Evolution of bilaterian central nervous systems: a single origin?

de Beer [121] was one of the first to recognize the hierarchical nature of homology when he noted similar morphological features in two different animals could develop under the control of different genes; a phenomenon now known as genetic piracy [122]. The converse is also known – where parts of homologous gene networks are involved in the development of apparently nonhomologous structures [123]. When such disconnects are discovered, some would pay more attention to structure [124], and others would pay more attention to the genes (as a deep homology) [125]. In discussing neural evolution in higher deuterostomes, Lowe and colleagues strongly favor genes over morphological features as arbiters of homology. Thus, they maintain that structures with very different morphology (for example, the proboscis of a hemichordate and the forebrain of a vertebrate) are not morphologically homologous [25,126] even though conserved gene expression patterns may indicate a common evolutionary origin. Thus, while those authors find that homologous genetic programs operate at the anterior end of the hemichordate proboscis and vertebrate ANR, at the boundary between the hemichordate proboscis and collar and at the vertebrate ZLI, and at the boundary between the hemichordate collar and trunk and at the vertebrate ISO located at the MHB, they do not refer to these three regions of the hemichordate ectoderm and vertebrate brain as homologs [9]. Even so, as discussed by Wagner [16], these regions could be homologous even though morphologically divergent.

Lowe and colleagues, in their series of papers concerning genes and structures involved in neural evolution of deuterostomes, have been somewhat inconsistent in their treatment of the subject of homology. In their initial work, which concerned expression of 22 genes with restricted ectodermal domains along the A/P axis (Table 1), they concluded that the surface ectoderm of S. kowalevskii and the chordate CNS have a common ancestry [24]. These authors noted that S. kowalevskii ‘shows pervasive neurogenesis with no large, contiguous non-neurogenic subregion, as occurs in chordates’ and concluded that the deuterostome ancestor had a diffuse ectodermal nerve net that evolved into the vertebrate CNS [24]. One difficulty with this argument is that, as Aronowicz and Lowe [25] later noted, the surface ectoderm of invertebrate chordates outside the neural tube contains widespread ectodermal sensory neurons (Figure 1) [127,128], while in vertebrates the large pan-placodal region outside the neural plate is highly neurogenic [68]. In addition, although the authors maintained that most of 22 genes they studied are not expressed in the “epidermal ectoderm” in chordates [24], in the 10 years since this paper was published, it has become clear that all except possibly Vax, Rx and Nkx2-1 are expressed in ectodermal sensory cells in chordates (Table 1) [68,129-131]; expression of Vax in amphioxus and of Nkx2-1 in the tunicate Ciona intestinalis have not been determined. For example, Sox1/2/3 (SoxB) and Hu/Elav are expressed in ectodermal sensory cells in chordates [129,130], Hox1, Hox3 and Hox4 are expressed in nested patterns in the amphioxus ectoderm and at especially high levels in ectodermal sensory cells [132], ascidian Hox1 is expressed in development of ectodermal sensory cells [133], and some Hox genes are expressed in placode derivatives in vertebrates [134].

In addition, Gbx is expressed in the “non-neural” ectoderm in amphioxus [135] and in developing placodes in vertebrates [136], while Gbx and Otx2 mutually repress one another in development of the otic placode as they do in A/P patterning of the CNS [137]. Therefore, as noted above, the ectodermal expression patterns of genes along the A/P axis do not clearly distinguish CNS from ectoderm outside the CNS. Indeed, Aronowicz and Lowe noted that genes such as Otx and Hox appear to be involved in patterning neurogenic tissues generally [25]. Thus, expression of these A/P patterning genes alone does not distinguish between theories that the ancestral bilaterian and ancestral deuterostome had an ectodermal nerve net or whether they both had a CNS. As noted above, genes expressed exclusively in the CNS of bilaterians (for example, earmuff/Fezf/Fezl and D/V patterning genes) are generally more informative for inferring the course of evolution of the CNS [64]. Indeed, domains of homologs of three genes mediating lateral to medial (D/V) patterning in the chordate and protostome CNSs (Hh, Nkx2.2, Msx) [138-140] are not expressed in comparable patterns in the hemichordate ectoderm [113]. Thus, if the chordate CNS evolved from a nerve net in the ancestral deuterostome, expression of these lateral to medial genes would represent convergent evolution in the CNSs of protostomes and chordates.

