Selection of viral capsids and promoters affects the efficacy of rescue of Tmprss3-deficient cochlea

Adeno-associated virus (AAV)-mediated gene transfer has shown promise in rescuing mouse models of genetic hearing loss, but how viral capsid and promoter selection affects efficacy is poorly characterized. Here, we tested combinations of AAVs and promoters to deliver Tmprss3, mutations in which are associated with hearing loss in humans. Tmprss3tm1/tm1 mice display severe cochlear hair cell degeneration, loss of auditory brainstem responses, and delayed loss of spiral ganglion neurons. Under the ubiquitous CAG promoter and AAV-KP1 capsid, Tmprss3 overexpression caused striking cytotoxicity in vitro and in vivo and failed to rescue degeneration or dysfunction of the Tmprss3tm1/tm1 cochlea. Reducing the dosage or using AAV-DJ-CAG-Tmprss3 diminished cytotoxicity without rescue of the Tmprss3tm1/tm1 cochlea. Finally, the combination of AAV-KP1 capsid and the EF1α promoter prevented cytotoxicity and reduced hair cell degeneration, loss of spiral ganglion neurons, and improved hearing thresholds in Tmprss3tm1/tm1 mice. Together, our study illustrates toxicity of exogenous genes and factors governing rescue efficiency, and suggests that cochlear gene therapy likely requires precisely targeted transgene expression.


INTRODUCTION
Nearly 1 in 500 children in the United States is born with hearing loss, 65% of which are caused by genetic mutations. 1More than 70% of genetic hearing loss is attributed to autosomal recessive, nonsyndromic deafness (ARNSD) mutations, 2 up to 10% of which are caused by transmembrane protease serine 3 (TMPRSS3) mutations.3][4] As one of the most common causal genes of hearing loss among adult cochlear implant recipients, 5 TMPRSS3 mutations currently lack a biological treatment that prevents or reverses the course of disease.
The mammalian cochlea is composed of distinct sensory and nonsensory cell types that are all essential for auditory function (Fig- ure 1A).As mechanoreceptors, hair cells convert mechanical stimuli to electrical signals, are intercalated by supporting cells, and relay auditory input centrally via spiral ganglion neurons.Mice deficient in Tmprss3 (Y260X), where a nonsense mutation results in a truncated protease domain, exhibit normal cochlear development followed by rapid hair cell degeneration during the onset of hearing at postnatal day (P) 12. 6 This leads to a complete loss of auditory brainstem responses (ABRs), implicating a requirement of Tmprss3 for hair cell survival and cochlear function.
8][9]13 Both epithelial sodium channels (ENaCs) and calcium-activated potassium (BK) channels are candidate downstream targets of Tmprss3, [9][10][11][12][13] yet loss of function of ENaC does not cause hearing loss in humans, while deletion of BK channels leads to hearing loss only after 8 weeks of age in mice, well after the onset of hearing. 14,15At present, the exact function of Tmprss3 and the pathogenesis of DFNB8/10 are unclear.
In the inner ear, adeno-associated virus (AAV) capsids have demonstrated low immunogenicity and variable degrees of transduction efficacy in multiple cell types. 16][19][20][21][22][23][24] However, whether this approach is effective for mutations affecting multiple cochlear cell types and whether selection of viral capsids and promoters affects the overall efficacy of rescue are not known.
Here, we used recombinant AAV (rAAV) vectors in an attempt to restore Tmprss3 expression and function in Tmprss3 tm1 knockout mice (Tmprss3 tm1/tm1 ).Our studies demonstrate that Tmprss3 is required for hair cell survival and auditory function.During rescue experiments, we discovered that the engineered KP1 capsid was capable of transducing cochlear cells with high efficiency. 25However, when administering the Tmprss3 transgene under control of the ubiquitous promoter (CAG) 26 we observed cytotoxicity both in vitro and in vivo.Treatment with a selective viral capsid (AAV-DJ 27 ) and decreasing vector dose reduced toxicity but failed to prevent hair cell loss or auditory dysfunction.In contrast, by using the EF1a core promoter, 28 hair cell survival and auditory function were partially rescued in Tmprss3 tm1 knockout mice.Together, we have established that the input capsid, promoter, and vector dose dictate cytotoxicity and efficacy in rescuing the Tmprss3-deficient cochlea.

Tmprss3 is expressed in multiple cochlear cell types
To characterize Tmprss3 expression in the mouse cochlea, we performed RNAScope in situ hybridization.In the embryonic (E) day 18, postnatal 1-and 5-day-old (P1 and 5) wild-type cochlea, Tmprss3 mRNA was robustly expressed in the organ of Corti, including the inner hair cells, outer hair cells, and supporting cell subtypes (Figures 1A, S1A, and S1B).We did not observe a tonotopic gradient in expression.To a lesser extent, Tmprss3 transcripts were also detected in the greater epithelial ridge, lesser epithelial ridge, interdental cells, lateral cochlear wall, and select spiral ganglion neurons (Figure 1A).This is consistent with previous single-cell RNA sequencing and in situ hybridization data. 11,12,25,29We next assessed the effects of Tmprss3 deficiency by examining the Tmprss3 tm1 mouse line (Figure 1B), which is a knockout model generated through targeted mutation by homologous recombination in exon 1. 30 Using probes designed to detect the deleted sequences (BaseScope), we found that Tmprss3-exon1-2 transcripts were absent in the P1 Tmprss3 tm1/tm1 cochlea, whereas Tmprss3 transcripts remained detectable in different cell types in the P1 wild-type cochlea (Figures S1E and S1F).These results indicate that Tmprss3 mRNA expression is effectively abolished in the Tmprss3 tm1/tm1 cochlea.

