Reconstruction of the human blood-brain barrier in vitro reveals a pathogenic mechanism of APOE4 in pericytes

Abstract

In Alzheimer's disease, amyloid deposits along the brain vasculature lead to a condition known as cerebral amyloid angiopathy (CAA), which impairs blood-brain barrier (BBB) function and accelerates cognitive degeneration. Apolipoprotein (APOE4) is the strongest risk factor for CAA, yet the mechanisms underlying this genetic susceptibility are unknown. Here we developed an induced pluripotent stem cell-based three-dimensional model that recapitulates anatomical and physiological properties of the human BBB in vitro. Similarly to CAA, our in vitro BBB displayed significantly more amyloid accumulation in APOE4 compared to APOE3. Combinatorial experiments revealed that dysregulation of calcineurin-nuclear factor of activated T cells (NFAT) signaling and APOE in pericyte-like mural cells induces APOE4-associated CAA pathology. In the human brain, APOE and NFAT are selectively dysregulated in pericytes of APOE4 carriers, and inhibition of calcineurin-NFAT signaling reduces APOE4-associated CAA pathology in vitro and in vivo. Our study reveals the role of pericytes in APOE4-mediated CAA and highlights calcineurin-NFAT signaling as a therapeutic target in CAA and Alzheimer's disease.

Extended Fig. 1. Characterization of human iPSC-derived cells

a and b iPSC-derived brain endothelial cells stained with CD144 (VE-Cadherin), CD31 (PECAM), ZO1 and GLUT1. These experiments were repeated at least 4 times with similar results. c and d, iPSC-derived astrocytes stained with GFAP, S100β and AQP4. e and f Comparative expression analysis of genes in iPSC-derived astrocytes from RNA-sequencing that are reported to be the most differentially upregulated in these experiments were repeated at least 4 times with similar results. e, fibroblasts and f, oligodendrocytes when compared to astrocytes from In vivo data in Vanlandewijck et al., 2018. Astrocytes are from 6 independent bulk RNA-sequencing g-i iPSC-derived mural cells stained with CD13, SM22, NG2, and SMA. These experiments were repeated at least 2 times with similar results. j. Comparative expression analysis of the top differentially upregulated genes in pericytes compared to smooth muscle cells (SMCs) from In vivo data in Vanlandewijck et al., 2018. iMCs are from 6 independent bulk RNA-sequencing. Expression is represented as FPKM values from bulk RNA-sequencing. Center values are mean and error bars are SD k, Comparative expression analysis of the top differentially upregulated genes in SMCs compared to pericytes from In vivo data in Vanlandewijck et al., 2018. iMCs are from 6 independent bulk RNA-sequencing. Expression is represented as FPKM values from bulk RNA-sequencing Center values are mean and error bars are SD l, Expression of the top three differentially upregulated genes in pericytes compared to fibroblasts. m, Expression of the top three differentially upregulated genes in fibroblasts compared to pericytes from In vivo data in Vanlandewijck et al., 2018. iMCs are from 6 independent bulk RNA-sequencing. Center values are mean and error bars are SD. n, Expression of mural cells and mesenchymal marker genes in iPSC-derived mural cells. For e, f, j, k, l, m, differential gene lists are based on analysis provided shown as average counts compared to FPKM from bulk RNA-sequencing of iPSC-derived astrocytes and mural cells. Center value and error bars are means and SD from RNA prepared from 6 independent wells. o, Identification of In vivo Brain Endothelial, Pericyte, and SMC cluster via marker gene expression from single-nucleus RNA-sequencing of the human post-mortem hippocampus from 82 individuals. p, iPSC-derived mural cells express markers of in vivo human brain pericytes. Center values and error bars are means and SD from independent RNA prepared from 3 separate wells. q, Global hierarchical clustering of transcriptomes (13,338 genes) demonstrates that iPSC-derived mural cells cluster with in vivo human hippocampal pericytes. Clustering was performed by average correlation with simple linkage.

