Proteomics reveal PTEN as a critical mediator of sustained mitochondrial dysfunction during chronic spinal cord injury

Abstract

Activity of the phosphatase and tensin homologue protein (PTEN) remains elevated in neurons chronically after spinal cord injury (SCI) and suppresses tissue repair. However, PTEN may also disrupt other neuronal functions not directly related to regeneration. To better understand the role of PTEN on neuronal functions in chronic SCI, neuronal-specific PTEN-KO was induced using spinal injections of retrogradely-transported AAVs (AAVrg) immediately after contusion SCI in mice. Spinal cords were harvested at 6 weeks post-injury and untargeted total proteomics was performed. Bioinformatics analyses revealed a downregulation of mitochondrial-associated proteins in chronic SCI that was reversed after PTEN-KO. We replicated the experimental conditions to validate the effects of chronic SCI±PTEN-KO on mitochondrial functions using ex vivo respiratory testing on whole-spinal cord mitochondrial isolates. Mitochondrial respiratory capacity was reduced in chronic SCI and was restored after PTEN-KO. Next, we evaluated the extent to which chronic SCI specifically affects neuronal mitochondria and whether PGC1α upregulation can restore respiratory capacity. We designed an AAVrg vector to enable a magnetic bead pulldown approach to isolate neuron-specific mitochondria with, or without, concurrent PGC1α upregulation. AAVrg vectors were delivered into the spinal cord at 15-weeks post-injury, and neuron-specific mitochondria were isolated 6-weeks later. Neuronal mitochondria present a ~50% loss of respiratory capacity in chronic SCI that was restored with PGC1α upregulation. Collectively, we demonstrate that mitochondrial respiratory abilities are significantly repressed chronically after SCI, that PTEN is a major contributor to sustained mitochondrial dysfunction, and that PGC1α upregulation can restore mitochondrial bioenergetic abilities during chronic SCI.

1. 0. Introduction:

Paralysis after a spinal cord injury (SCI) remains a permanent disruption to quality of life, for which there is no cure or available treatment options. Most clinical cases of SCI result in incomplete injuries that present with some spectrum of spared axons that can elicit limited voluntary motor functions or conscious sensations¹. Even conditions characterized as motor- and sensory-complete often present evidence of discomplete lesions that retain some anatomically-spared connections that are insufficient to elicit voluntary motor abilities or conscious perceptions^{2, 3}. The prospects of leveraging spared and intact neural architecture to improve functional abilities after SCI has been proven with recent clinical trials using epidural/transcutaneous stimulation^{4, 5}. Collectively, emerging evidence supports that targeting spared axon pathways can be leveraged to improve motor and sensory functions during chronic stages after SCI. However, regenerating damaged axons remains a coveted goal for the prospects of restoring normal to near-normal functions.

As an experimental approach to drive axon regeneration after SCI, inhibiting the phosphatase and tensin homologue protein (PTEN) has withstood the test of time as efficacious and reproducible in animal models. In our past work we leveraged neuronal-specific retrogradely transported Adeno-Associated Viral vectors (AAVrg’s) to limit the effects of PTEN-manipulations to only the spinal-projecting neurons that have been damaged or are spared from SCI^{6, 7}. To our surprise, we did not observe evidence of regeneration when AAVrg’s were used to knock out PTEN (PTEN-KO) in chronic SCI. Instead, locomotor functions improved within 2 weeks after treatment delivery with histological assessments revealing the ability for AAVrg’s to target spared axons tracts, suggesting non-regenerative mechanisms may underlie the observed treatment effects⁷.

PTEN is a potent lipid and protein phosphatase with the best described roles as an inhibitor of the mammalian target of rapamycin (mTOR) pathway^{8, 9}. Non-regenerative effects elicited by PTEN-KO include increasing neural excitability¹⁰, suppressing autophagy^{11, 12}, and augmenting cap-dependent protein translation which are mediated via mTOR pathway disinhibition¹³. Ultimately the effects of mTOR pathway augmentation in spared neural circuits after SCI have not been fully elucidated. What is known about PTEN regulation after SCI is that PTEN remains chronically hyperactive in neurons around the lesion which contributes to growth cone collapse and regenerative failure^{8, 9}. Due to the diverse cellular roles mediated by the mTOR pathway, we hypothesized that PTEN regulates broad molecular pathways that may affect neuronal functions independent of regeneration that can be manipulated to improve motor functions through spared axon pathways.

To begin interrogating the role of PTEN during chronic stages after SCI, we utilized untargeted total proteomics to evaluate how neuron-specific PTEN-KO affects cellular physiology at the protein level. Mitochondrial dysfunction emerged amongst the largest affected Gene Ontology (GO^{14, 15}) biological process affected by SCI that was subsequently reversed after PTEN-KO. We validated that the respiratory abilities of mitochondria obtained from the perilesional environment chronically after SCI remains dysfunctional and that neuron-restricted PTEN-KO significantly restores mitochondrial respiratory abilities. To determine the extent to which mitochondrial dysfunction occurs within neurons, we pulled down tagged neuronal mitochondria and revealed a ~50% loss of respiratory capacity in neuronal-specific mitochondria at 21-weeks post-injury. Finally, we demonstrate that overexpressing the peroxisome proliferator-activated receptor-γ coactivator 1-α (PGC1α) in neurons at 15-weeks post-SCI can restore mitochondrial bioenergetic abilities 6-weeks after treatment during chronic SCI.

2. 0. Materials and Methods:

2.1. Animals and study design.

Mice were obtained from the Jackson Laboratory either as wildtype C57BL6/J (#000664; age = 10-12 weeks) or PTEN-flox (PTEN^fl/fl; B6.129S4-PTEN^tm1hwu/j; #006440; bread in house and used between 12-20 weeks of age). PTEN^fl/fl mice were bred homozygous and both male (n=21) and female (n=22) mice were used for proteomics assessments and mitochondrial respiratory testing. For evaluating effects using neuron-specific mitochondrial isolation approaches (described below), with or without PGC1α overexpression, only female mice were used (n=33). Prior to performing PGC1α experiments we also utilized n=10 female mice to validate the ability to upregulate PGC1α using AAVrg’s in naïve mice. For proteomics testing, a total of 22 mice were used with attrition observed due to 1 death occurring during surgery and 5 mice being excluded due to evidence of absent or poor AAVrg labeling. For mitochondrial testing after PTEN-KO, n=21 mice were used with 3 deaths occurring during surgery. For neuron-specific mitochondrial testing, n=33 mice were used with n=7 mice dying either during the initial surgery, during post-operative periods, or during re-exposure surgery. No attrition was observed due to any delivered treatment. In total, n=86 mice were used prior to attrition to complete all experiments. All procedures were approved by the University of Kentucky’s Institutional Animal Care and Use Committee under protocol number 2023-4281.

In total we performed 3 major experiments including: 1) an interrogation into the effects of chronic SCI and PTEN-KO using total proteomics; 2) an interrogation into the effects of chronic SCI and PTEN-KO on mitochondrial respiratory abilities; and 3) an evaluation of the effects of chronic SCI on neuron-specific mitochondria with or without PGC1α upregulation applied during chronic SCI. We focused our proteomics analyses on two major comparisons first being naïve/uninjured mice against mice receiving a spinal contusion (Naïve vs SCI), as well as mice receiving SCI to mice receiving SCI and neuron-specific PTEN-KO (SCI vs PTEN-KO). PTEN-KO was not performed in uninjured mice.

