Spi1 regulates the microglial/macrophage inflammatory response via the PI3K/AKT/mTOR signaling pathway after intracerebral hemorrhage

Abstract

Preclinical and clinical studies have shown that microglia and macrophages participate in a multiphasic brain damage repair process following intracerebral hemorrhage. The E26 transformation-specific sequence-related transcription factor Spi1 regulates microglial/macrophage commitment and maturation. However, the effect of Spi1 on intracerebral hemorrhage remains unclear. In this study, we found that Spi1 may regulate recovery from the neuroinflammation and neurofunctional damage caused by intracerebral hemorrhage by modulating the microglial/macrophage transcriptome. We showed that high Spi1 expression in microglia/macrophages after intracerebral hemorrhage is associated with the activation of many pathways that promote phagocytosis, glycolysis, and autophagy, as well as debris clearance and sustained remyelination. Notably, microglia with higher levels of Spi1 expression were characterized by activation of pathways associated with a variety of hemorrhage-related cellular processes, such as complement activation, angiogenesis, and coagulation. In conclusion, our results suggest that Spi1 plays a vital role in the microglial/macrophage inflammatory response following intracerebral hemorrhage. This new insight into the regulation of Spi1 and its target genes may advance our understanding of neuroinflammation in intracerebral hemorrhage and provide therapeutic targets for patients with intracerebral hemorrhage.

Introduction

Intracerebral hemorrhage (ICH) accounts for 15–20% of all strokes and is associated with high rates of mortality, disability, and morbidity worldwide (Cordonnier et al., 2018; Krishnamurthi et al., 2020; Rosand, 2021). Thus, an integrated understanding of the mechanisms underlying brain injury after hemorrhage is urgently required to improve diagnosis and management. Previous studies have shown that neuroinflammation, which is caused by macrophage infiltration and local microglia, contributes to the development of multiphasic damage after ICH (Bai et al., 2020; Fang et al., 2020; Mei et al., 2021b). In response to hemorrhagic injury, resident microglia and macrophages rapidly migrate towards the injury site, which results in morphological and functional changes (Lan et al., 2011; Zhao et al., 2015; Poon et al., 2017). In the acute phase of ICH, microglia/macrophages promote secondary injury (Gao et al., 2022; Zhang et al., 2022) and hematoma expansion (Morotti et al., 2016) through the secretion of immunomodulatory molecules such as cytokines and chemokines (Wang, 2010; Mracsko and Veltkamp, 2014; Li and Barres, 2018), which further increases the risk of mortality (Adeoye et al., 2014; Walsh et al., 2015) and early disability (Hammond et al., 2014). Conversely, activated microglia/macrophages in the later phase of ICH phagocytose erythrocytes, dying cells, and debris resolve edema, and maintain homeostasis in the perihematomal microenvironment, which supports the recovery of injured neurons (Aguzzi et al., 2013; Chang et al., 2018; Xu et al., 2020; Li et al., 2021; Chen et al., 2022), restores white matter integrity, and promotes functional recovery after ICH (Wan et al., 2016; Lin et al., 2017). Thus, research into the regulation of microglial/macrophage function is critical for understanding ICH pathophysiology.

The E26 transformation-specific sequence family transcription factor gene Spi1, which encodes the PU.1 protein, plays a pivotal role in the development of myeloid and lymphoid lineage cells (Anderson et al., 1998; DeKoter et al., 1998). Several studies have suggested that Spi1 is essential for macrophage function and survival (Weigelt et al., 2009; Lawrence and Natoli, 2011). Spi1 promotes macrophage polarization and survival of mature macrophages (Karpurapu et al., 2011; Qian et al., 2015). In macrophages, Spi1 induces miR-233 and represses E2F1, which regulates cell cycle progression (Denechaud et al., 2016; Solomon et al., 2017). In the central nervous system (CNS), Spi1 expression is limited to microglia (Huang et al., 2017), and its expression levels affect microglial transcription, activation (Zhou et al., 2019), and phenotype (Cakir et al., 2022). Reducing Spi1 expression in microglia results in reduced phagocytic capacity (Smith et al., 2013; Huang et al., 2017; Rustenhoven et al., 2018), whereas increased Spi1 expression in microglia results in increased zymosan phagocytosis (Pimenova et al., 2021). In Alzheimer’s disease, Spi1 controls the expression of interleukin-33 (IL-33), which activates MHC-II⁺ phagocytic microglia and enhances microglia-mediated amyloid-beta clearance (Lau et al., 2020). Spi1 overexpression in human cortical organoids (mhCOs) generates microglia-like cells that protect the parenchyma from Aβ-associated cellular and molecular damage (Cakir et al., 2022). Although Spi1 has been shown to be expressed in the brain, its role in ICH remains uncharacterized. Understanding how Spi1 affects microglial/macrophage transcriptional responses could provide more specific targets for ICH therapies.

In recent decades, high-throughput technologies have been widely used to monitor transcriptomic and proteomic changes at the genome level (Kolodziejczyk et al., 2015; Reuter et al., 2015; Goodwin et al., 2016). In this study, we analyzed six datasets related to ICH from the NCBI Gene Expression Omnibus (GEO) to comprehensively explore the role of Spi1 in microglial/macrophage transcriptome changes and inflammation in the brain following ICH. The aim of the study was to systematically investigate the effects of Spi1 on microglial/macrophage function following ICH in order to advance our understanding of ICH pathophysiology and provide innovative opportunities for ICH treatment.

Methods

Data collection and identification of differentially expressed genes

The six publicly available RNA-Seq and ChIP-seq datasets of patients with ICH and experimental mouse models used in this study were obtained from GEO using the R package GEOquery (Davis and Meltzer, 2007). Detailed information on the datasets used in this study is provided in Additional Table 1. All matrix data were processed using RStudio (R version 4.1.2) (Team, 2022; Posit, 2023). The R packages DESeq2 (version 1.34.0) (Love et al., 2014) and limma (version 3.50.0) (Ritchie et al., 2015) were used to select differentially expressed genes (DEGs) with an adjusted threshold setting of P < 0.05; log2 fold change (log2FC) was specified in every dataset (Love et al., 2014; Ritchie et al., 2015). DEGs were visualized as volcano plots using the EnhancedVolcano R package (version 1.12.0; Blighe et al., 2021).