Extended Fig. 2. Characterization of human iBBB

a, Three-dimensional vascular network of endothelial cells stained with CD144 scale bar = 200 μm. b, one week after formation iMCs labeled with SM22 are homogeneously dispersed and rudimentary vessels started forming. After two weeks endothelial vessels formed and iMCs have homed to perivascular space. c, Astrocytes are dispersed throughout iBBB cultures. Experiments in a-c were repeated at least 3 times. d, mRNA expression of AQP4 in each cell type alone, pair-wise and combined. One-way ANOVA with Bonferroni’s multiple comparison. **, p = 0.0013, 0.002, 0.0035, and 0.0076 for BECs, iMCs, Astrocytes, and BECs + iMCs respectively. ****, p < 0.0001. Center values and error bars are means and SDs from 6 independent RNA isolations. e, iBBB without astrocytes do not stain for AQP4 (green). In iBBBs with astrocytes AQP4 densely stains along endothelial vessels. f, Immunostaining for LAMA4 showing that Matrigel does not contain LAMA4 however iBBB cultures remodel basement membrane surrounding endothelial vessels to contain LAMA4. Experiments in e-f were repeated at least 3 times. g, PLVAP mRNA expression is upregulated in BECs from iBBB cultures compared to BECs cultured alone. Center values and error bars are mean and SD from RNA harvested from six independent wells. h, PLVAP mRNA expression is downregulated in BECs from iBBB upon removal of VEGFA from culture media. Center values and error bars are mean and SD from RNA harvested from six independent wells. Two-sided unpaired student t-test p < 0.0001. i, iBBB cultured in trans-well format express high levels of BBB marker CLDN5 (green) and ZO1 (red). Experiments in e-f were repeated at least 2 times. j, Polarization of ABCG2 was measured by Hoechst transport for both a BECs monolayer and the iBBB from the apical to the basolateral surface and vice versa. Samples treated with the ABCG2 specific inhibitor KO143 were normalized to each respective non-inhibitor treated sample. Stars represent significance determined by two-sided multiple student’s t-test (FDR = 0.01; p = 0.0011) Center values and error bars are mean and SD from 6 independent transwells.

Extended Figure 3.. Validation of the iBBB as a model for CAA

a, iBBBs generated from a familial AD patient iPSC with duplication of the APP gene (APP1.1) do not inherently have higher amyloid levels than non-AD controls (AG09173). b, iBBBs generated from iPSCs with a familial AD-associated mutation (M146I) in the PSEN1 gene do not inherently have higher amyloid levels than its non-AD isogenic control. Experiments in a and b were repeated at least 3 times. c, Media conditioned by neuronal cells derived from familial AD patient has significantly higher Aβ(1–42). Two-sided Student t-test (p = 0.0022) Center values and error bars are mean and SD from 3 independent wells. d, Representative images depicting that iBBBs derived from APOE3/4 individuals exhibit high levels of amyloid accumulation relative to iBBBs generated from APOE3/3 individuals. e and f, Representative images depicting that iBBBs derived from isogenic APOE3/3 and APOE4/4 individuals exhibit high levels of amyloid accumulation assay with anti-amyloid antibody Thioflavin T (f) and 12F4 (e). Experiments in d-f were repeated at least 3 times. g and h, Representative images and quantification of amyloid accumulation in isogenic iBBBs exposed to 20 nM Aβ-FITC for 1–40 and 1–42 isoforms. The total area positive for amyloid was divided by total nuclei and then normalized to the mean amyloid/nuclei from all E3/3 samples such that the mean of E3/E3 is set to 100% for each isoform. Students t-test, 1–40 p = 0.0044; 1–42 p > 0.00001. Experiments in were repeated at least 3 times. Center values and error bars are means and SD from 12 independent iBBBs. i and j, Normalized amyloid accumulation in isogenic iMCs and BECs mono-culture for each APOE genotype. Two-sided Students t-test, iMural cells, p = 0.0002; BECs p = 0.0118. Center values and error bars are means and SD from 12 independent wells.

Extended Fig. 4. APOE4 Pericytes Increase CAA pathology in iBBB

a, Quantification of Aβ accumulation in deconstructed iBBBs. B/iMC/A3 and B/iMC/A4 indicate all APOE3/3 and APOE4/4 iBBBs respectively where B = BECs only, B/A = BECs and astrocytes, and B/iMC = BECs and iMural cells. Analysis was performed by One-way ANOVA with Bonferroni’s post-hoc analysis (p < 0.0001). Center values and error bars are means and SD from 5 independent iBBBs. B/iMCA3 v B/iMCA4, p = 0.0005; B/iMCA4 v: B3, p = 0.0001; B4, p = 0.0001; B/A3, p = 0.0064; B/A4, p = 0.0001; B/iMC3, p = 0.0026; B/iMC/A3 v B/iMC4, p < 0.0001. b, Exposing APOE4/4 astrocytes to APOE4/4 iMural cell conditioned media significantly increases amyloid accumulation compared APOE3/3 pericyte conditioned media. Unpaired two-sided Student t test, p = 0.0001. Center values and error bars are means and SD from 4 iBBBs. c, GO analysis from Toppfun (statistics described at https://toppgene.cchmc.org/enrichment.jsp) depicting biological processes associated with up-regulated and down-regulated genes. From RNA extracted from 3 independent wells of iMCs for each genotype.

