F₁F₀ ATP Synthase–Cyclophilin D Interaction Contributes to Diabetes-Induced Synaptic Dysfunction and Cognitive Decline

Large epidemiological studies have demonstrated that diabetes significantly increases the risk of dementia, cognitive decline, and Alzheimer’s disease (1–3). These central nervous system (CNS) complications, referred to as diabetic encephalopathy, are present in patients with type 1 and type 2 diabetes (4,5), with no effective treatment currently (6). Compared with other diabetic complications, including cardiovascular and renal failure, diabetic encephalopathy remains relatively unexplored. Investigations of the cellular basis of diabetic impairments in peripheral organs have identified abnormalities in mitochondrial dynamics, morphology, and function in a variety of cell types, including islet cells, hepatocytes, skeletal muscle cells, mononuclear blood cells, and endothelial cells (7–9). However, the implications of diabetes on mitochondrial dysfunction in the brain and its molecular basis have yet to be fully understood.

Our recent studies demonstrated that CNS neurons are highly susceptible to mitochondrial dysfunction in diabetes-affected mice (10). Here we seek to identify whether the ion channels on the mitochondrial membrane play a critical role in the pathology of mitochondria dysfunction and neuron degeneration in the diabetic brain. Several ion channels have been identified in the mitochondrial membrane: the ATP-sensitive K⁺ channel (11), the mitochondrial Ca²⁺ uniporter (MCU) (12), and, a high-conductance channel, the mitochondrial permeability transition pore (mPTP) (13). These play critical roles in the regulation of intracellular Ca²⁺ homeostasis, energy production, mitochondrial morphology, and cell death (14–16). Of these mitochondrial membrane channels, mPTP opening is a trigger for mitochondrial dysfunction and cell death under pathological conditions such as ischemic stroke (17,18), Alzheimer’s disease (19,20), Ullrich congenital muscular dystrophy (21), Parkinson’s disease (22,23), and others (24). In diabetes-affected cells, including β-cells (25), adipocytes (26), and skeletal muscle cells (27), cyclophilin D (CypD)–mediated mitochondrial permeability transition is involved in the dissipation of the mitochondrial inner membrane electrochemical gradient, reduction of β-cell mass, and development of insulin resistance. In addition, the mPTP is hypersensitive to Ca²⁺ in the diabetic rat heart and thereby contributes to cardiac dysfunction (28). Thus, investigation of diabetes-induced mPTP opening in brain mitochondria is highly valuable for the development of new therapeutic interventions for diabetic neurological complications, including cognitive decline. The single-channel patch clamp technique provides a direct way to determining the ion channel activities, including the channel conductance and probability of opening. Here we used the single-channel patch clamp to directly measure the abnormalities of single-channel currents in the brain mitochondria from diabetic mice.

Although mPTP has been considered as an inducible activity of profound disease relevance, the molecular identity of the mPTP is not well understood. CypD has been suggested as an important regulator based on the pharmacological and genetic determinant, showing that cyclosporine A (CsA), a CypD-specific inhibitor, was able to inhibit mPTP opening (29) and that mitochondria lacking CypD were desensitized to mPTP opening (14,15). A novel mechanism for mPTP formation involving regulation of mPTP structure and function via F₁F₀ ATP synthase–CypD interaction has recently been proposed (30–32). This newly proposed mPTP model might also provide a link between mitochondrial energy deficiency and mPTP opening, because the interaction between ATP synthase and CypD not only causes mPTP opening but also decreases the activity of F₁F₀ ATP synthase (31). These studies raise the following unexplored key questions: Does interaction between CypD and F₁F₀ ATP synthase occur under stress or disease conditions such as diabetes? If so, does blocking F₁F₀ ATP synthase–CypD interaction protect against diabetes-induced mPTP opening and mitochondrial degeneration? Is CypD-mediated mitochondrial dysfunction linked to hyperglycemia and diabetes induced synaptic alterations? Does blocking CypD-dependent mPTP restore synaptic structure and function insulted by diabetes? Thus, uncovering the mechanism by which diabetes alters this vital F₁F₀ ATP synthase/CypD-dependent mitochondrial process is essential.

Using a single-channel mitochondrial patch clamp and a newly generated CypD-deficient diabetes mouse model, we comprehensively evaluate the consequences of changes in mPTP properties and mitochondrial function associated with diabetes. We further assess whether blockade of F₁F₀ ATP synthase–CypD interaction by deletion of CypD suppresses mPTP opening and restores synaptic plasticity and cognitive function in diabetic mice. Our investigation provides new insights into the role of CypD in mitochondrial degeneration and synaptic pathology relevant to the pathogenesis of diabetes-associated dementia and cognitive decline.

