evolution of the echinoderm hox gene cluster

EVOLUTION & DEVELOPMENT

3:5, 302–311 (2001)

©

BLACKWELL SCIENCE, INC.

302

Evolution of the echinoderm

Hox

gene cluster

Suzanne Long* and Maria Byrne

Anatomy F13, University of Sydney, NSW 2006, Australia

*Author for correspondence (email: [email protected])

SUMMARY

Extant echinoderms are members of an an-cient and highly derived deuterostome phylum. The composi-tion and arrangement of their

Hox

gene clusters areconsequently of interest not only from the perspective of evo-lution of development, but also in terms of metazoan phylog-eny and body plan evolution. Over the last decade numerousworkers have reported partial

Hox

gene sequences from avariety of echinoderms. In this paper we used a combinedmethods approach to analyze phylogenetic relationships be-tween 68 echinoderm

Hox

homeodomain fragments, fromspecies of five extant classes—two asteroids, one crinoid,

one ophiuroid, one holothuroid, and three echinoids. Thisanalysis strengthens Mito and Endo’s (2000) proposition thatthe ancestral echinoderm’s

Hox

gene cluster contained atleast eleven genes, including at least four posterior paralo-gous group genes. However, representatives of all paralo-gous groups are not known from all echinoderm classes. Inparticular, these data suggest that echinoids may have lost aposterior group

Hox

gene subsequent to the divergence ofthe echinoderm classes. Evolution of the highly derived echi-noderm body plan may have been accompanied by class-specific duplication, diversification and loss of

Hox

genes.

INTRODUCTION

Hox

genes probably evolved in the metazoan ancestor andare very highly conserved phylogenetically (for review seeGellon and McGinnis 1998).

Hox

proteins are homeo-domain-containing transcription factors that assign differentidentities to body regions by the differential regulation of nu-merous downstream genes (Gellon and McGinnis 1998). De-finitive features of

Hox

genes, such as clustering, colinearity,and their close sequence similarity, are suggestive of an ori-gin by tandem gene duplication and divergence from a smallnumber of ancestral genes (Lewis 1978; Kappen et al. 1989;Schubert et al. 1993; Zhang and Nei 1996).

Hox

genes are important in specification of bilaterianbody plans (Carroll 1995; Hall 1996). Echinoderms, al-though presumed to have evolved from the same bilaterianancestor that gave rise to chordates, display a highly derivedmetazoan body architecture (Lowe and Wray 1997). Thesuite of autapomorphic changes within the phylum and therelationships between the five extant classes are reasonablywell understood (Littlewood et al. 1997; Janies and Mooi1999; but see Rowe et al. (1988) for discussion of the contro-versial concentricycloids), and in this context, a more gen-eral understanding of the connection between

Hox

gene evo-lution and body plan evolution might be possible. This hasled to numerous fishing expeditions for

Hox

genes in variousspecies of echinoderms, mainly echinoids and asteroids(e.g., Popodi et al. 1996; Mito and Endo 1997, 2000; Morriset al. 1997; Arenas-Mena et al. 1998; Martinez et al. 1999;

Long et al. 2000; Mendez et al. 2000). Most of the genes re-ported in these papers were located via PCR surveys usingdegenerate primers, so few conclusions could be drawn re-garding the evolution of the

Hox

cluster until Martinez et al.(1999) reported the complete cluster from the echinoid

Strongylocentrotus purpuratus

.Despite their pentamerically symmetrical body plans, the

Hox

cluster of echinoids is essentially similar to those of bilat-erally organized chordates (Martinez et al. 1999). As proposedby Popodi et al. (1996), echinoid

Hox

genes appear to be sin-gle-copy, as known in lower chordates (i.e., cephalochordatesand urochordates) and other invertebrate metazoans. The

Hox

cluster of

S. purpuratus

is

�

0.5 megabase in length and con-tains 10 genes, which on the basis of sequence similarity canbe assigned to the anterior, medial, and posterior paralogousgroups observed in

Hox

clusters from many phyla (Arenas-Mena et al. 1998; Pradel and White 1998; Martinez et al.1999). Anteriorly and medially, the echinoid genes are furtherclassifiable into paralogous groups (PG, following Martinez etal. 1999), reflective of possible homologies with vertebrate

Hox

PG but relationships between the posterior PG

Hox

genesof these two deuterostome groups are unclear.

Since then two important papers (Mendez et al. 2000,Mito and Endo 2000) have provided the first informationabout

Hox

genes from three other classes of echinoderms—holothuroids, crinoids, and ophiuroids. These data alongwith the nomenclature and framework established by Mar-tinez et al. (1999) made the present comparative analysis anddiscussion possible.

