Diabetes has adverse effects on the brain, especially the hippocampus, which is particularly susceptible to synaptic injury and cognitive dysfunction. The underlying mechanisms and strategies to rescue such injury and dysfunction are not well understood. Using a mouse model of type 2 diabetes (db/db mice) and a human neuronal cell line treated with high concentration of glucose, we demonstrate aberrant mitochondrial morphology, reduced ATP production, and impaired activity of complex I. These mitochondrial abnormalities are induced by imbalanced mitochondrial fusion and fission via a glycogen synthase kinase 3β (GSK3β)/dynamin-related protein-1 (Drp1)-dependent mechanism. Modulation of the Drp1 pathway or inhibition of GSK3β activity restores hippocampal long-term potentiation that is impaired in db/db mice. Our results point to a novel role for mitochondria in diabetes-induced synaptic impairment. Exploration of the mechanisms behind diabetes-induced synaptic deficit may provide a novel treatment for mitochondrial and synaptic injury in patients with diabetes.
Mitochondrial dysfunction, synaptic damage, and resultant impairment in cognitive function are pathological features of diabetes (1–11). Diabetes adversely affects the brain and confers increased risk for depression and dementia (3,12,13). In neurons, synaptic mitochondria are vital for maintenance of synaptic function and transmission through normal mitochondrial dynamics, distribution, and trafficking, as well as energy metabolism and synaptic calcium modulation. Imbalance in mitochondrial dynamics contributes to oxidative stress– and hyperglycemia-induced alterations in mitochondrial morphology and function (10,14,15). The molecular and cellular mechanisms regulating the precise distribution of mitochondria in neurons are poorly understood.
In addition to well-known adverse effects on the cardiovascular and peripheral nervous systems, the hippocampus is particularly susceptible to diabetes-induced episodic memory impairment (12) and synaptic plasticity deficits (5,7,8,11,16–20). For example, rats rendered diabetic by treatment with the pancreatic β-cell toxin streptozotocin (STZ) (a model of type 1 diabetes) exhibit spatial learning and memory impairment (5,8). Similar deficits have been reported in the insulin-resistant type 2 diabetes mouse model (7,8,16). Clinical investigations also documented impaired cognitive function in human subjects with type 1 (caused by insulin deficiency) as well as type 2 (mediated by insulin resistance) diabetes compared with age-matched nondiabetic subjects (21,22). In support of those findings, long-term potentiation (LTP) of synaptic transmission, believed to be a cellular mechanism of learning and memory, is impaired in the hippocampus of diabetic animal models (5,8,20). The decline of hippocampal LTP in diabetic individuals is partially due to increased corticosterone levels and/or impaired insulin signaling (8,19). Given that mitochondrial function is abnormal in many tissues negatively affected by diabetes and that mitochondria are important for maintenance of synaptic plasticity, as this organelle provides energy and regulates intrasynaptic metabolic homeostasis, we hypothesized that mitochondria-dependent mechanisms contribute importantly to diabetes-induced synaptic dysfunction such as hippocampal LTP deficit.
Mitochondria are dynamic organelles, which undergo continuous fission and fusion. Fission events are regulated by dynamin-related protein (Drp1), and fusion events are regulated by the large dynamin-related GTPases known as mitofusin (Mfn)1 and -2 as well as optic atrophy 1 (OPA1) (23–25). Alterations in mitochondrial dynamics significantly impact mitochondrial numbers and shape, respiratory enzyme activity, and ATP production. Imbalances of mitochondrial fission and fusion in diabetes result predominantly from upregulation of Drp1, which induces mitochondrial division and dysfunction (impaired respiration and ATP production) in a variety of cell types, including islet cells (26), hepatocytes (27), skeletal muscle cells (9,28), mononuclear blood cells (29), endothelial cells (30), and dorsal root ganglion neurons (15,31). Mitochondrial dysfunction is also suggested as a cause for development of insulin resistance in skeletal muscle cells and hyperglycemia in type 2 diabetes (32,33). Compared with other cell types, neurons are highly susceptible to mitochondrial dysfunction because neuronal transmission is regulated by energy homeostasis in synapses. In type 2 diabetes, hippocampal synaptic plasticity is widely believed to be impaired, but mechanisms for hippocampal mitochondrial damage and consequent neuronal dysfunction are unknown. The current study sought to elucidate how hippocampal mitochondria respond to diabetes-induced mitochondrial and synaptic alterations that result in synaptic injury.
Research Design and Methods
These studies were approved by the Animal Care and Use Committee of University of Kansas-Lawrence in accordance with the National Institutes of Health guidelines for animal care. Leptin receptor heterozygous mice (animals are on C57BLKS background) were purchased from The Jackson Laboratory. Heterozygous db/m mice were crossed with each other to generate three genotypes of mice: wild-type mice, leptin receptor heterozygous mice (db/m), and homozygous leptin-deficiency mice (db/db). The genotype of offspring was identified by PCR using primers (forward, ACA AGA ACA AAA AGC CTG AAA CC, and reverse, CCA AAC TGA ACT ACA TCA AAC C) for genotyping as previously described (34). All db/m and db/db male mice used for the described experiments were 5–6 months of age unless otherwise stated. The investigators were blinded to the mouse genotype and treatment conditions in performing the immunoprecipitation, electron microscopy, and immunohistochemistry experiments.
