Chronic inflammation in fat plays a crucial role in the development of obesity-related insulin resistance

Mouse models.

Male mice of the genetically obese/diabetic models ob/ob, db/db, agouti, and tubby in the C57BL/6J strain, along with littermate controls, were purchased from The Jackson Laboratory (Bar Harbor, Maine, USA). These mice were fed standard chow (Farmer’s Exchange, Framingham, Massachusetts, USA). For diet-induced obese (DIO) studies, starting at 4–5 weeks of age, wild-type C57BL/6J mice from The Jackson Laboratory were put on diets containing 10%, 45%, or 60% kcal from fat (Research Diets Inc., New Brunswick, New Jersey, USA). For transcriptional profiling studies, mice with early-onset genetic obesity (ob/ob and db/db) were sacrificed at 15 weeks of age; mice with late-onset genetic obesity (agouti and tubby) were sacrificed at 25 weeks of age; mice with diet-induced obesity (45% kcal from fat) were sacrificed at 20 weeks of age after 16 weeks of high-fat diet. Animals used to track progression to obesity were sacrificed after 0, 3, 6, 8, 11, 16, and 26 weeks on high-fat diet (60% kcal from fat). Tissues and body fluids collected for this work were epididymal WAT, liver, skeletal muscle (gastrocnemius and quadriceps), lung, spleen, and blood.

For rosiglitazone treatment, 8-week-old male ob/ob mice from The Jackson Laboratory were acclimated for 1 week. Rosiglitazone or vehicle (sterile water) was orally gavaged once a day at a dose of 15 mg/kg for 28 consecutive days. At the end of the study, mice were sacrificed by CO₂ inhalation, and epididymal fat pads were excised for RNA extraction. All animal experiments were approved by the Institutional Animal Care and Use Committee of Millennium Pharmaceuticals Inc.

Tissue collection, total-RNA preparation, and quantitative PCR.

Mice were sacrificed by CO₂ inhalation. Tissues were collected and frozen immediately in liquid nitrogen. For total-RNA preparation, tissues were ground into fine powder in liquid nitrogen and homogenized in Trizol solution (Invitrogen, Carlsbad, California, USA). RNA was isolated according to the manufacturer’s instructions. For WAT, the surface oil layer was removed prior to chloroform extraction to ensure better RNA quality. TaqMan real-time quantitative PCR was performed and analyzed according to the manufacturer’s instructions (Applied Biosystems, Foster City, California, USA).

Isolation of adipocytes, stromal-vascular cells, and primary macrophages.

Isolation of adipocytes and stromal-vascular cells was performed as described previously (16). Five days before macrophage isolation, male C57BL/6J mice were injected with 2 ml of BBL thioglycolate culture medium (Becton Dickinson and Co., Sparks, Maryland, USA). Mice were sacrificed by CO₂ inhalation, and the peritoneal cavities were washed with cold sterile PBS to collect macrophages. After removal of red blood cells, macrophages were resuspended in RPMI containing 10% heat-inactivated FBS and plated into 10-cm dishes. After 3 hours of incubation, cells were washed three times with PBS and Trizol solution was added for RNA extraction.

Immunohistochemistry and in situ hybridization.

Fresh frozen sections of WAT were cut for in situ hybridization. Probes were generated by RT-PCR from mRNA isolated from WAT with gene-specific primers tailed with T3 (sense) and T7 (antisense) promoter sequences. The hybridization was performed as described previously (17).

At the time of sacrifice, portions of WAT were either flash frozen in liquid nitrogen or fixed in 4% paraformaldehyde or Bouin’s fix (Poly Scientific R&D Corp., Bay Shore, New York, USA). After dehydration with increasing concentrations of ethanol, tissues were incubated in xylenes and then embedded in paraffin at 60°C. Seven-micron sections were cut on a microtome and mounted on slides. Samples were dewaxed, rehydrated, stained with toluidine blue O, destained, and dehydrated. Mounting solution and coverslips were added.

Paraffin sections of WAT were also stained with anti-mouse F4/80 antibody (Serotec Inc., Raleigh, North Carolina, USA). Primary antibody was biotinylated. Prior to incubation with antibody, sections were blocked with biotin and avidin. Endogenous peroxidase activity was blocked by pretreatment with 0.06% H₂O₂ in methanol. Endogenous Fc receptor was blocked with 20% FCS in PBS. Sections were incubated in primary antibodies (1:100 dilution) for 1 hour and then washed with PBS/0.05% Tween-20. HRP-conjugated streptavidin (DAKO Corp., Carpinteria, California, USA) was added and incubated for 30 minutes, and then slides were washed with PBS/0.05% Tween-20. Finally, peroxidase substrate diaminobenzidine (Vector Laboratories Inc., Burlingame, California, USA) was added and incubated for 15 minutes. Slides were rinsed and counterstained with Mayer’s hematoxylin. Mounting solution and coverslips were added.

