Inefficient Thermogenic Mitochondrial Respiration Due to Futile Proton Leak in a Mouse Model of Fragile X Syndrome

Abstract

Fragile X syndrome (FXS) is the leading known inherited intellectual disability and the most common genetic cause of autism. The full mutation results in transcriptional silencing of the Fmr1 gene and loss of fragile X mental retardation protein (FMRP) expression. Defects in neuroenergetic capacity are known to cause a variety of neurodevelopmental disorders. Thus, we explored the integrity of forebrain mitochondria in Fmr1 knockout mice during the peak of synaptogenesis. We found inefficient thermogenic respiration due to futile proton leak in Fmr1 KO mitochondria caused by coenzyme Q (CoQ) deficiency and an open cyclosporine-sensitive channel. Repletion of mitochondrial CoQ within the Fmr1 KO forebrain closed the channel, blocked pathological proton leak, restored rates of protein synthesis during synaptogenesis, and normalized key phenotypic features later in life. The findings demonstrate that FMRP deficiency results in inefficient oxidative phosphorylation during neurodevelopment and suggest that dysfunctional mitochondria may contribute to the FXS phenotype.

Introduction

Fragile X syndrome (FXS) is the leading known inherited intellectual disability and the most common genetic cause of autism (1). The full mutation is caused by a CGG triplet repeat expansion in the 5’ untranslated region of the Fmr1 gene, resulting in hypermethylation and Fmr1 transcriptional silencing (2). Consequently, expression of the Fmr1 gene product, fragile X mental retardation protein (FMRP), is lost. FMRP is an RNA binding protein thought to regulate translation of numerous mRNAs important for synapse development (3). In the last several years, enhanced knowledge of the biological complexities of FMRP has helped to elucidate a number of potential therapeutic targets within a variety of affected signaling pathways. However, the exact mechanisms that disrupt synapse maturation, impair cognition, and result in behavioral abnormalities in FMRP deficiency remain unknown.

The mitochondrion is a dynamic organelle that is primarily responsible for generating cellular energy via oxidative phosphorylation in vertebrate eukaryotic cells. Neurons within the developing brain rely predominantly on this bioenergetic process to meet the substantial metabolic demands and energetic costs of synapse formation, protein synthesis, and dendritic arborization during synaptogenesis (4-6). Prior to maturation, neural progenitor cells utilize aerobic glycolysis for adenosine triphosphate (ATP) generation, however, undergo a metabolic switch to aerobic respiration during neuronal differentiation, capitalizing on the greater efficiency of oxidative phosphorylation (7). Emerging evidence suggests that mitochondria may play a role in such metabolic reprogramming via regulation of the mitochondrial permeability transition pore (mPTP) and superoxide signaling (8). Because the immature brain is uniquely vulnerable during critical windows of maturation, impairments in mitochondrial efficiency during such periods have the potential to disrupt neurodevelopment.

Defects in neuroenergetic capacity are known to cause a variety of neurodevelopmental disorders (9, 10). Prior work described alterations in metabolism in Drosophila lacking FMRP and in mice with mutations of both Fmr1 and its paralog, Fxr2 (11, 12). Specifically, FXS mutants demonstrated increased electron transport chain (ETC) capacity and oxygen consumption (11, 12). Recent investigation also reported aberrant mitochondrial bioenergetics in the Fmr1 KO mouse cerebral cortex characterized by increased ETC activity and inefficient respiration (13). However, the effect of FMRP deficiency on oxidative phosphorylation has not been thoroughly assessed and the role of mitochondria in the manifestation of the FXS phenotype is unknown. Thus, we aimed to systematically interrogate the integrity of mitochondria in the forebrain of Fmr1 knockout mice during a critical period of neurodevelopment. Because of the importance of oxidative phosphorylation to neuron maturation, synaptogenesis, and synaptic protein translation, we hypothesized that Fmr1 mutant mitochondria would demonstrate discrete defects. Using a top-down approach, we found that Fmr1 KO mitochondria exhibited inefficient thermogenic respiration due to futile proton cycling and leak. This defect was caused by coenzyme Q (CoQ) deficiency and an open cyclosporine A (CsA)-sensitive channel. Repletion of mitochondrial CoQ within the Fmr1 KO forebrain closed the channel, blocked pathological proton leak, restored rates of protein synthesis during synaptogenesis, and normalized key phenotypic features later in life. The findings demonstrate that FMRP deficiency results in inefficient oxidative phosphorylation during the peak of synaptogenesis and suggest that dysfunctional mitochondria may contribute to the FXS phenotype.

Materials and Methods

Animals

The care of the animals in this study was in accordance with NIH and Columbia University Medical Center Institutional Animal Care and Use Committee guidelines and conformed to the provisions of the Animal Welfare Act (NIH/DHHS) and the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Fmr1 KO (FVB.129P2-Pde6b⁺ Tyr^c-ch Fmr1^tm1Cgr/J) and control (FVB.129P2-Pde6b⁺ Tyr^c-ch/AntJ) mice were acquired (Jackson Laboratory, Bar Harbor, ME) and bred to yield fully affected Fmr1 KO newborn male pups and pups that expressed wild-type FMRP. Assessments were performed on 10 day old male pups or 6-8 week old males (phenotyping). For FMRP expression, wild-type male and female pups were evaluated on P4, P7, P10, and P14. For protein synthesis, wild-type and Fmr1^−/Y mice were assessed on P10 or 8 weeks of age.

Coenzyme Q (CoQ) injection

For in vivo CoQ repletion, littermates were randomly injected (ip) on P9 with 160 mg/kg CoQ₁₀ (in intralipid) or equal volume vehicle (intralipid alone). CoQ₁₀ (Sigma-Aldrich C9538) was dissolved in 100% ethanol and heated at 65°C for 5 minutes. Dissolved CoQ₁₀ was then added to intralipid (Sigma-Aldrich I141) (2.9 mM) and heated for 5 minutes at 65°C (14). For dose-response testing, mouse pups were injected (ip) on P9 with 20, 40, 80, or 160 mg/kg CoQ₁₀ (in intralipid) and assessed on P10. Equal volume vehicle (intralipid alone) served as a control injectate.

Isolation of synaptic and non-synaptic forebrain mitochondria

Mitochondria were isolated as previously described (15). Forebrain was harvested and homogenized in ice-cold isolation buffer (225 mM mannitol, 75 mM sucrose, 1 mM EGTA, 5 mM HEPES-KOH (pH 7.2) and 1 mg/mL of fatty-acid-free bovine serum albumin (BSA)). Because FMRP is expressed in all cell types, within all regions of the developing brain, mitochondria were isolated from the entire forebrain (16). This approach resulted in adequate mitochondrial yield per animal to permit systematic assessment of mitochondrial integrity using high fidelity assays. The homogenate was spun at 1100 g for 5 min at 4 °C. Supernatant (0.5 mL) was removed and mixed with 0.07 mL 80 vol% Percoll solution and then layered on 0.7 mL 10% Percoll solution and centrifuged at 18,500 g for 10 min at 4 °C. The mitochondria fraction was collected and resuspended in 0.7 mL of washing buffer (250 mM sucrose, 5 mM HEPES-KOH (pH 7.2), 0.1 mM EGTA and 1 mg/mL of BSA). The suspension was centrifuged at 10,000 g for 5 min at 4 °C. The mitochondrial pellet was resuspended in 0.07 mL of washing buffer and mitochondrial protein concentrations were determined using the method of Lowry.

Western blotting and immunostaining

10μg samples of mitochondrial protein or forebrain homogenate were subjected to SDS-acrylamide gel electrophoresis and immunoblotting. Mitochondrial protein loading was assessed with a primary monoclonal antibody to mouse VDAC (Abcam ab15895) and cytosolic protein loading was assessed with a primary monoclonal antibody to mouse actin (ThermoFisher Scientific MA5-15739). Signal was detected with enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech, Piscataway, New Jersey, USA), and density was measured using scanning densitometry. For immunohistostaining, the brain was perfused with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) via left ventricle injection for 30 min and then post-fixed in additional fixative solution for 24 h at 4°C. 6-μm sections were immunostained with anti-FMRP antibody (Abcam ab17722), biotinylated secondary antibody, and developed with DAB.

Body weight and temperature

Mouse pup body weight was measured using an analytical balance (Mettler Toledo PB303-S, Columbus, OH). Rectal thermometry (Thermalert TH-5, Physitemp Instruments, Clifton, NJ) was utilized to determine body temperature. Temperature measurements were performed between 9:00 am and 11:00 am in a room maintained at approximately 20°C. All measurements were obtained in awake, hand-restrained mice.

Blood glucose and lactate

65 μL of blood was collected from each mouse pup and glucose and lactate levels were immediately determined (GEM Premier 4000, Instrumentation Laboratory, Bedford, MA).

Forebrain lactate and pyruvate

Levels in forebrain homogenate (2.5-4.5 mg) were determined using colorimetric/fluorometric assay kits (Sigma MAK064-1KT, MAK071-1KT, St. Louis, MO) following the manufacturer’s directions. Molar lactate-to-pyruvate ratios were calculated.

ATP levels

ATP levels in forebrain homogenate (100-200 mg) and isolated forebrain mitochondria (167 μg) were determined separately using a colorimetric/fluorometric ATP assay kit (Abcam ab83355, Abcam, Cambridge, MA) following the manufacturer’s directions.

