Male C57BL/6 mice (6–8 weeks old) were procured from Shanghai Lingchang Biotechnology and housed under a controlled temperature (21–23°C) with a 12-h light/dark cycle. All the animal protocols were performed according to the Guide for the Care and Use of Laboratory Animals issued by the National Institutes of Health (NIH) and were approved by the Animal Welfare and Ethics Group of the Department of Laboratory Animal Science of Fudan University (2020-Huashan hospital-JS190).

Cohort 1: Mice (6–8 weeks old) were assigned to either a normal diet (ND, n = 15) or high-fat diet (HFD, n = 30) for 5 months. Glucose levels were measured, and mice with fasting blood glucose ≥8.0 mmoL/L were classified as T2DM. (Roche, Switzerland). T2DM mice were randomly assigned to either the HFD (n = 14) or SP (n = 14) group, receiving either a vehicle (50% PEG-400 + 0.5% Tween-80 + 49.5% H₂O) (Syed et al., 2018) or SP (30 mg/kg/d) in drinking water for 1 month, respectively. The groups were designated as ND, HFD, and SP.

Cohort 2: Mice (6–8 weeks old) were fed either ND (n = 15) or HFD (n = 80) for 5 months. HFD mice randomly received bilateral hippocampal injections of AAV carrying either CAIII shRNA (HFD + AAV-CAIII sh, n = 40) or control shRNA (HFD + AAV-Con, n = 40) via stereotaxic surgery. CAIII knockdown was confirmed by real-time PCR and western blotting 3 weeks post-injection. Mice that received either AAV injection were subsequently randomized to receive either vehicle or SP (30 mg/kg/d) in drinking water for 1 month. Body weight, fasting blood glucose, and daily water/food intake were monitored (Supplementary Figure S1).

Mice were fasted for 12 h to test fasting blood glucose (FBG). For the intraperitoneal glucose tolerance test (IPGTT), mice underwent a 14-h fast followed by an intraperitoneal injection of 20% glucose (2 mg/kg body weight) in PBS. Blood glucose levels were measured at 0, 30, 60, 90, and 120 min post-injection, and the area under the curve (AUC) was calculated. For the intraperitoneal insulin tolerance test (IPITT), mice were fasted for 5 h before receiving an intraperitoneal insulin injection (1 U/kg body weight, Novo Nordisk, Denmark). Blood glucose was measured at 0, 30, 60, 90, and 120 min after injection, and AUC was calculated.

SP was synthesized by the Institute of Organic Chemistry, Chinese Academy of Sciences based on a previous study (Nelson et al., 2017). The molecular structures of 9-PAHSA and SP are illustrated in Figure 1A.

The Morris water maze (MWM) was used to evaluate mouse spatial learning and memory. The latency on the training days (d 1 to d 5) was defined as the first time to reach the platform within 60 s. Mice unable to find the platform within 60 s were gently guided to the platform for 15 s. On day 6 (probe trial day), the platform was removed, with trajectories recorded by the video and analyzed by EthoVision software (version 8.5).

Y maze was utilized to assess the short-term spatial working memory of mice. Mice were positioned at the end of the designated arm (start arm) and allowed to explore freely for 8 min. The course was recorded via video, with subsequent analysis performed by EthoVision software (version 14.0).

Concentrations of SP in tissues were determined using ultra-high performance liquid chromatography (UPLC-MS/MS). The procedure was conducted by Maixi Medical Technology Co. according to the instructions.

Mice from four groups (n = 4 per group) were perfused with PBS followed by 4% paraformaldehyde. The brains were postfixed in the same fixative overnight at 4°C. After progressively dehydrated, embedded in paraffin, and sliced into 4 μm-thick sections, brain morphology was observed by Hematoxylin and Eosin (HE) and Nissl staining as previously described (Yu et al., 2023). Sections were incubated with a primary antibody against neuronal nuclei (NeuN, 1:200 dilution, Beyotime, China) at 4°C overnight, followed by PBS washing. Then, a Cy3-conjugated secondary antibody (1:500, Servicebio, China) was applied for 1 h at room temperature. Fluorescence was detected under a fluorescence microscope (Nikon, Japan).

LDH (Beyotime, China), superoxide dismutase (SOD, Dojindo, Japan), catalase (CAT), glutathione peroxidase (GSH-Px), and hydrogen peroxide (H₂O₂) concentrations were measured by respective ELISA kits (Nanjing Jiancheng Biological Engineering Research Institute, China), following the provided protocols.

