Eya4 regulation of Na⁺/K⁺-ATPase is required for sensory system development in zebrafish

Eya molecules are evolutionarily conserved transcriptional co-activators that participate in the development of multiple organs, including the eye,pituitary gland, muscle, kidney, inner ear and heart(Hanson, 2001; Kawakami et al., 2000; Rebay et al., 2005; Schonberger et al., 2005). Eya proteins contain a highly conserved carboxyl domain with phosphatase activity and interaction sites for Sine oculis (Six) proteins, and a variable amino domain with unknown functions (Hanson,2001; Kawakami et al.,2000; Rebay et al.,2005; Schonberger et al.,2005). Interactions with Six proteins permit Eya proteins to translocate into the nucleus (Hanson,2001; Rebay et al.,2005) where phosphatase activity releases transcriptional repression caused by Six-Dachshund and other molecules(Li et al., 2003; Rayapureddi et al., 2003; Tootle et al., 2003). Human deletions in EYA4 cause a dominant form of sensorineural hearing loss(Pfister et al., 2002; Wayne et al., 2001), which sometimes is accompanied by late-onset dilated cardiomyopathy(Schonberger et al., 2000; Schonberger et al., 2005). The gene targets regulated by Eya4, Six and Dachshund in the ear or heart are unknown.

Na⁺/K⁺-ATPases comprise a plasma membrane enzymatic complex that regulates ion homeostasis in many eukaryotic tissues(Blanco and Mercer, 1998; Therien and Blostein, 2000). With ATP hydrolysis, the complex extrudes three Na⁺ ions and imports two K⁺ ions, therein establishing a chemical and electrical gradient. Na⁺/K⁺-ATPase participates in maintaining the delicate balance of high K⁺, low Na⁺ in the endolymphatic fluid that bathes the sensory epithelium in the membranous labyrinth of the inner ear and that is required for sensory transduction(Wangemann, 2002). In the heart, Na⁺/K⁺-ATPase cooperates with the Na⁺/Ca²⁺ exchanger to produce inotropic effects on myocytes (Ingwall and Balschi,2006; Schwinger et al.,2003). Heart failure is commonly treated with cardiac glycosides(e.g. digitalis), which bind the α subunit of the Na⁺/K⁺-ATPase and inhibit pump functioning so that intracellular Na⁺ increases, Ca²⁺ extrusion decreases and contractile performance is enhanced(Ingwall and Balschi, 2006; Schwinger et al., 2003).

The heterodimeric Na⁺/K⁺-ATPases complexes contain a catalytic α subunit and a stabilizing β subunit(Jorgensen, 1974); both are required for enzyme activity (Goldin,1977; Horowitz et al.,1990). There is considerable evolutionary diversity in the subunit isoforms of Na⁺/K⁺-ATPases. Human and rodent genomes contain four α subunit and three β subunit isoforms, while the zebrafish (Danio rerio) genome encodes nine α and six βisoforms (Blasiole et al.,2002; Levenson,1994; Malik et al.,1998; Rajarao et al.,2002; Rajarao et al.,2001; Shamraj and Lingrel,1994; Underhill et al.,1999). Isoform expression is spatially and temporally regulated(Peters et al., 2001). In human, rat and mouse, β2 subunit expression encoded by the Atp1b2 gene is restricted to the marginal cells of the stria vascularis in the cochlea (Peters et al.,2001; Wangemann,2002). Zebrafish have two β2 subunit genes: atp1b2ais expressed in the brain, spinal cord neurons and lateral line ganglia; atp1b2b predominates in the retina, neuromasts and otic vesicles(Rajarao et al., 2002). atp1b2b gene expression is required for otic vesicle development and otolith formation (Blasiole et al.,2006).

Na⁺/K⁺-ATPases are also essential for zebrafish heart development. Two zebrafish mutants, heart and mind(Shu et al., 2003) and small heart (Yuan and Joseph,2004), are caused by deficiency of atp1a1a.1 (also known as atp1α1B1) and have malformed hearts (due to defective primitive heart tube extension and cardiomyocyte differentiation)with reduced contractility and heart rate.

We characterized eya4 expression in the mechanosensory epithelia of the zebrafish otic vesicle and in neuromasts, sensory patches that are related to the mammalian inner ear(Whitfield et al., 1996; Whitfield et al., 2002), and demonstrated that eya4 antisense morpholino oligonucleotides reduced hair cell numbers in these organs. Hypothesizing that Eya4 regulated the expression of Na⁺/K⁺-ATPase, we examined subunit levels in eya4 morphant zebrafish and demonstrated the selective reduction of two subunits. Re-expression of the Na⁺/K⁺-ATPaseβ2b subunit rescued eya4 deficiency in morphant zebrafish. Taken together, these results indicate that Eya4 regulates Na⁺/K⁺-ATPase, and therein provides a mechanism by which human EYA4 mutations cause both hearing loss and heart disease.

