The tbx/bHLH transcription factor mga regulates gata4 and organogenesis



The mga gene encodes a unique transcription factor containing both TBOX and bHLHzip DNA-binding domains. Here we describe the structure, expression pattern, and loss-of-function phenotype for zebrafish mga. The mga gene is conserved with mammalian homologs for both DNA-binding domains. It is expressed maternally, and subsequently in the developing brain, heart, and gut, and its depletion causes morphogenetic defects in each of these organ systems. The heart and gut phenotypes are similar to those described previously for loss of gata4, and the mga morphant shows increased levels of gata4 transcripts in lateral mesoderm. Knockdown of gata4 rescues the early heart-looping defect (but not the gut defect), indicating that mga restricts the normal levels of Gata4 required for heart tube looping, while both genes are important for gut development. Transcript profiling experiments show that mga functions early to influence key regulators of mesendoderm, including tbx6, cas, and sox17. Developmental Dynamics 239:535–547, 2010. © 2009 Wiley-Liss, Inc.


During embryogenesis, tight spatial and temporal control of gene expression is achieved through functional networks of interacting transcription factor families. For example, Drosophila genetics established the identity of transcription factors required to specify cardiac progenitors, form a primordial heart tube, and complete the formation of a functional dorsal vessel, analogous to the vertebrate heart (Tao and Schulz, 2007). Homologous genes have been identified in vertebrates; for example, the GATA and TBOX families of transcription factors that are known to carry out a variety of functions during cardiogenesis (Olson, 2006). Thus, gata4, gata5, gata6, tbx5, and tbx20 are implicated from animal models in transcriptional programs controlling cardiac cell specification, looping of the heart tube, chamber formation, differentiation, and valvogenesis (Brown et al., 2005; Holtzinger and Evans, 2005, 2007; Plageman and Yutzey, 2004). These genes are required for normal cardiogenesis in humans, since specific mutations in TBX5, TBX20, or GATA4, for example, are associated with defined human syndromes and congenital heart defects (Basson et al., 1997; Garg et al., 2003; Kirk et al., 2007; Li et al., 1997).

Progress in understanding the genetic hierarchy amongst these genes has been a challenge in vertebrate systems, because the primordial genes underwent expansion into families comprised of highly related sister genes, encoding proteins with overlapping expression patterns, at least partial functional compensation, and similar DNA and protein-binding activities. Three genes, gata4, gata5, and gata6 are implicated in regulating heart and gut development (Peterkin et al., 2005) while the number of TBOX genes potentially regulating various aspects of cardiogenesis is even more extensive (Naiche et al., 2005). Loss-of-function analyses in zebrafish showed that gata4 is not required for cardiomyocyte specification or differentiation. Rather, embryos depleted of Gata4 protein have a heart tube looping defect at 48 hr post-fertilization (hpf) and develop a severe “heartstring” phenotype by 5 days post-fertilization (dpf), in addition to defects in endoderm-derived gut tissues (Holtzinger and Evans, 2005). Similar phenotypes, resulting in relatively late morphogenetic abnormalities, rather than earlier specification or differentiation defects, are described by the loss of TBOX factors, including Tbx5, or Tbx20 (Garrity et al., 2002; Szeto et al., 2002). It is presumed that the proteins are essential for earlier developmental functions, but that this is masked in loss-of-function studies by compensation from related genes, and this has been confirmed at least for GATA factors in both zebrafish (Holtzinger and Evans, 2007; Peterkin et al., 2007) and mouse (Zhao et al., 2008).

Another set of transcription factors, encoded by the MYC/MAX family, regulate genetic programs involved in proliferation and cell cycle, as well as cell survival and apoptosis (Meyer and Penn, 2008). This is also disease-relevant since c-MYC, for example, is over-expressed in a wide variety of human cancers (Adhikary and Eilers, 2005). MYC and MAX dimerize through their basic region/helix loop helix/leucine zipper (bHLHzip) domains and activate transcription of downstream genes by binding to consensus E-box elements on the DNA (Blackwood and Eisenman, 1991), whereas MAD or MXI dimerization with MAX leads to repression of downstream target genes either by loss of the MYC-dependent partner, or by binding E-box elements (Ayer and Eisenman, 1993).

MGA (MAX-gene-associated) is a transcription factor that belongs to both the TBOX and MYC families, encoding both a Tbox and a Myc-like bHLHzip domain. The murine MGA protein was initially identified in 2-hybrid experiments as a binding partner of MAX, and the large Mga gene was found to encode a protein of 3,006 amino acids (Hurlin et al., 1999). MGA binds E-box sequences in a MAX-dependent manner and T-box sequences in a MAX-independent manner; reporter assays confirmed the ability of MGA to regulate transcription from promoters containing either T-box or E-box binding sites. This study also showed that the levels of MAX determine whether MGA functions as an activator or repressor. In addition, forced expression of MGA inhibits foci formation by fibroblasts transfected with Myc and Ras, suggesting that MGA could act as a tumor suppressor (Hurlin et al., 1999). A loss-of-function phenotype for Mga has not been described.

Because of its distinctive structural features, MGA is a candidate to integrate regulation of cell survival, proliferation, and differentiation programs, based on its ability to function through TBOX- and MYC-directed transcriptional pathways. GATA factors may also function in the context of MYC regulatory networks (Rylski et al., 2003). Furthermore, we identified previously a cardiac enhancer for the gata4 gene that contains binding sites for proteins from both TBOX and MYC gene families (Heicklen-Klein and Evans, 2004). Therefore, we investigated if a functional relationship exists among MGA, TBOX, or GATA factors using the zebrafish model, taking advantage of loss-of-function assays, and the ability to carry out epistatic analyses. Here we present the first characterization of a non-mammalian mga gene, and describe its loss-of-function phenotype. We show that zebrafish mga is required for heart- and gut-derived organ development and that it functions to regulate a subset of early mesendoderm markers, including tbx6, and to define the appropriate levels of Gata4 needed for primitive heart tube looping.


