Blood-brain barrier-associated pericytes internalize and clear aggregated amyloid-β42 by LRP1-dependent apolipoprotein E isoform-specific mechanism

Mice were housed in plastic cages on a 12 h light cycle with ad libitum access to water and a standard laboratory diet. All procedures were approved by the Institutional Animal Care and Use Committee at the University of Southern California with National Institutes of Health guidelines. We studied APP^Sw/0 mice expressing human APP transgene with the K670 M/N671 L (Swedish) double-mutation under control of the hamster prion promoter [43]. Lrp^lox/lox mice [44, 45] were crossed with Cspg4-Cre mice [46] to generate Lrp^lox/lox;Cspg4-Cre mice with deletions of Lrp1 gene from pericytes and oligodendrocyte progenitor cells. To minimize confounding effects of background heterogeneity all experiments were performed using age-matched littermates. Both, male and female mice were used in the study. All animals were randomized for their genotype information. All experiments were blinded; the operators responsible for experimental procedure and data analysis were blinded and unaware of group allocation throughout the experiments.

Written consent was obtained and approved by the University of Rochester Medical Center for all studied human subjects prior to death. The postmortem interval ranged between 4 and 16 h. Postmortem brain tissue samples including frontal cortex (Brodmann area 9/10) were obtained from subjects with a definite diagnosis of AD confirmed by neuropathological analysis including Braak stages ≥ III; CERAD (Consortium to Establish a Registry for Alzheimer’s Disease) frequent, and neurologically intact controls with no AD pathology (Braak stages ≤ III; CERAD – negative). Six controls and six AD samples were used for the current study. Their demographic and clinical features were as we reported previously [47], and shown in Additional file 1: Table S1. Vascular risk factors such as atherosclerosis, hypertension and/or myocardial infarction were present in 4 out 6 AD patients and 6 out of 6 controls. The cause of death in all AD and control patients was either respiratory failure or cardiac failure. All tissues were paraformaldehyde (PFA)-fixed, paraffin-embedded and cut to 10 μm thick slices. Sections were deparaffinized with xylene and rehydrated to distilled water after serial ethanol washes. Subsequently, heat-induced antigen retrieval (HIAR) was performed following Dako’s protocol. The tissue sections were then blocked in 5% donkey serum (Jackson ImmunoResearch, West Grove, PA, USA) containing 0.3% Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA), and then incubated with the following primary antibodies overnight at 4 °C: goat anti-human PDGFRβ (1:100, R&D Systems, Minneapolis, MN, USA), mouse anti-human aminopeptidase N (CD13) (1:100, R&D Systems), rabbit anti-human Aβ (1:100, Cell signaling, Boston, MA, USA). Blood vessels were stained by DyLight 488 Labeled Lycopersicon esculentum (Tomato) Lectin (1:100, Vector Laboratories, Burlingame, CA, USA, # DL-1174) for 1 h at room temperature. Species-specific fluorochrome-conjugated secondary antibodies were incubated for 1 h at room temperature, including Alexa 568-conjugated donkey anti-goat (1:200, Invitrogen), Alexa 568-conjugated donkey anti-mouse (1:200, Invitrogen) and Alexa 647-conjugated donkey anti-rabbit (1:200, Invitrogen). Tissue sections were mounted and coverslipped using fluorescent mounting media (Dako, Carpinteria, CA, USA). All slices were scanned using Zeiss 510 confocal microscopy with Zeiss Apochromat water immersion objectives (Carl Zeiss MicroImaging Inc., Thornwood, NY, USA). All slices were scanned using Zeiss 510 confocal microscopy with Zeiss Apochromat water immersion objectives (Carl Zeiss MicroImaging Inc., Thornwood, NY, USA), with lasers and band-pass filter settings as the following: a 488-nm argon laser to excite Alexa Fluor and Dylight 488, and the emission was collected through a 500–550-nm bp filter; a 543 HeNe laser to excite Alexa Fluor 568 and Cy3 and the emission was collected through a 560–615-nm bp filter; a 633 HeNe laser to excite Alexa fluor 647 and the emission was collected through a 650–700-nm bp filter.

