Interaction of Tumor Necrosis Factor- - and Thiazolidinedione ...

0013-7227/04/$15.00/0 Printed in U.S.A.

Endocrinology 145(5):2214 –2220 Copyright © 2004 by The Endocrine Society doi: 10.1210/en.2003-1580

Interaction of Tumor Necrosis Factor-␣- and Thiazolidinedione-Regulated Pathways in Obesity KATHRYN E. WELLEN, K. TEOMAN UYSAL, SARAH WIESBROCK, QING YANG, HONG CHEN, ¨ KHAN S. HOTAMISLIGIL GO

AND

Division of Biological Sciences and Department of Genetics and Complex Diseases (K.E.W., K.T.U., S.W., G.S.H.), Harvard School of Public Health, Boston, Massachusetts 02115; and Millennium Pharmaceuticals (Q.Y., H.C.), Cambridge, Massachusetts 02142 Thiazolidinediones (TZDs) are potent insulin-sensitizing compounds and high-affinity ligands for the transcription factor peroxisomal proliferator-activated receptor ␥. The mechanism through which TZDs improve insulin sensitivity, however, is not clear. In this study, we asked whether the ability of TZD to suppress and antagonize TNF␣ is an underlying mechanism for its molecular and physiological effects, using obese (ob/ob) mice lacking TNF␣ function. We found that the lipid-lowering effects of TZD are completely independent of TNF␣ suppression, and the insulin-sensitizing effects of TZD are partially independent. TZD treatment improved insulin sensitivity in ob/ob mice both with and without functional TNF␣, albeit with different absolute potency. To characterize

the potential interdependency of TZD- and TNF␣-regulated pathways at the molecular level, we also performed four-way transcriptional profiling of white adipose tissue of TZD- and vehicle-treated ob/ob mice, with and without TNF␣ function. The majority of metabolic genes identified were regulated independent of the presence of TNF␣, whereas most effects on inflammatory mediators were dependent on TNF␣. This study demonstrates that the insulin-sensitizing action of TZD occurs partially through TNF-independent mechanisms, although a subset of the molecular effects of TZD treatment in adipose tissue depends on TNF␣. (Endocrinology 145: 2214 –2220, 2004)

T

to prove that regulation of TNF␣ or any of the other candidate factors is required for TZD action. Several lines of evidence suggest that suppression of TNF␣ may contribute to the insulin-sensitizing and other effects of these compounds. Treatment with TZD directly suppresses the activity of the TNF␣ promoter in cultured cells and reduces TNF␣ mRNA in the adipose tissue of obese mice (14, 15). Correspondingly, plasma levels of TNF␣ in obese rodents and humans are lowered by TZD treatment (16, 17). Moreover, TZD interferes with TNF␣ action. Pretreatment with TZD blocks the ability of TNF␣ to inhibit insulin receptor signaling and induce insulin resistance, both in cell culture and in vivo (4, 5, 18). The available data provide multiple lines of strong, but nevertheless indirect, evidence that the ability of TZD to block TNF␣ action might be critical to its ability to improve metabolism and insulin action. In this study, we therefore sought to address this question directly, using TNF␣ loss-of function mouse models. We compared metabolic responses and WAT gene expression in TZD-treated obese mice with and without functional TNF␣ (ob/ob-p55⫺/⫺p75⫺/⫺ and ob/ob-TNF␣⫺/⫺). This study demonstrates that TZD lowers plasma lipid levels in a completely TNF␣-independent manner and improves insulin sensitivity partially independent of TNF␣. The effects of TZD on metabolic gene expression appear to be for the most part independent of TNF␣. However, TZD does act through TNF␣ primarily to reduce production of inflammatory mediators in adipose tissue, an action which is likely to contribute to the ability of TZD to modulate various aspects of the metabolic syndrome, including insulin sensitivity.

HIAZOLIDINEDIONE (TZD) COMPOUNDS are potent insulin sensitizers and are currently used clinically to treat type 2 diabetes. TZDs were identified based on their antihyperglycemic activity, but they are also able to improve other abnormalities associated with type 2 diabetes, such as hyperlipidemia, atherosclerosis, hypertension, and chronic inflammation (1). Although it has been known for several years that these compounds are ligands of peroxisome proliferator-activated receptor ␥ (2), how they act to improve insulin sensitivity remains poorly understood. Many genes targeted by TZD have been proposed as contributing to metabolic actions of TZD, including several that are likely to mediate its effects on insulin sensitivity, such as adiponectin, leptin, Cbl-associated protein, IL-6, and TNF␣ (3– 8). One potential mechanism for TZD action involves its ability to suppress production and action of the inflammatory cytokine TNF␣. In cells, TNF␣ inhibits insulin signaling, at least in part by blocking insulin receptor tyrosine kinase activity and inducing serine phosphorylation of insulin receptor substrate-1 (9). In both obese mice and humans, TNF␣ is overexpressed in adipose tissue (10 –12). In genetic and dietary mouse models of obesity, null mutations in either the gene encoding TNF␣ or those encoding both of its receptors improve insulin sensitivity over obese mice with TNF␣ activity (13). To date, however, no direct evidence has been generated Abbreviations: BAT, Brown adipose tissue; CtsS, cathepsin S; FFA, free fatty acid; MME, macrophage metalloelastase; MMP, matrix metalloprotease; TNFR, TNF receptor; TZD, thiazolidinedione; WAT, white adipose tissue. Endocrinology is published monthly by The Endocrine Society (http:// www.endo-society.org), the foremost professional society serving the endocrine community.

