High-density lipoproteins suppress Aβ-induced PBMC adhesion to human endothelial cells in bioengineered vessels and in monoculture

Scaffold-directed human engineered vessels were generated under flow bioreactor conditions as described [17]. Briefly, non-woven polyglycolic-acid meshes (PGA; BMS) were coated with polycoprolactone (PCL, Purac) and polyglycolic acid (PLA, Purac) by dipping into a 1.75% (w/w/w) solution of PCL/PLA/tetrahydrofuran (Sigma Aldrich). A tubular shape (length 10 mm and inner diameter 2 mm) was obtained by heat welding before external coating with a 10% (w/w) PCL/tetrahydrofuran solution. After sterilization with 70% ethanol for 30 min followed by 3 washes with PBS, scaffolds were pre-incubated in advanced DMEM overnight before cell seeding. Human umbilical cord smooth muscle cells were isolated as described [17] and 2 × 10⁶ cells/cm² were seeded in the inner surface of the vascular scaffold using fibrinogen (10 mg/mL clottable protein, Sigma Aldrich) and thrombin (10 mU) as cell carriers [18]. After a short static incubation period of 3 days, vascular constructs were exposed to dynamic conditioning in a flow bioreactor, where the flow of nutrient medium (Advanced DMEM (Invitrogen) with 10% FBS; 0.05% Penicillin/Streptomycin, 1% L-glutamine and 1.5 mM L-ascorbic acid) was directed through the inner lumen of the bioreactor circulation loop at 6 mL/min. After 14 days of flow conditioning, vascular grafts were endothelialzed with HUVECs (1.5 × 10⁶ cells/cm²) seeded into the lumen followed by cultivation under static conditions for 5 days in complete endothelial growth medium (EBM-2; Lonza) supplemented with 10% foetal bovine serum (FBS) and growth supplements to form EGM complete media FBS. After the static phase, vascular grafts were placed back in the bioreactor for 14 additional days with increasing medium flow (4 to 6 mL/min) in complete EGM-2 with 10% FBS.

For histological characterization, engineered vessels were fixed in formalin (Fisher), dehydrated through a graded ethanol series using a Sakura Tissue Processor (Sakura), embedded in paraffin and sectioned at 7 μm thickness. Sections were deparaffinised, rehydrated through a graded ethanol series and stained using haematoxylin & eosin (SigmaAldrich) following the manufacture’s instructions. For immunofluorescence analyses, tissues were cryopreserved in Cryomatrix (ThermoFisher) and sectioned at 20 μm thickness using a cryotome (Leica). Immunostaining was performed as described [17] using antibodies specific for the EC marker CD31 (clone JC/70A Biolegend, 1:50) and α-smooth muscle actin (clone 1A4 SigmaAldrich, 1:200). Sections were counter-stained with DAPI and imaged with an inverted microscope (Carl Zeiss). Endothelium barrier integrity was analyzed by injecting Evans blue (Sigma Aldrich) at a final concentration of 0.5% in the circulation loop of the bioreactor for 10 min followed by continuous PBS washing for 20 min. Vessels were cut open longitudinally and en face preparations were analysed macroscopically with photo documentation.

All experiments were conducted under an approved clinical protocol (UBC Clinical Ethics Research Board H14–03357). Upon receipt of written informed consent, 100 mL of fasted blood was collected from normolipidemic healthy donors into vacutainer tubes. Plasma HDL (1.063–1.21 g/mL) was isolated by sequential potassium bromide gradient ultracentrifugation as described [19]. The purity of the HDL preparations was verified by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by Coomassie blue staining to ensure no low-density lipoprotein (LDL) or albumin contamination. Eight independent donors were used across the experiments, 6 isolated in-house and 2 commercially obtained (Leebioscience). Human-derived, lipid free apoA-I was a kind gift from CSL-Behring. Immortalized human THP1 monocytes (ATCC) were cultured in RPMI containing 10% FBS, 1% Pen/Strep, 2 mM L-glutamine and 0.1% β-mercaptoethanol. Primary human PBMC were isolated from healthy donors by centrifugation on a continuous density gradient (Lymphoprep™, Stemcell) following the manufacturer’s instructions. Freshly isolated PBMC were fluorescently labeled with 10 μM of Cell-Tracker Red for 30 min (Invitrogen) following the manufacturer’s recommendations.

