|Home | About | Journals | Submit | Contact Us | Français|
How adjacent organ fields communicate during development is not understood. Here, we identify a mechanism in which signaling within the forelimb field restricts the potential of the neighboring heart field. In zebrafish embryos deficient in retinoic acid (RA) signaling, the pectoral fins (forelimbs) are lost while both chambers of the heart are enlarged. We provide evidence that both of these phenotypes are due to RA signaling acting directly within the forelimb field. hoxb5b, an RA-responsive gene expressed within the forelimb field, is required to restrict the number of atrial cells arising from the adjacent heart field, although its function is dispensable for forelimb formation. Together, these data indicate non-autonomous influences downstream of RA signaling that act to limit individual chamber size. Therefore, our results offer new perspectives on the mechanisms regulating organ size and the possible causes of congenital syndromes affecting both the heart and forelimb.
The early vertebrate embryo can be thought of as a collage of organ fields, partially overlapping zones of developmental potential that envelop their respective pools of organ progenitor cells (Fishman and Chien, 1997; Huxley and deBeer, 1934; Jacobson and Sater, 1988). Over time, refinement of developmental potential narrows each field into the precise dimensions of the organ primordium. The gradual restriction of organ fields is therefore one important mechanism for controlling organ size. The proximity and overlap of developing fields suggests that communication between adjacent territories could contribute to organ field restriction, but little is known about whether neighboring fields coordinate their development.
The embryonic heart field is an interesting example of a zone of developmental potential that becomes restricted over time. Several studies have demonstrated that cardiac developmental potential extends beyond the boundaries of the region of the lateral plate mesoderm (LPM) that normally becomes myocardium. For example, Notch signaling plays a role in repressing cardiac specification in a lateral portion of the Xenopus LPM that normally forms mesocardium and pericardium (Raffin et al., 2000; Rones et al., 2000). In zebrafish, components of the vascular and hematopoietic specification pathways, particularly the transcription factors Scl and Etsrp, inhibit cardiac specification in rostral LPM (Schoenebeck et al., 2007). There appear to be multiple pathways that restrict the plasticity of the heart field, and it is likely that additional inhibitory mechanisms remain to be characterized.
Our recent studies have revealed that the retinoic acid (RA) signaling pathway plays a potent role in limiting cardiac specification. Zebrafish embryos lacking RA signaling exhibit a surplus of cardiomyocytes, a consequence of an excess of cardiac progenitor cells (Keegan et al., 2005). More recently, it has also been demonstrated that mouse embryos lacking the RA synthesis enzyme Raldh2 display an expansion of the second heart field (SHF), a territory that contributes to both the inflow and outflow poles of the amniote heart (Ryckebusch et al., 2008; Sirbu et al., 2008). It is not yet apparent how this newly identified early role of RA in restricting the SHF fits together with a previously hypothesized role of RA signaling during cardiac fate assignment in amniotes. Specifically, it has been proposed that RA signaling promotes atrial cell identity within the heart field, thereby defining the relative proportions of atrial and ventricular cells (Hochgreb et al., 2003; Xavier-Neto et al., 1999; Xavier-Neto et al., 2001). It remains unknown whether the proposed roles of RA in promoting atrial identity and in restricting the SHF are distinct or overlapping. Moreover, it is unclear where and how RA acts to restrict cardiac specification.
The impact of RA signaling on the forelimb field may provide clues to the mechanisms by which RA restricts the size of the heart field. In both zebrafish and mouse embryos that lack RA signaling, there is a converse relationship between heart and forelimb formation: forelimbs are lost, while the number of cardiac cells is increased (Keegan et al., 2005; Niederreither et al., 1999; Ryckebusch et al., 2008; Sirbu et al., 2008). Previous studies in both organisms have indicated that RA signaling induces formation of the forelimb field by promoting tbx5 expression (Begemann et al., 2001; Gibert et al., 2006; Mercader et al., 2006; Mic et al., 2004). However, it is not known whether this reflects a direct requirement for RA signaling within forelimb progenitor cells. It is tempting to speculate about a connection between the mechanisms that result in forelimb deficiency and cardiac surplus. One possibility is that reduction of RA signaling converts forelimb progenitors into heart progenitors, or perhaps effects of RA on one field have indirect consequences for the other. Alternatively, RA may impact the development of each field independently. The possibility of coordinate regulation of the heart and forelimb fields is likely to be relevant to the causes of human congenital defects that affect both tissues, including inherited heart-hand syndromes (Wilson, 1998).
