2.1 scRNA-Seq Data Preprocessing and Identification of Cellular Subpopulations

We accessed the single-cell dataset labeled GSE174574 from the GEO repository (https://www.ncbi.nlm.nih.gov/). Initially, we meticulously cleaned the dataset, excluding cells that exhibited gene counts below 200, above 5000, or with mitochondrial genes exceeding 25%. This stringent filtering process culminated in a refined dataset comprising 58,505 cells for subsequent analysis. To mitigate any batch effects that might skew the subsequent analysis due to sample origin, we deployed the “Harmony” R package (version 0.1.0) for batch effect correction [39]. Post the correction, we conducted a t-SNE (t-distributed stochastic neighbor embedding) analysis to delineate the cellular clusters. This was achieved by applying the FindNeighbors and FindClusters algorithms with a resolution parameter set to 0.1. For the identification of these clusters, we relied on established markers that have been documented in the existing scientific literature [40]. Specifically, Astrocyte was represented by Aldoc, Gfap, Aqp4, Slc1a3; Endothelial cell expressed Tie1, Tek, Vwf, Cldn5; Glial cell expressed Plp1, Sox10, Npy, Gja1; Microglial cell expressed Arpp21, Clec7a, Cx3cr1, Iba1, Ly6C, P2ry12, P2ry13, R3hdm1, Tmem119, Alf1, C1qa, Csf1r; Neutrophils expressed Cd44, Ccrl2, S100a9; Myeloid cell expressed Lyz2, Cd11b, Cx3cr1, Csf1r; neuron expressed Gad1, Gad2, Slc32a1, Slc17a6, Sp8, Tubb5; NK cell expressed Ncr1, Klrb1c, Klre1, Xcl1, Bcl2; Pericytes expressed Vtn, Rgs5, Kcnj8, Pdgfrb.

2.2 Calculation of Cell Hydrogen Sulfide and Autophagy Score

We employed the ReactomeGSA tool from the R programming environment to assess the functional enrichment for each distinct cell type and investigate the biological mechanisms that underpin the functions of each MCAO scRNA-Seq cell category. Additionally, the CytoTRACE software, specifically version 0.3.3, aided in forecasting the sequence of cell differentiation stages and offers valuable perspectives on cellular stemness [41]. To assess the Hydrogen sulfide (Systematic name: MM9615 and MM13529) and Positive regulation of autophagy (Systematic name: MM5430) enrichment score for each cell, we collected the genes from the GSEA database and calculated single-cell scores by jointly assessing signature mean and inferring enrichment (JASMINE) from the JASMINE package [42], incorporating 16 selected H₂S-associated and 161 positive regulation of autophagy-associated genes (Table S1). Finally, matrix bubble diagram and scatter diagram of the above two enrichment scores were performed through Hiplot (https://hiplot.com.cn/home/index.html).

2.3 Human Sample Collection

All human samples involved in the study were obtained according to the Eighty eighth Hospital of the People's Liberation Army ethics requirements, which was approved by the Medical Ethics Committee (No. 2016–010202). Blood samples from acute cerebral infarction patients were collected from the department of neurology and normal human samples were collected from department of physical examination center. Informed consent was obtained from all participants. For patients inclusion criteria, please refer to the “Eligibility Criteria for Participants” (S3).

2.4 H₂S Measurement

The serum was collected 48 h after the cerebral ischemia operation. The serum H₂S levels were measured according to a previous study [43]. Briefly, 30 μL serum was mixed with 10 μL Tris–HCL buffer (200 mM, pH 8.5) and 70 μL MBB solution (3.5 mM dissolved in acetonitrile). After shaking for 1 h in the dark, 10 μL 20% formic acid (v/v) was added and vortexed for 10 s. The supernatant was collected after 14,000 rpm for 10 min and stored at −70°C. Samples were measured by gas chromatography AB 4000Q TRAP tandem mass spectrometry detection [43]. Concentrations of H₂S were determined in accordance with the standard curve.

