Transmembrane protein 2 (Tmem2) is required to regionally restrict atrioventricular canal boundary and endocardial cushion development

During mammalian embryonic development, the primitive linear heart tube undergoes dynamic morphogenesis to give rise to the looped heart. This process begins as an asymmetric positioning of the heart tube followed by subsequent ballooning of the developing chambers and constriction of the atrioventricular canal (AVC). The AVC partitions the atrial and ventricular chambers and gives rise to the endocardial cushions (ECs), which later contribute to the developing valves and septa of the heart. The AVC remains in an undifferentiated state, in contrast to the differentiating chamber myocardium, and employs a defined genetic programme to establish and maintain this spatially discrete region (Moorman and Christoffels, 2003). Developmental defects arising from improper AVC formation give rise to atrioventricular septal defects. These malformations constitute the largest subset of human congenital heart malformations and can have devastating consequences for the affected individual (Lagendijk et al., 2010). Identifying and characterising the molecular regulators of AVC development are crucial to understanding and managing this group of diseases.

For more than a decade, the zebrafish has successfully been used as a model with which to identify regulators of cardiac morphogenesis, including the AVC. Mutants for genes such as ugdh (jekyll) (Walsh and Stainier, 2001), notch1 (Timmerman et al., 2004) and foxn4 (slipjig) (Chi et al., 2008) exhibit a loss or diminution of EC development. By contrast, mutations in apc (Hurlstone et al., 2003), tbx5 (heartstrings) (Camarata et al., 2010; Garrity et al., 2002) or hey2 (gridlock) (Rutenberg et al., 2006) result in an expansion of the AVC. In this latter category of mutants, the AVC-chamber boundary is compromised. Although the AVC defects in these mutants are phenotypically and molecularly diverse, each case is accompanied by cardiac looping defects, suggesting that appropriate patterning of the AVC is essential for correct looping and morphogenesis of the heart.

Here, we describe a novel zebrafish mutant, wickham (wkm), identified in a forward genetic screen by a defect in cardiac looping. We show that the wkm mutant harbours a mutation in tmem2, a previously undescribed regulator of cardiac development that is expressed in the zebrafish and mammalian heart.

Fish were kept under standard conditions as previously described (Westerfield, 1995). The tg(myl7:dsRed) and tg(kdrl:gfp) lines were described previously (Huang et al., 2003; Lawson and Weinstein, 2002; Mably et al., 2003). Pregnant wild-type CD1 mice were sacrificed and embryos dissected at the required stages.

ENU mutagenesis was performed as previously described (Wienholds et al., 2002). F1 progeny from mutagenised Tuebingen Longfin males were crossed to an AB strain to produce F2 families. Subsequent incrossing of F2 progeny generated F3 embryos that were screened for cardiac morphogenesis defects. All animal work conformed to ethical guidelines and was approved by the relevant animal ethics committees at the Hubrecht Institute and the University of Queensland.

The tmem2^hu4800 allele was mapped using standard meiotic mapping with simple sequence length polymorphisms (SSLPs). Unique SSLP primer sequences for the region (CA1-CA4) are listed in Table S1 in the supplementary material. Linkage for the tmem2^hu5935 allele was also performed using SSLP markers. Subsequent genotyping of the tmem2^hu5935 allele was performed by sequencing using the primer pair 5′-ATGTCTTGGACCTTCCTCAC-3′ (forward) and 5′-AGAACAGAGTATAAAGCCCTCTG-3′ (reverse). The genomic region was analysed with reference to Ensembl assembly zv8, release 55, December 2008.

A 1500 bp fragment of zebrafish spp1 (Image clone 5602411, Imagenes) was cloned into pBluescript+ (Stratagene) for RNA probe synthesis. Morpholino oligonucleotides (MOs) (Gene Tools) were dissolved in water and injected at 1 nl per embryo. The tmem2 MO was targeted to the exon 5 donor site with the sequence 5′-ACAAACCAAAGCCATCTCACCTTGA-3′. The splice-targeting bmp4 MO was as previously described (Chocron et al., 2007). The standard p53 control MO from Gene Tools was used.