In this study, we sought to address where and how RA signaling limits the number of cardiac progenitor cells in zebrafish. We find that RA signaling restricts the numbers of both atrial and ventricular cardiomyocytes, through independent effects on each lineage. Loss of RA signaling leads to a posterior extension of both atrial and ventricular progenitor cells into territory normally occupied by forelimb progenitor cells. We show that RA signaling is required cell-autonomously for forelimb formation. However, posterior extension of the heart field in embryos lacking RA signaling does not seem to reflect a transformation of forelimb progenitor cells into cardiac progenitor cells. Additionally, we demonstrate that expression of hoxb5b in the forelimb field is RA-responsive. Strikingly, hoxb5b plays an essential and non-autonomous role in restricting the number of atrial cardiomyocytes that emerge from the heart field. Together, our data indicate that RA signaling, through its direct effect on forelimb field formation, acts indirectly to limit the specification of cardiac progenitors within the heart field; thus, interactions between adjacent organ fields play an important role in the refinement of developmental potential. These findings provide a new perspective on the regulation of organ size and may yield insight into the etiology of human heart-hand syndromes.
We have recently demonstrated that RA signaling plays an important role in restricting the production of cardiac progenitor cells in the zebrafish embryo (Keegan et al., 2005). However, this study did not elucidate the mechanisms downstream of RA signaling that are responsible for heart field restriction. As a first step toward this goal, we sought to determine whether RA signaling has comparable effects on both the atrial and ventricular lineages. Therefore, we examined cardiac chamber formation in embryos that were treated with either the retinaldehyde dehydrogenase inhibitor DEAB (Russo et al., 1988) or the RA receptor antagonist BMS189453 (BMS) (Schulze et al., 2001) beginning at 40% epiboly, just before the onset of gastrulation.
At 48 hours post-fertilization (hpf), atrial and ventricular chambers were distinct and easily identifiable in wild-type embryos (Figure 1A), as well as in embryos treated with DEAB or BMS (Figures 1B and 1C). However, the hearts of DEAB-treated or BMS-treated embryos were highly dysmorphic and typically appeared enlarged. We counted the cardiomyocytes in each chamber to determine if aberrant morphology reflected a change in cell number. Both DEAB and BMS treatments led to a significantly increased number of atrial and ventricular cells (Figures 1D and 1E; Supplemental Table 1). Increases in atrial and ventricular populations were also apparent prior to heart tube formation, as indicated by the expression patterns of the earliest known chamber-specific genes, atrial myosin heavy chain (amhc) (Berdougo et al., 2003) and ventricular myosin heavy chain (vmhc) (Yelon et al., 1999) (Figures 1F–1K; Supplemental Table 2). The phenotypes resulting from treatments with BMS or DEAB resembled those resulting from treatment with Ro41-5253 (RO, an RARα-specific antagonist) or from loss of raldh2 gene function (Supplemental Figures 1 and 2; Supplemental Tables 1 and 2; JSW and DY, unpublished data). Together, these data show that RA signaling is required to restrict the numbers of both atrial and ventricular cells in zebrafish.
To refine our understanding of when RA signaling is required to limit the numbers of atrial and ventricular cells, we next compared the effects of initiating BMS treatment at different developmental stages. We found a progressive decrease in sensitivity to BMS over time. Both atrial and ventricular populations are significantly enhanced when BMS treatment occurs during gastrulation (Figures 1D and 1E; Supplemental Table 1), similar to our previous findings (Keegan et al., 2005). However, by the 6–8 somite stage, only atrial cells, not ventricular cells, are affected by BMS treatment (Figures 1D and 1E; Supplemental Table 1). This temporal difference in sensitivity to BMS treatment is compatible with the notion that ventricular differentiation normally precedes atrial differentiation (Berdougo et al., 2003) and suggests that RA signaling might limit atrial and ventricular cell number through different mechanisms.
To determine whether RA signaling has a similar impact on both atrial and ventricular progenitor specification, we next examined cardiomyocyte progenitor fate maps from wild-type and BMS-treated embryos (Supplemental Figure 3, Supplemental Table 3). Comparison of the wild-type and BMS-treated fate maps demonstrated that BMS treatment does not significantly disrupt the relative spatial organization of atrial progenitors (APs) and ventricular progenitors (VPs) (Figures 1L and 1M; Supplemental Figure 3; Supplemental Table 3). Importantly, we did not find VPs in the territory where normally only APs are found, or vice versa (Figures 1L and 1M; Supplemental Figure 3), suggesting that neither increase in cell number occurs at the expense of the other lineage.
