Amelioration of Leigh syndrome induced by mouse blastocyst complementation with a mutant human mitochondrial ATP synthase 6

Abstract

Background

Mutation of the m. 8993 T > G ATP6 subunit of ATP synthase causes a maternally inherited Leigh Syndrome (LS), a rapidly fatal encephalomyelopathy in childhood called and related Neuropathy, Ataxia and Retinitis Pigmentosa (NARP) in adults. There is no cure, and relevant animal models are few. Here we describe a novel ATP6 mouse model of LS and preclinical efficacy of a gene therapy approach to neutralize the ATP6 mutation by competition with a mitochondrial targeted adeno‐associated virus serotype 9 (AAV9) vector containing wild‐type ATP6.

Methods

LS transgenic mice with an ATP6 mutation were developed by microinjecting AAV2 modified by a COX8 mitochondrial targeting sequence (MTS) into mouse blastocysts to deliver the human ATP6 gene with a T > G mutation at position 8993 of the mitochondrial genome, responsible for clinical LS DNA, protein, and metabolic assays confirmed functional gene delivery and in depth physiological/clinical assessments supported germline transmission of an LS phenotype. Rescue of LS in this model was evaluated after intravenous injection of mito‐targeted AAV9 containing wild‐type ATP6 followed by reevaluation of phenotypes.

Results

LS transgenic mice expressed mutant ATP6 protein in multiple relevant tissues including brain, eye, liver. Mutant ATP6 incorporated into Complex V and conferred decreased ATP synthase efficiency. Mice exhibited hallmarks of the human disease with one or multiple systemic symptoms including early death, paralysis, hunching, vision loss and seizures over six generations. Necropsy revealed spongiform encephalopathy, retinal degeneration, and cardiomegaly. Rescue by mitochondrially directed AAV9 containing wild type ATP6 after disease onset conferred prolonged survival and reduced paralysis. When administered prior to disease onset all treated ATP6 mice survived with evidence of globally improved function.

Conclusions

The results are consistent with efficient delivery and expression of foreign genes in mitochondria by AAV vectors directed by MTS and support further development of such strategies via intravenous injection of mitochondrial targeted AAV9 to deliver corrective genes.

INTRODUCTION

Leigh syndrome (LS) is a progressive neurodegenerative disease that affects as many as 1 in 40 000 live births. It is caused by mutations in nuclear or mitochondrial genes that encode structural components or assembly factors of the electron transport chain and ATP synthase complexes I–V that are responsible for eukaryotic energy production by oxidative phosphorylation (OXPHOS). Over 75 loci on more than 60 genes affecting complexes I–V have been linked to Leigh Syndorme (LS) with about 25% of cases caused by mutations of mitochondrial DNA (mtDNA). LS-associated mtDNA mutations include six genes in complex I, cytochrome c oxidase (COX)-III in Complex IV and the ATP6 subunit of Complex V. Because all mammalian cells contain numerous mitochondria, and correspondingly numerous mt-genomes, mutated and wild type mtDNA can co-exist in conditions known as heteroplasmy. In general, symptoms of disease only manifest when the ratios of mutated to wild type mtDNA are increased to levels that impair OXPHOS sufficiently to restrict physiological functions.

The typical onset of LS is 3–12 months with 83% of cases presenting by age 2 years and evidence of congenital abnormalities in 22% of cases. Although there is more variability in adult-onset LS, the core features are similar. LS subjects typically present with closely similar brain magnetic resonance imaging (MRI) scans that identify symmetrical necrotizing lesions of the basal ganglia, thalamus, brainstem and cerebellum. Symptoms vary and can include psychomotor arrest or decline, hypotonia, spasticity, ataxia, dystonia, lethargy, seizures, infantile spasms, movement disorders, dysphagia, abnormal ocular movements, hearing loss and respiratory failure. Mouse models with mutations or deletions of OXPHOS-related genes that mimic the human LS condition have been described for both nuclear and mitochondrial origin genes. These mouse models have brain MRI profiles that closely mimic human LS, and symptoms that include progressive encephalopathy, growth retardation, hypothermia, ataxia, hypotonia, lethargy, failure to thrive, breathing irregularities and early death.

There is no cure for LS or indeed any of the up to 300 other mitochondrial diseases that have been described, although there is evidence for successful palliative pharmacological treatments, and molecular genetic therapies to correct or mitigate the effects of the mutations are in the pipeline. Mouse models have been developed to mimic LS associated with nuclear gene mutations in complexes I, II and IV, the most common being the complex I NDUFS4 knockout mouse that replicates multiple traits of human LS and has been instrumental in developing therapies. Our group recently described a mouse model that mimics Leber hereditary optic neuropathy (LHON), a mitochondrial disease caused by mutations of the mitochondrial encoded NADH ubiquinone oxidoreductase subunit 4 (ND4) gene. Mouse germline delivery of a mutant human ND4 gene in an adeno-associated virus (AAV) vector targeted to the mitochondria produced mice with hallmark symptoms of LHON that were significantly reversed by subsequent intravitreal delivery of the wild type ND4 gene.

Here, we describe mouse germline delivery of a mutant ATP6 gene, responsible for the most common form of maternally inherited LS (MILS) and related neuropathy, ataxia and retinitis pigmentosa (NARP). Progeny of mice created by blastocyst transduction of a human mitochondrial-targeted mutant ATP6 AAV containing a T to G transition that causes substitution of leucine for arginine at amino acid 156 of the ATP6 protein, developed multiple system disorders including premature death, paralysis, vision loss, seizures and cardiomegaly. To varying degrees, these hallmarks of human MILS were passed along the maternal lineage for six generations. Symptoms were partially mitigated by intravenous administration of a mito-targeted AAV9 containing the wild-type human ATP6 allele delivered either prior to or after disease onset.

MATERIALS AND METHODS

Ethics statement

This study was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All mouse studies were approved by the University of Miami Institutional Animal Care and Use Committee. A6 founder mice were generated by the University of North Carolina transgenic core facility, and A6 offspring were bred and housed at the animal facility of University of Miami.

Plasmids and constructs

A human mitochondrial ATP6 gene fused with a FLAG epitope was synthesized by GenScript, Inc. and cloned in a self-complementary (sc) AAV vector, in which expression was driven by the human mitochondrial heavy strand promoter (HSP). To clone the plasmid sc-HSP-mtATP6FLAG-mCherry, plasmids sc-HSP-mtATP6 and sc-HSP-ND4-mCherry were respectively digested by Xba I and BamH I and religated to create sc-HSP-mtATP6FLAG-mCherry, in which mCherry follows the mtATP6FLAG stop codon (see Figure ). The human mutant m. 8993 T > G ATP6 was generated by site directed mutagenesis (Quikchange II XL site-directed mutagenesis kit, Stratagene) with substitution of T to G at nucleotide position 8993. The primers used for site directed mutagenesis are listed in Table . Plasmid AAV9COX8 was constructed by a large fragment polymerase chain reaction (PCR) and seamless cloning of the cytochrome oxidase subunit 8 (COX8) mitochondrial targeting sequence (MTS) to the N terminus of VP2 of the AAV9 capsid. The forward primer was designed using the COX8 fragment following VP2, and the reverse primer was the fragment of complementary COX8 following the C terminus of VP1. The primers designed for cloning COX8 to AAV9VP2 are listed in Table .

Mito-targeted AAV vector production and purification

All mito-targeted AAV vector production and purifications were performed in-house. Mito-targeted AAV2-ATP6FLAG, AAV2-ATP6FLAG-mCherry and AAV2-mutantATP6FLAG-mCherry were respectively packaged using plasmids sc-HSP-mtATP6FLAG, sc-HSP-mtATP6FLAG-mCherry or sc-HSP-mutantATP6FLAG-mCherry, with pVP2COX8 and pDG at ratios 1:1:4. Mito-targeted AAV2/9-ATP6FLAG-mCherry was produced by combining four plasmids: AAV2/9, sc-HSP-ATP6FLAG-mCherry, VP2COX8 and pHelper (1:1:1:2) or three plasmids: AAV2/9COX8, sc-HSP-ATP6FLAG-mCherry and pHelper (1:1:2). Briefly, 293AAV cells (Cat. No. AAV-100, Cell Biolabs, Inc.) were cultured at 37°C to 50-60% confluency with 10% heat inactivated (HI) FBS DMEM containing 100 U/ml of penicillin and 100 μg/ml streptomycin before transfection. Calcium phosphate transfection was performed according to a standard procedure. Vectors were purified by iodixanol gradient ultracentrifuge and fast protein liquid chromatography using a GE AKTA Purifier. The viruses were concentrated, and buffer exchanged in phosphate buffered saline (PBS), with a Biomax 100 kDa concentrator (Millipore, Billerica, MA). Viral titers were determined by SYBR-Green real-time PCR relative to a standard curve. Viral purity was validated by Coomassie blue stained sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), assayed for sterility, divided into aliquots and stored at −80°C.

