CD36 mediates albumin transcytosis by dermal but not lung microvascular endothelial cells: role in fatty acid delivery

Abstract

In healthy blood vessels, albumin crosses the endothelium to leave the circulation by transcytosis. However, little is known about the regulation of albumin transcytosis or how it differs in different tissues; its physiological purpose is also unclear. Using total internal reflection fluorescence microscopy, we quantified transcytosis of albumin across primary human microvascular endothelial cells from both lung and skin. We then validated our in vitro findings using a tissue-specific knockout mouse model. We observed that albumin transcytosis was saturable in the skin but not the lung microvascular endothelial cells, implicating a receptor-mediated process. We identified the scavenger receptor CD36 as being both necessary and sufficient for albumin transcytosis across dermal microvascular endothelium, in contrast to the lung where macropinocytosis dominated. Mutations in the apical helical bundle of CD36 prevented albumin internalization by cells. Mice deficient in CD36 specifically in endothelial cells exhibited lower basal permeability to albumin and less basal tissue edema in the skin but not in the lung. Finally, these mice also exhibited a smaller subcutaneous fat layer despite having identical total body weights and circulating fatty acid levels as wild-type animals. In conclusion, CD36 mediates albumin transcytosis in the skin but not the lung. Albumin transcytosis may serve to regulate fatty acid delivery from the circulation to tissues.

INTRODUCTION

Albumin is the most abundant protein in the circulation and is also found in large amounts in lymph. Its molecular radius (18) precludes its passage between adjacent cells of an intact endothelial monolayer, which limits paracellular diffusion to molecules less than 3–5 nm in size (12, 34). However, the detection of albumin in interstitial and lymphatic fluid (5, 43)—to as much as 40–60% of the level in plasma—indicates that the protein is capable of leaving the lumen of the microvasculature, even in the absence of inflammation. Pioneering work by Palade and colleagues (35) established that albumin injected into the circulation of animals was later detected in intracellular capillary endothelial vesicles; in some cases, these vesicles appeared to be releasing the albumin into the interstitium although it was never observed crossing interendothelial junctions. This vesicular transport process, termed transcytosis, is mediated by caveolae; importantly, knockdown or deficiency of caveolin-1, which ablates caveolae, prevents albumin transcytosis (36, 46). Thus, under physiological conditions, transcytosis of albumin is thought to be the dominant route for albumin transport out of the circulation (35).

To date, however, little is known about its regulation; this is attributable in part to technical difficulties in its study. In particular, when measuring endothelial permeability by cultured cells it has been challenging to distinguish the contribution of paracellular leakage from bona fide transcytosis (3). Most work has focused on the lung endothelium, which has been reported to bind to albumin via the glycoprotein 60 (gp60) receptor in caveolae (45, 50); however, this low-affinity binding contributes only modestly to transcytosis in the lung, which is instead dominated by fluid-phase uptake (25). Despite its description ~30 years ago, no molecular reagents for gp60 appear to be commercially available, and it is possible that alternative receptors for albumin may exist on the endothelium, particularly outside of the lung. In fact, essentially nothing is known about albumin transcytosis in nonpulmonary tissues, including the skin. Given the marked heterogeneity of endothelial cells supplying different organs (1, 2), it is plausible that the mechanisms of albumin transcytosis are distinct between different tissues. There is now growing interest in delineating the mechanisms of albumin kinetics in the circulation given its potential utility as a drug carrier (30).

Another reason for the paucity of research on albumin transcytosis by endothelial cells is likely because of uncertainty as to its physiological importance; its purpose remains obscure. Although mice deficient in caveolin-1 possess no caveolae and exhibit decreased endothelial internalization of albumin, these animals also display compensatory endothelial paracellular leakage (47). Thus, it is not possible using this model to draw conclusions about the physiological function of albumin egress from the microcirculation by transcytosis. Given its numerous binding sites for fatty acids (10), one possibility is that albumin transcytosis is important for the regulated traffic of circulating fatty acids to downstream tissues. It is noteworthy that deficiency of albumin is associated with elevated circulating cholesterol and phospholipid levels (8, 15).

To elucidate the mechanisms of endothelial transcytosis, we recently devised a single-cell assay that uses total internal reflection fluorescence (TIRF) microscopy and automated image analysis using a MATLAB script (4, 21, 27, 29). TIRF uses an evanescent wave generated by total internal reflection to illuminate just the proximal 100 nm or so of the cell; thus, we can selectively image the basal membrane of a live endothelial cell with minimal confounding from the overlying cytoplasm and apical surface. Briefly, confluent endothelial monolayers are exposed to a fluorophore-tagged ligand at the apical cell surface while the basal membrane of the cell is imaged by TIRF. Cytoplasmic vesicles undergoing exocytosis with the basal membrane are directly visualized and quantified. This approach is not affected by paracellular gaps between endothelial cells in the monolayer (4) and is quantitative, thereby being well suited to mechanistic studies. Using this and complementary in vivo methods, we now systematically compare the transcytosis of albumin by primary human dermal and lung microvascular endothelial cells.

For the first time, we identify CD36 as a receptor capable of mediating albumin transcytosis by the skin but not the lung endothelium. In keeping with these findings, mice deficient in endothelial CD36 exhibit decreased basal vascular permeability to albumin in the skin but not in the lung. Furthermore, these mice exhibit decreased amounts of subcutaneous adipose tissue despite having similar circulating fatty acid levels and similar total body weight. These findings suggest that CD36 mediates endothelial transcytosis of albumin in the skin where its purpose is to regulate fatty acid delivery.