Additional Table 1.

Basic information regarding the datasets used in this study

GEO accession	Platform	Source	Group	Sample size	Reference
GSE149317	GPL24688	SD rats	ICH,	8	Yuan et al., 2020
		Brain	Contralateral	8
GSE163256	GPL18573	Human	Hematoma effluent,	15	Michael et al., 2021
		Brain/PBMC	Peripheral blood	15
GSE162526	GPL17021	Mouse	PU.1 knock-down	3	Pimenova et al., 2021
		BV2 cells	Control of PU.1knock-down	3
			PU.1 overexpression	3
			Control of PU.1overexpression	3
GSE100889 (GSM2695650, GSM2695651)	GPL16791	Human PBMC	PU.1 ChIP-seq of humandifferentiating macrophages	2	Czimmerer et al., 2018
GSE62826 (GSM1533906)	GPL13112	Mouse Microglia	Microglia PU.1 ChIP-seq of no treatment	1	Gosselin et al., 2014
GSE167593 (GSM5111160)	GPL24247	Mouse Brain	Hemorrhagic Stroke, Control	2	Shiet al., 2021

DEG enrichment and protein-protein interaction network analyses

To identify biological processes potentially underlying ICH pathology and physiology, we performed Gene Ontology (GO) (Ashburner et al., 2000; The Gene Ontology, 2019) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (Ogata et al., 1999; Kanehisa and Goto, 2000) ingenuity pathway enrichment analyses using the clusterProfiler (version 4.0.5) package (Yu et al., 2012; Wu et al., 2021) (with P < 0.05). To identify canonical pathways, we conducted gene set enrichment analysis (GSEA) (Subramanian et al., 2005) using the clusterProfiler package (cut-off Q = 0.05). The Search Tool for the Retrieval of Interacting Genes database (STRING; version 11.5; http://string-db.org/) was used to search for gene interactions (Szklarczyk et al., 2021). The protein-protein interaction (PPI) network was constructed using Cytoscape software (version 3.9.1) based on the output from STRING (Shannon et al., 2003).

Gene set variation analysis and single-sample gene set enrichment analysis

To detect subtle changes in pathway activity in the six datasets, we used gene set variation analysis (GSVA) (Hänzelmann et al., 2013). We downloaded pathway datasets from the Molecular Signatures Database (MSigDB) (version 7.4) (Liberzon et al., 2015) and AmiGO (Carbon et al., 2009). Specifically, the gene expression profiles were compared with the expression profiles of genes related to immunity found in the ImmPort database (https://www.ImmPort.org/) (Bhattacharya et al., 2018). Single-sample gene set enrichment analysis (ssGSEA) was performed on RNA-seq sample matrices with reference to all 50 hallmark gene sets. Graphs were generated using the R packages ggplot2 (version 3.3.5) and ggpubr (version 0.4.0).

Estimation of immune cell fractions

We calculated the immunocyte fractions based on the results from the dataset analysis using three methods. CIBERSORTx was used to deconvolute RNA-seq profiles, which were then used to estimate cell fractions (Newman et al., 2019). xCell was used to calculate the immune cell enrichment score for gene expression profiles (Aran et al., 2017). The Immune Cell Abundance Identifier database (ImmuCellAI) was used to estimate the abundance of 24 immune cell types (Miao et al., 2020). Furthermore, we performed GSVA of the gene expression matrix using cell markers downloaded from CellMarker and PanglaoDB (Franzén et al., 2019; Zhang et al., 2019). We compared the cell fractions obtained from the various methods using the ScaledMatrix package (version 1.4.1) in R (Lun, 2022). The intra-class correlation coefficients (ICCs) were calculated based on the results obtained from these three methods utilizing the irr package in R (version 0.84.1) to evaluate consistency (Gamer et al., 2012). Graphs were generated using the ComplexHeatmap R package (version 2.14.0; Gu et al., 2016).

Weighted gene co-expression network analysis and discovery of key transcription regulators

To explore the relationship between clinical traits and modules, weighted gene co-expression network analysis (WGCNA) of the GEO expression profiles was performed using the WGCNA R package (version 1.70-3) (Langfelder and Horvath, 2008). After calculating the degree of correlation between the modules and clinical traits, one module was selected for further analysis. Epigenetic landscape in silico deletion analysis (LISA) was used to identify transcription factors (TFs) and chromatin regulators (CRs) that potentially influence the significant expression changes observed in the DEGs (Qin et al., 2020). We gathered the protein expression information for these TFs and CRs from the GeneCards database (https://www.genecards.org/) and the Human Protein Atlas (http://www.proteinatlas.org) (Fishilevich et al., 2016; Stelzer et al., 2016; Sjöstedt et al., 2020). The prediction accuracy of TF and CR expression levels in perihematomal and contralateral samples was assessed using the pROC package (version 1.18.0) (Robin et al., 2011) in R. A receiver operating characteristic (ROC) curve was plotted, and the area under the curve (AUC) of key TFs and CRs was calculated.

Assessing PU.1 promoter binding sites

To identify potential targets of Spi1, microglial and macrophagial PU.1 ChIP-seq datasets were accessed using the GEO database. The gene promoters were identified using ChIPseeker (version 1.30.3) (Yu et al., 2015) in R and the mouse mRNA transcript database. After removing duplicates, the ChIP-seq datasets were compared to the DEGs using the GeneOverlap package (version 1.30.0) (Shen and Sinai, 2021), with Fisher’s exact test used to determine the significance of DEGs containing a PU.1 binding site. The VennDiagram package (version 1.7.7) (Chen and Boutros, 2011) was used to identify intersecting genes.

