Epigenetic Rewiring Connects Aging to Alzheimer’s Pathology in the Human Prefrontal Cortex

Epigenetic Rewiring Connects Aging to Alzheimer’s Pathology in the Human Prefrontal Cortex | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Epigenetic Rewiring Connects Aging to Alzheimer’s Pathology in the Human Prefrontal Cortex Jing Zhang, Dan Zhou, Wenli Zhai, Wenwei Fang, Yuan Zhou, Zuyun Liu, and 14 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7381745/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract The epigenetic mechanisms that connect aging to neurodegeneration remain incompletely understood, particularly in non-European populations. To address this gap, we constructed a comprehensive DNA methylation atlas of the human prefrontal cortex (PFC) from 1,057 postmortem brain samples (ages 1-106) as part of the China Brain Multi-Omics Atlas Project (CBMAP). This represents the largest DNA methylation dataset of East Asian brain tissue to date and provides a critical resource for understanding aging and Alzheimer’s disease (AD) in underrepresented populations. We found that over one-third of CpG sites exhibit significant age-associated methylation changes, with high concordance between Asian and European ancestries, but marked divergence between brain and peripheral blood methylation profiles. In the PFC, age-related CpGs were enriched near histone genes and associated with transcriptomic signatures of neuronal decline. Importantly, we identified 485 CpGs linked to AD neuropathologic change (ADNC), including 228 novel loci enriched in immune regulation, vesicle trafficking, and extracellular matrix remodeling—hallmark pathways of AD. Mediation analysis revealed that DNA methylation accounts for 27.3% of aging’s effect on ADNC, highlighting a key mechanistic link. Notably, CpG site cg09221482 mediates the relationship between aging and neurofibrillary tangle severity via MAP3K4 expression. Together, these findings uncover novel aging-associated epigenetic signatures that contribute to AD pathology and establish DNA methylation as a critical intermediary bridging aging and neurodegeneration, particularly in East Asian populations. Biological sciences/Genetics/Epigenomics Health sciences/Diseases/Neurological disorders/Neurodegenerative diseases/Alzheimer's disease Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Aging remains the strongest risk factor for Alzheimer’s disease (AD) and most neurodegenerative disorders, yet the biological mechanisms that link aging to AD pathology are still being uncovered 1,2 . Among candidate mechanisms, DNA methylation—an epigenetic modification known to change with age— has emerged as a potential contributor to age-related physiological decline and AD pathogenesis 3,4 . However, our understanding of how age-related DNA methylation in the human brain contributes to AD neuropathology remains limited. Previous studies have laid the foundation for epigenetic aging research 4,5 , leading to the development of age-predictive “epigenetic clocks” such as the blood-based Hannum Clock 6 , the multi-tissue Horvath Clock 3 , and a cortex-specific clock developed by Shireby and colleageus 7 . While informative, these efforts have not sufficiently explored the functional relevance of brain-specific methylation changes, nor have they adequately represented non-European populations. Additionally, most existing studies rely heavily on peripheral blood data or focus on narrow age ranges within postmortem brain donors, limiting their capacity to fully characterize brain aging and its relationship to AD across the lifespan. Recent integrative analyses 8 , 9,10 , 11 , such as those using data from Religious Orders Study and Memory and Aging Project (ROSMAP) have identified methylation loci associated with tau pathology, but these studies fall short in several key areas: they primarily focus on neurofibrillary tangle (tau) burden with limited analysis of amyloid-β (Aβ) pathology or comprehensive AD neuropathological changes (ADNC); functional insights are often limited by a lack of comprehensive integration with transcriptomic data; the narrow age range of subjects (mostly ≥70 years) restricts the ability to capture aging trajectories; and the overwhelming focus on individuals of European ancestry leaves a critical gap in understanding epigenetic aging in other populations. To address these limitations, we leveraged the China Brain Multi-Omics Atlas Project (CBMAP) 12 , profiling DNA methylation in the prefrontal cortex (PFC) of 1,057 individuals aged 1-106 years across multiple Chinese brain banks. This dataset—the largest of its kind in an East Asian population—also includes matched transcriptomic data and detailed neuropathological assessments, enabling a comprehensive interrogation of how age-related methylation patterns influence AD-related changes in the human brain. In this study, we aim to: (1) define age-associated DNA methylation signatures in the PFC across the lifespan; (2) identify CpG sites associated with ADNC, including both amyloid and tau pathology; (3) explore the functional relevance of these sites through integration with transcriptome; and (4) quantify the extent to which DNA methylation mediates the impact of aging on AD pathology. By filling a major gap in brain epigenomics for East Asian populations, our work provides novel insights into the epigenetic bridge between aging and AD. Results Sample Characteristics and DNA Methylation Profiling As part of the CBMAP, we profiled DNA methylation on PFC samples from 1,057 donors across eastern, northern, and central China. Of these, 401 (37.9%) were female. The donor ages ranged from 1 to 106 years, with a mean age of 74.9 years. Regional and age distributions are presented in Fig. 1a,b . Notably, compared to ROSMAP, CBMAP exhibits a markedly broader age distribution for the donors with methylation measurements. CBMAP samples underwent standardized neuropathological assessments, including evaluations for Aβ, Braak NFT stage, and CERAD neuritic plaque score. Following the NIA-AA criteria 13 and exclusion criteria (Methods), samples were classified into four ADNC categories: Control (n = 482), Low (n = 281), Intermediate (n = 205), and High (n = 30) ( Fig. 1c ). To examine how methylation contributes to AD pathogenesis, ADNC was treated as the primary outcome, while related pathological features, including Aβ deposits (represented by A score) and Braak NFT stage, were considered as secondary outcomes. DNA methylation was profiled using the Illumina HumanMethylationEPIC v2.0 array, targeting 930,140 CpG sites. After quality control and filtering (Methods), 907,506 CpG sites from 1,031 samples passed quality checks and were included in subsequent analyses ( Extended Data Fig. 1 ). We performed principal component analysis (PCA) on methylation data to examine potential variations associated with age and regions ( Extended Data Fig. 2a,b ). Although the PCA revealed no significant regional differences in global methylation patterns across the three studied regions, we identified 21,355 differentially methylated CpG sites between northern and southern China. Among the top 1,000 differentially methylated CpG sites between north and south ( Extended Data Fig. 2c ), significant pathway enrichments were observed in KEGG including “B cell receptor signaling pathway” ( P -value = 3.74×10 -6 ) and “Viral protein interaction with cytokine and cytokine receptor” when excluding the MHC region ( P -value = 5.95×10 -6 , Extended Data Fig. 2d,e , Supplementary Table 1 ). The broad age range and comprehensive pathological assessments make the CBMAP dataset a valuable resource for identifying methylation features associated with aging and early neuropathological changes related to ADNC. Epigenetic landscape of aging in the PFC To identify age-associated CpG sites, we fitted linear models adjusting for covariates including sex, estimated cell type proportions, and batch. As shown in Fig. 2a , over one-third of CpG sites (n = 402,503) exhibited significant age-related methylation levels (FDR < 0.05). We characterized all the age-related CpGs by gene set enrichment analysis (GSEA) to manifest the potentially regulated biological process or pathway. By integrating the gene expression from the same tissue samples, we identified 1,349 significant gene sets (FDR < 0.05) which led by gene sets featured by microglia-specifically expressed genes ( Supplementary Table 2 ). We also observed the downregulation of AD-related terms including extracellular matrix (NABA MATRISOME, P -value = 3.73×10 -15 ) and secretory granule (GOCC SECRETORY GRANULE, P -value = 1.63×10 -13 ). To investigate whether the effect of age on methylation varies across tissues and ancestries, we conducted a comparative analysis of age-associated methylation across three datasets: CBMAP (East Asian [EAS], PFC), ROSMAP (primarily European ancestries [EUR], PFC), and Hannum (primarily EUR, PBMC) 6 . First, we performed pairwise comparisons of effect sizes for age-CpG association among the three datasets. . As shown in Fig. 2b , we observed that the correlation of effect sizes between the two PFC datasets (CBMAP and ROSMAP) was high (r = 0.89), while the correlation between PFC (ROSMAP) and PBMC (Hannum) was relatively lower (r = 0.71). Further stratified analysis showed that CpG sites located in regulatory regions such as the 1stExon and promoter regions (TSS1500, TSS200, and 5'UTR) exhibited stronger effect size correlations across datasets, with the highest correlation observed in the 1stExon region (r = 0.91) ( Fig. 2c ). Next, we characterized the top 1,000 age-associated CpG sites in each dataset. As illustrated in the Venn diagram ( Fig. 2d ), CBMAP and ROSMAP, both PFC datasets, shared 410 of their top 1,000 CpG sites, representing the highest degree of overlap ( Supplementary Table 3 ). We then performed enrichment analysis for these 410 CpG sites. The potential target genes of these sites were enriched in numerous immune-related Gene Ontology (GO) terms and KEGG pathways ( Extended Data Fig. 3a,b ). After excluding the major histocompatibility complex (MHC) region, significant enrichment was observed in GO terms including structural constituent of chromatin ( P -value = 1.57×10 -27 ) and nucleosome ( P -value = 2.01×10 -22 ) ( Supplementary Table 4 ). We also identified CBMAP-specific age-associated CpG sites, including 27 sites among the top 1,000 in CBMAP that did not reach statistical significance (FDR > 0.05) in ROSMAP ( Supplementary Table 5 ). Notable examples included cg06144905 (located in PIPOX ; CBMAP P -value = 2.01×10 -86 , ROSMAP P -value = 0.36) and cg09580336 (located in ATP1A1 ; CBMAP P -value = 6.95×10 -81 , ROSMAP P -value = 0.17), suggesting ancestry-specific heterogeneity in age-related methylation. Of these 410 CpG sites, 186 did not reach statistical significance (FDR > 0.05) in Hannum (EUR PBMC) ( Supplementary Table 6 ), suggesting the brain-specificity of age-related DNA methylation. The enrichment analysis results showed that the potential regulatory genes of these 186 brain-specific CpGs were more significantly enriched in GO terms including structural constituent of chromatin ( P -value = 5.07×10 -37 ) ( Extended Data Fig. 3c,d , Supplementary Table 7 ). In summary, we identified a large amount of age-related CpGs which enriched in AD-relevant terms and uncovered the differences in age-related DNA methylation between blood and brain tissues, while showed concordance between EAS and EUR brains. Methylation-Mediated Effects of Aging on Transcriptome Regulation To explore the potential regulatory genes and biological functions of age-associated CpG sites, we focused on the 1,000 CpG sites most strongly associated with age ( Fig. 3a ). The methylation levels of these sites showed strong age associations, with individual site-specific linear models yielding coefficient of determination (R 2 ) ranging from 0.713 to 0.931 ( Extended Data Fig. 4 ). Enrichment analysis revealed that the potential target genes of these CpG sites were significantly enriched in GO terms related to structural constituent of chromatin ( P = 5.39×10 -18 ) and nucleosome ( P -value = 5.28×10 -16 ) ( Fig. 3b , Supplementary Table 8 ). These terms encompassed genes encoding histones, including H2AX , H2BC12, H2BC14, H2AC15, H2BC3, H3C7, H2AC17 and H3C10 . Mediation analysis integrating PFC transcriptomic data demonstrated that most of these CpG sites mediated the effect of age on the expression of histone-encoding genes ( Supplementary Table 9 and 10 ). For example, the relationship among age, cg03459567, and H2BC12 ( Fig. 3c ) showed that age was positively correlated with cg03459567 methylation ( P -value = 1.92×10 -113 ), cg03459567 methylation was negatively correlated with H2BC12 expression ( P -value = 1.53×10 -4 ), and age was negatively correlated with H2BC12 expression ( P -value = 0.001), with a significant mediating effect ( P -value = 2.65×10 -4 ). Similarly, nine other CpG sites followed a pattern where increased methylation with age led to reduced expression of histone-encoding genes ( Fig. 3d ). These findings align with prior studies of declining histone expression with age 14,15 , while our study further demonstrates that, in the PFC, this pattern is largely mediated by DNA methylation. Such methylation-driven reductions in histone expression may impair chromatin stability, promote chromatin remodeling, and contribute to transcriptional dysregulation. Beyond histone genes, we investigated the 186 age-associated CpG sites specifically in the PFC but not in PBMC. Mediation analysis revealed that cg20591728 mediated the negative association between age and NXPH3 expression ( P -value = 0.001) in PFC. Notably, NXPH3 was found to be associated with ADNC ( β = -0.044; P -value = 0.016) in CBMAP which is consistent with previous study that a lower NXPH3 expression is associated with dysregulated synaptic plasticity and impaired neurotransmitter release 16 . Additionally, cg26092675, an age-associated CpG site specific to the PFC, mediated the age-related increase in BTN3A2 expression ( P -value = 4.18×10 -4 ). BTN3A2 has been identified as a potential AD risk gene, with functional studies showing that its overexpression suppresses excitatory synaptic transmission in CA1 pyramidal neurons, potentially through molecular interactions with presynaptic neurexins 17 , differential gene expression analysis in CBMAP further supports this finding, showing a positive association between BTN3A2 and ADNC ( β = 0.125; P -value = 0.020). We also observed that cg26092675 mediated the positive association between age and HFE ( P -value = 0.027). HFE, a human homeostatic iron regulator protein, studies have shown that higher HFE expression disrupts iron/cholesterol homeostasis, causing neuronal iron overload, oxidative stress, and synaptic/myelin damage, which synergistically accelerate Aβ/tau pathology and neuroinflammation in AD 18 . A positive association was observed between HFE and ADNC ( β = 0.107; P -value = 0.001) ( Fig. 3e , Supplementary Table 11 ). We also explored CpG sites exhibiting non-linear age-related methylation trends and grouped the CpGs in four patterns ( Fig. 3f ). For instance, our analysis revealed an age-dependent association, with cg13084525 methylation showed a significant negative association with age exclusively in individuals aged over 50 years. Mediation analysis indicated that cg13084525 methylation significantly mediated the positive association between age and CD44 expression levels ( P -value = 9.15×10 -8 ) ( Fig. 3g ). CD44 plays a critical role in glial cells, contributing to neural development, injury repair, and immune responses under pathological conditions by mediating cell-matrix adhesion, migration, and inflammatory signaling 19,20 . To summarize, our findings demonstrate that age-associated CpG sites mediate the effects of aging on gene expression in the PFC, affecting both histone-encoding genes and genes typically implicated in neurodegenerative diseases. The Epigenetic Clock in the PFC Biological age often diverges from chronological age, with DNA methylation age (MethAge) serving as a widely used biomarker of biological aging. To investigate biological aging differences between PFC and PBMC, assess MethAge variations across ancestries, and evaluate the relevance of MethAge in cell type proportion and neurodegenerative diseases, we firstly constructed a CBMAP elastic-net clock model for the Chinese population. Using 501 samples from eastern and central China, we constructed an age prediction model via regularized regression with cross-validation to optimize hyperparameters (Methods). The predictive accuracy was evaluated in 530 samples from northern China, yielding a Pearson’s correlation coefficient (r) of 0.95 and a root mean square error (RMSE) of 4.48 ( Fig. 4a ). We also evaluated the Cortical Clock 7 —developed using European ancestry cortical samples—in the CBMAP northern China center, which showed r = 0.94 and RMSE = 6.81, comparable to the multi-tissue Horvath Clock 3 (r = 0.91, RMSE = 8.80). In contrast, a PBMC-based MethAge model derived from Chinese individuals 21 exhibited substantial bias in CBMAP brain tissue (r = 0.81, RMSE = 40.7), as did a European PBMC-based model (r = 0.91, RMSE = 26.23). Both PBMC models consistently underestimated age in older individuals, with greater deviations at higher chronological ages. Given the potential differences in MethAge model accuracy between ADNC cases and controls of CBMAP, we also constructed an ADNC-free MethAge model using samples without ADNC and tested it on samples with ADNC, achieving a correlation of r = 0.84 and an RMSE of 6.85. Using this model, we calculated age acceleration for each sample (Methods). We observed that higher age acceleration was significantly associated with more severe ADNC ( P -value = 0.03) and Braak NFT stage ( P -value = 0.01) ( Fig. 4b ), consistent with findings by Levine et al. using ROSMAP 22 . We further investigated changes in cell type proportions as a function of MethAge. Using a reference dataset of over 15,000 single cells from human frontal cortex 23,24 , we applied methylation-based deconvolution to estimate the proportions of seven major cell types (excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, microglia, oligodendrocytes, and oligodendrocyte precursor cells [OPCs]). The overall cell type distributions were similar between CBMAP and ROSMAP, except for excitatory neuron and astrocyte ( Fig. 4c ). We observed that the proportions of OPCs and oligodendrocytes increased with MethAge, while microglia, endothelial cell, and inhibitory neurons showed decreasing trends ( Fig. 4d,e ). These patterns were largely consistent with those observed in ROSMAP ( Extended Data Fig. 5 ). To account for the influence of ADNC status, we performed stratified analyses ( Fig. 4f ). Overall, the direction and significance of associations were largely consistent between control and ADNC groups. Notably, the effect sizes were generally larger in the ADNC case group, suggesting that ADNC pathology may amplify the impact of epigenetic aging on brain cellular composition, potentially reflecting disease-related shifts in cell-type dynamics with age. Considering MethAge and ADNC as the founder and primary outcome, respectively, we inferred conditional dependencies among MethAge, neurodegenerative conditions, and ADNC using a Bayesian network analysis framework (Methods). We hypothesized that MethAge influences ADNC risk by modulating three age-related neuropathologies (primary age-related tauopathy [PART], limbic-predominant age-related TDP-43 encephalopathy [LATE], and aging-related tau astrogliopathy [ARTAG]) and cerebrovascular pathologies (including atherosclerosis, arteriosclerosis, cerebral amyloid angiopathy, cerebral hemorrhage and cerebral infarction). Our analysis revealed significant associations between LATE and cerebral amyloid angiopathy (CAA) with ADNC, with consistent effects across MethAge strata ( Extended Data Fig. 6 ). This suggests that LATE and CAA may substantially increase ADNC risk in an age- and epigenetic-dependent manner. Notably, the effect of CAA on AD neuropathologic changes (ADNC) is unlikely to be disease-specific, as we have previously revealed 25 . Methylation Signatures of AD Neuropathologic Changes To characterize the methylation signatures of ADNC, we conducted an epigenome-wide association study (EWAS) in CBMAP samples (n = 1,031). ADNC severity was scored as 0 (Control), 1 (Low), 2 (Intermediate), and 3 (High). In addition to ADNC, we performed EWAS analyses for Braak NFT stage and Aβ deposits. Given nearly one million CpG sites on the array, multiple testing correction posed significant challenges, resulting in a limited number of sites passing stringent significance