Loss of Pdk1-Foxo1 Signaling in Myeloid Cells Predisposes to Adipose Tissue Inflammation and Insulin Resistance

All experimental protocols with mice were approved by the animal ethics committees of the Keio University School of Medicine (09134-1). To create macrophage-specific Pdk1 knockout mice, Pdk1^flox/flox mice (11) were crossed with LysMCre transgenic mice (14). The generation of R26^{floxneoCNFoxo1} mice was described previously (11). Only animals from the same generation of the mixed-background strain were compared. All mice studied were examined on a B6/129 mixed genetic background. Mice were obtained from two independent cohorts of independent breeders, and littermates were used for every in vivo study. Animals were housed in sterile cages in a barrier animal facility at 22–24°C with a 12-h light/dark cycle.

All antibodies used in the current study are available upon request.

For high-fat diet (HFD) studies, we used age-matched (28-week-old) mice. We started the HFD at age 4 weeks for the 24-week HFD and at age 24 weeks for the 4-week HFD. All of the HFD mice were compared with age-matched mice fed a normal chow diet (NCD). The HFD was described previously (15). Analysis was limited to male mice because they are more susceptible to insulin resistance and diabetes. We performed intraperitoneal glucose tolerance tests (IPGTTs) after an overnight fast and insulin tolerance tests (ITTs) after fasting for 3–5 h. The area under the curve (AUC) was calculated from the level of each measured point by the trapezoidal method.

Flow cytometric analysis was performed as described previously (16).

We measured glycogen content as described previously (17).

Double-positive cells were counted and marked digitally to prevent multiple counts with Adobe Photoshop CS4 EXTENDED and ImageJ software (National Institutes of Health, Bethesda, MD). Cells were counted in eight mice for each HFD duration. At least 300 cells were counted in each mouse.

Measurement of H₂O₂ production was performed as described elsewhere (18). Epididymal fat was dissected from age-matched male C57BL/6J mice on either an NCD or a 4-24 week HFD.

Measurement of number of crown-like structures (CLSs) was performed as described previously (16).

Adipocyte size was measured with FLVFS-LS software (Flovel, Tokyo, Japan) by manually tracing a minimum of 1,200 adipocytes for each mouse. We measured adipocytes in at least six mice of each genotype.

Isolation of bone marrow–derived macrophages (BMDMs) was performed as described elsewhere (19).

Transwell migration assays were performed as previously described (20).

Adenovirus constructs that encoded Foxo1 mutants are described elsewhere (21,22). RAW264.7 cells were infected with adenoviruses (10–100 multiplicity of infection [MOI]) and harvested after 48 h. For cotransductions, cells were first transduced with an adenovirus that encoded Flag-CNFoxo1 at the indicated MOI for 8 h. The virus was then removed from the culture dish, and the cells were transduced with another adenovirus that encoded HA-Δ256Foxo1 at the indicated MOI for 8 h.

The isolation of total RNA and real-time PCR were performed as described previously (15). All primer sequences are available upon request.

Western blotting was performed as described previously (15). Insulin-stimulated phosphorylation of insulin receptor substrates (IRSs) and Akt were performed as described elsewhere (23).

Several DNA fragments containing the mouse C-C motif chemokine receptor 2 (Ccr2) promoter were PCR-amplified from mouse genomic DNA. After verifying their nucleotide sequences by DNA sequencing, the Ccr2 promoter fragments were cloned into the luciferase reporter pGL3-Basic vector (Promega, Madison, WI). All primer sequences are available upon request.

The QuickChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used to alter the consensus Foxo1 binding elements in the Ccr2 promoter in PGL3-Basic vectors. Mutated nucleotides were confirmed with DNA sequencing. All primer sequences are available upon request.

The luciferase assay was performed as described previously (22).

Electrophoretic mobility shift assay (EMSA) and the super shift assay were performed as described previously (24).

Chromatin immunoprecipitation (ChIP) assay was performed as described previously (22).

We calculated descriptive statistics with ANOVA followed by Fisher test (Statview; SAS Institute Inc.). P < 0.05 was considered significant. Differences between two groups or among three groups were investigated with two-way repeated-measures ANOVA with an ad hoc multiple comparison method (Fisher least significant differences [LSD] test).

