Tmem88a mediates GATA-dependent specification of cardiomyocyte progenitors by restricting WNT signaling

Heart formation is a dynamic process involving cardiomyocyte progenitor (CP) specification, cell migration, cardiomyocyte differentiation and complex tissue morphogenesis (Brand, 2003; Evans et al., 2010; Harvey, 2002). Retrospective clonal lineage analysis has identified at least two major CP populations that contribute to the heart (Buckingham et al., 2005; Meilhac et al., 2004). The first wave of CPs comprises the first heart field (FHF) and differentiates into cardiomyocytes that form the left ventricle, atrioventricular canal and the atria. The second pool of CPs, or the second heart field (SHF), constitutes a highly proliferative cell population that contributes to the right ventricle and outflow tract (Black, 2007; Cai et al., 2003; Kelly et al., 2001; Mjaatvedt et al., 2001; Vincent and Buckingham, 2010; Waldo et al., 2001; Zaffran et al., 2004). The formation of the FHF and SHF and the regulatory network that coordinates CP specification is conserved in human, mouse, chick and zebrafish (Brand, 2003; Liu and Foley, 2011; Zhou et al., 2011). Therefore, molecular insight gleaned from model systems is relevant to the development of emerging targeted strategies to expand human CPs for therapeutic purposes.

The molecular drivers of CP specification include several well-studied transcription factor families and signaling pathways (Evans et al., 2010; Stainier, 2001). The generation of precardiac mesoderm requires the function of transcription factors MESP1 and MESP2. In chimeric mouse embryos, cells doubly mutant for Mesp1 and Mesp2 selectively fail to contribute to the heart (Saga et al., 2000). Downstream transcription factors believed to impact CP specification belong to the GATA, NK2, MEF2, SRF, ISL and TBX families. Various combinations of these factors selectively influence FHF or SHF development (Buckingham et al., 2005; Meilhac et al., 2004). Cooperating signaling molecules that enhance CP expansion include those from the BMP, FGF, HH, Nodal, retinoid and WNT families. BMPs and FGFs are secreted by cardiac-associated endoderm (Foley et al., 2006; Lough and Sugi, 2000), whereas HH and Nodal act cell-autonomously to stimulate specification (Keegan et al., 2004; Thomas et al., 2008). By contrast, retinoic acid (RA) inhibits CPs. RA delineates the forelimb organ field and induces Hox5b expression, which prevents caudal CP expansion into the neighboring forelimb territory (Waxman et al., 2008). RA also increases levels of the LIM protein Ajuba, which limits SHF formation by repressing the SHF-specific transcription factor Islet1 (Witzel et al., 2012). Rostrally, the domain of CP specification is limited by neighboring vascular and hematopoietic tissue. Hematovascular transcription factors SCL and ETSRP can override CP fate (Schoenebeck et al., 2007).

Understanding the function of the WNT pathway for CP specification has been more challenging, as canonical WNT signaling both promotes and restrains CP specification (Kwon et al., 2007; Lavery et al., 2008; Naito et al., 2006; Paige et al., 2010; Tzahor, 2007; Ueno et al., 2007). During gastrulation, WNT activity induces ventrolateral mesoderm to establish the precardiac territory. Subsequently, WNT signaling restricts CP specification. Other than being cell-autonomous, the molecular mechanism of WNT-mediated inhibition of specification is undefined. However, ectopic WNT activation correlates with reduced GATA factor expression (Afouda et al., 2008; Martin et al., 2010).

GATA factors are zinc-finger transcriptional regulators that control organ development (Patient and McGhee, 2002). Gata4, Gata5 and Gata6 each contribute to numerous aspects of vertebrate cardiogenesis. We showed previously that gata5 and gata6 function together as essential components for CP specification in zebrafish (Holtzinger and Evans, 2007). When both factors are simultaneously depleted, there is a complete failure in the formation of nkx2.5⁺ CPs, and the corresponding embryos develop without hearts. Loss-of-function studies in other species, including ascidians (Ragkousi et al., 2011), flies (Klinedinst and Bodmer, 2003) and mice (Singh et al., 2010; Zhao et al., 2008), confirm the requirement of GATA factors for cardiogenesis and, in particular, CP specification.

