Nck2 Deficiency in Mice Results in Increased Adiposity Associated With Adipocyte Hypertrophy and Enhanced Adipogenesis

In humans, obesity is a strong determinant condition for the development of metabolic disorders such as type 2 diabetes. Obesity is characterized by an excessive expansion of white adipose tissue (WAT) that relies on hypertrophy of preexisting adipocytes and the generation of mature adipocytes through growth and differentiation of preadipocytes after adipogenesis (1). At the molecular level, adipogenesis is regulated by a timely transcriptional network involving CCAAT-enhancer–binding protein (C/EBP) transcription factors, with C/EBPδ and β in the early stages of differentiation, whereas C/EBPα in concert with peroxisome proliferator–activated receptor-γ (PPARγ) promotes adipocyte maturation. A genome-wide association study for adiposity in the Framingham cohort found, as expected, single nucleotide polymorphisms in well-characterized adipocyte markers, such as PPARG and ADIPOQ (2). Of note, among significant single nucleotide polymorphisms associated with adiposity, this study also identified rs10496393 in the gene area encoding of the Src homology (SH) domain containing adaptor protein Nck2.

In mammals, two genes encode for the closely related Nck1 and Nck2 proteins, which are essentially composed of SH domains with three N-terminal SH3 and one C-terminal SH2 domains (3). Nck proteins are well-known to assemble molecular complexes that mediate canonical signaling from activated membrane receptors regulating cytoskeletal reorganization (4,5). In addition, Nck1 and Nck2 are implicated in noncanonical signaling pathways through their ability to regulate the unfolded protein response (UPR) (6–8). The UPR is initiated at the level of the endoplasmic reticulum (ER) by three ER transmembrane sensors: the double-stranded RNA-like ER kinase (PERK), Ser/Thr kinase/endoribonuclease inositol–requiring enzyme-1α (IRE1α), and activating transcription factor 6 (ATF6) (9). We and others implicated Nck in regulating the UPR through its interaction with the β-subunit of the eukaryotic initiation factor 2 (eIF2β) (8,10,11), Ser/Thr phosphatase PP1 (8,12), PERK (13,14), and IRE1α (7). We demonstrated that Nck1 is essential for sustained hepatic activation of the IRE1α-JNK pathway associated with impaired glucose homeostasis secondary to obesity in mice (15). We also have showed that Nck1 modulates PERK-dependent regulation of insulin biosynthesis and survival in pancreatic β-cells (13,14). In the current study, we report that Nck2 is required to regulate PERK activity and signaling during adipogenesis and in mature adipocytes to prevent abnormal WAT expansion and dysfunction associated with metabolic disorders.

Nck2^−/− and Nck2^+/+ mouse littermates were generated as previously described (16). Male mice were used in all experiments according to approved protocol 5069 by the McGill University animal care committee.

The antibodies used were as follows: Hsp90 (4877S), Akt (9272), pAkt Thr³⁰⁸ (9275L), pAkt Ser⁴⁷³ (9271L), eIF2α (9722S), fatty acid synthase (FAS) (3180), aP2 (3544), adiponectin (2789), perilipin (9349), acetyl-CoA carboxylase (ACC) (3676), PERK (3192), and β-actin (4967S) from Cell Signaling Technology; PPARγ (sc-7196), ATF4 (sc-200), pPERK Thr⁹⁸⁰ (32577), and RasGAP (sc-63) from Santa Cruz Biotechnology; human Nck2 (TA307351) from Origene, PEPCK (1002S) from Cell Applications; peIF2αSer⁵¹ (44728G) from Invitrogen; Flag (3165) from Sigma; and Nck1 and panNck as previously described (10,15). The 3T3-L1 cell line from ATCC was cultured as recommended by the manufacturer.

Fat mass was determined by DEXA (Lunar PIXImus II; GE Healthcare). Glucose (1–2 g/kg), insulin (0.75 units/kg), and sodium pyruvate (2 g/kg) were injected intraperitoneally, and blood glucose was quantified with an Accu-Chek glucometer (Roche). With use of a TSE PhenoMaster, metabolic parameters and locomotor activity were recorded according to the manufacturer. Commercial kits were used to determine serum insulin, tumor necrosis factor-α (TNF-α), interleukin (IL) 6, and adipokine levels (Meso Scale Discovery) as well as triglycerides (TGs) and nonesterified fatty acids (NEFAs) (Sigma). As recommended by the manufacturer, hepatic and skeletal muscle glycogen contents were evaluated using boiled tissue extracts and a glycogen assay kit from Sigma (MAK016).

