APOE4 promotes cerebrovascular fibrosis and amyloid deposition via a pericyte-to-myofibroblast transition

Abstract

Cerebrovascular disease is a major but poorly understood feature of Alzheimer's disease (AD). The strongest genetic AD risk factor, APOE4, is associated with cerebrovascular degeneration, including vascular amyloid deposition and fibrosis. To uncover how APOE4 promotes cerebrovascular pathology, we generated a single-cell transcriptomic atlas of human brain vasculature. In APOE4 carriers, pericyte abundance was significantly reduced and accompanied by the emergence of a myofibroblast-like cell population co-expressing contraction and extracellular matrix genes. Immunostaining confirmed non-vascular myofibroblasts in APOE4 human and mouse brains. We show that APOE4 pericytes transition into myofibroblasts that secrete fibronectin, which promotes vascular amyloid accumulation. Computational and experimental analyses identified elevated TGF-β signaling as the driver of this pericyte-to-myofibroblast transition. Inhibition of TGF-β restored pericyte coverage and reduced vascular fibrosis and amyloid to APOE3 levels, revealing a targetable mechanism linking APOE4 to cerebrovascular pathology in AD.

Keywords: APOE4; Alzheimer’s disease; TGF-β; cerebral amyloid angiopathy; cerebrovascular degeneration; fibronectin; fibrosis; iPSC modeling; myofibroblast; pericyte.

Extended Data Figure 1.. Metadata, quality control, and validation of single-nuclei analyses.

A. Flowchart describing the generation of the final integrated cerebrovascular atlas used for downstream analyses. Astrocytes, endothelial cells, pericytes, and smooth muscle cells were extracted from each individual. Within each dataset, propensity matching based on age, sex, Alzheimer’s disease (AD) status, and vascular nuclei contribution was performed using the MatchIt package. B. Quality control of propensity matching. Unpaired two-sample Student’s t-tests were used to compare age and average vascular cell yield. Chi-squared tests were used to compare age and AD status for the Sun dataset. These two variables were compared using Fisher’s Exact Test in the Yang and Haney datasets. C. UMAP of each dataset after integration. D. Annotation of blood brain barrier cell types (PECAM1 = endothelial cells, PDGFRB = pericytes, MYH11 = smooth muscle cells, AQP4 = astrocytes). E. Annotation of mural cell subclusters (SLC20A2 = Pericyte_1, COL4A1 = Pericyte_2, SLIT3 = SMC_1, TAGLN = SMC_2). F. Number of differentially expressed genes between APOE4 carriers and APOE3/3 mural cell subclusters. Analysis was performed using the MAST package, and differential expression was defined as adjusted p-value < 0.05 and |logFC| > 0.25. G. Validation of upregulated pathways in SMC_2 cells from APOE4 carriers using a pseudobulking approach (see Methods). After statistical analysis with the limma R package, genes were ranked according to sign(logFC) * −log10(P-value) and input into the fgsea package using the REACTOME database. Significantly upregulated pathways (NES > 0, FDR < 0.05) related to the extracellular matrix and contraction were then annotated, aligning with the results in Figure 1H. H. Feature map of mural cells annotated for expression of contraction or extracellular matrix (ECM) gene programs. Module scores were calculated using Seurat’s AddModuleScore() function based on the specified gene sets. A co-expression score was defined for each cell as the minimum of its contraction and ECM scores. I. Comparing SMC_2 contraction, ECM, and co-expression signature scores between APOE4 carriers and APOE3/3 individuals. Scores for SMC_2 cells calculated in H. were averaged per individual. Individuals contributing less than 5 cells were excluded from the analysis. Data points represent individuals (N = 16 per genotype). Bars are group means ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. J. Myofibroblast gene activity in SMC_2 cells by APOE genotype. SMC 2 gene expression was aggregated per individual. Individuals contributing less than 5 cells were excluded from the analysis. Gene set variation analysis (GSVA) was then performed using the GSVA R package. GSVA scores were assigned to each individual using previously described myofibroblast marker genes (see Methods). Data points represent individuals (N = 16 per genotype). Bars are group means ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests.