Tmprss3 deficiency leads to hair cell degeneration and cochlear dysfunction
To determine whether Tmprss3 is required for cochlear maturation, we first examined cochleae in P5 Tmprss3 tm1/tm1 mice and found no evidence of hair cell or supporting cell loss with cell counts comparable to those of Tmprss3 tm1/+ and wild-type mice (Table S1).Similarly, each turn of the P12 and P13.5 Tmprss3 tm1/tm1 cochleae showed comparable counts and organization of sensory hair cells and supporting cells to Tmprss3 tm1/+ and wild-type cochleae (Figures 1D-1F, 1M, S1K, and S1L; Table S1), suggesting that Tmprss3 is not required for hair cell patterning or survival from P5 to 13.5.
Shortly after the onset of hearing around P14-14.5, Tmprss3 tm1/tm1 cochleae showed rapid and extensive degeneration of both inner and outer hair cells (Figures S1M-S1P), corroborating previous results in Tmprss3 (Y260X) mice. 6By P21, all Myosin7a + hair cells had significantly degenerated (p < 0.0001) in the Tmprss3 tm1/tm1 cochleae, whereas wild-type and Tmprss3 tm1/+ cochleae showed similar cell counts and organization (Figures 1G-1I and 1M; Table S1).As expected from the severe hair cell loss, P21 Tmprss3 tm1/tm1 mice exhibited no detectable ABR at any frequency tested (Figure 1C), whereas both wild-type and Tmprss3 tm1/+ mice displayed robust responses.These results demonstrate that Tmprss3 is required for hair cell survival and cochlear function after the second postnatal week.
Degeneration of supporting cells and spiral ganglion neurons in mature Tmprss3 tm1/tm1 cochleae In the juvenile and mature Tmprss3 tm1/tm1 cochlea, sensory hair cell loss was the prominent feature.Sox2 + supporting cell subtypes, which expressed Tmprss3 mRNA between E18.5 and P5, appeared disorganized in the P21 Tmprss3 tm1/tm1 cochlea, likely as a result of severe hair cell loss (Figures 1G-1I).However, we did not detect significant degeneration of supporting cells (Figure 1N; Table S1) or spiral ganglion neurons at this age (Figures 1O-1Q and 1U; Table S1).By P120, in addition to hair cell loss, there was a moderate and variable degree of supporting cell loss in the Tmprss3 tm1/tm1 cochleae (Figures 1J-1L (C) At P21, Tmprss3 tm1/+ littermates had ABR thresholds that were indistinguishable from WT littermates, whereas Tmprss3 tm1/tm1 mice demonstrated no ABR responses across all frequencies tested.(D-F) Immunostaining of P12 cochleae showed no loss or disorganization of hair cells and supporting cells among WT, Tmprss3 tm1/+ , and Tmprss3 tm1/tm1 littermates prior to the onset of hearing.(G-L) Substantial inner and outer hair cell loss and disorganized supporting cells were observed in the P21 and P120 Tmprss3 tm1/tm1 mice.No cell loss in WT or Tmprss3 tm1/+ cochleae.(M-N) Quantification in the middle cochlear turn showing significant loss of hair cells in P21 and P120 Tmprss3 tm1/tm1 cochleae and medial supporting cell loss at P120. (O-Q) Cross sections of Rosenthal's canal at P21 showing no spiral ganglia neuron degeneration.(R-T) At P120, there was a noticeable loss of spiral ganglion neurons in the Tmprss3 tm1/tm1 cochleae.(U-V) Quantitative analysis showing a significant loss of TuJ1 + spiral ganglion neurons, but not Sox2 + glial cells, in P120 Tmprss3 tm1/tm1 cochleae.Data shown as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.Two-way ANOVA with Tukey's multiple comparison.n = 3-6.IHC, inner hair cell; OHC, outer hair cell; DC, Deiters' cell; IPhC, inner phalangeal cell; IPC, inner pillar cell; OPC, outer pillar cell; SGN, spiral ganglion neuron.
In the Rosenthal canal, no degeneration of spiral ganglion neurons or glia was observed in P21 Tmprss3 tm1/tm1 mice (Figures 1O-1Q, 1U, and 1V; Table S1).However, there were significantly fewer Tuj1 + spiral ganglion neurons, but not Sox2 + glia, in each turn of the P120 Tmprss3 tm1/tm1 cochlea relative to age-matched Tmprss3 tm1/+ and wild-type controls (Figures 1R-1V; Table S1).Together, these results indicate that Tmprss3 deficiency causes delayed loss of cochlear supporting cells and spiral ganglion neurons in the adult cochlea.

