During the formation of the Drosophila heart, a combinatorial network that integrates signaling pathways and tissue-specific transcription factors specifies cardiac progenitors, which then undergo symmetric or asymmetric cell divisions to generate the final population of diversified cardiac cell types. Much has been learned concerning the combinatorial genetic network that initiates cardiogenesis, whereas little is known about how exactly these cardiac progenitors divide and generate the diverse population of cardiac cells. In this study, we examined the cell lineages and cell fate determination in the heart by using various cell cycle modifications. By arresting the cardiac progenitor cell divisions at different developing stages, we determined the exact cell lineages for most cardiac cell types. We found that once cardiac progenitors are specified, they can differentiate without further divisions. Interestingly, the progenitors of asymmetric cell lineages adopt a myocardial cell fate as opposed to a pericardial fate when they are unable to divide. These progenitors adopt a pericardial cell fate,however, when cell division is blocked in numb mutants or in embryos with constitutive Notch activity. These results suggest that a numb/Notch-dependent cell fate decision can take place even in undivided progenitors of asymmetric cell divisions. By contrast, in symmetric lineages, which give rise to a single type of myocardial-only or pericardial-only progeny, repression or constitutive activation of the Notch pathway has no apparent effect on progenitor or progeny fate. Thus, inhibition of Notch activity is crucial for specifying a myogenic cell fate only in asymmetric lineages. In addition, we provide evidence that whether or not Suppressor-of-Hairless can become a transcriptional activator is the key switch for the Numb/Notch activity in determining a myocardial versus pericardial cell fate.
INTRODUCTION
How cell-fate diversity is generated is a central issue in developmental biology. Recently, this question has been more specifically asked during the development of individual organs. During Drosophila heart development, a limited number of progenitors generate a defined set of cell types that are arranged in a stereotyped and segmentally repeated pattern(Ward and Skeath, 2000; Han et al., 2002), making it an excellent model with which to address this question.
The Drosophila heart originates from bilaterally symmetrical rows of precursor cells in the dorsal mesoderm, which express the homeobox gene tinman (Bodmer, 1993; Azpiazu and Frasch, 1993). As a consequence of inductive signals from the adjacent ectoderm, tinmanexpression is confined to the dorsalmost region of the mesoderm, defined as the cardiac mesoderm (Frasch,1995; Wu et al.,1995; Lockwood and Bodmer,2002). Within this competence domain, cardiac progenitors emerge as cell clusters that express a distinct combination of transcription factors,including even-skipped (eve), ladybird(lb), seven-up (svp), odd-skipped(odd) and tinman itself, which are thought to be involved in conferring the appropriate differentiation pathway to these clusters(Ward and Skeath, 2000; Jagla et al., 2002; Han et al., 2002). The initiation of these cell clusters of different developing fates is probably mediated by extrinsic, inductive mechanism, which is determined by their position in the cardiogenic domain. Subsequently, a defined pattern of lineages generates the final diversity of cell types(Ward and Skeath, 2000). Some cardiac progenitors divide symmetrically, whereas others undergo stereotyped asymmetric cell divisions.
Two types of asymmetric cell division in the cardiac mesoderm have been described: the progenitors of the Svp lineage divide asymmetrically to generate Svp myocardial cells (SMC) and Svp-Odd pericardial cells (SOPC)(Ward and Skeath, 2000); the progenitors of the mesodermal Eve lineage generate Eve pericardial cells (EPC)and muscle founders DA1 and DO2 (Park et al., 1998; Halfon et al.,2000; Carmena et al.,2002; Han et al.,2002). These two types of asymmetric cell division are similar in the sense that they both generate myogenic versus non-myogenic sibling cell fates. The myocardial and muscle founder cells have many characteristics of muscle identity in common, which is excluded in pericardial cell differentiation. For example, they both express the muscle differentiation gene Mef2 (Bour et al.,1995; Lilly et al.,1995). Previous studies suggest that the Notch pathway plays an important role in determining these alternative cell fate decisions(Park et al., 1998; Carmena et al., 1998; Carmena et al., 2002; Ward and Skeath, 2000). In mutants of the Notch antagonist encoded by numb (reviewed by Jan and Jan, 1998), the number of pericardial SOPC and EPC is increased, accompanied by a loss of myocardial SMC and of DA1 muscles. Conversely, when numb is overexpressed in the mesoderm, EPCs are not formed, only DA1 muscles. Moreover, when a constitutively active form of Notch, Notch intracellular domain or N(icd), is expressed in the mesoderm, the mesodermal Eve lineage is almost completely eliminated, suggesting that Notch activity inhibits the formation of progenitors at an early stage. However, when a temperature-sensitive allele of Notch is used to eliminate Notch function at the time when the Eve progenitors divide, EPCs but not DA1 muscle fail to form(Park et al., 1998). Thus, Notch seems to have a dual function, as it is required both for progenitor specification and for asymmetric cell fate determination of the same lineage. It also appears that the Notch pathway is required for specifying pericardial as opposed to myocardial or muscle founder cell fate in both the Svp and the Eve lineages. It is not known, however, if the Notch-dependent cell fate decision is made after cell division in the progeny cells or if it is already initiated in the progenitors before division.
Studying these Notch-mediated asymmetric cell divisions in the context of the cell cycle may provide insights into the coordination of cell fate and cell division. Thus, preventing cell division of asymmetric cell divisions allows us to investigate whether precursors blocked in cell cycle progression are predetermined or biased in their cell fate decision, or whether this is a necessary prerequisite for alternative cell fate determination. Moreover, blocking the mesodermal cell division at progressively later stages will enable us to get a better understanding of the cell lineages in the heart. Previous studies have followed marker gene expression and used the Flp-FRT-based lineage tracing method to address this question (Park et al., 1998; Ward and Skeath, 2000; Carmena et al., 1998; Carmena et al., 2002), but some of the details of these lineages are still unclear.