MATERIALS AND METHODS

Cell culture.

Primary human microvascular endothelial cells of dermal origin (HDMECs) were isolated from discarded human foreskin as previously described (22) and were used throughout the study. Primary human microvascular pulmonary endothelial cells (HPMECs) were purchased from Lonza (HMVEC-L, CC-2527, Basel, Switzerland). All endothelial cells were cultured in EGM-2 BulletKit medium (Lonza, Switzerland), grown in 37°C at 5% CO₂, and were used from passages 3 to 8. Chinese hamster ovary (CHO) cells (ATCC) were cultured in DMEM/HAM’S F-12 50/50 mix with l-glutamine (MULTICELL, Wisent Bioproducts), supplemented with 10% fetal bovine serum and penicillin/streptomycin. The cells were subcultured every 5–7 days using trypsin, and the medium was changed every 2–3 days.

Immunofluorescence.

HDMECs and HPMECs were seeded on 0.1% gelatin-coated coverslips and grown until confluent. Cells were fixed in 4% paraformaldehyde for 20 min, then incubated with 0.15% glycine overnight at 4°C. Cells were blocked with 5% donkey serum for 45 min followed by incubation with anti-CD36 IgM (BD PharMingen, 555455) for 1 h. Cells were then washed with PBS and incubated with Cy3-conjugated donkey anti-mouse IgM (Jackson ImmunoResearch, 1:1,000) for 1 h. Coverslips were mounted on DAPI and imaged using spinning-disk confocal microscopy (Quorum Diskovery/Nipkow, Hamamatsu ImagEM X2 EM-CCD camera, at ×63 objective, numerical aperture 1.47) using the 561 laser for Cy3 and acquiring z-stacks at 0.3-μm intervals.

TIRF assay.

TIRF microscopy images were acquired on an Olympus cell TIRF Motorized Multicolor TIRF module mounted on an Olympus IX81 microscope (Olympus, Hamburg, Germany). Samples were imaged using a ×150/1.45 objective with 491-nm excitation lasers and Volocity acquisition software. Unless otherwise indicated, the penetration depth was set at 110 nm. For each cell, 150 TIRF images were acquired at a frame rate of 10 per second for a constant duration of 15 s, generating a video for every cell imaged. At least 10 randomly selected cells were imaged in each experimental replicate. Endothelial cells were subcultured and seeded on 25-mm glass coverslips until confluency was reached (24–48 h postseeding) and placed in serum-free EGM-2 media 2 h before every albumin TIRF assay. The coverslip to be imaged was placed inside an Attofluor cell chamber (Life Technologies, cat. no. A-7816). Alexa-Fluor 488 (AF488) albumin (ThermoFisher Scientific, A13100) was added to the cells, which were then incubated at 4°C for 10 min to allow apical membrane binding. Cells were rinsed twice in cold PBS(+) to remove unbound albumin and fresh warm media (HEPES-buffered RPMI) was added, along with nuclear stain NucBlue Live ReadyProbes (Molecular Probes, Oregon). The chamber was placed immediately on a heated (37°C) stage, and TIRF images of the basal membrane (to detect exocytosis of the apically applied ligand) were taken after 2 min of equilibration. Each field of view was imaged for 15 s, and then a new field of view was selected; each coverslip was imaged for 15 min. For the dose-response curves, AF488-albumin was dissolved in RPMI to concentrations of 5–50 μg/ml and added to coverslips of confluent HPMECs and HDMECs followed by analysis by TIRF. For the competition assay, 100-fold excess (by mass, i.e., 1,000 μg/ml) and 200-fold excess (2,000 μg/ml) of fatty acid-free BSA (Sigma, cat. no. A-7030) was dissolved in RPMI with 10 μg/ml of AF488-albumin; this solution was then added to cells followed by analysis by TIRF.

Inhibitor pretreatments.

To inhibit fluid-phase internalization by macropinocytosis, serum-starved HPMECs and HDMECs were treated with 1 mM of amiloride hydrochloride hydrate (Sigma, cat. no. A-7410) for 20 min at 37°C before TIRF. A 100-mM stock of amiloride in DMSO was diluted in serum-free EGM-2 media to a final concentration of 1 mM. Because amiloride is insoluble in aqueous solution, the media were heated at 45°C to ensure dissolving of the drug before addition to the cells. Because of the reversibility of amiloride’s effects, it was added during the membrane-binding step of AF488-albumin and to the warm media for imaging. To block CD36, 400 μM sulfo-N-succinimidyl oleate (SSO; Sigma) was applied to HPMECs and HDMECs. Following 2 h of serum starvation, HPMECs and HDMECs were treated with SSO for 30 min at 37°C, immediately following which albumin transcytosis was measured by TIRF.

Quantification of TIRF videos.

Blinded and automated quantification of the transcytotic events was performed using a tracking algorithm for MATLAB as previously reported (4). Briefly, the scripts correct the image for noise and local background using a Gaussian filter. Then putative vesicles are identified based on size (9–36 pixels², XY dimension, 73.5 nm/pixel), aspect ratio (>0.2; the ratio of the minor axis to the major axis), and intensity (threshold of 10% above mean image intensity). The tracking algorithm then tracks each moving vesicle based on a maximum-probability assessment of how closely those potential tracks resemble free and super-diffusive Brownian diffusion. The resulting tracks are analyzed for the duration of the vesicle being stationary in the TIRF field (vesicle docking), the speed of vesicular movement, and the degrees to which the particles’ movements deviate from free Brownian diffusion (γ). Vesicles undergoing fusion with the plasma membrane (exocytosis) are identified as those having a γ significantly less than that of an equivalent model population undergoing Brownian diffusion, typically 0 < γ < 0.873 and which undergo a decrease in fluorescence signal over the last 2 time points of their tracks equivalent to a drop of at least 2.5 standard deviations of vesicular intensity over the entire period the vesicle has been tracked. After disappearance, tracks are monitored for two additional frames to ensure they do not return.

