Disrupted energy metabolism is associated with retinal ganglion cell degeneration in autosomal dominant optic atrophy

Abstract

Autosomal dominant optic atrophy (ADOA) is a hereditary optic neuropathy caused by OPA1 variants, leading to retinal ganglion cell (RGC) degeneration and vision loss. The mechanisms behind RGC vulnerability to mitochondrial dysfunction remain unclear. We developed a patient-specific Opa1^V291D/+ knock-in mouse model to investigate mitochondrial dysfunction and retinal metabolism in ADOA. We observed that Opa1^V291D/+ mice exhibited anatomical and functional RGC abnormalities recapitulating the ADOA phenotypes. Reduced optic atrophy 1 (OPA1) protein levels were noted in Opa1^V291D/+ mice, accompanied by decreased protein stability. Moreover, mitochondrial function was compromised, as indicated by reduced Complex I activity, increased oxidative stress, and diminished adenosine triphosphate production in the retinas of Opa1^V291D/+ mice. Spatial metabolomics revealed energy deficits in the inner retina and heightened glycolysis in the outer retina. Immunostaining showed decreased expression of glycolytic proteins in the ganglion cell layer. Single-nucleus RNA sequencing disclosed significant down-regulation of energy-production genes in RGCs, while other retinal cell types remained unaffected. These findings emphasize the specific vulnerability of RGCs to bioenergetic crises, connecting disrupted energy homeostasis to their degeneration. By increasing the nicotinamide adenine dinucleotide (NAD⁺)/reduced form of NAD⁺ (NADH) redox ratio through the overexpression of mitochondrial-targeted Lactobacillus brevis NADH oxidase (MitoLbNOX) in RGCs, we demonstrated improved RGC function and survival through enhanced energy metabolism and reduced oxidative stress. These findings confirm that disrupted energy metabolism leads to RGC degeneration and emphasize the enhancement of the NAD⁺/NADH redox ratio as a promising treatment strategy to protect RGCs from degeneration in ADOA.

Opa1 variant disrupted mitochondrial energy metabolism and redox homeostasis, triggering RGC degeneration in ADOA.

INTRODUCTION

Autosomal dominant optic atrophy (ADOA) is the most common inherited optic neuropathy, with incidence rates ranging from 1 in 12,000 to 50,000 individuals (1, 2). It is a mitochondrial eye disease primarily characterized by the degeneration of retinal ganglion cells (RGCs) (3). This degeneration leads to progressive vision loss and is associated with variants in the nuclear DNA–encoded OPA1 mitochondrial dynamin like GTPase (OPA1) gene, which compromise mitochondrial function (4–6). Variants in the OPA1 gene alter the optic atrophy 1 (OPA1) protein, a key component of the inner mitochondrial membrane responsible for mitochondrial dynamics and fusion (7). In mammalian mitochondria, the OPA1 protein plays a vital role not only in the regulation of the fusion of the inner mitochondrial membrane but also in the shaping of mitochondrial cristae (7, 8), which are essential for the regulation of mitochondrial respiration, the stabilization of the electron transport chain (ETC), and the maintenance of oxidative stress homeostasis (9, 10). Understanding the impact of the OPA1 variant on RGC is crucial for clarifying the pathogenic mechanisms underlying ADOA.

The impact of OPA1 variants on cells has been studied in previous in vitro research. Those observations revealed that HeLa cells transfected with OPA1 variants exhibited fragmented mitochondria and impaired oxidative phosphorylation (OXPHOS) (11, 12). However, a separate study found that there was no decrease in mitochondrial adenosine triphosphate (ATP) production in ADOA human fibroblasts carrying different OPA1 variants (13), indicating a disparity between these observations and the in vitro findings collected from non-RGC cells. Recent studies demonstrated that introducing the Opa1^K301A and Opa1^R905* variants into mouse RGC cultures resulted in the autophagic degradation of mitochondria and a subsequent decrease in mitochondrial activity content (14). Despite these findings connecting the OPA1 variant to mitochondrial dysfunction, it is still unclear whether the degeneration of RGCs in ADOA is primarily due to a bioenergetic crisis, decreased antioxidant capacity, or a combination of both factors (15, 16).

In vivo models provide advantages over in vitro models regarding the evaluation of the impact of these variants on visual function. Three Opa1 gene–modified ADOA mouse models have been reported. These include mice carrying a nonsense variant (Opa1^Q285STOP) (17–20), a 4–base pair deletion causing a frameshift (Opa1^{c.2708_2711delTTAG}) (21, 22), and a splice-site variant (Opa1^c.1065+5G>A) (23). It is important to emphasize that all three mouse models express a truncated OPA1 protein. Currently, there are no reports of Opa1 mouse models containing missense variants, which are the most common protein-coding mutations identified in patients with ADOA, according to ClinVar and the Leiden Open Variation Database (24, 25). Furthermore, no studies have examined transcriptomes at the single-cell level in the Opa1 mouse model to understand why RGCs are more vulnerable than other retinal cells, especially since this nuclear-encoded protein is expressed universally in all cells. In addition, while photoreceptors have the highest density of mitochondria in the retina (26), the mitochondrial dysfunction associated with ADOA primarily affects RGCs, leaving photoreceptors largely unaffected. This disparity underscores the urgent need to investigate the effects of Opa1 variants on different retinal cells. Comprehensive studies are crucial to uncover the underlying disease mechanisms, identify factors contributing to RGC vulnerability, and develop targeted therapeutic interventions for RGC degeneration.

Nicotinamide has garnered attention in RGC degeneration as it is depleted in the plasma signatures of patients with ADOA and glaucoma (27–30). While nicotinamide adenine dinucleotide (NAD⁺) itself plays a crucial role in glycolysis, the tricarboxylic acid (TCA) cycle, and OXPHOS, the NAD⁺/reduced form of NAD⁺ (NADH) redox ratio serves as a key regulator of cellular energy metabolism and an indicator of cellular stress levels (30–32). Previous studies have demonstrated that oral administration of the NAD⁺ precursor nicotinamide and gene therapy promoting Nmant1 expression, a key NAD⁺-producing enzyme, halted RGC degeneration in the DBA/2J mouse model of glaucoma (33–35). Although these promising results highlight the antioxidant properties of vitamin B3 and its role in NAD⁺ synthesis, it remains uncertain whether similar strategies to increase NAD⁺ levels could enhance energy metabolism and support RGC survival in the Opa1 mouse model. Moreover, it is unclear whether directly converting NADH to NAD⁺ to boost the NAD⁺/NADH redox ratio would be an effective and efficient strategy for protecting RGCs.

In this study, we developed a novel mouse model by introducing a patient-derived missense variant of the OPA1 gene to investigate the pathophysiology of ADOA. We evaluated whether the model accurately replicated the clinical phenotypes of ADOA, focusing on functional deficits, anatomical alterations, OPA1 protein characteristics, and mitochondrial phenotypes. To evaluate the effect of the Opa1 variant on mitochondrial function, we analyzed energy metabolism and oxidative stress throughout the retina. Immunostaining and spatial metabolomics were used to assess histological changes and metabolic adaptations, particularly in the ganglion cell layer where RGCs reside. We used single-nucleus RNA sequencing (snRNA-seq) to uncover transcriptomic changes linked to ADOA at the single-cell level, aiming to identify the mechanisms contributing to the selective vulnerability of RGCs in ADOA. In addition, we examined the impact of increasing the NAD⁺/NADH redox ratio in RGCs on their survival in our ADOA mouse model.

RESULTS

Clinical and genetic profile of the patient with ADOA carrying an OPA1 missense variant

A 32-year-old woman visited our institution with a history of gradually declining vision. The results of an eye examination performed on this patient are displayed in Fig. 1. Fundus imaging indicated temporal pallor of the optic disc (Fig. 1A). Optical coherence tomography (OCT) of the optic disc revealed a reduction in retinal nerve fiber layer (RNFL) thickness (Fig. 1B). Electrophysiological testing showed normal rod and cone responses on full-field electroretinography (ERG), albeit with reduced pattern ERG (PERG) responses (Fig. 1C). Genetic testing confirmed the presence of a heterozygous variant in the OPA1 gene, i.e., c.1037T>A, p.V346D (NM_130837.3), thereby confirming the diagnosis of ADOA. This variant is classified as a missense variant, which is the most common type of mutation in ADOA, according to reports in ClinVar and the Leiden Open Variation Database (24, 25). This OPA1 missense variant was recently submitted by a reporter to the ClinVar database (ID: 447907). The OPA1^V346D variant is classified as likely pathogenic, on the basis of aggregated data from public databases, following American College of Medical Genetics and Genomics guidelines (table S1).

Fig. 1. Optic atrophy and visual function impairment in a patient with ADOA and the generation of an Opa1^V291D/+ mouse model.

(A) Fundus photography showing temporal disc pallor in the left eye, representative of both eyes. (B) OCT demonstrating decreased RNFL thickness, averaging 69.8 μm in the left eye. (C) Full-field ERG indicating normal rod and cone responses, with decreased PERG responses. (D) Targeting strategy used for generating the Opa1^V291D/+ knock-in mouse, with primers (F1 and R1) designed to detect exon 9 of Opa1. (E) Genotyping results for Opa1^+/+ and Opa1^V291D/+ tissues using the indicated primers. The knock-in allele includes an additional 83 nucleotides compared with the wild-type (WT) allele, incorporating the LoxP site and adjacent sequences. (F) Sequencing of the region between the indicated primers confirming the heterozygous T-to-A variant. (G) Body weight measurements of mice at different ages (total n = 306; independent t tests P = 0.8096, 0.3582, 0.2501, 0.0191, 0.0011, <0.0001, and < 0.0001 at P30, P90, P120, P180, P270, P360, and P450, respectively). Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ****P < 0.0001. n.s., not significant; Ex, exon; NeoR, neomycin resistance.

Generation and characterization of a patient-specific knock-in ADOA mouse model (Opa1^V291D/+)

Because of the absence of OPA1 mouse models carrying missense variants, we developed a patient-specific knock-in mouse model (Opa1^V291D/+) carrying a V291D variant equivalent to the V346D variant found in our patient with ADOA (Fig. 1D). The resulting knock-in mouse harbored the Opa1 c.871T>A variant, which changes the 291st amino acid of OPA1 from valine to aspartic acid. Mice that were homozygous for this variant exhibited embryonic lethality, consistent with observations in other ADOA mouse models. We verified the presence of the V291D variant through the polymerase chain reaction (PCR) amplification of exon 9 using forward and reverse primers (table S2), which confirmed variant heterozygosity in the mutant mice (Fig. 1E); this was further validated using Sanger sequencing (Fig. 1F). The mutant mice had lower body weights than their WT littermate controls after 180 days (Fig. 1G). Moreover, the mutant mice exhibited a hunched-back posture (fig. S1), which was suggestive of an illness or aging condition associated with the specific variant.