Several genes have been shown to arrest cell divisions at different stages and cell cycle number within the ectoderm. It is believed that most cells in Drosophila undergo three rounds of cell division after blastoderm that are partially synchronous; e.g. slightly earlier in the mesoderm than in the ectoderm (Foe, 1989; Campos-Ortega and Hartenstein,1997). Later, specialized embryonic tissues, such as the nervous system, undergo further cell divisions(Bodmer et al., 1989; Foe, 1989; Lu et al., 2000). The last round of global cell division, mitosis 16, is blocked in CyclinA(CycA) or Rca1 mutants(Knoblich and Lehner, 1993; Dong et al., 1997). By contrast, mitosis 16 is not obviously affected by CyclinB(CycB) mutants. However, double mutants of CycA and CycB act synergistically and arrest cell division of most ectodermal cells at the G2/M transition of cycle 15(Knoblich and Lehner, 1993). In string (stg, cdc25 in yeast) mutant embryos, mesodermal cells fail to enter metaphase of mitosis 14(Foe, 1989; Edgar and O'Farrell, 1989; Edgar and O'Farrell, 1990). In addition to these genes that are required for cell cycle progression, some other cell cycle genes have been shown to be required for cells to exit the cell cycle. These include dacapo (dap)(Lane et al., 1996) and fizzy related (fzr; rap – FlyBase)(Sigrist and Lehner, 1997). dap encodes a CDK inhibitor that is necessary for exiting the cell cycle at the appropriate time, whereas fzr negatively regulates the levels of cyclins A, B and B3, and is required for cyclin removal during G1(when cell proliferation stops). Loss of either gene causes cell division progression through an extra cycle. Conversely, premature overexpression of dap or fzr in transgenic embryos inhibits mitosis and results in cell division arrest (Lane et al., 1996; Sigrist and Lehner,1997).
In this study, we examined the formation of cardiac cell types in cell cycle mutants and in embryos in which cell cycle inhibitors are overexpressed. We found that the cardiac progenitors continue to differentiate in the absence of cell division. Interestingly, the progenitors of the asymmetric Eve and Svp lineages always adopt a myogenic cell fate, presumably because Notch activation is prevented in the progenitor because of the presence of Numb. To test this, we arrested cell division in numb mutants or overexpressed N(icd) in these progenitors, and found that they adopt a non-myogenic pericardial cell fate. These data indicate that genes normally involved in the cell fate decisions during asymmetric cell division also determine the progenitor cell fate. By contrast, cell fate decisions with symmetric cardiac lineages are not influenced by the presence or absence of Notch activity. Our data suggest that cell cycle genes act in concert with the Notch pathway to generate the diversity of cell types in the Drosophila heart, and that the bHLH transcription factor Suppressor-of-Hairless [Su(H)] mediates this activity. We speculate that an increase in cell type diversity can be achieved by adopting the Numb/Notch system to generate asymmetry within any lineage of an organism.
MATERIALS AND METHODS
Drosophila strains
The following mutant stocks were used: Rca103300, stg7 and numb1 are from the Bloomington Stock Center; CycAC8 and CycB2 (described by Knoblich and Lehner, 1993) are gifts from C. Lehner; Notchts has been described previously (Park et al.,1998). Overexpression of transgenes was accomplished by using the Gal4-UAS system (Brand and Perrimon,1993). The following fly lines were used: twi-Gal4 and 24B-Gal4 (conferring pan-mesodermal expression)(Greig and Akam, 1993; Brand and Perrimon, 1993),eme-Gal4 (which confers expression in the mesodermal Eve lineage)(Han et al., 2002),UAS-dap (Lane et al.,1996), UAS-fzr(Sigrist and Lehner, 1997),UAS-Su(H) and UAS-Su(H)-vp16(Furriols and Bray, 2000), and UAS-N(icd) and UAS-numb(Yaich et al., 1998; Park et al., 1998). Fly lines carrying emeA-lacZ (380 bp eve mesodermal enhancer directing lacZ expression) (Han et al.,2002) is used to label EPCs and the DA1 and DO2 founder cells. Oregon-R was used as the wild-type reference strain.
Immunohistochemistry and microscopy
Embryos from different lines were collected and stained with various antibodies as previously described (Han et al., 2002). The following primary antibodies were used:anti-β-galactosidase 1:300 (Promega); anti-Eve 1:10,000(Frasch et al., 1987);anti-Tin 1:500 (Venkatesh et al.,2000); anti-Mef2 1:1000 (Lilly et al., 1995); and anti-Lbe 1:40(Jagla et al., 1997). FITC- or Cy3-conjugated secondary antibodies (from Jackson Laboratories) were used to recognize the primary antibodies. Images were obtained with a Zeiss LSM510 confocal microscope.
RESULTS
Mesodermal Eve lineages and the fate of Eve progenitors in the absence of cell division
The Drosophila heart progenitors are specified at the dorsal mesodermal edge and are coincident with the cells that maintain expression of tinman at stage 11. These progenitor cells undergo symmetric or asymmetric cell divisions to generate a diverse set of cardiac cell types(Ward and Skeath, 2000). From stage 13 onwards, each cell type is positioned at a specific dorsoventral and anteroposterior location and is distinguished by the expression of a unique combination of transcription factors (see Han et al., 2002). Of these,the MADS-box transcription factor Mef2(Bour et al., 1995; Lilly et al., 1995) marks all the myocardial cells, whereas the homeobox gene even-skipped(eve) labels the EPCs (a pericardial cell type), DA1 (a dorsal muscle) and transiently the founder of the DO2 muscle(Fig. 1A)(Frasch et al., 1987; Carmena et al., 2002; Han et al., 2002). The lineage giving rise to the mesodermal eve-expressing cells has been studied intensely (e.g. Carmena et al.,1998; Park et al.,1998; Su et al.,1999; Halfon et al.,2000). Nevertheless, the exact lineage of these cells has not been unambiguously resolved, which is reflected in two different models(Fig. 1E,F)(Park et al., 1998; Carmena et al., 1998).