Gain-of-function experiments have shown that GATA factors can integrate into molecular complexes sufficient for specification (Ieda et al., 2010; Lou et al., 2011; Takeuchi and Bruneau, 2009). For example, combined expression of Gata4, Mef2c and Tbx5 in postnatal mouse fibroblasts was reported to reprogram fibroblasts into cardiomyocytes (Ieda et al., 2010). In embryonic mouse explants, ectopic expression of Gata4 and Baf60c (Smarcd3), which encodes a chromatin-remodeling protein, conferred cardiogenic fate onto non-cardiogenic mesoderm (Takeuchi and Bruneau, 2009). Cell transplantation studies in zebrafish further demonstrated that GATA factors and Baf60c (Smarcd3b) act cell-autonomously to impact cardiac fate (Lou et al., 2011).

Despite their importance, the mechanism by which GATA factors impact CP specification is unknown. Cardiac transcription factors, including GATA factors, cross-regulate each other at both the transcriptional and post-transcriptional levels (Zhou et al., 2012), but it is less clear how they integrate into the established signaling network that modulates CP specification. Here, we take advantage of the zebrafish model, which is dependent on gata5 and gata6 for CP formation, to screen for downstream transcriptional targets potentially involved in cardiac specification. One of the top candidates to emerge from this screen is tmem88a, which encodes a transmembrane protein and putative negative regulator of canonical WNT signaling. We show that GATA factors act through tmem88a to modulate WNT signaling and thereby regulate CP specification.

Embryos were raised at 28.5°C and morphologically staged as described (Kimmel et al., 1995). All experimental procedures were conducted in compliance with the Weill Cornell Medical College IACUC. Wild-type embryos were derived from crosses of the AB and TU lines. The following strains were used: tp53^M214K (Berghmans et al., 2005), cloche (Stainier et al., 1995), tg(myl7:eGFP) (Huang et al., 2003), tg(myl7:dsRed2-nuc) (Mably et al., 2003), tg(hsp70l:dkk1-eGFP) (Stoick-Cooper et al., 2007), tg(hsp70l:wnt8a-eGFP) (Weidinger et al., 2005) and TOPdGFP (Dorsky et al., 2002).

Translation-blocking (5′-GGAAGACTCATCTTGCCGTTCATCA) and splice-blocking (5′-TGACGCTGAACCTGTGGAACACAGA) MOs were used to target tmem88a (Gene Tools) at optimal doses of 6 ng and 8 ng, respectively. All experiments, apart from MO validation, were performed using the translation blocker. Standard control MO (5′-CCTCTTACCTCAGTTACAATTTATA) and previously described gata5 and gata6 MOs were also used (Holtzinger and Evans, 2007).

The tmem88a cDNA was PCR amplified and cloned into the pCS2+ vector (D. Turner, U. Michigan). A mutant tmem88a construct was also generated containing a frameshift mutation predicted to cause protein truncation. Capped mRNA was synthesized in vitro (mMESSAGE mMACHINE, Invitrogen), diluted in RNase-free water and microinjected at 200 pg.

Samples prepared with the Illumina mRNA-seq Sample Preparation Kit were sequenced on the Illumina Genome Analyzer IIx with single 36 bp reads. Alignment to the zebrafish Zv9.61 genome was performed with the Burrows-Wheeler Aligner (Li and Durbin, 2009). Differential expression analysis was conducted in DESeq/Bioconductor R (Anders and Huber, 2010) with DESeq adjusted P<0.05, RPKM >1.00 and fold-change >2.00. Biological significance was inferred by gene ontology annotation through the Database for Annotation, Visualization and Integrated Discovery (DAVID) (Huang et al., 2009a; Huang et al., 2009b). Analysis was facilitated by GobyWeb (Dorff et al., 2012). The raw data are available at Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/projects/geo/) with accession number GSE44026.

WISH was performed as described (Jowett, 1998). The tmem88a sequence was PCR amplified from cDNA clone MDR1734-9140265 (OpenBiosystems) using forward primer 5′-GAGTAGAGCAGGAAAGAAGGAAGA and reverse primer 5′-GAATTTTAAACAGGGGTTTATCCA and ligated into the pCS2+ vector. The riboprobe was synthesized from an EcoRI-linearized plasmid with T7 RNA polymerase. Additional probes were prepared as described previously: nkx2.5, myl7, amhc, vmhc (Reiter et al., 1999); scl, gata1 (Liao et al., 1998); gata4 (Heicklen-Klein and Evans, 2004); ntl (Schulte-Merker et al., 1992); pax2a (Krauss et al., 1991); and myod (Weinberg et al., 1996). Tyramide signal amplification (TSA) dual fluorescent in situ hybridization was performed as described (Brend and Holley, 2009). A fluorescein-labeled tmem88a probe and a digoxigenin-labeled mixture of nkx2.5 and ntl probes were hybridized overnight at 65°C. The TSA Plus Cyanine 5/Fluorescein System (Perkin Elmer) was used to develop the riboprobe signals.