Heme content

Mitochondrial heme content was calculated from the difference in spectra (dithionate/ascorbate reduced minus ferricyanide oxidized) of forebrain mitochondria (3-5 mg) solubilized in 10% lauryl maltoside (17). The contents of heme aa₃, heme b, and heme c were calculated using absorption coefficients of 24 mM⁻¹cm⁻¹ at 605-630 nm, 20 mM⁻¹cm⁻¹ at 562-575 nm, and 19.1 mM⁻¹cm⁻¹ at 550-535 nm, respectively (18).

Mitochondrial oxygen consumption

Forebrain mitochondria (0.1 mg) were added to 0.5 mL of respiration buffer (200 mM sucrose, 25 mM KCl, 2 mM K2HPO4, 5 mM HEPES-KOH (pH 7.2), 5 mM MgCl2, 0.2 mg/mL BSA). Oxygen consumption was measured using a Clark-type electrode (Oxytherm, Hansatech, UK) with Complex I-dependent substrates (10 mM glutamate and 5 mM malate) or Complex II-dependent substrate (10 mM succinate in the presence of 5 μM rotenone) at 32°C. State 2 respiration was determined following the addition of substrates. ADP (200 μM) was then added to initiate state 3 respiration. The respiratory control ratio (RCR) was calculated as state 3 respiration divided by the observed rate following ADP consumption (state 4). The ADP-to-oxygen ratio (ADP:O) was calculated from the quantities of ADP and oxygen consumed during state 3 respiration. State 4 respiration was separately induced with oligomycin (2.5 μg/mL) (state 4o) and maximal rate of uncoupled state 3 respiration was induced with dinitrophenol (DNP) (70 μM) (state 3u). DNP:oligo ratios were then calculated.

Krebs cycle and electron transport chain (ETC) enzyme activity

Citrate synthase activity was determined in isolated mitochondria (2 μg) spectrophotometrically using a 96-well microplate reader (Medical Devices, Spectrmax m2, San Jose, CA) and a citrate synthase assay kit (Sigma-Aldrich CS0720). Following the manufacturer’s directions, the colorimetric rate of 5-thio-2-nitrobenzoic acid (TNB) formation was measured at 412 nm using 13.6 mM⁻¹ cm⁻¹ as the extinction coefficient of TNB.

Malate dehydrogenase activity was determined in isolated mitochondria (40 μg) spectrophotometrically by measuring the rate of oxidation of NADH at 340 nm. Assays were executed in a 1-mL reaction volume containing 100 mM PO₄⁻² (pH 7.4), 20 mM KCN, antimycin A (400 μg), 1 mM rotenone, Triton X-100, and 0.5 mM NADH. Oxaloacetate was added at a concentration of 0.5 mM to initiate the reaction. Specific activity was calculated using 6.2 mM⁻¹ cm⁻¹ as the extinction coefficient of NADH at 340 nm.

ETC enzyme complex activities were measured spectrophotometrically in a 1-mL reaction volume as previously described (19-21). Rotenone-sensitive Complex I specific activity was measured in isolated mitochondria (40 μg) using 4.8 mM⁻¹ cm⁻¹ as the extinction coefficient of NADH at 340 nm with the reference wavelength of 380 nm. For Complex II, 2-thenoyltrifluoroacetone (TTFA)-sensitive succinate-ubiquinone reductase (SQR) and succinate dehydrogenase (SDH) activities were determined separately in isolated mitochondria (40 μg) using 19.1 mM⁻¹ cm⁻¹ as the extinction coefficient of 2,6-dichlorophenolindophenol (DCPIP) at 600 nm for SQR and 17 mM⁻¹ cm⁻¹ as the extinction coefficient of 2-(4,5-dimethyl-2-thiazolyl)-3,5-diphenyl-2H-tetrazolium bromide (MTT) at 570 nm for SDH. For Complexes III and IV, inhibitor-sensitive first-order rate constants were calculated in isolated mitochondria (4 μg and 2 μg, respectively) using 18.5 mM⁻¹ cm⁻¹ as the extinction coefficient of cytochrome c at 550 nm. Oligomycin-sensitive Complex V specific activity was measured in isolated mitochondria (40 μg) using 6.2 mM⁻¹ cm⁻¹ as the extinction coefficient of NADH at 340 nm. Rotenone-sensitive Complex I+III linked activity and antimycin A-sensitive Complex II+III linked activity were determined separately in isolated mitochondria (40 μg) using 18.5 mM⁻¹ cm⁻¹ as the extinction coefficient of cytochrome c at 550 nm. Complex II+III activity was also measured following the addition of the ubiquinone analogue, decylubiquinone (0.1 mM in ethanol), the control quinone, MitoQ (mitoquinone mesylate, 0.1 mM in ethanol), or 0.1 mM ethanol. ETC enzyme complex specific activities were normalized to citrate synthase or malate dehydrogenase activity.

Coenzyme Q quantification

Coenzyme Q (CoQ) was extracted from forebrain homogenate (200 mg) with 1-propanol (22). The lipid extract was injected in a high-performance liquid chromatography with electrochemical detection by use of a reverse-phase column and isocratic mobile phase (22). Ubiquinone was extracted from isolated forebrain mitochondria or cardiomyocyte mitochondria (1.5-3 mg) using Tween-20 (3%), cold (−20°C) methanol (300 μL), and light petroleum (450 μL) (23, 24). Samples were centrifuged at 2200 g for 15 minutes at room temperature. Supernatant was collected, evaporated in a glass tube, and resuspended in 100% ethanol. Total ubiquinone was calculated in 400 μL from the difference in spectra (oxidized minus sodium borohydride reduced) using an absorption coefficient of 12.25 mM⁻¹ cm⁻¹ at 275 nm (24).

Modular kinetics

The three interconnected modules relative to the proton motive force (substrate oxidation, ATP turnover, proton leak) were measured independently as previously described (25). Oxygen consumption and mitochondrial membrane potential were measured simultaneously using forebrain mitochondria (0.2 mg) in 1-mL of respiration buffer (200 mM sucrose, 25 mM KCl, 2 mM K2HPO4, 5 mM HEPES-KOH (pH 7.2), 5 mM MgCl2, 0.2 mg/mL BSA) containing 80 ng/mL nigericin (to collapse ΔpH) and 5 μM rotenone at 37°C. Membrane potential was determined using an ion sensitive electrode selective for the lipophilic cation, tetraphenylphosphonium (TPP+) (World Precision Instruments, Sarasota, FL), and calculated using the Nernst equation as previously described (26). Mitochondrial respiration was initiated with 5 mM succinate. To measure substrate oxidation kinetics, state 4o was induced with oligomycin (2.5 μg/mL) and respiration was titrated with DNP (up to 100 μM). For ATP turnover, state 3 was induced using an ADP-regenerating system (100 μM ATP, 20 mM glucose and 10 units/mL hexokinase [baker’s yeast, Sigma-Aldrich]) and titrated with malonate (up to 200 μM) to permit measurement of kinetics.

For proton leak kinetics, state 4o was induced with oligomycin (2.5 μg/mL) and respiration was titrated with malonate (up to 2 mM). In separate experiments, voltage-gated proton leak was induced using titrations of malonate (0.1 mM) during state 4o respiration. Cyclosporine A (CsA, 1 μM), carboxyatractyloside (cAT, 1 μM), and guanosine diphosphate (GDP, 0.75 mM) were added to specifically inhibit the mPTP, ANT, and UCPs, respectively to determine source of leak (27, 28). The effect of coenzyme Q₁ (CoQ₁, 100 μM) was also assessed and the ubiquinone analog, decylubiquinone (DUb, 100 μM), was used to antagonize CoQ₁. In a subset of experiments, atractyloside (ATR, 200 μM) was added to open the mPTP during state 4o respiration in order to assess the inhibitory effect of CsA and CoQ₁ on the mPTP and succinate (5 mM) was added to determine substrate effect following malonate-induced voltage-gated proton leak.

Calcium loading capacity

Calcium uptake and release and oxygen consumption were measured simultaneously using forebrain mitochondria (0.2 mg) in 1-mL of respiration buffer (200 mM sucrose, 25 mM KCl, 2 mM K2HPO4, 5 mM HEPES-KOH (pH 7.2), 5 mM MgCl2, 0.2 mg/mL BSA) containing 5 μM rotenone and oligomycin (2.5 μg/mL) at 37°C (29). Calcium concentration was determined using an ion sensitive selective electrode (World Precision Instruments, Sarasota, FL) and calculated based on calibration using sequential additions of CaCl₂ ( 20 μM, 20 μM, 40 μM, 80 μM, and 100 μM) (29). Mitochondrial respiration was initiated with 10 mM succinate. In a subset of experiments, calcium loading capacity was determined in the presence of cyclosporine A (CsA, 1 μM) or coenzyme Q₁ (CoQ₁, 100 μM).

Intravital microscopy

Neurons within the somatosensory neocortex were imaged through a cranial window using a two-photon microscope (30, 31). Briefly, surgical anesthesia was achieved with 20 mg/kg ketamine (i.p.) and 3 mg/kg xylazine (i.p.). The skull was exposed via midline scalp incision and the somatosensory cortex identified based on stereotaxic coordinates (1 mm lateral from midline, 1.7 mm posterior to bregma). The head was immobilized, and a cranial window was created by removing a circular area of skull (~200 μm in diameter) with fine forceps.