To explore the significance of insulin signaling pathway in ATM, we examined insulin receptor (InsR) protein expression in ATMs by double immunofluorescence with anti-InsR and anti-F4/80 antibodies. During the HFD, ATM InsR protein levels were significantly reduced by ∼50% compared with controls (Fig. 1A and B).

Next, we explored Pdk1 expression in ATMs under different diets. Immunofluorescence of epididymal fat from C57BL/6J mice on an HFD for 16 weeks revealed that cells positive for the macrophage marker CD68 were also positive for Pdk1 (Fig. 2A, top). Because Pdk1 activity depends on Ser 241 phosphorylation (25), we probed with an antiphospho-Pdk1 antibody. On an NCD, ∼80% of F4/80⁺ cells were stained with antiphospho-Pdk1. On an HFD for 24 weeks, the proportion of phospho-Pdk1⁺ ATMs gradually decreased from 80% to from 40 to 50% (Fig. 1C). These data confirm that the InsR-Pdk1 pathway was functionally regulated in ATMs during the HFD.

To explore the relative importance of Foxo family members in ATMs, we compared the expression of Foxo1, Foxo3a, and Foxo4 in M1 and M2 macrophages isolated from the stromal vascular fraction (SVF) of epididymal fat from C57BL/6J mice fed an HFD for 16 weeks. We defined F4/80⁺CD11c⁺CD206⁻ cells as M1 macrophages and F4/80⁺CD11c⁻CD206⁺ cells as M2 macrophages (16). M1 macrophages showed significantly increased Foxo1 expression compared with M2 macrophages. Foxo3a expression was also increased in M1 compared with M2 macrophages but not significantly. In contrast, M1 and M2 macrophages showed similar Foxo4 expression (Fig. 1D). These observations suggest that Foxo1 played an essential role in ATMs.

Because Foxo1 activity depends on its subcellular localization (26), we examined Foxo1 with immunofluorescence in ATMs from age-matched C57BL/6J mice fed an NCD or HFD. Under the NCD, ∼20% of Foxo1 was localized to the nucleus. After 24 weeks of an HFD, ∼45% of Foxo1 was localized to the nucleus (Fig. 1E and F). These data suggest that Foxo1 was functionally significant in ATMs.

Foxo1 is regulated by oxidative stress through H₂O₂ production and the Jun NH₂-terminal kinase (JNK)–mammalian Ste20-like kinase 1 (MST1) pathway, which induces Foxo1 nuclear translocation (27–30). The production of H₂O₂ significantly increased at ∼24 weeks of HFD (Fig. 1G). Furthermore, JNK and MST1 phosphorylation significantly increased after 24 weeks of HFD (Fig. 1H and I). These data suggest that both decreased Pdk1 phosphorylation and activation of the JNK-MST1 pathway may contribute to Foxo1 nuclear localization.

To clarify the function of Pdk1 in ATMs, we generated mice that lacked Pdk1 in macrophages/granulocytes (LysMPdk1^−/−). Efficient, specific Pdk1 deletion was evidenced by immunofluorescence (Fig. 2A) and Western blot analysis (Fig. 2B). Thus, we could study the effects of cell-specific Pdk1 deficiency.

The deletion of Pdk1 in ATMs was expected to cause nuclear localization of Foxo1. Immunofluorescence with an anti-Foxo1 antibody in epididymal fat revealed that ∼60–70% of Foxo1 was localized to the nuclei of ATMs in LysMPdk1^−/− mice (Fig. 2C). We assumed that Foxo1 was active in Pdk1-deficient ATMs and that this activity could be blocked with the dominant-negative form of Foxo1 (Δ256Foxo1), which lacked a COOH-terminal transactivation domain (31). To investigate this, we crossed R26^{floxneoΔ256FoxO1} (11) with LysMCre transgenic mice to generate R26^{floxneoΔ256FoxO1} LysMCre (Δ256Foxo1^LysM) double heterozygotes. Real-time PCR analysis and immunofluorescence confirmed the macrophage-specific expression of the transgene and the nuclear localization of FLAG-Δ256Foxo1, respectively (Supplementary Figs. 1 and 2). We crossed Δ256Foxo1^LysM with Pdk1^flox/+ to generate double mutant mice (Δ256Foxo1^LysMPdk1^+/−). Finally, these mice were crossed with Pdk1^flox/+ to generate Δ256Foxo1^LysMPdk1^−/− (Δ256LysMPdk1^−/−) mice (Supplementary Fig. 3). As expected, Δ256LysMPdk1^−/− mice showed excess nuclear Foxo1 in F4/80⁺ cells from epididymal fat (Fig. 2C).

