sign in sign up
<<
Home , Aschelminth, Bilateria, Eumetazoa, Rhabditophora

Syst. Biol. 49(1):130–142, 2000

Early of the

BERNHARD HAUSDORF Zoologisches Institut und Zoologisches Museum der Universität Hamburg, Martin-Luther-King-Platz 3, D-20146 Hamburg, Germany; E-mail: [email protected]

Abstract.—The phylogeny of the Bilateria and especially the early steps in the evolution of the bila- terian bauplan are still a controversial topic. In this context the relationships of the platyhelminths

and the play a crucial role. Previous molecular studies of the relationships of these Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 groups, which were based on 18S ribosomal DNA sequences, yielded conflicting results. In the present study a new framework is developed for the phylogenetic analysis of bilaterian relation- ships, using concatenated amino acid sequences of several nuclear genes. In this analysis, the rhab- ditophoran platyhelminths are probably the of all other analyzed Bilateria, the Eubila- teria, which are characterized by a one-way intestine with an . The Eubilateria are split into the lineage and the coelomates. The phylogenetic results of the present study indicate that genetic features found in the model organisms Caenorhabditis and Drosophila might be found in all Eubilateria. Estimations of the divergence times show that the major bilaterian phyla did not origi- nate in an explosive radiation during the but rather that the Bilateria have a several hun- dred million long history. [Bilateria; Coelomata; ; molecular phy- logeny; Nematoda; Platyhelminthes.]

The phylogeny of the Bilateria and espe- differences in the base composition among cially the early steps in the evolution of the taxa (Hasegawa and Hashimoto, 1993; bilaterian bauplan are still a controversial Abouheif et al., 1998), drastic differences in topic. In this context the relationships of the substitution rates among taxa that cause platyhelminths and the nematodes play a long branch attraction (Carmean and crucial role because of their supposed basal Crespi, 1995), and conflict between the trees position. But even if only the phylogenetic based on 18S rDNA and trees estimated by relationships of these groups and two other using other data (Huelsenbeck and Bull, major bilaterian phyla, the and 1996). Moreover, 18S rDNA does not have the , are considered, the tree enough informative positions for a robust topologies of the various phylogenetic re- reconstruction of the metazoan phylogeny constructions do not correspond, neither (Philippe et al., 1994), the alignment of the between those reconstructions based on rDNAs is often more ambiguous than that morphological characters (Figs. 1a–e) nor of protein-coding genes (Winnepenninckx between those based on 18S ribosomal and Backeljau, 1996), and the strong influ- DNA (rDNA) (Figs. 1f–j). ence of the rRNA secondary structure on Because of problems with the interpreta- the evolution of its sequence is difficult to tion and establishment of homology for model (Dixon and Hillis, 1993). morphological characters across metazoan Conservative protein-coding genes might phyla, the phylogenetic analysis of con- be more suitable for the reconstruction of served genes offers a promising approach. the metazoan phylogeny (Hasegawa and So far, only 18S rDNA has been used widely Hashimoto, 1993; Maley and Marshall, for phylogenetic analyses of the metazoan 1998), and a growing number of studies phyla (e.g., Riutort et al., 1993; Philippe et are based on individual protein-coding al., 1994; Raff et al., 1994; Winnepenninckx genes (e.g., Sidow and Thomas, 1994; et al., 1995; Aguinaldo et al., 1997). How- McHugh, 1997; Nikoh et al., 1997; Bor- ever, the inconsistent results of analyses chiellini et al., 1998). However, these stud- based on 18S rDNA (Figs. 1f-j) are disap- ies show that individual protein-coding pointing. The reliability of 18S rDNA as a genes do not include enough information phylogenetic marker for deep divergences for a robust reconstruction of metazoan has repeatedly been questioned because of phylogeny.

130 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 131

plications do not affect the topology of the tree, and one paralog was chosen randomly for further analyses; otherwise, the genes were discarded. Further analyses with Sac- charomyces as an additional outgroup were performed to test the robustness of the tree. The accession numbers of the analyzed se- quences and the number of amino acids FIGURE 1. Comparison of the topologies of platy- used for the phylogenetic analyses are helminths (P), nematodes (N), arthropods (A), and listed in Table 1. Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 chordates (C) in published trees based on morphologi- The protein sequences were aligned by cal characters (a–e) or on 18S rDNA sequences (f–j). (a) using the divide-and-conquer multiple se- Hennig, 1979. (b) Schram, 1991. (c) Eernisse et al., 1992. (d) Backeljau et al., 1993. (e) Nielsen, 1995. (f) Riutort et quence alignment program version 1.0 al., 1993. (g) Philippe et al., 1994 (in this analysis the (Tönges et al., 1996) with the BLOSUM 62 flies do not cluster with the other arthropods but with substitution matrix (Henikoff and Henikoff, the nematodes). (h) Raff et al., 1994. (i) Winnepen- 1992). All positions with gaps and all posi- ninckx et al., 1995. (j) Aguinaldo et al., 1997. tions adjacent to gaps to the first site with a conserved amino acid in all analyzed taxa In the present study a new framework for were excluded in the phylogenetic analy- the phylogenetic analysis of bilaterian rela- ses. The sequence alignments may be ob- tionships is developed, using concatenated tained from the author, or from the System- amino acid sequences of several nuclear atic web site (www.utexas.edu/ genes. Furthermore, the divergence times ftp/depts/systbiol/). of the major bilaterian groups are estimated The program package PHYLIP version by means of maximum likelihood trees con- 3.57c (Felsenstein, 1995) was used for maxi- structed with the clocklike evolving se- mum parsimony (Fitch, 1971) and neigh- quences of the data set. bor-joining (Saitou and Nei, 1987) analyses. The distances used for the neighbor-joining analyses were calculated with the PAM 250 MATERIALS AND METHODS matrix (Dayhoff et al., 1978). The program Because there are fewer sequences of package PUZZLE version 4.0 (Strimmer platyhelminths than of nematodes, arthro- and von Haeseler, 1996) was used for maxi- pods, or , the GenBank was mum likelihood (Felsenstein, 1981; Kishino screened for proteins from platyhelminths et al., 1990) analyses, either with the PAM first. The fluke is the platy- 250 matrix or with the BLOSUM 62 matrix. helminth with most entries. The sequences Unless otherwise stated, the calculations available from Schistosoma were screened were performed by assuming uniform rates for genes that are also known in Arabidopsis, over all sites. If rate heterogeneity was al- Caenorhabditis, Drosophila, Mus, and Homo. lowed for, the rates were assumed to be Arabidopsis is used as outgroup. Caenorhab- gamma-distributed with eight rate cate- ditis, Drosophila, and Mus and Homo are the gories; the gamma distribution parameter representatives of the nematodes, the arth- alpha was estimated from the data set. All ropods, and the chordates, respectively, other program options, unless noted, were having the most entries in the data bases. default. Confidence values for internal About 30 genes were found to be known in branches were established by bootstrap- all of the mentioned taxa and are so con- ping (Felsenstein, 1985; with 1,000 replica- served that an unambiguous alignment is tions) for maximum parsimony and neigh- possible. Many of these genes have several bor-joining analyses. Quartet puzzling paralogs, at least in some taxa. All genes provided the reliability values for maxi- with paralogs were analyzed phylogeneti- mum likelihood analyses (Strimmer and cally to determine whether the paralogs von Haeseler, 1996). Furthermore, alterna- from each form monophyletic tive topologies were evaluated by the groups. When this is the case, the gene du- Kishino-Hasegawa test (Kishino and Hase- TABLE 1. GenBank/EMBL accession numbers of the sequences and numbers of amino acids used for the phylogenetic analyses.

