Astrocytic GSTM2-STAT3 interaction ameliorates the diabetes associated cognitive dysfunction via targeting mitochondrial defects and oxidative stress

Abstract

Diabetes associated cognitive dysfunction (DACD) present a substantial challenge to the management of diabetic patient. The pathogenesis of DACD is intricately associated with mitochondrial defects and oxidative stress. However, the exact molecular pathways implicated have not been comprehensively elucidated. This study aimed to characterize the mechanisms by which dysregulation of glutathione S transferase mu2 (GSTM2) contributes to the pathogenesis of DACD. Astrocytic GSTM2 expression is markedly downregulated in DACD mice hippocampus employed proteomic sequencing. Hippocampal astrocytes specific GSTM2 overexpression alleviated cognitive dysfunction accompanied by inhibiting mitochondrial defects and oxidative stress in db/db mice. Conversely, astrocytic GSTM2 deficiency aggravated these dysfunctions. Immunoprecipitation-mass spectrometry and surface plasmon resonance were utilized to identify the GSTM2 - interacting proteins. The GSTM2 directly interacted with STAT3 and suppressed the phosphorylation of STAT3, thereby downregulating Drp1 signals and ultimately exerting a protective effect against mitochondrial defects and oxidative stress. Pharmacological intervention of STAT3 dysregulated mitochondrial function which influence the protective benefits conferred by GSTM2. Finally, we found that Icariin, which was explored by molecular docking and virtual screening from large-scale compound libraries, could activate the GSTM2/STAT3 pathway to improve cognitive impairment in DACD mice. Conclusively, the down regulation of GSTM2 impairs mitochondrial function and oxidative stress via the STAT3-Drp1 signaling pathway, further exacerbating DACD pathology. This discovery indicates that GSTM2 may serve as a promising and novel therapeutic target for the prevention and treatment of DACD.

Graphical abstract

Highlights

• •GSTM2 expression is down regulated in DACD.
• •Overexpression of GSTM2 alleviates cognitive dysfunction through targeting mitochondrial defects and oxidative stress.
• •GSTM2 directly interacts with STAT3 and suppresses its phosphorylation and subcellular localization.
• •Icariin modulates the GSTM2/STAT3 signaling pathway and ameliorates cognitive impairment in DACD.

1. Introduction

Globally, the incidence rates of diabetes are demonstrating a sharp upward trend. Disease spectrum prediction suggests that by the year 2045, 10.9% of the world's population will be affected by diabetes [1]. A notable neurological consequence of this epidemic is diabetes associated cognitive dysfunction (DACD), the disorder originating from the metabolic dysregulation intrinsic to diabetes. DACD profoundly impairs essential cognitive functions including learning and memory. The morbidity of DACD is 13.5% [2], and the morbidity may exceed 24.2% among people over 75 years old [3]. The development of DACD is attributed to a multifactorial and complex pathophysiological process. The pathogenesis of DACD is driven by complicated factors, including insulin resistance, oxidative stress, mitochondrial defects and inflammation [4]. Although the precise molecular and cellular mechanisms underlying the disease process remain incompletely understood. The neuropathological changes of structural damage and memory deficits are observed across multiple brain regions in rodent DACD models. Notably in the hippocampus, which is crucial for learning and memory, demonstrates particular susceptibility [5].

Astrocytes execute a diverse spectrum of crucial physiological functions within the central nervous system. These functions include the supply of metabolic substrates to neurons, the preservation of blood brain-barrier integrity, the modulation of synaptic activity, and the elimination of cellular metabolic byproducts [6]. The aforementioned functions of astrocytes depend on mitochondrial energy supply. Mitochondrial homeostasis impairment and increased reactive oxygen species (ROS) levels will induce neuronal death by inhibiting synaptogenesis, reducing phagocytosis and secreting inflammatory factors [7]. Mitochondrial dynamics, particularly the equilibrium between fission and fusion, play a pivotal role in the etiology of DACD [8,9]. The dysregulation of these processes is fundamentally associated with the emergence of mitochondrial defects, a key pathological feature. The regulatory machinery for these dynamics involves: mitochondrial fission is governed by proteins like Drp1, whereas Mfn2 is the principal controller of fusion. An imbalance toward excessive fission can trigger widespread mitochondrial fragmentation, a surge in ROS generation, and a subsequent exacerbation of mitochondrial defects [10]. Research indicates that caveolin-1 can improve DACD by alleviating mitochondrial morphological changes and neuronal damage depend on inhibition excessive mitochondrial fission and mitigate mitophagy damage [11]. Furthermore, the recovery of cognitive function in diabetic mice has been attributed to metformin administration linked to the suppression of mitochondrial fission, a reduction in oxidative stress originating from mitochondria, and the mitigation of neuronal loss within the hippocampus [12]. However, it remains unclear the mechanism by which astrocytes maintain normal mitochondrial function in the development of DACD.

As a Mu-class isoenzyme within the glutathione S-transferase (GST) superfamily, GSTM2 is distinguished from its counterparts by unique antioxidant and anti-inflammatory capabilities. GSTM2 participates in the fundamental detoxification processes characteristic of the GST family, which function as phase II enzymes by catalyzing the conjugation of glutathione (GSH) to a diverse array of exogenous compounds and endogenous metabolites [13]. GSTM2 could regulate autophagy-lysosomal pathway function to protect astrocytes from the amino pigments induced toxicity [14]. In addition, GSTM2 is released into the synaptic cleft and subsequently internalized into neurons to exert neuroprotective effects [15]. Based on the antioxidant properties, GSTM2 exerts beneficial effects as anti-oxidative stress and anti-inflammation on various central nervous system diseases [16,17]. We found that GSTM2 was significantly downregulated in hippocampal astrocytes and was associated with cognitive decline by proteomics analysis of hippocampal tissue from DACD mice. Although GSTM2 has impressive clinical value, to date, little is known about whether GSTM2 regulates mitochondrial defects in astrocytes and thereby regulates DACD, as well as the potential mechanisms involved is less well understood.

In the present study, we explore the mechanistic route through which GSTM2 provides protection against DACD. We have determined that the molecular mechanism through which GSTM2 maintains mitochondrial homeostasis is demonstrated to function via the GSTM2 - STAT3 axis in the hippocampus. This fundamental understanding not only elucidates the regulatory function of GSTM2 but also novel intervention targets for DACD treatments.

2. Material and methods

2.1. Animal studies

Eight weeks old male diabetic mice (BKS/DB−/−, db/db; 31 ± 2 g) and their age-matched nondiabetic counterparts (BKS/DB+/+, db/m; 20 ± 2 g) were procured from Beijing Huafukang Biotechnology Co., Ltd. (Beijing, China). Following acclimatization, animals were group-housed (three per cage) in transparent plastic cages. This environment was strictly maintained at a constant temperature of 22 °C and a 12-h light/dark cycle (lights on: 08:00–20:00). Standard rodent chow and tap water were available ad libitum throughout the experimental period.

2.2. Ethics statement

The design and execution of this study were rigorously guided by the principles of the 3Rs (Replacement, Reduction, and Refinement). Ethical compliance for all experimental procedures was formally granted through the approval of our research protocol by the Experimental Animal Ethics Committee of Jinzhou Medical University (Approval No. 2022050801). Furthermore, our methodologies were strictly aligned with the stipulations set forth in the Guide for the Care and Use of Laboratory Animals, operating under Permit Number: SYXK [Liao] 2019-0007.