2.5 SPRC Pharmacodynamics Study

All animal experimental protocols were approved by the institutional ethical committee according to internationally accepted ethical standards. The animals were supplied by the Laboratory Animal Centre, Fudan University, Shanghai, China. Male Sprague–Dawley (SD) rats (6–8 weeks), weighing between 280 and 320 g, were intravenously injected with SPRC (dissolved with Phosphate Buffered Saline) at three different dosage levels (6.25 mg/kg, 12.5 mg/kg and 25 mg/kg). The first tail vein injection was performed within 3 min after the surgery, followed by a repeat injection after 90 min. Blood samples were collected at 5, 15, 30, 60, 95, 105, 120, 180, and 240 min after injections. Serum was separated for H₂S measurement as described above.

2.6 Cerebral Ischemia Reperfusion Model for Rats and Mice

Healthy 6–8 weeks male SD rats were housed under diurnal lighting and fed either food or water. Transgenic mice were obtained using C57BL/6J mice. The Longa ligation method was used to clog the left cerebral middle artery of rats or mice to develop the cerebral ischemia–reperfusion (I/R) injury [44]. After 1.5 h of ischemia, the suture was removed, the artery was ligated, and the muscle and skin were sutured. Animals were placed in a warm and comfortable environment until awake. All experimental operations were carried out under the premise of alleviating the suffering of animals. The rats or mice in sham groups only received anesthesia and artery separation without ligation.

The middle cerebral artery occlusion (MCAO) model was monitored and confirmed by Doppler flow detection (moor LDI2-HIR, Moor Instruments, Britain) with a 2-D laser speckle flow imaging technique. When the line blot was plugged into the beginning of the middle cerebral artery, the blood flow values suddenly dropped to 20%–30% of the original baseline (Figure S1A,B). It remained at such low blood flow during the whole process of cerebral I/R injury and gradually recovered almost close to the baseline until the bolt was slowly pulled out, which indicated that MCAO was successfully built up.

2.7 Construction of CBS^+/− + 3-MST-Ckd Mice

Cystathionine-β-synthase (CBS) knockout mice in a C57BL/6J background were obtained from Shanghai Research Center for Model Organisms and backcrossed to a C57BL/6J background. Wild-type control (CBS^+/+) mice from the same litter were used at 8 and 10 weeks of age. All animals' genotyping was performed using primers and protocols provided by the supplier. The CBS-KO mice were generated by standard embryonic stem cell (ESC) based gene targeting techniques and verified by PCR and southern blot. Briefly, the targeting vectors contained homologous recombination CBS gene fragments (Figure S2A) were transferred into ESC, and then the ESC with homologous vectors were transplanted into the blastosphere to breed chimeric mice. CBS knockout mice were genotyped from tails using PCR (forward: 5′-CCCTGGCATAGTCTCACAA-3′; reverse: 5′-ACTAGAGCTTGCGGAACCC-3′) and southern blot. The identification bands indicated that non-transgenic mice (NGT, namely wild type) only presented at 355 bp, while the homozygous (CBS^−/−) mice presented at a 203 bp band. The heterozygous CBS^+/− mice presented at both 355 bp and 203 bp (Figure S2B). CBS^−/− mice all died at the embryo or within 24 h after birth, which was consistent with the previous research [45]. CBS^+/− mice were smaller than NGT mice in the first 4 weeks in size, but there was no difference after adulthood. No other obvious difference was observed in appearance and movement between the two groups. The brain tissues were stained with Indian ink, and no difference was found in cerebral arteries anatomy in NGT and CBS^+/− mice (Figure S2C).

The 3-MST shRNA lentivirus, purchased from OBiO Technology (Shanghai) Co. Ltd., was injected into the brains of CBS^+/− mice using a stereotaxic device as previously described [46]. We injected lentivirus into the cerebral cortex of mice with 1 μL lentivirus injected into each half of the brain at the coordinates (BLA, AP: −2 ± 0.5 mm; ML, ±1 mm; DV, −0.6 mm), and then we collected the cerebral cortex between the anterior and posterior fontanelles 2 weeks later for detection. The mice could be used for subsequent experiments.

The physiological characters of CBS^+/−-3MST-cerebral knockdown (CBS^+/− + 3-MST-ckd) were measured and compared to the NTG and CBS^+/− mice, and no significant difference was found in these mice (Figure S3D,E and Supporting Table S1).

