Human cardiomyopathy mutations induce myocyte hyperplasia and activate hypertrophic pathways during cardiogenesis in zebrafish

In the last two decades, genetic studies of inherited forms of human cardiomyopathy have offered important insights into the fundamental biology of myocardial hypertrophy and heart failure. Many of the causal genes for inherited hypertrophic cardiomyopathy (HCM) encode for proteins involved in the proper function of the cardiac sarcomere. However, substantial pleiotropy is seen for many cardiomyopathy genes. Even within families harboring the same mutation there is significant variation in the extent of hypertrophy, the risk of arrhythmias or the progression to heart failure (Alcalai et al., 2007; Arad et al., 2002). Furthermore, different mutations in a single gene can lead to either HCM or dilated cardiomyopathy (DCM) (Seidman and Seidman, 2001). The major modifiers regulating this phenotypic diversity have not been defined.

Evidence is emerging of significant overlap between the pathways regulating cell division and those driving cardiac hypertrophy (Ahuja et al., 2007). Nuclear division and polyploidy are observed in the later stages of human heart failure and resolve with ventricular unloading, but cellular dynamics earlier in hypertrophy have not yet been explored (Rivello et al., 2001). The limited capacity for cell division in adult mammalian cardiomyocytes leads to the chronic activation of cell division pathways; these same pathways are thought to drive the changes in cardiomyocyte differentiation seen in the hypertrophic process. The effects of manipulating cell division pathways on myocardial growth are diverse and might involve epigenetic mechanisms with complex inheritance patterns (Wagner et al., 2008).

In an effort to define the earliest elements of the response to hypertrophic stimuli, we have generated a zebrafish model of a human cardiac troponin T (TNNT2) mutation known to cause HCM. This model not only facilitates the study of the initial developmental effects of a primary cause of hypertrophic cardiomyopathy, but also allows the systematic study of genetic and environmental modifiers before the effects of chronic myocardial remodeling supervene.

In order to precisely recapitulate a human TNNT2 disease mutation in the zebrafish, we designed morpholino antisense oligonucleotides (oligos) that target the intron 13 splice donor site in the zebrafish cardiac troponin T (tnnt2) mRNA. The resultant sequence is orthologous to an established human TNNT2 splice mutant and its expression is importantly under the control of the native zebrafish tnnt2 promoter. Morphant hearts exhibited disrupted sarcomere formation and showed induction of a pathological transcriptional response during the earliest stages of cardiogenesis. In addition, in these embryonic hearts, we confirmed that Tnnt2 mutations lead to significant myocardial hyperplasia. We also demonstrate that this Tnnt2 mutation perturbs normal cardiomyocyte Ca²⁺ handling, suggesting that the arrhythmic risk in the context of sarcomeric protein mutations does not simply reflect the consequences of cellular disarray and scarring. Lastly, once the genetic stimulus is eliminated, sarcomerogenesis normalizes. Taken together, these data highlight the distinctive initial responses of the embryonic heart to a primary hypertrophic stimulus, despite evidence of substantial sharing of sarcomeric biology and transcriptional pathways with adult myocardium.

In order to precisely recapitulate an autosomal dominant hypertrophic cardiomyopathy mutation, we chose to model a mutation in the splice donor site of exon 15 in human TNNT2 (Thierfelder et al., 1994). The mutant TNNT2 mRNA splice products either exclude exon 15 or use a cryptic splice site that changes the reading frame and leads to a premature stop codon (Fig. 1A). We used a morpholino (TNNT2sp MO) to target the splice donor site in zebrafish tnnt2 exon 13, which is the ortholog of human TNNT2 exon 15 (detailed sequence data for zebrafish tnnt2 in supplementary material Fig. S1). The resulting morphant splice product excludes zebrafish tnnt2 exon 13 (Fig. 1B), causing a disruption in the C-terminus of the zebrafish Tnnt2 protein at the identical position to that mutated in humans (Fig. 1C). This morpholino allows the creation of a dominant mutation in Tnnt2 while retaining the full control of the native zebrafish tnnt2 promoter. In an effort to model the nature of the human disease as precisely as possible, the TNNT2sp MO was dosed so that the wild-type tnnt2 transcript was reduced by approximately 50% (Fig. 1B).