To determine whether the increases in ventricular and atrial cell number in BMS-treated embryos are due to increased numbers of VPs and APs, we compared the frequency of encountering each type of progenitor within their respective territories in the wild-type and BMS-treated fate maps. We encountered VPs almost twice as often within the ventricular territory of BMS-treated embryos as we did within the ventricular territory of wild-type embryos (p<0.005; Figures 1L-1N; Supplemental Figure 3; Supplemental Table 3). In contrast, the frequency of encountering APs within the atrial territory did not change in the BMS-treated fate map (Figures 1L-1N; Supplemental Table 3). These data indicate an increase in the number of VPs in the lateral margin of BMS-treated embryos at 40% epiboly, while the number of APs appears unaffected by BMS treatment at this stage.
Although BMS treatment did not affect the density of APs at 40% epiboly, we did observe a striking effect of BMS on the number of labeled progeny produced per AP. Our fate map data indicated that APs in BMS-treated embryos generated an increased number of atrial cardiomyocytes compared to wild-type APs (p<0.05; Figure 1O; Supplemental Table 3). In contrast, we did not observe an increased number of progeny from VPs in BMS-treated embryos compared to the number of progeny from wild-type VPs (Figure 1O; Supplemental Table 3). Therefore, our fate map data suggest that RA signaling restricts the number of atrial cells at a later stage than when it restricts the number of ventricular cells. These data are compatible with the temporal differences in atrial and ventricular sensitivity to BMS (Figures 1D and 1E), suggesting different timeframes for independent processes of ventricular and atrial specification.
As a next step toward understanding the mechanism through which RA signaling restricts cardiac chamber size, we wanted to identify RA-responsive genes that are expressed within the LPM during the time interval when RA signaling is required to limit cardiomyocyte production. Although there were some known targets of RA signaling, their relevance to cardiomyocyte production was unclear. Using microarrays followed by validation via in situ hybridization, we identified 12 RA-responsive genes expressed within the LPM; two examples that were positively regulated by RA signaling were hoxb5b and retinol short-chain dehydrogenase/reductase (retSDR1/dhrs3; referred to here as retd) (Figure 2; Supplemental Figure 4).
Our identification of RA-responsive genes within the LPM suggested locations of RA signaling that might be related to the role of RA in restricting cardiac cell number. However, none of these genes appeared to be expressed within the heart field; instead, they appeared to be in a more posterior portion of the LPM. For example, at the 6–8 somite stages, when robust expression of gata4 and nkx2.5 is first evident in the heart field (Schoenebeck et al., 2007), retd is found in an adjacent posterior portion of the LPM (Supplemental Figures 5D–5I), and hoxb5b is found at an even further posterior position (Supplemental Figures 5J–5L). This observation was confirmed by measuring the distances between the posterior and anterior limits of gene expression, relative to the tip of the notochord at the 8 somite stage (Figures 3A–3E). The posterior limits of gata4 and nkx2.5 expression were on average at 28 and 27 notochord cell diameters (ncds), respectively (Figures 3A and 3B; Supplemental Table 4), while the anterior limits of retd and hoxb5b expression were on average at 36 and 63 ncds, respectively (Figures 3C and 3D; Supplemental Table 4). The relatively posterior expression of RA-responsive genes correlates with the even more posterior location of raldh2 expression. At these stages, raldh2 is expressed at high levels in the somites, the most anterior of which is located on average at 72 ncds (Figure 3E; Supplemental Figures 5A and 5B). The distance between raldh2 and gata4 expression in zebrafish contrasts with data from mouse and chick embryos where raldh2 expression is reported to be adjacent to gata4 at comparable stages (Hochgreb et al., 2003). Although raldh2 begins to be expressed more anteriorly in the ectoderm in zebrafish at the 8 somite stage (Supplemental Figures 5C and 5D; Grandel et al., 2002), this expression begins at the end of the time period during which RA signaling is required to limit cardiac cell number (Figures 1D and 1E). Thus, the locations of RA-responsive genes and RA synthesis suggest that RA signaling is occurring in a portion of the LPM posterior to the heart field.
We next wanted to understand how the expression patterns of RA-responsive genes related to the various progenitor populations present within the LPM. Therefore, we constructed a fate map of the relevant portion of the LPM at the 8 somite stage in wild-type embryos (Supplemental Figure 6), a posterior extension of our prior fate map of the most anterior portion of the LPM (Schoenebeck et al., 2007). Our results indicated that the LPM territories containing AP and VP cells extend posteriorly to approximately 50 and 40 ncds, respectively (Figure 3F). Thus, the locations of cardiac progenitors (CPs) correlate relatively well with the expression patterns of gata4 and nkx2.5, although there may be a posterior population of CPs that does not express these genes at this stage. By comparison, pectoral fin (forelimb) progenitors (FPs) were found in a territory beginning at approximately 40 ncds and extending to 60–70 ncds, corresponding to the locations of retd and hoxb5b expression. Together, our fate map data indicate that RA-responsive genes in the LPM are expressed primarily posterior to the heart field and within the forelimb field.
