Müller Glia–Exclusive CLRN1 Expression Drives Non–Cell-Autonomous Photoreceptor Degeneration in Usher Syndrome Type 3A

Abstract

Purpose

Usher syndrome type 3A (USH3A), caused by CLRN1 mutations, leads to progressive deaf–blindness. Although CLRN1 expression has been localized to Müller glia (MG), the pathogenic mechanisms linking glial CLRN1 deficiency to photoreceptor death remain undefined. We utilized a novel large-animal model to elucidate cell-type–specific pathogenesis at single-nucleus resolution.

Methods

CLRN1 ^−/− rabbits were generated via CRISPR/Cas9. Retinal phenotyping included longitudinal optical coherence tomography (OCT), electroretinography (ERG), and histology for up to 3 years. Single-nucleus RNA sequencing (snRNA-seq) was performed at a pre-symptomatic stage (10 months, n = 3/genotype). Differentially expressed genes (DEGs) were identified using pseudobulk DESeq2 and validated via immunofluorescence.

Results

CLRN1 ^−/− rabbits exhibited progressive outer nuclear layer thinning and delayed ERG responses starting after 20 months, mimicking human clinical kinetics. The snRNA-seq revealed that CLRN1 is exclusively expressed in a discrete subset (∼24%) of MG. Mutant MG exhibited transcriptomic downregulation of the adherens junction component CTNNA2, which was validated at the protein level by immunofluorescence staining. Photoreceptors exhibit extensive non–cell-autonomous dysregulation (232 rod and 68 cone DEGs) characterized by loss of the cone synaptic organizer TENM2 and rod-specific splicing dysregulation (CWF19L2). Concurrently, inner retinal neurons mounted a robust proteostatic response via HSP90 family upregulation (HSP90AB1, HSP90AA1, HSP90B1), a defensive signature absent in vulnerable photoreceptors.

Conclusions

Our findings suggest an “anchor–shield” mechanism: MG CLRN1 loss compromises the outer limiting membrane (OLM) adhesion complex (“anchor”), and photoreceptors lack the proteostatic response (“shield”) seen in resilient inner neurons. Restoring CLRN1 in Müller glia or implementing photoreceptor neuroprotection represents promising therapeutic strategies for USH3A.

Usher syndrome represents the most prevalent genetic disorder resulting in concurrent auditory and visual impairment, characterized by a wide range of clinical presentations.¹ Within its classifications, Usher syndrome type 3A (USH3A) is attributed to mutations in the CLRN1 gene, which encodes clarin-1, a protein with four transmembrane domains crucial for preserving the structural and functional integrity of sensory cells.^2,3 Mutations in CLRN1 lead to vestibular dysfunction and progressive loss of both vision and hearing. Cochlear implantation alleviates hearing impairment to some degree, but no treatment exists to prevent the progressive photoreceptor cell death that leads to blindness. Although USH3A is relatively rare in most populations, it represents approximately 40% of USH cases in Finland and in Ashkenazi Jews.⁴

Recent advancements in transcriptomics and high-resolution in situ RNAscope imaging have resolved the longstanding debate regarding the cellular localization of clarin-1. Contrary to the initial hypothesis positing the presence of the protein in photoreceptor synapses,⁵ evidence from healthy human, macaque, and mouse retinas indicates that CLRN1 expression is primarily confined to Müller glia.³ This shift suggests that USH3A is a Müller glia–primary disorder, yet the mechanism by which a glial defect triggers the progressive degeneration of light-sensing neurons remains elusive.

Rodent models, specifically clrn1 knockout and N48K knock-in mice, have been instrumental in elucidating the function of clarin-1 by demonstrating its critical role in the maintenance of sensory hair cells.^6,7 Nonetheless, a significant limitation of these models is their inability to manifest a pronounced retinal degeneration phenotype or notable structural abnormalities, thereby hindering their capacity to fully replicate the human USH3A ocular condition in vivo. Zebrafish models of USH3A have yielded significant insights, demonstrating that the loss of clarin-1 results in disorganized actin-based structures in both Müller glia and photoreceptors and suggesting a structural support mechanism, possibly through direct interactions between Müller glia and photoreceptors.⁸ However, the cone-dominant retinas and inherent regenerative capacity of zebrafish constrain their effectiveness in modeling the chronic, progressive degeneration characteristic of retinitis pigmentosa.^8,9 To address this translational gap, there is a critical need for a large animal model that possesses anatomical and physiological similarities to the human eye.

In this study, we employed CRISPR/Cas9 technology to develop the inaugural CLRN1 null rabbit model. Our longitudinal evaluation over a 3-year period demonstrated that this model effectively recapitulated the dual sensory deficits characteristic of USH3A, including progressive hearing loss and manifestation of progressive photoreceptor degeneration and visual loss. Utilizing single-nucleus RNA sequencing (snRNA-seq) with single nucleotide polymorphism (SNP)-based demultiplexing and pseudobulk differential expression analysis, we propose a novel anchor–shield pathogenic cascade in the USH3A rabbit model. This cascade elucidates the mechanism by which a primary glial defect leads to non–cell-autonomous photoreceptor degeneration while preserving the inner retinal region. This CLRN1^−/− rabbit model provides a definitive mammalian platform for investigating USH3A pathophysiology and evaluating novel therapeutic strategies.

Methods

Animals

All procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Animal Care and Use Committee at the University of Michigan (PRO00011158). New Zealand White rabbits were bred and housed at the Center for Advanced Models and Translational Sciences and Therapeutics (CAMTraST, University of Michigan Medical School, Ann Arbor, MI) in specific pathogen-free conditions with a 12-hour light/dark cycle. The light fluence in the animal room was maintained at a mean of 61.08 lux (top cages), 33.93 lux (middle cages), and 21.11 lux (floor cages). Both male and female rabbits were used to account for sex-related variability.

