|Home | About | Journals | Submit | Contact Us | Français|
Somatic myogenesis in Drosophila relies on the reiterative activity of the basic helix-loop-helix transcriptional regulator, Twist (Twi). How Twi directs multiple cell fate decisions over the course of mesoderm and muscle development is unclear. Previous work has shown that Twi is regulated by its dimerization partner: Twi homodimers activate genes necessary for somatic myogenesis, whereas Twi/Daughterless (Da) heterodimers lead to the repression of these genes. Here, we examine the nature of Twi/Da heterodimer repressive activity. Analysis of the Da protein structure revealed a Da repression (REP) domain, which is required for Twi/Da-mediated repression of myogenic genes, such as Dmef2, both in tissue culture and in vivo. This domain is crucial for the allocation of mesodermal cells to distinct fates, such as heart, gut and body wall muscle. By contrast, the REP domain is not required in vivo during later stages of myogenesis, even though Twi activity is required for muscles to achieve their final pattern and morphology. Taken together, we present evidence that the repressive activity of the Twi/Da dimer is dependent on the Da REP domain, and the activity of the REP domain is sensitive to tissue context and developmental timing.
A central theme of developmental biology is the specification of different cell lineages from a common progenitor cell. A primary mechanism by which this occurs is the establishment of differential gene expression through transcriptional regulation. One way for cells to achieve precise transcriptional control during embryonic development is to differentially modulate the activity of transcriptional regulators. The question of how the regulation of these proteins is sufficiently dynamic to match their changing tissue environment remains unclear. In this study, we focus on a particular transcriptional regulator, Twist (Twi), and how its activity is controlled during multiple stages of Drosophila mesoderm development.
Twi activity is critical for sequential cell fate decisions within the mesoderm lineage. Twi is required first at gastrulation to activate genes critical for mesoderm development, such as snail (Ip et al., 1992; Leptin, 1991) and Drosophila myocyte enhancing factor 2 (Dmef2) (Cripps et al., 1998). Following gastrulation, Twi is necessary for the allocation of mesodermal cells to particular tissue fates, including somatic mesoderm, heart, visceral mesoderm, fat body and mesodermal glia (Baylies and Bate, 1996; Borkowski et al., 1995). During this period, Twi activates another set of genes that are necessary for myogenesis, including evenskipped (eve) (Sandmann et al., 2007). Subsequent to this allocation of mesodermal tissues, two populations of myoblasts within the somatic mesoderm are specified: founder cells (FCs) and fusion competent myoblasts (FCMs). Each FC contains the information necessary to form a somatic muscle, and the FCMs must fuse to FCs to achieve the final size, shape, and attachment site of each muscle (Bate, 1990; Beckett and Baylies, 2006). During this period, Twi directly regulates genes, such as blown fuse (Furlong et al., 2001; Sandmann et al., 2007), that are critical for muscle differentiation. Lastly, Twi is required to prevent premature differentiation of the adult muscle progenitor cells (Anant et al., 1998). Altogether, Twi’s activity profile corresponds to changes in the regulation of different sets of target genes through developmental time, as indicated by genetic analysis and microarray profiling studies (Azpiazu and Frasch, 1993; Cripps et al., 1998; Furlong et al., 2001; Sandmann et al., 2007; Shishido et al., 1993; Zeitlinger et al., 2007).
Dimer partner choice is a critical factor in determining Twi activity in both flies and vertebrates (Castanon et al., 2001; Connerney et al., 2006; Firulli et al., 2005). In flies, Twi, which belongs to the basic-helix-loop-helix (bHLH) family of transcriptional regulators, either homodimerizes or heterodimerizes with Daughterless (Da), another bHLH protein that is ubiquitously expressed throughout the embryo (Cabrera and Alonso, 1991; Campuzano et al., 1985; Caudy et al., 1988; Cronmiller and Cummings, 1993). Twi homodimers activate the transcription of myogenic genes, such as Dmef2, and direct cells to a somatic myogenic fate. In contrast, Twi/Da heterodimers lead to the repression of somatic myogenic genes, allowing for the development of other mesodermal tissues (Castanon et al., 2001).
In vertebrates, E proteins, the homologues of Da, are potent activators of transcription (Brunet and Ghysen, 1999; Markus et al., 2002). During vertebrate myogenesis, MyoD/E protein heterodimers are required for the activation of myogenic genes, whereas E protein homodimers are necessary for activating B-cell genes during B lymphocyte differentiation (Kadesch, 1992; Lassar et al., 1989). An analysis of E proteins in various vertebrate tissue culture cell lines has revealed an auto-regulatory domain, the repression (Rep) domain, which is required for MyoD/E proteins to mediate transcriptional activation of myogenic genes. The Rep domain is also required for preventing E protein homodimers from inappropriately activating myogenic genes within the B cell lineage (Markus et al., 2002). This study suggested that this domain, through its interactions with E protein activation domains, is crucial for differentially modulating the activity of E protein homodimers and heterodimers in specific tissue contexts.
Here, we provide insight to how Da dictates Twi activity in vivo during Drosophila myogenesis. We find that the repressive activity of Twi/Da dimers is sensitive to tissue context. We identify the REP domain in Da and show that Twi/Da transcriptional repression relies on the Da REP. The function of this domain is also dependent upon developmental timing. During the allocation of mesodermal cells to different fates, Twi/Da exerts its transcriptional repression through the REP domain. However, during the later stages of somatic myogenesis, where we show that Twi plays a role, Twi/Da repression does not require the REP domain. Altogether, our results reveal that the cis-regulatory REP domain determines Twi/Da activity during the allocation of mesodermal cells, but not during muscle differentiation. Our analysis of REP domain function highlights the dynamic regulation of Twi activity through developmental time. We speculate that similar mechanisms control Twi activity both in other species and in cancer.