Long and Byrne

Echinoderm Hox gene cluster evolution

303

Because of their fundamental importance to developmentalprocesses involved in body plan construction across the meta-zoa and their perceived independence from 18S rDNA data,

Hox

gene sequences are attractive to phylogeneticists inter-ested in elucidating gross relationships between metazoanphyla (Balavoine 1998; de Rosa et al. 1999); de Rosa et al.(1999) went so far as to state that the reconstruction of the evo-lutionary history of the

Hox

gene family was central to under-standing the evolution of bilaterian body plans. However,characteristic features of

Hox

genes, such as their sequenceconservation, the possibility of strict functional constraints onsequence variation, and their probable origin through tandemduplication events, could pose problems for phylogeneticanalysis (Irvine et al. 1997; Kourakis and Martindale 2000).Nevertheless, the accuracy of a phylogeny is greater when dif-ferent phylogenetic methods, each making trees according todifferent sets of criteria, agree on tree topology (Kim 1993;Kourakis and Martindale 2000). Combined methods ap-proaches are consequently necessary when dealing with aphylogenetically problematic data set such as this one (Kim1993; Irvine et al. 1997; Ferrier et al. 2000; Kourakis and Mar-tindale 2000). Here we compared parsimony and distance(neighbor-joining and maximum likelihood) methods in ouranalysis of phylogenetic relationships between 68 partial

Hox

homeodomains from five classes of echinoderms.

METHODS

Ferrier et al. (2000) considered a sequence of 72 amino acids, span-ning the complete homeodomain with six residues of flanking se-quence on either side, most phylogenetically informative for

Hox

genes. However, in the case of echinoderm genes, sequence outsideresidues 47–60 of the homeodomain was available for only 10genes, eight of which were from echinoids. Clearly, few conclusionsregarding

Hox

gene phylogeny could be drawn from an analysis us-ing only these examples (although this limited data set was used asa methods test; results not shown). In the present study selection ofa 24 amino acid partial homeodomain sequence enabled analysis of68 echinoderm

Hox

genes, collected from online sequence databasesand published papers. Species of five extant echinoderm classeswere represented, involving 14 asteroid genes [

Asterina minor

(

AM

)and

Patiriella exigua

(

Pe

)], 22 echinoid genes [

Heliocidaris eryth-rogramma

(

He

),

Holopneustes purpurescens

(

Hp

), and

Strongylo-centrotus purpuratus

(

Sp

)], nine holothuroid genes [

Holothuriaglaberrima

(

Hg

)], 11 ophiuroid genes [

Stegophiura sladeni

(

SS

)],and 12 crinoid genes [

Oxycomanthus japonicus

(

CJ

)]. The 68 se-quences and their sources were:

AM1

,

AM3

,

AM4

,

AM5

,

AM6

,

AM7

,

AM9

(Mito and Endo 1997);

PeHbox3

,

PeHbox4

,

PeHbox6

,

PeHbox7

,

PeHbox9

,

PeHbox15

,

PeHbox16

(Long et al. 2000);

HeHbox1

,

HeHbox6

,

HeHbox7

,

HeHbox9

,

HeHbox10

(Popodi et al.1996);

HpHbox1

,

HpHbox3

,

HpHbox4

,

HpHbox6

,

HpHbox7

,

HpHbox10

,

HpHbox11

(Morris et al. 1997);

SpHox1

,

SpHox2

,

SpHox3

,

SpHox4/5

,

SpHox6

,

SpHox7

,

SpHox8

,

SpHox9/10

,

SpHox11/13a

,

SpHox11/13b

(Martinez et al. 1999);

HgHbox1

,

HgHbox2

,

HgHbox3

,

HgHbox5

,

HgHbox9

,

HgHbox10

,

HgHbox11

,

HgHbox12

,

HgHbox13

(Mendez et al. 2000);

SSHox1

,

SSHbox1

,

SSHbox4

,

SSHbox6

,

SS9

,

SS9.2

,

SS9a

,

SSHbox10

,

SS11

,

SS12

,

SS13

(Mito and Endo 2000);

CJ28

,

CJHox1

,

CJHbox1

,

CJHox2

,

CJHox3

,

CJHbox3

,

CJHbox4

,

CJHbox4a

,

CJHbox6

,

CJHbox7

,

CJ9,CJHbox10 (Mito and Endo 2000). Included for comparison were hu-man representatives of all 13 vertebrate PG Hox homeodomains:HoxA1 (HSU10421), HoxB2 (X16176), HoxB3 (X16175), HoxA4(M74297), HoxA5 (M26679), HoxB6 (X58431), HoxA7 (P31268),HoxB8 (X16173), HoxA9 (U41813), HoxA10 (S69029), HoxC11(AJ000041), HoxC12 (S14933), and HoxA13 (S14932). Orthologsof engrailed gene products from three deuterostomes were also in-cluded as an outgroup, from the echinoid Heliocidaris erythro-gramma (Heengrai, GI5881326), the amphioxus Branchiostomafloridae (Bfengrai, AAB40144), and human (Hsengrai, E48423).