The LysMPdk1^−/− and Δ256LysMPdk1^−/− mice exhibited normal body weight when fed an NCD (Fig. 2D), and their epididymal fat tissue weight and adipocyte sizes were similar to those of control mice (Supplementary Fig. 4A and B). However, the IPGTTs revealed that LysMPdk1^−/−, but not Δ256LysMPdk1^−/−, mice exhibited glucose intolerance (Fig. 2E and F). Insulin secretion during the IPGTT was higher in LysMPdk1^−/− mice than in controls and Δ256LysMPdk1^−/− mice, but the difference was not significant (Fig. 2G). Furthermore, insulin tolerance significantly decreased in LysMPdk1^−/− mice compared with control and Δ256LysMPdk1^−/− mice (Fig. 2H). These data indicate that the deletion of Pdk1 deteriorates insulin sensitivity and that the ectopic expression of Δ256Foxo1 ameliorates insulin sensitivity.

To identify the tissues that are responsible for insulin resistance, we investigated insulin-stimulated phosphorylation of IRS1, IRS2, and/or Akt in epididymal fats, liver, and skeletal muscle from control, LysMPdk1^−/−, and Δ256LysMPdk1^−/− mice. In epididymal fat and liver, insulin-stimulated phosphorylation of IRS1 or IRS2 and Akt was significantly decreased in LysMPdk1^−/− mice compared with control mice (Fig. 2I). However, insulin-stimulated phosphorylation of IRS and Akt in epididymal fat and liver from Δ256LysMPdk1^−/− mice was similar to that of control mice (Fig. 2I). The expression of G6pc was significantly increased in liver from LysMPdk1^−/− compared with Δ256LysMPdk1^−/− mice (Fig. 2J); moreover, the hepatic glycogen content of LysMPdk1^−/− mice was significantly decreased compared with control and Δ256LysMPdk1^−/− mice (Fig. 2K). In contrast, Akt phosphorylation in skeletal muscle from LysMPdk1^−/− mice was similar to that of control and Δ256LysMPdk1^−/− mice (Fig. 2I). These data indicate that the deletion of Pdk1 in ATMs led to insulin resistance, mainly in adipose tissue and liver, and that ectopic expression of Δ256Foxo1 ameliorated insulin resistance in those tissues.

A CLS is the accumulation of immune cells around dead adipocytes (32). We found that the number of F4/80⁺ CLSs per field in epididymal fat was significantly higher in LysMPdk1^−/−mice than in control and Δ256LysMPdk1^−/− mice (Fig. 3A).

The SVF of adipose tissue from 20-week-old mice contained a substantially higher proportion of F4/80⁺ cells in LysMPdk1^−/− compared with control mice (Fig. 3B and C). Analysis of macrophage subpopulations in the SVF showed a higher proportion of F4/80⁺CD11c⁺CD206⁻ cells in LysMPdk1^−/− mice than in control mice (Fig. 3B and C). In contrast, the adipose tissue of Δ256LysMPdk1^−/− mice showed significantly reduced proportions of F4/80⁺ cells and F4/80⁺CD11c⁺CD206⁻ cells compared with LysMPdk1^−/− mice (Fig. 3B and C). These data suggest that the deletion of Pdk1 caused a significant increase in the proportion of M1 macrophages in epididymal fat, and the proportion was reduced with the overexpression of Δ256Foxo1.