Animal development

Microinjections of mitochondrial-targeted AAV2 vectors into fertilized oocytes were performed by the University of North Carolina transgenic core facility using a standard microinjection protocol. Briefly, hybrid (C57BL/6J × DBA/2J) F0 mice, referred to herein as B6D2F1 mice, were obtained from the Jackson Laboratory for production of fertilized oocytes. Superovulation was stimulated by i.p. injection of B6D2F1 females with 5 IU pregnant mare serum gonadotropin (National Hormone and Peptide Program, National Institute of Diabetes and Digestive and Kidney Diseases) after 46–48 h by injection of 5 IU human chorionic gonadotropin (HCG) (Sigma-Aldrich. St Louis, MO). Following HCG injection, females were mated with B6D2F1 stud males, and oocytes were harvested the following day. Approximately 1–2 pL of the mito-targeted AAV2-HSP-mutantATP6FLAG+mCherry virus (1.8 × 10¹² vector genomes [vg]/ml) was microinjected into fertilized oocytes using a continuous flow injection mode. Surviving eggs were implanted into the ampulla of pseud-pregnant Swiss Webster (Taconic) recipient females. After weaning, the resulting offspring were transferred to the University of Miami. A6 mice were bred and analyzed for the presence of the transgene. Viral injections resulted in the offspring listed in Table .

Cell culture, AAV transduction and cell survival assay

Homoplasmic m. 8993 T > G cybrids were cultured with high glucose Dulbecco's Modified Eagle Medium (DMEM) (Invitrogen, Cat.11995) with 10% HI-FBS and 1% penicillin and streptomycin containing 4.5 g/l glucose, 2 mM glutamine and 110 μg/ml sodium pyruvate, and supplemented with 50 μg/ml uridine (normal media). Note that 60%–80% confluent cybrids cells were transduced by mito-targeted AAV9/ATP6FLAG-mCherry (1.8 × 10¹² vg/ml) in DMEM without FBS for 2–3 h., followed by DMEM with 10% FBS for ∼36 h and selection in glucose-free media with 50 μg/ml galactose (selection media) for 5 days. Homoplasmic m. 8993 T > G cybrids in 96-well plates at density 2000 cells per well were infected with mito-targeted AAV containing wild-type ATP6 at multiplicities of infection (MOI) 0, 500, 1000, 5000 and 10 000 for 5 h. in triplicates. After 5 days of selection in glucose-free galactose media, viable cell numbers were counted using a BioRad, TC20.

ATP synthesis assays

The ATP synthesis rates of cybrids and freshly homogenized mouse brains were measured by a modified luciferin-luciferase assay. Luminescence is expressed as relative light unit (RLU). ATP standard curves (RLU relative to ATP concentration) were generated for each assay over a range of 0–100 nM ATP. Homoplasmic cybrids were seeded in 6-well plates and infected with mito-targeted AAV9-mtATP6FLAG at MOIs 0, 500, 1000, 5000 and 10 000 followed by selection for 5 days and culture to confluence in normal growth media (described above). Rinsed pellets of 5 × 10⁵ cells or chilled homogenized mouse brains were resuspended in ATP synthesis buffer A. RLU was recorded by chemiluminescence using a modified luciferin–luciferase assay in digitonin-permeabilized tissues with the complex I substrates malate and pyruvate in real-time (every second for 1–15 s) using a luminometer (Optocom I; MGM Instruments, Hamden, CT) in a total volume of 200 μl (freshly prepared) that included the following: 160 μl of Buffer A (150 mM KCL, 25 mM Tris-HCL, 2 mM EDTA, 0.1% BSA,10 mM Potassium phosphate, 0.1 mM MgCl₂, pH 7.4), 2.5 μl of 1 mM malate in buffer A, 2.5 μl of 1 mM pyruvate in buffer A, 5 μl of 0.1 mM ADP in Buffer A, 5 μl of 0.15 mM Adenosine pentaphosphate in water, 10 μl cybrids or homogenized mouse brain, 10 μl of buffer B (194.4 μl of solution A, 1.6 μl of 100 mM Luciferin in water and 4 μl of 1 mg/ml Luciferase in 0.5 M Tris-acetate pH7.75), 5 μl of digitonin (50 μg/ml) with or without 5 μl 0.05 mg/ml oligomycin in ethanol. The ATP synthesis rates (nmol/min/μg) were calculated from the slope of the curve using excel software with proteins quantification by Bicinchoninic acid (BCA) for each assay. Oxidative phosphorylation (OXPHOS) was calculated as total ATP (without oligomycin) minus the value with oligomycin. ATP rate (nmol/min/μg) normalized to protein concentration and ATP standard curve. ATP and OXPHOS indices are expressed as nmol/min/μg.

PCR, Sanger DNA sequencing and RT-PCR

mtDNA was extracted with a DNeasy Blood & tissues kit (QIAGEN Cat.69504). Sequences of forward and reverse primers to amplify the 604 bp ATP6 fragment were respectively GCT TCA TTC ATT GCC CCC AC, and AGG CGA CAG CAG TTT CTA GG. The 695 bp fragment was amplified with the human ATP6 GCT TCA TTC ATT GCC CCC AC as forward primer and Flag CTT GTC GTC ATC GTC TTT GTA GTC as reverse primer. The two sets of primers were used to amplify both human wild-type and mutant ATP6 fragments from cybrids followed by sequencing to confirm the identity. To determine relative contributions of wild type and m. 8993 T > G mutation ATP6, the signal intensities of the relevant DNA bands separated by agarose gel were measured using Image J.

Mitochondrial isolation, BN/PAGE, immunoblotting and complex V by in-gel activity assay

After washing with cold PBS, cells or tissues were suspended and homogenized in mitochondrial extract buffer (0.25 mM Sucrose, 20 mM HEPS, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA pH 7.4). Homogenized cells or tissues were centrifugated at low speed (750 g for 3 min, 3500 g for 3 min, 5,000 for 3 min) to remove debris and extremely large cellular organelles. Supernatant was centrifugated at 10 500 g for 15 min to isolate crude mitochondria pellets. Mitochondrial pellets were solubilized in Native Page Sample buffer with 2.5% digitonin and proteinase inhibitor, incubated on ice for 15 min, centrifuged at 20 000 g for 30 min at 4°C and the supernatants (mitochondrial protein) were aliquoted and stored in −80°C. Mitochondrial protein concentration were determined by BCA protein assay. Proteins separated on precast native PAGE 4%–16% Bis-Tris gels from life technologies (Cat. BN1002BOX). and processed as per manufacturer's instructions. Fifty μg of mitochondrial protein were used for one dimensional native blue gel Coomassie staining and complex V in-gel activity assays. Transfer membranes were fixed in 8% acetic acid, blocked and detected respectively with FLAG antibody (1:1000, Origen, Cat. No. TA100011), mCherry antibody (1:500, Abcam, Cat. No. ab125096), MT-ATP6 antibody (1:1000, Thermo Fisher Scientific, Cat. PA5-103625) and ATP5A antibody (1:1000, Abcam, ab110273). For two-dimensional SDS-PAGE western blot, 200 μg of mitochondrial protein were used and strips from the first dimensional gel were cut and denatured in 50 mM DTT 1 h before running the second dimensional gel. The gels were transferred to PVDF membrane and detected respectively with FLAG, mCherry, ATP5A and MT-ATP6 antibodies. For complex V in-gel activity, washed one-dimensional native blue gels were incubated in 50 mM Tris pH 8.6 for 1 h followed by incubation with 35 mM Tris, 270 mM glycine, 14 Mm MgSO₄, 5 mM ATP and 0.2% Pb (NO₃)₂ at 37°C overnight. After washing in water, gels were scanned and photographed. SDS-PAGE western blot membranes were respectively reacted with anti-FLAG, anti-mCherry, and anti-VDAC/Porin antibodies (Abcam, Cat.ab186321). Ten microlitre of mito-targeted AAV samples were separated on precast NuPAGE 4%–12% Bis-Tris gels (NP0321BOX) respectively stained by cox8 Antibody (Lifespan Scientific, Cat. LS-C353706 or LS-C664745) and AAV antibody (anti-AAV B1, Progen, Cat.65158, mouse monoclonal B1). ECL chemiluminescence was used to visualize the bands using a FUJI400 imager.

