Excessive Dpp signaling induces cardial apoptosis through dTAK1 and dJNK during late embryogenesis of Drosophila

Background To identify genes involved in the heart development of Drosophila, we found that embryos lacking raw function exhibited cardial phenotypes. raw was initially identified as a dorsal open group gene. The dorsal open phenotype was demonstrated to be resulted from the aberrant expression of decapentaplegic (dpp), a member of the tumor growth factor beta (TGF-β), signaling pathway. Despite the role of dpp in pattering cardioblasts during early embryogenesis of Drosophila have been demonstrated, how mutation in raw and/or excessive dpp signaling involves in the differentiating heart of Drosophila has not been fully elaborated at late stages. Results We show that raw mutation produced a mild overspecification of cardial cells at stage 14, but these overproduced cells were mostly eliminated in late mutant embryos due to apoptosis. Aberrant dpp signaling is likely to contribute to the cardial phenotype found in raw mutants, because expression of dpp or constitutively activated thickven (tkvCA), the type I receptor of Dpp, induced a raw-like phenotype. Additionally, we show that dpp induced non-autonomous apoptosis through TGFβ activated kinase 1 (TAK1), because mis-expression of a dominant negative form of Drosophila TAK1 (dTAK1DN) was able to suppress cell death in raw mutants or embryos overexpressing dpp. Importantly, we demonstrated that dpp induce its own expression through dTAK1, which also leads to the hyperactivation of Drosophila JNK (DJNK). The hyperactivated DJNK was attributed to be the cause of Dpp/DTAK1-induced apoptosis because overexpression of a dominant negative DJNK, basket (bskDN), suppressed cell death induced by Dpp or DTAK1. Moreover, targeted overexpression of the anti-apoptotic P35 protein, or a dominant negative proapoptotic P53 (P53DN) protein blocked Dpp/DTAK1-induced apoptosis, and rescued heart cells under the raw mutation background. Conclusions We find that ectopic Dpp led to DJNK-dependent cardial apoptosis through the non-canonical TGF-β pathway during late embryogenesis of Drosophila. This certainly will increase our understanding of the pathogenesis of cardiomyopathy, because haemodynamic overload can up-regulate TGF-β and death of cardiomyocytes is observed in virtually every myocardial disease. Thus, our study may provide possible medical intervention for human cardiomyopathy.

Results: We show that raw mutation produced a mild overspecification of cardial cells at stage 14, but these overproduced cells were mostly eliminated in late mutant embryos due to apoptosis. Aberrant dpp signaling is likely to contribute to the cardial phenotype found in raw mutants, because expression of dpp or constitutively activated thickven (tkv CA ), the type I receptor of Dpp, induced a raw-like phenotype. Additionally, we show that dpp induced non-autonomous apoptosis through TGFβ activated kinase 1 (TAK1), because mis-expression of a dominant negative form of Drosophila TAK1 (dTAK1 DN ) was able to suppress cell death in raw mutants or embryos overexpressing dpp. Importantly, we demonstrated that dpp induce its own expression through dTAK1, which also leads to the hyperactivation of Drosophila JNK (DJNK). The hyperactivated DJNK was attributed to be the cause of Dpp/DTAK1-induced apoptosis because overexpression of a dominant negative DJNK, basket (bsk DN ), suppressed cell death induced by Dpp or DTAK1. Moreover, targeted overexpression of the anti-apoptotic P35 protein, or a dominant negative proapoptotic P53 (P53 DN ) protein blocked Dpp/DTAK1-induced apoptosis, and rescued heart cells under the raw mutation background. Conclusions: We find that ectopic Dpp led to DJNK-dependent cardial apoptosis through the non-canonical TGF-β pathway during late embryogenesis of Drosophila. This certainly will increase our understanding of the pathogenesis of cardiomyopathy, because haemodynamic overload can up-regulate TGF-β and death of cardiomyocytes is observed in virtually every myocardial disease. Thus, our study may provide possible medical intervention for human cardiomyopathy.