2.8 Animal Experiment Groups for SPRC Therapy Study

Male SD rats were divided randomly into the following groups: sham-operated group, cerebral I/R injury model group, SPRC groups (6.25, 12.5, 25 mg/kg, n = 47 per dosage group), and positive control edaravone group (3 mg/kg, n = 20). The administration method of SPRC has been described in Section 2.3. Edaravone was administered as a single intravenous injection within 3 min after modeling at a dose of 3 mg/kg.

Male mice were divided randomly into the 8 following groups: NTG + sham-operated group (n = 6), NTG + CIRI model group (n = 7), CBS^+/− + I/R group (n = 8), CBS^+/− + 3-MST-ckd + I/R group (n = 6), CBS^+/− + 3-MST-ckd + I/R + SPRC low dosage group (20 mg/kg, n = 6), CBS^+/− + 3-MST-ckd + I/R + SPRC middle dosage group (40 mg/kg, n = 6), CBS^+/− + 3-MST-ckd + I/R + SPRC high dosage group (80 mg/kg, n = 7) and CBS^+/− + 3-MST-ckd + I/R + L-Cys (30 mg/kg, n = 6) as a positive control. The scale of surgery and dosing refers to Figure S2C.

2.9 Brain Infarction Size Measurement

48 h after reperfusion, animals were sacrificed and the whole brain was collected and quickly frozen at −80°C for 5 min. The brain tissues were coronally cut into a set of continuous 2 mm slices. The infarct area was observed by using 0.2% TTC staining at 37°C for 30 min, followed by fixing in 4% paraformaldehyde overnight. Normal brain tissues were stained fresh red, while the infarcted area was pale. The Image J software (NIH, Boston, MA) was used to analyze the infarct volume. Infarct volume (%) = 100% × infarct area/total area of the brain slice.

2.10 Behavioral Evaluation

The neuronal function impairment was evaluated based on a neurological deficit grading system with a scale of 0 to 7 as previously described [47] at 48 h after ischemic stroke.

Spontaneous motor activity was measured using a rota-rod treadmill for rats, under 10 speeds from 4 to 40 rpm for 5 min. The interval from the timepoint when the animal climbed the rod to that when it fell down was recorded as the performance time [48]. The animals were trained five times per day until 2 days before cerebral IR injury in order to obtain stable baseline values. The mean duration for 5 times' training was recorded.

Cat walk evaluation was carried out as the following steps. Rats were trained to walk across a glass aisle daily for 1 week before surgery. Training was resumed 24 h after surgery, and data were evaluated 48 h after surgery.

2.11 PET/CT Scanning

Animals were anesthetized by i.p. 60 mg/kg pentobarbital sodium 48 h after surgery and fixed on a plate. Intravenously injected with F-18-FDG, the animal was sent into PET-CT for 45 min for imaging. CT scanned for 10 min: 80 kV, 500 μA, exposure time 1100 ms, 600 s/round, Fov (mm): 768 × 768 × 512. PET scanned for 20 min: Fov (mm): 128 × 128 × 159. Using the 3D acquisition method and all the cross-sectional images were selected. Images were analyzed by utilizing Inveon Acquisition Workplace software.

2.12 Brain Water Content

48 h after surgery, rats or mice were anesthetized and their brains on the ischemic side were separated into hemispheres. The wet weight was immediately measured and recorded. After drying at 60°C for 48 h, the dry weight of the same samples was quantified by an electronic analytical balance. The formula for the brain water content calculation was as follows: (wet weight-dry weight) / wet weight × 100%.

2.13 Primary Neuronal Culture

Cell plates were precoated with 0.1 mg/mL poly-lysine in advance for 12–24 h and blowdried sterilely. Pregnant rats (16–18 days) were sacrificed, and fetal rats were removed to high glucose Dulbecco's modified Eagle medium (DMEM, Hyclone, USA) without fetal bovine serum (FBS). The cerebral cortex of fetal rats was separated in buffer and washed with hank's buffer 3 times. Hank's buffer was later replaced by trypsin. Tissues were cut with a scissors and incubated at 37°C for 30 min after being transferred into high glucose DMEM medium added with 10% FBS, and a pipette was used to gently blow into a single neuronal cell. After centrifuging at 800 rpm for 8 min, the precipitated cells were counted and plated in the prepared cell culture plates. The primary neurons were identified by microtubule-associated protein 2 (MAP-2) staining and the purity was > 95% (Figure S4).

