The Extracellular Signal–Regulated Kinase Isoform ERK1 Is Specifically Required for In Vitro and In Vivo Adipogenesis

We generated ERK1^−/− mice from heterozygotes issued from six backcrosses with C57BL/6 animals, as previously described (12), and control came from the same backcross. Mice were housed on a 12-h light/dark schedule and had free access to water and food. High-fat (45% fat, 35% carbohydrates, and 20% proteins) and standard diet (10% fat, 70% carbohydrates, and 20% proteins) were purchased from SAFE (Epinay/Orge, France). The experiments were conducted following standard ethical guidelines (European Union guidelines on animal laboratory care) and were approved by the Faculty of Medicine, University of Nice-Sophia Antipolis ethical committee.

Mouse embryonic fibroblasts were prepared on day 14 postcoitum as described previously (14). Preadipocytes from adult tissue were isolated from subcutaneous fat pads of 8-week-old mice (15). Next, 2 days’ postconfluent cells were treated for 48 h with the adipocyte differentiation cocktail containing 5 μg/ml insulin, 0.25 μmol/l dexamethasone, 0.5 mmol/l IBMX (3-isobutyl-1-methylxanthine), and 10 μmol/l thiazolidinedione. The medium was then replaced for 6 days by Dulbecco’s modified Eagle’s medium with insulin and thiazolidinedione only. Triglyceride accumulation has been measured using a kit (Triglyceride 100; ABX Diagnostics, Montpellier, France). Glycerol-6-phosphate dehydrogenase activity was measured as described (16).

Gene expression analysis was performed with an ABI Prism 7000 (Applied Biosystems) and SYBR Green reagents (Eurogentec, Seraing, Belgium). cDNAs were synthesized from 2 μg of total RNA using Superscript II reverse transcriptase (Invitrogen). The set of primers was designed according to the manufacturer software. Samples contained 1 × SYBR Green Master Mix, 0.5 μmol/l primers, and 1/50 synthesized cDNA in a 25 μl volume. PCR conditions were as follows: 10 min at 95°C, then 40 cycles of 15 s at 94°C, 30 s at 60°C, and 1 min at 72°C. We used 36B4 as an internal control.

Glucose and insulin tolerance tests were performed on 13- to 14-week-old animals at 7 weeks after the beginning of the diet. Glucose (2 g/kg) and insulin (0.75 IU/kg) were administrated by intraperitoneal injection in awake mice. Blood samples were withdrawn from the tail vein at the indicated time, and glycemia was determined using a biochemical assay (Glucose PAP 250; ABX Diagnostics) and Accu-Check active bands (Roche Diagnostics, Mannheim, Germany). Quantification of triglycerides and free fatty acids was performed using, respectively, Triglycerides 100 (ABX Diagnostics) and NEFA-C (Wako Chemicals, Neuss, Germany) biochemical kits.

Mice were placed for 22 h in a metabolic cage connected to an open-circuit, indirect calorimetry system, with the air flow adjusted to 0.5 l/min. The temperature in the metabolic cage (24 ± 1°C) was stable and controlled throughout the experiment. Oxygen consumption and carbon dioxide production were recorded at 10-s intervals, using a computer-assisted data acquisition program. Computer-assisted processing of respiratory exchanges and spontaneous activity signals made it possible to compute that part of the total metabolic rate devoted to fueling the energy cost of activity and thus (by continuously extracting the energy expended with activity) to compute the resting metabolism of these free-moving mice (17). The mice were housed in the metabolic cage at 1000h without food but with water. Postabsorptive resting energy expenditure (equivalent to basal metabolism) was measured between 1600 and 1800 h. At 1800 h, 1 g of the usual high-fat diet was given to the mice, and the thermogenic response to feeding (metabolism and respiratory quotient [RQ]) was computed for 8 h as the increase of metabolism and RQ above premeal baseline levels.

Tissue extracts were prepared using lysis buffer (9) and were analyzed by Western blot using antibodies against ERK phosphorylated on Thr202/Tyr204 (Cell Signaling Technology) and total ERK (Santa Cruz).