Rabbit CLRN1 Gene Editing

CLRN1 mutant rabbits were generated using CRISPR/Cas9 genome editing as previously described.¹⁰ Briefly, pronuclear-stage embryos were injected with high-fidelity (HiFi) Cas9/gRNA ribonucleoprotein complex (50 ng/µL each) targeting exon 1 of CLRN1 (target sequence is shown in Supplementary Table S1) and transferred to synchronized recipients. Founder animals were genotyped by polymerase chain reaction (PCR) amplification of genomic DNA from ear tissue followed by Sanger sequencing (primer sequences are shown in Supplementary Table S1).

CLRN1 Expression Analysis

Total RNA from wild-type (WT) rabbit retina and other tissues was isolated using RNeasy kits (QIAGEN, Hilden, Germany). Quantitative real-time PCR (RT-qPCR) was performed using SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA, USA) on a CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories) with primers spanning CLRN1 exons 1 to 2 (Supplementary Table S1). For in situ hybridization, paraffin-embedded retinal sections were processed using RNAscope Multiplex Fluorescent Reagent Kit v2 (Advanced Cell Diagnostics, Newark, CA, USA) with probes targeting rabbit CLRN1 (nucleotides 160–1229 of XM_002716269.4). Images were acquired using a DMi8 automated widefield fluorescence microscope (Leica Microsystems, Wetzlar, Germany). Western blot analysis was performed on retinal homogenates using anti-clarin-1 antibody (rabbit polyclonal, 1:3,000; 26630-1-AP; Proteintech, Rosemont, IL, USA).³ Cultured HEK293 cells were transfected with CLRN1 cDNA expression plasmid (hemagglutinin [HA] tagged) using the jetPRIME transfection reagent (Polyplus, Illkirch, France) according to the manufacturer's instructions and harvested 48 hours post-transfection for protein extraction using TRIzol reagent. Immunoblot analysis was performed using an anti-HA-tag antibody (mouse monoclonal, clone 6E2, 1:1,000; 2367; Cell Signaling Technology, Danvers, MA, USA). Anti–beta-actin antibody (mouse monoclonal, clone 8H10D10, 1:1000; 3700; Cell Signaling Technology) served as a loading control antibody. Retinal phenotyping auditory brainstem response (ABR), spectral-domain optical coherence tomography (SD-OCT), electroretinography (ERG), and histological analyses were performed as previously described.¹¹ Briefly, full-field ERGs were recorded under dark- and light-adapted conditions to assess rod and cone function. SD-OCT imaging was performed to measure outer nuclear layer (ONL) and outer segment thickness. Retinal histology was performed on paraffin-embedded sections stained with hematoxylin and eosin.

Immunofluorescence and Confocal Imaging

Retinal punches from the visual streak of 20-month-old WT and CLRN1^–/– rabbits (n = 4/genotype) were fixed in 4% paraformaldehyde in PBS for 30 minutes, cryoprotected in 30% sucrose overnight, and embedded in optimal cutting temperature compound (Tissue-Tek; Sakura Finetek USA, Torrance, CA, USA) for cryosectioning (10 µm). Sections were blocked in 5% normal donkey serum with 0.3% Triton X-100 in PBS for 1 hour at room temperature, then incubated overnight at 4°C with a rabbit anti-alpha-N-catenin primary antibody (F0920, clone K10P7, 1:50; Selleck Chemicals, Houston, TX, USA). Following PBS washes, sections were incubated for 2 hours with a donkey anti-rabbit Alexa Fluor 647 secondary antibody (A-31573, 1:300; Thermo Fisher Scientific, Waltham, MA, USA), followed by Alexa Fluor 488–conjugated phalloidin (R37110; Thermo Fisher Scientific) for 30 minutes to visualize F-actin. Slides were mounted using ProLong Diamond Antifade Mountant with DAPI (P36962; Thermo Fisher Scientific) for nuclear counterstaining. Representative confocal Z-stacks centered on the outer limiting membrane (OLM) were captured using a Nikon AX R NSPARC confocal microscope (Nikon Corporation, Tokyo, Japan) at approximately 56× effective magnification (20× objective with 2.8× digital zoom). To ensure valid fluorescence intensity comparisons, all samples were processed concurrently and imaged using identical laser power, gain, offset, and pinhole configurations.

Single-Nucleus RNA Sequencing

Retinal nuclei were isolated from WT (n = 3) and CLRN1^−/− (n = 3) rabbits at 10 months of age. Nuclei from three biological replicates per genotype were pooled and processed using the 10x Genomics Chromium platform (10x Genomics, Pleasanton, CA, USA). Libraries were sequenced on a NovaSeq 6000 (Illumina, San Diego, CA, USA).

Bioinformatic Analysis

Raw sequencing reads were aligned to the rabbit reference genome (OryCun2.0, Ensembl release 113) using Cell Ranger (10x Genomics). Quality control was performed using Seurat v5, retaining nuclei with 200 to 2500 detected genes and <5% mitochondrial reads.¹² Data were normalized using SCTransform, followed by principal component analysis and uniform manifold approximation and projection (UMAP) for dimensionality reduction. Cell clustering was performed using the Louvain algorithm, and cell types were annotated based on canonical retinal markers.

SNP-Based Demultiplexing

Pooled samples were deconvolved into individual biological replicates using cellsnp-lite v1.2.3¹³ for genotype calling at heterozygous SNP positions, followed by vireo v0.5.9¹⁴ for donor assignment (k = 3 donors per condition, probability threshold = 0.5). Doublets were excluded from analysis.

