Endothelial Cells From Visceral Adipose Tissue Disrupt Adipocyte Functions in a Three-Dimensional Setting: Partial Rescue by Angiopoietin-1

Vascularization is critical for white adipose tissue (WAT) development and homeostasis by providing oxygen and nutrients to this key organ (1). In close interactions with adipocytes, the capillary network contributes to the export of adipokines as leptin and adiponectin to targeted tissues. Endothelial cells (ECs) and pericytes are the main cell components of blood vessels. These mural cells, through direct interactions with ECs, promote the maturation and stabilization of blood vessels (2).

Obesity is characterized by a chronic and auto-inflammation of hypertrophied WAT. The inflamed WAT is the site of cellular remodeling, including adipocyte hypertrophy and altered immune cell and EC phenotypes. In turn, this remodeling leads to increased secretion of numerous inflammatory mediators, including interleukin (IL)-6, chemokine ligand (CCL)-2, IL-8, and CCL5, some of which are implicated in the development of insulin resistance (3).

The accumulation of visceral fat depots (visceral adispose tissue [VAT]) is associated with metabolic syndrome and adverse obesity comorbidities (4). VAT is more inflammatory and vascularized than subcutaneous depots (subcutaneous adipose tissue [SAT]) (5,6). ECs from VAT (VAT-ECs) play a key role in the inflammatory process, exhibiting a senescent phenotype and expressing inflammatory and angiogenesis-related molecules (5). Though adipocytes have been described to participate in the altered phenotype of ECs during obesity (5,7–9), the retroactive influence of ECs on adipocyte functions remains poorly described. Most studies have highlighted the requirement for ECs in adipose tissue growth playing a role in the proliferation and differentiation of adipocyte precursors (10,11). Thus, evaluating the impact of ECs on the metabolism of unilocular adipocytes is of particular importance in pathophysiological situations such as obesity.

Here, we focus on direct interactions between human adipocytes and ECs isolated from VAT of obese subjects by coculturing these cells in a three-dimensional (3D) setting. This is the first report showing an impact of VAT-ECs on adipocyte functions, identifying IL-6 and IL-1β as key factors in these alterations. Then, we propose an experimental strategy to reverse these effects by improving EC inflammation with Ang-1 treatment, since this molecule, constitutively produced by pericytes, is known to prevent EC inflammation (12).

Antibodies used in the study are listed in Supplementary Table 1. The human recombinant proteins IL-6 and IL-1β were obtained from Peprotech (Rocky Hill, NJ).

Adipocytes, ECs, and pericytes were isolated from WAT of 29 lean (SAT) and 87 obese subjects (SAT and VAT, omental depot) whose clinical parameters are described in Supplementary Table 2. SAT from lean subjects were obtained after elective surgery. All clinical investigations were performed according to the Declaration of Helsinki and approved by the ethics committee of Hôtel-Dieu Hospital (Paris, France). In an accepted protocol related to the pathophysiology of low-grade inflammation in obesity (Assistance Publique/Hôpitaux de Paris, Clinical Research Contract), obese subjects (mean BMI >30 kg/m²) were candidates for gastric surgery programs as previously described (6).