Having determined the relative positions of CPs and FPs within the wild-type LPM, we wanted to examine how loss of RA signaling and the consequential loss of FPs affects the distribution of CPs. First, we examined LPM gene expression patterns in embryos deficient in RA signaling. We found a significant lengthening of gata4 expression in the LPM of BMS-treated, DEAB-treated, and RO-treated embryos (Figures 4A–4D, Supplemental Table 5). In all cases, the change in gata4 expression is considerably more prominent posterior to the notochord tip than it is anterior to the notochord tip. We also found similar posterior extensions of nkx2.5 expression in treated embryos (JSW and DY unpublished data). Together, these results suggest that RA signaling restricts the posterior extent of cardiac gene expression within the LPM.
We were intrigued by this posterior extension of cardiac gene expression, as it suggested an enlargement of the heart field into territory that would normally contain the forelimb field. Consistent with this, we detected an altered distribution of tbx5 expression in RA signaling-deficient embryos (Figures 4E–4H). tbx5 is expressed in both the heart and forelimb (Ahn et al., 2002; Begemann and Ingham, 2000; Garrity et al., 2002). By the 12–14 somite stage, two slightly separated fields of tbx5 expression appear to approximate the locations of the CPs and FPs (Figure 4E and Ahn et al., 2002). In embryos deficient in RA signaling, the separation between populations is less evident, suggesting an enlargement of the anterior population and a reduction or loss of the posterior population (Figures 4F–4H). These results confirm and extend the previous observation of a loss of tbx5 expression in the forelimb field of raldh2 mutant embryos at the 12 somite stage (Begemann et al., 2001). To examine whether the posterior extension of cardiac markers corresponds to a posterior extension of CPs, we constructed LPM fate maps in RA signaling-deficient embryos. Consistent with our gene expression analysis, the fate map data indicated CPs residing in posterior LPM territories that are normally occupied by FPs (Figures 3G–3I). Therefore, our data indicate that RA signaling restricts posterior extension of the heart field within the LPM.
The posterior expansion of CPs into a region normally occupied by FPs suggested that a loss of RA signaling may transform FPs into CPs. One possible scenario, suggested by the expression patterns of RA-responsive genes, is that cells in the forelimb field may normally need to receive RA signaling in order to become FPs rather than CPs. We performed mosaic analysis to determine whether RA signaling-deficient cells would preferentially contribute to the heart rather than to the forelimb. We used two different strategies to deplete RA signal transduction in a cell-autonomous fashion: transgenic activation of a dominant negative RA receptor or BMS treatment (Supplemental Figure 7). After transplanting cells from Tg(hsp70:dnRARα) or BMS-treated donor embryos into wild-type host embryos, we assessed contributions of donor cells to the heart and forelimb (Figures 5A–5F). With either method of autonomously antagonizing RA signaling, we found a significant decrease in the frequency with which donor cells contribute to the fin mesenchyme (Figures 5G and 5H). Despite this indication of a cell-autonomous requirement for RA signaling during fin formation, we did not find a corresponding increase in the frequency of donor cells becoming cardiomyocytes, as would be expected if FPs were transformed into CPs (Figures 5G and 5H). Therefore, our data suggest that RA signal transduction is required cell-autonomously to promote FP formation and is most likely acting indirectly to restrict cardiac cell number.
