Amyloid beta peptides (Aβ) from Alzheimer’s disease neuronal secretome induce endothelial activation in a human cerebral microvessel model

2.1. APPSwe hiPSCs-CNs secretes Aβ in a genotype-dependent manner

To investigate the effect of AD neuron derived paracrine signaling on HBMECs, we differentiated hiPSCs from an established cell line harboring the APPSwe mutation into cortical neurons (CNs) (Penney et al., 2020b; Young et al., 2018) (Fig. 1A,B). Following a previously published protocol (Knupp et al., 2020; Rose et al., 2018), neural progenitor cells (NPCs) were first differentiated from hiPSCs for 3 weeks, followed by another 3 weeks of differentiation into neurons. At the end of the 3-week differentiation, CNs were enriched in the cultures, as marked by robust immunofluorescent staining for the neuron marker MAP2 (Fig. 1B). hiPSC lines homozygous (Swe^+/+) and heterozygous (Swe^+/WT) for the APPSwe mutation were used to generate hiPSC-CNs (Fig. 1A). Control hiPSC-CNs were made using the parental wild type (Swe^WT/WT) line for the APPSwe lines, and an APP knock out (APP^KO) cell line was used as an additional control (Young et al., 2018, 2015).

After 3-weeks of differentiation, CM was collected from hiPSC-CN cultures and measured for their Aβ 1–42 and 1–40 levels using an ELISA assay (Fig. 1C). CM collected from Swe^+/+ and Swe^+/WT hiPSC-CNs showed increased Aβ 1–42 and 1–40 levels compared to controls, with the highest mean in the Swe^+/+ line for Aβ 1–42 (487 pg/mL) and 1–40 levels (5431 pg/mL). The Swe^+/WT line showed significantly lower Aβ 1–42 (137 pg/mL) and Aβ 1–40 (1033 pg/mL) levels compared to Swe^+/+, but significantly higher than that collected from the Swe^WT/WT (Aβ 1–42 at 50 pg/mL and 1–40 at 385 pg/mL). These results confirm that the APPSwe neurons secrete increased levels of Aβ and may be a viable way to mimic paracrine signals from neurons in the AD brain.

2.2. Paracrine signaling from neurons induces endothelial activation and cytoskeletal reorganization in 2D monolayer culture

Under stress, endothelial cells may be activated, which leads to the formation of stress fibers (Peng et al., 2019). We next treated HBMECs in a 2D monolayer (Fig. 2A (i)) with CM obtained from hiPSC-CN cultures to evaluate the paracrine effect of neurons on HBMECs. Four media groups were tested to assess endothelial activation after a 6-h treatment: Swe^+/+ CM, Swe^+/WT CM, Swe^WT/WT CM, and unconditioned media (UCM). Upon treatment with Swe^WT/WT CM, HBMECs exhibited quiescent cytoskeletal structures characterized by a cortical actin rim around the peripheral junctions (Fig. 2A i and ii-arrows). In comparison, Swe^+/+ CM and Swe^+/WT CM treated HBMECs showed reorientation of actin fibers into elongated stress fibers with increased cell polarity (Fig. 2A iii and iv-arrowheads), suggesting their activated state. In particular, Swe^+/+ CM treated HBMECs show significant retraction of endothelial cells (Fig. 2A (iv-arrowheads)), characterized by degradation of vascular endothelial cadherin (VECAD) junctions in areas of retraction (Fig. 2A (viii-asterisk)), and increased gap formation between adjacent ECs.

