Zebrafish pax5 regulates development of the utricular macula and vestibular function



The zebrafish otic vesicle initially forms with only two sensory epithelia, the utricular and saccular maculae, which primarily mediate vestibular and auditory function, respectively. Here, we test the role of pax5, which is preferentially expressed in the utricular macula. Morpholino knockdown of pax5 disrupts vestibular function but not hearing. Neurons of the statoacoustic ganglion (SAG) develop normally. Utricular hair cells appear to form normally but a variable number subsequently undergo apoptosis and are extruded from the otic vesicle. Dendrites of the SAG persist in the utricle but become disorganized after hair cell loss. Hair cells in the saccule develop and survive normally. Otic expression of pax5 requires pax2a and fgf3, mutations in which cause vestibular defects, albeit by distinct mechanisms. Thus, pax5 works in conjunction with fgf3 and pax2a to establish and/or maintain the utricular macula and is essential for vestibular function. Developmental Dynamics 235:3026–3038, 2006. © 2006 Wiley-Liss, Inc.


The vertebrate inner ear is a conserved organ system comprising a series of interconnected chambers, each of which primarily mediates either vestibular or auditory function. Each chamber contains a sensory patch consisting of sensory hair cells and supporting cells. Hair cells synapse with neurons of the statoacoustic ganglion (SAG), or the VIIIth cranial nerve, axons of which project to processing nuclei in the hindbrain (reviewed by Lewis et al.,1985).

All sensory patches originate from a prosensory region in the ventromedial wall of the otic vesicle, and each sensory patch subsequently differentiates with specific structural and functional attributes (Lewis et al.,1985; Fekete and Wu,2002; Riley and Phillips,2003; Barald and Kelley,2004). Hair cells in the semicircular canals have very long cilia embedded in a gelatinous cupula that senses angular acceleration through fluid motion in the canal. Hair cells in the cochlea of mammals and birds have shorter cilia embedded in a tectorial membrane that transmits sound vibrations. Hair cells in the sensory maculae of the utricle, saccule, and lagena bear cilia that contact crystalline otoliths that transmit forces caused by linear acceleration, gravity and, in fish, sound vibrations (Fay and Popper,1980; Popper and Fay,1993).

Despite the dual sensory capacity of fish maculae, studies in zebrafish (Danio rerio) and the closely related goldfish (Carassius auratus) suggest that the functions of different maculae are not identical. Although it is likely that all maculae contribute to hearing, the saccule is the primary auditory sensor, particularly at frequencies above several hundred Hz (Popper et al.,2003). Zebrafish, like other ostariophysan fishes, use a series of bones (Weberian ossicles) that are thought to transmit sound vibrations from the swim bladder to the saccule (Popper et al.,2003). Disruption of the Weberian ossicles or swim bladder results in partial loss of hearing (Fay and Popper,1974; Bang et al.,2002; Zeddies and Fay,2005). The utricle probably also has some role in hearing, possibly in sound source localization (Popper et al.,2003). However, unlike the saccule, the utricle is essential for vestibular function in zebrafish larvae, as shown by analysis of monolith (mnl) mutants (Riley et al.,1997; Riley and Moorman,2000). These mutants usually form saccular otoliths but not utricular otoliths, which ablates all discernible vestibular function and is lethal during larval stages. However, experimental manipulations that restore utricular otoliths rescue both vestibular function and viability in mnl mutants, even if saccular otoliths are ablated instead. Thus, in zebrafish, the utricle is especially important for vestibular function, whereas the saccule has a more pronounced role in hearing. Sensory cristae within the semicircular canals are also devoted to vestibular function but these do not become functional until after 30 days postfertilization (dpf; Beck et al.,2004).

Differential gene activity in the otic vesicle presumably underlies development of the characteristic structure and function of each chamber and sensory epithelium. Indeed, many candidate genes have been identified that show expression in only one or a small subset of sensory patches. However, loss of function of such genes often causes severe morphogenetic defects that preclude assessment of functional output. For example, Otx1 is expressed in the presumptive lateral crista and Otx1−/− mutant mice do not produce the lateral crista or a normal lateral semicircular canal (Morsli et al.,1999). Thus, it is not clear whether Otx1 plays an ancillary role in programming the lateral crista or its associated neurons to specialize as a vestibular endorgan.

In zebrafish, sensory epithelia form at an early stage before extensive morphogenesis of the various chambers of the inner ear (Haddon and Lewis,1996; Whitfield et al.,2002). The nascent otic vesicle contains only two sensory patches corresponding to the utricular (anterior) and saccular (posterior) maculae. We have been interested in identifying genes required for regulating functional specialization of these two sensory epithelia. One candidate gene, pax5, is initially expressed in the anterior end of the nascent otic vesicle (Pfeffer et al.,1998) and later becomes localized to the utricular macula. This pattern suggested that pax5 might be involved in development, maintenance, or functional organization of the utricular macula. To investigate pax5 function, we cloned the full sequence of pax5 cDNA and performed loss of function studies using antisense morpholinos (MOs). Knocking down pax5 caused vestibular defects in zebrafish larvae without altering morphogenesis of the ear or the ability to hear. We show that vestibular deficits result from defects in maintaining utricular hair cells, with secondary defects in the pattern of SAG neuronal processes in the utricular macula.


Cloning of Zebrafish pax5

The known sequence for zebrafish pax5 cDNA was incomplete, with sequences missing from both the 5′ and 3′ cDNA ends (Pfeffer et al.,1998). We completed cloning of the pax5 sequence (see Experimental Procedures section). Analysis of multiple cDNA clones revealed two distinct splice isoforms, pax5-variant 1 (pax5-v1) and pax5-variant 2 (pax5-v2; Fig. 1A). pax5-v1 corresponds to full-length pax5 cDNA. pax5-v2 has a partial paired domain caused by splicing out the second exon (nucleotides 47–212). This splice variant is predicted to use an alternative translation start codon in exon 3. In mouse, six splice variant forms are known. The two zebrafish variants, pax5-v1 and pax5-v2, are homologous to mouse splice variants Pax-5a and Pax-5b, respectively, suggesting that mechanisms for alternative splicing of pax5 have been conserved (Zwollo et al.,1997). The relative abundance of cloned cDNAs suggests that pax5-v1 (8 of 10 clones) is more prevalent than pax5-v2 (2 of 10 clones).

