Abstract
The biological mechanisms underlying the increased prevalence of Alzheimer’s disease (AD) in women remain undefined. While previous case/control studies have identified sex-biased molecular pathways, the sex-specific relationships between gene expression and AD endophenotypes, particularly involving sex chromosomes, are underexplored. With bulk transcriptomic data across 3 brain regions from 767 decedents, we investigated sex-specific associations between gene expression and post-mortem β-amyloid and tau, as well as antemortem longitudinal cognition. Among 23,118 significant gene associations, 10% were sex-specific, with 73% of these identified in females and primarily associated with tau tangles and longitudinal cognition (90%). Notably, four X-linked genes, MCF2, HDAC8, FTX, and SLC10A3, demonstrated significant sex differences in their associations with AD endophenotypes (i.e., significant sex x gene interaction). Our results also uncovered sex-specific biological pathways, including a female-specific role of neuroinflammation and neuronal development, underscoring the importance of sex-aware analyses to advance precision medicine approaches in AD.
Similar content being viewed by others
Introduction
Women are disproportionately affected by Alzheimer’s disease (AD); two-thirds of all clinical cases of AD dementia in the United States are women1. Women exhibit greater vulnerability to AD neuropathology, with each additional unit of pathology increasing the odds of progressing to clinical AD by approximately 22-fold in women relative to 3-fold in men2. Women also experience more rapid cognitive decline in the presence of AD pathology3 and present with higher levels of AD pathology at autopsy, even among those deemed clinically normal prior to death2,4. However, the molecular factors that underlie these sex disparities remain poorly understood.
Sex hormones and the effects of menopause have been linked to some of the sex differences in AD. Previous studies have also identified sex-specific effects of genes known to be involved in AD such as APOE3,5 and BDNF6 (among others). New research on sex differences in AD continues to emerge, focusing on molecular mechanisms derived through large-scale human omics analyses. Genetic variants exhibiting sex-specific effects on AD on cerebrospinal fluid (CSF) amyloid-beta (Aβ) and tau have been identified in genome-wide association studies (GWAS)7. Transcriptome-wide studies have shown significant differential gene expression between the sexes in postmortem AD brains8,9,10,11, and highlighted sex-biased molecular pathways, including neuroinflammation and bioenergetic metabolism10. Others have identified differentially expressed genes between males and females with increasing resolution in specific cell types such as microglia and astrocytes12,13. However, gaps in knowledge remain. Most transcriptomic studies of sex differences in AD have largely focused on AD case vs. control, and few studies have explored sex-specific relationships between gene expression and the core AD neuropathologies, β-amyloid (Aβ) plaques and tau neurofibrillary tangles, or the clinical presentation of the disease outside of case/control analyses.
In addition, these studies have frequently excluded sex chromosomes due, in part, to technical considerations resulting from pseudo-autosomal regions or X inactivation, as examples14 resulting in few AD-related sex chromosome findings. Given the etiologic complexity of AD, targeted studies on sex chromosomes in AD offer an invaluable opportunity to better understand the underlying, and potentially sex-specific, biological pathways driving disease.
Leveraging bulk RNA sequencing (RNAseq) data from over 750 sex-matched decedents (spanning 1490 samples from the dorsolateral prefrontal cortex, posterior cingulate, and caudate nucleus) enrolled in the Religious Orders Study/Memory and Aging Project (ROS/MAP)15,16,17, we performed the largest and most comprehensive study of sex-specific transcriptomic associations with AD endophenotypes (i.e., Aβ plaques, tau tangles, and longitudinal cognition). By taking a targeted approach to analyze sex chromosomes, we captured both autosomal and sex chromosome-linked genes that exhibit sex-specific effects in AD.
Results
Study cohort characteristics
This study leveraged brain tissue samples taken at autopsy from ROS/MAP participants. Given the large sample size difference between males (n = 328) and females (n = 640), the sexes were propensity score matched using age of death, post-mortem interval (PMI), education, latency to death (i.e., difference between age at death and age at last cognitive visit), race, and APOE-ε4 allele count as criteria (see Methods) for each brain region: the dorsolateral prefrontal cortex (DLPFC), posterior cingulate cortex (PCC), and head of the caudate nucleus (CN). The final analytical dataset included 1490 samples across 767 unique decedents, whose characteristics are presented by sex and brain region in Table 1. Overall, the mean age at death of the entire sample was 88.3 years (Standard Deviation (SD) = 6.5, Range = 67.3–108.3), and the average number of annual follow-up visits prior to death was 7.4 (SD = 4.9, Range = 0–23) (Supplementary Table 1). Most participants self-identified as non-Hispanic White (98%), and 26% of participants carried at least one copy of APOE-ε4. There were no statistically significant sex differences in criteria used for matching across any brain region (p > 0.05), and no significant sex difference in incidence of mild cognitive impairment (MCI)/AD (p > 0.05, Supplementary Fig. 1). Notably, females had greater tau tangle burden in comparison to males in all brain regions, following previous reports (p < 0.05, Table 1)5,18,19.
Sex-specific gene expression signatures of AD neuropathology and clinical presentation
To identify sex-specific gene expression signatures of AD, we conducted both sex-stratified (i.e., male and female analyses separately) and sex x gene interaction analyses. We examined brain gene expression associations with continuous immunohistochemical measures of Aβ and tau burden at autopsy and with a longitudinal global cognition composite score. Expression of 17,091 autosomal, 645 X-linked genes, and 83 Y-linked genes derived from RNAseq of bulk brain tissue were tested using linear regression models for cross-sectional autopsy outcomes and mixed effect models for the longitudinal cognition outcome. For sex-stratified analyses, all models included age at death and post-mortem interval as covariates, while longitudinal cognitive models additional included years of education and latency to death. Y-linked analyses were only conducted in males. To specifically target findings from sex chromosomes, they were analyzed independently from the autosomal chromosomes. False discovery rate (FDR) correction was performed separately for autosomal chromosomes, the X chromosome, and the Y chromosome. Sex-specific gene associations were defined as related to a trait in one sex (pFDR ≤ 0.05) but not in the other sex in stratified analyses and showed suggestive evidence of a sex-modifying effect in interaction analyses (sex x gene interaction p < 0.05, uncorrected). Results from sensitivity and post hoc analyses (i.e. three-way interactions with APOE-ε4) were not corrected for multiple comparisons. FDR-corrected p-values will be notated as “pFDR” throughout; “p” will be used for uncorrected p-values.
Transcriptome-wide analyses yielded over 23,118 significant associations with Aβ, tau tangles, and longitudinal cognition (22,489 autosomal, 629 X-linked, and 0 Y-linked). 10,882 unique autosomal and 332 unique X-linked genes were represented in the 23,118 associations. Approximately 10% (2320 of 23,118) of associations were sex-stratified, significant in one sex but not the other, as illustrated in Fig. 1. The unique genes from significant sex-stratified associations were not enriched for X-linked genes (2032 autosomal, 75 X-linked, roughly 3%). The number of unique autosomal and X-linked genes for each outcome are presented in Supplementary Table 2. In general, the pattern of sex-specific associations was consistent across the DLPFC, PCC, and CN and both the autosomal and X chromosomes (Supplementary Fig. 2). In all three brain regions, we identified a total of 1697 female-stratified (1622 autosomal, 75 X-linked) and 623 male-stratified (609 autosomal, 14 X-linked) gene associations with Aβ, tangles, and longitudinal cognition (Fig. 1). Most associations were identified in the DLPFC (70%, 1619 of 2320 associations) and with tau tangles (39%, 908 of 2320) or longitudinal cognition (52%, 1182 of 2320), Supplementary Data 1).