Single-cell RNA-seq analysis

Using cellranger (version 6.1.2) (Zheng et al., 2017) and the Seurat R package (version 4.1.0) (Hao et al., 2021), we reanalyzed the single-cell RNA-seq data (GSE167593). Cells that expressed fewer than 400 or more than 4,000 genes were filtered out. Cells for which more than 30% of gene expression was accounted for by mitochondrial genes were also removed. The remaining cells were clustered and visualized in two dimensions using a Uniform Manifold Approximation and Projection (UMAP) approach (Becht et al., 2018). The microglia were divided into Spi1-Low (Spi1 expression < 1) and Spi1-High (Spi1 expression > 3) groups. The FindMarkers (Trapnell et al., 2014) function was used to identify DEGs between the Spi1-Low and Spi1-High groups. Microglia were labeled using the microglia marker Tmem119.

ICH model

All experimental protocols involving animals followed the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and were approved by the Institutional Animal Care and Use Committee of Zhejiang University on February 13, 2019 (approval No. 2019-157), and are reported in accordance with the ARRIVE 2.0 guidelines (Animal Research: Reporting of In Vivo Experiments) (Percie du Sert et al., 2020). We used male mice only to avoid potential data variability caused by the estrous cycle of female mice. Specific pathogen free (SPF)-grade naive C57BL/6 mice (6–8 weeks, 20–25 g) were purchased from Slac Laboratory Co., Ltd. (Shanghai, China, license No. SCXK (Hu) 2022-0004). Mice were raised four per cage under controlled temperature and humidity conditions with a 12/12-hour light/dark cycle. For this double-blind experiment, we randomized mice into collagenase and saline groups (n = 6 in each group). As previously described (Rynkowski et al., 2008; Mei et al., 2021a), the mice were anesthetized by intraperitoneal injection of sodium pentobarbital (40 mg/kg, Akorn, Lake Forest, IL, USA). For pre- and postoperative analgesia, mice were received buprenorphine (0.05 mg/kg, Buprenex, Reckitt Benckiser, Wallisellen, Switzerland) subcutaneously. A 1-mm-diameter burr hole was drilled in the skull (0.3 mm anterior to bregma and 2.5 mm right lateral to midline), then 0.5 μL of 0.2 U type VII-S collagenase (Sigma-Aldrich, St. Louis, MO, USA) was injected into the right basal ganglia (3.5 mm below the skull) with a microperfusion pump over the course of 5 minutes (Paxinos and Franklin, 2013). To prevent blood backflow, the needle was gradually removed 10 minutes after injection. For the saline group, the mice were injected with 0.5 μL saline in the same manner.

Immunofluorescence staining

Mice were euthanized with 4% pentobarbital (40 mg/kg) at 24 and 72 hours after ICH and transcardially perfused with 0.1 M phosphate-buffered saline (pH 7.4) and 4% paraformaldehyde (PFA). Then, the brains were removed and fixed in PFA for 24 hours at 4°C. After dehydration in 15% and 30% sucrose solutions, the cerebral hemispheres were embedded in OCT and snap-frozen in isopentane precooled with dry ice and liquid nitrogen. Then the samples were sliced into 10-μm-thick sections. After permeabilization with 0.3% Triton X-100 for 15 minutes, cryosections were blocked with 10% donkey serum (Abcam, Cambridge, UK, Cat# ab7475) for 1 hour and then incubated at 4°C overnight with the following antibodies: goat anti-ionized calcium binding adaptor molecule 1 (Iba1; 1:250, Abcam Cat# ab48004, RRID: AB_870576), rabbit anti-Spi1 (1:100, Abcam, Cat# ab227835, RRID: AB_2909597), mouse anti-neuronal nuclear antigen (NeuN; 1:200, Abcam, Cat# ab104224, RRID: AB_10711040), mouse anti-glial fibrillary acidic protein (GFAP; 1:200, Cell Signaling Technology, Boston, MA, USA, Cat# 3670, RRID: AB_561049), mouse anti-CC-1 (1:200, Millipore, Cat# OP80, RRID: AB_2057371, Darmstadt, Germany), mouse anti-TMEM119 (1:200, Proteintech, Philadelphia, PA, USA, Cat# 66948-1-Ig, RRID: AB_2882272). The cryosections were incubated with secondary antibodies for 1 hour and counterstained with 4′,6-diamidino-2-phenylindole at 25°C (Beyotime, Shanghai, China, Cat# P0131). Finally, fluorescence images were captured with a fluorescence microscope (Olympus, Tokyo, Japan). Spi1⁺ Iba1⁺ cells were quantified using the automated image analysis software program ImageJ (version 1.53c, National Institutes of Health, Bethesda, MD, USA; Schneider et al., 2012).

Statistical analysis and visualization

Statistical significance was determined by unpaired two-tailed Student’s t-test (normal distribution) and Wilcoxon test (non-normal distribution). We used the Kolmogorov-Smirnov test to check the normal distribution of datasets. We considered a P-value of less than 0.05 to be statistically significant. All analyses were performed using R software (version 4.1.2, Vienna, Austria) and RStudio.

Results

Microglia and macrophages are involved in neuroinflammation and brain repair after ICH

The scRNA-seq dataset GSE167593 (Shi et al., 2021) was downloaded to explore the major cell types infiltrating the brain parenchyma following ICH in control and hemorrhagic mice. Cells were clustered and projected using Uniform Manifold Approximation and Projection (UMAP), which delineated seven distinct clusters, as well as a cluster representing dividing cells, based on marker gene expression (Figure 1A and B). All seven clusters contained cells derived from both control and hemorrhagic mouse brains (Figure 1C). We observed that hemorrhagic mouse brain accounted for over 45% of the cells in the Microglia_Macrophage cluster, indicating that microglia and macrophages are major cell types that rapidly migrate towards the injury site following ICH.

Figure 1.

Functional annotation of differentially expressed genes, as well as cell type distribution, within ICH brain samples.

