MGL1 promotes adipose tissue inflammation and insulin resistance by regulating 7/4^hi monocytes in obesity

We have previously shown that the resident population of type 2 ATMs expresses high levels of MGL1 (Lumeng et al., 2008). We hypothesized that MGL1 might participate in the recruitment of these ATMs to fat. To test this, we examined ATMs from Mgl1^−/− mice fed a normal chow diet (ND) by immunofluorescence. ATMs were present in Mgl1^−/− mice with a morphology and density that was identical to WT controls (Fig. 1 A). ATMs in Mgl1^−/− and WT mice expressed CX₃CR1 and the M2 macrophage marker, CD206/mannose receptor. Quantitation of ATMs in epididymal fat pads from ND mice by flow cytometry showed no difference in the quantity of F4/80⁺ CD206⁺ ATMs between Mgl1^−/− mice and WT controls (Fig. 1 B). We also found that type 2 ATMs in lean mice from both genotypes express ST2 (Fig. 1 C), a member of the TLR-IL receptor superfamily that negatively regulates TLR4 signaling (Brint et al., 2004). Similar to MGL1, ST2 expression is restricted to type 2 ATMs in interstitial spaces of adipose tissue from obese mice and is not expressed in inflammatory type 1 ATMs in CLS (Isolectin⁺; Fig. 1 D). MGL1 staining confirmed the lack of MGL1 expression in Mgl1^−/− mice, and demonstrated specificity of the antibody for MGL1 (Fig. 1 C). Overall, these results demonstrate that MGL1 is not required for the proper localization and antiinflammatory characteristics of resident type 2 ATMs.

Although MGL1 does not influence the phenotype of resident ATMs in lean mice, we hypothesized that Mgl1 deficiency might alter inflammation and insulin resistance in the context of DIO. Therefore, we placed Mgl1^−/− and control mice on a HFD for 18 wk. Mgl1^−/− mice gained weight at a similar rate as WT controls until the last weeks of the HFD feeding when the weight of Mgl1^−/− mice was slightly reduced compared with controls (Fig. 2 A). There were no differences in weight between ND-fed Mgl1^−/− and Mgl1^+/+ mice (Fig. 2 B). No significant difference in food intake was observed between Mgl1^−/− mice and control mice that could explain the slight weight difference (Fig. 2 C). Respiratory quotient was also similar between the two genotypes (Figs. 2 D). A small, but statistically significant increase in oxygen consumption was seen in HFD-fed Mgl1^−/− mice, but this lost significance when corrected for lean body mass (Fig. 2 E).

Body composition analysis showed that HFD-fed Mgl1^−/− mice had less fat mass compared with controls, and this primarily accounted for the difference in weight (Fig. 2 F). Examination of the tissue weights showed that Mgl1^−/− mice had less subcutaneous white adipose tissue (inguinal fat pad) and lower liver weight. When expressed as a percentage of total body weight, liver weight was significantly decreased in Mgl1^−/− mice. Surprisingly, obese Mgl1^−/− mice had a different distribution of fat, with relatively more visceral white adipose tissue (epididymal fat pad) than HFD-fed control mice (Fig. 2 G). Overall, Mgl1^−/− mice showed slight protection from long-term HFD feeding, but had a greater accumulation of visceral fat compared with controls.

Glucose metabolism was evaluated in ND and HFD-fed Mgl1^−/− and WT mice. HFD-fed mice from both genotypes demonstrated fasting hyperglycemia compared with ND-fed mice, and there were no significant differences between genotypes (Fig. 3 A). Fasting insulin levels were elevated in HFD-fed mice compared with lean controls with a trend toward increased insulin levels in knockout versus WT controls (Fig. 3 B; P = 0.2). Glucose tolerance tests showed no differences between lean ND-fed Mgl1^−/− and controls (Fig. 3 C). However, obese Mgl1^−/− mice had significantly improved glucose tolerance compared with HFD-fed controls. Insulin tolerance tests demonstrated that HFD Mgl1^−/− mice were significantly more insulin sensitive than obese Mgl1^+/+ mice with a profile similar to ND mice (Fig. 3 D). These data show that Mgl1^−/− deficiency leads to improved glucose tolerance and insulin sensitivity with DIO.