Vascular grafts were perfused with complete EGM-2 with 2% FBS. 1 μM Aβ42 or Aβ40 monomers were injected directly into the graft chamber to mimic Aβ originating from the brain (antelumen) side of the vessel. At time points ranging from 2 to 72 h, THP1 cells were fluorescently labeled with Cell-Tracker Red as described above, injected in the graft circulation at a concentration of 1 × 10⁶ cells/mL and maintained under flow conditions for 3 h. For HDL experiments, vascular grafts were perfused with luminal HDL (200 μg/mL) for 2 h before injecting Aβ in the antelumen side for 8 h. Tissues were longitudinally cut open, washed extensively with PBS and fixed with 4% PFA. After 20 min, tissues were washed 3 times with PBS and mounted in Prolong Gold antifade reagent with DAPI. For each independently seeded tissue, adherent monocytes and monocytes undergoing diapedesis were counted in 3 random squares of 1.23 mm² using a z-stack covering the whole tissue thickness with a SP8 confocal microscope (Leica), averaged and expressed as percent of vehicle normalized to 100% for Fig. 1d and f or percent of Aβ normalized to 100% for Fig. 1 h and j, and Fig. 8a and b.

hCMEC/D3 (Fisher; passage 27–35), and HUVEC (passage 4–7, isolated as described [17]) cells were cultured using complete EGM-2 with 2% FBS, ECs were cultured in a humidified incubator at 37 °C at 5% carbon dioxide. For mechanistic experiments, ECs were treated with SR-BI blocking antibody (NB440–113 Novus, 1:500), the SR-BI inhibitor block lipid transport-1 (BLT-1, SigmaAldrich, 10 μM), the eNOS inhibitor L–NG-nitroarginne methyl ester (L-NAME, SigmaAldrich, 1 mM), the receptor associated protein (RAP, Oxford biome, 1 μM), the RAGE blocking antibody (176,902, R&D Systems, 1:50), heparin (10 mU), heparinase (SigmaAldrich, 0.2 mM) or the CD36 blocking antibody (JC63.1, ABCAM, 1:500) for 1 h before HDL priming. For miRNA experiments, cells were transiently transfected using Lipofectamine 2000 (Life Technology) 2 h before Aβ stimulation with the miR-223 mimetic (Life Technology, 4,464,066 (MC12301), 100 nM) or miR-223 inhibitor (Life Technology, 4,464,084 (MH12301), 100 nM) in EBM-2 containing 0.2% BSA.

ECs were seeded at 1 × 10⁵ cells/well in 24-well plates and cultured until confluent for 2 to 3 days. On the day of the assay, ECs were primed for 2 h with 100 μg/mL bovine serum albumin as vehicle control or 100 μg/mL HDL before stimulation with 1 ng/mL of TNF-α (Preprotech) or various concentrations of monomeric Aβ40 and Aβ42 (0.001–1 μM, California Peptide), prepared as described below. After 3 h, 5 × 10⁵ Cell-Tracker Red-labeled PBMC per well were added to ECs for 3 additional hours before washing 3 times with PBS to remove non-adherent PBMC. Cells were then fixed with 4% PFA for 15 min before 3 additional PBS washes and 4′,6-diamidino-2-phenylindole (DAPI) counterstaining. For each independent experiment, adherent monocytes were counted in 5 random squares of 7.84 mm² using a fluorescent inverted microscope (Zeiss), averaged and expressed as percent of vehicle normalized to 100%. DAPI counterstaining was used to ensure endothelial cell coverage.

Recombinant Aβ40 and Aβ42 peptides (California Peptide) were dissolved in ice-cold hexafluoroisopropanol (HFIP). The HFIP was removed by evaporation overnight. To prepare soluble monomers, the peptide film was reconstituted in DMSO to 5 mM, diluted further to 100 μM in DMEM and used immediately. Oligomers were prepared by diluting the 5 mM DMSO peptide solution in phenol red-free F12 medium (Life Technologies) to a final concentration of 100 μM and incubating for 24 h 4 °C. Fibrils were prepared by diluting the 5 mM peptide solution in 0.1 uM of HCl to a final concentration of 100 μM and incubating for 24 h at 37 °C. Aβ monomer, oligomer and fibril preparations were then either used to stimulate hCMEC/D3 monolayers, as given, or were analysed by Transmission Electron Microscopy (TEM) and dot blot. For TEM, 0.5 μL of 100 μM Aβ preparation was diluted in 2 μL filtered distilled water, spotted onto formvar-coated 200-mesh nickel grids (EM Sciences) and allowed to dry. Grids were then negatively stained with 0.5% aqueous uranyl acetate for 30 s, and viewed on a FEI Tecnai G2 Spirit Transmission Electron Microscope. For dot blot, aliquots of Aβ(0.1 μM) were added to PVDF membrane, which were dried and blocked in 3% skimmed milk PBST. After 1 h, blots were incubated with β-amyloid 1–16 antibody (6E10 Biolegend 1:500), amyloid A11 oligomeric antibody (AB9234 Millipore, 1:1000) or amyloid fibrils OC antibody (AB2286 Millipore, 1:1000) in blocking buffer for 16 h, washed extensively in PBST and incubated with anti-mouse or anti-rabbit (1:1000) secondary antibody in blocking buffer. After 1 h, blots were washed as above and developed using enhanced ECL and a ChemiDoc MP imager.

