Insulin desensitization occurs not only under the obese diabetic condition but also in the fasting state. However, little is known about the common secretory factor(s) that are regulated under these two insulin-desensitized conditions. Here, using database analysis and in vitro and in vivo experiments, we identified stromal derived factor-1 (SDF-1) as an insulin-desensitizing factor in adipocytes, overexpressed in both fasting and obese adipose tissues. Exogenously added SDF-1 induced extracellular signal–regulated kinase signal, which phosphorylated and degraded IRS-1 protein in adipocytes, decreasing insulin-mediated signaling and glucose uptake. In contrast, knockdown of endogenous SDF-1 or inhibition of its receptor in adipocytes markedly increased IRS-1 protein levels and enhanced insulin sensitivity, indicating the autocrine action of SDF-1. In agreement with these findings, adipocyte-specific ablation of SDF-1 enhanced insulin sensitivity in adipose tissues and in the whole body. These results point to a novel regulatory mechanism of insulin sensitivity mediated by adipose autocrine SDF-1 action and provide a new insight into the process of insulin desensitization in adipocytes.
Introduction
Adipose tissue has long been considered an energy-storage organ. However, accumulating data indicate that adipose tissue also functions as a metabolic organ involved in the regulation of systemic insulin sensitivity and glucose homeostasis. Lipodystrophy (failure of normal adipose tissue development) is associated with severe insulin resistance and hyperglycemia in humans and mice (1,2). Adipose-specific ablation of the insulin receptor or GLUT4, an insulin-responsive GLUT, impairs systemic insulin sensitivity and glucose homeostasis (3,4). Conversely, overexpression of GLUT4 in adipocytes improves systemic glucose disposal (5). Furthermore, selective enhancement of adipocyte-insulin sensitivity is reported to improve systemic glucose homeostasis (6).
Insulin resistance develops not only under obese conditions but also in the fasting state. Insulin resistance develops in obese individuals and is thought to be harmful in the context of being a major risk factor for type 2 diabetes and dyslipidemia (7). However, regulation of insulin action, especially insulin desensitization, is an important aspect of physiological metabolism. Insulin resistance also occurs in fasting healthy subjects, represented by a decreased rate of glucose disposal during hyperinsulinemic-euglycemic clamp (8). Fasting-induced insulin resistance suppresses glucose utilization in peripheral tissues, including adipose tissue, to spare glucose for use by other tissues, such as the brain, that require a large amount of glucose for cellular function and survival (8–10). Although several exogenous and endogenous factors are associated with insulin resistance in adipose tissues (7,11–13), little is known about the common factor(s) that are regulated under both obesity- and fasting-mediated insulin desensitized conditions.
Here, by integrating publicly available microarray data sets, we compiled gene lists abundantly present in these insulin-desensitized conditions, including fasting and obese adipose tissues. This analysis led to the identification of a previously unnoticed chemokine, stromal derived factor-1 (SDF-1), also called CXCL12. We describe here the crucial role of SDF-1 as an autocrine insulin-desensitizing factor in adipocytes, using in vitro adipocyte analysis and adipocyte-specific SDF-1 knockout (AdSDF-1 KO) mice.
Research Design and Methods
Animal Studies
Mice were housed in groups of one to three mice per cage, maintained in a room under controlled temperature (23 ± 1.5°C) and humidity (45 ± 15%) on a 12-h dark/12-h light cycle, and had free access to water and chow (MF; Oriental Yeast, Tokyo, Japan). SDF-1 flox mice (Stock No. 021773) (14) were purchased from The Jackson Laboratory, and other mice were purchased from Charles River Japan (Yokohama, Japan). Adiponectin-Cre mice were provided by E. Rosen (Beth Israel Deaconess Medical Center) (15). Male AdSDF-1 KO mice and their littermate control (SDF-1 flox/flox) mice were analyzed between 13 and 17 weeks of age after 8–12 weeks of a high-fat diet (HFD). Epididymal white adipose tissue (WAT), brown adipose tissue (BAT), liver, and gastrocnemius muscle were harvested and used for the study. A diet-induced obesity mouse model was established by feeding an HFD containing 60% of calories from fat for 8 to 12 weeks, starting at 5 weeks of age (HFD-60; Oriental Yeast).
For the intraperitoneal insulin tolerance test (ITT), mice were fasted for 5 h and injected with 0.7 units/kg body weight (BW) (those fed a normal diet) or 2.0 units/kg BW (those fed HFD) insulin. For the intraperitoneal glucose tolerance test (GTT), mice were fasted for 5 h before injection of glucose at 1 g/kg BW. Glucose values were measured by tail vein sampling at the indicated times using a portable glucose meter (Glutest Neo alpha; Sanwa Kagaku Kenkyusho, Nagoya, Japan). For measurement of insulin-mediated Akt phosphorylation, mice were fasted for 5 h, anesthetized, and injected intraperitoneally with 10 units/kg BW insulin. The relevant tissues were quickly harvested after 15 min and frozen immediately in liquid nitrogen. All mouse studies were approved by the Ethics Review Committee for Animal Experimentation of Osaka University, Graduate School of Medicine, and performed in accordance with the Osaka University Institutional Animal Care and Use Committee Guidelines.
Adipose Tissue Fractionation
Epididymal WAT was excised and minced in DMEM supplemented with 10% FBS and 1% AA (antibiotics and antimycotics). Collagenase (4,000 units/mL) and DNase (0.1 mg/mL) were added, and the tissue samples were incubated at 37°C for 30 min under constant shaking. The cell suspension was filtered through a 110-μm cell strainer and then centrifuged at 500g for 5 min to separate the stromal vascular fraction (SVF) pellet from the floating mature adipocytes fraction. Separated mature adipocytes and SVF cells were resuspended in different tubes and centrifuged at 500g for 5 min. The washing-centrifugation process was repeated twice.
Preparation of Mouse Primary Differentiated Adipocytes
Subcutaneous WAT was excised and minced in DMEM supplemented with 10% FBS and 1% AA. Collagenase (4,000 units/mL) and DNase (0.1 mg/mL) were added, and the tissue samples were incubated at 37°C for 30 min under constant shaking. The cell suspension was filtered through a 70-μm cell strainer and then centrifuged at 500g for 5 min to obtain the SVF pellet. The pellet was resuspended in DMEM containing 10% FBS and 1% AA and plated into an appropriate culture dish. At 4–6 h after seeding, the cells were washed with culture medium to remove superfluous cells and debris. At 2 days after 100% confluence, the cells were differentiated to adipocytes using differentiation medium containing 3-isobutyl-1-methylxanthine (0.5 mmol/L), dexamethasone (1 μmol/L), insulin (1 μmol/L), and pioglitazone (10 μmol/L). The cells were used in the experiment 5–7 days after differentiation.