Differential Expression Analysis

Pseudobulk differential expression analysis was performed to account for biological variability. For each cell type, raw counts were aggregated per biological replicate using the Seurat AggregateExpression function. Differential expression testing was performed using DESeq2 v1.40¹⁵ with design formula ∼condition. Genes with fewer than 10 total counts were excluded. Differentially expressed genes (DEGs) were defined as Benjamini–Hochberg adjusted P < 0.05. For CLRN1 expression comparison between genotypes, expression was averaged per animal (n = 3 per genotype) and compared using the Wilcoxon rank-sum test. Cross-cell-type expression validation was performed to assess DEG credibility. For each significant DEG, average expression across all cell types was evaluated to confirm enrichment in the expected cell type, mitigating artifacts from ambient RNA contamination or nuclei misassignment.

Gene Ontology Enrichment Analysis

Gene Ontology (GO) enrichment analysis was performed using clusterProfiler v4.8.¹⁶ Upregulated and downregulated genes (adjusted P < 0.05) were analyzed separately for Biological Process terms using human orthologs (org.Hs.eg.db) as reference due to limited rabbit annotations. For Müller glia, which had only one gene reaching the adjusted P value threshold, a relaxed significance cutoff (nominal P < 0.05) was applied to enable pathway-level analysis. Benjamini–Hochberg correction was applied, with adjusted P < 0.05 considered significant.

Statistical Analysis

Statistical analyses were performed using SPSS Statistics 22 (IBM, Chicago, IL, USA) or R 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Data are presented as mean ± SD. The Mann–Whitney U test was used for between-group comparisons, with P < 0.05 considered significant. For snRNAseq analyses, statistical methods are described in the respective subsections above.

Results

The Clarin-1 Protein Exhibits a Significant Degree of Conservation Between Rabbits and Humans

The rabbit CLRN1 gene is comprised of three exons that encode a protein consisting of 232 amino acids (Ensembl transcript ID: ENSOCUT00000013346.3). Sequence alignment analysis indicates a high degree of conservation between rabbit and human clarin-1 proteins, with an identity score of 94% and a positive score of 96% (Supplementary Fig. S1a). To investigate the expression profiles of CLRN1 in rabbits, RT-qPCR was conducted. The results revealed that CLRN1 transcripts are predominantly present in the retina (Supplementary Fig. S1b). RNAScope analysis further demonstrated that CLRN1 mRNA in the rabbit retina is primarily localized to the inner nuclear layer (INL) (Supplementary Fig. S1c), which is consistent with findings in other species.^3,8,17

Development of CLRN1 Null Rabbit Model

To develop a genetically modified model of USH3A, we utilized CRISPR/Cas9-mediated genome editing to create CLRN1 frameshift mutations in rabbits. The CRISPR/Cas9 system was employed to target exon 1 of the rabbit CLRN1 gene in New Zealand White rabbit embryos. A 23-bp deletion frameshift mutation (p.L45G*fs76), resulting in a premature stop codon, was selected to generate a null allele, which is predicted to eliminate clarin-1 protein production (Figs. 1a, 1b). The CLRN1 mutant rabbits demonstrated normal viability and development.

Figure 1.

Development of CLRN1 mutant rabbit. (a) Schematic representation of CRISPR/Cas9-mediated genome editing used to generate CLRN1 mutant rabbits, showing the CRISPR target sites relative to the CLRN1 exons. (b) Comparison of CLRN1 cDNA sequences between WT and −23-bp CLRN1 mutant alleles, along with the corresponding predicted protein sequences. (c) Retinal CLRN1 expression was analyzed by RT-qPCR using the primers RTF1/RTR1. (d) Cloning of full-length CLRN1 cDNA from WT and mutant (Mut) rabbit retinas for overexpression in HEK293 cells. Protein expression was detected by western blotting using an anti-HA antibody, and β-actin served as the loading control. (e) ABR testing demonstrated progressive hearing loss in CLRN1 null (CLRN1 KO) rabbits at 3 weeks and 1 year of age. (f) Inner ear pathology at 1 year, showing loss of outer hair cells (black arrows) and degeneration of spiral ganglion neurons (white arrows) in CLRN1 null (CLRN1 KO) rabbits.

The RT-qPCR analysis revealed no significant difference in CLRN1 mRNA expression between WT and mutant retinas (Fig. 1c). Following confirmation of the frameshift mutation via cDNA cloning and Sanger sequencing, we hypothesize that the observed transcript stability results from the mutant CLRN1 mRNA evading nonsense-mediated decay. This evasion may be attributed to the position of the premature termination codon or specific structural features within the shortened transcript.

Western blot analysis failed to detect clarin-1 protein in both WT and CLRN1 mutant rabbit retinas (data not shown). This absence of signal likely results from either a lack of cross-reactivity between the human-targeted antibody and the rabbit ortholog or endogenous expression levels that fall below the limit of detection of the assay. To confirm that the 23-bp deletion results in a loss of clarin-1 protein, full-length CLRN1 cDNAs were amplified from WT and mutant rabbit retinas (Fig. 1d, Supplementary Table S1). Sanger sequencing confirmed the 23-bp frameshift deletion within the mutant transcript. These cDNAs were then subcloned into a mammalian expression vector featuring a C-terminal HA tag (Fig. 1d). Upon transfection into HEK293 cells, the WT construct yielded robust clarin-1 expression as detected by western blot using an anti-HA antibody. In contrast, the mutant construct completely abolished protein expression (Fig. 1d), validating the functional knockout. Consequently, this rabbit line was designated as the CLRN1 null (CLRN1^−/−) rabbit model.