Albumin internalization assay.

CHO cells transfected with wild-type green fluorescent protein (GFP)-tagged CD36 or the GFP vector alone were incubated with 20 μg/ml of AF-555-albumin for 10 min at 37°C, following which cells were washed twice with PBS(+) and immediately fixed using 4% paraformaldehyde. After neutralization with glycine, coverslips were mounted on glass slides using mounting medium (Dako) supplemented with 1 μg/ml DAPI (Sigma-Aldrich). Coverslips were imaged using the spinning-disk microscope by spinning-disk confocal microscopy (Quorum Diskovery/Nipkow, Hamamatsu ImagEM X2 EM-CCD camera, at ×63 objective, numerical aperture 1.47) using the 488 laser for GFP and 561-nm laser for AF-555-albumin, and acquiring z-stacks at 0.3-μm intervals; randomly selected fields of transfected cells were imaged keeping the microscope settings constant. Albumin internalization was quantified in the merged z-stacks using the puncta analysis tool in the ImageJ software (NIH); this value was normalized to the total number of GFP-transfected cells in the image.

CD36 knockdown and mutant constructs.

Depletion of CD36 in both HPMECs and HDMECs was carried out by electroporation using the Neon Transfection System (Invitrogen by Life Technologies, Thermo Fisher Scientific, Inc.) and CD36 siRNA (FlexiTube GeneSolution GS948 for CD36, cat. no. 1027416) from Qiagen (Valencia, CA). The HPMECs were transfected at 1,650 V for 20 ms and with 1 pulse; the HDMECs were transfected at 1,200 V for 40 ms and with 1 pulse. TIRF experiments and cell lysis for Western blot analysis and quantitative polymerase chain reaction (qPCR) were performed 48 h posttransfection. To overexpress CD36 plasmids, CHO cells were transfected with wild-type GFP-CD36 or indicated GFP-tagged mutants (or GFP alone) using Lipofectamine 3000 (ThermoFisher Scientific); experiments were performed 48 h later.

GFP-CD36 mutants were generated by ligation-independent cloning using the In-Fusion HD Eco Dry Cloning kit from Clontech (cat. no. 639689) and verified by sequencing. All mutants were confirmed to retain targeting to the plasma membrane by fluorescent microscopy. Transfected cells were exposed to AF-555-albumin for 10 min, rinsed, fixed, and then imaged. Uptake of AF-555-albumin by transfected cells was measured using spinning-disk confocal microscopy with settings kept constant between conditions. In each experiment, 30 cells were imaged for each plasmid, and images were analyzed by ImageJ. The mean fluorescence intensity of AF-555-albumin per transfected cell was measured and normalized to the mean intensity of wild-type GFP-CD36-transfected cells.

Macropinocytosis assay.

The macropinocytosis assay was modified from Canton et al. (13). Briefly, HDMECs and HPMECs plated on 25-mm coverslips were serum-starved for 2 h, following which they were treated with 1 mM amiloride for 20 min in serum-free EGM-2. Cells were then incubated with 100 μg/ml of tetramethylrhodamine-conjugated 70-kDa dextran in serum-free media at 37°C for 15 min, washed twice with PBS(+), and imaged immediately after addition of the nuclear stain NucBlue. Randomly selected fields of cells were then imaged by spinning-disk confocal microscopy (Quorum Diskovery/Nipkow, Hamamatsu ImagEM X2 EM-CCD camera, at ×63 objective, numerical aperture 1.47) using a 561-nm laser and acquiring z-stacks at 0.3-μm intervals. Macropinosomes across all z-stacks were then counted using the puncta analysis tool in the ImageJ software (NIH), with the filter for size set to an area of 10–100 pixels² and circularity to 0.2–1.0. This value was normalized to the total number of nuclei in the image. Microscope settings were kept constant between conditions.

Western blot analysis.

Cell lysates were prepared with 10% SDS lysis buffer (62.5 mM Tris-HCl pH 6.8, 10% SDS, 10% glycerol, 10 mM DTT). Equal amounts of protein were run on an SDS-PAGE using 9% polyacrylamide gels at 120 V; proteins were transferred to nitrocellulose membranes (110 V for 70 min) and blocked for 1 h with 5% milk in Tris-buffered saline-Tween 20. Membranes were incubated with the following primary antibodies overnight at 4°C: anti-CD36 (Santa Cruz Biotechnology, cat. no. sc-7309) and anti-β-actin (Santa Cruz, cat. no. sc-47778). The following day membranes were washed in Tris-buffered saline-Tween 20 and incubated with anti-rabbit and anti-mouse horseradish peroxidase-conjugated secondary antibodies, respectively, at a 1/10,000 dilution for 1 h, washed, and then visualized by enhanced chemiluminescence (Amersham), imaged using the ChemiDoc Imaging System (Bio-Rad), and quantified using the ImageLab software (Bio-Rad).

qPCR.

