Concordia discors: duality in the origin of the vertebrate tail

Abstract

The vertebrate tail is an extension of the main body axis caudal to the anus. The developmental origin of this structure has been a source of debate amongst embryologists for the past century. Some view tail development as a continuation of the morphogenetic processes that shape the head and trunk (i.e. gastrulation). The alternative view, secondary development, holds that the tail forms in a manner similar to limb development, i.e. by secondary induction. Previous developmental studies have provided support for both views. Here I revisit these studies, describing caudal morphogenesis in select vertebrates, the associated genes and developmental defects, and, as a relevant aside, consider the developmental and evolutionary relationships of primary and secondary neurulation. I conclude that caudal development enlists both gastrulation and secondary induction, and that the application of recent high-resolution cell labelling technology may clarify how these discordant programmes interact in building the vertebrate tail.

Introduction

A post-anal tail distinguishes vertebrates and their fellow chordates from the remaining Bilateria. As an extension of the main body axis, the tail comprises a dorsal, hollow neural tube, and a notochord flanked by somitic mesoderm (from which segmental musculature and vertebral elements arise in some vertebrates). Naturally, an understanding of how these structures develop in the tail would provide valuable insight into processes that shape the entire vertebrate body. Over 70 years of study, however, have brought little consensus, instead polarizing developmental biologists into two camps. One camp contends that caudal development is a continuation of the processes occurring in the head and trunk, whereas the other suggests that the tail follows a distinct developmental programme (Fig. 1).

Fig. 1.

Vertebrate tail development by continued rostral development (model 1) and secondary development (model 2). According to model 1, morphogenetic movements (i.e. gastrulation) occurring within the head end of the embryo continue beyond the anus (dashed arrows). Secondary development (model 2), however, involves secondary inductive events, occurring within a tailbud comprising homogeneous mesenchyme, regulated by morphogenetic signals (white arrows) emanating from a thickened ectodermal ridge, i.e. VER. Secondary structures (neural tube, notochord, somites) arise directly from mesenchyme without germ layers as an intermediate. Whole arrows denote the progress of development.

An early and persistent perspective

The embryogenesis of the vertebrate tail first drew scientific interest and debate in the mid-1920s with the work of the German biologist D. E. Holmdahl (1925a, b, c). Holmdahl theorized that vertebrate development proceeds in two successive phases, which he termed primärer Körperentwicklung (‘primary body development’) and sekundärer Körperentwicklung (‘secondary body development’). In primary development, structures of the head and trunk differentiate from the primary germ layers established by gastrulation. Secondary development, which Holmdahl proposed takes place in the tail, does not involve germ layers as an intermediate. Instead, tail structures arise directly from a terminally located, blastema-like mass of mesenchymal cells in a manner comparable to limb bud development (Fig. 1).

The tailbud, particularly of mammals, seems to corroborate Holmdahl's claim. It is apparently a homogeneous mass of cells at the caudal limit of the embryo from which the tail arises (Griffith et al. 1992). Furthermore, it comprises loosely associated mesenchymal cells in most vertebrates (Schubert et al. 2001). This was first pointed out by von Kölliker (1879, 1884, 1889), who recognized that the caudal reach of the nervous system arose from mesoderm in a process referred to as secondary neurulation.

Secondary neurulation has since been described in many vertebrate groups, including lampreys (Lampetra japonica), neopterygian fishes (Teleostei, Lepidosteus, Amia), various sarcopterygian fishes (Lepidosiren, Protopterus), frogs (Xenopus laevis), birds (chick and quail), and mammals (rat, mouse, opossum, pig, hamster and human) (Kingsbury, 1932; Criley, 1969; Hughes & Freeman, 1974; Schoenwolf & DeLongo, 1980; Balinsky, 1981; Nakao & Ishizawa, 1984; Schoenwolf, 1984; Müller & O'Rahilly, 1987; Zehr et al. 1989; Griffith et al. 1992; Nievelstein et al. 1993; Papan & Campos-Ortega, 1994; Schmitz et al. 1994; Hall, 1998; Beck & Slack, 1999). In secondary neurulation, tail bud tissue condenses to form a solid medullary cord (or neural keel) that cavitates (i.e. hollows out) into a neural tube. Typically, cavitation involves the coalescence of several smaller cavities into the central canal or secondary neurocoel (Hughes & Freeman, 1974; Schoenwolf & DeLongo, 1980). This mode of neurulation accounts for the neural tube along the entire length of the body of neopterygian fishes and lampreys (Nakao & Ishizawa, 1984; Reichenbach et al. 1990; Schmitz et al. 1993; Papan & Campos-Ortega, 1994).

