Introduction

Developing sensory hair cells (HCs) in the inner ear undergo a complex polarization process to detect and interpret mechanical stimuli, including sound. Each mature HC is able to detect stimuli in a directional manner by developing an asymmetric brush of actin-based membrane protrusions, or stereocilia: the hair bundle. Neighboring HCs also adopt concerted orientations to align their hair bundles, a property known as planar cell polarity (PCP). One subclass of heterotrimeric guanine nucleotide-binding (G) protein was intimately associated with different levels of mouse HC polarization: inhibitory G alpha subunits (GNAI1, GNAI2, GNAI3 and GNAO; collectively GNAI or Gαi). However, several roles proposed for GNAI proteins have not been validated physiologically, and the individual contribution of each GNAI remains uncertain.

HC polarization along the epithelial plane starts with the off-center migration of the basal body and its associated primary cilium, the kinocilium (Denman-Johnson & Forge, 1999; Mbiene & Sans, 1986; Tilney et al, 1992). At that stage, Regulator of G protein Signaling 12 (RGS12) is required for GNAI and the G protein signaling modulator 2 (GPSM2) scaffold to form a polarized complex at the apical membrane on the side of the off-center basal body (Akturk et al, 2022; Ezan et al, 2013; Tarchini et al, 2013). GPSM2-GNAI is best known as a highly conserved protein complex orienting the mitotic spindle during progenitor divisions (Du et al, 2001; Schaefer et al, 2001; Schaefer et al, 2000; Woodard et al, 2010). In post-mitotic HCs, GPSM2-GNAI first excludes microvilli and microvilli-derived stereocilia from the portion of the HC surface where it resides, the expanding bare zone (Figure 1A). Bare zone expansion then pushes back the basal body/kinocilium from a more eccentric position near the lateral junction to a less eccentric position at the vertex of the forming hair bundle. Later, GPSM2-GNAI becomes enriched at the distal tip of row 1 stereocilia abutting the bare zone (Tarchini et al, 2016). In these stereocilia, GPSM2-GNAI is a module of the elongation complex that also comprises MYO15A, WHRN and EPS8 (Mauriac et al, 2017; Tadenev et al, 2019). GPSM2-GNAI is required at row 1 tips for boosting enrichment of other elongation complex partners, and presumably actin incorporation, compared to further stereocilia rows. GPSM2-GNAI thus confers row 1 its tallest identity and the hair bundle its asymmetric graded-height morphology.

Figure 1.

Pertussis toxin (ptx) has been extensively used as a tool to ADP-ribosylate the GNAI subunit and dissociate heterotrimeric G protein complexes from GPCRs to inactivate downstream signaling (Locht et al, 2011). In vivo expression of ptx catalytic subunit (ptxA) prevents normal enrichment and polarization of GPSM2-GNAI in developing HCs (Tadenev et al, 2019; Tarchini et al, 2013), suggesting that ADP-ribosylation directly or indirectly inhibits GPSM2-GNAI function as well. Ptx recapitulates severe stereocilia stunting and immature-looking hair bundles observed when GPSM2 or both GNAI2 and GNAI3 are inactivated (Beer-Hammer et al, 2018; Mauriac et al, 2017; Tadenev et al, 2019; Tarchini et al, 2016). In contrast, stereocilia height is more variably reduced in Gnai3 single mutants, with defects more severe at the cochlear base (Beer-Hammer et al, 2018; Mauriac et al, 2017). This can explain why hearing loss is more severe at high frequencies in Gnai3 mutants, but profound at all frequencies in a ptxA model and in mutants lacking GPSM2 or both GNAI2 and GNAI3 (Beer-Hammer et al, 2018; Mauriac et al, 2017; Tarchini et al, 2016). Surprisingly, before affecting hair bundle differentiation at postnatal stages, ptx also causes two distinct defects in HC polarization at embryonic stages.

First, one study reported that a high dose of ptx in cultured explants of the developing cochlea resulted in a low proportion of symmetrical HCs with a central kinocilium surrounded by a rounded hair bundle (Figure 1B) (Ezan et al, 2013). Because GPSM2-GNAI recruits partners to pull on astral microtubules during mitotic spindle orientation, the authors proposed that GPSM2-GNAI similarly orients the migration of the basal body when post-mitotic HCs break planar symmetry. However, this hypothesis has not been validated in vivo to date. Studies where Gpsm2 (Bhonker et al, 2016; Ezan et al, 2013; Tarchini et al, 2013), Gnai3 (Beer-Hammer et al, 2018; Ezan et al, 2013) or Gnai2; Gnai3 (Beer-Hammer et al, 2018) were inactivated did not report symmetrical HCs. In addition, mouse strains expressing ptxA in vivo also did not produce symmetrical cochlear HCs (Tarchini et al, 2013; Tarchini et al, 2016).

Second, ptx experiments induced striking HC misorientation. In the cochlea, misorientation manifested as a 180° inversion of outer HCs in the first and second row (OHC1-2) whereas inner HCs (IHCs) and OHC3 were much less affected (Figure 1B) (Ezan et al, 2013; Kindt et al, 2021; Tarchini et al, 2013). In vestibular organs, ptxA expression abrogated the line of polarity reversal and thus mirror-image HC organization in the maculae, the utricle and saccule (Jiang et al, 2017; Kindt et al, 2021). Similar defects were observed by inactivating two endogenous mouse proteins, the transcription factor EMX2 in the maculae (Jiang et al, 2017) and the orphan GPCR GPR156 in the cochlea and the maculae (Kindt et al, 2021). These recent studies uncovered an EMX2>GPR156>GNAI signaling cascade that regionally reverses HC orientation in macular organs to produce mirror-image HC organization. GNAI signals downstream of GPR156 to apparently reverse the orientation of basal body migration in Emx2-positive compared to Emx2-negative HCs (Tona & Wu, 2020) (reviewed in (Tarchini, 2021)). While ptx-induced HC misorientation is thus a physiologically relevant phenotype, HC misorientation was not reported in single Gnai3 or double Gnai2; Gnai3 mutants (Beer-Hammer et al, 2018; Mauriac et al, 2017).

In summary, the importance of the GPSM2-GNAI complex for hair bundle development is well-established, but multiple discrepancies cast a doubt on whether GNAI proteins also assume earlier polarization roles. Specific questions include whether GNAI proteins participate in the mechanism that pushes the basal body away from the cell center, and in the distinct EMX2>GPR156 mechanism that makes a binary decision on the direction of this push (Figure 1). Furthermore, it remains unclear whether GNAI2 and GNAI3 adopt similar or distinct distributions in HCs, and a potential role for GNAI1 or GNAO has not been addressed.

In this study, we embarked on a systematic analysis of single and combined Gnai1, Gnai2, Gnai3 and Gnao mutants to solve the actual role(s) of GNAI proteins during HC differentiation. Our results show that GNAI3 is the only GNAI/O protein required for normal hair bundle morphogenesis and normal auditory brainstem response (ABR) thresholds. In absence of GNAI3, GNAI2 can fully compensate at embryonic stages but is not enriched with GPSM2 long enough to ensure normal hair bundle morphogenesis at postnatal stages. We also directly demonstrate that GNAI proteins have two polarization roles independent of GPSM2 during embryogenesis. In sum, GNAI function is instrumental for HCs to a) break planar symmetry, b) adopt a proper binary orientation along the PCP axis downstream of EMX2 and GPR156, and c) elongate and organize stereocilia into a functional hair bundle with GPSM2.

Results

A near-comprehensive collection of Gnai/o mouse mutants

In order to interrogate the individual and combined roles of all inhibitory G proteins during HC differentiation, we obtained or generated mouse strains to build a collection of single and double Gnai/o mutants. Single Gnai1 and Gnai3 mutants were derived from the Gnai1^tm1Lbi; Gnai3^tm1Lbi double mutant strain (hereafter Gnai1^neo; Gnai3^neo) (Jiang et al, 2002) by segregating individual mutations upon breeding (see Methods and Supp. Table 1 for details). We generated a new constitutive Gnai2 mutant strain (Gnai2^del) carrying a deletion of exons 2 to 4 (Figure 2 Supp. 1A; see Methods). Finally, we obtained and derived two Gnao1 mutant strains: a constitutive inactivation where a neomycin cassette disrupts exon 6 (Gnao1^neo) (Jiang et al, 1998) and the consortium alleles Gnao1^{tm1c(EUCOMM)Hmgu} (conditional inactivation; hereafter Gnao1^flox). As simultaneous constitutive loss of GNAI1 and GNAI2 was reported as viable (Plummer et al, 2012), we established a Gnai1^neo; Gnai2^del double mutant strain in addition to Gnai1^neo; Gnai3^neo (Jiang et al, 2002). In contrast, double inactivation of Gnai2; Gnai3 is lethal around Embryonic (E) day 10.5, before HCs are born (Gohla et al, 2007). Consequently, we generated a new Gnai3^flox strain by flanking exons 2 and 3 with loxP sites (Figure 2 Supp.1B; see Methods). We then generated conditional FoxG1-Cre; Gnai2^del; Gnai3^floxdouble mutants where Gnai3 inactivation occurs as early as E8.5 in the otic vesicle (Hebert & McConnell, 2000). Investigating all Gnai/o strains in the same genetic background was unrealistic for feasibility and lethality reasons. We reasoned that the key steps in HC apical differentiation we study are probably hardwired and less likely to be influenced by genetic heterogeneity compared to susceptibility to disease, for example.