AAV-KP1 transduces multiple cochlear cell types with high efficacy
As Tmprss3 is expressed in multiple cochlear cell types, we postulated that a gene delivery approach that is ubiquitous and efficient would be needed to rescue the phenotype caused by its deficiency.The AAV-KP1 capsid was obtained from a screen of a shuffled AAV capsid library on primary human islet cells and shown to have transduction efficacy across multiple murine and human cell lines comparable with or higher than the AAV-DJ capsid. 31,32However, neither capsid had been systematically evaluated for its ability to transduce inner ear cell types.To characterize the tropism and efficacy of both chimeric capsids in the cochlea, we generated rAAV vectors carrying the tdTomato reporter expressed under a CAG promoter and packaged them using KP1 and DJ capsids (AAV-KP1-CAG-tdTomato and AAV-DJ-CAG-tdTomato). Viral capsids were injected via a posterior semicircular canal approach into P1 pups (1 mL injected over 3 min, 1.0 Â 10 9 vector genomes [vg]) (Figure 2A).
After injection with AAV-DJ-CAG-tdTomato capsid at P1, P21 wildtype mice demonstrated tdTomato-labeled hair cells and supporting cells, albeit at lower rates than with the KP1 capsid (Figures S2E and  S2F).In the middle turn, the transduction rates were 65.6% ± 29.3% in inner hair cells, 30.0%± 19.5% outer hair cells, 54.7% ± 5.0% Deiters' cells, 48.7% ± 15.9% inner pillar cells, 52.3% ± 30.2% outer pillar cells, and 76.4% ± 25.0% inner phalangeal cells (Figure S2F; Table S2).Similar to saline-injected animals, those injected with rAAV packaged with either KP1 or DJ capsids exhibited no detectable ABR threshold shifts at P21 (Figure 2F and S2G).Together, these data indicate that both AAV-KP1 and AAV-DJ capsids transduce cochlear hair cells and supporting cells with no adverse effects on cell survival or cochlear function, with the former more efficiently transducing cochlear cells in vivo.Furthermore, these data show the temporal differences in tdTomato expression where there is a delay in the onset of KP1-CAG-tdTomato expression in hair cells compared to that of DJ-CAG-tdTomato expression.

Overexpression of Tmprss3 is cytotoxic in vitro and in vivo
To begin examining the effects of exogenous Tmprss3, we generated a KP1 capsid-packaged rAAV vector expressing the mouse Tmprss3 construct under the control of a CAG promoter.During rAAV production, detachment and death of producer cells (HEK293T/17) were noted, requiring a shortening of the incubation period prior to rAAV harvest (30 h instead of 60-72 h).Cell proliferation decreased with increasing multiplicity of infection (MOI) of AAV-KP1-CAG-Tmprss3 but not for a control rAAV-factor IX expression vector or for non-transduced cells (Figures 3A-3D).Similarly, dose-dependent cytotoxicity was found with rAAV-KP1-CAG-Tmprss3 using HeLa cells (Figures S3A-S3D).
Additionally, AAV-KP1-CAG-Tmprss3 injection resulted in elevated ABR thresholds that were significantly higher than those of saline-injected P21 wild-type animals (p < 0.01; Figure 3N) and non-injected ears (Figure S3N).Halving the titers lessened, and a 10-fold dilution prevented, ABR threshold shifts (Figures 3O-3P).However, AAV-KP1-CAG-Tmprss3 at full or reduced titers failed to rescue ABR thresholds in Tmprss3 tm1/tm1 mice (Figures 3N-3P).Together, these results indicate that exogenous Tmprss3 causes cytotoxicity in vitro and in vivo and that decreasing transduction reduced toxicity but failed to prevent Tmprss3 deficiency-induced hair cell loss and auditory dysfunction.

