Correspondence Thomas Stach, Fakultät für Biologie, Zoologie – Systematik und Evolution der Tiere, Freie Universität Berlin, Königin-Luise-Straße 1-3, 14195 Berlin, Germany. Email: email@example.com
Traditional concepts of chordate phylogeny have recently been in turmoil: in a large-scale molecular study, the traditional hypothesis that cephalochordates are sister taxon to craniates was replaced by the hypothesis of a sister group relationship between tunicates and craniates. It was claimed that the morphological evidence that supported traditional phylogeny was weak and that morphological characters at least equally strong could be mustered in support of the ‘new phylogeny.’ In the present review, it is shown that the uncritical use of published codings of morphological characters in recent phylogenetic analyses is responsible for this perception. To ameliorate this situation, the main focus of the present publication is a review of the morphological evidence that has been deemed relevant in chordate phylogeny. Characters are presented in enough detail to allow readers to make self-reliant informed decisions on character coding. I then analyze these characters cladistically, and it is demonstrated that support of the traditional hypothesis is substantial. I briefly evaluate molecular systematic studies and criticize ‘evo-devo’ studies for lack of cladistic rigor in the evolutionary interpretations of their data by (1) failing to formally code their characters (2) failing to subject their data to the congruence test with other characters, the crucial test in phylogenetic analyses. Finally, a short and by necessity eclectic discussion of suggested evolutionary scenarios is presented.
Chordata, a taxon consisting of about 50 000 animals, includes Homo sapiens, our own species, and it might be for this latter reason that the interest in understanding the evolution of chordates never ceased. It is probably the former number that makes this understanding so difficult, because with this number, although low compared with some protostomian groups, comes an enormous range of life styles and corresponding body plans. These body plans fall into three easily recognized groups: Tunicata, Cephalochordata and Craniata (Different names exist for identical taxa. For example, Acrania is used instead of Cephalochordata, Urochordata instead of Tunicata or Vertebrata instead of Craniata. In German publications, Craniota is used instead of Craniata. Dohle (2004) presents a tabulation of different names in use for higher deuterostome taxa.). The main bulk in sheer number belongs to the Craniata and therein to the Osteognathostomata, with the primarily aquatic Osteichthyes (bony fishes) being first in number and arguably also in diversity. The Cephalochordata comprise a small confined group of about 27 species that are very similar in appearance, anatomy and life style. The Tunicata, on the other hand, while only mustering about 2500 strictly marine species, show a formidable range of habits, body plans and life histories. The most numerous group of tunicates is ascidians that are sessile filter feeders with tadpole larvae that can propagate sexually and asexually. A significant group, the Thaliacea, has mastered and explored the holoplanktonic realm by evolving a dizzying variety of adaptations including metagenesis, heteromorphy and bioluminescence. Another holoplanktonic life form belonging to the Tunicata is a small group of about 60 species named Appendicularia (=Larvacea), for they retain the tadpole tail for their entire life.
The three chordate taxa, Craniata, Cephalochordata and Tunicata, are clearly monophyletic. For a proper application of phylogenetic methodology, it is important to demonstrate that the taxa under consideration are indeed monophyletic. Therefore, the monophyly of the taxa under consideration will be discussed in the following paragraphs, before the interrelationship of these taxa is contemplated.
Monophyly of Chordata
Recently, it had been suggested that Chordata is not monophyletic by pointing out that there are similarities between tunicates, cephalochordates and protostomians regarding gene expression pattern (Raineri, 2006). While such a controversial interpretation might stimulate further research, there is no reason to assume that the similarities in gene expression pattern of genes involved in axis specification are synapomorphies of protochordates and protostomians. Rather, consistent with current knowledge, it can be concluded that these similarities are plesiomorphic characters within the Bilateria. Another currently controversial debate focuses on the taxonomic rank of Chordata and its three subordinate taxa. Cameron, Garey & Swalla (2000) proposed to elevate the Tunicata to phylum rank because ‘they have a unique adult body plan, are the only metazoan subphylum classified by their larvae, and are a monophyletic group that share specific synapomorphies, including the tunic and an open circulatory system’ (Cameron et al., 2000). This suggestion, however, is not helpful, as taxonomic ranks are unimportant human constructs modern taxonomy tries to abolish (e.g. de Queiroz, 2006; Rieppel, 2006). The Chordata itself is supported as a monophyletic taxon by numerous autapomorphies.
The notochord, or Chorda dorsalis, is a skeletal rod that serves as a central stiffening element that ensures length constancy in undulatory locomotion. The notochord stretches from the posterior tip of the tail to the trunk in tunicates, from the posterior tip of the tail to the posterior brain capsule in craniates, and from the posterior tip of the tail to the anterior tip of the animal in cephalochordates. In all taxa, it is derived from the dorsal epithelium of the archenteron during embryogenesis (e.g. Hausen & Riebesell, 1991; Stach, 2000; Munro & Odell, 2002). That the notochord is homologous in all chordates is also corroborated by the fact that the brachyury gene plays a similar role in determination of the notochord in all chordates. These similarities support the hypothesis that the notochords are homologous in all three chordate taxa, although the cells that make up the notochord differ drastically in the three clades (Welsch, 1968b; Welsch & Storch, 1969; Schmitz, 1998; Stach, 1999).
Dorsal neural cord with a neural canal, Reissner's fiber and neurenteric canal
Dorsal to the notochord lies a strand of nervous tissue that comprises the central nervous system in chordates. In all taxa, it has an anterior swelling that receives input from anterior sensory structures and stretches over the length of the entire body during one point in ontogeny (Meinertzhagen & Okamura, 2001; Lemaire, Bertrand & Hudson, 2002). In addition, all neural cords in chordates are similar in that they possess a central fluid-filled canal with cilia and a gel-like strand of unknown function, the so-called Reissner's fiber (Olsson, 1993). Moreover, the canal is continuous with the endodermal archenteron at its posterior end. This connection is called the neurenteric canal or the canalis neurentericus, and it obliterates during ontogeny in all chordates (Salvini-Plawen, 1998).
Epithelium of the pericard forms the musculature of the heart (Fig. 1)
The craniate heart is a complex organ situated ventrally immediately behind the branchial basket in primitive forms (e.g. Starck, 1982; Mickoleit, 2004). While it is known that the vascular system of craniates, including the heart, is lined by an endothelium, this is not the case in tunicates and cephalochordates (Rähr, 1981; Corley, 1995; Burighel & Cloney, 1997; Ruppert, 1997b). Nevertheless, a ventral major blood vessel exists in the branchial basket in all chordate taxa (Rähr, 1981; Starck, 1982; Corley, 1995). In addition, in all three taxa, a center for the propulsion of the blood fluid is present just behind the branchial basket. Moreover, the cells equipped with contractile filaments providing the contractive force for this propulsion are similarly derived from a coelomic epithelium, the pericard in tunicates and craniates (Hirakow, 1989) and the extensive rostral coelom in cephalochordates (Stach, 1998), which therefore could be homologous in its medio-ventral part to the pericard.
