The α₂Na⁺/K⁺-ATPase is critical for skeletal and heart muscle function in zebrafish

The Na⁺/K⁺-ATPase maintains Na⁺ and K⁺ gradients across the plasma membrane, which are crucial for various cell functions such as regulation of cell volume, pH, Na⁺-coupled secondary transport of molecules and neurotransmitters, maintenance of the resting membrane potential (RMP) of most cells, and the excitability of muscle and neuronal cells (Kaplan, 2002; Blanco, 2005). In mammals, the four α isoforms (α₁ through α₄) display distinct tissue-specific expressions (Kaplan, 2002; Lingrel et al., 2003): the α₁Na⁺/K⁺-ATPase is ubiquitously expressed and suggested to maintain fundamental cellular functions (James et al., 1999), the α₂Na⁺/K⁺-ATPase is predominantly expressed in skeletal muscle, but also in the heart and in astrocytes of the central nervous system, the α₃Na⁺/K⁺-ATPase is mainly neuron associated, and the α₄Na⁺/K⁺-ATPase expression testes specific (Orlowski and Lingrel, 1988; McGrail et al., 1991; Fink et al., 1996; Blanco et al., 1999; Lingrel et al., 2007; Bøttger et al., 2011).

Mutations in the human ATP1A2 gene were reported to cause a rare inherited subtype of migraine with aura, Familial hemiplegic migraine type 2 (FHM2) (De Fusco et al., 2003), and since then, multiple mutations in the ATP1A2 gene have been linked to FHM2 (Bøttger et al., 2012). Astrocytic α₂Na⁺/K⁺-ATPase plays an important role in clearance of K⁺ from the extracellular space during neuronal activity (D'Ambrosio et al., 2002), and is fundamental also for the clearance of released glutamate from the synaptic cleft, because active reuptake of glutamate into astrocytes is driven by Na⁺ and K⁺ gradients (Cholet et al., 2002; Ikeda et al., 2004; Rose et al., 2009). Consistently, mice lacking the α₂Na⁺/K⁺-ATPase die immediately after birth due to disruption of respiratory-related neuronal activity (Ikeda et al., 2004). Whereas, a missense α₂Na⁺/K⁺-ATPase knock-in FHM2 mouse model helped to identify the role of α₂Na⁺/K⁺-ATPase in cell electrophysiology, as cortical spreading depression (CSD, a wave of electrophysiological hyperactivity followed by a wave of inhibition) was facilitated in these animals (Leo et al., 2011). These and other mouse models contributed to our knowledge of the α₂Na⁺/K⁺-ATPase. However, major questions still remain about the biological consequences of the functional loss of the α₂Na⁺/K⁺-ATPase, e.g. in skeletal muscles and the heart. To address these questions, we employed zebrafish, Danio rerio, which has emerged as an excellent model organism for vertebrate biology. A single, highly conserved ortholog of the mammalian α₂ isoform was identified in zebrafish (Rajarao et al., 2001) and is required for embryonic cardiac patterning (Shu et al., 2003; Shu et al., 2007).

Skeletal muscle comprises fast and slow type fibers associated with different mechanisms, where fast-twitch muscles move five to ten times faster than slow-twitch muscles. Fast-twitch muscles gain their fast contraction time by a faster turnover of ATP compared to slow-twitch muscle. In the zebrafish embryo, the slow- and fast-twitch muscles are already distinguishable in the trunk from the onset of myogenesis (Devoto et al., 1996). At 17–19 hours post fertilization (hpf), spontaneous muscle contractions (coilings) start in the embryos and correlate with initial contact of primary motoneuron growth cones with muscle cells in anterior somites (Saint-Amant and Drapeau, 2001). This early spontaneous motor behavior precedes sensory touch responses, which can be evoked by 21 hpf. Later, at 27 hpf, the embryos develop swimming behavior, in line with the changes occurring in both the motor and sensory neural networks (Hirata et al., 2009). Of particular interest, glutamate signaling has been demonstrated to be required for response to touch in zebrafish (Pietri et al., 2009).

Our data has identified the α₂Na⁺/K⁺-ATPase as a key player in vertebrate skeletal and heart muscle function, implicated by decreased RMP in slow-twitch muscle cells, abrupt mechanosensory response and bradycardia, respectively.