Given the difficulties outlined above we took a cautious com-bined methods approach to phylogenetic analysis of the collecteddata. Parsimony, neighbor-joining, and maximum likelihood tree-building methods were employed and the resulting trees were com-pared. All analyses were conducted using software available on theBioNavigator by eBioinformatics Pty Ltd website, http://www.eBioinformatics.com. First a multiple alignment was produced usingClustalW(accurate) (Thompson et al. 1994). For maximum likeli-hood analysis Protml (Adachi and Hasegawa 1996) generated a max-imum likelihood tree using the Dayhoff transition model of Kishinoet al. (1990) and the OTU method of topology search. For neighbor-joining analysis, 1000 bootstrap replications of the alignment oc-curred via Seqboot (Felsenstein 1989), and protein distance matriceswere produced by Protdist (Felsenstein 1989) with a PAM-Dayhoffmatrix. Neighbor (Felsenstein 1989) constructed a thousand treesfrom these matrices, and a consensus tree was generated using Con-sense (Felsenstein 1989). Parsimony analysis occurred via Protpars(100 bootstrap replications) (Felsenstein 1989), and a consensus treewas generated using Consense (Felsenstein 1989).

RESULTS

The 84 partial homeodomains used in the analyses arealigned in Fig. 1. Figures 2, 3, and 4 present the parsimony,neighbor-joining, and maximum likelihood trees, respec-tively. Comparison of these four figures shows some strongconsistencies with regard to paralogous groupings. Whilenot every echinoderm homeodomain fragment can be un-equivocally assigned to a particular PG by each method,comparison between trees gives a gauge of the “robustness”of each PG—this is the power of the combined methods ap-proach. The following description of the results is organizedfollowing Martinez et al.’s (1999) work on gene order withinthe echinoid Hox gene cluster.

Anterior paralogous groupsPG1. All three trees agree that AM1, PeHbox16, HgHbox1,SpHox1, CJHox1, and SSHox1, along with human HoxA1,

304 EVOLUTION & DEVELOPMENT Vol. 3, No. 5, September–October 2001

are paralogues. It is clear from sequence similarities and pre-vious analyses (Martinez et al. 1999; Mito and Endo 2000)that these genes are representatives of PG1, the most anteriorgene of the Hox cluster.PG2. Only two candidates for PG2 genes are known fromechinoderms: SpHox2 (whose position second in the echi-noid Hox cluster has been confirmed by Martinez et al.1999), and CJHox2. HoxB2 clades with them in Figs. 2 and3 but is slightly separated in Fig. 4, although it is still withinthe PG2/PG3 branch.PG3. AM3, CJHox3, HpHbox11, HoxB3, SpHox3,HgHbox3, and HgHbox2 consistently clade together, form-ing PG3, which appears to be closely related to PG2. Of noteis the apparent presence within this PG of two holothuroidgene products, HgHbox2 and HgHbox3 [although the parsi-mony tree disagrees (Fig. 2)]. A paralogous ophiuroid genehas not yet been reported.

Medial paralogous groupsPG4/5. Martinez et al. (1999) described PG4/5 as possiblysynapomorphic for echinoderms, since partial homeo-domains of this group are known from both echinoids(SpHbox4/5, HeHbox9) and asteroids (PeHbox9, AM5). Be-yond these four genes, however, the trees disagree and thesituation becomes unclear. HgHbox5 is included in thisgroup in Figs. 2 and 3, but not in Fig. 4; CJHbox3 and HoxB8are part of this PG in Fig. 4.PG6. As with PG4/5, the echinoid (HpHbox3, SpHox6) andasteroid (PeHbox3, AM4) paralogues clade consistently to-gether in all trees, while the affinities of CJHbox3 and SS11(Figs. 2 and 3 compared to Fig. 4) are less clear. No potentialparalogues have been reported from holothuroids.PG7. Six gene products, from asteroids (AM7), echinoids(SpHox7, HpHbox1, HeHbox1), ophiuroids (SSHbox1), andcrinoids (CJHbox1), clade together in all trees, comprisingPG7. In Fig. 3 an additional crinoid gene, CJ28, is associatedwith this group, along with human HoxB8. No paralogue hasyet been reported from holothuroids.PG8. The genes that comprise PG8 clade are clearly to-gether in all three trees and include representatives from allechinoderm classes except holothuroids. In all three treesechinoderm PG8 was associated with a clade comprised ofmedial human PG genes, which were evidently more similarto each other than to echinoderm medial PG genes.