WAT biopsies were dissociated by collagenase treatment isolating unilocular adipocytes from the stromavascular fraction (SVF). The adipocytes were then washed three times in 10% sucrose solution. The contamination of adipocyte preparation by preadipocytes was <5% as assessed by the expression of Pref-1/Dlk1, a preadipocyte marker. WAT-ECs were isolated from the SVF of the same WAT biopsies as previously described (13) using immunoselection cocktails (EasySep Do-It-Yourself Selection kit; Stemcell Technologies, Grenoble, France) for the endothelial marker CD31 (M0823 DAKO). The endothelial phenotype of CD31-positive cells was verified by immunofluorescence analysis of an endothelial-specific marker, von Willebrand Factor (vWF) (Supplementary Fig. 1A), and tube formation assay on Matrigel-GF reduced (BD Biosciences, Franklin Lakes, NJ) (Supplementary Fig. 1B). The purity of the WAT-ECs was ~90% with an absence of macrophage contamination, as verified by measuring CD14 expression. To isolate pericytes, the CD31-negative cell fraction was incubated with NG2-positive selection cocktail (ab83508; Abcam). The phenotype of WAT pericytes was verified by comparing the expression of CD146 and RGS5 mRNA, two pericyte markers (14,15), among other cell types (WAT-ECs, preadipocytes, and skin fibroblasts [Supplementary Fig. 2A and B]) and by immunofluorescence analysis of the expression of pericyte markers chondroitin sulfate proteoglycan (NG)-2 and α smooth muscle actin (αSMA). NG2⁺ cells were distinguished from CD31⁺ cells in the WAT SVF (Supplementary Fig. 2C). The proportion of pericytes (CD31⁻NG2⁺) to ECs (CD31⁺) in SAT of lean and from SAT/VAT of obese subjects was evaluated by counting the corresponding cells.

Adipocytes and ECs both isolated from VAT of obese subjects or SAT of lean subjects were cocultured in the 3D setting. The difficulty getting access to VAT from lean subjects prevented us from performing coculture experiments. After sonication for 30 min to decrease viscosity, the hydrogel (Puramatrix; BD Biosciences) was diluted in 20% sucrose solution according to the manufacturer’s recommendations. Adipocytes were embedded in this gel at a concentration of 1 × 10⁴ cells/100 µL gel preparation into 96-well plates containing 150 µL endothelial cell basal medium (ECBM) (PromoCell, Heidelberg, Germany), 1% BSA, 1% antibiotics, and human insulin (50 nmol/L). The CD31⁺ cells isolated from the same individual were then incorporated into 96-well plates containing 3D hydrogel/adipocytes in the ratio of 2 × 10⁴ cells for 1 × 10⁴ adipocytes, which is representative of the proportion of ECs in the WAT vasculature (16). (See detailed protocol in Supplementary Fig. 3.) The culture medium was changed every other day and kept at −80°C for experimental measurements. Additional details can be found in the international patent (PCT/IB2011/052241) (17). The VAT-ECs were prone to migrating into the hydrogel in response to VAT adipocytes and then interacting directly. ECs also maintained their inflammatory profile in the 3D setting (Supplementary Fig. 4A–C).

A portion of VAT biopsies and VAT adipocytes/ECs/pericytes was fixed in 4% paraformaldehyde and processed for immunofluorescence analysis. Samples were incubated with the appropriate primary antibody and then the corresponding anti-IgG. The samples were examined with an OlympusBX41 fluorescence microscope (Olympus, Lake Success, NY) or Zeiss 710 confocal laser-scanning microscope (Carl Zeiss, Thornwood, NY).

ECs isolated from SAT of lean and SAT/VAT of obese subjects were assessed for senescence-associated β-galactosidase activity (SA-β-gal) according to the company’s instructions (Sigma, St. Louis, MO) and as previously described (18). Five phase-contrast images were recorded on a digital camera (Olympus, Tokyo, Japan). The number of SA-β-gal–positive cells was normalized to the total cell number counterstained with hematoxylin-eosin.

The adipocytes/hydrogel were incubated in Krebs-Ringer bicarbonate buffer supplemented with 3% BSA for 4 h at 37°C. The release of glycerol (R-Biopharm, Marshall, MI) and nonesterified fatty acids (Randox Laboratories, Antrim, U.K.) was evaluated after stimulation with forskolin (10 µmol/L), dibutyryl-cAMP (DcAMP) (0.5 mmol/L), and isoproterenol (1 µmol/L) (Sigma).