The similarities in patterning the ANR and the anterior part of the hemichordate proboscis, the ZLI and the boundary between the proboscis and collar, and the ISO and the boundary between the collar and trunk lead to the conclusion that in the ancestral deuterostome the role for these signaling centers, which ultimately gave rise to the vertebrate ANR, ZLI and ISO, was to regionalize the general body plan [9]. However, Pani and colleagues also proposed that amphioxus partially lost the ANR, and completely lost both the ZLI and ISO, while the tunicate C. intestinalis lost the ZLI and partially lost both the ANR and ISO [9]. They conclude that in ‘certain cases hemichordates will be a more informative group than basal chordates for reconstructing stem chordate characters and understanding the origins of vertebrate developmental genetic processes’ [9]. As this is quite an extreme view and is at odds with conclusions based on morphology, gene expression and gene function in amphioxus, tunicates and vertebrates, the evidence merits close examination.

The vertebrate ANR is characterized by expression of Fgf8, Six3, Pax6, Otx2, Sox2 with Dlx5 expressed in adjacent non-neural ectoderm and Bf-1 (FoxG-1) expressed in the rostral forebrain (Figure 4) [160], while Sfrp1a is expressed in the anterior/ventral part of the developing neural tube [161]. In addition, transplantation of the ANR laterally expanded the telencephalon and promoted expression of Bf-1 (FoxG1), demonstrating that the ANR is an organizer [162]. In S. kowalevskii, homologs of these genes are expressed in the proboscis ectoderm, although their relative domains are rather divergent from those of their homologs in the vertebrate forebrain. For example, Six3, Sfrp1/5 and Fgf8/17/18 are strongly expressed in the anterior proboscis ectoderm, and Sox1/2/3 is broadly expressed in the proboscis and anterior part of the collar [9]. However, Pax6 is not expressed at all in the anteriormost proboscis ectoderm but more posteriorly throughout much of the proboscis and collar, while Otx is chiefly expressed in the collar ectoderm with only very weak expression at the anterior end of the proboscis. In addition, Dlx is very patchily expressed in the proboscis ectoderm, but appears to be highly expressed where the dorsal nerve cord will form. Hh is expressed in ectoderm at the tip of the proboscis in S. kowalevskii[9], while in vertebrates it is expressed in the basal plate of the forebrain and midbrain and in the floor plate of the anterior hindbrain as well as in the prechordal plate [163].

Experimental evidence for homology of gene networks patterning the ANR and anterior proboscis ectoderm in S. kowalevskii came from manipulation of Fgf and Hh signaling. Although inhibition of Fgf or Fgf signaling scarcely affected expression of the proboscis domain of the anterior marker Rx, it did eliminate FoxG1 (BF-1) expression in the proboscis. In addition, knockdown of Hh eliminated expression of the anterior marker Fgf-Sk1. Thus, inhibition of Fgf or Hh signaling affects development of the proboscis and expression of some anterior markers. However, Green and colleagues found that the major role of Fgf8/17/18 in development appears to be in mesoderm induction [164]. Therefore, the effects of inhibition of Fgf signaling on anterior development could be secondary to those on mesoderm.

Whether this evidence is sufficient to indicate that the vertebrate ANR evolved from the anterior part of a diffuse ectodermal nerve net in the deuterostome ancestor is open to question. A major difficulty in identifying homologous gene networks is deciding how much of two gene networks must be conserved for them to be considered homologous [165]. This is a particular problem when the morphology is not conserved. Gene networks can include several thousand genes, and it is well known that core parts of signaling pathways are often coopted for patterning nonhomologous structures [166]. Thus, there are several alternative explanations for similarities in gene expression between the hemichordate proboscis ectoderm and the vertebrate ANR. One is that similar expression of a suite of genes including Fgfs in these two regions may simply be indicative of ancient roles in specification of the anterior end of embryos in general. For example, Sinigaglia and colleagues reported that Six3/6, FoxQ2, Irx, SoxB1 and Fgf are expressed in the aboral region in the cnidarian N. vectensis, while Wnts are expressed at the opposite end of the embryo [39]. Six3/6 is initially involved in specification of the anterior end of the embryo and later in neurogenesis. Fgf signaling is required for development of the apical tuft of cilia. Therefore, roles for these genes in patterning the hemichordate proboscis may simply reflect an inheritance from prebilaterians. Another possibility is that if, as comparisons of protostomes and chordates indicate, these genes were coopted for patterning the anterior CNS of the ancestral bilaterian, then the ancestral deuterostome might have had a more extensive CNS than modern hemichordates. If so, the forebrain may have been lost as the proboscis evolved, and the genetic pathways for anterior neural development coopted into the proboscis ectoderm.