Primary hepatocytes were prepared as previously described (17). Insulin dose response (10 and 100 nmol/L, 15 min) was performed 2 days after initial plating (18). To prepare the stromal vascular fraction (SVF), WAT depots were minced and digested with collagenase (1 mg/mL, C0130; Sigma). Digestion was stopped by adding ice-cold DMEM plus 10% FBS followed by successive centrifugation and filtration on prewet 70- and 40-μm cell strainers. SVFs were plated at 1 × 10⁵ cells in six-well plates. At confluence, differentiation was induced with 1 μmol/L dexamethasone, 1 μmol/L rosiglitazone, 0.5 mmol/L 3-isobutyl-1-methylxanthine (IBMX), and 3 μg/mL insulin for 3 days and then maintained in the same medium but without IBMX for 4 days.

Tissues and cell extracts were prepared as previously reported (15). Total proteins (20–50 μg) were resolved by SDS-PAGE, transferred onto polyvinylidene fluoride, and immunoblotted with indicated antibodies. For Nck distribution, tissue extracts from Nck1^−/− and Nck2^−/− mice were normalized for protein content (50 μg) and processed similarly using panNck antibodies.

Two days postconfluency, 3T3-L1 cells were differentiated by using a classical cocktail (1 μmol/L dexamethasone, 0.5 mmol/L IBMX, 1 μg/mL insulin) for 2 days and then incubated in DMEM plus 10% FBS containing only 1 μg/mL insulin for 2 days before being maintained in regular medium. For small interfering RNA (siRNA) experiments, 2 days before confluency, 3T3-L1 cells were reverse transfected with 1 nmol/L of Nck2 (Mouse)-3 unique 27mer siRNA duplexes (SR412820; Origene) using 7.5 μL Lipofectamine RNAiMAX Reagent (Invitrogen) in six-well plates. Lipid droplet formation was visualized using BODIPY 493/503 (Thermo Fisher Scientific) and confocal microscopy and quantified after oil red O (ORO) staining. For ORO staining, cells were fixed in 10% PBS-buffered formalin for 15 min, permeabilized using 60% isopropanol for 5 min, and stained with 0.18% ORO for 15 min. For quantification, ORO was eluted in 100% isopropanol for 10 min and read at 492 nm with a spectrophotometer (EnSpire 2300 Multilabel Plate Reader; PerkinElmer). For BODIPY 493/503, cells were incubated at 1 μg/mL for 10 min and wash twice before visualization with a confocal Zeiss microscope (LSM 510 META) or quantification using Infinite M200 PRO Tecan (excitation 500 nm, emission 550 nm). In vitro lipogenesis was assessed by using a classical oleate uptake protocol (19).

3T3-L1 cells Nck2 siRNA transfected or overexpressing Nck2 and their respective controls were plated at 5 × 10³ cells/well in 24-well plates. MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] activity was assessed to determine cell number on days 1–4 after plating as previously described (14). Freshly isolated WAT SVF from both mouse genotypes were washed twice and incubated for 30 min at 4°C with the following antibodies: PE/Cy7 CD29 (clone HMβ1-1; BioLegend 102221), allophycocyanin CD34 (clone HM34; BioLegend 128611), and Pacific Blue Sca-1 (clone D7; BioLegend 108119). After antibody incubation, cells were washed and analyzed on a BD FACSCanto II flow cytometer. For 3T3-L1 cells, after incubation with BODIPY (1 mg/mL) for 10 min at room temperature, cells were washed three times and analyzed using the same FACS analyzer.

Adipose tissues (7 μm) were embedded in paraffin and processed for hematoxylin and eosin (H&E) staining. Adipocyte density was quantified using ImageQuant (GE Healthcare Life Sciences). Hepatosteatosis was assessed by ORO staining. Livers were embedded in optimal cutting temperature matrix (CellPath) and kept at −80°C until analysis. Frozen liver sections (6 μm) were processed for ORO or H&E staining. Briefly, sections were dried for 10 min, rehydrated, and incubated for 30 min in 2% weight for volume ORO prepared in 50% acetone and 35% ethanol solution. Sections were counterstained with hematoxylin for 2 min.