Conditioned media obtained from different experimental conditions of adipocytes/WAT-ECs cocultured for 3 days in the 3D setting were analyzed using human cytokine/chemokine Panel I 39 plex and human from Millipore according to the manufacturer’s instructions. Multianalyte profiling was performed on the Luminex-200 system and Xmap Platform (Luminex, Austin, TX). Acquired fluorescence data were analyzed by Xponent software, version 3, using standard curves obtained with serial dilutions of standard cytokine mixtures (as previously described [18]). The detection threshold was fixed as >2 pg/mL. Heat maps were created using MEV MultiExperiment Viewer software (version 4.8; MeV, Boston, MA). Colorimetric ELISA kits were used to determine the concentrations of adipokines (leptin and adiponectin; Duoset, R&D Systems, Minneapolis, MN) and inflammatory factors (IL-6, G-CSF, and CCL2, Duoset, R&D Systems, and IL-8 and CCL5, Peprotech) in the medium from 3D adipocyte/EC cocultures according to the provider’s instructions.

The adipocyte/hydrogel was insulin deprived overnight. Next, the cells were stimulated with 10 nmol/L insulin for 15 min at 37°C. Insulin sensitivity was evaluated by the level of Ser⁴⁷³ phosphorylation of Akt (pS⁴⁷³ Akt) normalized to total Akt.

Cell extracts were prepared as previously described (19). Protein samples were separated by SDS-PAGE and blotted on nitrocellulose transfer membranes (GE Healthcare, Little Chalfont, U.K.). The membranes were probed overnight at 4°C with the corresponding primary antibodies. Specific signals were detected using the ECL detection solution (GE Healthcare) and immediately exposed to X-ray films. The signals were quantified by densitometry.

VAT-ECs (8 × 10⁴ cells) from obese subjects were cultured in 1 mL ECBM with 1% BSA for 7 days at 37°C (medium was changed twice). After washing, VAT-ECs were placed in 800 µL ECBM with 1% BSA for 24 h at 37°C, after which conditioned medium was collected and stored at −80°C until needed. VAT adipocytes cultured in hydrogel were treated for 5 days with conditioned medium prepared from VAT-ECs. After washing, adipocytes were placed in fresh medium for 48 h at 37°C, which was collected and stored at −80°C.

Cytotoxicity was evaluated by measuring the activity of lactate dehydrogenase (LDH) released from damaged cells according to the manufacturer’s instructions (LDH-Cytotoxicity Assay kit; BioVision Research Products, Mountain View, CA).

Adipocytes, ECs, and cocultured cells from VAT of obese subjects were treated with IgG1 (3 µg/mL) or with IL-6 (2.5 µg/mL) and IL-1β (0.5 µg/mL) neutralizing antibodies or with tumor necrosis factor (TNF)-α (0.5 µg/mL) neutralizing antibody during 3 days with every day changes. Recombinant IL-6 (10 ng/mL) and IL-1β (1 ng/mL) were incubated with adipocytes, ECs, and cocultured cells from SAT of lean subjects and cultured in the hydrogel during 3 days with everyday changes.

Isolated adipocytes (1 × 10⁴ cells) and ECs (2 × 10⁴ cells) from VAT were introduced in separate hydrogels (100 µL in 96-well plates) in control culture medium. ECs were incubated in the presence or absence of Ang-1 (100 ng/mL; R&D Systems) for 24 h at 37°C. Adipocytes and Ang-1–treated ECs were cocultured by associating the two hydrogels in control medium. The metabolic and secretory functions of adipocytes were then studied. (See detailed protocol in Supplementary Fig. 5.)

RNA extraction, reverse transcription, and real-time PCR were performed as previously described (20). The primers that were used are listed in Supplementary Table 3. All values were normalized with regard to 18S expression.

The experiments were performed at least five times using adipocytes and WAT-ECs from different obese and lean subjects. Statistical analyses were performed using GraphPad software (San Diego, CA). Values are expressed as means ± SEM. Comparisons between two conditions were analyzed using the Wilcoxon nonparametric paired test (adipocytes and adipocyte/EC cultures) and the Mann-Whitney nonparametric test (lean vs. obese pericytes). Comparisons among more than two groups were performed using one-way ANOVA followed by post hoc tests. Spearman coefficients were calculated to examine correlations. Differences were considered significant when P < 0.05.