Zebrafish mga Is Conserved for Both the Tbox and bHLHZip Domains

As part of a screen to evaluate the genetic interaction of Tbox proteins with the gata4 gene during zebrafish cardiogenesis, we sought to identify the zebrafish homolog of mga. The zebrafish mga locus was not annotated on the UCSC zebrafish genome browser. Therefore, we used BLAST analysis comparing mouse or human MGA protein sequence to map homology to chromosome 17, including a “predicted” open-reading frame (ORF) with several associated ESTs. A preliminary molecular analysis suggested that gaps or inversions were present in the genomic database (not shown). Therefore, we took advantage of the available EST sequences and performed 5′ RACE using cDNA from early embryos. In this manner, we identified a putative translational start site (ATG) for the zebrafish gene. Then, by standard RT-PCR “walking,” we generated several overlapping fragments to define an entire predicted coding sequence, including an in-frame putative stop codon. Finally, we confirmed this analysis using long-range PCR and primers that flank the entire ORF predicted by our walking data, to isolate and sequence a mga cDNA encoding 2,735 amino acids. The analysis confirmed errors in the genomic database including one false exon/intron prediction.

The overall domain structure of mga is schematized in Figure 1A. The predicted zebrafish protein has several structural characteristics conserved with the mouse homolog. Both the zebrafish and mouse proteins are large (298 and 333 KDa, predicted molecular weight, respectively), and contain two DNA-binding domains, an N-terminal Tbox domain with 68% identity (Fig. 1B), and a more C-terminal bHLHZip domain with 34% identity (Fig. 1C). Both genes are encoded by 23 exons. Although we cannot currently rule out alternative splicing, RT-PCR analysis consistently generated a single large transcript (Fig. 1D), and we failed at any time to identify by RT-PCR alternative cDNA isoforms.

Figure 1.

Zebrafish mga encodes a large protein with conserved T-box and bHLHzip domains. A: Schematic representation of the zebrafish mga cDNA showing the relative position of the two domains. B, C: Comparison between putative protein sequences for zebrafish (zMga) and mouse (mMga) domains, with conserved (identical) amino acids boxed. B: The T-box domain shows 68% identity. C: The bHLH domain is 34% identical. D: RT-PCR was used to confirm the size of the cDNA, including an ORF comprising 8.2 kb. The full cDNA sequence of zebrafish mga was submitted to Genbank (Accession number GU122854).

Zebrafish mga Is Expressed Maternally and Throughout Development

We analyzed the expression pattern of mga during zebrafish development by whole-mount in-situ hybridization using mga-specific antisense RNA probes. At the 1-cell zygote stage, mga transcripts are detected readily indicating that the gene is maternally expressed (Fig. 2A). During gastrulation (75% epiboly), mga is expressed widely at relatively high levels, although restricted to the embryo proper; transcripts are not apparent in the yolk syncytial layer, surrounding the yolk (Fig. 2B). During somitogenesis, the gene is still widely expressed, although transcript levels appear decreased, particularly in the mid-caudal regions (10 somites; Fig. 2C, D). By 24 hpf, transcript levels are much more restricted to the anterior/brain region, and are also evident in the endoderm/primordial gut tube (Fig. 2F–H). By 48 hpf, relatively high transcript levels are found throughout the anterior part of the embryo including the head, brain, heart, gut endoderm, and fin buds (Fig. 2I, J). Several different antisense mga probes were tested and each gave equivalent results (not shown), while control sense strand probes showed no signal (Fig. 2E, K). Expression of mga was also analyzed in a variety of adult organs by semi-quantitative RT-PCR and transcripts were readily detected at similar levels in many organs including testis, heart, intestine, brain, and kidney (data not shown). In summary, zebrafish mga is expressed maternally and widely during early embryogenesis, more restricted to anterior regions during somitogenesis, and also expressed in adult organs. Unfortunately, antibodies generated against mouse MGA (kindly provided by Dr. P. Hurlin) did not cross-react to the zebrafish protein.

Figure 2.

Zebrafish mga transcripts are expressed maternally and throughout embryogenesis, with eventual restriction to anterior regions. Shown are representative embryos following in-situ hybridization using probes specific to mga, or as indicated using a control sense strand probe. A: Maternal expression at the 1-cell stage. B: During late gastrulation, transcripts are widely expressed at relatively high levels, except in the yolk syncytial layer, shown here at 75% epiboly. C–E: There is fairly ubiquitous expression during somitogenesis stages (10 somite stage) shown in a (C) lateral view or a (D) dorsal view. E: A control sense probe at the 10 somite stage. F–H: mga transcripts become restricted by 24 hpf to the anterior region including head/brain, heart, and gut endoderm (arrow in F) as shown in a lateral view (F) or dorsal views (G, H) where H shows a representative flat-mounted embryo. I–K: mga transcripts are enriched at 48 hpf in the head/brain, gut tube (arrow in I), and fin buds as shown either in (I) a lateral view or (J) a dorsal view. K: A representative embryo hybridized with the control sense probe at 48 hpf. In all cases, anterior is to the left. H, head; T, tail; HT, heart; FR, forebrain; MB, midbrain; HB, hindbrain; EY, eye; FB, fin bud; GE, gut endoderm; GT, gut tube.