2.3. Viral injection

To manipulate GSTM2 expression specifically in astrocytes, stereotactic surgery was performed to deliver recombinant adeno-associated viruses (rAAVs) into the hippocampal regions. db/db and db/m mice received unilateral injections of either rAAV-GFAP-GSTM2-EGFP-WPRE-hGH-pA (rAAV-GSTM2) for overexpression or rAAV-GFAP- sh-GSTM2(rAAV-shGSTM2) for knockdown. Control animals from both genotypes were injected with rAAV-GFAP-EGFP-WPRE-hGH-pA (rAAV-NC) or rAAV-GFAP-Scrambled shRNA(rAAV-Scrambled) to serve as the baseline control group. rAAV9-GFAP-GSTM2 and rAAV9-GFAP-shGSTM2, were custom-designed and subsequently sourced from OBio Technology (Shanghai) Co., Ltd., China. The procedure for stereotactic viral delivery was executed as delineated below. Initially, mice were subjected to general anesthesia via intraperitoneal administration of pentobarbital sodium. Once a stable anesthetic plane was achieved, the head of each mouse was securely immobilized in a stereotactic frame (Stoelting, USA). In preparation for the microinjection, we first safeguarded the animal's eyes from desiccation and infection by administering erythromycin eye ointment to the corneal surface. To enable precise targeting, a small craniotomy was then conducted to reveal the Bregma landmark. Using this as a reference point, we directed microinjections to unilateral hippocampal region, utilizing the following stereotactic coordinates relative to Bregma: anteroposterior (AP) −2.06 mm, mediolateral (ML) ±1.5 mm, and dorsoventral (DV) −1.9 mm. The viral vectors were then infused using a 5 μl syringe equipped with a 33-gauge metal needle (65460-02, Hamilton, USA).

The viral vectors, with a titer of ≥2.00E + 12 vg/mL as per the manufacturer's specification, were microinjected. A total volume of 300 nL was delivered at a controlled flow rate of 30 nL/min for 10 min using an injection pump. To ensure adequate diffusion and minimize backflow, the needle was maintained in situ for an additional 10 min post-injection before being slowly retracted over a 5-min period. Following the procedure, the incision was sutured, and the mice were placed on an electric blanket for postoperative recovery and monitoring.

2.4. Pharmacological intervention

As detailed in references, cohorts of db/db mice received a 12-week treatment regimen. This included oral gavage of Stattic (a selective STAT3 inhibitor, 50 mg/kg/day, S7947, Sigma-Aldrich, Germany), gavage of Icariin (100 mg/kg/day, CRN1212, Hubei Cuiyuan Bio. Co., Ltd.) and intraperitoneal administration of Colivelin-TFA (a selective STAT3 activator, 1 mg/kg/day, HY-P1061A, MCE, USA) [[18], [19], [20]]. Throughout the experimental timeline, fasting blood glucose concentrations and body weight fluctuations were serially monitored at weekly intervals. Upon completion of the 12-week intervention, cognitive function was comprehensively assessed using both the Novel Object Recognition (NOR) test and the Morris Water Maze (MWM) test. Forty-eight hours post-assessment, the animals were euthanized, and tissues were harvested for subsequent experimental procedures.

2.5. Morris water maze (MWM) test

Spatial navigation and escape performance, key indicators of cognitive function, were evaluated in mice using the MWM, a task that requires subjects to learn the location of a hidden platform within a water-filled tank to escape from the aversive stimulus. Through learning, mice begin to comprehend the spatial layout of their surroundings and exhibit certain memory abilities during the process. The MWM protocol was conducted in two distinct phases: an acquisition phase and a probe trial. During the acquisition phase, which spanned five consecutive days, each mouse was subjected to four trials daily. For each trial, the animal was carefully released into the pool from one of the four cardinal quadrants, and its swimming path was automatically tracked for a maximum of 60 s. We quantified spatial learning using the principal dependent variable, escape latency, which was the duration required for a mouse to find the hidden platform. In instances where a mouse did not locate the platform within the 60-s trial period, an experimenter gently guided it to the platform, and a latency of 60 s was assigned for that trial. To evaluate the retention of this spatial memory, a single 60-s probe trial was administered on day six. During this trial, the platform was removed from the maze, and each animal was released into the water from the quadrant diagonally opposite the target. An automated video-tracking system was then utilized to measure critical indices of memory performance, specifically the across platform times and target quadrant retention time.

2.6. Novel object recognition (NOR) test

NOR test serves as a measure for an animal's capability to distinguish a new object within a familiar environment. On the 1st day, known as the habituation phase, each mouse is familiarized with the testing arena in a 10-min session without any behavioral stimulus. On the day designated for training, two identical objects were placed in the testing arena, and each mouse was allowed a 5-min exploration session to familiarize itself with the sample objects. On the third day, in the retention phase, we replaced one of the previously placed objects with a new one, allowing the mouse a 5-min period to investigate the novel object. The duration of object exploration by each mouse was quantified using an automated, video-assisted tracking system. Cognitive performance was subsequently assessed via a Preference Index (PI), a metric calculated with the following formula: PI (%) = [(Time spent with the novel object - Time spent with the familiar object)/Total exploration time] × 100%.

2.7. Transmission Electron Microscopy (TEM)

The hippocampal tissues were meticulously micro-dissected and immediately subjected to post-fixation in a 2.5% glutaraldehyde solution (G1102, Servicebio, Beijing, China) for a duration of 2 h at a constant temperature of 4 °C. In parallel, for in vitro cellular investigations, primary hippocampal astrocytes were subjected to distinct treatment protocols, including exposure to HG, OE-GSTM2, and si-STAT3. Following a 24-h postfixation in 2.5% Glutaraldehyde, the samples were thoroughly rinsed in phosphate-buffered saline (PBS) and subsequently subjected to postfixation in a 1% osmium tetroxide solution for 3 h at ambient temperature. Thereafter, the specimens were processed through a standard protocol encompassing graded ethanol dehydration, resin infiltration, and embedding. Ultra-thin sections (60-80 nm) were then prepared using an ultramicrotome. For contrast enhancement, these sections were double-stained with uranyl acetate and lead citrate prior to visualization and image acquisition under a Transmission Electron Microscope (Hitachi, HT7700) for ultrastructural analysis.

2.8. Golgi staining

The mice brains were fixed based on the protocol of PK401 Rapid GolgiStain TM Kit (#PK401, FDNeuroo Technologies, MD, USA). Briefly, brains were incubated in Solutions A and B at room temperature in the dark for 14 days, followed by a 7-day immersion in Solution C under identical conditions. Sequentiallyl, the transparentized brains were embedded in Optimal Cutting Temperature (OCT) compound and sectioned coronally at a thickness of 100 μm using a cryostat. The resulting sections underwent histological staining with Solutions D and E, followed by a standard dehydration and coverslipping procedure. Whole-slide imaging was then performed using a panoramic section scanner (Pannoramic MIDI, 3DHISTECH, Hungary). Finally, the acquired digital images were subjected to quantitative analysis utilizing the NeuroJ plugin within the ImageJ software environment.

2.9. Label free proteomic sequencing analysis

The hippocampal tissue samples were harvested from 24 weeks old DACD mice and db/m mice and subjected to a comprehensive proteomic profiling. Following protein extraction and a rigorous quality control (QC) assessment, the prepared samples were labeled and subsequently analyzed via liquid chromatography-tandem mass spectrometry (LC-MS/MS). For qualitative protein identification and quantification, the acquired spectral data were processed against the Proteome Discoverer 2.4 database. This analytical pipeline culminated in the identification of 65,414 total spectra, which were matched to 60,760 unique peptides, ultimately leading to the characterization of 7188 distinct proteins. Following the identification of differentially expressed proteins (DEPs), which were defined by a |Fold change| ≥ 1.2 and p < 0.05, their expression profiles were subjected to volcano plot and cluster analysis for visualization and pattern recognition. Subsequently, KEGG pathway enrichment analysis was employed to uncover the functional implications and pathway alterations associated with these DEPs.

2.10. Immunoprecipitation-mass spectrometry (IP/MS)

To identify GSTM2-interacting proteins, collect samples of 2 × 10⁷ primary astrocytes cultured in HG medium for lysis, followed by immunoprecipitation assays using an anti-GSTM2 primary antibody (1:100, ab196503, Abcam). The immunoprecipitated complexes were then resolved by SDS-PAGE and visualized via silver staining. Subsequently, protein bands that were uniquely present or exhibited differential abundance. These excised bands were subjected to in-gel trypsin digestion and the resulting peptides were analyzed by liquid chromatography-mass spectrometry (LC-MS), which was performed by Jiyun Biotechnology Co., Ltd. (Guangzhou, China).