Background
The Drosophila heart is a simple tubular organ located at the dorsal midline beneath the epidermis, and it is therefore alternatively termed the dorsal vessel. The fly heart consists of two major cell types, myocardial cells and pericardial cells, which arise from two bilateral rows of cardiac primordia at the leading edge of the migrating mesoderm. The contractile myocardial cells which form the lumen are arranged in a segmental repeat comprised of six cells per hemisegment in the mature embryonic heart. The pericardial cells, which are essential for normal cardiac function, are aligned alongside the myocardial cells. Despite its simple structure, fly heart has recently emerged as an excellent model system for dissecting the complex pathway that determines cardiogenic cell fate, and for investigating the physiologic function of the adult heart [1,2].
Extensive study has revealed that a combinatory action of extrinsic signaling and intrinsic transcription network is required for correct specification of cardial precursors and differentiation of mature heart (reviewed in [3]). Of all external signalings, Dpp, a member of the mammalian Transforming growth factor superfamily (TGF-β), has been shown to play a pivotal role during cardiogenesis of Drosophila [4]. The cardiogenic function of Dpp begins when it is expressed in the dorsal epidermis in a broad band along the anterior-posterior axis during germ band extension in Drosophila [5]. This spatiotemporal pattern of Dpp specifies the underling dorsal mesodermal cell fate by maintaining the expression of the transcription factor, tinman (tin) [4,[6][7][8]. Dpp also regulates the expression of several other cardiogenic transcription factors, including pannier (pnr) and dorsocross (doc) [9,10]. For further specification of the cardiogenic mesoderm, Wg signaling together with the combinatorial action of several transcription factors, including tin, pnr, doc and tailup, are required [11][12][13][14][15][16][17][18][19][20]. Around stage 10, Dpp expression in the dorsal ectoderm vanishes briefly, but reappears in the leading edge (LE) cells of the dorsal ectoderm at stage 11. This second round of Dpp expression in LE cells persists through stage 17 [21]. Interestingly, pMad, the activated Dpp signal transducer, can be detected in a subset of cardial progenitors in stages 12 to 14 [22]. This indicates that a second round of Dpp activity is required for further differentiation of Drosophila heart. Indeed, dpp mutants with alleles that affect the expression of Dpp in LE cells have impaired embryonic heart development and larval cardiac function [23,24]. These findings indicate a biphasic requirement for Dpp during cardiogenesis of Drosophila, in which it is required early for dorsal mesoderm patterning and later for differentiating heart cells.
Dpp regulates many developmental processes, including cell fate determination, alteration of cell shape, proliferation, and apoptosis. Morphogenic function of Dpp in cell fate determination has been shown to be mediated through the canonical pathway, in which it interacts with a type I receptor, Tkv, and a type II receptor, Punt. Upon formation of ligand-receptor complex, activated Punt phosphorylates Mad, which subsequently interacts with Medea. The resultant complex containing pMad, and Medea is then translocated into the nucleus where it activates transcription of Dpp target genes [25]. Other than the canonical pathway, it has been found that mammalian Dpp homolog, TGF-β transduce its signaling that is independent of Smad, a homolog of Drosophila pMad. The Smad-independent pathway is designated as the non-canonical TGF-β pathway. In the non-canonical pathway, TGF-β activated kinase 1 (TAK1) forms a multiple protein complex protein complex with TRAF6, TAB2, and TAB3. Upon binding of TGF-β to its receptor, TRAF6 exerts its E3 ubiquitin ligase activity together with ubiquitin-conjugating enzymes to catalyze Lys63-linked polyubiquitination. Subsequently, the Lys63-linked polyubiquitin chain associates with TAB2 which leads to autophosphorylation and activation of TAK1 [26,27]. Despite it is less clear in Drosophila, many functionally-conserved non-canonical signaling transducers, including DTAK1 and TAB2, have been identified in Drosophila [28,29]. Moreover, Dpp signaling has also been shown to control the viability of cells. A lack of Dpp signaling activates c-Jun N-terminal kinase (JNK)-dependent apoptosis in wing discs [30]. Interestingly, Dpp is also likely to function as a pro-apoptotic signal because increased Dpp activity leads to both non-autonomous JNK activation and cell death [31]. However, the link between Dpp signaling and JNK-mediated apoptosis is currently unclear in Drosophila.
Dpp is a downstream target of the JNK pathway, a conserved and pleiotropic signaling system whose function governs many different biological activities, including morphogenesis, differentiation, proliferation and apoptosis. Components of the DJNK pathway, including Djun and Raw, have been shown to participate in Drosophila heart development by modulating the expression of Dpp [14,16]. In Djun mutant embryos, the expression of dpp is not maintained at dorsal edge, which leads to downregulation of cardiac tin at later stages [16]. By contrast, pericardial cells are overspecified in raw loss-of-function mutant embryos [14]. The excessive differentiation of cardial cells has been attributed to the ectopic Dpp activity induced by dysregulated DJNK signaling.
In our effort to identify genes involved in the heart development of Drosophila, we observed that embryos lacking raw function exhibited cardial phenotypes, in which heart cells were overspecified in moderately degree during mid-embryogenesis, and that the overproduced heart cells had disappeared at late stages. We show here that the elimination of heart cells in late raw mutant embryos was a result of excessive apoptosis, and ectopic Dpp signaling was responsible for the cardial apoptosis phenotypes of raw mutant embryos. We also found that elevated Dpp can function as a pro-apoptotic signal to promote non-autonomous apoptosis in a dose-dependent manner. Importantly, ectopic Dpp auto-regulate its own expression through DTAK1. The autocrine Dpp further enhanced the expression of DJNK and consequently led to P53-dependent cell death. Our study defined a novel pathway which linked ectopic Dpp signaling and DJNK-dependent apoptosis during late cardiogenesis of Drosophila.

Plasmid construction and generation of transgenic flies
The UAS/Gal4 binary expression system was used to drive the expression of transgenic constructs [37]. For generation of the UAS-raw transgenic construct, the full-length raw cDNA was amplified using pNB3301 (obtained from A. Letsou) as template with a pair of primers (5'-CAC-CATGAAAACTGAAAGCAGCAGT-3' and 5'-GCAGC GGTCGCGGTTGTTGT-3') by PCR. The amplified DNA fragment was first cloned into a pENTR/D-TOPO vector (Invitrogen), and subsequently into a pTWF vector (The Drosophila Genomics Resource Center). The plasmid construct was confirmed by sequencing before germ-line transformation. The raw-RNA interference construct, UAS-raw-RNAi, was generated by amplification of a 693 bp DNA fragment using the following primers, 5'-GCTCTAGACCTGGAGCGCCAGAGTCTC-3' and 5'-GCTCTAGATGACGAAGAGCAACACTCG-3'. The amplified DNA fragment was digested with XbaI and cloned into the AvrII site of a pWIZ vector. The correct clone was used to clone the same XbaI-digested DNA fragment into the NheI site, as described elsewhere [38]. The orientation of the inverted DNA fragment was confirmed by restriction enzyme digestion. Flies carrying transgenic constructs were generated by P-elementmediated germ-line transformation procedure using w 1118 as the parental line [39].

