It is well known that some X chromosome genes escape XCI [14, 15, 41, 42], leading to gene dosage imbalance between males and females [43, 44], which could impact post-stroke inflammation and outcomes once a stroke occurs. The present study focused on two X chromosome escapee genes, Kdm6a and Kdm5c, and investigated their epigenetic modulation of IRF5/IRF4 via demethylation of H3K27Me3/H3K4Me3 in aged microglia after stroke. IRF5-IRF4 regulatory axis has been previously found to be the determinant pathway that regulates microglial pro-/anti-inflammatory responses [21, 22, 28, 45, 46], and is critical in mediating stroke injury. The current data showed that Kdm6a and Kdm5c signaling both impact on one end of the axis, i.e. IRF5, but in an opposite pattern. Two alleles of Kdm6a led to active transcription of IRF5; whereas two alleles of Kdm5c caused suppressive transcription of the pro-inflammatory factor. Two alleles of either Kdm6a or Kdm5c in microglia induced exacerbated pro-inflammatory responses after stroke, which led to worsened stroke injury. The different effect of the two Kdms on IRF5 transcripti on suggests that the two X escapee genes impact on stroke outcomes in the aged via different pathways.
Stroke is a sexually dimorphic disease [47, 48]; ischemic stroke sensitivity is mediated primarily by gonadal hormones in young population [49, 50] and by sex chromosomal complement in the aged [23, 51]. The contribution of the second X chromosome to stroke sensitivity in the aged has been observed in our previous study, with the two XCI escapee genes (Kdm6a/Kdm5c) involved [24]. The double expression of the two alleles of Kdm6a/Kdm5c due to the escape has been found also implicated in sex differences in cardiac infarction and adiposity [52, 53]. The current study utilized three animal models with different allele numbers of active Kdm6a or Kdm5c, and demonstrated the detrimental effects of both X escapee genes on stroke injury. Since the Kdm CKO female mice only has one allele of Kdm6a or Kdm5c and without Y chromosome, the comparison between CKO and Kdmfl/fl females is exclusively reflective of the effect of X chromosome dosage but none of Y effect. Therefore, the current data convincingly indicate that the escape of Kdm6a or Kdm5c plays a detrimental role in post-stroke inflammation and stroke injury, and support the rational that the Y chromosome has limited effect on the stroke sensitivity [23].
The current study focused on the effect of Kdm6a/5c escape from XCI in aged microglia on stroke, as microglia play important roles in initiating and perpetuating post-stroke neuroinflammation. The inducible CKO model utilized in the study makes it feasible to investigate gene escape in microglia specifically. Although CX3CR1-CreER system targets both microglia and infiltrating monocytes in the ischemic brain, we did not perform experiments until 6 weeks after TMX induction so that the microglia can be the sole target. Infiltrating monocytes have gone through ‘turnover”[54, 55] and no longer bear the TMX induced gene knockout after 6 weeks of TMX induction; whereas microglia still have the KO due to their longevity [56]. Gene escape from XCI is random, and has tissue and cell variability [57, 58]. In addition, gene escape from XCI may be affected by various biological homeostasis changes including aging and stroke injury. XCI becomes unstable with age, which is a frequently proposed explanation for the phenotype spectrum of disease in females [42, 59, 60], suggesting some X-linked genes escape more easily with aging. Our previous study has found Kdm6a/5c were significantly higher expressed in sorted aged female vs. male microglia from naïve mice, and the sex difference was lost when evaluated in whole brain tissue in sham mice but present in brain tissue homogenates after stroke [24]. These data suggest that Kdm6a/5c escape from XCI has cell variability, and is sensitive to stroke stimulus.