Independently of the defect in thymocyte maturation and enhanced long-term memory (12,13), no obvious phenotype has been described for ERK1^−/− mice. To determine whether ERK1 invalidation had consequences on adiposity in mice, total carcass lipid content was measured and found to be reduced by 33% in ERK1^−/− mice compared with ERK1^+/+ animals at 3 and 10 weeks of age (Fig. 1A). Most organs were roughly similar in size between the genotypes, and no difference in brown adipose tissue and liver weight was noticed (data not shown). However, fat depots were significantly smaller in ERK1^−/− animals compared with wild-type animals (Fig. 1B–E). As important differences in adipose tissue were observed, we analyzed adipocyte size and number. Adipocytes from knockout animals were slightly smaller (data not shown), and, more importantly, a diminution of 40–50% of the number of adipocytes from subcutaneous depots was observed in ERK1^−/− mice (Fig. 1F). Food intake may explain the reduction of fat depots. Therefore, we evaluated this parameter in ERK1^−/− and ERK1^+/+ mice. Food intake was similar for each group of mice (knockout animals versus control animals: 28.8 ± 2.4 vs. 31.7 ± 2.5 g · week⁻¹ · animal⁻¹, P = 0.76). Finally, our results indicate that the lack of ERK1 gene correlates in vivo with decreased adiposity and adipocyte hypocellularity.

To investigate the role of the ERK pathway in the development of obesity, mice were subjected to high-fat diet (45% of total calories come from fat). We asked whether high-fat diet induced the activation of the ERK pathway in white adipose tissue, liver, and muscle. As shown in Fig. 2 (and data not shown), high-fat diet had no effect on ERK1 and -2 expression in the three tissues. Furthermore, ERK activity is not modified in liver (Fig. 2A) and muscle (data not shown) on hypercaloric regimen. By contrast, total ERK activity was more than three times higher in white adipose tissue from high-fat-diet mice compared with animals on standard diet (Fig. 2B). Interestingly, adipocytes from obese patients with type 2 diabetes have higher ERK activity than control patients (18), suggesting an important correlation between elevated ERK activity and obesity.

To determine whether the invalidation of ERK1 prevents the occurrence of obesity, we analyzed the incidence of high-fat diet on wild-type and knockout mice. As expected, control mice became markedly obese on high-fat diet compared with their littermates on standard diet (10% of total calories from fat). In contrast, ERK1^−/− mice did not develop obesity (Fig. 3A). Indeed, on high-fat diet, the average weight gain per week was significantly higher in control mice than in ERK1^−/− mice (2.9 ± 0.4 vs. 1.7 ± 0.3 g/week, respectively; P < 0.008). In ERK1^−/− animals under high-fat diet, inguinal axilliary fat pads and subcutaneous fat thickness were reduced by 51 and 30%, respectively (Fig. 3B). Interestingly, we noticed no difference in epididymal fat pad weight under high-fat diet, suggesting that depending on the location of the depot, the lack of ERK1 may affect adipose tissue development differently. Alternatively, because ERK1 invalidation did not completely block the development of adipose tissues, we cannot exclude that epididymal fat pads but no other depots have reached their maximal development in both wild-type and ERK1^−/− mice. As observed for mice on standard diet, no difference of food intake was noticed between ERK1^−/− and ERK1^+/+ mice on high-fat diet (data not shown).

High-fat diet is frequently associated with insulin resistance. As expected, high-fat diet led to hyperglycemia in fed wild-type animals (246 ± 43 vs. 171 ± 15 mg/dl on high-fat vs. normal diet, respectively; P = 0.01). In contrast, fed ERK1^−/− mice did not develop hyperglycemia (168 ± 13 vs. 196 ± 25 mg/dl for high-fat vs. normal diet, respectively; P = 0.1). Fasted mutant and control mice did not develop hyperglycemia, and no difference was found concerning triglycerides and free fatty acids in both groups under normal or high-fat diet (Table 1). We then determined the insulin sensitivity of the animals, performed a glucose tolerance test, and performed an insulin tolerance test. Wild-type mice on high-fat diet became glucose intolerant, a defect that did not appear in ERK1^−/− animals (Fig. 3C). Likewise, the insulin tolerance test demonstrated severe insulin resistance in only wild-type mice on high-fat diet (Fig. 3D). Altogether, our results show that ERK1 invalidation protects from high-fat diet–induced obesity and glucose intolerance.

