NOV/CCN3: A New Adipocytokine Involved in Obesity-Associated Insulin Resistance

More than 300 million people are currently obese worldwide, and this noninfectious epidemic affects both Western and emerging countries (1–3). In the absence of adequate care, obesity induces many comorbidities and reduces life expectancy (4–7). Thus, there is a crucial need to better understand the pathophysiology of obesity, insulin-resistance, and type 2 diabetes. In this context, the discovery of new adipocytokines involved in the control of energy balance is not only a challenge for understanding the mechanisms responsible for the abnormal development of obesity or insulin resistance but also offers new opportunities to manage this disease and its complications.

The NOV/CCN3 gene (nephroblastoma overexpressed gene) (8) is a founder of the CCN (Cyr61/CCN1, CTGF/CCN2, NOV/CCN3) family. The CCN genes encode matricial multifunctional proteins that are involved in organogenesis (9). In addition, these proteins play key roles in inflammation (10), wound healing, fibrosis, and cancers (9). Several studies have also shown that NOV is involved in the adhesion, migration, proliferation, differentiation, and survival of different cell types (11–16). NOV also modulates the expression of inflammatory molecules and their effects (17–20). These different functions are mediated through the interaction of NOV with specific integrins and, in some cases, by the Notch pathway (9).

Measurements of plasma NOV concentrations in a large cohort of patients screened for cardiometabolic diseases, in which 60% of the patients had a BMI above 30 kg/m², revealed, for the first time, a strong relationship between plasma NOV concentrations and obesity (21). Plasma NOV levels were also correlated with weight loss in patients who had undergone bariatric surgery. Accordingly, we found that the NOV protein was synthesized and secreted by adipocytes and macrophages in adipose tissue from obese patients and in human primary cultures (21). Moreover, we demonstrated that the induction of NOV during adipose tissue expansion was not restricted to humans. Indeed, in mice fed a high-fat diet (HFD), plasma NOV levels and its expression in adipose tissue were also increased compared with mice fed a standard diet (SD) (21).

Even though a high BMI is often associated with impaired glucose tolerance and type 2 diabetes, we did not find any correlation between NOV and fasting glycemia. However, there was a positive correlation between plasma NOV and HbA_1c (21). Thus, it is conceivable that a relationship could exist between NOV and carbohydrate metabolism. In this context, a possible link between NOV and insulin was suggested by Shimoyama et al. (22), who reported that NOV expression was reduced in rat aortas after streptozotocin injection and was increased by insulin treatment. Moreover, the nov gene is localized on chromosome 8q24 (23), which is a susceptibility locus controlling β-cell function in linkage studies of patients with diabetes (24). Taken together, these results and a recent report (25) showing that NOV is a direct target of transcription factor FOXO1 and that NOV impairs insulin secretion in pancreatic β-cells reinforce the possibility that this protein could be involved in energy homeostasis. However, no data have been reported to date on the function of NOV in adipose tissue or in whole-body metabolism. Thus, the main focus of this study was to identify whether targeted disruption of the nov gene modulates energy balance in mice fed a normal or obesogenic diet.

NOV^−/− mice were generated by Shimoyama et al. (22) using a targeting vector to delete the genomic region encompassing exons 1, 2, and a part of 3 for complete elimination of the NOV transcription. Mice heterozygously transmitting the NOV-knockout genome were generated, and their descendants were backcrossed to C57Bl/6J for at least six generations and maintained as heterozygotes. C57Bl/6J heterozygous males were then mated with 129sv/PAS females. Embryos derived from these crossings were reimplanted in pseudopregnant females to get an F1 on a mixed background (C57Bl/6J-SV129) with the “free of specific pathogenic microorganisms” status.

C57Bl/6-SV129 NOV^−/− mice and the wild-type (WT) controls were derived from the same heterozygous crossings. Mice were kept on a 12-h light/12-h dark cycle with ad libitum access to food and tap water. Male mice (6 weeks old) were fed a SD (4% fat; Genestil 1326 Royaucourt, France) or an HFD (42% fat; TD.88137; Harlan Teklad, Gannat, France) for 16 weeks. Mice were bred according to the Guide for the Care and Use of Laboratory Animals (U.S. National Institutes of Health Publication No. 85-23, revised 1996). The animal facility was granted approval (B-75-12-01) given by the French Administration. All procedures were approved by a local ethic committee (No. Ce5/2012/091).

Mice genotypes were performed as previously described (22). The genotypes were further checked by NOV mRNA levels in epididymal white adipose tissue (eWAT) and subcutaneous WAT (scWAT), the plasma level of the protein and its expression in heart where NOV is normally highly expressed (Supplementary Fig. 1).

Body weight (BW) was measured using a laboratory scale (Ohaus CS 200 Series; Sigma-Aldrich, l’Isle d’Abeau, France). Body mass composition was analyzed using an EchoMRI 100 Whole Body Composition Analyzer (EchoMRI, Houston, TX) according to the manufacturer's instructions (26).