Exogenous Tmprss3 toxicity is associated with multiple AAV capsid types
To verify that cytotoxicity can be reduced by decreasing transduction rates, we also used AAV-DJ to overexpress Tmprss3 in vivo.AAV-DJ-CAG-Tmprss3 (2.0 Â 10 8 vg) was administered to P1 Tmprss3 tm1/tm1 and wild-type mice.In the injected P21 wild-type cochlea, no hair cell loss was detected (Figures 3Q, 3S, S4, and S4B), while a small but significant ABR threshold shift across several frequencies was observed, suggesting some cytotoxicity similar to the lower titers of AAV-KP1-CAG-Tmprss3 (p < 0.05; Figure 3T).Moreover, AAV-DJ-CAG-Tmprss3 administration failed to prevent hair cell loss or ABR threshold shifts in Tmprss3 tm1/tm1 mice (Figures 3R-3T, S4C, and S4D).Thus, Tmprss3-related cytotoxicity is likely dependent on transduction efficiency and can occur with both AAV-KP1 and AAV-DJ viral capsids.The presence of normal numbers of hair cells in DJ-CAG-Tmprss3-transduced wild-type cochleae as well as the presence of ABR thresholds, albeit significantly elevated, in these mice suggest that functional hair cells are present.Thus, viral transduction may be affecting supporting cells or cells outside of the organ of Corti.Additionally, using a capsid with lower transduction efficacy failed to prevent hair cell degeneration and auditory dysfunction caused by Tmprss3 deficiency.
Unlike AAV-KP1-CAG-Tmprss3, the AAV-KP1-EF1a-Tmprss3 construct did not diminish proliferation of 293T/17 cells at various MOIs, with rates similar to those of no-virus controls and those transduced with a control rAAV-factor IX prep (Figures S5C-S5F).Collectively, these results suggest that the use of the EF1a promoter abolished the cytotoxicity of exogenous Tmprss3 in vitro.We next administered the AAV-KP1-EF1a-Tmprss3 (6.5 Â 10 8 vg) vector into P1 wild-type mice (Figure 4D).At P21, we did not detect any degeneration of hair cells or supporting cells in any cochlear turns (Figures 4E-4H; Table S3).Collectively, these results suggest that the use of the EF1a promoter abolished the cytotoxicity of exogenous Tmprss3 in vitro and in vivo.
AAV-KP1-EF1a-Tmprss3 partially prevents cochlear degeneration and auditory dysfunction in Tmprss3 tm1/tm1 mice To assess its effects on hair cell degeneration and auditory dysfunction caused by Tmprss3 deficiency, we administered AAV-KP1-EF1a-Tmprss3 to P1 Tmprss3 tm1/tm1 mice (Figure 4D).At P7, BaseScope in situ hybridization detected robust expression of Tmprss3 transgene in all supporting cell subtypes and, to a lesser degree, inner and outer hair cells in the organ of Corti of AAV-KP1-EF1a-Tmprss3-injected Tmprss3 tm1/tm1 cochlea (Figures S6A-S6D 00 ).We observed preservation of both inner and outer hair cells in all three turns of the P21 Tmprss3 tm1/tm1 cochlea (Figures 5A-5C).Nearly complete hair cell survival in the middle and basal turns was observed, whereas, in the apical turn, survival was partial and variable, especially of the outer hair cells.Myo7a + cell counts in all three turns of treated Tmprss3 tm1/tm1 cochlea were significantly higher than those in contralateral, control cochlea (p < 0.001; Figure 5D; Table S3).Surprisingly, three of five Tmprss3 tm1/tm1 animals showed survival of some hair cells in the contralateral, control cochlea, suggesting a low level of transport of virus between ears (Figures S7A-S7D).
In AAV-KP1-CAG-Tmprss3-injected Tmprss3 tm1/tm1 cochlea, we detected robust expression of Tmprss3 transgene in both supporting cells and hair cells (Figures S6E-S6H 00 ).Relative to AAV-KP1-EF1a-Tmprss3-injected Tmprss3 tm1/tm1 cochlea, transgene appeared similarly expressed in supporting cells but higher among hair cells.How the level and pattern of transgene expression attributes to the lower cytotoxicity and ability of AAV-KP1-EF1a-Tmprss3 to prevent hair cell degeneration is currently unclear.
We further examined whether hair cell survival in P120-treated Tmprss3 tm1/tm1 cochleae persisted and whether degeneration of supporting cells and spiral ganglion neurons was also prevented.Similar to P21, sensory hair cells were consistently present in the middle and basal turns of the treated Tmprss3 tm1/tm1 cochlea, with survival being more variable in the apical turn.Treated Tmprss3 tm1/tm1 displayed significantly more hair cells than the contralateral, control cochlea (p < 0.01; Figures 5E-5H; Table S3).As seen at P21, one out of the five P120, Tmprss3 tm1/tm1 mice had a partial rescue of hair cells in the contralateral ear (Figures S7E-S7H).Both the counts and organization of Sox2 + supporting cells in the treated Tmprss3 tm1/tm1 cochlea were similar to those in the littermate and wild-type controls, while some disorganization was observed in the apex (Figures 5E-5G; Table S3).Furthermore, spiral ganglion neuron survival in the apical and middle turns was significantly higher in the treated Tmprss3 tm1/tm1 cochlea than in the contralateral cochlea (p < 0.01, Figures 5I-L; Table S3).
Finally, all five Tmprss3 tm1/tm1 mice injected with AAV-KP1-EF1a-Tmprss3 showed detectable ABR responses across a range of frequencies at P21 and P120, when none of the non-injected Tmprss3 tm1/tm1 mice or contralateral ears showed any responses at either age (Figures 5M-5O, S7M, S7N, and S7P; p < 0.0001).As expected, AAV-KP1-EF1a-Tmprss3 did not cause ABR threshold shifts in wild-type animals (Figure 5N).As a group, injected ears of Tmprss3 tm1/tm1 mice showed significantly higher thresholds than wild-type animals at P21 and P120 (Figures 5N, 5O, S7M, and S7P).Individually, each injected ear of Tmprss3 tm1/tm1 mice had ABR responses detected (Figures S7N and S7Q).However, only one of the three injected Tmprss3 tm1/tm1 ears showed detectable distortion product otoacoustic emission (DPOAE) responses, whereas none of the non-injected, mutant ears showed responses at P21 (Figure S7O).This suggests that outer hair cell function was not well restored even though most injected mutant ears showed improved survival of hair cells.In summary, these data indicate that treatment with AAV-KP1-EF1a-Tmprss3 partially prevents hair cell degeneration and cochlear dysfunction.