cDNA synthesis was carried out using the High-Capacity cDNA Reverse Transcription Kit according to the manufacturer’s instructions. For each sample, RNA was reverse-transcribed using T-Gradient Thermoblock (Biometra) according to the manufacturer’s directions. qPCR was conducted using Power SYBR Green PCR Master Mix (Applied Biosystems). cDNA was denatured at 95°C for 10 min followed by 40 cycles of 95°C for 15 s then 60°C for 1 min. qPCR was performed with the ABI Prism 7900HT (Applied Biosystems), and the data were analyzed with SDS software v2.1 (Applied Biosystems) and Microsoft Excel 2003 (Microsoft). Relative gene expression was compared using the comparative C_T method using 18S as the reference. Primer sets used for this study are as follows: CD36: (qPCR) forward 5′-TCTTTCCTGCAGCCCAATG-3′; reverse 5′-AGCCTCTGTTCCAACTGATAGTGA-3′; 18s rRNA: (qPCR) forward 5′-GATGGAAAATACAGCCAGGTCCTA-3′ and reverse 5′-TTCTTCAGTCGCTCCAGGTCTT-3′.

Mice.

Male mice with an endothelial cell-specific deletion of CD36 (Fl/FlCD36 Tie2eCre⁺; EC-CD36 KO) were generated by crossing CD36 floxed mice on a C57BL/6 background with C57BL/6 mice expressing Cre driven by the Tie2 5′ promoter and first intron enhancer element. Tie2eCre is distinct from Tie2Cre and is restricted to endothelia (52). Mice were genotyped by ear clippings and PCR using the RedExtract-N-Amp Tissue PCR Kit (Sigma, XNAT 100RXN), and primers for floxed CD36: FL1v(For): 5′-CCA CAC TGT ATG GGG AAA GTT TCA GG-3′, FL3v(Rev): 5′-CGC CCT ATC TAG TTT CTC CAC CC-3′, and the primers for Tie2eCre: (Tie2::Cre-FWD): 5′-CCC TGT GCT CAG ACA GAA ATG AGA (Tie2 FOR), (Tie2::Cre-REV): 5′-CGC ATA ACC AGT GAA ACA GCA TTG C (CRE REV) (11). Age and weight-matched male C57BL/6J mice were purchased from Jackson (Bar Harbor, ME) and used for controls. Mice were housed in the St. Michael’s Hospital Vivarium on a standard light-dark cycle and were given free access to food and water. Experimental procedures were conducted in accordance with the St. Michael’s Hospital Animal Care Committee guidelines and subject to approved animal protocols (ACC721/772 and ACC670).

Wet-to-dry wt ratio and miles assay.

To measure vascular permeability, a modified Miles assay was performed. Prior to the procedure, mice were anesthetized under inhaled isoflurane gas at 3%, dorsal hair was removed through shaving, and surgical plane of anesthesia was established by lack of response to toe pinch. Mice were then administered 5 mg/g of a 1% Evan’s blue dye solution in 0.9% NaCl through tail-vein injection via a 27-gauge needle. Twenty-four hours later, mice were placed under inhaled isoflurane gas at 4% and surgical plane of anesthesia was established by toe pinch; mice were then euthanized by cardiac puncture. The circulatory system was then flushed with 10 ml cold PBS by instillation through the left ventricle.

Both lobes of the lung were removed, patted dry, and then weighed for determination of wet weight. The dorsal skin of the mice was removed by gentle dissection and immediately weighed. The skin and lung tissues were dried for 24 h and then weighed for the dry weight. The dried skin and lung tissue were then placed in formamide at 50°C for 72 h to extract Evan’s blue dye. The absorbances at 620 (A620) and 740 nm were measured in the extractions. Dye content was calculated by correcting A620 for heme and converted to µg/ml Evan’s blue dye by comparing to a standard curve and normalized to tissue weight and corrected for the amount of dye injected.

Immunohistochemistry.

Skin samples were dissected from five standardized regions of the mouse dorsum (age-matched mice, 12–22 wk old) and embedded in either paraffin wax or optimal cutting tissue compound (OCT; Fisher Healthcare). For paraffin embedding, the skin sections were fixed in 10% formalin for 2–4 days, following which they were processed and embedded in paraffin wax. For OCT embedding, fresh skin samples were snap-frozen in liquid nitrogen and immediately placed in OCT compound, frozen, and stored at −80°C until cut. The paraffin and OCT blocks of skin tissue were cut in serial 5-μm-thick sections, mounted on glass slides, and stained with hematoxylin and eosin.

For immunofluorescence on lung and dermal tissue, paraffin-embedded tissue was deparaffinized then boiled for 20 min in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween-20, pH 6.0) for antigen retrieval. Slides were washed in TBS-0.1% Tween-20 and incubated with PBS-1% triton X-100 for 30 min at room temperature. Samples were circled using a hydrophobic pen and blocked in 10% rabbit serum with 1% BSA in 1× Tris-buffered saline for 1 h. Samples were then blocked with AffiniPure Fab Fragment Goat Anti-mouse IgG (Jackson ImmunoResearch, 13703) for 1 h at room temperature. Samples were incubated with primary antibody (1:400 rabbit anti-mouse Tie-2, Santa Cruz Biotechnology, H-176, sc-9026) and 1:50 anti-CD36 (BD PharMingen) overnight at 4°C; for some samples the primary antibodies were omitted as controls. Tissues were then incubated with conjugated secondary antibody Cy3-conjugated goat anti-mouse (Jackson ImmunoResearch) and AF-488-conjugated goat anti-rabbit (Jackson ImmunoResearch) for 1 h at room temperature. Slides were counterstained with DAPI.