2214

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity

Materials and Methods Animals Obese (ob/ob) mice deficient in each TNF receptor (TNFR) were generated by crossing mice with targeted null mutations at both TNFR1 and TNFR2 (p55⫺/⫺, p75⫺/⫺) or TNF␣ (TNF␣⫺/⫺) loci with Ob/ob mice (C57BL/6 background) to produce animals heterozygous at the TNFR1, TNFR2, and ob (p55⫹/⫺, p75⫹/⫺, Ob/ob) or TNF␣ and ob (TNF␣⫹/⫺, Ob/ob) loci. The resulting double or triple heterozygotes were then crossed with each other to produce OB/OB and ob/ob littermates with mutations at one or both TNFRs or TNF␣. This same cross also generated OB/OB and ob/ob wild-type littermates as controls (13). All mice are treated with the peroxisome proliferator-activated receptor ␥ agonist MCC-555 for 2 wk at 3 mg/kg䡠d dose orally between ages 10 and 12 wk after 2 wk of daily acclimation to oral gavage. All mice were males. Body weight was determined daily and blood samples were collected with 3-d intervals. At the end of the study, mice were killed, and tissues were dissected, weighed, and snap frozen in liquid nitrogen for future analyses. Each group consisted of at least eight mice. All animal experimentation was carried out in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals and was approved by an institutional review board.

Metabolic measurements Blood samples were collected after a 6-h fast every 3 d starting at 10 wk of age. Metabolic measurements were conducted as previously described (13, 19). Briefly, serum glucose concentrations were measured using glucoanalyzer blood glucose strips (MediSense, Abbott Laboratories, Abbott Park, IL). The triglyceride and free fatty acid levels in serum were determined using GPO Trinder (Sigma, St. Louis, MO) and NEFA C (Wako, Richmond, VA) assays, respectively. Insulin tolerance tests were performed on conscious mice after a 6-h fast by ip administration of human insulin (1 IU/kg; Eli Lilly, Indianapolis, IN) and measurement of blood glucose at 30, 60, 90, and 120 min.

Statistics For each animal, body weight was measured daily during the 15-d treatment period, and the mean of the daily weight gain were determined for each genotype and treatment group and plotted over the 15-d treatment period. Mean brown adipose tissue (BAT) and white adipose tissue (WAT) weights were determined for each group at the end of the experiment. Mean blood glucose, triglycerides, and free fatty acids (FFAs) for each group were determined at each indicated time point. For all means, sd and error were also determined. Data are expressed as mean ⫾ sem. Weight gain, metabolic measurements, and insulin tolerance tests were compared with ANOVA repeated measures and significance determined at the P ⱕ 0.05 level. Values for a given time of measurement between groups were also subjected to two-tailed Student’s t tests and significance determined at the P ⱕ 0.05 level.

RNA preparation and cDNA microarrays Upon killing of animals, tissues were dissected and immediately frozen in liquid nitrogen. Either RNAwiz (Ambion, Austin, TX) or Trizol (Life Technologies, Inc., Grand Island, NY) kits were used to isolate RNA samples from the sc and epididymal WAT of each of the four groups of mice. RNA samples from four mice were grouped for each experimental category and analyzed in triplicate. RNA was reverse transcribed (Superscript reverse transcriptase; Invitrogen, Carlsbad, CA) to obtain the oligo-dT30 primed, [33P]dCTP-labeled first-strand cDNA probe for microarray analysis. Hybridization experiments were performed on Millennium custom mouse DNA microarrays, which contained approximately 5000 annotated cDNAs and approximately 5000 expressed sequence tag clones. Duplicate filters per probe were used for hybridization as previously described (20). Dried filters were exposed on phosphor imaging plates (Fuji-Film, Tokyo, Japan), and median intensity plus sd for each probe in duplicate was calculated. An in-house developed self-organizing map analytical tool was used to analyze the profiling data by clustering genes into similar expression patterns (21). A gene was discarded if the coefficients of variation for its two relative expression intensities from duplicates was greater than 0.5.

Endocrinology, May 2004, 145(5):2214 –2220 2215

RNA preparation and Northern blot analysis of gene expression Denatured RNA (10 ␮g per lane) was separated on 1% agarose gels containing 3% formaldehyde and, after electrophoresis, transferred onto nylon membranes. ␣-32P-dCTP (NEN Life Science Products, Boston, MA)-labeled cDNA probes were hybridized to the membranes and blots exposed to Biomax films (Kodak, Rochester, NY) or imaged using a phosphor imager (Bio-Rad Laboratories, Hercules, CA).

