SHANK2 establishes auditory hair bundle architecture essential for mammalian hearing

Significance

Mammals have evolved distinct bundle geometries in their auditory sensory cells: linear bundles in inner hair cells (IHCs) for sound detection and V-shaped bundles in outer hair cells (OHCs) for sound amplification. Here, we identify SHANK2 as a key regulator of these characteristic bundle geometries. When SHANK2 or its upstream regulator RAP1 is absent, both IHCs and OHCs lose their distinct bundle architecture, and OHCs exhibit impaired high-frequency amplification, resulting in high-frequency-specific hearing loss. Over time, this architectural disruption leads to progressive bundle degeneration and further auditory decline. These findings demonstrate that the characteristic bundle architecture—particularly the V-shaped geometry of OHCs—is essential for high-frequency hearing and long-term cochlear integrity.

Abstract

The mammalian auditory system relies on the precise architecture of the hair cell stereociliary bundle for effective sound transduction. Each bundle consists of approximately 100 actin-filled stereocilia arranged in a three-row staircase pattern, forming a linear shape in inner hair cells (IHCs) and a V-shape in outer hair cells (OHCs), the latter geometry being a hallmark of the mammalian cochlea. While the initial development from uniformly distributed microvilli into stereociliary bundles is guided by lateral migration of the kinocilium, the mechanisms that establish the characteristic bundle architecture and its functional significance remain unclear. Here, we show that SHANK2, a protein implicated in synaptic function and autism spectrum disorders, is a critical regulator of bundle architecture. SHANK2 localizes to the medial apical surface of developing hair cells. This localization is regulated by the small GTPase RAP1, independently of known lateral (Gαi, GPSM2) or medial (aPKCζ, PARD6B) proteins. Hair cell-specific ablation of Shank2 or Rap1 disrupts bundle architecture while preserving key features essential for mechanotransduction. In particular, OHCs lose their unique bundle geometry and show impaired amplification, especially at high frequencies. Longitudinal studies further reveal that this architectural disruption leads to progressive bundle degeneration and hearing loss. These findings suggest that the characteristic bundle architecture, particularly the V-shaped geometry of OHCs, is essential for high-frequency hearing and long-term bundle integrity in the mammalian cochlea.

The mammalian cochlea contains specialized sensory hair cells (HCs) that are structurally and functionally optimized to convert sound vibrations into electrical signals. Each HC contains a stereociliary (hair) bundle erected on the apical surface, comprising approximately 100 actin-filled microvilli arranged in a three-row staircase pattern. The mammalian cochlea contains two types of HCs: inner hair cells (IHCs) and outer hair cells (OHCs). IHCs, which receive most of the afferent innervation, are responsible for transmitting sound information to the brain, and the bundle forms a linear or crescent shape. On the other hand, OHCs, which are responsible for amplifying and fine-tuning sound signals to enhance cochlear sensitivity and selectivity, form V-shape bundles, which is uniquely observed in the mammalian cochlea (1, 2). The kinocilium, a microtubule-based primary cilium of HCs, is located at the vertex of the V-shaped arrangement and is consistently oriented laterally along the cochlear duct.

During early development, nascent HCs have a symmetrical apical surface arrangement with the kinocilium centrally positioned among evenly distributed microvilli (3). As development proceeds, the kinocilium migrates laterally, creating a microvilli-free bare zone on the lateral surface (4, 5). The remaining microvilli on the medial surface then form the linear- or V-shaped bundle in a three-row staircase configuration (3). Deflection of the bundle toward the kinocilium opens the mechanotransduction channels, triggering electrical activity in the HCs (1, 2). This asymmetrical architecture, along with the uniform orientation of these bundles along the cochlear axis, is crucial for generating synchronous and directional responses to sound waves (6, 7).

Several factors regulate this uniform orientation of hair bundles along the cochlear duct including the planar cell polarity (PCP) signaling pathway. Mutations in core PCP components such as VANGL2, SCRB1, DVL1/2, and FZ3/6 cause misorientation of the kinocilium and hair bundles, but typically preserve individual bundle architecture, including the three-row staircase (6, 8–10). Another group of proteins, such as Gαi and GPSM2, are asymmetrically located on the lateral side of the apical surface of nascent HCs and play important roles in the formation of the individual V-shaped bundle architecture (4–6). In particular, Gαi regulates kinocilium migration via pulling forces on the microtubule cytoskeleton (5, 11). Mutations in these proteins affect kinocilium positioning, resulting in disrupted bundle architecture and loss of the three-row staircase pattern (4, 5). In addition, loss of function of Daple and PARD3, which are also located on the lateral junction and link PCP components to the microtubule network, disrupts both uniform orientation and individual bundle architecture (12–14).

In addition, ciliary proteins at the basal body or kinocilium also regulate bundle morphology through mechanisms distinct from lateral polarity pathways. Mutations in Kif3a, Ift88, and Ift20 disrupt ciliogenesis, leading to loss of the kinocilium and resulting in flattened and misoriented hair bundles (15–17). Notably, many of these mutants exhibit cochlear shortening and supernumerary rows of hair cells, phenotypes attributed to impaired sonic hedgehog signaling (15–18). The kinocilial link isoform Pcdh15-CD2 is also essential for establishing the V-shaped geometry by anchoring the tallest stereocilia to the kinocilium (19). Additional ciliary mutants, such as Bbs8 and Alms1, also display dysmorphic bundles and kinocilium mislocalization, further supporting a mechanical role of the kinocilium in shaping bundle morphology during development (16, 20).