Having established where RA signaling is acting, we next wanted to determine if any of the RA-responsive genes expressed in the forelimb field are required to limit cardiac cell number. Although anti-retd morpholinos (MOs) do not appear to affect fin formation or restrict cardiac cell number (JSW and DY, unpublished data), anti-hoxb5b MOs do seem to disrupt heart formation. To examine the function of hoxb5b, we used MOs designed to block proper splicing of hoxb5b transcripts (Figure 6A). Despite the high similarity in target sequences, anti-hoxb5b and anti-hoxb5a MOs specifically abrogated splicing of their respective transcripts (Figure 6A and 6F). hoxb5b morphants (anti-hoxb5b MO-injected embryos) exhibit enlarged hearts and pericardial edema (Figure 6G), reminiscent of embryos with defective RA signaling. However, hoxb5b morphants have overtly normal fins (Figure 6H), unlike embryos lacking RA signaling. Although hoxb5a MOs effectively block splicing and reduce levels of wild-type hoxb5a transcript (Figure 6A), hoxb5a morphants resembled wild-type embryos (Figures 6D and 6E). The enlarged hearts of hoxb5b morphants were especially intriguing (Figure 6G), so we assessed the number of cells in each cardiac chamber. Strikingly, hoxb5b morphants contain an increased number of atrial cells at 48 hpf and an increased number of cells expressing amhc prior to heart tube assembly (Figures 6I, 6J, 6L–6O, 6Q, and 6R; Supplemental Tables 1 and 2). However, neither ventricular cell number nor vmhc expression are affected in hoxb5b morphants (Figures 6I–6K, 6M–6P, and 6R; Supplemental Tables 1 and 2). Interestingly, the increases in atrial cell number were similar to those found in DEAB-treated and RO-treated embryos (Figures 1D and 1E and Supplemental Figures 2T and 2Y). We therefore wondered whether any APs are found within the forelimb field of hoxb5b morphants, as was observed in RA signaling-deficient embryos. Fate mapping of hoxb5b morphants revealed that atrial cells do occasionally originate within the forelimb field, although not nearly as frequently as in the fate maps of embryos deficient in RA signaling (Figures 3G–3J). Consistent with this, we did not observe a posterior expansion of gata4 or nkx2.5 at the 8 somite stage in hoxb5b morphants (Supplemental Table 6). This suggests that the majority of the excess atrial cells in hoxb5b morphants arise from the heart field.
The distance between hoxb5b expression and the heart field (Figures 3A, 3B, 3D, and 3F; Supplemental Figures 5J–5L) strongly suggested that hoxb5b acts through a non-autonomous mechanism to restrict atrial cell number. We performed two types of mosaic analysis to test this hypothesis. First, we examined whether a hoxb5b-deficient environment would affect the contribution of wild-type cells to the atrium. We transplanted cells from wild-type donor embryos into hoxb5b morphant or hoxb5a morphant host embryos, the latter of which served as controls (Figures 7A–7F). We found a 1.5-fold increase in the frequency with which donor cells contributed to the atrium in hoxb5b morphant hosts relative to control hoxb5a morphant hosts (Figure 7G), correlating with the 1.5-fold difference in atrial cell number between hoxb5b morphants and wild-type embryos. We did not detect a comparable difference in the frequency of donor cells contributing to the ventricle (Figure 7G). These results suggest that hoxb5b plays a cell non-autonomous role in restricting atrial cardiomyocyte formation.
Next, we tested whether ectopic sources of hoxb5b, outside of the heart field, can rescue the atrial cardiomyocyte surplus found in hoxb5b morphants. In these experiments, we transplanted cells from donor embryos injected with a hyperactive hoxb5b (vp16-hoxb5b) mRNA into hoxb5b morphant hosts; transplants of hoxb5b morphant donor cells into hoxb5b morphant hosts served as controls (Figures 7H–7K). We then selected chimeras in which donor cells contributed to anterior mesodermal tissues, such as the anterior somites, but did not contribute to the heart, and we assessed cardiac cell number in these embryos. As expected in the control experiments, donor cells from hoxb5b morphants had no effect on the hoxb5b morphant host phenotype; just as was seen in hoxb5b morphants, these chimeras exhibited an 1.5-fold increase in atrial cell number and no significant change in ventricular cell number, relative to wild-type (Figures 7L, 7M, 6O, and 6R). In contrast, donor cells expressing hyperactive hoxb5b mRNA caused a statistically significant (p< 0.05) reduction of the number of atrial cells in hoxb5b morphant hosts (Figures 7L and 7M; Supplemental Table 1). Thus, we find that expression of hoxb5b outside of the heart field can function to restrict atrial cell number. Altogether, we conclude that expression of hoxb5b, which is regulated by RA signaling within the forelimb field, plays an essential and cell non-autonomous role in restricting the production of atrial cardiomyocytes.
Synthesizing our data, we propose a model in which interaction between the forelimb field and the heart field acts to restrict the number of cardiac progenitor cells. The forelimb and heart fields are juxtaposed during early stages of LPM development, and RA signaling acts cell-autonomously within the forelimb field to induce formation of FPs and expression of RA-responsive genes, including hoxb5b (Figure 7N). These effects of RA signaling indirectly result in repression of both AP and VP formation within the heart field, and, in embryos deficient in RA signaling, the expanded AP and VP populations occupy the space created by the loss of the forelimb field (Figure 7O). Hoxb5b plays a key role in the restriction of CP formation; although Hoxb5b is not essential for forelimb formation, it is required to limit atrial cell number through a cell non-autonomous mechanism (Figure 7P). Thus, we conclude that communication from the forelimb field, mediated at least in part by genes downstream of Hoxb5b, instructs the heart field to limit production of atrial cardiomyocytes, and we speculate that an independent pathway downstream of RA signaling plays a similar role in limiting ventricular cell number.
