Transcriptional analyses identify pericyte-centered signaling programs altered by sex and brain region in Alzheimer’s Disease

Kareem El-Ghazawi; Yonathan T Aberra; Yashasvisai Veeramasu; Wisam A Fares; Brant E Isakson; Harald Sontheimer; Ukpong B Eyo; Shayn M Peirce

doi:10.1038/s42003-026-10446-y

. 2026 Jun 8;9:1216. doi: 10.1038/s42003-026-10446-y

Transcriptional analyses identify pericyte-centered signaling programs altered by sex and brain region in Alzheimer’s Disease

Kareem El-Ghazawi ^1,², Yonathan T Aberra ^1,³, Yashasvisai Veeramasu ¹, Wisam A Fares ¹, Brant E Isakson ^4,⁵, Harald Sontheimer ⁶, Ukpong B Eyo ^6,^7,^✉, Shayn M Peirce ^1,^✉

PMCID: PMC13586196 PMID: 42260160

Abstract

Pericytes are critical components of the neurovascular unit (NVU), regulating endothelial cell stability, blood-brain barrier (BBB) integrity, and neuroimmune signaling. Their role in Alzheimer’s Disease (AD), particularly regarding sex differences and brain region specificity, remains poorly defined. Using single-nucleus RNA sequencing analysis and cell-cell communication tools (LIANA, Tensor-cell2cell), we characterize pericyte transcriptional and signaling changes across the middle temporal gyrus (MTG) and dorsolateral prefrontal cortex (DLPFC) of AD and non-AD donors stratified by sex. We identify a pericyte-endothelial TGFβ signaling program upregulated in female AD donors specifically in the MTG, and a downregulated estrogen pathway involving pericyte-astrocyte ligand-receptor interactions, supported by increased pericyte-astrocyte separation in spatial transcriptomics. We also identify a microglia-to-pericyte program enriched for hypoxia and p53 pathways that is elevated in both sexes with regional specificity. Our findings implicate pericyte-driven communication as a mechanistic contributor to female-biased AD vulnerability, and support sex- and region-aware approaches in neurodegeneration research.

Subject terms: Cellular signalling networks, Computational neuroscience, Neuro-vascular interactions

snRNA-seq and cell-cell communication modeling reveal sex and brain region specific pericyte signaling changes in Alzheimer’s Disease, implicating TGFβ activation, estrogen pathway loss, and inflammatory programs in neurovascular dysfunction.

Introduction

Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and widespread brain pathology, including amyloid plaques, tau tangles, neuroinflammation, and cerebrovascular dysfunction¹. Mounting evidence suggests that vascular pathology is not simply a consequence of AD but may actively contribute to disease onset and progression^2,3. In particular, the neurovascular unit (NVU)—composed of endothelial cells (ECs), pericytes, smooth muscle cells, astrocytes, and microglia—plays a crucial role in maintaining blood-brain barrier (BBB) integrity, regulating cerebral blood flow, and coordinating neuroimmune interactions⁴. Dysregulation of NVU signaling has been implicated in early AD pathology⁵, yet the mechanisms by which specific NVU cell types contribute to these changes remain incompletely understood.

Pericytes, which enwrap capillaries and mediate communication with ECs and glia, are emerging as critical regulators of cerebrovascular health. Experimental studies have shown that pericyte dysfunction can lead to BBB breakdown^6,7, impaired perfusion⁸, and altered inflammatory responses—all features observed in AD brains⁹. However, much of our understanding of pericyte involvement in AD is derived from animal models or bulk tissue analyses, which obscure cell-type-specific and regionally localized interactions. Moreover, the extent to which pericyte signaling pathways differ between sexes remains underexplored, despite the well-established epidemiological observation that women are at greater risk for developing AD and may exhibit distinct pathological features¹⁰.

To address these gaps, we leveraged the Seattle AD Brain Consortium single-nucleus RNA-sequencing (snRNA-seq) data^11,12 from the middle temporal gyrus (MTG) and dorsolateral prefrontal cortex (DLPFC) of the same postmortem human donors to construct a comprehensive, sex-stratified atlas of NVU transcriptional and intercellular signaling profiles in AD. To infer communication networks, we applied Ligand-Receptor Inference Analysis (LIANA)^13–15, a consensus framework that integrates multiple published ligand-receptor resources and scoring methods to robustly infer cell-cell interactions based on transcriptomic data. This allowed us to compute communication scores between all NVU cell types using curated receptor-ligand pairs across a unified interface.

To uncover coordinated patterns of signaling across donors, we further applied Tensor-cell2cell^16,17, a dimensionality reduction framework that constructs a four-dimensional communication tensor—capturing variation across donors, ligand-receptor pairs, and sender-receiver cell-type pairs—and performs tensor decomposition to identify latent communication factors. Each factor represents a distinct pattern of cell-cell signaling activity that may underlie context-specific biological processes, such as disease state, sex, or brain region. We annotated these factors using gene set enrichment analysis (GSEA) and pathway-level footprint enrichment analysis (FEA) to gain mechanistic insight into the signaling pathways likely driving each communication program.

To evaluate the specific role of pericytes in these altered communication landscapes, we implemented a full NVU tensor decomposition in addition to a pairwise pericyte-centric tensor decomposition strategy, running LIANA + Tensor-Cell2cell separately for pericyte-astrocyte, pericyte-endothelial, and pericyte-microglial signaling axes. We then compared communication programs across brain regions and disease contexts. Additionally, we used spatial transcriptomics in a subset of donors to validate a key ligand-receptor pair that was present in both the snRNA-seq and MERFISH datasets, and its proximity relationship in situ.

Our findings reveal brain region- and sex-specific disruptions in pericyte signaling networks in AD, highlighting altered TGF-β and estrogen pathway activity as potential contributors to neurovascular dysfunction and sex-differential vulnerability. This study provides new insight into pericyte-driven communication changes in AD and underscores the importance of incorporating sex and brain region as biological variables in the study of neurodegeneration.

Results

A snRNA-seq dataset of the neurovascular unit from the MTG and DLPFC of the same patients

We utilized a previously published snRNA-seq provided by the Seattle Alzheimer’s Disease Brain Cell Atlas¹¹, which contains data from the MTG and DLPFC of the same 84 donors, 72 of which have a reported diagnosis (Fig. 1A). After filtering and creating subsets that only include astrocytes, pericytes, microglia, and ECs, we retained 57 donors (described in Methods) that included: females with AD (FAD), males with AD (MAD), females with no AD (FNA), and males with no AD (MNA). Given our focus on pericyte signaling programs, we included NVU cell types in contact with pericytes, excluding smooth cells. The percentage of donors from each group that were apolipoprotein E4 (APOE4) carriers did not noticeably change when comparing the original 72 donor dataset to the subsetted 57 donor dataset (Supplementary Fig. 1). Group representation across sex and diagnosis was balanced (Fig. 1B–D), as well as cognitive status and neuropathology between healthy and disease groups (Supplementary Fig. 2). Cell types were confirmed by their enrichment for canonical markers (Supplementary Fig. 3). Since pericytes are found on capillaries, we confirmed that the ECs we used were capillary endothelial cells through strong expression of brain capillary EC markers over arterial and venous markers^18–20 (Supplementary Fig. 4). The total number of cells and genes between the two brain regions was also found to be similar between the two brain regions (Fig. 1C–D). Distribution of cells was found to be similar between groups (Supplementary Table 1). We proceeded with differential gene expression and LIANA with Tensor-Cell2Cell analyses¹⁷ to investigate region- and sex-specific signaling alterations in the NVU. This was followed up with a pericyte-centric pairwise analysis for each of the other cell types (Fig. 1E).

Fig. 1 — A Cell and gene numbers from the full SEA-AD snRNA-Seq dataset from the MTG (green) and DLPFC (purple). B Distribution of donor ages in each group of the NVU cell subset. C Proportions of each NVU cell type in the MTG for each group. D Cell proportions in the DLPFC for each group. E Workflow of the study’s analyses adapted from Baghdassarian, Dimitrov et al.¹⁷and Created in Biorender El-Ghazawi, K. (2025) https://BioRender.com.

NVU cells show a sex specific differentially expressed gene (DEG) profile

In the MTG, we identified a strong sex-specific transcriptional response. Female AD (FAD) donors showed numerous differentially expressed genes (DEGs) in pericytes, ECs, and astrocytes compared to female non-AD (FNA) donors (Fig. 2A–D), while male AD (MAD) donors showed minimal changes. We do not see any overlap in AD changes between female and male donors from any of these cell types. Through overrepresentation analysis, upregulated genes in all cells of FAD donors were found to be enriched in oxidative phosphorylation (Fig. 2B–E, Supplementary Fig. 5) (e.g., MT-ATP6, MT-CO3, MT-ND3) with pericytes also showing increased cellular stress via an upregulation in WWOX²¹ (Fig. 2B). ECs and astrocytes in FAD donors also showed several downregulated genes. In ECs, these downregulated genes were enriched for neurovascular/synaptic signaling (e.g., ROBO1, NTM, SEMA5A)^22–25, cell adhesion (e.g., IL1RAPL1)^26,27, vascular integrity (ELN)²⁸, stress/apoptotic response (e.g., RGS2)²⁹, cytoskeletal/structural stability (e.g., MAP7, TRIM2)^30,31, and metabolism (e.g., TAFA5, SCD5, SLC27A4)^32–34 (Fig. 2C). Downregulated astrocyte genes in FAD donors additionally were associated with metabolism and energy production (PGAM2, ELOVL7)^35,36, cell growth and survival signaling (ROS1)³⁷, and membrane and structural protein/cell adhesion (TM4SF19)³⁸ (Fig. 2D). In contrast, microglia showed robust DEGs in both sexes. (Fig. 2E). We also compared FNA to MNA donors to identify DEGs at baseline between female and male donors (Supplementary Fig. 6). A few of these (e.g., KDM6A, PCDH11Y) directly interact with androgen signaling^39–41 while many others (e.g., DDX3Y, GSTM5)^42–44 are indirectly involved in androgen signaling. This comparison contextualizes some of the sex-specific findings in our subsequent analyses.

