|Home | About | Journals | Submit | Contact Us | Français|
Little is known about the ancient chordates that gave rise to the first vertebrates, but the descendants of other invertebrate chordates extant at the time still flourish in the ocean. These invertebrates include the cephalochordates and tunicates, whose larvae share with vertebrate embryos a common body plan with a central notochord and a dorsal nerve cord. Tunicates are now thought to be the sister group of vertebrates. However, research based on several species of ascidians, a diverse and wide-spread class of tunicates, revealed that the molecular strategies underlying their development appear to diverge greatly from those found in vertebrates. Furthermore, the adult body plan of most tunicates, which arises following an extensive post-larval metamorphosis, shows little resemblance to the body plan of any other chordate. In this review, we compare the developmental strategies of ascidians and vertebrates and argue that the very divergence of these strategies reveals the surprising level of plasticity of the chordate developmental program and is a rich resource to identify core regulatory mechanisms that are evolutionarily conserved in chordates. Further, we propose that the comparative analysis of the architecture of ascidian and vertebrate gene regulatory networks may provide critical insight into the origin of the chordate body plan.
Despite the diversity of their adult forms and of their early embryogenesis, all vertebrates go through a stage of development, called the ‘phylotypic stage’ or ‘pharyngula stage’, in which their embryos display a similar, tadpole-like organization . Several groups of marine invertebrates, including the cephalochordates and the tunicates, display a similar developmental stage and collectively with the vertebrates form the chordate phylum. The cephalochordates , commonly known as amphioxus, retain a prototypical chordate body plan throughout their adult life, and the few known species show relatively little morphological variation. In contrast, diversity is extreme in the tunicates, which number thousands of species, and include the ascidians, as well as several lesser-known groups of marine invertebrates. After a brief tadpole-like larval phase, tunicates undergo metamorphosis to become adults that are either pelagic (salps, doliolids and larvaceans ,) or sessile (ascidians) (Figure 1; ) (Table 1). In the adult form of tunicates, and ascidians in particular, the shared ancestry with vertebrates is difficult to recognize.
Traditional phylogenies, based on morphological characteristics such as the presence of segmented muscle, had quite reasonably placed amphioxus as the sister group to the vertebrates. Recent molecular analyses, however, strongly indicate that cephalochordates are the most basal extant chordate clade, and that the tunicates are the closest invertebrate relatives of the vertebrates [5,6]. These conclusions are consistent with morphological traits such as neurogenic placodes, which are present in vertebrates and ascidian embryos, but apparently not in amphioxus or other invertebrates [7,8]. The status of ascidians as the vertebrate sister group reinforces the conclusion that the sessile body form of adult ascidians is a derived trait and not ancestral to the chordates, as was speculated earlier .
Tunicates have long fascinated researchers, this group features anatomical, embryological and physiological traits shared with amphioxus and vertebrates, which thus are likely to have been present in the earliest chordates. At the same time, many traits differ greatly form those seen in vertebrates. Triggered by advances in molecular and developmental biology, the past twenty years have seen a resurgence in interest in tunicates. New embryological and genomic techniques, applied in particular to ascidians, offer an opportunity to study both types of ascidian traits in order to better understand chordate evolution. This review focuses on the potential of ascidians as model organisms (Box 1) in developmental biology, and on their status as nature’s extreme chordate experiment. We will review recent findings on the developmental biology of ascidians, and argue that the very differences in the developmental strategies used by ascidians and vertebrates may ultimately shed light on the emergence of the chordates.
The solitary ascidian species on which Edward Conklin worked out the ascidian cell-lineage. No longer used in present day labs.
An edible Japanese ascidian, which because of its industrial aquaculture, was simple to obtain in large quantities in Japanese labs. Played a major role in the emergence of ascidian molecular studies in the 1980's. Its eggs being quite large (280μm), this species is very suitable for embryological work including transfer of egg and blastomere cytoplasm. It's long generation time (2–3 years), unsequenced genome and the restriction of its habitat to north-east Asia constitute handicaps.