Amino acids used in analyses Without With Saccharomyces Saccharomyces Arabidopsis Saccharomyces Schistosoma Caenorhabditis Drosophila Mus Homo Aldolase 297 297 X53058 — L38658 D83738 M98351 Y00516 M11560 Calreticulin 253 253 U27698 — L24159 X59589 X64461 X14926 M84739 Elongation factor 1a 437 428 X16432 U51033 Y08487 U40935 X06869 L26479 J04617 Enolase 384 361 X58107 J01322 U33177 Z68318 X17034 X52379 M14328 Glyceraldehyde-3- 267 264 X98130 J01324 L09549 X52674 M11254 M32599 X01677 phosphate dehydrogenase 3-Hydroxy-3-methylglutaryl 168 160 L19261 M22002 M27294 U28991 M21329 M62766 M11058 coenzyme A reductase Phosphoglycerate kinase 342 332 U37701 J01342 L36833 U88169 Z14029 M15668 V00572 Ribosomal protein L13E 165 123 X75162 Z47071 U57003 U88308 X77926 U28917 X64707 t-Complex polypeptide-1a 511 433 D11351 X85021 U55769 U07941 M21159 D10606 X52882

Triose phosphate isomerase 223 201 U02949 J01366 M83294 U23081 X57576 X53333 X69723 Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 October 01 on guest by https://academic.oup.com/sysbio/article/49/1/130/1678667 from Downloaded 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 133 gawa, 1989) with use of the program pack- paralogs of the same phylum. The 10 genes age PUZZLE (Strimmer and von Haeseler, are fructose-bisphosphate aldolase (class I), 1996) and the PAM 250 matrix. calreticulin, elongation factor 1a , enolase, The hypothesis that the amino acid sub- glyceraldehyde-3-phosphate dehydrogenase, stitution rates are equal among lineages 3-hydroxy-3-methylglutaryl coenzyme A re- was tested with a likelihood ratio test ductase, phosphoglycerate kinase, riboso- (Felsenstein, 1981), again using the pro- mal protein L13E, t-complex polypeptide- gram PUZZLE for each gene and for differ- 1a , and triose phosphate isomerase. The ent substitution and rate heterogeneity aldolase and the enolase genes have been models. Only those genes for which the duplicated in the evolution of the chor- Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 clock hypothesis cannot be rejected at a sig- dates. Mice as well as humans exhibit sev- nificance level of 5% were used for the cal- eral clearly identifiable paralogs of these culations of divergence times. The calcula- genes. Therefore, the same ortholog from tions of divergence times were based on the mice and humans has been included in the maximum likelihood branch lengths of the analysis. No class I fructose-bisphosphate trees estimated under the clock hypothesis. aldolase and no calreticulin orthologs are The divergence between rodents and pri- known from Saccharomyces. Therefore, these mates was the only available reference for two genes were coded as unknown for the datings. Several recent molecular clock yeast in the analyses with Saccharomyces as estimations of the date of this divergence, an additional outgroup. based on different data and assumptions, The maximum likelihood and neighbor- vary from 100 million years before present joining analyses of the concatenated amino (MYBP) (Li et al., 1990), 104 MYBP (Hedges acid sequences of the 10 proteins resulted in et al., 1996), and 115 MYBP (Janke et al., the same tree topology (Figs. 2, 3). This tree 1997) to 125 MYBP (Janke et al., 1997). To topology is also found in the maximum get conservative estimates of divergence parsimony analysis with Arabidopsis as the times, the rodent-primate divergence is set only outgroup. If Saccharomyces is included at 100 MYBP. The 85-million--old fos- as an additional outgroup, Schistosoma and sils of hoofed (Archibald, 1996), Caenorhabditis are joined in the most-parsi- which originated only after the divergence monious tree. However, the majority-rule of rodents and primates, show that 100 consensus tree of 1,000 bootstrap replicates MYBP is a reasonable estimate of the diver- with Saccharomyces as an additional out- gence time of rodents and primates. group analyzed with the maximum parsi- Of course, the available data do not allow mony method corresponds to the tree exact datings because of the small number found by other methods and not to the of clocklike evolving genes known from all most-parsimonious tree. examined taxa and because only a single The quartet puzzling reliability values calibration point was available. Moreover, obtained in the maximum likelihood analy- the true divergence times might be 25% ses imply a strong support for all internal older than those calculated from the conser- branches of this tree. In contrast, the boot- vative estimate of the rodent-primate diver- strap values for the neighbor-joining and gence at 100 MYBP. parsimony analyses indicate that especially the monophyly of the group that includes the nematodes, arthropods, and chordates RELATIONSHIPS OF MAJOR needs further corroboration. According to BILATERIAN PHYLA the Kishino-Hasegawa test (Table 2), many Ten nuclear protein-coding genes were of the alternative topologies cannot be re- found in the GenBank that fulfill the fol- futed at the 5% significance level. However, lowing conditions: (1) They must be known the test shows that the log-likelihood of the in at least Arabidopsis, Schistosoma, Caeno- tree proposed by Aguinaldo et al. (1997) rhabditis, Drosophila, Mus, and Homo; (2) they (Fig. 1j) is significantly (p < 0.05) smaller must be so conserved that an unambiguous than that of the maximum likelihood tree, if alignment is possible; (3) no paralogs are Saccharomyces is included as an additional allowed that do not form with other outgroup. 134 SYSTEMATIC BIOLOGY VOL. 49 Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021

