1 Introduction
Hearing loss is one of the most common sensory disorders. According to the latest estimates by the World Health Organization (WHO), approximately 430 million people worldwide, more than 5% of the global population, suffer from disabling hearing loss, including 34 million children
[1]. Among every 1,000 newborns, 2–3 are affected, and about 60% of these cases are attributable to genetic causes
[2]. Although hearing aids and cochlear implants can partially compensate for functional deficits
[3], they do not achieve complete restoration, and effective biological or medical treatments for hereditary hearing loss remain unavailable. Given these unmet clinical needs, the development of curative therapeutic strategies is urgently required. Recently, gene therapy has demonstrated promising clinical outcomes in treating hereditary hearing loss, highlighting its potential as a fundamentally curative approach for genetic forms of auditory dysfunction
[4-
6].
Adeno-associated viruses (AAVs) have become one of the common viral vectors in gene therapy due to their favorable properties, including low pathogenicity and immunogenicity, stable gene expression, and the capacity to infect both dividing and non-dividing cells
[7]. To date, at least 13 naturally occurring serotypes and more than 100 AAV variants have been developed for gene therapy research
[8,
9]. However, naturally occurring AAV serotypes often exhibit suboptimal transduction efficiency or limited tissue specificity, and epidemiological studies indicate that 40%–80% of individuals possess pre-existing antibodies against natural AAV serotypes, which collectively pose substantial challenges for clinical translation
[10,
11]. The inner ear, by contrast, is a relatively enclosed compartment anatomically isolated from surrounding tissues. Following local administration, therapeutic agents can diffuse through circulating labyrinthine fluids and reach multiple cellular targets. These unique anatomical and physiological features create an ideal environment for gene therapy targeting hereditary hearing loss
[12]. Consequently, enhancing AAV tropism toward inner ear cell holds significant clinical value for improving gene therapy outcomes in genetic forms of deafness.
AAV belongs to the genus
Dependoparvovirus within the family
Parvoviridae. It consists of an icosahedral capsid approximately 26 nm in diameter and a linear single-stranded DNA genome
[13,
14]. As a replication-defective virus, its life cycle depends on the presence of a helper virus, such as adenovirus
[15]. The AAV genome contains two major genes,
rep and
cap, flanked by inverted terminal repeats (ITRs). The tropism of AAV variants is determined by the capsid proteins encoded by the
cap gene. Within the
cap gene, three open reading frames (ORFs) are present. The first encodes the three structural viral proteins (VPs)—VP1, VP2, and VP3
[16]. These proteins share an identical C-terminal region and assemble into the AAV capsid in a ratio of 1:1:10. The unique N-terminal sequence of VP1 is required for endosomal escape during infection; the shared N-terminal region of VP1 and VP2 contributes to nuclear entry; and the common C-terminal region of all three VPs is involved in receptor recognition, capsid assembly, and genome packaging
[17]. The second ORF encodes the assembly-activating protein (AAP), which facilitates the nuclear transport of VPs and is essential for capsid assembly
[18]. The final ORF encodes the membrane-associated accessory protein (MAAP), which is critical for AAV secretion
[19,
20]. Comparative structural analyses of different AAV serotypes have shown that only approximately 520 amino acids from the C-terminal regions of the three VPs form the external surface of the capsid, while the remaining residues are located internally. The surface topology of these regions is highly conserved across serotypes, comprising a conserved α-helix (
αA), a
βA strand, and a core eight-stranded antiparallel
β-barrel (
βB–βI) connected by large interstrand loops. Nine variable regions (VRs) reside within these loops, and these VRs are associated with receptor binding, tissue tropism, and infection efficiency
[17,
21,
22]. Thus, engineering these VRs to redesign the AAV capsid and enhance its tropism toward inner ear cells represents a promising strategy with potentially transformative implications for gene therapy targeting hereditary hearing loss.