Zebrafish studies were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC). Wild-type AB zebrafish were maintained, and embryos collected and staged, as previously described(Westerfield, 2000). Embryos were anesthetized with 0.16 mg/ml Tricaine (Sigma) in E3 solution prior to sacrifice.

Whole-mount in situ hybridization was performed as described(Jowett, 1999). Sections were hybridized with slight modifications(Schonberger et al., 2005). Zebrafish embryos were fixed with 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS), embedded in paraffin, and longitudinally sectioned to view the otic vesicle and heart. Using full-length eya4 and atp1b2b cDNAs as templates, RNA probes were synthesized and labeled using digoxigenin (DIG) RNA labeling kit (Roche). Isotope-labeled RNA probes were synthesized with ³⁵S-ATP and ³⁵S-UTP; signal was detected using liquid film emulsion autoradiography (Kodak) of Toluidine Blue-counterstained tissues.

All analyses were performed with the observers blinded to the morpholino oligonucleotide treatment status of the zebrafish embryos.

Zebrafish embryos were fixed in methanol-free, 4% PFA in PBS buffer for 24-48 hours (4°C), washed in PBST (PBS with 0.5% Tween-20), and permeabilized either by incubation in 1% Triton X-100 in PBS (7 hours) or by incubation in acetone -15°C (20 minutes). After washing in PBST, embryos were reacted (2.5 hours) with 50 ng/ml FITC or Alexa Fluor 488-labeled phalloidin in 1% Triton X-100/PBS. After washing three times with PBST (30 minutes) embryos were embedded in agar or 50% glycerol and visualized using a Leica TCS SP2 confocal microscope.

Neuromast hair cells were stained with vital dye, 1 mM DASPEI[4-(4-(diethylamino)-styryl)-N-methylpyridinium iodide] (Molecular Probes) in E3 solution (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄), for 10 minutes, then rinsed five times with E3 solution(Whitfield et al., 1996). Embryos were anaesthetized with 0.02% Tricaine in E3 solution and observed using a FM-1-43 filter set (Chroma).

Cell death was assessed by Acridine Orange staining (2 μg/ml) of zebrafish embryos as previously described(Blasiole et al., 2006). The otic vesicles were observed under a Zeiss Discovery V8 fluorescence dissecting microscope.

A three-dimensional otic vesicle surface model was reconstructed from serial confocal images of the otic vesicle acquired by differential interference contrast imaging using the Amira Advanced 3D Visualization and Volume Modeling Software (Version 4.1.2, Mercury Computer Systems).

Total RNA was isolated using Trizol Reagent (Invitrogen). RT-PCR and semi-quantitative RT-PCR reactions were carried out using One-Step RT-PCR kits(Qiagen), following the manufacturer's instructions(Ausubel et al., 2008). For semi-quantitative RT-PCR, 1 μg of total RNA was used as template where eya1 and atp1b2b cDNA were co-amplified for 25 cycles in a 50 μl reaction using 0.3 μM of each of the gene-specific primers(eya1, 5′ATGGAAATGCAGGATCTAGCTAGT3′ and 5′CTGCTGCTCATTGGCTCTGTTTTAA3′; atp1b2b,5′TGAGCGCACATTTAGTCCAG3′ and 5′GCACCACACGTGACATAAGG3′). Reactions (10 μl) were resolved on 2% agarose gels and band intensity was calculated using ImageJ software(NIH).

Full-length atp1b2b cDNA was amplified from total RNA extracted from zebrafish embryos (72 hpf) by RT-PCR using the primer set 5′GCACGAGGTCTCTCTCTCTCTCTCTCTCC3′ and 5′AAAAAAAAAACAAACACTCTTGTCTGTTCAATCTCTGG3′; cDNA was then cloned into pCS2+ between the EcoRI and XhoI sites. The cloned sequence was verified. atp1b2b mRNA was subsequently synthesized using the mMESSAGE mMACHINE High Yield Capped RNA Transcription Kit(Ambion).

All morpholino oligonucleotides were complimentary to splice donor sites and were injected into 1- to 2-cell-stage zebrafish embryos. Sequences and injected amounts of morpholino oligonucleotides against eya4 (see also Schonberger et al., 2005)and against atp1b2b (denoted β2b) are provided below. Numbers denote the exon-intron splice donor sites targeted by the morpholino (e.g. MO3 indicates that the morpholino target is the splicing donor site at exon 3). Mismatch morpholinos (e.g. MO3mis) differ from the related morpholino oligonucleotides at five bases.