Mesodermal tail musculature derived from dorsolateral pockets of archenteron
In craniates, the ventral nervous tissue of the thalamus region in the diencephalon is in close contact with the adenohypohysis, which is ontogenetically derived from the stomodaeal epithelium (Hartenstein, 2006). In cephalochordates, a similar close positional relationship exists between a ventral extension of the brain vesicle and a dorsal extension of the epithelium of the buccal cavity that is named Hatschek's pit (Gorbman, 1994, 1999; Gorbman, Nozaki & Kubokawa, 1999). The ontogeny of Hatschek's pit is complicated (Hatschek, 1881; Stach, 1996, 2002), but in addition to the aforementioned topological similarity different adenohypophyseal hormones have been demonstrated to be present in Hatschek's pit in cephalochordates (Olsson, 1969; Gorbman et al., 1983; Gorbman, 1995). In tunicates, a similar topological relationship exists between the neural gland and the ciliated funnel that opens into the dorsal anterior branchial basket (Ruppert, 1990). Again, several hormones found in craniate adenohypophysis have been demonstrated using immunocytological techniques (Romanov, 2000). While gene expression studies are less conclusive, with some key pituitary genes not being expressed in the neural gland or the ciliated funnel of tunicates, homology of the pituitary in chordates is supported by these studies as well (Sherwood, Adams & Tello, 2005; Candiani et al., 2008). Thus, the presence of a hypophysis system including a neural and an adenohypohyseal part with particular hormones has to be interpreted as an autapomorphy of Chordata.
The dorsal wall of the diencephalon of craniates forms a photoreceptor, the pineal eye that contains cells with stacks of lamellae modified from cilia that are identical in ultrastructure to lamellar cells in the dorsal wall of the cerebral vesicle of cephalochordates (Pu & Dowling, 1981; Lacalli, 1994, 2004). Several authors have argued that the ocellus of tunicates could also be homologous to the pineal eye (Eakin, 1973; Whittaker, 1997), which seems to be corroborated by sequence similarities in the opsin genes (Kusakabe et al., 2001). However, these similarities also exist to the opsins in the lateral eyes of vertebrates and to sequences isolated from cephalochordates (Terakita, 2005). Thus, if the arguments against the homology of the tunicate ocellus to the paired vertebrate eyes are considered, the hypothesis of homology to the pineal eye in vertebrates and the lamellar body of cephalochordates becomes a reasonable, yet controversial alternative. Thus, given these premonitions, the pineal eye can be assumed to be another synapomorphy of Chordata.
In addition to these unique morphological autapomorphies, Chordata share some gene regulatory pathways that are equally unique to Chordata. Summaries of the similarities in the expression pattern of homologous genes that seem to be unique to Chordata abound and can, for example, be found in Rowe (2004), Lemaire (2006), Schlosser (2007).
Monophyly of higher chordate taxa
Monophyly of Craniata
The Craniata, to begin with, show so many autapomorphies that only a few of the more complex ones will be mentioned in this paragraph.
Blood vessels in the invertebrate taxa are, as a rule, spaces within the extracellular matrix (Ruppert & Carle, 1983; Westheide & Rieger, 2007), with the exception of nemertines. While they can be either vast and more or less undefined as in the molluscs or anatomically highly stable and precisely organized as in many annelids and cephalochordates, they always lack a cellular endothelial lining (Ruppert & Carle, 1983; Goldschmid, 2007).
The craniate-type heart is, in the plesiomorphic condition, situated on the ventral side immediately behind the gill-bearing pharynx. Thus, in its primitive craniate condition the heart has inherited its position from the last common ancestor of the Chordata. It carries only unoxygenated blood. The ground plan of the plesiomorphic condition of the craniate heart can be stated more precisely. It is muscular and consists of distinct parts: an atrium, a ventricle and the conus arteriosus (Starck, 1982; Mickoleit, 2004).
Anterior sensory organs
This short phrase summarizes countless detailed autapomorphies of the Craniata. It comprises paired lateral eyes, organs with a complicated ontogeny that involves an evagination of the diencephalon and the induction of an epidermal lens (Starck, 1982; Hall, 1995; Mickoleit, 2004). It also comprises the labyrinth system that allows the fast-moving craniates to orientate in three-dimensional (3D) space (Starck, 1982; Streit, 2001; Mickoleit, 2004). Anterior sense organs present in the last common ancestor of the recent craniates also include the olfactory organs associated with the telencephalon, and the diencephalic dorsal parietal and pineal eyes (Starck, 1982; Mickoleit, 2004; Janvier, 2008).
Probably the most impressive autapomorphic trait of the craniates is the brain encased in a skeletal neurocranium. In all craniates, it consists of distinct parts that can be recognized as corresponding, and whose ontogeny is highly similar as well. Just to allude to the substance behind this claim, I mention the Telencephalon, Diencephalon, Mesencephalon and Rhombencephalon as easily recognized homologous parts of the craniate brain (Starck, 1975a, 1982; Romer & Parsons, 1986; Mickoleit, 2004; Murakami et al., 2005).
In craniates, nerves leave the spinal cord in each myomere on both sides to innervate musculature and connect to sensory cells (Starck, 1982; Romer & Parsons, 1986; Kelly Kuan et al., 2004; Mickoleit, 2004). They show a regular pattern in as much as in the primitive craniate condition, the dorsal nerves are connected to the peripheral nerves in a ganglion that is situated close to the spinal cord in the so-called spinal ganglion. While the dorsal counterparts of nerve roots are also found in cephalochordates, the cell bodies reside within the nerve cord and do not form a spinal ganglion (Bone, 1960a,b; Wicht & Lacalli, 2005).
There are countless more morphological (and molecular) characters that attest to the monophyly of Craniata, in fact too many to list here. The reader is kindly referred to the excellent accounts that are available on this subject, for example, Starck (1975a), Romer & Parsons (1986) and Mickoleit (2004).
Monophyly of Cephalochordata
Compared with craniates, the 27 species of cephalochordates are rather uniform and relatively simple organisms, and yet there are also too many autapomorphic traits that establish the monophyly of this clade to list them all in the present review.
This unique cell type that obviously serves an excretory function is unique in the animal kingdom. The coelomic cyrtopodocyte consists of a podocytic part, in which cytoplasmic extensions cover a blood vessel. These extensions interdigitate with similar extensions of the neighboring cells and a slit membrane is formed between the cells and above the blood vessel. The second part of the cyrtopodocyte resembles a protonephridic solenocyte; a central apical cilium is surrounded by 10 long, slender and stiff microvilli. These microvilli project into the excretory canal that leads to the outside, and they are covered by a sheath of extracellular material that forms a second filtration barrier (Brandenburg & Kümmel, 1961; Ruppert, 1996; Stach & Eisler, 1998).
Unique among the chordates, the main part of the notochord consists of cells that contain contractile fibers. These paramyosin fibers are arranged in a striated fashion and can contract the cells laterally, thus stiffening the notochord locally. The notochord cells send cytoplasmic extensions through pores in the strong notochordal sheath to the nerve cord, obviously contacting the nerve cord for innervation. In addition to this unique notochord cell type, the notochord contains at least one more cell type dorsally and ventrally, the Müller cells. Such a differentiation of notochord cells is also unique within the chordates (Welsch, 1968b; Flood, 1975; Ruppert, 1997a; Stach, 1999).
