A functional genomic screen for cardiogenic genes using RNA interference in developing Drosophila embryos

Abstract

Identifying genetic components is an essential step toward understanding complex developmental processes. The primitive heart of the fruit fly, the dorsal vessel, which is a hemolymph-pumping organ, has provided a unique model system to identify cardiogenic genes and to further our understanding of the molecular mechanisms of cardiogenesis. Using RNA interference in developing Drosophila embryos, we performed a genomewide search for cardiogenic genes. Through analyses of the >5,800 genes that cover ≈40% of all predicted Drosophila genes, we identified a variety of genes encoding transcription factors and cell signaling proteins required for different steps during heart development. Analysis of mutant heart phenotypes and identified genes suggests that the Drosophila heart tube is segmentally patterned, like axial patterning, but assembled with regional modules. One of the identified genes, simjang, was further characterized. In the simjang mutant embryo, we found that within each segment a subset of cardial cells is missing. Interestingly, the simjang gene encodes a protein that is a component of the chromatin remodeling complex recruited by methyl-CpG-DNA binding proteins, suggesting that epigenetic information is crucial for specifying cardiac precursors. Together, these studies not only identify key regulators but also reveal mechanisms underlying heart development.

Over the last decade significant progress has been made in our understanding of the molecular details of cardiac morphogenesis. These advances can be attributed to the fact that many embryonic events involved in cardiogenesis are remarkably conserved among distinct species (1–3). In the Drosophila embryo, bilaterally symmetric groups of cardiac precursor cells in the dorsal-most mesoderm, specified by positional information provided by intrinsic cell-autonomous factors and extrinsic signals, meet in the dorsal midline to form a linear heart tube (3). Unlike the vertebrate linear heart tube, which proceeds further to looping and multichamber formation, the Drosophila heart remains a contractile linear tube at maturity. Two major classes of cell observed within the Drosophila cardiogenic repertoire are cardial cells (cardioblasts), which form the lumen of the linear heart tube, and pericardial cells, which flank and are associated with cardial cells (4). Cardial cells eventually form the cardiac muscles and are the contractile cells of the heart, whereas pericardial cells do not express muscle proteins but may function in hemolymph filtration. The linear heart tube shows clear structural differences along its length after its formation in late embryogenesis. The posterior three segments constitute the heart, a contractile tube with a wide bore involved in the forward pumping of hemolymph. Three pairs of specialized inflow valves (ostia) are present in the heart to facilitate the lateral entry of hemolymph (5, 6). Anterior to the heart is a narrower section, termed the aorta, that encompasses four segments. Despite its apparently simple structure and function, the linear heart tube of Drosophila shares many similarities with the early-stage hearts of vertebrates (3, 7), providing a unique model system in which the cell dynamics of heart formation can be described with single-cell resolution.

Loss-of-function mutant phenotypes greatly help us understand how genes function during development and in specific signaling pathways. Unlike conventional mutagenesis, RNA interference (RNAi) (8), a phenomenon in which double-stranded RNAs (dsRNAs) silence gene expression through specific degradation of their cognate mRNAs, is a direct and an efficient way of producing and identifying the loss-of-function mutant phenotypes of predicted genes in a variety of organisms as a reverse genetic tool (9, 10). In this article, we describe an RNAi-based, genomewide loss-of-function screen for cardiogenic genes essential for the development of the Drosophila embryonic heart. From the screen, we have identified a variety of genes encoding cell signaling proteins and transcription factors required for different steps during heart development. Furthermore, we have found that epigenetic information is crucial for specifying cardiac precursors.

Materials and Methods

Preparation of dsRNAs. Master plates containing aliquots of individual dsRNAs for 5,849 genes were described previously (11). Briefly, individual template DNAs for in vitro transcription were prepared by PCR with general primers containing the T7 promoter sequence at the end. cRNAs for each strand then were transcribed in vitro. RNAs were purified, and equal amounts of cRNAs were annealed to generate dsRNAs (2–3 μg/μl), which were dissolved in injection buffer. For short dsRNAs, oligoribonucleotides (21-mer) were purchased from Dharmacon (Lafayette, CO) or synthesized by the Applied Biosystems oligonucleotide synthesizer, deprotected, and purified by using a PolyPak cartridge (Glen Research, Sterling, VA) as described in the manufacturers' protocols. dsRNA solutions (20 μM) were prepared by annealing complementary oligoribonucleotides. Every oligoribonucleotide contains two deoxythymidine residues at the 3′ end as described (12).

