Eight- to 10-week-old male C57BL/6 mice (SLAC Laboratory Animal Co., Ltd., Shanghai, China) were housed in a temperature- and humidity-controlled room under a 12-h day/night cycle and had free access to food and water. All protocols were approved by the Institutional Ethics Committee of the Second Affiliated Hospital, Zhejiang University School of Medicine. The animal experiments were performed according to the National Institutes of Health’s Guide for the Care and the Use of Laboratory Animals and the ARRIVE (Animal Research: Reporting in vivo Experiments) guidelines.

As previously described, PLX3397 (Selleckchem, Houston, TX, United States) was formulated in AIN-76 A standard chow at a concentration of 290 ppm. Mice were fed with PLX3397 chow for 21 consecutive days (Najafi et al., 2018) until the end of experiments.

The endovascular perforation model was established as previously described (Muroi et al., 2015). Briefly, mice were anesthetized with pentobarbital sodium (40 mg/kg) via intraperitoneal injection. Left carotid artery and its bifurcation were exposed. Then, 5-0 sharpened nylon suture was inserted into the internal carotid artery (ICA) from the external carotid artery. Then, the suture was pushed until the tip reached the intracranial bifurcation of anterior cerebral artery and middle cerebral artery. The suture was pushed 1 mm further to perforate the vessel. The mice in the sham group underwent the same procedures except perforation. Since all procedures were performed on the left side, we collected the left hemisphere for all downstream experiments. Additionally, we assessed the degree of SAH via the grading system as previously described (Sugawara et al., 2008).

Modified Garcia test (range, 0–18) was used to assess the short-term neurological performance, by evaluating spontaneous activity, climbing, forelimb stretching, spontaneous movements of all limbs, body proprioception, and response to vibrissae touch (Shi et al., 2018).

To access the motor coordination and sensory neglect after SAH, adhesive test was performed following previous studies (Bouet et al., 2009). Small adhesive tape strips (2 mm × 3 mm) were applied to mice forepaws. Contact time and removal time were recorded with a maximum observation time of 120 s.

Mice were anesthetized with 40 mg/kg of pentobarbital sodium and transcardially perfused with ice-cold PBS. Brain tissues were obtained, and cerebrum was dissected and separated into ipsilateral and contralateral hemispheres. Ipsilateral hemispheres were mechanically dissociated using a razor blade and placed in a 15-ml conical tube with digestion solution [0.6 mg/ml of collagenase D (Sigma)]. Then, the mixture were incubated for 30 min at 37°C. After that, a 70-μm strainer was used to generate a single-cell suspension (BD FALCON). Cells were isolated by centrifugation (30 min, 800 × g at 23°C) using 30–70% Percoll gradient solutions (GE Healthcare) (Agalave et al., 2020). Isolated cells were washed and resuspended in PBS with 0.01% bovine serum albumin (BSA) and then incubated with indicated anti-mouse antibodies for 30 min at 4°C [rat anti-mouse CD45 PerCP (BD Bioscience) and rat anti-mouse CD11b FITC (BD Bioscience)]. The population of microglia (CD45 positive and CD11b positive) was sorted.

In flow cytometry, rat anti-mouse Ly6G PE and rat anti-mouse Ly6C APC (BD Bioscience) antibodies were used and incubated with CD45 perCP and CD11b FITC.

Total RNA from microglia sorted by FACS was isolated using TRIzol (Invitrogen, CA, United States) according to the manufacturer’s protocol. The total RNA quantity and purity were checked by an Agilent 2100 bioanalyzer. High-quality samples (RIN number > 6.8) were used for downstream sequencing. Sequence libraries were constructed according to the standard SMART-seq protocol, and paired-end sequencing was performed with Illumina Novaseq 6000 (LC Bio) following the vendor’s recommended protocol. Prior to assembly, low-quality reads that contain sequencing adaptors, sequencing primers, or low-quality nucleotides, were removed. The sequence quality was also checked with FastQC. HISAT was used to align and map reads to the UCSC¹ GRCm38 mouse reference genome. The mapped reads of each sample were assembled using StringTie. Then, all transcriptomes from the samples were merged to reconstruct a comprehensive transcriptome using perl scripts. After the final transcriptome was generated, StringTie and edgeR were used to estimate the expression levels of all transcripts. StringTie was used to perform expression level for mRNAs by calculating Fragment per Kilobase of transcript per Million mapped reads (FPKM). All raw sequence data have been uploaded to GSE167957.