Recombinant AAV vectors encoding CAIII shRNA (5′-GGTTCACTGGAATCCAAAGTA-3′) or non-specific control shRNA (5′-CGCTGAGTACTTCGAAATGTC-3′) were synthesized by Shanghai Genechem Co., Ltd. Mice were anesthetized with 1.2% isoflurane and placed on a stereotaxic apparatus (RWD, China) with a heating pad to maintain body temperature at 37°C. A cranial incision was created by an electric drill (RWD, China), followed by the precise injection of 2 × 10^9 v.g. AAV into each hippocampus (AP: −1.85 mm, ML: ±1.60 mm, DV: −1.90 mm) at 0.08 μL/min via a glass micropipette. The syringe (Gaoge, China) remained for 10 min before being slowly withdrawn. After bilateral injections, the incision was sutured, and the mice were kept warm until recovery.

PC12 cells, obtained from Shanghai Fuheng Biotechnology, were authenticated before use through standard cell line identification methods. The CAIII knockdown (CAIII sh), CAIII overexpressing (CAIII OE), and negative control (NC) PC12 cells were transfected with lentiviral vectors. Cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum, 100 μg/mL penicillin, 100 μg/mL streptomycin, and 1 μg/mL puromycin, and maintained in a humidified 5% CO₂ incubator at 37°C. Previous studies indicated that PC12 cell growth was inhibited by elevated glucose and palmitic acid levels. A combination of 100 mM glucose and 200 μM palmitic acid (G100F200) was identified as the threshold for significant growth inhibition, establishing an in vitro diabetes model. The CAIII sh, CAIII OE, and NC cells were exposed to four conditions for 24 h: normal culture medium (Con group), high-glucose/high-lipid medium (G100F200 group), solvent control (DMSO group), and 60 μM SP (SP group).

Total RNA was extracted from PC12 cells using the RNA Extraction Kit (Takara, Japan) according to the manufacturer’s protocol. RNA quantity and integrity were evaluated with a NanoDrop ND-2000/Qubit 2.0, and Agilent 4,200 TapeStation (Agilent Technologies, USA). Only high-quality samples were used for transcriptome sequencing. Double-stranded cDNA libraries were constructed using the TruSeq® RNA Sample Preparation Kit (Illumina, USA), following the recommended protocol. In brief, mRNA containing polyadenylated tails was enriched using magnetic beads conjugated with poly-T oligonucleotides. The resulting libraries were quantified with a Qubit® 2.0 Fluorometer (Life Technologies, USA), and library quality—including fragment size and molar concentration—was assessed using the Agilent 2,100 Bioanalyzer (Agilent Technologies, USA). Sequencing clusters were generated on a cBot system after diluting the libraries to 10 pM, and paired-end sequencing was performed on the Illumina HiSeq X Ten platform (Illumina, USA). All steps related to library preparation and sequencing were conducted at Shanghai Biotechnology Corporation. Raw sequencing reads were initially subjected to quality control, during which rRNA sequences, adapter contaminants, short fragments, and low-quality reads were removed. Clean reads were then aligned to the human reference genome (GRCh38) using HISAT2 (version 2.0.4) (Kim et al., 2015), allowing up to two mismatches. Transcript assembly and quantification were performed with StringTie (version 1.3.0) (Pertea et al., 2015; Pertea et al., 2016) using the corresponding gene annotation to calculate FPKM (Fragments Per Kilobase of transcript per Million mapped reads) values for annotated genes. Differential gene expression was assessed using edgeR (Robinson et al., 2010), followed by functional annotation using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and Gene Ontology (GO) enrichment analyses. The significance threshold for multiple comparisons was controlled using the false discovery rate (FDR) method. Fold change values were calculated based on FPKM levels in individual samples. Genes were considered differentially expressed if they met both criteria: FDR ≤ 0.05 and fold change≥2. Data processing was conducted using R software (version 4.3.1).

Cell viability was measured by CCK-8 assay kit (Dojindo, Japan). The CAIII sh, CAIII OE, and NC PC12 cells were plated at 4000 cells/well in a 96-well plate and cultured for 24 h. They were then treated with normal culture medium (Con group), high-glucose and high-fat culture medium (100 mM glucose +200 μM palmitic acid, G100F200 group), solvent control (DMSO group), and 60 μM SP (SP group) for 24 h. Following treatment, 10 μL of CCK8 reagent was added to each well and incubated for 2 h in the dark. The absorption was determined at 450 nm using a microplate reader (Biotek, USA).

