The molecular mechanisms controlling inductive events leading to the specification and terminal differentiation of cardiomyocytes are still largely unknown. We have investigated the role of Cripto, an EGF-CFC factor, in the earliest stages of cardiomyogenesis. We find that both the timing of initiation and the duration of Cripto signaling are crucial for priming differentiation of embryonic stem (ES) cells into cardiomyocytes, indicating that Cripto acts early to determine the cardiac fate. Furthermore, we show that failure to activate Cripto signaling in this early window of time results in a direct conversion of ES cells into a neural fate. Moreover, the induction of Cripto activates the Smad2 pathway, and overexpression of activated forms of type I receptor ActRIB compensates for the lack of Cripto signaling in promoting cardiomyogenesis. Finally, we show that Nodal antagonists inhibit Cripto-regulated cardiomyocyte induction and differentiation in ES cells. All together our findings provide evidence for a novel role of the Nodal/Cripto/Alk4 pathway in this process.
Specification of vertebrate cardiac mesoderm is the first developmental step leading to heart formation occurring concurrently with gastrulation (Fishman and Chien, 1997). The formation of a vertebrate heart from an amorphous field of precursor cells takes place through multiple developmental steps, such as determination of the cardiac field in the mesoderm, differentiation of cardiac precursor cells into cardiomyocytes, and heart morphogenesis (Olson and Srivastava, 1996). No other early differentiation event is more vital for survival; it must occur in a short time to provide a working circulation system for the rapidly growing embryo (Rosenthal and Xavier-Neto, 2000). Precursor cells, destined for the heart, originate in the lateral epiblast and migrate through the primitive streak to emerge as cardiogenic mesoderm already specified as distinct cardiac lineages. Bilateral fields of cardiogenic mesoderm continue to migrate rostrally and eventually join at their anterior boundaries to form the cardiac crescent, where they commit to cardiac fate (Rosenthal and Xavier-Neto, 2000). The understanding of early cardiogenesis is of particular interest as cardiomyocyte loss in mammals is largely irreversible and frequently underlies the diminished cardiac function associated with heart diseases (Gepstein, 2002). Recently, some of the secreted factors required for cardiogenesis have been identified (McFadden and Olson, 2002). Members of the bone morphogenetic protein (BMP), Wnt, and EGF-CFC families have been implicated in vertebrate myocardial development. Zebrafish with mutations in the BMP2 homologue gene swirl and one-eyed pinhead (the zebrafish member of the vertebrate EGF-CFC family) exhibit severe defects in myocardial differentiation and reduced expression of two early markers of the myocardial precursors Nkx2.5 and GATA5 (Reiter et al., 2001). Results obtained in Xenopus and chick indicate that BMP signals from the endoderm induce cardiomyocyte fate, whereas Wnt-mediated signals from the underlying neural tube and notochord suppress cardiomyocyte specification (Schultheiss et al., 1997; Marvin et al., 2001; Tzahor and Lassar, 2001). It has been hypothesized that cardiac muscle cell specification is likely to depend on the location and duration of signals governing more general developmental decisions in the early embryo (Rosenthal and Xavier-Neto, 2000). In this scenario, the mouse cripto gene, the founding member of the EGF-CFC family, appeared to have a crucial role. In mouse embryos, the cripto expression profile is associated with the developing heart structures and is detected first in the precardiac mesoderm (Dono et al., 1993). Later on, at 8.5 dpc, cripto expression is found in the ventriculus, before being specifically restricted, at 9.5 dpc, to the truncus arteriosus of the developing heart (Dono et al., 1993). Notably, mouse cripto mutants exhibit defects in myocardial development, as evidenced by the absence of expression of terminal myocardial differentiation genes such as α-myosin heavy chain (αMHC) and myosin light chain 2v (MLC2v) (Ding et al., 1998; Xu et al., 1999). Accordingly, by using embryoid bodies (EBs) derived from Cripto−/− ES cells, it has been shown that cripto is essential for cardiomyocyte induction and differentiation (Xu et al., 1998). However, how cripto functions to regulate cardiogenesis is still unknown. To study this process, we took advantage of embryonic stem (ES) cells, which have been widely used as a model system of cardiogenesis, proven to be a powerful tool to study early events of cardiac induction (Doetschman et al., 1993; Monzen et al., 2001, 2002; Boheler et al., 2002). To create a system in which we could manipulate Cripto activity, we developed an assay in which recombinant Cripto protein restored cardiomyocyte differentiation in Cripto−/− ES cells. This approach allowed us to define the dynamics of Cripto signaling required for differentiation of cardiac precursor cells. We showed that Cripto is required in a precise moment during differentiation, after which it fails to specify the cardiac lineage. Moreover, we found that the absence of Cripto signaling in this early acting window of time resulted in a direct conversion of Cripto−/− EB–derived cells into a neural fate. This observation suggests that Cripto inhibits mammalian neuralization and supports the hypothesis that a default model for neural specification is operating in ES cells. Furthermore, we show that Cripto protein activates the Smad2 pathway during cardiomyocyte induction and, moreover, that overexpression of an activated form of type I receptor ActRIB restored the ability of Cripto−/− ES cells to differentiate into cardiomyocytes. Taken together, our results indicate that Cripto participates in heart development, regulating early events that lead to cardiac specification, and highlight a novel role for the Nodal/Cripto/Alk4 pathway in cardiomyogenesis.