For both experiments utilizing PTEN-KO, mice received a 100 kDyn SCI to reproduce a severe, discomplete injury, and AAVrg treatments were delivered immediately post-SCI to knockout PTEN. Mice were allowed to recover for up to 6-weeks before harvesting tissue for either proteomics or mitochondrial respiratory testing. Proteomics assessments were performed at the MaineHealth Proteomics and Lipidomics Core Facility, while mitochondrial respiratory testing was performed at the University of Kentucky’s Central Nervous System (C.N.S)-Metabolomics core. To determine the role of chronic SCI on mitochondrial functions specifically in neurons, as well as determine the ability for PGC1α upregulation to restore respiratory abilities, we treated wild-type mice with AAVrg’s at 15-weeks post-SCI. AAVrg’s delivered in our final experiment expressed an outer-mitochondrial membrane targeting eGFP with a fused HA tag for high-affinity bead pulldown, along with a second AAVrg to co-express either mCherry with or without PGC1α. Neuron-specific mitochondria were isolated at 21-weeks post-SCI and evaluated for respiratory capacity. Mice received a 90 kDyn SCI for our last experiment because our ongoing work has revealed that a 100 kDyn SCI leaves little-to no spared axons, which would prohibit the ability to test the effects of PGC1α on locomotor functions. All experiments included age- and genotype-matched naïve or laminectomy-only controls.

2.2. Surgical modeling.

Surgical modeling was performed as previously described¹⁶. Mice were anesthetized using ketamine (100 mg/kg) and xylazine (10 mg/kg) before removing hair overlying the thoracic vertebra. A sharp #11 scalpel was used to create a 1-cm lesion over the T9 vertebra. Fat, fascia, and muscles were dissected over the T9 vertebrae and the interspinous muscle connecting the T9 to T10 vertebra was cut. Spring-formed scissors were inserted under the lamina and bilateral cuts were made to complete the laminectomy. Forceps were inserted through the muscles around the T8 and T10 vertebra to suspend the mice and stabilize the spinal column for contusive SCI. A 100 or 90 kDyn contusion SCI was performed. For mice acutely treated with PTEN-KO, mice were transferred onto a stereotactic frame where their spinal columns were stabilized for midline injections to deliver AAVrg’s. For mice treated chronically after SCI, the mice were anesthetized as just described and the scar tissue overlying the spinal cord was gently dissected away, before stabilizing on a stereotactic frame. In total, 2 μL of 1 × 10¹³ gc/mL of AAVrg’s were injected midline rostral to the lesion using pulled-glass needles. AAVrg’s delivered a Cre-p2a-dTomato under a Synapsin 1 promoter (107738-AAVrg; Addgene, Watertown, MA) as previously described for PTEN-KO, or PGC1α-p2a-mCherry under a Synapsin 1 (Syn1) promoter uilizing concurrent injections with a CamKIIa promoter expressing an HA-tag-eGFP-3’ OMP25 fusion peptide to target mitochondria to the outer mitochondrial membrane. Control viruses utilized a Syn1-mCherry (114472-AAVrg; Addgene, Watertown, MA).

After SCI and/or AAVrg injections, the muscles, fascia, and fat were sutured using absorbable sutures and mice were given 1-mL saline, enrofloxacin (2.5 mg/kg), and buprenorphine extended-release formulation (1 mg/kg). All mice received daily manual bladder expressions twice per day for the duration of in vivo survival time. Male mice were single-housed post-SCI to prevent aggressive mutilation that often results in early mortality¹⁷.

2.3. Tissue harvesting and preparation.

Mice were anesthetized with ketamine (180 mg/kg) and xylazine (20 mg/kg) and received cardiac perfusions using saline or mitochondrial isolation buffer (described below) to remove blood. Approximately 0.7 cm above, but including, the lesion site was removed to assess total spinal-cord derived proteins and isolated mitochondria. For mice utilized for proteomics and mitochondrial assessment, 4% formaldehyde prepared from paraformaldehyde was perfused through the heart after spinal cord removal to fix the remaining tissue. Brains were removed for histological assessments as well as the lumbar spinal cords a to assess mitochondrial densities within transduced axons. Brains from mice for the proteomics assessment were evaluated for successful viral labeling within the brain to validate an efficient and successful injection of AAVrg’s into the spinal cord. Of the 6 starting mice within the PTEN-KO group, 2 failed to present with successful viral labeling and were omitted from proteomics assessments.

2.4. Viral vector construction, packaging, and validation.

Syn1-Cre-p2a-dTomato and Syn1-mCherry AAVrg’s were purchased from Addgene.org, while CamK2a-Ha-eGFP-3’-OMP25 vector was constructed and packaged at Vectorbuilder Inc. as described above. Syn1-PGC1α-p2a-mCherry was constructed in the Petersen Lab. The mouse isoform of the peroxisome proliferator-activated receptor-γ co-activator 1-alpha (PPARGC1A or PGC1α) sequence was synthesized and subcloned into a p2a-mCherry backbone. The AAVrg helper plasmid was used for AAV packaging to confer the retrograde pseudotype as previously described¹⁸ (Addgene plasmid #81070; Addgene, Watertown, MA). AAV particles were generated using plasmid triple transfection into HEK293T cells. At 3-days post-transfection HEK cells were harvested, lysed, AAV particles were precipitated using poly-ethylene glycol 8,000 mw (PEG 8,000), followed by purification using chloroform phase separation steps to remove PEG and lipid debris. Soluble AAV particles were dialyzed followed by concentration with a 100 kDa Amicon ultracentrifugation filter (UFC510024; MilliporeSigma). Purified AAV particles were collected and tittered for genome copies using qPCR.

AAVrg upregulation of PGC1α was validated by injecting 2 μL of 1 × 10¹³ gc/mL AAV particles into n = 10 naïve spinal cords using the constructed PGC1α-p2a-mCherry vector or mCherry alone. At 2-weeks post-injection, fresh spinal cord tissue was isolated as described above to be used for western blotting. Spinal cords were homogenized in RIPA buffer (20-188; Bio-Rad) containing protease inhibitors (cOmplete Mini, EDTA-Free; 11836170001; Sigma-Aldrich), and 30 μg of protein wa s added to each well of a 4-15% acrylamide SDS-PAGE gradient gel (4561086; Bio-Rad). Protein was transferred using the Bio-Rad Trans-Blot Turbo Semi-dry transfer system onto nitrocellulose membranes (1704158; Bio-Rad). Membranes were incubated in 5% milk for 1-hour before incubation in primary antibodies targeting PGC1α (Rb anti-PGC1α; 1:2,000; NBP1-04676; NovusBio), mCherry (Ck anti-mCherry; 1:5,000; MCHERRY-0100; Antibodies Inc.), and GapDH (Ms anti-GapDH; 1:5,000; G8795; Sigma-Aldrich). For the first round of imaging, PGC1α (~120 kDa) and mCherry (~27 kDa) were revealed using IRDye-700 and 800 (1:5,000; LICORbio) respectively, followed by adding secondary antibodies to reveal GapDH (~37 kDa) using IRDye-700. The fold-change of PGC1α was quantified, along with any difference observed between mCherry production between the two AAV viral preps.