(A) Uniform manifold approximation and projection (UMAP) visualization of major cell types present in the striatum in control and hemorrhagic mice. (B) Dotplot of the expression of representative marker genes for each cluster. (C) Barplot of the relative frequencies of cells in each cluster. (D) Volcano plot of differentially expressed genes following ICH. (E) Gene Ontology enrichment analysis of genes that are overexpressed following ICH. (F) Heatmap of hallmark gene set enrichment results for the perihematomal and contralateral areas. (G) Forest plot showing the consistency of brain cell and immunocyte counts among the different methods used. (H) Gene set variation analysis results for macrophage and microglia immune-related pathways. (I) Log2 fold-change of differentially expressed genes related to macrophage and microglia activation and function. *P < 0.05, **P < 0.01 (unpaired two-tailed Student’s t-test). ICH: Intracerebral hemorrhage.

We then downloaded dataset GSE149317 (Yuan et al., 2020), which comprises data from 16 paired ipsilateral and contralateral brain hemisphere samples taken from Sprague-Dawley rats 3 days after ICH, to explore neuroinflammation after ICH. The dataset was stringently screened to remove genes with low expression levels; only genes that were expressed in more than two samples (count > 10) were retained. DEGs between the ICH and control groups were identified through functional enrichment analysis. In total, 961 upregulated and six downregulated genes were identified. Volcano plots for the DEGs were generated using the EnhancedVolcano package (Figure 1D). GO database analysis indicated that the overexpressed genes were predominantly involved in immune responses, such as leukocyte activation, chemotaxis, and phagocytosis (Figure 1E). GSVA of all 50 hallmark gene sets revealed that the inflammatory response, apoptosis, complement, angiogenesis, coagulation, and hypoxia gene sets were upregulated after ICH (Figure 1F).

We further calculated immunocyte fractions using three methods: CIBERSORTx (Additional Figure 1 ^{(576.5KB, tif)} ), ImmuCellAI (Additional Figure 2 ^{(574.9KB, tif)} ), and xCell (Additional Figure 3 ^{(711.6KB, tif)} ). In addition, we performed GSVA of the gene expression matrix using cell markers obtained from CellMarker (Additional Figure 4 ^{(462.6KB, tif)} ) and PanglaoDB (Additional Figure 5 ^{(347KB, tif)} ). We then evaluated the consistency among the results obtained using the three methods by calculating the ICC in R. The results showed good uniformity regarding the proportion of monocytes (ICC value: 0.91, 95% confidence interval (CI): 0.79–0.97, P < 0.001), CD8⁺ T cells (ICC value: 0.86, 95% CI: 0.67–0.95, P < 0.001), and macrophages (ICC value: 0.77, 95% CI: 0.48–0.91, P < 0.001), implying that these three types of cells increased in number after ICH. The GSVA results revealed that microglial numbers (ICC value: 0.99, 95% CI: 0.972–0.997, P < 0.001) were significantly elevated in the ICH group samples, indicating that microglia are present at sites of ICH-induced brain injury, with a resulting decrease in the number of neurons at these sites (ICC value: 0.98, 95% CI: 0.951–0.994, P < 0.001; Figure 1G).

The above results suggest that inflammatory responses are elicited after ICH, with microglia and macrophages being the principal immunocytes mediating immunity. However, it is not clear how these cells types affect ICH-induced brain damage. Hence, GSVA was used to assess microglial/macrophage-related immune processes. As shown in Figure 1H, macrophage activation, chemotaxis, migration, and cytokine production were enriched in perihematomal samples. Likewise, microglia-mediated cytotoxicity, activation, and migration were significantly enriched in perihematomal samples. This suggests that microglia/macrophages are recruited and activated at the damaged site, thereby initiating secondary brain injury processes and subsequent brain repair. Figure 1I presents the expression changes of DEGs involved in immune-related pathways in macrophages and microglia.

WGCNA reveals Spi1 as a key transcriptional regulator during ICH

To further explore the co-expression of mRNAs involved in ICH, WGCNA was applied to the GSE149317 dataset. In total, 15,187 valid genes remained after filtering for low-abundance genes. These valid genes were log2-transformed and imported into WGCNA for co-expression analysis. We selected a soft thresholding power value of β = 9 (scale-free R² = 0.87) to ensure a scale-free network (Additional Figure 6 ^{(546.4KB, tif)} ). Each leaf in the hierarchical clustering tree represents a gene, and each color indicates a module (Figure 2A).

Figure 2.

Key gene modules and hub TRs identified by WGCNA and LISA.

(A) Dendrogram of hypervariable genes clustered by dissimilarity measures. (B) Correlation heatmap of modules and clinical traits. The turquoise ME module was chosen for further analysis. (C) Functional annotation of the genes in the turquoise module demonstrated enrichment in the immune response. (D) Boxplot showing TR RNA expression levels. *P < 0.05, **P < 0.01 (Wilcoxon test). (E) Barplot showing TR protein expression in healthy human tissue. (F) TR protein-protein interaction network. (G) SPI1 expression in the healthy human cerebral cortex and basal ganglia was relatively low, but SPI1 was highly expressed in the bone marrow, spleen, and lymph nodes. (H) The area under the receiver operating characteristic curve of Spi1 versus diagnosis of intracerebral hemorrhage was 0.969. LISA: Epigenetic landscape in silico deletion analysis; ME: module eigengene; TRs: transcriptional regulators; WGCNA: weighted gene co-expression network analysis.

The fractions of infiltrating immunocytes were set as phenotypic traits to estimate the most significant modules associated with ICH and immunocytes. The turquoise module was strongly correlated with the immunocyte fractions, indicating that this gene set may correspond to macrophage and microglial function (Figure 2B). The turquoise module was also associated with ICH, suggesting that macrophages and microglia may contribute to ICH pathophysiology. Functional annotation revealed that the turquoise module is related to immune responses, including leukocyte activation, cell-cell adhesion, and cytokine production (Figure 2C).

To detect the TFs and CRs that potentially regulate the expression of genes in the turquoise module, LISA was performed. Based on P-values, we identified 504 transcriptional regulators (TRs) that could be involved in this process (Additional Table 2 ^{(154.8KB, pdf)} ). Stringent screening strategies were employed to identify key TRs modulating the turquoise module, including kWithin above the median, LISA P-value < 0.01, and TRs encoded by genes within the turquoise module. Applying these strategies resulted in the identification of ten TRs: Spi1, Rela, Fli1, Irf8, Cebpb, Stat6, Ets1, Stat5a, Stat3, and Ikzf1. These ten TRs were highly expressed in the ICH group samples (Figure 2D), implying that they play a pivotal role in ICH pathophysiology.