Fly stocks used: twi-GAL4 (Baylies et al., 1995), da-GAL4 (Wodarz et al., 1995), twiID96/CyO, wg-LacZ (Simpson, 1983), Dmef2-LacZ (Cripps et al., 1998), twi:CD2 (Baylies and Bate, 1996; Borkowski et al., 1995), rP298-GAL4 (Menon and Chia, 2001), UAS-twi (Baylies and Bate, 1996), UAS-da (a gift from J. Campos-Ortega), UAS-twi-twi (Castanon et al., 2001), UAS-twi-da (Castanon et al., 2001), UAS-daΔ, UAS-twi-daΔ, UAS-twiRNAi, UAS-twi-daHA, and UAS-twi-daΔHA. The GAL4-UAS system (Brand and Perrimon, 1993) was used for expression studies. All genetic crosses were performed at 25°C. yw or OreR were used as wild type strains. Transgenic UAS-daΔ, UAS-twi-daΔ, UAS-twiRNAi, UAS-twi-daHA, and UAS-twi-daΔ HA flies were generated by injection of yw embryos as previously described (Baylies and Bate, 1996). Multiple independent transformant lines were obtained and tested from each construct.
To create the daΔ construct, two fragments of the da cDNA were amplified by PCR. Primers 5′CCGGTACCATGGCGACCAGTGAC3′ and 3′GCAGTAGCGTCTCGGGAATTCGG5′ containing a Kpn1 and an EcoRI site, respectively, amplified the portion of da corresponding to amino acids 1–493. Primers 5′CCGAATTCGGTGGTGGACACGCC3′ and 3′CAGCTTCCGCAATAGTCTAGACC5′ containing an EcoRI and a Xba1 site, respectively, amplified the portion of da corresponding to amino acids 530 to end. These fragments were ligated in frame and cloned into pCDNA3 (Invitrogen). To construct twi-daΔ linked dimers, a pCDNA3 vector containing the twi cDNA and a flexible linker were used (Castanon et al., 2001; Markus et al., 2002; Neuhold and Wold, 1993). The KpnI and ApaI restriction sites were used to clone daΔ in frame with twi cDNA and the linker. The pCDNA3 vector contains a CMV promoter and was used for all cell culture experiments. For P-element transformation, daΔ and twi-daΔ were subcloned into pUAST (Brand and Perrimon, 1993). HA tagged twi-da and twi-daΔ constructs were made by PCR amplification of twi-da and twi-daΔ linked dimer constructs using a primer that fuses a single HA transcript to the C-terminal end of the linked transcript. The forward primer, 5′ATATGCGGCCGCATGATGAGCGCTCGCTCGGTGTCG3′, contains a Not1 site and anneals to the 5′ end of the twi cDNA. The reverse primer 3′CCTGTGCGGGAAGGCGTTATGGGTATGCTACAAGGTCTAATGCGAATCG GTACCAGATCTCG′5, anneals to the 3′ end of the da cDNA, and contains the HA transcript, a Nco1 site and a Xba1 site. The PCR product from these primers, both twi-daHA and twi-daΔHA, were individually cloned into the pUAST vector using the Not1 and Xba1 sites (Brand and Perrimon, 1993). To create the twiRNAi construct, two copies of twi cDNA and a GFP linker were ligated and cloned into the p1180 vector. The twi cDNAs were ligated in forward and reverse orientation and separated by the GFP transcript linker (Llorens et al., 2007). Therefore, when expressed, this transcript forms a double-stranded RNA that undergoes RNAi processing (Piccin et al., 2001). The twiRNAi construct was then subcloned into the Bluescript SK vector (Stratagene) via the Hind III and Xba1 sites. For P-element transformation, the twiRNAi construct was subcloned into the pUAST vector using the Kpn1 and Xba1 sites (Brand and Perrimon, 1993). The integrity of all constructs was verified by sequencing.
For transfection assays, a reporter construct containing a 175bp Dmef2 enhancer (Cripps et al., 1998) cloned into the pGL2 Basic Vector (Promega) was used. Equal molar amounts of the pCDNA3 expression vector containing twi, da, twi-twi, twi-da, daΔ, or twi-daΔ was cotransfected with 3μ of the Dmef2 enhancer reporter plasmid and 3μ of the actin-LacZ plasmid, which served as a control for transfection efficiency (T. Lieber). The DNA concentration for each transfection was equalized by addition of pBluescript plasmid to a final concetration of 20 μg. Schneider Line 2 cells (S2; T. Lieber) were maintained and transfected as previously described (Castanon et al., 2001). Transfected cells were harvested and luciferase activity was assayed as previously described (Castanon et al., 2001). The data shown are mean values of at least three independent, triplicate transfections and are presented as the fold activation obtained in each sample over the luciferase activity generated in the absence of expression plasmid. Luciferase activity was normalized against β-galactosidase activity. DNA binding assays were performed as previously described (Castanon et al., 2001).
Immunocytochemistry was performed following standard techniques (Artero et al., 2003). Antibodies were preabsorbed (PA) against fixed yw embryos or in combination with the Tyramide Signal Amplification system (TSA; PerkinElmer Life Sciences). The following antibodies were used: anti-Mhc (1:10,000; TSA; D. Kiehart), anti-Mhc (1:200; gift of S. Abmayr), anti-Krüppel (1:2000; PA; J. Reinitz), anti-β-galactosidase (1:1000; Promega), anti-Fasciclin III (1:100; Developmental Studies Hybridoma Bank, Univ. of Iowa), anti-Zfh-1 (1:5000; Z. Lai), and anti-Eve (1:3000; PA; J. Reinitz). Biotinylated secondary antibodies (1:200) were used in combination with the Vector Elite ABC kit (Vector Laboratories, CA). Specimens were embedded in Araldite. Images were captured using an Axiocam digital camera (Zeiss). All mesodermal images are a merge of several focal planes and were combined into one image using Adobe Photoshop software. Fluorescent immunohistochemistry was conducted using anti-Eve (1:3000; PA), anti-β-galactosidase (1:1000; Promega), anti-HA (1:500; Roche), anti-Twi (1:200; PA, S. Roth), anti-Da (1:25; C. Cronmiller), anti-CD2 (1:10000; Serotech) and anti-Kr (1:2000, PA, J. Reinitz). Primary antibodies were detected using Alexa488 and Alexa555 conjugated secondary antibodies (1:500; Molecular Probes). Anti-Kr and anti-Da antibodies were detected by biotinylated secondary antibodies (1:200) and Alexa 488 conjugated streptavidin (1:200). Embryos were mounted in Vectashield (Vector Laboratories, CA). Embryos used for experiments that involved comparing protein levels by staining intensity, which includes β-galactosidase staining for the Dmef2-LacZ experiment, Twi staining for the twiRNAi experiment, and HA staining, were collected and processed together under identical conditions. Fluorescent images were acquired on a Zeiss LSM510 confocal microscope.