Extended Fig. 5. APOE expression is selectively elevated in APOE4 pericytes

a, Quantification and representative image of APOE protein expression in brain vascular pericytes (NG2-positive cells) and non-pericytes (NG2-negative) cells in APOE4 knock-in mouse. Two-sided Student t test, p < 0.0001 Scale bar = 50 um. Center values and error bars are means from 150 APOE-positive cells for each genotype. b, anti-APOE antibody (Abcam, ab52607) is specific for human APOE protein and does not react with mouse Apoe protein or other proteins present in wild-type mouse hippocampus as reported by the manufacturer. The staining was repeated using 3 different mice c and d, Expression of APOE in isogenic iMural cells (c) and astrocyte (d) measured by RNA sequencing each condition represents three biological replicates pericyte, q = 0.0003 astrocyte, q = 0.0006 statistics performed by Dseq2, Center values and error bars are mean and SD from RNA prepared from 3 independent wells for each genotype. e Violin plots depicting APOE expression in pericytes/endothelial cells isolated from post-mortem prefrontal cortex of APOE4-carriers (n = 7) compared to non-carriers (n = 18). Differential expression was measured using a two-tailed Wilcoxon rank sum test, considering cells with detected expression of APOE (p = 0.0026). f, Images and quantification of APOE protein expression in post-mortem human prefrontal cortex from APOE4 carriers and non-carriers. Unpaired two-tailed t test (p = 0.023). Center values and error bars are mean and SD from staining of 4 prefrontal cortex sections from 4 different individuals. g, Differential plot of representative maker genes showing that pericytes and endothelial cells isolated from human hippocampus segregated into distinct cellular clusters (n = 82 individuals). h, Violin plots depicting APOE expression in endothelial cells isolated from post-mortem hippocampus APOE4-carriers (n = 16) compared to non-carriers (n = 46). Differential expression was measured using a two-tailed Wilcoxon rank sum test, considering cells with detected expression of APOE. Center lines are mean and dashed line are SD. i, Images and quantification of APOE protein expression α-SMA positive SMC in post-mortem human hippocampus from APOE4 carriers and non-carriers. Unpaired two-tailed t test (p = 0.5814). Center values are mean and SD of hippocampal sections from 6 individuals for each genotype. j, Western blot of APOE and GAPDH protein in a isogenic iPSC line in which the APOE gene was truncated via CRISPR-editing to yield a line that does not express of APOE protein and therefore is termed APOEKO line. The experiment was repeated at least 3 time with similar results.

Extended Fig. 6. NFAT/Calcineurin signaling is selectively elevated in APOE4 pericytes

a, Increasing the soluble APOE concentration through the addition of recombinant APOE protein to iBBB culture increases amyloid accumulation. One-way ANOVA with Bonferroni’s post-hoc analysis. Center values and error bars are mean and SD from 4 independent iBBBs for each condition. One-way ANOVA with Bonferroni’s multiple comparison test. APOE3/3 (Parental) V: APOE4/4 (Isogenic), p = 0.02; E3/3 + rAPOE3, p = 0.0034; E3/3+ rAPOE4, p = 0.0144. b and c, Representative western blot and quantification depicting nuclear NFATc1 expression in isogenic APOE3 and APOE4 iMCs. Unpaired student t test, p = 0.0254. Experiment was repeated 3 times with similar results. (c) Center values and error bars are mean and SD from 3 independent lysate preparations for each genotype. d, Expression of calcineurin catalytic subunits measured by RNAseq. PPP3CA (q = 0.0003); PPP3CC (q = 0.0188). Analysis from DSEQ2. Center values and error bars are mean expression and SD of RNA prepared from 3 independent wells for each genoptype. e, Expression of negative Regulators of Calcineurin genes (RCANs) measured by RNAseq. RCAN2 (q = 0.0003); RCAN3 (q = 0.0123). ). Analysis from DSEQ2. Center values and error bars are mean expression and SD of RNA prepared from 3 independent wells for each genoptype. f, Expression of DYRKs kinases known to phosphorylate NFAT measured by RNAseq. DYRK4 (q = 0.0003). Analysis from DSEQ2. Center values and error bars are mean expression and SD of RNA prepared from 3 independent wells for each genoptype. g, Expression of predicted NFAT response gene, VCAM1 and ACTG2, in iMural cells. Expression is quantified by qRT-PCR and normalized to the average of E3/3 cells. Significance determined by One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison. For VCAM1 E3/3 (Parental) v: E4/4 (Isogenic), p < 0.0001; H9, p < 0.0001; 231, p = 0.0152; 332, p < 0.0001. For ACTG2 E3/3 (Parental) v: E4/4 (Isogenic), p = 0.0006; H9, p = 0.0108; 231, p = 0.0076; 332, p < 0.0001. Center values and error bars are mean expression and SD of RNA prepared from 4 independent wells for each cell line. h and i, Violin plots depicting NFATC1 (h) and NFATC2 (i) expression in pericytes isolated from post-mortem prefrontal cortex of APOE4-carriers (n = 16) compared to non-carriers (n = 46). Differential expression was measured using a two-tailed Wilcoxon rank sum test, considering cells with detected expression of APOE. Center lines are mean and dashed line are SD. j and k, Violin plots depicting NFATC2 expression in endothelial cells isolated from post-mortem prefrontal cortex of APOE4-carriers (n = 7 compared to non-carriers (n = 18). Differential expression was measured using a two-tailed Wilcoxon rank sum test, considering cells with detected expression of APOE (p = 0.035). Center lines are mean and dashed line are SD.

