Impairment of the blood-brain-barrier (BBB) is a key feature in Alzheimer's disease (AD), contributing to the entry of neurotoxic substances from the blood into the brain, triggering inflammatory responses and reducing cerebral blood flow. Accumulation of amyloid β (Aβ) deposits around cerebral vasculature is implicated as both a cause and consequence of BBB impairment, serving as an early biomarker of cognitive dysfunction, predicting cognitive decline, and contributing to AD pathogenesis and progression. However, the precise transcriptional changes within the gliovascular unit (GVU) of the BBB in AD and the molecular interactions between its main cell types – brain vascular cells and astrocytes – remain largely unestablished at a systems level. Single-cell RNA sequencing (RNAseq) offers a powerful approach to profile cell types and subtypes in AD and healthy brains, identifying cellular states, describing vulnerable populations, and elucidating perturbed genes and pathways. While most single-cell transcriptomic studies of AD brains have focused on neurons and abundant glial cells, relatively little is known about transcriptional changes in vascular cells (endothelia and pericytes) and their interaction with other CNS cells in the GVU. Recent studies have begun to reveal transcriptional profiles of human cerebrovasculature and detected differentially expressed genes (DEGs) in AD within enriched or unenriched nuclei. However, comprehensive studies that systematically investigate GVU transcriptional perturbations, validate interacting GVU molecules, and assess their effects on the BBB are crucial for identifying high-confidence therapeutic targets or biomarker candidates for BBB dysfunction. This study addresses this gap by employing a systematic approach to detect, prioritize, validate, and replicate GVU transcriptional perturbations in postmortem AD brains, investigating the association of the top perturbed vascular transcript (SMAD3) with antemortem outcomes, and performing in vitro and in vivo validations of SMAD3 interactions with its predicted astrocytic partner VEGFA.

The literature review section extensively cites previous research on blood-brain barrier (BBB) dysfunction in Alzheimer's disease (AD), highlighting the known breakdown of the BBB and its contribution to AD pathophysiology. Several studies using single-cell or single-nucleus RNA sequencing (sc/snRNAseq) are reviewed, focusing on the transcriptional changes in different brain cell types, including neurons, glia, and vascular cells. The authors point out the scarcity of studies specifically focusing on brain vascular cells, particularly pericytes and endothelial cells, due to their low abundance in the brain. They mention recent snRNAseq studies that have begun to profile cerebrovascular transcriptomes in AD, identifying differentially expressed genes (DEGs) in different vascular cell types. However, the review emphasizes the need for systematic studies that go beyond simply identifying DEGs to investigate the molecular interactions between the different cell types of the gliovascular unit (GVU) and validate these interactions using orthogonal methods and experimental models. This review sets the stage for the current study, which aims to fill these gaps in the literature.

This study employed a multi-faceted approach combining single-nucleus RNA sequencing (snRNAseq), quantitative PCR (qPCR), RNAscope, immunohistochemistry (IHC), immunofluorescence (IF), in vitro studies using induced pluripotent stem cell (iPSC)-derived pericytes, and in vivo studies using a zebrafish model. The snRNAseq analysis involved optimizing a nuclei isolation method to achieve high purity and the detection of rare cell types. Temporal cortex tissue from 12 AD patients and 12 age- and sex-matched controls was used. After quality control and filtering, the data was clustered and annotated using known cellular markers to identify various cell types, including excitatory and inhibitory neurons, oligodendrocytes, astrocytes, microglia, oligodendrocyte progenitor cells (OPCs), and three distinct vascular clusters (pericytes, endothelia, and perivascular fibroblasts). Differential expression (DE) analysis was performed to identify genes differentially expressed between AD and control groups within each cell cluster. Gene ontology (GO) enrichment analysis was used to identify enriched biological pathways for the DEGs. To discover potential interactions between astrocytes and vascular cells, the NicheNet platform was utilized to predict ligand-target interactions between significantly altered genes in astrocytic and vascular clusters. qPCR and RNAscope were employed to validate the expression of selected candidate genes. IHC and IF were used to validate protein expression. Human iPSC-derived pericytes from AD and control individuals were used in vitro to validate the identified interactions, with treatments to activate or inhibit VEGF signaling. A zebrafish model was utilized to perform in vivo validation of the interaction of the chosen gene pair and its effect on BBB integrity. Finally, the association of blood SMAD3 levels with AD-related neuroimaging outcomes was assessed in two large longitudinal cohorts, Mayo Clinic Study of Aging (MCSA) and Alzheimer's Disease Neuroimaging Initiative (ADNI), using blood SMAD3 expression data, genetic variants, and neuroimaging data (amyloid-β deposition and cortical thickness).