2.11. Western blot analysis

The expression levels of the target protein were evaluated by Western blotting. The following primary antibodies were used: GSTM2 (1:2000, ab196503), Drp1(1:1000, ab184247), p-Drp1(1:1000, ab314755), Mfn1(1:2000, ab126575), STAT3(1:2000, ab68153), p-STAT3(1:2000, ab32143) were from Abcam (MA, USA); Postsynaptic Density Protein 95 (1:5000, PSD95, 20665-1-AP), Synaptophysin (SYP, 1:5000, 17785-1-AP) were from Proteintech (PTG). HRP-conjugated Affinipure Goat Anti-Mouse IgG(H + L) (1:5000, SA00001-1, PTG); Rhodamine (TRITC)–conjugated Goat Anti-Rabbit IgG(H + L) (1:5000, SA00007-2, PTG).

2.12. Immunofluorescence(IF)

For the Hippocampal tissue: Subsequent to dewaxing, hydration and antigen retrieval. We subsequently permeabilized the sections by 0.3% Triton X-100 for a span of 15 min, then blocked them by 10% goat serum for 1 h at ambient temperature. Following these procedures, sections got incubated at 4 °C for a whole night with antibodies against GSTM2 (1:200, ab196503, Abcam), GFAP (1:800, ab279289, Abcam), IBA-1(1:500, Ab283319, Abcam), MBP (1:500, ab218011, Abcam), NeuN (1:800, ab104224, Abcam). Following a PBS wash, fluorescent secondary antibodies (Goat Anti-Mouse IgG H&L (Alexa Fluor® 647), ab150115, Abcam; Goat Anti-Mouse IgG H&L (Alexa Fluor® 488), ab150113, Abcam) got utilized and incubated for 2 h at room temperature, shielded from light. Primary astrocytes got subjected to the treatment process delineated earlier. 4% paraformaldehyde for a duration of 30 min at ambient temperature. Following this, the cells underwent incubation within 0.3% Triton X-100 for 15 min. After a blocking phase utilizing 10% goat serum at room temperature for 1 h, the cells got subjected to overnight incubation at 4 °C with specific antibodies against GSTM2 (1:500, ab196503, Abcam), GFAP (1:1000, ab279289, Abcam), STAT3 (1:500, ab68153, Abcam). Post PBS wash, the incubation process got carried out in the dark at room temperature for 2 h with fluorescent secondary antibodies. Images were captured as soon as possible using a confocal fluorescence microscope (BX53F, Olympus, Tokyo, Japan).

2.13. Coimmunoprecipitation (Co-IP)

Cell lysates were prepared by incubating the cells with an optimized volume of immunoprecipitation lysis buffer (Pierce, Thermo Scientific) supplemented with a protease inhibitor cocktail for 4 h at 4 °C. Following centrifugation, the clarified supernatants were harvested and subjected to overnight immunoprecipitation at 4 °C with the following primary antibodies: anti-GSTM2 (1:200, ab196503, Abcam) and anti-STAT3 (1:100, ab68153, Abcam). Subsequently, protein A/G magnetic beads (MedChemExpress, HY-K0202) were added to the antigen-antibody complexes and incubated for an additional 4 h at 4 °C. The bead-bound complexes were then washed three times with lysis buffer (50 mM Tris-HCl (pH 7.4) +150 mM NaCl+0.1% NP-40), and the immunoprecipitated proteins were eluted in loading buffer for subsequent immunoblotting analysis.

2.14. Primary astrocyte culture

Primary hippocampal tissues were isolated from C57BL/6J mice on day 2∼3 of postnatal (PND). Following euthanasia, whole brains were rapidly dissected out and placed in ice-cold Dulbecco's Modified Eagle's Medium (DMEM). Under a stereomicroscope, the meningeal layers were meticulously removed, and the hippocampus were carefully excised and finely minced with fine scissors. The minced tissue was then subjected to enzymatic digestion in 2 ∼ 3 mL of 0.25% trypsin for 10 min at 37 °C. Subsequently, the digestion was terminated, and the resulting cell suspension was filtered through a 75 μm nylon mesh to obtain a single-cell solution. Following enumeration, no need to wrap the culture dish, directly inoculate the cell suspension into a 10 cm culture dish at a density of 5 × 10 ⁵ cells/cm ², and add 8-10 mL of complete culture medium. Incubate in a 37 °C, 5% CO ₂ incubator for 1.5-2 h. After vaccination, replace the complete culture medium every 2-3 days and culture for 7-10 days Identification using immunofluorescence staining (GFAP positivity rate ≥95%). To genetically modulate primary astrocytes, plasmid-mediated transfection was employed to achieve GSTM2 overexpression and STAT3 knockdown. The OE-GSTM2 and si-STAT3 constructs utilized in this study were custom-designed and procured from OBio Technology Co. Ltd. (Shanghai, China).

2.15. Cell Counting Kit-8 (CCK8) assay and detection of lactate dehydrogenase (LDH) release test

For the assessment of cell viability and cytotoxicity, primary cultured astrocytes were seeded in 96-well plates (4 × 10³ cells/well). After reaching logarithmic growth, they were exposed to either high-glucose (HG) medium (50 mmol/L, A16828, Thermo Fisher) or mannitol (50 mmol/L, M8140-250, Beijing Solarbio Science & Technology Co., Ltd.) for 24, 48, or 72 h. Following treatment, 10 μL of CCK-8 solution was added to each well, and after a 1 h incubation at 37 °C, absorbance was measured to determine cell viability using a Cell Counting Kit-8 (CK04, Servicebio, Beijing, China). Concurrently, lactate dehydrogenase (LDH) activity in the supernatant was quantified with an LDH Cytotoxicity Test Kit (BC0685, Servicebio, Beijing, China), adhering strictly to the provided protocol.

2.16. ROS level detection

Primary cultured hippocampal astrocytes, initially seeded at a density of 3 × 10⁵ cells per well in a 6-well plate, are exposed to HG, OE-GSTM2, and si-STAT3 for 48 h. Following this treatment period, the cells are enzymatically dissociated using a 0.05% trypsin solution. After neutralization with culture medium supplemented with 2% Fetal Bovine Serum, the astrocytes were harvested and subjected to centrifugal separation at 3000 rpm for 5 min. Add 1 mL of probe working solution (DCFH-DA, 10 μmol/L), gently pipette to mix, and incubate in the dark at 37 °C in a 5% CO₂ incubator for 20–30 min. After incubation, centrifuge at 1000 rpm for 5 min and discard the probe working solution. Resuspend cells in 500 μL of pre-chilled PBS to form a single-cell suspension and immediately analyzed using flow cytometry techniques.

2.17. Detection of lipid peroxide and antioxidant capacity

The levels of malondialdehyde (MDA, SEKSM-0058) and glutathione (GSH, SEKSM-0053), as well as the activity of superoxide dismutase (SOD, BC5165), within the hippocampal tissues and cellular samples, were quantified utilizing respective ELISA or WST-1assay kits. All procedures were strictly performed in accordance with the protocols provided by the manufacturer (Servicebio, Beijing, China).

2.18. Detection of mitochondrial membrane potential

The assessment of mitochondrial membrane potential was carried out by calculating the ROS detector (JC-1, C2006, Beyotime Biotech Inc) by comparing the levels of green fluorescence (JC-1 monomer) at 530 nm with those of red fluorescence (JC-1 polymer) at 590 nm.

2.19. Seahorse extracellular flux analyzer assays

For the assessment of mitochondrial function, primary cultured astrocyteswere seeded at a density exceeding 10⁵ cells per well into specialized V-7 Seahorse plates. Subsequently, the primary cultured astrocytes underwent a pre-incubation period in a CO₂-free incubator with glucose-free Seahorse assay medium supplemented with 1 mM pyruvate. A mitochondrial stress test was then executed by sequentially introducing the following compounds via the instrument's injection ports: 20 mM glucose, 1.0 μM Oligomycin, 1.0 μM FCCP, 1.0 μM Rotenone, and 2.0 μM Antimycin A. Key parameters derived from this test, such as exogenous glucose oxidation, basal respiration, maximal respiration, spare respiratory capacity, ATP-linked respiration, proton leak, and non-mitochondrial respiration, were quantified by measuring the oxygen consumption rate (OCR).