Secreted Cripto retains its ability to rescue cardiomyocyte differentiation
Previous data on cultured ES cells lacking cripto have revealed an essential role of cripto for contractile cardiomyocyte formation. Cripto−/− ES cells selectively lose the ability to form beating cardiomyocytes, a process that can be rescued by expression of Cripto (Xu et al., 1998). As Cripto is a GPI-anchored membrane protein, we first determined if a secreted form of Cripto could restore cardiomyocyte differentiation in Cripto−/− ES cells (Fig. 1). To this end, we overexpressed a secreted derivative of Cripto, which lacks the hydrophobic COOH terminus region required for membrane anchorage (Minchiotti et al., 2000), in Cripto−/− ES cells and compared its activity to that of wild-type (wt) Cripto (Fig. 2 A). A pooled population of cells selected for resistance to puromycin were examined for the number of EBs containing beating areas, from day 8 to 12 of the in vitro differentiation assay. Spontaneous rhythmic contractile myocytes were observed in Cripto−/− ES cells expressing either the membrane-anchored or the secreted Cripto protein (Fig. 2 B). Moreover, similar results were obtained by expressing a secreted Cripto protein that lacks the NH2 terminus region (EGF-CFC; Fig. 2, A and B), thus indicating that not only membrane anchorage is dispensable for activity, but that the EGF-CFC domain alone is sufficient for Cripto activity in the cardiogenic induction. Previous data have shown that a refolded synthetic Cripto peptide containing the EGF-like domain had mitogenic and branching activity for mammary cell lines (Salomon et al., 1999). We thus went on to define whether the Cripto EGF-like domain alone was able to induce cardiogenesis similar to the EGF-CFC peptide. Two cripto cDNA deletion derivatives encoding either the EGF-like domain retaining the NH2 terminus region (EGF long) or just the EGF domain (EGF short; Fig. 2 A) were generated. No beating areas were observed in EBs derived from Cripto−/− ES cells expressing either EGF long or EGF short peptide (Fig. 2 B), thus indicating that both EGF and CFC domains of Cripto are essential for cardiogenic induction. Western blot analysis showed that the EGF long and EGF short peptides were produced and secreted as efficiently as EGF-CFC (Fig. 2 C and not depicted), thus demonstrating that their inability to rescue the mutant phenotype was not due to a difference in protein expression level. To support the morphological data observed, we examined the expression levels of the cardiac-specific αMHC and MLC2v, two major contractile proteins of cardiomyocytes. As expected, expression of both the αMHC and MLC2v genes was induced in wt ES cells but not in Cripto−/− cells from day 7 of in vitro differentiation (Fig. 2 D). Importantly, the expression pattern of αMHC and MLC2v genes in wt ES cells was reproduced in Cripto−/− cells expressing either wt Cripto or the secreted derivative, but not in cells expressing either EGF long or EGF short peptides (Fig. 2 E).
Timing and duration of Cripto activity in cardiomyocyte differentiation
To gain further insight into the functional role of Cripto in cardiogenic induction and differentiation, we first examined the timing of Cripto expression during ES cell differentiation. Western blot analysis performed with anti-Cripto antibodies on lysates from both wt and Cripto−/− ES cells revealed that Cripto was detectable as early as day 0 and peaked in expression by day 4 in wt EBs (Fig. 3). Importantly, the transient nature of Cripto accumulation suggested that its activity might be required at a defined step in cardiomyocyte differentiation. The time window of Cripto action could not be adequately investigated by means of transfection assays. Therefore, to directly address this issue, a recombinant soluble Cripto protein was used in which the hydrophobic COOH terminus was replaced by a 6xHis epitope (Cripto-His; Minchiotti et al., 2001). Based on our observation that secreted Cripto protein is able to promote cardiogenesis when expressed in the Cripto−/− ES cells (Fig. 2 B), experiments were performed where Cripto signaling was reconstituted by addition of recombinant secreted Cripto protein directly to the cells (Fig. 4). Addition of Cripto during the 0–2-d interval effectively restored the differentiation ability of Cripto−/− ES cells. Addition at later time points resulted in dramatically reduced cardiomyocyte differentiation (Fig. 4 A). Comparable results were obtained with two independent Cripto−/− ES clones (DE7 and DE14; Xu et al., 1998), thus excluding any phenotype difference due to clonal variation (Fig. 4 A). All together, these data indicated that stimulation in trans with soluble Cripto protein was fully efficient in promoting cardiomyocyte induction and differentiation and, more interestingly, defined exactly when Cripto activity was required to promote specification of the cardiac lineage. Furthermore, to define the optimal concentration of Cripto required to promote cardiogenesis, increasing amounts of purified recombinant Cripto protein were added directly to the culture medium of 2-d-old Cripto−/− EBs from either DE7 or DE14 cell lines for 24 h (Fig. 4 B). Increasing amounts of recombinant Cripto resulted in enhanced differentiation efficiency (Fig. 4 B), thus indicating that Cripto-mediated cardiogenic induction was dose dependent.