There was no evidence of non-cardiac or off-target morpholino effects in the TNNT2sp morphants and the embryos developed at a normal rate (Fig. 2A). Gross morphological examination of the morphants showed a smaller ventricle, dilated atria and pericardial edema compared with controls (Fig. 2B). Measurement of the ventricular internal chamber dimensions at end diastole [end-diastolic diameter (EDD; largest ventricular diameter)] and end systole [end-systolic diameter (ESD; smallest ventricular diameter)] confirmed substantial reductions in EDD in morphants compared with control embryos (control EDD 55±1.1 μm; TNNT2sp MO EDD 31±1.6 μm; P<0.005, n=5). ESD was also significantly decreased but to a lesser extent (control ESD 25±1.4 μm; TNNT2sp MO ESD 21±0.68 μm; P<0.04, n=5) (Fig. 2C). TNNT2sp morphants also exhibited significant reductions in ventricular fractional shortening (FS) (control FS 0.54; TNNT2sp MO FS 0.33; P<0.005, n=5) (Fig. 2D). Videos of control and TNNT2sp morphant hearts are included in supplementary material Movies 1 (control) and 2 (TNNT2sp morphant) (both at 72 hpf).

Mutant sarcomeric proteins frequently perturb sarcomere assembly. Therefore, we performed electron microscopy of control and TNNT2sp morphants to explore the ultrastructure of the embryonic heart. There was marked sarcomeric disarray in the TNNT2sp morphants, which was not present in control embryos at 96 hpf (Fig. 3A). The altered mRNA splicing induced by injection of the TNNT2sp MO is temporary and the morpholino effect is gradually diluted through cell division during development. We tested 8- and 21-day-old embryos that had recovered from the initial morpholino injection to determine whether disruption of early sarcomere structure persists despite the elimination of the primary genetic stimulus. Although, sarcomere disarray was apparent in 8-dpf embryos, by 21 dpf there were no ultrastructural differences observed between control- and TNNT2sp-MO-injected embryos (Fig. 3A).

Sarcomeric mutations are a stimulus for cardiomyocyte hypertrophy, which, in adult mammalian cardiomyocytes, seems to occur in isolation with no evidence of cardiomyocyte division (Ahuja et al., 2007). The effects of typical hypertrophic stimuli on embryonic cardiomyocytes remain unclear. We quantified the total number of cardiomyocytes in control and TNNT2sp morphant embryos using a cmlc2::DsRed-nuc transgenic zebrafish reporter line [red fluorescent protein (DsRed-nuc) expressed under the control of the cardiac myosin light chain-2 (cmlc-2) promoter] that fluorescently labels the nuclei of differentiated cardiomyocytes (Mably et al., 2003). These embryos were injected with TNNT2sp or control morpholino and cardiomyocyte nuclei were counted at 48 and 96 hpf. The nuclei numbers were similar at 48 hpf (control 197±22, n=4; TNNT2sp MO 204±14, n=5; P=NS) but, at 96 hpf, the TNNT2sp morphant hearts had approximately 26% more DsRed-positive nuclei than control hearts (control 247±7, n=4; TNNT2sp MO 311±14, n=3; P<0.03) (Fig. 3B). There was no evidence of bi-nucleated cardiomyocytes; therefore, total nuclear count correlates with total cardiomyocyte count. Ventricular cardiomyocyte cell size was measured to determine whether morphant cardiomyocytes underwent cellular hypertrophy secondary to the presence of mutant tnnt2 splice product. Morphant ventricular cardiomyocytes did not undergo hypertrophy but were actually slightly smaller than control cardiomyocytes (control 109±6 μm², n=15; TNNT2sp MO 94±4 μm², n=11; P<0.04) (Fig. 3C).