This phrase alludes to many unique features in the ontogeny of cephalochordates. Here, I want to mention three of them as they are specific, easy to distinguish characters: the left-sided oral papilla, the right-sided, obliquely v-shaped rudiment of the endostyle and the club-shaped gland situated immediately behind the larval endostyle in the pharynx (Conklin, 1932; Olsson, 1983; Stach, 2000).
Again, as for the craniates, there are too many morphological characters that attest to the monophyly of Cephalochordata to list here. The reader is kindly referred to the excellent accounts that are available on this subject, for example, Franz (1927), Ruppert (1997a) and Stach (2000).
Monophyly of the Tunicata
If one considers the diversity of life styles and anatomies in Tunicata, it is hardly surprising that the monophyly of this taxon is difficult to ascertain. Yet, even in the case of the Tunicata, there are unique, uncontroverted (Wenzel, 2002) autapomorphies that can be listed to back the hypothesis that this taxon is monophyletic.
The monolayered epidermis of tunicates is covered by a sheath of extracellular material, the so-called tunic. It can contain cells that seem to contribute to the integrity of the tunic in several ways. In addition, the tunic contains proteins, but more significantly cellulose fibers. This is the only case in the animal kingdom that cellulose is produced, and similarities in cellulose-synthetase genes suggest that this ability was inherited by lateral gene transfer from bacteria. It has been demonstrated in over 180 species from all major tunicate groups (Hirose et al., 1997; Kimura et al., 2001; Nakashima et al., 2004; Davison & Blaxter, 2005; Stach, 2007).
This is a convenient and at the same time informative apomorphic character for the Tunicata, because it can be seen in most species with the help of a dissecting microscope alone. All species examined so far show this phenomenon, where the blood is pumped anteriorly toward the pharynx for several minutes and, after a short pause, pumped posteriorly toward the intestinal organs (Skramlik, 1938; Kriebel, 1967).
If one considers the potential phylogenetic neighborhood of the Tunicata, that is, the deuterostome taxa, it becomes clear that hermaphroditism is the derived state. However, hermaphroditism as a character state is prone to homoplasy, as it highly depends on the ecology of the animals. For example, sessile organisms tend to be hermaphroditic, in order to increase their chances of successful reproduction, and it is quite probable that the last common ancestor of Tunicata was a sessile organism (Huus & Knudsen, 1950; Stach & Turbeville, 2002). While most planktonic appendicularians are hermaphroditic, the well-known species Oikopleura dioica is dioecious.
A morphologically distinct visceral ganglion is present in all tunicate tadpole larvae that have been examined so far, including the probably progenetic appendicularians (Burighel & Cloney, 1997; Søviknes, Chourrout & Glover, 2005). While at first glance, the unique presence of a distinct ganglion immediately anterior to the tail complex seems to be a clear autapomorphy of the Tunicata, this does not preclude that this part of the nervous system has a corresponding homologous part in other chordates, which is not morphologically distinct (Dufour et al., 2006).
With the monophyly of the three taxa established by arguments, I will proceed to the main task of this paper, that is, to review the possible interrelationships of these three taxa as can be inferred from the morphological evidence.
The competing hypotheses
With only three monophyletic groups, there are in theory only three possible ways to draw a rooted phylogenetic tree for the Chordata. These possibilities can be seen in Fig. 3. In the remainder of this paper, I will call the hypothesis that Tunicata is sister taxon to Cephalochordata ‘Atriozoa hypothesis,’ the hypothesis that Tunicata is sister taxon to Craniata is called ‘Olfactores hypothesis,’ while the hypothesis that Cephalochordata is sister taxon to Craniata is called ‘Notochordata hypothesis.’ For long, the ‘Notochordata hypothesis’ seemed to be widely accepted, with the ‘Atriozoa hypothesis’ being of merely historical interest and the ‘Olfactores hypothesis’ being only invoked because of the controversial interpretation of unusual fossils. In supporting the ‘Notochordata hypothesis,’ Ax even nicknamed the cephalochordates ‘honorary vertebrates’ (Ax, 2001). This picture has been upset recently, because of the analysis of large sequence datasets supporting the ‘Olfactores hypothesis’ (Delsuc et al., 2006; Dunn et al., 2008; Putnam et al., 2008; Swalla & Smith, 2008). Delsuc and coleagues suggested that the morphological evidence that led to the ‘Notochordata hypothesis’ was meager and at least rivaled by morphological data supporting the ‘Olfactores hypothesis.’ In order to give a more complete picture, I will review the morphological evidence available in some detail by explaining the morphological reasoning behind the three competing hypotheses. I will then proceed to compile and analyze a matrix, amenable for cladistic analysis including the characters advanced from proponents of all three hypotheses. Then, I will take a cursory look at the molecular evidence and the evolutionary interpretations of this combined evidence.
The Atriozoa hypothesis
The Atriozoa hypothesis suggests that Cephalochordata and Tunicata are sister taxa to the exclusion of Craniata. This hypothesis derives its name from the fact that it was originally based on the shared common possession of a water jacket around the branchial basket, that is, the atrium. The hypothesis was suggested by Minot (1897) and supported by Masterman (1898), but is also reflected in the name ‘Protochordata’ that, as a taxon, would also comprise Tunicata and Cephalochordata. However, the term Protochordata is more often used as a colloquial name and etymologically also hints at the possibility that it might describe a paraphyletic grouping. While this hypothesis hardly finds any support in recent scientific publications, proponents of this hypothesis suggested some potential synapomorphies that will be discussed in the following paragraphs.
In stark contrast to the craniates, the enteropneusts and the pterobranchs, tunicates and cephalochordates possess a water jacket around their branchial basket that is called an atrium or a peribranchial chamber (Dohle, 2004). This space is delimited by the branchial basket on the interior face and by a double layer of monolayered epithelium with mesodermal muscle cells in between on the exterior face. In both taxa, there is a single atrial opening that lies on the ventral side in cephalochordates, but dorsal in most tunicates. However, the atrial opening is posteriorly placed in Thaliacea, and an atrium is entirely lacking in Appendicularia. The atrium originates as a paired dorsal rudiment in Aplousobranchiata and part of the Phlebobranchiata, but as an unpaired dorsal rudiment in Stolidobranchiata and as an unpaired ventral room delimited by epidermal folds in cephalochordates. Therefore, the homology of the atria of Tunicata and Cephalochordata has been doubted (Pietschmann, 1962; Dohle, 2004).
Discoidal notochord cells arranged in single file
Ruppert (1997b) suggested that the arrangement of cells that comprise the notochord in a single row (‘stack of coins’) is common only to Tunicata and Cephalochordata to the exclusion of Craniata. In most tunicates, this ‘stack of coin’ stage is present only as a temporary stage during larval development and is lost in the swimming larva and is absent in adult tunicates. It remains present in adult cephalochordates, where the notochord cells are differentiated as muscle cells. In craniates, as Stach (1999) pointed out, the arrangement of the notochord cells in single file is also present during embryogenesis, but is lost later on. Thus, in the general form as suggested by Ruppert (1997b), this character is not supporting the closer relationship of Tunicata and Cephalochordata but instead is another autapomorphy of Chordata.