Here, we study the Eve and other lineages of the heart by arresting the cell cycle at various stages. We first examined formation of cardiac cell types in CycA mutant, which causes cell cycle arrest at mitosis 16 in the ectoderm (Knoblich and Lehner,1993). In CycA mutants, EPCs are absent as judged by the absence of eve expression, but DA1 muscles form as in wild type(Fig. 1A,B). A similar phenotype is observed in Rca1 mutants (arrest also at ectodermal cycle 16 as CycA) (Dong et al.,1997; Wai et al.,1999), but with somewhat less penetrance(Fig. 1C). In Rca1mutants, 86% (n=144) of hemisegments show no EPCs and 14% show one EPC; all of them have normal DA1 muscles(Fig. 1C). The one-EPC phenotype suggests that a common progenitor exists for the two EPCs in each hemisegment. This conclusion is further supported by examining embryos in which the cell cycle inhibitor dap or fzr is overexpressed specifically in the mesoderm [using the twist-Gal4 driver line(Greig and Akam, 1993) in conjunction with the UAS-Gal4 system(Brand and Perrimon, 1993)]. In these embryos (progeny from the cross between twist-Gal4 crossed UAS-dap, twi>dap) most hemisegments show only one EPC(94%, n=150; Fig. 1D)suggesting that the last division generating two EPCs did not occur. This division apparently is symmetrical. A similar phenotype is observed in 70%twi>fzr embryos (n=150, data not shown). Because it is thought that Dacapo causes arrest of the last division(Lane et al., 1996), and that CycA and Rca1 are required for the cell cycle 16, we propose that the EPCs are generated at cycle 17 by symmetric cell division and the progenitor of EPCs and a muscle founder are generated at cycle 16 by asymmetric cell division. Thus, these findings together with previous data support the model depicted in Fig. 1F (see also Fig. 9). More importantly,these data also suggest that in CycA or Rca1 mutants, the undivided Eve progenitors of cycle 16 adopt a myogenic cell fate.
Undivided progenitors of the Svp lineage also adopt a myocardial fate
In all the four cell cycle arrest situations examined so far (CycAand Rca1 mutants, overexpression of dap or fzr in the mesoderm), four instead of six myocardial cells form in each hemisegments of the abdominal segments, A2-A7 (Fig. 1B-D), indicating that the progenitors of myocardial cells differentiate as myocardial cells when the last cell division is blocked(Fig. 9). It also indicates that the last division of myocardial cells is mitosis 16, as twi>dap or twi>fzr embryos exhibit the same myocardial cell phenotype as CycA or Rca1 mutants. Previous studies suggest that in each hemisegment the four Tinman myocardiac cells (TMC) are generated from symmetrical divisions, whereas two Svp progenitors divide asymmetrically to give rise to two Svp myocardial cells (SMC) and Svp-Odd pericardial cells (SOPC; Fig. 2A,E) (Ward and Skeath,2000) (see also Fig. 9C). If this is the case, the number of TMC in CycA or Rca1 and twi>dap embryos should be two, and the number of SMC should be either two or none, depending of the fate of the undivided Svp precursors. Indeed, in these cycle 16 arrest situations, only two TMC per hemisegment form (Fig. 2B-D). Because the total number of myocardial cells is four(Fig. 1B-D), the other two myocardial cells are probably SMC. In order to test this, we examined the number of odd-expressing pericardial cells, two out of a total of four co-express svp (SOPC) in wild-type embryos(Fig. 2E). In CycAmutants, a single odd-expressing pericardial cell (OPC) is present per hemisegment (Fig. 2F, Fig. 9B), consistent with the prediction that the two Svp-Odd progenitors assume a myocardial fate when cycle 16 is blocked. A third Odd-only progenitor is likely to divide symmetrically, as suggested previously(Ward and Skeath, 2000).
tinman-expressing myocardial progenitors are likely to divide symmetrically
In each abdominal hemisegment of A2-A7 there are four ladybird-expressing heart cells(Fig. 9A,C), two that are myocardial, co-expressing tinman and Mef2 (TLMC), and the other two that are pericardial, co-expressing tinman only (LPC)(Jagla et al., 1997). These cells are adjacent to the Eve cells but never overlap with them(Fig. 2G)(Han et al., 2002). In CycA mutants, two of the myocardial cells express tinman(Fig. 2B), and of only one expresses ladybird (TLMC in Fig. 2H; Fig. 9B),suggesting that the two tinman-expressing myocardial progenitors divide symmetrically, but each assumes a distinct molecular identity distinguished at least in part by the presence or absence of ladybirdexpression (summarized in Fig. 9).
