The evolution of vertebrate gastrulation

Evolutionary and embryological studies have been intertwined throughout their history. The common elements of body plans can be visualized best at mid embryogenesis, the phylotypic stage, as noted initially for the vertebrates by von Baer (1828; Gould, 1977). At an earlier stage, gastrulation, the process by which morphogenetic movements of cell layers produce an embryo consisting of endoderm, mesoderm and ectoderm, appears to be very different in the various vertebrate classes.

In the zebrafish the most striking gastrulation movement is epiboly, in which cells of the embryo proper envelop a spherical yolk mass. In Xenopus, which has holoblastic cleavage, the endomesoderm invaginates through a circular blastopore. In the chick the main gastrulation movement is the invagination of the mesoderm and endoderm through a linear primitive streak. In the mouse the gastrula has the shape of a cup, whereas in the human, gastrulation takes place in a flat epithelial epiblast layer and is very similar to that of chick. Mammalian gastrulation is thought to have evolved from ovo viviparous ancestors that lost the yolk after nutrients could be obtained from the mother (Kerr, 1919), thus explaining the similarities observed among the amniotes.

Because the outcome of gastrulation in all the animals is a very similar body plan at mid embryogenesis, it is reasonable to assume that the underlying mechanisms are similar in all vertebrates. The availability of molecular markers now permits the identification of the common elements in vertebrate gas trulation.

In this paper we will compare the distribution of two gene markers, goosecoid and Brachyury, in Xenopus, zebrafish, chick and mouse embryos, which shed light on the evolution of the organizer and of mesoderm precursor cells. In addition we will examine the mechanism of development of the postanal tail, which arises as a continuation of gastrulation in Xenopus, and perhaps in other vertebrates as well.

An experiment that has had a permanent influence in the way we think about embryos was carried out by Spemann and Mangold in 1924. When the region at which involution starts in the amphibian gastrula, the dorsal lip of the blastopore, was transplanted to the opposite (ventral) side of a host embryo, a twinned embryo resulted. Because the transplanted tissue recruited cells from the host into the secondary axis, this region was called the organizer (Spemann, 1938; Hamburger, 1988). The organizer was able to induce neural tissue and segmented mesoderm such as somites and pronephros, while the graft itself differentiates predominantly into prechordal plate, notochord and part of the tailbud (Hamburger, 1988; Gont et al., 1993).

One of the recent advances has been the isolation of genes that mark the organizer. These include the genes for transcrip tion factors such as goosecoid (Blumberg et al., 1991; Cho et al., 1991), Xlim-1 (Taira et al., 1992), FKHD-1 (Dirksen and Jamrich, 1992) and Xnot (von Dassow et al., 1993; Gont et al., 1993) as well as those for the secreted protein noggin (Smith and Harland, 1992) and the membrane bound molecule integrin a3 (Whittaker and DeSimone, 1993).

The homeobox-containing gene goosecoid has been studied extensively and its expression pattern in Xenopus, zebrafish, chick and mouse is reviewed below. From a functional point of view, ectopic expression of goosecoid in ventral blastomeres of Xenopus mimics many of the functions of organizer grafts, such as formation of secondary axes (Cho et al., 1991; Stein beisser et al., 1993), recruitment of neighboring uninjected cells into new cell fates, and triggering of cell migration in the dorsoanterior direction (Niehrs et al., 1993). It also causes dose-dependent formation of dorsal tissues (Niehrs et al., 1994). While goosecoid has been studied most intensively, the other dorsal markers are also expected to play roles in the organizer phenomenon. In the case of noggin it has been shown that its ectopic expression produces complete axial rescue of Xenopus embryos ventralized by UV light (Smith and Harland, 1992). The analysis of many of the other genes is in progress, particularly by targeted inactivation of their mouse homo logues, and one can expect that much information will be forth coming. In this paper we concentrate on goosecoid, which is a marker of the head organizer region.