Abnormal cardiomyocyte Ca²⁺ handling is thought to be a stimulus for arrhythmias in patients with sarcomeric gene mutations (Baudenbacher et al., 2008; Knollmann et al., 2003). We performed high-resolution Ca²⁺ imaging in order to determine the effects of sarcomeric mutation on cardiomyocyte Ca²⁺ handling during early heart development. The most striking difference between TNNT2sp morphant and control embryos was the shortening of the Ca²⁺ transient duration (CTD50) in the TNNT2sp morphants. This shortening was observed in both atrium and ventricle (Fig. 4C) (supplementary material Table S1). Zebrafish null for Tnnt2 (TNNT2atg) were also analyzed to determine whether the reduction in CTD50 was a distinctive feature of the mutant RNA resulting from the TNNT2sp MO or was a result of loss or absence of wild-type Tnnt2 protein. TNNT2atg hearts, which never contract, exhibited a significant reduction of CTD50 in the atrioventricular canal, but did not develop significant reductions of CTD50 in the atrium or ventricle. In contrast to changes seen in CTD50, diastolic Ca²⁺ levels and Ca²⁺ transient amplitudes in TNNT2sp morphants were not significantly different to control hearts across multiple locations (Fig. 4A,B; supplementary material Fig. S2 and Table S1). However, TNNT2atg hearts showed significant changes compared with controls in these two parameters in multiple different locations (Fig. 4A,B; supplementary material Table S1). These data suggest that the splice mutant Tnnt2 protein results in an early ‘gain-of-function’ effect on embryonic cardiomyocyte Ca²⁺ handling, quite distinct from the effects of null tnnt2 alleles (TNNT2atg).

The typical gene expression signature induced in the setting of sarcomeric dysfunction in adult animals is referred to as reactivation of the ‘fetal gene program’. This group of genes includes the cardiac natriuretic peptides NPPA and NPPB, as well as isoform switching of myosin heavy chain genes. We were interested to determine whether these transcriptional pathways could be pathologically induced during cardiac development. We performed a microarray analysis of gene expression in TNNT2sp- and control-MO-injected morphants. Interestingly, we saw a significant induction of the zebrafish ortholog of NPPA (average 2.9-fold increase; Q value<0.03). A direct comparison of myosin heavy chain isoform changes from zebrafish to mammals cannot be made because, to date, no zebrafish ortholog of myh7 (beta-myosin heavy chain) has been identified. Two cardiac myosin heavy chain genes have been characterized in the zebrafish, myh6 (also known as atrial myosin heavy chain) and vmhc (ventricular myosin heavy chain). Zebrafish Myh6 and Vmhc proteins are more closely related to mouse Myh6 (α-myosin heavy chain) than mouse Myh7 (NCBI, Blast Protein Alignment). Notably, zebrafish myh6 gene expression was increased 1.73-fold in the TNNT2sp morphant embryos, whereas vmhc expression was not significantly changed (Fig. 5A, black bar; Q value=0.11). The other sarcomeric genes evaluated on the microarray are presented for comparison. Of note, one probe specifically bound to exon 13 of zebrafish tnnt2, allowing us to independently assess the efficacy of our TNNT2sp MO. The TNNT2sp morphants had a significant reduction in wild-type tnnt2 mRNA compared with embryos injected with a control morpholino [Fig. 5A, white bar labeled ‘tnnt2 (exon 13)’; Q value=0.006]. A second tnnt2 probe that bound the 3′ untranslated region of tnnt2 mRNA also showed reduced levels of expression in the TNNT2sp morphants [Fig. 5A, white bar labeled ‘tnnt2 (3′ UTR)’; Q value=0.06].

Given the significant reduction in CTD50 seen in our TNNT2sp morphants, we evaluated the list of differentially expressed genes to identify potential transcriptional differences that might explain these effects. In the morphants, we found upregulation of several genes that are known to regulate cardiomyocyte Ca²⁺ homeostasis – the cardiac sodium/calcium exchanger (slc8a1a), calsequestrin 2 (casq2), sarcalumenin (srl) and calpastatin (cast) (Harris et al., 2006; Kubalova et al., 2004; Yoshida et al., 2005) (Fig. 5B).

We next undertook formal analysis of the microarray datasets for pathway enrichment in the morphant using FuncAssociate (Berriz et al., 2003). The only genetic pathway that was significantly perturbed in TNNT2sp morphants was a group of genes classified under the gene ontology term ‘hormone activity’ (GO:0005179), all of which were coordinately downregulated. The specific genes associated with this pathway included growth hormone releasing hormone (ghrh), parathyroid hormone (pth1a), proopiomelanocortin (pomca) and urotensin2a (uts2a) (Fig. 5C).

To confirm that the microarray accurately reflects gene expression changes, we performed quantitative reverse-transcriptase PCR (qPCR) on select upregulated (>twofold increased) and downregulated (>50% decreased) genes identified on the microarray. The microarray and qPCR results show similar levels of up- or down-regulation (Fig. 5D).

