Posterior Semicircular Canal Gene Delivery in the Adult Mammalian Inner Ear


 Hearing loss is a common disability affecting the world’s population today. While several studies have shown that inner ear gene therapy can be successfully applied to mouse models of hereditary hearing loss to improve hearing, most of these studies rely on inner ear gene delivery in the neonatal age, when mouse inner ear has not fully developed. However, the human inner ear is fully developed at birth. Therefore, in order for inner ear gene therapy to be successfully applied in patients with hearing loss, one must demonstrate that gene delivery can be safely and reliably performed in the mature mammalian inner ear. The posterior semicircular canal approach has been shown to be an effective gene delivery method in the neonatal mouse inner ear. In this study, we examine the steps involved in posterior semicircular canal gene delivery in the adult mouse inner ear. We observe that the adult mouse inner ear is more susceptible to surgical trauma. We also find that the duration of perilymphatic leakage and injection rate have a significant effect on the post-surgical hearing outcome. Our results show that AAV2.7m8 is capable of transducing the adult mouse inner and outer hair cells with high efficiency.


Introduction
Hearing loss is a common disease process affecting the world's population today. Approximately 3 in every 1000 newborns are affected by hearing loss every year 1 . Over the past few years, several studies have shown that inner ear gene therapy is effective at improving the auditory function in mouse models of hereditary hearing loss 2 . In most of these studies, gene delivery is done in the neonatal age (< P5). One major difference between human and mouse ears is the fact that the auditory system is fully mature at birth in humans, whereas the onset of hearing is not until ~ P12 in mice 3 . The mouse inner ear is immature at birth and continues to undergo development after birth 3,4 . Therefore, in order for inner ear gene therapy to be successfully translated to patients with hereditary hearing loss, one needs to demonstrate that gene therapy can be effective when delivered to the mature mammalian inner ear. In addition, one also needs to identify viral vectors which can successfully transduce target cells in the mature mammalian inner ear.
Various gene delivery methods have been examined for delivering gene therapy to the inner ear in animal models of hearing loss 5 . Three delivery methods (cochleostomy, round window injection, and canalostomy) are commonly used in mice 6 . Cochleostomy allows for transgene delivery directly into the scala media, where the mechanosensory hair cells in the cochlea are located. Even though one study showed no hearing loss in adult mice with this approach 7 , other studies have shown that this surgical approach causes signi cant hearing loss, likely due to the trauma incited by drilling through the lateral wall of the cochlea [8][9][10] . Round window injection is another method for administering gene therapy into the inner ear. The round window is a membranous structure at the base of the cochlea which separates the middle ear and the cochlea. It can be accessed via the middle ear after the opening of the tympanic bulla. Despite being less invasive than cochleostomy, it can lead to middle ear effusion, which negatively affects hearing temporarily 11 . In addition, the transduction e ciency of round window injection is not evenly distributed throughout the cochlear turns, with a lower transduction rate in the apical turn, which is further away from the injection site (round window) 12,13 .
The canalostomy approach involves gene delivery through one of the semicircular canals located super cially in the temporal bone. It does not require opening the tympanic bulla, which minimized the chances for surgical trauma and middle ear effusion [14][15][16] . In rodents, the posterior semicircular canal (PSC) is the most prominent and easily accessible out of the three semicircular canals. Therefore, PSC approach is a commonly used surgical method for inner ear gene therapy studies in mice 17,18 . In adult mice, additional surgical aspects need to be considered when using this method, since the mouse otic capsule is initially cartilaginous and becomes ossi ed by postnatal week 2 4,19 , which potentially makes PSC gene delivery in adult mouse inner ear more challenging technically.
In this study, we examine the surgical steps involved in the PSC approach in order to re ne this surgical technique for safe and reliable gene delivery in the adult mouse inner ear. We nd that the adult mouse inner ear is more susceptible to surgical trauma compared to the neonatal mouse inner ear. In addition, we nd that the duration of perilymphatic leakage and injection rate have signi cant effect on hearing in the adult mouse inner ear. We also show that the synthetic AAV2.7m8 is capable of transducing the adult mouse inner and outer hair cells with high e ciency.