The protection from insulin resistance in Mgl1^−/− mice occurred in spite of the fact that Mgl1^−/− mice have proportionately more visceral fat. Examination of visceral adipose tissue showed that obese Mgl1^−/− had more intact adipocytes and fewer macrophage-containing CLS compared with HFD-fed Mgl1^+/+ controls (Fig. 3 E). Analysis of adipocyte size demonstrated that obese Mgl1^−/− mice had significantly larger adipocytes compared with obese controls (Fig. 3 F). Overall, the data showed that HFD-fed Mgl1^−/− mice had more visceral fat than controls correlating with more extensive adipocyte hypertrophy and less ATM accumulation.

These observations suggested that Mgl1^−/− mice are more efficient at storing lipid in adipose tissue with HFD. Analysis of serum-free fatty acid (FFA) levels confirmed that Mgl1^−/− mice had lower circulating levels of FFAs than obese controls (Fig. 3 G). Furthermore, liver histology demonstrated that Mgl1^−/− mice were protected from hepatic steatosis induced by HFD feeding, a finding that likely explains the difference in liver mass between the genotypes (Fig. 3 H). Overall, this demonstrates that obese Mgl1^−/− mice have more efficient lipid storage in adipose tissue compared with controls, thus providing protection from fatty liver.

The adipose tissue histology suggested that alterations in ATM-induced inflammation might explain the protection from insulin resistance in obese Mgl1^−/− mice. Examination of inflammatory gene expression in epididymal adipose tissue showed that obese Mgl1^−/− mice had significantly more Il10 expression compared with controls (Fig. 4 A). Il10 expression was not altered in the inguinal fat pads from the Mgl1^−/− mice, suggesting that Mgl1 deficiency primarily effects visceral fat inflammation. No other significant differences in inflammatory gene expression were seen between genotypes, although there was a trend toward a decrease in Itgax/Cd11c expression and an increase in Cx3cr1 expression in Mgl1^−/− mice. Mgl2 expression was not changed in adipose tissue with Mgl1 deficiency, demonstrating that this orthologue is not up-regulated to compensate for the lack of MGL1. No significant differences in gene expression were seen in inguinal (subcutaneous) fat (Fig. 4 B).

Because type 2 ATMs are a source of IL-10 in adipose tissue (Lumeng et al., 2008), we hypothesized that obese Mgl1^−/− mice may have an altered ratio of type 1 to type 2 ATMs. To test this, we evaluated ATM content in obese mice by flow cytometry using markers of type 1 (CD11c) and type 2 (CD206) ATMs (Fig. 4 C; Nguyen et al., 2007). Although the percentage of F4/80⁺ CD11b⁺ ATMs in Mgl1^−/− mice was similar to controls, the total number of ATMs isolated from epididymal fat pads was slightly decreased. The content of CD11c⁺ ATMs in obese Mgl1^−/− mice was reduced to ∼40% of WT levels that persisted with normalization to fat weight. This resulted in an overall reduction in the ratio of type 1 to type 2 (CD11c⁺/CD206⁺) ATMs in HFD-fed Mgl1^−/− mice (Fig. 4 D). CD206 staining was not different between ATMs from Mgl1^−/− mice and controls, demonstrating that the expression of this lectin is also not altered in the absence of MGL1.

Immunofluorescence microscopy of epididymal fat pads supported the flow cytometry data. There was a significant decrease in the number of type 1 ATM clusters in fat from Mgl1^−/− mice compared with Mgl1^+/+ controls (Fig. 4 E). This protection from type 1 ATM accumulation was more striking with long-term HFD exposure (26 wk) as Mgl1^−/− mice had very few Mac2⁺ CLS compared with controls (Fig. 4 F). Composite images demonstrated that CLS were not uniformly distributed in epididymal fat pads in HFD-fed WT mice.

