Restoration of defective oxidative phosphorylation to a subset of neurons prevents mitochondrial encephalopathy

We tested whether the restoration of a missing OXPHOS complex in neurons could protect mice against mitochondrial encephalopathy. Previously, our lab characterized neuron-specific mouse models of OXPHOS complex I and complex IV deficiency. These conditional neuronal knockouts (nKO) were created using CamKIIα-Cre and floxed Ndufs3 and Cox10 alleles, respectively. NADH-ubiquinone oxidoreductase iron-sulfur protein 3 (NDUFS3) is a nuclear encoded subunit of Complex I, critical to Complex I assembly and function (Procaccio et al, 2000; Vogel et al, 2007). COX10 is a nuclear encoded protein involved in heme a biosynthesis, and critical for the assembly of Complex IV (Mogi et al, 1994). Heme a is an indispensable cofactor for the maturation of COX1, a catalytic subunit of Complex IV. The Ndufs3- and Cox10-nKO mice experience shortened lifespans, progressive weight and neuronal loss, and impaired motor coordination and balance (Diaz et al, 2012; Peralta et al, 2020).

To replace the missing genes in neurons, we used recombinant AAV-PHP.eB viruses. This AAV capsid subtype has been shown to effectively deliver genes to neurons of C57BL6 mice (Chan et al, 2017). For simplicity, henceforth we will refer to AAV-PHP.eB as “AAV”. The recombinant genes contain the human synapsin promoter (hSYN) driving the neuronal expression of cDNA coding for Ndufs3, Cox10, or eGFP. Figure 1A illustrates the structure of the constructs used. The recombinant AAVs were systemically delivered to 18–21-day-old nKO mice via a single retro-orbital injection (i.v.), as previously described (Bacman et al, 2018; Pereira et al, 2020; Yardeni et al, 2011). We treated and analyzed five groups: Ndufs3-nKO mice receiving AAV-eGFP (Ndufs3-nKO+eGFP), Ndufs3-nKO mice receiving AAV-Ndufs3 (Ndufs3-nKO+Ndufs3), Cox10-nKO mice receiving AAV-eGFP (Cox10-nKO+eGFP), Cox10-nKO mice receiving the AAV-Cox10 (Cox10-nKO+Cox10), and CamKIIα-Cre hemizygous mice (WT). Both males and females were treated.

The different groups were analyzed for body weight, rotarod, and open field behaviors weekly after the injections. The timeline of the experiments is depicted in Fig. 1B. As observed during the initial characterization, nKO mice injected with AAV-eGFP exhibited hunched posture and began to lose weight between 4 and 4.5 months of age (Fig. 2A,C,E). On the other hand, after administration of the respective conditionally deleted genes, both OXPHOS-nKO models showed normal posture and no weight loss (Fig. 2A,C,E).

Ndufs3-nKO mice develop a severe encephalopathy and need to be sacrificed between 4.5 and 5 months of age (Peralta et al, 2020). Although WT and Ndufs3-nKO+Ndufs3 appeared normal and healthy at this age, they were also sacrificed at 5 months of age for molecular and histological comparisons with the nKO model. We allowed some mice to age beyond the 5-month timepoint and observed that AAV-Ndufs3 injections significantly extended survival (Fig. 2B). Three of these mice were sacrificed at 15 months of age with no detectable encephalopathy phenotype. A fourth mouse died at 6 months of age for reasons apparently unrelated to neurological problems.

Although Cox10-nKO mice can live up to 8–10 months, quality of life progressively declines around 5 months of age (Diaz et al, 2012). Similarly to the Ndufs3 model, mice were sacrificed and analyzed at 6 months of age. A small cohort of mice was allowed to age beyond the 6-month timepoint, and we observed that AAV- Cox10 injections greatly extended survival, to 15 months (Fig. 2B).

Ndufs3-nKO mice lose motor coordination and balance at 3 months of age (Peralta et al, 2020). In the current study, we analyzed the motor and coordination phenotypes through rotarod and open field tests. At 3.5 months of age, clear deficits were noticed in the Ndufs3-nKO and Cox10-nKO mice on the rotarod that became significant at 4 months of age and continued until mice were terminal. In contrast, Ndufs3-nKO+Ndufs3 behaved like WT mice (Fig. 2D). Cox10-nKO+Cox10 also had greatly improved motor balance (Fig. 2F).