Lastly, to explore the extent to which the embryonic zebrafish TNNT2sp transcriptional responses parallel those in adult mammalian HCM models, we compared our whole-embryo dataset with an HCM gene expression dataset from mouse ventricular tissue generated with Polony Multiplex Analysis of Gene Expression (P-MAGE) (Kim et al., 2007). The mouse ventricular tissue that was analyzed was harvested from alphaMHC403 heterozygote mice. These mice harbor a missense mutation in their alpha myosin heavy chain gene (myh6) that causes a change from arginine to glutamine at position 403 (Geisterfer-Lowrance et al., 1996). The expression of 122 genes was significantly changed in both our tnnt2 morphant zebrafish dataset and the mouse alphaMHC403 HCM P-MAGE dataset. Of these genes, 74 were regulated in a concordant fashion and 48 were regulated in a discordant fashion (cumulative binomial probability=0.012) (Fig. 6). Induction of the natriuretic peptide gene nppa and multiple genes encoding Ca²⁺ regulatory proteins was seen in both datasets (casq2, cast and srl). In addition, there was concordant downregulation of a number of genes important to cardiovascular biology (abcc9, foxo1 and hey1).

Our embryonic zebrafish model of a human sarcomeric gene mutation recapitulated many of the cellular and transcriptional responses seen in adult organisms with these mutations. In the context of a hypertrophic stimulus, even the initial embryonic cardiomyocytes exhibited sarcomeric disarray, a hallmark of sarcomeric gene mutations in humans and mammalian models. These cellular phenomena were accompanied by a robust transcriptional response closely paralleling the patterns of gene expression observed in a murine model of HCM. In addition, there was a significant perturbation of cardiomyocyte Ca²⁺ handling evident during cardiogenesis in this zebrafish model. Despite these similarities with adult HCM models, the embryonic zebrafish heart did not display cardiomyocyte hypertrophy in response to this sarcomeric mutation, but rather developed significant cardiomyocyte hyperplasia.

The recapitulation of this human TNNT2 splicing abnormality has dramatic effects on early cardiomyocyte sarcomerogenesis, and these seem to be reversible in the embryonic heart. Interestingly, the disruption in sarcomerogenesis persisted for a considerable period (several days) after the abnormal tnnt2 splice product was eliminated but, by 21 dpf, the sarcomere had fully remodeled and normal sarcomerogenesis was evident. These observations suggest that even the profound abnormalities of sarcomeric structure seen in human cardiomyopathic cells might be reversible if the genetic mutation can be ‘silenced’ through chemical or genetic means. Nevertheless, the sarcomere recovery that we have observed in morphant embryos might reflect distinctive characteristics of the zebrafish embryonic heart, and post-natal mammalian sarcomeres might not be so plastic. Indeed, recent data in a rabbit model of HCM suggest that, although chemical modifiers might be able to reverse the interstitial fibrosis and cardiomyocyte hypertrophy associated with longstanding HCM, they do not modify myocyte disarray (Lombardi et al., 2009).

Despite the phylogenetic distance between zebrafish and mammals, there is substantial conservation of the gene expression responses to sarcomeric gene mutations. One of the most highly upregulated genes in the TNNT2sp morphant embryos is nppa. This induction of the natriuretic peptide pathway is a consistent feature of mammalian models of HCM and acquired hypertrophy. There was robust correlation of gene expression in our morphant with published cardiac gene expression analysis in an adult mouse HCM model across multiple pathways. The transcriptional effects seen in our zebrafish embryos were evident despite the potential limitations of whole-embryo expression profiling. These data highlight the fundamental nature of the hypertrophic transcriptional program, which is activated virtually from the onset of cardiac function in this zebrafish model of HCM.

One advantage of whole-embryo expression profiling is that systemic responses to cardiac-restricted stimuli can be assessed. Developing an understanding of how the whole organism responds to altered cardiac function could provide useful insights into the pathophysiology of hypertrophy and the factors determining progression to the systemic syndrome of congestive heart failure. Comprehensive pathway analyses of gene expression in our embryonic zebrafish TNNT2sp model demonstrated downregulation of a group of hormonal genes. Interestingly, these same neuroendocrine genes have been shown to modify cardiomyocyte survival (ghrh and pth1a) (Cha et al., 2009; Granata et al., 2009), directly modulate the cardiomyocyte hypertrophic response (uts2) (Tzanidis et al., 2003) or are transcriptionally regulated by hypertrophic pathways (pomca) (Chalmers et al., 2008).