The expression matrixes were counter-checked to determine if there were any systematic errors or batch effects. The sva R package was used for identifying, estimating, and removing batch effects (Leek et al., 2012). The differentially expressed genes (DEGs) were selected with fold change > 2 or fold change < −2 and with statistical significance (Benjamini–Hochberg adjusted p-value < 0.01) by DESeq2 package (Love et al., 2014). FPKM were used for gene expression and were log2(x + 1) transformed. Principal components analysis (PCA) was performed on normalized counts. R package pheatmap was used to generate heatmap.

Functional enrichment analysis was conducted using Metascape² (Bhattacharya et al., 2018). All genes in mouse genome were used as background genes, and the default settings were used as enrichment criteria (minimum overlap = 3, P-value cutoff = 0.01, minimum enrichment score = 1.5). A gene ontology term was considered activated/increased with a z-score > 2 and a p-value < 0.01, and was predicted inhibited/decreased with a z-score < −2 and a p-value < 0.01.

Transcription factors (TFs) of selected DEGs were predicted using IRegulon plugin in Cytoscape (Janky et al., 2014). TFs were ranked by normalized enrichment score (NES), and NES > 3 was set as threshold.

The protein–protein interaction (PPI) network of selected DEGs was established using The Search Tool for the Retrieval of Interacting Genes (STRING³) (Szklarczyk et al., 2015) and then visualized in Cytoscape (Shannon et al., 2003). A combined score > 0.4 was set as a significant threshold. Furthermore, CytoHubba, a Cytoscape plugin, was used to explore hub genes in the constructed PPI network, and the top 10 genes were displayed based on degree method.

Immunofluorescence staining was performed as described previously (Lu et al., 2019). In brief, mice were sacrificed and transcardially perfused with 0.9% NaCl, followed by 4% paraformaldehyde. Brain tissues were harvested and immersed in 4% paraformaldehyde for 24 h and then cryoprotected in 30% sucrose solution. Frozen serial coronal brain sections (9 μm) were prepared and fixed on slides. Slices were blocked with 5% donkey serum for 1 h and then incubated at 4°C overnight with primary antibodies, including goat anti-Iba1 antibody (1:500, Abcam, ab5076), rabbit anti-Tlr2 antibody (1:100, Abcam, ab209216), and mice anti-Irf7 antibody (1:200, Santa Cruz, sc-74471). After washing, the cryosections were incubated at 37°C for 1 h with the following secondary antibodies: donkey-anti-goat IgG(H + L) Alexa Fluor 594 (1:500, Thermo Fisher, A-11058), donkey-anti-rabbit IgG (H + L) Alexa Fluor 488 (1:500, Thermo Fisher, A-21206), and goat-anti-mouse IgG (H + L) Alexa Fluor 488 (1:500, Thermo Fisher, A-11001). Finally, the sections were observed and images were taken using an Olympus fluorescence microscope (Olympus Co., Japan).

RNA-seq data were analyzed as mentioned above. For other data, the statistical analyses were conducted using R software (version 3.6.3) and GraphPad Prism (version 8.0.2). Student’s t-test and the Kruskal–Wallis test were employed in the two-group comparisons. A two-tailed P-value of <0.05 was considered statistically significant without specific annotation.

At first, we induced SAH in adult C57BL/6 mice; the neurological scores and behavior tests were assessed in both SAH and sham group (Figure 1A). At 24 and 72 h after SAH, the average neurological scores were 11.714 ± 0.881 and 14.167 ± 1.675, respectively. SAH mice showed worse neurological scores than the sham group (p < 0.001; Figure 1B). At 24 h after SAH, compared with the sham group, SAH mice showed significantly increased contact times and removal times, respectively (p < 0.001; Figures 1C,D). Similarly, at 72 h, increased contact times and removal times were observed in SAH mice compared to sham mice (p < 0.001; Figures 1C,D). Representative images of Iba1 staining in cortex, striatum, and hippocampus are shown in Figure 1E. Significant morphological changes, like larger bodies, thicker pseudopodia, and ameboid morphology, were displayed in SAH mice, indicating microglia activation at 72 h after SAH.