Progressive Hearing Loss in CLRN1 Null Rabbits

Given that USH3A impacts both auditory and visual functions, we evaluated the auditory capabilities in CLRN1^−/− rabbits. ABR testing indicated a significant hearing impairment at frequencies of 4, 12, and 16 kHz in CLRN1^−/− rabbits when compared to WT controls at 3 weeks of age. By 14 months, auditory responses were absent across all tested frequencies (n = 3, P = 0.0139) (Fig. 1e). Histological examination revealed that the severe hearing loss was associated with degeneration of cochlear hair cells and spiral ganglion neurons (Fig. 1f). No vestibular abnormalities were detected in the CLRN1^−/− rabbits.

Progressive Outer Retinal Degeneration in CLRN1 Null Rabbits

To evaluate structural retinal alterations associated with CLRN1 deficiency, we performed color fundus photography and longitudinal SD-OCT imaging on WT and CLRN1^−/− rabbits from 2 months to 3 years of age. Color fundus photography revealed normal retinal architecture and morphology of retinal vessels, choroidal vessels, nerve fiber layer, and optic nerve in both genotypes throughout the study period (Supplementary Fig. S2).

SD-OCT imaging of the visual streak region (3 mm inferior to the retinal vessels) demonstrated a biphasic disease course: Retinal architecture remained normal in CLRN1^−/− animals through 12 months (pre-symptomatic phase), followed by progressive ONL and outer retinal thinning beginning at 24 months, while inner retinal architecture remained preserved throughout (Fig. 2a). Quantitative analysis revealed a significant age-dependent decline in ONL thickness, with accelerated thinning in CLRN1^−/− rabbits compared to WT controls at 3 years (P < 0.05) (Fig. 2b). Similarly, outer segment (OS) thickness exhibited a progressive reduction in CLRN1^−/− animals, with significant changes by 3 years (P < 0.01) (Fig. 2c). In contrast, inner retina (IR) thickness remained stable across all time points with no significant differences between genotypes (not significant) (Fig. 2d). Overall outer retinal thickness (encompassing ONL, OLM, and OS) was significantly reduced in CLRN1^−/− rabbits at 3 years (P < 0.05) (Fig. 2e). These longitudinal findings demonstrate that CLRN1 deficiency causes selective, progressive outer retinal degeneration beginning after 12 months while sparing inner retinal structure.

Figure 2.

Time-dependent changes in retinal morphological features on OCT in CLRN1 null rabbits. (a) Representative SD-OCT images of WT and CLRN1^−/− rabbits 2 months to 3 years of age. GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OLM, outer limiting membrane; OS, outer segment; RPE, retinal pigment epithelium; IR, inner retina; OR, outer retina. (b–e) Quantitative analysis of ONL (b), OS (c), IR (d), and OR (e) thickness from 2 months to 3 years of age. CLRN1 KO represents CLRN1^−/−. Scale bar: 50 µm. *P < 0.05, **P < 0.01 (Mann–Whitney U test; n = 6).

Progressive Retinal Dysfunction in CLRN1 Null Rabbits

Full-field electroretinography (ffERG) was performed on WT and CLRN1^−/− rabbits from 2 months to 3 years of age (Supplementary Fig. S3). At 3 years, representative ERG waveforms demonstrated alterations in rod- and cone-mediated responses in CLRN1^−/− subjects (Fig. 3a). Under scotopic conditions, B-wave amplitudes did not exhibit significant genotype-dependent differences at −24 dB and 0 dB intensities. A-wave amplitudes at 0-dB scotopic stimulation were lower in CLRN1^−/− rabbits, although this difference did not reach statistical significance (Fig. 3b). In contrast, cone-mediated responses were more markedly affected. Although photopic B-wave amplitudes were comparable across groups, 30-Hz flicker responses, indicative of cone pathway integrity, were significantly impaired in CLRN1^−/− rabbits after 3 years (P < 0.01). These findings suggest that rod-driven responses are relatively preserved in CLRN1^−/− rabbits at 3 years, whereas cone function is selectively and progressively impaired.

Figure 3.

Time-dependent changes in retinal function in CLRN1 null rabbits. (a) Representative scotopic and photopic ERG waveforms at 3 years, recorded under dark-adapted conditions for scotopic −24-dB (rod) and 0-dB (combined) flashes and under light-adapted conditions for photopic 0-dB (cone) and 30-Hz (flicker) flashes. (b) Quantitative analysis of the amplitude and implicit time for scotopic and photopic responses from 2 months to 3 years. CLRN1 KO represents CLRN1^−/−. *P < 0.05, **P < 0.01 (Mann–Whitney U test; n = 6).

Delayed Retina Signaling in CLRN1 Null Rabbits

Implicit time analysis of ffERG recordings has identified deficits in synaptic transmission. Under scotopic conditions, CLRN1^−/− animals demonstrated significant delays in B-wave implicit times at −24 dB and 0 dB across multiple ages (P < 0.05 and P < 0.01, respectively) (Fig. 3b), indicating impaired transmission in the rod bipolar pathway within the retina. The A-wave implicit time at 0 dB remained consistent between genotypes, suggesting that photoreceptor hyperpolarization latencies are preserved. In photopic recordings, a significant delay in B-wave implicit time was observed in CLRN1^−/− animals by 7 months of age (P < 0.05), whereas A-wave latencies tended to be longer but did not reach statistical significance. Notably, implicit times for 30-Hz flicker responses were significantly prolonged in CLRN1^−/− rabbits at several time points (P < 0.05 and P < 0.01). Collectively, these findings suggest that CLRN1 deficiency compromises not only cone-driven response amplitudes but also the temporal fidelity of rod- and cone-mediated signaling, particularly affecting inner retinal transmission and cone flicker responsiveness.