Figure 2.

As Pertussis toxin produced two of the three types of defects associated with GNAI/O function (Figure 1B), Gnai/o strains above needed to be compared to a strain expressing ptxA in HCs. We used our R26^LSL-myc:ptxA strain (hereafter LSL-myc:ptxA) expressing N-terminal myc-tagged ptxA after Cre-recombination (Tarchini et al, 2016). Because a related R26^{LSL-ptxAa:myc} strain carrying a C-terminal myc tag (Regard et al, 2007) caused much milder HC misorientation than LSL-myc:ptxA in the vestibular system (Jiang et al, 2017; Kindt et al, 2021), we questioned whether the myc tag could weaken toxin activity even perhaps when located N-terminal. Consequently, we generated a new strain, R26^DIO-ptxA (hereafter DIO-ptxA), where untagged ptxA is flanked by double-inverted lox sites and flipped from the non-coding to the coding strand upon Cre recombination (Figure 2 Supp. 1C-D; see Methods) (Schnutgen et al, 2003). In DIO-ptxA, ptxA expression is driven by a strong artificial CAG promoter, and not by the endogenous R26 promoter as in previous strains (Regard et al, 2007; Tarchini et al, 2016). We bred LSL-myc:ptxA and DIO-ptxA either with FoxG1-Cre (Hebert & McConnell, 2000), or with Atoh1-Cre (Matei et al, 2005) to limit GNAI/O inhibition to post-mitotic HCs. Supplementary Table 1 summarizes the strains used in this study, their origin, genetic background and viability.

Only GNAI3 is required for hair bundle morphogenesis yet other GNAI proteins participate

To investigate the role of individual GNAI/O proteins in HC development, we imaged hair bundles in 3 week-old animals at mid-cochlear position using scanning electron microscopy. In Gnai3^neo mutants, some OHC hair bundles appeared truncated, and IHCs displayed variably shortened row 1 stereocilia as well as supernumerary stereocilia rows (Figure 2A-C), as previously described (Mauriac et al, 2017). Single inactivation of Gnai1 or Gnai2, and double Gnai1; Gnai2 inactivation did not produce overt defects, yet quantifications revealed a subtle reduction in row 1 height in all cases (Figure 2A, C). Accordingly, Gnai1^neo; Gnai3^neo double mutants appeared more severely affected than single Gnai3^neo mutants: IHC row 1 height was further reduced (from ∼32% to ∼47% reduced compared to littermate controls; Figure 2C), and the number and width of row 1 stereocilia were respectively only significantly increased or decreased in Gnai1^neo; Gnai3^neo mutants (Figure 2D-E). In contrast, a constitutive (Gnao1^neo) or conditional (Atoh1-Cre; Gnao1^flox) inactivation of GNAO did not produce overt apical HC defects, and conditional inactivation of GNAO did not enhance defects in the Gnai1^neo; Gnai3^neo background (Figure 2 Supp. 1E-F).

As reported previously, myc:ptxA expression in post-mitotic HCs did not only stunt stereocilia and produce extra rows, as seen in Gnai3^neoand Gnai1^neo;Gnai3^neo mutants (Figure 2A-F), but also inverted OHC1-2 (Figure 2A) (Kindt et al, 2021; Tarchini et al, 2013; Tarchini et al, 2016). This defect was proposed to reflect an earlier role for GNAI proteins downstream of the GPR156 receptor to define the binary orientation of HCs along the PCP axis (Kindt et al, 2021). Despite breeding for over two years, we could only obtain one single Gnai2^neo; Gnai3^flox adult double mutant due to postnatal lethality (FoxG1-Cre; Gnai2^del/del; Gnai3^flox/flox; see Supp. Table 1). Remarkably, this specimen recapitulated stereocilia stunting and extra rows observed in Gnai3^neo, Gnai1^neo;Gnai3^neo and LSL-myc:ptxA models (Figure 2A-F), but apparently also an inversion of OHC1-2 only observed to date in LSL-myc:ptxA and Gpr156 mutants (Figure 2A; see also below, Figure 8 and Figure 8 Supp. 1) (Kindt et al, 2021). IHC quantifications uncovered a Gnai3^neo < Gnai1^neo; Gnai3^neo < LSL-myc:ptxA < FoxG1-Cre; Gnai2^del; Gnai3^flox allelic series along which a) defects increased in severity, as manifested by increased variability and decreased averages (row 1 height and width; Figure 2C-D), and b) new defects appeared (excess row 1 stereocilia in Gnai1^neo; Gnai3^neo and more affected models; Figure 2E). In general, this phenotypic series moves towards increasingly immature-looking hair bundles, and increasingly mimics severe hair bundle defects in Gpsm2 mutants (Mauriac et al, 2017; Tadenev et al, 2019; Tarchini et al, 2016).

To quantify and compare hair bundle truncations in OHCs, we measured the length of half-bundle ‘wings’ on each side of the central vertex and plotted paired left and right values for each OHC (Figure 2G; Figure 2 Supp. 1G). We tested length variance using mutant strains where hair bundles retained a recognizable vertex (thus excluding LSL-myc:ptxA and FoxG1-; Gnai2^del; Gnai3^flox double mutants). In Gnai1^neo, Gnai2^deland Gnai1^neo; Gnai2^delmutants, the two wings of each hair bundle had similar lengths and variance was similar to control littermates. In contrast, variable truncation of one wing in Gnai3^neoand Gnai1^neo; Gnai3^neomutants resulted in significantly higher length variance than controls (Figure 2G; Figure 2 Supp. 1G). In both Gnai3^neo and Gnai1^neo; Gnai3^neo mutants, 49% of hair bundles had wings of more different lengths than the worst OHC outlier in littermate controls (see Source Data file).

In conclusion, GNAI proteins are not equally involved in hair bundle morphogenesis, with GNAI3 playing a particularly prominent role. GNAI2 makes a clear contribution since stereocilia defects increase in severity when GNAI loss extends from GNAI3 to both GNAI2 and GNAI3. These results confirm previous conclusions (Beer-Hammer et al, 2018). In addition, we largely rule out that GNAO is involved in apical HC differentiation, but uncover a possible contribution from GNAI1. Of note, as Gnai3^neois in mixed (129S1/SvImJ; C57BL/6J) whereas Gnai1^neo; Gnai3^neo is in pure (129S1/SvImJ) genetic background (Supp. Table 1), we cannot rule out that the more severe defects in Gnai1^neo; Gnai3^neo double mutants originate from genetic modifiers in 129S1/SvImJ. Finally, the one adult specimen obtained for FoxG1-Cre; Gnai2^del/del; Gnai3^flox/flox suggests that GNAI2 can fully compensate for the combined loss of GNAI1 and GNAI3 in Gnai1^neo; Gnai3^neo double mutants at embryonic stages, and secure proper OHC1-2 orientation in the EMX2>GPR156>GNAI signaling cascade. This important point is verified and expanded below where neonate HCs are analyzed.

