|Home | About | Journals | Submit | Contact Us | Français|
The development of a complex organ requires the specification of appropriate numbers of each of its constituent cell types, as well as their proper differentiation and correct positioning relative to each other. During Drosophila cardiogenesis, all three of these processes are controlled by jumeau (jumu) and Checkpoint suppressor homologue (CHES-1-like), two genes encoding forkhead transcription factors that we discovered utilizing an integrated genetic, genomic and computational strategy for identifying genes expressed in the developing Drosophila heart. Both jumu and CHES-1-like are required during asymmetric cell division for the derivation of two distinct cardiac cell types from their mutual precursor, and in symmetric cell divisions that produce yet a third type of heart cell. jumu and CHES-1-like control the division of cardiac progenitors by regulating the activity of Polo, a kinase involved in multiple steps of mitosis. This pathway demonstrates how transcription factors integrate diverse developmental processes during organogenesis.
The remarkable cellular diversity present within metazoan organs illustrates several important themes in developmental biology, including a requirement for the specification of appropriate numbers of distinct cell types, the proper differentiation of these cells and their correct positioning within the organ (Rosenthal and Harvey, 2010). Taken together, the existence of multiple organ-specific cell types implies that numerous biological processes must work in unison during development, and raises an intriguing question: how is the requisite integration of these diverse developmental pathways achieved?
The formation of the Drosophila embryonic heart provides a particularly amenable system for addressing this question (Bodmer and Frasch, 2010; Bryantsev and Cripps, 2009). An organ that pumps hemolymph throughout the body cavity, the Drosophila heart is composed of two groups of cells arranged in a metamerically repeated and stereotyped pattern (Figures 1A-1C): an inner group of Myocyte enhancer factor 2 (Mef2)-expressing contractile cardial cells (CCs) that form a linear tube, surrounded by a sheath of pericardin (prc) and Zn finger homeodomain 1 (zfh1)-expressing nephrocytic pericardial cells (PCs). Neither the CCs nor the PCs constitute a uniform population, as revealed both by their distinct cell lineages and by the complexity of their individual gene expression programs. From anterior to posterior, and named for the transcription factors they express, there are two Seven-up-CCs (Svp-CCs), two Tinman-Ladybird-CCs (Tin-Lb-CCs), and two CCs expressing only Tin (the posterior-most Tin-CCs) in each hemisegment. A larger number of PCs surround the cardial cells: 2 Svp-PCs and 2 Odd-skipped-PCs (Odd-PCs) are positioned laterally, 2 Even-skipped-PCs (Eve-PCs) are situated dorsolaterally, and a row of Tin-PCs and Tin-Lb-PCs runs immediately ventral to the CCs (Azpiazu and Frasch, 1993; Bodmer, 1993; Jagla et al., 1997; Ward and Skeath, 2000).
A stereotyped series of asymmetric and symmetric cardiac progenitor cell divisions gives rise to these eight differentiated cell types (Alvarez et al., 2003; Han and Bodmer, 2003). The differential expression of multiple genes, and both the distinct lineage and intricate but invariant positioning of the individual heart cell types, argue for a high degree of functional precision and regulatory complexity in the generation of the heart. This hypothesis is borne out by classical genetic studies, which showed that the development of the Drosophila heart from the dorsal-most region of the mesoderm, a tin-expressing domain referred to as the cardiac mesoderm (CM), is dependent on contributions from multiple signals and transcription factors that are conserved between flies and vertebrates (summarized in Figure 1D; reviewed by Bodmer and Frasch, 2010; Bryantsev and Cripps, 2009; Chien et al., 2008). Thus, the identification of genes that regulate cardiac development, and detailed investigations of their expression and function in Drosophila, are likely to provide considerable insight into the related mechanisms controlling cardiogenesis in vertebrates, including human.
Here, we describe an integrated strategy that we developed and applied to identify 70 genes expressed in the Drosophila CM or heart. We further show that one gene discovered with this approach, jumeau (jumu), plus its homolog, Checkpoint suppressor homologue (CHES-1-like)—both of which encode Fkh transcription factors—mediate both asymmetric and symmetric cardiac progenitor cell divisions by regulating a Polo-kinase dependent pathway.
As an initial step to identify regulators and effectors of heart development in Drosophila, we screened for genes expressed in the CM or differentiated heart using an integratedgenetic, genomic and computational strategy that we previously applied to study somatic muscle gene expression (Estrada et al., 2006). Two essential aspects of our approach are the use of specific genetic backgrounds to selectively perturb CM gene expression based on prior knowledge of cardiogenic pathways (Figures 1D-1E), and the availability of a training set of 40 genes already known to be expressed in the CM. Previous studies revealed that activation of the fibroblast growth factor receptor (FGFR)- and epidermal growth factor receptor (EGFR)-driven receptor tyrosine kinase (RTK)/Ras, Wingless (Wg) or Decapentaplegic (Dpp) pathways produces extra CM cells compared with wild-type, thereby elevating levels of CM gene expression (Azpiazu and Frasch, 1993; Bodmer, 1993; Carmena et al., 1998; Frasch, 1995; Gisselbrecht et al., 1996; Grigorian et al., 2011; Michelson et al., 1998; Staehling-Hampton et al., 1994). In contrast, activation of Notch results in fewer CM cells and thus reduced levels of CM gene expression (Hartenstein et al., 1992; Mandal et al., 2004). Moreover, since the CM arises from the tin-expressing dorsal mesoderm, CM genes will be highly enriched in cells purified from this subregion of the whole embryo.
