Main

Stereocilia are similar to other actin-based cell-membrane protrusions, such as microvilli and filopodia. These cellular projections have a core of unidirectionally-oriented actin filaments, with the barbed (plus) end oriented towards the tip of the protrusion1, which is the site of high-affinity actin polymerization and elongation7,8,9,10. Actin filaments within these protrusions vary in number and packing density, and have distinct sets of actin-bundling proteins11,12 and myosins12. In contrast to microvilli and filopodia, which continuously form and disassemble7,13, stereocilia bundles retain their mature staircase shape despite internal remodelling processes of the paracrystalline actin core10. Nevertheless, the general mechanism of cytoskeletal remodelling in stereocilia, microvilli and filopodia are assumed to be similar12.

During development of a mammalian hair cell, nascent stereocilia are indistinguishable from surrounding microvilli in length and width14,15,16. Subsequently, stereocilia undergo differential elongation with a simultaneous increase in thickness, culminating in the formation of the staircase architecture of the hair bundle3,15,16. The length of each stereocilium is predetermined and depends on the position within the bundle and on the location of the hair cell within the cochlear spiral or vestibular sensory epithelium15,16,17. The molecules that initiate the transformation of microvilli-like structures into stereocilia are largely unknown.

Alterations in the differential elongation of stereocilia cause hearing loss and balance disorders in humans and mice12. Abnormally short stereocilia have been reported in the shaker 2 and whirler strains of deaf mice and mutations of myosin-XVa also cause deafness in humans4,18. We have previously shown that myosin-XVa localizes to the tips of wild-type inner-ear hair-cell stereocilia19, whereas Myo15ash2 mice lack myosin-XVa immunoreactivity at stereocilia tips19,20, indicating that myosin-XVa may be involved in differential stereocilia elongation. The short stereocilia phenotype of whirler mice is caused by a deletion, which introduces a frameshift in Whrn and results in a premature translation termination codon upstream of the region encoding the third PDZ domain (PDZ3) of whirlin5,6. In humans, a nonsense mutation of Whrn that truncates whirlin, thereby eliminating the third PDZ domain, is associated with deafness6.

Considering the similar short stereocilia phenotype of shaker 2 and whirler mice and the presence of a predicted PDZ-ligand on the C terminus of myosin-XVa19, we investigated two questions. Does myosin-XVa interact with whirlin and transport it to the stereocilia tips? Is this interaction a critical step in differential elongation and in the formation of the staircase architecture of hair-cell stereocilia bundles?

We initiated this study by generating antisera to whirlin, as whirlin was previously localized to the hair-cell bundle but its precise position within a stereocilium was not reported6. We found that whirlin, similar to myosin-XVa19, is concentrated at the tips of cochlear and vestibular hair-cell stereocilia in adult and newborn wild-type mice (Fig. 1a). The abnormally short hair-cell stereocilia of homozygous Whrnwi and Myo15ash2 mice do not show specific immunoreactivity to whirlin (Fig. 1b), whereas myosin-XVa was localized to the apices of stereocilia of homozygous Whrnwi mice (Fig. 1b). Therefore, whirlin is not required for the targeting and tethering of myosin-XVa to the tips of stereocilia, although myosin-XVa is necessary for whirlin localization at the stereocilia apices. Targeting of whirlin and myosin-XVa to the tips of wild-type stereocilia was demonstrated by transfection of mouse and rat inner-ear sensory epithelial explants, mediated by a gene gun. GFP–whirlin accumulated at the tips of stereocilia of wild-type hair cells that were transfected with a GFP-Whrn cDNA construct (Fig. 1c). When gold particles coated with a mixture of GFP-Myo15a and DsRed-Whrn cDNA constructs (1:1 ratio) were used, both epitope-tagged proteins were co-localized at the tips of hair-cell stereocilia (Fig. 1d). The presence of both GFP–myosin-XVa and DsRed–whirlin at stereocilia tips indicate that epitope tags at the amino termini do not interfere with proper interaction or localization of these proteins in hair cells. Overexpression and accumulation of these proteins caused a bulging of the tips of stereocilia similar to that observed for myosin-XVa alone19, but did not induce over-elongation of wild-type stereocilia (Fig. 1c, d).