Postmortem brain materials from patients with diabetes and age-matched control subjects without diabetes were obtained from Banner Sun Health Institute (Sun City, AZ). Frozen and fixed human brain tissues for this study were from subjects who had enrolled in the Brain and Body Donation Program (33) at the Banner Sun Health Research Institute, which is approved by the Banner Health Institutional Review Board. All patients were given written informed consent. Detailed information for each of the subjects studied is reported in Supplementary Table 1. There was no statistically significant difference in PMI between brain tissues from healthy control subjects and from patients with diabetes.

Protocols involving animal use were approved by the University of Kansas-Lawrence Animal Care and Use Committee in accordance with the National Institutes of Health guidelines for animal care. CypD homozygous-null (Ppif^−/−) mice were gifts from Dr. Jeffery D. Molkentin (15). Diabetes was induced in 3-month-old wild-type (WT) or Ppif^−/− male mice by intraperitoneal administration of streptozotocin (STZ) (at 50 mg/kg, freshly dissolved in 100 mmol/L sterile citric acid, pH 4.5; Sigma-Aldrich) for 5 consecutive days, as previously described (34). Metabolic features are reported in Supplementary Table 2. After 2–3 months' duration of STZ-induced diabetes, these mice were evaluated with the Morris Water Maze (MWM) test. Brains, not including the olfactory bulb and cerebellum, were dissected, weighed, and snap-frozen in liquid nitrogen or used for cortical mitochondria isolation for further analysis, as described below.

Cerebral cortices and hippocampi were dissected and homogenized in cold mitochondria isolation buffer (225 mmol/L d-mannitol, 75 mmol/L sucrose, and 2 mmol/L K₂HPO₄, pH 7.2). Mitochondria were isolated by centrifugation, as previously described (10).

Mitochondrial swelling assay was performed on 100 µg of mitochondria in 1 mL of mitochondria swelling buffer (225 mmol/L d-mannitol, 75 mmol/L sucrose, 2 mmol/L K₂HPO₄, 5 μmol/L glutamate, and 5 µmol/L malate, pH 7.2). Swelling was triggered by the addition of 100 μmol/L CaCl₂. In some experiments, mitochondria were incubated with CsA for 5 min on ice before the swelling assay. Swelling was monitored with light scattering at 540 nm on an Ultrospec 3100 pro Spectrophotometer (Amersham Biosciences) for 12 min, including a 2-min baseline before Ca²⁺ addition.

Ca-induced cytochrome c release was detected in the mitochondria using previously described methods (15). Briefly, isolated mitochondria were incubated in the isolation buffer at room temperature in the presence of CaCl₂ with indicated concentrations for 30 min. Then the mitochondria were centrifuged at 8,000g for 10 min, and the resulting mitochondrial pellets and supernatant were subject to immunoblotting assay.

CypD membrane translocation (19), mitochondrial oxygen consumption rate (OCR) (19,35), and ATP synthase activity (36) were measured as previously described.

Mitochondrial single-channel patch clamp was performed, as previously described (37,38), with modifications. Briefly, 5 µL solution of freshly isolated mitochondria (about 2–5 µg mitochondria) was added to the bath recording solution (150 mmol/L KCl, 20 mmol/L K-HEPES, and 1 mmol/L CaCl₂, pH 7.2) in a glass bottom recording chamber, which was then left undisturbed for several minutes to allow mitochondria to settle onto the bottom of the chamber. Patch pipettes were pulled in two steps from glass capillaries (World Precision Instrument) using a Narishige puller and polished by heating (Narishige) to yield a resistance of 20–50 MΩ when filled with pipette solution (150 mmol/L KCl and 20 mmol/L K-HEPES, pH 7.2). Mitochondria were approached by means of an electrically driven micromanipulator (MPC-200, Sutter Instrument). Recordings were collected using a MultiClamp 700B amplifier, a digitizer (Digidata 1440A), and pClamp 10.3 software (Molecular Devices) after obtaining a seal resistance of 1 to 2 GΩ. Single-channel currents were recorded at the fixed holding potential indicated in the respective figures. P_o represents the average channel opening probability and was determined by all-point analysis.

