Adipocyte Glucocorticoid Receptor Activation With High Glucocorticoid Doses Impairs Healthy Adipose Tissue Expansion by Repressing Angiogenesis

Glucocorticoids (GCs) and their synthetic analogs are a potent treatment for inflammatory diseases and are therefore among the most prescribed drugs. Today, 1–1.2% of the general population may undergo systemic GC therapy for months or years (1). However, prolonged GC exposure, as observed in patients with Cushing syndrome, may lead to metabolic adverse events, including insulin resistance, hyperglycemia, dyslipidemia, and lipodystrophy (2–4), suggesting a role for GC signaling in the metabolic syndrome.

GCs have diverse effects on adipose tissue (AT) biology, ranging from adipocyte precursor differentiation to the regulation of metabolic and endocrine functions in mature adipocytes (2,3). The biological effects of cortisol (in humans) and corticosterone (in rodents) are mediated by both the mineralocorticoid receptor and the glucocorticoid receptor (GR), the latter being prominently involved in the metabolic alterations induced by GCs (5,6). These data strengthen the interest for studying the impact of GR activation on AT biology and metabolism.

Over the past decade, independent studies have reported the effect of constitutive adipocyte-specific GR invalidation in mice fed a high-fat diet (7–12) leading to contrasting phenotypes. The discrepancies may relate to the experimental methods used and to the developmental process adapted to constitutive ablation of the GR. To overcome potential compensatory mechanisms resulting from constitutive GR deficiency during AT development in mice, we generated an inducible model of selective adipocyte GR invalidation (AdipoGR-KO) in adult mice to enable time- and cell-specific ablation of adipocyte GR and avoid any potential confounding effects that GR might have on adipocyte differentiation and development. We treated AdipoGR-KO mice with the natural ligand of the murine GR, corticosterone (CORT), which led to hypercortisolism throughout the day, while retaining GC circadian oscillations (13). As expected, the hypercorticism occurring during the diurnal period led to glucose intolerance and insulin resistance in control mice, whereas these alterations were markedly reduced in GR-deficient mice (13). This metabolic improvement was associated with reduced hepatic steatosis and improved lipid profile. Interestingly, these beneficial effects were associated with a healthy and massive expansion of subcutaneous AT (SCAT) and gonadal AT (GAT), reflecting the beneficial adaptations in adipocyte lipid storage and inflammatory profile (13). These results were also consistent with higher adiposity and insulin sensitivity observed in constitutive GR-deficient mice, despite GC treatment (12,14). Thus, selective GR deletion within adipocytes alleviated some adverse GC effects by favoring a healthy expansion of adiposity in AdipoGR-KO mice.

AT is a highly active tissue in which expansion is accompanied by the development of an appropriate vascular network adapted to the tissue size and nutritional challenges (15–17). Blood flow ensures metabolic and oxygen exchanges and delivers hormones and growth factors involved in whole-body metabolic homeostasis (15,16,18). The vascular system controls the AT microenvironment, which influences preadipocyte differentiation, AT function, and plasticity (18). Blood vessel growth and angiogenesis are partly promoted by mild hypoxia through induction of the transcription factor hypoxia-inducible factor 1α (HIF-1α) (19), which acts with its partner HIF-1β on hypoxia response elements present in the promoter region of the gene coding for vascular endothelial growth factor A (VEGFA), a key angiogenic factor (17,19). During obesity, the rapid increase in fat mass exceeds the vascular network expansion capacities, triggering AT inflammation and insulin resistance and contributing to systemic metabolic alterations (17,19,20). Adipose-specific VEGFA overexpression improves AT vascularization and protects mice from deleterious effects of a high-fat diet, while suppressing VEGFA in AT reduces the adipose vascular network and promotes inflammation and the development of severe insulin resistance (21–23). Angiogenesis adaptation therefore seems to be an early and crucial process for healthy AT expansion.

The aim of the current study is to show that adipocyte GR activation could directly preclude AT vascular network adaptation and, in turn, modulate whole-body metabolic homeostasis. By combining in vivo and in vitro approaches, we unveil a new mechanism in which the adipocyte GC/GR signaling restrains AT vascularization through the regulation of the HIF-1α/VEGFA pathway, therefore participating in the deleterious metabolic effects of GCs.