Posterior paralogous groupsPG9/10. Figures 2 and 4 failed to resolve potential paral-ogues of PG9/10 into a single clade. In these trees the twoechinoid genes, SpHox9/10 and HpHbox4, form a group dis-tinct from the other PG9/10 genes. However, all five classesof echinoderms are represented in this group in Fig. 3, alongwith human HoxA9. Two of these genes that clade togetherin Fig. 2 are from the same crinoid, CJHbox4 and CJHbox4a.Mito and Endo (2000) concluded that these two genes wereallelic variants and thus not representative of different paral-ogous groups. SS12’s association with this group is ex-tremely weak (Fig. 3).PG11/13a. Six genes from four classes of echinoderms areincluded in this PG in Figs. 2 and 3. Interestingly, this groupis associated with different human Hox genes, HoxA9 in Fig.2 and HoxA10 in Fig. 3. Maximum likelihood analysis (Fig.4) has grouped only four of these gene products together. Nopotential paralogues are known from asteroids.PG11/13b. The same seven genes, from four echinodermclasses, clade together to form PG11/13b with all three meth-ods of analysis. The ophiuroid gene SSHbox10 is included inthis clade in the maximum likelihood analysis (Fig. 4) butnot in either of the others. Two holothuroid gene productsare associated with this PG.PG15. Mito and Endo (2000) found that asteroid AM9, oph-iuroid SS9 (along with SS9a and SS9.2) and crinoid CJ9

Fig. 1. Multiple alignment of all 84 homeodomain fragments (24aa each) used in the analyses, showing consensus and similarity.Alignment by CLUSTALW(accurate) using BLOSUM30. Spe-cies abbreviations: AM, Asterina minor (asteroid); Bf, Branchio-stoma floridae (amphioxus); CJ, Oxycomanthus japonicus (crinoid);He, Heliocidaris erythrogramma (echinoid); Hg, Holothuria glab-errima (holothuroid); Hp, Holopneustes purpurescens (echinoid);Hs, Homo sapiens (human); Pe, Patiriella exigua (asteroid); Sp,Strongylocentrotus purpuratus (echinoid); SS, Stegiophiura sladeni(ophiuroid).

Long and Byrne Echinoderm Hox gene cluster evolution 305

genes formed a group they termed HboxP9. Our analysesadd another asteroid gene (PeHbox15) and a holothuroidgene (HgHbox13) to this apparently new PG, which we havegiven the interim name of PG15 (until this gene’s position inthe echinoderm Hox cluster can be determined with cer-tainty). The three ophiuroid gene products that clade withthis group, SS9, -9a, and -9.2, might be trans-paralogues(Mito and Endo 2000). There is no echinoid paralogue.

DISCUSSION

A combined method approach to analysis of echinoderm par-tial Hox homeodomain sequences revealed strong consisten-cies between trees, particularly with regard to compositionof the echinoderm PGs. Given the problems inherent in phy-logenetic analysis of such a short, highly conserved, andfunctionally constrained set of sequences, this strong consis-

Fig. 2. Unrooted consensustree (parsimony analysis, 100bootstrap replicates) describ-ing relationships among 68echinoderm Hox homeodo-main amino acid sequencesand 13 vertebrate Hox geneproducts, with three deutero-stome engrailed gene prod-ucts as outgroup. See Methodsfor details of tree construc-tion. Numbers indicate thepercentage of times a parti-tion was supported by thebootstrap (unlabeled nodes�30%). See Fig. 1 captionfor species abbreviations ingene product names.


tency is encouraging. Topological similarity between treesconstructed using different methods of estimation was con-sidered by Kim (1993) to be indicative of accurate recon-struction. We conclude from these analyses that there are atleast 11 Hox PGs in echinoderms, only 10 of which occur inechinoids (Martinez et al. 1999).

Conventionally accepted bootstrap support values (50–70%) might be too stringent to resolve relationships betweenhighly conserved sequences such as Hox genes, particularlyin parsimony analyses where only differing amino acid posi-tions are informative (Kourakis and Martindale 2000). Addi-tionally, the data set is very large, and as a general rule the

Fig. 3. Unrooted phylogram(neighbor-joining analysis,1000 bootstrap replicates) ofrelationships among 68 echi-noderm Hox homeodomainamino acid sequences and 13vertebrate Hox gene prod-ucts, with three deuterostomeengrailed gene products asoutgroup. See Methods fordetails of tree construction.Numbers indicate the percent-age of times a partition wassupported by the bootstrap(unlabeled nodes �30%).Branch lengths are drawn toscale; bar indicates changesper amino acid position. SeeFig. 1 caption for species ab-breviations in gene productnames.


more sequences involved in an analysis, the greater numberof possible clading patterns and consequently a generallylower bootstrap support for particular nodes (Kourakis andMartindale 2000). Hence we believe that the generally lowbootstrap values given in Figs. 2 and 3 should not be takenas indicators of inaccuracy.