Chord diagrams summarizing the relative number of both autosomal and X-linked genes associated with Aβ, tau tangles, or longitudinal cognition in the Dorsolateral prefrontal cortex (DLPFC) or Posterior Cingulate Cortex (PCC) or Caudate Nucleus (CN) with chord size corresponding to number of significant genes. Green chords represent the number of main effect associations (not sex-specific) we observe in the data. Pink chords represent female-specific gene associations and blue chords represent male-specific associations. The right panel further highlights the significant male- and female-specific gene associations in sex-stratified analyses. Source data are provided as a Source Data file.
No genes showed significance across all three brain regions. Tables 2 and 3 highlights in bold those sex-stratified associations observed in at least two brain regions and demonstrated consistent directions of effect. Featured in this list are six male-stratified autosomal gene associations with Aβ (GSR) and tau tangles (C16orf74, TDH-AS1, PNPLA6, ATG4B, Novel Transcript (ENSG00000262691)); Table 2). Greater expression of GSR, TDH-AS1, and C16orf74 were associated with lower levels of neuropathology, whereas ATG4B, ENSG00000262691, and PNPLA6 were associated with higher pathological burden. GSR20, PNPLA621,22, and ATG4B23 have previously been implicated in neurodegenerative disorders, including AD. Additionally, two female-stratified autosomal genes were associated with tau tangles consistently across 2 brain regions (MYO15A, NNT), and a further 7 genes with longitudinal cognition (PEPD, CTNNBL1, MVB12A, XKR8, WBP1L, GALNT7-DT, GCA; Table 3). Higher levels of NNT, MYO15A, and GCA were protective in females; greater expression of PEPD, CTNNBL1, MVB12A, XKR8, WBP1L, and GALNT7-DT were associated with faster cognitive decline in females. MYO15A24, NNT25 and GCA26 have previously been implicated in AD. XKR827,28 and WBP1L29 are related to other disorders, such as schizophrenia, and CTNNBL1 has previously been associated with memory in a young, cognitively healthy cohort30. GRM5 was significantly associated with tau tangles in the DLPFC (β = −0.16, pFDR = 0.03) and PCC (β = 0.28, pFDR = 0.02) in female-stratified analyses but exhibited different effects in each brain region. GRM5 encodes the metabotropic glutamate receptor 5, which plays a role in synaptic signaling and has been linked to Aβ and AD31.
To determine whether APOE-ε4, the strongest known common genetic risk factor for AD, played a role in our sex-stratified associations with Aβ or tau tangles, we performed sensitivity analyses additionally covarying for APOE-ε4 allele status. All sex-stratified gene associations with Aβ and tau in two brain regions remained significant. Similarly, models in which cognition was the outcome were additionally covaried for Aβ or tau burden, as pathologic burden is known to independently effect cognition. All sex-stratified gene associations with cognition in two brain regions remained significant in sensitivity analyses (Supplementary Data 2). Finally, we added diagnosis as a covariate to all models. All sex-stratified gene associations in two brain regions remained significant (Supplementary Data 3).
Genes with multi-level evidence of sex-specificity
Here, we focus on genes that met sex-specific gene association criteria and additionally had a sex x gene interaction that met significance set a priori at pFDR ≤ 0.10. This interaction p-value was chosen following literature precedence of studying sex differences32.
No autosomal genes met these criteria. When focusing on X-linked genes, sex significantly moderated MCF2, HDAC8, SLC10A3, and FTX on tau tangle burden or cognitive decline in the DLPFC (Table 4, Fig. 2). Higher expression of MCF2 was significantly associated with lower levels of tau tangles in females (β = −0.18, pFDR = 0.02, Fig. 2A), but not in males (β = 0.10, pFDR = 0.26). By contrast, higher expression of HDAC8 (Fig. 2B) was associated with higher tau tangle burden in females (β = 0.22, pFDR = 0.01), but not in males (β = −0.09, pFDR = 0.37). Higher SLC10A3 (Fig. 2C) expression was associated with faster cognitive decline in females (β = −0.02, pFDR = 0.04), but not in males (β = 0.02, pFDR = 0.24). Alternatively, FTX (Fig. 2D) expression was significantly associated with slower cognitive decline in males (β = 0.02, pFDR = 0.0496), while no relationship was found in females (β = −0.02, pFDR = 0.13).
A–D Plots demonstrating the relationship between gene expression (x-axis) and AD endophenotype (y-axis). The central diagonal lines represent the line of best fit as determined by linear regression and shaded regions represent the 95% confidence interval. Females are colored in orange and males are colored in blue. E A plot demonstrating the relationship between FTX and cognitive decline colored by APOE-ε4 status and sex. Females are colored in yellow (ε4-) and orange (ε4+ ); males are colored in green (ε4-) and blue (ε4+ ). The central diagonal lines represent the line of best fit as determined by linear regression and shaded regions represent the 95% confidence interval and shaded regions represent the 95% confidence interval. Source data are provided as a Source Data file.
Similar to the sex-stratified models, the interaction models for MCF2, HDAC8, SLC10A3, and FTX were additionally adjusted for APOE-ε4 allele positivity and/or AD neuropathology, where relevant, to see whether the genes had effects on tau tangles and cognition independent of APOE-ε4 and neuropathology. All four associations remained significant when covarying for APOE-ε4 status and neuropathological burden (Supplementary Table 3) suggesting that these genes do have an additional impact on AD endophenotypes beyond APOE-ε4 and neuropathology. The sex-stratified associations for these four genes also remained significant in sensitivity analyses.
The interaction models for MCF2, HDAC8, SLC10A3, and FTX were not significant in the PCC or CN, though we looked whether the pattern of effects for the interaction beta coefficients were consistent (i.e., all negative or all positive) between brain regions. FTX was the only gene for which interaction beta coefficients were consistent between all three brain regions (β > 0, Supplementary Data 1). Briefly, interaction beta coefficients were negative in the PCC and CN for MCF2 whereas DLPFC (significant) is positive. Similarly, for HDAC8: beta coefficients are both negative for DLPFC and PCC, but positive for CN. The coefficients for SLC10A3 are both positive in the DLPFC and CN, however it is flipped for the PCC. These findings suggest limited regional concordance, though interpretation is constrained by lack of significance outside the DLPFC.
