Introduction

The prevalence of obesity is increasing worldwide, and will be one of the most serious public health concerns in the future [1]. Obesity is frequently accompanied by insulin resistance and type 2 diabetes [2] and is characterised by increased fat mass caused by an increased adipocyte number (hyperplasia) and/or size (hypertrophy) [3]. When the demand for fat storage exceeds capacity, adipocyte number increases through proliferation and/or differentiation from pre-adipocytes [4]. Adipogenesis is mediated by a well-programmed sequence of transcriptional events beginning with the induction of two transcription factors in the CCAAT/enhancer binding protein (C/EBP) family: C/EBPβ and C/EBPδ [5, 6]. Subsequently, these proteins activate transcription of the peroxisome proliferator-activated receptor gamma (PPARγ) and C/EBPα, two central adipogenic regulators that positively control each other and cooperate to orchestrate expression of the full adipogenic programme, including induction of additional transcription factors, suppression of growth-associated genes and stimulation of insulin-dependent glucose transport [7].

When endoplasmic reticulum (ER) homeostasis is disrupted, unfolded proteins accumulate in the ER lumen, which subsequently activates the unfolded protein response (UPR) through dissociation of binding immunoglobulin protein/78 kDa glucose-regulated protein (BiP/GRP78) from protein kinase RNA-like ER kinase (PERK), inositol-requiring enzyme 1 alpha (IRE1α) and activating transcription factor 6 alpha (ATF6α) [8, 9]. Activated PERK phosphorylates eIF2α at Ser51 leading to rapid and transient attenuation of protein synthesis [10, 11]. Paradoxically, under these conditions translation of Atf4 mRNA is selectively enhanced, which induces transcription of Chop (also known as Ddit3) [12]. Activated IRE1α elicits an endoribonuclease function that initiates unconventional splicing of Xbp1 mRNA [13] to produce a novel translation product X-box binding protein 1 (XBP1), which induces expression of genes encoding functions including ER protein chaperones, lipid biosynthetic enzymes and ER-associated protein degradation (ERAD) [14, 15]. Upon ER stress, ATF6α traffics to the Golgi apparatus where it is cleaved by the processing enzymes S1P and S2P to liberate a fragment that migrates into the nucleus to induce genes encoding ER protein chaperones and ERAD functions [16, 17].

There have been some studies suggesting that ER stress might be important for adipogenesis [18–20]. However, it is not well understood how ER stress and subsequent activation of the UPR affects adipogenesis and weight gain. Interestingly, mice deleted in p58 IPK (also known as Dnajc3), a co-chaperone DNAJ family member for BiP/GRP78 in the ER, as well as mice heterozygous for Grp78 (also known as Hspa5), gain less fat mass compared with wild-type mice [21, 22]. Since p58 IPK deficiency and Grp78 heterozygosity decrease ER chaperone activity and increase ER stress, the reduced fat mass in these mice might result from activation of UPR signalling pathways. In addition, we reported previously that high-fat diet (HFD)-fed mice with a heterozygous mutation at the phosphorylation site in Eif2a (Ser51Ala, S/A) became significantly more obese than HFD-fed wild-type mice [23]. This prompted us to investigate the role of eIF2α phosphorylation and IRE1α signalling in adipocyte differentiation using 3T3-L1 cells, mouse embryonic fibroblasts (MEFs) and genetically engineered mice.

Methods

Animals

Animal use was in compliance with the Institute of Laboratory Animal Research Guide for the Care and Use of Laboratory Animals and approved by the University Committee on Use and Care of Animals at the University of Michigan. Chop −/− [24] and Eif2a S/A [23] mice were previously described. The genetic background of Chop −/− and Eif2a S/A mice is C57BL/6J. Lepr db/+ mice (C57BKS.Cg-m +/+ Lepr db, JAX mice) were bred with Chop −/− mice to generate heterozygous F1 mice that were intercrossed to obtain Chop −/− Lepr db/db and Chop +/+ Lepr db/db mice. For HFD studies, 10- to 14-week-old male mice were provided with free access to HFD (catalogue D12451; Research Diets, New Brunswick, NJ, USA) or regular chow (catalogue D12450; Research Diets) for the indicated times. Body weights were measured between 15:00 and 17:00 hours. Food intake was analysed by measurement of food mass each day for 4 days and intake calculated (kJ/day) using a conversion rate of 19.79 kJ/g. For all studies, age-matched siblings were used as controls.

Cell culture and adipocyte differentiation

3T3-L1 pre-adipocytes were maintained in DMEM supplemented with 10% calf serum [25]. Two days after confluency, cell differentiation into adipocytes was induced with DMEM containing 1 μg/ml insulin (Invitrogen, Carlsbad, CA, USA), 0.5 mmol/l 3-isobutyl-1-methylxanthine (Sigma, St Louis, MO, USA) and 1 μmol/l dexamethasone (Sigma) for 4 days, and then with DMEM supplemented with 10% FBS and 1 μg/ml insulin only for another 3 days. After induction, cells were fed every other day with DMEM containing 10% FBS. Next, Oil red O (Sigma) working solution was added to formalin-fixed cells and incubated for 10 min at room temperature. Images were taken using an Olympus microscope system (Olympus, Center Valley, PA, USA). For quantification, absorbance was measured at 500 nm using a spectrophotometer (SpectraMAX plus; Molecular Devices, Sunnyvale, CA, USA).