The visual functional phenotype of the novel Opa1^V291D/+ mice recapitulated the features of ADOA

To analyze the visual functional presentation of the Opa1^V291D/+ mouse model, we performed several electrophysiological tests, including PERG, photopic negative responses (PhNRs), scotopic threshold response (STR), and serial-intensity scotopic and photopic flash ERGs. At 180 days, the PERG revealed a significant decrease in amplitude between P1 and N2, which continued to decline up to 630 days, indicating the presence of a degenerative process in this mouse model (Fig. 2A). Opa1^V291D/+ mice exhibited a significantly reduced PhNR amplitude at 180 days (Fig. 2B). In terms of STR, a significant reduction in the negative STR was observed at all three intensities (−5.6, −5.3, and −5.0 log cd·s/m²) at 180 days (Fig. 2C). The scotopic and photopic ERGs did not display significant differences in both a- and b-wave amplitudes across all intensities between the Opa1^V291D/+ and their littermate-control WT mice at 360 days (Fig. 2D). Our Opa1^V291D/+ mice exhibited abnormal results in electrophysiological tests specific to RGC, whereas the function of photoreceptors remained unaffected. These findings were consistent with those observed in human patients with ADOA.

Fig. 2. Opa1^V291D/+ variant in mice recapitulates the RGC-specific visual function deficits of patients with ADOA.

(A) Representative PERG recordings showing the amplitude measured from N2 to P1 (total n = 161; independent t tests P = 0.8478, 0.0097, 0.0021, <0.0001, 0.0265, and 0.0462 at P90, P180, P270, P460, P450, and P630, respectively). (B) Representative PhNR recordings showing the amplitude measured from the baseline to the trough (total n = 15; independent t tests P = 0.0026). (C) Representative STR recordings showing the amplitude measured from the baseline to the positive STR (pSTR) and negative STR (nSTR) (total n = 51; independent t tests P = 0.0021, <0.0001, and 0.0006 at nSTR –5.6, −5.3, and −5.0 log cd·s/m², respectively; P = 0.6499, 0.9336, and 0.9042 at pSTR –5.6, −5.3, and −5.0 log cd·s/m², respectively). (D) Representative serial scotopic and photopic ERG recordings at different intensities at 360 days (total n = 9; linear regression model P for interaction = 0.420, 0.887, 0.201, and 0.117 in scotopic a-wave, photopic a-wave, scotopic b-wave, and photopic b-wave, respectively). Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Reduction in RNFL thickness and RGC count in the retinas of the Opa1^V291D/+ mice

In addition to assessing their functional phenotype, we used in vivo spectral domain (SD)–OCT and immunostaining to elucidate the anatomical phenotype in the retinas of the Opa1^V291D/+ mice. The SD-OCT examination revealed a decreased RNFL thickness in Opa1^V291D/+ mice compared with their littermate WT controls. This was observed in both female and male mice at 90 days (Fig. 3A). The number of RGCs was examined by immunostaining of whole-mounted retinas using an anti-BRN3A antibody, as shown in Fig. 3B. This analysis revealed a reduction in RGC numbers in Opa1^V291D/+ mice compared with their littermate WT controls, which was correlated with the decrease in RNFL thickness. Notably, the RGC counts in mutant mice were significantly decreased after 180 days and continued to decline up to 420 days.

Fig. 3. The Opa1^V291D/+ variant leads to RGC loss and mitochondrial ultrastructure alterations in the retina and optic nerve.

(A) SD-OCT at 90 days (total n = 32; independent t tests P = 0.0360, 0.0197, and 0.0228 in male, female, and total groups, respectively). (B) Representative images showing BRN3A-positive RGC counts in 20 squares from three different zones of a whole-mounted retina. Analysis of the RGC counts per 20 squares at 180, 360, and 420 days (n = 4, 6, and 3 mice per group at P180, P360, and P420, respectively; independent t tests P = 0.0105, 0.0031, and 0.0028 at P180, P360, and P420, respectively). (C) Representative image showing confocal microscopy with super-resolution imaging of the optic nerves. Violin plot of the mitochondrial sphericity in the optic nerves at 360 days (n = 4 mice in each group; independent t tests P = 0.0356). (D) Representative images showing optic nerve ultrastructure in TEM. Analysis of the number of myelinated axons in optic nerves at 50 days (n = 3 mice in each group; independent t tests P = 0.0009) and 360 days (n = 4 mice in each group; independent t tests P < 0.0001). (E) Separation of the inner mitochondrial membranes, loss of cristae, and mitochondrial vacuolation were also observed in Opa1^V291D/+ mice. Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

The optic nerves of Opa1^V291D/+ mice showed alterations in axonal and mitochondrial structure

To examine in greater detail the anatomical features of the myelinated sheath and mitochondrial morphology in the optic nerve, which contains the axons of RGCs, super-resolution imaging, including spinning disk confocal microscopy (SDCM) with super-resolution radial fluctuations (SRRFs) and transmission electron microscopy (TEM), was applied to mouse optic nerves. Our SDCM with SRRF imaging analysis detected the presence of altered mitochondrial shapes in Opa1^V291D/+ mice at 360 days, which exhibited a greater number of spherical and less variable mitochondria compared with control mice (Fig. 3C), indicating the presence of mitochondrial fragmentation caused by impaired mitochondrial fusion. To further delineate regional differences in mitochondrial dynamics, we performed additional imaging to assess mitochondria across the prelaminar region, the unmyelinated optic nerve head, and the myelinated optic nerve (36). Increased mitochondrial sphericity was consistently observed in mutant mice across all three regions (fig. S2A). TEM analysis revealed a loosened myelinated sheath and a significantly reduced number of myelinated axons in mutant mice at ages 50 and 360 days (Fig. 3D), reflecting chronic RGC degeneration (37). Moreover, TEM imaging of the optic nerve revealed changes in mitochondrial morphology in the Opa1^V291D/+ mice, including the separation of the inner mitochondrial membranes, the loss of cristae, and mitochondrial vacuolation (Fig. 3E). Examination of RGC somata in the ganglion cell layer also revealed disrupted mitochondria and the accumulation of mitophagosomes (fig. S2B). To determine whether mitochondrial genomic alterations accompanied these structural abnormalities, we analyzed mitochondrial DNA (mtDNA) copy number and integrity in retinal tissues. Quantitative PCR (qPCR) revealed a significant increase in mtDNA copy number in Opa1^V291D/+ retinas compared with WT controls, possibly reflecting impaired fusion, accumulation of mitophagosome, and compensatory mitochondrial turnover (fig. S3A). In contrast, qPCR-based mtDNA damage assays and long-extension PCR showed no detectable differences in mtDNA deletions or damage between Opa1^V291D/+ and WT mice (fig. S3B). These results indicate that, although mtDNA copy number is elevated, the overall integrity of the mitochondrial genome remains intact, suggesting that the observed mitochondrial defects are primarily functional and structural rather than due to mtDNA instability. Together, these findings indicate that Opa1^V291D/+ mice exhibit reduced RGC numbers in the retina, thinner RNFL on SD-OCT, fewer myelinated axons in the optic nerve, and mitochondrial structural abnormalities accompanied by increased mtDNA copy number but preserved mtDNA integrity. These combined changes demonstrate RGC degeneration associated with impaired mitochondrial fusion and respiratory dysfunction.

Opa1^V291D/+ mice showed decreased stability of the OPA1 protein

To investigate whether the V291D variant affects the expression of the OPA1 protein in the retina, we performed Western blot analyses, which revealed a significantly reduced level of the OPA1 protein in the retinas of Opa1^V291D/+ mice (Fig. 4A). In contrast, qPCR analyses revealed no differences in Opa1 mRNA expression in retinal cells between the mutant and littermate-control WT mice, indicating that the down-regulation of OPA1 protein levels was not due to a decrease in the transcription of the corresponding gene (Fig. 4B). Because reduced protein levels are often linked to protein destabilization and degradation via the ubiquitin-proteasome system (38), we further examined OPA1 protein expression and ubiquitination in Opa1^WT- and Opa1^V291D-transfected human embryonic kidney (HEK) 293 cells. In the presence of N-carbobenzyloxy-l-leucyl-l-leucyl-l-leucinal (MG132) (25 μM), which is a proteasome inhibitor, we observed polyubiquitinated OPA1 substrates in the lysates of Opa1^V291D-transfected cells (Fig. 4C), suggesting that increased degradation through the ubiquitin-proteasome pathway contributed to the down-regulation of the OPA1 protein. Moreover, treating HEK293 cells with MG132 (25 μM) for 4 and 6 hours resulted in a mild restoration of OPA1 accumulation after 6 hours, especially the short form of the protein, in Opa1^V291D-transfected cells (Fig. 4D), indicating that the ubiquitin-proteasome pathway contributes to, but does not fully account for, OPA1 depletion. To further assess the impact of the V291D variant on OPA1 isoform processing, we analyzed the ratio of long OPA1 (l-OPA1) and short OPA1 (s-OPA1) in retinal lysates. Both isoforms were significantly reduced in Opa1^V291D/+ mice compared with WT controls, with a disproportionately greater decrease in the short (soluble) form (fig. S4). This pattern suggested that the V291D variant caused reduced overall OPA1 protein stability and impaired proteolytic processing, leading to selective depletion of s-OPA1. Because s-OPA1 acts together with l-OPA1 in crista remodeling, mitochondrial fusion, and restoration of energy efficiency (39), its preferential loss likely aggravates crista disorganization and compromises OXPHOS efficiency. Together, these results demonstrate that the Opa1^V291D variant leads to decreased protein stability, enhanced proteasomal degradation, and impaired isoform processing, resulting in reduced OPA1 function. This combination of effects provides a mechanistic link between the mutation, disrupted mitochondrial structure, and the OXPHOS dysfunction underlying RGC degeneration in ADOA.

Fig. 4. Decreased OPA1 protein levels in the Opa1^V291D/+ mouse retinas and reduced OPA1 protein stability in Opa1^V291D-transfected cells.