AdipoGR-KO mice were generated as previously described (13). Cre recombinase was activated in 10-week-old male mice by intraperitoneal injection of tamoxifen (60 mg/kg/day) (MP Biomedicals, Illkirch-Graffenstaden, France) for 5 consecutive days. Fourteen-week-old AdipoGR-KO and GR-Flox control littermate (hereafter defined as Ctrl) mice were treated for 4 days or 4 weeks with CORT (100 μg/mL) (Sigma-Aldrich, St. Louis, MO) or vehicle (VEH) (1% ethanol) added to drinking water (1). For the VEGFA blockade experiment, the soluble decoy receptor aflibercept (AFLI) (4 mg/kg; Sanofi, Gentilly, France) or diluent (PBS) was injected intraperitoneally twice a week throughout the CORT or VEH treatment for 4 weeks. At sacrifice, tissues were snap frozen in liquid nitrogen and kept at −80°C. Blood was drawn by intracardiac puncture with heparin-moistened syringes. Plasma was obtained after centrifugation at 5,000g for 10 min at 4°C. Body mass composition is detailed in the Supplementary Material.

Insulin tolerance test and insulin pulse experiments were performed as previously described (13) in 5-h fasted mice injected intraperitoneally with 1 IU/kg and 1.5 IU/kg of human insulin (Actrapid Penfill; Novo Nordisk, Paris, France), respectively. Hepatic triglycerides (TGs) were extracted and measured with a colorimetric diagnostic kit according to the manufacturer’s instructions (Triglycerides FS; DiaSys, Grabels, France).

Paraffin-embedded liver and AT were cut into 5-μm-thick sections and stained with hematoxylin and eosin. For CD31 staining, Novolink Polymer Detection Systems Kit was used according to the manufacturer’s recommendations (Leica, Buffalo Grove, IL) (Supplementary Table 1). The specificity of the CD31 signal was verified with a nonspecific mouse IgG1 antibody (Supplementary Fig. 1A–D). CD31 intensity was determined for 5–10 fields of SCAT and GAT per section covering the entire tissue surface at original magnification ×10 (IX83 Inverted Microscope; Olympus, Rungis, France) using QPath and ImageJ software (https://rsbweb.nih.gov/ij/). Results are expressed as region of interest per adipocyte or as endothelial cells per adipocyte.

When indicated, adipocyte and stromal vascular fractions (SVFs) were isolated from SCAT and GAT of treated Ctrl and AdipoGR-KO mice as previously described (13).

Freshly isolated SVFs were incubated with fluorescent-conjugated antibodies or their respective controls (Supplementary Table 1). Flow cytometry was performed with a CytoFLEX cytometer (Beckman Coulter, Villepinte, France), and data analysis was conducted using FlowJo software (Becton, Dickinson and Company, Franklin Lakes, NJ).

3T3-F442A preadipocytes (24) were maintained and differentiated as described in the Supplementary Material. Mature adipocytes were treated with indicated dexamethasone (Dex) (Sigma-Aldrich) concentrations, with or without 100 μmol/L of RU-486 (Sigma-Aldrich), under normoxic (21% O₂, 5% CO_2, and 74% N₂) or hypoxic (1% O₂, 5% CO₂, and 94% N₂) conditions for 16 h.

Total RNA extraction from human and mouse tissues and adipocyte/SVFs are detailed in the Supplementary Material.

Details are provided in the Supplementary Material.

Chromatin immunoprecipitation (ChIP) assays were performed on mature 3T3-F442A adipocytes treated with 100 nmol/L Dex under normoxic or hypoxic conditions for 16 h with the Acetyl-Histone H4 Immunoprecipitation (ChIP) Assay Kit according to the manufacturer’s instructions (Sigma-Aldrich) and as previously described (25) and detailed in Supplementary Tables 1 and 2.