The open field test is less sensitive to detect behavioral changes in these models because the OXPHOS nKO models undergo a period of hyperactivity that masks decrease in activity or vertical counts. Nonetheless, such impairment becomes evident by 4–4.5 months depending on the model (Appendix Figure S1). Improved performance was clear, as Ndufs3-nKO+Ndufs3, Cox10-nKO+Cox10 and WT mice did not show differences in exploratory or rearing patterns over time (Appendix Figure S1).

To quantify changes in NDUFS3 protein in Ndufs3-nKO mice, we performed western blots of cortex and hippocampal homogenates of 5-month-old male mice. In cortices of 5-month-old Ndufs3-nKO+eGFP mice, NDUFS3 levels were depleted to ~40%, which were restored to ~90% with the AAV-Ndufs3 treatment (Fig. 3A,B). We observed a similar increase of NDUFS3 levels in the hippocampus, from 17 to 92% (Fig. 3D,E). We also analyzed another subunit of Complex I, NDUFB8, in both the cortex and hippocampus. NDUFB8 is known to be unstable if Complex I is not assembled and thereby is a useful surrogate to Complex I holoenzyme (Lazarou et al, 2009; Peralta et al, 2020). NDUFB8 was depleted to approximately 34% and 48% in cortex and hippocampus, respectively, of Ndufs3-nKO mice. NDUFB8 levels were restored to 69% and 89% in these brain regions with gene therapy (Fig. 3B,E). In cortex and hippocampal homogenates of 5-month-old female mice, NDUFS3 levels increased from 32% to 60% and 36% to 42%, respectively (Fig. EV1A,B,E,F). A bigger change was observed with the marker NDUFB8 in both cortex (36% to 71%) and hippocampus (44% to 81%) (Fig. EV1A,B,E,F).

Western Blot analyses for proteins of complexes II, III, and IV in the Ndufs3 nKO model did not show major changes in cortex and hippocampus (Fig. 3C,F). An increase of COX1 (CIV) in the hippocampus of Ndufs3-nKO mice was reverted by the gene replacement (Fig. 3F). In the cortex of female mice, we detected increases in SDHA (CII) and UQCRC1 (CIII) that were resolved with AAV treatment (Fig. EV1A,C). A similar trend was observed with hippocampal levels of SDHA, however, UQCRC1 levels were decreased in both nKO groups (Fig. EV1E,G). Examining another mitochondrial marker (VDAC), an outer mitochondrial membrane protein, did not suggest an increase in mitochondrial mass in cortex; however, there was an increase in male hippocampus homogenates and a decrease in female hippocampal homogenates (Figs. 3G and EV1D,H). Finally, Ndufs3-nKO mice showed a trend towards an increase in mtDNA copy number (Appendix Figure S2).

To verify the reintroduction of NDUFS3 protein in cortical neurons, we performed immunofluorescence imaging with anti-NDUFS3 and NeuN. Approximately 50% of NeuN-positive cells in Ndufs3-nKO+Ndufs3 cortex had clear, mitochondrial localization of NDUFS3 staining, compared to 10% in nKO+eGFP and 80% in WT mice (Fig. 3H). Therefore, after normalization to the levels observed in WT samples, ~34% NeuN-positive somas express NDUFS3 from the AAV-PHP.eB.

To quantify changes in COX10 protein in Cox10-nKO mice, we performed Western blot of cortex and hippocampal homogenates of 5-month-old male mice. As a surrogate for COX10 expression, we probed for COX1, as COX10 expression is essential for COX1 maturation and stability (Mogi et al, 1994) (Fig. 4A,E). Although we typically normalize to total protein loading, we observed a significant increase of VDAC in the cortex of Cox10-nKO+eGFP mice (Fig. 4D). This would suggest an increase in mitochondrial mass which would be reflected in the levels of OXPHOS subunits; therefore, we compared Complex IV subunits to VDAC after normalization for total protein loading. In the cortex, COX1 and COX4 (a nuclear-encoded Complex IV subunit) levels are clearly depleted in Cox10-nKO+eGFP mice and restored in treated mice (Fig. 4A,B). In the hippocampus of Cox10-nKO+eGFP animals, COX1 was comparable to WT levels, while COX4 was depleted (Fig. 4E,F). Interestingly, VDAC was not increased in hippocampus samples, whereas mtDNA trended higher, but it was not significant (Fig. 4H,I). Hippocampal homogenates of Cox10-nKO+Cox10 mice show comparable levels of COX1 without the accompanying increase in mtDNA levels (Fig. 4E,F).