APOE-ε4 and diagnosis further modify the interaction between sex and gene expression on cognition
Sex differences in the effect of APOE-ε4 on AD risk and on AD endophenotypes are well-established3,18. Therefore, we performed post hoc analyses examining the possible modifying effects of both sex and APOE-ε4 for the four X-linked genes (MCF2, HDAC8, SLC10A, and FTX) by leveraging a three-way interaction model. We observed a significant APOE-ε4 x sex x FTX interaction on longitudinal cognition (p = 0.03, Fig. 2E, Supplementary Table 4), whereby the male-specific protective effect of FTX expression on cognitive decline was driven by APOE-ε4 carriers (β = 0.05, p = 0.01, Fig. 2E). Due to the known sex differences in tau accumulation and cognitive decline among individuals with higher pathological burden, we also looked to see whether there were gene x Aβ positivity interactions on tau and gene x tau positivity interactions on cognition for MCF2, HDAC8, SLC10A, and FTX. There were no moderating effects of Aβ or tau (Supplementary Table 5).
As the sample includes individuals who are cognitively normal (NC) as well as those who have MCI or AD, we also looked for moderating effects of diagnosis. We observed a significant SLC10A3 x sex x diagnosis interaction on longitudinal cognition (p = 0.01) such that higher expression of SLC10A3 is associated with slower cognitive decline among males with AD (β = 0.03, p = 0.01). No other interactions were significant (Supplementary Table 6, Supplementary Fig. 6).
Differential expression of MCF2, HDAC8, SLC10A, and FTX
One challenge of studying the X chromosome is the accounting for X chromosome inactivation, a biological process during which one X chromosome in females is randomly silenced to prevent gene overdosing in females in comparison to males. Due to this phenomenon, we wanted to examine whether the moderating effects of sex on MCF2, HDAC8, SLC10A, and FTX were due to sex differences in gene expression between males and females. FTX and HDAC8 were expressed more highly in males than females (Wilcoxon p < 0.0001), SLC10A3 was expressed more highly in females than males (Wilcoxon p = 0.047), and MCF2 expression did not significantly differ between sexes (Supplementary Fig. 3). As gene expression can change during disease progression, we also examined differences in gene expression based on diagnosis (i.e., NC, MCI, AD) separately in each sex. In males there are no significant differences in gene expression across diagnoses, in females, however, MCF2 expression is decreased in females with MCI and AD in comparison to NC females (Wilcoxon p < 0.05). Female HDAC8 and FTX expression are increased in females with MCI and AD in comparison to NC females (Supplementary Figs. 4 and 5).
Exploratory analysis of the Y chromosome
In males, we also examined the association between 83 Y-linked genes (Supplementary Data 4) and Aβ, tau tangles, and longitudinal cognition in the DLPFC, PCC, and CN. No genes survived correction for multiple comparisons at a pFDR ≤ 0.05.
Post-hoc pathway analysis of sex-specific associations
To connect our sex-stratified findings to broader biological mechanisms that may be implicated in the neuropathological progression of AD in one sex but not the other, we completed gene set enrichment analyses using summary statistics from our sex-stratified analyses. Briefly, summary statistics for each sex-stratified regression model were pre-ranked by p-value and beta coefficient. Enrichment analyses were performed for autosomal and X-linked genes separately, and no gene thresholding based on p-value was performed (see number of genes in Methods). Gene Ontology: Biological Process (GO:BP)33,34 terms between the size of 15 to 500 genes were used for annotation, and all results were FDR-corrected to account for multiple comparisons.
Enrichment analyses yielded a total of 4198 significant GO:BP pathways (1594 unique) across all brain regions and outcomes. Roughly 66% of all significant pathways (2774) were from female-stratified analyses. Most pathways were significant in the DLPFC (2234 of 4198) and were significantly enriched for autosomal genes (4190 significant autosomal terms vs. 8 X-linked). In females, 161 unique pathways and enrichment scores were consistent across two or more brain regions for amyloid, 116 for tau tangles, and 52 for longitudinal cognition. In males, 74 unique pathways were consistent across regions for amyloid, 42 for tau tangles, and 115 for longitudinal cognition. All gene set enrichment results are presented in Supplementary Data 5 and 6.
Focusing on different pathways between sexes, we defined sex-specific biological pathways as GO:BP terms that were significantly enriched in one sex, but not the other, and were oppositely regulated in each sex (e.g., upregulated in females, but negative in males). Figure 3 illustrates the top 6 most significantly enriched pathways by magnitude of normalized enrichment score (NES; 3 positive, 3 negative) for each outcome and sex. Briefly, pathways related to leukocyte differentiation and tissue morphogenesis were upregulated in females (NES > 0) in relation to amyloid whereas serotonin signaling and lipid catabolism were downregulated (NES < 0). MAPK signaling and (chemo)taxis were upregulated in males with relation to amyloid, whereas cytoplasmic translation was downregulated (Fig. 3A). In relation to tau tangles, biological processes related to axonal development were upregulated in females, though neurotransmission-related terms were downregulated. Vesicular transport was downregulated in males, though endocytosis and cell development, growth, and differentiation (Smoothened signaling pathway) were upregulated (Fig. 3B). Finally, translation and innate immunity were downregulated in females in relation to cognition, though receptor signaling was upregulated. Cell cycle-related terms were downregulated in males whereas endosomal organization was upregulated (Fig. 3C). While GO:BP terms are often broad and overlapping, these results provide evidence that biological pathways in AD differ between males and females.
A–C Dot plots demonstrating selected GO:BP terms from gene set enrichment analyses using fgsea (v 1.26.0) for each AD endophenotype. FDR correction was used for multiple comparisons. Depicted terms were selected using the following criteria: An FDR-corrected p-value significant in one sex, but not the other, and opposite signs of normalized enrichment scores (NES). The plots include the top 3 highest and lowest enrichment scores for each sex. Females are denoted by yellow points, and males in blue. Filled circles represent a significant FDR-corrected p-value while an empty circle does not. Brain regions are included at the end of each GO:BP term: Dorsolateral prefrontal cortex (DLPFC) or Posterior Cingulate Cortex (PCC) or Caudate Nucleus (CN). Source data are provided as a Source Data file.
In X-linked gene enrichment analyses, few GO:BP pathways survived correction for multiple comparisons. In the DLPFC, pathways associated with stress and circulatory system development were upregulated in response to amyloid in females whereas processes related to RNA processing were upregulated in males primarily in the CN (Fig. 4). Interestingly, these processes in the CN were downregulated in females. No X-linked gene set enrichment terms for tau tangles or longitudinal cognition survived correction for multiple comparisons. Though the most significant pathways for tau tangles were “microtubule cytoskeleton organization” and “cellular component disassembly” for females and males, respectively. “Anatomical structure formation involved in morphogenesis” for males and “organonitrogen compound biosynthetic process” in females were most significant in relation to longitudinal cognition (Supplementary Data 6).
Dot plots demonstrating selected GO:BP terms from gene set enrichment analyses using fgsea (v 1.26.0) for X-linked genes and each AD endophenotype. FDR correction was used for multiple comparisons. Depicted terms were selected using the following criteria: An FDR-corrected p-value significant in one sex, but not the other. Females are denoted by yellow points, and males in blue. Filled circles represent a significant FDR-corrected p-value while an empty circle does not. Brain regions are included at the end of each GO:BP term: Dorsolateral prefrontal cortex (DLPFC) or Posterior Cingulate Cortex (PCC) or Caudate Nucleus (CN). Source data are provided as a Source Data file.