Cell Culture
3T3-L1 cells were differentiated into adipocytes using differentiation medium containing 3-isobutyl-1-methylxanthine (0.5 mmol/L), dexamethasone (1 μmol/L), and insulin (1 μmol/L). The cells were used in experiments 7 days after differentiation. In all experiments, adipocytes were cultured in a serum-free DMEM to avoid unknown serum effects. PlatE cells were used to produce retrovirus, followed by transfection in 3T3-L1 cells. Stable 3T3-L1 cells expressing tetracycline (tet)-inducible SDF-1 (3T3-L1–tet–SDF-1 cell), IRS-1 (3T3-L1–tet–IRS-1 cell), or control cells (3T3-L1–tet–empty cell) were produced using the Retro-X Tet-On Advanced system according to the protocol supplied by the manufacturer (Clontech, Mountain View, CA). The coding region of mouse SDF-1 or mouse IRS-1 (16) was subcloned into the expression vector, pRetroX-Tight-Pur or pRetroX-Tight-hygro, respectively. Retroviral particles were generated using pRetroX-Tight-Pur-mSDF1, pRetroX-Tight-hygro-mIRS-1, or pRetroX-Tight-hygro-empty and pRetroX-tet-on advanced vectors. Infected 3T3-L1 cells were selected in 400 μg/mL G418 and 5 μg/mL puromycin or 200 μg/mL hygromycin. Recombinant proteins and other materials were as described; murine SDF-1 (R&D Systems, Minneapolis, MN,), murine tumor necrosis factor-α (TNF-α) (PeproTech, Rocky Hill, NJ), U0126 (Sigma-Aldrich, St. Louis, MO), pertussis toxin (Sigma-Aldrich), and TC14012 (R&D Systems and Cayman Chemical Company, Ann Arbor, MI).
Small Interfering RNA
The differentiated 3T3-L1 adipocytes or mouse primary differentiated adipocytes (day 5–7) in 10-cm dish were treated with trypsin-EDTA and incubated at 37°C for 2 min. The cells were washed in a 50 mL conical centrifuge tube and centrifuged at 500g for 5 min. In the meantime, the small interfering (si)RNA mixture of Opti-MEM, siRNA solution (Qiagen, Valencia, CA; AllStars Negative Control siRNA and Flex tube siRNA), and RNAiMAX (Invitrogen, Carlsbad, CA) was prepared according to the instructions provided by the manufacturer. The cell pellet was gently resuspended in the culture medium (106 cells/mL), plated onto 12-well dish with the siRNA mixture, and incubated for 2 days. At 2 days after siRNA, the cells were maintained in a serum-free medium for 12–36 h to allow SDF-1 accumulation in the medium.
Western Blot Analysis
Cultured cells or tissue samples were lysed in lysis buffer (20 mmol/L Tris/HCl [pH 7.4], 1.0% Triton ×100, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA) containing 1 mmol/L phenylmethylsulfonyl fluoride, 1.6 g/mL aprotinin, 10 g/mL leupeptin, and protease inhibitor cocktail (Nacalai Tesque, Kyoto, Japan). Protein concentration was determined by the bicinchoninic acid method (Pierce, Rockford, IL). The samples were used for Western blot analysis after concentrated sample buffer was added and the samples were heated for 5 min at 95°C.
Equal amounts of protein were separated by SDS-PAGE and transferred electrophoretically to polyvinylidene difluoride membranes. The membranes were blocked for 1 h at room temperature using Tris-buffered saline (137 mmol/L NaCl, 20 mmol/L Tris–HCl, pH 7.6) containing 0.05% Tween-20 (TBS-T) and 5% skim milk. After triple washing with TBS-T, each for 10 min, the membranes were incubated overnight at 4°C with primary antibodies against phosphorylated (phospho)–extracellular signal–regulated kinase (Erk; Thr202/Tyr204; Cell Signaling Technology, Danvers, MA), total Erk (Cell Signaling), phospho-Akt (Ser473; Cell Signaling), total Akt (Cell Signaling), total IRS-1 (Upstate, Charlottesville, VA), total IRS-2 (Cell Signaling), phospho–IRS-1 (Ser636/639; Cell Signaling), adiponectin (R&D Systems), β-actin (Sigma-Aldrich), GAPDH (Cell Signaling), and α-tubulin (Cell Signaling) in TBS-T and 5% skim milk. After triple washing with TBS-T, each for 10 min, the membranes were incubated for 1 h at room temperature with enhanced chemiluminescence horseradish peroxidase–linked secondary antibodies (GE Healthcare, Piscataway, NJ) in TBS-T and 5% skim milk. After extensive triple washing in TBS-T, the immunoreactive bands were visualized by Pierce Western Blotting Substrate Plus. Quantification was conducted by densitometry using ImageJ software (National Institutes of Health).
ELISA
ELISA assay kits for SDF-1 (R&D Systems), MCP-1 (R&D Systems), and insulin (Morinaga, Yokohama, Japan) were purchased, and analysis was performed according to the instructions provided by the manufacturer.
RNA Isolation and Quantitative PCR
Total RNA was isolated from cells or tissues using TRI reagent (Sigma-Aldrich). RT-PCR was performed using the Transcriptor First Strand cDNA Synthesis Kit (Roche, Indianapolis, IN). Quantitative PCR was performed with LightCycler-DNA Master SYBR Green I mix (Roche). All procedures were performed using the instructions provided by the manufacturer. The specific primers were purchased from Sigma-Aldrich, and the sequences are available upon request.
DNA Isolation and PCR
Total DNA was isolated from tissues with TRI reagent (Sigma-Aldrich) using the procedure recommended by the manufacturer.
Immunohistochemistry
Epididymal WAT and liver were excised and fixed in 10% formalin. After paraffin embedding and sectioning, the sections were stained using anti-Mac2 antibodies (Abcam).