2.20. SPR assay

The binding affinity of recombinant GSTM2 (ab99435, Abcam) for recombinant STAT3 (ab68153, Abcam) was determined via Surface Plasmon Resonance (SPR) on a BIAcore T200 system (Biacore, Cytiva), with the equilibrium dissociation constant (KD) calculated directly by the system's analytical software.

2.21. Virtual screening (VS)

The chemical library employed in this docking investigation was assembled from a pool of 20,723 entities, comprising Traditional Chinese Medicine Active Compound Library (http://selleckchem.com, L8300), Antioxidant Compound Library(https://www.tsbiochem.com/). All molecular structures were prepared using the LigPre module integrated into the Maestro 11.9 environment, during which they were subjected to cleaning, filtration, protonation, and energy minimization under the OPLS3e force field. In parallel, the GSTM2 target protein structure was predicted de novo from its amino acid sequence with AlphaFold 3 v3.0.0 (https://alphafoldserver.com/). This structural model was then refined and prepared for simulation via the Protein Preparation Wizard in Schrödinger2019, executing a series of preprocessing, optimization, and minimization steps. VS and optimization were performed with the Glide module of Schrödinger Maestro. A two-tiered docking strategy was implemented: an initial high-throughput screen was conducted with the Standard Precision (SP) mode to process the entire dataset, followed by a refined evaluation of the highest-ranked compounds using the Extra Precision (XP) mode. The XP protocol was selected for its enhanced fidelity in correlating ligand binding poses with their calculated scores. Finally, the excellent molecules obtained by XP are screened using MMGBSA, and the most valuable compounds are confirmed by combining the binding mode and docking score with manual visual inspection.

2.22. Statistical analysis

Data are expressed as mean ± standard error of the mean (SEM) or standard deviation (SD) (SD was used in Fig. 3R–U and Fig. 5K–N). All statistical evaluations were carried out utilizing SPSS 19.0 software. Graphical presentations were executed using GraphPad Prism 10.0 software. The differences between groups were determined by Student's t-test, one-way analysis of variance (ANOVA) or two-way ANOVA followed by Tukey's post-hoc test. P value < 0.05 was established as the threshold for statistical significance.

Fig. 5.

Inhibition of STAT3/Drp1 signaling defends HG-induced oxidative stress and mitochondrial defects. (A) Western blotting of STAT3 protein levels in primary cultured astrocytes, and quantification data. n = 3. (B) Representative images of ROS content by flow cytometry analysis, and quantification data. n = 5. (C-E) MDA, GSH and SOD contents in primary cultured astrocytes. n = 5. (F) Representative images of the morphology of mitochondria by TEM in primary cultured astrocytes. Scale bar, 500 nm. (G) Mitochondrial aspect ratio, n = 3. (H) Mitochondrial cristae density, n = 3. (I) Western blotting of Mfn1, p-Drp1 and Drp1 protein levels in primary cultured astrocytes. (J) Representative images of JC-1 staining by flow cytometry analysis used to assess mitochondrial membrane potential. (K–N) OCR, ATP production, Basal respiration and Maximal respiration in primary cultured astrocytes. n = 5. si-NC (scrambled negative control siRNA); si-STAT3 (si-STAT3 primer). (High glucose, 50 mmol/L treated 48 h). Data are shown as mean ± SEM or SD. Statistical significance was defined as ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001.

3. Results

3.1. GSTM2 expression in astrocytes is significantly downregulated in the hippocampus of DACD mice models

Based on comprehensive behavioral evaluation including the NOR and MWM tests, 24 weeks old db/db mice appeared marked cognitive deficits compared to db/m controls. The complete sequence of these assessments is detailed in the experimental timeline (Fig. 1A). Subsequent assessment via the novel object recognition paradigm provided evidence of cognitive compromise, showing that db/db mice exhibited a lower preference index than controls (Fig. 1B and C). This impairment was particularly evident within the MWM paradigm, where the DACD cohort demonstrated significantly prolonged escape latency, fewer platform crossings and a diminished percentage of time in the target quadrant (Fig. 1D–G). To rule out motor dysfunction as a confounding factor, we confirmed that the total distance traveled of NOR and swimming speeds of MWM for db/db mice did not differ from their control counterparts (Supplementary Fig. S1A and B). Next, employing a comparative proteomic profiling strategy, we analyzed hippocampal tissues from DACD mice relative to controls. This investigation successfully revealed a distinct proteomic signature in the DACD hippocampus, comprising 198 proteins that exhibited increased expression and 107 proteins that showed decreased expression. Volcano plots and cluster analysis showed that antioxidant protein GSTM2 was highly enriched in hippocampal tissue and significantly reduced in DACD mice (Fig. 1H and I). The complete datasets for these proteins with differential abundance is provided in Supplementary Table 1. KEGG pathway enrichment analysis emphasized that these proteins participate in metabolic pathways, glutathione metabolism, mitophagy, valine, leucine and isoleucine degradation et al. (Fig. 1J). Based on recent research, our investigation specifically targeted the protein expression dynamics within the glutathione metabolism pathway, a decision predicated on the well-established role of oxidative stress as a central pathogenic driver in DACD. Therefore, we focused on GSTM2 in subsequent experiments. Immunofluorescence staining and immunoblotting confirmed that GSTM2 protein expression (Fig. 1K and L) and mRNA expression levels (Supplementary Fig. S1C) were significantly reduced in the hippocampus of DACD mice and were mainly in astrocytes (Fig. 1K). A significant reduction in both the protein (Fig. 1M and N) and mRNA (Supplementary Fig. S1D) levels of GSTM2 was observed in primary cultured astrocytes following exposure to a high-glucose environment. In addition, with the increase of high sugar concentration, GSTM2 showed a dose-dependent decrease (Supplementary Fig. S1E), providing direct in vitro evidence for its dysregulation. To ascertain that this effect was attributable to glucose toxicity rather than osmotic stress, a parallel experiment was conducted using mannitol as an osmotic control. The findings confirmed that mannitol treatment at an equivalent concentration did not compromise cell viability or induce LDH release (Supplementary Fig. S1F and G). These in vitro results align with our in vivo observations, where the validity of the DACD mouse model, was rigorously confirmed through weekly monitoring of blood glucose and body weight (Supplementary Fig. S1H and I). Collectively, this compelling evidence, linking the downregulation of GSTM2 to DACD in both cellular and animal models, strongly implicates GSTM2 as a key player in the pathogenesis of DACD.

Fig. 1.

GSTM2 expression in astrocytes is significantly downregulated in the hippocampus of DACD mice models. (A) Experimental paradigms. (B) The represents of the exploration trajectory of db/m and db/db mice in the NOR test. (C) The preference index in the NOR test. n = 10. (D) The represents of the swimming trajectories in the MWM test. (E-G) The escape latency, across platform times and target quadrant retention time. n = 10. (H) Cluster analysis of proteomic sequencing. (I) Volcano plot volcanic map of proteomic sequencing. Red dots are raised, and green dots are decreased. Gstm2 was downregulated in hippocampus of db/db mice compared to db/m mice. (J) KEGG pathway enrichment analysis showed Glutathione metabolism was the key metabolic pathways. (K) Representative images of the cellular localization and expression of GSTM2 by immunofluorescence in astrocytes (GFAP), microglia (Iba-1), neurons (NeuN), oligodendrocytes (MBP) and nucleus (DAPI) of the CA1 region in hippocampus. Scale bar, 50 μm. And quantification data. n = 3. (L) Western blotting of GSTM2 protein expression in hippocampus, and quantification data. n = 3. (M) Representative images of the localization and expression of GSTM2 by immunofluorescence in HG primary cultured astrocytes (50 mmol/L concentration glucose treated for 48 h) compared to control. Scale bar, 50 μm. And quantification data. n = 3. (N) Western blotting of GSTM2 protein expression in primary cultured astrocytes, and quantification data. n = 3. HG (High glucose, 50 mmol/L treated 48 h). Data are shown as mean ± SEM. Statistical significance was defined as ∗∗P < 0.01 and ∗∗P < 0.001.

