In vivo genetic manipulation of inner ear connexin expression by bovine adeno-associated viral vectors

We have previously shown that in vitro transduction with bovine adeno–associated viral (BAAV) vectors restores connexin expression and rescues gap junction coupling in cochlear organotypic cultures from connexin–deficient mice that are models DFNB1 nonsyndromic hearing loss and deafness. The aims of this study were to manipulate inner ear connexin expression in vivo using BAAV vectors, and to identify the optimal route of vector delivery. Injection of a BAAV vector encoding a bacterial Cre recombinase via canalostomy in adult mice with floxed connexin 26 (Cx26) alleles promoted Cre/LoxP recombination, resulting in decreased Cx26 expression, decreased endocochlear potential, increased hearing thresholds, and extensive loss of outer hair cells. Injection of a BAAV vector encoding GFP-tagged Cx30 via canalostomy in P4 mice lacking connexin 30 (Cx30) promoted formation of Cx30 gap junctions at points of contacts between adjacent non-sensory cells of the cochlear sensory epithelium. Levels of exogenous Cx30 decayed over time, but were still detectable four weeks after canalostomy. Our results suggest that persistence of BAAV-mediated gene replacement in the cochlea is limited by the extensive remodeling of the organ of Corti throughout postnatal development and associated loss of non-sensory cells.


Results
Canalostomy is a viable and safe route for gene delivery to the mouse cochlea. To validate canalostomy as a potential route to target cochlear structures, we delivered bromophenol blue or fluorescent wheat germ agglutinin dissolved in PBS in the semicircular canal of P25 or P4 wild type mice (C57BL6/N), resulting in widespread distribution of the injected agent without visible damage to cochlear duct structures (Fig. 1a,b). In a subsequent set of experiments, we injected Dulbecco's Modified Eagle Medium/Nutrient Mixture F−12 (DMEM/ F12) via canalostomy in wild type mice at P25 (2.5 μl) or P4 (1.0 μl). Four weeks later, we assessed hearing performance by recording auditory brainstem responses (ABRs), which are electrical signals evoked from the brainstem following the presentation of sound stimuli 62 . We measured the IV wave thresholds of the ABR for click and tone burst stimuli of 8,14,20,26,32 kHz and found no differences in injected mice compared with non injected controls (Fig. 1c). These results indicate that delivery of DMEM/F12 to the inner ear via canalostomy causes no adverse effects on hearing and could function as an effective route of vector delivery.

In vivo transduction with BAAV vectors via canalostomy achieves widespread expression of a reporter gene in non-sensory cells of the mouse cochlear duct.
To evaluate the efficacy of in vivo transduction via canalostomy, we delivered a previously tested reporter gene vector, BAAVβ−actin−GFP 56,58 prepared in DMEM/F12, to the inner ear of P25 wild type mice. Four weeks later, we processed the cochlea for confocal immunofluorescence imaging (Fig. 2). No β−actin−GFP signal was detected in cochlear sensory hair cells, whereas diffuse expression was clearly visible in spiral limbus and spiral prominence, two structures populated by fibrocytes, as well as in non−sensory cells of the cochlear sensory epithelium. In particular, the majority of Hensen's and Claudius cells, some pillar cells and inner sulcus cells expressed β−actin−GFP. Extensive transgene expression was also evident in the lateral wall of the cochlea, namely in the spiral ligament, stria vascularis and supra-strial zone (Fig. 2).
Prior work in adult guinea pigs used cochleostomy as route of administration and reported a higher efficiency of expression of BAAVβ-actin-GFP when delivered into scala media compared to a scala tympani approach; however the scala media approach resulted in hair cells loss 58 . We replicated these experiments in mice injected at P25 and noted expression of the transgene not only in scala media but also in scala tympani and scala vestibuli, accompanied by an alteration of the cochlear structure and the Reissner's membrane, highlighting the limitation of this technique for clinical applications ( Figure S1).
