The data is available from National Center for Biotechnology Information and Gene Expression Omnibus, where allowed researchers to download and reuse public datasets for scientific purposes without ethics approval. The data was filtered based on sex, age, type of stroke, and the sequencing platform used. To analyze gender differences in humans, we selected the transcriptomic profiles from peripheral monocytes, neutrophils, and whole blood from 38 IS patients and 18 control individuals from the dataset published by Carmona-Mora (Carmona-Mora et al., 2023). To analyze differences in mice, we used the GSE137482 dataset, that included 3- and 18-months old C57BL/6 mice subjected to permanent middle cerebral artery occlusion to model cerebral ischemia (N = 6/group) (Androvic et al., 2020). In addition, the GSE174574 single-cell RNA-seq data published by Zheng was reanalyzed (N = 3/group) (Zheng et al., 2022). Gene expression data from the GSE137482 and GSE174574 datasets were downloaded for validation purposes. The data was imported into R (version 4.0.2) for downstream analysis. RStudio (2022.07.1) was used for statistical analyses and to create image plots. Sample information and the allocation of figures is summarized in Supplementary Table S1.

The analysis of gene expression was performed in R and RStudio. DEGs analysis was carried out using the DESeq2 algorithm (v 1.32.0) (Love et al., 2014; Zhu et al., 2018). For the enrichment analysis of DEGs transcripts, we utilized the ClusterProfile package (v 3.18.1) (Wu et al., 2021; Yu et al., 2012) accepting a p-adjusted value <0.05 as the cutoff for biologically relevant enrichment results. All analyses were conducted in RStudio, with the results being plotted using ggplot2.

Data were presented as mean ± S.D. Difference was assessed by multiple t-test for two individual groups, and by two-way ANOVA with Tukey post hoc tests among multiple groups (GraphPad Prism Software Inc., San Diego, CA). In order to control the confounding variables, a multiple regression was conducted in analyzing human data, in which the dependent variable was gene expression variable, while independent variables were the risk factors affecting stroke incidence and outcome, including gender, age, diabetes, hypercholesterolemia and hypertension. R (version 4.0.2) was used for statistical analyses and for generating image plots. Statistical significance was defined as p < 0.05.

To track the long-term impact of sex on trends of stroke incidence worldwide, we analyzed new case statistics, mortality, and hospitalization rates of stroke according to the patients’ sex and type of stroke (Data source: https://nccd.cdc.gov/DHDSPAtlas/?). Between 2005 and 2020, mortality rates from all strokes, including cerebral ischemia and hemorrhage, declined from 500 to 550 to 400–450 per 100,000 population. Invariably, males have a higher mortality rate compared to females regardless of age, race, and ethnicity (Figure 1A). Previous studies largely agreed that IS was 10 times more common than HS in Western countries and carried a lower mortality risk (Namale et al., 2020; Andersen et al., 2009), although contradictory data was also reported (Abdu et al., 2021). We analyzed the mortality between 2005 and 2020 in the US (Figure 1B) and found that the likelihood of death was two-fold higher after IS during this period and this difference was not influenced by the sex of the patients. However, there was a trend toward a continuously declining mortality for all forms of stroke, reflecting the improved post-stroke management and treatment strategies.

Hospitalization rate is generally accepted to reflect the early detection and prompt treatment of stroke. However, it also provides a picture of stroke incidence. As shown in Figure 1C, the hospitalization rate for IS declined slightly between 2005 and 2019. Conversely, the hospitalization rate of HS showed an upward trend in both sexes during the same period. The underlying reasons are complex, including age-related hypertension, cerebral small vessel disease, changes in lifestyle, improved diagnostic techniques, and enhanced public awareness. In addition, females had a higher incidence of strokes, especially IS (Figure 1D). Furthermore, differences between males and females in stroke prevalence, disability-adjusted life years, and mortality show similar trends worldwide (Feigin et al., 2022), with higher rates of IS in females, and a higher incidence of HS in males (Figures 1E–G).