DyLight 649–labeled tomato lectin from Lycopersicon esculentum (Vector Laboratories, Peterborough, U.K.) was systemically administered to the mice via the retro-orbital sinus 5 min before sacrifice (Supplementary Table 1). The detailed methodology is provided in the Supplementary Material.

Patients were diagnosed with endogenous ACTH-independent Cushing syndrome caused by benign adrenal gland adenoma. The Cushing syndrome diagnosis was made according to at least one of the following three parameters: an increase in urinary free cortisol, altered plasma cortisol after a 1-mg Dex suppression test, or a high midnight plasma cortisol associated with a low ACTH plasma level. Sample collection, processing, and morphometric and metabolic parameters are described in Supplementary Tables 3 and 4.

All values are expressed as the mean ± SEM. Normality was tested by Shapiro-Wilk test. For comparisons of two groups, data were analyzed using Student t test. For multiple group comparisons, data were analyzed using one-way ANOVA followed by Bonferroni post hoc analysis when appropriate or Kruskal-Wallis test. Correlations between human gene expression data and clinical and anthropometric parameters were analyzed by calculating Spearman rank correlation coefficients. Test analyses were performed using Prism 8 software (GraphPad Software, La Jolla, CA). P < 0.05 was considered statistically significant.

Data sets and resources are available upon request.

To analyze GR’s role in angiogenesis during CORT-induced AT expansion, we measured Vegfa and Hif-1α mRNA expression in whole SCAT and GAT or, alternatively, in isolated adipocyte and SVFs collected from AdipoGR-KO and Ctrl mice treated with VEH or CORT for 4 weeks. CORT treatment drastically decreased Vegfa and Hif-1α mRNA levels in whole SCAT and GAT of Ctrl mice and in the corresponding adipocyte fractions (Fig. 1A and B). In agreement, VEGFA protein content was reduced by 40–45% in the SCAT and GAT of CORT-treated Ctrl mice (Fig. 1C). Strikingly, adipocyte GR deficiency almost completely prevented the CORT-related downregulation in Vegfa and Hif-1α mRNA expression in AT and more specifically, in their adipocyte fractions (Fig. 1A and B and Supplementary Fig. 2A). Accordingly, VEGFA protein concentration was increased by 2–4.5-fold in AT of CORT-treated AdipoGR-KO mice compared with Ctrl mice (Fig. 1C) without impacting VEGFA plasma concentration (Supplementary Fig. 2B). Furthermore, GR deficiency in AT did not affect Vegfa and Hif-1α mRNA expression in the liver and skeletal muscles of mice under VEH and CORT treatment (Supplementary Fig. 2C). Notably, the increased Vegfa mRNA and protein expression in CORT-treated AdipoGR-KO mice was consistent with their massive AT expansion and preserved insulin sensitivity compared with Ctrl mice (Fig. 1D and E).

Interestingly, when comparing SCAT and GAT weight gain to Vegfa and Hif-1α expression after only 4 days of CORT treatment in Ctrl and AdipoGR-KO mice, we found elevated fat depot weights (Supplementary Fig. 3A) associated with higher expression of Vegfa and Hif-1α mRNA in fat depots of AdipoGR-KO mice compared with Ctrl mice (Supplementary Fig. 3B). Altogether, these data demonstrate a functional link between adipocyte GR signaling and an early angiogenesis and AT expansion.

To investigate the regulation of adipose Vegfa and Hif-1α mRNA expression by GC, we treated 3T3-F442A adipocytes with different doses of Dex alone or in association with the GR antagonist RU-486 (Fig. 2A). As expected, the expression of the GR target gene Fkbp5 was strongly induced by Dex in a dose-dependent manner (up to 20-fold at 100 nmol/L) and conversely, was inhibited in the presence of RU-486 (Fig. 2A). Vegfa and Hif-1α expression decreased in a dose-dependent manner in Dex-treated 3T3-F442A adipocytes, down to a 50% level for Vegfa and 40% for Hif-1α. RU-486 totally prevented the inhibitory action of Dex treatment on both genes (Fig. 2A), supporting the fact that Vegfa and Hif-1α regulation depends on GR activation.