A large solitary ascidian found mainly in the Atlantic and Mediterranean. As each adult can produce up to several hundreds of thousands of optically clear embryos, P. mammillata is an excellent model system for cell biological work. The small egg size (~120μm) allows whole embryo confocal imaging. It has a rather long generation time (>8 months) and a compact, but not sequenced, genome.
The cosmopolitan solitary ascidian that has become the major model species and for which the most advanced molecular tools have been developed. Its small (~160Mb) genome has been sequenced, and complemented by large scale EST and in situ hybridization projects. DNA constructs can be easily electroporated. Generation time is around 3 months allowing forward genetics approaches. Egg diameter is around 140μm, which still allows some classical embryological work.
A close sister species to Ciona intestinalis, whose genome has also been sequenced, and which can also be electroporated. Embryos have similar characteristics to those of C. intestinalis. Found mainly in the Pacific ocean. Generation time is around 3 months, allowing forward genetics approaches.
A cosmopolitan colonial ascidian, that has been raised in the laboratory for more than 50 years, reproducing both sexually and asexually. B. schlosseri is the major model system for the study of blastogenesis, colony fusion, and regeneration. Some molecular tools are available, though not as many as in Ciona.
Transfer of egg and blastomere cytoplasm, isolation and recombination of identified blastomeres.
Over/mis expression by injection of synthetic mRNA.
Knockdown by injection of antisense morpholino oligonucleotides (all species), or RNAi (B. schlosseri only).
Electroporation of DNA to make transient transgenic animals (Ciona sp. and P. mammillata).
Establishment of transgenic lines by genomic insertion with aid of meganuclease or Minos transposon (Ciona sp.).
Forward genetic screens
Tunicates have evolved sexual and asexual reproductive strategies not found in other chordates. To begin with, almost all tunicates are hermaphrodites. This trait probably arose early during tunicate evolution, perhaps in response their sessile and pelagic lifestyles. During ascidian spawning, eggs and sperm are usually shed simultaneously, and self-fertilization is possible in some but not all species. An even more surprising feature found in compound and colonial ascidians is their ability to bypass the chordate phylotypic stage altogether and produce adults via asexual reproduction by budding — a process called ‘blastogenesis’. Buds form from a small coelomic stem cell population and blastogenesis is accompanied by the apoptosis of parental zooids, such that adult tissue is cyclically replaced on an approximately weekly schedule in the colonial ascidian Botryllus schlosseri . These dual reproductive strategies of ascidians may thus offer an opportunity to study how the same adult structures can be generated via alternative developmental programs.
Most ascidians start life as free-swimming larvae whose sole task is finding an appropriate substrate to attach themselves via adhesive papillae at their rostral end, thus beginning the sessile, feeding phase of their lives. The larval sensory, central nervous, and muscular systems are all specialized for this one task. The two larvae that have been characterized in greatest detail, those of Ciona intestinalis and Halocynthia roretzi, are characteristic of the simpler ascidian larval forms [11, 12], in which the digestive system, heart, filter-feeding structures and blood exist only as undifferentiated rudiments. What is most striking about these larvae is the small number of cells they are composed of: in the case of C. intestinalis or H. roretzi tadpoles only approximately 3000 cells. The notochord consists of 40 cells arranged in a single column that is surrounded in the tail by either 36 (C. intestinalis) or 42 (H. roretzi) muscle cells. The larval nervous system is also very simple with less than 130 neurons and 230 glial cells in Ciona, including centrally- and peripherally-located light, gravity, mechano- and possibly chemoreceptors [13, 14].
Although Ciona and Halocynthia are members of the two major groups of ascidians, the Stolidobranchia and the Phlebobranchia, respectively, they are not representative of all ascidians. The larvae of some species, such as Molgula occulta, have been simplified even further and have lost the ability to swim; they have, in fact, lost their tails entirely, along with most larval sensory organs . Conversely, the larvae of some colonial species are much more complex and their structures more differentiated (Figure 1A). The larva of Ecteinascidia turbinata, for instance, contains several thousand muscle cells. Unfortunately there are currently few studies on these ascidians.