TSA fluorescent in situ hybridization for digoxigenin-labeled nkx2.5 and ntl probes was combined with TMR-Red In Situ Cell Death Detection (Roche) or phospho-Histone H3 (PH3) immunohistochemistry (Millipore). For TSA/PH3 staining, embryos were fixed for 2 hours in 4% paraformaldehyde and riboprobe hybridization was performed at 55°C. The primary anti-PH3 antibody was diluted 1:200 dilution and detected with goat anti-rabbit IgG Alexa 546 (Invitrogen) diluted 1:500.

RNA isolated using Trizol Reagent (Invitrogen) was reverse transcribed with the Superscript III First-Strand Synthesis System (Invitrogen). The tmem88a transcript was detected with forward primer 5′-TCATCTTTGGCCTCATCCTC-3′ and reverse primer 5′-AAAGCAACACAGCCTCCATC-3′. Gene expression was normalized to β-actin levels (Holtzinger and Evans, 2007). Relative fold change was determined as described (Livak and Schmittgen, 2001).

Immunohistochemistry on zebrafish hearts was performed as described (Yang and Xu, 2012). Anti-DsRed2 (Clontech) was diluted 1:100 and detected with goat anti-rabbit IgG Alexa 546 (Invitrogen). S46 (DSHB) was diluted 1:20 and detected with goat anti-mouse IgG1 Alexa 647 (Invitrogen). Secondary antibodies were diluted 1:500.

Live embryos were imaged using a Nikon SMZ1500 with an Insight Firewire 2 digital camera and SPOT advanced imaging software. WISH preparations were mounted in glycerol and similarly imaged. TSA-treated embryos and embryonic hearts were imaged using the Zeiss LSM 510 confocal microscope with Zen software. The 40× and 20× objectives were used, respectively. Images were analyzed in ImageJ (NIH) and Adobe Photoshop. Student’s t-test was used to assess statistical significance. All data were graphed with error bars corresponding to one standard deviation of the mean.

Embryos were heat shocked at 39°C and subsequently screened for eGFP expression to confirm the presence of the transgenic allele in either tg(hsp70l:dkk1-eGFP) or tg(hsp70l:wnt8a-eGFP) background.

Gata5 and Gata6 independently regulate heart morphogenesis, but together they act to coordinate CP specification (Holtzinger and Evans, 2007). Based on this observation, we developed a strategy to identify gata5 and gata6 (gata5/6) co-dependent candidate genes involved in CP specification. A next-generation sequencing screen was performed comparing RNA samples from gata5/6 double-morphant embryos with wild-type controls and individual gata5 or gata6 morphants. Transcription profiles were evaluated to identify genes uniquely dysregulated in gata5/6 double morphants (Fig. 1A). Samples were collected at the bud or 6-somite stage of development. Bud stage corresponds to the end of gastrulation and onset of precardiac patterning in the lateral mesoderm. At the 6-somite stage, the expression pattern of the CP marker nkx2.5 has previously been used to reliably assess CP formation (Chen and Fishman, 1996; Hami et al., 2011; Schoenebeck et al., 2007). In wild-type embryos, the expression of nkx2.5 initiates between the bud and 6-somite stages of development, whereas in gata5/6 morphants CP specification fails at this time (Fig. 1B). The screen was therefore designed to identify transcriptional components closely associated with the loss of both Gata5 and Gata6 at the time when CP specification fails. The screen does not distinguish between direct or indirect GATA targets.

Using defined criteria, we identified 225 upregulated and 304 downregulated genes specific to the gata5/6 double morphants (fold change >2; Δgata5/6 > Δgata5 + Δgata6; P<0.05) (Fig. 1C; supplementary material Table S1). Upregulated genes included a subset of parvalbumins, vitellogenins and serpins. Downregulated genes were mainly associated with hematovascular and cardiac development. In particular, the known CP markers nkx2.5, tbx20 and gata4 were identified by the screen (Fig. 1D), validating the experimental strategy.

To select genes for further study, we focused on candidates that are expressed in the precardiac mesoderm during embryogenesis. The published whole-mount in situ hybridization (WISH) data in the Zebrafish Model Organism Database (ZFIN) was used as a reference to select these candidates (Sprague et al., 2006; Sprague et al., 2008). From the initial set of downregulated genes, 12 were identified (Table 1) as expressed in, or near, the anterior lateral plate mesoderm (ALPM), including tmem88a, a gene that encodes a putative transmembrane protein of unknown function.