Branching order within clades is probably not phyloge-netically significant (Irvine et al. 1997), and ideally eachclade should be considered a group of paralogues. While inmost cases paralogy assignment and hence composition ofthe various echinoderm PGs is fairly clear, the relationshipsbetween echinoderm PGs are not. Sequence outside the ho-

Fig. 4. Unrooted maximumlikelihood tree describing re-lationships among 68 echi-noderm Hox homeodomainamino acid sequences and13 vertebrate Hox gene prod-ucts (without species prefix),with three deuterostomeengrailed gene products asoutgroup. See Methods fordetails of tree construction.See Fig. 1 caption for speciesabbreviations in gene prod-uct names.


meodomain—information not currently available from suffi-cient echinoderm taxa to be phylogenetically informative—will be necessary to give further insight into evolutionary re-lationships between paralogous groups of echinoderm Hoxgenes.

According to these analyses CJHbox3, SSHbox10,CJHbox10, HgHbox2, CJ28, SS11, SS12, and SS13 are ofuncertain paralogy. It is possible that this uncertainty resultsfrom the PCR-based methods used to obtain the sequences orthe phylogenetic techniques employed in the current analy-sis. Some alternative explanations are discussed below.

CJ28 and CJHbox3 may be the result of class-specific du-plication of the PG6 paralogue in the crinoid ancestor (Mitoand Endo 2000). CJHbox3, remaining identifiably medial,may continue the PG6 function, while CJ28 shows some se-quence divergence and hence possibly functional divergenceor loss of function (Sharkey et al. 1997; Mito and Endo2000). The vague association of CJ28 with different paralo-gous groupings in the different trees is indicative of its po-tentially pseudogenic nature. Alternatively Mito and Endo(2000) proposed that CJ28 may represent a new medialparalogous group, one that does not exist in echinoids andthat has not yet been identified in any other deuterostome.

SS11 is another medial-group Hox gene product whoseposition within the Hox cluster is difficult to determine. Mitoand Endo (2000) speculated that SS11 may represent a de-scendant of the ancestral deuterostome PG4 or PG5, one ofwhich has been lost in echinoids and possibly in other echi-noderm classes also. Two other ophiuroid Hox fragments,SS12 and SS13, do not consistently clade with any otherechinoderm Hox gene products. Their paralogy is unknown,although their partial homeobox sequences suggest that theyare probably posterior. Mito and Endo (2000) felt that theymight represent ophiuroid-specific expansions of the poste-rior Hox cluster, or alternatively that a second Hox gene cluster,possibly with only a few complements, exists in ophiuroids.This might also account for the apparent trans-paralogy ofSS9 and SS9.2 (Mito and Endo 2000).

HgHbox2 was classed as a PG2 gene by Mendez et al.(2000). Maximum likelihood and neighbor-joining methodsgroup it with PG3 genes (Figs. 3 and 4), although not asclosely as HgHbox3, while parsimony associates HgHbox2with SS13 (Fig. 2). It is possible that HgHbox2 is a divergedholothuroid PG2 gene, but more sequence will be necessaryto confirm its paralogy. Similarly it seems likely thatCJHbox10 and SSHbox10 are paralogues of PG11/13a (as inFigs. 2 and 3), despite the ambiguity presented by Fig. 4.

The failure of paralogues of PG9/10 to form a single co-herent PG in two of the three trees (Figs. 2 and 4) is interest-ing. It is possible that this incoherency may be indicative of theexistence of two very closely related PGs, with very similarhomeodomains, only one of which is apparently known fromechinoids (Martinez et al. 1999). Consistent with this possi-

bility is the occurrence of two genes from the same ho-lothuroid, HgHbox11 and HgHbox12, within PG11/13b,with all three methods of analysis. Mendez et al. (2000) in-terpreted HgHbox11 and HgHbox12 as representatives oftwo different posterior holothuroid PGs. Alternatively,HgHbox11 and HgHbox12 may be descendants of a recentduplication event of the holothuroid PG11/13b gene. Thesespeculations will be settled by the advent of more sequencedata.

Adding weight to the validity of these phylogenetic anal-yses is the intelligible clading of vertebrate Hox gene prod-ucts with their echinoderm paralogues—HoxA1 with PG1,HoxB2 with PG2, HoxB3 with PG3; medial human Hox ho-meodomains with medial echinoderm Hox homeodomains;and posterior human Hox homeodomains with posteriorechinoderm Hox homeodomains. It is noticeable that whilethe anterior echinoderm PGs are each clearly paralogouswith their counterpart vertebrate anterior PGs and medialechinoderm PGs are partially so, the relationships betweenspecific posterior echinoderm PG and specific posterior ver-tebrate PG are not at all clear.