Histological Evidence of Photoreceptor Cell Loss

Histological analysis was performed to evaluate the integrity of photoreceptor cells in WT and CLRN1^−/− rabbits at the ages of 7 months, 1.5 years, and 2.5 years (Supplementary Fig. S4). At 7 months, the morphology of photoreceptors appeared similar between the genotypes, and quantitative analysis revealed no significant differences in the nuclear counts of the ONL. By 1.5 years, the retinal architecture remained comparable, with no significant differences in ONL cell numbers. However, by 2.5 years, there was a significant reduction in the number of ONL nuclei in CLRN1^−/− rabbits compared to WT controls, indicating photoreceptor loss (P < 0.05 and P < 0.01, respectively) (Supplementary Figs. S4b–S4d).

CLRN1 Expression Exclusively in Müller Glia

The snRNA-seq of pooled retinal samples from WT (n = 3) and CLRN1^−/− (n = 3) rabbits yielded 6542 high-quality nuclei after quality control filtering was performed. Unsupervised clustering identified seven major retinal cell populations: rod photoreceptors (3679 nuclei), amacrine cells (986), retinal ganglion cells (RGCs; 287), cone photoreceptors (268), bipolar cells (197), horizontal cells (174), and Müller glia (172) (Fig. 4a, Supplementary Fig. S5, Supplementary Dataset S1). Cell type annotations were validated by the expression of canonical markers, including RHO (rods), ARR3 (cones), RBPMS (RGC), GLUL (Müller glia), SLC6A9 (amacrine cells), VSX2 (bipolar cells), and ONECUT1 (horizontal cells) (Supplementary Fig. S5). SNP-based demultiplexing using cellsnp-lite and vireo successfully recovered three biological replicates per condition, enabling a rigorous pseudobulk statistical analysis.

Figure 4.

The snRNA-seq of 10-month-old USH3A rabbit retina. (a) UMAP plot of retinal nuclei from WT rabbits, identifying seven major retinal cell types. (b, c) CLRN1 expression by cell type. Feature plot and dot plot show that CLRN1 was exclusively expressed in Müller glia (23.8% of cells), with negligible expression in rods and cones. This finding identified USH3A as a Müller glia–driven disorder. (d) CLRN1 mRNA levels in Müller glia were comparable between WT (n = 3 animals, 124 cells) and CLRN1^−/− (CLRN1 KO; n = 3 animals, 48 cells) (pseudobulk Wilcoxon test, P = 0.7), indicating that the frameshift mutation does not affect transcript stability.

Analysis of CLRN1 expression across retinal cell types revealed exclusively and sparsely expression in Müller glia (mean log-normalized expression = 0.44, detected in 24% of cells), with negligible expression in rods and cones (Figs. 4b, 4c). Within CLRN1-expressing Müller glia, mean expression was 1.85 (range, 0–2.9), indicating robust expression in this cell population (Fig. 4d). This expression pattern identifies Müller glia as the primary—and exclusive—site of clarin-1 function in the retina. These data also explain the absence of a clarin-1 signal in western blots of WT retina, as the low endogenous expression levels (mean = 0.44 in Müller glia, 23.8% of cells) might fall below the limit of detection of the assay.

Importantly, CLRN1 mRNA levels in Müller glia were comparable between WT (n = 3 animals) and CLRN1^−/− (n = 3 animals) (pseudobulk Wilcoxon test accounting for biological replication, P = 0.7) (Fig. 4d), indicating that the 23-bp frameshift deletion does not trigger nonsense-mediated decay or reduce transcript stability, consistent with the RT-qPCR data (Fig. 1c). The clinically faithful retinal phenotype in CLRN1^−/− animals therefore results from loss of functional clarin-1 protein despite normal mRNA levels.

Cell-Type-Specific Transcriptional Responses

Pseudobulk DESeq2 analysis identified DEGs (adjusted P < 0.05) in six of the seven retinal cell types (Fig. 5). Rod photoreceptors exhibited the greatest number of DEGs (232), followed by amacrine cells (200), cones (68), horizontal cells (45), RGCs (41), and Müller glia (1). No genes reached statistical significance in the bipolar cells. Cross-cell-type expression validation confirmed the specificity of key DEGs to their respective cell types, including CTNNA2 (α-N-catenin) in Müller's glia, TENM2 in cones, and CWF19L2 in rods, verifying that the observed transcriptional changes reflect genuine cell-type-specific biology rather than technical artifacts.

Figure 5.

Differential gene expression analysis revealed cell-type-specific transcriptional responses in CLRN1 null retina. (a–g) Volcano plots showing DEGs in seven retinal cell types (CLRN1^−/− vs. WT, n = 3 animals/group; pseudobulk DESeq2). Blue indicates downregulated (adjusted P < 0.05); red indicates upregulated (adjusted P < 0.05); and gray indicates not significant. Key genes are labeled. (h) Bar chart showing DEG counts per cell type. Müller glia showed only one DEG (CTNNA2), whereas photoreceptors and other neurons showed extensive transcriptional changes (41–232 DEGs), demonstrating non–cell-autonomous degeneration mechanisms. CLRN1 KO represents CLRN1^−/−.

Müller Glia: Primary Site of CLRN1 Dysfunction

Given that CLRN1 is expressed exclusively in Müller glia, the transcriptional changes observed in this cell type are of central mechanistic importance. The only gene that reached significance after multiple testing correction was CTNNA2 (log₂FC = −1.78, adjusted P = 0.0016), a critical component of adherens junctions that links cadherins to the actin cytoskeleton. Downregulation of CTNNA2 suggests potential adhesion defects (anchor failure) at the OLM, a structural barrier formed by Müller glia–photoreceptor junctions. To corroborate this bioinformatic inference, we performed immunohistochemical validation in 20-month-old WT and CLRN1^–/– rabbit retinas. These assays confirmed a pronounced physical reduction of both alpha-N-catenin protein and its associated F-actin scaffolding along the OLM boundary (Supplementary Fig. S6). GO analysis of nominally significant genes (P < 0.05) revealed coherent downregulation of glutamate receptor signaling (adjusted P = 0.031), receptor localization to synapses (adjusted P = 0.031), and GABAergic transmission regulation (adjusted P = 0.034), indicating progressive failure of Müller glia synaptic support functions (Fig. 6a; Supplementary Datasets S2, S3). The limited number of significant DEGs likely reflects the low number of Müller glia nuclei captured per biological replicate (13–52 cells), reducing statistical power despite biologically meaningful changes.