ECs were seeded at 5 × 10⁵ cells/well in 6-well plates and cultured for 2 days until confluent in complete EBM-2. Sixteen hours before the assay, ECs were serum-depleted in EBM-2 containing 0.2% FBS. On the day of the assay, ECs were incubated with 100 μg/mL of HDL in serum-depleted EBM-2 medium containing 1 μM of 4,5-diaminofluorescein diacetate (DAF2, Caymanchem) at 37 °C. After 6 h, ECs were washed with PBS, trypsinized, and triazolofluorescein fluorescence was measured (excitation wavelength of 485 nm, emission 538 nm), using an Infinite M200Pro plate reader (Tecan). In addition to DAF2 measurement, the phosphorylation of eNOS at Ser1177 was compared to total eNOS by immunoblotting (below) in cell lysates harvested 15 min after HDL treatment.

ECs were seeded at 5 × 10⁵ cells/well in 6-well plates and cultured for 2 days until confluent. On the day of the assay, ECs were treated with HDL, Aβ or TNF-α as above. After stimulation, EC monolayers were washed twice with ice cold PBS (pH 8), cooled on ice for 15 min and biotinylated with 250 μg/mL EZ-link™-sulfo-NHS-biotin (ThermoFisher Scientific) in PBS (pH 8) at 4 °C. After 1 h the reaction was stopped with a 5 min incubation in DMEM with 10% FBS. After two additional PBS washes, ECs were lysed in RIPA buffer. Following quantification of protein concentration using a BCA assay (Thermofisher Scientific), at least 100 μg of protein were incubated with streptavidin-conjugated sepharose beads (Pierce) at 4 °C overnight. Beads were washed 3 times with RIPA buffer and the recovered proteins were resolved on a SDS-PAGE.

ECs were seeded at 3 × 10⁵ cells/well in 12-well plates and cultured until confluent for 2 to 3 days. On the day of the assay, ECs were primed for 2 h with 100 μg/mL HDL before stimulating with 0.1 μM of Aβ40 and Aβ42 monomers at 37 °C for total association, or at 4 °C for cell surface binding. After 3 h, hCMEC/D3 were washed 3 times with PBC and lysed in RIPA buffer containing 10 mM Tris pH 7.4, 150 mM NaCl, 1.0% NP-40, 1.0````% sodium deoxycholate, 0.1% SDS and cOmplete protease inhibitor with EDTA (Roche). Aβ40 (KHB3442, Life Tech) and Aβ42 (KHB3482, Life Tech) were quantified using commercial ELISAs and normalized to total protein concentration. For Aβ uptake, hCMEC/D3 were seeded at 1 × 10⁵ cells/well in 24-well plates and cultured to confluence for 2 to 3 days. On the day of the assay endothelial, ECs were primed for 2 h with 1 mg/mL of HDL before stimulating with 1 μM monomeric FITC-Aβ40 and FITC-Aβ42 (Bachem) prepared as described above. After 3 h at 37 °C, hCMEC/D3 were washed 3 times with PBS and fixed in 4% PFA for 20 min. After one Tris-HCl and two PBS washes, hCMEC/D3 were mounted in Prolong antifade reagent.

For mRNA, cells were lysed in Trizol (Invitrogen) and RNA was extracted and treated with DNase I (Invitrogen) according to the manufacturer’s protocol. cDNA was generated using oligo-dT primers and Taqman reverse transcription reagents (Applied Biosystems). Real-time quantitative PCR was done using FastStart Universal SYBR Green Master reagent (Roche) on a Light Cycler 96 system (Roche) to quantify gene expression relative to vehicle using specific primer against ICAM-1 (fwd: ATGGCAACGACTCCTTCTCG; rev: CGCCGGAAAGCTGTAGATGG) and VCAM-1 (fwd: TGTTTGCAGCTTCTCAAGCTTTT; rev: GATGTGGTCCCCTCATTCGT) and normalized to GAPDH (fwd: CCTGCACCACCAACTGCTTA; rev: CATGAGTCCTTCCACGATACCA). MiRNA was isolated using an miRNeasy mini kit (Qiagen) following the manufacturer’s instructions, respectively. Reverse transcription was performed using 2 μg of total RNA with specific miRNA primers (Life Technology). MiR223 and U6 were quantified using specific TaqMan probes (Life Technology).