Discussion
AD disproportionately affects women in comparison to men. Women at all stages of disease have greater tau burden than men as measured at autopsy, via tau-PET and cerebrospinal fluid (CSF) measures5,18, and this difference in tau burden is even more apparent if they carry APOE-ε43,5,35 or have abnormal amyloid burden5. Studies also demonstrate that women experience faster cognitive decline in response to tau pathology, especially if they are Aβ-positive and carry APOE-ε43,36. Though several hypotheses behind these sex differences in AD have been put forward, there is much more to be understood about the biological pathways behind these differences.
Large-scale human omics studies have provided an invaluable opportunity to better understand the molecular basis behind these sex differences in AD. They have discovered genetic variants that exhibit sex-specific associations7, sex-biased molecular pathways10, and with increasing cell-type resolution12,13, changes in gene expression between the sexes in AD, though most of these studies have focused on AD case-control status. Our study builds upon these previous findings by identifying sex-specific transcriptomic associations with AD endophenotypes in three brain regions.
Approximately 10% of our observed autosomal and X-linked gene expression associations with AD endophenotypes exhibited sex-stratified significance. 89 of the 2320 (3.8%) significant sex-specific associations were X-linked, which is consistent with the proportion of protein-coding genes (4% X-linked) within the entire genome37. Most sex-specific associations were female-stratified (73.1%) and were predominantly associated with post-mortem tau tangles and ante-mortem cognitive trajectories. The direction of effects suggested increased gene expression was associated with both risk for and protection against AD. Though 16 genes were significant in two brain regions, no genes exhibited sex-stratified significance across all three brain regions.
Few genes are expressed on the Y chromosome, and many are pseudogenes. We did not identify any Y-linked gene associations with AD endophenotypes in our study, though the Y chromosome has previously been implicated in AD. Mosaic loss of the Y chromosome and downregulation of Y-linked gene expression have been associated with increased AD risk in men38,39. These studies calculated the loss of expression across the entire Y chromosome, suggesting that different approaches may be more appropriate for studying the Y chromosome than the ones used in this study.
A technical challenge in studying the X chromosome is that X chromosome dosage may influence the expression of X-linked genes. In females, one X chromosome is silenced at random to equalize dosage between males and females40. The silencing is not complete; up to a third of genes escape inactivation, which can result in increased expression in females41,42. X chromosome inactivation escapism (XCI-e) is also heterogeneous between individuals and reported to change with aging43, representing further complexity. Despite these challenges, the X chromosome remains a font of knowledge for AD research. X-linked genes are highly expressed in the brain44, with many of them involved in immune function45. Previous studies have implicated X-linked genes such as KDM6A46, USP1147, TLR748 in AD and compelling evidence suggests that the presence of a second X chromosome is protective in aging46.
Our study is the largest transcriptomic investigation of the X chromosome and AD endophenotypes to date. We identified 4 gene associations that were moderated by sex, and 89 other X-linked gene associations, consisting of 75 unique genes, were sex-stratified (62 in female-stratified analyses, 13 male). 26 of 62 (42%, Supplementary Data 7) unique genes identified in female-stratified analyses have been reported to escape XCI suggesting a potential enrichment for these XCI-e in contributing to sex differences in the progression of AD. Some of these genes, such as TSPAN649 and SEPTIN650,51, have previously been implicated in AD while others such as JPX are directly involved in the process of XCI52, warranting further study.
MCF2, HDAC8, SLC10A3, and FTX, displayed evidence of moderation by sex. These genes conferred both risk and protection, with greater MCF2 and FTX expression associated with less tau tangles in females and slower cognitive decline in males, respectively. Further, protective FTX effects appeared to be driven by male APOE-ε4 carriers. By contrast, greater HDAC8 and SLC10A3 expression was associated with greater tau tangle burden and faster cognitive decline, respectively, in females only. However, greater SLC10A3 expression in the DLPFC was associated with slower cognitive decline among males diagnosed with AD putting forth that there are disease-stage specific effects, though our sample is underpowered. These four X-linked associations survived additional adjustment from APOE-ε4, the most common genetic risk factor for sporadic AD dementia positing that these genes have an effect beyond APOE-ε4 carriership. These effects were not consistent across brain regions, perhaps suggesting that these genes are playing region-specific roles though X-linked gene expression also generally differs across the brain. The absence of significance beyond the DLPFC limits our ability to draw firm conclusions about regional specificity.
There has been no previous association between MCF2 and Alzheimer’s disease endophenotypes, though it has been associated with autism spectrum disorders53 and cortical neuron migration54. HDAC8 and SLC10A3 are members of protein families that have been implicated in AD55,56. HDAC8 encodes a histone deacetylase; epigenetic deacetylation has been associated with AD55,57. Similarly, SLC10A3 is a sodium-bile acid cotransporter and thought to be a housekeeping gene due to its ubiquitous expression58; higher levels of bile acids have been observed in individuals with AD and MCI59. Recently, another member of the SLC super family, SLC9A7 was identified by Belloy et al., as a risk locus in an X-chromosome wide association study60. Though we see a protective effect in males, FTX is a long non-coding RNA that is heavily involved in XCI61, which males do not undergo, warranting further study. We posit that some of the protective effects seen on X-linked genes for men, particularly on FTX, may be due to a demographically skewed sample. However, these protective effects could also be observed on the maternally inherited X chromosome46. Both FTX and HDAC8 showed higher expression in males compared to females despite being X-linked and related to X inactivation61,62, though their expression may not be increased compared to females in other datasets.
We identified biological pathways that were significantly enriched in one sex, but not the other underscoring the thesis of distinct sex-specific pathways that may lead each sex to higher risk of AD. Females exhibited upregulation in neuronal development and immune-related pathways. A study by Meyer et al. also identified an upregulation of genes involved in neurogenesis in induced pluripotent stem cells derived from individuals with sporadic AD, supporting our findings63. They suggested that premature neuronal differentiation as a result of this overexpression may contribute to the onset of AD. Additionally, neuroinflammation and innate immunity has been implicated in the development of AD64,65,66. Females also exhibited downregulation in neurotransmission-related pathways. Disruption in neurotransmission has long been linked to AD, perhaps resulting from neuronal death during the progression of disease67. In contrast, males exhibited upregulation in endocytosis pathways. Cellular aging is suggested to facilitate increased amyloid precursor protein (APP) uptake, resulting in the creation of more Aβ deposits68,69. Similarly, tau hyperphosphorylation has been linked to aberrant endocytosis, both of which may contribute to the development of AD68. Our study also highlighted changes to the cell cycle in males, described in the two-hit hypothesis of AD70 such that abnormal neuronal reentry into the cell cycle results in cell death71. We further highlight biological processes that are upregulated in one sex and downregulated in the other (Fig. 3) perhaps suggesting that one intervention may not be effective in both sexes. These data may point toward the development of precision therapeutics that may function differently for each sex on the same pathway.