FIGURE 2. Bilaterian phylogeny based on amino acid sequences of 10 nuclear genes. The branch lengths shown are maximum likelihood estimations calculated with the PAM 250 matrix and assuming uniform rates over all sites. The numbers on the internal branches represent (from top to bottom) the quartet puzzling reliability values for the maximum likelihood analyses with the PAM 250 matrix and the BLOSUM 62 matrix and the bootstrap values for neighbor-joining and maximum parsimony analyses. (a) With Arabidopsis as only outgroup. (b) With both Arabidopsis and Saccharomyces as outgroups.

Considering the sequence length, the low bootstrap values may appear disillusion- ing. However, one must consider that the divergences between the few taxa for which the analyzed sequences are available are very deep. Therefore, the terminal branches are very long and, thus, many multiple and parallel substitutions may have occurred along these branches. The statistical sup- port for the internal branches may increase if additional taxa will be sampled that break the long terminal branches (Gray- beal, 1998). The next steps in a suitable sam- pling strategy might be to include a rhab- ditophoran “turbellarian”, a nematomorph, a mollusc or an , and an in the analyses. The poor taxon sampling is the major dis- advantage of the amino acid sequence data set presented in comparison with the avail- able 18S rDNA data set. The major advan- tage of the concatenated amino acid se- quences is that they might include enough informative positions for a robust recon- struction of the metazoan phylogeny, if se- quences from additional taxa become avail- able. If the current data set does not include enough informative positions, it can easily be expanded by inclusion of the amino acid sequences of hundreds of other nuclear FIGURE 3. Temporal frame of the bilaterian phy- logeny. The estimations of the divergence times are protein-coding genes. In contrast, the 18S based on the maximum likelihood branch lengths cal- rDNA does not provide enough informa- culated with the PAM 250 matrix and assuming a uni- tion for a robust reconstruction of the meta- form rate over all sites (see Table 4). The name Coelo- zoan phylogeny despite a dense taxon sam- mata is used only as a taxon name and should not necessarily imply homology of the cavities. = pling (Philippe et al., 1994; Abouheif et al., Cambrian; V = Vendian; MYBP = million years before 1998), and that data set can be expanded present. without problems by only the few genes of 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 135

TABLE 2. Maximum likelihood analyses (with the PAM 250 matrix) of the 15 possible topologies of the platy- helminths (P), nematodes (N), arthropods (A), and chordates (C). D log L, log-likelihood differences between the maximum likelihood tree (for which the log-likelihood is given in parentheses) and the 14 other topologies as well as the standard deviation (s.d.) of these differences. Differences that are statistically significant at the 5% level according to the Kishino–Hasegawa test are marked with an asterisk.

Analyses without Saccharomyces Analyses with Saccharomyces D log L s.d. D log L s.d. (P,(N,(A,C))) (2 23428.95) (2 24538.82) (P,(A,(N,C))) 2 13.58 21.89 2 23.10 21.14 Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 (P,(C,(A,N))) 2 19.44 21.25 2 22.01 21.27 (A,(N,(P,C))) 2 72.71* 28.96 2 67.79* 28.25 (A,(P,(N,C))) 2 45.12 30.12 2 56.00* 28.09 (A,(C,(P,N))) 2 38.35 27.39 2 44.64 27.12 (N,(A,(P,C))) 2 61.71* 26.72 2 49.41 26.91 (N,(P,(A,C))) 2 12.57 16.96 2 4.15 15.88 (N,(C,(P,A))) 2 14.12 30.72 2 12.03 30.12 ((P,N),(A,C))) 2 19.63 16.36 2 0.58 16.44 ((P,A),(N,C)) 2 30.76 31.66 2 27.28 30.75 ((N,A),(P,C)) 2 72.40* 28.54 2 54.63 28.85 (((P,N),A),C) 2 43.56 27.42 2 52.95* 26.60 (((P,A),N),C) 2 36.78 32.40 2 44.82 30.60 (((N,A),P),C) 2 56.01 30.03 2 61.94* 28.29