Thus echinoderms seem to conform to the previously ob-served chordate pattern of anterior conservation and poste-rior diversification of gene sequence within the Hox cluster(van der Hoeven et al. 1996; Holland and Garcia-Fernandez1996; Sharkey et al. 1997; Mito and Endo 2000; Ferrier et al.2000). This apparently deuterostome phenomenon has beentermed “posterior flexibility” (Holland and Garcia-Fernan-dez 1996; Mito and Endo 2000; Ferrier et al. 2000) and is instriking contrast to the protostome situation, in which reso-lution of posterior PG Hox genes is possible across phyla(Ferrier et al. 2000). Posterior flexibility could be an evolu-tionary consequence of another interesting feature observedin chordate posterior PGs, termed “laxitas terminalis” by vander Hoeven et al. (1996): the more posterior a PG, the greatersequence divergence there will be between its members.Echinoderms also appear to exhibit laxitas terminalis (Long,unpublished). The functional and evolutionary consequencesof both posterior flexibility and laxitas terminalis in deu-terostomes are intriguing topics for speculation. Ferrier et al.(2000) proposed that functional variability might be toler-ated in posterior PGs because, in higher vertebrates at least,posterior PG Hox gene expression is not linked to patterningof fundamental vertebrate features. Conversely, Holland andGarcia-Fernandez (1996) speculated that posterior PG diver-sification among vertebrates could have been instrumental indiversification of the vertebrate body plan.

The question of whether duplication and diversificationof posterior PG genes has been important in evolution of thehighly derived echinoderm body plan is currently under in-vestigation. Arenas-Mena et al. (1998) examined the expres-sion of posterior PG echinoid Hox genes during rudiment de-velopment in larvae of Strongylocentrotus purpuratus, and


they found many examples of apparent co-options of poste-rior Hox genes to echinoid-specific functions. As discussedby Mooi and David (1998), echinoids have the most radicalmetamorphosis of all echinoderms, discarding nearly all oftheir larval (extra-axial) tissues and forming a juvenile bodycomposed almost entirely of axial elements. It is possiblethat a clearer picture of the ancestral patterns of echinodermHox expression might be gained from examination of devel-opment of one of the less-derived classes, such as the aster-oids or crinoids, in which the metamorphosed juvenile bodyis composed of both axial and extra-axial (larval-derived) el-ements.

Putative Hox gene clusters for all of the echinodermclasses, as described by Figs.1 through 4, are summarized inFig. 5. The complete complement is known only for echi-noids (Martinez et al. 1999), and the possibility that as yetunidentified Hox genes exist in clusters from the other fourclasses is indicated by question marks in Fig. 5. From previ-ous work and from the trees presented in this paper it is clearthat there are at least 11 PGs in the echinoderm Hox cluster.

This is a minimum estimate of the number of PGs; prob-lematic echinoderm PGs such as PG4/5, PG9/10, and moreposterior groups may well prove (with more sequence data)to be polyphyletic. HgHbox2, SS11, CJ28, SS12, and SS13are of uncertain exact paralogy but are depicted with ante-rior, medial, or posterior affinities, as described by their po-sition in the trees (Figs. 2 through 4). The tentative groupingdesignated HboxP9 by Mito and Endo (2000) is shown bythese analyses to be present in four of the five echinodermclasses, and is given the interim designation PG15. This newposterior PG does not appear from phylogenetic analyses toinclude Ferrier et al.’s (2000) sixth amphioxus posterior PGHox gene, AmphiHox14 (Long, personal observation), andthus may be an echinoderm-specific PG.

The ancestral deuterostome may have possessed two pos-terior Hox genes, one of the PG9/10 type and one of thePG11/13 type (Martinez et al. 1999). It is thought that thefive posterior PG observed in vertebrates today resulted fromduplication and diversification of these ancestral deutero-stome genes (Garcia-Fernandez and Holland 1994). The lackof obvious correspondence between vertebrate and echino-derm posterior PG genes indicates that at least some of thisexpansion and diversification has occurred subsequent to theseparation of echinoderms from the chordate lineage.

Mapping of the echinoderm posterior PG genes onto Lit-tlewood et al.’s (1997) echinoderm class phylogeny (based

Fig. 5. Hox gene clusters in echinoderms (graphic summary ofdata presented in Figs. 1–4). PG nomenclature follows Martinezet al. (1999). Examples of the genes for each PG for each echino-derm class are given. Question marks indicate paralogues that arecurrently unknown but may be present in a class.


on both molecular and morphological data) reveals that theancestral echinoderm may have possessed at least four pos-terior PG genes (Fig. 6; Mito and Endo 2000). This conclu-sion hinges upon the fact that although crinoids are consid-ered basal group echinoderms, and holothuroids morederived; both classes possess paralogues of PG9/10, PG11/13a, PG11/13b, and PG15 (Fig. 5). Examination of Fig. 5demonstrates that the suites of posterior PG genes currentlyknown to be possessed by asteroids, ophiuroids, and echi-noids are comprised of different combinations of the ances-tral four genes.