Calcein-acetoxymethyl (AM) ester (Sigma-Aldrich) and tetramethylrhodamine ethyl ester (TMRE) (Sigma-Aldrich) were prepared in DMSO as 4 mM and 1 mM stock solutions, respectively. Cobalt chloride (CoCl₂) (Sigma-Aldrich) was prepared in artificial cerebral spinal fluid (ACSF [NaCl 129 mM, KCl 3 mM, NaH₂PO₄ 1.25 mM, MgSO₄ 1.8 mM, CaCl₂ 1.6 mM, NaHCO₃ 26 mM, and glucose 10 mM (pH 7.4)]). Calcein-AM (400 μM freshly diluted in ACSF) and TMRE (20 μM freshly diluted in ACSF) were applied directly to the cortical surface. The solutions were then washed liberally with ACSF following 1-hour incubation. Mice were placed under a two-photon microscope (Scientifica Hyperscope, East Sussex, UK) while still under anesthesia and temperature maintained with a heating pad (~37°C). The two-photon laser was tuned to 940 nm. All experiments were performed using a 1.05–numerical aperture (NA) 25× objective lens immersed in ACSF to obtain 100X high-magnification (104 μm × 104 μm; 512 × 512 pixels; 1-μm step) images for neuronal analysis. Neurons within layer II/III of the somatosensory neocortex (90-100 μm depth) were easily identified by their large round green fluorescent nuclei. CoCl₂ (2 mM) was then applied directly to the cortical surface to quench calcein fluorescence. Time-lapsed images were obtained at 5, 20, 30, and 60 minutes post-application of CoCl₂. Neurons in 3-4 imaged fields per mouse were identified in the aligned time series and region of interest boundaries were drawn around each neuronal soma (ImageJ). Arbitrary calcein and TMRE fluorescence were quantified (ImageJ) (range, 0-255, 8-bit) within each region of interest. TMRE fluorescence was expressed in arbitrary units relative to FVB control values and calcein fluorescence expressed as a ratio relative to baseline fluorescence.

Protein synthesis

A puromycin incorporation assay was used to measure global rates of protein synthesis (32). Mice were injected (ip) with puromycin (20 mg/kg, Sigma-Aldrich P8833) and forebrain was harvested 45 minutes later (32). Steady-state levels of newly synthesized puromycin-labeled peptides were quantified in forebrain homogenate via immunoblot analysis using a mouse primary monoclonal antibody to puromycin (Sigma-Aldrich MABE343). Protein loading was determined using a primary monoclonal antibody to mouse actin (ThermoFisher Scientific MA5-15739).

Dendritic spine density and morphology

Mouse forebrain was harvested and stained with Golgi-Cox solution (FD Rapid GolgiStain Kit PK401, FD NeuroTechnologies, Inc. Columbia, MD) following the manufacturer’s directions. Coronal sections (150 μm) were cut using a vibratome (Precisionary Instruments LLC, Greenville, NC), slide mounted, and images were obtained at 100X (Nikon A1 Confocal Microscope System, Nikon Instruments, Inc., Melville, NY). The primary somatosensory cortex (barrel cortex) was defined in accordance with atlases of the developing and adult mouse brain and 6-7 sequential sections were imaged for each animal (33, 34). Stack images were acquired (NIS Elements Confocal Microscope Imaging Software, Nikon Instruments, Inc., Melville, NY) and consisted of 150 optical sections, separated axially by 1 μm with a resolution of 0.1 μm/pixel. Images were analyzed (ImageJ, NIH) in a blinded manner and spines on apical dendrites of layer V, localized 50 to 100 μm from the soma of neurons, were counted and morphologically characterized as previously described (35).

Behavioral test battery

All behavioral tests were conducted within the Columbia University Mouse NeuroBehavior Core facility.

Elevated plus maze

Mice were tested for anxiety-like behavior as previously described (36). The elevated plus maze consisted of two open arms (30 X 5 cm) and two closed arms (30 X 5 X 15 cm) that extended from the central area (5 X 5 cm). The test was initiated by placing the mouse in the center region facing a closed arm. Mice were permitted to freely explore the maze for 5 minutes. Time spent in open and closed arms were quantified in a blinded manner.

Open field test

Hyperactivity and locomotor activity were tested as previously described (36). Exploratory locomotion was assessed in a blinded manner for 60 minutes in 5-minute bins using a VersaMax Animal Activity Monitoring System (AccuScan). Horizontal distance traveled was quantified by adjacent beam breaks in lower photocell panels and vertical activity was determined by beam breaks in the z upper photocell panels.

Acoustic startle threshold and pre-pulse inhibition of acoustic startle

Acoustic startle and pre-pulse inhibition of acoustic startle were quantified in a blinded manner using the SR-Laboratory System (San Diego Instruments) as described previously (36). Mice were placed in a Plexiglas holding cylinder for 5 minutes for acclimation. Baseline response to no stimulus was recorded. Startle responses to 40 ms sound bursts at 80, 90, 100, 110, or 120 dB were measured. Each trial was presented in a pseudorandom order and only presented once within each trial block. Intertrial interval was 10-20 seconds in duration. Startle amplitude was quantified every 1 ms over a 65 ms period.

For pre-pulse inhibition, a 20 ms pre-pulse stimulus tone at 74, 78, 82, 86, or 92 dB intensity was presented 100 ms prior to a 110 dB startle stimulus. Each trial was presented in a pseudorandom order and only presented once within each trial block. Intertrial interval was 10-20 seconds in duration. Startle amplitude was quantified every 1 ms over a 65 ms period.

Marble burying test

16 marbles were place equidistant (4 x 4) in a cage with fresh bedding. Mice were then placed in the cage for 30 minutes. At the end of the testing period, the number of marbles that were covered more than two-thirds with bedding were quantified by a blinded observer.

Repetitive behaviors

Mice were observed for 1-hour and the number of stereotypic behaviors (grooming, jumping, circling, digging, rearing, climbing), hyperlocomotion events, and resting counts were quantified as previously described (36). Sessions were videotaped and counts were quantified in a blinded manner.

Social interaction

The social interaction test was conducted using the Noldus PhenoTyper Observer 3000 chamber (25 X 25 X 35 cm) as previously described (36). An age and sex matched C57Bl/6 was placed in the chamber simultaneously with the test mouse. Social interactions (nose-nose, nose-genital, follow, front approach, push crawl) were videotaped for 10 min and quantified by a blinded observer.

Quantification and Statistical Analysis

Statistical analysis was performed using GraphPad Prism 7 software (GraphPad Software, La Jolla, CA). Data was assessed for normality by examining histograms and box plots and are presented in the figures as means ± SD unless otherwise specified in the figure legends. The sample number of mice (n) studied for each experiment is indicated for each figure. Differences between the two strains were assessed using two-tailed, unpaired Student t test. One-way ANOVA with Tukey’s post hoc test was utilized to compare 3 or more groups considering only one variable. Two-way ANOVA with Tukey’s post hoc test was used to assess differences between 3 or more groups with two variables. Significance was set at p < 0.05.

Results

The Metabolic Phenotype of Neonatal Fmr1 KO Mice Suggests Defective Mitochondria

FMRP is a ubiquitous protein, expressed predominantly within the brain and testis (37). Although FMRP is thought to be expressed within neurons throughout the life cycle, there is evidence of cell-specific temporal expression in various regions of the developing brain (16, 38). Thus, in order to identify the ideal time point for investigation, we assessed for changes in wild-type FMRP expression over time within the immature forebrain of FVB controls in the first two weeks of life (Figure S1). Using immunoblot analysis, we identified a relative peak in steady-state FMRP levels at 10 days of life in controls (Figure S1A). FMRP is known to be expressed in neurons, astrocytes, microglia, and oligodendrocyte precursors at this time point and we found widespread expression in all brain regions examined (16) (Figure S1B). Based on these findings, we performed our analyses on postnatal day 10 (P10) with a focus on the forebrain in male mice (given the male predominance of FXS).

We first attempted to define general features of the metabolic phenotype of Fmr1 KO mice and FVB controls on P10 to gather indirect evidence of mitochondrial dysfunction in mutants. Although there was no significant difference in body weight, blood glucose, or forebrain ATP levels between strains, Fmr1 KOs demonstrated significantly elevated rectal temperature, blood lactate levels, and forebrain lactate-to-pyruvate ratios compared to controls (Figure 1). Increased lactate and lactate-to-pyruvate ratios and low-normal levels of mitochondrial ATP suggested a defect in oxidative phosphorylation, necessitating aerobic glycolysis in Fmr1 KOs to maintain cellular ATP content while elevated body temperature implied an underlying thermogenic process given the relative locomotor inactivity and underdeveloped means of thermoregulation in newborn rodents (39, 40). These findings raised suspicion for uncoupled oxidative phosphorylation in Fmr1 KOs, reminiscent of the metabolic features of brown adipose tissue (BAT) mitochondria, which exhibit inefficient respiration to yield heat via non-shivering thermogenesis and rely on aerobic glycolysis to generate ATP (41).

Figure 1. Indirect Evidence of Mitochondrial Dysfunction in Fmr1 KO mice.

General features of the metabolic phenotype were assessed in 10-day old Fmr1 KO mice and FVB controls. Graphical representations of (A) body weight (n = 68 FVB controls, 69 Fmr1 KOs) (B) blood glucose (n = 7 FVB controls, 8 Fmr1 KOs) (C) rectal temperature (n = 4 per group) (D) blood lactate (n = 7 FVB controls, 8 Fmr1 KOs) and brain lactate-to-pyruvate ratio (n = 6 per group) (E) adenosine triphosphate (ATP) levels in forebrain and forebrain mitochondria (n = 4 per group) are depicted. Values are expressed as means ± SD. p values were calculated by Student’s t test. *p < 0.05, †p < 0.01.