Fig. 2 — A UMAP of all NVU cells in the MTG. B DEGs of pericytes from FAD vs. FNA donors and MAD vs. MNA donors. C DEGs of ECs from FAD vs. FNA donors and MAD vs. MNA donors. D DEGs of astrocytes from FAD vs. FNA donors and MAD vs. MNA donors. E DEGs of microglia from FAD vs. FNA donors and MAD vs. MNA donors. P-values were adjusted using the Benjamini-Hochberg method, and thresholds of adjusted-p > 0.05 and |log2FoldChange | > 0.3 were used.

Overall, we report pericyte, EC, astrocyte, and microglial DEGs in the MTG of FAD donors but only microglial DEGs in the MTG of MAD donors. In contrast, in the DLPFC, pericytes show no DEGs in either sex, and ECs express only four DEGs across both sexes (Supplementary Fig. 7). These findings highlight a female-specific, MTG-localized transcriptional vulnerability in pericytes, ECs, and astrocytes in AD, suggesting that NVU dysfunction in women may be associated with coordinated metabolic and structural changes across multiple vascular-associated cell types.

NVU cells show a sex specific disruption of communication patterns

Tensor-cell2cell decomposition of NVU-wide ligand-receptor activity in the MTG revealed two communication programs—Factor 3 and Factor 9—that were significantly altered across sex and diagnosis groups. Factor 3 was characterized by signaling from microglia, astrocytes, and pericytes directed toward ECs (Fig. 3A). Defining this Factor was done through gene set enrichment analysis (GSEA). This identified strong enrichment for serine/threonine kinase signaling, TGFβ signaling, cell migration, and membrane remodeling (Fig. 3B). Factor 3 activity was significantly higher in FAD donors compared to FNA donors (Fig. 3C), suggesting enhanced vascular-targeted signaling in FAD brains. To investigate pathway activity, we performed footprint enrichment analysis (FEA) using PROGENy gene sets on ligand-receptor (LR) loadings derived from Tensor-cell2cell. This analysis revealed strong TGFβ pathway activity (Fig. 3D). To pinpoint upstream ligands from sender cells that may drive this activity via EC receptors, we constructed LR-pathway gene sets based on PROGENy footprints (Fig. 3E).

Fig. 3 — A Tensor factorization of Factor 3. B GSEA of Factor 3 using GO terms, NES, and log(p-value). C Comparison of Factor 3 communication using a t-test between FAD and FNA and between MAD and MNA with Benjamini-Hochberg correction **p < 0.01). D FEA PROGENy barplot of factor 3 with active pathways in red and inactive pathways in blue. E PROGENy plot of LR pairs contributing to TGFß activity, where each dot is a unique LR pair. N = 30 FAD, 4 FNA, 19 MAD, 4 MNA independent donor samples.

This analysis highlighted several extracellular matrix (ECM)-associated ligand-receptor interactions as key contributors of TGFβ pathway activation, including CCN2^ITGA5, FN1^ITGAV-ITGB1, and COL4A1^ITGA1-ITGB1 (Table 1).

Table 1.

LR pairs contributing to TGFß and estrogen pathway activation in the MTG

Factor + Pathway	LR Pair	PROGENy Weight
F3 TGFß	CCN2^ITGA5	4.58
F3 TGFß	COL4A1^ITGA1_ITGB1	3.5
F3 TGFß	FN1^ITGA4_ITGB1	3.12
F3 TGFß	FN1^ITGAV_ITGB1	3.00
F9 estrogen	LAMA3^SDC2	4.25
F9 estrogen	VEGFA^CD44	3.07
F9 estrogen	THBS1^SCARB1	1.60
F9 estrogen	NLGN3^NRXN1	-0.37

Open in a new tab

Factor 9 was also seen to be altered in AD. Factor 9 is involved in signaling from each of the four NVU cells to one another equally (Fig. 4A). GSEA identified this communication pattern to be involved in synapse regulation, assembly, and organization as well as cell-cell adhesion (Fig. 4B). This factor is significantly reduced in FAD donors as compared to FNA donors (Fig. 4C). FEA identified both the WNT and estrogen pathways as highly active in Factor 9 (Fig. 4D). Given the female-specific downregulation of this communication pattern, we chose to focus on the estrogen pathway due to its known roles in synaptic support, vascular integrity, and AD resilience^45,46. LR pathway gene set mapping revealed the estrogen signaling in this factor was shaped by both activating LR pairs involved in ECM remodeling (e.g., LAMA3^SDC2), angiogenesis, and immune modulation (e.g., VEGFA^CD44 and TBS1^SCARB1), as well as inhibitory interactions in synaptic communication (e.g., NRLGN3^NRXN1) (Fig. 4E, Table 1). The reduction in Factor 9 in FAD donors thus reflects a loss of both estrogen-promoting and estrogen-repressing signals, suggesting a broader disruption in pericyte-associated homeostatic signaling.

In the full NVU analysis of the DLPFC, no communication pattern was found to be altered between groups. More specifically, factors that were similar to Factors 3 and 9 of the MTG showed no differences between groups in the DLPFC (Supplementary Fig. 8). Together, these findings suggest that MTG-specific, female-enriched communication programs involving ECM-activated activated TGFβ and estrogen signaling are selectively disrupted in AD, implicating both vascular and synaptic signaling deficits in NVU dysfunction.

Pericyte-EC signaling drives increased TGFß activity in AD relative to FNA donors

To identify how pericytes contribute to the altered communication patterns observed at the NVU level, we conducted a series of pairwise Tensor-cell2cell analyses, modeling pericyte interactions with each of the other NVU cell types in the MTG. This approach allowed us to isolate pericyte-specific signaling dynamics across cell-type pairs. Among these, the pericyte-EC interaction revealed a distinct communication program—Factor 1—that was significantly altered across disease groups. In this factor, pericytes primarily acted as signal senders to EC receivers (Fig. 5A).

Fig. 5 — A Tensor factorization of Factor 1. B GSEA of Factor 1 using GO terms, NES, and log(p-value). C Comparison of Factor 1 communication using a t-test between FAD and FNA and between MAD and MNA with Benjamini-Hochberg correction ***p < 0.001 (**p < 0.01, *p < 0.05). D FEA PROGENy barplot of factor 1 with active pathways in red and inactive pathways in blue. E PROGENy plot of LR pairs contributing to TGFß activity, where each dot is a unique LR pair. F Example LR pairs from Factor 1 with their associated weight. G Expression of *FN1* in pericytes in each group in both MTG and DLPFC using psuedobulk (**p < 0.01). N = 30 FAD, 4 FNA, 19 MAD, 4 MNA independent donor samples.

GSEA of Factor 1 revealed enrichment for pathways associated with kinase signaling, ECM organization, and migration, mirroring the functional categories of Factor 3 from the full NVU model (Fig. 5B, see also Fig. 3B). Factor 1 activity was significantly elevated in both FAD donors compared to FNA, indicating that pericyte-driven EC signaling is upregulated in AD, specifically in females (Fig. 5C). The activated TGFß and related LR pairs also directly mirrored what we observed in Factor 3 of the full NVU analysis (Fig. 5D–F, see also Fig. 3D and Table 1). We found no such changes in the DLPFC when investigating this axis. When evaluating expression levels of the top ligands triggering TGFß activation, we found FN1 overexpression in pericytes of the FAD group relative to FNA donors unique to the MTG (Fig. 5G). These results indicate that the increased Factor 3 activity observed at the NVU level is at least partially driven by pericyte-derived signals to ECs, highlighting a pericyte-specific contribution to vascular ECM-driven TGFβ pathway activation in the MTG that is more pronounced in FAD donors via FN1 signaling to EC integrins. We then performed the same analysis in a separate snRNAseq dataset of the MTG from AD and non-AD male and female donors⁴⁷. Although in a smaller subset of donors, we observed this same trend in a group of donors from a separate cohort (Supplementary Fig. 9).

Pericyte-Astrocyte signaling disruption impairs estrogen activity in FAD donors

We next examined pericyte-astrocyte communication to identify additional pericyte-driven changes relevant to AD. Pairwise Tensor-cell2cell analysis between these cell types revealed a single altered communication program (Factor 9) involving bidirectional signaling between pericytes and astrocytes, as was seen in Factor 9 of the full NVU analysis (Fig. 6A). Consistent with the full NVU analysis, GSEA revealed enrichment for processes including synapse regulation, assembly, and cell-cell adhesion (Fig. 6B). Additionally, Factor 9 activity was significantly reduced in FAD donors compared to FNA donors (Fig. 6C), echoing patterns observed in the full NVU analysis (Fig. 4).