For measurement of energy expenditure, respiratory exchange ratio (RER), and spontaneous locomotor activity, mice were placed in a LabMaster indirect calorimetry system (TSE Systems GmbH, Bad Homburg, Germany).

Oral glucose tolerance tests (OGTTs) and insulin tolerance tests (ITTs) were performed after 14 and 15 weeks of the diet, respectively, on food-deprived (6 h and 4 h, respectively) nonanesthetized mice. For the OGTT, animals were weighed and orally fed using gavage needles (Ref 18060-020; Phymep, Paris, France) with a 1 g/kg BW of a 20% glucose solution. Whole-tail vein blood (30 µL) was sampled at baseline and 15 min to measure plasma insulin. For the ITT, animals were weighed and intraperitoneally injected with human insulin (0.75 mU/kg). For both tests, glucose levels were obtained from whole-tail vein blood using an automatic Accu-Chek Performa glucometer (Roche, Meylan, France) at indicated times. For insulin-signaling experiments, mice were intraperitoneally injected with 0.5 units of regular human insulin (Actrapid Penfill; Novo Nordisk, Paris, France). Tissues were snap frozen in liquid nitrogen 10 min later.

Total cholesterol, HDL and LDL cholesterol, triglycerides, and nonesterified fatty acids (NEFA) levels were determined by enzymatic colorimetric assays (SOBIODA, Montbonnot–Saint-Martin, France). Leptin, adiponectin, and NOV levels were measured using ELISA-specific kits (R&D Systems, Lille, France).

The pancreatic and islet insulin content was extracted at −20°C in acidic ethanol (1.5% [vol/vol] HCl in 75% [vol/vol] ethanol). Pancreatic islets from NOV^−/− and WT mice were isolated after collagenase digestion and cultured in RPMI-1640 containing 10% FBS at 37°C for 24 h. Batches of 30 islets were incubated in Krebs buffer containing 2.8 mmol/L glucose, 16.7 mmol/L glucose, or 50 mmol/L KCl for 1 h at 37°C. Plasma insulin, total pancreatic and islet insulin content, and insulin secreted in response to glucose were measured using the Ultra Sensitive Insulin ELISA kit (Crystal Chem, Inc., Downers Grove, IL).

Paraffin-embedded adipose sections were analyzed after hematoxylin and eosin staining. Five fields were randomly observed at original magnification ×10 by an optical microscope (BX61 Olympus). Adipocyte quantification was performed using ImageJ software (http://rsbweb.nih.gov/ij/) on ∼500–600 cells per mouse (4 animals per group and per genotype were analyzed).

Considering that adipocytes constituted 90% of the adipose tissue volume, the number of cells was estimated by dividing the tissue mass by the mean volume of cells. The results are expressed in cells per gram of tissue.

Sections (3-µm thick) were stained with Sirius Red, and interstitial fibrosis was assessed semiquantitatively by randomly selecting 10 fields per section that covered the entire surface at original magnification ×200. Fibrosis was then quantified using computer-based morphometric Analysis software (Olympus) that allowed the formation of a binary image in which the stained area could be automatically calculated as a percentage of the image area (27).

eWAT from WT and NOV^−/− mice was removed, minced, and digested with collagenase A (1 mg/mL in PBS; Sigma-Aldrich) at 37°C for 45 min. Homogenates were passed through a 70-µm mesh, and the effluent was centrifuged at 500g for 3 min. Cells from pellets defined as the stromal vascular fraction (SVF) were resuspended in FACS buffer (PBS with 5% FBS). Extracellular staining was preceded by incubation with FcR-blocking antibody (2.4G2; BD Biosciences, Franklin Lakes, NJ) to avoid nonspecific staining. For lymphocytes cytometry analysis, cells were first stained with monoclonal antibodies to cell-surface markers purchased from BD Biosciences; fluorescein isothiocyanate–conjugated anti-CD3 (145-2C11), allophycocyanin-conjugated anti-CD4 (RM4-5), and peridinin-chlorophyll protein–conjugated anti-CD8 (53-6.7). Cells were then fixed, permeabilized, and incubated with phycoerythrin (PE)-conjugated anti-Foxp3 (FJK-16s, eBioscience) for intracellular staining according to the manufacturer’s specifications. For staining analysis of macrophages, cells were surface-stained with BV421-conjugated anti-F4/80 (T45-2342), Alexa Fluor 647–conjugated anti-CD206 (19.2), and biotin-conjugated anti-CD11b (M1/70) revealed with PE-conjugated streptavidin all from BD Bioscience. Fluorescein isothiocyanate-conjugated anti-CD11c (N418) and anti–NOS2-PE cyanine 7 were from eBioscience. Stained cells were analyzed using a Gallios flow cytometer (Beckman Coulter, Inc., Villepinte, France), and data were processed using Kaluza software (Beckman Coulter).

eWAT from WT and NOV^−/− mice was removed and digested as described above. Homogenates were centrifuged, and isolated adipocytes were obtained after centrifugation. Glycerol secreted from isolated adipocytes was used as an index of lipolysis and measured with the Free-Glycerol reagent (Sigma-Aldrich). Triacylglycerols were extracted from adipocytes by the Dole method (28) and from liver by acetone and then tested with the triglycerides reagent (Sigma-Aldrich).