This study had many strengths, including a large sample size, the examination of multiple brain regions, and gene expression from both X- and Y-chromosomes. Nevertheless, this study also possessed some weaknesses. First, there are approximately 500 X-linked genes in our study in comparison to 20,000 autosomal genes. Given one of our aims was to highlight possible contributions of the X chromosome to AD-related biology, a relatively understudied area, we made an a priori decision to analyze them separately to ensure that potential X-linked associations were not obscured by multiple testing correction across the full transcriptome. We acknowledge that this may have introduced bias in our results. Second, most individuals in the sample identify as non-Hispanic white and are highly educated, which may not be generalizable to a larger population. In addition, participants were, on average, 88 years at their time of death. At such an advanced age, it is very possible that brain-derived gene expression might be indicating different disease processes than what might be observed in vivo at younger ages (i.e., below 70). Our study is also limited by the cross-sectional nature of data derived from post-mortem brain tissue. We are unable to examine changes in gene expression over time and how they may relate to AD endophenotypes, nor can we infer causality or timing of disease progression. Further, we age-matched females and males in this sample, which may result in a male sample who is less representative of the general population, given the average male life expectancy is 73.5 years as of 202172. Given the increase in AD risk attributed to APOE-ε4 carriership73, male APOE-ε4 carriers living to an average age of 88 likely required a considerable level of gene-environment protective effects to survive into advanced age74. Taken together, similar studies in the future should consider a sample that is more representative of the general population. Finally, AD patients at older age are likely to have comorbidities (e.g., Lewy bodies, cardiovascular or cerebrovascular disease) that may also affect gene expression as well as cognitive outcomes. Though our study focused on AD, the contribution of these comorbidities to gene expression and cognition should be considered in future analyses75.
Our study is the largest brain transcriptomic study of sex differences in AD and provides a comprehensive look at the autosome and sex chromosomes. We identified 2320 sex-specific transcriptomic associations with amyloid, tau, and longitudinal cognition, with 89 associations on the X chromosome, and four of whom are significant moderators. Our findings support the involvement of processes such as endocytosis, neuronal development, and neurotransmission in the progression and development of AD, and suggest that biological pathways to AD may differ by sex.
Though we have successfully identified sex-specific gene associations with AD endophenotypes, we have yet to characterize the mediating effects of genes, pinpoint cell-type specific mechanisms, and assess the impact of X-chromosome inactivation escapism on our observed associations. Since these contributions remain unclear, the present study forms a strong foundation for future work with longitudinal and cell-specific data, which are critical to continue characterizing and uncovering sex differences in AD. Overall, these findings highlight the importance of precision medicine approaches that consider sex-specific biological pathways to select new targets for AD therapeutic intervention.
Methods
Study population
Data were obtained from the Religious Orders Study and Rush Memory and Aging Project (ROS/MAP). These studies enrolled older adults free of dementia who agreed to annual clinical evaluations and brain donation at death15. All participants gave written informed consent, an Anatomic Gift Act, and signed a repository consent allowing their data to be shared. The Rush Institutional Review Board (IRB) approved all protocols and the Vanderbilt University Medical Center IRB approved secondary analyses of existing data. Data from this cohort is accessible either online on the Accelerating Medicines Partnership – Alzheimer’s Disease (AMP-AD) Knowledge Portal (https://adknowledgeportal.synapse.org/Explore/Programs/DetailsPage?Program=AMP-AD, syn3219045) or via the Rush Alzheimer’s Disease Center Resource Sharing Hub (https://www.radc.rush.edu/).
Neuropsychological composite scores
Cognitive measures in ROS/MAP have been described previously76. Cognitive assessments are administered annually to participants. A global cognition composite was calculated by averaging z-scores from 17 tests across 5 domains of cognition (episodic, semantic, and working memory, perceptual orientation, and perceptual speed).
Bulk mRNA sequencing
Dorsolateral prefrontal cortex (DLPFC), posterior cingulate cortex (PCC), and the head of the caudate nucleus (CN) tissue were acquired at autopsy. These regions were selected by the parent study based on tissue availability and because the regions are affected by numerous age-related disease conditions. Full details on RNA extraction, library preparation, and RNA sequencing have been described previously16,17. Briefly, RNA was extracted from each brain region and ribosomal depletion was formed (RiboGold (Illumina, 20020599). Sequence libraries were sequenced using 2 x 150 bp paired end reads on an Illumina NovaSeq 6000 targeting an average of 40 to 50 million reads.
RNAseq alignment and quality control
Sequence alignment and gene counting followed a published protocol77,78 including STAR (version 2.5.2b)79 alignment to the Ensembl (GRCh38 release 99)80 reference genome. Gene counts were obtained using the featureCounts function from the Subread package (v.2.0.0)81. Alignment metrics were calculated using Picard tools (version 2.18.27, http://broadinstitute.github.io/picard/)82. Prior to quality control, the number of samples was as follows: nDLPFC = 1126, nPCC = 531, nCN = 726.
For quality control (QC) after alignment, samples with RNA integrity number (RIN) < 4 or post-mortem interval (PMI) > 24 h were removed. At this time, genes were split into three groups by chromosome: autosomal (chromosomes 1-22), X, and Y chromosome (males only) and underwent QC separately. Prior to quantile normalization using the cqn package (version 1.48.0)83, genes with <1 count per million (CPM) in 50% of individuals per diagnosis (i.e., normal cognition (NC), MCI, AD) were removed. Additionally, genes missing gene length or GC-content measurements were removed. After quantile normalization, gene expression values with >5 standard deviations (SD) from the mean were set to 0. Samples that were principal component outliers (greater than 5 SD from the mean), whose principal components did not align to participants’ self-reported sex, or who were missing RIN or demographic variables (e.g., age, sex, cognition), were also removed. To confirm observed effects were not due to technical variation, RNAseq data underwent further iterative normalization using the limma package (version 3.60.4)84,85 adjusting for batch, post-mortem interval (PMI), RIN, and percentage of coding, intronic, and intergenic bases resulting in the following datasets: DLPFC [n = 926, genes = 17,088 (autosomal), 562 (X-linked), 63 (Y-linked)]; PCC [n = 518, genes=18,385 (autosomal), 598 (X-linked), 71 (Y-linked)]; CN [n = 706, genes = 17,928 (autosomal), 591 (X-linked), 76 (Y-linked)]. A visual workflow including the number of samples dropped at each step is included as Supplementary Figs. 3, 4.
Measures of Alzheimer’s disease pathology
Aβ load and tau tangle density were measured via immunohistochemistry, as detailed previously2. Aβ load (cortical) and tau tangle density was quantified as the average area occupied by pathology across 8 brain regions: hippocampus, angular gyrus, and entorhinal, midfrontal, inferior temporal, calcarine, anterior cingulate, and superior frontal cortices. Aβ and tau tangle values were square-root transformed to approximate a normal distribution for analyses. Aβ positivity was determined using CERAD scores (1 or 2 = positive, 3 or 4 = negative). Tau positivity was determined using Braak stages (1, 2, or 3 = negative, 4, 5, and 6 = positive).