Quantification of subcutaneous fat layer thickness.

The hematoxylin-eosin-stained sections were imaged using the Nikon Upright E800 Microscope at ×4 magnification, and images were acquired and analyzed using the NIS-elements software. Blinded quantification for the thickness of the subcutaneous fat layer was done using the measurement tool on the NIS-element software; 10 randomly selected, evenly distributed vertical measurements were taken across the entirety of every skin section. The 10 measurements were averaged to obtain the mean thickness of the skin sections.

Blood chemistry analysis.

Blood samples were collected from age-matched mice (10–20 wk old); serum and plasma were sent to The Centre for Phenogenomics (Toronto, Ontario) for analysis of total cholesterol, HDL, triglycerides, LDL, glucose, and albumin. The measurement for blood plasma fatty acid levels was done using the NEFA-HR (2) Free Fatty Acid detection kit (Wako Diagnostics) in accordance with the manufacturer’s instructions.

Statistical analysis.

Statistical analysis was performed using GraphPad Prism software (GraphPad Prism 5.0; GraphPad Software, La Jolla, CA). Student’s t-tests and one-sample t-tests (GraphPad, La Jolla, CA) were used to determine the significance of raw or normalized data, respectively, from two groups. A one-way ANOVA and post hoc t-tests were conducted for experiment with three groups or more. All experiments were performed at least five times on different batches of cells; data are presented as means ± SE.

RESULTS

Albumin transcytosis is saturable in dermal but not lung microvascular endothelial cells.

Using TIRF, we compared transcytosis by primary human dermal and lung microvascular endothelial cells of Alexa 488-conjugated albumin added to the apical surface. Dermal endothelial cells exhibited a plateau in transcytosis events when the dose of fluorescent albumin was increased, whereas lung endothelial cells performed transcytosis at an increasing rate (Fig. 1A). Consistent with the saturability data, competition with excess unlabeled and defatted albumin significantly attenuated transcytosis by dermal but not lung microvascular endothelia (Fig. 1, B and C). Thus, transcytosis of albumin by dermal and lung microvascular endothelial cells displays different kinetics, demonstrating saturability in the skin but not in the lung.

Fig. 1.

Albumin transcytosis is saturable in dermal but not lung microvascular endothelial cells. Dose responsiveness of human pulmonary microvascular endothelial cells (HPMECs; lung) and human dermal microvascular endothelial cells (HDMECs; skin) to Alexa-Fluor 488 (AF488)-albumin transcytosis determined by TIRF microscopy (n = 5 independent experiments for each concentration) over 15 s of observation (A). Competition of AF-488-albumin (10 μg/ml) transcytosis by 100- and 200-fold excess unlabeled and defatted BSA (B). Representative stills from the TIRF videos (i.e., images of the basal membrane) are shown on the right (C); size bar is 11 μm. Single cells (10–15) were imaged for each n in A and C. *P < 0.05; **P < 0.005. TIRF, total internal reflection fluorescence.

CD36 mediates albumin transcytosis by dermal microvascular endothelial cells.

The saturability and competition data suggested the presence of a receptor on dermal endothelial cells capable of mediating albumin transcytosis. The scavenger receptor CD36 is expressed on capillary endothelial cells of the skin (17), although its expression in the lung endothelium has been controversial (32, 48). Given reports that it can bind albumin in epithelial cells (7), we considered that it might contribute to albumin transcytosis. Dermal and lung microvascular endothelial cells expressed CD36 in whole-cell lysates, whereas the receptor was absent from CHO cells (Fig. 2A). Inducing its expression by transient transfection in CHO cells was sufficient to increase albumin internalization well over twofold. Knockdown of CD36 by siRNA in the dermal endothelial cells depleted the protein by almost 80% and reduced albumin transcytosis by almost 40% (Fig. 2B). In contrast, a 70% reduction in CD36 in the lung endothelial cells had no effect on albumin transcytosis (Fig. 2C). Thus, CD36 is both necessary and sufficient for albumin internalization and transcytosis by dermal but not lung microvascular endothelial cells.

Fig. 2.

CD36 mediates albumin transcytosis by dermal microvascular endothelial cells. Western blot showing expression of CD36 in HDMECs (dermal) and HPMECs (lung) but its absence in CHO cells (left). Transfection of CHO cells with GFP-CD36 significantly increases internalization of Alexa 555-conjugated (AF555)-albumin (middle and right, respectively); n = 5; 8 randomly acquired images for each n; size bar is 20 μm (A). Knockdown of CD36 expression by siRNA significantly reduces Alexa 488-conjugated (AF488)-albumin transcytosis in HDMECs (B) but not HPMECs (n = 5; 10–15 cells were imaged for each n and each point represents one TIRF video) (C). ***P < 0.001, ****P < 0.0001. CHO, Chinese hamster ovary; GFP, green fluorescence protein; HDMEC, human dermal microvascular endothelial cells; HPMEC, human pulmonary microvascular endothelial cell; NS, not significant; TIRF, total internal reflection fluorescence.

Pinocytosis contributes to albumin transcytosis in lung but not dermal microvascular cells.