As an initial step towards understanding the α₂Na⁺/K⁺-ATPase function in zebrafish embryos, the expression of Atp1a2 mRNA was analyzed. The Atp1a2 transcript was barely detected between 1k-cell (3 hpf) and 6-somite (12 hpf) stages, while a substantial increase in expression was observed between prim-6 (25 hpf) and pec fin (60 hpf) stages, reaching levels similar to and higher than those seen in adult fish, respectively (Fig. 1A). In situ hybridization further demonstrated abundant expression of the Atp1a2 transcript in skeletal muscle and in the brain of embryos at 35 and 60 hpf (Fig. 1B). At 60 hpf, expression in the brain was localized to the olfactory bulb, ventral thalamus, and midbrain tegmentum. Interestingly, expression in the cardiac region was also evident at this stage, as was the expression in muscle-rich regions, e.g. the pectoral fin. Furthermore, combined in situ hybridization and immunostaining for slow-twitch and fast-twitch muscles demonstrated that Atp1a2 mRNA is predominantly expressed in the slow-twitch muscles (Fig. 1C).

To functionally assess the role of α₂Na⁺/K⁺-ATPase in zebrafish development, we used an antisense morpholino oligonucleotide (MO)-induced approach to target the Atp1a2 mRNA. At 60 hpf, α₂-MO-injected [Atp1a2-knockdown (KD)] embryos displayed tail contortion and loss of the V-shape of the somites (Fig. 2A), both indicators of impaired skeletal muscle development. Moreover, the Atp1a2-KD embryos showed pericardial edema and bradycardia, indicating heart muscle dysfunction, when compared to non-injected control embryos or embryos injected with a control MO (std-MO; Fig. 2A,B). Also, the Atp1a2-KD embryos displayed heart tube malformation in the form of left-right asymmetry loss (Fig. 2A). In some instances of the latter phenotype, the pumping function of the heart was reduced dramatically, severely compromising blood circulation. To assess the bradycardia phenotype, representative hearts of noninjected (see supplementary material Movie 1), and Atp1a2-KD (see supplementary material Movie 2) embryos were video recorded and analyzed. Compared to noninjected or std-MO-injected embryos, Atp1a2-KD embryos showed a highly significant reduction in the heart rate (117±2.4/110±2.5 versus 79±5 beats/min; Fig. 2B).

To confirm the specificity of the Atp1a2-KD phenotype, embryos were co-injected with in vitro synthesized Atp1a2 mRNA to rescue the phenotypes. To prevent binding of the α₂-MO, six silent mutations were introduced in the Atp1a2 mRNA (mut-Atp1a2 mRNA). In summary, an average of 62% of the Atp1a2-KD embryos had swollen pericardium, 82% displayed contorted tail, and 83% had bradycardia. These phenotypes were significantly (P<0.05) less frequent, namely 24%, 45% and 65%, respectively, in the rescued embryos (Fig. 2C). Some of the mut-Atp1a2 mRNA-co-injected Atp1a2-KD embryos were fully rescued, and revealed a complete rescue of the somitic structures comparable to control embryos (Fig. 2A). Finally, co-injection with a p53-targeting morpholino (p53-MO) to block non-specific MO effect (Robu et al., 2007), resulted in similar phenotypes, ruling out the possibility that the phenotypes are due to interference with the apoptotic pathway (Fig. 2A).

The Na⁺/K⁺-ATPase controlled Na⁺ and K⁺ ion fluxes over the plasma membrane contribute, e.g. to extrusion of Ca²⁺ coupled to the Na⁺/Ca²⁺ exchanger (NCX), which has been found to be important for heart functions in rodents (James et al., 1999; Dostanic et al., 2003; Dostanic et al., 2004; Swift et al., 2008), and laterality in zebrafish (Shu et al., 2007). To determine if NCX might contribute to the Atp1a2-KD phenotypes (Fig. 2A–C), we microinjected zebrafish embryos with a MO against Ncx4a (Ncx4a-MO), the NCX isoform, which is expressed in both brain, heart and muscles, and thus, resembles the expression profile of the Atp1a2 the more than other NCX isoforms (Shu et al., 2007). Intriguingly, similar phenotypes, i.e. a swollen pericardium, a contorted tail and decreased heart rate, were observed in Ncx4a-MO-injected (Ncx4a-KD) embryos, although at lower frequencies (Fig. 2D,E). The heart rate was significantly reduced in the Ncx4a-KD embryos with bradycardia phenotype, compared to the noninjected embryos (98±2.5 versus 125±1.6 beats/minute) (Fig. 2F).