While the ascidian tadpole is vertebrate-like in its overall organization, it is less clear to what extent the similarity extends to the individual cell types found in these larvae. The vacuolated notochord and striated muscle cells appear quite similar to those of vertebrates, and some homology of cell types in the anterior nervous system has also been reported . In contrast, the epidermal cells of tunicates secrete a thick cellulose tunic surrounding the animal. The tunicates are unique among animals in their ability to make cellulose and probably acquired this biosynthetic pathway by horizontal gene transfer [17, 18].
Let us a start a detailed comparison of the ascidian and vertebrate strategies leading to the formation of similar tadpoles by comparing the cleavage patterns and fate maps at early stages of development (Figure 2). Early development of ascidians is marked by rapid and invariant cleavage divisions. The cell lineages are stereotyped and well documented, much as in Caenorhabditis elegans (Figure 2; ). The cell fates of blastomeres in early embryos are highly conserved between the distantly related ascidian genera, Ciona and Halocynthia, and, in contrast to certain nematodes , there is no evidence that any ascidian develops without an invariant lineage. Gastrulation starts with only 110 cells, and at this stage, the cell fates of most blastomeres have already become restricted to a single tissue type.
Figure 2C shows schematic fate maps of Xenopus and ascidians shortly before the onset of gastrulation [19, 21]. The presumptive tissues show a high degree of topographic similarity between the two fate maps, although each territory differs in the relative proportion of the blastula it occupies. In both ascidians and vertebrates, ectoderm is derived from the animal hemisphere, endoderm from the vegetal pole region, and mesoderm from the equator of the embryo. This organization, which is not found in non-chordate deuterostomes, is conserved among ascidians, amphibians  and basal ray-finned fish such as sturgeons , while the amphioxus fate map remains controversial . The fate maps of ascidians and vertebrates also show similarity along a second axis, which runs orthogonal to the animal-vegetal axis. In this article, we will refer to this second axis as the ‘circum-notochord/contra-notochord’ axis to facilitate comparison between (for the discrepancy in definition of embryonic axes between ascidian and amphibian early embryos, see ). The circum-notochord side marks the area from which the notochord forms and which is conventionally referred to as ‘dorsal’ and ‘anterior’ in amphibian and ascidian embryos, respectively. The contra-notochord side is centered on the opposite pole of the embryo. In both ascidian and vertebrate fate maps, the CNS and notochord are derived from the circum-notochord side, and the other mesodermal tissues originate from the contra-notochord side as well as the lateral region of the equator of the embryo. Therefore, in spite of the determinate cell lineage of ascidians, but not vertebrates, both the early fate maps at the onset of gastrulation and the final tadpole body plan are very similar in ascidians and vertebrates (represented by sturgeons and amphibians) .
Many of the cellular and molecular mechanisms that drive development up through the gastrula stage in ascidians have been deciphered, with insight coming from several species, in particular Ciona intestinalis, Halocynthia roretzi, and Phallusia mammillata. When the same process has been dissected in more than one ascidian species, it has been found that the mechanisms are similar.
The presence of maternal activities required for the definition of the animal-vegetal axis in ascidians was first revealed by experimental removal or transfer of egg cytoplasm (Figure 3A; ). These experiments showed that vegetal determinants are highly concentrated at the vegetal pole just after fertilization, and that their distribution expands during the second mitotic cell cycle to the entire vegetal hemisphere. In addition, a maternal animal-determining activity was also mapped to the animal pole . Microarray analysis gave no indication of mRNAs specifically localized to the vegetal or animal pole of ascidian eggs after the first cell cycle , but two maternal proteins were shown to be involved in the establishment of this axis. In response to as yet uncharacterized localized vegetal determinants, β-catenin gradually accumulates during cleavage stages in vegetal nuclei (Figure 3A, bottom), where it is crucial to transcriptionally activate the zygotic vegetal program [29,30]. The involvement of nuclear β-catenin in establishment of the animal-vegetal axis is an ancient mechanism that ascidians share with echinoderms [31, 32], but that has been lost in vertebrates. Conversely, localized maternal vegetal determinants found in vertebrates, such as the transcription factor VegT in Xenopus (Figure 3, ), have no ascidian orthologs. At the opposite pole of the embryo, the Ciona ortholog of the GATA4/5/6 family of vertebrate transcription factors is required for animal development (Figure 3A; ). This protein is ubiquitously distributed but its activity is gradually repressed in the vegetal hemisphere by accumulating nuclear β-catenin. This maternal animalizing role of GATA4/5/6 has so far not been found in any other organism and may thus be an ascidian invention.