Generation of stable 3T3-L1 cell lines

The pBabe vector encoding Fv2E-PERK was obtained from D. Ron (University of Cambridge) [26]. For Chop, Ire1a (also known as Ern1) and Ire1aK907A constructs, cDNAs were amplified by PCR and cloned into pLVX-tight-puro (Clonetech, Mountain View, CA, USA). Tet-inducible 3T3-L1 cells, described previously [25], were transduced with viruses, followed by puromycin (2 μg/ml) selection for stable inducible expression of the transgene.

Measurement of body composition

Body composition was analysed in the conscious state with a Minispec LF90 II (Bruker Optics, Billerica, MA, USA), a nuclear magnetic resonance-based whole-body composition analyser at the mouse phenotyping core at the University of Michigan.

MEF generation

Primary MEFs were generated as described previously [27]. For differentiation, only primary MEFs at passage two were used. Differentiation was induced as described for 3T3-L1 cells except that 10 μg/ml of insulin and rosiglitazone (50 nmol/l; Cayman Chemical, Ann Arbor, MI, USA) was added for the first 4 days.

RNA extraction and real-time PCR analysis

Total RNA was extracted from cells using RNAeasy (Qiagen, Valencia, CA, USA). The relative amounts of mRNAs were calculated from the comparative threshold cycle (Ct) values relative to β-actin. Real-time primer sequences are shown in electronic supplementary material (ESM) Table 1. Data was analysed using the \( {2^{{-\Delta \Delta {{\mathrm{C}}_{\mathrm{t}}}}}} \) method.

Immunohistochemistry and cell size measurement

Epididymal fat pads were obtained from mice at indicated times, then fixed in formalin and paraffin embedding. Five-micrometre sections were deparaffinised and stained with haematoxylin and eosin. For each mouse two or three representative images of each slide were obtained and cell sizes were measured using ImageJ 1.43u software (http://rsb.info.nih.gov/ij).

Serum analysis

Serum levels of NEFA and total cholesterol were measured using HR Series NEFA-HR kit and Cholesterol E (Wako, Richmond, VA, USA) according to the manufacturer’s instructions. Triacylglycerol levels were measured using Infinity Triglycerides Reagent (Thermo, Middletown, VA, USA) according to the manufacturer’s instructions. Serum samples were collected after a 4–6 h fast.

Insulin tolerance tests

Mice were fasted for 6 h, followed by i.p. injection of insulin (0.75 U/kg body weight). Blood samples were collected via the tail vein and blood glucose levels were measured using a OneTouch Ultra glucometer (LifeScan, Milpitas, CA, USA).

Western blotting

Cell lysates were obtained in cell lysis buffer (50 mmol/l Tris HCl [pH 7.4], 150 mmol/l NaCl, 1% [vol./vol.] Triton X-100, 0.1% [wt/vol.] SDS, 1% [wt/vol.] sodium deoxycholate and protease inhibitors) (Roche Diagnostics, Indianapolis, IN, USA) and total protein concentration in each sample was measured using the Lowry protein assay kit (Biorad, Hercules, CA, USA). Primary antibodies were as follows: p-eIF2α (Invitrogen), eIF2α (Invitrogen), p-IRE1α (Novus, Littleton, CO, USA), IRE1α (Cell Signaling, Danvers, MA, USA), KDEL which detects GRP78 and GRP94 (Abcam, Cambridge, MA, USA), ATF4 (a kind gift from M. Kilberg, University of Florida), CHOP (Santa Cruz Biotechnology, Santa Cruz, CA, USA), PPARγ (Cell Signaling), C/EBPα (Santa Cruz Biotechnology), C/EBPβ (Santa Cruz Biotechnology), C/EBPδ (Santa Cruz Biotechnology) and tubulin (Sigma). Chemiluminescence detection was performed using ECL Western Blot detection reagents (GE Healthcare, Pittsburgh, PA, USA). Membranes were exposed to imaging film and developed using a Kodak X-OMAT processor (Kodak, Rochester, NY, USA).

Statistical analysis

All data are presented as means ± SEM. The difference between groups was evaluated using Student’s t test; p < 0.05 was considered significant.

Results

The UPR is activated during adipogenesis

During differentiation of 3T3-L1 cells, the production of proteins encoded by the adipocyte-specific genes Pparg, Cebpa, Cebpb and Cebpd was increased (Fig. 1a). In parallel, mRNA expression of Adipoq, Fabp4, Cebpa and Cebpb was increased coinciding with the progression of adipogenesis (Fig. 1b–f). In contrast, the expression of the pre-adipocyte-specific gene Pref1 (also known as Delk1) decreased (Fig. 1g), indicating that this cell line was well differentiated. Next, we investigated the levels of UPR markers during adipogenesis (Fig. 1h–m). Interestingly, the ratio of phosphorylated eIF2α to total eIF2α was reduced at day 2 but restored at days 4 and 7 (Fig. 1a, ESM Fig. 1a). Consistent with this pattern, the production of CHOP, a transcription factor induced by eIF2α phosphorylation, decreased at day 2 and subsequently increased between days 4 and 7 (Fig. 1a, i). In addition, the level of both phosphorylated and unphosphorylated IRE1α increased during adipocyte differentiation (Fig. 1a, ESM Fig. 1b), where the levels of spliced Xbp1 (Xbp1-s), as well as total Xbp1 (Xbp1-t), mRNAs were upregulated during the later period of adipocyte differentiation (Fig. 1j, k).

