Choice of vector and surgical approach enables efficient cochlear gene transfer in nonhuman primate

Vector design and validation in mouse model

AAV constructs were designed to include an optimized Chicken β-actin promoter with early CMV-enhancer (CB7) and chimeric intron (CI) given its prior use in NHP and human studies^32,33,34. Due to the novelty of the construct and the promoter combination in the cochlea, prior to injection in NHPs, AAV2/1.CB7.CI.eGFP.RBG and AAV2/Anc80L65.CB7.CI.eGFP.RBG were tested in wild-type mice by systemic and cochlear injection.

In line with expectations for AAV1 and Anc80L65, differential levels of enhanced green fluorescent protein (eGFP) expression were detected in the murine livers with both vectors after intravenous delivery (Supplementary Fig. 1a)²⁸. To determine whether the CB7 promoter contributed to a different tropism or cellular expression compared to the previously used CMV promoter^22,24, neonatal mouse cochleae were injected with CMV-driven eGFP and CB7-driven eGFP vectors. No differences were found in the expression or tissue distribution (Fig. 1).

Development of surgical injection method

The surgical procedure was developed on cadaveric macaque heads. These dissections revealed that it was not feasible to utilize a transcanal approach as the external auditory canal is too small to allow introduction of surgical instruments, including the smallest available endoscope. We evaluated a transmastoid posterior tympanotomy approach, which achieved excellent exposure of the round window, and exposure of the stapes.

Based on the optimizations performed, the eventual procedure can be described as a transmastoid facial recess exposure of the RWM with round window injection following oval window venting. Specifically, a left postauricular incision was made and dissection of the soft tissue was performed down to the level of the periosteum. The periosteum was incised and elevated to expose the mastoid bone. A cortical mastoidectomy was performed with a combination of high-speed cutting and diamond drill burs (Fig. 2a, b and Supplementary Movie 1). The facial recess was then opened, allowing for adequate visualization of the round window and oval window. Fenestration of the oval window was performed using a Rosen needle (Fig. 2c and Supplementary Movie 1), because we hypothesized that this would allow potential fluid displacement and enable an improved vector flow toward the apex of the cochlea after the injection of a larger volume without damage of the internal cochlear structure. Vector injection through the RWM was then performed (Fig. 2d and Supplementary Movie 1). The soft tissue was closed in layers with buried, interrupted, 3-0 monocryl sutures (Ethicon). The skin was re-approximated with running subcuticular 4-0 monocryl suture (Ethicon).

Procedure and in-life phase

To evaluate the transduction efficiency in the cochlear sensory epithelium and auditory neurons, we sought to inject one ear of two animals with AAV1 and one ear of three additional animals with Anc80L65 (Table 1). Rhesus macaque was selected as a species given their established use in gene therapy, their genetic, metabolic, and immunological proximity and relevance to human applications, and the size of the cochlea, which is approximately 30% of the human inner ear^35,36.

Three- to four-year-old, female animals were enrolled in the study based on a clean health record, no experimental treatment history, and serum neutralizing antibody titers against AAV1 or Anc80L65 below 1:10.

The study was designed to navigate several unknowns, given the limited precedent of NHP cochlear injections, particularly of an AAV, including dose, volume, potential for inflammatory responses to vector and/or transgene that may harm the animal and the interpretation of the histology, in vivo surgical feasibility, and kinetics of expression. The study design staggered the injections over multiple days to assess surgical feasibility and short-term tolerability. The endpoint of the study was 1 week upon injection to mitigate noxious effects of any potential immunological adaptive responses. One animal dosed with Anc80L65 was monitored for 14 days following injection to preliminarily assess the kinetics of expression and whether longer-term expression was tolerated. All animals received an absolute dose of 2.5 × 10¹¹ vector genome copies (GC) in 30 μl buffered saline (at a concentration of 8.33 × 10¹² GC/ml).

