CD36 Is a Marker of Human Adipocyte Progenitors with Pronounced Adipogenic and Triglyceride Accumulation Potential

Abstract

White adipose tissue (WAT) expands in part through adipogenesis, a process involving fat cell generation and fatty acid (FA) storage into triglycerides (TGs). Several findings suggest that inter-individual and regional variations in adipogenesis are linked to metabolic complications. We aimed to identify cellular markers that define human adipocyte progenitors (APs) with pronounced adipogenic/TG storage ability. Using an unbiased single cell screen of passaged human adipose-derived stromal cells (hADSCs), we identified cell clones with similar proliferation rates but discordant capabilities to undergo adipogenic differentiation. Transcriptomic analyses prior to induction of differentiation showed that adipogenic clones displayed a significantly higher expression of CD36, encoding the scavenger receptor CD36. CD36+ hADSCs, in comparison with CD36-cells, displayed almost complete adipogenic differentiation while CD36 RNAi attenuated lipid accumulation. Similar findings were observed in primary CD45-/CD34+/CD31-APs isolated from human WAT where the subpopulation of MSCA1+/CD36+ cells displayed a significantly higher differentiation degree/TG storage capacity than MSCA1+/CD36-cells. Functional analyses in vitro and ex vivo confirmed that CD36 conferred APs an increased capacity to take up FAs thereby facilitating terminal differentiation. Among primary APs from subcutaneous femoral, abdominal and visceral human WAT, the fraction of CD36+ cells was significantly higher in depots associated with higher adipogenesis and reduced metabolic risk (i.e., femoral WAT). We conclude that CD36 marks APs with pronounced adipogenic potential, most probably by facilitating lipid uptake. This may be of value in developing human adipocyte cell clones and possibly in linking regional variations in adipogenesis to metabolic phenotype.

Introduction

White adipose tissue (WAT) is remarkably plastic and expands by increasing both fat cell size (hypertrophy) and number (hyperplasia). Adipogenesis is the cellular process resulting in the formation of novel lipid-containing adipocytes from nondifferentiated precursor cells present in the stromal vascular fraction (SVF) of WAT. Several observations suggest that increased adipogenesis (resulting in many small fat cells), particularly in the subcutaneous depot, plays a major role in buffering fatty acid (FA) excess [1, 2]. Hence, triglyceride (TG) accumulation in specific WAT regions may reduce ectopic lipid deposition in the liver, skeletal muscle, and vessels. In line with this, stimulation of adipogenesis, for example, by ligands activating peroxisome proliferator-activated receptor-γ (PPARγ), the master transcription factor in adipogenesis, improves several metabolic parameters including insulin sensitivity [3]. Conversely, conditions with attenuated adipogenesis (resulting in few but large fat cells), for example, mediated by genetic manipulations in mice models or lipodystrophic syndromes in man are linked to changes in whole body energy homeostasis characterized by pronounced insulin resistance, hyperlipidemia, and atherosclerosis [4, 5]. Therefore, inter-individual variations in the ability to expand the fat mass through adipogenesis might constitute an important mechanism contributing to metabolic complications associated with obesity. It is well recognized that accumulation of fat mass in the central part of the body (abdominal subcutaneous and visceral adipose tissue) is deleterious while increased fat depots in the lower part of the body (subcutaneous gluteal and femoral) correlates with reduced cardiometabolic risk. Over-feeding studies in man suggest that femoral, but not abdominal subcutaneous WAT expands by increased proliferation [6]. Whether these effects are linked to depot-specific differences in adipogenic potential and/or TG accumulation remain to be established although a number of reports indicate that this may indeed be the case [7, 8].

White adipocyte differentiation involves a number of transcription factors (apart from PPARγ also C/EBPβ and –α) which together control the expression of genes involved in lipogenesis (e.g., lipoprotein lipase, FA transporters [FATs]), lipid droplet formation (e.g., perilipin, adipose differentiation-related protein [ADRP]) as well as lipid binding (FA binding proteins [FABPs]). The concerted action of the corresponding proteins promote the formation of a large intracellular TG-containing droplet, which together with the secretion of specific proteins (e.g., adiponectin, leptin) constitute a hallmark of mature adipocytes [9]. The cellular origin of adipocytes has been the focus of intense studies primarily in murine models [10], while much less is known about human cells [11]. Cell tracking approaches in mice [10, 12] as well as fluorescence-activated cell sorting (FACS) in murine and human adipose tissue [13–15] have characterized the phenotype of the cell fraction with adipogenic potential present in the stromal-vascular compartment of WAT. This has demonstrated that there is considerable heterogeneity among adipocyte progenitors (APs) in terms of adipogenic potential and their regional distribution. Furthermore, it is still unclear whether human fat cells arise from one, few or several progenitor pools [16]. This major caveat in understanding adipogenesis constitutes an obstacle for both in vitro and in vivo experiments as there are currently no immortalized adipocyte cell lines available (with the exception of a cell line derived from an infant with the Simpson-Golabi-Behmel syndrome, SGBS [17]). Identifying cellular markers of human APs with pronounced adipogenic/TG storage ability could enable the establishment of valuable cellular models for advanced mechanistic studies and could, at least theoretically, be of clinical value in future treatments of severe forms of lipodystrophy.

Several studies using FACS have identified the native human APs as a fraction of CD45-/CD34+/CD31- [11]. Using additional cell surface markers, we recently described that this cell population contains distinct cellular subsets [14]. Our results suggested that progenitor cells expressing mesenchymal stem cell antigen 1 (MSCA1, encoded by the ALPL gene) constituted a cell fraction with pronounced adipogenic potential [14].