Figure 1.

cDNA structure and expression of pax5. A: General structure of pax5 splice variants. Brackets indicate exon boundaries. Conserved functional domains, paired (PD), octapeptide (OP), homeo (HD), transactivation (TAD), and inhibitory (ID) are marked. Putative translation start sites (M) are indicated. Binding sites for translation-blocking (TB1 and 2) and splice-blocking (SB1, 2, and 3) morpholinos are shown. Newly identified 5′ (i) and 3′ (ii) sequences are shown in comparison with fugu pax5. Zebrafish and fugu sequences are 100% identical at the amino acid level. B–E: Expression of pax5 in the otic placode at 17 hours postfertilization (hpf; B), in the otic vesicle at 24 hpf (C) and in the utricular macula at 48 hpf (D,E). E: Enlarged view of boxed area in D. Hair cell (hc) supporting cell layers are marked. Arrow, weak expression in the saccule. (A) Dorsal, (B) dorsolateral, and (C,D) lateral views, with anterior to the left. Scale bar = 30 μm in A, 40 μm in B,C, 12.5 μm in D.

Expression of pax5 in the Otic Vesicle

pax5 is first detected in the anterior end of the otic placode at approximately 17 hours postfertilization (hpf), just before formation of the otic vesicle (Fig. 1B). By 24 hpf, the anterior quarter of the otic vesicle shows a uniformly high level of pax5 expression (Fig. 1C). Expression is subsequently restricted to the anterior (utricular) macula and remains in the macula until at least 72 hpf (Fig. 1D,E). At these later stages, all cells in the utricular macula express pax5, but hair cells show higher expression than supporting cells (Fig. 1E). In addition to the predominant domain in the utricle, a small number of saccular hair cells also express pax5 (Fig. 1C). This posterior expression is maintained through at least 48 hpf (not shown).

Vestibular Defects in pax5-Depleted Larvae

To study pax5 function in the inner ear development, two morpholinos were designed to block translation from two putative translation start sites and three were designed to block splicing of sequences encoding the paired domain or the homeodomain (Fig. 1A). Each of these morpholinos, used individually, disrupted vestibular function (discussed below) but varied in efficiency. However, a cocktail of all five morpholinos proved most efficient and was used for the remainder of this study. Embryos injected with pax5-MO cocktail (pax5 morphants) show no obvious morphological defects. The otic vesicle is normal in size, and otoliths form in the correct positions at the right time. Because of the predominant expression of pax5 in the utricle, we assayed the vestibular-dependent functions of balance, motor coordination, and swim bladder inflation (Riley and Moorman,2000). For comparison we also examined mnl mutants, which show a severe and permanent loss of vestibular function due to the lack of utricular otoliths (Riley and Grunwald,1996; Riley et al.,1997; Riley and Moorman,2000). By all three assays, pax5 morphants are delayed by a day or more in development of vestibular function (Fig. 2A–C). This finding does not reflect a general developmental delay because morphological and molecular milestones occur on time (see below). Although many pax5 morphants eventually display normal vestibular behavior, approximately 20% never do so and continue to show severely impaired vestibular function through at least 7 dpf (Fig. 2). These data support the hypothesis that pax5 is required for development and/or function of the vestibular system.

Figure 2.

A–C: Assessment of vestibular and auditory function. Development of balance (A), motor coordination (B), and swim bladder inflation (C) in wild-type (n = 173), pax5-morphant (n = 330), lia/lia (n = 110) and mnl/mnl (n = 238) embryos between 3 and 7 days postfertilization (dpf). Data show the means and standard errors of two independent experiments. D: Frequency range and sensitivity of hearing in wild-type larvae and pax5 morphants at 5 dpf. Only pax5 morphants with severe vestibular deficits were used in D.

In contrast to vestibular function, acoustic function appears normal in pax5 morphants. Zeddies and Fay (2005) recently described an assay to test acoustically evoked behavioral responses (AEBRs) in zebrafish larvae and adults. Beginning at 5 dpf, wild-type larvae respond at the same levels over the same frequency range as adult zebrafish. Because pax5 shows only minor expression in the saccular macula, we hypothesized that hearing should be relatively normal. Indeed, even pax5 morphants with severe and persistent vestibular deficits appear to respond normally to sound on 5 dpf and later (Fig. 2D, and data not shown). These data suggest that the vestibular deficits in pax5 morphants are not caused by global or nonspecific defects but instead reflect a specific requirement in the utricular epithelium for vestibular function.

Otic Vesicle Patterning in pax5 Morphants

The vestibular defects in pax5 morphants could be caused by perturbation of general patterning of the otic vesicle. To test this possibility, we examined several markers of otic vesicle patterning. nkx5.1, which marks the anterior end of the otic vesicle, is expressed normally in pax5 morphants (Fig. 3A,B), as is zp23, a marker of the posterior medial wall adjacent to rhombomere (r) 5 and r6 of hindbrain (Fig. 3C,D). Patterning of dorsoventral and mediolateral axes also appear normal as demonstrated by expression of a dorsomedial marker, dlx3b, and a ventrolateral marker, otx1 (Fig. 3E–H). In addition, sensory maculae and cristae appear to form on time and express appropriate markers (Fig. 3I–L). Several aspects of inner ear patterning, including otic expression of pax5, depend on Fgf3 from the hindbrain (Kwak et al.,2002), so we also examined hindbrain patterning. Expression of fgf3, as well as other hindbrain markers such as krox20, are normal in pax5 morphants (Fig. 3M–P, and data not shown). Thus, hindbrain patterning and general features of otic vesicle patterning appear normal in pax5 morphants.

Figure 3.