Extended Data Figure 2.. Additional validation of single-nuclei analyses.

A. UMAP of significantly differentially abundant (DA) cell neighborhoods in mural cells of APOE4 carriers vs. non-carriers using the miloR package (spatial FDR < 0.05, positive logFC indicates enrichment in APOE4 carriers). B. Representative marker genes of APOE4-enriched SMC_2 cells and non-enriched SMC_2 cells identified via DAseq in Figure 1J. The MAST package was utilized to identify differentially expressed genes (|logFC| > 0.25, adjusted p < 0.05) between these two subregions. C. The top three enriched GO pathways in APOE4-enriched cells compared to the remaining SMC_2 cells. The differentially expressed genes identified in B. were input into ClusterProfiler. D. Top cell type annotations of APOE4-enriched SMC_2 cells. The differentially expressed genes upregulated in APOE4 SMC_2 cells from B. were input into EnrichR with the CellMarker database. Redundant cell types were merged to enhance readability. E. Binomial generalized linear model identifying predictors of myofibroblasts in non-AD individuals. The proportion of myofibroblasts relative to total SMC_2 cell count per non-AD individual (P_{myofibroblast}) was modeled using a binomial generalized linear model with a logit link. Predictor variables included APOE genotype, age (modeled using a natural spline with 3 degrees of freedom), sex, and dataset. Model diagnostics were evaluated using the DHARMa package. Odds ratios (points) and 95% confidence intervals (error bars) are shown for each predictor. F. Correlation of APOE4 dosage with myofibroblast gene activity in APOE-TR mice. A processed bulk RNA sequencing dataset from the cerebral cortices of 4-month-old APOE3/3, APOE3/4, and APOE4/4-TR was downloaded from the source publication. Gene set variation analysis (GSVA) was performed using the GSVA R package. GSVA scores were assigned to each sample using the APOE4 myofibroblast marker genes identified in Extended Data Figure 1B. Pearson correlation analysis was subsequently performed between GSVA scores and APOE4 genotype dosage, encoded as 0 for APOE3/3, 1 for APOE3/4, and 2 for APOE4/4. G. Representative images of large cerebral vessels stained for lectin and αSma in age- and sex-matched APOE3/3 and APOE4/4-TR mice.

Extended Data Figure 3.. Validation of APOE4-induced myofibroblast accumulation in other iPSC lines.

A. Characterization of iPSC lines utilized in this study. B. Schematic of mix-and-match miBrain experiments performed with additional isogenic pairs of iPSC-derived mural cells (iMCs). C-D. miBrains were immunostained for αSMA, NG2, and PECAM. Scale bar, 50 µm. The αSMA and NG2 area ratio was quantified and expressed relative to APOE3/3. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. E-G. Representative images of isogenic iMCs in monoculture stained for NG2, αSMA, and fibronectin. MSSM3 scale bar, 50 µm. ADRC scale bar, 100 µm. SAD scale bar, 25 µm. Intracellular a-SMA and fibronectin areas were quantified as the positively stained regions within the NG2 mask, normalized to the total NG2 area, and expressed relative to APOE3/3 levels. Data points represent mean values from individual wells (n = 6 images per well, n = 4–6 wells per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests.

Extended Data Figure 4.. Additional computational validation of the APOE4 pericyte-to-myofibroblast transition.