SK Cell Culture and Transfections
Human SK-N-SH cells were cultured on 24-well culture plates at low density and incubated with DMEM with 10% FBS and 5.5 mmol/L d-glucose. Cells were then transfected with plasmids containing constitutively active forms of hemagglutinin–glycogen synthase kinase-3β (HA-GSK3β) (HA-GSK3βS9A) or a dominant-negative form of GSK3β (HA-GSKK85A) (Addgene), green fluorescence protein (GFP)-tagged wild-type Drp1 or Drp1K38A (provided by Dr. Yi-Ren Hong, Kaohsiung Medical University Hospital, Taiwan), or empty control vectors (GFP or HA) using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions. Forty-eight hours after transfection, cells were treated with indicated reagents and assessed for changes in mitochondrial morphology and function as well as molecular signaling pathways.
Electrophysiological recordings were performed on coronal hippocampal slices (400 μm in thickness) as previously described (35). Hippocampal slices were recovered at 37°C for at least 1 h and then maintained in an interface chamber at 29°C and perfused with artificial cerebral spinal fluid (124 mmol/L NaCl, 4.4 mmol/L KC, 1 mmol/L Na2HPO4, 25 mmol/L NaHCO3, 2 mmol/L CaCl2, 2.0 mmol/L MgSO4, and 10 mmol/L glucose) continuously bubbled with 95% O2 and 5% CO2. Field-excitatory postsynaptic potential (fEPSPs) was recorded from the CA1 region of the hippocampus by placing both the stimulating and the recording electrodes in the CA1 stratum radiatum. Basal synaptic transmission (input-output curve) was assayed by plotting the stimulus voltage (V) against slopes of fEPSP to generate input-output relations. A 30-min baseline recording was established using low-frequency stimulation (0.033 Hz; 0.1 ms impulse duration) and then adjusted intensity that induced fEPSPs with ∼30% of the maximal fEPSP amplitude. LTP was induced by θ-burst stimulation (four pulses at 100 Hz, with the bursts repeated at 5 Hz, and each tetanus, including three 10-burst trains separated by 15 s). In experiments using inhibitors, drugs were continuously perfused over slices starting at least 1 h before LTP induction. Values of fEPSP slope were expressed as the percentage change relative to their mean baseline amplitude.
Drugs were prepared as stock solutions and were diluted to the final concentration immediately before use. Incubation of hippocampal slices with drugs was performed in either a recovery chamber or interface-recording chamber as needed. Drugs were added to culture plate wells. Final concentrations and sources of the drugs were as follows: 4-benzyl-2-methyl-1,2,4-thiadiazolidine-3,5-dione (TDZD8) (5 μmol/L; Sigma-Aldrich), mdivi-1 (10 μmol/L; Sigma-Aldrich), and glucose (50 mmol/L; Sigma-Aldrich). The final concentration of vehicle control DMSO was less than 0.5% in all experiments. For chronic in vivo administration of mdivi-1, the drug was dissolved in DMSO (10 or 25 mg/mL) as stock solution and further diluted in saline (DMSO: saline = 1:9) immediately before daily injection in diabetic or nondiabetic male mice at a dosage of 10 or 25 mg/kg i.p. Drug was administered once a day beginning at 19 weeks of age and continued for 2 weeks. The same amount of DMSO dilution in saline was prepared and then administered to age-matched male nondiabetic and diabetic littermates as controls. One day after the end of treatment, animals were subjected to either electrophysiological studies or biochemistry assays.
Brain mitochondria were isolated from whole brain cortex of mice or SK cultures as previously described (36). Briefly, samples were placed in 9 mL ice-cold mitochondria isolation buffer (225 mmol/L mannitol, 75 mmol/L sucrose, and 2 mmol/L K2HPO4 [pH 7.2]) and homogenized (10 strokes) using a Douce homogenizer (Kontes Glass Co.). Homogenate was centrifuged at 1,300g for 5 min at 4°C. The resultant supernatant was then centrifuged at 34,000g for 10 min after layering on 15% Percoll. After centrifugation, the homogenate was resuspended and incubated for 5 min on ice in 20 mL mitochondria isolation buffer with 0.02% digitonin and centrifuged at 8,000g for 10 min. The pellet was washed twice in 1.5 mL mitochondria isolation buffer and centrifuged again at 8,000g for 10 min. The final pellets were resuspended in 200 µL mitochondria isolation buffer. Total protein concentration of isolated mitochondria fraction was determined by Bradford protein assay (Bio-Rad Laboratory).
Immunoprecipitation and Immunoblotting Assays
For immunoprecipitation, hippocampal slices were lysed in nondenaturing lysis buffer containing 50 mmol/L Tris HCl (pH 8), 150 mmol/L NaCl, 1 mmol/L EDTA, 0.5% NP-40, and protease inhibitors (EMD Millipore). After centrifugation at 12,000g for 10 min at 4°C, the supernatant was precleaned with protein A/G-agarose beads (Thermo Scientific) for 2 h at 4°C, and 500 μg precleaned protein extracts was incubated with antibody to GSK3β or Drp1 for 16 h at 4°C. Precipitated complexes were washed in lysis buffer and bound proteins analyzed by immunoblots. For immunoblot experiments, hippocampal slices, cultured cells, or isolated mitochondrial/cytosol fractions were homogenized in cell lysis buffer (Cell Signaling), and then samples were separated in SDS-PAGE gels and transferred onto membranes. Polyclonal rabbit anti–phospho-GSK3β (Ser9) (Cell Signaling), anti-GSK3β (3D10) (Cell Signaling), anti–phospho-Drp1 (Ser616) (Cell Signaling), anti-Drp1 (Thermo Scientific), mouse anti–heat shock protein 60 (HSP60) (Enzo), and mouse anti–α-tubulin (Cell Signaling) were used as the primary antibodies. Binding sites of primary antibodies were visualized with horseradish peroxidase–conjugated anti-rabbit IgG antibody (Life Technology) or anti-mouse IgG antibody (Life Technology) followed by the addition of enhanced chemiluminescence substrate (GE Healthcare). Band relative optical density was determined by NIH ImageJ software (public domain) and normalized with α-tubulin levels.