Distinct lineages of the tinman-expressing myocardial cells in the anterior two segments of the heart
By comparing the Mef2 and tinman expression in both wild-type and cell cycle mutants, we noticed that the pattern of cardiac cell types is different in the anterior two segments (T3 and A1). The embryonic heart is formed approximately within segment T3 to A8, and is composed of six myocardial cells per hemisegment, except in A8, where only four are present(Fig. 3A). Therefore, the total number of the myocardial cells expressing Mef2 is 104[(6×8+4×1)×2] (number of cells per hemisegment ×number of segments × 2 sides) (Fig. 3A). Two of the myocardial cells in each A2-A8 hemisegment express svp but not tinman (SMC). Therefore, the total number of SMC is 28 (2×7×2) (see Gajewski et al., 2000; Lo et al.,2002). In the anterior two segments, all myocardial cells express tinman (Fig. 9A),which adds up to a total number of 76 TMCs[(6×2+4×6+2×1)×2],(Fig. 3C; Fig. 9A) (Alvarez et al.,2003). In embryos in which dap is expressed throughout the mesoderm,thus blocking the last division, the number of myocardial cells is reduced in T3-A1 (average 3.2 per hemisegment, n=20), as it is in the A2-A7 abdominal segments (two per hemisegment, n=76)(Fig. 3B,D). In CycAor Rca1 mutant embryos, however, which are normally blocked in cell cycle 16, the number of TMCs in T3-A1 remains unchanged (six per hemisegment, n=26), unlike in A2-A7 where the number of TMCs is reduced to half(two per hemisegment, n=82) (Fig. 3E,F; Fig. 9B). This suggests that in the anterior two segments, the TMC precursor divisions are already complete after cycle 15, thus not affected in CycA and Rca1 mutants, or the anterior myocardial lineages are different, as suggested by a recent lineage study (Alvarez et al., 2003), or both. Whatever turns out to be the case, these conclusions are consistent with recent observation that the T3-A1 heart precursors are specified under homeotic control that is distinct from that of the other abdominal segments(Lovato et al., 2002; Lo et al., 2002). Not only is the myocardial cell number in T3-A1 unaffected in CycA or Rca1 mutants, but as observed in a late stage embryos, these anterior myocardial cells assemble into a tube as in wild type(Fig. 3F). By contrast, the fewer than normal abdominal myocardial cells do not align properly, suggesting that a reduction in myocardial cell number adversely affects heart tube morphogenesis.
CycA;CycB double mutant arrests mesodermal division at mitosis 15
In order to study the cardiac lineages further, we examined CycA;CycB double mutants, which are thought arrest cell division before cycle 16. CycA and CycB function synergistically during the G2-M transition (Knoblich and Lehner,1993). Even though CycB mutants do not seem to affect the three post-blastoderm divisions, CycA;CycB double mutants are more severe than CycA mutant alone, in that ectodermal cell division is arrested at cycle 15 (Knoblich and Lehner, 1993). In the heart, CycA;CycBdouble mutants exhibit a further reduction in the number of myocardial cells when compared with CycA alone(Fig. 4A-D). Instead of six myocardial cells in wild type and four myocardial cells in CycAmutants, only two myocardial cells are observed in 75% of 120 observed hemisegments in CycA;CycB double mutants(Fig. 4B). In addition, only one TMC forms in 64% of 120 observed hemisegments(Fig. 4D), indicating that the cell divisions of myocardial lineages are indeed blocked at cycle 15 in CycA;CycB double mutants. These data also suggest that the four TMC that are formed in each hemisegment originate from one super progenitor (TSP) and the two SMC plus two SOPC originate from another super progenitor (SSP, Fig. 9C). Both super progenitors appear to be specified before cycle 15. Although one of the two TMC per hemisegment in CycA mutants expresses ladybird early, this homeobox gene product is no longer detected in double mutants of CycA and CycB (Fig. 4E,F). These observations are consistent with the idea that initially one TMC super-progenitor is specified in each hemisegment, which divides asymmetrically during cycle 15, giving rise to a Ladybird-positive and Ladybird-negative TMC progenitor.
In segments T3-A1, the number of TMCs is not altered by
CycA or CycB mutants. By contrast, CycA;CycBdouble mutants exhibit a reduction of TMC from six to three in 70% of observed hemisegments (n=46; Fig. 4G,H). These data are consistent with the hypothesis that the T3-A1 TMC progenitor divisions occur during cycle 15, which also fits with the idea that the identity of these two cardiac segments is specified differently(Lo et al., 2002; Alvarez et al., 2003). Interestingly, the DA1 muscles are specified similarly in CycA single or in CycA;CycB double mutants(Fig. 1B; Fig. 4B,D), suggesting that Eve progenitor specification is unaffected by arresting mitosis at cycle 15. This is reminiscent of the observation made previously in stg mutants, in which the cell division is arrested at cycle 14, the first post-blastoderm division: up to two Eve progenitors are specified per hemisegment(Carmena et al., 1998; Su et al., 1999), one of which will give rise to the DA1 muscle founder and the other to the immediate precursor of two EPCs per hemisegment (Fig. 9C). Re-examining the cardiac phenotype in stg mutants we found that the eve-expressing cells often appear in pairs, although in some segments tinman- or eve-expressing progenitors fail to form in this early arrest mutant. The pairs of eve-expressing cells seem to adopt a muscle founder cell fate, as most of them express Mef2 but not tinman in stage 13/14 embryos(Fig. 4I,J). However, unlike CycA;CycB double mutants, stg mutants show severe segmentation and other patterning defects, presumably aggravated by the paucity of cells that are formed, which precludes further interpretation of the cardiac lineages.