Mga Is Required for Normal Development of the Brain, Heart, and Gut

In order to analyze the function of mga during embryogenesis, we injected mga-specific morpholinos into fertilized eggs to block production of functional Mga protein. To ensure that the phenotypes are specific to loss of Mga, several distinct morpholinos were used, including those predicted to block either translation or splicing of the endogenous mRNA. MO1 targets specifically the 5′UTR and initiation ATG, efficiently blocking the in vitro translation of mga sequenced fused to a reporter (data not shown). MO2 is also a putative translation blocker that targets a distinct sequence downstream of the initiation ATG. MO3 is a splicing blocker that targets the intron1-exon2 boundary. In addition, we used at the same concentration a mis-match morpholino (relative to MO1) that served as a control.

Injection of 5–10 ng MO1 (the translational blocker targeting the ATG start site) results in a highly reproducible phenotype in 100% of the embryos, which is first clearly evident by 24 hpf. Although the overall body plan is grossly normal compared to control embryos, there is a significant disruption of normal brain morphology. Thus, compared to uninjected or control (mis-match) injected embryos, there is a failure in the definition for the boundaries of the forebrain, midbrain, and hindbrain in morphant embryos (Fig. 3A, B). The eye is present but reduced in size. By 48 hpf, the brain defect is severe, particularly in the hindbrain region, which appears largely missing or edemic (Fig. 3C–I). At this point, there are two additional obvious developmental defects: a failure in normal heart tube looping with an associated edema, and a regression of the yolk stalk extension, implying a potential endoderm/gut defect (Fig. 3C–E). At 48 hpf, the morphant embryos are approximately 25% smaller, lack pigmentation, and have reduced levels of circulating erythrocytes, compared to controls. All three morpholinos (MO1, MO2, and MO3) generate very similar phenotypes (Fig. 3E–G), although they vary in level of penetrance (see Fig. 3 legend). MO2 may be less effective because it targets a sequence considerably downstream of the ATG (see Experimental Procedures section; morpholinos are thought to be less effective distal to the ATG). MO3 was used only at relatively low levels, above which it was not tolerated. However, it may also be less active since it would not block activity of any maternal Mga protein. We used RT-PCR and primers that flank the target site to test if splicing was disrupted. In some experiments, a larger cDNA consistent with intron inclusion was seen (not shown), although in most experiments there was a complete loss of detectable transcript, suggesting that MO3 destabilizes mga RNA (Fig. 3J). By 5 days post-fertilization (dpf) morphant embryos develop a large cardiac edema in the context of a thin linear “heart-string,” very similar to the phenotype of the tbx5/heartstrings mutant (Garrity et al., 2002) in which the heart tube becomes thin, collapsed, and distended by 72 hpf. Morphant embryos retain a large yolk mass that fails to be absorbed, likely due to a failure in normal gut development. Most morphant embryos die by 7 dpf.

Figure 3.

The mga morphants show defects in organogenesis of the brain, heart, and gut. Shown are representative control (wildtype: WT, or mismatch) or morphant (MO1, MO2, MO3) embryos as indicated. A,B: At 24 hpf, compared to control embryos, the morphant fails to partition normally the forebrain, midbrain, and hindbrain structures, and eye development is significantly delayed. C–I: At 48 hpf, mga morphants show a non-looping heart within a cardiac edema (asterisks), an edemic hindbrain cavity (arrowheads in E, F, G, and I), and a regression of the yolk stalk extension (large arrowheads in E–G), indicative of an endoderm defect. The morphant phenotype (brain, heart, and yolk stalk defects) shown here by representative embryos was documented for MO1 (100%, 320/320 embryos), MO2 (75%, 170/228 embryos), and MO3 (29%, 46/161 embryos). Views are lateral, dorsal to the top, and anterior to the right. HB, hindbrain; MHB, midbrain-hindbrain boundary; MB, midbrain; FB, forebrain; E, eye; HT, heart; YSE, yolk stalk extension. J: RT-PCR analysis demonstrates that the MO3 morpholino disrupts the production of normal transcript as shown by the absence of mga cDNA (left). The cDNA was prepared from control wildtype (WT) or morphant (MO3) embryos. Right panel shows equal amplification of β-actin for both WT and MO3 cDNA samples.

In morphant embryos, the primordial heart tube appears to form normally, the heart tube beats, and chambers form. This is consistent with normal specification and differentiation, based on normal expression of cardiomyocyte markers according to in situ hybridization experiments, including for nkx2.5, tbx5, and tbx20 (data not shown). To evaluate the cardiac phenotype further, we injected MO1 into fertilized eggs derived from cmlc2:gfp transgenic reporter fish, which marks the developing myocardium with GFP. As shown in Figure 4, the failure in tube looping by 48 hpf in the morphant is obvious (Fig. 4A, B), and the linear heart tube remains GFP+. The primary defect is, therefore, during heart tube looping, and this results in a severe heartstring phenotype at 5 dpf (Fig. 4C, D).

Figure 4.

Mga is required for normal heart tube looping. Shown are representative embryos derived from cmlc2:gfp transgenic reporter fish. A–D: Comparison of uninjected control (A, C) and embryos injected with MO1 (B, D) at 48 hpf. E, F: The same comparison, except at 5 dpf, showing the extreme heartstring phenotype and cardiac edema. A, B and E, F are lateral views, dorsal to the top, and anterior to the right. C, D are ventral views, anterior to the right.