2.14 MTT Assay

Neurons were planted and incubated in a 96-well plate after 7 days' primary culture, and the supernatant was replaced with DMEM without glucose or FBS to develop an oxygen glucose deprivation (OGD) model. 3-Methyladenine (3-MA), the specific inhibitor of autophagic/lysosomal protein, was used to suppress autophagy. The OGD condition and the corresponding treatment groups were kept at the settled different time points and changed to normal medium for additional 2 h incubation. In each well, 20 μL MTT (final concentration is 0.5 mg/mL) were added and incubated for 4 h at 37°C with 5% CO₂. After removal of supernatant, samples were washed with D-Hank's solution once followed by 100 μL DMSO and 10 min slight shaking. The absorbance was detected by a microplate reader at 490 nm (Infinite 1000; Tecan, Switzerland).

2.15 Lactate Dehydrogenase (LDH) Assay

Primary neurons were planted at 10⁴ cells/ml in a 96-well plate on the seventh day. After overnight incubation, the medium was replaced with DMEM with 0.5% FBS. Cells were incubated in low serum medium and then exposed to the corresponding reagents for 30 min. The plates were centrifuged at 400 g for 5 min and the supernatants were transferred into a new 96-well plate for detection following the protocol from the manufacturer (LDH cytotoxicity assay kit, Beyotime Biotechnology, Shanghai, China) at 490 nm, and calculation was performed in accordance with the instructions.

2.16 Western Blot Analysis

Tissue lysates were prepared using RIPA lysis buffer (Beyotime Biotechnology, Shanghai, China) containing protease and phosphatase inhibitors. Immunoblot was used to analyze proteins with a standard procedure. Briefly, electrophoresis was used to separate proteins in a sodium dodecyl sulfate polyacrylamide gel and then transferred to a nitrocellulose membrane. The membrane was incubated with different primary antibodies at 4°C overnight after blocking. After incubation with peroxidase-conjugated secondary antibody (HRP fluorescent anti-mouse or anti-rabbit from Cell Signaling Technology) for 2 h at 37°C, the membrane was imaged using the Odyssey two-color infrared laser imaging system (LI-COR Biosciences, Lincoln, NE, USA). The relative intensity was calculated with ImageJ (NIH).

2.17 Immunofluorescent Staining

Sample was washed three times with PBS and incubated in 0.3% Triton-X 100 solution for 15–30 min, and an additional three times washing with PBS was performed again. Non-specific binding points were blocked by blotting buffer for 1 h, and then incubated with primary antibody at 4°C overnight. After washing with PBS for three times, the sample was incubated with 1:200 diluted fluorescent secondary antibody (Alexa Fluor 488 or 594) for 2 h before being stained with 1 μg/mL DAPI dye for 5 min. Cyto-ID autophagy detection kit obtained from Enzo Life Sciences (Farmingdale, NY, USA) was used as described in the manufacturer's instruction. The slide was sealed after dehydration and was observed with a confocal microscope (Becton, Dicknson and Company, New Jersey, USA).

2.18 Transmission Electron Microscopy

The primary neurons were analyzed with a JEM 1230 transmission electron microscope (JEOL, USA Inc). Micrographs were taken at ×5,000 or ×20,000 magnification.

2.19 Statistical Analysis

Statistical analysis of the rat and mouse experimental data was conducted utilizing SPSS version 25 (IBM Corporation, USA), while GraphPad Prism version 9.0 (GraphPad Software Inc.) was employed for the generation of statistical charts. Data presentations were standardized to the format of mean values with their corresponding standard error of the mean (SEM). The Shapiro–Wilk test served as the method for assessing data normality. When data complied with the criteria for both variance homogeneity and normal distribution, the analysis was conducted by one-way ANOVA with Tukey's HSD was implemented. Statistical significance was determined at a threshold of p < 0.05.