Primary preadipocytes derived from SVF of scWAT or eWAT were plated on six-well culture dishes and cultured in DMEM (4.5 g/L d-glucose) with 10% FCS (Eurobio, Les Ulis, France). Cell proliferation was assessed by BrdU incorporation according to the manufacturer’s instructions (Cell Signaling, Danvers, MA). Adipocyte differentiation was initiated in confluent cultures of primary and 3T3-L1 preadipocytes with DMEM supplemented with 10% FCS and 0.1 mmol/L 3-isobutyl-1-methylxanthine, 200 nmol/L dexamethasone, and 100 nmol/L insulin (Sigma-Aldrich) for 3 days. Adipocytes were further maintained for 6 days in DMEM supplemented with 10% FCS and 100 nmol/L insulin.

Bone marrow mononuclear (BMM) phagocytic precursor cells were isolated from femurs and tibiae of HFD-fed WT and NOV^−/− mice, as previously described (29). These precursors were differentiated into adherent mature macrophages (BMM) for 6 days in noncoated Petri dishes in DMEM (4.5 g/L d-glucose) with 20% decomplemented FCS (Eurobio) containing 20 ng/mL macrophage colony-stimulating factor (PeproTech, Neuilly-sur-Seine, France). BMM were seeded at 2 × 10⁶ cells/well in six-well plates and polarized toward a M1 phenotype with 10 ng/mL lipopolysaccharide (LPS; Sigma-Aldrich) or toward a M2 phenotype with 10 ng/mL interleukin 4 (IL-4; PeproTech) for 24 h. Cytokine secretion into the cell culture medium of BMM was subsequently analyzed using tumor necrosis factor α (TNF-α) for M1 or IL-10 for M2 phenotype ELISA kit (PeproTech).

Peritoneal macrophages (PEM) were derived from HFD-fed WT and NOV^−/− mice after an intraperitoneal administration of PBS 1× (5 mL). The peritoneal wash was recovered from injected mice, and PEM were collected by centrifugation at 600g at 4°C for 5 min. PEM, at a density of 10⁶ cells/well in a six-well plate, were polarized toward an M1 or an M2 phenotype as described for BMM. BMM or PEM were then processed as mentioned for FACS analysis.

Total RNA was extracted from mouse tissues, primary preadipocytes, adipocyte cultures, and from 3T3-L1 cells using RNeasy Lipid Mini or RNeasy Mini Kits (Qiagen, Les Ulis, France). Reverse transcription and real-time semiquantitative PCR (qPCR) amplification on an ABI 7300 apparatus (Applied Biosystems) were performed as previously described (13). Specific primers (Proligo; Sigma-Aldrich) used for the amplification of target genes are presented in Supplementary Table 1. The comparative Ct method (30) was used to calculate gene expression values. The ribosomal S26 was used for normalization as a housekeeping gene.

Small interfering RNAs (siRNA) sequences predesigned by Qiagen are listed in Supplementary Table 2. 3T3-L1 cells plated at 9 × 10³ cells/cm² in DMEM containing 10% FCS were transfected, 1 day later, with siRNA (120 pmol in 5 µL) using the RNAi Max reagent (Qiagen). Differentiation was initiated 72 h after transfection (day 0). On day 3 after the induction of differentiation, total RNA extracted from the cells was subjected to real-time qPCR. For the study of insulin signaling, 3T3-L1 cells were transfected on day 3 with siRNA (240 pmol in 10 µL); then, on day 5, the cultures were deprived of serum and insulin and were further treated on day 6 with insulin (10⁻⁸ mol/L) for 8 min. RNA was obtained from parallel cultures and underwent real-time qPCR.

3T3-L1 adipocytes or mouse tissues were lysed, and equal protein amounts (20 µg) were separated by SDS-PAGE. The conditioned medium corresponding to 10⁶ 3T3-L1 adipocytes was used for ELISA tests or for NOV expression.

Recombinant human NOV protein was obtained as previously described (15,17). For NOV protein detection, we used a homemade affinity-purified rabbit antibody (referred as CT-μ, anti-mouse NOV antibody) (31).