A. Cell plot used for density graph generation in Figure 2G. B. Binomial generalized linear model identifying predictors of intermediate cells (ACTA2+/CSPG4+) in non-AD individuals. The proportion of intermediate cells relative to total mural cell count per non-AD individual (P_intermediate) was modeled using a binomial generalized linear model with a logit link. Predictor variables included APOE genotype, age (modeled using a natural spline with 3 degrees of freedom), sex, and dataset. Model diagnostics were evaluated using the DHARMa package. C. Odds ratios (points) and 95% confidence intervals (error bars) are shown for each predictor. D. Blood vessel distance measurements for pericytes, intermediate cells, and myofibroblasts in APOE4/4 miBrains and APOE4/4-TR mice. Cells were classified as described in Figure 2H. The morph() function in CellProfiler was used to calculate the distance of each pixel from the nearest PECAM+/lectin+ vascular region. For each classified cell, vascular distance was defined as the median pixel distance of all pixels within that cell. Distances were then averaged per miBrain or per mouse. Data points represent mean values from individual APOE4/4 miBrains or APOE4/4-TR mice (n = 6 miBrains, n = 3 mice). P-values were calculated using a repeated measures one-way ANOVA with a post-hoc test for a linear trend. E. CIBERSORTx deconvolution of bulk RNA-seq data from iPSC-derived mural cells (iMCs) using the mural cell subclusters from the single-nuclei cerebrovascular atlas as the reference. F. Predicted relative abundance of mural cell subclusters in iMC bulk RNA-seq data via CIBERSORTx deconvolution. G. Relative abundance of mural cell subclusters stratified by APOE genotype. Bars are mean logFC values ± SEM in APOE4/4 identity scores relative to APOE3/3. Data points represent individual bulk RNA-sequencing replicates (n = 3 per genotype). P-values were calculated using multiple unpaired two-sample Student’s t-tests with the Holm-Sidak correction for multiple comparisons. H. Representative myofibroblast marker genes upregulated in APOE4/4 iMCs (|logFC| > 0.25, adjusted p < 0.05 calculated from limma package). I. Top dysregulated GO pathways in APOE4/4 vs. APOE4/4 iMCs. Differentially expressed genes from H. were input into ClusterProfiler.

Extended Data Figure 5.. Additional experiments and computational analysis of fibronectin and TGF-B signaling in the APOE4 pericyte-to-myofibroblast transition.

A. Fibronectin deposition near αSMA staining in APOE4/4-TR mice. The percentage of fibronectin area within 2.5 µm of αSMA-positive area relative to the total fibronectin-positive area was quantified. Data points represent mean values from individual mice (n = 6 images per mouse, n = 4 mice). Bars are mean group values ± SEM. P-values were calculated using a one-sample t-test comparing the APOE4/4-TR mean to 50%. B. Comparison of FN1 gene expression in fibroblasts versus myofibroblasts. Gene expression was aggregated per individual. Data points represent individuals. Bars are group means ± SEM. P-values were calculated using a paired two-sample Student’s t-test. C. Association of FN1 expression in mural cells with AD phenotypes. Mural cell FN1 expression was aggregated per individual from the Haney dataset. A global pathology score was calculated for each individual by averaging their amyloid plaque, neurofibrillary tangle, infarct, and Lewy body scores reported in the source metadata. These pathology scores were classified as low (bottom 75%) or high (top 25%). Mural cell FN1 expression was then compared between APOE genotypes (n = 8 per genotype) and between the low (n = 11, one outlier removed via Grubbs’ test) and high (n = 4) pathology groups. For each comparison, FN1 expression was normalized to APOE3/3 or the low pathology group, respectively. Data points represent individuals. Bars are group means ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. The association between individual mural cell FN1 expression and age at AD diagnosis (n = 16) was assessed via Pearson correlation analysis. D. miBrain experiments performed with additional isogenic pairs of iPSC-derived mural cells (iMCs) as previously described in Extended Figure 3B. ADRC scale bar, 25 µm. SAD scale bar, 50 µm. The area positive for fibronectin was normalized to nuclei and expressed relative to APOE3/3. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 4–6 miBrains per genotype). E. Validation of FN1 shRNA knockdown. qRT-PCR was performed on APOE4/4 iMCs in monoculture. Data points represent mean values from individual wells of iMCs normalized to scramble (3 qRT-PCR replicates per well, n = 3 wells per condition). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. F-I. Additional outputs of NicheNet analysis from Figure 4A, B. J. Representative images of iMCs with and without TGFβ1 stimulation. Scale bar, 25 µm. For the control group, APOE3/3 and APOE4/4 iMCs were left untreated. For the experimental group, APOE3/3 iMCs were treated with 50 ng/mL TGFβ1 for 96 hours. iMCs were then fixed and stained for PDGFRβ and SMAD2. SMAD2 intensity inside the nuclear mask was quantified and expressed relative to APOE3/3 control. Data points represent mean values from individual wells of iMCs (6 images per well, 4 wells per condition). Bars are mean group values ± SEM. P-values were calculated using a one-way ANOVA with Dunnett’s test for multiple comparisons. K. Representative images of myofibroblast phenotypes in iMC monocultures after TGF-β inhibition. Scale bar, 50 µm. For the control group, APOE4/4 iMCs were left untreated. For the experimental group, APOE4/4 iMCs were treated with 10 µM SB431542 for 96 hours. iMCs were then fixed and stained for NG2, αSMA, and fibronectin. Intracellular αSMA and fibronectin areas were quantified as the positively stained regions within the NG2 mask, normalized to the total NG2 area, and expressed relative to APOE4/4 levels. Data points represent mean values from individual wells (n = 6 images per well, n = 4 wells per condition). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests.