Statistical analyses
Statistical analyses were completed using R (version 4.2.1). Participant sex was determined based on self-report as well as gene expression (XIST in females, UTY in male). Due to a larger number of female participants, we subsampled female participants by matching to male participants using propensity scores based on age of death, PMI, education, latency to death (i.e., difference between age of death and age at last cognitive visit), race, and APOEε4 allele count. After matching, males and females were statistically compared across these measures to ensure no differences remained between groups.
Gene expression, Aβ, tau tangles, and cognitive measures were mean-centered prior to analyses and were treated as continuous variables. Linear regression and mixed-effects models were used to assess the association between gene expression and AD endophenotypes for each gene in both pooled and sex-stratified analyses. All p-values for discovery analyses were adjusted using the false discovery rate (FDR) method to correct for multiple comparisons across each model, brain region, sex-stratification, and outcome. FDR-correction was performed for genes expressed on autosomal chromosomes, the X chromosome, and Y chromosome separately so that smaller effects on sex chromosomes may be observed. Significance in sex agnostic (i.e., pooled) analyses was set a priori to FDR-corrected p ≤ 0.05. Sex-specific gene associations were defined as related to a trait in one sex (pFDR ≤ 0.05) but not in the other sex in stratified models and showed suggestive evidence of a sex-modifying effect in interaction models (sex x gene interaction p < 0.05, uncorrected). Gene x sex interaction significance was set a priori to pFDR ≤ 0.10. Age at death, PMI and sex were included as covariates where relevant (i.e., sex in male-female pooled analyses) in the cross-sectional models examining Aβ load and tau tangle density. Linear mixed-effects models tested the association between gene expression and longitudinal global cognition where the intercept and time interval between the final cognitive visit and the visit of interest were entered as both fixed and random effects. Mixed-effects models included age at death, PMI, latency to death, education, and sex (where relevant) as covariates.
APOE-ε4 status (positive/negative) was included as an additional covariate in sensitivity analyses; neuropathological measures were also added as covariates when cognition was used as an outcome. A linear mixed-effects model including the aforementioned covariates was used to assess the three-way sex x gene x APOE-ε4 status on cognition as a post hoc analysis. Additional post hoc analyses examined the sex x gene x Aβ positivity interaction on tau, the sex x gene x tau positivity interaction on amyloid, and the moderating effects of diagnosis (sex x gene x diagnosis; normal, MCI, AD). Sensitivity and post hoc analyses were not corrected for multiple comparisons. Wilcoxon tests were used to compare mean normalized gene expression of between males and females. FDR-corrected p-values will be notated as “pFDR” throughout; “p” will be used for uncorrected p-values.
Gene set enrichment analysis
We performed pre-ranked gene set enrichment analyses with the R package fgsea (version 1.26.0)86 on all tested autosomal and X-linked genes per sex, brain region, and outcome. Analyses were limited to Gene Ontology: Biological Process (C5)33,34 terms of 15 to 500 genes. Gene rankings were based on the sex-stratified beta coefficients from each linear regression model. All enrichment results were FDR-corrected to account for multiple comparisons.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All bulk brain RNA sequencing and phenotypic data are available through the Accelerating Medicines Partnership – Alzheimer’s Disease (AMP-AD) Knowledge portal https://adknowledgeportal.synapse.org/Explore/Programs/DetailsPage?Program=AMP-AD) under accession code syn3219045. Data are available under controlled use conditions due to privacy laws. Access can be obtained by requesting a data use agreement on the AMP-AD portal or from the Rush Alzheimer’s Disease Center Resource Sharing Hub (https://www.radc.rush.edu/requests.htm). Source data for figures are provided with this paper. Source data are provided with this paper.
Code availability
Publicly available software was used for all analyses. Code for RNAseq processing, regression analyses, and gene set enrichment analyses can be found on Zenodo (https://doi.org/10.5281/zenodo.16879086) or are available upon request from Logan.C.Dumitrescu@vumc.org or mseto@bwh.harvard.edu.
References
Rajan, K. B. et al. Population estimate of people with clinical Alzheimer’s disease and mild cognitive impairment in the United States (2020-2060). Alzheimers Dement 17, 1966–1975 (2021).
Barnes, L. L. et al. Sex Differences in the Clinical Manifestations of Alzheimer Disease Pathology. Arch. Gen. Psychiatry 62, 685–691 (2005).
Buckley, R. F. et al. Sex, amyloid, and APOE ε4 and risk of cognitive decline in preclinical Alzheimer’s disease: Findings from three well-characterized cohorts. Alzheimers Dement 14, 1193–1203 (2018).
Filon, J. R. et al. Gender Differences in Alzheimer Disease: Brain Atrophy, Histopathology Burden, and Cognition. J. Neuropathol. Exp. Neurol. 75, 748–754 (2016).
Buckley, R. F. et al. Sex Mediates Relationships Between Regional Tau Pathology and Cognitive Decline. Ann. Neurol. 88, 921–932 (2020).
Li, G. D. et al. Female-specific effect of the BDNF gene on Alzheimer’s disease. Neurobiol. Aging 53, 192.e111–192.e119 (2017).
Deming, Y. et al. Sex-specific genetic predictors of Alzheimer’s disease biomarkers. Acta Neuropathol. 136, 857–872 (2018).
Sun, L. L., Yang, S. L., Sun, H., Li, W. D. & Duan, S. R. Molecular differences in Alzheimer’s disease between male and female patients determined by integrative network analysis. J. Cell Mol. Med. 23, 47–58 (2019).
López-Cerdán, A. et al. An integrated approach to identifying sex-specific genes, transcription factors, and pathways relevant to Alzheimer’s disease. Neurobiol. Dis. 199, 106605 (2024).
Guo, L., Zhong, M. B., Zhang, L., Zhang, B. & Cai, D. Sex Differences in Alzheimer’s Disease: Insights From the Multiomics Landscape. Biol. Psychiatry 91, 61–71 (2022).
Fass, S. B. et al. Relationship between sex biases in gene expression and sex biases in autism and Alzheimer’s disease. Biol. Sex. Differences 15, 47 (2024).
Belonwu, S. A. et al. Sex-Stratified Single-Cell RNA-Seq Analysis Identifies Sex-Specific and Cell Type-Specific Transcriptional Responses in Alzheimer’s Disease Across Two Brain Regions. Mol. Neurobiol. 59, 276–293 (2022).
Soudy, M., Bars, S. L. & Glaab, E. Sex-dependent molecular landscape of Alzheimer’s disease revealed by large-scale single-cell transcriptomics. Alzheimers Dement 21, e14476 (2025).
Sun, L., Wang, Z., Lu, T., Manolio, T. A. & Paterson, A. D. eXclusionarY: 10 years later, where are the sex chromosomes in GWASs?. Am. J. Hum. Genet 110, 903–912 (2023).
Bennett, D. A. et al. Religious Orders Study and Rush Memory and Aging Project. J. Alzheimers Dis. 64, S161–S189 (2018).