Tissues were dissected and immediately snap frozen in liquid nitrogen for further analysis. RNA was extracted using TRIzol reagent (Invitrogen) according to manufacturer instructions. cDNA synthesis was performed by using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Briefly, 1 μg of RNA was reverse transcribed in a master mix solution containing reverse transcriptase, random primers, deoxynucleotide (dNTPs), and RNase inhibitor. The reaction was carried out at 25°C for 10 min, 37°C for 120 min, and terminated at 85°C for 5 min. Quantitative real-time PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) and run on a ViiA 7 thermal cycler system (Applied Biosystems) using specific primers. Briefly, PCRs were performed for 40 cycles following 95°C for 15 s, 60°C for 30 s, and 72°C for 30 s. Expression levels were calculated using the ΔΔCt method, and data were normalized to housekeeping gene cyclophilin B, the expression of which did not vary among treatments. Specific primers for PCR amplification of targeted genes were used, and their sequences are available upon request.

For both mouse genotypes, epididymal WAT (eWAT) total RNA (n = 2/genotype) was prepared as aforementioned and purified using RNeasy columns (QIAGEN). Expression libraries were generated using cBot clusters, and deep sequencing was performed using Illumina TruSeq RNA Sample Preparation Kit, Illumina TruSeq SR Cluster Kit v2, and Illumina TruSeq SBS Kit v2 (50 cycles) according to the manufacturer’s procedures. Sequencing was performed at the Génome Québec Innovation Centre (McGill University) by using the Illumina HiSeq 2000 platform. Reads were aligned to the GRCm38 genome with TopHat v2.0.10, and the raw alignment counts were calculated with HTSeq v0.5.3. The differential expression measurements were performed with DESeq2 v1.4.5. The generation of the Nck2^−/− signature was performed using DAVID v6.7 (Database for Annotation, Visualization and Integrated Discovery) (20,21) and GSEA v2.1.0 (Gene Set Enrichment Analysis) (22). For DAVID analyses, genes included in the studies had an adjusted P < 0.10 and an average differential expression of at least twofold. Principal component analyses of the gene expression data confirmed that the duplicates of the Nck2^+/+ and Nck2^−/− samples respectively cluster together, whereas between genotypes, the samples were well separated from one another.

Blood insulin levels before and after glucose (2 g/kg, 10 min) and insulin signaling (0.75 units/kg i.p., 15 min) in indicated tissues were assessed in overnight-fasted mice. Isolated pancreatic islets, as previously described (23), were subjected to an acid-ethanol extraction before determination of insulin content using radioimmunoassay (Linco Research) and expressed as normalized to DNA determined by SYBR Green.

Upon transfection using Lipofectamine 2000 (Invitrogen) and G418 selection, stable pools of mock or Flag-Nck2 expressing 3T3-L1 cells were submitted to differentiation as aforementioned. Transient PERK wild-type overexpression was achieved after pcDNA3.1-PERK transfection using Lipofectamine 3000 (Invitrogen) 48 h before adipocyte differentiation. PERK inhibitor (GSK2606414) was added at 10 nmol/L during differentiation of 3T3-L1 cells.

Human subcutaneous WAT (scWAT) and omental WAT (oWAT) biopsy specimens from male subjects (BMI 35.5–69.8 kg/m²) paired for age and date of bariatric surgery were obtained from the Biobank of Institut Universitaire de Cardiologie et de Pneumologie de Québec (IUCPQ), where written informed consent was obtained from the subjects according to institutionally approved management modalities. Study approval was obtained from the ethics committees of both IUCPQ and McGill institutions.

Data from each group were compared by unpaired Student t test or two-way ANOVA using Prism software (GraphPad Software), and P < 0.05 was considered significant.

Analysis of Nck protein expression in mouse tissues by Western blotting revealed that Nck1 is expressed in all tissues tested (Fig. 1A). In contrast, Nck2 is highly expressed in both lungs and eWAT and at very low levels in pancreas, brown adipose tissue (BAT), testis, and spleen, whereas it was not detected in other tissues tested (Fig. 1A). In addition, Nck2 is preferentially expressed in WAT compared with BAT, with the highest level in eWAT compared with scWAT (Fig. 1B).

