How our complex body structure is generated from a single cell, the fertilized egg, has fascinated people for centuries. Although our understanding of the process continues to increase owing to recent advances in molecular and cellular biology, we still need to clarify the mechanisms by which developmental fates are specified during animal embryogenesis. Early analyses, including tracing of cell lineages (Conklin,1905a) and experimental manipulation of embryonic development (Chabry,1887; Roux,1888; Driesch and Morgan,1895–1896; Wilson,1904; Conklin,1905a,b), led to the idea that the egg is organized by way of localized and inherited maternal factors that are responsible for the development of parts of the embryo and adult. This mode of development is called mosaic development and occurs in the embryos of many organisms. In contrast to this concept, other experiments (Driesch,1892; Spemann,1903) demonstrated that some surgically isolated parts of an embryo can compensate and make a full larva, in a process of self-regulation. By extending this observation, Spemann and Mangold (1924) later demonstrated that one part of the embryo, when transplanted, is capable of inducing development in surrounding cells. This experiment led to the concept of induction and of an inducer known as an “organizer.” The two modes of development, although they appear to be different, are not necessarily incompatible, but rather both cell- autonomous and non–cell-autonomous modes are active during embryogenesis.
Ascidians are marine invertebrate chordates (Fig. 1L–N) with several characteristics that make them useful for studying processes of cell fate specification (Lemaire et al.,2002; Nishida,2005). These characteristics have attracted many researchers over the past decade, and ascidians have emerged as one of the most popular model organisms in developmental biology. A big advantage of using ascidian embryos to understand the mechanisms of cell fate specification is that the ascidian tadpole larva, with a body plan that is characteristic of all chordates, consists of a relatively small number of cells. This feature makes it possible to draw a comprehensive picture of cell fate specification in an embryo at the single-cell level. Because cell fates are mostly restricted in a single cell and determined by the 110-cell stage, it is important to look at events that happen by this stage. Recent advances in studies of ascidian development depend largely on four evolutionarily distant species (Ciona intestinalis [Fig. 1M] and savignyi, Halocynthia roretzi [Fig. 1L], and Phallusia mammillata [Fig. 1N]). Genome-wide analyses on Ciona, micromanipulation techniques on Halocynthia, and imaging techniques on Phallusia have contributed complementarily. Here, we review current knowledge on how developmental fates are specified in the early embryo, starting with a brief description of embryogenesis and cell lineages. Although ascidian development has long been referred to in textbooks as a typical example of mosaic development, recent studies show that it also involves inductive events and that both modes interact with each other. The recent application of molecular biology techniques has also unraveled common molecular mechanisms between ascidians and other organisms, in particular vertebrates.
Embryogenesis and Larval Structure
Ascidian embryogenesis proceeds rapidly and involves a small number of cells. The first cleavage occurs 1 hr 45 min after fertilization in Halocynthia roretzi at 13°C and 1 hr after in Ciona intestinalis at 18°C. After two more synchronous and four asynchronous cleavages, the embryo reaches the 110-cell stage (9 hr in Halocynthia, 5 hr in Ciona), when gastrulation starts. During gastrulation, a single layer of endoderm and mesoderm cells in the vegetal hemisphere invaginates into the interior as the ectoderm layer migrates toward the vegetal pole to form a surface around the embryo (Fig. 1A–C). The spatial relationship between the cells derived from the vegetal hemisphere is dramatically changed during morphogenetic cell movements. Neurulation begins soon after gastrulation is complete (13 hr in Halocynthia, 7 hr in Ciona). As in vertebrates, in ascidian neurulation, the neural plate is curled up dorsally to form a tube-like structure known as the neural tube (Fig. 1D–G). The formation of the tube-like structure progresses from posterior to anterior. Once it is completely closed, the tail becomes distinguishable (18 hr in Halocynthia, 9 hr in Ciona). With tail bud formation (Fig. 1H), the embryo enters a relatively long-lasting stage, the tail bud stage, during which the tail continues to elongate until the embryo is ready to hatch (35 hr in Halocynthia, 18 hr in Ciona). Inside the embryo, the notochord cells converge and extend during neurulation and at the early tail bud stage (Munro and Odell,2002). After the notochord cells are all intercalated and stacked in a single row, they further extend individually along the anterior–posterior (A–P) axis and provide a driving force for tail elongation. Cellular mechanisms of convergence and extension are conserved between ascidians and vertebrates (Munro and Odell,2002).
The anatomy of the ascidian larva (Fig. 1I,J) is similar to that of its amphibian counterpart. Both possess a notochord flanked by blocks of muscle and paralleled by a hollow neural tube located dorsally. The major tissues that constitute the ascidian tadpole are the epidermis, nervous system, notochord, muscle, mesenchyme, trunk lateral and ventral cells (TLCs and TVCs), and endoderm (Figs. 1I,J, 4D). Only the nervous system is complex, in that it has several different cell types (Lemaire et al.,2002). It is composed of the central nervous system (CNS), with fewer than 100 neurons and 250 glial cells (Meinertzhagen and Okamura,2001), and the peripheral nervous system, with 20 to 30 epidermal sensory neurons, which have been mapped precisely (Takamura,1998; Ohtsuka et al.,2001a). The ascidian larva can be subdivided into the trunk and the tail. The tail region is adsorbed after metamorphosis and never contributes to the adult structure (Hirano and Nishida,1997), with the exception of epidermal cells and putative germ cells, which reside in an endodermal structure known as the endodermal strand (Fujimura and Takamura,2000; Takamura et al.,2002; Shirae-Kurabayashi et al.,2006), which runs ventral to the notochord along the entire length of the tail. The tail region, therefore, seems to be used only to propel the tadpole to the substrate for metamorphosis. The notochord, which runs from anterior to posterior in the center of the tail (Figs. 1I,J, 4D), consists of exactly 40 cells in most solitary ascidian species, and provides a supportive structure when the flanking muscle contracts and extends for tail movement. The muscle cells are aligned in three rows on each side of the notochord (Fig. 4D; 21 cells on each side in Halocynthia and 18 in Ciona) and generate waves of motion of the tail in a bilateral direction. They are striated but never fused like those in vertebrates. The posterior neural tube, known as the nerve cord in ascidians, runs dorsal to the notochord in the tail region (Figs. 1I,J, 4D, 7C), and consists of only four rows (1 dorsal, 2 lateral, and 1 ventral) of ependymal cells and bundles of axons (Fig. 7D).
