Fractalkine Is a Novel Human Adipochemokine Associated With Type 2 Diabetes

Each clinical study was performed with the approval of the University of Pennsylvania (UPenn) Institutional Review Board after written informed consent was obtained from all research participants.

As previously described (16–18), healthy volunteers aged 18–40 years and with BMI 18–30 kg/m² were recruited. Exclusions included inflammatory disease, pregnancy, medication, substance, or supplement use. Serial whole-blood samples were collected before and 2, 4, 6, 8, 12, 16, and 24 h following intravenous bolus of either 3 ng/kg (moderate dose) or 0.6 ng/kg (low dose) U.S. standard reference endotoxin (lipopolysaccharide [LPS], lot no. CC-RE-LOT-1 + 2; Clinical Center, Pharmacy Department at the National Institutes of Health, Bethesda, MD). Subcutaneous adipose samples were collected by core needle aspiration through a 4-mm gluteal incision from distinct sites 30 min before and 4, 12, and 24 h after LPS and stored at −80°C.

Adipose from lean (n = 11; BMI mean 24.3, 95% CI 22.5–26.1) subjects was obtained by aspiration biopsy as described above and at bariatric surgery for obese patients (n = 13; BMI mean 48.3, 95% CI 43.0–53.7). Samples were stored at −80°C for protein isolation. In a subset of lean subjects (n = 8), a portion of the adipose was minced and digested immediately with type I collagenase (Sigma-Aldrich, St. Louis, MO) for 1 h, centrifuged, and mature adipocytes and stromal vascular fraction (SVF) were isolated.

This analysis in the University of Pennsylvania Coronary Artery Calcification Study (PennCAC) sample focused on subjects of European ancestry recruited from 3 separate parallel community-based studies at UPenn: the Study of Inherited Risk of Coronary Atherosclerosis (SIRCA; n = 799), the Penn Diabetes Heart Study (PDHS; n = 782), and the Philadelphia Area Metabolic Syndrome Network (PAMSyN; n = 480) (19). SIRCA is a cross-sectional study of factors associated with CAC in asymptomatic nondiabetic subjects aged 30–75 years and their families (20,21). PDHS is an ongoing, cross-sectional community-based study of subjects aged 35–75 years with clinical type 2 diabetes (clinical history of type 2 diabetes, use of oral hypoglycemic agents/insulin in a subject older than age 40 years, or fasting blood glucose >126 mg/dL, 2-h postprandial glucose >200 mg/dL, and no evidence of type 1 diabetes) (21,22). PAMSyN is a cross-sectional study of nondiabetic subjects aged 18–75 years with one or more metabolic syndrome risk factors (19). In all studies, known clinical atherosclerotic cardiovascular disease (CVD) and overt kidney disease were excluded. SIRCA, PDHS, and PAMSyN used the same clinical research center, nursing staff, and research laboratories; these details and measurement of biochemical and metabolic parameters have been described elsewhere (19,21,23).

An age- and gender-matched case-control study of type 2 diabetes was selected from study resources at UPenn. The diabetic subjects (n = 281) were selected randomly from the ongoing PDHS described and referenced above. Age- and gender-matched nondiabetic subjects (n = 274) were chosen from PAMSyN, also described above. Subjects in this case-control study did not overlap with the genetic study above. Frozen plasma samples, stored at −80°C, were used to assay CX3CL1 levels.

For the exploration of the cellular source of CX3CL1, adipocyte and SVF of fresh human adipose tissue were prepared for RNA and protein isolation as below. For in vitro experiments, preadipocytes from SVF were isolated, cultured, and differentiated to adipocytes as previously described (16). Briefly, cells were grown to confluence in oxidation fermentation medium with 20% FBS and differentiated in serum-free medium with high insulin and PPARγ agonist (16,24). For assessment of inflammatory induction of CX3CL1, mature adipocytes were washed and treated ± 100 ng/mL LPS for 2, 4, 12, and 24 h (in triplicate for two independent experiments) or cytokines (overnight ± 10 ng/mL tumor necrosis factor-α [TNF-α] and 5 ng/mL interferon-γ [IFN-γ]; R&D Systems, Minneapolis, MN) with and without 24-h pretreatment with PPARγ agonist at 10 μmol/L concentration (GW347845; gift from GlaxoSmithKline, King of Prussia, PA).

As described (16,18), human monocytes were isolated from healthy donor blood following apheresis and elutriation at the UPenn Center for AIDS Research. These cells, as well as the monocytic THP-1 cell line (American Type Culture Collection, Manassas, VA) were grown in RPMI-1640 (Cambrex Bio Science, Verviers, Belgium) supplemented with 10% FBS and 1% penicillin/streptomycin. To promote differentiation to macrophages, cells were then supplemented with 100 ng/mL macrophage-colony stimulating factor (Sigma, St. Louis, MO) for 7 days.

Fully differentiated human adipocytes were washed and placed in Dulbecco’s modified Eagle’s medium/F12 + 0.5% BSA. THP-1 monocytes were washed and resuspended in PBS, labeled with calcein (Invitrogen, Carlsbad, CA) 15 μl/mL for 30 min, and then washed twice before plating 2 × 10⁵ cells per well of adipocytes (also grown at approximately 2 × 10⁵ cells per well). Anti-CX3CL1, anti–MCP-1, both anti-CX3CL1 and anti–MCP-1 (all from R&D Systems), or control antibody (rabbit IgG) were added to each well at a concentration of 20 μg/mL. Cells were incubated together at 37°C for 2 h after which they were washed three times with PBS. Adherent cells were immediately counted under a fluorescent microscope with five random high power fields per well and four wells per condition. The experiment was performed three times, on three different days, using the same THP-1 and human adipocyte lines.

