the presence of a classic RA-signaling

ystem in protostomes, and raises the possibility of functionalonservation of the RA-signaling system between chordate andon-chordate species. In this regard, it has been reported that inhe lophotrochozoan Lymnaea stagnalis (great pond snail), RA actss a trophic factor and a chemotactic molecule [41] and affects eyeormation [42]. Similarly, RA induces neurogenesis and neurite dif-erentiation, and participates in eye morphogenesis in vertebrates43–46].

RA machinery has been lost or modified beyond recognitionn many metazoan species. First, none out of seven ecdysozoanpecies possess Cyp26 and RAR orthologs; second, Cyp26 and RARenes have also been lost in the lophotrochozoans H. robusta (phy-um Annelida) and S. mansoni (phylum Platyhelminthes); and third,ldh1a, Cyp26 and RAR genes have been lost during the evolution ofhe chordate Oikopleura dioca [6]. The phylogenetic distribution ofhese species implies that such losses have occurred independentlyuring animal evolution. Moreover, the absence of the RA toolkit

mplies that RA does not play the classical morphogenetic role inhese animals. This prediction has been experimentally tested in O.ioca, in which it has been shown the lack of the classical role for RAn anterior-posterior axial patterning during embryo development47,48].

It is worth mentioning that although some protostome speciesack Cyp26 and RAR genes, most of them retain Aldh1a orthologs. Inertebrates, in addition to the classical RAR-mediated RA-signaling,

A participates in other functions by the direct binding to other pro-eins (e.g., protein kinase C alpha) [49,50] for which a source of RA

ight be required. Moreover, Nowickyj et al. have recently shownhat in vitro, RXR from the primitive insect Locusta migratoria binds-cis-RA and all-trans-RA with high affinity, and that embryonic

ologica

etpttttdtadhbtt

gtgal

C

A

BC

lrmRa

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

R. Albalat, C. Canestro / Chemico-Bi

xtracts contain significant amounts of endogenous 9-cis- and all-rans-RA [51]. Therefore, although we do not yet know whether RAlays similar non-RAR mediated roles in other invertebrate species,his possibility needs to be considered when analyzing the func-ionality of the Aldh1a enzymes in protostomes. Moreover, it haso be remembered that orthology does not necessarily imply func-ional conservation, and that Aldh1a enzymes could have acquiredifferent activities other than RA synthesis [52]. In this regard,he �-crystallins of molluscs are paradigmatic examples becauselthough they belong to the Aldh1a family, they lack retinaldehydeehydrogenase activity, do not bind NAD or NADP cofactors, andave been recruited to be preferentially expressed in the inverte-rate eye lens [30,53–55]. Therefore, the biochemical activities ofhe different protostome Aldh1a identified here, and whether or nothey are involved in RA metabolism, deserve further investigation.

In summary, our data provide robust evidence that the RA-enetic machinery is an old evolutionary device that arose beforehe divergence of the major animal phyla, and reveal that the RA-enetic machinery has been differently preserved, lost, recruitednd modified during animal evolution to befit the diverse physio-ogical and developmental requirements of different organisms.

onflict of interest

The authors declare that there are no conflicts of interest.

cknowledgements

This material is based on work supported by Universitat dearcelona Grant MC064448 to RA and NSF Grant IOB-0719577 toC.

In agreement with our conlusions, Campo-Paysaa and col-eagues reviewed the RA signalling in several invertebrate groups,eaching to similar inferences on the ancient origin of the RAachinery: F. Campo-Paysaa, F. Marlétaz, V. Laudet, M. Schubert.

etinoic acid signaling in development: Tissue-specific functionsnd evolutionary origins. Genesis (in press).

eferences

[1] S.A. Ross, P.J. McCaffery, U.C. Drager, L.M. De Luca, Retinoids in embryonaldevelopment, Physiol. Rev. 80 (2000) 1021–1054.

[2] K. Niederreither, P. Dollé, Retinoic acid in development: towards an integratedview, Nat. Rev. Genet. 9 (2008) 541–553.

[3] A.J. Durston, J.P. Timmermans, W.J. Hage, H.F. Hendriks, N.J. de Vries, M. Heide-veld, P.D. Nieuwkoop, Retinoic acid causes an anteroposterior transformationin the developing central nervous system, Nature 340 (1989) 140–144.

[4] S. Fujiwara, Retinoids and nonvertebrate chordate development, J. Neurobiol.66 (2006) 645–652.