The Brachyury gene has also been intensively studied (reviewed by Beddington and Smith, 1993). Named after its short tailed phenotype in heterozygous mice, it was one of the first embryonic lethal mutations to be identified (Dobrovol skaia-Zavadskaia, 1927; Gluecksohn-Schoenheimer, 1938). The expression pattern of Brachyu,y has been studied in mouse (Herrmann, 1991), Xenopus (Smith et al., 1991) and zebrafish (Schulte-Merker et al., 1992). Brachyury provides a marker for both ventral and dorsal mesoderm. At mid gastrula Brachywy marks the notochord and uncommitted mesoderm and, together with goosecoid, which marks the prechordal plate at this stage, permits one to identify the main territories of the gastrula. At later stages Brachyury is expressed in the tailbud and, together with Xnot-2 (Gont et al., 1993), permits the analysis of the developing tail.

goosecoid mRNA is present at low levels in the unfertilized egg (De Robertis et al., l 992; Lemaire and Gurdon, 1994), but shortly after mid blastula the levels of its transcripts increase, reaching a maximum just before the start of gastrulation at stage 10. At this stage goosecoid mRNA is localized to the dorsal side of the marginal zone, forming agradient over tissue lhat at this stage is already fated to form dorsal mesoderm (Fig. I A,B). By culturing fragments of marginal zone (Dale and Slack, 1987) it has been shown that dorsal mesodermal tissues (notochord and muscle) are formed by the organizer (dorsal lip) region, that intermediate regions give rise to pronephros and muscle, and that the rest of the marginal zone produces ventral mesoderm (mesothelium, mesenchyme and blood), as indicated in Fig. I A. At later stages the more lateral regions of lhe marginal zone become induced to become muscle under the influence of a ‘horizontal’ signal emanating from the organizer (reviewed by Slack, 1991), but at the early gastrula only the goosecoid-expressing region is specified to form dorsal mesoderm. This early gradient of expression is thought to be involved in mesodermal patterning, because two-fold increments in the amounts of microin jected synthetic goosecoid mRNA can trigger lhe expression of increasingly dorsal molecular markers, forming several sharp thresholds in explanted ventral marginal zones (Niehrs et al., 1994).

Once gastrulation starts, goosecoid is down-regulated, and the cells that continue to express it leave the dorsal lip with the involuting endomesoderm (Fig.2A), eventually becoming restricted to the small group of cells that form the chordal plate or head mesoderm (Stein beisser and De Robertis, 1993), as shown in Fig. 2B. Not all genes expressed in the dorsal lip follow thispattern.For example, the homeobox gene Xnot-2 is also expressed dorsally, but its expression stays localized in the dorsal lip, and at neurula stages is expressed in the notochord and eventually becomes restricted to the tail of the embryo (Fig. 2C,D).

Al the early gastrula stage Xenopus Brachyu,y (Xbra) is expressed in a band or ring encompassing lhe entire marginal zone (Fig. 2E). In other words, Xbra is expressed in the progenitors of both dorsal and ventral mesoderm (Smith et al., 1991). Once cells involute through lhe circular blastopore, they lose Xbra expression (except in the notochord which is tran siently positive), so that at later stages the only region expressing Xbra is the tailbud, as shown in Fig. 2F. Two regions of the tailbud express Brachyury, one of dorsal origin (lhe chordoneural hinge, which al o expresses Xnot-2) and the other deriving from the lateral marginal zone (the posterior wall). These two populations are separated by the neurenteric canal (indicated by an arrowhead in Fig. 2F). The role of these two regions in late gas trulation (Gont et al., 1993) will be discussed below.

We conclude that in the early Xenopus gastrula goosecoid marks most cells specified to become dorsal mesoderm, including notochord and muscle, but that as gastrulation proceedsonly those cells forming the prechordal plate continue to express goosecoid. Xbra i expressed in all mesoderm prog enitors at early gastrula, but al later stages becomes restricted to posterior notochord and the tailbud.