Figure 1: Localization of whirlin and myosin-XVa in hair-cell stereocilia of wild-type (wt), Myo15ash2 and Whrnwi mutant mice.
figure 1

(a) Localization of whirlin (green) at the tips of hair-cell stereocilia of the adult (left) and postnatal day 7 (P7, right) wild-type mice using anti-whirlin antisera HL5141 and HL5136, respectively. (b) Whirlin (green) was not detected at the tips of hair-cell stereocilia of homozygous Whrnwi mice (left panel) and homozygous Myo15ash2 mice (middle panel), but myosin-XVa was detected at the tips of hair-cell stereocilia of homozygous Whrnwi mice (right panel). Left panel: P7 Whrnwi mouse organ of Corti stained with anti-whirlin HL5141 antibody. Middle panel: P12 Myo15ash2 utricle stained with anti-whirlin HL5140 antibody. Right panel: P7 Whrnwi utricle stained with anti-myosin-XVa TF1 antibody. (c) Accumulation of green fluorescent protein (GFP)–whirlin at the tips of hair-cell stereocilia after transfection of inner-ear sensory epithelia explants, using a Helios gene gun. Left panel: transfected and neighbouring non-transfected P5 rat auditory hair cells after 3 d in culture and 21 h post-transfection. Right panel: a P3 wild-type mouse vestibular hair cell after 3 d in culture and 65 h post-transfection. (d) Simultaneous transfection of GFP-Myo15a and DsRed-Whrn into the vestibular hair cell of a wild-type mouse. GFP-Myo15a (left), DsRed-Whrn (middle) and merged image (right). (e) Direct proportionality between the fluorescence of GFP–myosin-XVa and DsRed–whirlin in hair-cell stereocilia after simultaneous transfections, as shown in d. Each point represents relative fluorescence intensities measured as an average value over the circular area of 0.7 μm in diameter, with a centre at the tip of each individual stereocilium (n = 37). Dashed line shows a linear fit through the data. Correlation coefficient and statistical significance are indicated at the top of the graph. (f) Control transfections of wild-type hair-cell sensory epithelia explants with GFP-Myo1c (left, utricle), GFP-Myo7a (middle, organ of Corti), GFP-MYO6 (right, ampulla). Insets show green channel representing GFP fluorescence of each transfected cell. Cytoskeletal actin was visualized by rhodamine-phalloidin staining (red) in all panels except panel d, in which it was stained with Alexa Fluor 633 phalloidin (blue). Scale bars, 5 μm.

Full size image

Endogenous myosin-X accumulates at the tips of filopodia of HeLa cells in direct proportion to the length of filopodia21. Similarly, quantitative analysis of the fluorescence at the tips of stereocilia of different lengths indicates that the amount of DsRed–whirlin was directly proportional (r = 0.97, p <0.0001) to that of GFP–myosin-XVa (Fig. 1e), which is directly proportional to the length of a stereocilium19,22. The molecular mechanism that establishes this direct relationship is unknown.

Myosins Ic, VI and VIIa are also implicated in normal hearing and are essential for stereocilia structure and function23. Transfection of cDNA constructs of these myosins into wild-type hair cells showed that GFP–myosin-Ic and GFP–myosin-VIIa are localized along the length of stereocilia and are slightly concentrated towards the tips (Fig. 1f), whereas GFP–myosin-VI was localized in the pericuticular necklace and hair-cell body and was not detected within stereocilia (Fig. 1f). Therefore, myosin-XVa may be the only motor in hair cells that selectively delivers cargo to the tips of stereocilia.

To explore whether myosin-XVa is a plus-end directed motor, which actively moves towards the barbed ends of actin filaments, we performed time-lapse recordings of COS7 cells that were transfected with either wild-type GFP-Myo15a or mutant GFP-Myo15a[C592Y]19 with a Myo15ash2 motor-domain mutation4. During live-cell recordings, GFP–myosin-XVa fluorescence progressively increased at filopodia tips until some fluorescent patches detached and moved towards the cell body, initiating a new cycle of GFP–myosin-XVa accumulation at the tip. We did not observe patches of GFP–myosin-XVa moving from the cell body to the filopodia tips, probably because GFP–myosin-XVa molecules move towards the tips as singlets or in small groups. When the entire length of a filopodium was photobleached, we observed the recovery of myosin-XVa fluorescence at the filopodium tip (Fig. 2a, b, c, and see Supplementary Information, Movie 1). These data indicate that myosin-XVa molecules move continuously towards the barbed ends of actin filaments at filopodia tips. Myosin-XVa with the C592Y mutation was unable to translocate towards the tips of filopodia (data not shown). Also, the velocity of retrograde movement of GFP–myosin-XVa patches (Fig. 2c) was about 20 nm s−1 (Fig. 2d), which is similar to the speed of the actin treadmill in vitro24 and in filopodia at 37 °C (ref. 9), indicating passive removal of myosin-XVa from filopodia tips. Together, these data indicate that myosin-XVa is a plus-end directed motor that moves along actin filaments and accumulates at filopodia tips.