Western Blot analyses for subunits of complexes I, II and III showed a mild increase in SDHA (CII) and UQCRC1 (CIII) in the cortex and hippocampus of nKO+eGFP animals, which was partially resolved in Cox10-nKO+Cox10 animals, as SDHA still appeared to be elevated in the hippocampus (Fig. 4C,G).

To verify reintroduction of COX10 protein in cortical neurons, we performed immunofluorescence imaging with anti-COX1 and NeuN. Approximately 60% of NeuN-positive cells in Cox10-nKO+Cox10 cortex had clear, mitochondrial localization of COX1 staining, compared to 30% nKO+eGFP and 70% in WT mice (Fig. 4J).

Our group had previously observed that lack of NDUFS3 significantly impaired Complex I assembly and supercomplex formation. To analyze Complex I levels, we analyzed cortical homogenates on BN-PAGE and probed for NDUFA9, a subunit of Complex I. The Ndufs3-nKO+eGFP had a marked decrease in Complex I levels in the supercomplex CI + CIII and on its own (~49% compared to controls). With AAV-Ndufs3 treatment, CI levels were restored to 93% in males, and 77% in females (Figs. 5A,B and EV1I,J). Probing for UQCRC1, a subunit of Complex III, the disruption of the CI + CIII supercomplex is also clear in Ndufs3-nKO+eGFP mice (Figs. 5A,B and EV1I,J). In addition to recovery of Complex III levels, Complex IV was increased in Ndufs3-nKO+Ndufs3 mice (Fig. 5B, right). To further confirm functionality of assembled Complex I, we performed an in-gel activity (Fig. 5C). We observed a marked decrease in functional Complex I in the nKO+eGFP (48% ± 10.97) that was rescued by the gene therapy (91.47% ± 9.56) (Fig. 5D, left). Similar results were observed with female animals (47% to 77%) (Fig. EV1K,L). There were no differences in Complex IV activity between the groups, in either sex (Figs. 5D and EV1L).

COX10 depletion led to a progressive Complex IV defect, significantly impairing Complex IV enzymatic activity (Fig. 5E,F). Interestingly, in addition to a reduction in Complex IV levels in Cox10-nKO+eGFP mice, we observed a decrease in the CI + CIII supercomplex when probing for NDUFA9 and an increase in “free” Complex III when probing for UQCRC1 (Fig. 5E,F). Therefore, Complex IV deficiency led to an instability of supercomplexes (Novack et al, 2020). Complex IV levels were restored from 59% to 127% with AAV-PHP.eB-Cox10 treatment (Fig. 5F, left). Complex I (Fig. 5F, middle) and Complex III (Fig. 5F, right) levels were also restored to WT levels in Cox10-nKO+Cox10 mice. Functionality of Complex IV was confirmed in the rescued brains, from 71.15% ± 11.78 to 110.6% ± 21.4 (Fig. 5G,H).

In previous studies, we showed that deletion of Ndufs3 was associated with neuronal death in the hippocampus at 4 months, despite no major differences in brain weight (Diaz et al, 2012; Peralta et al, 2020). Additionally, Ndufs3-nKO mice displayed marked glial activation in the cortex and hippocampus. In the current study, we performed immunofluorescence imaging with antibodies against glial acidic fibrillary protein (GFAP, activated astrocyte marker) and IBA1 (microglia marker) on cortical slices of 5-month-old animals. GFAP and IBA1 staining was decreased in Ndufs3-nKO+Ndufs3 slices compared to nKO+eGFP (Fig. 6A,B). Western blot analysis supported these data, with Ndufs3-nKO+Ndufs3 mice displaying significantly decreased GFAP and IBA1 levels compared to nKO+eGFP (Figs. 6C,D and EV2A,B). To analyze neuronal loss, we measured TUJI levels by western blot and found slight decreases in both cortex and hippocampus that were mitigated in Ndufs3-nKO+Ndufs3 mice (Fig. 6C,D). We did not detect any changes in TUJI levels in Ndufs3-nKO female mice (EV2A,B). As previously reported, no changes were observed in the brains weight of all three Ndufs3 groups at 5 months (Fig. 6E).