Extended Fig. 7. Inhibiting NFAT/Calcineurin signaling reduces APOE expression and CAA pathology

a, Chemical structures of Cyclosporine A (CsA), FK506, and INCA6 showing highly dissimilar structures. b, Expression of PGK1, HPRT, and GAPDH in pericytes after two weeks with DMSO, CsA, FK506 or INCA6. One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison. Center values and error bars are mean expression and SD of RNA prepared from 4 independent wells for each cell line. c and d, Representative immunofluorescence imaging of APOE protein staining in pericytes after two weeks of treatment with chemicals. Scale bar, 50 μm. Experiments were repeated at least 3 times with similar results. e DEGs and associated GO terms for up-regulated and down-regulated genes in E3 and E4 CsA-treated pericyte from RNA-sequencing of RNA prepared from 3 independent wells for each condition. DEGs were determined by DSEQ2 and GO analysis was performed with Toppfun. f and g, Representative imaging and quantification depicting APOE protein expression in the APOE4KI mouse cortical slices following treatment with cyclosporine A (CsA) for one week. Unpaired, two tailed t test (p = 0.0009). Experiments were repeated with similar results using at least 3 slice preparations for each condition. Center values and error bars are mean intensity and SD from 12 independent measurements. h, Quantification of amyloid APOE4KI mouse cortical slices treated with either CsA or FK506 for one week and then exposed to 20 nM Ab for 48 hours. One-way ANOVA (p = 0.0105) with Bonferroni’s multiple comparison. Control v: CsA, p = 0.0188; FK506, p = 0.0245. Center values and error bars are mean and SD from slices prepared from 3 different mice. i, APOE mRNA expression in primary pericytes isolated from brain microvasculature of APOE4 knock-in mice treated with DMSO, Cyclosporine A, or FK506. One-way ANOVA (p = 0.0139) with Bonferroni’s multiple comparison. For DMSO v: CsA, p = 0.0221; FK506, p = 0.0367. Center values and error bars are mean and SD from pericytes prepared from 3 different mice. j, Representative image of immunostaining for APOE in hippocampal pericytes from APOE4 KI x 5xFAD mice treated with cyclosporine A or vehicle for one week. k, Representative images of vascular amyloid in the hippocampus following treatment of 6-month-old APOE4KI x 5XFAD female mice with either vehicle or CsA. Amyloid was detected and quantified with two independent anti-amyloid antibodies (6e10 and 12F4). These experiments were repeated 2 times with similar results.

Fig. 1.. Reconstruction of Anatomical and Physiological Properties of the Human Blood-brain-barrier in vitro (iBBB).

a, Schematic of iBBB formation from iPSCs. b, iBBB stained for endothelial cell marker CD144 (red) Scale bar, 50 μm. c, iPSC-derived Mural Cells (iMural cells of iMCs) localize to endothelial vessels after two weeks in culture. Scale bar, 50 μm. d, iMural Cells are labeled with NG2 (green) and BECs with CD144 (red). e, Astrocytes surround endothelial vessels after two weeks in culture. Astrocytes are labeled in green with GFAP. BECs are labeled with CD144 in red. Scale bar, 50 μm. f, Aquaporin 4 (AQP4) green, is expressed on BEC vessels labeled with ZO-1, red, pan-astrocyte marker S100β is blue. Scale bar, 50 μm. Experiment in b-f were repeated at least 3 times each with similar results. g, qRT-PCR measuring the expression of common BBB-associated genes. All expression is normalized to pan-endothelial marker PECAM to account for potential differences in BEC cell number. Mean expression and SD of RNA harvested from 6 independent samples for each condition. Differences were analyzed by one-way ANOVA with Bonferroni’s post-hoc analysis CLDN5, p < 0.0001 for BECs v BA and iBBB; JAMA iBBB v: B+iMCs p < 0.0001, B+A, p = 0.0009, B only, p < 0.0001, B v B+A, p = 0.0013, B v B+iMCs, p = 0.0246; For PgP, iBBB v: B, p = 0.0079, B+A, p = 0.0001, B+iMCs, p 0.0019; LRP1, iBBB v: B, p < 0.0001, B+A, p < 0.0001, B+iMCs, p < 0.0001, B v B+A, p = 0.0029; RAGE iBBB v: B, p < 0.0001, B+A, p = 0.0357; GLUT1, iBBB v: B, p = 0.0026. h, qRT-PCR measuring the expression of transporters, adhesion molecules, and efflux-pumps, and tight-junctions found in the BBB. All expression levels are normalized to BECs alone. Y-axis is the expression level in BECs isolated from the iBBB normalized to BECs cultured alone. X-axis is BECs co-cultured with astrocytes normalized to BECs cultured alone. Blue circles represent means from three biological replicates and three PCR replicates. i, Cartoon depicting transwell setup for measuring iBBB permeability j, Representative image of BECs (ZO-1, red), iMural cells (SM22, blue) and astrocytes (S100β, green) co-cultured on transwell membrane. Transwell imaging experiments were repeated twice k, Trans-endothelial electrical resistance (TEER) measurements from HuVECs, HuVECs plus iMural Cells (MC) and astrocytes (A), BECs only, and the iBBB. Circles represent single measurements from individual transwells. Experiment consisted of six independent transwells for each condition (n = 6). Dots are the means from three readings from one transwell. Bar are means. Error bars are SD. Differences were analyzed by one-way ANOVA with Bonferroni’s post-hoc analysis. BECs only v iBBB, p = 0.0182; BECs and iBBB v HuVECs samples p < 0.0001. The TEER experiment was repeated three times with similar results l, Permeability of fluorescently labeled molecules for BECs alone or iBBB. All values are reported as a percent of each molecule’s permeability across a blank transwell membrane. Dots are the means from three permeability readings from one transwell. Bar are means. Stars represent significance determined by Two-sided multiple student’s t-test (FDR = 0.01). Cadaverine p = 4.14 e-008; 4 kDa Dextran p = 2.8 e-005; 10 kDa Dextran p = 3.39 e-006; Transferin p = 0.006; 70 kDa Dextran p = 0.0006. m, BBB properties of the iBBB require cooperative interaction of iMural cells and astrocytes. The permeability of 4 kDa dextran was quantified in the iBBB and compared to BECs with 2x iMural Cells, 2x astrocytes, or BECs with mouse embryonic fibroblasts (MEFs). Permeability is normalized to BECs alone. Each dot is the mean of three permeability readings from one transwell. N = 12 transwells per a condition. Center values are mean, error bars are SD. One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparisons. n, ABCG2 expression is up-regulated in the iBBB. One-way ANOVA with Bonferroni’s post-hoc analysis (p < 0.0001). n = 6 biological replicates, each dot is the PCR mean from one biological replicate. Center value is the mean of biological replicatres. Error bars are SD. o, Polarization of Pgp was measured by rhodamine 123 transport for both a BECs monolayer and the iBBB from the apical to basolateral surface and vice versa. Inhibitor-treated samples were normalized to each respective non-inhibitor-treated sample. Each dot is the mean of three permeability readings from one transwell. N = 6 transwells per a condition. Center values are mean, error bars are SD. Stars represent significance determined by two-sided multiple student’s t-tests (p = 0.015) (FDR = 0.01).