Cell death detection
Apoptosis was detected by acridine orange (AO) staining or terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL). For AO staining, the same protocol was followed as used elsewhere [40]. Briefly, dechorionated embryos were stained by placing in an equal volume of n-heptane and 1 × PBS containing 5 mg/ml of acridine orange (Sigma) for 5 minutes on a shaking platform. For TUNEL analysis, an in situ cell death kit was used according to the manufacturer's instructions (Roche Applied Science). To detect cardiac apoptosis in embryos, a TUNEL and immunoflourescence double-labeling protocol was followed as described [41]. Stained embryos were mounted either with mineral oil (Sigma) or series 700 Halocarbon oil (Halocarbon Products, Hackensack, NJ). Samples were viewed either with a fluorescence or TCS SP2 confocol microscope (Leica Microsystems).

Image acquisition and processing
Epifluorescence images were acquired using a digital camera (CoolSnap 5.0, Photometrics) steered by the Northern Eclipse 6.0 software (EMPIX Imaging, Mississauga, Ontario, Canada). When necessary, Z-series of optical or fluorescent images were acquired at 2 μm increments with a piezo-electric motor (LVDT, Physik Instruments). The Helicon Focus program was applied to combine the focused images (Helicon Soft Ltd. Kharkov, Ukraine). All the figures were arranged in Adobe Illustrator CS3 (Adobe Co.).

Loss of raw function impairs heart and muscle development
In our effort to identify genes involved in the heart development of Drosophila, we found that many cardial cell types were missing in raw mutant embryos. For instance, using anti-Eve antibodies we found that Evenskipped positive pericardial cells (EPCs) were aligned normally along the dorsal vessel in wild-type embryos ( Figure 1A). However, these EPCs almost completely disappeared in the raw mutant at stage 16 ( Figure 1D). Similarly, absence of pericardial cells was observed in raw mutant using anti-EC11 and anti-Odd antibodies, which labelled extracellular matrix and Odd-skipped pericardial cells (OPCs), respectively, of pericardial cells ( Figure 1B, E and Additional file 1, Fig. S1). Additionally, using a heart-specific enhancer trap line, E2-3-9, we found that svp-expressing myocardial cells were reduced and/or missing under the raw mutant background (Figure 1C vs. 1F, arrow). To further investigate whether raw is involved in the early cardiogenesis of Drosophila, we used Tin as a marker because it is expressed initially in all cardial progenitors and later in four of six cardioblasts per hemisegment as well as a subset of pericardial cells in the mature embryonic heart ( Figure 1H and 1I; see also [46][47][48][49]). We found that the expression of Tin in the dorsal mesoderm was normal in both wild-type and raw mutant embryos at stage 12 ( Figure 1G and 1J). However, Tin-expressing heart cells were mildly overproduced in raw mutants at stage 14 ( Figure 1K, braces). Nevertheless, these overproduced heart cells were reduced and/or absent in raw mutants during late embryogenesis ( Figure 1L). Since the above markers label most of the cardial and pericardial cell types, our results suggest that the raw mutation affects all the cardial cell types in developing Drosophila heart. Based on the above observations, we thus concluded that heart cells were over-specified mindedly during mid-embryogenesis, and that these overproduced heart cells disappeared at late stages in raw mutant embryos.
To further confirm our findings, we used a him-GFP reporter in which a cardial-specific enhancer of him was placed upstream of nuclear GFP [33]. The GFP reporter was expressed in the precursors of both muscle and heart cells at stage 12 and its cardiac expression persisted till late embryogenesis in the differentiated heart cells under the control of tin, while its muscle expression was greatly reduced at stage 14 ( Figure 1M and 1N; and M-T. Su unpublished results). We found that the expression of him-GFP was increased in both heart and musculature at stage 14 under the raw mutation background ( Figure 1P). Consistent with the above data, cardiac expression of him-GFP decreased significantly in mutant embryos at stage 16 ( Figure 1O and 1Q). Taken together, these data show that heart precursors are overspecified during mid-embryogenesis, but are missing in late mutant embryos.

Down-regulation of raw causes cardial apoptosis
How could the cardial cells disappear in the raw mutant embryos at later stages? One possibility is that loss of raw function induces programmed cell death (PCD). To investigate whether cardial apoptosis occurred in raw mutant embryos, we stained embryos with a vital dye, acridine orange (AO), which provides a rapid visual assessment of apoptosis in live Drosophila [40]. Epifluorescence micrography showed apoptotic cells, mainly in the cephalic ganglia and in head regions in wild-type embryos at stage 14 ( Figure 2A). A similar AO staining pattern was detected in raw 1 mutant embryos at the same stage ( Figure 2D). Excessive cell death in the dorsal mesoderm was observed in raw 1 mutant embryos at stage 16 ( Figure 2B vs. 2E). To verify the above findings, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) was applied. Indeed, TUNEL positive nuclei were identified in the dorsal mesoderm of raw 1 mutant embryos at stage 16 ( Figure 2C vs. 2F). We then made use of RNA interference to specifically knock down the expression of raw ( Figure 2G and 2H), and the results were consistent with the above findings, founding that silencing the expression of endogenous raw in ectoderm using 69B-gal4 successfully induced localized apoptosis in the dorsal mesoderm of embryos ( Figure 2H).
Since the embryonic heart of Drosophila is located in the dorsal mesoderm, which was also the region where cell death was present in the raw mutants, we suspected that PCD is the cause of eliminating cardial cell types in the raw mutant. To test this, we double-labeled raw mutant embryos with TUNEL and heart-specific him-GFP reporter ( Figure 2I-L). By confocal microscopy analysis, the results showed that many him-GFP expressing cells were co-labeled with TUNEL, indicating that mesodermally-derived tissues did undergo apoptosis ( Figure  2I, yellow nuclei surrounded with green cytoplasma). Since the expression of him-GFP was mainly in cardial and muscle cells under raw mutation background at late stages ( Figure 1Q), these results indicate that raw mutation does induce death of heart cells. Apart from the him-GFP-expressing cardial and muscle cells, we noticed few unidentified TUNEL-positive cells ( Figure 2I, red, arrowheads). Since heart is the major defective tissue in the mesoderm of raw mutant, we have focused our attention on how loss of raw function leads to cardial apoptosis in Drosophila during late embryogenesis.