the rDNA cluster, because only these genes mertodermatida as well have to be sepa- evolve according to the same model as the rated from the main group of the platy- 18S rDNA. helminths, the . However, If the individual protein-coding genes of the inconclusiveness of the 18S rDNA data the present data set are analyzed separately, is shown by the analysis of Campos et al. the phylogenetic results differ from gene to (1998), who favor the monophyly of the gene, similar to the analyses of Wang et al. Platyhelminthes and include the (1999). This confirms that the information and the on the basis of almost conserved in single protein-coding genes is the same data. In the present analysis, only insufficient for the phylogenetic analysis of a representative of the Rhabditophora, metazoan relationships and that the only Schistosoma, is included. Therefore, no con- promising approach to a robust resolution clusions about the monophyly of the platy- of metazoan relationships is to analyze a set helminths can be drawn, and the conclu- of several sequences from many taxa. sions presented about the phylogenetic According to the maximum likelihood relationships of the “plathyhelminths” ap- tree based on the present set of amino acid ply, strictly speaking, only to the Rhabdi- sequences, the rhabditophoran platyhel- tophora. minth lineage is the sister group of all other As are the and the , analyzed Bilateria, the Eubilateria (Figs. 2, which belong to the stem group of the 3). Recently, several analyses based on mor- Metazoa, the platyhelminths are character- phological characters and 18S rDNA se- ized by a compact (acoelomate) organi- quences have challenged the monophyly of zation (Ruppert, 1991), a blind-ending in- the platyhelminths (Haszprunar, 1996; Car- testine, and protostomy. Thus, the basal ranza et al., 1997; Zrzavy´ et al., 1998; Little- position of the platyhelminths within the wood et al., 1999; Ruiz-Trillo et al., 1999). Bilateria is consistent with the hypothesis According to these analyses, the Acoela and that the mentioned character states are an- perhaps also the Catenulida and the Ne- cestral within the Bilateria and that the one- 136 SYSTEMATIC BIOLOGY VOL. 49 way intestine is a derived character state of lineage (probably comprising some of the the Eubilateria (Hennig, 1979; Ax, 1985). smaller groups formerly included in the as- It is not necessary to assume that platy- chelminths) and in the arthropods. helminths are derived from ancestors with The phylogenetic relationships of the spacious coelom cavities by a secondary re- platyhelminths and of the nematodes could duction of these cavities in the adult (Siew- not be resolved on the basis of the 18S ing, 1980) or by progenesis from larval or rDNA sequences (Figs. 1f-j), partly because juvenile stages (Rieger, 1985; Balavoine, of the unusually high substitution rates of 1998). the 18S rDNA in platyhelminths and nema- Often the platyhelminths have been clas- todes (Aguinaldo et al., 1997; Balavoine, Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 sified with several other phyla (e.g., mol- 1997). Aguinaldo et al. (1997) tried to solve luscs, , arthropods) as or this problem by looking for platyhelminths Eutrochozoa because of the spiral quartet and nematodes with apparently low substi- cleavage and the formation by tution rates. They calculated the number of the 4d mesentoblast or one of its daughter substitutions per position from the last cells (Siewing, 1980; Nielsen, 1995; Valen- common ancestor of , using a tine, 1997). However, the tree favored in the cnidarian and an echinoderm as outgroups. present analyses is not consistent with a That means they assumed the protostomes Spiralia. If this tree is correct, the spi- are monophyletic and the are ral quartet cleavage is either a synapomor- an outgroup to the protostomes. However, phy of the Bilateria (perhaps except the several previous phylogenetic analyses— Acoela) or has been achieved convergently morphological (Hennig, 1979; Ax, 1985; in the platyhelminths and the Eutrochozoa. Schram, 1991) as well as molecular (Riutort The convergent origin of the spiral cleavage et al., 1993; Philippe et al., 1994; Sidow and is less likely, even if one considers that this Thomas, 1994; Winnepenninckx et al., cleavage pattern results in the thermody- 1995)—have shown that the protostomes namically most stable arrangement of the are not monophyletic and that the echino- blastomeres, because the formation of the derms are not an outgroup of the proto- mesoderm by the same blastomere can stomes. This conclusion is also supported hardly be explained. Therefore, a determi- by the present data. Therefore, the dis- nate spiral quartet cleavage was probably tances calculated by Aguinaldo et al. (1997) ancestral within the Bilateria and has secon- are not distances between the last common darily been modified in several lineages of ancestor of the protostomes (which is the the Eubilateria (Ax, 1985). last common ancestor of all Bilateria) and In the maximum likelihood tree, the Eu- the respective taxa but instead reflect to a bilateria split into the nematode lineage certain degree the convergent similarities of with a compact organization on the one the analyzed platyhelminths and nema- hand and the lineage of arthropods and todes with the annelids and priapulids, chordates on the other. This result is also which Aguinaldo et al. used for the calcula- supported by a four-cluster analysis of con- tions. Their results, that the Protostomia are catenated amino acid sequences of 18 nu- monophyletic, is a consequence of their bi- clear genes by Wang et al. (1999). ased taxon selection and reflects their basic The arthropods, the chordates, and sev- assumption. The exclusion of the appar- eral other bilaterian phyla (e.g., molluscs, ently more rapidly evolving taxa may have annelids) often have been classified as resulted in systematic errors in tree recon- Coelomata (Hennig, 1979). However, it is struction. In accordance with other recent still questionable whether the various analyses (Abouheif et al., 1998), the result of coelom cavities (or their rudiments) are the present study—of several nuclear pro- homologous (Ruppert, 1991; Nielsen, 1995). tein-coding genes that are characterized by Ecdysis, which has been taken for a more homogeneous substitution rates across synapomorphy of a supposed major clade the analyzed taxa—suggests that the 18S of the Metazoa, the (Eernisse et rDNA sequences alone are insufficient for a al., 1992; Aguinaldo et al., 1997), has proba- stable phylogenetic reconstruction of the bly evolved convergently in the nematode deep branches within the Metazoa. 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 137

The phylogenetic results of the present TIME FRAME OF BILATERIAN EVOLUTION analysis imply that genetic features found in the model organisms Caenorhabditis and Although some divergence times of met- Drosophila might be found in all Eubilateria, azoan groups have been estimated by means if the features have not been modified sec- of the molecular clock (Wray et al., 1996; ondarily. In contrast, common features of Feng et al., 1997; Nikoh et al., 1997; Ayala et Drosophila and mouse or human are not nec- al., 1998; Gu, 1998; Wang et al., 1999), the di- essarily present in other metazoans. This is vergence time between platyhelminths and true, even if the phylogenetic position of the eubilaterians has not been investigated so platyhelminths in the favored tree is incor- far. However, this divergence is of extraor- Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 rect. These important predictions of the gen- dinary importance, because it marks the be- erality of genetic features found in model ginning of the bilaterian radiation. There- organisms are contrary to the predictions of fore, datings of this and other evolutionary Aguinaldo et al. (1997) and highlight how divergences were estimated on the basis of important the knowledge of the correct the clocklike evolving genes of the present topology of the tree of life is. data set. The phylogenetic results also allow some The results of the likelihood ratio tests for conclusions on the evolution of the Hox equal amino acid substitution rates among complex. At least seven genes in the Hox lineages are listed in Table 3. In the analyses cluster of the rhabditophoran platyhel- with Arabidopsis as the only outgroup, the minths are orthologous to cognate groups clock hypothesis cannot be rejected for five of the coelomates (Bartels et al., 1993; Bal- or seven genes, the specific genes depend- avoine, 1997, 1998). Thus, there were at ing on the chosen substitution and rate het- least seven genes in the Hox cluster of the erogeneity models. In the analyses with common ancestor of the rhabditophoran Saccharomyces as an additional outgroup, platyhelminths and the coelomates, be- the aldolase and the calreticulin genes have cause the rhabditophoran platyhelminths to be omitted, because they are not known are inferred to be the sister group of the from Saccharomyces. Considering the re- other analyzed Bilateria (Fig. 2). Conse- maining genes, the clock hypothesis cannot quently, at least three Hox genes must have be rejected for only two, three, or five been lost in the lineage of the nematode genes. This is partly because of an acceler- Caenorhabditis, of which the Hox cluster in- ated substitution rate in Saccharomyces, cludes only four genes (Bürglin et al., 1991; which becomes apparent also in the phylo- Kenyon and Wang, 1991). These conclu- gram (Fig. 2b). Because of the few suitable sions are in agreement with the results of a genes available, the analyses that included phylogenetic analysis of the homeodomain Saccharomyces are less reliable than those sequences of the Hox genes of nematodes without Saccharomyces and are not consid- and coelomates (Zhang and Nei, 1996). Al- ered further. though the nematodes are more closely re- The estimations of the divergence times lated to the coelomates than are the platy- based on the maximum likelihood branch helminths (Fig. 2), the distances between lengths of the tree estimated under the the homeodomain sequences of the Caenor- clock hypothesis with the amino acid se- habditis Hox genes and the Hox genes of ei- quences of the clocklike evolving genes ther the platyhelminths or the coelomates (Table 4) vary widely, depending on the are greater than those between the respec- chosen substitution and rate heterogeneity tive sequences of the platyhelminths and models. Nevertheless, all estimates agree in the coelomates (Bartels et al., 1993). The loss two important points: The major diver- of Hox cluster genes and the divergent evo- gences occurred in the Proterozoic before lution of the remaining Hox genes in the ne- the Vendian, and these divergences were matode lineage might be the consequence spread over several hundred million of a different selection acting on the Hox years. The estimations based on uniform cluster as a result of the acquisition of a dif- rates among sites and those based on the ferent, strongly determinative development BLOSUM 62 substitution matrix and gamma- mode in the nematodes. distributed rates suggest that the diver- 138 SYSTEMATIC BIOLOGY VOL. 49