ISH analysis of zebrafish was carried out as previously described (Smith et al., 2009). For sectioning, embryos were mounted in Technovit 8100 (Hereaus Kulzer) and sectioned at 7 μm. Whole-mount ISH analysis of mouse embryos was as described previously (Fowles et al., 2003). Immunohistochemistry was performed as previously described (Smith et al., 2008). Hyaluronic acid staining was performed on paraffin sections as described previously (Bakkers et al., 2004).

BAC recombineering was performed as previously described (Hogan et al., 2009), integrating the Cherry coding sequence in place of the tmem2 stop codon in CHORI BAC clone CH73-204I21. Primer sequences are detailed in Table S1 in the supplementary material.

Heartbeats of 15 (alternating) sibling and mutant embryos were counted over a 1-minute period under a dissecting microscope.

Confocal imaging was performed using Leica SP2 and Zeiss 510 confocal laser-scanning microscopes with 40× magnification. Embryos were mounted in ProLong Gold (Invitrogen). Imaris (Bitplane) (Figs 1, 2) and Volocity (Improvision) (Figs 3, 4) software were used for image analysis. ImageJ software was employed to remove myocardial Alcama expression in Figs 3 and 4. Statistical analyses were performed in Excel (Microsoft).

To identify genetic regulators of cardiac morphogenesis and EC development, a forward genetic screen for zebrafish mutants with defective heart looping was performed. Embryos were screened at 26-32 hours post-fertilisation (hpf) to ensure normal formation and asymmetry of the cardiac tube, and at 48-54 hpf, to identify mutants with defective looping. Two alleles of the recessive lethal wkm (wkm^hu4800 and wkm^hu5935) mutation were identified that resulted in mutants that were indistinguishable from siblings at 26 hpf but displayed an unlooped heart at 48 hpf (Fig. 1A-C′). In heterozygous crosses the alleles failed to complement (27% affected transheterozygotes), confirming that these two mutations are allelic (Fig. 1I,I′). The gross morphology of wkm mutants was relatively normal, as compared with siblings. The cardiac looping defect was accompanied by cardiac oedema and an accumulation of blood above the ventral tail fin. By 4 days post-fertilisation (dpf), massive pericardial oedema was observed (data not shown) and, from 5 dpf onward, lethality resulted.

The morphology of the heart was examined by in situ hybridisation (ISH) analysis. A reduction in the S-loop shape of the heart accompanied by diminished constriction of the AVC were observed in both whole-mount ISH and frontally sectioned embryos stained for cardiac myosin light chain 2 (cmlc2; also known as myl7) and atrial myosin heavy chain (amhc; also known as myh6) (Fig. 1C-D′). Chamber myocardium was specified normally in wkm mutants as determined by ventricular myosin heavy chain (vmhc) and amhc expression (Fig. 1E-F′). Pacemaker activity was also unchanged in wkm mutants, when beats per minute at 50 hpf were measured in mutant and sibling embryos (see Fig. S1A in the supplementary material). At later stages (>56 hpf), heart rate and cardiac contractility became impaired (data not shown); however, this occurred subsequent to the looping defect and was, therefore, secondary to the morphogenetic defect. No alteration in overall myocardial cell number was observed at 50 hpf in wkm mutants as compared with sibling embryos (see Fig. S1B in the supplementary material).

The overall morphology of wkm hearts was examined by confocal imaging of embryos immunostained for Alcama [also known as Dm-grasp; which labels the myocardium and ECs (Beis et al., 2005)] on the transgenic background kdrl:GFP (GFP expressed in all endocardial cells). The myocardium and endocardium were intact in the wkm heart, forming distinct layers. The presence of the endocardium was further confirmed by ISH for nfatc1 (see Fig. S2 in the supplementary material), and the overall laterality of the embryo was normal, as determined by the direction of gut looping (see Fig. S2 in the supplementary material).

Taken together, these data indicate that myocardial patterning, cell number and function and endocardial specification were unaffected in wkm mutants at 50 hpf, suggesting that the cardiac morphogenesis defect was highly specific.