[5] T.F. Schilling, R.D. Knight, Origins of anteroposterior patterning and Hox generegulation during chordate evolution, Philos. Trans. R. Soc. Lond. B Biol. Sci. 356(2001) 1599–1613.

[6] C. Canestro, J.H. Postlethwait, R. Gonzàlez-Duarte, R. Albalat, Is retinoic acidgenetic machinery a chordate innovation? Evol. Dev. 8 (2006) 394–406.

[7] F. Marletaz, L.Z. Holland, V. Laudet, M. Schubert, Retinoic acid signaling and theevolution of chordates, Int. J. Biol. Sci. 2 (2006) 38–47.

[8] M.S. Simões-Costa, A.P. Azambuja, J. Xavier-Neto, The search for non-chordateretinoic acid signaling: lessons from chordates, J. Exp. Zool. B Mol. Dev. Evol.310 (2008) 54–72.

[9] S. Reijntjes, A. Blentic, E. Gale, M. Maden, The control of morphogen signalling:regulation of the synthesis and catabolism of retinoic acid in the developingembryo, Dev. Biol. 285 (2005) 224–237.

10] G. Duester, F.A. Mic, A. Molotkov, Cytosolic retinoid dehydrogenases governubiquitous metabolism of retinol to retinaldehyde followed by tissue-specificmetabolism to retinoic acid, Chem. Biol. Interact. 143–144 (2003) 201–210.

11] N.A. Sophos, V. Vasiliou, Aldehyde dehydrogenase gene superfamily: the 2002

update, Chem. Biol. Interact. 143–144 (2003) 5–22.

12] S. Pittlik, S. Domingues, A. Meyer, G. Begemann, Expression of zebrafish aldh1a3(raldh3) and absence of aldh1a1 in teleosts, Gene Exp. Patterns 8 (2008)141–147.

13] H. Fujii, T. Sato, S. Kaneko, O. Gotoh, Y. Fujii-Kuriyama, K. Osawa, S. Kato,H. Hamada, Metabolic inactivation of retinoic acid by a novel P450 dif-

[

[

l Interactions 178 (2009) 188–196 195

ferentially expressed in developing mouse embryos, EMBO J. 16 (1997)4163–4173.

14] J.A. White, B. Beckett-Jones, Y.D. Guo, F.J. Dilworth, J. Bonasoro, G. Jones,M. Petkovich, cDNA cloning of human retinoic acid-metabolizing enzyme(hP450RAI) identifies a novel family of cytochromes P450, J. Biol. Chem. 272(1997) 18538–18541.

15] J.A. White, H. Ramshaw, M. Taimi, W. Stangle, A. Zhang, S. Everingham,S. Creighton, S.P. Tam, G. Jones, M. Petkovich, Identification of the humancytochrome P450, P450RAI-2, which is predominantly expressed in the adultcerebellum and is responsible for all-trans-retinoic acid metabolism, Proc. Natl.Acad. Sci. U.S.A. 97 (2000) 6403–6408.

16] M. Taimi, C. Helvig, J. Wisniewski, H. Ramshaw, J. White, M. Amad, B. Korczak, M.Petkovich, A novel human cytochrome P450, CYP26C1, involved in metabolismof 9-cis and all-trans isomers of retinoic acid, J. Biol. Chem. 279 (2004) 77–85.

17] K. Nagatomo, S. Fujiwara, Expression of Raldh2, Cyp26 and Hox-1 in normal andretinoic acid-treated Ciona intestinalis embryos, Gene Exp. Patterns 3 (2003)273–277.

18] H. Escriva, F. Delaunay, V. Laudet, Ligand binding and nuclear receptor evolu-tion, Bioessays 22 (2000) 717–727.

19] S. Bertrand, F.G. Brunet, H. Escriva, G. Parmentier, V. Laudet, M. Robinson-Rechavi, Evolutionary genomics of nuclear receptors: from twenty-fiveancestral genes to derived endocrine systems, Mol. Biol. Evol. 21 (2004)1923–1937.

20] H. Escriva, S. Bertrand, P. Germain, M. Robinson-Rechavi, M. Umbhauer, J. Cartry,M. Duffraisse, L. Holland, H. Gronemeyer, V. Laudet, Neofunctionalization invertebrates: the example of retinoic acid receptors, PLoS Genet. 2 (2006) e102.

21] M. Howard-Ashby, S.C. Materna, C.T. Brown, L. Chen, R.A. Cameron,E.H. Davidson, Gene families encoding transcription factors expressed inearly development of Strongylocentrotus purpuratus, Dev. Biol. 300 (2006)90–107.

