Epigenetic control of microglial mitochondrial immunity by KAT7 drives Alzheimer’s disease pathogenesis

SUMMARY

Mitochondrial DNA (mtDNA)-driven innate immune signaling sustains chronic neuroinflammation in neurological diseases such as Alzheimer’s disease (AD), yet how this pathway is regulated in microglia remains poorly understood. Here, we identify the histone acetyltransferase KAT7 (HBO1) as a central epigenetic regulator that links chromatin remodeling to mitochondrial immune activation. KAT7 and its histone mark H3K14ac are elevated in microglia from 5×FAD mice and human AD brains. Integrative transcriptomic and epigenomic analyses reveal that KAT7 activates transcription of Cmpk2, a mitochondrial kinase essential for mtDNA synthesis. Loss of KAT7 reduces Cmpk2 expression, impairs mtDNA replication and release, and consequently suppresses cGAS-STING and NLRP3 signaling. Importantly, both microglia-specific deletion and pharmacological inhibition of KAT7 mitigate cytosolic mtDNA-induced neuroinflammation, decrease amyloid-β burden, restore synaptic plasticity, and improve cognitive function in 5×FAD mice. Together, these findings uncover an epigenetic-mitochondrial axis sustaining microglial pathogenicity and establish KAT7 as a potential therapeutic target for AD.

Graphical Abstract

eTOC Blurb:

Microglia-mediated neuroinflammation drives Alzheimer’s disease progression. In this article, Liu et al. show that the histone acetyltransferase KAT7 promotes mitochondrial DNA-dependent inflammation in microglia. Inhibiting KAT7 reduces amyloid-β accumulation, neuroinflammation, and cognitive deficits in Alzheimer’s mouse models, revealing a potential epigenetic target to counteract disease-associated inflammation.

INTRODUCTION

Alzheimer’s disease (AD) is the most common cause of dementia, affecting tens of millions worldwide and imposing an ever-growing societal and economic burden¹. It is neuropathologically characterized by the accumulation of extracellular amyloid-β (Aβ) plaques and intracellular neurofibrillary tangles of hyperphosphorylated tau². Yet, decades of research targeting these hallmark proteins have yielded only modest clinical benefit, indicating that additional mechanisms underlie AD pathogenesis³. Accumulating evidence implicates chronic neuroinflammation as a central disease mechanism^4–6. Genome-wide association studies have linked many AD risk loci to microglial pathways, highlighting microglia, the brain-resident immune cells, as key players in AD^7–9. While microglia initially protect the brain by clearing misfolded proteins, chronic activation drives their transition into a proinflammatory state marked by excessive cytokine production and loss of homeostatic functions¹⁰. This maladaptive state fuels a self-perpetuating cycle of inflammation and neurodegeneration.

Mitochondrial dysfunction is increasingly recognized as a critical driver of chronic microglial inflammation^11–13. In both aging and AD brains, microglia exhibit elevated levels of mitochondrial DNA (mtDNA) in the cytosol, where it acts as a potent damage-associated molecular pattern^14–16. Cytosolic mtDNA is sensed by cyclic GMP-AMP synthase (cGAS), which produces the second messenger cyclic GMP-AMP (cGAMP) to activate STING (stimulator of interferon genes)¹². Activation of this pathway promotes phosphorylation of TBK1 (TANK-binding kinase 1) and IRF3 (interferon regulatory factor 3), leading to type I interferon induction and the release of proinflammatory cytokines¹⁷. Persistent cGAS-STING activation in microglia sustains a maladaptive inflammatory state that drives AD progression^15,16. Notably, genetic or pharmacological inhibition of this signaling pathway mitigates microglial activation and alleviates AD-related pathology^15,18, underscoring its pathogenic role. However, the upstream mechanisms governing this cytosolic mtDNA-initiated inflammatory cascade in microglia remain largely unknown.

Epigenetic regulation provides a critical layer of control over gene expression, enabling transient stimuli to be converted into long-lasting transcriptional programs^19–21. Among these mechanisms, histone acetylation plays a pivotal role: acetylation of lysine residues on histone tails generally relaxes chromatin structure and facilitates transcription²². This places histone acetyltransferases (HATs), also known as lysine acetyltransferases (KATs), at the core of transcriptional reprogramming²²; however, their contribution to microglial inflammation and AD pathogenesis remains poorly understood. Here, through gene expression profiling, we identify KAT7 (also known as HBO1), a member of the MYST family of HATs²³, as a key epigenetic regulator of neuroinflammation. By coupling histone acetylation to enhanced mitochondrial DNA synthesis and cGAS-STING activation in microglia, our work uncovers a previously unrecognized epigenetic mechanism driving chronic neuroinflammation and highlights KAT7 as a promising therapeutic target for AD.

RESULTS

Expression of the KAT7 complex is upregulated in microglia from both 5×FAD mouse and human AD brains

To determine whether epigenetic regulators of histone acetylation are altered during microglial activation, we analyzed public RNA sequencing (RNA-seq) datasets from primary mouse microglia²⁴ and human monocyte-derived microglia-like cells²⁵. Jade2, which encodes a scaffold subunit of the KAT7 complex, was selectively upregulated in response to lipopolysaccharide (LPS) stimulation (Figures 1A, 1B and S1A). We next performed RNAseq on LPS-treated BV2 cells, a mouse microglia-derived cell line, and also observed increased expression of Jade2 in response to LPS (Figure 1C), whereas expression of other HAT complexes remained largely unchanged. This was further validated by qPCR and western blot analyses (Figures 1D and S1B–S1D). Given that neuroinflammation is a hallmark of AD⁵, we examined whether expression and activity of the KAT7 complex are also elevated in this context. To this end, we isolated microglia from 6-month-old 5×FAD mice, a well-characterized AD model carrying five familial AD mutations²⁶, using CD11b microbeads (Figures S1E and S1F). qPCR analysis revealed that the expression of Kat7 and two of its scaffold subunits, Jade2 and Brpf2, was increased in microglia from 5×FAD mice compared with age-matched wild-type (WT) littermate controls (Figures 1E and S1G). RNAscope analysis further showed upregulation of Kat7 and Jade2 specifically in microglia, but not in neurons or astrocytes (Figures S2A and S2B). To extend these findings to humans, we analyzed an RNA-seq dataset from post-mortem superior frontal gyrus tissue of AD patients and healthy controls²⁷. While expression of KAT7 and its other subunits was unchanged, the scaffolds JADE2 and BRPF3 were upregulated in microglia from AD patients (Figures 1F and S2C). Because suitable antibodies for the KAT7 complex proteins were unavailable, we leveraged the fact that KAT7 is the primary enzyme responsible for H3K14 acetylation (H3K14ac) in cells^28–30 and performed immunostaining with an anti-H3K14ac antibody on brain sections. Microglial H3K14ac levels were low in WT controls but markedly elevated in 5×FAD mice, particularly in microglia near Aβ plaques (Figure 1G). In contrast, no changes in H3K14ac were observed in neurons or astrocytes (Figures S2D and S2E). As expected, microglial H3K14ac signals in 5×FAD mice co-localized with Jade2 mRNA detected by RNAscope (Figure S2F). Importantly, H3K14ac signals were also strongly increased in microglia from postmortem human AD brains compared with healthy controls (Figure 1H). Together, these results demonstrate that upregulation of the KAT7 complex and its histone acetylation mark accompanies microglial activation, implicating KAT7 in neuroinflammation and AD pathogenesis.

Figure 1. Expression of the KAT7 complex is elevated in microglia from both 5×FAD mice and human AD patients.

(A) Heatmap of HAT components from a public RNA-seq dataset (GSE90046, n=3 per group) in mouse primary microglia treated with LPS.

(B) Upregulation of JADE2 in human monocyte-derived microglial cells following LPS treatment, based on a public RNA-seq dataset (GSE187452, n= 6 per group).

(C) Heatmap of HAT components from RNA-seq data (n=3 per group) in BV2 cells treated with LPS.

(D) qPCR analysis of KAT7 complex in BV2 cells treated with LPS. n=4 per group.

(E) qPCR analysis of KAT7 complex in the isolated microglia from 6-month-old 5×FAD mice. n=4 per group.

(F) Upregulation of JADE2 and BRPF3 in microglia from AD patients based on GSE125050 dataset. n= 14 controls and 10 AD patients.

(G) Left: Representative images of H3K14ac co-stained with Aβ plaques and microglia (Iba1) in the cortex of 6-month-old WT and 5×FAD mice. Scale bar, 20 μm (left), 4 μm (right). Right: Quantification of H3K14ac intensity in microglia. n=100 cells from 4 mice per group.

(H) Left: Representative images of H3K14ac co-stained with Aβ plaques and microglia in the frontal lobe of AD patients and healthy controls. Scale bar, 20 μm (left), 6 μm (right). Right: Quantification of H3K14ac intensity in microglia. n=80 cells from 4 samples per group.

White arrowheads indicate H3K14ac in microglia from WT/control groups or plaque-adjacent microglia in AD samples. Yellow arrowheads indicate H3K14ac in plaque-distant microglia in AD samples. Unpaired student’s t test (B and D-F). Mann-Whitney test (G and H). *p<0.05, **p<0.01. Data are presented as mean ± SEM.

See also Figure S1, Figure S2 and Table S1.