Senescence level of WAT-ECs was evaluated in lean (SAT) and obese (SAT and VAT) subjects. Although no significant difference among depots were observed in WAT of obese subjects, an increased β-gal activity was detected in VAT-ECs from obese subjects compared with SAT-ECs of lean subjects (Fig. 1A). Moreover, VAT-ECs exhibit an increased secretion of several cytokines and chemokines compared with lean individuals (Fig. 1B). Based on these observations, we attempted to identify the potential impact of ECs on the metabolic and secretory properties of adipocytes in VAT from obese subjects. We cocultured adipocytes and ECs isolated from the VAT of the same individual to reproduce this inflammatory environment. Cocultures experiments with noninflammatory SAT-ECs were also performed as a control.

To have a longer follow-up of biological activities, we previously developed an original 3D culture of human adipocytes using a self-assembling peptidic hydrogel device (17) (Supplementary Fig. 6 [international patent PCT/IB2011/052241]).

We investigated the influence of VAT-ECs from obese subjects on adipocyte secretions in this 3D setting. From the third day of coculture, the VAT-ECs provoked a decreased secretion of the adipokines leptin (−50%, n = 6, P < 0.01) and adiponectin (−20%, n = 6, P < 0.001) (Fig. 2A). Lipolytic activity was also altered with decreased production of glycerol and nonesterified fatty acid using different stimulators: isoproterenol (1 µmol/L) (∼30%, n = 6, P < 0.001), forskolin (10 µmol/L) (∼30%, n = 6, P < 0.001), and DcAMP, the nonhydrolyzable cAMP analog (0.5 mmol/L) (∼20%, n = 5, P < 0.01) (Fig. 2B; Supplementary Fig. 7A). Next, we explored the possibility that ECs alter insulin sensitivity. The ratio of pS⁴⁷³ AKT to total AKT was decreased in adipocytes/ECs cocultures (−41.7%, n = 4, P < 0.05) (Fig. 2C; Supplementary Fig. 7B). We also observed that gene expression of three key markers of ER stress, ATF4, HSPA5, and CHOP, was increased (∼1.5-fold, P < 0.05) after 3 days of coculture (Fig. 2D). However, alterations in adipocyte functions were not associated with the cytotoxicity induced by ECs, as indicated by an unchanged LDH activity in cocultures compared with cells cultured alone (Supplementary Fig. 8).

Using multiplex and ELISA assay, we quantified the secretion of inflammation-related molecules in cocultures. Among the 39 screened molecules, 26 were significantly detected in adipocytes and ECs (>2 pg/mL) (Fig. 3A). CXCL_1/2/3, IL-8, IL-6, and G-CSF were highly secreted in adipocytes. The significant changes in the secretion of the inflammatory molecules are presented for each experimental condition in Fig. 3B. For most of the molecules, secretion was enhanced in the presence of adipocytes and ECs in an additive manner. However, we observed a synergistic effect on CXCL_1/2/3 (6-fold, P < 0.01), IL-6 (7.6-fold, P < 0.01), IL-8 (3-fold, P < 0.05), G-CSF (3.8-fold, P < 0.01), CCL2 (5.7-fold, P < 0.01), fractalkine (4.3-fold, P < 0.05), and γ-interferon (IFN0γ) (5.6-fold, P < 0.05). Increased release of soluble forms of the adhesion molecules intracellular adhesion molecule-1 (4.4-fold, P < 0.01) and E-selectin (3.3-fold, P < 0.05) by cocultured cells was also observed (Fig. 3B). However, the cell origin of these inflammatory mediators could not be identified in our coculture experiments exploring paracrine interaction, but the observation strongly suggests a direct effect of VAT-ECs on adipocyte function. In fact, cocultures associating adipocytes with an increased number of VAT-ECs exhibited a strong dose-dependent effect on adipocyte functions (lipolytic activity and adipokine secretions) and the secretion of inflammatory mediators (IL-6 and IL-8) (Supplementary Fig. 8).