We then explored activity and energy expenditure in ERK1^−/− mice. No difference in spontaneous activity was noticed in ERK1^−/− compared with ERK1 wild-type mice (data not shown). Under fasted conditions, we observed no significant difference in the basal metabolic rate in ERK1^−/− mice compared with wild-type animals (Fig. 4A). Basal RQ was identical in both groups (Fig. 4B), indicating that the ratio of glucose to lipid used to fuel basal metabolism was the same in mutant and control mice. We then investigated the thermogenic response to a calibrated test meal (1 g) in mice. The ERK1^−/− mice displayed a higher postprandial increase in metabolic rate as well as in RQ (Fig. 4A and B), the higher RQ indicating that the increased thermogenic response to feeding was fueled by an increased rate of glucose oxidation (data not shown). Therefore, it can be extrapolated that under standard conditions (ad libitum feeding), energy expenditure may be significantly higher in ERK1^−/− mice, which may contribute to the resistance to high-fat diet–induced obesity.

To gain insight into cellular mechanisms involved in reduced adiposity and resistance to high-fat diet–induced obesity, we analyzed adipocyte differentiation in ERK1^−/− cells. Mouse embryo fibroblasts isolated from ERK1^−/− and wild-type embryos were induced to differentiate, as described in research design and methods. After differentiation, the accumulation of cytoplasmic triacylglycerol as lipid droplets, a marker of adipocyte differentiation, was strongly reduced in ERK1^−/− cells (Fig. 5A). Concomitantly, the activity of glycerol-6-phosphate dehydrogenase, a lipogenic enzyme expressed in adipocyte, was reduced by half (Fig. 5B). Furthermore, the expression of specific adipocyte markers such as aP2, peroxisome proliferator–activated receptor-γ (PPAR-γ), adiponectin, and leptin was diminished by >50% in knockout mouse embryo fibroblasts (Fig. 5C). On the contrary, myogenin, a marker of the myocyte lineage, was identically expressed in wild-type and ERK1^−/− cells (Fig. 5D). This result clearly shows that ERK1 gene invalidation specifically impairs adipocyte differentiation and does not seem to interfere with myogenesis. This is in agreement with our data obtained in embryonic stem cells showing that inhibition of the ERK pathway interferes specifically with adipogenesis and not with myogenesis (9). The recruitment of new adipocytes from precursor cells takes place all life long. To determine whether adipocyte differentiation is affected in ERK1^−/− adult tissues, cells from the stromal vascular fraction of white adipose tissue were isolated and induced to differentiate into adipocytes. As found in embryonic cells, after Oil Red O staining, we observed a strong reduction of adipocyte formation in ERK1^−/− cells (Fig. 5E). Indeed, the total amount of triglyceride in ERK1^−/− cells was reduced by 80% compared with wild-type cells (Fig. 5F). Our results demonstrate that ERK1 is required for adipocyte differentiation and plays a critical role in adipogenesis, from embryonic stages through adulthood.

Impaired adipogenesis may be caused either by a decrease in the number of adipocyte precursors or by a defect in terminal differentiation. Pref-1 is an inhibitor of differentiation and a specific preadipocyte marker not expressed in mature adipocytes (19–21). Therefore, we assume that the level of Pref-1 expression, investigated in mouse embryo fibroblasts and preadipocytes before differentiation, is a direct reflection of the pool of adipogenic precursors. We analyzed its expression in mouse embryo fibroblasts and cells of the stromal vascular fraction before differentiation and also in white adipose tissue. Interestingly, compared with control, Pref-1 expression was markedly decreased in ERK1^−/− cells and tissues (Fig. 5G). This result suggests that impaired adipogenesis observed in ERK1^−/− cells may result from a reduction of the pool of preadipocytes.