In contrast to Ndufs3-nKO mice, Cox10-nKO mice exhibit a significant decrease in brain mass with obvious cortical atrophy at 6 months of age (Diaz et al, 2012). Treatment with AAV-PHP.eb-Cox10 prevented cortical mass loss (Fig. 7A,B). Cox10-nKO mice also experience dramatic glial activation in the cortex and hippocampus, as seen via immunofluorescent imaging, and supported by western blot quantifications (Fig. 7C–F). Although mice receiving AAV-PHP.eB-Cox10 displayed a decrease in glial markers in cortex homogenates, GFAP and IBA1 remain elevated in the hippocampus (Fig. 7F). Despite notable differences in appearance and mass of the cortex, Cox10-nKO+eGFP mice did not display drastic changes in TUJI levels on western blot analysis; however hippocampal homogenates displayed a mild decrease in both Cox10-nKO groups (Fig. 7E,F). Via NeuN immunofluorescent staining, we observed hippocampal degeneration, but no apparent differences in staining throughout the cortex (Fig. 7G).

As previously mentioned, we followed a small cohort of nKO mice and their WT littermates to 15 months of age. These mice maintained a healthy weight and posture (Fig. EV3A). On the rotarod, both nKO+OXPHOS mice performed in a manner comparable to WT controls, excluding one Cox10-nKO+Cox10 mouse who had some difficulties due to increased weight (Fig. EV3B). In the open field analysis, Ndufs3-nKO+Ndufs3 mice tended to have more exploratory behavior compared to both WT and Cox10-nKO+Cox10 mice, as noted by vertical counts (Fig. EV3C). Following euthanasia and dissection, we found no significant differences between the groups in terms of brain weight or appearance.

To quantify levels of the restored protein, we performed western blots for NDUFS3 and COX1 in both cortex and hippocampal homogenates. In both models, the targeted protein was comparable to WT levels, suggesting effective long-term expression of the AAV (Fig. EV3D,E). To confirm sustained stability and functionality of the targeted Complexes, we performed BN-PAGE and in-gel activity assays for Complexes I and IV. Complex I and IV levels and enzymatic activity were comparable to WT levels (Fig. EV3F–I).

Immunofluorescent staining for neuroinflammatory markers GFAP and IBA1 also did not show major differences between treated and WT animals (Fig. EV4A,B). Western blots showed that Ndufs3-nKO+Ndufs3 animals do have a mild, though not significant increase in GFAP levels compared to their WT counterparts (Fig. EV4C,D). Additionally, Cox10-nKO+Cox10 animals had an increase in hippocampal IBA1 (Fig. EV4C,D). No significant differences were observed concerning neuronal content or cortical mass (Fig. EV4C–E). This data confirms that the neuroinflammation seen at 5–6 months of age is resolved and maintained in older ages of treated mice.

Our replacement gene constructs (Ndufs3 and Cox10) do not contain a tag, as it could affect folding, complex assembly or enzymatic function, therefore we used the AAV-eGFP virus as a surrogate to determine the extent of viral mediated expression of AAV-PHP.eB. We performed western blot analysis, probing for GFP, on homogenates of brain cortex, heart, liver, and kidney. As expected, we observed that GFP expression was only observed in the cortex sample in nKO+eGFP animals (Fig. 8A,B). We also quantified the number of viral particles delivered to cortex and hippocampus. This was performed by digital PCR with a custom primer/probe targeting the hSYN promoter, which is present only in the AAV, but not in the mouse genome. As expected, non-injected homogenates did not have hSYN DNA in either OXPHOS nKO model. Although similar AAV titers were injected, mice injected with AAV-Ndufs3 had a higher AAV copy number compared to mice injected with AAV-eGFP (Fig. 8C, left), whereas this trended in the opposite direction for the Cox10-nKO mice (Fig. 8C, right). These variabilities are not uncommon and reflect the percentage of viable viral particles in different preps. Because AAV copy number does not directly correlate to the number of infected cells, we performed immunohistochemistry on AAV-eGFP-injected animals using anti-GFP and anti-NeuN (Fig. 8D). We determined the percentage of GFP-/NeuN-double positive cells in the motor and piriform cortices and the hippocampus. Approximately 23% and 28% of NeuN-positive cells expressed GFP in Ndufs3-nKO and Cox10-nKO mice, respectively (Fig. 8E).