Fig. 1
figure 1

Expression of ER stress genes is upregulated during adipogenesis in 3T3-L1 cells. (a) Protein expression profiles for adipocyte-related genes and UPR-related genes. Cell lysates were collected at the indicated times after the start of adipocyte differentiation and analysed by western blotting analysis. The white and black arrows indicate larger and smaller isoforms of C/EBPα (* indicates a non-specific band). p-eIF2α and p-IRE1α indicate phosphorylated forms of eIF2α and IRE1α, respectively. KDEL antibody can detect GRP78 and GRP94 proteins (b–m) Gene expression profiles for adipocyte-related and UPR-related genes. Total RNA was isolated from 3T3-L1 cells at the indicated times during adipogenesis and mRNA levels were measured by quantitative real-time RT-PCR. Xbp1-s and Xbp1-t indicate spliced and total Xbp1, respectively. D0, D2, D4 and D7 indicate day 0, day 2, day 4 and day 7 after initiation of adipogenesis. Data are presented as means ± SEM of three independent experiments with triplicates

Full size image

ER stress represses adipocyte differentiation

Next, we asked whether exogenous ER stress affects adipogenesis. Induction of ER stress by treatment with a low dose of tunicamycin (Tm), an inhibitor of N-linked glycosylation that is not cytotoxic [28], significantly inhibited adipogenesis quantified by Oil red O staining (Fig. 2a, b) and expression of mature adipocyte marker genes (Fig. 2c–f). We also investigated the effects of a physiologically relevant inducer of ER stress, hypoxia [29], on adipogenesis. As early as 4 h under hypoxic conditions, eIF2α was phosphorylated and subsequently, after 12 h of hypoxia, its downstream target, CHOP, was upregulated as were other UPR-induced genes (Fig. 2g, h). When hypoxia was introduced during adipogenesis, adipocyte differentiation was significantly attenuated, with reduced expression of mature adipocyte marker genes (Fig. 2i–n). Since mice deficient in p58 IPK, which assists chaperone-mediated protein maturation in the ER [30, 31], display reduced adipose mass [22], we also investigated the effect of p58 IPK deletion on adipocyte differentiation. We observed that UPR-related genes were upregulated to a significantly greater degree in response to ER stress in p58 IPK−/− compared with p58 IPK+/+ pre-adipocytes (Fig. 2o), indicating that p58 IPK-deficient MEFs are more susceptible to ER stress. Consistent with the in vivo observation, adipogenesis in p58 IPK−/− pre-adipocytes was reduced compared with p58 IPK+/+ pre-adipocytes in the presence or absence of ER stress (Fig. 2p, q), with reduced expression of mature adipocyte marker genes (Fig. 2r–u).

Fig. 2
figure 2

ER stress suppresses adipogenesis in 3T3-L1 cells. (a, b) Suppression of adipocyte differentiation by tunicamycin (Tm) treatment. (a) Confluent 3T3-L1 cells were cultured in differentiation media in the presence of Tm at the indicated concentrations. On day 10, cells were fixed and stained with Oil red O. Representative images of Oil red O staining are shown. (b) Quantification of Oil red O staining in (a). **p < 0.01 vs vehicle (c–f) Gene expression profiles for markers of mature adipocytes. Total RNA was isolated from 3T3-L1 cells at the indicated times during adipogenesis in the absence or presence of Tm (50 ng/ml) and mRNA levels were measured by quantitative real-time RT-PCR. (g–n) Suppression of adipogenesis by hypoxia. (g) Induction of eIF2α phosphorylation and CHOP production. Cell lysates were collected at the indicated times after hypoxic treatment and analysed by western blotting. Tg indicates thapsigargin at 300 nmol/l. (h) Gene expression profiles after hypoxic treatment for 24 h. Total RNA was isolated from 3T3-L1 cells at the indicated times during adipogenesis and mRNA levels were measured by quantitative real-time RT-PCR. White bar, normoxic conditions; grey bar, hypoxia. (i) Oil red O staining of 3T3-L1 cells after adipogenesis under normoxic (21%) or hypoxic (1%) conditions. (j) Quantification of Oil red O staining in (i). (k–n) Gene expression profiles for markers of mature adipocytes under normoxic or hypoxic conditions. (o–u) Suppression of adipogenesis by p58 IPK deletion. (o) Gene expression profiles in p58 IPK+/+ (white bars) or p58 IPK−/− cells (grey bars) in the presence of Tm (1 μg/ml). Total RNA was isolated from p58IPK +/+ and p58IPK −/− pre-adipocytes at 24 h after treatment with Tm and mRNA levels were measured by quantitative real-time RT-PCR. Data were plotted as fold induction relative to the vehicle-treated sample. (p) Oil red O staining of p58 IPK+/+ and p58 IPK−/− MEFs after differentiation in the absence or presence of Tm (50 ng/ml). (q) Quantification of Oil red O staining in (p). White bars, vehicle; grey bars, Tm. (r–u) Gene expression profiles for markers of mature adipocytes in p58 IPK+/+ and p58 IPK−/− MEFs in the absence or presence of Tm (50 ng/ml). Data are presented as means ± SEM of three independent experiments with triplicates. *p < 0.05, **p < 0.01. AU, arbitrary units