For immunoprecipitation, isolated mitochondria were lysed in PBS containing 0.5% NP-40, 1 mmol/L phenylmethylsulfonyl fluoride, and protease inhibitor mixture (Calbiochem). Mitochondria lysates (1 mg/mL, 300 µL) were then incubated overnight under wheel rotation at 4°C in the presence of anti-ATP synthase antibody (Cat# ab109867; Abcam) or nonimmune IgG. Complexes were precipitated by A/G-Agarose beads (#20421; Pierce) and washed in PBS. Immunoprecipitation supernatants were separated by SDS-PAGE, followed by immunoblotting analysis. For immunoblotting assays, brain tissues, isolated mitochondrial fractions, or protein samples were homogenized in extraction buffer (10 mmol/L Tris-HCl [pH 7.4], 100 mmol/L sodium chloride, 1 mmol/L EDTA, 1 mmol/L EGTA, 1 mmol/L sodium fluoride, 20 mmol/L sodium pyrophosphate, 2 mmol/L sodium orthovanadate, 1% Triton X-100, 10% glycerol, 0.1% SDS, and 0.5% deoxycholate) containing 1 mmol/L phenylmethylsulfonyl fluoride and protease inhibitor mixture (Calbiochem).

Samples were separated under denaturing conditions on 10% SDS-PAGE and transferred to a nitrocellulose membrane (Amersham). Membranes were then incubated overnight at 4°C with primary antibodies: anti-mouse CypD (Cat# ab110324; Abcam), Rabbit anti-human CypD (generated in our laboratory) (19), Hsp60 (Cat# ADISPA806D; Enzo), N-methyl-d-aspartate receptor subtype 2B (NR2B) (Cat# ab28373; Abcam), postsynaptic density protein 95 (PSD95) (Cat# ab16495; Abcam), synapsin I (Cat# PA1-4673; Affinity Bioreagents), synaptophysin (Cat# MAB5258; Chemicon), cytochrome c (Cat# 4272s; Cell Signaling), cyclooxygenase IV (Cat#A21348; Invitrogen), or β-actin (Cat# A5441; Sigma-Aldrich). Membranes were incubated with horseradish peroxidase–conjugated anti-rabbit or anti-mouse IgG (Life Technology), visualized using enhanced chemiluminescence detection reagents (Thermo Fisher Scientific), and scanned by FluorChem HD2 (Cell Biosciences). The intensity of immunoreactive bands was quantified by ImageJ software (National Institutes of Health).

Mouse brains were fixed with 4% paraformaldehyde at 4°C for 2 days after saline perfusion. Fixed brains were cut into 30-μm-thick sections on a Vibratome (Leica VT1000S). Precut free-floating temporal cortex sections (40 µm thick) of postmortem human brains were obtained from Banner Sun Health Institute (Sun City, AZ) for immunohistochemistry. The following primary antibodies were used for immunostaining: CypD (Cat# ab110324; Abcam), MAP2 (Cat# 1281959; Boehringer Mannheim), PSD95 (Cat# 16495; Abcam), synaptophysin (Cat# MAB5258; Chemicon), synapsin I (Cat# PA1-4673; Affinity Bioreagents), anti–4-hydroxy-2-nonenal (4-HNE) IgG (1:100, Abcam). Alexa Fluor 488 or 594 secondary antibodies (Invitrogen) or biotin-conjugated goat anti-rabbit IgG (Sigma-Aldrich) were used. Sections were imaged under a Leica DMI4000B confocal microscope, and images were analyzed by the Universal MetaMorph Image Program (Molecular Devices).

Formalin-fixed pancreas sections underwent antigen retrieval in boiling citrate buffer (pH 6.0) for 10 min before labeling with antibodies against insulin (Cat# ab7842; Abcam). Sections were incubated with horseradish peroxidase–conjugated IgG (Sigma-Aldrich), followed by staining using 3-amino-9-ethylcarbazole (AEC) substrate. Serum was collected after 6-h fasting, and insulin levels were measured by ELISA kit (Crystal Chem), following the manufacture’s instruction.

We measured tissue reactive oxygen species (ROS) levels in mouse brains using electron paramagnetic resonance (EPR) assay (39,40) and MitoSOX Red (Molecular Probes) staining, as previously described (41).

WT or Ppif^−/− hippocampal neurons were cultured from day 1 newborn mice in Neurobasal B27 media containing 25 mmol/L d-glucose (Invitrogen). Tetramethylrhodamine methyl ester (TMRM) was used to stain 12–14 day in vitro neurons, as previously described (19).

The MWM test was performed as previously described (34). The training protocol for the hidden platform version of the MWM consisted of four trials (60-s maximum) each day for 6 consecutive days. The probe trial was done on day 7. The trajectories were recorded with a video tracking system. Investigators were blinded to mouse genotypes and treatments until the behavioral tests were finished.