Fig. 2.. APOE4 increases amyloid accumulation in the iBBB.

a, Cartoon depicting the experimental paradigm for exposing iBBBs to exogenous amyloid-β. b, Amyloid selectively accumulates on non-AD iBBBs exposed to media conditioned by iPSC-derived neuronal cells from a familial AD patient with an APP-duplication (APP1.1). iBBB derived from APOE3/3 iPSC line (E3/3 parental) from a healthy 75-year-old female. 6e10 antibody, green, recognizes Aβ1-16 epitope. Scale bar, 50 μm. Experiment was repeated at least 3 times with similar results. c, The APOE3/3 parental iPSC line was genetically edited to an isogenic APOE4/4 allowing the generation of genetically identical iBBBs. Isogenic APOE4/4 iBBBs accumulated more amyloid compared to the parental APOE3/3 iBBB when simultaneously exposed to APP1.1 conditioned media for 96 hours. Scale bar, 50 μm. Experiment was repeated at least 3 times with similar results. d, Representative quantification of amyloid accumulation in two isogenic iBBBs with reciprocal genetic editing strategies. Arrows indicate direction of genetic editing where the right-facing arrow indicates editing from APOE3/3 to APOE4/4 and the left-facing arrow indicates editing from APOE4/4 to APOE3/3. Total area positive for amyloid was divided by total nuclei and then normalized to the mean amyloid/nuclei from all E3/3 samples such that the mean of E3/E3 is set to 100%. Blind automated image analysis was performed with ImageJ. Two-sided Student t-test (p = 0.0114). For each condition 12 iBBBs were quantified for each condition. Dots represent mean amyloid across 4 images for each iBBB. Center values are means and error bars are SD. e, APOE3/4 heterozygous iBBBs accumulate significantly more amyloid than APOE3/3 iBBBs. Quantification performed as described in d. For each condition 9 iBBBs were quantified for each condition. Dots represent mean amyloid across 4 images for each iBBB. Center values are means and error bars are SD. ANOVA with Bonferroni’s multiple comparison test (p < 0.0001). (E3/3 v E4/4, p < 0.0001; E3/3 v sAD231, p = 0.0055; E3/3 v sAD332, p = 0.0002; E3/3 v. H9, p < 0.0001) f, Representative images depicting that iBBBs derived from isogenic APOE3/3 and APOE4/4 individuals exhibit high levels of amyloid accumulation assay with anti-amyloid antibody D54D2. Experiment was repeated at least 3 times with similar results. g, Quantification of amyloid in isogenic iBBBs for Thioflavin T (p = 0.0258), and two different amyloid antibodies D54D2 (p =0.0020) and 12F4 (p = 0.0054). Two-sided student t test. Each dot represents the mean of 3 images from one iBBB, 6 iBBBs were analyzed for this experiment. Bars center values represent mean and error bars represent SD. Experiment was repeated 2 times with similar results. h, Quantification of soluble versus in soluble Aβ 1-40 in remaining in the iBBB culture media 96 hours after inoculation with 20 nM Aβ 1-40 (p = 0.0319). Two-sided student t test. Each dot represents the mean of 3 images from one iBBB, 6 iBBBs were analyzed for this experiment. Bars center values represent mean and error bars represent SD. Experiment was repeated 2 times with similar results. i, Representative three-dimensional IMARIS renderings depicting vascular amyloid accumulation in APOE3/3 and APOE4/4 iBBBs. iBBBs were allowed to mature for 1 month and then simultaneously exposed to neuronal conditioned media from the fAD APP1.1 line. Three-dimensional surfaces of 6e10 and VE-Cadherin staining were created using IMARIS software. The total area of 6e10 within 20 μm of the VE-Cadherin surfaces was measured. This was normalized to the total area of the Vecad surfaces Scale bar, 10 μm. Experiment was repeated 2 times with similar results. j, Quantification of vascular (< 20 μm from BEC vessel) (p = 0.0055) and non-vascular (> 20 μm from BEC vessel) (p= 0.0062) using IMARIS software. Amyloid area was normalized to total vascular area for each image. Each dot represents the mean of 3 images from one iBBB, 4 iBBBs were analyzed for this experiment. Bars center values represent mean and error bars represent SD. k, Representative image depicting amyloid accumulation in non-vascular cells positive for astrocyte marker S100β. Scale bar 50 μm. Experiment was repeated 2 times with similar results. l, Quantification showing the number of astrocytes positive for amyloid for each isogenic genotype. (p= 0.0003) two-sided student t-test. Each dot represents the mean of 4 images from one iBBB, 4 iBBBs were analyzed for this experiment. Bars center values represent mean and error bars represent SD.