We further analyzed TR protein expression in the central nervous and immune systems using the GeneCards database. These ten TR proteins were expressed at high levels in immunocytes (Figure 2E). Though Spi1 expression Spi1 was low in the normal brain, it increased significantly after ICH. The PPI network constructed from these TRs (Figure 2F) shows that Spi1 is a hub gene that was expressed at higher levels than the other TRs.

Spi1 expression levels in the brain were further confirmed by analyzing immunohistochemistry images of various human tissues obtained from the Human Protein Atlas (Sjöstedt et al., 2020). SPI1 expression in the cerebral cortex and basal ganglia was relatively low. In contrast, SPI1 was highly expressed in the bone marrow, spleen, and lymph nodes (Figure 2G). In addition, the area under the ROC curve of Spi1 versus ICH diagnosis was 0.969 (Figure 2H).

To further validate the changes in Spi1 expression and determine its subcellular localization, mice were injected with collagenase and sacrificed 72 h later for immunofluorescence staining (n = 6). Consistent with the above results, Spi1 was expressed at low levels in normal brains (Figure 3A and Additional Figures 7 ^{(911.9KB, tif)} –9 ^{(910.2KB, tif)} ). Conversely, the Spi1 signals colocalized with Iba1-positive cells in the perihematomal area and dramatically increased after ICH (Figure 3A and B). These results suggest that Spi1 regulates microglial/macrophage activation and infiltration following ICH.

Figure 3.

Changes in Spi1 expression in the brain after ICH.

(A) Iba1/Spi1 double immunofluorescence staining 72 hours after ICH in the perihematomal area (scale bars: 750 μm in 50× magnification and 100 μm in 200× magnification). Arrows indicate the needle track. (B) Quantification of Spi1⁺ Iba1⁺ cells. **P < 0.01 (unpaired two-tailed Student’s t-test). n = 6 in each group. Experiments were repeated three times. Dapi: 4′,6-diamidino-2-phenylindole; Iba1: ionized calcium binding adaptor molecule 1; ICH: intracerebral hemorrhage.

Spi1 acts as a master regulator of monocyte/macrophage function

To decipher the mechanisms by which monocytes/macrophages induce brain injury after hemorrhage, the GSE163256 dataset (Askenase et al., 2021) was obtained from the GEO database. This dataset contains gene expression profiles of CD14⁺ monocytes/macrophages obtained from the peripheral blood and hematoma effluent of ICH patients. We analyzed data from day 3 of the trial for 15 paired hematoma and blood samples.

Differential gene expression analysis revealed 734 upregulated genes and 1136 downregulated genes (Figure 4A). GO database analysis showed that the overexpressed genes were significantly enriched for cytokine activity, leukocyte migration, wound healing, cell chemotaxis, response to hypoxia, and coagulation (Figure 4B). GSVA of the 50 hallmark gene sets showed that the angiogenesis, inflammatory response, apoptosis, TGF-β signaling, coagulation, and hypoxia gene sets were upregulated (Figure 4C).

Figure 4.

Functional annotation and comparison of DEGs with the Spi1 ChIP-seq dataset in CD14⁺ monocytes/macrophages.

(A) Volcano plot of DEGs. (B) Gene Ontology enrichment analysis of overexpressed genes in monocytes/macrophages obtained from hematoma samples. (C) ssGSEA of all hallmark gene sets in hematoma samples and blood monocytes/macrophages. (D) Gene set variation analysis scores of DEGs involved in immune-related pathways. (E) DEG expression levels are associated with monocyte/macrophage immune function. (F) Overlap between all protein coding genes (grey) and DEGs (orange) that had an Spi1 binding site. (G) Overlap between DEGs (orange) and genes that had Spi1 binding sites within their promoter region (grey). (H) Gene Ontology enrichment analysis of the 263 genes predicted to be regulated directly by Spi1. *P < 0.05, **P < 0.01 (unpaired two-tailed Student’s t-test). ChIP-seq: Chromatin immunoprecipitation sequencing; DEGs: differentially expressed genes; ssGSEA: single-sample gene set enrichment analysis; TSS: transcriptional start site.

Furthermore, we demonstrated activation of immune-related pathways by performing GSVA of the expression profiles of data from the ImmPort database. Cytokines, chemokines, interleukins, and TGF-β family members were significantly enriched in the hematoma effluent (Figure 4D). As Figure 4E shows, the GSVA scores for monocyte/macrophage chemotaxis and differentiation, monocyte aggregation, and macrophage migration in the hematoma samples were all higher than in the peripheral blood.

Next, the ChIP-seq dataset GSE100889 (Czimmerer et al., 2018) was used to identify potential direct targets of Spi1/PU.1. Figure 4F shows that the promoter regions of 59% of the DEGs were enriched in PU.1 binding motifs (P value of overlap = 5.6 × 10^–12), suggesting that Spi1 might regulate the expression of these genes. Moreover, 14% of genes in the ICH dataset exhibited a PU.1 binding site (Figure 4G, P value of overlap = 4.6 × 10^–6). GO database analysis showed that these 263 genes were significantly enriched for cytokine production, myeloid leukocytes, B cells, and macrophage activation (Figure 4H).

Spi1 alters the microglial cell cycle and immune response

Next, we explored how changes in Spi1 expression affect the microglia transcriptome using the GSE162526 dataset obtained from the GEO database. This dataset includes data from FACS-sorted BV2 cells with transient shRNA-mediated Spi1 knockdown (KD) or Spi1 cDNA overexpression (OE).

In the Spi1 KD and scrambled control groups, 320 genes were upregulated and 351 were downregulated (Figure 5A). According to Gene Ontology analysis, the upregulated genes were significantly enriched in chromosome segregation, organelle fission, nuclear division, and spindles (Figure 5B). The downregulated genes were mainly enriched in cytokine production, leukocyte migration, and cell chemotaxis (Figure 5C). GSEA of the KEGG pathway analysis indicated that the significantly enriched functions in the Spi1 KD group included positive regulation of the spliceosome and cell cycle and negative regulation of the chemokine signaling pathway, autophagy, phagosome, and apoptosis (Figure 5D).