Confocal and Electron Microscopy to Assess Mitochondrial Morphology
Brains were fixed by perfusion with 4% paraformaldehyde, and after additional fixation for 72 h at 4°C, coronal sections (25 μm) were cut with a Vibratome (Leica VTS1000). After blockade with 5% goat serum in PBS, slices were incubated with primary antibody (anti–superoxide dismutase 2 [SODII], Enzo Life Sciences; anti–cytochrome c oxidase [COXIV], Abcam; anti–microtubule-associated protein 2 [MAP2], Chemicon; or anti–Drp-1, BD Transduction Laboratories) overnight at 4°C. Secondary antibody (Alexa Fluor 488–conjugated goat anti-rabbit IgG and/or Alexa Fluor 594–conjugated goat anti-mouse IgG [Invitrogen]) was applied for 1 h at room temperature. Nuclei were stained by DRAQ5 (5 μmol/L; Cell Signaling) for 5 min at room temperature. Blocking serum and primary and secondary antibodies were applied in PBS with 0.3% Triton X-100. Brain sections incubated with preimmune serum or second antibody alone were used as negative controls. Mitochondria in cultured cells were visualized with MitoTracker red (100 nmol/L applied for 30 min at 37°C before fixation; Molecular Probes). Brain slices and cell cultures were examined under confocal microscope (Leica TCS SPE) with a 63 × 1.4 NA Apochrome objective (Carl Zeiss MicroImaging, Inc.) at 488 nm, 594 nm, or 543 nm wavelengths, respectively. For quantification of mitochondrial staining, four brain sections from each animal (five animals in each group) were randomly selected at different rostro-caudal hippocampal levels. More than twenty randomly selected hippocampal pyramidal neurons from each slice or SK cells in three independent experiments were used for quantification of mitochondrial length and density using MetaMorph software as we previously described (37). For analysis of electron microscopic images, we used our previously published method (35,38). Briefly, animals were perfused with 4% paraformaldehyde and 0.1% glutaraldehyde, and brains were fixed in the same fixative buffer overnight at 4°C. Small pieces of hippocampi were embedded in LR white medium (Electron Microscopy Laboratory). Ultra-thin sections were counterstained with uranyl acetate and examined using a JEOL 100S transmission electron microscope. Mitochondrial parameters such as major/minor axis length, perimeter, area, and density were measured using Metamorph software. The form factor was calculated as perimeter2/4π × area. The investigators were blinded to mouse genotype and treatment conditions in performing morphological analysis and immunohistochemical experiments.
Mitochondrial Respiration Complex Activity and ATP Measurement
Mitochondrial respiration complex activity was measured in hippocampi or in SK cell homogenates as previously described (37,39). Briefly, hippocampi or cultured cells were homogenized and sonicated in the isolation buffer containing 250 mmol/L sucrose; 20 mmol/L HEPES, pH 7.2; and 1 mmol/L EDTA. NADH-ubiquinone oxidoreductase (COXI) enzyme activity was determined in 25 mmol/L potassium buffer containing KCl, Tris-HCl ,and EDTA (pH 7.4). The change in absorbance was monitored at 340 nm wavelength every 20 s for 6 min using an Amersham Biosciences Ultrospect 3100 pro spectrophotometer. For homogenized samples (50 μg protein), the oxidation of NADH was recorded for 3 min after the addition of 2 μg/mL antimycin, 5 mmol/L MgCl2, 2 mmol/L KCN, and 65 μmol/L coenzymes Q1 to the assay mixture, and then 2 μg/mL rotenone was added to the mixture. The absorbance of samples was measured for another 3 min. Enzyme activities in complex II, complex III, complex IV, and citrate synthase activity were determined as previously described (37). We also determined ATP content in hippocampus using an ATP Bioluminescence Assay Kit (Roche) according to manufacturer’s instructions with a Luminescence plate reader (Molecular Devices) with an integration time of 10 s.
Quantitative Real-Time PCR
Total RNA was extracted from hippocampal tissue using TRIzol reagent (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer’s protocol as described in our previous study (40). cDNA synthesis was performed using random hexamer primers and the TaqMan reverse transcription kit (Applied Biosystems, Foster City, CA). The TaqMan probes and primers Mm00802465_m1 were used for the standard TaqMan PCR procedures in the ABI Prism 7700HT Sequence Detection System (Applied Biosystems). Each treatment was performed in quadruplicate, and data were analyzed with the comparative Ct method (41) and normalized to GAPDH RNA of the corresponding sample.