To assess possible electrophysiological effects in the slow-twitch muscles of Atp1a2-KD embryos, RMP in 48 hpf embryos were recorded. We also included Atp1a3a-KD embryos as a control, which is another catalytic subunit of Na⁺/K⁺-ATPase and not expressed in slow-twitch muscle (Rajarao et al., 2001). We found that slow-twitch muscle cells from Atp1a2-KD embryos were significantly (P<0.001) more depolarized (mean RMP: −48.2±2 mV), compared to normal values measured in noninjected (mean RMP: −66.6±1.4 mV) embryos, embryos injected with std-MO (mean RMP: −67.1±1.2 mV) and Atp1a3a-KD (mean RMP: −64±0.9 mV) embryos, which had comparable (P>0.05) RMP values (Fig. 3).

Locomotion in zebrafish embryos is a result of interplay between the nervous system and skeletal muscle (Granato et al., 1996). Considering the morphological defects and the changes in RMP in Atp1a2-KD embryos, we addressed the effect of deficient α₂Na⁺/K⁺-ATPase in embryonic motility. For this, touch responses of noninjected, std-MO-injected and Atp1a2-KD embryos were assessed at 60 hpf. The noninjected and std-MO-injected embryos (see supplementary material Movies 3, 4) responded with burst swimming to touch as expected from embryos at this stage. In contrast, Atp1a2-KD embryos displaying morphological phenotypes lacked burst swimming in response to touch, but rather displayed recoil or circling (see supplementary material Movie 5, parts I–II). Moreover, half of the embryos displayed notably delayed response to touch (see supplementary material Movie 5, part II). Representative frame shots of an Atp1a2-KD embryo compared to noninjected embryos are shown in Fig. 4.

Abrupt mechanosensory response, e.g. circling, was also observed in Ncx4a-KD embryos, whereas no delay in the response was noted in these embryos (see supplementary material Movie 6).

It is known that the mammalian astrocyte α₂Na⁺/K⁺-ATPase regulates the level of K⁺ and glutamate in the synaptic cleft. Since the glial EAAT2b glutamate receptor was recently shown to be important for motor behaviors in zebrafish (Pietri et al., 2009; McKeown et al., 2012), and considering the mechanosensory response defects in Atp1a2-KD embryos (Fig. 4), we wanted to determine whether α₂Na⁺/K⁺-ATPase is astrocyte associated, and whether knockdown of Atp1a2 would cause a detectable astrocytic phenotype. First, we tested the presence of Atp1a2 transcripts in cells expressing the astrocytic marker, glial fibrillary acidic protein (GFAP) (Eng and Shiurba, 1988; Bernardos and Raymond, 2006). A combination of Atp1a2 RNA in situ hybridization and immunostaining using a GFP primary antibody, performed on a transgenic zebrafish line Tg(Gfap:GFP) at 60 hpf, showed that Atp1a2 transcripts are expressed in the GFAP-positive astrocytes (Fig. 5A). However, no co-expression of Atp1a2 and GFAP was observed in the trunk (see supplementary material Fig. S2). To analyze the GFAP-positive astrocytes in Atp1a2-KD embryos, we performed Gfap in situ hybridization on Atp1a2-KD and std-MO-injected embryos (Fig. 5B, upper panel). The control group showed normal distribution of Gfap mRNA within distinct CNS structures, such as the forebrain ventricular zone, the posterior midbrain, the hindbrain, and the dorsal spinal cord. In contrast, structural distinctions of Gfap-expressing regions could not be recognized in Atp1a2-KD embryos, and the forebrain, the midbrain, and the hindbrain appeared conglutinated. Although the Atp1a2 transcripts were not detected in GFAP-expressing cells in the spinal cord, the distribution of Gfap mRNA on the dorsal spinal cord appeared to be affected. The distorted distribution of Gfap mRNA was further supported by imaging of Tg(Gfap:GFP) embryos (Fig. 5B, lower panel).

The cellular protein networks connected to the α₂Na⁺/K⁺-ATPase are largely unknown. We therefore carried out proteomics analysis to identify candidate proteins affected by Atp1a2-KD, using both two-dimensional gel electrophoresis (2D-GE) and isobaric tags for relative and absolute quantitation (iTRAQ). Identified candidate proteins were grouped according to tissue/pathway association (see supplementary material Table S1), and selected regulated proteins associated with skeletal muscle (e.g. pyruvate kinase), heart [e.g. calmodulin (CaM)], and brain (e.g. GFAP) were listed according to their possible relevance for the Atp1a2-KD phenotypes (Table 1). It is important to note that the identified proteins in the proteomics assay need further verification. We assessed the relative mRNA expressions of the four representative candidates from the proteomics assay (Pkm2b, Cam2l, Gfap and Ckbb). We found that these transcripts are also downregulated in Atp1a2-KD embryos, supporting the proteomics data at the transcript level (see supplementary material Fig. S1A).