TABLE 3. Results of the likelihood ratio tests for equal amino acid substitution rates among lineages. The first symbol refers to the analyses without Saccharomyces, the second to those with Saccharomyces. +, the clock hypoth- esis has not been rejected on a significance level of 5%; 2 , the clock hypothesis has been rejected on a significance level of 5%; *, data not available.

Gamma-distributed rates Uniform rate, all sites (8 categories) PAM 250 BLOSUM 62 PAM 250 BLOSUM 62 Aldolase +/* +/* +/* +/* Calreticulin +/* +/* +/* +/* Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 Elongation factor 1a +/2 +/2 2 /2 +/2 Enolase 2 /+ 2 /2 2 /2 2 /+ Glyceraldehyde-3- +/+ +/+ +/+ +/+ phosphate dehydrogenase 3-Hydroxy-3-methylgutaryl +/2 +/2 2 /2 2 /2 coenzyme A reductase Phosphoglycerate kinase 2 /2 2 /2 2 /2 2 /2 Ribosomal protein L13E +/+ +/+ +/+ +/+ t-complex polypeptide-1a +/2 +/+ 2 /2 +/+ Triose phosphate isomerase 2 /2 2 /2 +/2 +/+ Clocklike evolving genes 7/3 7/3 5/2 7/5

gence between the rhabditophoran platy- vative fixing of the rodent-primate diver- helminths and the Eubilateria is more than gence at 100 MYBP. Thus, the general con- a billion years old, that the divergences be- clusion of a Proterozoic origin of the Bila- tween the nematode lineage and the coe- teria long before the Cambrian and a lomates occurred at the beginning of the prolonged radiation of bilaterian phyla is Neoproterozoic, and that the divergences reliable, even if the confidence intervals between the and the might be large because of uncertainties in lineages occurred in the middle Neo- the calibration, which could not be calcu- proterozoic. The estimations based on the lated because only a single calibration point PAM 250 substitution matrix and gamma- was available. distributed rates with only five clocklike The estimation of a middle Neoprotero- genes suggest even older divergence times. zoic divergence between the arthropod and The estimations are underestimates rather the chordate lineages is in accord with than overestimates because of the conser- some of the previous datings based on

TABLE 4. Estimation of divergence times (6 s.d.), MYBP, based on the maximum likelihood branch lengths of the trees estimated under the clock hypothesis with the amino acid sequences of the clocklike evolving genes. The given standard deviations are only the errors in the estimation of the branch length and do not take into account uncertainties in the dating of the divergence between rodents and primates, which was used as reference.