Microinjection and Analysis of Embryos. The yw strain containing the D-mef2-lacZ transgene (13) was used as the source of embryos for microinjections. Embryos were collected over a 30-min period at 25°C, dechorionated by rolling onto double-stick tape with a forceps, and attached to a slide coated with double-stick tape. Embryos then were desiccated and covered with 700 halocarbon oil (Sigma). A Femtojet microinjector (Eppendorf) was used for microinjection of dsRNA solutions into the embryos at or before the syncytial blastoderm stage. The average injection volume was 65 pl. After injection of dsRNAs, embryos were incubated at 18°C under oil overnight for further development. Injected embryos were recovered and fixed for 5-bromo-4-chloro-3-indolyl β-d-galactoside (X-Gal) staining by using a standard protocol (14).

Antibodies and Immunohistochemistry. For further immunostaining, embryos injected with dsRNAs were first devitellinized by hand-peeling and subjected to immunohistochemistry. (Devitellinization by hand-peeling results in the best recovery of injected embryos for immunostaining.) For immunohistochemical analysis of embryos either from flies overexpressing transgenes or from mutant flies, standard protocols were used (15). The following primary antibodies and concentrations were used: rabbit anti-Mef2, 1:2,000 (gift of B. Patterson, National Institutes of Health); rabbit anti-Tin, 1:900 (gift from M. Frasch, Mount Sinai School of Medicine, New York); mouse anti-Pericardin, 1:2 (EC11; gift from D. Gratecos, Laboratoire de Génétique et Physiologie du Développement, Institut de Biologie du Développement de Marseille/Centre National de la Recherche Scientifique/Institut National de la Santé et de la Recherche Médicale/Université de la Mediterranée, Marseille, France); rabbit anti-Odd, 1:800 (gift of J. Skeath, Washington University School of Medicine, St. Louis); mouse anti-Lbe, 1:1 (gift of K. Jagla, Institut National de la Santé et de la Recherche Médicale, Clermont-Ferrand, France); and mouse anti-Eve and anti-Wg, 1:10 and 1:300, respectively (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City). Rabbit anti-SIMJ and guinea pig anti-MBD antibodies were generated by using His-tagged, full-length SIMJ and MBD protein as an immunogen, respectively. Fluorescent secondary antibodies were either Alexa Fluor 488 anti-rabbit or Alexa Fluor 594 anti-mouse (Molecular Probes), and nonfluorescent antibody detection was performed by using the VECTASTAIN Elite kit (Vector Laboratories). (See Fig. 5, which is published as supporting information on the PNAS web site.)

Whole-Mount in Situ Hybridization. Embryos were collected every 3 h for 18 h and fixed by using standard procedures. A master embryo mix combining equal amounts of embryos from all time periods was prepared for in situ hybridization analysis. Wholemount in situ hybridizations were performed with digoxigenin-labeled RNA probes as described previously (16). To complement in situ analysis of the identified cardiogenic genes, we also collected expression pattern data from a public database (www.fruitfly.org/cgi-bin/ex/insitu.pl).

Transgenic and Mutant Fly Lines. The Gal4 driver lines 24B-gal4, twi-gal4, and dpp-gal4 and the UAS lines UAS-abd-A, UAS-Abd-B, and UAS-Ubx were obtained from the Bloomington Stock Center (Bloomington, IN). Homeotic genes (abd-A, Abd-B, and Ubx) were overexpressed in the mesoderm by using either 24B-gal4 or twi-gal4 driver lines. Embryos were collected and analyzed by immunostaining with anti-Mef2, anti-Wg, and anti-Pericardin antibodies. P element insertion lines in the simj or the mbd-like gene locus were collected from the Bellen laboratory (Baylor College of Medicine, Houston), the Bloomington Stock Center, and Exelixis (South San Francisco, CA) and the l(3)01814 line was further characterized. The UAS transgenic lines harboring a GFP-tagged simj cDNA were established by using standard procedures as described previously (15). For construction of the UAS-simj vector, a simj cDNA (2.7-kb EcoRV/XbaI DNA fragment) was initially subcloned into SmaI/XbaI sites of pEGFP-C1, and the resulting GFP-tagged simj DNA fragments excised by double digestion with NheI and XbaI were subcloned into the XbaI site of the P element vector pUAST.