The study revealed widespread transcriptomic alterations in the gliovascular unit (GVU) of Alzheimer's disease (AD) brains. Single-nucleus RNA sequencing (snRNAseq) identified three distinct vascular cell clusters (pericytes, endothelia, and perivascular fibroblasts), with pericytes showing the most significant transcriptional changes in AD. A total of 1562 genes were upregulated and 64 genes were downregulated in AD pericytes. NicheNet analysis predicted interactions between astrocytic ligands and vascular targets, with SMAD3 (upregulated in AD pericytes) showing the strongest interactions with multiple astrocytic ligands, including VEGFA (downregulated in AD astrocytes). This SMAD3-VEGFA interaction was prioritized for further validation. qPCR validated the differential expression of six selected genes (SMAD3, STAT3, AHNAK, ANGPT2, ECE1, and TSC22D3) in AD vs. control samples, confirming the snRNAseq findings. RNAscope and immunofluorescence confirmed the expression of SMAD3 in vascular cells and VEGFA in astrocytes in human brain tissue. Increased phospho-SMAD3 immunoreactivity was observed in AD pericytes, suggesting increased SMAD3 signaling. Analysis of external snRNAseq datasets replicated the findings of upregulated pericytic SMAD3 and downregulated astrocytic VEGFA in AD across multiple brain regions. In ADNI and MCSA cohorts, genetic variants associated with higher blood SMAD3 levels were associated with decreased brain infarct size, and higher blood SMAD3 levels correlated with less brain amyloid-β deposition and less cortical atrophy. In vitro studies using iPSC-derived pericytes demonstrated an inverse relationship between VEGF and SMAD3 expression levels: VEGF treatment reduced SMAD3, and VEGF receptor inhibition increased SMAD3. In vivo zebrafish studies showed that Aβ treatment reduced astrocytic vegfaa (VEGF ortholog) expression, and blocking VEGF signaling increased pSMAD3 (active form) levels and impaired BBB integrity.

The study's findings strongly support a model where reduced VEGFA signaling and expression in the presence of Aβ (and possibly other AD neuropathologies) leads to increased SMAD3 levels, signaling, and ultimately, BBB disintegrity. The precise role of elevated SMAD3—whether it's detrimental or a compensatory/protective response—remains to be fully elucidated. The antemortem data suggests that higher SMAD3 levels might be protective against vascular, Aβ, and neurodegenerative outcomes in AD. However, this contrasts with findings in mouse models, where Smad3 inhibition reduced brain Aβ and inflammation. Further research is needed to clarify the functional consequences of SMAD3 in AD. The study highlights the usefulness of the combined approach of in silico prediction, in vitro validation using iPSCs, and in vivo validation in zebrafish models in translating big omics data into high-confidence molecular targets for complex diseases such as AD.

This study provides a comprehensive and systematic investigation of gliovascular transcriptional perturbations in Alzheimer's disease (AD), identifying a prioritized set of perturbed molecular interactions, particularly the inverse relationship between pericytic SMAD3 and astrocytic VEGFA. These findings were validated using multiple independent methods and replicated across different species and cohorts. The association of blood SMAD3 levels with AD-related neuroimaging outcomes warrants further exploration of SMAD3 as a potential therapeutic target or biomarker. Future research should focus on elucidating the precise role of SMAD3 in BBB integrity, determining the downstream effectors of SMAD3 signaling and investigating the potential therapeutic implications of modulating SMAD3-VEGFA interactions in AD.

The study focused primarily on the predicted interactions between astrocytes and brain vascular cells, neglecting potential interactions with other cell types such as neurons, oligodendrocytes, and OPCs. While the SMAD3-VEGFA interaction was prioritized, other predicted interactions warrant future investigation. The study mainly focused on late-stage AD cases and a single brain region (temporal cortex), limiting generalizability. The relatively small sample size of the primary snRNAseq cohort may also impact statistical power. While external datasets were used for replication, inherent differences in methodologies and cohorts may influence comparability. The correlation between blood SMAD3 levels and neuroimaging outcomes, while suggestive, doesn't prove direct causality and further investigations are needed to establish clear functional links.