KAT7 regulates LPS- and Aβ-induced inflammatory responses in microglia

To investigate the role of KAT7 in neuroinflammation, we generated Kat7-knockout (KO) BV2 microglial cells using CRIPSR-Cas9 (Figure 2A) and employed LPS stimulation as a well-established model of inflammatory activation. Kat7 deletion markedly reduced LPS-induced iNOS expression and secretion of the inflammatory cytokine IL-6 (Figures 2B–2D). Conversely, overexpression of WT KAT7 enhanced IL-6 production, whereas the catalytically inactive mutant (KAT7-E508Q) failed to do so (Figures 2E and 2F), indicating that the enzymatic activity of KAT7 is required for its pro-inflammatory function. We next examined the role of the scaffold subunit JADE2 in neuroinflammation. Overexpression of JADE2 increased KAT7 protein levels, suggesting that JADE2 stabilizes the KAT7 complex (Figure S3A). Moreover, JADE2 promoted the expression of pro-inflammatory factors in a manner dependent on its interaction with KAT7 (Figure S3B). To validate these findings in primary cells, we cultured microglia from neonatal mice and transfected them with Kat7-specific siRNAs (Figure 2G). Consistently, Kat7 knockdown significantly attenuated LPS-induced IL-6 production (Figures 2H–2J). Similarly, pharmacological inhibition of KAT7 with WM-3835, a potent small-molecule inhibitor²⁸, suppressed LPS-induced Nos2 and Il6 expression in a dose-dependent manner (Figure 2K) without obvious cytotoxicity (Figure S3C). Beyond LPS, we examined whether KAT7 regulates inflammation driven by aggregated Aβ. Treatment of primary microglia with oligomeric Aβ robustly induced Il6 and Nos2 expression, which was markedly attenuated by Kat7 knockdown and, to a lesser extent, by Jade2 knockdown (Figures S3D–S3G). Similarly, pharmacological inhibition of KAT7 with WM-3835 attenuated Aβ-induced production of the pro-inflammatory cytokines IL-6 and IL-1β (Figures 2L–2O). We further generated human induced pluripotent stem cells (iPSCs)-derived microglia-like cells (iMGs). Following doxycycline (DOX) induction, iPSCs efficiently differentiated into highly pure Iba1⁺ iMGs (Figures S3H–S3J). WM-3835 dose-dependently reduced Aβ-induced IL-6 secretion in iMGs (Figure 2P). Collectively, these results establish the KAT7-JADE2 complex as a critical epigenetic driver of microglial inflammatory responses to both LPS and Aβ.

Figure 2. KAT7 regulates LPS- and Aβ- induced inflammatory responses in microglia.

(A) Generation of Kat7-KO BV2 cells.

(B) Western blot analysis of KAT7 and iNOS levels in Kat7-KO BV2 cells treated with LPS.

(C-D) qPCR and ELISA analyses of IL-6 levels in Kat7-KO BV2 cells treated with LPS. n=3.

(E) Western blot analysis of KAT7 overexpression in BV2 cells.

(F) ELISA analysis of IL-6 secretion in BV2 cells treated with LPS. n=4.

(G) qPCR analysis of Kat7 level in mouse primary microglia transfected with siRNA. n=4.

(H-I) qPCR analysis of Nos2 and Il-6 levels in primary microglia treated with siRNA and LPS. n=3.

(J) ELISA analysis of IL-6 secretion in primary microglia. n=3.

(K) qPCR analysis of Nos2 and Il-6 levels in primary microglia treated with WM-3835 and LPS. n=3.

(L) Schematic diagram of mouse primary microglia treated with WM-3835 and oligomeric Aβ42.

(M) qPCR analysis of Il-6 levels in primary microglia treated with WM-3835 and oligomeric Aβ42. n=4.

(N-O) ELISA analysis of IL-6 secretion (6 h after Aβ42 treatment) and IL-1β (24 h after Aβ42 treatment) in primary microglia treated with WM-3835 and Aβ42. n=4.

(P) Left: Schematic diagram of induced human microglia (iMG) treated with oligomeric Aβ. Right: ELISA analysis of IL-6 secretion in iMGs treated with WM-3835 and Aβ42. n=3.

*p<0.05, **p<0.01, ***p<0.001. One-way ANOVA test (G, K, and M-P). Two-way ANOVA test (C, D, F, and H-J). Data are presented as mean ± SEM.

See also Figure S3.

Combined transcriptomic and epigenomic profiling reveals Cmpk2 as a critical transcriptional target of KAT7

To elucidate the molecular mechanisms by which KAT7 regulates neuroinflammation, we performed RNA-seq to profile transcriptomic changes in WT and Kat7 KO BV2 cells with or without LPS treatment (Figures 3A and S4A). Using an adjusted p-value < 0.05 and |log₂ fold change| > 1 as cutoffs, we identified 1,074 upregulated genes in response to LPS stimulation in WT cells (Figure 3B). Deletion of Kat7 attenuated the induction of 110 of these genes (Figures 3C, 3D and S4B). Gene ontology (GO) analysis revealed that these KAT7-dependent genes were significantly enriched in pathways related to interferon signaling and inflammatory responses (Figure 3D). In contrast, of the 276 genes downregulated by LPS in WT cells, only five were reversed by Kat7 deletion (Figure S4C), indicating that KAT7 has minimal impact on LPS-induced transcriptional repression. Given that KAT7 is primarily responsible for H3K14 acetylation, a histone mark associated with transcriptional activation³⁰, we next performed CUT&Tag (Cleavage Under Targets and Tagmentation) analysis to map the genome-wide distribution of H3K14ac (Figure 3E). Genomic distribution analysis revealed that most differential H3K14ac peaks were located in promoter regions: 38% (6,121 peaks) between WT_LPS and WT cells, and 40% (5,776 peaks) between KO_LPS and WT_LPS cells (Figures S4D and S4E). Among the genes showing increased H3K14ac enrichment at promoters in WT cells upon LPS stimulation, 244 exhibited reduced acetylation in Kat7 KO cells (Figures 3E and S4F). Integrative analysis of RNA-seq and CUT&Tag datasets identified 17 genes as potential direct transcriptional targets of KAT7 in response to LPS stimulation (Figures 3F and 3G). The limited overlap likely reflects the stringent thresholds applied. Validation by qPCR and quantitative chromatin immunoprecipitation (qChIP) using KAT7 and H3K14ac antibodies (with H3K23ac antibody as a negative control) confirmed the profiling results (Figures 3H and S4G–S4J). Given that the KAT7-JADE complex mediates H3K14ac in a chromatin context-dependent manner³¹, particularly in the presence of H3K4me3, we performed CUT&Tag for H3K4me3 in BV2 cells and observed its strong enrichment at the promoters of KAT7 target genes (Figure S4H). Among KAT7-regulated genes, Cmpk2 (Cytidine/uridine monophosphate kinase 2) was one of the most dramatically suppressed in Kat7 KO cells in response to LPS, highlighting it as a key downstream target (Figures 3H and S4G). Importantly, oligomeric Aβ induced KAT7-dependent enrichment of H3K14ac, but not H3K23ac, at the Cmpk2 promoter in primary mouse microglia (Figure 3I). Aβ-induced Cmpk2 expression was also markedly attenuated by Kat7 knockdown (Figures 3J and S4K). Together, these findings identify Cmpk2 as a direct transcriptional target of KAT7 and suggest that it functions as a shared downstream effector in microglial inflammatory responses to both LPS and Aβ.

Figure 3. Integrated transcriptomic and epigenomic analyses identify Cmpk2 as a key target of KAT7.

(A) Heatmap of RNA-seq data from WT and Kat7-KO BV2 cells with or without LPS stimulation. n=3 replicates per group.

(B) Volcano plot showing differentially expressed genes between WT_LPS and WT BV2 cells.

(C) Volcano plot showing differentially expressed genes between Kat7-KO_LPS and WT_LPS BV2 cells.

(D) Left: Venn diagram of overlapping genes among those downregulated in KO vs WT, downregulated in KO_LPS vs WT_LPS, and upregulated in WT_LPS vs WT. Right: GO pathway analysis of the 110 overlapping genes.

(E) Heatmap of CUT&Tag data using IgG and H3K14ac antibodies in WT and Kat7-KO BV2 cells with or without LPS stimulation.

(F) Top: Venn diagram showing overlapped genes between RNA-seq (110 genes) and CUT&Tag (244 genes). Bottom: Heatmap of the 17 overlapped genes in RNA-seq.

(G) Representative CUT&Tag tracks of H3K14ac at Cmpk2. Green box indicated proximal promoter. TSS, transcriptional start site.

(H) qPCR analysis of Cmpk2 in BV2 cells treated with LPS. n=3.

(I) qChIP analysis of the Cmpk2 promoter using the indicated antibodies in primary microglia treated with Aβ. n=3.

(J) qPCR analysis of Cmpk2 in primary microglia treated with Kat7 siRNA and oligomeric Aβ42. n=3.

*p<0.05, **p<0.01, ***p<0.001. One-way ANOVA test. Data are presented as mean ± SEM.

See also Figure S4 and Table S6.

KAT7 drives microglial innate immune signaling via CMPK2-dependent mtDNA synthesis

CMPK2 encodes a mitochondrial nucleotide monophosphate kinase that acts as the critical rate-limiting enzyme, ensuring dNTP precursor availability and driving the dramatic upregulation of mitochondrial DNA (mtDNA) synthesis during macrophage activation³². CMPK2-dependent mtDNA synthesis facilitates the release of mtDNA into the cytoplasm, which subsequently activate the cGAS-STING signaling pathway and the NLRP3 inflammasome^32–34. Given that KAT7 regulates Cmpk2 expression during microglia activation, we tested whether it modulates mtDNA replication, release, and downstream innate immune responses. We first performed 5-ethynyl-2’-deoxyuridine (EdU) labeling, which preferentially incorporates into newly synthesized mtDNA and appears as bright cytoplasmic puncta in non-proliferating primary mouse microglia (Figures 4A and 4B). Kat7 knockdown impaired LPS-induced mtDNA replication, an effect rescued by WT CMPK2 but not by a catalytically inactive CMPK2 mutant (CMPK2-D330A) (Figures 4A, 4B and S5A–S5C), indicating that KAT7 promotes CMPK2-dependent mtDNA synthesis in response to LPS. To assess mtDNA release, we primed microglia with LPS followed by ATP treatment, a common method to induce mitochondrial stress and trigger innate immunity³⁴. qPCR analysis of cytosolic fractions using primers specific to the mitochondrial D-loop region revealed elevated mtDNA levels in control microglia (Figure 4C), indicating increased release from damaged mitochondria. Notably, Kat7 knockdown reduced cytosolic mtDNA levels (Figure 4C). Consistently, it led to reduced phosphorylation of TBK1 (Ser172, p-TBK1) and IRF3 (Ser396, p-IRF3) (Figures 4D–4G and S5D), as well as diminished IL-1β production (Figure 4H), indicating attenuated cGAS-STING activation and NLRP3 inflammasome signaling. Importantly, CMPK2 overexpression restored cytosolic mtDNA levels and innate immune signaling in KAT7-deficient cells, whereas the catalytically inactive mutant failed to do so (Figures 4C–4H). In addition, we validated these findings in an Aβ context, showing that Kat7 knockdown impaired Aβ-induced mtDNA replication and reduced p-TBK1 levels, both of which were rescued by CMPK2 overexpression (Figures S5E and S5F). Together, these results demonstrate that KAT7 orchestrates CMPK2-dependent mtDNA synthesis and release, establishing it as a critical epigenetic regulator of innate immune pathways during microglial activation.