Results and Discussion

Identification of Cardiogenic Genes by RNAi in Developing Drosophila Embryos. For the screen, we have microinjected dsRNAs into a host Drosophila embryo that harbors a β-galactosidase (β-gal) marker gene expressed in the cardial cells and a subset of ventral muscle founder cells (Fig. 1 Left) (13). Because the β-gal marker gene is expressed in cardiac cells throughout heart development, simple X-Gal staining followed by monitoring of the β-gal expression pattern and heart morphology in injected embryos enabled us to successfully identify potential cardiogenic genes. As a dsRNA source, master plates were prepared (11). The master plates (96-well plates) contained a total of 5,849 individual long dsRNAs, representing all of the genes of the Drosophila Gene Collection (DGC Release 1.0) released by the Berkeley Drosophila Genome Project. For quality control of dsRNAs, both PCR products used for template DNAs for in vitro transcriptions and annealed individual dsRNAs were checked by agarose gel electrophoresis (11). In addition, we also synthesized short dsRNAs (21-mer) for a group of genes that are not included in DGC Release 1.0. The synthetic short dsRNAs can induce very potent and specific RNAi in developing Drosophila embryos as shown by immunostaining and Western blot analysis (see Fig. 6, which is published as supporting information on the PNAS web site). Initially, we injected embryos (40–60 embryos per pool) with a pool of dsRNAs of three different genes. From the X-Gal staining patterns in the injected embryos, a positive pool was identified, and then the individual dsRNA from a positive pool was injected again to determine which gene was silenced and thereby caused the phenotype (Fig. 1 Right). Because we often further characterized injected embryos by immunostaining with specific antibodies, several hundred embryos were injected with dsRNAs to identify one specific gene that causes a reproducible mutant heart phenotype. Also, to ensure that mutant heart phenotypes generated by RNAi of long dsRNAs indeed resulted from specific gene silencing, secondary experiments using short dsRNAs (21-mer) specific for identified individual genes were performed.

Fig. 1.

Host embryo and strategy of the screen for cardiogenic genes by RNAi in the developing Drosophila embryo. (Left) The WT host embryo for microinjection of dsRNAs stained with X-Gal, showing the expression pattern of the D-mef2-lacZ reporter gene and normal heart morphology during embryogenesis. Sn, embryonic stage n; CC, cardial precursor cells; SC, a subset of somatic muscle founder cells; DV, dorsal vessel (heart). (Right) A schematic of the screen for cardiac genes. Embryos were initially injected with dsRNAs from a pool of three different genes, incubated at 18°C overnight for development, and subjected to X-Gal staining. Stained embryos were examined under a microscope to select a positive pool that caused abnormal heart development. Each dsRNA from the positive pool was injected again into embryos for the identification of individual genes that affect heart development. N, negative; P, positive.

Using this strategy, we analyzed the >5,800 genes that cover ≈40% of all predicted Drosophila genes (17). Among them, a total of 132 genes showed heart mutant phenotypes in dsRNA-injected embryos, indicating that these genes are either directly or indirectly involved in heart development. The number is not necessarily an absolute one, specific for heart development, because many genes are also shown to be involved in the development of other systems. Genes identified in this screen are listed in Table 1, which is published as supporting information on the PNAS web site. Using bioinformatics in the FlyBase (http://flybase.bio.indiana.edu/), we found that among the cardiogenic genes identified, mutants for 81 genes are unknown. Approximately 80% of identified genes either showed mammalian homologues or contained a coding region for known functional domains. Mutants for 51 genes are known. Of those, genes such as heartless, pannier, and zfh1, which were shown to be involved in heart development, were recovered in our screen and verify its validity. However, we do not exclude the possibility that some genes are missing in our screen, especially when the effects of some genes are silent because of compensatory or redundant pathways or when genes affect heart development at different stages than those examined here. We also found that the role in heart development of most genes with known mutants had not been characterized. Our screen, using RNAi in developing Drosophila embryos, identifies a variety of cardiogenic genes and describes the functions in heart development of known genes.