Only GNAI3 is required for normal auditory brainstem thresholds yet other GNAI proteins participate

To pair morphological defects with auditory function, we next tested Auditory Brainstem Response (ABR) in mutants and control littermates at 3 to 4 weeks of age. Anesthetized animals were presented with pure tone stimuli of decreasing sound pressure intensity (dB SPL) at 8, 16, 32 and 40 kHz and ABRs were recorded with a subcutaneous probe (see Methods). Mirroring their overtly normal apical HC morphology, Gnai1, Gnai2 and Gnai1; Gnai2 mutants showed thresholds comparable to littermate controls at all frequencies (Figure 3A-C). Similarly, constitutive (Gnao1^neo) or conditional (Atoh1-Cre; Gnao1^flox) inactivation of GNAO did not alter ABR thresholds (Figure 3 Supp. 1A-B). In contrast, Gnai3^neo mutants were profoundly deaf at 32 and 40kHZ and displayed significantly elevated thresholds at 8 and 16 kHz compared to littermate controls (Figure 3D). Gnai1^neo; Gnai3^neo double mutants shared a similar ABR profile as Gnai3^neo single mutants, with apparently higher thresholds at 8 and 16 kHz although these two strains were not compared as littermates (Figure 3D). Littermate controls for Gnai1^neo; Gnai3^neodouble mutants were homozygote for Gnai1^neo (Gnai1^neo/neo; Gnai3^neo/+) and had apparently higher thresholds than littermate controls for Gnai3^neo mutants (Gnai3^neo/+) whereas loss of GNAI1 did not impact auditory thresholds on its own (Figure 3A). These results match more severe hair bundle defects when GNAI1 is inactivated along with GNAI3 (Figure 2C-F), and here again, could reflect either a limited role for GNAI1 in hearing or differences in genetic background.

Figure 3.

In summary, we confirm that GNAI3, but not GNAI2, is essential for proper auditory thresholds, as previously proposed (Beer-Hammer et al, 2018). We raise the possibility that GNAI1 may have a limited role, and clarify that GNAO does not participate. Beer-Hammer and colleagues previously showed that inactivating GNAI2 worsens hearing loss in the Gnai3 null background, as tested in FoxG1-Cre; Gnai2^flox/flox; Gnai3^flox/flox adults (Beer-Hammer et al, 2018). Using a comparable yet distinct model (FoxG1-Cre; Gnai2^del/del; Gnai3^flox/flox), we were largely unable to obtain young adults and thus could not perform ABR. This suggests that our model is more severely affected, and probably lacks ABRs similar to the FoxG1-Cre; Gnai2^flox/flox; Gnai3^flox/flox (Beer-Hammer et al, 2018) and LSL-myc:ptxA (Tarchini et al, 2016) models.

Distribution of individual GNAI proteins in neonate hair cells

Next, we attempted to detect individual GNAI/O proteins in neonate HCs and define how they localize. Because GNAI/O proteins are close paralogs, validating and using specific antibodies is challenging. Instead, we used one commercial antibody raised against GNAI3 (scbt”GNAI3”; expected to detect GNAI3>GNAI1>GNAI2) and one antibody raised against GNAI2 (pt”GNAI2”; expected to detect GNAI2>GNAI1>GNAI3; see Methods) and systematically immunolabeled our collection of mouse models to tease apart protein-specific behavior.

Both antibodies produced the familiar GNAI pattern of enrichment at the bare zone and at row 1 stereocilia tips (Figure 4) (Tarchini et al, 2013; Tarchini et al, 2016). Both revealed generally normal GNAI enrichment in Gnai1^neo (Figure 4A), Gnai2^del(Figure 4B), and Gnai1^neo; Gnai2^del double mutants (Figure 4C). This indicates that pt”GNAI2” is not specific for GNAI2, can also detect GNAI3, and should thus be able to detect GNAI1, a better target than GNAI3. In Gnai3^neo single and Gnai1^neo; Gnai3^neo double mutants, both antibodies still revealed GNAI protein at the bare zone and at stereocilia tips although some HCs displayed incomplete enrichment (Figure 4D-E; further detailed below). This indicates that scbt”GNAI3” is not specific for GNAI3, can also detect GNAI2, and should thus be able to detect GNAI1, a better target than GNAI2. Finally, neither antibody detected consistent signal over background in Gnai2; Gnai3 double mutants (Figure 4F; FoxG1-Cre; Gnai2^del/del; Gnai3^flox/flox). Together, these results indicate that a) GNAI2 and GNAI3 share the same polarized distribution pattern at the bare zone and at stereocilia tips, and b) GNAI1 is either present in very low amounts hidden by background signal, or absent at the HC apical surface. Loss of GNAO in the Gnao1^neo or Atoh1-Cre; Gnao1^flox models did not alter signals obtained with scbt”GNAI3” (Figure 4 Supp. 1A-B). Moreover, a GNAO antibody produced unpolarized apical signals that we deemed unspecific because they were unchanged in Atoh1-Cre; Gnao1^flox/flox mutants (Figure 4 Supp. 1C). GNAO thus likely does not contribute to HC polarization, as also suggested by normal hair bundles and normal ABR thresholds in Gnao1 mutants.

Figure 4.

GNAI2 only partially spans the bare zone and stereocilia tips and partially rescues hair bundle development in postnatal hair cells lacking GNAI3

We next took a closer look at incomplete GNAI enrichment in Gnai3^neo and Gnai1^neo; Gnai3^neo mutant HCs (Figure 4D-E). In both models, we observed an identical outcome at the P0 cochlear base where remaining GNAI was unable to fully and consistently occupy the HC sub-domains where GNAI3 was missing (Figure 5A-B; bare zone, arrows; stereocilia tips, arrowheads). In Gnai1^neo; Gnai3^neo double mutants (Figure 5B), the GNAI protein detected is by default GNAI2 and, based on results above (Figure 4), we conclude that GNAI2 likely also forms the bulk of remaining GNAI in Gnai3^neo mutants (Figure 5A). Because in all cases GPSM2 co-localized with GNAI2 (Figure 5A-B), these results demonstrate that GNAI2, and not only GNAI3, can form a complex with GPSM2 in HCs. Intriguingly, the absence of GPSM2-GNAI2 on one side of the bare zone at P0 appeared to coincide with its absence at stereocilia tips on the corresponding side (Figure 5A-B). We thus quantified GNAI2 intensity at the bare zone and at stereocilia tips in half-OHCs in Gnai1^neo; Gnai3^neo mutants. While as expected control OHCs showed little variation in GNAI enrichment in either compartment (Figure 5 Supp. 1A), mutant OHCs showed highly variable signals that were significantly correlated between the bare zone and tips in the same half-OHC (Figure 5C). Analyzing HCs at different stages and tonotopic positions (gradient of increasing HC maturity from cochlear apex to base) clarified that GNAI2 is progressively unable to compensate for missing GNAI3. At E18.5, the GPSM2-GNAI2 complex could still occupy the totality of the bare zone in Gnai1^neo; Gnai3^neo mutants (Figure 5 Supp. 1B), suggesting that GNAI2 fully compensates for the loss of GNAI3 in embryonic HCs. Partial complementation by GNAI2 observed at P0 (Figure 5A-C) evolved into a lack of complementation at P6 at the cochlear base, where GNAI2 was no longer detected at the tips of stunted stereocilia in mutant IHCs (Figure 5D). In contrast, more immature P6 IHCs at the cochlear mid position retained partial GNAI2 enrichment correlated between the bare zone and tips (Figure 5E), as seen at P0 (Figure 5A-B). Time and position-based differences suggest that GNAI2 is progressively unable to occupy the full bare zone and tip compartments during HC differentiation.

Figure 5.

The progressive inability of GNAI2 to cover for GNAI3 in individual HCs helps explain the unique profile of apical surface defects in Gnai3^neo and Gnai1^neo; Gnai3^neo mutant HCs. Loss of GPSM2-GNAI2 in one wing of the OHC hair bundle led to stereocilia degeneration by P8 (Figure 5F). One-sided loss of global GNAI function at stereocilia tips is thus likely the origin of truncated hair bundle wings observed in adults Gnai3^neo and Gnai1^neo; Gnai3^neo mutant OHCs (Figure 2G). We divided the P8 OHC apical surface in two halves based on the position of the basal body at the hair bundle vertex, and measured the length of each hair bundle wing as well as the total apical surface area in the same half in Gnai1^neo; Gnai3^neo mutants (Figure 5F). We found a significant correlation between these two values (Figure 5G; control graph in Figure 5 Supp. 1C), providing new evidence that loss of stereocilia prompts a corresponding loss of flat HC surface area on the same OHC side (Etournay et al, 2010). Finally, we asked whether in time GNAI2 is lost at all stereocilia tips and along the entire cochlea in Gnai1^neo; Gnai3^neomutants. This proved not to be the case, as GNAI2 could still be detected at stereocilia tips in P28 OHCs and IHCs, although in low and variable amounts compared to GNAI tip signals in littermate controls (Figure 5 Supp. 1D-E).