Despite significant progress in understanding hair bundle development, two fundamental questions remain unclear: What mechanisms act on the medial surface to establish the characteristic bundle architecture, and what is its functional significance in mammalian hearing? The V-shaped geometry of OHC bundles is uniquely observed in mammals, suggesting its potential importance for mammalian-specific hearing capabilities. However, most existing mouse models with disrupted bundle morphology also exhibit concurrent defects in HC orientation, staircase organization, or cochlear extension (5, 12–18, 21, 22). These concurrent defects result in hearing loss unrelated to bundle architecture, making it difficult to determine the specific role of the unique bundle architecture in vivo.

In this study, we identified SHANK2, a protein associated with synaptic function and autism spectrum disorders (ASD) (23), as a critical regulator of bundle architecture in the mammalian cochlea. SHANK2 specifically localizes to the medial apical surface of developing HCs, providing an insight into medial surface regulation of bundle morphogenesis. Notably, loss of SHANK2 specifically disrupts bundle architecture while preserving key features essential for hearing, such as uniform HC orientation, the three-row staircase organization, and proper kinocilium positioning. This selective disruption allowed us to demonstrate that the unique bundle geometry of OHCs is essential to effectively amplify high-frequency sounds and maintain long-term bundle integrity and auditory function.

Results

SHANK2 Is Specifically Expressed On the Medial Surface of Developing HCs.

We performed in situ hybridization and whole-mount immunostaining to determine the localization of SHANK2 in the developing mouse cochlea (Fig. 1 and SI Appendix, Fig. S1). At embryonic day 16.5 (E16.5), Shank2 mRNA was weakly expressed in the sensory region of the basal cochlea and in spiral ganglion neurons (SGNs) (SI Appendix, Fig. S1 A–D). By postnatal day 1 (P1), we detected stronger Shank2 expression in HCs, with a base-to-apex gradient, and weak expression in SGNs in all cochlear turns (SI Appendix, Fig. S1 E–H). Shank2 expression decreased from the basal turn at P7 and was barely detectable by P14 (SI Appendix, Fig. S1 I–P).

Fig. 1.

Spatiotemporal localization of SHANK2 during hair cell development. (A) SHANK2 and Gαi immunostaining at E15.5. Neither protein is detected in immature IHCs (Aa). Both emerge simultaneously: Gαi from lateral and SHANK2 from medial edges (Ab). In more differentiated IHCs, Gαi localizes to the bare zone, while SHANK2 occupies the medial side (Ac-Ad). (B–E) At P1, SHANK2 covers the medial surface while Gαi and GPSM2 localize to the bare zone (B and C, white dashed line). Other medial proteins, aPKCζ and PARD6B, show similar restriction to the medial surface (D and E). (F) At P5, SHANK2 remains medial, with the strongest expression close to the medial junction and some expression at stereocilia base. (G) Schematic of medial and lateral protein localization during HC cytoskeletal morphogenesis. [Scale bar, A, 5 μm; Aa-Ad, 2.5 μm; B–E (Left panels), 5 μm; B–E (Middle/Right panels), 2.5 μm; F (Left panel), 5 μm; F (IHC/OHC), 2.5 μm.]

Immunostaining revealed that SHANK2 protein specifically localizes to the medial side of the apical surface in differentiating HCs (SI Appendix, Fig. S2A). The specificity of SHANK2 immunoreactivity was confirmed by its absence in Shank2^–/– cochlea (SI Appendix, Fig. S2B). At E15.5, SHANK2 appeared on the medial surface simultaneously with the lateral appearance of Gαi (Fig. 1, Ab) (5). As differentiation progressed, SHANK2 expanded laterally without overlapping with Gαi (Fig. 1, Ac–Ad). At P1, SHANK2 showed distinct medial localization, opposite to the lateral localization of Gαi and GPSM2 (Fig. 1 B and C). Unlike other medial proteins aPKCζ and PARD6B (4, 5), which broadly cover the medial surface except the lateral bare zone (Fig. 1 D and E), SHANK2 showed more restricted medial localization. This specific pattern persisted at P5, with SHANK2 more restricted to the medial apical surface, whereas aPKCζ showed strong expression near the base of the stereocilia (Fig. 1 F and G). Costaining with the tight junction marker ZO-1 showed that SHANK2 was consistently localized to the medial apical surface, positioned nonoverlapping with the ZO-1-positive apical junctional boundary at both P1 and P5 (SI Appendix, Fig. S2C). SHANK2 maintained similar complementary localization to Gαi in vestibular HCs (SI Appendix, Fig. S2D).

SHANK2 Regulates Hair Bundle Architecture Independently of Kinocilium Positioning.

Given the medial localization of SHANK2 opposite to Gαi and GPSM2, which regulate bundle arrangement from the lateral surface (4, 5), we investigated the role of SHANK2 in hair bundle development. In E17.5 wild-type mice (Shank2^+/+), developing hair bundles showed an inverted V or crescent shape at the base, with less pronounced shapes toward the apex, reflecting the basal-to-apical gradient of HC differentiation (Fig. 2A). In Shank2^–/– mutants, both IHC and OHC bundles appeared wavy or fragmented (Fig. 2A). Scanning electron microscopy (SEM) at P0 and P7 confirmed these defects in bundle architecture (Fig. 2 B and C). Notably, we found that vestibular HC bundles remained intact in Shank2^–/– mutants (SI Appendix, Fig. S2E).