(A) Western blot (WB) showing the OPA1 protein expression in retinas (n = 6 in each group, independent t tests P < 0.0001). (B) qPCR of the Opa1 mRNA expression in retinas (n = 6 in each group, independent t tests P = 0.7744). (C) Lysates from HEK293 cells transfected with Opa1^WT and Opa1^V291D were treated with MG132. Immunoprecipitation (IP) revealed the presence of polyubiquitinated OPA1 in the Opa1^V291D-transfected cells. (D) Levels of the OPA1 protein after treatment with MG132 (25 μM) at baseline, 4 hours, and 6 hours in the Opa1^V291D-transfected HEK293 cells (n = 3 in each group; one-way analysis of variance (ANOVA) with Tukey’s test P = 0.9431 and 0.0241 in 0 versus 4 hours and 0 versus 6 hours). Data are presented as means ± SEM. *P < 0.05, ****P < 0.0001.

Reduction of mitochondrial Complex I activity (NADH/ubiquinone oxidoreductase) in the retinas of Opa1^V291D/+ mice

On the basis of our examination of the changes in the shape and structure of mitochondria in Opa1^V291D/+ mice, we investigated how these alterations affect mitochondrial function in the retinas of these mice. Specifically, we performed tests to measure mitochondrial respiration and ATP hydrolysis in the retinas using frozen tissue samples [referred to as the respirometry in frozen sample (RIFS) and hydrolysis in frozen sample (HyFS) assays, respectively (Fig. 5A) (40–42). The results of these assays revealed a significant decrease in the activity of Complex I in terms of both the protein-normalized and the MitoTracker Deep Red (MTDR)–normalized oxygen consumption rates in Opa1^V291D/+ mice (Fig. 5B). This normalization helped account for potential variations in mitochondrial content between samples, thereby ensuring that the observed differences reflect true functional deficits rather than changes in mitochondrial abundance. In addition to Complex I dysfunction, we also observed a decrease in Complex IV activity, as evidenced by the increased Complex II/IV activity ratio without significant change in Complex II activity in Opa1^V291D/+ mice (Fig. 5C). This suggests that both Complex I and IV activities are diminished in Opa1^V291D/+ retinas, consistent with the role of OPA1 in maintaining mitochondrial crista integrity, which is essential for the stability and function of respiratory complexes. The altered ratio further indicates a compensatory adjustment in the respiratory chain to preserve energy production despite dual impairment. In the HyFS assay, a trend toward reduced protein-normalized ATP hydrolytic capacity was observed in Opa1^V291D/+ mice (Fig. 5D), although this result did not reach statistical significance. These findings underscore the presence of ETC defects with Complex I dysfunction in the retina of Opa1^V291D/+ mice.

Fig. 5. Mitochondrial dysfunction, oxidative stress, reduced energy production, and glycolytic shift in Opa1^V291D/+ mouse retinas.

(A) Representative bioenergetic profile, as determined using the RIFS protocol in frozen retinas. (B) Optimized RIFS analysis of mitochondrial Complex I, II, and IV activities normalized to total protein and mitochondrial content (MTDR; n = 6 per group). (C) Ratios of Complex I/IV, II/IV, and I/II activities. Optimized RIFS results normalized to total protein and mitochondrial content using MTDR (n = 6 per group). (D) ATP hydrolytic capacity assessed by HyFS (n = 6 per group) (E) The GSH/GSSG ratio and total GSH level in retinal lysate (n = 7 per group. (F) The SOD activity in mouse retinas (n = 7 per group). (G) Representative immunostaining of 4-HNE in retinal sections showing increased fluorescence intensity in the Opa1^V291D/+ mouse retina, particularly in the ganglion cell layer. Bar chart of the 4-HNE fluorescence intensity in retinal immunostaining (n = 5 per group). (H) The NAD⁺/NADH ratio, the quantity (picomol) of NAD⁺ per amount (milligram), and the quantity (picomol) of NADH per amount (milligram) of protein in mouse retinas (n = 6 per group). (I) The quantity (nmol) of ATP per amount (milligram) of protein in mouse retinas (n = 6 per group). (J) The level of lactate per amount (milligram) of protein in mouse retinas (n = 5 per group). (K) Western blot of the phospho-PFKFB3, phospho-GLUT1, HK1, and HK2 in mouse retinas lysates with quantification (n = 6 to 8 per group). Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001. AA, antimycin A; Rot, rotenone; Asc, ascorbate.

Reduced antioxidant capacity and increased oxidative stress in the retinas of Opa1^V291D/+ mice

Because of the potential for the induction of oxidative stress by Complex I impairment (43), we also examined the antioxidant and oxidative stress profiles in mouse retinas. The glutathione (GSH) levels and the ratio of GSH to its oxidized form (GSSG) were significantly lower in Opa1^V291D/+ retinas (Fig. 5E), indicating a disrupted redox balance and heightened oxidative stress. In addition, superoxide dismutase (SOD) activity was significantly lower in the Opa1^V291D/+ retinas (Fig. 5F), suggesting a diminished antioxidant capacity that may worsen oxidative stress by facilitating the accumulation of superoxide radicals. Immunostaining of retinal tissues for 4-hydroxynonenal (4-HNE), which is a crucial marker of oxidative stress (44), showed significantly elevated 4-HNE levels in Opa1^V291D/+ mice, particularly in the inner retina (Fig. 5G), further confirming the accumulation of oxidative stress. These findings link oxidative stress with the defective ETC and impaired Complex I activity observed in Opa1^V291D/+ retinas.

Decreased NAD⁺/NADH redox ratio and ATP levels but increased glycolysis in the retinas of Opa1^V291D/+ mice

Given the defects in ETC observed in Opa1^V291D/+ mouse retinas, we next examined retinal energy metabolism by measuring NAD⁺ and its reduced form (NADH) and ATP levels using assay kits. The results indicated that the NAD⁺/NADH redox ratio and NAD⁺ levels were reduced in Opa1^V291D/+ mouse retinas, whereas NADH levels remained comparable between the mutant and control groups (Fig. 5H), which was consistent with the impairment in Complex I activity noted in Opa1^V291D/+ mice. The ATP levels were significantly decreased in Opa1^V291D/+ mouse retinas (Fig. 5I), thus corroborating ETC dysfunction and the resulting bioenergetic crisis within the retina. Since glycolysis serves as an alternative energy source when OXPHOS is impaired in the retina (45, 46), we further assessed lactate levels in mouse retinas. Lactate assays revealed increased lactate production in Opa1^V291D/+ retinas (Fig. 5J), indicating an adaptive metabolic response. In addition, immunoblot analysis demonstrated significant up-regulation of phospho–6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 3 (PFKFB3), phospho–glucose transporter 1 (GLUT1), hexokinase 1 (HK1), and HK2 proteins in Opa1^V291D/+ retinas (Fig. 5K), suggesting a metabolic shift toward glycolysis to compensate for impaired ETC function.

Decreased ATP with accumulation of adenosine monophosphate in the inner retinas, while increased glycolytic metabolites in the outer retinas of Opa1^V291D/+ mice

To further characterize metabolic alterations in Opa1^V291D/+ mouse retinas, we conducted matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) analysis, which revealed substantial ATP depletion and adenosine monophosphate (AMP) accumulation, particularly in the inner retinal layers where RGCs reside, indicating a severe energy crisis in these regions (Fig. 6A). In contrast, MALDI results demonstrated significantly elevated signal intensities of glycolysis metabolites, including glucose-6-phosphate (G6P) and pyruvate, predominantly in the outer retinal layers, where photoreceptors are located (Fig. 6B). These findings suggest a metabolic shift toward glycolysis as a compensatory mechanism in response to energy deficits in Opa1^V291D/+ retinas, particularly in the photoreceptor-rich outer retina, while the inner retinal layers, including RGCs, do not exhibit this change.

Fig. 6. Reduced energy production and metabolic shift toward glycolysis in Opa1^V291D/+ mouse retinas, with decreased glycolytic activity in the ganglion cell layer.

(A) Representative hematoxylin and eosin (H&E)–stained retinal sections, corresponding MALDI MS images, and manual image segmentation from WT and Opa1^V291D/+ mice at 180 days. Bar charts of ATP signal intensity in positive ion mode (n = 3 per group; independent t test P = 0.0423, 0.0496, and 0.0649 in whole retina, inner retinal layer, and outer retinal layer) and AMP signal intensity in negative ion mode (P = 0.0451, 0.0280, and 0.0697). (B) Representative MALDI MS images of G6P and pyruvate in WT and Opa1^V291D/+ mouse retinas at 180 days. Bar chart of G6P signal intensities in negative ion mode (P = 0.0188, 0.0744, and 0.0260 in whole retina, inner retinal layer, and outer retinal layer) and pyruvate signal intensities in negative ion mode (P = 0.0335, 0.0502, and 0.0367). (C) Representative immunostaining of phospho-AMPKα, phospho-PFKFB3, phospho-GLUT1, HK1, LDHB, and IDH3 in retinal sections from WT and Opa1^V291D/+ mice. Bar charts of the fluorescence intensity of phospho-AMPKα (n = 5 per group; independent t test P = 0.0013 and 0.0396 in ganglion cell layer and photoreceptor layer, respectively), phospho-PFKFB3 (P = 0.0454 and 0.0029), phospho-GLUT1 (P = 0.0595 and 0.0029), HK1 (P = 0.0018 and 0.0394), LDHB (P = 0.0038 and 0.8227), and IDH3 (P = 0.0006 and 0.3595) in mouse retinas. Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Reduced glycolytic activity in the ganglion cell layer contrasted with the photoreceptor layer

To assess cellular responses to the bioenergetic crisis at the histological level, we analyzed phospho–AMP-activated protein kinase α (AMPKα) expression using immunostaining. We observed increased phospho-AMPKα fluorescence intensity in both the ganglion cell and photoreceptor layers of Opa1^V291D/+ retinas, indicating AMPK pathway activation under metabolic stress (Fig. 6C). Previous studies have demonstrated distinct preferences regarding the energy metabolism between the retinal layers, with outer retinal layers relying on glycolysis to compensate for ATP deficiencies (45–48), whereas inner retinal cells, including RGCs, primarily depend on mitochondrial ETC and OXPHOS, exhibiting lower glycolytic activity (45, 46). Given these differences and building on the results of our immunoblot analysis, which indicate a glycolytic shift in retinal metabolism in response to a bioenergetic crisis, we further investigated the expression of glycolytic enzymes at the histological level to evaluate metabolic changes across different retinal layers. Immunostaining revealed a significantly reduced fluorescence intensity for phospho-PFKFB3, HK1, lactate dehydrogenase B (LDHB), and isocitrate dehydrogenase 3 (IDH3) in the ganglion cell layer of Opa1^V291D/+ mice. In contrast, the fluorescence intensities of phospho-PFKFB3, phospho-GLUT1, and HK1 were significantly elevated in the photoreceptor inner and outer segment layers (Fig. 6C). These findings suggest that, in response to the bioenergetic crisis caused by defective ETC, the compensatory energy metabolism via glycolysis and TCA cycle was impaired in the ganglion cell layer, where RGCs reside. This disruption of energy homeostasis in the ganglion cell layer may suggest the selective vulnerability of RGCs in Opa1^V291D/+ mice.