Results Effects of TZD on body and adipose tissue weight in obese mice with and without functional TNF␣

To examine a potential interaction between TNF␣- and TZD-regulated pathways, obese (ob/ob) mice with and without functional TNFRs (ob/ob-p55⫺/⫺p75⫺/⫺) were treated for 15 d with either TZD or vehicle alone. Animals of both the ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺ genotypes gained similar amounts of total body weight during the course of the TZD treatment (change in weight for ob/ob-veh ⫽ 0.2 ⫾ 0.5 g; ob/ob-p55⫺/⫺p75⫺/⫺veh ⫽ ⫺0.4 ⫾ 0.6 g; ob/ob-TZD ⫽ 2.6 ⫾ 0.4 g; ob/ob-p55⫺/⫺p75⫺/⫺TZD ⫽ 4.1 ⫾ 0.4 g) (Fig. 1A). WAT and BAT weights of each of the four groups of mice were also determined. BAT increased significantly with TZD treatment in both genotypes, with the TNFR null mice exhibiting a greater gain in BAT weight (ob/ob-veh ⫽ 0.45 ⫾ 0.070 g; ob/ob-p55⫺/⫺p75⫺/⫺veh ⫽ 0.545 ⫾ 0.050 g; ob/obTZD ⫽ 0.865 ⫾ 0.114 g; ob/ob-p55⫺/⫺p75⫺/⫺TZD ⫽ 1.167 ⫾ 0.100 g). In contrast, ob/ob mice had more WAT than ob/obp55⫺/⫺p75⫺/⫺ mice under control conditions (2.962 ⫾ 0.125 g and 2.256 ⫾ 0.111 g, respectively) and gained additional WAT weight on TZD treatment (3.423 ⫾ 0.200 g and 2.200 ⫾ 0.12 g, respectively) (Fig. 1B). The same experimental design was also applied to ob/ob and ob/ob-TNF␣⫺/⫺ mice to independently confirm that any differences in the effects of TZD on the two genotypes were in fact due to loss of TNF function. A similar pattern of body and adipose tissue weight gain was also observed in these animals. Both ob/ob and ob/ob-TNF␣⫺/⫺ mice gained weight on TZD treatment (change in weight for ob/ob-veh ⫽ 0.9 ⫾ 0.7 g; ob/ob-TNF␣⫺/⫺veh ⫽ ⫺0.4 ⫾ 0.5 g; ob/ob-TZD ⫽ 3.9 ⫾ 0.6 g; ob/ob-TNF␣⫺/⫺TZD ⫽ 2.4 ⫾ 0.6 g) (Fig. 1C). Both genotypes gained BAT weight on TZD treatment, although the TNF␣⫺/⫺ mice experienced a greater increase in BAT weight (ob/ob-veh ⫽ 0.568 ⫾ 0.090 g; ob/ob-TNF␣⫺/⫺veh ⫽ 0.640 ⫾ 0.050 g; ob/ob-TZD ⫽ 0.860 ⫾ 0.114 g; ob/obTNF␣⫺/⫺TZD ⫽ 1.385 ⫾ 0.050 g). The ob/ob mice had higher WAT weights than the ob/ob-TNF␣⫺/⫺ mice under both control and TZD-treated conditions (ob/ob-veh ⫽ 3.79 ⫾ 0.115 g; ob/ob-TNF␣⫺/⫺veh ⫽ 2.92 ⫾ 0.211 g; ob/ob-TZD ⫽ 3.93 ⫾ 0.200 g; ob/ob-TNF␣⫺/⫺TZD ⫽ 3.25 ⫾ 0.120 g) (Fig. 1D). TZD effects on plasma lipids and glucose metabolism in obese mice lacking TNF␣ function

To examine the role of TNF␣ function in the effects of TZD treatment on lipid metabolism, we determined plasma triglyceride, FFA, and blood glucose levels during TZD treatment. Both ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺ mice had identical plasma levels of triglyceride (42.62 ⫾ 1.61and 45.64 ⫾ 2.29 mg/dl, respectively) and FFA (0.396 ⫾ 0.022 and 0.356 ⫾

2216

Endocrinology, May 2004, 145(5):2214 –2220

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity

FIG. 1. Body and adipose tissue weight of TZD-treated obese mice with and without TNF function. Ten-week-old male ob/ob and ob/obp55⫺/⫺p75⫺/⫺ or ob/ob-TNF␣⫺/⫺ mice were treated for 15 d with the TZD MCC555 at 3 mg/kg䡠d or with vehicle. Body weight was determined daily throughout the treatment (A and C). At the end of the treatment period, the animals were killed, and WAT and BAT were collected and weighed (B and D). White circles, ob/ob mice, vehicle treated; black circles, ob/ob-p55⫺/⫺p75⫺/⫺ (A) or ob/ob-TNF␣⫺/⫺ mice (C), vehicle treated; white squares, ob/ob mice, TZD treated; black squares, ob/ob-p55⫺/⫺p75⫺/⫺ (A) or ob/ob-TNF␣⫺/⫺ mice (C), TZD treated; black bars, ob/ob mice, vehicle treated; vertical stripes, ob/ob-p55⫺/⫺p75⫺/⫺ (B) or ob/ob-TNF␣⫺/⫺ mice (D), vehicle treated; white bars, ob/ob mice, TZD treated; horizontal stripes, ob/ob-p55⫺/⫺p75⫺/⫺ (B) or ob/ob-TNF␣⫺/⫺ mice (D), TZD treated. The data are presented as mean ⫾ SEM. Statistical significance is indicated by asterisks (*, P ⱕ 0.05; **, P ⱕ 0.005; and ***, P ⱕ 0.0005).