We performed long-term potentiation (LTP) recordings as previously described (42,43). Slices were prepared using a chopper (Electron Microscopy Sciences) and maintained at 29°C in an interface chamber (Warner Instruments) with artificial cerebrospinal fluid containing (in mmol/L): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, 24 NaHCO₃ and 15 glucose, bubbled with 95% O₂/5% CO₂. Slices were allowed to recover for at least 1.5 h before recording. Field-excitatory postsynaptic potentials were recorded from the CA1 region of the hippocampus by placing the bipolar silver stimulating electrode at the level of the Schaeffer collateral fibers, whereas the recording electrode was placed in the stratum radium area of CA1. Extracellular responses were acquired using Clampex Software 10.4 (Molecular Devices) and a microamplifier (IE-210; Warner Instruments). LTP was induced by ɵ-burst stimulation (4 pulses at 100 Hz, with the bursts repeated at 5 Hz and each tetanus including three 10-burst trains separated by 15 s). Data collection and analysis were performed by investigators blinded to the experimental conditions.

All data are expressed as means ± SEM. Statistical analysis with two-tailed Student t test and one-way ANOVA was performed using StatView statistics software and GraphPad Prism. Pearson correlation analysis was performed using GraphPad Prism. P < 0.05 was considered significant.

To determine ion channel abnormalities in diabetes-affected brain mitochondria, we first recorded single-channel events and analyzed the difference in channel opening probability of mitochondria from STZ-administered mice versus vehicle controls. Black, well-separated, and round vesicles of ∼1 µm in diameter were approached and gigasealed by pipette tip (Fig. 1A). In the mitochondria-attached configuration, we recorded 61 single-channel currents in vehicle-mitochondria and 45 currents in STZ-mitochondria, with detection channel conductance ranging from 100 ps to 1.5 ns. Most of them in the vehicle-mitochondria and STZ-mitochondria displayed ∼250 ps (Supplementary Fig. 1A), 500–700 ps (Supplementary Fig. 1B), or large-channel conductance (>800 ps) (Fig. 1B). In the subgroup of channels with large channel conductance, STZ-mitochondria showed significantly increased single-channel opening probability (P₀) compared with vehicle-mitochondria, but no significant difference in P₀ was found in the subgroups of ∼250 ps or 500–700 ps events (Fig. 1C and Supplementary Fig. 1A and B).

mPTP has been established as a high conductance ion channel with up to 1.3 ns whose opening can be induced by increased cytosolic Ca²⁺ concentrations or specifically blocked by CsA (29). According to CsA sensitivity, the large conductance channels were subdivided into two groups: mPTP, with CsA sensitivity (Fig. 1D) and non-mPTP channel, without CsA sensitivity (Supplementary Fig. 1C). Large conductance channels with CsA sensitivity displayed increased P₀ differentials between vehicle- and STZ-mitochondria (Fig. 1E, upper panel), but no significant difference in P₀ was observed in non-mPTP channels (Fig. 1E, lower panel), suggesting that enhanced mPTP opening contributes to the increased P₀ in the subgroup of large conductance channels in STZ-mitochondria. No significant difference in mPTP channel conductance was observed between vehicle- and STZ-mitochondria (Fig. 1F).

mPTP regulates solute exchange between the mitochondrial matrix and the cytoplasm. Prolonged mPTP opening induced by Ca²⁺ overload leads to mitochondrial swelling and cytochrome c release (14,15). Consistent with patch clamping results showing increased mPTP opening, brain mitochondria from STZ diabetic mice displayed increased swelling (Fig. 2A) and cytochrome c release (Fig. 2B) when stimulated by Ca²⁺, indicating that more STZ-mitochondria transform from the closed to the open state under Ca²⁺ stimulation compared with vehicle-mitochondria. In addition, CsA, the CypD inhibitor, suppressed overswelling of STZ-mitochondria in a dose-dependent manner (Fig. 2C and D). These data further illustrated the enhancement of CypD-dependent mPTP opening in diabetic mitochondria.

The enhanced mPTP opening in response to elevated Ca²⁺ led us to examine whether these changes are associated with increased expression of CypD in diabetic mitochondria. STZ mice exhibited significant elevation of CypD expression (Fig. 3A) by 1.5- to 2.5-fold in the hippocampus and cortex, regions related to memory formation, compared with nondiabetic littermates. Considering that increased CypD levels might be caused by increased mitochondria proliferation, we next examined the CypD levels in isolated brain mitochondria. Compared with vehicle-mitochondria, STZ-mitochondria displayed higher CypD contents (Fig. 3B), suggesting that it is unlikely that increased CypD expression in diabetic brain mitochondria was caused by mitochondrial proliferation. Hsp60 protein levels in mitochondrial samples from vehicle or STZ mice were comparable and thus served as a loading reference. Accordingly, in situ immunohistochemistry staining showed increased density of CypD in the hippocampus CA1 region and cerebral cortex (Fig. 3C) in diabetic mice. These results demonstrate that CypD expression is increased in the diabetic brain, which likely contributes to the enhancement of mPTP opening.