Figure 1.. A myofibroblast-like mural cell population is enriched in the APOE4 brain.

A. Schematic of cerebrovascular atlas generation from single nuclei RNA-sequencing of the post-mortem human brain in APOE4 carriers and non-carriers. B. UMAP of harmonized blood-brain barrier cell types. C. Significant differentially abundant (DA) cell neighborhoods per cell type in APOE4 carriers vs. non-carriers using the miloR package (spatial FDR < 0.05, positive logFC indicates enrichment in APOE4 carriers). Data points are individual cell neighborhoods. Solid vertical lines represent mean logFC for each cell type. A one-sample Wilcoxon signed-rank test with the Benjamini-Hochberg correction was performed on the distribution of logFC values for each cell type to test whether the median logFC significantly differed from 0. D. Representative images of post-mortem human hippocampus stained for markers of pericytes (NG2) and blood vessels (VE-CAD) in APOE3/3 and APOE4 carriers. Pericyte coverage for each genotype was calculated by quantifying the overlapping area of pericyte and blood vessel staining, expressed relative to APOE3/3. Data points represent individuals (N = 13 APOE3/3 and N = 7 APOE4 carriers). Bars are group means ± SEM. P-values were calculated using an unpaired two-tailed Student’s t-test. E. Representative images of APOE4 carrier and non-carrier post-mortem human hippocampus stained for the smooth muscle cell marker αSMA. The area positive for αSMA was quantified, normalized to nuclei, and expressed relative to APOE3/3. Data points represent individuals (N = 11 APOE3/3 and N = 9 APOE4 carriers). Bars are group means ± SEM. P values were calculated using an unpaired two-tailed Student’s t-test. F. UMAP of mural cell types after subclustering. G. Differential gene expression between APOE4 carriers and non-carriers for each mural cell type was performed using the MAST package. Genes were then ranked according to sign(logFC) * −log10(P-value) and input into the fgsea package using the REACTOME database. The number of significant pathways for each cell type was then calculated (FDR < 0.05). H. The top significantly upregulated and downregulated pathways in APOE4 carrier SMC_2 cells compared to APOE3/3 (FDR < 0.05, ranked by normalized enrichment score [NES]). Redundant pathways were merged to enhance readability. I. Top cell type annotations of SMC_2 cells from APOE4 carriers. The differentially expressed genes (p < 0.05, logFC > 0.25) upregulated in APOE4 SMC_2 cells from the pseudobulk approach in Extended Data 1G were input into EnrichR with the CellMarker database. Redundant cell types were merged to enhance readability. J. Differential abundance analysis on the mural cell clusters was performed with DAseq. Red and blue indicate enrichment and depletion in APOE4 carriers, respectively. Given the overlap of the APOE4-enriched SMC_2 region with ECM and contraction co-signatures, and the analyses performed in Extended Data 2B–D, this region was annotated as myofibroblasts. K. Subject-level predicted proportion of myofibroblasts in non-AD individuals, as estimated by the binomial generalized linear model described in Extended Data 2E. Data points represent individuals (N = 13 per genotype). Bars are group means ± SEM. P-values are derived from the binomial GLM. L. Representative images of small vessels (<6 μm) in the corpus callosum of age- (9 to 10 months of age) and sex-matched APOE3/3 and APOE4/4-TR mice stained for markers of blood vessels (lectin), myofibroblasts (αSma), and pericytes (Ng2). Scale bar, 10 µm. The area positive for αSma was quantified, normalized to nuclei, and expressed relative to APOE3/3. Pericyte coverage was measured by quantifying the overlapping Ng2 and lectin area, normalized to APOE3/3. Data points represent mean values from individual mice (n = 6 images per mouse, n = 3 mice per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. M. Schematic of the miBrain model. N. Representative images of isogenic APOE3/3 and APOE4/4 miBrains stained for markers of blood vessels (CD144), myofibroblasts (αSMA), and pericytes (NG2). Scale bar, 75 µm. Three-dimensional pericyte coverage was measured by the overlapping NG2 and CD144 volume, normalized to the total CD144 volume, and expressed relative to APOE3/3. The area positive for αSMA was quantified, normalized to nuclei, and expressed relative to APOE3/3. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests.