Mass spectrometry

Mass spectrometry (MS) was performed by the Harvard University Mass Spectrometry core facility. Protein bands were excised from one-Dimensional blue native gels, digested with trypsin and subject to MS. Briefly, capillary reverse phase HPLC separation of protein digests were performed on a 10-cm 75-m inner diameter PepMap C18 column (LC Packings, San Francisco, CA) in combination with an in-house capillary HPLC system operated at a flow rate of 200 μl/min. In-line mass spectrometric analysis of the column eluate was accomplished by a quadrupole ion trap instrument (LCQ; ThermoFinnigan, San Jose, CA) equipped with a nanoelectrospray source. Fragment ion data generated by data-dependent acquisition via the LCQ were searched against the NCBI nr sequence database using the SEQUEST (Thermo Finnigan) and Mascot (Matrix Science, Boston) database search engines. In general, the score for SEQUEST protein identification was considered significant when dCn was equal to 0.08 or greater and the cross-correlation score was greater than 2.2. MASCOT probability-based MOWSE scores above the default significant value were considered for protein identification in addition to validation by manual interpretation of the tandem mass spectrometry data.

MRI

High-resolution (4.7 T magnet) MRI (Oxford Instruments Oxford, UK) was performed on mildly symptomatic F0 and F2 A6 mice maintained under 1.5%–2% isoflurane anesthesia. All animals were placed in a prone position with heads held firmly by a purpose-built head coil. T2-weighted images of the brain were acquired.

Laser microdissection and human MT-ATP6 copy number analysis

Frozen blocks of mCherry-stained or unstained retina, spinal cord, brain of transgenic A6 mice were cut into 8–18 μM. Laser capture microdissection was performed using a Leica LMD6500. mtDNA was extracted from mCherry positive cells. Taqman RT-PCR assays were performed to quantify human and mouse mitochondrial ATP6 copy numbers with MT-ATP6 plasmid DNA which was standardized by CHOP standard vector for clinical standard vector titration as a standard curve (Thermo scientific mouse MT-ATP6: Cat. Mm03649417_g1; human MT-ATP6: Hs02596862_g1).

A6 mouse phenotypes and histopathology

Phenotypes of A6 mice including weight loss, off balance, ocular defects, hunching, seizures, paralysis and mortality were observed and recorded over six generations. A6 mice were weighed and anesthetized using ketamine/xylazine (i.p. 80 mg/kg and 20 mg/kg body weight). Tissues of A6 mice including brain, spinal cord, eye, and heart were fixed in 10% formalin. Hematoxylin and eosin (H&E) staining was performed by the histology core laboratory at the University of Miami. Images were captured using a Leica fluorescence microscope (Buffalo Grove, IL).

Transmission EM of transduced cybrids and brain sections

For electron microscopy (EM), cybrids were cultured on gold mesh EM grids or 6 well plate and respectively infected with mito-targeted AAV9 containing wild-type ATP6 at MOI of 5000. Grids were examined for AAV particles 2–4 h. later and fixed by 2% PFA for 10 min and stained with AAV B1 antibody 1:100 (Progen, Cat. No. 65158) and goat anti-mouse 5 nm (gold) as secondary antibody without counterstaining. For 6 well plate 4 h. later, cells were suspended in PBS and fixed in 4% formaldehyde and 1% glutaraldehyde in 0.1 M PBS (pH 7.4) by equal volume of fixative. Cells were spined for 10 min, taken off fixative and added fresh fixative for overnight, then replaced fixative with 8% sucrose in 0.1 M PBS overnight at 4°C. Brains were immersion fixed in 2% paraformaldehyde, 2% glutaraldehyde in 0.1 M PBS buffer. Fixed cells and brains were post fixed in 1% osmium tetroxide and 0.1 M sodium cacodylate-HCl buffer (pH 7.4). Then fixed cell or brains were dehydrated through ethanol to propylene oxide, infiltrated, and embedded in epoxy resin. Semithin longitudinal sections (0.5 μm to 1 μm) of the brain were stained with toluidine blue and placed on nickel grids for ultrastructural examination using a JEOL JEM 1400 TEM and GATAN camera (Orius SC200 CCD). Image quantifications were performed by image J.

Rotarod assay

A6 and wild-type C57BL/6 mice motor and balance were evaluated using an LE8500 rotarod (Panlab, Spain) as per manufacturer's instructions. Time to fall was recorded on the instrument screen with a starting speed of 4 rpm that was gradually increased. Each mouse was given one trial (maximum time 10 min). The time of latency-to-fall and speed were recorded for each trial.

Pattern and flash electrophysiology

Mice were weighed, anesthetized with ketamine/xylazine (i.p. 80 mg/kg and 20 mg/kg body weight) placed on a heating pad (37°C) and restrained using a bit bar and a nose holder with unobstructed vision. A pattern electroretinogram (PERG) electrode, diameter of 0.25 mm, was placed on the corneal surface by means of a micromanipulator and positioned to encircle the pupil without limiting the field of view. Reference and ground electrodes were stainless steel needles inserted under the skin of the scalp and tail, respectively. Before recording, a small drop of balanced salt solution was topically applied to the cornea to prevent drying. A visual stimulus of contrast-reversing (1 Hz, 2 reversal/s) horizontal bars generated by a programmable graphic card (VSG-Cambridge Research Systems, Rochester, UK) on a cathode-ray tube display (Sony Multiscan 500) was aligned with the projection of the pupil. At a viewing distance of 15 cm, the stimulus field covered an area of 69.4 × 63.4 deg. Patterns had a fixed mean luminance of 50 cd/m2. Retinal signals were amplified (10 000-fold) and band-pass filtered (1-30 Hz). Three consecutive responses to 600 contrast reversals each were recorded. The responses were superimposed to check for consistency and then averaged. To obtain a corresponding index of outer retinal function, flash electroretinogram (FERG) was also recorded in response to strobe flashes of 20 cd/m²/s superimposed on a steady background light of 12 cd/m² and presented within a Ganzfeld bowl. Under these conditions, rod activity is largely suppressed while cone activity is minimally suppressed. Averaged PERG and FERG were automatically analyzed and evaluated the major positive and negative waves by Sigma Plot (Systat software Inc., San Jose, CA).

Virus administration

Death was not an intentional endpoint, and mice were euthanized when moribund. Mice were sedated by inhalation with 1.5% to 2% isoflurane. A local anesthetic (proparacaine HCl) was applied topically to the cornea and 10 μl of mito-targeted AAV9 was administrated using retro-orbital injection into mice by the Hamilton syringe (a 32-gauge needle). Mice samples were collected and analyzed as described above.

Statistics

All data plotted with error bars are expressed as the mean ± SEM (standard error of the mean). The p values were generated by analyzing data with student's t-tests, one-way analysis of variance (ANOVA) or two-way ANOVA using GraphPad Prism 7 software. Cumulative proportional survival rate statistical analysis was performed by Kaplan–Meier analyses and statistical analyses were performed with log-rank test.