Function of Numb and Notch in the asymmetric cardiac cell lineages
Previous studies have shown that the asymmetric cell divisions in both the Svp lineage and the Eve lineage are numb dependent and involve the Notch pathway (Park et al.,1998; Carmena et al.,1998; Carmena et al.,2002; Ward and Skeath,2000). In numb mutants, the number of myocardial cells in segments A2-A7 is reduced from six to four per hemisegment(Fig. 5A,B; Fig. 9B), whereas the number of myocardial cells in T3-A1 is not affected (data not shown) (Alvarez et al.,2003). The reduction of myocardial cells is paralleled by an increase in EPC number from two to four in 80% of 200 hemisegments counted, accompanied by a loss of DA1 muscles (Fig. 5B; Fig. 9B)(Park et al., 1998). By using a Gal4 driver under the control of the mesodermal eve enhancer(eme-Gal4) (Han et al., 2002),Numb and constitutively active N(icd), were expressed exclusively in the mesodermal Eve lineage (Fig. 5C,D). In eme>numb embryos, no EPCs are formed, only the DA1 muscles (Fig. 5C). By contrast, no DA1 muscles are formed in eme>N(icd) embryos, and 60%of 160 hemisegments exhibit an increase in EPC number from two to three or four (Fig. 5D), supporting a cell autonomous action of Numb and Notch in this lineage. Interestingly, the remaining segments show no expression of eve, suggesting the corresponding eve progenitors have not formed, as observed previously with pan-mesodermal overexpression of N(icd)(Park et al., 1998). It is likely that the variation in phenotype is the result of slight regional differences in the onset of N(icd) expression: earlier expression eliminates the progenitor, while later expression results in a sibling fate transformation opposite to that observed in numb mutants.
Notch pathway functions in the cardiac mesoderm through Su(H)
As Su(H) is the only transcription factor that has been characterized so far in mediating Notch activation, we wanted to know if it has a similar function in the cardiac lineages. Unexpectedly, overexpression of Su(H) in the mesodermal Eve lineage or throughout the mesoderm does not alter the normal cardiac/muscle fate determination (Fig. 5E; data not shown). Recent data suggest that Su(H) is converted from a transcription repressor to an activator when the Notch receptor is activated (reviewed by Bray and Furriols,2001). Therefore, the mere presence of more wild-type Su(H)protein may not cause activation of the Notch pathway in the mesoderm, as it does in parts of the nervous system (Wang et al., 1997; Nagel et al.,2000). To circumvent this limitation, we expressed an activated form of Su(H), Su(H)vp16, which has been generated by fusing wild-type Su(H)to vp16 (Furriols and Bray,2000), a strong transcriptional activation domain. Interestingly,eme>Su(H)vp16 embryos exhibit a phenotype similar to that of eme>N(icd): in some segments no Eve cells form, whereas in others more EPCs and no DA1 muscles form (Fig. 5F). These data suggest that the dual function of Notch,lateral inhibition and asymmetric cell fate determination, are both mediated by a mechanism that involves the conversion of Su(H) into a transcriptional activator.
As the function of Numb is thought to interfere with Notch activation in the sibling cell it segregates into (Fig. 9C), we wanted to examine if this activity is at the level of Notch itself or downstream in the pathway at the level of Su(H). We reasoned that if Numb acts at the level of activated Notch, overexpression of numb attenuates the N(icd) overexpression phenotype, but not that of Su(H)vp16. Indeed, it seems that increasing the level of Numb protein by eme-Gal4-mediated expression counteracts the effect of N(icd), in that more DA1 muscles and less EPCs form(Fig. 5G). By contrast,overexpression of numb together with Su(H)vp16 generates a phenotype similar to that of Su(H)vp16 overexpression alone(Fig. 5H), which suggests that Numb functions upstream of activated Su(H).
The phenotype of eme>Su(H)vp16 embryos is not as strong as with eme>N(icd), in that some DA1 muscles are still formed in some segments. This allowed us to address the question, whether or not DA1 muscle founders are siblings of EPCs, from a new angle: if they derived from the same precursor, DA1 muscle formation would never occur simultaneously with EPC duplication. By contrast, we did occasionally observe four EPCs and one DA1 muscle in the same hemisegment (Fig. 5F), consistent with an independent lineage of EPC and DA1 progenitors (Fig. 1F; Fig. 9C).
emeA-lacZ as a marker for sibling cell fates of the mesodermal Eve lineage
A recent study suggests that the sibling cell of the EPC progenitor is the DO2 muscle founder (Carmena et al.,2002). However, muscle founders fuse with surrounding myoblasts and eve expression disappears in the DO2 founder after the asymmetric cell division is completed, which makes it difficult to follow cell fate transformations between the proposed siblings. To circumvent this problem, we took advantage of a mesodermal eve enhancer, emeA(Han et al., 2002), which labels the EPCs and both the DA1 and DO2 founder nuclei, and founder expression persists even after myotubes begin to form (but, unlike with eve, the myoblast nuclei that have fused with the founders are unlabeled; Fig. 6A). In eme>dap embryos, both DA1 and DO2 founders form, but the number of EPCs is reduced to half (Fig. 6B), similar to the phenotype observed with dapoverexpression throughout the mesoderm(Fig. 1D). In the CycAmutant, only the two muscle founders are present in each hemisegment(Fig. 6C), suggesting that when the asymmetric divisions of cycle 16 are blocked both progenitors adopt muscle founder cell fate. In numb mutants, neither DA1 and nor DO2 founders form; instead, four EPCs are present (Fig. 6D). We note that the occasional formation of a DO2 founder(identified by position and absence of Eve protein) is always accompanied by the presence of two instead of four EPCs (data not shown), consistent with a lineage relationship between DO2 and EPC(Fig. 1F; Fig. 9C). Conversely, in eme>numb embryos, two pairs of DA1 and DO2 muscle founders, but no EPCs, are observed in 28% of 120 hemisegments counted(Fig. 6E). In the remaining segments, only one DO2 is observed accompanied always by two EPC with or without a duplicated DA1 founder (data not shown). This suggests again that DO2 rather than DA1 is related by lineage to the EPCs. Finally, N(icd)expression in the mesodermal Eve lineage results in a loss of DO2 and DA1 founders accompanied by the concomitant formation of four EPC in 60% of 120 hemisegments counted (Fig. 6F),or no eve or lacZ expression in the remaining segments (see also Fig. 5D).