2.3. RNA sequencing (RNAseq) reveals endothelial activation signatures in CM treated HBMECs

To understand the transcriptomic changes in endothelial cells following CM treatment, we sequenced RNA collected from 2D HBMECs that were treated with two batches of Swe^+/+ CM, Swe^WT/WT CM, and UCM. Principal component analysis (PCA) showed that individual UCM, Swe^WT/WT, and Swe^+/+ CM treated groups clustered together (Fig. 2B). In particular, the UCM cluster is distinctly separated from the Swe^WT/WT and Swe^+/+ clusters, suggesting that neuronal secretomes had a large effect on EC function compared to UCM. Clusters also form within the same media batch groups, with distinct cluster presentations for Swe^WT/WT and Swe^+/+ groups. Batch-to-batch variation appeared to exist in the CM obtained, leading to overlaps among clusters, which is likely due to the inherent heterogeneity of hiPSC-CNs and secretome production (Supplemental Fig. 1). For analysis of RNAseq, we combined both batches for unbiased representation of transcriptomes that were differentially expressed between Swe^WT/WT and Swe^+/+ treated ECs. Results showed 10 genes that were differentially expressed between HBMECs treated with Swe^WT/WT CM and Swe^+/+ CM (Fig. 2C). This includes upregulation of genes associated with endothelial adhesion molecules: SELE (~7-fold change), VCAM-1 (~6-fold change) and ICAM-1 (~3-fold change) (Fig. 2D and E). CCL2 and SDCBP2 were also upregulated, which supports the presence of higher levels of inflammatory cytokines in Swe^+/+ CM. Down-regulated genes include: INPP5J, TSPYL5, SOCS3 and PDZD7 (Fig. 2C and D). Downregulation of the TSPYL5 gene, known to be associated with endothelial proliferation and tube formation, suggests angiogenic ability may be attenuated in ECs with AD exposure (Na et al., 2019). GO term and KEGG pathway analysis showed upregulation of pathways related to leukocyte adhesion, cellular extravasation, and endothelial inflammation (Fig. 2F). 24-h treatment of HBMECs with Swe^+/+ CM and Swe^WT/WT CM confirmed increased VCAM-1 expression on the endothelial surface upon stimulation with Swe^+/+ CM (Supplemental fig. 2).

Overall, both phenotypic and transcriptomic analyses of HBMECs suggest that ECs are activated by neuronal secretomes in AD CM through cytoskeletal reorganization, and transcriptional alteration. These data indicate that paracrine signaling from AD neurons elicits a pro-inflammatory, pro-thrombotic, and anti-angiogenic environment in brain ECs.

2.4. Swe^+/+ conditioned media induces endothelial activation in 3D flow-mediated engineered cerebral microvessels

The endothelium in vivo forms an intact lumen and 3D network structure, responding to flow and cytokines synchronically. When activated, the endothelium is known to release granules, such as VWF, which can extend and multimerize into large fibers, under flow, to bind platelets and other blood cells (Zheng et al., 2015). Such activation denotes a shift of the endothelium toward a prothrombotic phenotype and has been shown as a signature phenomenon in 3D microvessels (Zheng et al., 2015). We examined whether Swe^+/+ CM would induce rapid activation of 3D flow-mediated engineered cerebral microvessels. Using the engineered microvessel method previously developed by our lab (Zheng et al., 2012), we created devices with a double grid geometry (8 × 8) for microvessels with a feature height of 150 μm (Fig. 3A) and seeded with HBMECs at a concentration of 7 × 10⁶ cells/mL. After culture for 5–7 days under gravity-driven flow, a continuous and confluent endothelium was established in 3D engineered cerebral microvessels (Fig. 3A).

We then perfused cerebral microvessels with different CM for 1 h under gravity-driven flow (Fig. 3A) and evaluated the formation of VWF fibers in the vessel lumen. Treatment of vessels with Swe^+/+ media led to the formation of large transluminal VWF fibers throughout the vessels (Fig. 3B (green) and 3C), which was not observed upon perfusion with CM from the Swe^WT/WT line or UCM. The extended VWF fibers in the vessel lumen expose binding sites for platelets, which could initiate thrombosis in vessels (Drakeford and O’Donnell, 2017). We next perfused washed platelets isolated from whole blood, stained for the platelet marker CD41a, through the microvessels. Upon perfusion, platelets readily bound to VWF fibers in the Swe^+/+ condition, but there was minimal binding in vessels with no VWF fibers (UCM and Swe^WT/WT) (Fig. 3B (red)). Using VWF staining and platelet perfusion, two types of VWF fibers were identified in perfused microvessels, including transluminal fibers, seen as large fibers in the center of the vessel lumen, and wall fibers, as small fibers found on the wall of the vessels that are marked by platelet strings (Supplemental Fig. 3). Brightfield and scanning electron microscopy of the Swe^+/+ CM treated microvessels showed that large aggregates of platelets are bound to ultra-large transluminal VWF fibers (Fig. 3C(i)) whereas single platelet binding occurred on wall string fibers. (Fig. 3C(ii) and 3D). Quantification of wall and transluminal fibers in vessels perfused with Swe^+/+ CM showed a significant increase in transluminal and total fibers, and a trend toward increased wall fiber formation across several replicates compared to controls (UCM and Swe^WT/WT) (Fig. 3E).