0.013 mm, respectively) when treated with vehicle, and both genotypes also exhibited similar decreases in plasma lipid on treatment with TZD for 13 d (triglyceride 25.88 ⫾ 1.74 and 24.79 ⫾ 2.34 mg/dl, respectively; FFA 0.261 ⫾ 0.013 and 0.295 ⫾ 0.018 mm, respectively) (Fig. 2A,B). In contrast, blood glucose was decreased by loss of TNF activity alone (ob/ob ⫽ 350.88 ⫾ 18.4 mg/dl; ob/ob-p55⫺/⫺p75⫺/⫺ ⫽ 314.57 ⫾ 12.05 mg/dl), although not to the extent as with TZD treatment. Blood glucose levels of ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺ mice further were lowered to similar levels on TZD treatment (284.2 ⫾ 22.14 and 276 ⫾ 15.28 mg/dl, respectively) (Fig. 3A). Similarly, when insulin tolerance tests were performed, loss of TNF function alone allowed significantly improved response to insulin (blood glucose at 120 min after insulin stimulation: ob/ob ⫽ 112.1 ⫾ 10.4 mg/dl e; ob/obp55⫺/⫺p75⫺/⫺ ⫽ 88.5 ⫾ 6.5 mg/dl), although TZD further improved the response in both genotypes (ob/ob ⫽ 40 ⫾ 4 mg/dl; ob/ob-p55⫺/⫺p75⫺/⫺ ⫽ 51 ⫾ 3.9 mg/dl) (Fig. 3B). Plasma lipid and glucose levels were also measured and insulin tolerance tests performed in ob/ob-TNF␣⫺/⫺ mice. The effects of loss of TNF function and TZD treatment in these mice were very similar to those in ob/obp55⫺/⫺p75⫺/⫺ mice (Figs. 2, C and D, and 3, C and D; FFA at d 13 of treatment, ob/ob-veh ⫽ 0.389 ⫾ 0.051 mm; ob/obTNF␣⫺/⫺veh ⫽ 0.395 ⫾ 0.085 mm; ob/ob-TZD ⫽ 0.292 ⫾ 0.036 mm; ob/ob-TNF␣⫺/⫺TZD ⫽ 0.318 ⫾ 0.023 mm; triglyceride at d 13, ob/ob-veh ⫽ 32.84 ⫾ 5.52 mg/dl; ob/obTNF␣⫺/⫺veh ⫽ 33.07 ⫾ 4.32 mg/dl; ob/ob-TZD ⫽ 22.89 ⫾ 2.87 mg/dl; ob/ob-TNF␣⫺/⫺TZD ⫽ 28.09 ⫾ 5.56 mg/dl; blood glucose at d 10, ob/ob-veh ⫽ 337 ⫾ 23 mg/dl; ob/obTNF␣⫺/⫺veh ⫽ 293 ⫾ 14 mg/dl; ob/ob-TZD ⫽ 276 ⫾ 18 mg/dl; ob/ob-TNF␣⫺/⫺TZD ⫽ 252 ⫾ 13 mg/dl; blood glucose after 120 min insulin stimulation, ob/ob-veh ⫽ 86.5 ⫾ 11.5 mg/dl; ob/ob-TNF␣⫺/⫺veh ⫽ 61 ⫾ 5.5 mg/dl; ob/ob-TZD ⫽ 49.3 ⫾ 13 mg/dl; ob/ob-TNF␣⫺/⫺TZD ⫽ 42.5 ⫾ 3 mg/dl). Because the total improvement in insulin sensitivity was of smaller magnitude in animals lacking functional TNF␣ due to their already improved insulin sensitivity, it is possible that suppression of TNF␣ could be required for the full effects of TZD. However, within this physiological setting, it

was not possible to generate further evidence to support or refute this possibility. TNF-dependent and -independent effects of TZD

We sought to capture TNF␣-dependent actions of TZD by analyzing differential regulation of gene expression by TZD in animals with and without TNF function. We performed four-way transcriptional profiling of the WAT of each of the four groups of mice: ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺, treated with TZD or vehicle. As expected, many genes were identified that were regulated by TZD treatment independent of genotype (Fig. 4A, supplementary table). In line with the regulation pattern of blood lipids, most lipid metabolism genes identified in this study, such as fatty acid transport protein 1, pyruvate carboxylase, phosphoenolpyruvate carboxykinase, and stearoyl coenzyme A desaturase 1, were regulated by TZD similarly in both genotypes. In fact, our results indicate that many components of the fatty acid and triglyceride synthetic pathways are up-regulated by TZD and that the action of TZD on these pathways is probably entirely independent of TNF␣ (Fig. 4A, supplementary table). Several genes regulated differentially by TZD in the two genotypes were also identified (Fig. 4B; supplementary table). By Northern blot analyses, we confirmed that several typical macrophage and lysosomal genes, including macrophage metalloelastase (MME), Mac-1, and cathepsin S (CtsS), were down-regulated by TZD only in mice with intact TNF␣ function. Expression of another extracellular matrix-related gene, secreted protein acidic and rich in cysteine, was downregulated by TZD only in the animals lacking functional TNF␣. In addition to these genes, our experimental results indicate that TZD treatment specifically suppresses a large number of inflammatory genes typical of macrophages in a TNF-dependent manner (supplementary table). We also examined expression of several genes, including fatty acid transport protein 1, pyruvate carboxylase, MME, CtsS, and Mac-1 in the ob/ob-TNF␣⫺/⫺ animals. These experiments