ECs were lysed in RIPA buffer containing either cOmplete protease inhibitor or Phosphostop (Roche) and quantified using a BCA assay. Equal amounts of total protein were separated by SDS-PAGE followed by electrophoretic transfer to polyvinylidene fluoride (PVDF) membranes (Millipore). After blocking membranes for 1 h with 5% skim milk powder in PBST, or 5% BSA in TBST for phosphoproteins, ICAM-1, (EP1442Y Abcam, 1:1000), VCAM-1 (EPR5047 Novus, 1:1000), peNOS (ser1177 Cellsignalling, 1:1000), eNOS (M221 ABCAM, 1:1000), AnxA1 (D5V2T, Cellsignalling, 1:1000), p-Akt (Ser473 D9E, Cellsignalling, 1:2000), Akt (9272, Cellsignalling, 1:1000), p-NF-κB p65 (Ser536 93H1, Cellsignalling, 1:1000), NF-κB p65 (D14E12, Cellsignalling, 1:1000), p-SAPK/JNK (Thr183/Tyr185 G9, Cellsignalling, 1:1000), JNK (2C6, Cellsignalling, 1:1000), p-p42/44 MAPK (Erk1/2) (Thr202/Tyr204 D13.14.4E, Cellsignalling, 1:1000), p42/44 MAPK (Erk1/2) (137F5, Cellsignalling, 1:1000), p-Stat3 (Tyr705, Cellsignalling 1:1000), Stat3 (124H6, Cellsignalling, 1:1000) and GAPDH (MAB374 1:10,000, Millipore) were immunodetected by incubating for 16 h in primary antibody in blocking buffers. Membranes were washed extensively with PBST or TBST, and incubated with anti-mouse or anti-rabbit (1:1000–10′000, Jackson ImmunoResearch) secondary antibody in blocking buffer. After 1 h, membranes were washed as above and developed using enhanced chemiluminescence (ECL, Amersham) and a ChemiDoc MP imager (Biorad). Densitometric images were captured with ImageJ and band intensity normalized to GAPDH as a loading control.

Cell-free Thioflavin-T fibrillization assays were performed on an Infinite M2000 Pro plate reader (Tecan) as described (Truran 2015). Briefly, 10 μM monomeric Aβ40 or Aβ42 were incubated in a buffer consisting of 20 mM of Thioflavin-T in 150 mM NaCl and 5 μM of HEPES at pH 7.4, with and without 1 mg/mL of HDL, at 37 °C with 20 s of orbital shaking (3 mm amplitude) every 5 min in a black 96-well plate. Formation of fibrillar β-amyloid pleated sheets over time was monitored by excitation at 440 nm and measuring emission intensity at 490 nm every 5 min up to 12 h in total.

Frozen brain tissues (cortex Brodmann area 9 and cerebellum) were provided by the Harvard Brain Tissue Resource Center under an approved UBC protocol (C04–0595) and extracted with 8 volumes of ice-cold carbonate buffer (100 mM Na₂CO₃, 50 mM NaCl, pH 11.5) containing cOmplete protease inhibitor (Roche Applied Science) by manual homogenization with a tissue probe. After incubating on ice for 10 min, lysates were clarified by centrifugation at 16,600 rcf for 45 min at 4 °C. The supernatant was removed and neutralized by adding 1.5-volumes of 1 M Tris-HCl pH 6.8 to give a final pH of approximately 7.4. Brain tissues from all samples were extracted in an identical manner, and fractions were aliquoted and immediately frozen at −80 °C until analysis. Protein concentrations were determined using the Lowry Protein Assay (Biorad). ICAM-1 (ab174445, ABCAM), VCAM-1 (ab187393, ABCAM), Aβ40 (KHB3442, Life Tech) and Aβ42 (KHB3482, Life Tech) in carbonate extracts were quantified using commercial ELISAs. ELISA data were normalized to total protein concentration of the extract. For immunofluorescent staining, brains were sectioned at 20 μm, rehydrated in PBS and blocked in 4% paraformaldehyde (PFA). After 20 min, sections were washed once with Tris-HCl (0.5 mM, pH 8), twice with PBS, and blocked in blocking buffer (5% goat serum and 1% BSA). After 60 min, section were incubated at 4 °C with antibodies against CD31 (WM59 Biolegend, 1:50), ICAM-1 (EP1442Y ABCAM 1:200), and VCAM-1 (EPR5047 ABCAM 1:200) in blocking buffer overnight. Sections were washed 3 times in PBS and incubated at RT with Alexa-488 anti-mouse and Alexa-594 anti-rabbit fluorescently labeled secondary antibodies (LifeTechnologies, 1:600). After 45 min and 3 additional PBS washes, sections were mounted in Prolong Gold antifade reagent with DAPI (LifeTechnologies) and imaged on an inverted fluorescent microscope (Zeiss)