All the animals successfully recovered from the surgical procedure with minor physical impairments or discomfort expected after anesthesia and/or ear surgery (imbalanced, scratching of incision site, or decreased activity) that resolved within the first 24 to 48 h. No changes in food and water intake were observed during the in-life phase of the study. Animal RA3131 however developed unilateral facial paresis ipsilateral to the injection with incomplete left eye closure (no lower facial paralysis was observed) a day following surgery. The animal was preventively treated with a lateral tarsorrhaphy 24 h after the procedure. No signs of corneal ulcer, inflammation, or irritation were detected in this animal during the in-life phase of the study. Animals were closely monitored after the procedure for surgical complications; no obvious adverse effects or signs of acute toxicity related to the injection of AAV1 or Anc80L65 viral vectors were observed (Supplementary Table 1).

Cochlear tropism and transduction efficiency

Cochleae were harvested seven days after the surgery except for animal RA3128, for which the endpoint was 14 days after the procedure. AAV transduction was evaluated by immunohistochemistry using an anti-GFP antibody. eGFP expression in the IHCs was detected in four monkeys; no expression in any cell type was detected in RA3120, which is why it is not depicted in the Figures. While it was not quantified, sporadic outer hair cell damage was observed in injected and contralateral uninjected ears for all the animals. Inner hair cells were quantified at each frequency (a total of 20 cells per frequency region along the length of the cochlea) and no cell damage was observed with either serotype (data non included). Minimal tissue damage at the most basal portion of the cochlea was detected in a few specimens, most likely related with the specimen post-mortem dissection or the injection procedure performed. No eGFP signal was observed in the histological samples of the contralateral cochlear tissue in any of the treated animals (Supplementary Fig. 1B).

Frequency mapping of the injected cochleae was carried out using images acquired with an epifluorescence microscope equipped with a 10x objective. To evaluate the transduction efficiency in hair cells and supporting cells (SCs), 63x images of each mapped frequency region were taken with a confocal microscope and quantified (Fig. 3a, b). eGFP-positive hair cells were counted, and values were plotted as a percentage of the total number of hair cells. The highest number of cells expressing eGFP was found at the apex, while the lowest number of eGFP-positive cells was observed at the base of the cochlea (Fig. 3a). Transduction efficiency of IHCs by Anc80L65 and AAV1 was different; high transduction of the IHCs was detected for Anc80L65 in all cochlear regions (Fig. 3b). However, only minimal outer hair cell (OHC) transduction was observed for both serotypes when evaluating cochlear whole mounts using confocal microscopy. Specifically, one positive outer hair cell was detected in the 32.0 kHz region of ear RA3009 and three appeared in the 22.6 kHz region of ear RA3128. Overall, no transduction of the SGNs and auditory nerve sections from different regions of the cochlea was detected (Fig. 4). eGFP signal in the modiolus was only observed in one small area of an apical section from animal RA3109 that was injected with AAV1.

In addition to the IHCs, both serotypes were able to transduce SCs (Supplementary Fig. 2). While the GFP fluorescence intensity was lower in the SCs than the IHCs, a large number of SCs were transduced in the animals included in the study. The SCs contribute to the development, structural maintenance, and function of the hair cells; some mutations in genes expressed in these cells result in deafness³⁷. A qualitative analysis of the number of eGFP-positive SCs was performed to assess the transduction efficiency (Supplementary Fig. 2a). AAV1 and Anc80L65 transduced SCs at similar efficiency from apex to base. More transduction was detected in the apical region encoding lower frequencies, with a decrease of signal towards the base of the cochlea with both AAV1 and Anc80L65 (Supplementary Fig. 2b).