MO3eya4 (1 ng), TACTGATGTTACCTGTTGTCTACTG;

MO10eya4 (1 ng), TAATATGTATACCTGGCATCTGATA;

MO3mis (1 ng), TAATGATTTTAACTGTTCTCTAATG;

MO10mis (1 ng), TATTATCTATAGCTGGGATCTCATA;

MO1β2b (2.5 ng), AGCTAGTCTTACCCCAACTGCTCGC; and

MO4β2b (0.6 ng), TGTTTCTCATCTTACACGGTTGAGC.

Observers who were blinded to the morpholino oligonucleotide treatment status of the zebrafish embryos assessed heart phenotypes. Cardiac dimensions and contraction cycle lengths (to deduce heart rate) were measured using quantitative high-speed image analysis as described(Schonberger et al.,2005).

Cardiac expression of atp1b2b was assessed using the pEGFP-N1 plasmid (Invitrogen). In brief, a 2.5-kb promoter fragment of the atp1b2b gene was cloned into pEGFP-N1. The modified plasmid (100 pg;denoted pβ2bprom) was injected into one-cell-stage embryos and fluorescence was examined daily using a Zeiss Discovery V8 microscope.

Individual wild-type or morphant fish (72 hpf) with normal head morphology and tails that were not curved ventrally or dorsally were placed in a 100 mm Petri dish filled with 25 ml E3 solution. Startle reflexes were evoked by a single tap to a petri dish, with simultaneous video recording of movements. Zebrafish that moved outside of the camera field via coordinated swimming were scored as normal (see Movie 3 in the supplementary material). Zebrafish that either did not swim or moved only in concentric tight circles were scored as abnormal (see Movie 4 in the supplementary material). Startle reflexes were evoked at least three times in each zebrafish.

Student's t-tests were used to compare hair cells in wild-type and morphant fish. Fisher exact tests were used to compare cardiac phenotypes and startle reflexes in wild-type and morphant fish. P-values <0.05 were considered significant. Means±standard error (s.e.m.) are reported.

Previous studies (Schonberger et al.,2005) demonstrated that adult zebrafish express eya4 in the otic vesicle and neuromast sensory organs. To characterize developmental expression, we studied zebrafish embryos and tissue sections of the developing otic vesicle using eya4 antisense probes. By 14 hours post-fertilization (hpf), eya4 transcripts were detected at two distinct sites in the otic placode region (see Fig. S1A in the supplementary material). This pattern persisted through further otic placode development assessed at 16 and 18 hpf (see Fig. S1B,C in the supplementary material). At 18 hpf, an additional cluster of cells, anterior to the otic placode, which may give rise to ganglion neuroblasts (see below), also expressed eya4 (see Fig. S1C in the supplementary material). By 24 hpf, eya4 expression was most prominent along the anterior and posterior axis of the ventral side of the otic vesicle(Fig. 1A; see also Fig. S1D in the supplementary material), from where the progenitors of sensory epithelial cells originate (Whitfield,2002; Whitfield et al.,2002). Neuroblasts in the statoacoustic ganglion, adjacent to the otic vesicle, also expressed eya4(Fig. 1A; see also Fig. S1D in the supplementary material). At 36 hpf this expression pattern was maintained,albeit with some broadening of ventral eya4 expression(Fig. 1B). By 48 hpf, eya4 expression was diffuse, and included the anterior macula(Fig. 1D,E) and pharyngeal arches (Fig. 1D,E), as well as several ganglia adjacent to the otic vesicle(Fig. 1D). eya4expression patterns in the otic vesicles were similar at 60 and 72 hpf(Fig. 1C; see also Fig. S1E;data not shown): eya4 expression was restricted to the sensory patches (including hair cells and supporting cells) in the three cristae(anterior, lateral and posterior; see Fig. 1C, and Fig. S1E in the supplementary material) and in the two maculae (anterior and posterior; see Fig. S1E in the supplementary material;data not shown).

To consider eya4 roles in otic vesicle maturation, we examined the developmental effects of two eya4 antisense morpholino oligonucleotides (MO3eya4 and MO10eya4), which are complementary to the splice donor sites of exon 3 and exon 10, respectively, in over 200 zebrafish embryos. Previous analyses using these morpholino oligonucleotides showed selective attenuation of eya4 expression without changing eya1 expression (Schonberger et al., 2005). Although maturation of the zebrafish otic vesicle is not complete until 120 hpf (Whitfield,2002; Whitfield et al.,2002), analyses later than 72 hpf were not performed because morpholino oligonucleotides were diluted at these later developmental stages.

By acquiring differential interference contrast and serial confocal images of zebrafish otic vesicles at 72 hpf, we produced three-dimensional reconstructions of the otic vesicles (Fig. 2; see also Movies 1, 2 in the supplementary material). At 72 hpf, eya4 morphant fish had smaller (cf. Fig. 2A and 2B) and misshaped (compare the DIC images of Fig. 3B-D) otic vesicles compared with wild-type embryos.