Altogether, these experiments suggest canalostomy as a preferred route for BAAV-mediated transgene delivery to the mouse cochlea. SCIeNTIFIC RePoRts | 7: 6567 | DOI:10.1038/s41598-017-06759-y BAAV-driven Cre-Lox recombination abates Cx26 in the cochlea of adult Cx26 loxP/loxP mice. Inner ear connexins play a crucial developmental role and are essential for the maturation of sensory hair cells, despite the fact that hair cells do not express any connexin 63 . The conventional gene knockout approach is unsuitable for postnatal studies on Cx26 because homozygous knockout mice die in utero due to insufficient transplacental uptake of glucose 64 . Here, we investigated the role of Cx26 in the maintenance of sensory cells utilizing BAAV for the timed and localized knockout of Cx26 based on the Cre/loxP system 65, 66 using a canalostomy route of delivery. Cre-Lox recombination is an irreversible process which does not require sustained protein expression, as it needs to take place only once and requires a limited amount of Cre recombinase 65,66 . We engineered a BAAV vector encoding GFP-tagged bacterial Cre recombinase under the CMV promoter (BAAVCre-IRESGFP). We prepared the vector in DMEM/F12 and injected it via canalostomy to the inner ear of P25 Cx26 loxP/loxP mice. Confocal immunofluorescence imaging of cochlear midmodiolar sections obtained four weeks after canalostomy showed BAAVCre-IRESGFP caused a dramatic reduction of Cx26 immunofluorescence signals in the lateral wall of Cx26 loxP/loxP mice (n = 4) (Fig. 3a), but had no effect on tissue morphology, cell viability and connexin expression in the inner ear of wild type C57BL/6 N mice (n = 3) used as controls (Fig. 3b). Cre-mediated excision of Cx26 was comparatively less effective in the organ of Corti, as indicated by the presence of residual Cx26 expression ( Fig. 3c and d for control). q-PCR analysis of whole cochlea samples confirmed a 43.5 ± 2.2% overall reduction of Cx26 transcript levels with respect to the contra lateral non-injected ear (n = 4; p < 0.001, paired t test). Based on these immunofluorescence results, we conclude that the majority of Cx26 loss occurred in fibrocytes of the spiral ligament. Consistent with this conclusion, the endocochlear potential was significantly lower in Cx26 loxP/loxP mice injected with BAAVCre-IRESGFP at P25 (26 ± 10 mV, n = 3) compared to non-injected Cx26 loxP/loxP controls (106 ± 3 mV, n = 11; p < 0.001, ANOVA). Inner hair cells, pillar cells and Deiters' cells were maintained in all transversal sections, spanning the whole length of the cochlear duct. Instead, we observed a 92 ± 16% (n = 3) loss of outer hair cells ( Fig. 3e and f for control). In agreement with these results, we found remarkably increased hearing thresholds in Cx26 loxP/loxP mice injected with BAAVCre-IRESGFP at P25, whereas thresholds in C57BL/6 N mice injected at P25 with BAAVCre-IRESGFP were indistinguishable from non-injected Cx26 loxP/loxP controls (Fig. 4). Therefore, we conclude that: (1)   Cx26 loxP/loxP mice. Although this is useful to explore the function of inner ear connexins in hearing, restoration of connexin expression in DFNB1 mouse models is clearly of primary importance for future therapeutic applications.
Targeted ablation of Cx26 in the mouse inner ear (Cx26 Sox10Cre , Cx26 OtogCre ) leads to irreversible cell loss in the organ of Corti either before or around the onset of hearing 5,67 , which in mice occurs at P12 68 . A similar problem affects also Cx30 −/− mice, which show hearing loss at all frequencies accompanied by complete absence of endocochlear potential 69,70 . Therefore it is clear that any intervention must be scheduled before irreversible damage to the organ of Corti, due to lack of connexin expression, takes place. Previously, we succeeded in restoring connexin expression and rescued intercellular coupling in vitro, by transducing cochlear organotypic cultures from P5 Cx26 Sox10Cre or Cx30 −/− mice with BAAV vectors encoding respectively Cx26 or Cx30 4, 5 . Here, we used a BAAVCx30GFP vector 4 , which encodes GFP-tagged mouse Cx30 under the control of the CMV promoter, prepared in DMEM/F12, and injected it to P4 Cx30 -/mice via canalostomy. Auditory thresholds of injected mice, measured four weeks after surgery, remained super imposable to those of untreated Cx30 -/mice (p < 0.001, ANOVA) ( Figure S2a).