In frogs, birds and mammals, however, secondary neurulation occurs only in the caudal end following primary neurulation at the head and trunk (Schoenwolf & DeLongo, 1980; Schoenwolf, 1984; Nievelstein et al. 1993; Beck & Slack, 1999). The primary neural tube forms by the folding of the neural plate and fusion of the flanking neural folds (Schoenwolf & Smith, 1990; Gilbert, 1997). With the closure of the posterior neuropore (typically at the sacral region), primary neurulation is completed and secondary neural tube formation begins (Criley, 1969; Nievelstein et al. 1993). Criley (1969) noted an overlap zone between primary and secondary neurulation at the level of the posterior neuropore in the chick. During this time, the primary neural tube lies immediately dorsal to the forming secondary tube. Overlap does not occur in mammals; the secondary neural tube forms from tailbud cells clustered at the terminus of the internalized primary tube (Schoenwolf, 1984; Nievelstein et al. 1993). Eventually, in both the chick and mammal, the primary and secondary neurocoels become continuous and a single neural tube extends from head to tail.

The histogenetic potential of the tailbud is presumed to be great. Early studies showed, for example, that the tailbud could give rise to generalized tail structures and, under special circumstances, the neural tube following extirpation and heterotopic grafting (Seevers, 1932; Fox, 1949; Criley, 1969). More recently, Schoenwolf (1977, 1978, 1979) and colleagues (Schoenwolf & DeLongo, 1980) used tritiated-thymidine-labelled grafts and quail-chick chimeras to show that all neural, muscular, vascular and skeletal components of the avian tail arise from presumably undifferentiated mesenchyme. Similar developmental potential has been observed in chick tail buds cultured in vitro (Griffith & Sanders, 1991). These studies have since been revisited using higher resolution techniques and will be revisited in kind later in this paper.

Gene expression distinguishes the tailbud from the head and trunk of the embryo. A growing number of genes are expressed predominantly within the caudal end of vertebrates as the tail develops (see Table 1 for list). I should point out that none of these genes, however, has an anterior expression boundary that corresponds with the level of the anus (i.e. trunk–tail boundary), as strict adherence to Holmdahl's model would predict.

Table 1. Developmental and genetic characteristics of post-anal tail in select chordate groups.

	Mode of caudal:
Chordate taxon	Notogenesis	Neurulation	Somitogenesis	Level of tailbud mosaicism	Implicated genes (references)
Urochordate (various Ciona)	Posterior extension	1°	No somites	Tail from extension of anterior structures; not true tailbud	As-T, bobcat, HrCdx, Manx (1–3)
Cephalochordate (Branchiostoma floridae)	Accretion/posterior extension (?)	1° (?)	TBM; No PSM	Genetically mosaic	AmBra, AmphiCdx, AmphiEvxA, AmphiNotch, AmphiWnt3/5/6/8/11 (1, 4–8)
Lamprey (Lampetra japonica)	Posterior extension (?)	2° (neural keel)	PSM	Histologically mosaic	–
Teleost (Danio rerio)	Posterior extension	2° (neural keel)	PSM	Histologically mosaic	BMP4, caudal, eve1, mfn, ntl, snail, tolloid, Wnt5/8 (9–16)
Amphibian (Xenopus laevis)	Posterior extension	2°	PSM	Histologically mosaic	BMP2/4, derrière, FGF8, Fgfr1, Lfng, Xbra, Xcad2/3, X-delta-1 Xhox3, Xnot2, Xwnt3a/5a/11 (17–21)
Avian (Gallus domesticus; Coturnix coturnix japonica)	Accretion and extension	2°	PSM	Histologically mosaic (VER)	cCDX-B, Ch-T, Ch-Tbx6L, Gnot1 (22–23)
Mammal (Mus musculus; Rattus norvegicus)	Accretion and extension	2°	PSM	Genetically mosaic (VER)	Brachyury, Cdx1/2, Evx1/2, Fgfr1, Gdf11, Lef1, Lfng, Notch1/2, Tbx6, Tcf1, Wnt3a/5a (24–35)

Abbreviations: TBM, tail bud mesoderm; PSM, presomitic mesoderm; VER, ventral ectodermal ridge.