By 72 hpf, wild-type fish develop projections of epithelial pillars that grow and fuse in the center of the otic vesicle to provide scaffolding and to shape the developing semicircular canals (anterior, lateral and posterior),and each canal is associated with a corresponding crista(Whitfield et al., 2002) (see Fig. 2A, Fig. 3A,B; see also Movie 1 in the supplementary material). In eya4 morphant embryos, the smaller-sized otic vesicle contained foreshortened, misshaped and, usually,disconnected epithelial pillar protrusions, so that portioning of the otic vesicle failed to occur (Fig. 2B, Fig. 3C,D; see also Movie 2 in the supplementary material). This led to malformed semicircular canals and the disruption of their association with the diminutive sensory cristae in eya4 morphant fish(Fig. 2B, Fig. 3C,D; see also Movie 2 in the supplementary material).

The formation of otoliths (crystalline deposits of calcium carbonate and protein that are visible by light microscopy within otic vesicles) appeared unaffected in eya4 morphant fish. Mismatch morpholino oligonucleotides against eya4 (MO3mis and MO10mis) did not affect otic vesicle development (data not shown) in more than 100 zebrafish.

Sensory hair cell development in the otic vesicle was assessed by fluorescence-labeled phalloidin at 72 hpf. Staining of wild-type embryos(n=10) revealed five to seven delicate actin-rich stereocilia of hair cells projecting into each ampulla from the underlying anterior, lateral and posterior cristae (Fig. 4A),and an average of 38.3±0.9 hair cells in the anterior macula(n=6). Hair cells in the posterior macula were not assessed due to different imaging depths. Embryos treated with MO3mis or MO10mis had normal hair cell numbers (data not shown). By contrast, fluorescence labeled-phalloidin staining of eya4 morphant fish (n=13)showed few or no stereocilia on cristae (anterior cristae, n=1.7±0.7; lateral cristae, n=1.4±0.6;posterior cristae, n=1.3±0.4; Fig. 4B), and fewer hair cells in the anterior macula (n=24.5±2.5; P≤0.0013).

To determine whether reduced hair cell numbers in eya4 morphants reflected death of these specialized cells, we stained zebrafish embryos(n>15, each genotype) with Acridine Orange at 24 and 48 hpf (see Fig. S2 in the supplementary material). At 24 hpf, most otic vesicles from wild-type embryos had few if any Acridine Orange-stained cells. The average numbers of stained cells per otic vesicle were 0.2±0.6 (24 hpf) and 0.4±1.2 (48 hpf). In eya4 morphant fish, there were comparable average numbers of stained cells per otic vesicle: 0.3±0.7 (24 hpf; P=0.9 versus wild type) and 0.1±0.7 (48 hpf; P=0.4 versus wild type). We interpreted these data to indicate that eya4deficiency did not promote cell death.

Because eya4 is also expressed in neuromasts of the lateral line sensory system (Schonberger et al.,2005), we assessed neuromast hair cells labeled with the vital dye DASPEI (Whitfield et al.,1996). At 72 hpf, wild-type zebrafish (n=13) had 10-17 neuromasts on the lateral line sensory system of each flank, and each neuromast contained approximately five hair cells(Fig. 5A,C). By contrast, eya4 morphant fish (MO3eya4, n=6; MO10eya4, n=7)had fewer than six neuromasts along the lateral line sensory system of each flank, and each neuromast contained only one or two hair cells (MO3eya4, P=2×10^-21; MO10eya4, P=5×10^-29; Fig. 5B,C). Neither MO3mis (n=4 fish) nor MO10mis(n=6 fish) altered the total number of neuromasts or the hair cell numbers per neuromast (P=0.11, MO3mis versus wild type; Fig. 5C).

The otic vesicle and the lateral line sensory system are required for a normal startle reflex in zebrafish, or a `tail-flip' escape response that is elicited by clicks, tapping or vibrational stimuli(Whitfield et al., 2002). Wild-type zebrafish exhibit this response by 72 hpf, and swim with coordinated movements away from startle stimuli (see Movie 3 in the supplementary material). We assessed the individual responses of 80 wild-type and morphant fish (72 hpf; see Table S1 in the supplementary material). All 80 wild type and 25 eya4 morphant fish had normal startle reflexes. However, 70%(n=55) of the eya4 morphant fish had abnormal startle reflexes (P=2.1×10^-23) and swam with un-coordinated movements in small circles (see Movie 4 in the supplementary material). Normal startle responses were preserved in MO3mis- or MO10mis-treated fish(n>100 morphant fish; data not shown).