In this study, we aimed to identify the cell population in the SVF of human WAT that display the highest adipogenic/TG storage capacity. Based on two distinct approaches including an unbiased screening in cultured human adipose-derived stem cells (hADSCs) as well as studies in primary APs from human WAT, our joint analyses suggest that the highest degree of differentiation capacity in vitro is observed in cells expressing the scavenger receptor/FAT CD36. Our data also demonstrate that in non-obese women, MSCA1+/CD36+ progenitors exhibit a depot-specific distribution with significantly higher numbers in the femoral compared with abdominal and visceral fat regions.

Materials and Methods

Subjects

WAT samples were obtained within clinical studies approved by the regional institutional review boards. All subjects provided written informed consent. Protocols are registered at ClinicalTrials.gov (GLUTAB: NCT01605578; SENADIP: NCT01525472). In brief, paired subcutaneous femoral and abdominal WAT biopsies (GLUTAB study) were obtained from non-obese healthy female volunteers (n = 17, age [mean ± SD]: 34 ± 7 years, range 25–44; BMI: 24.8 ± 1.5 kg/m², range 21.4–27.5) and paired subcutaneous and visceral adipose tissues (SENADIP study) were obtained from non-obese women undergoing gynecological surgery for non-malignant disorders (n = 24, age: 44 ± 9 years, range 21–59; BMI: 24.1 ± 2.7 kg/m², range 18.7–28).

Culture Selection and In Vitro Cell Differentiation

hADSCs were isolated from subcutaneous abdominal WAT from a male donor (16 years old, BMI 24 kg/m²) as described in detail previously [18]. In brief, WAT was dissociated for 50 minutes in Krebs-Ringer-Phosphate buffer containing 0.05% collagenase, and 4% bovine serum-albumin (BSA). The crude SVF was separated from the adipocyte fraction by 200 µM nylon filters followed by low speed centrifugation (200g, 10 minutes). Cells from the pelleted SVF were seeded onto uncoated tissue culture plates at a density of 3,500 cells/cm², followed by medium change after 12–14 hours. Fast-adherent cells were collected as hADSCs and were cultured and differentiated into adipocytes using described protocols [18–20], further detailed below. For the experiments described in the present work, cells from passage 11 were used to initiate proliferation. They were cultured in Dulbecco’s modified Eagle’s medium (DMEM 1 g/L Glucose, Lonza, Basel, Switzerland, http://www.lonza.com/) supplemented with 10% fetal calf serum (FCS, HyClone, GE Healthcare Life Sciences, Logan, UT, https://promo.gelifesciences.com/gl/hyclone/index.html), 2.5 ng/ml FGF2 (Sigma-Aldrich, St. Louis, MO, http://www.sigmaaldrich.com) and 100 U/ml penicillin/streptomycin. After reaching 90% confluence, adherent cells were dissociated in 0.05% trypsin-EDTA (Thermo Fisher Scientific, Waltham, MA, https://www.thermofisher.com) and seeded into new flasks at a density of 4,500 cells per cm². Cultures were maintained at 37°C in a standard humidified gassed incubator with 5% CO₂. Media were changed every other day. For adipocyte differentiation, cells were plated at high density (30,000 cells/cm²) and cultured in DMEM/Ham’s F12 media supplemented with 10 μg/ml transferrin (T8158, Sigma-Aldrich), 0.86 μM insulin (I9278, Sigma-Aldrich), 0.2 nM triiodothyronine (T6397, Sigma-Aldrich), 1 μM dexamethasone (DEX, D1756, Sigma-Aldrich), 100 μM isobutyl-methylxanthine (IBMX, I5879, Sigma-Aldrich), and 1 μM rosiglitazone (Cayman Chemical Company, Ann Arbor, MI, https://www.caymanchem.com) for the first 3 days. After this, cell cultures were incubated in similar medium except that DEX and IBMX were omitted. Cells were isolated at different time points during differentiation. Staining of nuclei and neutral lipids at the end of differentiation was performed as described previously [21].

Single-Cell Cloning by Limiting Dilution

Human ADSCs cells from passage 11 were trypsinized at passage 12 and plated in proliferation medium at a dilution of 1 cell/200 µl medium, corresponding to one cell/well in three 96-well cell culture plates. According to Poisson statistics, the used dilution provided 36.8% of the wells with 1 cell/well. After 12 days of proliferation, each well was examined to identify the presence of cell colonies. Wells containing more than 60 cells were treated with 50 µl 0.05% trypsin EDTA (Gibco, Thermo Fisher Scientific, Waltham, MA) and the cells were transferred into 12-well plates and continued to be cultured in proliferation medium. When they reached 90% confluence, the cells were expanded through a series of expansion steps in T25 flasks, 10 cm dishes, and T175 flasks, respectively (a schematic representation of the selection process is detailed in Fig. 1B). Clones that reached 90% confluence in T175 flasks were selected for further analysis. Cell doubling times for each clone were calculated using the online tool Doubling Time Computing (http://www.doubling-time.com/compute.php).

Targeting CD36 with siRNA

CD36 silencing was performed in hADSCs at passage 13 by reverse transfection using the ON-TARGET plus Human CD36 siRNA SMART pool (L-010206–00-0005) and ON-TARGET plus Nontargeting Pool siRNA (D-001810–10-05) from Dharmacon (Little Chalfont, UK, http://dharmacon.gelifesciences.com/). Transfections were performed in nondifferentiated hADSCs using DharmaFECT transfection reagents (Dharmacon) in 24- and 96-well cell culture plates. In brief, 100,000 cells/well in 24 well plates and 10,000 cells/well in 96-well plates were plated in proliferation medium where FGF2 had been removed one day before transfection. Final siRNA concentration in each well was 50 nM. This siRNA transfection protocol has been optimized in our hADSCs and results in >90% transfection efficiency (unpublished observations). This was confirmed herein by marked knockdown of CD36 both at the mRNA and protein level. Adipogenic differentiation was initiated 24 hours after transfection.

