Development and Evolution of the Ascidian Cardiogenic Mesoderm

Abstract

The heart and other blood pumping organs are close to being universally essential in the animal kingdom. These organs present a large anatomical, morphological, and cellular diversity, which is thought to have arisen by building developmental modules on a conserved core of ancestral heart regulatory units. In this context, studies using the ascidian model system Ciona intestinalis offer a distinctive set of theoretical and experimental advantages, which we herein discuss in details. Development of the heart and related muscles in Ciona has been analyzed with a cellular to subcellular resolution unprecedented in Chordate model systems. Unique derived developmental characters of the cardiogenic mesoderm appear to be shared between Ciona and vertebrates. Notably, accumulating evidence point to an early Chordate origin of the cardiopharyngeal population of mesoderm cells that may have provided the foundation for the emergence of the second heart field in higher vertebrates.

Introduction

Young biologists can only be awestruck upon learning about the intricacy and refinement of the Amniote heart organization and physiology. This complexity arises from a modular developmental program, the alterations of which lead to congenital heart disease in about 1% of the new born humans (reviewed, e.g., in Buckingham et al., 2005; Vincent and Buckingham, 2010). The four-chambered heart that permits directional blood flow within independent systemic and pulmonary circulations exists only in Amniotes; however, anatomical innovations that accompanied the diversification of the heart in higher vertebrates elaborated on a preexisting chordate developmental program (Koshiba-Takeuchi et al., 2009). Indeed, the core gene regulatory program for heart development is conserved throughout the animal kingdom, and studies using the Drosophila model system have provided novel insights into the molecular bases of heart development in all animals including humans (Olson, 2006; Reim and Frasch, 2010).

More recently, a Tunicate, the ascidian Ciona intestinalis has emerged as a simple model system to study the cellular and molecular basis of early heart development in Chordates (Davidson, 2007). Molecular phylogenomic studies have shown that Tunicates, including ascidians, are the closest living relatives to vertebrates (Delsuc et al., 2006; Putnam et al., 2008). As such, ascidian larvae display the features of a typical chordate body plan: they possess a notochord that is restricted to the tail (hence the name Urochordates), flanked dorsally by a hollow central nervous system that forms by neurulation, laterally by skeletal muscle cells, and ventrally by the endoderm; during metamorphosis, the anterior endoderm forms pharyngeal gills slits. The ascidian heart does not function before metamorphosis, and in adults, it consists of a simple U-shaped tube that beats peristaltically in a reversible orientation to pump blood through an open circulatory system (reviewed in Davidson, 2007).

In spite of sharing chordate synapomorphies, ascidians have changed markedly since their divergence from the vertebrate lineage. One striking feature is the ascidian-specific metamorphosis that profoundly affects the body plan and transforms a swimming nonfeeding chordate larva into a sessile filter--feeding adult. Another phylum-specific feature is a drastic reduction of the cell numbers of the early embryo, which is accompanied by fixed cleavage patterns that are very stable over long evolutionary distances among ascidians. Ascidians initiate gastrulation with only about 110 cells, they develop rapidly, hatch, and form swimming larvae with only about 2600 cells. This cellular simplicity allows developmental studies to be carried out with cellular to subcellular resolution (Hotta et al., 2007; Tassy et al., 2006). Lineage tracing showed that a single pair of blastomeres, called B7.5 in the pregastrula 64-cell stage embryo, gives birth to the rudiment of the heart in the juvenile (Hirano and Nishida, 1997; Satou et al., 2004). By the time of metamorphosis, descendents of the B7.5 lineage have given rise to anterior tail muscles (ATM), atrial siphon muscles (ASM), longitudinal body wall muscles (LoM), and the heart rudiment (Hirano and Nishida, 1997; Satou et al., 2004; Stolfi et al., 2010).