Peptide insertion is a common strategy for AAV capsid engineering
[7], which involves inserting short random peptide sequences into specific regions of the AAV capsid to generate capsid libraries, thereby altering the tropism of natural AAV serotypes. VR-IV and VR-VIII, which are located on prominent protrusions on the AAV capsid surface, serve as ideal insertion sites. Previous studies employing peptide insertion–based directed evolution of AAV9 have generated several enhanced variants—including AAV-PHP.B, AAV-PHP.eB, and AAV.CAP-B10—that exhibit markedly improved ability to cross the blood–brain barrier in mice and non-human primates, as well as enhanced neuronal transduction specificity
[23-
25]. Current AAV vectors used for inner ear gene therapy generally transduce adult mouse inner hair cells (IHCs) efficiently, yet exhibit extremely low or negligible transduction of other cochlear and vestibular cell types. Importantly, in humans, the auditory system is fully developed at birth, whereas in mice, cochlear maturation begins postnatally and is completed at approximately one month of age
[26]. Therefore, developing gene therapy strategies that are effective in adult mice carries greater translational relevance. Enhancing the tropism of AAV vectors for various inner ear cells in adult animals will substantially accelerate the translation of biological discoveries into clinical therapeutic applications.
In this study, we leveraged a peptide-insertion-based directed evolution strategy to engineer a high-diversity AAV1 capsid library. Following three iterative rounds of in vivo selection in the adult mouse cochlea, we isolated variants with significantly enhanced inner ear tropism. The transduction efficiency and safety profiles of these highly enriched AAV variants were systematically evaluated to establish an efficient toolkit for inner cochlear gene delivery.
2 Methods and materials
2.1 Animal
All experiments were approved by the Institutional Animal Care and Use Committee of ShanghaiTech University and all efforts were made to minimize the number of mice used and their suffering. Mice of the C57BL/6 strain, aged P35–P40 at the time of injection, were purchased from Shanghai Lingchang Biotechnology Co., Ltd. (Shanghai, China) and Shanghai Jihui Laboratory Animal Breeding Co., Ltd. (Shanghai, China). Animals were housed in the Model Animal Facility of ShanghaiTech University under a standard 12-hour light/12-hour dark cycle at a controlled temperature of 22 ± 1 °C, with food and water provided ad libitum.
2.2 Plasmids
A random 9-amino acid peptide library was designed to be displayed on the surface of the AAV1 capsid. An oligonucleotide encoding the 9-mer peptide (5'-MNNMNNMNNMNNMNNMNNMNNMNNMNN-3', where M is A or C, and N is A, T, C, or G) was synthesized, flanked by sequences homologous to the AAV1 cap gene for subsequent recombination-based cloning. The peptide-encoding DNA fragment was amplified via PCR using Phanta Max Super-Fidelity DNA Polymerase (Vazyme, P505) according to the manufacturer's instructions. The reaction utilized primers PD1-9-F and PD1-9-R (forward, 5'-GGCGTCACAACCATCGCTAA-3'; Reverse, 5'-CATAGCATGCACATCTCCGGTCGCAGGGTCTGTMNNMNNMNNMNNMNNMNNMNNMNNMNNGCTGCTGCTCTGGAAATT-3') and the plasmid PLib-RCA-AAV1-588-fragment as template. The PCR program was as follows: 95 °C for 3 min; 30 cycles of 95 °C for 15 s, 55 °C for 15 s, and 72 °C for 30 s; and a final extension at 72 °C for 5 min. The resulting ~850 bp PCR product was verified by 1% agarose gel electrophoresis and purified using the QIAquick Gel Extraction Kit (QIAGEN, 28706).
The recipient plasmid, PLib-RCA-AAV1-588-vector-bar3, was digested with Sgs I and Pac I restriction enzymes (NEB) for 3 hours at 37 °C. The linearized about 7000 bp vector backbone was isolated by gel electrophoresis and purified. The purified peptide-encoding PCR fragment and the linearized vector were assembled using the NovoRec® plus One step PCR Cloning Kit (novo protein, NR005-01B) with a 1:2 (vector: insert) molar ratio. The reaction was incubated at 50 °C for 30 min, and the resulting library plasmid was purified using the QIA quick PCR Purification Kit (QIAGEN, 28106).