Western Blot

Analyses of protein levels in CD36+ and CD36- hADSCs as well as in CD36-silenced cells was performed by Western blot as described previously [22]. In brief, 8–10 µg of total protein lysates were separated by standard 10% sodium dodecyl sulphate-polyacrylamide gel electrophoresis and transferred to polyvinylidine fluoride membranes (GE Healthcare Life Sciences, Uppsala, Sweden, http://www.gelifesciences.com). Blots were blocked for 1 hour at room temperature in Tris-buffered saline with 0.1% Tween-20 (TBS-T) and 3% amersham enhanced chemiluminescence (ECL) blocking agent (GE Healthcare Life Sciences) followed by an overnight incubation at 4°C with primary rabbit antibodies against CD36 (HPA002018, Sigma-Aldrich), FABP4 (HPA002188, Sigma-Aldrich) and Actin (A2066, Sigma-Aldrich). Proteins were detected by chemiluminescence using a secondary anti-rabbit antibody conjugated to horseradish peroxidase (Sigma-Aldrich) and ECL Select™ western blotting detection reagent (GE healthcare life sciences). Blots were analysed using a Chemidoc XRS system (Bio-Rad Laboratories Inc., Hercules, CA, http://www.bio-rad.com).

Fluorescence Activated Cell Sorting of Human ADSCs and SVF Cells

The cultured hADSCs at passage 11 were dissociated by 0.05% Trypsin-EDTA (Gibco, Thermo Fisher Scientific, Waltham, MA) and re-suspended in phosphate-buffered saline (PBS) supplemented with 10% FCS. Cells were first incubated with purified ChromPure Mouse IgG (Jackson ImmunoResearch, West Grove, PA, https://jireurope.com) to block unspecific binding. For analyses of cell surface marker expression, the cells were stained with monoclonal anti-human antibodies specific for CD29-APC, CD31-PECY7, CD34-APC, CD36-PE, CD44-APCCY, CD45-Pacific blue, CD49B-FITC, CD49F-APC, CD90-PECY5, CD105-APC, CD106-PECY5, CD140A-PE, CD146-PE, Glycophorin A/CD235a-Pacific blue, and CD271-FITC for 15 minutes on ice. After washing with PBS + 10% FCS, the cells were resuspended in PBS + 10% FCS and analyzed by a LSRII Flow Cytometer (BD Biosciences, San Jose, CA, http://www.bdbiosciences.com) using the FlowJo software (TreeStar Inc. San Carlos, CA, https://www.flowjo.com/). Dead cells were excluded by propidium iodide staining. For cell differentiation and expansion assays, CD36+ and CD36- hADSCs were sorted using a FACSAriaIII Cell Sorter (BD Biosciences). The gates for analyzing the cell surface marker expression by FACS were set according to fluorescence-minus-one controls.

For analyses of primary human cells from the SVF of human subcutaneous WAT, samples were obtained as previously described [14]. To analyze cell subsets, a panel of cell surface marker antibodies (BD Biosciences, Le Pont de Claix, France) were used in addition to appropriate isotype control panels and included: anti-CD14-PEvio, CD31-BV450, CD34-PerCP, CD36-APC-Cy7, CD45-BV510, MSCA1-PE, and CD271-APC. For lipid staining, cells were incubated with 500 ng/ml Bodipy 493/503 (Life Technologies, Courtaboeuf, France, https://www.thermofisher.com) for 30 minutes at 37°C. After washes, anti-CD31-APC and anti-CD34-PerCP were added. For FATs, cells were stained with a panel of cell surface marker antibodies (anti-CD14-PEvio, CD31-BV450, CD34-PerCP, CD36-APC-Cy7, CD45-BV510), fixed and permeabilized using cytofix/cytoperm kit (BD Biosciences) and stained using FATP1-PE and FATP4-APC (BD Biosciences). Appropriate isotype control panel was performed in the same condition. Following incubation with the appropriate panel of antibodies and washing steps, 100,000 cells were analyzed by flow cytometry using a FACS Canto^TM II flow cytometer and Diva Pro software (BD Biosciences). See Supporting Information Table S1 for detailed information about the antibodies used in the study.

Lipolysis in Human Subcutaneous Adipose Tissue Explants

Subcutaneous abdominal WAT (30–40 g) was minced and washed in Krebs-Ringer-bicarbonate-HEPES (KRBH) buffer containing 0.5% BSA free fatty acid, pH = 7.4, then washed in endothelial cell basal medium (ECBM) supplemented with 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) and 0.5% BSA free fatty acid, pH = 7.4. WAT explants were incubated in cell culture cassettes (Clinicell, Mabio International, Tourcoing, France, http://www.mabio.net) filled with ECBM/HEPES/0.5% BSA free fatty acids, pH = 7.4, and lipolytic agents were added (1µmol/L isoprenaline, Sigma-Aldrich; 500 nmol/L human recombinant atrial natriuretic peptide [ANP], Bachem AG, Bubendorf, Switzerland, http://www.bachem.com) supplemented with 10 µmol/L Thiorphan (Sigma-Aldrich, St Quentin Fallavier, France). After 24 hours, explants were removed and conditioned media were filtered through 0.22 µm membranes, aliquoted and stored at −20°C. Non-esterified fatty acids (NEFA) were determined using a commercial kit (NEFA determination kit, WAKO, Sobioda, Montbonnot St. Martin, France, http://www.sobioda.com/). Explants were digested as detailed in reference[14] and stromal-vascular cells were analyzed by flow cytometry as described above or plated for adipogenic differentiation.