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity

FIG. 2. Plasma lipid levels are lowered by TZD in a TNF-independent manner. During the course of the TZD treatment, blood samples were drawn from ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺ (A and B) or ob/obTNF␣⫺/⫺ (C and D) mice every 3 d. Plasma FFA (A and C) and triglyceride (B and D) levels were measured. White circles, ob/ob mice, vehicle treated; black circles, ob/ob-p55⫺/⫺p75⫺/⫺ mice (A and B) or ob/ob-TNF␣⫺/⫺ mice (C and D), vehicle treated; white squares, ob/ob mice, TZD treated; black squares, ob/ob-p55⫺/⫺p75⫺/⫺ mice (A and B) or ob/ob-TNF␣⫺/⫺ mice (C and D), TZD treated. Data are shown as mean ⫾ SEM. Statistical significance is indicated by asterisks (*, P ⱕ 0.05 and **, P ⱕ 0.005).

independently confirmed the patterns of gene expression due to lack of TNF function (data not shown). Considering that both TNF␣ and TZD have dramatic ef-

Endocrinology, May 2004, 145(5):2214 –2220 2217

FIG. 3. Blood glucose levels and insulin tolerance are improved by both loss of TNF function and TZD treatment in obese mice. Blood samples were drawn every 3 d from each mouse and blood glucose levels measured (A and C). Insulin tolerance tests were performed at the end of the treatment period (B and D). Both blood glucose and insulin tolerance improved significantly in ob/ob-p55⫺/⫺p75⫺/⫺ (A and B) and ob/ob-TNF␣⫺/⫺ (C and D) mice in comparison with ob/ob mice. In both genotypes, TZD treatment further improved blood glucose and insulin tolerance. White circles, ob/ob mice, vehicle treated; black circles, ob/ob-p55⫺/⫺p75⫺/⫺ mice (A and B) or ob/obTNF␣⫺/⫺ mice (C and D), vehicle treated; white squares, ob/ob mice, TZD treated; black squares, ob/ob-p55⫺/⫺p75⫺/⫺ mice (A and B) or ob/ob-TNF␣⫺/⫺ mice (C and D), TZD treated. Data are presented as mean ⫾ SEM. Statistical significance is indicated by asterisks (* P ⬍ 0.05 and **, P ⱕ 0.005).

fects on metabolism, it might be expected that expression of at least some genes that are regulated by both pathways would also be affected by obesity. To test this possibility, we isolated RNA from the WAT of three genetic mouse models of obesity, ob/ob, db/db, and Ay, and their lean controls, and compared expression of several genes found to be regulated by TZD in a TNF-dependent fashion. Interestingly, genes down-regulated by TZD through suppression of TNF␣, in-

2218

Endocrinology, May 2004, 145(5):2214 –2220

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity

FIG. 4. TZD regulates gene expression through both TNF-dependent and -independent mechanisms. RNA was isolated from the WAT of ob/ob and ob/ob-p55⫺/⫺p75⫺/⫺ mice (A and B) or from three different obese mouse models and their wild-type controls (C) and used for Northern blot analysis to confirm expression patterns generated by the microarray analysis and examine expression in obesity. Many genes, including most of the metabolic genes identified in this study, were regulated by TZD similarly in both genotypes (A). Several identified genes, especially those typical of macrophages or involved in inflammation, were regulated by TZD in a TNF␣-dependent manner (B).

cluding MME, CtsS, and Mac-1, were found to also be highly up-regulated in obesity in white adipose tissue (Fig. 4C). Discussion