Isolated Forebrain Fmr1 KO Mitochondria Demonstrate Alterations in Oxygen Consumption

We next measured oxygen consumption using a Clark-type electrode in freshly isolated forebrain mitochondria with the goal of identifying uncoupled respiration in Fmr1 KOs (Figure 2A, B). Although certain rates of Fmr1 KO oxygen consumption were altered, mitochondrial respiration appeared to be coupled without obvious evidence of increased proton leak (i.e., increased state 4 or oligomycin-induced state 4 respiration rates). Taken at face value, these findings seemed to suggest defective, but coupled, oxidative phosphorylation in Fmr1 mutant mitochondria. However, such an interpretation contradicts observations from Drosophila lacking FMRP and Fmr1/Fxr2 double KO mice, making this simple conclusion unlikely (11, 12). Importantly, without parallel measurement of mitochondrial membrane potential, precise and unambiguous interpretation of changes in rates of oxygen consumption is impossible (42).

Figure 2. Oxygen Consumption in Fmr1 KO Mitochondria.

Representative tracings for (A) Complex I-dependent oxygen consumption using glutamate/malate (n = 14 controls, 20 KOs) and (B) Complex II-dependent oxygen consumption using succinate (n = 11 per group) are depicted. Graphical representation of state 2 respiration (following addition of substrate), state 3 respiration (following addition of adenosine diphosphate (ADP)), state 4 respiration, uncoupled state 3 respiration (state 3u; following addition of dintrophenol (DNP)), and oligomycin-induced state 4 (state 4o) are shown below the respective tracings. Graphical depiction of adenosine diphosphate-to-oxygen ratios (ADP:O), respiratory control ratios (RCR), and DNP-to-oligomycin ratios (DNP:oligo) are also shown. The specific activities of (C) Krebs cycle enzymes and (D) normalized activities of the ETC complexes are depicted (n = 6 per group for Complex I and Complex II; n = 5 controls and 6 KOs for Complexes III, IV, and V). The succinate-ubiquinone reductase (SQR) and succinate dehydrogenase (SDH) activities of Complex II are shown individually. First order rate constants were determined for Complexes III and IV and expressed as turnover number (TN). Values are expressed as means ± SD. p values were calculated by Student’s t test. *p < 0.05, †p < 0.01, ‡ p < 0.001.

In order to gain further insight, we next measured the kinetic activity of each of the ETC enzyme complexes and two Krebs cycle enzymes (citrate synthase and malate dehydrogenase) within forebrain mitochondria (Figure 2C, D). The only divergence in enzyme activity between KOs and controls was seen in Complex II (the succinate dehydrogenase component) and Complex V (Figure 2D). In both cases, Fmr1 KO mitochondria demonstrated significantly increased kinetic activity compared to controls (Figure 2D). Although interesting, these abnormalities did not help to provide an obvious explanation for the changes in Fmr1 mutant oxygen consumption. Likewise, assessment of expression of several nuclear and mitochondrial encoded ETC complex subunits and related proteins, known mitochondrial targets of various microRNAs (miRNAs), and heme content within forebrain mitochondria yielded findings that were functionally insignificant (Figures S2, S3). Thus, differences in oxygen consumption in Fmr1 KO mitochondria were not on the basis of altered ETC complex protein expression.

Fmr1 KO Forebrain is Deficient in Coenzyme Q

Ubiquinone or coenzyme Q (CoQ) was the final component of the ETC to be assessed. CoQ is a lipophilic molecule that carries electrons from Complexes I and II to Complex III (43). We first measured the linked kinetic activities of Complex I+III and Complex II+III within forebrain mitochondria to indirectly quantify the CoQ pool. The activities of both Complex I+III and Complex II+III were significantly decreased in Fmr1 KO mice compared to controls, suggesting CoQ deficiency (Figure 3A). We next directly quantified CoQ content via HPLC and spectrophotometry and found significantly lower levels of forebrain CoQ₁₀ and mitochondrial ubiquinone in Fmr1 KOs versus controls (Figure 3B). In vitro addition of the CoQ analog, decylubiquinone (DUb), increased Complex II+III activity in both strains (Figure 3C). These findings were in concordance with data demonstrating that DUb increases linked Complex II+III activity in vitro by enhancing electron transport (44, 45). However, the fact that DUb-mediated Complex II+III activity in Fmr1 mutants did not quite reach or exceed control levels and there was no compensatory increase in Complex II SQR or Complex III activity (Figure 2) suggested that endogenous CoQ deficiency in Fmr1 KO mice was not rate limiting with regard to electron transport (45).

Figure 3. Coenzyme Q is Deficient in Fmr1 KO Forebrain and Mitochondria.

(A) The CoQ pool was assessed indirectly by measuring the linked kinetic activities of Complex I+III and Complex II+III within forebrain mitochondria. The specific enzyme activities were normalized to citrate synthase activity (n = 5 per group). (B) CoQ₉ and CoQ₁₀ were measured directly in forebrain via HPLC and total ubiquinone was measured directly within forebrain mitochondria via spectrophotometry (n = 5 per group). (C) In vitro Complex II+III activity response to an exogenous ubiquinone analogue. Normalized Complex II+III activities were determined in isolated mitochondria following the addition of the ubiquinone analog (Ub), decylubiquinone (DUb), the control quinone analog (MitoQ), or ethanol (EtOH) solvent alone (n = 5 per group). Values are expressed as means ± SD. p values were calculated between strain by Student’s t test and within strain using one-way ANOVA. *p < 0.05, ‡ p < 0.001. n.s. = nonsignificant.

Fmr1 KO Mitochondria Demonstrate Pathological Proton Leak

Although the slower Fmr1 KO mitochondria oxygen consumption rates we observed using a Clark-type electrode were in accordance with a recently published report, we rejected the interpretation that decreased respiration simply represented impaired electron transport (13). This is because such an inference contradicts prior findings from two different FXS models and would be in conflict with our data suggesting uncoupled respiration in Fmr1 mutant mice (11, 12). Because accurate and precise interpretation of changes in rates of oxygen consumption requires parallel measurement of mitochondrial membrane potential, use of in depth quantitative measures, such as modular kinetics, is often necessary to provide a more complete assessment of the functional state of mitochondria and permit identification of exact mechanisms of mitochondrial dysfunction (42).

Thus, we next assessed the three interconnected modules (substrate oxidation, ATP turnover, proton leak) relative to the proton motive force separately (Figure 4A) in freshly isolated forebrain mitochondria from Fmr1 mutants and controls. Although the substrate oxidation curve was somewhat flatter in Fmr1 KO mitochondria versus controls, there were no significant differences in oxygen consumption rates over the range of membrane potentials between strains (Figure 4B). This indicated that the ability of Fmr1 mutants to generate the proton motive force was not impaired and endogenous CoQ deficiency was not rate-limiting with regard to oxidation of ETC substrates. We next assessed the ATP turnover module and found that the curve in Fmr1 KOs was right-shifted relative to controls and the rate of mitochondrial respiration was significantly lower in mutants at the highest common membrane potential (Figure 4C). The data indicated that Fmr1 KO mitochondria were relatively hyperpolarized during state 3 respiration, with lower rates of ATP turnover compared to controls. Thus, hyperpolarization rendered Fmr1 KO mitochondria relatively more efficient in vitro, decreasing the demand for substrate oxidation to generate the proton motive force.

Figure 4. Modular Kinetic Analysis of Isolated Forebrain Mitochondria.

(A) Schematic of the three interconnected modules relative to the proton motive force (ΔΨ). (B-D) Rates of oxygen consumption using succinate as a substrate were measured over a range of mitochondrial membrane potentials. For substrate oxidation (B), state 4o was induced with oligomycin (data point with highest membrane potential) and respiration was titrated with serial additions of dintrophenol. For ATP turnover (C), state 3 respiration was induced using an ADP-regenerating system (data point with highest membrane potential) and titrated with serial additions of malonate. For proton leak (D), state 4o was induced with oligomycin (data point with highest membrane potential) and respiration was titrated with serial additions of malonate. n = 4 per group. Values are expressed as means ± SEM. p values for the rate of oxygen consumption at the highest common membrane potential were calculated by Student’s t test. *p < 0.05, †p < 0.01.

In separate experiments, we confirmed hyperpolarization during state 3 respiration in Fmr1 KOs and identified a possible role for the hydrolytic activity of the ATP synthase (reverse enzyme activity to generate the proton gradient). Evidence supporting this was seen in the increase in Complex V kinetic activity (Figure 2D), the failure to regain state 2 membrane potential levels in the presence of oligomycin (Figure S4A), and the decline in membrane potential following the addition of oligomycin in Fmr1 KOs (Figure S4B). Such a hyperpolarized state explains why state 3 rates of Fmr1 KO mitochondrial respiration were decreased in our initial experiments (Figure 2A, B) given that ATP turnover and substrate oxidation contribute equally to state 3 respiration (42). The findings also explain why respiration was decreased in Fmr1 KO cortical mitochondria in a recent report (13). Thus, decreased rates of Fmr1 KO state 3 seen with conventional polarography (Figures 2A, B) did not reflect impaired electron transport.