Similarly, we evaluated gut-derived organ growth in morphants using the ef1a:gfp reporter strain (Field et al., 2003), which labels the gut, liver, and pancreas with GFP. By 48 hpf, these organs are readily apparent in both wildtype and morphant embryos as a looping intestinal bulb, with the liver and pancreas buds emerging from the left and right sides, respectively (Fig. 5A, B). By 3 dpf, the buds have undergone significant growth in wild type or control embryos (Fig. 5C), but not in the morphant embryos (Fig. 5D). Strong GFP+ organs are developed by 5 dpf in control embryos, while in morphant embryos distinct liver and pancreatic tissues are not evident (Fig. 5E, F). We also evaluated the expression of specific markers for the gut and liver using in situ hybridization, since this can provide a more sensitive assay for the presence or absence of tissue compared to GFP. As shown in Figure 6A–D, the liver marker transferrin is visible in a small patch of tissue (compared to wildtype controls) but even this low level is extinguished by 4 dpf in the morphant, suggesting that the initial liver bud regresses or fails to thrive. Similarly, expression of ifabp (an intestinal/gut marker) is missing in 4-dpf morphants.

Figure 5.

Gut organogenesis fails in the mga morphant embryos. Shown are representative control or morphant (MO1) embryos derived from ef1a:gfp transgenic fish used to image endoderm/gut morphogenesis. A, B: The comparison at 48 hpf shows relatively normal budding of the pancreas and liver suggesting that early organ budding is normal. C, D: By 3 dpf, the mga morphants show a significant decrease in the growth of liver and intestine tissue. E, F: By 5 dfp, the mga morphant embryos completely lack GFP-labeled endoderm-derived organs including the pancreas (dashed line in E), liver (dashed line in F), intestine, and swim bladder. Views are: (A, B) dorsal, anterior to the left, (C, D) lateral, dorsal to the top and anterior to the left, (E) lateral, dorsal to the top and anterior to the right, (F) lateral, dorsal to the top and anterior to the left. L, liver; P, pancreas; IB, intestinal bulb; SB, swim bladder.

Figure 6.

Gut organogenesis fails as shown by organ-specific markers. Shown are representative embryos labeled by in situ hybridization for the liver-specific marker transferrin (A–D) at 3 dpf (A, B) or 4 dpf (C, D) comparing wildtype (WT) control embryos with morphant (MO1) embryos (B, D). At 3 dpf, there is a small patch of transferrin-positive cells (arrow in B) but this is eliminated or regressed by 4 dpf. E,F: Shown are representative embryos labeled for the gut-specific marker ifabp at 4 dpf, comparing wildtype (WT) control (E) to morphant (F). Gut tissue is barely detectable in the morphant embryos (arrow in F). All views are lateral, dorsal to the top, and anterior to the left.

Mga Regulates gata4 Transcript Levels

To gain insight into genetic pathways that are altered in the mga morphant, we considered known factors of the transcriptional network expected to interact with mga, given that it contains both Tbox and bHLH domains. For this purpose, we screened members of the TBOX, GATA, and MYC families. The loss-of-function phenotype for some of these genes, for example tbx5, tbx20, and gata4, show similar cardiac morphogenetic defects compared with the mga morphants. We used quantitative real-time RT-PCR to screen the levels of transcripts in the morphants compared to control embryos. Several MYC family members are up-regulated modestly in the morphant embryos, including c-myc, l-myc, mxi, and mad4 (data not shown). Likewise gata6 and tbx20 transcript levels are enhanced 2-fold (Fig. 7A). However, these levels did not increase significantly over time. More striking, transcript levels for gata4 are enhanced 2-fold at the 5-somite stage compared to controls, and the difference is increased to 4-fold by 24 hpf (Fig. 7A).

Figure 7.

Gata4 transcript levels are increased for mga morphants in lateral plate mesoderm and the heart tube. A: Results of quantitative RT-PCR assays comparing gene expression levels for Mga morphants at 5 somites (blue) and 24 hpf (green) for the indicated cardiogenic genes. Shown is the average fold change in expression calculated from at least 3 independent experiments, with samples analyzed each time in triplicate. Samples were normalized to transcript levels for β-actin, and morphant transcript levels were quantified considering wildtype levels as 1. Student's t-test was performed comparing standard differences of the mean. The asterisks indicate statistically significant differences (P ≤ 0.05) compared to control (at both stages in the case of gata4). B, D: In-situ hybridization using a gata4 probe shows increased gata4 transcript levels in the lateral plate mesoderm (LPM) comparing WT (B) and MO1 morphant embryos (D) at 5 somites. Note in particular the increase in the gata4 levels in the caudal region of the lateral plate mesoderm (denoted with a solid line). C, E: The gata4 transcript levels at 24 hpf appear increased in the heart (HT) of morphants (E), but not changed in the gut/endoderm (END). Views (B, D) are dorsal, anterior to the top. Views (C, E) are dorsal, anterior to the right.