In this study, we demonstrate that the ability of TZD to improve insulin sensitivity and regulate metabolic gene expression is not dependent of TNF␣. Physiological experiments in TZD-treated ob/ob mice with and without TNF function clearly show that most or all of the antihyperlipidemic and at least part of the antihyperglycemic effects of TZD are independent of TNF␣. It is important to note, however, that, whereas the maximal antihyperglycemic action of TZD is similar in mice with and without TNF function, the antidiabetic effects of TZD are of greater overall magnitude in animals with TNF function. Thus, it remains possible that suppression of TNF␣ may contribute to the insulin-sensitizing action of TZD. Analysis of gene expression data indicates that the effects of TZD on transcription of genes involved in lipid and energy metabolism are primarily, if not entirely, independent of the ability of TZD to suppress TNF␣. In contrast, the ability of TZD to suppress inflammatory pathways in adipose tissue, particularly those involving extracellular matrix dynamics and cell adhesion, appears to be largely dependent on its ability to interact with the TNF␣ pathway. Although suppression of TNF␣ by TZD does not appear to be a primary requirement for insulin sensitization, the importance of reducing the chronic inflammation characteristic of obesity is critical. Increasingly, evidence is emerging that inflammation is a key player in the development of metabolic syndrome and that the extracellular environment is essential for adipocyte function. For example, in addition to TNF␣, several other inflammatory cytokines are elevated in obesity, including IL-6, IL-1␤, and IL-8. IL-6 in particular has been linked with insulin resistance. IL-6 is secreted from adipocytes, and production of IL-6 is correlated with obesity and degree of insulin resistance (22–24). Like TNF␣, IL-6 can induce insulin resistance in adipocytes, and it has been recently demonstrated that IL-6-induced insulin resistance can be reversed by TZD treatment (8, 25). Interestingly, IL-6 also seems to promote weight loss (26, 27). Both TNF␣ and IL-6 appear to play important roles in obesity-linked adipose tissue inflammation and adipocyte insulin resistance. As exemplified by the fact that IL-6 expression can be induced by

TNF␣ in adipocytes (22), it is likely that the roles of these and possibly other cytokines are interregulated and to some degree complementary. In this study, however, no changes in expression of inflammatory cytokines were detected during transcriptional profiling among any of the four conditions, and likewise, by Northern blot analyses we did not find regulation of IL-6 or IL-1␤ gene expression, suggesting that these genes are not likely to be mediating effects attributed here to TNF␣ (data not shown). In addition to antiinflammatory cytokines, several of the matrix metalloproteases (MMPs), including MME (MMP12), are regulated in dietary and genetic mouse models of obesity (Fig. 4C) (28, 29). In adipose tissue, proteases may be involved in remodeling of the extracellular matrix as obesity develops to create space for and vascularize the increased tissue mass. Evidence for the relationship between extracellular matrix remodeling and adipocyte function has been seen both in vitro and in vivo, as administration of MMP inhibitors impairs differentiation of cultured adipocytes and decreases weight gain in ob/ob mice and adipose tissue weight in mice fed a high-fat diet (29 –32). Additionally, these enzymes are enriched in atherosclerotic plaques and have been implicated in contributing to plaque instability (33, 34). Thus, it seems likely that MME and other proteases contribute to the development of both obesity and atherosclerosis. Mac-1, the counterreceptor for intercellular adhesion molecule-1 expressed on monocytes and macrophages, is also up-regulated in inflammation and may contribute to atherosclerosis. Mac-1 is also clearly elevated in at least three mouse models of obesity (Fig. 4C). Interestingly and apparently paradoxically, both mac-1⫺/⫺ and intercellular adhesion molecule-1⫺/⫺ mice develop obesity on standard diets, suggesting that these proteins may play a role in modulation of adipose tissue and may be protective against obesity (35). Perhaps TZD down-regulation of Mac-1, whereas possibly reducing inflammation and improving insulin sensitivity, may contribute to the gain in WAT weight observed during TZD treatment. To conclude, this study directly demonstrates for the first time that inhibition of TNF␣ is not an absolute requirement for insulin sensitization by TZD, although suppression of inflammatory gene expression by TZD appears to be mediated in large part through TNF-dependent mechanisms. These data suggest that TZD- and TNF␣-regulated pathways

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity

Endocrinology, May 2004, 145(5):2214 –2220 2219

This work was supported by Grant DK52539 from the National Institute of Diabetes and Digestive and Kidney Diseases (to G.S.H.). K.E.W. received support from the National Institute of Environmental Health Sciences (Training Grant T32 ES07155).

References

FIG. 5. Model for TNF-dependent and -independent action of TZD. TZD acts through both TNF-dependent and TNF-independent mechanisms to improve metabolic abnormalities. The ability of TZD to improve hyperlipidemia appears to be mostly, if not entirely, independent of TNF␣. However, inhibition of TNF␣ is probably quite important to the modulatory action of TZD on inflammatory and cell environment-related pathways. Suppression of TNF␣ may be a less important mechanism for direct improvement of insulin sensitivity. However, chronic low-grade inflammation contributes to the development of insulin resistance; thus, suppression of TNF␣ may contribute to improved insulin sensitivity through the reduced activity of connected inflammatory pathways.