The molecular nature of the mPTP remains the subject of protracted scientific debate, but progress has recently been made with the identification of the F₁F₀ ATP synthase as the pore-forming component of mPTP, which provides a novel molecular mechanism underlying the modulation of pore formation through F₁F₀ ATP synthase–CypD interaction (30,32). Enhanced mPTP opening in response to elevated Ca²⁺ and increased expression of CypD in the diabetic brain led us to examine whether these are linked by increased interaction of CypD with F₁F₀ ATP synthase. First, we examined translocation of CypD from the mitochondrial matrix to the inner membrane in response to Ca²⁺ overload. STZ-mitochondria displayed increased translocation of CypD to the mitochondrial inner membrane after Ca²⁺ (100 μmol/L) treatment (Fig. 3D). Second, we examined the interaction between F₁F₀ ATP synthase and CypD by coimmunoprecipitation assay using brain mitochondria. As shown in Fig. 3E, CypD was present in F₁F₀ ATP synthase immunoprecipitates, which is consistent with previous reports (31,32). Furthermore, the amount of CypD in F₁F₀ ATP synthase immunoprecipitates was increased in STZ-mitochondria compared with vehicle-mitochondria (Fig. 3F). These results suggest that the diabetes-induced elevation of CypD provides more available CypD protein for recruitment to the mitochondrial inner membrane and subsequently interacts with F₁F₀ ATP synthase, leading to mPTP formation under Ca²⁺ overload insulted by diabetes.

Given that increased CypD interaction with F₁F₀ ATP synthase was associated with mPTP dysfunction in the brain affected by diabetes, we next determined whether CypD deficiency is able to rescue diabetes-associated mitochondria and neuronal perturbation by using Ppif ^−/− mice. In Ppif ^−/− mice, the absence of CypD was verified by PCR and immunoblotting for CypD protein expression (Supplementary Fig. 2A–C). Ppif ^−/− mice do not have any detectable phenotypic anomalies, as we previously described (19). To validate the effect of CypD deficiency on mitochondrial membrane potential under physiological condition, we monitored TMRM signals in primary cultured hippocampal neurons derived from Ppif ^−/− mice by staining of live images with TMRM, a fluorescent probe widely used to measure the mitochondrial membrane potential. No significant difference in TMRM signals was observed between CypD-deficient neurons and WT neurons (Supplementary Fig. 2D). Blood glucose levels, body weight, β-cell loss in pancreatic islets, and serum insulin levels were comparable between WT-STZ mice and CypD-deficient STZ mice, suggesting that loss of CypD does not affect STZ-induced metabolic changes (Supplementary Table 2 and Supplementary Fig. 3).

Cortical mitochondria isolated from diabetic mice showed greater swelling induced by 100 μmol/L Ca²⁺ than normal mitochondria (Fig. 4A), whereas Ppif ^−/− mitochondria isolated from nondiabetic and diabetic mice were resistant to Ca²⁺-induced swelling (Fig. 4A). No difference in Ca²⁺-induced cytochrome c release from mitochondria was observed between Ppif^−/− mitochondria with and without diabetes (Fig. 4B). In addition, mitochondria lacking CypD restored decreased F₁F₀ ATP synthase activity in STZ brains (Fig. 4C). We next evaluated mitochondrial function by measuring mitochondrial oxygen consumption rate (OCR). Compared with nondiabetic WT mice, STZ-treated WT mice exhibited a significant reduction in OCR, whereas STZ administration did not affect the OCR in Ppif ^−/− mice (Fig. 4D). Given mitochondrial respiratory deficits were usually accompanied by increased ROS production (44), we determined whether CypD deficiency inhibits oxidative stress in STZ mouse brains. Brain ROS levels were measured by highly selective and sensitive EPR, which detects the presence of unpaired electrons and provides an excellent technique for measuring intracellular free radicals (39,40). The peak of the EPR signal detected was considerably higher in diabetic WT brains than in nondiabetic WT brains, whereas diabetic Ppif ^−/− brains showed lower EPR peaks, indicating much lower ROS levels in CypD-deficient brain tissues (Fig. 4E and F).