To get insights into molecular mechanisms underlying the defect of adipogenesis in ERK1^−/− cells, we measured ERK activation upon addition of the adipocyte differentiation media to mouse embryo fibroblasts (Fig. 6A). As in standard growing conditions (12), the ERK activity of mouse embryo fibroblasts induced to differentiate was mainly caused by ERK2. Indeed, ERK1 activity represents an average of only 23 ± 11% of the total ERK activity throughout the time course of activation (quantification of Fig. 6A). Of note, the activity of ERK2 was not significantly changed in ERK1^−/− cells (Fig. 6A), showing that there is no compensation in gene expression or kinase activation in these cells. We then analyzed the contribution of the ERK pathway to cell proliferation during adipocyte differentiation. Indeed, during the early stages of adipocyte differentiation, opposite results have been found concerning the requirement of cell division after the cells have reached confluence (6,22,23). Therefore, we tested the hypothesis that the reduction of the adipocyte precursor cell number in ERK1^−/− cell cultures is caused by a slower growth rate of these cells. During the adipocyte differentiation protocol, we assessed cell proliferation by counting the cells after confluence (Fig. 6B). In this model, proliferation of wild-type cells was inhibited by the addition of differentiation medium, showing that cell division is not necessary for adipogenesis in these cells (Fig. 6B). The ERK1^−/− mouse embryo fibroblasts had growth rates identical to ERK1^+/+ cells in standard or differentiation media (Fig. 6B). This result demonstrates that ERK1 gene deficiency does not affect cell growth. Interestingly, the addition of U0126 inhibited cell growth similarly in both knockout and wild-type cells (Fig. 6B). Therefore, ERK2 seems to play an important role in mouse embryo fibroblast proliferation independently of ERK1.

We then investigated the role of the ERK2 isoform in adipocyte differentiation. Because ERK2 invalidation results in early embryonic death (embryonic day 6.5–8.5), thus avoiding the isolation of mouse embryo fibroblasts (10,11), we analyzed the effect of the addition of U0126 on wild-type mouse embryo fibroblasts and on residual adipocyte differentiation in ERK1^−/− cells. U0126 led to a 45% inhibition of adipogenesis in wild-type mouse embryo fibroblasts, which is identical to the decreased adipogenesis observed in knockout cells. Importantly, the inhibitor of the ERK pathway did not significantly affect residual adipocyte formation of ERK1^−/− cells (Fig. 6C). Therefore, while still presenting 80% of total ERK activity, ERK1^−/− cells had impaired adipogenesis, and the addition of U0126 had no effect on this adipogenesis. These results suggest that ERK1 plays a specific role in adipogenesis, whereas the ERK2 isoform is not involved. The latter is required for cell proliferation after confluence, a process not involved in the adipocyte differentiation of these cells (Fig. 6B).

Because c-Jun NH₂-terminal kinase 1 (JNK1)-deficient mice are resistant to the development of obesity on high-fat diet (24), we determined whether the JNK pathway interferes with adipocyte differentiation by treating cells with the specific JNK inhibitor SP600125. We found that the inhibitor did not affect adipocyte formation in both wild-type and ERK1^−/− mouse embryo fibroblasts (Fig. 6C). Thus, this result suggests that the JNK pathway is not required for optimal adipocyte differentiation. Indeed, no evidence for impaired adipogenesis was described in JNK1^−/− animals (24). We also found that JNK is not involved in the early stages of adipocyte differentiation (9).

Many studies performed in preadipocyte cell lines showed that the ERK/mitogen-activated protein kinase pathway is implicated in adipocyte differentiation (2–9,25). However, these studies lead to controversial results and were carried out using specific inhibitors of the ERK pathway such as the MEK inhibitors (PD98059 or U0126) or antisense oligonucleotides. Importantly, these approaches do not discriminate between ERK1 and ERK2. Here we show, using a genetic approach, that disruption of ERK1 in embryonic fibroblasts and adult preadipocytes leads to impaired adipocyte differentiation. We measured the expression of Pref-1 in two cell types before differentiation, at the preadipocyte stage. Pref-1 mRNA level is low in ERK1^−/− cells, which differentiate poorly in adipocytes. We therefore suggest that the low level of Pref-1 in mouse embryo fibroblasts and stromal vascular cells is caused by a reduced number of adipocyte precursors. Compared with wild-type cells, Pref-1–reduced expression suggests that this defect appears early in adipogenesis and could be caused by a reduced pool of preadipocytes. This result is in good agreement with our previous observation in embryonic stem cells (9). The positive role for the ERK pathway early in adipocyte differentiation can be compatible and may precede an opposed negative function later on in terminal differentiation. This hypothesis reconciles a positive role of ERK1 in adipogenesis (our results and others) to the late inhibitory effect of ERK activation on the adipogenic transcription factor PPAR-γ (2). This latter study demonstrates that phosphorylation of PPAR-γ by ERK inhibits adipocyte differentiation. We observed no difference in PPAR-γ transcriptional activities between ERK1^−/− and wild-type cells (data not shown), indicating that invalidation of ERK1 has no effect on PPAR-γ transactivation.