To generate a compendium of gene expression profiles associated with CM development, we used flow cytometry to purify the entire mesoderm from both wild-type stage 11 embryos and similar embryos from 9 informative genetic backgrounds, as well as tin-expressing dorsal mesoderm from equivalently staged wild-type embryos (Figure 1E). We then used a statistical meta-analysis method (Estrada et al., 2006) fitted to the training set of 40 known CM genes to rank all Drosophila genes by their likelihood of being expressed in the CM based on their collective behavior in this expression profiling compendium. Any gene that (i) is upregulated with activation of the RTK/Ras, Wg or Dpp pathways, upregulated with Dl loss-of-function, downregulated with Notch activation, and downregulated with wg loss-of-function, and (ii) is enriched in tin-expressing mesoderm relative to the entire mesoderm, has a high probability of being expressed in the CM (Table S1A, Experimental Procedures).
To validate the predictions of this meta-analysis, we used large-scale whole-embryo in situ hybridizations to assess the in vivo expression patterns of highly ranked CM gene candidates. Of 136 randomly selected genes that provided informative in situ hybridization results among the top-ranked 400 candidates, and which did not include any of the training set genes, 70 were expressed in the CM and/or heart. Thus, the meta-analysis predicted cardiac genes with an accuracy of 51.4% (Table S1B). Further analyses revealed that 37 genes are expressed in both the CM and the mature heart, 22 genes are expressed in the CM but not in the heart, and 11 genes are expressed in the heart but not in the CM (Table S1B and Figure S1).
To gain insight into the biological processes in which these genes are involved, the 110 training set plus newly validated CM and heart genes were queried for the relative enrichment of Gene Ontology (GO) terms. Overrepresented terms (Table S1C) include categories associated with mesoderm development, cardiac differentiation, cell fate specification, transcriptional regulation, migration, tube morphogenesis and the RTK/Ras pathway. Another enriched category was nervous system development, which likely reflects the pleiotropic effects of many developmental regulators, and the fact that many of the identified genes are also expressed in the nervous system (data not shown). Among the unexpected overrepresented categories were cytokinesis and cell division, the relevance of which became apparent upon a detailed analysis of the cardiogenic functions of jumu, CHES-1-like and polo.
Previous studies have shown a striking conservation of transcription factors involved in both Drosophila and vertebrate cardiogenesis. Genes encoding transcription factors were also overrepresented among the 110 CM- and heart-expressed genes. One such gene is jumu, which encodes a Fkh subclass N transcription factor (Lee and Frasch, 2004) that is continuously expressed in the CM and differentiating heart from embryonic stages 11 to 13 (Figures 2A and S1Y-S1Y’’’). In addition, we examined the expression pattern of the only other Drosophila Fkh subclass N gene, CHES-1-like, and found that it is also expressed in the CM during stages 11 and 12 (Figures 2B and S1Z-S1Z’’’).
Given the presence of these two Fkh transcription factors in the embryonic CM, and the fact that this class of proteins is involved in mammalian cardiogenesis, we next used a whole embryo RNA interference (RNAi) assay to assess whether jumu and CHES-1-like play a role in Drosophila cardiac development. RNAi directed against either jumu or CHES-1-like resulted in incorrect numbers and an uneven distribution of both CCs and PCs (Figures 2C-2E), indicating that both of these Fkh factors are essential for normal heart development.
We undertook a more detailed analysis of the cardiogenic effects of jumu and CHES-1-like by examining the phenotypes associated with loss-of-function mutations in these genes. Staining with antibodies against the nuclear protein Mef2 (which is expressed in CCs of the heart, as well as in somatic myoblasts) revealed that the uniform and symmetrically aligned distribution of CCs seen in wild-type embryos (Figure 2F) is markedly disrupted in embryos homozygous for jumu hypomorphic mutations (jumu06439 and jumuDf2.12, Figure 2G-2H), a jumu null deficiency (Df(3R)Exel6157, Figure 2I) which deletes both jumu and another gene not involved in heart development (Cheah et al., 2000; Strodicke et al., 2000; see also Table S2), and a null mutation that we generated in CHES-1-like (Df(1)CHES-1-like1, Figures 2J and S2; see Experimental Procedures).