Previous studies have revealed that Twi has potent somatic myogenic capabilities in both the mesoderm and the ectoderm (Baylies and Bate, 1996; Castanon et al., 2001). To determine which dimer combination was responsible for this myogenic capability, a linked dimer technique was employed, which tethers two bHLH proteins together by a flexible serine/glycine linker and results in preferential dimerization (Castanon et al., 2001; Neuhold and Wold, 1993). Linked dimer proteins are denoted by a hyphen (Twi-Da), as opposed to unforced dimer pairs (Twi/Da). Overexpression of Twi-Twi linked homodimers were shown to have similar effects as overexpressed Twi monomers in activating somatic myogenesis in cardiac and visceral mesoderm as well as converting ectoderm to a myogenic program (Baylies and Bate, 1996; Castanon et al., 2001). In contrast, overexpression of Twi-Da linked heterodimers in the mesoderm led to the repression of myogenic genes and somatic myogenesis (Castanon et al., 2001).
To further investigate how Da interaction affects Twi activity, we expressed Twi-Da dimers in both the mesoderm and the ectoderm, using the da-GAL4 transgene. Da is ubiquitously expressed in both the mesoderm and the ectoderm until late embryogenesis (Cronmiller and Cummings, 1993; Fig. S1). The somatic or body wall muscles of the wild type embryo are segmentally repeated in a stereotypic manner and lie beneath ectodermally-derived tissues. Each hemisegment gives rise to 30 distinct muscles that can be visualized using antibodies that recognize Myosin heavy chain (Mhc; Bate, 1990; Fig. 1A,AA). Overexpression of Twi-Da linked dimers in the mesoderm using the twi-GAL4 driver resulted in severe patterning defects and muscle losses (Castanon et al., 2001; Fig. 1B,BB). Ectopic expression of Twi or Twi linked homodimers using da-GAL4 caused the activation of myogenic genes and conversion of ectoderm into somatic muscle (Baylies and Bate, 1996; Castanon et al., 2001). When Twi-Da was expressed in this manner, we found that Twi-Da dimers also caused an activation of both myogenic genes and the somatic muscle program in the ectoderm (Fig 1C,CC). Similar to ectopic expression of Twi or Twi linked homodimers (Baylies and Bate, 1996; Castanon et al., 2001), no ectodermally-derived structures such as cuticle were formed. Taken together, these data indicated a conversion of the ectoderm into mesoderm (Fig 1C,CC; data not shown). These results suggested that Twi/Da activity is dependent on tissue context: Twi-Da dimers led to the repression of myogenic transcription in the mesoderm, but activated myogenic genes in the ectoderm.
We next asked what could contribute to the tissue specific activity of Twi/Da heterodimers. Previous studies conducted on the vertebrate homologue of Da, the E proteins, identified the Rep domain, which modulates the transcriptional activity of E proteins in mouse cell lines (Markus et al., 2002). Based on the Rep domain sequence in E proteins, we identified a similar domain in Da and found that this domain is conserved between all sequenced Drosophila species and in both vertebrate and invertebrate species (Fig. 2A, B; data not shown). The Da REP domain consists of a 36 amino acid stretch (residues 494 to 530) located 21 amino acids N-terminal to the bHLH motif and shares 16% homology with the E protein Rep domain. This region is rich in polar and charged amino acids, specifically serine, threonine, and lysine. From studies conducted on E proteins, we hypothesized that the Da REP domain modulates Twi/Da transcriptional activity.
To determine the effect of the Da REP domain on the transcriptional activity of Da and, subsequently, Twi/Da, the Da REP domain was removed from Da (DaΔ) (Fig. 2C). We then assayed the ability of DaΔ to activate transcription from the Dmef2 enhancer in tissue culture. This enhancer contains two E boxes, which are bHLH binding sites, one of which is targeted by Twi (Cripps et al., 1998). Transfection of da cDNA resulted in a 1.8-fold activation of the reporter (Fig. 2D). In contrast, cells that were transfected with twi resulted in a much higher activation level, as seen with a 9-fold difference in Dmef2 reporter activity (Fig. 2D). Cotransfection of both twi and da caused a minimal activation (1.9-fold), as Twi/Da dimers led to the repression of the Dmef2 enhancer (Castanon et al., 2001). Transfection of daΔ also resulted in a low 2.1-fold activation, a level similar to that achieved by Da alone. However, the cotransfection of twi and daΔ resulted in a 6.4-fold activation of the Dmef2 reporter plasmid (Fig. 2D). Because the removal of the REP domain resulted in reporter activation, these data suggested that the Da REP domain normally prevents transcriptional activation. Based on this finding, we hypothesized that the Da REP conferred transcriptionally repressive activity to the combined activity of Twi and Da; more specifically, the Twi/Da heterodimer.
We next generated a twi-daΔ tethered dimer construct to confirm that the activity of the REP domain is specific to Twi/Da dimers. When the twi-daΔ tethered dimer was transfected into S2 cells, it resulted in a 5.6-fold activation of the Dmef2 reporter. By comparison, transfection of twi-da tethered dimer resulted in a minimal 1.5-fold change in reporter activity (Fig. 2D). These data reinforced our hypothesis that the Da REP domain is necessary for Twi/Da repressive activity. Electophoretic mobility shift analysis revealed that Twi-DaΔ forms heterodimers and binds DNA (Fig. S2), indicating that the removal of the Rep domain does not affect the ability of these proteins to dimerize or bind DNA. Taken together, these results demonstrated that, in S2 cells, the failure of Twi/Da dimers to activate the transcription of Dmef2, a myogenic gene, is dependent on the Da REP domain.