References:

¹Holland et al. (1995);

²Katsuyama et al. (1999);

³Swalla et al. (1999);

⁴Brooke et al. (1998);

⁵Ferrier et al. (2001);

⁶Holland et al. (2001);

⁷Schubert et al. (2001);

⁸Holland (2002);

⁹Joly et al. (1992);

¹⁰Joly et al. (1993);

¹¹Thisse et al. (1993);

¹²Schulte-Merker et al. (1994);

¹³Kelly et al. (1995);

¹⁴Kanki & Ho (1997);

¹⁵Rauch et al. (1997);

¹⁶Connors et al. (1999);

¹⁷Amaya et al. (1991);

¹⁸Beck & Slack (1998);

¹⁹Beck & Slack (1999);

²⁰Sun et al. (1999);

²¹Tada & Smith (2000);

²²Knezevic et al. (1998);

²³Ehrman & Yutzey (2001);

²⁴Bastin & Gruss (1990);

²⁵Herrmann et al. (1990);

²⁶Dollé et al. (1994);

²⁷Takada et al. (1994);

²⁸Gofflot et al. (1997);

²⁹Chapman & Papaioannou (1998);

³⁰Partanen et al. (1998);

³¹Galceran et al. (1999);

³²Gamer et al. (1999);

³³Yamaguchi et al. (1999);

³⁴Catala (2002);

³⁵van den Akker et al. (2002)

Genes: AmBra, As-T, Ch-T, Ch-Tbx6L, Ntl (no tail), Tbx6, Xbra–Brachyury-related and T-box genes (mesoderm specification); AmphiNotch, Gnot1, Notch 1/2, Xnot2 – homologues of Notch (segmentation clock); Lfng (Lunatic fringe), X-delta-1 – encode Notch regulators; AmphiCdx, caudal, cCDX-B, Cdx1/2, HrCdx, Xcad3 – homologues of Drosophila caudal gene (posterior development); AmphiEvxA, eve1, Evx1/2, Xhox3 – homologues of Drosophila even-skipped (posterior development); AmphiWnt3/5/6/8/11, Wnt3a/5a, Wnt5/8, Xwnt3a/5a/11 – homologues of Drosophila wingless (tailbud formation/elongation); Lef1, Tcf1 – encode LEF-1 (Lymphoid enhancer factor-1)/TCF (T cell factor) transcription factors (Wnt signalling); BMP2, BMP4, derrière, Gdf11 – TGF-β genes (mesoderm determination); tolloid, mfn (mini-fin) – encode BMP regulators; FGF8, Fgfr1 – encode Fibroblast growth factors/receptors (posterior development); Manx, bobcat – Ascidian zinc finger genes (tail formation).

Many of the caudally expressed genes also appear to play a crucial role in the development of the tail. Mutants for T (or Brachyury; Herrmann et al. 1990), the mesoderm regulators Wnt3a (Takada et al. 1994; Greco et al. 1996) and Wnt5a (Yamaguchi et al. 1999), related genes Lef1 and Tcf1 (Galceran et al. 1999), various FGF receptors (Amaya et al. 1991; Partanan et al. 1998), the ‘segmentation clock’ genes Notch1 (Conlon et al. 1995) and Lunatic fringe (Evrard et al. 1998; Zhang & Gridley, 1998), and caudal genes (van den Akker et al. 2002) exhibit relatively typical head and trunk regions, but severe posterior disruptions. Thus, morphological, histological, and genetic data provide support for Holmdahl's idea.

The alternative view

Walther Vogt (1926) regarded caudal development in a different light than his contemporary Holmdahl. He emphasized the ‘unity and continuity’ of the vertebrate body plan and proposed that the tail arises from the morphogenetic processes of gastrulation. Vogt's model was further elaborated by the French biologist Jean Pasteels, who posited that the tailbud is not the homogeneous blastema predicted by Holmdahl, but a mosaic of distinct cellular populations having their origins more anteriorly in the embryo (Fig. 1). Specifically, Pasteels (1939, 1942, 1943) identified the ‘charnière chordoneurale’ or chordoneural hinge (CNH), which contributes to the notochord and spinal cord of the tail, and is the direct descendant of the dorsal blastoporal lip. Bijtel (1931) had previously shown that neural and mesodermal tissues were in close apposition at the site of tailbud development.