numb directs a myogenic cell fate of undivided progenitors only in asymmetric cardiac lineages
The loss-of-function phenotype of numb is opposite to that of CycA within the EPC lineage, in that numb mutants produce only pericardial EPCs, whereas in CycA mutants only myogenic fates are specified (Fig. 6A,C,D; Fig. 7A-C). We wished to determine the epistatic relationship between these genes in the mesodermal Eve lineage by examining numb;CycA double mutants. In such double mutant embryos, we observe formation of one non-myogenic, eve-expressing cell per hemisegment, apparently at the expense of DA1 muscle formation (Fig. 7D, 100%penetrance, n=110). This phenotype is different from a EPC-only numb phenotype (Fig. 7B, 100% penetrance, n=100) or a muscle founder-only CycA phenotype (Fig. 7C, 100% penetrance, n=96), indicating that numbfunction is required in the asymmetric Eve lineages to specify a myogenic fate regardless whether the progenitor divides or not. Similarly, when N(icd) is expressed after the initial specification of mesodermal cardiac progenitors using a heatshock promoter [Hs-N(icd)], it can generate a similar cell fate transformation as that of numb or a numb;CycA mutant (Fig. 7E, 84% penetrance, n=64; Fig. 7F, 90% penetrance, n=42) (see Park et al.,1998). As other evidence suggests that the EPC progenitor is the sibling of the DO2 founder, which undergoes an asymmetric cell division at cycle 16, it is likely that the observed non-myogenic eve-expressing cells of numb;CycA or of Hs-N(icd);CycA mutants are the undivided EPC/DO2 progenitors, rather than the siblings of the DA1 founders, which apparently no longer express eve (see Carmena et al., 2002). This DA1 sibling is likely to die, because eme-lacZ expression, unlike in DO2, does not perdure in this DA1 sibling cell (Figs 6, 9).
When myocardial cell lineages were examined, we found that the number of myocardial cells in either numb or CycA mutants is four per hemisegment in segment A2-A7 (Fig. 7B,C). However, the identities of these four myocardial cells in numb mutants are different from those in CycA mutants, as indicated by Tinman expression (Fig. 8B,C; see also Fig. 2F and Fig. 9)(Ward and Skeath, 2000): in CycA mutants, two out of the four myocardial cells are SMC and the other two are TMC (Fig. 7A; Fig. 8B); whereas in numb mutants, all four myocardial cells are TMCs(Fig. 8C). In numb;CycA double mutant embryos, only two Mef2-expressing myocardial cells are specified in each A2-7 hemisegment (Fig. 7D, 94%penetrance, n=66), and both of them express tinman(Fig. 8D), suggesting the symmetrically dividing TMC progenitors maintain their myogenic fates in the absence of numb function, whereas the asymmetrically dividing SMC progenitors may assume a pericardial, non-myogenic fate. In agreement with this conclusion, the number of odd-expressing pericardial cells increases from four to six per A2-7 hemisegment of numb mutants(Fig. 8E, average 5.8, n=48), but only one is specified in the CycA mutant(Fig. 2F, average 1.1, n=55; see also Fig. 9B). In numb;CycA double mutants, the average number of odd-expressing cells per hemisegment is three(Fig. 8F, average 3.2, n=42), consistent with the two SMC progenitors acquiring a non-myogenic SOPC-type fate in these double mutants (see also Fig. 9B). Similarly, expression of N(icd) in CycA mutants also generates two TMC per hemisegment (Fig. 7H, 94%penetrance, n=86; Fig. 8H, 92% penetrance, n=62), instead of four in a otherwise wild-type background (Fig. 8G,91% penetrance, n=46). Taken together, these data suggest that numb is only required to inhibit Notch activity in asymmetric lineages that produce both myogenic and non-myogenic cell fates to promote the myogenic pathway of differentiation in the sibling that inherits Numb. By contrast, the presence or absence of Notch activation has no influence on cell fates of the symmetric myogenic-only lineages (summarized in Fig. 9).
DISCUSSION
Postblastoderm cell divisions in the cardiac mesoderm
All mesodermal cells go through three postblastoderm cell cycles (cycle 14-16) (Campos-Ortega and Hartenstein,1997). The mesodermal cells enter the first postblastoderm division (mitosis 14) at 210 minutes AED as domain 10(Foe, 1989). They are the first embryonic cells to go through the second postblastoderm division at about 250 minutes AED. The mesodermal cells are thought to divide in an approximately synchronous fashion in the first two postblastoderm cell divisions. The third division (mitosis 16) takes place during late stage 10 to early stage 11 between 280-300 minutes AED. During this division, a continuous longitudinal zone that may generate the heart precursors and part of the visceral mesoderm appears as a subdomain of mitosis 16(Campos-Ortega and Hartenstein,1997). It is likely that tinman is involved in generating this subdomain because it is specifically expressed in such a continuous longitudinal zone at stage 10 (Bodmer et al., 1990).
Mesodermal Eve lineages
We show that the asymmetric cell divisions of the Eve lineage that generates the EPC progenitors and DO2 founders are arrested in CycAor Rca1 mutants, in which cell cycle 16 is blocked(Knoblich and Lehner, 1993; Dong et al., 1997), but not in the twi>dap embryos, in which the subsequent division of the EPC progenitor is inhibited (cycle 17). Based on the effects of these genes in the ectoderm, it is likely that CycA mutant blocks the progression of cell cycle also in the mesoderm, but ectopic dap induces early exit from the cell cycle. Therefore, in CycA mutants mitosis may be blocked at a certain time point during development (such as at G2/M transition of mitosis 16), but the ectopic dap may induce skipping of the last division. It has been shown that the Numb crescent in the precursor P2, which generates the DO2 founder and the EPC progenitor, appears at late stage 10 and the division happens between late stage 10 and early stage 11(Carmena et al., 1998),consistent with this being mitosis 16 of the mesodermal cells. Therefore, it seems that many cells of the cardiac mesoderm go through three postblastoderm cell divisions (mitosis 14-16), but some of them (such as the EPC lineage)undergo an additional division (mitosis 17).