We exposed 3T3-F442A adipocytes under hypoxia, a condition that stabilizes HIF-1α protein, and tested whether Dex could affect Vegfa and Hif-1α expression in this permissive condition. As expected, hypoxia induced glucose transporter Slc2a1 and Vegfa mRNA expression (26,27) (Fig. 2B). In this context, Dex treatment markedly reduced hypoxic induction of Slc2a1 and Vegfa mRNA (Fig. 2B). In agreement with the known posttranslational regulation of HIF-1α protein, hypoxia alone had little impact on Hif-1α mRNA expression (Fig. 2B) but increased the overall protein level (Fig. 2C). Dex decreased both Hif-1α mRNA and protein expression, suggesting a transcriptional and posttranscriptional GR-mediated HIF-1α regulation (Fig. 2B and C). ChIP assay showed that Dex treatment decreased hypoxia-induced HIF-1α recruitment onto the Vegfa promoter (Fig. 2D), suggesting a reduced HIF-1α transcriptional activity. These results indicate that under Dex stimulation, GR downregulates Vegfa expression through a decrease in Hif-1α expression and activity.

To explore whether changes in Vegfa mRNA and protein levels affected AT angiogenesis and the vascular network in CORT-treated AdipoGR-KO mice, we first performed flow cytometry analysis of CD31⁺/CD45⁻ endothelial cells from SCAT and GAT. In Ctrl mice, CORT treatment decreased endothelial cell number compared with VEH-treated mice (Fig. 3A), consistent with CORT-induced downregulation of Vegfa expression (Fig. 1A). Strikingly, adipocyte GR deficiency completely prevented this decrease in GAT, leading to a threefold increase in endothelial cell number compared with CORT-treated Ctrl mice (Fig. 3A). Although not significant, we observed a trend toward a higher number of endothelial cells in the SCAT of CORT-treated AdipoGR-KO mice (Fig. 3A). These results were confirmed by immunohistochemistry showing an increased endothelial cell (CD31) labeling in AT of AdipoGR-KO mice compared with Ctrl mice (Fig. 3B). To assess whether these changes were associated with vascular network remodeling, we performed three-dimensional (3D) fluorescence imaging of the cleared SCAT and GAT vascular network from Ctrl and AdipoGR-KO mice (Fig. 3C–E). We observed a higher vascular volume in the AT of CORT-treated AdipoGR-KO compared with Ctrl mice (Fig. 3D and Supplementary Movies 1–8) along with a higher mean adipocyte volume (Fig. 3E). A mathematical estimation of the number of adipocytes relative to AT weight (Fig. 3F and G) showed that the difference in fat mass observed between CORT-treated AdipoGR-KO and Ctrl mice was mostly due to adipocyte hypertrophy. Vascular and adipocyte volumes were not modified in AdipoGR-KO receiving VEH. These findings support the idea that adipocyte GR activation restrains AT vascularization and fat cell size.

To determine the role of VEGFA in the CORT-induced AT expansion in AdipoGR-KO mice, we treated animals at the same time as CORT with AFLI, which binds VEGFA and blocks its action (28). AFLI efficiency on VEGFA/VEGF receptor 2 signaling was validated by the expression of the target gene endothelial cell-specific molecule 1 (Esm1) (29). Esm1 was upregulated in SCAT and GAT of CORT-treated AdipoGR-KO mice compared with Ctrl mice and was decreased, as expected, by AFLI (Fig. 4A).

We then explored the consequences of altered VEGFA signaling on endothelial cell number by flow cytometry and immunohistochemistry. Blocking VEGFA action in CORT-treated AdipoGR-KO mice drastically reduced endothelial cell number per adipocyte compared with mice treated with CORT alone (Fig. 4B and C). Moreover, AFLI and CORT–combined treatment attenuated the greater body weight gain, fat mass, and AT expansion observed in AdipoGR-KO mice treated with CORT alone (Fig. 4D–G). Of note, AFLI alone had no significant impact on body and AT weights in CORT-treated Ctrl mice (Supplementary Fig. 4). Therefore, these findings demonstrate that the crosstalk between GR-deficient adipocytes and endothelial cells through VEGFA secretion is necessary for AT expansion in AdipoGR-KO mice.