Another aspect of the pleiotropy of HCM is the variable risk of cardiac arrhythmias. In animal models of HCM, arrhythmias can develop before significant fibrosis occurs, suggesting that primary alterations in cardiomyocyte electrical properties, in addition to fibrotic scarring, underlie the arrhythmogenesis (Baudenbacher et al., 2008; Wolf et al., 2005). Our TNNT2sp morphants had significant shortening of CTD50 in both atrium and ventricle. Interestingly, complete loss of Tnnt2 resulted in distinctive effects on diastolic Ca²⁺ concentrations and on the amplitude of the Ca²⁺ transient. These differences suggest that competing effects on Ca²⁺ handling result from the deletion of different domains within Tnnt2, and these non-overlapping roles might explain the disparate effects of individual troponin mutations on Ca²⁺ sensitivity in vitro (Szczesna et al., 2000). Among the potential mechanisms for the reduced CTD50 observed in the TNNT2sp morphants are effects on sarcomeric buffering of cytoplasmic Ca²⁺ or the upregulation of a number of Ca²⁺ handling proteins – slc8a1a, casq2 and srl. Increased expression of slc8a1a and the sarcoplasmic reticulum (SR) genes casq2 and srl would facilitate extrusion of Ca²⁺ from the cytoplasm to the extracellular space or SR, respectively (Harris et al., 2006; Kubalova et al., 2004; Yoshida et al., 2005). The role of cast is less clear, but it has been shown to bind and modulate the activity of Cav1.2, the cardiac L-type calcium channel (Hao et al., 2000). The precise mechanisms by which reduction in CTD50 might result in arrhythmogenesis are not known. However, abnormalities in CTD50 are seen in animal models of catecholaminergic polymorphic ventricular tachycardia (CPVT) (Song et al., 2007).

Cellular hypertrophy is a fundamental response of the stressed heart. Many of the pathways associated with cell division are known to be activated in this hypertrophic response, leading to the hypothesis that cardiomyocyte hypertrophy might be the result of a fundamental block in karyokinesis and cytokinesis, by which the adult cardiomyocyte is unable to disassemble sarcomeres, uncouple from neighboring cells and divide (Ahuja et al., 2007). Nuclear division and polyploidy are observed in the latest stages of human heart failure and reverse with ventricular unloading, but cellular dynamics earlier in the pathophysiological cascade have not yet been explored (Rivello et al., 2001). In the zebrafish model that we generated, we were able to explore the effects of disrupted sarcomere gene function during cardiogenesis. We found that when the embryonic heart is exposed to a hypertrophic stimulus, in contrast to the response in the adult mammalian heart, a hyperplastic response occurs. Whereas pressure or volume overload seem to increase cardiomyocyte proliferation in the fetal mammalian heart, the effects of such hypertrophic stimuli on the relevant molecular pathways have not been identified (Saiki et al., 1997; Sedmera et al., 2003). We were able to show that, despite the hyperplastic nature of the embryonic response to a primary hypertrophic stimulus, the classic hypertrophic transcriptional stress response networks are activated in this zebrafish model. Cross-species comparison of the distinctive molecular networks regulating cardiomyocyte hypertrophy and hyperplasia could offer insights into the factors determining these differential responses to a single stimulus.

It is interesting to compare our findings with those in a murine model with transgenic overexpression of a comparable truncated TNNT2 protein under the control of the alphaMHC(myh6) promoter (Tardiff et al., 1998). In both fish and mouse there was evidence of reduced ventricular chamber dimensions and some impairment of contractile function. In the transgenic mouse there was evidence of reduced cell number, but in the fish we observed increased cell number, despite the reduction in chamber dimensions, with a parallel reduction in cell size. Importantly, this particular transgenic mouse model did not exhibit evidence of induction of the typical transcriptional markers of the hypertrophic response, unlike other vertebrate cardiac troponin T models (Tardiff et al., 1999). It remains to be seen whether these differences between mouse and fish models result from the onset of mutant allele expression in different time windows (the alphaMHC promoter does not become active until late in cardiac development), or from intrinsic differences between murine and zebrafish cardiomyocytes.