Lipid Uptake Assays

SVF cells obtained from human subcutaneous WAT were plated at 100,000 cells/cm² and maintained in ECGM-MV for three days. Cells were washed with PBS and maintained for 4 hours in HBSS supplemented with 0.1% BSA free fatty acid before adding 2 µmol/L Bodipy FLC12 (Thermofisher Scientific) for the indicated time. Pretreatments with 25 µmol/L succinimidyl oleate (SSO, Santacruz, Clinisciences, Montrouge, France, https://www.clinisciences.com/) were performed 30 minutes before adding Bodipy FLC12. Cells were then washed, recovered with trypsin/EDTA 0.1% and further analyzed by flow cytometry.

Transduction of Primary Human Adipose Tissue Progenitors

Cell sorting of progenitor cells CD45-/CD31-/CD34+/MSCA1+ was performed on WAT SVF isolated by flow cytometry (BD Influx, BD Biosciences). Freshly isolated MSCA1+ progenitor cells from three subjects were plated overnight (25,000 cells/cm²) in ECGM-MV medium (Promocell, Heidelberg, Germany, http://www.promocell.com) supplemented with antibiotics (100 U/ml penicillin/streptomycin, Sigma Aldrich, St Quentin Fallavier, France). Cells were transduced with human telomerase and human papilloma virus E6/E7 genes (lentivirus pLV-EF1-hTERT-WPRE and lentivirus pLV-EF1-E6E7-WPRE, Vectalys, Toulouse, France, http://www.vectalys.com) in the presence of 4µg/ml Polybrene for 6 hours. After infection, cells were maintained in ECBM medium (Promocell) supplemented with 5% FCS, 2mM L-glutamine and antibiotics for at least three passages. MSCA1+/CD36+ and MSCA1+/CD36- cells were then selected by cytometer cell sorting (BD Influx, BD Biosciences).

Adipogenic Differentiation of Human Adipose Tissue Cells

Cells were plated at high density (120,000 cells/cm²) in ECBM containing penicillin/streptomycin and supplemented with 10% FCS for 48 hours. Medium was changed into defined basal adipogenic medium in ECBM containing penicillin/streptomycin, 66 nM insulin, 1 nM triiodothyronine, 0.1 µg/ml transferrin, and 100 nM cortisol supplemented or not with 3 µM rosiglitazone. After 3 days, the media was replaced with basal defined adipogenic medium (without rosiglitazone) for the following 7 days.

Osteogenic Differentiation

MSCA1+/CD36+ and MSCA1+/CD36- cells were plated at 120,000 cells/cm² in osteogenic induction medium (ECBM high glucose [4.5 g/l] supplemented with 10% FCS, 10 nM DEX, 100 μM ascorbic acid-2 phosphate, 100 µM, β-glycerophosphate, and 10 nM 1α, 25-dihydroxyvitamin D3) and cultured for 3 weeks. After this, cultures were rinsed with PBS, fixed with 10% formaldehyde, and stained with Alizarin Red S to visualize calcium deposits.

Chondrogenic Differentiation

MSCA1+/CD36+ and MSCA1+/CD36- cells were cultured in pellet (250,000 cells/pellet) in chondrogenic induction medium (ECBM high glucose [4.5 g/l] supplemented with 10% FCS, ITS+ (1X), 100 nM DEX, 200 μM ascorbic acid-2 phosphate, and 10 ng/ml of TGFβ3 and BMP6) and cultured for three weeks. After this, cultures were rinsed with PBS, fixed with 10% formaldehyde, embedded in paraffin and stained with Alcian Blue.

TG Quantification

Cell TG concentration was determined using a commercial TG reagent kit (Sigma-Aldrich). DNA concentration was quantified using Picogreen dsDNA quantification Kit (Invitrogen, Molecular Probes).

Global Transcriptional Profiling

Gene expression profiling in hADSCs was performed using GeneChip^® Human Transcriptome Array-2.0 (Affymetrix, Santa Clara, CA, https://www.thermofisher.com). Data were analyzed with packages available from Bioconductor (http://www.bioconductor.org). Normalization and calculation of gene expression was performed with the Robust Multichip Average expression measure using oligo package [23]. Prior to further analysis, a nonspecific filter was applied to include genes with expression signal >30 in at least 25% of all samples. Limma package [24] was used to identify differentially expressed genes. Differentially expressed genes (false discovery rate (FDR) adjusted p value < .25) between clone 1–2 and 3–4 were subjected to pathway enrichment analysis. Identified genes were interrogated for their functional classes and importance in stem cell development using PathVisio [25]. In PathVisio, the gene database Hs_Derby_Ensembl_83.bridge and the pathway collection from WikiPathways [26] were used. A permutation test was performed to calculate p values in PathVisio. Pathways with a permuted p value < .05 were significantly enriched for differentially expressed genes.

Transcriptional Analysis of Human Adipose Tissue Progenitors

Total RNA was isolated from progenitor cells and mature adipocytes using the RNeasy kit (Qiagen, Hilden, Germany, https://www.qiagen.com). cDNA synthesis was performed using Superscript^® II kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA). The amplification reaction was performed on 25 ng cDNA using the following TaqMan Primers^®: PLIN1: Hs00160173_m1; LPL Hs00173425_m1; CD36 Hs00169627_m1; ALPL Hs01029144_m1; CEBPA Hs00269972_s1; FABP4 Hs01086177_m1; PPARG2 Hs01115510_m1; PLIN2 Hs00605340_ m1; ACACA Hs00167385_m1; SLC2A1 Hs00197884_m1; SLC2A4 Hs00168966_m1; SCD Hs01682761_m1; and FASN Hs01005622_ m1. Expression was determined using the Viia7 detection system and software (Applied Biosystems, Foster City, CA, https://www.thermofisher.com). Results were normalized to 18S rRNA levels.