In zebrafish goosecoid expression starts at the blastula stage, where it forms a striking dorsoventral gradient (Schulte-Merker et al., 1994), as shown in Fig. lC,D. These cells then cover the yolk through the movement of epiboly, which is accompanied by convergence towards the dorsal midline (forming the ‘embryonic shield’ on the dorsal side). Zebrafish goosecoid transcripts are maximal at 50% epiboly (Stachel et al., 1993; Schulte-Merker et al., 1994), the stage at which invo lution of the endomesoderm starts; expression is then down regulated and follows the involuting cells in the anterior-most regions. The notochord is negative at all times, and by the end of gastrulation expression can be seen in the endomesoderm of the prechordal plate (Schulte-Merker et al., 1994), as shown in Fig. IE. Patches of expression can be seen in the CNS and branchial arches at later stages (Schulte-Merker et al., 1994).

Zebrafish Brachyury (also called notail due to the phenotype of its mutant) has been studied in considerable detail, as an excellent antibody is available (Schulte-Merker et al., 1992). Expression starts on the dorsal side and forms a ring along the entire marginal zone, as in Xenopus. Interestingly, double staining experiments using goosecoid RNA probes and Brachyury antibodies have shown that both genes are expressed in the same cells of the dorsal side before involution starts. As gastrulation proceeds, the goosecoid signal follows the prechordal plate, which becomes negative for Brachyury. Brachyury is found in the notochord and tailbud (Schulte Merker et al., 1992, 1994). Although both genes are expressed initially in the same cells, in Brachyury (notail) mutants the expression of goosecoid is normal (Schulte-Merker et al., 1994).

We conclude that despite the predominance of epiboly movements in zebrafish gastrulation, the expression of goosecoid is not unlike that of its Xenopus homologue: it starts as a dorso-ventral gradient, has maximal expression at the start of involution of the endomesoderm and is then down regulated, following the prechordal plate.

The chick embryo offers many advantages to the experimen talist, one of which is that the embryo develops from a flat epithelial sheet, the epiblast, from which all germ layers derive. This facilitates the study of pre-gastrulation stages. Chick goosecoid (Izpisua-Belmonte et al., 1993) is first detectable in the unincubated egg, which as it is laid by the hen already contains several thousand cells. Expression starts in a thicken ing of the posterior marginal zone called Koller’ s sickle, where it is confined to a group of cells located in a middle layer of cells between epiblast and the forming hypoblast. The existence of this cell population had been overlooked, although Koller’ s “sickle had been noted almost 100 years ago, but was revealed by the goosecoid marker. It is a very interesting group of cells, for lineage tracing with the hydrophobic dye DiI showed that they migrate anteriorly to the goosecoid-express ing region of Hensen’s node, indicating that this group of cells constitutes the earliest mesendodermal cells of the chick gastrula (Izpisua-Belmonte et al., 1993). The Koller sickle region can also induce secondary axes when grafted. The main conclusion from these studies was that in the chick gastrula tion starts much earlier than previously thought, in the unin cubated egg (Izpisua-Belmonte et al., 1993). Fig. IF shows goosecoid expression in Koller’ s sickle in an embryo that has been incubated for a few hours (stage XII, when the forming sheet of hypoblast covers 50% of the area pellucida).