The recent emergence of molecular tools and resources to observe and manipulate gene function in vivo in ascidians has sparked a renewed interest for ascidian developmental biology (Satoh, 2003). The genome sequence of C. intestinalis and its sibling species Ciona savignyi were published in the early 2000s (Dehal et al., 2002; Hill et al., 2008; Small et al., 2007; Vinson et al., 2005); comprehensive annotations were compiled using large amounts of express sequence tags (ESTs) and state-of-the-art databases created to provide user-friendly access to this wealth of sequence data (Satou et al., 2002a, Satou et al., 2002b, Satou et al., 2005, Satou et al., 2006; Tassy et al., 2010). Gene expression and function could also be analyzed using whole mount in situ hybridization, microinjection of antisense oligonucleotides, and electroporation of synthetic plasmids to drive expression of recombinant proteins in the cells of interest (Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d; Corbo et al., 1997; Satou et al., 2001a, Satou et al., 2001b, Satou et al., 2001c). Whole genome expression and tiling arrays, fluorescence-activated cell sorting and chromatin immunoprecipitation techniques are now being used to investigate cell- or lineage-specific developmental gene regulatory networks (Azumi et al., 2003; Christiaen et al., 2008, Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d; Kubo et al., 2010).

Here, we present an in-depth discussion of our current understanding of the cellular and molecular mechanisms that determine the heart versus muscle precursors fate specification and that control the behavior of heart progenitor cells within the cardiogenic mesodermal lineage during ascidian embryogenesis.

Lineage studies in the solitary ascidians Halocynthia roretzi and C. savignyi showed that the heart is derived from the B7.5 blastomeres in the 64-cell embryo (Hirano and Nishida, 1997; Satou et al., 2004). These cells also give rise to anterior tail muscles (ATMs), atrial siphon muscles (ASMs), and LoMs (Hirano and Nishida, 1997; Stolfi et al., 2010). Large-scale in situ hybridization screens identified hundreds of genes showing expression in the B7.5 blastomeres but only Mesp is expressed exclusively in the B7.5 blastomeres of early gastrula-stage embryos (Fujiwara et al., 2002; Imai et al., 2004; Miwata et al., 2006; Tassy et al., 2010). Mesp encodes a bHLH transcription factor orthologous to vertebrate Mesp1, Mesp2, and pMesogenin1 (Satou et al., 2003, Satou et al., 2004). In the mouse, Mesp1 is required in the early embryo for subsequent heart development (Kitajima et al., 2000; Saga et al., 1999, Saga et al., 2000). In Ciona, Mesp knockdown using antisense morpholino oligonucleotides (morpholino, or MOs) inhibited heart development and the B7.5 blastomeres gave birth exclusively to anterior tail muscles (Satou et al., 2004). These data showed that Mesp is required for heart but not for primary muscle development in the B7.5 lineage.

Mesp is required for heart development but does not seem sufficient since half of the B7.5 progeny gives birth to anterior primary muscle cells (it has not been determined whether ectopic expression of Mesp in the ascidian embryo would be sufficient to promote mislocalized formation of heart tissue in different embryonic lineages). These data suggest that Mesp confers the competence to form heart tissue (see Section 2 for a discussion of the heart inducer(s) in ascidians). Therefore, in ascidian embryos, the competence to form heart tissue is restricted to a single pair of blastomeres as early as the 110-cell stage.

The initial study of Mesp function and regulation revealed that both β-catenin and the muscle determinant macho-1, the main maternal factors that specify cell fates in the vegetal hemisphere, are required for Mesp activation in the B7.5 blastomeres (Figs. 4.1D, 4.2; Satou et al., 2004). Previous studies revealed that the effects of macho-1, which is encoded by a localized mRNA that is excluded from B7.5 cells following asymmetric divisions (Fig. 4.2), are mediated by some of the macho-1 primary targets, the T-box family transcription factors Tbx6b and -c (Yagi et al., 2005). In keeping with a prevalent role for Tbx6 downstream of macho-1, a minimal B7.5 enhancer of Mesp was shown to contain three candidate Tbx6 binding sites termed B to D, which were all able to bind a Tbx6c recombinant protein in electromobility shift assays (Davidson et al., 2005). Sites B and C were also required for reporter gene expression in transient transfection assays (Davidson et al., 2005). These data strongly suggest that Tbx6b and/or -c function downstream of macho-1 to activate Mesp in the B7.5 blastomeres.