Six Na⁺/K⁺-ATPase subunits are expressed in the zebrafish otic vesicle early in embryogenesis(Blasiole et al., 2003): atp1b2b, atp1a1a.1, atp1a1a.2, atp1a1a.4, atp1a1a.5 and atp1b1a. Five of these genes have developmental expression patterns in the otic vesicle that differ from eya4 expression patterns,whereas the temporal and spatial expression pattern of atp1b2b(Blasiole et al., 2003)mirrored eya4 expression (24 and 72 hpf). Hypothesizing that eya4 might regulate expression of this β2b subunit of Na⁺/K⁺-ATPase, we first confirmed atp1b2bexpression in the otic vesicle by whole-mount in situ hybridization (72 hpf). We also identified atp1b2b expression in the lateral line neuromasts,as well as in the retina and somites (Fig. 6A,B,D). Expression of atp1b2b in the heart was low and was detected only by section in situ hybridization, with ventricular expression of atp1b2b exceeding atrial expression(Fig. 6D). To independently validate cardiac atp1b2b expression, the plasmid pβ2bprom(containing a 2.5-kb promoter fragment of atp1b2b upstream of EGFP) was injected into zebrafish embryos. Cardiac fluorescence was detected at 48 hpf and 72 hpf (Fig. 6E, data not shown), therein confirming atp1b2bexpression in the heart.

Levels of atp1b2b mRNA (assessed by semi-quantitative RT-PCR) in eya4 morphant fish were approximately 50% of wild-type levels(Fig. 6F). Levels of atp1a1a.1, atp1a1a.2, atp1a1a.4 and atp1b1a mRNAs were comparable in eya4 morphant fish and wild-type fish(Fig. 6F). Although the atp1a1a.5 mRNA level was also reduced(Fig. 6F) in eya4morphant fish, further studies did not focus on this gene because of the different spatial pattern of expression of atp1a1a.5(Blasiole et al., 2003) and eya4 (Fig. 1).

eya4 morphant atp1b2b expression was assessed by whole-mount in situ hybridization. Wild type and eya4 morphant atp1b2b expression was similar in the retina, somite, pectoral fin and mature neuromasts (data not shown). However, atp1b2b signal was notably missing or reduced in the sensory epithelium of the malformed eya4 morphant otic vesicle (compare Fig. 6B with 6C).

If eya4 regulation of atp1b2b expression contributed to the observed developmental defects, we reasoned that attenuated expression of this Na⁺/K⁺-ATPase subunit would mimic the otic vesicular, neuromast and cardiac phenotypes found in eya4 morphant fish. Two antisense morpholino oligonucleotides directed against splice donor sites at the junction of exon 1-intron 1 (MO1β2b) and exon 4-intron 4(MO4β2b) of the atp1b2b gene were constructed and studied(Fig. 7). RT-PCR and sequencing analysis confirmed abnormal atp1b2b splicing (data not shown),resulting in the deletion of 346 nucleotides and excision of the translation initiation site by MO1β2b, and the deletion of 32 nucleotides, which produced a frameshift in exon 4, by MO4β2b.

Zebrafish injected with MO1β2b (n=42) or MO4β2b(n=70) had significantly reduced atp1b2b expression (data not shown) and indistinguishable phenotypes. atp1b2b morphant fish demonstrated diminutive, malformed otic vesicles and under-developed semicircular canals resembling the anatomical defects found in eya4morphant fish (Fig. 2, Fig. 3E,F). In addition, about 43% of atp1b2b morphant fish lacked one or both otoliths, whereas wild-type fish and all eya4 morphant fish had two normally developed otoliths (Fig. 7A-C). The sensory cristae in 25 atp1b2b morphant fish contained an average of three to four hair cells per cristae (comparable to the average of <3 hair cells per cristae in eya4 morphants). Among the 27 atp1b2bmorphants in which neuromast hair cells were identified(Fig. 7D), there were significantly fewer hair cells than in wild-type neuromasts(P=0.001).

Acridine Orange staining was used to assess cell death in the atp1b2b morphants at 24 and 48 hpf (see Fig. S2 in the supplementary material). Only background level staining was observed at both stages (24 hpf,average number of stained cells=0.1±0.3 per otic vesicle, n=16 fish, P=0.3 versus wild type; 48 hpf, average number of stained cells=0.2±0.5 per otic vesicle, n=20 fish, P=0.4 versus wild type; Fig. S2 in the supplementary material), indicating that hair cell death was not increased in atp1b2b morphant fish.

Startle responses (see Movie 5 in the supplementary material) were abnormal in more than 85% (n=61) of atp1b2b morphant fish(P=7.2×10^-11 versus wild type). In 52 atp1b2b morphant fish, startle responses were completely absent,while two morphant fish had uncoordinated movements similar to that observed in eya4 morphant fish.