Consistent with the above findings, the expression of chemokine (C-C motif) ligand 2 (Ccl2) (also known as monocyte chemoattractant protein-1 [Mcp-1]) and Cd68 in epididymal fat (Fig. 3D) and of Ccr2 and Tnfa in SVF from LysMPdk1^−/− mice were significantly increased compared with control and Δ256LysMPdk1^−/− mice (Fig. 3E). Furthermore, the expression level of IL-1 receptor antagonist, which is a naturally occurring antagonist of IL-1β and produced by adipose and other tissues (33), in SVF from LysMPdk1^−/− mice was significantly decreased compared with control mice (Fig. 3E). These data support the notion that the deletion of Pdk1 increased the recruitment of M1 macrophages to adipose tissues.

To clarify the function of Foxo1 in ATMs, we generated macrophage-specific CNFoxo1 transgenic mice. We crossed Rosa26-CNFoxo1 (11) with LysMCre (CNFoxo1^LysM) mice. Real-time PCR revealed that the transgene was expressed exclusively in the spleen, liver, hypothalamus, and lung and in ATMs from the epididymal fat (Supplementary Fig. 5). These tissues have tissue-specific macrophages, which include the cells in the sinusoidal lining of the spleen, Kupffer cells in the liver, microglia in the hypothalamus, and alveolar macrophages in the lung (14,34,35). Therefore, resident macrophages likely account for the increased expression of the transgene in these tissues. Immunofluorescence of the epididymal fat showed that FLAG-CNFoxo1 was exclusively localized in the nucleus of F4/80⁺ macrophages (Supplementary Fig. 6). Furthermore, immunofluorescence revealed that ∼50% of F4/80⁺ cells in epididymal fat of CNFoxo1^LysM mice were positive for FLAG (Fig. 4A) and that the percentages of nuclear Foxo1⁺ cells in adipose tissue of CNFoxo1^LysM fed an HFD for 16 weeks was significantly increased compared with control mice fed an HFD or CNFoxo1^LysM fed an NCD (Fig. 4B). These results show that CNFoxo1^LysM mice were an appropriate model for studying the specific effects of overexpressing Foxo1 in ATMs.

On an NCD, CNFoxo1^LysM mice exhibited normal body weight, glucose tolerance, insulin secretion, and insulin sensitivity (Supplementary Fig. 7A–D). On an HFD, the body and tissue weights of CNFoxo1^LysM mice were similar to those of control mice. However, adipocyte size in the epididymal fat of CNFoxo1^LysM mice tended to be larger than that in control mice (Fig. 4C and Supplementary Fig. 8A–C). Although, on the HFD, CNFoxo1^LysM and control mice exhibited similar glucose tolerance (Fig. 4D), the AUC of the IPGTT was significantly increased in CNFoxo1^LysM compared with control mice (Fig. 4E). Furthermore, the CNFoxo1^LysM mice exhibited significantly increased insulin secretion and decreased insulin sensitivity (Fig. 4F and G). These data suggest that the CNFoxo1 in ATMs caused insulin resistance.

Adipocyte size and CLS density exhibit a positive correlation (32,36). Indeed, under HFD conditions, CNFoxo1^LysM mice had a significantly higher number of CLSs in epididymal fat than control mice (Fig. 4H). Phenotypic analysis of ATMs revealed significantly more F4/80⁺ cells in the SVF of CNFoxo1^LysM mice compared with control mice (Fig. 4I and J). Further analysis showed that CNFoxo1^LysM mice had a significantly higher percentage of F4/80⁺CD11c⁺CD206⁻ and F4/80⁺CD11c⁻CD206⁺ cells compared with control mice (Fig. 4I and J). These data suggest that the CNFoxo1^LysM mice have increased numbers of macrophages in adipose tissues under HFD conditions.

To investigate how CNFoxo1 increased the M1 macrophage subpopulation in adipose tissue, we analyzed gene expression in the SVF of epididymal fat from mice fed an HFD. Real-time PCR demonstrated that CNFoxo1^LysM mice expressed significantly higher levels of Ccr2 and Tnfa mRNAs than control mice (Fig. 5A). Furthermore, the level of Ccr2 expression in F4/80⁺CD11c⁺CD206⁻ cells was significantly increased in CNFoxo1^LysM mice compared with control mice (Fig. 5B). To examine whether CNFoxo1 directly induces Ccr2 expression, we infected RAW264.7 cells with an adenovirus encoding β-galactosidase or CNFoxo1. Overexpression of CNFoxo1 in RAW264.7 cells significantly increased endogenous Ccr2 expression (Fig. 5C–E). These data suggest that the overexpression of CNFoxo1 in ATMs increased Ccr2 expression.