Figure 4. Kat7 knockdown in microglia limits mtDNA replication and release by repressing Cmpk2 transcription.

(A-B) Representative images (A) and quantification (B) of EdU-labeled newly synthesized mtDNA in primary microglia treated with LPS. Scale bar, 5 μm. n=20 cells from 3 replicates per group.

(C) Quantification of cytosolic mtDNA (cmtDNA) by qPCR (normalized to nuclear B2m DNA) in primary microglia treated with LPS plus ATP. The D-loop refers to a specific fragment within mtDNA. n=3.

(D-E) Representative images (D) and quantification (E) of p-TBK1 in primary microglia treated with LPS plus ATP. Scale bar, 5 μm. n=10 cells from 2 replicates per group.

(F-G) Representative images (F) and quantification (G) of p-IRF3 in nuclei of primary microglia treated with LPS plus ATP. Scale bar, 3 μm. n=10 cells from 2 replicates per group.

(H) ELISA analysis of IL-1β secretion from primary microglia treated with LPS plus ATP. n=3.

**p<0.01, ***p<0.001. One-way ANOVA test (E and G). Two-way ANOVA test (B, C and H). Data are presented as mean ± SEM.

See also Figure S5.

Microglia-specific Kat7 deletion attenuates neuroinflammation in 5×FAD mice

Given our findings of elevated microglial KAT7 complex expression in AD, and its critical role in regulating neuroinflammation in vitro, we hypothesized that KAT7 promotes AD pathogenesis in vivo by driving cytosolic mtDNA-induced innate immune signaling. To test this, we generated Kat7 floxed mice using the efficient additions with single-strand DNA inserts CRISPR (Easi-CRISPR) method³⁵. Exon 2 of Kat7 was flanked by loxP sites, and its deletion induces a frameshift mutation resulting in functional KO (Figures S6A–S6C). These mice were crossed with Cx3cr1-CreER mice³⁶ to achieve microglia-specific knockout (Kat7 cKO) and subsequently bred with 5×FAD mice (Figure S6D). Tamoxifen was administered at 2 months of age to avoid perturbing microglial development, and biochemical, pathological, and behavioral assessments were performed at 6 months. Microglia isolated from Kat7 cKO mice exhibited efficient Kat7 deletion (Figure 5A), whereas its expression was not altered in non-microglial cells (Figure S6E). Immunofluorescence analysis confirmed that Kat7 deletion markedly reduced H3K14ac levels in microglia (Figure S6F). Notably, Cmpk2, which was upregulated in 5×FAD microglia, was reduced by Kat7 deletion (Figure 5B). Given KAT7’s role in mtDNA synthesis and release, we assessed cytosolic mtDNA levels by qPCR analysis and found them elevated in 5×FAD microglia but significantly attenuated in Kat7 cKO; 5×FAD mice (Figure 5C). Correspondingly, phosphorylation of TBK1 and IRF3, key mediators of the cGAS-STING pathway, was markedly reduced in KAT7-deficient microglia, along with decreased levels of the proinflammatory cytokines IL-1β and IL-6 (Figures 5D–5F and S6G). Further analysis revealed that microglial activation was suppressed in Kat7 cKO; 5×FAD mice, as evidenced by diminished levels of phosphorylated p65 (p-p65) (Figure 5G), a key NF-κB effector downstream of cGAS-STING signaling³⁷, and a decreased number of Iba1⁺ microglia due to reduced proliferation (Figures 5H and S6H–S6J).

Figure 5. Microglia-specific Kat7 deletion inhibits neuroinflammation in 5×FAD mice.

(A) Knockout efficiency of Kat7 in microglia was confirmed by western blot (left) and qPCR analysis (right).

(B) qPCR analysis of Cmpk2 and Mx2 in microglia isolated from 6-month-old mice. n=4.

(C) Quantification of cytosolic mtDNA by qPCR (normalized to nuclear B2m DNA) in microglia isolated from 6-month-old mice. D-loop and Nd4 indicate mtDNA fragments. Tert and B2m indicate nuclear DNA. n=3.

(D) Representative images of p-TBK1 staining in the cortex of 6-month-old mice. Scale bar, 10 μm (left), 4 μm (right). Right: Quantification of mean p-TBK1 intensity per microglia per mouse. n=4 mice per group.

(E) Representative images of p-IRF3 staining in the cortex of 6-month-old mice. Scale bar, 5 μm (left), 2 μm (right). Right: Quantification of mean p-IRF3 intensity per microglia per mouse. n=4 mice per group.

(F) ELISA analysis of IL-1β and IL-6 production in cortex region of 6-month-old mice. n=3 mice per group.

(G) Immunoblot analysis of p65 and p-p65 in the cortex of 6-month-old mice (left), with quantification normalized to β-actin (right). n=4 mice per group.

(H) Representative images (left) and quantification (right) of Iba1⁺ microglia in 6-month-old mice. Scale bar, 500 μm (left), 100 μm (right). n=4 mice per group.

(I) UMAP plots of microglia clusters identified by scRNA-seq in 7-month-old mice.

(J) Relative proportions of microglial subclusters from scRNA-seq in indicated groups.

(K) Volcano plot showing differentially expressed genes in major DAM clusters (clusters 3 and 4).

(L) UMAP plots showing reduced microglial IL-6 expression in cKO;AD mice compared with AD mice.

Unpaired student’s t test (A and H). One-way ANOVA test (D-G). Two-way ANOVA test (B and C). *p<0.05, **p<0.01, ***p<0.001. Data are presented as mean ± SEM.

See also Figure S6 and Figure S7.

To assess the role of KAT7 in regulating microglial subpopulations, we performed single-cell RNA-seq (scRNA-seq) of CD11b⁺/CD45^low myeloid cells isolated from 7-month-old mouse brains (Figure S7A). We analyzed approximately 11,000 microglia from control (WT), Kat7 cKO, 5×FAD, and Kat7 cKO;5×FAD mice. Uniform Manifold Approximation and Projection (UMAP) analysis identified eleven transcriptionally distinct subclusters (Figures 5I and S7B). These subclusters were annotated based on marker gene expression profiles and differences between WT and 5×FAD mice, including homeostatic microglia (P2ry12, Tmem119, and Cx3cr1; clusters 1 and 2) and disease-associated microglia (DAM; clusters 3 and 4) (Figures 5J and S7C). Notably, the homeostatic microglial subcluster 2 was markedly reduced in 5×FAD mice, decreasing from 33.7% to 13.2%, but partially recovered to 18.8% in cKO;5×FAD mice (Figure 5J). Although the proportion of major DAM clusters (clusters 3 and 4) showed only a slight decrease in cKO;AD mice compared with AD mice, the expression of inflammatory genes was markedly reduced, including Il6, Nox2 and Ifitm2 (Figures 5J–5L, S7D and S7E). GO analysis revealed that these downregulated genes were mainly enriched in interferon signaling and inflammatory response pathways (Figure S7F). Although few cells expressed Cmpk2, likely due to limited sequencing depth, its expression in major DAM clusters was elevated in AD microglia compared to WT and attenuated by Kat7 deletion (Figures S7G and S7H). Collectively, these results demonstrate that microglial Kat7 deletion suppresses neuroinflammation in 5×FAD mice.

Microglia-specific Kat7 deletion reduces Aβ pathology and improves cognition in 5×FAD mice

Growing evidence indicates that microglia-driven neuroinflammation promotes Aβ deposition and disrupts synaptic activity, ultimately contributing to cognitive decline^5,38,39. Our scRNA-seq data show that microglial Kat7 deletion not only reduces inflammatory gene expression but also upregulates genes involved in Aβ clearance, such as Scara1 (Figures 5K and S7E), a key receptor mediating Aβ uptake⁴⁰. To investigate whether the effects of microglial Kat7 deletion on Aβ pathology depend on the timing of gene ablation, tamoxifen was administered at 2 or 4 months of age, and tissues were harvested at 6 months. At both induction time points, Kat7 cKO;5×FAD mice exhibited a significantly reduced Aβ plaque burden compared with 5×FAD controls, as quantified by thioflavin S-positive plaques (Figures 6A–6D). Notably, Kat7-f/f;Cx3cr1-CreER;5×FAD mice that did not receive tamoxifen showed no reduction in Aβ plaques (Figures 6A–6D), suggesting that the heterozygous Cx3cr1-CreER allele alone does not account for the phenotype. To evaluate synaptic plasticity, we performed field potential recordings to measure long-term potentiation (LTP) at Schaffer collateral-CA1 synapses in acute hippocampal slices. The LTP deficits characteristic of 5×FAD mice were significantly rescued by microglial Kat7 deletion (Figures 6E and 6F). We next assessed hippocampus-dependent spatial learning and memory using the Morris water maze. In line with restored synaptic function, Kat7 cKO; 5×FAD mice demonstrated accelerated learning during training (Figure 6G) and superior memory retention in the probe trial, spending more time in the target quadrant and crossing the former platform location more frequently compared to 5×FAD controls (Figures 6H–6J). Notably, swimming speed was comparable among groups (Figure 6K), ruling out motor deficits as a confounding factor. Collectively, these results demonstrate that microglial Kat7 deletion reduces Aβ accumulation, restores synaptic function, and improves cognitive performance in 5×FAD mice.

Figure 6. Microglia-specific Kat7 deletion ameliorates Aβ pathology and improves cognitive function in 5×FAD mice.

(A) Representative images of Thioflavin-S (TS) staining in brain sections from 6-month-old mice. Tamoxifen (TAM) was administered at 2 months of age. Right: Enlarged images of the cortical region. Scale bar, 0.5 mm (left), 0.1 mm (right).