Furthermore, we observed that in brain microvessels perfused with Swe^+/+ CM, leukocytes were able to bind to the endothelial surface and initiate a tethering and rolling response. This suggests that endothelial cells were activated by CM with higher Aβ concentrations which facilitated immunoadhesion within the microvessels (Supplemental Fig. 4; Supplemental Video 1). These results suggest that perfusion of engineered microvessels with CM with high Aβ levels can induce EC activation and VWF fiber formation that is not present in vessels perfused with CM from cell lines with low Aβ production.

2.5. Knocking out APP and inhibiting the beta-secretase enzyme BACE1 with a small molecule inhibitor (BACEi) attenuate endothelial cell activation and VWF release

To test whether the increased level of Aβ in our APPSwe CM is a driving factor for endothelial activation, we perfused microvessels with CM from Swe^+/+, Swe^+/WT, and Swe^WT/WT hiPSC-CNs treated with a BACE1 inhibitor (BACEi). BACEi inhibits BACE1 function and prevents the amyloidogenic cleavage of APP into Aβ (Fig. 4A) (Chow et al., 2010; Satir et al., 2020). CM was collected from cultures of hiPSC-CNs after treatment with either BACEi (25 nM) or vehicle control (DMSO) for 72 h (Fig. 4B). ELISA measurements showed a statistically significant 44% decrease in Aβ 1–42 levels (DMSO: 170 pg/mL, BACEi: 75.4 pg/mL) and 45% decrease in Aβ 1–40 levels (DMSO: 1504 pg/mL, BACEi: 683 pg/mL) in Swe^+/+ BACEi treated cultures compared to DMSO controls (Fig. 4C and Supplemental Fig. 5). In Swe^+/WT cultures, BACEi treatment led to a statistically significant 38% decrease in both Aβ 1–42 (DMSO: 101 pg/mL, BACEi: 38.1 pg/mL) and 1–40 (DMSO: 794 pg/mL, BACEi: 298 pg/mL) levels compared to controls. In Swe^WT/WT cultures, BACEi treatment led to a 28% reduction in Aβ 1–42 levels (DMSO: 32.4 pg/mL, BACEi: 9.14 pg/mL) and 26% reduction in Aβ 1–40 levels (DMSO: 264 pg/mL, BACEi: 69.1 pg/mL).

Furthermore, perfusion of cerebral microvessels with Swe^+/+ and Swe^+/WT BACEi CM led to the formation of only a few VWF fibers, in contrast to the significantly larger number of fibers in Swe^+/+ and Swe^+/WT DMSO CM conditions (Fig. 4D). Quantification of the fiber number in each media condition confirms this consistent and significant decrease across multiple replicates in BACEi treated conditions compared to the corresponding DMSO controls (Fig. 4F). In our Swe^WT/WT control line, there was little to no fiber formation in both the BACEi and DMSO conditions as expected (Fig. 4D).

To test whether VWF fibers would form in the presence of neuronal secretomes but in the complete absence of Aβ or other components of amyloidogenic APP processing, CM was collected from APP^KO hiPSC-CNs (Young et al., 2018). Analysis of Aβ 1–42 and 1–40 levels showed that CM from the APP^KO line did not secrete detectable levels of Aβ peptides as expected (Fig. 4C and Supplemental Fig. 4). When APP^KO CM was perfused through brain microvessels, no transluminal VWF fibers were observed in the vessel lumen (Fig. 4E and F).

Notably, the number of VWF fibers in the HBMEC microvessels appears to have a positive linear correlation with the levels of Aβ 1–42, across all the media groups (Fig. 4G). For example, Swe^+/+ DMSO CM had Aβ 1–42 levels of 170 pg/mL and led to maximal fiber production of, on average, 45 VWF fibers per vessel. In contrast, Swe^+/WT DMSO CM had 41% less Aβ 1–42 (101 pg/mL) and 46% less average VWF fiber production (24 VWF fibers per vessel) compared to Swe^+/+ DMSO CM, demonstrating a strong correlation between Aβ 1–42 levels and VWF fiber production (Fig. 4G).