Apoptosis was analyzed by flow cytometry using dual staining with Annexin V-FITC (to detect early apoptotic cells) and propidium iodide (PI) (to identify late apoptotic and necrotic cells) (Thermo Scientific, USA). The quadrants were interpreted as follows: Q1 (upper left): PI⁺/Annexin V⁻ (necrotic cells), Q2 (upper right): PI⁺/Annexin V⁺ (late apoptotic cells), Q3 (lower right): PI⁻/Annexin V⁺ (early apoptotic cells), and Q4 (lower left): PI⁻/Annexin V⁻ (viable cells).

Mitochondrial membrane potential was assessed by TMRE (tetramethylrhodamine ethyl ester) staining in PC12 cells seeded at 5000 cells/well on 12-well plates with coverslips and incubated for 24 h. Following treatment with normal culture medium, G100F200, DMSO, or SP, cells were incubated with TMRE working solution (Beyotime, China) for 10 min. Cells were then stained with 10 μg/mL Hoechst 33258 (Yeasen, China) for 10 min. Fluorescent images were captured using a fluorescence microscope (Nikon, Japan). Mitochondrial ROS levels were measured using MitoSOX Red (Thermo Scientific, USA). After treatment with the respective media, cells were incubated with 5 μM MitoSOX for 10 min, followed by 0.1 μM Mito-Tracker Green (Beyotime, China) for 30 min and 10 μg/mL Hoechst 33258 for 10 min. After washing with HBSS, fluorescence was analyzed under a fluorescence microscope (Nikon, Japan).

Total RNA from the brain and the cells were extracted using an RNA Extraction Kit (Takara, Japan) followed by the manufacturer’s instructions. The cDNA was synthesized from 1,000 ng RNA using the 1st Strand cDNA Synthesis SuperMix (Yeasen, China). mRNA levels were quantified via SYBR Green (Yeasen, China) real-time PCR and normalized to β-actin using the 2^−∆∆Ct method. Primer sequences: mouse CAIII (F 5′-GAATCTCAGCACTCCTACTTTCA-3′; R 5′-GTCCGCATACTCCTCCATAC-3′), mouse β-actin (F 5′-CCTCTATGCCAACACAGT-3′; R 5′-AGCCACCAATCCACACAG-3′), rat CAIII (F 5′-TGCCGGGACTATTGGACCTA-3′; R 5′-ATTCTCTGCACTGGCGAACA-3′), rat β-actin (F 5′-CCTCTATGCCAACACAGT-3′; R 5′-AGCCACCAATCCACACAG-3′).

Proteins from the brains and the cells were extracted using radioimmunoprecipitation assay (RIPA) buffer as previously described (McCrimmon et al., 2012). Protein concentrations were determined by BCA method (Thermo Scientific, USA). Proteins were separated by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE), transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA), and probed with anti-CAIII (1:500; Santa Cruz, USA), anti-p-AMPK (1:1000; Cell Signaling, USA), anti-AMPK (1:1000; Cell Signaling, USA), anti-Sirt1 (1:1000; Abcam, UK), anti-PGC-1α (1:1000; Cell Signaling, USA), anti-Bcl-2 (1:1000; Cell Signaling, USA), anti-Bax (1:1000; Cell Signaling, USA), cleaved-Caspase 3 (1:1000; Cell Signaling, USA), Bcl-xl (1:1000; Cell Signaling, USA), and anti-β-actin (1:2000; Cell Signaling, USA) antibodies. ECL (Bio-Rad, USA) was used for imaging, and the protein bands were analyzed using Image J software.

Data were described as mean ± standard error of mean (SEM). Statistical comparisons were made using Student’s t-test for two-group analysis and one-way ANOVA for multiple groups. Analysis and graphing were conducted with GraphPad Prism 9.0, EthoVision, Image J, and R software. p < 0.05 was considered statistically significant.

The structure of 9-PAHSA and S-9-PAHSA (SP) are depicted in Figure 1A, and the experimental timeline is shown in Figure 1B. The hippocampus is an essential region for memory and the loss of hippocampal neurons is the main pathological feature of cognitive impairment. HE staining revealed that the ND group displayed clearly, normally arranged neurons with evenly distributed basophilic granules (Nissl bodies) (Figures 1C,D). Conversely, HFD-fed mice exhibited disorganized neurons in the CA1 and CA3 regions, with irregular morphology, reduced basophilic granules, highly eosinophilic cytoplasm and dense black basophilic nuclei, indicative of neuronal damage. SP treatment ameliorated this injury in diabetic mice, as evidenced by improved cellular organization (Figures 1C–G). Nissl staining confirmed that the significant Nissl body reduction in the HFD group was significantly reduced, suggesting neurodegeneration, which was attenuated by SP (Figures 1D–G). The results implied that SP might rescue the HFD-induced hippocampal neuronal damage.

Y maze test was used to assess working memory, and representative trajectories are shown in Figure 2A. HFD-fed mice showed reduced total distance, movement velocity, and alternation percentage compared to ND-fed mice, while SP treatment reversed these deficits (Figures 2B–D, all p < 0.05). This indicated that SP enhanced working memory in diabetic mice.

Next, we found that SP increased Bcl-2 (Figures 2E,F, p < 0.01) and Bcl-xl (Figures 2I,J, p > 0.05) expression, decreased Bax expression (Figures 2G, p > 0.05) and cleaved-Caspase 3 (cl-casp3) (Figures 2K,L, p > 0.05), and remarkably elevated the Bcl-2/Bax ratio (Figure 2H, p < 0.05) in the hippocampus, attenuating neuronal apoptosis in the hippocampus.

In addition, AMPK phosphorylation was significantly diminished in diabetic mice (Figures 2M–P, p < 0.01). SP intervention rescued the inhibition of AMPK phosphorylation in the HFD group (Figures 2M,N, p < 0.05), significantly increased downstream Sirt1 expression (Figure 2O, p < 0.05), and slightly elevated PGC1α expression (Figure 2P, p > 0.05).

Mice on a high-fat diet showed significantly increased body weights (44.51 ± 5.06 g vs. 30.97 ± 1.19 g) and fasting blood glucose levels (9.82 ± 1.39 mmoL/L vs. 5.95 ± 0.98 mmoL/L) compared to those on a normal diet (Figures 3A,B, all p < 0.0001). Of the 30 HFD-fed mice, 28 had fasting glucose levels exceeding 8.0 mmoL/L, resulting in a 93.3% success rate for diabetes modeling. These findings confirm that chronic HFD feeding substantially elevated both body weight and blood glucose. In HFD-fed mice treated with SP, serum and brain SP levels were increased by 34.8 and 21.7%, respectively, while levels in the small intestine and colon were reduced by 36.7 and 83.6%, respectively (Figures 3C–F, p > 0.05). The results suggested that SP is absorbed through the intestine and can reach the brain. Lactate dehydrogenase (LDH), a key marker of tissue damage, showed significantly increased activity in the serum of HFD mice compared to ND controls, which was attenuated by SP treatment (Figure 3G). Additionally, hippocampal reactive oxygen species (ROS) levels were lower in SP-treated mice compared to the HFD group (Figure 3K). In both serum and hippocampus, antioxidant enzyme activities of SOD and CAT were reduced in the HFD group but were elevated by SP treatment (Figures 3H,I,L,M). SP also tended to increase GSH-Px activity (Figures 3J,N), although it had no effect on H₂O₂ (Figure 3O). These results suggest that SP may alleviate oxidative stress and exert protective effects.

We observed a reduction in CAIII expression in the hippocampus of HFD-fed mice, whereas treatment with SP effectively mitigated the HFD-induced downregulation of CAIII (Supplementary Figures S2A,B, p < 0.01), suggesting that CAIII may be involved in the protective effects of SP. To further investigate this possibility, we performed bilateral hippocampal knockdown of CAIII in T2DM mice following the protocol (Figure 4A). HFD mice were bilaterally injected with AAV-Con sh or AAV-CAIII shRNA. After 3 weeks, mRNA and protein expression of CAIII in the hippocampus, cortex, and remaining brain regions (excluding the hippocampus and cortex) were significantly reduced in the AAV-CAIII sh group compared to the AAV-Con sh group (Figures 4B–G; Supplementary Figures S3A–C). These findings confirm the effective knockdown of CAIII in these brain areas following AAV-CAIII shRNA injection. AAV injection had no observable effect on weekly body weight or fasting blood glucose levels in any group (Supplementary Figures S3D,E). Weekly body weight and fasting blood glucose of HFD-fed mice were significantly higher compared to ND-fed mice, with no significant variations observed within the HFD group (Figures 4H,I). Daily water intake and food intake showed no difference across all groups (Figures 4J,K, p > 0.05). In IPGTT, blood glucose peaked 30 min post-injection, with HFD-fed mice showing higher levels than ND-fed mice at all time points, but no AUC differences among HFD-fed groups (Figures 4L,M). IPITT revealed no significant differences in insulin sensitivity among HFD-fed groups (Figures 4N,O). These results indicated that SP had no significant effect on glucose tolerance and insulin sensitivity.

Y-maze test showed that compared to the ND group, the total distance (p > 0.05), movement velocity (p > 0.05), and alternation percentage (p < 0.05) decreased in the HFD group and HFD + AAV-Con+vehicle group (Figures 5A–D). SP administration improved cognitive performance (Figure 5D, p < 0.01), except for no change in alternation percentage between HFD + AAV-CAIII+vehicle and HFD + AAV-CAIII+SP groups. It suggested that SP ameliorates spatial memory in diabetic mice, with CAIII knockdown in the hippocampus negating this benefit.

The trajectories of MWM test are shown in Figure 5E. HFD-fed mice demonstrated increased latency compared to ND controls across training days (Figure 5F). On the probe trial day, the escape latency elevated in the HFD and the HFD + AAV-Con+vehicle group, which was remarkably reduced by SP treatment, but this effect was abolished with CAIII knockdown (Figure 5G, p < 0.05). HFD group and the HFD + AAV-Con+vehicle group also showed significant fewer platform crossings and decreased time in the platform and the target quadrant versus the ND group, improvements that were observed with SP administration (Figures 5H–J, p < 0.05). No obvious difference in total distance and velocity were noted among groups (Figure 5K,L, p > 0.05). These findings further confirmed that CAIII may mediate the protective effect of SP on cognitive function of diabetic mice.

Hippocampal ROS levels were significantly elevated in the HFD and HFD + AAV-Con+vehicle groups but were notably reduced with SP treatment. This SP-induced reduction in ROS was attenuated in CAIII knockdown mice (Figure 6A). SP also enhanced SOD activity and CAT, effects not seen in CAIII knockdown mice (Figures 6B,C). A similar trend was also found in GSH-Px (Figure 6D, p > 0.05), while SP had no significant effect on H₂O₂ (Figure 6E, p > 0.05). These findings indicate that SP’s neuroprotection may involve CAIII-mediated modulation of oxidative stress. NeuN staining revealed disorganized neurons in the CA1 and CA3 regions of the HFD + AAV-Con+vehicle group, which were more orderly in the HFD + AAV-Con+SP group. This beneficial effect of SP was not observed in CAIII knockdown mice (Figure 6F). Additionally, the proportion of NeuN-positive cells was higher in both the CA1 (Figure 6G, p < 0.05) and CA3 regions (Figure 6H, p > 0.05) in the HFD + AAV-Con+SP group compared to the HFD + AAV-Con+vehicle group, with CAIII knockdown reducing this protective effect (Figures 6G–I, p > 0.05). Furthermore, the SP-induced increases in Bcl-2/Bax ratio and Bcl-xl were diminished in CAIII knockdown mice (Figures 6J–O), while the cleaved-Capase 3 decreased by SP was increased after CAIII knockdown (Figures 6P,Q, p > 0.05). Likewise, the inhibitory impact of SP on the AMPK pathway in diabetic mice was also decreased (Figures 6R–U, p > 0.05). These results suggested that CAIII played a crucial role in mediating the neuroprotective and anti-apoptotic effects of SP.

Consistent with the in vivo findings, our in vitro experiments revealed that a diabetes-mimicking environment led to a reduction in CAIII expression, which was alleviated by SP treatment (Supplementary Figures S2C,D, p < 0.05). Lentiviral transfection successfully generated CAIII knockdown (CAIII sh), overexpressing (CAIII OE), and their negative control (NC) PC12 cells, confirmed at both the mRNA and protein levels (Supplementary Figures S4A–F, p < 0.05). Cell viability was reduced in the G100F200 and DMSO groups but increased in NC and CAIII OE cells (Figures 7A,C,D, p < 0.05). SP did not significantly affect cell viability in CAIII sh cells under diabetic conditions (Figures 7B, p > 0.05). Elevated LDH release was observed under high glucose and lipid conditions in NC cells (Figures 7E–H, p < 0.001). SP intervention significantly lowered LDH release in NC cells exposed to diabetic conditions (Figure 7E, p < 0.05), but this inhibitory effect was negated by CAIII knockdown (Figure 7F, p > 0.05). Additionally, there was a tendency for SP to increase LDH in CAIII OE cells (Figure 7H, p > 0.05). Flow cytometric analysis further demonstrated that SP significantly attenuated apoptosis induced by high-fat and high-glucose conditions (Figures 7I,J, p < 0.01). In NC cells, antioxidant enzymes including SOD, CAT, and GSH-Px were significantly reduced in G100F200 and DMSO group compared to controls (Figures 7K–U, p < 0.05). SP treatment notably reversed this decline (Figures 7K,O,S, p < 0.05). However, in CAIII knockdown cells, SP showed no significant effect (Figures 7L,P,T, p > 0.05). In contrast, in CAIII OE cells, SP continued to restore antioxidant enzyme levels (Figures 7N,R,V, p < 0.05). Therefore, CAIII might mediate the protective effects on neuronal viability and cellular damage mitigation by SP in vitro.

Our in vivo experiments indicated that CAIII influences hippocampal ROS production. Given mitochondria’s central role in ROS generation, we detected ROS levels in mitochondria. Mitochondrial ROS were markedly elevated in NC cells under high-glucose and high-fat stimulation but were attenuated by SP (Figures 8A,B, p < 0.05). However, SP had no significant effect on mitochondrial ROS in CAIII knockdown cells (Figures 8C,D, p > 0.05). Subsequently, we assessed the mitochondrial membrane potential by TMRE staining, which revealed high intensity in controls and diminished intensity under diabetic conditions. SP treatment restored the loss of mitochondrial membrane potential in the diabetic environment (Supplementary Figures S5A,B, p < 0.0001). This beneficial effect of SP was absent in CAIII knockdown cells (Supplementary Figures S5C,D, p > 0.05). These results suggest that CAIII is essential for the impact of SP on mitochondrial function and ROS reduction in diabetic conditions.

Volcano plots and heatmaps were used to visualize differentially expressed genes (DEGs) between DMSO- and SP-treated groups. Thirteen genes were significantly altered in NC cells following SP treatment, including five upregulated and eight downregulated transcripts (Supplementary Figures S6A–D, FDR-adjusted p < 0.05). The most prominently upregulated genes included Pfkfb3, Mrfap1, RT1-M6-1, 1810024B03Rik, and Fbxo41. Among these, Pfkfb3 has been implicated in multiple biological processes, particularly in the regulation of the AMPK signaling pathway. In CAIII sh cells, five genes were differentially expressed in the SP group, with 3 upregulated (Srm, Cebpb, Oca2) and 2 downregulated (LOC361108, Mgp) (Supplementary Figures S6A–D, FDR-adjusted p < 0.05), involving in cellular process and various pathways. The UpSet plot further illustrates the intersections of DEGs in DMSO- or SP-treated NC cells and CAIII sh cells (Supplementary Figure S6E).

GO analysis revealed that in NC cells, SP and DMSO treatments induced differential gene expression across various biological processes, cellular components, and molecular functions (Figure 9A). In contrast, in CAIII sh cells, gene expression shifted, with a decrease in genes related to signaling and metabolic processes, an increase in organelle-associated genes, and a reduction in catalytic activity genes (Figure 9B). These changes suggest that SP modulates neuronal processes via CAIII. GO enrichment analysis showed that SP in NC cells significantly impacted terms related to multicellular organism development, anatomical structure development, response to stimulus, and intracellular membrane-bounded organelles (Figure 9C). In CAIII sh cells, SP’s effects on protein binding and membrane-bounded organelles were greatly diminished (Figure 9D), indicating that CAIII mediates SP’s regulation of cellular processes in PC12 cells.

Pathway analysis revealed that SP influenced several pathways in NC cells, including immune response, endocrine signaling, carbohydrate metabolism, transcription, signal transduction, and cell growth and death (Figure 9E). However, following CAIII knockdown, the impact of SP on immune and signal transduction pathways was reduced, while effects on endocrine signaling, carbohydrate metabolism, and cell growth were less pronounced (Figure 9F). KEGG analysis further identified that SP affected pathways involved in diabetic complications, such as AGE-RAGE signaling, endocrine resistance, type 1 diabetes, and HIF-1 signaling (Figures 9G,H). CAIII knockdown reduced the number of pathways influenced by SP, reinforcing the role of CAIII in mediating SP’s effects on oxidative stress, cell proliferation, and apoptosis in PC12 cells.