Our screen is based primarily on morphological defects in the heart generated by targeted gene silencing by RNAi. It is, therefore, not easy to classify genes into specific categories. However, the individual mutant heart phenotypes facilitate understanding of the organogenesis of the linear heart tube, because we know from bioinformatics the nature of the individual genes that caused specific mutant heart phenotypes. Fig. 2 shows panels of discernible mutant heart phenotypes generated by loss-of-function of individual genes [marked with Computed gene (CG) number]. From our screen, we identified several known genes involved in the early steps of cardiogenesis. This category, which presumably functions in generation of the heart field, includes genes encoding signaling molecules such as the FGF receptor (CG7223), EGF receptor (CG10079), components of the Dpp and Wnt signaling pathways (CG7904 and CG11579), and the transcription factor Tinman. In fact, mutations of these genes caused the heartless phenotype (e.g., Fig. 2, CG10079). The CG1412 gene is classified in this category because RNAi for this gene also caused the heartless phenotype (Fig. 2, CG1412). The function of the CG1412 gene during cardiogenesis was previously unknown. This gene is maternally expressed, and its zygotic expression is transient during early stages of mesoderm development. The predicted molecular function of the gene product is RhoGAP. Thus, it is likely that, similar to the heartless (CG7223) mutant phenotype, the RhoGAP mutant heart phenotype may be caused by cell migration defects involving the FGF/FGFR signaling pathway. No additional genes encoding transcription factors, other than tinman in this category, were recovered in our screen.

Fig. 2.

Various heart phenotypes generated by RNAi. Examples of X-Gal-stained embryos at late stage 16, representing abnormal heart phenotypes caused by RNAi of individual dsRNAs, are shown. Computed gene (CG) number corresponds to individual genes that caused mutant heart phenotypes by RNAi. Dorsal views of embryos with anterior on the left are shown. WT, control embryo showing the normal heart phenotype.

Cardiac precursors within the heart field are further classified as one of two main cell types, cardial and pericardial. The heart phenotypes generated by defects in specification of a subset of cardiac cells usually showed a broken-heart phenotype, in which there are gaps in X-Gal staining in the heart structure (e.g., Fig. 2, CG3978, a pannier mutant phenotype). Identified genes in this category are diverse and include CG6268, CG9941, and CG14641 (Fig. 2). The CG6268 gene encodes a receptor tyrosine kinase signal mediator containing a KH domain and ankyrin repeat. This gene was shown to be involved in photoreceptor differentiation in the developing eye disk (18). The CG9941 gene encodes a protein containing a RING finger domain. The CG14641 gene, which is highly expressed in the trunk mesoderm primordium, encodes a protein containing an RNA binding domain. The predicted molecular functions of these genes suggest that a variety of regulatory molecules acting in different signaling pathways are involved in specification of a subset of cardial cells. These phenotypes, generated by RNAi for genes in this category, are worthy of further characterization, to identify which subset of cardiac cells is affected (see below).

Specified cardiac cells eventually move to the dorsal midline to assemble the linear heart tube. A number of genes affecting these steps were identified (Table 1). The assembly process is accompanied by dorsal closure. Thus, genes involved in dorsal closure obviously were recovered in our screen. However, we found clear distinctions in the assembly of the anterior and posterior structures of the heart. For example, in mutant embryos generated by RNAi for genes CG4744 and CG9425, the anterior portion (the aorta) of the heart tube is normally assembled, whereas the assembly of the posterior portion (the heart) is incomplete and delayed (Fig. 2). The CG9425 gene encodes a potential RNA helicase containing zinc finger domains. The human homologue of this gene has been identified (GenBank accession no. D29677), and the mouse gene CHAMP, encoding a potential RNA/DNA helicase, is specifically expressed in the developing mouse heart (19). The CG4744 gene encodes a protein with no identifiable functional domain except a glutamine-rich domain. In the case of the CG13739 gene, the aorta portion of the heart tube is missing (Fig. 2). These results suggest that, in conjunction with the general machinery of dorsal closure, separate regulatory networks may operate for the assembly of different parts of the heart tube during development.

We also identified genes that caused patterning defects in the heart tube. In this class of mutant heart phenotype, dorsal closure is normal, but specific segments of the heart tube are missing. This category of identified genes includes segmental patterning genes that play roles in axial patterning. As shown in Fig. 3, we found that RNAi for the pair-rule genes prd and ftz-f1 generates half-size hearts (Fig. 3 D, E, G, and H). Further analysis of early-stage embryos indicates that the phenotypes are generated by deletions of alternating segments (Fig. 3 F and I). Also, RNAi for the pair-rule gene opa (CG1133) caused a similar mutant heart phenotype. Thus, these results suggest that regulatory inputs from activities of ectodermal segmentation genes also govern the developmental program of the heart. Although it has been shown that several homeotic genes are expressed in the heart tube, homeotic gene function during heart development was not appreciated until recently (6, 20, 21). From our screen, in addition to segmentation genes, we also found that abd-A function is required for the development of the posterior portion (the heart) of the heart tube (Fig. 3 J and K). In the case of the Abd-B gene, we found that it affects development of the terminal portion (A8, abdominal segment 8) of the heart tube (data not shown). Consistent with this result, from our screen we recovered the lines gene (CG11770) (Table 1) that was shown to be required for specific functions of the Abd-B protein (22). Together, these results indicate that the Drosophila heart tube is segmentally patterned, like axial patterning. Of note, our results suggest that the Drosophila heart tube is assembled with different structural modules, as proposed in vertebrate heart organogenesis (23). In supporting this notion, several single-gene mutations that caused a deletion of the specific portion of the heart tube were recovered in our screen (Table 1).