2.5. Total protein isolation and preparation for proteomics and mass spectrometric analysis.

2.5.1. Sample preparation for mass spectrometry.

Spinal cords isolated for proteomics analyses were quickly frozen on dry ice and stored at −80°C until shipment to the Proteomics and Lipidomics Core at MaineHealth. Tissue samples were solubilized using plastic disposable tissue homogenizers (1.5 ml tube and pestle style, Fisher Scientific) in ice-cold 50 mM Tris buffer (pH = 8.5) containing 2% sodium dodecyl sulfate (SDS, Molecular biology grade, MP Biomedicals) and 1 mM phenylmethylsulfonyl fluoride (PMSF, MP Biomedicals) added immediately before lysis from a 150 mM stock solution prepared in isopropanol and stored at −80°C. Each sample was then subject to probe-tip ultrasonication on ice (3 × 30 second bursts, 25% power, 30% duty cycle, Branson Ultrasonifier 250, Brookfield, CT) before centrifugation (4°C, 10,000 × g) to remove debris. Protein concentrations were measured using the Pierce bicinchoninic acid (BCA) assay (Thermo Fisher Scientific).

For each sample, 20 μg protein were transferred to a 2-ml tube and brought to 5 mM tris-(2-carboxyethyl) phosphine hydrochloride (TCEP, Gold Biotechnology) from a 300 mM stock solution prepared in 1.0 M Tris base (Ultra-pure grade, Alfa Aesar) and stored at −80°C. Samples were reduced for 25 minutes at 56°C before cooling to room temperature. Each was brought to 15-20 mM iodoacetamide (I-AcNH2, MP Biomedicals) from a 500 mM solution in water stored at −80°C, and alkylation allowed to proceed in the dark for 25 minutes. Each reaction was then quenched with 1 μL β-mercaptoethanol (Sigma Aldrich).

A total of 400 μg of silica beads were added to each sample (polydisperse silica microspheres 2.2 g/cc d50 = 4-8 um 400 ml, Cospheric) in a 5 mg/ml suspension in water with agitation. This was followed by acetonitrile addition at a volume equal to or greater than the total reduction-alkylation reaction volume. Precipitation was allowed to proceed for 5 minutes at room temperature with gentle agitation, then samples were centrifuged at 12,000 × g for 5 minutes. Overlays were removed and discarded, then samples washed with 80% ethanol (3 × 500 μL) at room temperature. Trypsin was added to each sample (1 μg trypsin, proteomics-grade; Trypsin Gold, Promega Corporation) in 50 mM Tris (pH = 8.5) containing 10% methanol, and samples shaken vigorously (1,400 rpm) at either 47°C for 2 hours, or at 37°C overnight.

Each sample was brought to 5% formic acid (Optima grade, Fisher Scientific) and agitated. Peptides were freed from buffers and trace contaminants using C18 cartridges prepared in-house using glass fiber filters as frits in 200 μL pipet tips, and 25 μm C18-coated silica particles (SiliaSphere PC, 90 Å pore, SiliCycle). After peptide loading, cartridges were washed with 5% acetonitrile containing 0.1% formic acid, then eluted using 100 μL of 40% acetonitrile containing 0.1% formic acid directly into MS sample tubes. They were taken to dryness in a centrifugal vacuum concentrator and each re-dissolved in 5% acetonitrile containing 5% formic acid.

2.5.2. Mass spectrometric analysis.

Peptide separations were performed on an Eksigent Ekspert 425 nanoLC system (Sciex, Framingham, MA) The separation column was fabricated in-house (ReproSil-Pur C18-AQ, 5 μm particle, 120 Å pore, Dr. Maisch GmbH, Ammerbuch, Germany; Column dimensions 50 μm I.D. × 15 cm length). Gradients were performed with Burdick & Jackson LC-MS-grade solvents (Honeywell, Charlotte, NC) using Optima-grade formic acid (Fisher Chemical, Pittsburgh, PA). Channel A pumped 0.1% formic acid, while channel B ran 0.1% formic acid in acetonitrile. The separation column was equilibrated at 45°C with 5% B at 350 nL/min. In all cases, approximately 1 μg of sample was loaded via the autosampler fitted with a 10 μL sample loop. After 10 minutes loading time, the loop was taken out of the flow path and a linear gradient to 30% B at 80 minutes was executed, followed by a gradient to 50% B at 110 minutes. This condition was held for two minutes before stripping the column at 92% B for 5 minutes. The system was then returned to starting conditions and equilibration performed for 10 minutes before the next sample was run.

Column effluent was directed into a Sciex 6600 Triple TOF mass spectrometer fitted with a nano source and silica emitter (360 um OD × 20 um ID × 7 cm length, CoAnn Technologies, Richland, WA). In all cases, the source voltage was kept at 2,400 V and the temperature at 150°C with a curtain gas flow rate of 23 L/min and a nebulizer nitrogen pressure of 3 psi.

LC-MS/MS data-dependent acquisition (DDA) experiments were performed in high-sensitivity mode utilizing a MS parent ion scan from 400-1250 amu with an accumulation time of 250 milliseconds (msec). Criteria required for candidate selection in the MS/MS analyses were a minimum target peak intensity of 350 counts per second (cps), a charge state of 2-4, and selection of 100 candidates per cycle. After MS/MS analysis, also in high-sensitivity mode, candidate parent ions (50 mDa mass tolerance) were excluded from subsequent selection and sequencing for 10 seconds. MS/MS spectra were acquired from 100-1250 amu using an accumulation time of 30 msec and rolling collision energies determined by parameters defined by the instrument manufacturer. These experiments were used to generate spectral libraries for subsequent quantitative analyses.

Peptide quantification was performed by data-independent analysis (DIA, SWATH) on the same instrumentation as was used for the DDAs. While chromatographic and mass spectrometer source conditions were identical to those used in DDA mode, SWATH analyses were started with a parent ion scan in high-sensitivity mode employing an accumulation time of 200 msec. Precursor mass windows of differing sizes were created using a SWATH variable window calculator available on the SCIEX software download website. Collision energies used were optimized for doubly-charged ions of the corresponding mass window and a 5 V collision energy spread (CES) was used in all cases. Accumulation time was set to 40 msec for each window. The PRIDE (PRoteomics IDEtifications Database) was used to upload and share raw data (Identifier: PXD067597).

2.5.3. Mass spectrometry data analysis.

Protein identification for ion library generation was determined for six representative DDA runs using the Paragon algorithm ¹⁹ in the ProteinPilot software package (Sciex, Version 5.0.2) searching a Uniprot reviewed human database with a <5% false discovery rate (FDR) at the peptide level. Quantification data was generated using the SWATH Acquisition MicroApp (version 2.0.1) within the Sciex PeakView software (version 2.2.0). A maximum of 6 peptides per protein were used for quantification with each analyte being characterized by a maximum of 6 transitions. Each peptide peak was extracted over a 10-minute window at a resolution of 75 parts per million (ppm). A peptide confidence threshold of at least 95% was required, as was a false discovery rate threshold of 5%. Relative protein expression levels from SWATH data and Principle Component Analysis (PCA) were determined using MarkerView (Sciex) and were subject to most-likely ratio (MLR) normalization. P-values for fold-change were calculated using Welch’s t-test in MarkerView.

2.5.4. Bioinformatics analysis for proteomics.

Raw counts were normalized by global median scaling to correct for systematic technical variation across samples. Limma p-values were calculated, along with false discovery rate (FDR) adjusted q values after a 20% set threshold. Data normalization, statistical analysis, and data exploration were performed using Omics Playground (BigOmics Analytics) ²⁰. Normalized data values were exported for downstream analysis. Log fold-change values were calculated between Naïve vs SCI-controls, as well as SCI vs PTEN-KO groups. Log fold-change values were assessed using Gene Set Enrichment Analysis^{21, 22} (GSEA) and evaluated for Gene-Ontology enrichment using the Biological Process pathways^{14, 15} (GO:BP), WikiPathways databases²³, as well as Drug Connectivity Map²⁴. GO enrichment pathway analysis was performed using R (v. 4.5.1) and significant identified pathways were simplified using the “simplify” function in the R package clusterProfiler, to remove highly similar and redundant pathways. Graphical displays corresponding to WikiPathways analyses and analysis using Drug Connectivity Profiling were obtained using Omics Playground²⁰. Data from peptides enriched within the top GO-annotations were extracted and matched to the peptide-level contrasts between groups and presented as fold-change or log fold-change for visualization. Peptide level contrasts are presented as both pre- and post-FDR corrected values using the Benjamini and Hochberg method²⁵, as well as differentially expressed (DE) peptides defined as a 2-fold increase relative to the respective control at an FDR correction value of 0.2.