Flow Cytometry
FACS analysis was performed as described previously. Briefly, cells in the SVF from epididymal WAT were suspended in FACS buffer and incubated with anti-mouse CD16/CD32 (93; BioLegend, San Diego, CA) for 15 min. Then, the cells were rinsed and resuspended in FACS buffer and stained for 25 min with anti-CD45 (30F-11; BioLegend), anti-CD11b (M1/70; BioLegend), anti-MHC class II (M5/114.15.2; eBioscience, San Diego, CA), anti-F4/80 (BM8; BioLegend), CD11c (N418; BioLegend), and anti–siglec-F (E50-2440; BD Pharmingen, San Diego, CA) for macrophages, dendritic cells, and eosinophils. For B cells, CD4+ T cells, CD8+ T cells, natural killer (NK) cells, and NK T cells, SVF was incubated with anti-B220 (RA3-6B2; BioLegend), anti-CD19 (6D5; BioLegend), anti-NK1.1 (PK136; BioLegend), anti-CD8 (53-6.7; BioLegend), anti-CD4 (RM4-5; BioLegend), and anti-CD3 (17A2; BioLegend). For preadipocytes, endothelial cells, and hematopoietic cells, SVF was incubated with anti-CD45, anti-CD31 (390; BioLegend), anti-CD34 (RAM34; eBioscience), and anti–platelet-derived growth factor receptor-α (APA5; BioLegend). The SVF was washed twice and resuspended with 400 μL FACS buffer, and 20 μL precision count beads (BioLegend) was added as internal control and analyzed with FACSVerse (BD Biosciences, San Diego, CA). The absolute cell count was determined according to instructions provided by the manufacturer.
Glucose Uptake Assay
2-Deoxyglucose (2-DG) uptake kit (Cosmo Bio, Tokyo, Japan) was purchased and analysis was performed according to the instructions provided by the manufacturer.
Ex Vivo Glucose Uptake Assay
Epididymal WAT was dissected from mouse, and cut into 50- to 100-mg pieces, washed, and incubated for 1 h with Krebs Ringer Phosphate HEPES buffer at 37°C. Insulin (1 nmol/L) was added and incubated for 20 min, followed by treatment with 1 mmol/L 2-DG, further incubated for 20 min, and then washed four times with PBS. Tissue samples were lysed by sonication in 10 mmol/L Tris-HCl pH 8.0 buffer and intracellular 2-DG-6-phosphate levels were measured according to the 2-DG uptake kit instructions provided by the manufacturer (Cosmo Bio).
Statistical Analysis
All data are presented as mean ± SEM. Differences between two groups were examined for statistical significance by the Student t test. A P value <0.05 denoted the presence of a statistically significant difference.
Results
SDF-1 Gene Expression Correlates With Insulin-Desensitized Conditions in Adipocytes
To search for factors involved in “insulin desensitization” common to fasting and obesity, we compiled and intercrossed gene lists from four independent microarray data sets related to insulin-desensitized or -sensitized conditions: 1) fasting-induced genes in mouse adipose tissue (GSE46495); 2) obesity-induced genes in human adipose tissue (GDS3602); 3) TNF-α–induced genes in 3T3-L1 adipocytes (GSE62635) as upregulated genes in an insulin-resistance model in vitro (17); and 4) peroxisome proliferator–activated receptor-γ (PPARγ) agonist–reduced genes in rat adipose tissue (GDS3850) as downregulated genes in an insulin-sensitive model in vivo (18). Comprehensive analysis of the data sets yielded two overlapping genes, Cyp1b1 and SDF-1 (Fig. 1A). The Cyp1b1 gene has been well described previously, and its deficiency ameliorates glucose intolerance induced by an HFD (19). SDF-1 is a secretary protein classified as a CXC chemokine (20). Its gene expression is exceptionally high compared with other chemokines in RNA sequencing data sets of 3T3-L1 adipocytes (GSM2322563 and GSE50612) (Fig. 1B and Supplementary Fig. 1A). Quantitative PCR and ELISA data showed high expression of SDF-1 in the culture media of 3T3-L1 adipocytes and mouse primary differentiated adipocytes (Supplementary Fig. 1B–E).
SDF-1 gene expression correlates with insulin-desensitized conditions in adipocytes. A: Schematic diagram of microarray analysis to identify insulin-resistance factors expressed in adipose tissue and adipocytes. The following Gene Expression Omnibus DataSets were used for the analysis: fasting-induced genes in mouse adipose tissue (GSE56248, fold-change >1.5, P < 0.05; 781 genes), obesity-induced genes in human adipose tissue (GDS3602, fold-change >3, P < 0.05; 86 genes), TNF-α–induced genes in 3T3-L1 adipocytes (GSE62635, fold-change >3, P < 0.05; 375 genes), and PPARγ agonist–reduced genes in rat adipose tissue (GDS3850, fold-change <0.6, P < 0.05; 477 genes). B: Chemokine gene expressions in 3T3-L1 adipocytes (GSM2322563). RPKM, reads per kilobase million. C: SDF-1 gene expression in epididymal WAT after 24 h feeding or fasting (n = 6). D: SDF-1 gene expression in epididymal WAT of control C57BL/6J mice (WT, wild-type) or ob/ob mice (n = 4–6). E: SDF-1 gene expression with or without TNF-α (10 ng/mL) for 12 h in 3T3-L1 adipocytes (n = 3). F: SDF-1 gene expression with or without pioglitazone (Pio; 10 μmol/L) in 3T3-L1 adipocytes (n = 3). Data are mean ± SEM. **P < 0.01; ***P < 0.001.
SDF-1 gene expression correlates with insulin-desensitized conditions in adipocytes. A: Schematic diagram of microarray analysis to identify insulin-resistance factors expressed in adipose tissue and adipocytes. The following Gene Expression Omnibus DataSets were used for the analysis: fasting-induced genes in mouse adipose tissue (GSE56248, fold-change >1.5, P < 0.05; 781 genes), obesity-induced genes in human adipose tissue (GDS3602, fold-change >3, P < 0.05; 86 genes), TNF-α–induced genes in 3T3-L1 adipocytes (GSE62635, fold-change >3, P < 0.05; 375 genes), and PPARγ agonist–reduced genes in rat adipose tissue (GDS3850, fold-change <0.6, P < 0.05; 477 genes). B: Chemokine gene expressions in 3T3-L1 adipocytes (GSM2322563). RPKM, reads per kilobase million. C: SDF-1 gene expression in epididymal WAT after 24 h feeding or fasting (n = 6). D: SDF-1 gene expression in epididymal WAT of control C57BL/6J mice (WT, wild-type) or ob/ob mice (n = 4–6). E: SDF-1 gene expression with or without TNF-α (10 ng/mL) for 12 h in 3T3-L1 adipocytes (n = 3). F: SDF-1 gene expression with or without pioglitazone (Pio; 10 μmol/L) in 3T3-L1 adipocytes (n = 3). Data are mean ± SEM. **P < 0.01; ***P < 0.001.