Fig. 3.. Mural Cells are required for increased amyloid accumulation in the iBBB.

a, Representative images depicting combinatorial screen of APOE3/3 and APOE4/4 isogenic cell-types reveals that APOE4 expression in iMural cells is required for increased iBBB amyloid accumulation. These experiments were repeated 3 times. b, Quantification of amyloid accumulation in isogenic iBBBs for each permutation of combinatorial matrix. Each dot represents mean amyloid of four images. 6 iBBBs were analyzed for each condition. Center values represent mean and error bars are SD. Analysis was performed by one-way ANOVA with Bonferroni’s multiple comparison test where BECs/iMCs/Astrocytes: 3/3/3 v 4/4/4, p < 0.0001; 4/4/4 v 4/3/3, p < 0.0001; 4/4/4 v 4/3/4 p = 0.0004, 4/4/4 v 3/3/4 p < 0.0001; 3/3/3 v 3/4/3, p = 0.0391; 3/3/3 v 3/4/4, p = 0.050; 3/3/3 v 4/4/3 p = 0.0023. c, Segregating each isogenic permutation based on relative amyloid levels (low or high), reveals that APOE3/3 and APOE4/4 BECs and astrocytes are equally represented between the two conditions, however, iMural cells are not. For the low amyloid condition only APOE3/3 iMural cells are present. In contrast, for the high amyloid condition, only APOE4/4 iMural cells are present. d, Quantification of Aβ accumulation in iBBBs derived from APO3/3 (3), H9 is APOE3/4 heterozygous and 210 is APOE3/3 homozygous. Each dot represents mean amyloid of four images. 6 iBBBs were analyzed for each condition. Center values represent mean and error bars are SD. Analysis was performed by one-way ANOVA with Bonferroni’s multiple comparison test where BECs/iMCs/Astrocytes: 3/3/3 v 4/4/4 p < 0.0001; 4/4/4 v 4/3/4 p = 0.0004; 3/3/3 v H9/H9/H9 p = 0.0018; 3/3/3 v 3/H9/3, p = 0.002. e, Quantification of Aβ accumulation in isogenic iBBBs and APOE3/3 iBBBs treated with iMural cell conditioned media from either APOE3/3 (parental) or APOE4/4 (isogenic) iMural cell. Media was conditioned for 48 hours and added iBBBs with 1:1 ratio of fresh media and 20 nM Aβ-FITC for 96 hours. Each dot represents mean amyloid of 3 images. 12 iBBBs were analyzed for each condition. Center values represent mean and error bars are SD. Analysis was performed by two-sided student t-test where iBBB genotype/iMC conditioned media: 4/NA v 3/NA, p = 0.0095; 3/3 v 3/4, p = 0.001.