RESULTS

Development of transgenic ATP6 (A6) mice by blastocyst injection of mitochondrially directed AAV2 carrying mutant human m. 8993 T > G ATP6

The mitochondrial double membrane has proven to be a formidable barrier to manipulating the mitochondrial genome. Still, viruses, linearized and double-stranded DNA appear to traverse them and express exogenous mitochondrial genes. To generate a mutant A6 mouse, we created the mutant human m. 8993 T > G ATP6 gene by site-directed mutagenesis of an sc-HSP-ATP6 plasmid, in which the ATP6 gene was synthesized and fused to a FLAG epitope followed by a stop codon (without a tRNA) and mCherry, using the mitochondrial genetic code, into a sc AAV serotype 2 (scAAV2) plasmid (sc-HSP-mutATP6FLAG+mCherry) (Figure , described in methods). To direct the vector to mitochondria, we appended a cytochrome oxidase subunit 8 (COX8) pre-sequence into the N terminus of one of the three AAV capsid proteins (VP2). Three plasmids including modified pVP2COX8, sc-HSP-mutATPFLAG+mCherry and pDG with mutations at positions 444, 500 and 730 (Y444F, Y500F, Y730F) to induce robust transduction were applied to produce the mitochondrially targeted scAAV2-mutATP6FLAG+mCherry viral vector. After packaging, mito-targeted scAAV2 aliquots were microinjected into mouse blastocysts and implanted into pseudo-pregnant females (Figure ). Of fifty-three founder mice generated by such blastocyst injections (Table ), six females with high ocular expression of mCherry, visualized by laser scanning ophthalmoscopy, were used to generate a total of 462 additional F1-F6 A6 mice. (Figure ). The final A6 colony included 252 females, 203 males and 60 unsexed pups that died within 8 days of birth. High early mortality rates were a major characteristic of F0 to F5 generations (Table ).

[IMAGE OMITTED. SEE PDF]

Transgenic A6 mice develop human-like LS phenotypes

Transgenic A6 mice presented a spectrum of human-like LS or NARP phenotypes that included premature death, weight loss, paralysis, hunching, seizures, motor disfunction and ocular defects. From the entire colony of 515 transgenic mice, F0 through F6, such human-like LS/NARP features were recorded in ∼70% of animals (Table ). Of these, 167/515 (32%) died before 15 months including 101/515 (20%) that died before 6 months, revealing highly significant rates of premature death (Figure , Figure and Table ). Six of 53 F0 mice died before 6 months of age, including four males that died before 3 months. F1 mice incurred the highest 6-month mortality rate of 56%, significantly greater than all other groups (n = 54/96), followed by an F2 mortality rate of 23% that was also significantly greater than the F3-F5 groups. Premature death was generation-dependent, highly significant in F0-F2, reduced but still significant in F3-F5, and absent (not significant) in F6. Of F0 to F6 mice that survived > 15-months, at least 50% displayed one or more phenotypes of hunching, paralysis, motor disfunction (off-balance), visual loss or seizures (Table ). Hunching was observed in 11% (38 of 350) A6 mice with a mean age 19.9 months (Figure ). Three percent of all A6 mice developed severe hinder limb paralysis with a mean age 19.6 months (Figure , Table and Video ). For motor dysfunction, we focused on F0 and F1 generations, (Table ). Accelerating rotarod testing of ambulation and balance revealed latency-to-fall times for F0-F1 mice that were significantly less than the corresponding times observed for similarly aged wild type mice (Figure ). Of 68/101 (67%) randomly selected F0 and F1 mice aged between 6–12 months, 59% showed significant motor disfunction (Table ). Seizures were observed in 0.8% of total A6 mice, with mean age 22.5 months (Figure , Table and Video ). As noted in the Table legend, such seizure estimates are likely low because mice were not under constant surveillance. Reduced weight gain was another significant clinical feature (Table ). A6 mice that succumbed to premature death at age 3–4 months had average body weights of 17 g, significantly less than the corresponding mean weights of age-matched wild types (25 g), (Figure ). Combined assessment of ocular defects by pattern electroretinogram (PERG), flash electroretinogram (ERG) and H&E revealed significantly defective ocular functions in 35 of 43 (80%) of A6 mice selected from F0-F2 generations (Table ), that included declines in 10 of 11 by PERG, (Figure , relative to wild types), and 5 of 11 founders by FERG, (Figure , compared with wild types). Together the results support the presence of pronounced but variable symptoms of A6 mice that mimic those reported for human MILS and NARP. In addition, at least 7% of A6 mice developed tumors with a mean age 20.1 months (Table and Figure ). While not a recognized phenotype of LS, tumors are known to be associated with mitochondrial mutations, including ATP6.

Basal ganglia lesions, retinopathy and central nervous system neuronal degeneration of A6 mice

To determine whether A6 mice develop lesions of the basal ganglia, a condition seen in many human cases of MILS, we imaged five A6 brains and one wild type by MRI. In one 4-month-old, female F2 A6 mouse with quadriparesis and severe weight loss (Figure  and Video ), we observed an increased T2 signal at the head of the hippocampus relative to the wild type, that was bilateral and symmetric (Figure ). By the same MRI technique, the other four mice with only mild symptomatic LS disease were similar to controls (Figure ). Because of the high mortality rate observed in A6 mice, postmortem histological examination was primarily used to characterize anatomical pathologies. Ocular histopathology of deceased A6 mice revealed retinas with normal ganglion cell, inner nuclear and outer nuclear layers (Figure ) and others with complete loss of the photoreceptor and outer nuclear layers (Figure  and Figure ). The thickness of A6 mouse retina outer and/or inner nuclear layers were significantly thinner relative to controls (Figure , ***p < .001). Loss of the retinal photoreceptor layers is characteristic of RP of human NARP and MILS. Light microscopy of A6 mouse brains revealed foci of spongiform encephalopathy where neurons were lost, also a characteristic of human MILS (Figure , Figure ). Transmission EM of A6 mouse brains revealed, in addition to normal deep white matter with intact axons and myelin (Figure ), clear cystic spaces at the lesion foci where fibers and neurons were decreased about two-fold relative to control brains (Figure , **p < .01). Examination of the cerebral cortices revealed increased prevalence of electron dense structures in neuronal mitochondria that coexisted with normal mitochondria, reminiscent of the mitochondrial degeneration that has been reported in MILS necropsy specimens (Figure ). Abnormal cristae of neuronal mitochondria were also apparent with cross-sectional areas up to two-fold wilder than those of adjacent mitochondria (Figure , ***p < .001). Loss of ATP synthase dimerization can induce profound disturbances of mitochondrial cristae ultrastructure, as reported in human LS patients.

[IMAGE OMITTED. SEE PDF]

Heteroplasmy of mutant human m. 8993 T > G ATP6 in A6 mice

Next, we characterized expression of the human mutant ATP6 gene in mouse mitochondria of A6 founder and offspring mice. PCR analysis of DNA extracted from transgenic A6 mouse brain, spinal cord or liver mitochondria using human ATP6 specific primers revealed the presence of the mutant human ATP6 gene (Figure ). Sanger DNA sequencing confirmed that the PCR fragments were mutant human m. 8993 T > G ATP6 (Figure ). SDS-PAGE western blot analysis showed that A6 mouse mitochondrial proteins reacted with FLAG and/or mCherry antibodies (Figure ). Immunofluorescence imaging of mutant human ATP6 fused FLAG revealed a perinuclear pattern in A6 mouse brain and spinal cord neuronal cells indicative of mitochondrial localization (Figure ). Then, we examined the ratio of mutant human ATP6 DNA relative to the endogenous mouse allele from whole tissues by semi-qPCR and quantitative real time (q-rt) PCR. The data indicate low ratios of human mutant ATP6 to mouse ATP6 (Figure ). Such ratios declined significantly in brains from subsequent F3 to F6 progeny (Figure ). Laser capture microscopy of founder mouse brain and retinal layers followed by qPCR revealed that human ATP6 levels were close to 10% in some areas of the brain, and up to 40% in the retinal outer layer (Figure  and Figure ). By applying similar methods of quantification, the corresponding ratios from F1 early death brains ranged from 10%–20% (Figure ), while some of those from F2 and/or later generation of mice with severe LS symptoms were as high as 70% (Figure ). Although higher mutant loads have been described in MILS and NARP autopsy specimens, low mutant loads have also been associated with shortened lifespan phenotypes, consistent with proposals that factors in addition to overall mutant mtDNA load play important roles in determining disease phenotypes. In addition, based on the uneven distribution of mCherry fluorescence described above, it seems probable that the mutant ATP6 DNA loads are dependent on the anatomical location of the affected tissues, particularly neurons (Figure upper) and may, for example be influenced by the high bioenergetic requirements and larger mitochondrial densities of neuronal tissues.