We validated the original microarray data sets by quantitative PCR. Fasting increased SDF-1 gene expression in epididymal WAT (Fig. 1C). Treatment with culture medium mimicking fasting conditions, such as insulin depletion or glucose starvation, increased SDF-1 expression in 3T3-L1 adipocytes (Supplementary Fig. 2A and B). We also confirmed that SDF-1 expression was augmented in epididymal WAT of various obese mouse models compared with the control mice, including ob/ob mice (Fig. 1D), KKAy mice (Supplementary Fig. 2C), and HFD-fed mice (Supplementary Fig. 2D), as reported previously (21,22). In the human microarray data set, SDF-1 expression was higher in epididymal WAT of obese subjects than in nonobese subjects (GDS3602) (Supplementary Fig. 2E). Furthermore, SDF-1 expression tended to be higher in adipose tissues of obese insulin-resistant subjects with diabetes compared with age- and BMI-matched normal glucose-tolerant subjects (GDS3665) (Supplementary Fig. 2F), suggesting the involvement of adipose–SDF-1 in human obesity and diabetes as well. Furthermore, the addition of TNF-α to cultured 3T3-L1 adipocytes markedly increased SDF-1 gene expression (Fig. 1E), whereas pioglitazone, a PPAR-γ agonist, decreased SDF-1 gene expression in 3T3-L1 adipocytes (Fig. 1F). These data suggest the upregulation of SDF-1 gene under insulin-desensitized conditions and its downregulation in insulin-sensitive states in adipose tissues and adipocytes.
SDF-1 Directly Induces Insulin Desensitization in Adipocytes With Reduced IRS-1 Protein Levels
To assess the direct effects of SDF-1 on adipocyte insulin sensitivity, we conducted in vitro experiments using cultured adipocytes and evaluated SDF-1–related factors. IRS-1 is an important signaling molecule that regulates insulin action in adipocytes (23,24). IRS-1 deficiency induces insulin resistance in adipocytes with decreased insulin-induced Akt phosphorylation and glucose uptake (25,26). SDF-1 markedly reduced IRS-1 protein levels in 3T3-L1 adipocytes (Fig. 2A) in a manner similar to TNF-α, as reported previously (27). On one hand, IRS-1 protein levels were significantly reduced after 9-h treatment with SDF-1 (Supplementary Fig. 3A and B). On the other hand, SDF-1 did not alter IRS-2 protein levels significantly (Supplementary Fig. 3A and C).
SDF-1 directly desensitizes adipocytes to insulin action with decreased IRS-1 protein levels. A: IRS-1 protein levels in 3T3-L1 adipocytes treated with the vehicle (Control), TNF-α (10 ng/mL), or SDF-1 (500 ng/mL) for the indicated duration (1, 3, or 6 h). B: IRS-1 protein and Akt phosphorylation levels in 3T3-L1 adipocytes pretreated with SDF-1 (500 ng/mL) for the indicated duration (0, 1, 3, or 9 h), followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 5, or 15 min). C: Relative glucose uptake in 3T3-L1 adipocytes pretreated with or without SDF-1 (1,000 ng/mL) in serum-free medium for 18 h, followed by treatment with insulin (0, 0.1, or 1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 3–4). D: SDF-1 gene expression in 3T3-L1–tet–SDF-1 adipocytes treated with (+) or without (−) 30 μg/mL doxycycline (Dox) for 48 h (n = 5, 4). E: IRS-1 protein and Akt phosphorylation levels in 3T3-L1–tet–SDF-1 adipocytes pretreated with or without 30 μg/mL of Dox for 48 h, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, 30, or 60 min). F: Relative glucose uptake in 3T3-L1–tet–SDF-1 adipocytes pretreated with or without 30 μg/mL Dox for 48 h, incubated with serum-free medium for 18 h, followed by treatment with insulin (0.1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 6). Data are mean ± SEM. *P < 0.05; ***P < 0.001.
SDF-1 directly desensitizes adipocytes to insulin action with decreased IRS-1 protein levels. A: IRS-1 protein levels in 3T3-L1 adipocytes treated with the vehicle (Control), TNF-α (10 ng/mL), or SDF-1 (500 ng/mL) for the indicated duration (1, 3, or 6 h). B: IRS-1 protein and Akt phosphorylation levels in 3T3-L1 adipocytes pretreated with SDF-1 (500 ng/mL) for the indicated duration (0, 1, 3, or 9 h), followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 5, or 15 min). C: Relative glucose uptake in 3T3-L1 adipocytes pretreated with or without SDF-1 (1,000 ng/mL) in serum-free medium for 18 h, followed by treatment with insulin (0, 0.1, or 1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 3–4). D: SDF-1 gene expression in 3T3-L1–tet–SDF-1 adipocytes treated with (+) or without (−) 30 μg/mL doxycycline (Dox) for 48 h (n = 5, 4). E: IRS-1 protein and Akt phosphorylation levels in 3T3-L1–tet–SDF-1 adipocytes pretreated with or without 30 μg/mL of Dox for 48 h, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, 30, or 60 min). F: Relative glucose uptake in 3T3-L1–tet–SDF-1 adipocytes pretreated with or without 30 μg/mL Dox for 48 h, incubated with serum-free medium for 18 h, followed by treatment with insulin (0.1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 6). Data are mean ± SEM. *P < 0.05; ***P < 0.001.