Pasteels (1943) also showed that cell proliferation and growth rates in the amphibian tail are consistent with those of the rest of the body and not significantly higher as predicted by a secondary inductive model, i.e. development from a blastema. Davis & Kirschner (2000), who followed the development of fluorescently labelled cells in Xenopus tail buds, recently validated this finding. Similarly, typical cell proliferation rates have been noted in the zebrafish (Kanki & Ho, 1997) and chick tail buds (Schoenwolf, 1977; Mills & Bellairs, 1989).

Neither Pasteels nor Bijtel's studies were included in a 1992 review by Griffith and colleagues on tailbud development in vertebrates. Relying largely on the work of Schoenwolf and colleagues along with their own studies of the avian tailbud (Griffith & Sanders, 1991), they concluded that the vertebrate tailbud is indeed the homogeneous mass of undifferentiated cells originally described by Holmdahl (Griffith et al. 1992).

Tailbud development: Xenopus

Vogt's and Pasteels' interpretation, however, found new life in the work of Gont et al. (1993; also reviewed in De Robertis et al. 1994), who followed the expression of the Xenopus transcription factors Xbra (T) and Xnot2 from the dorsal blastoporal lip into two distinct cell populations in the Xenopus tailbud. These populations are the CNH (first identified by Pasteels) and the posterior wall (formed by the fusion of the lateral blastoporal lips). The two are separated by a neurenteric canal, and contribute to distinct structures within the tail, namely the notochord, the ventral spinal cord (both derived from the CNH) and the somites (from the posterior wall).

Gont et al. (1993) went on to show definitively that the tailbud is a direct descendant of the late blastoporal lip and that the tail tip retains organizer activity, i.e. recruitment of cells into an axis. Therefore, the posterior body of Xenopus develops by the continuation of developmental mechanisms beginning at the rostral end. This idea is further supported by Tucker & Slack (1995a, b), who used fate and specification maps to demonstrate that the Xenopus tail does not originate exclusively from the post-anal bud. Paradoxically, trunk cells (located anterior to the anus) also contribute to caudal development. The reassignment of trunk cells to a caudal fate occurs as the future anus is positioned more rostrally during tail extension. Along with the tailbud, these converted trunk cells constitute the Xenopus‘tail-forming region’. Similarity between Tucker and Slack's maps of the region also indicates that the axial components of the tail (neural, notochordal, and myotomal tissue) are embryologically mosaic (i.e. arise from cell populations established by gastrulation).

Slack and colleagues (Tucker & Slack, 1995a, b; Beck & Slack, 1999) have also proposed a model to account for tail formation in Xenopus. According to their ‘NMC model’, the posterior neural plate (including both N and M regions) must interact with underlying dorsal mesoderm (C) to initiate the tailbud. These regions are brought together by gastrulation and appear to be genetically discrete. Gene expression studies by Beck & Slack (1998) and Gawantka et al. (1998) have identified at least 10 discrete domains within the Xenopus tailbud.

Tailbud development: avians

The next significant contribution to the debate came from Catala et al. (1995) in France. They repeated the earlier chick studies of Schoenwolf and colleagues using grafts of selective regions of the tailbud instead of the entire bud. At this higher resolution, the chick tailbud emerges as a heterogeneous accumulation of predominantly mesenchymal cells laid down by gastrulation. The recent advent of single-cell electroporation techniques (Haas et al. 2001, 2002) opens the door to re-examination of cell lineage and development in the tailbud of the chick, Xenopus, and other vertebrates. The unparalleled resolution afforded by this technique may uncover more complicated developmental processes at work.

Catala et al. (1995) also described a structure equivalent to the CNH of Xenopus (Pasteels, 1939, 1942, 1943; Gont et al. 1993). It is derived from Hensen's node and, as in Xenopus, gives rise to the notochord and the ventral spinal cord. Knezevic et al. (1998) confirmed Catala et al.'s findings by tracing the expression of several molecular markers (e.g. T, Gnot1) from the primitive streak into distinct regions within the tailbud. They further showed that gastrulation-like ingressive movements from the surface continue in the early chick tailbud, and that the bud retains organizer activity.