The identification of gata4 as a candidate downstream of mga is interesting, given that depletion of either gene generates a phenotype with several common features. Specifically, both morphants display a non-looping heart tube with associated cardiac edema, and regression of the yolk stalk extension at 48 hpf with subsequent defects in the growth of endoderm-derived organs including the intestine, liver, and pancreas (Holtzinger and Evans, 2005). The phenotypes are not identical. For example, Gata4-depleted embryos display a distinctive small kink at the tip of the tail (Holtzinger and Evans, 2005), and do not have the degenerative feature of the mga morphant brain. We compared the RNA transcript patterns for gata4 in wildtype and mga morphant embryos using in-situ hybridization. At the 5-somite stage, which is around the time when gata4 transcripts are first readily detected in the lateral mesoderm, there is a consistent alteration in the pattern. Transcript levels for gata4 in the mga morphant appear modestly higher with expression in more cells (a wider domain), particularly in the more caudal aspect of the lateral mesoderm. The data are consistent with Mga acting as a repressor to negatively regulate gata4 expression in lateral mesoderm (Fig. 7B, D). At 24 hpf, the gata4 pattern appears similar in the region of the developing gut tube region, whereas transcript levels associated with the linear heart tube are again modestly enhanced in the morphant compared to controls (Fig. 7C,E). Although the in situ analysis is not quantitative, the data are consistent with the qPCR data, suggesting that mga negatively regulates gata4 specifically in lateral mesoderm.

Depletion of Gata4 Can Rescue the Heart-Looping Phenotype of the mga Morphant

We next tested whether the increased levels of gata4 transcripts detected in the mga morphant have functional relevance for the phenotype. For this purpose, we carried out experiments co-injecting a gata4 morpholino with the mga morpholino, in order to reduce the levels of Gata4, potentially closer to normal levels. We found that 8 ng of mga morpholino MO1 is sufficient to generate efficiently with high penetrance the mga morphant phenotype. For the purpose of the rescue, we sought to reduce enhanced levels of gata4, but not to reduce levels excessively, since this would generate a gata4 morphant phenotype. By titration experiments, we found that injection of 8 ng gata4 morpholino was below the threshold required to generate a heart defect, consistent with previous studies (Holtzinger and Evans, 2005). Therefore, for each experiment we injected a cohort of embryos with 8 ng mga morpholino and then, subsequently, injected a subset of these mga morphants with 8 ng of the gata4-specific morpholino. Phenotypes were scored in the mga and mga+gata4 morphants at 2, 3, or 5 dpf. As expected, embryos injected with only the mga morpholino always fail to generate a looped heart tube and display severe cardiac edema. In contrast, a significant number (35%) of mga morphants that had been co-injected with the gata4 morpholino are rescued for the looping/edema phenotype. Representative examples are shown in Figure 8A–C. We scored looping in the cmlc2:gfp reporter strain, and similarly found approximately 1/3 of the mga morphants are rescued for looping by knockdown of gata4 (Fig. 8D–I). Normal heart development was never rescued, although by 5 dpf, heart morphogenesis of co-injected embryos are significantly advanced at similar rates (35–40%; Fig. 8J–L). These results show that mga interacts genetically with gata4, acting as an upstream negative regulator of gata4 with functional relevance to normal heart tube morphogenesis.

Figure 8.

Knockdown of gata4 can rescue partially the mga morphant heart tube phenotype. A–C: Shown are representative embryos that are wildtype (A), injected with only the mga-specific MO1 (B), or injected first with MO1 and subsequently with the gata4 morpholino (C). The ratios shown represent the number of embryos with an apparently normal (non-edemic) phenotype in a representative experiment (which was reproduced several times). Note that the cardiac edema present in 100% of the single morphants (B) is rescued in a significant percentage (35%) of the co-injected embryos. The brain defect is also largely rescued, although in contrast, the regression in the yolk stalk extension persists. D–L: Similar experiments were carried out and imaged using embryos derived from cmlc2:gfp transgenic fish. In these experiments, initial heart looping was rescued in double-knockdown embryos compared to the mga morphants as shown in F (51/128, 40%) or I (32/77, 42%) at 3 dpf. J–L: Although it is not normal, a similar rescue (35/105, 33%) of heart looping was scored at 5 dpf. All views are lateral, dorsal to the top, and anterior to the right except for G–I, which are ventral, anterior to the right. HT, heart; YSE, yolk stalk extension.

Depletion of Mga Activates p53-Dependent Apoptosis in the Brain

The mga morphant embryos show degeneration of the brain at a relatively early stage (24 hpf), suggesting a possible cell survival function. TUNEL assays revealed no increased apoptosis at 5 somites, but increased apoptosis in the brain of mga morphants compared to controls at 12 somites (Fig. 9A–D), and in the brain and neural tube at 24 hpf (Fig. 9E, F). We tested if this increase in apoptosis is p53-dependent by co-injecting the mga MO1 morpholino together with a morpholino targeting the p53 protein. This showed a significant decrease in the amount of apoptotic cells detected by TUNEL staining (Fig. 10).

Figure 9.

Depletion of mga results in increased apoptosis in the head/brain region. Shown are representative TUNEL assays analyzing control wildtype (WT) or morphant (MO1) embryos at the stages indicated. There is no apparent apoptosis at early somitogenesis stages (A, B), while at the 12-somite or 24-hpf stage there is a significant increase in TUNEL-positive cells (asterisks). Note that apoptosis was not detected in heart or gut/endoderm. Views are: (A,B) lateral, dorsal to the top, and anterior to the right; (C,D) lateral, dorsal to the left, and anterior to the top; (E,F) dorsal, anterior to the top. H, head.

Figure 10.

Knockdown of p53 can rescue the mga morphant apoptosis in the head/brain. A–D: Shown are representative morphant embryos injected with (A) 7 ng or (C) 10 ng of mga MO1 morpholino followed by TUNEL staining at 48 hpf. Note that there is a significant and dose-dependent accumulation of apoptosis in the brain mainly in the midbrain (MB) and hindbrain (HB) regions. B, D: Apoptosis is either (B) fully or (D) significantly reduced when embryos are co-injected with 1 ng of p53 morpholino.