We next determined the activity of DaΔ in vivo. Previous data indicated that the effects of Da overexpression in the mesoderm could be enhanced in a twi heterozygous (twiID96/+) background (Castanon et al., 2001). While twiID96/+ embryos have a wild type phenotype (Simpson, 1983), mesodermal overexpression of Da using the twi-GAL4 transgene in twiID96/+ embryos resulted in missing muscles (compare Fig. 3A,AA,D,DD to Fig. 3B,BB,E,EE; Table S1). In contrast, the majority of twiID96/+ embryos overexpressing DaΔ had no missing muscles, but exhibited muscles with altered shapes and aberrant attachments to tendon cells (compare Fig. 3A,AA,D,DD to Fig. 3C,CC,F,FF; Table S1). These in vivo data suggested that Da and DaΔ have distinct effects on myogenesis. Analysis of embryos carrying other GAL4 and UAS-daΔ lines showed similar phenotypes, indicating that these effects are not due to differing levels of protein expression (data not shown).
We next analyzed founder cell (FC) specification to determine whether the phenotypes observed in the final muscle pattern were reflected at earlier stages of myogenesis. FC specification was evaluated using antibodies raised against Krüppel (Kr), which is expressed in a subset of FCs located dorsally, laterally and ventrally throughout the mesoderm in stage 12 embryos (Ruiz-Gomez et al., 1997; Fig. 3G). In twiID96/+ embryos that overexpressed Da, we observed losses of Kr-positive cells (Fig. 3H). By comparison, DaΔ overexpression in twiID96/+ embryos resulted in increased numbers of Kr FCs as compared to wild type (Fig. 3I). These data reinforced our observations that Da and DaΔ have different activities in vivo and supported our hypothesis that DaΔ contributes to the activation of early myogenic genes required for somatic myogenesis.
The increased numbers of FCs generated by overexpressing DaΔ suggested that mesodermal cells are diverted to a somatic muscle fate at the expense of other mesodermal tissues, an effect that has been reported for Twi overexpression (Baylies and Bate, 1996; Castanon et al., 2001). To test this hypothesis, we analyzed the specification of visceral and cardiac mesoderm. We used an antibody that recognizes Fasciclin III (Fas III), which is expressed at stage 11 in a continuous garland of cells that give rise to the gut muscles of the embryo (Baylies and Bate, 1996; Fig. 3J). Embryos that overexpressed DaΔ displayed gaps in Fas III expression (Fig. 3L), indicating losses in visceral mesoderm. These disruptions in the visceral mesoderm do not occur when Da is overexpressed in a similar manner (Fig. 3K). Cardiac tissue, in contrast, was affected similarly by Da and DaΔ. To analyze cardiac development, we used an antibody raised against Zinc Finger Homeodomain 1 (Zfh1), a protein that marks cardioblasts and pericardial cells (Lai et al., 1991). In stage 16 embryos, these cells are dorsally located and are arranged in four rows (Fig. 3M). Overexpression of Da or DaΔ caused a reduction of Zfh1-positive cells in both embryos with a wild type background and those with a twi heterozygous background (Figs. 3N,O; S3). Moreover, this phenotype is reflected in the final stages of embryogenesis, as these embryos exhibit missing cardiac tissue as visualized by Mhc stainings (Fig. 3E,EE,F,FF; Fig. S4). These data suggested that this cardiac phenotype is specifically due to Da activity that is independent of the REP domain. Taken together, the overexpression of DaΔ promotes somatic myogenesis through the allocation of mesodermal cells to FCs, an effect opposite to that of Da overexpression. These data demonstrated that the REP domain is required for the ability of Twi/Da heterodimers to mediate transcriptional repression of myogenic genes, both in cell culture and in vivo. These data also indicated that tissue context (visceral versus cardiac) is an important determinant of REP domain activity.
To determine directly whether Twi/Da dimer mediated repression of myogenesis is specifically dependent on the Da REP domain in vivo, we generated transgenic flies carrying Twi-DaΔ tethered dimers. To assay the transcriptional activity of the tethered Twi-DaΔ in vivo, we overexpressed these dimers in the mesoderm of embryos that carry a β-galactosidase (β-gal) reporter transgene under the control of the same Dmef2 muscle enhancer used in the S2 cell culture assay (Cripps et al., 1998). Basal activation of the reporter transgene was visualized in control embryos using antibodies raised against β-gal (Fig. 4A,A″). Overexpression of Twi-Da dimers resulted in reduced β-gal expression in embryos of the same stage (Fig. 4B,B″). Notably, we observed an increase in reporter gene expression in embryos that overexpressed Twi-DaΔ (Fig. 4C,C″). Additionally, analysis of embryos carrying other GAL4 and UAS-twi-daΔ lines, and embryos that express HA tagged UAS-twi-daHA and UAS-twi-daΔHA linked dimers revealed similar phenotypes and comparable HA levels, respectively (data not shown; Fig. S5). These data suggested that the differing effects of Twi-Da and Twi-DaΔ are not due to differing levels of protein expression. Therefore, consistent with the S2 cell culture transfection assays, these results indicated that Twi-DaΔ dimers activate transcription from the Dmef2 enhancer in vivo, providing a direct target gene that is required for somatic myogenesis.
We next determined the effect of mesodermal overexpression of Twi-DaΔ on somatic myogenesis. We analyzed FC specification using two FC markers, Eve and Kr (Carmena et al., 1998; Ruiz-Gomez et al., 1997). Eve marks a single FC and pericardial cells in the dorsal region of each hemisegment (Fig. 5A). Whereas overexpression of Twi-Da resulted in the reduction of Eve-positive FCs in each cluster (Fig. 5B), Twi-DaΔ overexpression caused increased Eve FCs compared to wild type (Fig. 5C). The same effect was observed for Kr FCs: as compared to wild type embryos (Fig. 5D), Twi-Da overexpression caused reductions in Kr FCs (Fig. 5E), yet Twi-DaΔ overexpression resulted in increased numbers of Kr FCs (Fig. 5F). These data suggested that the effects of Da and DaΔ activity in the mesoderm are more severe when dimerized to Twi. Moreover, these phenotypes were similar to those observed in twiID96/+ embryos that overexpress Da and DaΔ, respectively. We also determined that the reduced FC number in embryos that overexpressed Twi-Da is not due to apoptosis, as we did not observed ectopic expression of cleaved Caspase-3 (data not shown). Additionally, the increased FC number in embryos that overexpressed Twi-DaΔ was not due to ectopic cell divisions, as analyzed by phospho-Histone 3 expression (data not shown) and consistent with the data below.