To evaluate the effects of MGL1 deficiency on systemic inflammation in obesity, we examined serum cytokine levels in the mice (Fig. 4 G). Inflammatory cytokines RANTES/CCL5 and TNF were increased with HFD in control mice. However, the levels of these chemokines were decreased in HFD Mgl1^−/− mice compared with controls. IL-10 was similarly decreased in Mgl1^−/− mice consistent with previous observations (Saba et al., 2009). CCL2/MCP-1 levels were increased in HFD-fed Mgl1^−/− mice compared with controls. Overall, our data suggest that the improved glucose and insulin tolerance in Mgl1^−/− mice is related to a decrease in the accumulation of type 1 inflammatory ATMs in visceral adipose tissue, less adipose tissue inflammation, less systemic inflammation, and more intact adipocytes, which led to less hepatic steatosis.

The prior results were unexpected, as it appeared that MGL1 was required for the accumulation of type 1 ATMs with low MGL1 expression. Additionally, despite elevated levels of CCL2 in obese Mgl1^−/− mice, ATM trafficking to fat was not increased and was, in fact, suppressed. We hypothesized that these observations might relate to an undescribed role for MGL1 in monocytes in the bone marrow or blood that give rise to ATMs. To evaluate this, we examined the quantity of 7/4^hi and 7/4^mid monocytes, monocyte subclasses with different inflammatory functions, in the blood of Mgl1^−/− and Mgl1^+/+ mice on different diets (Fig. 5 A). In ND mice, Mgl1^−/− mice had similar amounts of 7/4^hi monocytes compared with controls, but had an increased percentage of 7/4^mid monocytes in the blood. Similar to other studies (Tsou et al., 2007), we observed an increase in circulating 7/4^hi monocytes in control mice with HFD. However, in Mgl1^−/− mice, DIO did not induce an increase in 7/4^hi inflammatory monocytes. 7/4^mid monocytes were not significantly different between HFD-fed Mgl1^−/− and control mice. Examination of 7/4^hi and 7/4^mid monocytes in the bone marrow demonstrated no significant differences in the monocyte subclasses between Mgl1^−/− and control mice (Fig. 5 B).

These results suggest that MGL1 plays a role in the induction and/or maintenance of circulating 7/4^hi monocytes in response to DIO that ultimately attenuates the accumulation of type 1 inflammatory ATMs. Consistent with this, we found that MGL1 was expressed only on 7/4^hi monocytes and not on 7/4^mid monocytes in lean animals (Fig. 5, C and D). HFD exposure induced low MGL1 expression in 7/4^mid monocytes.

The identification of MGL1 expression on monocytes suggests that MGL1 deficiency alters the properties of circulating monocytes. This would provide a mechanism to explain the decrease in 7/4^hi monocytes in the blood, but not the bone marrow, of knockout mice. It can also explain the attenuation of ATM accumulation in visceral fat despite adipocyte hypertrophy. To assess this, we performed adoptive transfer experiments to compare the circulating half-life of monocytes from wild-type and Mgl1^−/− mice. CFSE-labeled monocytes from the two genotypes were injected IV into C57BL/6 recipients and circulating CD11b⁺ monocytes were quantified in the blood 24 and 72 h after injection. This demonstrated that fewer monocytes were recovered from the circulation after transfer of Mgl1^−/− cells compared with control cells at both time points (Fig. 6 A). Assessment of transferred cells in the spleen at 72 h showed equal numbers of monocytes and did not suggest that sequestration in the spleen explained the peripheral monocyte phenotype (unpublished data). This suggests that Mgl1 deficiency alters the intrinsic properties of monocytes that helps maintain them in the circulation.