Serum Insulin and Corticosterone Measurement
Animals were fasted for 6 h before obtaining blood samples. Blood samples were centrifuged at 2,500 rpm (4°C) for 5 min. Serum was then collected and stored at −80°C. ELISA kits for insulin and corticosterone were purchased from Abnova. Manufacturer’s instructions for measurement of serum insulin and corticosterone concentrations were followed.
Data are presented as means ± SEM. Statistics were analyzed using Statview software (SAS Institute). For comparison between the two groups, a two-tailed independent Student t test was used. For comparisons between multiple groups, one-way ANOVA was used followed by individual post hoc Fisher tests when applicable. P < 0.05 was considered statistically significant.
Alterations in Neuronal Mitochondrial Morphology and Function in Hippocampi of Diabetic Mice
To determine whether type 2 diabetes alters brain mitochondrial dynamics, architecture, and function, we examined the changes in hippocampus, which is particularly susceptible to adverse effects of diabetes (hippocampal damage results in episodic memory impairment  and synaptic plasticity deficits [5,7,8,11,16–20]). Using confocal microscopy and SODII (mitochondrial protein) immunostaining, we first examined mitochondrial morphology in CA1 neurons in db/db hippocampus compared with control db/m mice. SODII immunopositive mitochondrial density in db/db hippocampal neurons was significantly lower than that of db/m control mice (Fig. 1A and B). We next performed immunostaining with another mitochondrial marker, COXIV, to further examine the mitochondria density. As shown in Fig. 1C–D, the COXIV immunofluorescence intensity was significantly decreased in CA1 pyramidal neurons of db/db hippocampus compared with control db/m mice. To provide further evidence, we used electron microscopy to evaluate morphologic changes in hippocampal mitochondria (Fig. 1E) such as density, mitochondrial major axis length (mitochondrial length), the ratio of mitochondrial major axis length to minor axis length (mitochondrial width), and perimeters and area for calculation of form factor (indicating mitochondrial branching). The mitochondrial density (Fig. 1F), the length of mitochondria (major axes) (Fig. 1G), the ratio of mitochondrial length to width (Fig. 1H), and mitochondrial form factor (Fig. 1I) were significantly reduced within db/db hippocampus compared with db/m control hippocampus, suggesting decreased mitochondria branching and increased fragmentation in diabetic hippocampus. It has been reported that changes in mitochondrial numbers and shape can lead to impaired respiratory enzyme activity and energy production in tissue of diabetic mice (15,27,30,42–45). Therefore, we next evaluated hippocampal mitochondrial function by measuring key enzyme activity associated with respiratory chain and ATP levels. Complex I activity was significantly reduced in db/db hippocampal mitochondria compared with age-matched nondiabetic littermates (Fig. 1J); complex II and IV activities were unchanged (Supplementary Fig. 1). Similarly, hippocampal ATP abundance in db/db mice was significantly lower than in db/m control mice (Fig. 1K). Our data demonstrate that mitochondrial shape changes were paralleled with decreases of mitochondrial respiratory enzyme activity and ATP abundance in diabetic hippocampus.
Given that imbalance of mitochondrial fission and fusion plays a critical role in maintenance of morphology, distribution, and function, we next investigated whether mitochondrial dynamics properties were altered in diabetes. Immunoblotting of hippocampal homogenates revealed that levels of the mitochondrial fission protein Drp1 were increased in diabetic mice (Fig. 1L and O), but levels of mitochondrial fusion proteins Mfn2 and OPA1 in diabetic db/db hippocampus were not significantly different from db/m control levels (P > 0.05 db/m vs. db/db) (Supplementary Fig. 2A). While the effect of Drp1 upregulation/activation on mitochondrial fusion/fission is sometimes still controversial, accumulated evidence suggests that the Drp1-dependent mitochondrial elongation or fragmentation is closely associated with its phosphorylation status at different amino acid sites. Among those phosphorylated residues, phosphorylation of Ser616 was reported to be associated with mitochondrial fission (46). Indeed, we observed an increase in Ser616 phosphorylation in parallel with an increase in total Drp1 expression in hippocampus of db/db mice (Fig. 1M–O). An elevated level of Drp1 was found in diabetes-affected CA1 neurons (Supplementary Fig. 2B and C). Furthermore, quantitative real-time PCR assay revealed that the gene expression of Drp1 was increased∼1.3 fold in db/db animals compared with that of db/m controls (Supplementary Fig. 2D). Because the mitochondrial translocation of phosphorylated Drp1 at Ser616 is necessary to promote fission (46), we then determined the levels of Drp1 in mitochondria. Immunoblotting of the isolated brain mitochondria displayed increased levels of Drp1 (Fig. 1P and S) and of phosphorylation of Drp1 (Fig. 1Q–S) in diabetes-affected brain compared with db/m controls. These results suggest increased mitochondrial fission in diabetic mice, which may subsequently impair mitochondrial morphology and function.
Mitochondria Fission Is Increased in Diabetes Cell Culture Model
Next, using a human neuronal SK cell line, we elucidated the role of Drp1 in mitochondrial alterations under diabetic conditions. We first investigated whether there was an increased mitochondrial translocation of Drp1 in diabetic conditions. Treatment with high glucose (50 mmol/L d-glucose for 1 h) lowered the cytosol Drp1 level (Fig. 2A) but increased mitochondrial Drp1 protein (Fig. 2B) compared with that in control cells treated with l-glucose, a nonmetabolizable form of glucose (data not shown). In addition, immunostaining of Drp1 demonstrated localization of Drp1 in mitochondria treated with high glucose as demonstrated by overlaying a Drp1 image with the mitochondrial marker MitoTracker (Fig. 2C). These data suggested that high-glucose exposure induce mitochondrial translocation of Drp1 in cultured SK cells, which was similar to what we observed in the type 2 diabetes–affected animal brain (Fig. 1).