FEA identified the estrogen signaling pathway as active in Factor 9 (Fig. 6D), with contributions from both positive regulators—such as LAMA3^SDC2, VEGFA^CD44, and THBS1^SCARB1—and negative regulators, including NLGN3^NRXN1 (Fig. 6E–F). To determine whether any of these ligand-receptor interactions showed transcriptional changes consistent with the overall disruption of Factor 9 signaling, we examined the expression of all ligands and receptors contributing to estrogen pathway activity. Among these, NLGN3 was the only component with significant expression changes across disease groups. Notably, NLGN3—part of the negatively weighted NLGN3^NRXN1 interaction—was significantly reduced in pericytes from FAD donors relative to FNA, specifically in the MTG (Fig. 6G). While this reduction would imply decreased repression of the estrogen pathway, its presence alongside reduced overall Factor 9 activity suggests a broader collapse of pericyte-astrocyte communication balance. This pattern was not observed in the DLPFC, supporting a region-specific perturbation. Moreover, the selective expression of NLGN3 in pericytes and NRXN1 in astrocytes supports directional signaling from pericytes to astrocytes in this context.

To further explore the NLGN3 signaling pathway, we turned to spatial transcriptomic data from a subset of the same MTG donors. Although NLGN3 is not included in the MERFISH panel, its homolog NLGN1 is. Like NLGN3, NLGN1 is similarly reduced in the FAD group relative to FNA donors (Fig. 6H). Moreover, NLGN1+ pericytes are closest in proximity to astrocytes in FNA donors as compared to other groups (Fig. 6I); confirming the reduced signaling between the cell types in those other contexts. Together, these findings support a model in which loss of pericyte-derived NLGN3 weakens inhibitory control over estrogen signaling within Factor 9, contributing to the broader collapse of pericyte-astrocyte communication observed in the full NVU analysis. This disruption may impair estrogen-mediated synaptic maintenance mechanisms and contribute to the sex-specific vulnerability of the MTG in AD.

Given only two overlapping FNA and MNA donors had spatial transcriptomic data available, no statistical tests were performed, and these results should be interpreted with caution.

Pericyte-Microglia communication pattern is conserved between brain regions and impaired with sex specificity

We also investigated pericyte-microglia communication to determine whether immune-related signaling contributes to disease-associated NVU changes. Pairwise Tensor-cell2cell analysis revealed a single altered communication program—Factor 2—between these cell types in the MTG, with microglia primarily acting as senders and pericytes as receivers (Fig. 7A). GSEA indicated that this factor was enriched for pathways involved in cell migration, immune activation, and inflammatory signaling (Fig. 7B). Factor 2 activity was significantly elevated in FAD donors compared to FNA (Fig. 7C), suggesting enhanced microglia-to-pericyte communication in the FAD brain. FEA identified hypoxia and p53 signaling as among the most active pathways in this factor (Fig. 7D). LR pairs activating hypoxia pathway included ECM interactions such as ADAM17^ITGA5, whereas p53 activation is through lipid related signaling like APP^LRP10 and PLTP^ABCA1 (Fig. 7E–F, Table 2).

Fig. 7 — A Tensor factorization of Factor 2 of the MTG. B GSEA of Factor 2 using GO terms, NES, and log(p-value). C Comparison of Factor 2 communication using a t-test between FAD and FNA and between MAD and MNA with Benjamini-Hochberg correction (**p < 0.01). D FEA PROGENy barplot of factor 2 with active pathways in red and inactive pathways in blue. E PROGENy plot of LR pairs contributing to hypoxia activity, where each dot is a unique LR pair and the same for F p53 activity. M Example LR pairs from both factors with their associated weight. N = 30 FAD, 4 FNA, 19 MAD, 4 MNA independent donor samples.

Table 2.

Microglia-Pericyte LR pairs activating hypoxia and p53 pathways in both MTG and DLPFC

Factor + Pathway	LR Pair	PROGENy Weight
F2 Hypoxia: MTG	ADM^PTH1R	9.21
F5 Hypoxia: DLPFC	ADM^PTH1R	9.21
F2 Hypoxia: MTG	VEGFA^EGFR	6.28
F5 Hypoxia: DLPFC	VEGFA^EGFR	6.28
F2 Hypoxia: MTG	ADAM17^ITGA5	3.54
F5 Hypoxia: DLPFC	ADAM17^ITGA5	3.54
F2 p53: MTG	PLTP^ABCA1	5.22
F5 p53: DLPFC	PLTP^ABCA1	5.22
F2 p53: MTG	APP^LRP10	2.93
F5 p53: DLPFC	APP^LRP10	2.93
F2 p53: MTG	APOE^ABCA1	2.32
F5 p53: DLPFC	APOE^ABCA1	2.32

Open in a new tab

To evaluate whether this immune-pericyte signaling program was specific to the MTG or conserved across brain regions, we analyzed the DLPFC using the same approach. Remarkably, the same communication pattern with microglia sending to pericytes and enriched for the same inflammatory pathways—was also observed in the DLPFC (here as Factor 5) (Fig. 8A–F). However, in this region, Factor 5 was elevated in only in MAD donors compared to MNA donors, indicating that the female-specificity seen in the MTG is not preserved in the DLPFC (Fig. 8C).

Fig. 8 — A Tensor factorization of Factor 5 of the DLPFC. B GSEA of Factor 5 using GO terms, NES, and log(p-value). C Comparison of Factor 5 communication using a t-test between FAD and FNA and between MAD and MNA with Benjamini-Hochberg correction (**p < 0.01). D FEA PROGENy barplot of factor 5 with active pathways in red and inactive pathways in blue. E PROGENy plot of LR pairs contributing to Hypoxia activity, where each dot is a unique LR pair and the same for F p53 activity. N = 30 FAD, 4 FNA, 19 MAD, 4 MNA independent donor samples.

In both regions, hypoxia and p53 signaling remained highly active, and the contributing LR interactions and weights were nearly identical between MTG and DLPFC (Fig. 8D–F, Table 2). Together, these findings suggest that microglia-driven inflammatory signaling to pericytes is a conserved feature of AD pathology across these two brain regions, but that the magnitude and sex specificity of this communication varies by brain region. These results reveal a conserved microglia-to-pericyte inflammatory signaling program—absent in the full NVU analysis but uncovered through pairwise modeling—that is enriched for hypoxia and p53 activity and may contribute to pericyte dysfunction in AD. While this communication axis is present in both MTG and DLPFC, its female specificity in the MTG and in males in the DLPFC suggest that microglia-driven pericyte signaling may contribute to sex- and region-specific patterns of neurovascular vulnerability.

Discussion

In this study, we applied a multimodal computational framework to dissect cell-cell communication within the human NVU in AD, with a focus on sex- and region-specific differences. Using snRNA-seq data from the MTG and DLPFC of the same donors, we combined DEG analysis with ligand-receptor inference via LIANA, latent communication modeling via Tensor-cell2cell, pathway activity estimation via FEA, and spatial transcriptomics. This integrative approach enabled us to identify transcriptional and communication-level shifts that would not be evident from gene expression alone⁴⁸.

Our DEG analysis revealed a sex- and region-specific pattern: FAD donors exhibited widespread transcriptional changes in pericytes, astrocytes, and ECs in the MTG, while MAD donors only showed changes in microglia of the MTG. In the DLPFC, changes were primarily seen in microglia with fewer in astrocytes of MAD donors. Building on these transcriptional patterns, we identified two latent intercellular communication programs of the full NVU—Factor 3 and Factor 9—that were significantly altered in the MTG of FAD donors, but not in the DLPFC.

The marked asymmetry in DEGs between female and male donors in the MTG underscores a pronounced sex-specific molecular response in AD. Although not evaluated in MTG, recent studies have also shown different transcriptional profiles between female and male AD, with female AD patients having more DEGs^49–51. In our findings in the MTG, pericytes, astrocytes, and ECs from FAD donors exhibited coordinated transcriptional changes, including upregulation of oxidative phosphorylation and stress-response genes and downregulation of adhesion and metabolic regulators. In contrast, these cell types in MAD donors showed minimal DEG signatures in the MTG. Previous work has similarly identified female-specific DEG profiles for EC, microglia, and astrocytes; although pericytes were not evaluated⁵². While DEG analysis in the DLPFC revealed limited transcriptional changes overall, we did observe a set of DEGs in astrocytes from MAD donors, suggesting a cell-type and sex-specific signal in that region. This transcriptional landscape provided the foundation for the subsequent communication modeling, where altered signaling programs were found exclusively in the MTG and primarily in FAD donors.

The alignment between cell-type-specific DEG patterns and Tensor-cell2cell factor changes suggest that latent communication alterations are not artifacts of modeling but reflect underlying shifts in gene expression relevant to ligand and receptor availability. For example, the observed upregulation of ECM components such as FN1 in FAD pericytes corresponded with the increased activity of a pericyte-EC signaling program driving TGFβ activation. Conversely, the loss of NLGN3 expression in pericytes paralleled the reduction in an estrogen-associated pericyte-astrocyte communication pattern. These findings illustrate how transcriptional dysregulation in AD—particularly in FAD donors and in the MTG—manifests at the level of intercellular signaling and reinforces the biological relevance of stratifying analyses by both sex and brain region. These differences can be attributed to the androgen-driven factor differences that we see in baseline male and female donors (Supplementary Fig. 3).

Among the altered communication programs identified in the MTG, Factor 3 was characterized by signaling from pericytes, astrocytes, and microglia to ECs and was strongly enriched for TGFβ signaling, cell migration, and ECM remodeling. These processes have been suggested as contributors to neurovascular dysfunction in AD⁵³. Importantly, this signal was selectively increased in FAD donors, suggesting that dysregulated signaling toward the vascular endothelium may be sex-specific and regionally confined. FEA further confirmed robust activation of the TGFβ pathway, pointing to ECM-integrin interactions as primary upstream drivers.