ERK1 and -2 share an overall 75% identity at the amino acid level and are activated by the same stimuli. However, unlike ERK1^−/− mice, ERK2 knockout animals are not viable, suggesting that the two isoforms have distinct biological functions and are not redundant (10–12). Regarding adipocyte differentiation, exposure of mouse embryo fibroblasts, to the differentiating medium equally activates both isoforms, as observed in 3T3-L1 preadipocyte cell lines (8). Here, we show that ERK1 invalidation does not alter cell growth, whereas ERK2 inhibition blocks ERK1^−/− cell proliferation. By contrast, ERK2 inhibition does not further affect the residual adipogenesis observed in ERK1-deficient cells. Therefore, by revealing that ERK1 deficiency affects specifically adipocyte differentiation, whereas ERK2 is required for proliferation, our results extend to adipogenesis the notion of distinct biological functions for the two genes. The molecular basis for such divergent activities is not yet understood. One hypothesis is that ERK1 might have preferential substrates implicated in adipogenesis.

We show here that the defect observed in vitro correlates with the reduced adiposity observed in ERK1^−/− mice. Interestingly, as opposed to other tissues examined where ERK1 activity is weaker than ERK2, in white adipose tissues ERK1 and -2 are activated at the same level, suggesting an important role for this isoform. According to this observation, ERK1-deficient mice have reduced adiposity and fewer adipocytes than wild-type animals. Furthermore, as for preadipocyte and mouse embryo fibroblasts, we observed no compensatory expression of ERK2 in ERK1^−/− adipose tissue (data not shown). Recently, several studies have shown that the adipose tissue contains pluripotent stem cells (26–31), and they validate the concept that recruitment of new adipocytes from stem cells occurs from birth to the adult stage. White adipose tissues and cells from the stroma vascular fraction of ERK1^−/− animals have reduced expression of the preadipocyte marker Pref-1. This suggests that, as observed in vitro, the pool of preadipocytes is diminished and that ERK1 is required during early stages of in vivo adipogenesis.

Obesity is characterized by the hypertrophy and the hyperplasia of adipocytes, and new cells are recruited through adipocyte differentiation. High-fat diet induces a strong activation of the ERK pathway, specifically in white adipose tissue and not in other tissues. This activation is required for the development of obesity because we found that ERK1^−/− animals are markedly resistant to high-fat diet–induced obesity. Therefore the defect observed in the development of the white adipose tissue of these animals may be sufficient to partly protect them from obesity.

We showed in this study that basal metabolism on a per-mouse basis was not different between mutant and control mice. Furthermore, we found that the thermogenic response to ingestion of a meal was higher in ERK1^−/− mice, which may contribute, in addition to impaired adipogenesis, to a resistance to obesity. We also demonstrated that the increased thermogenic response to feeding was fueled by an increased rate of glucose oxidation. It has been shown that meal-induced elevated thermogenesis may result from a higher insulin sensitivity in peripheral tissues (32–34), leading to increased glucose oxidation. These data are thus in agreement with the better insulin sensitivity observed in mutant mice.

In conclusion, our results clearly link the invalidation of ERK1 to reduced adiposity and resistance to high-fat diet–induced obesity due to impaired adipocyte differentiation and higher postprandial metabolism. Furthermore, we have identified specialized and separate functions for the two isoforms: ERK1 in adipocyte differentiation and ERK2 in mouse embryo fibroblast proliferation. Further studies will be required to understand, at the molecular level, the diverging pathways between the two kinases at the level of their substrates. Our data suggest that targeting specifically the ERK1 isoform and not ERK2 would be of particular interest to fight obesity and insulin resistance.

This work was funded by grants from INSERM, the Bettencourt-Schueller foundation, and the Association pour la Recherche pour le Cancer (ARC 4525). M.A. and L.C. were supported by a fellowship from INSERM-Région Provence Alpes Côte d’Azur and La lique nationale centre le cancer.

We thank Myriam Bost for her support. We are indebted to Jean François Tanti, Phillippe Gual, and Thierry Grémeaux for their helpful contribution to this work. We also thank the Animal Facility of the Faculté de Medicine of Nice.