Each mutant exhibited different hemisegments having localized increases or decreases in CC number, occasional enlarged CC nuclei, or CCs which were misaligned with other CCs within a hemisegment or with their counterparts across the dorsal midline. Similar phenotypes were also observed when either jumu or CHES-1-like activity was knocked down by CM-targeted RNAi directed by the Hand-GAL4 and tinD-GAL4 drivers (Figure 2K-2L), indicating that the requirement of these Fkh genes for correct heart development is autonomous to the cardiac mesoderm. Embryos doubly homozygous for both the jumu null deficiency and the CHES-1-like null mutation exhibited a more severe phenotype, often missing entire hemisegments of CCs (Figure 2M). Taken together, these results suggest a role for abnormal cell division as the origin of the jumu and CHES-1-like mutant heart phenotypes, which is consistent with the known involvement of jumu in nervous system development (Cheah et al., 2000).
Two asymmetric progenitor cell divisions generate all the Svp-expressing heart cells, with each division producing one Svp-CC and one Svp-PC per hemisegment (yellow and red cells respectively in Figure 3A; Gajewski et al., 2000; Ward and Skeath, 2000). In contrast, a pair of symmetric cell divisions gives rise to the four Tin-CCs in each hemisegment, the two Tin-Lb-CCs and the two posterior-most Tin-CCs (green cells in Figure 3A; Han and Bodmer, 2003). These lineage relationships are shown in Figure 3G.
We took advantage of this ability to distinguish the products of asymmetric and symmetric cardiac progenitor cell divisions to determine whether cell division defects are responsible for the heart phenotypes seen in jumu and CHES-1-like mutants. Indeed, one source of the localized increase in CC number in embryos lacking jumu or CHES-1-like function is an abnormal asymmetric cell division that causes a Svp progenitor cell to yield two Svp-CCs instead of one Svp-CC and one Svp-PC (phenotype I in Figures 3B, 3F and 3G). Conversely, in some cases a Svp progenitor produces two Svp-PCs instead of a Svp-CC and a Svp-PC, resulting in a localized reduction in CC number in jumu mutants (phenotype II in Figures 3B and 3G).
Occasional karyokinesis defects also occurred during the asymmetric division of Svp progenitor cells in both jumu and CHES-1-like mutants (phenotype III in Figures 3C, 3F and 3G). This finding is more clearly illustrated in a three-dimensional reconstruction of microscopic images corresponding to the two highlighted opposing hemisegments in Figure 3C (also see Movie S1). Note that the posterior-most Svp-CC nuclei in each hemisgment are arrested in the process of dividing, with each appearing to possess two nuclei that are unable to completely dissociate. The karyokinesis defects did not change the number of Svp-CCs, but there was a reduction in the number of associated Svp-PCs. In addition, depending on when the karyokinesis arrest occurred, some of the Svp-CC nuclei appeared larger than normal. Mutations in jumu and CHES-1-like also caused karyokinesis defects in the symmetrically dividing Tin-CCs, which resulted in a localized reduction in the number of these cells (phenotype IV in Figures 3C, 3D, 3F and 3G).
We also observed localized increases in the number of Tin-CCs in jumu and CHES-1-like mutant embryos (phenotype V in Figures 3D, 3F and 3G). Additional cell division is the likely source of these extra Tin-CCs since in mutant embryos some hemisegments had wild-type numbers of Svp-CCs and Tin-CCs but one or more Tin-CCs were arrested in the process of undergoing extra cell division (Figure 3E).
Finally, a small fraction of hemisegments in both jumu null and CHES-1-like null mutant hearts exhibit two phenotypes which cannot be explained by any of the previously considered mechanisms: (i) hemisegments containing only one Svp-CC and one Svp-PC (Figure S3A-S3B), and (ii) hemisegments with a total of six Svp-expressing cells (Figure S3C-S3D). Defects in the earlier round of cell divisions that give rise to the Svp progenitors can explain both of these phenotypes. In the first case, this mechanism would produce only one Svp progenitor cell in a hemisegment—which, in turn, could give rise to only two Svp heart cells—and in the second case, it would generate three Svp progenitor cells which subsequently divide to yield six Svp cardiac cells. A quantitative summary of the jumu and CHES-1-like mutant phenotypes, the statistical significance of each class, and the mechanisms by which they arise are found in Tables S2A-S2B.
Figures 3D and 3F illustrate one possible reason for incorrectly positioned CCs in jumu and CHES-1-like mutants. When one hemisegment contains as many as eight CCs, and its counterpart across the dorsal midline has as few as five such cells, keeping the hemisegments aligned requires one of the rows of CCs to bulge out (Figure 2G). Alternatively, some of the excess CCs may be displaced from their normal linear arrangement (Figures 3D and 3F). This latter model is supported by the observation that, in jumu and CHES-1-like mutants, segments in which opposing hemisegments have unequal numbers of CCs exhibit significantly more incorrectly positioned cells than do segments with hemisegments containing the same number of CCs (Tables S2C-S2D).
