Animal experiments were approved by the Local Ethical Committee in accordance with the Guide for the Care and Use of Laboratory Animals. Wild-type male Sprague–Dawley rats (180–200 g weight) were provided by SiPeiFu Biotechnology (Beijing, China) and housed in pathogen-free condition. Middle cerebral artery occlusion (MCAO) was used to induce IS as previously described (Wang P. et al., 2021). Briefly, rats were anesthetized by intraperitoneal injection of pentobarbital sodium (40 mg/kg) and then fixed in a supine position to expose the neck area. An incision was made along the median neck to separate the common carotid artery (CCA), internal carotid artery, and external carotid artery (ECA) from adjacent tissues and vagus nerve. After the distal ECA was tied off and opened by arteriotomy, a nylon monofilament was inserted and advanced upwards about 1 cm past the CCA bifurcation. After 90 min of occlusion, the suture was removed for reperfusion (N = 6, Model group). During the surgery, all rats were maintained at 37°C to avoid death. Similar surgery without occlusion was performed in the Sham group (N = 6).

Recombinant adenoviruses carrying shRNA-ATF3 (ad-shATF3-1 and - 2) and corresponding negative control (ad-shNC) were synthesized in RiboBio (Guangzhou, China). Rats were randomized to model, ad-shATF3-1, ad-shATF3-2, and ad-shNC groups (N = 6 each group). The recombinant adenoviruses (2 × 10¹² vg) were intraparenchymal injected into rats at 1 week before MCAO. Subsequently, the MCASO or sham surgery was performed as mentioned above. After the reperfusion for 24 h, rats were finally sacrificed by intraperitoneal injection of an overdose of pentobarbital sodium (150 mg/kg), and the brain tissues were collected.

The brain tissues were frozen at −20°C for 30 min and then sliced into 2 mm coronal sections. Subsequently, the sections were incubated with 2% TTC solutions (Solarbio, China) for 15 min at 37°C in the dark. After fixed in 4% paraformaldehyde for 24 h, the brain samples were observed under a stereomicroscope (Olympus, Japan). The unstained white area presented the infarcted tissues, and stained red area presented non-infarcted tissues.

For HE staining, the brain tissues were fixed in 4% paraformaldehyde, dehydrated in graded ethanol, paraffin-embedded, and sliced into 5 μM sections. The sections were then dewaxed in xylene, rehydrated in graded ethanol, and stained with Hematoxylin for 5 min and with Eosin for 2 min. After dehydration and vitrification, the stained sections were observed under an inverted microscope (Olympus).

For IHC staining, the paraffin-embedded sections were prepared as mentioned above. After dewaxing and dehydration, the sections received 10 min of microwave irradiation in 10 mM citrate buffer for antigen retrieval. The sections were then blocked with goat serum in PBS, incubated with anti-caspase-3 (1:1000, Cell Signaling Technology, Danvers, MA, United States) overnight at 4°C, and continuous incubated with horseradish peroxidase-conjugated secondary antibody (1:3000, Cell Signaling Technology) for 1 h at 37°C. Subsequently, the sections were stained with diaminobenzidine for 10 min and counterstained with hematoxylin for 1 min. Followed by dehydration and vitrification, the stained sections were observed under an inverted microscope (Olympus).

The TUNEL assay was carried out using a TUNEL Apoptosis Assay kit (Beyotime, Beijing, China). Briefly, the deparaffined/rehydrated tissue sections were incubated with DNase-free Proteinase K for 20 min, and then with TUNEL solution for 1 h in the dark. After counterstained with DAPI for 10 min, the apoptotic cells were observed and counted under a fluorescence microscope (Olympus).

Brain tissue samples from the Sham and Model groups were used for transcriptome sequencing. Briefly, total RNAs isolated from the samples were qualified by Agilent 2,100 Bioanalyzer. mRNAs were purified from total RNAs via polyA-selection, and randomly fragmented into 200–300 bp using metal ion-catalyzed hydrolysis. Subsequently, cDNA library (about 450 bp) was constructed and sequenced on Illumina Novaseq 6,000 platform. DEGs (|log2FoldChange| > 1 and p < 0.05) were clustered by Pheatmap, and functional enriched by Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses.

Total RNA was extracted from the brain tissues using TRIzol reagent (Beyotime). cDNA was reverse transcribed from total RNA using FastKing RT Kit (TIANGEN, Beijing, China). Quantitative real-time PCR (qRT-PCR) was performed using SYBR Green PCR Master Mix (Bio-Rad, Hercules, CA, United States) on a LightCycler® 480 System (Roche, Basel, Switzerland) with standard procedures. Relative expression level was calculated following the 2^−ΔΔCt method using GAPDH as the internal control. The primer sequences were shown in Supplementary Table S1.

Protein samples were extracted from the brain tissues by lysing in RIPA buffer (Beyotime). The isolated proteins were separated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidenefluoride membranes. The membranes were then blocked with 5% non-fat milk for 1 h, and incubated with primary antibody (anti-ATF-3, -GPX4, -COX2, and -GAPDH; 1:1000, Cell Signaling Technology) for 12 h at 4°C. After incubated with horseradish peroxidase-conjugated secondary antibody (anti-rabbit IgG, 1:3000, Cell Signaling Technology) for 1 h at 37°C, the protein blots were visualized using Electrochemiluminescence Kit (Beyotime) and analyzed by a Gel Imaging System (Tanon 5,200, Shanghai, China) with ImageJ software.