Immunofluorescence Staining

Cells were fixed with a 4% paraformaldehyde solution at room temperature for 10 minutes. Samples were incubated with Bodipy 493/503 (10 µg/ml, Life Technologies, Courtaboeuf, France) for 1 hour at room temperature. Nuclei were stained by 4′,6-diamidino-2-phenylindole (DAPI) or Hoechst 33342. Images in Figures 5 and 6 were taken with a fluorescent microscope (Nikon Eclipse TE300, software NIS-Elements 2.5 BR, Nikon^®, Tokyo, Japan, http://www.nikon.com). Images in Figures 1 and 3 were obtained using a Nikon A1R confocal microscope system equipped with a Nikon ECLIPSE Ti inverted microscope and 10x oil objective, under the control of NIS-Elements AR version 4 software. Images were taken with montage to cover the whole well in 96-well plates and analyzed with Image J. Nuclei and intracellular lipids were segmented based on Hoechst 33342 and Bodipy signal, respectively. The proportion of fat cells was calculated per well using the number of lipid-containing cells divided by the number of nuclei. Three replicates were performed for each condition.

Statistical Analyses

Unless otherwise stated, comparisons were performed using Student’s paired t test or analysis of variance (ANOVA). Associations were evaluated using Pearson correlations coefficient. A p value < .05 was considered statistically significant. Error bars in figures are SEM. Analyses were performed using standard software packages.

Results

Isolation of Single-Cell-Derived Clones with Adipogenic Potential

Using standard adipogenic protocols, hADSC-derived cells display a differentiation degree of around 70% indicating a heterogenous cell population (Fig. 1A). In order to isolate cell clones with more pronounced adipogenic potential, hADSCs at passage 12 from a single donor were plated at a dilution corresponding to 1 cell/well followed by several proliferation steps (a schematic representation of the cloning strategy is shown in Fig. 1B). Out of 288 individually plated wells, cells from 12 colonies could be propagated through several passages but only four showed a robust proliferation capacity at passage 16 (i.e., ≥ 90% confluency in 10 cm dishes). Following adipogenic induction at passage 17, two of the clones displayed an almost complete differentiation into adipocytes (Fig. 1C, clone 1 and 2). In contrast, two clones showed no significant differentiation capacity (Fig. 1C, clone 3 and 4) despite displaying similar doubling times as clones 1–2 (Supporting Information Fig. S1A).

Transcriptomic Analysis of Cell Clones with Differential Adipogenic Capacity

To identify genes associated with adipogenic capacity in hADSCs, global gene expression of clones 1–4 (at passage 18) was determined by microarray before and after adipocyte differentiation. As expected, a principal component analysis showed that clones 1–2 were significantly separated from clones 3–4 after differentiation (Fig. 2A). However, there were also clear differences between the clones before differentiation, suggesting that specific gene sets may associate with the capacity to differentiate into adipocytes and store TGs. A comparison of the differentially expressed genes in clones 1–2 versus 3–4 prior to differentiation, revealed 1,929 transcripts (1,136 of which were annotated to a gene symbol) at an FDR level of 0.25 (Supporting Information Table S2). Gene set enrichment analysis identified several upregulated and downregulated pathways in clones 1–2 versus 3–4 (Fig. 2B; Supporting Information Table S3). Interestingly, with the exception of PPARG (encoding PPARγ) and ALPL (MSCA1), there was no enrichment of genes linked to adipogenesis in the clones with strong adipogenic capacity (Supporting Information Table S2). In order to identify cell surface markers linked to adipogenic potential, we focused on individual genes encoding cell surface proteins that were significantly higher expressed in clones 1–2 versus 3–4 before differentiation. CD36 was the strongest candidate fulfilling these criteria. Analysis of gene array data confirmed that CD36 was markedly higher in clone 1–2 both before and after differentiation (Fig. 2C, lower panels). In comparison, the differences in ALPL (MSCA1) expression, albeit significant, were quantitatively less pronounced between the clones (Fig. 2C, upper panels). Importantly, the differences in adipogenic capacity between the clones were not secondary to cellular senescence as they displayed similar expression levels of markers such as CDKN1A and CDKN2A [27] (Supporting Information Fig. S1B).

CD36 Expressing hADSC Cells Display High Adipogenic Capacity

The notion that CD36 may be a marker for precursor cells with marked adipogenic/TG storage capacity was further tested by performing FACS of total hADSCs at passage 13 using antibodies directed against several proposed adipocyte precursor markers [10, 28, 29] (Supporting Information Fig. S2A). This revealed that most markers were rather uniformly expressed in hADSCs except CD36 which could divide hADSC into a population that was negative (76.3%) and one that was positive (16.0%) (Supporting Information Fig. S2A). Fractionating the hADSCs solely based on CD36 (Supporting Information Fig. S2B) confirmed that the CD36+ hADSC displayed a significantly more pronounced adipogenic differentiation than CD36- cells determined by gene expression (Fig. 3A, including PPARγ-target genes ADIPOQ, PLIN1, and PPARG itself), FABP4 protein levels (Fig. 3B, lower left panel) and lipid accumulation (Fig. 3B, upper and lower right panel). CD36 mRNA increased significantly during in vitro differentiation in both CD36+ and CD36- cells albeit from different levels. This is in accordance with previous data demonstrating that CD36 is a PPARγ-response gene [30]. Still, CD36 protein levels were considerably lower in CD36- cells at the end of differentiation (Fig. 3B, lower left panel).