Tbx6b transcripts are detected, at least transiently, in all of the B-line mesenchyme and primary muscle precursors starting in 16-cell embryos and until the early gastrula stage (Fig. 4.2A; Imai et al., 2004; Takatori et al., 2004; Yagi et al., 2005). Therefore, transcriptional activation by Tbx6b/c alone is not sufficient to explain the restricted expression of Mesp. Morpholino-mediated knockdown of β-catenin, a key determinant in the vegetal hemisphere, inhibited Mesp expression (Satou et al., 2004). Similarly, knockdown of the β-catenin target Fgf9/16/20 provoked a mild (less than twofold) reduction of Mesp transcripts’ abundance in late gastrula embryos (Imai et al., 2006). Therefore, it has been proposed that Fgf9/16/20 mediates the effects of β-catenin on Mesp expression. Using misexpression of Tbx6b and β-catenin in the animal hemisphere, we observed that the a- and b-line neural cells, which are known to receive an FGF9/16/20 signal (Bertrand et al., 2003), expressed higher levels of ectopic Mesp than the a- and b-epidermal precursors (Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d). This observation is compatible with an activating role of FGF9/16/20 for Mesp expression. However, in these misexpression assays, Tbx6b alone was not sufficient to induce ectopic activation of Mesp, even though the cells presumably perceived the FGF signal; the effects of β-catenin on Mesp expression appeared to be cell-autonomous and the expression pattern of Fgf9/16/20 in early gastrula embryos is inconsistent with an instructive role restricting Mesp expression to the B7.5 blastomeres. For example, FGF9/16/20 induces Twist-like-1 expression in the B7.7 blastomeres, which also express Tbx6b but do not turn on Mesp (Imai et al., 2002, Imai et al., 2003). Thus, it seems that FGF9/16/20 upregulates Mesp expression in B7.5 cells but does not contribute to restricting it to the B7.5 blastomeres.

The LIM homeobox gene Lhx3 is a primary β-catenin target that mediates its effects in the endoderm precursors (Satou et al., 2001a, Satou et al., 2001b, Satou et al., 2001c). Lhx3 is also expressed in B7.5 cells but not in mesenchyme or primary muscle precursors (Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d, Satou et al., 2001a, Satou et al., 2001b, Satou et al., 2001c; Imai et al., 2004). Misexpression of Tbx6b throughout the vegetal hemisphere caused ectopic activation of Mesp primarily in the endoderm precursors (Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d). A combination of misexpression and gene knockdown assays showed that a Tbx6b and Lhx3 synergy activated Mesp specifically in B7.5 blastomeres, which are the only cells that coexpress Lhx3 and Tbx6b in the 110-cell stage embryo, when Mesp expression starts (Fig. 4.2; Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d).

The minimal Mesp enhancer contains putative Lhx3 binding sites that overlap previously identified Tbx6 sites and were required for reporter gene expression in the B7.5 cells (Christiaen et al., 2009a, Christiaen et al., 2009b, Christiaen et al., 2009c, Christiaen et al., 2009d). These data indicate that cooperative DNA binding may contribute to the observed synergy between the two necessary trans-activators of Mesp, thus encoding a developmental logic—the restricted overlap of two necessary activators—in the cis-regulatory DNA of an essential heart specification gene.

Careful examination of the Tbx6b and Lhx3 expression patterns indicates that they may be transiently coexpressed in other blastomeres of the early embryo (Fig. 4.2A). For instance, in 32-cell stage embryos, Lhx3 starts to be expressed in B6.1 where Tbx6b transcripts appear to be fading away following an early expression in B5.1 (the mother cell of B6.1); similarly, in 64-cell stage embryos, Lhx3 starts to be expressed in B7.3 when Tbx6b transcripts are vanishing following expression in B6.2 (the mother of B7.3; Fig. 4.2). Therefore, exquisite temporal regulation of Tbx6b and Lhx3 expression may be required to achieve restricted activation of Mesp in the B7.5 cells. Why would Lhx3 activation be delayed compared to Tbx6b in earlier stage blastomeres?