To determine the gene responsible for the wkm phenotype, positional cloning using traditional genetic mapping with SSLP markers was performed. The mutation responsible for one of the wkm alleles (wkm^hu4800) was localised to a genetic interval of 0.3 Mb on chromosome 5 containing five genes (Fig. 1H); however, sequencing of the coding regions failed to identify any causative mutations. For the second allele (wkm^hu5935), linkage on chromosome 5 was confirmed. Sequencing the coding regions of all five genes in the genetic interval for wkm^hu5935 revealed an A-to-T transition at position 985 in the third exon of the 23-exon gene transmembrane protein 2 (tmem2; accession NM_001099449). This mutation is predicted to introduce a stop codon (K329X) leading to a prematurely truncated protein (Fig. 1J). To confirm that the lesion in tmem2 is responsible for the wkm phenotype, morpholino (MO) knockdown of tmem2 was performed and robustly reproduced the wkm cardiac phenotype (Fig. 1K). Thus, genetic linkage, the identification of a premature stop codon and the MO phenocopy confirm that disruption of the tmem2 gene results in the wkm phenotype (see Fig. S3 in the supplementary material).

The 4173 nt tmem2 transcript is predicted to encode a 1390 amino acid protein that harbours a single transmembrane domain in close proximity to the N-terminus, and a G8 domain, which is hypothesised to be involved in ligand binding and catalysis (Fig. 1L) (He et al., 2006). The long C-terminal region of the protein is predicted to reside extracellularly (Hogan et al., 2003). Until now, no functional role had been ascribed to Tmem2.

tmem2 expression was examined in zebrafish and found to be present at the 2-cell stage (data not shown), indicating maternal deposition, and at the 22-somite stage, when it is ubiquitously expressed (Fig. 2A). By 48 hours, expression is restricted to the brain and head mesenchyme, otic vesicles, fin buds, caudal vein and the heart (Fig. 2B). Within the heart, tmem2-positive regions included the outflow tract, ventricular regions and outer curvature of the atrium (Fig. 2C). Sectioned hearts counterstained with tropomyosin (to label the myocardium) revealed both myocardial and endocardial expression (Fig. 2D).

To expand on these analyses, a recombineered bacterial artificial chromosome (BAC) containing the tmem2 locus, with an in-frame insertion of the cherry coding sequence replacing the tmem2 stop codon (resulting in a fusion of Cherry to the Tmem2 C-terminus), was created. Injection of BAC DNA into 1-cell stage zebrafish embryos resulted in mosaic expression of Tmem2-Cherry protein. In the developing heart, fusion protein expression localised to both myocardial and endocardial cells at 48 hpf, consistent with tmem2 RNA expression at the same stage (Fig. 2E,F). These approaches confirm that tmem2 is transcribed and expressed in both endocardium and myocardium during morphogenesis of the zebrafish heart.

To investigate possible conservation of function in mammals, we investigated expression of Tmem2 by whole-mount ISH analysis in mouse embryos. At early stages of development, Tmem2 showed ubiquitous expression (Fig. 2G,H), although enriched, punctate staining was evident in the developing heart, particularly from E9.5 (Fig. 2J). Sagittal sections of stained E9.5 embryos revealed that expression was restricted to endocardial cells lining the ventricles and atria (Fig. 2K,K′). In an analogous manner to zebrafish tmem2, Tmem2 expression became progressively restricted at later stages of mouse development (E11.0; Fig. 2I).

Although the morphology of zebrafish and mammalian hearts ultimately differ, many of the underlying molecular and cellular patterning events are conserved. The finding that Tmem2 has enriched expression in the mouse heart suggests a potential conservation of function.

The tbx5 mutant has been shown to exhibit a specific cardiac looping defect that is similar to wkm mutants (Garrity et al., 2002). To determine whether the looping defect observed in wkm mutants is related to the tbx5 genetic pathway, reciprocal expression analysis was performed on both mutants. No obvious difference in tbx5 expression was observed in wkm mutants as compared with sibling embryos (see Fig. S5A,A′ in the supplementary material). Likewise, no difference in tmem2 expression was observed in tbx5 mutant embryos (see Fig. S5B,B′ in the supplementary material), suggesting that these genes operate in distinct genetic pathways to affect cardiac looping.