This model of posterior PG Hox gene evolution in echin-oderms indicates that in the 500 million years since the fiveextant echinoderm classes diverged (Paul and Smith 1984;Littlewood et al. 1997), those four ancestral posterior Hoxgenes appear to have evolved in different directions in thedifferent classes. Specifically, ophiuroids may have dupli-cated one or more of the genes, which have subsequently di-verged in sequence and potentially also in function (generat-ing SS13 and SS12), while echinoids appear to have lost aHox gene entirely (PG15; see Fig. 5). Loss of a Hox gene hasbeen speculatively associated with a change in body plan inmetazoan evolution and has been illustrated by the apparentlack of Abd-B and corresponding absence of abdominal seg-ments in barnacles (Mouchel-Vielh et al. 1998). However, itis difficult to correlate the loss of a posterior Hox gene withloss of any structures in echinoids. Given the potential forco-option and redeployment of homeobox genes discussedby Lowe and Wray (1997) and Davidson and Ruvkun(1999), among others, such speculation is likely to be fruit-less without expression data.

Wray and Lowe (2000) suggested that the evolution ofechinoderms has occurred via modifications in roles and ex-pression domains for a pre-existing set of developmentalregulatory genes. The analysis presented in this paper posits

a further mechanism for evolution in echinoderms, via class-specific duplication, divergence, and loss of Hox genes.

AcknowledgmentsWe thank Bernie Degnan and Buz Wilson for valuable commentson a draft. This work was supported by an Australian PostgraduateAward (S.L.) and a Large ARC (M.B.).

REFERENCES

Adachi, J., and Hasegawa, M. 1996. Molphy Version 2.3. Programs for Mo-lecular Phylogenetics Based on Maximum Likelihood. Computer Sci-ence monographs No. 28. Institute of Statistical Mathematics, Tokyo.

Arenas-Mena, C., Martinez, P., Cameron, R. A., and Davidson, E. H. 1998.Expression of the Hox gene complex in the indirect development of asea urchin. PNAS 95: 13062–13067.

Balavoine, G. 1998. Are Platyhelminthes coelomates without a coelom? Anargument based on the evolution of Hox genes. Amer. Zool. 38: 843–858.

Carroll, S. B. 1995. Homeotic genes and the evolution of arthropods andchordates. Nature 376: 479–485.

Davidson, E. H., and Ruvkun, G. 1999. Themes from a NASA workshopon gene regulatory processes in development and evolution. J. Exp.Zool. 285: 104–115.

de Rosa, R., et al. 1999. Hox genes in brachiopods and priapulids and pro-tostome evolution. Nature 399: 772–776.

Felsenstein, J. 1989. PHYLIP—Phylogeny inference package (Version3.2). Cladistics 5: 164–166.

Ferrier, D. E. K., Minguillon, C., Holland, P. W. H., and Garcia-Fernandez,J. 2000. The amphioxus Hox cluster: deuterostome posterior flexibilityand Hox14. Evol. Dev. 2: 284–293.

Garcia-Fernandez, J., and Holland, P. J. 1994. Archetypal organization ofthe Amphioxus Hox gene cluster. Nature 370: 563–566.

Gellon, G., and McGinnis, W. 1998. Shaping animal body plans in develop-ment and evolution by modulation of Hox expression patterns. BioEs-says 20: 116–125.

Hall, B. K. 1996. Baupläne, phylotypic stages, and constraint: why there areso few types of animals. Evolutionary Biology 29: 215–261.

Holland, P. W. H., and Garcia-Fernandez, J. 1996. Hox genes and chordateevolution. Devl Biol. 173: 382–395.

Irvine, S. Q., Warinner, S. A., Hunter, J. D., and Martindale, M. Q. 1997.A survey of homeobox genes in Chaetopterus variopedatus and analysisof polychaete homeodomains. Mol. Phyl. Evol. 7: 331–345.

Janies, D., and Mooi, R. 1999. Xyloplax is an asteroid. In M. D. CandiaCarnevali and F. Bonasoro (eds.). Echinoderm Research 1998. BalkemaPress, Rotterdam, pp. 311–316.

Kappen, C., Schughart, K., and Ruddle, F. H. 1989. Two steps in the evolutionof Antennapedia-class vertebrate homeobox genes. PNAS 86: 5459–5463.

Kim, J. H. 1993. Improving the accuracy of phylogenetic estimation bycombining different methods. Syst. Biol. 42: 331–340.

Kishino, H., Miyata, T., and Hasegawa, M. 1990. Maximum likelihood in-ference of protein phylogeny and the origin of chloroplasts. J. Mol.Evol. 30: 151–160.

Kourakis, M. J., and Martindale, M. Q. 2000. Combined-method phylogeneticanalysis of Hox and ParaHox genes of the Metazoa. J. Exp. Zool. (Mol.Dev. Evol.) 288: 175–191.