Figure 2: Green fluorescent protein (GFP)-tagged wild-type myosin-XVa accumulates continuously at the tips of filopodia in COS7 cells.
figure 2

(a) Time-lapse images of filopodia before (top-left image), during (top-middle image), immediately after (top-right image) and following recovery from photobleaching (bottom images). Photobleaching area is outlined by a dashed circle (top-middle image). (b) % change of fluorescence intensity in the tip regions, outlined by coloured circles in a for filopodia within (red, cyan) and outside (green) photobleaching region. Time scale indicates when the images shown in a were captured. Photobleaching interval of 40 s is indicated by arrows. Note that the fluorescence signal at the tips recovered in both cases: after photobleaching (0 min) and after spontaneous removal of a large myosin patch from the tip of the filopodium (8.2 min, 'green' filopodium). (c) Intensity of GFP–myosin-XVa fluorescence along a representative filopodium of another cell before (left), immediately after (middle) and 6 min after photobleaching (right). The intensity of fluorescence was represented relative to the fluorescence of the filopodium tip before photobleaching. (d) Overlay of bright-field and fluorescent images shows a patch of myosin-XVa at the filopodium tip (left) and during its backwards movement along the filopodium core (right). (e) Velocity of individual myosin-XVa patches before (negative times) and after (positive times) dislocating from filopodia tips. The zero time point corresponds to the moment when a patch is dislocated from the tip. All analysed individual filopodia were tethered to the glass. After dislocating from the tip, myosin-XVa patches move towards the cell body, with a constant velocity of ∼20 nm s−1, which is typical for actin treadmill-driven retrograde flow9,24. The mean velocity for five individual filopodia is shown. Scale bars, 5 μm.

Full size image

To exclude the possibility that the C592Y mutation 'locks' mutant myosin-XVa tightly to actin, blocking its active or passive movement towards the tips of filopodia or stereocilia, we used a myosin-XVa construct with two missense mutations (R167A; G388A) of conserved amino acids of the motor domain. These mutations have been shown to abolish motility and to 'lock' myosin in a weak actin-binding state25,26. When homogenates of COS7 cells that were transfected with GFP-Myo15a, GFP-Myo15a[C592Y] or GFP-Myo15a[R167A;G388A] were sedimented with actin in the presence of ATP, the predominant fraction of mutated myosins was observed in the supernatant in approximately the same amounts as that of wild-type myosin-XVa. This strongly indicates that these two mutant myosins, which fail to localize to filopodial tips, are not trapped by cytoskeletal actin (data not shown).

To evaluate whether GFP–myosin-XVa could interact with DsRed–whirlin and deliver it to the tips of filopodia, full-length cDNA expression constructs of GFP-Myo15a and DsRed-Whrn were co-transfected into COS7 cells (Fig. 3a). Both of the corresponding proteins were colocalized at the tips of filopodia (Fig. 3b). However, when GFP-Myo15a[C592Y] or GFP-Myo15a[R167A;G388A] were co-transfected with DsRed-Whrn, neither of the corresponding proteins were localized at the tips of filopodia (Fig. 3c and data not shown). Similarly, DsRed- or GFP-Whrn expression constructs transfected alone did not accumulate at the tips of filopodia (data not shown). These results indicate that GFP–myosin-XVa and DsRed–whirlin can interact in heterologous cells and that a functional motor domain of myosin-XVa is required to deliver whirlin to filopodia tips.

Figure 3: Localization of recombinant wild-type and mutant myosin-XVa and whirlin in COS7 cells.
figure 3

(a) Schematic representation of full-length DsRed-tagged whirlin (top) and green fluorescent protein (GFP)-tagged myosin-XVa (bottom) and their domains. Horizontal red lines indicate the location of the amino-acid residues of antigens, which were used to immunize rabbits. Pairs of Myo15a and Whrn expression constructs used in co-transfection of COS7 cells are shown on the left side of the panels b–e. (b) Co-transfection of GFP-Myo15a (left panel) with DsRed-Whrn (middle panel) reveals colocalization of these proteins at the tips of filopodia (right panel). (c) Co-transfection of GFP-Myo15a[C592Y] (left panel) with DsRed-Whrn (middle panel) shows no fluorescence signal at the tips of filopodia (right panel). (d) Co-transfection of wild-type GFP-Myo15a (left panel) with DsRed-Whrn[−PDZ3] (middle panel) reveals accumulation of GFP–myosin-XVa at the tips of filopodia, which is not accompanied by DsRed–whirlin[−PDZ3] (right panel). (e) Co-transfection of GFP-Myo15a[−PDZL] (left panel) with DsRed-Whrn (middle panel) shows concentration only of GFP–myosin-XVa[−PDZL] at the tips of filopodia without colocalization with DsRed–whirlin (right panel). Scale bar. 10 μm.