In summary, all of the heart phenotypes observed in jumu and CHES-1-like mutants can be accounted for by defects in different aspects of the asymmetric or symmetric division of cardiac progenitor cells.
Membrane-associated Numb protein localizes on one side of asymmetrically dividing neural precursor cells and segregates to only one of the two daughter cells where it antagonizes the activity of Notch, leading to differences in progeny cell fates (Rhyu et al., 1994; Spana and Doe, 1996). Although Numb expression has not previously been examined in Svp cardiac progenitor cells, the identification of supernumerary Svp-PCs in numb mutants was used in a prior study to infer that numb plays a similar role in the Svp progenitors, with the daughter cell which inherits most of Numb protein assumed to adopt a Svp-CC fate (Ward and Skeath, 2000). We pursued this hypothesis in more detail by both genetic interaction and Numb protein localization experiments.
If the Svp progenitor cell division defects in jumu and CHES-1-like mutants is a consequence of the wild-type functions of these genes being mediated via Numb localization during asymmetric cell division, then strong pairwise genetic interactions should occur between numb and each of jumu and CHES-1-like alleles. To examine this possibility, the heart phenotypes of single mutant heterozygotes of these three genes were quantitated and compared with those of embryos that are doubly heterozygous either for mutations in both jumu and numb, or for mutations in both CHES-1-like and numb (Figures 4A-4B and Tables S2A-S2B). Double heterozygotes for both jumu and the numb null mutations exhibit asymmetric cell division defects in Svp-expressing cells that are significantly more severe (p = 0.0018) than the additive effects of each of the two single heterozygotes. In contrast, defects in the symmetric cell divisions that yield the Tin-CCs in the double heterozygotes are not significantly different (p = 0.7198) from the additive effects of the single jumu and numb heterozygotes. A similar synergistic genetic interaction between CHES-1-like and numb occurs for asymmetric (p = 0.0124) but not for symmetric (p = 0.5863) cardiac cell divisions. Together, these results are consistent with jumu and CHES-1-like acting through numb to regulate the asymmetric cell division of Svp cardiac progenitor cells.
To directly test whether Numb mislocalization is associated with jumu and CHES-1-like mutant cardiac cell fate phenotypes, we first stained wild-type embryos carrying the svp-lacZ enhancer trap for expression of both Numb and β-galactosidase. Numb protein is asymmetrically localized in a crescent at one pole of normal Svp progenitor cells (Figure 4C). Thus, only one of the two daughter cells should inherit most of this protein and adopt a Svp-CC fate.
In contrast, in embryos homozygous for single or double null mutations of jumu and CHES-1-like, Numb protein is found in a more diffuse halo surrounding most of the nuclei in all dividing Svp progenitor cells (Figure 4D-4F). This finding implies that, after cell division, both progeny cells inherit roughly equal amounts of Numb protein, resulting in an inability to distinguish one cell from the other and with both taking on the same fate. Of note, similar Numb localization defects are also detected in some dividing Svp progenitor cells from embryos doubly heterozygous for mutations in the Fkh genes and numb, but not in numb heterozygotes (Figure 4G-4J).
Proper asymmetric localization of Numb protein in the Svp progenitor cells during asymmetric cell division requires its physical interaction and colocalization with phosphorylated Partner of Numb (Pon) protein (Lu et al., 1998; Wang et al., 2007). Intriguingly, synergistic genetic interactions are also observed between jumu and pon, and between CHES-1-like and pon, during asymmetric, but not during symmetric, cell divisions of the Svp progenitor cells (Figures 4K-4L and Tables S2A-S2B). These findings suggest that the utilization of numb by jumu and CHES-1-like during asymmetric cell division also involves pon function.
The requirement of both jumu and CHES-1-like for the proper localization of Numb during the asymmetric cell division of cardiac progenitors led us to consider that other regulators of mitosis might be involved in the effects of these Fkh transcription factors. One plausible candidate is polo, which encodes a kinase that phosporylates Pon, the protein that serves as an adapter for Numb during its asymmetric cellular localization (Wang et al., 2007), and which, as noted previously, exhibits synergistic genetic interactions with both jumu and CHES-1-like during asymmetric division of Svp cardiac progenitors. Intriguingly, Polo kinase not only regulates asymmetric cell division but also has been implicated in multiple steps of mitosis, meiosis and cytokinesis (Archambault and Glover, 2009), observations that correlate with the other jumu and CHES-1-like mutant cardiac phenotypes. Furthermore, a polo ortholog plays a role in cardiac myocyte proliferation during zebrafish heart regeneration (Jopling et al., 2010). Of additional significance, polo ranked very highly in our statistical meta-analysis for identifying genes expressed in the CM (rank position 66; Table S1A). Moreover, in situ hybridization revealed that the polo transcript is indeed transiently detected in the CM during embryonic stages 11 to 12 when cardiac progenitor cells divide (Figure 5A).