In conclusion, GNAI2 still provides a low dose of GNAI protein at the tip of stereocilia that do not degenerate when GNAI3 is missing. Variable GNAI2 amounts likely explain why IHCs stereocilia have variably reduced heights in absence of GNAI3 (Figure 2B-C; Figure 5D-E), unlike in Gpsm2 or LSL-myc:ptxA mutants where they are more uniformly stunted (Beer-Hammer et al, 2018; Mauriac et al, 2017; Tadenev et al, 2019; Tarchini et al, 2016). Interestingly, correlated loss of GNAI2 signal at the bare zone and at stereocilia tips on the same HC side adds to previous evidence suggesting that bare zone enrichment is essential for GPSM2-GNAI transfer to adjacent row 1 stereocilia (Akturk et al, 2022; Jarysta & Tarchini, 2021; Tarchini et al, 2016).

Combined loss of GNAI2 and GNAI3 delays and de-polarizes bare zone expansion with drastic consequences on stereocilia distribution

Since GNAI2 and GNAI3 show functional redundancy and are the most important GNAI/O proteins for hair bundle differentiation, we next focused on characterizing early HC development in Foxg1-Cre; Gnai2^del/del; Gnai3^flox/flox mutants and comparing defects to those observed previously with Pertussis toxin. Unlike at adult stage, Gnai2; Gnai3 double mutants were obtained in close to Mendelian proportion at P0. For this purpose, we used the new DIO-ptxA allele in case the myc tag hindered ptxA activity in the LSL-myc:ptxA allele. We started with validating the new Gnai3^flox and DIO-ptxA alleles (Figure 2 Supp. 1B-C). As expected, GNAI signals at the bare zone and stereocilia tips were normal in Gnai3^flox/flox homozygotes but showed a distinctive incomplete pattern along with stunted stereocilia upon Cre recombination (Figure 6 Supp. 1A), as in constitutive mutants (Figure 5A). The new DIO-ptxA model produced in fact identical apical HC defects to the earlier LSL-myc:ptxA strain in single HCs (Kindt et al, 2021; Tarchini et al, 2016) (Figure 6 Supp. 1B). The fraction of inverted OHC1 in the DIO-ptxA model was lower than in the LSL-myc:ptxA model when using the later Atoh1-Cre (Figure 6 Supp. 1C-D), but identically encompassed 100% of OHC1 with FoxG1-Cre. Similar apical HC defects between strains indicate that the N-terminal myc and mild R26 promoter do not limit ptxA activity in LSL-myc:ptxA. For comparison purposes, we used the FoxG1-Cre driver to inactivate GNAI2/GNAI3 and to express ptxA, the same driver used by Beer-Hammer and colleagues (Beer-Hammer et al, 2018).

Figure 6.

Inactivating GNAI2 and GNAI3 abolished the bare zone at E17.5 based on the absence of GNAI labeling that was obvious in controls, and abnormally uniform F-actin signals at the HC surface (Figure 6A-B; arrows point to the bare zone). In contrast, E17.5 DIO-ptxA HCs had developed a distinct bare zone (Figure 6C, arrows). Measuring its surface area by HC type confirmed that the bare zone is virtually absent in conditional Gnai2; Gnai3 double mutants and reduced in ptxA-expressing OHCs but not IHCs (Figure 6D). By P0, most Gnai2; Gnai3 mutant HCs at the cochlear base had developed a region lacking microvilli or stereocilia, but its position at the apical surface was highly irregular, complementing extremely dysmorphic hair bundles (Figure 6E-F; arrows point to bare regions). In contrast, P0 DIO-ptxA HCs displayed largely coherent hair bundles and a distinct bare zone (Figure 6G, arrows). Quantifications revealed a significantly reduced bare area in P0 Gnai2; Gnai3 mutants compared to littermate controls (Figure 6H), showing that bare zone emergence and expansion is greatly delayed and deregulated in this model (compare P0 in Figure 6H and E17.5 in Figure 6D). In line with delayed bare zone expansion upon inhibiting GNAI function, bare zone surface area was back to normal in P0 DIO-ptxA HCs, except in the least mature OHC3 (compare P0 in Figure 6H and E17.5 in Figure 6D).

These results best illustrate to date the importance of GPSM2-GNAI for bare zone emergence, expansion and polarized positioning. While both mutant models consistently delay bare zone expansion, differences in timing and severity suggest that ptxA expression does not achieve as extended a loss of GNAI function as the Gnai2; Gnai3 double mutant. This is further underscored by severe stereocilia distribution defects observed in Gnai2; Gnai3 but not DIO-ptxA mutants. As reported previously (Kindt et al, 2021; Tarchini et al, 2013; Tarchini et al, 2016), the orientation of OHC1-2 expressing ptxA was reversed compared to controls (Figure 6C, G), and we expected to observe a similar pattern of misorientation in Gnai2; Gnai3 mutants. However, HC orientation proved challenging to assess in Gnai2; Gnai3 mutants due to highly dysmorphic hair bundles (Figure 6A-B; E-F). We used acetylated tubulin (AcTub) and pericentrin (PCNT) as markers for the kinocilium and the position of the basal body, respectively. This showed that in Gnai2; Gnai3, but not in DIO-ptxA mutants, the basal body and kinocilium were sometimes in an approximately central position surrounded partially or entirely by stereocilia (Figure 6E-F). Rounded perinatal hair bundles are a hallmark of HCs where the early off-center migration of the basal body, hence symmetry breaking, is defective.

To distinguish defects affecting symmetry-breaking versus HC orientation and to uniformize analysis of HC polarization, we next used the position of the basal body at the base of the kinocilium to derive both HC eccentricity and HC orientation. This strategy helped compare the two mouse models, and identified two early functions for GNAI before its association with GPSM2 for stereocilia elongation.

GNAI proteins drive the off-center migration of the basal body

We used the position of the basal body to measure HC eccentricity as a metric for cytoskeleton asymmetry. Eccentricity was calculated as a ratio reporting how far away from the cell center the basal body was positioned (see schematic in Figure 7). A perfectly symmetrical HC with a central basal body would thus have an eccentricity ratio close to 0 whereas a ratio close to 1 would indicate that the off-center basal body is juxtaposed to the apical junction. Because HC maturation progresses in time and along the tonotopic axis (cochlear apex to base gradient of increasing maturity), we measured eccentricity at E17.5 (base and mid cochlear position) and at P0 (base, mid and apex) to understand how eccentricity progressed at the HC population level.

Figure 7.

Control HCs at E17.5 had already broken symmetry at the mid position, with eccentricity averaging ∼0.5 and increasing at the more mature base position (Figure 7A). At E17.5 mid and base positions, Gnai2; Gnai3 mutants showed significantly reduced eccentricity, suggesting a delay in the off-center migration of the basal body. We arbitrarily defined 0.25 as a threshold below which HCs were considered “symmetrical”. Symmetrical HCs were not observed in controls but represented up to ∼29% of IHCs and ∼34% of OHCs in mutants at E17.5 (detailed in Figure 7A by stage, position and HC type, red highlights). By P0, the proportion of symmetrical cells in mutants had decreased at all positions for OHCs (3-15%), but remained high for IHCs (21.5%; Figure 7A). P0 OHC eccentricity generally remained lower in mutants than in controls, with a trend towards less significant differences in OHCs at the cochlear base where HCs are most mature. In contrast, IHCs did not appear to recover by P0, as eccentricity was much lower than in controls, even at the cochlear base.

The situation was radically different in the DIO-ptxA model. First, no symmetrical HC was observed at any stage or position in mutants (Figure 7B). In fact, eccentricity was instead higher in DIO-ptxA OHCs compared to littermate controls, with a tendency for this difference to become less significant along with OHC differentiation (see below for interpretation). IHCs had overall a normal distribution of eccentricity compared to controls in the DIO-ptxA model.

Defects in bare zone expansion (Figure 6D, H) can help explain the distinct eccentricity defects in the DIO-ptxA and Gnai2; Gnai3 mouse models. At E17.5 and P0, the position of the basal body is the sum of two opposite movements (Figure 1A): a) the early off-center migration that brings it in the vicinity of the lateral junction, and b) the subsequent relocalization towards the cell center that brings it in contact with the forming hair bundle upon bare zone expansion (Tarchini et al, 2013). In DIO-ptxA, we interpret transiently increased eccentricity in OHCs (Figure 7B) as the outcome of normal off-center basal body migration yet delayed bare zone expansion (Figure 6D, H). In Gnai2; Gnai3 mutants, however, decreased eccentricity in OHCs (Figure 7A) stems from defective basal body off-center migration compounded with a first absent (E17.5), and then reduced (P0), bare region (Figure 6D, H). When an unpolarized bare region eventually emerges in Gnai2; Gnai3 mutants (Figure 6H), it impacts basal body position without directionality, leading to highly variable eccentricity values compared to controls (Figure 7A), and thus variably dysmorphic hair bundles. The apparent increase in symmetrical IHCs over time in Gnai2; Gnai3 mutants (Figure 7A) may result from the delayed expansion of the bare region, which will relocalize the basal body centrally in a proportion of IHCs.