De Jager, P. L. et al. A multi-omic atlas of the human frontal cortex for aging and Alzheimer’s disease research. Sci. Data 5, 180142 (2018).
Mostafavi, S. et al. A molecular network of the aging human brain provides insights into the pathology and cognitive decline of Alzheimer’s disease. Nat. Neurosci. 21, 811–819 (2018).
Hohman, T. J. et al. Sex-Specific Association of Apolipoprotein E With Cerebrospinal Fluid Levels of Tau. JAMA Neurol. 75, 989–998 (2018).
Oveisgharan, S. et al. Sex differences in Alzheimer’s disease and common neuropathologies of aging. Acta Neuropathol. 136, 887–900 (2018).
Foret, M. K. et al. Early oxidative stress and DNA damage in Aβ-burdened hippocampal neurons in an Alzheimer’s-like transgenic rat model. Commun. Biol. 7, 861 (2024).
Zuchner, S. et al. Mutations in PNPLA6 Cause a Range of Neurodegenerative Phenotypes (P6.060). Neurology 82, P6.060 (2014).
Mathys, H. et al. Single-cell multiregion dissection of Alzheimer’s disease. Nature 632, 858–868 (2024).
Shin, J. H. et al. Down-regulated TMED10 in Alzheimer disease induces autophagy via ATG4B activation. Autophagy 15, 1495–1505 (2019).
Bellenguez, C. et al. New insights into the genetic etiology of Alzheimer’s disease and related dementias. Nat. Genet. 54, 412–436 (2022).
Ghosh, D., Levault, K. R. & Brewer, G. J. Relative importance of redox buffers GSH and NAD(P)H in age-related neurodegeneration and Alzheimer disease-like mouse neurons. Aging Cell 13, 631–640 (2014).
Zhou, R. et al. Bone Marrow-Derived GCA(+) Immune Cells Drive Alzheimer’s Disease Progression. Adv. Sci. (Weinh.) 10, e2303402 (2023).
Huang, Z. Evidence that Alzheimer’s Disease Is a Disease of Competitive Synaptic Plasticity Gone Awry. J. Alzheimers Dis. 99, 447–470 (2024).
Yu, Y., Chen, R., Mao, K., Deng, M. & Li, Z. The Role of Glial Cells in Synaptic Dysfunction: Insights into Alzheimer’s Disease Mechanisms. Aging Dis. 15, 459–479 (2024).
Li, C., Yang, T., Ou, R. & Shang, H. Overlapping Genetic Architecture Between Schizophrenia and Neurodegenerative Disorders. Front. Cell Dev. Biol. 9, 797072 (2021).
Papassotiropoulos, A. et al. A genome-wide survey and functional brain imaging study identify CTNNBL1 as a memory-related gene. Mol. Psychiatry 18, 255–263 (2013).
Wang, J. et al. Metabotropic glutamate receptor 5 (mGluR5) is associated with neurodegeneration and amyloid deposition in Alzheimer’s disease: A [(18)F]PSS232 PET/MRI study. Alzheimers Res. Ther. 16, 9 (2024).
Buckley, R. F. et al. Menopause Status Moderates Sex Differences in Tau Burden: A Framingham PET Study. Ann. Neurol. 92, 11–22 (2022).
Ashburner, M. et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet. 25, 25–29 (2000).
Consortium, T. G. O. et al. The Gene Ontology knowledgebase in 2023. Genetics 224, iyad031 (2023).
Tsiknia, A. A. et al. Sex differences in plasma p-tau181 associations with Alzheimer’s disease biomarkers, cognitive decline, and clinical progression. Mol. Psychiatry 27, 4314–4322 (2022).
Koran, M. E. I., Wagener, M. & Hohman, T. J. Sex differences in the association between AD biomarkers and cognitive decline. Brain Imaging Behav. 11, 205–213 (2017).
Ross, M. T. et al. The DNA sequence of the human X chromosome. Nature 434, 325–337 (2005).
Dumanski, J. P. et al. Mosaic Loss of Chromosome Y in Blood Is Associated with Alzheimer Disease. Am. J. Hum. Genet 98, 1208–1219 (2016).
Caceres, A., Jene, A., Esko, T., Perez-Jurado, L. A. & Gonzalez, J. R. Extreme downregulation of chromosome Y and Alzheimer’s disease in men. Neurobiol. Aging 90, 150.e151–150.e154 (2020).
Panning, B. X-chromosome inactivation: the molecular basis of silencing. J. Biol. 7, 30 (2008).
Oliva, M. et al. The impact of sex on gene expression across human tissues. Science 369, eaba3066 (2020).
Tukiainen, T. et al. Landscape of X chromosome inactivation across human tissues. Nature 550, 244–248 (2017).
Gadek, M. et al. Aging activates escape of the silent X chromosome in the female mouse hippocampus. Sci. Adv. 11, eads8169 (2025).
Nguyen, D. K. & Disteche, C. M. Dosage compensation of the active X chromosome in mammals. Nat. Genet. 38, 47–53 (2006).
Libert, C., Dejager, L. & Pinheiro, I. The X chromosome in immune functions: when a chromosome makes the difference. Nat. Rev. Immunol. 10, 594–604 (2010).
Davis, E. J. et al. A second X chromosome contributes to resilience in a mouse model of Alzheimer’s disease. Sci. Transl. Med. 12, eaaz5677 (2020).
Yan, Y. et al. X-linked ubiquitin-specific peptidase 11 increases tauopathy vulnerability in women. Cell 185, 3913–3930.e3919 (2022).
Lopez-Lee, C. et al. Tlr7 drives sex differences in age- and Alzheimer’s disease-related demyelination. Science 386, eadk7844 (2024).
Guix, F. X. et al. Tetraspanin 6: a pivotal protein of the multiple vesicular body determining exosome release and lysosomal degradation of amyloid precursor protein fragments. Mol. Neurodegeneration 12, 25 (2017).
Liu, S., Pei, H., Zeng, D., Deng, Y. & Xie, W. Septin6 as a new approach for AD treatment. Acta Materia Medica 3, 309–311 (2024).
Princen, K. et al. Pharmacological modulation of septins restores calcium homeostasis and is neuroprotective in models of Alzheimer’s disease. Science 384, eadd6260 (2024).
Tian, D., Sun, S. & Lee, J. T. The long noncoding RNA, Jpx, is a molecular switch for X chromosome inactivation. Cell 143, 390–403 (2010).
Piton, A. et al. Systematic resequencing of X-chromosome synaptic genes in autism spectrum disorder and schizophrenia. Mol. Psychiatry 16, 867–880 (2011).
Molinard-Chenu, A. et al. MCF2 is linked to a complex perisylvian syndrome and affects cortical lamination. Ann. Clin. Transl. Neurol. 7, 121–125 (2020).
Pereira, M., Cruz, M. T., Fortuna, A. & Bicker, J. Restoring the epigenome in Alzheimer’s disease: advancing HDAC inhibitors as therapeutic agents. Drug Discov. Today 29, 104052 (2024).
Jia, Y. et al. Alteration in the Function and Expression of SLC and ABC Transporters in the Neurovascular Unit in Alzheimer’s Disease and the Clinical Significance. Aging Dis. 11, 390–404 (2020).