Figure 6.

Pathway-level analysis reveals cell-type-specific functional responses to CLRN1 loss. (a) GO enrichment analysis showing dysregulated biological pathways by cell type (dot size = –log₁₀ P_adj). Colors indicate functional categories: extracellular matrix (ECM)/fibrosis (red), glial support (blue), stress/metabolic (orange), synaptic/visual (green). (b) Heatmap of log₂FC for representative genes across cell types. Color indicates expression change: red, upregulated; blue, downregulated. *P_adj < 0.05, **P_adj < 0.01, ***P_adj < 0.001. Photoreceptors show downregulation of synaptic and visual pathways (**), and inner retinal neurons show compensatory upregulation of stress response genes, consistent with outer retinal vulnerability and inner retinal protection. N = 3 biological replicates per genotype; pseudobulk DESeq2 analysis. CLRN1 KO represents CLRN1^−/−.

Photoreceptor Dysfunction Through Non–Cell-Autonomous Mechanisms

Despite the absence of CLRN1 expression in photoreceptors, both rods and cones exhibited extensive transcriptional alterations. In rods, the most significantly downregulated gene was CWF19L2 (log₂FC = −1.83, adjusted P = 6.4 × 10⁻³⁴), which encodes a pre-mRNA splicing factor, suggesting broad gene regulatory disruption. GO analysis of upregulated rod genes revealed significant enrichment of extracellular matrix organization (adjusted P = 2.4 × 10⁻⁶), connective tissue development (adjusted P = 3.7 × 10⁻⁵), and cartilage development (adjusted P = 7.6 × 10⁻⁶). Combined with the significant downregulation of ciliary machinery and phototransduction components (e.g., CFAP54, GNGT1, CCDC3), these results indicate a pathological reprogramming where rods lose their neural identity and adopt a profibrotic remodeling response to cellular stress (Fig. 6a; Supplementary Datasets S2, S3).

In cones, TENM2 (teneurin-2), which encodes a synaptic adhesion molecule essential for neuronal connectivity, was dramatically downregulated (log₂FC = −5.76, adjusted P = 5.5 × 10⁻¹⁵). GO analysis revealed enrichment of synapse organization (adjusted P = 0.006), sensory perception of light (adjusted P = 0.006), visual perception (adjusted P = 0.026), and retinal cone cell development (adjusted P = 0.028), indicating both synaptic disassembly and impaired phototransduction (Fig. 6a; Supplementary Datasets S2, S3). The downregulation of TENM2 and synaptic genes provides a molecular basis for the early ERG implicit time delays observed in CLRN1^−/− rabbits (Fig. 3).

Convergent HSP90-Mediated Stress Responses in Inner Retinal Neurons

Amacrine cells, RGCs, and horizontal cells, none of which expresses CLRN1, exhibited remarkably convergent cellular stress response patterns (Fig. 6). All three HSP90 family members (HSP90AB1, HSP90AA1, and HSP90B1) were significantly upregulated in amacrine cells (log₂FC = 1.79, 1.47, and 1.38, respectively; all adjusted P < 2.4 × 10⁻¹¹), RGCs (log₂FC = 1.08, 0.90, and 1.02, respectively; all adjusted P < 0.011), and horizontal cells (log₂FC = 1.34, 1.44, and 1.57, respectively; all adjusted P < 0.015) (Fig. 6b; Supplementary Datasets S2, S3). In amacrine cells, this extended to the comprehensive activation of the unfolded protein response, with significant upregulation of additional endoplasmic reticulum chaperones, including HSPA5/BiP (adjusted P = 6.5 × 10⁻⁴), HSPD1 (adjusted P = 9.8 × 10⁻⁴), CLU (adjusted P = 1.2 × 10⁻³), CANX (adjusted P = 4.9 × 10⁻⁵), and PDIA3 (adjusted P = 5.9 × 10⁻⁴). Glycolytic pathway genes (LDHA, ALDOA, GAPDH, ENO2) were also upregulated in amacrine cells (GO adjusted P = 1.4 × 10⁻⁴), and RGCs showed similar metabolic reprogramming with enrichment of aerobic respiration (adjusted P = 7.1 × 10⁻⁵) (Fig. 6; Supplementary Datasets S2, S3).

Critically, this HSP90-mediated proteostatic response was absent in photoreceptors, and none of the three HSP90 family members was significantly upregulated in either rods or cones (Fig. 6b). Bipolar cells showed a consistent directional trend in HSP90 upregulation (HSP90AB1 log₂FC = 1.48, P = 0.004) but did not reach significance after multiple testing correction, likely due to the limited cell numbers. These pathway-level findings reinforce the cell-type-specific results and support a model in which Müller glia dysfunction propagates distinct stress signatures across the retinal network, with fundamentally divergent outcomes for outer versus inner retinal neurons.