Fig. 2.

Shank2^–/– mice show hair bundle defects in developing cochlea. (A) Hair bundle morphology visualized by phalloidin staining at E17.5. Wild-type mice show inverted V-shaped bundles at the base and developing bundle shapes in the middle and apex. Shank2^–/–mice show split or wavy bundles. (B and C) SEM analysis of stereocilia. Bundle defects persist in both IHCs and OHCs at P0 (B) and P7 (C). [Scale bar, A (low magnification), 5 μm; A (IHC/OHC), 2 μm; B (low magnification), 10 μm; B (IHC/OHC), 1 μm; C (low magnification), 10 μm; C (IHC/OHC), 1 μm.]

In contrast to mutations in the lateral proteins Gαi and GPSM2, which disrupt both bundle structure and kinocilium positioning (4, 5), Shank2^–/– mutants maintained normal kinocilium and basal body positioning (SI Appendix, Fig. S3). Immunofluorescence staining for γ-tubulin (basal body) and ARL13B (kinocilium) at P1 showed consistent lateral localization of both structures in control and Shank2^–/– cochleae (SI Appendix, Fig. S3 A–C). These findings suggest that SHANK2 regulates hair bundle architecture independently of the mechanisms that control kinocilium biogenesis or migration.

Mature Hair Bundle Architecture Is Disrupted in Shank2 Mutants.

We next examined the morphology of the hair bundles in the mature cochlea. At 3 wk, wild-type mice showed a characteristic bundle architecture: linear in IHCs and V-shaped in OHCs, both arranged in a three-row staircase pattern (Fig. 3A and SI Appendix, Fig. S4A). In Shank2^–/– mice, HCs showed disrupted bundle architecture along the cochlear duct (Fig. 3B). Quantitative analysis revealed differential effects: While most OHCs lost their V-shaped architecture, IHCs showed fragmentation in about half of the cells (Fig. 3 C–E). Despite the disrupted bundle shapes, these cells retained their three-row staircase arrangement (SI Appendix, Fig. S4A) and essential stereociliary links, including tip-links and tectorial membrane (TM) imprints that connect OHC stereocilia to the TM (24) (SI Appendix, Fig. S4 B and C). These results suggest that SHANK2 regulates bundle architecture without affecting other structural features critical for mechanotransduction.

Fig. 3.

Shank2^–/– mice show disrupted bundle architecture and impaired high frequency hearing at 3 wk. (A and B) SEM images showing hair bundles along the cochlear duct: linear IHC and V-shaped OHC bundles in Shank2^+/+ mice (A) and fragmented or wavy hair bundles in Shank2^–/– mice (B). (C) Representative examples of OHC bundle defect categories. (D and E) Quantification of bundle defects in OHC (D) and IHC (E) in Shank2^+/+ (WT), Shank2^+/– (HET), and Shank2^–/– (KO) mice. (F and G) ABR measurements in 3 wk. Shank2^–/– mice show elevated ABR thresholds at high frequencies (F) and reduced ABR wave I amplitudes specifically at 30 kHz, but not lower frequencies (G). (H–K) DPOAE measurements. Shank2^–/– mice exhibit elevated DPOAE thresholds at 30 kHz, with reduced 2f₁-f₂ DPOAE amplitudes at 30 kHz (J) but not at 18 kHz (I). DPOAE amplitudes show progressive reduction toward higher frequencies (K). [Scale bar, A and B (low magnification), 10 μm; A and B (IHC/OHC), 1 μm.] Data are presented as means ± SD. Statistical comparisons were made using two-way ANOVA with Bonferroni corrections for multiple comparisons (n.s., nonsignificant, *P < 0.05, **P < 0.01, and ***P < 0.001).

Loss of SHANK2 Specifically Affects High-Frequency Amplification.

We next assessed auditory function at the onset of hearing (3 wk). Despite widespread bundle defects along the cochlear duct (Fig. 3 D and E), Shank2^–/– mice showed ABR threshold elevations only at high frequencies, with normal thresholds at lower frequencies or to click stimuli (Fig. 3F).

To investigate the underlying mechanism of high-frequency-specific hearing deficits, we analyzed ABR wave I input/output (I/O) functions, which measure the magnitude of synchronized auditory nerve responses to sounds of increasing intensity, reflecting the efficiency of synaptic transmission between IHCs and SGNs (25). Shank2^–/– mutants showed significantly reduced wave I amplitudes specifically at high frequency (30 kHz), but normal amplitudes at lower frequencies (6, 18, and 24 kHz) (Fig. 3G). While the overall response magnitude was reduced at 30 kHz, the slope of the I/O function was preserved (Fig. 3G), suggesting that IHCs retain their ability to encode sound intensity differences. Normal interpeak latencies of ABR waves I–IV and I–V indicated that central auditory processing is unaffected in Shank2^–/– mutants (26) (SI Appendix, Fig. S5 A–E).

To determine whether the reduced auditory nerve responses were due to impaired OHC amplification, we measured distortion product otoacoustic emissions (DPOAEs). Shank2^–/– mutants showed elevated DPOAE thresholds specifically at high frequencies (Fig. 3H), which correlated with ABR threshold shifts (Fig. 3F). Frequency spectrum analysis revealed a significantly reduced 2f₁-f₂ DPOAE amplitude at 30 kHz but not at 18 kHz (Fig. 3 I and J). The I/O functions of the DPOAE amplitudes showed progressive reduction toward higher frequencies (Fig. 3K). These results suggest that loss of the characteristic bundle architecture impairs OHC amplification specifically at high frequencies, resulting in high-frequency-specific hearing loss.