DISCUSSION
TMPRSS3 mutations cause DFNB8 and DFNB10 and are found in up to 11% of patients with sensorineural hearing loss. 4,36Previously, Guipponi and colleagues showed that Tmprss3 null mice exhibit severe degeneration of cochlear hair cells between P12 and 14. 8 Here, we found that Tmprss3 tm1 mutant mice also demonstrated extensive inner and outer hair cell loss around P14, 32 confirming that Tmprss3 is required for hair cell survival after the onset of hearing.Strikingly, exogenous Tmprss3 delivered via an rAAV vector using the KP1 capsid and the CAG promoter was toxic to multiple cell lines, caused hair cell loss in wild-type cochlea in vivo, and failed to prevent hair cell degeneration in Tmprss3 tm1/tm1 mice.Toxicity was diminished when using lower viral titers or by replacing the KP1 capsid with DJ, neither of which prevented hair cell degeneration in Tmprss3 tm1/tm1 mice.Finally, by using the EF1a promoter, exogenous Tmprss3 was no longer cytotoxic in vitro and in vivo, prevented hair cell degeneration, and partially restored auditory function in Tmprss3 tm1/tm1 mice.Recently, Du and colleagues reported similar findings of toxicity after overexpression of mouse Tmprss3 in the cochlea. 37prss3 is broadly expressed in the embryonic and neonatal cochlea, with expression spanning sensory hair cells, a subset of spiral ganglion neurons, and non-sensory cells in the cochlea.][10][11][12]25 Although TMPRSS3 is known to be a serine protease, Figure 5. AAV-KP1-EF1a-Tmprss3 partially prevents degeneration and auditory dysfunction in Tmprss3 tm1/tm1 mice (A-C) After AAV-KP1-EF1a-Tmprss3 injection (6.5 Â 10 8 vg) in P1 Tmprss3 tm1/tm1 mice, most IHCs and OHCs were present in the middle and basal turns, while most IHCs and some OHCs remained in the apical turn at P21. (A 0 -C 0 ) High-magnification images from (A)-(C).(D) Myo7a + cell counts in each turn of treated Tmprss3 tm1/tm1 cochlea were significantly higher than untreated Tmprss3 tm1/tm1 cochlea and were similar to WT controls.(E-G) Hair cell survival persisted in each turn of P120-treated Tmprss3 tm1/tm1 cochlea.(H) Each turn of P120-treated Tmprss3 tm1/tm1 cochlea displayed significantly more hair cells than untreated Tmprss3 tm1/tm1 cochlea and similar to WT controls.(I-K) Many SGNs were preserved in each turn of P120-treated Tmprss3 tm1/tm1 cochlea, particularly the middle and basal turns.(L) Treated Tmprss3 tm1/tm1 cochlea had higher SGN counts than untreated Tmprss3 tm1/tm1 cochlea.(M) Raw ABR waveforms of P21 WT, untreated, and treated Tmprss3 tm1/tm1 mice at 16kHz.(N and O).All treated Tmprss3 tm1/tm1 ears demonstrated detectable ABR responses at P21 and P120, whereas untreated ears showed no responses.Data shown as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.Two-way ANOVA with Tukey's multiple comparison.n = 3-5.
its targets in the inner ear and exact function remain elusive. 15In this and previous studies, extensive loss of hair cells in Tmprss3-deficient mice indicates that Tmprss3 is critical for their survival.
While we cannot rule out dysfunction of other Tmprss3-expressing cell types (e.g., supporting cells) in Tmprss3 tm1/tm1 mice, their survival is likely less dependent on Tmprss3, as we did not detect significant supporting cell degeneration.Moreover, spiral ganglion neuron degeneration observed at P120 is likely attributed to a loss of trophic support from hair cell loss rather than Tmprss3 deficiency.
In our study, high levels of exogenous Tmprss3 induced cytotoxicity in multiple cell lines in vitro and in the juvenile cochlea in vivo.9][40][41][42][43] Several type II transmembrane serine proteases similar to TMPRSS3 are implicated in the development and progression of different types of cancer. 44Mechanistically, Tmprss3 overexpression can decrease E-cadherin levels and thereby disrupt cell-cell adhesion, leading to tumor invasiveness and metastasis. 38,45Based on these studies, we postulate that Tmprss3 overexpression similarly disrupts cell-cell adhesion, causing hair cells to be susceptible to degeneration after, but not prior to, the onset of hearing.Our model system has established a foundation to further investigate this potential mechanism.
Our study reveals that AAV-KP1-EF1a-Tmprss3 was effective in partially preventing hair cell loss and improving auditory function in Tmprss3 tm1/tm1 mice, although outer hair cell function was not well restored.In situ hybridization showed that expression of Tmprss3 transgene was robust in supporting cells and remarkably lower in hair cells with the EF1a promoter.It is interesting that Tmprss3 transgene expression in hair cells appeared higher after treatment with AAV-KP1-CAG-Tmprss3 than AAV-KP1-EF1a-Tmprss3 with expression in supporting cells comparable between the two promoters.While AAV-KP1-EF1a-Tmprss3 partially rescued hearing function, AAV-KP1-CAG-Tmprss3 did not and moreover was cytotoxic to cochlear cells.One interpretation is that the Tmprss3 transgene is most critical for hair cell survival in the early postnatal cochlea, and the use of EF1a, and not CAG, promoter was able to restore Tmprss3 expression early.Alternatively, it is possible that hair cell survival does not directly depend on expression of Tmprss3 transgene within hair cells themselves but relies on an optimal level of expression in surrounding supporting cells, possibly via its effects on cell-cell junctions.Lastly, it is conceivable that a higher level of Tmprss3 transgene expression is needed to restore outer hair cell function.By increasing the titer or using a promoter (other than CAG) that drives Tmprss3 expression in outer hair cells, this may help to restore outer hair cell and overall cochlear function.Future work is necessary to delineate the function and downstream targets of Tmprss3 and the use of other promoters to drive Tmprss3 transgene expression in the cochlea.
9][50] The current study suggests that overexpression of Tmprss3 in supporting cells, rather than hair cells, may be more important for hair cell survival.However, this approach only partially prevents auditory function and fails to maintain outer hair cell function.While several studies have advocated the use of viral capsids with high transduction efficiency (e.g., AAV-ie, Anc80L65, and AAV2.7m8), [51][52][53] it is unclear whether (1) broad transduction is necessary to rescue all cells expressing the gene of interest, (2) ectopic expression of a gene of interest can have deleterious effects in the inner ear, and (3) the level of transgene expression affects the efficacy of rescue.Even though AAV-KP1 robustly transduces multiple sensory and non-sensory cell types in the cochlea at rates comparable to several other viral capsids, [51][52][53] AAV-KP1-CAG-Tmprss3 caused death of cochlear hair cells.Both decreasing viral titers and packaging with the less efficient AAV-DJ capsid reduced cell death and cochlear dysfunction.While these results suggest that cytotoxicity depends on the high transduction efficiency of cochlear cells, both approaches failed to rescue the function and cellular morphology of Tmprss3 tm1 mutant cochlea.By using an EF1a core promoter, we eliminated cytotoxicity in cell lines and in the cochlea in vivo and partially rescued hair cell degeneration and cochlear dysfunction.The findings that rescue of hair cell survival and auditory function was partial and not complete are likely multifactorial and may include differences in spatiotemporal expression between exogenous and endogenous Tmprss3 as well as differences in the level of expression in cells of interest (e.g., hair cells).Nevertheless, since AAV-EF1a-Tmprss3 led to sustained hair cell survival at least to P120, this approach is promising and should guide future studies to further optimize rescue of hearing loss caused by Tmprss3 deficiency.
While some studies employed the CAG promoter within their viral vector constructs to successfully prevent cochlear dysfunction and degeneration in mouse models of hearing loss, 22,46,54 this approach to drive Tmprss3 expression leads to cytotoxicity, especially when used with the broadly transducing viral capsid KP1.In the retina, the input dose, viral capsid, the encoded gene, the promoter driving transgene expression, and target cells have been demonstrated to govern cellular toxicity related to AAV administration. 55Moreover, broadly active promoters were found to be more toxic to the retinal pigment epithelium, and a weaker photoreceptor-specific promoter attenuated the toxicity. 56Here, there are several possible contributing factors to cytotoxicity of AAV-KP1-CAG-Tmprss3 in the cochlea, including ectopic transgene expression (e.g., stria vascularis, Reissner's membrane), higher-than-native expression (hair cells and supporting cells), and differences in temporal expression.It is noteworthy that CAG-Tmprss3 contained the woodchuck hepatitis virus posttranscriptional regulatory element (WPRE), while EF1a-Tmprss3 did not.As WPRE has been reported to increase transgene expression, 57 this sequence may further contribute to the toxicity observed.Lastly, Tmprss3 appears more highly expressed in hair cells after treatment with AAV-KP1-CAG-Tmprss3 than AAV-KP1-EF1a-Tmprss3, while that in supporting cells appears comparable between the two approaches.Thus, it is possible that the level of transgene expression in hair cells is more critical for hair cell function and in supporting cells for hair cell survival.The exact role of these factors warrants further investigation in future studies.
In summary, our results have demonstrated that the AAV-KP1 capsid has high transduction efficacy, but exogenous Tmprss3 under the CAG promoter led to cytotoxic effects in vitro and in vivo.Tmprss3-induced cytotoxicity was ameliorated by reducing transduction efficacy across cochlear cells using the AAV-DJ capsid and by using the EF1a core promoter.Collectively, our data indicate that precise spatial and temporal control of Tmprss3 expression is necessary for hair cell survival and cochlear function and further supports the need for a tailored approach to viral capsid and promoter selection to optimize gene therapy.These results may have important implications for the selection of viral capsids and promoters in human inner ear gene therapy.