To mimic the complexity of native cell-cell and/or cell-matrix interactions observed in native human vessels, we used innovative tissue engineering technology to generate in vitro 3D human vessels composed of primary human umbilical vein ECs (HUVEC) and primary human smooth muscle cells (SMC) to maximize translational relevance of our studies relative to in vitro studies that solely use traditional static cell culture models. Our system uses a scaffold-directed dynamic, semi-pulsatile flow bioreactor system, on which primary human cells are sequentially seeded into 2 mm diameter biodegradable polyglycolic-acid (PGA)/polycaprolactone (PCL) composite matrices. After a short static incubation, vascular constructs are exposed to dynamic flow in a bioreactor, where nutrient medium is directed through the lumen of the bioreactor circulation loop to mimic native blood flow. We previously demonstrated that these scaffold-directed engineered vessels are useful for studies of endothelial activation in 3D culture [17]. A schematic of the bioreactor system is presented in Fig. 1a. Under our flow bioreactor conditions, haematoxylin and eosin staining confirmed a dense tissue formation composed of cells with extracellular matrix on the luminal side of the scaffold (Fig. 1b-d). Immunofluorescent staining confirmed a SMC phenotype of cells in the inner layers and an EC monolayer on the luminal side (Fig. 1c,d). Integrity of the endothelial barrier was functionally assessed by injecting Evans blue dye into the bioreactor circulation loop, which was excluded after EC seeding, demonstrating a functionally tight endothelial barrier (Fig. 1e).

Having established a 3D human vascular model, we injected 1 μM Aβ40 or Aβ42 directly into the graft chamber to mimic Aβ coming from the brain anteluminal side, followed by injection of human THP-1 monocytes into the lumen chamber to mimic circulating monocytes. Both Aβ40 and Aβ42 led to THP1 adhesion to endothelium in the engineered vessels, with the most robust response appearing 8 h after Aβ injection (Fig. 1f-h). We then tested the ability of HDL, isolated from healthy human donors using KBr density gradient ultracentrifugation, to suppress Aβ-induced monocyte adhesion to ECs in engineered vessels by circulating either media alone or 200 μg/mL of HDL for 2 h prior to injecting 1 μM Aβ into the graft chamber. In this engineered vascular model, HDL robustly suppressed monocyte adhesion to endothelium by both Aβ40 (5 fold, p = 0.0286) and Aβ42 (10 fold, p = 0.0016) (Fig. 1i-l).

To define the mechanisms by which HDL attenuates Aβ-induced monocyte binding to ECs, we used static monolayer EC cultures and peripheral blood mononuclear cells (PBMC) isolated by continuous density gradient ultracentrifugation from healthy human donors. We first confirmed that Aβ induces PBMC adhesion in monotypic HUVEC cultures and that HDL attenuates this activity (Fig. 2a,b). As Aβ originates within the brain and the primary pathways by which Aβ is cleared from the brain involve cerebral vessels [7], we also showed that the human brain microvascular endothelial cell line hCMEC/D3, a commonly used cell line for in vitro studies of the blood brain barrier (BBB) [20], also exhibit increased PBMC adhesion upon Aβ treatment and that this too is attenuated by HDL (Fig. 2c,d). As HUVEC and hCMEC/D3 give nearly identical results, we focused on hCMEC/D3 cells for all subsequent mechanistic experiments.

Our first question was to determine if lipid-free apoA-I was functionally equivalent to mature HDL in its ability to suppress Aβ-mediated PBMC binding to hCMEC/D3. Our results clearly showed that pre-incubation of hCMEC/D3 with lipid-free apoA-I (50 and 100 μg/mL) did not alter Aβ-induced PBMC adhesion (Fig. 2e,f), demonstrating that mature HDL particles are required to attenuate Aβ-induced PBMC adhesion to brain-derived ECs.

To establish the experimental conditions to investigate how HDL suppresses Aβ-mediated PBMC adhesion to hCMEC/D3, we performed PMBC adhesion assays in hCMEC/D3 after a 3 h exposure to increasing concentrations human Aβ40 and Aβ42 monomers. Compared to baseline, Aβ40 concentrations of 0.01, 0.1 and 1.0 μM led to significant 175%, 169% and 179% increases in PMBC adhesion (p = 0.033, 0.002 and 0.019 respectively, Additional file 1a), The same concentrations of Aβ42 also led to significant 184%, 161% and 223% increases in PMBC adhesion (p = 0.019, 0.0406 and 0.0421, respectively) relative to baseline (Additional file 1b). Having demonstrated that 0.1 μM of monomeric Aβ40 or Aβ42 is sufficient to induce PMBC adhesion in hCMEC/D3, we then tested whether pre-treating hCMEC/D3 for 2 h with 0 to 400 μg/mL HDL could attenuate Aβ-induced PBMC adhesion. Significant suppression of PBMC adhesion was observed at concentrations of HDL from 100 μg/mL and above. At 100 μg/mL HDL there was an 80% and 60% reduction of adhered PBMCs in cells induced by 0.1 μM Aβ40 (p = 0.017) or 0.1 μM Aβ42 (p = 0.018) respectively (Additional file 1c,d).