A role for tmem88a in cardiac development has not been described. However, gain-of-function experiments have implicated the human homolog, TMEM88, in the negative regulation of canonical WNT signaling (Lee et al., 2010). Nuclear magnetic resonance and Xenopus explant experiments showed that TMEM88 can bind and regulate disheveled 1 (Dvl1), a key player in the WNT pathway. Binding occurs between a highly conserved TMEM88 C-terminal Val-Trp-Val motif and the Dvl1 PDZ domain. Since WNT signaling is an essential mediator of CP specification, we hypothesized that tmem88a, co-regulated by Gata5 and Gata6, might potentiate CP specification by modulating WNT activity during embryonic cardiogenesis.

Based on our screen, tmem88a is expressed at both the bud and 6-somite stages of development in wild-type zebrafish embryos (Fig. 2A). In gata5/6 morphants, tmem88a expression levels at the bud and 6-somite stages are reduced by ∼70% and ∼45%, respectively. Using quantitative reverse-transcription PCR (qPCR), we confirmed that tmem88a expression is reduced in double morphants, compared with controls, by ∼50% (Fig. 2B). WISH analysis of tmem88a transcript localization in gata5/6 morphants and wild-type controls confirmed the gata5/6 dependency of tmem88a expression, particularly within the ALPM (Fig. 2C,D). The gata5/6-dependent expression pattern of tmem88a implicates a role for this gene in the regulation of lateral mesoderm development.

To evaluate the role of Tmem88a in development, the tmem88a expression pattern was analyzed in wild-type embryos between the 8-somite stage and 24 hours post-fertilization (hpf), relative to defined markers. At the 8- and 14-somite stages, tmem88a transcripts were identified throughout the lateral plate mesoderm (Fig. 3A-D). Notably, tmem88a transcripts formed a distinctive ‘flare’ pattern in the ALPM that was reminiscent of nkx2.5⁺ CP fields (Fig. 3A-D, insets). The notochord tip, labeled by ntl, was used as a morphological landmark for locating the adjacent CPs. Fluorescent dual in situ hybridization showed that tmem88a and nkx2.5 transcripts colocalize in 8- and 14-somite stage embryos, confirming CP-specific tmem88a expression (Fig. 3E,F). Additionally, WISH for tmem88a was performed in similarly staged cloche mutant embryos, which lack progenitors for the hematovascular lineages. Transcript localization was identical in the ALPM of cloche mutants and wild-type embryos, indicating that, in the vicinity of nkx2.5⁺ cells, tmem88a is expressed in precardiac fields and not in the presumptive endothelium (supplementary material Fig. S1). By 23-somites, however, the tmem88a transcript pattern becomes restricted to vascular tissue, including the endocardium and head vessels (supplementary material Fig. S2). Endothelial-specific expression persists until 48 hpf, after which point tmem88a transcripts are undetectable by WISH (data not shown). The expression of tmem88a in emerging CPs is consistent with an early role in cardiogenesis.

Loss-of-function studies were performed to determine whether tmem88a regulates cardiogenesis. Two morpholinos (MOs), a translation-blocker (tbMO) and a splice-site blocker (ssMO), were designed to target the tmem88a transcript (supplementary material Fig. S3A). Antibodies for validating the tbMO are not available. However, the effectiveness of the ssMO was confirmed by semi-quantitative RT-PCR. The native tmem88a transcript appeared to be eliminated in embryos injected with 8 or 12 ng ssMO (to the limit that we can detect using this strategy), and an aberrant unspliced variant accumulated (supplementary material Fig. S3B). The aberrant transcript was sequenced and found to encode a truncated and presumably dysfunctional protein. Individual injection of tbMO or ssMO into single-cell embryos caused 100% penetrant, identical and highly reproducible cardiac defects (supplementary material Fig. S4A). Co-injection of the MOs, each at subthreshold levels that did not individually perturb cardiogenesis, fully recapitulated this phenotype, confirming the specificity of the tmem88a knockdown (supplementary material Fig. S4B). Morphologically, the tmem88a-related cardiac phenotype was first evident in 2-day-old embryos. At 28 hpf, control and tmem88a morphants were indistinguishable (supplementary material Fig. S5). However, by 48 hpf the tmem88a morphants had shortened trunks, reduced cranial features and pericardial edema (Fig. 4A). Closer examination revealed a lack of blood, a translucent heart-string and slowed heart rate (Fig. 4B,C; supplementary material Fig. S6).

Morphants were generated in the tg(myl7:eGFP) background, which expresses cardiomyocyte-specific eGFP. At 72 hpf, control hearts were fully looped, with adjacently positioned cardiac chambers of comparable size (Fig. 4D). By contrast, Tmem88a-deficient hearts were linear, and the cardiac chambers were reduced in size (Fig. 4E). These defects persisted until 7 days post-fertilization, at which point the embryos died. Similar cardiac defects, along with the absence of blood, were observed in tp53^M214K mutants injected with tmem88a MOs (supplementary material Fig. S7), indicating that Tmem88a is required specifically for proper cardiogenesis and blood formation. We note that the cranial and truncal abnormalities were less pronounced in these embryos, so that partial off-target effects cannot be ruled out for those phenotypes.

Cardiogenesis is an elaborate process that requires CP specification, cardiomyocyte differentiation, proliferation and morphogenesis. Cardiomyopathies can arise from failure at any of these stages. To better understand the requirement of Tmem88a for heart development, we analyzed heart cellularity in control and tmem88a morphants, derived from the tg(myl7:DsRed2-nuc) line, which expresses nuclear cardiomyocyte-specific DsRed2. At 51 hpf, embryonic hearts were dissected, immunostained and imaged to quantify cardiomyocyte cell numbers (Fig. 5A,B). Atrial, ventricular and total cardiomyocytes were counted (Fig. 5C). Depletion of Tmem88a causes a 30% reduction in total cardiomyocyte number. Atrial cardiomyocytes decrease by 20% and ventricular cells by 40%. The vertical length of the cardiac chambers was also measured (Fig. 5D). Morphant hearts were, on average, 30% shorter than those of controls.

The heart defects observed in tmem88a morphants could result from aberrations in CP specification, proliferation or cardiomyocyte differentiation. To assess cardiomyocyte differentiation, WISH was performed for differentiation markers in tmem88a and control morphants. atrial myosin heavy chain (amhc; myh6 - Zebrafish Information Network) and ventricular myosin heavy chain (vmhc) were used to monitor atrial and ventricular development, respectively. At the 23-somite stage, control embryos showed the wild-type expression patterns of both markers (Fig. 6A,B). By contrast, tmem88a morphants had disrupted amhc and vmhc transcript patterns (Fig. 6C,D). The area of vmhc staining was reduced by 35%, and amhc staining was decreased by 45% (Fig. 6E,F). Relative expression analysis, as measured by qPCR, confirmed the reduction of both markers in Tmem88a-depleted embryos (Fig. 6G,H).

To determine whether the changes in differentiation marker expression were a consequence of defective CP specification, WISH was performed for the CP marker nkx2.5. As before, the rostral tip of the notochord was used as a landmark for orienting the normal nkx2.5 expression domains. At the 8-somite stage, control morphants displayed the expected bilateral nkx2.5 expression pattern (Fig. 7A). By contrast, identically staged tmem88a morphants had a significant reduction in the nkx2.5⁺ domains (Fig. 7B). Using area measurements, we established that the nkx2.5⁺ CP field was 60% smaller in tmem88a morphants than in controls (Fig. 7C). Relative expression analysis of nkx2.5 by qPCR confirmed this reduction (Fig. 7D).

Since impaired CP specification might result from a more general mesodermal defect or loss of ALPM, we assessed the integrity of the ALPM in tmem88a morphants. Genes expressed throughout the ALPM include gata4, gata5 and tbx20 (Schoenebeck et al., 2007; Waxman et al., 2008). WISH revealed that the gata4 expression pattern was identical in control and tmem88a morphants (Fig. 7E,F). In particular, the posterior boundary of the gata4⁺ domains was located in the same position, as determined by measuring the distance from this boundary to the notochord tip, regardless of MO injection (Fig. 7G). Moreover, the relative expression level of gata4 was unchanged in tmem88a morphants compared with controls (Fig. 7H). Given that the ALPM is unperturbed in Tmem88a-depleted embryos, the loss of nkx2.5 expression implies that these embryos have a primary defect in CP specification.

To establish whether the impaired CP specification in tmem88a morphants is due to defective cell cycle dynamics, apoptosis and cell proliferation assays were performed. Increased apoptosis was observed in the developing brain of tmem88a morphants compared with control embryos (data not shown). By contrast, no difference in apoptosis or cell proliferation was detected in the cardiogenic ALPM of control and tmem88a morphants (Fig. 8A-F).

The possibility of cell fate conversion was also considered. The heart develops in association with the hematovascular and forelimb organ fields (Schoenebeck et al., 2007; Van Handel et al., 2012; Waxman et al., 2008). During normal development, these non-cardiac domains limit the range of CP expansion. Moreover, overexpression of the hematovascular transcription factor scl (tal1 - Zebrafish Information Network) has been shown to convert precardiac mesoderm into blood and endothelium (Schoenebeck et al., 2007), indicating that fate conversion between neighboring organ fields is possible. To establish whether reduction in CP specification in tmem88a morphants is due to concomitant expansion of either hematovascular or forelimb tissues, we performed WISH for scl and tbx5. Neither the blood nor forelimb field was expanded in tmem88a morphants (Fig. 8G-J), indicating that the Tmem88a-related CP deficiency does not result from fate conversion to these lineages.

Since the expression pattern of tmem88a in 8-somite stage embryos extends in the lateral mesoderm beyond the CP fields, we examined whether this gene is implicated in the development of other mesodermal tissues. WISH for the axial mesoderm marker ntl, paraxial mesoderm marker myod (myod1 - Zebrafish Information Network), intermediate mesoderm marker pax2a and hematovascular marker scl indicated that the expression pattern of these genes was indistinguishable in tmem88a and control morphants (supplementary material Fig. S8A-H). However, the erythroid marker gata1 was essentially undetectable in 75% of tmem88a morphants (supplementary material Fig. S8I,J). Analysis of additional hematopoietic and vascular markers showed that transcripts for gata1 and the myeloid regulatory gene pu1 (spi1b - Zebrafish Information Network) are reduced in tmem88a morphants compared with control embryos (supplementary material Fig. S8K). These findings are consistent with a recent publication documenting the requirement of Tmem88a in erythromyeloid development (Cannon et al., 2013). The selective loss of gata1 and pu1 expression in 8-somite stage morphants indicates an early role for Tmem88a in erythromyeloid precursor specification and/or blood lineage commitment downstream of the stem/progenitor regulatory gene scl.

Hematopoietic and cardiac development are regulated by WNT signaling. In both cases, a precisely timed sequence of WNT activation and inhibition is required for lineage commitment (Naito et al., 2006; Ueno et al., 2007). For instance, CP specification depends on WNT activity at or around gastrulation followed by WNT inhibition at the onset of somitogenesis. Interestingly, human TMEM88 was shown to negate a WNT overexpression phenotype when expressed in Xenopus explants (Lee et al., 2010). To establish whether zebrafish Tmem88a negatively regulates WNT activity during CP specification, we monitored the relative expression of a genetically engineered WNT sensor in response to Tmem88a expression. Using transgenic TOPdGFP reporter fish, WNT activity (GFP) was found to be elevated in tmem88a morphants and reduced in embryos injected with tmem88a mRNA, compared with controls (Fig. 9A,B). Similarly, the relative expression of the WNT target gene dkk1 increased in tmem88a morphants and decreased in embryos injected with tmem88a mRNA, compared with controls (Fig. 9C,D). By contrast, control morphants and embryos injected with mutated tmem88a mRNA, which encodes a truncated protein lacking the C-terminal Val-Trp-Val motif, were unchanged for WNT-related gene expression. Therefore, active WNT signaling negatively correlates with Tmem88a levels during early somitogenesis.

To evaluate whether the correlation between Tmem88a levels and WNT activity is significant to cardiogenesis, functional interactions were tested in the context of CP specification. Since CP specification requires the sequential activation and inhibition of WNT signaling, inappropriate modulation of the pathway should interfere with CP formation. For example, forced inhibition of WNT signaling by dkk1 overexpression prior to 3.5 hpf reduces nkx2.5 expression in 8-somite stage embryos (Ueno et al., 2007). Similarly, forced expression of Tmem88a by microinjection of mRNA caused a significant reduction in the nkx2.5 expression pattern compared with wild-type controls (Fig. 10A,B) or control embryos injected with mutated tmem88a mRNA (Fig. 10C). By 48 hpf, embryos injected with wild-type tmem88a mRNA developed a discernable cardiomyopathy, whereas control embryos had morphologically normal hearts (supplementary material Fig. S9).

We reasoned that if the CP deficiency associated with tmem88a overexpression results from reduced levels of WNT signaling, forced WNT induction should rescue this phenotype. Rescue was attempted by injecting tmem88a mRNA into tg(hsp70l:wnt8-eGFP) embryos, which carry a heat-shock-inducible wnt8 gene. Control and injected embryos were heat shocked at 3 hpf for 30 minutes and analyzed for nkx2.5 expression at the 8-somite stage of development. Early induction of wnt8 in wild-type embryos resulted in increased nkx2.5 expression (Fig. 10D), as documented previously (Ueno et al., 2007). Interestingly, in embryos overexpressing tmem88a, forced WNT activation largely rescued the Tmem88a-induced CP deficiency (Fig. 10E,F). Since the deleterious effect of tmem88a overexpression on CP specification was reversed by increasing WNT signaling, Tmem88a and the WNT pathway appear to be antagonistic.

The tmem88a mRNA overexpression studies established antagonism between Tmem88a and WNT signaling, but did not determine the normal physiological relevance of this interaction. Expression of tmem88a is virtually undetectable until the end of gastrulation, a stage that coincides with procardiogenic inhibition of WNT activity. To establish whether tmem88a contributes to this inhibitory regulation, a combined MO and genetic approach was adopted. We reasoned that if Tmem88a is required for CP specification because it represses WNT activity, then increasing WNT signaling should sensitize embryos to the depletion of Tmem88a. To test this hypothesis, embryos derived from the tg(hsp70l:wnt8-eGFP) line were heat shocked for 30-60 minutes at 75% epiboly. This developmental stage was chosen because it follows mesoderm establishment but precedes the emergence of nkx2.5⁺ CPs. A 60 minute heat-shock treatment caused significant depletion of nkx2.5 in 8-somite stage control morphants, as previously reported (Ueno et al., 2007). By contrast, embryos treated for 30 minutes appeared identical to non-heat-shocked controls (Fig. 11A,B). Partial depletion of Tmem88a was accomplished using subthreshold quantities of tmem88a MO (1 ng). These embryos had only a moderate (25%) reduction in nkx2.5 staining (Fig. 11C). A 30 minute heat-shock treatment of the partial morphants caused a profound (60-70%) deficiency in the nkx2.5 expression domains, which was much more than the additive effect of the individual manipulations (Fig. 11D). These findings were quantified by measuring the area of nkx2.5 expression under each condition, and using qPCR to determine nkx2.5 relative expression levels (Fig. 11E,F). The synergistic repression of CP formation by wnt8 overexpression and tmem88a depletion is consistent with a functional interaction between WNT signaling and Tmem88a activity.

If loss of Tmem88a leads to enhanced WNT activity that blocks CP specification, then inhibition of WNT signaling just prior to specification should restore CP generation in the absence of Tmem88a. To test this idea, WNT signaling was downregulated using the tg(hsp70l:dkk1-eGFP) line to express the WNT inhibitor dkk1 in a temporally specific manner. Control or tmem88a morphant embryos were heat shocked for 1 hour at 75% epiboly and CP formation was monitored by WISH for nkx2.5. In control embryos, WNT inhibition at this stage had no effect on the generation of nascent CPs (Fig. 12A,B). By contrast, the CP domains were rescued in 80% of the heat-shocked tmem88a morphants (Fig. 12C,D). These embryos had, on average, a 42% expansion of the nkx2.5⁺ territories compared with non-heat-shocked tmem88a morphants, as quantified by area measurements and nkx2.5 relative expression levels (Fig. 12E; supplementary material Fig. S10).

In addition to rescuing CP specification, dkk1 overexpression resulted in complete recovery of total cardiomyocyte number in tmem88a morphants. Heat-shocked and non-heat-shocked control embryos had equivalent numbers of cardiomyocytes at 48 hpf (Fig. 12F,G). The 46% cardiomyocyte deficiency measured in tmem88a morphants was rescued upon dkk1 overexpression (Fig. 12H-J). Cardiac morphology remained abnormal in these embryos, suggesting a separate or fine-tuned role for Tmem88a in morphogenesis. However, the rescue by dkk1 expression was sufficient to cause a marked overall improvement in the gross craniocardiac morphology of the tmem88a morphants (Fig. 12K-N).

The WNT-dependent reversal of CP and cardiomyocyte loss in tmem88a morphants further supports a model whereby Tmem88a negatively modulates WNT signaling to create a permissive environment for cardiac progenitor specification, following an earlier WNT-dependent phase of cardiac mesoderm commitment.