As development continues, the primitive streak forms in the posterior end. As the extension of the streak progresses in the anterior direction, goosecoid is expressed at the tip of the primitive streak. Transcripts become more abundant as the streak elongates, and by stage 3+, when the streak develops a groove but has not yet reached its maximal extension, goosecoid reaches its maximal expression (Fig. lG). Tran scripts are maximal anteriorly and decrease gradually, like the tail of a comet, towards the posterior primitive streak (Izpisua Belmonte et al., 1993). This stage may be considered homol ogous to stage 10 (beginning of dorsal lip formation) in Xenopus (Fig. lB). This is also the stage in which the anterior end of the primitive streak, the young Hensen’s node, has its maximal inductive activity. Hensen’s node is classically con sidered to be the equivalent of the amphibian organizer (Waddington, 1933). Once the primitive streak reaches its full extension at stage 4, goosecoid is down-regulated, and this correlates with a decrease in inducing activity by the node (Storey et al., 1992). By stage 4+, goosecoid RNA is found in the cells that fan out from the node ingressing to form the endomesoderm of the head, as shown in Fig. lH. At later stages (not shown) goosecoid expression is found in foregut endoderm, prechordal plate, and ventral diencephalon. The expression pattern of Brachyury in the chick has not yet been reported, but from what is known from Xenopus and mouse one would expect this gene to be active throughout the length of the primitive streak, where the forming mesoderm is located.

We conclude that while two of the regions in which chick goosecoid is found, Hensen’s node and prechordal plate, have clear homologues in Xenopus and zebrafish, the earliest one, Koller’ s sickle, is more difficult to fit into the picture of a gen eralized gastrulation mechanism.

The mouse gastrula develops from a cup-shaped epiblast, the egg cylinder. Expression is first detected at the posterior end of the egg cylinder at 6.4 days after fertilization. It is seen as a spot on the side of the egg cylinder on the initial group of cells that leave the epiblast in the initial epithelial-mesenchy mal transition that starts formation of the primitive streak (Blum et al., 1992). This early expression may be homologous to that found in Koller’ s sickle in the chick. Once the primitive streak is formed, goosecoid expression is found at its anterior end (Fig. ll,J). Expression is maximal at day 6.7, also at the anterior of the streak. This early phase of expression is brief, lasting for about 10 hours. Transcripts were not detectable in the prechordal plate or other regions. The whole-mount in situ hybridization signal of mouse goosecoid is less intense than in other organisms, perhaps explaining why the signal is lost. After day 10.5 mouse goosecoid has a late phase of expression that has been described in some detail (Gaunt et al., 1993). goosecoid transcripts are found in limb buds (in the proximal, ventral and anterior region), neural crest of pharyngeal arch 1 (mandibular and maxillary processes) and anterior third of arch 2 (hyoid), the floor of the diencephalon, and other sites (Gaunt et al., 1993). The late expression pattern of goosecoid might be similar in other organisms, but systematic studies have not been carried out.

The studies in the mouse suggest that the anterior primitive streak at day 6.7, when goosecoid expression is maximal, should be homologous to the early dorsal lip of the Xenopus early gastrula. Mouse embryo fragments containing this region have inducing activity when transplanted into Xenopus embryos by the Einsteck procedure (Blum et al., 1992). The mouse node, which at 7.5 days can induce trunk structures when grafted to mouse embryos (Beddington, 1994) would be the equivalent to the late (regressing) Hensen’s node of the chick or the late dorsal lip in amphibians, which have trunk tail organizer activities but do not express goosecoid.

The Brachyury gene was initially isolated from the mouse, and it has been studied in detail. It is expressed first in the entire primitive steak and then in the notochord and tailbud (Herrmann, 1991; Beddington and Smith, 1993). In mouse, expression of Brachyury in the notochord is more persistent than in Xenopus.

It has been observed recently that the formation of the postanal tail results from a continuation of gastrulation in Xenopus and is induced by cells originating from the late dorsal lip (Gont et al., 1993). These findings will be summarized here and then discussed in the context of the evolution of vertebrate gastru lation. Three lines of evidence support the view that tail formation is a continuation of gastrulation: (1) studies with the gene markers Xnot-2 and Brachyury, (2) lineage tracing studies and (3) tail-organizer transplantations (Gont et al., 1993).

The gene marker Xnot-2 is found first in the dorsal lip (Fig. 2C) and then in a part of the tailbud that forms a hinge between the notochord and floor plate (Fig. 2D). Brachyury is expressed as a ring in the marginal zone of the early gastrula (Fig. 2E) but is also found in the tailbud at late stages, although not exclusively in the hinge region, but also in cells located more posteriorly (Fig. 2F). The identification of these two cell pop ulations, called the chordoneural hinge and the posterior wall cells, was our first indication that the tailbud is indeed hetero geneous. Both cell populations are separated by the neurenteric canal. Fifty years earlier Pasteels had reached similar conclu sions based on careful examination of histological sections (Pasteels, 1943) but we were not aware of this at the time.

Lineage tracing of the late blastopore lip revealed that the chordoneural hinge is derived from the organizer. As shown in Fig. 3A, after the goosecoid-expressing cells that will form the prechordal plate leave the organizer, the predominant movement at mid-gastrula is that of involution (driven by con vergence and extension), which leads to the formation of the trunk notochord and somites. At the early neurula (stage 13), the ectoderm and mesoderm attach to each other and the two layers undergo posterior extension movements instead of invo lution (Fig. 3B). It is these posterior movements of the organizer region (driven by continuing cell intercalations in the neural plate and notochord) that cause tail formation. When marks of Dil were placed in the closing blastopore at the slit stage (Fig. 4) and the embryos were left to develop, it was observed that: (1) the dorsal lip becomes the chordoneural hinge and gives rise to tail notochord and ventral spinal chord, (2) the lateral lip gives rise mostly to posterior wall and the tail somites that derive from it and, (3) the ventral lip gives rise to lateral plate mesoderm as well as tissue spanning the region between the anus and the tailbud. The neurenteric canal, which connects the gut and the spinal cord cavity, is formed by the fusion of the lateral lips and the closure of the neural plate so that the lateral lips become the posterior wall (Fig. 4; Gont et al., 1993). As shown in Fig. 5, at tailbud stages the neurenteric canal separates the chordoneural hinge (which expresses Xnot-2 and Brachyury) from the posterior wall (which expresses only Brachyury).

Transplantation of the chordoneural hinge into host gastrulae by the Einsteck procedure showed that it has potent organizing activity, inducing tails with structures such as somites, spinal cord and fins, while the graft itself gives rise mostly to chor doneural hinge and notochord (Gont et al., 1993).

The main conclusion is that the organizer region retains its activity at tailbud stages and occupies a distinct region of the tailbud. This has implications on the way we think about Xenopus gastrulation. For example, the ‘horizontal’ signal emanating from the organizer that induces somite formation (Slack, 1991) might continue well into tailbud stages. The source of inductive signals would be the chordoneural hinge, which as it develops could recruit uncommitted cells from the Brachyury-positive posterior wall into the myotome pathway. In addition, the experiments in Xenopus tell us that the gastru lation process continues much longer than previously thought in the tail region (Gont et al., 1993). We will discuss below whether other vertebrates might share a similar mechanism of late gastrulation.

Despite their considerable anatomical differences, the common elements of gastrulation in the various vertebrate classes are highlighted by analyzing the expression of two markers, goosecoid and Brachyury. The organizer is marked by goosecoid. The period of highest goosecoid expression corre sponds to that of maximal organizer activity in Xenopus and chick, the two organisms in which detailed transplantation analyses have been possible.

The future mesoderm, both dorsal and ventral, is marked by Brachyury. In Xenopus and zebrafish, this can be seen as a ring encircling the spherical embryo, the marginal zone. In the mouse embryo, it is expressed in the entire primitive streak. Thus, the marginal zone ring has become flattened, forming two parallel accumulations of cells in the primitive streak in the amniote embryo. At the anterior end of the early primitive streak lies the organizer, which expresses goosecoid. We conclude that the primitive streak is homologous to the cir cumblastoporal marginal zone present in Xenopus and zebrafish.

It might be worthwhile to attempt to identify the equivalent stages of gastrulation from the expression patterns. The maximal expression of goosecoid correlates with maximal inducing activity, tested by grafting, in Xenopus and chick. These stages may therefore be considered truly homologous. Furthermore, maximal expression of goosecoid is followed by the migration of cells that form the endomesoderm of the head in Xenopus, zebrafish, chick and mouse. These anterior cells continue to express goosecoid, while the gene is down regulated in other cells that will become a number of other dorsal tissues. The region of maximal expression corresponds to the dorsal marginal zone in Xenopus, the embryonic shield in zebrafish, the early Hensen’s node in chick and the anterior primitive streak in the mouse. Maximal expression of goosecoid is achieved just before dorsal lip formation in Xenopus (stage 10, Fig. IB), before the primitive streak is fully extended in the chick (stage 3+, Fig. 1G), at 50% epiboly in zebrafish (Stachel et al., 1993; Schulte-Merker et al., 1994), and at day 6.7 in the mouse (Blum et al., 1992). We propose that these stages are equivalent in the vertebrate gastrula.

Xenopus embryologists usually consider the beginning of gastrulation as the start of involution of cells through the dorsal lip at stage 10. However, if we consider this stage to be homol ogous to the chick embryo as it reaches full extension of the primitive streak (stage 3+ to 4), it is evident that in the chick gastrulation starts much earlier. The extensive morphogenesis involved in the formation of the primitive streak, starting from Koller’ s sickle, is being ignored in most current Xenopus work. However, there are indications that also in Xenopus, pregastru lation movements are much more extensive than commonly thought (Hausen and Riebsell, 1991; Bauer et al., 1994). The challenge for the future lies in discovering whether homologous events exist at the very early stages of vertebrate gastrulation.

In Xenopus inducing activity is first found in dorsal-vegetal blastomeres, in a region designated the Nieuwkoop center (Gerhart et al., 1989). As early as the 32-to 64-cell stage the vegetal cells are thought to emit a signal that induces organizer activity in the overlying marginal zone. Nieuwkoop center cells do not themselves participate in the axis, but remain as yolky endodermal cells (Smith and Harland, 1991). The next step for those working in zebrafish, chick or mouse would be to identify the source of the equivalent Nieuwkoop center signal in these animals, if one exists. In zebrafish we may assume that the early gradient of goosecoid transcripts that is observed at blastoderm (Fig. 1C,D) reflects the intensity of the inductive signal from the Nieuwkoop center equivalent. This signal could emanate for example from the yolk region.

In chick, the goosecoid-positive cells of Koller’s sickle have been shown by fate mapping to contribute to the main body axis, in particular Hensen’s node and its derivatives (Izpisua Belmonte et al., 1993). Therefore these cells are not homolo gous to the Nieuwkoop center, whose cells can induce but do not themselves form part of the axis. If a Nieuwkoop center exists in the chick, one would venture that it should signal before the chick egg is laid and that it would be located in the posterior marginal zone, in the vicinity of the region where Koller’ s sickle is formed. Another possibility is that Koller sickle cells may not require a separate localized inducing center.

In Xenopus there is recent evidence suggesting that the redis tribution of egg cytoplasmic determinants may be all that is required for the activation of goosecoid. Lemaire and Gurdon (1994) marked eggs on the dorsal or ventral side with vital dyes and then cultured dissociated blastomeres. Even in the absence of any cell-cell interactions, goosecoid was turned on in dorsal blastomeres and Xwnt-8, a ventral marker, was activated in ventral blastomeres. One possibility worth exploring is that perhaps the inductive activity found in vegetal-dorsal blas tomeres (Nieuwkoop center) is just a manifestation of a diffuse organizer present from the earliest stages in the dorsal side of the Xenopus embryo. Perhaps the vegetal cells do not partici pate in the axis proper because they are too bulky to migrate, while smaller more dorsal cells might be able to migrate and still have similar inductive properties. While the possibilities mentioned in the last two paragraphs are entirely hypothetical, they were raised here to illustrate how a comparative approach to the Nieuwkoop center/Koller’s sickle problem helps us think about how the vertebrate embryo is formed.

Unlike the studies on goosecoid and Brachyury, the experi ments reviewed here on the development of the tail apply only to Xenopus. Given the conservation of developmental mecha nisms, we would propose that the mechanism of tail formation will tum out to be very similar in all vertebrates. However, it could be argued that what happens in Xenopus might be the exception, because the tadpole is under strong selective pressure to develop a tail rapidly in order to swim for survival. The currently prevailing view states that the tailbud of a chick or a mouse is a proliferation blastema of undifferentiated cells, from which the various tissues of the tail arise (Griffith et al., 1992).

The comparative approach could be very productive in advancing knowledge in this case. If the Xenopus tail formation mechanism (Gant et al., 1993) operates in amniotes, several strong predictions can be made. For example, a popu lation of cells derived from the organizer should regress through the primitive streak (the region equivalent to the lateral blastopore lip and the posterior wall of Xenopus). This popu lation of cells would be homologous to the chordoneural hinge of Xenopus and should be endowed with potent organizing properties, inducing somites and other tail structures as it comes in contact with pluripotent primitive streak cells. Some observations on mouse somite formation (Tam and Tan, 1992) may be considered to be at least consistent with the mechanism proposed here. Gene markers will be useful, but ultimately definitive answers will require the traditional transplantation and lineage tracing tools available to the embryologist.

In the case of chick and mouse, the tail formation mechanism might be masked by the extensive cell prolifera tion required for the growth of the embryo which, unlike Xenopus, greatly increases in size at late gastrulation. Perhaps the largest obstacle is the lack of a visible neurenteric canal, which serves as a useful landmark, in chick and mouse. However, this canal must be an ancestral chordate characteris tic, for it is present in Amphioxus, shark, turtle and human embryos (see Gant et al., 1993).

Thus in the immediate future the task will be to establish whether a common mechanism exists in the development of the postanal tail, which is a structure found only in the chordates. By focusing attention on this problem it is hoped that more will be learnt about the mechanisms that control late gastrulation in the vertebrates.

We have emphasized throughout this paper how developmen tal mechanisms that control gastrulation in the vertebrates appear to be conserved when examined through the narrow prism of gene markers. One might argue that a productive area of research might be to extend these comparisons by studying homologues of organizer-specific genes in invertebrates, e.g., hemichordates, echinoderms and even in protostomes. Such studies might help answer the question of whether the organizer phenomenon is restricted only to the vertebrates. On the other hand, one could also argue that understanding how differences in body shape arise will be of greater importance in evolutionary studies. After all, mutations in the genes that regulate development may provide one of the main sources of the variation upon which natural selection acts in metazoans (reviewed by De Robertis, 1994).

The choice of emphasizing either similarities or differences can be seen in articles throughout this volume. But this is not something new in evolutionary biology. In 1830 a famous debate took place in the French Academy between Cuvier and Geoffroy Saint-Hilaire (see Appel, 1987). The latter defended the principle of unity of composition among animals. Those who today choose to emphasize homologies marvel at the unity of composition of the mechanisms of development. But others marveled before. One of the observers of the 1830 debate sum marized his views in the following way:

“Il n’y a qu’un animal. Le createur ne s’est servi que d’un seul et meme patron pour tous les etres organises. “

“There is but one animal. The Creator used only a single pattern for all organisms. “ Honore de Balzac in “La Comedie Humaine”, 1842

We are indebted to Stephan Schulte-Merker, Martin Blum, Juan Carlos Izpisua-Belmonte and Claudio Stem for their contributions to the study of expression patterns of goosecoid. H. S. was supported by a DFG postdoctoral fellowship and A. F. was recipient of an American Cancer Society International Cancer Research Fellowship. This work was funded by grants of the NIH (HD21502-09) and HFSPO.