To better understand the extent of recovery by AAV-PHP.eB, we determined the levels of gene ablation over time, starting shortly after time of injection, using a three primer PCR to detect floxed, deleted and WT alleles. At 1-month of age, the percentage of Ndufs3 recombination in the cortex and hippocampus homogenates was 47% and 59%, respectively. Recombination peaked at 3 months of age at about 60% in both cortex and hippocampus homogenates (Appendix Fig. S3A,B). Although Cre recombination plateau, we observed a steady decrease in NDUFS3 protein levels to approximately one-third of WT levels, as estimated by the ratio of NDUFS3 to SDHA (Appendix Fig. S3C,D). Together this implies that at least 50% of the cells in these regions have undergone complete Cre recombination. Because homogenates contain a mixture of neuronal and glial subtypes, we can assume that most CamKIIα-positive neurons have both alleles deleted, leading to the severe Complex I or IV deficiency.

We next did a dose-response study of AAV.PHP.eB-hSYN-eGFP in 3-month-old mice. We retro-orbitally injected either a low (1.5 × 10¹¹ vg), medium (1.5 × 10¹² vg), or high (7.5 × 10¹² vg) dose and collected the brain 8 weeks post injection. Our data showed that the low dose, which was most similar to the doses used in the rescued mice, led to an average of 13% GFP+ NeuN+ while medium and high doses had comparable GFP+/NeuN+ expression with ~30% of NeuN+ cells expressing GFP (Appendix Figure S4A,B). There was a similar number of double positive cells even though AAV copy number in cortex was increased between the medium- and the high-titer dose (Appendix Figure S4C).

Our gene replacement results suggested that a subset of OXPHOS-positive neurons protects the CNS from developing an encephalopathy. We further explored this concept by producing OXPHOS conditional nKO with an alternative, less active, CamKIIα-Cre strain. We found that the Cre-mediated flox ablation by B6.Cg-Tg(Camk2a-cre)T29-1Stl/J mice (known as T29-1) was less robust than the one by B6.Cg-Tg(Camk2a-cre)3Szi/J mice, the strain used in all the experiments described above. We followed the phenotype and floxed allele deletion of Ndufs3-nKO produced with the weaker T29-1 CamKIIα-cre driver. Even though the levels of Ndufs3 gene ablation reached 20 and 35%, in cortex and hippocampus homogenates, respectively, the Ndufs3-nKO mice showed no behavioral phenotypes up to 6 months of age (Fig. EV5A–E). Western blot analysis of this new model showed 25 and 40% reduction in NDUFS3 in cortex and hippocampus homogenates, respectively (Fig. EV5F,G,J–K). Considering that these homogenates are not exclusively CamKIIα-positive neurons (i.e., they contain glia and other types of neurons), it is safe to assume that the Ndufs3 deletion was substantially higher in neurons than the assays revealed. Therefore, the data with the weaker T29-1 CamKIIα-cre independently corroborate our AAV gene therapy rescue findings, showing that even a minority of OXPHOS-positive neurons prevents CNS failure, even when a substantial number of neurons are OXPHOS deficient.

Ndufs3-nKO and Cox10-nKO mice were previously described in Peralta et al (2020) and Diaz et al (2012), respectively. Briefly, we generated neuron-specific KO mice by mating Ndufs3^fl/fl or Cox10^fl/fl males with respective heterozygous, CaMKIIα-Cre positive females. When possible, controls and conditional KO mice were obtained from the same litters. The presence of the WT, floxed genes, and the Cre transgene was detected by PCR from tail DNA (Appendix Table S2).

Cre recombination efficiency was calculated using a three primer PCR approach. A standard curve was created using different ratios of the purified products of floxed and deleted bands from a Ndufs3^fl/del sample. 10 ng of homogenate-derived DNA was used for samples.

All animals used in this work had a C57BL/6J background and were backcrossed for at least 10 generations. Mice were housed in a virus antigen-free facility at the University of Miami, Division of Veterinary Resources, in a 12-h-light/dark cycle at room temperature and fed ad libitum.