During specification of the circum-/contra-notochord axis, different strategies and molecules are employed in ascidians and amphibians (Figure 3B), whereby the mechanisms responsible for the relocation of circum-/contra-notochord determinants appear more conserved than the determinants themselves. In both groups, the sperm entry point breaks the initial radial symmetry of the egg at fertilization and leads to a microtubule-dependent translocation of the determinants from their initial position at the vegetal pole to one side of the embryo before the first cleavage [25, 35]. In both ascidians and amphibians, the localized mRNAs tightly associate with the endoplasmic reticulum at the oocyte cortex [36,37]. However, these vegetally localized determinants (Figure 3B) move towards opposite sides with respect to the future fate maps of ascidian and anamniote eggs: in both Xenopus and zebrafish embryos, they move towards the circum-notochord side and promote nuclear import of cytoplasmic β-catenin [38,39], which results in the activation of axial zygotic target genes such as Siamois/Twin or Goosecoid. By contrast, in ascidians, axial maternal determinants are translocated to the contra-notochord side of the egg and there is no evidence of circum-notochord maternal determinants. Consistently, a comprehensive survey of localized maternal RNAs by microarray in Ciona embryos revealed no circum-notochordal mRNAs, but identified a large number of maternal mRNAs localized to the surface of the cortex of the contra-notochord side of fertilized eggs . These postplasmic/PEM RNAs (Figure 3B) encode a variety of proteins ranging from transcription factors, signaling molecules, RNA binding proteins, such as Vasa, to a glucose transporter . Some of these PEMs play pivotal roles in the early specification of the contra-notochord side. The POPK1 kinase, for instance, is important for the proper translocation of PEM RNAs to the cortex , while the orphan PEM-1 gene product is necessary for the positioning of successive contra-notochord-specific cell division planes . In addition, macho-1, a Zic-family zinc-finger transcription factor, is essential to promote contra-notochord identity and repress circum-notochord cell fates [43, 44]. This comparison shows that, although both ascidians and anamniote vertebrates rely on maternally supplied localized determinants, their identity and mechanism of action as well as their localization are remarkably different.
By the 8-cell stage in ascidian embryos, maternal cues have defined the embryonic axes and specified the four founder lineages. The animal circum-notochordal a-line and contra-notochordal b-line will form the head and tail ectoderm, respectively (Figure 2B), while the vegetal circum-notochordal A-line and contra-notochordal B-line will give rise to the mesodermal and endodermal derivatives as well as to the posterior nervous system. These founder lineages are further patterned during cleavage stages, such that by the onset of gastrulation the fates of most individual blastomeres are restricted to a single tissue type.
Ascidian embryogenesis has long been regarded as a typical example of ‘mosaic development’, in which a highly stereotyped pattern of cell divisions partitions distinct sets of localized maternal determinants to the different blastomeres. This is indeed what happens for a few tissues, such as the epidermis and primary muscle lineage. Local cell interactions, however, also play a major role in cell fate decisions, particularly during the specification of the mesodermal and neural blastomeres located close to the equator of the embryo (Figure 2B). Three major pathways have so far been shown to be involved in the early embryonic patterning of ascidians: the FGF, Nodal and BMP pathways.