CORT-treated AdipoGR-KO mice were, as expected, more sensitive to insulin than CORT-treated Ctrl animals (Figs. 5A and 1F). However, AFLI- and CORT-treated AdipoGR-KO mice reached an insulin resistance level similar to CORT-exposed Ctrl mice (Fig. 5A). We performed insulin pulse experiments to investigate the effect of AFLI treatment on the activation of insulin signaling in SCAT, GAT, and liver and measured Akt phosphorylation (P-Akt) (Fig. 5B–E). Immunoblot analysis indicated that AdipoGR-KO mice treated with CORT were protected from CORT-induced insulin resistance in SCAT and GAT, but not in the liver, compared with Ctrl mice (Fig. 5B–E). Importantly, AFLI treatment drastically impaired insulin-induced P-Akt in GAT and to a lesser extent in SCAT, but not in the liver, of CORT-treated AdipoGR-KO mice (Fig. 5B–D). Overall, our data support the concept that AdipoGR-KO mice are less protected from CORT-induced insulin resistance when treated with AFLI and show that VEGFA is a prerequisite for healthy AT expansion, which is required to sustain insulin sensitivity.

We previously demonstrated that GR-deficient mice were protected from hepatic steatosis and had lower TG concentrations in their liver despite CORT exposure (13). To investigate whether AFLI treatment could impact CORT-induced hepatic steatosis, we performed hematoxylin and eosin staining on liver sections. As expected, CORT treatment led to liver steatosis in Ctrl mice, which was strongly reduced in AdipoGR-KO mice (Fig. 5F). GR-deficient mice treated with combined AFLI and CORT treatment were no longer protected from liver steatosis (Fig. 5F). Accordingly, hepatic TG concentration was drastically reduced in the liver of CORT-treated AdipoGR-KO compared with Ctrl mice, an effect that was abrogated under AFLI treatment (Fig. 5G). Thus, adipocyte GR plays a key role in the regulation of healthy AT expansion and, therefore, the protection against hepatic steatosis in AdipoGR-KO mice under CORT treatment.

We assessed the pathophysiological relevance of GC-regulated VEGFA expression in a cohort of 20 patients with hypercortisolism adrenal adenoma (Pattern of Gene Expression in Adipose Tissue From Patients With Cushing Syndrome [LIPOCUSH]; clinical trial reg. no. NCT01688349, ClinicalTrials.gov). Patients’ clinical, metabolic, and anthropometric parameters are shown in Supplementary Tables 3 and 4. We determined whether VEGFA expression was correlated with a healthier metabolic profile. As observed in Fig. 6A and Supplementary Table 5, VEGFA expression was negatively correlated with BMI and HOMA of insulin resistance (HOMA-IR) and positively correlated with the percentage of fat mass in the upper and lower limbs. We then divided the LIPOCUSH cohort into two subgroups according to their low or high level of VEGFA expression in SCAT, based on the stratified median (Fig. 6B). Patients with the highest expression of VEGFA displayed lower levels of insulin resistance (assessed by HOMA-IR index) and HbA_1c (Supplementary Table 3). No further changes were observed regarding cortisol production, plasma TGs, and free fatty acids between the groups. Patients with higher VEGFA presented an increase in the percentage of fat mass in upper and lower limbs despite a similar percentage of total and trunk fat mass compared with patients expressing the lowest VEGFA level (Supplementary Table 4). Thus, higher VEGFA expression in patients with Cushing syndrome was associated with an improved metabolic phenotype and enhanced expansion of AT in the limbs, likely acting as a protective metabolic sink through its increased capacity of lipid buffering (30,31).