Expression of Rx, Fgf8/17/18 and other genes in the hemichordate proboscis is compatible with either theory. In chordates, Rx expression is apparently restricted to the anterior end of the forebrain (it is not expressed in sensory ectoderm) [69], while Fgf8/17/18 is expressed in the ANR and telencephalon of vertebrates as well as throughout the forebrain of amphioxus. In contrast, Hh is expressed in the anterior tip of the notochord in amphioxus and in the floorplate but not in the anterior CNS. Given that Fgf8 and Hh have been coopted in vertebrate limbs, which evolved in gnathostomes, for patterning the apical ectodermal ridge (AER) and zone of polarizing activity (ZPA) [167], it is certainly possible that these genes plus Rx and other genes involved in neuronal specification may have been coopted for patterning the proboscis ectoderm of hemichordates.

Another possibility, proposed by Nomaksteinsky and colleagues [88], is that the ectoderm of developing S. kowalevskii may represent a transient larval nervous system unrelated to that of the adult. Evidence is that SoxB genes, which mark the neural plate in chordates, are broadly expressed in the ectoderm of indirectly developing embryos of the hemichordate Ptychodera flava, with expression becoming localized during development to the stomodaeum, ciliary bands and apical ciliary tuft [114]. Dlx is expressed in the aboral ectoderm, with a concentration towards the apical tuft, and Fz5/8 and FoxQ2 are expressed in the apical tuft [168]. Similarly in S. kowalevskii, Six3, Fz5/8 and Dlx are broadly expressed in the proboscis ectoderm and Bf-1 (FoxG1) in the apical region [9,24]. The tornaria larva of indirect developing hemichordates has much in common with the dipleurula type of echinoderm larvae, which also have an apical tuft and a band of cilia around the mouth. Moreover, gene expression in the apical tuft and surrounding ectoderm is highly conserved in the sea urchin and hemichordate larvae, and there is considerable similarity with expression in the proboscis ectoderm of S. kowalevskii embryos [169]. In sea urchins, FoxQ2, Fzl5/8, Sfrp1/5, Dkk3 and Six3 are expressed at the apical end of the larva, with Wnt signaling acting to restrict expression of these genes to the anterior end of the larva [169,170]. Similarly, in S. kowalevskii knocking down the Wnt receptor Fz5/8 results in a posterior expansion of the domains of anterior markers [9], indicating that Wnt signaling is also regulating expression of genes in the proboscis ectoderm as it is in the apical ectoderm of sea urchin larvae.

The vertebrate ZLI is positioned and regulated, at least in part, by anterior expression of Otx and Fezf abutting a posterior domain of Irx[171,172]. Organizer properties are conferred by Hh, which is expressed at the ZLI and which is regulated by Wnt8b (Figure 4) [173]. In fish, knockdown of Fezf2, which is expressed anterior to Irx, eliminates the prethalamus and causes mis-specification of the ZLI. Correspondingly, knockdown of Otx inhibits expression of Shh at the ZLI and reduces expression of Ptc1 and Wnt8b[172]. Evidence for organizer properties of the vertebrate ZLI comes from grafting experiments and implants of Fgf8-coated beads, which induce the rostral neuroepithelium to develop an ectopic and polarized mesencephalon/metencephalon (reviewed in [174]).