functionally interact to modify the inflammatory status of adipose tissue. Several of these TNF- and TZD-regulated proteins have already been associated with atherosclerosis, in which inflammation also plays a crucial role (36). If these proteins are also relevant to insulin resistance, this would further support the role of inflammation as a critical link between obesity and insulin resistance (Fig. 5). In this scenario, as obesity develops, adipose tissue becomes chronically inflamed, producing and secreting inflammatory proteins, including cytokines such as TNF␣ and proteases such as MME. These inflammatory proteins, along with other factors such as aberrant lipid metabolism, contribute to insulin resistance and development of diabetes as well as other metabolic complications. Our study suggests that the primary effect of suppression of the TNF␣ pathway is to reduce inflammation, which is likely to have secondary effects on reversing insulin resistance. In addition, if factors downstream of TNF␣ are also activated through TNF-independent mechanisms in obesity, and TZD is able to interfere with such factors, as has been recently demonstrated for nuclear factor ␬-B (37), this study may underestimate the importance of this pathway to the effects of TZD. Future studies aimed at understanding the interaction of TZD-regulated pathways with both TNF-regulated factors and other pathways independent of TNF␣ should clarify these issues and yield further insight into the pathways linking inflammation, obesity, and insulin resistance. Acknowledgments Received November 21, 2003. Accepted January 27, 2004. Address all correspondence and requests for reprints to: Go¨ khan S. Hotamisligil, M.D., Ph.D., Harvard School of Public Health, 665 Huntington Avenue, Boston, Massachusetts 02115. E-mail: ghotamis@hsph. harvard.edu.

1. Willson TM, Lambert MH, Kliewer SA 2001 Peroxisome proliferator-activated receptor ␥ and metabolic disease. Annu Rev Biochem 70:341–367 2. Lehmann JM, Moore LB, Smith-Oliver TA, Wilkison WO, Willson TM, Kliewer SA 1995 An antidiabetic thiazolidinedione is a high affinity ligand for peroxisome proliferator-activated receptor ␥ (PPAR␥). J Biol Chem 270:12953– 12956 3. Maeda N, Takahashi M, Funahashi T, Kihara S, Nishizawa H, Kishida K, Nagaretani H, Matsuda M, Komuro R, Ouchi N, Kuriyama H, Hotta K, Nakamura T, Shimomura I, Matsuzawa Y 2001 PPAR␥ ligands increase expression and plasma concentrations of adiponectin, an adipose-derived protein. Diabetes 50:2094 –2099 4. Miles PDG, Romeo OM, Higo K, Cohen A, Rafaat K, Olefsky JM 1997 TNF-␣-induced insulin resistance in vivo and its prevention by troglitazone. Diabetes 46:1678 –1683 5. Peraldi P, Xu M, Spiegelman BM 1997 Thiazolidinediones block tumor necrosis factor-␣-induced inhibition of insulin signaling. J Clin Invest 100:1863– 1869 6. Nolan JJ, Olefsky JM., Nyce MR, Considine RV, Caro JF 1996 Effect of troglitazone on leptin production. Studies in vitro and in human subjects. Diabetes 45:1276 –1278 7. Ribon V, Johnson JH, Camp HS, Saltiel AS 1998 Thiazolidinediones and insulin resistance: peroxisome proliferator activated receptor ␥ activation stimulates expression of the CAP gene. Proc Natl Acad Sci USA 95:14751–14756 8. Laguthu C, Bastard J-P, Auclair M, Maachi M, Capeau J, Caron M 2003 Chronic interleukin-6 (IL-6) treatment increased IL-6 secretion and induced insulin resistance in adipocyte: prevention by rosiglitazone. Biochem Biophys Res Commun 311:372–379 9. Hotamisligil GS, Peraldi P, Budavari A, Ellis R, White MF, Spiegelman BM 1996 IRS-1-mediated inhibition of insulin receptor tyrosine kinase activity in TNF-␣- and obesity-induced insulin resistance. Science 271:665– 668 10. Hotamisligil GS, Shargill NS, Spiegelman BM 1993 Adipose expression of tumor necrosis factor-␣: direct role in obesity-linked insulin resistance. Science 259:87–90 11. Hotamisligil GS, Arner P, Caro JF, Atkinson RL, Spiegelman BM 1995 Increased adipose tissue expression of tumor necrosis factor-␣ in human obesity and insulin resistance. J Clin Invest 95:2409 –2415 12. Kern PA, Saghizadeh M, Ong JM, Bosch RJ, Deem R, Simsolo RB 1995 The expression of tumor necrosis factor in human adipose tissue. J Clin Invest 95:2111–2119 13. Uysal KT, Wiesbrock SM, Marino MW, Hotamisligil GS 1998 Protection from obesity-induced insulin resistance in mice lacking TNF-␣ function. Nature 389:610 – 614 14. Jiang C, Ting AT, Seed B 1998 PPAR-␥ agonists inhibit production of monocyte inflammatory cytokines. Nature 391:82– 86 15. Hofmann C, Lorenz K, Braithwaite SS, Colca JR, Palazuk BJ, Hotamisligil GS, Spiegelman BM 1994 Altered gene expression for tumor necrosis factor-␣ and its receptors during drug and dietary modulation of insulin resistance. Endocrinology 134:264 –270 16. Murase K, Odaka H, Suzuki M, Tayuki N, Ikeda H 1998 Pioglitzaone timedependently reduces tumour necrosis factor-␣ level in muscle and improves metabolic abnormalities in Wistar fatty rats. Diabetologia 41:257–264 17. Katsuki A, Sumida Y, Murata K, Furuta M, Araki-Sasaki R, Tsuchihashi K, Hori Y, Yano Y, Gabazza EC, Adachi Y 2000 Troglitazone reduces plasma levels of tumour necrosis factor-␣ in obese patients with type 2 diabetes. Diabetes Obes Metab 2:189 –191 18. Szalkowski D, White-Carrington S, Berger J, Zhang B 1995 Antidiabetic thiazolidinediones block the inhibitory effect of tumor necrosis factor-␣ on differentiation, insulin-stimulated glucose uptake, and gene expression in 3T3–L1 cells. Endocrinology 136:1474 –1481 19. Hirosumi J, Tuncman G, Chang L, Gorgun CZ, Uysal KT, Maeda K, Karin M, Hotamisligil GS 2002 A central role for JNK in obesity and insulin resistance. Nature 420:333–336 20. Herman MP, Sukhova GK, Libby P, Gerdes N, Tang N, Horton DB, Kilbride M, Breitbart RE, Chun M, Schonbeck U 2001 Expression of neutrophil collagenase (matrix metalloproteinase-8) in human atheroma: a novel collagenolytic pathway suggested by transcriptional profiling. Circulation 104:1878 – 1880 21. Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E, Lander ES, Golub TR 1999 Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc Natl Acad Sci USA 96:2907–2912 22. Rotter V, Nagaev I, Smith U 2003 Interleukin-6 (IL-6) induces insulin resistance in 3T3–L1 adipocytes and is, like IL-8 and tumor necrosis factor-␣,