SHANK2 Function in Hair Cells Is Essential for Normal Hearing.

SHANK2 serves as a critical synaptic scaffolding protein in the brain (23). However, our analyses indicate that SHANK2’s role in hearing is primarily related to HC function, particularly OHCs, rather than to auditory nerve activation or central processing (Fig. 3 and SI Appendix, Fig. S5 A–E). This functional specificity is consistent with the distinct expression patterns of SHANK family genes (Shank1, Shank2, and Shank3) in the cochlea (27–30). RNA-sequencing data show that Shank2 is predominantly expressed in embryonic HCs, whereas Shank1 is most abundant in SGNs with minimal expression of Shank2 and Shank3 (SI Appendix, Fig. S6A) (29). In the mature cochlea, all three SHANK genes are downregulated in HCs (SI Appendix, Fig. S6B) (27, 28), whereas Shank1 maintains expression in all SGN subtypes (SI Appendix, Fig. S6C) (30), consistent with the observed localization of SHANK1, but not SHANK2 or SHANK3, in the postsynaptic densities of SGNs (31).

To confirm the cell-specific role of SHANK2 in hearing, we generated HC-specific (Gfi1^Cre; Shank2^lox/lox) and SGN-specific (Bhlhe22^Cre; Shank2^lox/lox) Shank2 conditional knockout (cKO) mutants. At 3 wk, HC-specific, but not SGN-specific, Shank2 mutants phenocopied the bundle defects of systemic Shank2^–/– mutants (SI Appendix, Figs. S7 A–D and S8B). Consistently, HC-specific Shank2 mutants showed high-frequency-specific elevation in ABR and DPOAE thresholds (SI Appendix, Fig. S7 E and G). Similar to systemic Shank2 mutants (Fig. 3), HC-specific mutants showed reduced ABR wave I amplitudes only at 30 kHz (SI Appendix, Fig. S7F), normal interpeak latencies of ABR waves I–IV and I–V (SI Appendix, Fig. S5F), and progressive reduction of DPOAE amplitudes toward high frequencies (SI Appendix, Fig. S7H). In contrast, SGN-specific Shank2 mutants showed no significant differences compared to controls (SI Appendix, Fig. S8 C–F), consistent with minimal Shank2 expression in SGNs (SI Appendix, Fig. S6) (30). These results demonstrate that SHANK2 functions specifically in HCs to establish the bundle architecture essential for high-frequency amplification.

Disrupted Bundle Architecture Leads to Progressive Hearing Loss.

To assess the long-term consequences of disrupted bundle architecture, we performed a longitudinal auditory measurements in Shank2 mutants for up to 16 wk (SI Appendix, Fig. S9). Both systemic and HC-specific Shank2 mutants showed progressive expansion of high-frequency threshold shifts toward low frequencies (Fig. 4 A and B and SI Appendix, Fig. S9 A–D), whereas SGN-specific mutants maintained normal thresholds (SI Appendix, Fig. S9 E and F).

Fig. 4.

Progressive hearing loss and bundle degeneration in Shank2 mutants. (A and B) Auditory function at 16 wk. Both systemic (Shank2^–/–) and HC-specific (Gfi1^Cre; Shank2^lox/lox) mutants show elevated ABR and DPOAE thresholds at mid to high frequencies (A and B). (C) SEM images showing severe hair bundle degeneration in systemic and HC-specific Shank2 mutants, but not in SGN-specific (Bhlhe22^Cre; Shank2^lox/lox) mutants. (D) Examples of OHC bundle defect categories. (E–H) Quantification of bundle defects in OHCs (E and F) and IHCs (G and H) of 16-wk-old systemic and HC-specific mutants. [Scale bar, C (low magnification), 10 μm; C (IHC/OHC), 1 μm.] Data are presented as means ± SD. Statistical comparisons were made using two-way ANOVA with Bonferroni corrections for multiple comparisons (n.s., nonsignificant, *P < 0.05, **P < 0.01, and ***P < 0.001).

Consistent with progressive hearing loss, systemic and HC-specific mutants exhibited significant bundle degeneration at 16 wk, whereas SGN-specific mutants maintained normal bundles (Fig. 4C). Bundle loss was extensive in basal OHCs but minimal in the middle and apical regions (Fig. 4 D–F). IHCs showed moderate bundle loss in the base region but minimal loss in the middle and apical regions (Fig. 4 G and H). These findings suggest that loss of the characteristic bundle architecture predisposes disrupted bundles to progressive degeneration, more so in the basal cochlea. Interestingly, the progression of threshold shifts did not directly correlate with regions of bundle degeneration: Despite significant threshold shifts at 18 kHz (Fig. 4 A and B), the corresponding tonotopic region (45 to 55% from the base) showed minimal degeneration (Fig. 4 E–H) (32, 33), suggesting that additional factors contribute to progressive hearing loss.