A–H: Inner ear and hindbrain patterning in pax5 morphants. Expression of nkx5.1, zp23 (arrows mark otic domain), dlx3b, and otx1 in the otic vesicle of uninjected control embryos (A,C,E,G) and pax5-morphants (B,D,F,H) at 24 hpf. I,J: Macular expression of fgf8 in control (I) and pax5-morphant (J) at 48 hours postfertilization (hpf). K,L: Expression of msxC in cristae (arrowheads) of control (K) and pax5-morphant (L) at 72 hpf. M–P:krox20 and fgf3 expression at 13.5 hpf (nine-somite stage) in the hindbrain of uninjected control embryos (M,O) and pax5 morphants (N,P). Images show dorsolateral (A–F), lateral (I–L), and dorsal (G,H,M–P) views with anterior to the left. MHB, midbrain–hindbrain border; r3, rhombomere 3; r4, rhombomere 4. Scale bar = 35 μm in A–J, 65 μm in K,L, 160 μm in M–P.

Formation of SAG Neurons in pax5 Morphants

Another possible cause of vestibular deficits is failure to form the SAG. SAG neuroblasts are initially specified in the ventral region of the otic vesicle in a region that partially overlaps with the domain of pax5 expression (Haddon and Lewis,1996). Neuroblasts delaminate from the otic vesicle and migrate to a position between the anteromedial wall of the otic vesicle and hindbrain. neurogenin1(ngn1) encodes a basic helix–loop–helix transcription factor required for SAG specification and is first expressed at 18 hpf (Andermann et al.,2002). ngn1 is expressed normally in pax5 morphants (Fig. 4A,B). After delamination, SAG neuroblasts express nkx5.1 (Adamska et al.,2000) and this pattern is also normal in pax5 morphants (Fig. 3A,B). Similarly, the number and position of SAG neuroblasts is normal at 30 hpf as shown by anti-Islet staining (Fig. 4C,D). Thus, depletion of pax5 does not alter production or migration of SAG neurons.

Figure 4.

Development of the statoacoustic ganglion (SAG). A,B: Expression of ngn1 in the otic vesicle (arrowheads) and SAG (arrows) in a control embryo (A) and pax5-morphant (B) at 24 hours postfertilization (hpf). C,D: Anti–Islet-1/2 staining of SAG neuroblasts (outlined) at 30 hpf. An average of 16.5 ± 4.2 neuroblasts were detected in control embryos (C) compared with 15.7 ± 3.5 in pax5 morphants (D). Arrows mark otoliths. E,F: Utricular maculae of a 48 hpf control embryo (E) and pax5 morphant (F), showing acetylated tubulin (green) relative to Pax2 in hair cell nuclei (red). White arrows in E mark axonal process projecting to hair cells and broader regions of staining (yellow arrows) are observed at the basal regions of hair cells, possibly associated with synapses. Specimen in F shows a misplaced hair cell (arrowhead) associated with a single thick SAG process (arrow). G,H: Central projections of SAG neurons visualized by injecting DiI (1,1-dioctadecyl-3,3,3,3 -tetramethylin-docarbocyanine perchlorate) into the utricular macula at 72 hpf. Schematic in G shows the site of DiI injection (orange arrow) and SAG projections relative to the ear. Wild-type larvae show either two discrete axonal bundles in the hindbrain (G, type-1) or more diffuse projection patterns (H, type-2), including smaller secondary branches indicated by arrows. I: Table 1, percentage of larvae showing type-1 or type-2 projection patterns in control embryos, pax5 morphants, mnl/mnl mutants, and lia/lia mutants. A–K: Images show dorsolateral (A,B,E,F), dorsal (C,D), and lateral (G–K) views, with anterior to the left. Scale bar = 50 μm in A,B,E,F, 30 μm in C,D, 12.5 μm in G–K.

Neuronal Targeting of SAG Neurons in pax5 Morphants

SAG neurons are bipolar neurons, sending processes into the hindbrain and sensory patches of the ear. Axonal processes to the hindbrain were visualized by injecting a lipophilic tracer, DiI (1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate), into utricular maculae at 3 dpf or later. Utricular SAG neurons initially extend their axons in a bundle to the hindbrain in a dorsoposterior direction. This axonal bundle splits into two main branches in the hindbrain, one ascending and the other descending (Fig. 4G). In some specimens, ascending and descending branches are compact and well organized (Fig. 4G, type-1). In approximately 60% of control larvae, the main branches are more diffuse and there are several additional minor branches projecting in parallel to the main branches (Fig. 4H, type-2). A similar distribution of type-1 vs. type-2 patterns is seen at 7 dpf, even though all control larvae show fully integrated vestibular function by 5 dpf (Fig. 2). Moreover, virtually identical patterns are observed in mnl mutants (Fig. 4I; Table 1), which are null for vestibular function (Riley and Moorman,2000). Thus, the distribution of type-1 vs. type-2 patterns is not influenced by the status of vestibular signaling or early maturation of the larval hindbrain. Similar SAG projections are seen in pax5 morphants, although type-2 patterns are slightly more frequent than in control larvae (Fig. 4I; Table 1). At present, we do not understand the significance of the two different projection patterns. Nevertheless, vestibular deficits in pax5 morphants do not appear to be caused by aberrant projections in the hindbrain.

We also examined SAG processes in developing maculae. Acetylated tubulin is localized in the cortex and cilia of hair cells, as well as axonal processes of SAG neurons (Fig. 4G,H). Acetylated tubulin staining is especially prominent in the basal part of hair cells where SAG neurons synapse (Fig. 4E). Whereas some pax5 morphants appear normal, half or more show a variety of defects in neural patterning in the utricle. For example, approximately half of pax5 morphants show loss of putative synapse staining on utricular hair cells and the number of SAG processes is reduced (Fig. 4F). Occasionally, processes can be seen projecting at oblique angles and fail to innervate any hair cells (not shown). In addition, approximately 20% of pax5 morphants show thick bundles of dendrites reaching to the luminal surface without contacting any hair cells (Fig. 4F). Anti–nerve cell adhesion molecule staining shows similar patterns of SAG axonal processes (data not shown). Innervation of the saccular macula is difficult to visualize because of its close proximity to the brightly stained hindbrain. However, SAG innervation of hair cells in cristae is normal in pax5 morphants (data not shown). Thus, variable defects in hair cell innervation primarily affect the utricle in pax5 morphants and could contribute to the observed vestibular deficits in pax5 morphants.