Figure 2.. APOE4 promotes a pericyte-to-myofibroblast transition.

A. Schematic of the hybrid genetic experiment in which APOE3/3 iPSC-derived mural cells (iMCs) were integrated into an APOE4/4 miBrain. B. Representative images of isogenic APOE3/3, APOE4/4, and APOE4/4 miBrains integrated with APOE3/3 iMCs stained for PECAM, αSMA, and NG2. Scale bar, 50 µm. Three-dimensional pericyte coverage was measured by the overlapping NG2 and CD144 volume, normalized to the total CD144 volume, and expressed relative to APOE3/3. The area positive for αSMA was quantified, normalized to nuclei, and expressed relative to APOE3/3. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per genotype). Bars are mean group values ± SEM. P-values were calculated using a one-way ANOVA with Dunnett’s multiple comparisons test. C. Pseudotime analysis of the mural cell subclusters identified in Figure 1 using the slingshot R package. D. Generalized additive models (GAMs) were fit along each lineage using curated myofibroblast- and pericyte-associated gene signatures (see Methods) to validate the Slingshot-inferred trajectories. Signature score trajectories were then compared between lineages by computing the area under the curve (AUC) across pseudotime, and significance was assessed using a permutation-based p-value. E. Lineage bias between APOE4 carriers and APOE3/3 individuals. Distributions of Slingshot-derived curve weights were plotted for each lineage, stratified by APOE genotype. A value of 0 indicates a cell is exclusive to the pericyte-to-pericyte lineage. A value of 1 indicates a cell is exclusive to the pericyte-to-myofibroblast lineage. The p-value was calculated using the differentiationTest function in the condiments R package (see Methods). F. Schematic of ACTA2 and CSPG4 gene expression gradient during the pericyte-to-myofibroblast transition. G. Density plot of mural cells in non-AD individuals stratified by ACTA2 and CSPG4 expression. Cutoffs for gene expression were set at a log-normalized count value of 1.5. Subject-level predicted proportion of intermediate cells (ACTA2+/CSPG4+) in non-AD individuals was plotted as estimated by the binomial generalized linear model described in Extended Data 4B–C. Data points represent individuals (n = 25 APOE3/3, n = 24 APOE4 carriers). Bars are group means ± SEM. P-values are derived from the binomial GLM. H. Representative images of the pericyte-to-myofibroblast transition in APOE4/4 miBrains and TR-mice. APOE3/3 and APOE4/4 miBrains and TR-mice were stained for a vascular marker (CD144/lectin), αSMA, and NG2. Cells were classified into pericyte (vascular NG2+/ αSMA-), intermediate (NG2+/ αSMA+), or myofibroblast (nonvascular NG2-/ αSMA+) categories using intensity-based thresholds in CellProfiler. The number of cells in each category was normalized to the total number of classified cells. Data points represent mean values from individual miBrains or mice (n = 6 images per miBrain/mouse, n = 6 miBrains and n = 3 mice per genotype). P-values represent comparisons between genotypes for each cell state. P-values were calculated using a two-way repeated measures ANOVA with Geisser-Greenhouse correction and Tukey’s multiple comparisons test.