We emphasized the benefits of our 3D experimental design for direct coculture of adipocytes and VAT-ECs. The conditioned media prepared from VAT-ECs had no effects on adipocytes similar to those of direct cocultures. The inflammatory response of adipocytes to conditioned VAT-EC media remained lower than the response in direct cocultures and failed to induce alterations in adipokine secretion (Supplementary Fig. 9). Taken together, these findings support a putative role of labile factors or direct cell contact in adipocyte dysfunction.

To identify the potential factors involved in the inflammatory response and adipocyte alterations, we tested the effects of neutralizing antibodies of two candidates, IL-6 and IL-1β, in adipocytes and VAT-ECs cocultured or alone in the 3D setting. After 3 days of culture, the presence of the neutralizing IL-6/IL-1β antibodies significantly rescued some deleterious effects induced by cocultures as lipolytic activity (n = 5, P < 0.01) (Fig. 4A) and leptin secretion (n = 7, P < 0.01) (Fig. 4B). Insulin-stimulated Akt phosphorylation was improved by this treatment (n = 5, P < 0.05), while the presence of neutralizing TNF-α antibody failed to rescue the insulin response (Fig. 4C; Supplementary Fig. 10A). In the presence of the neutralizing IL-6/IL-1β antibodies, the inflammatory response of cocultured cells was reduced with decreased levels of IL-6 level (−84.8%, P < 0.001), G-CSF (−79%), and chemokines (CXCL_1/2/3, −76%, and IL-8, −67.8%) (n = 7, P < 0.001) (Fig. 4D). However, benefits of IL-6 and IL-1β neutralizing antibody treatments on inflammation and adipocyte functions could be attributed to an effect on both adipocytes and VAT-ECs. Indeed, these treatments 1) tended to increase leptin production by adipocytes and 2) decreased G-CSF secretion by either adipocytes or VAT-ECs cultured alone in the 3D setting (−32% and −49%, respectively; P < 0.001) (Supplementary Fig. 10B–D). Importantly, no significant effects were observed with a treatment of neutralizing IL-6 antibody alone (data not shown), suggesting the importance of a combined action of the two cytokines in WAT inflammation and dysfunctions. Taken together, these results underlined important effects of these inflammatory molecules overproduced in cocultures on inflammation and adipocyte dysfunctions.

To establish the importance of the inflammatory environment induced by WAT-ECs on adipocyte dysfunctions, we cocultured adipocytes and ECs isolated from the SAT of lean nondiabetic subjects who did not display an inflammatory/senescent profile. Under these conditions, the secretion of leptin and adiponectin and lipolytic activity was not altered (n = 4) (Fig. 5A and B). Only 10 molecules were significantly detected by the multiplex analysis for which secretion remained much lower than cocultures with cells from VAT (n = 6) (Fig. 5C; Supplementary Figure 11). Although CCL7 and MIP-1α chemokines were induced in SAT cocultures, a synergistic effect of VAT cocultures on CXCL_1/2/3, IL-6, IL-8, G-CSF, and CCL2 secretions was not observed in SAT cocultures. Compared with cells from obese subjects, lean cocultures did not exhibit enhanced secretion of soluble adhesion molecules (data not shown). We next confirmed the role of the cytokines IL-6 and IL-1β as key factors in alterations of cocultured adipocytes. Treatment of cocultures by the recombinant cytokines are efficient to induce the inflammatory burst and lipolysis alterations, as observed in cocultured VAT cells (Fig. 6).