Having shown that the timing and dose of Cripto signaling activation were crucial to promote cardiomyocyte induction and differentiation, we thus went on to define whether the duration of Cripto signaling was crucial for its biological response. 2-d-old EBs from DE7 or DE14 Cripto−/− ES cells were treated with 10 μg/ml of recombinant Cripto for various lengths of time, washed to remove unbound Cripto, and then cultured for the remaining days. An effective Cripto response required a minimum induction of 24 h, while shorter inductions showed markedly reduced activity (Fig. 4 C). Taken together, our data demonstrated that the amount, timing, and duration of Cripto signaling were all crucial factors to achieve cardiogenic induction and differentiation.
Direct conversion of Cripto−/− EB–derived cells into a neural fate
Our observation that initiation of Cripto signaling in an early acting window of time is crucial for priming differentiation of ES cells to cardiac fate prompted us to gain further insight into the functional role of Cripto at an early phase of ES cell differentiation. Interestingly, when Cripto−/− EBs were plated onto an adhesive substrate, a population of cells with a neuron-like morphology was observed that produced a network surrounding the aggregates. This characteristic morphology was never observed either in wt EBs or in Cripto−/− EBs treated with effective doses of Cripto protein. To confirm that those cells were indeed neurons, immunofluorescence analysis was performed on both wt and Cripto−/− EBs, by using antibodies that recognize the neuron-specific form of class βIII tubulin. These antibodies stained clusters of cells in Cripto−/− EBs, revealing the presence of a dense network of neurons (Fig. 5 A). Neurons were detected in 71% of Cripto−/− EBs, whereas βIII-tubulin–positive cells were never detected in both wt EBs and rescued Cripto−/− EBs that, on the contrary, showed extensive areas of MF-20–positive cardiomyocytes (Fig. 5 A). To gain insight into this issue, we used our controlled differentiation assay to modulate Cripto signaling and to eventually score EB-derived cells for either cardiomyocyte or neuron differentiation, by using morphological criteria as well as immunofluorescence analysis. Addition of Cripto protein during the 0–2-d interval rescued, as expected, the cardiac phenotype of Cripto−/− ES cells (Fig. 5 B), but also resulted in a dramatic inhibition of neural differentiation (Fig. 5 B). Conversely, addition of recombinant Cripto at later time points (i.e., 3–6-d interval) resulted in progressive impairment of cardiac differentiation (see previous paragraph and Fig. 5 B) and, at the same time, increased competence of the EB-derived cells to acquire a neural phenotype, resulting in close to 70% of Cripto−/− EBs that show extensive areas of βIII-tubulin–positive cells. All together our results support the hypothesis that Cripto signaling represses neural differentiation in ES cells and, moreover, show that the restricted time window of Cripto signaling required to achieve proper terminal cardiac differentiation of Cripto−/− ES cells correlates with the competence window for those cells to become committed to a neuronal phenotype.
Cripto activates a Smad2 pathway associated with cardiomyocyte differentiation
Findings in mice, Xenopus, and zebrafish point to a strong functional link between the EGF-CFC proteins and TGFβ ligand Nodal (Shen and Schier, 2000; Adamson et al., 2002). Accordingly, recent studies have shown that Cripto can associate with type I receptor ActRIB (Alk4) and can form a complex together with Nodal and type II receptor ActRIIB (Reissmann et al., 2001; Yeo and Whitman, 2001; Bianco et al., 2002; Yan et al., 2002). Activation of Smad proteins by phosphorylation is a universal signal transduction event following activation of Alk receptors. To ask whether Cripto activates the Smad2 pathway during cardiomyocyte induction and differentiation, 2-d-old Cripto−/− EBs were starved in low serum for 3 h and then stimulated with recombinant soluble Cripto protein for 30, 60, or 120 min. Western blot analysis revealed that phosphorylation of Smad2 significantly increased after treatment with recombinant Cripto (Fig. 6). Smad2 phosphorylation was detectable already after 30-min treatment, persisting at comparable levels even after prolonged exposure to Cripto protein. An anti-Smad2/3 antibody applied to the same blot was used to normalize for total amount of protein (Fig. 6). In vitro studies on mammary cell lines have suggested that Cripto is involved in the Ras/Raf/MEK/MAPK pathway (Salomon et al., 1999). When we looked for activation of the MAP kinase ERK by using an anti–phospho-ERK antibody, recombinant Cripto was unable to activate MAP kinase (unpublished data); thus indicating that the Smad2 pathway was selectively activated during cardiomyocyte induction and differentiation induced by Cripto.