Figure 5.

Functional annotation and comparison of DEGs with the Spi1 ChIP-seq dataset in Spi1 KD and OE microglia.

(A) Volcano plot of DEGs in the Spi1 KD and scrambled control groups. (B) Gene Ontology enrichment analysis of upregulated genes in Spi1 KD microglia. (C) Gene Ontology enrichment analysis of downregulated genes in Spi1 KD microglia. (D) Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of genes in Spi1 KD microglia. (E) Protein-protein interaction network of the 145 downregulated genes in Spi1 KD microglia. (F) Volcano plot of DEGs in the Spi1 overexpression and scrambled control groups. (G) Heatmap of canonical pathways affected by Spi1 expression changes in microglia. *P < 0.05, **P < 0.01 (unpaired two-tailed Student’s t-test). ChIP-seq: Chromatin immunoprecipitation sequencing; DEGs: differentially expressed genes; KD: knockdown; OE: overexpression.

Next, a microglial PU.1 ChIP-seq dataset (GSM1533906) (Gosselin et al., 2014) was compared with the DEGs. Additional Figure 10 ^{(410KB, tif)} shows that the promoter regions of 49% of the downregulated genes were enriched in PU.1 binding motifs (P value of overlap = 7.1 × 10^–3, Additional Figure 10 ^{(410KB, tif)} ). Then, a PPI network of the 141 genes downregulated in the KD group was established, and the hub genes were identified by Cytoscape based on the STRING database (Figure 5E). The six hub genes that were identified––Fcgr1, Itgam, Itgb2, Syk, Hck, and Trem3––are known effectors of the microglial pathogen phagocytosis pathway.

In the Spi1 overexpression and scrambled control groups, 41 genes were upregulated and 24 were downregulated (Figure 5F). ssGSEA of hallmark gene sets revealed that a diverse array of cell cycle checkpoint pathways were activated in Spi1 KD microglia, such as the MYC target V1, G2/M checkpoint, and E2F target pathways. The inflammatory response, apoptosis, complement, TNFα-NF-κB signaling, and hypoxia were enriched in the paired control group (Figure 5G).

Spi1 promotes brain recovery via phagocytosis, glycolysis, and autophagy

To investigate the role of Spi1 in microglial function, a single-cell RNA-seq count matrix (GSE167593) (Shi et al., 2021) of mouse hemorrhagic stroke was reanalyzed using the Seurat package. A total of 1303 microglial cells were divided into two groups according to Spi1 expression: Spi1-Low (n = 651 cells) and Spi1-High (n = 652 cells).

As shown in Figure 6A, 535 and 422 genes were upregulated and downregulated, respectively, in the Spi1-High group compared with the Spi1-Low group. ssGSEA of hallmark gene sets showed that TGF-β, mTORC1, and PI3K/AKT/mTOR signaling were activated in Spi1-High microglia (Figure 6B). Spi1-High microglia also displayed changes in the expression of genes related to a broad array of remodeling changes, such as glycolysis, angiogenesis, coagulation, and apoptosis. Additionally, the GSVA scores of phagocytosis-related pathways were significantly higher in the Spi1-High group than in the Spi1-Low group (Figure 6C). This indicates that a high level of Spi1 expression promotes microglial phagocytosis. As shown in Figure 6D, 57% of upregulated genes in the Spi1-High group had binding sites for PU.1 within the promoter region (P value of overlap = 3.5E–23).

Figure 6.

Functional comparison of microglia expressing high versus low levels of Spi1 after ICH.

(A) Dotplot of Log2 fold-changes in gene expression plotted against cell percentage differences (Δ = percentage difference). (B) Heatmap of hallmark gene set enrichment results from microglia expressing high versus low levels of Spi1. (C) Gene set variation analysis results of phagocytosis-related pathways. (D) Proportion of upregulated genes (orange) that contained Spi1 binding sites within their promoter. (E) Spi1 may be involved in brain repair following ICH via phagocytosis, glycolysis, and autophagy. *P < 0.05, **P < 0.01, ****P < 0.0001 (unpaired two-tailed Student’s t-test). Abca1: ATP binding cassette subfamily A member 1; AKT: AKT serine/threonine kinase; AMPK: AMP-activated protein kinase; APOE: apolipoprotein E; CSF1: colony stimulating factor 1; CSF1R: colony stimulating factor 1 receptor; FCER1G: Fc fragment of IgE receptor Ig; FCGR1: Fc fragment of IgG receptor I; HIF1α: hypoxia-inducible factor 1α; ICH: intracerebral hemorrhage; ITGAM: integrin subunit alpha M; ITGB2: integrin subunit beta 2; LAT: linker for activation of T cells; mTOR: mechanistic target of rapamycin kinase; mTORC1: mechanistic target of rapamycin complexes 1; NCKAP1LI: NCK associated protein 1 like; PI3K: phosphatidylinositol 3-kinase; PLCG2: phospholipase C gamma 2; SYK: spleen associated tyrosine kinase; TGFβ: transforming growth factor beta; TGFβR1: transforming growth factor beta receptor 1; TGFβR2: transforming growth factor beta receptor 2; TREM2: triggering receptor expressed on myeloid cells 2; TYROBP: transmembrane immune signaling adaptor; VAV: Vav guanine nucleotide exchange factor.

Taken together, our findings suggest that high Spi1 expression is associated with many pathways that promote phagocytosis, glycolysis, and autophagy, including the TGF-β and PI3K/AKT/mTOR signaling pathways (Figure 6E). By engaging TREM2/TYROBP, FCGR1/FCER1G, CSF1/CSF1R, and TGFβ/TGFβR, Spi1 may regulate phagocytosis through PI3K/AKT/mTOR signaling cascades, thereby participating in a series of processes associated with recovery of cellular function, such as enhanced debris clearance and sustained remyelination.