The region of the hemichordate ectoderm expressing homologs of ZLI genes lies at the boundary between the proboscis and the collar [9] and is marked by the posterior limit of Rx expression and bands of Fgf8/17/18, FoxG, Otx, Wnt8, Hh, Ptch, FoxA and Dlx. Congruent expression of Wnt8 and Hh in S. kowalevskii is consistent with a possible role of Wnt8 regulating Hh as it does at the ZLI in vertebrates. However, domains of Fezf and Irx, which abut at the ZLI in vertebrates, do not abut at the proboscis/collar boundary in S. kowalevskii. Expression of Fezf was not shown, but the major domain of Irx appears to be congruent with that of engrailed in the posterior part of the collar [24].

For S. kowalevskii, the experimental evidence presented by Pani and colleagues for the proboscis/collar boundary region acting as a signaling center was that knockdown of Otx expression reduced the intensity of the stripe of Hh expression just posterior to the proboscis/collar boundary, while inhibition of Hh signaling downregulated Dlx both at the proboscis/collar boundary and more anteriorly and also reduced the size of the proboscis [9]. These results show that Otx may act upstream of Hh at the proboscis/collar boundary, while Hh is vital for normal development. However, Otx and Shh also interact in patterning the vertebrate midbrain, not just the ZLI, and knockdown of Otx results in dorsal expansion of the Shh domain and a dorsal and anterior rotation of the MHB [175]. Thus, the relationship between Otx and Hh and perhaps some other parts of the gene network may be evolutionarily conserved between S. kowalevskii and vertebrates, but whether tissue at the proboscis/collar boundary in an ancestral deuterostome evolved into the ZLI is open to question given the markedly different expression of key genes such as Fezf and Irx as well as the lack of anatomical similarity.

In the vertebrate CNS, the MHB or isthmus functions as an organizer, and is therefore termed the isthmic organizer or ISO. The MHB is positioned by opposition of anterior Otx and posterior Gbx. Organizer properties are conferred by the action of a suite of genes including En, Fgf8/17/18, Wnt1 and Pax2/5/8 (Figure 4) [176]. Otx and Fgf8 are expressed in the midbrain and Gbx in the anterior hindbrain, with the Wnt1, Engrailed and Pax2/5/8 domains spanning the boundary; mutual repression of Otx and Gbx positions the ISO, and Wnt signaling is required for expression of Engrailed. Pani and colleagues also presented evidence for a homologous gene network in S. kowalevskii operating at the boundary between the collar and trunk [9]. Engrailed, Fgf8/17/18, Wnt1 and Gbx are expressed near this boundary. In addition, clonal suppression of β-catenin effectively inhibited Engrailed expression, suggesting that Wnt signaling regulates Engrailed, as it does at the ISO in vertebrates, while suppression of Fgf8/17/18 reduced Engrailed expression at the collar/trunk boundary. However, despite these similarities, there are several problems in interpreting the gene network at the collar trunk boundary in S. kowalevskii as homologous to that patterning the vertebrate ISO. Importantly, as the authors note, the ‘spatial arrangements of Otx and Wnt1, and Gbx and Fgf8/17/18 are reversed in S. kowalevskii compared to the ISO in vertebrates’ [9]. Moreover, suppression of Fgf signaling had no effect on expression of Pax2/5/8. A problem in interpreting these experiments is that in vertebrates, Fgf8 has an early role in induction of neural tissue, while patterning the ISO is a relatively late role. In S. kowalevskii, Fgf8/17/18 is expressed from the blastula stage, and by the neurula stage expression is restricted to anterior ectoderm.

Knockdown experiments show that Fgf8/17/18 is required for mesoderm induction [164]. Therefore, effects of Fgf8/17/18 inhibition in S. kowalevskii on Engrailed expression may be secondary to the loss of mesoderm and not directly related to effects on A/P patterning. Additional evidence for the common ancestry of the collar/trunk boundary and the vertebrate ISO was that Hox genes are expressed in nested patterns in the trunk ectoderm. However, it appears that at least Hox1 and perhaps Hox4 are also expressed in the dorsal part of the collar (Figure five in [24]), indicating that Hox expression in the ectoderm is probably not congruent with that in the collar, and raising the question of the genetic mechanisms patterning the collar nerve cord and how they might compare with those in chordates.