At first sight, the role of FGF in ascidians and vertebrates appears relatively similar. As in vertebrates , the FGF/ras/ERK pathway has an important function in the induction of ascidian mesoderm and anterior neural tissue. It is required from the late 32-cell stage onwards for notochord, mesenchyme, tail-tip muscle and anterior nervous system development (Figure 2) [25, 46]. The major FGF-family member acting in these tissues is FGF9/16/20, which is transcriptionally activated directly or indirectly by β-catenin in most vegetal blastomeres . The enhancers conferring FGF responsiveness to several direct transcriptional targets of FGF have been analyzed in ascidians [49–51]. These studies revealed that the activation of distinct transcriptional targets of FGF signaling results from the synergy between a ubiquitous maternal factor, ETS1/2, and tissue-specific transcription factors, such as FoxA and Zic (notochord), macho-1 (mesenchyme), and GATA4/5/6 (anterior neural plate). A majority of the transcription factors cooperating with ETS1/2 in ascidians do not appear to act in the FGF pathway of vertebrates, and thus the cis-regulatory logic acting downstream of FGF in ascidians and vertebrates may differ.
In addition, the range of action of FGF may also differ between vertebrates, where fields of cells are induced, and ascidians, where mainly single cells are induced. This short range of action requires strategies for restricting the response to the signal to selected cells (Figure 4). In the notochord, mesenchyme and trunk lateral cell lineages, the mother cell is polarized by the inducing signal(s), and cleaves asymmetrically giving rise to an induced and an uninduced daughter [52–54]. In the anterior neural plate, by contrast, the mother cell cleaves symmetrically to give two equi-competent daughter cells that establish different degrees of contact with the inducing cells . Only the daughter with the largest area of contact is induced. It is currently unclear whether similar control mechanisms are also acting in vertebrates.
Unlike FGF, the roles of Nodal and BMP differ significantly between ascidians and vertebrates. Nodal is the main inducer of mesodermal and endodermal cell fates in vertebrate embryos . In Ascidians, nodal transcription is transiently activated in all vegetal cells during cleavage stages. However, Nodal plays no role in endoderm induction in ascidians and is only involved in the formation of a small minority of the mesoderm derivatives [57, 58]. (Figure 2B). Towards the end of the cleavage stages of ascidian embryos, Nodal also patterns the neural plate by promoting lateral neural fates at the expense of medial ones [59, 60] — an activity opposite to what is found in vertebrates at later developmental stages.
Also BMP signaling is used differently in the two groups. In vertebrates, repression of BMP signaling is necessary for notochord and neural plate formation , while in ascidians BMP signaling promotes rather than antagonizes notochord formation and has no major effect on the size of the neural plate [61, 62]. These results are particularly surprising because Nodal and BMP pathways are central to the formation and function of the vertebrate organizer, a transient structure crucial for the formation of the circum/contra-notochord axis. The different, sometimes even opposite, functions of these pathways in vertebrates and ascidians suggest that ascidians may lack a structure homologous to the vertebrate organizer. Consistently, the expression patterns of most organizer genes are not conserved in ascidian embryo genes [62, 63]. Although functional evidence for the existence of a organizer in cephalochordates is not firmly established, the expression patterns of organizer genes are overall much more conserved in cephalochordates, suggesting that this structure may have been present in the common chordate ancestor and was lost in ascidians . In any case, ascidian studies demonstrate that the presence of an organizer is not required for the formation of a tadpole body plan, a very unanticipated result given the importance placed on the organizer in models of chordate evolution .
Thus, despite the similarity of topography in ascidian and vertebrate fate maps, the early developmental logic is quite different between the two. Looking within the vertebrate clade, this trend is also observed, though to a lesser extent. Early fish, chick, amphibian and mammalian embryos are morphologically very different, and the maternal information that leads to their patterning differs largely — in fact, mammalian embryos are thought to lack localized maternal determinants. Yet, the strategies converge shortly before gastrulation, and all vertebrate embryos have an organizer equivalent that expresses similar molecules.