To determine the predominate cellular source of this signaling, we performed pericyte-centric Tensor-cell2cell pairwise analyses and identified a pericyte-EC communication program (Factor 1) that mirrored the functional and pathway enrichment profiles of Factor 3. This pairwise analysis confirmed that pericytes are the dominant senders, engaging EC integrins through ligands such as FN1, COL4A1, and CCN2. Expression analysis revealed that FN1 was significantly upregulated in pericytes from FAD donors, but only in the MTG, supporting a region-specific pericyte-driven mechanism of TGFβ activation. These findings position pericytes as key initiators of vascular remodeling in AD, particularly in female donors, and highlight the potential for ECM-integrin signaling axes to mediate neurovascular dysfunction via TGFß activation. ECM-integrin interactions are known to coordinate the release of TGFß from latent TGFß binding proteins in the ECM⁵⁴, and this ECM driven activation of TGFß has been previously described in the vascular context for Vascular Dementia⁵⁵, but not directly in AD or with any indication of sex and region specificity. It has also yet to be described in the context of pericyte-EC interactions in the brain.

The second major communication pattern altered in the MTG of FAD donors was Factor 9, which involved distributed signaling between all NVU cell types but was particularly enriched for pericyte-astrocyte interactions and processes related to synapse organization and cell-cell adhesion. FEA of this factor revealed activity of both the estrogen and WNT signaling pathways. Estrogen plays well-established roles in supporting synaptic plasticity, neuroprotection, and cerebrovascular integrity—functions that closely align with the processes enriched in this communication pattern^56–59.

Interestingly, Factor 9 included both activating and inhibitory ligand-receptor pairs contributing to estrogen signaling. To prioritize the interactions most relevant to the observed reduction in Factor 9 activity in FAD donors, we examined the expression of all estrogen-associated ligands and receptors. NLGN3^NRXN1 emerged as the only pair showing a transcriptional change consistent with the communication loss: NLGN3 was significantly downregulated in pericytes from all disease groups compared to FNA donors, specifically in the MTG. NLGN3 carried a negative weight in the PROGENy model—suggesting it acts as a repressor of estrogen signaling—its downregulation occurred alongside a broader reduction in Factor 9 activity and in other estrogen-associated LR pairs. This suggests a general collapse of estrogen-linked signaling, rather than a compensatory disinhibition of the pathway.

To further support this model, we leveraged spatial transcriptomic data. While NLGN3 was not present in the MERFISH panel, we used its close homolog, NLGN1, as a proxy. NLGN1 was similarly downregulated in pericytes from all disease groups relative to FNA donors, and NLGN1+ pericytes were located closer to astrocytes in FNA donors than in AD groups. These findings provide spatial and transcriptional evidence for reduced pericyte-astrocyte neuroligin-neurexin interactions in AD. Together, these results point to a selective loss of estrogen-associated pericyte-astrocyte communication in the MTG of female AD donors, potentially contributing to sex-specific deficits in synaptic maintenance and neurovascular support. The estrogen pathway has been suggested to play a role in AD prevention⁶⁰, but its relation to neurovascular functions is still not well understood. Although NLGN1 has previously been shown to be decreased in female AD models⁶¹, its specific role in pericyte-astrocyte signaling is not known, nor is the role of NLGN3 in AD. More work is needed to understand how this signaling pathway relates to the estrogen pathway in AD.

In addition to the sex-specific communication programs identified in the MTG, our analysis uncovered a conserved inflammatory signaling axis between microglia and pericytes through Factor 2. In both the MTG and DLPFC, this factor was characterized by microglia acting as dominant senders and pericytes as receivers. GSEA and FEA linked this communication program to immune activation, cell migration, and signaling through hypoxia and p53 pathways—processes associated with vascular injury, stress response, and neuroinflammation in AD.

Although the structure of this communication pattern was conserved across regions, its magnitude and sex specificity were variable. In the MTG, Factor 2 activity was elevated specifically in FAD donors, while in the DLPFC, the same pattern (Factor 5) was elevated in MAD donors. This variation may reflect differential microglial reactivity, regional vulnerability of the vasculature, or local environmental cues influencing immune-vascular crosstalk. Microglia have been shown to have a sex- and regional- specific transcriptional profile and injury response in the brain^62–64. They are also known to be in contact with pericytes⁶⁵, but more work is needed to understand the signaling pathways between these cells and their specificity to sex and brain region.

Both p53 and hypoxia are central microglial pathways in AD, which makes sense why they are seen to be conserved across regions^66–70. Conservation of these pathways across regions and contexts also reinforces the robustness of our modeling framework and highlights pericytes as key downstream targets of microglial-derived inflammatory signals, motivating avenues of future work to determine pericyte involvement with microglia in sex- and region- specific facets.

Collectively, our findings reveal that pericyte-centered communication programs in the MTG are selectively disrupted in FAD donors. These include a loss of estrogen-linked pericyte-astrocyte signaling, enhanced ECM-driven pericyte-EC TGFβ activation and increased immune cell-to-pericyte signaling from microglia. While estrogen pathway disruption may contribute to synaptic vulnerability in women, the enhanced TGFβ signaling may reflect a parallel sex-biased mechanism of vascular remodeling. Together, these communication programs may help explain the greater burden of neurovascular and cognitive decline observed in women⁷¹.

Our results also emphasize the importance of regional stratification in neurodegenerative research. The MTG emerged as a uniquely affected region, with both transcriptional and cell-cell communication changes that were not observed in the DLPFC. While the DLPFC is commonly included in transcriptomic studies of AD^72–75, our data suggest that the MTG may be more sensitive to NVU dysfunction, especially in females. These findings support the need for future studies to more deeply investigate the MTG, particularly in relation to sex-specific pathways, neurovascular health, and synaptic integrity. More broadly, our work underscores the value of integrating single-cell, communication modeling, and spatial approaches to resolve subtle but disease-relevant signaling alterations that span both transcriptional and intercellular scales. This work can be expanded with the use of experimental validations and with computational tissue-level modeling, as our group has previously performed in other disease conditions^76,77, to further identify key contextual contributors in AD.

While our analysis provides new insights into context-specific neurovascular signaling in AD, several limitations should be acknowledged. First, our inference of cell-cell communication is based on transcriptomic data and curated ligand-receptor databases, which do not account for post-transcriptional regulation, protein abundance, or ligand activation states. Consequently, predicted interactions may not fully represent functional signaling. Second, although we applied both pairwise and full NVU models, these remain computational and do not capture spatial proximity or biophysical constraints in vivo. Third, the analysis is based on post-mortem human brain tissue from the MTG and DLPFC, representing a static snapshot of signaling and limiting generalizability to other brain regions or disease stages. Moreover, while donor-matched samples help control for inter-individual variability, sampling remains limited in size and diversity. Finally, although we observed strong convergence between pericyte-centric and broader NVU signaling programs, we did not perform experimental validation in human tissue at either the transcriptomic or proteomic levels. Future studies integrating spatial transcriptomics, proteomics, and functional validation will be critical to confirm the directionality and functional consequences of these interactions.

This study integrates snRNA-seq differential expression, ligand–receptor inference, latent communication modeling, and spatial transcriptomics to uncover sex- and region-specific neurovascular signaling changes in AD. It identifies pericytes as central mediators of vascular remodeling and synaptic vulnerability in the female MTG through TGFβ activation and estrogen-linked signaling loss. These findings underscore the importance of stratifying analyses by brain region and sex to resolve context-specific mechanisms of neurovascular dysfunction in AD.

Methods

Donor samples and dataset description

We analyzed publicly available single-nucleus RNA-sequencing (snRNA-seq) data from the Seattle Alzheimer’s Disease Brain Cell Atlas, which includes matched samples from the middle temporal gyrus (MTG) and dorsolateral prefrontal cortex (DLPFC) of 72 postmortem human donors. We focused on a subset of 57 donors who passed quality control and for whom all relevant metadata (sex, diagnosis, brain region) were available. Donors were originally classified based on Alzheimer’s Disease Neuropathologic Change (ADNC) scores into four categories: No, Low, Intermediate, and High. We excluded donors with Low ADNC and combined Intermediate and High ADNC groups into a single AD group, while retaining the No ADNC group as Not AD. Final groups were then constructed based on sex and AD status, resulting in four groups: female AD (FAD), female not AD (FNA), male AD (MAD), and male not AD (MNA). A subset of these donors (N = 16) also had matched spatial transcriptomic data from the MTG using MERFISH, including 2 FNA, 2 MNA, 5 FAD and 7 MAD donors.

snRNAseq preprocessing and cell type filtering

We performed preprocessing using the Scanpy Python package (v1.10.2). We retained only astrocytes, pericytes, microglia, and ECs based on annotations provided in the original dataset. Marker gene expression across cell types was used to confirm the annotations. Donors with less than four of any cell type were removed. Cells were filtered using standard quality control metrics, removing those with fewer than 200 detected genes or more than 15% mitochondrial content. Doublets were removed by removing cells with a high number of unique molecular identifiers ( > 5500). Data were normalized using total-count normalization and log1p transformation. Batch effect was accounted for by removing batches contained less than two cells. Highly variable genes were identified to guide dimensionality reduction and visualization, but all genes were retained for differential expression and communication analyses.

Differential gene expression analysis

Pseudobulk expression profiles were generated for each donor, brain region, and cell type by summing raw gene counts across cells. Differential expression analysis was performed using PyDESeq2 (v0.4.12), comparing FAD vs. FNA and MAD vs. MNA within each brain region. Wald tests were used to assess statistical significance of model coefficients, and p-values were adjusted using the Benjamini-Hochberg method. Genes were considered differentially expressed if they met a threshold of adjusted p < 0.05 and absolute log2 fold-change > 0.3. GSEA was used to annotate differentially expressed genes using Gene Ontology (GO) Biological Process terms.