thresholds. To enhance statistical power, we performed a meta-analysis combining CBMAP results with those from an identical analysis in ROSMAP (n = 734). To account for potential heterogeneity between studies, we first assessed effect size differences using a heterogeneity test. For CpG sites without heterogeneity, a fixed-effects model was applied, while a random-effects model was used for sites with heterogeneity to identify CpG sites robustly associated with ADNC across both studies. The meta-analysis identified 204, 181, and 266 CpG sites significantly associated with ADNC, Braak NFT stage, and A-score, respectively (FDR < 0.05) ( Supplementary Table 12 ). These corresponded to 485 unique CpG sites, of which 257 were previously reported in a meta-analysis of four datasets with predominantly European ancestries 11 , and 228 were novel discoveries in this study. Among the 204 CpG sites associated with ADNC, 171 showed increased methylation with ADNC progression, while 33 exhibited decreased methylation ( Fig. 5a ). As shown in Extended Data Fig. 7a,b , the mapped genes of the 204 CpG were primarily enriched in (1) granule membrane and vesicle-related terms (cell periphery, plasma membrane, specific granule membrane, cytoplasmic vesicle), (2) oxidative phosphorylation (NADPH oxidase complex), and (3) immune regulation (positive regulation of CD8-positive, alpha-beta T cell differentiation, immune effector response) ( Supplementary Table 13 ), as well as MAPK and phagosome pathway of KEGG database ( Supplementary Table 14 ). The most significantly upregulated CpG site was cg07883124, with P -values of 2.23×10 -5 and 5.95×10 -8 in CBMAP and ROSMAP, respectively, and a meta-analysis P -value of 6.62×10 -12 ( Fig. 5b ). cg07883124 maps to MCF2L , which encodes a guanine nucleotide exchange factor that interacts with GTP-bound Rac1 and modulates Rho/Rac signaling pathways, implicated in cytoskeletal regulation and cell migration 26 . Among downregulated sites, cg00464927 was one of the most significant, with P -values of 3.71×10 -5 and 6.51×10 -2 in CBMAP and ROSMAP, respectively, and a meta-analysis P -value of 1.89×10 -5 ( Fig. 5b ). This CpG site showed a stronger effect in CBMAP and has not been previously associated with AD, suggesting potential ancestry-specificity. It maps to MRGPRF , encoding a G protein-coupled receptor primarily expressed in sensory neurons and immune cells 27 , with no reported role in neurodegenerative diseases. Next, we dug into the 181 CpG sites associated with Braak NFT stage ( Fig. 5c,d ). Their potential target genes were enriched in pathways related to antigen binding, immune response, and inflammation, including peptide antigen binding, immune system process, immune response, and lymphocyte activation ( Extended Data Fig. 7c,d , Supplementary Table 15 ). Among the 157 significantly upregulated CpG sites, cg07883124 (also associated with ADNC) was the most significant, with a meta-analysis P -value of 3.11×10 -10 . The second most significant site, cg23968456, located in the exonic region of VSIR , had P -values of 6.95×10 -4 and 1.87×10 -6 in CBMAP and ROSMAP, respectively, and a meta-analysis P -value of 1.80×10 -8 . Notably, its methylation level increased markedly in early Braak NFT stages ( Fig. 5d ). VSIR is highly expressed in microglia and in mouse inflammation models 28 . VSIR knockout was found to be associated with elevated pro-inflammatory cytokine levels from T cells and myeloid cells 29,30 . Among the 24 CpG sites negatively associated with Braak NFT stage, cg09470754, located in LAIR1 , was one of the most significant (meta P -value = 1.13×10 -6 ). Limited reports link LAIR1 to AD, but its expression has been implicated in microglial states, inflammation, and neuronal injury 31 . For the 266 CpG sites associated with Aβ deposits ( Fig. 5e,f ), their potential target genes were enriched in immune response and vesicle-related functions (endocytic vesicle, endocytic vesicle membrane, phagosome) ( Extended Data Fig. 7e,f , Supplementary Table 16 ). Among the 143 significantly upregulated sites, cg00601836 was prominent, with P -values of 5.68×10 -4 , 2.10×10 -5 , and 5.28×10 -8 in CBMAP, ROSMAP, and the meta-analysis, respectively. This site, unreported in prior AD studies, maps to ESR1 , encoding estrogen receptor α. Among sites negatively associated with Aβ deposits, cg14010720 ( SLC17A9 ) was one of the most significant, showing a linear decrease in methylation with increasing Aβ deposits in CBMAP ( P -value = 2.53×10 -4 ) and ROSMAP ( P -value = 5.85×10 -6 ). SLC17A9 encodes a transmembrane protein of the solute carrier family 17, mediating vesicular uptake, storage, and secretion of ATP and other nucleotides, and is highly enriched in lysosomes, contributing to cell viability and lysosomal function 32 . Sensitivity analyses adjusting for Lewy body disease (LBD) and cerebrovascular disease (CVD) confirmed that the associations of most highlighted CpG sites with ADNC remained significant ( Supplementary Table 17 ). In summary, our EWAS identified numerous novel methylation sites associated with AD neuropathological changes, with potential target genes enriched in phagocytosis-related processes (primarily linked to Aβ deposits), immune regulation (primarily linked to Braak NFT stage), and oxidative phosphorylation. These findings highlight the critical role of DNA methylation in AD pathology. Co-regulation of ADNC-Associated CpG Sites To investigate whether ADNC-associated CpG sites exhibit co-expression patterns and to identify potential regulatory pathways linked to ADNC-related functional modules, we performed weighted gene co-expression network analysis (WGCNA) 33 on 23,199 nominally significant ADNC-associated CpG sites. We identified 39 co-regulation modules, with each module containing 33 to 1,073 CpG sites ( Fig. 6a ). Using the first principal component of each module as a representative feature, regression analysis revealed 11 modules significantly associated with ADNC ( Fig. 6b ). Most of these modules were also associated with at least one of Braak NFT stage or Aβ deposits. Subsequent enrichment analysis of the hub CpG sites within these modules identified potential target gene groups ( Fig. 6c ). Module 1, associated with both Braak NFT stage and Aβ deposits, was significantly enriched in functions related to secretory granule membrane, dendritic cell migration, and dendritic cell chemotaxis. The top-ranked hub CpG in Module 1 mapped to the potential target gene ANK1 . Module 4, linked to Aβ deposits, was enriched in extracellular matrix (ECM)-related functions (cell periphery, extracellular matrix structural constituent) and potassium ion transmembrane transporter activity. Module 8 was linked to molecular transducer activity, signaling receptor activity, transmembrane signaling receptor activity, and immune effector process, with its potential hub regulator being FOXC2 . Module 18, associated with both ADNC and age, was significantly enriched in chromatin, nucleosome, and protein-DNA complex functions. Module 28 was mapped to neurotransmitter transmembrane transporter activity, GABA-A receptor activity, and GABA-gated chloride ion channel activity ( Supplementary Table 18 ). Overall, co-methylation analysis identified 11 ADNC-associated modules, encompassing functions related to granule membranes, immune regulation, nucleosome activity, dendritic cell migration, ECM, and neurotransmitter transport. These findings replicated the key functional categories identified in the ADNC enrichment analysis and revealed more specific modules underlying ADNC pathology. ADNC-Specific and ADNC-/Age-Shared CpG Sites To investigate the functional similarities and differences between CpG sites associated with age and ADNC, we performed KEGG and GO enrichment analyses ( Extended Data Fig. 8 ). The potential target genes of CpG sites associated with both ADNC (or Braak NFT stage/Aβ) and age were predominantly enriched in immune signaling pathways, such as antigen processing and presentation. In contrast, CpG sites associated exclusively with AD neuropathological changes (ADNC level, Braak NFT stage, or Aβ deposits) were enriched in functions related to the ECM, NADPH oxidase complex, vesicle, calcium ion binding, and the MAPK signaling pathway ( Supplementary Table 13-16 ). While these patterns outline broad functional tendencies, several CpG subsets exhibited distinct enrichment profiles. For instance, cell activation involved in immune responses was enriched among potential target genes of CpG sites associated solely with Braak NFT stage ( Extended Data Fig. 8e ). Additionally, CpG sites linked to both age and Aβ deposits were enriched in the GO term “regulation of lamellipodium assembly” ( Extended Data Fig. 8c ). Lamellipodia are critical precursor structures for dendritic spine formation during synaptic development. Studies in AD model animals have demonstrated impaired lamellipodium formation and reduced dendritic spine density 34 . These results suggest that CpG sites associated with both ADNC and age are primarily linked to immune regulation, whereas those exclusively associated with AD neuropathological changes are more frequently enriched in signaling pathways including ECM, oxidative phosphorylation, and vesicle-related processes. Pseudotime trajectories of ADNC-Associated Methylation Sites To characterize the age-related trajectories of ADNC-associated CpG sites, we used loess regression to fit pseudotime trajectories and performed unsupervised k-means clustering on these trajectories ( Fig. 7a,b ). Functional enrichment analysis was conducted for each of the clusters ( Fig. 7c ). Clusters 1 and 2 showed increasing methylation levels with age. Cluster 1 was enriched in cell adhesion, calcium ion binding, and phagocytosis, while Cluster 2 was enriched in TRAIL binding pathways, where TRAIL, a TNF family cytokine, induces programmed cell death 35 . Cluster 3 exhibited decreasing methylation with age and was enriched in granule membrane and signal transduction pathways. Cluster 4, associated with the MHC region, remained relatively stable with age ( Supplementary Table 19 ). Similar analyses for Braak NFT stage ( Fig. 7d-f ) and Aβ deposits ( Fig. 7g-i ) revealed that Cluster 2 for Braak NFT stage, with methylation levels positively correlated with age, was enriched in myeloid cell and immune cell differentiation functions, consistent with reports that bone marrow-derived cells may enter the brain and function similarly to microglia 36 ( Supplementary Table 20 ). For Aβ deposits, Cluster 1 showed methylation levels positively correlated with age, with the steepest increase around age 75, and was enriched in actin, TNFα, cell differentiation, and phagocytosis functions ( Supplementary Table 21 ). These results elucidate the age-dependent patterns of ADNC-associated methylation changes. Methylation Mediated the Effect of Aging on ADNC To assess whether methylation mediates the effect of age on ADNC, we performed mediation analysis on the 204 ADNC-associated CpG sites. Of these, 140 (68.6%) showed partial mediation of the age-ADNC relationship, with an average mediation proportion of 8.28% ( Fig. 8a ). The highest mediation was observed for cg19803550, which mediated 19.7% of age’s effect on ADNC ( Fig. 8b,c ). cg19803550 is located in WDR81 , encoding a multi-domain transmembrane protein 37 . The CpG site most strongly associated with ADNC, cg07883124, mediated 18.4% of age’s effect. To evaluate the cumulative mediation effect of these 140 sites, we computed the first 10 principal components (PCs) of their methylation levels (PCs are orthogonal, simplifying cumulative contribution assessment) and performed mediation analysis on each of the PCs. The mediation proportions for these 10 PCs ranged from -4.23% to 33.50%, with a cumulative mediation proportion of 27.27%. Interestingly, some PCs, including PC3, exhibited negative mediation proportions. Analysis of CpG sites with high loading scores in PC3 identified sites with negative mediation effects. For instance, the methylation level of cg19229215 was lower in elder donors ( β < 0, P -value 0, P -value = 1.2×10 -3 ), contributing a negative mediation proportion of -5.3%. This site was strongly associated with NCF1 expression ( β > 0, P -value = 3.1×10 -7 ), a component of NADPH oxidase linked to reactive oxygen species production and AD-related oxidative stress 38 , suggesting that high NCF1 expression may exacerbate neuronal damage. These results indicate that while methylation predominantly mediates age’s positive effect on ADNC, certain CpG sites like cg19229215 may counteract neurodegenerative changes with increasing age. For Braak NFT stage, 165 of 181 associated CpG sites (91.1%) exhibited significant mediation effects ( Fig. 8d-f ). The highest mediation was observed for cg09221482, which mediated 15.5% of age’s effect on Braak NFT stage and was associated with MAP3K4 expression ( P -value = 6.3×10 -4 ). MAP3K4 mediated 10.7% of the cg09221482-Braak NFT stage association, suggesting a pathway where increased cg09221482 methylation with age upregulates MAP3K4 expression, leading to higher Braak NFT stages ( P -value = 6.0×10 -5 ). Other sites, such as cg15821544 (intergenic) and cg05714396 (located in ATG10 , encoding an E2 enzyme involved in autophagosome formation 39 ), also showed mediation effects. The first 10 PCs of these 165 sites collectively mediated 18.33% of age’s effect on Braak NFT stage. For Aβ deposits, 77 of 266 associated CpG sites (28.9%) showed potential mediation effects, with the highest mediation proportions observed for cg19803550 ( WDR81 ), cg24647108 ( SUN1 ), and cg14010720 ( SLC17A9 ) ( Fig. 8g-i ). The first 10 PCs of these 77 sites collectively mediated 11.53% of age’s effect on Aβ deposits. In summary, a substantial proportion of ADNC-associated methylation sites, particularly those linked to Braak NFT stage, mediate age’s effect on ADNC. These findings underscore the critical role of methylation in modulating age-related ADNC risk. Discussion We constructed the first large-scale brain methylation atlas for the East Asian population, comprising over 1,000 PFC samples, addressing a critical gap in human brain epigenetic data in EAS. Our study elucidates the landscape of age-related methylation sites in the PFC, highlighting both similarities and differences between brain and blood, as well as between East Asian and European ancestry samples. We identified 228 novel CpG sites associated with ADNC and characterized their potential functional roles. Through integrative analysis of age, methylation, and ADNC, we delineated the age-dependent dynamics of ADNC-associated methylation sites and quantified the proportion of age-ADNC associations mediated by methylation. These findings provide valuable insights into the epigenetic mechanisms underlying brain aging and ADNC. CBMAP is a highly valuable data resource for the study of aging, DNA methylation, and early neuropathological changes. On one hand, compared to ROSMAP, CBMAP offers a larger sample size and a broader age range, making it particularly suited for studying age-related methylation patterns. On the other hand, as CBMAP samples were sourced from a national brain bank, they encompass a diverse range of causes of death and were not specifically selected for cognitive statuses. This allowed inclusion of numerous samples with neurodegenerative changes while in absence of overt cognitive impairment, facilitating the investigation of early-stage molecular features of AD and related pathologies. Our results revealed that age-related methylation effects in the PFC were largely consistent in populations of East Asian and European ancestry, although some ancestry-specific signals were observed. In contrast, substantial differences were observed between age-associated methylation in the brain and blood. Analysis of MethAge confirmed a degree of cross-ancestry conservation in cortical methylation clock models, but marked discrepancies between PBMC and PFC methylation clocks, aligning with findings by Shireby et al. 7 Among age-related methylation sites, those most strongly influenced by age were frequently located near histone-related genes. While histone expression is known to decline with age, our mediation analyses further demonstrated that this downregulation in the PFC is largely driven by DNA methylation. WGCNA analysis specified an ADNC-related co-regulation module featured by histone-related methylation sites, suggesting a potential mechanism whereby age influences methylation levels, which in turn regulate histone gene expression, impacting chromatin stability and remodeling, and contributing to neurodegenerative changes. We found that ADNC-associated methylation sites were significantly enriched in pathways related to immune response, vesicles trafficking, and granule membranes—recent research foci of AD research. Neuroinflammation in AD has a dual role, while chronic inflammation exacerbates pathology, protective immune functions, such as pathological protein clearance, offer therapeutic benefits 40,41 . Dysfunction in granule membranes and vesicles is implicated in impaired protein clearance, disrupted neurotransmitter transport, and pathological protein propagation, likely playing a critical role in early neuronal dysfunction in AD 42,43 . Our large-scale epigenomic study reinforces the importance of these pathways in early AD neuropathological changes and identifies a set of novel CpG sites. Co-expression module analysis further refined and prioritized functional changes related to immune cell activation, ECM, oxidative phosphorylation, and neurotransmitter transporter activity, which may drive neuroinflammation, Aβ and tau accumulation, mitochondrial dysfunction, and synaptic impairment, ultimately exacerbating neuronal damage and cognitive decline. Beyond well-established AD-related pathways, we uncovered novel evidence, including the identification of cg00601836. Cg00601836 is a methylation site associated with Aβ deposits, mapping to ESR1 , which encodes estrogen receptor α (ERα). Beyond its expression in reproductive tissues, ERα is predominantly expressed in the hippocampus and PFC and was found to play a substantial role on memory and cognitive function 44 . Reduced ERα activity may trigger neuroinflammatory pathways, acting as an upstream driver of AD 45 . Notably, genetic polymorphisms near ESR1 are associated with increased AD risk, potentially by influencing cholesterol metabolism and promoting Aβ accumulation 46 , complementing our observation that ESR1 expression may be regulated by methylation, impacting Aβ deposits. This provides convergent genetic and epigenetic evidence for ESR1’s role in AD pathology. Integrative analyses of age, methylation, and ADNC revealed that the majority of ADNC-associated methylation effects can be attributed to aging, with DNA methylation mediating 27.3% of the effect of age on ADNC. This proportion may be underestimated, as we only considered CpG sites significantly associated with ADNC in the CBMAP-ROSMAP meta-analysis. While most age-related methylation changes promote ADNC progression, a subset, including a CpG site regulating NCF1 (linked to oxidative stress), exhibited protective effects against neurodegeneration 38 . Further mechanistic exploration is warranted. Additionally, integrative transcriptomic analysis established a pathway whereby increased methylation of cg09221482 with age upregulates MAP3K4 expression, correlating with higher Braak NFT stages. MAP3K4 is associated with glial cell activation, with studies by Shen et al. 47 showing that MAP3K4 interacts with GADD45G to activate neuroimmune signaling in astrocytes, and others indicating that Fra-1 upregulates Map3k4 transcription to enhance microglial activity during neuroinflammation. Key strengths of this study include CBMAP’s broad age range, enabling robust analysis of age-related methylation changes. Using ADNC as the primary outcome, rather than clinical cognitive impairment, facilitated the identification of early molecular signatures. Integration of methylation and transcriptomic data clarified the mapping of methylation sites to genes, enhancing the accuracy of pathway identification. Comprehensive analyses of age, methylation, and ADNC systematically quantified methylation’s mediation of age-related ADNC risk and its underlying patterns. Limitations include the relatively small number of samples under age 50, though donors in this age group are scarce. This will probably be addressed in CBMAP Phase II as we keep increasing the inclusion. In addition, the ADNC-associated methylation sites we identified require further functional investigation. In conclusion, our study comprehensively profiled DNA methylation for understanding brain aging and ADNC