Full size image

Phosphorylation of eIF2α represses adipogenesis

The experiments above suggest that UPR activation appears to decrease adipogenesis. Since all three UPR subpathways are activated under conditions of ER stress, it is difficult to attribute an inhibitory effect to any specific UPR subpathway. Therefore, we investigated the effect of mutations that singly inactivate each of the UPR subpathways. First, we generated stable 3T3-L1 cell lines producing Fv2E-PERK (3T3-L1-Fv2EPERK), which forms dimers in the presence of the drug AP20187, to pre-emptively phosphorylate eIF2α without an ER stress signal [26] (ESM Fig. 2). After AP20187 treatment, eIF2α phosphorylation was dramatically increased at 1 h then slightly declined up to 24 h (Fig. 3a). ATF4 induction was observed at 2 h and CHOP was also upregulated at 2–4 h after AP20187 treatment (Fig. 3a). Addition of AP20187 induced eIF2α phosphorylation to a level comparable with that observed in MEFs treated with thapsigargin (ESM Fig. 3) and significantly reduced adipogenesis in the 3T3-L1-Fv2EPERK stable cell line compared with vehicle-treated cells as detected by reduced lipid accumulation (Fig. 3b, c) and expression of mature adipocyte marker genes (Fig. 3d–g).

Fig. 3
figure 3

eIF2α phosphorylation represses adipogenesis. (a) Protein expression profile in a stable 3T3-L1 cell line that expresses chimeric Fv2E-PERK. Cell lysates were collected at the indicated times after AP20187 treatment (5 nmol/l) for western blot analysis (* indicates a non-specific band). (b) Oil red O staining. Adipogenesis was stimulated in 3T3-L1 cells expressing Fv2E-PERK in the absence (vehicle) or presence of AP20187. On day 10, cells were fixed and stained with Oil red O. Representative images from indicated cell lines are shown. Scale bar indicates 100 μm. (c) Quantification of Oil red O staining in (b). (d–g) Gene expression profiles for markers of mature adipocytes in Fv2E-PERK 3T3-L1 cells in the absence or presence of AP20187 (5 nmol/l). (h) Oil red O staining of Eif2a S/S and Eif2a A/A MEFs after adipogenesis. Differentiation was induced in primary MEFs from wild-type (Eif2a S/S) or homozygous mutant (Eif2a A/A) for 14 days. Cells were then fixed and stained with Oil red O. Representative images are shown and scale bar indicates 100 μm. (i) Quantification of Oil red O staining in (h). (j–n) Gene expression profiles during adipocyte differentiation of primary MEFs. Total RNA was isolated from cells at the indicated times during adipogenesis, and mRNA levels were measured by quantitative real-time RT-PCR. Circles, Eif2a S/S; squares, Eif2a A/A. D0, D2, D4, D7, D10 indicate day 0, day 2, day 4, day 7 and day 10 after initiation of adipocyte differentiation. Data are presented as means ± SEM of three independent experiments with triplicates. **p < 0.01, *p < 0.05 vs Eif2a A/A. AU, arbitrary units

Full size image

To analyse the effect of impaired eIF2α phosphorylation on adipogenesis, primary MEFs were isolated from mouse embryos with wild-type eIF2α alleles (Eif2a S/S) or with two mutant eIF2α alleles that had alanine substitutions at Ser51 to prevent phosphorylation (Eif2a A/A) [11]. When induced for adipogenesis, there was significantly more lipid accumulation in the Eif2a A/A MEFs compared with the Eif2a S/S MEFs (Fig. 3h, i). Similar results were obtained in two additional independent preparations of Eif2a S/S and Eif2a A/A primary MEFs. The lipid accumulation correlated with significantly reduced induction of Chop in the Eif2a A/A mutant MEFs during the entire period of adipocyte differentiation compared with wild-type Eif2a S/S MEFs (Fig. 3j). In contrast, expression of the main adipogenic regulators Pparg and Cebpa in Eif2a A/A MEFs was significantly increased at day 10 (Fig. 3k–n). These results suggest that eIF2α phosphorylation and subsequent activation of downstream signalling pathways represses adipocyte differentiation.

Eif2aS/A mice fed HFD become more obese than wild-type mice through an increase in adipocyte number

We then investigated the effect of eIF2α phosphorylation in vivo using heterozygous Eif2a S/A and wild-type mice (Eif2a S/S) since homozygous Eif2a A/A is perinatal lethal. The Eif2a S/A mice gained more body weight than Eif2a S/S mice fed an HFD [23]. Although there were no differences in food intake (ESM Fig. 4a), the Eif2a S/A mice displayed increased body fat and less lean body mass compared with Eif2a S/S mice (Fig. 4a–f), indicating that most of the weight gain was due to increased fat mass. The size distribution of adipocytes was not significantly different between the genotypes (Fig. 4g, h). In addition, there was no significant difference in the number of cells or their rate of proliferation from the stromal vascular fraction (SVF) of epididymal fat tissues between Eif2a S/A and Eif2a S/S mice (Fig. 4i, j). These results suggest that the increased fat mass in the HFD-fed Eif2a S/A mice results from increased adipocyte number. Although obesity can cause hyperlipidaemia, the levels of serum NEFA, total cholesterol, adiponectin and triacylglycerol, as well as liver triacylglycerol and insulin sensitivity, were not significantly different between the genotypes fed an HFD (Fig. 4k–o; ESM Fig. 4b).