As for the ANR and ZLI, drawbacks to interpreting the boundary of the collar and trunk in hemichordates with the vertebrate ISO as having common ancestry not only include differences in expression of key genes expressed in these regions and also the general phenomenon of cooption of parts of gene networks for new functions. For example, in the AER of the vertebrate limb bud, Wnt3a induces expression of Fgfs, while ectopic Engrailed (En-1) induces ectopic Fgf8 expression [177,178]. Moreover, in amphioxus, engrailed is co-expressed with Wnt8, but not at the MHB, suggesting that coexpression of Wnt and engrailed is not wedded to the MHB. The questions therefore become: how conserved must gene networks be in order for one to be reasonably certain that two morphologically rather different structures have a common ancestry; and if they do share an ancestry, can one distinguish whether portions of the gene networks operating in the CNS of a deuterostome ancestor were transferred to the hemichordate ectoderm as the hemichordate CNS became reduced or whether the ancestral deuterostome lacked a CNS and used these gene networks to partition the A/P axis?

In addition to asserting that the S. kowalevskii ectoderm is not only homologous to the vertebrate CNS but also has homologs of the ANR, ZLI and ISO, Pani and colleagues maintain that the invertebrate chordates, amphioxus and tunicates, have lost all or part of these three regions [9]. Here we draw attention to data that are not consistent with such a view (Figure 4).

Although transplantation experiments in vertebrates demonstrated that the ANR functions as an organizer [163], such transplantation experiments are not feasible for hemichordate, amphioxus or C. intestinalis embryos due to their small size. However, it is clear that much of the gene network for specification of the ANR and conferring organizer properties upon it is present in amphioxus. Fgf8/17/18 is expressed in the entire forebrain of amphioxus, and Bf-1 (FoxG1) is expressed at the tip of the forebrain [145] as are Pax2/5/8 and Six3/6[43]. Otx and Pax6 are expressed in comparable patterns with strong expression in the anterior forebrain [135,179], and the Wnt antagonist Sfrp1/2/5 is expressed in the anteriormost dorsal ectoderm, including the most anterior neuroectoderm [180]. Moreover, as in vertebrates, Dlx is expressed in ectoderm outside the neural tube as well as in the edges of the anterior neural plate [70], while Hh is expressed in the underlying tip of the notochord [181,182]. Taken together, these expression patterns indicate that amphioxus has most of the components for an ANR comparable to that in vertebrates.

Tunicates have undergone some radical changes to their genome and their anatomy in evolution. For example, the Gbx gene has been lost, and there are relatively few neurons – none in the tail nerve cord of ascidians. Otx is expressed anteriorly in the CNS, but the evidence for an ANR in C. intestinalis is relatively weak. In addition to Otx, Pax6 is expressed in the anterior CNS, but not Fgf8/17/18 and Pax2/5/8, Hh, or Gli[183]. Expression of Bf-1 is not known. Therefore, it may be that as tunicates adopted early decision of cell fates and decreased the number of cells in the nerve cord, the need for an anterior brain organizer diminished.

For the ZLI, patterns of gene expression indicate that amphioxus probably has much of the genetic mechanism in place. In amphioxus, Irimia and colleagues showed that anterior expression of Fezf abuts posterior expression of IrxB about the midpoint of the forebrain [64] – approximately where Wnt8 is expressed [184]. Similarly, Wnt8b expression at the ZLI in vertebrates is flanked by Lfng, while in amphioxus the Fng domain appears to be posterior to and possibly abutting that of the Wnt8 domain at the late neurula stage [184,185]. Dlx, Nkx2-2 and Gli, which mediates Hh signaling, are also expressed in this region, while Fgf8/17/18 is expressed throughout the forebrain [70,71,186]. Hh is expressed in the floor plate as it is in vertebrates, but it is unclear at the early neurula stage whether it is expressed congruently with Engrailed and Wnt8 or a little more posteriorly. Later expression is limited to a zone posterior to the forebrain [181]. Inhibition of Fgf signaling at the late blastula does not inhibit expression of neural plate markers, but eliminates that of Otx in the cerebral vesicle (forebrain) and reduces its size [187], indicating that Fgf signaling is essential for development of the forebrain. Similarly, upregulation of Wnt/β-catenin signaling reduces expression of Otx in the forebrain and eliminates expression of the anterior marker FoxQ2[188], showing that suppression of Wnt/β-catenin signaling by inhibitors such as Sfrp1/2/5 [180] is essential for forebrain development.