Gamma-distributed rates Uniform rate, all sites (8 categories) PAM 250 BLOSUM 62 PAM 250 BLOSUM 62 Rodents–primates 100 100 100 100 Arthropods–chordates 715 6 24 700 6 23 1,171 6 53 810 6 28 Nematodes–coelomates 941 6 23 914 6 22 1,445 6 48 1,002 6 26 Platyhelminths–eubilaterians 1,114 6 27 1,061 6 25 1,678 6 60 1,156 6 31 – 1,334 6 37 1,263 6 35 2,575 6 123 1,544 6 53 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 139 other data, namely, 730 MYBP (Feng et al., a given gene. The more stringent the crite- 1997), 610 MYBP (Ayala et al., 1998), 736 ria for the exclusion of genes that are not MYBP (Ayala et al., 1998), and 830 MYBP clocklike evolving, the less the described (Gu, 1998). However, still other studies re- bias will influence the estimation of diver- sulted in older divergence times. The diver- gence times. gence of arthropods and chordates has been The global tree-based likelihood ratio test estimated at 993 MYBP by Wang et al. used in this study to examine the clocklike (1999) and , 1,200 MYBP by Wray et al. evolution of a gene is a stringent test (San- (1996). The divergence times of the proto- derson, 1998), because it examines the sub- stomes and estimated by stitution rates in all lineages at the same Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 Bromham et al. (1998), using 18S rDNA se- time, whereas the relative rate tests used in quences, are even distinctly higher. many other studies (Wray et al., 1996; Gu, The estimation of an early Neoprotero- 1998; Wang et al., 1999) compare only the zoic divergence between the nematode lin- rates in two lineages at one time and have eage and the coelomates is intermediary be- low statistical power (Ayala et al., 1998). tween the estimate by Feng et al. (1997), 815 The differences between the results of the MYBP, and that by Wang et al. (1999), 1,177 various studies of metazoan divergence MYBP. The same is true for the estimation times can partly be explained by the differ- of a middle or early Mesoproterozoic ani- ent exclusion criteria for the genes used in mal- divergence, which was estimated these studies. The studies using DNA se- at 1,200 MYBP by Feng et al. (1997) and at quences of the 18S rDNA and the mitochon- 1,576 MYBP by Wang et al. (1999). How- drial genes, which are known not to evolve ever, the reliability of the present estimate in a clocklike way (Ayala et al., 1998), re- of the divergence time of animals and sulted in the oldest divergence dates (Wray plants is questionable, because of the inabil- et al., 1996; Bromham et al., 1998), whereas ity to test whether the outgroup evolved at the estimates based on amino acid se- the same rate as the ingroups. quences of nuclear genes (Feng et al., 1997; The divergence time estimates calculated Nikoh et al., 1997; Ayala et al., 1998; Gu, by Wang et al. (1999) that are based on 1998; Wang et al., 1999), which have been those genes for which their relative rate examined for clocklike substitution rates, tests could not show rate heterogeneity are in better agreement with the estimates among lineages are always smaller than of the present study. The estimates based on their estimates based on all genes, i.e., in- rDNA are additionally biased by the corre- cluding genes shown to evolve with differ- lated evolution of paired sites, which prob- ent rates in different lineages. This indicates ably results in an increase in the count of in- a bias in the evolution of the substitution dependent changes between two sequences rates. The substitution rates within the ver- and thus causes an overestimation of diver- tebrates, which were used for the calibra- gence times (Bromham et al., 1998). tions, are generally smaller than those in The present divergence time estimations the other groups examined (Ayala et al., differ from previous estimations in that 1998). This can also be seen in the phylo- they are based on branch length. The use of grams based on all 10 amino acid sequences branch length has the advantage, that devi- used in this study (Fig. 2) and in compari- ations from the assumption of rate homo- son of the branch length between the trees geneity between lineages are corrected for calculated without clock assumption and by the maximum likelihood method. If di- with clock assumption, if only the clocklike vergence time estimations are based on evolving genes are considered. If the substi- pairwise distances, as in most previous tution rates within the vertebrates are gen- studies, differences between the pairwise erally smaller than those in the other exam- distances relevant for one divergence time ined groups, the divergence times based on (e.g., Mus-Drosophila, Homo-Drosophila) can calibrations with divergences are be corrected for only by simple averaging. overestimations. The overestimation will be Calculations based on branch length result the greater, the greater the difference be- in coherent estimates of divergence times, tween vertebrate and rates for whereas other methods may result in para- 140 SYSTEMATIC BIOLOGY VOL. 49 doxical estimates. For example, Gu (1998) radiation of the arthropods, some eutrocho- found that if the hydroxy-3-methylglutaryl zoans, and the deuterostomes (Runnegar, coenzyme A reductase is calibrated with an 1982). -fungus clock, the calculated diver- gence of arthropods and vertebrates is older than the animal-fungus divergence. ACKNOWLEDGMENTS Similarly, the average of the estimated di- I am grateful to the referees and the editors for help- vergence times of echinoderms and verte- ing me to develop a more critical outlook to the results and conclusions of this study. Furthermore, I thank brates based on mitochondrial DNA in the Martin Hingston (Hamburg) for correcting the English study of Bromham et al. (1998) is greater text. Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 than the average of their estimations based on mitochondrial DNA for the split of pro- tostomes and deuterostomes. REFERENCES Despite the differences in the applied ABOUHEIF, E., R. ZARDOYA, AND A. MEYER. 1998. Limi- methods and the differences in the absolute tations of metazoan 18S rRNA sequence data: Impli- cations for reconstructing a phylogeny of the animal time estimates, the previous molecular and inferring the reality of the Cambrian clock studies are compatible with the two explosion. J. Mol. Evol. 47:394–405. major conclusions of the present study, AGUINALDO, A. M. A., J. M. TURBEVILLE, L. S. LINFORD, namely, the occurrence of the divergences M. C. RIVERA, J. R. GAREY, R. A. RAFF, AND J. A. LAKE. between the major lineages of the Bilateria 1997. Evidence for a clade of nematodes, arthro- pods and other moulting animals. Nature 387:489– in the Proterozoic before the Vendian, and 493. the extension of these divergences over sev- ARCHIBALD, J. D. 1996. Fossil evidence for a late Cre- eral hundred million years. taceous origin of “hoofed” mammals. Science 272: The divergence time of the rhabdito- 1150–1153. AX, P. 1985. The position of the Gnathostomulida phoran platyhelminths and the Eubilateria and Platyhelminthes in the phylogenetic system of more than a billion years ago represents the the Bilateria. Pages 168–180 in The origins and rela- minimum age of the start of the bilaterian tionships of lower (S. Conway Morris, radiation, because the evolution of the Bila- J. D. George, R. Gibson, and H. M. Platt, eds.). teria probably began with a radiation of Clarendon Press, Oxford, UK. AYALA, F. J., A. RZHETSKY, AND F. J. AYALA. 1998. Ori- compact (acoelomate) phyla, of which only gin of the metazoan phyla: Molecular clocks confirm a few (e.g., rhabditophoran platyhelminths) paleontological estimates. Proc. Natl. Acad. Sci. USA are still extant. 95:606–611. The early emergence of bilaterians is con- BACKELJAU, T., B. WINNEPENNINCKX , AND L. DE BRUYN. 1993. Cladistic analysis of metazoan relationships: sistent with the interpretation of trace fos- A reappraisal. Cladistics 9:167–181. sils more than one billion year old as evi- BALAVOINE, G. 1997. The early emergence of platy- dence for the presence of bilaterians (Breyer helminths is contradicted by the agreement between et al., 1995; Seilacher et al., 1998) and with 18S rRNA and Hox genes data. C.R. Acad. Sci. Paris the hypothesis that the decline in the diver- Life Sci. 320:83–94. BALAVOINE, G. 1998. Are Platyhelminthes coelomates sity of stromatolites starting about one bil- without a coelom? An argument based on the evolu- lion years ago was due to grazing bilateri- tion of Hox genes. Am. Zool. 38:843–858. ans (Walter and Heys, 1985). Furthermore, BARTELS, J. L., M. T. MURTHA, AND F. H. RUDDLE. 1993. the estimated divergence time of the arthro- Multiple Hox/HOM-class in Platyhel- pods and chordates more than 700 MYBP is minthes. Mol. Phylogenet. Evol. 2:143–151. BORCHIELLINI , C., N. BOURY-ESNAULT, J. VACELET, AND consistent with the interpretation of some Y. LE PARCO. 1998. Phylogenetic analysis of the fossils as arthropods and echino- Hsp70 sequences reveals the monophyly of Metazoa derms (Conway Morris, 1993). and specific phylogenetic relationships between ani- The realization of a more than 500- mals and fungi. Mol. Biol. Evol. 15:647–655. BREYER, J. A., A. B. BUSBEY, R. E. HANSON, AND E. C. ROY million-year-long Precambrian history of III. 1995. Possible new evidence for the origin of the Bilateria exposes the Cambrian explo- metazoans prior to 1 Ga: Sediment-filled tubes from sion as the near coincidence of the appear- the Mesoproterozoic Allamoore Formation, Trans- ance of large soft-bodied animals, the min- Pecos Texas. Geology 23:269–272. eralization of preexisting organic substrates BROMHAM, L., A. RAMBAUT, R. FORTEY, A. COOPER, AND D. PENNY. 1998. Testing the in already complex soft-bodied animals of hypothesis by using a molecular dating technique. diverse bilaterian lineages, and a secondary Proc. Natl. Acad. Sci. USA 95:12386–12389. 2000 HAUSDORF—EARLY EVOLUTION OF THE BILATERIA 141