Our studies offer insight into the mechanisms by which RA signaling both promotes forelimb development and restricts heart size. The direct impact of RA signaling on the forelimb appears relatively straightforward: RA signaling promotes FP formation in the forelimb field. This role is consistent with prior work in zebrafish and mouse suggesting that RA produced by the paraxial mesoderm initiates forelimb development and promotes tbx5 expression in the forelimb field, although previous studies did not demonstrate the cell-autonomy of this role of RA signaling (Begemann et al., 2001; Capdevila and Izpisua Belmonte, 2001; Gibert et al., 2006; Grandel et al., 2002; Mic et al., 2004; Robert and Lallemand, 2006). It is unclear what happens to aspiring FPs when RA signaling is inhibited. CPs and FPs originate from nearly the same zone of the 40% epiboly embryo (Keegan et al., 2004), and the anterior-posterior organization of CPs and FPs emerges during or shortly after gastrulation as the LPM forms. In RA signaling-deficient embryos, the posterior extension of CPs into the region normally occupied by FPs does not represent a canonical anterior-posterior transformation, since FPs do not seem to be transformed into CPs. Instead, we envision that CPs could come to occupy a more posterior position in the LPM by passively filling the FP void, which may be why we do not see a major posterior displacement of APs in hoxb5b morphants. Perhaps the aspiring FPs transform into another, non-myocardial lineage, although we have not discerned increases in other cell types, including the endocardium (JSW and DY, unpublished data), in RA signaling-deficient embryos, consistent with our previous analysis of the endocardial lineage in BMS-treated embryos (Keegan et al., 2005). Alternatively, FPs may simply fail to thrive without early receipt of RA signaling; we have not found evidence of decreased proliferation or cell death within the forelimb field of RA signaling-deficient embryos (Supplemental Table 7), but a gradual loss of cells would be difficult to detect.
In contrast to the direct role of RA within the forelimb field, the indirect impact of RA signaling on the heart field relies on the effects of RA-responsive genes, like hoxb5b, that are expressed in the forelimb field. The pectoral fins appear intact in hoxb5b morphants that exhibit a surplus of atrial cells, again emphasizing that the forelimb field is not likely to be the source of extra cardiac cells in this scenario. Additional CPs may form via fate transformation of a yet unidentified lineage or via increased proliferation of CPs. The latter model is particularly appealing to consider for the origin of extra atrial cells, since our 40% epiboly fate maps suggest that RA signaling limits the number of cardiomyocytes produced by individual APs (Figure 1O). We have not been able to detect differences in proliferation within the heart fields of RA signaling-deficient embryos through examination of phospho-histone H3 modification or BrdU incorporation (Supplemental Table 7; JSW and DY, unpublished data), but these negative data cannot rule out a modest or gradual impact on CP cell cycle length. Alternatively, the emergence of additional CPs may reflect utilization of otherwise dormant developmental potential that resides within the heart field.
Our data indicate that RA acts via separable mechanisms to limit the numbers of both ventricular and atrial cardiomyocytes. Several aspects of our conclusions are consistent with previous studies in amniotes. Our observation of a posterior extension of the heart field in zebrafish deficient in RA signaling is reminiscent of the posterior extension of the SHF in mice lacking Raldh2 (Ryckebusch et al., 2008; Sirbu et al., 2008). Likewise, our fate maps at the 8 somite stage (Figures 3F–3I) are comparable to chick fate maps demonstrating that reduced RA signaling causes a posterior expansion of VPs into territory normally containing APs (Hochgreb et al., 2003). In the chick embryo, this expansion is thought to be at the expense of atrial progenitors, although a tradeoff with non-cardiac lineages remains a possibility (Hochgreb et al., 2003).
Despite similarities in the roles of RA across species, our data do not support a model in which RA signaling is necessary to partition the heart by promoting formation of atrial cardiomyocytes at the expense of ventricular cardiomyocytes, as suggested by previous analyses of amniotes (Hochgreb et al., 2003; Simoes-Costa et al., 2005; Xavier-Neto et al., 1999; Xavier-Neto et al., 2001; Yutzey and Kirby, 2002; Yutzey et al., 1994). It is clear from our studies in zebrafish that loss of RA signaling increases atrial cell number. Furthermore, although reduced RA signaling causes increased ventricular cell number in zebrafish, this increase does not occur at the expense of the atrium. It is interesting to note that the excess SHF progenitors in Raldh2 mutant mice fail to differentiate normally (Ryckebusch et al., 2008). Therefore, we speculate that RA may play similar roles in fish and amniotes at early stages, during restriction of progenitor specification, while also playing different roles during later steps of differentiation and morphogenesis (Chen et al., 1998; Chen et al., 2002; Kastner et al., 1997; Niederreither et al., 1999; Niederreither et al., 2001). Future studies in amniotes should help to resolve the distinctions between roles of RA at different stages.