Sharma, V. K., Mehta, V. & Singh, T. G. Alzheimer’s Disorder: Epigenetic Connection and Associated Risk Factors. Curr. Neuropharmacol. 18, 740–753 (2020).
Claro da Silva, T., Polli, J. E. & Swaan, P. W. The solute carrier family 10 (SLC10): beyond bile acid transport. Mol. Asp. Med. 34, 252–269 (2013).
Ehtezazi, T., Rahman, K., Davies, R. & Leach, A. G. The Pathological Effects of Circulating Hydrophobic Bile Acids in Alzheimer’s Disease. J. Alzheimers Dis. Rep. 7, 173–211 (2023).
Belloy, M. E. et al. Role of the X Chromosome in Alzheimer Disease Genetics. JAMA Neurol. 81, 1032–1042 (2024).
Furlan, G. et al. The Ftx Noncoding Locus Controls X Chromosome Inactivation Independently of Its RNA Products. Mol. Cell 70, 462–472.e468 (2018).
Chakrabarti, A. et al. HDAC8: a multifaceted target for therapeutic interventions. Trends Pharmacol. Sci. 36, 481–492 (2015).
Meyer, K. et al. REST and Neural Gene Network Dysregulation in iPSC Models of Alzheimer’s Disease. Cell Rep. 26, 1112–1127.e1119 (2019).
Bettcher, B. M. & Kramer, J. H. Longitudinal inflammation, cognitive decline, and Alzheimer’s disease: a mini-review. Clin. Pharm. Ther. 96, 464–469 (2014).
Friedberg, J. S. et al. Associations between brain inflammatory profiles and human neuropathology are altered based on apolipoprotein E ε4 genotype. Sci. Rep. 10, 2924 (2020).
King, E. et al. Peripheral inflammation in prodromal Alzheimer’s and Lewy body dementias. J. Neurol., Neurosurg. amp; Psychiatry 89, 339 (2018).
Kaur, S., DasGupta, G. & Singh, S. Altered Neurochemistry in Alzheimer’s Disease: Targeting Neurotransmitter Receptor Mechanisms and Therapeutic Strategy. Neurophysiology 51, 293–309 (2019).
Zadka, Ł. et al. Endocytosis and Alzheimer’s disease. Geroscience 46, 71–85 (2024).
Burrinha, T. & Guimas Almeida, C. Aging impact on amyloid precursor protein neuronal trafficking. Curr. Opin. Neurobiol. 73, 102524 (2022).
Zhu, X., Raina, A. K., Perry, G. & Smith, M. A. Alzheimer’s disease: the two-hit hypothesis. Lancet Neurol. 3, 219–226 (2004).
Moh, C. et al. Cell cycle deregulation in the neurons of Alzheimer’s disease. Results Probl. Cell Differ. 53, 565–576 (2011).
Harris, E. Life Expectancy Gap Grows Between Men and Women in US. JAMA 330, 2241–2241 (2023).
Reiman, E. M. et al. Exceptionally low likelihood of Alzheimer’s dementia in APOE2 homozygotes from a 5,000-person neuropathological study. Nat. Commun. 11, 667 (2020).
Holmes, R. et al. How are APOE4, changes in body weight, and longevity related? Insights from a causal mediation analysis. Front Aging 5, 1359202 (2024).
Beach, T. G. & Malek-Ahmadi, M. Alzheimer’s Disease Neuropathological Comorbidities are Common in the Younger-Old. J. Alzheimers Dis. 79, 389–400 (2021).
Wilson, R. S. et al. Temporal course and pathologic basis of unawareness of memory loss in dementia. Neurology 85, 984–991 (2015).
Logsdon, B. A. et al. Meta-analysis of the human brain transcriptome identifies heterogeneity across human AD coexpression modules robust to sample collection and methodological approach. bioRxiv, 510420 (2019).
Seto, M. et al. RNASE6 is a novel modifier of APOE-ε4 effects on cognition. Neurobiol. Aging 118, 66–76 (2022).
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
Howe, K. L. et al. Ensembl 2021. Nucleic Acids Res. 49, D884–D891 (2021).
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 30, 923–930 (2014).
Institute, B. Picard Toolkit. Broad Institute, GitHub Repository (2019).
Hansen, K. D., Irizarry, R. A. & Wu, Z. Removing technical variability in RNA-seq data using conditional quantile normalization. Biostatistics 13, 204–216 (2012).
Law, C. W., Chen, Y., Shi, W. & Smyth, G. K. voom: precision weights unlock linear model analysis tools for RNA-seq read counts. Genome Biol. 15, R29 (2014).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47–e47 (2015).
Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. 102, 15545–15550 (2005).
Acknowledgements
The results published here are in whole or in part based on data obtained from the AD Knowledge Portal (https://adknowledgeportal.synapse.org). Study data were provided by the Rush Alzheimer’s Disease Center, Rush University Medical Center, Chicago, Illinois, USA. ROS/MAP is supported by the following grants from the National Institute on Aging: P30AG10161, R01AG15819, R01AG17917, R01AG30146, R01AG36042, RC2AG036547, R01AG36836, R01AG48015, RF1AG57473, U01AG32984, U01AG46152, U01AG46161, U01AG61356 as well as the Illinois Department of Public Health, and the Translational Genomics Research Institute. Data can be requested at www.radc.rush.edu. Additional support includes: 24AARF-1201281 [M.S.], R01AG073439 [L.D.], DP2AG082342 [R.F.B.], R00AG061238 [R.F.B.], K01AG049164 [T.J.H.], R01AG059716 [T.J.H.], R01AG061518 [T.J.H.], HHSN311201600276P [T.J.H], K24AG046373 [A.L.J.], R01AG034962 [A.L.J.], R01NS100980 [A.L.J.], R01AG056534 [A.L.J.], R01AG15819 [A.L.J.], the Vanderbilt Clinical Translational Science Award (UL1TR000445), Vanderbilt High Performance Computer Cluster (S10OD023680), and the Vanderbilt University Alzheimer’s Disease Research Center (P20AGAG068082).
Author information
Authors and Affiliations
Contributions
M.S., M.C., R.F.B., and L.D. conceptualized the project. R.F.B. and L.D. supervised the project. M.S., M.C., and M.L.G. performed all analysis. K.A.G., A.L.J., P.L.D., D.A.B., Y.W., L.L.B., J.A.S., and T.J.H. arranged and provided data for the project. M.S. and M.C. wrote the original draft. All authors (M.S., M.C., M.L.G., G.C., K.A.G., A.L.J., P.L.D., D.A.B., Y.W., L.L.B., J.A.S., T.J.H., R.F.B., L.D.) reviewed and edited the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Polymnia Georgiou and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Seto, M., Clifton, M., Gomez, M.L. et al. Sex-specific associations of gene expression with Alzheimer’s disease neuropathology and ante-mortem cognitive performance. Nat Commun 16, 9466 (2025). https://doi.org/10.1038/s41467-025-64525-5
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1038/s41467-025-64525-5