The hearing loss phenotype exhibited by Cx30 −/− mice depends on the cumulative effect of deletion of Cx30 and 3' insertion of the lacZ and neo genes 6 , which are associated with dramatically reduced Cx26 levels 4 . Previous work showed restoration of hearing and prevention of hair cell death in Cx30 −/− mice in which extra copies of the Cx26 gene were transgenically expressed from a modified bacterial artificial chromosome 71 . Therefore we injected Cx30 -/mice with a BAAV vector espressing Cx26CFP via canalostomy at P4, but also this treatment remained ineffective ( Figure S2a). q-PCR analyses using primers specific for GFP tag to monitor transgene expression at 2, 3, 6, and 30 days after canalostomy, showed that transcript levels of Cx26CFP were maximal 2 days after gene delivery and thereafter decreased continuously. Thirty days after canalostomy, transgene expression was only 15.5 ± 4.5% (p < 0.005, ANOVA) of its initial peak ( Figure S2b).
These negative outcomes might depend, at least in part, on disruption of the cytoarchitecture of the sensory epithelium and the stria vascularis in the Cx30 −/− mouse model 69,70 . Therefore, we decided to also test the Cx30 Δ/Δ Cx30 knock out strain. Despite being ubiquitously deprived of Cx30, hearing is normal in Cx30 Δ/Δ mice 6 and their sensory epithelium is well preserved ( Figure S3). We administered the BAAVCx30GFP vector prepared in DMEM/F12 to P4 Cx30 Δ/Δ pups via canalostomy, and four weeks later we measured ABR thresholds. Compared to non-injected age-matched Cx30 Δ/Δ controls, threshold values were comparable or even slightly lower in Cx30 Δ/Δ mice injected with BAAVCx30GFP (Fig. 5). Therefore, we conclude that delivery of exogenous Cx30 gene via BAAV and canalostomy does not cause any hearing loss.
The cochleae of these injected mice were studied by immunofluorescence four weeks after vector delivery via canalostomy. In order to avoid cross-reactivity of the anti Cx30 antibody with the residual Cx26 expressed by  Cx30 Δ/Δ mice, we used an anti-GFP antibody that selectively targets the transgene product delivered by the BAAV vector. Confocal imaging showed extensive gap junction plaques formed by the Cx30GFP fusion protein at points of contacts between adjacent non-sensory cells of the sensory epithelium (Figs 6 and 7). Cx30GFP signals were also detected in the supra-strial zone (Fig. 7); we speculate that BAAV spread by transcytosis 57 to accumulate in that zone traversing the lateral wall. Clearly, this viral fraction was lost as far as Cx30GFP expression in the critical sensory epithelium is concerned. Of note, all hair cells (both inner and outer) were preserved in this tissue after viral transduction and did not express Cx30GFP. These results indicate that recombinant Cx30GFP proteins traffic correctly to the plasma membrane of cochlear non-sensory cells not only in vitro 4 but also in vivo (Figs 6 and 7). However, Cx30GFP immunofluorescence signals in transduced cochlear epithelia were considerably weaker than the corresponding signals due to native Cx30 in age-matched non-injected wild type C57BL/6 N control mice (Figs 6 and 7). To address this issue, we quantified the viral genome copy number still present in the injected cochleae using primers specific for the CMV promoter. We found that the CMV signal was almost undetectable 6 days after transgene delivery via canalostomy (Fig. 8).

Discussion
Viral transduction performed in adult Gjb2 conditional knock out mice failed to correct hearing function 72,73 , whereas an early intervention in newborn mice produced limited Cx26 reinstatement and only partial rescue of hearing 73 . Here, we have assayed the potential of BAAV vectors encoding Cx30 to rescue protein expression in Cx30 KO mice in vivo by transduction at P4. To our knowledge, this is the first report of a clear and widespread expression of exogenous Cx30 (tagged with GFP) that persisted in cells of the cochlear sensory epithelium four weeks after virus delivery via canalostomy. However, Cx30GFP expression was visibly lower than native Cx30 in wild type controls. This is partly due to the spread of BAAV particles to type V fibrocytes present in the supra-strial zone 74 , probably via transcytosis 57 from scala media through the later wall, reducing the therapeutic index of the vector dose. Furthermore, we noted a decrease in BAAV genome level following transduction, which was almost undetectable already 6 days after delivery via canalostomy at P4. In contrast, BAAV-mediated gene delivery via canalostomy in wild type mice injected at P25 achieved extensive expression of a reporter gene (β−actin−GFP) that was maintained at high levels for at least one month.