Lewis, E. B. 1978. A gene complex controlling segmentation in Drosophila.Nature 276: 565–570.

Littlewood, D. T. J., Smith, A. B., Clough, K. A., and Emson, R. H. 1997.The interrelationships of the echinoderm classes: morphological andmolecular evidence. Biol. J. Linn. Soc. 61: 409–438.

Long, S. C., Morris, V. B., and Byrne, M. 2000. Seven Hox gene sequencesfrom the asterinid starfish Patiriella exigua (Echinodermata: Asteroi-dea). Hydrobiologia 420: 95–98.

Lowe, C. J., and Wray, G. A. 1997. Radical alterations in the roles of ho-meobox genes during echinoderm evolution. Nature 389: 718–721.

Martinez, P., Rast, J. P., Arenas-Mena, C., and Davidson, E. H. 1999. Or-ganisation of an echinoderm Hox gene cluster. PNAS 96: 1469–1474.

Fig. 6. Model of posterior PG Hox gene evolution in echinoderms,following Littlewood et al.’s (1997) phylogeny of echinoderm classrelationships. Because crinoids and holothuroids have the same setof at least four posterior PG Hox genes, the ancestral echinodermmust also have had at least four. Note that the asteroid, ophiuroid,and echinoid suites of three genes contain different combinationsof the ancestral four genes. � indicates that undetected posteriorPG Hox genes may exist in the class; there are only three posteriorPG Hox genes in echinoids (Martinez et al. 1999).


Mendez, A. T., Roig-Lopez, J. L., Santiago, P., Santiago, C., and Garcia-Arraras, J. E. 2000. Identification of Hox gene sequences in the sea cu-cumber Holothuria glaberrima Selenka (Holothuroidea: Echinoderma-ta). Marine Biotechnology 2: 231–240.

Mito, T., and Endo, K. 1997. A PCR survey of Hox genes in the sea star As-terina minor. Mol. Phyl. & Evol. 8: 218–224.

Mito, T., and Endo, K. 2000. PCR survey of Hox genes in the crinoid andophiuroid: evidence for anterior conservation and posterior expansionin the echinoderm Hox gene cluster. Mol. Phyl. Evol. 14: 375–388.

Mooi, R., and David, B. 1998. Evolution within a bizarre phylum: homolo-gies of the first echinoderms. Am. Zool. 38: 965–974.

Morris, V. B., Brammall, J., Byrne, M., and Frommer, M. 1997. Hox-typeand non-Hox homeobox gene sequences in genomic DNA of the sea ur-chin Holopneustes purpurescens. Gene 201: 107–110.

Mouchel-Vielh, M., Rigolot, C., Gibert, J.-M., and Deutsch, J. S. 1998.Molecules and the body plan: the Hox genes of cirripedes (Crustacea).Mol. Phyl. Evol. 9: 382–389.

Paul, C. R. C., and Smith, A. B. 1984. The early radiation and phylogeny ofechinoderms. Biol. Rev. 59: 443–481.

Popodi, E., Kissinger, J. C., Andrews, M. E., and Raff, R. A. 1996. Sea ur-chin Hox genes: insights into the ancestral Hox cluster. Mol. Biol. Evol.13: 1078–1086.

Pradel, J., and White, R. A. H. 1998. From selectors to realizators. Int. J.Dev. Biol. 42: 417–421.

Rowe, F. W. E., Baker, A. N., and Clark, H. E. S. 1988. The morphology,development and taxonomic status of Xyloplax Baker Rowe & Clark1986 (Echinodermata: Concentricycloidea) with the description of anew species. Proc. Roy. Soc. London B 233: 431–459.

Schubert, F. R., Nieselt-Struwe, K., and Gruss, P. 1993. The Antennapedia-type homeobox genes have evolved from three precursors separatedearly in metazoan evolution. PNAS 90: 143–147.

Sharkey, M., Graba, Y., and Scott, M. P. 1997. Hox genes in evolution: pro-tein surfaces and paralog groups. Trends in Genetics 13: 145–151.

Thompson, J. D., Higgins, D. G., and Gibson, T. J. 1994. CLUSTAL W: im-proving the sensitivity of progressive multiple sequence alignmentthrough sequence weighting, position-specific gap penalties and weightmatrix choice. Nuc. Acids Res. 22: 4673–4680.

van der Hoeven, F., Zakany, J., and Duboule, D. 1996. Gene transpositionsin the HoxD complex reveal a hierarchy of regulatory controls. Cell 85:1025–1035.

Wray, G. A., and Lowe, C. J. 2000. Developmental regulatory genes andechinoderm evolution. Systematic Biology 49: 28–51.

Zhang, J., and Nei, M. 1996. Evolution of Antennapedia-class homeoboxgenes. Genetics 142: 295–303.

evolution of the echinoderm hox gene cluster

Documents