We also examined whether the high-glucose treatment in vitro induced mitochondrial defects comparable with those we demonstrated in vivo. Here, we observed decreased mitochondrial length (Fig. 2D and E), density (Fig. 2D and F), and complex I enzyme activity (Fig. 2D and G) in SK cells subjected to high-glucose conditions (50 mmol/L) for 1 h. Notably, the nonmetabolizable sugar, mannitol or l-glucose, used as an osmotic control, did not induce any noticeable alterations, meaning that the mitochondrial changes were specifically dependent on the high glucose–induced diabetic insult (Supplementary Fig. 3). Importantly, high glucose–induced deleterious effects were significantly prevented by pretreatment with mdivi-1 (10 μmol/L), an inhibitor of Drp1 activity (47) (Fig. 2D–G). For determination of a specific effect of Drp1 on mitochondria pathology, cells were transfected with inactivated Drp1 and evaluated for mitochondria morphology and function under diabetic conditions. In this case, we transfected SK cells with GFP-tagged dominant negative mutant Drp1 (DN-Drp1) to determine the effect of DN-Drp1 on high glucose–induced mitochondrial morphology changes. Compared with the control vector–transfected cells, inactivation of Drp1 significantly preserved mitochondria against deleterious effects of high-glucose treatment (Fig. 2H–K). Additionally, knockdown of Drp1 with small interfering RNA (siRNA) protected mitochondrial morphology, dynamics, and function from adverse effects of high-glucose exposure (Fig. 2L-O). No significant difference of mitochondrial dynamics and function were found in cells transfected with DN-Drp1 (Supplementary Fig. 4A–C) or Drp1 siRNA (Supplementary Fig. 4D–F) compared with cells transfected with GFP vector alone or control siRNA under the vehicle-treated normal glucose condition. These results indicate that Drp1 potentiates neuronal mitochondria abnormalities under diabetic conditions in vitro as well.
GSK3β Directly Regulates Drp1 in Diabetic Condition
It is known that GSK3β is activated in db/db mice owing to dephosphorylation on Ser9 residue (48–50) and it can use Drp1 as one of its substrates in an in vitro system (51–53). We thereby evaluated whether GSK3β could directly regulate Drp1 in diabetes-affected brain. Immunoprecipitation of Drp1 followed by immunoblotting with GSK3β showed a GSK3β immunoreactive band. Conversely, immunoprecipitation of GSK3β showed Drp1 immunoreactive bands (Fig. 3A), suggesting the formation of a complex of Drp1 with GSK3β in a mouse model of type 2 diabetes. GSK3β is a serine/threonine kinase, and phosphorylation on Ser9 residue (p-ser9 GSK3β) inactivates it (50). Most interestingly, p-ser9 GSK3β levels are decreased in both diabetic human subjects and rodent models of diabetes (48,49), indicating increased GSK3β activation. In the current study, we also observed an increased GSK3β activation (reduced levels of p-ser9 GSK3β) in db/db mice compared with db/m control littermates (Fig. 3B). Treatment with TDZD8 (5 μmol/L for 1 h), an inhibitor of GSK3β (54), resulted in a significant increase of p-ser9 GSK3β levels in db/db hippocampus (Fig. 3B), suggesting that TDZD8 treatment may be sufficient to inhibit GSK3β activation in diabetic hippocampal tissue. Further, in the presence of TDZD8, we observed a decreased Drp1 protein expression (P < 0.001, vehicle-perfused db/db hippocampal slices vs. TDZD8-perfused db/db slices [Fig. 3C]) and decreased Drp1 Ser616 phosphorylation (P < 0.05, vehicle-perfused db/db hippocampal slices vs. TDZD8-perfused db/db slices [Fig. 3C and D]) in db/db mouse hippocampal tissue.
To investigate direct effect of GSK3β on Drp1-mediated mitochondrial alterations under diabetic conditions, human neuronal SK cells were treated with GSK3β inhibitor TDZD8 or transfected with a dominant negative form of GSK3β (DN-GSK3β) in the presence of high glucose. Clearly, pharmacological blockade (Fig. 3E–G) or genetic inhibition of GSK3β (Fig. 3H–I) prevented high glucose–induced Drp1 changes. In addition, transfection of SK cells with the GSK3βS9A (constitutively active form of GSK3β) resulted in increased Drp1 expression and its phosphorylation (Fig. 3J and K). Additionally, we performed triple immunostaining for mitochondria (red), Drp1 (green), and GSK3β (blue) to determine whether there was a colocalization of GSK3β and Drp1 in mitochondria under a diabetic condition. As shown in Fig. 3L, the high-glucose treatment induced a colocalization (white) of GSK3β/Drp1 within mitochondria in SK cultures.