[IMAGE OMITTED. SEE PDF]

Mutant human m. 8993 T > G ATP6 protein assembles into murine complex V and impairs ATP synthesis and hydrolysis

High magnification immunostaining revealed that the transgenic mutant human ATP6-fused flag protein complex co-localized with the mitochondrial marker VDAC1/Porin in A6 mouse brain and spinal cord mitochondria (Figure ). This finding supports the expression of mutant human ATP6 in murine mitochondria. To determine whether mutant human ATP6 protein assembled into the murine complex V, we performed one-dimensional blue native PAGE (BN/PAGE) and two-dimensional SDS-PAGE western blots. One-dimensional blots probed with an anti-FLAG antibody identified a band corresponding to the FLAG-mutant human ATP6 fusion protein that migrated with complex V (∼550-597 kDa, Figure , upper panel). Two-dimensional SDS-PAGE western blots also probed with anti-FLAG and anti-mCherry using a different blot, identified ∼60 kDa bands that represent ATP6FLAG-mCherry (∼56 kDa, Figure -1,2). We then reacted another gel with anti-ATP5A antibody, and confirmed reactivity with the ∼60 kDa band (MW ∼55 kDa), identified as ATP6FLAG-mCherry, migrating as complex V (Figure -3). With increased denaturing time the larger undenatured bands were decreased or lost, consistent with our results from SDS-PAGE western blots that show only the ∼60 kDa, ATP6FLAG-meCherry band (Figures and ). A possible explanation for a fusion protein is that the stop codon between ATP6FLAG and mCherry allows read-through. Another possibility is the mutant human ATP6 fusion (hATP6) lacks the normal secondary structure required for precise endonucleolytic cleavage of mitochondrial mRNAs by murine mitochondrial RNase P. Generation of fusion proteins were also described in our parallel mito-ND4 mouse model. To confirm the identity of mutantATP6FLAG incorporated into complex V, the relevant FLAG-positive band from one-dimensional blue native gel was excised from the gel and analyzed by mass spectroscopy (MS) sequencing. MS and bioinformatics analyses of such extracts revealed the presence of more than 240 unique ATPase peptides that included 4 mt-ATP8 peptides (n = 3, Figure and Table ). The results confirmed identity of the ATP synthase and therefore assembly of ATP6FLAG into complex V.

To investigate the effects of mutant human ATP6 on complex V function, we implemented in-gel complex V activity assays. The natural mitochondrial ATP synthase assembles into dimers with optimal adenosine triphosphatase (ATPase) activity. As shown in Figure  and Figure , we found that specific ATP hydrolysis activity was impaired in parallel with lost dimer bands from A6 mouse brain, spinal cord and muscle ATPase. In addition, the rate of ATP synthesis and oxidative phosphorylation (OXPHOS) in A6 brain extracts were decreased significantly, by ∼50% compared to wild type brains (Figure ). Taken together, these findings are consistent with assembly of mutant human ATP6 subunits into murine complex V, with corresponding loss of ATPase activity in brain, spinal cord and muscle of A6 mice and defective OXPHOS and mitochondrial ATP generation in A6 mouse brains.

Construction of mito-targeted AAV9 with a COX8 signal sequence to deliver wild-type ATP6 into mitochondria

To investigate the feasibility of a gene therapy strategy to ameliorate or preempt the disease phenotypes in this A6 mouse model of human MILS/NARP, we constructed a rescue vector by replacing the mutant human ATP6 with wild-type human ATP6 (Figure ). Unlike AAV2, AAV9 efficiently crosses the blood-brain barrier following intravenous injection therefore we used AAV9 serotype for the rescue. A COX8 signal sequence (mitochondrial targeting) was added to the N terminal of the VP2 sequence of AAV to target to mitochondria. Western blots of the packaged mito-targeted AAV9 vector stained respectively with anti-COX8 and AAV antibodies confirmed the presence of VP2COX8 in mito-targeted AAV9 with the predicted slower mobility relative to the native VP2 (Figure , left and right panel).

[IMAGE OMITTED. SEE PDF]

Transduction of homoplasmic mutant m. 8993 T > G ATP6 cybrids with mito-targeted AAV9 delivering wild-type ATP6

The presence of both mutant and wild type human ATP6 DNA in AAV9-transduced homoplasmic mutant m. 8993 T > G ATP6 cybrids was confirmed by amplifying DNA from transduced and non-transduced cells and digesting fragments with Sma I after 5-day selection in galactose media. The relative representation of wild type and mutant ATP6 can be determined because wild-type ATP6 lacks a Sma I site. As shown in Figure , whereas DNA from non-transduced cybrids was completely digested by Sma I, the fragments from AAV9-transduced cells were incompletely digested. The identity of the undigested band was confirmed by excision from the gel and digestion with BstN1 that cuts wild-type ATP6, but not mutant human ATP6 (Figure , lane 4). Using Sanger DNA sequencing, we further confirmed that the PCR fragments derived from transduced cells contained both mutant and wild-type human ATP6 alleles (Figure , upper), whereas the amplificants obtained from uninfected cells had only the T > G allele (Figure , bottom). The ratio of wild type to mutant ATP6 in mito-targeted AAV9 transduced cybrids, determined by relative band intensities was 25%–47%. Immunofluorescence microscopy of transduced cybrids after staining for FLAG (present as ATP6FLAG in the AAV9 rescue vector) revealed perinuclear FLAG expression that co-localized with MitoTracker Red (Figure ). As a control, immunofluorescent images of homoplasmic m. 8993 T > G cybrid cells transduced with AAV9 containing wild-type ATP6FLAG without the COX8 targeting signal, showed evidence of ATP6FLAG expression in nuclei and cytoplasm, and many cells without signal, markedly different from the COX8 signal group (Figure ). Nuclear expression of the human ATP6 gene encoded under the mitochondrial genetic code was not expected and the relatively low fluorescence observed in some nuclei may represent readthrough of the MT-ATP6 tryptophan stop codons and/or background autofluorescence, often an artifact with such fluorescence staining. Transmission EM of cybrids 2–4 h after infection with mito-targeted AAV9 revealed viral particles associated with membrane-bound organelles (Figure , Figure ). One-dimensional blue native/PAGE western blots of mitochondria from transduced cybrids confirmed reactivity of a 550 kDa complex V band with anti-FLAG antibody as well as with an anti-MT-ATP6 antibody (Figure ). Complex V in-gel assays comparing uninfected and transduced cybrids showed monomer and dimer bands from transduced cybrid cells with stronger ATPase activity especially associated with monomers (Figure ). Of note, the sizes of the major FLAG-positive bands in western blots shown in Figure  are smaller than the corresponding monomer/dimer bands revealed by in-gel ATPase activity assay (Figure ).  The former represents F0F1 subunits, the latter holocomplex V monomers and dimers.  Note also the presence of an enzymatically active F1 band at ∼380 kDa in Figure , right panels.  We attribute the differences to variations of extraction procedure and the unstable nature of complex V that is highly sensitivity to conditions of detergent solubilization, an effect that is exacerbated by the presence of a mutated subunit. Although expression of the wild type ATP6-FLAG variant ameliorates stability of complex V, the ATP6-FLAG is mainly immunodetected in subcomplex, what indicates that the FLAG tag in ATP6 induces its instability. Together, the findings indicate successful transfer and expression of wild type human ATP6 by mito-targeted AAV9 in the mitochondria of homoplasmic m. 8993 T > G cybrid cells and augmentation of complex V activity.

Rescue of homoplasmic mutant m. 8993 T > G ATP6 cybrids phenotypes by mito-targeted AAV9 carrying wild type ATP6

Transduction of homoplasmic mutant m. 8993 T > G ATP6 cybrids with mito-targeted AAV9 delivering wild type ATP6 significantly increased the viability of cells cultured in glucose-free, galactose media relative to uninfected cybrids (Figure ). Luciferin-luciferase assays to measure ATP with malate and pyruvate substrates, revealed significantly increased rates of ATP synthesis of 2.82 and 3.49 nmol/min/μg (Figure ) in cybrids transduced respectively at MOIs of 500 and 1,000 compared with 1.3 nmol/min/μg of uninfected cybrids. Accordingly, the derived rate of OXPHOS was also significantly increased, respectively to 2.44 and 2.79 nmol/min/μg at MOIs of 500 and 1,000 relatives to 0.93 nmol/min/μg of uninfected cybrids (Figure ).