The reduction in IRS-1 protein was associated with attenuated insulin-mediated Akt phosphorylation (Fig. 2B and Supplementary Fig. 3D). SDF-1 also inhibited insulin-mediated glucose uptake in 3T3-L1 adipocytes (Fig. 2C). Similarly, doxycycline-induced overexpression of SDF-1 in 3T3-L1 adipocytes (Fig. 2D and Supplementary Fig. 3E) decreased IRS-1 protein levels, insulin-mediated Akt phosphorylation, and glucose uptake (Fig. 2E and F). We generated 3T3-L1 adipocytes that overexpressed IRS-1 (Supplementary Fig. 3F) to examine the importance of IRS-1 expression level on SDF-1–induced insulin desensitization. SDF-1 inhibited insulin-mediated glucose uptake in 3T3-L1 adipocytes, which was partially reversed by overexpression of IRS-1 (Supplementary Fig. 3G). These results clearly show that SDF-1 directly induces insulin desensitization in adipocytes with reduced IRS-1 protein level and its downstream effects.
SDF-1 Induces Insulin Desensitization in Adipocytes via CXCR4/Erk/IRS-1 Axis
We next investigated the molecular mechanism of SDF-1–induced reduction of IRS-1 protein level in adipocytes. With regard to gene expression, SDF-1 did not change the IRS-1 mRNA level (Supplementary Fig. 3H). Previous studies showed that IRS-1 is degraded by proteasome activity in adipocytes (28). Pretreatment with lactacystin, a proteasome inhibitor, significantly restored SDF-1–induced reduction of IRS-1 protein in 3T3-L1 adipocytes (Fig. 3A and Supplementary Fig. 4A and B), suggesting the involvement of SDF-1 in activation of the IRS-1 proteasome degradation pathway. SDF-1 activates Erk signals (29), although this has not been examined in adipocytes. Furthermore, Erk participates in IRS-1 degradation via serine phosphorylation (30). In our experiments, 3T3-L1 adipocytes and mouse primary differentiated adipocytes cultured in the presence of SDF-1 showed marked activation of Erk (Supplementary Fig. 4C), and the effect was dose-dependent (Supplementary Fig. 4D). SDF-1–induced Erk activation reached a peak level at 5 min, followed by a gradual decrease (Supplementary Fig. 4E), and was concomitantly associated with phosphorylation of IRS-1 at Ser636 (Fig. 3B). Inhibition of Erk signal by U0126, an upstream MEK inhibitor, completely blocked SDF-1–induced IRS-1 phosphorylation (Fig. 3C and Supplementary Fig. 4F) and IRS-1 degradation (Fig. 3D) in 3T3-L1 adipocytes and mouse primary differentiated adipocytes. Furthermore, U0126 markedly reversed SDF-1 action on impaired glucose uptake in the adipocytes (Supplementary Fig. 4G). These data show that SDF-1–induced IRS-1 degradation in adipocytes is Erk signal dependent.
SDF-1 degrades IRS-1 protein via CXCR4/Erk signaling in adipocytes. A: IRS-1 protein levels in 3T3-L1 adipocytes pretreated with lactacyctin (Lacta) (250 nmol/L) for 30 min, followed by treatment with SDF-1 (1,000 ng/mL) for 18 h. B: Erk1/2 and IRS-1 (Ser636) phosphorylation in 3T3-L1 adipocytes treated with SDF-1 (500 ng/mL) for the indicated intervals (0, 5, 15, 30, or 60 min). C: Erk1/2 and IRS-1 (Ser636) phosphorylation in 3T3-L1 adipocytes pretreated with U0126 (10 μmol/L) for 30 min, followed by treatment with SDF-1 (500 ng/mL) for the indicated intervals (0, 5, or 15 min). D: IRS-1 protein levels in 3T3-L1 adipocytes pretreated with U0126 (250 nmol/L) for 30 min, followed by treatment with SDF-1 (1,000 ng/mL) for 18 h. E: Erk1/2 and IRS-1 (S636) phosphorylation levels in mouse primary differentiated adipocytes pretreated with the vehicle (Control) or TC14012 (200 μg/mL) for 30 min, followed by treatment with SDF-1 (500 ng/mL) for the indicated intervals (0, 10, or 20 min). F: IRS-1 protein levels in mouse primary differentiated adipocytes pretreated with vehicle (Control) or TC14012 (TC) (200 μg/mL) for 30 min, followed by treatment with the vehicle or SDF-1 (500 ng/mL) for 12 h. G: Relative glucose uptake in mouse primary differentiated adipocytes pretreated with vehicle (Control) or TC14012 (200 μg/mL) for 30 min, treated with the vehicle or SDF-1 (1,000 ng/mL) for 12 h, followed by treatment with insulin (0.1 nmol/L) for 20 min (n = 3). Data are mean ± SEM. **P < 0.01.
SDF-1 degrades IRS-1 protein via CXCR4/Erk signaling in adipocytes. A: IRS-1 protein levels in 3T3-L1 adipocytes pretreated with lactacyctin (Lacta) (250 nmol/L) for 30 min, followed by treatment with SDF-1 (1,000 ng/mL) for 18 h. B: Erk1/2 and IRS-1 (Ser636) phosphorylation in 3T3-L1 adipocytes treated with SDF-1 (500 ng/mL) for the indicated intervals (0, 5, 15, 30, or 60 min). C: Erk1/2 and IRS-1 (Ser636) phosphorylation in 3T3-L1 adipocytes pretreated with U0126 (10 μmol/L) for 30 min, followed by treatment with SDF-1 (500 ng/mL) for the indicated intervals (0, 5, or 15 min). D: IRS-1 protein levels in 3T3-L1 adipocytes pretreated with U0126 (250 nmol/L) for 30 min, followed by treatment with SDF-1 (1,000 ng/mL) for 18 h. E: Erk1/2 and IRS-1 (S636) phosphorylation levels in mouse primary differentiated adipocytes pretreated with the vehicle (Control) or TC14012 (200 μg/mL) for 30 min, followed by treatment with SDF-1 (500 ng/mL) for the indicated intervals (0, 10, or 20 min). F: IRS-1 protein levels in mouse primary differentiated adipocytes pretreated with vehicle (Control) or TC14012 (TC) (200 μg/mL) for 30 min, followed by treatment with the vehicle or SDF-1 (500 ng/mL) for 12 h. G: Relative glucose uptake in mouse primary differentiated adipocytes pretreated with vehicle (Control) or TC14012 (200 μg/mL) for 30 min, treated with the vehicle or SDF-1 (1,000 ng/mL) for 12 h, followed by treatment with insulin (0.1 nmol/L) for 20 min (n = 3). Data are mean ± SEM. **P < 0.01.