Excluding the lungs, Nck2 is typically detected at high levels only in WAT, we used Nck2 knockout mice to explore whether Nck2 plays a role in WAT. Nck2^−/− mice displayed significantly larger WAT depots, whereas body weight, mouse size, and other tissues weight were comparable to Nck2^+/+mice (Fig. 2A and B and data not shown). Accordingly, DEXA confirmed a significant increase in total fat and fat mass in Nck2^−/− mice, but lean mass was not changed (Fig. 2C). Total bone area was slightly, but not significantly, reduced in Nck2^−/− mice compared with Nck2^+/+ mice (data not shown), potentially masking the impact of increased fat mass on total body weight. Increased adiposity in Nck2^−/− mice was progressive (Fig. 2D), and analysis of eWAT at 24 weeks postweaning revealed adipocyte hypertrophy and reduced adipocyte density (Fig. 2E). Concomitantly, circulating levels of leptin and adiponectin were increased in Nck2^−/− mice at 28 and 32 weeks postweaning, respectively (Fig. 2F and G). Of note, increased adiposity in Nck2^−/− mice was not accompanied by inflammation as supported by identical TNF-α and IL-6 circulating levels compared with Nck2^+/+ mice. Furthermore, Tnfa, Adrge1 (F4/80), and Itgax (Cd11c) mRNA levels were comparable between eWAT of both mouse genotypes (Fig. 2H). F4/80 staining showed no evidence of infiltrating macrophage-dependent formation of crown-like structure in expanded eWAT of Nck2^−/− mice (data not shown). Collectively, the data demonstrate that WAT expansion in Nck2^−/− mice is accompanied by adipocyte hypertrophy and enhanced release of adipokines but without any sign of inflammation.

Increased adiposity in Nck2-deficient mice prompted us to assess whether WAT expansion in obese humans correlates with decreased Nck2 expression in WAT. Compared with both protein and mRNA levels in moderately obese subjects, Nck2 expression was reduced in oWAT of severely obese subjects (Fig. 3A and B). Nck2 protein levels were also lower in scWAT from severely obese subjects but did not reach statistical significance due to the limited number of samples analyzed. In contrast, Nck1 expression was comparable between groups and WAT depots (Fig. 3A–C), suggesting differential regulation of Nck proteins expression in WAT related to obesity. Altogether, these findings suggest a novel role for Nck2 in limiting WAT expansion in mice and humans.

To gain insight into the mechanisms that promote adiposity in Nck2^−/− mice, we established differential gene expression profiles of eWAT between mouse genotypes by RNA sequencing (RNASeq). We focused on differentially expressed genes showing more than twofold significant change (P < 0.05): 1,420 and 2,472 genes were found upregulated and downregulated, respectively, in eWAT of Nck2^−/− mice (National Center for Biotechnology Information Gene Expression Omnibus accession number GSE63510). Clustering function analysis of upregulated genes revealed changes in distinct functional networks related to adipocyte, lipid metabolism, obesity, and extracellular matrix (ECM) (Fig. 4A). mRNA levels of PPARγ (Pparg) and the proadipogenic transcription cofactor Hairless (Hr) were significantly increased in Nck2^−/− mouse eWAT. Accordingly, PPARγ protein levels tend to increase in Nck2^−/− mouse eWAT (Fig. 4B). RNASeq also revealed that the expression of PPARγ target genes (Cebpa, Fabp4, and Fgf21) and adipokine genes (Lep, Adipoq, and Rbp4) was induced in Nck2^−/− mouse eWAT. In contrast, expression of early adipocyte differentiation regulators C/EBPβ and δ was not affected, suggesting that Nck2 regulates late events of adipocyte differentiation.

We next determined that the expression of genes involved in lipid metabolism, such as diacylglycerol acetyltransferase enzyme type 2 (Dgat2), PEPCK (Pck1), and several lipases (Hsl, Atgl, Lpl, and Lipf), and lipid droplet formation (Cidec/FSP27, Cidea, and Plin4) were significantly upregulated in Nck2^−/− mouse eWAT. Furthermore, mesoderm-specific transcript (Mest) involved in adipocyte size regulation and adipose tissue expansion (24) was among the top upregulated genes in Nck2^−/− eWAT. Genes previously identified as high-potential obesity candidates, including Deptor, Thbs1, Tuba1a, Npr3, and Gys2 (25,26), were also significantly upregulated in Nck2^−/− mouse eWAT. Consistent with the absence of inflammation in Nck2^−/− mouse eWAT, TNFa and IL6 mRNA levels were comparable in both mouse genotypes. RNASeq also revealed changes in the canonical Wnt pathway, which inhibits early steps of adipocyte differentiation (27,28). Expression of Wnt ligands was strongly downregulated, whereas Sfrp5 encoding a Wnt signaling inhibitor (29) was greatly increased in eWAT of Nck2^−/− mice. Increased levels of Adipoq gene expression in Nck2^−/− eWAT were also consistent with increased circulating adiponectin levels in Nck2^−/− mice (Fig. 2G). In agreement with an important role for ECM remodeling in regulating preadipocyte commitment, numerous genes encoding ECM proteins (Col5a1, Col15a1, Sparc, and Mfap5) were significantly upregulated in Nck2^−/− mice. Finally, we validated RNASeq data by quantitative real-time PCR for a selected set of genes and found concordance between both approaches (data not shown). Collectively, the findings strongly support a significant role for Nck2 in regulating the WAT adipogenic program.