Figure 3.. Myofibroblast-derived fibronectin drives amyloid accumulation in APOE4 models.

A. Representative images of small vessels in the corpus callosum of APOE3/3-TR and APOE4/4-TR mice stained for lectin, αSma, and fibronectin. Scale bar, 10 µm. B. Representative images of APOE3/3 and APOE4/4 miBrains stained for PECAM and fibronectin. Scale bar, 50 µm. C. Quantification of FN1 expression across cell types in the cerebrovascular atlas. FN1 expression was aggregated per cell type for each individual. Individuals missing more than one cell type were excluded from the analysis. Data points represent individuals (n = 57). Bars are group means ± SEM. P-values were calculated using a repeated measures mixed-effects analysis with the Geisser-Greenhouse correction and Dunnett’s multiple comparisons test. D. Representative images of isogenic pairs of APOE3/3 and APOE4/4 miBrains stained for three amyloid antibodies targeting different epitopes of the amyloid-β peptide. Scale bar, 25 µm. E. Representative images of isogenic APOE3/3, APOE4/4, and APOE4/4 miBrains integrated with APOE3/3 iMCs stained for PECAM, fibronectin, and amyloid (D54D2). Scale bar, 25 µm. Data represent two independent experiments. In the first experiment, APOE3/3 and APOE4/4 were compared using an unpaired two-sample Student’s t-test. In the second experiment, APOE4/4 was compared to APOE4/4 with APOE3/3 iMCs using an unpaired two-sample Student’s t-test. P-values shown reflect these independent comparisons. For clarity, only the APOE4/4 group from the second experiment is shown. Representative images are from the corresponding experiment in each condition. F. Schematic of FN1 knockdown experiment in APOE4/4 miBrains. G. Representative images of APOE4/4 miBrains integrated with either iMCs transduced with scramble shRNA or shRNA targeting FN1 stained for fibronectin and amyloid (12F4). Scale bar, 25 µm. H. Treatment of encapsulated APOE3/3 iPSC-derived endothelial cells (iECs) with fibronectin and amyloid mixtures. Scale bar, 50 µm. For the control group, iECs were left untreated. For the experimental groups, iECs were treated with either 20 nM amyloid-β 1–40 and 1–42 or 20 nM amyloid-β 1–40 and 1–42 mixed with 50 ng/mL fibronectin for 96 hrs. iECs were then fixed and stained for amyloid (D54D2). For all panels, the area positive for αSMA, fibronectin, or amyloid immunoreactivity was quantified, normalized to nuclei, and expressed relative to APOE3/3 (panels B-D, H) or APOE4/4 (panels E, G). Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per condition) or mice (n = 6 images per mouse, n = 4 mice per genotype). Bars are mean group values ± SEM. Unless stated otherwise, p-values were calculated using unpaired two-sample Student’s t-tests (panels B-D, G) or a one-way ANOVA with Dunnett’s multiple comparisons test (panel H).

Figure 4.. Increased TGF-β signaling in the APOE4 brain drives the pericyte-to-myofibroblast transition.