Mouse genotyping
The Tmprss3 tm1/Lex mouse (MMRRC lab, stock # 032680, background C57/Bl6) strain was used.The Tmprss3 tm1/tm1 mouse was described as having an absence of a startle response to 120 dB (prepulse inhibition assay). 30Mice of both genders were used.Genomic DNA was prepared from mouse tail tips.The genomic DNA template was produced by adding 180 mL of 50 mM NaOH to tissue biopsies and incubating at 98 C for 1 h and then 15 C for 2 min.Next, 20 mL of 1 M Tris-HCl was added, and the samples were vortexed.The following primers were used: Tmprss3 mutant forward (Fwd) (5 0 GCA GCG CAT CGC CTT CTA TC), Tmprss3 mutant reverse (Rev) (5 0 CAG AGC CTT AAC TCT CCA CG), and Tmprss3 wild-type Fwd (5 0 TTC TAG GAC TTT GCT ATG ACC).All experiments were approved by the Institutional Animal Care and Use Committee (protocol #18606) at Stanford University.
In situ hybridization Previously published protocols were followed. 58,59Briefly, temporal bone tissues harvested from P1-P6 mice were fixed in 4% paraformaldehyde (in PBS, pH 7.4, Electron Microscopy Services) for 22 h at 4 C.The tissues were then cryoprotected using a serial sucrose gradient over 2 days starting at 15%, 20%, 30% sucrose solution and then gradually increasing the optimal cutting temperature compound (OCT) (Tissue Tek) gradient of 30% sucrose: 50% OCT, to 30:70, and finally to 100% OCT.Next, tissues were stored at À80 C until further use.The sections were cut at 10 mm thickness and placed on Superfrost Plus slides (Fisher).Tissue sections were hybridized with commercial probes from Advanced Cell Diagnostics (ACDbio) and counterstained with hematoxylin (Sigma-Aldrich) according to the manufacturer's instructions for fixed frozen sections with colorimetric detection.Briefly, sections were washed in PBS (1Â) for 5 min and then treated with H 2 O 2 for 10 min.Next, sections were permeabilized using target retrieval reagent (ACDbio) and proteinase before hybridization.RNAScope Red v2.5 kit (catalog #323350) was used with the following probes: DapB (catalog #310043), Polr2a (catalog #312471), and Sox2 (catalog #401041).BaseScope v2 Red kit (catalog #323900) was used with the following probes: BaseScope Probe BA-Mm-Tmprss3-E1E2 (catalog #716911), BA-Mm-Ppib-1zz (catalog # 712351), BA-Dapb-1zz (catalog #701021) (ACDbio).The BaseScope Tmprss3 probe was diluted 1:20 and BaseScope step Amp 7 was performed for only 10 min to reduce signal intensity.RNA Polymerase II (Polr2) and Peptidylprolyl Isomerase B (PPib) are ubiquitously expressed in the cochlea and were selected as the positive controls.DapB gene is expressed by the Bacillus subtilis strain SMY, a soil bacterium, and not in mammalian tissues and was selected as a negative control (Figures S1C, S1D, S6G, and S6H).Wild-type and mutant cochleae were processed in parallel, with sections collected on the same slide and subjected to mRNA detection under identical conditions.
For cell counting of whole-mount preparations, confocal images were analyzed using ImageJ software (NIH).Representative z stack images were taken on individual turns, and cells were counted from stacks and analyzed with ImageJ and Photoshop CS6 (Adobe Systems).For spiral ganglion neuron counting in sections, three sections per cochlea were used for quantification and averaged out (spiral ganglion neurons per 15,000 mm 2 ).