Formation of Hair Cells

We hypothesized that pax5 might also regulate development of utricular hair cells. To test this, embryos were stained with anti-Pax2 antibody, which labels nuclei of mature hair cells (Riley et al.,1999). In pax5 morphants, hair cells are produced normally in the utricular and saccular maculae at 24 hpf, but at later stages, the number of utricular hair cells is consistently reduced by 20–30% relative to uninjected controls (P < 0.05; Fig. 5A–C). In contrast, the number of saccular hair cells is normal through at least 72 hpf (Fig. 5C). To confirm these results, we used two other markers to stain hair cell cilia, anti-acetylated tubulin and phalloidin. The number of utricular hair cells detected by phalloidin or acetylated tubulin staining is slightly greater than Pax2 staining at all time points (Fig. 5G), probably because cilia form before high level accumulation of Pax2 in differentiating hair cells. However, these markers confirmed a 20–30% decrease in utricular hair cells in pax5 morphants (Fig. 5G). In the saccule, the number of Pax2-postivie cells does not change after 30 hpf (Fig. 5C), yet the number of hair cells detected by acetylated tubulin or phalloidin staining increases steadily (Fig. 5D–G). At 72 hpf, for example, there are only two to four Pax2-positive cells in the saccule, whereas 28 ± 4.8 hair cells are detected by phalloidin staining (Fig. 5G). We do not know the functional significance of the small number of Pax2-positive hair cells in the saccule but note that the pattern of Pax2-staining is similar to the pattern of pax5 expression. In any case, the number of saccular hair cells in pax5 morphants is not significantly different from the control (Fig. 5G). Thus, the deficiency of hair cells in pax5 morphants is limited to the utricle.

Figure 5.

Assessment of hair cell development. A,B: Anti-Pax2 staining in the otic vesicle in a control embryo (A) and pax5 morphant (B) at 48 hours postfertilization (hpf). C: Number of Pax2+ hair cells at the indicated stages (means and standard errors of at least three experiments, with at least 15 specimens/time-point/experiment). P values for the comparison of control embryos vs. pax5 morphants are as follows: utricle, P = 0.042 (30 hpf), P = 0.009 (36 hpf), P = 0.0007 (48 hpf), P = 0.017 (60 hpf); and saccule, P = 0.136 (30 hpf), P = 0.138 (36 hpf), P = 0.05 (48 hpf), P = 0.28 (60 hpf). D,E: Rhodamine–phalloidin staining in the saccular macula of a control embryo (D) and pax5 morphant (E) at 48 hpf. In control embryos (n = 23), there were 21.5 ± 5.0 hair cells in the utricular macula and 17.8 ± 9.7 in the saccular macula. In pax5 morphants (n = 42), there were 16.8 ± 4.5 hair cells in the utricle and 14.8 ± 7.5 in the saccule. F: Saccular maculae stained with anti-acetylated tubulin (green) and anti-Pax2 (red) in a pax5 morphant at 48 hpf. Only two hair cells are Pax2-positive (arrows). G: Hair cell numbers detected by anti-Pax2 or phalloidin staining in pax5 morphants and uninjected controls at 72 hpf. Data bars are color-coded as in C. H,I: Enlarged view of the utricular macula stained with anti-Pax2 at 48 hpf in a control embryo (H) and pax5 morphant (I). Basal edges of hair cell (hc) and supporting cell (sc) layers are indicated. J: Otic vesicle of pax5 morphant stained with anti-Pax2 at 36 hpf. Arrowheads mark misplaced hair cells. The ventral limit of the otic vesicle (ov) is indicated. Images show dorsolateral (A,B,D–F,J) and lateral (H,I) views with anterior to the left. Scale bar = 40 μm in A,B,D–F, 12.5 μm in H,I, 25 μm in J.

Anti-Pax2 staining also demonstrates that pax5 morphants have irregular arrangements of hair cells in utricular macula. In a variable fraction (22.9 ± 8.3%) of pax5 morphants, one or two hair cell nuclei are localized in the basal supporting cell layer or even outside of the otic vesicle (Fig. 5I,J). Of interest, misplaced hair cells are usually accompanied by the appearance of abnormal SAG processes (Fig. 4F). Ejection of hair cells undergoing apoptosis has been previously described in several species. In mouse and guinea pig, for example, apoptotic hair cells sink to the basal layer within the sensory epithelium (Sobkowicz et al.,1992,1997; Quint et al.,1998). Similarly, hair cells in zebrafish mind bomb (mib) mutants begin to die after 36 hpf and are extruded from otic vesicle to the underlying mesenchyme (Haddon et al.,1998). Therefore, the reduced number and misplaced position of hair cells in pax5 morphants could reflect elevated apoptosis.

pax5 and Cell Death in the Utricle

To examine the pattern of cell death, embryos were stained with the vital dye acridine orange (AO). Control embryos show very little AO staining in the otic vesicle between 30 and 72 hpf (Fig. 6A). Summing the patterns of 30 embryos shows a “hot spot” of cell death near the anteromedial wall of the otic vesicle, although some of these cells may lie outside the otic vesicle (Fig. 6G). Staining in other regions is very sparse. Only 5.4 to 7.7% of control embryos (depending on the stage) show AO-positive cells in the utricle. The overall pattern of AO staining is very similar in pax5 morphants, except that there are roughly fivefold more labeled cells in the utricular macula (Fig. 6G; 34 labeled cells in pax5 morphants vs. 7 in control embryos). On average, 31.2 to 37.1% of pax5 morphants (at 30 hpf and 48 hpf, respectively) show labeled cells in the utricular macula (Fig. 6B,E). Similarly, whole-mount terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) assays show that 40% of pax5 morphants have apoptotic cells in utricle at 48 hpf (Fig. 6C,D). The saccular macula shows little cell death in either uninjected embryos or pax5 morphants (Fig. 6G), showing that pax5-depletion specifically affects the utricular macula.