3.2. Restoration of hippocampal astrocytic GSTM2 ameliorated cognitive function and impairment of neuronal synaptic plasticity in DACD models

Then we elucidate the function of GSTM2 in the hippocampus astrocyte of db/db mice. A targeted genetic intervention was performed on 12 weeks old db/db mice. Using a stereotaxic injection technique, we administered rAAV9 to the hippocampus, which was engineered to drive the overexpression of GSTM2 specifically in astrocytes under the control of the GFAP promoter (rAAV-GSTM2). The control group received a null vector (rAAV-NC) injection. Following the intervention, all mice were housed until they were 24 weeks old. A cohort of behavioral tests was subsequently adopted to evaluate cognitive function. The evaluation protocol included the NOR test followed by the MWM test, conducted at the 21 weeks and 22 weeks respectively. The complete sequence of these assessments is detailed in the experimental timeline (Fig. 2A). The efficiency of GSTM2 overexpression in the hippocampus was validated by Western blotting (Fig. 2B) and Immunofluorescence (Supplementary Fig. S2A). Western blot analysis corroborated the successful intervention, revealing a significant upregulation of GSTM2 protein within the hippocampal tissue of db/db mice following rAAV-GFAP-GSTM2 administration (Fig. 2B). This molecular enhancement translated into a partial, yet discernible, amelioration of the cognitive deficits observed in db/db mice according to their performance both in the NOR and MWM paradigms. This is evidenced by increased in preference index (Fig. 2C and D), shortened escape latency (Fig. 2E and F), increased across-platform times (Fig. 2G) and target quadrant retention time (Fig. 2H). Although total distance traveled of NOR and swimming speeds of MWM for mice did not differ from their control counterparts (Supplementary Fig. S2B and C). Next, we examined the injuring neuronal synaptic plasticity using Golgi staining, TEM and immunoblotting. As shown in Fig. 2I and J, the average length of dendrites was significantly reduced. Additionally, the number of intersection points of concentric circles (Fig. 2K), dendritic spine density (Fig. 2L), and PSD length (Fig. 2M) were all significantly decreased in db/db mice. Targeted delivery of rAAV-GFAP-GSTM2 into the hippocampus yielded partial restoration of neuronal synaptic plasticity in db/db mice. Concomitant with these improvements, a marked upregulation was observed in the protein levels of the presynaptic marker SYP and the postsynaptic marker PSD95 (Fig. 2N). Taken together, the rescued cognitive performance, strongly suggests that the targeted reinstatement of GSTM2 in the hippocampus astrocyte is the underlying driver for the amelioration of both cognitive deficits and impaired synaptic plasticity in DACD models.

Fig. 2.

Restoration of astrocytic GSTM2 ameliorates cognitive dysfunction and neuronal synaptic plasticity impairment in DACD mice models. (A) Experimental paradigms. (B) Western blotting of GSTM2 protein expression in hippocampus, and quantification data. n = 3. (C) The represents the exploration trajectory of db/m and db/db mice in the NOR test. (D) The preference index in the NOR test. n = 10. (E) The represents of the swimming trajectories in the MWM test. (F–H) The escape latency, across platform times and target quadrant retention time. n = 10. Finally, to investigate the underlying neuronal mechanisms of the hippocampus of db/m and db/db mice with or without GSTM2 overexpression, morphological analyzes were conducted. (I) Representative images of the neuronal dendritic architectures by Golgi staining. Scale bar, 20 μm. (J) Quantification data of the average length of dendrite in Fig. 2I n = 10. (K) Quantification data of intersection points of concentric circles in Fig. 2I n = 10. (L) Representative images of the high-resolution micrographs of dendritic spines by Golgi staining. Scale bar, 10 μm. And quantification data of dendritic spine density. n = 9. (M) Representative images of the synaptic deficits at a subcellular level by TEM in hippocampus. Scale bar, 500 nm. And quantification data of postsynaptic density (PSD) length. n = 9. (N) Western blotting of PSD95 and synaptophysin (SYP) protein expression in hippocampus, and quantification data. n = 3. rAAV-NC (rAAV-GFAP-EGFP-WPRE-hGH pA); rAAV-GSTM2 (rAAV-GFAP-GSTM2-EGFP-WPRE-hGH pA). Data are shown as mean ± SEM. Statistical significance was defined as ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001.

3.3. GSTM2 inhibits oxidative stress and mitochondrial defects in DACD

GSTM2 is robustly expressed within astrocytes. Astrocytes are dependent on the energy provided by mitochondria to execute their regular functions, and disruptions in mitochondrial function can precipitate oxidative stress, which subsequently impairs mitochondrial integrity. To explore whether GSTM2 affects mitochondrial function, we assessed oxidative stress and mitochondrial morphology to study the role of astrocytic GSTM2 overexpression in DACD.

We identified significant mitochondrial abnormalities within the hippocampal astrocytes of db/db mice, primarily manifesting as cristae loss and cellular swelling using TEM analysis. The forced expression of GSTM2 by rAAV-GFAP-GSTM2 markedly attenuated both the swelling and the degeneration of cristae in db/db mice while yielded no discernible changes in control db/m mice (Fig. 3A–C). db/db mice exhibited marked oxidative stress in hippocampus compared to db/m mice, while GSTM2 overexpression significantly reduced oxidative stress, as evidenced by decreased MDA levels (Fig. 3D) and increased GSH (Fig. 3E) and SOD levels (Fig. 3F). To elucidate the mitochondrial protective capacity of GSTM2 in astrocytes, an in vitro model was established by exposing primary cultured astrocytes to high glucose (HG 50 mmol/L, 48h) milieu. The astrocytes were transfected with plasmid engineered to over express GSTM2 (OE-GSTM2). Western blotting showed that GSTM2 was successfully over expressed in primary cultured astrocytes treated with or without HG (Fig. 3G). Then, we assessed oxidative stress levels, which were significantly increased after HG treatment. GSTM2 overexpression significantly suppressed oxidative stress, as evidenced by a marked decrease in ROS levels (Fig. 3H) and MDA content (Fig. 3I), and a significant increase in GSH (Fig. 3J) and SOD levels (Fig. 3K) compared to vector control. TEM analysis revealed abnormal mitochondria morphology in HG cultured astrocytes, characterized by cristae loss and swelling. GSTM2 overexpression significantly reduced mitochondrial swelling and cristae loss in HG treated astrocytes (Fig. 3L–N). Furthermore, immunoblotting was employed to assess the expression of critical mediators of mitochondrial dynamics. In contrast to the control group, those astrocytes exposed to high glucose exhibited a pronounced imbalance, characterized by significantly elevated phosphorylation of Drp1(p-Drp1/Drp1) and diminished expression of Mfn1. Notably, the OE-GSTM2 effectively counteracted these alterations, restoring Mfn1 levels and suppressing Drp1 phosphorylation (Fig. 3O). Moreover, we performed JC-1 staining to assess the effect of GSTM2 on mitochondrial membrane potential. GSTM2 significantly enhanced the mitochondrial membrane potential levels in HG cultured astrocytes (Fig. 3P and Q). To further elucidate the functional consequences of GSTM2 on mitochondrial bioenergetics, we conducted a comprehensive assessment of the cellular oxygen consumption rate. HG treatments were found to profoundly disrupt key mitochondrial respiratory parameters, such as ATP synthesis, basal respiration, and maximal respiratory capacity. However, these detrimental effects were successfully mitigated by the overexpression of GSTM2 (OE-GSTM2) (Fig. 3R–U). In summary, these data robustly supported that reinstating GSTM2 expression in astrocytes could ameliorate the mitochondrial defects and oxidative stress responsible for the DACD pathogenesis.

Fig. 3.