To further demonstrate that Nck2 regulates adipogenesis, we compared in vitro differentiation of primary adipocyte precursors in isolated SVF from both mice genotypes. Nck2-deficient WAT SVF, which contains equivalent percentages of adipocyte precursors (CD29⁺CD34⁺Sca-1⁺) compared with Nck2^+/+ mice (Fig. 4C), showed enhanced differentiation as characterized by increased Pparg and Fabp4 mRNA levels and ORO staining after 7 days of differentiation compared with Nck2^+/+ WAT SVF (Fig. 4D). We also confirmed that Nck2 regulates adipogenesis by silencing Nck2 in 3T3-L1 preadipocytes using siRNA, which decreased Nck2 mRNA by ∼70% for at least 4 days of differentiation. In fact, silencing Nck2 significantly promoted Pparg and Fabp4 mRNA expression at day 4 of differentiation (Fig. 4E). In agreement, significantly higher levels of mature adipocyte markers: PPARγ₂, FABP4, adiponectin, and ACC were found in siRNA Nck2 3T3-L1 adipocytes (Fig. 4F). Evidence of increased fatty acid synthesis and lipid storage were suggested by higher levels of FAS, perilipin, and lipid droplet formation as assessed by ORO staining in Nck2-silenced 3T3-L1 cells (Fig. 4F and H). In addition, flow cytometry analysis demonstrated that the percentage of cells accumulating lipids was almost double upon silencing Nck2 in 3T3-L1 cells (Fig. 4I). Finally, silencing Nck2 in 3T3-L1 cells did not alter proliferation (Fig. 4G).

To assess whether Nck2 gain of function also affects adipogenesis, Nck2 was stably overexpressed in 3T3-L1 preadipocytes (Fig. 4J). In contrast to enhanced adipogenesis observed after silencing Nck2 in 3T3-L1, overexpression of Nck2 reduced 3T3-L1 differentiation as shown by a significant decrease in lipid droplet formation (Fig. 4K) and expression of Fabp4 mRNA (Fig. 4L). As for silencing Nck2, stable overexpression of Nck2 in 3T3-L1 preadipocytes did not affect proliferation (Fig. 4M). Altogether, these results provide strong evidence in favor of a cell-autonomous role for Nck2 in regulating adipocyte differentiation.

To further support RNASeq data showing increased expression of genes involved in lipid metabolism in Nck2-deficient adipocytes, we compared expression of genes associated with adipocyte function in Nck2-deficient WAT SVF and Nck2-silenced 3T3-L1 adipocytes. mRNA levels of lipogenic enzymes, such as FAS (Fasn), ACC (Acaca), stearoyl-CoA desaturase 1 (Scd1), and fatty-acid elongation enzyme (Elovl6), were upregulated in Nck2^−/−-differentiated WAT SVF (Fig. 5A). Similarly, mRNA levels of these enzymes were increased in differentiated Nck2-silenced 3T3-L1 adipocytes (Fig. 5B). In addition, silencing Nck2 in 3T3-L1 preadipocytes significantly enhanced oleate-induced lipid droplet formation as monitored following BODIPY C16 uptake (Fig. 5C), indicating that Nck2 deficiency already affects lipogenic control in adipocyte precursor cells. Moreover, increased expression of lipid transporter genes Cd36 and Fabp4 in RNASeq of Nck2^−/− mice eWAT, along with increased lipid droplet formation during differentiation of Nck2-deficient WAT SVF (Fig. 4D) and Nck2-silenced 3T3-L1 cells (Fig. 4H), strongly suggests a role for Nck2 in mature adipocyte lipid metabolism.