We further evaluated mitochondrial ROS generation by in situ staining with MitoSOX Red, a unique fluoroprobe highly selective for the detection of superoxide production in the mitochondria of live cells (41,45). The area occupied by MitoSOX Red staining in acute brain slices was significantly increased in the cerebral cortices and hippocampi of diabetic WT mice compared with nondiabetic WT controls. In contrast, diabetic Ppif ^−/− mice showed considerably less MitoSOX Red staining than diabetic WT mice (Fig. 4G and H). MitoSOX Red signals were extensively colocalized with mitochondrial Hsp60, confirming the presence of mitochondria-derived ROS (Fig. 4I). Oxidative posttranslational protein modifications in the brain were assessed histochemically using antibody against 4-HNE, a major product of endogenous lipid peroxidation and a reliable marker of oxidative stress. Levels of 4-HNE were significantly decreased in the cortices of diabetic Ppif ^−/− mice compared with diabetic WT mice (Fig. 4J). Together, these data suggest that blockade of F₁F₀ ATP synthase–CypD interaction by genetic deletion of CypD confers protection against diabetes-induced mitochondrial respiratory deficits and oxidative stress in the brain.

To further determine whether CypD deficiency affords a protective effect on diabetes-induced cognitive decline, we used the MWM test to assess spatial learning and memory in STZ-induced diabetic WT and Ppif ^−/− mice. During the spatial acquisition phase, nondiabetic WT mice demonstrated a day-to-day decrease in escape latency, whereas diabetic WT mice exhibited longer escape latencies, suggesting impaired spatial learning (Fig. 5A). Compared with nondiabetic WT mice, diabetic mice crossed the platform location fewer times and spent significantly less time in the target quadrant during the probe test in which the escape platform was removed (Fig. 5B–D). Notably, the impairment in learning and memory exhibited by diabetic WT mice was ameliorated in diabetic Ppif ^−/− mice, as indicated by shorter escape latency, increased target quadrant occupation, and more target zone crossings (Fig. 5A–D). The performance of nondiabetic Ppif ^−/− mice did not differ from that of nondiabetic WT mice in the MWM test (Fig. 5A–D), suggesting that CypD deficiency has no effect on hippocampus-dependent spatial memory under physiological condition. There were no observable differences in swimming speed among the four groups of mice, indicating that the learning and memory improvement exhibited by Ppif ^−/− mice were not likely caused by differences in motor activity (Supplementary Fig. 4A).

To assess the effect of CypD deficiency on diabetes-induced synaptic dysfunction, we assessed hippocampal synaptic plasticity, which represents the molecular basis of new memory formation and is often assessed by LTP (46,47). Although LTP was significantly lower in diabetic WT mice, inhibition of CypD-involved mPTP opening by CsA significantly blocked LTP suppression in hippocampal slices from diabetic WT mice (Fig. 5E and G). Consistent with the results from spatial learning and memory performance, CypD deficiency restored LTP in diabetic mice (Fig. 5F and G). LTP was not altered in nondiabetic Ppif ^−/− mice (Fig. 5F and G). No changes in basal synaptic transmission were found in any mice groups (Supplementary Fig. 4B).

We next evaluated whether CypD-involved synaptic dysfunction was reflected in alterations in synaptic structure. To explore possible beneficial effects of CypD ablation on synaptic perturbation in the diabetes-affected brain, we measured expression levels of synapse-related proteins including post- and presynaptic proteins. Immunoblotting of brain homogenates revealed that levels of two major postsynaptic proteins, NR2B and PSD95, were significantly decreased in diabetic WT mice, whereas Ppif ^−/− mice displayed significantly restored levels of these two proteins (Fig. 5H). Levels of synapsin I and synaptophysin, which are both presynaptic proteins, were not significantly different between diabetic WT mice and nondiabetic WT mice (Fig. 5H). These results were further confirmed by in situ staining of PSD95 puncta in mouse brain sections. Decreased amounts of PSD95-positive puncta on the dendrites of pyramidal neurons were observed in the cerebral cortex and hippocampus CA1 region in diabetic WT mice compared with nondiabetic WT mice (Fig. 5I). Notably, the number of PSD95-positive puncta was increased in diabetic Ppif ^−/− mice compared with diabetic WT mice (Fig. 5I). There was no difference in the quantity of synaptophysin-positive puncta in the cerebral cortex and hippocampus of diabetic mice compared with nondiabetic WT mice (Supplementary Fig. 5A and B). In addition, there was a slight but not significant decrease in the brain weight of diabetic mice compared with their nondiabetic counterparts (Supplementary Table 2), suggesting that the decreased synaptic protein level was unlikely due to global neuronal loss. Taken together, these behavioral, electrophysiological and biochemical studies indicate that blockade of F₁F₀ ATP synthase–CypD interaction by CypD ablation protects against diabetes-induced synaptic dysfunction and memory loss.