The purified library plasmid was electroporated into MegaX DH10B T1R Electrocomp™ Cells (Invitrogen, C640003). The transformed bacteria were allowed to recover for 1 hour in SOC medium and then plated on LB/Amp plates at various dilutions to determine library diversity. The remaining culture was expanded in 500 mL of LB/Amp medium for 14 hours. Plasmid DNA was extracted using the HiPure Plasmid EF Maxi Kit (Magen, P1114-03). Library diversity was calculated as: Diversity = (Number of colonies) × (Dilution factor) × 10 × (Total recovery volume / Plated volume) × (Sequencing positive rate). Twenty random clones were picked for Sanger sequencing to validate the diversity and correct insertion of the peptide library.
2.3 AAV production and purification
HEK293T cells were maintained in DMEM (HyClone) supplemented with 10% fetal bovine serum (LOSERA) and 1% Penicillin-Streptomycin (Gibco). For AAV production, cells at about 90% confluence in 15-cm dishes were co-transfected using the calcium phosphate method. For peptide-library AAV packaging, each dish received 100 ng library plasmid and 42.4 µg helper plasmid. For standard AAV vectors, capsid plasmid, transgene plasmid, and pHelper were co-transfected at defined ratios. Six to eight hours post-transfection, cultures were switched to production medium (DMEM with 1% FBS). Supernatants were collected at 48-hour, and cells were harvested at 96 hours post-transfection.
AAV particles were precipitated overnight at 4 °C by adding NaCl (to 1 mol/L) and PEG8000 (10% w/v). The precipitate was collected by centrifugation, resuspended, and treated with DNase I and RNase A. A crude lysate was obtained after chloroform extraction. The virus was purified by iodixanol (BioVision) density gradient ultracentrifugation. A discontinuous gradient was prepared with 15%, 25%, 40%, and 60% iodixanol layers in an ultracentrifuge tube. The clarified lysate was loaded on top and centrifuged at 58,400 rpm for 2 hours and 25 minutes at 18 °C (SW 41 Ti rotor, Beckman Coulter). The virus-containing fraction at the 40%–60% interface was collected. The purified virus was concentrated and buffer-exchanged into PBS containing 0.001% Pluronic F-68 using 100 kDa molecular weight cut-off centrifugal concentrators (Amicon). The concentrated virus was aliquoted and stored at −80 °C.
Viral genome (VG) titers were determined by qPCR. Viral samples were treated with DNase I to remove free DNA, followed by protease digestion to release the viral genome. qPCR was performed on a Quant Studio system using primers Rep2-qP-F and Rep2-qP-R targeting the rep gene. Standard curves were generated for absolute quantification, and titers were calculated according to established formulas (GC/mL).
2.4 In vivo screening in mice
AAV peptide-insertion libraries were injected into the cochlea of P35–P40 C57BL/6 mice via a posterior semicircular canal approach. Microinjections were performed using custom-pulled glass pipettes under anesthesia. The fenestration and surgical wound were sealed with tissue adhesive (3 mol/L). Mice were euthanized 5 days post-injection. Cochleae were harvested and homogenized using a high-speed tissue grinder (KZ-II). Genomic DNA, including the packaged AAV genomes, was extracted from the homogenates using the DNeasy Blood & Tissue Kit (QIAGEN, 69506).
2.5 Analysis of enriched capsid sequences
Capsid sequences from the in vivo screened library were identified from cochlear genomic DNA by PCR using Phanta Max Master Mix and primers PD1-9-recover-F (5'-GGCGTCACAACCATCGCTAA-3') and PD1-9-recover-R (5'-GCTCCCATAGCATGCACATCTCC-3'). The PCR program was: 95 °C for 3 min; 30 cycles of 95 °C for 15 s, 56 °C for 15 s, and 72 °C for 30 s; final extension at 72 °C for 5 min. The ~850 bp product was gel-purified.
The second PCR was performed using Q5® Hot Start High-Fidelity 2X Master Mix (NEB, M0494S) and primers AAV1-NGS-F (5'-AAGCCACTAACCCTGTGGC-3') and AAV1-NGS-R (5'-AATGGGACCCTGCAGGTACA-3') to generate ~180 bp amplicons. The program was: 98 °C for 30 s; 20 cycles of 98 °C for 10 s, 55 °C for 30 s, and 72 °C for 15 s; final extension at 72 °C for 2 min. The resulting NGS library was purified and sequenced on an Illumina platform (GENEWIZ).
