Acknowledgements We thank S. Akira for the MyD88-deficient mice L. Kopp and C. Pasare for discussion and reading the manuscript C. Annicelli for maintaining the mouse colony L. Evangeliate for technical assistance L. Alexopoulou and K. Kobayashi for technical advice and the Yale Cancer Center Transgenic Mouse and Gene Targeting Resource for blastocyst injections. This work was supported by the NIH, the Searle Foundation (R.M.) and the Howard Hughes Medical Institute (R.M., T.H., G.B. and R.F.). Competing interests statement The authors declare that they have no competing financial interests. Correspondence and requests for materials should be addressed to R.M. (e-mail: ruslan.medzhitov@yale.edu). .............................................................. A central role for JNK in obesity and insulin resistance Jiro Hirosumi*��, Gurol �� Tuncman*��, Lufen Chang���, Cem Z. Gorgun*, �� �� K. Teoman Uysal*, Kazuhisa Maeda*, Michael Karin��� & Gokhan �� S. Hotamisligil* * Division of Biological Sciences and Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, Massachusetts 02115, USA ��� Laboratory of Gene Regulation and Signal Transduction, Department of Pharmacology, University of California San Diego, School of Medicine, 9500 Gilman Drive, La Jolla, California 92093, USA �� These authors contributed equally to this work ............................................................................................................................................................................. Obesity is closely associated with insulin resistance and estab- lishes the leading risk factor for type 2 diabetes mellitus, yet the molecular mechanisms of this association are poorly under- stood1. The c-Jun amino-terminal kinases (JNKs) can interfere with insulin action in cultured cells2 and are activated by inflammatory cytokines and free fatty acids, molecules that have been implicated in the development of type 2 diabetes3,4. Here we show that JNK activity is abnormally elevated in obesity. Furthermore, an absence of JNK1 results in decreased adiposity, significantly improved insulin sensitivity and enhanced insulin receptor signalling capacity in two different models of mouse obesity. Thus, JNK is a crucial mediator of obesity and insulin resistance and a potential target for therapeutics. Obesity and type 2 diabetes are the most prevalent and serious metabolic diseases they affect more than 50% of adults in the USA5. These conditions are associated with a chronic inflammatory response characterized by abnormal cytokine production, increased acute-phase reactants and other stress-induced molecules3. Many of these alterations seem to be initiated and to reside within adipose tissue, an unusual site for inflammation3. Elevated production of tumour necrosis factor (TNF)-a by adipose tissue decreases sensi- tivity to insulin and has been detected in several experimental obesity models and obese humans6,7. Free fatty acids (FFAs) are also implicated in the aetiology of obesity-induced insulin resist- ance, although the molecular pathways involved in their action remain unclear4. Because both TNF-a and FFAs are potent JNK activators8���11, we asked whether obesity is associated with altera- tions in stress-activated and inflammatory responses through this signalling pathway and whether JNKs are causally linked to aberrant metabolic control in this state. We examined JNK activity in liver, muscle and adipose tissues of various models of obesity compared with lean controls to determine whether obesity activates the JNK pathway. In both dietary and genetic (ob/ob) models of obesity, there was a significant increase in total JNK activity in all tissues tested (Fig. 1a). In these tissues there was no apparent difference in the expression of either JNK1 or JNK2 polypeptides, suggesting that the activity of one or both of these kinases is increased in response to obesity. To test the functional significance of this alteration in the pathogenesis of obesity, insulin resistance and type 2 diabetes, we induced obesity in mice lacking either JNK1 (Jnk12/2) or JNK2 (Jnk22/2). Jnk12/2 or Jnk22/2 mice and their control littermates (Jnk1��/�� or Jnk1��/2 and Jnk2��/�� or Jnk2��/2) were placed on a high-fat (50% of total calories derived from fat) and high-caloric diet (5286 kcal kg21 Bioserve, Frenchtown, NJ, USA) along with a control group of each genotype on a standard diet. On the high-fat diet, both controls and Jnk22/2 mice developed marked obesity in comparison with mice kept on standard diet (Fig. 1b and c). Weight gain in Jnk2��/��, Jnk2��/2 and Jnk22/2 animals was indistinguish- able on either standard or high-fat diet. However, weight gain on both standard and high-fat diets was significantly decreased for the Jnk12/2 group (Fig. 1d and e). Although animals with one targeted allele of Jnk1 (Jnk1��/2) had a body weight intermediate between that of wild-type and Jnk12/2 mice maintained on either diet, these differences did not reach statistical significance (Fig. 1e). We assessed whether these differences in weight gain were related to alterations in adiposity. Sections of adipose tissue from Jnk12/2 mice exhibited decreased adipocyte size relative to wild-type con- trols (Fig. 2a). This was not observed in Jnk22/2 adipose tissue. The fat pad weights of Jnk1��/��, Jnk1��/2 and Jnk12/2 mice were similar in the lean group at both subcutaneous and epididymal fat depots (data not shown). However, in the obese group, the average weight Figure 1 Total JNK activity and development of diet-induced obesity. a, Total JNK activity and protein concentrations in liver, muscle and adipose tissues of lean and obese mice. Obese HF, dietary model (high-fat) ob/ob, genetic model for leptin deficiency. In JNK immunoblots, JNK1 and JNK2 have relative molecular masses of 56,000���54,000 and 46,000���43,000, respectively. Lower panels show means ^ s.e.m. of the quantified and normalized activities. b���e, Development of diet-induced obesity in Jnk2 2/2 (b, c) and Jnk1 2/2 (d, e) mice. All mice were male, 16 weeks old and on C57Bl/6J background. In b and d, n �� 10 in each group. c, e, Means ^ s.e.m. of body weights of male mice. Asterisk, statistical significance (P , 0.05) in a two-tailed Student t-test comparing Jnk1 2/2 or Jnk2 2/2 mice with controls. letters to nature NATURE | VOL 420 | 21 NOVEMBER 2002 | www.nature.com/nature 333 �� 2002 Nature Publishing Group

of the subcutaneous fat depot was decreased by 33% in Jnk12/2 mice compared with wild-type controls (Fig. 2b). Surprisingly, the weight of the epididymal fat pad was higher in the obese Jnk12/2 group than in the wild-type controls, indicating a redistribution of adipose depots (Fig. 2b). No difference in fad pad weight was evident between Jnk22/2 and wild-type mice in either condition (data not shown). To investigate systemic alterations in adiposity, we next examined total body composition. These studies demon- strated significantly decreased total body adiposity in Jnk12/2 mice compared with controls (Fig. 2c). In contrast, the body composition of the Jnk22/2 group was indistinguishable from wild-type controls (data not shown). To address alternative causes for decreased body weight in Jnk12/ 2 mice, we compared lipid metabolism, food intake, intestinal lipid absorption and core body temperature of Jnk12/2 and Jnk1��/�� mice. No significant differences were observed in plasma triglycer- ide, cholesterol and FFA concentrations between genotypes (data not shown). Examination of faecal lipid content also did not reveal any differences, thus excluding changes in intestinal lipid absorp- tion (Fig. 2d). There was a small and statistically insignificant decrease in daily food intake (0.46 g d21) and increase (0.32 8C) in core body temperature in obese Jnk12/2 mice compared with wild- type mice (Supplementary Fig. 1a and b). Although we cannot rule out the possibility that these small changes might contribute to decreased weight gain, the results strongly suggest that the deficiency in JNK1 is associated primarily with decreased adipocyte size, decreased adiposity and adipose redistribution in the context of dietary obesity. Adipose tissue can have a substantial impact on systemic glucose homeostasis through production of bioactive molecules. We exam- ined serum concentrations of adipocyte-derived secreted proteins with postulated roles in obesity and insulin action12���15. ACRP30 (30-kDa adipocyte complement-related protein)/adiponectin con- centrations in the obese Jnk12/2 mice were significantly higher than in Jnk1��/�� controls (Fig. 2e). In contrast, the concentrations of resistin were lower in Jnk12/2 mice than in Jnk1��/�� animals (Fig. 2f). Because recent studies have indicated a role for adipo- nectin as a mediator of fatty-acid oxidation and hepatic insulin Figure 3 Metabolic effects of JNK1-deficiency. a���d, Examination of glucose homeostasis by fasting plasma glucose (a) and insulin (b) concentrations and insulin (c) and glucose (d) tolerance tests in lean and obese Jnk1 2/2 and control male mice at 16 weeks of age. Investigation of the dynamics of the responses to the tolerance tests were done by analysis of variance repeated-measures analysis (Statview 4.01, Abacus Concepts, Berkeley, CA, USA). e���g, Body weight (e) and blood glucose (f) and plasma insulin (g) in ob/ob-Jnk1 ��/�� and ob/ob-Jnk1 2/2 mice. Body weight and blood measurements for ob/ob mice were performed at 4 and 8 weeks of age and after a 6-h daytime food withdrawal. Asterisk, statistical significance (P , 0.001 in c and d P , 0.05 in e���g). WT, wild type. Figure 2 Adipose tissue morphology and adiposity in Jnk1 2/2 mice and wild-type controls. a, b, Histological sections of epididymal fat pads (original magnification ��50) (a) and subcutaneous (SC) and epididymal (EPI) fat pad weights (b) of 16-week-old male Jnk1 2/2 and Jnk1 ��/�� mice (n �� 3 in a, n �� 9 in b). c���f, Total body composition (c), faecal lipid content (d), serum adiponectin concentration (e) and resistin concentration (f) in Jnk1 2/2 and Jnk1 ��/�� mice. Representative immunoblots are shown in insets. Total carcass lipid analysis was performed25 to determine fat mass of individual mice (n �� 6 in each group). Asterisk, statistical significance (P , 0.05) in a two-tailed Student t-test comparing Jnk1 ��/�� and Jnk1 2/2 mice. letters to nature NATURE | VOL 420 | 21 NOVEMBER 2002 | www.nature.com/nature 334 �� 2002 Nature Publishing Group