Figure 6.

Analysis of cell death in pax5 morphants. A,B: Acridine orange (AO) staining in the otic vesicle of a control embryo (A) and in the utricle of a pax5 morphant (B) at 48 hours postfertilization (hpf). White arrows indicate AO-positive cells, black arrows show otoliths. C,D: Terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) staining in the utricle of a control embryo (C) and a pax5 morphant (D) at 48 hpf. E: The arrowhead shows a TUNEL-positive cell. E,F: A pax5-morphant stained with acridine orange at 48 hpf (E) and subsequently stained with anti-Pax2 (F). An AO-positive cell appears in the same position as a misplaced hair cell (white arrows). The utricular macula is outlined in A–F. G: Cumulative data (n = 30) representing the frequency and distribution of AO-labeled cells in the otic vesicle of a wild-type control embryo, a pax5 morphant, and a noi/noi mutant at 48 hpf. The positions of labeled cells (red spots) were projected onto schematic maps of the otic vesicle. Positions of the utricular macula (u), saccular macula (s), anterior crista (ac), lateral crista (lc), and posterior crista (pc) are indicated. All images show lateral views, with anterior to the left. Scale bar = 40 μm in A, 25 μm in B–D, 30 μm in E,F, 50 μm in G.

We hypothesized that dying cells seen in the utricle of pax5 morphants correspond to misplaced hair cells. To test this idea, AO-stained embryos were photographed at 48 hpf to record positions of dying cells, then fixed and stained for Pax2. pax5 morphants with no cell death show normal hair cell arrangements (n = 14). In contrast, misplaced Pax2-positive hair cells were frequently detected in the corresponding position where AO-positive cells had been detected (12 of 19 embryos, Fig. 6E,F). The remainder of AO staining was detected in normally positioned hair cells in the utricular macula (3 of 19) or Pax2-negative cells within the basal layer (presumptive support cells or nascent hair cells, 4 of 19; data not shown). These data support the hypothesis that utricular hair cells in pax5 morphants undergo an elevated rate of apoptosis and that dying hair cells are ejected from the utricular macula.

Elevated AO staining persists through 3 dpf in pax5 morphants but declines to normal by 4 dpf (not shown). This finding probably reflects diminishing capacity of the injected morpholinos (MOs) to knock down pax5 function.

pax5 mRNA Rescues Early Defects in pax5 Morphants

To confirm specificity of gene knockdown by pax5-MO, we coinjected pax5-MO with pax5-v1 and pax5-v2 mRNAs to try to rescue pax5 morphants. These mRNAs are impervious to splice-blocking MOs and when injected at high levels can overwhelm the effects of translation-blocking MOs. In control (nonmorphant) embryos, misexpression of pax5 has no effect on morphology. However, pax5 mRNA restores hair cell numbers to normal in pax5 morphants at 32 hpf (P = 0.47, wild-type vs. rescued embryos; Table 2). The fraction of embryos showing cell death in the utricular macula (21.7%) is reduced to half of the level otherwise seen in pax5 morphants, a significant difference (P ≤ 0.032). At 48 hpf, the effects of pax5 mRNAs are less evident (Table 2). This finding is probably because injected RNAs rarely persist beyond 24 to 30 hpf, whereas morpholinos often continue to function for at least 3 days. The ability to rescue balance and coordination could not be evaluated due to the limited stability of injected mRNA. However, the finding that pax5 mRNA rescues early defects in the pax5 morphants validates the specificity of pax5-MO.

Table 2. Rescue of pax5 Morphants by pax5 mRNA Injectiona
  Controlpax5-MOpax5-MO +pax5 RNApax5 RNA
  • a

    P values are based on t-tests in comparison with control (*) or with pax5 morphants (‡).

  • b

    Based on the number of Pax2-positive hair cells; mean of two to three experiments ± standard error (n, total number of embryos examined). Percentages in boldface reflect values relative to the control.

  • c

    Mean (± standard error) of two experiments, except for 48 hpf control and 32 hpf mRNA-injection, which are means of one experiment each.

  • MO, morpholino; hpf, hours postfertilization; AO, acridine orange.

No. of hair cells in the utricular maculab32 hpf100%79.5%101.5%107.4%
 6.8 ± 0.8 (n = 92)5.4 ± 0.4 (n = 90)6.9 ± 0.7 (n = 35)7.3 ± 0.7 (n = 48)
  *P = 0.008*P = 0.474*P = 0.254
   P = 0.013 
48 hpf100%70.5%79.5%Not determined
 14.6 ± 1.1 (n = 30)10.3 ± 0.7 (n = 28)11.6 ± 0.8 (n = 19) 
  *P = 0.023*P = 0.045 
   P = 0.118 
No. of hair cells in the saccular maculab32 hpf100%100%100%100%
 2.3 ± 0.3 (n = 92)2.3 ± 0.2 (n = 90)2.3 ± 0.1 (n = 35)2.3 ± 0.3 (n = 48)
  *P = 0.492*P = 0.434*P = 0.493
   P = 0.242 
48 hpf100%104.8%95.2%Not determined
 2.1 ± 0.1 (n = 30)2.2 ± 0.3 (n = 28)2.0 ± 0.2 (n = 19) 
  *P = 0.386*P = 0.242 
   P = 0.242
% of embryos with AO-positive cells in the utricular maculac32 hpf5.4%37.1%21.7%4.0% (n = 35)
 ± 1.77 (n = 39)± 4.91 (n = 59)± 10.9 (n = 50) 
   P = 0.032 
48 hpf7.7%31.2%26.2%Not determined
 (n = 13)± 11.4 (n = 44)± 4.5 (n = 52) 
   P = 0.268 

Distinct Roles for pax2a and fgf3 in Regulating pax5 and Vestibular Function

Previous studies identified pax2a and fgf3 as upstream regulators of pax5. Knocking down fgf3 by morpholino injection diminishes expression of pax5 in the ear (Kwak et al.,2002). noi (pax2a) null mutants show a complete loss of pax5 expression in the ear (Pfeffer et al.,1998; and Fig. 7A). Therefore, we speculated that these mutants might display defects similar to those of pax5 morphants.