As Aβ has been reported to bind HDL-like particles in both cerebrospinal fluid (CSF) and blood [21], and apoA-I has been reported to affect Aβ aggregation [13, 22], we reasoned that one mechanism by which HDL may attenuate PBMC adhesion to hCMEC/D3 could be by affecting Aβ structure to prevent the formation of toxic higher order species [23]. A cell-free Thioflavin-T reporter assay was used to determine whether HDL affects the fibrillization kinetics of Aβ40 or Aβ42 over a 12 h period, using 10 μM Aβ and 10 mg/mL HDL (ratio: 1:1), which is the same Aβ:HDL ratio used for our cellular PBMC adhesion assays and allows detection of fluorescence signal using a plate reader (Fig. 2c,d). Under these conditions, in the absence of HDL, the onset of Aβ fibrillization requires at least 3 h for Aβ42 and at least 10 h for Aβ40. Notably, in the presence of HDL, we observed delayed onset of fibrillization for both Aβ42 and Aβ40, and a slower rate of Aβ42 fibrillization (Fig. 3a). As HDL can suppress PBMC adhesion to hCMEC/D3 induced by both Aβ40 and Aβ42 within 3 h of Aβ addition, we conclude that suppression of Aβ fibrillization cannot be the only mechanism by which HDL functions to protect ECs from the detrimental effects of Aβ.

Several experiments were then performed to further test whether Aβ structure affects HDL’s anti-inflammatory activity on hCMEC/D3. We varied the structural species of the Aβ preparation added to hCMEC/D3, reasoning that if HDL acts primarily through maintaining Aβ in a soluble monomeric state, the protective effect of HDL should be diminished if ECs are stimulated with preformed Aβ oligomers or fibrils. Soluble monomeric and oligomeric, and insoluble fibrillar Aβ were prepared as described [24] and structures were confirmed using a dot blot assay as well as electron microscopy (Fig. 3b,c). Interestingly, monomeric and oligomeric Aβ40 and Aβ42 preparations induced comparably robust PBMC adherence, whereas the insoluble fibrillary form failed to significantly increase PBMC adhesion. Importantly, the ability of HDL to suppress PMBC adhesion was unaffected by input of either Aβ monomers or oligomers (Fig. 3d-e). These results suggest that delayed Aβ fibrillization may not be the major pathway by which HDL suppresses PBMC adhesion to ECs under our experimental conditions as HDL maintained its protective effect when ECs are treated with pre-formed oligomers. To confirm this, we also varied the timing of HDL and Aβ addition to hCMEC/D3 such that, in addition to a 2 h HDL pre-incubation as above, HDL was added at the same time as Aβ (co-incubation) or 1 h after Aβ addition (post-incubation). We reasoned that if HDL acts primarily through maintaining Aβ solubility, pre-incubation and co-incubation designs should give equivalent and maximal suppression of Aβ-induced PBMC adhesion as, in both scenarios, Aβ is always in the presence of HDL and aggregation could be delayed. In contrast, as the post-incubation scenario would allow Aβ aggregation to begin prior to addition of HDL, the protective effect of HDL should be reduced if added after the Aβ stimulus if Aβ structure is the key driver of PBMC adhesion to ECs. We observed that PMBC adhesion was significantly and similarly reduced regardless of when HDL was added to hCMEC/D3 relative to Aβ40 or Aβ42 providing additional support that HDL’s ability to attenuate Aβ-mediated PBMC adhesion to hCMEC/D3 is independent of whether Aβ input is monomeric or oligomeric (Fig. 3f,g).

These two lines of evidence suggest that, although HDL can affect Aβ structure in cell-free conditions, these effects are neither rapid nor robust enough to fully explain how HDL suppresses Aβ mediated PBMC adhesion to hCMEC/D3. We next evaluated signalling pathways implicated in HDL’s anti-inflammatory effects on ECs.

Stimulation of NO synthesis by HDL is reported to reduce arterial EC activation [25]. We therefore tested whether HDL induces NO production in hCMEC/D3 using the cell permeable indicator 4,5-Diaminofluorescein diacetate (DAF2), which reacts with NO to produce highly fluorescent trizaolo-fluorescein. Compared to baseline conditions, HDL-treated hCMEC/D3 showed a significant 130% increase (p = 0.01) in NO production (Fig. 4a), as well as significantly elevated phosphorylated endothelial NO synthase (eNOS) on serine 1177 (161%, p = 0.047) (Fig. 4b,c). These results confirm that HDL can induce NO production in hCMEC/D3, as observed in ECs derived from other origins.

