To better understand how Eps8 influences stereocilia lengths, we investigated the localization of Eps8 in hair cells. Immunofluorescence labeling of mouse and rat cochlear and vestibular sensory tissue shows localization of Eps8 at the tips of stereocilia (Figures 1b and S1). This tip-labeling persisted from the early stages of stereocilia elongation at birth through to adulthood, similar to MyoXVa [7] and whirlin [8]. Quantification of Eps8 immunofluorescence shows that Eps8 appears at stereocilia tips in concentrations proportional to stereocilia length (Figures 1b and S1), much like MyoXVa [7]. Exogenous expression of GFP-Eps8 or cherry-Eps8 (Figure S1) in transfected hair cells from four days old rats confirms the length proportional amounts of Eps8 at stereocilia tips. This also demonstrates that Eps8 is continuously targeted to stereocilia tips in concentrations that are actively maintained after stereocilia have already elongated and reached steady-state lengths. It is noteworthy that other than Eps8, espin-1, a MyoIIIa cargo [9, 10], is the only other actin-regulatory protein reported to be at the tips of stereocilia in amounts proportional to length. Interestingly, actin polymerization rates in stereocilia also show a similar relationship to stereocilia length [7, 11].

The striking similarity in both localization of MyoXVa [7], whirlin [12], and Eps8 labeling and the similar shortening of stereocilia when these proteins are absent or non-functional led us to hypothesize that MyoXVa transports Eps8 to stereocilia tips. Using the same antibody to Eps8 on shaker-2 and whirler mouse tissue, we found that Eps8 labeling was absent or reduced (respectively) at stereocilia tips in cochlear (Figure 1b) and vestibular (data not shown) hair cells. This indicates that Eps8 tip-localization is completely dependent on the presence of MyoXVa but can localize to the tips, albeit in smaller amounts, in the absence of whirlin. Interestingly, we found that MyoXVa labeling at the tips of whirler mouse stereocilia was reduced by ~50%, similarly to Eps8 (Figure S2). This result, along with the gradation in stereocilia length from whirler to Eps8 KO to shaker-2 mice, suggests that the scaffolding protein whirlin facilitates stereocilia elongation by stabilizing the MyoXVa:Eps8 complex at the tips of stereocilia.

Further supporting the hypothesis that MyoXVa elongates stereocilia by transporting Eps8 to stereocilia tips, overexpression of GFP-MyoXVa in wild-type hair cells resulted in stereocilia elongation and concomitant enrichment of Eps8 at stereocilia tips (Figure 2). When compared to neighboring non-transfected control cells, hair cells transfected with MyoXVa showed slight elongation of the tallest row of stereocilia (9% increase, n=55, p<0.05) and even more elongation of the middle row of stereocilia (39%, n=48, p<0.0001). Thus, in hair cells overexpressing GFP-MyoXVa, the ratio of heights between the tallest and middle row of stereocilia significantly decreases (Figure 2a–b), demonstrating that the staircase patterning of the hair cell bundle is sensitive to the dynamic sorting of myosin concentrations in each row. These data suggest that not only is MyoXVa essential for the elongation of stereocilia during development [12], but MyoXVa can also dynamically regulate stereocilia length after elongation has terminated [13]. A previous study showed that there is no detectable increase in stereocilia length beyond wild-type lengths when MyoXVa is overexpressed in shaker-2 mutant vestibular hair cells [12]. However, shaker-2 stereocilia do not develop normally and have structural and mechanotransduction defects [14] caused by the lack of a functional myoXVa expression during their initial elongation. Furthermore, Shaker-2 hair cells express a motor-dead MyoXVa, which potentially has dominant-negative effects.