To further assess the importance of MGL1 in promoting monocyte trafficking to inflammatory sites, we looked at the accumulation of thioglycollate (TG)-elicited peritoneal macrophages (PMs) in Mgl1^−/− mice. TGPMs had increased MGL1 expression compared with resident PMs, consistent with other reports of MGL1 induction upon stimulation of migration (Fig. 6 B; Dupasquier et al., 2006). Analysis of the accumulation of PMs after TG injection showed that Mgl1^−/− mice had fewer cells recruited 18 h after injection compared with WT controls (Fig. 6 C). At the 72 h time point, there was no significant difference between the two genotypes. Ccr2^−/− controls showed little PM recruitment after TG injection. In addition, TG-elicited PMs from Mgl1^−/− mice had decreased TNF expression with LPS treatment compared with controls, suggesting that the recruited Mgl1^−/− PM have less inflammatory potential (Fig. 6 D). Overall, these results support a role for MGL1 in generating the maximal recruitment of inflammatory mononuclear phagocytes to sites of inflammation.

Another property of monocytes that might explain the protection from ATM infiltration is altered interactions between adipocytes and MGL1-deficient monocytes. To assess this hypothesis, we compared the adhesion of monocytes from control and Mgl1^−/− mice to 3T3-L1 adipocytes, a well established adipocyte cell line. In this assay, the adhesion of Mgl1^−/− monocytes was significantly decreased compared with wild type (Fig. 7 A). This difference was not seen when monocytes were incubated with an endothelial cell line (bEnd.3) in the presence or absence of TNF (Fig. 7 B). This suggests that ligands for MGL1 are present on adipocytes and may regulate the trafficking and/or retention of macrophages or monocytes in adipose tissue.

MGL1 binds Lewis X trisaccharides with high-affinity and specificity (Singh et al., 2009). This led us to hypothesize that the DIO would induce or alter the expression of Lewis X motifs in adipose tissue that might direct the trafficking of MGL1⁺ 7/4^hi monocytes. Lewis X was not detected in visceral adipose tissue from lean mice (Fig. 7 C). However, in obese mice, Lewis X was detected as highly concentrated in the isolectin⁺ CLS (Fig. 7 D). This indicates that MGL1 binding targets are induced in CLS with DIO, which may play a role in 7/4^hi monocyte trafficking to these regions.

Our study reveals a novel and unexpected function for the murine scavenger receptor MGL1 in the regulation of the inflammatory response to obesity. Because MGL1 is highly expressed in type 2 ATMs that secrete IL-10 and have low inflammatory activity, our initial hypothesis was that Mgl1 knockout mice would have a higher ratio of type 1 ATMs to type 2 ATMs, and thus more inflammation. Surprisingly, we found that Mgl1^−/− mice are protected from HFD-induced insulin resistance and steatosis, which correlates with attenuation of inflammatory ATM accumulation in visceral fat despite having larger adipocytes. The mechanism behind this phenotype is related to the observation that MGL1 regulates the properties of circulating monocytes that ultimately impact the appearance of ATMs in fat. MGL1 is expressed on inflammatory 7/4^hi monocytes in the blood, a suspected precursor of type 1 ATMs, and these cells are decreased in the circulation of obese animals. The absence of MGL1 alters the properties of these monocytes decreasing their survival in the circulation. MGL1 also participates in mediating the adhesion between monocytes and adipocytes, and we have identified Lewis X binding sites for MGL1 that are increased in adipose tissue with DIO.

These observations demonstrate that MGL1 is among a growing number of leukocyte receptors and chemokines that regulate the appearance and stability of monocytes in the circulation. CX₃CR1 regulates 7/4^mid Gr-1^lo monocyte numbers in mice by enhancing their survival in the circulation (Landsman et al., 2009). CCR2 is now recognized to be critical for the migration of monocytes from the bone marrow into the circulation via interactions with CCL2/MCP-1 and CCL7/MCP-3 (Serbina and Pamer, 2006; Tsou et al., 2007). Ccr2^−/− mice have a profound loss of 7/4^hi Ly6c⁺ monocytes with preservation of circulating 7/4^mid monocytes in the blood (Tsou et al., 2007). This translates into a lack of accumulation of type 1 ATMs in obese Ccr2^−/− mice, the normal recruitment of type 2 ATMs, and protection from obesity-induced insulin resistance (Weisberg et al., 2006; Lumeng et al., 2007b, 2008).