in an East Asian population, identifying novel CpG sites involved in immune response, vesicle membrane, ECM, and estrogen receptor–related functions and signaling pathways. We revealed that DNA methylation mediates 27.3% of the effect of aging on ADNC, underscoring its pivotal role as a molecular bridge between aging and neurodegeneration. Methods Sample collection The human brain tissue samples used in this study were obtained from the CBMAP initiative. The first phase of this project was launched by the China Human Brain Bank Consortium, adhering to a unified standard for brain tissue collection and processing. Participating institutions include the Human Brain Tissue Resource Center for Health and Disease at Zhejiang University (ZJU) in eastern China, the National Brain Bank for Development and Function at Peking Union Medical College (PUMC) in the north, and the Xiangya Brain Bank at Central South University (CSU) in central China. Informed consent was obtained either from the donors themselves or their families through voluntary donation agreements, allowing the use of both biological specimens and associated data for research purposes. Ethical approval for this study was granted by the Ethics Committee of Zhejiang University School of Medicine (Approval Nos. 2020-005 and 2024-007). In Phase I of the CBMAP, we specifically focused on the prefrontal cortex (PFC) region. Anatomically, tissue samples were dissected from the region encompassing Brodmann Area 9 (BA9) within the superior frontal gyrus. We performed DNA methylation profiling on PFC samples from a total of 1,057 donors. Neuropathological diagnosis and inclusion/exclusion criteria As detailed in the CBMAP study profile and our recent comorbidity study 25 , all donor brains underwent standardized postmortem evaluations. Following fixation and coronal sectioning, representative brain regions were systematically sampled for immunohistochemical (IHC) analysis. Uniform assessment criteria were applied across all sites. IHC results were interpreted based on established neuropathological staging systems, including Thal phasing for Aβ deposits (converted to A-score), Braak neurofibrillary tangle staging for hyperphosphorylated tau (B-score), and the CERAD neuritic plaque score (C-score). The severity of Alzheimer’s disease neuropathological changes (ADNC) was classified using the ABC scoring system, which designates overall pathology as Not (control), Low, Intermediate, or High 13 . In addition, diagnoses were also made for Lewy body disease (LBD), cerebrovascular disease (CVD), primary age-related tauopathy (PART), aging-related tau astrogliopathy (ARTAG), and limbic-predominant age-related TDP-43 encephalopathy (LATE). In the study identifying ADNC-associated CpG sites, we excluded individuals with schizophrenia, amyotrophic lateral sclerosis, epilepsy, as well as those with severe congenital neurodevelopmental abnormalities or brain tumors. For the ADNC control group, in addition to requiring the ADNC classification to be “Not”, we further excluded samples from individuals with documented cognitive decline, a clinical diagnosis of AD or PD, or a postmortem pathological diagnosis of LBD or ARTAG (Gray matter, Perivascular, or Subpial). DNA Methylation Measurement and Quality Control Genomic DNA was extracted from the CBMAP prefrontal cortex (PFC) samples using a nuclear lysis buffer, which included 10% sodium dodecyl sulfate (SDS) and DNA Extraction Kit. DNA methylation levels were measured for 930,140 CpG sites using the HumanMethylationEPIC v2.0 array. A pilot batch of 68 samples was initially processed, followed by 989 additional samples analyzed on the same platform with identical arrays, totaling 1,057 samples. Quality control (QC) involved sequential filtering for duplicated probes, low-quality probes, SNP-containing probes, probes with excessive missing data, and sex mismatches ( Extended Data Fig. 1 ). After QC, 907,506 CpG sites from 1,031 samples were retained for downstream analysis. Beta values, representing the proportion of methylation at each CpG site, were calculated and used for subsequent analyses. The ROSMAP data was accessed from the AD Knowledge Portal (ID: 9603055) and the RADC Research Resource Sharing Hub (ID: 6014). The data process was consistent with CBMAP. Cell-type Proportion Estimation Cell-type proportions were estimated using reference-based deconvolution with single-cell methylome sequencing data from Luo et al. 23 , comprising more than 15k cells from human frontal cortex, identifying seven major cell types (excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, microglia, oligodendrocytes, and OPCs). Cell-type-specific markers were selected based on extreme beta values, resulting in 99-983 markers per cell type. Marker validity was confirmed through visualization and consistency checks across reference and bulk samples. The estimation was conducted using the pipeline from Gandal lab (https://github.com/gandallab/brain_CTP_deconv). Age-Related Methylation and Cross-Tissue, Cross-Ancestry Comparisons To identify age-associated methylation sites, we employed linear regression models with methylation levels at each CpG site as the dependent variable, age (in years) as the independent variable, and sex, sample plate, and cell type proportions as covariates. To facilitate cross-tissue and cross-ancestry comparisons, similar analyses were conducted on 734 samples from ROSMAP (primarily PFC of European ancestry) 48 and 656 samples from the Hannum et al. study 6 (hereafter referred to as 'Hannum', primarily whole blood of European ancestry). CpG-gene mapping and KEGG/GO/GSEA Enrichment Analysis Potential target genes regulated by methylation sites were identified in CBMAP samples by analyzing associations between methylation levels and the expression of nearby genes. Gene expression was quantified using RNA sequencing (RNA-seq) of PFC samples. Briefly, RNA was extracted from tissue using TRIzol reagent. RNA libraries were prepared using the N406-01 Ribo-off rRNA Depletion Kit (Human/Mouse/Rat, Novogene) to remove ribosomal RNA. Sequencing was performed on the BGI DNBSEQ platform with paired-end 150 bp (PE150) reads. After quality control, raw counts were normalized to transcripts per million (TPM) and log-transformed for analysis. Associations between methylation levels and gene expression (considering a window 1 Mb both sides of the CpG) were assessed using linear regression models, adjusting for covariates such as age, sex, and the proportion of neuronal cells. A significance threshold of FDR < 0.05 to identify significant methylation-gene expression pairs. Functional enrichment of potential target genes was performed using the R package “clusterProfiler”. Gene Set Enrichment Analysis (GSEA) 49 , Gene Ontology (GO) 50 terms (Biological Process, Molecular Function, and Cellular Component) and Kyoto Encyclopedia of Genes and Genomes (KEGG) 51 pathways were tested, with significance set at FDR < 0.05. Enrichment analyses were conducted for CpG sites associated with age, ADNC, Braak NFT stage, and Aβ deposits. Mediation Analysis To investigate whether methylation mediates the effects of age on gene expression, age on ADNC we conducted mediation analyses using the R package “bruceR” and adjusted for sex and proportion of neuronal cells. For assessing the overall mediation effect of methylation on the age-ADNC relationship, we performed principal component analysis (PCA) on all methylation sites mediating age-ADNC effects, retaining the top 10 principal components (PCs). Mediation analysis was conducted for each PC to estimate its contribution to the age-ADNC relationship. The cumulative mediation proportion was calculated by summing the mediation effects of significant PCs. Mediation analyses used a bootstrap resampling (1,000 simulation samples) to estimate mediation proportions and confidence intervals. The mediation analysis was also applied to estimate the role of gene expression on methylation-ADNC association. Methylation Clock and Age Acceleration A PFC-specific methylation age (MethAge) model was developed using 501 samples from eastern and central China via regularized regression (elastic net) and a combination of principal component analysis and regularized regression, with 10-fold cross-validation to optimize hyperparameters. The model was validated in 530 samples from northern China, with prediction accuracy assessed using Pearson’s correlation coefficient (r) and root mean square error (RMSE). The Cortical clock 7 and PCBrainAge 52 based on cortical tissue of European ancestry, MultiClock 3 based on multiple tissues of European ancestry, Hannum Clock 6 , BloodClock 53 and PhenoClock 54 based on blood tissue, and ICAS-DNAmAge 21 model from blood tissue of Chinese ancestry were also evaluated in the CBMAP for cross-ancestry and cross-tissue comparisons. Age acceleration was quantified using a ratio-based method, whereby the residual from a regression of predicted age on chronological age was divided by the chronological age to account for individual differences in age scale. Bayesian Network Analysis To infer conditional dependencies between MethAge, ADNC, and neuropathological variables (e.g., PART, LATE, ARTAG, cerebrovascular pathologies), we employed a Bayesian network framework using the R package “bnlearn” 55 . A hill-climbing algorithm 56 was used to learn the network structure, with constraints to ensure ADNC as the primary outcome and MethAge as an upstream variable. We compiled the Bayesian network into a concatenated tree using the “gRain” package and recomputed the conditional distribution of ADNC given its parent nodes in the Bayesian network. For ADNC, we estimated conditional probability tables (CPTs) using maximum likelihood estimation (MLE) based on the observed frequency of variable configurations in the data. Identification of ADNC-Associated Methylation Sites and Meta-Analysis Epigenome-wide association studies (EWAS) were conducted for ADNC, Braak NFT stage, and Aβ deposits using linear regression models, with methylation levels as the dependent variable and ADNC scores (0-3), Braak NFT stage (0-6), or Aβ scores (0-3) as independent variables, adjusted for age, sex, sample plate and the proportion of neuronal cells. Multiple testing correction was applied using FDR < 0.05. To increase statistical power, a meta-analysis was performed combining CBMAP and ROSMAP results. Heterogeneity was assessed using Cochran’s Q test 57 ; fixed-effects models 58 were used for CpG sites without heterogeneity ( P > 0.05), and random-effects models 59 were applied otherwise. The R package “meta” was used for meta-analysis. WGCNA Analysis Weighted gene co-expression network analysis (WGCNA) was performed using the R package WGCNA on 23,199 nominally significant ADNC-associated CpG sites ( P 0.8). Modules were identified using dynamic tree cutting, and module eigengenes (first principal components) were correlated with ADNC, Braak NFT stage, and Aβ scores via regression analysis. Hub CpG sites were identified based on high module membership, and their potential target genes were subjected to KEGG/GO enrichment analysis as described above. Pseudotime Trajectory Fitting and Unsupervised Hierarchical Clustering To model age-related trajectories of ADNC-associated methylation sites, we fitted pseudotime trajectories using loess regression with methylation levels as the dependent variable and age as the independent variable. Trajectories were clustered using the k-means method based on their curve shapes, implemented in the R package "stats". Clusters were functionally annotated by performing KEGG/GO enrichment analysis on the potential target genes of CpG sites within each cluster, as described above. Declarations Data availability The omics and metadata of CBMAP can be accessed in the OMIX database (https://ngdc.cncb.ac.cn/omix/) under the accession number OMIX010874. Code availability The codes for data analysis are available at https://github.com/zdangm/CBMAP_methylation_aging_AD/ Acknowledgements We are deeply grateful to all brain tissue donors and pay our heartfelt tribute to them and their families. We thank the National Health and Disease Human Brain Tissue Resource Center (Zhejiang University), the National Human Brain Bank for Development and Function (Peking Union Medical College), and the Xiangya Brain Bank (Central South University) for providing the brain tissue samples used in this project. Funding The project was supported by the Science Innovation 2030 - Brain Science and Brain-Inspired Intelligence Technology Major Project #2021ZD0201100 (Brain Tissue Resource Repository and Brain Bank Collaboration Network Platform, tasks 2 [2021ZD0201102] and 6 [2021ZD0201106]) from the Ministry of Science and Technology (MOST) of China, the National Natural Science Foundation of China (82020108012 [J.Z.], 82204118 [D.Z.], 82022024 [C.C.], 32270656 [D.Z.], and 82370612 [Z.S.]), the Key R&D Program of Zhejiang (2024C03098, [J.Z.]), and the CAMS Innovation Fund for Medical Sciences (CIFMS) #2021-1-I2M-025 [C.M., W.Q.]. Author information These authors contributed equally: Dan Zhou, Wenli Zhai, Wenwei Fang. Contributions J.Z., D.S., B.A., C.M., and D.Z. conceived the study. W.Z., W.F., Y.Z., Z.L., C.Y. were involved in study design. W.Z. and W.F. processed DNA methylation data and performed the data analyses. D.Z., W.Z. and W.F. wrote the first draft. J.Z., Y.Z., D.D.Z., S.D., Z.L, L.W., C.Y., Y.H., H.L., K.Z., Y.S., L.W., X.Y., A.B., W.Q., and C.M. commented and revised the manuscript. All authors reviewed and agreed to submit the manuscript. Ethics declarations Competing interests The authors declare no competing interests. References Querfurth, H. W. & LaFerla, F. M. Alzheimer’s Disease. New England Journal of Medicine 362 , 329–344 (2010). Hou, Y. et al. Ageing as a risk factor for neurodegenerative disease. Nat Rev Neurol 15 , 565–581 (2019). Horvath, S. DNA methylation age of human tissues and cell types. Genome Biology 14 , 3156 (2013). Booth, L. N. & Brunet, A. The Aging Epigenome. Mol Cell 62 , 728–744 (2016). Raddatz, G. et al. Aging is associated with highly defined epigenetic changes in the human epidermis. Epigenetics & Chromatin 6 , 36 (2013). Hannum, G. et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 49 , 359–367 (2013). Shireby, G. L. et al. Recalibrating the epigenetic clock: implications for assessing biological age in the human cortex. Brain 143 , 3763–3775 (2020). Horvath, S. et al. Aging effects on DNA methylation modules in human brain and blood tissue. Genome Biology 13 , R97 (2012). De Jager, P. L. et al. Alzheimer’s disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci. Nat Neurosci 17 , 1156–1163 (2014). Watson, C. T. et al. Genome-wide DNA methylation profiling in the superior temporal gyrus reveals epigenetic signatures associated with Alzheimer’s disease. Genome Medicine 8 , 5 (2016). Zhang, L. et al. Epigenome-wide meta-analysis of DNA methylation differences in prefrontal cortex implicates the immune processes in Alzheimer’s disease. Nat Commun 11 , 6114 (2020). Zhou, D. et al. The China Brain Multi-omics Atlas Project (CBMAP). bioRxiv (2025) doi:10.1101/2025.04.30.651445. Montine, T. J. et al. National Institute on Aging-Alzheimer’s Association guidelines for the neuropathologic assessment of Alzheimer’s disease: a practical approach. Acta Neuropathol 123 , 1–11 (2012). Yi, S.-J. & Kim, K. New Insights into the Role of Histone Changes in Aging. Int J Mol Sci 21 , 8241 (2020). Wang, K. et al. Epigenetic regulation of aging: implications for interventions of aging and diseases. Sig Transduct Target Ther 7 , 374 (2022). Beglopoulos, V. et al. Neurexophilin 3 Is Highly Localized in Cortical and Cerebellar Regions and Is Functionally Important for Sensorimotor Gating and Motor Coordination. Mol Cell Biol 25 , 7278–7288 (2005). H, Y. et al. Mendelian randomization integrating GWAS and eQTL data revealed genes pleiotropically associated with major depressive disorder. Translational psychiatry 11 , (2021). Ali-Rahmani, F., Schengrund, C.-L. & Connor, J. R. HFE gene variants, iron, and lipids: a novel connection in Alzheimer’s disease. Front. Pharmacol. 5 , (2014). Dzwonek, J. & Wilczynski, G. M. CD44: molecular interactions, signaling and functions in the nervous system. Frontiers in Cellular Neuroscience 9 , 175 (2015). Jordan, A. R., Racine, R. R., Hennig, M. J. P. & Lokeshwar, V. B. The Role of CD44 in Disease Pathophysiology and Targeted Treatment. Front. Immunol. 6 , (2015). Zheng, Z. et al. DNA methylation clocks for estimating biological age in Chinese cohorts. Protein Cell 15 , 575–593 (2024). Levine, M. E., Lu, A. T., Bennett, D. A. & Horvath, S. Epigenetic age of the pre-frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer’s disease related cognitive functioning. Aging (Albany NY) 7 , 1198–1211 (2015). Luo, C. et al. Single nucleus multi-omics identifies human cortical cell regulatory genome diversity. Cell Genom 2 , 100107, 100106 (2022). Yap, C. X. et al. Brain cell-type shifts in Alzheimer’s disease, autism, and schizophrenia interrogated using methylomics and genetics. Sci Adv 10 , eadn7655. Wang, X. et al. Prevalence of mixed neuropathologies in age-related neurodegenerative diseases: A community-based autopsy study in China. Alzheimers Dement 21 , e14369 (2025). Fitzpatrick, E. R., Hu ,Tinghui, Ciccarelli ,Bryan T & and Whitehead, I. P. Regulation of vesicle transport and cell motility by Golgi-localized Dbs. Small GTPases 5 , e972860 (2014). Gour, N. & Dong, X. The MRGPR family of receptors in immunity. Immunity 57 , 28–39 (2024). Borggrewe, M. et al. VISTA expression by microglia decreases during inflammation and is differentially regulated in CNS diseases. Glia 66 , 2645–2658 (2018). Ceeraz, S. et al. VISTA Deficiency Accelerates the Development of Fatal Murine Lupus Nephritis. Arthritis Rheumatol 69 , 814–825 (2017). Li, N. et al. Immune-checkpoint protein VISTA critically regulates the IL-23/IL-17 inflammatory axis. Sci Rep 7 , 1485 (2017). Carroll, K. R. et al. Lupus autoantibodies initiate neuroinflammation sustained by continuous HMGB1:RAGE signaling and reversed by increased LAIR-1 expression. Nat Immunol 25 , 671–681 (2024). Ayka, A. & Şehirli, A. Ö. The Role of the SLC Transporters Protein in the Neurodegenerative Disorders. Clin Psychopharmacol Neurosci 18 , 174–187 (2020). Zhang, B. & Horvath, S. A general framework for weighted gene co-expression network analysis. Stat Appl Genet Mol Biol 4 , Article17 (2005). Kommaddi, R. P. et al. Aβ mediates F-actin disassembly in dendritic spines leading to cognitive deficits in Alzheimer’s disease. J Neurosci 38 , 1085–1099 (2018). Wang, S. & El-Deiry, W. S. TRAIL and apoptosis induction by TNF-family death receptors. Oncogene 22 , 8628–8633 (2003). Simard, A. R., Soulet, D., Gowing, G., Julien, J.-P. & Rivest, S. Bone marrow-derived microglia play a critical role in restricting senile plaque formation in Alzheimer’s disease. Neuron 49 , 489–502 (2006). Yalçın, H. Y. et al. Expression of Autophagy-Related Proteins in Microlissencephaly Associated with a Novel Variant in the WDR81 Gene. Mol Syndromol 16 , 1–10 (2025). Marie-Laure Pinel-Marie, Sparfel, L., Desmots, S. & Fardel, O. Aryl hydrocarbon receptor-dependent induction of the NADPH oxidase subunit NCF1/p47 phox expression leading to priming of human macrophage oxidative burst. Free Radic Biol Med 47 , 825–834 (2009). Hong, S. B., Kim, B.-W., Kim, J. H. & Song, H. K. Structure of the autophagic E2 enzyme Atg10. Acta Crystallogr D Biol Crystallogr 68 , 1409–1417 (2012). Kiraly, M., Foss, J. F. & Giordano, T. Neuroinflammation, Its Role in Alzheimer’s Disease and Therapeutic Strategies. J Prev Alzheimers Dis 10 , 686–698 (2023). Amelimojarad, M., Amelimojarad, M. & Cui, X. The emerging role of brain neuroinflammatory responses in Alzheimer’s disease. Front. Aging Neurosci. 16 , (2024). Blackstone, C., Elwood, F., Plun-Favreau, H. & Lewis, P. A. Vesicle trafficking and pathways to neurodegeneration. Molecular Neurodegeneration 16 , 56 (2021). Xue, M., Cao, Y., Shen, C. & Guo, W. Computational Advances of Protein/Neurotransmitter-membrane Interactions Involved in Vesicle Fusion and Neurotransmitter Release. J Mol Biol 435 , 167818 (2023). Almey, A., Milner, T. A. & Brake, W. G. Estrogen receptors in the central nervous system and their implication for dopamine-dependent cognition in females. Horm Behav 74 , 125–138 (2015). Liu, J., Yuan, S., Niu, X., Kelleher, R. & Sheridan, H. ESR1 dysfunction triggers neuroinflammation as a critical upstream causative factor of the Alzheimer’s disease process. Aging (Albany NY) 14 , 8595–8614 (2022). Cheng, D., Liang, B., Hao, Y. & Zhou, W. Estrogen receptor α gene polymorphisms and risk of Alzheimer’s disease: evidence from a meta-analysis. Clin Interv Aging 9 , 1031–1038 (2014). Shen, T. et al. GADD45G operates as a pathological sensor orchestrating reactive gliosis and neurodegeneration. Neuron S0896-6273(25)00345–9 (2025) doi:10.1016/j.neuron.2025.04.033. Huo, Z. et al. DNA methylation variability in Alzheimer’s disease. Neurobiol Aging 76 , 35–44 (2019). Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences 102 , 15545–15550 (2005). Ashburner, M. et al. Gene Ontology: tool for the unification of biology. Nat Genet 25 , 25–29 (2000). Kanehisa, M. & Goto, S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28 , 27–30 (2000). Thrush, K. L. et al. Aging the brain: multi-region methylation principal component based clock in the context of Alzheimer’s disease. Aging (Albany NY) 14 , 5641–5668 (2022). Zhang, Q. et al. Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing. Genome Med 11 , 54 (2019). Levine, M. E. et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 10 , 573–591 (2018). Nagarajan, R., Scutari, M. & Lèbre, S. Bayesian Networks in R: With Applications in Systems Biology . (Springer, New York, NY, 2013). doi:10.1007/978-1-4614-6446-4. Tsamardinos, I., Brown, L. E. & Aliferis, C. F. The max-min hill-climbing Bayesian network structure learning algorithm. Mach Learn 65 , 31–78 (2006). Higgins, J. P. T. & Thompson, S. G. Quantifying heterogeneity in a meta-analysis. Stat Med 21 , 1539–1558 (2002). Rice, K., Higgins, J. P. T. & Lumley, T. A Re-Evaluation of Fixed Effect(s) Meta-Analysis. J R Stat Soc Ser A Stat Soc 181 , 205–227 (2018). Riley, R. D., Higgins, J. P. T. & Deeks, J. J. Interpretation of random effects meta-analyses. BMJ 342 , d549 (2011). Additional Declarations There is NO Competing Interest. Supplementary Files methylationagingADsupplementary0808.docx Supplementary Figure SupplementaryData.xlsx Supplementary Data Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7381745","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":506962657,"identity":"1718b2ff-3523-4217-a8a8-5ff65e2cb72a","order_by":0,"name":"Jing Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/ElEQVRIiWNgGAWjYBACxgYeECUhx8/efIBBAsQ+QJwWC2PJnmMJxGlhYABrqUjccCPHACJASAtz/9mDnwt+SSTOnJHz+YNlG4Mc340Exs8FeB12Lll6Zp+EcT/P2w0Gkm0MxpI3EpilZ+DT0thjIM3bIyE7sz13QwJQC9CFCWzMPPi0NPMY/wZqYdxwIOfBAaCWesJa2njMpHl+SChuOJHD2ADUkmBAUEsPj5k1b4MEKJCNGSTOSRjOPPOwWRqfFsP+M8a3ef7UgaLy8WeJMht5vuPJBz/j1dIAdh2EwywBjkzGBjwaGBjkweQfqCs/4FU7CkbBKBgFIxUAAC48TMUP2nhVAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-7222-8317","institution":"Zhejiang University","correspondingAuthor":true,"prefix":"","firstName":"Jing","middleName":"","lastName":"Zhang","suffix":""},{"id":506962658,"identity":"6d0ed048-7e73-4e9e-9737-dcfb4dcc0dee","order_by":1,"name":"Dan Zhou","email":"","orcid":"https://orcid.org/0000-0002-5313-8164","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Dan","middleName":"","lastName":"Zhou","suffix":""},{"id":506962659,"identity":"e30505e3-0736-4db2-b802-fe68eed69e03","order_by":2,"name":"Wenli Zhai","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wenli","middleName":"","lastName":"Zhai","suffix":""},{"id":506962660,"identity":"041f0226-0b04-4d9d-b6b2-8cbfe925fe77","order_by":3,"name":"Wenwei Fang","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Wenwei","middleName":"","lastName":"Fang","suffix":""},{"id":506962661,"identity":"f867c537-da77-45ea-ba0f-c4e62523ec7a","order_by":4,"name":"Yuan Zhou","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Zhou","suffix":""},{"id":506962662,"identity":"40d1fe49-531b-42fb-9c7d-9c95d159579a","order_by":5,"name":"Zuyun Liu","email":"","orcid":"https://orcid.org/0000-0001-6120-5913","institution":"Department of Big Data in Health Science School of Public Health, Zhejiang Key Laboratory of Intelligent Preventive Medicine, Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zuyun","middleName":"","lastName":"Liu","suffix":""},{"id":506962663,"identity":"000deb84-7bf6-4d1f-8e4a-d41cb49bebf6","order_by":6,"name":"Changzheng Yuan","email":"","orcid":"https://orcid.org/0000-0002-2389-8752","institution":"School of Public Health, the Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China","correspondingAuthor":false,"prefix":"","firstName":"Changzheng","middleName":"","lastName":"Yuan","suffix":""},{"id":506962664,"identity":"9aad4dff-2a6b-4b00-8d70-0944e2da6f52","order_by":7,"name":"Lang Wu","email":"","orcid":"https://orcid.org/0000-0001-9938-3627","institution":"UNIVERSITY OF HAWAII AT MANOA","correspondingAuthor":false,"prefix":"","firstName":"Lang","middleName":"","lastName":"Wu","suffix":""},{"id":506962665,"identity":"95d80b3c-618a-4989-a90d-98f50b93b4df","order_by":8,"name":"Yu-Ting Hu","email":"","orcid":"https://orcid.org/0000-0002-8707-3995","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yu-Ting","middleName":"","lastName":"Hu","suffix":""},{"id":506962666,"identity":"39078cc1-68b6-465c-9f18-40b50a957a41","order_by":9,"name":"Jiayao Fan","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jiayao","middleName":"","lastName":"Fan","suffix":""},{"id":506962667,"identity":"900090c2-c952-45ea-a69b-88b00b0768e2","order_by":10,"name":"Shiao Zhou","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shiao","middleName":"","lastName":"Zhou","suffix":""},{"id":506962668,"identity":"ea91f74a-d0e1-4c6f-afbd-7882e17e9acb","order_by":11,"name":"Huaqing Liu","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Huaqing","middleName":"","lastName":"Liu","suffix":""},{"id":506962669,"identity":"4abf38cf-8ad0-4147-8072-0f293226697e","order_by":12,"name":"Keqing Zhu","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Keqing","middleName":"","lastName":"Zhu","suffix":""},{"id":506962670,"identity":"e7ac4ec9-dc2c-49be-b1c9-80f769e919cc","order_by":13,"name":"Yi Shen","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Shen","suffix":""},{"id":506962671,"identity":"bfa553df-ae03-41ef-8825-72c8a9ca6bff","order_by":14,"name":"Xiaoxin Yan","email":"","orcid":"https://orcid.org/0000-0001-6654-982X","institution":"Department of Anatomy and Neurobiology, Central South University Xiangya Medical School","correspondingAuthor":false,"prefix":"","firstName":"Xiaoxin","middleName":"","lastName":"Yan","suffix":""},{"id":506962672,"identity":"4c040c43-6f92-4693-840a-8a1f563a7d47","order_by":15,"name":"Wenying Qiu","email":"","orcid":"","institution":"Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"Wenying","middleName":"","lastName":"Qiu","suffix":""},{"id":506962673,"identity":"151d3ca8-c388-4a45-a564-56c062610f7e","order_by":16,"name":"Dandan Zhang","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Dandan","middleName":"","lastName":"Zhang","suffix":""},{"id":506962674,"identity":"0230bdb6-40b6-47c3-a5a3-2f424933660f","order_by":17,"name":"Chao Ma","email":"","orcid":"","institution":"Peking Union Medical College","correspondingAuthor":false,"prefix":"","firstName":"Chao","middleName":"","lastName":"Ma","suffix":""},{"id":506962675,"identity":"26996a6e-f28f-445a-a8f6-f4064cef9c5c","order_by":18,"name":"Ai-Min Bao","email":"","orcid":"https://orcid.org/0000-0002-8755-1950","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Ai-Min","middleName":"","lastName":"Bao","suffix":""},{"id":506962676,"identity":"f8582f17-18db-40ca-8cbc-14c5febd47e6","order_by":19,"name":"Shumin Duan","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shumin","middleName":"","lastName":"Duan","suffix":""}],"badges":[],"createdAt":"2025-08-15 13:36:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7381745/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7381745/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90169800,"identity":"542d0c7e-ebd1-49f6-98fe-795abe59d4ab","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":164726,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSample collection and study overview. a\u003c/strong\u003e, The CBMAP phase I centers. \u003cstrong\u003eb\u003c/strong\u003e, Violin plots show the age distribution of CBMAP and ROSMAP samples with DNA methylation profiled. \u003cstrong\u003ec\u003c/strong\u003e, Based on the ADNC staging, samples were classified into Control (n = 482), ADNC Low (n = 281), Intermediate (n = 205), and High (n = 30) groups. Integrative analysis was performed linking aging, DNA methylation, and ADNC.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/9d9e0cc5b2b66dc66420215c.png"},{"id":90169801,"identity":"39d02f10-e63b-4af9-ba0b-3cb729b098f8","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":344247,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAge-associated DNA methylation sites and their cross-dataset comparisons. a,\u003c/strong\u003e Manhattan plot showing genome-wide associations between age and DNA methylation in CBMAP, with over 40% of CpG sites significantly associated with age (FDR \u0026lt; 0.05). \u003cstrong\u003eb, \u003c/strong\u003ePairwise correlations of age effect sizes on DNA methylation among CBMAP (EAS, PFC), ROSMAP (EUR, PFC), and Hannum (EUR, PBMC) datasets. Color intensity indicates the strength of correlation. \u003cstrong\u003ec,\u003c/strong\u003e Heatmaps of effect size correlations across datasets, stratified by CpG site genomic annotations. \u003cstrong\u003ed,\u003c/strong\u003e Venn diagram showing overlap among the top 1,000 age-associated CpG sites in each dataset.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/b76291ee68db359d59f2ab95.png"},{"id":90170118,"identity":"01729ea4-d295-45ec-9d80-45066835a64f","added_by":"auto","created_at":"2025-08-29 11:07:32","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":552036,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDNA Methylation Mediates the Effect of Aging on Transcriptional Regulation. a, \u003c/strong\u003eMethylation levels of top 1000 age-related CpGs (strongest positive/negative) and their association with chronological age \u003cstrong\u003eb,\u003c/strong\u003eFunctional enrichment analysis of genes linked to the most age-related CpGs. \u003cstrong\u003ec,\u003c/strong\u003ecg03459567 mediates age-associated \u003cem\u003eH2BC12\u003c/em\u003e downregulation. Scatter plots show the negative methylation-\u003cem\u003eH2BC12\u003c/em\u003e association and the negative age-\u003cem\u003eH2BC12\u003c/em\u003eassociation. \u003cstrong\u003ed,\u003c/strong\u003e Age-related hypermethylation CpGs mediates the negative association between aging and histone expression. \u003cstrong\u003ee,\u003c/strong\u003e Mediation path diagrams for representative CpGs illustrating the indirect effect of age on \u003cem\u003eNXPH3\u003c/em\u003e, \u003cem\u003eBTN3A2\u003c/em\u003e, \u003cem\u003eHFE\u003c/em\u003e via methylation. \u003cstrong\u003ef,\u003c/strong\u003e Nonlinear age-related methylation dynamics across chronological age. \u003cstrong\u003eg,\u003c/strong\u003e cg13084525 mediates age-associated \u003cem\u003eCD44\u003c/em\u003e upregulation. Scatter plots show the negative methylation-\u003cem\u003eCD44\u003c/em\u003e association and the positive age-\u003cem\u003eCD44\u003c/em\u003e association.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/851e5391545766a7112c453f.png"},{"id":90170119,"identity":"f1061060-e577-4dea-b1a9-136e6dffac81","added_by":"auto","created_at":"2025-08-29 11:07:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":335690,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDNA methylation age estimation and its association with AD pathology and brain cell-type composition. a,\u003c/strong\u003e Performance of various DNA methylation clocks in predicting chronological age based on PFC methylation data. Each panel shows the correlation and RMSE between predicted and chronological age using different clock models. \u003cstrong\u003eb,\u003c/strong\u003e DNA methylation age acceleration stratified by ADNC severity and Braak NFT stage. Increased age acceleration is associated with more severe neuropathological burden. \u003cstrong\u003ec,\u003c/strong\u003eEstimated proportions of seven major brain cell types in CBMAP and ROSMAP datasets using DNA methylation deconvolution. \u003cstrong\u003ed,\u003c/strong\u003e Associations between MethAge and estimated cell-type proportions in CBMAP. \u003cstrong\u003ee,\u003c/strong\u003e Scatter plot showing a negative association between age and microglia proportion in CBMAP. \u003cstrong\u003ef,\u003c/strong\u003eStratified associations between MethAge and cell-type proportions in ADNC case and control groups.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/69a5f04e686eccc1676c1d70.png"},{"id":90169804,"identity":"eee00e8f-a75c-4546-8deb-3a36fed56903","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":265184,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMeta-analysis identifies CpG sites associated with AD-related neuropathological features. \u003c/strong\u003eVolcano plots showing meta-analysis results for CpG associations with ADNC (\u003cstrong\u003ea\u003c/strong\u003e), Braak NFT stage (\u003cstrong\u003ec\u003c/strong\u003e), and A score (\u003cstrong\u003ee\u003c/strong\u003e), respectively. Boxplots showing methylation levels of representative CpG sites across pathology severity levels in CBMAP and ROSMAP for ADNC (\u003cstrong\u003eb\u003c/strong\u003e), Braak NFT stage (\u003cstrong\u003ed\u003c/strong\u003e), and A score (\u003cstrong\u003ef\u003c/strong\u003e).\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/7504ade2b28ba193eeb0b98e.png"},{"id":90169805,"identity":"c74587e6-3d4d-4427-af02-cf89e07cef7c","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":435269,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCo-regulated modules of ADNC-associated CpG sites and their functional characterization. a,\u003c/strong\u003e Bar plot showing the number of CpG sites in each of the 39 co-methylation modules identified by WGCNA based on 23,199 ADNC-associated CpG sites. \u003cstrong\u003eb,\u003c/strong\u003e Heatmap displaying associations between module eigengenes and AD-related traits including age, ADNC, Braak NFT stage, Aβ deposits, and C-score. Red and blue colors indicate positive and negative associations, respectively. Significance levels are denoted as * (\u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.05), ** (\u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.01), *** (\u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.001). \u003cstrong\u003ec,\u003c/strong\u003e Gene Ontology (GO) enrichment results for ADNC-related modules based on hub CpG-associated genes. Dot plots show enriched GO terms, with dot size representing gene ratio and color indicating adjusted \u003cem\u003eP\u003c/em\u003e-value.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/d3879f275a1af98d79b406e0.png"},{"id":90170121,"identity":"8e9a1ae3-0d64-42c2-846d-bbe32922e385","added_by":"auto","created_at":"2025-08-29 11:07:32","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":652843,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAge-related Trajectories of Methylation Changes in ADNC, Braak NFT, and Aβ Deposits. a,\u003c/strong\u003e Trajectories of methylation levels across chronological age for ADNC-associated CpG sites, visualized with loess regression. The heatmap shows the z-score of methylation levels across age. \u003cstrong\u003eb,\u003c/strong\u003eUnsupervised k-means clustering identified four distinct clusters of CpG sites based on their age-related methylation patterns. \u003cstrong\u003ec,\u003c/strong\u003e Gene ontology (GO) enrichment analysis for the identified clusters reveals key biological pathways associated with the age-related methylation changes in ADNC-associated CpG sites. \u003cstrong\u003ed,\u003c/strong\u003e Methylation trajectories for CpG sites associated with Braak NFT stage. \u003cstrong\u003ee,\u003c/strong\u003e Unsupervised k-means clustering identifies three distinct clusters of Braak NFT-associated CpG sites based on their methylation patterns across age. \u003cstrong\u003ef,\u003c/strong\u003e GO enrichment analysis for the Braak NFT-associated CpG clusters reveals the major biological pathways linked to their age-dependent methylation changes. \u003cstrong\u003eg,\u003c/strong\u003e Methylation trajectories for CpG sites associated with Aβ deposits. \u003cstrong\u003eh,\u003c/strong\u003e Unsupervised k-means clustering identifies three distinct clusters of CpG sites associated with Aβ deposits based on their age-related methylation patterns. \u003cstrong\u003ei,\u003c/strong\u003e GO enrichment analysis for the Aβ deposit-associated CpG clusters reveals the biological pathways enriched in each cluster, reflecting their role in age-related changes.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/10da19757c2f636cd2b8d802.png"},{"id":90170120,"identity":"8bdadcf9-4a7b-40ec-8053-576f3160902a","added_by":"auto","created_at":"2025-08-29 11:07:32","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":258868,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMediation analyses reveal AD-associated CpG sites linking age to AD Neuropathology. \u003c/strong\u003eVolcano plots showing mediation analysis results for CpG sites linking age to ADNC (\u003cstrong\u003ea\u003c/strong\u003e), Braak NFT stage (\u003cstrong\u003ed\u003c/strong\u003e), and A score (\u003cstrong\u003eg\u003c/strong\u003e), respectively. The x-axis represents mediation proportion. Boxplots of representative CpG sites showing methylation levels across ADNC (\u003cstrong\u003eb\u003c/strong\u003e), Braak stage (\u003cstrong\u003ee\u003c/strong\u003e), and A score (\u003cstrong\u003eh\u003c/strong\u003e) levels in CBMAP (top row) and ROSMAP (bottom row). Mediation path diagrams for representative CpG sites illustrating the indirect effect of age on ADNC (\u003cstrong\u003ec\u003c/strong\u003e), Braak stage (\u003cstrong\u003ef\u003c/strong\u003e), and A score (\u003cstrong\u003ei\u003c/strong\u003e) via methylation.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/a4ee6bbf4ed23430e5ad5a77.png"},{"id":90170930,"identity":"367a71ef-6894-4ada-aaac-efd7eebaa233","added_by":"auto","created_at":"2025-08-29 11:15:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4683931,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/ef4ace8c-7448-47e0-9301-1bd52a35763b.pdf"},{"id":90169807,"identity":"c4dfebc2-e144-4707-9ee4-ffc253b33009","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1758633,"visible":true,"origin":"","legend":"Supplementary Figure","description":"","filename":"methylationagingADsupplementary0808.docx","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/9f2f999b3f9969fe6b52878d.docx"},{"id":90169809,"identity":"0e954841-096b-4d50-a95f-426c9024e04c","added_by":"auto","created_at":"2025-08-29 10:59:32","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":7967240,"visible":true,"origin":"","legend":"Supplementary Data","description":"","filename":"SupplementaryData.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7381745/v1/7d47107d2125c35f8a48c49a.xlsx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Epigenetic Rewiring Connects Aging to Alzheimer’s Pathology in the Human Prefrontal Cortex","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAging remains the strongest risk factor for Alzheimer\u0026rsquo;s disease (AD) and most neurodegenerative disorders, yet the biological mechanisms that link aging to AD pathology are still being uncovered\u003csup\u003e1,2\u003c/sup\u003e. Among candidate mechanisms, DNA methylation\u0026mdash;an epigenetic modification known to change with age\u0026mdash; has emerged as a potential contributor to age-related physiological decline and AD pathogenesis\u003csup\u003e3,4\u003c/sup\u003e. However, our understanding of how age-related DNA methylation in the human brain contributes to AD neuropathology remains limited.\u003c/p\u003e\n\u003cp\u003ePrevious studies have laid the foundation for epigenetic aging research\u003csup\u003e4,5\u003c/sup\u003e, leading to the development of age-predictive \u0026ldquo;epigenetic clocks\u0026rdquo; such as the blood-based Hannum Clock\u003csup\u003e6\u003c/sup\u003e, the multi-tissue Horvath Clock\u003csup\u003e3\u003c/sup\u003e, and a cortex-specific clock developed by Shireby and colleageus\u003csup\u003e7\u003c/sup\u003e. While informative, these efforts have not sufficiently explored the functional relevance of brain-specific methylation changes, nor have they adequately represented non-European populations. Additionally, most existing studies rely heavily on peripheral blood data or focus on narrow age ranges within postmortem brain donors, limiting their capacity to fully characterize brain aging and its relationship to AD across the lifespan.