Cell-Cell communication inference: LIANA + Tensor cell2cell

To infer cell-cell communication programs, we used log-normalized expression with LIANA (v1.4.0) with the consensus scoring method and the Omnipath ligand-receptor resource. The resulting interaction scores were formatted into a 4D tensor (samples × ligand-receptor pairs × sender cell types × receiver cell types) and decomposed using Tensor-cell2cell (v0.7.4) to extract latent communication factors as previously described¹⁷. Separate tensor decompositions were performed for the MTG and DLPFC using the full NVU cell set.

Gene set enrichment and footprint enrichment analyses

We used GSEA (via gseapy v1.0.3) to annotate communication factors with GO Biological Process terms. To identify signaling pathway activity, we performed FEA using PROGENy gene sets. Ligand-receptor pairs were weighted by their loadings in each factor, and enrichment was assessed using a weighted Kolmogorov-Smirnov test with 1000 permutations. Pathways with adjusted p < 0.05 were considered significantly enriched.

Pairwise pericyte-centric Tensor cell2cell analysis

To evaluate pericyte-specific signaling axes, we created three pairwise subsets of the SEA-AD dataset, each containing pericytes and one of the other NVU cell types (astrocytes, ECs, or microglia). For each pairwise subset, we ran the same LIANA + Tensor-cell2cell pipeline as was used in the full NVU analysis. Log-normalized expression was used as input for LIANA, followed by formatting of interaction scores into a 4D tensor (samples × ligand-receptor pairs × sender × receiver). Tensor decomposition was then applied to extract latent communication patterns. Factor annotations were made using GSEA and FEA and compared with NVU-level factors by correlating LR loading vectors and overlapping enriched pathways.

Spatial transcriptomic analysis

We analyzed spatial transcriptomic data from the same donors using MERFISH measurements of the MTG. Only genes included in the MERFISH panel (180 genes) were used, and NLGN3 was not among them. Instead, NLGN1 was used as a proxy. Gene expression was normalized to 10,000 counts per cell and log-transformed prior to group-level comparisons. To assess pericyte-astrocyte spatial relationships, we calculated the distance between each astrocyte and its nearest NLGN1+ pericyte using spatial transcriptomics data from the MTG. NLGN1+ pericytes were defined as pericytes with nonzero expression of NLGN1. Spatial coordinates were extracted from the MERFISH dataset, and distances were computed using Euclidean distance via a k-d tree implementation (scipy.spatial.cKDTree). For each astrocyte, the shortest distance to the nearest NLGN1+ pericyte was recorded. Group-wise comparisons were visualized using violin and strip plots. Statistical testing was not performed, as MNA and FNA groups contained only 2 donors each.

Statistical analysis and reproducibility

All statistical analyses were conducted using Python packages, including scanpy, scipy, and statsmodels. Visualizations were generated using matplotlib, seaborn, and plotnine. Significance was defined as p < 0.05 unless otherwise indicated. All enrichment p-values were corrected for multiple testing using the Benjamini-Hochberg method. Post hoc pairwise comparisons were performed using Dunn’s test when applicable. FAD donors (N = 30) were compared directly to FNA donors (N = 4) for the MTG and DLPFC independently, and MAD donors (N = 19) were compared directly to MNA donors (N = 4) for the MTG and DLPFC independently as described above.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplemental Information^{(3.4MB, pdf)}

Reporting summary^{(2MB, pdf)}

Acknowledgements

We would like to acknowledge and thank the donors who are a part of SEA-AD consortium. We thank the Allen Brain Map Community Forum for help with data access and interpretation. Funding sources included: UVA Brain Institute Pilot Transformative Neuroscience Grant (K.E., S.M.P., U.B.E.), HL137112 (BEI), and HL165143 (BEI). Additionally, we thank Dr. Le Zhang and Dr. Stephen Strittmatter at Yale University for making the additional MTG data available and for providing the necessary metadata.

Author contributions

K.E.: conceived the project, designed the study, wrote the code, conducted analyses, interpreted analyses, created visualizations, wrote the initial draft, and reviewed and edited subsequent drafts. Y.T.A.: contributed to the initial code, contributed to study design, and revised the initial draft. Y.V. and W.A.F.: contributed to interpretation of analyses, contributed to the initial draft, and revised the initial draft. B.E.I. and H.S.: revised the initial draft. U.B.E. and S.M.P: supervision, contributed to study design, contributed to visualization design, and edited the original draft.

Peer review

Peer review information

Communications Biology thanks Leyla Anne Akay and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Benjamin Bessieres.

Funding

K.E., S.M.P., and U.B.E. disclose support for the research from the UVA Brain Institute Pilot Transformative Neuroscience Grant. K.E. discloses support for the research from the NIH T32 Biotechnology for Human Health Predoctoral Grant (NIH-T32GM136615). U.B.E. discloses support for the research from R01 #5R01NS122782-05.

Data availability

RNAseq and spatial transcriptomic datasets were obtained from the Seattle Alzheimer’s Disease Brain Cell Atlas through their public registry of open data on AWS: https://registry.opendata.aws/allen-sea-ad-atlas/. The additional MTG data is available through GEO: GSE188545. Any additional data is available by request.

Code availability

The LIANA x Tensor-cell2cell tutorial is available online: https://ccc-protocols.readthedocs.io/en/latest/index.html. An annotated version of the custom script used for this is available on Zenodo⁷⁸: https://doi.org/10.5281/zenodo.16760785.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Contributor Information

Ukpong B. Eyo, Email: ube9q@virginia.edu

Shayn M. Peirce, Email: smp6p@virginia.edu

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10446-y.