Considering the importance of the inflammatory profile of VAT-ECs in adipocyte dysfunctions, we proposed to reverse these alterations directly by influencing VAT-ECs. Pericytes are mural cells that promote the maturation and stabilization of blood vessels, in particular through the secretion of Ang-1, exerting prosurvival and anti-inflammatory actions on ECs (12). Therefore, we broadened the topic of this study by analyzing pericyte phenotype in VAT and tried to determine a possible link with EC dysfunction in the context of obesity.

The pericytes, visualized using the specific marker NG2 in WAT from obese subjects (Fig. 7A), were isolated from SAT of lean and from SAT/VAT of obese subjects. The pericyte-to-EC ratio was evaluated in the different conditions. This ratio was strongly reduced in SAT and VAT of obese subjects compared with lean SAT (Fig. 7B). Pericytes from obese WAT displayed an inflammatory profile with increased secretion of G-CSF (∼60-fold), CXCL_1/2/3 (∼10-fold), IL-6 (∼30-fold), IL-8 (∼3-fold), and MIP-1α (∼8-fold) compared with lean SAT. No differences were observed between pericytes from SAT and VAT from obese subjects (Fig. 7C). Moreover, Ang-1, secreted by pericytes from lean SAT (37.7 ± 11 pg/mL, n = 4), was not detected in culture media of pericytes from obese pericytes (SAT and VAT).

We observed that pericytes, displaying an altered phenotype during obesity, did not rescue inflammatory profile of ECs when cocultured in the 3D setting (Supplementary Fig. 12A). Addition of pericytes from VAT in adipocytes/ECs cocultures did not further increase the inflammatory response (Fig. 7A) or adipocyte dysfunctions (lipolysis and adipokine secretions [Supplementary Fig. 12D]). However, when pericytes were pretreated with dexamethasone (Dex), their inflammatory profile was reduced (Supplementary Fig. 12B), leading to a global decrease of inflammation in adipocytes/ECs cocultures (Fig. 7D). However, addition of Dex-treated pericytes failed to rescue adipocyte functions in these cocultures (Supplementary Fig. 12D).

Because Ang-1 production was decreased in pericytes during obesity and because its secretion was not rescued by Dex treatment (Supplementary Fig. 12C), we tested the effects of recombinant Ang-1 on VAT-ECs from obese subjects (inflammatory profile) and the consequences on cocultured adipocytes. Briefly, VAT-ECs were incubated in the presence or absence of Ang-1 (100 ng/mL) for 24 h before coculture with unilocular adipocytes. Cells were then cocultured in the 3D setting for 3 days as described in Supplementary Fig. 5.

A 24-h treatment of VAT-ECs with Ang-1 (Ang-1ECs) significantly reduced their secretion of CXCL_1/2/3 (−55.8%, P < 0.01) and IL-6 (−37.3%, P < 0.05), whereas it tended to decrease G-CSF and IL-8 secretions (n = 5) (Fig. 8A). Interestingly, when adipocytes were cocultured with Ang-1 ECs, the adipocytes recovered their lipolytic functions (50%, n = 3, P < 0.05) and insulin sensitivity (39%, n = 5, P < 0.05) compared with control cocultures, whereas adipokine secretion remained unchanged (Fig. 8B and C; Supplementary Fig. 13). Moreover, coculturing adipocytes and Ang-1ECs led to reduced secretion of several cytokines and chemokines compared with a control coculture of visceral adipocytes/ECs (n = 5): G-CSF (−42.9%, P < 0.05), IL-6 (−44.1%, P < 0.05), CXCL_1/2/3 (−38.6%, P < 0.01), and IL-8 (−47.8%, P < 0.05) (Fig. 8D). Notably, Ang-1 treatment had no direct effect on adipocyte function (data not shown). We did not observe any benefits of Ang-1 when it was added directly to cocultures without WAT-EC pretreatment (data not shown). Taken together, these findings suggest that Ang-1 partially reverses adipocyte dysfunction and inflammation through a specific action on EC inflammation.