\u003c/p\u003e\n\u003cp\u003eRecent integrative analyses\u003csup\u003e8\u003c/sup\u003e\u003csup\u003e,\u0026nbsp;\u003c/sup\u003e\u003csup\u003e9,10\u003c/sup\u003e\u003csup\u003e,\u003c/sup\u003e\u003csup\u003e11\u003c/sup\u003e, such as those using data from Religious Orders Study and Memory and Aging Project (ROSMAP) have identified methylation loci associated with tau pathology, but these studies fall short in several key areas: they primarily focus on neurofibrillary tangle (tau) burden with limited analysis of amyloid-\u0026beta; (A\u0026beta;) pathology or comprehensive AD neuropathological changes (ADNC); functional insights are often limited by a lack of comprehensive integration with transcriptomic data; the narrow age range of subjects (mostly \u0026ge;70 years) restricts the ability to capture aging trajectories; and the overwhelming focus on individuals of European ancestry leaves a critical gap in understanding epigenetic aging in other populations.\u003c/p\u003e\n\u003cp\u003eTo address these limitations, we leveraged the China Brain Multi-Omics Atlas Project (CBMAP)\u003csup\u003e12\u003c/sup\u003e, profiling DNA methylation in the prefrontal cortex (PFC) of 1,057 individuals aged 1-106 years across multiple Chinese brain banks. This dataset\u0026mdash;the largest of its kind in an East Asian population\u0026mdash;also includes matched transcriptomic data and detailed neuropathological assessments, enabling a comprehensive interrogation of how age-related methylation patterns influence AD-related changes in the human brain.\u003c/p\u003e\n\u003cp\u003eIn this study, we aim to: (1) define age-associated DNA methylation signatures in the PFC across the lifespan; (2) identify CpG sites associated with ADNC, including both amyloid and tau pathology; (3) explore the functional relevance of these sites through integration with transcriptome; and (4) quantify the extent to which DNA methylation mediates the impact of aging on AD pathology. By filling a major gap in brain epigenomics for East Asian populations, our work provides novel insights into the epigenetic bridge between aging and AD.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eSample Characteristics and DNA Methylation Profiling\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAs part of the CBMAP, we profiled DNA methylation on PFC samples from 1,057 donors across eastern, northern, and central China. Of these, 401 (37.9%) were female. The donor ages ranged from 1 to 106 years, with a mean age of 74.9 years. Regional and age distributions are presented in \u003cstrong\u003eFig. 1a,b\u003c/strong\u003e. Notably, compared to ROSMAP, CBMAP exhibits a markedly broader age distribution for the donors with methylation measurements. CBMAP samples underwent standardized neuropathological assessments, including evaluations for A\u0026beta;, Braak NFT stage, and CERAD neuritic plaque score. Following the NIA-AA criteria\u003csup\u003e13\u003c/sup\u003e and exclusion criteria (Methods), samples were classified into four ADNC categories: Control (n = 482), Low (n = 281), Intermediate (n = 205), and High (n = 30) (\u003cstrong\u003eFig. 1c\u003c/strong\u003e). To examine how methylation contributes to AD pathogenesis, ADNC was treated as the primary outcome, while related pathological features, including A\u0026beta; deposits (represented by A score) and Braak NFT stage, were considered as secondary outcomes. DNA methylation was profiled using the Illumina HumanMethylationEPIC v2.0 array, targeting 930,140 CpG sites. After quality control and filtering (Methods), 907,506 CpG sites from 1,031 samples passed quality checks and were included in subsequent analyses (\u003cstrong\u003eExtended Data Fig. 1\u003c/strong\u003e). We performed principal component analysis (PCA) on methylation data to examine potential variations associated with age and regions (\u003cstrong\u003eExtended Data Fig. 2a,b\u003c/strong\u003e). Although the PCA revealed no significant regional differences in global methylation patterns across the three studied regions, we identified 21,355 differentially methylated CpG sites between northern and southern China. Among the top 1,000 differentially methylated CpG sites between north and south (\u003cstrong\u003eExtended Data Fig. 2c\u003c/strong\u003e), significant pathway enrichments were observed in KEGG including \u0026ldquo;B cell receptor signaling pathway\u0026rdquo; (\u003cem\u003eP\u003c/em\u003e-value = 3.74\u0026times;10\u003csup\u003e-6\u003c/sup\u003e) and \u0026ldquo;Viral protein interaction with cytokine and cytokine receptor\u0026rdquo; when excluding the MHC region (\u003cem\u003eP\u003c/em\u003e-value = 5.95\u0026times;10\u003csup\u003e-6\u003c/sup\u003e, \u003cstrong\u003eExtended Data Fig. 2d,e\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 1\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eThe broad age range and comprehensive pathological assessments make the CBMAP dataset a valuable resource for identifying methylation features associated with aging and early neuropathological changes related to ADNC.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEpigenetic landscape of aging in the PFC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify age-associated CpG sites, we fitted linear models adjusting for covariates including sex, estimated cell type proportions, and batch. As shown in \u003cstrong\u003eFig. 2a\u003c/strong\u003e, over one-third of CpG sites (n = 402,503) exhibited significant age-related methylation levels (FDR \u0026lt; 0.05). We characterized all the age-related CpGs by gene set enrichment analysis (GSEA) to manifest the potentially regulated biological process or pathway. By integrating the gene expression from the same tissue samples, we identified 1,349 significant gene sets (FDR \u0026lt; 0.05) which led by gene sets featured by microglia-specifically expressed genes (\u003cstrong\u003eSupplementary Table 2\u003c/strong\u003e). We also observed the downregulation of AD-related terms including extracellular matrix (NABA MATRISOME, \u003cem\u003eP\u003c/em\u003e-value = 3.73\u0026times;10\u003csup\u003e-15\u003c/sup\u003e) and secretory granule (GOCC SECRETORY GRANULE, \u003cem\u003eP\u003c/em\u003e-value = 1.63\u0026times;10\u003csup\u003e-13\u003c/sup\u003e).\u003c/p\u003e\n\u003cp\u003eTo investigate whether the effect of age on methylation varies across tissues and ancestries, we conducted a comparative analysis of age-associated methylation across three datasets: CBMAP (East Asian [EAS], PFC), ROSMAP (primarily European ancestries [EUR], PFC), and Hannum (primarily EUR, PBMC)\u003csup\u003e6\u003c/sup\u003e. First, we performed pairwise comparisons of effect sizes for age-CpG association among the three datasets. . As shown in \u003cstrong\u003eFig. 2b\u003c/strong\u003e, we observed that the correlation of effect sizes between the two PFC datasets (CBMAP and ROSMAP) was high (r = 0.89), while the correlation between PFC (ROSMAP) and PBMC (Hannum) was relatively lower (r = 0.71). Further stratified analysis showed that CpG sites located in regulatory regions such as the 1stExon and promoter regions (TSS1500, TSS200, and 5\u0026apos;UTR) exhibited stronger effect size correlations across datasets, with the highest correlation observed in the 1stExon region (r = 0.91) (\u003cstrong\u003eFig. 2c\u003c/strong\u003e). Next, we characterized the top 1,000 age-associated CpG sites in each dataset. As illustrated in the Venn diagram (\u003cstrong\u003eFig. 2d\u003c/strong\u003e), CBMAP and ROSMAP, both PFC datasets, shared 410 of their top 1,000 CpG sites, representing the highest degree of overlap (\u003cstrong\u003eSupplementary Table 3\u003c/strong\u003e). We then performed enrichment analysis for these 410 CpG sites. The potential target genes of these sites were enriched in numerous immune-related Gene Ontology (GO) terms and KEGG pathways (\u003cstrong\u003eExtended Data Fig. 3a,b\u003c/strong\u003e). After excluding the major histocompatibility complex (MHC) region, significant enrichment was observed in GO terms including structural constituent of chromatin (\u003cem\u003eP\u003c/em\u003e-value = 1.57\u0026times;10\u003csup\u003e-27\u003c/sup\u003e) and nucleosome (\u003cem\u003eP\u003c/em\u003e-value = 2.01\u0026times;10\u003csup\u003e-22\u003c/sup\u003e) (\u003cstrong\u003eSupplementary Table 4\u003c/strong\u003e). We also identified CBMAP-specific age-associated CpG sites, including 27 sites among the top 1,000 in CBMAP that did not reach statistical significance (FDR \u0026gt; 0.05) in ROSMAP (\u003cstrong\u003eSupplementary Table 5\u003c/strong\u003e). Notable examples included cg06144905 (located in \u003cem\u003ePIPOX\u003c/em\u003e; CBMAP \u003cem\u003eP\u003c/em\u003e-value = 2.01\u0026times;10\u003csup\u003e-86\u003c/sup\u003e, ROSMAP \u003cem\u003eP\u003c/em\u003e-value = 0.36) and cg09580336 (located in \u003cem\u003eATP1A1\u003c/em\u003e; CBMAP \u003cem\u003eP\u003c/em\u003e-value = 6.95\u0026times;10\u003csup\u003e-81\u003c/sup\u003e, ROSMAP \u003cem\u003eP\u003c/em\u003e-value = 0.17), suggesting ancestry-specific heterogeneity in age-related methylation. Of these 410 CpG sites, 186 did not reach statistical significance (FDR \u0026gt; 0.05) in Hannum (EUR PBMC) (\u003cstrong\u003eSupplementary Table 6\u003c/strong\u003e), suggesting the brain-specificity of age-related DNA methylation. The enrichment analysis results showed that the potential regulatory genes of these 186 brain-specific CpGs were more significantly enriched in GO terms including structural constituent of chromatin (\u003cem\u003eP\u003c/em\u003e-value = 5.07\u0026times;10\u003csup\u003e-37\u003c/sup\u003e) (\u003cstrong\u003eExtended Data Fig. 3c,d\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 7\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eIn summary, we identified a large amount of age-related CpGs which enriched in AD-relevant terms and uncovered the differences in age-related DNA methylation between blood and brain tissues, while showed concordance between EAS and EUR brains.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethylation-Mediated Effects of Aging on Transcriptome Regulation\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo explore the potential regulatory genes and biological functions of age-associated CpG sites, we focused on the 1,000 CpG sites most strongly associated with age (\u003cstrong\u003eFig. 3a\u003c/strong\u003e). The methylation levels of these sites showed strong age associations, with individual site-specific linear models yielding coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e) ranging from 0.713 to 0.931 (\u003cstrong\u003eExtended Data Fig. 4\u003c/strong\u003e). Enrichment analysis revealed that the potential target genes of these CpG sites were significantly enriched in GO terms related to structural constituent of chromatin (\u003cem\u003eP\u003c/em\u003e = 5.39\u0026times;10\u003csup\u003e-18\u003c/sup\u003e) and nucleosome (\u003cem\u003eP\u003c/em\u003e-value = 5.28\u0026times;10\u003csup\u003e-16\u003c/sup\u003e) (\u003cstrong\u003eFig. 3b\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 8\u003c/strong\u003e). These terms encompassed genes encoding histones, including \u003cem\u003eH2AX\u003c/em\u003e, \u003cem\u003eH2BC12, H2BC14, H2AC15, H2BC3, H3C7, H2AC17\u0026nbsp;\u003c/em\u003eand \u003cem\u003eH3C10\u003c/em\u003e. Mediation analysis integrating PFC transcriptomic data demonstrated that most of these CpG sites mediated the effect of age on the expression of histone-encoding genes (\u003cstrong\u003eSupplementary Table 9 and 10\u003c/strong\u003e). For example, the relationship among age, cg03459567, and \u003cem\u003eH2BC12\u003c/em\u003e (\u003cstrong\u003eFig. 3c\u003c/strong\u003e) showed that age was positively correlated with cg03459567 methylation (\u003cem\u003eP\u003c/em\u003e-value = 1.92\u0026times;10\u003csup\u003e-113\u003c/sup\u003e), cg03459567 methylation was negatively correlated with \u003cem\u003eH2BC12\u003c/em\u003e expression (\u003cem\u003eP\u003c/em\u003e-value = 1.53\u0026times;10\u003csup\u003e-4\u003c/sup\u003e), and age was negatively correlated with \u003cem\u003eH2BC12\u003c/em\u003e expression (\u003cem\u003eP\u003c/em\u003e-value = 0.001), with a significant mediating effect (\u003cem\u003eP\u003c/em\u003e-value = 2.65\u0026times;10\u003csup\u003e-4\u003c/sup\u003e). Similarly, nine other CpG sites followed a pattern where increased methylation with age led to reduced expression of histone-encoding genes (\u003cstrong\u003eFig. 3d\u003c/strong\u003e). These findings align with prior studies of declining histone expression with age\u003csup\u003e14,15\u003c/sup\u003e, while our study further demonstrates that, in the PFC, this pattern is largely mediated by DNA methylation. Such methylation-driven reductions in histone expression may impair chromatin stability, promote chromatin remodeling, and contribute to transcriptional dysregulation.\u003c/p\u003e\n\u003cp\u003eBeyond histone genes, we investigated the 186 age-associated CpG sites specifically in the PFC but not in PBMC. Mediation analysis revealed that cg20591728 mediated the negative association between age and \u003cem\u003eNXPH3\u003c/em\u003e expression (\u003cem\u003eP\u003c/em\u003e-value = 0.001) in PFC. Notably, \u003cem\u003eNXPH3\u003c/em\u003e was found to be associated with ADNC (\u003cem\u003e\u0026beta;\u0026nbsp;\u003c/em\u003e= -0.044; \u003cem\u003eP\u003c/em\u003e-value = 0.016) in CBMAP which is consistent with previous study that a lower \u003cem\u003eNXPH3\u003c/em\u003e expression is associated with dysregulated synaptic plasticity and impaired neurotransmitter release\u003csup\u003e16\u003c/sup\u003e. Additionally, cg26092675, an age-associated CpG site specific to the PFC, mediated the age-related increase in \u003cem\u003eBTN3A2\u003c/em\u003e expression (\u003cem\u003eP\u003c/em\u003e-value = 4.18\u0026times;10\u003csup\u003e-4\u003c/sup\u003e). \u003cem\u003eBTN3A2\u003c/em\u003e has been identified as a potential AD risk gene, with functional studies showing that its overexpression suppresses excitatory synaptic transmission in CA1 pyramidal neurons, potentially through molecular interactions with presynaptic neurexins\u003csup\u003e17\u003c/sup\u003e, differential gene expression analysis in CBMAP further supports this finding, showing a positive association between \u003cem\u003eBTN3A2\u003c/em\u003e and ADNC (\u003cem\u003e\u0026beta;\u003c/em\u003e = 0.125; \u003cem\u003eP\u003c/em\u003e-value = 0.020). We also observed that cg26092675 mediated the positive association between age and \u003cem\u003eHFE\u003c/em\u003e (\u003cem\u003eP\u003c/em\u003e-value = 0.027). HFE, a human homeostatic iron regulator protein, studies have shown that higher \u003cem\u003eHFE\u003c/em\u003e expression disrupts iron/cholesterol homeostasis, causing neuronal iron overload, oxidative stress, and synaptic/myelin damage, which synergistically accelerate A\u0026beta;/tau pathology and neuroinflammation in AD\u003csup\u003e18\u003c/sup\u003e. A positive association was observed between \u003cem\u003eHFE\u003c/em\u003e and ADNC (\u003cem\u003e\u0026beta;\u0026nbsp;\u003c/em\u003e= 0.107; \u003cem\u003eP\u003c/em\u003e-value = 0.001) (\u003cstrong\u003eFig. 3e\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 11\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eWe also explored CpG sites exhibiting non-linear age-related methylation trends and grouped the CpGs in four patterns (\u003cstrong\u003eFig. 3f\u003c/strong\u003e). For instance, our analysis revealed an age-dependent association, with cg13084525 methylation showed a significant negative association with age exclusively in individuals aged over 50 years. Mediation analysis indicated that cg13084525 methylation significantly mediated the positive association between age and \u003cem\u003eCD44\u003c/em\u003e expression levels (\u003cem\u003eP\u003c/em\u003e-value = 9.15\u0026times;10\u003csup\u003e-8\u003c/sup\u003e) (\u003cstrong\u003eFig. 3g\u003c/strong\u003e). \u003cem\u003eCD44\u003c/em\u003e plays a critical role in glial cells, contributing to neural development, injury repair, and immune responses under pathological conditions by mediating cell-matrix adhesion, migration, and inflammatory signaling\u003csup\u003e19,20\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo summarize, our findings demonstrate that age-associated CpG sites mediate the effects of aging on gene expression in the PFC, affecting both histone-encoding genes and genes typically implicated in neurodegenerative diseases.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe Epigenetic Clock in the PFC\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBiological age often diverges from chronological age, with DNA methylation age (MethAge) serving as a widely used biomarker of biological aging. To investigate biological aging differences between PFC and PBMC, assess MethAge variations across ancestries, and evaluate the relevance of MethAge in cell type proportion and neurodegenerative diseases, we firstly constructed a CBMAP elastic-net clock model for the Chinese population. Using 501 samples from eastern and central China, we constructed an age prediction model via regularized regression with cross-validation to optimize hyperparameters (Methods). The predictive accuracy was evaluated in 530 samples from northern China, yielding a Pearson\u0026rsquo;s correlation coefficient (r) of 0.95 and a root mean square error (RMSE) of 4.48 (\u003cstrong\u003eFig. 4a\u003c/strong\u003e). We also evaluated the Cortical Clock\u003csup\u003e7\u003c/sup\u003e\u0026mdash;developed using European ancestry cortical samples\u0026mdash;in the CBMAP northern China center, which showed r = 0.94 and RMSE = 6.81, comparable to the multi-tissue Horvath Clock\u003csup\u003e3\u003c/sup\u003e (r = 0.91, RMSE = 8.80). In contrast, a PBMC-based MethAge model derived from Chinese individuals\u003csup\u003e21\u003c/sup\u003e exhibited substantial bias in CBMAP brain tissue (r = 0.81, RMSE = 40.7), as did a European PBMC-based model (r = 0.91, RMSE = 26.23). Both PBMC models consistently underestimated age in older individuals, with greater deviations at higher chronological ages. Given the potential differences in MethAge model accuracy between ADNC cases and controls of CBMAP, we also constructed an ADNC-free MethAge model using samples without ADNC and tested it on samples with ADNC, achieving a correlation of r = 0.84 and an RMSE of 6.85. Using this model, we calculated age acceleration for each sample (Methods). We observed that higher age acceleration was significantly associated with more severe ADNC (\u003cem\u003eP\u003c/em\u003e-value = 0.03) and Braak NFT stage (\u003cem\u003eP\u003c/em\u003e-value = 0.01) (\u003cstrong\u003eFig. 4b\u003c/strong\u003e), consistent with findings by Levine et al. using ROSMAP\u003csup\u003e22\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eWe further investigated changes in cell type proportions as a function of MethAge. Using a reference dataset of over 15,000 single cells from human frontal cortex\u003csup\u003e23,24\u003c/sup\u003e, we applied methylation-based deconvolution to estimate the proportions of seven major cell types (excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, microglia, oligodendrocytes, and oligodendrocyte precursor cells [OPCs]). The overall cell type distributions were similar between CBMAP and ROSMAP, except for excitatory neuron and astrocyte (\u003cstrong\u003eFig. 4c\u003c/strong\u003e). We observed that the proportions of OPCs and oligodendrocytes increased with MethAge, while microglia, endothelial cell, and inhibitory neurons showed decreasing trends (\u003cstrong\u003eFig. 4d,e\u003c/strong\u003e). These patterns were largely consistent with those observed in ROSMAP (\u003cstrong\u003eExtended Data Fig. 5\u003c/strong\u003e). To account for the influence of ADNC status, we performed stratified analyses (\u003cstrong\u003eFig. 4f\u003c/strong\u003e). Overall, the direction and significance of associations were largely consistent between control and ADNC groups. Notably, the effect sizes were generally larger in the ADNC case group, suggesting that ADNC pathology may amplify the impact of epigenetic aging on brain cellular composition, potentially reflecting disease-related shifts in cell-type dynamics with age.