Further examination at 16 wk revealed more extensive functional deficits compared to 3 wk (SI Appendix, Fig. S10). At 3 wk, auditory nerve function was reduced only at high frequencies, with reduced ABR wave I amplitudes specifically at 30 kHz (Fig. 3). However, by 16 wk, both systemic and HC-specific Shank2 mutants showed significantly reduced ABR wave I amplitudes at all frequencies examined (6, 18, 24 kHz) (SI Appendix, Fig. S10 A and B). Similarly, while initial DPOAE deficits were restricted to high frequencies at 3 wk (Fig. 3), 16-wk-old mutants showed a much greater reduction in DPOAE amplitudes, particularly in the mid-to-high frequency range (SI Appendix, Fig. S10 D and E). SGN-specific Shank2 mutants retained normal ABR and DPOAE amplitude at 16 wk (SI Appendix, Fig. S10 C and F). These results suggest that disturbed bundle morphology initially impairs high-frequency hearing through defective OHC amplification but leads to a progressive deterioration of both mechanical and neural auditory function across frequencies.

SHANK2 Regulates Bundle Architecture Independently of known Morphogenesis Regulators.

The lateral proteins Gαi and GPSM2 are essential regulators of hair bundle morphogenesis (4, 5, 21). Their loss leads to fragmented bundles similar to Shank2 mutants (4, 5) but also causes medial proteins such as aPKCζ to expand into the lateral compartment (4, 5). Since SHANK2 interacts with aPKC to establish epithelial polarity in intestinal cells (34), we investigated whether SHANK2 regulates the localization of known regulators of bundle formation.

In Shank2^–/– HCs lacking medial SHANK2 (SI Appendix, Fig. S11 A–D, asterisk), the lateral proteins Gαi and GPSM2 maintained their localization in the bare zone (SI Appendix, Fig. S11 A and B, arrows). Similarly, the medial proteins aPKCζ and PARD6B remained correctly restricted to the medial surface (SI Appendix, Fig. S11 C and D, arrows). Conversely, in Gpsm2^–/– mutants, which showed reduced Gαi (SI Appendix, Fig. S11E, asterisk) and expanded aPKCζ expression into the lateral bare zone (SI Appendix, Fig. S11E, yellow arrow), SHANK2 maintained its medial localization (SI Appendix, Fig. S11F, arrowheads). These results suggest that SHANK2 shapes bundle architecture by a mechanism independent of these known regulators of bundle morphogenesis.

Bundle Defects in Shank2 Mutants Are Not Caused by Cuticular Plate or Apical Junction Disruption.

Since the loss of SHANK2 did not affect the localization of known lateral or medial regulators, we next examined whether other structural components of the hair cell apical surface, specifically the cuticular plate and tight junctions, might contribute to the bundle abnormalities observed in Shank2 mutants (35, 36). Immunostaining of β-spectrin and LMO7, two established markers of the cuticular plate (35, 37), showed continuous and well-defined labeling beneath the stereociliary bundles in both control and Shank2^–/– cochleae, irrespective of bundle morphology (SI Appendix, Fig. S12), indicating that the gross structure of the cuticular plate was preserved.

We also assessed apical junctional organization by ZO-1 immunostaining. At P1, when bundle defects were already evident, Shank2^–/– HCs displayed circumferential ZO-1 labeling comparable to that of controls (SI Appendix, Fig. S13A). By 3 wk, most HCs continued to exhibit intact ZO-1 staining, although occasional contour irregularities were observed in OHCs with severely deformed bundles (SI Appendix, Fig. S13B). Together, these results suggest that the bundle abnormalities in Shank2 mutants are not attributable to disruptions in either cuticular plate architecture or apical junction integrity.

Identification of RAP1 as a Potential Regulator of SHANK2 Localization.

Since SHANK2 functions independently of known apical regulators of bundle morphogenesis, we employed yeast two-hybrid screening to identify potential SHANK2-interacting partners in developing HCs that might explain its role in bundle morphogenesis. The SHANK2 protein contains multiple domains and is expressed as three major isoforms (SHANK2A, SHANK2B, and SHANK2E) (SI Appendix, Fig. S14A) (38, 39). Using qPCR analysis, we found that Shank2e, which contains the N-terminal SPN-ANK domain, is the predominant isoform in the organ of Corti (SI Appendix, Fig. S14B). This SPN-ANK domain is known to regulate protein localization at cell–cell contacts and tight junction formation in renal epithelial cells (34). Using the SPN-ANK domain as a bait, we screened inner ear cDNA libraries from E16.5 and P2-P6 mice—ages corresponding to hair bundle formation (Fig. 1). RAS-related protein 1B (RAP1B) emerged as a potential SHANK2-interacting protein with the highest confidence score from both cDNA libraries (SI Appendix, Tables S1 and S2). Notably, previous studies have shown that RAP1–SHANK interaction through the SPN-ANK domain regulates the morphology and stability of actin-based cytoskeletal structures, and RAP1 loss affects the localization of apicobasal polarity proteins and tight junctions (34, 40).

To investigate the role of RAP1 in bundle formation, we treated E14.5 cochlear explants with the RAP1 inhibitor GGTI298 for six days (34). We observed that the treated explants exhibited severe bundle fragmentation similar to Shank2 mutants (SI Appendix, Fig. S15A). Importantly, SHANK2 was lost from the medial surface of both IHCs and OHCs in GGTI298-treated explants (SI Appendix, Fig. S15B, asterisk). In contrast, other medial (aPKCζ and PARD6B) and lateral (GPSM2 and Gαi) proteins maintained their localization (SI Appendix, Fig. S15 C–F, arrows). These results suggest that RAP1 specifically regulates SHANK2 localization and thus bundle morphogenesis.

RAP1 Is Required for SHANK2 Localization in Developing Hair Cells.