Epigenetic regulation of genes has been widely studied including DNA methylation [61], histone [62] and non-coding RNAs [63] modifications, with growing interest in exploring the related regulatory mechanisms underlying neuroinflammation in stroke [11, 64, 65]. Techniques such as chromatin immunoprecipitation (ChIP) [62, 66], and CUT&RUN [67–69] have accelerated the advance of epigenetic studies, by elucidating gene-protein interactions and the downstream targets. Epigenetics involves histones which serve as “gatekeepers” to modulate DNA replication/transcription and gene expression [70]. Kdm6a and Kdm5c are demethylases for H3K27Me3 [71] and H3K4Me3 [72], and the demethylation of the two histones induces active and a transcriptive effect on gene transcription [73–76], respectively. Recently we have demonstrated by ChIP that the inflammatory transcription factors, IRF 5/4, bind to H3K27Me3 or H3K4Me3, suggesting the two IRFs are subjected to the epigenetic modulation of the histones [24]. Histones contain five components: H1, H2A, H2B, H3, and H4 [12], and undergo post-translational modifications of the N-terminal tail by acetylation, methylation, phosphorylation, ubiquitination, demethylation, and lactylation [77–80]. The modifications of histone tails affect the interaction of histones and DNA, and alter the structure and stability of chromatin [81], and regulate the gene transcription through modulating the affinity of transcription factors and structural gene promoters [14]. X chromosome-linked genes have been shown to play important roles in epigenetic modification of genes related to post-stroke inflammation [23, 82]. Our data show that kdm6a/5c both regulate IRF5 transcription as in (Fig. 2&3), however in an opposite pattern (active vs. suppressive) through different histone demethylation, reflecting the complex nature of histone chromatin accessibility to transcriptional elements of the IRF5 gene after stroke. Epigenetic mechanisms after stroke are critical in the molecular pathophysiology of the disease, and are potential therapeutic targets [64, 83] to salvage the hypoperfused ischemic penumbra that has not yet evolved into infarcted tissue [84]. The present study provided potential epigenetic avenues to target XCI escapee genes to regulate the expression of the pro-inflammatory transcription factor IRF5.
IRF5 is a well-established pro-inflammatory transcription factor responsible for mediating microglial production of inflammatory cytokines [21]. Of note, our data demonstrated that Kdm6a and Kdm5c signaling have opposite effects on IRF5 transcription (Figs. 2&3). However, the escape of both Kdms from XCI has pro-inflammatory effects including promoting microglial pro-inflammatory response (Fig. 4), increasing plasma/brain levels of pro-inflammatory cytokines (Fig. 5&6), and both led to exacerbated stroke injury (Figs. 7&8). The active effect of Kdm6a on IRF5 transcription is logic to the downstream pro-inflammatory response and worsened stroke injury, but the suppressive effect of Kdm5c on IRF5 seems irrelevant to the downstream outcomes. Different Kdm family proteins finetune the switch of gene expression by manipulating active or repressive histone methylation markers, thus participating in various links of immune cells and inflammatory activities [85, 86]. Kdm6a is a demethylase for H3K27Me3 [87], whereas Kdm5c is responsible for demethylation of H3K4Me3 [88]. Our data are consistent with this as H3K4-IRF5 axis was not affected by Kdm6a (Fig. 2D-F) and K3K27-IRF5 not changed by Kdm5c (Suppl. 2 A-C). The specific histone target for the two Kdms might be the reason why they have different effect on IRF5 transcription. It is likely that Kdm5c suppresses transcription of some anti-inflammatory genes to confer detrimental effects on neuroinflammation.
The current study has some caveats that we should keep in mind when interpreting the data. We examined the Kdm-histone-IRF axis in aged microglia only at the acute phase of stroke (3 days after MCAO), and did not include a chronic stage cohort study which is still on-going (years of work). However, our acute study has already elucidated the mechanistic link between Kdm6a/5c and post-stroke inflammation, which will be further confirmed in the following experiments. Another caveat of the study is that we did not examine Kdm-IRF5 signaling in infiltrating monocytes. It has been reported [89] that demethylation of H3K27Me3 by Kdm6a markedly increased IL-1β expression through a Caspase-1 pathway in macrophages. The infiltrating monocytes in the ischemic brain also express IRF5 [45, 90, 91]. Nevertheless, our previous study has already suggested that the central (microglia) IRF signaling is more important than the IRFs expressed on peripheral immune cells in post-stroke inflammation [92].
In summary, the present study investigated the demethylating effects of Kdm6a/5c on H3K27Me3/H3K4Me3-IRF5/4 signaling in microglia, and assessed their impact on stroke outcomes in aged mice. Our findings reveal that the escape of microglial Kdm6a/5c from XCI exacerbates post-stroke inflammation and worsens outcomes. IRF5 signaling plays a critical role in mediating the deleterious effect of Kdm6a (Fig. 9); whereas Kdm5c’s effect is independent of IRF5. The epigenetic modification of histones by X escapee genes is a novel mechanism in inducing sex differences in stroke among the elderly, highlighting new, sex-specific therapeutic targets for this devastating disease.