To our knowledge, no data are available on the expression profile of all components of the Alk4/ActRIIB/Nodal complex during the differentiation of ES cells; thus, we first measured by RT-PCR the expression of Nodal, Alk4, and ActRIIB in EBs derived from both wt and Cripto−/− ES cells. Nodal, Alk4, and ActRIIB were expressed in all analyzed stages (Fig. 7 A). If Cripto signaling in cardiomyocyte differentiation acts via the Alk4 receptor, overexpression of a constitutively active type I receptor would be expected to compensate for the lack of Cripto signaling in promoting cardiomyocyte differentiation. We overexpressed in Cripto−/− ES cells the wt or constitutively activated form (ca) of either human HA-tagged Alk4 or its zebrafish counterpart Taram-A (Renucci et al., 1996). Type I receptor serine/threonine kinases can be activated in a ligand- and type II receptor–independent manner by replacing an acidic residue for a specific threonine within the juxtamembrane region of the intracellular domain, a segment known to be involved in kinase regulation (Wieser et al., 1995). Overexpression of either Alk4 ca or Taram-A ca partially restored the ability of Cripto−/− ES cells to differentiate into cardiomyocytes (Table I). On the contrary, overexpression of the wt receptors, either Taram-A or Alk4, had no significant activity despite their similar expression levels (Fig. 7 B). Furthermore, addition of recombinant Cripto to Alk4 ca–expressing cells restores cardiomyocyte differentiation close to controls (Table II). In accordance with the morphological data, expression of the αMHC gene was only detected in Cripto−/− ES cells expressing the activated form of the receptors (Fig. 7 C). Altogether, our results provide an intriguing link between Alk4/Smad pathway activation by Cripto and cardiomyocyte development.
Analysis of Cripto mutants identifies key residues both in the EGF and in the CFC domain
We showed above that the EGF-CFC domain is sufficient to promote cardiogenic induction when overexpressed in Cripto−/− ES cells, whereas the EGF domain alone is unable to rescue the biological activity. To determine the contribution of the EGF and CFC domains, single amino acid substitutions were introduced into the cripto cDNA (Fig. 8 A), and the activity of the corresponding mutant proteins was compared with the wt in the cardiomyocyte assay. Although each mutant was expressed at levels comparable to wt Cripto (Fig. 8 B), three of them were completely inactive or showed a strongly reduced activity in our assay (Table III). Similar results were obtained with two independent Cripto−/− ES clones (Table III). To support the observed morphological data, the expression of the αMHC and the MLC2v genes was examined by RT-PCR on total RNA prepared from EBs derived from Cripto−/− ES cells overexpressing Cripto mutant derivatives (Fig. 8 C). Expression of αMHC and MLC2v genes was either absent or reduced in cells overexpressing G71N, F78A, or W107G cripto mutants, whereas it was restored in Cripto−/− cells transfected with wt cripto. Together these data show that critical amino acid residues are located in both EGF and CFC domains, thus indicating the requirement of both domains for Cripto activity in cardiogenic induction.
Recent reports have shown that Cripto is modified by the addition of sugar residues. N-linked glycosylation was shown to affect Cripto biological activity in the zebrafish assay (Minchiotti et al., 2001). More recently, an O-linked fucosylation of Cripto has been reported to be required for Cripto signaling activity in cotransfection assay in mammalian cells (Schiffer et al., 2001; Yan et al., 2002). To assess if posttranslational modifications were required for Cripto activity in cardiogenic induction, two alanine substitutions were generated, corresponding to either the N-glycosylation site (N63I) or the O-linked fucosylation site (T72A). The activities of the corresponding mutant proteins were tested in the differentiation assay and compared with wt Cripto. Based on the percentage of EBs containing beating areas, both mutant proteins had comparable ability in promoting cardiomyocyte differentiation, compared with wt Cripto (Table III), thus suggesting that addition of sugar residues was not strictly required for Cripto activity in ES cells.