When comparing Eps8 labeling at the tips of the tallest row of stereocilia in GFP-MyoXVa transfected vs. non-transfected cells (Figure 2c), we found that Eps8 immunofluorescence is enhanced in transfected cells (transfected: relative intensity RI ± S.E.M. = 8.5 ± 1.8, n=13; non-transfected: 4.7 ± 0.9, n=20). We also observed that occasionally when MyoXVa is overexpressed, it appears in a “bulbous” pattern at stereocilia tips (Figure 2c) and when this happens, Eps8 also appears with the same pattern further suggesting that these two proteins strongly colocalize. Finally, when shaker-2 hair cells were transfected with GFP-MyoXVa, Eps8 tip-localization was rescued (Figure 2c).

Heterologous expression of GFP-MyoXVa in organ of Corti and vestibular supporting cells increases the amount of Eps8 labeling at the tips of their microvilli (Figure 3a). Similarly, Eps8 often decorates the tips of filopodia in Cos-7 cells, but when cells were transfected with GFP-MyoXVa Eps8 labeling was increased (Figure 3b) from (RI ± S.E.M.) 1.1 ± 0.12 (n_filopodia=20, n_cells=19) to 3.0 ± 0.30 (n_filopodia=25, n_cells=14). In contrast, GFP-MyoX [15], an unconventional myosin implicated in actin protrusion elongation that also possesses a tail with a MyTH4-FERM domain, similar to the two MyTH4-FERM domains found in MyoXVa, failed to enhance Eps8 accumulation at filopodia tips (RI = 0.6 ± 0.14, n_filopodia=20, n_cells=16) (Figure 3b). GFP-MyoIIIa, a hair cell myosin that elongates actin protrusions by transporting espin-1 to stereocilia tips [10], also failed to increase Eps8 amounts at filopodia tips (RI = 0.3 ± 0.15, n_filopodia=10, n_cells=6). The enhanced accumulation of Eps8 at filopodia tips specifically in the presence of GFP-MyoXVa suggests that MyoXVa motor activity can maintain higher concentrations of Eps8 at the tips of actin protrusions. This should be particularly important in longer protrusions such as stereocilia, in which case passive diffusion over much longer distances could result in much lower concentrations of Eps8 available for binding at the tip, especially since Eps8 can also bind to the sides of actin filaments [16].

Consistent with the hypothesis that Eps8 and MyoXVa cooperate to elongate stereocilia in hair cells, we found that co-overexpression of these two proteins in Cos-7 cells (Figure 3c) elongates actin protrusions in a cooperative manner. We found that Cos-7 cells overexpressing cherry-Eps8 and GFP-MyoXVa display actin protrusions almost twice as long (mean length ± S.E.M. = 3.0 ± 0.2 μm, n_filopodia=84, n_cells=15) as cells overexpressing either cherry-Eps8 alone (1.7 ± 0.1 μm, n_filopodia=66, n_cells=12) or GFP-MyoXVa alone (1.6 ± 0.1 μm, n_filopodia=81, n_cells=16) (Figure 3c). In contrast, both GFP-MyoIIIa and GFPMyoX failed to further elongate filopodia when co-expressed with cherry-Eps8 (MyoIIIa: 1.4 ± 0.08 μm, n_filopodia=46, n_cells=11; MyoX: 1.4 ± .05 μm, n_filopodia=53, n_cells=19). This is consistent with the inability of these myosins to target Eps8 to filopodia tips, and is also consistent with the similar finding that GFP-MyoXVa failed to cooperate with the MyoIIIa cargo espin 1 in filopodia elongation [10]. It should be noted that our Eps8 overexpression experiments are within the context of endogenous Eps8 as well as Eps8 regulators, e.g. Abi-1, which could be influencing Eps8-overexpression mediated elongation activity via activation of Eps8 capping activity [17].