Total proteomics DE analyses were performed within Omics Playground and volcano plots of log-fold change and uncorrected p-values were generated using GraphPad (v 11.0). Data visualization was performed using R (ggPlot2) to produce simplified GO:BP dot plot charts, while peptide-level visualization was performed in GraphPad.

2.6. Mitochondrial isolation and preparation.

Mitochondrial isolation from whole spinal cords was performed as previously described^26-28. In short, spinal cords were isolated without cardiac perfusions and homogenized using a Dounce homogenizer in mitochondrial isolation buffer including: 215 mM Mannitol, 75 mM Sucrose, 0.1% BSA, 1 mM ethyleneglycolbis(β-aminoethyl)-N,N,N',N'-tetraacetic acid (EGTA), and 20 mM 4-(2-hydroxyethyl)-1-piperazine-ethanesulfonic acid (HEPES) at pH 7.2. For the first experiment performed on chronic SCI mice with/without PTEN-KO, a crude total mitochondrial isolation was performed. For crude mitochondrial isolation, tissue lysate was centrifuged at low speed (1,300 × g for 3 minutes at 4°C) to pellet solid debris, followed by transfer of the mitochondrial-containing supernatant to a new tube and centrifugation at 13,000 × g for 10 minutes at 4°C to pellet mitochondria and other remaining soluble debris. Mitochondrial pellets were resuspended in cellular respiration buffer containing: 125 mM KCl, 0.1% bovine serum albumin (BSA), 20 mM HEPES, 2 mM MgCl2, and 2.5 mM KH2PO4, adjusted pH 7.2. Total protein per sample was evaluated using the BCA and a total of 6 μg of protein was added into each well of a 96-well Seahorse plate for analysis.

To assess neuron-specific mitochondrial fractions with or without PGC1α upregulation, we developed an AAVrg to express an HA-tagged eGFP that was fused to the 3’ end of the OMP25 protein to enable targeting to the outer mitochondrial membrane that was expressed under a CamKIIa promoter for neuron specificity. Gene sequences for the 3’ OMP25 mitochondrial targeting sequence were obtained from Addgene ID: 0326745^{29, 30}, and the plasmid and viral preparations were constructed at VectorBuilder. Mitochondrial isolation using a bead-pulldown method was optimized prior to collection from study tissue. In short, spinal cords were perfused using isolation buffer (215 mM Mannitol; 75 mM Sucrose; 900 nM EGTA; and 20 mM HEPES buffer). Isolated spinal cords were homogenized in mitochondrial isolation buffer and tissue lysate was incubated with 50 μL HA-tag magnetic beads (#130-091-122; Miltenyi Biotec) for 30 minutes on ice. Bead separation was performed by first passing the lysate through a 20 μm LS column pre-filter (130-101-812; Militenyi Biotec) before pushing the lysate through the magnetic column using LS columns (130-042-401; Militenyi Biotec). The columns were washed using a total of 4 mL of cellular respiration buffer before removing the columns and eluting the mitochondria-tagged magnetic beads. In total 2 SCI mice were pooled per sample to obtain sufficient mitochondrial isolation while singular spinal cords provided sufficient mitochondria from naïve controls. A total of 2 of the 5 pooled samples yielded insufficient mitochondria for testing in the same group (SCI-only) providing only n=3 replicates of 2 pooled samples in our neuron-specific chronic SCI experiments.

2.7. Seahorse assay for mitochondrial bioenergetic analysis.

In total 6 μg of isolated protein from crude mitochondrial extracts and from purified mitochondria pulldowns was used for bioenergetic testing on the Seahorse XF96 Flux analyzer (Agilent Technologies, Santa Clara, CA, USA) as previously published^{27, 31-34}. Starting 24 hours prior to testing a 96-well dual-analyte Seahorse sensor cartridge was hydrated at 37°C. All chemical reagents used to test mitochondrial bioenergetics were diluted in respiration buffer. Reagents used during the seahorse assay include a final concentration of: 5 mM pyruvate, 2.5 mM malate, and 2 mM adenosine diphosphate (ADP; used to generate State III respiration), 2.5 μM oligomycin A (used to generate State IV respiration), 4 μM carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; used to generate State V-complex I (V_CI) respiration), and 0.8 μM rotenone with 10 mM succinate (used to generate State V-complex II (V_CII) respiration).

The 6 μg of mitochondria was added to respiration buffer in each well to reach a final volume of 175 μL per well. The plate was centrifuged at 3900 rpm for 6 minutes at 4°C to adhere mitochondria to the bottom of each well. Oxygen consumption rate was assessed during each respiratory state driven by reagents described above. For our assessments State III respiration was engaged by providing pyruvate, malate, and ADP to activate oxidative phosphorylation and ATP production. State IV respiration was initiated using oligomycin to inhibit the ATP synthase which blocks the electron transport chain, and in turn, enables the assessment of baseline oxygen consumption rate and proton leak. State V_CI respiration was induced via mitochondrial uncoupling using FCCP to drive maximal electron transport chain activity. Respiratory State V_CII utilized rotenone to inhibit Complex I while providing succinate to activate and analyze Complex II mediated respiration.

2.8. Locomotor Functional Analysis.

The locomotor abilities of mice were evaluated in our final study after PGC1α upregulation. Locomotor abilities were evaluated using the Basso Mouse Scale of locomotor recovery³⁵ (BMS). For scoring, two raters who were blinded to group assignment performed the assessment. Mice were allowed to explore an unfilled open sandbox for up to 4 minutes during the assessment period. Raters discussed functional abilities until a score was obtained. In the event rater decisions were split on a score assignment, scoring to the deficit was employed as standard practice.

2.9. Immunohistochemistry.

After formaldehyde fixation, spinal cords and brains were acclimated to 30% sucrose for at least 1 week prior to freezing in cryomolds for sectioning. Spinal cord segments and brains were blocked at up to 15 spinal cords or brains per cryomold in M1-embedding matrix (#1310; FisherScientific) and cut at 20- or 30-μm in thickness, respectively, at −15°C. Sections were thaw-mounted and sets of every 7^th section were mounted on the same slide to provide a representative distribution across the spinal cord and brain length. For immunohistochemistry, mCherry (no antibody) or Neurofilament 200 kDa (Chicken anti-NFH; 1:2,000; NFH; Aves Labs) was used to visualize transduced neurons and axons. eGFP was present and tagged to mitochondria but the signal was amplified for better visualization using immunolabeling against HA-tag (Rabbit anti-HA; 1:2,000; #3724; Cell-Signaling Technologies; or Chicken anti-HA; 1;2,000; ET-HA100; Aves Labs).