As was true for heart looping, we found that the degenerative brain phenotype is also rescued for many of the embryos in mga morphants co-injected with the gata4 morpholino, shown by comparing the morphology of the brain in mga and mga+gata4 morphants (Fig. 11A–C) and by rescue of the enhanced TUNEL-staining seen in the brain region of mga single morphants (Fig. 11D–F). In contrast, the yolk stalk extension remains regressed in both the single and double-injected embryos (seen in Fig. 8B–C) and the expression of transferrin-expressing liver or ifabp-expressing gut fails to recover (Fig. 11G–O), suggesting that mga-mediated up-regulation of Gata4 is not the cause of the gut phenotype.

Figure 11.

Knockdown of gata4 can rescue partially the mga morphant brain degeneration, but not the gut phenotype. A–C: Shown are representative embryos at 24 hpf that are wildtype (A), morphant injected with only the mga-specific MO1 (B), or injected first with MO1 and subsequently with the gata4 morpholino (C). The ratios shown represent the number of embryos with a grossly normal brain structure in a representative experiment (which was reproduced several times). Note that the brain degeneration present in 100% of the single morphants (B) is rescued in a significant percentage (33%) of the co-injected embryos (C). D–F: Similar results showing that the morphological rescue correlates with an abrogation of apoptosis measured by TUNEL staining. The representative 24-hpf embryos are WT (D), MO1 morphants (E), or co-injected with MO1 and gata4 morpholino (F). G–L: Shown are representative embryos stained for transcripts encoding the liver marker transferrin at 3 dpf (G–I) or 4 dpf (J–L). Embryos are wildtype (WT) control (G, J), MO1-injected (H, K), or co-injected with MO1 and gata4 morpholino (I, L). M–O: A similar analysis was performed at 4 dpf using a probe for ifabp to stain the gut. In contrast to the rescue for brain morphogenesis, knockdown of gata4 failed to rescue the liver or gut phenotype in the mga morphant background. Views (A–F) and (M–O) are lateral, dorsal to the top, anterior to the left). G–L: Views are dorsal, anterior to the left. L, liver; G, gut.

An Early Role Is Indicated for mga in Mesendoderm Development

We performed a genome-wide microarray analysis comparing cDNA levels derived from RNA isolated at the 12-somite stage from wildtype or mga morphant embryos (not shown). While many genes are altered in mga morphants, we evaluated the gene-set for common properties. A set of selected genes that showed significant changes includes lefty2, bambi, sizzled, and sestrin3, which suggests that mga regulates multiple genes involved in the control of TGFβ-family signaling. Lefty2 encodes a growth factor that inhibits nodal signaling and regulates left/right asymmetry, and qPCR analysis verified that lefty2 transcript levels are significantly enhanced by the 5-somite stage (Fig. 12). We analyzed embryos for heart laterality defects at 24 hpf, but failed to document a significant difference between wildtype and mga morphant embryos (data not shown). Thus, the heart tube morphogenetic defect is subsequent to initial lateral side choice. We also performed a qPCR screen for known downstream targets of nodal signaling comparing transcript levels between wildtype and mga morphant embryos at 50% epiboly (gastrulation) and 5 somites (early somitogenesis). Most of the genes analyzed showed no significant difference, at either stage, indicating that nodal signaling is not globally disturbed. However, transcript levels for several early regulators of mesendoderm development, namely, sox17, tbx6, and cas are decreased approximately 2-fold at 50% epiboly (Fig. 12). Sox17 is an early endoderm marker, while both tbx6 and cas are expressed during gastrulation at the margin; tbx6 is expressed in both mesoderm and endoderm, while cas transcripts are more restricted to the endoderm layer. Therefore, normal expression of mga in the margin during gastrulation appears to be important upstream of sox17, tbx6, and cas, key regulators of the tissue that ultimately contributes to both mesoderm and endoderm derivatives.

Figure 12.

A subset of early mesendoderm regulatory genes is affected in the mga morphant. Shown are the results of quantitative RT-PCR assays for the indicated genes, which are all known targets of nodal signaling, important for mesendoderm development. Transcript levels compared wildtype embryos with mga morphants at 50% epiboly (gastrulation, solid bars, left for each pair) or at 5 somites (early somitogenesis, diffused bars, right for each pair). Shown are the average fold changes in expression calculated from at least 3 independent experiments, with samples analyzed each time in triplicate. Samples were normalized to transcript levels for β-actin, and morphant transcript levels were quantified considering wildtype levels as 1. Student's t-test was performed comparing standard differences of the mean. The asterisks indicate statistically significant differences (P ≤ 0.05) compared to control (at both stages in the case of tbx6).


Our studies show that the TBOX transcription factor mga is essential for zebrafish embryogenesis and regulates normal development of the brain, heart, and gut, consistent with its normal expression pattern. We believe that our approach has targeted successfully the mga transcript, since multiple morpholinos of distinct sequences generate equivalent phenotypes. One limitation of the approach is that morpholinos deplete Mga protein from the earliest stages of development. Since mga is expressed maternally and transcripts are expressed widely during early development, some or all aspects of the morphant phenotype might be due to relatively indirect effects from deregulation of early patterning programs. However, our observations suggest that mga functions differently in distinct organ systems, which might reflect the unique structural characteristic of the protein, with potential to interact with both TBOX- and MYC-directed programs.