To further test the hypothesis that MyoXVa transports Eps8 to the tips of actin protrusions, we imaged live Cos-7 cells co-expressing cherry-MyoXVa and a GFP-Eps8 construct lacking the SH3, as well as actin capping and bundling activity (GFP-Eps8ΔSH3ΔCΔB). GFP-Eps8ΔSH3ΔCΔB, like full-length Eps8 (Figure 3b), colocalized with cherry-MyoXVa in transfected cells, and we observed co-migration of green and red fluorescent puncta in filopodia from the tip to the base of the filopodia at ~25 nm/s (Figure S3). Removing the actin-binding domains of Eps8 allowed us to rule out the possibility that co-migration of cherry-Eps8 and GFP-MyoXVa is due only to Eps8 binding to the treadmilling actin filaments. Removing the SH3 domain disrupts Eps8 binding to a number of potential interactors [18, 19]. Thus, the co-transport of these two proteins, even when Eps8 lacks actin-binding activity and its SH3 domain, further supports the hypothesis that MyoXVa transports Eps8.

The extreme tip localization of Eps8 led us to also question whether Eps8 is self-targeted to stereocilia tips by the activity of its actin capping domain which has a high affinity for F-actin barbed-ends [5, 17]. However, a GFP-Eps8 construct lacking its capping activity (Eps8ΔCapping [5]) or its SH3 domain (not shown) was still targeted to stereocilia tips in a length-dependent fashion (Figure S4), consistent with MyoXVa-dependent transport.

Notably, whirlin may also contribute to Eps8 targeting. Consistent with a model in which the N-terminus of Eps8 binding to whirlin enhances stereocilia tip targeting, a GFP-Eps8 construct lacking the first 280 amino acids in the Eps8 N-terminus (GFP-Eps8ΔPTBΔPR) targeted stereocilia tips much less efficiently (Figure S4). This result suggests that the first 280 amino acids of Eps8, which contains the phosphotyrosine binding domain (PTB) and the N-terminus proline rich domain (PR) of Eps8, is required for efficient targeting to stereocilia tips, perhaps via binding to whirlin or another protein.

In order to determine whether interactions between Eps8, whirlin, and MyoXVa can occur in vitro, we performed GST-pull-down experiments on Eps8 fragments with GFP-MyoXVa and GFP-whirlin (Figure 4). We found that MyoXVa specifically binds to the C-terminus of Eps8, and perhaps weakly binds to the N-terminus. A yeast two-hybrid screen of human proteins revealed an interaction between the second MyTh4-FERM domain of the MyoXVa tail and Eps8 [20]. A MyoXVa construct with a tail truncated just past the SH3 domain was generated. This construct, which lacks the second MyTh4-FERM domain (cherry-MyoXVaΔ) as well as the C-terminus PDZ-binding domain (PBD) still interacted with Eps8 in our GST-pull-down experiment, suggesting that another region of the MyoXVa tail can interact with Eps8. We also found that the N-terminus of Eps8 interacts with whirlin. Consistent with previous results showing that MyoXVa needs its C-terminus PBD to interact with whirlin [12], we found that the co-expression of MyoXVa: whirlin, but not MyoXVaΔ:whirlin, allowed both the C- and N-termini of Eps8 to interact with the complex. These results suggest that whirlin and Eps8 interact with each other and with MyoXVa independently of one another, but that all three proteins can also form a complex (Figure 4e).

Taken together, our results show that the MyoXVa:whirlin:Eps8 complex at stereocilia tips is essential for stereocilia elongation. The similarity between whirler and Eps8 KO mouse stereocilia (Figure 1) suggests that Eps8 is the molecule directly involved in whirlin's role in regulating stereocilia actin. Our results further suggest that one of the functions of whirlin in this process is to serve as a scaffold for the proper assembly, stability, and targeting of the MyoXVa:Eps8 complex. The identification of Eps8 as a key component of the MyoXVa:whirlin motor:cargo complex lends novel insight towards the physiopathology of the human deafness associated with mutations in MYO15A (DFNB3 [1]) and WHRL (DFNB31/USH2D [2, 21]) genes and furthermore explains the heretofore inexplicable effect of Myo15 and Whrl mutations on stereocilia length in the shaker-2 [3] and whirler [2] mice, respectively. However, it still remains to be elucidated how exactly Eps8 regulates actin protrusion length.

01