Rabbit polyclonal antibodies against total Akt, phosphorylated (p)-Akt (P-Ser 473), total Erk, and p-Erk, were from Cell Signaling; uncoupling protein 1 (UCP-1), peroxisome proliferator-activated receptor γ (PPAR-γ) coactivator 1-α (PGC1-α), poly(ADP-ribose) polymerase 1 (Parp-1) were from Abcam (Paris, France), and mouse monoclonal antibodies against β-actin, β-tubulin, and GAPDH were from Sigma-Aldrich.

The results are expressed as the mean ± SEM. Variance equality was analyzed by an F test (Excel; Microsoft Corp., Issy-Les-Moulineaux, France). Comparisons between groups in the animal studies were carried out using a parametric Student t test or by a nonparametric Mann-Whitney-Wilcoxon test (Minitab, Paris, France). A P value of <0.05 was considered statistically significant.

We addressed the effect of NOV deficiency on weight gain in mice fed the SD or HFD for 16 weeks. No difference was observed in mice fed the SD between the two groups of mice (Fig. 1A and B). However, NOV^−/− mice fed the HFD gained less weight than WT controls (Fig. 1C). The BW gain of NOV^−/− mice was markedly lower compared with WT mice (Fig. 1D). NOV^−/− mice fed the HFD also displayed a decrease in fat and relative fat mass (Fig. 1E and F) and a slight increase in relative lean body mass compared with their respective controls; however, the absolute lean mass was unchanged (Fig. 1G and H), and there was no difference in mouse-tail length (96 ± 0.83 mm for WT and 96.5 ± 1.32 mm for NOV^−/−).

Indirect calorimetry was used to intensively analyze energy intake and energy expenditure of the mice. No significant change in food intake (Fig. 1I) or in spontaneous locomotor activity (Fig. 1K and Supplementary Fig. 2C) was observed. A slight increase in energy expenditure was detected in NOV^−/− mice but was limited to the daylight period (Fig. 1J and Supplementary Fig. 2B). The RER (VCO₂-to-VO₂ ratio) did not change between NOV^−/− and WT mice (Fig. 1L and Supplementary Fig. 2D).

To examine whether the marked differences in BW gain and adiposity between the two genotypes were associated with changes in carbohydrate metabolism, we evaluated glycemia and insulinemia under basal and dynamic conditions (Fig. 2). During the OGTT, HFD-fed (Fig. 2B) but not SD-fed (Fig. 2A) NOV^−/− animals showed improved glucose tolerance compared with WT mice. This correlated with a significant decrease in the area under the curve for NOV^−/− mice compared with controls (Fig. 2B, upper panels). Of note, fasting glucose levels were slightly lower in HFD-fed NOV^−/− mice compared with their WT controls. We then conducted an intraperitoneal ITT (Fig. 2C and D), which demonstrated an improved insulin sensitivity in HFD-fed NOV^−/− mice compared with their WT littermates (Fig. 2D). The improved glucose tolerance observed in HFD-fed NOV^−/− mice was not accompanied by an increased insulinogenic index after the oral glucose challenge compared with the WT controls (Fig. 2E). Because a previous study reported decreased insulin release subsequent to NOV overexpression in a rat β-cell line (25), insulin secretion was tested in islets isolated from WT and NOV^−/− mice fed the HFD. Insulin secretion similarly increased in control and NOV^−/− islets when exposed to 16.7 mmol/L glucose or 50 mmol/L KCl compared with 5.5 mmol/L glucose (Fig. 2F). The pancreatic and islet insulin contents were also similar between the two genotypes. Further pancreatic morphometry analysis showed that β-cell fraction, β-cell mass, mean islet density, and mean islet size were similar in NOV^−/− and WT mice fed the HFD (Supplementary Fig. 3).

Collectively, these observations show that on an HFD, NOV deficiency decreased weight gain and improved glucose tolerance and insulin sensitivity, while insulin production remained unchanged.

On the SD, the organ weights–to–BW ratios were not significantly different between the two genotypes (Table 1). However, on the HFD, a striking and similar decrease of these ratios was observed in NOV^−/− mice compared with the WT controls for all the WAT depots. We also assessed the adipocyte size in hematoxylin and eosin–stained eWAT and scWAT tissue from WT and NOV^−/− mice fed the HFD. Of interest, we observed that NOV deficiency influenced the adipocyte size distribution in both fat tissues (Fig. 3). The mean adipocyte surface area in eWAT and scWAT was markedly reduced in NOV^−/− compared with WT mice (Fig. 3A and B). Moreover, χ² analysis on more than 1,500 independent determinations showed that in both eWAT and scWAT, NOV^−/− displayed a higher proportion of small adipocytes compared with those of the WT mice (P < 0.0001) (Fig. 3C and D). Adipocyte cellularity was also evaluated in eWAT and scWAT with the assumption that adipocytes represent ∼90% of adipose tissue volume (32). The results revealed no difference in the number of adipocytes in scWAT between WT and NOV^−/− mice (3.10 ± 0.41 and 3.1 ± 0.46 × 10⁶ cells/g of tissue, respectively). In contrast, the adipocyte number in eWAT was significantly increased in NOV^−/− mice (3.3 ± 0.47 vs. 2.1 ± 0.32 × 10⁶ cells/g of tissue, respectively; P < 0.05).