GGTI298 has been shown to inhibit not only RAP1 but also other small G-proteins such as RhoA (34, 41) and explant cultures limited the analysis of the functional effects of RAP1 inhibition. To investigate the specific role of RAP1 in vivo, we generated two Rap1 cKO models targeting both Rap1a and Rap1b due to their functional redundancy (42): an otic-specific Rap1 cKO (Slc26a9^Cre; Rap1a^lox/lox; Rap1b^lox/lox) deleting both genes from E9.5 otocyst (43) and a HC-specific Rap1 cKO (Gfi1^Cre; Rap1a^lox/lox; Rap1b^lox/lox) that ablates them in differentiating hair cells from E16.5 (44).

In otic-specific Rap1 mutants, the medical localization of SHANK2 (Fig. 5A, arrowhead) was completely absent in both HC types at E16.5 and P1 (Fig. 5 A and B, asterisks). Developing hair bundles appeared severely fragmented with disorganized organ of Corti patterning (Fig. 5 A and B). In contrast, HC-specific Rap1 mutants showed initial medial SHANK2 localization at E16.5 but lost it by P1 (Fig. 5 A and B, yellow arrowhead and asterisk), probably because Gfi1^Cre-mediated Rap1 ablation occurs after initial SHANK2 localization. In both mutants, mislocalized SHANK2 proteins formed aggregates at P1 (Fig. 5B), indicating preserved SHANK2 expression despite its mislocalization. Consistent with explant findings (SI Appendix, Fig. S15), both Rap1 mutants maintained normal medial aPKCζ and lateral Gαi localization (Fig. 5 C and D, arrows). In addition, similar to Shank2 mutants (SI Appendix, Fig. S3), kinocilium formation and positioning, assessed by ARL13B staining, remained unaltered in HC-specific Rap1 mutants, indicating that RAP1 loss does not impair kinocilium morphogenesis and migration in developing HCs (SI Appendix, Fig. S16 A and B). These results demonstrate that RAP1 specifically regulates the medial localization of SHANK2, controlling both initial targeting and subsequent maintenance.

Fig. 5.

RAP1 regulates SHANK2 localization and hair bundle morphology. (A and B) SHANK2 localization in Rap1 mutant cochlea. At E16.5, SHANK2 localizes to the medial apical surface in control and HC-specific (Gfi1^Cre; Rap1a/b^lox/lox) mutants (A, arrowheads) but is absent in otic-specific (Slc26a9^Cre; Rap1a/b^lox/lox) mutants (A, asterisk). By P1, SHANK2 remains medial in controls (B, arrowhead) but is lost in both mutants (B, asterisks). (C and D) Both Rap1 mutants maintain medial aPKCζ (C, arrows) and lateral Gαi (D, arrows) localization. (E) SEM images of basal hair bundles at 3 wk. Controls show V-shaped OHC bundles, while otic-specific Rap1 mutants exhibit severe fragmentation and HC-specific mutants display rounded U-shaped bundles. (F–H) Auditory function at 3 wk. Otic-specific mutants show threshold shifts across all frequencies (F), while HC-specific mutants show shift only at high frequencies (G) with progressive reduction in DPOAE amplitudes toward higher frequencies (H). [Scale bar, A–D (low magnification), 5 μm; A–D (high magnification) 2 μm; E (low magnification), 10 μm; E (high magnification), 1 μm.] Data are presented as means ± SD. Statistical comparisons were made using two-way ANOVA with Bonferroni corrections for multiple comparisons (n.s., nonsignificant, *P < 0.05, **P < 0.01, and ***P < 0.001).

Loss of RAP1 in Hair Cells Disrupts Hair Bundle Architecture and High-Frequency Hearing.

We next examined how SHANK2 mislocalization in Rap1 cKO mutants affects hair bundle morphology. Otic-specific Rap1 mutants showed severe disruption of hair bundles at 3 wk, with fragmented morphology and loss of the characteristic linear or V-shaped architecture (Fig. 5E and SI Appendix, Fig. S16C). In addition, these mutants exhibited a disorganized organ of Corti and complete OHC degeneration in the apex and only two rows in the mid-to-basal regions (Fig. 5E and SI Appendix, Fig. S16C). In contrast, HC-specific Rap1 mutants showed milder defects, mainly showing rounded U-shaped rather than V-shaped OHC bundles across the cochlear duct (Fig. 5E and SI Appendix, Fig. S16C). To quantify this shape difference, we measured the ratio between the triangular area formed by connecting the vertex to bundle ends and the actual bundle area (SI Appendix, Fig. S16D). While control OHCs maintained a V-shaped architecture with ratios close to 1, HC-specific Rap1 mutants showed ratios greater than 1, reflecting their U-shaped morphology (SI Appendix, Fig. S16D).

We next assessed auditory function in both Rap1 mutants at 3 wk, matching the timepoint used for Shank2 mutants (Fig. 3). Otic-specific Rap1 mutants exhibited profound hearing loss across all frequencies in both ABR and DPOAE measurements (Fig. 5F). HC-specific Rap1 mutants, characterized by U-shaped OHC bundles, showed hearing deficits specifically at high frequencies in both ABR and DPOAE (Fig. 5G), similar to Shank2 mutants (Fig. 3). DPOAE amplitudes decreased progressively toward higher frequencies (Fig. 5H). ABR wave I amplitude was significantly reduced at 30 kHz but not at 18 and 24 kHz (SI Appendix, Fig. S16E). We also observed reduced wave I amplitudes at 6 kHz in response to high stimulus intensities, suggesting an additional role of RAP1 in the apical cochlea. These results demonstrate that proper bundle geometry regulated by the RAP1–SHANK2 pathway is essential for OHC amplification, especially at high frequencies.