Since Drp1 plays a key role in maintenance of mitochondrial structure and function and is directly regulated by GSK3β, we next tried to determine whether modulation of GSK3β causes changes in mitochondrial morphology under diabetic conditions. Pharmacological blockade of GSK3β with TDZD8 significantly suppressed mitochondrial fission (Fig. 4A and B) and increased mitochondrial density (Fig. 4A and C) and complex I activity (Fig. 4D) in the presence of high-glucose treatment. Genetic inactivation with DN-GSK3β also preserved mitochondria against high-glucose insult (Fig. 4E–H). Neither pharmacological (Supplementary Fig. 4G–I) nor genetic (Supplementary Fig. 4J–L) inhibition of GSK3β had any significant effect on the mitochondrial morphology and function in vehicle-treated cells. As expected, activation of GSK3β by transfecting cells with GSK3βS9A elicited mitochondrial impairments (Fig. 4I–L). These results demonstrate that GSK3β-Drp1 interaction mediates mitochondrial alteration under diabetic condition.
To validate that GSK3β acts as an upstream Drp1 signaling in diabetic conditions, we cotransfected SK cells with GSK3βS9A with and without DN-Drp1 plasmids or DN-GSK3β with or without Drp1 plasmids. Transfection with DN-Drp1 prevented effects of GSK3βS9A-mediated alterations in mitochondrial morphology and complex I enzyme activity (Fig. 5A–D). However, transfection with DN-GSK3β did not completely restore Drp1 transfection–induced mitochondria impairments (Fig. 5E–H). These results indicate that Drp1 was required for GSK3β activation–mediated mitochondrial dysfunction. In addition, suppressing Drp1 did not change GSK3β activation that was induced by either high-glucose treatment in cells or type 2 diabetic hippocampus. As showed in Supplementary Fig. 5A and B, reduced GSK3β phosphorylation in cells with high-glucose treatment was not restored by modulation of Drp1 via transfection with DN-Drp1 or Drp1 siRNA. Similarly, administration of mdivi-1, a potent cell-permeable inhibitor of Drp1 (47), to db/db mice failed to alter levels of p-ser9 GSK3β (P > 0.05, vehicle-treated db/db vs. mdivi-1 treated db/db hippocampal slices [Supplementary Fig. 5C]). These data suggest that GSK3β is a likely upstream signaling molecule that controls Drp1 activity under diabetic conditions.
Mitochondrial Impairment Is Responsible for Synaptic Plasticity Deficit
We further sought to investigate the role of Drp1 in mediating mitochondrial degeneration and energy supply deficiency in diabetes-affected hippocampus by assessing whether suppression of Drp1 expression and activity in diabetic brain would have favorable effects. Treatment with Drp1 inhibitor mdivi-1 (25 mg/kg or 10 mg/kg once daily for 2 weeks) resulted in significantly increased mitochondrial content (Fig. 6A and B), complex I enzymatic activity (Fig. 6C), and ATP levels (Fig. 6D) in db/db mouse hippocampus. When the same mdivi-1 treatment was administered to db/m control animals, we found no significant changes in mitochondrial density (Fig. 6A and B), complex I activity (Fig. 6C), or ATP levels (Fig. 6D). Electron microscopy confirmed that mdivi-1 treatment (10 mg/kg) significantly preserved mitochondrial morphology by maintaining mitochondrial length and shape (Fig. 6E–I).
Because mitochondrial morphology and function are crucial for synaptic plasticity and function (55–58), we next investigated mitochondria-dependent mechanisms of synaptic plasticity in hippocampus of the db/db mouse model of type 2 diabetes. Using electrophysiological analysis, we observed reduced LTP (Fig. 7C), whereas basal neuronal transmission (input-output relationship [Fig. 7A]) or short-term plasticity (pair-pulse facilitation [Fig. 7B]) was not changed in db/db hippocampus. Treatment with mdivi-1 significantly reversed the decline in hippocampal LTP from diabetic hippocampus (P < 0.05, mdivi-1–treated db/db mice vs. vehicle-treated db/db hippocampal slices [Fig. 7G and H]) without changes in basic neuronal transmission or pair-pulse facilitation (Fig. 7E and F). The treatment of nondiabetic db/m mice with mdivi-1 did not significantly affect hippocampal LTP (Fig. 7D and H), which was coincident with the unaffected mitochondrial structure and function (Fig. 6). Taken together, these results indicate a protective effect of inhibition of diabetes-induced Drp1 activation on synaptic dysfunction.
Given that GSK3β activation modulated Drp1 signaling, we next evaluated whether treatment of hippocampus slices with GSK3β inhibitor TDZD8 or LiCl (4 mmol/L) recapitulated the protective effects of mdivi-1 on synaptic function in vivo. Indeed, inhibition of GSK3β significantly protected hippocampus LTP against diabetes insult (Fig. 7I–K) without changing basal synaptic function in db/m hippocampus. Since insulin (19) and corticosterone (8) signaling pathways were reported to mediate LTP impairment in rodent models of diabetes, we explored their reaction to mdivi-1 treatment. Consistent with the previous report (8,19), we did find increased blood levels of insulin and corticosterone (P < 0.05 vs. db/m) in db/db mice (Supplementary Fig. 6A and B). Mdivi-1 treatment did not significantly change insulin and corticosterone levels (P > 0.05) (Supplementary Fig. 6A and B) or blood glucose content in control or db/db mice (Supplementary Table 1). These findings suggest that the protective effect of mdivi-1 on LTP restoration in diabetic hippocampus is not likely due to insulin or corticosterone signaling pathways.