Isolated stromal-vascular cells were seeded in chamber slides (Nalge Nunc International, Naperville, Illinois, USA) at a concentration of 5 × 10⁵ per chamber in DMEM containing 20% FCS (HyClone Laboratories, Logan, Utah, USA) and 20 ng/ml recombinant mouse G-CSF (R&D Systems Inc., Minneapolis, Minnesota, USA) for overnight culture. To prepare for fixation, the cells were rinsed five times with PBS+ (PBS supplemented with 1 mM MgCl₂ and 0.1 mM CaCl₂) and fixed in PBS+ containing 3% paraformaldehyde for 30 minutes at room temperature. Cells were then incubated in PBS+ containing 50 mM NH₄Cl for 5 minutes at room temperature and rinsed three times with PBS+.

Immunohistochemistry on stromal-vascular cells was performed as described below. After fixing with paraformaldehyde, stromal-vascular cells were used for immunohistochemical staining with anti-mouse F4/80 antibody. Primary antibody was diluted 1:50 for the experiments. Alkaline phosphatase–conjugated secondary antibody (Serotec Inc.) was used (1:300 dilution). Levamisole was used to block endogenous alkaline phosphatase activity. Alkaline phosphatase substrate was purchased from Vector Laboratories Inc. Slides were counterstained with Mayer’s hematoxylin. Mounting solution and coverslips were added.

Oil red O was used to stain lipids in fixed stromal-vascular cells. Cells were washed three times in PBS+ and then once in 60% isopropanol. Oil red O was added, and the cells were incubated for 10 minutes at room temperature. Cells were washed once in 60% isopropanol and three times in PBS+. Slides were rinsed and counterstained with Mayer’s hematoxylin. Mounting solution and coverslips were added.

Expression levels of genes in inflammatory pathways are significantly upregulated in WAT of obese mice.

To study obesity and obesity-induced insulin resistance, we performed global transcriptional profiling studies with various tissues (WAT, brown adipose tissue, muscle, liver, stomach, hypothalamus, small intestine, and pancreas) taken from genetically obese mice, including ob/ob, db/db, tubby, agouti, and DIO mice. Notably, we found that many of the most significantly upregulated genes in WAT were not known to be involved in adipocyte biology; instead, they could be broadly categorized as macrophage- or inflammation-related genes. Of the genes upregulated more than twofold in at least four of these five models, 59% (50/85) could be counted as inflammation genes, as determined by their known functions. The remaining genes were involved in diverse molecular pathways, including fat storage, cholesterol metabolism, DNA modification, transcription, cell division, signal transduction, and unknown functions (see Supplemental Table 1 for details; http://www.jci.org/cgi/content/full/112/12/1821/DC1). These results were consistent with an earlier study on leptin-deficient ob/ob mice (18); however, with multiple models of genetic and diet-induced obesity, our data suggest that the inflammatory response is a general phenomenon of the obese state, independent of the availability of the leptin protein. We also noticed that this phenomenon was WAT-specific and was not observed in any other tissues we profiled.

To understand this apparent inflammatory response in WAT in detail and to explore its role in obesity and insulin resistance, we focused our follow-up studies on two genetic mouse models (ob/ob and db/db) and the diet-induced obesity model. We chose the following six macrophage or inflammation genes for further study: ADAM8, a disintegrin-like metalloproteinase strongly expressed in monocytic lineage (19); macrophage inflammatory protein-1α (MIP-1α), a gene derived from mononuclear cells and involved in the acute inflammatory state in the recruitment and activation of polymorphonuclear leukocytes (20); MCP-1, which is a member of the small inducible cytokines family and which plays a role in the recruitment of monocytes to sites of injury and infection (21); macrophage antigen-1 (MAC-1) (CD11b), an integrin found predominantly on monocytes, macrophages, neutrophils, and NK cells (22); F4/80, a classical macrophage-restricted surface glycoprotein (23); and CD68 (macrosialin), a heavily glycosylated transmembrane protein expressed specifically in macrophages and macrophage-related cells (24).