Next, we investigated whether Δ256Foxo1 could block Foxo1-induced Ccr2 expression. We cotransduced RAW264.7 cells with adenoviruses that encoded Flag-CNFoxo1 and HA-Δ256Foxo1. We found that the presence of Δ256Foxo1 inhibited the expression of endogenous Ccr2 in a dose-dependent manner (Fig. 5F and G). These data indicate and confirm that the Δ256Foxo1 construct had a dominant negative effect on Foxo1-induced Ccr2 expression.

To determine whether Foxo1 regulation of Ccr2 expression was involved in insulin signaling, we tested whether insulin or IL-4 inhibited Ccr2 expression in a Foxo1-dependent manner. Both insulin and IL-4 could significantly inhibit Ccr2 expression in BMDMs from control and Δ256LysMPdk1^−/− mice but not in BMDMs from LysMPdk1^−/− and CNFoxo1^LysM mice (Fig. 5H). In contrast, insulin and IL-4 did not affect expression of other genes specifically expressed in M1 macrophages, including Tnfa and Il1b (Fig. 5H). These data indicate that Foxo1-induced Ccr2 expression was regulated by both insulin and IL-4.

To analyze the functional effects of Pdk1 deficiency in macrophages, we performed transwell migration assays with BMDMs. Pdk1-deficient BMDMs exhibited significantly more migration than BMDMs from control and Δ256LysMPdk1^−/− mice (Fig. 5I). Furthermore, BMDMs from CNFoxo1^LysM mice exhibited significantly increased MCP-1–stimulated migration capacity compared with control BMDMs (Fig. 5J). These data confirm that a Pdk1 deficiency and/or Foxo1 activation in macrophages resulted in increased migration as a result of increased expression of Ccr2.

To characterize the Foxo1 response element (FRE) in the Ccr2 promoter, we constructed different versions of the mouse Ccr2 promoter by progressively deleting portions of the upstream region. The transcriptional activity of each mutant promoter in response to CNFoxo1 binding was examined in HEK293T cells (Fig. 6A). Ccr2 promoters with deletions up to −291 nucleotides (nt) responded to Foxo1 transactivation. However, further deletions, up to −208 nt, completely abolished transcription of the reporter (Fig. 6A). Thus, the FRE was confined to a small nucleotide region between −291 and −208 in the mouse Ccr2 promoter. Consistent with this observation, the promoter region contained several putative Foxo response elements (FREs), including GTAAAT from −254 to −249 nt and AAACA from −215 to −211 nt (Fig. 6A). It is interesting that the former region is conserved among human, mouse, and rat Ccr2 promoters (Supplementary Fig. 9). To confirm this finding, we generated one additional truncated mutant promoter (237Ccr2), which had the latter FRE but not the former. The 237Ccr2 promoter did not respond to Foxo1 induction. These data suggest that the AAACA sequence from −215 to −211 was unnecessary for Foxo1 activation of the Ccr2 promoter. We also generated two additional mutant Ccr2 promoters, one harboring nucleotide substitutions between −254 and −249 (254mut) and one with substitutions between −215 and −211 (215mut). Foxo1 induced transcription from the 215mut but not from the 254mut Ccr2 promoter (Fig. 6A). These data suggest that the GTAAAT sequence from −254 to −249 nt in the mouse Ccr2 promoter was the functional FRE.

To examine the ability of this putative FRE to bind Foxo1, we conducted an EMSA. Foxo1 caused significant retardation of the FRE DNA (Fig. 6B, lane 1). Inclusion of the anti-cMyc antibody resulted in a supershifted DNA band (Fig. 6B, lane 2). The same EMSA was performed using a mutant DNA containing five base substitutions within the FRE motif as a control. Alterations in the consensus FRE motif abrogated its ability to bind Foxo1 (Fig. 6B, lane 4). Incubating nuclear extracts from cells expressing cMyc-tagged Foxo1 with a probe encoding the 31–base pair FRE DNA sequence yielded a slower complex that was competed out by excess cold probe (Fig. 6C, lanes 1–4) but not mutant probe (Fig. 6C, lanes 5–8).