GSTM2 alleviates oxidative stress and mitochondrial defects in DACD models. (A) Representative images of the morphology of astrocytes mitochondria by TEM in hippocampus. Scale bar, 1.0 μm. (B) Mitochondrial aspect ratio, n = 3. (C) Mitochondrial cristae density, n = 3. (D-F) MDA, GSH and SOD contents in hippocampus. n = 5. (G) Western blotting of GSTM2 protein expression in primary cultured astrocytes, and quantification data. n = 3. (H) Representative images of ROS content by flow cytometry analysis, and quantification data. n = 5. (I–K) MDA, GSH and SOD contents in primary cultured astrocytes. n = 5. (L) Representative images of the morphology of primary astrocytes mitochondria by TEM. Scale bar, 500 nm. (M) Mitochondrial aspect ratio, n = 3. (N) Mitochondrial cristae density, n = 3. (O) Western blotting of Mfn1, p-Drp1 and Drp1 protein levels in primary cultured astrocytes, and quantification data. n = 3. (P, Q) Representative images of JC-1 staining by flow cytometry analysis used to assess mitochondrial membrane potential. (R–U) OCR, ATP production, Basal respiration and Maximal respiration in primary cultured astrocytes. n = 5. rAAV-NC (rAAV-GFAP-EGFP-WPRE-hGH pA); rAAV-GSTM2 (rAAV-GFAP-GSTM2-EGFP-WPRE-hGH pA). Vector (pEGFP-C1 plasmid empty vector); OE-GSTM2 (pEGFP-C1-GSTM2 plasmid vector). HG (High glucose, 50 mmol/L treated 48 h). Data are shown as mean ± SEM or SD. Statistical significance was defined as ∗P < 0.05, ∗∗P < 0.01 and ∗∗P < 0.001.

To confirm the contribution of astrocytic GSTM2 to the pathogenesis of DACD, an in vivo knockdown model was established. For this purpose, 12 weeks old DACD mice received a stereotaxic injection into the hippocampus of rAAV-GFAP-shGSTM2 or null control, driven by the GFAP promoter to achieve astrocyte-specific targeting. Following the intervention, all mice were housed until they reached 24 weeks old. To evaluate cognitive outcomes, we subjected the animals to the NOR test at 21 weeks old, followed by the MWM test one week later (Fig. S3A). Western blotting confirmed the significant reduction of GSTM2 expression in the hippocampus (Fig. S3B). Mice administered rAAV-GFAP-shGSTM2 demonstrated varying deterioration of cognitive impairments, especially in the db/db mice injection with rAAV-GFAP-shGSTM2 mice. The aggravates cognition were marked decrease in preference index (Fig. S3C–E) and evident in significantly delayed escape latency (Fig. S3F–H), decreased across-platform times (Fig. S3I), reduced target quadrant retention time (Fig. S3J). Total distance traveled of NOR and swimming speeds of MWM for mice did not differ from their control counterparts (Fig. S3D and H). TEM results also revealed hippocampal astrocytes in mice administered with rAAV-GFAP-shGSTM2 exhibited varying degrees of mitochondrial abnormalities, characterized by cristae loss and swelling, with the most severe mitochondrial damage observed in the db/db mice injection with rAAV-GFAP-shGSTM2(Fig. S3K–M). Subsequently, we measured oxidative stress levels in hippocampus. Similarly, mice administered with rAAV-GFAP-shGSTM2 exhibited varying degrees of oxidative stress, manifested as increased MDA levels (Fig. S3N) and reduced GSH (Fig. S3O) and SOD (Fig. S3P) levels. All above results indicated that knockdown of GSTM2 aggravates cognitive function and mitochondrial defects in DACD mice models.

3.4. GSTM2 suppresses the phosphorylation of STAT3

To elucidate the molecular mechanisms underlying the role of GSTM2 in DACD pathogenesis, IP-MS was employed to identify GSTM2 interacting proteins. By using GSTM2 antibody to precipitate the corresponding protein-protein complexes in primary astrocytes, we separated and purified the proteins, and identified proteins interacting with GSTM2 by mass spectrometry (MS) (Fig. 4A). IP-MS results supported a comprehensive profile of GSTM2-binding proteins, which was then interrogated with emphasis on candidates associated with mitochondrial defects (Supplementary Table 2). The findings indicated that STAT3 was the most prominent in the list of precipitated proteins, as determined by the intensity of the FC, which is linked to mitochondrial defects (Fig. 4B). To confirm the direct interaction between GSTM2 and STAT3, we used SPR analysis to detect the affinity with the KD equals to 112 nM (Fig. 4C). Moreover, Co-IP results indicated GSTM2 interacts with STAT3 in primary astrocytes treated with or without HG treatment (Fig. 4D and E). Immunofluorescence consistently showed colocalization of GSTM2 and STAT3 in astrocytes (Fig. 4F). These results indicated that under normal conditions, there exist direct interaction between GSTM2 and STAT3. According to recent study, Ser727 phosphorylation of STAT3 is closely associated with its mitochondrial defects [[21], [22], [23]]. Therefore, we analyzed STAT3 phosphorylation levels in HG cultured primary astrocytes transfected with si-GSTM2 primer (si-GSTM2) or scrambled siRNA (si-NC). Compared with the si-NC group, si-GSTM2 significantly increased STAT3 phosphorylation in Ser727 sites. Conversely, OE-GSTM2 significantly inhibited STAT3 Ser727 phosphorylation (Fig. 4G). Futhermore, we investigated the influence of GSTM2 on the subcellular localization of p-STAT3. Compared with the si-NC group, si-GSTM2 significantly increased nuclear STAT3 levels under HG condition. In contrast, OE-GSTM2 significantly inhibited nuclear STAT3 levels (Fig. S4A). In addition, compared with the control group(si-NC or Vector), si-GSTM2 and OE-GSTM2 significantly inhibited cytoplasmic p-STAT3 under HG condition(Fig. S4B).

Fig. 4.

GSTM2 suppresses the phosphorylation of STAT3. (A) Experimental paradigms. (B) Top 15 list of GSTM2-binding proteins identified through GSTM2 IP-MS analysis in primary cultured astrocytes. STAT3 was identified with high level of interaction intensity. (C) Detection of the binding affinity between GSTM2 and STAT3 by SPR method. (D-E) Co-IP analysis for validation of the endogenous interaction between GSTM2 and STAT3 in primary cultured astrocytes. (F) Representative images of the cellular localization and expression of GSTM2 and STAT3 by immunofluorescence in primary cultured astrocytes. Scale bar, 50 μm. GSTM2 (red), STAT3 (green) and DAPI (blue). (G) Western blotting of p-STAT3(Ser727), STAT3 and GSTM2 protein levels in primary cultured astrocytes, and quantification data. n = 3. si-NC (scrambled negative control siRNA); si-GSTM2 (si-GSTM2 primer). Vector (pEGFP-C1 plasmid empty vector); OE-GSTM2 (pEGFP-C1-GSTM2 plasmid vector). HG (High glucose, 50 mmol/L treated 48 h). Data are shown as mean ± SEM. Statistical significance was defined as ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001.

3.5. Inhibition of STAT3/Drp1 signals ameliorates HG-induced oxidative stress and mitochondrial defects

Drp1 is identified as the key target downstream of STAT3, thereby we positioned Drp1 as a potential mediator in the pathogenic cascade of GSTM2-STAT3 pathway. Additionally, we have previously confirmed that overexpression of GSTM2 significantly downregulated Drp1. Therefore, we hypothesized that the GSTM2-STAT3 pathway drives mitochondrial defects by Drp1 targeting. We analyzed mitochondrial function and oxidative stress in primary astrocytes transfected with si-STAT3 primer (si-STAT3) and scrambled negative control siRNA (si-NC) with HG treatment (Fig. 5A). The results showed that STAT3 knockdown effectively suppressed oxidative stress levels, as evidenced by a significant decrease in ROS levels (Fig. 5B) and MDA content (Fig. 5C), and a significant increase in GSH (Fig. 5D) and SOD levels (Fig. 5E). TEM analysis results showed that STAT3 knockdown significantly reduced mitochondrial swelling and cristae loss in astrocytes cultured in HG conditions (Fig. 5F–H). STAT3 knockdown also inhibited Drp1 phosphorylation levels and promoted Mfn1 expression (Fig. 5I). JC-1 staining results showed that STAT3 knockdown significantly rescued mitochondrial membrane potential levels in HG treated astrocytes (Fig. 5J). Furthermore, we measured cellular oxygen consumption rate (OCR) (Fig. 5K) and found that STAT3 knockdown increased ATP production (Fig. 5L) basal respiration (Fig. 5M) and maximal respiration (Fig. 5N). Collectively, these results indicate that STAT3 promotes oxidative stress and mitochondrial defects by influencing mitochondrial quality control, especially enhancing Drp1 signals.