snRNA-seq and spatial transcriptomics revealed the down-regulation of energy metabolism–related genes in the RGCs of Opa1^V291D/+ mice

To understand further the molecular mechanisms underlying ADOA at the single-cell resolution, we used snRNA-seq to analyze transcriptomic changes in RGCs and other retinal cell types between Opa1^V291D/+ and WT mouse retinas at 360 days. In a total of 19,315 nuclei, the snRNA-seq and unsupervised clustering analysis identified 10 clusters corresponding to nine retinal cell types, as assessed on the basis of the expression of specific cell markers (table S3), together with an additional cluster comprising other cells, as shown in Fig. 7A. Two distinct RGC clusters, RGC-1 and RGC-2, were identified. A comparative analysis revealed that both clusters expressed pan-RGC markers, including Rbpms, Slc17a6, and Thy1 (49, 50). There were no statistically significant differences between RGC-1 and RGC-2 in the expression of Pou4 genes, despite minor variations in relative expression levels observed in the plots (Fig. 7B). The differential gene expression analysis of these 10 clusters showed significant down-regulation of genes linked to the ETC, Complex I biogenesis, and glycolysis, particularly in the RGC-2 cluster of Opa1^V291D/+ mice compared with the littermate WT mice (Fig. 7C). In contrast, no significant changes in the expression of genes associated with energy-production pathways were detected in other retinal cell types, such as rods and cones, between Opa1^V291D/+ and WT mice. A dot plot analysis displayed the log expression and percent expression of genes related to ETC and glycolysis across different cell types in Opa1^V291D/+ and WT mice (Fig. 7D). It revealed that RGC-2 had high energy demands in WT mice and showed more pronounced differences between Opa1^V291D/+ and WT compared with the other cell clusters. These snRNA-seq findings aligned with the results of the spatial metabolomics and immunostaining reported above, which indicated energy depletion and impaired glycolysis predominantly in the ganglion cell layer. In addition, the snRNA-seq revealed decreased expression of genes related to mitophagy and autophagy pathways specifically in the RGC-2 cluster (fig. S5A), whereas these pathways were preserved in photoreceptors and other retinal cells. To spatially validate these observations, we performed high-resolution spatial transcriptomics on mouse retinal sections. Consistent with the snRNA-seq data, spatial transcriptomic analysis revealed markedly reduced expression of ETC- and glycolysis-related genes in cells within the ganglion cell layer of Opa1^V291D/+ retinas compared with WT (Fig. 7, E and F). Spatial transcriptomics also confirmed decreased expression of autophagy-related genes specifically in RGC-rich regions, whereas mitophagy-related transcripts showed a downward trend but did not reach statistical significance (fig. S5B). These pathway-specific deficits were not observed in photoreceptors. Together, these integrated transcriptomic datasets demonstrate that RGCs exhibit coordinated down-regulation of ETC, glycolysis, and mitochondrial turnover pathways. This cell type–specific impairment in metabolic and mitochondrial quality-control responses likely contributes to the selective vulnerability of RGCs in Opa1^V291D/+ mice and underlies their progressive degeneration in ADOA.

Fig. 7. Down-regulation of genes involved in the ETC, complex I biogenesis, and glycolysis in RGCs of Opa1^V291D/+ mice.

(A) A cluster analysis of the results from snRNA-seq of retinal cells from WT and Opa1^V291D/+ mice at 360 days identified 10 retinal cell types, including two distinct RGC clusters, RGC-1 and RGC-2, via unsupervised clustering. (B) A heatmap of the RGC markers in the RGC-1 and RGC-2 clusters in WT and Opa1^V291D/+ mice. Both clusters expressed pan-RGC markers, with no statistically significant differences in Pou4 markers between the clusters. (C) Heatmaps of pathway analyses highlighting multiple down-regulated genes in RGC-2 from Opa1^V291D/+ mice compared with WT controls, particularly in pathways related to ETC (n = 5 mice per group; adjusted P < 0.0001, q < 0.0001, WikiPathways database), Complex I biogenesis (adjusted P < 0.0001, q < 0.0001, REACTOME database), and glycolysis (adjusted P = 0.0081, q = 0.0064, REACTOME database). (D) A dot plot illustrating differential gene expression in the ETC and glycolysis pathways across various retinal cell types. RGC-2 displayed more significant differences in gene expression between Opa1^V291D/+ and WT mice compared with other retinal cell clusters. (E) Representative image of high-resolution spatial transcriptomics analyzed using QuPath cell segmentation. Cells from the ganglion cell layer (GCL) were selected and clustering distinguished GCL-derived cell populations in WT and Opa1^V291D/+ retinas at 280 days. (F) Heatmaps from spatial transcriptomic pathway analysis showing decreased expression of ETC genes (adjusted P = 0.0079, q = 0.1215; WikiPathways database) and glycolysis genes (adjusted P < 0.0001, q < 0.0001; WikiPathways database) in RGC-rich regions of Opa1^V291D/+ retinas.

Enhanced RGC function and survival in Opa1^V291D/+ mice following MitoLbNOX overexpression

Considering that our Opa1^V291D/+ mouse model displayed Complex I dysfunction, a reduced NAD⁺/NADH redox ratio, elevated oxidative stress, and decreased ATP production, we investigated whether increasing the NAD⁺/NADH redox ratio could promote RGC survival in our Opa1^V291D/+ mice. To increase the NAD⁺/NADH ratio, our approach was to use Lactobacillus brevis (Lb) NOX (51), a bacterial water-forming NADH oxidase, to directly increase NAD⁺ by oxidization of NADH to NAD⁺. Both LbNOX and MitoLbNOX, the latter containing the mitochondrial targeting sequence, have been shown to lower cytosolic NADH levels in HeLa cells, as demonstrated by the SoNar sensor and lactate/pyruvate ratios (51). However, MitoLbNOX not only had a more significant effect on the mitochondrial NAD⁺/NADH ratio but also doubled the total cellular NAD⁺/NADH ratio, while LbNOX does not significantly affect the total cellular NAD⁺/NADH ratio because most of the NADH within the cell is located in the mitochondria, and MitoLbNOX directly acts in this compartment (51, 52). Therefore, we used MitoLbNOX to effectively boost the NAD⁺/NADH redox ratio in RGC mitochondria. We generated Opa1^V291D/+; Rosa26^{LSL-MitoLbNOX/+} (hereafter, V291D-MitoLbNOX) mice that could conditionally overexpress MitoLbNOX when crossed with an RGC-specific Cre reporter line (Vglut2^Cre;Rosa26^LSL-MitoTag; hereafter VG2-MitoTag) (Fig. 8A). To verify the specificity of Cre-loxP–mediated conditional overexpression, we examined green fluorescent protein (GFP) expression within the MitoTag cassette in VG2-MitoTag mice, confirming localized GFP expression in RGCs (Fig. 8B). GFP expression remained stable in both Opa1^V291D/+;Vglut2^Cre/+;Rosa26^{LSL-MitoTag/LSL-MitoLbNOX} (V291D-VG2-MitoTag-MitoLbNOX) and Opa1^V291D/+;Vglut2^Cre/+;Rosa26^{LSL-MitoTag/+} (V291D-VG2-MitoTag) mouse retinas (Fig. 8B). Next, we analyzed the functional outcomes and survival of RGCs from the V291D-VG2-MitoTag-MitoLbNOX and V291D-VG2-MitoTag mice. PERG recordings at 180 days demonstrated significantly larger amplitudes in the V291D-VG2-MitoTag-MitoLbNOX mice compared with their littermate control V291D-VG2-MitoTag mice, indicating improved RGC function (Fig. 8C). In addition, whole-mounted retina immunostaining revealed a greater count of RGCs per peripheral square in V291D-VG2-MitoTag-MitoLbNOX mice (Fig. 8D), suggesting that MitoLbNOX overexpression enhances RGC survival. On the basis of prior studies, boosting the NAD⁺/NADH redox ratio through the overexpression of MitoLbNOX could improve energy metabolism via the TCA cycle while also playing a crucial role in oxidative stress regulation (31, 53). Thus, we evaluated TCA cycle activity and oxidative stress levels at the histological level. Immunostaining of the retinal section revealed elevated pyruvate dehydrogenase E1 component (PDHE1) and IDH3 expression in the ganglion cell layer of V291D-VG2-MitoTag-MitoLbNOX mice, indicating heightened TCA cycle activity. In addition, the fluorescence intensity of 4-HNE in the ganglion cell layer significantly decreased in V291D-VG2-MitoTag-MitoLbNOX mice, indicating a reduction in oxidative stress following MitoLbNOX overexpression (Fig. 8E). To further explore whether the integrated stress response (ISR) contributes to the pathological phenotype, we performed additional immunofluorescence staining for eukaryotic translation initiation factor 2A (eIF2α), phosphorylated eIF2α (p-eIF2α), and activating transcription factor 4 (ATF4). No significant differences in either marker were detected among WT, V291D-VG2-MitoTag, and V291D-VG2-MitoTag-MitoLbNOX retinas, suggesting that the canonical ISR pathway is not prominently activated under these conditions. In contrast, nuclear factor erythroid 2-related factor 2 (NRF2) expression was markedly reduced in the ganglion cell layer of V291D-VG2-MitoTag retinas and restored to near-normal levels following MitoLbNOX overexpression (fig. S6). This NRF2 restoration aligns with the 4-HNE findings and indicates that MitoLbNOX mitigates oxidative stress by normalizing redox signaling rather than suppressing the ISR. To further assess the metabolic impact of MitoLbNOX overexpression, we performed MALDI analysis on V291D-VG2-MitoTag and V291D-VG2-MitoTag-MitoLbNOX retinas (fig. S7). These analyses revealed a trend toward increased ATP abundance in the inner retinal layer of V291D-VG2-MitoTag-MitoLbNOX mice, consistent with improved mitochondrial energy output. Together, these findings demonstrate that MitoLbNOX overexpression restores mitochondrial redox balance, enhances metabolic capacity, reduces oxidative stress, and ultimately protects RGCs from degeneration in Opa1^V291D/+ mice.

Fig. 8. MitoLbNOX overexpression enhanced RGC function, survival, TCA cycle, and reduced oxidative stress in V291D-VG2-MitoTag-MitoLbNOX mice.

(A) Schematic diagram of the strategy used to generate V291D-VG2-MitoTag and V291D-VG2-MitoTag-MitoLbNOX mice. (B) Immunostaining of retinal sections from the VG2-MitoTag, V291D-VG2-MitoTag, and V291D-VG2-MitoTag-MitoLbNOX mouse models, showing GFP fluorescence colocalized with RBPMS⁺ RGC. (C) Analysis of PERG recordings at 180 days (n = 13 mice per group; one-way ANOVA with Tukey’s test P = 0.0023, 0.6653, and 0.0227 for Opa1^+/+ (WT) compared to V291D-VG2-MitoTag, WT compared to V291D-VG2-MitoTag-MitoLbNOX, and V291D-VG2-MitoTag compared to V291D-VG2-MitoTag-MitoLbNOX, respectively). (D) Quantification of RGCs in the peripheral zone of whole-mounted retinas at 180 days (n = 5 mice per group; one-way ANOVA with Tukey’s test P = 0.0032, 0.6147, and 0.0175 for WT compared to V291D-VG2-MitoTag, WT compared to V291D-VG2-MitoTag-MitoLbNOX, and V291D-VG2-MitoTag compared to V291D-VG2-MitoTag-MitoLbNOX, respectively). (E) Representative immunostaining images of PDHE1, IDH3, and 4-HNE in retinal sections from WT, V291D-VG2-MitoTag, and V291D-VG2-MitoTag-MitoLbNOX mice. Analysis of the fluorescence intensity of PDHE1 (n = 5 mice per group; one-way ANOVA with Tukey’s test P = 0.0027, 0.5287, and 0.0192, for WT compared to V291D-VG2-MitoTag, WT compared to V291D-VG2-MitoTag-MitoLbNOX, and V291D-VG2-MitoTag compared to V291D-VG2-MitoTag-MitoLbNOX, respectively), IDH3 (P = 0.0028, 0.6565, and 0.0006), and 4-HNE (P = 0.0135, 0.7149, and 0.0033) immunostaining in the ganglion cell layer. Data are presented as means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

DISCUSSION

In this study, we developed a novel patient-specific Opa1^V291D/+ knock-in mouse model to replicate the most common type of mutation, the missense mutation, found in human patients with ADOA. This model accurately recapitulated the anatomical and functional phenotypes of ADOA, reflecting those observed in patients. Our findings showed that the V291D variant affected mitochondrial structure, disrupted OXPHOS complexes and redox state, and increased oxidative stress in Opa1^V291D/+ mice. Furthermore, our study revealed that the RGCs in the Opa1^V291D/+ mouse model did not shift their energy metabolism to glycolysis, unlike other retinal cells, which adapted to compensate for the bioenergetic crisis caused by the defective ETC function. These findings provide a potential explanation for the selective vulnerability of RGCs observed in ADOA. To explore potential therapeutic strategies, we overexpressed MitoLbNOX in RGC mitochondria and observed enhanced TCA cycle activity, reduced oxidative stress, and restored RGC function and survival in Opa1^V291D/+ mice. These findings highlight the critical role of bioenergetic crisis and oxidative stress in RGC degeneration and suggest that targeting NAD⁺/NADH homeostasis with MitoLbNOX overexpression could serve as a promising therapeutic strategy for ADOA.

The genetics of OPA1-related ADOA are more complex and diverse than initially recognized. Many of these variants lead to the premature truncation of the open reading frame, pinpointing haploinsufficiency as the primary disease mechanism. In contrast, missense variants, which are primarily clustered in the guanosine triphosphatase (GTPase) domain, are believed to exert a dominant-negative effect and are strongly associated with an increased risk of developing the more severe ADOA “plus” phenotype (14, 54–56). In our study, the V346D variant identified in our patient with ADOA and the V291D variant from our novel mouse model are located within the leading portion of the GTPase domain (57). Our Opa1^V291D/+ mouse model exhibited significantly reduced OPA1 protein levels. Similarly, cultured cells transfected with the V291D variant showed diminished levels and stability of the OPA1 protein. In turn, treatment with MG132 only partially restored the OPA1 levels, indicating that its degradation is not fully reliant on the ubiquitin-proteasome system and suggesting the involvement of additional regulatory mechanisms that contribute to the instability of the OPA1 protein. Our findings indicate that the OPA1 protein is highly unstable in the presence of this variant, supporting the hypothesis that the V291D missense variant causes haploinsufficiency. Similarly, patient-derived fibroblasts carrying missense variants in the GTPase domain of OPA1 display haploinsufficiency characterized by decreased OPA1 protein expression and a shortened protein half-life (58, 59). Therefore, missense variants in the GTPase domain could lead to haploinsufficiency or a dominant-negative effect. Further experiments, including an active GTPase pull-down assay, are necessary to confirm this hypothesis.

Changes in OPA1 protein levels can disrupt the communication between mitochondria and the cell nucleus, resulting in significant transcriptional changes in neurons (60). These alterations in mitochondrial dynamics can lead to a loss of coordination between mitochondrial and nuclear gene expression, particularly in the context of pathways that are involved in energy metabolism (61). The cellular environment and energy-usage status can also affect the expression of both mitochondrial and nuclear-encoded energy metabolism transcripts, thereby highlighting the importance of a synchronized modulation between the nucleus and mitochondria in response to energy deficits and nutrient shifts (62, 63). Furthermore, when mitochondrial dynamics are disturbed by OPA1 protein mutations, mitochondrial dysfunction can trigger retrograde signaling, in which stress signals are transmitted from the mitochondria to the nucleus (64, 65). This signaling cascade can lead to changes in the expression of nuclear-encoded mitochondrial genes. A similar phenomenon is observed during the development of neurodegenerative disorders, such as Parkinson’s, Alzheimer’s, and Huntington’s diseases, in which mitochondrial abnormalities are closely associated with a significant down-regulation of the nuclear-encoded ETC and OXPHOS proteins, thus contributing to cellular aging in neural tissues (66–68). Increased oxidative stress, which can damage nucleic acids, likely plays a role in this premature aging (68, 69). Furthermore, cells may activate apoptosis in response to oxidative stress, which can affect the expression of genes related to mitochondrial biogenesis (70). The coordinated down-regulation of ETC and OXPHOS regulation in both mitochondria and the nucleus results in impaired mitochondrial metabolism, diminished energy production, and heightened oxidative stress, thereby creating a detrimental cycle that further promotes apoptosis (71). This cascade aligns with the down-regulation of ETC-related genes indicated by our snRNA-seq results, thus highlighting a potential mechanism for RGC-specific vulnerability in ADOA.

It is widely acknowledged but not well understood that RGCs are more susceptible than other retinal cells to mitochondrial dysfunction, although photoreceptors have the highest density of mitochondria in the retina. In addition, it remains unclear whether this vulnerability is primarily caused by a bioenergetic crisis, oxidative stress, or a combination of both (1, 15, 72). A previous study introduced a mouse model of Leber hereditary optic neuropathy (LHON), which is a mitochondrial optic neuropathy caused by a variant in the ND6 gene, a key subunit of Complex I, and found that increased oxidative stress is likely a primary pathogenic factor in this disease, whereas ATP production was not affected (73). Consequently, the accumulation of oxidative stress from impaired mitochondria is one of the major causes of RGC degeneration in LHON; thus, many studies have focused on antioxidants as potential treatments (16, 33). Previous research has shown that idebenone, which bypasses defective Complex I and acts as an antioxidant, is a promising candidate that is currently an approved therapy for LHON (74, 75). However, about half of the patients did not respond to this treatment, suggesting that oxidative stress alone is not the sole issue in LHON (76). Furthermore, another report showed that multiple therapeutic targets affect mitochondria and demonstrated that pathways beyond oxidative stress, including energy metabolism, mitochondrial biogenesis, and mitophagy, also play significant roles in fibroblasts derived from patients with LHON (77). Here, we performed several experiments to assess the energetic and oxidative stress profiles of the Opa1^V291D/+ mouse retina. Our findings revealed an increase in oxidative stress and a reduction in ATP levels in Opa1^V291D/+ mouse retinas. These results suggest that both a bioenergetic crisis and oxidative stress contribute to the development of RGC degeneration in ADOA. Although both LHON and ADOA lead to RGC degeneration due to mitochondrial dysfunction, their underlying mechanisms may vary.

Furthermore, our results suggest that the increased oxidative stress and reduced ATP production observed in the Opa1^V291D/+ retina are likely attributable to compromised Complex I activity. Complex I not only plays a significant role in maintaining the balance of oxidative stress but also functions as the entry point for electrons in the ETC and as a proton pump to create a proton gradient (43, 78). Although electrons can still enter the ETC through Complex II via the reduced form of flavin adenine dinucleotide (FADH₂) if there is damage to Complex I, this affects the efficiency of OXPHOS and ATP production because Complex II does not contribute to proton translocation (79). The decrease in Complex I activity observed in the Opa1^V291D/+ retina may be partly attributed to abnormalities in the inner mitochondrial membrane. The OPA1 protein, which is primarily responsible for inner mitochondrial membrane fusion, plays a critical role in maintaining the structure of mitochondrial cristae. Defective OPA1 protein can disrupt crista remodeling, destabilize respiratory complexes, and ultimately impair Complex I function (80). Consistent with this mechanism, our blue-native polyacrylamide gel electrophoresis revealed a trend toward reduced levels of Complex I–containing supercomplexes in Opa1^V291D/+ retinas (fig. S8), suggesting subtle alterations in supercomplex stability. Although these changes did not reach statistical significance, they align with prior in vitro evidence that OPA1 variants can affect the structural organization of Complex I (81, 82). Therefore, the relationship among OPA1 variants, mitochondrial structural changes, and Complex I dysfunction is closely interconnected, with each factor influencing the others in the pathogenesis of ADOA.

Although the V291D variant impaired energy production and increased oxidative stress throughout the retina, the functional impairments in Opa1^V291D/+ mice were limited to RGCs. This selective degeneration may be attributed to their heightened vulnerability to energy deficits, which are driven by their high energy demands, long axons, and lack of a myelinated sheath before the lamina cribrosa (83, 84). In contrast, photoreceptors, which have the highest density of mitochondria in the retina, prefer glycolysis for energy production and can use lipids to compensate for ATP deficiencies (45, 47, 48, 85, 86), whereas inner retinal cells, including RGCs, rely heavily on mitochondrial ETC and OXPHOS and exhibit a lower glycolytic activity (45, 46). This greater reliance on ETC and OXPHOS renders RGCs particularly sensitive to mitochondrial dysfunction, explaining their susceptibility to degeneration in ADOA (83, 84, 87). Although ATP production was generally decreased in the retina of the Opa1^V291D/+ mouse model, a metabolic shift toward glycolysis was observed, particularly in the outer retinal layers, suggesting that photoreceptors compensate for ATP deficiency by up-regulating glycolysis, consistent with previous findings (48, 88, 89). This highlights the relationship between altered energy metabolism and the metabolic flexibility of retinal cell types (90). The inability of RGCs to adapt to defective ETC function, in contrast to the metabolic flexibility of photoreceptors, underscores the significant role of compromised energy metabolism in RGC degeneration associated with ADOA. This deficiency in energy production further elevates oxidative stress, creating a harmful cycle that worsens neuronal degeneration (91).

Our Opa1^V291D/+ missense variant mouse model showed RGC abnormalities, both anatomically and functionally, closely matching the clinical presentation of human patients with ADOA. Although previous mouse models with truncated OPA1 proteins revealed changes in the shape and structure of mitochondria in the whole mouse retina and optic nerve (17, 21, 23), transcriptomic changes in the retina at single-cell resolution remain unexplored. Moreover, the selective vulnerability of RGCs, with photoreceptors remaining largely unaffected, has yet to be fully understood. Our study revealed a significant down-regulation of glycolytic proteins in the ganglion cell layer of the Opa1^V291D/+ retinas, as assessed using immunostaining; furthermore, our snRNA-seq analysis identified down-regulated energy production–related genes, including those involved in ETC and glycolysis, specifically in the RGC cluster. However, we did not detect significant changes in these genes related to energy-production pathways in other retinal cell types between WT and Opa1^V291D/+ mice, including cones and rods, thus providing a potential explanation for the lack of significant photoreceptor dysfunction in our patient and mouse model. This impaired metabolic adaptation in RGCs likely exacerbates the bioenergetic crisis, ultimately contributing to their selective degeneration.

Although the cause-and-effect relationship between oxidative and metabolic stress is not fully understood in the pathogenesis of ADOA, we believe that both factors contribute to RGC degeneration in ADOA and that interrupting this vicious cycle could serve as a potential therapeutic target for the condition. In our study, we demonstrated that increasing the NAD⁺/NADH redox ratio by MitoLbNOX overexpression could improve energy metabolism via the TCA cycle and reduce oxidative stress in the Opa1^V291D/+ mouse model. This, in turn, promoted neuronal survival and successfully mitigated the detrimental effects of the Opa1 variant, restoring both functional integrity and survival in RGCs of Opa1^V291D/+ mice. In mitochondria, NAD⁺ serves as a coenzyme for three rate-limiting enzymes in the TCA cycle, where it is reduced to NADH, generating ATP for direct energy supply and producing FADH₂ as an alternative electron donor for Complex II in the ETC (92). Beyond our findings, a previous showed that MitoLbNOX overexpression could activate the TCA cycle by increasing the NAD⁺/NADH redox ratio in m.3243A>G fibroblasts (53). Moreover, evidence from other disease models has shown that replenishing NAD⁺ levels can increase energy metabolism, reduce oxidative stress, and prolong survival across various cell types, including those in the heart, liver, and inflammatory cells (93–96). Last, our findings following MitoLbNOX overexpression reaffirmed the critical role of bioenergetic crisis and oxidative stress, driven by Complex I dysfunction, in RGC degeneration, highlighting NAD⁺/NADH homeostasis as a promising therapeutic target for preventing RGC loss in ADOA.

Despite evidence that Opa1^V291D/+ RGCs exhibit impaired metabolic compensation and heightened vulnerability to mitochondrial dysfunction, the precise mechanisms underlying this cell type–specific susceptibility remain incompletely understood. Although our data show that Complex I–driven NAD⁺/NADH imbalance selectively disrupts glycolytic and TCA cycle rewiring in RGCs, the reason this effect is confined to inner retinal neurons rather than photoreceptors remains unresolved. A previous publication highlighted that mitochondria display distinct “mitotypes” across cell types, reflecting specialized structural and functional adaptations to unique energetic demands (78). In this context, RGCs may depend more heavily on Complex I–linked redox balance, whereas photoreceptors may have greater metabolic flexibility or alternative substrate usage that buffers against OXPHOS perturbations. Nevertheless, the molecular determinants of this selective vulnerability remain to be fully elucidated.

In conclusion, we developed the Opa1^V291D/+ missense mouse model, which recapitulated ADOA phenotypes. The V291D variant reduced OPA1 protein stability and expression, supporting a haploinsufficiency mechanism. It impaired mitochondrial morphology and Complex I function, leading to oxidative stress, ATP depletion, and an energetic crisis. As a compensatory response, the retina exhibited a metabolic shift toward glycolysis, but RGCs failed to up-regulate glycolytic proteins. Spatial metabolomics, immunostaining, and snRNA-seq revealed pronounced bioenergetic crisis and down-regulated energy-production genes in RGCs, highlighting their selective vulnerability in ADOA. Notably, increasing mitochondrial NAD⁺/NADH redox ratio by MitoLbNOX overexpression in RGC could improve energy metabolism, reduce oxidative stress, and enhance RGC survival, underscoring the therapeutic potential of targeting mitochondrial metabolism in ADOA.

MATERIALS AND METHODS

Study design

The objective of this study was to investigate the impact of a patient-derived Opa1 missense variant on RGC degeneration, as well as to determine why RGCs are particularly vulnerable to mitochondrial dysfunction in ADOA. To achieve this, we generated a novel patient-specific knock-in Opa1^V291D/+ mouse model and conducted survival experiments to analyze functional phenotypes, as well as nonsurvival experiments for anatomical phenotyping and molecular assessments. Immunostaining, spatial metabolomics, and snRNA-seq were performed to examine the impact of the Opa1 variant at both the tissue and cellular levels. In addition, we examined how increasing the NAD⁺/NADH redox ratio in RGCs affects their survival in our ADOA mouse model. This study was approved by the Institutional Review Board of Columbia University (no. AAAV3523) and adhered to the principles of the Declaration of Helsinki. Because of the retrospective nature of the study and the use of deidentified historical data, the Institutional Review Board granted a waiver of informed consent. All animal experiments were approved by the Institutional Animal Care and Use Committee of Columbia University (no. AC-AABQ7582).

Patients with ADOA and mouse models

Patients with clinically diagnosed ADOA were reviewed, and their genetic testing reports were assessed at the Columbia University Irving Medical Center. An OPA1 missense variant was identified in one patient and was used to generate a knock-in mouse model. The patient-specific Opa1^V291D/+ mouse model was created by C.-S.L. The V291D point variant was introduced using the GalK pop-in-pop-out method into a bacterial artificial chromosome (BAC) clone (RP23-229C8) from the BACPAC Resources Center (https://bacpacresources.org). A gene-targeting vector was prepared using the BAC recombineering method and electroporated into KV1 (129S6 hybrid) embryonic stem (ES) cells, to generate targeted ES clones via homology recombination; the method showed an absence of aberrant splicing donor or acceptor activity. This knock-in mouse harbored a T-to-A missense variant, which converted the 291st amino acid of OPA1 from valine to aspartic acid. These mice were backcrossed to the C57BL/6J strain (JAX no. 000664, the Jackson Laboratory) for five generations and then genotyped, which confirmed the absence of the rd8 variant (97). All mice analyzed in this study were heterozygous Opa1^V291D/+ mice exhibiting normal longevity and fertility. In subsequent experiments, littermate-control WT mice (Opa1^+/+) were used for comparisons with Opa1^V291D/+ mice. To label mitochondria and assess their morphological features in these mice, we crossed the Opa1^V291D/+ mice with mito::mKate2 reporter mice (JAX no. 032188, the Jackson Laboratory), to express the fluorescent mKATE2 protein specifically in mitochondria. Housing for these animals was provided by the animal care facility of the Institute of Comparative Medicine at Columbia University.

Pattern electroretinography

The PERG was conducted as described in prior publications (98, 99). In brief, we used the PERG Animal System (Jorvec Corp, Miami, FL) for our recordings. The PERG signals from each eye were desynchronized using a phase-locking averaging method with two noncorrelated frequencies (right eye, every 492 ms; left eye, every 496 ms) and then averaged over three consecutive session blocks (98). To assess the RGC-specific function, we measured the P1N2 amplitude from the peak positive waves (P1) to the lowest negative waves (N2) recorded in the grand-average PERG waveforms.

Flash electroretinography

Flash ERG assessments were conducted according to previous publications (100) using an Espion system coupled with a Ganzfeld stimulator (Colordome, Diagnosys LLC, Lowell, MA), to measure scotopic and photopic serial intensities. To assess the STR, the light intensities of the stimuli that were used for scotopic serial-intensity ERG were −5.6, −5.3, and −5.0 log cd·s/m² in sequence. After a 10-min period of light adaptation, PhNRs were elicited using three different stimulus intensities, i.e., 0, 1, and 2 log cd·s/m², against a 10-cd·s/m² rod-saturating green background. For each intensity level, an average of 25 flashes was calculated, with an interstimulus interval of 3000 ms. The positive and negative STRs were measured at 100 and 233 ms, respectively. To specifically evaluate the RGC function, PhNR amplitudes were measured from the baseline to the PhNR trough.

Spectral domain–optical coherence tomography

We performed live imaging to measure the thickness of the RNFL using an SD-OCT imaging device (Envisu UHR2210, Bioptigen, Durham, NC, USA), which provides an axial resolution of 1.75 μm in tissue, according to previously established protocols (101). A rectangular scan of 1.8 mm in length and width was performed, with 0° angle adjustments and no horizontal or vertical offsets. The scan settings included 1000 A-scans per B-scan, 100 B-scans, and 10 frames per B-scan, with 80 inactive A-scan lines per B-scan and one volume captured (fig. S9). The resulting 10-frame OCT images were averaged using the Bioptigen InVivoVue (v2.4) software and then further processed with the Bioptigen Diver (v.3.4.4) software, to obtain measurements of RNFL thickness.