Two different models have been proposed for the mesodermal Eve lineages. One model suggested that each EPC share a progenitor with a muscle founder(Park et al., 1998). The other model suggested that the two EPCs per hemisegment share a progenitor, which in turn share a progenitor with one of the muscle founder cells (DO2), whereas the other DA1 muscle founder derives from the second progenitor(Carmena et al., 1998). The data presented in this paper strongly support the latter model (depicted in Fig. 9). The most direct evidence derives from pan-mesodermal dap overexpression, which results in the formation of a single EPC, probably because of a block the last division. By contrast, the first model predicts formation of either no EPC or two EPCs, clearly not what is observed. These conclusions are also supported by recent lineage tracing experiments (A. Alvarez and J. B. Skeath,unpublished).
Myocardial lineages along the anterior-posterior axis
Recent studies have suggested that cardiac specification along the anterior-posterior axis is under the control of homeotic genes(Lovato et al., 2002; Ponzielli et al., 2002; Lo et al., 2002) (A. Alvarez and J. B. Skeath, unpublished). For example, Svp myocardial cells are only present in the abdominal segments of the heart, but not in the thoracic segments (Gajewski et al.,2000; Lo et al.,2002); probably under the control of Antennapedia which is active in these segments (A. Alvarez and J.B. Skeath, unpublished). In this study we found that in addition to cell identity differences between the anterior two and the posterior heart segments, the lineage of the T3-A1 myocardial cells are also distinct. In cycle 16 blocking CycA or Rca1 mutants, the last cycle of myocardial divisions in T3-A1 is not arrested unlike it is the case for the posterior myocardial cells. The anterior myocardial progenitors may either undergo the last division during cycle 15 or they may be less susceptible to a loss-of-CycA-function. The first possibility is consistent with the observation that embryos with overexpression of last division inhibitor dap, but not cycle 16 arrested CycA mutants, exhibit a reduction of TMCs in T3-A1. Alternatively, it is possible that all T3-A1 myocardial lineages are asymmetric, as the Svp lineages more posterior, except they all express tinman, due to the lack of svp in these two segments(Lo et al., 2002). Thus,blocking division in CycA mutants would not alter the number of TMC in these two segments. Recent lineage tracing data are consistent with this view (A. Alvarez and J. Skeath, unpublished), but further experiments are needed to elucidate these lineages.
Extension of cardiac lineages
The second postblastoderm cell division of the mesodermal cells (mitosis 15) seems to be arrested if both CycA and CycB functions are lost. In CycA;CycB double mutants, only two of the normally six myocardial cells are formed in A2-A7 segments, one exhibiting TMC and the other SMC characteristics. Therefore, we propose that in each hemisegment two myocardial superprogenitors are specified(Fig. 9): the TMC superprogenitor (TSP) divides twice symmetrically, whereas the SMC superprogenitor (SSP) first divides symmetrically and then asymmetrically. Present and previous studies suggest there are probably five progenitors in each hemisegment that give rise to 14 heart-associated cells (six myocardial and eight pericardial): the TSP gives rise to four myocardial cells (two of which are TLMC), the SSP generates two myocardial and two SOPC, the EPC progenitor to two EPCs; the remaining four pericardial cells, two OPC and two LPC, probably derive from two symmetrically dividing precursors, although their lineage is not as well understood.
Asymmetric cell division and cell cycle progression
Asymmetric divisions have previously been studied in the context of cell cycle progression in the Drosophila PNS and CNS (Vervoort et al.,1997; Wai et al., 1999; Tio et al., 2001). In the PNS,Notch activity is required for specification of a type I versus type II neuronal fate. When sensory organ progenitor cell division is blocked in stg- mutants, the undivided precursor adopts a type II neuronal fate, whereas in numb;stg double mutants, a type I fate is chosen (Vervoort et al., 1997). In the CNS, Notch is required for specification of the sib cell fate versus the RP2 cell fate of the GMC1 asymmetric cell division. In Rca1 mutants, the undivided GMC1 adopts a RP2 fate, whereas in numb;Rca1 double mutants, the undivided GMC1 often adopts the sib cell fate(Buescher et al., 1998; Wai et al., 1999; Lear et al., 1999). Both experimental outcomes are analogous to what we observe in CycAmutants: the undivided P2 progenitor adopts a pericardial fate in the absence of numb function instead of a myogenic fate in a wild-type background. Taken together, these observations suggest that arrest of an asymmetric cell division leads the undivided progenitor to adopt the fate of the daughter cell that inherits Numb, and in the absence of Numb the alternative fate is chosen.
Notch activity promotes autonomously a non-myogenic fate in asymmetric lineages of the cardiac mesoderm irrespective of cell division
Previous studies suggested that Notch activity controls two distinct processes during the specification of cardiac cell fates(Park et al., 1998). First, it is required to single initial progenitors out of a field of competence by supporting the selection and inhibiting surrounding cells from adopting the same fate (Culi and Modolell,1998). Subsequent to the progenitor specification, Notch is required again for the specification of alternative cell fates of sibling cells produced during asymmetric cell divisions (reviewed by Jan and Jan, 1998). In this study, we have examined the cell autonomy of Notch, by using eme-Gal4 to drive activated forms of Notch and Su(H) exclusively in the mesodermal Eve lineages. We also used conditional ubiquitous expression of activated Notch to examine its lineage-specific function in other cardiac lineages. We find that Notch is required for specification of a non-myogenic fate in both the Eve and the Svp lineages of the cardiac mesoderm. By contrast, activation or inhibition of the Notch pathway did not affect cell fate decisions within the symmetric lineages. This suggests a mechanism by which cell type diversity may be increased during evolution by co-opting the Notch pathway during cell division to distinguish between alternative fates of the daughter cells. The inability of activated Su(H) to autonomously influence cell fates in symmetric cardiac lineages further suggests that other factors or activities, not present in symmetric lineages, are crucial for the asymmetric lineage-specific functions of Notch and Su(H).