To explore if there was a difference in the expression of XCI escape genes between the sexes, we reanalyzed the transcriptomic profiles of human peripheral whole blood, monocytes and neutrophils in the RNA-seq dataset published by Carmona-Mora (Carmona-Mora et al., 2023). We focused on 19 genes that have been implicated in the onset and maintenance of XCI, including KDM5C, KDM6A, EIF2S3, ATRX, CNKSR2, DDX3X, FMR1, HDAC8, MECP2, MID1, MORF4L2, MSL3, MAGEC1, OGT, PHF6, SMC1A, SYP, USP9X, and TMEM47 (Supplementary Figures S1, S2). The datasets were already filtered and normalized by the original authors (TPM normalized, non-log, filtered features with maximum ≤40 reads were excluded). We herein explored whether stroke event could affect the sexual dimorphism expression of these 19 X chromosome-escaping genes. We found that, compared to males, females performed significantly higher expression of KDM5C and KDM6A in whole blood of healthy cohort (Figures 2A, B). This difference seemed further exacerbated by ischemic stroke, which was evidenced by that more significant differences were observed between sexes. Interestingly, EIF2S3 level was comparable between sexes in the whole blood of healthy cohort, but become significantly higher in females after stroke (Figure 2C). Importantly, a multiple regression analysis indicated that other risk factors of stroke, including age, diabetes, hypercholesterolemia and hypertension, were not associated with the sex difference of EIF2S3 level (Supplementary Table S2). It is known that circulating monocytes and neutrophils infiltrate the injured brain and mediate the inflammatory process that develops after stroke (Carmona-Mora et al., 2021; Shichita et al., 2012; Planas, 2018a). However, the expression levels of KDM5C, KDM6A and EIF2S3 were comparable between sexes with or without ischemic stress (Figures 2D–I).

It was previously reported that gene expression profiles in specific cell types change with increasing age (French et al., 2017; Chow et al., 2012; Brinkmeyer-Langford et al., 2016). Therefore, we explored whether KDM5C, KDM6A, or EIF2S3 displayed age-dependent expression changes in males and females and whether such age-dependent expression changes could be altered by ischemic stroke (Figures 2J–R). We found that the expression of KDM5C and KDM6A showed relatively smooth fluctuations in the whole blood, monocytes, and neutrophils. However, EIF2S3 expression showed a trend of increase in the monocytes of the 91–100-year-old cohort (Figure 2O). In neutrophils, IS caused an increased trend of EIF2S3 at most age groups, irrespective of the sex of the patient (Figure 2R). The level of EIF2S3 mRNA in female neutrophil is very close to 0 throughout ages, although it is clearly detectable in female brain and liver (Xu et al., 2006), suggesting that the tissue- and cell type-preferential expression of EIF2S3 might be sexual dimorphic. It is worthy to note that this observation is lack of statistical support as the sample size is insufficient in separated age bracket. More data is required in the future.

To explore X chromosome-encoded DEGs according to sex and cerebral ischemia, we reanalyzed the dataset published by Androvic (GSE137482) (Androvic et al., 2020). This was derived by a bulk RNA-seq analysis of four groups, including young (3-month-old) and old (18-month-old) mice suffered sham operation and ischemia. We respectively investigated the aging effect to genes alternation under either sham (Figure 3A) or stroke (Figure 3B) condition, and that the effect of stroke insult to genes alternation under either aged (Figure 3C) or young (Figure 3D) cohort. We found that accepting a fold change of >2 with a p < 0.05 statistical cut off, aging led to 223 DEGs upregulated and 140 downregulated in sham animals (Figure 3A), and that aging led to 389 DEGs upregulated and 94 downregulated in stroke animals (Figure 3B). In the aged population, stroke insult led to 5,516 DEGs upregulated and 105 downregulated (Figure 3C), whereas 4,452 DEGs upregulated and 91 downregulated in young population (Figure 3D). In addition, we explored the locations of the most prominently up- or downregulated DEGs in individual chromosome. As shown in Figures 3E, F, changes associated with aging were distributed relatively evenly along autosomal chromosomes, with hardly any DEGs localizing to the X chromosome. In contrast, stroke insult led to a marked increase of DEGs on the X chromosome regardless age (Figures 3G, H).

Next, we explored DEGs localizing to the X chromosome in more detail (Figures 3I–L; Table 1). When we compared sham operated old and young animals, none of the DEGs were encoded on the X chromosome (Figure 3I), while only one gene, Nrk, significantly increased in aged stroke animals compared to young stroke group (Figures 3J, M). These findings are in line with the analysis presented above (Figures 3E, F). It is worth noting that the amplitude of expression changes was higher for genes located on the X chromosome, with −2.5 to +7.5 folds changes in both young and ages. In contrast, age alone caused −3 to +3 folds change regardless stroke insult (Figures 3K, L). Importantly, in the aged population, stroke induced 37 X chromosome-encoded DEGs upregulated, whereas stroke only induced 31 DEGs upregulated in the young population (Table 1), in which 26 DEGs overlapped with that in the aged cohort indicated by Venn diagram (Figures 3N, O). These data collectively suggest that stroke is a more robust factor affecting gene regulation than age, and this is particularly true for DEGs located on the X chromosome of mice.