2.6 Auditory function test via ABR
Hearing function in injected mice was assessed by tory Brainstem Response (ABR). Anesthetized mice were placed in a sound-attenuating chamber. Subdermal electrodes were inserted at the vertex (active), beneath the pinna of the test ear (reference), and near the contralateral ear (ground). Brainstem responses to click and pure-tone stimuli (5.6–32 kHz) at descending sound pressure levels were recorded using BioSigRZ software (TDT).
2.7 Immunofluorescence and confocal microscopy
Cochleae were harvested, fixed in 4% PFA, and decalcified in 0.5 mol/L EDTA. The basilar membrane and stria vascularis were micro-dissected, blocked with 10% donkey serum, and incubated overnight at 4 °C with a primary antibody against Myosin VIIa (Proteus Biosciences, 25-6790, 1:500). Tissues were then incubated with an Alexa Fluor 555-conjugated secondary antibody (Invitrogen, A31570, 1:500) and counterstained with DAPI. Samples were mounted with Fluoromount-G® (EMS) and imaged using a laser scanning confocal microscope (TI-E AI STROM, Nikon).
2.8 Statistical analysis
Statistical analyses and data visualization were performed using Microsoft Excel and GraphPad Prism 6.0. Data are presented as the mean or mean ± SEM as indicated. For ABR threshold comparisons across multiple frequencies, two-tailed unpaired Student’s t-tests were applied, and the resulting P-values were corrected for multiple comparisons using the Bonferroni method. A Bonferroni-adjusted P < 0.05 was considered statistically significant. For analyses involving more than two groups, one-way ANOVA followed by Bonferroni post hoc testing was performed where appropriate. Sample sizes (n values), the exact statistical tests applied, and the definitions of error bars are provided in the respective figure legends. All experiments were conducted with at least three biological replicates unless otherwise specified, and no data points were excluded from analysis.
3 Results
3.1 Construction and In vivo screening of the AAV1 nine-peptide insertion library
Directed evolution relies on generating a diverse library of capsid sequence variants followed by multiple rounds of
in vivo or
in vitro selection. A widely used method is the insertion of random peptide sequences into variable loop regions of the AAV capsid, such as VR-IV or VR-VIII
[27]. Competition among variants within the library enables the identification of enriched peptide sequences that enhance capsid tropism toward specific target tissues. Most studies employ natural AAV serotypes or their derivatives, such as AAV1, AAV2, or AAV2/1. These serotypes typically transduce IHCs efficiently
[28,
29]，but show extremely low or negligible transduction of outer hair cells (OHCs) and supporting cells. In AAV1, the VR-VIII region surrounding amino acid position 588 is located on a prominent surface-exposed protrusion, making it a commonly selected site for peptide insertion–based capsid engineering
[30]. We constructed an AAV1 capsid library by inserting a randomized nine-peptide at position 588 (Figure 1). Electro transformation and Sanger sequencing of single colonies estimated the final plasmid library diversity to be 2.6 × 10
7 (Figure 2). The library plasmids were packaged into AAV particles and subjected to three rounds of
in vivo selection. For each round, the library virus was injected into the cochleae of P35-P40 C57BL/6 mice via the posterior semicircular canal. In addition, under the same experimental conditions we compared the established AAV variants with Anc80L65 to illustrate the improvements and enhanced performance of our engineered capsids. Anc80L65 is a synthetic AAV capsid that integrates advantageous features from multiple serotypes (e.g., AAV1, AAV2, AAV8, AAV9). It exhibits high transduction efficiency and low toxicity across diverse inner ear cell types, and has demonstrated robust transduction performance and tolerability in both adult murine models and non-human primates
[31,
32]. After five days, AAV genomes were identified from the cochleae, analyzed by next-generation sequencing (NGS), and used to reconstruct the library for the subsequent selection round. The highly enriched variants identified from the final AAV library screening were designated Wild Adult Cochlea AAV1 9-1 to -10 (WAC19-1 to WAC19-10).
The heatmap of amino acid distribution showed a significant shift from the initial library to the post-selection pool, with a notable enrichment of proline at positions 5, 6, 7, and 8 of the inserted peptides (Figure 3).