Interestingly, this influence of the Notch pathway on cell fate decision in asymmetric cardiac Eve and Svp lineages is not altered when cell division is arrested. Thus, cell division is not essential to distinguish between alternative cell fates. The data also suggest that the default cell fate of a asymmetrically dividing cardiac precursor in Drosophila is determined to assume a myogenic fate, owing to Numb-mediated inhibition of Notch, unless that fate is switched by the activation of target genes downstream of Su(H). Moreover, in a double mutant of Notch and numb we would expect to observe the same lineage phenotype as of Notch alone, i.e. a myogenic cell fate, as the primary role of Numb is to inhibit Notch signaling (see also Spana and Doe,1996). Unfortunately, analysis of such double mutants is complicated by the earlier role of Notch in lateral inhibition.
Another unresolved issue is the source of the Notch ligand that activates signal transduction within asymmetric cardiac lineages. If the myogenic cell were to produce the ligand for Notch activation in its pericardial sibling,then the undivided progenitor would have to secrete its own Notch ligand. This is unlikely, as production of the ligand is usually inhibited within the cell that experiences Notch signaling (see Culi and Modellel, 1998). In the asymmetric MP2 lineage of the Drosophila CNS, for example, ligand production appears to be required in cells outside the MP2 lineage (Spana and Doe, 1996). A similar scenario may be operating in the asymmetric cardiac lineages.
Numb acts at the level of Notch in preventing Su(H) activation
Within the Eve lineages, Notch activation is mimicked by Su(H) fused to the VP16, a potent transcriptional activation domain. Recent studies suggest that in the absence of Notch activity, DNA-bound Su(H) prevents activators from promoting transcription. When Notch ligands, such as Delta, bind to its receptor, Notch is cleaved to produce an intracellular domain fragment,N(icd), which is thought to enter the nucleus and interact directly with Su(H)to recruit transcriptional co-activators and alleviate Hairless-mediated repression, thus promoting transcription (for a review, see Bray and Furriols, 2001; Barolo and Posakony, 2002). In support of this model, we find that Su(H) overexpression can mimic Notch activation only when linked directly to a transcriptional activator, but not in its wild-type form when it presumably associates with co-repressors, such as Hairless and Groucho (Barolo et al.,2002), that prevent Su(H)-dependent transcriptional activation in the absence of Notch signaling.
The role of the PTB-containing, membrane-associated protein Numb in preventing Notch activation in the nervous system is well established (for a review, see Kopan and Turner,1996; Jan and Jan,1998; Jafar-Nejad et al.,2002). To explore at which level Notch signaling is inhibited by Numb in the cardiac lineages, we overexpressed numb simultaneously with N(icd) or Su(H)vp16 within the mesodermal Eve lineages. Excess Numb was able to counteract activated Notch but not activated Su(H)function, suggesting that Numb can inhibit Notch activity after it has been cleaved, possibly by preventing its nuclear translocation, but is unlikely to prevent the transcriptional activator function of Su(H) directly. Recent data suggest that Numb is involved in stimulating endocytosis of Notch, thus removing it from the cell surface and inhibiting its function(Berdnik et al., 2002). It is not clear, however, if this inhibition by endocytosis is at the level of the entire Notch receptor, or (also) at the level of N(icd) after it is cleaved off. Our experiments provide strong evidence that Numb can indeed interfere with N(icd) function, but it remains to be determined if endocytosis is an obligatory intermediate in this inhibition of activated Notch.
Notch activity may specify a pericardial cell fate in both Drosophila and vertebrates by similar mechanisms
A recent study in Xenopus suggests that the Notch pathway may also have a role in vertebrates in specifying pericardial and other non-myogenic cell fates within the dorsolateral cardiogenic region of the anterolateral plate mesoderm (Rones et al.,2000). As in the Eve and Svp lineage of the Drosophilaheart, activation of the Notch pathway decreased myocardial gene expression and increased expression of a pericardial marker, whereas inhibition of Notch signaling resulted in an increase of cardiac myogenesis. Similar results were obtained with an activated form of RBP-J [a vertebrate homolog of Drosophila Su(H) fused to vp16, as in our study](Rones et al., 2000). These data indicate that the Notch pathway may play a role in the specification of myocardial versus pericardial cell fates in both Drosophila and vertebrates. This raises the question of whether the mechanism of Notch mediated cell identity determination is also conserved between vertebrates and flies. Because it is not yet known if (Numb-controlled) asymmetric cell divisions are also involved in vertebrate heart development, the answer awaits future studies. However, recent studies on the role of Numb during cortical development suggest that it is likely to have a similar control function in cell fate specification in vertebrates as it does in flies(Shen et al., 2002).
Acknowledgements
We thank C. Lehner for helpful suggestions and discussions; the Bloomington stock center and C. Lehner for sending Drosophila stocks; J. Skeath for sharing data prior to publication; and C. Lehner, M. Mercola and J. Skeath for comments on the manuscript in preparation. This work was supported by an American Heart Association Predoctoral Fellowship to Z.H. and by grants from the NHLBI of NIH to R.B.