Brain slices were prepared from 2-month old C57BL6 mice or Lrp^lox/lox;Cspg4-Cre mice. Following the urethane anesthesia, brains were quickly removed from the cranial cavity and sectioned with McIlwain tissue chopper (Ted Pella, Inc., Redding, CA, USA) into 200 μm thick slices containing cortex and hippocampus, as previously described [48]. Slices were recovered for 30 min in a submersion chamber filled with the per-warmed (at 37 °C) artificial cerebrospinal fluid (aCSF; 126 mM NaCl, 2.5 mM KCl, 1.25 mM Na₂PO₄, 26 mM NaHCO₃, 1 mM MgCl₂, 2 mM CaCl₂, 0.5 mM ascorbic acid, 2 mM sodium pyrurate, and 10 mM glucose, saturated with 95% O₂ and 5% CO₂). Prior to incubation with aCSF containing 5 μg/ml Cy3-Aβ42 for 30 min, slices were pre-incubated with either non-immune IgG (NI-IgG) or LRP1-specific blocking antibody [22, 29] (anti-LRP1 N20, 50 μg/ml; Santa Cruz Biotech., Santa Cruz, CA, USA) for 20 min at 37 °C. At the end of the experiment, brain slices were washed with phosphate buffer saline (PBS) and fixed with 4% PFA for 30 min. The same immunostaining procedure as for human brain tissue without antigen retrieval was performed on mouse brain slices. Primary antibodies included goat anti-mouse aminopeptidase N (CD13) (1:100, R&D Systems), and goat anti-mouse PDGFRβ (1:100, R&D Systems). Secondary antibodies were: Alexa 488-conjugated donkey anti-goat (1:200, Invitrogen). Images were obtained with Zeiss 510 confocal microscopy (Carl Zeiss MicroImaging Inc.) with lasers and band-pass filter settings as described above. The distribution of Cy3-Aβ42 in pericytes and capillary endothelium was analyzed with the NIH Image J software [49].

Brain capillaries were isolated from mouse cortices pooled from 3 to 4 mouse brains for each biological replicate. In experiments in Figs. 4 and 5, we used 3–4 independent biological replicates (cultures), as indicated in the legends to these respective figures. Capillaries were isolated as we have described previously [49, 50]. In brief, mouse cortices were macroscopically dissected, and all visible white matter was discarded in ice-cold PBS containing 2% fetal bovine serum (FBS). The brain was homogenized in PBS containing 2% FBS and centrifuged at 6,000 g for 15 min after addition of Dextran (70 kDa, Sigma). The capillary pellet was collected and sequentially filtered through a 45 μm cell strainer (BD Falcon). The remaining pellet on top of the 45 μm cell strainer was collected in PBS and digested for 12 h at 37 °C with collagenase A (Roche, 10103586001), as we previously described. The cells were washed with PBS and then plated in complete medium containing DMEM, 10% FBS, 1% non-essential amino acids, 1% vitamins and 1% antibiotic/antimycotic on plastic (non-coated) tissue culture plates. After 6 to 12 h the non-adherent cells were washed away and fresh medium was replaced every 2–3 days. Cultures were confirmed to be morphological consistent with pericyte cultures and PDGFRβ-positive, SMA-positive, Desmin-positive, GFAP-negative, AQP4-negative, MAP2-negative, NeuN-negative, VWF-negative, and Iba1-negative, as previously described [37, 49]. Transfections were performed with Neon transfection system (Invitrogen, Grand Island, NY, USA) following the manufacturer’s protocol. After optimization, the transfection efficiency with pEGFP plasmid was > 80%. For siRNA transfection, 15 pmol of siRNA in 1 μl was electroporated into 1 × 10⁵ cells in a total volume of 10 μl.