Next, we performed immunoprecipitation (IP) of Aβ peptides in Swe^+/+ and Swe^WT/WT CM to investigate whether the presence of Aβ peptides in Swe^+/+ was necessary for inducing EC activation and VWF fiber formation in engineered brain microvessels. Aβ peptides in Swe^WT/WT and Swe^+/+ CM were precipitated using the anti-Aβ 6E10 antibody. An ELISA assay following IP confirmed successful removal of Aβ peptides from Swe^+/+ and Swe^WT/WT CM. (Supplemental fig. S6). Next, control Swe^WT/WT and Swe^+/+ CM (control CMs) and immunoprecipitated Swe^WT/WT and Swe^+/+ CM (IP CMs) were perfused in engineered brain microvessels. VWF staining in the microvessels revealed a reduced number of fibers formed in Swe^+/+ IP CM compared to its Swe^+/+ control CMs. Furthermore, perfusion of both Swe^WT/WT control and IP CMs resulted in a lack of VWF fibers, further supporting that Aβ peptides were an important component of the CM for inducing EC activation and VWF fiber formation.

Taken together, the number of fibers formed in BACEi treatment, APP^KO CM perfusion and Aβ-depleted CM perfusion relative to Aβ peptide levels suggest that Aβ is an essential component of the CM for inducing EC activation and VWF fiber formation in engineered cerebral microvessels.

2.6. Analysis of pro-inflammatory cytokines and pTau from the neuronal secretome further supports increased Aβ levels as a primary factor in endothelial cell activation

To determine whether secretion of pro-inflammatory cytokines into our CM acts as a confounding factor driving EC activation, independent of Aβ, we performed a multiplexed cytokine ELISA assay. We screened untreated CM from APP^KO, Swe^WT/WT, Swe^+/WT, and Swe^+/+ hiPSC-CNs and CM from DMSO and BACEi treated Swe^WT/WT, Swe^+/WT, and Swe^+/+ hiPSC-CNs. APP^KO CNs were not treated with DMSO or BACEi due to the lack of APP production and therefore were not included in those analyses. Our cytokine panel included 9 pro-inflammatory cytokines and chemokines: TNF-α, IP-10, IL-8, IL-6, IL-15, IL-1β, IFN-β, IFN-α2a, and Eotaxin.

We observed that there was no trend toward increased cytokine secretion in untreated Swe^+/+ and Swe^+/WT CM as compared to Swe^WT/WT, or APP^KO for any of the measured cytokines (Supplemental Fig. 6). This suggests that the increased number of VWF fibers observed in microvessels treated with Swe^+/+ and Swe^+/WT CM was not driven by the cytokines present in CM. Surprisingly, for the majority of cytokines, IP-10, IL-8, IL-6, IFN-β, IFN-α2a, and Eotaxin, the highest cytokine levels were observed in our control CM (Swe^WT/WT or APP^KO). Additionally, across all measured cytokines, we saw no trend suggesting BACEi CM had decreased cytokine levels compared to DMSO CM (Supplemental Fig. 7).

To rule out secreted phosphorylated tau (pTau) as a potential driver of the EC activation, we performed a pTau and total Tau (tTau) ELISA assay to quantify pTau levels in our CM. We screened media from APP^KO, Swe^WT/WT, Swe^+/WT, and Swe^+/+ hiPSC-CNs. We found that pTau secretion into the CM was negligible for all of the measured lines and the levels of pTau within CM did not correlate with the EC activation we observed in microvessels treated with Swe^+/+ and Swe^+/WT CM (Supplemental fig. S8).

Overall, we conclude that that the APPSwe mutation does not drive increased pro-inflammatory cytokine secretion from neuronal cultures, treatment of CNs with a BACEi does not lead to reduced cytokine secretion, and pTau secretion in conditioned media is negligible across all lines and does not correlate with EC activation. Taken together, our results suggest that the endothelial activation observed in our studies is strongly correlated to the Aβ levels in neuronal secretomes.

4.1. Endothelial cell culture

Human brain microvascular endothelial cells (HBMECs) (CC# ACBRI 376 V; Cell systems) at passage number 3–6 were used for all experiments. HBMECs were maintained in culture flasks coated with Attachment Factor (CC# 4Z0-210; cell systems) and fed growth medium (GM) (CC# 3202; EGM-2 MV Microvascular Endothelial Cell Growth Medium-2 BulletKit (CC# 3156; EBM-2 Basal Medium), 5% fetal bovine serum (FBS), hydrocortisone, human fibroblast growth factor-beta (hFGF-B), vascular endothelial growth factor (VEGF), R3-insulin-like growth factor-1 (R3-IGF-1), ascorbic acid, human epidermal growth factor (hEGF), GA-1000 (gentamycin and amphotericin)). Cells were maintained at 37 °C in a 5% CO2 incubator and media was replaced every 3 days.