A. APOE4 myofibroblast marker genes identified in Extended Data 2B were input into the R package NicheNet. B. Cirkos plot of top ligands and their predicted target genes from NicheNet. C. A generalized additive model (GAM) was fitted along each lineage using the hallmark TGF-β signaling gene set (see Methods). Signature score trajectories were then compared between lineages by computing the area under the curve (AUC) across pseudotime, and significance was assessed using a permutation-based p-value. D. TGF-β expression scores of pericytes from APOE4 carriers and APOE3/3 individuals. A TGF-β module scores were calculated using Seurat’s AddModuleScore() function based on the specified genes. The scores from Pericyte 1 and Pericyte 2 cells were then averaged per individual. Data points represent individuals (n = 106 APOE3/3, n = 103 APOE4 carriers). Bars are group medians. P-values were calculated using the Wilcoxon rank-sum test. E. Representative images of APOE3/3 and APOE4/4 miBrains and TR-mice stained for a vascular marker (PECAM/lectin) and pSMAD2. Scale bar, 25 µm. Nuclei along the vasculature were classified as pSMAD2+ using an intensity-based threshold in CellProfiler, normalized to the total number of vascular nuclei, and expressed relative to APOE3/3. Data points represent mean values from individual miBrains or mice (n = 6 images per miBrain/mouse, n = 6 miBrains and n = 4 mice per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests. F. Upregulated TGF-β signaling GO pathways in APOE4/4 iPSC-derived mural cells (iMCs). Differentially expressed genes in the iMC bulk RNA-seq dataset (|logFC| > 0.25, adjusted p < 0.05 calculated from limma) were input into ClusterProfiler. The significantly upregulated TGF-β signaling GO pathways (FDR < 0.05) were then extracted for graphical representation. G. TGFβR expression in isogenic pairs of iMCs from two donor lines. MSSM3 line: FPKM values from iMC bulk RNA-seq dataset normalized to APOE3/3. Data points represent individual bulk RNA-sequencing replicates (n = 3 per genotype). Bars are mean group values ± SEM. Adjusted p-values were calculated using the R package limma to account for multiple comparisons. sAD line: TGFβR expression in iMCs via qRT-PCR relative to APOE3/3. Data points represent mean values from individual wells of iMCs (3 qRT-PCR replicates per well, 4 wells per genotype). Bars are mean group values ± SEM. P-values were calculated using unpaired two-sample Student’s t-tests.

Figure 5.. TGF-β inhibition reverses APOE4-driven myofibroblast accumulation, cerebrovascular fibrosis, and amyloid deposition.

A. Timeline of TGF-β inhibition in miBrains. B. Representative images of myofibroblast phenotypes after TGF-β inhibition in miBrains. Scale bar, 25 µm. For the control group, APOE3/3 and APOE4/4 miBrains were treated with a DMSO vehicle for two weeks. For the experimental group, APOE4/4 miBrains were treated with 10 µM SB431542 for two weeks. miBrains were then fixed and stained for PECAM, αSMA, and NG2. Three-dimensional pericyte coverage was measured by the overlapping NG2 and PECAM volume, normalized to the total PECAM volume and expressed relative to APOE4/4. The area positive for αSMA was quantified, normalized to nuclei, and expressed relative to APOE4/4. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per condition). Bars are mean group values ± SEM. P-values were calculated using a one-way ANOVA with Dunnett’s multiple comparisons test. C. Representative images of cerebrovascular fibrosis and amyloid deposition in miBrains after TGF-β inhibition. Scale bar, 25 µm. For the control group, APOE3/3 and APOE4/4 miBrains were treated with a DMSO vehicle for 4 weeks. For the experimental group, APOE4/4 miBrains were treated with either 50 µM SB431542 or 50 µM galunisertib for 4 weeks. miBrains were then fixed and stained for PECAM, fibronectin, and amyloid (D54D2). The data represent two independent experiments. In the first experiment, APOE3/3 DMSO, APOE4/4 DMSO, and APOE4/4 treated with galunisertib were compared using a one-way ANOVA with Dunnett’s multiple comparisons test. In the second experiment, APOE4/4 DMSO was compared to APOE4/4 treated with SB431542 using an unpaired two-sample Student’s t-test. P-values shown reflect these independent comparisons. For clarity, only the APOE4/4 DMSO group from the first experiment is shown. Representative images are from the corresponding experiment in each condition. Data points represent mean values from individual miBrains (n = 6 images per miBrain, n = 6 miBrains per condition). Bars are mean group values ± SEM. D. Mechanistic model of the TGF-β-mediated pericyte-to-myofibroblast transition in APOE4 carriers.