Full size image

The tail of myosin-XVa has two pairs of myosin tail homology 4 (MyTh4) and band 4.1/ezrin/radixin/moesin (FERM) domains, a Src-homology-3 domain (SH3)27 and a predicted class I PDZ-ligand28 (ITLL) at the C terminus19 (Fig. 3a). None of the other mammalian myosins that are known to be localized to hair-cell stereocilia are predicted to have a PDZ-ligand at the C terminus. Whirlin has three PDZ domains and a proline-rich domain6 (Fig. 3a). PDZ domains are protein–protein recognition modules that are important in the assembly of submembrane supramolecular signalling complexes28. As the phenotype of whirler mice can be largely rescued by a BAC transgene encoding a whirlin isoform that includes the proline-rich and PDZ3 domains6, we postulated that PDZ3 of whirlin might interact with the PDZ-ligand of myosin-XVa. To test this hypothesis, epitope-tagged truncation constructs of whirlin and myosin-XVa were first evaluated in COS7 cells. DsRed–whirlin lacking PDZ3 (DsRed-Whrn[−PDZ3]) failed to colocalize with wild-type GFP–myosin-XVa at the tips of filopodia (Fig. 3d), indicating that the PDZ3 domain of whirlin is necessary for the interaction with myosin-XVa. In the reciprocal experiment, GFP-Myo15a lacking the PDZ-ligand (GFP-Myo15a[−PDZL]) was co-transfected with wild-type DsRed-Whrn into COS7 cells. GFP–myosin-XVa[−PDZL] accumulated at the tips of filopodia in a manner identical to the wild type, but was not accompanied by DsRed–whirlin, which remained in the cell body (Fig. 3e). These results indicate that both the PDZ-ligand of myosin-XVa and PDZ3 of whirlin are required for localization of whirlin at filopodia tips.

We also used live-cell imaging to corroborate that whirlin is a cargo of myosin-XVa. When wild-type GFP-Myo15a and DsRed-Whrn were co-transfected into COS7 cells, we observed simultaneous accumulation of green and red fluorescence at the filopodia tips, as well as patches of green and red fluorescence moving together towards the cell body (Fig. 4a and Supplementary Information, Movie 2). In control co-transfection experiments, GFP–myosin-XVa[−PDZL] accumulated alone at the tips of filopodia and relocated in patches towards the cell body unaccompanied by DsRed–whirlin, which remained in the cell body (data not shown). The co-transfection of mutant DsRed-Whrn[−PDZ3] with wild-type GFP-Myo15a confirmed that not only the PDZ-ligand of myosin-XVa, but also the PDZ3 of whirlin, are required for this interaction (Fig. 4b and Supplementary Information, Movie 3). However, we cannot rule out the possibility of an intermediary adaptor protein being involved, linking the PDZ-ligand of myosin-XVa with the PDZ3 of whirlin.

Figure 4: Concurrent movement of myosin-XVa and whirlin along the filopodia of live COS7 cells.
figure 4

(a) Simultaneous retrograde movement of green fluorescent protein (GFP)–myosin-XVa (left) and DsRed–whirlin (right) along the filopodium. (b) The absence of concurrent movement of GFP–myosin-XVa (left) and DsRed–whirlin[−PDZ3] (right). Both a and b panels show time-lapse sequence of image pairs. Images of each pair (green/red) were recorded almost simultaneously (within 1–1.5 s, see Methods). Time in minutes from the beginning of observation is indicated on the right side of each image pair. COS7 cells were co-transfected with the corresponding constructs and observed at 37 °C 24 h after transfection. Note the progressive translocation of big patches that is indicated by arrows. Scale bars, 5 μm. (c) Co-transfection of COS7 cells with DsRed-Whrn and GFP-Myo1c (first panel), GFP-Myo7a (second panel), GFP-MYO6 (third panel) and GFP-Myo10 (fourth panel) reveals no colocalization of DsRed–whirlin with these myosins in filopodia. Scale bars, 10 μm.