Intercellular spaces between notochordal cells
Another character suggested by Ruppert (1997b) as supportive of the ‘Atriozoa hyothesis’ is the presence of intercellular spaces between notochordal cells. In tunicates, such intercellular spaces can be limited to small pockets between subsequent cells of the ‘stack of coin,’ but in many species, these spaces enlarge, become confluent and form a fluid-filled canal. In this way, the cytoplasms of notochordal cells become restricted to the walls of the aforementioned canal and in character become epithelial, fulfilling the definition of a coelom as given by Bartolomaeus (1993). In cephalochordates, the intercellular spaces originate in a similar way as small pockets between subsequent cells. These pockets expand, but do not form a continuous central canal (Stach, 1999, 2000). However, they are confluent by way of two narrow canals dorsally and ventrally of the notochord cells. The notochord cells of craniates are large, numerous and ‘turgescent.’ In the electron microscopic aspect, the turgor-providing vacuoles contain flocculent material and are membrane bound, with intermediate filaments strengthening the membrane (Schmitz, 1998).
This character is the main synapomorphy on which the ‘Olfactores hypothesis’ was originally founded. It is based on the detailed interpretation of the enigmatic fossils named carpoids in the hypothesized stem lineages of tunicates and craniates (Jefferies, 1986). The character is not present in recent chordates; therefore, independent reduction is required for tunicates and craniates. While I have not examined these fossils myself, I want to point out that their interpretation is controversial (Jollie, 1982; Peterson, 1995; Jefferies, 1997; Ruta, 1999). Accepting the interpretation of Jefferies (Jefferies, 1986, 1997; Jefferies, Brown & Daley, 1996; Dominguez Alonso & Jefferies, 2001) also makes it necessary to suppose the independent loss of calcareous mesodermal skeletal structures in tunicates, cephalochordates and craniates alike. In addition, it also follows that the peculiar stereome structure of calcitic hard parts of the skeleton, which is found only in Echinodermata in extant fauna, evolved in the last common ancestor of all Deuterostomia.
Hatschek's nephridium is a single large nephridium that is situated in the dorsal roof of the preoral cavity in amphioxus. Cephalochordates, in addition, possess segmental nephridial structures situated along the dorsolateral sides of the branchial basket. This situation is similar to the one found in enteropneusts, where a single large excretory organ is present anterior to the mouth and repeated coelomic units with parts of their epithelium differentiated as podocytes are present along the dorsolateral sides of the brancial basket (Stach, 2002). In the primitive craniate condition, segmental nephridial structures are present along the most part of the body, but no single excretory organ can be found anterior to the mouth (Starck, 1975b, 1982; Kluge & Fischer, 1990, 1991). In tunicates, mesodermal excretory structures involving ultrafiltration through basement membranes held by podocytic cell extensions are not present (Burighel & Cloney, 1997). Thus, reduction in the entire typical deuterostome excretory system has to be assumed in Tunicata and the loss of the anterior preoral kidney can be inferred for Craniata.
Septate junctions are present as an apical barrier within the apical junctional complexes of epidermis cells in almost all animals. However, they are occasionally found in craniates (Gobel, 1971; Friend & Gilula, 1972; Sotelo & Llinas, 1972; Altorfer & Hedinger, 1975) but have never been found in tunicates. On the other hand, there are several cases of cell junctions that are ultrastructurally indiscernible from tight junctions in other invertebrates such as amphipods, bivalves or crayfish (Lane & Chandler, 1980). Thus, while there are several homologous genes necessary for the correct expression of septate junctions in Drosophila melanogaster and tight junctions in craniates (Willot et al., 1993; Stevenson & Keon, 1998), this is no proof for homology of the structure itself, especially because both types do occur together in the same species.
Swimming musculature, partly a functional syncytium (Ruppert, 2005)
This is a peculiar character. Historically, it had been discussed whether the myofibrils in the tail of ascidian larvae were continuous (Berrill & Sheldon, 1964) between adjoining cells. However, detailed electron microscopic investigations have shown that the fibrils (=cells) were indeed separate, but that the cells are connected by special cell junctions (Cavey & Cloney, 1976; Cavey, 1982; Cavey & Strecker, 1985; Burighel & Cloney, 1997). In vertebrates, the lateral muscle cells of the myotome fuse to form a true syncytium (Starck, 1975b; Sachidamamdam & Dhawan, 2003; Kielbowna & Daczewska, 2005). Thus, while it is correct to state that the myofibrils are functionally connected in tunicates and craniates, a syncytium is only found in craniates. The character state with separate, mononucleate muscle cells present in ascidian larvae is the plesiomorphic condition in Chordata and is present in cephalochordates as well as other deuterostomes. Whether the connection by special cell junctions in ascidian larvae is homologous to the fact that myofibrils are formed inside of syncytia in craniates is doubtful. At least, this character therefore should be coded as two separate ones: (1) muscle cells forming a syncytium and (2) myofibrils of originally separate cells functionally coupled.
The homology of coronet cells found in the brain of O. dioica with cells from the Saccus vasculosus in craniates has been originally suggested by Olsson (1975). As similarity between the two cell types, this author mentions that the vertebrate cells possess numerous apical globuli, whereas the tunicate cell possesses a single globus. Both types of cells are rich in endoplasmic reticulum, Golgi complexes and vesicles and seem to function in secretion. Despite these similarities, there are differences as well. The globus in O. dioica cells possesses a rudimentary cilium. No such structure exists in the globuli of the craniate cells. In addition, the coronet cells in craniates are embedded in a tall epithelium and connected by an apical junction with the neighboring cells. The neighboring cells in O. dioica coronet cells, on the other hand, are extremely flat compared with the coronet cells, and no cell junctions are depicted in Olsson's publication.
The suggested homology of sensory cells found in tunicates with the neuromast cells of craniates was suggested by Bone & Ryan (1978). These authors found primary sensory cells in the atrium of Ciona intestinalis, which showed a deeply inserted apical cilium surrounded by microvilli and were covered by a mucus dome. Sensory cilia with an apical cilium surrounded by microvilli are common among deuterostomes, but a cupula is uniquely present in the atrium of tunicates and in the acustico-lateralis neuromast organs in craniates. However, the case for homology of the two systems is not straightforward. The neuromasts in craniates are secondary sensory cells and are innervated by a separate nerve, whereas the sensory cells in the cupular organs of C. intestinalis are primary sensory cells with their own efferent axon. To complicate matters even more, Burighel et al. (2003) suggested the homology of the acustico-lateralis sensory structures in craniates to another group of cells present on the oral tentacles in Botryllus schlosseri, another tunicate species. These possible homologs to the neuromast cells also possess apical cilia surrounded by microvilli, and are secondary sensory cells. However, they do not possess a mucus cupula. In Branchiostoma sp., there are several sensory cell types (primary and secondary) that possess apical cilia surrounded by microvilli, but no mucus cupula has been described (Bone & Best, 1978; Baatrup, 1981; Stokes & Holland, 1995; Lacalli & Hou, 1999).