(B) Quantification of TS-positive Aβ plaque number and area in 6-month-old AD mice. n=4 mice per group.

(C) Representative images of TS staining in brain sections from 6-month-old mice. Tamoxifen (TAM) was administered at 4 months of age. Right: Enlarged images of the cortical region. Scale bar, 0.5 mm (left), 0.1 mm (right).

(D) Quantification of TS-positive Aβ plaque number and area in 6-month-old AD mice. n=4 mice per group.

(E) TBS-induced LTP at Schaffer collateral to CA1 synapses in 6-month-old mice. Arrow indicates LTP induction. Representative traces show fEPSP recorded before (1) and 50 min after (2) TBS.

(F) Averaged fEPSP slopes during 50 to 60 min after the stimulation. n=10–11 slices from 5 mice per group.

(G) Time spent before reaching the hidden platform during training days in the Morris water maze test. Two-way ANOVA test.

(H-K) Representative swim paths (H), time spent in target quadrant (I), number of platform zone crossings (J) and swimming speed (K) during the probe test. n=11–15 mice per group.

One-way ANOVA test (B, D, F, and I-K). *p<0.05, **p<0.01. ns, not significant. Data are presented as mean ± SEM.

Pharmacological inhibition of KAT7 mitigates neuroinflammation, reduces Aβ burden, and improves cognition in 5×FAD mice

To determine whether pharmacological inhibition of KAT7 confers protection against AD, we first examined the effects of the KAT7 inhibitor WM-3835 in primary mouse microglia treated with oligomeric Aβ42. Consistent with our knockdown results (Figures 3J, 4A and 4B), WM-3835 treatment suppressed Aβ-induced Cmpk2 upregulation and markedly reduced CMPK2-dependent mtDNA synthesis (Figures 7A and 7B). Before evaluating KAT7 inhibition in vivo, we examined whether Kat7 deletion impacts brain function beyond early development, given its neuronal expression and essential roles in neural stem cell differentiation and cortical development³⁰. We generated excitatory neuron-specific Kat7 cKO mice using CamKII-Cre (Figures S8A and S8B), which induces deletion beginning at 2–3 weeks of age⁴¹. Unlike mice with embryonic Kat7 deletion³⁰, these cKO mice were viable, healthy, and fertile. Electrophysiological recordings from hippocampal CA1 pyramidal neurons in acute slices revealed normal miniature excitatory postsynaptic current (mEPSC) amplitude and frequency, as well as intact LTP at Schaffer collateral-CA1 synapses (Figures S8C and S8D), indicating preserved synaptic transmission and plasticity. These results suggest that loss of KAT7 is well tolerated in mature neurons. We then evaluated the therapeutic potential of WM-3835 in vivo. Five-month-old 5×FAD mice received intracerebroventricular (ICV) infusion of WM-3835 for four weeks via an osmotic pump (Figure 7C). Post-treatment analysis confirmed robust target engagement, evidenced by a pronounced reduction in H3K14ac levels across the brain (Figures 7D and 7E). Notably, WM-3835 significantly decreased microgliosis (Figure 7F), lowered cytosolic mtDNA levels (Figure 7G), and attenuated cGAS-STING pathway activation in microglia (Figure 7H). Consistent with these effects, WM-3835 treatment also markedly reduced Aβ plaque burden in the hippocampus and cortex (Figure 7I). Behaviorally, WM-3835-treated 5×FAD mice exhibited improved spatial learning and memory in the Morris water maze, with faster acquisition during training and superior performance in the probe trial (Figures 7J–7L). These cognitive benefits closely mirrored those observed in microglia-specific Kat7 cKO mice, although we cannot exclude potential contributions from other cell types. Together, our results demonstrate that pharmacological inhibition of KAT7 mitigates microglial activation, reduces Aβ pathology, and improves cognitive function in 5×FAD mice, providing preclinical evidence for WM-3835 as a potential therapeutic strategy for AD.

Figure 7. Pharmacological inhibition of KAT7 reduces neuroinflammation and Aβ burden in 5×FAD mice.

(A) qPCR analysis of Cmpk2 levels in primary microglia treated with WM-3835 and oligomeric Aβ42. n=4.

(B) Representative images (left) and quantification (right) of EdU-labeled newly synthesized mtDNA in primary microglia treated with WM-3835 and Aβ42. Scale bar, 4 μm. n=30 cells from 3 replicates per group.

(C) Schematic diagram illustrating intracerebroventricular (ICV) delivery of vehicle or WM-3835 via osmotic pump in 5-month-old 5×FAD mice.

(D-F) Representative images (D) and quantification of H3K14ac intensity in microglia (E) and Iba1⁺ microglia number (F) in the cortex of 6-month-old AD mice treated with vehicle or WM-3835. Scale bar, 20 μm. n=4 mice per group.

(G) Quantification of cytosolic mtDNA by qPCR (normalized to nuclear B2m DNA) in microglia isolated from 6-month-old AD mice treated with vehicle or WM-3835. D-loop and Nd4 indicate mtDNA fragments. B2m indicates nuclear DNA.

(H) Representative images (left) and quantification (right) of p-TBK1 in microglia from 6-month-old AD mice treated with vehicle or WM-3835. Scale bar, 10 μm. n=4 mice per group.

(I) Representative images (left) and quantification (right) of TS staining in the cortex and hippocampus of 6-month-old AD mice treated with vehicle or WM-3835. Scale bar, 0.4 mm (left), 0.1 mm (middle and right). n=4 mice per group.

(J) Time to reach the hidden platform during training in the Morris water maze test.

(K-L) Time spent in each quadrant (K) and swimming speed (L) during the probe test. n=10 mice per group.

*p<0.05, **p<0.01, ***p<0.001. Unpaired student’s t test (E, F, H and L). One-way ANOVA test (A-B). Two-way ANOVA test (G, and I-K). Data are presented as mean ± SEM.

See also Figure S8.

DISCUSSION

Cytosolic mtDNA has emerged as a potent driver of neuroinflammation through activation of innate immune signaling. Here, we identify the histone acetyltransferase KAT7 as a critical upstream regulator of this mtDNA-initiated inflammatory cascade in microglia (Figure 8). Mechanistically, KAT7 promotes H3K14 acetylation at the Cmpk2 promoter during microglial activation, thereby enhancing Cmpk2 expression, increasing mtDNA synthesis and release, and fueling downstream inflammatory responses. Multiple lines of evidence indicate that this pathway is active in AD. First, the expression of KAT7 complex and its histone mark H3K14ac are elevated in microglia from 5×FAD mice and human AD brains. Second, Aβ oligomers induce Cmpk2 expression and mtDNA replication in primary microglia, processes that depend in part on KAT7 expression or activity. Third, Cmpk2 expression and cytosolic mtDNA levels are elevated in microglia from 5×FAD mice, both of which are reduced upon Kat7 deletion. Finally, both microglia-specific Kat7 deletion and pharmacological inhibition suppress innate immune signaling and neuroinflammation, leading to reduced Aβ burden and improved cognitive function in 5×FAD mice. Our findings thus reveal an epigenetic-mitochondrial axis that mechanistically links Aβ deposition to chronic immune activation. Disrupting this self-perpetuating loop offers a potential therapeutic strategy for AD and related neurodegenerative disorders.

Figure 8. Summary diagram.

Our results support a model in which KAT7 acts as a central epigenetic driver of neuroinflammation by promoting H3K14ac-dependent Cmpk2 expression. Through this mechanism, KAT7 integrates nuclear and mitochondrial responses to amplify neuroinflammatory signaling, thereby contributing to Aβ accumulation and synaptic dysfunction in AD.

Microglial activation requires extensive metabolic reprogramming to meet heightened bioenergetic and biosynthetic demands^42,43. KAT7-mediated upregulation of CMPK2, a key enzyme for mtDNA replication, may help fulfill these needs. However, while essential for sustaining microglial activation, excessive mtDNA replication renders mitochondria vulnerable to stress-induced mtDNA release, thereby amplifying innate immune signaling. Upon pro-inflammatory activation, microglia undergo a metabolic shift from oxidative phosphorylation to aerobic glycolysis^44,45. This glycolytic reprogramming can promote mitochondrial stress, leading to the release of mtDNA into the cytosol⁴⁶. In turn, cytosolic mtDNA activates innate immune pathways such as cGAS-STING and NLRP3, which further reinforce glycolytic flux while suppressing oxidative phosphorylation^12,33. In this context, CMPK2-driven mtDNA replication may potentiate this feed-forward loop by increasing the availability of mtDNA for release under stress conditions. Notably, a similar CMPK2-mtDNA pathway drives cGAS-STING and NLRP3 activation in macrophages during inflammation^32,34,47, underscoring its conserved role in innate immunity across myeloid lineages. Future studies will be required to directly define the role of CMPK2 and its downstream mtDNA replication in neuroinflammation and AD. Of note, a recent study showed that CMPK2 is also upregulated in microglia during ischemic stroke⁴⁸, where it promotes neuroinflammation and brain injury. Therefore, the KAT7-CMPK2 pathway may represent a general mechanism by which microglia couple metabolic reprogramming to innate immune activation across diverse chronic inflammatory conditions.

Histone acetylation in neurons has long been associated with learning and memory^49–51, and several altered histone acetylation marks have been reported in AD mouse models and patients^52–54. However, the functional relevance of individual histone modifications in AD remains poorly defined. Here, we show that KAT7-dependent H3K14ac is low in microglia from WT mice and healthy controls but markedly and specifically elevated in AD brains. Functional studies further reveal a microglia-specific pathogenic role for KAT7 in 5×FAD mice. These findings highlight the importance of examining histone acetylation and other epigenetic modifications in a cell type-specific manner. Notably, KAT7 has also been implicated in ageing and cellular senescence, where it enhances transcription of the cyclin-dependent kinase inhibitor p15^INK4b to promote cell cycle arrest⁵⁵. Downregulation of Kat7 in aged mice reduces hepatocyte senescence, alleviates liver inflammation, and extends lifespan⁵⁵. Given its beneficial effects in both AD and aging contexts, KAT7 inhibition may represent a promising strategy to mitigate neurodegeneration and promote brain resilience across the lifespan.