Cardiac phenotypes in atp1b2b morphant fish also recapitulated those in eya4 morphant fish (n=9): ventricular chambers were smaller than wild-type hearts (P=0.05 versus wild-type systolic diameter; P=0.004 versus wild-type diastolic diameter) and pericardial effusions were present in 77% (99 of 128) atp1b2bmorphant fish (Fig. 7B,C,E). Heart rates were also significantly slower (P=0.003) in atp1b2b morphants (n=9) than in wild-type fish(Fig. 7E) or eya4morphant fish, which have heart rates comparable to wild type(Schonberger et al.,2005).

We reasoned that if eya4 regulated atp1b2b (either directly or indirectly), then eya4-atp1b2b double morphants should replicate and perhaps accentuate the phenotypes of either an eya4 or atp1b2b single morphant. By contrast, if these molecules functioned in independent developmental pathways, much more severe phenotypes would be expected in the double morphants than in either of the single morphant fish.

Lower doses of eya4 and atp1b2b morpholino oligonucleotides than those previously used (so as to minimize non-specific morpholino toxicity and to minimize the potential for lethal heart failure)were co-injected into embryos and resultant phenotypes were assessed. Dose titration of MO3eya4 and MO1β2b indicated that fish injected with either 0.5 ng of MO3eya4 or 1.8 ng of MO1β2b were viable and developed 12-15 neuromasts on each flank (n>9), with significantly reduced(average=3) hair cells per neuromast (MO3eya4, P=2×10^-6 versus wild type; MO1β2b, P=7×10^-6 versus wild type). Normal hair cell numbers(six or seven per cristae) were observed in the mono-morphant otic vesicles(MO3eya4, n=22; MO1β2b, n=15; P=not significant; see Table S1 in the supplementary material). In fish(n=76) treated with MO1β2b (1.8 ng), seven morphants lacked one otolith and one morphant lacked both otoliths. Small pericardial effusions were found in mono-morphant fish (MO3eya4, n=31; MO1β2b, n=60), and only one MO1β2b mono-morphant demonstrated severe heart failure (versus 77% observed with a higher MO1β2b dose).

Embryos co-injected with both 0.5 ng of MO3eya4 and 1.8 ng of MO1β2b had hair cell numbers that were not significantly different from fish injected with identical doses of single morpholinos. Double-morphant embryos had 12-16 neuromasts on each flank with an average of three hair cells per neuromast(n=13, P=0.66 versus MO3eya4 and P=0.68 versus MO1β2b) and an average of six to seven hair cells per cristae,(n=17, P=not significant; see Table S1 in the supplementary material). Of 64 double morphants, two lacked one otolith and one lacked both otoliths. Only three double morphants showed severe heart failure.

Because the double-morphant studies supported the model that eya4regulated atp1b2b, we determined whether overexpression of atp1b2b mRNA could rescue phenotypes produced by eya4deficiency. Embryos injected with in vitro transcribed atp1b2b mRNA(50-70 pg) showed no malformations (data not shown). Embryos were therefore co-injected with MO10eya4 and 70 pg atp1b2b mRNA, and otic vesicle morphology, cardiac function and startle reflexes were characterized(Table 1).

Otic vesicle formation was markedly improved in fish (72 hpf) co-injected with MO10eya4 and atp1b2b mRNA. Unlike the severely malformed otic vesicle in eya4 morphant fish, approximately 50% of co-injected fish had otic vesicles that were smaller but otherwise indistinguishable from those of wild-type fish. Notably, the development of epithelial pillars and the formation of semicircular canals was rescued(Fig. 3G-I). Co-injected fish had normal numbers of sensory hair cells in the otic vesicle (n=5)and neuromasts (n=17) that were not significantly different from wild-type fish (Table 1). Startle responses in co-injected fish were also improved: 83% (103/123) of fish treated with MO10eya4 and atp1b2b mRNA had normal swimming patterns (see Movie 6 in the supplementary material; Table 1).

Cardiac structure and function was also improved in fish co-injected with MO10eya4 and atp1b2b mRNA (Fig. 8A,B). Atrial and ventricular sizes were normal and pericardial effusions were completely absent in 59% (72/123) of co-injected embryos,whereas only 5% (5 out of 80) of eya4 morphant fish had normal heart structure (P=1.8×10^-8; Table 1).

We report an essential role for eya4 expression in the development of the zebrafish otic vesicle sensory epithelia and neuromasts. Eya4 deficiency attenuated hair cell numbers, disrupted maturation of the otic vesicle and perturbed normal sensory responses. We suggest that the atp1b2b gene is a direct or indirect downstream target of Eya4 - a conclusion based on the colocalized temporal and spatial expressions of eya4 and atp1b2b, the diminished atp1b2b expression in eya4 morphant fish, and the similar phenotypes found in zebrafish treated with either atp1b2b or eya4 antisense morpholino oligonucleotides. Moreover, eya4-atp1b2b double morpholino oligonucleotides did not demonstrate a synergistic effect and overexpression of atp1b2b mRNA partially compensated for eya4 deficiency,restoring both sensory hair cells, and cardiac development and function.