Mitochondrial co-labeling between HA/eGFP⁺ and axons and neurons (NFH⁺ labeling) were imaged using confocal microscopy. Microscope settings included using a 100× objective and z-stacks of 0.4 μm step size through 10 total μm of tissue. Maximum intensity projection images were obtained for presentation and 3-D rendering was performed using Imaris software (v Imaris 11.0; Bitplane, Zurich, Switzerland). To our surprise, mCherry expression under the Syn1 promoter had been suppressed by 6-weeks post-treatment in PGC1α-treated mice but not mCherry-alone controls, similar to our prior observations using the Syn1 promoter to knockout PTEN^{6, 7}. The CamKIIa promoter remained stable and mitochondria within the cell bodies and axons were visible within the brain and spinal cord. However, due to sparse labeling of spared axons below the lesion, a loss of mCherry in PGC1α-treated mice, and an unquantifiable density of mitochondria within the neuronal somas, we were unable to reliably quantify differences in mitochondrial morphologies. For this manuscript, the microscopy images serve to demonstrate the neuron-specificity and targeting of spinal-projecting neurons in the brain.

2.10. Statistics.

Statistical approaches for bioinformatics analysis were provided above. For all other outcomes, between-group effects were analyzed using a Welch’s one-way ANOVA or two-way ANOVA with repeated measures for locomotor outcomes and further evaluated using the Benjamini and Hochberg (BH) approach to control false discoveries during pairwise comparisons.

3. 0. Results:

3.1. Total proteomics identify mitochondrial and synaptic dysfunctions as core pathophysiological mechanisms in chronic SCI that are reversed after PTEN-KO.

3.1.1. Mitochondrial dysfunction is a top dysregulated biological process sustained during chronic SCI and is improved after PTEN-KO.

To better understand the chronically injured spinal cord environment with or without PTEN-KO, we performed spinal cord contusions at the T9 vertebra in 3–5-month-old PTEN^fl/fl mice and knocked out PTEN from spinal-projecting neurons at the time of SCI. Spinal cords were harvested for protein 6-weeks after injury. Protein was obtained from uninjured naïve mice (n=6), mice receiving an SCI with a retrogradely transported AAV (AAVrg) to deliver an RFP as a vector control (n=5), along with mice receiving an SCI with an AAVrg to deliver cre-recombinase and an RFP to knockout PTEN (n=4). The Syn1 promoter was used to limit transgene expression to neurons. Prior to performing proteomics, the brains of mice were evaluated for evidence of successful AAV injections through the existence of RFP labeling in neurons in the motor cortex.

Total proteomics analysis was performed as an exploratory and hypothesis-generating approach to better understand the chronic SCI environment with/without neuronal-specific PTEN-KO. Our proteomics data was largely underpowered (n= 4-6/group) and too variable to detect DEGs surviving FDR corrections when comparing both Naïve vs SCI and SCI vs PTEN-KO. However, nominal p-value level analyses did demonstrate some significantly affected proteins when comparing Naïve vs SCI (16/3393 proteins), and SCI vs PTEN-KO (377/3396 proteins). Of all significantly affected proteins none were 2-fold changed in the Naïve vs SCI comparisons, while 16/377 exceeded 2-fold change in expression levels when comparing SCI vs PTEN-KO, of which all were increased in PTEN-KO relative to SCI (Fig. 1). Due to the low-powered analysis for detecting robustly upregulated proteins at the peptide level, we focused our analysis on broader biological processes.

Figure 1: Mitochondrial- and synaptic-associated pathways are reciprocally regulated between mice with chronic SCI and after treating SCI with PTEN-KO.

Total proteomics was performed on 12-20 week old PTEN^fl/fl mice to compare uninjured conditions to mice with chronic SCI (Naïve vs SCI), and mice with chronic SCI to mice treated with neuron-specific PTEN knockout after SCI (SCI vs PTEN-KO). Total protein was analyzed 6 weeks after SCI. (A) Volcano plots present nominal p values due to an absence of any singular protein identified as significant after FDR corrections. Heatmaps present the top up and down regulated proteins that were significantly affected at the level of nominal p values. (B) GSEA enrichment analysis of simplified GO:BP pathways reveal mitochondrial and synaptic-related pathways at the most enriched when comparing Naïve to chronic SCI, as well as SCI to PTEN-KO conditions. (C) GSEA Word Cloud analyses of the most abundantly identified terms across all gene and protein databases assessed within the Omics Playground identified Mitochondrial and Synaptic terms as associated with the most downregulated pathways when comparing Naïve to chronic SCI, with the same terms being the most upregulated when comparing SCI vs PTEN-KO conditions.

GSEA enrichment analysis provided a strong assessment of biological processes shifting within the total dataset. First, processes found to be most affected within the multi-database analysis used in Omics Playground were highlighted in a Word Cloud visualization of the data, providing a high-level assessment of the most-impacted biological processes. The terms “Mitochondria” and “Synapse” were the most strongly and negatively associated in the comparison of Naïve vs chronic SCI and were the most positively associated when comparing SCI vs PTEN-KO. Within the GO:BP pathway enrichment database, the most impacted biological processes identified in total proteomics were those related to mitochondrial and synaptic functions for both the Naïve vs SCI and SCI vs PTEN-KO comparisons after simplifying pathway enrichment terms (Fig. 1). We selected enrichment plots from GO:BP-Cellular Respiration to present representative shifts in mitochondrial and metabolic pathways within the total dataset (Fig. 2). Cellular Respiration-associated proteins were significantly enriched in both Naïve vs SCI (Normalized Enrichment Score (NES) = −2.60; q = 1.41 × 10⁻⁸) and SCI vs PTEN-KO (NES = 2.62: q = 7.07 × 10⁻⁹) comparisons. We isolated protein IDs within several GO annotations related to mitochondria and metabolic functions, removed duplicate IDs, and evaluated the directional regulation compared to respective controls for both comparisons. Nearly all mitochondria- and metabolic-related proteins were reduced after SCI compared to Naïve controls, and nearly all mitochondrial and metabolic related proteins were upregulated after neuron-specific PTEN-KO. Proteins that were significantly affected after PTEN-KO at the nominal p-value level were plotted, along with the fold-change of all mitochondria-related proteins that were obtained from pooling GO-annotations (Fig. 2).

Figure 2: Analysis of mitochondrial-associated biological pathways reveal a significant downregulation of mitochondrial-related proteins chronically after SCI that is reversed after PTEN-KO.

Total proteomics was performed on 12-20 week old PTEN^fl/fl mice to compare uninjured conditions to mice with chronic SCI (Naïve vs SCI), and mice with chronic SCI to mice treated with neuron-specific PTEN knockout after SCI (SCI vs PTEN-KO). Total protein was analyzed 6 weeks after SCI. GSEA enrichment was performed on ranked log-fold change values and assessed using GO:BP pathway analysis. (A) Enrichment plots from the GO:BP Cellular Respiration pathway was chosen to represent the global shift in mitochondrial related proteins affected by both chronic SCI and after PTEN-KO. Mitochondrial-related proteins experience a global downregulation chronically after SCI compared to Naïve controls, while PTEN-KO robustly restores the abundance of mitochondrial proteins. (B) The Log-Fold Change values of individual proteins detected using SWATH LC-MS/MS were compiled across several mitochondrial-related GO:BP pathways and plotted to visualize the distribution against the respective controls. (C) Proteins that were significantly affected by PTEN-KO at the nominal p value were plotted and compared for the directional effect and magnitude of effect between the Naïve vs SCI and SCI vs PTEN-KO comparisons. In most cases, the magnitude of downregulation occurring chronically after SCI was upregulated to a greater extent after PTEN-KO. (D) The WikiPathway Electron Transport Chain OXPHOS System in Mitochondria graphic was obtained from the Omics Playground which demonstrates the directionality of individual proteins affected in the Naïve vs SCI and SCI vs PTEN-KO comparisons. With few exceptions, most proteins downregulated chronically after SCI were reciprocally upregulated after PTEN-KO. (D) Blue represents a downregulation and Red represents an upregulation of the annotated proteins.