Analysis of the visceral mesoderm showed that, while Twi-Da overexpression resulted in wild type Fas III expression pattern, Twi-DaΔ overexpression caused disruptions and gaps in Fas III expression (Fig. 5G–I). Mhc staining revealed that the Twi-DaΔ overexpression resulted in conversion of cardiac tissue into somatic muscles (Fig. 6F,FF) and also caused ectopic multinucleated muscles in the ventral region of the embryo (Fig. 5L,LL), a location where muscles normally do not form (Fig. 5J,JJ). Again, these data indicated that in embryos overexpressing Twi-DaΔ, cells that normally develop into visceral or cardiac mesoderm are being diverted to the somatic muscle fate. Additionally, the overexpression of Twi-DaΔ in the ectoderm, like Twi-Da and Twi-Twi, resulted in the conversion of non-mesodermal cells into somatic muscles (data not shown). This result indicated that the REP domain did not affect the activation ability of Twi-Da dimers in the ectoderm.
Because Twi-DaΔ overexpression in the mesoderm resulted in an activation of somatic myogenesis, we next determined whether this effect could also be observed in the final somatic muscle pattern. As previously reported, mesodermal expression of Twi-Da resulted in severe muscle losses, unfused myoblasts, aberrant muscle morphology and abnormal attachment sites. These embryos exhibited an overall decrease of cells allocated to the somatic muscle fate, resulting in fewer and smaller muscles (Castanon et al., 2001). In contrast to Twi-Da activity, mesodermal expression of Twi-Twi caused minor somatic muscle patterning defects (Fig. S5), but resulted in losses of cardiac tissue and visceral muscle defects (Castanon et al., 2001). Surprisingly, embryos that overexpressed Twi-DaΔ revealed a range of muscle defects similar qualitatively to Twi-Da (compare Fig. 6B,BB,E,EE,H,HH to Fig. 6C,CC,F,FF,I,II). This result was unexpected compared to what we observed at earlier stages of somatic myogenesis. These embryos exhibited large numbers of unfused myoblasts, abnormal muscle shapes and incorrect attachment sites (Fig. 6C,CC,F,FF,I,II). Taken together, these data indicated that Twi-DaΔ activated target genes involved with early somatic myogenesis, such as Dmef2, causing the allocation of more cells to the somatic muscle fate. However, continued expression of Twi-DaΔ in the mesoderm disrupted later stages of somatic myogenesis and genes that are involved with muscle fusion and morphogenesis. These results suggested that early Twi-Da transcriptional repression relies on the activity of the Da REP domain, but, during the later stages of myogenesis and muscle differentiation, Twi-DaΔ dimers, like Twi-Da dimers, disrupt these processes.
Because the maintained expression of Twi-DaΔ tethered dimers resulted in disruptions in somatic muscle differentiation, we asked whether this effect is due to an unexplored role of Twi/Da dimers during muscle morphogenesis. Previous work has shown that Da is uniformly expressed throughout the mesoderm at all stages (Cronmiller and Cummings, 1993; Fig. S1). The role of Twi, however, is less clear during these later stages of somatic myogenesis.
We first analyzed late Twi expression in wild type embryos. We observed Twi expression in a subset of FCs and developing myotubes from stage 12 to stage 14 embryos (Fig. 7A-A″; data not shown), which is consistent with a requirement for Twi activity during these stages of myogenesis. To determine whether Twi has a role during muscle morphogenesis, we used a loss-of-function and a gain-of-function approach to analyze Twi activity during this stage. To reduce Twi levels in specific tissues and at specific developmental stages, we generated a UAS-twiRNAi construct. Using anti-Twi antibodies to visualize Twi protein levels, we tested the efficacy of the UAS-twiRNAi transgene in vivo. We observed that mesodermal expression of two copies of the UAS-twiRNAi transgene (UAS-twiRNAi 2X) in twiID96/+ embryos resulted in a reduced level of Twi expression in comparison to control embryos (Fig. 7B,C). This result indicated that knockdown of Twi protein levels can be achieved with the expression of the UAS-twiRNAi construct. We then wanted to determine whether the specific reduction of Twi protein levels in FCs and myotubes would affect muscle differentiation. Hence, we expressed UAS-twiRNAi 2X using the rP298-GAL4 line. Unlike twi-GAL4, which mirrors endogenous Twi expression from late stage 5 to stage 14, the rP298-GAL4 transgene drives expression specifically in progenitor cells, FCs and myotubes, from stage 10 until the end of embryogenesis (Menon and Chia, 2001). The expression pattern of the rP298-GAL4 transgene bypasses earlier myogenic steps and enables the direct analysis of the effects of UAS-twiRNAi during the later stages of muscle development. Expression of UAS-twiRNAi2X using rP298-GAL4 resulted in specific duplications of Lateral Transverse (LT) muscles 1–3 (Fig. 7E,EE). For gain-of-function analysis, we overexpressed two copies of the UAS-twi transgene (UAS-twi 2X) using the rP298-GAL4 transgene. Embryos with increased levels of Twi protein had a similar phenotype, in which LT muscles 1–3 were duplicated (Fig. 7F,FF). Taken together, these data indicated that Twi has a role in muscle identity specification and differentiation, and that these processes are sensitive to Twi levels.
To determine the activity of Twi/Da and Twi/DaΔ during somatic muscle differentiation, we overexpressed these tethered dimers using the rP298-GAL4 line. Expression of either of these tethered proteins resulted in missing muscles, incorrect muscle attachment, aberrant muscle morphology and unfused myoblasts (Fig. 8A–CC), indicating that the Da REP domain is not required for these processes. These data further supported our hypothesis that Twi/Da transcriptional repression is mediated through the Da REP domain during early myogenic processes, but is not dependent on the REP domain during muscle differentiation.