References

1.Iadecola, C. & Gottesman, R. F. Cerebrovascular alterations in Alzheimer’s disease: incidental or pathogenic? Circ. Res.123, 406–408 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Strickland, S. Blood will out: vascular contributions to Alzheimer’s disease. J. Clin. Invest.128, 556–563 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Eisenmenger, L. B. et al. Vascular contributions to Alzheimer’s disease. Transl. Res. J. Lab. Clin. Med.254, 41–53 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Schaeffer, S. & Iadecola, C. Revisiting the neurovascular unit. Nat. Neurosci.24, 1198–1209 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Gerrits, E. et al. Neurovascular dysfunction in GRN-associated frontotemporal dementia identified by single-nucleus RNA sequencing of human cerebral cortex. Nat. Neurosci.25, 1034–1048 (2022). [DOI] [PubMed] [Google Scholar]
6.Benarroch, E. What Are the Roles of Pericytes in the Neurovascular Unit and Its Disorders? Neurology100, 970–977 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Li, M. et al. Microvascular and cellular dysfunctions in Alzheimer’s disease: an integrative analysis perspective. Sci. Rep.14, 20944 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Hartmann, D. A. et al. Brain capillary pericytes exert a substantial but slow influence on blood flow. Nat. Neurosci.24, 633–645 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.El-Ghazawi, K., Eyo, U. B. & Peirce, S. M. Brain microvascular pericyte pathology linking Alzheimer’s disease to diabetes. Microcirculation31, e12877 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Moutinho, S. Women twice as likely to develop Alzheimer’s disease as men — but scientists do not know why. Nat. Med.31, 704–707 (2025). [DOI] [PubMed] [Google Scholar]
11.Gabitto, M. I. et al. Integrated multimodal cell atlas of Alzheimer’s disease. Nat. Neurosci.27, 2366–2383 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Seattle Alzheimer’s Disease Brain Cell Atlas (SEA-AD) - Registry of Open Data on AWS. https://registry.opendata.aws/allen-sea-ad-atlas/.
13.Dimitrov, D. et al. Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. Nat. Commun.13, 3224 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Dimitrov, D. et al. LIANA+ provides an all-in-one framework for cell–cell communication inference. Nat. Cell Biol.26, 1613–1622 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.LIANA: a LIgand-receptor ANalysis frAmework. https://saezlab.github.io/liana/.
16.Armingol, E. et al. Context-aware deconvolution of cell–cell communication with Tensor-cell2cell. Nat. Commun.13, 3665 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Baghdassarian, H. M., Dimitrov, D., Armingol, E., Saez-Rodriguez, J. & Lewis, N. E. Combining LIANA and Tensor-cell2cell to decipher cell-cell communication across multiple samples. Cell Rep. Methods4, (2024). [DOI] [PMC free article] [PubMed]
18.Vanlandewijck et al. A molecular atlas of cell types and zonation in the brain vasculature. Nature554, 475–480 (2018). [DOI] [PubMed] [Google Scholar]
19.Yang, A. C. et al. A human brain vascular atlas reveals diverse mediators of Alzheimer’s risk. Nature603, 885–892 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Chen, M. B. et al. Brain endothelial cells are exquisite sensors of age-related circulatory cues. Cell Rep.30, 4418–4432.e4 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Liu, S.-Y., Chiang, M.-F. & Chen, Y.-J. Role of WW domain proteins WWOX in development, prognosis, and treatment response of glioma. Exp. Biol. Med.240, 315–323 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Gangaraju, S. et al. Cerebral endothelial expression of Robo1 affects brain infiltration of polymorphonuclear neutrophils during mouse stroke recovery. Neurobiol. Dis.54, 24–31 (2013). [DOI] [PubMed] [Google Scholar]
23.Venkannagari, H. et al. Highly conserved molecular features in IgLONs contrast their distinct structural and biological outcomes. J. Mol. Biol.432, 5287–5303 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Duan, Y. et al. Semaphorin 5A inhibits synaptogenesis in early postnatal- and adult-born hippocampal dentate granule cells. eLife3, e04390 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Sadanandam, A., Rosenbaugh, E. G., Singh, S., Varney, M. & Singh, R. K. Semaphorin 5A promotes angiogenesis by increasing endothelial cell proliferation, migration, and decreasing apoptosis. Microvasc. Res.79, 1–9 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Il1rapl1 interleukin 1 receptor accessory protein-like 1 [Mus musculus (house mouse)] - Gene - NCBI. https://www.ncbi.nlm.nih.gov/gene/331461.
27.Valnegri, P. et al. The X-linked intellectual disability protein IL1RAPL1 regulates excitatory synapse formation by binding PTPδ and RhoGAP2. Hum. Mol. Genet.20, 4797–4809 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Duca, L. et al. Matrix ageing and vascular impacts: focus on elastin fragmentation. Cardiovasc. Res.110, 298–308 (2016). [DOI] [PubMed] [Google Scholar]
29.Wang, C.-H. J. & Chidiac, P. RGS2 promotes the translation of stress-associated proteins ATF4 and CHOP via its eIF2B-inhibitory domain. Cell. Signal.59, 163–170 (2019). [DOI] [PubMed] [Google Scholar]
30.Adler, A. et al. A structural and dynamic visualization of the interaction between MAP7 and microtubules. Nat. Commun.15, 1948 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Ylikallio, E. et al. Deficiency of the E3 ubiquitin ligase TRIM2 in early-onset axonal neuropathy. Hum. Mol. Genet.22, 2975–2983 (2013). [DOI] [PubMed] [Google Scholar]
32.Wesołek-Leszczyńska, A., Pastusiak, K., Bogdański, P. & Szulińska, M. Can adipokine FAM19A5 be a biomarker of metabolic disorders?. Diab. Metab. Syndr. Obes.17, 1651–1666 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Zhang, Q., Sun, S., Zhang, Y., Wang, X. & Li, Q. Identification of Scd5 as a functional regulator of visceral fat deposition and distribution. iScience25, 103916 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Li, Z. et al. SLC27A4-mediated selective uptake of mono-unsaturated fatty acids promotes ferroptosis defense in hepatocellular carcinoma. Free Radic. Biol. Med.201, 41–54 (2023). [DOI] [PubMed] [Google Scholar]
35.Li, Z. et al. Targeting the metabolic enzyme PGAM2 overcomes enzalutamide resistance in castration-resistant prostate cancer by inhibiting BCL2 signaling. Cancer Res.83, 3753–3766 (2023). [DOI] [PubMed] [Google Scholar]
36.Wang, X., Yu, H., Gao, R., Liu, M. & Xie, W. A comprehensive review of the family of very-long-chain fatty acid elongases: structure, function, and implications in physiology and pathology. Eur. J. Med. Res.28, 532 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Acquaviva, J., Wong, R. & Charest, A. The multifaceted roles of the receptor tyrosine kinase ROS in development and cancer. Biochim. Biophys. Acta BBA - Rev. Cancer1795, 37–52 (2009). [DOI] [PubMed] [Google Scholar]
38.Ding, L. et al. TM4SF19 aggravates LPS-induced attenuation of vascular endothelial cell adherens junctions by suppressing VE-cadherin expression. Biochem. Biophys. Res. Commun.533, 1204–1211 (2020). [DOI] [PubMed] [Google Scholar]
39.Hua, C. et al. KDM6 demethylases and their roles in human cancers. Front. Oncol.11, 779918 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Nickerson, M. L. et al. TET2 binds the androgen receptor and loss is associated with prostate cancer. Oncogene36, 2172–2183 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Pecoraro, G. et al. Co-modulation of a circular form of PCDH11Y during neuroendocrine differentiation of prostate cancer. Front. Oncol.15, 1502405 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Zeng, Q., Yi, H., Huang, L., An, Q. & Wang, H. Reduced testosterone and Ddx3y expression caused by long-term exposure to arsenic and its effect on spermatogenesis in mice. Environ. Toxicol. Pharmacol.63, 84–91 (2018). [DOI] [PubMed] [Google Scholar]
43.Vellky, J. E., McSweeney, S. T., Ricke, E. A. & Ricke, W. A. RNA-binding protein DDX3 mediates posttranscriptional regulation of androgen receptor: A mechanism of castration resistance. Proc. Natl. Acad. Sci. USA117, 28092–28101 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Martínez-Rivera, F. J. et al. Differential protein expression profile in the hypothalamic GT1-7 cell line after exposure to anabolic androgenic steroids. PLoS One12, e0180409 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Zhao, L., Woody, S. K. & Chhibber, A. Estrogen receptor β in Alzheimer’s disease: from mechanisms to therapeutics. Ageing Res. Rev.24, 178–190 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Jett, S. et al. Endogenous and Exogenous Estrogen Exposures: How Women’s Reproductive Health Can Drive Brain Aging and Inform Alzheimer’s Prevention. Front. Aging Neurosci. 14, 831807 (2022). [DOI] [PMC free article] [PubMed]
47.Zhang, L. et al. Single-cell transcriptomic atlas of Alzheimer’s disease middle temporal gyrus reveals region, cell type and sex specificity of gene expression with novel genetic risk for MERTK in female. 2023.02.18.23286037 Preprint at 10.1101/2023.02.18.23286037 (2023). [DOI] [PMC free article] [PubMed]
48.Su, J. et al. Cell–cell communication: new insights and clinical implications. Signal Transduct. Target. Ther.9, 196 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.National Academies of Sciences, E. et al. Transcriptomic evidence for sex differences in neurodevelopmental and neurodegenerative disorders. in Sex Differences in Brain Disorders: Emerging Transcriptomic Evidence: Proceedings of a Workshop (National Academies Press (US), 2021). [PubMed]
50.Soudy, M., Bars, S. L. & Glaab, E. Sex-dependent molecular landscape of Alzheimer’s disease revealed by large-scale single-cell transcriptomics. Alzheimers Dement21, e14476 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.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). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Sanfilippo, C. et al. Middle-aged healthy women and Alzheimer’s disease patients present an overlapping of brain cell transcriptional profile. Neuroscience406, 333–344 (2019). [DOI] [PubMed] [Google Scholar]
53.Grammas, P. & Ovase, R. Cerebrovascular transforming growth factor-β contributes to inflammation in the Alzheimer’s disease brain. Am. J. Pathol.160, 1583–1587 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Saharinen, J., Hyytiäinen, M., Taipale, J. & Keski-Oja, J. Latent transforming growth factor-beta binding proteins (LTBPs)-structural extracellular matrix proteins for targeting TGF-beta action. Cytokine Growth Factor Rev.10, 99–117 (1999). [DOI] [PubMed] [Google Scholar]
55.Kandasamy, M. et al. TGF-β signaling: a therapeutic target to reinstate regenerative plasticity in vascular dementia?. Aging Dis.11, 828–850 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Russell, J. K., Jones, C. K. & Newhouse, P. A. The role of estrogen in brain and cognitive aging. Neurotherapeutics16, 649–665 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Zárate, S., Stevnsner, T. & Gredilla, R. Role of Estrogen and Other Sex Hormones in Brain Aging. Neuroprotection and DNA Repair. Front. Aging Neurosci. 9, 430 (2017). [DOI] [PMC free article] [PubMed]
58.Sheppard, P. A. S., Choleris, E. & Galea, L. A. M. Structural plasticity of the hippocampus in response to estrogens in female rodents. Mol. Brain12, 22 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Bustamante-Barrientos, F. A. et al. The impact of estrogen and estrogen-like molecules in neurogenesis and neurodegeneration: beneficial or harmful? Front. Cell. Neurosci. 15, 636176 (2021). [DOI] [PMC free article] [PubMed]
60.Wharton, W. et al. Potential role of estrogen in the pathobiology and prevention of Alzheimer’s disease. Am. J. Transl. Res.1, 131–147 (2009). [PMC free article] [PubMed] [Google Scholar]
61.Dufort-Gervais, J. et al. Neuroligin-1 is altered in the hippocampus of Alzheimer’s disease patients and mouse models, and modulates the toxicity of amyloid-beta oligomers. Sci. Rep.10, 6956 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Brandi, E. et al. Brain region-specific microglial and astrocytic activation in response to systemic lipopolysaccharides exposure. Front. Aging Neurosci. 14, 910988 (2022). [DOI] [PMC free article] [PubMed]
63.Grabert, K. et al. Microglial brain region-dependent diversity and selective regional sensitivities to ageing. Nat. Neurosci.19, 504–516 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Barko, K. et al. Brain region- and sex-specific transcriptional profiles of microglia. Front. Psychiatry13, 945548 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Morris, G. P. et al. Microglia directly associate with pericytes in the central nervous system. Glia71, 1847–1869 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Davenport, C. M., Sevastou, I. G., Hooper, C. & Pocock, J. M. Inhibiting p53 pathways in microglia attenuates microglial-evoked neurotoxicity following exposure to Alzheimer peptides. J. Neurochem.112, 552–563 (2010). [DOI] [PubMed] [Google Scholar]
67.Aloi, M. S., Su, W. & Garden, G. A. The p53 Transcriptional Network Influences Microglia Behavior and Neuroinflammation. Crit. Rev. Immunol.35, 401–415 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Jebelli, J., Hooper, C. & Pocock, J. M. Microglial p53 activation is detrimental to neuronal synapses during activation-induced inflammation: Implications for neurodegeneration. Neurosci. Lett.583, 92–97 (2014). [DOI] [PubMed] [Google Scholar]
69.Butturini, E., Boriero, D., Carcereri de Prati, A. & Mariotto, S. STAT1 drives M1 microglia activation and neuroinflammation under hypoxia. Arch. Biochem. Biophys.669, 22–30 (2019). [DOI] [PubMed] [Google Scholar]
70.March-Diaz, R. et al. Hypoxia compromises the mitochondrial metabolism of Alzheimer’s disease microglia via HIF1. Nat. Aging1, 385–399 (2021). [DOI] [PubMed] [Google Scholar]
71.Levine, D. A. et al. Sex differences in cognitive decline among US adults. JAMA Netw. Open4, e210169 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Mathys, H. et al. Single-cell multiregion dissection of Alzheimer’s disease. Nature632, 858–868 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Mathys, H. et al. Single-cell atlas reveals correlates of high cognitive function, dementia, and resilience to Alzheimer’s disease pathology. Cell186, 4365–4385.e27 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Olayinka, O. A. et al. Single nucleus RNA sequencing reveals interaction between microglia and endothelial cells in mixed Alzheimer’s disease and vascular pathology. Alzheimers Dement20, e089628 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Wang, X. et al. Deciphering cellular transcriptional alterations in Alzheimer’s disease brains. Mol. Neurodegener.15, 38 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.Leonard-Duke, J. et al. Multiscale computational model predicts how environmental changes and treatments affect microvascular remodeling in fibrotic disease. PNAS Nexus4, pgae551 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Rikard, S. M. et al. Mathematical model predicts that acceleration of diabetic wound healing is dependent on spatial distribution of VEGF-A mRNA (AZD8601). Cell. Mol. Bioeng.14, 321–338 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
78.El-Ghazawi, K., Aberra T. Y. & Peirce M. S. Full NVU Tensor C2C of SEA-AD DLPFC. https://zenodo.org/records/16760785 (2025).