Finally, we evaluated the proton leak module. Although the starting point in oligomycin-induced state 4 (state 4o) respiration (~ 200 mV) was similar between groups, striking differences in oxygen consumption rates emerged as the membrane potential declined (Figure 4D). First, Fmr1 KO mitochondrial respiration increased significantly above state 4o rates with the first dosage of malonate, which was highly abnormal (Figure 4D). Second, the Fmr1 KO curve shifted to the left relative to controls, such that the rate of oxygen consumption at the highest common membrane potential was markedly and significantly increased in Fmr1 mutant mitochondria (Figure 4D). Thus, the proton leak module unmasked a voltage-gated leak in Fmr1 KO mitochondria (within the physiological range of mitochondrial membrane potentials) and indicated pathologically increased proton conductance in mutants reminiscent of the inefficient, futile proton cycling seen in BAT mitochondria (46). Therefore, modular kinetic analysis uncovered the presence of uncoupled respiration due to a pathological proton leak in Fmr1 mutant mitochondria. We confirmed the presence of this voltage-gated leak channel in Fmr1 KO mitochondria by simultaneously monitoring oxygen consumption and membrane potential during state 4o respiration over several minutes (Figure S4B). As the membrane potential declined following the addition of oligomycin, respiration increased; indicating a sudden increase in proton leak (Figure S4B). Importantly, the threshold for the excessive leak in Fmr1 KOs was well below the membrane potentials generated during the ATP turnover module (Figures 4C, D; Figure S4B). This explains why Fmr1 KO mitochondria were relatively more efficient in vitro during state 3 respiration.

A CsA-sensitive Channel is the Source of the Pathological Proton Leak in Fmr1 KO Mitochondria

Physiological proton leak within mitochondria can be constitutive or inducible and is thought to be mediated by the adenine nucleotide translocase (ANT) and uncoupling proteins (UCPs) (45). Uncoupled hydrogen ion conductance can also occur through other channels in the inner mitochondrial membrane such as the mPTP (47). In order to determine the etiology of proton leak in both groups, we first assessed for protein expression of the ANT and the various UCP isoforms and found no difference in steady-state levels within forebrain mitochondria between strains (Figure S5). Next, we utilized specific inhibitors of the ANT, UCPs, and the mPTP (carboxyatractyloside (cAT), guanosine diphosphate (GDP), cyclosporine A (CsA), respectively) to delineate the source(s) of proton leak in each group (Figure 5A) (27, 28). In separate experiments, we confirmed the ability of CsA to inhibit mPTP-mediated proton conductance in control mitochondria following opening of the pore with atractyloside (Figure S6A).

Figure 5. Identification of the mPTP as a Source of Proton Leak.

(A-D) Inhibition of proton leak in isolated forebrain mitochondria. Oligomycin-induced state 4 was initiated using succinate as a substrate. (A) Cyclosporine A (CsA), carboxyatractyloside (cAT), and guanosine diphosphate (GDP) were added to specifically inhibit proton leak via the mitochondrial permeability transition pore (mPTP), the adenine nucleotide translocase (ANT) and uncoupling proteins (UCPs), respectively (27, 28). (B-D) Representative tracings of oxygen consumption (above) with simultaneous mitochondrial membrane potential measurement (below) are depicted. (B and C) The voltage-gated proton leak was induced in Fmr1 KO mitochondria using malonate. Opening of the leak channel is evident by an increase in respiration with a concomitant decline in membrane potential. CsA, cAT, and GDP were added to determine the source of leak in (B) Fmr1 KO and (C) FVB mitochondria. Numbers and dashed lines indicate rates of oxygen consumption (nmol•mL⁻¹•min⁻¹•mg mitochondrial protein⁻¹). Green numbers are rates of oxygen consumption at comparable membrane potentials (green arrows) prior to and after addition of CsA or coenzyme Q₁ (CoQ₁). (D) The ubiquinone analog, decylubiquinone (DUb), was added to antagonize CoQ₁. Experiments were repeated in 3 different animals. (E-K) In vitro and in vivo assessment of the mPTP. (E) Calcium Loading Capacity. Calcium-induced calcium release from mitochondria was measured and quantified during state 4o succinate-dependent respiration. The amount of calcium required to induce opening of the mPTP was defined as the calcium loading capacity (29). (F) Graphical representation of calcium loading capacity. n = 3 per group. Values are expressed as means ± SD. p values were calculated by Student’s t test. *p < 0.05. (G) Representative tracings of calcium uptake and release are depicted. Numbers indicate calcium loading capacity (nmol•mg mitochondrial protein⁻¹). Tracings in the presence of mPTP inhibitors (CsA or CoQ₁) are also shown. (H) Calcein-Cobalt Method. Calcein diffuses into all cellular compartments and fluoresces bright green. Cobalt chloride (CoCl₂) quenches cytosolic and nuclear calcein fluorescence without affecting mitochondrial entrapped calcein due to the impermeable nature of the inner mitochondrial membrane. Opening of the mPTP permits cobalt to enter mitochondria and abolish intramitochondrial calcein fluorescence. (I-J) Intravital microscopy was performed to image the living murine cortex. Following 1-hour incubation with calcein-AM and tetramethylrhodamine ethyl ester (TMRE), green fluorescent neurons (calcein) and actively respiring mitochondria (TMRE) within layer II/III of the somatosensory neocortex were identified. (I) Representative images obtained at 100X magnification, 60 minutes after quenching with CoCl₂. Scale bar is 10 μm. Intramitochondrial calcein (green) and actively respiring mitochondria (red) are easily seen. Nuclei of various cell types within the neocortex can be seen as circular voids. Neurons were identified by their large circular nuclei and peri-nuclear mitochondria (arrowheads). Neuronal somata are outlined for clarity. (J) Time-lapsed imaging. Representative magnified serial images of a single neocortical neuron imaged at 100X are shown. Scale bar is 10 μm. An FVB neuron is depicted over time in the top row; an Fmr1 KO neuron is depicted in the bottom row. Somata of neurons are outlined for clarity. The large round nuclei and peri-nuclear mitochondria are easily seen. Calcein fluorescence was quenched with CoCl₂ and imaged over time (up to 60 minutes). Arrowheads in FVB control neurons indicate loss of calcein fluorescence over time within individual mitochondria. (K) Graphical quantification of neuronal TMRE fluorescence after 1-hour incubation is shown. FVB control values were arbitrarily set to 1. Values are expressed as means ± SEM. n = 3 biological replicates per strain, TMRE fluorescence in 5-8 neurons within 3-4 imaged fields per replicate. (L) Graphical quantification of neuronal calcein fluorescence over time is shown. Fluorescence at the 5-minute time point post CoCl₂ was established as baseline. Values are expressed as means ± SD. n = 3 biological replicates per strain, neuronal fluorescence in 3-4 imaged fields per replicate. p values were calculated using two-way ANOVA. *p < 0.05 vs. 5-minute time point within strain. †p < 0.001 vs. 5-minute time point within strain. ‡p < 0.001 vs. time-matched FVB control values and 5- and 20-minute time points within strain.

State 4o respiration was initiated in freshly isolated forebrain mitochondria and the voltage-gated proton leak was unmasked in Fmr1 KO mitochondria using malonate (Figure 5B). CsA caused a rapid repolarization of the membrane potential in Fmr1 mutant mitochondria with a concomitant slowing of the oxygen consumption rate (Figure 5B). Comparison of mitochondrial respiration at identical membrane potentials before and after addition of CsA demonstrated a slower rate of oxygen consumption post-CsA, indicating decreased proton conductance (Figure 5B). Both cAT and GDP decreased mitochondrial respiratory rate further, however, had minimal effect on the membrane potential (Figure 5B). In controls, CsA also decreased the rate of oxygen consumption, however, had little influence on mitochondrial membrane potential (Figure 5C). In contrast to Fmr1 KOs, the respiratory rate in control mitochondria at comparable membrane potentials prior to and after CsA was relatively unchanged (Figure 5C). Both cAT and GDP reduced respiration further in controls, however, cAT caused an immediate increase and stabilization of the membrane potential (Figure 5C). The data indicated that proton leak was mediated, in part, by a CsA-sensitive channel, ANT, and UCPs in both strains based on the decline in oxygen consumption following the addition of each inhibitor. However, the ANT was the predominant source of physiological leak in controls given the combined effects of cAT on oxygen consumption and membrane potential. On the other hand, a CsA-sensitive channel was the source of the pathological proton leak in Fmr1 KOs given that CsA caused complete repolarization of the mitochondrial membrane potential and reduced respiration to a rate below state 4o.

To evaluate mitochondrial integrity and the CsA-sensitive channel further, we next measured calcium loading capacity in Fmr1 KO and control forebrain mitochondria. Calcium-induced calcium release from mitochondria was measured and quantified during state 4o respiration to determine the threshold for mPTP opening (Figure 5E) (29). The amount of calcium required to induce opening of the mPTP was defined as the calcium loading capacity (29). Calcium loading capacity was significantly lower in Fmr1 KO mitochondria compared to controls, indicating earlier opening of the mPTP in Fmr1 mutants (Figure 5F, G). Importantly, this finding was not due to altered expression of the mitochondrial calcium uniporter (Figure S7). CsA increased the amount of calcium required to induce mPTP opening in both strains, however, the effect was more pronounced in Fmr1 KOs (Figure 5G). The results corroborated the proton leak findings and provided more evidence for pathological opening of the CsA-sensitive channel within Fmr1 KO mitochondria.