Function of raw is required in ectoderm
Previous study showed that mutation of raw resulted in dorsal open phenotype [35]. Ventral denticle belts are also missing in embryos homozygous for the null raw 1 allele ( Figure 3B, see also [50]). As demonstrated above, lack of raw function leads to apoptosis of dorsal mesodermal tissues, including heart. These findings suggest that raw is a pleiotropic gene which is required for the normal development of multiple tissues in Drosophila.
To determine the spatial requirement of raw, we performed rescue experiments by expressing raw using 69B-or 24B-gal4 driver, which direct the expression of UASraw in ectoderm or mesoderm respectively (69B>raw or 24B>raw). Consistent with a previous study, we found that the cuticular phenotypes of raw mutant could be rescued by targeted expression of Raw protein under the control of 69B-gal4 ( Figure 3A, C, E, see also [50]). In fact, 55% of the raw mutant flies survived to adulthood after rescue of the cuticular phenotype (data not shown). By contrast, expression of the UAS-raw transgene driven by 24B-gal4 failed to restore the dorsal open or the loss of denticle belt phenotypes ( Figure 3D and 3E). Conversely, we were able to replicate the cuticular phenotype by knocking down endogenous raw using a 69B-gal4 driver ( Figure 3F). However, no cuticle defect was observed when the raw-RNAi was driven with the pan-mesoderm driver 24B-gal4 ( Figure 3G).
Rescue experiments were also conducted to determine whether raw functions cell-autonomously for the survival  of mesodermally-derived tissues. We found that epidermal expression of raw suppressed cardial apoptosis under raw mutant background ( Figure 3H). However, mesodermallyoverexpressed raw was unable to completely inhibit cardial apoptosis in the raw mutant ( Figure 3I, brackets). Additionally, transgenic raw-RNAi construct driven by 69B-gal4, but not 24B-gal4, produced a cardial apoptosis phenotype ( Figures 2H, 3J vs. Figure 3K, brackets). These results strongly suggest that ectodermal Raw is sufficient for the proper development of Drosophila.

Raw affects Dpp and Wg signalings
Having established that ectodermal Raw functions in a cell non-autonomous manner to affect the viability of dorsal mesodemal cells, we speculated that ectodermallysecreted factors might be responsible for the cardial apoptosis observed in the raw mutant. In this regard, Wg and Dpp are good candidates because they are essential for patterning dorsal mesodermally-derived tissues, including cardial progenitor cells. Previous study has shown that Dpp is ectopically expressed at the dorsal epidermis of raw mutant embryos [35]. Using a dpp-LacZ reporter, we confirm that Dpp signaling is ectopically activated in raw mutant embryos at stage 14 ( Figure 4A, B vs. Figure 4C, D). To further determine if Dpp signaling is altered in raw mutants, we conducted immunocytochemistry experiments using a monoclonal antibody against pMad [43], a Dpp-activated Smad protein. pMad immunoreactivity was detected as a broad band in the dorsal ectoderm of both wild-type and raw mutant embryos at stage 13 ( Figure 4E and 4G). The broad band expression pattern disappeared in wild-type embryos but not in the dorsal ectoderm of raw mutants at stage 14 ( Figure 4F vs. 4H, brackets). Moreover, optical sectioning through the stained embryo showed the same broad band expression pattern of pMad in mutant but not in wild-type embryos ( Figure 4I and 4K, brackets). At late stages, pMad was not detected in the dorsal ectoderm of either mutant or wild-type embryos ( Figure 4J and 4L).
Immunostaining was also used to investigate if Wg signaling is affected by the raw mutation. We found that Wg was expressed in transverse striped domains of the ectoderm along the anteroposterior axis in wild type embryos at stages 13 (Additional File 2, Fig. S2). The expression pattern of Wg was not altered in raw mutants by stage 13. However, the expression level of Wg had significantly decreased in raw mutants at stage 14 (Additional File 2, Fig. S2). At stage 16, residual Wg was detected in the ectoderm of wild-type embryos, but was completely absent in raw mutant embryos (Additional File 2, Fig. S2).