As shown in Figure 1a, the mRNA levels of these six genes were consistently and significantly upregulated in WAT of ob/ob, db/db, and DIO mice that had been on a high-fat diet for 16 weeks, confirming the microarray experiments. Regulation for three genes was further confirmed by Northern blot analysis using independent samples from the same experiment (Figure 1b). The fold increase for each gene among different obesity models was not directly comparable. DIO mice were studied separately from the other two genetic models. The DIO mice used here were 20 weeks old, fed on a high-fat diet (60% kcal from fat) for 16 weeks, with an average body weight of about 60 g. Each obese model was compared with its respective lean control. The absolute expression levels of the genes of interest were lower in the lean controls of the DIO model than in the wild-type controls of the genetic models because of the difference in diets (10% low-fat diet for the former vs. 16% regular chow for the latter). This resulted in an exaggerated fold change for the DIO model. Excluding the diet factor and experimental variations, however, it was still apparent that different genes were upregulated differently among obese models even when the expression levels were normalized, as shown in Figure 1b. For example, ADAM8 was most dramatically upregulated in DIO mice fed on 60% high-fat diet. In short, upregulation of inflammation genes is observed in WAT of genetic and DIO mouse models, although the expression levels for each gene appear to be regulated differently from one model to another. Both diet and lack of leptin-signaling pathways might have played a role here. Consistent with this possibility, regulation of inflammation genes was less dramatic with DIO mice fed with a 45% fat diet (Supplemental Table 1, http://www.jci.org/cgi/content/full/112/12/1821/DC1). The absence of leptin in genetically obese mice might even be protective, since leptin has been reported to increase cholesterol ester synthesis in cultured macrophages (25).

To determine whether the upregulation of these genes occurs prior to the development of systematic insulin resistance, which is characterized by hyperinsulinemia, we tracked the expression levels of these genes in WAT of mice with high-fat diet–induced obesity at multiple time points for 26 weeks. The body weight increased steadily over this period, as did the fasting blood glucose level, although the latter remained within the normal range (<120 mg/dl) until sometime after 16 weeks (Figure 2a). Meanwhile, we observed an increase in expression of some of these inflammation genes as early as 3 weeks on high-fat diet (Figure 2b). Around 16 weeks on high-fat diet, a much more dramatic upregulation of these transcripts occurred, which correlated closely with a marked increase in fasting blood insulin levels (Figure 2). It appears that the adipose inflammatory response increases with an increase of adiposity, prior to the increase of fasting insulin level, but intensifies at the onset of hyperinsulinemia.

On the other hand, in muscle and liver, two other important tissues for insulin action, the mRNA expression of these genes was barely detectable and essentially unchanged in these obese mice (data not shown). To find out whether similar inflammatory activities are present in other tissues rich in macrophages, we compared expression of ADAM8, MIP-1α, MCP-1, MAC-1, F4/80, and CD68 in lung, spleen, and WAT in ob/ob and db/db mice with that in their lean littermate controls. Overall, there was little change in these genes in lung and spleen in the obese state (Figure 3). In the progression study, after 26 weeks on high-fat diet, a significant upregulation of CD68 was observed in the liver, although the absolute expression level was much lower compared with that in fat.

The expression of inflammation genes is restricted to, or enriched in, macrophages.

To sort out the cell types in which these genes were expressed, we separated cells in WAT into two fractions: adipocytes and stromal-vascular cells. As shown in Figure 4, TaqMan analysis revealed that the inflammation genes under investigation were predominantly expressed in the stromal-vascular fraction, where macrophages might reside. These genes were expressed at much lower levels in the adipocyte fraction. Notably, TNF-α and IKKβ, two inflammation genes linked to obesity-induced insulin resistance by loss-of-function studies, were also more abundant in the stromal-vascular fraction. To control for the quality of RNA from adipocytes, leptin, an adipocyte-specific hormone, was used as a marker to show that the fractionation was done properly. The expression pattern of the inflammation genes was further examined in many tissues including adipose and primary macrophages isolated from the peritoneal cavity of normal mice. Consistent with existing evidence, these genes were expressed at the highest levels in the primary-macrophage population, followed by WAT, demonstrating that WAT is normally enriched in macrophages (data not shown).