The mesoderm consists of two distinct parts in tunicate larvae: the locomotory larval musculature and the larval mesenchyme that develops into heart, blood cells and body musculature in the adult (Nishida, 1987; Hirano & Nishida, 1997). It is not known whether the mesenchyme in tunicates forms an epithelium at any stage of development. In cephalochordates, true dorsolateral somites are formed, embedded in a continuous extracellular matrix and possessing a myocoel that separates the myotome from the dermatome (Stach, 2000). The medial ventral blood vessel is accompanied by a contractile coelomic ‘vessel’ that is probably homologous to the tunicate and craniate heart. A ventral portion of the mesoderm extends in a segmental fashion ventrally around the intestine and later during ontogeny forms the transverse muscle. Dorsal of the ventral mesoderm and at the ventral border of the dorsolateral somites, the mesoderm forms segmental nephridia in amphioxus (Ruppert, 1996; Stach & Eisler, 1998). The mesoderm in craniates also forms true dorsolateral somites that, besides myocoel, myotome and dermatome, possess a sclerocoel and a sclerotome, structures involved in forming the skeletal system. Thus, the cells that produce novel structures in craniates undergo the well-known epithelial–mesenchymal transition (Thiery & Sleeman, 2006). No such transition has so far been documented in tunicates, although Katz (1983), in the only complete anatomical study of a tunicate larva based on serial transmission electron microscopy, suggests that the trunk mesoderm consists of two ‘pockets’ rather than a mesenchyme. In addition, the ‘novel structures’ that, according to Ruppert, should develop from the larval mesenchyme in tunicates and craniates are not specified. If he refers to the pericardium/heart this structure is not novel in an evolutionary sense, because it was already present in the last common ancestor of the Chordata (and maybe before). If he refers to the body wall musculature and musculature of the siphons in the adult sessile ascidians, it is not clear to what these muscles can be homologized in craniates. If he refers to the hemocytes in tunicates as being homologous to blood cells of craniates, the cells found in the blood vessels of cephalochordates (Welsch, 1975; Rähr, 1981) should also be considered. Thus, the homology and the level of possible homology of this character implied by Ruppert's coding is unclear.
Many features that are craniate apomorphies are derived ontogenetically from the neural crest (Gans, 1989; Northcutt, 1996). A peculiarity of neural crest cells in craniates is that they actively migrate away from their original position close to the ventro-lateral lines of the early nervous system to form different tissues and structures in adult craniates. One population of migratory neural crest cells forms melanocytes in the epidermis of craniates and is thus responsible for the color pattern visible in many primitively aquatic craniates. In a phlebobranch ascidian, Ecteinascidia turbinata, Jeffery and colleagues (Jeffery et al., 2004; Jeffery, 2007) demonstrated that migratory cells are responsible for the bright orange coloration observed in larval and adult individuals of this species. These authors suggested that the cells are homologous to migratory neural crest cells in craniates. It has to be noted that based on the similar expression of Pax genes, a population of cells in the ventro-lateral area of the neural system of cephalochordates can probably be homologized with neural crest cells in craniates (Holland & Holland, 2001). However, these cells are not reported to migrate away from their position in the neural system.
Ruppert suggested that the musculature in craniates and tunicates is not epithelial and a plesiomorphic epithelial condition is retained in cephalochordates. As mentioned above, the locomotory musculature of all three major chordate clades develops from dorsolateral archenteron cells. While in cephalochordates, the medial cells of the developing somites differentiate into muscle cells, the epithelial somites in craniates go through an epithelial–mesenchymal transition, after which the medial cells differentiate into muscle cells (Thiery & Sleeman, 2006). In both taxa, a myocoel separating the medial (prospective) muscle cells from a lateral cell population exists. In tunicates, there is no somitic stage with medial and lateral epithelial cells separated by a myocoel. The differentiated muscle cells in craniates and tunicates are apolar and fulfill the definition of a mesenchyme (Bartolomaeus, 1993). The situation found in cephalochordates, however, is not the classical epithelial one. The plate-like muscle cells are in contact with the extracellular material covering the nerve tube and with the myocoel. However, each cell spans through the entire length of the somite, and the cells are not connected by apical adherens junctions. Thus, coding the character according to Ruppert's suggestions leads to conflict with the ontogenetic similarities seen in the development of craniates and cephalochordates, and the distinction of the character states is rather unconventional.
Notochord differentiates beyond the stack of coin stage (Ruppert, 2005)
Again, it is necessary to describe the respective characters in the three taxa, in order to judge the primary homology hypothesis implied by the short description given by Ruppert. In cephalochordates, the notochordal cells that are arranged like a stack of coin during early ontogeny roughly remain in this arrangement. However, the cells become more irregular in shape, vacuolated and differentiate into muscle cells. In addition, several intercellular spaces develop and additional cells with different characteristics are present as part of the cephalochordate notochord (Welsch, 1968b; Ruppert, 1997a; Stach, 1999). In primarily aquatic craniates, the notochord is surrounded by a tough sheath and the cells that are originally arranged like a stack of coins continue to proliferate until an irregular arrangement is found (Boeke, 1908; Schinko et al., 1992; Grotmol et al., 2006). It is interesting to note that the cells become vacuolated like the cells in cephalochordates. In tunicates, the stack of coin stage in many (although not all) species develops into a situation that fulfills the definition of a coelom (Bartolomaeus, 1993). However, the ontogenetic mechanism is peculiar, as it seems likely that the mesenchymal cells become fenestrated and a central canal is formed (Burighel & Cloney, 1997). The suggested homology of a notochord differentiating beyond a stack of coin stage seems to be purely linguistic.
Hemal system with functionally distinct circulating corpuscles (Ruppert, 2005)
While the overall anatomy of the hemal system of cephalochordates is strikingly similar in cephalochordates and craniates (Rähr, 1979), only a single cell type has been demonstrated from the hemal system of cephalochordates (Welsch, 1975; Rähr, 1981), whereas several types of cells are present in the blood vessels of tunicates and craniates (Sawada, Zhang & Cooper, 1993; Cima et al., 2001). However, it is difficult to pinpoint more detailed similarities besides the mere differentness of the cell types that could be interpreted as evidence for homology of specific cell types.
Several epithelia, such as the ependymal cells in the central nervous system, the epithelium of the lungs or the epidermis of some early ontogenetic stages, possess multiciliated cells in craniates. In tunicates, multiciliated cells can be found, for example, in the branchial basket, the intestine and in the ciliated funnel that is associated with the neural gland (Burighel & Cloney, 1997). Multiciliated cells, however, have also been observed in the margin of the larval mouth (Lacalli, Gilmour & Kelly, 1999; Wicht & Lacalli, 2005) of cephalochordates. Also, even structures that are supposed to be homologous among the three chordate taxa show different character states in this respect. For example, Hatschek's pit is most probably homologous to the ciliated funnel in tunicates and the adenohypophysis of craniates. Yet, Hatschek's pit consists of monociliated cells, the ciliated funnel of multiciliated cells and the adenohypophysis in craniates can be without cilia [Myxine glutinosa (Fernholm, 1972); Petromyzon marinus (Wright, 1989)], multiciliated [Canis lupus (Nunez & Gershon, 1976)] or monociliated cells [Scyllium canicula (Alluchon-Gérard, 1978); Carassius auratus (Yamamoto et al., 1982)]. Besides the heterogenous distribution of the character in chordates, it has to be kept in mind that one of the potential sister taxa of Chordata, the Enteropneusta, has several multiciliated epithelia as well. Thus, a complicated character distribution plus the possibility that the presence of multiciliar and of monociliar epithelia could be the plesiomorphic condition in Chordata makes this character difficult to evaluate.