Fig. 3.

Axial patterning genes play roles in Drosophila heart development. Embryos were injected with control dsRNA (A–C), the ftz-f1 dsRNA (D–F), and short dsRNA (21-mer) of prd (G–I) and abd-A (J and K) and subjected to X-Gal staining (A–J) followed by immunostaining with an anti-Mef2 antibody (B, E, and H). The half-size heart was generated by loss of either prd or ftz-f1 gene function (D, E, G, and H). The shape of the heart is maintained, but its size is reduced by half. At late stage 11 (C, F, and I), the embryo injected with the ftz-f1 (F) or the prd (I) dsRNA shows segmental deletion of both ventral muscle founder cells (arrowheads or stars) and cardiac precursor cells (arrow). A subset of muscle founder cells at the abdominal segments A1 and A3 is marked with an arrowhead, and stars indicate either muscle founder cells at segment A2 (C) or corresponding cells missing in the injected embryo (F and I). The embryo injected with the short dsRNA of abd-A shows transformation of the heart portion into the aorta, which is demonstrated by heart morphology (J) and cardial cell staining (K). Note that in the embryo injected with abd-A dsRNA (K), the posterior heart region (bracket) has the same narrower width as the aorta.

We also identified genes that affect cardiogenesis at late steps (e.g., Fig. 2, CG6721 and CG11914). In this category of mutant heart phenotypes, the overall pattern of the heart tube looks intact, but the heart tube is thicker (CG11914) or thinner (CG6721) than normal (Fig. 2). The predicted molecular function of the CG6721 gene product is RasGAP, whereas the CG11914 gene encodes a protein containing six Lim domains. A member of the cytoskeletal Lim-domain protein superfamily, Alp, is expressed during the growth of the embryonic ventricular chambers and was shown to interact with α-actinin (24). Furthermore, targeted gene ablation of Alp disrupted cardiomyocyte cytoarchitecture, leading to dilated cardiomyopathy (24). Because no change in the number of cardial cells (cardiomyocytes) in mutant hearts was observed, it is speculated that these genes may be involved in the signaling pathways required for proper development of cardiomyocyte cytoarchitecture. In fact, the CG11914 gene was not expressed until late stage 14 in the embryonic muscle system. It is not known, however, whether the CG11914 gene product can interact with cytoskeletal proteins. Included among this category are genes CG8815 (Sin3A) and CG8318 (Nf1), which encode a transcriptional corepressor and the neurofibromin 1 protein, respectively (Table 1).