Like for lipogenic enzymes, genes encoding lipolytic enzymes, such as hormone-sensitive lipase (Lipe), adipose triglyceride lipase (Atgl), and monoacylglycerol lipase (Mgl), were strongly upregulated in both Nck2^−/− WAT SVF and Nck2-silenced 3T3-L1 adipocytes (Fig. 5D and E). To determine whether increased lipolytic enzyme gene expression in Nck2-deficient adipocytes affect WAT homeostasis, we subjected mice to prolonged fasting periods and followed changes in body weight and weight of adipose tissue depots. Upon 24-h fasting, Nck2^−/− mice showed greater loss in body weight and weight of eWAT and scWAT compared with Nck2^+/+ mice (Fig. 5F). However, these differences were less robust upon 48-h fasting, and no difference in BAT loss was noticed between mouse genotypes at any time. Thus, despite bigger WAT depots, Nck2^−/− mice display accelerated loss of WAT upon fasting, suggesting increased in vivo lipolysis in agreement with increased expression levels of lipolytic enzymes. Moreover, fed plasma TGs (Fig. 5G) and NEFAs (Fig. 5H) were significantly elevated in Nck2^−/− mice, revealing failure in lipid storage and/or increased lipolysis in Nck2^−/− mice. Altogether, these data provide strong evidence that Nck2 deficiency in mice, in addition to promoting adipogenesis, alters adipocyte function by increasing lipogenesis and lipolysis, both of which potentially contribute to elevated circulating lipid levels.

The UPR is essential during adipogenesis (30–32). We and others previously demonstrated that Nck1, which shares a high level of identity with Nck2, regulates the IRE1α and PERK arms of the UPR (6–8). Therefore, we hypothesized that enhanced adipogenesis induced by Nck2 deficiency in adipocyte precursors is UPR dependent. Accordingly, Atf4 and Ddit3 (CHOP) mRNA levels were significantly further increased in Nck2-silenced 3T3-L1 cells after 4 days of differentiation, whereas uXbp1 and sXbp1 mRNAs remained unchanged compared with control 3T3-L1 cells (Fig. 6A). Higher levels of ATF4 and activated PERK (Thr⁹⁸⁰) were detected in Nck2-silenced 3T3-L1 adipocytes (Fig. 6B). Furthermore, eIF2αSer⁵¹ phosphorylation, a classical marker of PERK activation associated with increased ATF4, was significantly increased in Nck2^−/− mouse eWAT (Fig. 6C). Altogether, these data demonstrate that Nck2 deficiency is accompanied by increased activity of the PERK-peIF2α-ATF4 pathway, which could mediate Nck2 deficiency effect on adipocyte differentiation. Supporting this hypothesis, overexpression of PERK at low levels in 3T3-L1 preadipocytes mimicked the effects of Nck2 silencing by enhancing Pparg and Fabp4 mRNA levels at day 6 of differentiation (Fig. 6D and E). To further demonstrate that Nck2 deficiency involves enhanced PERK activity and signaling in promoting adipogenesis, we followed control and Nck2 siRNA 3T3-L1 cell differentiation in the presence of a potent specific PERK inhibitor (GSK2606414) at a dose that only partially inhibits thapsigargin-induced PERK activation (10 nmol/L, data not shown). As aforementioned, silencing Nck2 in 3T3-L1 cells increased expression of Pparg and Fabp4 and lipid droplet formation after 4 days of differentiation (Fig. 6F and G). However, PERK inhibitor added during differentiation of 3T3-L1 cells prevented the effects of Nck2 silencing on enhanced Pparg and Fabp4 mRNA levels and lipid droplet formation, but it had no effect in control 3T3-L1 cells. Nck2 mRNA quantitative PCR analysis established that the PERK inhibitor effects were not due to Nck2 expression recovery (Fig. 6G). Taken together, these results provide strong evidence that the mechanism through which Nck2 deficiency leads to enhanced adipogenesis in vitro and adiposity in vivo is likely associated with primed physiological PERK activity and signaling.

We investigated whether enhanced adiposity accompanied by adipocyte dysfunction affects glucose homeostasis in Nck2^−/− mice. Glucose tolerance tests revealed that Nck2^−/− mice displayed glucose intolerance at 16 weeks postweaning, which was exacerbated in 1-year-old mice (Fig. 7A). As revealed by insulin tolerance tests, insulin resistance in Nck2^−/− mice was already apparent 16 weeks postweaning and became significantly established at 24 weeks (Fig. 7B). Impaired glucose disposal in Nck2^−/− mice was not due to failure of pancreatic β-cells to provide enough insulin because in vivo glucose-stimulated insulin secretion (GSIS) in Nck2^−/− mice (Fig. 7C) and insulin content in islets from Nck2-deficient mice (Nck2^+/+ 16.4 ± 2.1 vs. Nck2^−/− 21.3 ± 1.1 ng insulin/ng DNA, P < 0.05) were increased compared with respective controls. In addition, Nck2^−/− mice displayed hyperinsulinemia from 24 weeks (Fig. 7D), which probably contributed to maintaining normal glycemia in Nck2^−/− mice (Fig. 7E).