Studies of the chick have also shed light on the relationship between primary and secondary neurulation. Traditionally, the two events have been perceived as occurring by distinct molecular and developmental mechanisms (Schoenwolf, 1979; Schoenwolf & DeLongo, 1980; Griffith et al. 1992; Beck & Slack, 1999; Hall, 2000). Catala (2002) suggested, however, that they might be continuous with one another. Catala et al. (1996) and Le Douarin et al. (1998) demonstrated using quail-chick marking that the medullary cord arises from epithelial cells of the neural plate. It is this same structure that curls up to form the neural folds and, in turn, the primary neural tube within the main body of the embryo. A recent single-cell electroporation study (Mathis et al. 2001) has revealed similar developmental contiguity. A population of neural progenitor cells within the regressing Hensen's node contributes to both primary and secondary neural structures and is maintained by fibroblast growth factor (FGF) signalling (Mathis et al. 2001).

Given these findings, it is reasonable to conclude that caudal development in avians, as in Xenopus, relies largely on the same mechanisms that occur more rostrally in the embryo (Catala et al. 1995, 1996).

Tailbud development: zebrafish

The parallels continue with zebrafish, which have a heterogeneous tailbud, and cell movements consistent with gastrulation occurring at the caudal terminus (Kanki & Ho, 1997). More convincing evidence of rostral–caudal contiguity comes from the work of Kimmel & Warga (1987a, b), which showed that notochord and muscle cell lineages within the trunk continue into the tail.

Gene expression in the tailbud, however, implies that secondary development may also be at work. The genes snail (Thisse et al. 1993), eve1 (Joly et al. 1993) and caudal (Joly et al. 1992) are expressed almost exclusively in the tailbud. Furthermore, silencing the zebrafish T homologue ntl results in an otherwise typical embryo bearing massive caudal defects and no notochord along its entire length (Schulte-Merker et al. 1994). Similar perturbations have been noted in mouse T mutants and the corresponding Xenopus phenocopy (Herrmann et al. 1990; Conlon et al. 1996).

Tailbud development: mammals

The mammalian tailbud, like that of zebrafish, exhibits developmental features congruent with both Holmdahl's and Vogt's ideas. The bud consists of a ball of histologically homogeneous tail bud mesenchyme overlain by a ventral ectodermal ridge (VER) at the caudal tip continuous with a more transient ectodermal groove rostrally (Grüneberg, 1956; Gajovic & Kostovic-Knezevic, 1995). Grüneberg (1956) was the first to point out the histological similarity between the tailbud and embryonic limb bud, and, by extension, the possibility that secondary inductive events (as seen in the limb; reviewed in Capdevila & Izpisua Belmonte, 2001) were involved in tail formation.

There is growing evidence to suggest that this is indeed the case. Grüneberg & des Wickramaratne (1974) described a mouse mutant (vestigial tail) bearing an undeveloped tailbud with only traces of an ectodermal ridge. This phenotype is copied by mutants for the genes T (Herrmann et al. 1990) and Wnt3a (Takada et al. 1994; Greco et al. 1996). More recently, Hall (2000) and Goldman et al. (2000) showed that ablating the VER results in tailbud regression and prevents somitogenesis and chondrogenesis in the tail.

The VER is also a genetically discrete structure, expressing specifically Msx1, Wnt5a, BMP2 and FGF17 (Lyons et al. 1992; Gofflot et al. 1997, 1998; Goldman et al. 2000). Although inductive functions have yet to be determined for each of these genes, their localized expression suggests that the VER, like the AER, is a source of molecular signals that regulate underlying bud mesenchyme. The VER, however, does not appear to employ extensive FGF signalling (as seen in the AER) and, accordingly, cannot regulate development when grafted onto limb bud explants (Goldman et al. 2000).

It is unlikely, however, that secondary development is the sole mechanism of tail formation in mammals. Several mapping studies have demonstrated that tailbud materials originate from the primitive streak and Hensen's node, suggesting that rostral developmental mechanisms may continue into the caudal portion of the embryo (Tam & Tan, 1992; Wilson & Beddington, 1996). Gofflot et al. (1997) provided more convincing evidence of this. They followed the expression of several gastrulation marker genes (Wnt5a, T, Hoxb1 and others) from the primitive streak and Hensen's node into distinct regions of the tailbud.