We also analyzed several plasma metabolic parameters, which are summarized in Table 1. On the SD, no differences could be detected between the two genotypes. In WT mice, the HFD induced a significant increase in the concentrations of triglycerides, cholesterol (C), HDL-C, NEFA, and leptin. In NOV^−/− mice, the HFD also provoked an increase in cholesterol, triglycerides, HDL-C, and leptin, but not NEFA. However, compared with WT, the increases in cholesterol, HDL-C, and leptin were significantly reduced in NOV^−/− mice. A trend (P < 0.1) toward a lower concentration of LDL-C was also observed in these mice.

Basal or induced lipolysis was also investigated in adipocytes isolated from eWAT fat depots from HFD-fed mice. As shown in Supplementary Fig. 4, basal and isoproterenol-stimulated lipolysis was stimulated in adipocytes from NOV^−/− compared with those of WT animals.

The liver phenotype of WT and NOV^−/− mice fed the HFD was also characterized. Analysis of hepatic histology, illustrated in Supplementary Fig. 5A, revealed less steatosis in NOV^−/− mice. Indeed, the hepatic triacylglycerol content was reduced by approximately threefold (Supplementary Fig. 5B). This observation was consistent with decreased mRNA levels of the fatty acid transporter CD36 and its main transactivator PPAR-γ (Supplementary Fig. 5C). By contrast, no significant difference was detected for genes involved in fatty acid oxidation (PPAR-α, ACOX1), lipogenesis (SREBP-1c, ACC) or neoglucogenesis (PEPCK1, G6Pase). Interestingly, we did not detect any change in several liver inflammation markers, including chemokine (C-C motif) ligand 2 (CCL2), TNF-α, CD68, and F4/80 (Supplementary Fig. 5D).

Because a difference in adipogenesis could at least partly explain the reduced adipocyte size and adipose mass found in HFD-fed NOV^−/− mice, we first investigated NOV expression in differentiating 3T3-L1 cells. As shown in Supplementary Fig. 6A, NOV mRNA and protein expression was strongly induced during differentiation. In agreement with this observation, there was an ∼10-fold increase in the NOV transcript level after adipocyte differentiation in primary cultures (Supplementary Fig. 6B). These findings suggested that NOV could be involved in adipogenesis. Thus, we then analyzed the expression of the three major transcription factors induced during adipogenesis in eWAT, perirenal WAT (prWAT), and scWAT. As shown in Fig. 4A–C, we did not find any significant differences in the expression of SREBP-1c, PPAR-γ, and C/EBP-α between WT and NOV^−/− mice in the three fat depots, but the pattern of these mRNAs did not exclude the possibility that NOV could modulate adipogenesis. To examine whether an alteration in adipocyte differentiation was an intrinsic property of NOV-deficient mice, we isolated SVFs from eWAT and scWAT and tested their in vitro adipogenic capacity. Interestingly, we showed (Fig. 4D and E) that preadipocytes originating from NOV^−/− mice had a greater adipocyte differentiation ability than those from WT mice, as assessed by the increase in SREBP-1c, PPAR-γ, and C/EBP-α expression. In line with this result, when NOV expression was suppressed in 3T3-L1 preadipocytes by transfection with an siRNA specific for NOV, the expression of SREBP-1c, PPAR-γ, and C/EBP-α was also increased compared with the same cells transfected with a nonsilencing siRNA (Fig. 4F). We did not observe any difference in the proliferation capacity of preadipocytes derived from WT or NOV^−/− mice (Supplementary Fig. 6C).

These results show that under HFD, the relative resistance to obesity and insulin resistance of NOV^−/− mice might be related to a higher proportion of small adipocytes that likely exhibit better intrinsic adipogenic ability in the absence of NOV.

NOV has been involved in the regulation of various inflammatory molecules (10,20) that are known to impair insulin signaling and promote insulin resistance in metabolic tissues. Consistent with enhanced insulin action and sensitivity (Fig. 2), we observed a specific pattern of altered inflammatory gene expression in different adipose tissues from HFD-fed NOV^−/− mice (Fig. 5). In eWAT and prWAT, transcript levels of TNF-α, CCL2, and F4/80 were decreased (Fig. 5A and B), whereas the expression of these genes did not differ from WT in scWAT (Fig. 5C). We also detected a decrease in leptin levels in eWAT and scWAT in NOV^−/− animals, consistent with their reduced leptin plasma levels (Table 1). Thus, in NOV^−/− mice fed the HFD, the reduced proinflammatory pattern of gene expression essentially affects the visceral fat depots, which are generally considered to be crucial in the control of metabolic status. Consistent with these data, we found that treatment of 3T3-L1 cells with NOV recombinant protein led to a striking increase in CCL2 expression at both the mRNA and protein level (Fig. 5D and E).