Discussion

Mammals have evolved specialized auditory hair cells with distinctive bundle architectures, including the V-shaped geometry of OHC bundles. Our study identifies a molecular pathway that shapes this architecture and demonstrates its significance for auditory function. SHANK2, acting downstream of the small GTPase RAP1, localizes to the medial apical surface of developing HCs and contributes to proper bundle formation. Using cell type–specific genetic ablation, we show that disruption of SHANK2-dependent bundle architecture leads to impaired high-frequency hearing through reduced OHC amplification. Over time, these architectural defects are associated with progressive bundle degeneration and worsening auditory thresholds. These findings provide fundamental insights into both the molecular control of hair bundle morphogenesis and the functional importance of mammalian-specific bundle architecture. A comparative summary of protein localization and hair bundle phenotypes across multiple mutants is provided in Fig. 6.

Fig. 6.

Schematic summary of protein localization and hair bundle architecture in each mutant. Wild-type cochlear hair cells show normal protein localization and V-shaped OHC bundles. In Gpsm2^–/– mice, SHANK2 remains medial, but lateral Gαi is reduced and medial aPKCζ expands, resulting in short, numerous stereocilia rows with disrupted staircase pattern. Shank2^–/– mice lack medial SHANK2 but maintain normal medial (aPKCζ and PARD6B) and lateral (Gαi and GPSM2) proteins, showing disrupted V-shaped architecture but preserved staircase pattern. Both Rap1 mutants maintain medial/lateral proteins but lack SHANK2, resulting in bundle defects: Otic-specific (Slc26a9^Cre; Rap1a/b^lox/lox) mutants never localize SHANK2 medially and show severe bundle defects, while HC-specific (Gfi1^Cre; Rap1a/b^lox/lox) mutants initially localize but fail to maintain SHANK2, resulting in rounded U-shaped bundles.

The mechanism by which SHANK2 contributes to hair bundle morphogenesis appears distinct from that of established lateral regulators such as Gαi and GPSM2. These proteins control bundle morphogenesis through kinocilium migration and stereocilium elongation (4, 5, 21, 22), and their loss result in multiple defects including disrupted bundle morphology, random kinocilium positioning, and loss of staircase arrangement (21, 22). In contrast, SHANK2 localizes to the medial apical surface, complementary to the distribution of Gαi and GPSM2 (Fig. 1 and SI Appendix, Fig. S2), and specifically regulates the formation of bundle architecture while preserving staircase arrangement and kinocilium positioning (Figs. 2 and 3 and SI Appendix, Figs. S3 and S4). The preserved localization of both lateral (Gαi, GPSM2) and medial (aPKCζ, PARD6B) proteins in Shank2 mutants—and the reciprocal retention of SHANK2 in Gpsm2 mutants—suggest that SHANK2 acts independently of these known regulators (SI Appendix, Fig. S11). Furthermore, bundle abnormalities in Shank2 mutants are not attributable to disruptions of other apical structures. The cuticular plate and apical junctions remain intact (SI Appendix, Figs. S12 and S13), and kinocilium formation and positioning are unaffected in both Shank2 and Rap1 mutants (SI Appendix, Figs. S3 and S16). These findings make it unlikely that the observed bundle defects arise as secondary consequences of apical surface structural defects. SHANK proteins have shown to regulate actin dynamics across diverse cellular contexts. In neurons, SHANK2’s proline-rich domain recruits cortactin to promote actin polymerization at developing dendritic spines (45, 46), while SHANK3 binds directly to actin through its SPN-ANK domain to regulate filopodia in osteosarcoma cells and dendritic spine morphology (40). These conserved actin-related functions raise the possibility that SHANK2 shapes hair bundle architecture by modulating actin dynamics at the medial surface of developing hair cells—a hypothesis that warrants further investigation.

Through screening of embryonic and neonatal inner ear cDNA libraries, we identified RAP1 as a potential SHANK2-interacting partner (SI Appendix, Tables S1 and S2). RAP1 is known to control the localization of junctional proteins and to promote apicobasal polarity in various epithelial systems (47, 48). In renal epithelial cells, RAP1 recruits SHANK2 via the SPN-ANK domain to sites of tight junction assembly (34). Consistent with this, pharmacological inhibition of RAP1 in cochlear explants disrupted SHANK2 localization at the medial apical surface without affecting other medial (aPKCζ, PARD6B) or lateral (Gαi, GPSM2) proteins, resulting in bundle disorganization that phenocopied Shank2 mutants (SI Appendix, Fig. S15). In vivo, both otic-specific and HC-specific Rap1 mutants showed loss of SHANK2 at the medial surface (Fig. 5), confirming that RAP1 is required for the proper localization and function of SHANK2 during bundle development.

Importantly, RAP1 appears to have broader roles in cochlear development beyond SHANK2 regulation. Otic-specific Rap1 mutants exhibited disorganization of the organ of Corti and OHC degeneration by 3 wk, resulting in profound hearing loss across all frequencies (Fig. 5 and SI Appendix, Fig. S16). HC-specific Rap1 mutants showed a reduced ABR wave I amplitudes at 6 kHz without corresponding changes in DPOAE thresholds, suggesting possible defects in auditory nerve function. These additional phenotypes may reflect RAP1’s known roles in cytoskeletal regulation such as cofilin dephosphorylation (49, 50) and activity-dependent synaptic plasticity (51, 52). Further studies will be needed to define RAP1’s full contribution to cochlear patterning, hair cell maintenance, and afferent function.