3.1 Retrieving and Preprocessing of MCAO scRNA-Seq Data

After preprocessing the data, we employed the harmony algorithm to integrate the samples and effectively remove any potential batch effects. Figure S1A illustrates the representation of the integrated dataset consisting of 3 MCAO and 3 control samples after implementing the harmony algorithm. Following the standard steps of Seurat, we successfully identified a total of 16 clusters, which were visualized using t-SNE, as shown in Figure S1C,D. Then, we identified marker genes in the cell-type populations (Figure S1B). We observed 9 cell clusters (Endothelial Cells: Itm2a, Cldn5, Slco1a4; Microglial cell: Hexb, P2ry12, Selplg; Myeloid cells: Lyz2, Cd74, H2-Ab1; Astrocyte: Aldoc, Clu, Mt3; Pericytes: Acta2, Tagln, Myl9; Glial cell: Plp1, Ptgds, Cldn11; Neutrophils: S100a8, S100a9, Retnlg; NK cells: Ccl5, Nkg7, Ms4a4b; neuron: Mgp, Igfbp6, Nov).

3.2 Profiling of Cells in MCAO at the scRNA Transcript Level and Cell Stemness Analysis

The distribution of cell types was compared between the MCAO and control samples. A significant difference was detected in various types, including astrocytes, glial cells, Endothelial Cells, and myeloid cells (Figure 1A). Figure 1B showed that MCAO progression was mainly associated with NEIL3-mediated resolution of ICLs, Regulation of thyroid hormone activity, TWIK-related potassium channel (TREK), Degradation of GABA, Potassium transport channels, Alanine metabolism, Tandem pore domain halothane-inhibited K⁺ channel (THIK), FGFR1c and Klotho ligand binding and activation, and sterols are 12-hydroxylated by CYP8B1, ATP-sensitive Potassium channels.

By comparing the differences in CytoTRACE scores of cells between MCAO and control groups, we observed that Astrocyte, Endothelial cell, Microglial cell, and neuron showed a lower differentiated state and possessed a higher degree of stemness in MCAO than in control groups (Figure 1C,E). In addition, we identified the top10 genes with the highest correlation out of CytoTRACE scores, including Ctsb, Ftl1, Fcer1g, Pfn1, Tyrobp, Itm2a, Bsg, Slco1a4, Tsc22d1, Spock2 (Figure 1D).

3.3 Calculation of Cell H₂S and Autophagy Score

Since studies indicated that H₂S and autophagy played a significant role in neuroprotection in stroke, we further explored whether H₂S and autophagy define the different cell fate in MCAO. Figure S1 showed that Neutrophils, Myeloid cells, Microglial cells, Astrocytes, and Glial cells displayed a higher hydrogen sulfide level than other cell types. In Neutrophils, Myeloid cells, and Microglial cells, the H₂S level of the MCAO group was significantly lower than that of the control group.

Furthermore, through the correlation analysis between H₂S and positive regulation of autophagy scores, we found the positive correlation in NK cells of sham (r = 0.26, p = 3.26e-03, Figure 2A) and Microglial cell of MCAO (r = 0.09, p = 1.54e-12, Figure 2B). Moreover, the significant increase in correlation was greater in neurons of MCAO (r = 0.26, p = 0.07) compared to the control (r = −0.19, p = 0.14) (Figure 2B), suggesting the crosstalk of H₂S and positive regulation of autophagy plays a key role in neurons.

3.4 H₂S Involves in the Pathophysiology of Ischemic Stroke

Homocysteine (Hcy) has been recognized as a risk factor for stroke [49], and increased reactive oxygen species (ROS) is both a foreshadow and a pathological marker of stroke [49]. Although H₂S has been demonstrated as a pivotal signaling molecule in the central nervous system, no study has shown the age-dependent changes of H₂S in vivo so far. Due to the lipid solubility of H₂S, the serum level of H₂S can sufficiently reflect its concentrations in the brain [50]. Herein, in order to verify the above bioinformatics result and whether H₂S is a potential risk factor like Hcy and ROS or a pathophysiological modulator in ischemic stroke, we measured the content of serum H₂S, Hcy, and ROS in normal subjects and rats with different ages. As shown in Tables 1 and 2, H₂S, Hcy, and ROS in normal subjects and rats exhibited an appreciable age-dependent up-regulation, seemingly implying H₂S might be a risk factor like Hcy and ROS. However, in contrast to the age-dependent up-regulation of Hcy and ROS in ischemic patients and cerebral I/R injury rats (Figure 3A,B), the H₂S level declined dramatically in acute ischemic stroke patients and cerebral I/R injury rats (Figure 3C,D, *p < 0.05, **p < 0.01), indicating that H₂S may act as a pathophysiological modulator rather than a risk factor like Hcy and ROS in ischemic stroke.