For ascidians, expression of Irx and Fezf is not known. C. intestinalis has two Wnt genes that are possibly related to Wnt8 but which have long branches in phylogenetic analyses and for which assignation remain unclear [189]. Expression of these genes has not been characterized. However, Gli, Hh and Fgf8/17/18 are not expressed in the anterior CNS. Even so, inhibition of Fgf signaling blocks expression of the anterior marker Six3/6, indicating that Fgf signaling is required for development of the anterior brain [190]. Thus, while C. intestinalis likely lacks most of the gene network specifying the vertebrate ZLI and conferring organizer properties, until the expression patterns of Irx and Fezf are known the presence of part of the network for specification of the ZLI cannot be ruled out.

Amphioxus also appears to have in place the genetic mechanism for positioning the ISO, but this region does not express homologs of the genes that confer organizer properties on the vertebrate ISO. In the amphioxus brain, as in that of vertebrates, the domain of Otx expression in the forebrain/midbrain abuts that of Gbx in the hindbrain [135]. In vertebrates, organizer properties are conferred on the ISO by expression of Fgf8, Wnt1, Engrailed and Pax2/5/8 genes at this boundary. In amphioxus, Wnt1 and Engrailed are not expressed between the Otx and Gbx domains, although Fgf8/17/18 is expressed anterior to and abutting this boundary, while Pax2/5/8 is expressed posterior to and abutting it [187,191,192]. Engrailed is expressed together with Wnt8 approximately at the boundary between the Fezf and IrxB domains. Interestingly, there is also a stripe of Engrailed in the general ectoderm that is in register with that in the CNS [193].

The tunicate C. intestinalis appears to have more of the ISO network in place than amphioxus does, but it has lost a key gene – Gbx. Pax2/5/8 and Fgf8/17/18 are expressed in the neck region in a few cells just posterior to the posterior limit of Otx. Engrailed is expressed in two domains, one near the Pax2/5/8 domain and the other overlapping with that of Otx[194].

The emerging picture is that amphioxus has most, if not all of the genetic mechanism for the ANR in place and part of that for the ZLI and ISO while the ancestral tunicate probably added to these gene networks but subsequently lost parts of them as the CNS became simplified. The most parsimonious explanation is that the ancestral bilaterian had a CNS in which portions of these gene networks were used to establish divisions in the CNS and perhaps to some extent in the ectoderm generally, and these gene networks became modified in the protostome and deuterostome lineages. Under this scenario, the basic gene networks for specifying the ANR, ZLI and MHB began to be established in the ancestral chordate or even earlier in the ancestral deuterostome or ancestral bilaterian, and additional genes were recruited to these networks before tunicates branched off the lineage leading to vertebrates. In the tunicate lineage, in conjunction with shrinking genomes, a switch from regulative to determinate development and a reduction of cell numbers in the CNS, some parts of the ancestral tunicate networks for these brain regions were lost. Modifications occurring in the hemichordate lineage may have involved changes in the CNS and/or cooption of additional genes and portions of gene networks to the surface ectoderm.

The bilaterian CNS typically develops from a longitudinal strip of ectoderm that is then internalized. Although the remainder of the ectoderm is usually termed “non-neural”, it also contains numerous neurons – often sensory neurons – which send signals to the CNS. Therefore, the ectoderm really should be thought of as “more neural” and “less neural” rather than “neural” and “non-neural”. This is highly relevant for discussion of how a CNS evolved from an ectodermal nerve net, since part of the genetic mechanism(s) for patterning the bilaterian CNS are shared with the ectoderm as a whole.