\u003c/p\u003e\n\u003cp\u003eConsidering MethAge and ADNC as the founder and primary outcome, respectively, we inferred conditional dependencies among MethAge, neurodegenerative conditions, and ADNC using a Bayesian network analysis framework (Methods). We hypothesized that MethAge influences ADNC risk by modulating three age-related neuropathologies (primary age-related tauopathy [PART], limbic-predominant age-related TDP-43 encephalopathy [LATE], and aging-related tau astrogliopathy [ARTAG]) and cerebrovascular pathologies (including atherosclerosis, arteriosclerosis, cerebral amyloid angiopathy, cerebral hemorrhage and cerebral infarction). Our analysis revealed significant associations between LATE and cerebral amyloid angiopathy (CAA) with ADNC, with consistent effects across MethAge strata (\u003cstrong\u003eExtended Data Fig. 6\u003c/strong\u003e). This suggests that LATE and CAA may substantially increase ADNC risk in an age- and epigenetic-dependent manner. Notably, the effect of CAA on AD neuropathologic changes (ADNC) is unlikely to be disease-specific, as we have previously revealed\u003csup\u003e25\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethylation Signatures of AD Neuropathologic Changes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo characterize the methylation signatures of ADNC, we conducted an epigenome-wide association study (EWAS) in CBMAP samples (n = 1,031). ADNC severity was scored as 0 (Control), 1 (Low), 2 (Intermediate), and 3 (High). In addition to ADNC, we performed EWAS analyses for Braak NFT stage and A\u0026beta; deposits. Given nearly one million CpG sites on the array, multiple testing correction posed significant challenges, resulting in a limited number of sites passing stringent significance thresholds. To enhance statistical power, we performed a meta-analysis combining CBMAP results with those from an identical analysis in ROSMAP (n = 734). To account for potential heterogeneity between studies, we first assessed effect size differences using a heterogeneity test. For CpG sites without heterogeneity, a fixed-effects model was applied, while a random-effects model was used for sites with heterogeneity to identify CpG sites robustly associated with ADNC across both studies. The meta-analysis identified 204, 181, and 266 CpG sites significantly associated with ADNC, Braak NFT stage, and A-score, respectively (FDR \u0026lt; 0.05) (\u003cstrong\u003eSupplementary Table 12\u003c/strong\u003e). These corresponded to 485 unique CpG sites, of which 257 were previously reported in a meta-analysis of four datasets with predominantly European ancestries\u003csup\u003e11\u003c/sup\u003e, and 228 were novel discoveries in this study.\u003c/p\u003e\n\u003cp\u003eAmong the 204 CpG sites associated with ADNC, 171 showed increased methylation with ADNC progression, while 33 exhibited decreased methylation (\u003cstrong\u003eFig. 5a\u003c/strong\u003e). As shown in \u003cstrong\u003eExtended Data Fig. 7a,b\u003c/strong\u003e, the mapped genes of the 204 CpG were primarily enriched in (1) granule membrane and vesicle-related terms (cell periphery, plasma membrane, specific granule membrane, cytoplasmic vesicle), (2) oxidative phosphorylation (NADPH oxidase complex), and (3) immune regulation (positive regulation of CD8-positive, alpha-beta T cell differentiation, immune effector response) (\u003cstrong\u003eSupplementary Table 13\u003c/strong\u003e), as well as MAPK and phagosome pathway of KEGG database (\u003cstrong\u003eSupplementary Table 14\u003c/strong\u003e). The most significantly upregulated CpG site was cg07883124, with \u003cem\u003eP\u003c/em\u003e-values of 2.23\u0026times;10\u003csup\u003e-5\u003c/sup\u003e and 5.95\u0026times;10\u003csup\u003e-8\u003c/sup\u003e in CBMAP and ROSMAP, respectively, and a meta-analysis \u003cem\u003eP\u003c/em\u003e-value of 6.62\u0026times;10\u003csup\u003e-12\u003c/sup\u003e (\u003cstrong\u003eFig. 5b\u003c/strong\u003e). cg07883124 maps to \u003cem\u003eMCF2L\u003c/em\u003e, which encodes a guanine nucleotide exchange factor that interacts with GTP-bound Rac1 and modulates Rho/Rac signaling pathways, implicated in cytoskeletal regulation and cell migration\u003csup\u003e26\u003c/sup\u003e. Among downregulated sites, cg00464927 was one of the most significant, with \u003cem\u003eP\u003c/em\u003e-values of 3.71\u0026times;10\u003csup\u003e-5\u003c/sup\u003e and 6.51\u0026times;10\u003csup\u003e-2\u003c/sup\u003e in CBMAP and ROSMAP, respectively, and a meta-analysis \u003cem\u003eP\u003c/em\u003e-value of 1.89\u0026times;10\u003csup\u003e-5\u003c/sup\u003e (\u003cstrong\u003eFig. 5b\u003c/strong\u003e). This CpG site showed a stronger effect in CBMAP and has not been previously associated with AD, suggesting potential ancestry-specificity. It maps to \u003cem\u003eMRGPRF\u003c/em\u003e, encoding a G protein-coupled receptor primarily expressed in sensory neurons and immune cells\u003csup\u003e27\u003c/sup\u003e, with no reported role in neurodegenerative diseases.\u003c/p\u003e\n\u003cp\u003eNext, we dug into the 181 CpG sites associated with Braak NFT stage (\u003cstrong\u003eFig. 5c,d\u003c/strong\u003e). Their potential target genes were enriched in pathways related to antigen binding, immune response, and inflammation, including peptide antigen binding, immune system process, immune response, and lymphocyte activation (\u003cstrong\u003eExtended Data Fig. 7c,d\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 15\u003c/strong\u003e). Among the 157 significantly upregulated CpG sites, cg07883124 (also associated with ADNC) was the most significant, with a meta-analysis \u003cem\u003eP\u003c/em\u003e-value of 3.11\u0026times;10\u003csup\u003e-10\u003c/sup\u003e. The second most significant site, cg23968456, located in the exonic region of \u003cem\u003eVSIR\u003c/em\u003e, had \u003cem\u003eP\u003c/em\u003e-values of 6.95\u0026times;10\u003csup\u003e-4\u003c/sup\u003e and 1.87\u0026times;10\u003csup\u003e-6\u003c/sup\u003e in CBMAP and ROSMAP, respectively, and a meta-analysis \u003cem\u003eP\u003c/em\u003e-value of 1.80\u0026times;10\u003csup\u003e-8\u003c/sup\u003e. Notably, its methylation level increased markedly in early Braak NFT stages (\u003cstrong\u003eFig. 5d\u003c/strong\u003e). \u003cem\u003eVSIR\u003c/em\u003e is highly expressed in microglia and in mouse inflammation models\u003csup\u003e28\u003c/sup\u003e. \u003cem\u003eVSIR\u003c/em\u003e knockout was found to be associated with elevated pro-inflammatory cytokine levels from T cells and myeloid cells\u003csup\u003e29,30\u003c/sup\u003e. Among the 24 CpG sites negatively associated with Braak NFT stage, cg09470754, located in \u003cem\u003eLAIR1\u003c/em\u003e, was one of the most significant (meta \u003cem\u003eP\u003c/em\u003e-value = 1.13\u0026times;10\u003csup\u003e-6\u003c/sup\u003e). Limited reports link \u003cem\u003eLAIR1\u003c/em\u003e to AD, but its expression has been implicated in microglial states, inflammation, and neuronal injury\u003csup\u003e31\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFor the 266 CpG sites associated with A\u0026beta; deposits (\u003cstrong\u003eFig. 5e,f\u003c/strong\u003e), their potential target genes were enriched in immune response and vesicle-related functions (endocytic vesicle, endocytic vesicle membrane, phagosome) (\u003cstrong\u003eExtended Data Fig. 7e,f\u003c/strong\u003e,\u003cstrong\u003e\u0026nbsp;Supplementary Table 16\u003c/strong\u003e). Among the 143 significantly upregulated sites, cg00601836 was prominent, with \u003cem\u003eP\u003c/em\u003e-values of 5.68\u0026times;10\u003csup\u003e-4\u003c/sup\u003e, 2.10\u0026times;10\u003csup\u003e-5\u003c/sup\u003e, and 5.28\u0026times;10\u003csup\u003e-8\u003c/sup\u003e in CBMAP, ROSMAP, and the meta-analysis, respectively. This site, unreported in prior AD studies, maps to \u003cem\u003eESR1\u003c/em\u003e, encoding estrogen receptor \u0026alpha;. Among sites negatively associated with A\u0026beta; deposits, cg14010720 (\u003cem\u003eSLC17A9\u003c/em\u003e) was one of the most significant, showing a linear decrease in methylation with increasing A\u0026beta; deposits in CBMAP (\u003cem\u003eP\u003c/em\u003e-value = 2.53\u0026times;10\u003csup\u003e-4\u003c/sup\u003e) and ROSMAP (\u003cem\u003eP\u003c/em\u003e-value = 5.85\u0026times;10\u003csup\u003e-6\u003c/sup\u003e). \u003cem\u003eSLC17A9\u003c/em\u003e encodes a transmembrane protein of the solute carrier family 17, mediating vesicular uptake, storage, and secretion of ATP and other nucleotides, and is highly enriched in lysosomes, contributing to cell viability and lysosomal function\u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eSensitivity analyses adjusting for Lewy body disease (LBD) and cerebrovascular disease (CVD) confirmed that the associations of most highlighted CpG sites with ADNC remained significant (\u003cstrong\u003eSupplementary Table 17\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eIn summary, our EWAS identified numerous novel methylation sites associated with AD neuropathological changes, with potential target genes enriched in phagocytosis-related processes (primarily linked to A\u0026beta; deposits), immune regulation (primarily linked to Braak NFT stage), and oxidative phosphorylation. These findings highlight the critical role of DNA methylation in AD pathology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCo-regulation of ADNC-Associated CpG Sites\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate whether ADNC-associated CpG sites exhibit co-expression patterns and to identify potential regulatory pathways linked to ADNC-related functional modules, we performed weighted gene co-expression network analysis (WGCNA)\u003csup\u003e33\u003c/sup\u003e on 23,199 nominally significant ADNC-associated CpG sites. We identified 39 co-regulation modules, with each module containing 33 to 1,073 CpG sites (\u003cstrong\u003eFig. 6a\u003c/strong\u003e). Using the first principal component of each module as a representative feature, regression analysis revealed 11 modules significantly associated with ADNC (\u003cstrong\u003eFig. 6b\u003c/strong\u003e). Most of these modules were also associated with at least one of Braak NFT stage or A\u0026beta; deposits.\u003c/p\u003e\n\u003cp\u003eSubsequent enrichment analysis of the hub CpG sites within these modules identified potential target gene groups (\u003cstrong\u003eFig. 6c\u003c/strong\u003e). Module 1, associated with both Braak NFT stage and A\u0026beta; deposits, was significantly enriched in functions related to secretory granule membrane, dendritic cell migration, and dendritic cell chemotaxis. The top-ranked hub CpG in Module 1 mapped to the potential target gene \u003cem\u003eANK1\u003c/em\u003e. Module 4, linked to A\u0026beta; deposits, was enriched in extracellular matrix (ECM)-related functions (cell periphery, extracellular matrix structural constituent) and potassium ion transmembrane transporter activity. Module 8 was linked to molecular transducer activity, signaling receptor activity, transmembrane signaling receptor activity, and immune effector process, with its potential hub regulator being \u003cem\u003eFOXC2\u003c/em\u003e. Module 18, associated with both ADNC and age, was significantly enriched in chromatin, nucleosome, and protein-DNA complex functions. Module 28 was mapped to neurotransmitter transmembrane transporter activity, GABA-A receptor activity, and GABA-gated chloride ion channel activity (\u003cstrong\u003eSupplementary Table 18\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eOverall, co-methylation analysis identified 11 ADNC-associated modules, encompassing functions related to granule membranes, immune regulation, nucleosome activity, dendritic cell migration, ECM, and neurotransmitter transport. These findings replicated the key functional categories identified in the ADNC enrichment analysis and revealed more specific modules underlying ADNC pathology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eADNC-Specific and ADNC-/Age-Shared CpG Sites\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate the functional similarities and differences between CpG sites associated with age and ADNC, we performed KEGG and GO enrichment analyses (\u003cstrong\u003eExtended Data Fig. 8\u003c/strong\u003e). The potential target genes of CpG sites associated with both ADNC (or Braak NFT stage/A\u0026beta;) and age were predominantly enriched in immune signaling pathways, such as antigen processing and presentation. In contrast, CpG sites associated exclusively with AD neuropathological changes (ADNC level, Braak NFT stage, or A\u0026beta; deposits) were enriched in functions related to the ECM, NADPH oxidase complex, vesicle, calcium ion binding, and the MAPK signaling pathway (\u003cstrong\u003eSupplementary Table 13-16\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eWhile these patterns outline broad functional tendencies, several CpG subsets exhibited distinct enrichment profiles. For instance, cell activation involved in immune responses was enriched among potential target genes of CpG sites associated solely with Braak NFT stage (\u003cstrong\u003eExtended Data Fig. 8e\u003c/strong\u003e). Additionally, CpG sites linked to both age and A\u0026beta; deposits were enriched in the GO term \u0026ldquo;regulation of lamellipodium assembly\u0026rdquo; (\u003cstrong\u003eExtended Data Fig. 8c\u003c/strong\u003e). Lamellipodia are critical precursor structures for dendritic spine formation during synaptic development. Studies in AD model animals have demonstrated impaired lamellipodium formation and reduced dendritic spine density\u003csup\u003e34\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThese results suggest that CpG sites associated with both ADNC and age are primarily linked to immune regulation, whereas those exclusively associated with AD neuropathological changes are more frequently enriched in signaling pathways including ECM, oxidative phosphorylation, and vesicle-related processes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePseudotime trajectories of ADNC-Associated Methylation Sites\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo characterize the age-related trajectories of ADNC-associated CpG sites, we used loess regression to fit pseudotime trajectories and performed unsupervised k-means clustering on these trajectories (\u003cstrong\u003eFig. 7a,b\u003c/strong\u003e). Functional enrichment analysis was conducted for each of the clusters (\u003cstrong\u003eFig. 7c\u003c/strong\u003e). Clusters 1 and 2 showed increasing methylation levels with age. Cluster 1 was enriched in cell adhesion, calcium ion binding, and phagocytosis, while Cluster 2 was enriched in TRAIL binding pathways, where TRAIL, a TNF family cytokine, induces programmed cell death\u003csup\u003e35\u003c/sup\u003e. Cluster 3 exhibited decreasing methylation with age and was enriched in granule membrane and signal transduction pathways. Cluster 4, associated with the MHC region, remained relatively stable with age (\u003cstrong\u003eSupplementary Table 19\u003c/strong\u003e). Similar analyses for Braak NFT stage (\u003cstrong\u003eFig. 7d-f\u003c/strong\u003e) and A\u0026beta; deposits (\u003cstrong\u003eFig. 7g-i\u003c/strong\u003e) revealed that Cluster 2 for Braak NFT stage, with methylation levels positively correlated with age, was enriched in myeloid cell and immune cell differentiation functions, consistent with reports that bone marrow-derived cells may enter the brain and function similarly to microglia\u003csup\u003e36\u003c/sup\u003e (\u003cstrong\u003eSupplementary Table 20\u003c/strong\u003e). For A\u0026beta; deposits, Cluster 1 showed methylation levels positively correlated with age, with the steepest increase around age 75, and was enriched in actin, TNF\u0026alpha;, cell differentiation, and phagocytosis functions (\u003cstrong\u003eSupplementary Table 21\u003c/strong\u003e). These results elucidate the age-dependent patterns of ADNC-associated methylation changes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethylation Mediated the Effect of Aging on ADNC\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess whether methylation mediates the effect of age on ADNC, we performed mediation analysis on the 204 ADNC-associated CpG sites. Of these, 140 (68.6%) showed partial mediation of the age-ADNC relationship, with an average mediation proportion of 8.28% (\u003cstrong\u003eFig. 8a\u003c/strong\u003e). The highest mediation was observed for cg19803550, which mediated 19.7% of age\u0026rsquo;s effect on ADNC (\u003cstrong\u003eFig. 8b,c\u003c/strong\u003e). cg19803550 is located in \u003cem\u003eWDR81\u003c/em\u003e, encoding a multi-domain transmembrane protein\u003csup\u003e37\u003c/sup\u003e. The CpG site most strongly associated with ADNC, cg07883124, mediated 18.4% of age\u0026rsquo;s effect. To evaluate the cumulative mediation effect of these 140 sites, we computed the first 10 principal components (PCs) of their methylation levels (PCs are orthogonal, simplifying cumulative contribution assessment) and performed mediation analysis on each of the PCs. The mediation proportions for these 10 PCs ranged from -4.23% to 33.50%, with a cumulative mediation proportion of 27.27%. Interestingly, some PCs, including PC3, exhibited negative mediation proportions. Analysis of CpG sites with high loading scores in PC3 identified sites with negative mediation effects. For instance, the methylation level of cg19229215 was lower in elder donors (\u003cem\u003e\u0026beta;\u003c/em\u003e \u0026lt; 0, \u003cem\u003eP\u003c/em\u003e-value \u0026lt; 2.2\u0026times;10\u003csup\u003e-16\u003c/sup\u003e) but was positively associated with ADNC (\u003cem\u003e\u0026beta;\u003c/em\u003e \u0026gt; 0, \u003cem\u003eP\u003c/em\u003e-value = 1.2\u0026times;10\u003csup\u003e-3\u003c/sup\u003e), contributing a negative mediation proportion of -5.3%. This site was strongly associated with \u003cem\u003eNCF1\u003c/em\u003e expression (\u003cem\u003e\u0026beta;\u003c/em\u003e \u0026gt; 0, \u003cem\u003eP\u003c/em\u003e-value = 3.1\u0026times;10\u003csup\u003e-7\u003c/sup\u003e), a component of NADPH oxidase linked to reactive oxygen species production and AD-related oxidative stress\u003csup\u003e38\u003c/sup\u003e, suggesting that high \u003cem\u003eNCF1\u003c/em\u003e expression may exacerbate neuronal damage. These results indicate that while methylation predominantly mediates age\u0026rsquo;s positive effect on ADNC, certain CpG sites like cg19229215 may counteract neurodegenerative changes with increasing age.\u003c/p\u003e\n\u003cp\u003eFor Braak NFT stage, 165 of 181 associated CpG sites (91.1%) exhibited significant mediation effects (\u003cstrong\u003eFig. 8d-f\u003c/strong\u003e). The highest mediation was observed for cg09221482, which mediated 15.5% of age\u0026rsquo;s effect on Braak NFT stage and was associated with \u003cem\u003eMAP3K4\u003c/em\u003e expression (\u003cem\u003eP\u003c/em\u003e-value = 6.3\u0026times;10\u003csup\u003e-4\u003c/sup\u003e). \u003cem\u003eMAP3K4\u003c/em\u003e mediated 10.7% of the cg09221482-Braak NFT stage association, suggesting a pathway where increased cg09221482 methylation with age upregulates \u003cem\u003eMAP3K4\u003c/em\u003e expression, leading to higher Braak NFT stages (\u003cem\u003eP\u003c/em\u003e-value = 6.0\u0026times;10\u003csup\u003e-5\u003c/sup\u003e). Other sites, such as cg15821544 (intergenic) and cg05714396 (located in \u003cem\u003eATG10\u003c/em\u003e, encoding an E2 enzyme involved in autophagosome formation\u003csup\u003e39\u003c/sup\u003e), also showed mediation effects. The first 10 PCs of these 165 sites collectively mediated 18.33% of age\u0026rsquo;s effect on Braak NFT stage. For A\u0026beta; deposits, 77 of 266 associated CpG sites (28.9%) showed potential mediation effects, with the highest mediation proportions observed for cg19803550 (\u003cem\u003eWDR81\u003c/em\u003e), cg24647108 (\u003cem\u003eSUN1\u003c/em\u003e), and cg14010720 (\u003cem\u003eSLC17A9\u003c/em\u003e) (\u003cstrong\u003eFig. 8g-i\u003c/strong\u003e). The first 10 PCs of these 77 sites collectively mediated 11.53% of age\u0026rsquo;s effect on A\u0026beta; deposits.\u003c/p\u003e\n\u003cp\u003eIn summary, a substantial proportion of ADNC-associated methylation sites, particularly those linked to Braak NFT stage, mediate age\u0026rsquo;s effect on ADNC. These findings underscore the critical role of methylation in modulating age-related ADNC risk.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe constructed the first large-scale brain methylation atlas for the East Asian population, comprising over 1,000 PFC samples, addressing a critical gap in human brain epigenetic data in EAS. Our study elucidates the landscape of age-related methylation sites in the PFC, highlighting both similarities and differences between brain and blood, as well as between East Asian and European ancestry samples. We identified 228 novel CpG sites associated with ADNC and characterized their potential functional roles. Through integrative analysis of age, methylation, and ADNC, we delineated the age-dependent dynamics of ADNC-associated methylation sites and quantified the proportion of age-ADNC associations mediated by methylation. These findings provide valuable insights into the epigenetic mechanisms underlying brain aging and ADNC.