Next, we assessed for open probability of forebrain mPTP in vivo. To achieve this, we performed intravital microscopy in P10 mice to image calcein fluorescence within neurons of the somatosensory neocortex over time. Calcein readily gains access to all cytosolic compartments and becomes trapped within mitochondria (Figure 5H) (48). Cobalt chloride was used to visualize mitochondrial entrapped calcein because it quenches cytosolic and nuclear calcein fluorescence without affecting mitochondrial emission (Figure 5H) (48). However, pore opening permits cobalt to enter the mitochondrion and abolish intramitochondrial calcein fluorescence (Figure 5H) (48). Thus, cobalt-mediated quenching of intramitochondrial fluorescence indicates that the mPTP is open (48).

The exposed living cortex of Fmr1 KO mice and controls was incubated with calcein-AM and tetramethylrhodamine ethyl ester (TMRE) and imaged via multiphoton laser microscopy. After 1-hour, green fluorescent neurons with peri-nuclear mitochondria (red) within layer II/III of the somatosensory neocortex were easily identified (Figure 5J). Importantly, there was no significant difference in relative TMRE fluorescence between strains, indicating that Fmr1 KO neurons were capable of maintaining physiological mitochondrial membrane potential in vivo (Figure 5K). Calcein fluorescence was then quenched with cobalt chloride, imaged in a time-lapsed manner, and quantified. A gradual and steady decline in neuronal calcein fluorescence was observed over time in FVB controls, with a loss of ~30% fluorescence from baseline and evidence of quenching of fluorescence in specific mitochondria (Figure 5J, L). In Fmr1 KO mice, neuronal calcein fluorescence decreased markedly 30 minutes post-cobalt and remained significantly lower than time-matched control values (Figure 5J, L). In addition, there was evidence of overt calcein quenching within numerous mitochondria of Fmr1 KO neurons with a loss of approximately 70% fluorescence from baseline (Figure 5J, L). Markedly less punctate intramitochondrial calcein fluorescence was seen 60 minutes post-cobalt in Fmr1 KOs compared with controls due to quenching (Figure 5I). These results provided in vivo evidence of mPTP opening within neurons of the somatosensory neocortex of 10-day old mice in both strains. However, taken together, the data indicated that the open mPTP compromised mitochondrial integrity and was the source of pathological proton leak in Fmr1 KOs given the magnitude of cobalt-mediated quenching of intramitochondrial fluorescence, the threshold for calcium-induced calcium release, and the effect of CsA on calcium loading capacity and the mitochondrial membrane potential in the setting of excessive proton leak.

A Coenzyme Q Analogue Blocks Proton Leak in Fmr1 KO Mitochondria

Ubiquinone analogues have been shown to modulate the mPTP by inhibiting or inducing pore opening (49). Certain quinones have also been shown to antagonize the inhibitory effects of other ubiquinone analogues (49). To determine the activity of CoQ on newborn forebrain mPTP, we first assessed the in vitro effect of the ubiquinone analogue, CoQ₁, in control mitochondria following opening of the pore with atractyloside (Figure S6B). We utilized CoQ₁ and DUb in in vitro experiments because, unlike CoQ₁₀, their short side chains permit rapid diffusion across the mitochondrial outer membrane (50). Similar to CsA, CoQ₁ stabilized the mitochondrial membrane potential and slowed the rate of oxygen consumption (Figure S6B). Thus, CoQ₁ inhibited the mPTP in newborn forebrain mitochondria.

We next assessed the effect of CoQ₁ on the pathological proton leak in Fmr1 KO mitochondria. State 4o respiration was initiated in freshly isolated forebrain mitochondria and the voltage-gated proton leak was induced in Fmr1 KO mitochondria using malonate (Figure 5D). CoQ₁ caused an immediate increase and stabilization of the membrane potential (Figure 5D). Concomitantly, following an initial increase in oxygen consumption, there was a decline in the rate of respiration (Figure 5D). The initial CoQ₁-induced increase in respiration was likely due to substrate effect as was seen with the Complex II-dependent substrate, succinate, in separate experiments (Figure S6C). However, comparison of mitochondrial respiration at identical membrane potentials before and after the addition of CoQ₁ demonstrated a slower rate of oxygen consumption following CoQ₁, indicating decreased proton leak (Figure 5D). Administration of a CoQ₁-antagonizing quinone, DUb, caused an immediate decline in membrane potential and an increase in respiration, suggesting re-opening of the channel (Figure 5D). Subsequent administration of CsA resulted in repolarization of the membrane potential and decreased oxygen consumption, indicating closure of the pore once again (Figure 5D). The data indicated that a CoQ analogue was capable of blocking the pathological proton leak in Fmr1 KO mitochondria. To confirm quinone-mediated inhibition of the mPTP, we next assessed the effect of CoQ₁ on calcium loading capacity in Fmr1 KO mitochondria. Calcium-induced calcium release from mitochondria was quantified during state 4o respiration. CoQ₁ increased Fmr1 KO calcium loading capacity to levels that approached FVB control values, confirming quinone-mediated inhibition of the mPTP (Figure 5G).

Repletion of Forebrain Mitochondrial CoQ Blocks the Pathological Proton Leak, Restores Protein Synthesis, and Acutely Induces Morphologic Changes and Density of Dendritic Spines in Fmr1 KOs

Given the deficiency of endogenous ubiquinone in Fmr1 KO forebrain mitochondria and that a CoQ analogue blocked the CsA-sensitive channel in vitro, we next assessed the ability of exogenous CoQ₁₀ to close the proton leak in vivo. We utilized CoQ₁₀ in the in vivo experiments given the deficiency of this specific quinone species in the developing Fmr1 KO forebrain. First, we determined if it was possible to replete CoQ within Fmr1 KO forebrain mitochondria using various doses of CoQ₁₀. Both strains of mice were intraperitoneally (ip) injected with CoQ₁₀ or vehicle on P9 and assessed 24 hours later, on P10. A dose-response relationship between CoQ₁₀ and the linked kinetic activities of Complex I+III and Complex II+III within forebrain mitochondria was demonstrated in Fmr1 KOs and the highest CoQ₁₀ dosage restored ubiquinone levels to control values (Figure S8).

We next confirmed that exogenous CoQ₁₀ blocked the CsA-sensitive proton leak in forebrain Fmr1 KO mitochondria following in vivo injection. Fmr1 KOs and controls were injected with the highest dosage of CoQ₁₀ or vehicle on P9 and then evaluated on P10. To assess the effect of CoQ₁₀ on thermogenic uncoupling, we first measured rectal temperature and forebrain lactate-to-pyruvate ratios in injected animals. As with our initial findings, we found that vehicle-injected Fmr1 KO mice demonstrated significantly elevated rectal temperature and increased brain lactate-to-pyruvate ratios compared to vehicle-injected controls (Figure 6A, B). However, CoQ₁₀ injection normalized body temperature and brain lactate-to-pyruvate ratios to control values in Fmr1 mutant mice, suggesting improved mitochondrial coupling (Figure 6A, B).

Figure 6. In vivo Blockade of Pathological Proton Leak in Fmr1 KOs with Coenzyme Q₁₀.

Mice were injected (ip) with CoQ₁₀ or vehicle on P9 and assessed on (A-E) P10 or at (F-M) 6-8 weeks of age. (A) Rectal temperature. n = 6 animals per FVB cohort, 7 vehicle-injcted Fmr1 KOs, 11 CoQ₁₀-injected Fmr1 KOs. †p < 0.01. (B) Forebrain lactate-to-pyruvate ratio. n = 3 animals per cohort. *p < 0.05. (C) Proton leak kinetics. State 4o was initiated in isolated forebrain mitochondria using succinate as a substrate. n = 4 per group. The rates of oxygen consumption at the highest common membrane potential prior to and after the addition of malonate were compared. †p < 0.01 vs. vehicle-injected FVB controls, ‡p < 0.001 vs. all other cohorts, *p < 0.01 vs. vehicle-injected FVB controls and p < 0.001 vs. vehicle-injected Fmr1 KOs. (D) Rates of protein synthesis. Immunoblot of puromycin-labeled peptides in forebrain is depicted. Actin was used as a loading control. Graphical representation of relative densities is shown. Values are expressed as means ± SD. Vehicle-injected FVB control values were arbitrarily set to 1. n = 3 per group. *p < 0.05. (E) Dendritic spine density and morphology on P10. Representative images of high magnification Golgi-stained sections of primary somatosensory cortex are depicted. Scale bar = 5 μm. Graphical representation of spine density is shown to the right. Checkered pattern represents density of immature spines while solid color represents density of mature spines. n = 3 biological replicates per cohort, 6-7 neurons per replicate. ^p < 0.05 vs. vehicle-injected FVB controls. *p < 0.05. (F) Dendritic spine density in 6-8 week old mice. Representative images of high magnification Golgi-stained sections of primary somatosensory cortex are depicted. Scale bar = 5 μm. Graphical representation of spine density is shown to the right. n = 4 biological replicates per cohort, 6-7 neurons per replicate. *p < 0.05, ‡p < 0.001. (G-M) Battery of behavioral tests in 6-8 week old mice. n = 10 per FVB cohort, 12 per Fmr1 KO cohort. (G) Elevated plus maze. Percentage of time spent in open and closed arms is depicted. (H) Open field test. Horizontal distance traveled and number of vertical counts are shown. *p < 0.05 vs. time-matched vehicle-injected FVB control, †p < 0.01 vs. time-matched vehicle-injected FVB control, ‡p < 0.001 vs. time-matched vehicle-injected FVB control, ^p < 0.01 vs. time-matched vehicle-injected FVB control, #p < 0.01 vs. time-matched CoQ-injected Fmr1 KO, ^ap < 0.001 vs. time-matched vehicle-injected FVB control, p < 0.05 vs. time-matched CoQ-injected Fmr1 KO, ^bp < 0.05 vs. CoQ-injected cohorts. (I) Startle response. Response to various stimulus intensities is shown. *p < 0.05 vs. vehicle-injected FVB control. †p < 0.01 vs. CoQ-injected FVB control. (J) Pre-pulse inhibition. %Response to various pre-pulse inhibition intensities is depicted. *p < 0.05. (K) Marble burying. Representative images of buried marbles are shown. The number of marbles buried over a 30-min period were quantified. *p < 0.05, †p < 0.01. (L) Repetitive behaviors. Behaviors were observed over 1-hour. The number of stereotypic behaviors (grooming, jumping, circling, digging, rearing, climbing), hyperlocomotion events, and resting counts were quantified. †p < 0.01. (M) Social interaction. The total number of social interactions (nose-nose, nose-genital, follow, front approach, push crawl) were quantified. All values are expressed as means ± SEM unless otherwise indicated. All p values were calculated using two-way ANOVA.