Between the gastrula and tadpole stages, the body plans of ascidians and vertebrates are morphologically most similar. One might thus expect the strongest conservation of developmental strategies among chordates during this period. Indeed, at the morphogenetic level, gastrulation, particularly on the circum-notochord side, is similar between ascidians and vertebrates , and ascidian neurulation, as in most vertebrates, proceeds by the folding of the neural plate. During the tailbud stages, the tail elongates markedly in both groups. In vertebrates, this elongation is due to the combined action of a posterior growth zone at the tip of the elongating axis, and of convergence and extension movements in axial and paraxial mesoderm along the axis. Ascidians lack a posterior growth zone, and axis elongation occurs with minimal cell division, but is initially driven by convergence and extension movements of notochord cells . Subsequently, the notochord stiffens and its volume increases dramatically as extracellular material accumulates between notochord cells . The identification of a mutant in Ciona savignyi -for a Planar Cell Polarity (PCP) gene, prickle, revealed that this pathway is involved, as in vertebrate embryos, in ascidian notochord morphogenesis . However, even in the absence of the PCP pathway considerable convergence and elongation of the notochord was observed in Ciona, driven by a presumed boundary effect . Also in contrast to vertebrates , the narrowing and elongation of the caudal neural tube in ascidians is apparently PCP-independent, as Prickle is not involved .
At the level of transcriptional control, the expression patterns of crucial transcription factors are reasonably well conserved between ascidians and vertebrates. It thus seems that while ascidians and vertebrate embryos use different strategies for cell fate specification steps, they may use some similar molecules to trigger terminal differentiation once cell fate has been specified. For example, MyoD, Tbx6 and MEF2 are specifically expressed in muscle; Brachyury in the notochord; Pitx in the presumptive oral region and in the left side epidermis, where it overlaps with Nodal; Otx in the fore and mid-brain, and Gata4/5/6 in the endoderm and heart precursors. It will be interesting to test whether this commonality of expression profiles is underlain by a commonality of cis-regulatory mechanisms. The lack of conservation between vertebrates and ascidians of blocks of non-coding sequence , and the few examples where ascidian cis-regulatory sequences were tested in vertebrates (e.g., ) suggest that the precise cis-regulatory logic of individual genes may not be well conserved. Furthermore, many regulatory genes show a great divergence of expression amongst chordates. This includes in particular the Hox genes, proposed to play a central role in the maintenance of the phylotypic stage in vertebrates . Tunicate Hox genes have lost their genomic clustering and became atomized , temporal colinearity is absent and some spatial colinearity is only detected in the larval central nervous system and the juvenile gut .
A firm conclusion on the evolutionary conservation of developmental mechanisms requires a functional approach. In comparison to early development, few functional studies have addressed the mechanisms that control post-gastrula development. The argument for evolutionary conservation is strongest in the heart, which in ascidians is derived from trunk ventral cells (Figure 2A). These cells are specified during the gastrula stages but their terminal differentiation only occurs after metamorphosis . Recent molecular work allowed a direct comparison of the Ciona, Drosophila and vertebrate gene regulatory networks [77, 78]. In all three organisms, formation of the heart involves a highly conserved subnetwork, or kernel , including Gata4/5/6, Hand, and nkx2.5 orthologs . bHLH transcription factors of the Mesp family act upstream of this kernel in ascidians and vertebrates, but not in Drosophila. The Mesp network thus defines a chordate-specific regulatory program. Less detailed evidence exists for the molecular conservation of other processes of chordate development including the role of the BMP pathway in the definition of the dorsal neural tube identity  and ventral median fin .
These examples, however sparse, give hope that the morphogenetic and transcriptional mechanisms acting at post-gastrula stages in ascidians may be more comparable to those of vertebrates than cleavage-stage ones. It should, however, be remembered that the tadpole stage is not a prerequisite to form an adult tunicate. We saw above that asexually reproducing tunicates bypass this stage. In addition, most pelagic tunicates, including doliolids, salps and pyrosomes, have a different larval body plan, suggesting that an adult pelagic life style decreases the selective pressure to keep the tadpole stage .