Fig. 4
figure 4

Mice with mutation in eIF2α (Eif2a S/A) display increased body weight compared with wild-type Eif2a S/S mice without differences in adipocyte size or serum lipid levels. (a–f) Weight gain, body fat mass, body lean mass, fluid mass, blood glucose and insulin levels. Mice were fed an HFD for 13 weeks and metabolic variables were measured using an NMR-based analyser. Eif2a S/A (n = 12) and Eif2a S/S (n = 7). (g, h) No difference in adipocyte size between genotypes. After HFD for 13 weeks, epididymal fat tissues were obtained for histochemistry and representative images of epididymal fat pads are shown in (g). Scale bar indicates 100 μm. (h) Histogram shows the distribution of adipocyte size. The x-axis represents the logarithm of cell size in pixels, while the y-axis shows the percentage of cells having a given size for each group of mice. White bars, Eif2a S/S; grey bars, Eif2a S/A. (i) The number of cells from the SVF of adipose tissues from Eif2a S/A (n = 4) and Eif2a S/S (n = 4) mice at age of 6–8 weeks on a regular diet. (j) Proliferation rate of cells from SVF. Cells from SVF were plated in replicate and the number of cells was counted after 4 days of culture using an automated cell counter. Each suspension was counted twice. The y-axis represents the percentage of the initial cell number. Plasma levels of non-esterified NEFA (k), cholesterol (l) and triacylglycerol (m) after 13 weeks of HFD. Sera from Eif2a S/A (n = 11) and Eif2a S/S (n = 9) mice were collected and measured for NEFA and cholesterol as indicated in methods. (n) Hepatic triacylglycerol levels. Liver samples were collected for analysis after 13 weeks of HFD. (o) Insulin tolerance tests (ITTs). After 13 weeks of HFD, blood glucose levels were measured at the indicated times after insulin injection (0.75 U/kg body weight). Eif2a S/A (n = 12, squares) and Eif2a S/S (n = 9, circles). *p < 0.05. BW, body weight

Full size image

CHOP production inhibits adipogenesis

Since CHOP is induced through eIF2α phosphorylation and inhibits adipogenesis by interfering with other C/EBP family members [32], we examined the effect of CHOP on adipogenesis more intensively using an inducible expression system. First, conditional 3T3-L1 cell lines were generated that express CHOP under the control of a tetracycline-inducible promoter (ESM Fig. 5). It is notable that CHOP production itself is not apoptotic unless there is a stress signal [33]. When CHOP production was induced during adipocyte differentiation, adipogenesis was significantly inhibited in a doxycycline-dose-dependent manner (Fig. 5a). The level of C/EBPβ was slightly attenuated at days 2 and 4 after doxycycline treatment, and substantially reduced by day 7 (Fig. 5b). Upon doxycycline treatment, C/EBPα and PPARγ were barely detected during adipogenesis, suggesting that adipogenesis was significantly impaired by overproduction of CHOP (Fig. 5b). Next, we identified the most critical time point for CHOP to exert its inhibitory effect on adipogenesis. Whereas doxycycline treatment during the entire period of adipogenesis almost completely blocked differentiation (Fig. 5c-3, d, e), doxycycline treatment during the earlier period (Fig. 5c-1, c-2, d, e) or the later period (Fig. 5c-7, d, e) had negligible effects on adipogenesis. However, induction of CHOP by doxycycline treatment from day 2 to day 4 significantly repressed adipogenesis (Fig. 5c-6, d, e) and doxycycline treatment from day 0 to day 4 further repressed adipocyte differentiation in the CHOP-inducible 3T3-L1 cell line (Fig. 5c-5, d, e), indicating that CHOP production during day 2 to day 4 is critical for its inhibitory effect. Consistent with this observation, transient expression of Chop from day 2 to day 4 during adipogenesis significantly attenuated induction of Cebpa and Pparg mRNAs (Fig. 5f–i) and their related proteins (Fig. 5j). However, treatment with doxycycline from day 4 to day 7 did not alter expression of these genes. These results indicate that CHOP inhibits adipogenesis through suppression of C/EBPα and PPARγ production and the most critical time period is around day 2 when the production of CHOP is reduced during adipogenesis (Fig. 1a). To study the requirement for CHOP in suppressing adipogenesis in response to ER stress, we studied adipogenesis in Chop +/+ and Chop −/− MEFs in the absence or presence of ER stress induced by Tm treatment (50 ng/ml). Whereas Tm treatment significantly reduced adipogenesis in Chop +/+ MEFs, adipogenesis was not significantly reduced in Chop −/− MEFs, indicating that CHOP is essential for ER stress-mediated suppression of adipogenesis (Fig. 5k–p). However, there was no significant difference either in cAMP levels or in insulin sensitivity, during adipogenesis, between Chop +/+ and Chop −/− MEFs (ESM Fig. 6).