We aimed to reproduce the adipocyte/EC dialog participating in the complex intercellular network of the inflamed VAT in a 3D setting. Better knowledge of this dialog could help in the development of new therapeutic strategies targeting inflammatory WAT. We cocultured unilocular adipocytes and ECs isolated from VAT from obese subjects in a 3D hydrogel in which cells cultured alone maintain their functions and their inflammatory phenotype for at least 7 days. Direct cocultures in the 3D setting allow cell-cell contact and the paracrine dialog between cell populations, maintaining labile signals. By following adipocyte functions in the 3D cocultures, we identified an inflammatory cross-talk between adipocytes and VAT-ECs that leads to alterations in lipolytic activity and insulin sensitivity. Adipokine secretion was also impaired in adipocyte-EC cocultures. The decreased secretion of adiponectin and an inflammatory environment are in accordance with most studies exploring EC dysfunctions (8) and insulin resistance (21). In contrast, the influence of inflammation on decreased lipolysis remains a subject of debate. Several studies have described increased basal lipolysis in inflammatory 3T3-L1 adipocytes (3,22), suggesting different mechanisms in unilocular human adipocytes. Interestingly, the effects were observed under β-adrenergic stimulation, which is impaired in obese subjects (23).

In our coculture setting, we observed increased secretion of many relevant inflammatory molecules, particularly IL-6, G-CSF, CXCL_1/2/3/8, and IFN-γ, which are proposed to contribute to obesity-related complications, such as type 2 diabetes and atherosclerosis. Overall, these overproduced chemokines and cytokines could aggravate inflammation in VAT by promoting the accumulation and inflammation of immune cells. For example, fractalkine is involved in the recruitment of monocytes and T lymphocytes in atherosclerosis, type 2 diabetes, and obesity (24). Interestingly, adipocytes were recently identified as antigen-presenting cells in response to IFN-γ in high-fat-diet mice with T-lymphocyte activation (25). The adipocyte production of IFN-γ could perpetuate, by a paracrine loop, WAT inflammation. CCL5, which is overproduced by VAT in human obesity, promotes the accumulation and survival of macrophages in WAT (26).

Although being performed with human WAT cells, our 3D setting presents some technical limitations, notably in reproducing the complex inflammatory environment of obese WAT. While we focused on adipocytes and ECs, other cell types could also influence this cross-talk (i.e., preadipocytes, neutrophils, or macrophages). Our team showed that inflammatory macrophages induced an inflammatory and profibrotic phenotype of human preadipocytes (27). More recently, we showed that neutrophils, interacting with VAT-ECs of obese subjects, provoked their inflammation/senescence (18).

Obesity provokes an ER stress in adipocytes (28). Notably, ER stress induces a decreased secretion of leptin and adiponectin, an alteration of insulin signaling and lipolysis, and inflammation in these cells (29). Here, we observed an ER stress in cocultured cells, which could be linked to the adipocyte dysfunctions.

Chronically elevated levels of IL-6, combined with increased IL-1β, favor development of type 2 diabetes (30). As with IL-6 (31), IL-1β secretion is increased in WAT during obesity (32), and both cytokines are known to act in concert in diverse biological process (33). Il-1β is also implicated in a “secondary” inflammatory response, regulating the production of cytokines and chemokines as G-CSF, IL-6, or IL-8 (34–36). Interestingly, our study showed that neutralization of IL-6 and IL-1β in cocultures led to a decreased inflammatory response and a rescue of some metabolic and secretory alterations of adipocytes (i.e., G-CSF, CXCLs, and IL-8). Conversely, treatment of SAT cells with recombinant IL-6 and IL-1β provoked roughly the same cell dysfunctions. This strongly suggests that IL-6 and IL-1β are important adipocyte/EC-derived factors provoking cell dysfunctions.

