Serrate and Notch specify cell fates in the heart field by suppressing cardiomyogenesis

The vertebrate heart arises from paired primordia of cardiogenic mesoderm specified during gastrulation, in part by signals from the adjacent endoderm (Sater and Jacobson, 1989, 1990a; Nascone and Mercola, 1995; Schultheiss et al., 1995). While several diffusible factors, including members of the family of bone morphogenetic proteins (BMPs), are known to be involved in specification of the heart field (Lough et al., 1996; Schultheiss et al., 1997), little is understood of the subsequent steps required for the determination of the diverse cell types that arise within the field. Recently, chamber-or tissue-specific transcription factors, such as MEF2 family members, the bHLH proteins eHAND and dHAND, and the homeodomain factor Irx4, have been shown by molecular genetic analyses to be involved in the regional control of contractile protein gene expression within the developing myocardium (Lyons, 1996; Mohun and Sparrow, 1997). Yet, despite these advances, the molecular nature of local cell-cell interactions that define the identities of distinct cardiac tissues have not been described.

In Xenopus, the heart primordia originate in dorsal mesoderm flanking Spemann’s organizer, migrate anteriorly during gastrulation (stage 10-12) to underlie the prospective hindbrain, and then migrate laterally until they fuse at the ventral midline by the neural tube stage. By the early tailbud stage (stage 22), the cardiogenic mesoderm is demarcated by Nkx2.5 expression and forms a crescent extending upwards from the ventral midline towards the neural tube. During late tailbud stages (stage 28-30), the sheet of cardiogenic mesoderm folds. Shortly thereafter (stage 32-33) a linear myocardial tube lined by endothelial cells has formed and is suspended from the roof of the pericardial cavity by the mesocardium (Fig. 1Ap; Fishman and Chien, 1997; Raffin et al., 2000). Recent fate-mapping studies have shown that mesocardial and pericardial roof cells are derived from the dorsolateral portions of the crescent of cardiogenic mesoderm at stage 22 while the myocardium originates from the intervening ventral region (Raffin et al., 2000). Despite their final fate, the dorsolateral portions of the Nkx2.5/heart field form beating heart tubes with lumens when isolated experimentally, either as explants or by heterotopic transplantation (Copenhaver, 1924; Ekman, 1925; Jacobson and Duncan, 1968; Sater and Jacobson, 1990b; Raffin et al., 2000). Thus, the entire field is specified initially as cardiomyogenic. However, between stages 22 and 28, signals from prospective myocardium and neurogenic tissue suppress cardiomyogenesis in the dorsolateral portions of the heart field and thereby subdivide the Nkx2.5/heart field into ventral myogenic and dorsolateral non-myogenic domains (Raffin et al., 2000). Similarly, fate-mapping experiments in chick and zebrafish have suggested that the region with myocardial potency extends beyond the cells fated to form myocardial tissue (Lee et al., 1994; Cohen-Gould and Mikawa, 1996; Serbedzija et al., 1998).

Notch signaling mediates a wide array of cell fate decisions in both invertebrates and vertebrates (Artavanis-Tsakonas et al., 1995, 1999; Weinmaster, 1997, 1998). Notch receptors are single-pass transmembrane proteins of approximately 300 kilodaltons (kDa) containing a series of highly conserved domains in both their extracellular and intracellular regions. Notch orthologues have been isolated from Xenopus, chick, zebrafish, mice and humans (Coffman et al., 1990; Ellisen et al., 1991; Weinmaster et al., 1991, 1992; Franco del Amo et al., 1992; Reaume et al., 1992; Bierkamp and Campos-Ortega, 1993; Lardelli et al., 1994; Williams et al., 1995; Myat et al., 1996; Uyttendaele et al., 1996; Westin and Lardelli, 1997). Both the Delta and Serrate/Jagged families of ligands are also single-pass transmembrane proteins with conserved functional domains. Orthologues of these genes have been identified in numerous species and are often expressed coincidentally with various Notch family members (Bettenhausen et al., 1995; Chitnis et al., 1995; Henrique et al., 1995; Lindsell et al., 1995; Myat et al., 1996; Shawber et al., 1996a; Dunwoodie et al., 1997; Jen et al., 1997; Haddon et al., 1998). Interaction with either Delta or Serrate induces a cleavage within the cytoplasmic domain of Notch and permits the intracellular domain (ICD) to associate with the transcription factor CBF-1/RBP-J/Suppressor of Hairless [Su(H)]. Translocation of this complex to the nucleus then activates downstream genes that include members of the Enhancer of split (HES or ESR in vertebrates) family that encode basic helix-loop-helix (bHLH) transcription factors. Notch activation is complex and responsiveness is also influenced by additional factors including members of the Fringe family and post-translational modifications of Notch (Logeat et al., 1998; Schroeter et al., 1998; Irvine, 1999).