Results
Adult mouse inner ear is more susceptible to hearing loss than neonatal inner ear for inner ear gene delivery Posterior semicircular canal approach (PSC) has been shown to be a safe and effective surgical approach for inner ear gene delivery in the neonatal mouse inner ear 17,20 . In addition, some studies have also shown that it can be safely implemented in the adult mouse inner ear 15,21 . However, our initial attempts at using the PSC approach for gene delivery in the adult mouse inner ear showed signi cant ABR threshold elevation in many mice (Fig. 1a). Therefore, we decided to investigate the implementation of PSC approach in adult mice more carefully. The three main surgical steps in the PSC approach are, 1) fenestration of PSC, 2) insertion of injection tubing into the PSC, and 3) injection of uid into the PSC (Fig. 1b). We decided to examine each of these surgical steps involved with PSC gene delivery to see if we could re ne this surgical technique to minimize trauma to the adult mouse inner ear.
The duration of perilymphatic leakage after PSC fenestration negatively affects hearing in adult mouse inner ear We rst examined the effect of PSC fenestration on the adult mouse inner ear. The fenestration of PSC is performed using a small 27-gauge needle to expose the canal lumen. Observation of perilymphatic leakage is used as con rmation for successful access to the PSC lumen. In this experiment, we performed PSC fenestration on adult CBA/J mice. The PSC fenestration was left open for several minutes to allow for perilymphatic leakage to occur, and then sealed off using a muscle plug. We found that some mice developed signi cant hearing loss after PSC fenestration while others didn't ( Fig. 2a). We decided to examine the PSC fenestration more closely by timing the duration of PSC opening and perilymphatic leakage. We separated our animals into three groups based on various durations of PSC opening and perilymphatic leakage: 2, 5, and 10 minutes. We found that mice with 2-minute and 5-minute PSC opening had minimal ABR threshold elevation compared with non-surgery control mice. However, mice in the 10-minute PSC opening group had signi cant ABR threshold elevation compared to the non-surgery control mice ( Fig. 2b; p=0.0153 for 4kHz, p=0.0006 for 8kHz, p=0.0082 for 16kHz, and p=0.0010 for 32kHz, t-test). This indicates that prolonged PSC opening time and perilymphatic leakage can adversely affect the hearing outcome in adult mouse inner ear.
Tube insertion into the PSC had no effect on hearing in the adult mouse inner ear Next, we compared ABR thresholds in mice with or without insertion of injection tubing into the PSC to determine whether this surgical step would adversely affect auditory function. It is important to remember that mice undergoing tube insertion will have to undergo PSC fenestration. Therefore, this surgical step cannot be evaluated on its own, but must be evaluated after the PSC fenestration has been created. The PSC opening time was kept below 5 minutes to minimize perilymphatic leakage. We found that there was no signi cant difference in ABR thresholds between mice that underwent tube insertion and non-surgery control mice when the PSC opening was kept below 5 minutes (Fig. 3).
The rate of injection has signi cant effect on hearing in the adult mouse inner ear Lastly, we evaluated the effect of injection rate in the adult mouse inner ear via the PSC. Again, it is important to remember that in order for mice to be injected with gene therapy, the PSC must be fenestrated rst, and then the injection tubing must be inserted into the PSC lumen in order for the injection to take place. In neonatal mice, we have shown previously that we could deliver approximately 1 µl of uid volume within a span of 30 seconds into the inner ear without any ABR threshold elevation compared to non-surgery control mice 18 . However, when we used the same injection rate in adult mice, signi cant ABR threshold elevation was observed (Fig. 4a). Substantial IHC and OHC loss was found throughout the cochlear turns (Fig. 4b). In the basal turn of the cochlea, all IHCs and OHCs were damaged, suggesting the hearing loss observed resulted from hair cell damage after injection (Fig. 4c). Therefore, we decided to examine the effect of injection rate in the adult mouse inner ear more carefully. The micro-injector that we use allows us to set the uid volume per injection (e.g. 13.8 nl, 27.6 nl, 46 nl, etc.), and the injection interval can be spaced out as determined by the investigators. We assessed the following three different injection regimens: 72 injections of 13.8 nl per injection every 10 sec (for a total volume of 993.6 nl), 36 injections of 27.6 nl per injection every 10 sec (for a total volume of 993.6 nl), and 20 injections of 46 nl per injection every 10 sec (for a total volume of 920 nl). We found that there was no signi cant difference in the average ABR thresholds between mice in the 13.8 nl per injection and 27.6 nl per injection groups compared to non-surgery control mice (Fig. 4d). However, mice in the 46 nl per injection group exhibited signi cantly higher ABR thresholds compared to non-surgery control mice (Fig.  4d). The differences in ABR threshold were signi cant at all tested frequencies except 4kHz ( p=0.0641 for 4kHz, p=0.0027 for 8kHz, p=0.0258 for 16kHz, and p=0.0318 for 32kHz, t-test).