In contrast to the tail region, the trunk region remains after metamorphosis, and consists of several types of undifferentiated tissues that are used for adult structures, although larvae of some compound ascidian species already have differentiated gill slits and digestive ducts in the trunk region. Such undifferentiated tissue types include mesenchyme, the TLCs, the TVCs, and endoderm. The mesenchyme cells, situated bilaterally in the posterior region of the trunk (Figs. 1I,J, 4D), give rise to tunic cells in the adult (Hirano and Nishida,1997; Tokuoka et al.,2005). The TLCs and TVCs are small groups of cells located bilaterally dorsal and ventral, respectively, to the mass of mesenchyme cells (Fig. 4D). The TLCs become body wall (oral siphon and longitudinal mantle) muscle and blood cells, while the TVCs give rise to body wall (atrial siphon and latitudinal mantle) muscle, heart, and pericardium (Hirano and Nishida,1997; Satou et al.,2004). The endoderm cells, which look homogeneous and occupy two thirds of the trunk region anteriorly (Figs. 1I,J, 4D), differentiate into several endodermal organs after metamorphosis, including endostyle, branchial sac, peribranchial epithelium, and digestive organs (Hirano and Nishida,2000). The larva does not feed until after metamorphosis. The trunk region also contains a brain (a.k.a. sensory vesicle) in the dorsal region (Figs. 1I,J, 4D, 7C) and three palps as adhesive organs at the anterior tip of the trunk. The brain, at the anterior-most part of the CNS, contains a light-sensing organ known as an ocellus and a gravity-sensing otolith, which are used to control larval behavior before metamorphosis (Tsuda et al.,2003). The CNS consists of several parts along the A–P axis (Lemaire et al.,2002), including the visceral ganglion, posterior to the brain, and the posterior nerve cord, in the tail region (Figs. 1I,J, 7C).
Establishment of Embryonic Axes
The first axis to be established in ascidian development is the animal–vegetal (A–V) axis. The animal pole is defined by the position in the egg where the polar bodies form, and the vegetal pole by the position opposite the animal pole. Although the polar bodies become visible after fertilization, the A–V axis is already established during oogenesis. In Xenopus, several developmentally important maternal mRNAs, such as Xcat2, Vg1, and VegT, are localized to the vegetal cortex during oogenesis (King et al.,2005). The polarization along the A–V axis in the ascidian oocyte is evident by the migration of the meiotic spindle to the egg cortex at the animal pole following the germinal vesicle breakdown (GVBD), with concomitant exclusion of mitochondria and maternal mRNAs known as postplasmic/PEM (posterior end mark) RNAs (Yoshida et al.,1996; Satou and Satoh,1997; Sasakura et al.,1998a,b,2000; Satou,1999; Makabe et al.,2001; Nakamura et al.,2003,2006; see another article by F. Prodon et al. in this issue) from the site where the spindle reaches (Prodon et al.,2006). These materials were distributed throughout the germinal vesicle (GV) stage oocyte at the cortex (Prodon et al.,2006). The mechanism by which the meiotic spindle migrates is not known, but this process could be the first symmetry-breaking event in the ascidian egg. However, in Halocynthia, unlike other species such as Ciona and Phallusia mammillata, the GV is already situated in an eccentric position (Numakunai,2001), and the meiotic spindle migrates to the closest cortex (F. Prodon and H. Nishida, unpublished observation). The mature oocyte is spawned as an unfertilized egg with the meiotic spindle situated at the cortex of the animal pole. The A–V axis is defined as the future ventral–dorsal axis of the tadpole larva.
The high concentration of maternal mRNAs at the vegetal pole, like that observed in the Xenopus oocyte (King et al.,2005), is achieved in ascidians shortly after fertilization by a cytoplasmic and cortical reorganization known as the first phase of ooplasmic movements (Fig. 2B,E,H; Conklin,1905a; Sardet et al.,1989; Roegiers et al.,1999; see another article by C. Sardet in this issue). During the movements, the mitochondria and postplasmic/PEM RNAs, which show a biased distribution along the A–V axis in the mature oocyte (Nishida and Sawada,2001; Prodon et al.,2005; Fig. 2A,D,G), become further and highly concentrated at the vegetal pole in a microfilament-dependent manner.