4.2. Cell lines and hiPSC neuronal differentiation

To generate AD mimicking neurons from hiPSCs we used previously established gene-edited hiPSC lines heterozygous (Swe^+/WT) and homozygous (Swe^+/+) for the APP Swedish mutation as well as the parental control (Swe^WT/WT) described in Young et al. (Young et al., 2018). These cell lines were generated from a previously published and characterized CV background human induced pluripotent stem cell line (Knupp et al., 2020). This is a male cell line with an APOE ε3/ε4 genotype (Levy et al., 2007). A previously published APP knockout (APP^KO) line was used as an additional control for conditioned media (CM) experiments (Young et al., 2018). This cell line was made from a parental APP duplication hiPSC line and was engineered using CRISPR/Cas9 genome editing to lack APP production. Neurons from these cell lines were differentiated from hiPSCs using dual-SMAD inhibition techniques as previously described (Knupp et al., 2020; Rose et al., 2018). For the experiments presented here, we started with neural precursor cells (NPCs) previously differentiated from hiPSCs and frozen. NPCs for all the APPSwe lines (Swe^+/WT, Swe^+/+, Swe^WT/WT) and the APP^KO line were thawed and plated on a Matrigel (Growth factor reduced basement membrane matrix; CC# 356231; Corning) coated 6-well plates. NPCs were cultured in Basal Neural Maintenance Media (BNMM) + FGF (1:1 DMEM/F12 + Glutamine (CC# 11320–033; Gibco) and Neurobasal media (CC# 21103–049; Gibco), 0.5% N2 Supplement (CC# 17502–048; Gibco), 1% B27 Supplement (CC# 17504–044; Gibco), 0.5% GlutaMax (CC# 35050061; Thermo Fisher Scientific), 0.5% insulin-transferrin‑selenium (CC# 41400045; Thermo Fisher Scientific), 0.5% NEAA (CC# 11140050; Thermo Fisher Scientific), 0.2% b-mercaptoethanol (CC# 21985023, Life Technologies) + 20 ng/mL FGF (R&D Systems, Minneapolis, MN)). The NPCs were allowed to reach 70% confluence for 1–3 days with media changes every other day. Once the NPCs were sufficiently confluent, cells were dissociated with Accutase (CC# 07930, STEMCELL Technologies) and passaged into Matrigel-coated 10 cm culture plates for neural differentiation. 24 h after replating, the cells were switched to Neural Differentiation media (BNMM +0.2 mg/mL brain-derived neurotrophic factor (CC# 450–02; PeproTech) + 0.2 mg/mL glial-cell-derived neurotrophic factor (CC# 450–10; PeproTech) + 0.5 M dbcAMP (CC# D0260; Sigma Aldrich). Media was changed twice a week for 21 days at which point the differentiation is considered finished. Cells were maintained at 37 °C in a 5% CO2 incubator.

4.3. Conditioned media preparation

Following 21 days of differentiation, Neural Differentiation media was refreshed and was then kept on the cells for at least 72 h with no media change to generate conditioned media (CM) for 2D and 3D treatment experiments. After 72 h, the media was collected and stored at 4 °C for short-term storage and −20 °C for long-term storage. To generate CM from BACEi treated APPSwe neurons, neurons were dissociated from 10 cm plates using Accutase following 21 days of differentiation and split into 3 wells of a Matrigel-coated 6 well plate. This was repeated for all cell lines. At least 3 days later, neurons were treated with either 25 nM β-secretase inhibitor (CC# HY-13240; BACEi LY2886721; MedChemExpress) or dimethyl sulfoxide (DMSO, as vehicle control) or left untreated every 24 h 72 h with no media change. After 72 h of treatment, media from untreated, DMSO, and BACEi treated neurons was harvested for perfusion through microvessels, treatment in 2D, and quantification of Aβ1-40 and Aβ1-42 peptides secreted by neurons.