The Atp1a2 mRNA expression is notable at the prim-6 stage (25 hpf), which correlates with the end of somitogenesis and the time when the zebrafish heart becomes visible as a cone-shaped tube, prominently occupying a pericardial sac on the anterior-most region of the yolk (Kimmel et al., 1995). The higher Atp1a2 expression in the later developmental stages is consistent with the increasing Na⁺/K⁺-ATPase activity in the developing chick heart (Sperelakis, 1972), mouse (Cougnon et al., 2002) and rat (Kjeldsen et al., 1982) skeletal muscles from infant to adult.

Compared to the mammalian α₂Na⁺/K⁺-ATPase, which is highly abundant in both fast- and slow-twitch muscles (Orlowski and Lingrel, 1988; Bonilla et al., 1991; Fowles et al., 2004), the zebrafish Atp1a2 mRNA is also predominantly expressed in skeletal muscles, but particularly in slow-twitch muscles. Currently, there is no study reporting the expression of different Na⁺/K⁺-ATPases in different muscle fibers of zebrafish. Distinct isoforms of the Na⁺/K⁺-ATPase could indeed be expressed in the same cell; one or more isoforms could be required to maintain the more housekeeping role of the Na⁺/K⁺-ATPase in ion homeostasis and one or more isoforms could carry out more specialized functions in the cell. The expression of the α₂Na⁺/K⁺-ATPase, particularly in the slow-twitch muscles, might indicate that this isoform is associated with slow-twitch muscle-specific functions, such as maintenance of the posture. Consistent with the Atp1a2 expression, analysis of the somitic structures in the Atp1a2-KD embryos revealed skeletal muscle abnormalities. Analysis at the cellular level by electrophysiological recordings demonstrated that the slow-twitch muscle cells of Atp1a2-KD embryos were more depolarized than the control groups, supporting the role of α₂Na⁺/K⁺-ATPase in maintaining ion homeostasis in these cells. This correlates well with the predicted decrease in the intracellular K⁺ level due to reduced Na⁺/K⁺-ATPase activity, which, subsequently, would give rise to a more depolarized state. Indeed, studies in rat show a decrease in RMP by the specific Na⁺/K⁺-ATPase inhibitor ouabain, both in fast and slow-twitch muscles (Tyapkina et al., 2009). Such cellular modification might have affected skeletal muscle function directly or via its influence on certain pathways that lead to particular protein regulations as detected in our proteomics assay. For instance, several muscle-associated factors, such as muscle kinases, parvalbumin and enolase, are downregulated in Atp1a2-KD embryos, although the direct interaction between these candidates and the α₂ isoform remains to be investigated. Due to the difficulties in accessing to fast-twitch muscle cells, located more profoundly than the superficial slow-twitch muscle cells, we were unable to measure RMP in those cells.

Although we did not include a long recording of spontaneous swimming behavior of the Atp1a2-KD zebrafish, most of these embryos are significantly immotile, and approximately half of the Atp1a2-KD zebrafish display notable delay in mechanosensory responses compared to control groups. This might relate to previous observations, where α₂^+/−-KO mice were reported to be hypoactive (Lingrel et al., 2007; Moseley et al., 2007), and α₂^−/−-KO mice were characterized by lack of spontaneous body movement and response to pinch (Moseley et al., 2003). Considering that the zebrafish motility is controlled by both the CNS and skeletal muscles (Granato et al., 1996), reduced touch response of the Atp1a2-KD embryos might be a consequence of the combined skeletal muscle and brain defects in the Atp1a2-KD zebrafish. For the other half of the Atp1a2-KD embryos showing abnormal motility, it is tempting to speculate that their phenotype might result from skeletal muscle defects. However, further studies are necessary to distinguish between neuronal and muscle defects in the mechanosensory response of the Atp1a2-KD embryos.

In skeletal muscle, excitability and contractility depend on the membrane potential and the Na⁺ and K⁺ gradients across the plasma membrane, mediated by the Na⁺/K⁺-ATPase. In mice, a reduction of α₂Na⁺/K⁺-ATPase in skeletal muscles results in an increased isometric force, which could be the result of the inhibition of the NCX activity that causes an increase of the sarcoplasmic/endoplasmic reticulum Ca²⁺ (He et al., 2001). We found that the Ncx4a-KD zebrafish, similar to Atp1a2-KD zebrafish, had distorted tail, and in connection with that displayed abrupt swimming in response to touch. Thus, we suggest that the Ca²⁺ homeostasis can be affected by inhibition of NCX activity in the Atp1a2-KD zebrafish, and may, along with the increased isometric force, contribute to the aberrant responses to touch. In line with this, proteins associated with Ca²⁺ signaling and sequestering, such as parvalbumin and CaM, were altered in Atp1a2-KD zebrafish as detected by proteomics.