However, a role of these modifications in the modulation of Cripto signaling might be masked in our assay due to overexpression of the proteins. To overcome this limitation, we purified a recombinant Cripto T72A mutant protein from conditioned medium of transfected 293 cells, and its activity was compared with the wt Cripto. When used in the cardiomyocyte differentiation assay, the Cripto T72A mutant protein resulted in close to a 30% reduction in the numbers of Cripto−/− EBs displaying beating cardiomyocytes, compared with the wt Cripto (Fig. 9). A similar reduction was observed when using Cripto T72A in the Smad2 phosphorylation assay, indicating that doses higher than those used for wt Cripto were required to achieve equivalent induction (unpublished data).
Nodal antagonists inhibit Cripto activity in cardiomyogenesis
To gain direct evidence that Nodal signaling is indeed required to support Cripto-regulated cardiac induction and differentiation in ES cells, we sought to determine whether inhibition of Nodal signaling might interfere with Cripto ability to promote cardiomyogenesis. To directly address this point, 2-d-old Cripto−/− EBs were treated with increasing amounts of recombinant Cripto (1–10 μg/ml) in a media containing the supernatant collected from a transiently transfected 293T cell line producing Cerberus protein. This multifunctional antagonist inhibits Nodal as well as BMP and Wnt signaling. However, a truncated form of Cerberus, named Cerberus-Short (Cerberus-S), is a specific Nodal antagonist (Piccolo et al., 1999). The presence of either Cerberus or Cerberus-S supernatant resulted in a significant inhibition of Cripto ability to prime cardiomyocyte differentiation compared with control supernatant (Fig. 10 A). Furthermore, we treated Cripto−/− EBs with 10 μg/ml of recombinant Cripto in the presence of increasing amounts of Cerberus-S–containing medium, thus being able to show that inhibition of Cripto activity by Cerberus-S was indeed dose dependent (Fig. 10 B). Finally, as an additional control, we used Cripto−/− ES cells transfected with either Cerberus or Cerberus-S expression vectors, before treatment of the derived EBs with recombinant Cripto. In accord with the results obtained with conditioned media, expression of either Cerberus or Cerberus-S resulted in a significant inhibition of Cripto activity (Table IV). Together, these results show that Cerberus and Cerberus-S can act as effective antagonists of Cripto signaling in ES cell differentiation and provide evidence for a functional role of the Nodal pathway in Cripto-mediated specification of the cardiac lineage.
Role of secreted Cripto as a priming factor for cardiomyogenesis
Cripto is a GPI-anchored protein; however, recent data in zebrafish have shown that Cripto protein could be provided in a soluble form to enable proper Nodal signal propagation (Minchiotti et al., 2001). Moreover, previous data on chimeric mouse embryos established from a combination of wt and Cripto−/− ES cells suggested that Cripto acts nonautonomously during development (Xu et al., 1998). Due to the absence of analysis of cellular genetic markers, and because only late-stage chimeric embryos were analyzed, the cell autonomous or nonautonomous activity of Cripto is an issue that still remains unsolved (Rosa, 2002). Expression of different deletion mutant derivatives of cripto cDNAs in Cripto−/− ES cells led us to demonstrate that secreted Cripto could efficiently induce cardiogenesis and, moreover, that expression of the EGF-CFC domain was sufficient to restore the ability of Cripto−/− ES cells to differentiate into cardiomyocytes. Furthermore, we showed that the Cripto EGF domain is required, but not sufficient, for cardiomyogenesis. In contrast, it was previously demonstrated that a refolded synthetic human Cripto peptide comprising only the EGF-like domain displayed high affinity binding and mitogenic activity on mammary cell lines (Salomon et al., 1999). Considering the high percentage of amino acid identity in the EGF domain between human and mouse Cripto (Dono et al., 1993), this observation strengthened the hypothesis that there may be divergent Cripto signaling pathways, depending both on different Cripto domains and on specific cell types, that remain to be explored in the future.
Dose and temporal regulation of Cripto signaling in the commitment of ES cells to cardiac fate
EB differentiation is currently considered a powerful model system, reproducing many aspects of in vivo tissue formation during the crucial, but less accessible, stages of mammalian embryogenesis, thus providing access to early cell populations that develop in a normal fashion (Keller, 1995). The timing of initiation of Cripto signaling and the strength and duration of the signal are interdependent variables that have not yet been resolved experimentally, mainly due to the complexity of in vivo analysis. Western blot analysis performed on total lysates, prepared at different times of ES cell in vitro differentiation, showed a regulated accumulation of Cripto protein whose expression was restricted to the very beginning of the differentiation program.