We performed a ChIP assay to determine the association between Foxo1 and the Ccr2 promoter in RAW264.7 cells. Because of low levels of Foxo1 expression in RAW264.7 cells, we transduced cells with adenovirus encoding CNFoxo1. Using primers flanking the FRE motif within the Ccr2 promoter (Fig. 6D), we detected a sequence-specific DNA corresponding to the proximal region (−474/9 nt) of the Ccr2 promoter in immunoprecipitates obtained with anti-FLAG antibody (Fig. 6E). We also performed PCR analysis using a pair of off-target primers flanking distal regions (−1857/−1667 and −1425/−1177 nt). No specific DNA was amplified in the immunoprecipitates using normal mouse IgG or anti-FLAG antibody (Fig. 6E). These data confirm that Foxo1 directly binds the Ccr2 promoter and that Ccr2 is a target gene of Foxo1.

To determine whether the Pdk1-Foxo1 pathway was essential for alternative activation of macrophages, we analyzed macrophage signatures in insulin- or IL-4–stimulated BMDMs from control, LysMPdk1^−/−, CNFoxo1^LysM, and Δ256LysMPdk1^−/− mice. The signature genes, including Arg1, Cd163, Il10, and Mr, were significantly induced by insulin or IL-4 in BMDMs from control mice (Fig. 6F). In contrast, Pdk1 deficiency or constitutive Foxo1 activation completely abolished insulin- or IL-4–stimulated induction of the genes necessary for alternative macrophage activation (Fig. 6F). It is interesting that the expression of transactivation-defective (Δ256) Foxo1 rescued IL-4–induced, but not insulin-induced, gene expression (Fig. 6F). These data indicate that the Pdk1-Foxo1 pathway was required for the activation of macrophages via the alternative pathway.

To determine whether blocking Foxo1 transactivation by expressing Δ256Foxo1 in ATMs would alleviate insulin resistance, we compared glucose homeostasis and insulin sensitivity in wild-type and Δ256Foxo1^LysM mice fed an HFD for 24 weeks. We observed no differences in body weight, glucose tolerance, or insulin secretion between genotypes (Fig. 7A–C). Furthermore, the Δ256Foxo1^LysM mice showed a weak but significant improvement in insulin sensitivity compared with wild-type mice (Fig. 7D and E).

After a 24-week HFD, Δ256Foxo1^LysM and wild-type mice had similar proportions of F4/80⁺, F4/80⁺CD11c⁺CD206⁻, and F4/80⁺CD11c⁻CD206⁺ cells in adipose tissues (Fig. 7F). Moreover, in epididymal fat, no differences were observed in the gene expression profiles of M1 macrophages, including Ccr2, Il1b, Tnfα, and Il6. However, there was a significant increase in Arg1 expression in Δ256Foxo1^LysM compared with control mice (Fig. 7G). Taken together, these data show that overexpression of Δ256Foxo1 in macrophages did not prevent glucose intolerance, but it did partially alleviate insulin resistance.

In the current study, we demonstrate that Pdk1 in ATMs inhibits recruitment of M1 macrophages into adipose tissues, while Foxo1 antagonizes these processes. These findings suggest that the Pdk1-Foxo1 signaling pathway in ATMs is important for regulation of chronic inflammation and insulin sensitivity in vivo (Fig. 8).