Transduction of the vestibular organ and cochlear lateral wall

The peripheral vestibular system and cochlear lateral wall are relevant targets for gene transfer applications. For example, loss of type II fibrocytes in the cochlear lateral wall as well as degeneration of stria vascularis and spiral ligament contribute to noise-induced hearing loss³⁸. A qualitative analysis of the eGFP expression in whole mounts of the vestibular organ is shown in Fig. 5a, b. Only animal RA3131, which received Anc80L65 showed substantial eGFP expression throughout the peripheral vestibular system; the fluorescent signal was detected in utricle, saccule, ampulla, and membranous semicircular ducts (Fig. 5a, b). Weak fluorescent signal was detected in small areas of the ampullae, vestibule, and saccule of the animals injected with AAV1, RA3009 and RA3109. No expression was detected in animal RA3128, despite the fact that this animal had been injected with Anc80L65 and showed high eGFP signal in the cochlea. Therefore, no direct correlation was found between serotype and transgene expression in the vestibular system or other areas of the inner ear.

The lateral wall was dissected in 7 sections, and whole mounts were examined. A qualitative evaluation of the eGFP expression in the 7 sections of the lateral wall was performed in all the specimens obtained from the four animals where we found any eGFP-positive cells in the inner ear. eGFP signal was detected in at least 3 to 4 sections of the lateral wall of all NHPs examined. Signal distribution and intensity were similar between all animals, and no differences were noticed between serotypes (Fig. 5c). Microglia was stained using an anti-Iba1 antibody, transduction of the microglia (colocalization between eGFP and Iba1 signal) was not observed with either Anc80L65 or AAV1.

Systemic neutralizing response to AAV

Neutralizing antibody (NAB) titers against AAV capsid can be developed in humans and NHP as a result of exposure to the wild-type virus or after administration of AAV vectors. The presence of pre-existing NAB can preclude AAV transduction when the vector is delivered by systemic or a local route of administration. Hence, antibody titers against AAV1 and Anc80L65 in serum were determined by neutralization assays at multiple time points – prior to animal purchase, following arrival to the Animal Facility, the day of surgery (Day 0) and at the endpoint (Day 7 and Day 14, respectively). In addition, antibody titers in cerebrospinal fluid (CSF) were determined at the endpoint only. Animals were negative for NAB or showed low antibody titers prior to the surgery, but all the NHP seroconverted against the serotype injected after the surgery (Table 1). Titers were higher when animals were injected with Anc80L65, but there was no direct correlation between the titers and the levels of eGFP expression. No neutralizing antibodies were detected in CSF for either serotype (Table 1).

Acute vector tolerability and biodistribution after cochlear injection

In addition to efficient gene transfer, the use of viral vectors for clinical applications requires evaluation of the tolerability profile for future therapeutic use. To evaluate systemic toxicity or inflammatory processes, hematoxylin and eosin staining was performed in different organs and several regions of the central nervous system and a postmortem histopathology analysis was completed for both serotypes (Supplementary Fig. 3). No morphological or pathological changes were detected, and no signs of inflammation or fibrosis were detected in animals injected with AAV1 or Anc80L65.

In addition, animals were phlebotomized pre-surgery (Day 0) and at the endpoint (Day 7 and Day 14) and blood was analyzed for Serum Chemistry (Supplementary Table 2). No evidence of toxicity was observed in the blood chemistry profiles of NHPs treated with AAV1 or Anc80L65. Values were mostly within the normal species-specific reference range, and although a few differences were detected in some parameters, these alterations had no clinical significance.

Finally, biodistribution and shedding of vector genomes in CSF, cerebellum, several cerebral lobes, spinal cord, and liver were determined by droplet-digital PCR (Fig. 5d and Supplementary Table 3). Genome copies detected in cerebellum, cerebral lobes, and spinal cord for both serotypes were considered negative since levels were around background level. Higher vector DNA copies were detected in CSF and liver in animals injected with AAV1 compared with Anc80L65 monkeys, with the exception of animal RA3120, which was injected with Anc80L65 but did not express eGFP in the cochlea.

Overall, we did not identify any signs of systemic acute toxicity or inflammation related to viral vector injections, and both serotypes showed favorable profiles.