We further explored whether expression of key markers of GC signaling and adipocyte differentiation and function were correlated to VEGFA expression (Fig. 6 and Supplementary Table 5). VEGFA expression was positively correlated with NR3C1 (human GR gene) and adipocyte marker–related genes, while negatively correlated with the expression of 11-β hydroxysteroid dehydrogenase 1 (HSD11B1), an enzyme responsible for local production of cortisol. The stratified analysis highlighted that patients with higher VEGFA displayed an increased NR3C1 expression, while HSD11B1 mRNA levels decreased (Fig. 6C). Furthermore, mRNA expression of early (C/EBPB, C/EBPD) and late (peroxisome proliferator–activated receptor γ [PPARG], C/EBPA) adipogenic transcription factors (Fig. 6D and E) and lipogenic markers (SREBF1 and fatty acid synthase [FASN]) (Fig. 6F) were significantly increased in patients with the highest VEGFA expression, supporting an improved adipocyte differentiation and lipid storage capacity. Finally, according to the decreased insulin resistance observed in patients with high VEGFA, ADIPOQ, but not LEP, mRNA expression was elevated (Fig. 6G). Collectively, these data show that patients with high cortisol levels and lower VEGFA expression in SCAT exhibited a worsened metabolic profile compared with patients with higher VEGFA expression, highlighting a combined action of VEGFA and GC/GR pathway in human AT and metabolic homeostasis.

We previously demonstrated that the selective deletion of the GR in AT counteracts several deleterious GC-induced metabolic effects, leading to insulin resistance and diabetes (13). This improvement was associated with a massive expansion of AT in mice, suggesting a detrimental role of the GR in the healthy AT adaptation, an effect still poorly documented. In this study, we demonstrate for the first time that adipocyte GR acts as a central player of adipose vascular flexibility by repressing Vegfa expression and downregulating VEGFA protein, which is known as a cornerstone of AT plasticity (17). Our data are consistent with the work of Hayashi et al. based on transcriptomic analysis suggesting an altered VEGF signaling pathway in patients with Cushing's syndrome (14). Hence, our main conclusion is that the adipocyte GC/GR signaling pathway negatively regulates AT vascularization, thereby precluding healthy AT expansion and whole-body metabolic homeostasis.

AT exhibits remarkable plasticity and expansion capacities, requiring the development of new vessels to support an adequate supply of oxygen and nutrients (15,32,33). However, an altered synergy between AT expansion and vascularization, as observed during obesity, may lead to hypoxia, inflammation, and impaired AT function (34). VEGFA is crucial to regulating this process, as previously documented by others (21–23). In the current study, CORT treatment in Ctrl mice led to deleterious metabolic effects, which coincide with decreased Vegfa expression in AT. These effects were counteracted by the absence of adipocyte GR, suggesting a GC- and GR-dependent Vegfa expression. Surprisingly, despite a strong induction of VEGFA in AT, plasma levels remained stable in AdipoGR-KO mice, indicating a paracrine rather than an endocrine effect of adipocyte VEGFA on endothelial cells to promote vascular network and AT expansion. These results are consistent with the work of Elias et al. (23), who showed that adipocyte Vegfa overexpression had no impact on VEGFA plasma levels. Moreover, our data are in line with the increased adipocyte Vegfa expression observed in total HSD11B1-deficient obese mice (35,36). Nevertheless, because of our unique inducible adipose-specific mouse model, we show for the first time that adipocyte GR signaling exerts a significant role on SCAT and GAT vascular adaptation and plasticity through VEGFA. The healthier expansion of GAT in addition to SCAT protects other lean tissues (e.g., the liver) from lipid accumulation and, therefore, improves whole-body insulin sensitivity.