We then used soluble Cripto protein on Cripto−/− ES cells to modulate Cripto signaling and measured its effect on cardiomyocyte differentiation. Kinetic experiments performed by adding recombinant Cripto protein directly to the culture medium of Cripto−/− ES cells demonstrated that stimulation in trans with a soluble Cripto peptide was capable of promoting cardiac induction and, strikingly, revealed an early acting window of time in which the initiation of Cripto signaling is crucial for priming differentiation of ES cells to cardiac fate. Having defined the timing of Cripto signaling, we next examined two more variables, dose and duration of signaling. Either different amounts of Cripto protein for a defined length of time or a high dose of Cripto peptide for various lengths of time was used on Cripto−/− EBs. Both protein dose and signaling duration were crucial parameters. Worth noticing, our results show that transient presence of Cripto is inadequate and that sustained Cripto signaling is strictly required to promote cardiogenesis. We thus propose that Cripto activity at a given time, strength of the signal, and length of the time of exposure are critical parameters for the correct specification and differentiation of the cardiac lineage.
Neural cell fate is established from Cripto−/− EB–derived cells in the absence of retinoic acid
Several studies demonstrated that neural differentiation from ES cells relies upon an aggregation step followed by exposure to specific inducing factors, such as retinoic acid (Bain et al., 1995, 1996; Okabe et al., 1996). Our present findings indicate that in the absence of retinoic acid, Cripto−/− EB–derived cells spontaneously differentiate into neurons. Furthermore, we show that the timing of Cripto signaling required for priming differentiation of cardiac cells resembles the competence window of EB-derived cells to acquire a neural character. All together, our results indicate that Cripto signaling is strictly required in an early acting window to negatively regulate neural differentiation and, at the same time, to permit differentiation of ES cells to cardiac fate.
Recent studies have investigated the mechanisms underlying neural specification in uncommitted ES cells (Tropepe et al., 2001). However, the role of the pathways that have been implicated in neural generation in the context of stem cells is still under investigation (Munoz-Sanjuan and Brivanlou, 2002). Although a better understanding of the molecular events that mediate the acquisition of neural fates of ES cells in the absence of Cripto signaling is required, our controlled differentiation paradigm could represent an attractive system to further investigate this issue.
Cripto signaling pathway in cardiomyogenesis
Both genetic and biochemical data indicate a role for Cripto and, more generally, for the EGF-CFC factors in Nodal signaling (Gritsman et al., 1999; Reissmann et al., 2001; Schiffer et al., 2001; Yan et al., 2002). More recently, findings in Xenopus and zebrafish have also shown that the TGFβ Vg1/GDF1-like signals depend on EGF-CFC proteins (Cheng et al., 2003).
We observed a significant increase in Smad2 phosphorylation consequent to treatment of Cripto−/− ES cells with recombinant Cripto protein for different lengths of time. This suggests that Cripto signaling acts through the Smad2 pathway to promote cardiac induction and reveals a potential role of Nodal signaling in cardiogenesis. Acute stimulation by Cripto, although competent in activating Smad2, is insufficient to achieve proper terminal cardiac differentiation, again highlighting the importance of signal duration for cardiomyogenesis.
Activation of Alk4/Tar-A signaling and cardiac induction
One critical function of Cripto during development is to render Activin type I receptor competent for activation by Nodal and to potentiate Nodal-triggered Alk7 signaling activity (Gritsman et al., 1999; Reissmann et al., 2001; Yeo and Whitman, 2001; Yan et al., 2002). Here we show that activated forms of either Alk4 or zebrafish Taram-A partially compensate for the lack of Cripto in the cardiomyocyte differentiation assay, suggesting that, indeed, this receptor family is involved in Cripto-mediated cardiomyogenesis. The incomplete rescue could be due to different reasons. As overexpression experiments do not allow the modulation of receptor signaling both in terms of timing and signal strength, we measured the effect of constitutive, but not transient (acute), activation of Alk4-mediated signaling. Cripto may also interact either with other Alk or TGFβ receptor family members or still unknown molecules to promote cardiomyocyte induction and differentiation in ES cells.