Both obese Mgl1^−/− mice and Ccr2^−/− mice demonstrate that attenuation of inflammatory 7/4^hi Ly6c⁺ monocytes can block ATM accumulation in adipose tissue, decrease obesity-induced inflammation in fat, and improve glucose tolerance. These observations provide added support to a model where differential trafficking of monocyte subpopulations generates the increased ratio of type 1 to type 2 and the “phenotypic switch” to a proinflammatory environment with obesity. Our results suggest that a critical nexus of interaction between obesity and inflammation may reside at the level of circulating monocytes, and support the results of other studies showing activation of peripheral monocytes in obesity (Ghanim et al., 2004; Degasperi et al., 2009). This perspective suggests that the elevations in circulating CCL2/MCP-1 seen in obesity and type 2 diabetes (Sartipy and Loskutoff, 2003; Takahashi et al., 2003) may in fact play a primary role in promoting egress of inflammatory monocytes from the bone marrow into the circulation, rather than just the recruitment of these cells from the circulation into adipose tissue. Redundant signals from CCL2 and CCL7 to promote the migration of CCR2⁺ monocytes from the bone marrow could explain why examination of ATMs in obese Ccl2^−/− mice has led to variable results (Kanda et al., 2006; Inouye et al., 2007; Kirk et al., 2008).

Evaluation of glucose metabolism in obese Mgl1^−/− mice demonstrates a significant protection from insulin resistance and type 1 ATM accumulation that is out of proportion to the modest difference in body weight observed. It is possible that differences in glucose metabolism observed are caused by the slight weight loss seen in the Mgl1^−/− mice. However, Mgl1^−/− mice actually have more visceral adipose tissue than controls with DIO, which is partially caused by larger adipocytes. The unique dissociation between adiposity and glucose intolerance is associated with a profound lack of ATM infiltration in Mgl1^−/− mice, further supporting a primary role for ATMs in the regulation of glucose metabolism. The deficiency of type 1 ATMs in the knockout mice results in more intact adipocytes, improved triglyceride storage in fat, decreased circulating FFA, and protection from hepatic steatosis.

These observations emphasize a property of adipose tissue that is still poorly understood: how visceral adipose tissue expansion may actually be metabolically beneficial in certain contexts. This agrees with recent observations that overexpression of adiponectin in leptin-deficient animals led to adipose tissue expansion without inflammation and maintained insulin sensitivity despite profound obesity (Kim et al., 2007).

Composite images (Fig. 4 F) demonstrate that the ATM accumulation with DIO is not uniform as CLS are distributed in patches. This regional variation in ATM accumulation may explain the variation in RT-PCR analysis of gene expression in small fat biopsy samples. Our observation of increased IL-10 expression and a decrease in the ratio of type 1 to type 2 ATMs in visceral fat from obese Mgl1^−/− mice provides further support for the idea that ATM activation state is an important variable in the generation of adipose tissue inflammation. Changes in IL-10 were not observed in subcutaneous fat in the knockout animals which may be caused by the fact that type 1 ATMs do not accumulate in this depot as significantly as in visceral fat (Lumeng et al., 2007b). Such differences may also be caused by other sources of IL-10 such as newly identified adipose tissue regulatory T cells (Feuerer et al., 2009).

We also found that MGL1-deficient monocytes had decreased adhesion to adipocytes in vitro which can contribute to the significant decrease in type 1 ATM accumulation in obese Mgl1^−/− mice. Less efficient binding of the Mgl1^−/− 7/4^hi monocytes to adipocytes may work in concert with their lower levels to attenuate inflammation. Although glycan arrays have shown that MGL1 binds to Lewis X and Lewis A trisaccharides, the full range of binding partners in vivo has yet to be revealed. We speculated that DIO and hyperglycemia promote the formation of a Lewis X containing ligands for MGL1 and observed the induction of Lewis X expression in CLS with DIO, an observation that may provide a means for circulating MGL1⁺ 7/4^hi monocytes to traffic to CLS.