Figure 7.

Otic development in noi (pax2a) and lia (fgf3) mutants. A,E,F: Otic expression of pax5 in noi/noi (A) and lia/lia (E,F) mutants at 24 hours postfertilization (hpf). Arrowheads mark the pax5 expression domain. B: Statoacoustic ganglion (SAG) projections in a noi/noi mutant labeled by injecting DiI (1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate) into the utricular macula at 72 hpf. C,G: Acridine orange (AO) staining in noi/noi (C) and lia/lia (G) mutants at 48 hpf. D,H: terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling (TUNEL) staining in noi/noi (D) and lia/lia (H) mutants at 48 hpf. White arrowheads indicate apoptotic cells in C and D, and black arrows mark otoliths. Utricular maculae are outlined in C, D, G, and H. I,J: Rhodamine–phalloidin labeling of the utricular macula (outlined) in wild-type (I) and noi/noi (J) embryos at 48 hpf. K,L: Anti-Pax2 staining of the otic vesicle in wild-type (K) and lia/lia (L) embryos at 30 hpf. I–L: White arrows indicate hair cell patches. A–L: Images show dorsolateral (A,B,E,F,I–L) and lateral (C,D,G,H) views, with anterior to the left. Scale bar = 30 μm in A,E,F,K,L, 50 μm in B, 25 μm in C,D,G,H,I,J.

noi mutants initially produce more hair cells than normal due to weakened lateral inhibition (Riley et al.,1999). Thus, noi mutants produce an average of 6.0 ± 0.8 utricular hair cells by 30 hpf, compared with 4.9 ± 0.8 in the wild-type. However, noi mutants later show a deficit of utricular hair cells similar to that seen in pax5 morphants: At 48 hpf, noi mutants have 16 ± 3.4 utricular hair cells (n = 29), a 27% decrease compared with the wild-type (22 ± 2.7 hair cells, n = 22; Fig. 7I,J). Moreover, approximately 35% of noi mutants show dying cells in the utricular macula at 48 hpf (Fig. 7C). Summing data from 30 noi mutants shows nearly a fourfold increase in the number of AO-stained cells in the utricular macula (Fig. 6G; 26 labeled cells in noi/noi vs. 7 in +/+ embryos). noi mutants also show elevated cell death in primordia of the posterior and lateral cristae (Fig. 6G). TUNEL assays give similar results (Fig. 7D). SAG projections to the utricular macula are difficult to discern and, when present, are highly disorganized (data not shown). SAG projections to the hindbrain are also disorganized (Fig. 7B). Because the morphology of noi mutants is severely altered and embryos begin to die by 3 dpf, balance and coordination cannot be tested. Thus, there are several similarities between noi mutants and pax5 morphants, but noi mutants have a wider range of defects. This finding is probably because noi mutants lack expression of pax5 as well as numerous other downstream genes.

A null mutation in fgf3, lim absent (lia) was recently identified (Herzog et al.,2004). Consistent with the results of fgf3-MO injection (Kwak et al.,2002), lia mutants display decreased pax5 expression in the otic vesicle (Fig. 7E) and in some cases, pax5 transcripts are almost ablated (Fig. 7F). lia mutants produce fewer hair cells in the utricular macula (2.6 ± 0.5, 30 hpf; n = 11) than wild-type (5 ± 0.6, 30 hpf; n = 38; Fig. 7K,L). Projections of SAG neurons to the hindbrain are similar to wild-type (Fig. 4; Table1). Despite having strongly reduced pax5 expression, lia mutants do not show increased cell death in the utricular macula (Fig. 7G,H) or misplaced hair cells (data not shown), and SAG projection patterns in the utricle are normal. The reduced number of hair cells in lia probably reflects a reduced rate of production, as Fgf3 is implicated in hair cell specification (Kwak et al.,2002). Vestibular function is more severely impaired in lia mutants than pax5 morphants (Fig. 2). However, we note that utricular and saccular otoliths fuse in lia mutants by 48 hpf. Combined with the reduced number of utricular hair cells, the late stage otolith defects are likely to contribute to the severe vestibular deficits in lia mutants. Thus, loss of fgf3 perturbs vestibular function by a mechanism distinct from that seen in noi mutants and pax5 morphants. The implications of these findings are discussed below.


In this report, we have shown that pax5 regulates the maintenance and function of the utricular macula, which is essential for vestibular function during larval development (Riley and Moorman,2000). In pax5 morphants, utricular hair cells appear to form normally but begin to die by 30 hpf, resulting in variable disorganization of the macula. There is not a wholesale loss of hair cells, possibly because of ongoing developmental expansion of the macula or regeneration of lost hair cells. In addition, it is likely that the phenotype is ameliorated by loss of pax5-MO activity during later stages of development, as suggested by a return to normal rates of cell death by 4 dpf. Globally misexpressing pax5 rescues early cell death in the utricular macula of pax5 morphants, resulting in restoration of utricular hair cell number. Together, these data suggest that pax5 regulates maintenance rather than formation of hair cells, and its function specifically affects the utricular macula. In addition, pax5 morphants show variable defects in the pattern of SAG processes in the utricle, the severity of which correlates with the degree of hair cell disorganization and death. It seems likely that SAG mispatterning is a secondary consequence of hair cell loss. Together, these changes in utricular architecture seem sufficient to explain the disruption of vestibular function seen in pax5 morphants.

Some of the above defects could also reflect changes in support cells, which are thought to be necessary for hair cell survival (Haddon et al.,1998). However, there are no substantial deficits in support cells based on the morphology of the sensory epithelia in pax5 morphants. In addition, approximately 80% of the dying cells detected in the utricle by AO staining appear to be Pax2-postive hair cells. The remainder are Pax2-negative cells in the basal layer of the utricular sensory epithelium. These cells could be either newly specified hair cells in the earliest stages of differentiation or support cells (for which there are no markers in zebrafish). For now, the involvement of support cells remains an open question.