Our data establish CMPK2 as a key target of KAT7, but additional downstream genes, including several interferon-responsive genes, are also likely to contribute to chronic neuroinflammation. While KAT7 is best known for catalyzing H3K14ac, it also mediates other histone acylations, such as propionylation and crotonylation⁵⁶, and can modify both alternative histone sites and non-histone substrates^23,57. These activities may further diversify its regulatory functions and shape the transcriptional landscape of activated microglia. Regardless, pharmacological inhibition of KAT7 is expected to suppress all of these activities, providing a unified strategy to blunt its proinflammatory effects in AD. While the 5×FAD model used in this study recapitulates robust Aβ pathology, it does not develop tau pathology, a defining feature of human AD. Future studies employing tauopathy or mixed pathology models will therefore be essential to fully delineate the role of KAT7 in AD pathogenesis. Nevertheless, our findings identify the KAT7-CMPK2 axis as a critical regulator of microglial mitochondrial immunity, linking epigenetic control to neuroinflammation and AD progression. These insights highlight the epigenetic-mitochondrial axis as a promising therapeutic target not only for AD but also for other neurological diseases characterized by chronic neuroinflammation.

STAR METHODS

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human brain samples

Formalin-fixed paraffin-embedded (FFPE) human postmortem brain samples from AD patients and controls (aged-matched and died of certain cause unrelated to dementia) were obtained from Johns Hopkins Brain Resource Center. Subject demographics were listed in Tables S1. The study using patient samples/data was approved by the Johns Hopkins University School of Medicine Office of Human Subjects Research Institutional Review Boards.

Mice

All procedures related to animal care and treatment were approved by the Johns Hopkins University Animal Care and Use Committee and met the guidelines of the National Institute of Health Guide for the Care and Use of Laboratory Animals. All animals were group housed in a standard 12-hour light/dark cycle with ad libitum access to food and water. The following mouse lines were used for the experiments: C57BL/6J (Jackson Laboratory, 000664), 5×FAD (Jackson Laboratory, 034848), Cx3cr1-Cre^ER (Jackson Laboratory, 021160, EYFP expression is weak and does not interfere with immunofluorescence staining or flow cytometry analysis), Camk2a-Cre (Jackson Laboratory, 005359). Kat7 floxed mice were generated at Transgenic Core of Johns Hopkins University. Male mice were used for all experiments except the behavioral tests, where both male and female mice were included, and no sex-specific effects were observed. Mice with Kat7 specific knockout in microglia were induced by tamoxifen (S1238, Selleck). Briefly, tamoxifen was dissolved in corn oil (C8267, Sigma) to a final concentration of 20 mg/ml. Mice at 2 or 4 months of age received tamoxifen or vehicle via intraperitoneal injection at 100 mg/kg for five consecutive days and were sacrificed at 6–7 months of age.

Generation of Kat7 floxed mice

Kat7 floxed mice were generated at Transgenic Core of Johns Hopkins University using the Easi-CRISPR method, as previously described³⁵. Two single-guide RNAs (sgRNAs) were designed by http://crispor.tefor.net/. The sequences were as follows: sgRNA #1 (reverse strand), AAGTACCAAGTTCCAACATAAGG; and sgRNA #2 (forward strand), GATACTGCTCCTGAGCTTGATGG. Two crRNAs containing each sgRNA and ssDNA donor containing the homology arms and floxed exon sequences were custom synthesized from IDT company. The annealed crRNA and tracrRNA (IDT) were diluted in microinjection buffer (0.25 mM EDTA and 10 mM Tris-HCl, pH 7.4) and mixed with Cas9 protein (30 ng/μl; IDT) to obtain ctRNP complexes. One-cell embryos of C57BL/6J mice were microinjected with a mixture of floxing ssDNA donors and two ctRNP complexes and were transferred into the oviducts of pseudopregnant ICR females (Charles River Laboratories). Successful insertions of two LoxP sites were detected by PCR genotyping of mouse tails and confirmed by Sanger sequencing. The primers used are provided in Table S2.

BV2 cell culture

The mouse microglial BV2 cell line was a gift from Dr. Tony Wyss-Coray’s laboratory at Stanford University⁵⁸. Cells were cultured in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin and maintained in an incubator at 37 °C with 5% CO₂. Adherent cells were split using 1× TrypLE (Gibco).

KAT7-KO BV2 cells were generated using CRISPR-Cas9 method. Guide RNA (GACTCGGGCAGATCGGCGCG) targeting mouse Kat7 was cloned into LentiCRISPR-v2-Puro (Addgene, #52961)⁵⁹. The primers used to design the single-guide RNA (sgRNA) targets were (5’ to 3’) Kat7 forward CACCGGACTCGGGCAGATCGGCGCG and Kat7 reverse AAACCGCGCCGATCTGCCCGAGTCC. Lentiviral particles containing Kat7 sgRNA were packaged using the 3rd generation lentivirus system and used to infect BV2 cells. One day after infection, the medium was changed to fresh DMEM containing 10% FBS and 1% P/S. Cells were then treated with puromycin (4 μg/ml) for 5 days to select for successfully transduced cells. Single clones were obtained using limiting dilution and were analyzed by western blotting and Sanger sequencing to confirm KAT7 deletion. A scrambled gRNA control was also used as negative control (5’-GCGCCAAACGTGCCCTGACG-3’).

For the overexpression in BV2 cells, the GFP in the lentivirus vector pLenti-EF1a-GFP-P2A-Puro (a gift from Dr. Shuying Sun lab at Johns Hopkins University) was replaced by mouse CMPK2, CMPK2-D330A, human KAT7, KAT7-E508Q and JADE2 DNA fragments at the AgeI and BamHI site using NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs, #E5520). Lentiviral particles were packaged using the 3rd generation lentivirus system and used to infect BV2 microglia. Cells were then treated with puromycin (4 μg/ml) to select the transduced cells.

Human induced microglia (iMG) differentiation

Differentiation of KOLF2.1J CLYBL 6-TF-iMG iPSC line (JIPSC002072, Jackson Laboratory) into i-Microglia was performed with minor modifications to a previously described protocol^60,61. hiPSCs were maintained in mTeSR Plus medium (StemCell Technologies) supplemented with Y-27632 (StemCell Technologies) for at least two passages prior to differentiation. On day 0, cells were plated onto double-coated culture plates. The first coating consisted of Poly-D-Lysine (0.1 mg/mL, Gibco) diluted in culture-grade water (Gibco) and incubated at 37°C overnight. This was followed by a second coating with Matrigel (Corning) diluted in F-12 (1x, Gibco) and incubated at 37°C for 2 hours. Cells were seeded in mTeSR Plus medium at a density of 250,000 cells per 60 mm dish. On day 2, the medium was replaced with Advanced DMEM/F12 (Gibco) supplemented with doxycycline (2 μg/ml, Sigma) 1% GlutaMAX (Gibco), 100 ng/ml IL-34 (BioLegend), and 10 ng/ml GM-CSF (BioLegend). On day 4, a full medium change was performed, and 50 ng/ml of M-CSF (BioLegend) and 50 ng/ml of TGF-β1 (PeproTech) were added. The medium was then replaced every 2–3 days thereafter. i-Microglia differentiated for 12 days were stimulated with oligomeric Aβ42 (1 μM, rPeptide) for 6 h. Following stimulation, cells or conditioned medium were collected for experiments.

METHOD DETAILS

Primary microglial cell preparation and stimulation

Primary microglial cells were prepared from neonatal mice (day 0–2). Briefly, brain tissues were quickly removed, and the meninges were carefully stripped in ice-cold HBSS. The cortex and hippocampi were then digested with 0.25% trypsin (Quality Biological) at 37 °C for 15 min and gently pipetted to generate single cells with DMEM containing 10% heat-inactivated fetal bovine serum (FBS; Avantor) and 1% penicillin/streptomycin (P/S; Quality Biological), followed by plating on poly-D-lysine-coated T75 flasks. After 11–14 days, primary microglia were separated from the mixed glial culture using a shake-off method (90 rpm for 2 hours). The collected microglia were seeded in the poly-D-lysine-coated plates and its purity was confirmed by Iba1 immunostaining.

For LPS stimulation, microglial cells were cultured in basic DMEM without FBS and treated with LPS (0.2 μg/ml, Sigma) for 6 h. For the assessment of mtDNA release, microglia were primed with LPS (0.2 μg/ml, Sigma) for 6 h, followed by ATP (2 mM, Sigma) treatment for 30 min. For oligomeric Aβ42 stimulation, microglial cells were cultured in basic DMEM without FBS and treated with oligomeric Aβ42 (1 μM, rPeptide) for 6 h or 24 h.

Isolation of microglia from adult mouse brain

6-month-old mice were anesthetized with isoflurane and perfused transcardially with cold saline. Brain tissue was freshly harvested, cut into small pieces, and digested with collagenase (Type IV, 5 mg/ml, Sigma) and DNase I (50 μg/ml, Sigma) for 1 hour at 37 °C with 250 rpm. The digested brain tissues were transferred to a 15 ml Dounce homogenizer and homogenized gently on ice. Brain tissue homogenates were suspended in HBSS, filtered with cell strainers (70 μm), and centrifuged at 500g for 5 min (4 °C) to collect the cell pellets. Then, 90% Percoll solution was prepared using absolute Percoll (Cytiva) and 10× HBSS (9:1, v/v), and further diluted (v/v) to 70, 37, and 30% with 1× HBSS. Cell pellets were suspended in a 37% Percoll solution. Microglia were isolated by density gradient centrifugation. Density gradient was added into 15 ml tubes, by layers of Percoll solution from bottom to top containing: 70%, 37%, and 30% Percoll solution and HBSS. Centrifugation was carried out in a horizontal centrifuge at 2000g for 30 min (4 °C). Microglia were converged on the interphase between the 37% and 70% Percoll solution. Isolated microglia were washed with 10× volumes of PBS and centrifuged at 500g for 5 min (4 °C). For MACS-based separation, microglia were further purified using CD11b MicroBeads (Miltenyi Biotec, #130-093-634) according to the manufacturer’s protocol. For FACS-based separation, cells were stained with fluorophore-conjugated antibodies against CD11b (BioLegend, #101256) and CD45 (BioLegend, #103116) and subsequently sorted using a SONY MA900 cell sorter.

siRNA transfection

siRNAs were purchased from Dharmacon and transfected into primary microglia with the final concentration of 40 nM using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer’s instructions. A negative control siRNA (siControl) was included to account for potential transfection effects. This control induced only a modest increase in Il6 (~1.5-fold), while Tnf levels remained unchanged. In contrast, Il6 induction in response to LPS or Aβ reaches thousands-fold, indicating that the minor increase observed with siRNA transfection is negligible for downstream analyses. The sequences of siRNA used in this study are as follows: siControl sense, UGGUUUACAUGUCGACUAA; mouse Kat7 siRNA#1 sense, GAACCGAAGAUUCCGAUUU; siRNA#2 sense, UGUUUGAAGUAGACGGCAA; mouse Jade2 siRNA sense, UGGAUGAGAUCGAUGCGUA.