Adenohypophysis consisting of an ectoderm and an endoderm (Ruppert, 2005)
The ontogenetic derivation of the adenohypophysis is not uniform in craniates. While it seemed to be established textbook knowledge that the adenohypophysis in craniates derives from the endoderm and the ectoderm (Romer & Parsons, 1986), it is not easy to decide whether Rathke's pouch, the ontogenetic rudiment of the adenohypophysis, derives from the endoderm or the ectoderm, because no clear distinguishing morphological (or other) criteria exist to characterize the early germ layer affiliation. More modern investigations seem to suggest that adenohypophysis cells derive from ectodermal placodes in Gnathostomata (Toro & Varga, 2007) and Hyperoartia (Uchida et al., 2003), leaving the basally branching Hyperotreta as sole taxon, in which the adenohypophysis probably derives from the endoderm (Gorbman, 1983). In addition, while the neural gland complex, the structure considered homologous to the adenohypophysis in Tunicata, seems to possess endodermal and ectodermal portions (Ruppert, 1990), this can also be said about the preoral pit/Hatschek's pit in cephalochordates (Stach, 1996). Thus, the suggested coding of character states as synapomorphic for Tunicata and Craniata can be refuted based on alternative hypotheses of primary homology.
Pax1/9 (Urochordata) or Pax1 and Pax9 (Vertebrata) expression in developing pharynx and musculature (somites) (Ruppert, 2005)
There are two corresponding genes in craniates to the orthologous Pax1/9 gene that is found in tunicates and cephalochordates. These genes are expressed in the pharynx of all chordate taxa and in enteropneusts as well (Ogasawara et al., 1999; Lowe et al., 2003). In addition, Pax1 and Pax9 are expressed in the mesodermal somites of Gnathostomata (Brent & Tabin, 2002) but not Hyperoartia (Ogasawara et al., 2000). I could not find published evidence that Pax1/9 is expressed in the tail musculature of tunicates, and although the caption of this paragraph is taken unaltered from the caption of the cladogram in Ruppert (2005), Ruppert neither describes nor cites evidence for this character in the text.
The Notochordata hypothesis
Phosphocreatine as the sole phosphogen in muscle cells (Hennig, 1984)
Watts (1975) demonstrated that the energy required for muscle contraction in chordates is stored in different molecules. In tunicates, these are phosphoarginine and phosphocreatine. In cephalochordates and craniates, on the other hand, phosphocreatine is the only molecule used to store energy (Watts, 1975; Livingstone, 1991). Because phosphoarginine and phosphocreatine are also used for storage of energy in many other marine invertebrates (Rockstein, 1971), it can be inferred that the situation with two different molecules is plesiomorphic within chordates. The specialization to the usage of only phosphocreatine in cephalochordates and craniates can thus be interpreted as a shared, derived character state in notochordates. This interpretation is consistent with more recent studies of the evolution of the molecules involved in energy transfer in the respective groups, the phosphoargininekinases and phosphocreatinekinases, respectively (Ellington, 2001; Bertin, 2006; Uda et al., 2006).
Somites are complex structures consisting not just of muscle cells (see e.g. Starck, 1975b; Brent & Tabin, 2002), and as such are only present in craniates and cephalochordates. Some of the characters can occur independently in the two groups. For example, the sclerotome is only found in craniates (Mahadevan, Horton & Gibson-Brown, 2004). A lateral group of cells that is ultrastructurally different from the medial muscle cells is present in both taxa. This group of cells is similar in its lateral position and is separated by a coelomic compartment, the myocoel from the muscle cells. The entire arrangement can therefore be homologized as lateral dermatome, medial myotome and separating myocoel (Ruppert, 1997a; Stach, 2000).
The liver caecum in cephalochordates is a ventral anteriorly pointing extension of the intestinal tract behind the branchial complex. The cells in the epithelially organized caecum are probably responsible for the production of digestive enzymes but also for the storage of energy-rich components (Welsch, 1975). While the liver in adult craniates is still connected with the intestinal tract, its function is more diverse (Romer & Parsons, 1986). However, the ontogeny of the liver in most craniates, but especially in the larval lamprey, demonstrates the origin as a ventral epithelial extension of the intestine behind the branchial basket (Damas, 1944). These similarities are unique to craniates and cephalochordates, and the liver caecum can therefore be interpreted as a homologous structure, a conclusion in agreement with recent gene expression studies (Jiang et al., 2007; Tian et al., 2007).
Notochord along major part of body
This character was suggested by Nielsen (2001) and is the character after which the taxon Notochordata was named by the same author for the group comprising craniates and cephalochordates. While a notochord is present in the tails of all chordates, such a skeletal element is only present in the anterior trunk of craniates and cephalochordates.
In an elegant study, Rähr (1979) visualized the organization of the circulatory system of cephalochordates. While there are differences in the organization of the circulatory systems of primarily aquatic craniates (Romer & Parsons, 1986; Mickoleit, 2004), the overall similarity is remarkable. In order to acknowledge these similarities, three major blood vessels are chosen as representative for detailed homologies supported by Rähr's study: the dorsal posterior aorta dorsalis, which transports oxygenated blood to the caudal fin; the portal liver vein, vena portae hepatica, which connects the two capillary systems around the intestine and the liver caecum; the main vessel that collects anoxic blood before it is transported to the endostylar artery in the ventral midline in the branchial basket, the ductus cuvieri (Fig. 6).
While all chordates possess muscles that function together with fins and a notochord to propel the animal by undulatory movement, only in cephalochordates and craniates are consecutive muscle blocks separated by an extracellular material containing mainly collagen and proteoglycans (Stach, 2000; Gemballa, Weitbrecht & Sanchéz-Villagra, 2003). These muscle blocks contain several phylogenetic informative characters, as they are composed of recognizable subunits, such as dermatome, myotome or myocoel. As lucidly discussed by Ruppert (1997a), these complex muscle blocks should be called myomeres.
Shape of myomere with an anterior dorsal point
3D-reconstructions of the shape of the myosepta in Branchiostoma lanceolatum (Gemballa et al., 2003) and in hagfish and lamprey (Vogel & Gemballa, 2000) differ in many details, but are similar in that they possess a prominent anterior-pointing cone in the dorsal third of the animals.
Red and white muscle cells
Two different types of muscle cells differentiate during ontogeny in the somatic musculature in craniates and cephalochordates but not in tunicates (Starck, 1975b; Burighel & Cloney, 1997; Ruppert, 1997a). These types of muscle cells are known as fast or white and enduring or red fibers in craniates (e.g. Devoto et al., 1996). They correspond to the deep (white) and superficial (red) lamellae in cephalochordates (Flood, 1968). Only a single type of muscle cells is present in tunicates (Burighel & Cloney, 1997) that probably corresponds to the deep (white) muscle cells in notochordates.