SDF-1–induced Erk activation and its biological functions are mediated mainly by CXCR4, a G protein-coupled receptor (GPCR) that uses the Gαi subunit for its signal transduction (20,31,32). Blockade of the Gαi subunit by pertussis toxin markedly reduced SDF-1–induced Erk activation in adipocytes (Supplementary Fig. 4H). TC14012, a peptidomimetic CXCR4 antagonist (33), inhibited SDF-1–induced Erk activation, but not the epidermal growth factor-induced one, in C2C12 myocytes (Supplementary Fig. 4I), as reported previously (34). Under similar conditions, TC14012 significantly inhibited SDF-1–induced phosphorylation of Erk1/2 and IRS-1 at Ser636 and downregulation of IRS-1 in mouse primary differentiated adipocytes, without changing IRS-1 mRNA levels (Fig. 3E and F and Supplementary Fig. 4J). Also, TC14012 markedly reversed SDF-1 action on impaired glucose uptake in the adipocytes (Fig. 3G). CXCR7, another receptor for SDF-1, does not use the Gαi subunit for its signal transduction, but rather β-arrestin (35). Knockdown of CXCR7 did not affect the SDF-1–induced Erk activation in 3T3-L1 adipocytes and mouse primary differentiated adipocytes (Supplementary Fig. 4K–M). These results indicate that CXCR4 mainly mediates the insulin-desensitizing action of SDF-1 in adipocytes.
Autocrine SDF-1 Action Controls Insulin Sensitivity in Adipocytes
Based on the above findings, we hypothesized that adipocyte-derived SDF-1 acts in an autocrine manner to control adipocyte insulin sensitivity. To confirm this hypothesis, we estimated the autocrine action of SDF-1 in adipocytes in vitro. Knockdown of SDF-1 resulted in a marked increase in IRS-1 protein level, insulin-mediated Akt phosphorylation, and glucose uptake in 3T3-L1 adipocytes (Fig. 4A–C and Supplementary Fig. 4N) and mouse primary differentiated adipocytes (Fig. 4D and E), confirming the autocrine action of SDF-1. Moreover, blockade of its receptor CXCR4 with TC14012 alone, without exogenous SDF-1 treatment, augmented IRS-1 protein level and insulin-mediated Akt phosphorylation in mouse primary differentiated adipocytes (Fig. 4F). These results confirm that adipocyte-derived SDF-1 controls adipocyte-insulin sensitivity in an autocrine manner.
Autocrine SDF-1 action downregulates insulin sensitivity in adipocytes. A: SDF-1 concentration in conditioned medium of 3T3-L1 adipocytes transfected with control siRNA or SDF-1 siRNA (n = 3). B: IRS-1 protein and Akt phosphorylation levels in 3T3-L1 adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 5, 15, 30, or 60 min). C: Relative glucose uptake in 3T3-L1 adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (0.1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 6). D: SDF-1 concentration in conditioned medium of mouse primary differentiated adipocytes transfected with control siRNA or SDF-1 siRNA (n = 3). E: IRS-1 protein and Akt phosphorylation levels in mouse primary differentiated adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, 30, or 60 min). F: IRS-1 protein and Akt phosphorylation levels in mouse primary differentiated adipocytes pretreated with the vehicle (Control) or TC14012 (200 μg/mL) for 12 h, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, or 30 min). Data are mean ± SEM. *P < 0.05; ***P < 0.001.
Autocrine SDF-1 action downregulates insulin sensitivity in adipocytes. A: SDF-1 concentration in conditioned medium of 3T3-L1 adipocytes transfected with control siRNA or SDF-1 siRNA (n = 3). B: IRS-1 protein and Akt phosphorylation levels in 3T3-L1 adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 5, 15, 30, or 60 min). C: Relative glucose uptake in 3T3-L1 adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (0.1 nmol/L) for 20 min. 2-DG uptake for 10 min was measured (n = 6). D: SDF-1 concentration in conditioned medium of mouse primary differentiated adipocytes transfected with control siRNA or SDF-1 siRNA (n = 3). E: IRS-1 protein and Akt phosphorylation levels in mouse primary differentiated adipocytes transfected with control or SDF-1 siRNA, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, 30, or 60 min). F: IRS-1 protein and Akt phosphorylation levels in mouse primary differentiated adipocytes pretreated with the vehicle (Control) or TC14012 (200 μg/mL) for 12 h, followed by treatment with insulin (1 nmol/L) for the indicated intervals (0, 10, or 30 min). Data are mean ± SEM. *P < 0.05; ***P < 0.001.
Adipocyte-Specific SDF-1 Ablation Enhances Insulin Sensitivity in Adipose Tissue
To access the effect of SDF-1 on insulin sensitivity in vivo, we generated AdSDF-1 KO mice by crossing SDF-1 flox/flox mice (14) with adiponectin-Cre mice (36). In contrast to embryonic lethality of whole-body KO of SDF-1, AdSDF-1 KO mice were born at the expected Mendelian ratio and appeared grossly normal, with no apparent differences in body weight (Supplementary Fig. 5A), organ weight (Supplementary Fig. 5B), and food intake (Supplementary Fig. 5C) compared with control flox/flox mice. In AdSDF-1 KO mice, Cre-mediated recombination occurred specifically in WAT and BAT, but not in other tissues such as the liver or muscle (Supplementary Fig. 6A). Fractionation data showed specific SDF-1 gene ablation in mature adipocytes of KO mice but not in the SVF (Supplementary Fig. 6B and C).
We also evaluated the effects of adipocyte-specific SDF-1 ablation on whole-body insulin sensitivity and glucose metabolism. In mice fed the normal diet, fasting insulin levels were significantly lower in AdSDF-1 KO mice than in the control, with normal fasting glucose levels (Fig. 5A and B). Systemic insulin sensitivity (Fig. 5C) and glucose tolerance (Fig. 5D) were also better in AdSDF-1 KO mice than the control mice. These data suggest that adipose SDF-1 contributes to physiological insulin desensitization in adipocytes as well as in the whole body.