Since the stromal-vascular fraction from fat contains several cell types including preadipocytes, it was important to determine which cells within the stromal-vascular fraction significantly express these inflammation genes. To address this question, two cultured preadipocyte cell lines, 3T3-L1 and 3T3-F442A, were stimulated to undergo adipogenesis. TaqMan analysis showed that ADAM8, MIP-1α, F4/80, and MAC-1 were not detectable in either preadipocytes or adipocytes. CD68 had only a very low level of expression in adipocytes, while MCP-1 showed moderate expression in preadipocytes (data not shown).

Recent evidence has indicated that preadipocytes have the potential to “transdifferentiate” into macrophages (26). To explore the possibility that preadipocytes could express these inflammation and macrophage genes in an inflammatory environment, we used cytokines upregulated in the obese state, TNF-α, MIP-1α, and MCP-1, as well as general inflammatory stimuli (LPS and PMA) to stimulate 3T3-L1 preadipocytes for 16 hours. While treatment with these inflammatory stimuli did not increase the expression of macrophage-marker genes (MAC-1, F4/80, and CD68) in 3T3-L1 preadipocytes, TNF-α did significantly stimulate the expression of ADAM8, MIP-1α, and MCP-1 in 3T3-L1 preadipocytes (Supplemental Figure 1, http://www.jci.org/cgi/content/full/112/12/1821/DC1). This suggests that obesity-related inflammatory stimuli can trigger preadipocytes to increase expression of some, but not all, inflammation genes under certain conditions.

Morphological changes of WAT in obesity.

The apparent upregulation of gene expression could be a consequence of both increased transcription in existing macrophages and increased macrophage infiltration into WAT of obese mice. To determine the source of this signal, adipose tissue sections from wild-type and ob/ob mice were compared. The morphological differences were striking. Besides the well-documented size difference of adipocytes from wild-type and ob/ob mice, there were clusters of what appeared to be small, nucleated cells in the interstitial spaces between adipocytes in ob/ob fat tissues, in contrast to the near absence of such cells in the control-mouse sections (Figure 5a). As the obese mice aged (5 months), these cell clusters became more prevalent and adipocytes showed early features of lipolysis manifested by multifocal cell shrinkage (Figure 5b). In situ hybridization and immunohistochemical staining using F4/80 RNA probes and antibody illustrated that F4/80 expression was restricted to these clusters of small, nucleated cells (Figure 5, c and d). Similar results were seen with other inflammation genes as well, confirming the findings with F4/80 (data not shown). Lack of costaining with platelet-endothelial cell adhesion molecule (PECAM), a vascular endothelium marker, excluded involvement of endothelial cells (data not shown).

To further determine that these cells were not preadipocytes, stromal-vascular fractions from wild-type and ob/ob adipose tissues were seeded on slides and stained with anti–F4/80 antibody. Most of the stromal cells from wild-type WAT appeared to be fibroblasts, possibly preadipocytes, and were F4/80 negative. In the stromal-vascular fraction from ob/ob mice, on the other hand, there were numerous F4/80–positive cells. Many of these cells contained multiple nuclei, reminiscent of giant cells (Figure 5e). Giant cells are formed by the fusion of multiple macrophages and are seen in instances of chronic inflammation and granuloma (27). In mice at 5 months of age, it was determined (by positive F4/80 staining) that 2% of wild-type and 33% of ob/ob stromal-vascular cells were macrophages. As shown in Figure 5f, F4/80–positive cells also contained many oil red O–staining vesicles, indicating intracytoplasmic lipid accumulation. This observation is consistent with histiocytic (phagocytic) activity. We also stained the stromal cells for the T cell–specific marker CD3, the B cell–specific marker CD72, and the granulocyte-specific marker Ly6G (data not shown). Of the four cell types tested, only macrophages were observed in WAT. From this, we conclude that there is significant macrophage accumulation in WAT during the development of obesity.

Rosiglitazone downregulates the mRNA level of the inflammation genes.

PPARγ is a member of the nuclear hormone receptor family. It is the decisive transcription factor regulating adipocyte differentiation (28, 29). PPARγ agonists, including troglitazone, pioglitazone, and rosiglitazone, are a new and highly efficacious class of insulin-sensitizing drugs known as thiazolidinediones (TZDs). To explore the relationship of insulin resistance and inflammation in adipose tissue from a different perspective, we treated ob/ob mice with rosiglitazone, a PPARγ agonist. As shown in Figure 6, ADAM8, MAC-1, F4/80, and CD68 showed statistically significant reduction in expression levels after the treatment. MIP-1α and MCP-1 expression also trended down but did not reach statistical significance.