To test the hypothesis that jumu and CHES-1-like function in heart development by a polo-mediated pathway, we initially examined the cardiac expression of Mef2 in polo mutants. Embryos homozygous for either of two strong polo hypomorphic mutations, polo9 and polo10 (Donaldson et al., 2001), exhibit localized increases or decreases in Mef2-positive CC number, larger than normal CC nuclei and incorrectly positioned CCs (Figures 5B-5D), all of which phenocopy jumu and CHES-1-like mutants. Furthermore, the same five classes of cell division defects as previously described for jumu and CHES-1-like mutants occur with polo loss-of-function (Figures 5E-5I and Table S2A). These observations suggest that jumu and CHES-1-like act through a polo-mediated pathway to regulate the division and fates of cardiac progenitor cells, a possibility that we examined with the following series of additional experiments.
If jumu, CHES-1-like and polo function together during cardiogenesis, they might exhibit strong genetic interactions. To assess this possibility, the heart phenotypes of single heterozygotes for mutations in each of these three genes were quantitated and compared with those of embryos that are doubly heterozygous for mutations in all pairwise combinations of these genes. Our results (Figures 6A-6B and S4A, Tables S2A-S2B) demonstrated that synergistic genetic interactions indeed occur between jumu and polo, between CHES-1-like and polo, and between jumu and CHES-1-like during both asymmetric and symmetric cell divisions, suggesting that all three genes regulate heart development by functioning together in the same genetic pathway. Consistent with this model, many of the dividing Svp progenitor cells in these double heterozygotes also exhibit defective Numb localization (Figures 4J and 6E-6F).
We next considered the possibility that mutations in jumu and CHES-1-like might alter the expression level or subcellular distribution of Polo kinase. Since Polo normally localizes at the centrosomes of mitotically dividing cells where it is required for proper spindle assembly (Archambault and Glover, 2009; Moutinho-Santos et al., 1999), whether Jumu and CHES-1-like affect Polo protein can be assessed by determining if there are centrosomes lacking Polo in dividing cardiac cells that are mutant for jumu and CHES-1-like. Thus, we simultaneously stained wild-type and appropriate mutant embryos containing a svp-lacZ enhancer trap with antibodies against β-galactosidase (to detect Svp progenitor cells), phospho-histone H3 (to detect dividing cells), Pericentrin-like protein (PLP, a centriole/centrosome marker), and Polo.
In all 15 of the dividing wild-type Svp progenitor cells examined, Polo protein co-localized with PLP (Figures 5J-5J’’’). In contrast, in 15 dividing Svp progenitor cells in each of jumu and CHES-1-like null mutant embryos, Polo was not detected at the centrosome in 3 and 2 cells, respectively (Figures 5K-5L’’’). The incomplete penetrance of this effect is consistent with the observation that not all cardiac progenitors exhibit cell division defects in these mutants (Figure 3 and Table S2A). The Polo staining results could reflect either a defect in Polo protein localization to centrosomes or an overall reduction in Polo protein level. In either case, our data suggest that Jumu and CHES-1-like act upstream of the mitotic regulatory kinase encoded by polo.
To test the hypothesis that the Fkh genes act upstream of polo, polo was expressed under the control of a ubiquitin promoter in wild-type embryos and in embryos homozygous for either jumu or CHES-1-like null mutations. If polo acts downstream of the Fkh transcription factors, then ubiquitous expression of polo should at least partially rescue the cardiac phenotypes associated with the Fkh gene mutants. While in a control experiment ubiquitous expression of polo does not affect heart development in wild-type embryos, ectopic polo significantly reduces the severity of the cardiac defects associated with either single Fkh gene mutant alone (Figures 6C-6D and S4B-S4F, Tables S2A-S2B). The effect of ectopic polo is so efficient that the hearts in 4 out of 17 and 6 out of 16 of the rescued jumu and CHES-1-like mutant embryos, respectively, appeared completely wild-type (Figures S4D and S4F). Partial rescue of the Fkh mutant phenotypes was also obtained by driving polo expression specifically in the cardiac mesoderm by a tinD-GAL4 driver (Figures 6C-6D and Tables S2A-S2B), thereby indicating that the requirement of polo for correct asymmetric and symmetric cardiac progenitor cell divisions is autonomous to the heart.
The ubiquitous expression of polo in either jumu or CHES-1-like mutants also restored the proper asymmetric localization of Numb protein in some, but not all of the dividing Svp progenitor cells (Figures 6G-6H). Whereas Polo protein was not detected in the centrosomes of a small fraction of the dividing Svp progenitors in either jumu or CHES-1-like mutants (Figures 5K-5L’’), Polo protein was localized in the centrosomes of all 15 of the dividing Svp progenitors that were examined for each Fkh mutant in which polo was constitutively expressed under the control of the ubiquitin promoter (Figure S4I-S4J’’’).