2220

23. 24.

25. 26. 27. 28. 29.

Endocrinology, May 2004, 145(5):2214 –2220

overexpressed in human fat cells from insulin-resistant subjects. J Biol Chem 278:45777– 45784 Vozarova B, Weyer C, Hanson K, Tataranni PA, Bogardus C, Pratley RE 2001 Circulating interleukin-6 in relation to adiposity, insulin action, and insulin secretion. Obes Res 9:414 – 417 Bastard JP, Maachi M, Van Nhieu JT, Jardel C, Bruckert E, Grimaldi A, Robert JJ, Capeau J, Hainquoe B 2002 Adipose tissue IL-6 content correlates with resistance to insulin activation of glucose uptake both in vivo and in vitro. J Clin Endocrinol Metab 87:2084 –2089 Hotamisligil GS, Murray DL, Choy LN, Spiegelman BM 1994 Tumor necrosis factor ␣ inhibits signaling from the insulin receptor. Proc Natl Acad Sci USA 91:4854 – 4858 Wallenius V, Wallenius K, Ahren B, Rudling M, Carlsten H, Dickson SL, Ohlsson C, Jansson J-O 2002 Interleukin-6-deficient mice develop matureonset obesity. Nat Med 8:75–79 Wallenius K, Wallenius V, Sunter D, Dickson SL, Jansson JO 2002 Intracerebroventricular interleukin-6 treatment decreases body fat in rats. Biochem Biophys Res Commun 293:560 –565 Maqoui E, Munaut C, Colige A, Collen D, Lijnen HR 2002 Modulation of adipose tissue expression of murine matrix metalloproteinases and their tissue inhibitors with obesity. Diabetes 51:1093–1101 Chavey C, Mari B, Monthouel MN, Bonnafous S, Anglard P, Van Obberghen E, Tartare-Deckert S 2003 Matrix metalloproteinases are differentially expressed in adipose tissue during obesity and modulate adipocyte differentiation. J Biol Chem 278:11188 –11196

Wellen et al. • TNF␣- and TZD-Regulated Pathways in Obesity 30. Lijnen HR, Maquoi E, Hansen LB, Van Hoef B, Frederix L, Collen D 2001 Matrix metalloproteinase inhibition impairs adipose tissue development in mice. Arterioscler Thromb Vasc Biol 22:374 –379 31. Rupnick MA, Panigrahy D, Zhang C-Y, Dallabrida SM, Lowell BB, Langer R, Folkman MJ 2002 Adipose tissue mass can be regulated through the vasculature. Proc Natl Acad Sci USA 99:10730 –10735 32. Bouloumie A, Sengenes C, Portolan G, Galitzky J, Lafontan M 2001 Adipocyte produces matrix metalloproteases 2 and 9: involvement in adipose differentiation. Diabetes 50:2080 –2086 33. Curci JA, Liao S, Huffman MD, Shapiro SD, Thompson RW 1998 Expression and localization of macrophage elastase (matrix metalloproteinase-12) in abdominal aortic aneurysms. J Clin Invest 102:1900 –1910 34. Galis ZS, Khatri JJ 2002 Matrix metalloproteinases in vascular remodeling and atherogenesis: the good, the bad, and the ugly. Circ Res 90: 251–262 35. Dong ZM, Guttierrez-Ramos J-C, Coxon A, Mayadas TN, Wagner DD 1997 A new class of obesity genes encodes leukocyte adhesion receptors. Proc Natl Acad Sci USA 94:7526 –7530 36. Ross R 1999 Atherosclerosis—an inflammatory disease. N Engl J Med 340: 115–126 37. Ruan H, Powell HJ, Lodish HF 2003 Troglitazone antagonizes tumor necrosis factor-␣-induced reprogramming of adipocyte gene expression by inhibiting the transcriptional regulatory functions of NF-␬B. J Biol Chem 278:28181–28192

Endocrinology is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the endocrine community.