The multi-spot glass slides (Thermo Scientific, 9991090, Waltham, MA, USA) were coated with Cy3- Aβ42 [29, 51] at 1 μg per spot without cells or with brain capillary pericytes treated with non-immune IgG (NI-IgG) or LRP1-specific blocking antibody [22, 29] (anti-LRP1; 50 μg/ml) or) or apoE-specific blocking antibody (anti-apoE, HJ6.3, 50 μg/ml), scrambled siRNA (si.Control) or LRP1 siRNA (si.Lrp1) (Dharmacon, E-040764-00-0010) or apoE siRNA (si.Apoe) (Dharmacon, E-040885-00-0005), and 500 nM receptor-associated protein (RAP) [52] (EMD Bioscience), lipidated apoE3 (40 nM) or apoE4 (40 nM) [49]. Cells (5,000 per spot) were incubated for 5 days in DMEM containing 10% heat inactivated FBS, penicillin and streptomycin (Invitrogen). Cells were labeled with Cell Tracker Green CMFDA (Invitrogen C7025), and then fixed with 4% paraformaldehyde. Slides were scanned using Zeiss 510 confocal microscopy (Carl Zeiss MicroImaging Inc.). The Cy3-Aβ42 relative intensity was analyzed with the NIH Image J software.

Lipidated apoE isoform particles were purified from culture media of human apoE3 or apoE4 overexpressing immortalized astrocytes using an affinity column, as we described previously [53]. Briefly, astrocytes were cultured in advanced DMEM (Invitrogen) with 10% FBS. After 90–95% confluency, cells were washed by PBS and further incubated in advanced DMEM with N-2 Supplement (Invitrogen) and 3 mM 25-hydroxycholesterol (Sigma) for 3 days. Collected culture media were applied onto mouse monoclonal antibody against a human apoE (WU E-4) column. Lipidated apoE particles were eluted from the column with 3 M sodium thiocyanate, concentrated using Apollo centrifugal quantitative concentrators (QMWL: 150 kDa, Orbital Biosciences), and dialyzed against PBS.

Cell samples transfected with scrambled siRNA or LRP1 siRNA were washed in cold PBS and lysed with RIPA buffer. Proteins were quantified with BCA protein assay kit (Pierce). An equal amount of protein sample was loaded for SDS-PAGE. We used the following primary antibodies: rabbit anti-mouse LRP1 (1: 20,000, Abcam, Cambridge, MA, USA), β-actin (1:10,000, Sigma), and Aβ antibody (6E10, 1:1000, Covance). Images were scanned using ChemiDoc™ MP imager with LEDs and quantified using Image Lab™ software (Bio-Rad, Hercules, CA, USA).

Paraformaldehyde-fixed, paraffin embedded brain tissue sections from APP^Sw/0 mice were sectioned at a thickness of 10 μm. Immunofluorescent detection of pericytes (CD13-positive), Aβ and endothelial-specific lectin fluorescence was conducted as described above. The DeadEnd Fluorometric TUNEL system (Promega) was then completed as described by the manufacturer.

All quantified data represent as mean ± s.e.m. Student’s t-test or one-way analysis of variance (ANOVA) followed by Bonferroni post-hoc test was used to determine statistically significant differences. A P value < 0.05 was considered statistically significant.

Using 3-channel confocal microscopy, here we show that Aβ accumulates in > 30% of CD13+ and/or PDGFRβ+ pericytes on lectin+ brain endothelial capillary profiles (< 6 μm in diameter) in brain cortical sections from AD patients compared to barely detectable Aβ accumulation in pericytes in age-matched controls (see demographic Additional file 1: Table S1; Fig. 1a-f). To confirm this observation, we immunostained cortical sections of 8.5-month-old APP^Sw/0 mice for human Aβ, at a disease stage when these mice begin depositing Aβ [21]. We found that nearly 60% of CD13+ pericytes were positive for human Aβ (Fig. 1g-h). Collectively, these findings indicate that pericytes in AD and APP^Sw/0 mouse brains both accumulate Aβ. Additionally, we found that Aβ also accumulates in brain endothelium of AD patients (Additional file 1: Figure S1A, B), and in lectin+ brain endothelial cells of 8.5-month-old APP^Sw/0 mice (Additional file 1:Figure S1C).