Discussion

USH3A presents with progressive retinitis pigmentosa typically manifesting between childhood and late teens.^18–20 Although Clrn1 mutant mice successfully model auditory deficits, they fail to develop retinal degeneration, thus limiting mechanistic studies. The genotypic landscape of USH3A is marked by distinct founder mutations that appear to influence the initial temporal sequence of sensory loss.^18–20 Our CRISPR-engineered CLRN1^−/− rabbit, carrying a frameshift truncating mutation, exhibits progressive dual-sensory deficits that closely mirror human USH3A. The slow disease kinetics in this model—marked by significant structural and functional decline beginning at 20 months—parallels the Finnish USH3A cohort harboring the p.(Tyr176*) truncating variant, where retinitis pigmentosa diagnosis averages 17 years with visual decline continuing into the fourth decade.²⁰ This stands in contrast to the Ashkenazi Jewish N48K missense mutation, which often presents with a “vision-first” clinical course where night blindness may precede detectable auditory threshold shifts.¹⁸ Throughout disease progression, the inner retina remained intact in the rabbit models, consistent with clinical USH3A imaging.^21–23 This progressive phenotype, validated across OCT, ERG, histology, and single-nucleus transcriptomics, establishes the first large-animal platform for mechanistic studies and therapeutic testing in USH3A.

The rabbit's anatomical similarities to human retina, including a visual streak analogous to the macula, likely explain successful disease recapitulation where rodent models failed. The chronological lifespan of the animal is also a critical factor in modeling slow-onset, progressive retinal degenerations such as USH3A. It is highly probable that the mouse visual system does not possess a sufficient biological window for the cumulative metabolic and structural stresses to surpass the threshold of overt cell death. This highlights the primary value of our rabbit model: With a natural lifespan of 8 to 12 years, the rabbit provides the necessary “long runway” to observe the protracted kinetics characteristic of human USH3A (particularly the Finnish founder phenotype).The failure of rodent models likely stems from a combination of these chronological constraints and species-specific differences in Müller glia–photoreceptor metabolic load.

CLRN1 mutant pigs generated via CRISPR/Cas9 represent another critical large-animal platform (Dinculescu A, et al. IOVS 2022;63:ARVO E-Abstract 1926). Compared to the CLRN1 mutant pig model, which shows early rod-specific decline from 6 months, the rabbit model offers advantages for studying temporal signaling deficits through prominent cone vulnerability and delayed ERG implicit times. Furthermore, comprehensive longitudinal data beyond 6 months remain unavailable for the pig model, and the prohibitive cost of multi-year pig colony maintenance favors the rabbit as a more cost-effective platform for studying progressive USH3A retinopathy.

Anchor–Shield Model of USH3A Pathogenesis

Although CLRN1 expression in Müller glia has been reported in human and primate transcriptomic atlases,³ mechanisms linking glial CLRN1 loss to photoreceptor degeneration have remained unknown. Our data support an anchor–shield model integrating three findings (Fig. 7): (1) CLRN1 expression restricted to Müller glia, (2) transcriptional dysregulation across six of seven retinal cell types despite restricted expression, and (3) robust HSP90-mediated stress responses in inner retinal neurons but not photoreceptors.

Figure 7.

The anchor–shield model of USH3A retinal degeneration. Schematic illustration of the proposed three-stage pathogenic cascade in CLRN1-deficient retina. Stage 1 (anchor failure): CLRN1 loss in Müller glia leads to CTNNA2 downregulation (adjusted P = 0.0016), destabilizing adherens junctions at the OLM and disrupting glutamate signaling support. Stage 2 (stress propagation): Non-cell-autonomous stress propagates across the retinal network with divergent responses. Outer retinal photoreceptors, lacking the HSP90 shield, undergo synaptic disassembly (cones: TENM2 log₂FC = −5.76) and fibrotic remodeling (rods: CWF19L2 log₂FC = −1.83). Inner retinal neurons mount a coordinated HSP90 chaperone response (all adjusted P < 0.015). Stage 3 (divergent outcomes): Photoreceptor degeneration (ONL thinning at 2.5–3 years) but inner retinal thickness is preserved. (This image was created with the assistance of the generative AI tool Nano Banana by Google.)

Anchor Component (Müller Glia Dysfunction)

CTNNA2 (α-N-catenin) was the sole differentially expressed gene in CLRN1-deficient Müller glia (log₂FC = 1.78, adjusted P = 0.0016). α-Catenin links cadherins to the actin cytoskeleton and is essential for adherens junction integrity at the OLM, where Müller glia form heterotypic junctions with photoreceptor inner segments.^24–26 CTNNA2 downregulation suggests OLM junction failure, consistent with zebrafish USH3A models showing disrupted N-cadherin distribution and Müller glia microvilli organization.⁸ This structural anchor failure destabilizes the metabolic niche compartmentalizing photoreceptor segments. Pathway analysis additionally revealed downregulated glutamate receptor signaling and synaptic support functions, indicating broader homeostatic dysfunction beyond adhesion defects.

Differential Proteostatic Stress Response (Hypothesized Shield Model)

Despite lacking CLRN1 expression, photoreceptors exhibited extensive transcriptomic dysregulation (232 DEGs in rods, 68 in cones), confirming a non–cell-autonomous pathogenic mechanism. Strikingly, inner retinal neurons—including amacrine cells, RGCs, and horizontal cells—mounted a highly coordinated upregulation of HSP90-family chaperones (HSP90AB1, HSP90AA1, and HSP90B1) alongside unfolded protein response activation and metabolic reprogramming toward glycolysis. While the current data do not definitively establish causality, we hypothesize that this robust inner retinal reaction represents a potential compensatory or stress-response signature. This signature may serve as a molecular shield to buffer cellular stress propagating from dysfunctional Müller glia.^27,28 In stark contrast, vulnerable photoreceptors completely failed to activate this adaptive proteostatic machinery, a disparity that may underlie their selective vulnerability to degeneration. In cones, TENM2 downregulation (a synaptic adhesion molecule critical for cone synapse structure) alongside enrichment of synapse organization pathways indicates synaptic dysfunction, providing a molecular basis for early ERG implicit time delays. In rods, CWF19L2 (pre-mRNA splicing factor) downregulation coupled with ECM organization and fibrotic pathway upregulation represents maladaptive remodeling contributing to progressive ONL thinning.²⁹ Future functional validation will be required to determine whether manipulating this differential stress-response pathway can causally confer neuroprotection.