Repeated mesodermal coelomic podocytes with apical cilia covering certain blood vessels
The segmental excretory organs in cephalochordates and craniates are complex structures that consist of several subunits that are recognizable and have been formalized as a character in the preceding paragraph. In cephalochordates as well as in craniates, each of the segmental excretory organs consists of podocytic cells that cover the extracellular matrix surrounding a blood vessel and that possess apical cilia (Lacy, Castellucci & Reale, 1987; Kluge & Fischer, 1990, 1991). While all these structures are clearly recognizable in cephalochordates, the apical cilium is, in addition, surrounded by 10 slender microvilli (Ruppert, 1994a,b, 1996; Stach & Eisler, 1998; Westheide & Rieger, 2007). This similarity to protonephridia in lophotrochozoan taxa is clearly due to convergence and the peculiar microvilli are not present in similar form in any other deuterostome species.
Segmented ventro-lateral trunk mesoderm
Ventral to the ‘Somitenstiel,’ the somites extend most probably in the form of segmented mesodermal bands in cephalochordates (Hatschek, 1881; Stach, 2000). A similar condition can also be seen in the early ontogenetic stages of lampreys (Damas, 1944; Shimeld & Holland, 2000; Kusakabe & Kuratani, 2005). In lamprey, later during ontogeny, the ventral mesoderm becomes an unsegmented mesodermal cell mass (Damas, 1944). This is the case in other craniates from the beginning where the ventral mesodermal cell mass differentiates into the lateral plate (‘Seitenplatte’) and forms the extensive body coeloms (Funayama et al., 1999).
The extracellular matrix in craniates and cephalochordates is extensive and contains a high amount of collagen, whereas the extracellular matrix in tunicates is more feeble. In addition, fibroblasts, a specific cell type situated within the extracellular matrix, can be found in craniates (Romer, 1972) and cephalochordates (Welsch, 1968a; Rähr, 1981). In both taxa, fibroblasts produce collagen molecules. While tunicates possess orthologous FGF genes, they lack the special cell type, fibroblasts, within the extracellular matrix, although it might correspond to an extra-zooidal cell type in the tunic (Swift & Robertson, 1991).
Larval cephalochordates possess a photoreceptor consisting of rows of ciliated receptors and pigment cells that are situated at the anterior tip of the neural cord, with the receptor cells situated ventrally and the pigment cells dorsally (Lacalli, Holland & West, 1994; Lacalli, 1996). This is also the initial location of the embryonic eye rudiment in craniates (Starck, 1975b; Li et al., 1997). Tunicate ocelli, on the other hand, are situated at the posterior wall of the cerebral vesicle and they possess only a single, ventrally located cup-shaped pigment cell (Burighel & Cloney, 1997). Tunicate ocelli are also unique in the possession of glycogen-filled lens cells (Eakin & Kuda, 1972), while in some species, receptor cells are similar to the ones in the remaining chordates in that they are primary receptor cells with apical ciliary structures (Gorman, McReynolds & Barnes, 1971; Barnes, 1974; Eakin, 1979). Recent investigations into the gene expression pattern of Rx genes have been taken as evidence for the homology of the larval ascidian ocellus to vertebrate eyes (D'Aniello et al., 2006). This would render the frontal anterior eye a synapomorphy of Chordata. However, orthologs of Rx genes are known from many invertebrates, including Ecdysozoa (Eggert et al., 1998) and Lophotrochozoa (Arendt et al., 2004) and even plants (Bendahmane, Kanyuka & Baulcombe, 1999). Morevover, Rx-gene expression has also been reported from the pineal eye in vertebrates (Mathers & Jamrich, 2000), supporting the hypothesis of primary homology between the tunicate ocellus and the pineal eye in vertebrates suggested above.
In the current review, recourse has been made to gene expression studies on several occasions in order to substantiate hypotheses of primary homology. While an exhaustive review of gene expression studies relating to chordate evolution is beyond the scope of this review, a few comments are necessary, because the results using this technique are extensively cited by morphologists and are utilized to support molecular phylogenies (Nielsen, 2003; Ruppert, 2005; Lacalli, 2006). Whereas gene expression studies have added an exciting insight and new levels to our understanding of embryological processes and as such are relevant for phylogenetic considerations, these data have rarely been subjected to an established cladistic methodology even where used in a phylogenetic context (reviewed e.g. in Svensson, 2004; Jenner, 2006). For example, it is largely agreed upon that homology is a relationship of two characters in different individuals that is based on the reality of a shared historical process (e.g. Hall, 1995; Bolker & Raff, 1996; Nielsen & Martinez, 2003; Svensson, 2004; Cracraft, 2005). This reality is discovered and suggested as a scientific hypothesis in a two-step process. Based on similarities, the hypothesis of primary homology is proposed and is subsequently subjected to the congruence test, where it is evaluated under an optimality criterion (e.g. parsimony or maximum likelihood; de Pinna, 1991; Kitching et al., 1998; Rieppel & Kearney, 2002; Richter, 2005; Rieppel, 2005). This is not done in most evo-devo studies as has been rightfully criticized in earlier publications (Nielsen & Martinez, 2003; Svensson, 2004; Jenner, 2006). Nevertheless, gene expression studies can be incorporated into phylogenetic analyses as evidence for primary homology hypotheses, if gene expression similarities are observed. It must be emphasized here that the reverse conclusion is not true: if no similarity is observed, this is no evidence against homology, especially not if other (e.g. morphological) similarities exist. The basis for this logical asymmetry has been noted early on (Darwin, 1859) and has been lucidly discussed by Jenner (2006). Thus, while gene expression studies can be informative in a phylogenetic context, if they are supportive of primary homology hypotheses, Chordata might be the taxon, in which the second step – coding and analyzing gene expression patterns as characters – could also be gone. This optimistic expectation derives from the fact that within Chordata, a relatively high taxon sampling in gene expression studies exists. This fortunate situation has already led to the observation that molecular developmental pathways can dramatically differ even in the specification of homologous structures (Lemaire, 2006) and should be explored further for phylogenetic purposes.