A link between GR signaling and VEGFA suppression has been suggested in AT and non-AT (37,38); however, the molecular mechanism still remains elusive. We raised several hypotheses to explain Vegfa modulation by GC/GR signaling. No glucocorticoid-responsive element consensus site has been reported on Vegfa promoter (39). Therefore, HIF-1α, one of the major transcription regulators of VEGFA, appears as a relevant molecular mediator of GC action. Deletion of adipocyte prolyl-hydroxylase domain protein 2 (Phd2) in mice results in stabilization of HIF proteins, leading to a pseudo-hypoxia condition and an increased adiposity associated with a beneficial AT vascularization (40). Similarly, we showed a significant induction of Hif-1α closely related to Vegfa expression in adipose depots of CORT-treated AdipoGR-KO mice. Experiments on mature 3T3-F442A adipocytes showed a concomitant regulation of the Vegfa and Hif-1α contents by Dex, involving transcriptional and posttranslational HIF-1α changes and a reduced binding on Vegfa promoter. Collectively, these in vivo and in vitro approaches indicate that GR signaling decreases HIF-1α activity on Vegfa promoter, leading to a reduced Vegfa expression. However, the absence of glucocorticoid-responsive element consensus sites in Hif-1α promoter suggests a more complex level of regulation by GC/GR pathway (41). Alternative mechanisms could be considered, such as potential crosstalk between GR and PPARγ to modulate the HIF-1α/VEGFA pathway and therefore AT angiogenesis (42).

Our thorough study supports that the improved metabolic profile in AdipoGR-KO mice is partly due to an enhanced SCAT and GAT vascularization. While CORT decreased the number of CD31 endothelial cells in AT of Ctrl mice, adipocyte GR deletion prevented CORT-induced endothelial cell decline. Through the innovative technique of whole-AT clearing coupled to 3D volume fluorescence imaging (18,43,44), we showed that these data were associated with a higher density of the adipose vascular network, characterized by an increased vessel volume rather than a change in their length in the AT of AdipoGR-KO mice. The overall higher density of capillaries may favor the interaction surface with the surrounding adipocytes and, therefore, contributes to AT expansion by increasing adipocyte hypertrophy rather than hyperplasia.

To determine the contribution of VEGFA in the healthy AT expansion of AdipoGR-KO mice, we used AFLI combined with either VEH or CORT treatment to block VEGFA action. AFLI is used in human colorectal cancer or diabetic retinopathy (28,45,46). This recombinant protein was engineered by the fusion of two extracellular domains of VEGF receptor 1 and 2 to the heavy-chain portion of human IgG1. AFLI prevents VEGFA, VEGFB, and to a lower affinity, placenta growth factor binding to their receptors on endothelial cells. Blocking VEGFA action on endothelial cells impedes AT growth and development and results in the loss of the beneficial effects of GR deficiency on insulin resistance and liver steatosis. Although we are aware that a double-adipocyte Gr-Vegfa deletion in mice would be a relevant model to carry out this demonstration, our complementary and convergent approaches strongly suggest a major role for VEGFA in the healthy expansion of AT in mice with suppressed adipocyte GR.

Finally, we assessed the relevance of our findings in a cohort of patients with Cushing syndrome due to hypercortisolism adrenal adenoma. Patients with the highest VEGFA expression showed an improved glycemic profile associated with an increased expression of the insulin-sensitizing adipokine adiponectin. These patients also exhibited a higher expression of the main markers of adipocyte differentiation and lipogenesis in SCAT, suggesting a healthy and insulin sensitive AT. Although our human study was restricted because of a small number of selected patients, these data strengthen our in vitro and in vivo conclusions in mice regarding the role of VEGFA and angiogenesis in the control of AT expansion and metabolic homeostasis under hypercortisolism. Given that VEGFA is also known to promote vascular sympathetic innervation and to increase sensory nerve density, and given that AT vascularization and innervation share a remarkably close anatomical and functional relationship (47), we believe that it would be interesting to investigate in a future study the role of adipocyte GR in AT innervation.

This study has several limitations. First, we used only male mice. As several articles have shown, female mice are less susceptible to CORT treatment–induced adverse effects, such as insulin resistance and AT accumulation, compared with male mice (48,49). Thus, sex effects, including the importance of androgens and estrogens, are currently being investigated in our AdipoGR-KO model in an independent work to analyze the sex-dependent effects of GC. Another limitation of our study is the use of tamoxifen in our mouse model of invalidation. Although this estrogen receptor agonist can remain present in AT and act on its vascularization as shown in several articles (50,51), our mice were studied at a distance (2 months) from tamoxifen injections, which reduces the risk of potential confounding effects.