Ndufs3 (Gene ID: 68349) and Cox10 (Gene ID: 70383) cDNA were cloned into an AAV-PHP.eb plasmid containing the hSYN promoter via In-Fusion cloning (Takara). Plasmids were transformed into Stbl3 chemically competent E. coli (Invitrogen, C737303) and prepped with the NucleoBond Xtra Maxi EF kit (Macherey-Nagel). The plasmids were sent to the University of Iowa Viral Core Facility, which produced virus with 4.07 × 10¹² vg/mL and 6.65 × 10¹² vg/mL for Ndufs3 and Cox10 viruses, respectively. The AAV-eGFP was obtained from Penn Vector Core, containing the same promoter and backbone, at a concentration of 1.46 × 10¹⁴ vg/mL. Juvenile mice (2.5–3 weeks of age) were injected retro-orbitally as previous described (Bacman et al, 2018; Yardeni et al, 2011). Ndufs3-nKO mice received 2 × 10¹¹ vg of either AAV-Ndufs3 or AAV-eGFP. Cox10-nKO mice received 3.32 × 10¹¹ vg of either AAV-Cox10 or AAV-eGFP. Considering their average weight at this age, the doses were 2.8 × 10¹³ vg/kg for Ndufs3-nKO and 3.87 × 10¹³ vg/kg for Cox10-nKO mice. Injections were performed retro-orbitally, which immediately drain into the venous system (Yardeni et al, 2011). Retro-orbital injections are more consistent than tail vein injection as the latter tends to collapse if the injection is not precise.

All animal procedures were approved by the University of Miami Animal Care and Use Committee.

Open field (Med Associates Inc.) is a sensitive method for measuring gross and fine locomotor activity. It consists of a chamber and a system of 16 infrared transmitters that record the position of the animal in the 3D space. This system records both horizontal and vertical movement. For our study, the animals were placed in the chamber and the locomotor activities were recorded for 15 min.

Mouse motor coordination was tested at different ages using a Rotarod (Ugo Basile 47650) set at a constant speed of 10 rpm over 180 s. The test consisted of 3 trials performed for each animal at the corresponding age, and the latency to fall was recorded. Mice that completed the task received a final latency time of 180 s. Animals were trained in the rotarod twice per trial, before their first test.

Anesthetized mice were transcardially perfused with PBS. The brains were isolated, and cortex and hippocampus were dissected from one hemisphere for protein extraction. The second hemisphere was fixed O.N. in 4% PFA, cryoprotected in 30% sucrose, and frozen in OCT. The brains were cut at a 20 μm thickness with a cryostat and stored at −80 °C (Leica).

Antigen retrieval was performed using a 10 mM sodium citrate buffer (pH 6.0) preheated to 80 °C in a water bath for 30 min. Slides were washed three times in PBS. Sections were permeabilized with 1% Triton, blocked with 2% BSA for 1 h at room temperature, and incubated with primary antibody diluted in 1% BSA and 0.1% Triton overnight at 4 °C. Slides were incubated with Alexa Fluor secondary antibody for 1 h at room temperature and mounted with Vectashield (Biotium 23004). Images were captured with a Leica confocal microscope.

QuPath Cell Detection software was used to detect and count NeuN and GFP-positive cells. For NDUFS3- and COX1-positive NeuN calculations, 3–5 images, captured at 63× with a pixel size of 180 nm, for an average of 200 neurons, were used per mouse. For GFP-postiive NeuN calculations, 3–5 images, captured at 20× with a pixel size of 500 nm, were used per mouse.

Protein extracts were prepared from the cortex and hippocampus regions and homogenized in PBS containing a protease and phosphatase inhibitor mixture (Roche). Samples were sonicated for ~5 s. Protein concentration was determined by the Lowry assay using the DC kit (Bio-Rad). Approximately 20–50 μg protein was separated by SDS-PAGE in 4–20% acrylamide gels (Bio-Rad). Gels were run in Tris-glycine buffer (25 mM Tris-Cl, 250 mM glycine, 0.1% SDS) at 100 V until completion. Gels were transferred to 0.2 μm PVDF membranes (Bio-Rad) with Trans-Blot Turbo Transfer System and 1X Trans-Blot Turbo Transfer Buffer (Bio-Rad), and imaged after for total protein loading. For IBA1, gels were transferred to 0.2 μm nitrocellulose membranes, stained with Ponceau for total protein loading, and otherwise treated the same as PVDF membranes.