Recent work with AAV serotypes indicates that restoration of hearing is within reach of cochlear gene therapy [53][54][55] . However, targeting the post-natal 3500 or so inner hair cells, or even the three to four-fold more numerous outer hair cells, is facilitated by the fact that both types of postnatal sensory cells are post-mitotically differentiated 75,76 and numerically stable from P0 onward [77][78][79][80] . Achieving early, stable and widespread expression of exogenous connexins in the far more abundant, heterogeneous and developmentally evolving populations of non-sensory cells in the sensory epithelium and lateral wall of the post-natal cochlea is more problematic. To better appreciate the inherent difficulties, one should consider that both sensory and non-sensory cells of the cochlear sensory epithelium originate from postmitotic progenitor cells, which undergo terminal mitosis between E12 and E14 75,76 . A single row of inner hair cells develop from the lateral margin of the greater epithelial ridge (GER), whereas three to four rows of outer hair cells, and lateral non-sensory cells, develop from the lesser epithelial ridge (LER) [77][78][79][80] . At birth (P0), differentiated hair cells are found from the basal to the apical cochlear turn and thereafter form a life-long stable population [75][76][77][78][79][80] , however postnatal development continues 81 . Programmed cell death [82][83][84][85] (type-I, or apoptosis, as well as type II, or autophagic cell death) plays a crucial role in tissue remodeling, specification of cell fate and differentiation [86][87][88] . Recently, autophagy-related genes Becn1, Atg4g, Atg5 and Atg9 were shown to be expressed in the mouse cochlea from late embryonic developmental stages (E18.5) to adulthood, and up-regulated as the postnatal inner ear gains functional maturity 89 . Whereas cytological changes that occur in the LER have not been studied extensively 75 , the abundant non-sensory cells that populate the GER are replaced by the far fewer cuboidal cells of the inner sulcus during the first two postnatal weeks [77][78][79][80] . Cells positive for activated caspase-3, one of the cysteine proteases that play essential roles in programmed cell death, were detected in the GER between P7 and P13, and appeared progressively along the cochlear duct from base to apex 90 . The GER persisted throughout all turns of the cochlea in 2-week-old mice lacking caspase-3, resulting in hyperplasia of supporting cells, degeneration of hair cells, and severe hearing loss, suggesting that caspase-3-dependent apoptosis is necessary for the development and formation of a properly functioning auditory system in mammals 91 . Administration of thyroid hormone (T3) to wild type mice on P0 and P1 advanced the overall program of apoptosis and remodeling by about 4 days, suggesting initiation of apoptosis by a receptor-mediated process in conjunction with other unknown signals 90,92 . Given the extensive remodeling and cell turn over, near 100% transduction of non-sensory cells will be required in therapeutic intervention for connexin related disorders.
Various laboratories reported different values for the time required to reach maximal transgene expression level after viral delivery to the developing inner ear 72,93 , possibly reflecting differences in the transduction pathway of the different AAV serotypes used in those studies. Recombinant AAV vectors are well known to remain mostly episomal and be lost quickly even after one round of cell replication 94 .
In summary, using three different viral vectors in wild type and transgenic mice, we show here that BAAV works efficiently as a gene manipulation tool and permits to control gene expression in non-sensory cells of the inner ear, in vivo. BAAV viral transduction provides efficient in vivo gene delivery to cochlear non−sensory cells via canalostomy, however rescue of DFNB1 phenotype requires early and stable expression of connexins, which is not afforded by the current generation of these viral vectors. Our results suggest that the persistence of BAAV-mediated gene replacement in the cochlea is limited by the extensive non-sensory cell loss, which occurs in the organ of Corti throughout postnatal development 88 . Future work is needed to improve vector production and reformulation that will allow higher concentrations of the BAAV vectors to achieve the near 100% transduction necessary in this application in the critical postnatal period that precedes acquisition of hearing in mice. Mice of either sex were used. Cx26 loxP/loxP mice 67 and Cx3 Δ/Δ mice 6 were maintained on a pure C57BL/6 N background. Cx26 loxP/loxP mice were genotyped by screening for the presence of the loxP insertions on extracted mouse tail tips using the following primers:
Genotyping of the Cx30 Δ allele was performed using the Gjb6F and a primer binding downstream of the third loxP site Gjb6ΔR: 5′-CCCACCATCAAGGTTGAACT-3′.