The key finding of the current study was that Foxo1 targeted Ccr2 expression in macrophages. Ccr2 is the primary receptor for Mcp1/Ccl2, a member of the chemokine family of proteins. Ccr2 is expressed on circulating monocytes and ATMs, where it serves as a crucial monocyte recruitment factor by directing macrophages to sites of injury and inflammation. Furthermore, Ccr2 is important in the regulation of insulin sensitivity in vivo. Obesity increases the production of Ccl2 in adipose tissues, which leads to an accumulation of Ccl2-bound macrophages. When recruited macrophages are classically activated, they secrete proinflammatory cytokines, which leads to insulin resistance in various insulin-responsive tissues (2). Indeed, Ccr2 deletion ameliorated insulin resistance in HFD-induced insulin resistance (37). Therefore, our observation of increased Ccr2 expression in SVF M1 macrophages in LysMPdk1^−/− and CNFoxo1^LysM mice was an important cue that insulin resistance had developed. Thus, the current study directly demonstrates that ATM Foxo1 played a pivotal role in regulating insulin sensitivity in vivo.

Nuclear accumulation of Foxo1 suddenly increased at 24 weeks of HFD, although phosphorylation of Pdk1 was not changed. These findings suggest that another signaling pathway may be involved in subcellular localization of Foxo1 in ATMs. One of the candidates is MST1, which mediates oxidative stress, phosphorylates FOXO proteins at a conserved site within the forkhead domain, disrupts their interaction with 14–3-3 proteins, and promotes FOXO nuclear translocation (27). Furthermore, JNK is known to phosphorylate and activate MST1 (30). HFD increased oxidative stress (18), leading to activation of JNK, MST1, and nuclear accumulation of Foxo1 (27). Of interest, we observed that H₂O₂ significantly increased at 24 weeks of HFD and that phosphorylation of MST1 also significantly increased at the same time. Therefore, oxidative stress in HFD may contribute to nuclear accumulation and activation of Foxo1. The findings suggest that nuclear accumulation of Foxo1 contributes to recruitment of M1 macrophages into adipose tissue during HFD.

However, we observed that expression of Δ256Foxo1 just partially protected against diet-induced insulin resistance and could not rescue Ccr2 expression in mice fed an HFD for 24 weeks. Furthermore, the current study demonstrates that nuclear localization of Foxo1 started to occur at 24 weeks of HFD. Therefore, it is possible that nuclear localization of Foxo1 plays a role specifically in the late progression of diet-induced insulin resistance. From the current study, nuclear accumulation of Foxo1 in ATMs is only 40–45% of all F4/80⁺ at 24 weeks of HFD, which means that an HFD cannot activate Foxo1 in ATMs completely. In contrast, the percentages of nuclear Foxo1 in ATMs of LysMPdk1^−/− and CNFoxo1^LysM fed an HFD are ∼70%. Therefore, the effect of loss of transactivation of Foxo1 on Ccr2 expression in an HFD is small compared with LysMPdk1^−/− mice. Alternatively, nuclear Foxo1 in myeloid cells may promote insulin resistance by other mechanisms than its role in the control of Ccr2 gene expression. Furthermore, the CNFoxo1^LysM mice fed an NCD did not exhibit insulin resistance, while LysMPdk1^−/− mice exhibited insulin resistance. These findings suggest that Foxo1 per se is not sufficient to cause HFD-induced insulin resistance, although Foxo1 may enhance the negative effect of an HFD on insulin sensitivity.

Our results provide direct evidence for the notion that ATM cell autonomous Pdk1-Foxo1 signaling regulates adipose tissue inflammation and insulin sensitivity in vivo. This finding may suggest a new target for pharmacological intervention that could lead to novel therapeutic strategies for treating insulin resistance and type 2 diabetes.

This work was supported by a grant from Nippon Boehringer Ingelheim Co., Ltd. to H.I. and a grant from Keio University Grant-in-Aid for Encouragement of Young Medical Scientists to Y.K. No other potential conflicts of interest relevant to this article were reported.

Y.K. researched data. J.N. conceived the hypothesis, designed and researched data, supervised the analyses, and wrote the manuscript. N.W., S.F., K.I., R.S., Y.H., and K.T. researched data. M.K. and T.N. generated and provided tissue-specific Pdk1 knockout mice. A.Y. provided LysMCre mice and helpful discussion regarding experiments. M.O. researched data and assisted with data interpretation. H.I. supervised all experiments and assisted with preparation of the manuscript. J.N. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Takahito Kaji (Training Department, Customer & Commercial Excellence, MSD K.K.) for analysis of data by two-way repeated-measures ANOVA.