The proposed role for RA signaling in partitioning the amniote heart has also led to an evolutionary model that the two-chambered fish heart evolved via subdivision of a single chamber (Simoes-Costa et al., 2005; Xavier-Neto et al., 2001; Yutzey and Kirby, 2002). However, it has also been proposed that the atrium and ventricle arose from independent evolutionary origins (Fishman and Chien, 1997). Our data regarding the independent regulation of atrial and ventricular lineages in zebrafish lend support to the latter model, in which atria and ventricles arose independently, rather than as RA-mediated subdivisions of a single field (Simoes-Costa et al., 2005; Xavier-Neto et al., 2001; Yutzey and Kirby, 2002). Our data therefore suggest that RA signaling does not seem to be an ancestral, conserved mechanism for partitioning the heart field into AP and VP populations.
Our demonstration that hoxb5b acts downstream of RA signaling to restrict cardiac cell number suggests that the effects of RA on the heart field are mediated indirectly via genes expressed elsewhere, although this does not exclude the possibility that RA signaling may also have additional roles within the heart field. We favor the interpretation that hoxb5b regulates heart field size from its site of expression in the forelimb field, but we cannot rule out that it influences the heart via its expression in the neural tube. However, numerous other hox genes, which may function redundantly with hoxb5b, are expressed together with hoxb5b in the ectoderm (Prince et al., 1998a; Prince et al., 1998b), while we have yet to find another hox gene with similar expression in the forelimb field. We suspect that the non-autonomous effect of hoxb5b on the heart field involves downstream transcriptional targets that allow the forelimb field to communicate with the adjacent heart field. One possibility is that this communication occurs through the control of a morphogen gradient, as has been shown for Hox control of organ size in Drosophila (Crickmore and Mann, 2006, 2007). The specific impact of hoxb5b function on atrial, but not ventricular, cell numbers in zebrafish supports a model in which RA signaling limits atrial and ventricular cell numbers through independent downstream pathways. Identification of the relevant targets of Hoxb5b and elucidation of the pathway limiting ventricular cell number will both be important future endeavors.
It is important to consider whether the role of a Hox gene in delimiting the heart field may be conserved in amniotes. Although mouse Hoxb5 mutants display homeotic defects in vertebral patterning and a shift in position of the shoulder girdle, they do not appear to exhibit heart defects (McIntyre et al., 2007; Rancourt et al., 1995). Perhaps subtle cardiac defects are indeed present in Hoxb5 mutants; addition of extra atrial cells might not be immediately obvious or lethal. Alternatively, other mouse Hox genes may be contributing to the role played by zebrafish hoxb5b. Multiple Hox genes are expressed in or near the amniote heart field (Ryckebusch et al., 2008; Searcy and Yutzey, 1998), and the mouse Hoxa3 mutant has been reported to have defects in cardiac development (Chisaka and Capecchi, 1991; Lo and Frasch, 2003). Hoxa3 mutants display a number of intriguing phenotypes that could be related to alteration of cardiac cell number, including atrial hypertrophy and enlargement of the great veins, as well as other phenotypes such as patent ductus arteriosus, stenosis of the aortic valve, and improperly formed pulmonary valves. Although these defects have been attributed to problems with neural crest formation (Chisaka and Capecchi, 1991; Lo and Frasch, 2003), it is tempting to speculate that Hoxa3 could also influence the heart fields.
Overall, our data provide evidence for an unexpected mechanism of interaction between the forelimb and heart fields during their development. Despite the eventual distance between the heart and the forelimbs, the juxtaposition of their respective organ fields at early stages has a potent effect on the restriction of heart size. Although the precise spatial relationship between the heart and forelimb fields in amniotes is not yet clear, it is likely that they are juxtaposed as they are in zebrafish (Wilson, 1998). The interaction between heart and forelimb fields is particularly germane to the greater than 100 heritable congenital syndromes that feature both heart and forelimb defects (Wilson, 1998). Only a few of these heart-hand syndromes have been linked to their causative genes; one notable example is Holt-Oram syndrome, which is caused by mutation of Tbx5 (Basson et al., 1997; Li et al., 1997). Since Tbx5 is expressed in both the heart and forelimb fields, it is thought to autonomously affect both tissues. Our demonstration that genes in the forelimb field can influence the heart field predicts an alternative, non-autonomous mechanism for the etiologies of heart-hand syndromes: initial defects in forelimb development may indirectly impact heart development. Given this interaction, we speculate that there may be reciprocal influences of the heart field on the forelimb field. Altogether, the identification of the restriction of heart field potential by the forelimb field suggests a general mechanism by which communication between adjacent organ fields contributes to the regulation of organ size.