Assessment of hearing function
ABRs and DPOAEs were measured as previously described. 60riefly, mice were anesthetized (100 mg/kg ketamine and 10 mg/kg xylazine) and injected intraperitoneally.Three needle electrodes were placed as follows: one inferior to the tympanic bulla, referenced to an electrode on the vertex of the head, with a ground electrode placed in the hindlimb.Tone pip stimuli were delivered at frequencies ranging from 4 to 46 kHz (4.0, 5.7, 8.0, 11.3, 16.0, 22.3, 32, 46.1 kHz) up to an 80-dB sound pressure level (SPL) in 10-dB steps.In all, 512 trials at each frequency and intensity were conducted and averaged.Distortion product otoacoustic emissions were measured by a probe tip microphone placed in the auditory canal.The sound stimuli used to elicit the DPOAE were two 1-s sine wave tones of differing frequencies (F2/F1 ratio = 1.22).Frequencies ranged from 5.7 to 32 kHz and the two tones were stepped up from 20 to 80 dB SPL in 10-dB increments.The amplitude of the cubic distortion product was 2xF1-F2.The threshold was calculated as a DPOAE of two standard deviations above the noise floor for each frequency.For analysis of ABR and DPOAE, thresholds were manually scored in a blinded fashion by two individuals (K.A.A. and P.J.A.).A lack of a response was designated at the highest sound level, 80 dB SPL.

Vector constructs
The long isoform of Tmprss3 cDNA is 2,874 bp and is therefore within the packing capacity ($4.8 kbp) of rAAV vectors.The mouse Tmprss3 gene was obtained from Origene (#MC216545) and cloned into the rAAV CAG-FLuc vector 32 (plasmid #83281, Addgene) in place of the luciferase coding sequence using standard molecular biology techniques.The resulting construct (CAG-Tmprss3) contained AAV2 ITRs flanking the Tmprss3 sequence driven by a CAG promoter.A WPRE sequence between the coding sequence and the S40 late poly(A) signal was included to enhance expression.
rAAV production 293T/17 cells (ATCC #CRL-11268) were transfected with rep2-capKP1, 32 the respective ITR-containing rAAV vector plasmid and pAd5 using the CaPO4 transfection method, 32 or Transporter-5 transfection reagent (Polysciences, #26008).Virus was obtained from cell lysates 2 or 3 days post transfection and either purified using two rounds of CsCl centrifugation as previously described 32 or using the AAV Pro All Serotype purification kit (Takara, #6666) according to the manufacturer's instructions.Some rAAV preparations were further concentrated using Ultracel-100 spin columns (Millipore-Sigma, #UFC510008).Virus preparations were stored in aliquots at À80 C until use.
Viral genomes were isolated using the MinElute Virus Spin kit (Qiagen, #57704), and vector genome titers were determined using qPCR.For the Tmprss3-expressing rAAV preps, primers Tmp-qPCR-F (CACAGCAAGTACAAGCCAAAG) and Tmp-qPCR-R (GCTGGATGGT CTCGTCAAA) were used, while primer sets Td-qPCR-F (ATTACCTGGTGGAGTTCAAGAC) and Td-qPCR-R (GTCCTCGTTGTGTGAAGTGATA) were used to titer tdTomatoexpressing rAAV preps.Copy number standards consisting of linearized and serially diluted (10 8 -10/mL) plasmids were included on each qPCR plate.All AAVs used in this study were made in the same facility (Kay lab), with batches with higher titers diluted down to match those with lower titers.
In vitro transduction and transfection experiments 293T/17 cells or HeLa cells (ATCC #CCL-2) were seeded in 24-well plates, and, when they had reached a confluency of approximately 60%-70%, cells were transduced with rAAV at the MOI as indicated.Images were taken 3 days post transduction using a microscope with a built-in camera (Evos M5000, Invitrogen).