It was intriguing that although both skin and lung microvascular endothelial cells express abundant amounts of CD36, the receptor appears to perform albumin transcytosis only in the skin. By immunofluorescence, the subcellular distribution of the receptor appeared similar on both lung and dermal microvascular endothelial cells (Fig. 3A); we also confirmed that the receptor is expressed by endothelial cells in both lung and dermal tissue in vivo (Fig. 3, B–C). Instead, we reasoned that a high rate of constitutive pinocytosis by lung endothelial cells might outweigh the contribution of plasmalemmal CD36 to albumin internalization. To test this notion, we incubated skin and lung endothelial cells with 70-kDa rhodamine-conjugated dextran to measure macropinocytosis (13). As a control, we incubated the cells with amiloride, which inhibits macropinocytosis by lowering the submembranous pH (26). Lung endothelial cells internalized several fold more dextran than skin endothelial cells, and this was significantly attenuated by amiloride (Fig. 4A). Similarly, amiloride significantly inhibited albumin transcytosis by lung endothelial cells but had no effect on albumin transcytosis by the dermal microvascular endothelium (Fig. 4B); the inhibition by amiloride in the lung endothelial cells persisted even after depletion of CD36 by siRNA (not shown). Thus, it is likely that a high basal rate of macropinocytosis obviates a meaningful contribution of CD36 to albumin transcytosis in the lung.

Fig. 3.

Lung and dermal endothelium express CD36. Primary human lung (HPMEC) and dermal (HDMEC) microvascular endothelial cells were immunostained with CD36 and analyzed by confocal microscopy (A). Representative XY and YZ images are shown; no difference in subcellular distribution was observed. Size bar is 6 μm for XY images and 5 μm for YZ images. Lung sections from C57BL/6 mice were probed for CD36 (red) and Tie2 (green) (B). Skin sections from C57BL/6 mice were probed for CD36 and Tie2 (C). Size bar is 6 μm.

Fig. 4.

Pinocytosis contributes to albumin transcytosis in lung but not dermal microvascular cells. HPMECs (lung) perform more macropinocytosis relative to HDMECs (skin) determined by internalization of 70 kDa tetramethylrhodamine (TMR)-dextran (red); amiloride is a known inhibitor of macropinocytosis (n = 5); nuclei are stained with NucBlue (blue). Scatterplot shows the quantification of TMR-dextran internalization (punctae) normalized to the number of cells per field (n = 5; 10 randomly selected fields for each n) (A). Inhibition of macropinocytosis by amiloride attenuates Alexa Fluor 488 (AF488)-albumin transcytosis in HPMECs but not HDMECs, as measured by TIRF; DMSO is the solvent control (n = 5; 10 single cells were imaged for each n and each point represents one TIRF video) (B). ***P < 0.001, ****P < 0.0001. HDMEC, human dermal microvascular endothelial cells; HPMEC, human pulmonary microvascular endothelial cell; NS, not significant; TIRF, total internal reflection fluorescence.

SSO and CD36 mutants define a putative binding region for albumin.

SSO binds irreversibly to lysine 164 in the extracellular loop of CD36, inhibiting its binding to long-chain fatty acids and oxidized LDL (28). Incubation with SSO for 30 min significantly attenuated albumin transcytosis by dermal microvascular endothelial cells but had no effect on the lung endothelial cells (Fig. 5A). These data suggested that the binding site for albumin was likely overlapping with that reported for long-chain fatty acids and other ligands. To test this by a more specific approach, we generated GFP-tagged CD36 mutants spanning the ligand-binding extracellular domain of the protein and verified that none of the mutations interfere with delivery of the receptor to the plasma membrane. We next transfected the mutant CD36 constructs into CHO cells, taking advantage of the lack of endogenous receptor; the cellular internalization of Alexa-555-conjugated albumin by transfected cells was then compared. Mutations in the α-helical bundle near the apex of the protein (38) (e.g., amino acid residues 153 and 156) greatly decreased albumin uptake relative to the wild-type receptor (Fig. 5B), despite their unimpaired localization to the plasma membrane (Fig. 5C). In contrast, a mutation just outside of the putative ligand-binding region (amino acid residue 78) had a more modest effect on albumin internalization.

Fig. 5.

SSO and CD36 mutants define a region important for albumin internalization. HPMEC (lung) and HDMEC (dermal) were incubated with sulfo-N-succinimidyl oleate (SSO, 400 μM) or solvent control for 30 min before measurement of albumin transcytosis by TIRF; data are from five independent experiments with each point representing one TIRF video, *P < 0.05 (A); CHO cells were transfected with GFP vector alone or GFP-tagged wild-type (WT) or mutant CD36 constructs (mutated amino acid residue as indicated on the x-axis) and allowed to internalize Alexa-Fluor 555-conjugated albumin for 10 min (B). Internalization of albumin was quantified; n = 5 independent experiments with 30 transfected cells counted per construct, per experiment. Data are normalized to WT CD36-GFP and are presented as means and SD; ***P < 0.001 by one-way ANOVA and Tukey’s multiple comparison post hoc test. C: representative images of CHO cells transfected with mutant CD36 constructs or GFP vector alone; white arrow indicates targeting of the CD36 construct to the plasma membrane. Scale bar is 60 μm. CHO, Chinese hamster ovary; GFP, green fluorescence protein; HDMEC, human dermal microvascular endothelial cells; HPMEC, human pulmonary microvascular endothelial cell; NS, not significant.

Endothelial-specific loss of CD36 leads to decreased albumin permeability and tissue edema in the skin.