After a brief swimming period, less than one day in Ciona, the ascidian larva attaches to a substrate via adhesive papillae or palps, located at its rostral end. This initiates a series of poorly characterized events, involving EGF-repeat containing signaling proteins, like Hemps, and their targets [81–84]. The structures necessary for larval locomotion and sensation, including the notochord, muscle, and most of the nervous system degenerate in a wave of programmed cell death involving the NO and ERK/JNK kinases [85–88]. In parallel, the visceral organs undergo a 90° rotation —which is central to the development of the adult ascidian body form and filter-feeding ability— and adult organs and tissues rapidly develop. In the simple larvae of C. intestinalis and H. roretzi, adult organ formation is initiated at metamorphosis from undifferentiated or developmentally arrested rudiments: the adult branchial sac and digestive organs derive from larval endoderm, the body wall muscle originates from the trunk lateral and ventral cells and the heart from the trunk ventral cells (Figure 5; [89, 90]). Analysis of a cellulose synthase mutant revealed that adult organ formation can occur independently of tail retraction . Little is known at present on the mechanisms that arrest and re-initiate adult organ development.
Fate map analyses identified the larval rudiments that give rise to the main adult organs [89, 90, 92], giving the opportunity to compare in ascidians and vertebrates the adult structures generated by homologous larval territory (Figure 5). If one focuses on the minority of larval territories that escape cell death, several scenarios areobserved. In some cases, there is a continuity of conservation of the developmental program through metamorphosis: the larval heart precursors in vertebrates and ascidians express similar developmental programs, and give rise to homologous adult organs. The larval stomodeum, or mouth precursor, gives rise to adult structures that are morphologically very different, but perform similar functions: the vertebrate mouth and ascidian oral siphon . Functional conservation is however not observed when comparing the adult structures derived from the evolutionary related ascidian atrial and vertebrate otic placodes [94–97]. The otic placodes contribute vertebrate inner ear structures, while the atrial placodes give rise to the adult ascidian atrial siphon, which, although it includes sensory cilia, is not involved in hearing. Instead, it provides an exit for feces, gametes and seawater that passes through the pharynx and gill slits. Thus, homologous structures at the pharyngula stage can give rise to organs that fulfill radically different functions in the adult.
Until the resurgence of ascidians as subjects of modern developmental biology about 10 years ago, most studies on chordate development focused on vertebrate embryos. These studies led to the concept of the developmental ‘hourglass’ or ‘egg timer’ [98, 99]: although significant differences are observed during very early and late developmental stages, a central period extending from the early gastrula to the onset of organogenesis and including the phylotypic stage is very well conserved during vertebrate development. This period corresponds to the action of the organizer, and to the period during which temporal and spatial colinearity of Hox gene expression is established, two events crucial for the formation of vertebrate body plan at the phylotypic stage. These vertebrate rules were generally considered to also apply to invertebrate chordates.
One of the main lessons from ascidian work is that this view needs to be revisited, and that the strategies leading to a tadpole-like body plan appear more diverse than expected. The previous sections illustrated that it may still be possible to define an hourglass model for the whole chordate phylum, but the absence of an organizer and the much reduced spatial and temporal colinearity observed for Hox genes, suggest that other events may cause the stability of the phylotypic stage. The challenge ahead is to identify these causes.
The stability of the chordate body plan could result from a combination of external and internal constraints. During the Cambrian, this body plan was probably invented to facilitate swimming. This environmental constraint is still active in most modern ascidians, cephalochordates and anamniote vertebrates. In some cases, this initial external constraint may have been replaced by a different one applying a similar pressure on body organization. Appendicularians, for instance, retain the chordate body plan throughout their life but their beating tail is utilized to generate a water current to bring food to their mouth, not to swim . There are also instances where the body plan has been retained although external constraints appear to be much reduced. This includes most notably the amniotes that have lost the requirement for larval or adult swimming but have retained the ancestral chordate body plan for more than 300 million years, although the relative size of the notochord, an essential organ for larval swimming, has been greatly reduced. In this case, internal developmental constraints must have, in part at least, relayed external constraints to ensure the stability of the body plan.
Where could these internal developmental constraints come from? While the stability of the body plan requires a conservation of the major cell types, it is their spatial organization that defines the shape of the organism. A coordinated displacement of the major cell types during and after gastrulation is needed to achieve the precise geometrical organization of tissues that define the body plan. Indeed, explants containing a single fate-determined tissue do not usually adopt the proper geometry when cultured in isolation. In contrast, combinations of tissues can recapitulate, to some extent at least, normal morphogenetic processes . The fact that embryos of ascidians and many vertebrates share similar fate maps at the onset of gastrulation and undergo similar gastrulation movements may thus be more important for the conservation of the body plan than the precise molecular cascades that specify individual cell fates.