Ectopic Dpp signaling promotes cell death
To investigate if down-regulation of Wg signaling causes cardial apoptosis, we conducted a temperature shift experiment using a temperature-sensitive wg IL114 allele. Inactivation of Wg at 9-15 hr, the time window during which Wg expression is missing in raw mutant embryos, did not cause the cardial apoptosis seen in raw mutant embryos, and overexpression of wg did not suppress the apoptosis phenotype under raw mutation background (Additional File 2, Fig. S2). We thus concluded that defective Wg signaling is not the cause of cardial apoptosis.
We then turned our attention to the question of whether ectopic Dpp signaling can induce cardial apoptosis, and found that overexpression of dpp, using 69B-gal4 (69B>dpp), phenocopied the raw mutant phenotypes ( Figure 5A-F). For instance, him-GFP expressing heart cells were overproduced at stage 14, but they had disappeared in 69B>dpp embryos at stage 16 ( Figure  5A-B and Figure 5D-E). Naked cuticle and dorsal open phenotypes were also observed in 69B>dpp embryos ( Figure 5C). However, unlike raw mutant embryos in which only him-GFP expressing cardioblasts were lost, both heart and muscle cells were completely abolished in 69B>dpp embryos at stage 16 ( Figure 1Q and 5E). This loss of him-GFP-expressing myoblasts and cardioblasts in 69B>dpp embryos is very likely to have resulted from apoptosis, because extensive cell death was detected throughout the entire 69B>dpp embryos, whereas dead cells were located mainly in the dorsal mesoderm in raw mutants ( Figure 2E and 5F). The difference in the apoptosis phenotype between raw mutant and 69B>dpp embryos might reflect the fact that the ectopic dpp was mainly found in the dorsal ectoderm of raw mutant whereas dpp was expressed in the entire ectoderm under the control of 69B-gal4. These findings reinforce our hypothesis that Dpp activity is correlated with cardial apoptosis.
Dpp is a secretory protein which can be transduced to the underlying mesoderm. If the apoptotic phenotype is associated with the ectopic Dpp activity, it would expect that mesodermal overexpression of Dpp should replicate the apoptosis phenotypes seen in 69B>dpp embryos. Indeed, mis-expression of dpp in mesoderm using a pan mesodermal driver, 24B-gal4, induced excessive cell death accompanied with elimination of him-GFP-expressing cells during late developmental stages (Additional File 3, Fig. S3). These results suggest that excessive dpp in ectoderm or mesoderm can cause apoptosis and removal of cardial cells.
Dpp transduces its signaling by binding to a heteromeric type I/type II transmembrane serine/threonine kinase receptor complex, encoded by tkv and punt. To further demonstrate that ectopic Dpp signaling is the cause of apoptosis in raw mutants, we targeted overexpression of a constitutively-active form of tkv, tkv CA , using a 24B-gal4 driver (24B> tkv CA ). We found that  24B>tkv CA embryos exhibited the same phenotype as 69B>dpp or 24B> dpp embryos, in which him-GFPexpressing cardial cells were increased in 24B> tkv CA embryos at stage 14 ( Figure 5G), while him-GFP-expressing heart and muscle cells were eliminated at stage 16 ( Figure 5H). Extensive cell death was also detected throughout 24B> tkv CA embryos ( Figure 5L).
In a complementary experiment, we depleted the expression of tkv by RNAi using a transgenic fly carrying inverted repeats corresponding to tkv, under the control of a UAS sequence inducible by 24B-gal4 (24B>tkv-RNAi, Additional file 4, Fig. S4). Down-regulation of mesodermal tkv induced a moderate degree of apoptosis in which the AO-positive cells were randomly distributed in the entire 24B>tkv-RNAi embryos ( Figure  5I vs. 5K). A similar pattern of scattered cell death was also detected in loss-of-function of tkv mutant embryos ( Figure 5J). These data suggest that Dpp is required for cell survival during embryogenesis in Drosophila. To test whether inhibiting Tkv activity can block excessive Dpp-induced apoptosis, we compared the AO staining pattern in raw, tkv and raw;tkv double mutant embryos. Both tkv and raw;tkv double mutant embryos exhibited the same scattered AO staining pattern, unlike the cardial apoptosis phenotype observed in raw mutant (Figure 2E vs. 5J and 5M). Moreover, mesodermal knockdown of the expression of tkv was able to suppress cardial apoptosis under the raw mutant background ( Figure 5N). These results indicate that ectopic Dpp induced apoptosis is tranduced through Tkv.
As above-mentioned, ectopic tkv expression promoted apoptosis in 24B>tkv CA embryos, indicating that Dpp can function as a pro-apoptotic signaling. To directly test this hypothesis, Dpp was constitutively overexpressed under the control of the ubiquitous Act5C-gal4 driver (Act5C>dpp). In control Act5C-gal4 embryos, AO staining revealed that apoptotic cells were mainly present in the ventral nerve cord and head regions at stage 16 ( Figure 5P-R). Raising the temperature slightly increased the amount of cell death in the central nervous system of late Act5C-gal4 embryos ( Figure 5O-R). Constitutive overexpression of Dpp caused a remarkable degree of apoptosis in Act5C>dpp embryos ( Figure 5S-U). Notably, increasing the dpp expression level by raising the temperature significantly increased the number of dying cells ( Figure 5O-U). This result suggests that Dpp can function as a pro-apoptotic signaling in a dosedependent manner.

Dpp induced apoptosis is mediated through DTAK1
As shown above, reducing Tkv activity successfully blocked dpp-mediated cardial apoptosis in raw mutants ( Figure 5M and 5N). This data encouraged us to further examine whether overexpression of brinker (brk), a transcriptional repressor of Dpp, could inhibit the cardial apoptosis phenotype in raw mutants. Overexpression of brk using 24B-gal4 driver (24B>brk) caused a moderate degree of apoptosis ( Figure 6A). The AO staining pattern in 24B>brk embryos was similar to the pattern in tkv 7 and 24B>tkv-RNAi embryos (Figures 6A,  5J, and 5K), suggesting that ectopic brk-induced apoptosis was likely to be a result of the inhibition of Dpp signaling. Unexpectedly, mis-expression of brk in mesoderm failed to suppress the cardial apoptosis phenotype of the raw mutant ( Figure 6B, brackets), suggesting that the suppressor activity of Brk may not be strong enough to block Dpp-mediated cardial apoptosis. Nevertheless, it is equally possible that Dpp-induced apoptosis is mediated through a distinct pathway which can not be repressed by Brk.
A previous study showed that TGF-β activated kinase 1 (TAK1), a member of the JNKK kinase superfamily that activates the JNK cascade, can transduce TGF-β signaling and induce apoptosis in vertebrates [51]. To determine if TAK1 induces apoptosis in response to Dpp signaling, a Drosophila TAK1 homolog, DTAK1, was mis-expressed using either Act5C-or 24B-gal4. Ectopic DTAK1 activity was sufficient to induce apoptosis in embryos ( Figure 6C and 6D). In contrast, ectodermal or mesodermal overexpression of a dominant negative form of dTAK1 (dTAk1 DN ) was capable of suppressing developmental programmed cell death ( Figure  6E and 6F). Constitutive overexpression of dTAK1 DN also effectively inhibited ectopic Dpp-induced apoptosis ( Figure 6G). Moreover, mesodermally overexpressed dTAK1 DN suppressed cardial apoptosis under the raw mutation background ( Figure 6H). Taken together, our results clearly demonstrate that DTAK1 is a downstream effector of Dpp-mediated cardial apoptosis.