Analysis of the NGS data revealed a dynamic enrichment process (Figure 4a). Each bubble represents a unique amino acid sequence, and bubble size reflects the fold enrichment of each AAV capsid variant relative to the parental plasmid library. The top-performing variant, WAC19-1, was enriched by nearly two million-fold after the third round of selection. This trend was further supported by the sequence logo of the top 50 enriched variants, in which proline was consistently overrepresented at positions 5, 6, 7, and 8 of the inserted nine-peptide, consistent with earlier observations (Figure 4b). Amino acid colors correspond to those used in the bubble plot. Variants with amino acid frequencies greater than 2% in the third-round sequencing results were selected as candidate sequences. Based on their high abundance following the final selection round (Figure 4c), nine candidates (WAC19-1 to WAC19-8 and WAC19-10) were chosen for further validation.
3.2 Packaging efficiency of candidate AAV variants
The candidate capsids were packaged with a CMV-EGFP expression cassette. Viral titers, determined by qPCR, revealed significant differences in packaging efficiency among the variants (Table 1). While most candidates packaged robustly, yielding titers between 4.78 E12 to 2.93 E13 GC, WAC19-8 was an extreme outlier, producing a very low titer (1.15 E10 GC) despite using a large amount of starting material. Consequently, WAC19-8 was excluded from subsequent functional analyses.
3.3 Distinct tropism of candidate AAVs for cochlear cell types
To evaluate the ability of the AAV1 nine-peptide-insertion library candidates to transduce cochlear cells in adult mice, we used the CMV promoter and packaged EGFP under the capsids of WAC19-1, WAC19-2, WAC19-3, WAC19-4, WAC19-5, WAC19-6, WAC19-7, and WAC19-10. To assess cell-specific tropism, each candidate AAV was injected into the cochleae of adult mice (in total 5 E9 GC). Fourteen days after injection, cochlear tissues were collected and examined by immunofluorescence staining.
3.4 Hair cell tropism
WAC19-2 emerged as the most efficient vector for cochlear hair cells, achieving 100% transduction of both IHCs and OHCs across the apical, middle, and basal turns, although EGFP expression in OHCs gradually decreased from apex to base. Six other candidates (WAC19-1, -3, -5, -6, -10) also reached 100% IHC transduction. WAC19-3 transduced nearly 100% of OHCs in the apical and middle turns, but EGFP fluorescence intensity was very weak and transduction failed in the basal turn. WAC19-5, -6, and -10 only transduced OHCs in the apical turn. WAC19-4-CMV-EGFP transduced approximately 85% of IHCs in the apical turn and 50−60% in the middle and basal turns. WAC19-7-CMV-EGFP achieved full IHC transduction in the apical turn, ~70% in the middle turn, and ~10% in the basal turn. Notably, WAC19-1, -4, and -7 showed no tropism for OHCs (Figures 5 and 6).
3.5 Supporting cell tropism
Supporting cells can be classified into four types: border cells (BCs), phalangeal cells (PhCs), pillar cells (PCs), and Hensen’s cells (HeCs)
[33]. Tropism for supporting cells was generally more restricted (Figure 7). Among the candidates, only WAC19-7 exhibited broad transduction, infecting outer phalangeal cells (OPhCs) across all three cochlear turns. WAC19-2 and WAC19-10 displayed limited tropism, transducing OPhCs only in the apical turn with weak EGFP expression. The remaining candidates showed no detectable transduction of supporting cells.
3.6 Safety assessment: impact on hearing
The ototoxicity of the candidate vectors was evaluated by measuring ABR thresholds 14 days post-injection (Figure 8). Using the CMV promoter, EGFP was packaged into the capsids of WAC19-1, WAC19-2, WAC19-3, WAC19-4, WAC19-5, WAC19-6, WAC19-7, and WAC19-10. Each virus was injected into the cochleae of mice via the posterior semicircular canal, with PBS-injected mice serving as negative controls. Four variants—WAC19-1, WAC19-5, WAC19-7, and WAC19-10—showed no statistically significant hearing loss at any tested frequency compared to wild-type mice, indicating an excellent safety profile. WAC19-2 caused a slight but significant hearing impairment at 5.6 kHz. In contrast, WAC19-3 was highly ototoxic, causing approximately 35 dB of hearing loss across all frequencies. WAC19-4 impaired low-frequency hearing, and WAC19-6 impaired both low- and high-frequency hearing.