4.4. Immunoprecipitation of Aβ

For immunoprecipitation of Aβ from our CM, 4 μg/mL of mouse anti-Aβ 6E10 antibody (CC# 803001, BioLegend) was added to CM samples from the Swe^+/+ and Swe^WT/WT lines as previously described (Dewji et al., 2006; Oberstein et al., 2021; Welzel et al., 2014). Media containing the antibody was incubated tilting, overnight at 4 °C. 50 μl of M-280 Sheep Anti-Mouse IgG Dynabeads^™ (CC# 11202D, Invitrogen) for every 4 μg of antibody were added to each CM sample according to the manufacturer’s instructions and the samples were incubated tilting at 4 °C for 1 h. The Dynabeads^™ covalently coupled to 6E10 labeled Aβ were removed from the media using a magnet according to the manufacturer’s instructions. Supernatant was removed from the magnetically separated beads and CM lacking Aβ was collected and stored at −80 °C for downstream analysis and perfusion through engineered microvessels.

4.5. Amyloid beta measurements

Aβ1-40 and Aβ1-42 peptides were measured using the methods previously described by Young et al., 2015. (Young et al., 2015) Conditioned media collected from the Swe^+/+, Swe^+/WT, Swe^WT/WT, and APP^KO cell lines after a 3-week differentiation were frozen at −80 °C until use. The media was run on an Aβ Triplex ELISA plate (CC# 151200E-2; MesoScale Diagnostics).

4.6. Meso-Scale Discovery (MSD) U-PLEX platform to measure extracellular levels of cytokines and chemokines

Pro-inflammatory proteins in the media were measured using the MSD U-PLEX platform (MesoScale Diagnostics). For this assay, conditioned media collected from the Swe^+/+, Swe^+/WT, Swe^WT/WT, and APP^KO cell lines after a 3-week differentiation were frozen at −80 °C until use. 25 μl of media was used to run an MSD U-plex assay as per manufacturer’s protocol and MSD Quick plex SQ120 instrument was used to detect analytes. This MSD U-Plex plate detected a panel of pro-inflammatory cytokines including IL-1β, IL-6, IL-8, IFN-β, TNF-α, IL-15, IFN-α2a and chemokines including Eotaxin, and IP-10, known to be elevated in AD patients.

4.7. Total and phospho-tau measurements

Samples were prepared by plating differentiated neurons from the Swe^+/WT, Swe^+/+, Swe^WT/WT and APP^KO lines in a 96 well plate at 300,000 cells per well. Conditioned media and cell lysate were harvested and pooled from six wells per line. Conditioned media was incubated on the cells for at least 72 h prior to collection. pTau (Thr231) and tTau levels were measured in these samples using pTAUThr231/tTAU ELISA plates (CC# K15121D-1, Meso Scale Discovery).

4.8. 3D cerebral microvessel fabrication

Engineered microvessels were fabricated using soft lithography and injection molding techniques previously described by Zheng et al., 2012. Briefly, type I rat tail collagen was dissolved in 0.1% acetic acid at a stock concentration of 15 mg/mL (Zheng et al., 2012). The collagen was neutralized and diluted to 7.5 mg/mL on ice with 1 M NaOH (20 mM final), 10× M199 (CC# 11825015; Thermo Fisher Scientific), and GM. A microfabricated polydimethylsiloxane (PDMS) stamp was used to define a double-grid vessel network with a feature height of 150 μm. Collagen was molded around this PDMS stamp. A flat collagen gel was also generated. After gelation at 37 °C, the channels were incubated with the flat collagen gel to generate closed off microchannels that can be perfused and seeded with endothelial cells (ECs). These two gels merged through additive bonding upon incubation at 37 °C. Once the gel was fabricated, HBMECs were lifted from culture flasks with trypsin (CC# MT-25-052-CI; Thermo Fisher Scientific) and resuspended in GM at a concentration of 7–10 × 10⁶ cells/mL. Using a gel loading pipette tip, 10 μL of this cell suspension is injected into the inlet of the microvessel and allowing them to circumferentially attach to the collagen gel under static conditions for 1 h prior to the start of gravity-driven flow. Microvessel constructs were cultured under gravity-driven flow for 5–7 days with media replenishment approximately every 12 h.