In conclusion, the loss of GNAI activity achieved in Gnai2; Gnai3 mutants impairs symmetry-breaking (Figure 7) before it impairs bare zone emergence and expansion (Figure 6). PtxA-based downregulation of GNAI activity is sufficient to delay bare zone expansion, but not to affect symmetry-breaking. As only ∼3-15% of OHCs and ∼21% of IHCs in Gnai2; Gnai3 mutants are symmetrical at the P0 cochlear base, there is likely sufficient GNAI activity remaining in this model as well for most HCs to break planar symmetry.

GNAI proteins orient hair cells laterally

GNAI proteins were proposed to signal downstream of the GPR156 receptor to regulate HC orientation in the auditory and vestibular organs (Jiang et al, 2017; Kindt et al, 2021). In HCs expressing the transcription factor EMX2, GPR156 reverses the orientation of the off-center basal body migration, and thus HC orientation, compared to Emx2-negative HCs (Tona & Wu, 2020). To date, however, the HC misorientation profile observed in Emx2 or Gpr156 mutants was only recapitulated using ptxA (Figure 1B). If ptxA-based orientation defects are physiologically relevant to GNAI function, they should be observed in Gnai2; Gnai3 mutants that achieve a further loss of GNAI activity.

To test this assumption, we first excluded symmetrical HCs (eccentricity< 0.25) in the E17.5 and P0 datasets since an asymmetric cytoskeleton is pre-requisite for HCs to exhibit a defined orientation. Next, we used the position of the off-center basal body in the remaining HCs to infer their orientation. We measured the angle formed by a vector running from the HC center to the basal body relative to the cochlear longitudinal axis (α, see schematic in Figure 8). Angles were plotted in circular histograms where 0° pointed to the cochlear base along the longitudinal axis, and 90° to the lateral edge.

Figure 8.

At E17.5 at the cochlear mid position, DIO-ptxA already showed the graded pattern of OHC misorientation described previously, with inverted OHC1-2 and imprecisely oriented IHCs and OHC3 (Figure 8A). At this stage and position, Gnai2; Gnai3 mutant HCs above threshold showed abnormal and highly variable orientations without a clear trend towards OHC1-2 inversion yet. Strikingly however, more mature E17.5 OHC1 at the base were largely inverted (Figure 8B), as also observed at P0 at all positions (Figure 8C-D; Figure 8 Supp. 1A). OHC2 were still largely oriented laterally at the E17.5 mid (Figure 8B) and P0 apex and mid (Figure 8C; Figure 8 Supp. 1A), but clearly adopted an inverted medial orientation at the P0 base (Figure 8D). These results thus confirm for the first time that endogenous GNAI proteins are integral to proper HC orientation. They also clarify that the level of GNAI activity required for proper HC orientation is higher than for symmetry-breaking since both DIO-ptxA and Gnai2; Gnai3 mutants showed HC misorientation. It should be noted that concomitant deficits in HC eccentricity and the arbitrary eccentricity cut-off at 0.25 were expected to blur the misorientation pattern, explaining why OHC1-2 inversion only becomes apparent at later stages of differentiation in Gnai2; Gnai3 compared to DIO-ptxA mutants. Interestingly, unlike in DIO-ptxA mutants, IHC and OHC3 were also severely misoriented in Gnai2; Gnai3 mutants at all stages and positions analyzed (Figure 8; Figure 8 Supp. 1A). We thus conclude that a lower level of GNAI activity is sufficient to properly orient IHC and OHC3 compared to OHC1-2 (see Model Figure 9).

Figure 9.

Discussion

By examining a large collection of single and combined mutations in Gnai genes (Gnai1, Gnai2, Gnai3, Gnao1), we assign here three distinct and successive roles for GNAI proteins during the apical polarization of a developing HC (Figure 9): a) drive the centrifugal migration of the basal body when a prospective HC breaks planar symmetry, b) orient this migration along the PCP axis in a binary manner, and c) position and elongate stereocilia during hair bundle morphogenesis. Key results in the study include demonstrating that endogenous GNAI proteins indeed serve early functions (a) and (b), since to date only hair bundle defects (c) were reported in knock-out mouse models where Gnai genes were directly targeted (Beer-Hammer et al, 2018; Mauriac et al, 2017). Importantly, previous studies indicated that GNAI proteins partner with different regulators in these different functions, organizing and elongating stereocilia by binding to the scaffold GPSM2 (c), and reversing HC orientation in Emx2-expressing HCs via heterotrimeric signaling downstream of the GPR156 receptor (b). A GNAI regulator or mechanism for basal body off-center migration (a) remains to be identified.

Different GNAI/O identity, dosage and timing underlie different functions

Interestingly, besides involving different regulators, each function also involves different dosage and specific identity of the underlying GNAI protein(s) at different times during HC development (Figure 9). For postnatal hair bundle morphogenesis (c), GNAI proteins are sequestered in the GDP state by GPSM2, forming a complex highly enriched at the bare zone and at row 1 stereocilia tips that can be reliably immunodetected (Akturk et al, 2022; Kindt et al, 2021). In these two sub-cellular compartments, the identity of the GNAI proteins at work appears to matter, with GNAI3 playing a necessary and prominent role while GNAI2 is important yet not required. We cannot rule out that GNAI1 participates as well in a minor way because Gnai1 mutants appear to have more variable stereocilia height than littermate controls (Figure 2C), and we observed more severe stereocilia defects and more elevated ABR thresholds in Gnai1; Gnai3 compared to single Gnai3 mutants (Figure 2C, E; Figure 3D). The latter observations might instead reflect differences in the genetic background, however. In stark contrast, GNAI2 can fully rescue normal off-center basal body migration (a) and normal HC orientation (b) when both GNAI1 and GNAI3 are constitutively absent (Figure 9). In FoxG1-Cre; Gnai2; Gnai3 double mutants, a majority of HCs eventually break symmetry, strongly suggesting that GNAI1 and/or GNAO can participate in the off-center migration mechanism. Finally, even high expression of untagged ptxA in the DIO-ptxA model does not affect symmetry-breaking in cochlear HCs, whereas ptxA severely disrupts HC orientation (Figure 9). We thus conclude that i) ptxA downregulates but does not abolish GNAI/O activity in HCs, and ii) the GNAI dosage required to ensure each function is graded (a)<(b)<(c), with probably several GNAI/O proteins able to participate in embryonic functions (a, b). A low dose of GNAI/O dynamically cycling between the GDP and GTP-bound state in heterotrimeric G protein complexes could explain why GNAI proteins are not immunodetected along with GPR156 or other regulators besides GPSM2 (Akturk et al, 2022; Kindt et al, 2021).

A strict dosage- and time-dependence for correct GNAI functions explains why our conditional Gnai2; Gnai3 double mutant model is more severely affected than a comparable conditional Gnai2; Gnai3 model where only hair bundle defects (c) were reported (Beer-Hammer et al, 2018). While Beer-Hammer and colleagues used FoxG1-Cre as a driver as we did, combined GNAI2 and GNAI3 inactivation demanded Cre recombination at four floxed loci (Gnai2^flox/flox; Gnai3^flox/flox). In contrast, our model demands recombination at two loci only (Gnai2^del/del; Gnai3^flox/flox). We thus conclude that our model achieves an earlier and greater loss of GNAI activity. This explanation is supported by extensive perinatal lethality in our model, whereas Beer-Hammer and colleagues could analyze Gnai2; Gnai3 double mutants as adults. Our work thus usefully reconcile all previous reports about GNAI functions, notably showing that inactivating endogenous GNAI proteins does produce early defects (a,b) so far only observed with ptx. We also validate all HC defects observed with ptx as physiologically relevant and specific to GNAI/O function.