We did not find evidence that other lectin receptors such as MGL2 or CD206 are up-regulated in macrophages in response to MGL1 deficiency. Human MGL promotes interactions between DCs and endothelial cells that inhibit migration and promote retention of DCs in lymphoid organs (van Vliet et al., 2008b). Glycosylation-dependent interactions between leukocytes and endothelial cells play a significant role in the recruitment of inflammatory cells to atherosclerotic plaques via interactions between PSGL-1 and P-selectin (Lowe, 2003; Martinez et al., 2005). Our observations with Lewis X are consistent with these observations, and the report of alterations in glycosylation at Lewis X residues on α1-acid glycoprotein in type I diabetes (Poland et al., 2001). Further studies will be needed to assess if mMGL1, mMGL2, or hMGL are potential targets for inhibitors that could specifically target the obesity-induced trafficking of inflammatory monocytes and attenuate the formation of type 1 ATMs.

We have identified ST2 as a novel marker of type 2 ATMs. The expression of this receptor can inhibit signaling from TLR receptors suggesting another mechanism by which type 2 ATMs maintain a suppressed inflammatory state. Lack of ST2 in type 1 ATMs may help maintain proinflammatory signals from TLR4 activated by free fatty acids that are concentrated in adipose tissue (Shi et al., 2006; Nguyen et al., 2007; Coenen et al., 2009).

We have previously referred to ATM subtypes as M1 and M2 ATMs; however, this is imprecise and likely overly simplistic because these states were defined in vitro (Gordon and Taylor, 2005). Additionally, how the “M1/M2 paradigm” relates to human macrophages is still unclear (Zeyda et al., 2007; Bourlier et al., 2008). We propose a new nomenclature for these ATM subtypes: type 1 ATMs are F4/80⁺ MGL1⁻ CD206⁻ ST2⁻ CD11c⁺ and are concentrated in CLS. These ATMs express many genes seen in classically activated M1 macrophages. Type 2 ATMs are F4/80⁺ MGL1⁺ CD206⁺ ST2⁺ CD11c⁻ and represent the resident ATM population localized to interstitial spaces between adipocytes. These cells express many, but not all, markers of alternatively activated M2 macrophages. This redefinition leaves room to define new ATM types as the examination of ATMs is broadened to include different fat depots and to define the properties of human ATMs.

Mgl1^−/− mice on a C57BL/6 background were provided by the Consortium for Functional Glycomics Core F. Male C57BL/6 mice were obtained from Jackson ImmunoResearch Laboratories. Ccr2^−/− mice were provided by S. Kunkel (University of Michigan, Ann Arbor, MI). DIO was induced by feeding male mice a HFD consisting of 60% of calories from fat (Research Diets, Inc.) starting at 6–8 wk of age for 16–18 wk. Control mice were fed a ND consisting of 4.5% fat (5002; LabDiet). Animals were housed in a specific pathogen–free facility with a 12 h light/12 h dark cycle and given free access to food and water. All 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.

Anti-CD206/Mannose receptor, anti-CD301, and anti-7/4 Neutrophil were obtained from AbD Serotec. Anti-CD301/MGL1 and anti-ST2 antibodies were obtained from Abcam. Anti-F4/80, biotinylated anti-Lewis X (SSEA-1), and anti-Mac2 antibodies were obtained from eBioscience. Fc-block and anti-Ly6G antibodies were obtained from BD Biosciences. Isolectin-Alexa Fluor 647 was obtained from Invitrogen.

Immunofluorescence was performed as previously described (Lumeng et al., 2008). In brief, mice were perfused with 1% paraformaldehyde for fixation before dissecting fat pads. Samples were incubated en bloc in primary and secondary antibodies in 5% BSA in PBS. Imaging analysis was performed on an Olympus FluoView confocal microscope or an Olympus inverted microscope for epifluorescence. Image capture was performed with FluoView software or an Olympus DP72 charge-coupled device camera. Confocal images were pseudocolored in FluoView or ImageJ (National Institutes of Health). Composite images were generated using Adobe Photoshop. H&E-stained sections were imaged and adipocyte size was assessed by manually delineating adipocyte membranes and calculating cross-sectional area in ImageJ.