The V-shaped hair bundle of OHCs is a distinctive feature of the mammalian cochlea, contrasting with the linear or crescent-shaped bundles observed in nonmammalian species (53). Mammalian OHCs uniquely function as cochlear amplifiers, enhancing hearing sensitivity and dynamic range (2, 54). Since mammals generally detect higher frequency sounds (55), their hair bundles experience greater mechanical stress and hydrodynamic drag, particularly in the basal cochlea (56). The V-shaped geometry is thought to minimize fluid flow resistance during high-frequency deflections (56–58), a hypothesis supported by computational modeling, which predicts an optimal V angle of approximately 100°—consistent with bundle geometry observed in the cochlear base (57, 59). However, testing the functional significance of this geometry has been hindered by the lack of appropriate animal models. Most mouse models with disrupted bundle morphology also exhibit confounding defects such as HC misorientation, loss of staircase pattern, cochlear shortening, or extra HC rows—as seen in mutants of PCP components, lateral polarity regulators, and ciliary proteins (4, 5, 8–10, 12–17, 21, 22). Other models, such as Bbs8 and Alms1 mutants, show only mild and localized bundle disorganization in a limited subset of OHCs (~20%) without measurable auditory deficits (16, 20).

In contrast, Shank2 mutants represent a rare model for examining how disrupted bundle architecture alone affects auditory function, without accompanying defects in orientation, staircase organization, or cochlear patterning. Loss of SHANK2, or disruption of its apical localization in Rap1 mutants, leads to widespread bundle disorganization and selective impairment of high-frequency amplification (Figs. 3 and 5 and SI Appendix, Fig. S16). While SHANK2 localizes opposite to Gαi in both auditory and vestibular HCs, its loss affects bundle shape only in auditory HCs (SI Appendix, Fig. S2). Given that the vestibular system is considered more ancestral than the cochlea (60), SHANK2-dependent shaping of OHC bundle geometry may represent a mammalian adaptation for high-frequency hearing. Supporting this idea, subterranean naked mole-rats, which hear in a much lower frequency range (125 Hz to 8 kHz) than mice and rats, possess U-shaped rather than V-shaped OHC bundles (61). Although their limited high-frequency hearing has been attributed to multiple factors, including impaired OHC function, HC orientation defects, and amino acid substitutions (61), these observations further suggest a function link between bundle geometry and high-frequency hearing capability.

Our longitudinal assessment indicates that proper bundle architecture is critical not only for high-frequency amplification but also maintaining long-term bundle integrity and auditory function. Shank2 mutants exhibited progressive hearing loss and bundle degeneration with age (Fig. 4 and SI Appendix, Figs. S9 and S10). Although additional roles for SHANK2 in the mature cochlea cannot be excluded, its reduced expression in neonatal HCs (SI Appendix, Fig. S1) and near absence in mature HCs (27, 28, 31) (SI Appendix, Fig. S6) suggest that early architectural defects are the likely driver of delayed auditory decline. These findings expand the importance of bundle architecture beyond its immediate role in cochlear amplification to its contribution to structural maintenance over time. At 16 wk, Shank2 mutants showed elevated ABR and DPOAE thresholds in the mid-to-high frequency range (Fig. 4 and SI Appendix, Fig. S9). Unlike younger animals, these aged mice also exhibited reduced ABR wave I amplitudes, suggesting additional deterioration of IHC function (Fig. 3 and SI Appendix, Figs. S7 and S10). The inability of mammalian HCs to regenerate underscores the significance of developmental bundle architecture in preserving auditory capacity over the lifespan.

SHANK2 organizes multiprotein complexes at postsynaptic densities in the central nervous system and has been implicated in ASD (23). However, our results reveal a distinct role for SHANK2 in the peripheral system. HC-specific Shank2 mutants recapitulated the auditory deficits observed in systemic knockouts (Fig. 3 and SI Appendix, Fig. S7), whereas SGN-specific mutants exhibited normal auditory function (SI Appendix, Figs. S8–S10). This is consistent with previous findings in Shank1 mutants, which showed no major auditory or postsynaptic defects in the cochlea (31), and suggests that SHANK proteins have cell type–specific roles in the inner ear distinct from their central synaptic functions. Although SHANK2 mutations are not directly linked to peripheral hearing loss in humans, some individuals with ASD exhibit hearing abnormalities that may exacerbate sensory and communicative challenges (62–64). While these deficits are typically attributed to central processing dysfunction, altered ABR and otoacoustic emissions have been reported in some cases (65–67), consistent with the peripheral impairments observed in Shank2 mutant mice. These findings raise the possibility that peripheral auditory abnormalities may contribute to sensory phenotypes in a subset of individuals with SHANK2-related neurodevelopmental disorders.

Materials and Methods

The mice used in this study were generated as previously described (4, 43, 44, 68–72). Immunofluorescence staining, in situ hybridization, scanning electron microscopy, ABR, and DPOAE measurements were performed as previously described (59, 73). Expression analysis of Shank2 in embryonic and mature cochlea was performed using published RNA-sequencing datasets (27–30). Yeast two-hybrid screening was performed using Shank2 (amino acids 1 to 395) as bait to screen inner ear cDNA libraries from E16.5 and P2-P6 mice. A more detailed account of the materials and methods used in this study appears in the SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.