Hair cell break of symmetry

Although cochlear HCs expressing ptxA undergo normal symmetry-breaking in vivo (Kindt et al, 2021; Tadenev et al, 2019; Tarchini et al, 2013; Tarchini et al, 2016), a fraction of HCs expressing ptxA in the utricle and saccule have abnormally central basal bodies (Kindt et al, 2021). This suggests that a higher dose of GNAI is required for off-center migration of the basal body in vestibular compared to cochlear HCs, making vestibular HCs more susceptible to ptxA. “Ciliary” proteins are required for ciliogenesis, including kinocilium formation and maintenance, and also play non-ciliary functions (May-Simera et al, 2015; Sipe & Lu, 2011). Inactivation of intraflagellar transport protein IFT88 was reported to result in a central basal body and circular hair bundles in ∼10% of OHCs (Jones et al, 2008), but the underlying mechanism was not elucidated. A low proportion of symmetrical HCs were also reported in absence of the CD2 isoform of the protocadherin PCDH15 that forms inter-stereocilia and kinocilium-stereocilia fibrous links during embryogenesis (Webb et al, 2011). It remains unclear whether GNAI activity is involved or not in Ift88 or Pcdh15 mutant defects, or whether GNAI could be active at the basal body or the kinocilium.

Hair cell orientation

The PCP axis is defined by opposite asymmetric enrichment of the core PCP transmembrane proteins VANGL2 and FZD3/6 and their specific cytosolic partners at HC apical junctions (Deans, 2013; Montcouquiol & Kelley, 2019; Tarchini & Lu, 2019). Previous work showed that regional Emx2 expression in HCs reverses how the basal body migrates relative to early-set core PCP landmarks, establishing the line of polarity reversal in the utricle and saccule (Jiang et al, 2017; Tona & Wu, 2020). EMX2 activates the GPCR GPR156 by triggering its polarized enrichment at the apical HC junction, where downstream GNAI signaling appears to repel the basal body (Figure 9, b) (Kindt et al, 2021). It is important to note that HCs lacking GPR156 have a normal apical cytoskeleton, including a normally formed hair bundle, and only show orientation defects (Kindt et al, 2021). This indicates that the centrifugal move of the basal body and the orientation of this move that defines early HC orientation are two distinct molecular mechanisms that both involve GNAI/O proteins (Figure 9; a-b). HC orientation is regulated by core PCP proteins defining an axis of polarization, and by the EMX2>GPR156>GNAI cascade defining binary orientation along that axis. Following the same logic, HCs lacking GPSM2 are not inverted in orientation as in Emx2, Gpr156, Gnai2; Gnai3 or ptxA mutants (Bhonker et al, 2016; Ezan et al, 2013; Tarchini et al, 2013). This shows that bare zone enrichment of GPSM2-GNAI is a mechanism to shape and polarize the growth of the hair bundle, but not to migrate off-center or orient the basal body. It is worth noting that GPSM2-GNAI is polarized at the HC apical membrane on the side of the off-center basal body, and not at the apical junction where core PCP proteins regulate HC orientation (Siletti et al, 2017; Tarchini et al, 2013).

While both GPR156 (Greene et al, 2023) and GPSM2 (Doherty et al, 2012; Walsh et al, 2010) have been reported as human deafness genes, there is currently no evidence implicating any GNAI/O protein. This is most likely because even GNAI3, which is more specifically required for hair bundle morphogenesis (Figure 9), plays ubiquitous and critical signaling roles in the context of heterotrimeric protein signaling across many cell types.

Methods

Mouse strains and husbandry

All mouse strains used in this work and strain-related information is summarized in Supplementary Table 1. All primers used for genotyping are indicated in Supplementary Table 2 by strain.

Strains from The Jackson Laboratory

Gnai1^neo; Gnai3^neo (Gnai1^tm1Lbi; Gna3^tm1Lbi; MGI: 5560183) (Jiang et al, 2002) carries a neo cassette in Gnai1 exon 3 and a neo cassette replacing part of intron 5 and exon 6 in Gnai3 in the 129S1/SvlmJ background. The single Gnai1^neo and single Gnai3^neo alleles were segregated from Gnai1^neo; Gnai3^neo by breeding with C57BL/6J mice and are consequently on a mixed 129S1/SvlmJ: C57BL/6J background. Gnao1^neo (Gnao1^tm1Lbi; MGI: 2152685) carries a neo cassette in Gnao1 exon 6 in the 129S1/SvImJ background (Jiang et al, 2002). The two Cre strains used in this work are Atoh1-Cre (Tg(Atoh1-cre)1Bfri; MGI: 3775845) expressing Cre in HCs from E14.5 (Matei et al, 2005) and FoxG1-Cre (Foxg1^tm1(cre)Skm; MGI: 1932522) expressing Cre in the prospective otic vesicle from E8.5 (Hebert & McConnell, 2000).

Consortium strain

Gnao1^flox (Gnao1^{tm1c(EUCOMM)Hmgu}) is a derivative of the EUCOMM strain Gnao1^{tm1a(EUCOMM)Hmgu} (MGI: 4456727; “KO first, conditional ready”). The Gnao1^flox conditional allele was produced via FLP-mediated recombination (FLP strain MGI: 4830363) to remove the FRT-flanked LacZ-neo cassette, leaving exon 3 floxed. Gnao1^flox is on a C57BL/6N background.

Constitutive Gnai2^del inactivation (see Figure 2 Supp. 1A).

Gnai2^del (Gnai2^em1Btar; MGI: 6466534) was generated using direct delivery of CRISPR-Cas9 reagents in mouse zygotes via electroporation. The following guide RNAs (gRNA) were used to delete exon 2, exon 3 and part of exon 4 in the C57BL/6J background. Upstream gRNA: CTGCCCTCTGTTCCAGGTGC, downstream gRNA: ATGCTTCCTGAAGACCTGTC. The resulting 681 bp deletion encompassed TGCTGGAGAGTCAGGGAAGA…CCTGAAGACCTGTCCGGTGT. The electroporation mixture consisted of the gRNAs with AltR-Streptococcus pyogenes Cas9 (SpCas9) V3 (Integrated DNA Technologies #1081059) in embryo-tested TE buffer (pH 7.5). Electroporation of zygotes was performed as described in (Qin et al, 2015). In order to segregate away potential non-specific mutations, founders were bred for two generations with C57BL/6J animals to generate a N2 heterozygote stock. Because Gnai2^del homozygotes showed low viability, this strain was used in a mixed C57BL/6J:FVB/J background that improved the proportion of homozygotes (see Supp. Table 1).

Conditional Gnai3^flox inactivation

(see Figure 2 Supp. 1B). A plasmid-based donor vector was cloned to flank Gnai3 exon 2 and 3 with loxP sites using the same CRISPR/Cas9 system described above for Gnai2^del. A fragment carrying a loxP, a genomic region including exon 2-3, the restriction site ClaI, a second loxP and the restriction site XhoI was synthetized (Genscript) and cloned between 5’ and 3’ homology arms amplified by PCR using Gibson assembly. The following gRNAs were used and define the extremities of the floxed region: upstream, AGCTCACCAAAATTCCCATT, TAGGGGATATAGATCCAAAT, downstream, TTCCAGGACTCTGCATGCGT, TACCGACGCATGCAGAGTCC. The gene editing reagents (gRNAs, donor vector, SpCas9) were microinjected in zygotes as described in (Qin et al, 2015). To confirm insertion at the Gnai3 locus, we performed long-range PCRs with a genomic primer located outside the homology arm on each side: 5’ long-range: F(external)_ TACTGAGATGAGAGACTGAGGG and R (internal)_TGGCTGACATCCTTTGATGGAC. The 3070 bp product was digested with XhoI, producing 2,551 + 519 bp fragments upon donor insertion (flox allele). 3’ long-range: F (internal)_TGAAAGGTAAAGGCAACGTGAG and R (external)_TGTGAGACAGGGTCTCTCTTTG. The 2,966 bp product was digested with XhoI, producing 2,000 + 966 bp fragments upon donor insertion (flox allele). In order to segregate away potential non-specific mutations, founders were bred for two generations with C57BL/6J animals to generate a N2 heterozygote stock.