RGC counting in flat-mounted retinas

Immunolabeling and fluorescent staining of flat-mounted retinas were performed as previously described (102, 103). Eyecups for flat-mounted retinas were fixed in cold 4% paraformaldehyde in phosphate-buffered saline for 1 hour. To identify RGCs, a mouse anti-BRN3A antibody (1:50, MAB1585, Millipore) was used, followed by incubation with a secondary donkey anti-mouse antibody (1:200, 715-225-151; Jackson ImmunoResearch). RGCs were quantified using flat-mounted retinas, as described previously (102, 104). We obtained 4, 4, and 12 squares with a size of 300 μm by 300 μm from each central, midperipheral, and peripheral retinal zone, respectively. The RGC counts from all squares were then totaled and analyzed. All images were acquired using a Nikon Ti Eclipse inverted confocal microscope. BRN3A⁺ cells were counted semiautomatically and quantitatively using the ImageJ software (https://imagej.net/ij/).

Confocal microscopy assessment of mitochondrial morphology in mouse optic nerves

To analyze mitochondrial characteristics, we used SDCM with SRRFs in both WT and Opa1^V291D/+ mice. Cryosections of optic nerves were prepared from both groups, and mitochondria were visualized through mKate2 expression, which enabled red fluorescence excitation (561 nm/594 nm) using an SDCM system (Dragonfly 600, Oxford Instruments Andor) with an iXon 888 Life EMCCD camera. Super-resolution images were captured using a 100× oil objective and the Andor FUSION software (Oxford Instruments Andor), which operates the SRRF function. After acquiring the images, we applied deconvolution techniques and analyzed the data using the Surface Rendering Model provided in the iMaris software (v10.2) to thoroughly compare mitochondrial characteristics between the mouse models.

Transmission electron microscopy

We used TEM to examine the morphology of mitochondria and the myelination of axons. Ultrathin cross sections were obtained from the optic nerve and stained with uranyl acetate and lead citrate for contrast enhancement. These sections were imaged using a Hitachi 7100 transmission electron microscope (TEM instrument; Hitachi, Tokyo, Japan) equipped with an advanced digital camera system for microscopy techniques.

Immunoblotting

Mouse retinas were dissected at 180 days of age and homogenized with radioimmunoprecipitation assay (RIPA) lysis and extraction buffer (89900, Thermo Fisher Scientific), supplemented with protease and phosphatase inhibitor cocktails (P0044 and P8340, MilliporeSigma). This process was followed by sonication using an SLPe Digital Sonifier (Branson Ultrasonics, Brookfield, CT). The resulting supernatant was collected for protein quantification and subsequent Western blot analysis of total retinal proteins. Protein concentrations were determined with a Pierce BCA assay kit (23225, Thermo Fisher Scientific). For electrophoresis, proteins were denatured and separated using a Mini Blot system (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA). The separated proteins were transferred onto polyvinylidene fluoride membranes (PB5240, Invitrogen) with a Power Blotter system (PB0012, Invitrogen). The membranes were incubated in blocking buffer for 30 min, followed by applying primary antibodies and incubating at 4°C overnight. Secondary antibodies were applied at room temperature for 2 hours. Details of the primary and secondary antibodies, as well as other materials, are provided in table S4. Signals were visualized using an iBright 1500 Imaging System (Invitrogen, Thermo Fisher Scientific), and data were analyzed with the iBright Analysis Software (v5.2.1).

Quantitative real-time PCR for Opa1

To assess the gene expression levels of Opa1 between mutant and control mice, total RNA was extracted from mouse retinas using RNeasy kits (QIAGEN), following the manufacturer’s guidelines. cDNA was synthesized using the SuperScript VILO cDNA Synthesis Kit (Invitrogen), following the provided instructions. qPCR was then conducted using dye-based techniques with specifically designed primers (table S2), and samples were run in technical triplicate. The qPCR mixtures were prepared using the PowerTrack SYBR Green Master Mix (Thermo Fisher Scientific). A CFX Connect Real-time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA) was used to monitor and analyze gene expression.

Cell culture for protein stability testing

HEK293 cells (BCRC 60019; Bioresource Collection and Research Center, Hsinchu, Taiwan) were cultured to form a monolayer in a medium supplemented with 10% fetal bovine serum. Lipid-based transfections (Lipofectamine 3000, Invitrogen) were conducted using cytomegalovirus plasmid vectors (pcDNA3.1, GenScript) that carried a 3X Flag tag to insert the Opa1^WT and Opa1^V291D genes. After transfection, the cells were lysed, and immunoprecipitation was performed to isolate the ubiquitinated OPA1 protein. MG132 (25 μM) was used to inhibit protease activity. The isolated proteins were analyzed to evaluate the ubiquitination status and stability of the OPA1 protein in the lysates obtained from HEK293 cells transfected with Opa1^WT and Opa1^V291D.

Coimmunoprecipitation

For coimmunoprecipitation assays, HEK293 cells were transiently transfected with either an empty control vector or a Flag-tagged OPA1 expression construct using Lipofectamine 3000 (Thermo Fisher Scientific), according to the manufacturer’s protocol. After 48 hours, cells were harvested and lysed in ice-cold RIPA buffer supplemented with protease inhibitors. Clarified lysates were incubated overnight at 4°C with anti-Flag M2 agarose beads (Sigma-Aldrich, M8823). Bound proteins were washed, eluted, and subjected to immunoblot analysis. Experimental procedures were performed following previously published protocols (105).

Analysis of mitochondrial respiratory and hydrolytic function in retinas

To evaluate mitochondrial function in the Opa1^V291D/+ mouse model, we used the RIFS and HyFS assays on a Seahorse XF analyzer (Agilent Technologies, Cedar Creek, TX, USA), as described elsewhere (40–42). Retinal and heart tissues were harvested and immediately frozen at −80°C and then sent for analysis. Frozen tissues were placed in tubes containing four 3-mm zirconium beads and homogenized in mitochondrial assay solution [MAS buffer: 70 mM sucrose, 220 mM mannitol, 5 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, and 2 mM Hepes (pH 7.4)] using a bead homogenizer (Benchmark Scientific, Sayreville, NJ, USA) for 30 s at 6.5 m/s. The homogenates were then centrifuged at 1000g for 5 min at 4°C, and the supernatants were collected. Protein concentrations were determined, with retinal samples showing optimal responses to substrates at concentrations of 10 μg in the assays. This procedure yields a mixed mitochondrial homogenate containing disrupted mitochondria and submitochondrial particles with varying membrane orientations. Because the samples were previously frozen, exogenous NADH can access the matrix-facing NADH-binding site of Complex I. Thus, NADH was used directly as the substrate (1 mM) to assess Complex I–linked respiration, rather than pyruvate/malate. Complex II–linked respiration was measured using 5 mM succinate in the presence of 2 μM rotenone to inhibit Complex I. To inhibit the ETC upstream of Complex IV, 4 μM antimycin A (a Complex III inhibitor) and 2 μM rotenone (a Complex I inhibitor) were used. Complex IV activity was assessed by supplying electrons through 0.5 mM N,N,N',N'-tetramethyl-p-phenylenediamine (TMPD; maintained in a reduced state by 1 mM ascorbate), with 50 mM azide serving as a Complex IV inhibitor. Oxygen consumption rates were accurately measured and normalized to protein content and mitochondrial density using MTDR, to account for variations in sample processing or intrinsic mitochondrial differences. Because of limited retinal material, two complexes were typically measured per well. In the HyFS assay, the hydrolytic capacity of Complex V (ATP synthase) was assessed under uncoupled conditions. The assay was initiated with 5 mM succinate and 2 μM rotenone to measure respiratory capacity through Complex II. The ETC was then shut down with 2 μM antimycin A, and 1 μM carbonyl cyanide p-trifluoromethoxyphenylhydrazone was added to ensure complete uncoupling. Subsequently, 20 mM ATP was injected to drive ATP synthase in the reverse (hydrolytic) direction, while 5 μM oligomycin was added to inhibit Complex V activity. Because the mitochondrial membranes are disrupted, ATP freely accesses the matrix-facing catalytic site of Complex V, allowing direct measurement of ATP hydrolysis–driven oxygen consumption independent of ADP/ATP translocase function. The output data from three technical replicates were averaged for analysis. The ATP hydrolytic capacity measurements were normalized to Complex V expression, as determined using immunoblotting for ATP5A1.

MALDI-TOF MS imaging

To investigate metabolomics changes in the mouse retina, MALDI-TOF MS imaging was performed at the MALDI MS Imaging Facility, Advanced Science Research Center, The City University of New York. Mouse eyeballs were harvested at 200 days of age, embedded in 4% CMC (no. 419273, Sigma-Aldrich) at −10°C, and snap frozen on dry ice. Cryosections (10-μm thickness) were prepared using a CryoStar NX70 (Thermo Fisher Scientific), mounted on indium tin oxide–coated slides (no. 8237001, Bruker Daltonics), and desiccated under vacuum for 30 min. Matrix deposition was performed with an HTX M5 sprayer (HTX Technologies) using 2,5-dihydroxybenzoic acid (DHB) (no. D2933, TCI Chemicals) 40 mg/ml in methanol/water, 70/30 at 85°C for 8 cycles or N-(1-naphthyl) ethylenediamine dihydrochloride (NEDC, no. 222488, Sigma-Aldrich) 10 mg/ml in isopropanol/water, 70/30 at 80°C for 30 cycles. The same spray parameters were used for both matrices: velocity of 1300 mm/min; track spacing of 2 mm; N₂ pressure of 10 psi (68.95 kPa); flow rate of 3 liters/min; and nozzle height of 40 mm. Initial spectra acquisition was conducted using a MALDI-TOF MS Autoflex (Bruker Daltonics) in positive ion (DHB) or negative ion (NEDC) mode, which was calibrated with red phosphorus (no. 343242, Sigma-Aldrich). The following settings were used for both ion modes: raster width of 25 μm, laser smartbeam of “minimum,” laser frequency of 500 Hz, 500 shots per position, and mass/charge ratio (m/z) range of 60 to 1200. Ion images were processed using FlexImaging (v3.0) and SCiLS Lab (v2015b), normalized via root mean square, and a bin width of ±0.10 to ±0.20 according to peak width at a certain m/z. The spectra were interpreted manually, and the analytes were assigned according to a method described previously (106). To validate and extend metabolic coverage, high-resolution imaging was subsequently performed using a timsTOF fleX MALDI-2 instrument (Bruker Daltonics) in both positive (DHB) and negative (NEDC) ion modes. The instrument was operated with the following settings: raster width 20 μm, SmartBeam laser in “Single” mode, laser frequency 10,000 Hz, 200 shots per pixel (positive mode), 250 shots per pixel (negative mode), and an m/z acquisition range of 50 to 1000. Data were acquired using timsControl software and processed with SCiLS Lab using the same normalization strategy described above. Key metabolites were detected as follows: AMP at m/z 346.1 as [AMP-H]⁻, G6P at m/z 171.0 as [G6P-H]⁻, pyruvate at m/z 87.0 as [pyruvate-H]⁻, and ATP at m/z 508.0 as [ATP + H]⁺. Quantification was performed within defined regions of interest in the tissue.