To test whether NO generation is required for HDL to suppress Aβ-mediated PBMC adhesion to hCMEC/D3, we treated cells with the eNOS inhibitor L-NAME, followed by the addition of HDL, and finally stimulation with either Aβ40, Aβ42 or TNF-α. We observed that blocking eNOS with L-NAME had no effect on the ability of HDL to suppress either Aβ- or TNF-α-mediated PBMC adhesion to hCMEC/D3 (Fig. 4d-f). Additional experiments confirmed the potency of L-NAME to block eNOS in HDL-stimulated HUVEC (Additional file 2a), yet this also had no effect on the ability of HDL to suppress either Aβ- or TNF-α- mediated PBMC binding in HUVEC (Additional file 2b–d). We further ruled out a role for the NO pathway in hCMEC/D3 by measuring the effect of HDL in the presence of VPC23019, an antagonist of sphingosine-1 phosphate receptor 1 (S1P1) and S1P3, receptors that are required for eNOS activity [26], and again found no effect on the ability of HDL to suppress either Aβ- or TNF-α-mediated PBMC adhesion to hCMEC/D3 (Fig. 4g-i) or HUVEC (Additional file 2e–h). Taken together, these results provide strong support that the eNOS pathway is not involved in the ability of HDL to suppress PBMC adhesion to Aβ- or TNF-α-activated hCMEC/D3 and HUVEC.

In contrast to a recent report suggesting that HDL up-regulates Anx1, which affects HDL’s ability to suppress ECs activation [27], we observed no significant change in Anx1 protein levels in hCMEC/D3 after either HDL or Aβ treatment (Additional file 2i).

In addition to numerous proteins and lipid species, HDL particles carry microRNAs (miRNAs) that can correlate with vascular disease risk [28]. One of the most abundant miRNAs in HDL is miR-233, which has recently been shown to affect gene expression in human coronary artery ECs including suppressing ICAM-1 expression [29]. Three lines of evidence suggest that HDL-mediated suppression of Aβ-mediated PBMC binding to hCMEC/D3 is independent of miR-233. First, direct transfection of a miR-223 mimetic did not prevent Aβ40 or TNF-α-mediated PBMC adhesion to hCMEC/D3, although a statistically significant reduction with Aβ42 was observed (Additional file 3a–c). Second, priming hCMEC/D3 with HDL in the presence of a specific miR-223 inhibitor did not diminish the ability of HDL to suppress PMBC adhesion by Aβ40 , Aβ42 or TNF-α (Additional file 3d–f). Third, in the 5 h period where hCMEC/D3 were exposed to HDL (2 h pre-treatment then 3 h with Aβ), no significant transfer of miR-223 was detected (see Additional file 3g), consistent with a previous report that after 6 h, miR-223 was not transferred from HDL to several types of ECs including HUVEC [30]. These results suggest that, although the miR-233 mimetic can suppress Aβ42-mediated PBMC binding to hCMEC/D3 when added directly, miR-233 is unlikely to underlie the ability of HDL to suppress PBMC adhesion to hCMEC/D3 triggered by Aβ40, Aβ42 or TNF-α.

ICAM-1 and VCAM-1 are canonical markers of EC activation that mediate PBMC adhesion by classical inflammatory stimuli [31]. Several studies have reported elevated microvascular inflammatory markers in the AD brain [32, 33]. Using ELISA, we confirmed that protein levels of intercellular adhesion molecule 1 (ICAM-1) but not vascular cell adhesion molecule 1 (VCAM-1) are increased in the cortex of AD patients compared to either cerebellum from the same individual or the cortex from non-cognitively impaired (NCI) control subjects (Additional file 4). Increased ICAM-1 levels were further confirmed using immunofluorescence revealing both vascular and glial ICAM-1 reactivity whereas VCAM-1 remained unchanged (Additional file 5a–b). As expected, soluble Aβ40 and Aβ42 levels were elevated in the cortex of AD patients compared to their corresponding cerebellum or to the cortex of NCI subjects (Additional file 4). Because ICAM-1 is a classical marker of EC activation and HDL can attenuate PBMC adherence by reducing ICAM-1 levels in aortic derived ECs [34], we also evaluated the roles of ICAM-1 and VCAM-1 in hCMEC/D3 activation. Using Western blotting, we observed that pretreatment of monotypic hCMEC/D3 cultures with HDL significantly suppressed TNF-α-mediated induction of ICAM-1 (from 1343% to 920%, p = 0.041) and VCAM-1 (from 2838% to 1928% p = 0.012), relative to untreated cells (Additional file 6a–c), results that are consistent with HDL’s anti-inflammatory effects on peripherally derived ECs. By contrast, Aβ treatment did not affect either mRNA levels or total protein levels of VCAM-1 and ICAM-1 in hCMEC/D3 (Additional file 6d–m). Although total ICAM-1 levels showed no significant change in response to either Aβ40 or Aβ42 treatment, biotinylation assays revealed significantly increased cell surface ICAM-1 levels, but not VCAM-1 levels, in response to 0.1 μM of Aβ40 (p = 0.009) and a strong trend in response to Aβ42 (p = 0.06) (Additional file 6 h–j). Unexpectedly, however, we observed that cell surface ICAM-1 levels were not reduced by a 2 h exposure to HDL (Additional file 6 k–m), showing that cell surface ICAM-1 levels do not correlate with PMBC adhesion in Aβ-activated hCMEC/D3. We next analysed the effects of Aβ on the phosphorylation of multifunctional serine/threonine protein kinases that are involved in EC activation, survival, apoptosis, proliferation and migration. TNF-α activated NFκB p-P65 (p = 0.002), SAPK/JNK (p = 0.005), MAPK/ERK (p = 0.01), but not Akt or STAT2 pathways, whereas Aβ did not alter any of these pathways in hCMEC/D3 (Additional file 7). Taken together, these results suggest that Aβ does not activate any of classical inflammatory pathways known to increase PBMC adhesion through ICAM-1 or VCAM-1.