Although the precise molecular intermediary linking glial CLRN1 deficiency to CTNNA2 transcriptional downregulation remains to be fully elucidated, their structural intersection at the OLM provides a compelling spatial rationale. Clarin-1 is a tetraspan protein localized within specialized plasma membrane microdomains (tetraspanin-enriched microdomains [TEMs]), where it functions as a structural scaffold to stabilize cell-adhesion complexes and organize transmembrane signaling networks. We hypothesize that loss of functional glial CLRN1 destabilizes this specialized junctional architecture at the OLM, triggering either compensatory transcriptional responses or accelerated protein turnover of its intracellular scaffolding partners, particularly α-N-catenin (CTNNA2). Consequently, CLRN1 loss disrupts OLM structural integrity at the molecular level long before overt architectural collapse occurs. Future studies employing proximity-labeling proteomics or live-cell imaging will be essential to delineate the precise molecular pathway and temporal dynamics linking CLRN1 to CTNNA2 regulation.

Distinction From Other Usher Syndrome Subtypes

This non–cell-autonomous, glia-first mechanism fundamentally distinguishes USH3A from USH1 and USH2. USH1 proteins (MYO7A, USH1C, CDH23, PCDH15, USH1G) function primarily in photoreceptor calyceal processes and ribbon synapses,^30,31 whereas USH2 proteins (USH2A, ADGRV1, WHRN) localize to the periciliary membrane complex³²—both representing cell-autonomous photoreceptor defects. Our data indicate that USH3A is primarily a gliopathy where photoreceptor degeneration is secondary to Müller glia dysfunction. This may explain the characteristically slower progression of USH3A, as gradual deterioration of glial support leads to cumulative rather than acute photoreceptor damage.

Translational Impact and Therapeutic Window

The anchor–shield model suggests dual therapeutic strategies: (1) AAV-mediated CLRN1 gene replacement in Müller glia to restore structural and metabolic support (anchor repair), and (2) pharmacological enhancement of photoreceptor stress responses or direct neuroprotection to compensate for absent HSP90 induction (shield augmentation). Combination therapy may prove more effective than either approach alone. The extended pre-symptomatic window (10 months) of the rabbit model defines an optimal therapeutic intervention period before irreversible photoreceptor degeneration occurs.

Our identification of sparse, Müller glia–restricted CLRN1 expression (mean = 0.44, 24% of cells) has critical implications for therapeutic strategy. Recent zebrafish studies revealed that Müller glia–specific, but not photoreceptor-specific, clrn1 re-expression rescues photoreceptor death—validating our target identification.⁸ Critically, the zebrafish work revealed that rescue efficacy correlated with clrn1 expression levels, but excessive expression proved deleterious, causing enhanced cell death in both WT and mutant animals.⁸ This dose-dependent toxicity presents a significant challenge for conventional adeno-associated virus–mediated gene augmentation approaches, which typically drive constitutive transgene expression from strong promoters and cannot recapitulate the sparse, regulated expression pattern we observed (24% of Müller glia). These constraints strongly favor gene editing–based therapeutic strategies, such as CRISPR-mediated gene editing repair, which would restore endogenous CLRN1 expression under native regulatory control, preserving the physiological expression pattern and avoiding overexpression toxicity. The failure of photoreceptor-directed clrn1 expression to provide rescue in zebrafish,⁸ despite photoreceptors being the degenerating cell type, strongly supports our non–cell-autonomous model and indicates that CLRN1 functions through Müller glia–photoreceptor structural interactions, likely involving the adherens junctions we identified as disrupted through CTNNA2 downregulation.

Analytical Rigor and Limitations

Our analytical pipeline—integrating SNP-based demultiplexing, pseudobulk DESeq2 modeling, and cross-cell-type expression validation—ensures data integrity. Cross-validation confirmed that primary DEGs (CTNNA2, TENM2, CWF19L2) were predominantly localized to their identified clusters, providing confidence that the transcriptional changes reflect genuine biological responses rather than technical artifacts. Several limitations warrant consideration. First, limited cell numbers per sample for some populations (particularly Müller glia, bipolar, and horizontal cells) reduced statistical power; the single significant Müller glia DEG likely underestimated true transcriptional changes in this critical cell type. Second, snRNA-seq was performed at a single time point (10 months) prior to significant structural degeneration (observed at 2.5–3 years); temporal transcriptional progression requires additional time-point analysis. Third, functional validation of key candidates (CTNNA2, TENM2, and CWF19L2) is necessary to establish causal relationships within the proposed anchor–shield cascade. Testing the anchor–shield model requires: (1) immunohistochemical and electron microscopic validation of OLM junction integrity, (2) assessment of whether CLRN1 gene therapy restores CTNNA2 expression, and (3) determination of whether HSP90 induction protects photoreceptors in vivo.

Conclusions

The CLRN1^−/− rabbit model recapitulates hallmark features of USH3A-associated dual sensory deficits with progressive hearing and vision loss. Single-nucleus transcriptomics identified Müller glia as the primary therapeutic target and revealed non–cell-autonomous mechanisms of photoreceptor degeneration. This model provides a translational platform bridging basic USH3A research and clinical applications, enabling development of targeted interventions for this form of inherited deaf–blindness.