The α₂ isoform has previously been associated with heart function in human, guinea pig, rat, mouse and zebrafish (Shamraj et al., 1991; Gao et al., 1999; James et al., 1999; Rajarao et al., 2001; Canfield et al., 2002). We observed subtle embryonic expression of the Atp1a2 mRNA in the heart at 60 hpf. The pericardial swelling indicated a cardiac defect in the Atp1a2-KD zebrafish, and the morphology of the heart within the swollen pericardium was subsequently found to be disrupted, with loss of the heart tube LR asymmetry. Defects in the Atp1a2-KD heart muscles were marked by a significant bradycardia. Bradycardia is most often connected to the electrical pathways involving intracellular Na⁺ and hence Ca²⁺ concentrations of the heart (James et al., 1999; Berry et al., 2007), and it is known to lead to low blood pressure (Walker et al., 1990). In the Atp1a2-KD embryos with severe phenotypes, efficient blood body circulation could not be maintained, probably as a direct consequence of the bradycardia. Such phenotypes are in agreement with the known roles of α₂Na⁺/K⁺-ATPase in control of myocardial contractility and blood pressure regulation (Zhang et al., 2005; Blaustein et al., 2006; Swift et al., 2007; Blaustein et al., 2009). Cardiac defects observed also in Atp1a1a.1-KD zebrafish (Shu et al., 2003), indicate that the role of α₂Na⁺/K⁺-ATPase in the heart is most likely based on its pump function to maintain ion homeostasis. However we cannot rule out the contributions from α₂ isoform-specific interactions.

The fluxes of both Na⁺ and K⁺ ions affect several other membrane proteins required for ion homeostasis, such as the NCX. It is well known that NCX and Na⁺/K⁺-ATPase cooperate to maintain a low intracellular Ca²⁺ level in resting cells (Blanco and Mercer, 1998). In zebrafish, the α₂ isoform and the NCX4a were also reported to be functional partners, in e.g. establishment of cardiac laterality (Shu et al., 2007). Intriguingly, the Ncx4a-KD embryos, also displayed bradycardia, although to a lower extend compared to the Atp1a2-KD embryos, indicating that the disruption of functional coupling between these two proteins in maintaining the intracellular Ca²⁺ concentration due to a disturbed Na⁺ gradient has an impact at some level on the heart phenotypes observed in Atp1a2-KD zebrafish, which was suggested for hypercontractile hearts and elevated blood pressure in α₂^+/−-KO mice and congestive heart failure in a rat model (James et al., 1999; Zhang et al., 2005; Swift et al., 2008). In our proteomics analysis of Atp1a2-KD embryos, we further detected altered proteins associated with cardiac function, which might have contributed to the observed high prevalence heart phenotypes in these embryos.

In line with this, several studies have addressed the role of the Na⁺/K⁺-ATPase in relation to heart failure, such as systemic vascular hypertension modeled in rat hypertensive models (Herrera et al., 1988), increased ouabain hypersensitivity to failing human hearts (Shamraj et al., 1993) and significant positive inotropic effects in intact mouse hearts (Xu et al., 2006). In addition, a specific reduction of the α₂ isoform in hypertrophied rat hearts (Charlemagne et al., 1994; Sweadner et al., 1994), α₂ isoform specific ouabain-induced cardiac contractility in mice (Dostanic et al., 2003) and specific contributions of the α₂ isoform in hypertension in mice (Rindler et al., 2011; Van Huysse et al., 2011) were also reported.