It has been strongly argued that changes in the architecture of gene regulatory networks constitute a major drive for the evolution of animal form . These networks are modular and can be broken down into subnetworks that regulate distinct developmental processes. The extent of evolutionary conservation of individual subnetworks is highly variable. Only a small minority of subnetworksare conserved over long periods of evolutionary time. The bilaterian heart, echinoderm endoderm or eye subnetworks constitute examples of such conserved kernels [78, 102, 103]. The architectures of these kernels share two features: Signaling proteins are excluded from them so that the kernels exclusively link transcription factors, and each gene receives multiple regulatory inputs from other genes of the kernel. Because of this recursive wiring, affecting the function of a single element is sufficient to disrupt the function of the whole kernel and, in consequence, of the process it controls. Although kernels were initially proposed to underlie the conservation of body organization at the phylum level , it should be noted that the small number of kernels identified so far all control cell fates, rather than morphogenesis. Because of the importance of cell–cell interactions to coordinate the morphogenesis of adjacent tissues, morphogenesis kernels may include signaling proteins as well as transcription factors.
A possible explanation for the conservation of the chordate body plan could thus be that an initially strong environmental pressure for more efficient swimming led to the emergence of robust networks of cross-regulating genes controlling body plan organization. The recursive wiring of these networks in turn prevented small adaptive changes, which would in effect generally “lock” the body plan even when environmental constraints are reduced (e.g., as in amniotes), while allowing rare drastic changes such as those observed in certain ascidians  and thaliaceans that have lost their larval tail and locomotory apparatus altogether. If this scenario were true, the search for recursively wired regions in the ascidian developmental gene regulatory network should point to the limited number of kernels conserved throughout the chordate phylum and thus likely to be crucial for the stability of the chordate body plan.
Ascidians are particularly well suited for the reconstruction of gene regulatory networks and the analysis of their cellular outputs (Box 1) . In contrast to vertebrates, their genome is not duplicated and is remarkably compact, with an intergenic distance of only 5 kb on average . The recent sequencing of two genomes in the Ciona genus makes phylogenetic footprinting an efficient way to identify cis-regulatory modules, which can be rapidly tested by electroporation . Finally, gene knock down can be performed by injection of morpholino antisense oligonucleotides into eggs. These features were recently put to use to reconstruct the first metazoan whole embryo gene regulatory network up to gastrula stages . This network is being extended by integrating all published data and currently counts over 120 nodes and 360 edges. Consistent with most examples provided in this review, a majority of edges are not conserved with vertebrates. This network is likely, however, still very incomplete and only extends up to the gastrula stage, the period of development that is least conserved. One of the challenges ahead is to reconstruct the Ciona network, up to the more conserved tadpole stages, to an extent sufficient for meaningful mathematical analysis. This will involve the large-scale identification of the cis-regulatory elements driving the expression of most transcription factors, complemented by the identification by techniques such as chromatin immunoprecipitation of a large fraction of the target genes of these transcription factors.
In summary, while the morphologies of adult vertebrates and ascidians differ greatly, they develop through a core conserved larval body plan common to nearly all chordates. It appears, however, to be naive to expect that the conservation at the morphological level is necessarily reflected in an equally strong conservation of molecular mechanisms. Rather, one finds varying degrees of similarity between ascidians and vertebrates (and for that matter, among ascidians and vertebrates themselves). Surprisingly, developmental pathways for structures common between vertebrates and ascidians often involve the same molecules, but used in vastly different ways. Thus recent work in the ascidian are challenging many long-held assumptions concerning the conservation of the link between molecular mechanisms of development and phenotype, and the allowable degree of plasticity in regulatory pathways.
A wealth of genome, protein, anatomical, and embryological information, mainly on Ciona intestinalis and Halocynthia roretzi can be accessed from the Tunicate portal (http://www.tunicate-portal.org/).