3.5 SPRC Blocks Deterioration of Cerebral Function After Rat Cerebral I/R Injury

As an endogenous H₂S modulator, SPRC was widely studied in cardiovascular diseases including ischemic heart disease in rat models. But the therapeutic efficacy of SPRC on cerebral I/R injury has yet to be elucidated. In order to verify whether SPRC could protect against cerebral I/R injury, rats were treated with designed dosage SPRC at defined timepoints (Figure 4A). As shown in Figure 4B, compared to the sham group, rats in the I/R group gained an apparently high neurological deficient score, but SPRC at medium and high dosages could significantly reduce the ethological score like Edaravone treatment when compared with the I/R group (^&p < 0.05, ^&&p < 0.01). Similarly, another ethological indicator, body speed, was also accelerated by SPRC, especially at high dosage (Figure 4C, ^&&p < 0.01). Furthermore, retention time was largely reduced by SPRC and Edaravone treatment after training (Figure 4D, ^&p < 0.05, ^&&p < 0.01). In addition, the Micro PET results disclosed that SPRC, especially at medium and high dosages, significantly increased the standard uptake value (SUV) in the brain compared with the I/R model group, indicating that SPRC could promote the glucose metabolism (Figure 4E,F, ^&&p < 0.01). These data in living body demonstrated the protective potential of SPRC in cerebral I/R injury rats.

3.6 SPRC Ameliorates Neurological Injury Induced by Cerebral I/R Injury in Rats

To further assess the neuroprotective potential of SPRC against ischemic stroke, we next evaluated other indicators at autopsy. Infarct volume was assayed by TTC staining. As presented in Figure 5A,B, when compared to the sham group, the infarct area and percentage of infarct volume in the I/R group were remarkably increased, while SPRC administration could significantly decrease the infarct volume (^&&p < 0.01), which was similar to the effect of positive control Edaravone (^&&p < 0.01). Detection of brain swelling and water content can indirectly reflect the degree of injury after I/R; in the I/R group, the percentage of brain swelling and water content were both largely increased compared to the sham group; however, after SPRC treatment, these two indicators were greatly reduced (Figure 5C,D, ^&p < 0.05, ^&&p < 0.01). We finally further evaluated the ultrastructure changes of brain tissues using transmission electron microscopy. As the graphs in Figure 5E show, in the sham group, the neuronal nucleus in the rat hippocampus was intact and regular in morphology with homogeneous nuclear chromatin, and the neurons were rich in synapses, mitochondria, endoplasmic reticulum, and ribosomes. However, after I/R injury, all of these ultrastructure manifestations SPRC administration could strongly salvage the ultrastructure deterioration. Summarily, these data further demonstrated the definite neuroprotection of SPRC against cerebral I/R injury.

3.7 Neuroprotection of SRPC Is Not Completely Dependent on Synthetase CBS and 3-MST

As above mentioned, in mammals, endogenous H₂S is mainly produced by three synthetases: cystathionine-γ-lyase (CSE), cystathionine-β-synthase (CBS) and 3-mercaptopyruvat sulfurtransferase (3-MST). To once again confirm that CBS and 3-MST are the two main synthetases in the brain to biosynthesize H₂S, we detected their protein expression as well as another synthetase CSE in the lifetime of rats and mice. Our results showed that CBS and 3-MST were indeed expressed stably in the life (from newborn to 96 weeks) of rats (Figure 6A), but conditions were opposite regarding the expression of CSE, though slight upregulated expression was found after 4.5 h I/R injury (Figure 6B). Similar results were verified in mice (Figure 6D). Besides, only CBS and 3-MST were successfully located by immunofluorescence in rat brain tissues (Figure 6C). These data from Figure 6A through D just demonstrated that CBS and 3-MST are indeed the two main synthetases that produce H₂S in the brain.