Fig. 4.. APOE and NFAT/Calcineurin-signaling are up-regulated in APOE4 mural cells with pericyte-like properties in vitro and selectively in pericytes in the human brain.

a, Heat map depicting differentially expressed genes between isogenic APOE3/3 and APOE4/4 iPSC-derived iMural cells. (q = 0.01) RNA and sequencing was performed on three independent wells for each isogenic cell line b, APOE gene expression is significantly up-regulated in APOE4/4 iMural cells whereas it is down-regulated in APOE4/4 astrocytes. Expression values from qRT-PCR than used for RNAseq experiment center value are means. two-sided student t-test Astrocyte (p = 0.0009), iMural Cells (p < 0.0001) (n = 9). c, Immunofluorescence staining and quantification of APOE in isogenic iMural cells. Scale bar, 50 μm. Dots are mean APOE fluorescence intensity from four independent images from a single well. Four wells were measured for each genotype. Unpaired Two-tailed t test (p = 0.0005). center values are means. d, Western blot and quantification for APOE protein in APOE isogenic iMural cells. Two constitutively expressed proteins in iMural cells are included smooth muscle actin (SMA) and GAPDH. Dots are band intensity from three independent lysates. Center values are means and error bars are SD. Unpaired Two-tailed t test (p = 0.0033) e, qRT-PCR showing APOE gene expression is also up-regulated in an additional isogenic pair that was edited from E4/4 to E3/3 and three APOE3/4 heterozygous iMural cells from iPSC lines derived from individuals with sporadic AD and H9 hESC line. Arrows indicated the direction of genetic editing. All values are normalized to the mean expression in all APOE3/3 (n = 4) iMural cells. Center values are mean and error bars are SD. Significance determined by One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison test to APOE3/3 iMural cells E3/3 v E4/4, p = 0.0002; E3/3 v sAD231, p = 0.0315; E3/3 v sAD332, p <0.0001; E3/3 v H9, p =0.0005. f, Violin plots depicting APOE expression in pericytes isolated from post-mortem hippocampus of APOE4 carriers (n = 16 individuals) Non-carriers (n = 46 individuals) Dashed line is mean. Differential expression was measured using a two-tailed Wilcoxon rank sum test, considering cells with detected expression of APOE. g, Representative images and quantification depicting the expression of APOE protein in hippocampal NG2-positive pericytes in post-mortem brains from APOE4-carriers (n= 6) and non-carriers (n = 6). For each genotype more than 250 NG2-positive pericytes were identified. Center values are means and error bars are SD. Two-sided Unpaired t test, p = 0.0068. h, Isogenic iBBBs that are deficient for APOE by genetically knocking-out (KO) display similar amyloid accumulation to E3/3 iBBBs. (n = 9) Center values are mean and error bars are SD. Significance displayed as One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison test. ** = 0.0091; *** = 0.0006. i, Immunodepleting APOE from APOE4 iMural cells conditioned media significantly reduces amyloid accumulation in the APOE3 iBBB. (n = 12) Center values are mean and error bars are SD. One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison test where iBBB genotype/ iMC conditioned media/ immunodepletion antibody: 3/NA/NA v 4/NA/NA, p = 0.0051; 3/3/NA v 3/4/NA, p 0.0011; 3/3/NA v 3/4/IgG, p 0.0002; 3/4/NA v 3/4/αAPOE, p 0.0001; 3/4/IgG v 3/4/αAPOE, p < 0.0001. j, Transcription factors differentially expressed between APOE3/3 and E4/4 isogenic pairs (q < 0.05). Blue denotes down-regulated transcription factors and red dots denote up-regulated transcription factors. The five transcription factors highlighted are reported to bind APOE gene regulatory elements. Analysis and samples the same as described in 4a. k, APOE isogenic iMural cells stained for NFATc1 (green) and SM22 (red). NFATc1 is present in both cytoplasm and nucleus. Dephosphorylation of NFAT by calcineurin leads to NFAT translocation to the nucleus. Quantification of NFATc1 staining per nuclei for each APOE3/3 and APOE4/4. 150 cells were analyzed for each genotype. Significance determined by two-sided students t-test, (p < 0.0001). l, Violin plots depicting NFATC1 and NFATC2 expression in endothelial cells isolated from post-mortem hippocampus of APOE4-carriers (n = 16 individuals) and non-carriers (n = 46 individuals) Differential expression was measured using a two-tailed Wilcoxon rank sum test p = 3 e-06. Dashed line is mean.

Fig. 5.. Inhibition of Calcineurin reduces APOE expression and ameliorates Aβ deposition