In Drosophila, Notch signaling governs cell fate decisions in all three germ layers (Corbin et al., 1991; Hartenstein et al., 1992). Similarly, Notch signaling in vertebrates influences differentiation of many tissues, including neural ectoderm, somites, cartilage, retina and T-lymphocytes (Austin et al., 1995; Conlon et al., 1995; Chitnis and Kintner, 1996; Robey et al., 1996; de la Pompa et al., 1997; Dorsky et al., 1997; Hrabe de Angelis et al., 1997; Jen et al., 1997, 1999; Wettstein et al., 1997; Wang et al., 1998; Crowe et al., 1999). Although different combinations of ligand, receptor and downstream genes are involved in each of these systems, all involve the initial expression of a Notch receptor and its ligand throughout a field of equivalently specified cells. Resolution of this pattern such that ligand and receptor become expressed preferentially by different cells establishing the basis for the intercellular signaling that eventually subdivides the field into distinct cell types (for review, see Artavanis-Tsakonas et al., 1999). The asymmetry that triggers progression from an equivalence state is poorly understood but, once manifest, is amplified by a feedback loop within each cell such that activation of the Notch receptor results in a decrease in ligand production (see Artavanis-Tsakonas et al., 1999). Studies of the developing nervous system and somites have demonstrated that activation of Notch generally inhibits differentiation and can serve to maintain a precursor population of cells (Austin et al., 1995; Dorsky et al., 1995, 1997; Weinmaster, 1997; Wang et al., 1998). Accordingly, experimental perturbations of Notch signaling can change the fate of cells and/or alter the borders that normally distinguish tissue compartments. Expectedly, mutations in components of the Notch pathway have severe clinical repercussions. For example, translocations of the human Notch1 locus (TAN1) result in leukemia, mutations in human Notch3 result in complex neurological defects, and mutations in human Jagged1/Serrate1 are associated with Alagille syndrome, an autosomal dominant disorder involving multiple organs, including the heart (Ellisen et al., 1991; Joutel et al., 1996; Li et al., 1997; Oda et al., 1997).

Previous studies have shown expression of Notch and Delta homologues in the heart region of chick and zebrafish embryos both before and after formation of a linear heart tube (Myat et al., 1996; Westin and Lardelli, 1997). In addition, null mutations in either Notch1 or RBPJ in the mouse lead to complex defects that include pericardial edema and embryonic lethality (Swiatek et al., 1994; Conlon et al., 1995; Oka et al., 1995). However, it is unclear whether these malformations are caused by aberrant heart field patterning or are secondary to other defects, or both. These observations prompted us to determine if Notch signaling plays a direct role in patterning the vertebrate heart. We report that transcripts of Notch1 (also known as Xotch) and Serrate1 are expressed in the heart field of tailbud stage Xenopus embryos during the period when the dorsolateral portions lose cardiomyogenic potency. We also show that activation or inhibition of Notch signaling through Su(H) leads to opposite effects on the size of the myocardial domain and the ability of individual cells to contribute to myocardium. We propose that signaling through the Serrate-Notch-Su(H) pathway is responsible for the normal loss of myogenic potency in the dorsolateral domains of the Nkx2.5/heart field and is essential for generating distinct cardiomyogenic and non-myogenic domains from an initial field of cardiomyogenic potency. Interestingly, Notch has been implicated in regulating the number of cardioblasts in the dorsal vessel of Drosophila independently of its effect on neural tissue (Corbin et al., 1991; Hartenstein et al., 1992). The similarity between the Drosophila data and our model indicate that the cardiogenic function of Notch has been remarkably well conserved throughout evolution.

Embryos were dejellied in 2% cysteine-HCl (pH 7.8), washed and maintained in 0.1× Marc’s Modified Ringer’s solution (MMR). Embryos were reared at 14-22°C and staged according to Nieuwkoop and Faber (1994).

Capped mRNA for microinjection was synthesized in vitro using the mMessage Machine kit (Ambion). Embryos were injected into one dorsal-vegetal blastomere at the 8-cell stage. cDNAs encoding the inducible constructs, GR-Su(H)VP16, GR-Su(H)^DBM and GR-NotchICD were the gift of Robert Davis. GR-Su(H)VP16 was made by fusing the activating domain of VP16 to the carboxyl terminus of Xenopus Su(H) (Wettstein et al., 1997) and the ligand binding domain of the human glucocorticoid receptor (GR) to the amino terminus (as in Kolm and Sive, 1995). GR-Su(H)^DBM is similar except that it lacks the VP16 activation domain and the Su(H) coding region contains a mutation in the DNA-binding domain (provided by Chris Kintner; Wettstein et al., 1997). Notch-ICD encodes the intracellular domain of Notch, and has been shown previously to activate Notch signaling constitutively. An inducible version was made by fusion of GR to the amino terminus. 250-500 pg of the GR-Su(H) mRNAs, or 500-1000 pg of GR-NotchICD mRNA were injected along with 50-100 pg of nuclear β-galactasidase mRNA as a lineage tracer. Injected embryos were cultured in 0.1× MMR until they reached the desired stages. Induction of GR constructs was by addition of dexamethasone (10μM) (Sigma). Dexamethasone was changed daily when embryos were cultured longer than 24 hours.

Embryos were fixed for 1 hour at room temperature in MEMFA (0.1 M MOPS pH 7.4, 2 mM EGTA, 1 mM MgSO⁴, 3.7% formaldehyde), rinsed in 1× PBS + 2 mM MgCl², and incubated in staining solution at 37°C with either X-gal or magenta-gal in order to detect β-galactosidase activity. Embryos were then rinsed in 1× PBS, postfixed in MEMFA for 2 hours and dehydrated in methanol. For in situ hybridization, embryos were rehydrated and processed using digoxigenin-coupled cRNA probes (Harland, 1991). Probes included: cardiac troponin-I (TnIc) (Drysdale et al., 1994), cardiac actin (Mohun et al., 1984), MHCα (Logan and Mohun, 1993), Nkx2.5 (Tonissen et al., 1994), Gata4 (Jiang and Evans, 1996), Bmp4 (Nishimatsu et al., 1992), Serrate1 (gift of Chris Kintner), and Notch1 (also known as Xotch and most similar to mouse Notch1; Coffman et al., 1990). For double in situ hybridization analysis, the second probe was FITC-coupled. In some cases, following in situ hybridization, embryos were postfixed and dehydrated through a methanol series, transferred to JB4 and embedded according to manufacturer’s directions (Polysciences).

We surveyed the spatial and temporal expression pattern of the previously identified Xenopus genes Notch1, Serrate1, Delta1 and Delta2. Delta1 and Delta2 mRNAs were not detected in the heart region by in situ hybridization between stages 20 and 30 (Fig. 1C). Notch1 and Serrate1 expression were first detected during the early tailbud stages (Fig. 1Ab,c,e,f). Expression of Notch1 and Serrate1 between stages 22 and 25 overlap and span an extensive portion of the Nkx2.5 field (Fig. 1Aa,d). This region comprises the future myocardium as well as extending more dorsolaterally to include the regions that give rise to the dorsal mesocardium and pericardial roof. Expression of both Serrate1 and Notch1 is preceded by expression of Nkx2.5 and precede that of markers of myocardial differentiation, such as TnIc, by several hours (Fig. 1Ag,k,o).

As development progresses, the expression patterns of both Notch1 and Serrate1 become more refined (Fig. 1Ai,j,m,n,q,r). By stage 30, just prior to heart tube fusion, Notch1 transcripts are present primarily in the developing myocardium whereas Serrate1 mRNAs are present primarily in the future dorsal mesocardium but also in the dorsalmost portion of the developing myocardium where they overlap Notch1 transcripts (Fig. 1Am,n,q,r). The resolution of the dynamic expression patterns of Notch1 and Serrate1 occurs during the developmental stages when the dorsolateral portions of the heart field lose myocardial potency (Raffin et al., 2000) and the prospective myocardium begins to express genes indicative of terminal myocardial differentiation. This can be seen most clearly in double in situ hybridization analysis of TnIc and Serrate1 transcripts where TnIc marks exclusively the developing myocardium and Serrate1 has become localized primarily to the dorsal mesocardium and roof of the pericardium (Fig. 1B). The timing and spatial patterns of expression suggest a model in which Notch activity feeds back on Serrate1, as in other systems (see Artavanis-Tsakonas et al., 1999). In addition to shifting Serrate1 expression dorsolaterally, Notch activity might suppress cardiomyocyte differentiation in the dorsolateral domains of the heart field. These possibilities are examined below.

To study the consequences of Notch signaling in the developing heart (during the second day of development) without affecting any earlier developmental steps in which Notch functions, we used inducible forms of the transcription factor Su(H). Activated and dominant negative forms of Su(H) were fused to the human glucocorticoid receptor ligand binding domain (Fig. 2; Materials and Methods). GR fusions have been employed to create transcription factors that are maintained in an inactive complex until activated by the addition of the glucocorticoid dexamethasone to the culture media (Kolm and Sive, 1995). By allowing signaling to be manipulated at a time consistent with the endogenous expression of Notch1 and Serrate1 in the heart region, this approach circumvented any confounding effects that early constitutive alteration of Notch signaling may cause.

The constructs were tested for both function and inducible regulation by injecting into animal blastomeres and examining the effect on primary neurogenesis, as indicated by β-tubulin expression. We obtained results identical to those of Wettstein et al. (1997), who observed specific alterations in the pattern and number of primary neural precursors. Thus, we concluded that our constructs are inducible and function normally (data not shown).

To address the possibility that a feedback loop refines the expression of Serrate1 in the heart field, we injected embryos with mRNA encoding either the activated or dominant negative form of Su(H). Embryos were injected into one dorsal-vegetal blastomere at the 8-cell stage with mRNA encoding either GR-Su(H)VP16 [activated Su(H)] or GR-Su(H)^DBM [dominant negative Su(H)]. These embryos were cultured in the absence of dexamethasone (uninduced) until stages 18-19, and then maintained in media containing dexamethasone (induced) until fixation at stages 28-31. This time period was chosen to yield functional GR-Su(H)VP16 or GR-Su(H)^DBM at a time coincident with endogenous Notch1 and Serrate1 expression (as shown in Fig. 1A).

Activation of Notch signaling with GR-Su(H)VP16 decreased the expression of Serrate1 on the injected side of the embryo compared to that on the uninjected side, which served as an internal control (Fig. 3A-C). A specific decrease was observed in 76.4% of injected embryos (n=144) treated with dexamethasone. In contrast, suppression of Notch signaling by GR-Su(H)^DBM increased expression of Serrate1 in 61.5% of injected embryos (n=78) (Fig. 3D-F). Sibling embryos injected with GR-Su(H)VP16 or GR-Su(H)^DBM but not incubated with dexamethasone were affected either minimally or not at all (data not shown). Moreover, no change in gene expression was observed in uninjected embryos cultured in dexamethasone, indicating that the hormone itself has no effect (data not shown). As for GR-Su(H)VP16, injection of mRNA for NotchICD decreased Serrate1 expression (80.5%, n=87, data not shown). Thus, we conclude that a feedback loop might be responsible for the resolution of Notch1 and Serrate1 expression in the heart field.

We next asked whether Notch signaling in the heart affects myocardial-specific genes. We examined several genes encoding contractile proteins including troponin Ic (TnIc), myosin heavy chain (MHCα) and cardiac actin (c. actin). These genes are expressed identically throughout the developing myocardium and are considered markers of myocardial differentiation (Mohun et al., 1984; Logan and Mohun, 1993; Drysdale et al., 1994). 60-80% of embryos injected with GR-Su(H)VP16 and treated with dexamethasone after stage 18-19 showed severely reduced or no expression of each of these genes (Fig. 4B,D; summarized in Fig. 6). In contrast, control embryos cultured in the absence of dexamethasone showed symmetrical expression on both the injected and the uninjected sides (Figs 4A,C, 6). A decrease in contractile protein gene expression could be explained by the apoptotic elimination of myocardial cells. However, TUNEL analysis employed to detect fragmented chromatin revealed no change in the pattern of apoptotic cells (data not shown). This argues against apoptotic elimination as does previous work demonstrating that the activation of Notch signaling can protect cells from apoptosis (Deftos et al., 1998; Jehn et al., 1999; Shelly et al., 1999). Additionally, we examined the effect of conditionally activating Notch signaling using GR-NotchICD, as has been done previously (Wettstein et al., 1997). As for GR-Su(H)VP16, GR-NotchICD caused a decrease in the myocardial marker TnIc (Fig. 4J; 47.7%, n=130).

In contrast to the effects of GR-Su(H)VP16 and GR-NotchICD, targeted expression of GR-Su(H)^DBM increased expression of TnIc, MHCα and cardiac actin on the injected side of the embryo (Fig. 4F,H; summarized in Fig. 6). As before, the effects were observed only on the injected side of dexamethasone-treated embryos. Interestingly, while the extent of myocardial gene expression was expanded, it never extended outside the domain of Nkx2.5 expression, even when GR-Su(H)^DBM was injected outside of this region. Instead, expression appeared to expand to occupy the dorsolateral regions of the Nkx2.5 domain, which normally form mesocardium and pericardial roof tissues. This may reflect the requirement of Nkx2 family members (in particular Nkx2.3 and Nkx2.5) for myocardial gene expression (Fu et al., 1998; Grow and Krieg, 1998). The opposite effects on myocardial gene expression elicited by the activating and inhibiting Su(H) constructs suggest that endogenous Notch signaling suppresses cardiomyogenesis while lack of Notch signaling permits myocardial differentiation.

Both the timing of endogenous Notch1 and Serrate1 expression (Fig. 1A) and the effects of GR-Su(H)VP16, GR-NotchICD and GR-Su(H)^DBM on myocardial gene expression (Fig. 4) are consistent with a role for Notch signaling in cardiogenesis after the early heart field is established. We therefore examined two markers of the early heart field, Nkx2.5 and Gata4 (Tonissen et al., 1994; Jiang and Evans, 1996). Unlike the pronounced effects on myocardial genes, dexamethasone activation of GR-Su(H)VP16, GR-Su(H)^DBM or GR-NotchICD after stage 18-19 had no or minimal effect on Nkx2.5 and Gata4 (Figs 5, 6). Thus, the observed effects on myocardial gene expression do not reflect a generalized perturbation of mesodermal patterning or an alteration in the specification of the heart field, including the amount of cardiogenic mesoderm present in the embryo. Instead, we conclude that endogenous activation of Notch controls cardiac cell fate after the specification of the Nkx2.5/heart field.

The observed effects of experimental manipulation of Su(H) activity on expression of myocardial genes are consistent with a role for Notch in selection of myocardial versus non-myocardial lineages. To examine this model further, we followed the cellular progeny of injected blastomeres to determine whether alterations in Notch signaling correlate with a change in contribution to myocardial tissue. This analysis extends the marker study by examining alterations of fate of individual cells rather than early patterns of gene expression, which might provide an imprecise prognosis of cell identity. Embryos were co-injected with mRNAs encoding either GR-Su(H)VP16 or GR-Su(H)^DBM along with β-galactosidase, induced with dexamethasone at stage 18-19, and then fixed and stained with X-gal at stage 40 to identify the progeny of injected cells. The percentage of embryos with β-galactosidase-positive cells in the myocardium was scored (Fig. 7). As expected, frequent contribution to myocardial and non-myocardial tissues was observed in embryos that were not induced with dexamethasone, indicating that the injected mRNA was properly targeted (Fig. 7Aa,b). However, induction with dexamethasone diminished the myocardial contribution of β-galactosidase-positive cells in embryos where Notch signaling was activated by GR-Su(H)VP16 (Fig. 7Ac). In these embryos, β-galactosidase-positive cells persisted in the embryo but were found predominately in non-myocardial tissue. In addition, the myocardium of these embryos appeared smaller and overall morphology was disorganized. In contrast, embryos expressing the dominant negative Su(H) and treated with dexamethasone showed robust contribution of β-galactosidase-positive cells to the myocardium (Fig. 7Ad). These hearts often appeared larger than those of controls, which had been injected but not treated with dexamethasone. Thus, perturbations of Notch signaling affect not only gene expression but also the fate of cells within the heart field.

One implication of the opposite effects of GR-Su(H)VP16 and GR-Su(H)^DBM on cardiomyogenesis, consistent with the spatial patterns of Serrate1 and Notch1, is that endogenous Notch signaling mediates a lineage choice between myocardium and the non-myocardial cells that give rise to the mesocardium and pericardial roof. To explore this possibility, we asked if the observed changes in myocardial gene expression are accompanied by commensurate alterations in genes specific for mesocardial or pericardial roof tissue. Two markers for these tissues were examined: Serrate1, which is affected by Notch signaling (as shown in Fig. 3), and Bmp4 (Nishimatsu et al., 1992). In situ hybridization showed that Bmp4 transcripts are localized specifically in the prospective mesocardium extending into the pericardial roof but not elsewhere in the heart region at stage 28 (Fig. 8Aa-c). In embryos injected with GR-Su(H)VP16 mRNA and cultured in dexamethasone from stage 18-19 until fixation, Bmp4 expression increased on the injected side of the embryo (Fig. 8B). This increase was noted in 62.7% (n=370) of injected embryos treated with dexamethasone. Bmp4 expression expanded either ventrally (into tissue which normally forms myocardium) or more laterally. In contrast, GR-Su(H)^DBM decreased expression of Bmp4 on the injected side of the embryo (Fig. 8B). This phenotype was observed in 52.9% (n=121) of injected, dexamethasone-treated embryos. The decrease in gene expression observed in GR-Su(H)^DBM-injected embryos was not attributable to apoptotic elimination as determined by TUNEL staining (data not shown). The magnitude of the alteration in Bmp4 expression, elicited by either GR-Su(H)VP16 or GR-Su(H)^DBM, was less than that for either of the myocardial genes examined; however, the percentage of embryos affected was similar. Possible reasons for this observation are discussed below. Nonetheless, the inverse response of myocardial genes and Bmp4 to both activating and inhibitory Su(H) constructs is consistent with the model that endogenous Notch signaling influences the selection between myocardial and mesocardial/pericardial roof cell fates within the Nkx2.5/heart field.

An early differentiation event in the cardiogenic mesoderm is the establishment of distinct myocardial and non-myocardial domains. Although the entire crescent of the Xenopus Nkx2.5-expressing cardiac mesoderm is specified initially as myogenic, this potency (but not Nkx2.5 expression) is lost in the dorsolateral portions of the field and these cells eventually adopt mesocardial and pericardial roof fates. The experiments described here show that activation of Notch signaling by GR-Su(H)VP16 and NotchICD caused diminished expression of Serrate1 (Fig. 3) and the myocardial genes TnIc, MHCα and cardiac actin (Figs 4, 6) and a corresponding increase in expression of Bmp4. Inhibition of endogenous Notch signaling by GR-Su(H)^DBM resulted in the opposite of these effects. In addition, GR-Su(H)VP16 diminished contribution of individual cells to the myocardium (Fig. 7). Importantly, these changes in gene expression and cell fate were not accompanied by alterations in the expression of Nkx2.5 and Gata4. We conclude that endogenous Notch signaling regulates cell fate downstream of the establishment of the early heart field and that, after stage 20, Notch signaling does not alter expression of these early heart field markers. In addition, our observed regulation of Serrate1 expression upon injection of Su(H) suggests that the resolution of the largely overlapping patterns

Our results suggest that Su(H) is both necessary and sufficient to mediate the suppression of cardiomyogenic differentiation. However, Su(H)-independent signaling from Notch has been demonstrated (Shawber et al., 1996b; Matsuno et al., 1997; Nofziger et al., 1999; Wilson-Rawls et al., 1999). Whether or not such a pathway also functions during cardiogenesis has not yet been addressed. In particular, the expression of Notch1 and Serrate1 after the differentiation of myocardial and non-myocardial domains leaves open the possibility that the Su(H)-independent pathway might be involved during later steps of heart development.

Two models are consistent with our finding that Notch controls the ability of cardiogenic mesoderm to differentiate (Fig. 9). In a binary cell fate choice model, the presence or absence of Notch signaling causes an Nkx2.5-positive cell to adopt either a myocardial or a mesocardial/pericardial roof fate. Alternatively, a differentiation delay model would predict that Notch signaling maintains or prolongs cardiac cells in an undifferentiated state. This is similar to studies of the Xenopus retina, where cells transfected with a gene encoding the constitutively active, intracellular domain of Xenopus Notch1 do not immediately shift fate but remain transiently undifferentiated (Dorsky et al., 1997). The maintenance or prolongation of a stable precursor population by Notch signaling may be a general feature of the developing nervous system (Austin et al., 1995; Dorsky et al., 1995, 1997; Wang et al., 1998). In the avian neural retina, successive rounds of ligand expression and of Notch1 and Serrate1 mRNA expression at stage 22-25 to the minimally overlapping pattern seen at stage 29-30 (Fig. 1), occurs via feedback inhibition of Serrate1 by Notch activation. Finally, an important conclusion is that endogenous Serrate-Notch-Su(H) signaling inhibits cardiomyogenesis. This is evident by comparison of the opposite changes in myocardial gene expression (Figs 4, 6) as well as the diversion of cardiomyogenic cells to non-myocardial fates by activation of GR-Su(H)VP16 (Fig. 7). These results, taken together with the spatial and temporal patterns of Notch1 and Serrate1 expression (Fig. 1), suggest that endogenous Serrate1/Notch1/Su(H) signaling is involved in the suppression of cardiomyogenesis that normally occurs in the dorsolateral portions of the Nkx2.5/heart field between stages 22 and 28. lateral inhibition are proposed to lead to a controlled exodus of cells from the precursor pool to take on primary, secondary and tertiary differentiated fates (Austin et al., 1995). By analogy, endogenous Notch signaling may block responsiveness to a myocardial differentiation signal in the dorsolateral portions of the Nkx2.5/heart field after stage 22-25, when these cells have overlapping Serrate1 and Notch1 mRNA expression and lose myocardial potency. However, precise knowledge of the spatial and temporal patterns of endogenous Notch activity as well as more precise temporal control over experimental perturbations of Notch activity will be needed to distinguish between the differentiation delay and simple lineage choice models.

Expression of Bmp4 was examined to explore whether the effects on myocardial differentiation caused by GR-Su(H)VP16 and GR-Su(H)^DBM were accompanied by a compensatory change in mesocardial and pericardial roof tissue. This possibility is predicted by the lineage choice model, but is also possible in the differentiation delay model if the activity of the GR-Su(H) constructs were not sustained. This latter scenario is possible since the injected mRNAs decay over time. In support of the idea that Notch regulates a balance between differentiation as myocardium versus mesocardium/pericardial tissues, we found that Bmp4 and myocardial genes were inversely regulated by the Su(H) constructs (compare Figs 4 and 8). Although the magnitude of altered Bmp4 expression (which marks mesocardium and pericardial roof tissue) was less than that seen for the myocardial markers, the percentage of embryos affected was similar. Several factors could account for the moderate magnitude of change in Bmp4 expression. First, the ability to form myocardium may be restricted to only a fraction of Bmp4-positive cells, either because they lack competence or are too distant from any pro-myocardial factors that might exist in or near the prospective myocardium. Second, expression of Bmp4 itself may not be an ideal marker of mesocardial and pericardial roof specification. We are unaware of additional markers (other than Serrate1, which appears regulated by Notch signaling as part of the feedback loop that resolves the overlapping Notch1/Serrate1 expression). The identification of additional markers will be helpful to corroborate our tentative conclusion that suppression of cardiogenesis by Notch activity coincides with an increase in mesocardial and pericardial roof fates.