Evolutionary considerations

The developing tails of lampreys, zebrafish, frogs, chicks and mice are not homogeneous masses of cells, but mosaics of either genetically, developmentally or histologically distinct cell populations (Table 1). Therefore, it can be reasonably inferred that the common ancestor of all vertebrates had a heterogeneous tailbud. The histological homogeneity of the mammalian tailbud and the implicated secondary inductive interactions (involving the VER) undoubtedly arose secondarily given the derived nature of the mammals. The presence of a VER-like structure (for which a function has yet to be defined) in the chick tailbud suggests that these traits may at least be common to amniotes (Knezevic et al. 1998). Furthermore, the persistence of gastrulation caudally, as seen in Xenopus, birds, zebrafish and perhaps mice, suggests that their earliest common ancestor initiated its tailbud and formed its tail in the same manner (as opposed to secondary development).

Schubert et al. (2001) recently arrived at a similar conclusion upon studying the expression of three Wnt genes in amphioxus, the closest living proxy to the ancestral vertebrate. They concluded that the tailbud of this cephalochordate and, by extension, the presumed vertebrate ancestor, arises directly from the dorsal blastoporal lip and comprises the CNH and the posterior wall. These same cell populations have been described in the lamprey, zebrafish, frog and chick (Nakao & Ishizawa, 1984; Griffith et al. 1992; Gont et al. 1993; Catala et al. 1995; Knezevic et al. 1998; Dheen et al. 1999). In each group, the posterior wall typically contributes to the roof of the neural tube, while the hinge contributes to the floorplate of the neural tube and the notochord (Schubert et al. 2001).

The manner by which the CNH contributes to the elongation of these structures is unclear for amphioxus, but better understood for vertebrates (Table 1). Two mechanisms are thought to be at work (either singly or together). Accretion, the gradual addition of new cells (from involution) to the caudal end of the notochord and the neural floorplate, is the primary driving force behind tail elongation in birds (i.e. chick and quail) and the mouse (Sausedo & Schoenwolf, 1993, 1994). The alternative mechanism, posterior extension, drives tail outgrowth by the lengthening of structures (due to cell division, intercalation or elongation) rostral to the tailbud. This mechanism acts in concert with accretion in birds and mammals, but by itself in more basal vertebrate classes. Urochordates (sea squirts) have taken posterior extension to the extreme as they rely exclusively on the rearrangement of pre-existing cell populations for tail elongation (Satoh, 1994; Katsuyama et al. 1999). For this reason, urochordates are not considered to have a true tailbud, and I have deliberately not dealt with them in this paper.

Gont et al. (1993) showed that the tail of the frog Xenopus elongates by the intercalation of cells of the neural plate and notochord. Accretion is ruled out as involution movements terminate at the early neurula stage. Low cell proliferation rates at the tail tip of the zebrafish (Kanki & Ho, 1997) and in the tailbud of the lamprey (Nakao & Ishizawa, 1984) similarly rule out accretion in these species. In contrast, Schubert et al. (2001) favour an accretive origin for the neural tube floorplate and notochord in amphioxus, though they do not rule out posterior extension.

Lastly, Schubert et al. (2001) suggest that the ancestral vertebrate tailbud would have formed somites (segmented, paraxial mesoderm that gives rise to musculature and vertebrae) directly from bud mesoderm. In vertebrates, however, somitogenesis occurs indirectly through mesenchymatous presomitic mesoderm lying between the tailbud and recently formed somites (Table 1). This band of mesenchyme probably arose early in the course of vertebrate evolution.

Evolution of primary and secondary neurulation

As a relevant aside, I will now consider the evolutionary and embryological relationships of primary and secondary neurulation, and their implication to the larger question at hand.

I suggest here that primary neurulation is the ancestral vertebrate condition. It is arguably the predominant mechanism of neural tube formation in vertebrates, given its presence in all studied orders (Fig. 2; Kingsbury, 1932; Balinsky, 1981; Nakao & Ishizawa, 1984; Gilbert, 1997). What is more, primary neurulation has been described in urochordates (Nicol & Meinertzhagen, 1988), the basal chordate group, and likened to neural tube formation in amphioxus (Hatschek, 1893; Balinsky, 1981; Schubert et al. 2001). In this diminutive chordate, a thickened neural plate folds up to form a tube. Neural tube formation follows overgrowth of the epidermal ectoderm in amphioxus, whereas the two processes occur concurrently in typical primary neurulation (Holland et al. 1996; Schubert et al. 2001). Holland et al. (1996) have suggested that the uncoupling of the neuro- and epidermal ectoderm in the vertebrate ancestor may have opened the door to the invention of the neural crest (NC). The observation of NC-like cells that border the neural plate in amphioxus and express such NC marker genes as slug, snail, AmphiFoxD, Distal-less, Pax3/7, and Zic supports this idea (Holland et al. 1996; papers cited in Yu et al. 2002).