Full size image

In control experiments, wild-type DsRed-Whrn was co-transfected with either mouse GFP-Myo1c, mouse GFP-Myo7a, human GFP-MYO6 or bovine GFP-Myo10. In COS7 cells, GFP–myosin-Ic was distributed along the length of the filopodia and concentrated at their tips, but was not accompanied by DsRed–whirlin (Fig. 4c). GFP–myosin-VIIa was observed along the length of filopodia but was not concentrated or colocalized with whirlin at filopodia tips (Fig. 4c). GFP–myosin-VI was not targeted to the tips of filopodia and we observed no colocalization with DsRed–whirlin (Fig. 4c). The co-expression of GFP-Myo10, together with DsRed-Whrn, showed that myosin-X, which is localized to the tips of filopodia29, also did not transport DsRed–whirlin to filopodia tips (Fig. 4c). These data indicate that whirlin is not a promiscuous cargo of unconventional myosins, but binds selectively to the PDZ-ligand of myosin-XVa for transport to filopodia tips.

The stereocilia phenotypes of shaker 2 and whirler mice indicate that myosin-XVa and whirlin are not required for the emergence, initial elongation or increase in thickness of stereocilia within the hair bundle. Our data indicate that myosin-XVa contributes specifically to the differential elongation of stereocilia by delivering whirlin, and possibly other cargo, to the tips of stereocilia. Indeed, after transfection of GFP-Myo15a into the hair cells of Myo15ash2 sensory epithelial explants, we observed rescue of the normal length and shape of the abnormally short hair bundles (Fig. 5a). After 48 h, stereocilia of a representative transfected hair cell were approximately twofold longer than their neighbouring, control non-transfected hair cells, and developed a staircase architecture similar to the bundles of wild-type hair cells (Fig. 5a). After 67 h, we observed apparently complete rescue of the hair-bundle staircase architecture in transfected vestibular hair cells from Myo15ash2 mice (Fig. 5b). Over-elongation of the restored bundles was not observed, indicating that an excess amount of myosin-XVa does not perturb the tightly regulated programme of stereocilia elongation. Moreover, supernumerary rows of stereocilia were also resorbed so that the bundle acquires a mature wild-type appearance (Fig. 5b). In agreement with these data, GFP-Myo15a[−PDZL] that was transfected into Myo15ash2 hair cells did not stimulate the elongation of stereocilia but did concentrate at the apices of the abnormally short stereocilia (Fig. 5c). In Myo15ash2 hair cells transfected with GFP-Myo15a[−PDZL] and stained with anti-whirlin antibody, endogenous whirlin remained in the cell body and did not colocalize with GFP–myosin-XVa[−PDZL] at the tips of stereocilia (Fig. 5c, right).

Figure 5: Re-initiation of programmed stereocilia elongation and restoration of hair-bundle staircase architecture after gene-gun-mediated transfection of epitope-tagged Myo15a or Whrn into hair cells from homozygous Myo15ash2 or Whrnwi mice, respectively.
figure 5

(a) An auditory hair cell of a Myo15ash2 mouse transfected with GFP-Myo15a and neighbouring non-transfected control cells (48 h post-transfection). Left panel is a merge of rhodamine-phalloidin staining (middle) with the GFP–myosin-XVa image (right). (b) Restoration of the staircase shape of a stereocilia bundle of a Myo15ash2 vestibular hair cell 67 h after GFP-Myo15a transfection. The average height of stereocilia bundles: control homozygous Myo15ash2 cells, 2.9 ±0.6 μm, n = 24; transfected cells, 7.3 ±2.2 μm, n = 12; normal bundles of heterozygous Myo15ash2 cells: 9.6 ±1.5 μm, n = 22; t-test p<0.00001. (c) Stereocilia of Myo15ash2 auditory hair cells (first three panels) and vestibular hair cells (fourth panel) do not elongate after transfection with GFP-Myo15a[-PDZL] (control homozygous Myo15ash2 cells: 2.9 ±0.6 μm, n = 24; transfected cells: 2.8 ±0.7 μm, n = 6; p>0.5). In the fourth panel, staining with anti-whirlin antibody (ab) is shown in blue. (d) GFP–whirlin (left), GFP–myosin-XVa[C592Y] abbreviated 'sh2' (middle), and GFP–myosin-XVa[R167A, G388A] abbreviated 'dm' (right) are concentrated in the cell bodies and are not targeted to stereocilia tips in vestibular hair cells of Myo15ash2 mouse (left) and wild-type mouse (middle, right). (e) Restoration of the staircase shape of a stereocilia bundle in a Whrnwi vestibular hair cell 48 h after transfection with GFP-Whrn (control homozygous Whrnwi cells: 3.1 ±0.5 μm, n = 37; transfected cells: 7.0 ±1.9 μm, n = 18; normal bundles of heterozygous Whrnwi cells: 9.5 ±1.4 μm, n = 21; p<0.00001). (f) Colocalization of GFP–whirlin (green, left panel) and HL5137 anti-whirlin antibody staining (blue, middle panel) at stereocilia tips of a Whrnwi vestibular hair cell transfected with GFP-Whrn, but not in neighbouring non-transfected cells (right panel, merged image). (g) Exogenous GFP–myosin-XVa (green, left panel) recruits endogenous whirlin stained with HL5136 antibody (blue, middle panel) to stereocilia tips of a Myo15ash2 vestibular hair cell transfected with GFP-Myo15a (right panel, merged image). There is no anti-whirlin immunoreactivity in the stereocilia of neighbouring non-transfected hair cells. Cytoskeletal actin is visualized by rhodamine-phalloidin (red) in all panels. Sensory explants were harvested at P2–P5 and transfected the next day. Scale bars, 5 μm.