Fig. 5
figure 5

CHOP production inhibits adipogenesis in 3T3-L1 cells. (a) Inhibition of adipogenesis by CHOP. 3T3-L1 cells producing CHOP under the control of tetracycline-inducible promoter (Tet-CHOP) were differentiated in the presence of doxycycline (Dox) at various concentrations. On day 10, cells were fixed and stained with Oil red O. (b) Protein production profile during adipogenesis in the absence (−) or presence (+) of doxycycline. Cell lysates were collected at indicated time points and subjected to western blotting. D0, D2, D4 and D7 indicate day 0, day 2, day 4 and day 7 after initiation of adipocyte differentiation. The white and black arrows indicate larger and smaller isoforms of C/EBPα, respectively. (c–e) Identification of critical time period for inhibitory effect of CHOP on adipogenesis. Adipocyte differentiation was stimulated in Tet-CHOP cells in the presence of doxycycline for various time intervals as indicated (d). On day 10, cells were stained with Oil red O and representative images from each condition are shown (c). Staining was quantified and presented as means ± SEM of three independent experiments with triplicates (e). Scale bar indicates 100 μm. (f–j) Effect of transient CHOP overproduction on (f–i) gene expression and (j) protein production during adipogenesis. Tet-CHOP cells were differentiated in the absence (white bars) or presence of doxycycline during the entire differentiation period (black bars) or transient doxycycline treatment (grey bars) as indicated. D–2 indicates 2 days before the adipocyte differentiation; D0, D2, D4 and D7 indicate day 0, day 2, day 4 and day 7 after initiation of adipogenesis. (f–i) At the days indicated, total RNAs were isolated for analysis by real-time RT-PCR. Data are presented as means ± SEM of three independent experiments with triplicates. *p < 0.05 and **p < 0.01 vs control. (j) Protein production was analysed by western blot under the same conditions as (f–i). (k) Oil red O staining. Adipogenesis was stimulated in primary Chop +/+ and Chop −/− MEFs in the absence (vehicle) or presence of Tm (50 ng/ml). On day 10, cells were fixed and stained with Oil red O. Representative images from indicated MEF lines are shown. Scale bar indicates 100 μm. (l) Quantification of Oil red O staining in (k). White bars, Chop +/+; grey bars, Chop −/−. (m–p) Gene expression profiles for markers of mature adipocytes in Chop +/+ (white bars) and Chop −/− MEFs (grey bars) in the absence or presence of Tm (50 ng/ml). *p < 0.05 and **p < 0.01. AU, arbitrary units

Full size image

Chop deletion increases obesity in vivo

Next, we investigated the effect of Chop deletion in mice. After being fed with an HFD, Chop −/− mice gained more body weight, had a higher percentage of body fat and a lower percentage of lean mass compared with Chop +/+ mice (Fig. 6a–f). Analysis of the distribution of adipocyte sizes demonstrated no significant difference between genotypes of HFD-fed mice (Fig. 6g, ESM Fig. 7a). In addition, there was no significant difference in number of cells, or their proliferation rates, from the SVF of epididymal fat tissues between Chop +/+ and Chop −/− mice (Fig. 6h, i). These results suggest that the increased fat mass in the HFD-fed Chop −/− mice results from increased adipocyte number. In addition, there was no significant difference in the levels of serum NEFA, cholesterol, adiponectin or triacylglycerol, as well as liver triacylglycerol (Fig. 6j–m; ESM Fig. 7b) between the genotypes. Similar results were obtained from 1-year-old mice, where the body weight and fat mass was significantly greater in Chop −/− compared with wild-type mice, without significant differences in the size distribution of adipocytes (ESM Fig. 8).

Fig. 6
figure 6

Chop deletion increases body fat mass in mice fed a HFD or in genetically obese Lepr db/db mice. (a–f) Weight gain, body fat mass, body lean mass, fluid mass, blood glucose and insulin levels. Chop +/+ (n = 10) and Chop −/− (n = 8) mice were fed an HFD for 13 weeks and weight gain, body fat mass, body lean mass and fluid mass were measured using an NMR-based analyser. (g) Histogram showing the distribution of adipocyte size in mice fed an HFD for 13 weeks. The x-axis represents the logarithm of cell size in pixels, while the y-axis shows the percentage of cells having a given size for each group of mice. White bars, Chop +/+; black bars, Chop −/−. (h) The number of cells from the SVF of adipose tissues from Chop +/+ (n = 4) and Chop −/− (n = 4) mice at the age of 6–8 weeks on regular diet. (i) Proliferation rate of cells from SVF. Cells from SVF were plated in replicate and the number of cells was counted after 4 days using an automated cell counter. Each suspension was counted twice. The y-axis represents the percentage of the initial cell number. Serum levels of NEFA (j), cholesterol (k) and triacylglycerol (l) from mice fed an HFD for 13 weeks. Sera from Chop +/+ (n = 10) and Chop −/− (n = 8) mice were collected and analysed. (m) Hepatic triacylglycerol levels. Liver samples were collected after 13 weeks of HFD for analysis. (n) Body weights of Chop +/+ Lepr db/db (n = 5) and Chop −/− Lepr db/db (n = 5) mice at 6 months of age. (o) Representative images of epididymal fat pads from Chop +/+ Lepr db/db and Chop −/− Lepr db/db mice at 6 months of age. Scale bar indicates 100 μm. (p, q) ITTs. Chop +/+ Lepr db/db (n = 5, circles) and Chop −/− Lepr db/db (n = 5, squares) mice at 6 months of age were challenged with insulin (0.75 U/kg body weight) and blood glucose levels were measured. Data are presented as means ± SEM. **p < 0.01, *p < 0.05. BW, body weight