To determine whether our in vitro findings could be replicated in vivo, mice deficient in CD36 specifically in the endothelium (EC-CD36KO) were generated and used for a modified Miles assay. In this procedure, Evan’s blue dye is injected into the circulation where it binds tightly to albumin. Extravasation of Evan’s blue-bound albumin is often used as an indicator for increased endothelial leakage in response to an inflammatory stimulus (39). By injecting Evan’s blue into healthy wild-type and EC-CD36KO mice—and without administering any inflammatory mediator—we measured the flux of albumin out of the circulation under basal conditions. In this setting, transcytosis is thought to account for the majority of albumin efflux from the blood (35). We observed significantly less Evan’s blue accumulation in the skin of EC-CD36KO mice compared with wild-type; importantly, this difference was not observed in the lungs (Fig. 6A). As albumin is the main contributor of oncotic force and edema formation, we then measured the wet-to-dry wt ratio for the skin and lungs of the animals. Although the wet-to-dry wt ratio in lungs was the same between knockout and wild-type mice, it was significantly lower in the skin of the EC-CD36KO animals (Fig. 6B). Thus, CD36 contributes to albumin transcytosis across the endothelium both in vitro and in vivo, in mice.

Fig. 6.

Endothelial-specific loss of CD36 leads to decreased albumin permeability and tissue edema in the skin. A: Fl/FlCD36 Tie2eCre⁺ (EC-CD36 KO) and age-matched C57BL/6 WT mice were injected intravenously with 1% Evan’s blue dye (EBD); 24 h later, the accumulation of dye in the skin (left) and lungs (right) was compared. Dye is normalized to dry weight of tissue (n = 6 for WT mice, n = 7 for EC-CD36 KO mice); photographs display representative images of shaved dorsal skin region of mice 24 h after EBD. B: basal tissue fluid content was determined by calculating the wet-to-dry wt ratio; note the lower ratio in the skin but not the lungs of the EC-CD36 KO mice relative to WT (n = 6 for WT mice, n = 6 for EC-CD36 KO mice). *P < 0.05; **P < 0.005. Each point represents one animal. NS, not significant; WT, wild-type.

Endothelial-specific loss of CD36 leads to decreased fat deposition in the skin.

Given that albumin possesses multiple binding sites for fatty acids (16), we hypothesized that albumin transcytosis might play an important and as-of-yet undescribed role in fatty acid metabolism. Blinded pathological examination of the skin of age-matched mice revealed a significantly thinner subcutaneous fat layer in the knockout mice (Fig. 7A), despite indistinguishable circulating fatty acid and albumin levels (Fig. 7, B–C). The difference in subcutaneous fat thickness was not because of a lower total body weight, as both groups of animals displayed similar body weights and weight gain over time (Fig. 7D). Circulating triglycerides tended to be lower in the knockout mice, but this was not statistically significant; LDL and HDL levels were unchanged (Fig. 7E). Thus, deficiency in albumin transcytosis is associated with decreased deposition of subcutaneous fat despite similar circulating fatty acid levels. This suggests a role for albumin transcytosis in the traffic of circulating fatty acids to downstream tissues such as the skin.

Fig. 7.

Endothelial-specific loss of CD36 leads to decreased fat deposition in the skin. H&E-stained dorsal skin sections (5 μm) from EC-CD36 KO and age/weight-matched C57BL/6 mice were measured for thickness of the subcutaneous fat layer: a, epidermis; b, subcutaneous fat; c, muscle; data reflect sections from five mice in each group, colors represent age-matched pairs (blue = 22 wk; yellow = 20 wk; other colors are 12-wk-old pairs). Size bar is 1,000 μm (A). Circulating free fatty acid levels and serum albumin in EC-CD36 KO and WT mice (B and C). Total body weight in EC-CD36 KO and WT mice measured over two weeks (n = 5 for each) (D). Plasma triglycerides, HDL, LDL, and total cholesterol levels in EC-CD36 KO and WT mice (n = 5 per group) (E); **P < 0.005. EC-CD36 KO, Fl/FlCD36 Tie2eCre⁺ mice; H&E, hematoxylin-eosin; NS, not significant; WT, wild-type.

DISCUSSION

The regulation and the purpose of albumin transcytosis by the endothelium has remained a largely unanswered question in vascular biology. Serial electron microscopic studies in animals have demonstrated that intracellular vesicles transport albumin from the vascular lumen to the interstitial space, but although elegant, this approach is not well suited to addressing mechanistic questions. Work to date has reported that a receptor named gp60 mediates low-affinity binding of albumin by the lung endothelium (50) but also showed that most transcytosis of albumin in this organ is performed via fluid-phase uptake (25). Almost nothing is known about the mechanisms of albumin transcytosis in other organs. Organ-specific mechanisms of endothelial permeability are not without precedent (1, 2); endothelial cells from different tissue beds are extremely heterogeneous, with the same receptor exhibiting different downstream signaling in different vascular beds (20, 44).

Using a highly sensitive TIRF microscopy-based approach, we report for the first time that albumin transcytosis by human dermal microvascular endothelial cells is saturable, suggesting a receptor-mediated process. In contrast [and as reported by others (25)], albumin transcytosis by lung microvascular endothelial cells is not saturable and is dominated by fluid-phase uptake. We extend these observations to report a novel role for the scavenger receptor CD36 in albumin transcytosis. The receptor is expressed in high levels at the endothelial membrane and its transfection into CHO cells, which lack endogenous CD36, is sufficient to increase albumin uptake. Depletion or pharmacologic inhibition of CD36 in dermal (but not lung) microvascular endothelial cells significantly attenuates albumin transcytosis. Finally, mice deficient in CD36 in endothelial cells exhibit reduced accumulation of Evan’s blue in the skin, confirming a reduction in basal albumin traffic.