In the current study, we report a novel and pivotal role of mitochondrial dysfunction in diabetes-induced synaptic impairment. Our data suggest a GSK3β/Drp1-dependent connection between mitochondrial dysfunction in diabetic neurons and LTP deficit. We believe that such a link is rational because provision of neurons with access to energy/metabolism homeostasis is the prerequisite for synaptic transmission. Understanding mitochondrial alterations in animal models of diabetes will likely generate novel approaches for the treatment of diabetes-associated neurological disorders.
Perturbations in maintenance of mitochondrial morphology and function are involved in several adverse effects of diabetes in many organs including pancreas (26,32), liver (27), skeletal muscle (9,28), and the vascular system (29,30). Brain is highly enriched with mitochondria and strongly reliant on oxidative phosphorylation for energy production. The current literature suggests a connection between diabetes and mitochondrial biogenesis and fission within mouse dorsal root ganglia (15,31). One of our recent publications suggests that mitochondrial dysfunction is associated with synaptic deficit in an Alzheimer disease mouse model superimposed with type 1 diabetes (36). However, it is still unclear whether dysfunctional synaptic plasticity (resulting in observed episodic memory impairment) links to mitochondrial defects in type 2 diabetes.
Staining of db/db mouse hippocampal pyramidal neurons for the mitochondrial marker SODII or COXIV revealed a significant reduction of mitochondria compared with control(s). This reduction in neuronal mitochondria is suggestive of a diabetes-mediated interference with mitochondrial biogenesis and fission. In addition, these mitochondrial morphology changes were confirmed by electron microscopy. Measurement of hippocampal mitochondrial functional capacity in this mouse model of diabetes indicated significantly reduced activity of mitochondrial respiration complex I and lowered ATP content. Our results are consistent with findings of a similar study probing the skeletal muscle of patients with type 2 diabetes (33), in which there are impairments in energy metabolism and respiratory complex enzyme activity. Balance of mitochondrial fission and fusion processes is crucial for maintenance of mitochondrial morphology and function (59,60). In diabetic hippocampus, the mitochondrial fission protein, Drp1, was significantly increased and its phosphorylation was coincidentally elevated. On the other hand, levels of Mfn2, OPA1, and other auxiliary mitochondrial proteins such as HSP60 (data not shown) were not altered in diabetic hippocampus. Our findings are consistent with previous investigations of diabetic neuropathy, which included reports of increased Drp1 expression and mitochondrial fission in dorsal root ganglion neurons of 6-month-old db/db mice (15,31). In contrast to the greater numbers of mitochondria in dorsal root ganglion neurons (31), we note that hippocampal neurons in 5- to 6-month-old db/db mice displayed smaller numbers of mitochondria, while there was no sign showing mitochondrial decline among subjects younger than 3 months (data not shown). Differences in sensitivity to hyperglycemia between hippocampal neurons and dorsal root ganglion cells may account for the differences in our findings compared with previous studies (31). In addition, increased mitochondrial fission may eventually result in mitochondrial degeneration if the balance of fission and fusion is not maintained. We further tested whether mitochondrial enzyme (COXI) activity and ATP content were changed at 3 months and 6 months of age. We found that complex I enzyme activity significantly declined by 15% and 35% at 3 and 6 months of ages, respectively, though no significant alteration in ATP content was found at 3 months of age (Supplementary Fig. 7A and B). Similarly, the LTP significantly declined by 16% in 3-month-old db/db hippocampus compared with that of age-matched db/m control mice (Supplementary Fig. 7C–E). These data suggest age-dependent defects in mitochondrial and synaptic function. Pharmacologic or genetic inactivation of Drp1 significantly prevented mitochondrial morphology and function changes in db/db hippocampus or human SK cells under high-glucose conditions, indicating a role of Drp1 in diabetes-induced mitochondrial morphology and functional changes.
There is at least one mechanism by which diabetes elevates Drp1 expression and phosphorylation resulting in mitochondrial alterations. To our knowledge, Drp1 is able to be phosphorylated at several sites by different kinases. It was first identified that Drp1 can be phosphorylated at Ser637 by protein kinase A (61,62), which was coincident with mitochondrial elongation. CaMK I can also phosphorylate the Ser637 site to promote mitochondrial fission (63). While the role of phosphorylated Drp1 at the Ser616 site still remains paradoxical, increasing evidence points out that Ser616 Drp1 is related to the mitochondrial fragmentation; for example, cyclin-dependent kinase 1/cyclin B–dependent phosphorylation at Ser616 of Drp1 induced mitochondrial fission during mitosis (64). Furthermore, our recent studies suggest a role of extracellular signal–related kinase 1/2 in oxidative stress–induced increased phosphorylation of Drp1 (37). In our present study, we observed mitochondrial degeneration/fragmentation along with elevated levels of Drp1 phosphorylation at Ser616 in diabetic conditions both in vivo and in vitro. GSK3β is activated in db/db mice in vivo owing to dephosphorylation on Ser9 residue (48,49) and can use Drp1 as one of its substrates in in vitro systems (51–53). We therefore focused on the role of GSK3β-mediated mitochondrial defects by modulating Drp1 Ser616 phosphorylation. We have clearly demonstrated that GSK3β activation was elevated in either type 2 diabetes–affected hippocampus or high glucose–treated human neuronal cell lines. While the underlying mechanisms for the alterations in GSK3β activation in in vivo and in vitro diabetic conditions are not clear, diabetes-induced Akt dephosphorylation (Supplementary Fig. 8A and B) may contribute, at least in part, to the GSK3β activation in addition to the aforementioned activation of other kinases. Indeed, we observe that pharmacological inhibition of GSK3β successfully prevented Drp1 phosphorylation at Ser616 in the diabetes-affected mouse brain and high glucose–treated human cell line. These results further support the role of GSK3β in diabetes-induced mitochondrial defects via Drp1-mediated abnormal mitochondrial fission, though our results would not exclude the possibility of the involvement of extracellular signal-regulated kinases 1/2 in high glucose–induced Drp1 Ser616 phosphorylation and mitochondrial fragmentation (65). Furthermore, we demonstrated that genetic activation of GSK3β without high-glucose treatment can also promote mitochondrial fragmentation. The inactivation of GSK3β prevents high glucose–induced mitochondrial alterations. Taken together, these data suggest that GSK3β is one of the mechanisms underlying diabetes-induced mitochondrial abnormalities.