Collectively, bioinformatics analysis of total proteomics studies implicated mitochondria dysfunction as a novel pathology that is sustained chronically after SCI around the lesion. Further, while not hypothesized, neuron-specific PTEN-KO was predicted to reverse mitochondrial dysfunction at a protein level. Data for normalized and log-transformed counts, protein-level p values, and GO pathway enrichment statistics, are available in an Excel spreadsheet as Supplemental Materials 1.

3.1.2. Total proteomics predicts pre-synaptic functions as a novel mechanism mediated downstream of PTEN-KO.

Similar to mitochondria-related pathways, pre-synaptic-associated pathways emerged as another top-affected biological process that was downregulated in chronic SCI and similarly reversed with PTEN-KO. GO:BP-Trans-Synaptic Signaling enrichment analysis demonstrates a significant downregulation of synapse-related proteins when comparing Naïve vs SCI (NES = −2.07, q=8.91 ×10⁻⁹) with a subsequent upregulation after PTEN-KO (NES = 2.73, q = 4.69 × 10⁻⁹). Synapse-associated proteins that were significantly upregulated after PTEN-KO within the GO:BP-Trans-Synaptic Signaling pathway were isolated and plotted for visualization (Fig. 3).

Figure 3: Analysis of synaptic-associated biological pathways reveal a significant downregulation of synaptic-related proteins chronically after SCI that is reversed after PTEN-KO.

Total proteomics was performed on 12-20 week old PTEN^fl/fl mice to compare uninjured conditions to mice with chronic SCI (Naïve vs SCI), and mice with chronic SCI to mice treated with neuron-specific PTEN knockout after SCI (SCI vs PTEN-KO) log-fold change values and assessed using GO:BP pathway analysis. (A) Enrichment plots from the GO:BP trans-synaptic signaling pathway was chosen to represent the global shift in synaptic related proteins affected by both chronic SCI and after PTEN-KO. Synaptic-related proteins experience a global downregulation chronically after SCI compared to Naïve controls, while PTEN-KO robustly restores the abundance of synaptic proteins. (B) The Log-Fold Change values of individual proteins detected using SWATH LC-MS/MS were compiled across several synaptic-related GO:BP pathways and plotted to visualize the distribution against the respective controls. (C) Proteins that were significantly affected by PTEN-KO at the nominal p value were plotted and compared for the directional effect and magnitude of effect between the Naïve vs SCI and SCI vs PTEN-KO comparisons. In most cases, the magnitude of downregulation occurring chronically after SCI was upregulated to a greater extent after PTEN-KO. (D) The WikiPathway Synaptic Vesicle Pathway graphic was obtained from the Omics Playground which demonstrates the directionality of individual proteins affected in the Naïve vs SCI and SCI vs PTEN-KO comparisons within the pre-synaptic space. With few exceptions, most proteins downregulated chronically after SCI were reciprocally upregulated after PTEN-KO. (D) Blue represents a downregulation and Red represents an upregulation of the annotated proteins.

3.1.3. mTOR pathway inhibition is the best described molecular consequence of the chronic SCI environment.

Drug connectivity profiling is useful to compare how genetic or proteomic changes occurring between two conditions correlate with changes elicited by specific drug targets. In the context of comparing Naïve vs SCI, while no treatments were delivered, L1000/activity analysis identified the chronic SCI environment to be most strongly comparable to conditions elicited by mTOR-inhibiting compounds (NES = 2.28). mTOR inhibitors as well as other inhibitors associated with the mTOR pathway such as IGF-1- and PI3K-inhibitors were among the top comparable profiles that best describe the chronic SCI environment. Drug connectivity profiling data emphasizes the role of mTOR pathway dysregulation as a key molecular driver of pathology during chronic stages after injury. As anticipated, mTOR-pathway inhibitors were inversely correlated with effects after PTEN-KO (NES = −2.58), with the strongest positive correlations being comparable to protein-synthesis- and HDAC-inhibitors. Collectively, insights from drug connectivity profiling associate mTOR dysregulation as a top pathological descriptor of the chronic SCI environment, for which, neuron-specific PTEN-KO exerts a corrective restoration (Fig. 4).

Figure 4: The chronic SCI environment most strongly compares to conditions of mTOR-pathway inhibition.

Total proteomics was performed on 12-20 week old PTEN^fl/fl mice to compare uninjured conditions to mice with chronic SCI (Naïve vs SCI), and mice with chronic SCI to mice treated with neuron-specific PTEN knockout after SCI (SCI vs PTEN-KO). Total protein was analyzed 6 weeks after SCI. (A) GSEA assessment using the L1000/activity Drug Prediction Analysis of our total proteomics data within the Omics Playground most strongly compares the chronic SCI environment to conditions elicited by mTOR-pathway inhibitors. (B) As expected, PTEN-KO was least associated with mTOR-pathway inhibition. While no drugs were provided in the Naïve vs chronic SCI conditions, the qualitative nature of the L1000/activity analysis emphasizes the mTOR pathway as a key molecular pathology sustained during chronic SCI.

3.2. Mitochondrial bioenergetic abilities are chronically dysregulated after SCI and are partially reversed after neuron-specific PTEN-KO, validating proteomic assessments.

The in vivo experimental conditions used to obtain proteomics assessments were replicated to determine if mitochondria exhibit functional deficits chronically after SCI and/or improvements after PTEN-KO that were predicted from proteomics analyses. Mice were treated immediately after SCI with an AAVrg-Cre-p2a-dTomato or control vectors immediately after injury and allowed to survive for 6 weeks. Whole spinal cords were used for crude mitochondrial isolation at 6-weeks post-SCI. Significant main effects were observed for each respiratory state after Seahorse respiratory testing (State III F_{(2.00, 6.82)}= 12.37, p=0.005; State IV F_{(2.00, 7.97)}= 11.21, p=0.005; State V_CI F_{(2.00, 6.21)}= 17.97, p=0.005; State V_CII F_{(2.00, 6.23)}= 42.01, p=0.005).

SCI induced a significant reduction in mitochondrial respiratory abilities in every respiratory state (State III q=0.026; State IV q =0.01; State V_CI q =0.018; State V_CI q =0.003). Neuron-specific PTEN-KO applied at the time of SCI resulted in a significant improvement in respiratory abilities for State III and V_CI, but not State IV and V_CII (State III q=0.032; State IV q =0.296; State V_CI q =0.028; State V_CII q =0.056) (Fig. 5). Data from Seahorse respiratory testing supports a consistent conclusion derived from our proteomics work, specifically identifying a sustained mitochondrial pathology chronically after SCI that is reversed after PTEN-KO, with a large effect on Complex I-driven respiration. Raw data from bioenergetic testing is available in Supplemental Materials 1.

Figure 5: Isolating mitochondria from chronic SCI conditions validates a sustained respiratory dysfunction that is improved after PTEN-KO.