Expression of wild type human ATP6 in mouse brain following intravenous delivery of mito-targeted AAV9

Because AAV9 crosses the blood-brain barrier and has been reported to rescue mice with LS caused by a nuclear gene knockout, we tested AAV9 transduction first in control mice by intravenously injecting 10 μl of sc mito-targeted AAV9 (1.86 × 10¹²vg/ml) expressing wild-type human ATP6FLAG+mCherry into the right orbital venous sinus. Using laser-scanning ophthalmoscopy focused on the inner retina (Figure ), we detected ATP6FLAG+mCherry protein in both right (Figure ) and left eyes (Figure ) of live injected mice but not in noninjected control mice. Relative to noninjected mice (Figure ), confocal microscopy of the brains of injected mice further revealed fluorescence of the ATP6FLAG+mCherry protein in the brain (lower magnification, arrow shown in Figure ). Punctate and perinuclear ATP6FLAG+mCherry stained by FLAG colocalized with VDAC1/Porin, a mitochondrial marker (arrow shown in Figure  of Figure  and at higher magnification in Figure  of Figure ). These findings suggest that the AAV9 vector traversed the blood-brain barrier into mitochondria of central nervous system (CNS) neurons. Therefore, we tested whether delivery of our AAV9 vector was effective in preventing and/or reversing symptoms of MILS in this mouse model.

[IMAGE OMITTED. SEE PDF]

Amelioration of LS symptoms by intravenous delivery of mito-targeted AAV9 containing wild-type human ATP6

Transgenic A6 mice that developed symptoms of severe paralysis or hunching and that were assigned as high risk for imminent death, were administered intravenous injections of 1.86 × 10¹⁰vg of mito-targeted AAV9 containing wild-type human ATP6. Figure  shows examples of the locomotion of a high responder rescued mouse that had severe hind limb paralysis prior to treatment. Symptoms appeared to stabilize within 1 week post injection, and the mouse survived to Day 65 (Video ). Rescues were initiated in two groups and survival was tracked at progressive times post injection (D0). Group 1 included similarly aged A6 mice with severe hind limb paralysis. Log-rank tests showed a 28-day mean survival (range 5–65 days), of such mice in the treatment group (n = 11), while 5 untreated A6 mice with similar hind limb paralysis died within 4 days of D0 (Figure ). In group 2, that included similarly aged A6 mice with severe hunching, using the same protocol, the mean survival of treated mice (n = 14) was 16 days (range 5–47 days), compared with < 3 days for the non-rescued hunching group (n = 12) (Figure ). Complex V in-gel assay reveals the intensity of the bands in the treated A6 mice were slightly stronger than untreated A6 mice (Figure ). The results are consistent with the rapid expression, and phenotypic complementation by the scAAV9-hATP6 vector seen in transduced cybrids (Figure ), and indicate significantly improved survival of treatment groups relative to non-rescued controls.

[IMAGE OMITTED. SEE PDF]

Suppression of LS phenotypes and enhanced Complex V activity by early rescue

Preventative treatment was initiated in non-symptomatic 3-month F2 A6 mice that were selected from the progeny of high-mortality F1 dams. Thirty-four mice from 3 different A6 mice were divided randomly into 10 rescue and 24 non-rescue controls. The treatment group received intravenous injections of 1.86 × 10¹⁰vg mito-targeted AAV9 rescue vector while controls received PBS. All rescue animals (100%) survived through 21 months, while untreated A6 mice sustained 8 of 24 premature deaths between 5–12 months and a cumulative survival rate of 67% (Figure ). Log-rank tests after 15+ months, indicate 56% (9 out of 16) of untreated A6 mice presented severe LS phenotypes including paralysis (1), hunching (4) and weight loss (3), also including one A6 mouse with tumors, whereas 80% of the treatment group were non-symptomatic. The exceptions were 2 treatment mice that showed significant weight loss (Figure ). Three months after injection, visual function measured by pattern electroretinogram (ERG), was significantly improved in treated A6 mice over that of the untreated group (Figure ). Histopathological analysis confirmed that the brains and spinal cords of rescued A6 mice (Figure ) were similar to wild-type mice with markedly less vacuolization and cell loss (Figure ) compared to the severe loss seen in the non-rescue group (Figure ). In addition, non-rescued survivors lost an average of 30% body weight at age 1-year (n = 25), whereas the mean weights of rescued mice were no different from age-matched wild-type controls (Figure 7J). The ratio of heart to body weight of hunching non-rescued A6 mice was almost twice that of wild-type or rescued A6 mice (Figure ). Cardiomegaly, a component of MILS, was also apparent only in the non-rescued A6 hearts (Figure ). Immunofluorescence images of mutant human ATP6 fused FLAG in heart sections revealed a perinuclear expression pattern consistent with presence of the mutant human ATP6 protein in heart mitochondria (Figure , bottom panel). In-gel activity measurements of ATP hydrolysis to assess complex V activity revealed significantly improved ATPase activity of rescued A6 mice (brain) relative to untreated A6 mice (Figure ). The results are consistent with efficient AAV9-mediated delivery of wild type ATP6 by intravenous injection, complementation of the defective mutant ATP6 and effective amelioration of the MILS phenotypes by early gene delivery in this model.

[IMAGE OMITTED. SEE PDF]

Heteroplasmy of human ATP6 alleles in rescued A6 mice

We used PCR and restriction fragment length polymorphism (RFLP) to quantify heteroplasmy of mitochondria in the A6 rescue mouse brain, spinal cord or liver mtDNA. Sequencing of PCR amplified DNA obtained from mitochondria of rescued A6 mice revealed the presence of both wild type and mutant human ATP6 alleles (Figure ). To determine the relative contributions of wild type and m. 8993 T > G mutation of ATP6, the signal intensities of the relevant DNA bands separated by agarose gels were measured using image J and the corresponding percentages calculated for each sample. As shown in Figure , the ratio of wild type to mutant ATP6 in rescued mice was close to 50%. Western blots confirmed ∼10-fold increased intensity of the ATP6FLAG+mCherry protein band in samples from rescued versus non-rescued A6 mice from brains (Figure ).

DISCUSSION

We used a mitochondrial targeted AAV2 vector microinjected into mouse blastocysts to mediate germline transmission of mitochondrial encoded subunit 6 of the human ATPase gene variant m. 8993 T > G. Female founder mice with high ocular expression of the transgene were backcrossed with wild-type C57BL/6 males for 6 generations to generate a total of 53 F0 and 446 progeny F1 to F6 mice, 70% of which displayed one or more hallmark symptoms of LS/NARP. Presence of the m. 8993 T > G variant DNA and protein in transgenic animals, localization to mitochondria and assembly into complex V with dysfunctional ATPase activity were confirmed by molecular, immunoblot and immunostaining analyses of brain, spinal cord, ocular or liver mitochondria. Because of the prevalence of ocular defects and high vulnerability of neuronal ocular tissues to mitochondrial variants that cause LS including the m. 8993 T > G mutation, we focused our molecular genetic analyses primarily on these tissues in F0, F1 and F2 mice. Combined laser capture and RT-qPCR analyses revealed variable heteroplasmy between different retinal locations that approached 40% in the outer nuclear layer and 20%-25% in the ganglion cell inner nuclear layers and close to 10% in founder brain. This compared with 10%-20% in F1 brain sections. Representation of the m. 8993 T > G variant decreased markedly in F3-F6 generations in favor of the wild type allele, and there was evidence of positive correlations between transgene expression and retinal cell loss. Sustained disease phenotypes of F1 and F2 generation mice suggests significant transgene replication and may also be attributed to our selection of high expression founder mice for F0 backcrossing, tissue-specific responses to mutation load, and the apparent absence of tight associations between disease phenotypes and degrees of heteroplasmy often seen in human mitochondrial disease (see below). The results are also consistent with previous studies from our group where we reported similarly variable tissue heteroplasmy after germ line delivery of a human ND4 m. 11778 G > A gene variant to mouse mitochondria. In these studies, we reported a maximum transgene expression of 20% in the retinal RGC layer, 0.08% in the inner nuclear layer and undetectable expression in the outer nuclear layer. Mice with the variant mitochondrial gene displayed robust disease phenotypes that closely mimicked symptoms of human LHON disease despite the relatively low transgene penetration. LS-related symptoms observed in 70% of F0-F6 mice included premature death, paralysis, seizures, weight loss, off balance and visual loss. Inclusive of all offspring, 32% of mice died before 15 months of age with F1 mice showing the highest mortality rate of 50% at 6-months, followed by 23% mortality of F2 mice at 6 months. Of surviving mice, ∼50% became hunched, paralyzed, developed visual loss and/or seizures and weight loss with varying severity and age of onset. Together the results indicate widespread and pronounced but variable disease phenotypes of A6 mice that in many respects mimic human MILS. The presence of the most severe disease phenotypes in F0-F2 mice and loss or reduced severity and prevalence of such phenotypes in subsequent generations in parallel with decline of mutant gene loads is consistent with the expected role of gene load in disabling mitochondrial functions and promoting disease expression in this model.