Full size image

We also investigated the effect of Chop deletion in the leptin receptor-deficient mouse, Lepr db/db. Since Lepr db/db mice lose appetite control due to deficiency of the leptin receptor, these mice show severe obesity without an HFD. Chop −/− mice were crossed with Lepr db/db mice to generate Chop −/− Lepr db/db and Chop +/+ Lepr db/db mice. Strikingly, Chop deletion increased body weight in the Lepr db/db mice (Fig. 6n), without a significant difference in adipocyte size (Fig. 6o). However, there was no difference in insulin sensitivity between the genotypes despite the increased body weight in Chop −/− Lepr db/db mice (Fig. 6p, q).

IRE1α signalling does not alter adipogenesis

Since the expression and phosphorylation of IRE1α and Xbp1 mRNA splicing were increased during adipogenesis (Fig. 1a, b), we investigated the requirement for IRE1α in adipogenesis. For this purpose, stable 3T3-L1 cell lines (Tet-IRE) were generated which induce production of IRE1α under the control of a doxycycline-responsive promoter (Fig. 7a). Forced production of IRE1α by doxycycline treatment induced splicing of Xbp1 mRNA, as well as Atf6α and Grp78 mRNA expression (Fig. 7b). Induction of IRE1α by doxycycline treatment in Tet-IRE cells did not affect adipocyte differentiation, indicating that pre-emptive activation of IRE1α had little effect on adipogenesis (Fig. 7c).

Fig. 7
figure 7

IRE1α signalling is not required for adipogenesis in vitro or in vivo. (a, b) Generation of 3T3-L1 cells that produce wild-type IRE1α under the control of a tetracycline-inducible promoter (Tet-IRE). Total protein lysates and total RNA were collected at indicated time points after doxycycline (5 μg/ml) treatment and analysed by western blot (a) and quantitative real-time PCR (b). White bars, day 0 (D0); grey bars, day 1 (D1); black bars, day 2 (D2). **p < 0.01, *p < 0.05 vs D0. (c) IRE1α overproduction does not affect adipogenesis. Differentiation was induced in Tet-IRE cells in the presence of doxycycline (Dox) at various doses. On day 10, cells were stained with Oil red O. Representative images from each condition are shown. (d–g) Gene expression profile for stable 3T3-L1 cells expressing dominant-negative mutant Ire1aK907A under the control of a tetracycline-inducible promoter (Tet-IREKA) in the absence or presence of Tm-induced ER stress. Data are presented as means ± SEM of three independent experiments with triplicates. (h) Images of Oil red O staining for 3T3-L1 cells expressing mutant Ire1aK907A under the same conditions as (d–g). (i) Quantitative real-time RT-PCR to verify IRE1α inhibition. 3T3-L1 cells were treated with Tm (2 μg/ml) in absence (grey bars) or presence of IRE1α inhibitor (30 μmol/l, black bars) for 16 h for analysis of gene expression. ‘Mock’ indicates DMSO treatment (white bars). (j) Oil red O staining. Adipogenesis was stimulated in 3T3-L1 cells in the absence (vehicle) or presence of IRE1α inhibitor (30 μmol/l). On day 10, cells were fixed and stained with Oil red O. Representative images are shown. Scale bar indicates 100 μm. (k) Quantification of Oil red O staining is shown in right panel. Data are presented as means ± SEM. **p < 0.01, *p < 0.05. AU, arbitrary units

Full size image

Next, we investigated the requirement for IRE1α in adipogenesis. First, a stable 3T3-L1 cell line (Tet-IREKA) was generated with inducible expression of the dominant-negative Ire1aK907A RNase mutant (ESM Fig. 9). The Ire1α Κ907Α mutant cannot initiate Xbp1 mRNA splicing [15] and inhibits Xbp1 splicing by wild-type IRE1α, even in the presence of ER stress (Fig. 7d–g). Overexpression of the K907A mutant during adipogenesis did not alter adipogenesis in the 3T3-L1 cells (Fig. 7h). Second, we treated 3T3-L1 cells with an IRE1α inhibitor [34] to inhibit Xbp1 mRNA splicing (Fig. 7i) during adipogenesis. Consistent with the above results, the IRE1α inhibitor did not affect adipocyte differentiation in 3T3-L1 cells (Fig. 7j, k).