Pertussis toxin catalytic subunit (ptxA)-expressing strains

(see Figure 2 Supp. 1C-D). The LSL-myc:ptxA strain at the ROSA26 (R26) locus was described previously (Gt(ROSA)26Sor^{em1(ptxA)Btar}; MGI: 6163665) (Tarchini et al, 2016). The new CAG-DIO-ptxA strain line (see Figure 2 Supp. 1C) was generated at the R26 locus using the Bxb1 attP(GT) integrase technology and C57BL/6J-Gt(ROSA)26Sor^/Mvw (MGI: 6757188) as the host strain (Low et al, 2022). The donor vector consisted in a CAG promoter followed by the flipped ptxA coding sequence flanked by double inverted lox sites (loxP and lox2272; Figure 2 Supp. 1C) and a bGHpA sequence. In order to exclude the prokaryotic vector backbone, the donor vector was prepared as a minicircle using the MC-Easy Minicircle Production kit (SystemBio, MN920A-1). Briefly, the donor plasmid insert was cloned into a Minicircle Cloning Vector using PhiC31 integrase before transformation into the ZYCY10P3S2T E. coli minicircle producer strain, which after induction with arabinose, results in the generation of the donor minicircle. To eliminate parental plasmid contamination, a restriction digest was performed, followed by MC-safe DNase treatment. The minicircle DNA was then purified by phenol-chloroform extraction and reconstituted in microinjection buffer (10 mM Tris; 0.1 mM EDTA pH7.5). The CAG-DIO-ptxA strain was generated by pronuclear microinjection of the Bxb1 Integration reagents directly into zygotes of the host strain. These reagents included the Bxb1 mRNA (Trilink) at 100ng/µl, RNasin (Promega) at 0.2U/µl and the donor minicircle DNA at 10ng/µl combined in microinjection buffer. To confirm successful integration of the 3,212bp transgene, as well as identify random transgenics, the initial screening was performed using a four-PCR strategy. The In/Out Left and Right (IOL, IOR) PCRs, which bridge the recombined attachment sites, each contain one primer specific to the R26 locus and a second primer designed against the inserted sequence. The Transgene (TG) PCR amplifies a non-genomic sequence in the insert, and the Off-Target Integration (OTI) PCR was designed to detect non-recombined minicircle integration, presumably outside of the R26 locus.

IOL PCR (F1/R1), F1: GTCGCTCTGAGTTGTTATCAGT, R1: GCCAAGTAGGAAAGTCCCATAA (720 bp, WT=no band). IOR PCR (F2/R2), F2: GGTGATGCCGTTGTGATAGA, R2: TGTGGGAAGTCTTGTCCCTCCAAT (1,013 bp, WT=no band). TG PCR (F2/R3) F2: 5’-GGTGATGCCGTTGTGATAGA, R3: CCACTTCATCGGCTACATCTAC (214 bp, WT=no band). OTI PCR (F3/R1) F3: GGGAGGATTGGGAAGACAATAG, R1: GCCAAGTAGGAAAGTCCCATAA (578 bp, WT=no band). 33 mice were born following 6 transfers totaling 117 injected embryos (28% survival), and 1/33 (3%) was identified as a founder and crossed to C57BL6/J to generate N1 offspring. The In/Out PCR confirmed that a copy of the transgene was inserted at the R26 locus, but the OTI strategy also gave a product, even after breeding for multiple generations. As breeding would have segregated a separate, random integration of the transgene, we performed Nanopore-based Cas9-targeted sequencing at the R26 locus (Gilpatrick et al, 2020; Low et al, 2022). This revealed that, as seen in some cases previously (Low et al, 2022), tandem insertions had occurred in this founder Specifically, 3 consecutive copies of the CAG-DIO-ptxA transgene were inserted at the R26 locus in this strain. This did not impact specific Cre-based expression of ptxA because phenotypes observed in this new strain were comparable to the previous LSL-myc:ptxA strain (Figure 6 Supp.1 B-D).

Experimental animals in the study ranged in age between E17.5 and P29 as indicated in each figure. Male and females were systematically included but sex was not tracked except for Auditory Brainstem Recordings because there is no evidence that sex influences HC orientation or hair bundle morphogenesis. Animals were maintained under standard housing conditions (14h light / 10h dark cycle, ambient temperature and normal humidity). All animal work was reviewed for compliance and approved by the Animal Care and Use Committee of The Jackson Laboratory (Animal Use Summary AUS #14012).

Scanning Electron Microscopy (SEM)

Temporal bones were isolated, punctured at the cochlear apex, and fixed by immersion for at least one overnight at 4°C in 2.5% glutaraldehyde (Electron Microscopy Sciences; 16200) and 4% paraformaldehyde (Electron Microscopy Sciences; 15710) in 1mM MgCl₂, 0.1M Sodium Cacodylate, 20mM CaCl₂. Samples were rinsed and decalcified overnight in 4% EDTA. The auditory epithelium was then dissected into 3 pieces (cochlear base, mid, and apex) before progressive dehydration in an ethanol series (30-50-70-80-90-100%, at least 20 minutes per step) and chemical drying with hexamethyldisilazane (HMDS; Electron Microscopy Sciences 50-243-18). Dry samples were mounted on aluminum stubs using double-sided carbon tape and sputter-coated with gold-palladium before imaging on a Hitachi 3000N VP electronic microscope at 20kV.

Immunofluorescence and antibodies

For embryonic and postnatal stages, temporal bones were either immediately dissected to expose the sensory epithelium and fixed in paraformaldehyde (PFA 4%; Electron Microscopy Sciences; 15710) for 1h at 4°C. After fixation, the tectorial membrane was removed, and samples were permeabilized and blocked in PBS with 0.5% Triton-X100 and bovine serum albumin (1%) for at least at 1 hour at room temperature. For adult stages, temporal bones were isolated and punctured at the cochlear apex to facilitate access for the fixative. Samples were then immersion-fixed in PFA 4% for 1 hour at 4°C, rinsed in PBS, and incubated overnight in 4% EDTA for decalcification. Cochleae were next dissected in 3 pieces (cochlear base, mid, and apex), before permeabilization and blocking as described for earlier stages. Primary and secondary antibodies were incubated overnight at 4°C in PBS with 0.025% sodium azide. Fluorescent dye-conjugated phalloidin was added to secondary antibodies. Samples were washed 3 times in PBS + 0.05% Triton-X100 after each antibody incubation and post-fixed in PFA 4% for at least 1 hour at room temperature. Samples were then mounted flat on microscopy slides (Denville M102) using Mowiol as mounting medium (Calbiochem/MilliporeSigma 4759041), either directly under a 18×18mm #1.5 coverglass (VWR 48366-045) (postnatal cochleae) or using one layer of office tape as a spacer (adult cochleae). Mowiol (10% w/v) was prepared in 25%(w/v) glycerol and 0.1M Tris-Cl pH8.5. Primary antibodies used were:

Rabbit anti-GNAI2, pt”GNAI2” (Proteintech, 11136-1-AP); the antigen is the full human GNAI2 protein, and GNAI1 is a closer paralog to GNAI2 compared to GNAI3. Rabbit anti-GNAI3, scbt”GNAI3” (Santa Cruz Biotechnology, sc-262); the C-terminal rat antigen was not disclosed but the company indicated that this product detects preferentially GNAI3>GNAI1>GNAI2. Rabbit anti-GNAO (Proteintech, 12635-1-AP) Mouse anti-acetylated alpha tubulin (Santa Cruz Biotechnology scbt-23950) Rabbit anti-GPSM2 (Sigma, A41537) Goat anti-GPSM2 (Thermofisher, PA5-18646) Rabbit anti-Pericentrin/PCNT (Biolegend/Covance, PRB-432C) Rat anti-ZO1 (Developmental Studies Hybridoma Bank, R26.4C)

Secondary antibodies from ThermoFisher Scientific were raised in donkey and conjugated to Alexa Fluor (AF) 488, 555, or 647 (donkey anti-rabbit AF555 (A-31572), AF647 (A-31573), donkey anti-mouse AF647 (A-31571), donkey anti-rat AF488 (A-21208), donkey anti-goat AF555 (A-21432), AF647 (A-21447)). Fluorescent conjugated phalloidins used to reveal F-actin were from ThermoFisher Scientific: AF488 (A12379); AF555 (A34005) and Sigma-Aldrich: FITC (P5282).

Sample cohorts, image acquisition and analysis

All quantifications included at least three animals per genotype. All graphs or their legends indicate the animal cohort size (N) as well as the number of HC or stereocilia (n) analyzed. Further details can be found in a single Supplementary Source Data file where tabs are referencing figure panels in order. When an experimental outcome was not quantified, at least 3 mutant and 3 control littermates encompassing two or more litters were analyzed, and figure panels illustrate a representative outcome observed in all samples of the same genotype.