Of note, the more severe phenotype in which cardiac cells were entirely missing from some hemisegments in a significant fraction (8/50; p = 0.00054) of embryos doubly homozygous for both jumu and CHES-1-like null mutations could not be rescued by ubiquitous expression of polo (11 out of 50 double mutants with the Ubi-GFP-polo transgene exhibited similar defects; Figures S4G-S4H). This result suggests that jumu and CHES-1-like regulate heart development by additional, polo-independent mechanisms.
Collectively, the failure to detect Polo in the centrosomes of dividing heart cells in the Fkh gene mutants, and the partial rescue of cardiac phenotypes in both jumu and CHES-1-like mutants by either ubiquitously expressed or CM-targeted polo, argue strongly that both Fkh genes act upstream of polo in a regulatory pathway governing cardiac progenitor cell divisions.
A possible explanation for the partial rescue of the cardiac phenotypes of the Fkh gene mutants is that Jumu and CHES-1-like directly or indirectly control the transcription of polo. We tested this hypothesis by overexpressing Jumu throughout the mesoderm with a twi-Gal4 driver, and by using Affymetrix microarrays to quantitate the effects of this genetic perturbation on polo expression levels in stage 11-12 mesodermal cells that were purified by flow cytometry to enrich for mesoderm-specific responses. This strategy offers the additional advantage of simultaneously measuring the transcriptional responses of other mesodermal genes to ectopic jumu on a genome-wide scale. Compared to wild-type, jumu overexpression is associated with a significantly elevated level of polo expression (>1.6-fold enrichment; p < 0.05; Table S3A), suggesting that jumu indeed activates polo transcription.
Furthermore, this genome-wide expression profiling experiment identified a total of 374 genes whose expression levels are significantly elevated when jumu is overexpressed in the entire mesoderm (Table S3A). Of note, 24 of the 110 cardiac genes previously described in this study (Table S1) are included among the 374 jumu-upregulated genes (Table S3A). This number represents a statistically significant overrepresentation of cardiac genes among those upregulated by jumu (p < 10-14 by the hypergeometric distribution).
To gain insight into the biological processes in which the jumu-upregulated mesodermal genes are involved, this gene set was queried for the relative enrichment of Gene Ontology (GO) terms (Table S3B). The overrepresented GO terms include multiple categories associated with asymmetric and symmetric cell division, cell cycle and cytokinesis, suggesting that the regulation of these processes by Jumu also involves genes independent of but having functions related to that of polo. Prominent examples of such potentially synergistically acting Jumu-responsive genes include abnormal spindle, Inner centromere protein, pavarotti and borealin-related.
The results described above illustrate how the two transcription factors encoded by jumu and CHES-1-like act through polo to ensure that the differentiated heart acquires the requisite types, numbers and arrangement of cardiac cells. Other transcription factors known to play critical early roles in cardiogenesis include the NK homeodomain Tin, the three T-box factors encoded by the Dorsocross genes, Doc1, Doc2 and Doc3, and the GATA factor encoded by pannier (pnr) (Alvarez et al., 2003; Azpiazu and Frasch, 1993; Bodmer, 1993; Reim and Frasch, 2005). Thus, we undertook genetic interaction experiments to determine whether these previously characterized cardiogenic transcription factors also participate in the developmental pathways governed by the Fkh proteins Jumu and CHES-1-like in the heart.
Our results (Tables S2A-S2B) demonstrate synergistic genetic interactions between the Fkh genes and tin during both asymmetric and symmetric cell divisions (Figures 7A-7B), synergistic interactions between the Doc genes and the Fkh genes only during symmetric cell divisions (Figures 7C-7D), and no genetic interactions between either Fkh gene and pnr (Figures 7E-7F). Collectively, these data suggest that tin and the Fkh genes act together during both asymmetric and symmetric cell divisions, that the Doc genes and the Fkh genes are closely associated only during symmetric cell divisions, and that pnr regulates heart development by mechanisms not involving either of the Fkh genes.
In this study, we used an integrated strategy to discover genes expressed in a complex organ and its progenitor cells by combining informative genetic perturbations of development, a statistical analysis of the genome-wide gene expression profiles of purified primary cells of interest, and the large-scale validation of predicted gene expression patterns by whole-embryo in situ hybridization. This approach offers two significant advantages over other studies undertaken to identify genes involved in a particular developmental process. First, isolating the cells of interest eliminates the potentially confounding effects of genetic perturbations in the rest of the embryo and increases the sensitivity of genome-wide expression profiling. Second, by examining perturbations of not one, but multiple convergent developmental pathways, any bias associated with the manipulation of a single genetic pathway is reduced and the relative contribution of each pathway to gene expression is taken into consideration in the statistical meta-analysis, thereby increasing the accuracy of the gene predictions.