\u003c/p\u003e\n\u003cp\u003eCBMAP is a highly valuable data resource for the study of aging, DNA methylation, and early neuropathological changes. On one hand, compared to ROSMAP, CBMAP offers a larger sample size and a broader age range, making it particularly suited for studying age-related methylation patterns. On the other hand, as CBMAP samples were sourced from a national brain bank, they encompass a diverse range of causes of death and were not specifically selected for cognitive statuses. This allowed inclusion of numerous samples with neurodegenerative changes while in absence of overt cognitive impairment, facilitating the investigation of early-stage molecular features of AD and related pathologies.\u003c/p\u003e\n\u003cp\u003eOur results revealed that age-related methylation effects in the PFC were largely consistent in populations of East Asian and European ancestry, although some ancestry-specific signals were observed. In contrast, substantial differences were observed between age-associated methylation in the brain and blood. Analysis of MethAge confirmed a degree of cross-ancestry conservation in cortical methylation clock models, but marked discrepancies between PBMC and PFC methylation clocks, aligning with findings by Shireby et al.\u003csup\u003e7\u003c/sup\u003e Among age-related methylation sites, those most strongly influenced by age were frequently located near histone-related genes. While histone expression is known to decline with age, our mediation analyses further demonstrated that this downregulation in the PFC is largely driven by DNA methylation. WGCNA analysis specified an ADNC-related co-regulation module featured by histone-related methylation sites, suggesting a potential mechanism whereby age influences methylation levels, which in turn regulate histone gene expression, impacting chromatin stability and remodeling, and contributing to neurodegenerative changes.\u003c/p\u003e\n\u003cp\u003eWe found that ADNC-associated methylation sites were significantly enriched in pathways related to immune response, vesicles trafficking, and granule membranes—recent research foci of AD research. Neuroinflammation in AD has a dual role, while chronic inflammation exacerbates pathology, protective immune functions, such as pathological protein clearance, offer therapeutic benefits\u003csup\u003e40,41\u003c/sup\u003e. Dysfunction in granule membranes and vesicles is implicated in impaired protein clearance, disrupted neurotransmitter transport, and pathological protein propagation, likely playing a critical role in early neuronal dysfunction in AD\u003csup\u003e42,43\u003c/sup\u003e. Our large-scale epigenomic study reinforces the importance of these pathways in early AD neuropathological changes and identifies a set of novel CpG sites. Co-expression module analysis further refined and prioritized functional changes related to immune cell activation, ECM, oxidative phosphorylation, and neurotransmitter transporter activity, which may drive neuroinflammation, Aβ and tau accumulation, mitochondrial dysfunction, and synaptic impairment, ultimately exacerbating neuronal damage and cognitive decline.\u003c/p\u003e\n\u003cp\u003eBeyond well-established AD-related pathways, we uncovered novel evidence, including the identification of cg00601836. Cg00601836 is a methylation site associated with Aβ deposits, mapping to \u003cem\u003eESR1\u003c/em\u003e, which encodes estrogen receptor α (ERα). Beyond its expression in reproductive tissues, ERα is predominantly expressed in the hippocampus and PFC and was found to play a substantial role on memory and cognitive function\u003csup\u003e44\u003c/sup\u003e. Reduced ERα activity may trigger neuroinflammatory pathways, acting as an upstream driver of AD\u003csup\u003e45\u003c/sup\u003e. Notably, genetic polymorphisms near \u003cem\u003eESR1\u003c/em\u003e are associated with increased AD risk, potentially by influencing cholesterol metabolism and promoting Aβ accumulation\u003csup\u003e46\u003c/sup\u003e, complementing our observation that \u003cem\u003eESR1\u003c/em\u003e expression may be regulated by methylation, impacting Aβ deposits. This provides convergent genetic and epigenetic evidence for ESR1’s role in AD pathology.\u003c/p\u003e\n\u003cp\u003eIntegrative analyses of age, methylation, and ADNC revealed that the majority of ADNC-associated methylation effects can be attributed to aging, with DNA methylation mediating 27.3% of the effect of age on ADNC. This proportion may be underestimated, as we only considered CpG sites significantly associated with ADNC in the CBMAP-ROSMAP meta-analysis. While most age-related methylation changes promote ADNC progression, a subset, including a CpG site regulating \u003cem\u003eNCF1\u003c/em\u003e (linked to oxidative stress), exhibited protective effects against neurodegeneration\u003csup\u003e38\u003c/sup\u003e. Further mechanistic exploration is warranted. Additionally, integrative transcriptomic analysis established a pathway whereby increased methylation of cg09221482 with age upregulates \u003cem\u003eMAP3K4\u003c/em\u003e expression, correlating with higher Braak NFT stages. \u003cem\u003eMAP3K4\u003c/em\u003e is associated with glial cell activation, with studies by Shen et al.\u003csup\u003e47\u003c/sup\u003e showing that \u003cem\u003eMAP3K4\u003c/em\u003e interacts with \u003cem\u003eGADD45G\u003c/em\u003e to activate neuroimmune signaling in astrocytes, and others indicating that Fra-1 upregulates Map3k4 transcription to enhance microglial activity during neuroinflammation.\u003c/p\u003e\n\u003cp\u003eKey strengths of this study include CBMAP’s broad age range, enabling robust analysis of age-related methylation changes. Using ADNC as the primary outcome, rather than clinical cognitive impairment, facilitated the identification of early molecular signatures. Integration of methylation and transcriptomic data clarified the mapping of methylation sites to genes, enhancing the accuracy of pathway identification. Comprehensive analyses of age, methylation, and ADNC systematically quantified methylation’s mediation of age-related ADNC risk and its underlying patterns. Limitations include the relatively small number of samples under age 50, though donors in this age group are scarce. This will probably be addressed in CBMAP Phase II as we keep increasing the inclusion. In addition, the ADNC-associated methylation sites we identified require further functional investigation.\u003c/p\u003e\n\u003cp\u003eIn conclusion, our study comprehensively profiled DNA methylation for understanding brain aging and ADNC in an East Asian population, identifying novel CpG sites involved in immune response, vesicle membrane, ECM, and estrogen receptor–related functions and signaling pathways. We revealed that DNA methylation mediates 27.3% of the effect of aging on ADNC, underscoring its pivotal role as a molecular bridge between aging and neurodegeneration.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eSample collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe human brain tissue samples used in this study were obtained from the CBMAP initiative. The first phase of this project was launched by the China Human Brain Bank Consortium, adhering to a unified standard for brain tissue collection and processing. Participating institutions include the Human Brain Tissue Resource Center for Health and Disease at Zhejiang University (ZJU) in eastern China, the National Brain Bank for Development and Function at Peking Union Medical College (PUMC) in the north, and the Xiangya Brain Bank at Central South University (CSU) in central China. Informed consent was obtained either from the donors themselves or their families through voluntary donation agreements, allowing the use of both biological specimens and associated data for research purposes. Ethical approval for this study was granted by the Ethics Committee of Zhejiang University School of Medicine (Approval Nos. 2020-005 and 2024-007).\u003c/p\u003e\n\u003cp\u003eIn Phase I of the CBMAP, we specifically focused on the prefrontal cortex (PFC) region. Anatomically, tissue samples were dissected from the region encompassing Brodmann Area 9 (BA9) within the superior frontal gyrus. We performed DNA methylation profiling on PFC samples from a total of 1,057 donors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNeuropathological diagnosis and inclusion/exclusion criteria\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAs detailed in the CBMAP study profile and our recent comorbidity study\u003csup\u003e25\u003c/sup\u003e, all donor brains underwent standardized postmortem evaluations. Following fixation and coronal sectioning, representative brain regions were systematically sampled for immunohistochemical (IHC) analysis. Uniform assessment criteria were applied across all sites. IHC results were interpreted based on established neuropathological staging systems, including Thal phasing for A\u0026beta; deposits (converted to A-score), Braak neurofibrillary tangle staging for hyperphosphorylated tau (B-score), and the CERAD neuritic plaque score (C-score). The severity of Alzheimer\u0026rsquo;s disease neuropathological changes (ADNC) was classified using the ABC scoring system, which designates overall pathology as Not (control), Low, Intermediate, or High\u003csup\u003e13\u003c/sup\u003e. In addition, diagnoses were also made for Lewy body disease (LBD), cerebrovascular disease (CVD), primary age-related tauopathy (PART), aging-related tau astrogliopathy (ARTAG), and limbic-predominant age-related TDP-43 encephalopathy (LATE).\u003c/p\u003e\n\u003cp\u003eIn the study identifying ADNC-associated CpG sites, we excluded individuals with schizophrenia, amyotrophic lateral sclerosis, epilepsy, as well as those with severe congenital neurodevelopmental abnormalities or brain tumors. For the ADNC control group, in addition to requiring the ADNC classification to be \u0026ldquo;Not\u0026rdquo;, we further excluded samples from individuals with documented cognitive decline, a clinical diagnosis of AD or PD, or a postmortem pathological diagnosis of LBD or ARTAG (Gray matter, Perivascular, or Subpial).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDNA Methylation Measurement and Quality Control\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGenomic DNA was extracted from the CBMAP prefrontal cortex (PFC) samples using a nuclear lysis buffer, which included 10% sodium dodecyl sulfate (SDS) and DNA Extraction Kit. DNA methylation levels were measured for 930,140 CpG sites using the HumanMethylationEPIC v2.0 array. A pilot batch of 68 samples was initially processed, followed by 989 additional samples analyzed on the same platform with identical arrays, totaling 1,057 samples. Quality control (QC) involved sequential filtering for duplicated probes, low-quality probes, SNP-containing probes, probes with excessive missing data, and sex mismatches (\u003cstrong\u003eExtended Data Fig. 1\u003c/strong\u003e). After QC, 907,506 CpG sites from 1,031 samples were retained for downstream analysis. Beta values, representing the proportion of methylation at each CpG site, were calculated and used for subsequent analyses.\u003c/p\u003e\n\u003cp\u003eThe ROSMAP data was accessed from the AD Knowledge Portal (ID: 9603055) and the RADC Research Resource Sharing Hub (ID: 6014). The data process was consistent with CBMAP.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell-type Proportion Estimation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCell-type proportions were estimated using reference-based deconvolution with single-cell methylome sequencing data from Luo et al.\u003csup\u003e23\u003c/sup\u003e, comprising more than 15k cells from human frontal cortex, identifying seven major cell types (excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, microglia, oligodendrocytes, and OPCs). Cell-type-specific markers were selected based on extreme beta values, resulting in 99-983 markers per cell type. Marker validity was confirmed through visualization and consistency checks across reference and bulk samples. The estimation was conducted using the pipeline from Gandal lab (https://github.com/gandallab/brain_CTP_deconv).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAge-Related Methylation and Cross-Tissue, Cross-Ancestry Comparisons\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify age-associated methylation sites, we employed linear regression models with methylation levels at each CpG site as the dependent variable, age (in years) as the independent variable, and sex, sample plate, and cell type proportions as covariates. To facilitate cross-tissue and cross-ancestry comparisons, similar analyses were conducted on 734 samples from ROSMAP (primarily PFC of European ancestry)\u003csup\u003e48\u003c/sup\u003e and 656 samples from the Hannum et al. study\u003csup\u003e6\u003c/sup\u003e (hereafter referred to as \u0026apos;Hannum\u0026apos;, primarily whole blood of European ancestry).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCpG-gene mapping and KEGG/GO/GSEA Enrichment Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePotential target genes regulated by methylation sites were identified in CBMAP samples by analyzing associations between methylation levels and the expression of nearby genes. Gene expression was quantified using RNA sequencing (RNA-seq) of PFC samples. Briefly,\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRNA was extracted from tissue using TRIzol reagent. RNA libraries were prepared using the N406-01 Ribo-off rRNA Depletion Kit (Human/Mouse/Rat, Novogene) to remove ribosomal RNA. Sequencing was performed on the BGI DNBSEQ platform with paired-end 150 bp (PE150) reads. After quality control, raw counts were normalized to transcripts per million (TPM) and log-transformed for analysis. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAssociations between methylation levels and gene expression (considering a window 1 Mb both sides of the CpG) were assessed using linear regression models, adjusting for covariates such as age, sex, and the proportion of neuronal cells. A significance threshold of FDR \u0026lt; 0.05 to identify significant methylation-gene expression pairs.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFunctional enrichment of potential target genes was performed using the R package \u0026ldquo;clusterProfiler\u0026rdquo;. Gene Set Enrichment Analysis (GSEA)\u003csup\u003e49\u003c/sup\u003e, Gene Ontology (GO)\u003csup\u003e50\u003c/sup\u003e terms (Biological Process, Molecular Function, and Cellular Component) and Kyoto Encyclopedia of Genes and Genomes (KEGG)\u003csup\u003e51\u003c/sup\u003e pathways were tested, with significance set at FDR \u0026lt; 0.05. Enrichment analyses were conducted for CpG sites associated with age, ADNC, Braak NFT stage, and A\u0026beta; deposits.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMediation Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate whether methylation mediates the effects of age on gene expression, age on ADNC we conducted mediation analyses using the R package \u0026ldquo;bruceR\u0026rdquo; and adjusted for sex and proportion of neuronal cells. For assessing the overall mediation effect of methylation on the age-ADNC relationship, we performed principal component analysis (PCA) on all methylation sites mediating age-ADNC effects, retaining the top 10 principal components (PCs). Mediation analysis was conducted for each PC to estimate its contribution to the age-ADNC relationship. The cumulative mediation proportion was calculated by summing the mediation effects of significant PCs. Mediation analyses used a bootstrap resampling (1,000 simulation samples) to estimate mediation proportions and confidence intervals. The mediation analysis was also applied to estimate the role of gene expression on methylation-ADNC association.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethylation Clock and Age Acceleration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA PFC-specific methylation age (MethAge) model was developed using 501 samples from eastern and central China via regularized regression (elastic net) and a combination of principal component analysis and regularized regression, with 10-fold cross-validation to optimize hyperparameters. The model was validated in 530 samples from northern China, with prediction accuracy assessed using Pearson\u0026rsquo;s correlation coefficient (r) and root mean square error (RMSE). The Cortical clock\u003csup\u003e7\u003c/sup\u003e and PCBrainAge\u003csup\u003e52\u003c/sup\u003e based on cortical tissue of European ancestry, MultiClock\u003csup\u003e3\u003c/sup\u003e based on multiple tissues of European ancestry, Hannum Clock\u003csup\u003e6\u003c/sup\u003e, BloodClock\u003csup\u003e53\u003c/sup\u003e and PhenoClock\u003csup\u003e54\u003c/sup\u003e based on blood tissue, and ICAS-DNAmAge\u003csup\u003e21\u003c/sup\u003e model from blood tissue of Chinese ancestry were also evaluated in the CBMAP for cross-ancestry and cross-tissue comparisons. Age acceleration was quantified using a ratio-based method, whereby the residual from a regression of predicted age on chronological age was divided by the chronological age to account for individual differences in age scale.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBayesian Network Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo infer conditional dependencies between MethAge, ADNC, and neuropathological variables (e.g., PART, LATE, ARTAG, cerebrovascular pathologies), we employed a Bayesian network framework using the R package \u0026ldquo;bnlearn\u0026rdquo;\u003csup\u003e55\u003c/sup\u003e. A hill-climbing algorithm\u003csup\u003e56\u003c/sup\u003e was used to learn the network structure, with constraints to ensure ADNC as the primary outcome and MethAge as an upstream variable. We compiled the Bayesian network into a concatenated tree using the \u0026ldquo;gRain\u0026rdquo; package and recomputed the conditional distribution of ADNC given its parent nodes in the Bayesian network. For ADNC, we estimated conditional probability tables (CPTs) using maximum likelihood estimation (MLE) based on the observed frequency of variable configurations in the data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIdentification of ADNC-Associated Methylation Sites and Meta-Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEpigenome-wide association studies (EWAS) were conducted for ADNC, Braak NFT stage, and A\u0026beta; deposits using linear regression models, with methylation levels as the dependent variable and ADNC scores (0-3), Braak NFT stage (0-6), or A\u0026beta; scores (0-3) as independent variables, adjusted for age, sex, sample plate and the proportion of neuronal cells. Multiple testing correction was applied using FDR \u0026lt; 0.05. To increase statistical power, a meta-analysis was performed combining CBMAP and ROSMAP results. Heterogeneity was assessed using Cochran\u0026rsquo;s Q test\u003csup\u003e57\u003c/sup\u003e; fixed-effects models\u003csup\u003e58\u003c/sup\u003e were used for CpG sites without heterogeneity (\u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05), and random-effects models\u003csup\u003e59\u003c/sup\u003e were applied otherwise. The R package \u0026ldquo;meta\u0026rdquo; was used for meta-analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWGCNA Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWeighted gene co-expression network analysis (WGCNA) was performed using the R package WGCNA on 23,199 nominally significant ADNC-associated CpG sites (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05). A soft-thresholding power was selected to achieve a scale-free topology (R\u0026sup2; \u0026gt; 0.8). Modules were identified using dynamic tree cutting, and module eigengenes (first principal components) were correlated with ADNC, Braak NFT stage, and A\u0026beta; scores via regression analysis. Hub CpG sites were identified based on high module membership, and their potential target genes were subjected to KEGG/GO enrichment analysis as described above.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePseudotime Trajectory Fitting and Unsupervised Hierarchical Clustering\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo model age-related trajectories of ADNC-associated methylation sites, we fitted pseudotime trajectories using loess regression with methylation levels as the dependent variable and age as the independent variable. Trajectories were clustered using the k-means method based on their curve shapes, implemented in the R package \u0026quot;stats\u0026quot;. Clusters were functionally annotated by performing KEGG/GO enrichment analysis on the potential target genes of CpG sites within each cluster, as described above.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u0026nbsp; \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe omics and metadata of CBMAP can be accessed in the OMIX database (https://ngdc.cncb.ac.cn/omix/) under the accession number OMIX010874.