It is now largely recognized that rodent or human obesity is associated with a low-grade inflammation of adipose tissue and its infiltration by immune cells, including lymphocytes and macrophages, that shift from a M2-like anti-inflammatory to a M1-like proinflammatory phenotype (33). In addition, these changes in macrophage phenotypes are likely involved in obesity-linked insulin resistance (34). Given the metabolic phenotype and the profile of gene expression in deep fat depots of NOV^−/− animals, we further characterized the immune cell populations present in eWAT of NOV^−/− and WT mice. The absence of NOV was associated with an increase in a population of macrophages expressing high levels of the M2-like marker CD206 (Fig. 5F) and with a trend to fewer macrophages displaying an apparent M1-like profile (CD11c⁺). No difference in the overall frequency of CD3⁺ lymphocytes population was detected between the two genotypes (data not shown). Although no difference was observed in the frequency of CD8⁺T cells, the percentages of CD4⁺ lymphocytes among infiltrating T cells were strongly decreased in NOV^−/− compared with WT (Fig. 5G, left panel). This was related to significantly reduced frequency of conventional CD4⁺Foxp3^– T cells, with no difference in percentages of CD4⁺Foxp3⁺ regulatory T cells (Fig. 5G, right panel). Interestingly, a large proportion of infiltrating CD3⁺ lymphocytes were identified CD4^–CD8^– (double-negative) T cells, the frequency of which was significantly increased in NOV^−/− compared with WT mice (Fig. 5G, left panel).

We further analyzed whether NOV deletion can affect M1 versus M2 polarization in myeloid/macrophage cells. Macrophages were derived from BMM (or PEM) of WT and NOV^−/− HFD-fed mice. NOV expression was reported in human macrophages (21); however, its expression in macrophages from HFD-fed WT mice has not been investigated so far. In our experimental conditions, the levels of NOV mRNA in BMM or PEM derived from HFD-fed WT mice were below the threshold of detection (data not shown). Nevertheless, because a transient expression of NOV during macrophage maturation or differentiation could affect their susceptibility to proinflammatory molecules, we subjected WT and NOV^−/− BMM or PEM cultures to polarization to a M1-like or M2-like phenotype with LPS and IL-4, respectively (32). Compared with untreated control cultures, the mRNA levels of M1-like proinflammatory markers, such as TNF-α and CCL2, were significantly induced in BMM cultures treated with LPS, whereas IL-4 markedly enhanced the M2-like marker CD206 (Supplementary Fig 7A). We also observed in our experimental conditions, as previously described (35), that the mRNA levels of the M2-like marker IL-10 was considerably induced by LPS. Analysis of TNF-α and IL-10 proteins confirmed the mRNA results (Supplementary Fig. 7B).

Consistent with these results, we also found that among the CD11b⁺F4/80⁺ population of BMM treated with LPS there was a significant increase of macrophages expressing the M1-like marker inducible nitric oxide synthase 2, whereas IL-4 markedly enhanced the percentage of cells expressing the M2-like marker CD206 (Supplementary Fig. 7C). Overall, mRNA, secretion, or FACS analyses detected no significant difference between WT and NOV^−/− BMM under these different treatments (Supplementary Fig. 7A–C). Similar results were also obtained with PEM (data not shown). Thus, whereas NOV^−/− were protected against inflammation in adipose tissue, it does not seem to be due to the absence of NOV in macrophages.

Because limited inflammation and reduced fibrosis were previously described in NOV^−/− mice in the unilateral ureteral obstruction experimental model of nephropathy (36) and because fibrosis in adipose tissue is also linked to insulin-resistance (37), we also studied the effect of NOV deletion on fibrosis of the adipose tissue. Analysis of the interstitial collagen accumulation, assessed by Sirius Red coloration, revealed a trend for decreased fibrosis in eWAT of NOV^−/− animals and no significant difference between WT and NOV^−/− mice in the scWAT (Supplementary Fig. 8).

A slight increase in energy expenditure was observed in NOV ^−/− mice (Fig. 1J). Because an increased turnover of adipose cells in these mice could be energetically more expensive, we first analyzed markers of apoptosis in adipose tissue. Western blots did not reveal any significant differences in the Parp-1 cleavage in scWAT of both genotypes on the HFD, whereas Parp-1 cleavage in eWAT was even decreased in NOV^−/− animals (Supplementary Fig. 9). Thus, these results could not account for the lower BW gain of NOV^−/− mice.

We next analyzed the expression of genes in the brown adipose tissue (BAT) and WAT known to be involved in energy expenditure. In HFD-fed NOV^−/− mice, the expression of UCP-1 was increased in BAT and eWAT at both mRNA and protein levels (Fig. 6). The expression of PGC1-α, a major coregulator that cooperates with PPAR-γ to transactivate the UCP-1 gene promoter (38), was also significantly increased in the BAT of these mice (Fig. 6).

