Animals.

Cephalodiscus gracilis and Ptychodera bahamensis were identified and collected from Bermuda by C.B.C. and stored in 70% ethanol. Saccoglossus species (C.B.C.) and Harrimania species (C.B.C.) were collected subtidally from Barkley Sound, British Columbia and have not yet been described (33). B.J.S.'s lab extracted DNA and entirely sequenced complete 18S rDNA from the hemichordates and the ascidians Ascidia ceratodes, Herdmania curvata, Phallusia mammillata, and Molgula oculata (ref. 16; Table ​Table22).

Relationships Within Deuterostomia.

The complete alignment of 28 deuterostome species and 5 outgroup taxa contained 2,161 sites including those with gaps or 1,472 sites after all sites with a gap at any taxon were deleted. Initial results with the entire data set of 28 deuterostome taxa by using MP, ME, and NJ analyses were mixed (Table ​(Table1).1). In nearly all analyses, each of the four deuterostome taxa (echinoderms, hemichordates, urochordates, vertebrates + cephalochordates) were well-supported monophyletic groups (Table ​(Table1).1). In some cases, hemichordates and echinoderms grouped together, but chordates and urochordates rarely grouped together (Table ​(Table1).1). Testing of alternate MP and ME topologies by using macclade revealed that there was virtually no difference in the length of trees with different topological arrangements of the four main taxa (data not shown).

Table 1

Bootstrap support for deuterostome phylogenetic trees

Tree algorithm	Uro	Chor	Hemi	Echin	Hemi + Echin	Chor + Uro	Basal Deut?
Full dataset (29 deuterostomes)
NJ, Kimura	100	60	94	100	67	—	Chor
NJ, Kimura with γ	100	65	89	99	—	—	Echin
MP	96	83	93	89	80	59	2 + 2
ME, Kimura	100	68	92	99	56	—	Chor
ME, Kimura with γ	99	66	93	97	—	—	Chor
ME, Paralinear	99	56	95	97	—	—	Chor
≤0.1400 subs/site (19 deuterostomes)
NJ, Kimura	100	70	100	100	—	—	Chor
NJ, Kimura with γ	100	72	100	100	—	—	Echin
MP	100	79	93	88	66	54	2 + 2
ME, Kimura	100	71	99	98	—	—	poly
ME, Kimura with γ	100	68	100	98	—	—	poly
ME, Paralinear	100	58	99	100	—	—	poly
≤0.1200 subs/site (16 deuterostomes)
NJ, Kimura	100	71	100	100	80	—	Chor
NJ, Kimura with γ	100	67	100	99	64	—	Chor
MP	100	67	100	96	86	65	2 + 2
ME, Kimura	100	63	100	100	83	—	Chor
ME, Kimura with γ	100	74	100	100	70	—	poly
ME, Paralinear	100	63	100	100	89	—	Chor
ML	100	53	100	100	100	—	Chor
gambit and paralinear outgroup
Nephtys (0.1518)	100	71	98	90	86	83	2 + 2
Glycera (0.1432)	100	74	98	94	78	85	2 + 2
Artemia (0.2478)	100	68	86	99	86	—	poly
Tenebrio (0.2327)	100	—	100	87	89	—	poly
Anemonia (0.3373)	99	59	93	98	73	—	Chor

Open in a separate window

Summary of analyses of the full dataset (28 deuterostomes), a subset including those taxa with a branch length ≤0.1400 substitution per site, and a deuterostome subset with branch lengths of ≤0.1200 substitution per site. Bootstrap values are shown for the monophyly of urochordates (Uro), chordates (including cephalochordates, Chor), hemichordates (Hemi), and echinoderms (Echin). Bootstrap values for a monophyletic hemichordates + echinoderms and a monophyletic chordates + urochordates are shown when greater than 50%. The basal deuterostome in each analysis is shown unless there was a polytomy (poly). Instances where hemichordates + echinoderms was a sister group to chordates + urochordates are shown labeled “2 + 2.” Methods include NJ and ME trees with Kimura two-parameter, Kimura two-parameter with γ correction, or paralinear distances, and MP. ML and gambit analysis were carried out only for the smallest dataset (≤0.1200 subs per site). gambit was used with paralinear distances and a correction for site-to-site variation by using a variety of outgroups. The Kimura two-parameter γ corrected distances between each outgroup and Antedon are shown in parentheses. The most reliable analyses are shown in bold. 

Four-cluster analysis revealed the branch length from an internal node to each deuterostome taxon (Table ​(Table2).2). This shows that the number of substitutions per site varies considerably among the different deuterostome taxa used in this study. Results reported in Table ​Table22 were used to produce two subsets of the alignment including only those sequences with branches ≤0.1400 substitution per site (19 taxa), or ≤0.1200 substitution per site (16 taxa). This approach of excluding taxa with the most rapidly evolving 18S rDNA sequences has been used previously to minimize unequal rate effects (43, 44). The bootstrap support for key nodes of phylogenetic trees generated with the different deuterostome subsets is shown in Table ​Table1.1. As in the complete alignment, each of the four main deuterostome taxa (echinoderms, hemichordates, urochordates, vertebrates + cephalochordates) were generally well supported monophyletic groups, although the topology among the four groups varied (Table ​(Table1).1). With the subset of taxa with ≤0.12 substitution per site, there is moderate to strong support for a sister-group relationship between hemichordates and echinoderms (Table ​(Table1)1) with all methods (NJ, MP, ME, and ML). There was very little support for a sister-group relationship between chordates and urochordates.