Discussion

In this study, we demonstrate that eIF2α phosphorylation increases CHOP production to repress adipocyte differentiation in response to ER stress in vitro and in vivo. Our conclusion is supported by the following observations. First, activation of the UPR inhibited adipogenesis in 3T3-L1 cells. Second, pre-emptive phosphorylation of eIF2α by forced dimerisation of Fv2E-PERK inhibited adipogenesis in 3T3-L1 cells. Third, homozygous Ser51Ala mutation in eIF2α to prevent phosphorylation significantly enhanced adipogenesis in MEFs compared with wild-type MEFs. Fourth, heterozygous Ser51Ala mutation in eIF2α increased obesity and adipocyte number in HFD-fed mice. Fifth, transient induction of CHOP was sufficient to inhibit adipogenesis in 3T3-L1 cells. Sixth, Chop deletion increased obesity and adipocyte number upon HFD feeding, as well as in Lepr db/db mice. Finally, the IRE1α pathway had little effect on adipogenesis.

Several lines of evidence suggest that the UPR is activated during cellular differentiation [18, 35–39]. Consistent with these findings, during adipogenesis, we observed eIF2α phosphorylation and IRE1α activation. However, we also found that the ratio of phosphorylated eIF2α to total eIF2α was reduced at day 2, suggesting protein synthesis is increased during the early phase of adipogenesis, possibly due to increased synthesis of proteins required for differentiation. An elevated rate of protein synthesis would, in turn, increase phosphorylation of eIF2α later in differentiation and lead to the induction of downstream target molecules, including CHOP, as observed in our study.

Our results are consistent with the findings of Basseri et al that showed both eIF2α phosphorylation and the total amount of eIF2α were reduced at days 1–2, increased at days 3–7 and again reduced in the late period of 3T3-L1 differentiation [18]. In addition, the induction pattern of GRP78 and CHOP was also similar to what we observed. However, in contrast to our conclusion, they suggested that ER stress is required for adipocyte differentiation because the chemical chaperone 4-phenyl butyric acid (PBA) reduced ER stress and inhibited adipogenesis [18]. The difference might result from the source of ER stress and mechanism of UPR activation. Physiological UPR activation that occurs during differentiation might facilitate adipogenesis, whereas more severe UPR activation caused by exogenous stimuli might inhibit adipogenesis. It is also possible that PBA is doing something other than improving ER protein folding. For example, it is notable that PBA is a histone deacetylase inhibitor [40].

Our findings also show that the amount of both phosphorylated and total IRE1α increases during adipogenesis. Induction of IRE1α appeared as early as day 2 and was sustained at high levels up to day 7. Xbp1 mRNA splicing correlated with increased phosphorylation of IRE1α. However, these kinetics were quite different from those of eIF2α phosphorylation, which was reduced at day 2, suggesting that each subpathway of the UPR is regulated independently. Interestingly, either increasing IRE1α activation via IRE1α overexpression or inhibiting IRE1α via expression of a dominant-negative Ire1a RNase mutant, did not alter adipocyte differentiation. Therefore, it remains unclear why IRE1α is activated during adipogenesis. As recent studies identified a role for XBP1 and IRE1α in lipid metabolism [41–43], it is possible that IRE1α affects lipid metabolism in the differentiated mature adipocyte. Our findings conflict with those of Sha et al that suggest IRE1α activation and subsequent Xbp1 mRNA splicing is required for adipogenesis [19]. They proposed that physiological UPR activation occurs during the early period of adipogenesis and is maintained at a relatively low level in mature adipocytes. Although it is not possible to know the reason(s) for the different observations, adipogenesis is very inefficient and variable in immortalised MEFs. This may be due to variability of the differentiation state in immortalised MEFs that may change with passage number and growth conditions [44]. Our studies analysed the role of the IRE1α pathway in the well-established system of 3T3-L1 cells. We believe the role of IRE1α and XBP1 in adipogenesis deserves further investigation.

In obesity, mature adipocytes are exposed to elevated NEFA, inflammation and nutrient and oxygen deprivation [14, 29, 45, 46]. Since these factors can cause ER stress to activate the UPR, they likely influence pre-adipocytes and/or adipogenic stem cells [14]. As the demand for fat storage increases during the progression of obesity, a defect in adipose tissue expansion would result in storage of excessive NEFA in other peripheral tissues such as liver and muscle. The deposition of lipid in liver and muscle would increase insulin resistance [47–50]. Therefore, increasing the number of functional adipocytes would preserve insulin sensitivity by providing a storage depot for fatty acids. Under conditions of extreme obesity, UPR activation in pre-adipocytes would inhibit the generation of new adipocytes. This would limit the capacity to store excessive lipids inside fat depots, which in turn would exacerbate insulin resistance in peripheral tissues. In support of this notion, we found that although HFD-fed Eif2a S/A mice and HFD-fed Chop −/− mice were more obese, there was no significant increase in the serum NEFA, cholesterol or triacylglycerol, or in hepatic triacylglycerol or insulin resistance compared with HFD-fed Eif2a S/S and Chop +/+ littermates, respectively.

In conclusion, our findings show that eIF2α phosphorylation and CHOP production are reduced during the early phase of adipogenesis in order to permit adipocyte differentiation. Factors that would cause ER stress, such as inflammation, nutrient deprivation and elevated NEFA, would increase eIF2α phosphorylation and CHOP production. In this manner, eIF2α phosphorylation would exacerbate the metabolic consequences of obesity by inhibiting adipogenesis and limiting lipid storage in adipose tissue. Therefore, tight regulation of eIF2α phosphorylation is required to optimise adipogenesis, preserve metabolic homeostasis and limit lipid deposition in liver and muscle.