The human brain shows particularly high expression of Na⁺/K⁺-ATPases (Clausen, 1998); it utilizes twenty percent of the body's total energy (Raichle and Gusnard, 2002), with the majority consumed by the Na⁺/K⁺-ATPases. We detected the zebrafish Atp1a2 transcript in the brain, particularly in the ventral thalamus, the tegmentum and the olfactory bulb – areas important for the relay of sensory information to the cerebral cortex (Swenson 2006). It is known that the more rostral part of the hindbrain is important for touch-evoked escape responses (Pietri et al., 2009). Thus, defects in Atp1a2 expression areas might affect the sensory and motor systems in the Atp1a2-KD embryos, which could contribute to the altered touch response in these embryos. The mechanisms underlying the pathology of FHM2 are related to the impaired clearance of K⁺ and glutamate, and subsequent development of CSD, the phenomenon believed to trigger aura before a migraine attack (D'Ambrosio et al., 2002; De Fusco et al., 2003; Rose et al., 2009). This highlights the significance of the role of α₂ isoform in astrocytes. Our results suggest that the α₂ isoform is important for the integrity and organization of GFAP-positive astrocytes, and our proteomic data revealed that GFAP is indeed downregulated in Atp1a2-KD embryos. Therefore, this indicates for the first time a direct link between the α₂ isoform and GFAP as an integrity measure for astrocyte functions. Considering the recent evidence, suggesting that the glial EAAT2b glutamate receptor is important for motor behaviors in zebrafish (McKeown et al., 2012), we speculate that the disturbed glutamate clearance in astrocytes upon knockdown of Atp1a2, may, together with the skeletal muscle defects, have an impact on the reduced/lack of mechanosensory response seen in these embryos. Moreover, a change in astrocytic Ca²⁺ concentration can cause the release of glutamate, which triggers signaling pathways within the functional nervous system (Parpura and Haydon, 2000). Thus, the abnormal mechanosensory response observed in Ncx4a-KD embryos further supports the glutamatergic association of the mechanosensory response, and suggests an impact of the Ca²⁺ homeostasis in the abrupt mechanosensory response of the Atp1a2-KD embryos.

Through a combination of expression and functional assessments, we have attempted to dissect the key roles of α₂Na⁺/K⁺-ATPase, and found that α₂Na⁺/K⁺-ATPase is essential for vertebrate skeletal and heart muscle development and functions. The phenotypes observed suggest that impaired α₂Na⁺/K⁺-ATPase pump function might interfere with essential steps already during mesoderm formation. Somites are derived from the presomitic paraxial mesoderm during gastrulation (5.25–10 hpf) and segmentation (10–24 hpf), respectively, where the Atp1a2 was slightly detected, and thus, it is tempting to predict an essential role for the α₂Na⁺/K⁺-ATPase during early embryogenesis, where heart and skeletal muscle progenitor cells derive from distinct regions of the mesoderm (i.e. lateral plate mesoderm (LM) and paraxial mesoderm (PM), respectively) (Parker et al., 2003; Tirosh-Finkel et al., 2006) (Fig. 6). Similarly, the α₂Na⁺/K⁺-ATPase could be required during and/or after the development of the CNS from the neural tube (NT), which could affect the GFAP- and α₂Na⁺/K⁺-ATPase-positive astrocytes. Through proteomic analysis, novel protein associations of α₂Na⁺/K⁺-ATPase were revealed. Such interactions are interpreted to have a significant and specific impact on the phenotypes observed in the Atp1a2-KD deficient zebrafish embryos, some of which were also noted in FHM2 patients and mouse models.

Tübingen (TU) zebrafish strain (Nüsslein-Volhard lab, Max Planck Institute, Tübingen, Germany) and Tg(Gfap:GFP) transgenic line (Zebrafish International Resource Center, University of Oregon, USA) were used in the experiments. Embryos were maintained and staged as previously described (Kimmel et al., 1995; Westerfield, 1995). For optical clarity embryos were raised in the presence of 0.2 mM 1-phenyl-2-thiourea (PTU) following gastrulation (Westerfield, 1995).

Reverse transcription (RT) and quantitative RT-polymerase chain reaction (qRT-PCR) protocols were applied on cDNAs from adult zebrafish and embryos at early developmental stages (1k-cell, 50% epiboly, 75% epiboly, 6-somite, prim-6, prim-22, pec fin) as previously described (Doğanlı et al., 2010). Primers used for Atp1a2 (accession number: AF286373.1), Actb2 (accession number: NM_181601.3), Cam2l (accession number: XP_001919221), Ckbb (accession number: BC049529.1), Gfap (accession number: Q58EE9) and Pkm2b (accession number: NM_001003488.1; see supplementary material Table S2) were designed using an exon-spanning approach from Pearl Primer software (Marshall, 2004) or Roche Probe Finder software. PCR product sizes and identities were verified by gel electrophoresis (see supplementary material Fig. S1B) and DNA sequencing. Primer efficiencies were approved to be ∼100% for each primer pair used in qRT-PCR. Transcription levels were quantified using the relative quantification method based on comparative threshold cycle values (Ct).