To assess whether the differences between CD36+ and CD36- cells were linked to intrinsic variations in PPARγ activation state, nonsorted hADSC (at passage 13) were transfected with siRNAs, targeting CD36 or random sequences, followed by adipogenic induction. While the reduction in CD36 expression was pronounced with siRNAs (∼90% at the induction of differentiation, Supporting Information Fig. S3), CD36 mRNA levels increased during differentiation under both conditions albeit at different levels (Fig. 3C). At the end of differentiation, the expression of CD36 mRNA and protein was markedly lower in cells transfected with siRNA targeting CD36 (Fig. 3C, 3D, lower left panel). In contrast, gene expression of PPARγ and PPARγ-target genes as well as protein expression of FABP4 in CD36-silenced cells was not different from that observed in control cells. Nevertheless, with reduced CD36 expression, TG accumulation (Fig. 3D, upper and lower right panel) and glycerol release (Fig. 3E, right panel), but not adiponectin secretion (Fig. 3E, left panel), was significantly attenuated. Altogether, this suggests that CD36 is important for TG accumulation but does not in itself affect the intrinsic activity of PPARγ.

CD36 Expression on Human Native APs Is Linked to Higher FA Transport

Our results so far were based on hADSC propagated for several passages which in theory could affect gene expression and thereby the levels of cell surface proteins. To characterize CD36 expression on the native AP population, we performed flow cytometry analyses on SVF from subcutaneous WAT obtained from healthy non obese women (GLUTAB cohort, see methods). The total pool of progenitor cells, identified as CD45-/CD34+/CD31- cells (Fig. 4A), was further subdivided according to cell surface expression of MSCA1 (Fig. 4B) [14]. CD36 surface expression on native progenitor cells was four times higher in the MSCA1+ compared to the MSCA1- cells (Fig. 4C). Subdividing the MSCA1+ cells into CD36+ or CD36- (Fig. 4D) revealed that there was no difference in the cell surface expression of MSCA1, CD34, or CD271 between the two fractions. We next analyzed the expression of FATs (FATP1 and FATP4) on human subcutaneous APs (CD45-/CD34+/CD31-) by flow cytometry approaches. The CD36+ cells expressed 1.5 fold more FATP1 and 1.4 fold more FATP4 than the CD36- cells (Fig. 4E). To assess whether this could confer any differences in lipid storage capacity, we set up a FA uptake assay. When subcutaneous APs were incubated with fluorescent C12-FA (FLC12) at 4°C, fluorescence detected by flow cytometry was low and unaltered over time suggesting nonspecific staining (Fig. 4F). In contrast, at 37°C, a significantly higher and time-dependent increase in FA uptake was observed (Fig. 4F). Treatment with SSO, an inhibitor of CD36 [31], led to a marked reduction in CD36 cell surface expression and a concomitant attenuation of FA uptake (Fig. 4G). When analyzing all experiments in APs together, there was a significant positive correlation between CD36 surface expression and FA uptake (Fig. 4H).

CD36 Expression In Situ Is Linked to FA Uptake and Terminal Differentiation

To investigate the role of CD36 in FA handling in situ, subcutaneous WAT explants were incubated with pro-lipolytic agents (ANP or isoprenaline) resulting in an increased local release of FAs. After 24 hours, FA levels were determined in conditioned media. As expected, compared to control conditions, FA levels were higher in the presence of ANP or isoprenaline where the latter resulted in the most pronounced increase (Fig. 5A). Stromal-vascular cells (CD45-/CD31-/CD34+) were recovered from the explants and analyzed by flow cytometry. In the presence of lipolytic agents, APs exhibited a marked shift in their granularity assessed by side scatter (SSC, Fig. 5B), a marker of lipid uptake as shown by TG staining using Bodipy (Fig. 5C). Lipid accumulation into APs associated with cell surface expression of MSCA1 and CD36 (Fig. 5D) with a positive correlation between CD36 expression and Bodipy staining (Fig. 5E). Profiling AP subsets according to SSC showed that isoprenaline induced a marked decrease in cells with low granularity/lipids and an increase in cells with high granularity/lipids (Fig. 5F). Cells from ANP-treated explants displayed an intermediate SSC distribution (Fig. 5F). In order to determine whether FA uptake affected terminal differentiation, stromal-vascular cells isolated from explants incubated with or without isoprenaline were maintained in vitro under basal adipogenic condition or primed for 3 days with a PPARγ-agonist (rosiglitazone). After 7 days of incubation, an increased number of lipid laden cells were observed in isoprenaline-exposed cells compared with control conditions, both with or without rosiglitazone (Fig. 5G). Taken together, this suggests that an increased proportion of lipid-containing APs (expressing MSCA1/CD36) results in increased terminal differentiation.

Mesenchymal Differentiation Capacity of MSCA1+/CD36+ and MSCA1+/CD36- Progenitor Cells

The adipogenic potential of MSCA1+/CD36+ and MSCA1+/CD36- cells was determined by transducing native immunoselected MSCA1+ progenitor cells, from human subcutaneous WAT of three separate donors, with human telomerase and human papilloma virus E6/E7 genes. Transduced MSCA1+ cells were then sorted according to the expression of CD36 followed by in vitro expansion for further characterization. As expected, CD36 expression levels by qPCR were 11 fold higher in the MSCA1+/CD36+ compared to MSCA1+/CD36- cells while ALPL was similar between the two cell subsets (Fig. 6A). In the undifferentiated state, the gene expression of adipogenic proteins such ADRP (PLIN2) and C/EBPα (CEBPA) as well as the glucose transporter-1 (GLUT1, encoded by SLC2A1) and de novo lipogenesis and FA modification pathway enzymes including acetyl-CoA carboxylase alpha (ACACA), stearoyl-CoA desaturase (SCD) and fatty acid synthase (FASN) were similar in MSCA1+/CD36+ and CD36- cells while PPARγ2, FABP4, GLUT4 (SLC2A4), and perilipin (PLIN1) were very low or nondetectable in both fractions (Fig. 6A). Following adipogenic induction, there was a clear difference in adipogenic potential depending on the presence of CD36. The number of lipid-accumulating cells assessed by Bodipy staining or intracellular TG quantification was markedly higher in the MSCA1+/CD36+ compared to MSCA1+/CD36- (Fig. 6B, upper panels) while a mixture of both CD36+ and CD36- MSCA1 subsets (in a 1:1 proportion) displayed an intermediate phenotype (data not shown). The expression of the adipogenic markers LPL and PLIN1 determined in cells cultured under adipogenic conditions were significantly and positively correlated with the initial expression of CD36 in the undifferentiated state (Fig. 6B, lower left panels). In contrast, up-regulation of GLUT4 and genes encoding enzymes involved in de novo lipogenesis (ACACA, SCD, and FASN) as well as a concomitant downregulation of GLUT1, was independent of CD36 expression (Fig. 6B, lower right panel). To assess whether the presence or absence of CD36 impacted on the differentiation capacity into other mesenchymal lineages, MSCA1+/CD36+ and MSCA1+/CD36- transduced progenitor cells were cultured under chondrogenic and osteogenic conditions. Using established staining protocols, the amount of acidic polysaccharide (chondrocytes) or calcium (osteocytes) deposits were indistinguishable between CD36+ and CD36- MSCA1 subsets (Fig. 6C). Overall, these results suggest that differences in CD36 expression only affect adipogenic differentiation and that this appears to be independent of alterations in de novo lipogenesis pathways.