22] S.F. Altschul, T.L. Madden, A.A. Schaffer, J. Zhang, Z. Zhang, W. Miller, D.J. Lipman,Gapped BLAST and PSI-BLAST: a new generation of protein database searchprograms, Nucleic Acids Res. 25 (1997) 3389–3402.

23] D.P. Wall, H.B. Fraser, A.E. Hirsh, Detecting putative orthologs, Bioinformatics19 (2003) 1710–1711.

24] H. Bannai, Y. Tamada, O. Maruyama, K. Nakai, S. Miyano, Extensive feature detec-tion of N-terminal protein sorting signals, Bioinformatics 18 (2002) 298–305.

25] J.D. Thompson, T.J. Gibson, F. Plewniak, F. Jeanmougin, D.G. Higgins, TheCLUSTAL X windows interface: flexible strategies for multiple sequence align-ment aided by quality analysis tools, Nucleic Acids Res. 25 (1997) 4876–4882.

26] G. Perrière, M. Gouy, WWW-query: an on-line retrieval system for biologicalsequence banks, Biochimie 78 (1996) 364–369.

27] S. Guindon, O. Gascuel, A simple, fast, and accurate algorithm to estimate largephylogenies by maximum likelihood, Syst. Biol. 52 (2003) 696–704.

28] D.T. Jones, W.R. Taylor, J.M. Thornton, The rapid generation of mutation datamatrices from protein sequences, CABIOS 8 (1992) 275–282.

29] J. Horwitz, L. Ding, V. Vasiliou, M. Cantore, J. Piatigorsky, Scallop lens Omega-crystallin (ALDH1A9): a novel tetrameric aldehyde dehydrogenase, Biochem.Biophys. Res. Commun. 348 (2006) 1302–1309.

30] E. Carosa, Z. Kozmik, J.E. Rall, J. Piatigorsky, Structure and expression of thescallop Omega-crystallin gene. Evidence for convergent evolution of promotersequences, J. Biol. Chem. 277 (2002) 656–664.

31] M. Wiens, R. Batel, M. Korzhev, W.E. Muller, Retinoid X receptor and retinoicacid response in the marine sponge Suberites domuncula, J. Exp. Biol. 206 (2003)3261–3271.

32] C. Larroux, B. Fahey, D. Liubicich, V.F. Hinman, M. Gauthier, M. Gongora, K.Green, G. Worheide, S.P. Leys, B.M. Degnan, Developmental expression of tran-scription factor genes in a demosponge: insights into the origin of metazoanmulticellularity, Evol. Dev. 8 (2006) 150–173.

33] Z. Kostrouch, M. Kostrouchova, W. Love, E. Jannini, J. Piatigorsky, J.E. Rall,Retinoic acid X receptor in the diploblast, Tripedalia cystophora, Proc. Natl. Acad.Sci. U.S.A. 95 (1998) 13442–13447.

34] A.E. Sluder, C.V. Maina, Nuclear receptors in nematodes: themes and variations,Trends Genet. 17 (2001) 206–213.

35] A. Antebi, Nuclear Hormone Receptors in C. elegans, WormBook, 2006, pp. 1–13.36] N. Ollikainen, C. Chandsawangbhuwana, M.E. Baker, Evolution of the thyroid

hormone, retinoic acid, ecdysone and liver X receptors, Integr. Comp. Biol. 46(2006) 815–826.

37] B.F. Lang, C. O’Kelly, T. Nerad, M.W. Gray, G. Burger, The closest unicellularrelatives of animals, Curr. Biol. 12 (2002) 1773–1778.

38] N. King, S.B. Carroll, A receptor tyrosine kinase from choanoflagellates: molec-ular insights into early animal evolution, Proc. Natl. Acad. Sci. U.S.A. 98 (2001)15032–15037.

39] N. King, M.J. Westbrook, S.L. Young, A. Kuo, M. Abedin, J. Chapman, S. Fairclough,U. Hellsten, Y. Isogai, I. Letunic, M. Marr, D. Pincus, N. Putnam, A. Rokas, K.J.Wright, R. Zuzow, W. Dirks, M. Good, D. Goodstein, D. Lemons, W. Li, J.B. Lyons,A. Morris, S. Nichols, D.J. Richter, A. Salamov, J.G. Sequencing, P. Bork, W.A.Lim, G. Manning, W.T. Miller, W. McGinnis, H. Shapiro, R. Tjian, I.V. Grigoriev,D. Rokhsar, The genome of the choanoflagellate Monosiga brevicollis and the

origin of metazoans, Nature 451 (2008) 783–788.