Given that age and sex are major non-modifiable risk factors for stroke (Egorova et al., 2019; Curb et al., 1996), we compared the sex differences of genome-wide DEGs using the brains of 36-monthes male and female mice exposed to experimental hypoxia. In this analysis, the GSE137482 RNA-seq dataset was carried out, and that the significant difference was considered by a fold change of >6 with p < 0.05 statistical cut off (Figure 4A). We found that 392 transcripts were sexual dimorphic on X chromosome identified by volcano plot, in which 185 were upregulated and 207 were downregulated (fold change>2, p < 0.05, Figure 4B). To conduct a more accurate view of the chromosomal localization of the affected genes, we calculated the average gene expression levels within 1 kb windows tiled across the X chromosome. This analysis showed that the localization of DEGs on X chromosome was fairly evenly distributed (Figure 4C). Next, we analyzed up- and downregulated genes in male and female ischemic brains using a heatmap showing DEGs with more than 6-fold change. The upregulated X chromosome genes in males included Fam120c, Abcd1, Cd99l2, Taf9b, Nap1l3, Gm8822, Gm7803, and Gm26617; those in females were Chm, Enox2, Fundc2, Msl3, Ddx26b, Nkrf, Snx12, Armcx6, Pin4, Xist, and Gm26992 (Figure 4D). The individual variations of these DEGs in male and female brains were analyzed and visually presented in Figures 4E, F.

It has been documented that, after a cerebral stroke, some immune cells start to infiltrate the brain and, together with brain resident immune cells, primarily the microglia, contribute to the development of brain injury (Wang H. et al., 2023; Al et al., 2019). Microglia are generally considered as “first responders” after hypoxia showing divergent features between the two sexes (Ngwa et al., 2021; Ngwa et al., 2022; Ugidos et al., 2022; Han et al., 2021a). Therefore, we analyzed the transcriptomic datasets of microglial and peripheral infiltrating T cells, to explore the expression of X chromosome-encoded DEGs in 36-months-old mouse brains after cerebral ischemia.

First, we performed a volcano plot analysis to examine differences between male and female microglial mRNA abundance. With a p < 0.05 and 2-fold change cutoff, we identified 168 DEGs. Of these 120 were upregulated and 48 were downregulated in female (Figure 5A). When looking at the chromosomal distribution of the DEGs, only a limited number of them were significantly altered with more than 2-folds change from the X chromosome (Figure 5B). Specifically, 6 genes, including Mbnl3, Tlr8, Dmd, Fancb, Rbm3os and Gm14567, were upregulated and 4 genes, including Gm9670, Xist, Tmsb15b2 and Gm26992, were downregulated in males compared to females (Figures 5C–E).

Signaling via activated microglia initiates the migration of peripheral blood T lymphocytes into the injured area as early as 24 h after the initial ischemic insult (Jander et al., 1995; Brait et al., 2012). This process has been documented to exacerbate the neuroinflammatory process and contributes to the ensuing brain damage (Zhang et al., 2021; Benakis et al., 2022). However, little is known about the sex dichotomy of T cell features after stroke. In the T cells from the ischemic brain of 36-month-old mice, we detected a total of 429 DEGs, in which 198 DEGs was increased and 231 was decreased in female compared to male cells (Figure 5F). Interestingly, the amplitude of change of DEGs in T cells was more pronounced than that seen in microglia. As shown in Figure 5G, some DEGs showed as high as 10-fold increase or decrease between sexes. Apart from the robust changes in the gene expression presenting on autosomes and the X chromosome, T cells also exhibited dramatic changes in the Y chromosome-encoded genes (Figure 5G). Among the 429 DEGs, 20 were located on X chromosome (Figure 5H). Of these, 9 were upregulated in males (Figure 5I), and 11 genes were upregulated in females (Figure 5J). These observations indicate that X chromosome-encoded DEGs performed more remarkable sexual dimorphism on activated T cells than on activated brain resident microglial cells.