The function of mga in brain development appears related to cell survival. Brain morphogenesis is disrupted in morphant embryos and this correlates with widespread p53-dependent apoptosis. This brain degeneration phenotype could be caused by off-target effects, which have been documented using some morpholinos (Robu et al., 2007). However, for several reasons we believe it may reflect a normal function of mga. First, the pattern of apoptosis coincides precisely with the normal expression pattern of mga transcripts in the developing brain. Second, the same phenotype is recapitulated using three morpholinos with distinct sequences, targeting either translation or splicing, and is not seen using a mismatch control morpholino. Third, we detect no increase in the expression of the N-terminal truncated p53 isoform, which has been shown to be associated with off-target effects (data not shown). Finally, the phenotype can be rescued by co-injection of the gata4 morpholino. This suggests that apoptosis is caused in the brain through suppression of a mga-regulated program that normally restricts gata4. Formal proof for this hypothesis will require RNA rescue experiments, which are particularly challenging for this very large gene.

The effect of Mga-depletion in the gut and heart is not associated with apoptosis. The function of mga in gut-derived organs is likely due to early endoderm defects since there is a consistent failure in morphants at 48 hpf to maintain the yolk stalk extension. Interestingly, this feature phenocopies the gata4 morphant. We did not detect obvious changes in gata4 transcript levels in the primitive gut tube of mga morphants. Neither does depletion of gata4 rescue the yolk stalk defect. This suggests that mga and gata4 both function in early gut organogenesis, and that mga is not a negative regulator of the gata4 gene in this context. In contrast, at least some aspects of the cardiac phenotype can be attributed to the loss of normal control for gata4 levels. By reducing excess Gata4, the early heart-looping defect can be rescued in the mga morphant. It is not known if Mga protein functions directly on gata4 regulatory elements, although both TBOX- and E-BOX-binding sites are present in a known gata4 cardiac regulatory element (Heicklen-Klein and Evans, 2004). Mga could function indirectly by regulating the function of other TBOX proteins, since several are co-expressed with gata4 in the developing embryo.

The microarray analysis showed a significant increase in transcripts for lefty2, a negative regulator of nodal signaling. This was validated by qPCR data, which showed that by the 12 somite stage, lefty2 transcript levels are 30-fold increased compared to controls (not shown). However, we could not find strong evidence that nodal signaling is broadly disrupted in the mga morphants, since many of the known nodal target genes are not altered in expression. Rather, a subset of key regulators for mesendoderm development are deregulated, including sox17, cas, and tbx6. This early alteration may lead to the subsequent changes in gata4 levels and endoderm defects. It is interesting to note that tbx6, in addition to interacting genetically with mga, is also the TBOX protein most closely related evolutionarily to mga (Lardelli, 2003).

In summary, mga is required for normal organogenesis and functions in the GATA/TBOX regulatory network. The mga and gata4 loss-of-function organ phenotypes are remarkably similar, yet mga depletion leads to enhanced gata4 transcript levels. This suggests that depletion or over-expression of Gata4 generates similar phenotypes. This is a known characteristic of TBOX proteins (Zweier et al., 2007), but had not been previously associated with GATA factors. Testing this directly will require the means to target gata4 specifically to the mga expression domain. An added feature is the potential for mga to integrate, in addition, regulatory programs of the MYC pathway, thereby linking cell survival, proliferation, and differentiation programs. We measured increased transcript levels for some MYC family members in the mga morphant, although preliminary attempts to rescue the phenotype by depletion of MAX (thereby in principle lowering MYC activity) were not successful (data not shown). However, MGA was found within a multimeric protein complex that includes MAX, E2F-6, DP-1, HP1γ, and YAF2, as well as histone methyltransferases, implicating MGA with functions related to multiple signaling pathways (Ogawa et al., 2002). With respect to cell survival, it is interesting that MGA and YAF2 are found in a co-complex, since the RYBP/YAF2 protein family has been shown to encode pro-apoptotic functions. In zebrafish, yaf2 inhibits caspase 8–mediated apoptosis (Stanton et al., 2006). Future structure-function studies should determine whether mga functions in distinct or overlapping programs associated with TBOX- and MYC-regulated pathways.



Zebrafish embryos were maintained at 28°C and staged as described (Westerfield, 1995). Wildtype fish are hybrids derived from a cross between AB and TU strains. The cmlc2:gfp strain was provided originally by H.J. Tsai (Huang et al., 1997), and we generated equivalent lines independently. The ef1a:gfp strain was kindly provided by Didier Stainier's laboratory.

Mga Cloning

Total RNA was isolated from wildtype embryos at defined stages of development and purified using Tri-Reagent according to the manufacturer's instructions (Molecular Research Center, Cincinnati, OH). First-strand cDNA synthesis was performed using the Superscript III First-Strand Synthesis System (Invitrogen, Carlsbad, CA). To isolate sequences from the 5′ end of the mRNA, 5′-RACE was performed as specified according to the GeneRacer kit (Invitrogen). Five overlapping fragments that were together predicted to encompass an entire ORF were cloned by RT-PCR (for primer details, see Supp. Table S1, which is available online) into the pCRII-TOPO vector (Invitrogen), and the sequence of each was determined. In subsequent genome browser updates, the sequence verified our isolated fragments. Also, in order to confirm the entire sequence, full-length mga was amplified using primers to generate a single product that encompassed the entire cDNA, using LongRange PCR (Qiagen, Valencia, CA). The zebrafish mga Genbank Accession number is GU122854.