Analysis of adipose tissue stromal vascular fraction (SVF) by flow cytometry was performed as previously described (Lumeng et al., 2008). In brief, after collagenase digestion, adipose tissue cell suspensions were spun at 500 g for 5 min to separate floating adipocytes from the SVF pellet. After erythrocyte lysis, cells were incubated with Fc Block (BD) and stained according to manufacturer recommendation before analysis. TG elicited PMs were treated with 0.1 µg/ml LPS or vehicle for 6 h before collection. Cells were fixed, permeabilized (AbD Serotec), and stained with anti-TNF antibodies (BD) before analysis.

Real-time quantitative RT-PCR was performed as previously described (Lumeng et al., 2007b, 2008).

Glucose tolerance tests were performed after IP injection with 0.7 g/kg dextrose. Insulin tolerance tests were performed after injection with 1 U/kg insulin. Glucose measurements were made with a glucometer (FreeStyle; Abbot Laboratories). Serum fasting insulin levels were determined by ELISA (Crystalchem Inc.). Oxygen consumption (VO2), carbon dioxide production, spontaneous motor activity, and food intake were measured using the Comprehensive Laboratory Monitoring System (CLAMS; Columbus Instruments) in the University of Michigan Animal Phenotyping Core. Body composition was determined using a Minispec NMR analyzer (Bruker Optics). FFA levels were assessed in duplicate using colorimetric assays (Zen-Bio Inc.).

Thioglycollate-elicited PMs were isolated after i.p. injection of 3 ml of 3% Brewers TG by lavage with ice-cold PBS. Cells were enriched by adhesion to plastic and cultured in DMEM with 10% heat-inactivated fetal bovine serum (Invitrogen). 3T3-L1 adipocytes were grown and differentiated as previously described (Baumann et al., 2000). For in vitro adhesion assays, 0.5 × 10⁶ CFSE-labeled monocytes from either genotype were added to differentiated adipocytes or confluent bEnd.3 monolayers in 12-well dishes for 30 min at 37°C. In some experiments, bEnd.3 cells were treated for 3 h with 50 ng/ml recombinant TNF (Sigma-Aldrich) before the addition of monocytes. After adhesion, wells were washed three times in PBS, and then fixed in 10% formalin. Adherent cells were quantitated after DAPI staining and imaging at 200× to enumerate monocytes in direct contact with adipocytes and normalized to the number of adipocytes per high power field.

Monocytes were isolated from single-cell suspensions from spleen by Ficoll gradient (GE Healthcare). After washing, cells were labeled with by incubation with 5 µM CFSE in PBS and quenched with serum containing media. After washing in PBS, cells were counted and 2 × 10⁶ cells were injected I.V. (retroorbital) in 200 µl total volume. 24 and 72 h after injection, blood was obtained by tail bleed in heparinized capillary tubes and stained for flow cytometry after erythrocyte lysis. CFSE⁺ 7/4^hi CD11b⁺ cells in the monocyte gate were quantitated and expressed as a percentage of all 7/4^hi CD11b⁺ cells.

Statistical comparisons made using Student's t tests with significance defined as a P value <0.05. Graphs presented as mean and errors bars represent standard error of the mean.

This work was funded by National Institutes of Health grants DK078851 (C.N. Lumeng) and DK060591 (A.R. Saltiel) and the Ferrantino Investigator Award from the University of Michigan Dept. of Pediatrics and Communicable Diseases (C.N. Lumeng). This work utilized the Animal Phenotyping and Cell and Molecular Biology Cores of the Michigan Diabetes Research and Training Center funded by DK020572 from the National Institute of Diabetes and Digestive and Kidney Diseases. The Mgl1 KO mice were provided by The Consortium for Functional Glycomics funded by NIGMS - GM62116.

The authors have no conflicting financial interests.