Discussion

The findings from our study, in which we analyzed datasets of transcriptomes following ICH and investigated a mouse model of ICH, suggest that Spi1 regulates neuroinflammation and neurofunctional recovery after ICH by modulating microglial/macrophage function. We further showed that Spi1 may alter the microglial/macrophage transcriptome and promote phagocytosis by regulating PI3K/AKT/mTOR signaling after ICH. In addition, microglia with a high level of Spi1 expression displayed a broader range of remodeled cellular processes associated with functional recovery, including glycolysis, autophagy, remyelination, angiogenesis, and apoptosis, than microglia with low levels of Spi1 expression.

Previous studies have suggested that ICH triggers activation of microglia/macrophages, which are a source of ongoing neuroinflammation (Zhu et al., 2019; Fu et al., 2021). ICH-induced neuroinflammation can have both detrimental and beneficial effects, depending on the timepoint after injury (Zille et al., 2022). In a variety of circumstances, microglia/macrophages phagocytose debris and remove hematoma, thereby promoting functional neurological recovery (Zhang et al., 2017). Our results demonstrate that microglia/macrophages were the major cell type infiltrating the brain parenchyma after ICH. In addition, the expression of genes related to the immune response, such as those involved in phagocytosis, apoptosis, and complement activation, were enriched in the perihematomal area. This suggests that microglial/macrophage activation and polarization are key cellular events following ICH (Lan et al., 2017; Li et al., 2018).

Monocyte infiltration of the brain parenchyma peaks in the days following ICH, and monocytes outnumber other immunocytes at this timepoint (Hammond et al., 2014; Mracsko and Veltkamp, 2014; Hu et al., 2023). Macrophages derived from these infiltrating monocytes exhibit higher phagocytic capacity than microglia in the ICH-affected brain (Chang et al., 2021). Our findings revealed that the genes that were overexpressed in monocytes/macrophages present within the hematoma were significantly enriched in wound healing, response to hypoxia, and coagulation. Previous studies have suggested that enhanced angiogenesis following ICH may improve neurofunctional recovery in animal models and patients (Lei et al., 2013; Pías-Peleteiro et al., 2017). Here, we observed that the expression of genes involved in TGF-β signaling, angiogenesis, and apoptosis was significantly altered in CD14⁺ monocytes/macrophages from hematoma effluent compared to those from peripheral blood. Taylor et al. (2017) found that TGF-β1 alleviates microglia-mediated neuroinflammation and improves neurofunctional recovery after ICH. Thus, macrophages may promote hematoma clearance and recovery after ICH through TGF-β signaling and angiogenesis (Chang et al., 2018).

WGCNA analysis of the mRNA datasets revealed that a key module was significantly correlated with immunocytes, especially microglia and macrophages, in ICH. LISA identified ten TRs that modulated this module (the turquoise module), including Spi1, Rela, Fli1, Irf8, Cebpb, Stat6, Ets1, Stat5a, Stat3, and Ikzf1. The transcription factor Spi1 is involved in immune cell differentiation and maturation (Turkistany and DeKoter, 2011; Li et al., 2020), and we identified Spi1 as a key transcription factor in the PPI network constructed in this study. In addition, Spi1 expression levels were significantly different between brains affected by ICH and control brains, highlighting the essential role of Spi1 in regulating the immune response after ICH. Three of the other key TRs identified in this study––Rela, Fli1, and Ets1––have been reported to function in angiogenesis and wound healing (Sun et al., 2019; Matrone et al., 2021). In microglia, Irf8, Cebpb, Stat3, and Stat5 contribute to microglial activation, polarization, and modulation of microglia-dependent neuroinflammation (Kierdorf et al., 2013; Pulido-Salgado et al., 2015; Zheng et al., 2021). Furthermore, Stat6 facilitates innate hematoma absorption and improves functional recovery following ICH (Xu et al., 2020). In line with these findings, our results demonstrated that the ten key TRs were highly expressed in perihematomal samples, and their protein expression was enriched in immune cells, implying that they participate in ICH pathophysiology, at steps such as apoptosis, angiogenesis, and immune system regulation, and thus may be potential therapeutic targets for clinical use.

We showed that higher Spi1 expression in microglia is associated with upregulation of Trem2 and Apoe. Previous studies have suggested that Trem2 induces Apoe signaling (Krasemann et al., 2017), which upregulates Spi1 expression (He et al., 2021) and mediates the emergence of neurodegenerative microglia following phagocytosis of apoptotic neurons. Numerous studies have shown that Spi1 is important for microglial/macrophage differentiation and function (Zakrzewska et al., 2010; Jego et al., 2014; Satoh et al., 2014). One study of an Spi1-deficient mouse model identified DEGs that are involved in brain microglia differentiation and maturation (Beers et al., 2006). Spi1 overexpression promotes the generation of microglia-like cells that develop an intact complement system (Cakir et al., 2022). In this study, we explored how modest changes in Spi1/PU.1 expression levels affect the microglial transcriptome. Briefly, Spi1 KD upregulated genes related to the spliceosome and cell cycle pathways and downregulated genes involved in the autophagy, phagosome, and apoptosis pathways. Our results suggest that approximately 49% of the downregulated genes contained PU.1 binding sites, and these genes were significantly enriched within the microglial pathogen phagocytosis pathway, implying that Spi1 upregulation may lead to alterations in the microglia transcriptome and microglia-mediated phagocytosis. Furthermore, the expression of 59% of the DEGs identified in macrophages within the hematoma was likely directly modulated by Spi1. Previous studies have reported that Spi1 target genes are enriched in macrophage functions, including endocytosis and Fc receptor-mediated phagocytosis (Satoh et al., 2014). Based on our results, PU.1/Spi1 is involved in the microglia/macrophage-mediated inflammatory response by modulating the expression of its target genes in microglia/macrophages.