(A) Mitochondria were isolated from 12-20 week old PTEN^fl/fl mice to evaluate effects of SCI with or without PTEN-KO on bioenergetic abilities ex vivo. (B) Mitochondria were obtained 6-weeks after SCI from whole-spinal cord homogenates. Mitochondria obtained chronically after SCI display a significant reduction in State III and V_CI respiratory capacity compared to naïve controls that is improved after PTEN-KO. (C) Next, we sought to determine the effects of chronic SCI on respiratory capacity of mitochondria obtained specifically from neurons, and to determine if augmenting biogenesis using PGC1α expression would restore respiratory capacity. (C,E) Wild-type C57BL6/J mice received SCI or sham-injury controls at 10-12 weeks of age and were treated with AAVrg’s to express mCherry with or without PGC1α at 15-weeks post-injury. (E,F) Further, mice also received a spinal injection of AAVrg’s to label mitochondria with an HA-eGFP-mitoTag under the CamKIIa promoter to enable magnetic bead pulldowns of neuron-specific mitochondria. Mitochondria were isolated 6-weeks post-injection at 21-weeks post-SCI. (D) Western blotting was used to validate the upregulation of PGC1α and the mCherry reporter within the spinal cords of naïve mice 1.5-weeks after AAVrg injections. (G) Mitochondria isolated from neurons display an ~50% reduction in State III and V_CI respiratory capacity at 21-weeks post-SCI, while PGC1α expression delivered at 15-weeks and sustained for 6-weeks post-treatment restored Complex I-driven bioenergetic abilities. (H) Locomotor rating scores were significantly reduced after 90 kDyn contusive injuries and PGC1α overexpression using AAVrg’s did not restore motor abilities in near-complete chronic SCI. Error bars = S.E.M. *p < 0.05, **p < 0.01, ***p < 0.005. This figure was created in https://BioRender.com.

3.3. Neuronal mitochondria experience a ~50% reduction in respiratory capacity chronically after SCI that can be reversed with PGC1α upregulation.

To determine the extent to which neuron-specific mitochondria are affected during chronic SCI, we applied an AAVrg vector that expressed a HA-tagged eGFP that was fused at the C’-terminal end to the 3’-OMP25 outer mitochondrial membrane signaling peptide under the neuron-restricted CamK2a promoter. eGFP expression localizes to mitochondria in neurons throughout the brain and allows visualization of mitochondria in spinal-projecting axons in the white matter and neurons in the grey matter (Fig. 5). We utilized a modified magnetic bead pulldown approach to isolate purified neuronal mitochondria for ex vivo respiratory testing. Mice received a 90 kDyn SCI and at 15-weeks post-injury, received spinal injections of AAVrg’s that delivered our mitochondrial-targeting eGFP with or without a second AAVrg to delivery Syn1-PGC1α-p2a-mCherry or mCherry alone. Neuronal mitochondria were isolated for respiratory analysis at 21-weeks post-SCI.

At 21-weeks post-injury, main effects of Seahorse respiratory testing revealed a significant main effect for all respiratory states except State V_CII (State III F_{(2.00, 6.99)}= 6.93, p=0.02; State IV F_{(2.00, 7.743)}= 4.60, p=0.048; State V_CI F_{(2.00, 7.044)}= 6.12, p=0.028; State V_CI F_{(2.00, 5.595)}= 1.77, p=0.552). Neuronal mitochondria displayed a ~50% loss of State III and V_CI respiratory capacity compared to sham-injured controls (State III q=0.027; and State V_CI q=0.028). No differences were observed for respiratory State IV or State V_CII. Effects of PGC1α expression in neurons demonstrated an intermediate restorative effect on State III (State III q=0.08) and State V_CI (State V_CI q=0.135) respiration compared to mice with SCI, having restored the mean values approximately back to sham-injured controls but were not significantly elevated over mice with SCI. No differences were observed for State IV or V_CII conditions after PGC1α expression. While each replicate was comprised of n = 2 mice, unfortunately, we lost sufficient mitochondria from n = 2 pooled samples from the SCI-treatment group restricting our power and analysis to n = 3 for the SCI-only group (Fig. 5). Raw data from bioenergetic testing is available in Supplemental Materials 1.

3.4. AAVrg-mediated PGC1α expression did not restore locomotor functions in chronic near-complete SCI conditions.

While we assessed hindlimb functional abilities throughout this PGC1α experiment, the mean BMS score at the time of treatments was 1, representing an injury severity with few, if any, spared axons whereby improving mitochondrial abilities would be able to improve functional outcomes. There was no improvement of PGC1α upregulation on BMS scores when applied in near-complete chronic SCI (F_(1,16)=0.0253, p=0.87; Fig. 5). Raw data from BMS testing is available in Supplemental Materials 1.

4.0. Discussion:

Our work highlights several pathological discoveries that persist in chronic SCI that are mediated by PTEN signaling, the most important of which is a sustained dysregulation of mitochondria in neurons around the lesion. We demonstrate that both neuron-specific PTEN-KO and/or PGC1α upregulation can restore mitochondrial bioenergetic abilities. Ultimately, our work supports several non-regenerative mechanisms that can putatively elicit functional improvements after PTEN-KO in spared axon tracts that may provide novel therapeutic directions to safely treat chronic SCI, including mitochondrial and metabolic pathways.

Observing sustained mitochondrial dysfunction during chronic stages after SCI was unexpected. Our previous work found that injury-induced impairments to mitochondrial respiratory abilities arise acutely after SCI but return to pre-injury levels by 1-week post-injury³⁶. Only recently has a report demonstrated a sustained impairment for up to 2-weeks post-SCI³⁷. Here we demonstrate that not only is mitochondrial respiratory capacity impaired chronically after injury, but we have found that neuron-specific mitochondria experience a ~50% loss of respiratory capacity as far out as 21-weeks post-SCI. Ultimately, the cause and consequence of mitochondrial dysfunction in chronic SCI remains unknown, but both our proteomics data and respiratory testing support a critical role for PTEN signaling as an active player in mitochondrial pathology.

Prior work has supported that the mTOR pathway is chronically suppressed in neurons after SCI³⁸. Proteomics data presented here more than validates prior work, but qualifies mTOR pathway impairment as one of, if not the most, strongly impaired signaling pathway that is sustained after SCI. There is a clear a need to develop safe approaches for mTOR pathway manipulations. While the power of our proteomics analyses was too low to reveal strong evidence of DEGs between comparisons, bioinformatics assessments revealed both mitochondria and synaptic functions as dysregulated chronically after injury and reversed after PTEN-KO. Results provide credence to the significant role of PTEN in the chronic SCI pathology. Collectively, our data supports a vital role of PTEN as a key player in the chronic SCI pathology that maintains mTOR pathway dysfunction.

The mTOR pathway plays a major role in certain elements of metabolism ^{11, 39-41}, such as the mediation of translation¹³, inhibition of autophagy¹¹, and as a regulator of mitochondrial functions^{11, 40, 42}. However, PTEN itself also affects mitochondria through mTOR-independent pathways^{43, 44}. Specifically, PTEN-KO increases AMP-activated protein kinase (AMPK) activation which is a major regulator of cellular metabolism ^{45, 46}. AMPK upregulates PGC1α which is a transcriptional co-activator of many metabolic transcription factors that regulate most processes involved in mitochondrial biogenesis^{47, 48}. In our work we directly upregulated PGC1α during chronic stages after SCI to determine the potential for augmenting mitochondrial biogenesis in spinal-projecting neurons to improve hindlimb motor abilities. Unfortunately, determining the extent to which PGC1α upregulation can restore motor or sensory functions in chronic SCI has evaded our ability to conclude due to the severity of injury used in this work.