Since GSTM2 protective effects relay on STAT3-Drp1 pathway inhibition, we investigated whether pharmacological intervention of STAT3 activation could also influence the GSTM2 function in DACD. db/db mice were administered the STAT3 inhibitor stattic via intraperitoneal injection alone while STAT3 activators colivelin TFA (C-TFA) were administered via intraperitoneal injection in mice that had already received rAAV-GSTM2 interfered. NOR and MWM test results showed both of stattic and GSTM2 overexpression improved the cognitive impairments, such as significant increase preference index (Fig. S5A–C) and shortened escape latency (Fig. S5D–F), increased across platform times (Fig. S5G) and target quadrant retention time (Fig. S5H). Furthermore, we observed the injury of neuronal synaptic plasticity by Golgi staining, TEM and immunoblotting analysis. The results showed that stattic and rAAV-GSTM2 significantly protected neuronal synaptic plasticity, as evidenced by increased values for average length of dendrites (Fig. S5I and J), intersection points of concentric circles (Fig. S5K), dendritic spine density (Fig. S5L) and PSD length (Fig. S5M). Additionally, the expression of presynaptic SYP and postsynaptic PSD95 were significantly increased (Fig. S5N). However, C-TFA partially reversed the protective effect of rAAV-GSTM2. In conclusion, these above confirm that the involvement of STAT3 was integral to the ameliorative effects of GSTM2 overexpression on neuronal synaptic plasticity and cognitive processes.

3.6. Icariin improved cognitive function through the GSTM2/STAT3 axis in DACD

As a prospective therapeutic target for DACD, the activation of GSTM2 confers advantages. We accepted the efficient computational method virtual screening (VS) to identify potential GSTM2 activators from large-scale compound libraries based on molecular docking and binding energy assessment. The complete sequence of these assessments is detailed in the experimental timeline (Fig. 6A). The compound library primarily consists of a total of 20,723 traditional Chinese medicine monomers, bioactive compounds and antioxidant compounds in this study. The data were ranked based on energy scores and the top 284 compounds were selected after docking. Furthermore, 30 compounds were ultimately identified through MMGBSA, binding energy scoring and assessment of key residues at the active site (Fig. 6B–Supplementary Table 3). Based on docking scores, Tubuloside A, Icariin, and Militarine were listed as candidate compounds (Fig. 6C–F). Among these three, Icariin is reported to have better blood-brain barrier permeability and to enhance cognitive function impaired by various diseases. Therefore, we chose Icariin for further exploration. A 3-month intervention with Icariin through oral gavage was performed on db/db mice starting at 12 weeks of age. Immunoblotting results showed that Icariin significantly upregulates the expression level of GSTM2 and inhibited STAT3 phosphorylation levels (Fig. 6G). The therapeutic efficacy of Icariin against cognitive decline in db/db mice. Icariin administration markedly ameliorated impairments in cognitive function. The effects were evidenced by significant increase in the preference index (Fig. 6H and I) and shortened escape latency (Fig. 6J and K), increased across-platform times (Fig. 6L) and target quadrant retention time (Fig. 6M). Total distance traveled of NOR and swimming speeds of MWM for mice did not differ from their control counterparts (Supplementary Fig. S6A and B). TEM analysis further indicated that Icariin treatment reduced mitochondrial cristae loss and swelling in hippocampal astrocytes from mice (Fig. 6N–P). Finally, we assessed oxidative stress levels in the hippocampus of db/db mice. Similarly, oxidative stress in hippocampal tissue from mice were gavaged with Icariin was significantly reduced, as decreased MDA levels (Fig. 6Q) and increased GSH (Fig. 6R) and SOD (Fig. 6S) levels. These results indicate Icariin regulates the GSTM2/STAT3 axis to successfully improve cognitive function and mitochondrial defects in DACD models.

Fig. 6.

Icariin improves cognition through the GSTM2/STAT3 axis in DACD mice models.

(A) Experimental paradigms. (B) Schematic representation of the virtual screening strategy utilized to discover potential small-molecule binding to GSTM2 protein. (C) Binding affinities of Tubuloside A, Icariin and Militarine by docking scores. (D) Visualization of the predicted binding pose of Tubuloside A docked to the GSTM2 protein. (E) Visualization of the predicted binding pose of Icariin to the GSTM2 protein. (F) Visualization of the predicted binding pose of Militarine to the GSTM2 protein. (G) Western blotting of p-STAT3(Ser727), STAT3 and GSTM2 protein levels in hippocampus, and quantification data. n = 3. (H) The represents of the exploration trajectory of db/m and db/db mice in the NOR test. (I) The preference index in the NOR test. n = 10. (J) The represents of the swimming trajectories in the MWM test. (K-M) The escape latency, across platform times and target quadrant retention time. n = 10. (N) Representative images of the morphology of astrocytes mitochondria by TEM in hippocampus. Scale bar, 500 nm. (O) Mitochondrial aspect ratio, n = 3. (P) Mitochondrial cristae density, n = 3. (Q-S) MDA, GSH and SOD contents in hippocampus. n = 5. Vehicle (gavage, 20% PEG400 solution); Icariin (gavage, 100 mg/kg/day). Data are shown as mean ± SEM. Statistical significance was defined as ∗P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001.

4. Discussion

In this study, we firstly revealed the GSTM2 protects synapse plasticity and mitochondrial function through the STAT3-Drp1 signaling pathway, further ameliorating DACD pathogenesis. Moreover, we found Icariin could activate GSTM2-STAT3 pathway to improve cognitive impairment in DACD mice models. Importantly, the therapeutic potential of targeting this axis is demonstrated by the finding that cognitive impairment in diabetes can be effectively rescued through either GSTM2 overexpression, direct STAT3 inhibition or treatment with Icariin, all of which converge on the amelioration of mitochondrial defects via Drp1 suppression. Therefore, this study broadens our understanding of DACD progression, and maintaining astrocytes GSTM2 protein levels may serve as a promising therapeutic strategy for DACD.

As a cornerstone of the body's phase II detoxification system, the glutathione S-transferase superfamily orchestrates the conjugation of glutathione with a diverse array of xenobiotics and endogenous metabolites. A critical subgroup within this family, the Mu-class GSTs, is particularly distinguished by its capacity to neutralize free radicals and peroxides. The neuroprotective capacity of GSTM2 is evidenced by its role in shielding dopaminergic neurons, particularly those containing neuroleukopigment, from degeneration; this protection is mediated through exosomes secreted by astrocytes [21,22]. Furthermore, the cytoprotective function of GSTM2 extends to the preservation of subcellular integrity, as it safeguards glioblastoma cells from aminopyrine-induced toxicity by preventing mitochondrial defects, autophagy, and lysosomal dysfunction [14,15,23]. Given this established ability of GSTM2 to mitigate neuronal and cellular damage across different pathological contexts, it is highly potentiated that it exerts a protective influence on the progression of DACD. A significant reduction in the protein levels of GSTM2 originating from astrocytes was identified for the first time in our study. GSTM2 downregulation was consistently observed in vivo and in vitro models: the hippocampal tissue of db/db mice and primary astrocytes subjected to hyperglycemic conditions. The discovery, substantiated by an integrated approach combining proteomics sequencing with both in vitro and in vivo validations, highlights GSTM2 as a key factor altered in the diabetic milieu.