Because we observed missing muscles, we verified that these muscle losses were not due to disruptions in FC specification (Fig. 8D–F). We observed normal Kr expression in embryos that overexpressed Twi-Da or Twi-DaΔ with this GAL4 line, which suggested that the losses of these muscles are not due to missing FCs, but due to problems encountered during muscle fusion and morphogenesis, processes that are essential for muscle differentiation. We also determined that these muscle losses were not due to apoptotic FCs, as we did not detect ectopic cleaved Caspase-3 expression in these embryos (data not shown).
We conducted a detailed analysis of these muscle loss phenotypes by determining the presence or absence of specific body wall muscles in abdominal segments 2 through 4 (A2–A4) in stage 16 embryos. This analysis revealed that certain subsets of muscles were more sensitive to either Twi-Da or Twi-DaΔ activity (Fig. 8G, Table S3). For example, the ventral oblique 6 (VO6) and ventral acute 3 (VA3) muscles were missing from the final muscle pattern in over 80% and 50% of hemisegments analyzed in rP298-GAL4 > UAS-twi-da and rP289-GAL4 > UAS-twi-daΔ embryos, respectively. By comparison, the dorsal transverse 1 (DT1), dorsal oblique 1 (DO2), and VO4 muscles were absent from these hemisegments in less than 5% of hemisegments analyzed. Although Twi-Da and Twi-DaΔ expression caused phenotypes that differ in severity, there is correlation in the losses of certain muscles. For example, the losses of VO6, VA3 and lateral oblique 1 (LO1) muscles are closely correlated between embryos expressing Twi-Da or Twi-DaΔ, but not with ventral longitudinal 4 (VL4) and DO3 muscles (Fig. 8G, Table S3). Additionally, muscle losses were observed dorsally, laterally, and ventrally, which indicated that muscle losses are not associated with a specific dorsal-ventral location, and therefore are not due to the disruption of positionally-specific factors, such as Dpp (Lee and Frasch, 2005) or Pox meso (Bopp et al., 1989; Duan et al., 2007).
We also overexpressed Da, DaΔ, and Twi-Twi tethered homodimers using the rP298-GAL4 line to ensure that the phenotypes we observed in embryos overexpressing Twi-Da and Twi-DaΔ tethered dimers were specific to these tethered dimers (Fig. S6). We observed minor defects in the final muscle pattern of embryos that overexpressed Da, and embryos that overexpressed DaΔ or Twi-Twi appeared wild type. These minor defects were different from the phenotypes observed in embryos overexpressing Twi-Da and Twi-DaΔ (Fig. S6), which indicated that Twi-Da and Twi-DaΔ dimers have distinct activities when overexpressed in FCs.
Taken together, these results demonstrated that the repressive activity of Twi/Da relies on the Da REP domain during the early stages of somatic myogenesis, specifically during mesodermal subdivision and FC specification. However, during the later stages of muscle differentiation, the Da REP domain is not required for Twi/Da repressive activity. Therefore, the repressive activity of Twi/Da relies on different protein domains through the course of somatic myogenesis. In conclusion, the identification of the Da REP has enabled us to address the tissue specific activity of Twi/Da and uncover the sensitivity of Twi/Da activity to developmental timing.
Here, we explore the regulation of Twi activity through mesoderm development and somatic myogenesis in Drosophila. We focus on how Twi, a bHLH transcriptional regulator, is modulated by its dimer partner, Da. Our examination of Twi/Da dimers revealed that the activity of these dimers is acutely sensitive to their tissue environment: both between germ layers (the ectoderm versus the mesoderm), and within cell lineages (early mesoderm versus somatic muscle). This sensitivity is determined, in part, by the activity of the Da REP domain, which is critical for Twi/Da activity during mesodermal subdivision and FC specification, but is not required for the later activity of Twi/Da during muscle differentiation (Fig. 9). This work provides insight to the mechanism of Twi/Da activity and calls attention to the effect of tissue context and developmental timing on bHLH protein regulation.
One of the most striking aspects of this study is the role of the Da REP domain in switching Twi/Da behaviour between a repressor and an activator function. This “switchable” behaviour of Twi/Da activity was initially observed by its ability to inhibit myogenesis in the mesoderm, but activate myogenesis in the ectoderm. Notably, the deletion of the REP domain from Da has little effect on Da activity in the absence of Twi, as demonstrated by cell culture transcriptional assays. However, the activity of Twi-DaΔ tethered dimers has a distinct effect on the mesoderm. Overexpression of these dimers had the greatest effect on somatic myogenesis during the process of mesodermal subdivision. The detection of increased numbers of FCs, which appear to be specified normally, indicated an increased number of mesodermal cells being allocated to a somatic muscle fate at the expense of cardiac and visceral mesoderm.
An outstanding question is how the Da REP domain functions to modulate Twi/Da activity. Since Twi/Da dimers bind DNA and therefore may actively regulate the transcriptional state of a target gene, we initially postulated that the REP domain must directly interact with transcriptional corepressors or factors that were expressed solely in the mesoderm and therefore were required for the repressive activity of Twi/Da in that tissue context. We have conducted exhaustive studies to identify these factors but have been unable to identify a protein that satisfies all necessary criteria.
Deletion analysis of the E protein Rep domain suggested that this domain is required for the repression of the E protein activation domains, AD1 and AD2. Like the Da REP domain, the E protein Rep domain has specific activities depending on its dimer partner and tissue context (Markus et al., 2002). Informed by this work, we interpret our data to suggest that the Da REP domain is a cis-acting repressor, which functions to repress both Da AD1 and AD2 when Da is dimerized to Twi and bound to myogenic enhancers. Moreover, the effect of the Da REP domain is not restricted to the E protein/Da protein family. Our work suggests that the Da REP domain also represses Twi’s activation domains, Twi-AD1 and Twi-AD2 (Chung, 1996; Gonzales, 2005), in Twi/Da dimers. We propose that the Da REP domain acts to mask the activation domains in both Twi and Da. Therefore, the net effect of the Da REP domain results in the recruitment of corepressors to myogenic enhancers by Twi/Da dimers. Alternatively, Twi/Da dimers may not actively repress target myogenic genes: instead, these dimers could compete for myogenic E boxes (Castanon et al., 2001) or transcriptional cofactors and machinery. In this model of passive repression, the Da REP domain could function to stabilize interactions with Twi or other factors that are required to properly mediate repression of myogenic target genes. These aspects of Da REP domain repression are currently being evaluated.