Proteomic analysis of PTEN knockout in both neurons and other cell types has previously been reported to affect mitochondrial protein expression. Similar to our work, mitochondrial related GO annotations, including “ATP synthesis Coupled Electron Transport”, “Cytochrome Complex Assembly”, and “Mitochondrial Translation” are amongst the top upregulated GO:BP annotations after PTEN-KO in primary neuronal cells⁴⁹. How exactly PTEN regulates mitochondrial related proteins is likely multi-factorial and includes transcriptional independent, and dependent mechanisms as just described. For example, Goo and colleagues (2012) observed that the activating role of the mTOR pathway on 4E-BP1, a translation initiating factor, is sufficient to drive an increase in the production of mitochondrial-related proteins independent of transcriptional responses in mouse embryonic fibroblasts⁵⁰. Further, effects of PTEN-KO on mitochondrial proteins via 4E-BP1 increases mitochondrial respiratory capacity in embryonic fibroblasts similar to our observations within the spinal cord. However, PTEN-KO may also drive increased transcriptional responses that mediate long-term metabolic reprogramming as well. Increasing AKT activity via PTEN-KO activates the transcription factor CREB which regulates the expression of PGC1α, that in turn modulates the transcription of diverse metabolic and mitochondrial related proteins⁵¹.

Mitochondrial dysregulation in other models of neurological disorders result in several well-defined pathologies including neurodegeneration^52-54, decreased synaptic functions⁵², as well as neuropathic pain^55-57, all of which are characteristic of chronic SCI. Indeed, use of metformin has gathered interest as a prospective analgesic to treat diabetic neuropathy as well other neuropathic pain-related conditions^58-60. Metformin activates AMPK, resulting in PGC1α upregulation and the stimulation of mitochondrial biogenesis⁴³. The emerging literature for the role of metformin and mitochondrial-targeting therapeutics to act as analgesics against neuropathic pain, along with our observation of sustained mitochondrial dysfunction after SCI, supports a potential alternate indication for using metformin to treat pain in chronic SCI and should be investigated. In support for the role of mitochondria and metabolic interventions to treat chronic SCI pain, prior literature has found that activating PPARγ using pioglitazone ameliorates pain in chronic SCI mouse models^{27, 61}. PGC1α is a necessary transcriptional co-activator of PPARγ. The prospects of targeting mitochondrial dysfunction during chronic SCI may prove to have strong and clinically viable prospects to treat pathologies arising from neuronal dysfunction within the spinal cord chronically after injury.

There are important factors to consider when interpreting the presented proteomics data. First, PTEN-KO drives activity through the mTOR pathway which inhibits autophagy, stimulates protein translation, as well as affects transcriptional responses. Increasing total protein may be a consequence of suppressing protein recycling along with stimulating translational processes, not necessarily through transcriptional modulations. Increasing total protein in a neuron-specific manner may have increased the representation of neuronal protein in the whole-cord homogenates, which may lead to misleading interpretations of some observations. For example, our proteomics findings would suggest that PTEN-KO induced a strong anti-inflammatory effect, as several of the downregulated proteins after treatment were inflammatory in nature. Concluding an anti-inflammatory effect is likely incorrect as our prior work that utilized PTEN-KO in chronic SCI did not observe a decrease in inflammatory cells after treatment⁶. Because many downregulated proteins observed in our proteomics dataset are non-neuronal in nature, they may be observed as downregulated simply due to a disproportional increase of neuron-specific proteins. For these reasons, we focused our conclusions more heavily on findings that presented an upregulation after PTEN-KO.

4.1. Conclusions.

We sought to identify non-regenerative mechanisms that may mediate improvements in locomotor functions when PTEN is knocked out of spared axon tracts after SCI. Proteomics data highlighted the extent to which dysregulation of the mTOR pathway plays a key role in the pathophysiology of chronic SCI, and the extent to which PTEN itself plays a vital role in mediating the major ongoing perturbations to cell signaling chronically after injury. We identified that the most impacted biological process that sustains dysregulated during chronic SCI is mitochondrial function, and that PTEN-KO almost completely reverses the dysregulation at a proteomic level. Ex vivo mitochondrial respiratory testing validated a robust and significant decrease in respiratory capacity chronically after SCI that was partially reversed after neuron-specific PTEN-KO, suggesting that the extent of mitochondrial dysfunction may extend to both neuronal and non-neuronal cells. After developing an approach to isolate neuron-specific mitochondria, we observed a ~50% loss in respiratory capacity that was sustained up to 21-weeks post-SCI around chronic lesions. Finally, we validated that the suppressed respiratory abilities can be rescued by upregulating neuronal PGC1α during chronic stages after SCI. Our findings have identified sustained mitochondrial dysfunction as a core pathophysiology during chronic SCI, for which the potential implications may be significant and multi-modal across several cellular and behavioral functions.

Significance Statement:

Chronic spinal cord injury (SCI) is hallmarked by sustained motor and sensory dysfunction with little potential for repair. The chronic SCI environment limits the excitability of spared neural circuits and significantly reduces the regenerative potential of exogenously applied therapeutics. Through a series of experiments, we have derived a novel and significant observation that neuronal mitochondria exhibit a ~50% loss of respiratory abilities chronically after SCI in mice. Moreover, by knocking out PTEN, a protein known to be chronically hyperactive after SCI, we demonstrate the ability to restore mitochondrial respiratory abilities. Our discoveries highlight a novel and vital pathological mechanism that is sustained chronically after SCI that is mediated by neuronal PTEN activity.

Highlights:

1. Sustained mTOR pathway dysfunction is a defining pathology of chronic SCI.

2. Mitochondria obtained from neurons respire at 50% efficiency in chronic SCI.

3. Neuron-specific PTEN-knockout restores mitochondrial respiratory capacity.

Abbreviations:

AAV

adeno-associated virus

AAVrg

retrograde psuedotyped adeno associated virus

ADP

adenosine diphosphate

AMP

adenosine monophosphate

AMPK

AMP-activated protein kinase

ANOVA

analysis of variance

ATP

adenosine triphosphate

BCA

bicinchoninic acid protein estimation assay

BSA

Bovine Serum Albumin

BH

Benjamini Hochberg method

BP

Biological Process

BMS

Basso Mouse Scale

CNS

central nervous system

DEG

Differentially Expressed Gene

eGFP

enhanced green fluorescence protein

EGTA

Ethylene glycol-bis(-aminoethyl ether)-tetraacetic acid

FCCP

carbonyl cyanide-p-trifluoromethoxyphenylhydrazone

FDR

False Discovery Rate

GO

Gene Ontology

GSEA

Gene Set Enrichment Analysis

HEPES

4-(2-hydroxyethyl)-1-piperazine-ethanesulfonic acid

KO

genetic knockout

mTOR

mammalian target of rapamycin

NES

Normalized Enrichment Score

NFH

Neurofilament Heavy 200kd

PBS

phosphate buffered saline

PEG

Poly-ethylene Glycol

PGC1α

peroxisome proliferator-activated receptor-γ coactivator 1-α

PTEN

phosphatase and tensin homolog protein

SCI

spinal cord injury

Syn1

synapsin 1 promoter

Data Sharing:

Raw proteomics data has been uploaded to the ProteomeXchange database under the title “Therapeutic effects of neuronal PTEN-knockout reveal a critical role of mitochondrial dysfunction in sustaining locomotor deficits chronically after spinal cord injury” with the dataset identifier PXD067597. Of note, raw data for a phosphoproteomics dataset obtained from the same experimental samples in this work is also available at ProteomeXchange but has not yet been analyzed nor reported under this publication. Normalized values, log-transformed values, Limma p-statistics, and bioinformatics databases evaluated using GSEA analyses are provided as supplemental materials 1. Data for mitochondrial seahorse testing and BMS data is provided in Excel format in supplemental materials 1.