DTAK1 induces JNK dependent apoptosis
Although our data suggest that dTAK1 acts downstream of dpp to promote apoptosis, it is interesting to note that dTAK1 stimulates dpp expression, suggesting that dpp is a downstream target of dTAK1 [29]. Despite the discrepancy in these epistatic relationships, these observations strongly suggest that dTAK1 and dpp act in the same genetic pathway. Using a dpp-LacZ reporter, we found that dTAK1 activated the transcription of dpp to the same degree as basket (bsk), the Drosophila Jun amino-terminal kinase (DJNK) homolog encoded gene which activates the expression of dpp in the dorsalmost epidermal cells ( Figure 7A-C). These results indicate that dTAK1 acts upstream of dpp. Since dpp is also activated by bsk, these results suggest that dTAK1 activates bsk and thereby dpp. Moreover, we found that overexpression of dTAK1 using 24B-gal4 driver stimulated the expression of pMad in the entire mesoderm ( Figure 7D vs. Figure 7E), indicating that the ectopic Dpp induced by DTAK1 can transduce its own signaling through Tkv and result in the expression of ectopic pMad. As Dpp can transduce its signaling through DTAK1 to promote apoptosis ( Figure 6G), the ectopic Dpp induced by DTAK1 is likely to autoregulate itself through DJNK. In other words, a positive autoregulatory loop for Dpp expression is formed when DTAK1 is activated.
The JNK cascade is a well-known pro-apoptotic signaling that participates in stress-related apoptosis in Drosophila. Together with the fact that TAK1 can activate the JNK cascade, the autocrine Dpp is expected to enhance the activity of DTAK1 and thereby hyperactivitates the DJNK pathway. In other words, the ectopic dpp-induced cell death could be a consequence of hyperactivated JNK signaling. To test this hypothesis, bsk was overexpressed in the entire embryo or specifically in mesoderm using either Act5C-or 24B-gal4 driver (Act5C>bsk; 24B>bsk). Ectopic bsk activity dramatically increased levels of apoptosis in the embryos ( Figure 7F and 7G). Conversely, overexpression of a dominant negative bsk, bsk DN , again using either Act5Cor 24B-gal4 driver, was able to suppress ectopic Dpp/ DTAK1-induced apoptosis ( Figure 7H and 7I). These results demonstrate that Dpp exerts its pro-apoptotic functions through DTAK1 as well as through DJNK.

Overexpression of anti-apoptotic P35 or dominant negative P53 blocks Dpp induced apoptosis
As demonstrated above the missing of cardial cells in raw mutants is likely to be a consequence of excessive dpp induced apoptosis. If this is the case, we expect that blocking cell death would reverse the cardial apoptosis phenotype in raw mutant embryos. The baculovirus P35 protein has been shown to suppress normal and induced apoptosis by inhibiting caspases in animals [36]. We found that overexpression of the anti-apoptotic P35 could prevent apoptotic cell death induced by either ectopic Dpp or DTAK1 ( Figure 8A and 8C). P35 also significantly reduced the death of heart cells under the raw mutation background ( Figure 8E). Moreover, it has been shown that aberrant JNK-induced apoptosis is mediated through P53 (reviewed in [52]). To further demonstrate if Dpp/DTAK1-induced apoptosis is P53dependent, a dominant negative form of p53 (p53 DN ) was ectopically expressed using either Act5C-or 24B-gal4 driver (Act5C>p53 DN or 24B> p53 DN ). We found that overexpresion of p53 DN suppressed Dpp-or DTAK1-induced apoptosis successfully ( Figure 8B and 8D), suggesting that Dpp/DTAK1-induced apoptosis is mediated through P53. We also showed that mesodermal overexpression of P53 DN significantly reduced the death of heart cells in raw mutant embryos ( Figure 8F). Consistent with the above observation, immunostaining experiments revealed that mesodermally-overexpressed P53 DN was able to rescue Tin-positive cardial cells in raw mutant embryos ( Figure 8G and 8H). In fact, superfluous Tin-positive cells were usually observed when cardial apoptosis was blocked by ectopic P53 DN expression ( Figure 1I vs. Figure 8H, arrow), suggesting that the overspecified cardial cells during mid-embryogenesis in raw mutants survived when apoptosis was blocked.
These results also indicate that the Dpp-DTAK1-DJNK mediated apoptosis pathway is likely to be P53dependent.