Although zebrafish models of complete loss of TNNT2 or truncation of the TNNT2 protein exist, we designed our approach to precisely recapitulate a known human sarcomeric gene mutation while maintaining comprehensive native regulation and wild-type hemizygous gene dose. Zebrafish homozygous for a null tnnt2 mutation (silent heart), with severe truncation of Tnnt2 protein or lack of translation, develop ultrastructural characteristics of HCM, mainly sarcomeric disarray. However, these recessive zebrafish tnnt2 mutants, as well as the cognate homozygous tnnt2-null mouse, never exhibit myocardial contraction (Sehnert et al., 2002; Ahmad et al., 2008). Similarly, C-terminal truncations of zebrafish tnnt2 (exclusion of exon 12; see supplementary material Fig. S1 for exon numbering) almost completely abolish myocardial contraction and are associated with a total loss of wild-type tnnt2 transcript (Huang et al., 2009). We sought to model a dominant human tnnt2 mutation, avoiding the non-physiological elimination of contractile function and its resultant effects on cardiogenesis.

In conclusion, recapitulation of human hypertrophic cardiomyopathy gene mutations in the zebrafish facilitates access to the earliest effects of these genetic perturbations. We were able to confirm the primary defects in sarcomere assembly and establish the activation of a subset of hypertrophic pathways as well as the presence of abnormal Ca²⁺ handling during initial cardiogenesis. We also obtained definitive evidence for a hyperplastic cardiomyocyte response to a genetic hypertrophic stimulus during development. Although this might simply reflect the plastic environment of the embryo or a general property of the zebrafish heart, it will be important to re-evaluate the role of regional perturbations in cell number in the unique features of HCM, such as asymmetric septal hypertrophy. In addition, understanding the comparative biology of hyperplastic and hypertrophic responses across a range of species may inform efforts to manipulate terminally differentiated cardiomyocytes or cardiac progenitor cell populations in human disease. Ultimately, our work suggests that the pleiotropy of human hypertrophic cardiomyopathy may be influenced by selection pressures exerted during development.

The investigation conforms to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996).

Morpholinos (Gene Tools, LLC, Philomath, OR) were resuspended in 1× Danieu’s solution and 5 ng injected into fertilized eggs from TuAB fish at the single-cell stage. The tnnt2 exon 13 splice (TNNT2sp) MO sequence was: 5′-TAGACACAGATGAACT-CACAATTTC-3′. There is a polymorphism at the morpholino-binding site in exon 13 that is highlighted in red in supplementary material Fig. S1. The 5 bp mismatch control morpholino sequence was (base changes are lower case): 5′-TAcACAgAGATcAACTCA-gAATaTC-3′. To determine whether the morpholino altered splicing of the tnnt2 mRNA, we designed primers that flanked tnnt2 exon 13 and then performed RT-PCR. Primer sequences are TNNT2ex11F: 5′-AGAAGAAGATTCTCGGTGATCG-3′ and TNNT2ex14R: 5′-CAACAGTGGTCAGCTCCTCTC-3′. We sequenced the PCR product to confirm that exon 13 was completely excluded. The TNNT2atg (‘silent heart’) morpholino sequence is 5′-CATGTTTGCTCTGATCTGACACGCA-3′ (Sehnert et al., 2002).

Embryos were fixed at 96 hpf, 8 dpf and 21 dpf, embedded in Epon 812 (Polysciences) and sectioned. Thin sections were cut on a Reichert Ultracut E ultramicrotome and collected onto Formar-coated slot grids. Sections were post-stained with uranyl acetate and lead citrate, and viewed in a Philips CM10 electron microscope at 80 keV.

Embryos were laterally positioned and allowed to acclimate. Video microscopy was performed with an Axioplan (Zeiss) upright microscope and 544 frames were digitally captured at identical levels of magnification (10×) and frame rate (250 frames per second). Sequential still frames were analyzed to identify ESD and EDD. FS was calculated using the formula (EDD – ESD)/(EDD).

cmlc2::DsRed-nuc embryos were injected with control or TNNT2sp MO at the one-cell stage. At 48 or 96 hpf the embryos were euthanized and fixed in 4% PFA for 12 hours at 4°C. The embryos were transferred to PBS and stored at 4°C. Whole embryos were mounted in 1% low-melt agar and were imaged using a Zeiss LSM5 Pascal confocal microscope and 40× water-immersion lens. z-stack images were collected and a three-dimensional image projection was created using ImageJ (NIH). DsRed-positive nuclei were then counted using ImageJ software.