An essential feature of the molecular and cellular system that regulate temporal gene expression in the posterior vegetal blastomeres of the early ascidian embryo is a generalized repression of transcription in cells that inherit postplasmic mRNAs, including macho-1 and PEM-1 (Nishida and Sawada, 2001; Sardet et al., 2003; Tomioka et al., 2002; Yoshida et al., 1996). PEM-1 has recently been shown to mediate global repression of transcription by inhibiting Serine 2 phosphorylation in the C-terminal domain of RNA polymerase II in both C. intestinalis and H. roretzi (Kumano et al., 2011; Shirae-Kurabayashi et al., 2011). As a result, macho-1 can activate Tbx6b expression following asymmetric division of B4.1 and only in the B5.1 daughter cells, since its sister B5.2 cell inherits the repressive PEM-1-containing postplasm (Fig. 4.2B). It is of note that the asymmetric segregation of macho-1 mRNAs prevents further protein synthesis in B5.1, thus possibly explaining why Tbx6b is only transiently expressed in this blastomere (if the macho-1 protein is rapidly degraded, e.g., Fig. 4.2B).

On the other hand, nuclear localization of β-catenin, which is required for Lhx3 activation, is barely detectable at the 16-cell stage and becomes evident by the late 32-cell stage (Imai et al., 2000; Fig. 4.2). Thus, only the B7.5 blastomeres can turn on simultaneously Tbx6b and Lhx3 at the 64-cell stage presumably because (1) asymmetric division of B6.3 excludes the postplasm thus permitting transient activation of Tbx6b and (2) high levels of β-catenin have accumulated in the nucleus by the time transcriptional repression is released, which permits synchronous activation of Lhx3 (Fig. 4.2B). In summary, the competence to form heart tissue is restricted to the B7.5 cells at the 110-cell stage through a combination of localized maternal factors, and a timed response to maternal factor effectors dependent on progressive segregation of germline transcriptional silencers.

In early gastrula-stage embryos, the B7.5 blastomeres undergo a first round of symmetrical division that gives rise to a pair of cells on each side of the embryo (Fig. 4.1A). These cells have been referred to as the heart founder cells (Cooley et al., 2011). Mesp expression is maintained in the founder cells throughout gastrulation but becomes undetectable by standard in situ hybridization toward the end of neurulation (Satou et al., 2004). Morpholino knockdown combined with systematic qRT-PCR and in situ hybridization analyses suggested that Mesp inhibits its own expression, which could explain why it becomes downregulated toward the end of neurulation (Imai et al., 2006; Satou et al., 2004). In situ hybridization assays also indicate that Lhx3 and Tbx6b expressions rapidly disappear from B7.5 blastomeres during gastrulation, which may also contribute to limiting the duration of Mesp expression (Imai et al., 2004; Kobayashi et al., 2010; Takatori et al., 2004; Yagi et al., 2005). Further studies would be required to determined whether Mesp feeds back negatively on itself and/or its upstream activators and also if Tbx6b and Lhx3 expressions fail to be maintained because their upstream activators are no longer available in B7.5 lineage cells.

During neurulation, pairs of B8.10 and B8.9 founder cells on each side of the embryo undergo an asymmetric division that gives rise to prospective trunk ventral cells (TVCs) and ATMs (Figs. 4.1A and 4.3; Davidson and Levine, 2003; Davidson et al., 2006; Satou et al., 2004). Shortly thereafter, the TVCs migrate to the ventral side of the trunk where part of their progeny forms the heart (see Section 4 for details). When the B7.5 lineage is followed with Di-I in embryos that have been injected with an anti-Mesp MO, no TVCs are observed, and the juvenile heart does not form. Instead, the entire B7.5 lineage remains in the anterior tail, where all cells differentiate into tail muscles (Satou et al., 2004). Further, expression of marker genes is lost specifically in the TVCs after Mesp MO knockdown. These genes include NK4, GATAa, hand, and Notrlc/Hand-like, the Ciona homologs of conserved “heart kernel” regulators—Nkx2.5/tinman, GATA4/5/6/Pannier, hand (Satou et al., 2004)—as well as additional genes, such as FoxF, Bmp2/4, Tolloid, ATP2A1/2/3, and an uncharacterized EST 00152, found to be expressed in TVCs and other tissues by large-scale in situ hybridization screens (Imai et al., 2004, Imai et al., 2006; Satou et al., 2001a, Satou et al., 2001b, Satou et al., 2001c, Satou et al., 2004). These data clearly indicate that Mesp function is required for proper TVC migration and subsequent heart development.