There is substantial evidence that Eya proteins are important for auditory system development in multiple vertebrate species. Human EYA4mutations cause sensorineural hearing loss either in isolation or with cardiomyopathy; EYA1 mutations cause Branchio-Oto-Renal syndrome(BOR) with sensorineural hearing loss, craniofacial and kidney defects(Abdelhak et al., 1997),phenotypes that are largely recapitulated in Eya1-deficient mice(Xu et al., 1999). In zebrafish, eya1 gene mutations cause the dog-eared phenotype(Whitfield et al., 1996),which is characterized by small otic vesicles, malformed semicircular canals,reduced numbers of hair cells in the otic vesicle and the lateral line neuromasts, and diminutive jaw structure(Kozlowski et al., 2005). The otic vesicle and neuromast malformations are remarkably similar to those found in eya4 morphant fish, a result that we interpreted to indicate participation by eya1 and eya4 in a common regulatory network during development of the zebrafish sensory system. However, the broader pattern of eya1 expression, diffusely throughout the ventral part of the otic vesicle at 24 hpf (Sahly et al., 1999), in contrast to the restricted focal eya4expression in regions that will form the sensory epithelium, implied that each of these molecules had specific roles in otic vesicle development. In support of these distinct functions, we note that unlike eya1-deficient animals, eya4 morphants did not show ectopic cell death in the developing otic vesicle. Selective functions by these transcriptional co-activators is further evident from their distinct extra-sensory phenotypes that involve the heart (eya4)(Schonberger et al., 2005) or mandible (eya1) (Kozlowski et al., 2005).

Eya proteins do not bind DNA directly, but require both Six and Dachshund transcription factors to mediate regulatory effects. Because the zebrafish genome encodes 10 six genes(Bessarab et al., 2004; Kobayashi et al., 2000; Seo et al., 1999; Seo et al., 1998a; Seo et al., 1998b),identification of the specific Six proteins that interact with Eya4 has been problematic. Two gene family members, six1 and six4.1, are expressed in the zebrafish otic vesicle(Bessarab et al., 2004; Kobayashi et al., 2000), but neither has an expression pattern that matches that of eya4. At 24 hpf, six1 is diffusely expressed in the ventral edge of the otic vesicle and expression decreases subsequently(Bessarab et al., 2004); six4.1 is expressed in the maculae and the vestibular/acoustic ganglia at 37 hpf (Bessarab et al.,2004; Kobayashi et al.,2000). six4.1 is also weakly expressed in the semicircular canals (at 60 hpf) (Bessarab et al., 2004; Kobayashi et al., 2000), but not within the cristae where eya4expression localizes.

Because insufficient information about Eya4-binding partners hindered the identification of target genes, we selected candidate genes for study. Because Na⁺/K⁺-ATPase has important roles in hearing and cardiac physiology, organ systems perturbed by human EYA4 mutations,zebrafish genes encoding α and β subunits were evaluated. Six of fifteen Na⁺/K⁺-ATPase α and β subunit genes are expressed in the zebrafish otic vesicle(Blasiole et al., 2003), but only two, atp1b2b and atp1a1a.5, had altered RNA levels in eya4 morphant fish (Fig. 6). The expression pattern of atp1a1a.5 was distinct from that of eya4. At 24 hpf, atp1a1a.5 expression was found diffusely within the anteroventral portion of the otic vesicle, whereas eya4 expression was punctuate(Fig. 1A). Later in development, atp1a1a.5 is found in the dorsolateral septum and the anterior, posterior and lateral semicircular pillars(Blasiole et al., 2003),regions that do not express eya4. Taken together, we concluded that eya4 regulation of atp1a1a.5 occurred in tissues other than the otic vesicle. By contrast, atp1b2b expression colocalized with eya4, and, based on previous studies indicating that atp1b2bdeficiency delayed development of the semicircular canal(Blasiole et al., 2006), we concluded that eya4 regulated atp1b2b during otic vesicle development.

This model was supported by shared phenotypes in eya4 and atp1b2b morphant fish: attenuated hair cell development in otic vesicles and neuromasts, and malformations of the semicircular canals. Consistent with these data, atp1b2b morphant fish, like eya4morphant fish, had sensory function abnormalities (see Movie 5 in the supplementary material). In addition, the phenotypes in eya4-atp1b2b double morphants resembled those of eya4 and atp1b2b mono-morphants. To corroborate that eya4 and atp1b2b function in a shared developmental pathway,we performed studies that paralleled the rescue experiments of atp1a1a.1-deficient zebrafish by exogenous zebrafish atp1a1a.1 mRNA or rat Atp1a1 mRNA(Blasiole et al., 2006), and assessed whether exogenous atp1b2b could rescue eya4morphant fish. Because of zebrafish atp1b2b is only 65% identical to its mammalian homolog Atp1b2, we overexpressed zebrafish atp1b2b in eya4 morphant fish and found both the cardiac and sensory system phenotypes were rescued(Fig. 8, Table 1). Taken together, we concluded that atp1b2b is a direct or indirect target of eya4.