Function of Raw in DJNK signaling pathway
Analysis of the amino acid sequence of Raw revealed that it does not comprise any specific functional domain or motifs. Despite the fact that the C-terminus of Raw protein is rich in glutamine residues, a characteristic feature of some transcription factors, Raw protein was mainly detected in cytoplasma, suggesting that it may not play a role in transcriptional regulation. The most prominent feature of Raw protein is that it contains two duplicated domains with a 32 amino acid core repeat (designated as Raw repeats) that is highly conserved in invertebrates [35]. The structure of Raw is currently unavailable, so no further information regarding the function of Raw can be deduced through sequence analysis.
Epistasis analysis has demonstrated that raw negatively regulates jra activity in parallel to DJNK signaling in the epidermis of Drosophila [35,50]. As demonstrated above, Dpp-induced apoptosis is mediated through activation of DJNK signaling ( Figure 7H). If raw can negatively regulate DJNK, it would be expected that forced expression of raw would inhibit the DJNK cascade and reverse the apoptotic phenotype in the raw mutant. Ectodermal overexpression of raw did rescue the cardial apoptosis in raw mutants ( Figure 3H). Nevertheless, targeted expression of raw in the mesoderm was unable to prevent cell death under the raw mutation background ( Figure 3I). Why is the effect of mesodermal Raw so different from that of ectodermal Raw? As demonstrated above, Dpp can trigger its own expression in the mesoderm through DTAK1 and DJNK ( Figure 7A-E). This auto-regulated Dpp is expected to further enhance DJNK activity, which may out-compete the suppressive activity of Raw. This might explain why mesodermal expression of raw was unable to suppress raw mutationinduced apoptosis. However, if the initial Dpp signaling in ectoderm is suppressed then it will not initiate the autocrinal Dpp signaling in mesoderm and result in cell death. Consistent with our notion, we found that ectodermally-overexpressed raw suppressed the expression of pMad in raw mutants. In contrast, targeted expression of raw in mesoderm did not inhibit the expression of pMad under the raw mutation background (Additional File 5, Fig. S5).

Dpp functions as survival and pro-apoptotic signals
The morphogenetic function of Dpp patterns cell fates across the developing field by forming a gradient which provides position identity for the receiving cells. Similarly, it seems that Dpp controls the viability of cells in the same concentration-dependent manner. Mutant cells deprived of Dpp signaling are lost from wing disc epithelium due to DJNK activation and apoptosis [30,53]. These observations have suggested that Dpp functions as a survival factor by preventing activation of the DJNK-dependent apoptotic pathway. Additionally, down-regulation of mad activated JNK and caspase-3, indicating that Dpp functions as a survival factor mediated through Mad [54]. Consistent with these findings, we found that down-regulation of tkv or overexpression of brk induced moderate cell death in embryos ( Figure 5J, K and 6A). Nevertheless, Dpp seems to act as a double-edged sword because increased Dpp signaling induces DJNK-mediated apoptosis in the proximal wing [30]. Similarly, in this study, we show that ectopic dpp or tkv CA expression promotes DJNK-mediated apoptosis in embryos ( Figure 5F and 5L). Moreover, the apoptotic propensity of Dpp is proportional to its own expression level ( Figure 5O and 5S-U). Taken together, these results suggest that Dpp can act as both survival and death signals and thus, an appropriate expression level of Dpp is indispensible for the survival of cells during development.
TAK1 is a key transducer that mediates ectopic Dpp induced apoptosis TAK1 is also a member of the MAPKKK family, which was originally identified as a mediator of TGF-β signaling pathway [55]. It has also shown to regulate a great variety of cellular processes through activating many down-stream kinase cascades, including I-kappa B kinase complex (IKK), p38 MAPK, JNK, and AMP-activated protein kinase (AMPK) (reviewed in [56]). Unlike the canonical pathway in which members of TGF-β family elicit phosphorylation of Smad proteins, activation of TAK1 was found to function in a receptor kinase-independent manner [26]. The existence of the non-canonical pathway may explain why mesodermal overexpression of brk was unable to block cardial apoptosis in raw mutants ( Figure 6B), because it seems that brk only suppresses Dpp target genes containing Mad consensus binding sites [57]. Interestingly, mesodermal pMad was increased in raw mutant or in embryos overexpressing dTAK1 ( Figure 4K, and 7E), indicating that both canonical and non-canonical pathways were activated simultaneously in response to ectopic Dpp signaling. The fact that apoptotic cell numbers increased dramatically in raw mutants or in embryos overexpressing dTAK1 ( Figure 2E, F, 6C and 6D) suggests that the non-canonical pathway suppresses the canonical pathway when cells are exposed to excessive Dpp levels. Together with the fact that the apoptotic propensity of Dpp is dose-dependent ( Figure 5O), these data imply that higher levels of Dpp elicit stronger DTAK1 activity and result in apoptosis. In support of this argument, homozygous raw mutant animals had significant apoptosis in the dorsal-most tissues, such as the heart, where Dpp activity is at its peak ( Figure 2E and 2F), and global overexpression of Dpp signaling led to ubiquitous cell death ( Figure 5F, L and 5S-U).