In conclusion, this new facet of adipocyte GR action on the physiology and pathophysiology of AT will help to define future therapeutic strategies aimed at reducing deleterious metabolic effects of GC. Research to date has focused on developing selective GR modulators that retain the anti-inflammatory properties of GC but limit their metabolic or bone side effects (52,53). However, due the complexity of GR signaling regulation at the molecular level and based on our results, it seems relevant to consider innovative approaches to selectively antagonize the GR, specifically in adipocytes, to prevent GC metabolic adverse effects such as insulin resistance and hepatic steatosis.

Acknowledgments. The authors are grateful to Dr. B. Gaugler (INSERM, Saint-Antoine Research Center, Sorbonne University) for helpful assistance on flux cytometry and M. Auclair, A. Mevel, and D. Moret (all from INSERM, Saint-Antoine Research Center, Sorbonne University) for technical assistance. The authors thank L. Dinard, A. Guyomard, T. Coulais, and Q. Pointout (Animal Housing Facility); B. Solhonne (Histomorphology Platform); R. Morrichon (Cell Imaging and Confocal Microscopy Platform); and A. Munier (Cytometry Platform) of the Saint-Antoine Research Center (INSERM, Sorbonne University) for excellent support. The authors thank A. Larsen and M. Sabbah (INSERM, Saint-Antoine Research Center, Sorbonne University) and E. Girault (Assistance Publique des Hôpitaux de Paris, Hôpital Saint-Antoine) for providing AFLI and L. Louadj (INSERM, Saint-Antoine Research Center, Sorbonne University) for helping with cultures under hypoxic conditions. The authors thank the patients who participated to the LIPOCUSH study and the LIPOCUSH investigators (Prof. P. Chanson, Prof. P. Kamenicky, Dr. E. Khun, and Dr. B. Parier from Bicêtre Hospital, Le Kremlin Bicêtre; Prof. J. Bertherat, Dr. L. Bricaire, and Dr. L. Guignat from Cochin Hospital; Dr. C. Ajzenberg from Mondor Hospital; Dr. S. Gaujoux, Prof. F. Menegaux, and Dr. C. Jublanc from Pitié Salpêtrière Hospital; Prof. A. Tabarin from Haut-Lévêque Hospital; Prof. S. Christin-Maitre from Saint-Antoine Hospital; and Prof. P. Sebe from Tenon Hospital and Diaconèses Hospital). The authors warmly thank K.E. Davis (UT Southwestern Medical Center) for helpful reading and discussion.

Funding. This work was supported by grants from INSERM, Sorbonne University, Société Francophone du Diabète (allocation no. RAK18021DDA), and Fondation pour la Recherche Médicale [FRM] (EQU201903007868). A.V., H.D., and A.L. were supported by doctoral fellowships from Ministère de l’Enseignement Supérieur and Institut National de la Santé et de la Recherche Médicale and an additional 1-year grant from FRM for H.D. This work was also supported by a grant from the French National Research Agency (ANR) through the Investments for the Future Labex SIGNALIFE (ANR-11-LABX-0028-01) to J.G., M.C., and J.-F.T.

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

Author Contributions. A.V., H.D., A.L., J.G., M.C., J.-F.T., B.F., A.G., and M.M. contributed to manuscript preparation, writing, analysis of the data, generation of all figures, data discussion, and manuscript editing. A.V., H.D., A.L., and M.G. performed and analyzed gene expression and protein experiments in culture and conditional KO mice. A.V., H.D., A.L., M.G., C.B., N.R., T.L., B.B., A.G., and M.M. performed the treatment, metabolic exploration, and biochemical analysis of the conditional KO mice. A.V., M.G., and N.R. performed and analyzed AT and liver histology. A.V., C.D., J.C., C.V., and B.F. performed and analyzed studies of the LIPOCUSH cohort. A.V. and K.P. performed and analyzed flux cytometry experiments. J.G. achieved AT clearing and analyses for AT clearing. E.H. performed ChIP studies on cultured adipocytes. T.L. performed animal injection of fluorescent (tomato) lectin. C.V. and B.F. recruited patients for the LIPOCUSH cohort. M.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.