To date, various transcriptional regulators have been shown to have different activities and target genes in different tissues and be modulated by dimerization partners. Recently, ChIP-on-chip analyses have identified almost 500 direct Twi targets throughout mesodermal development (Sandmann et al., 2007; Zeitlinger et al., 2007). Our study, however, is one of the first that focuses on how Twi activity is dynamically modulated through multiple developmental stages of a specific cell lineage, and how this regulation affects expression of Twi target genes.
One gene that is regulated by Twi dimers throughout somatic myogenesis is Dmef2. Dmef2 protein is expressed throughout and necessary for all stages of myogenesis (Bour et al., 1995; Cripps et al., 1998; Lilly et al., 1995; Taylor et al., 1995). Sandmann et al. 2006 find that Dmef2 coordinates multiple processes necessary for proper somatic myogenesis. Moreover, these authors have suggested that Dmef2 is required in combination with Twi to regulate the expression of a subset of Twi target genes in a feed-forward mechanism (Sandmann et al., 2007). Our data support these arguments, as we observe mesodermal phenotypes in Twi/DaΔ (activation) or Twi/Da (repression) overexpressing embryos that mirror those of embryos overexpressing Dmef2 or in Dmef2 mutant embryos, respectively. For example, we observe increased Dmef2 reporter gene expression and increased numbers of FCs in embryos that overexpress Twi/DaΔ panmesodermally. Consistent with these observations, Dmef2 has been shown to regulate components of the Ras/MAPK and Notch pathways, which are both required for the proper specification of FCs, and the expression of a subset of FC identity genes (Sandmann et al., 2006). Dmef2 has also been shown to regulate a subset of genes that are required for myoblast fusion and muscle attachment, processes required for proper muscle morphogenesis (Sandmann et al., 2006). We find that Twi/Da and Twi/DaΔ dimers disrupt myoblast fusion and muscle differentiation, which is likely due to these dimers repressing Dmef2 expression. In agreement with this observation, our muscle analysis revealed that embryos overexpressing Twi-Da and Twi-DaΔ dimers have muscle phenotypes that are similar to those observed in Dmef2424 hypomorph embryos (Ranganayakulu et al., 1995) and Dmef222.21 null embryos that have been partially rescued by UAS-Dmef2 transgenes (Gunthorpe et al., 1999) (Fig. 8G, green box, Fig. S7). Taken together, these results supported our conclusions of the pivotal regulation of Dmef2 activity by Twi dimers throughout myogenesis (Fig 9).
Another notable question is how the Da REP domain is required for Twi/Da mediated transcriptional repression during mesodermal subdivision, but not during muscle morphogenesis. One possibility is that during somatic muscle differentiation, the repressive activity of Twi/Da relies on a different protein domain. Another possibility includes the changes in Twi/Da target genes through the course of somatic myogenesis. Studies conducted on chromatin remodeling have emphasized the specificity involved with the transcriptional regulation of a single gene. Therefore, it is likely that the regulation of multiple sets of genes through time would rely on the modular nature of transcriptional regulators. The Da REP domain may be required for the repression of a subset of Twi/Da target genes, whereas other target genes are unresponsive to this domain’s repressive activity.
In summary, our results suggest that the regulation of Dmef2 by Twi/Da throughout myogenesis and the subsequent feed-forward mechanism by which Dmef2 and Twi regulate myogenic genes is critical for the coordination of the various disparate processes—mesodermal subdivision, FC specification, and muscle differentiation—necessary for somatic myogenesis (Fig 9).
Twi proteins are conserved across species [mouse (Gitelman, 1997), chicken (Tavares et al., 2001), C. elegans (Corsi et al., 2000), and jellyfish (Spring et al., 2000)] and have been shown to dimerize with Da homologs (Connerney et al., 2006; Corsi et al., 2002; Spicer et al., 1996), suggesting that REP domain regulation of Twi activity is conserved. Similarly to flies, Mouse Twi1 (MTwi1) heterodimerizes with E proteins to compete with MyoD/E proteins for binding sites on myogenic enhancers (Spicer et al., 1996). In this manner, MTwi1/E protein heterodimers act like Twi/Da dimers to repress myogenesis. In other tissues, however, MTwi1/E protein heterodimers have been identified as an activator of targets, such as thrombospondin-1 during cranial suture formation (Connerney et al., 2006). Therefore, like Twi/Da, MTwi1/E protein heterodimers are sensitive to tissue contexts. Of particular interest would be the examination of the E protein Rep domain in vivo. The function of this domain has been studied in mammalian cell culture (Markus et al., 2002), but not yet investigated in developmental processes. Moreover, the function of the E protein Rep domain has not been addressed in MTwi1/E protein dimers.
Notably, Twi proteins have also been implicated in a variety of tumourigenic processes, such as the inhibition of apoptosis (Puisieux et al., 2006) and the coordination of metastasis (Yang et al., 2004; Yang et al., 2006). Mouse models and correlative data from human tumour samples suggest that MTwi1 and human Twi1 (HTwi1), respectively, direct epithelial-to-mesenchymal transitions (EMT) during breast cancer metastasis (Yang et al., 2004). The involvement of Twi1 in the complex process of cancer has many similarities to the developmental processes that Twi directs in the fly mesoderm, which include cell proliferation and cell migration, processes that have been recently revealed to be directly regulated by Twi (Sandmann et al., 2007). The role of the Da REP domain in directing Twi/Da transcriptional repression, and the tissue specificity of this domain’s activity has illuminated various aspects of Twi regulation. We anticipate that these findings will shed light on mammalian Twi1 activity and the Twi family of proteins in development and disease.
(A–C″) Confocal micrographs of stage 9 (A–A″), stage 10 (B–B″), and stage 15 (C–C″) twi-CD2 embryos. All embryos have been stained for CD2 (A,B,C) to show twi expressing cells, and Da (A′,B′,C′) to show the broad expression pattern of Da in the mesoderm. Merged images (A″,B″,C″) show that Da is expressed in twi positive cells mesodermal cells in stage 9, 10 and 15 (Cabrera and Alonso, 1991; Campuzano et al., 1985; Caudy et al., 1988; Cronmiller and Cummings, 1993). These data suggest that Twi and Da are coexpressed during mesodermal development and that Twi/Da dimers can be formed and have an effect on mesoderm and muscle development during these stages. Scale bar represents 20 μm.