CD36 is a transmembrane protein composed of a single extracellular loop with two relatively short cytoplasmic tails (41). It is expressed on numerous cell types, including phagocytes, myocytes, and capillary endothelial cells (19); for the purposes of our study, it is interesting to note that it is absent from dermal lymphatics (23). Several putative ligands have been described, including collagen, oxidized LDL (19), and long-chain fatty acids (24). We now add albumin to this list; the specificity of the interaction is supported by the observation that defatted albumin significantly attenuated albumin transcytosis, effectively excluding confounding by fatty acids. Although kidney epithelial cells have been reported to bind to albumin via CD36 (7), to our knowledge this is the first description of it mediating endothelial albumin transcytosis.

The CD36 inhibitor SSO binds irreversibly to lysine 164 (28) in the extracellular loop in a region that has been reported to interact with long-chain fatty acids (6). Our data using both SSO and mutants of CD36 with single-codon substitutions in the apical helical domain (38) suggest that this region is also important for binding or for internalizing albumin. Almost nothing is known about postreceptor signaling during albumin transcytosis. Binding of oxidized LDL to CD36 initiates Src kinase-dependent signaling with effects on MAPK activation and focal adhesion kinase (49); whether albumin binding and transcytosis by CD36 requires a similar or a distinct signal transduction cascade remains to be elucidated.

The binding of CD36 to long-chain fatty acids has been a focus of much research, with mutations in the receptor being associated with defects in fatty acid utilization by adipose tissue, the myocardium, and other tissues (14). CD36 has been postulated to bind to fatty acids at the cell surface, facilitating their integration into the cell by a “flip-flop” mechanism. However, nonesterified fatty acids cannot circulate freely in plasma and are bound to albumin (40). As albumin possesses up to five binding sites for medium- or long-chain fatty acids (16), rather than simply binding fatty acids at the cell surface, CD36 and albumin likely serve to shuttle fatty acids out of the vasculature in a regulated fashion. Our observation that mice selectively deficient in endothelial CD36 exhibit reduced subcutaneous fat is therefore intriguing, given that circulating lipid levels were comparable to control animals and both groups displayed similar total body weights. This would be consistent with a defect in lipid transport or metabolism at the interface between the endothelium and the skin. Consistent with this hypothesis, analbuminaemic rats exhibit marked hypercholesterolemia (37) whereas patients congenitally deficient in albumin display elevated cholesterol and serum phospholipid concentrations (8, 15) that return to normal transiently following intravenous albumin infusions. Although deficiency of or mutations in CD36 have not previously been linked to a phenotype in the skin (31), previous studies have examined the effect of global deficiency of CD36 rather than endothelial cell-specific deletion; this may have confounded the phenotype. A more detailed characterization of the dermal adipose tissue in wild-type and EC-CD36KO mice is the subject of our ongoing work.

Although our data suggest an intriguing phenotype related to CD36 in the skin, we leave unanswered the question of the role of CD36 on the lung microvascular endothelium; although it does not appear to mediate albumin transcytosis, previous studies have suggested that lung endothelial CD36 is involved in the response to inhaled pollutants (42) or to infection (51).

Lastly, our data also suggest that additional receptors are likely to contribute to albumin transcytosis by the dermal microvascular endothelium. Several candidates exist, although little is known about their tissue distribution or downstream signaling (33). The development of a quantitative and sensitive assay for albumin transcytosis in primary endothelial cells using TIRF microscopy (4, 9, 29) may facilitate their study. There is growing interest in using albumin as a drug carrier (33), taking advantage of its abundance in the circulation and its long half-life. It is evident that understanding the mechanisms of albumin transcytosis in different tissues is likely to be critical to the success of this approach.

In conclusion, we describe differing mechanisms of albumin transcytosis between lung and skin microvascular endothelia and report a novel role for the CD36 receptor in mediating albumin transcytosis across human dermal microvascular endothelial cells. In mice, loss of endothelial CD36 is associated with decreased dermal fat and diminished albumin traffic out of the circulation, suggesting a potential function for albumin transcytosis in fatty acid metabolism.

GRANTS

This work was supported by Natural Sciences and Engineering Research Council of Canada Grant RGPIN-2015-05802 and Canada Foundation for Innovation Grant 34769 (both to W. Lee), as well as Natural Science Foundation of China Grant 31770938 and Key Program of Zhejiang Provincial Natural Science Foundation of China Grant LZ-16C-050001 (to D. Neculai). W. Lee is supported by a Canada Research Chair in Mechanisms of Endothelial Permeability.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

W.L.L. conceived and designed research; H.R., S.G., N.K., V.M., D.A., C.W., Y.H.K., M.H., D.N., M.F., and W.L.L. performed experiments; H.R., S.G., N.K., B.M., B.H., and W.L.L. analyzed data; H.R., S.G., N.K., C.W., B.M., H.-K.S., G.F., D.N., M.F., and W.L.L. interpreted results of experiments; H.R., N.K., and W.L.L. prepared figures; W.L.L. drafted manuscript; H.R., S.G., N.K., V.M., D.A., C.W., Y.H.K., B.M., H.-K.S., M.H., G.F., D.N., M.F., B.H., and W.L.L. edited and revised manuscript; H.R., S.G., N.K., V.M., D.A., C.W., Y.H.K., B.M., H.-K.S., M.H., G.F., D.N., M.F., B.H., and W.L.L. approved final version of manuscript.