Fig. 2.

A phylogeny of the chordates (based on Zardoya & Meyer, 1996; Mallatt & Sullivan, 1998; Cotton & Page, 2002) with data for primary and secondary neurulation (see text for cited literature). Primary neurulation, as seen in ascidians and the majority of vertebrate taxa, probably represents the ancestral chordate condition. Neural plate folding in amphioxus, a process consistent with primary neurulation, has been uncoupled from epidermal overgrowth. This event is thought to be related to the advent of the neural crest in craniates (Holland et al. 1996). Secondary neurulation, the derived condition in vertebrates, is presumed to have arisen by parallel substitution events in lampreys, neopterygian fishes, dipnoans (Lepidosiren and Protopterus with the exception of Ceratodus), and possibly tetrapods. Taxa for which data are missing are indicated in grey. Branches marked with asterisks are unresolved and have been represented as trichotomies.

Secondary neurulation, widely considered to be a vertebrate novelty, may have arisen by parallel substitution in lampreys, neopterygian fishes (ganoid and teleost fish), dipnoans and tetrapods (Fig. 2). Admittedly, this scheme is not parsimonious; however, it is certainly favourable to the alternative scenario by which secondary neurulation is ancestral and has been either partially or entirely forsaken by one or more members of all major vertebrate orders (Fig. 2). Furthermore, as I will argue below, the leap from primary to secondary neurulation may not have been a big one for vertebrates.

A similar morphogenetic mechanism appears to be at work in secondary neural tube formation in the lamprey, zebrafish (and other teleosts) and the chick. In each, the medullary cord (or neural keel) arises from the epithelial neural plate and not from a condensation of mesenchyme as seen in mammals (Kingsbury, 1932; Piavis, 1971; Schoenwolf, 1984; Reichenbach et al. 1990; Schmitz et al. 1993; Papan & Campos-Ortega, 1994; Catala, 2002). The common embryological origin of secondary and primary neural tubes from the neural plate implies that the two processes may not be diametrically opposed as traditionally thought. Indeed, secondary neurulation may be a derived form of primary neurulation.

Kingsbury, a contemporary of Holmdahl and Vogt, also highlighted the overlap of primary and secondary neurulation. He suggested that the formation of the solid neural keel in teleosts is a ‘masked folding’ event, wherein the folding of the neural plate and formation of the neurocoel is obscured by the massive crowding of material at the dorsal surface of the embryo (Kingsbury, 1932). Reichenbach et al. (1990) later described neurulation in a teleostean embryo proceeding by the tight folding of the neural plate. Papan & Campos-Ortega (1994) confirmed the epithelial genesis of the keel and showed that the location of cells within the zebrafish neural plate corresponds to their ultimate location in the completed neural tube (as is the case for primary neurulation).

They also demonstrated that the number of primordial neural cells at the anterior neural plate increases by nearly three-fold during neurulation – twice that seen in the frog Xenopus laevis. This lends credence to Kingsbury's ‘masked folding’ hypothesis and also to his suggestion that neurulation in amphibians (urodeles and frogs) represents an intermediate between primary and secondary neural tube formation. Upon fold fusion in frogs and the salamander Ablystoma, the neural tube is solid except for a tiny neurocoel at the ventral floorplate (Kingsbury, 1932). The neurocoel eventually expands dorsally by a process called relumination, which is driven by the medially directed migration and radial intercalation of cells in the neural tube (Davidson & Keller, 1999). This process greatly resembles cavitation in teleosts and appears to be reversed in secondary neurulation in birds and mammals, with lumen expansion proceeding from roofplate to floorplate (Davidson & Keller, 1999). This implies that crowding of materials at the neural plate may also occur in amphibians. Harris & Hartenstein (1991) provided compelling support for this by demonstrating that neurulation in Xenopus can proceed normally even in the presence of an antimitotic agent.

If secondary neurulation is a derived form of primary neurulation, then this could potentially account for the extraordinary capacity of vertebrates to exchange or couple these divergent developmental mechanisms (Fig. 2) in creating a dorsal, hollow neural tube.