40] H. Escriva, S. Bertrand, V. Laudet, The evolution of the nuclear receptor super-family, Essays Biochem. 40 (2004) 11–26.

41] J.M. Dmetrichuk, R.L. Carlone, G.E. Spencer, Retinoic acid induces neurite out-growth and growth cone turning in invertebrate neurons, Dev. Biol. 294 (2006)39–49.

1 ologica

[

[

[

[

[

[

[

[

[

[

[

[

[

[

96 R. Albalat, C. Canestro / Chemico-Bi

42] R. Créton, G. Zwaan, R. Dohmen, Specific developmental defects in molluscsafter treatment with retinoic acid during gastrulation, Dev. Growth Differ.(1993) 357–364.

43] M. Maden, G. Keen, G.E. Jones, Retinoic acid as a chemotactic molecule in neu-ronal development, Int. J. Dev. Neurosci. 16 (1998) 317–322.

44] M. Maden, Retinoid signalling in the development of the central nervous sys-tem, Nat. Rev. Neurosci. 3 (2002) 843–853.

45] F.A. Mic, A. Molotkov, N. Molotkova, G. Duester, Raldh2 expression in opticvesicle generates a retinoic acid signal needed for invagination of retina duringoptic cup formation, Dev. Dyn. 231 (2004) 270–277.

46] A. Molotkov, N. Molotkova, G. Duester, Retinoic acid guides eye morphogeneticmovements via paracrine signaling but is unnecessary for retinal dorsoventralpatterning, Development 133 (2006) 1901–1910.

47] C. Canestro, J.H. Postlethwait, Development of a chordate anterior-posterioraxis without classical retinoic acid signaling, Dev. Biol. 305 (2007)522–538.

48] C. Canestro, H. Yokoi, J.H. Postlethwait, Evolutionary developmental biology

and genomics, Nat. Rev. Genet. 8 (2007) 932–942.

49] W.F. Ochoa, A. Torrecillas, I. Fita, N. Verdaguer, S. Corbalán-García, J.C. Gomez-Fernandez, Retinoic acid binds to the C2-domain of protein kinase C(alpha),Biochemistry 42 (2003) 8774–8779.

50] A. Radominska-Pandya, G. Chen, P.J. Czernik, J.M. Little, V.M. Samokyszyn, C.A.Carter, G. Nowak, Direct interaction of all-trans-retinoic acid with protein

[

l Interactions 178 (2009) 188–196

kinase C (PKC). Implications for PKC signaling and cancer therapy, J. Biol. Chem.275 (2000) 22324–22330.

51] S.M. Nowickyj, J.V. Chithalen, D. Cameron, M.G. Tyshenko, M. Petkovich, G.R.Wyatt, G. Jones, V.K. Walker, Locust retinoid X receptors: 9-Cis-retinoic acidin embryos from a primitive insect, Proc. Natl. Acad. Sci. U.S.A. 105 (2008)9540–9545.

52] L.C. Hsu, W.C. Chang, I. Hoffmann, G. Duester, Molecular analysis of two closelyrelated mouse aldehyde dehydrogenase genes: identification of a role for Aldh1,but not Aldh-pb, in the biosynthesis of retinoic acid, Biochem. J. 339 (1999)387–395.

53] S.I. Tomarev, R.D. Zinovieva, J. Piatigorsky, Crystallins of the octopuslens. Recruitment from detoxification enzymes, J. Biol. Chem. 266 (1991)24226–24231.

54] R.D. Zinovieva, S.I. Tomarev, J. Piatigorsky, Aldehyde dehydrogenase-derivedomega-crystallins of squid and octopus. Specialization for lens expression, J.Biol. Chem. 268 (1993) 11449–11455.

55] J. Piatigorsky, Z. Kozmik, J. Horwitz, L. Ding, E. Carosa, W.G. Robison Jr., P.J.

Steinbach, E.R. Tamm, Omega-crystallin of the scallop lens. A dimeric alde-hyde dehydrogenase class 1/2 enzyme-crystallin, J. Biol. Chem. 275 (2000)41064–41073.

56] W. Wu, E.G. Niles, P.T. LoVerde, Thyroid hormone receptor orthologues frominvertebrate species with emphasis on Schistosoma mansoni, BMC Evol. Biol. 7(2007) 150.