As our results show that Aβ activates hCMEC/D3 through a mechanism that is independent of canonical EC activation intracellular signalling pathways, we hypothesised that binding or uptake of Aβ to ECs might influence PBMC adhesion. To test this hypothesis, we first used temperature modulation experiments to investigate how HDL affects the interactions of Aβ with hCMEC/D3 and observed that HDL pre-treatment significantly reduced total association measured at 37 °C (Fig. 5a,b), cell surface binding measured at 4 °C measured by ELISA (Fig. 5c,d) and intracellular uptake of fluorescently labeled Aβ40 and Aβ42 (Fig. 5e). Second, we tested whether blocking RAGE and LRP1, known receptors that modulate Aβ binding and uptake in ECs [35], might reduce PBMC adhesion. Blocking RAGE with a specific antibody abrogated PBMC adhesion to both Aβ40 and Aβ42 but not TNF-α stimulated hCMEC/D3(Fig. 5f-h). Similarly, blocking LRP1 with receptor associated protein (RAP) abolished Aβ40 and Aβ42 but not TNF-α effects (Fig. 5i-k). Finally, as heparin sulphate proteoglycans (HSPG) are also involved in binding and uptake of Aβ in neuron [36], we then saturated or removed HSPG by treating hCMEC/D3 with heparin or heparinase III, respectively. Both treatments abolished the ability of Aβ40 and Aβ42 to induce PBMC binding whereas TNF-α induced-PBMC adhesion remained significant (Fig. 5l-n). Together, these observations show that, unlike TNF-α, Aβ induces PBMC adhesion to ECs through a pathway that requires interactions with or internalization through Aβ receptors on the EC surface. We also observed that HDL does not decrease the expression of RAGE or LRP1 (Additional file 8), suggesting that down-regulation of these receptors cannot explain the protective effect of HDL on Aβ-induced PBMC binding to hCMEC/D3.

Scavenger receptor BI (SR-BI) is the principal HDL receptor on ECs and activates several HDL signalling pathways in addition to mediating selective cholesterol uptake upon HDL docking [37]. We observed that inhibiting SR-BI binding with a specific blocking antibody abolished the ability of HDL to suppress both Aβ- and TNF-α-mediated PBMC adhesion to hCMEC/D3 (Fig. 6a-c). Selectivity to SR-BI was confirmed by demonstrating that blocking scavenger receptor CD36 with a specific antibody had no effect on the ability of HDL to suppress either Aβ- or TNF-α-mediated PBMC adhesion to hCMEC/D3 (Fig. 6d-f). Intriguingly, inhibiting SR-BI-mediated selective cholesterol uptake with block lipid transfer 1 (BLT1) did not affect the ability of HDL to suppress either Aβ- or TNF-α-mediated PBMC adhesion to hCMEC/D3 (Fig. 6g-i). We further observed that inhibiting SR-BI with a specific blocking antibody abolished the ability of HDL to suppress Aβ uptake in hCMEC/D3 (Fig. 6j). These results demonstrate that HDL requires SR-BI-mediated signalling to inhibit Aβ uptake and subsequently suppress PBMC adhesion to hCMEC/D3 in response to Aβ- or TNF-α in a manner independent of lipid uptake.

Having used monotypic static cell cultures to demonstrate that HDL suppresses Aβ-induced PBMC adhesion to endothelial cells through a mechanism that requires SR-BI and is independent of ICAM-1 expression, we confirmed that this mechanism also explains the results in our 3D bioengineered vessels. We observed that inhibiting SR-BI with a specific antibody circulated through the vessel lumen abolished the ability of HDL to suppress Aβ-mediated monocyte adhesion to the vascular endothelium (Fig. 7a-b) without reducing ICAM-1 expression (Fig. 7c-d).