BÜRGLIN, T. R., G. RUVKUN, A. COULSON, N. C. HAWKINS, HUELSENBECK, J. P., AND J. J. BULL. 1996. A likelihood J. D. MCGHEE, D. SCHALLER, C. WITTMANN, F. ratio test to detect conflicting phylogenetic signal. MÜLLER, AND R. H. WATERSTON. 1991. Nematode Syst. Biol. 45:92–98. cluster. Nature 351:703. JANKE, A., X. XU, AND U. ARNASON. 1997. The com- CAMPOS, A., M. P. CUMMINGS, J. L. REYES, AND J. P. LA- plete mitochondrial genome of the wallaroo (Macro- CLETTE. 1998. Phylogenetic relationships of Platy- pus robustus) and the phylogenetic relationship helminthes based on 18S ribosomal gene sequences. among Monotremata, Marsupialia, and Eutheria. Mol. Phylogenet. Evol. 10:1–10. Proc. Natl. Acad. Sci. USA 94:1276–1281. CARMEAN, D., AND B. J. CRESPI. 1995. Do long branches KENYON, C., AND B. WANG. 1991. A cluster of Anten- attract flies? Nature 373:666. napedia-class homeobox genes in a nonsegmented CARRANZA, S., J. BAGUÑÀ, AND M. RIUTORT. 1997. Are animal. Science 253:516–517. the Platyhelminthes a monophyletic primitive group? KISHINO, H., AND M. HASEGAWA. 1989. Evaluation of Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 An assessment using 18S rDNA sequences. Mol. the maximum likelihood estimate of the evolution- Biol. Evol. 14:485–497. ary tree topologies from DNA sequence data, and CONWAY MORRIS, S. 1993. The fossil record and the the branching order in Hominoidea. J. Mol. Evol. early evolution of the Metazoa. Nature 361:219– 29:170–179. 225. KISHINO, H., T. MIYATA, AND M. HASEGAWA. 1990. DAYHOFF, M. O., R. M. SCHWARTZ, AND B. C. ORCUTT. Maximum likelihood inference of protein phy- 1978. A model of evolutionary change in proteins. logeny and the origin of chloroplasts. J. Mol. Evol. Pages 345–352 in Atlas of protein sequence and 31:151–160. structure, Volume 5, Suppl. 3 (M. O. Dayhoff, ed.). LI, W.-H., M. GOUY, P. M. SHARP, C. O’HUIGIN, AND National Biomedical Research Foundation, Wash- Y.-W. YANG. 1990. Molecular phylogeny of Roden- ington, D.C. tia, Lagomorpha, Primates, Artiodactyla, and Car- DIXON, M. T., AND D. M. HILLIS. 1993. Ribosomal nivora and molecular clocks. Proc. Natl. Acad. Sci. RNA secondary structure: compensatory mutations USA 87:6703–6707. and implications for phylogenetic analysis. Mol. LITTLEWOOD, D. T. J., K. ROHDE, AND K. A. CLOUGH. Biol. Evol. 10:256–267. 1999. The interrelationships of all major groups of EERNISSE, D. J., J. S. ALBERT, AND F. E. ANDERSON. 1992. Platyhelminthes: Phylogenetic evidence from mor- Annelida and Arthropoda are not sister taxa: A phy- phology and molecules. Biol. J. Linnean Soc. 66: logenetic analysis of spiralian metazoan morphol- 75–114. ogy. Syst. Biol. 41:305–330. MALEY, L. E., AND C. R. MARSHALL. 1998. The coming FELSENSTEIN , J. 1981. Evolutionary trees from DNA of age of molecular systematics. Science 279:505–506. sequences: A maximum likelihood approach. J. Mol. MCHUGH, D. 1997. Molecular evidence that echiu- Evol. 17:368–376. rans and pogonophorans are derived annelids. Proc. FELSENSTEIN , J. 1985. Confidence limits on phyloge- Natl. Acad. Sci. USA 94:8006–8009. nies: An approach using the bootstrap. Evolution NIELSEN, C. 1995. Animal evolution. Interrelation- 39:783–791. ships of the living phyla. Oxford Univ. Press, Ox- FELSENSTEIN , J. 1995. PHYLIP, version 3.57c. Depart- ford, UK. ment of Genetics, Univ. Washington, Seattle. NIKOH, N., N. IWABE, K. KUMA, M. OHNO, T. SUGIYAMA, FENG, D.-F., G. CHO, AND R. F. DOOLITTLE. 1997. Deter- Y. WATANABE, K. YASUI, Z. SHI-CUI, K. HORI, Y. mining divergence times with a protein clock: Up- SHIMURA, AND T. MIYATA. 1997. An estimate of di- date and reevaluation. Proc. Natl. Acad. Sci. USA vergence time of Parazoa and and that of 94:13028–13033. Cephalochordata and Vertebrata by aldolase and FITCH, W. M. 1971. Toward defining the course of triose phosphate isomerase clocks. J. Mol. Evol. 45: evolution: Minimum change for a specific tree topol- 97–106. ogy. Syst. Zool. 20:406–416. PHILIPPE, H., A. CHENUIL, AND A. ADOUTTE. 1994. Can GRAYBEAL, A. 1998. Is it better to add taxa or charac- the Cambrian explosion be inferred through mo- ters to a difficult phylogenetic problem? Syst. Biol. lecular phylogeny? Development 1994 (suppl.): 47:9–17. 15–25. GU, X. 1998. Early metazoan divergence was about RAFF, R. A., C. R. MARSHALL, AND J. M. TURBEVILLE. 830 million years ago. J. Mol. Evol. 47:369–371. 1994. Using DNA sequences to unravel the Cam- HASEGAWA, M., AND T. HASHIMOTO. 1993. Ribosomal brian radiation of the animal phyla. Annu. Rev. Ecol. RNA trees misleading? Nature 361:23. Syst. 25:351–375. HASZPRUNAR, G. 1996. Plathelminthes and Plathel- RIEGER, R. M. 1985. The phylogenetic status of the minthomorpha-Paraphyletic taxa. J. Zool. Syst. Evol. acoelomate organization within the Bilateria: A his- Res. 34:41–48. tological perspective. Pages 101–122 in The origins HEDGES, S. B., P. H. PARKER, C. G. SIBLEY, AND S. KUMAR. and relationships of lower invertebrates (S. Conway 1996. Continental breakup and the ordinal diversi- Morris, J. D. George, R. Gibson, and H. M. Platt, fication of and mammals. Nature 381:226– eds.). Clarendon Press, Oxford, UK. 229. RIUTORT, M., K. G. FIELD, R. A. RAFF, AND J. BAGUÑA. HENIKOFF, S., AND J. G. HENIKOFF. 1992. Amino acid 1993. 18S rRNA sequences and phylogeny of Platy- substitution matrices from protein blocks. Proc. helminthes. Biochem. Syst. Ecol. 21:71–77. Natl. Acad. Sci. USA 89:10915–10919. RUIZ-TRILLO, I., M. RIUTORT, D. T. J. LITTLEWOOD, E. A. HENNIG, W. 1979. Wirbellose I (ausgenommen HERNIOU, AND J. BAGUÑÀ. 1999. Acoel : Gliedertiere). Taschenbuch der Speziellen Zoologie, Earliest extant bilaterian metazoans, not members of 4th edition, Volume 2. Fischer, Jena, Germany. Platyhelminthes. Science 283:1919–1923. 142 SYSTEMATIC BIOLOGY VOL. 49