GSK3β most likely acts as an upstream signaling mechanism for Drp1 upregulation in diabetes-induced mitochondrial dysfunction. Phosphorylation of Ser9 on GSK3β in db/db hippocampus was decreased, resulting in higher levels of the active form of GSK3β. Inhibition of GSK3β activity was associated with suppressed phosphorylation of Ser616 on Drp1, a critical step enabling Drp1 GTPase activity for mitochondrial fission (46). Currently, there are no inhibitors available for prevention of Drp1 Ser616 phosphorylation; however, treatment with mdivi-1 (inhibiting the GTPase activity of Drp1) maintained mitochondrial morphology and function, though GSK3β activation was unaltered. GSK3β not only acts as an upstream signaling mechanism for Drp1 but also directly interacts with Drp1, which, in turn, is responsible for phosphorylation of Drp1 Ser693 and leads to decreased GTPase activity and an enhancement of mitochondrial fusion/elongation (53). We also determined whether an inactive form of GSK3β (p-ser9 GSK3β) interacts with Drp1. We found that pull-down GSK3β (active form) yielded Drp1 immunoreaction bands in db/db animals, but no interaction was found between p-ser9 GSK3β and Drp1 (data not shown).
Using a human neuronal cell line, we directly modulated GSK3β or Drp1 activation by transfection with different targeted vectors. Inactivation of GSK3β prevented GSK3β-involved activation of Drp1 signaling and changes in mitochondrial morphology and function under diabetic conditions. However, inactivation of Drp1 did not alter GSK3β activation even though it preserved mitochondrial morphology under high-glucose conditions, suggesting a role for GSK3β in diabetes-induced activation of Drp1 signaling involved in alterations in mitochondrial morphology.
Another major novel finding of the current study is that diabetes-induced synaptic transmission deficit, at least in part, depends on GSK3β/Drp1-mediated mitochondrial abnormalities. Consistent with the results of other researchers (5,7,8,11,16–20), we observed LTP reduction in the hippocampal CA1 region in db/db mice. Because complete absence of Drp1 in mice is lethal, we were not able to generate diabetic db/db mice with Drp1 knocked out at this point; however, we successfully restored diabetes-induced mitochondrial alterations by treating mice with the Drp1 inhibitor, mdivi-1. Treatment with mdivi-1 restores mitochondrial function along with the improvement in LTP in the diabetic hippocampus. Notably, a recent investigation demonstrates that mdivi-1 also inhibits the rapidly activating delayed-rectifier potassium channel (66) in murine cardiomyocyte cell lines. While the inhibitory effect of mdivi-1 on potassium channels has not yet been tested in neurons, mdivi-1 might cause an elevated excitatory activity in the central nervous system if a potential effect on potassium ion channel exists. This requires further investigation. The beneficial effect of inhibiting Drp1-mediated aberrant mitochondrial and synaptic function by mdivi-1 treatment is unlikely owing to an insulin- or corticosterone-dependent mechanism (8). Indeed, corticosterone levels are increased in diabetic animals compared with nondiabetic littermates; however, inhibition of Drp1 using mdivi-1 failed to prevent diabetes-induced corticosterone increases but did ameliorate diabetes-induced hippocampal LTP deficits.
In summary, our study offers new insights into the role of GSK3β/Drp1 interaction in diabetes-induced mitochondrial and synaptic dysfunction. We clearly demonstrated alterations in mitochondrial structure and function in diabetes-affected hippocampus. Impaired structural or functional capacity of mitochondria is attributable to type 2 diabetes–induced impairment of synaptic plasticity. Inhibition of GSK3β/Drp1-mediated deficits in mitochondrial architecture and resultant functional changes ameliorate LTP deficit. Blockade of GSK3β activation or inhibition of mitochondrial division significantly improves mitochondrial and synaptic function under diabetic conditions. These studies suggest targets for the development of interventions to prevent or treat diabetes-induced mitochondrial deficits and accompanying neurodegeneration.
Funding. This study was supported by grants from the National Institute on Aging (R37AG037319 and R01AG044793) and National Institute of Neurological Disorders and Stroke (R01NS065482).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.H., X.G., and D.F. performed experiments and statistical analysis. Y.W. supervised and analyzed all the experimental data, wrote the manuscript, and performed experiments and statistical analysis. C.Z. and L.W. performed experiments. G.H. prepared constructs. A.A.S. and G.M.M. conducted electron microscope experiments. H.Y. performed statistical analysis. S.S.Y. directed, designed, and supervised this study and wrote the manuscript. 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.