The 70 genes found to be expressed in the Drosophila CM and differentiated heart by this strategy provide a substantial set of candidates that can be examined for possible roles in cardiac development, as well as in mechanistic studies of gene regulation. Here, we focused on the cardiogenic functions of jumu and CHES-1-like, both of which encode Fkh transcription factors and function upstream of polo to control multiple processes in the developing heart, including both the symmetric and asymmetric division of cardiac progenitors, karyokinesis, cell fate specification and the proper positioning of CCs within the mature heart. The serine-threonine kinase encoded by Drosophila polo—and its orthologs in mammals, yeast, frogs and nematodes—are known to play essential roles in a number of conserved biological processes involved in both mitotic and meiotic cell divisions (Archambault and Glover, 2009). These functions include: (1) the initial entry into M phase, a defect in which could yield the karyokinesis phenotype seen in our study; (2) centrosome maturation and spindle formation, defects in which could result in problems with spindle orientation and assembly that might also account for the karyokinesis phenotype, as well as the observed increase in Tin-CCs and abnormalities in cell positioning; (3) determination of cell fates during asymmetric cell division by the Polo-mediated phosphorylation of Pon, errors in which could generate improper sibling identities in the progeny of the Svp progenitor cell; and (4) exit out of mitosis and the promotion of cytokinesis, flaws in which could additionally explain the karyokinesis abnormality. Of note, spindle assembly problems are also seen with loss-of-function of jumu and CHES-1-like in Drosophila S2 cells (Goshima et al., 2007). Thus, failure to correctly regulate the activity of the downstream gene polo could explain the entirety of the cardiac phenotypes that occur in jumu and CHES-1-like single mutant embryos. In contrast, the severity of the double compared with the single mutant jumu and CHES-1-like cardiac phenotypes, and the results of the polo rescue experiments, suggest that partially redundant polo-independent pathways must also regulate the cardiogenic functions of these two Fkh factors.
Given current knowledge about Polo function, there are at least three mechanisms by which polo activity could be controlled by Jumu and CHES-1-like. First, since Polo kinase is known to be activated by phosphorylation of its T-loop (Jang et al., 2002; Qian et al., 1999), the Fkh factors could affect this regulatory step. We do not favor this mechanism since mutations in either Fkh gene affect localization of Polo protein at centrosomes and/or the level of Polo protein, and either ubiquitous or CM-targeted polo expression rescues jumu and CHES-1-like mutants. Second, Jumu and CHES-1-like could influence the well-established Ubiquitin-dependent proteolysis of Polo (Lindon and Pines, 2004). In this context, it is worth noting that absence of COP9 complex homolog subunit 4, a member of a protein complex regulating Ubiquitin-mediated protein degradation, exhibits similar cardiac phenotypes as those observed for loss of jumu, CHES-1-like and polo functions (Tao et al., 2007). Further experiments will be required to assess whether the Drosophila Fkh transcription factors regulate such a pathway.
A third alternative is that Jumu and CHES-1-like directly or indirectly control the transcription of polo. The observation that Fkh transcription factors are required for the expression of polo orthologs in other species (Buck et al., 2004; Laoukili et al., 2005; Zhu et al., 2000) is consistent with the possibility that jumu and CHES-1-like could have a similar regulatory effect in Drosophila. We found additional evidence supporting this hypothesis through our determination that polo expression levels are appreciably increased when jumu is ectopically over-expressed throughout the entire mesoderm, and that other genes having functions related to polo are similary upregulated by ectopic Jumu. Furthermore, chromatin immunoprecipitation data from the modENCODE project (Negre et al., 2011) indicate that Jumu protein binds to the polo genomic region during the stage of embryogenesis when the CM develops, suggesting that this Fkh transcription factor might directly control polo transcription in cardiac progenitor cells.
The observation that FoxM1, a subclass M Fkh protein, transcriptionally regulates a polo ortholog in mammalian cell lines (Laoukili et al., 2005) is also of interest since FoxM1 homozygous knockout mice die in the perinatal period with dilated hearts (Korver et al., 1998). Moreover, histological analysis shows that the orientation of cardiomyocytes in FoxM1 mutant hearts is highly irregular, and the nuclei of these cardiomyocytes are enlarged, consistent with being polyploid (Korver et al., 1998). These phenotypes of FoxM1 mutant mice are remarkably similar to those of jumu and CHES-1-like mutants in Drosophila, suggesting that the cardiogenic roles of both Fkh genes and polo have been evolutionarily conserved.
In summary, our genetic analysis of the roles played by two mesodermally expressed Fkh transcription factors in mediating Polo-dependent cardiac progenitor cell divisions, our characterization of numerous jumu-responsive mesodermal genes, and the interactions we uncovered between the Fkh proteins Jumu and CHES-1-like and other classes of cardiogenic transcription factors emphasize the elaborate architecture of the regulatory network that is required for the precise orchestration of multiple developmental events during the formation of an organ comprising multiple differentiated cell types. Furthermore, the present findings illustrate how the coordination of diverse biological processes is achieved during development of the Drosophila embryonic heart through the localized control of a ubiquitous cell cycle regulator by the spatially and temporally restricted activities of two Fkh transcription factors.