Preparation of oligomeric Aβ

HFIP (hexofluoro-isopropanol) treated human Aβ_1–42 peptides (rPeptide, A-1163) were first dissolved in dimethyl sulfoxide (DMSO) to a final concentration of 5 mM, and this solution was then diluted with cold phenol-free basal culture media to a final concentration of 250 μM. Oligomeric Aβ_1–42 was prepared by incubation for 24 h at 4 °C, and this solution was aliquoted and stored at −80°C before use.

Enzyme-linked immunosorbent assay (ELISA)

The collected cultured medium of BV2 cells, primary microglia or human induced microglia was centrifuged at 500 g for 5 min (4°C) and the supernatant was processed for analysis with ELISA kits of mouse IL-6 (431301, BioLegend), mouse IL-1β (432601, BioLegend), and human IL-6 (430501, BioLegend), according to the manufacturer’s instructions. For cortical IL-6 and IL-1β detection, cortex tissues were dissected from the indicated mice and homogenized in RIPA buffer (Sigma) containing protease inhibitor cocktails (Roche). Total IL-6 and IL-1β protein levels were quantified by the ELISA kit and normalized first to total protein concentration determined using the Pierce BCA Protein Assay Kit (Thermo), and then normalized to the floxed (WT) group.

Western blotting

Proteins were isolated from cultured cells or brain tissues with RIPA buffer (Sigma) containing protease inhibitor cocktails (Roche). Samples were separated on Novex Tris-Glycine Mini Protein Gels (4 to 20%, Invitrogen) and transferred to nitrocellulose membranes (Bio-Rad), which were incubated with appropriate antibodies for overnight at 4°C. Primary antibody concentrations were as follows: anti-KAT7 (rabbit, 1:1000, Cell Signaling Technology, #58418), anti-JADE2 (rabbit, 1:2000, Proteintech, 11513–1-AP), anti-iNOS (rabbit, 1:3000, GeneTex, GTX130246), anti-Iba1 (rabbit, 1:1000, Wako, #019–19741), anti-p65 (rabbit, 1:1000, Cell Signaling Technology, #8242), anti-phospho-p65 (Ser536) (rabbit, 1:1000, Invitrogen, #MA5–15160), anti-CMPK2 (rabbit, 1:1000, Novus Biologicals, # NBP1–80653), anti-GAPDH (mouse, 1:5000, Proteintech, 60004–1-Ig), anti-phospho-TBK1 (rabbit, 1:1000, Cell Signaling Technology, #5483), anti-TBK1 (rabbit, 1:1000, Cell Signaling Technology, #3504), anti-β-actin (mouse, 1:5000, Proteintech, 66009–1-Ig), anti-H3K14ac (rabbit, 1:2000, Millipore, #07–353), anti-H3 (rabbit, 1:3000, Proteintech, 17168–1-AP). After wash, the membranes were incubated horseradish peroxidase (HRP)-conjugated secondary antibody (Cytiva, 1:5000). Immunoreactive bands were visualized using Western Chemiluminescent HRP Substrate (Millipore, #WBKLS0500) and analyzed with ImageJ.

Real-time qPCR

Total RNA was isolated from samples with TRIzol reagents (Invitrogen) and was reverse transcribed into cDNA using the HiScript III RT SuperMix for qPCR (+gDNA wiper) kits (R323–01, Vazyme). Relative quantitation was determined using the QuantStudio 6 Flex detection system (Applied Biosystems) that measures real-time SYBR green fluorescence and then calculated by means of the comparative Ct method (2^−ΔΔCt) with the expression of Gapdh or β-actin as an internal control. The sequences of primers used are provided in Table S3.

EdU staining

To measure newly synthesized mtDNA in primary microglia, in the presence of 10 μM EdU, the cells were treated with LPS (0.2 μg/mL) or oligomeric Aβ42 (1 μM) for 6 hours and then were incubated with MitoTracker (250 nM, Invitrogen) for 30 minutes. After cell fixation, permeabilization and blocking, EdU staining was performed according to the manufacturer’s protocol using a Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen). The nucleus was stained with DAPI for 5 minutes. Images were collected with a Zeiss LSM 900 confocal microscope and analyzed using ImageJ software (NIH).

Immunofluorescence

For cells immunofluorescence, primary microglia were washed with PBS and fixed with 4% PFA for 20 minutes at room temperature. After PBS wash, cells were permeabilized with 0.2% Triton X-100 and blocked with blocking buffer (2% donkey serum plus 1% BSA in PBS). Cells were incubated with primary antibodies overnight at 4°C. On the next day, cells were washed with PBS and incubated with secondary antibodies (1:100, Jackson ImmunoResearch) for 1 hour at room temperature.

For immunofluorescence of mouse brain cryosections, anesthetized mice were perfused transcardially with PBS, followed by 4% cold PFA in PBS. Brains were removed and fixed in 4% PFA at 4°C overnight. After dehydration by 30% sucrose, brains were embedded in OCT (Tissue-Tek) and cut into 30-μm-thick sections on cryostat microtome (Leica). Sections were permeabilized and blocked with 0.3% Triton X-100 and 5% donkey serum in PBS for 1 hour at room temperature, and incubated with primary antibodies at 4°C overnight. After washing three times with PBS, slices were incubated with secondary antibodies (1:100, Jackson ImmunoResearch) for 2 hours at room temperature.

For immunofluorescence of formalin-fixed paraffin-embedded (FFPE) human patient tissues, the brain tissue sections were deparaffinized in a 60°C oven for 2 hours, followed by xylene washes twice, each for 10 minutes at room temperature. Tissues were then rehydrated in a graded series of ethanol washes. Slides were rinsed with deionized water twice and transferred into sodium citrate buffer (10 mM, adjust pH to 6.0) for antigen retrieval at 120 °C for 20 minutes. After cooling down to room temperature, sections were washed with PBS and permeabilized with 0.3% Triton X-100 in PBS for 20 minutes at room temperature, then blocked with blocking buffer (20% donkey serum plus 1%BSA in PBS) for 1 hour at room temperature, and incubated with primary antibodies at 4°C overnight. On the next day, samples were incubated with secondary antibodies (1:100, Jackson ImmunoResearch) for 2 hours at room temperature. After washing three times with PBS, tissues were incubated with 0.1% Sudan Black B in 70% ethanol for 30 minutes to quench autofluorescence.

After samples were stained with DAPI and washed with PBS, samples were mounted using an aqueous mounting medium (Aqua-Poly/Mount, Polysciences). Images were obtained with Zeiss LSM900 confocal microscope and analyzed with ImageJ. Primary antibody concentrations were as follows: anti-Iba1 (goat, 1:200, Novus Biologicals, NB100–1028), anti-H3K14ac (rabbit, 1:100, Millipore, #07–353), anti-Aβ (mouse, 1:200, Biolegend, #803004), anti-phospho-TBK1 (rabbit, 1:100, Cell Signaling Technology, #5483), anti-phospho-IRF3 (rabbit, 1:100, Cell Signaling Technology, #4947), anti-GFAP (mouse, 1:400, Invitrogen, 14-9892-82), anti-NeuN (mouse, 1:200, Millipore, MAB377).

RNAscope in situ hybridization

Fixed brains were embedded in OCT (Tissue-Tek) and sectioned at a thickness of 14 μm. RNAscope Multiplex Fluorescent Reagent Kit v2 (ACD, Advanced Cell Diagnostics) was used following the manufacturer’s manual. Probe targeting mouse Kat7 (#1126701) or Jade2 (#1725601) was purchased from ACD. Images were collected with a Zeiss LSM 900 confocal microscope and analyzed using ImageJ software (NIH).

Thioflavin S (TS) staining

Aβ plaques were labeled by Thioflavin S staining on brain sections that were stained with 0.01% thioflavin S (T1892, Sigma) in 50% ethanol for 10 min. Then, sections were washed twice with 50% ethanol and three times with PBS. Brain sections were mounted for imaging and analyzed using a Zeiss LSM 900 confocal microscope.

Measurement of cytosolic mtDNA

Microglia were resuspended in digitonin buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, and 25 μg/ml digitonin) and incubated for 10 min at room temperature, followed by centrifugation at 2000g for 10 min at 4 °C. The supernatant containing cytosolic mtDNA (cmtDNA) was used for qPCR. The pellet was used for nuclear DNA extraction with QIAamp DNA Mini Kit (Qiagen) according to the manufacturer’s instructions. The cmtDNA in the supernatant was normalized to the nuclear DNA (B2m gene) in the pellet for each sample. D-loop and Nd4 were used to assess mtDNA expression. B2m and Tert was used to assess nuclear DNA expression. The sequences of primers used are provided in Table S4.