The basic A/P patterning mechanism in bilaterians with Six3/6 expressed anteriorly and Wnt genes posteriorly was inherited from an ancestral cnidarian-like animal. Additional genes (for example, Hox genes) were recruited to pattern the A/P axis of the ancestral bilaterian. Consequently, even though the ectoderm in annelids and arthropods is segmented, and the ectoderm of hemichordates and chordates is not, several genes that mediate A/P patterning of the CNS are expressed in similar patterns along the A/P axis of the non-neural ectoderm in early embryos/larvae of both protostomes, chordates and hemichordates. For example, the patterns of Otx, Gbx and Hox genes in the early ectoderm of the annelid P. dumerilii are conserved in the neuroectoderm of amphioxus embryos with an anterior Otx domain abutting that of Gbx, anterior to nested domains of Hox expression [48]. Expression of Hox genes and Gbx in the “non-neural ectoderm” in amphioxus is similar to that in the CNS [131,134,192]. In contrast, in the hemichordate ectoderm, while Otx and Gbx domains lie anterior to nested Hox domains, that of Gbx lies in between two domains of Otx[9].

More critical for distinguishing a CNS from the remainder of the ectoderm is the expression of the lateral/medial patterning genes (msh/Msx, ind/Gsh1, vnd/Nkx2.2), because expression of these genes is similar in the CNS of protostomes and chordates, but they are not expressed in comparable patterns in the “non-neural” ectoderm of either group.

The idea that the hemichordate ectoderm has organizing centers patterned by gene networks equivalent to those operating at the ANR, ZLI and ISO in the vertebrate CNS is based on similarities in gene expression as well as in some conserved gene interactions. However, some key genes involved in establishing these organizing centers in the vertebrate CNS are not similarly expressed in the hemichordate ectoderm. For example, in the S. kowalevskii ectoderm, the Fezf and Irx domains, which mark the vertebrate ZLI, do not abut at the collar/proboscis interface, while patterns of Otx and Wnt1 as well as those of Gbx and Fgf8/17/18, which mark the ISO, are reversed. Moreover, since parts of gene networks are often coopted for roles in different tissues, the fact that two genes interact does not suffice to demonstrate that a given region acts as an organizer. Thus, without transplantation experiments or at least the demonstration that a given morphogen such as Fgf8/17/18 can induce a change in cell fate when ectopically expressed, conclusions that homologs of these organizing centers are present in hemichordates are premature.

In addition, since several genes involved in specification of the ISO and ZLI are expressed in comparable positions in the amphioxus and vertebrate CNS as well as in the protostome CNS, claims that amphioxus has completely lost these regions are unwarranted. A more parsimonious explanation is that amphioxus has part of the machinery in place, upon which tunicates and vertebrates elaborated. Tunicates do appear to have more of the genetic machinery for specification of the ISO than amphioxus does, but have lost some key components, such as the Gbx gene.

The Ambulacraria (hemichordates and echinoderms), which are the sister group to chordates in the deuterostomes, are interesting in regard to what evolution can do. Both hemichordates and echinoderms have ectodermal nerve nets plus nerve cords. Gene expression indicates that the nerve cords of echinoderms, which have evolved pentamerous symmetry, are probably not homologous to chordate or protostome nerve cords. Hox genes, for example, are expressed in the developing coeloms of the adult, but not in the nerve cords or ectoderm [48,51]. It has been suggested that one or the other nerve cord of hemichordates is evolutionarily related to the chordate CNS, with historical opinions generally favoring the dorsal cord. However, dorsal expression of BMPs suggests that this homolog would be the ventral nerve cord even though it is the dorsal nerve cord that neurulates. If this is true, then a D/V inversion would have occurred at the base of the chordate lineage.

In sum, similar expression patterns of developmental genes involved in both A/P and D/V patterning in the protostome and chordate nerve cords as well as anatomical and functional similarities support the view that the ancestral bilaterian had a CNS. In the Ambulacraria, echinoderms may have lost this CNS, while it remains to be seen whether either of the hemichordate nerve cords is homologous to that of protostomes and chordates. Similarities of gene expression in the hemichordate ectoderm and chordate CNS may reflect conservation of A/P patterning in the ectoderm as a whole and in the CNS of bilaterians and/or a merging of the anterior portion of the brain and the proboscis ectoderm during evolution of hemichordates. A comprehensive study of expression of developmental genes in the hemichordate nerve cords would help to resolve these questions. In any event, similarities in gene expression between the CNS of amphioxus and vertebrates indicate that amphioxus has at least part of the genetic mechanisms for the ANR, ZLI and MHB in place and suggest that vertebrates subsequently elaborated upon these gene networks.