Astrocytes, the most abundant cell type in the CNS, form tri-synaptic structures with neurons and extensively surround synapses [24]. Under normal physiological conditions, astrocytes promote synaptic formation, maturation, and pruning by releasing glial neurotransmitters and synaptogenic molecules. In diabetic conditions, astrocytes become activated as reactive astrocytes, undergoing morphological and functional changes [25]. A phenotypic switch occurs in which these entities relinquish their protective role in synapse formation and instead develop potent neurotoxic properties, a transformation that ultimately leads to the impairment of neuronal synaptic plasticity [26,27]. This study found that GSTM2 is primarily expressed in astrocytes, and overexpression of GSTM2 in astrocytes effectively improved neuronal synaptic plasticity damage and cognitive decline. Additionally, knockdown of GSTM2 in astrocytes exacerbated cognitive dysfunction and synaptic plasticity damage induced by DACD. Interestingly, a prior study reported that under 25 or 2 mM glucose culture conditions, GLP-1R expression significantly increased, astrocyte-specific knockout of GLP-1R enhanced glucose uptake, improving systemic glucose homeostasis and memory function [28]. This seems to present a different perspective from the central role of GSTM2 downregulation in DACD. Unlike this, in this study, GSTM2 expression was significantly reduced at 50 mM glucose. Based on the antioxidant function of GSTM2, overexpression of GSTM2 effectively improved mitochondrial function in astrocytes. Furthermore, consistent with this study, multiple studies have shown that GLP-1R agonists can significantly improve DACD [29]. This apparent difference may be associated with different glucose concentrations, or signaling contexts. The pathogenesis of DACD is intimately linked to oxidative stress, a deleterious process counteracted in part by the antioxidant protein GSTM2. In this study, overexpression of GSTM2 in hippocampus astrocytes suppressed abnormalities in oxidative stress-related indicators, such as increased GSH and SOD and decreased ROS and MDA.

Elevated levels of reactive oxygen species (ROS) are a primary instigator of oxidative stress and subsequent cell death, constituting a key pathogenic factor in diabetes [30,31]. A principal target of this oxidative damage is the mitochondrion, whose functionality can be compromised through multiple pathways [32]. Given that mitochondrial defects is a critical contributor to the pathogenesis of DACD [33], it is noteworthy that this impairment is intricately linked to the regulatory processes of mitochondrial dynamics, encompassing both fission and fusion events. Mitochondrial fission is regulated by mitochondrial dynamic proteins such as Drp1, while mitochondrial fusion is primarily controlled by Mfn2. The integrity of the mitochondrial network is critically undermined by excessive fission, a pathological process that precipitates massive mitochondrial rupture, a surge in reactive oxygen species (ROS) production, and a consequent exacerbation of mitochondrial defects [24,25]. Ultimate balance between these opposing fission and fusion forces is critically governed by GST, which functions as a key intracellular regulator of mitochondrial function [26]. Although there are few reports on GSTM2 regulating mitochondrial function, there are numerous reports on GSTM2 isoenzymes having important regulatory roles in mitochondrial function, such as GSTA4 [27] and GST-14 [28]. Therefore, this study further examined mitochondrial function following GSTM2 overexpression. The study found that GSTM2 overexpression inhibited mitochondrial swelling and cristae loss in astrocytes cultured under high-glucose conditions, while also suppressing Drp1 phosphorylation levels and mitochondrial membrane potential levels, increasing Mfn1 expression, and elevating cellular oxygen consumption rates.

To elucidate the downstream signaling cascades through which GSTM2 exerts its critical influence on mitochondrial defects in DACD astrocytic cells, we conducted an unbiased proteomic screening using immunoprecipitation followed by mass spectrometry (IP-MS). Among the candidate molecules identified, STAT3 stood out as a potential key mediator of GSTM2's pathogenic effects. This molecule is of particular interest due to its canonical function as a signaling transducer that integrates diverse pathological stimuli, acting as a crucial conduit for transmitting signals from the cellular periphery and cytoplasm to the nucleus to modulate gene expression. Recent studies have extensively explored the critical role of STAT3 in diabetes and its complications, such as diabetic nephropathy [29], diabetic cardiomyopathy [30], and DACD. Persistent STAT3 activation is closely associated with DACD [31,32]. STAT3 activation is mediated by tyrosine residue(Tyr705) phosphorylation of Janus kinase membrane recruitment or serine residue(Ser 727) phosphorylation of the Ras-MAPK-ERK cascade. Multiple studies have demonstrated that Ser727 phosphorylation of STAT3 is required for mitochondrial function and oxidative stress signals, whereas Tyr705 phosphorylation of STAT3 mainly mediates nuclear transcriptional activity and inflammation signals [33,34]. In this study, we explored the effects of GSTM2 on STAT3 Ser727 phosphorylation relay on mitochondrial function and oxidative stress pathway. While, recent research has showed Serinc2 could promote mitochondrial bioenergetics by simultaneously regulating the phosphorylation of STAT3 on both Tyr705 and Ser727 in doxorubicin-induced cardiotoxicity models [35]. In the present work, we provide direct experimental validation for this hypothesis by demonstrating that GSTM2 physically interacts with STAT3. This direct binding serves to alter the phosphorylation status of STAT3 and ultimately precipitates significant mitochondrial defects. Similarly, treatment with si-STAT3 and Stattic reduces STAT3 phosphorylation, while the STAT3 activator (C-TFA) partially reverses the protection. Although C-TFA has been reported to exhibit neuroprotective effects, this may be due to different disease models, different stages of onset, and the different roles of STAT3 in the disease. In the LPS induced cognitive impairment mouse model, C-TFA partially reversed the neuroprotective effect of Indole 3-propionic acid (IPA) [18]. In addition, in the model of diabetes cardiomyopathy in mice, C-TFA significantly inhibited the myocardial protective effect of Dapagliflozin (DAPA) on db/db mice [36]. This study also indicates that Drp1 is a key effector in GSTM2-STAT3-mediated mitochondrial defects. STAT3 activation increases Drp1 phosphorylation levels and mitochondrial fragmentation. Finally, we identified Icariin as a potential GSTM2 activator through virtual screening technology. Relevant literature confirms that Icariin has certain neuroprotective potential against DACD [37]. In addition, multiple experiments have confirmed that icariin can cross the blood-brain barrier and enter the hippocampus to exert antioxidant and anti-inflammatory effects [38,39], this provides an experimental basis for this study. In this study, we found that Icariin could improve DACD through the GSTM2-STAT3 axis. This process primarily relies on the upregulation of icariin on the GSTM2 expression and STAT3 phosphorylation inhibition, thereby ameliorating cognitive impairment by reducing mitochondrial dysfunction and oxidative stress levels.

Although the present study has elucidated the relationship between the GSTM2-STAT3-Drp1 axis and DACD for the first time, subsequent research should conduct a more in - depth exploration of its underlying potential mechanisms. Moreover, the specific mechanism of interaction between Icariin and GSTM2 also need further investigation.

5. Conclusion

Our investigation delineates a novel mechanistic axis whereby GSTM2 exerts its protective effects against astrocytes mitochondrial defects. We demonstrate that GSTM2 achieves this by directly engaging with STAT3, an interaction that effectively curtails STAT3's phosphorylation. The disruption of STAT3 signaling subsequently results in the suppression of Drp1 phosphorylation levels. Consequently, by establishing this GSTM2-STAT3-Drp1 regulatory cascade, our work not only solidifies the critical role of GSTM2 in the pathogenesis of DACD but also illuminates this pathway as a promising and previously unexplored avenue for therapeutic intervention in this debilitating condition.

CRediT authorship contribution statement

Wenqiang Liu: Conceptualization, Data curation, Writing – original draft. Yufei Wang: Data curation, Methodology, Software, Writing – original draft. Yunshuang Zhao: Data curation, Methodology, Visualization. Bingxue Song: Data curation, Formal analysis, Software. Linqin Luo: Methodology, Software, Validation. Khan Muhammad Zahir: Software, Visualization. Miaomiao Song: Formal analysis, Methodology. Yong Wang: Investigation, Methodology. Li Zhang: Methodology, Visualization. Zhongfu Zuo: Project administration, Supervision, Writing – review & editing. Lu Yao: Project administration, Supervision, Writing – review & editing.

Declaration of competing interest

The authors have declared no conflict of interest.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.