ChIP-qPCR

ChIP experiments were performed according to the procedure described previously⁶². BV2 cells or primary microglia were fixed with 1% formaldehyde for 15 min at room temperature. The fixed cells were lysed in lysis buffer (1% SDS, 5 mM EDTA, 50 mM Tris-HCl, pH 8.1) containing protease inhibitor cocktail. The lysates were then sonicated to generate chromatin fragments of ~500 bp in length. Cell debris was removed by centrifugation and supernatant were collected. A dilution buffer (150 mM NaCl, 2 mM EDTA, 1% Triton X-100, and 20 mM Tris-HCl, pH 8.1) containing protease inhibitor cocktail was subsequently applied (1:9 ratio) and the chromatin solution (40 μl aliquot as the input) was then incubated with specific antibodies (2 μg) at 4°C overnight with mild rotation. 30 μl Protein A magnetic beads (Invitrogen) were added for incubation of 2 hours. Beads were sequentially washed with the following buffers: TSE I (150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1% Triton X-100, 20 mM Tris-HCl, pH 8.1), TSE II (500 mM NaCl, 2 mM EDTA, 0.1% SDS, 1% Triton X-100, 20 mM Tris-HCl, pH 8.1), buffer III (0.25 M LiCl, 1% Nonidet P-40, 1 mM EDTA, 1% sodium deoxycholate, and 10 mM Tris-HCl, pH 8.1), and Tris-EDTA buffer. The input and the precipitated DNA-protein complex were de-crosslinked at 65°C for 12 hours in elution buffer (1% SDS, 0.1 M NaHCO₃) with RNase A and Proteinase K. Then DNA was purified using QIAquick PCR Purification Kit (Qiagen). Quantification of the precipitated DNA fragments were performed with real-time PCR using primers listed in Table S5.

Flow cytometry (FACS) analysis

Single-cell suspensions from mouse brain were resuspended in cell staining buffer (BioLegend, 420201). For flow cytometric analysis of microglial cell death and proliferation, cells were stained with Live/Dead Fixable Aqua Dead Cell Stain (Invitrogen, L34957) in PBS according to the manufacturer’s instructions. For FACS sorting of microglia for single-cell RNA-seq, Vybrant DyeCycle violet (Invitrogen, R37172) was used to identify viable cells. Fc receptors were blocked by incubation with TrueStain FcX anti-mouse CD16/32 antibody (BioLegend, 156604) at 4°C for 20 min. Cells were then stained with fluorophore-conjugated antibodies against CD11b (PE/Dazzle 594, BioLegend, 101256) and CD45 (APC/Cy7, BioLegend, 103116) for 20 min at 4°C. For intracellular Ki67 detection, cells were fixed and permeabilized using the True-Nuclear Transcription Factor Buffer Set (BioLegend, 424401), followed by staining with Ki67 antibody (AF647, BioLegend, 151206) according to manufacturer’s instructions. Data were acquired on a CytoFLEX LX (Beckman Coulter) and analyzed using FlowJo (BD).

RNA sequencing (RNA-seq)

Three biological replicates were sequenced per group. For each sample, RNA was extracted from BV2 cells with TRIzol reagents (Invitrogen). High-throughput RNA sequencing (RNA-seq) was performed by Illumina NovaSeq 6000 at Novogene (CA, USA). The raw sequencing data were aligned to the mouse preference genome (GRCm39, mm39) using HISAT2 (v2.0.5). Reads on each GENCODE annotated gene were counted using HTSeq, and then differential gene expression analysis was performed using DESeq2 R package. GO pathway analysis was conducted with DAVID tools (https://davidbioinformatics.nih.gov/). Differentially expressed genes were listed in Table S6.

Single-cell RNA-seq (scRNA-seq)

FACS purified live microglia (CD11b⁺/CD45^low) were sequenced for each group at the Single Cell and Transcriptomics Core of Johns Hopkins University. scRNA-seq libraries were constructed using the 10x Genomics Chromium GEM-X Universal 3’ Gene Expression v4 4-plex On-Chip Multiplexing (OCM) kit (CG000768, Rev A) according to the manufacturer’s instructions. Libraries were sequenced on an Illumina NovaSeq X platform to a depth of ~20,000 reads per cell, targeting ~10,000 cells per sample. Raw sequencing data were processed with Cell Ranger (v10.0.0) against the mm39 reference. scRNA-seq analyses were conducted in R (v4.3.1) using Seurat (v5.4.0). Filtered gene-barcode count matrices from four experimental groups were generated by Cell Ranger and then processed and integrated in Seurat. Cells expressing fewer than 200 genes and genes detected in fewer than 3 cells were excluded; additional filtering removed cells with detected genes <200 or >6,000, total UMI counts <500 or >50,000, or mitochondrial reads ⩾ 10%. Data were normalized with LogNormalize; 2,000 highly variable genes were selected via variance-stabilizing transformation (vst); and datasets were integrated by canonical correlation analysis (CCA). Principal component analysis (PCA) was computed (50 PCs), and the top 20 PCs were used to construct a shared nearest neighbor (SNN) graph for Louvain clustering; visualizations were generated with Uniform Manifold Approximation and Projection (UMAP). Microglia were identified by canonical markers Cx3cr1, P2ry12, Tmem119, and Hexb; non-microglia clusters were removed and UMAP was recomputed on the microglia subset. Differential expression was assessed using the Wilcoxon rank-sum test with thresholds of |log₂FC| > 1, adjusted p < 0.05, and expression in >0.5% of cells in both groups. Dot plots were generated in R using ggplot2 (v4.0.2) from per-condition summaries computed with dplyr (v1.2.0). Volcano plots were generated with ggplot2 (v4.0.2) from Seurat FindMarkers outputs, with selected genes labeled using ggrepel (v0.9.6).

Cleavage Under Targets & Tagmentation (CUT&Tag)

CUT&Tag was performed with Hyperactive Universal CUT&Tag Assay Kit for for Illumina Pro (TD904, Vazyme) according to the manufacturer’s instructions. In brief, BV2 cells were collected and counted same number for each group. Nuclei were isolated from BV2 cells and bounded to Concanavalin A (ConA)-coated beads. Subsequently, Nuclei were resuspended in antibody buffer and incubated with primary antibodies against H3K14ac (rabbit, Cell Signaling Technology, #7627), H3K4me3 (rabbit, Cell Signaling Technology, #9751) and IgG (rabbit, Cell Signaling Technology, #66362) at 4 °C overnight. On the next day, samples were incubated with goat anti-rabbit secondary antibodies (1:50, Cell Signaling Technology, #35401). The samples were incubated with pA/G-Tn5 transposase. After transposon activation and fragmentation, 0.5 pg Spike-in DNA was added to each sample and total DNA was isolated, amplified, and purified to construct library. The library for sequencing was constructed with TruePrep Index Kit V2 for Illumina (TD202, Vazyme) and VAHTS DNA Clean Beads (N411, Vazyme) were used for purification steps. The library was sequenced on an Illumina NovaSeq (PE 150) at Novogene. Raw sequencing reads were trimmed using Cutadapt 5.0 (https://cutadapt.readthedocs.io/en/stable/). Trimmed reads were then aligned to the mouse reference genome mm10 with the Spike-in sequence using Bowtie2 (version 2.3.5.1). Bam files with low-quality reads were filtered and duplicates were removed using Samtools v1.18. Reads were then normalized to Spike-in using Bedtools v2.31.0. Peaks were then called with SEACR v1.3. Differential peak analysis was analyzed by MAnorm2 and annotated by CHIPseeker with a p<0.05 cutoff.

Intracerebroventricular (ICV) injection

5-month-old 5×FAD mice were used for ICV injection of WM-3835 (S9805, Selleck). Briefly, 20 mice were randomly separated into two groups (WM-3835 and vehicle, 10 mice per group), then deeply anesthetized with isoflurane and immobilized using a stereotactic device. To implant osmotic pumps in the mice, osmotic pumps (1004W for 4 weeks infusion, RWD) matched with Brain infusion kit (Bic-3, RWD) were loaded according to the manufacturer’s instructions. 100 μL of vehicle (5% DMSO, 40% PEG300, and 55% saline) or WM-3835 (1 mM) was filled in the osmotic pump, and the Bic-3 kit/tubing (2 cm) was backfilled before the two parts were connected. In addition to the incision on the scalp, the pocket for the osmotic pump was obtained by stretching the space between the skin and the muscle in the back with sterile forceps. The detachable top part of the infusion cannula was held with a holder. A 0.5–1 mm burr hole was drilled in the skull, and the cannula tip was gently implanted into the lateral ventricle (coordinates, bregma: anterior/posterior, −0.5 mm; medial/lateral, 1.0 mm; and dorsal/ventral: −2.3 mm). The osmotic pump was slowly positioned in the pocket under the back skin simultaneously. The position of the cannula was secured with instant adhesives, and the skin was sutured with suture thread. ICV infusion was performed for four weeks. Then, the mice were sacrificed and processed for pathology analyses.

Acute brain slice electrophysiology

Mice were anesthetized with isoflurane and intracardially perfused with ice-cold oxygenated cutting solution containing (in mM): 110 choline chloride, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 0.5 CaCl₂, 7 MgCl₂, 10 glucose, bubbled with 95% O₂ / 5% CO₂. The brain was removed rapidly and immersed in ice-cold oxygenated cutting solution. Transverse hippocampal slices (350 μm) were cut in the cutting solution using a vibratome (VT-1200S, Leica) and transferred to artificial cerebrospinal fluid (aCSF) containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 2 CaCl₂, 2 MgCl₂, 10 glucose, saturated with 95% O₂ / 5% CO₂. The slices were recovered for 20 min at 35 °C and then maintained at room temperature for 1 hour. Slices were subsequently transferred to a submerged recording chamber containing aCSF solution maintained at 34 °C. Picrotoxin (100 μM) was added to block inhibitory transmission. mEPSCs were recorded at a holding potential of −70 mV in the presence of 1 μM tetrodotoxin (TTX). fEPSPs were evoked in the CA1 stratum radiatum by stimulating the Schaffer collateral with a concentric bipolar electrode and recorded with a glass pipette (1–3 MΩ) filled with aCSF. The stimulus intensity was adjusted to evoke 40%−50% of the maximal response. LTP was induced by theta burst stimulation (TBS) consisting of two trains of 5 bursts at 5 Hz, and each burst contained 4 pulses at 100 Hz. Recordings were made with MultiClamp 700B amplifier (Molecular Devices) and data acquisition was performed with pClamp 10.7 software (Molecular Devices).