Late Dpp signaling in heart development
Compared to the early function of Dpp in patterning cardiogenic mesoderm, the function of Dpp during late cardiogenesis is less explored. Previously, studies had showed that numbers of various pericardial cell types, but not cardial cells, were increased in fly embryos with the dpp d6 mutant allele, whose expression was not maintained in the dorsal ectoderm during germ-band retraction [23,24]. The expression of the mitosis marker phospho-histone 3 was concomitantly increased in the dpp d6 mutant, suggesting that Dpp restricts the proliferation of pericardial cells specifically during late cardiogenesis [24]. As shown above, dpp signaling is ectopically-activated at stages 14-15 in raw mutants ( Figure 4D, E and 4K, see also [58]); this also gives us a good opportunity to decipher late Dpp function in the developing heart of Drosophila. As demonstrated above, ectopic Dpp signaling leads to both overproliferation of cardial cells during mid-embryogenesis and cardial apoptosis during late embryogenesis because overexpression of dpp or tkv CA replicated the raw mutant phenotypes ( Figure 5A-H). Interestingly, both cardial and pericardial cells were eliminated in either raw mutant, 69B>dpp or 24B>tkv CA embryos ( Figure 1D-F, L, O, Q, 5E and Figure 5H). Our data contradict the observation that targeted overexpression of tkv CA under the control of cardioblast-specific tinCΔ4-gal4 did not reduce the number of cardial cells significantly [24]. The disparate results may be due to differences in the temporospatial expression of the transgene driven by different gal4 drivers. Alternately, if the pro-apoptotic propensity of Dpp is concentration-dependent ( Figure 5O and 5S-U), as discussed above, it is also possible that the expression of tkv CA driven by tinCΔ4-gal4 is not strong enough to trigger the apoptosis response in cardioblasts.

Model for raw mutation mediated apoptosis
Based on our data, we propose a model to depict the genetic pathways involved in raw mutation-mediated apoptosis (Figure 9). At stage 14, Dpp activity in leading edge cells activates cardiogenic factors in the underlying mesoderm which are essential for the differentiation of dorsal mesodermally derived tissues, including the heart. Deficits in Raw function cause overexpression of dpp which increases the activities of cardiogenic factors and results in overgrowth of cardial cells at stage 14. At stage 15, the expression of Dpp in the LE cells is maintained by the DJNK cascade, and heart cells are continuously differentiated by the function of cardiogenic factors in wild-type embryos. In raw mutant embryos, ectopic Dpp activates DTAK1 which triggers the expression of Bsk as well as Dpp. The induced Dpp functions in an autocrine manner to further enhance activity of Bsk and eventually lead to P53-mediated apoptosis.
In support of our model, it has been found that TGFβ/SMAD signaling exerts its apoptotic function in an autocrine loop manner in rat cardiomyocytes [59]. Induction of cardiomyocyte apoptosis by caspase overexpression has been shown to cause lethality and dilated cardiomyopathy in mice [60]. In contrast, inhibition of cardiomyocyte apoptosis by treating with a caspase Figure 9 The genetic network that leads to cardiac apoptosis in raw mutation. At stage 14, Raw restricts bsk which limits the expression of dpp at ectodermal leading edge cells. Minimal Dpp activity maintains the activation of cardiogenic factors which are essential for the differentiation of cardial cells. In raw mutation, Bsk is de-regulated which causes overexpression of dpp in dorsal ectoderm at stage 14. Ectopic dpp signaling increases the expression of cardiogenic factors which results in overgrowth of cardial cell types at stage 14. At stage 15, while the expression of Dpp in the LE cells gradually decreases, heart cells are continuously differentiated with the function of cardiogenic factors. At stage 15, DTAK1 is activated by the ectopic Dpp signaling in raw mutant. DTAK1 activates DJNK and thereby Dpp. Autocrine Dpp further enhances DJNK and eventually leads to P53-dependent apoptosis.
inhibitor reduced apoptosis, improved cardiac function, and delayed progression of heart failure in a cardiomyopathy animal model [61]. Prolong haemodynamic overload can up-regulate TGF-β [62,63], and death of cardiomyocytes is observed in virtually every myocardial disease (reviewed in [64]. The pathway unraveled in this study is the first report that links ectopic Dpp and DJNK-dependent cardial apoptosis through the noncanonical pathway and dTAK1 activation. Our findings may thus suggest possible medical interventions for human cardiomyopathy.

Conclusion
By analyzing the heart defect phenotype of raw mutant embryos, we demonstrate that overexpression of dpp lead to cardial apoptosis during late embryogenesis of Drosophila. We also demonstrate that Dpp induces its own expression through dTAK1. The activation of dTAK1 causes the hyperactivation of Drosophila JNK (DJNK) thereby cardial apoptosis. This is the first report that links ectopic Dpp and DJNK-dependent cardial apoptosis through the non-canonical Dpp signaling pathway and dTAK1 activation. Since haemodynamic overload usually up-regulates TGF-β, a mammalian homolog of Dpp, and death of cardiomyocytes, the pathway delineated in this study may suggest possible medical interventions for human cardiomyopathy. However, its expression decreased significantly in raw mutant embryos at stage 14. (C, F) At stage 16, residual Wg staining was detected in the dorsal epidermis of wild-type embryos, but its expression was completely lost in raw mutants at stage 16. (G) raw mutation shows cadial apoptosis phenotype (brackets). (H) wg IL114 is a temperature-sensitive allele that mimics the null wg allele at non-permissive temperatures. Removal of wg function does not lead to cardial apoptosis phenotype in temperature shift experiment using wg IL114 allele. Incubation times (9-15 hr) were normalized to development at 25°C. (I) Ectopic wg expression driven by 69B-gal4, did not suppress cardial apoptosis in raw mutants (brackets). Additional file 4: Fig. S4. The expression of the endogenous tkv was specifically silenced as compared to rps17 mRNA in the embryos. RT-PCR products were resolved in 1.5% agarose gel and visualized with EtBr. WT, wild-type embryos.

Additional material
Additional file 5: Fig. S5. Ectodermally, but not mesodermlly overexpression of raw suppresses the ectopic pMad in raw mutant. (A) pMad was detected as a broad dorsal band in raw mutant (brackets). (B) Targeted expression of raw using 69B-gal4 inhibited ectopic pMad in raw mutant. (C) Forced expression of raw using 24B-gal4 can not inhibit ectopic pMad in raw mutant (brackets).