simj Gene Function Is Required for the Specification of a Subset of Cardiac Cells. One advantage of our screen is that we can rapidly get information on the molecular function of the identified genes. Of particular interest are genes encoding histone-modifying enzymes (e.g., CG3025, CG8887, and CG9007) that may function in establishing and/or translating the histone code through chromatin remodeling (25). These genes are maternally expressed, but their zygotic expressions are ubiquitous. Silencing of these genes by RNAi disrupted normal heart development (Fig. 2). The CG3025 gene encodes a histone acetyltransferase, whereas both the CG8887 and CG9007 genes encode histone methyltransferases containing SET domains (26). Recently, it has been shown that the mammalian gene Bop, which contains a SET domain, is essential for cardiac differentiation and morphogenesis (27). Additionally, we found that a mutation of the CG7983 gene specifically perturbed specification of a subset of cardiac cells within each segment (Fig. 4). Initially, this gene was identified from the analysis of broken-heart phenotypes in dsRNA-injected embryos (Fig. 4A Inset). Further analysis of embryos with specific antibodies revealed that some of the Mef2-positive cardial cells were missing (Fig. 4A, arrowheads). One homozygous lethal line containing a P element insertion within the CG7983 gene, l(3)01814, was identified from public stocks and further characterized. We called the CG7983 gene simj (simjang). In the WT embryo, six Mef2-positive cardial cells were present within each hemisegment (Fig. 4 B and C). Of those, four cells were Tin-positive (Fig. 4 C, E, and G, arrowheads) and two were Seven-up (SVP)-positive cardial cells (Fig. 4C, star) that are also known to express Pericardin. In the simj mutant embryo, only four cardial cells were detected (Fig. 4 B and D), suggesting that two of the Mef2-positive cardial cells were missing. Further analysis with an anti-Tin antibody showed that only two of the Tin-positive cells were present in each hemisegment (Fig. 4 F and H, arrows), indicating that the two missing Mef2-positive cardial cells were the Tin- but not the SVP-positive cardial cells (Fig. 4B). These two cells, which are Mef2-positive and Tin-negative in the simj mutant embryo (Fig. 4 D and F, stars), were indeed detected by the anti-Pericardin antibody, confirming these results (data not shown). In the simj mutant embryo, a subset of pericardial cells was also missing, and the size of the individual nucleus was bigger than normal (Fig. 4 D and F). The simj gene (CG7983) encodes a protein containing a zinc finger domain. Surprisingly, a mammalian homologue of the SIMJ protein, p66, was identified recently as a component of the MeCP1 complex that is recruited by methyl-CpG-DNA binding proteins MBD2 and MBD3 (28). In fact, we found that SIMJ and MBD proteins interact and are coexpressed in cardiac tissues (Fig. 5). Also, in the Drosophila mbd-like gene (CG8208) mutant embryo or in the embryos overexpressing the simj gene, cardiac patterning was perturbed (Fig. 4J). Together, these results indicate that simj function is required for specifying a subset of cardiac precursors. It is known that combinatorial actions of inductive signals and cardiogenic transcription factors, which involve a complex program of genetic control, are essential for heart specification and determination (3). However, it was not known whether an epigenetic program was involved in heart specification. Thus, our results provide evidence that epigenetic information is crucial for specification of cardiac precursors within each segment.

Fig. 4.

simj gene function is required for the specification of a subset of cardiac cells. (A) Embryo injected with the simj dsRNA is stained for Mef2. Arrowheads indicate missing cardial cells. (Inset) Embryo stained with X-Gal. (B) Schematic diagram showing loss of cardiac cells in the simj mutant embryo. CC, cardial cell; PC, pericardial cell; T, Tin-positive cardial or pericardial cell; S, SVP-positive cardial cell; E, Eve- and Tin-positive pericardial cell. Only Tin-positive pericardial cells are depicted. (C–F) Embryos collected from either WT or simj mutant flies are subjected to immunostaining with the indicated antibody. Dorsal view of stage 16 WT (C and E) and simj mutant (D and F) embryos stained for Mef2 (C and D) or Tin (E and F). Arrowheads indicate Tin-positive cardial cells, and stars indicate SVP-positive cardial cells. In simj mutant embryos, only four cardial cells within each segment are stained for Mef2. Two are stained for Tin (arrows). Lateral view of stage 14 WT (G) and simj mutant (H) embryo fluorescently stained for Tin (green), showing that two Tin-positive cardial cells in each hemisegment are missing in the simj mutant embryo. (I and J) Forced expression of simj resulted in overspecification of cardial cells. Embryos were collected from either WT or transgenic flies ectopically expressing simj driven by dpp-gal4 (UAS-simj) and subjected to immunostaining with an anti-Mef2 antibody. Lateral view of stage 14 embryos is shown. Arrowhead indicates supernumerary Mef2-expressing cardial cells.

Identifying genetic components is an essential step toward understanding complex developmental processes. The Drosophila system has often been used as a powerful genetic tool to this end (29). In addition, due to the Drosophila genome sequencing (17), recent studies have shown that large-scale gene expression data using DNA microarrays can be used to identify genes that are expressed in particular tissues (30) and during the life cycle of Drosophila (31) or genes involved in specific biological processes (32). In this study, we performed a genomewide functional screen for cardiogenic genes using RNAi in the developing Drosophila embryo as a reverse genetic tool. These studies not only identified many key regulators but also revealed mechanisms underlying heart development. Given that the identified cardiogenic genes in this screen are well conserved in mammals, and combined with results from the conventional forward genetic screen in zebrafish (2) and studies from targeted gene ablation in the mouse (1), the information obtained from these studies should help us understand the molecular mechanisms underlying cardiac development.