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Information^{(3.4MB, pdf)}

Reporting summary^{(2MB, pdf)}

Data Availability Statement

[CR1] 1.Iadecola, C. & Gottesman, R. F. Cerebrovascular alterations in Alzheimer’s disease: incidental or pathogenic? Circ. Res.123, 406–408 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Strickland, S. Blood will out: vascular contributions to Alzheimer’s disease. J. Clin. Invest.128, 556–563 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Eisenmenger, L. B. et al. Vascular contributions to Alzheimer’s disease. Transl. Res. J. Lab. Clin. Med.254, 41–53 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR4] 4.Schaeffer, S. & Iadecola, C. Revisiting the neurovascular unit. Nat. Neurosci.24, 1198–1209 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Gerrits, E. et al. Neurovascular dysfunction in GRN-associated frontotemporal dementia identified by single-nucleus RNA sequencing of human cerebral cortex. Nat. Neurosci.25, 1034–1048 (2022). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Benarroch, E. What Are the Roles of Pericytes in the Neurovascular Unit and Its Disorders? Neurology100, 970–977 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Li, M. et al. Microvascular and cellular dysfunctions in Alzheimer’s disease: an integrative analysis perspective. Sci. Rep.14, 20944 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Hartmann, D. A. et al. Brain capillary pericytes exert a substantial but slow influence on blood flow. Nat. Neurosci.24, 633–645 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.El-Ghazawi, K., Eyo, U. B. & Peirce, S. M. Brain microvascular pericyte pathology linking Alzheimer’s disease to diabetes. Microcirculation31, e12877 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Moutinho, S. Women twice as likely to develop Alzheimer’s disease as men — but scientists do not know why. Nat. Med.31, 704–707 (2025). [DOI] [PubMed] [Google Scholar]

[CR11] 11.Gabitto, M. I. et al. Integrated multimodal cell atlas of Alzheimer’s disease. Nat. Neurosci.27, 2366–2383 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Seattle Alzheimer’s Disease Brain Cell Atlas (SEA-AD) - Registry of Open Data on AWS. https://registry.opendata.aws/allen-sea-ad-atlas/.

[CR13] 13.Dimitrov, D. et al. Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. Nat. Commun.13, 3224 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Dimitrov, D. et al. LIANA+ provides an all-in-one framework for cell–cell communication inference. Nat. Cell Biol.26, 1613–1622 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.LIANA: a LIgand-receptor ANalysis frAmework. https://saezlab.github.io/liana/.

[CR16] 16.Armingol, E. et al. Context-aware deconvolution of cell–cell communication with Tensor-cell2cell. Nat. Commun.13, 3665 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Baghdassarian, H. M., Dimitrov, D., Armingol, E., Saez-Rodriguez, J. & Lewis, N. E. Combining LIANA and Tensor-cell2cell to decipher cell-cell communication across multiple samples. Cell Rep. Methods4, (2024). [DOI] [PMC free article] [PubMed]

[CR18] 18.Vanlandewijck et al. A molecular atlas of cell types and zonation in the brain vasculature. Nature554, 475–480 (2018). [DOI] [PubMed] [Google Scholar]

[CR19] 19.Yang, A. C. et al. A human brain vascular atlas reveals diverse mediators of Alzheimer’s risk. Nature603, 885–892 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR20] 20.Chen, M. B. et al. Brain endothelial cells are exquisite sensors of age-related circulatory cues. Cell Rep.30, 4418–4432.e4 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR21] 21.Liu, S.-Y., Chiang, M.-F. & Chen, Y.-J. Role of WW domain proteins WWOX in development, prognosis, and treatment response of glioma. Exp. Biol. Med.240, 315–323 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Gangaraju, S. et al. Cerebral endothelial expression of Robo1 affects brain infiltration of polymorphonuclear neutrophils during mouse stroke recovery. Neurobiol. Dis.54, 24–31 (2013). [DOI] [PubMed] [Google Scholar]

[CR23] 23.Venkannagari, H. et al. Highly conserved molecular features in IgLONs contrast their distinct structural and biological outcomes. J. Mol. Biol.432, 5287–5303 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Duan, Y. et al. Semaphorin 5A inhibits synaptogenesis in early postnatal- and adult-born hippocampal dentate granule cells. eLife3, e04390 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Sadanandam, A., Rosenbaugh, E. G., Singh, S., Varney, M. & Singh, R. K. Semaphorin 5A promotes angiogenesis by increasing endothelial cell proliferation, migration, and decreasing apoptosis. Microvasc. Res.79, 1–9 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Il1rapl1 interleukin 1 receptor accessory protein-like 1 [Mus musculus (house mouse)] - Gene - NCBI. https://www.ncbi.nlm.nih.gov/gene/331461.

[CR27] 27.Valnegri, P. et al. The X-linked intellectual disability protein IL1RAPL1 regulates excitatory synapse formation by binding PTPδ and RhoGAP2. Hum. Mol. Genet.20, 4797–4809 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Duca, L. et al. Matrix ageing and vascular impacts: focus on elastin fragmentation. Cardiovasc. Res.110, 298–308 (2016). [DOI] [PubMed] [Google Scholar]

[CR29] 29.Wang, C.-H. J. & Chidiac, P. RGS2 promotes the translation of stress-associated proteins ATF4 and CHOP via its eIF2B-inhibitory domain. Cell. Signal.59, 163–170 (2019). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Adler, A. et al. A structural and dynamic visualization of the interaction between MAP7 and microtubules. Nat. Commun.15, 1948 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Ylikallio, E. et al. Deficiency of the E3 ubiquitin ligase TRIM2 in early-onset axonal neuropathy. Hum. Mol. Genet.22, 2975–2983 (2013). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Wesołek-Leszczyńska, A., Pastusiak, K., Bogdański, P. & Szulińska, M. Can adipokine FAM19A5 be a biomarker of metabolic disorders?. Diab. Metab. Syndr. Obes.17, 1651–1666 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Zhang, Q., Sun, S., Zhang, Y., Wang, X. & Li, Q. Identification of Scd5 as a functional regulator of visceral fat deposition and distribution. iScience25, 103916 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Li, Z. et al. SLC27A4-mediated selective uptake of mono-unsaturated fatty acids promotes ferroptosis defense in hepatocellular carcinoma. Free Radic. Biol. Med.201, 41–54 (2023). [DOI] [PubMed] [Google Scholar]

[CR35] 35.Li, Z. et al. Targeting the metabolic enzyme PGAM2 overcomes enzalutamide resistance in castration-resistant prostate cancer by inhibiting BCL2 signaling. Cancer Res.83, 3753–3766 (2023). [DOI] [PubMed] [Google Scholar]

[CR36] 36.Wang, X., Yu, H., Gao, R., Liu, M. & Xie, W. A comprehensive review of the family of very-long-chain fatty acid elongases: structure, function, and implications in physiology and pathology. Eur. J. Med. Res.28, 532 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Acquaviva, J., Wong, R. & Charest, A. The multifaceted roles of the receptor tyrosine kinase ROS in development and cancer. Biochim. Biophys. Acta BBA - Rev. Cancer1795, 37–52 (2009). [DOI] [PubMed] [Google Scholar]

[CR38] 38.Ding, L. et al. TM4SF19 aggravates LPS-induced attenuation of vascular endothelial cell adherens junctions by suppressing VE-cadherin expression. Biochem. Biophys. Res. Commun.533, 1204–1211 (2020). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Hua, C. et al. KDM6 demethylases and their roles in human cancers. Front. Oncol.11, 779918 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Nickerson, M. L. et al. TET2 binds the androgen receptor and loss is associated with prostate cancer. Oncogene36, 2172–2183 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR41] 41.Pecoraro, G. et al. Co-modulation of a circular form of PCDH11Y during neuroendocrine differentiation of prostate cancer. Front. Oncol.15, 1502405 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Zeng, Q., Yi, H., Huang, L., An, Q. & Wang, H. Reduced testosterone and Ddx3y expression caused by long-term exposure to arsenic and its effect on spermatogenesis in mice. Environ. Toxicol. Pharmacol.63, 84–91 (2018). [DOI] [PubMed] [Google Scholar]