Control and TNNT2sp morphant embryos were euthanized and fixed in 4% PFA for 12 hours at 4°C. They were then transferred to 100% methanol and stored at −20°C. The embryos were then rehydrated in PBST. After rehydration, the embryos were washed in a blocking buffer (2 mg/ml BSA, 2% goat serum, 1% DMSO in PBST) for 2 hours at 25°C. They were then incubated with mouse anti-ZN8 (binds to ventricular cardiomyocyte cell surface) [Developmental Studies Hybridoma Bank (DSHB); 1:50] at 4°C for 12 hours, and then with Alexa Fluor 546 (Invitrogen; 1:1000) for 3 hours at 25°C. Whole embryos were mounted in 1% low-melt agar and were imaged using a Zeiss LSM5 Pascal confocal microscope and 40× water-immersion lens. ImageJ analysis software was used to measure the cell surface area of control and TNNT2sp morphant cardiomyocytes.

Hearts were isolated from zebrafish embryos at 72 hpf and placed in normal Tyrode’s solution (NT), which contained (in mM) Na⁺ (136), K⁺ (5.4), Mg²⁺ (1.0), PO₄³⁻ (0.3), Ca²⁺ (1.8), glucose (5.0) and HEPES (10.0) at pH 7.4. The chamber was mounted on the stage of an inverted microscope (TE-2000, Nikon). Excitation light was generated by a 120 W metal halide lamp (X-Cite 120, Exfo), transmitted through a 525/50 nm excitation filter and reflected onto the preparation by a 560-nm-cutoff dichroic mirror. Fluorescence emission was passed through a 635 nm long-pass emission filter and reflected by a mirror towards the camera, which was mounted at a side-port. For ratiometric Ca²⁺ transient recordings, hearts were loaded for 15 minutes with 50μM of the Ca²⁺-sensitive dye Fura-2, AM (Invitrogen). Hearts were then incubated in NT solution at room temperature for 30–45 minutes to allow complete intracellular hydrolysis of the esterified dye and placed in a perfusion bath (Warner Instruments) that contained NT solution supplemented with 30 μM of the excitation-contraction uncoupler blebbistatin (Calbiochem) to inhibit motion. A high-speed monochromator (Optoscan, Cairn Research, UK) was used to rapidly switch the excitation wavelength between 340 nm and 380 nm with a bandwidth of 20 nm and at a rate of 500 second⁻¹. Wavelength switching was synchronized with the camera to ensure that fluorescence acquisition occurred at well-defined time intervals, i.e. when the monochromator had reached the two target wavelengths. Each ratio acquisition required four frames, thus resulting in a final ratio rate of 125 second⁻¹. The excitation light was reflected by a 400-nm-cutoff dichroic mirror and fluorescence emission was collected by the camera through a 510/80 nm emission filter. For the measurement of fluorescence intensities, a high-speed 80×80-pixel CCD camera (CardioCCD-SMQ, RedShirtImaging, LLC) with 14-bit resolution was used. Using a 20×/0.75 NA objective and a 0.5× C-mount adapter, the final magnification was 10×, resulting in a pixel-to-pixel distance of 2.2 μm.

Acquired fluorescence images were exported as TIFF stacks and analyzed using Matlab (Mathworks) software. Regions of interest (ROIs) 16×16 pixels in size, covering an area of about 35×35 μm of epicardial tissue, were defined within the atrium, atrioventricular junction, and the inner curvature (IC) and the outer curvature (OC) of the ventricle. The size of the ROI was the same for all measurements. Errors resulting from variations in the positions of these ROIs within each target area (e.g. atrium, ventricle) of the same heart were less than 10% and smaller than the naturally occurring variations of Ca²⁺ concentrations between different hearts of the same genotype. Individual Ca²⁺ transients were extracted from each pixel contained within each ROI to calculate minimum (diastolic) and maximum (systolic) fluorescence ratios. Ca²⁺ transient amplitude was defined as the difference between the systolic and diastolic ratios, and transient duration (CTD50) as the time difference between the upstroke and the decay phases, measured at 50% of the amplitude. Diastolic ratios, transient amplitudes and transient durations were calculated and measurements averaged for all pixels within each ROI.

Control- and TNNT2sp-MO-injected embryos were collected at 96 hpf in groups of 40 embryos, and RNA was extracted using Trizol (Invitrogen) and further purified with RNeasy columns (Qiagen). Four biological replicates of control and TNNT2sp embryos were then analyzed using a two-color zebrafish-specific microarray (Agilent). Primary microarray data files and matrix files are available at NIH GEO repository – GSE20179.