Microarray-based gene expression profiles for GFP-positive and GFP-negative cells isolated by fluorescence-activated sorting of a single cell suspension prepared from homozygous tinD-GFP stage 11 embryos were obtained as described previously (Estrada et al., 2006). These data were combined with prior microarray results derived for the 9 genetic perturbations shown in Figure 1E (Estrada et al., 2006) in order to facilitate the cardiac meta-analysis undertaken in this study.
The statistical methods for combining multiple related microarray datasets to predict genes with expression similar to a training set of known co-expressed genes has previously been described in detail (Estrada et al., 2006). In brief, each gene is assigned a “combined significance statistic” T which is the weighted sum of the CyberT t-statistics for that gene from each condition-to-control comparison performed, as calculated using the Goldenspike R package (Choe et al., 2005) and multiplied by a “sign” term (1 or −1) reflecting the expected direction of change in expression of target genes in a given genetic condition. Weight profiles are systematically assessed for the ability to detect genes from the training set among the top-ranked genes at a variety of q-value cutoffs; those combinations of weights that were among the top 10% at all cutoffs examined were averaged to produce the final combination of weights used to generate the T scores on which genes were ultimately ranked (Figures 1F-1H).
Microarray-based gene expression profiles for GFP-positive cells isolated by fluorescence-activated sorting of single cell suspensions prepared from stage 11-12 twi-GAL4 UAS-2EGFP/UAS-jumu embryos and stage 11-12 twi-GAL4 UAS-2EGFP embryos were also obtained by Affymetrix microarray hybridization. These array data were analyzed using the affy (Gautier et al., 2004) and limma (Smyth, 2004) Bioconductor package. Raw intensities were normalized separately using the mas5 and rma commands with default settings; linear models were fitted separately to each normalized dataset in limma. Probesets were filtered for log2(fold change) > 0.5 and adjusted p-value < 0.1 after correction for multiple hypothesis testing, and only those genes represented by probesets meeting both criteria in both tests were considered further (Table S3).
CHES-1-like was deleted by FLP-catalyzed recombination (Parks et al., 2004) using the FRT-containing transposons e02377 and e04245 that flank the gene (Thibault et al., 2004). The absence of the gene in the resulting deficiency, Df(1)CHES-1-like1, was confirmed by PCR analysis (Figure S2). Homozygotes and hemizygotes for Df(1)CHES-1-like1 are viable and fertile, but every mutant embryo exhibits the heart phenotypes described in this study.
At the time that the targeted deletion strategy was designed and the crosses were undertaken, CHES-1-like was the only known gene that was predicted to be deleted. However, a more recent annotation of the Drosophila genome places another gene of unknown function, CG43287, in the deleted interval. While we cannot rule out a contribution of this second gene to the mutant phenotype associated with Df(1)CHES-1-like1, the similarity between the CHES-1-like RNAi and Df(1)CHES-1-like1 heart phenotypes strongly suggests that loss of CHES-1-like function is the primary cause of the cardiac defects.
Whole embryo RNAi assays were performed as previously described (Estrada et al., 2006). RNAi assays targeted to specific embryonic cell types were carried out by expressing the UAS-inverted repeat constructs GD4099 (jumu dsRNA) and KK101264 (CHES-1-like dsRNA) (Dietzl et al., 2007) in the cardiac mesoderm using both the tinD-GAL4 and Hand-GAL4 drivers simultaneously. Mef2-stained hearts of stage 16 embryos with the genotypes UAS-Dcr-2; Hand-GAL4/jumuGD4099; tinD-GAL4/+ and UAS-Dcr-2; Hand-GAL4/CHES-1-likeKK101264; tinD-GAL4/+ raised at 29°C were compared with those of siblings lacking the inverted repeat constructs as negative controls.
The FuncAssociate 2.0 web application (Berriz et al., 2009) was utilized to query for relative enrichment of Gene Ontology terms both in genes identified as being expressed in the CM or heart and in genes with elevated expression levels when jumu was ectopically expressed throughout the entire mesoderm.
We thank A. Hofmann, B. Paterson, B. Durand, J. B. Skeath, B. Lu, R. S. Hawley, Y. Xiaohang, I. Reim, Z. Han, J. Lipsick, the Bloomington Drosophila Stock Center, the Exelixis Collection at the Harvard Medical School, the Vienna Drosophila RNAi Center and the Developmental Studies Hybridoma Bank for fly lines and reagents; Y. Kim and X. Zhu for helpful discussions; G. C. Rogers for technical assistance with Polo antibody production; and C. Sonnenbrot, L. Phun and the NHLBI Flow Cytometry and Gene Expression Core Facilities for assistance with experiments. This work was supported by the NHLBI Division of Intramural Research (A.M.M. and N.M.R.) and by an American Heart Association Postdoctoral Fellowship (S.M.A.).
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.
Microarray data utilized in this study are available from the Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo) with the accession numbers GSE3854, GSE29573 and GSE34946.