Mobility shift assays with rho E box (CATATG; (Ip et al., 1992; Leptin, 1991) were performed with Twi (lanes 2, 5, 14, 15; Castanon et al., 2001), Twi-Twi linked homodimers (lane 3; Castanon et al., 2001), Twi and Da (lane 6; Castanon et al., 2001), Da (lane 7; Castanon et al., 2001), Twi-Da linked dimers (lane 9, 10; Castanon et al., 2001), and Twi-DaΔ linked dimers (lane 11, 12). Asterisks mark the Twi homodimer (lane 3) and the Twi/Da heterodimer (lane 6). Arrow points to “dimers of dimers” present in Twi-Twi linked homodimer shifts, but absent in Twi-Da and Twi-DaΔ linked dimers. The unprogrammed lysate is included for each gel (lane 1, 4, 8). Probe is in excess in all lanes shown. See Castanon et al., 2001 for additional controls.
(A,D) Wildtype embryos, (B,E) twi-GAL4 > UAS-da, (C,F) twi-GAL4 > UAS-daΔare shown. Stage 11 embryos stained with anti-Eve (A–C) and stage 16 embryos stained with anti-Zfh1 (D–F) are shown. Eve clusters are both expanded (arrows) and reduced in size (arrowheads) in embryos that express UAS-da or UAS-daΔ(B,C). Zfh1 expression reveals losses of pericardial cells (arrowheads, E,F). In a wild type background, overexpression of Da and DaΔ proteins can result in various dimerizations with endogenous HLH proteins. However, in a sensitized twiID96/+ background, overexpression of Da and DaΔ proteins result in phenotypes that are more similar to those observed in embryos that overexpress Twi-Da and Twi-DaΔ linked dimers, respectively (compare Fig. 3 and Fig. 5). All scale bars represent 20 μm.
Wild type (A,AA,D,DD,G,GG), twi-GAL4 > UAS-da (B,BB,E,EE,H,HH), and twi-GAL4 > UAS-daΔ (C,CC,F,FF,I,II) embryos are at stage 16 and have been stained with anti-Mhc. Lateral views (A–C,AA–CC), dorsal views (D–F,DD–FF), and ventral views (G–I,GG–II) are shown. Expression of UAS-da or UAS-daΔ causes muscle losses (BB,CC,II; asterisks), defective muscle morphology (BB,CC,HH; arrows), and loss of cardiac tissue (EE,FF; arrowheads). These data suggested that, in a wild type background cardiac tissue is particularly sensitive to Da activity. All scale bars represent 20 μm.
Overexpression of UAS-twi-twi causes muscle duplications and losses. (A–B) UAS-twi-da and UAS-twi-daΔ constructs are expressed at comparable levels. Confocal micrographs of stage 11 twi-GAL4 > UAS-twi-daHA and twiGAL4 > UAS-twi-daΔHA embryos stained with anti-HA (green) are shown. Scale bar represents 10 μm. Expression of the two linked dimers as visualized by HA staining reveals that the linked dimers are expressed at similar levels. (C–DD) Stage 16 wild type and twi-GAL4 > UAS-twi-twi embryos stained for Mhc are shown. Scale bars represent 20 μm. A missing muscle is indicated by an asterisk and muscle duplications are indicated by arrowheads.
Gunthorpe et al. have shown that too much Dmef2 activity disrupts muscle differentiation (Gunthorpe et al., 1999). These findings suggested that the phenotype caused by overexpression of Twi-DaΔ dimers using the rP298-GAL4 driver may be due to the inappropriate activation of Dmef2 expression. However, we show that the similar overexpression of Twi (Fig. 7) or Twi-Twi homodimers (Fig. S3), which are potent activators of Dmef2 expression, did not cause defects in muscle differentiation that are similar to those in embryos that overexpress Twi-DaΔ.
Lateral views of stage 16 embryos that have been stained for anti-Mhc. Wild type (A,AA), rP298-GAL4 > UAS-da (B,BB), rP298-GAL4 > UAS-daΔ (C,CC), and rP298-GAL4 > UAS-twi-twi 2X (D,DD), embryos are shown. All embryos appear wild type, except for specific muscle duplications observed in rP298-GAL4 > UAS-da and rP298-GAL4 > UAS-twi-twi embryos (BB, DD; arrows). All scale bars represent 20 μm.
(A,B,C) Schematic of the 30 somatic muscles in a hemisegment. (A–C) Muscles shown in blue are present in more than 50% of hemisegments, those depicted in orange are missing in more than 50% of hemisegments. (A) Representation of Dmef2424 hypomorphic mutant embryos (Ranganayakulu et al., 1995). (B) Representation of missing muscles in rP298-GAL4 > UAS-twi-da embryos. Muscles depicted in pink are missing in over 50% of hemisegments in rP298-GAL4 > UAS-twi-da embryos and are missing in over 20% of hemisegments in rP298-GAL4 > UAS-twi-daΔ embryos. (C) Depiction of muscles that are missing in Dmef222.21 null embryos that have been partially rescued with a weak UAS-Dmef2 transgene (Gunthorpe et al., 1999). Different somatic muscles require different levels of Dmef2; for example, the DT1 and ventral acute 1 (VA1) muscles require low levels, but the LO1 muscle requires high levels of Dmef2 to form properly, as observed in our work, Dmef2 hypomorphs (Ranganayakulu et al., 1995), and Dmef2 rescue experiments (Gunthorpe et al., 1999).
We thank L. Selleri, Y. Nibu, C. Rushlow, and the members of the Baylies lab for discussions and critical reading of the manuscript and D. Soffar for technical support. We also thank S. Abmayr, D. Kiehart, Z. Lai, J. Reinitz, S. Roth, C. Cronmiller, and the Developmental Hybridoma Bank for antibodies. This work was supported by the Sloan-Kettering Institute and a NIH grant (GM 586989) to M.K.B.
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.