To determine the correlation between CypD expression and diabetes-associated cognitive disorders, we evaluated the expression of CypD in brains from postmortem patients with diabetes by immunoblotting and immunohistochemistry. The study randomly selected 14 individuals (Supplementary Table 1) from subjects who had enrolled in the Brain and Body Donation Program of the Banner Sun Health Research Institute (33). These subjects received cognitive assessments until death, including Mini-Mental State Examination (MMSE) (48,49). Analysis of CypD protein levels using immunoblotting revealed significant increased expression of CypD in the inferior temporal cortex of the human brain samples (Fig. 6A and B), which correlates with MMSE scores. Analysis of CypD immunohistochemistry staining in the temporal cortex confirmed the negative correlation between MMSE scores and CypD expression (Fig. 6C and D). The difference in PMI between brain tissues from healthy control subjects and those from patients with diabetes was not statistically significant (P > 0.05) (Supplementary Table 1). These results suggest that the alternation in CypD expression levels is unlikely due to the variable degrees of autolysis of brain tissues.

Mitochondrial dysfunction and synaptic damage are pathological features of diabetes-affected brains (10,50–53). Diabetes imposes deleterious effects on the brain and increases the risk of depression, dementia, and cognitive decline (54,55). In neurons, mitochondria at synapses are vital for maintenance of synaptic function and transmission through normal mitochondrial dynamics, proper distribution and trafficking, energy metabolism, and synaptic Ca modulation. The perturbation of mitochondrial energy metabolism and the balance of mitochondrial dynamics contribute to alterations in mitochondrial structure and function induced by oxidative stress and hyperglycemia (10,40). In the current study, we have demonstrated for the first time the involvement of CypD-mediated mPTP formation in diabetes-induced brain mitochondrial and synaptic injury through F₁F₀ ATP synthase–CypD interaction in an STZ-induced diabetes mouse model. The STZ-induced diabetes mouse model is widely used for studying the mechanistic basis of diabetes-induced complications, including diabetic encephalopathy. Our previous and present studies have demonstrated that STZ-induced diabetic mice exhibit mitochondrial and synaptic abnormalities along with cognitive dysfunction, which reflect with cognitive decline in patients with diabetes (34). These data indicate the clinical relevance of the STZ-induced diabetes mouse model to human disease such as diabetes-involved CNS complications and dementia. Thus, our findings generated from Ppif ^−/− with STZ-induced diabetes provide substantial evidence that CypD/ATP synthase–mediated brain mitochondrial and synaptic pathology contributes to the cognitive decline and dementia that were seen in patients with diabetes.

Application of the single-channel patch clamp on the mitochondrial membrane revealed that diabetes specifically increased mPTP opening probability (P₀) in brain mitochondria. Consistent with this observation, isolated diabetic mitochondria displayed increased swelling and cytochrome c release during Ca²⁺ overload. These data indicate that diabetic brain mitochondria are hypersensitive to mPTP opening in response to stressors such as Ca²⁺ overload. Although the molecular properties of mPTP remain elusive, regulation of the mPTP in vivo and in vitro was extensively investigated. Matrix pH, membrane potential, oxidation-reduction state, Ca²⁺, p53, as well as the Bax and Bcl-2 family could modulate mPTP opening (17,38,56–58). These modulations of mPTP opening are CypD-dependent because they can be blocked by CsA or CypD ablation (15,17,38), suggesting that CypD functions as a switch controlling the transition of mPTP from transient opening to prolonged opening in response to mitochondrial stress. The newly proposed mPTP model based on the F₁F₀ ATP synthase–CypD interaction provides more molecular detail of CypD-dependent mPTP regulation. The new model is further supported by our findings that interaction of CypD with ATP synthase correlates with increased expression of CypD in diabetes-affected mitochondria, which significantly elevates mPTP formation and aberrant mitochondrial function (mitochondrial swelling, cytochrome c release, energy deficiency, and oxidative stress) and suppresses ATP production in diabetes. Consequently, extensive ATP synthase–CypD interaction augments oxidative stress as shown by the protective effect of CypD deficiency against mitochondrial ROS accumulation and mitochondrial respiratory function, all in conjunction with the resistance of brain mitochondria to permeability transition. We therefore concluded that increased interaction of CypD with ATP synthase, resulting in increased mPTP opening and decreased ATP production, contributes to mitochondrial dysfunction in diabetic brains.