AAV2.7m8 transduced adult cochlear hair cells with high e ciency
We previously showed that AAV2.7m8 is a powerful viral vector for gene delivery in the neonatal mouse inner ear 18 . However, it has been shown that AAV transduction e ciency can be different between neonatal and adult mouse inner ears 16,21,22 . Therefore, we assessed the transduction pattern and e ciency of AAV2.7m8 in the adult mouse inner ear using our newly re ned PSC approach. When 1 µl of AAV2.7m8 was delivered via PSC approach by 72 injections of 13.8 nl per injection every 10 sec, IHC and OHC transduction rates were 65.3±10.1 and 37.9±7.4% in the apical turn, 69.2±10.8 and 35.2±7.7% in the middle turn, and 40.3±7.8 and 10.8±6.3% in the basal turn of the cochlea (Fig. 5a). Even though the overall transduction rate is lower compared to our previous study in neonatal mouse inner ears, there were some adult mice that had very high rates of IHC and OHC transduction, comparable to neonatal ears. The reduction in overall IHC and OHC transduction rates in the adult mouse inner ear is likely due to the increased technical challenge with adult mouse inner ear gene delivery compared to neonatal ears.
To determine whether overall injection volume affects transduction e ciency, we injected some mice with 2 µl of AAV2.7m8-GFP using 72 injections of 27.6 nl per injection every 10 sec (Fig. 5b). IHC and OHC transduction e ciencies with 2 µl were 91.2±0.9% and 86.3±4.3% in the apical turn, 91.0±2.5% and 56.9±12.7% in the middle turn, and 83.7±6.3% and 21.6±9.8% in the basal turn of the cochlea (Fig. 5c). When compared with 1 µl injection, the overall transduction rate was higher across the cochlear turns, and the difference in transduction rate was signi cant in the basal turn of the cochlea for IHCs, and the apical turn of the cochlea for OHCs. This indicates that AAV2.7m8 is capable of transducing cochlear IHCs and OHCs at high levels, and the transduction e ciency in adult mice increases in a dose-dependent manner.
Hearing is preserved in adult mouse inner ear after 2 µl injection using the PSC approach Even though AAV2.7m8 is capable of transducing cochlear IHCs and OHCs at high levels in the adult mouse inner ear, we had to increase the total injection volume to 2 µl in order to match the transduction e ciency seen in the neonatal mice. Since the adult mouse inner ear is more vulnerable to surgical manipulation and injection volume, we assessed whether a 2 µl injection would have any effect on auditory function in these animals. We found that neither 1 µl nor 2 µl injection volume caused signi cant ABR threshold shift compared to non-surgery control mice, as long as the injection parameters were kept below 27.6 nl every 10 seconds. We also found that the average ABR threshold was not signi cantly different between mice injected with 1 µl and 2 µl (Fig. 6). These results demonstrate that up to 2 µl of uid volume can be safely injected into the adult mouse inner ear using the PSC approach without causing any signi cant ABR threshold elevation.