SDF-1 deletion in adipocytes enhances insulin sensitivity and glucose homeostasis. Metabolic analysis of AdSDF-1 KO mice maintained on a normal diet (A–E) or HFD (F–J). Fasting glucose (n = 5) (A) and fasting insulin (n = 10) (B) levels of flox/flox and AdSDF-1 KO mice at 10 weeks of age. Results of ITT at 12 weeks of age (C) and GTT at 14 weeks of age (D) of flox/flox and AdSDF-1 KO mice (n = 6). E: IRS-1 protein and insulin-mediated Akt phosphorylation levels in epididymal WAT. After a 5-h fast, control (flox/flox) and AdSDF-1 KO (SDF-1 KO) mice fed the normal diet were injected intraperitoneally with the vehicle (Insulin−) or insulin (10 units/kg BW) (Insulin+). The samples were collected for analysis after 15 min. A diet-induced obesity mouse model was established by feeding an HFD containing 60% calories from fat for 8–12 weeks starting at 5 weeks of age. Fasting glucose (n = 5, 4) (F) and fasting insulin (n = 12) (G) levels of flox/flox and AdSDF-1 KO mice fed the HFD at 13 weeks of age. ITT at 15 weeks of age (H) and GTT at 17 weeks of age (I) of flox/flox and AdSDF-1 KO mice (n = 8). J: IRS-1 protein and insulin-mediated Akt phosphorylation levels in epididymal WAT. After a 5-h fast, control (flox/flox) and AdSDF-1 KO (SDF-1 KO) fed the HFD were injected intraperitoneally with the vehicle (Insulin−) or insulin (10 units/kg BW) (Insulin+).The samples were collected for analysis after 15 min. Data are mean ± SEM. *P < 0.05; **P < 0.01.
SDF-1 deletion in adipocytes enhances insulin sensitivity and glucose homeostasis. Metabolic analysis of AdSDF-1 KO mice maintained on a normal diet (A–E) or HFD (F–J). Fasting glucose (n = 5) (A) and fasting insulin (n = 10) (B) levels of flox/flox and AdSDF-1 KO mice at 10 weeks of age. Results of ITT at 12 weeks of age (C) and GTT at 14 weeks of age (D) of flox/flox and AdSDF-1 KO mice (n = 6). E: IRS-1 protein and insulin-mediated Akt phosphorylation levels in epididymal WAT. After a 5-h fast, control (flox/flox) and AdSDF-1 KO (SDF-1 KO) mice fed the normal diet were injected intraperitoneally with the vehicle (Insulin−) or insulin (10 units/kg BW) (Insulin+). The samples were collected for analysis after 15 min. A diet-induced obesity mouse model was established by feeding an HFD containing 60% calories from fat for 8–12 weeks starting at 5 weeks of age. Fasting glucose (n = 5, 4) (F) and fasting insulin (n = 12) (G) levels of flox/flox and AdSDF-1 KO mice fed the HFD at 13 weeks of age. ITT at 15 weeks of age (H) and GTT at 17 weeks of age (I) of flox/flox and AdSDF-1 KO mice (n = 8). J: IRS-1 protein and insulin-mediated Akt phosphorylation levels in epididymal WAT. After a 5-h fast, control (flox/flox) and AdSDF-1 KO (SDF-1 KO) fed the HFD were injected intraperitoneally with the vehicle (Insulin−) or insulin (10 units/kg BW) (Insulin+).The samples were collected for analysis after 15 min. Data are mean ± SEM. *P < 0.05; **P < 0.01.
We also assessed the insulin-desensitizing action of SDF-1 in vivo adipose tissue. AdSDF-1 KO mice showed specific enhancement of insulin sensitivity in epididymal WAT and BAT, with a marked increase in IRS-1 protein level and insulin-mediated Akt phosphorylation (Fig. 5E and Supplementary Fig. 7A, B, and D). Furthermore, insulin-mediated glucose uptake was enhanced in epididymal WAT from AdSDF-1 KO mice compared with control mice (Supplementary Fig. 7C). These differences were not observed in other insulin-target organs, such as the liver and muscle (Supplementary Fig. 7E and F), again confirming the autocrine action of SDF-1 in adipose tissue.
Because SDF-1 was also increased in obese adipose tissue, another insulin desensitizing tissue, we analyzed the effect of lack of adipocyte SDF-1 in the HFD-induced obese insulin resistance condition. AdSDF-1 KO mice maintained on the HFD exhibited slightly lower body and organ weights, including adipose tissue and liver weights, than the control mice, without changes in food intake (Supplementary Fig. 5D–F). In addition, the AdSDF-1 KO mice had lower fasting glucose and insulin levels and improved insulin sensitivity and glucose tolerance compared with the control mice (Fig. 5F–I). AdSDF-1 KO mice showed specific enhancement of insulin sensitivity in adipose tissue, with marked elevation of IRS-1 protein level and insulin-mediated Akt phosphorylation under the HFD condition (Fig. 5J). These data clearly demonstrated that adipocyte SDF-1 attenuated systemic insulin sensitivity in obese conditions as well.
Lastly, we assessed the function of SDF-1 as a chemoattractant factor in AdSDF-1 KO mice fed the normal chow diet. Immunostaining of macrophages from AdSDF-1 KO mice showed crown-like structures in epididymal WAT, similar to the control mice (Supplementary Fig. 8A and B). Also, quantitative PCR indicated that SDF-1 ablation did not change immune cell marker genes in epididymal WAT (Supplementary Fig. 8C). Finally, we quantitatively profiled SVF components in epididymal WAT by FACS analysis. SDF-1 knockout did not alter the proportion or number of preadipocytes (CD45−CD31−PDGFRα+) (37), vascular endothelial cells (CD45−CD31+), or hematopoietic cells (CD45+) in epididymal WAT (Supplementary Fig. 8D–F). In addition, there were no differences in the proportion and number of adipose-immune cells between AdSDF-1 KO and control mice (Supplementary Fig. 8G and H). These results clearly show that adipocyte-derived SDF-1 does not affect immune cell profile in adipose tissue and rule out the involvement of adipose-immune cells in the observed enhanced insulin sensitivity in AdSDF-1 KO mice, at least under the chow diet condition.
Discussion
In the current study, we identified SDF-1 as an autocrine insulin-desensitizing factor in adipocytes. Fasting and obesity both induced SDF-1 expression in adipocytes. Its autocrine action activated Erk signaling, which concomitantly induced serine phosphorylation of IRS-1 protein, degraded IRS-1 protein in adipocytes, and attenuated insulin-mediated Akt phosphorylation and glucose uptake (Fig. 6).