Full size image

In Myo15ash2 mice, GFP–whirlin remained in the hair-cell body and was not detected at the tips of abnormally short stereocilia (Fig. 5d, left), confirming the requirement of a functional myosin-XVa motor to translocate whirlin to the stereocilia tips. Consistent with this observation, GFP–myosin-XVa mutants with a disabled motor domain — either due to C592Y (Fig. 5d, middle) or R167A;G388A (Fig. 5d, right) — remained in the bodies of transfected wild-type hair cells, and did not target the tips of stereocilia. GFP–myosin-Ic and GFP–myosin-VIIa were distributed along the length of abnormally short stereocilia of Myo15ash2 hair cells (data not shown). None of these control myosins stimulated stereocilia elongation (data not shown).

In inner-ear sensory epithelial explants from homozygous Whrnwi mice, GFP-Whrn also restored stereocilia bundle shape and length to the wild-type pattern and did not cause overelongation, indicating that whirlin is also a component of programmed stereocilia elongation (Fig. 5e). When Whrnwi inner-ear sensory epithelial explants, transfected with GFP-Whrn, were stained with anti-whirlin antibody, whirlin immunoreactivity was colocalized with GFP–whirlin at the tips of transfected hair-cell stereocilia and no immunoreactivity was observed at the tips of stereocilia of non-transfected hair cells. This confirmed the specificity of the antibody (Fig. 5f). Myo15ash2 hair cells transfected with GFP-Myo15a and stained with anti-whirlin antibody show that exogenous myosin-XVa recruits endogenous whirlin to the stereocilia tips, enabling differential elongation of stereocilia (Fig. 5g).

In addition to its motor and carrier function, myosin-XVa could have a structural role in the assembly and/or maintenance of the hair-cell mechanotransduction apparatus, which is also located at or near the tips of stereocilia. The restoration of stereocilia bundles, and perhaps their mechanosensitivity in postnatal mice — through the introduction of exogenous myosin-XVa and whirlin into hair cells with abnormally short stereocilia — indicates that various gene-based therapies for some forms of hereditary deafness in humans may be feasible.

Our results demonstrate that programmed stereocilia elongation2 occurs only when myosin-XVa interacts with whirlin and recruits it to stereocilia tips. Perhaps myosin-XVa and whirlin are part of a macromolecular complex that is involved in modulating the growth of stereocilia actin bundles by regulating capping and/or capping-antagonist activity or other molecules that are required for actin polymerization and bundling30 at the stereocilia tips. Although whirlin is the only cargo of myosin XVa that has been identified, the other myosin-XVa tail domains27 and PDZ domains of whirlin6 may provide surfaces for additional protein–protein interactions. Identification of all of the protein partners of whirlin and myosin-XVa should provide a more complete molecular model of programmed stereocilia elongation.

Methods

Genotyping.

To genotype homozygous and heterozygous Whrnwi mice, we took advantage of the Whrnwi deletion6. PCR reactions were performed with primers (Supplementary Information, Table 1 online) within exons 6 and 9 of Whrn. These primers amplified a 171 base-pair product from genomic DNA of homozygous or heterozygous Whrnwi mice but not from wild-type animals, as the amplimer is too long (14,523 bp) for conventional PCR conditions. In the same PCR reaction, we detected the wild-type allele (544 bp) using primers within introns 6 and 7 of Whrn. Myo15ash2 mice were genotyped, as previously described20.

Antibodies.