Total RNA was isolated from a male and a female zebrafish by TRIzol® method (Invitrogen) (Chomczynski and Mackey, 1995). RT-PCR was performed on total RNA by using SuperScript^TM III RT-PCR system (Invitrogen) according to manufacturer's instructions with Atp1a2-specific primer pairs tagged with restriction enzyme recognition sites. The program for RT-PCR was 1 cycle for 30 min at 50°C and 2 min at 94°C followed by 40 cycles of 15 sec at 94°C, 30 sec at 55°C and 3.5 min at 68°C and a final extension for 15 min at 68°C. The amplified fragments were purified from the gel using gel extraction kit (Macherey-Nagel), cloned into pTZ57R vector (InsTAclone^TM PCR Cloning Kit, Fermentas) following the manufacturers' protocols, and subjected to DNA sequencing. The same primer pairs were used to amplify full length open reading frame of Atp1a2.

The templates for antisense and sense riboprobes were generated by either RT-PCR on adult zebrafish RNA or for the case of Atp1a2 by PCR on pTZ57R vector harboring Atp1a2 probe coding sequence, introducing T7 priming site in both scenarios. Digoxigenin (DIG)-labeled riboprobes were synthesized from purified PCR products using the DIG RNA labeling mix (Roche) and the T7 RNA polymerase (Roche). Embryos at 35 hpf and 60 hpf (refers to 55–60 hpf) were fixed with freshly prepared 4% formaldehyde in phosphate-buffered saline (PBS; pH 7.4) overnight and whole-mount in situ hybridization was performed as previously described (Thisse and Thisse, 2008) with minor modifications. Embryos were imaged under an inverted microscope, Olympus IX71.

Embryos, TU WT (for F59 and F310) and Tg(Gfap:GFP; for GFP), were raised in embryo medium with PTU. In situ hybridization was adapted from a previously published method (Schulte-Merker et al., 1992), and immunostaining was performed subsequently as previously described (Novak and Ribera, 2003) with minor modifications. Atp1a2 antisense riboprobe was used for in situ hybridization, which was labeled using Fast Red tablets (Sigma) following the manufacturer's protocol. Stained embryos were fixed in 4% paraformaldehyde, blocked in 10% HIGS/PBST and incubated with primary monoclonal anti-GFP (Santa Cruz; 1∶100), F59 or F310 (Hybridoma Bank; 1∶10), at 4°C, overnight. Following the washes, embryos were incubated in secondary donkey anti-rabbit (for GFP) and goat anti-mouse (for F59 and F310) antibodies conjugated to Alexa Fluor 488 (1∶1000), for 4 hours at room temperature. Embryos were washed thoroughly and gradually transferred to 100% glycerol.

Embryos were embedded and oriented in 5% agar, so they could be cut transversely with regard to the length direction following dehydrating in ascending alcohol solutions (70% twice, 96% twice and 99% twice) and infiltration and embedding in glycolmethacrylate (Technovit 7100). Using a Microm 355, 20-µm-thick sections were cut and every tenth section was taken by systematic sampling. Whole-mount embryos and sections were observed using an inverted microscope, Olympus IX71 (for F59 and F310) or Zeiss LSM 710 T-PMT confocal microscope (for GFP).

Morpholinos (Gene Tools, LLC) for microinjection was diluted in sterile-filtered 2 mM Hepes, pH 7.0 with phenol red and microinjected into embryos at the 1- to 4-cell stage. Translation blocking α₂-MO (5′-TTTCATGTCCGTACCCTTTCCCCAT-3′) and control std-MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were injected at concentrations of 8 ng/embryo. 4 ng translation blocking p53-MO (5′-GCGCCATTGCTTTGCAAGAATTG-3′) was co-injected with α₂-MO as part of the control experiments. Atp1a3a-KD and Ncx4a-KD were mediated by injections of 11 ng/embryo translation blocking α_3a-MO (5′-CTTTCTTCAGTCTGTCAAACGGCGT-3′) and 12 ng/embryo translation blocking Ncx4a-MO (5′-AAAGGCGCAGATGAAACATGGTGGC-3′), respectively. Specificity of all the MOs used was confirmed by sequence alignments, and searches by basic local alignment search tool (BLAST) supported that no other mRNA than the intended will be targeted by the MOs. Embryos were grown for 60 h in embryo medium with PTU and observed under inverted microscope, Olympus IX71.

Silent mutations were introduced by PCR to the Atp1a2 coding sequence (mutations highlighted in bold: 5′-TAAGGGCTATGGTCAC-3′) on the α₂-MO binding site and mut-Atp1a2 was subcloned into the pTZ57R vector. mRNA was in vitro synthesized from gel purified mut-Atp1a2 PCR product by using the mMessage mMachine T7 ULTRA kit (Ambion) following the manufacturer's instructions. The RNA products were gel purified (Qiagen) and each embryo was co-injected with 120 pg of mRNA together with 8 ng α₂-MO. The experiment was repeated at least three times.