Most molecular phylogenetic studies utilized 18S rDNA sequences, and these studies date back to the beginning of the 1990s. From these early studies, it could be inferred that at least for chordates molecular sequences do not deliver only one answer. While Turbeville et al. (1994) found support for the traditional Notochordata hypothesis, other authors recovered quite unexpected groupings. In the analysis of Halanych (1998), for example, Tunicata was a sister taxon to the remaining triploblasts to the exclusion of flatworms and nematodes, and the study by Swalla et al. (2000) found Notochordata monophyletic, but Tunicata sister taxon to Ambulacraria, a taxon consisting of Echinodermata and Hemichordata, a result also found by Bourlat et al. (2003). Analyzing also 18S rDNA data, Giribet et al. (2000) found support for the Olfactores hypothesis, whereas Zeng, Jacobs & Swalla (2006) recovered an unresolved trichotomy and yet favored the traditional Notochordata hypothesis. Based on the very results of Zeng and colleagues, Swalla & Smith (2008) supported the Olfactores hypothesis. Stach & Turbeville (2002) also favored the Notochordata hypothesis; however, these authors, primarily concerned with Tunicata as the ingroup, did not address this question, but rooted their optimal trees after unordered and unrestricted analyses. In the first spectral analysis of a deuterostome dataset of 18S rDNA sequences, Wägele (2001) concluded that these data contain an exceptionally low signal-to-noise ratio and these data therefore became a textbook example for missing phylogenetic signal. Several studies expanded on the 18S rDNA datasets in terms of adding additional molecular sequences, while sacrificing taxon sampling at the same time. Winchell et al. (2002) added 28S RNA sequences and did not recover a monophyletic Chordata in their analyses. Instead, these authors recovered a sister-group relationship between Notochordata and Ambulacraria, and Tunicata as the sister taxon to this latter grouping. Naylor & Brown (1998), who phylogenetically analyzed the first complete mitochondrial genome of a cephalochordate, were also unable to support a monophyletic Chordata and suggested that the widely held opinion that adding longer sequences to converge on the true phylogeny might not always be justified. The final twist in results from molecular studies came recently, when the study of Delsuc et al. (2006), based on limited taxon sampling (14 deuterostome taxa) but an impressive 33 800 amino acid positions (c. 14 000 informative sites), supported the Olfactores hypothesis. These authors were also unable to recover a monophyletic Chordata and found the cephalochordate to be related to the sea urchin in their analysis, a finding that was considered unreliable in subsequent analyses (Bourlat et al., 2006). While the analysis by Delsuc and colleagues is based on a high amount of sequence data, it has to be kept in mind that the authors used concatenated operational taxonomic units, which might be problematic (Bapteste et al., 2008). Moreover, the conclusions are primarily drawn from model-based analyses, which are known to be sensitive to model selection (Pol & Siddall, 2001; Kolaczkowski & Thornton, 2004; Piontkivska, 2004), which is not satisfactorily explored for amino acid alignments (see e.g. (Abascal, Zardoya & Posada, 2005). In general, we can only judge which of comparatively few models best fits a given alignment, but we have no indication how close this model is to reality, which, after all, comprises several hundreds of million years of evolution in millions of taxa. In this context, the bon mot of Mishler (2005) should also be recalled, who, with respect to genomic scale phylogenetic studies, stated that complicated models for tree building can be seen as attempts to compensate for marginal data.
Phylogenetic analysis: a matrix based on all combined arguments proposed by adherents of the competing hypotheses
For the present review, I have combined all characters that have been discussed above in favor of one or the other hypothesis with the intention to make them accessible for a formal cladistic analysis. The matrix can be found as Supporting Information Appendix S1, while coding follows the arguments just presented. When primary character homology was not refuted outright, the character coding suggested by other authors was included, in order to test these homology hypotheses, even if doubts have been expressed in the presentation above. It contains the 34 characters discussed above, coded as binary characters (four are parsimony uninformative) for the three higher monophyletic chordate taxa plus outgroup, for which all characters could be coded as zeros. Analyzed in PAUP* (Swofford, 2003) under the parsimony criterion, utilizing the exhaustive search command, with all characters unordered, this matrix results in a single most parsimonious tree shown in Fig. 7 rooted with the outgroup. Tree length is 47, consistency index=0.72, homoplasy index=0.28, CI excluding uninformative characters=0.70, HI excluding uninformative characters=0.30, retention index=0.57 and rescaled consistency index=0.41. Notochordata is supported by 91.7% in a bootstrap search with 1000 branch-and-bound replicates and a Bremer support index of 7. The next best tree favors the Olfactores hypothesis, but is seven steps longer and this grouping barely receives 8.3% support in the bootstrap analysis. Support for a monophyletic Atriozoa is not measurable in this data matrix.
Discussion of the possible evolutionary sequences
A reliable phylogeny of chordate and deuterostome taxa as well as a basic understanding of the morphological characters and their functional roles for the organisms is a prerequisite to begin a serious discussion of the evolutionary events that can be reconstructed credibly for the higher monophyletic chordate taxa. At the same time, this reasoning validates the characters used beyond plotting them on a more or less reliable tree derived from molecular sequences. Unfortunately, at the present state, the evolution of chordates from a remote ancestor they did share with other deuterostomes has to remain elusive on many details. Yet, it is possible to tentatively gauge some of the scenarios that have been suggested so far for the evolutionary origin of chordates in light of the evidence discussed above.
The major methodological advance in cladistics had been the realization that no present-day animal species can serve as a surrogate for ancestral species, but that each character had to be evaluated independently. The evidence compiled here suggests that several characteristics evolved in the stem lineage of notochordata, such as a notochord stretching over the most part of the body, ventral and dorsal fins, somites with dermatome, myotome, myocoel and fast and slow muscle fibers, a distinct liver caecum, a highly sophisticated circulatory system and an inverse anterior eye. All of these characters are related to more efficient locomotion and the demands this puts on the general metabolism. This is in general agreement with major parts of scenarios by Barrington (1965), Berrill (1955, 1987a,b), Gans (1989) and Northcutt (1996), Nielsen (1994, 1999), Ruppert (1994a, 1997a,b) and others (see also reviews by Denison, 1971; Gee, 1996 or Jefferies, 1986). At the same time, the evidence derived from comparative morphological studies makes it difficult to substantiate hypotheses in which tunicates are derived from segmented animals equipped with somites (Kimmel, 1996; Delsuc et al., 2006; Rychel et al., 2006; Dunn et al., 2008; Putnam et al., 2008).
The scenario proposed by Garstang (1928) entails that the evolutionary lineage leading from the ancestor of all deuterostomes to the last common ancestor of notochordates did possess a sessile adult stage. Whether this claim can be supported by cladistic argumentation relies on the phylogenetic reconstruction of tunicates. While the phylogeny of tunicates is even less well resolved as the phylogeny of chordates, current morphological evidence supports the hypothesis that the last common ancestor of tunicates had a sessile adult stage (Stach, 2005, 2007; Stach & Kirbach, in press). Thus, Garstang's hypothesis can still be supported in this respect. Tunicate evolution is important in another sense as well. In Garstang's argument, progenesis plays an important role (see Stach & Turbeville, 2005 for a review). This argument can still be upheld, but not in its original form, as detailed comparisons of developmental sequences can so far only be used to support evolutionary acceleration in the stem lineage of tunicates and within tunicates in the stem lineage of larvaceans (Ruppert, 1997b; Stach et al., 2008). This, however, might be correlated with the newly acquired ability to produce a protective covering consisting of cellulose perhaps through lateral gene transfer (Mathysse et al., 2004; Nakashima et al., 2004; Davison & Blaxter, 2005). This ability might have profound effects on larval development, because there might be a conflict between the need to feed and the advantage of being protected. While this is another problem that deserves further investigation, it logically entails that the heterochrony observed in tunicates is probably independent from the potential neoteny postulated by Garstang in the stem lineage of notochordates. In conclusion, if the goal is a better understanding of the evolution of chordates, Tunicata is still the key taxon. Moreover, it will be necessary to establish the sister group of Chordata and fill the existing gaps in detailed anatomical knowledge of all life-cycle stages to obtain precise reconstructions of the last common ancestors of supported monophyletic groups.
While the present paper was under review, two phylogenetic studies based on large amounts of sequence data but limited taxon sampling were published by Dunn et al. (2008) and Putnam et al. (2008). Both supported the Olfactores hypothesis.
I want to thank Peter Adam for his skillful line drawings. Financial support for this study through grants Sta 655/1-1&2 and 655/2-1of the DFG is gratefully acknowledged.