However, the role of Mesp in TVC specification and migration is probably, at least in part, indirect because Mesp is expressed throughout the B7.5 lineage while only the TVCs express heart markers such as NK4, GATAa, and Notrlc/hand-like. Therefore, it is thought that Mesp renders the founder cells competent to form heart tissue. For instance, an FGF signal, mediated by an ERK MAP Kinase and the transcription factor Ets1/2 is required for TVC fate induction (Davidson et al., 2006; see Section 1.1). In this regard, it is possible that Mesp activity determines the TVC-specific response to the FGF-ERK-Ets1/2 pathway. It has been proposed that Mesp directly upregulates Ets1/2 expression in the founder cells (Davidson, 2007), but a direct regulatory connection between Mesp and Ets1/2 has yet to be established experimentally.

Nevertheless, this alleged regulatory connection would not account for all the roles of Mesp in the heart founder cells. First, upregulation and activation of Ets1/2 is not sufficient to induce the TVC fate in the absence of Mesp. Indeed, it is not known whether Ets1/2 is normally expressed and activated in Mesp morphant embryos, but it is clear that the FGF-ERK-Ets1/2 pathway is employed reiteratively in the early ascidian embryo to induce the a- and b-line neural precursors (Bertrand et al., 2003), the notochord (Yasuo and Hudson, 2007), the endoderm (Shi and Levine, 2008), and the various mesenchyme lineages (Miya and Nishida, 2003). In this regard, Mesp probably acts as a B7.5 lineage-specific selector gene to provide TVC-specific response to an otherwise pleiotropic signal.

Second, forced expression of modified versions of Mesp did not mimic the Mesp morphant phenotype. In these studies, the DNA binding basic helix–loop–helix (bHLH) domain of Mesp was fused to the VP16 trans-activation domain and the resulting chimera was overexpressed in the B7.5 lineage cells using the Mesp cis-regulatory DNA (Davidson et al., 2005). Characterization of the Mesp:VP16 expressing cells showed that TVCs failed to migrate in a majority embryos, but these cells always retained or showed ectopic expression of Notrlc/Hand-like. Further, Mesp:VP16 caused the formation of ectopic beating heart tissue in 13% of the analyzed juveniles, showing that heart specification could be uncoupled from TVC migration (Davidson et al., 2005). These observations contrast with the phenotypes observed by targeted expression of a constitutive activator form of Ets1/2, Ets:VP16, which caused ectopic activation of Hand-like and the four B7.5 lineage cells to migrate into the trunk (Davidson et al., 2006). Therefore, the Mesp:VP16 phenotype cannot be explained by upregulation of Ets1/2 and additional roles for Mesp must be invoked.

Raldh2 (retinaldehyde dehydrogenase 2) encodes a key enzyme for retinoic acid synthesis. In Ciona, Raldh2 is expressed specifically in the B7.5 lineage cells during gastrulation and becomes restricted to the ATMs by the tailbud stage (Nagatomo and Fujiwara, 2003; Fig. 4.1C). This B7.5-specific expression pattern opens the possibility that Raldh2 activation requires Mesp function. It is not known whether Raldh2 expression is reduced in Mesp morphants, but it was disrupted in Mesp:VP16 expressing embryos (Davidson et al., 2005). Thus, because Mesp:VP16 is thought to function exclusively as a transcription activator, it appears that wild-type Mesp can have opposite effects on B7.5 lineage gene expression and TVC migration. In keeping with this possibility, whole genome transcription profiling of B7.5 lineage cells expressing Mesp:VP16 showed a significant downregulation of TVC-specific regulators and effectors of cell migration such as GATAa or the small Rho GTPase RhoDF (see Christiaen et al., 2008; Section 3.2).

The simplest hypothesis is that Mesp functions as a dual transcriptional regulator: working as a transcriptional activator upstream of Ets1/2 and TVC markers such as Hand-like or a transcriptional repressor upstream of genes involved more specifically in TVC migration and ATM markers such as Raldh2. It is of note that, in mouse embryonic stem cells, Mesp1 directly binds to and activates conserved cardiogenic regulatory genes such as hand2, Nkx2.5, and GATA4, while it appears to directly repress endodermal (Nodal, Goosecoid, FoxA2) and primitive streak (FGF8, Brachyury/T) markers following inducible overexpression (Bondue et al., 2008).