Hematoxylin and eosin staining

Hematoxylin and eosin staining was performed on tissue sections after MALDI imaging, to access the histology of the MALDI images. The residual matrix was removed by rinsing slides with 95% ethanol, after which the sections were stained with Hematoxylin Gill No. 1 and Eosin Y (Sigma-Aldrich) according to the manufacturer’s instructions. The stained sections were imaged using a Leica Aperio CS2 slide scanner at ×20 magnification with a 0.75–numerical aperture Plan Apo objective. These images provided anatomical context for mass spectral data, allowing the establishment of precise correlations between molecular and histological features. Quantification was performed within defined regions of interest in the tissue.

Immunostaining

To assess protein expression distribution in mouse retinal histology, immunofluorescence was performed on cryosections of mouse retinas at 360 days of age according to previously established protocols (100). Briefly, slides were prepared using mouse retinas embedded in optimal cutting temperature compound (Tissue-Tek O.C.T. Compound, Sakura Finetek). The primary and secondary antibodies listed in table S4 were used for staining. Imaging was carried out using a Zeiss LSM 900 microscope equipped with an Airyscan super-resolution image scanning system (Carl Zeiss, Germany). Z-stack images spanning 5 μm with a step size of 0.3 μm were acquired from all retinal sections. Postacquisition processing and deconvolution were performed using the Airyscan Joint Deconvolution feature in the ZEN Blue software (v3.7). Images from matched mutant and WT samples were captured during the same experimental session under identical imaging settings. The fluorescence intensity in each maximum projection image was manually segmented and quantitatively measured using the ImageJ software (https://imagej.net/ij/).

ATP measurements from mouse retinas

ATP levels were measured in the retinas using a commercially available kit [ab83355, ATP Assay Kit (Colorimetric), Abcam] according to the manufacturer’s instructions. Fresh retinal tissue from both eyes of each mouse was carefully dissected and homogenized in the assay buffer. The homogenate was centrifuged at 13,000g for 5 min at 4°C, and the resulting supernatant was collected for protein quantification and subsequent analysis. To prevent enzyme interference in the assay, deproteinization was performed using a kit (ab204708, Deproteinizing Sample Preparation Kit, Abcam). After a 30-min incubation, the ATP assay was conducted, and optical density readings were taken at 570 nm using a microplate reader.

NAD⁺ measurements from mouse retinas

To assess the levels of NAD⁺ and NADH in mice, we used a commercially available kit [ab65348, NAD⁺/NADH Assay Kit (Colorimetric), Abcam] following the manufacturer’s instructions. We collected retinas from each mouse, homogenized them, and centrifuged the mixture at 14,000g for 5 min at 4°C. Next, we transferred the supernatant to a 10-kDa spin column (ab93349, 10kD Spin Column, Abcam) and centrifuged it at 10,000g for 20 min at 4°C. The filtrate was collected for protein quantification and the NAD assay. Optical density readings were taken at 450 nm using a microplate reader at room temperature 1 hour after the procedure.

GSH measurements from mouse retinas

Total GSH and reduced GSH levels were measured using a commercially available kit [ab239709, GSH+GSSG/GSH Assay Kit (Colorimetric), Abcam], following the manufacturer’s instructions. Retinal tissues were collected from both eyes of each mouse and homogenized in the buffer supplied with the kit. Protein quantification was conducted before adding 5% 5-sulfosalicylic acid to precipitate the proteins in the samples. Next, the reaction mix and substrate solution were added to the samples and incubated for 10 min. Optical density readings were taken at 415 nm using a microplate reader at room temperature 10 min after the procedure. The levels of GSH and GSSG were calculated on the basis of the optical density readings.

SOD measurements from mouse retinas

SOD levels were measured using a commercial kit [ab65354, Superoxide Dismutase Activity Assay Kit (Colorimetric), Abcam] following the manufacturer’s instructions. Retinal samples were homogenized in ice-cold immunoprecipitation lysis buffer (no. 87787, Thermo Fisher Scientific) that contained 1 mM phenylmethylsulfonyl fluoride protease inhibitor (no. 36978, Thermo Fisher Scientific). The homogenates were then centrifuged at 14,000g for 5 min at 4°C, and the supernatants were collected for analysis. The SOD assay was performed by mixing the supernatant with the working solution provided in the kit, followed by incubation at 37°C for 20 min. Optical density readings were obtained at 450 nm using a microplate reader to quantify SOD activity.

Lactate measurements in mouse retinas

The levels of lactate in the retinas were measured using a commercially available kit [ab65331, l-Lactate Assay Kit (Colorimetric), Abcam] according to the manufacturer’s instructions. Fresh retinal tissue from both eyes of each mouse was carefully dissected and homogenized. The homogenate was then centrifuged at 14,000g for 5 min at 4°C, and the resulting supernatant was collected. Deproteinization (ab204708, Deproteinizing Sample Preparation Kit, Abcam) was carried out to prevent lactate degradation by endogenous LDH. The deproteinized supernatant was then used for the assay. After a 30-min incubation at room temperature, the optical density was measured at 450 nm on a microplate reader.

Single-nucleus RNA sequencing

To investigate the impact of this Opa1 variant on the retinal transcriptomes at the single-cell level, we performed snRNA-seq on pooled frozen retinal tissues. Nucleus extraction was performed using the Miltenyi Nuclei Extraction Buffer (Miltenyi Biotec) according to the manufacturer’s guidelines. Upon isolation, the nuclei were counted using trypan blue and a Countess III Automated Cell Counter (Thermo Fisher Scientific, Waltham, MA, USA). snRNA libraries were prepared using the Chromium Single Cell 3′ kit (10x Genomics) and sequenced on an Illumina platform using standard protocols. After obtaining the sequencing data, we used Cell Ranger (v8.0) with default parameters to generate a filtered_feature_bc_matrix.h5 file containing cell barcodes and transcript counts for each sample. The data were aggregated using the Cell Ranger aggr program. The integrated dataset was first imported into the Rosalind platform (www.rosalind.bio/) for dimension reduction and unsupervised clustering using Cell Ranger Graph Based Clustering (10x Genomics). The dataset was then loaded into R (v4.2) and the Seurat package (v5.0) (107). Cell types were annotated using SC-type (v1.0) (108) with cell markers for major retinal cells (table S3). A pathway enrichment analysis was performed using clusterProfiler (v4.10.1) (109) with the REACTOME (110) and WikiPathways (111) databases. The results of differential gene expression analyses were visualized using heatmaps and dot plots wrapped in the Seurat package, and normalization was performed using log₂ transformation. The snRNA-seq data have been deposited into the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus repository (GSE292269).

High-resolution spatial transcriptomics of the mouse retinas

Mouse eyes from 280-day WT and Opa1^V291D/+ mice were enucleated after euthanasia. Whole eyecups were fixed in 10% neutral-buffered formalin for 12 to 24 hours, dehydrated, and paraffin embedded using standard histological procedures. Retinal sections (10-μm thickness) were collected onto 10x Genomics Visium HD FFPE Spatial Gene Expression slides. Sections were deparaffinized, stained with hematoxylin and eosin, and imaged to document tissue morphology and orientation. Target retrieval, probe hybridization, and on-slide chemistry were performed according to the 10x Genomics Visium HD FFPE protocol, with minor optimizations for retinal tissue integrity. Spatial gene expression libraries were constructed per manufacturer instructions, sequenced on an Illumina platform, and processed using Space Ranger (10x Genomics) for alignment, segmentation, and feature quantification. Annotation of ganglion cell–enriched regions was done by using QuPath (112). Downstream spot-level analysis and clustering were performed in Seurat package (v5.0) (107).

Generation of RGC-specific MitoLbNOX overexpression in Opa1^V291D/+ mice

In this study, we generated Rosa26^{LSL-MitoLbNOX} (LoxP-Stop-Lox[LSL]-MitoLbNOX) mice using a method similar to that used for the Rosa26^LSL-MitoTag line (JAX no. 032290, the Jackson Laboratory), which incorporates 3XHA-EGFP-OMP25 (MitoTag cassette) into the Rosa26 locus for targeted mitochondrial EGFP expression (113). We constructed a targeting vector containing a CAG promoter, a loxP-flanked reversed neomycin cassette, an SV40 poly-adenylation sequence, and cDNA encoding MitoLbNOX from the pUC57-mitoLbNOX plasmid (Addgene plasmid no. 74448), which was linearized and targeted to intron 1 of the mouse Rosa26 gene. To achieve conditional mitoLbNOX overexpression in Opa1^V291D/+ mice, we crossed LSL-MitoLbNOX mice with Opa1^V291D/+ mice, generating Opa1^V291D/+; Rosa26^{LSL-MitoLbNOX/+} offspring (V291D-MitoLbNOX). For the RGC-specific mitochondrial reporter Cre line, we created double homozygous Vglut2^Cre; Rosa26^LSL-MitoTag (VG2-MitoTag) mice by crossing Vglut2-Ires-Cre mice (JAX no. 28863, the Jackson Laboratory) with MitoTag reporter mice (JAX no. 032290, the Jackson Laboratory) over two generations. Last, to compare mice with and without mitoLbNOX overexpression in the RGC of Opa1^V291D/+ mice, we crossbred V291D-MitoLbNOX mice with VG2-MitoTag mice and selected V291D-VG2-MitoTag and V291D-VG2-MitoTag-MitoLbNOX offspring for experiments (Fig. 8A).

Statistical analysis

Study mice were matched for sex and age between the littermate-controlled WT and mutant groups. Statistical analyses were conducted using GraphPad Prism (v10.4), SPSS Statistics (v21), and R (v4.2). Unpaired independent t tests or linear regression analyses were used to compare the continuous parameters between the two groups. One-way analysis of variance (ANOVA) was used for comparisons of the continuous parameters between three groups. Continuous variables are expressed as the means ± SEM in the plots. P values derived from multiple testing were corrected using the Benjamini-Hochberg method. A two-tailed P value of <0.05 and a q value of <0.1 indicated statistical significance.

Supplementary Materials

The PDF file includes:

Figs. S1 to S9

Tables S1 to S4

Legends for supplementary Excel files

Other Supplementary Material for this manuscript includes the following:

Supplementary Excel Files