4.9. Conditioned media treatment

For the treatment of HBMECs with CM in 2D, HBMECs were plated at a density of 0.03 M/well in a glass bottom chamber slide (CC# 80827, Ibidi) and allowed to recover for 48 h prior to treatment. Prior to treatment, the cells were washed with EBM-2 basal media (serum-free) for 5 mins. Then, 250 μL of CM per well was then added to the cells for 8 h, following which the cells were fixed in 4% PFA. To treat our engineered microvessels in 3D, vessel fabrication was performed as described above. Following vessel fabrication and culture for 5–7 days under gravity-driven flow, the vessels were treated with CM from all of our APPSwe lines and CM from BACEi treated neurons. Vessels were washed with EBM-2 basal media (serum-free) for 20 mins. Vessels were then perfused with CM for 1 h. Media was replenished every 20 mins to continue gravity-driven flow throughout the duration of treatment. Following 1 h of perfusion, the vessels were fixed with 4%PFA for 20 mins for downstream analysis and washed with phosphate buffered saline (PBS) three times for 20 mins each. For RNAseq analysis, vessels were treated for 8 h with media replenished every 2 h.

4.10. RNA isolation and RNAseq analysis

RNA was collected from two different culture subtypes (2D cultures and 3D microvessels) for three different media conditions (unconditioned BNMN media, conditioned control media and conditioned Swe^+/+ media). RNA from 2D cultures was collected following media treatment from 24 well plates by addition of RLT lysis buffer and pipette homogenization. RNA from 3D vessels was collected by perfusion of RLT lysis buffer in the inlets and collection of the lysates in the outlets. Lysates from three wells in 2D cultures were pooled as one sample replicate, and two vessels from 3D cultures were pooled as one replicate. Total RNA was isolated from the lysate using on-column digestion of genomic DNA via RNeasy Micro Kit (CC# 74004; Qiagen). RNA quality was assessed using High sensitivity RNA TapeStation 4200 (Agilent Technologies) and total RNA samples with RNA integrity number > 8 were used to proceed to library preparation. For sequencing, cDNA preparation was performed using Takara SMART-Seq v4 Ultra Low Input RNA Kit (SMARTv4), library preparation using Illumina Nextera XT DNA Library Prep Kit (Nextera XT) and sequenced with Illumina NovaSeq 6000 (paired-end 50 cycles). Sequenced reads were aligned using STAR2 aligner and gene-level raw counts was obtained using STAR2’s internal quantification method. Gene ontology analysis was performed using iDEP.951 for pathway analysis of differentially expressed genes.

4.11. Isolated platelet and peripheral blood mononuclear cell (PBMC) perfusion in microvessels

Whole blood was drawn from normal healthy donors into a blood collection tube (BD, CC# 364606; Vacutainer) containing 1.5 mL of acid citrate dextrose (ACD). Platelets were isolated from whole blood using an existing protocol provided by the manufacturer. Briefly, collected blood was centrifuged at 200 g for 20 min (acceleration 4; deceleration 0) at room temperature for platelet-rich plasma (PRP) extraction. PRP was subsequently centrifuged at 1200 g for 10 min to obtain pellets containing the isolated platelets. Pellets were resuspended in CGS buffer containing 120 mM NaCl, 13 mM Sodium Citrate and 30 mM Glucose supplemented with Prostaglandin E1 (sigma P5515-1MG) at 1:10,000 (vol: vol) to inhibit aggregation of platelets. Isolated platelets rinsed in CGS/PGE1 buffer were counted under a bright-field microscope using a hemocytometer and centrifuged again using the previous setting. Pellets were resuspended carefully in Tyrode’s buffer containing 5.5 mM glucose, 10 mM HEPES, 138 mM NaCl, 12 mm HaHCO3, 2.9 mM KCL and 0.36 mM NaH2PO4 and stained with CD41a (CC# 555467; BD) conjugated with phycoerythrin (PE) for 30 min. Following staining, platelet concentration was adjusted to 5 × 10^{^}7 platelets/mL by supplementing an additional Tyrode’s buffer.

Isolated platelets were perfused in conditioned media treated microvessels to assess the functional activation of endothelial cells. CD41a labeled isolated platelets were perfused in HBMEC microvessels treated with conditioned media for 1 h through transfer pipettes to avoid shear-induced activation. Approximately 150 μl of platelets resuspended in Tyrode’s buffer was pipetted into the inlets of microvessels and incubated at 37C for 20 min. Following platelet treatment, vessels were washed three times, each rinse for five minutes, to eliminate residual unbound platelets within the vessel lumen. Immediately after the wash step, microvessels were fixed in 4% paraformaldehyde solution and rinsed with PBS three times. Platelet adhesion analysis was performed after performing immunofluorescent staining of VWF and confocal imaging.