All statistical analysis was performed using the R software package (2.9.2; www.cran.r-project.org). Agilent two-color microarray data were analyzed using the limma package. Limma employs a generalized linear model with empirical Bayesian methods to ‘borrow information’ across genes for more stable estimates of significance and magnitude of change, especially when the number of arrays is small. Unadjusted P-values for association were input into the fdrtool package for empirical tail-based false discovery rate (FDR) calculation (Strimmer, 2008). An arbitrary FDR threshold of 0.15 was selected – at this threshold, we expect that >85% of the ‘significant’ probe sets represent true positives. We have found that these relaxed thresholds combined with further pathway analysis methods can be useful for identifying enriched pathways. The FuncAssociate program was used for examining GO term enrichment (supplementary material Table S2).

A previously published hypertrophic cardiomyopathy P-MAGE data set (all tags significant at P=0.05) was downloaded for comparison of significantly changed orthologs (Kim et al., 2007). Mouse orthologs for zebrafish genes were identified using a combination of the Inparanoid, Zfin and Ensembl (www.ensembl.org) databases (Ostlund et al., 2010; Sprague et al., 2008). An additional number of orthologs were identified by matching symbol names and performing bidirectional BLASTN to ensure that the corresponding ortholog represented the top BLAST hit in each case. A total of 122 orthologs were significant at P=0.05 (P-MAGE) and the FDR was 0.15 (zebrafish array). We compared the direction of change for each of these and found 74 that demonstrated concordant change. Statistical significance was assessed as follows. Under the null model, we would expect a 50% probability for the concordance for any given pair of orthologs. The probability that 74 or more of these 122 orthologs are concordant is thus given by the cumulative binomial distribution with n=122 and P=0.5, analogous to the probability of 74 heads in 122 flips of a fair coin. To identify similarities in patterns of change for different probe sets, the limma coefficients for the 74 probe sets with concordant change in the P-MAGE and microarray data sets were depicted in a heatmap, using the heatmap function in R.

FuncAssociate uses a hypergeometric test and correction for multiple hypothesis testing to formally evaluate statistical significance for pathway enrichment (Berriz et al., 2009). Upregulated and downregulated genes were analyzed separately and ranked by magnitude of change. For some genes, multiple probe sets were available, usually with differing magnitude and significance of change. We operated under the assumption that differing effect sizes for multiple probe sets arises from some combination of chance variation in probe binding efficiencies, differential regulation of individual isoforms and measurement error. Because a single probe set had to be selected for subsequent analyses, for those genes mapping to multiple probe sets, the probe set with the lowest P-value was selected, thus making the optimistic assumption that the lowest P-value was capturing the true, maximal change in response to the perturbation. Although this assumption might introduce noise in further analyses, it should not introduce any systematic bias in terms of direction of change.

qPCR was performed using RNA extracted from whole embryos using Trizol reagent (Invitrogen) and reverse transcribed with the QuantiTect Reverse Transcription Kit (Qiagen). Gene-specific primers and SYBR Green reagent (Applied Biosystems) were used to perform the PCR portion of the experiment. The ΔΔC_t method was used to normalize the gene of interest to the endogenous housekeeping gene Rpl13alpha. Then the experimental sample was divided by the control sample to determine the level of induction. qPCR primer sequences: nppa: F 5′-GATGTACAAGCGC-ACACGTT-3′, R 5′-TCTGATGCCTCTTCTGTTGC-3′; ghrh: F 5′-AGACTTGGACCAGAAACTGAAAGC-3′, R 5′-TGTGGGT-CTCTGTAACTTTGTTCA-3′; 13alpha: F 5′-TCTGGAGGACT-GTAAGAGGTATGC-3′, R 5′-AGACGCACAATCTTGAGAG-CAG-3′.

To compare two continuous variables a Student’s t-test was used for normally distributed data. For data that were not normally distributed, a Wilcoxon rank-sum test was used. All data are expressed as mean ± s.e.m., unless otherwise noted.

We would like to thank Mary McKee for the electron microscopy technical assistance and Jennifer Love for the microarray technical assistance. We would also like to thank Michael Fifer and Barry London for their advice and encouragement. This work was supported by an institutional grant from the Hypertrophic Cardiomyopathy Center at Massachusetts General Hospital (C.A.M.) and T32HL007208 (J.R.B., R.C.D.).