Discussion
Inner ear gene therapy has been shown to be effective at improving the auditory function of several mouse models of hereditary hearing loss 2,23−28 . While these proof-of-concept studies are very promising, most of these studies require gene delivery to be performed at the neonatal age (before P5), when the mouse inner ear is still not fully mature. In contrast, the human inner ear begins to have auditory perception by ~ 19-week gestation and is fully developed at birth 29 . Therefore, in order to increase the probability of success in translating inner ear gene therapy from mouse models of hearing loss to patients with hearing loss, one must try to demonstrate successful gene delivery in the mature mammalian inner ear. While PSC approach is a well-established surgical approach for gene delivery in neonatal mouse inner ear, few studies have examined the impact of PSC approach on hearing outcome in the adult mouse inner ear 14,16,21 . Even though the anatomy between the neonatal and adult mouse inner ear is similar, one major difference is the fact that the adult otic capsule is completely ossi ed, whereas the neonatal otic capsule is still cartilaginous 4,19 . This difference makes accessing the adult inner ear much more challenging, and potentially more traumatic. Indeed, we found that adult mouse inner ear is more susceptible to hearing loss after PSC gene delivery compared to neonatal ears 18 . Therefore, we decided to see if we could re ne the surgical techniques of PSC gene delivery in the adult mouse inner ear to minimize hearing loss. We examined the three main surgical steps involved with PSC gene delivery: 1) fenestration of PSC, 2) insertion of injection tubing into the PSC, and 3) injection of uid into the PSC. We found that prolonged perilymphatic leakage and injection rate have signi cant effects on hearing in the adult mouse inner ear.
Fluid leakage from the fenestra is a good indicator of having obtained access to the perilymphatic space during canalostomy approach. In order to decrease the uid volume and pressure in the perilymphatic space to accommodate for gene therapy injection, Suzuki et al. recommended waiting for 5 minutes after the fenestra of the PSC wall is opened to allow perilymph to leak out 21 . Similarly, in a study by Yoshimura et al., they recommended fenestrating the PSC when performing round window gene delivery to allow perilymphatic leakage in order to decrease the pressure in the inner ear 13 . In the present study, we found that a major factor for hearing loss in adult mice undergoing PSC gene delivery is perilymphatic leakage.
We found that mice with a fenestra opening time of 10 minutes prior to securing the injection tubing exhibited signi cantly higher ABR thresholds compared to non-surgery control mice (Fig. 2b). The concept of minimizing the duration of time for perilymphatic leakage is well known in the otologic surgery literature. In cholesteatoma surgery, when the cholesteatoma has eroded through the otic capsule (most commonly involving the horizontal semicircular canal), it is generally recommended to leave the cholesteatoma matrix on the perilymphatic stula to avoid perilymphatic leakage and exposure of the inner ear 30,31 . In cases where a decision is made to open the eroded horizontal semicircular canal wall, it is usually recommended to seal the stula immediately in order to minimize perilymphatic leakage and preserve inner ear function 32,33 . Therefore, our data suggest that it is important to minimize the leakage of perilymph after the fenestra on the posterior semicircular canal is opened in order to minimize hearing loss. We recommend trying to insert and secure the injection tubing as soon as the fenestra on the posterior semicircular canal is opened.
The injection uid into the inner ear could potentially cause barotrauma and mechanical trauma to the inner ear. In mice injected with 1 µl of viral vectors, a larger volume of uid per injection was signi cantly associated with ABR threshold shift (Fig. 4c). In addition, mice that received fast injection rate (20 injections in 30 seconds) exhibited signi cant hair cell loss and ABR threshold elevation (Fig. 4a). Both large uid volume per injection and shortened duration between injections can induce large and fast displacement of perilymph, which may lead to increased pressure within the cochlea, causing damage. Clinical experience with hearing preservation cochlear implant surgery supports our ndings. Thick diameter of cochlear implant array is analogous to large volume per injection in this study. It has been shown that larger diameter cochlear implant array leads to higher insertion force and results in increased risk of loss of residual hearing during cochlear implant surgery 34 . Similarly, an increase in implant insertion speed is analogous to injection speed (the time interval between injections) in the present study. It has also been shown that increased cochlear implant insertion speed can cause adverse effects on residual hearing during cochlear implant surgery 35,36 . Therefore, our data suggest that a smaller volume of uid per injection with slower injection speed (longer time interval between injections) offers the best chance for hearing preservation in the adult mouse inner ear with PSC gene delivery.