Whole Mount In Situ Hybridization

Whole-mount in situ hybridization was performed essentially as described (Alexander et al., 1998). Briefly, embryos were treated with 0.003% phenylthiourea (PTU) to prevent pigmentation. After fixation, embryos older than 24 hr were treated with 10 μg/ml proteinase K. Hybridization was performed at 70°C, in 57% formamide buffer with digoxigenin-labeled RNA anti-sense probes. The probes used for in situ hybridization were prepared using either Sp6 or T7 polymerase and linearized templates derived from fragments (300–500 bp) of mga obtained from RT-PCR products described above. At least 3 different probes mapping to different regions of the mga cDNA were used and these generated the same expression patterns. Probes used for Figure 2 are derived from exon 1 and represent the first 406 bp of the ORF, starting at the ATG initiation codon. Sense probes were generated similarly and used as controls for background. Probes for gata4 and transferrin have been described (Holtzinger and Evans, 2005). The probe for ifabp was generated by isolating a specific fragment by RT-PCR using specific primers that have been described (Holtzinger and Evans, 2005). The fragment was cloned and sequence verified.

Morpholino Design and Microinjection

Three morpholino oligomers were designed to target specifically mga sequences and were purchased from the manufacturer (Gene Tools, LLC). MO1 (5′-CATGGGAATTAAGTCATC TGGATAG-3′) is complementary to the 5′UTR of the mature mga transcript (a translation blocker). MO2 (5′-AACTCAGTCTGAGCAAAACTGA GAG-3′) targets sequences +746–770 relative to the initiation ATG). This site was chosen initially as a putative splice-blocker, based on the gene structure predicted by the UCSC zebrafish genome browser. However, we discovered by sequencing cDNAs that the predicted structure is wrong, and that the MO2 target site is not an exon-intron junction, but lies fully within the first exon. It may be less effective compared to MO1 because it is distal to the ATG. MO3 (5′-TCTTGTTACTGTTAGAGAAAGAGAG-3′) targets the boundary between the true intron1 and exon2 of the mga pre-mRNA (a splicing blocker). Blast searches predicted that the morpholinos should be specific for mga. Morpholinos were injected into 1-cell-stage embryos in various doses to establish a threshold response. In control experiments, a mis-match morpholino was used, which is based on the MO1 sequence but contains 5 mis-matches (5′CATCGCAATTAACTCATGTGCATAG-3″ bold indicates mis-matches). Using the same concentration range as MO1, this failed to generate any apparent phenotype. The amount of each morpholino used for injection was: MO1, 5–10 ng (the range was equivalent); MO2, 15 ng; MO3, 1 ng; mis-match control, 5–10 ng (equivalent to MO1). For the rescue experiments, 8 ng of MO1 was injected into cohorts of embryos followed by injection of 8 ng of the gata4 morpholino (5′-TCCACAGGTGAGCGATTATTGC TCC-3′). This is a gata4 translation blocker that was described and validated previously (Holtzinger and Evans, 2005). For the p53 rescue experiments, either 7 or 10 ng of MO1 were injected into cohorts of embryos followed by injection of 1 ng of the p53 morpholino into a subset. Embryos were then raised and fixed at 48 hpf using 4% PFA followed by TUNEL staining. The p53 morpholino targets the 5′UTR and serves as a translation blocker (5′- TTGATTT TGCCGACCTCCTCTCCAC; Open Biosystems, Huntsville, AL).

TUNEL Assays

Equivalently staged embryos were fixed overnight in 4% PFA. After washing with PBT, they were dechorionated and dehydrated with 100% methanol. Prior to staining, the embryos were rehydrated into PBT, post-fixed for 1 hr in 4% PFA, blocked for 1 hr in 1 × TdT buffer, and incubated with a mixture of the TdT enzyme (150 U/ml; Invitrogen) and digoxigenin-labeled dUTP (0.5 μM; Roche Applied Science, Nutley, NJ) overnight at room temperature. The embryos were then washed first in PBT/1 mM EDTA at 65°C for 2 hr and then in PBT/bovine serum albumin. Embryos were incubated with anti-digoxigenin Fab antibody conjugated to alkaline phosphatase (Roche Applied Science) overnight at 4°C. The signal was detected using alkaline phosphatase NBT/BCIP staining and the embryos were then cleared in a BB:BA (benzyl benzoate/benzyl alcohol) 2:1 solution.

Quantitative Real-Time PCR

RNA was extracted from staged wildtype or morphant embryos using the RNeasy mini kit (Qiagen). First-strand cDNA synthesis was performed as described above. The cDNA was analyzed with Qiagen QuantiTect SYBR Green Mix (Qiagen) by quantitative reverse transcription (RT)–PCR using the Opticon DNA Engine 3 real-time PCR machine (Bio-Rad Laboratories, Hercules, CA) or Light Cycler 480II (Roche). All samples were prepared in triplicate, and each experiment was repeated at least 3 times using independent batches of embryos. The PCR cycle conditions were 95°C for 15 min (94°C for 15 sec, 60°C for 30 sec, and 72°C for 30 sec) for 40 cycles. A complete list of primers is found in Supplementary Table S2. The Ct value data were analyzed using the 2-ΔΔT method (Livak and Schmittgen, 2001).


We thank Didier Stainier and H.J. Tsai for transgenic reporter fish. The authors are grateful for the assistance of David Heck, who provided excellent technical assistance to initiate the project using 5′RACE, and Kellie McCartin, who provided essential technical assistance and zebrafish husbandry. This work was supported by a grant to T.E. from the NIH (R01 HL064282). A.R. designed the study, carried out experiments, and helped write the report. T.E. designed the study and wrote the report.