To further delineate the molecular mechanisms involved in the improved insulin sensitivity in NOV^−/− mice, we analyzed the effect of NOV deficiency on insulin signaling. We found that in 3T3-L1 adipocytes transfected with two different siRNAs specific for NOV, the decrease in NOV expression was accompanied by a marked increase in insulin-stimulated p-Akt and p-Erk compared with the same cells transfected with a nonsilencing siRNA as a control (sc) (Fig. 7A–C). In vivo, we also observed that an acute exposure to insulin resulted in a significant increase in p-Akt in eWAT of NOV^−/− compared with WT mice (Fig. 7D and E) but not in scWAT (data not shown).

In this study, we report for the first time that NOV deficiency protects mice fed an HFD against obesity, likely through an increase in energy expenditure. We also report that in NOV^−/− mice on an HFD, the inflammation of the eWAT and prWAT was reduced compared with WT mice. Taken together, these differences between WT and NOV^−/− mice likely account for the improved glucose tolerance, insulin sensitivity, and metabolic profile of NOV^−/− mice.

The increased population of small adipocytes in eWAT could be due to the better proliferation or the better differentiation ability of the resident preadipocytes. Our data are in favor of the second hypothesis because we did not observe any difference in the proliferation rate of primary preadipocytes derived from WT or NOV^−/− mice. In contrast, the absence of NOV had a positive effect on the differentiation ability of these cells. Moreover, different data have reported that NOV could play an effective role in the control of cell differentiation because it can inhibit myoblast and osteoblast differentiation (13,39,40). It is also crucial for primitive hematopoietic progenitor cell activity (14). However, although the improved ability to differentiate in NOV^−/− preadipocytes was observed in both eWAT and scWAT, the cellularity was only increased in eWAT. This observation is in accordance with recent data (41) showing that in mice fed an HFD for over 1 month, adipogenesis is preferentially initiated in eWAT compared with scWAT. This potentially improved ability for preadipocyte differentiation, in which NOV is disrupted or inhibited, is apparently contradictory with the decreased fat mass in NOV^−/− mice and with the increase in NOV expression in vitro during the differentiation process. Our results suggest that the absence of NOV favors the recruitment of smaller adipocytes that are more sensitive to insulin. However, a longer follow-up of WT and NOV^−/− mice would be required to examine whether the recruitment of small adipocytes maintains the limited expansion of fat mass in NOV^−/− mice or whether this is overcome over time by increased cellularity.

Taking into account the whole body, this fat mass loss occurs while there is no decrease in food intake, and preliminary analysis of lipids in feces did not reveal any significant difference in gut absorption of lipids between NOV^−/− and control mice (Supplementary Fig. 10). However, there is a slight but significant increased energy expenditure during the daylight period. Remarkably, this enhanced energy expenditure is associated with an induction in BAT and eWAT in the expression of UCP1 and PGC1-α, two key effectors of the thermogenic pathway. This could represent a major phenomenon to protect NOV^−/− mice from HFD-induced obesity and insulin resistance. However, the molecular mechanisms by which the absence of NOV leads to UCP1 induction remain unknown. For instance, the increased lipolytic activity of NOV^−/− adipocytes could also promote substrate availability for thermogenesis.

In this study, we report that the effect of the absence of NOV is only detectable when mice are challenged with an HFD. Indeed, two reports describing the phenotype of NOV^−/− mice revealed that they are viable and apparently normal and fertile (22,40). However, after a lesion of the femoral artery, the null mice presented an enhancement of neointimal thickening compared with WT mice during vascular repair (22). Another example showed that after an injury to the femur, bone regeneration in NOV^−/− mice was accelerated compared with WT mice (40). More recently, we also found that during the progression of obstructive nephropathy, NOV^−/− mice displayed limited inflammation and renal interstitial fibrosis compared with WT mice (36).

An HFD is known to induce low-grade inflammation in adipose tissue (42) and is associated with the increased expression of NOV mRNA in adipose tissue and NOV plasma levels (21). Under these specific nutritional conditions, the absence of NOV not only limits adipose tissue expansion but is also associated with improved glucose tolerance. This is primarily related to better insulin sensitivity, as supported by ITTs and by the normalization of plasma NEFA levels in NOV^−/− mice fed the HFD, suggesting the restoration of the antilipolytic effect of insulin. These results are strengthened by in vitro and in vivo experiments showing that NOV deficiency in adipocytes promotes insulin signaling.

Our results also strongly suggest that NOV could be a key player in adipose tissue inflammation and that this in turn could promote insulin resistance. Recent studies have shown that CD4⁺ and CD8⁺ T cells play a major role in promoting adipose tissue inflammation by secreting cytokines and recruiting macrophages during HFD (43). NOV deficiency led to a strong decrease in the frequency of infiltrating conventional CD4⁺ T cells, whereas no difference was observed for CD8⁺ or CD4⁺ regulatory T cells.