Transcription factors and signaling pathways are important components of the cardiogenic network that drives CP specification, but the molecular details of the cross-regulation and balance among cues to ensure normal numbers of CPs remain largely unknown. WNT activation has previously been associated with reduced GATA factor expression (Afouda et al., 2008; Martin et al., 2010). In this study, we implicate GATA factors in limiting WNT signaling. We identify tmem88a as a GATA-dependent gene that is required for the specification of normal numbers of cardiomyocytes. Either excessive or reduced tmem88a expression results in the disruption of nkx2.5⁺ CP field size. Timed activation or inhibition of the WNT pathway rescues the tmem88a overexpression or loss-of-function phenotype, respectively. These observations provide strong evidence for a functional interaction between Tmem88a and the WNT pathway that is responsible, at least in part, for appropriate CP specification.

Given the major effect of tmem88a knockdown on CP specification, as measured by nkx2.5 expression, it is somewhat surprising that later indicators of cardiac development, including differentiation markers and cardiomyocyte numbers, although altered significantly, were not more affected in tmem88a morphants. Since our morphant model is not null, one caveat is the transient nature of the tmem88a MO and potential recovery of Tmem88a levels. However, given the limited developmental window of action, this explanation seems unlikely. Another possibility is that neighboring cell populations compensate for the loss of nkx2.5⁺ CPs in the developing morphants. Lineage-tracing studies have shown that gata4⁺ nkx2.5^- head mesenchyme progenitors, positioned anterior to the nkx2.5⁺ CPs, can replace ablated CPs in 18-somite stage or younger embryos (Serbedzija et al., 1998). Since the gata4⁺ ALPM is intact in tmem88a morphants, these head mesenchyme progenitors might compensate and partially rescue the post-specification stages of cardiogenesis. Since tissue-specific markers for these cells are not available, the size of the head mesenchyme progenitor field cannot be quantified using WISH, and lineage-tracing experiments will be needed to test this hypothesis. Another unanswered question relates to the fate, in tmem88a morphants, of the gata4⁺ cells that are normally cardiogenic. We did not observe increased cell death or decreased cell proliferation in or near the nkx2.5⁺ ALPM of tmem88a morphants. Nor was there any expansion in the hematovascular or forelimb fields that are normally adjacent to the precardiac mesoderm. Again, with the development of appropriate reporter lines, lineage-tracing experiments might help to identify the fate of the cells that fail to express nkx2.5 in embryos lacking Tmem88a function.

Tmem88a is the second small transmembrane protein of unknown function (TMEM) to be implicated in cardiogenesis. Recent forward genetic screens identified Tmem2 as a protein involved in cardiomyocyte migration and atrioventricular canal formation (Smith et al., 2011; Totong et al., 2011). Tmem2 was shown to act by negatively modulating BMP signaling. Although Tmem2 and Tmem88a share no homology, they are both transmembrane proteins that inhibit important signaling pathways, most likely by sequestering key pathway components at the plasma membrane. Our findings suggest that Tmem88a is much less effective at WNT repression than Dkk1, a soluble WNT inhibitor. The effect of Tmem2 on BMP activation also seems moderate (Smith et al., 2011; Totong et al., 2011). TMEM proteins might therefore constitute a novel regulatory layer that is responsible for the fine-tuning of major signaling cascades in cardiogenesis.

Although we identified tmem88a as a regulator of CP specification, this gene is also involved in the development of erythromyeloid tissue. Cannon et al. recently showed that erythroid and myeloid cell numbers are significantly decreased in tmem88a morphants at 48 hpf, compared with wild-type embryos (Cannon et al., 2013). We observed that 8-somite stage tmem88a morphants had markedly reduced expression levels of the erythroid progenitor marker gata1 and the myeloid progenitor marker pu1. Thus, Tmem88a is likely to be required for the specification or differentiation of a common erythromyeloid precursor cell. Interestingly, dkk1 overexpression did not rescue the blood deficiency in tmem88a morphants, and so the mechanism of action of Tmem88a in hematopoietic development remains unclear.

We thank Timothy Hla for access to the LSM510 confocal microscope and the staff at the WCMC Optical Microscopy Core for imaging support; Wolfram Goessling for providing the cloche mutants and the tg(hsp70l:wnt8-eGFP), tg(hsp70l:dkk-eGFP) and TOPdGFP transgenic lines; Yariv Houvras for providing the tp53^M214K mutant line; Fabien Campagne and Christopher Mason for advice during the next generation sequencing screen; members of the Caroline Burns and Wolfram Goessling laboratories for invaluable technical advice; Yariv Houvras and members of the T.E. laboratory for thoughtful critiques of the work; and the Weill Cornell/Rockefeller/Sloan-Kettering MD-PhD Program for support and mentorship.