RUNNEGAR, B. 1982. The Cambrian explosion: Ani- VALENTINE, J. W. 1997. Cleavage patterns and the mals or fossils? J. Geol. Soc. Aust. 29:395–411. topology of the metazoan tree of life. Proc. Natl. RUPPERT, E. E. 1991. Introduction to the Acad. Sci. USA 94:8001–8005. phyla: A consideration of mesoderm, body cavities, WALTER, M. R., AND G. R. HEYS. 1985. Links between and cuticle. Pages 1–17 in Microscopic of the rise of the Metazoa and the decline of stromato- invertebrates, Volume 4 (F. W. Harrison and E. E. lites. Precambrian Res. 29:149–174. Ruppert, eds.). Wiley-Liss, New York. WANG, D. Y.-C., S. KUMAR, AND S. B. HEDGES. 1999. SAITOU, N., AND M. NEI. 1987. The neighbor-joining Divergence time estimates for the early history of method: A new method for reconstructing phyloge- animal phyla and the origin of plants, animals and netic trees. Mol. Biol. Evol. 4:406–425. fungi. Proc. R. Soc. London B 266:163–171. SANDERSON, M. J. 1998. Estimating rate and time in WINNEPENNINCKX , B., AND T. BACKELJAU. 1996. 18S molecular phylogenies: beyond the molecular clock? rRNA alignments derived from different secondary Downloaded from https://academic.oup.com/sysbio/article/49/1/130/1678667 by guest on 01 October 2021 Pages 242–264 in Molecular systematics of plants II. structure models can produce alternative phyloge- DNA sequencing (D. E. Soltis, P. S. Soltis, and J. J. nies. J. Zool. Syst. Evol. Res. 34:135–143. Doyle, eds.). Kluwer, Boston. WINNEPENNINCKX , B., T. BACKELJAU, L. Y. MACKEY, J. M. SCHRAM, F. R. 1991. Cladistic analysis of metazoan BROOKS, R. DE WACHTER, S. KUMAR, AND J. R. GAREY. phyla and the placement of fossil problematica. 1995. 18S rRNA data indicate that Aschelminthes Pages 35–46 in The early evolution of Metazoa and are polyphyletic in origin and consist of at least the significance of problematic taxa (A. M. Simon- three distinct clades. Mol. Biol. Evol. 12:1132–1137. etta and S. Conway-Morris, eds.). Cambridge Univ. WRAY, G. A., J. S. LEVINTON, AND L. H. SHAPIRO. 1996. Press, Cambridge, UK. Molecular evidence for deep precambrian diver- SEILACHER, A., P. K. BOSE, AND F. PFLÜGER. 1998. gences among metazoan phyla. Science 274:568–573. Triploblastic animals more than 1 billion years ago: ZHANG, J., AND M. NEI. 1996. Evolution of Antenna- evidence from India. Science 282:80–83. pedia-class homeobox genes. Genetics 142:295–303. SIDOW, A., AND W. K. THOMAS. 1994. A molecular evo- ZRZAVY´, J., S. MIHULKA, P. KEPKA, A. BEZDEKÆ , AND D. TIETZ. lutionary framework for eukaryotic model organ- 1998. Phylogeny of the Metazoa based on morpho- isms. Curr. Biol. 4:596–603. logical and 18S ribosomal DNA evidence. Cladistics SIEWING, R. 1980. Das Archicoelomatenkonzept. Zool. 14:249–285. Jahrb. Anat. 103:439–482. STRIMMER, K., AND A. VON HAESELER. 1996. Quartet puzzling: A quartet maximum likelihood method for reconstructing tree topologies. Mol. Biol. Evol. 13:964–969. TÖNGES, U., S. W. PERREY, J. STOYE, AND A. W. M. DRESS. 1996. A general method for fast multiple sequence Received 8 September 1998; accepted 23 February 1999 alignment. Gene 172:GC33–GC41. Associate Editor: S. Edwards