The Proportion of MSCA1+/CD36+ Cells Is Fat Depot-Specific in Non-Obese Women

To assess the relevance of the MSCA1+/CD36+ progenitor cells in vivo, flow cytometry analyses were performed on native SVF cells from paired subcutaneous femoral and abdominal adipose tissues (n = 17) and paired subcutaneous abdominal and visceral fat depots (n = 24) from non-obese women. MSCA1 cell surface expression on APs (defined as CD45-/CD34+/CD31-) was significantly higher in femoral and visceral compared to abdominal WAT (Fig. 7A). Similar results were observed regarding the proportion of MSCA1+ APs (Fig. 7B). Interestingly, CD36 surface expression exhibited a distinct profile with the highest expression in femoral APs and the lowest in visceral cells (Fig. 7C). Moreover, the largest proportion of MSCA1+/CD36+ cells was found in femoral WAT while the lowest was observed in the visceral depot (Fig. 7D). Thus, CD36 expression is highest in APs from WAT depots associated with lower cardiometabolic risk.

Discussion

The aim of this work was to identify specific markers for human APs associated with pronounced adipogenic differentiation and TG storage capacity. Using an unbiased single cell approach, we found that WAT-derived precursor cells (hADSCs) expressing CD36 displayed a good proliferation capacity and a remarkable propensity to undergo adipocyte differentiation and lipid accumulation. These findings were corroborated by analyses of primary cells from human WAT suggesting that the findings in hADSCs were not secondary to the in vitro culture process. CD36 can therefore serve as a marker for human APs with a particular susceptibility to undergo terminal adipogenic differentiation, a phenotype which appears to depend on increased FA uptake. The observation that there was a regional variation in the proportion of CD36+/MSCA1+ positive cells in different WAT depots indicates that CD36 expression may be linked to metabolic phenotypes involving WAT distribution.

CD36 is membrane glycoprotein that belongs to a larger family of scavenger receptors currently subdivided into eight subclasses [32]. While scavenger activity was first defined as the ability to internalize oxidized low-density lipoproteins, results in the last decades have demonstrated that these proteins recognize and internalize an array of different endogenous and exogenous factors with subsequent impact on both metabolism and immune function. CD36 displays a broad expression and is present in, for example, platelets, endothelial cells, monocytes/macrophages, fibroblasts, and cardiomyocytes [33]. The exact role of CD36 in different cell types is still elusive but recent investigations have identified a number of endogenous ligands including thrombospondin-1, long chain fatty acids, oxidized lipid products, and collagen that upon binding induce multiple signaling pathways. In both adipocytes [34, 35] and macrophages [36, 37] (as well as several other cell types), CD36 has been shown to be important for lipid uptake (hence the alternative name FA translocase [33]). It has also been implicated in glucose uptake and humans with CD36-deficiency display a phenotype of hyperlipidemia and insulin resistance [38]. The mechanisms through which CD36 mediates transmembrane transport are not understood as CD36 does not display a classic structure for transporter proteins. Studies primarily in non-adipose cells suggest that other membrane proteins, associated with CD36 may account for this (including FATPs), but how these membrane complexes are assembled, particularly in different cell types is currently unknown.

CD36 is one of several established markers of adipocyte differentiation [39] and CD36-deficiency in mice attenuates adipocyte differentiation, size, lipid accumulation, and increases hepatic TG content as well as insulin resistance [40]. Our present data extends these results by demonstrating that CD36 expression in immature precursor cells associates with robust adipogenic differentiation without affecting multilineage capacity in vitro. Thus, in both passaged hADSCs and primary APs in vitro, CD36 expression was positively correlated with lipid accumulation while CD36 inhibition by either RNAi or chemical compounds resulted in reduced FA uptake. How does CD36 mediate these effects? Some insight can be obtained by understanding the mechanisms promoting TG accumulation in vitro. Using defined adipogenic media (without the addition of exogenous lipids), lipid storage is a complex process which depends on metabolic changes related to the acquisition of the adipocyte phenotype. At the initial phase of differentiation, the FA moieties in TGs originate from de novo lipogenesis and FA modification pathways while the glycerol moiety of TGs is produced by glycolysis which in turn depends on glucose uptake [41]. Later during the differentiation process, additional pathways impact on lipid content, for example, recycling pathways involving lipolysis and subsequent re-esterification through increased FA uptake [41]. As we did not observe any differences in the gene expression of key de novo lipogenesis enzymes and glucose transporters between MSCA+/CD36+ and MSCA+/CD36- cells either before or after differentiation (Fig. 6A, 6B), CD36 does not appear to affect de novo lipogenesis and/or glucose transport. In contrast, CD36 expression in native APs associated with increased expression of FATs (Fig. 4). The role of CD36 in FA uptake was further demonstrated ex vivo, using WAT explants (Fig. 5). In situ, human APs are present in a microenvironment containing non-esterified FAs originating from either lipoprotein hydrolysis or adipocyte lipolysis. Thus, adipogenic differentiation of these cells is not necessarily dependent on de novo lipogenesis. Lipolysis stimulation, resulting in pronounced release of endogenous FAs stored in mature adipocytes, lead to increased lipid storage in APs with high cell surface expression of MSCA1 and CD36. This in turn, facilitated subsequent differentiation of lipid-laden adipocytes in vitro. Taken together, our data therefore suggest that CD36 promotes TG accumulation primarily through FA uptake.