The iron content in brain tissues was measured using Iron Assay Kit (Sigma, St. Louis, MO, United States) in accordance with the manufacturer’s instructions. Simply, the tissue samples were homogenized on ice and centrifuged at 13,000 g for 10 min. Subsequently, the supernatants were incubated with iron detection buffer for 30 min at 25°C. The optical density at 593 nm was measured using a microplate reader. The iron content was finally calculated according to the standard curves.

The GSH and MDA in brain tissues were measured using GSH assay kit (Solarbio) and MDA assay kit (Solarbio), respectively. Tissue homogenates were used for measurements strictly in accordance with the manufacturer’s instructions. The contents of GSH and MDA were finally calculated according to the standard curves.

Data were expressed as mean ± standard deviation (SD), and statistical analyzed by GraphPad Prism 8.0 (GraphPad, San Diego, CA, United States). Student’s t-test was used for the comparisons between two groups. One-way ANOVA followed by Tukey’s post-hoc test was used for the comparisons among multiple groups. A p value less than 0.05 presented statistically significant.

A rat model of IS was established by MCAO. TTC staining was performed to determine the infarcted area in brain tissues, which could react with active dehydrogenases in normal tissues (red), but not with inactive dehydrogenases in infarcted tissues (white). Consistent with the expectation, the Model group showed an obvious white straining compared with the Sham group (Figure 1A). HE staining revealed interstitial edema, inflammatory cell infiltration, and disordered cell shape in the Model group in contrast to the Sham group (Figure 1B). In addition, more cells positive to caspase-3 were observed by IHC staining in the Model group than that in the Sham group, reflecting the elevated cell apoptosis in the model rats (Figure 1C). The enhanced apoptosis was further determined by TUNEL staining in the Model group when compared with the Sham group (p < 0.01; Figure 1D). All the results above proved the successful establishment of IS model in rats.

Following the characterization of the rat model of IS, we focused on the DEGs at the transcriptional level. Total 1,160 DEGs were identified by transcriptome sequencing between the Model and Sham groups, including 699 up-regulated DEGs and 461 down-regulated DEGs (Figures 2A,B). The top 30 up-regulated genes and top 10 down-regulated genes were shown in Table 1. Subsequently, the isolated DEGs were enriched by GO and KEGG analyses. GO functional analysis showed that the DEGs were mainly enriched in system development, regulation of multicellular organismal process, positive regulation of biological process, and anatomical structure development, etc (Figures 2C,D). KEGG pathway analysis revealed the enrichment of DEGs in cytokine-cytokine receptor interaction, neuroactive ligand-receptor interaction, transcriptional misregulation in cancer, MAPK signaling pathway, etc (Figures 2E,F). These results above indicated that IS is related with the expression changes of multiple functional genes involving in a variety of biological processes.

Several DEGs relating with ischemic injury were selected from Table 1 for validation. qRT-PCR showed that the mRNA expression of Hspa1b, Tfpi2, Ptx3, Atf3 were significantly higher in the Model group compared with those in the Sham group (p < 0.05; Figure 3A). On the contrary, the mRNA expression of Smyd1 and Tacr2 were significantly lower in the Model group than those in the Sham group (p < 0.05; Figure 3B). These results of qRT-PCR were just consistent with those of transcriptome sequencing, and indicated these DEGs might be candidates for further studies on IS.

Among the validated DEGs, only ATF was related with both IS and ferroptosis as previously described. Therefore, we focused on ATF3 for functional studies. Western blot further identified a significantly higher protein expression of ATF3 in the Model group than that in the Sham group (p < 0.01; Figure 4A). Subsequently, ATF3 was knocked down in the model rats via injection of Ad-shATF3-1/2. As expected, the mRNA expression of ATF3 was significantly decreased by the injection of Ad-shATF3-1 and − 2 in the model rats (p < 0.01; Figure 4B). Interestingly, the infarcted area (white) in the model rats was obviously reduced by the injection of Ad-shATF3 (Ad-shATF3-1 was used in following assays), but not by the infection of Ad-shNC (Figure 4C). Besides, the interstitial edema, inflammatory cell infiltration, and disordered cell shape were recovered in the Ad-shATF3 group, closing to histomorphology in the Sham group (Figure 4D). Caspase-3 positive cells were also significantly reduced in the Ad-shATF3 group compared with that in the Model group (Figure 4E). The alleviated cell apoptosis in the Ad-shATF3 group was further confirmed by TUNEL assay (p < 0.01; Figure 4F).

Whether the action mechanism of ATF3 in IS is associated with the ferroptosis was further analyzed. COX2 (a negative regulator of ferroptosis) and GPX4 (a central positive regulator of ferroptosis) are two indicators of ferroptosis. The Model group showed higher protein expression of COX2 and lower protein expression of GPX4 compared to the Sham group (p < 0.01). The intervention of Ad-shATF3 significantly down-regulated COX2 and up-regulated GPX4 in the model rats (p < 0.01; Figure 5A). Excess intracellular iron is another indicator for ferroptosis, which can elevate reactive oxygen species levels and subsequent lipid peroxidation. The Model group showed a higher iron content than the Sham group (p < 0.01), and the abnormal increased iron content in the model rats was alleviated by the intervention of Ad-shATF3 (p < 0.01; Figure 5B). Furthermore, a lower content of GSH was observed in the Model group than that in the Sham group (p < 0.01). After knockdown of ATF3 in the model rats (the Ad-shATF3 group), the low GSH level was increased to some degrees (p < 0.01; Figure 5C). In contrast to GSH, the MDA level was higher in the Model group compared to the Sham group (p < 0.01). The intervention of Ad-shATF3 weakened the elevation of MDA level in the model rats (p < 0.01; Figure 5D). All the results above reflected that knockdown of ATF3 inhibited the ferroptosis in the model rats, contributing to the remission of IS.