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe codes for data analysis are available at https://github.com/zdangm/CBMAP_methylation_aging_AD/\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are deeply grateful to all brain tissue donors and pay our heartfelt tribute to them and their families. We thank the National Health and Disease Human Brain Tissue Resource Center (Zhejiang University), the National Human Brain Bank for Development and Function (Peking Union Medical College), and the Xiangya Brain Bank (Central South University) for providing the brain tissue samples used in this project.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe project was supported by the Science Innovation 2030 - Brain Science and Brain-Inspired Intelligence Technology Major Project #2021ZD0201100 (Brain Tissue Resource Repository and Brain Bank Collaboration Network Platform, tasks 2 [2021ZD0201102] and 6 [2021ZD0201106]) from the Ministry of Science and Technology (MOST) of China, the National Natural Science Foundation of China (82020108012 [J.Z.], 82204118 [D.Z.], 82022024 [C.C.], 32270656 [D.Z.], and 82370612 [Z.S.]), the Key R\u0026amp;D Program of Zhejiang (2024C03098, [J.Z.]), and the CAMS Innovation Fund for Medical Sciences (CIFMS) #2021-1-I2M-025 [C.M., W.Q.].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThese authors contributed equally:\u0026nbsp;Dan Zhou, Wenli Zhai, Wenwei Fang.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eContributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJ.Z., D.S., B.A., C.M., and D.Z. conceived the study. W.Z., W.F., Y.Z., Z.L., C.Y. were involved in study design. W.Z. and W.F. processed DNA methylation data and performed the data analyses. D.Z., W.Z. and W.F. wrote the first draft. J.Z., Y.Z., D.D.Z., S.D., Z.L, L.W., C.Y., Y.H., H.L., K.Z., Y.S., L.W., X.Y., A.B., W.Q., and C.M. commented and revised the manuscript. All authors reviewed and agreed to submit the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eQuerfurth, H. W. \u0026amp; LaFerla, F. M. Alzheimer\u0026rsquo;s Disease. \u003cem\u003eNew England Journal of Medicine\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, 329\u0026ndash;344 (2010).\u003c/li\u003e\n\u003cli\u003eHou, Y. \u003cem\u003eet al.\u003c/em\u003e Ageing as a risk factor for neurodegenerative disease. \u003cem\u003eNat Rev Neurol\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 565\u0026ndash;581 (2019).\u003c/li\u003e\n\u003cli\u003eHorvath, S. DNA methylation age of human tissues and cell types. \u003cem\u003eGenome Biology\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 3156 (2013).\u003c/li\u003e\n\u003cli\u003eBooth, L. N. \u0026amp; Brunet, A. The Aging Epigenome. \u003cem\u003eMol Cell\u003c/em\u003e \u003cstrong\u003e62\u003c/strong\u003e, 728\u0026ndash;744 (2016).\u003c/li\u003e\n\u003cli\u003eRaddatz, G. \u003cem\u003eet al.\u003c/em\u003e Aging is associated with highly defined epigenetic changes in the human epidermis. \u003cem\u003eEpigenetics \u0026amp; Chromatin\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 36 (2013).\u003c/li\u003e\n\u003cli\u003eHannum, G. \u003cem\u003eet al.\u003c/em\u003e Genome-wide methylation profiles reveal quantitative views of human aging rates. \u003cem\u003eMol Cell\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 359\u0026ndash;367 (2013).\u003c/li\u003e\n\u003cli\u003eShireby, G. L. \u003cem\u003eet al.\u003c/em\u003e Recalibrating the epigenetic clock: implications for assessing biological age in the human cortex. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e143\u003c/strong\u003e, 3763\u0026ndash;3775 (2020).\u003c/li\u003e\n\u003cli\u003eHorvath, S. \u003cem\u003eet al.\u003c/em\u003e Aging effects on DNA methylation modules in human brain and blood tissue. \u003cem\u003eGenome Biology\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, R97 (2012).\u003c/li\u003e\n\u003cli\u003eDe Jager, P. L. \u003cem\u003eet al.\u003c/em\u003e Alzheimer\u0026rsquo;s disease: early alterations in brain DNA methylation at ANK1, BIN1, RHBDF2 and other loci. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 1156\u0026ndash;1163 (2014).\u003c/li\u003e\n\u003cli\u003eWatson, C. T. \u003cem\u003eet al.\u003c/em\u003e Genome-wide DNA methylation profiling in the superior temporal gyrus reveals epigenetic signatures associated with Alzheimer\u0026rsquo;s disease. \u003cem\u003eGenome Medicine\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 5 (2016).\u003c/li\u003e\n\u003cli\u003eZhang, L. \u003cem\u003eet al.\u003c/em\u003e Epigenome-wide meta-analysis of DNA methylation differences in prefrontal cortex implicates the immune processes in Alzheimer\u0026rsquo;s disease. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 6114 (2020).\u003c/li\u003e\n\u003cli\u003eZhou, D. \u003cem\u003eet al.\u003c/em\u003e The China Brain Multi-omics Atlas Project (CBMAP). \u003cem\u003ebioRxiv\u003c/em\u003e (2025) doi:10.1101/2025.04.30.651445.\u003c/li\u003e\n\u003cli\u003eMontine, T. J. \u003cem\u003eet al.\u003c/em\u003e National Institute on Aging-Alzheimer\u0026rsquo;s Association guidelines for the neuropathologic assessment of Alzheimer\u0026rsquo;s disease: a practical approach. \u003cem\u003eActa Neuropathol\u003c/em\u003e \u003cstrong\u003e123\u003c/strong\u003e, 1\u0026ndash;11 (2012).\u003c/li\u003e\n\u003cli\u003eYi, S.-J. \u0026amp; Kim, K. New Insights into the Role of Histone Changes in Aging. \u003cem\u003eInt J Mol Sci\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 8241 (2020).\u003c/li\u003e\n\u003cli\u003eWang, K. \u003cem\u003eet al.\u003c/em\u003e Epigenetic regulation of aging: implications for interventions of aging and diseases. \u003cem\u003eSig Transduct Target Ther\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 374 (2022).\u003c/li\u003e\n\u003cli\u003eBeglopoulos, V. \u003cem\u003eet al.\u003c/em\u003e Neurexophilin 3 Is Highly Localized in Cortical and Cerebellar Regions and Is Functionally Important for Sensorimotor Gating and Motor Coordination. \u003cem\u003eMol Cell Biol\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 7278\u0026ndash;7288 (2005).\u003c/li\u003e\n\u003cli\u003eH, Y. \u003cem\u003eet al.\u003c/em\u003e Mendelian randomization integrating GWAS and eQTL data revealed genes pleiotropically associated with major depressive disorder. \u003cem\u003eTranslational psychiatry\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, (2021).\u003c/li\u003e\n\u003cli\u003eAli-Rahmani, F., Schengrund, C.-L. \u0026amp; Connor, J. R. HFE gene variants, iron, and lipids: a novel connection in Alzheimer\u0026rsquo;s disease. \u003cem\u003eFront. Pharmacol.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, (2014).\u003c/li\u003e\n\u003cli\u003eDzwonek, J. \u0026amp; Wilczynski, G. M. CD44: molecular interactions, signaling and functions in the nervous system. \u003cem\u003eFrontiers in Cellular Neuroscience\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 175 (2015).\u003c/li\u003e\n\u003cli\u003eJordan, A. R., Racine, R. R., Hennig, M. J. P. \u0026amp; Lokeshwar, V. B. The Role of CD44 in Disease Pathophysiology and Targeted Treatment. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eZheng, Z. \u003cem\u003eet al.\u003c/em\u003e DNA methylation clocks for estimating biological age in Chinese cohorts. \u003cem\u003eProtein Cell\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 575\u0026ndash;593 (2024).\u003c/li\u003e\n\u003cli\u003eLevine, M. E., Lu, A. T., Bennett, D. A. \u0026amp; Horvath, S. Epigenetic age of the pre-frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer\u0026rsquo;s disease related cognitive functioning. \u003cem\u003eAging (Albany NY)\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 1198\u0026ndash;1211 (2015).\u003c/li\u003e\n\u003cli\u003eLuo, C. \u003cem\u003eet al.\u003c/em\u003e Single nucleus multi-omics identifies human cortical cell regulatory genome diversity. \u003cem\u003eCell Genom\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 100107, 100106 (2022).\u003c/li\u003e\n\u003cli\u003eYap, C. X. \u003cem\u003eet al.\u003c/em\u003e Brain cell-type shifts in Alzheimer\u0026rsquo;s disease, autism, and schizophrenia interrogated using methylomics and genetics. \u003cem\u003eSci Adv\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, eadn7655.\u003c/li\u003e\n\u003cli\u003eWang, X. \u003cem\u003eet al.\u003c/em\u003e Prevalence of mixed neuropathologies in age-related neurodegenerative diseases: A community-based autopsy study in China. \u003cem\u003eAlzheimers Dement\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, e14369 (2025).\u003c/li\u003e\n\u003cli\u003eFitzpatrick, E. R., Hu ,Tinghui, Ciccarelli ,Bryan T \u0026amp; and Whitehead, I. P. Regulation of vesicle transport and cell motility by Golgi-localized Dbs. \u003cem\u003eSmall GTPases\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, e972860 (2014).\u003c/li\u003e\n\u003cli\u003eGour, N. \u0026amp; Dong, X. The MRGPR family of receptors in immunity. \u003cem\u003eImmunity\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 28\u0026ndash;39 (2024).\u003c/li\u003e\n\u003cli\u003eBorggrewe, M. \u003cem\u003eet al.\u003c/em\u003e VISTA expression by microglia decreases during inflammation and is differentially regulated in CNS diseases. \u003cem\u003eGlia\u003c/em\u003e \u003cstrong\u003e66\u003c/strong\u003e, 2645\u0026ndash;2658 (2018).\u003c/li\u003e\n\u003cli\u003eCeeraz, S. \u003cem\u003eet al.\u003c/em\u003e VISTA Deficiency Accelerates the Development of Fatal Murine Lupus Nephritis. \u003cem\u003eArthritis Rheumatol\u003c/em\u003e \u003cstrong\u003e69\u003c/strong\u003e, 814\u0026ndash;825 (2017).\u003c/li\u003e\n\u003cli\u003eLi, N. \u003cem\u003eet al.\u003c/em\u003e Immune-checkpoint protein VISTA critically regulates the IL-23/IL-17 inflammatory axis. \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 1485 (2017).\u003c/li\u003e\n\u003cli\u003eCarroll, K. R. \u003cem\u003eet al.\u003c/em\u003e Lupus autoantibodies initiate neuroinflammation sustained by continuous HMGB1:RAGE signaling and reversed by increased LAIR-1 expression. \u003cem\u003eNat Immunol\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 671\u0026ndash;681 (2024).\u003c/li\u003e\n\u003cli\u003eAyka, A. \u0026amp; Şehirli, A. \u0026Ouml;. The Role of the SLC Transporters Protein in the Neurodegenerative Disorders. \u003cem\u003eClin Psychopharmacol Neurosci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 174\u0026ndash;187 (2020).\u003c/li\u003e\n\u003cli\u003eZhang, B. \u0026amp; Horvath, S. A general framework for weighted gene co-expression network analysis. \u003cem\u003eStat Appl Genet Mol Biol\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, Article17 (2005).\u003c/li\u003e\n\u003cli\u003eKommaddi, R. P. \u003cem\u003eet al.\u003c/em\u003e A\u0026beta; mediates F-actin disassembly in dendritic spines leading to cognitive deficits in Alzheimer\u0026rsquo;s disease. \u003cem\u003eJ Neurosci\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1085\u0026ndash;1099 (2018).\u003c/li\u003e\n\u003cli\u003eWang, S. \u0026amp; El-Deiry, W. S. TRAIL and apoptosis induction by TNF-family death receptors. \u003cem\u003eOncogene\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 8628\u0026ndash;8633 (2003).\u003c/li\u003e\n\u003cli\u003eSimard, A. R., Soulet, D., Gowing, G., Julien, J.-P. \u0026amp; Rivest, S. Bone marrow-derived microglia play a critical role in restricting senile plaque formation in Alzheimer\u0026rsquo;s disease. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 489\u0026ndash;502 (2006).\u003c/li\u003e\n\u003cli\u003eYal\u0026ccedil;ın, H. Y. \u003cem\u003eet al.\u003c/em\u003e Expression of Autophagy-Related Proteins in Microlissencephaly Associated with a Novel Variant in the WDR81 Gene. \u003cem\u003eMol Syndromol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 1\u0026ndash;10 (2025).\u003c/li\u003e\n\u003cli\u003eMarie-Laure Pinel-Marie, Sparfel, L., Desmots, S. \u0026amp; Fardel, O. Aryl hydrocarbon receptor-dependent induction of the NADPH oxidase subunit NCF1/p47 phox expression leading to priming of human macrophage oxidative burst. \u003cem\u003eFree Radic Biol Med\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, 825\u0026ndash;834 (2009).\u003c/li\u003e\n\u003cli\u003eHong, S. B., Kim, B.-W., Kim, J. H. \u0026amp; Song, H. K. Structure of the autophagic E2 enzyme Atg10. \u003cem\u003eActa Crystallogr D Biol Crystallogr\u003c/em\u003e \u003cstrong\u003e68\u003c/strong\u003e, 1409\u0026ndash;1417 (2012).\u003c/li\u003e\n\u003cli\u003eKiraly, M., Foss, J. F. \u0026amp; Giordano, T. Neuroinflammation, Its Role in Alzheimer\u0026rsquo;s Disease and Therapeutic Strategies. \u003cem\u003eJ Prev Alzheimers Dis\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 686\u0026ndash;698 (2023).\u003c/li\u003e\n\u003cli\u003eAmelimojarad, M., Amelimojarad, M. \u0026amp; Cui, X. The emerging role of brain neuroinflammatory responses in Alzheimer\u0026rsquo;s disease. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eBlackstone, C., Elwood, F., Plun-Favreau, H. \u0026amp; Lewis, P. A. Vesicle trafficking and pathways to neurodegeneration. \u003cem\u003eMolecular Neurodegeneration\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 56 (2021).\u003c/li\u003e\n\u003cli\u003eXue, M., Cao, Y., Shen, C. \u0026amp; Guo, W. Computational Advances of Protein/Neurotransmitter-membrane Interactions Involved in Vesicle Fusion and Neurotransmitter Release. \u003cem\u003eJ Mol Biol\u003c/em\u003e \u003cstrong\u003e435\u003c/strong\u003e, 167818 (2023).\u003c/li\u003e\n\u003cli\u003eAlmey, A., Milner, T. A. \u0026amp; Brake, W. G. Estrogen receptors in the central nervous system and their implication for dopamine-dependent cognition in females. \u003cem\u003eHorm Behav\u003c/em\u003e \u003cstrong\u003e74\u003c/strong\u003e, 125\u0026ndash;138 (2015).\u003c/li\u003e\n\u003cli\u003eLiu, J., Yuan, S., Niu, X., Kelleher, R. \u0026amp; Sheridan, H. ESR1 dysfunction triggers neuroinflammation as a critical upstream causative factor of the Alzheimer\u0026rsquo;s disease process. \u003cem\u003eAging (Albany NY)\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 8595\u0026ndash;8614 (2022).\u003c/li\u003e\n\u003cli\u003eCheng, D., Liang, B., Hao, Y. \u0026amp; Zhou, W. Estrogen receptor \u0026alpha; gene polymorphisms and risk of Alzheimer\u0026rsquo;s disease: evidence from a meta-analysis. \u003cem\u003eClin Interv Aging\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1031\u0026ndash;1038 (2014).\u003c/li\u003e\n\u003cli\u003eShen, T. \u003cem\u003eet al.\u003c/em\u003e GADD45G operates as a pathological sensor orchestrating reactive gliosis and neurodegeneration. \u003cem\u003eNeuron\u003c/em\u003e S0896-6273(25)00345\u0026ndash;9 (2025) doi:10.1016/j.neuron.2025.04.033.\u003c/li\u003e\n\u003cli\u003eHuo, Z. \u003cem\u003eet al.\u003c/em\u003e DNA methylation variability in Alzheimer\u0026rsquo;s disease. \u003cem\u003eNeurobiol Aging\u003c/em\u003e \u003cstrong\u003e76\u003c/strong\u003e, 35\u0026ndash;44 (2019).\u003c/li\u003e\n\u003cli\u003eSubramanian, A. \u003cem\u003eet al.\u003c/em\u003e Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, 15545\u0026ndash;15550 (2005).\u003c/li\u003e\n\u003cli\u003eAshburner, M. \u003cem\u003eet al.\u003c/em\u003e Gene Ontology: tool for the unification of biology. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 25\u0026ndash;29 (2000).\u003c/li\u003e\n\u003cli\u003eKanehisa, M. \u0026amp; Goto, S. KEGG: kyoto encyclopedia of genes and genomes. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 27\u0026ndash;30 (2000).\u003c/li\u003e\n\u003cli\u003eThrush, K. L. \u003cem\u003eet al.\u003c/em\u003e Aging the brain: multi-region methylation principal component based clock in the context of Alzheimer\u0026rsquo;s disease. \u003cem\u003eAging (Albany NY)\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 5641\u0026ndash;5668 (2022).\u003c/li\u003e\n\u003cli\u003eZhang, Q. \u003cem\u003eet al.\u003c/em\u003e Improved precision of epigenetic clock estimates across tissues and its implication for biological ageing. \u003cem\u003eGenome Med\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 54 (2019).\u003c/li\u003e\n\u003cli\u003eLevine, M. E. \u003cem\u003eet al.\u003c/em\u003e An epigenetic biomarker of aging for lifespan and healthspan. \u003cem\u003eAging (Albany NY)\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 573\u0026ndash;591 (2018).\u003c/li\u003e\n\u003cli\u003eNagarajan, R., Scutari, M. \u0026amp; L\u0026egrave;bre, S. \u003cem\u003eBayesian Networks in R: With Applications in Systems Biology\u003c/em\u003e. (Springer, New York, NY, 2013). doi:10.1007/978-1-4614-6446-4.\u003c/li\u003e\n\u003cli\u003eTsamardinos, I., Brown, L. E. \u0026amp; Aliferis, C. F. The max-min hill-climbing Bayesian network structure learning algorithm. \u003cem\u003eMach Learn\u003c/em\u003e \u003cstrong\u003e65\u003c/strong\u003e, 31\u0026ndash;78 (2006).\u003c/li\u003e\n\u003cli\u003eHiggins, J. P. T. \u0026amp; Thompson, S. G. Quantifying heterogeneity in a meta-analysis. \u003cem\u003eStat Med\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, 1539\u0026ndash;1558 (2002).\u003c/li\u003e\n\u003cli\u003eRice, K., Higgins, J. P. T. \u0026amp; Lumley, T. A Re-Evaluation of Fixed Effect(s) Meta-Analysis. \u003cem\u003eJ R Stat Soc Ser A Stat Soc\u003c/em\u003e \u003cstrong\u003e181\u003c/strong\u003e, 205\u0026ndash;227 (2018).\u003c/li\u003e\n\u003cli\u003eRiley, R. D., Higgins, J. P. T. \u0026amp; Deeks, J. J. Interpretation of random effects meta-analyses. \u003cem\u003eBMJ\u003c/em\u003e\u003cstrong\u003e342\u003c/strong\u003e, d549 (2011).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7381745/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7381745/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe epigenetic mechanisms that connect aging to neurodegeneration remain incompletely understood, particularly in non-European populations. To address this gap, we constructed a comprehensive DNA methylation atlas of the human prefrontal cortex (PFC) from 1,057 postmortem brain samples (ages 1-106) as part of the China Brain Multi-Omics Atlas Project (CBMAP). This represents the largest DNA methylation dataset of East Asian brain tissue to date and provides a critical resource for understanding aging and Alzheimer’s disease (AD) in underrepresented populations.\u003c/p\u003e\n\u003cp\u003eWe found that over one-third of CpG sites exhibit significant age-associated methylation changes, with high concordance between Asian and European ancestries, but marked divergence between brain and peripheral blood methylation profiles. In the PFC, age-related CpGs were enriched near histone genes and associated with transcriptomic signatures of neuronal decline. Importantly, we identified 485 CpGs linked to AD neuropathologic change (ADNC), including 228 novel loci enriched in immune regulation, vesicle trafficking, and extracellular matrix remodeling—hallmark pathways of AD.\u003c/p\u003e\n\u003cp\u003eMediation analysis revealed that DNA methylation accounts for 27.3% of aging’s effect on ADNC, highlighting a key mechanistic link. Notably, CpG site cg09221482 mediates the relationship between aging and neurofibrillary tangle severity via \u003cem\u003eMAP3K4\u003c/em\u003e expression. Together, these findings uncover novel aging-associated epigenetic signatures that contribute to AD pathology and establish DNA methylation as a critical intermediary bridging aging and neurodegeneration, particularly in East Asian populations.\u003c/p\u003e","manuscriptTitle":"Epigenetic Rewiring Connects Aging to Alzheimer’s Pathology in the Human Prefrontal Cortex","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-29 10:59:27","doi":"10.21203/rs.3.rs-7381745/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-aging","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"nataging","sideBox":"Learn more about [Nature Aging](https://www.nature.com/nataging/)","snPcode":"","submissionUrl":"","title":"Nature Aging","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"64cd782b-d9ac-4009-b5f4-879cf30e20a6","owner":[],"postedDate":"August 29th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":53850616,"name":"Biological sciences/Genetics/Epigenomics"},{"id":53850617,"name":"Health sciences/Diseases/Neurological disorders/Neurodegenerative diseases/Alzheimer's disease"}],"tags":[],"updatedAt":"2026-04-15T08:21:01+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-29 10:59:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7381745","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7381745","identity":"rs-7381745","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}