Finally, we compared energy metabolism and physical activity in both mouse genotypes. Clearly, no difference was observed in daily food intake, energy expenditure, and locomotor activity between genotypes (Fig. 7F, G, and I). However, the respiratory exchange ratio (RER) showed that Nck2^−/− mice were significantly less effective at shifting from dark-cycle carbohydrate breakdown to lipid β-oxidation during daylight (Fig. 7H), making them metabolically less flexible.

Correlating with Nck2^−/− mice spontaneously developing whole-body insulin resistance and hyperinsulinemia, we discovered that although this was not observed at a younger age (10 weeks), Nck2^−/− mice at 32 weeks postweaning display hepatic steatosis (Fig. 8A). Hepatic steatosis in Nck2^−/− mice was not supported by changes in expression of genes regulating hepatic lipid metabolism (Fig. 8B). Furthermore, in vivo insulin-induced Akt phosphorylation was significantly reduced in the liver of Nck2^−/− mice (Fig. 8C), whereas no change in insulin-induced Akt phosphorylation was detected in muscle and eWAT (data not shown). In agreement with hepatic insulin resistance, Nck2^−/− mice also displayed increased hepatic gluconeogenesis in the pyruvate tolerance test (Fig. 8D), higher expression of the PEPCK implicated in gluconeogenesis (Fig. 8E), and lower glycogen content (Fig. 8F). Lipid accumulation in skeletal muscle was not apparent (data not shown), and glycogen content was unchanged (Fig. 8F), further supporting unaltered insulin sensitivity in this tissue.

To determine whether hepatic insulin resistance in Nck2^−/− mice is due to impaired autonomous hepatocyte function, we assessed insulin-induced phosphorylation of Akt in primary hepatocytes in culture. Of note, insulin response in Nck2^−/− primary hepatocytes appeared to be increased compared with control hepatocytes (Fig. 8G), indicating that hepatic insulin resistance in Nck2^−/− mice results from a systemic effect instead of being hepatocyte autonomous. Globally, Nck2 deficiency in mice induces progressive metabolic disorders that are consistent with enhanced adiposity and circulating lipids associated with dysregulated adipogenesis and adipocyte function.

In this study, we uncovered an unexpected role for the SH domains containing adaptor Nck2 in regulating adipogenesis and adipocyte function associated with the control of glucose homeostasis. We identified Nck2-deficient mice as a model displaying spontaneous increased adiposity and dysfunctional adipocytes concomitant with progressive glucose intolerance, insulin resistance, and hepatic steatosis. Correlating with high Nck2 expression in WAT, Nck2-deficient mice were found to be more susceptible to expanded WAT. Consistently, we demonstrate reduced Nck2 expression in oWAT of severely obese human subjects, supporting the existence of a negative relationship between WAT Nck2 expression and BMI in humans. The success of existing pharmacological approaches combined with caloric restriction and increased physical activity to facilitate weight loss in humans is still limited and accompanied by major secondary effects that dampen their clinical potential in treating obesity. The current study in mice and humans suggests that moving forward in developing strategies to modulate Nck2 expression could be a valuable alternative to specifically control the pathological visceral WAT expansion that leads to obesity.

In Nck2-deficient mice, increased adiposity is characterized by adipocyte hypertrophy, but increased levels of circulating lipids indicate dysfunctional adipocytes. Upregulated expression of lipogenic enzymes in Nck2-deficient adipocytes suggest that an increased rate of lipogenesis could contribute to adipocyte hypertrophy and exceed the maximum adipocyte lipid storage capacity. Moreover, enhanced lipolysis in Nck2-deficient mice could significantly contribute to increased levels of circulating lipids, explaining why Nck2-deficient mice develop hepatic steatosis. In parallel, RNASeq of differentially expressed genes in Nck2^−/− eWAT mice suggests that upregulated expression of ECM genes, promoting WAT ECM stiffness and fibrosis (33), could contribute to mature adipocyte dysfunction. Although ECM gene regulation is interesting, fold induction of ECM genes appears less pronounced than other genes reported. Therefore, further investigation is required to assess whether Nck2 deficiency affects mature adipocyte function through ECM-related fibrosis.

The current study provides strong evidence that increased fat mass in Nck2-deficient mice is associated with not only adipocyte hypertrophy resulting from impaired lipid metabolism in adipocytes, but also enhanced adipogenesis. Genes encoding players of the Wnt signaling reported to inhibit adipocyte differentiation (34) were strongly downregulated in Nck2-deficient eWAT, which thereby could significantly contribute to enhanced adipogenesis. Consistent with a role for Nck2 in limiting adipogenesis, Nck2 deficiency promotes differentiation of preadipocytes into adipocytes in vitro.