Cultured mouse pericytes [37] and human brain pericytes [54] die when treated with excess Aβ. To verify this result in APP^Sw/0 mice in vivo, we performed quadruple TUNEL, CD13, lectin and human Aβ staining, which revealed a significant increase in TUNEL+ CD13+ pericytes in 8.5-month old, but not 3-month old, APP^Sw/0 mice compared to 8.5-month old littermate controls (Fig. 1i-j), consistent with findings that 3-month old APP^Sw/0 mice show barely detectable Aβ deposition in the brain compared to 8.5-month old mice, which begin depositing Aβ at that stage [55].

To confirm Aβ uptake by pericytes, we next conducted Aβ uptake experiments on freshly isolated acute cortical slice cultures, as previously reported [48]. In this experiment, the mouse brains slices were left to equilibrate in aCSF for 20 min at 37 °C with aeration, that was followed by incubation with Cy3-Aβ42 (1 μM) for 30 min or 2 h. Confocal microscopy was performed to analyze Aβ internalization by pericytes (Fig. 2a). We found that uptake of Cy3-Aβ42 by pericytes was rapid and time-dependent. Compared to 30 min, at 2 h the amount of internalized Cy3-Aβ42 in CD13+ pericytes was further significantly increased (Fig. 2b-c). In these studies, Aβ42 preparation contained a mixture of Aβ42 monomers, dimers, trimers, tetramers, and small molecular weight oligomers (Additional file 1: Figure S2). Both low (Fig. 2b) and high (Fig. 2d) magnification imaging analysis showed strong Cy3-Aβ42 fluorescence signal in CD13+ pericytes and/or PDGFRβ+ pericytes (Additional file 1: Figure S3A-B). In addition to pericytes, Cy3-Aβ42 uptake was observed by endothelial cells, astrocytes and microglia (Additional file 1: Figure S3C-E). The intracellular Cy3-Aβ42 levels in pericytes were significantly reduced in the presence of an anti-LRP1-specific antibody [22, 29] and the receptor associated protein (RAP) [52], an LRP1 antagonist which inhibits ligand binding to LRP1 (Fig. 2d), as shown by 17.1% and 21.7% of the Cy3-Aβ42+ area occupying CD13+ pericyte profiles, respectively, compared to > 80% in control slices treated with vehicle or non-immune immunoglobulin G (IgG) (Fig. 2d-e).

To further investigate the role of LRP1 in Cy3-Aβ42 uptake by pericytes, we generated conditional knockout mice with LRP1 deficiency in pericytes (Lrp1^lox/lox; Cspg4-Cre) by crossing Lrp1^lox/lox mice [44, 45] with Cspg4-Cre mice [46] (Fig. 3a). As expected, CD13+ pericytes on isolated brain microvessels from Lrp1^lox/lox; Cspg4-Cre mice did not show a detectable LRP1 immunoreactivity in contrast to pericytes from control Lrp1^lox/lox mice, whereas LRP1 remained expressed in lectin+ endothelium in both Lrp1^lox/lox; Cspg4-Cre and Lrp1^lox/lox microvessels (Fig. 3b), confirming selective deletion of LRP1 from brain capillary pericytes. Both, Lrp1^lox/lox and Lrp1^lox/lox; Cspg4-Cre mice had comparable CD13+ pericyte coverage (Additional file 1: Figure S4). Next, we studied Cy3-Aβ42 uptake on brain slices from these mice, as described above. Two hours after incubation, Cy3-Aβ42 uptake by CD13+ pericytes was reduced by > 80% in Lrp1 ^lox/lox; Cspg4-Cre mice compared to control Lrp1 ^lox/lox mice (Fig. 3c-e). As expected, Aβ uptake by other cell types shown collectively as CD13- cells was not affected in this Lrp1 conditional knockout model (Fig. 3f). In summary, these data demonstrate LRP1-depedent uptake of Aβ42 species by pericytes.