[CR43] 43.Vellky, J. E., McSweeney, S. T., Ricke, E. A. & Ricke, W. A. RNA-binding protein DDX3 mediates posttranscriptional regulation of androgen receptor: A mechanism of castration resistance. Proc. Natl. Acad. Sci. USA117, 28092–28101 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Martínez-Rivera, F. J. et al. Differential protein expression profile in the hypothalamic GT1-7 cell line after exposure to anabolic androgenic steroids. PLoS One12, e0180409 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR45] 45.Zhao, L., Woody, S. K. & Chhibber, A. Estrogen receptor β in Alzheimer’s disease: from mechanisms to therapeutics. Ageing Res. Rev.24, 178–190 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.Jett, S. et al. Endogenous and Exogenous Estrogen Exposures: How Women’s Reproductive Health Can Drive Brain Aging and Inform Alzheimer’s Prevention. Front. Aging Neurosci. 14, 831807 (2022). [DOI] [PMC free article] [PubMed]

[CR47] 47.Zhang, L. et al. Single-cell transcriptomic atlas of Alzheimer’s disease middle temporal gyrus reveals region, cell type and sex specificity of gene expression with novel genetic risk for MERTK in female. 2023.02.18.23286037 Preprint at 10.1101/2023.02.18.23286037 (2023). [DOI] [PMC free article] [PubMed]

[CR48] 48.Su, J. et al. Cell–cell communication: new insights and clinical implications. Signal Transduct. Target. Ther.9, 196 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.National Academies of Sciences, E. et al. Transcriptomic evidence for sex differences in neurodevelopmental and neurodegenerative disorders. in Sex Differences in Brain Disorders: Emerging Transcriptomic Evidence: Proceedings of a Workshop (National Academies Press (US), 2021). [PubMed]

[CR50] 50.Soudy, M., Bars, S. L. & Glaab, E. Sex-dependent molecular landscape of Alzheimer’s disease revealed by large-scale single-cell transcriptomics. Alzheimers Dement21, e14476 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.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). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Sanfilippo, C. et al. Middle-aged healthy women and Alzheimer’s disease patients present an overlapping of brain cell transcriptional profile. Neuroscience406, 333–344 (2019). [DOI] [PubMed] [Google Scholar]

[CR53] 53.Grammas, P. & Ovase, R. Cerebrovascular transforming growth factor-β contributes to inflammation in the Alzheimer’s disease brain. Am. J. Pathol.160, 1583–1587 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR54] 54.Saharinen, J., Hyytiäinen, M., Taipale, J. & Keski-Oja, J. Latent transforming growth factor-beta binding proteins (LTBPs)-structural extracellular matrix proteins for targeting TGF-beta action. Cytokine Growth Factor Rev.10, 99–117 (1999). [DOI] [PubMed] [Google Scholar]

[CR55] 55.Kandasamy, M. et al. TGF-β signaling: a therapeutic target to reinstate regenerative plasticity in vascular dementia?. Aging Dis.11, 828–850 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Russell, J. K., Jones, C. K. & Newhouse, P. A. The role of estrogen in brain and cognitive aging. Neurotherapeutics16, 649–665 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Zárate, S., Stevnsner, T. & Gredilla, R. Role of Estrogen and Other Sex Hormones in Brain Aging. Neuroprotection and DNA Repair. Front. Aging Neurosci. 9, 430 (2017). [DOI] [PMC free article] [PubMed]

[CR58] 58.Sheppard, P. A. S., Choleris, E. & Galea, L. A. M. Structural plasticity of the hippocampus in response to estrogens in female rodents. Mol. Brain12, 22 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Bustamante-Barrientos, F. A. et al. The impact of estrogen and estrogen-like molecules in neurogenesis and neurodegeneration: beneficial or harmful? Front. Cell. Neurosci. 15, 636176 (2021). [DOI] [PMC free article] [PubMed]

[CR60] 60.Wharton, W. et al. Potential role of estrogen in the pathobiology and prevention of Alzheimer’s disease. Am. J. Transl. Res.1, 131–147 (2009). [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Dufort-Gervais, J. et al. Neuroligin-1 is altered in the hippocampus of Alzheimer’s disease patients and mouse models, and modulates the toxicity of amyloid-beta oligomers. Sci. Rep.10, 6956 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Brandi, E. et al. Brain region-specific microglial and astrocytic activation in response to systemic lipopolysaccharides exposure. Front. Aging Neurosci. 14, 910988 (2022). [DOI] [PMC free article] [PubMed]

[CR63] 63.Grabert, K. et al. Microglial brain region-dependent diversity and selective regional sensitivities to ageing. Nat. Neurosci.19, 504–516 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Barko, K. et al. Brain region- and sex-specific transcriptional profiles of microglia. Front. Psychiatry13, 945548 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR65] 65.Morris, G. P. et al. Microglia directly associate with pericytes in the central nervous system. Glia71, 1847–1869 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Davenport, C. M., Sevastou, I. G., Hooper, C. & Pocock, J. M. Inhibiting p53 pathways in microglia attenuates microglial-evoked neurotoxicity following exposure to Alzheimer peptides. J. Neurochem.112, 552–563 (2010). [DOI] [PubMed] [Google Scholar]

[CR67] 67.Aloi, M. S., Su, W. & Garden, G. A. The p53 Transcriptional Network Influences Microglia Behavior and Neuroinflammation. Crit. Rev. Immunol.35, 401–415 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR68] 68.Jebelli, J., Hooper, C. & Pocock, J. M. Microglial p53 activation is detrimental to neuronal synapses during activation-induced inflammation: Implications for neurodegeneration. Neurosci. Lett.583, 92–97 (2014). [DOI] [PubMed] [Google Scholar]

[CR69] 69.Butturini, E., Boriero, D., Carcereri de Prati, A. & Mariotto, S. STAT1 drives M1 microglia activation and neuroinflammation under hypoxia. Arch. Biochem. Biophys.669, 22–30 (2019). [DOI] [PubMed] [Google Scholar]

[CR70] 70.March-Diaz, R. et al. Hypoxia compromises the mitochondrial metabolism of Alzheimer’s disease microglia via HIF1. Nat. Aging1, 385–399 (2021). [DOI] [PubMed] [Google Scholar]

[CR71] 71.Levine, D. A. et al. Sex differences in cognitive decline among US adults. JAMA Netw. Open4, e210169 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Mathys, H. et al. Single-cell multiregion dissection of Alzheimer’s disease. Nature632, 858–868 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Mathys, H. et al. Single-cell atlas reveals correlates of high cognitive function, dementia, and resilience to Alzheimer’s disease pathology. Cell186, 4365–4385.e27 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR74] 74.Olayinka, O. A. et al. Single nucleus RNA sequencing reveals interaction between microglia and endothelial cells in mixed Alzheimer’s disease and vascular pathology. Alzheimers Dement20, e089628 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR75] 75.Wang, X. et al. Deciphering cellular transcriptional alterations in Alzheimer’s disease brains. Mol. Neurodegener.15, 38 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR76] 76.Leonard-Duke, J. et al. Multiscale computational model predicts how environmental changes and treatments affect microvascular remodeling in fibrotic disease. PNAS Nexus4, pgae551 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR77] 77.Rikard, S. M. et al. Mathematical model predicts that acceleration of diabetic wound healing is dependent on spatial distribution of VEGF-A mRNA (AZD8601). Cell. Mol. Bioeng.14, 321–338 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR78] 78.El-Ghazawi, K., Aberra T. Y. & Peirce M. S. Full NVU Tensor C2C of SEA-AD DLPFC. https://zenodo.org/records/16760785 (2025).

PERMALINK

Transcriptional analyses identify pericyte-centered signaling programs altered by sex and brain region in Alzheimer’s Disease

Kareem El-Ghazawi

Yonathan T Aberra

Yashasvisai Veeramasu

Wisam A Fares

Brant E Isakson

Harald Sontheimer

Ukpong B Eyo

Shayn M Peirce

Abstract

Introduction

Results

A snRNA-seq dataset of the neurovascular unit from the MTG and DLPFC of the same patients

Fig. 1. Dataset and demographic information.

NVU cells show a sex specific differentially expressed gene (DEG) profile

Fig. 2. DEG analyses in the MTG.

NVU cells show a sex specific disruption of communication patterns

Fig. 3. Communication factor 3 from the full NVU in the MTG across all groups.

Table 1.

Fig. 4. Communication factor 9 from the full NVU in the MTG across all groups.

Pericyte-EC signaling drives increased TGFß activity in AD relative to FNA donors

Fig. 5. Pairwise communication analysis of pericyte-EC communication in the MTG.

Pericyte-Astrocyte signaling disruption impairs estrogen activity in FAD donors

Fig. 6. Pairwise communication analysis of pericyte-astrocyte communication in the MTG.

Pericyte-Microglia communication pattern is conserved between brain regions and impaired with sex specificity

Fig. 7. Pairwise communication analysis of pericyte-microglia communication in the MTG.

Table 2.

Fig. 8. Pairwise communication analysis of pericyte-microglia communication in the DLPFC.

Discussion

Methods

Donor samples and dataset description

snRNAseq preprocessing and cell type filtering

Differential gene expression analysis

Cell-Cell communication inference: LIANA + Tensor cell2cell

Gene set enrichment and footprint enrichment analyses

Pairwise pericyte-centric Tensor cell2cell analysis

Spatial transcriptomic analysis

Statistical analysis and reproducibility

Reporting summary

Supplementary information

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Code availability

Competing interests

Footnotes

Contributor Information

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases