Wild-type male C57BL/6 mice (6-8 weeks old) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd (Shanghai, China), and housed in ventilated cages with free access to water and food in a light/dark cycle (12/12 h), temperature and humidity-controlled environment. All animal procedures were approved by the Institutional Animal Care and Use Committee of Shanghai Medical College of Fudan University. All efforts were made to minimize the suffering.

The diabetic/hyperglycemic mice model (DM) was established as described previously (25–27). In brief, the mice were fed adaptively with ordinary diet for 1 week, and then randomly grouped in the control group and the DM group. The DM mice were then fed with high-fat diet (FBSH, Shanghai, China) for 8 weeks, while the control mice with the ordinary diet for 8 weeks. After that, the mice were injected intraperitoneally with 40 mg/kg streptozotocin (STZ, Sigma-Aldrich, USA) or a citrate buffer for 5 days. One week later, those with fasting blood glucose higher than 11.1 mmol/L were defined as the successful diabetic/hyperglycemic mice (28–30).

The transient middle cerebral artery occlusion (MCAO) model was performed essentially as previously described (31–33). Briefly, DM or control mice were kept on a feedback-controlled heating pad at 37 ± 0.5°C, anesthetized with 3% isoflurane in 30% O₂ and 70% N₂O for induction and 1.5% isoflurane for maintenance. A nylon monofilament coated with silicone rubber (CINONTECH, Beijing, China) was introduced into the internal carotid artery through the external carotid artery to occlude the middle cerebral artery (MCA) and removed 30 min later to allow reperfusion. Sham-operations were subject to the same procedure except for the MCA occlusion. Regional cerebral blood flow (rCBF) was monitored at the surface of ipsilateral parietal cortex during ischemia by using Laser Doppler Flowmetry (Periflux System 5000, Perimed) as previously described (31, 33). The mice were kept for further experiments and statistics analysis if blood flow perfusion dropped and maintained at ~20% of the preischemia baseline during ischemia. We performed MCAO surgery 1 week after the hyperglycemia onset in the Experiment 1, and 1 month thereafter in the Experiment 2 (Figures 1, 2), with a relatively longer diabetes duration.

The recombinant adeno-associated virus (AAV) shRNA expressing vectors containing NRSF-shRNA (5′-GCATGAAGTGACCCGACAT-3′) or non-specific control-shRNA (5′-TTCTCCGAACGTGTCACGT-3′) were constructed by Shanghai GeneChem using the pAKD-CMV-bGlobin-eGFP vector as previously reported (24). The mice were anesthetized with isoflurane, fixed in a stereotaxic apparatus, and 0.8-μl AAV solution was slowly injected unilaterally into the ipsilateral hippocampus (AP: −1.85 mm, ML: −1.60 mm, DV: −1.90 mm) at an infusion rate of 0.08 μl/min with a 5-μl capacity micro syringe (Gaoge, Shanghai, China), and then kept for 10 min after the injection. After suturing the scalp, the mice were placed in an incubator until recovery.

All the tests and analyses were performed according to the experiment schedule (Figures 1, 2) and blind to the observers.

Immunofluorescence staining was performed as described previously with minor alterations (42, 43). Briefly, the mice brain was coronally cut into 30-μm-thick sections by using frozen microtome (Leica, Germany) and stored in a cryoprotectant solution at −20°C. Frozen sections were permeabilized with 0.5% Triton X-100, blocked by 10% normal goat/donkey serum (Jackson, USA) and 0.3% Triton X-100 in PBS for 1 h, and incubated with specific primary antibodies (CD31, 1:200, goat, R&D systems; MAP2, 1:50, mouse, Abcam; NRSF, 1:200, rabbit, Biorbyt; PDGFR-β, 1:100, rabbit, Abcam; GFAP, 1:200, mouse, CST) at 4°C for 48 h. The primary antibodies were visualized by incubating with fluorophore-conjugated secondary antibodies as follows: Alexa Fluor 488 goat anti-mouse IgG (1:800, Jackson, USA); Alexa Fluor 488 donkey anti-goat IgG (1:800, Jackson, USA); Alexa Fluor 488 donkey anti-mouse IgG (1:800, Jackson, USA); Cy3 donkey anti-goat IgG (1:800, Jackson, USA); Cy3 donkey anti-rabbit IgG (1:800, Jackson, USA); Alexa Fluor Plus 594 goat anti-rabbit IgG (1:1,000, Invitrogen, USA). DAPI (1:500, CST, USA) were used for nuclei staining. Images were captured using a fluorescence microscope (Nikon, Japan) or a confocal microscope (Olympus, Japan) and analyzed by using Image J software. Pericyte coverage was quantified as the percentage of PDGFR-β⁺ CD31⁺ double-positive pericyte area/CD31⁺ area (total vascular area) × 100% as described previously (44). The vascular length and the area were measured using image analysis software Image J as described previously (45), and results were normalized to the mean value of the sham group.

TTC staining was performed 1 day after ischemia/reperfusion (46, 47). Freshly cut 1-mm-thick brain sections were stained with 2% 2,3,5-triphenyltetrazolium chloride (TTC, Sigma-Aldrich, USA) for 15 min at 37°C while protected from light, and then fixed in 4% PFA overnight and photographed. The infarcted area and the brain area in TTC-stained brain slices were measured using Image J software. The corrected percentage of infarction was calculated as the formula: infarction (%) = (the area of contralateral hemisphere – the area of non-infarcted ipsilateral hemisphere)/(2 ^* the area of contralateral hemisphere) ^*100% (48).

Nissl staining was performed as previously reported (49, 50). The 30-μm frozen brain slices were mounted on glass slides coated with gelatin, dried for 48 h at 37°C, and then dehydrated and rehydrated with graded ethanol solutions (75, 85, 95, 100%), and stained with 0.25% cresyl violet (Sigma-Aldrich, USA) for 12 min. After that, sections were decolorized (95% ethanol containing 0.3% glacial acetic acid) and dehydrated in graded serried of ethanol. After being transparentized by xylene and sealed with neutral resin, the slices were observed under the light microscope (Nikon, Japan), and the number of surviving neurons per mm² was counted.

Fluoro-Jade B (FJB) staining was performed in accordance with the manufacturer's instructions (Millipore, USA) to assess the neurodegeneration after MCAO.

BBB permeability was evaluated with an Evans blue (EB) assay and IgG staining as previously reported (51, 52). At 6 h before sacrificed and 18 h after MCAO, the mice were intravenously injected with 2% Evans blue (0.4 g/kg body weight; Sigma-Aldrich, USA) in sterile saline via femoral vein. For IgG staining, brain sections were incubated with IgG (Alexa Fluor 488 goat anti-mouse IgG, 1:500, Jackson, USA), and brain images were captured using the fluorescence microscope (Nikon, Japan).

Western blotting analysis was performed as reported previously with minor alterations (53). Total protein of brain tissue was extracted by tissue homogenization in a RIPA lysis buffer (EpiZyme, Shanghai, China) with protease inhibitor and phosphatase inhibitor cocktail (Rocher, Switzerland). The protein concentrations were assessed by the BCA Protein Assay Kit (Thermo, USA). Equal amounts of protein were loaded onto 7.5% SDS-PAGE gels to electrophoresis, and then transferred to the polyvinylidene difluoride membrane (PVDF; Millipore, USA), blocked with blocking solution (Epizyme, Shanghai, China), incubated with primary antibodies overnight at 4°C (NRSF, 1:1,000, Novus; HDAC1, 1:1,000, CST; mSin3A, 1:1,000, SAB; NRP-1, 1:1,000, Abcam; β-TrCP, 1:1,000, CST; β-actin, 1:20,000, Proteintech). Membranes were incubated with HRP-conjugated goat anti-rabbit and goat anti-mouse (1:10,000, CST) secondary antibodies after being washed in TBST. Proteins were visualized using ECL Kit (Epizyme, Shanghai, China). The gray value of protein bands was quantified with Image Lab software (Bio-Rad, USA) and normalized to β-actin.

The proteins were extracted by tissue homogenization in 0.01 mmol/L PBS with protease inhibitor and phosphatase inhibitor cocktail (Rocher, Switzerland). Vascular endothelial cell growth (VEGF) and vascular endothelial cell growth factor receptor 2 (VEGFR2) levels in mice hippocampus were assessed by an ELISA kit (DiLab, Shanghai, China) according to the manufacturer's instructions.

The polyethylene glycol-associated solvent system (PEGASOS) tissue-clearing method was performed as previously described (54, 55). The mice were intravenously injected with DyLight 649-labeled Tomato Lectin (1.25 mg/kg body weight; Vector Labs, USA) via femoral vein to label blood vessels after anesthetization. The brain was removed 5 min later and perfused with PBS containing 10 U/ml heparin sodium and 4% PFA, and then fixed in 4% PFA for 24 h, decolorized in 25% ethylenediamine for 2 days, and the solution was refreshed daily, and then washed with PBS for 30 min, delipidated with graded tert-Butanol (tB; 30, 50, 70%) solution, dehydrated with tB-PEG solution [tB: PEG-MMA-500 (polyethylene glycol methyl ether methacrylate average Mn500): ethylenediamine = 70:30:3, V/V], transparentized with benzyl benzoate (BB)-PEG solution (BB: PEG-MMA-500: ethylenediamine = 75:25:3, V/V) until the tissue turned transparent.

Clearing brain images were acquired with the light-sheet microscopy (LS18, Nuohai, Shanghai, China). 3D reconstruction and analysis were performed with Imaris software version 9.7 (Bitplane, Switzerland). The Imaris Surface algorithm was used to semi-manually determine the identity of the hippocampus; the Filament algorithm was used for tracing blood vessels of hippocampus. Hippocampal volume, vascular volume, vascular length, vascular branch points were calculated by the Imaris software. The vascular volume density was quantified as hippocampal vascular volume/hippocampal volume. Vascular length density was quantified as hippocampal vascular length/hippocampal volume. Vascular branch points density was quantified as hippocampal vascular branch points/hippocampal volume.

Statistical analysis was performed using GraphPad Prism 7.0 software. Distributed continuous variables were expressed as mean ± SEM, and non-parametric data were presented as median (interquartile range). Data were compared with unpaired two-tailed Student's T-test or Mann–Whitney U-test between two groups, one-way ANOVA followed with Dunnett's or Tukey's multiple comparisons test among multiple groups, and two-way ANOVA followed by unpaired two-tailed Student's T-test between the two groups.

To investigate whether diabetic/hyperglycemic conditions exacerbate the ischemic-induced brain injury, we performed MCAO operation at Day 1 after the DM mice model was successfully established. After TTC staining, our results confirmed that the diabetic/hyperglycemic ischemic mice (DM + MCAO) had significantly larger brain infarction when compared with the ischemia control (MCAO) mice (Figures 3A,B, p < 0.01). In the MCAO mice, the focal ischemia induced mild infarction mainly in ipsilateral cortex and striatum, while, in the DM + MCAO mice, the infarction mostly enlarged to the entire hemisphere, including hippocampus as well.

When we observed the ipsilateral hippocampus, data from Nissl staining and FJB staining showed that the DM + MCAO mice had significantly induced neuronal loss and increased FJB-positive cells (Figures 3C,D). In the MCAO mice, only sparse FJB-positive cells appeared in CA1 but was not evident in CA2 and CA3 areas. However, in the DM + MCAO mice, remarkably increased FJB-positive cells appeared in not only CA1 but also CA2 and CA3. These results suggested that diabetic ischemia induced much more drastic neural damage in the ipsilateral hippocampus than nondiabetic ischemia.

In addition, we assessed the neurobehavioral deficits in the mice. Previous studies found that motor functional disorders were more apparent in acute MCAO injury (56, 57). As shown in Figures 3E-G in our study, the DM + MCAO mice also exhibited more serious sensorimotor disorders. The neuro-deficit score notably increased in the MCAO mice and further upregulated in the DM + MCAO group 1 day after the MCAO (Figure 3E, p < 0.05, vs. the MCAO group). Rotarod test and grid walking test were performed at the 3rd and 5th day after surgery, respectively. Results showed that the mice in the DM + MCAO group spent less time on the rotating rod (Figure 3F, p < 0.0001, vs. the MCAO group) and performed more forelimb faults (Figure 3G, p < 0.001, vs. the MCAO group). Moreover, since diabetic ischemia induces greater damage in the ipsilateral hippocampus, we also assessed the hippocampus-dependent spatial learning and memory ability by performing Y maze test at the 28th day after the MCAO surgery. The percentage of spontaneous alternations was significantly decreased in the DM + MCAO group (Figure 3H, p < 0.05, vs. the MCAO group), indicating significantly impaired learning and memory function in the DM + MCAO mice.

Cerebral ischemia leads to not only neuronal injury but also vasculature disorders, we further detected the blood-brain barrier (BBB) permeability by using the EB assay and IgG staining. As shown in Figures 4A,B, the EB leakages and the IgG-fluorescent area were greatly increased in ischemic hemisphere of the DM + MCAO mice compared with that of the MCAO mice. In consistent with the severer hippocampal neuronal damage in the DM + MCAO mice, the BBB leakage is also obvious in the hippocampus of the DM + MCAO mice. Next, we observed abnormal alterations of the BBB components. GFAP is the specific marker of astrocytes. As shown in Figures 4C,D, GFAP⁺ cells appeared apparently swollen, and the number of GFAP⁺ cells increased significantly in ipsilateral hippocampus of the DM + MCAO mice when compared with the sham mice (p < 0.01), suggesting that activated astrocytes increased in the DM + MCAO mice. Studies have found that pericyte coverage is negatively correlated with vascular maturation and BBB permeability (58–60), we further detected pericyte coverage on the blood vessels after DM + MCAO application. PDGFR-β is one of the specific markers of pericyte. As shown in Figures 4E,F, the PDGFR-β⁺ coverage on blood vessels was substantially decreased in the ipsilateral hippocampus of the DM + MCAO mice (p < 0.01, vs. the MCAO group). Moreover, our data showed that, in the ipsilateral hippocampus of the DM + MACO mice, CD31-labeled brain vascular length (Figures 4G,H, p < 0.01, vs. the MCAO group) and density (Figures 4G,I, p < 0.01, vs. the MCAO group) were remarkably reduced. These results suggested that diabetes/hyperglycemia worsened ischemia-induced BBB integrity and BBB vasculature disorders.

Therefore, the above results collectively demonstrated that the diabetic/hyperglycemic condition could further worsen cerebral ischemia-induced neuronal damage and vasculature disorders and deteriorate neurobehavioral disabilities.

We previously reported that NRSF elevation participates in neuronal injury under hyperglycemia environment (24), and then we detected NRSF levels in the MCAO and DM + MCAO mice by using the immunofluorescent staining in the present study. As shown in Figure 5A, the NRSF fluorescent particle was obviously increased in the hippocampus of the MCAO mice, and further enhanced in the DM + MCAO mice. Observing the subcellular distribution of NRSF, we further found that, in some neurons, the NRSF particles only localized in the nucleus, while, in other neurons, the NRSF particles localized both in the nucleus and cytoplasm or almost exclusively clustered in the extranuclear area (see white arrows in Figure 5A). Statistical analysis revealed that, in the hippocampus of the DM + MCAO mice, the integrated fluorescence intensity of NRSF was significantly enhanced (Figure 5B, p < 0.01, vs. the MCAO group).

Western blotting analysis showed accordant results that the NRSF protein level was markedly increased in the hippocampus of the DM + MCAO mice (Figures 5C,D, p < 0.05, vs. the MCAO group). Interestingly, we found that the expression of NRSF cofactors, HDAC1, and mSin3A also increased significantly in the DM + MCAO mice, following the same pattern as that of NRSF (Figures 5C,E,F). Therefore, these results demonstrated that diabetes/hyperglycemia enhanced MCAO-induced elevation of NRSF and its corepressors, which indicated the deepened activation of NRSF in the DM + MCAO mice.

β-TrCP is a well-known E3 ubiquitin ligase, which could mediate proteasomal degradation that regulates the expression of NRSF by ubiquitination (21). Thus, we detected the hippocampal β-TrCP level in the MCAO and DM + MCAO mice. Consequently, the expression of β-TrCP was reduced in the MCAO group and the DM + MCAO group (Figures 5C,G, p < 0.05, vs. the sham group).

Patients with stroke with chronic diabetes commonly present deficits in learning/memory abilities, which are closely related to the functions of hippocampus (61–66). Since we observed that the neural and vascular injuries increased and the NRSF is elevated in the hippocampus of the DM + MCAO mice, we investigated whether the enhanced hippocampal NRSF participates in diabetic ischemia-worsened brain damage and learning/memory abilities. Here, we intra-hippocampally injected NRSF shRNA (sh-NRSF) in the DM + MCAO mice to knockdown the expression of NRSF at 1 week after the hyperglycemia onset and performed the MCAO surgery at 1 month after the hyperglycemia onset. As shown in Figures 6A-G, the green fluorescence of sh-NRSF precisely localized in the ipsilateral hippocampus and effectively knocked down the NRSF expression, which coincides with our previous study (24). By whole-brain clearing, we semi-manually determined the identity of the hippocampus and assessed hippocampal volume by Imaris Surface algorithm. As shown in Figures 6B,D, although the contralateral hippocampus volume of the two groups was approximately equal (DM + MCAO sh-control: 1.01 × 10¹⁰ mm³; DM + MCAO sh-NRSF: 1.02 × 10¹⁰ mm³), the volume of ipsilateral ischemic hippocampus in the DM + MCAO sh-NRSF group was about 1.5 times larger than that of the DM + MCAO sh-control group (DM + MCAO sh-control: 4.31 × 10⁹ mm³; DM + MCAO sh-NRSF: 6.09 × 10⁹ mm³), which suggested that the knockdown of NRSF might ameliorate diabetic ischemic hippocampal atrophy. Meanwhile, the Nissl staining confirmed that the NRSF knockdown rescued neuronal loss, shrinks, and deformations in hippocampus of the MCAO and DM + MCAO mice, and the number of surviving neurons was also increased after the NRSF knockdown (Figure 6H, Supplementary Figure 1).

To confirm that this beneficial effect is not due to the surgery-induced stroke severity differences between sh-control and sh-NRSF groups, we also assessed the sensorimotor deficits 1 day after the MCAO surgery in these mice. As shown in Figure 6I, no significant differences were observed between the two groups in the MCAO mice or in the DM + MCAO mice. Previous studies have found that the sensorimotor deficits of the animals surviving up to 24 h after MCAO are positively correlated with the infarct size in the brain (67), thus, these results indicated that the hippocampally injection does not affect the acute MCAO severity in the sh-control and sh-NRSF mice. In addition, we evaluated hippocampus-dependent spatial learning and memory ability in the mice after the knockdown of NRSF by Y maze and MW maze. The Y maze test was performed at 42nd day, and the MW maze test was performed 43rd-49th days after MCAO. Our data showed that there was no significant difference in swimming speed among the four groups (Figure 6J). However, the times of platform crossing (Figure 6K) and percentage of time in the target quadrant (Figure 6L) were increased in the sh-NRSF mice when compared with the sh-control mice. The percentage of spontaneous alternation was increased as well in the MCAO sh-NRSF and DM + MCAO sh-NRSF mice (Figure 6M). These results indicated that the NRSF knockdown in hippocampus rescued spatial memory of the mice in the MCAO sh-NRSF and DM + MCAO sh-NRSF groups.

Therefore, the above results demonstrated that the NRSF knockdown attenuates neuronal injury in hippocampus, which is induced by diabetic ischemic stroke.

By performing whole brain clearing with PEGASOS, we investigated whether the knockdown of NRSF could also rescue hippocampal vasculature impairment in the DM + MCAO mice. As shown in Figures 7A,B, the ipsilateral ischemic hippocampus of the DM + MCAO sh-control mice exhibited abnormal fluorescence patterns in the CA1-CA2 and CA3 regions. In hippocampus of the DM + MCAO sh-NRSF mice, the fluorescence distribution was more uniform (Figures 7C,D). Next, we assessed the ipsilateral ischemic hippocampal vascular volume, length, and branch points by Imaris Filament algorithm. Our data showed that the vascular volume (Figures 7E,F, 8.02 × 10⁸ mm³ in sh-control vs. 2.07 × 10⁹ mm³ in sh-NRSF), vascular length (Figures 7G,H, 3.76 × 10⁶ mm in sh-control vs. 6.08 × 10⁸ mm in sh-NRSF), vascular branch points (Figures 7I,J, 1.53 × 10⁶ in sh-control vs. 2.44 × 10⁶ in sh-NRSF) were all greatly increased in hippocampus of the DM + MCAO mice after the NRSF knockdown. Meanwhile, we also calculated the density of vascular volume, length, and branch points in ipsilateral hippocampus and found that the densities (for vascular volume: 0.34 in sh-NRSF vs. 0.19 in sh-control; for vascular length: 9.98 × 10⁻⁴ in sh-NRSF vs. 8.72 × 10⁻⁴ in sh-control; for vascular branch points: 4.00 × 10⁻⁴ in sh-NRSF vs. 3.54 × 10⁻⁴ in sh-control) were substantially improved as well.

The above results suggested that the NRSF knockdown protected microvasculature against diabetic ischemic injury. To elucidate the possible mechanisms underlying, we detected Neuropilin-1 (NRP-1)/VEGF signaling by using western blotting analysis and ELISA. As shown in Figures 8A,B, the NRP-1 level further decreased in the DM + MCAO mice (p < 0.05, vs. the sham group). The expression of VEGF and VEGFR2 was detected by ELISA. The results showed that their expression decreased as well in the DM + MCAO mice when compared with the MCAO mice (Figures 8C,D, both p < 0.05 vs. the MCAO group). After the knockdown of NRSF, NRP-1 levels significantly increased in both the MCAO and DM + MCAO mice (Figures 8E,F). The findings of ELISA showed that both VEGF and VEGFR2 were significantly upregulated after the NRSF knockdown (Figures 8G,H). These results suggested that the NRSF knockdown ameliorates the abnormal hippocampal vascularization probably by upregulation of NRP-1/VEGF signaling. Moreover, we found that corepressors HDAC1 and mSin3A levels were remarkably reduced after the NRSF-knockdown (Figures 8E,I,J) reconfirmed the inactivation of NRSF.