In our previous study, we showed that AAV2.7m8 is capable of transducing neonatal cochlear IHCs and OHCs at high levels 18 . In this study, we tested the transduction e ciency of AAV2.7m8 in the adult mouse cochlea. We found that AAV2.7m8 was also capable of transducing the adult cochlear IHCs and OHCs, but the overall transduction rate was lower than what we observed in the neonatal inner ears 18 . The decrease in transduction e ciency in the adult mouse inner ear has been reported in other studies 9,37,38 . In a study comparing the transduction e ciency of several AAV serotypes between neonatal and adult mouse inner ears, Shu et al. found that the viral transduction e ciency in the adult mouse inner ear was signi cantly lower than the neonatal mouse inner ear 9 . In another study, when exogenous Tmc1 gene was delivered into the inner ear of Tmc1 de cient mice using Anc80L65, infected hair cell rates decreased as a function of injection age from 93% at P1 to 3% at P14 38 . A recent study in which AAV9-PHP.B was used as the viral vector for inner ear gene delivery in a mouse model of Usher syndrome type 3A, the authors observed that AAV9-PHP.B was able to transduce both OHCs and IHCs in neonate mice (P0-P1), while adult mice (P28) exhibited transduction in only IHCs 37 . Although this decrease may be due to the maturation of cellular architecture of the inner ear that prevents the diffusion of AAV to infect hair cell 9 , the exact mechanism involved is still unclear. Therefore, further study is needed to better understand the mechanism behind this phenomenon.
We observed that the transduction rates of IHCs and OHCs increased in a dose-dependent manner with AAV2.7m8, with no signi cant ABR threshold shift up to a total injection volume of 2 µl. The average volume of perilymphatic space in mice is 0.62-1.72 µl 39,40 . Therefore, it is interesting that an injection volume of 2 µl did not negatively impact auditory function. In addition to our results, other studies have also shown that up to 2 µl of uid volume can be delivered into the mouse inner ear without any signi cant effect on the auditory function 14,41−43 . The ability of the mouse inner ear to accommodate a large uid volume may be explained by the presence of a relatively large and patent cochlear aqueduct, which communicates between the perilymphatic and subarachnoid spaces 18,44 . The cochlear aqueduct essentially acts as an out ow valve which allows excess uid volume to escape into the subarachnoid space. The fact that up to 2 µl of uid volume can be safely delivered to the adult mouse inner ear is particularly useful in gene therapy studies requiring the use of multiple viral vectors (e.g. dual-AAV approaches for delivering large cDNA), since these studies often require larger injection volume.
In conclusion, our study shows that although the adult mouse inner ear is more susceptible to surgical manipulation, the auditory function can be preserved by reducing the PSC opening time to minimize perilymphatic leakage, as well as utilizing a slower uid injection rate. In addition, the synthetic AAV2.7m8 is capable of transducing adult mouse IHCs and OHCs with high e ciency. It is our hope that the detailed methods described in this study can be utilized for safe and e cient gene delivery in the adult mouse inner ear.

AAV vector construction
The AAV2.7m8-CAG-EGFP (9.75 x 10 12 GC/mL) was produced by the Research Vector Core at the Center for Advanced Retinal and Ocular Therapeutics (University of Pennsylvania). The production method for these viruses have been previously described 45 . Animal surgery Animal surgery was approved by the Animal Care and Use Committee at the National Institute on Deafness and Other Communication Disorders (NIDCD ASP1378-18). All animal procedures were done in compliance with the ethical guidelines and regulations set forth by the Animal Care and Use Committee at NIDCD. The study was performed in compliance with the ARRIVE guidelines for animals. Adult (P30-90) CBA/J mice were used in this study. Anesthesia was induced using iso urane gas (Baxter, Deer eld, IL) through a nose cone at a ow rate of 0.5 L/min. Gene delivery was done using the PSC approach. A post-auricular incision was made using small scissors. The soft tissues were bluntly dissected to expose the PSC. To expose lumen of the ossi ed canal, a 27-guage hypodermic needle was used. A small hole was created by rotating the needle with gentle pressure. After creating a hole, perilymph leakage from the PSC was identi ed and the hole was kept open for 2, 5 or 10 minutes to determine whether the opening time affects post-injection hearing results. A Nanoliter Microinjection System (Nanoliter2000, World Precision Instruments, Sarasota, FL) was used in conjunction with a polyethylene tube attached with glass micropipette to load viral vector. AAV2.7m8-CAG-EGFP (9.75 x 10 12 GC/mL) was injected according to following different volume and rates to de ne best option for uid injection: 13.8 nl × 72 injections every 10 seconds, 27.6 nl × 36 injections every 10 seconds, 46 nl × 20 injections every 10 seconds, 46 nl × 20 injections in 30 seconds total, and 27.6 nl × 72 injections every 10 seconds. Incision was closed with 5-0 vicryl sutures.