4 Discussion
Congenital deafness remains a significant global public health challenge, with genetic defects being a major cause
[1]. Adeno-associated virus (AAV)-mediated gene therapy offers a promising curative strategy
[4-
6]. However, targeted gene delivery is complicated by the cochlea’s structural complexity and relative isolation. While several AAV vectors have been reported (e.g., Anc80L65, AAV2.7m8, AAV-ie), their encouraging performance is primarily limited to neonatal mouse models
[32,
34,
35]. Critically, they show limited applicability in adult cochleae, and generally demonstrate poor transduction of non-hair cell populations. The significance of transducing adult cochlear cells is paramount, as mice only develop mature auditory behaviors after the first postnatal month.
In this study, we report the first systematic effort to construct and screen an AAV library via peptide insertion in the adult mouse cochlea. This
in vivo directed evolution strategy is a crucial testing stone that moves the field toward translationally relevant gene therapy. By inserting randomized nine-peptide into the AAV1 capsid’s VR-VIII loop and performing three iterative rounds of selection, we successfully isolated a panel of variants with significantly enhanced and diversified tropism, achieving transduction efficiencies that surpass those of many engineered capsids reported in recent years, including several AAV variants evolved through directed selection in the brain or retina
[36]. This approach, demonstrated in a fully developed auditory system, is vital not only for basic auditory research but for developing effective clinical therapeutics.
Our screening strategy yielded a spectrum of AAV variants with distinct and highly selective cochlear tropisms, demonstrating that peptide insertion can finely tune cellular targeting beyond what existing serotypes provide. Notably, one class of variants preferentially targeted supporting cell populations such as OPhCs, a feature that is particularly valuable for regenerative gene therapy approaches that depend on precise modulation of non-sensory cell types, including transcription-factor-based reprogramming and Notch-pathway inhibition
[37]. These findings illustrate the broader principle that capsid engineering can be leveraged not only to enhance efficiency but also to diversify the cellular specificity required for next-generation therapeutic strategies.
Our safety assessment further revealed that enhanced cellular entry is not inherently synonymous with biological compatibility. Variants that exhibited strong transduction in specific cochlear cell types did not uniformly demonstrate favorable functional outcomes, underscoring the multifaceted nature of vector safety. While several engineered capsids maintained normal auditory thresholds across all tested frequencies, others were associated with measurable hearing impairment, highlighting the importance of dissociating transduction potency from physiological tolerability. Rather than indicating a limitation of the approach, this diversity of outcomes provides an important foundation for iterative refinement. The ability to identify both well-tolerated and potentially ototoxic variants within the same library underscores the usefulness of our platform for guiding the rational design of vectors that balance targeted potency with long-term cochlear safety.
More work is clearly needed to refine these vectors. Nevertheless, this work successfully validates the strategy of in vivo directed evolution in the adult cochlea. It will undoubtedly stimulate a massive increase in research interest, as it provides a new, functional toolkit for precision cochlear gene delivery and sets the stage for advanced therapeutic applications, including gene replacement, RNA editing, and CRISPR-based interventions. This trend toward highly specific, safe, and effective AAV engineering for the adult inner ear must be true to bring curative treatments to patients.
5 Limitation
This study is an exploratory investigation applies a nine-amino acid peptide insertion strategy to construct an AAV capsid library and identify variants with enhanced hair-cell tropism and transduction efficiency. Despite its novelty, several limitations should be acknowledged. First, although WAC19-2 exhibited robust and widespread cochlear hair-cell transduction, the quantitative interpretation of fluorescence intensity is constrained by imaging settings. Second, the safety evaluation focused on acute auditory outcomes. ABR thresholds measured 14 days post-injection provide only short-term information, without assessing potential long-term ototoxicity. Extended follow-up—including multi-month hearing assessments together with histological analyses such as hair-cell counts, synaptic integrity, and inflammatory responses—will be essential to determine the durability and structural safety of WAC19-series-mediated transduction. Third, each experimental groups including only three animals, which may limit statistical power and increase variability, studies with larger sample sizes are necessary to validate the robustness of the observed effects.
The Author(s) 2025. This article is available under open access at journal.hep.com.cn.
PDF
(23505KB)