Schematic diagram of autocrine SDF-1 as an insulin-desensitizing factor in adipocytes. 1) Fasting or obesity induces SDF-1 expression in adipocytes. 2) Its autocrine action activates Erk signaling. 3) SDF-1–induced Erk signal concomitantly induces serine phosphorylation of IRS-1 protein and degrades IRS-1 protein. This attenuates 4) insulin-mediated Akt phosphorylation and 5) glucose uptake.
Schematic diagram of autocrine SDF-1 as an insulin-desensitizing factor in adipocytes. 1) Fasting or obesity induces SDF-1 expression in adipocytes. 2) Its autocrine action activates Erk signaling. 3) SDF-1–induced Erk signal concomitantly induces serine phosphorylation of IRS-1 protein and degrades IRS-1 protein. This attenuates 4) insulin-mediated Akt phosphorylation and 5) glucose uptake.
SDF-1 is a ubiquitously expressed and highly conserved secretary factor with 99% homology between human and mouse. It plays important roles in development, tissue regeneration, hematopoiesis, immunity, and carcinogenesis (20). SDF-1 contributes to the pathogenesis, progression, and diverse pathological effects of type 2 diabetes, such as insulitis, nephropathy, and adipose tissue inflammation (22,38,39). Furthermore, plasma SDF-1 levels correlated with the type 2 diabetes disease state (40). However, the lethal phenotype of global SDF-1 KO mice has made it difficult to assess the metabolic functions of SDF-1 in adult mice (20,41). In the current study, we successfully generated viable AdSDF-1 KO mice with an insulin-sensitive phenotype.
Insulin plays a key role in maintaining glucose homeostasis. Its action is regulated by endogenous signaling molecules and exogenous counterregulatory factors in fasting and obese conditions. Fasting induces counterregulatory hormones, such as catecholamines, cortisol, and growth hormone, and reduces insulin signaling in the target organs, including liver, skeletal muscle, and adipose tissue (11–13). In the development of obesity, infiltrating macrophages secrete proinflammatory cytokines, such as TNF-α, and induce inflammation and insulin resistance in adipose tissues (3,7,42). In this study, we found SDF-1 as a common factor responsible for fasting- and obesity-related insulin desensitization in adipose tissue. To our knowledge, this is the first evidence that insulin target cells, in this case adipocytes, secrete an autocrine factor that desensitizes insulin action.
Liver and skeletal muscle are also important insulin target organs regulating glucose metabolism. IRS-1 is responsible for insulin-mediated suppression of gluconeogenesis in the liver (43) and insulin-mediated glucose uptake in the skeletal muscle (44). In addition, the liver and skeletal muscle express SDF-1 and its receptor CXCR4 (45,46), and SDF-1 induced Erk phosphorylation in H4CII hepatocytes (data not shown) and C2C12 myocytes (Supplementary Fig. 4F), suggesting the possible role of SDF-1 in insulin signaling in these tissues as well. However, regulation of SDF-1 expression was different among these tissues. SDF-1 expression was augmented in adipose tissues under both insulin-desensitized conditions in fasting and obesity. In contrast, its expression showed no difference in the liver and was reduced in the skeletal muscle in the fasting condition (data not shown). Furthermore, it was decreased in liver and exhibited no difference in skeletal muscle of ob/ob mice compared with control mice (data not shown). Based on these results, we suspect that SDF-1 has a specific role in adipocytes as a common factor responsible for both obesity- and fasting-mediated insulin desensitization.
Previous studies suggested that SDF-1 acts as a pathological chemoattractant factor in obese adipose tissues (21,22). Treatment with AMD3100, a potent antagonist of CXCR4, reduced macrophage infiltration and inflammation in adipose tissues under the HFD condition (22). The study concluded that systemic inhibition of SDF-1 signaling ameliorated macrophage infiltration into adipose tissues, resulting in improvement of whole-body insulin resistance (22). In contrast, the current study showed that AdSDF-1 KO mice exhibited adipose tissue-specific enhancement of insulin sensitivity without changes in adipose-immune cell profiles, which clearly differed from the chemokine/inflammation/insulin resistance scenario in the adipose tissue. One possible explanation for this point is the distinct function of SDF-1 according to the cell type. In addition to adipocytes, adipose tissue contains the SVF, including endothelial cells, preadipocytes, and immune cells. Considering that SVF also expresses SDF-1 (GDS2428) (47), SDF-1 from adipocytes acts on adipocytes, and SDF-1 from other cell types in SVF should be related to the chemotactic property on macrophages.
SDF-1 has received little attention in the metabolic and physiological research fields. Our findings, however, shed new light on its function as a crucial autocrine insulin-desensitizing factor in adipocytes.
Article Information
Acknowledgments. The authors thank all members of the Shimomura Laboratory, Interdisciplinary Program for Biomedical Sciences (IPBS), T. Nagasawa and T. Sugiyama (Laboratory of Stem Cell Biology and Developmental Immunology, Graduate School of Frontier Biosciences, Osaka University) for the helpful discussion, E. Rosen (Department of Genetics, Harvard Medical School) and J. Eguchi (Division of Nephrology, Diabetology and Endocrinology, Okayama University) for providing adiponectin-Cre, and J. Zimmermann and T. Hunter (Molecular and Cell Biology Laboratory, Salk Institute for Biological Studies) for providing pcDNA3-mIRS-1 vectors.
Funding. This work was partly supported by the Japan Society for the Promotion of Science Grant-in-aid for Scientific Research (C) (grant no. 17K09829), and the Osaka University Institute for Academic Initiatives. J.S. is supported by a Japanese Government Ministry of Education, Culture, Sports, Science and Technology scholarship.
Duality of Interest. This work was partially supported by Sanofi (grant), AstraZeneca (grant), and Merck Sharp & Dohme (grant). A.F. and S.K. belong to a department endowed by Takeda Pharmaceutical Company; Sanwa Kagaku Kenkyusho Co., Ltd.; Rohto Pharmaceutical Co., Ltd.; Fuji Oil Holdings Inc.; and Roche DC Japan.
The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Author Contributions. J.S. designed and performed the experiments and acquired data. J.S. and A.F. interpreted the data and wrote the manuscript. T.O. performed FACS analysis. S.K., C.Y., M.O., and I.S. supervised the project and wrote the manuscript. A.F. 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 data analysis.