a and b Expression of APOE in isogenic (a) and heterozygous (b) iMural cells after two weeks treatment with DMSO, CsA, FK506 or INCA6. One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison. N = 6 RNA samples prepared from independent wells. Center values are means and error bars are SD. (a) For CsA: E3/3, p = 0.0002; E4/4, p = 4.52 e-005; sADE3/3, p = 1.41 e-006; sADE4/4, p = 4.717 e-006. For FK506: E3/3, p = 0.006; E4/4, p =0.018; sADE3/3, p =0.142; sADE4/4, p = 0.0005. For INCA6: E3/3, p = 0.226; E4/4, p = 0.002; sADE3/3, p = 0.052; sADE4/4, p = 0.0001. (b) For CsA: sAD330, p = 384 e-006; sAD369, p = 0.006; sAD332, p =0.0004; sAD231, p = 2.17 e-008. H9, p = 0.025. For CsA: sAD330, p = 0.021; sAD369, p = 0.387; sAD332, p = 0.029; sAD231, p = 8.08 e-008. H9, p = 0.046. For CsA: sAD330, p = 0.57; sAD369, p = 0.095; sAD332, p = 0.008; sAD231, p = 5.6 e-007. H9, p = 0.015. c, Soluble APOE protein is significantly reduced following two-week treatment with calcineurin inhibitor CsA. APOE concentration in iMural cell conditioned media was quantified using ELISA from three separated biological replicates. Center value is mean and error bars are SD. Two-sided Multiple Student t-tests. Discovery determined using FDR method with Benjamini and Hochberg with Q = 1%. sAD369, p = 0.004; E3/3, p = 0.06; E4/4, p = 0.008; E4/4 clone 2, p = 0.007; sAD E4/4, p = 0.015; H9, p = 0.035. d and e, Expression of NFATc1 (d) and APOE (e) is down-regulated in iMural cells by CsA treatment. Bars are mean value from 3 biological replicates One-way ANOVA For NFATc1, E3/3 v E4/4 DMSO p = 0.0012; E4/4 DMSO v E4/4 CsA, p = 0.0068. For APOE, p < 0.0001) with Bonferroni’s multiple comparison Center values are mean expression from RNA prepared from 3 separate wells and error bars are SD f, Heat map depicting differentially expressed genes between isogenic APOE3/3 iMural cells treated with DMSO and APOE4/4 iMural cells treated with DMSO, or 2 μM CsA. Genes and organized by hierarchical clustering using Spearmann’s Rank correlation with average linkage. Boxes outline genes clustering together. The total genes for each cluster are presented on the right side of the heatmap depicted values are mean normalized counts from 3 independent biological replicates g, Representative images of E4/4 iMural cells treated with DMSO, CsA, or FK506 for two weeks and then exposed to 20 nM Aβ-FITC for 96 hours. Experiments were repeated at least 3 times (Scale bar = 10 μm) h, Quantification of amyloid accumulation in iBBBs treated with DSMO, CsA, or FK506. iBBBs were pre-treated with chemicals for two weeks and then exposed to 20 nM Aβ for 96 hours. Center values are mean and error across 10 iBBBs for each condition. Significance determined via One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison. E3/3 DMSO v E4/4 DMSO, p < 0.0001; E4/4 DMSO v E4/4 CsA, p < 0.0001; E4/4 DMSO v E4 FK506, p p < 0.0001; sADE3/3 v sADE4/4, p = 0.0001; sADE4/4 DMSO v sADE3/3 CsA and FK506, p < 0.0001; sADE4/4 DMSO v sADE4 CsA, p 0.0003; sADE4/4 DMSO v sADE4/4 FK506, < 0.0001. i, Quantification of amyloid accumulation in APOE3/4 heterozygous iBBBs treated with DSMO, CsA, or FK506. iBBBs were pre-treated with chemicals for two weeks and then exposed to 20 nM Aβ for 96 hours. Center values and error bars are mean and SD amyloid accumulation across 10 iBBBs for each condition. Significance determined via One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparison. For sAD231: DMSO v CsA, p < 0.0001; DMSO v FK506, p = 0.0023. For sAD332: DMSO v CsA, p < 0.0001; DMSO v FK506, p < 0.0001. For H9: DMSO v CsA, p < 0.0001; DMSO v FK506, p < 0.0001. j, Quantification of Aβ accumulation in iBBBs treated with conditioned media from APOE4/4 iMural cells that were treated with calcineurin inhibitors for one at least week prior media harvesting. Center values and error bars are mean and SD amyloid accumulation across 12 iBBBs for each condition. One-way ANOVA (p < 0.0001) with Bonferroni’s multiple comparisons compared to E3/3. E3/3 v: E3/DMSO, p < 0.0001; E3/CsA, p = 0.3413; E3/FK506 > 0.9999; E3/INCA6, p = 0.3112. k, APOE protein concentration in the hippocampus of mice treated with either cyclosporine A or vehicle. APOE was measured by ELISA. Each dot represents mean APOE concentration from one mouse. Center value are means from 4 independent wells. Error bars are SD. Unpaired two-tailed t test (p = 0.0456). l, Representative image and quantification of immunostaining for APOE in cortical pericytes from APOE4 KI x 5xFAD mice treated with cyclosporine A or vehicle. Unpaired two-tailed t test (p = 0.0427). Center values are means from four mice. Error bars are SD. m, Representative image of concurrent reduction of vascular APOE protein and amyloid following a three-week treatment with CsA. This experiment was repeated in 4 mice with similar results. n, Representative images and quantification of vascular amyloid in the hippocampus following treatment of 6-month-old APOE4KI x 5XFAD female mice with either vehicle or CsA for three weeks. Amyloid was detected and quantified with two independent anti-amyloid antibodies (6e10 and 12F4). Center values are means from four mice for each condition. Error bars are SD. This experiment was repeated twice with similar results. Unpaired two-tailed t test (6e10, p = 0.0055; 12F4, p = 0.0242). (Scale Bars = 25 μm).