The A–P axis is established perpendicular to the A–V axis after fertilization. The second phase of ooplasmic movements (Fig. 2C,F,I; Conklin,1905a; Sardet et al.,1989; Roegiers et al.,1999; see another article by C. Sardet in this issue) translocates the vegetally localized cytoplasm/cortex, which contains mitochondria, cortical endoplasmic reticulum, and postplasmic/PEM RNAs, to the future posterior pole (Sardet et al.,2005), in a region called the posterior vegetal cytoplasm/cortex (PVC). This second reorganization, unlike the first, is driven by microtubules emanated from the centrosome supplied by the sperm. How the mechanism by which the cytoplasm/cortex is reorganized is switched from microfilament-dependent during the first phase to microtubule-dependent during the second is not known. The establishment of the A–P axis depends on the PVC, which exhibits a posteriorizing activity. The removal or transplantation of the PVC to the future anterior region of the normal egg causes duplication or deficiency of the anterior structure in the original posterior or anterior region, respectively (Nishida,1994). These results suggest that the anterior fate may arise by the absence of the posteriorizing activity. Some of the localized postplasmic/PEM RNAs within the PVC represent the posteriorizing activity. PEM (Yoshida et al.,1996), first identified as a postplasmic/PEM RNA, regulates the posterior-specific cleavage pattern (see below; T. Negishi, K. Sawada, H. Nishida, unpublished observations), while Macho-1 (Fig. 2G–I) plays an essential role in the formation of the posterior-specific tissues such as the primary muscle and mesenchyme (Nishida and Sawada,2001; Satou et al.,2002; Kobayashi et al.,2003). Knockdown of other postplasmic/PEM RNAs by morpholino antisense oligo (MO) injection has indicated that the postplasmic/PEM RNAs function over and influence a broader region of the embryo than just individual tissues (Nakamura et al.,2005,2006). This mechanism by which an axis is formed by the inheritance of maternal localized factors is conserved in many other organisms as well (Pellettieri and Seydoux,2002; Weaver and Kimelman,2004; Minakhina and Steward,2005). In ascidians, the conventional A–P axis does not precisely correspond to that of larvae, because cells do not stay stationary during gastrula cell movements (Nishida,2005).
When the A–P axis forms, the third axis, the mediolateral axis, is automatically established. The left–right asymmetry in the ascidian embryo is morphologically recognizable during tail elongation and in the brain positioning. The tail bends leftward, while elongating in most cases in Halocynthia (Morokuma et al.,2002). Ascidian orthologues of Nodal and Pitx2, which are essential for the establishment of the left–right asymmetry in vertebrates (Boorman and Shimeld,2002; Raya and Belomonte,2006; Hirokawa et al.,2006) and sea urchins (Duboc et al.,2005), are expressed only on the left side of the epidermis at the tail bud stage (Morokuma et al.,2002) and, therefore, might be responsible for this bending. The brain also becomes asymmetric in that it is rotated clockwise when viewed from the posterior pole (Kats,1983; Nicol and Meinertzhagen,1991). Cell lineage analysis shows that this asymmetry takes place between the late tail bud and larval stages (Taniguchi and Nishida,2004).
The cleavage pattern of the ascidian embryo is invariant between individuals (Conklin,1905a; Satoh,1979), unlike in vertebrates. The first cleavage divides the embryo into left and right halves. The embryo continues its bilaterally symmetrical cell division at least until gastrulation. The cleavage patterns in the anterior and posterior halves, on the other hand, are significantly different. This difference is largely because the posterior-most blastomeres in the vegetal hemisphere undergo three consecutive cleavages that are unequal in cell size. The B4.1 blastomere of the 8-cell stage embryo divides unequally to produce a smaller daughter B5.2 at the posterior-most position of the 16-cell stage embryo. Likewise, B5.2 produces a smaller B6.3, which, in turn, yields a smaller B7.6 posteriorly. The B7.6 blastomere, the smallest at the 64-cell stage, never divides again until the tail bud stage, when it ends up in a single cell situated in the endodermal strand (Nishida,1987), although it is recently shown in Ciona embryos that B7.6 divides into the B8.11 and B8.12 blastomeres during gastrulation with the latter blastomere undergoing two more divisions at the late tail bud stage (Shirae-Kurabayashi et al.,2006). The B8.12 blastomeres are considered to become germ cells after metamorphosis (Fujimura and Takamura,2000; Takamura et al.,2002; Shirae-Kurabayashi et al.,2006).
The mechanism by which these unequal cleavages take place involves a structure called the centrosome-attracting body (CAB; Fig. 3; Hibino et al.,1998; Nishikata et al.,1999; Iseto and Nishida,1999), which is situated at the posterior cortex of and, thus, is always inherited by the posterior-most blastomeres: B4.1, B5.2, B6.3, and B7.6. The CAB begins to form at the two-cell stage, and its formation requires the PVC (Nishikata et al.,1999). In each of the unequal cleavages, the CAB in the posterior-most position appears to shorten the bundle of microtubules that connect it and one of the centrosomes; consequently, the nucleus, the mitotic apparatus, and the cleavage plane shift posteriorly (Fig. 3; Nisihkata et al.,1999). The CAB is where the postplasmic/PEM RNAs are concentrated during the unequal cleavages (Sasakura et al.,2000; Nakamura et al.,2003,2005; Sardet et al.,2003). Knockdown of PEM results in the posterior-most cells undergoing equal cleavages (T. Negishi, K. Sawada, H. Nishida, unpublished observations). Interestingly, the CAB, which is inherited by the B7.6 (the putative germline), contains an electron-dense matrix that resembles the germ plasm in other organisms (Iseto and Nishida,1999). Thus, the CAB appears to be a multifunctional structure that is important for understanding the ascidian development.
Cells normally divide with their mitotic spindle oriented perpendicular to the axis of the previous cell division, because after the previous division, the centriole replicates and the two centrosomes migrate by 90 degrees in opposite directions. This finding is sometimes not the case, however, when asymmetric cell division in cell fates is involved, in which the two centrosomes in the perpendicular orientation further rotate 90 degrees, mostly at interphase to ensure that the polarized difference in the mother cell is partitioned correctly (Goldstein,2000; Roegiers and Jan,2004; Cowan and Hyman,2004). In this case, the cell divides in the same direction as the previous one. This pattern of cleavage is also observed during the cleavage stages in the vegetal hemisphere of the ascidian embryo. For example, the A5.1 blastomere of the 16-cell stage embryo divides in the meridional direction and produces the A6.1 blastomere vegetally and the A6.2 more animally. At the next division into the 64-cell stage, the A6.2 again divides in the same direction. This second cell division is asymmetric, producing two daughter cells with different fates (notochord vegetally and nerve cord more animally; Nishida,1987). The notochord fate is induced by fibroblast growth factor (FGF) signaling (Nakatani and Nishida,1994; Nakatani et al.,1996; Shimauchi et al.,2001; Imai et al.,2002a; Kumano et al.,2006), with the nerve cord fate as default (Minokawa et al.,2001; see below); however, the signal does not seem to be involved in the positioning of the mitotic spindle during this asymmetric cell division at least in Halocynthia, as the orientation of the cell division is not altered when FGF signaling is blocked (personal observation by G. Kumano and H. Nishida). This finding contrasts with the situation of the asymmetric cell division of the EMS blastomere in C. elegans, in which Wnt signaling from an adjacent posterior cell induces the fate of one of the daughter cells and at the same time orients the mitotic spindle such that the EMS divides in the same direction as the previous one (Goldstein,1992,1995; Rocheleau et al.,1997; Thorpe et al.,1997; Schlesinger et al.,1999).
Cell lineages of the ascidian embryo are well established and are invariant between individuals (Fig. 4A,C; Conklin,1905a; Nishida,1987). One of the most prominent features of ascidian embryogenesis extracted from the cell lineages is that developmental fates are restricted at very early stages (Fig. 4B). Most of the blastomeres at the 110-cell stage (102 of 110 cells), just before gastrulation, are already fated to give rise to a single type of tissue (Fig. 4B,C). Importantly, the fates are not only restricted but also determined in that even if these blastomeres are isolated at this stage, they differentiate into tissues according to their normal fates (Reverberi and Minganti,1946; Nishida,1992). Each blastomere at the fate-restricted stages has unique properties in terms of the cell type and the position of its descendants at the larval stage, and the number of cell cycles it will undergo until fully differentiated. One of the presumptive muscle cells, B7.4, at the 64-cell stage, for example, divides three more times to become eight muscle cells situated in the middle part of the tail. Developmental fates and the number of cell divisions might be tightly linked. The presumptive nerve cord cells when isolated and treated with FGF signaling (changing their fates to notochord) stop dividing at the time the normal notochord cells do (Nakatani et al.,1996). Experimental manipulation of C. elegans transforms both the fate of a cell and its own cell division pattern (Goldstein,1993; Draper et al.,1996).
Another feature of ascidian embryogenesis is that cells that constitute a tissue do not necessarily originate from a single lineage (Fig. 4A; Nishida,1987). For example, the primary muscle cells originate from both the B5.1 and the B5.2 lineages, which are separated at the division to the 16-cell stage. The B5.1 and B5.2 are fated to become not only the primary muscle cells but also other cells, such as endoderm and mesenchyme. Because the formation of the primary muscle cells relies on the inheritance of localized maternal determinants (Conklin,1905a; Nishida,1992), either the determinants would be inherited by the other endoderm and mesenchyme cells as well, or they would have to continue to be distributed asymmetrically at every cell cycle until the primary muscle fate is restricted. As discussed below in the Endoderm Formation section, the muscle determinant Macho-1 (Nishida and Sawada,2001) seems to be present in all the descendants of the B5.1 and B5.2 lineages (Kondoh et al.,2003). Therefore, a later event must take place to confine the primary muscle fate to the appropriate cells. That tissues, as we have seen in the case of the primary muscle, originate from multiple lineages (Fig. 4A), but tissue-forming territories in the fate map (see below) are assembled in one place bilaterally or singly (Fig. 4C), indicates that specification of developmental fates depends on spatial cues rather than on lineages.
It is notable that notochord and muscle cells have primary and secondary lineages (Nishida,1987). The primary lineages constitute the anterior parts of the respective tissues in the tail region (32 of 40 notochord cells and 28 of 42 muscle cells in both Halocynthia and Ciona), while the secondary lineages make up the posterior parts (8 notochord cells in both species, and 14 and 8 muscle cells in Halocynthia and Ciona, respectively). Although the primary and secondary lineage cells are functionally equivalent at the larval stage, the mechanisms by which the fates of these cells are specified are different (Nishida and Sawada,2001; Darras and Nishida,2001a; Imai et al.,2002b; Hudson and Yasuo,2005,2006).
The fate map of the 110-cell stage ascidian embryo is shown in Figure 4C. As in vertebrates (Dale and Slack,1987; Moody,1987; Lane and Smith,1999), ectoderm and endoderm in ascidians arise from cells in the animal and vegetal poles, respectively, and mesoderm forms in between at the equatorial region. The similarity between ascidians and vertebrates is also evident in the distribution of major tissue-forming territories, including epidermis, neural tissues, notochord, muscle, and endoderm, which appears to correspond closely with that of the same tissues in the Xenopus fate map (Kourakis and Smith,2005; Nishida,2005; Lane and Sheets,2006). This finding suggests that the body plan shared between ascidians and vertebrates is based upon a generalized fate map of the chordates. Although it was considered in the past that ascidian development depends largely on maternal localized factors, whereas vertebrate development depends on inductive events, recent studies have accumulated evidence that both modes of development are equally used in both ascidians and vertebrates, as discussed in the next section.
SPECIFICATION OF DEVELOPMENTAL FATES
Endoderm formation relies on the inheritance of localized maternal factors by cells of the vegetal hemisphere. In Xenopus, mRNA for the T-box transcription factor VegT is localized to the vegetal hemisphere during oogenesis and is inherited by vegetal cells (Zhang and King,1996; Stennard et al.,1996), where it activates its target genes for endoderm formation (Zhang et al.,1998). In ascidians, maternal β-catenin sits at the top of a hierarchy so far that leads to endoderm formation (Fig. 5; Imai et al.,2000). The maternal β-catenin activates the expression of several genes, including lhx-3 and ttf-1 (Fig. 5; Satou et al.,2001), which encode key transcription factors for endoderm formation (Nishida,2005). Lhx-3 is necessary and sufficient even without β-catenin for the expression of a late endoderm differentiation marker, alkaline phosphatase (AP) (Satou et al.,2001), while TTF-1 is sufficient for AP expression (Ristoratore et al.,1999: Satou et al.,2001). These downstream target genes become expressed only in cells of the vegetal hemisphere (Fig. 5), because β-catenin goes into nuclei only in those cells at the cleavage stage (Imai et al.,2000), although the protein is uniformly present in the entire embryo. Despite its role in initiating endoderm formation, the ubiquitous expression of the protein is not consistent with that of what might be a maternal cytoplasmic determinant for endoderm formation. Results from earlier cytoplasmic transfer experiments suggested that it is localized to the vegetal pole of the fertilized egg and inherited by the cells of the vegetal hemisphere (Nishida,1993). Therefore, a genuine localized endoderm determinant appears to be an upstream activator of β-catenin and translocates β-catenin into cell nuclei.
β-Catenin enters nuclei not only of the presumptive endoderm cells, but also of other vegetal cells that are not fated to become endoderm (Imai et al.,2000). Several other downstream genes of β-catenin are, in fact, expressed in endoderm and mesoderm cells. Those include FGF9/16/20 and the forkhead transcription factors FoxD and FoxA (Fig. 5; Imai et al.,2002a,b; Kumano et al.,2006). Even the expression of lhx-3 is not specific to the endoderm lineage (Satou et al.,2001). It seems that an endoderm determinant is active or potentially active in other vegetal cells than the presumptive endoderm cells as well. Therefore, it is important for the embryo to have a mechanism by which to ensure that endoderm does not form in the mesoderm lineage. How mesendoderm separates into endoderm and mesoderm has been investigated in other organisms. It has been shown in vertebrates and echinoderms that cell–cell communication by means of FGF or Notch signaling plays an essential role in the separation between these two germ layers (Rodaway et al.,1999; Contakos et al.,2005). In ascidians, FGF9/16/20 signaling activates the basic helix–loop–helix (bHLH) transcription factor twist-like-1 (Imai et al.,2003), which prevents endoderm from forming in the mesenchyme lineage (Imai et al.,2003; Tokuoka et al.,2004). FoxD and its downstream target, the zinc finger protein Zic (Imai et al.,2002c; Wada and Saiga,2002), also suppress endoderm formation in the notochord lineage (Imai et al.,2002b; Kumano et al.,2006). Accordingly, β-catenin regulates the expression of several genes that are important for endoderm (lhx-3 and ttf-1) and mesoderm (FGF9/16/20, twist-like-1, FoxD, and Zic) formation. It would be interesting to clarify how these downstream target genes are expressed in the different regions of the embryo after activation by β-catenin. In sea urchin, a gradient of β-catenin stability along the A–V axis controls cell specification (Weitzel et al.,2003). Whether ascidians use the same mechanism or other factors are involved remains to be seen.
Ascidian endoderm is subdivided into anterior and posterior parts with regard to their origins (Kondoh et al.,2003). The anterior endoderm is originated from the A4.1, the A–V blastomeres of the eight-cell stage embryo, and occupies the anterior part of the trunk endoderm of the tail bud embryo. The posterior endoderm, on the other hand, is derived from the B4.1, the P–V blastomeres, and develops into the posterior part of the trunk endoderm and the endodermal strand, which runs ventral to the notochord along the length of the tail. Whereas the anterior endoderm depends solely on a yet-unidentified localized maternal determinant for its formation, it is known in Halocynthia that the posterior endoderm additionally requires cell–cell communication to properly form; otherwise, it becomes muscle (Kondoh et al.,2003). This ectopic muscle formation is dependent on the function of Macho-1, a maternal localized determinant for muscle formation (see below), which is inhibited by cell–cell communication by means of FGF signaling (Kondoh et al.,2003). This phenomenon is another example where the function of a maternal cytoplasmic determinant is inhibited in cells that happen to inherit it. It seems that later events are elaborated to sharpen further the boundaries between tissues that have been roughly established by localized determinants.
Finally, endoderm also serves as the source of mesoderm-inducing signals (Nishida,2005; Kimelman,2006). The ascidian endoderm expresses FGF9/16/20 (Imai et al.,2002a; Kumano et al.,2006) and induces adjacent cells into mesoderm, such as the notochord and mesenchyme (Nakatani and Nishida,1994; Nakatani et al.,1996; Kim and Nishida,1999,2001; Kim et al.,2000; Shimauchi et al.,2001; Imai et al.,2002a,2003; Kumano et al.,2006). In the next section, we discuss mesoderm formation/induction in ascidians.
Mesoderm in ascidians is induced in the equatorial region (marginal zone of the vegetal hemisphere) mainly by a signal from the vegetal cells (Fig. 6A; Nishida,2005), as has been shown in vertebrates (Kimelman,2006). Whereas members of the transforming growth factor-β superfamily induce mesoderm in vertebrates (Feldman et al.,1998; Kofron et al,1999; Agius et al.,2000; Takahashi et al.,2000), FGF signaling represented by FGF9/16/20 plays a major role in mesoderm induction in ascidian embryos (Imai et al.,2002a,2003; Kumano et al.,2006). The FGF9/16/20 signal emanating from endoderm cells induces the notochord and mesenchyme in the anterior and posterior regions of the cleaving embryo, respectively (Fig. 6A; Imai et al.,2002a,2003; Kumano et al.,2006). As observed in the case of the notochord and mesenchyme induction, mesoderm in many organisms is induced in different cell types by the same signal (Cui et al.,1996; Kim et al.,2000; Kumano et al.,2001; Kobayashi et al.,2003). This finding is because regionalization within the mesoderm-arising territory takes place before the induction and provides different competence to respond to the inducing signal. In Halocynthia, maternal Macho-1 is responsible for this preinduction regionalization (Fig. 6C). This is another function of this protein, after its function as a muscle determinant as mentioned above. Maternal transcripts for Macho-1 become localized to the future posterior pole by means of ooplasmic movements after fertilization (Fig. 2G–I, see the article by Prodon et al. in this issue) and are inherited by the posterior blastomeres, and the protein confers posterior-specific responsiveness to the FGF9/16/20 signal at the 32-cell stage or later, when induction occurs (Fig. 6C; Kobayashi et al.,2003). A combination of Macho-1 and FGF9/16/20 signaling appears to lead to the expression of the twist-like-1 gene in the presumptive mesenchyme cells, which regulates all the mesenchymal genes known to date that are expressed at later stages (Imai et al.,2003). In the anterior region, on the other hand, competence to respond to the FGF9/16/20 signal appears to arise because of the absence of Macho-1 (Fig. 6C; Kobayashi et al.,2003). Zygotically expressed FoxA and Zic have recently been identified as intrinsic competence factors for notochord induction (Wada and Saiga,2002; Kumano et al.,2006). Maternally derived Macho-1, therefore, might suppress the function of FoxA and Zic in the posterior region. The FGF9/16/20 signal also induces the secondary muscle and the secondary notochord by means of activation of nodal signaling in cells (a pair of b6.5 blastomeres) in the animal hemisphere (Hudson and Yasuo,2005,2006).
Molecular cascades that lead to the expression of the notochord-specific brachyury gene (Yasuo,1993,1994) at the 110-cell stage have been extensively studied. Maternal β-catenin activates both intrinsic and extrinsic factors that are required for brachyury expression. FoxD and FoxA, both activated by β-catenin as early as the 16-cell stage (Imai et al.,2002b,2006; Kumano et al.,2006), are required for the production of Zic (Imai et al.,2002c,2006; Kumano et al.,2006) in the primary lineage, which together with FoxA confers notochord-specific responsiveness to FGF signaling (Wada and Saiga,2002; Kumano et al.,2006). The expression domains of FoxA and Zic overlap only in the presumptive notochord cells (Shimauchi et al.,1997; Wada and Saiga,2002; Imai et al.,2002c) in the anterior half of the embryo. The expression of the notochord inducer FGF9/16/20 is also activated by β-catenin (Imai et al.,2002a; Kumano et al.,2006). The secondary notochord is induced by Notch/Delta signaling in Ciona (Corbo et al.,1998; Imai et al.,2002b,2006; Hudson and Yasuo,2006), although this may not be the case in Halocynthia (Darras and Nishida,2001a; Akanuma et al.,2002). A recent study shows that the nodal signal from b6.5, both directly and indirectly by means of activation of Delta2, regulates brachyury expression in the secondary lineage (Hudson and Yasuo,2006).
Some other types of mesoderm are induced by signals from nonendodermal cells. In Xenopus, the ventral blood islands, where blood cells originate, are induced by bone morphogenetic protein (BMP) signaling from ectoderm cells (Maeno et al.,1994; Kumano et al.,1999). In ascidians, the TLCs, precursors of blood cells, are induced at the 16-cell stage by a signal from ectoderm (Kawaminami and Nishida,1997). Recent studies suggest that this signal may be represented by nodal, which emanates from b6.5 (Imai et al.,2006; Hudson and Yasuo,2006). In addition to the signal, the TLC also requires several other factors, including FoxD and NoTrlc (Imai et al.,2003,2006), whose function is probably intrinsic. The TVCs, precursors of heart in ascidians, are also induced. Heart precursors are induced by FGF signaling in two cells (a pair of B8.9 blastomeres) of the four competent cells (daughters of the B7.5 blastomeres, B8.9 and B8.10) that constitute the heart-forming field and are characterized by expression of the bHLH gene Mesp (Davidson and Levin,2003; Satou et al.,2004; Davidson et al.,2005,2006). Without the induction, the presumptive heart becomes larval tail muscle (Nishida,1992; Davidson et al.,2005,2006). Details about the formation of blood and heart, including the events that happen at later stages, are reviewed in other articles in this issue.
Asymmetric cell divisions are observed during ascidian embryogenesis, as mentioned in the previous section “2-3 Cleavage pattern.” In organisms that are known for their invariant embryogenesis with a relatively small number of embryonic cells, such as C. elegans and ascidians, cell fates are often induced and determined in a cell cycle before the fates are restricted (Nishida,1996). In ascidian notochord or mesenchyme induction, cells are fated to become both notochord and nerve cord, or mesenchyme and muscle, respectively, at the time they are induced at the 32-cell stage (Fig. 4A). It is the next division to the 44-cell stage when the cells asymmetrically divide to provide one daughter that gives rise to notochord or mesenchyme as induced fates and the other that becomes nerve cord or muscle as default fates (Fig. 6B). The notochord or mesenchyme cells arise closer to the endoderm (the source of the FGF9/16/20 signal), while the nerve cord or muscle cells arise away from it (Fig. 6B). This asymmetric division depends largely on external signals (Nakatani et al.,1996; Kim and Nishida,1999; Minokawa et al.,2001). Both daughter cells could adopt induced fates (notochord and mesenchyme) when the entire surface of the mother cells is exposed to FGF signaling, but take on default fates (nerve cord and muscle) in the absence of the signal (Fig. 6C). No intrinsic factors have been identified so far as being localized on one side of the mother cells before the division. It is, therefore, likely that the position from which FGF9/16/20 signaling is presented determines the asymmetry. In C. elegans, Wnt signaling can function as positional cues in establishing the polarity in the EMS cell (Goldstein et al.,2006), in which endoderm is induced before its fate is restricted in the E cell after the next division (Goldstein,1992). It is interesting that a common strategy to induce tissues is used in two distant species, ascidians and C. elegans. Whether this is also the case in other organisms remains to be seen.
Cell-autonomous formation of primary muscle.
In contrast to the mesoderm tissues described above, the formation of the primary muscle in ascidians does not require cell–cell communication and is executed cell-autonomously. Maternal mRNAs for Macho-1 is present in a localized region of the egg and, at the eight-cell stage, are inherited by the B4.1 blastomeres, which give rise to the primary muscle and other tissues (Fig. 4A; Nishida and Sawada,2001). Macho-1 encodes a transcription factor and is necessary and sufficient for the expression of several genes such as tbx6 and myoD that are important for muscle formation (Nishida and Sawada,2001; Yagi et al.,2004,2005; Sawada et al.,2005). Mesoderm formation by maternally localized factors may be a common strategy in vertebrates. Maternal β-catenin restricted to the prospective mesoderm region is involved in early mesoderm induction in Xenopus embryos (Schohl and Fagotto,2003). This region is marked by Spemann organizer-independent early expression of myoD and is fated to become muscle (Kumano and Smith,2002). As mentioned above, the function of Macho-1 as a muscle determinant must be inhibited in other cells than the primary muscle cells. The Macho-1 protein is supposed to be present in every descendant of B4.1 (Kondoh et al.,2003). Among the cell types that are derived from B4.1, posterior endoderm, mesenchyme, and secondary notochord cells require FGF signaling to undo the function of Macho-1 as a muscle determinant (Kim et al.,2000; Kim and Nishida,2001; Kondoh et al.,2003). Macho-1, however, is required in the presumptive mesenchyme cells as an intrinsic competence factor for mesenchyme induction (Kobayashi et al.,2003). It is, therefore, likely that the function of Macho-1 as a transcriptional activator (Sawada et al.,2005) per se is retained at least in the presumptive mesenchyme cells and that only the activation of muscle-specific genes is inhibited by FGF signaling.
Recent studies with chicks (Gallus gallus) and Xenopus have challenged the default model for neural specification (Streit et al.,1998,2000; Streit and Stern,1999; Wilson et al.,2000,2001; Delaune et al.,2005), in which a default neural state is preserved within the ectoderm unless BMP signaling is unopposed, which would lead to epidermis formation (Munoz-Sanjuan and Brivanlou,2002). These studies highlight the role of FGF signaling as an initiator of neural specification and suggest that the signal functions even in a BMP-independent manner (Streit et al.,2000; Wilson et al.,2000; Delaune et al.,2005). BMP inhibition, however, is required for maintaining the neural identity at later stages during gastrulation (Stern,2005), although it is also suggested that it plays an early role at the pregastrula stage in Xenopus neural induction (Baker et al.,1999; Wessely et al.,2001; Kuroda et al.,2004). In ascidians, the FGF9/16/20 signal also induces neural tissues (Inazawa et al.,1998; Hudson and Lemaire,2001; Darras and Nishida,2001b; Bertrand et al.,2003; Miyazaki et al.,2007). BMP inhibition is not involved in this process, and a chordin vs. BMP antagonism takes place at a later stage to specify pigment cells within a group of cells that have been specified as neural tissues (Darras and Nishida,2001b). Therefore, it could be argued that BMP inhibition might play a permissive role for further differentiation of neural cells in chordates (Kourakis and Smith,2005).
The ascidian CNS develops from three different lineages: the brain (sensory vesicle) mainly from the animal a-line, the dorsal-most row of ependymal cells in the nerve cord from the animal b-line, and the lateral and ventral rows from the vegetal A-line (Fig. 7; Nishida,1987). The a-line neural tissues are induced by FGF9/16/20 signaling from vegetal cells (Bertrand et al.,2003; Miyazaki et al.,2007). The brain induction in the a-line requires prolonged exposure to the signal from the 32-cell stage to the 110-cell stage (Reverberi et al.,1960; Nishida and Satoh,1989; Nishida,1991; Wada et al.,1999, Akanuma and Nishida,2004). In cells of the animal hemisphere, Ets and GATAa transcription factors are activated by the FGF9/16/20 signal (Bertrand et al.,2003). Ets is also activated in cells of the vegetal hemisphere and is required for notochord, mesenchyme, and heart induction (Miya and Nishida,2003; Davidson et al.,2006). GATAa, on the other hand, is activated only in the animal hemisphere, because it is translated only there at the 32-cell stage, when induction occurs (Bertrand et al.,2003), although its transcripts are present ubiquitously (Bertrand et al.,2003). Accordingly, Ets, together with GATAa in response to the FGF signal, activates the expression of the brain-specific otx gene only in the animal hemisphere (Bertrand et al.,2003), which leads to brain formation. It is proposed that the size of the contact area between individual animal cells and FGF9/16/20-expressing vegetal cells plays an important role in selecting the induced cells within the animal cells (Tassy et al.,2006).
The A–P patterning of the CNS has also been extensively studied in vertebrates, and two major models, the idea of separate head and trunk/tail organizers (Spemann,1931,1938; Mangold,1933) and the activation/transformation hypothesis (Nieuwkoop et al.,1952; Nieuwkoop and Nigtevecht,1954), have been put forward. Although it is currently unclear which model is favored in the ascidian CNS patterning, recent studies show that it involves a multistep process. The process includes a pre-existing A–P patterning in the ectoderm region before induction represented by FoxAa expression at the eight-cell stage (Lamy et al.,2006), and specification of posterior identity within the A-line nerve cord precursors by FGF signaling (Hudson et al.,2003). Occurrence of a pre-existing A–P patterning is reported in zebrafish embryos (Koshida et al.,1998). Posteriorizing activity of FGF in addition to Wnt and retinoic acid as a transformation signal is well known (Altmann and Brivanlou,2001).
Cell-autonomous formation of nerve cord.
In contrast to the a-line neural tissues, the A-line nerve cord precursor cells acquire neural fate cell-autonomously without cell–cell communication (Minokawa et al.,2001). The default nerve cord state is preserved, because FGF signaling is not activated in these cells (Fig. 6C). Intrinsically, Zic is necessary and sufficient for the expression of the nerve cord marker ETR-1 (Wada and Saiga,2002; Imai et al.,2002c), although it also plays an essential role in its sister lineage, the notochord, as mentioned above in the Mesoderm Formation section.
The nerve cord cells are patterned along the future dorsal–ventral axis at the 110-cell stage, such that the medial four cells (pairs of A8.7 and A8.8) give rise to the ventral-most row, while the lateral four cells (pairs of A8.15 and A8.16) give rise to the lateral rows (Fig. 7) (Nishida,1987). The ventral-most cells are thought to correspond to the vertebrate floor plate, judging from the expression of orthologues of FoxA and Hh in both the vertebrate floor plate and the ascidian counterpart (Di Gregorio et al.,2001; Takatori et al.,2002). Whereas it has been believed that the floor plate in vertebrates is induced by signals from the notochord after the neural tube forms (Placzek,1995; Placzek et al.,2000), recent studies have suggested that the floor plate and notochord share a common origin (Le Douarin et al.,1998; Teillet et al.,1998; Le Douarin and Halpern,2000) and that Notch signaling (which activates Hh signaling) is involved in fate decision between the two tissues before the mid-gastrula stage (Appel et al.,1999; Lopez et al.,2003,2005). A similar mechanism may exist in ascidians as well, in that fates of the nerve cord, although it involves not only the ventral but also lateral cells, and notochord are shared in the same cell until the 32-cell stage, as mentioned above, and that the fate decision between these two tissues involves binary choice of alternative fates by FGF9/16/20 signaling (Fig. 6C; Minokawa et al.,2001). Therefore, the last common ancestor of ascidians and vertebrates might have had a common origin of notochord and floor plate.
Mechanisms by which the ventral nerve cord in ascidians and the floor plate in vertebrates arise within the neural tube, on the other hand, seem to be different. The fate of the vertebrate floor plate is induced by Hh and Nodal signaling (Strahle et al.,2004), whereas the ventral fate within the ascidian nerve cord is restricted to the medial cells as a result of Nodal signaling promoting the lateral, but not ventral, fate (Hudson and Yasuo,2005). Therefore, different mechanisms generate a conserved structure such as the neural tube in ascidians and vertebrates.
Neural crest cells.
There were thought to be no morphologically recognizable delaminating and migratory neural crest cells in ascidians (Holland and Holland,2001). A recent study, however, shows that a colonial ascidian has migratory cells that become pigment cells (Jeffery et al.,2004), suggesting that the neural crest cells may not be a vertebrate invention. The peripheral nervous system is another tissue that is derived from the neural crest cells in vertebrates. Some of the ascidian epidermal sensory neurons, which probably sense mechanical stimuli, originate from cells near the boundary between the neural plate and epidermis (Pasini et al.,2006), and transiently expresses an ascidian orthologue of Pax3/7 (Wada et al.,1997; Ohtsuka et al.,2001b; Yagi and Makabe,2001), a neural crest cell marker in vertebrates. These cells, however, do not delaminate or migrate. Accordingly, ascidians may have a genetic program that generates neural crest-like cells, although they may not have genuine neural crest cells.
CONCLUSIONS AND PERSPECTIVES
With extensive studies over the past two decades, we are now beginning to understand molecular mechanisms of cell fate specification in each tissue of the ascidian embryo. Moreover, these studies, together with those using different vertebrate species, have uncovered several similarities in the mechanisms used between ascidians and vertebrates. Cell-autonomous specification of the primary muscle cells and the A-line neural tissues (nerve cord) was once considered to be a perfect example of how ascidians and vertebrates develop differently; however, even in the formation of these tissues, as discussed above, there might be similar mechanisms. Perhaps more significantly, however, ascidians and vertebrates occasionally use different strategies even when a similar body plan is generated, and use common mechanisms that generate a different body plan. It will be as important to study these differences as it has been to study the similarities during embryogenesis. To give an example of a difference, molecular cascades that involve maternal β-catenin are used to specify the A–V axis in ascidians (and echinoderms) but to specify an axis (the organizer-contra-organizer axis) perpendicular to it in vertebrates. In other words, how does a yet-unidentified endoderm determinant stay in ascidians at the vegetal pole during the second phase of ooplasmic movements? Another important difference is that vegetally localized maternal factors are translocated to the future posterior side in ascidians (e.g., postplasmic/PEM RNAs), which is opposite from where the notochord forms, and to the future organizer side in vertebrates (e.g., organizer side-determinants acting upstream of β-catenin), where the notochord arises, despite the similarity in relative allocation of tissue-forming territories in their fate maps. One area that has been largely missing in studies of cell fate specification in ascidians is how the endoderm and mesenchyme cells that are preserved undifferentiated in the tadpole larva become fully differentiated after metamorphosis. Future studies of events during metamorphosis and further analyses of the early embryo will give us insights into common and different mechanisms by which developmental fates are specified in ascidians and vertebrates, and even other organisms. Eventually, we might be able to explain generally how the basic chordate body plan is generated.
One key feature of ascidians is that ascidian embryogenesis involves a small number of cells and tissue types and a basic, nonredundant set of developmental genes. Because of this finding, it is possible to understand the mechanisms of cell fate specification in every tissue type of the embryo, and to interpret each tissue not only individually but also as an integrated part of the whole. Once we understand every single event and the set of all events, we might be able to understand development not only at the molecular and cellular levels, but also at the individual level. One such attempt has already begun, in which comprehensive in situ hybridization analysis and gene disruption assays (Imai et al.,2004,2006) describe gene networks consisting of more than 3,000 epistatic relationships of gene expression profile in the early embryo. The combination of this kind of genome-wide analysis with the simple features of ascidians will be a powerful tool to better explain cell fate specification in chordates.