Mutational dissection of Cripto
The mutational dissection of Cripto enabled us to define that both the EGF and the CFC domains are crucial for Cripto activity in cardiogenesis. Remarkably, the biological activities of the Cripto mutants in cardiogenic induction correlate well with their effects on Alk4/Nodal signaling. First, the two amino acids located in the EGF domain whose mutation significantly reduces or completely abolishes Cripto activity, namely G71 and F78, also appeared to be strictly required to rescue cell competence to respond to Nodal signaling in the zebrafish assay (Minchiotti et al., 2001). Interestingly, the impaired activity of mutant Cripto protein was dependent on the amino acids chosen for the substitution. In fact, while substitution of phenylalanine to alanine (F78A) significantly reduced protein activity, a tryptophan in the same position (F78W) preserved Cripto ability to promote cardiogenesis. Worth noting, F78 is fully exposed in the 3D model of Cripto and has been hypothesized to be involved in protein binding (Lohmeyer et al., 1997; Minchiotti et al., 2001). Second, receptor reconstitution experiments in Xenopus have indicated that the EGF domain of Cripto is crucial for Nodal binding to the Alk4/ActRIIB receptor complex (Yeo and Whitman, 2001), while the CFC domain was required for Cripto to interact with the Alk4 receptor. Specifically, either double or triple mutations in the CFC domain, including the amino acid W107, have been reported to impair Alk4-dependent Cripto activity (Yeo and Whitman, 2001; Yan et al., 2002). Here, we show that the single amino acid substitution of residue W107 in the CFC domain severely impairs the ability of Cripto to promote cardiac induction in Cripto−/− ES cells. Finally, several reports have described the modification of Cripto by the addition of sugar residues, including a rare case of fucosylation, suggesting that the activity of Cripto may be controlled by the extent of its glycosylation or fucosylation (for review see Rosa, 2002). Here we show that an alanine substitution in the site of O-fucosylation (T72A; Yan et al., 2002) generates a Cripto mutant protein that is still competent to promote cardiomyocyte differentiation, although showing a reduced activity compared with the wt. Although T72A modification of Cripto has been previously shown to be completely inactive in facilitating Nodal signaling in Xenopus (Schiffer et al., 2001) and in coculture assay (Yan et al., 2002), recent data showed that mutant embryos lacking O-fucosyltransferase do not resemble the cripto knockout phenotype, thus suggesting a less stringent requirement for O-fucose on Cripto activity in vivo than in reporter assay (Shi and Stanley, 2003).
Nodal signaling is required for Cripto-regulated cardiomyogenesis
Results reported herein suggested that Nodal signaling was required for Cripto-regulated cardiac induction and differentiation. To obtain more direct evidence to support this hypothesis, we performed loss-of-function experiments by using Nodal antagonists in our controlled differentiation assay. To this end, either Cerberus or Cerberus-S proteins were used, either by transfecting Cripto−/− ES cells with corresponding expression vectors or by using conditioned media containing the recombinant proteins. In both cases, the presence of either Cerberus or Cerberus-S results in a strong inhibition of Cripto activity in the differentiation assay, thus supporting the idea that Nodal is indeed required to mediate Cripto-dependent cardiomyocyte induction and differentiation of ES cells.
Understanding the early events of lineage segregation during differentiation of mammalian cells is crucial for the prospects of controlling stem cell differentiation for biomedical application. Although ES cells represent a viable source of donor cells for transplantation and gene delivery, the successful use of ES-derived donor cells would require the generation of essentially pure cultures of specific cell types (Boheler et al., 2002). In this respect, our results open new insights into the understanding of the molecular mechanisms by which cripto regulates cardiogenesis, and will hopefully contribute to the characterization of the molecular signals that control both cardiac and neuronal differentiation of stem cells as the first step in the ongoing efforts to employ these cells in regenerative medicine.
Materials And Methods
Plasmids and mutants
The pallino βA vector was derived from the pallino vector (provided by S. Chiocca, European Institute of Oncology, Milan, Italy) by replacing the CMV promoter with the chicken β-actin promoter (pCXN2 vector; Niwa et al., 1991). Restriction sites were blunt ended using Klenow polymerase. All the cripto mutant derivatives were obtained as previously described (Minchiotti et al., 2001). The Cripto-His here renamed “secreted Cripto” and the EGF-CFC derivatives have been previously described (Minchiotti et al., 2001). The cripto EGF long (nucleotide −5 to +288 of cripto cDNA; Dono et al., 1993), cripto EGF short (nucleotide −5 to +75 fused to +157/+288 of cripto cDNA), wt and activated (ca) Alk4, wt and activated (ca) Taram-A, Cerberus, and Cerberus-S cDNAs were all subcloned into pallino βA vector.
Cell cultures and ES differentiation
Human embryonic kidney 293 and 293EBNA cells and undifferentiated ES cells were cultured as previously described (Xu et al., 1999; Minchiotti et al., 2001). For in vitro differentiation, ES cells were cultivated in EBs essentially as previously described (Wobus et al., 1991; Maltsev et al., 1993; Fig. 1). The EBs were plated separately onto gelatin-coated 48-well plates for morphological analysis or onto 100-mm tissue culture plates for RT-PCR and Western blot.
Cell transfections and proteins
Undifferentiated ES cells (107/ml) were electroporated with linearized DNA (30 μg) at 400 V, 250 μF. 1 wk after selection with 2 μg/ml puromycin, resistant clones were pooled, expanded, and subjected to the differentiation assay. Transfection of 293EBNA cells was performed as previously described (Minchiotti et al., 2000).