Auditory brainstem response
Auditory brainstem response (ABR) testing was used to evaluate hearing sensitivity at ~P30. Animals were anesthetized with ketamine (100 mg/kg) and dexmedetomidine (0.375 mg/kg) via intraperitoneal injections and placed on a warming pad inside a sound booth (ETS-Lindgren Acoustic Systems, Cedar Park, TX). The animal's temperature was maintained using a closed feedback loop and monitored using a rectal probe (CWE Incorporated, TC-1000, Ardmore, PN). Sub-dermal needle electrodes were inserted at the vertex (+) and test-ear mastoid (-) with a ground electrode under the contralateral ear. Stimulus generation and ABR recordings were completed using Tucker Davis Technologies hardware (RZ6 Multi I/O Processor, Tucker-Davis Technologies, Gainesville, FL, USA) and software (BioSigRx, v.5.1). Click and tone-burst ABR thresholds were measured at 4, 8, 16, and 32 kHz using 3-ms, Blackman-gated tone pips presented at 29.9/sec with alternating stimulus polarity. At each stimulus level, 512-1024 responses were averaged. Thresholds were determined by visual inspection of the waveforms and were de ned as the lowest stimulus level at which any wave could be reliably detected. A minimum of two waveforms was obtained at the threshold level to ensure repeatability of the response. Physiological results were analyzed for individual frequencies, and then averaged for each of these frequencies from 4 to 32 kHz.

Immunohistochemistry and quanti cation
After completion of functional testing, mice were euthanized by CO2 asphyxiation followed by decapitation. Temporal bones were harvested and xed overnight with 4% paraformaldehyde followed by decalci cation in 120mM EDTA for 4 days. The vestibular organs and cochlear sensory epithelia were micro-dissected, blocked, and labeled with rabbit anti-myosin 7a antibody to label hair cells (1:200, product # 25-6790, Proteus BioSciences, Ramona, CA), and chicken anti-GFP antibody to label GFP (1:1000, product # ab13970, abcam, Cambridge, MA), and Hoechst stain (1:500, product # 62249, Life Technologies, Carlsbad, CA) to label nuclei. Primary and secondary antibodies were diluted in PBS. Images were obtained using a Zeiss LSM780 confocal microscope at 10x and 40x using z-stacks.
For quanti cation of cochlear hair cell and supporting cell infection e ciency, two 40x images were taken at the apex, middle turn, and base of cochlea. The number of hair cells and supporting with GFP expression was counted and averaged at each location along the cochlea. Each 40x image contains ~30 IHCs and ~90 OHCs. The overall transduction rate was calculated by averaging the transduction rates obtained from the entire cochlea. For quanti cation of utricular hair cell infection e ciency, two 40x images (each containing ~300 vestibular hair cells) were taken per utricle specimen and the number of hair cells with GFP expression was counted and averaged.

Statistics
Student's t-test (two-tailed) was used to assess differences in transfection e ciency. It has been shown that different AAV serotypes can have different transfection e ciencies in different regions of the cochlea 43 . Therefore, transfection e ciencies from each region of the cochlea (apex, middle turn, and cochlear base) were treated as separate measurements in the calculation of mean, standard error, and statistical signi cance. For ABR threshold, Student's t-test was used to assess differences in the thresholds. The p-value of <0.05 indicates statistical signi cance.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.