To analyze the effect of age on hypoxia-induced changes of X chromosome-encoded DEGs on the microglial and T cells, we conducted additional comparisons to exclude the confounding effect of the differences between sexes, only data from male brains was utilized (Rebekah et al., 2022). The volcano plot analysis indicated that 61 DEGs were upregulated and 76 were downregulated in brain microglia of the 36-month-old male mice compared to the young group (Figure 6A). A similar analysis detected 229 DEGs upregulated and 290 downregulated in T cells (Figure 6B). Limiting the analysis to genes encoded on the X chromosome identified 7 upregulated DEGs in the microglia (Figure 6C), 10 upregulated and 14 downregulated DEGs in the T cells of the older animals (Figure 6D). These findings are in line with our previous observations indicating that hypoxia induces more pronounced changes of gene expression in T cells. The individual variations of these DEGs in aged microglia and T cells were visually presented in Figures 6E–G.

A major challenge in understanding changes in hypoxia-induced gene expression patterns is the heterogeneous cellular composition of the brain. Single-cell RNA-seq is a powerful approach to gain more granular information in this context. To explore the effects of cerebral ischemia on X chromosome-encoded DEGs in individual brain cell populations, we reanalyzed the open access single-cell RNA-seq dataset (GSE174574) published by Zheng (Zheng et al., 2022). Using cell type specific gene markers, authors identified 17 principal clusters in the mouse brain and explored their transcriptional changes 24 h after permanent ischemia. We filtered and reanalyzed the expression of genes located on the X chromosome in the whole brain, astrocytes, microglia (MG), oligodendrocytes (OLG), ependymocytes (EPC), perivascular fibroblast-like cells (Fanning et al., 2019), lymphocytes (LYM), neutrophils (NEUT), pericytes (PC), venous endothelial cells (vEC), vascular smooth muscle cells (SMC), central nervous system (CNS) border-associated macrophages (CAM), choroid plexus epithelial cells (CPC), and dendritic cells (DC) (Figure 7; Supplementary Figure S2), which were also summarized (Table 2) and visualized (Supplementary Figure S3). It was notable that similar trends of gene expression were apparent across several cell types. For instance, the abundance of the transcript of Itm2a was significantly decreased in astrocytes, MG, and LYM (Figures 7B, C; Supplementary Figure S2C), but increased in OLG, EPC, PC, vEC, SMC, and CAM (Figure 7D; Supplementary Figures S2A, E, F, G., H). To summarize hypoxia-induced changes in the expression of X chromosome-encoded genes in the 14 distinct cell populations, we used an UpSet plot to indicate the detected trends (Figure 7E). These data reinforce the complexity of the expression of X chromosome-encoded genes in brain hypoxia and illustrates the challenges of understanding the differences in stroke responses between the two sexes.

Next, the biologic function of the hypoxia-induced DEGs of X chromosome (refer to Figure 3) was explored by performing gene classification analyses (Dataset GSE174574). All X chromosome DEGs were uploaded into the g:Profiler (https://biit.cs.ut.ee/gprofiler/gost) functional enrichment analysis that provides known functional information sources and detects significantly enriched terms (Raudvere et al., 2019; Reimand et al., 2007). The biologic process terms were summarized in the Figure 8A. The X-axis showed the functional terms grouped by colors, in which functional enrichments were further branched. The number indicates the intersection size of query (Available in Supplementary Material).

The most relevant enriched GO terms are reported in Figure 8. Among the 26 DEGs selected (refer to Figure 3), 4 were associated with GO molecular function, cytokine binding, including IL13ra1, IL2rg, Bgn and Plp2. Notably, 18 genes associated with GO biological processes, such as response to organic substance, cellular response to organic substance, tumor necrosis factor production, regulation of tumor necrosis factor production, regulation of tumor necrosis factor superfamily cytokine production, tumor necrosis factor superfamily cytokine production, and response to cytokine. Notably, in addition to their ontology, these genes are involved in multiple terms including Reactome (REAC), Human Protein Atlas (HPA), and Human Phenotype Ontology (HP). In this enrichment analysis, stroke-induced upregulation of Toll-like receptor (TLR) 7 was associated with mucosal lymphoid cells (HPA), X-linked inheritance (HP), abnormal lymphocyte count and morphology (HP) regardless age (Available in supplementary file), which is consistent with previous report about the TLR related poor outcome and greater inflammatory response after acute ischemic stroke (Brea et al., 2011), suggesting that antagonizing TLR7 might be promising in the treatment of ischemic stroke and brain injury (Gesuete et al., 2014; Lu et al., 2020).