The increasing support for a hemichordate + echinoderm clade from the alignment subset with the slowest evolving taxa (≤0.1200 substitution per site) suggested that unequal evolutionary rate was a significant source of error in the analyses. Further reduction of the data set (e.g., ≤0.1000 substitution per site) was not useful because it completely eliminated all of the vertebrate sequences from analyses. Therefore, bootstrapper's gambit (42), a method known to be insensitive to unequal evolutionary rates and site-to-site variation in evolutionary rate, was used for the final analysis on the ≤0.1200 substitution per site subset (16 deuterostome taxa). gambit revealed that hemichordates + echinoderms formed a sister group to chordates + urochordates with high bootstrap support (Table ​(Table11 and Fig. ​Fig.2).2). This analysis may be sensitive to the length of the branch leading to the nondeuterostome outgroup, therefore, the branch length from the slow evolving deuterostome Antedon serrata to each outgroup was calculated and is shown at the bottom of Table ​Table1.1. The two annelid sequences (Nephtys hombergii and Glycera americana) were the slowest evolving outgroups and, as expected, yielded the most highly supported topology. When longer-branched outgroups such as the arthropods Artemia salina and Tenebrio molitor or the diploblast Anemonia sulcata were used, the resolution of the tree degraded, and support for chordates + urochordates was lost (Table ​(Table1).1).

Open in a separate window

Figure 2

Phylogenetic tree of the deuterostomes when sequences with similar evolutionary rates (16 taxa ≤0.12 substitutions per site) were analyzed with gambit. Key characters are mapped to the deeper nodes. The deuterostomes form two great clades, one containing the hemichordates and echinoderms and the other made up of urochordates and chordates (cephalochordates and vertebrates). Major differences in adult body plan between Cephalochordata + Vertebrata (myotomes) and Urochordata (tunic) are marked. These results, combined with morphological data, suggest that Chordata should be restricted to Cephalochordata + Vertebrata and that Urochordata is an independent phylum and the sister group to Chordata. Note that the tripartate coelom of hemichordates is considered homologous to the three pairs of echinoderm coeloms.

Our analysis of the slow evolving rDNA sequences of the 16 deuterostome taxa by using gambit as shown in Fig. ​Fig.22 is the first molecular study to define clearly the evolutionary relationships among all four major groups of deuterostomes. We found Vertebrata to be monophyletic and a sister clade to Cephalochordata, as suggested by previous studies (45). Urochordata is monophyletic as shown previously (16) and forms a sister group to Vertebrata + Cephalochordata. The sister relationship between urochordates and vertebrates + cephalochordates is supported by morphological evidence. The presence of a notochord unites urochordates, cephalochordates, and vertebrates (9–13, 26, 45, 46), although some authors differentiate the notochord of vertebrates + cephalochordates from the urochord of urochordates (15). The characters that set vertebrates + cephalochordates apart from urochordates are the presence of myotomes in vertebrates + cephalochordates and the presence of the tunic in the urochordates (Fig. ​(Fig.2;2; 9–12, 26, 45–48). Urochordates are considered members of Chordata because the tadpole larva exhibits the chordate body plan. However, differences in adult body plan and life histories of urochordates compared with vertebrates + cephalochordates justify a reappraisal of the inclusion of urochordates within the phylum Chordata despite the presence of a notochord.

The chordate features of ascidian tadpole larvae develop during embryogenesis from morphological and genetic pathways similar to chordates (49, 50). Recent phylogenetic evidence from 18S rDNA data shows that the appendicularians, or larvaceans, are likely to be a sister group to the ascidians and thaliacians (16, 51). In light of these results, the ancestral urochordate may have been either pelagic or sessile. Urochordate adults are morphologically distinct from vertebrates and cephalochordates (46–48). Both sessile ascidians (46, 47) and pelagic tunicates (48) contain a characteristic extracellular coat, or tunic, that protects the adult. There is also a fundamental difference in life history traits between urochordates and other chordates. Both vertebrates and cephalochordates are solitary and sexual organisms, whereas urochordates have evolved a colonial lifestyle several times independently (16, 32, 52). All colonial urochordates can also reproduce asexually by budding or may reproduce sexually (46, 47). Although this is common in urochordates, it is virtually absent from other chordates. This morphological and ecological evidence, coupled with our molecular results (16), suggests that Urochordata should be considered an independent phylum that is the sister group to Vertebrata + Cephalochordata. The phylum name Chordata should be restricted to Cephalochordata + Vertebrata, and Urochordata should be raised to phylum status.

This study found strong support for hemichordate monophyly and for an echinoderm + hemichordate clade. The echinoderm + hemichordate clade, obtained by molecular data here and elsewhere (2–4, 53), is also strongly supported by larval morphological evidence. For many years, the enteropneust tornaria was considered the larva of an echinoderm, in particular an auricularia of a holothoroid or bipinnaria of an asteroid (11, 17). These large gelatinous larvae share a preoral feeding band that creates an upstream feeding current by using monociliated cells (21, 54) and a perioral ciliated band that manipulates food into the esophagus. The three coelomic sacs in hemichordates and echinoderms are both organized anterior to posterior as protocoel (echinoderm axocoel), mesocoels (echinoderm hydrocoels), and metacoels (echinoderm somatocoels) (9, 21). For an extensive comparison of an echinoderm auricularia to hemichordate tornaria see (9, 21, 22).