BAAV Production and Quantification. Hek293T cells, grown in DMEM/F12 supplemented with 5% FBS and 1% P/S, were transfected with three or four required plasmids (transgene vector, pAd12, and bovine adenoassociated virus (BAAV)-RepCap or AAV2-Rep plus BAAV-Cap). BAAVCx30GFP vector plasmids contained AAV-5 inverted terminal repeats (ITRs) whereas BAAVβ-actinGFP and BAAVCre-IRESGFP plasmids contained AAV-2 ITRs. Forty-eight hours after transduction, cells were harvested by scraping in TD buffer (140 mM NaCl, 5 mM KCl, 0.7 mM K 2 HPO 4 , 25 mM Tris-HCl pH 7.4) and the cell pellet was concentrated by low-speed centrifugation. Cells were lysed in TD buffer containing 0.5% deoxycholate and 100 U/ml DNase (Benzonase, Sigma) and incubated for 30 min at 37 °C. Following 10 min low speed centrifugation, viral particles were purified by CsCl gradient and dialyzed in DMEM/F12 (vehicle) with 10 K MWKO dialysis cassette (Pierce, Cat. No. 66383). Biological activity was confirmed on packaging cells 95 . Particle titers, determined by q-PCR, were in the range of 10 12 −10 13 particles/ml. For viral titration, a dilution of the viral preparation was added to a q-PCR reaction mixture containing 1× SYBR Green Master Mix (Applied Biosystems/Applera, Milan, Italy) and 0.25 pmol/μl forward and reverse primers. Amplification was measured using a sequence detector (ABI 7700, Applied Biosystems). Specific primers for CMV were designed with the Primer Express program (Applied Biosystems): CMV f 5′-CATCTACGTATTAGTCATCGCTATTACCAT-3′, CMV r 5′-TGGAAATCCCCGTGAGTCA-3′. Canalostomy. For canalostomy, as well as all other surgical procedures described in this article, mouse body temperature was kept at 38 °C by a feedback-controlled heating pad. P25 mice were anaesthetized with an intraperitoneal injection of zolazepam (25 mg/g) and xylazine (10 mg/g) whereas P4 pups were anaesthetized with xylazine 0.45 µg/g and zolazepam 0.15 µg/g diluted in physiological solution. Supplemental doses were administered as needed. After induction of anesthesia, mice were placed under a dissection microscope and the posterior (P25) or lateral (P4) semicircular canal was exposed by a dorsal post-auricular approach. For P25 mice, a hole was made with the tip of a 33 G needle, softly removing a part of the bony shell of the canal. Bromophenol-blue (SIGMA, product number B−5525), Alexa594 ® conjugated wheat germ agglutinin (10 μg/ml; ThermoFisher, catalogue # W11262) dissolved in phosphate buffered saline (PBS) or viral solution diluted in DMEM/F12 (vehicle; Gibco ® , catalogue # 11320074) were injected into the endolymphatic space (2.5 μl injected volume, 3 nl/s injection speed) with a micropump-controlled micro syringe (WPI, art.no. NANOFIL-100) equipped with a 36 G needle (WPI, art.no. NF36BV-2). For every transduction experiment, we injected ~10 9 viral particles. To avoid fluid leakage during injection, the needle inserted in the semicircular canal of P25 mice was sealed with a drop of dental cement (Temrex Interface Light Cured Cavity Liner and Base, Product n. 7100), which was rapidly cured with blue light from a LED source (Mini Led, Satelec, F02641). Ten minutes after injection, the needle was slowly removed and the hole was closed with dental cement. To test the efficacy of delivery (Fig. 1), animals were sacrificed 10 minutes after injection and the excised cochlea was examined by light and confocal microscopy.
In vivo electrophysiology. Endochlear potential was measured 30 days after surgery both in injected and contralateral ear. Mice were anaesthetized with 0.01 ml/g body weight of 20% urethane (SIGMA, product number 94300) and the potential difference between the scala media and a reference silver/silver chloride pellet under the dorsal skin was recorded 96 .