Membranes were blocked with 5% nonfat milk in 0.1% Tween-20 in PBS and subsequently incubated with specific antibodies, which were incubated overnight at 4 °C or 1 h at room temperature. Blots were incubated with secondary antibodies conjugated to horseradish peroxidase (Cell Signaling Technology) for 1 h at room temperature. The reaction was developed by chemiluminescence using SuperSignal Pico West reagent (Thermo Fisher Scientific, #34578). Blots were visualized with ChemiDoc Imaging System (Bio-Rad). Optical density measurements were taken by software supplied by Bio-Rad. Total protein loading (TPL) is used as control. Blots and their respective gel staining can be found in Appendix Figure S5.

Cortex homogenates were prepared by centrifuging at 10,000 × g for five minutes to isolate a purer mitochondrial fraction. Pellets were resuspended in a 1.5 M aminocaproic acid, 50 mM Tris-HCl buffer. Homogenates were solubilized with 4 mg of digitonin (Sigma, #300410) per 1 mg of protein. To determine the levels of isolated respiratory complexes and supercomplexes, 50 μg of protein was separated on by BN-PAGE in 3–12% acrylamide gradient gels (Invitrogen) in xCell SureLock Mini-Cell tank containing 1X Running Buffer (Invitrogen, BN2001). Gels were run in Dark Cathode buffer (1X Running Buffer with 5% Cathode Buffer Additive (Invitrogen, BN2002)) for 30–45 min at 150 V. Dark Cathode Buffer was exchanged for Light Cathode Buffer (1X Running Buffer with 0.5% Cathode Buffer Additive) and run until completion at 220 V. Gels were transferred to PVDF membrane (Bio-Rad) via wet transfer in bicarbonate buffer (10 mM sodium bicarbonate, 3 mM sodium carbonate), for 1 h at a constant current of 300 mA. Blots were blocked in 5% milk in 0.1% Tween-20 in PBS and incubated with antibodies against several subunits of mitochondrial respiratory complexes overnight at 4 °C. Blots were incubated with secondary antibody conjugated to horseradish peroxidase (Cell Signaling Technology) for 1 h at room temperature. The reaction was developed by chemiluminescence using SuperSignal Pico West reagent (Thermo Fisher Scientific, #34578). Blots were visualized with ChemiDoc Imaging System (Bio-Rad). Optical density measurements were taken by software supplied by Bio-Rad. CII band density is used as control.

To detect the activity of Complex I in gel, mitochondrial complexes treated with digitonin and separated in 3–12% gels were incubated overnight at room temp, in a buffer of 14 mM NADH (Calbiotech, #481913), 1 mg/mL nitroblue tetrazolium (Sigma, N6876), and 5 mM Tris-HCl pH 7.4. To detect the activity of Complex IV in gel, the above-described gels were incubated overnight at room temp, in a buffer of 50 mM Potassium Phosphate, 1 mg/mL DAB (ThermoScientific, #34001), 24 U/mL catalase (Sigma, #C9322), 1 mg/mL cytochrome c (Sigma, #C2506), and 75 mg/mL sucrose. Optical density measurements were taken in FIJI. CII band density was used as control.

Genomic DNA was extracted from cortex and hippocampal tissue using standard proteinase K, phenol, chloroform extraction, and isopropyl alcohol precipitation. Digital PCR reactions used IDT custom primer and probe sets (Appendix Table S3) in combination with QIAcuity Probe PCR kit (Qiagen) and were performed on QIAcuity One (Qiagen). The mtDNA/nDNA ration was determined using 1 ng of genomic DNA and probes for ND1 to 18S. The AAV viral titer was estimated using 50 ng of genomic DNA and the ratio of hSYN to TTR.

Sample sizes were determined based on previous publications, and independent biological replicates range from 3 to 4 for all experimental modalities used in this study. No data were excluded from the analysis. A few replicates are missing due to a failure of acquisition of the image after dPCR. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. Analysis was performed with Prism GraphPad. Exact p-values can be found in Appendix Table S1.