Tg(cmlc2:DsRed2-nuc)f2 embryos (Mably et al., 2003) were fixed at 48 hpf with 1% formaldehyde in PBS. Embryos were stained with the anti-Amhc monoclonal antibody S46 (Stainier and Fishman, 1992; gift of F. Stockdale), and S46 was detected using a goat anti-mouse IgG1 secondary antibody conjugated to FITC (Southern Biotechnology Associates). To determine the number of cells in each chamber, embryos were gently flattened under a cover slip, and the fluorescent nuclei in each chamber were counted.
In situ hybridization was performed as described by Oxtoby and Jowett (1993). All probes used were previously described, with the exceptions of retd (IMAGE Clone ID - IRAKp961D08178Q2), hoxb5a (IMAGE Clone ID - IMAGp998E1515251Q1), and hoxb5b (IMAGE Clone ID - IRALp962P2160Q2). Areas of expression of amhc and vmhc were measured using ImageJ (http://rsb.info.nih.gov/ij/index.html).
For all three antagonists of RA signaling, we employed doses that cause heart, hindbrain, forelimb, and pharyngeal endoderm phenotypes resembling those caused by genetic reduction of RA signaling in zebrafish raldh2 mutants (Begemann et al., 2004; Gibert et al., 2006; Keegan et al., 2005; Kopinke et al., 2006; Maves and Kimmel, 2005). For treatment details, see Supplemental Experimental Procedures.
We compared gene expression profiles from wild-type, BMS-treated, DEAB-treated, and RA-treated embryos at tailbud and 8 somite stages using Affymetrix zebrafish GeneChips. For further details, see Supplemental Experimental Procedures.
Fate mapping at 40% epiboly was performed as previously described (Keegan et al., 2005; Keegan et al., 2004). Fate mapping at the 8 somite stage was performed as previously described (Schoenebeck et al., 2007), with slight modifications. See Supplemental Experimental Procedures for further details.
The anti-raldh2 MO was reported previously (Begemann et al., 2001) and was injected with an anti-p53 MO to reduce toxicity (Robu et al., 2007). To knockdown hoxb5a or hoxb5b, we injected 25 ng of either anti-hoxb5a MO (5′-aatacgtatgtaccatggctaatat-3′) or anti-hoxb5b MO (5′-agatgtttataccatggctaatgtg-3′).
For ectopic expression of hoxb5b, embryos were injected with 30 pg of vp16-hoxb5b mRNA. This construct was made by fusing the vp16 activation domain directly to the N-terminus of zebrafish hoxb5b. Overexpression of vp16-hoxb5b mRNA posteriorizes embryos in a manner similar to that reported for wild-type hoxb5b mRNA (Supplemental Figure 8; Bruce et al., 2001). We consider vp16-hoxb5b to be a hyperactive form of hoxb5b, since it appears to be more potent than wild-type hoxb5b.
We subcloned a human dominant-negative RARα cassette (hRARΔ403; provided by Charles Sagerström; Damm et al., 1993; Roy and Sagerström, 2004) behind the zebrafish hsp70 promoter (Halloran et al., 2000) and generated a stable line of transgenic animals. For further details, see Supplemental Experimental Procedures.
We used the antibody H6409 (Sigma-Aldrich) to detect phospho-histone H3 and the ApopTag kit (Millipore) for TUNEL staining. Stained cells were detected using an alkaline phosphatase-conjugated secondary antibody and NBT/BCIP.
Embryos were examined using Zeiss M2Bio and Axioplan microscopes. Photographs were taken using a Zeiss Axiocam. Zeiss AxioVision 3.0.6 software and Adobe Photoshop were used to process images.
We thank A. Schier, A. Lekven, C. Moens, J. Nance, C. Loomis, and members of the Yelon lab for critical input, B. Conrad for assistance with statistical analysis, C. Zusi for providing BMS189453, and K. Niederreither and S. Zaffran for sharing data prior to publication. This work was supported by grants from the National Institutes of Health to DY and KDP and from the American Heart Association to DY. JSW received support from NIH F32 HL083591 and K99 HL091126. BRK received support from the NYU Graduate Training Program in Developmental Genetics (NIH T32 HD007520) and the Medical Scientist Training Program of NYU School of Medicine. The authors have no financial conflicts of interest to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.