How pax5 functions is not yet clear. The early expression of pax5 in the utricular primordium suggests that it might regulate an essential aspect of early differentiation, without which cells later die. A similar cell death phenotype is seen in noi (pax2a) mutants (Fig. 6). Because pax2a is required for expression of pax5 in the ear (Pfeffer et al.,1998; Fig. 7A), the effect of noi on cell survival in the utricle could be mediated specifically by loss of pax5 expression. Alternatively, because pax2a and pax5 are likely to be partially redundant (Bouchard et al.,2000), pax5 might serve to supplement pax2a in promoting hair cell survival in the utricle. In either case, these findings raise the question of why utricular hair cells alone require pax5 for survival. Presumably the basic ground plan of the sensory epithelium can be modified by specific combinations of multiple factors (combinatorial codes), possibly including Pax5, that confer unique properties and requirement to the utricular macula.

The importance of a combinatorial code is also suggested by the phenotype of lia (fgf3) mutants. lia mutants show strong reduction in pax5 expression yet do not show cell death or SAG mispatterning in the utricle. This finding shows that cell death is not an inescapable consequence of reducing pax5 function in the utricle. Fgf3 presumably regulates other genes that could influence utricular development in conjunction with pax5. Indeed, Fgf3 is required to block expression of the posterior marker zp23 in the utricular macula, consistent with a general role for Fgf3 in specifying anterior identity (Kwak et al.,2002). Based on the general morphology of the utricular macula and its SAG projections, lia mutants do not show a wholesale conversion of the utricle into a saccule. However, partial readjustment of the combinatorial code in lia (e.g., loss of pax5, ectopic expression of zp23) could subtly alter regional identity, thereby making hair cell survival independent of pax5.

In mouse, several other genes have been shown to differentially regulate survival of hair cells in different regions. Brn-3c is required for the survival of hair cells in all epithelia, although auditory hair cells are affected more severely (Xiang et al.,1998). In a gene expression profiling experiment, Gfi1 was identified as a downstream target of Brn-3c (Hertzano et al.,2004). Although Gfi1 is expressed in all hair cells, ablation of this gene causes cell death only in cochlear hair cells (Wallis et al.,2003). The cochlear phenotype of Gfi1−/− mutants is very similar to that of Brn-3c−/− mutants, suggesting that Brn-3c regulates maintenance of cochlear hair cells through Gfi1 function. That Gfi1 is dispensable for hair cell survival in other sensory patches again suggests that each region is regulated by a specific combination of differentiation and maintenance factors.

Pax2 and Pax5 Functions in Other Vertebrates

In mouse and chick, the expression and function of Pax2 have been studied most extensively in the cochlea, although it is also expressed in hair cells in other sensory epithelia (Lawoko-Kerali et al.,2002; Burton et al.,2004; Li et al.,2004; Sanchez-Calderon et al.,2005). However, its role in hair cell maintenance per se cannot be addressed in the cochlea of Pax2 null mice because of severe agenesis of this region. The sensory epithelia that do form (utricular macula and cristae) have not been examined in sufficient detail to determine whether there are defects in hair cell patterning or survival.

Pax5 expression in the otic vesicle has been reported in Xenopus (Heller and Brandli,1999) and recent gene expression profiling data for the chick ear indicate Pax5 expression in the adult utricle and cochlea (http://hg.wustl.edu/lovett/projects/nohr/inner_ear_ratio.html). In contrast, Pax5 is not detected in the mouse ear during embryonic development, and Pax5 null mice have no obvious defects in hearing or balance (Urbanek et al.,1994). It is possible that mouse represents a derived state wherein Pax5 is no longer used in otic development.

The Pattern of Utricular SAG Projections

DiI injections into the utricle revealed two patterns of central projections of the SAG, type-1 with discretely organized ascending and descending branches in the hindbrain, and type-2 with diffuse primary branches and several smaller secondary branches. These patterns are independent of age through 7 dpf and do not require vestibular activity, as shown by analysis of pax5 morphants and mnl and lia mutants. Variation in neural patterning is often seen during early development and is later corrected by pruning (reviewed by Maklad and Fritzsch,2003). It therefore seems likely that the type-2 pattern eventually resolves into a more cohesive pattern. Nevertheless, it is remarkable that vestibular activity and motor coordination are so effectively integrated in young larvae despite variation in SAG projection patterns.

A fundamental question remaining is how SAG neurons make connections between a given sensory epithelium to the appropriate processing center in the brain. A recent study by Satoh and Fekete (2005) showed that neuroblasts from one region of the ear often innervate sensory patches in another. This finding suggests that SAG targeting can be regulated after neuroblasts delaminate from the vesicle. For example, SAG neurons might project randomly to different sensory patches, after which regional signals from the sensory patch program the neuron to make appropriate central projections. Changes in the utricle caused by disruption of pax5 or fgf3 did not affect central projections, but noi mutants showed severely distorted central projections. At present, it is not clear whether this finding reflects changes in otic vesicle or hindbrain. Further analysis of noi and other zebrafish mutants could help resolve this issue.


Fish Strains and Staging

Wild-type zebrafish strains were derived from the AB line (Eugene, OR). Mutant alleles used in this study include noitu29a (Lun and Brand,1998), liat24152 (Herzog et al.,2004), and mnlz2 (Riley and Grunwald,1996). Embryos and larvae were maintained in an incubator at 28.5° C and staged as described by Kimmel et al. (1995). Ages are denoted as hours postfertilization or days postfertilization. Embryos normally hatch by 3 dpf, after which they are referred to as larvae. In some case, 0.2 mM phenylthiourea (PTU) was added to prevent melanin formation.

Cloning of pax5

Comparison of zebrafish (http://www.ensembl.org) and Fugu genomic sequences identified the putative missing 5′ and 3′ ends of zebrafish pax5. The identified sequence was confirmed by reverse transcriptase-polymerase chain reaction followed by cloning and sequencing using primers for the putative full-length open reading frame: pax5(-6), 5′GGGAATTCAACACGATGGAAATCCACTG3′; pax5(1128), 5′GGTCTAGATTATTTCGTGCCTCCCACTC3′.