Confocal images were captured with a LSM800 line scanning confocal microscope, a 63x NA1.4 oil objective, the Airyscan detector in confocal mode (except in Figure 5 Supp.1 D, E where the Airyscan detector was used in Airyscan mode) and the Zen 2.3 or Zen 2.6 software (Carl Zeiss AG). Raw Airyscan images in Figure 5 Supp.1 D, E were processed in Zen 2.6 selecting for automatic strength. Unless stated otherwise in the legends, images show a single optical z plane. To quantify HCs eccentricity (Figure 7), images were captured with a Leica DM5500B widefield microscope, a 63x oil objective, a Hamamatsu ORCA-Flash4.0 sCMOS camera and the Leica Application Suite (LasX) software (Leica Microsystems). All images in the same experiment were acquired using the same laser intensity and gain, and were then processed in Adobe Photoshop (CC2020) where the same image treatment was applied across conditions.

To measure stereocilia length and width in IHCs (Figure 2C-D), SEM samples were imaged laterally with a 10,000x magnification and an appropriate tilt (from 0 to 30°) to bring stereocilia parallel to the imaging plane and minimize parallax. To quantify the number of rows in adult IHCs and the number of stereocilia in their first row (Figure 2E, F), SEM samples were imaged medially with a 5,000x magnification. To measure the length of OHC hair bundle wings (Figure 2G, Figure 2 Supp. 1G), OHCs were imaged from the top at 5,000x. Stereocilia width was measure at half-length and all measurements were done with the straight-line tool in Fiji.

To quantify GNAI signal intensity in Gnai1^neo; Gnai3^neo mutants (Figure 5C, Figure 5 Supp. 1A), Z-stack series were acquired at the one-turn cochlear position. A single Z slice was chosen at the bare zone level, another one at the stereocilia tip level, each based on strongest signal in that compartment. The vertex of the V-shaped hair bundle was used to divide the HC apical surface in two equal halves. In each half (left and right), regions of interest (ROIs) were drawn using the polygon selection tool in Fiji to encompass all signals in that apical compartment, and mean gray values were measured. For each image, background signal was measured and averaged, and subtracted from all measurements in the same image.

To quantify surface area of the bare zone or total surface area in half-OHC as well as the length of half hair bundles (Figure 5G, Figure 5 Supp. 1C; Figure 6D, H), Z-stack series were acquired at the cochlear base and a single Z slice was selected at apical junction level using ZO1 (Figure 5G, Figure 5 Supp. 1C) or F-actin (Figure 6D, H) as reference. To measure total apical surface area and bundle length in half-OHCs, the pericentrin-stained basal body was used to divide the apical surface in two halves (Figure 5G, Figure 5 Supp. 1C). The ZO1-positive cell outline along with the dividing line at the basal body level were used as reference to measure surface area with the polygon selection tool in Fiji. To quantify the bare zone surface area (Figure 6D, H), a ROI was drawn around the total apical surface lacking phalloidin (F-actin) signals. The length of the F-actin stained hair bundle was measured using the straight-line tool in Fiji.

To determine HC eccentricity (Figure 7), the geometrical center of the HC apical surface was determined as the intersection of two orthogonal lines representing the maximal diameter of the cell along the cochlear longitudinal and radial axes. A vector (BB) was drawn from the center to the pericentrin-labeled basal body, and eccentricity was calculated as the ratio of BB length over the cell radius (r) along the same trajectory using F-actin-labeled apical junction as landmark (see Figure 7). To determine cell orientation (Figure 8 and Figure 8 Supp. 1), the angle (α) separating the longitudinal axis of the organ of Corti from the BB vector was measured with the angle tool in Fiji (see Figure 8). Both right and left cochleae were used and angles were measured so that 0° pointed toward the cochlear base and 90° towards the cochlear periphery (lateral). Eccentricity and angles were measured at the positions indicated (base ∼20%, mid ∼50%, apex ∼80% of the cochlear length starting from the base).

Auditory Brainstem Response (ABR) tests

All tests were performed in a sound-attenuating chamber, and body temperature of the anesthetized animals was maintained at 37°C using a heating pad (FHC Inc.). Animals from all strains except Atoh1-Cre; Gnao^flox (Figure 3 Supp. 1B, see below) were anesthetized with a mix of ketamine and xylazine (1 mg and 0.8 mg per 10g of body weight, respectively) and tested using the RZ6 Multi-I/O Processor System coupled to the RA4PA 4-channel Medusa Amplifier (Tucker-Davis Technologies). ABRs were recorded after binaural stimulation in an open field by tone bursts at 8, 16, 32, and 40 kHz generated at 21 stimuli/second, and a waveform for each frequency/dB level was produced by averaging the responses from 512 stimuli. Subdermal needles were used as electrodes, the active electrode inserted at the cranial vertex, the reference electrode under the left ear and the ground electrode at the right thigh.

Atoh1-Cre; Gnao1^flox animals (Figure 3 Supp. 1B) were anesthetized with tribromoethanol (2.5mg per 10g of body weight) and tested with the Smart EP evoked potential system from Intelligent Hearing Systems (IHS). ABR were recorded after single ear stimulation (right ear), using ear tubes speakers (ER3C insert earphones, IHS) delivering tone bursts at 8, 16, and 32 kHz generated at 40 stimuli/second. Electrodes were positioned as previously described except for the ground electrode that was placed at the base of the tail. ABR thresholds were obtained for each frequency by reducing the sound pressure level (SPL) by 5 decibels (dB) between 90 and 20 dB to identify the lowest level at which an ABR waveform could be recognized. We compared waveforms by simultaneously displaying 3 or more dB levels on screen at the same time.

Statistical Analyses

All data were plotted in Prism 9 (GraphPad), except for circular diagrams of HC orientation (Figure 8; Figure 8 Supp. 1) that were generated with R (4.2.2) and Rstudio (2022.12.0+353).

All data except for angles in circular diagrams (Figure 8 and Figure 8 Supp. 1) were plotted individually. Distribution was framed with 25-75% whisker boxes where exterior lines show the minimum and maximum, the middle line represents the median, and + represents the mean. Potential differences in data distribution between genotypes were tested for significance using Mann-Whitney U test (non-parametric unpaired test), except for ABR thresholds (Figure 3 and Figure 3 Supp. 1) where a 2-way ANOVA with Sidak’s multiple comparison post-hoc test was used. OHC half-bundle lengths (Figure 2G, Figure 2 Supp. 1G) were plotted as paired left and right values in the same HC and a potential difference in data variance between genotypes was tested using a F-test. GNAI signal intensity at the OHC bare zone and stereocilia tips (Figure 5C, Figure 5 Supp. 1A) as well as OHC surface area and half-bundle length (Figure 5G, Figure 5 Supp. 1C) were plotted and a simple linear regression curve was calculated and drawn for each pair of datasets. A correlation between variables was addressed with Pearson’s correlation test. Exact p-values were indicated on each graph when non-significant (p>0.05) and otherwise summarized a follows: p<0.0001 ****, p<0.001 ***, p<0.01**, p<0.05. Details of statistical results for all figures can be found in the Supplementary Data file.

Angle frequency distribution (Figure 8 and Figure 8 Supp. 1) was plotted in circular diagrams using the R package dyplr and the coord_polar function of the ggplot2 package to organize data and produce the graphs, respectively. The angle formed by the red line indicates the circular mean and the length of the arc at the end of the red line indicates the mean circular deviation. Both values were obtained using the colstats function in the R circular package. All scripts used in this work will be posted on github.

Acknowledgements

We are grateful to Elli Hartig for reading and commenting on the manuscript. We thank Simon Lesbirel for nanopore-based Cas9-targeted sequencing of the DIO-ptxA strain. We are grateful to The Jackson Laboratory Genome Engineering Technology and Reproductive Science services for their help with generating the Gnai2^del and Gnai3^flox strains, and cryorecovering the Gnai1^neo; Gnai3^neo and Gnao1^neo strains, respectively. A.J. was supported by a postdoctoral fellowship from Fondation Pour l’Audition (2018-2020; FPA RD-2018-3). This work was supported by the National Institute on Deafness and Other Communication Disorders grants R01DC015242 and R01DC018304 (to B.T.).

Author contribution

B.T. conceived the study and designed most experiments. A.J and A.L.T designed, performed and analyzed experiments. M.D. provided technical help with mouse colony maintenance, sample processing and genotyping. B. K. wrote the R script to generate the circular histograms. B.E.L and M.V.W. helped generate the DIO-ptxA strain using BxB1 recombination technology. B.T supervised the work. B.T. and A.J. wrote the manuscript. B.T. secured funding.

Competeting interests

The authors declare no competing interest.

Data availability

The data in all graphical representations are included in a single Source Data file. Original material used for quantifications and to build figures will be deposited in a publicly accessible repository.