The importance of gene load has also been described in patients with MILS/NARP, and whereas there is a general association between degree of heteroplasmy and phenotype severity, precise correlations of clinical phenotype with mutation load are rare with few clear thresholds corresponding to a given phenotype, and many instances where severe symptoms are associated with lower mutant gene loads and vice versa. For example, Stendel et al., recently quantified heteroplasmy and clinical presentation of 132 subjects with known pathogenic MT-ATP6 mutations and related the genotype with neurological manifestation of the disease. They reported high degrees of heteroplasmy but wide variations in both symptomatic and non-symptomatic individuals. For symptomatic subjects, the mean heteroplasmy was 95% with a range of 20%–100%, while for unaffected individuals the mean was 73%, also with a range of 20%–100%. Similarly, while most cases of LHON are homoplasmic, severe symptoms have been documented with as little as 33% mutated mtDNA, and a m. 5545 C > T mutation affecting tRNA specificity that usually carries a 70%–90% threshold was reported in one case study to confer a severe disease phenotype at less than 25% in affected tissues, and between 4% and 8% in cybrids generated using patient mitochondria. The apparent clinical dominance of such low copy number mitochondrial mutations has also been reported in rodent models, and may reflect tissue-specific and/or environmental effects including local bioenergetic demands and levels of antioxidants. Additionally, the basal rate of metabolism, or mass-specific-metabolic rate of mice is seven-fold greater than that of humans, placing correspondingly greater demands on mitochondrial function and energy production especially in high energy demand tissues. Consequently, higher rates of ROS production, major drivers of mitochondrial disease including LS, and predicted to increase with higher mass-specific-metabolic rates, are more likely to destabilize redox homeostasis in mouse models, increasing susceptibility to defective mitochondrial function. Such activities are also thought to contribute to the reduced lifespan and increased vulnerability of mice to age-related tumorigenesis relative to humans, both key characteristics of clinical LS. The apparent dominant negative actions of such mitochondrial mutations can account for the persistence of severe phenotypes in our F3 to F6 progeny despite dilution of the m. 8993 T > G variant in favor of wild type ATP6. It would also be predicted that phenotypes of ATP6 mice created in the laboratory would be more homogeneous than human MILS/NARP but subject to similar selection pressures for gene expression and tissue distribution, including dependence on tissue-specific energy requirements and this may contribute to the observed relation and sometimes apparent discordance between heteroplasmy and phenotype. In agreement with our previous study, we found no evidence for integration of AAV-encoded ATP6 DNA into the mitochondrial genome (ATP6 data not shown) and we attribute transgene dilution to a predicted less efficient replication of episomal AAV transgene DNA that is not fully in synchrony with that of mtDNA, as previously reported.

Whereas the correlations between disease phenotype, variant m. 8993 T > G load and defective ATPase activity, including the markedly higher retinal cell loss associated with variant load in the outer retinal layers of F0–F2 mice are consistent with molecular pathologies reminiscent of clinical MILS/NARP, our demonstration that the disease phenotype was rescued by AAV9 delivery of a wild type ATP6 gene not only supports the link between the disease phenotypes and mutant ATP6 load, but also provides a basis for treatment of LS by gene therapy. By using the same strategy described previously by our group for rescue of LHON mice, we delivered a mito-targeted wild type ATP6 gene to m. 8993 T > G transgenic mice by orbital venous sinus injections of AAV9 to achieve efficient penetration of the blood brain barrier and gene targeting to mitochondria. As also described in the previous study, rescue was robust especially when gene delivery was implemented prior to symptom onset. Our molecular analyses of heteroplasmy in rescued mice confirmed the co-presence of variant and rescue ATP6 DNA and protein in rescued mitochondria as well as significantly improved ATPase activity as assessed by in-gel activity of ATP hydrolysis in early rescue tissue samples. Rescued phenotypic parameters of mice containing the m. 8993 T > G mutation treated at age 3-months included 100% survival through 15 months without severe symptoms, compared with a 12-month 33% mortality of non-rescued transgenic mice, significantly improved visual function as well as normalized body weights and resolved cardiomegaly. In addition, histopathology confirmed markedly less vacuolization of brain and spinal cords in rescued animals and reduced cell loss. For mice that were already at end stage disease with severe paralysis or hunching, a stage where deterioration of the brain motor cortex and spinal cord, and significant neuronal loss is expected (see Figure ), rescue conferred remarkably enhanced functional recovery and survival. In one mouse, alleviation of symptoms was apparent within 4 days of AAV delivery that is consistent with in vitro transduced cells that show improved survival under selection only 36 h after gene delivery. Whereas a more rapid response time in expected for scAAV versus conventional ssAAV, the mouse in question had milder symptoms than others in the group and may represent an outlier. While we did not directly measure ATP production after rescue, because all mice in these groups were allowed to survive until they eventually succumbed to the disease, it is expected that gene therapy provided at least partial restoration of mitochondrial function and energy homeostasis, such that further neuronal loss was reduced, and neuromuscular functions temporarily restored.

The absence of animal models that accurately reflect genotypes and corresponding disease phenotypes of mitochondrial diseases, including MILS/NARP have hindered development of effective therapies. Nuclear localization and transcription (allotropic expression) of the ATP6 m. 8993 T > G mutation followed by cytoplasmic translation and transport into mitochondria was reported to mimic mitochondrial disease in mice. Homoplastic models can be approximated by gene knockout and such models have been used to demonstrate amelioration of disease phenotypes by AAV-based complementation of an Ndufs4 gene knockout in mouse models of LS. Similarly, mutant mtDNA loading of mice with a genetically manipulated heteroplasmic m. 5024 C > T gene mutation was markedly reduced by AAV9-mediated delivery of mito-targeted TALENs that selectively degraded mutant mtDNA and restored normal levels of the affected mitochondrial alanine tRNA gene product.

CONCLUSIONS

To our knowledge, studies reported here represent the first to directly target mitochondria with a human m. 8993 T > G mutation, establish and characterize a disease phenotype that is consistent with clinical presentation of patients with the same mutation, and confirm gene therapy rescue of affected animals by complementation with a similarly targeted AAV9 vector expressing wild type ATP6. The results support further development of gene therapy approaches for MILS/NARP and other mitochondrial diseases perhaps in combination with newer pharmacology based on amelioration of oxidative stress caused by excessive ROS leakage from defective mitochondria and known to play a central role in cytotoxicity and cell death during mitochondrial disease. The studies support and extend our previous work on gene therapy for LHON that are now in clinical development.