We noted a correlation (Table 1, Fig. 4A; see also Table S1 in the supplementary material; r>0.4 for all two-way comparisons) between the numbers of hair cells in different otic vesicle cristae and the anterior macula of eya4 morphant fish (data not shown), but whether this correlation reflects independent or related developmental processes is unclear. Perhaps eya4 is essential for development of the primordial cells that give rise to sensory hair cells throughout the otic placode, a model that is supported by expression early in development (14 hpf; see Fig. S1A in the supplementary material). Alternatively, eya4 expression could be required for independent processes that occur in the maculae and cristae and impact on hair cell development.

The diminutive size of the otic vesicle in eya4 and atp1b2b morphant fish, as well as the maldeveloped epithelium pillars and semicircular canals (Figs 2, 3), implicate roles for eya4 beyond development of the sensory epithelium. That broad consequences would result from the disruption of eya4, a transcriptional co-activator that probably regulates many genes, was not entirely unexpected. However, that multiple structural defects arose because of the dysregulation of a Na⁺/K⁺-ATPase subunit indicates that the ionic milieu is crucial for multiple aspects of otic vesicle development.

In situ hybridization experiments and a reporter assay also demonstrated atp1b2b expression in the embryonic zebrafish heart(Fig. 6D,E). Three other Na⁺/K⁺-ATPase subunit genes are expressed in the zebrafish heart: atp1a1a.1, atp1a2 and atp1b1a(Cheng et al., 2003; Shu et al., 2003; Yuan and Joseph, 2004). Whereas atp1a2 is believed to regulate cardiac laterality(Shu et al., 2003), the function of atp1b1a remains elusive(Cheng et al., 2003). Heart and mind and small heart phenotypes (diminutive hearts with slow beating rates) result from atp1a1a.1 mutations(Shu et al., 2003; Yuan and Joseph, 2004). atp1b2b morphant fish also had small hearts and slower beating rates,raising the possibility that atp1b2b interacts with the atp1a1a.1 subunit in the heart. This subunit pairing has also been suggested to be important for otic vesicle development(Blasiole et al., 2006). Although the function of Na⁺/K⁺-ATPase in cardiac biology remains incompletely understood, we suggest that human EYA4mutations affect the heart in part by altering the expression of Na⁺/K⁺-ATPase.

Some atp1b2b morphant fish had two additional phenotypes that were not observed in eya4 morphant fish. Most atp1b2b morphant fish lacked either one or both otoliths(Fig. 7B,C), whereas otolith agenesis was never observed in eya4 morphant fish. Previous studies(Blasiole et al., 2006) of atp1b2b morphant fish reported the presence of at least one otolith,a difference that may reflect the selective efficacy of atp1b2bmorpholino oligonucleotides that targeted splicing (this report) or translation initiation. In addition, a slower heart rate was observed in the atp1b2b morphant fish (Fig. 7E) but not in the eya4 morphant fish(Schonberger et al., 2005). Although other explanations are possible, we suggest that the additional phenotypes in atp1b2b morphant fish are due to a considerably greater(>90%) reduction in atp1b2b mRNAs than is seen in eya4morphants (50% reduction of atp1b2b mRNAs).

We are intrigued by the observation that zebrafish Eya4 regulates the same Na⁺/K⁺-ATPase subunit in both the heart and the hair cell in the sensory system. In humans, EYA4-deficiency impacts both of these organs, raising the possibility that the regulation of Na⁺/K⁺-ATPase subunits in mammals has been conserved throughout evolution. Despite the improved phenotypes in eya4morphant fish produced by atp1b2b mRNA, we expect that other genes expressed in these tissues are also regulated by eya4. Further understanding of the composition of Eya4-Six-Dach complexes and their DNA recognition sites should yield more insights into the regulatory pathways influenced by Eya4 that account for human sensorineural hearing loss and dilated cardiomyopathy.

We greatly appreciate the scientific comments from David A. Conner and his critical reading of the manuscript. We are grateful to Jost Schönberger for his valuable discussions. We thank José Rivera-Feliciano, Jenna Galloway, Sarah Keller, Haobing Wang, Yariv Houvras, Bruce Barut and Susanne Bartlett for their technical assistance. This work was supported by grants from the Howard Hughes Medical Institute (C.E.S.) and the National Institutes of Health (J.G.S. and C.E.S.).