Morris water maze test

Morris water maze tests were performed at Behavioral Core of Johns Hopkins University. In brief, we used a maze consisted of a round pool (diameter, 120 cm) filled with water that was at 24 °C and made opaque with nontoxic white paint. A circular plastic platform (diameter, 10 cm) was placed at the center of the target quadrant and submerged 1 cm below the surface of the water. Four local cues were provided to allow spatial map generation. In brief, we trained the mice for four trials per day with different start points for five consecutive days. Mice were gently placed into the water facing the wall of the pool and allowed to freely explore the whole maze for 1 min. Mice were then guided to the rescue platform if they did not find it. Mice were allowed to take a rest on the platform for 10 s and then retrained from a different start position with the same procedure. The latency for each animal to find the platform (at least 3 s stay) was recorded. On day 6, the platform was removed, and animals searched freely for 1 min starting from the opposite quadrant. The entries into the platform area, total time spent in the target quadrant, and the total distance travels were recorded using the ANY-maze software.

Quantification and statistical analysis

All data are presented as mean ± SEM. Statistical analyses were performed using GraphPad Prism 9 (GraphPad Software, CA). Prior to analysis, normality of the data was assessed using the Shapiro-Wilk test. Comparisons between two groups were performed using unpaired two-tailed Student’s t test or the Mann-Whitney test. Comparisons among three or more groups were made using one- or two-way ANOVA followed by Bonferroni’s post hoc multiple comparison test. A significance threshold of p < 0.05 was used. Details of statistical tests, sample sizes (n), and p values are provided in the corresponding figure legends.

Supplementary Material

2

Table S6. Differentially expressed genes from RNA-seq in BV2 cells, related to Figure 3 and S4.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-KAT7	Cell Signaling Technology	Cat #58418; RRID: AB_2799547
Rabbit anti-JADE2	Proteintech	Cat #11513-1-AP; RRID: AB_2877772
Rabbit anti-iNOS	GeneTex	Cat #GTX130246; RRID: AB_2886221
Rabbit anti-p65	Cell Signaling Technology	Cat #8242; RRID: AB_10859369
Rabbit anti-p-p65 (Ser536)	Thermo Fisher Scientific	Cat #MA5-15160; RRID: AB_10983078
Rabbit anti-TBK1	Cell Signaling Technology	Cat #3504; RRID: AB_2255663
Rabbit anti-p-TBK1 (Ser172)	Cell Signaling Technology	Cat #5483; RRID: AB_10693472
Rabbit anti-p-IRF3 (Ser396)	Cell Signaling Technology	Cat #4947; RRID: AB_823547
Rabbit anti-Iba1	Wako	Cat #019-19741; RRID: AB_839504
Goat anti-Iba1	Novus Biologicals	Cat #NB100-1028; RRID: AB_3148646
Rabbit anti-CMPK2	Novus Biologicals	Cat #NBP1-80653; RRID: AB_11037163
Rabbit anti-H3K14ac	Cell Signaling Technology	Cat #7627; RRID: AB_10839410
Rabbit anti-H3K14ac	Millipore	Cat #07-353; RRID: AB_310545
Rabbit anti-H3K23ac	Millipore	Cat #07-355; RRID: AB_310546
Rabbit anti-H3K4me3	Cell Signaling Technology	Cat #9751; RRID: AB_2616028
Rabbit anti-H3	Proteintech	Cat #17168-1-AP; RRID: AB_2716755
Rabbit anti-IgG	Cell Signaling Technology	Cat #66362; RRID: AB_2924329
Mouse anti-GAPDH	Proteintech	Cat #60004-1-Ig; RRID: AB_2107436
Mouse anti-β-actin	Proteintech	Cat #66009-1-Ig; RRID: AB_2782959
Mouse anti-Aβ	Biolegend	Cat #803004; RRID: AB_2715854
Mouse anti-GFAP	Thermo Fisher Scientific	Cat #14-9892-82; RRID: AB_10598206
Mouse anti-NeuN	Millipore	Cat #MAB377; RRID: AB_2298772
HRP linked Donkey anti-Rabbit IgG	Cytiva	Cat #NA934; RRID: AB_772206
HRP linked Sheep anti-Mouse IgG	Cytiva	Cat #NA931; RRID: AB_772210
Alexa Fluor 488 Donkey anti-Mouse IgG	Jackson ImmunoResearch	Cat #715-545-150; RRID: AB_2340846
Rhodamine Donkey anti-Rabbit IgG	Jackson ImmunoResearch	Cat #711-025-152; RRID: AB_2340588
Alexa Fluor 647 Donkey anti-Goat IgG	Jackson ImmunoResearch	Cat #705-605-147; RRID: AB_2340437
Biological samples
Mouse brain tissue	This paper	N/A
Human brain samples	Johns Hopkins Brain Resource Center, see Table S1 for details	N/A
Chemicals, peptides, and recombinant proteins
Thioflavin S	Sigma-Aldrich	Cat #T1892
Beta-Amyloid (1-42), HFIP	rPeptide	Cat #A-1163
Tamoxifen	Selleck	Cat #S1238
WM-3835	Selleck	Cat #S9805
LPS	Sigma-Aldrich	Cat #L2630
Collagenase Type IV	Sigma-Aldrich	Cat #C4-BIOC
DNase I	Sigma-Aldrich	Cat #DN25
Picrotoxin	Sigma-Aldrich	Cat #P1675
TTX	Tocris	Cat #1069
FBS	Avantor	Cat #97068-085
Percoll	Cytiva	Cat #17089101
Matrigel	Corning	Cat #354230
Y-27632	StemCell Technologies	Cat #72304
Doxycycline	Sigma-Aldrich	Cat #D9891
Critical commercial assays
RNAscope Multiplex Fluorescent Reagent Kit v.2	Advanced Cell Diagnostics	Cat #323100
TSA Plus Fluorescein Evaluation Kit	PerkinElmer	Cat #NEL741
Click-iT EdU Alexa Fluor 488 Imaging Kit	Invitrogen	Cat #C10337
HiScript III RT SuperMix for qPCR Kit	Vazyme	Cat #R323-01
Pierce BCA Protein Assay Kit	Thermo	Cat #23227
ZymoPURE II Plasmid Midiprep kit	Zymo Research	Cat #D4201
Mouse IL-6 ELISA Kits	BioLegend	Cat #431301
Mouse IL-1β ELISA Kits	BioLegend	Cat #432601
Human IL-6 ELISA Kits	BioLegend	Cat #430501
CUT&Tag Assay Kit	Vazyme	Cat #TD904
TruePrep Index Kit V2 for Illumina	Vazyme	Cat #TD202
CD11b MicroBeads	Miltenyi Biotec	Cat #130-093-634
Osmotic pumps	RWD	Cat #1004W
Brain infusion kit	RWD	Cat #Bic-3
Deposited data
RNA-seq data from BV2 cells	This paper	GEO: GSE327604
CUT&Tag data from BV2 cells	This paper	GEO: GSE327608
scRNA-seq data from FACS-	This paper	GEO: GSE327611
puried microglia
Experimental models: Cell lines
HEK293T	ATCC	CRL-3216; RRID: CVCL_0063
BV2	A gift from Tony Wyss-Coray58	N/A
Human CLYBL 6-TF-iMG iPSC	Jackson Laboratory	Cat #JIPSC002072
Experimental models: Organisms/strains
Mouse: Kat7f/f	This paper	N/A
Mouse: C57BL/6J	Jackson Laboratory	JAX: 000664; RRID: IMSR_JAX:000664
Mouse: 5xFAD mice	Jackson Laboratory	JAX: 034848; RRID:MMRRC_034848-JAX
Mouse: Cx3cr1-CreER	Jackson Laboratory	JAX: 021160; RRID: IMSR_JAX:021160
Mouse: Camk2a-Cre	Jackson Laboratory	JAX: 005359; RRID: IMSR_JAX:005359
Oligonucleotides
Primers for mouse genotyping, see Table S2	This paper	N/A
Primers for qPCR, see Table S3	This paper	N/A
Primers for mtDNA qPCR, see Table S4	This paper	N/A
Primers for qChIP, see Table S5	This paper	N/A
Recombinant DNA
pLentiCRISPR-v2-Puro vector	Sanjana et al.59	RRID: Addgene_52961
pLentiCRISPR-Kat7-sgRNA	This paper	N/A
pLenti-EF1a-P2A-Puro vector	Origene	Cat #PS100142
pLenti-EF1a-KAT7	This paper	N/A
pLenti-EF1a-KAT7 (E508Q)	This paper	N/A
pLenti-EF1a-JADE2	This paper	N/A
pLenti-EF1a-CMPK2	This paper	N/A
pLenti-EF1a-CMPK2 (D330A)	This paper	N/A
Software and algorithms
BioRender	BioRender	https://www.biorender.com/
GraphPad Prism version 9.0	GraphPad Software	RRID: SCR_002798
ImageJ	NIH	RRID: SCR_003070
Adobe Illustrator	Adobe	RRID: SCR_010279
pCLAMP 10.7	Molecular Devices	RRID:SCR_011323
RStudio	Posit Software	RRID: SCR_000432
FlowJo	BD Biosciences	https://www.flowjo.com
ZEN	ZEISS	RRID: SCR_013672
Other
mTeSR Plus Medium	StemCell Technologies	Cat #100-0276
Advanced DMEM/F12	Gibco	Cat #12634-010
F-12 Medium	Gibco	Cat #11765-054
Poly-D-Lysine	Gibco	Cat #A38904-01
Lenti-X Concentrator	Takara Bio	Cat #631232

Highlights:

KAT7 links histone acetylation to mitochondrial immune activation in microglia

KAT7 drives CMPK2 expression to promote mtDNA synthesis and release

KAT7 deletion or pharmacological inhibition reduces neuroinflammation and Aβ load

Targeting KAT7 restores synaptic plasticity and cognitive function in AD mice

RESOURCE AVAILABILITY

Lead contact:

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Zhaozhu Qiu (zhaozhu@jhmi.edu).

Materials availability:

All unique/stable reagents and cell lines generated in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability:

All the raw sequencing data have been deposited at the Gene Expression Omnibus (GEO) with the accession number: RNA-seq (GSE327604), CUT&Tag (GSE327608) and scRNA-seq (GSE327611). All analyses were performed using publicly available software packages, as detailed in the Methods section. Additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.