STUDY LIMITATIONS

We have not defined mechanisms or molecular pathways of rAAV2 activation or transmission of the ATP6 transgene in microinjected zygotes. AAV activation occurs classically by clathrin-dependent endocytosis, pH-dependent release into the cytoplasm followed by intracellular trafficking and uncoating in the nuclear compartment. AAV2 processing can occur uniquely via an alternative pleiomorphic CLIC/GEEC pathway, independently of clathrin, caveolin and dynamin that requires remodeling of the actin cytoskeleton and cytoplasmic transport by Rac-1- and PI3K-independent endosomal pathways. Our results are consistent with capsid processing by an alternate pathway(s), intracytoplasmic transport that is independent of classical plasma membrane endocytosis and delivery of capsids/DNA to mitochondria. Such alternative processing and delivery are also implicated respectively by the work of Kaeppel et al. and Yasuzaki et al. Indeed, multiple pathways to enhance DNA delivery to mitochondria in settings of gene therapy have been described and are the subjects of recent comprehensive reviews. Early studies described entry of genetic material including naked DNA and viral particles into mitochondria and possible mechanisms of transport that included involvements of the voltage-dependent anion channel and mitochondrial permeability transition pore respectively to traverse outer and inner membranes. Replication of integration deficient viral DNA, including AAV and Lentivirus in episomal configurations has also been demonstrated, although the molecular pathways remain speculative. Involvement of the inverted terminal repeats sequences that can also enhance transgene expression, and links with the mitochondrial replication/DNA repair apparatus have been suggested, but unconfirmed. Close links between mitochondria and replication of viruses such as SARS-CoV-2 (COVID-19) and Flock House Virus have been reported, and a class of viruses known as mitoviruses has been described that replicate exclusively in the mitochondria, although again the mechanisms are unknown. In studies described here, as well as our previous work on LHON, we propose that a critical step that determines localization of rAAV to mitochondria, and indeed the success of the strategy, involves mitochondrial targeting of the capsid using the COX8 MTS. In the absence of such targeting, foreign genes encoded by the mitochondrial genetic code with mitochondria-specific promoters would be delivered to the cell nuclei where they cannot be expressed, and result in abortive gene transfer. The requirement of the MTS for ND4 gene transfer and translation of mitochondrial encoded genes was unambiguously confirmed by our previous work. In the present study, we did not investigate the mechanism of AAV replication or germline transmission of the ATP6 transgene, however direct evidence for efficient replication of the ND4 transgene associated with mitochondria was presented in our previous model of LHON wherein we describe carryover of a DNA fragment of the hND4 gene selectively in the mitochondria of progeny transgenic, but not control WT mice.

AUTHOR CONTRIBUTIONS

H. J. Y was responsible for creating and characterizing the LS mouse model, rescue vector construction and production, data compilation, data analysis, and manuscript writing. K.A.W contributed data analysis, interpretation, and manuscript writing. M.T.B., W.W.H., and A.S.L., contributed proof reading and reagents. J. G. directed the study, performed hands on application of mouse MRI, EM, and AAV injections, and contributed importantly to the initial compilation of the manuscript.

ACKNOWLEDGEMENTS

The work was conceived and directed by Dr. John Guy who died shortly after the studies were completed. We remember John as an outstanding biomedical scientist with a lifelong commitment to mitochondrial disease therapy. We thank Dr. Eric A Schon for generously providing Cybrids (NARP) cells, Dr. Hong Yu for providing the sc-HSP-ATP6 plasmid, Dr. Chuanhui Dong and William J. Feuer for expert advice on statistical analyses.

FUNDING INFORMATION

This study was supported by the National Eye Institute R01EY027414 (Guy), R01EY 017141 (Guy), R24EY028764 (Webster), R43EY031238 (Webster), P30 EY014801 Bascom Palmer Eye Institute Core Grant and the University of Miami Electron Microscopy Core IS100D012061 (Salas). Institutional support to BPEI was from a Research to Prevent Blindness Unrestricted Grant (GR004596) and an NEI Center Core Grant (EY014801).

CONFLICT OF INTEREST

John Guy is inventor of US patent 8,278,428 B2: Mitochondrial nucleic acid delivery systems. All other authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported. The paper was handled by editors and has undergone a rigorous peer-review process.

DATA AVAILABILITY STATEMENT

Data and reagents are available to the scientific community on request to the corresponding author.

ETHICAL APPROVAL

This study was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All mouse studies were approved by the University of Miami Institutional Animal Care and Use Committee (IACUC). A6 founder mice were generated by the University of North Carolian transgenic core facility and A6 offspring were bred and housed at the animal facility of University of Miami.

References

Gerards M, Sallevelt SC, Smeets HJ. Leigh syndrome: resolving the clinical and genetic heterogeneity paves the way for treatment options. Mol Genet Metab. 2016;117:300‐312.

Darin N, Oldfors A, Moslemi AR, Holme E, Tulinius M. The incidence of mitochondrial encephalomyopathies in childhood: clinical features and morphological, biochemical, and DNA abnormalities. Ann Neurol. 2001;49:377‐383.

Ruhoy IS, Saneto RP. The genetics of Leigh syndrome and its implications for clinical practice and risk management. Appl Clin Genet. 2014;7:221‐234.

Reynaud‐Dulaurier R, Benegiamo G, Marrocco E, et al. Gene replacement therapy provides benefit in an adult mouse model of Leigh syndrome. Brain. 2020;143:1686‐1696.

Kirby DM, Crawford M, Cleary MA, Dahl HH, Dennett X, Thorburn DR. Respiratory chain complex I deficiency: an underdiagnosed energy generation disorder. Neurology. 1999;52:1255‐1264.

Quintana A, Zanella S, Koch H, et al. Fatal breathing dysfunction in a mouse model of Leigh syndrome. J Clin Invest. 2012;122:2359‐2368.

Kruse SE, Watt WC, Marcinek DJ, Kapur RP, Schenkman KA, Palmiter RD. Mice with mitochondrial complex I deficiency develop a fatal encephalomyopathy. Cell Metab. 2008;7:312‐320.

Di Meo I, Marchet S, Lamperti C, Zeviani M, Viscomi C. AAV9‐based gene therapy partially ameliorates the clinical phenotype of a mouse model of Leigh syndrome. Gene Ther. 2017;24:661‐667.

Yu H, Koilkonda RD, Chou TH, et al. Consequences of zygote injection and germline transfer of mutant human mitochondrial DNA in mice. Proc Natl Acad Sci U S A. 2015;112:E5689‐E5698.

Yu H, Koilkonda RD, Chou TH, et al. Gene delivery to mitochondria by targeting modified adenoassociated virus suppresses Leber's hereditary optic neuropathy in a mouse model. Proc Natl Acad Sci U S A. 2012;109:E1238‐47.

Yu H, Ozdemir SS, Koilkonda RD, et al. Mutant NADH dehydrogenase subunit 4 gene delivery to mitochondria by targeting sequence‐modified adeno‐associated virus induces visual loss and optic atrophy in mice. Mol Vis. 2012;18:1668‐1683.

Rahman S. Emerging aspects of treatment in mitochondrial disorders. J Inherit Metab Dis. 2015;38:641‐653.

Silva‐Pinheiro P, Cerutti R, Luna‐Sanchez M, Zeviani M, Viscomi C. A single intravenous injection of AAV‐PHP.B‐hNDUFS4 ameliorates the phenotype of Ndufs4 (‐/‐) mice. Mol Ther Methods Clin Dev. 2020;17:1071‐1078.

Al Khazal F, Holte MN, Bolon B, White TA, LeBrasseur N, Iii LJM. A conditional mouse model of complex II deficiency manifesting as Leigh‐like syndrome. FASEB J. 2019;33:13189‐13201.

Jain IH, Zazzeron L, Goldberger O, et al. Leigh syndrome mouse model can be rescued by interventions that normalize brain hyperoxia, but not HIF activation. Cell Metab. 2019;30:824‐832.

El‐Desouky S, Taalab YM, El‐Gamal M, Mohamed W, Salama M. Animal model for Leigh syndrome. Methods Mol Biol. 2019;2011:451‐464.

Cogliati S, Frezza C, Soriano ME, et al. Mitochondrial cristae shape determines respiratory chain supercomplexes assembly and respiratory efficiency. Cell. 2013;155:160‐171.

Campanella M, Casswell E, Chong S, et al. Regulation of mitochondrial structure and function by the F1Fo‐ATPase inhibitor protein, IF1. Cell Metab. 2008;8:13‐25.

Drew B, Leeuwenburgh C. Method for measuring ATP production in isolated mitochondria: ATP production in brain and liver mitochondria of Fischer‐344 rats with age and caloric restriction. Am J Physiol Regul Integr Comp Physiol. 2003;285:R1259‐R1267.

Fontanesi F, Diaz F, Barrientos A. Evaluation of the mitochondrial respiratory chain and oxidative phosphorylation system using yeast models of OXPHOS deficiencies. Curr Protoc Hum Genet. 2009. [DOI: https://dx.doi.org/10.1002/0471142905.hg1905s63]