Next, we assessed the effect of CoQ₁₀ or vehicle on mitochondrial proton leak 24 hours post-injection. To accomplish this, we performed modular kinetic analysis of the proton leak module in forebrain mitochondria isolated on P10. Mitochondria from vehicle-injected Fmr1 KO mice demonstrated a voltage-gated leak as evidenced by a left-shifted curve relative to vehicle-injected controls and significantly increased oxygen consumption rates at the highest common membrane potential following the addition of malonate (Figure 6C). CoQ₁₀ injection shifted the FVB curve slightly to the right compared to vehicle-injected controls, leading to significantly decreased rates of respiration at the highest common membrane potential shared by both FVB cohorts (Figure 6C). In Fmr1 KO mice, CoQ₁₀ injection shifted the curve markedly to the right, leading to significantly lower rates of oxygen consumption at the highest common membrane potentials prior to and after the addition of malonate (Figure 6C). The findings indicated that CoQ₁₀ inhibited the pathological thermogenic leak and futile proton cycling in Fmr1 KO mice in vivo, resulting in more efficient mitochondrial respiration.

Next, we explored the link between dysfunctional mitochondria and murine FXS phenotypic features. Characteristics that have been consistently described in the Fmr1 KO mouse include dysregulated forebrain protein synthesis, increased dendritic spine density and immature spine morphology, learning and memory deficits, increased locomotor activity, and heightened reactivity to sensory stimuli (49-53). To determine if uncoupled mitochondrial respiration contributes to the FXS phenotype we first assessed specific outcome measures immediately following in vivo inhibition of the pathological thermogenic proton leak. Thus, mice were injected with CoQ₁₀ or vehicle on P9 and assessed 24 hours later, on P10. However, little is known about the phenotypic features of mutants on P10. Furthermore, the only outcomes that can be reliably tested at this age include rates of protein synthesis and dendritic spine density and morphology. Therefore, we assessed each of these features, 24 hours post injection of CoQ₁₀ or vehicle.

Rates of protein synthesis were determined by measuring steady-state levels of newly synthesized puromycin-labeled peptides in forebrain via immunoblot analysis following puromycin injection. Vehicle-injected Fmr1 KO mice demonstrated significantly lower rates of protein synthesis compared to vehicle-injected controls (Figure 6D). This finding appeared to contrast the well-known paradigm seen in adult mice (54, 55). So, we next compared rates of protein synthesis between 10 day old mice and young adults (Figure S9). As expected, protein synthesis was aberrantly increased in 8 week old Fmr1 KO forebrain (Figure S9). However, protein synthesis was lower in P10 Fmr1 KOs compared to young adult Fmr1 KOs and age-matched controls (Figure S9). Consistent with developmental increases in protein synthesis, rates in P10 FVB forebrain were significantly greater than young adult controls (Figure S9) (56, 57). Thus, protein synthesis, although relatively excessive in the mature Fmr1 mutant forebrain, was relatively impaired in the developing Fmr1 KO forebrain. Importantly, CoQ₁₀ injection had little effect on rates of protein synthesis in FVB mice, however, significantly increased levels of puromycin-labeled peptides in Fmr1 KO forebrain on P10 (Figure 6D). Thus, blocking the pathological mitochondrial leak with CoQ₁₀ restored rates of protein synthesis in immature Fmr1 mutant mice (Figure 6D).

Next, we assessed dendritic spine density and morphology. The classic finding in adult Fmr1 KO mice is increased density and relative immature spine appearance (58). However, on P10, there is no difference in spine density or length of dendritic protrusions between mutant and wild-type mice (59). Consistent with this, we found no differences in spine density or morphology in the primary somatosensory cortex between vehicle-injected Fmr1 KO mice and controls (Figure 6E). CoQ₁₀ injection had little effect on the dendritic spines of FVB controls (Figure 6E). However, in Fmr1 KO mice, CoQ₁₀ significantly increased the total density of dendritic protrusions, the number of immature spines compared to vehicle-injected controls, and the number of mature spines versus vehicle-injected mutants (Figure 6E). These characteristics could be consistent with the dendritic spine densities and morphology seen in older, more developed wild-type mice (i.e., 3-4 weeks of age) (51, 59). Thus, inhibiting excessive proton leak with CoQ₁₀ in Fmr1 KO mice acutely induced changes in morphology and density of dendritic spines 24 hours post injection.

Blocking the Pathological Proton Leak During Synaptogenesis in Fmr1 KOs Rescues Key FXS Phenotypic Features in the Juvenile and Young Adult.

We next explored the link between excessive proton leak within developing forebrain mitochondria and key murine FXS phenotypic features later in life. Fmr1 KOs and controls were injected with the highest dosage of CoQ₁₀ or vehicle on P9 and then assessed at 6-8 weeks of age. First, we evaluated dendritic spine density in the primary somatosensory cortex in young adult mice. Consistent with the known histopathological nature of FXS, we found significantly increased spine density in vehicle-injected Fmr1 KO mice (Figure 6F). CoQ₁₀ injection had little effect on the dendritic spines of FVB controls, however, normalized the density of spines in Fmr1 KOs, suggesting a link between excessive proton leak within forebrain mitochondria and dendritic spine maintenance during brain development (Figure 6F).

Next, we assessed mouse behavior using a battery of tests. We first quantified anxiety-related behavior using the elevated plus maze and found no effect of strain or injectate on percent time spent in either the open or closed arms of the maze (Figure 6G). We then assessed for hyperactivity using the open field test. Consistent with the hyperactive FXS phenotype, vehicle-injected Fmr1 KO mice demonstrated significantly increased horizontal (distance traveled) and vertical activity versus vehicle-injected controls (Figure 6H) (60). In contrast, we found no significant difference between each CoQ-injected cohort and vehicle-injected FVB controls with regard to horizontal distance traveled and CoQ₁₀ injection significantly decreased vertical activity in Fmr1 KOs (Figure 6H). Thus, closing the pathological mitochondrial proton leak with CoQ₁₀ during synaptogenesis reduced Fmr1 KO hyperactivity later in life.

We next assessed stimulus sensitivity and sensorimotor gating via startle response and pre-pulse inhibition, respectively. Both Fmr1 KO cohorts demonstrated significantly reduced startle response at the highest stimulus intensities compared to injectate-matched controls (Figure 6I). Pre-pulse inhibition testing yielded inconsistent results, with both Fmr1 KO cohorts demonstrating increased responses at a single noise intensity only (Figure 6J). Thus, although there were abnormalities in startle response and pre-pulse inhibition in both Fmr1 KO cohorts, CoQ appeared to have no effect on stimulus sensitivity or sensorimotor gating. Next, we evaluated for repetitive behaviors. Compared to vehicle-injected FVB controls, vehicle-injected Fmr1 KO mice buried significantly more marbles, demonstrated significantly more stereotypical repetitive behaviors and locomotion, and spent significantly less time resting (Figure 6K, L). CoQ₁₀ injection normalized the number marbles buried and stereotypical repetitive behaviors, significantly decreased hyperlocomotor activity, and significantly increased resting time in Fmr1 KO mice (Figure 6K, L). Thus, inhibition of excessive proton leak in developing forebrain mitochondria improved the repetitive behavioral phenotype of young adult FXS mice. Finally, we evaluated social behavior, but found no significant differences in interaction based on strain or injectate (Figure 6M).

Discussion

In this study, we identified a discrete defect within Fmr1 KO murine forebrain mitochondria during the period of peak wild-type FMRP expression. Specifically, Fmr1 KO mitochondria demonstrated uncoupled thermogenic respiration due to futile proton cycling and pathological leak caused by CoQ deficiency. Blocking the excessive proton leak within Fmr1 KO forebrain mitochondria during synaptogenesis restored rates of protein synthesis within the developing forebrain, induced changes in morphology and density of dendritic spines acutely, normalized dendritic spine density later in life, and decreased hyperactivity and repetitive behavior in the young adult. This establishes a link between a pathologically open mPTP within the developing brain and certain key features of murine FXS. Thus, our findings reveal novel consequences of FMRP deficiency during the peak of synaptogenesis and suggest that dysfunctional mitochondria may contribute to the FXS phenotype.