Since SPRC is a putative endogenous H₂S modulator, based on the above results, we next intended to test whether the neuroprotective role of SPRC is dependent on CBS and 3-MST. Therefore, we evaluated the role of SPRC in CBS and/or 3-MST transgenic mice with I/R injury. As the results showed, compared with the non-transgenic mice with I/R injury, the neurological score, brain swelling, and water content of CBS^+/− mice suffering from I/R slightly increased, and conditions were much more serious in CBS^+/− + 3-MST-cerebral knock down (CBS^+/− + 3-MST-ckd) mice after I/R because these indicators were further significantly heightened. However, SPRC treatment could still greatly degrade these damage parameters in CBS^+/− + 3-MST-ckd mice (**p < 0.01). In TTC staining, similar results were observed that SPRC significantly reduced the infarct volume compared with the CBS^+/− + 3-MST-ckd I/R group (**p < 0.01). These data suggested that knockdown of H₂S synthetases CBS and 3-MST could aggravate cerebral I/R injury, and the neuroprotective role of SPRC seemed independent of the synthase activity of CBS and 3-MST.

3.8 SPRC Protects Cerebral I/R Injury via Preserving the Endogenous Level of H₂S

Since CBS and 3-MST were dispensable for the neuroprotective role of SPRC, we doubted whether SPRC could still improve the level of H₂S to exert its neuroprotective effects. Therefore, we next detected the H₂S level after SPRC administration in normal rats. Results in Figure 7A manifested that serum H₂S level rapidly increased in a dosage-dependent manner after SPRC administration, and the concentration peak was reached 1 h later after SPRC administration. Meanwhile, the H₂S level was also assayed in I/R rats with/without SPRC treatment. The results showed that the H₂S level in I/R rats was significantly decreased in the first 6 h compared to that in the sham group; while the H₂S concentration surprisingly presented a gradual increase 6 h after I/R (Figure 7B,C). However, after SPRC treatment, the H₂S level could be upregulated in the first 6 h and then downregulated after 6 h (Figure 7B,C, *p < 0.05, **p < 0.01, ***p < 0.001). These data preliminarily demonstrated that SPRC could preserve the H₂S level after I/R probably through direct H₂S releasing.

3.9 SPRC Promotes Protective Autophagy to Block Neurological Injury

Autophagy has been demonstrated as an important pathophysiological factor of ischemic stroke [30]. In order to explore the cellular and molecular mechanisms under the neuroprotective role of SPRC, we tried to assay the impact of SPRC on autophagy. LC 3-II is a typical hallmark of autophagy; results showed that the expression of LC 3-II was gradually decreased with the aging process under physiological conditions (Figure 8A, *p < 0.05), but could be sharply upregulated after rat I/R injury and oxygen glucose deprivation (OGD) in neurons (Figure 8B,C, *p < 0.05, **p < 0.01 and ***p < 0.001), implying that autophagy may play a role as a guard against ischemic stroke in our research. The following data in Figure 8G,H further verified that autophagy was protective after ischemic injury, as after inhibition of autophagy by 3-MA, cell viability was remarkably weakened while LDH release was strongly increased after OGD injury of neurons (compared with the control group, **p < 0.01, ***p < 0.001). Furthermore, SPRC treatment could significantly improve cell viability and block LDH release after neuron OGD injury (compared to the model, *p < 0.05, **p < 0.01, ***p < 0.001). Although neuronal injury was exacerbated, presented by lower cell viability and much more LDH release after autophagy inhibition (Figure 8G,H, ***p < 0.001), SPRC was still able to alleviate neuronal injury compared to the 3-MA group (**p < 0.01, ***p < 0.001). Additionally, in morphology, after OGD, neurons were observed shrinking and lacking synapses under the optical microscope (Figure 8D), and autophagosomes were apparently activated and increased under the transmission electron microscope and fluorescent microscope (Figure 8E,F). However, SPRC could mitigate these adverse morphological changes and improve autophagy activation, whether autophagy was inhibited by 3-MA or not (Figure 8D through F). Finally, results of protein expression showed that SPRC indeed can further increase the expression of LC3-II and decrease the expression of p62, though autophagy was inhibited by 3-MA (Figure 8I, *p < 0.05, ***p < 0.001).