Of note, NOV^−/− mice displayed a significantly increased subpopulation of CD3⁺ T lymphocytes that were CD4^–CD8^–. These particular T cells that lack the expression of both CD4 and CD8 T-cell coreceptors have been studied in mice and humans for their contribution to peripheral tolerance and disease prevention (44). Our current findings are consistent with these observations. The reduced TNF-α and F4/80 expression in two visceral WAT (eWAT and prWAT) depots from HFD-fed NOV^−/− mice reflects the decreased inflammation in these fat depots and is in agreement with the involvement of these tissues in insulin sensitivity (45–47). Interestingly, a decrease in TNF-α and F4/80 expression was not detected in scWAT, which is generally considered to have an accessory role in the insulin resistance phenotype. FACS analysis of eWAT suggests that under HFD, the absence of NOV was associated with a switched overbalance from M1-like to M2-like macrophage population. However, our in vitro analysis revealed that the absence of NOV did not influence the polarization of macrophages toward a M1 or M2 phenotype. This strongly suggests that the switch from M1-like to M2-like macrophage population does not result from an autocrine action of NOV on macrophages but rather from an indirect role of NOV on the expression of proinflammatory molecules by adipocytes and possibly also by other cells present in the adipose tissue. Given the documented protective role of such M2-like populations against obesity-linked insulin resistance, this suggests that these changes may also contribute to the improved insulin sensitivity and glucose tolerance of NOV^−/− mice.

Our in vitro and in vivo experiments revealed that CCL2 expression is modulated by NOV. This regulation may be considered to be pivotal in the development of insulin resistance because targeted deletions in mice for the CCL2 and/or CCR2 genes result in reduced macrophage content, decreased inflammation in fat, and protection from HFD-induced insulin resistance; conversely, mice overexpressing CCL2 in adipose tissue have increased local macrophage infiltration along with insulin resistance (47–49). The specific macrophage marker F4/80 expression is reduced in NOV^−/− mice WAT. This suggests that the absence of NOV could decrease CCL2 with a subsequent reduction in macrophage number within the adipose tissue. However, the regulation of CCL2 by NOV could also have an effect on insulin sensitivity through additional pathways because it has also been shown that an increase in the plasma levels of CCL2 is sufficient to induce insulin resistance independent of macrophage infiltration into adipose tissue (50). Accordingly, CCL2, through its angiogenic effect on endothelial cells, could also contribute to the expansion and remodeling of adipose tissue (51).

During the OGTT, we did not observe a higher plasma insulin concentration in NOV^−/− mice compared with WT controls. Insulin release in isolated islets was also comparable between the two genotypes, and the total content of insulin in the whole pancreas or in isolated islets was similar in NOV^−/− and WT mice. These results are apparently discordant with those reported by Paradis et al. (25). Nevertheless, the experimental conditions used by that study and ours substantially differ because they used a rat cell line overexpressing NOV. Taken together, our results suggest that the increase in insulin sensitivity is the major mechanism by which glucose tolerance is improved in NOV^−/− mice.

In summary, the current study demonstrates that NOV participates in the development of obesity-induced insulin resistance. These results indicate that the use of counteracting agents against NOV represents a potential strategy for new therapies against obesity and its metabolic comorbidities.

Acknowledgments. The authors acknowledge L. Dinard from the animal facilities platform (PHEA Saint-Antoine) for her technical support, A. Munier from the cytometry platform (UMS, Lumic 030) for her help in FACS analysis, R. Morichon from the image platform (UMS, Lumic 030) for his help in microscopy analysis, and S. Dumont and F. Merabetene for their contribution in adipose tissue histomorphological analysis (UMS, Lumic 030). The authors also acknowledge the technical platform Functional and Physiological Exploration (FPE) of the Unit “Biologie Fonctionnelle et Adaptative” (Sorbonne Paris City University, Paris Diderot University, BFA, CNRS, Paris, France) for metabolic and behavioral analysis.

Funding. This work was financed by INSERM, Pierre et Marie Curie University, and by a grant from the French Society of Diabetes. T.T.H.D. received financial support from the French Society of Endocrinology and the French Embassy in Vietnam, and P.-O.M. received financial support from ANRT (National Association of Research and Technology).

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

Author Contributions. C.M. and B.F. designed the research. M.G., T.T.H.D., B.A., M.M., G.D., C.K., M.A., M.B., T.L., P.-O.M., M.F., A.B., B.B., and R.G.D. performed the research. H.K., S.H., and C.E.C. contributed new reagents/analytic tools. C.M., B.A., M.M., G.D., B.B., R.G.D., S.L., and B.F. analyzed data. C.M. and B.F. wrote the paper. B.F. 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.