We previously reported that the PERK-peIF2α-ATF4 pathway is damped following overexpression of either Nck1 or Nck2 in various mammalian cells (12,13), revealing that both Nck1 and Nck2 are negative regulators of PERK signaling in multiple models. Accordingly, Nck2 deficiency enhances physiological PERK activation and signaling during adipocyte differentiation and in mature adipocytes. Consistent with PPARγ being a direct target of ATF4 and ATF4 overexpression in 3T3-L1 preadipocytes enhancing differentiation (35), the current findings that PPARγ and phosphorylation of eIF2αSer⁵¹ are significantly increased in Nck2-deficient eWAT further support enhanced PERK activation and signaling. We provided strong mechanistic evidence that physiological PERK activity and signaling regulate adipocyte differentiation and mediate Nck2 deficiency effect on adipogenesis in vitro and adipogenesis leading to increased adiposity in vivo (Fig. 8H). Previous studies reported that the UPR is activated in adipose tissue of obese patients without diabetes in relation with BMI (36) and in obese subjects with insulin resistance (37), suggesting a link between activated UPR and WAT mass expansion. We confirmed activation of PERK and IRE1α signaling during adipocyte differentiation. However, silencing Nck2 in preadipocytes has no further effect on the IRE1α-XBP1 pathway, suggesting that although Nck2 specifically modulates the PERK-peIF2α-ATF4 pathway during adipocyte differentiation, it has no control over IRE1α signaling in this context.

The exact mechanism by which silencing Nck2 promotes PERK activation and signaling during adipocyte differentiation still remains to be determined. We have identified Nck1 as a negative regulator of PERK activation through its direct interaction with PERK (13). In agreement with Nck2 also directly interacting with PERK (13), we provided strong evidence that silencing Nck2 in preadipocytes promotes adipogenesis by further enhancing PERK activation and signaling during this process (30).

Earlier studies proposed that Nck1 and Nck2 are functionally redundant because double knockout of Nck1 and Nck2 in mice is embryonically lethal, whereas individual Nck knockout shows no apparent phenotype (16). Although Nck1 and Nck2 share common functions and interacting proteins (38), studies have reported specific functions and exclusive binding partners for Nck proteins (39–41). In agreement, we showed that Nck1-deficient mice display normal glucose homeostasis (15), and in the current study, we report that Nck2-deficient mice spontaneously develop progressive increased adiposity concomitant with impaired glucose homeostasis, insulin resistance, and hepatic steatosis. Although Nck2^−/− mice were generated a while ago and have been reported with no obvious phenotype (16), progressive adiposity was probably missed because it requires careful WAT investigation given that mouse body weight was not affected.

In summary, this study using in vivo Nck2 deficiency and in vitro Nck2 silencing in preadipocytes argues for an unanticipated role of Nck2 in controlling the susceptibility of developing adiposity. Furthermore, the study unveils the importance of Nck2 and the regulation of PERK as potential new avenues for managing obesity in humans.

Acknowledgments. The authors thank T. Pawson (Mount Sinai Hospital, Toronto, Ontario, Canada) for providing Nck2^+/− mice several years ago. From McGill University and the McGill University Health Centre Research Institute, the authors also thank Victor Dumas for technical assistance, E.C. Davis for histological analysis, M. Kokoeva for metabolic studies, S. Chevalier for expertise in NEFA determination, the immunophenotyping platform for flow cytometry expertise, and S.A. Laporte and J.J. Bergeron for critical reading of the manuscript. Finally, the authors acknowledge the surgery team, bariatric surgeons, and Biobank staff of the IUCPQ.

Funding. J.D. was supported by postdoctoral fellowships from the Fonds de la Recherche du Québec en Santé and the Canadian Diabetes Association. J.-F.C. is a recipient of a senior career award from the Fonds de la Recherche du Québec en Santé. J.-F.C. and L.L. have received funds from the Canadian Institutes of Health Research (MOP-144425 and MOP-115045, respectively).

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

Author Contributions. J.D. contributed to the experimental design, data collection, data interpretation and analysis, preparation of figures, and writing of the manuscript. B.L. contributed to the experiments and critical reading of the manuscript. N.H. contributed to the experiments. M.-A.G. contributed to the RNASeq analysis. J.-F.C. contributed to the RNASeq analysis and critical reading of the manuscript. L.L. contributed to study design and final editing of the manuscript. L.L. 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.