In this study, we used rabbit polyclonal anti-myosin-XVa antibody (TF1) that we characterized previously19,20,27. To raise antibodies to whirlin (HL5136, HL5137, HL5140, HL5141; see also Fig. 1 legend), cDNA constructs encoding amino-acid residues 362–509 and 711–815 of mouse whirlin (NCBI accession number AY739114) were introduced individually into pGEX5.1 (Amersham Biosciences, Piscataway, NJ) and expressed in Escherichia coli (BL21 Gold DE3 pLysS; Stratagene, La Jolla, CA). Fusion proteins were isolated by incubating with glutathione sepharose 4B (Amersham Biosciences), and the purity confirmed by denaturing PAGE. Each fusion protein was used to immunize three rabbits (Covance, Denver, PA). cDNAs encoding amino acids 362–509 or 711–815 of mouse whirlin were also introduced into pMAL-c2x (New England Biolabs, Beverely, MA), transformed into E. coli Rosetta (DE3; Novagen, Madison, WI) and induced to express the corresponding maltose binding (MBP) fusion protein. The expressed fusion proteins were isolated using amylose resin and then bound to a column of 4% beaded agarose (AminoLink Plus; Pierce, Rockford, IL). Antisera from the immunized rabbits were affinity purified using the corresponding MBP–whirlin fusion protein.

Immunocytochemistry.

The inner ears of adult and postnatal C57BL/6 (Charles River Labs, Wilmington, MA), Myo15ash2 and Whrnwi mice (Jackson Laboratories, Bar Harbor, ME) — ranging from P0 to P30, as well as adult (120–150 g) and postnatal P0–P10 Sprague-Dawley rats (Taconic, Germantown, NY) — were dissected from temporal bones and fixed with 4% paraformaldehyde in phosphate-buffered saline for 2 h. Cultured inner-ear sensory epithelia explants were fixed overnight with 4% paraformaldehyde at 4°C. Auditory and vestibular sensory epithelia were dissected, immunostained and mounted using ProLong Antifade Kit (Molecular Probes, Eugene, OR). Immunostaining — using TF1, and HL5136, HL5137, HL5140 and HL5141 antibodies — was performed, as described previously19. F-actin was visualized by rhodamine-phalloidin or Alexa Fluor 633 phalloidin (Molecular Probes) staining19.

Myosin-XVa expression constructs.

There are two main full-length isoforms of Myo15a in the mouse inner ear: one with and one without exon 2, with the isoform without exon 2 being the predominant form19. GFP-tagged cDNA constructs containing the entire open reading frame (ORF) of each isoform were generated previously19. Due to the low efficiency of transfection with an exon-2-containing GFP-Myo15a cDNA expression construct19, this study used a full-length cDNA of the Myo15a isoform without exon 2 (NCBI accession number AY331133), which we previously designated [−N]Myo15a-GFP19 and refer to here as GFP-Myo15a. The two GFP-Myo15a expression constructs containing missense mutations within the motor domain (C592Y and R167A;G388A) were generated using a QuikChange Site-Directed Mutagenesis kit (Stratagene). To generate the GFP-Myo15a construct lacking the PDZ-ligand (deletion of sequence encoding amino acids 2303–2306; NCBI accession number AY331133), the ORF was PCR amplified using Pfu Ultra DNA polymerase (Stratagene) from the aforementioned cloned Myo15a cDNA that was inserted into the EcoRI and SalI sites of GFP-C2 (Clontech, Palo Alto, CA) and verified by DNA sequencing. The sequence of PCR primers can be found as Supplementary Information, Table 2 online.

Myosins Ic, VI, VIIa and X expression vectors.

The ORF of mouse Myo1c was PCR amplified from IMAGE clone 5344331 using Pfu Ultra DNA polymerase (Stratagene). The PCR product was cloned into the HindIII and SacII sites of GFP-C3 (Clontech) and sequence-verified. A full-length Myo7a ORF (NCBI accession number AY821853) was PCR amplified from mouse P5 inner-ear cDNA using LA Taq DNA Polymerase (Takara Mirus, Madison, WI) and cloned into the EcoRI and SalI sites of GFP-C2 vector (Clontech).

The ORF of human MYO6 was PCR amplified from human kidney cDNA and cloned into the EcoRI and SalI sites of GFP-C2 (Clontech) and sequence-verified (Supplementary Information, Tables 2 and 3 online). A bovine GFP-Myo10 expression construct29 was a gift from Dr Richard Cheney.

Whirlin expression constructs.