Zebrafish heart beats were recorded under an Olympus IX71 microscope at 10 frames/sec. Heart rates were determined by manual counting of ventricular and atrial beats in 20 second periods. At least three measurements were taken for each embryo. Data are presented as means ± s.e.m.

48 hpf embryos were sacrificed, skinned and mounted laterally as previously described (Moreno and Ribera, 2009). Sharp microelectrodes were pulled (P-97 microelectrode puller, Sutter Instruments, Novato, CA), filled with 3 M CH₃CO₂K and had resistance of 15–20 MΩ. The bath solution consisted of (in mM): 125 NaCl, 3 KCl, 10 CaCl₂, 5 HEPES, pH 7.4. Membrane potentials of superficial (slow) muscle cells were recorded using a 705 Electrometer (World Precision Instruments, Inc., Sarasota, FL). Data are presented as means ± s.e.m.

Embryos at 60 hpf were videorecorded under an Olympus IX71 microscope at 25 frames/sec. Mechanosensory stimulation was delivered to the trunk with a needle.

The principles for running 2D gels have been described in detail previously (Urbonavicius et al., 2009; Mandal et al., 2011). 30–50 of std-MO and α₂-MO-injected embryos, 60 hpf, were collected in Eppendorf tubes, were lysed in lysis buffer (pH 3–10 NL; 9 M urea, 2% Triton X-100, 0.13 M dithiothreitol, and 2% IPG buffer) with the help of a homogenizer and centrifuged at 14,000 rpm for 10 min. 50 µg of sample was loaded in each well with up to 150 µl lysis and 175 µl rehydrating buffer and the first dimension was run on precast immobilized pH gradient strips (18 cm; pH 3–10 NL) with the Multiphor II system (Amersham Biosciences, Inc., Buckinghamshire, UK). The second dimension was run on polyacrylamide gels (12% T, 3% C) that were fixed in 50% (v/v) ethanol, 12% (v/v) acetic acid, 0.0185% (v/v) formaldehyde for at least 1 hour or overnight. Subsequently, silver staining was performed. The dried gels were scanned on a GS-710 Imaging Densitometer from Bio-Rad (Hercules, CA, USA) using the Quantity One software package. Spot analysis was performed using PDQuest 2D-PAGE analysis software. Significant differences among the selected spots were calculated using a two-sample Student's t-test (P<0.05). Only well-focused spots showing significant correlation (between two groups, including four gels per group) were excised for in-gel digestion and identification by liquid chromatography–tandem mass spectrometry (LC-MS/MS), as previously described (Mandal et al., 2011) with the following modifications. The D. rerio part of the Swiss-Prot database was searched using the online version of the Mascot MS/MS Ions Search Facility (Matrix Science, Ltd) (Perkins et al., 1999). Searches were performed with doubly and triply charged ions with up to two missed cleavages, a peptide tolerance of 20 ppm, one variable modification, Carbamidomethyl-C, and a MS/MS tolerance of 0.05 Da. Contaminating peptides such as keratins were disregarded.

Cell lysate for iTRAQ LC-MS/MS analysis was prepared as described above with a different lysis buffer composition [0.02 M Tris, 0.137 M NaCl, 1% NP-40, 10% Glycerol, 1 mM PMSF, 1× Complete Protease Inhibitor Cocktail tablet (Roche)]. Cell lysate was analyzed by iTRAQ LC-MS/MS as previously described (Beck et al., 2011). Briefly, the cell lysate was in-solution digested using trypsin; labeled with the iTRAQ (reporter ions m/z 114 and 115) followed by fractionation of the peptide mixture using hydrophilic interaction chromatography and LC-MS/MS analysis of the peptide fractions. Raw data files were processed using the Proteome Discoverer software (version 1.3) integrated with the MASCOT database search program (version 2.2.3). Data were searched against the nrDB database restricted to D. rerio. Some of the tissue associations of the identified proteins were provided according to http://zfin.org/ or http://www.ncbi.nlm.nih.gov/unigene.

We thank C. Romberg, Department of Physiology and Biophysics, University of Colorado, Denver, for assistance in setting up the electrophysiology instrumentation for the sharp electrode and to Mona Hansen, Department of Biomedicine, Aarhus University, for expert technical assistance on 2D-GE.