Furthermore, we demonstrated that blockade of the F₁F₀ ATP synthase–CypD interaction by deleting the binding partner CypD not only restored mitochondrial function but also alleviated deficits in synaptic plasticity and improved learning and memory insulted by diabetes. These observations are in line with the protective effect of suppression of CypD-mediated mPTP on Ca homeostasis, oxidative stress, and mitochondrial properties. Investigation of human brains affected by diabetes suggests that CypD elevation plays an important role in the mild cognitive dysfunction or early cognitive dysfunction in diabetes. This correlation between CypD expression and cognitive decline was also evident in STZ-induced diabetic mice. CypD expression was increased in the hippocampus and cerebellar cortex of diabetic brains, which expectedly displayed significant deficits in spatial memory and synaptic plasticity. Notably, CypD deficiency protected diabetic mice from synaptic plasticity deficits and spatial memory decline as well as restored synaptic proteins. Expression levels of CypD are significantly associated with mitochondrial properties, including the capacity for mPTP opening and mitochondrial function. Low expression of CypD in brain mitochondria increases resistance to mPTP opening induced by Ca (59), whereas high expression of CypD results in greater vulnerability to mPTP opening and higher required levels of CsA (an inhibitor of CypD) to inhibit mPTP opening (19,60,61). Importantly, elevated CypD expression levels observed in the temporal cortices of patients with diabetes and STZ-induced diabetic brains point to the clinical relevance of CypD in the pathogenesis of diabetes-involved neurodegeneration. The biochemical and functional data presented in our study reflect a causal link between CypD-mediated mitochondrial abnormalities and cognitive impairment relevant to diabetes.

Another potential mechanism underlying diabetes-induced mitochondrial and synaptic dysfunction in diabetic mice may be involved in amyloid-β (Aβ) accumulation in parallel with elevation of CypD expression. The interaction of Aβ with CypD elicits CypD-containing mPTP formation in an Aβ-enriched environment, consequently causing brain mitochondrial oxidative stress, mitochondrial swelling, synaptic injury, and cognitive impairment (19). Hyperglycemia could interfere with Aβ or amyloid precursor protein metabolism in diabetic transgenic mice overexpressing human amyloid precursor protein and Aβ (62–64), suggesting that diabetes exacerbates Alzheimer’s disease–like pathology such as cerebral Aβ accumulation. However, compared with chronic diabetes with Aβ accumulation, no significant changes in Aβ levels were found in the acute STZ-induced diabetic brain compared with the vehicle-treated brain (34,50,65). Thus, the increased interaction of CypD with ATP synthase, caused by increased CypD expression, might represent an alternative molecular mechanism underlying cognitive dysfunction in diabetic encephalopathy and dementia.

In summary, our studies clearly demonstrate that lack of CypD remarkably improved cognitive and synaptic function in mice with STZ-induced type 1 diabetes. In the diabetes-affected brain, elevated CypD expression increases the interaction of CypD with ATP synthase, leading to mPTP opening, disruption of mitochondrial integrity, impairment of ATP production, and increased ROS generation. Notably, CypD ablation strengthens mitochondria against diabetes-induced mitochondrial and synaptic perturbation, at least in part, by maintaining the integrity of mPTP, as shown by improved brain mitochondrial function and decreased ROS production, as well as by rescued synaptic plasticity and memory deficits in diabetic CypD knockout mice. Although the pathological events from an animal model of short-term untreated diabetes may not be comparable to that from years of diabetes or treated patients with diabetes, our findings clearly demonstrate that CypD is a mediator of diabetes-induced brain dysfunction, which provides a foundation for preclinical effect via targeting CypD/ATP synthase–perturbed mPTP as a potential therapeutic approach for diabetes-associated dementia and cognitive impairment, a medical condition for which few effective therapeutic options are currently available.

Funding. This study was supported by grants from the National Institute of Neurological Disorders and Stroke (R01-NS-089116) and National Institute on Aging (R37-AG-037319, R01-AG-044793). Support for the EPR instrumentation was provided by National Science Foundation Chemical Instrumentation Grant (0946883).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.Y. designed and performed experiments, analyzed data, and wrote the paper. F.D. and Q.Y. performed experiments and analyzed data on mitochondrial function, EPR, and immunostaining. L.W. conducted the STZ injections and performed mouse behavioral tests. Z.Z. isolated the mitochondria and performed the mitochondrial swelling assay. C.Z. performed the coimmunoprecipitation assay. Y.W. assisted with the mitochondrial patch clamp, manuscript preparation, and discussion. L.-F.L. and D.G.W. provided the human tissues and the information of the case subjects. J.T.D. assisted with EPR experiments. S.S.Y. initiated and supervised the research, designed experiments, developed the concept, and wrote the paper. S.S.Y. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.