There is reason to believe that secondary neurulation may be more specialized in some vertebrates than in others. Caudal neural tube formation in mammals, for example, is consistently described as occurring from a condensation of mesenchymal cells, which subsequently undergo a transition into epithelium (Schoenwolf, 1984; Müller & O'Rahilly, 1987; Nievelstein et al. 1993). Developmental uncoupling of rostral and caudal neurulation has also been demonstrated in the frog. Beck & Slack (1999) produced secondary tails containing neural tube-like structures by grafting animal pole explants (from blastulae) containing active Notch in the posterior neural plate. This is strong support for the autonomy of caudal neurulation in Xenopus.

Concordia discors

Caudal development in vertebrates appears to be both continuous with axial development in the head and trunk yet distinct from it. Mapping and in situ hybridization studies have shown that developmental processes and expression of various genes (e.g. T) are often continuous between the two ends of the embryo. In contrast, the seeming independence of secondary neurulation from primary neurulation in mammals, Xenopus and possibly other vertebrates, and the localized occurrence of many genes and phenotypic disruptions in the caudal end imply just the opposite. Therefore, the vertebrate tailbud can be thought of as Horace's concordia discors (Epistles, see Ferry, 2001), a harmony of two discordant developmental processes.

The distinction between continuous rostro-caudal and secondary development in the tail has been further obscured by the recent work of Davis & Kirschner (2000). They marked small neighbourhoods of cells (≥ 3) in the Xenopus tailbud using photoactivation of fluorescence and noted that interspersed with specified cells (laid down by gastrulation) there are mesenchymal cells capable of differentiating into a variety of tissues (e.g. neural tube, notochord, somitic mesoderm). Despite their ‘multipotentiality’, these cells typically differentiate in line with adjacent cells, giving a very clear-cut fate map at the tissue level. Thus, the Xenopus tailbud cannot be the rigid mosaic originally proposed by Vogt and Pasteels nor the homogeneous blastema described by Holmdahl.

I suspect that similar developmental ambiguity – perhaps better described as ‘discordant harmony’ – will be revealed in other vertebrates with the application of high-resolution labelling techniques (as per Davis & Kirschner, 2000; Mathis et al. 2001).

Genetic basis of caudal development and defects

As efforts continue to resolve the nature of the tailbud, we are accumulating a wealth of data on the genes and molecular pathways that contribute to caudal development. Not surprisingly, the same genes appear to be involved in tailbud outgrowth in all of the vertebrates studied. Some of the key players include the homologues of T, caudal (e.g. cdx1/2) and even-skipped (e.g. Evx1), the Wnt genes Wnt3a and Wnt5a, various Notch genes, their ligands (e.g. Delta), and related genes (e.g. Lunatic fringe) (see Table 1).

A task for medical researchers is to elucidate the remaining genes and their signalling pathways, identify the homologues in humans and determine how they contribute to developmental defects. It is worth clarifying here that while the human tail is present only in early ontogeny, its formation and subsequent regression (by cell death) are critical events and, when awry, can result in a variety of caudal defects and neural tube defects (NTDs; Müller & O'Rahilly, 1987; Nievelstein et al. 1993; van Straaten & Copp, 2001; Catala, 2002).

Indeed, Holmdahl (1925a, b) saw the implications of his work for NTDs, specifically spina bifida, a failure of the spinal neural tube to close (occurring predominantly at the caudal terminus). Similar developmental malformations are seen in the curly tail mouse mutant, which was originally described by Hans Grüneberg (1954). This mutant has proved to be a valuable model for studying the genetic and cellular basis of spina bifida and other NTDs in humans (reviewed in van Straaten & Copp, 2001; Hall et al. 2001). The genetic basis of another mouse mutant bearing an NTD, Loop-tail, has just recently been elucidated (Murdoch et al. 2001). Sacral agenesis, an axial defect characterized by missing posterior vertebrae, has been linked to mutations for the gene HLXB9 (Catala, 2002). Furthermore, the Notch genes and their ligands (Delta, Jagged-1/2) are involved in a host of human defects, including CADASIL, Alagille Syndrome and spondylocostal dysostosis (SCDO) (Oda et al. 1997; Bulman et al. 2000; Joutel et al. 2000).

Clearly, progress has been made, and continued investigation of vertebrate caudal development and the associated molecular pathways is justified.