Either undifferentiated ES cells or EBs were lysed either in a buffer containing 10 mM Tris/Cl, pH 8, 140 mM NaCl, 2 mM EDTA, pH 8, 1% NP-40 or dissolved in Laemmli lysis buffer (Laemmli, 1970) and analyzed by Western blot using the Trans-Blot Semi-dry System (Bio-Rad Laboratories). The anti-HA (12CA5) antibody (ROCHE), anti-Porin 31HL antibody (Calbiochem), anti-Smad2/3, and anti–phospho-Smad2 (Ser465/467) (Upstate Biotechnology) were used according to the manufacturer's instructions.
RNA preparation and RT-PCR
Total RNA from either undifferentiated ES cells or EBs was extracted with TRIzol kit (Life Technologies) according to the manufacturer's instructions and reverse transcribed to cDNA with SuperScript II reverse transcriptase (Life Technologies) and random hexamers (as primers). cDNA samples synthesized from 100 ng of total RNA were subjected to PCR amplification with specific primers. The primers and the PCR conditions used were as follows: Nodal, 5′-TTCCTTCTCAGGTCACGTTTGC-3′ (forward) and 5′-GGTGGGGTTGGTATCGTTTCA-3′ (reverse), annealing temperature 58°C, cycles 35, producing a 518-bp fragment; ALK-4, 5′-AAGGATCCAGGCTCTGCTGTGTGCC-3′ (forward) and 5′-ACGGATCCATGTCCAACCTCTGGCGG-3′ (reverse), annealing temperature 60°C, cycles 30, 411-bp fragment; ActRIIB, 5′-ATGTGCCGTGGTGTCGTGGT-3′ (forward) and 5′-GACCTCCTGATCAGGGATAC-3′ (reverse), annealing temperature 58°C, cycles 30, 54-bp fragment; MLC2v, 5′-GCCAAGAAGCGGATAGAAGGCGGG-3′ (forward) and 5′-CTGTGGTTCAGGGCTCAGTCCTTC-3′ (reverse), annealing temperature 70°C, cycles 33, 490-bp fragment; cardiac αMHC, 5′-GGAAGAGTGAGCGGCGCATCAAGG-3′ (forward) and 5′-CTGCTGGAGAGGTTATTCCTCG-3′ (reverse), annealing temperature 65°C, cycles 30, 301-bp fragment. A set of primers for hypoxanthine phosphoribosyltransferase (HPRT), 5′-CCTGCTGGATTACATTAAAGCACTG-3′ (forward) and 5′-CCTGAAGTACTCATTATAGTCAAGG-3′ (reverse), annealing temperature 58°C, cycles 25, 369-bp fragment, was used as a control.
Immunofluorescence of EBs
Adherent EBs were fixed either in methanol/acetone (7:3; MF-20 antibody) or in 4% paraformaldehyde in PBS (β-tubulin isotype III). EBs were treated with 0.1% Triton X-100 (Sigma-Aldrich), 10% normal goat serum (DakoCytomation) in PBS and incubated with primary antibodies in 10% NGS, 1× PBS at the following working dilutions: anti–neurofilament-M (1:400; Chemicon International), anti–β-tubulin isotype III (1:400; Sigma-Aldrich), and anti–sarcomeric myosin (MF-20, 1:50; monoclonal supernatant obtained from the Developmental Studies Hybridoma Bank, University of Iowa). After washing, EBs were incubated with secondary antibodies, either fluorescein (Boehringer) or rhodamine conjugated (Jackson ImmunoResearch Laboratories), in 10% NGS, 1× PBS. After PBS wash, EBs were counterstained with DAPI and mounted in Vecta Shield medium (Vector Laboratories). Labeling was visualized by epifluorescent illumination using an Axioskop 2 microscope, and images were acquired on an Axiocam ARC camera (Carl Zeiss MicroImaging, Inc.).
We thank Mrs. M. Terracciano and Salvatore Ponticelli for technical assistance, Miss M. D'Agostino for correcting and typing the manuscript, and Frederic Rosa, Daniel Constam, and Stefano Piccolo for their gifts of plasmids. We are grateful to Scott Frank for his thoughtful comments on the manuscript and Drs. Frederic Rosa, Umberto di Porzio, and Mark Mercola for their valuable comments and helpful suggestions. S. Parisi wishes to thank F. Volpicelli for helpful advice and for providing reagents.
This work was supported by grants from the Department of Defense Breast Cancer Research Program, U.S. Army Medical Research and Materiel Command (to E.D. Adamson), the Associazione Italiana Ricerca sul Cancro, and BioGeM s.c.a.r.l. (to M.G. Persico). D. D'Andrea was supported by a fellowship from the Fondazione Italiana Ricerca sul Cancro.
Abbreviations used in this paper: BMP, bone morphogenetic protein; ca, constitutively activated; Cerberus-S, Cerberus-Short; EB, embryoid body; ES, embryonic stem; HPRT, hypoxanthine phosphoribosyltransferase; MHC, myosin heavy chain; MLC, myosin light chain; wt, wild type.