To more clearly define the cardiac defect in wkm mutants, the expression of a number of regionalised cardiac markers was examined. The myocardial chamber marker anf (also known as nppa) was expressed in the chambers but excluded from the AVC of wkm mutants (Fig. 3A,A′), although the area of anf-negative myocardial cells appeared to be broader than in siblings. Strikingly, the myocardial AVC markers bmp4 and tbx2b were upregulated and expanded, with bmp4 expression notably expanded throughout the entire ventricle in wkm mutants (Fig. 3B-C′). As with markers of the AVC myocardium, the AVC endocardial marker hyaluronan synthase 2 (has2) was expanded into the chambers (Fig. 3D,D′). Concordantly, an increase in hyaluronic acid (HA) in the chambers was also observed (see Fig. S5A,A′ in the supplementary material). notch1b expression in the AVC was less condensed but otherwise unchanged in mutants (see Fig. S5B,B′ in the supplementary material); however, expression of the Notch target hey2 was expanded (see Fig. S5C,C′ in the supplementary material). Immunofluorescence analysis of the EC marker Alcama also showed a dramatic expansion in ECs (Fig. 3E,E′), confirming an expansion of the EC region in wkm mutants. Intriguingly, whereas the suite of early AVC developmental markers showed expansion in their expression domains, spp1 (also known as osteopontin or zgc:111821) showed a substantial reduction in expression (Fig. 3F,F′). spp1 marks the developing cardiac valves at later stages of formation (Peal et al., 2009), suggesting that, although the EC primordium is expanded, there is a failure to progress to valve maturation.

Bmp signalling is essential for the formation of the AVC (van Wijk et al., 2007). bmp2/4 have restricted expression in the AVC and outflow tract and have been demonstrated to promote cardiac jelly formation, epithelial-to-mesenchymal transition and patterning of the AVC (Ma et al., 2005). Given this role, we hypothesised that the ectopic expression of bmp4 in wkm mutants might be responsible for the expanded AVC phenotype. To test this, Bmp4 was depleted in mutant and sibling zebrafish embryos by MO knockdown and ECs examined. has2 expression was unchanged in sibling and mutant embryos injected with control p53 MO in that an expansion of has2 was evident (Fig. 4A). By contrast, whereas injection of bmp4 MO had no effect on has2 expression in sibling embryos, rescue of the ectopic expression of has2 in wkm mutants was observed. A similar rescue was observed for Alcama staining in ECs (Fig. 4A), indicating that the expanded EC phenotype in wkm mutants is Bmp4 dependent. It is noteworthy that the AVC expression of has2 and Alcama are not abolished in bmp4 morphants, suggesting that this region is not under the sole regulation of Bmp4.

The expanded AVC phenotype of wkm mutants is phenotypically indistinguishable from the defect seen in zebrafish tbx5, apc and hey2 mutants. In these cases, an expansion of bmp4 expression throughout the ventricular myocardium is observed and is associated with EC marker expansion (Camarata et al., 2010; Garrity et al., 2002; Hurlstone et al., 2003; Rutenberg et al., 2006). Here, we show that, downstream of tmem2, ectopic bmp4 expression is functionally responsible for the expansion of the immature AVC. Bmp signalling is known to activate tbx2 expression in the AVC (Singh et al., 2009; Yamada et al., 2000) and, in turn, tbx2 inhibits the chamber myocardial differentiation programme via the repression of genes such as anf (Habets et al., 2002). The boundary between the AVC and chambers is kept in check by chamber-specific genes, such as Hesr1 and Hesr2 (Kokubo et al., 2007). Clearly, tbx5, apc, hey2 and, now, tmem2 (Fig. 4B) are instrumental in determining this AVC boundary and regionally restricting EC development.

We thank the Hubrecht Screen Team and Jeroen Korving for technical assistance and Dr David Pennisi for useful discussions. Confocal microscopy was performed at the ACRF Cancer Biology Imaging Facility at the IMB, established with the generous support of the Australian Cancer Research Foundation (ACRF).