Behavioral Analysis

Vestibular function was assayed by three tests between 3 and 7 dpf as previously described (Riley and Moorman,2000). Balance was assessed by the ability of larvae to rest with their dorsal sides up 1 min after initiating a startle response by tapping Petri dishes containing larvae. Each specimen was tested three times and was scored as negative if it failed all three trials. To test motor coordination, individual larvae were observed after a startle response, induced by tapping the plate or gentle physical stimulation of the tail. Normally, larvae rapidly traverse a 6-cm Petri dish in a straight line. Larvae with vestibular dysfunction swim in circles, vertical loops, spirals, or in erratic zigzags. Specimens failing three consecutive trials were scored as negative. Swim bladder inflation was observed under a dissecting microscope (Riley and Moorman,2000). Normally, larvae must swim to the surface to obtain air for swim bladder inflation. Vestibular deficits impede this motion and thereby prevent swim bladder inflation. AEBRs were tested in larval fish 7–12 dpf as described by Zeddies and Fay (2005). Briefly, larval fish were placed in the 8 central wells (one fish per well) of a 24-well plate affixed to a plastic platform. A TDT System 3 (TDT, Inc., Gainesville, FL) was used to deliver tonal pulses to the platform by means of a vertically oriented Bruel & Kjaer type 4810 shaker. Eight frequencies from 100 to 1,200 Hz were tested at seven different levels ranging from 0 to −42 dB re 1g (decibels relative to the force from acceleration at 1g). Five-second-long video sequences of the larvae were digitally recorded such that the tone was presented 2.5 sec after the start of the video recording. A frame-by-frame subtraction method was then used to analyze the video sequences and compare the movement of the larval fish in quiet with their movement when the stimulus is present. A positive response was registered when the movement during the tone was significantly different (P < 0.0001) than in quiet. A complete experiment consisted of the presentation of two randomized presentations of the 56 trials. The lowest level that resulted in a positive response at each frequency was then considered the threshold level for that frequency.

Morpholino Injection

Splice- and translation-blocking morpholino oligomers (Nasevicius and Ekker,2000; Draper et al.,2001) were generated to knock down pax5. The translation blocker for splice variant 1 (TB1)was 5′CAGTGGATTTCCATCTGTTTTAAA3′; the translation blocker for splice variant 2 (TB2) was 5′CTCGGATCTCCCAGGCAACATGGT3′; the splice blocker for the exon–intron boundary of exon 2 (SB1) was 5′TACTCATAACTTACCTGCCCAGTA3′; the splice blocker for the exon–intron boundary of exon 3 (SB2) was 5′ATGTGTTTTACACACCTGTTGATTG3′; and the splice blocker for the exon–intron boundary of exon 5 was 5′TTGACCCTTACCTAAATTATGCGCA3′. A cocktail of all five morpholinos was prepared in Danieaux solution (Nasevicius and Ekker,2000) to a concentration of 12 μg/μl (3 μg/μl each TB1 and TB2; and 2 μg/μl each SB1, 2, and 3). Approximately 1 nl was injected into wild-type zebrafish embryos at one-cell stage to generate pax5 morphants.

Injection of pax5 RNAs

Two splice variants, pax5-v1 and pax5-v2 were cloned in pCS2p+. RNAs for both variants were synthesized in vitro and ∼200 pg of pax5-v1 and pax5-v2 RNA mixture (100 pg each) was injected into embryos at the one- to two-cell stage.


Embryos raised in PTU were fixed and processed as previously described (Riley et al.,1999). Primary antibodies were as follows: mouse anti-Pax2 (Berkeley Antibody Company, 1:100 dilution), anti-acetylated tubulin (Sigma T-6793, 1:100), and anti–Islet-1/2 (Developmental Studies Hybridoma Bank 39.4D5, 1:100). Secondary antibodies were as follows: Alexa 546 goat anti-rabbit IgG (Molecular Probes A-11010, 1:50) and Alexa 488 goat anti-mouse IgG (Molecular Probes A-11001, 1:50).

Rhodamine–Phalloidin Staining

Larvae raised in PTU were fixed between 3 and 7 dpf. Fixed larvae were rinsed in phosphate buffered saline (PBS) containing 0.1% Triton X-100 for 15 min and then permeabilized by incubation in PBS containing 2–3% Triton X-100 for 4 hr at room temperature and then overnight at 4°C. Permeabilized embryos were incubated in Rhodamine–Phalloidin (Molecular Probes R415, 1:30 dilution in 1% bovine serum albumin in PBS) for 2 hr at room temperature, washed four times in PBS with 0.5% Triton X-100 for 30 min each.

DiI Labeling

Larvae were fixed between 3 and 7 dpf and then washed in PBS. Fixed larvae were mounted in 0.6% low-melting-temperature agarose made in PBS. To examine the neuronal projections from the statoacoustic ganglion (SAG), DiI (Molecular probes D-282, 4 mg/ml in 100% ethanol) was injected into the utricular macula. Glass micropipettes were backfilled with the DiI solution and directed to the utricular macula using a micromanipulator. Injected larvae were incubated at 33°C overnight and observed.

Whole-Mount In Situ Hybridization

Whole-mount in situ hybridizations were carried out as described previously (Phillips et al.2001) using the following riboprobes: nkx5.1 (Adamska et al.,2000), otx1 (Li et al.,1994), zp23 (Hauptmann and Gerster,2000), dlx3b (Ekker et al.,1992a), krox20 (Oxtoby and Jowett,1993), msxC (Ekker et al.,1992b), pax5 (Pfeffer et al.,1998), fgf8 (Reifers et al.,1998), and fgf3 (Kiefer et al.,1996).

Cell Death Analysis

Embryos were dechorionated and incubated in AO (1 μg/ml) in PBS for 1 hr at room temperature and washed twice (10 min each) in PBS before observation. In situ TUNEL assay was performed as suggested by the manufacturer (Promega TUNEL assay kit).


We thank Mark Zoran for thoughtful discussion of the data. S.J.K, S.V., B.B.R., D.Z., and A.P. were supported by the National Institutes of Health, NIDCD; S.J.M. was supported by a grant from NASA.