Previous studies have suggested that Spi1 modulates the microglial transcriptome and promotes phagocytosis (Jones et al., 2021; Pimenova et al., 2021). It is well established that mTOR activation is a key factor in triggering phagocytosis and also stimulates autophagy (Thomson et al., 2009; Zhao et al., 2018). Our findings demonstrate that high Spi1 expression is associated with many pathways that promote phagocytosis, glycolysis, and autophagy, such as the PI3K/AKT/mTOR and AMPK signaling pathways. We further found that Spi1 may continuously engage TREM2 and FCGR1 to induce mTOR signaling via upstream activators, such as PI3K and Akt, which are recruited by TYROBP and FCER1G (Weigelt et al., 2007; Ford and McVicar, 2009; Peng et al., 2010; Zhang et al., 2013). This is further evidenced by the finding that the expression of TGF-βR1/TGF-βR2 and CSF1R increased to a similar degree in Spi1-High microglia from ICH mice. TGF-βR and CSF1R stimulate PI3K/AKT/mTOR and AMPK signaling, promoting phagocytic and autophagic activity of microglia (Avila, 2020), which results in enhanced debris clearance and hematoma resolution. In contrast, Li et al. (2017) demonstrated that suppression of CSF1R signaling eliminated microglia and attenuated brain injury following ICH. Higher Spi1 expression in microglia led to increased cellular stress responses, such as oxidative phosphorylation, via the PI3K/AKT/mTOR signaling cascades. Oxidative phosphorylation plays a role in sustained remyelination and recovery of cognitive function post-stroke (Song et al., 2022). Additionally, Spi1-High microglia displayed changes in the expression of genes associated with a broad array of remodeling processes, including angiogenesis, coagulation, and apoptosis.

This study had several limitations. First, we did not determine the influence of Spi1 knockdown and overexpression on microglia/macrophage function after ICH. Thus, further research is required to investigate the relationship between modest changes in Spi1 expression and ICH-induced neuroinflammation. Secondly, our results were determined by relatively simple experiments mostly based on transcriptome data, which do not fully reflect immune responses following ICH. Consequently, future studies are needed to confirm the reliability of our findings and to investigate Spi1-dependent processes after ICH. Thirdly, the RNA-seq and ChIP-seq datasets are inherently limited by the methods used to obtain RNA and identify target genes specific to human ICH pathology. Therefore, there may have been some selection bias in the present study. Fourthly, there are some limitations to the collagenase mouse models; for example, slight accumulation of microglia along the needle track was observed in the brains of mice from the saline group (which we surmised to be owing to mechanical trauma caused by the needle). Finally, the dataset used in this study was small. Hence, the reliability and validity of our findings must be verified in further longitudinal studies.

Overall, our study showed that Spi1 is a key immune-related intervention target in ICH, that Spi1 alters the microglial/macrophage transcriptome and promotes phagocytosis after ICH, and that PI3K/AKT/mTOR signaling might participate in this process. These findings should be confirmed in future studies that investigate the consequences of Spi1 knockout and overexpression on neuroinflammation and other relevant outcomes in ICH.

Additional files:

Additional Figure 1 ^{(576.5KB, tif)} : CIBERSORTx evaluation of immunocyte abundance.

Additional Figure 1

CIBERSORTx evaluation of immunocyte abundance.

Additional Figure 2 ^{(574.9KB, tif)} : ImmuCellAI evaluation of immunocyte abundance.

Additional Figure 2

ImmuCellAI evaluation of immunocyte abundance.

Additional Figure 3 ^{(711.6KB, tif)} : xCell evaluation of immunocyte abundance.

Additional Figure 3

xCell evaluation of immunocyte abundance.

Additional Figure 4 ^{(462.6KB, tif)} : CellMarker evaluation of brain cell and immunocyte abundance.

Additional Figure 4

CellMarker evaluation of brain cell and immunocyte abundance.

Additional Figure 5 ^{(347KB, tif)} : PanglaoDB evaluation of brain cell and immunocyte abundance.

Additional Figure 5

PanglaoDB evaluation of brain cell and immunocyte abundance.

Additional Figure 6 ^{(546.4KB, tif)} : Network topology analysis of adjacency matrices with different soft threshold powers.

Additional Figure 6

Network topology analysis of adjacency matrices with different soft threshold powers.

Additional Figure 7 ^{(911.9KB, tif)} : Representative images of co-labeling for Iba1, Spi1, and NeuN at 72 hours following ICH in the ipsilateral basal ganglia.

Additional Figure 7

Representative images of co-labeling for Iba1, Spi1, and NeuN at 72 hours following ICH in the ipsilateral basal ganglia.

Scale bars: 250 μm. Iba1: Ionized calcium binding adaptor molecule 1; ICH: intracerebral hemorrhage; NeuN: neuronal nuclear antigen.

Additional Figure 8 ^{(913.5KB, tif)} : Representative images of co-labeling for Iba1, Spi1, and GFAP at 72 hours following ICH in the ipsilateral basal ganglia.

Additional Figure 8

Representative images of co-labeling for Iba1, Spi1, and GFAP at 72 hours following ICH in the ipsilateral basal ganglia.

Scale bars: 250 μm. GFAP: Glial fibrillary acidic protein; Iba1: ionized calcium binding adaptor molecule 1; ICH: intracerebral hemorrhage.

Additional Figure 9 ^{(910.2KB, tif)} : Representative images of co-labeling for Iba1, Spi1, and CC-1 at 72 hours following ICH in the ipsilateral basal ganglia.

Additional Figure 9

Representative images of co-labeling for Iba1, Spi1, and CC-1 at 72 hours following ICH in the ipsilateral basal ganglia.

Scale bars: 250 μm. Iba1: Ionized calcium binding adaptor molecule 1; ICH: intracerebral hemorrhage.

Additional Figure 10 ^{(410KB, tif)} : The proportion of downregulated genes containing an Spi1 binding site within their promoter region.

Additional Figure 10

The proportion of downregulated genes containing an Spi1 binding site within their promoter region.

Additional Table 1: Basic information regarding the datasets used in this study.

Additional Table 2 ^{(154.8KB, pdf)} : Transcriptional regulators identified by landscape in silico deletion analysis.

Additional Table 2

Transcriptional regulators identified by landscape in silico deletion analysis