Cell proliferation assays
293T/17 cells or HeLa cells were seeded in 96-well plates at a density of 10 4 cells/well in 100 mL of medium and allowed to attach for 5 h.Cells were then transduced with rAAV diluted in 100 mL of medium in triplicate and assayed for proliferation at days 1, 2, 3, and 4 post transduction using the CellTiter 96 A queous One Solution assay according to the manufacturer's instructions (Promega, #G3581).Standard curves were obtained using nontransduced cells seeded at various densities.Cells that had been transduced with a huFIX-expressing rAAV packaged with the KP1 capsid as well as nontransduced cells were included as controls.

In vivo gene transfer
To perform gene transfer experiments in vivo, P1 pups were anesthetized and injected via the posterior semicircular canal (PSCC) technique as previously described. 62Briefly, injection was performed using beveled glass microinjection pipettes, which were pulled from capillary glass on a P-2000 pipette puller (Sutter Instruments).Pups were anesthetized by rapid induction of hypothermia for 3-4 min on ice until loss of consciousness, and this state was maintained on a cooling platform for 10-15 min during the surgery.The surgical site was disinfected by scrubbing with Betadine and wiping with 70% ethanol.A postauricular incision was made to expose the PSCC and penetrate the tip of the micropipette.A total volume of 1 mL of either virus was unilaterally introduced at a rate of 300 nL/ min into the left ear.The skin incision was closed using superglue.Body temperature was maintained on a 37 C warming pad for 30 min after surgery and before reintroduction into the parental cage.

Statistical analyses
Data were analyzed using Microsoft Excel (Microsoft) and GraphPad Prism (GraphPad).Two-tailed Student's t tests or analysis of variance with post hoc tests were used to calculate statistical significance.p < 0.05 was considered statistically significant.Data are shown as mean ± SD.For all experiments, n values represent the number of animals examined.

Figure 1 .
Figure 1.Tmprss3 deficiency causes cochlear hair cell degeneration and hearing loss (A) In situ hybridization (RNAScope) of P5 wild-type (WT) cochlea (middle turn and counterstained with hematoxylin) revealed Tmprss3 mRNA expression in hair cells and supporting cells, interdental cells, inner phalangeal cells, lesser epithelial ridge, outer sulcus, and Rosenthal's canal.Schematic depicting hair cell and supporting cell subtypes.(B) The TMPRSS3 protein consists of 453 amino acids, with a transmembrane (TM) domain, a low-density lipoprotein receptor class A (LDRA), a scavenger receptor cysteine-rich domain (SRCR), and a C-terminal serine protease.The mutation was generated by targeted mutation through homologous recombination in exon 1.(C)At P21, Tmprss3 tm1/+ littermates had ABR thresholds that were indistinguishable from WT littermates, whereas Tmprss3 tm1/tm1 mice demonstrated no ABR responses across all frequencies tested.(D-F) Immunostaining of P12 cochleae showed no loss or disorganization of hair cells and supporting cells among WT, Tmprss3 tm1/+ , and Tmprss3 tm1/tm1 littermates prior to the onset of hearing.(G-L) Substantial inner and outer hair cell loss and disorganized supporting cells were observed in the P21 and P120 Tmprss3 tm1/tm1 mice.No cell loss in WT or Tmprss3 tm1/+ cochleae.(M-N) Quantification in the middle cochlear turn showing significant loss of hair cells in P21 and P120 Tmprss3 tm1/tm1 cochleae and medial supporting cell loss at P120. (O-Q) Cross sections of Rosenthal's canal at P21 showing no spiral ganglia neuron degeneration.(R-T) At P120, there was a noticeable loss of spiral ganglion neurons in the Tmprss3 tm1/tm1 cochleae.(U-V) Quantitative analysis showing a significant loss of TuJ1 + spiral ganglion neurons, but not Sox2 + glial cells, in P120 Tmprss3 tm1/tm1 cochleae.Data shown as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001.Two-way ANOVA with Tukey's multiple comparison.n = 3-6.IHC, inner hair cell; OHC, outer hair cell; DC, Deiters' cell; IPhC, inner phalangeal cell; IPC, inner pillar cell; OPC, outer pillar cell; SGN, spiral ganglion neuron.

Figure 2 .
Figure 2. Tropism and efficacy of AAV-KP1-CAG-tdTomato in the cochlea (A) Schematic showing AAV injection at P1 in WT pups and examination at P6 and P21.(B) Robust tdTomato expression in supporting cells (bottom) but not hair cells (top) in the P6 cochlea (middle turn shown).(C) Quantification of labeled hair cells and supporting cells.(D) Robust tdTomato expression in both hair cells and supporting cells (top and bottom) at P21. (E) Quantitative analysis of tdTomato-labeled hair cells and supporting cell subtypes.(F) Both saline-and AAV-KP1-CAG-tdTomato-injected animals showed normal ABR thresholds at P21. (G) Only a subset of Tuj1 + spiral ganglion neurons were tdTomato labeled at P21. (H) Spiral ganglion neurons were partially transduced in all three cochlear turns, with the apex showing the highest rate.Data shown as mean ± SD. ***p < 0.001.Two-way ANOVA with Tukey's multiple comparison.n = 3-5.