To examine whether microglia play a detrimental role in SAH, we administrated the CSF1R inhibitor PLX3397 or control diet in chow to deplete microglia (Figure 1F). The number of microglia (CD45^low/CD11b⁺) was assessed through flow cytometry. Administration of PLX3397 to sham mice resulted in an 65% reduction in microglia compared to control mice (p < 0.0001; Figure 1G). Next, we induced SAH in mice pretreated with PLX3397 or control diet, and maintained their chow until sacrifice (Figure 1F). Immunofluorescence showed that the number of microglia in the PLX3397-fed group was significantly lower than the control diet group at 72 h after SAH (Figure 1H). PLX3397 treatment had no apparent effect on the neurological scores at 72 h after SAH (Figure 1I). In the adhesive removal test, there was no difference in average contact times between the PLX3397-fed group and the control diet group (Figure 1J). However, PLX3397-fed mice spent less time removing the tape at 72 h after SAH compared with the control diet group (p < 0.05; Figure 1K). Taken together, these data suggested that microglia were activated after SAH, and they may positively correlate with neurological dysfunction.

To elucidate the functional roles of microglia after SAH, we used FACS to purify the microglia population (CD45⁺/CD11b⁺) from the brains of mice 72 h after SAH or sham operation. Sorted microglia cells were subjected to bulk RNA-seq (Figure 2A). We confirmed the high expression level of microglial marker in FACS-sorted cells, such as Itgam (encoding CD11b), Cx3cr1, Aif1 (encoding Iba1), Csf1r, Mertk, Tmem119, Siglech, and P2ry12 (Figure 2B). The established markers of neurons (Map2 and Rbfox3), astrocytes (S100b and Gfap), oligodendrocytes (Tppp3, Apc, and Cspg4), and other immune cells (Cd3e, Cd19, Ly6g, Mrc1, and Cd163) were all expressed at low level (Figure 2C). The result of PCA performed on RNA-seq data showed that samples in the same group clustered together, and samples in different groups separated clearly (Figure 2D), indicating robust transcriptomic differences between the SAH and Sham group.

Next, to determine the transcriptomic changes of microglia induced by SAH, differential expression analysis was performed. A total of 1576 genes were identified as DEGs (fold change > 2 or <−2, Benjamini–Hochberg adjusted p-value < 0.01), which contained 928 upregulated genes and 648 downregulated genes in post-SAH microglia compared with sham hemispheres (Figure 2E and Supplementary Table 1), indicating strong alterations of the genome (6.85% of total ∼23,000 genes). To elucidate the functional alterations in post-SAH microglia, pathway enrichment analysis was performed on the DEGs by an online tool Metascape. Regulation of biological process (21.79%), immune system process (18.18%), and chemotaxis and adhesion (9.36%) accounted for a major proportion of GO terms obtained from Metascape (Figure 2F). The top 20 clustered GO terms and specific GO terms obtained by Metascape are shown in Supplementary Tables 2, 3, respectively.

We analyzed the biological processes related to immune inflammatory responses in microglia after SAH, and the Top 20 GO terms predicted to be activated (z-score > 2) are shown in Figure 3A. The inflammatory response, defense response, innate immune response, adaptive immune response, and neuroinflammatory response were all involved.

As chemotaxis was a large cluster that participates in immune inflammatory response, we further compared the pathways that related to cytokines and chemokines. 17 GO terms were exhibited in Figure 3B, all of them were predicted to be activated (z-score > 2). We also screened the expression level of a panel of cytokine and chemokine genes (Bhattacharya et al., 2018), and there were 17 significantly upregulated genes. The upregulated genes involved seven chemokine-encoding genes (Cxcl1, Cxcl2, Cxcl3, Cxcl10, Cxcl11, Cxcl14, and Cxcl16), and 10 cytokine-encoding genes, such as Ccl6, Ccl9, Il1b, Il23a, Ifng, Il33, and Mif (Figure 3C). These data indicated that a variety of pro-inflammatory cytokines and chemokines were released from post-SAH microglia, to recruit peripheral immune cells into brain parenchyma, and it also suggested that myeloid cells were the main target of upregulated cytokines and chemokines (Figure 3C). Then, we detected the different populations of myeloid cells in ipsilateral hemispheres of SAH and sham mice by flow cytometry. Compared to sham mice, there were significant increases in the amount of infiltrating immune cells in post-SAH hemispheres, such as CD45⁺Ly6C⁺Ly6G^– monocytes (p < 0.01) and CD45⁺Ly6C^–Ly6G⁺ neutrophils (p < 0.05) (Figures 3D–F).