To determine whether pericytes can clear aggregated Aβ, we used a slightly modified Aβ clearance model with pericytes seeded on multi-spot glass slides pre-coated with aggregated Cy3-Aβ42, as we reported previously in studies with primary VSMCs [29] (Fig. 4a). Five days after culture, the majority of Aβ42 aggregates (> 70%) were removed from the multi-spot surface by pericytes in the presence of vehicle, control non-immune IgG or scrambled short interfering si.Control (Fig. 4b-c), when compared to cell-free control (Fig. 4a). In contrast, only 22.8%, 27.3% and 12.9% of Cy3-Aβ42 was cleared in the presence of anti-LRP1 [22, 29], RAP [52] and si.Lrp1 (Fig. 4c), respectively. In silencing experiment, si.Lrp1 efficiently downregulated LRP1 in pericytes by ~ 90% when compared to si.Control (Additional file 1: Figure S5A). We also found a time-dependent increase in cell death of pericytes cultured on Aβ pre-coated slides, from 8.2% at 3 DIV to 18.7% at 5 DIV, whereas inhibiting LRP1 by an anti-LRP1 antibody or LRP1 silencing (si.Lrp1), not only significantly reduced Aβ uptake, but also diminished pericyte cell death by approximately 3 and 4-fold at 3 and 5 DIV, respectively (Fig. 4d), suggesting that reducing Aβ uptake decreases Aβ toxicity consistent with previous studies with soluble Aβ [37] or Dutch Aβ peptides [54].

To check whether apoE is required for LRP1-mediated clearance of Aβ aggregates, we studied clearance by pericytes in the presence of an apoE specific blocking antibody compared to non-immune IgG, and after silencing mouse apoE (si.Apoe) compared to scrambled si.Control (Additional file 1: Figure S5B; Fig. 5a). After 5 days, we found substantially reduced Aβ clearance with either pharmacologic or genetic inhibition of murine apoE in pericytes as illustrated by representative confocal microscopy images (Fig. 5a), quantification of time-dependent Cy3-Aβ42 clearance by mouse pericytes cultured for 1, 3 and 5 days after silencing mouse endogenous apoE (si.Apoe) compared to si.Control (Fig. 5b), and quantification analysis of Cy3-Aβ42 relative signal intensity on multi-spot slides after 5 days of culture with pericytes under different experimental conditions (Fig. 5c). These data suggest that pericyte-derived apoE is required for clearance of Aβ aggregates by cultured pericytes, which we show is mediated by LRP1, an apoE receptor (see Fig. 2d-e; Fig. 3c-e and Fig. 4b-c). Besides the self-autonomous effect of apoE in regulating clearance of Aβ aggregates by pericytes (Fig. 5a-c), to mimic in vivo situation we next studied the non-autonomous effects of astrocyte-derived apoE by adding lipidated human apoE3 or apoE4 particles prepared from immortalized astrocytes, as previously described [53]. Previous work has shown that astrocyte-derived apoE provides a major source of apoE in the brain in vivo and signals pericytes via LRP1, but not LRP2, very low density lipoprotein receptor (VLDLR), low density lipoprotein receptor (LDLR) or apoER2 [49]. The effects of apoE3 and apoE4 on Aβ clearance by pericytes was studied after silencing mouse apoE (si.Apoe). ApoE3, but not apoE4, almost completely reversed Aβ clearance by pericytes with inhibited mouse apoE (Fig. 5a-c), suggesting that astrocyte-derived apoE exerts a non-autonomous isoform-specific effect on Aβ clearance by pericytes. To additionally confirm that apoE3 effect on pericytes requires LRP1 as previously shown [49], we performed the same experiment in the presence of an anti-LRP1 antibody. As expected, this experiment showed that anti-LRP1 inhibits apoE3’s ability to reverse the Aβ clearing capability of mouse pericytes with silenced mouse apoE (si.Apoe) (Fig. 5a, c).