Proton leak is a physiological process that permits oxidative phosphorylation to be carried out in an incompletely coupled manner, allowing certain cell types to generate heat, maintain carbon flux, and alter nutrient response in the setting of specific metabolic demands (46). In addition, proton leak modulates the mitochondrial membrane potential to prevent oxidative stress (46). Thus, leak is important for mitochondrial homeostasis and cellular health. Proton leak, however, can also be pathologic and can compromise mitochondrial efficiency. Here, we identified a large leak in Fmr1 KO forebrain mitochondria which uncoupled respiration and necessitated higher rates of oxygen consumption to defend the mitochondrial membrane potential. Evidence for the pathological nature of this mitochondrial defect was found in the relative hyperthermia, elevated lactate levels, and increased brain lactate-to-pyruvate ratios observed in Fmr1 KOs. Thus, Fmr1 mutants were relying, to some degree, on glycolysis to generate ATP in the setting of inefficient oxidative phosphorylation. Although we have not ruled out other tissue sources of thermogenic respiration, the proton leak kinetics seen in Fmr1 KO forebrain mitochondria were reminiscent of the characteristics of BAT mitochondria (46). However, unlike BAT mitochondria which have evolved to purposefully utilize proton leak to generate heat, neuronal mitochondria rely on well coupled oxidative phosphorylation to meet the bioenergetic demands of the developing brain (4, 5, 41). Thus, excessive mitochondrial proton leak in the immature Fmr1 mutant forebrain has the potential to disrupt neurodevelopment.

The voltage-gated nature of the proton leak in Fmr1 KO mice was informative and important. Here, conventional polarographic measurement failed to detect pathological leak. This is because Fmr1 KO mitochondria were hyperpolarized during state 3 and remained hyperpolarized during state 4 for a period of time. It is known that excess exogenous substrates hyperpolarize isolated mitochondria in the presence of oligomycin (42, 61). However, Fmr1 KO mitochondria appeared to be utilizing the reverse activity of the ATP synthase to pump hydrogen ions from the matrix into the intermembrane space to generate the hyperpolarized state in vitro. Reliance on the hydrolytic activity of the ATPase as an auxiliary proton pump within the ETC could be an adaptation of Fmr1 KO mitochondria to maintain the membrane potential in the setting of futile leak. Regardless of the etiology, the relatively high, supra-physiologic Fmr1 KO membrane potentials generated in vitro caused the proton leak channel to favor a closed formation. This phenomenon explains why hyperpolarized Fmr1 KO mitochondria appeared relatively more efficient during state 3 respiration and why the leak channel was masked. Yet, because of its sensitivity to changes in voltage, the channel opened during state 4o respiration as the membrane potential declined following the addition of either malonate or oligomycin.

Using a variety of specific inhibitors and assessment of calcium-induced calcium release, we were able to identify the mPTP, a non-selective inner mitochondrial membrane mega-channel, as the source of the pathological proton leak in Fmr1 KO mitochondria. Identification was based on knowledge that the mPTP is voltage-dependent, can conduct protons, and is specifically inhibited by CsA (62-64). Pore opening uncouples oxidative phosphorylation, reduces the efficiency of mitochondrial respiration, and generates heat (65-68). The relative hyperthermia, elevated lactate, and increased brain lactate-to-pyruvate ratios that we observed in Fmr1 KO mice provided evidence for non-shivering thermogenesis, uncoupled mitochondrial respiration, and open probability of the mPTP in vivo. Quenching of intramitochondrial calcein fluorescence within Fmr1 KO neocortical neurons confirmed mPTP opening in vivo. Importantly, we detected open probability of the mPTP in FVB control mitochondria as well, however, such opening appeared to be physiological during development given the limited effects of closing the pore on various outcome measures. In contrast, the open mPTP in the developing Fmr1 KO forebrain was pathological given the magnitude of intramitochondrial fluorescence quenching, the reduced threshold for calcium-induced calcium release, the effect of CsA on calcium loading capacity and the mitochondrial membrane potential in the setting of excessive proton leak, and the neurodevelopmental consequences of blocking the pore during synaptogenesis.

Physiological regulation of the mPTP is important for cellular health and development (66, 69). In the developing heart, for example, closure of the mPTP induces mitochondrial maturation and cardiomyocyte differentiation (70). In the developing nervous system, transient opening of the mPTP generates short bursts or flashes of mitochondrial superoxide which suppress neural progenitor cell proliferation and promote differentiation (71). These superoxide flashes also mediate the metabolic switch from aerobic glycolysis to oxidative respiration (71). Thus, neuron maturation is regulated, in part, by the mPTP. We found that closing the pore with CoQ₁₀ on P9 acutely induced changes in morphology and density of dendritic spines in Fmr1 KOs. Because Fmr1 KO spines are pathologically unresponsive to novel sensory stimuli in the early postnatal period, our findings indicated modulation of spine dynamics following closure of the mPTP, similar to the long-term potentiation-induced spinogenesis seen shortly after theta-burst stimulation in the normally developing rodent hippocampus (72, 73). Importantly, blocking the pathological leak on P9 normalized Fmr1 KO dendritic spine density several weeks later. This suggests that the mPTP may regulate synapse maturation and development and that a pathologically open pore maintains the spines in an immature state.

Excessive protein synthesis and loss of stimulus-induced translation are believed to disrupt synapse function and cause the myriad of symptoms seen in FXS (52, 74). In contrast to the widely accepted paradigm, however, recent work identified decreased protein translation and impaired developmental capacity in Fmr1-deficient Drosophila oocytes resulting in defective embryo neurodevelopment (75). These provocative findings suggested that FMRP may enhance protein translation in a development-specific manner during the period of its peak expression (76). Consistent with this, we found significantly lower rates of protein synthesis in the 10-day old Fmr1 KO mouse forebrain during synaptogenesis. It is known that Fmr1^-/Y synaptoneurosomes fail to synthesize proteins in response to stimulation and that the number of synapses with polyribosomal aggregates are markedly reduced in Fmr1 KO mice in the first few weeks of life (77, 78). Because the rate of protein synthesis in the immature rodent brain is expected to be relatively increased during synaptogenesis, decreased protein synthesis on P10 likely reflected translational failure in the developing Fmr1 KO forebrain (54, 55). Oxidative phosphorylation has recently been shown to be necessary for synapse plasticity and stimulus-induced synaptic protein translation (6). In fact, depletion of mitochondria within a local dendritic compartment abolishes the ability of stimulated synapses to undergo morphological change and synthesize new proteins (6). In our work, we found that closing the pathological mitochondrial leak in Fmr1 mutant mice with CoQ₁₀ restored rates of protein synthesis. This suggests that inefficient oxidative phosphorylation prevents neurons within the developing Fmr1 KO brain from meeting the substantial metabolic demands of synapse formation and protein synthesis during synaptogenesis (4, 5).

Our data indicate that CoQ₁₀-mediated regulation of the mPTP is important in the developing brain and that deficiency results in a pathologically open pore. However, we acknowledge that our work has several limitations. First, we evaluated forebrain mitochondria as a conglomerate. Although FMRP expression is widespread on P10 and occurs in neurons, astrocytes, microglia, and oligodendrocyte precursors, future investigation will need to focus on determining mitochondrial function in a cell-specific or even sub-cellular manner (16). Also, we have yet to elucidate exactly how loss of FMRP causes CoQ deficiency. FMRP binds to the mRNA of all of the known enzymes in the CoQ biosynthesis pathway (3, 79). Thus, it is possible that FMRP deficiency disrupts endogenous CoQ production directly. On the other hand, indirect mechanisms are just as feasible and will need to be identified in future work. Despite an incomplete understanding, however, the data clearly demonstrate a link between futile mitochondrial proton leak within the developing brain and key features of the FXS phenotype.

Our findings are significant because they introduce an entirely new direction for FXS investigation and elucidate a previously unknown therapeutic target. Furthermore, they create an opportunity for translation to the human condition. For example, elevated core body temperature and blood lactate levels could be developed as biomarkers of excessive proton leak in FXS patients. Secondly, mitochondrial inefficiency could be targeted during synaptogenesis in affected children using a variety of relatively safe therapeutic agents, such as CoQ₁₀ (80). Although CoQ₁₀ has low bioavailability and must be administered in high doses to achieve brain penetration, such an agent could be seamlessly incorporated into future multimodal therapeutic approaches to treat FXS (81, 82).

Because adequate mitochondrial function is essential for synaptic protein translation during plasticity, synaptogenesis may represent, on one hand, a critical period of vulnerability for FMRP deficient neurons and, on the other hand, a window for therapeutic intervention (6). Synaptogenesis in the human brain peaks within the first two years of life, however, treatment strategies for infants with FXS currently do not exist (83, 84). This is largely because children do not present with signs and symptoms of FXS until about 3-4 years of life (85). Although there has been a push for the institution of prenatal or newborn screening for the disease, the lack of targeted and effective therapy for infants has served as a major barrier (84). Thus, development of effective treatments for infants and young children with FXS, such as CoQ₁₀, could bolster the case for newborn screening. With further investigation and enhanced knowledge, we hope to gain a better understanding of how FMRP deficiency causes mitochondrial dysfunction in the developing brain and results in the FXS phenotype. Ultimately, we may be able to explore a novel class of pharmacological agents designed to restore mitochondrial efficiency, rescue the processes of stimulus-induced synaptic protein translation and synapse plasticity, and improve behavior in FXS.

Nonstandard Abbreviations

FXS

Fragile X syndrome

FMRP

Fragile x mental retardation syndrome

CoQ

coenzyme Q

mPTP

mitochondrial permeability transition pore

ETC

electron transport chain

CsA

cyclosporine A

BAT

brown adipose tissue

DNP

dinitrophenol

Oligo

oligomycin

Dub

decylubiquinone

cAT

carboxyatractyloside

GDP

guanosine diphosphate

TMRE

tetramethylrhodamine ethyl ester