Next, we investigated the DEGs involved in cytokine- and chemokine-related pathways. These upregulated products could also be defined as receptors (e.g., TLR2), cytokines (e.g., IFNG and IL1B), and chemokines (e.g., CXCL1 and CXCL10). The top 15 genes, including Ifng, Cxcl1, Cxcl10, Tlr2, and Il1b, were associated with at least three functional subcategories of cytokine- and chemokine-related pathways (Figure 4A). In addition, we also screened out the genes that contribute multiple steps in immune inflammatory response, including inflammatory response, regulation of innate immune response, neuroinflammatory response, and so on (Figure 4B). The products of these genes predicted to be active included TFs (e.g., NR1H3 and IRF7), receptors on the membrane (e.g., CD40 and TLR2), nuclear receptor (e.g., PLSCR1), enzyme (e.g., DNASE1L3), and inflammatory-related proteins (e.g., ADAM8, RSAD2, and ZBP1).

In order to investigate the upstream regulators of these biological processes, the interactions between all DEGs, which consisted of 985 nodes and 11,456 edges, were constructed based on STRING database (Supplementary Table 5). Then, the nodes related to high-frequency DEGs in Figures 4A,B were retrieved, out of which 11 unigenes were TFs (Figure 4C). In our dataset, the expression of Irf7, Spib, and Parp12 were significantly upregulated, whereas Nfkb1 was downregulated in post-SAH microglia. The upstream regulator analysis was performed by IRegulon, a plugin in Cytoscape. The top 10 potential TFs that target the selected DEGs are shown in Table 1. According to the transcriptional level and the results from IRegulon, IRF7 was predicted to be a potent master regulator of the transcriptional processes related to immune inflammatory response and chemotaxis. In the TF target gene network, IRF7 was predicted to target 19 genes in the selected high-frequency DEGs (e.g., Cd40, Il1b, Cxcl10, and Il23a) (Figure 4D). Moreover, Figure 4E demonstrated that IRF7 involved a GO term that was predicted to be activated. Using dual immunofluorescence staining of Iba1 and IRF7, we confirmed that IRF7 expression was induced in microglia at 72 h after SAH (Figure 4F). IRF7 immunosignal was detected in the cortex from sham mice, while the IRF7 was hardly colocalized with Iba1.

Following SAH, microglia were activated to cope with the disruption of brain homeostasis. We examined the GO terms related to “cellular response” and found that response to stimulus (e.g., response to virus and response to external stimulus) and response to cytokines (e.g., response to interferon and response to macrophage colony-stimulating factor) were enriched in post-SAH microglia (Figure 5A). We also examined SAH-induced gene expression changes of receptor on the membrane in microglia (Figure 5B), most of which were upregulated (e.g., Bst1, Gpnmb, Fgr, and Tlr2). Figure 5B also showed that several receptors participated in microglia “cellular response,” including response to external stimulus and response to interferon. Additionally, hub genes of the selected high-frequency genes were identified by CytoHubba, and the top 10 genes were calculated based on Degree algorithm (Figure 5C and Supplementary Table 4). All of the hub genes were upregulated except Tlr3, and Irf7 was the only one identified as TF. In addition, Tlr2 was also regarded as a hub gene, and the interaction between TLR2 and IRF7 indicates that the TLR2/IRF7 signaling axis may mediate the immune inflammatory response and chemotaxis in post-SAH microglia. Furthermore, immunostaining confirmed that the expression of TLR2 was elevated and colocalized with Iba1 in post-SAH brain (Figure 5D).