We chose to perturb Notch signaling at the level of the downstream mediator Su(H) in part to overcome the complexity generated by multiple Notch receptors and ligands. Although Delta1 and Delta2 transcripts were not detected in the early heart field, the broad specificity of GR-Su(H)VP16 and GR-Su(H)DBM leaves open the possibility that Notch receptors and/or ligands, in addition to Serrate1 and Notch1, influence the myocardial/non-myocardial lineage decision. Moreover, it is likely that Notch signaling plays additional roles during cardiogenesis, either prior to the specification of Nkx2.5 or after the initial determination of myocardial and non-myocardial tissues. Such a hypothesis is similar to the conclusions drawn from studies on somitogenesis and neurogenesis, in which Notch activity has been shown to act at multiple stages to contribute to complex tissue patterning (Fuerstenberg and Giniger, 1998; Chitnis, 1999; Sestan et al., 1999). A role for endogenous Notch signaling in the heart prior to establishment of the Nkx2.5/heart field could contribute to the slight effect that activation of the Su(H) constructs had on expression of Nkx2.5 and Gata4 (Fig. 6). Expression of Notch receptor and ligand mRNA and protein in the developing heart of the mouse, chick and zebrafish also point to possible later functions, in particular in the developing myocardial tube, outflow tract and endocardium (Franco del Amo et al., 1992; Reaume et al., 1992; Bierkamp and Campos-Ortega, 1993; Williams et al., 1995; Myat et al., 1996; Westin and Lardelli, 1997). This would be consistent with cardiac defects that characterize the preponderance of patients with Alagille syndrome caused by autosomal dominant mutations in the Jagged1 gene (Krantz et al., 1997; Li et al., 1997; Oda et al., 1997).

Our studies of Notch extend the current model for the subdivision of the heart field. We propose that signaling through the Serrate-Notch-Su(H) pathway mediates the suppression of the cardiomyogenic program in the dorsolateral portion of the Nkx2.5/heart field. Although Notch acts to inhibit differentiation, as shown in a number of in vivo and in vitro systems, and we have shown here that inhibition of Notch correlates with an increase in cardiomyogenesis, the impetus to differentiate may be more complex than lack or abrogation of a Notch signal. By analogy with its role in neurogenesis, Notch may regulate competence of cells within the cardiogenic mesoderm to adopt myocardial, mesocardial or pericardial roof fate. The absence of Notch signaling in the ventral portion of the heart field, inferred from our results, may confer competence to differentiate as myocardium. Also, Notch activity (at least transiently) may be important for differentiation of the mesocardium and pericardial roof. However, cell fate acquisition may also require specific local cues. Therefore, it will be of interest to examine whether locally expressed proteins, such as BMP4, function in concert with Notch to regulate cell fate within the cardiogenic mesoderm.

The authors thank Josh Rutenberg for helpful discussions and help with in situ hybridization during the early stages of this work; and Valerie Schneider for radioactive BMP4 in situ hybridization data. We also acknowledge other members of the Mercola lab, and Drs. Spyros Artavanis-Tsakonas, Karen Symes and Barbara Osborne for insightful comments on the manuscript. Additionally, we thank Drs Chris Kintner, Paul Krieg, Tim Mohun and, in particular, Robert Davis for generous contribution of reagents and Chris Simpson for expert histology. The authors gratefully acknowledge funding from the NIH (RO1 HL59502 to M. M. and F32 HL09740-03 to K. M.).