Using mouse P5 vestibular cDNA as template, full-length Whrn cDNA, including the 5'-untranslated region (UTR) and 3'-UTR was PCR amplified with LA Taq DNA Polymerase (Takara Mirus), cloned into PCR-XL-TOPO (Invitrogen, Carlsbad, CA), transformed into XL10 Gold cells (Stratagene) and the entire insert was sequence verified. The predominant vestibular Whrn cDNA isoform encoded the expected 907 amino-acid protein (NCBI accession number AY739114). Additionally, nine novel Whrn isoforms encoding predicted protein products of 156, 366, 403, 465, 476, 550, 906, 911 and 918 amino-acid residues were found (NCBI accession numbers AY739115–AY739122).

To construct a wild-type cDNA encoding the 907 amino-acid isoform of whirlin, the entire ORF was PCR amplified from cloned Whrn cDNA using LA Taq DNA Polymerase (Takara Mirus) and inserted into the EcoRI and SalI sites of GFP-C2 (Clontech). We also constructed a mutant cDNA encoding the whirlin protein that lacks the third PDZ domain [−PDZ3] equivalent to the R778X mutant allele, as reported by Mburu and coauthors6. Wild-type DsRed-Whrn and DsRed-Whrn[−PDZ3] expression constructs were generated using an identical procedure, as described above and inserted, into the EcoRI and SalI sites of DsRed2-C1 (Clontech). PCR primer sequences are given in Supplementary Information, Table 2 online.

Culture and transfection of inner-ear sensory epithelium.

Inner-ear sensory epithelium cultures were prepared from organ of Corti, saccule, utricle and ampulae of P1–P4 C57Bl/6, Myo15ash2 and Whrnwi mice, and from organ of Corti of P2–P3 rats, as described previously19. Cultures were then transfected using a Helios gene gun (Bio-Rad Labs, Hercules, CA). Gold particles (1.0 μm; Bio-Rad) were coated with plasmid DNA at a ratio of two μg plasmid DNA to one μg of gold particles and precipitated onto the inner wall of Tefzel tubing, which was cut into individual cartridges containing ∼1 μg of plasmid DNA. For some experiments, cartridges were prepared using a mix of GFP-Myo15a and DsRed-Whrn precipitated on the same gold particles. Samples were bombarded with the gold particles from one cartridge per culture, using 120 psi of helium. After an additional 8 h to 4 days in culture, samples were fixed in 4% paraformaldehyde, stained with rhodamine-phalloidin or Alexa Fluor 633 phalloidin and observed using a LSM510 confocal microscope (Zeiss, Thornwood, NY) equipped with a x 100, 1.45 NA objective.

In most sensory epithelium explants, we were able to find 2–8 transfected hair cells. Restoration of mutant hair bundles after transfection of sensory epithelial explants was evaluated by comparison of the height of mutant transfected and mutant non-transfected hair cells, as well as normal hair bundles from explants of sensory epithelia from heterozygous mice of the corresponding strain and age. Heights of the hair-cell bundles were measured as the length of the tallest stereocilia within a bundle, using LSM510 software (Zeiss). Data were presented as an average value ± standard deviation. Statistical significance of the differences of the bundle heights in transfected and non-transfected hair cells was estimated using a student t-test.

COS7 cell culture and time-lapse imaging.

African green monkey kidney cells (COS7) were cultured at 37 °C and 5% CO2 in DMEM supplemented with 10% fetal bovine serum and 10 mM HEPES. Using Lipofectamine 2000 (Invitrogen), cells were transfected with wild-type and mutant GFP-Myo15a expression constructs and wild-type and mutant GFP- and DsRed-Whrn expression constructs. In fixed tissue samples, colocalization of myosin-XVa and whirlin were analysed using LSM510 software (Zeiss).

For time-lapse imaging, COS7 cells were grown and transfected in the glass-bottom Petri dishes. Cells were observed with a Nikon TE300 inverted microscope equipped with a ×100, 1.3 NA objective. Temperature was maintained at 37 °C (±2 °C) using an objective heater and a heating stage. Epifluorescent and bright-field images were acquired every 10 or 20 s with a ORCA-II-ER cooled charge-coupled device camera (Hamamatsu Co., Hamamatsu City, Japan) controlled with MetaMorph software (Universal Imaging Co., Downington, PA). The same software controlled fast exchange of the fluorescent filters, acquiring GFP and DsRed images sequentially within 1–1.5 s, which covered no more than 15% of the shortest time-lapse interval. Photobleaching was performed by computer-controlled fast exchange of an additional neutral density filter (ND 2.0) to the pinhole. Particle tracking and fluorescent intensity measurements were performed using MetaMorph software (Universal Imaging Co.).

Note: Supplementary Information is available on the Nature Cell Biology website.