The observation that CD36 gene knockdown in hADSCs did not affect the expression of adipogenic genes, supports a notion wherein CD36 is a marker of APs with strong adipogenic capability but that CD36 per se does not affect the expression of PPARγ-target genes. Admittedly, the fact that CD36 is a PPARγ-responsive gene implied that rosiglitazone (a PPARγ agonist present in the adipogenic medium) increased CD36 also in knockdown cells which could impact our interpretation on the role of CD36 in adipogenesis. Thus, complete knockout with, for example, CRISPR/Cas9 would be of interest, techniques that are not yet established in hADSCs.

Our analyses of primary SVF cells in human WAT depots demonstrated that CD36+/MSCA1+ progenitor cells were particularly abundant in the femoral region and significantly lower in visceral WAT. An 8-week long overfeeding study in healthy subjects reported that adipocyte hyperplasia (an indirect sign of adipogenesis) only occurred in the lower- (i.e., femoral/gluteal) but not in the upper-body (i.e., subcutaneous abdominal) fat region [6]. It is therefore tempting to speculate that this may, at least in part, depend on the higher proportion of the CD36+/MSCA1+ progenitor cells in femoral WAT. This notion would of course need to be determined in controlled studies where the proportion of CD36+ cells is determined and set in relation to changes in fat cell number and WAT depot growth. The observation that single nucleotide polymorphisms in CD36 are linked to obesity in adolescents [42], lends further support to the notion that CD36 expression may play a patophysiologically important role in WAT expansion. Nevertheless, it is still a matter of debate whether there are intrinsic differences in APs between different depots [43, 44] or if it is the regional microenvironment that influences the phenotype [45].

The adipose expression of CD36 has primarily been studied in intact WAT where increased mRNA and protein levels have been reported in obese and/or type 2 diabetes subjects [46–49]. However, given the differential expression of CD36 in specific cell types, it is clear that analyses in whole tissue lysates provide little insight into the role of CD36 in WAT. It is therefore interesting that a recent study in 53 nondiabetic obese subjects showed that WAT expression of adipocyte-specific CD36 transcripts was positively correlated with insulin sensitivity and FA homeostasis [50]. This underlines the importance of determining CD36 expression in specific cell types and implies that efforts to link CD36 expression to metabolic outcome must include determination of CD36 expression (and preferably function) in adipocytes, progenitors, macrophages, and possibly other cells present in WAT.

Most adipose in vitro studies are performed on immortalized murine cell lines such as 3T3-L1. Because there is no available human adipocyte cell line, research is confined to primary SVF cells or hADSCs (including SGBS cells) which display variable degrees of differentiation and cannot be passaged indefinitely. If cellular markers of highly adipogenic cells can be identified, this could facilitate isolation of progenitor cell populations with very high differentiation capacity and possibly enable future development of immortalized human adipocytes. The presently described isolation procedure of SVF cells and hADSCs based on CD36 expression appears to be suitable for this purpose.

Taken together, our work identifies CD36 as a marker of human APs with high adipogenic potential. We suggest that selection of human preadipocytes from the heterogeneous mixture of SVF cells should include CD36 in order to maximize adipogenic induction efficiency and shorten selection time. The pathophysiological relevance of CD36 in human preadipocytes in vivo needs to be evaluated in future studies.

Acknowledgments

This work was supported by grants from the INSERM, Région Midi-Pyrénées, Swedish Research Council, The Swedish Diabetes Foundation, the Diabetes Theme Center at Karolinska Institutet, the Stockholm County Council, the Novo Nordisk Foundation including the Tripartite Immuno-metabolism Consortium (TrIC), Grant Number NNF15CC0018486 and the MSAM consortium NNF15SA0018346 as well as in part by a grant from Sanofi R&D. We acknowledge the support of the transcriptomics, histology and flow cytometry facilities and the technical support of Pauline Decaunes and Chloé Belles from I2MC and the Bioinformatics and Expression Analysis core facility at KI. Parts of this study was performed at the Live Cell Imaging unit, Department of Biosciences and Nutrition, Karolinska Institutet, Sweden, supported by grants from the Knut and Alice Wallenberg Foundation, the Swedish Research Council, the Centre for Innovative Medicine and the Jonasson donation to the School of Technology and Health, Royal Institute of Technology, Sweden. We thank Dr. Christel Björck for taking the images in Fig. 1A. Hui Gao is supported as a Naomi Berrie fellow in Diabetes research.

Author Contributions

A.B., J.G., H.G., and M.R.: Conception and design, provision of study material, collection and assembly of data, data analysis and interpretation, manuscript writing and final approval of the manuscript. D.E., J.C.G., N.M., H.Q., L.S., F.V., and G.Å.: Collection and assembly of data, final approval of the manuscript. P.A., T.N., S.L., and C.T.: Provision of clinical study material, final approval of the manuscript. A.B. and M.R.: Financial support and guarantors of the data.

Disclosure of Potential Conflicts of Interest

The authors indicated no potential conflicts of interest.

References

Author notes