The functional properties and activation state of WAT-ECs play a crucial role in the promotion of both adipocyte alterations and inflammatory cross-talk. WAT-ECs isolated from lean individuals were not able to promote adipocyte alterations or inflammatory responses, highlighting the specificity of the cross-talk of inflammatory WAT cells in the context of obesity. An important purpose of the current study was to reduce WAT-EC inflammation in order to reverse the adipocyte dysfunctions. We chose Ang-1, which is constitutively produced by pericytes and known to prevent EC inflammation (12). Low levels of Ang-1 and a decreased number of pericytes are both hallmarks of type 2 diabetes (37,38). Ang-1 treatment of VAT-ECs tended to decrease the inflammatory profile and reversed, at least in part, the inflammatory response and restored lipolysis and insulin sensitivity in the coculture system. Our results are supported by previous studies showing protective effects of Ang-1 on the vasculature. In in vitro models, Ang-1 has been shown to prevent EC apoptosis and reduce the inflammation of ECs through decreased surface expression of adhesion molecules (39,40). In mice, Ang-1 deficiency exaggerates the wound-healing response to injury, leading to profound damage, such as fibrosis and vascular abnormalities (41). Conversely, Ang-1 treatment of diabetic mice improves vascular density, reduces adipocyte size, and ameliorates metabolic disorders relative to obesity (42).

Several studies have described reduced tissue abundance or altered pericyte phenotype in pathological conditions such as diabetic retinopathy and systemic sclerosis, respectively (43,44). For the first time, to the best of our knowledge, we provide information on the features of pericytes in the WAT from obese subjects. We highlighted a strong decrease of the pericytes-to-EC ratio in obese WAT, suggesting a defect in the vascular protection by pericytes of WAT-ECs. Moreover, these pericytes displayed an altered phenotype with increased expression of several inflammatory molecules (CCL2, CCL5, and TNF-α) and a decreased expression Ang-1 compared with pericytes from lean subjects. Thus, the inflammation-associated senescence of VAT-ECs, which was attributed to adipocyte secretions (5), may also result from pericyte dysfunctions in WAT from obese subjects. Notably, dysregulation of VAT-ECs may also result from disrupted contacts with pericytes (45). However, more information is needed on the direct links between adipocyte, EC, and pericyte dysfunctions in the VAT of obese subjects. The kinetics of these events is not fully understood and cannot be resolved by the present in vitro study. Kinetics studies with high-fat-diet mice models regarding interactions between adipocyte metabolism and EC/pericyte alterations within VAT would be critical to support our conclusions.

Finally, the inflammatory cross-talk between adipocytes and VAT-ECS could position them as important actors in the maintenance of VAT inflammation. Amelioration of VAT-EC dysfunctions by acting on pericyte phenotype or targeting the inflammatory secretions of ECs could prevent the adipocyte alterations responsible for obesity-related complications, such as atherosclerosis and type 2 diabetes.

Acknowledgments. The authors acknowledge patients and the physician Dr. Christine Poitou of the Nutrition Department of Pitié Salpêtrière for patient recruitment. They also acknowledge Christophe Klein from imaging facilities (Cordeliers Research Center). For cellular studies, ethics authorization was obtained from Comité de protection des personnes Pitié Salpêtrière. Human adipose tissues pieces were obtained thanks to a clinical research contract (Assistance Publique/Direction de la Recherche Clinique AOR 02076). The manuscript was edited for language and style by San Francisco Edit.

Funding. This work was supported by a grant from the European Community seventh framework program, Adipokines as Drug to combat Adverse Effects of Excess Adipose tissue project (contract number HEALTH-FP2-2008-201100); Assistance Publique Hopitaux de Paris Clinical Research Contract, Emergence program, University Pierre et Marie Curie (to V.P.); and French National Agency of Research (French government grant, “Investments for the Future,” grant ANR-10-IAHU, ANR AdipoFib).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. V.P. and D.L. conceived and performed the experiments, analyzed data, and wrote the manuscript with the input of all the coauthors. C.R. conceived and performed the experiments and analyzed data. N.V. collected adipose tissue samples of obese patients during bariatric surgery. K.C. wrote the manuscript with the input of all the coauthors. K.C. and D.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.