PBMCs were also isolated from freshly drawn blood samples collected in ACD blood collection tube (BD CC# 364606; Vacutainer). PBMCs were isolated using density gradient centrifugation in a 50 mL SepMate tube (StemCell Technologies) following manufacturer’s protocol. PBMCs were stained with pan-leukocyte marker CD45 (CC# 555485; BD biosciences) for live imaging of leukocyte adhesion on Swe^+/+ CM treated HBMEC microvessel for 1 h.

4.12. Scanning electron microscopy imaging of platelet bound VWF fibers

Swe^+/+ perfused microvessels were fixed in ½ strength Karnovsky’s fixative (2% paraformaldehyde and 2.5% glutaraldehyde) overnight at 4C. The top and bottom layers of the collagen matrices were gently teased apart and rinsed with 0.1 M cacodylate buffer. Samples were dehydrated with graded series of alcohols and critical point dried (Autosamdri, Tousimis Corp). Microvessel layers were sputter coated with Au/Pd alloy (Denton Desk IV, Denton Vacuum) and imaged using a JSM 6610 LV scanning electron microscope (JEOL) at an accelerating voltage of 5 kV.

4.13. Flow cytometry analysis

To characterize vascular cell adhesion molecule 1 expression on the surface of endothelial cell following CM treatment, multiparameter FACS analysis of CM treated human brain endothelial cells was performed. BD FACS Canto II was used to analyze VCAM-1 positive cells using FTIC-anti-VCAM-1 antibody (CC# ab24853; Abcam) and the number of positive cells were quantified using with FlowJo software (Tree Star). Unstained controls were used to compensate and adjust voltage.

4.14. Immunofluorescent staining and image analysis

At the end of each experimental time point, both microvessels and 2D endothelial cultures were fixed with 4% paraformaldehyde for 20 min and washed with PBS three times. For immunofluorescent staining of microvessels, reagents were perfused through the microchannels through the inlet of the acrylic jigs. PBS containing 2% bovine serum albumin (BSA) was used to block microvessels and 2D cultures for analysis of extracellular VWF. Following 1 h of blocking, FITC conjugated VWF antibody (1:100; Abcam CC# ab8822) was added to samples for 1 h at room temperature and washed three times with PBS before proceeding to stain with the remaining antibody of interest. Treatment with 2% BSA with 0.1% Triton-x preceded before staining with antibodies including CD41a (CC# 555467; BD), phalloidin (CC# A12380; Thermo Fisher), CD144 (CC# 136008; BioLegend) for cell permeabilization and blocking non-specific binding. All stained microvessels and 2D cultures were imaged using Yokogawa W1 spinning disk confocal microscope or Leica SP8 laser confocal microscope. For microvessels, the entire collagen construct was imaged using a 5 by 5 large z stack image stitch (10% blending stitch) under a 10× objective with a z-step size of 5.5 μm. 3D reconstruction and cross-sectional images of the microvessels were rendered using ImageJ software.

4.15. VWF fiber quantification

VWF fibers were quantified into two distinct wall-fiber and transluminal fiber categories. Transluminal fibers were quantified from fibers formed cross-sectionally within the microchannels and are characteristically thicker and straighter than wall fibers. Wall fibers were quantified from string fibers found circumferentially on the endothelial surface and are smaller and tortuous by nature. For classification of the number of fibers found in each group, stitched z stack Images of each vessel were divided into three layers: top, center lumen and bottom. The top and bottom layers were constructed from compilation of initial and final 4 z-stacks (~20 μm) of each image construct respectively. The luminal layer was constructed from compiled z stack images excluding the initial and final 4 layers.

4.16. Statistical analysis

Statistical analysis of fiber quantification was performed with either one-way ANOVA or two-way ANOVA for comparison between groups followed by Tukey’s or Šídák’s post hoc test for multiple comparisons between groups. Statistical analysis on the RNAseq was conducted with Fischer’s Exact test. Significance was defined as a value of p < 0.05. For all imaging experiments the data was analyzed in a blinded manner. Statistical analysis was performed using GraphPad Prism software. Information on statistical details of individual experiments can be found in the respective figure legends.