Whole adipose tissue was homogenized or for in vitro experiments, adipocytes and monocytes were washed and lysed for RNA isolation using TRIzol reagent (Invitrogen). RNA (500 ng) was reverse transcribed to cDNA using the High-Capacity cDNA Archive Kit (ABI, Foster City, CA). Expression of genes was determined by quantitative real-time PCR (7900 Real-Time PCR System; ABI) using TaqMan Universal PCR MasterMix and primers and probes from Applied Biosystems. To control for between-sample differences, mRNA levels were normalized to β-actin for each sample by subtracting the Ct for β-actin from the Ct for the gene of interest, producing a ΔCt value. The ΔCt for each posttreatment sample was compared with the mean ΔCt for all pretreatment samples in a single individual or experiment using the relative quantitation 2^-ΔΔCt method to determine fold change from baseline (25).

Whole adipose tissue was placed in radioimmunoprecipitation assay (RIPA) buffer, homogenized, and protein concentration quantified using Pierce bicinchoninic acid kit (Thermo Fisher Scientific, Rockford, IL). Samples then were diluted with additional RIPA buffer to equal concentrations (1.5 μg/μL). Primary adipocytes, stromal vascular cells, and cultured adipocytes were lysed for protein extraction with RIPA buffer × 30 min at 4°C; cell media was collected immediately and frozen. Human CX3CL1 ELISA (R&D Systems; Minneapolis, MN) was performed in duplicate according to manufacturer instructions. If duplicates had >10% coefficient of variation (CV), the sample was repeated. Intra- and interassay CV were 5.7% and 5.3%. The lower and upper limits of detection of the ELISA were 0.156 ng/mL and 10 ng/mL.

For Western blotting, monocytes, macrophages, and adipocytes were lysed and protein quantified as above. Equal concentrations (20 μg) of lysate was reduced, separated by SDS-PAGE electrophoresis, transferred to nitrocellulose membranes, and blocked in 5% milk in TBS-tween prior to overnight incubation (at 4°C) with 1:1000 dilution of CX3CR1 antibody (R&D Systems). Blots were washed in Tris-buffered saline (TBS)-tween, incubated with rabbit anti-goat secondary antibody (BioRad), and visualized by chemiluminescence. The blots were stripped with stripping buffer (Pierce Endogen, Rockford, IL) and reincubated using the same protocol as above with 1:5000 β-actin antibody (Santa Cruz Biotechnology, Santa Cruz, CA).

A total of 2,096 Caucasian PennCAC study participants were genotyped using the HumanCVD BeadChip V2 (Illumina, San Diego, CA), which includes ∼50,000 single nucleotide polymorphisms (SNPs) within ∼2,000 CVD candidate genes (26). Chip genotype calls were generated with the cluster file provided by Illumina BeadStudio software. Individuals with call rates less than 95% were excluded. SNPs with a low call rate (<95%) were excluded. Recently, Sirois-Gagnon et al. (27) showed that two SNPs (rs3732378 and rs3732379) in CX3CR1 were associated with obesity traits in 900 French Canadians. Therefore, we examined the association of rs3732378 and rs3732379 with multiple obesity-related and metabolic traits in PennCAC (19). The call rates for rs3732378 and rs3732379 were 100%. The observed distribution of genotypes did not deviate significantly from Hardy-Weinberg equilibrium.

Continuous variables were compared using Student t test or the Mann-Whitney U test depending on whether normality could be assumed. Fisher exact test was used for proportions. Unless otherwise specified, data are reported as mean ± SD. Geometric means of the independent replicated samples and 95% CIs were used to estimate the fold change in mRNA for the RT-PCR data. To assess differences among time points and between groups, our analysis of CX3CL1 protein levels in adipose and cells used repeated-measures ANOVA with a Bonferroni correction for multiple post hoc comparisons. One-way ANOVA was used to assess differences in monocyte adhesion to adipocytes in adhesion assay.

The rs3732378 and rs3732379 SNPs in CX3CR1 were tested for association in the PennCAC sample using linear regression for quantitative traits and logistic regression for binary phenotypes with genotype included as an independent variable, coded as an additive genotype effect, with age, sex, and smoking included as covariates. The analysis was carried out using PLINK 1.06 (http://pngu.mgh.harvard.edu/purcell/plink/) (28).

The association of CX3CL1 with type 2 diabetes was examined using logistic regression, adjusting for covariates, including medication use and other cardiometabolic risk factors with known or suspected influence on fractalkine levels, as described in RESULTS. Analyses were performed using Stata 10 (StataCorp LP, College Station, TX).

Moderate dose (3 ng/kg) and low dose (0.6 ng/kg) endotoxin increased adipose (Fig. 1A and B) as well as blood (Fig. 1C) CX3CL1 in vivo. At 4 h, adipose tissue CX3CL1 mRNA increased by 33-fold (95% CI 26.8–39.9, P < 1 × 10⁻⁵) with 3 ng/kg LPS and 13-fold (95% CI 4.9–31.6, P < 0.0001) with 0.6 ng/kg LPS (Fig. 1A). Adipose CX3CL1 protein levels increased at 4 h by over 40-fold (0.040 ± 0.127 to 1.733 ± 1.197 ng/mL, P = 0.006) and eightfold (0.152 ± 0.121 to 1.254 ± 0.870 ng/mL, P = 0.004), respectively, with 3 ng/kg and 0.6 ng/kg LPS (Fig. 1B).

Levels of CX3CL1 protein in subcutaneous adipose were significantly higher in obese (n = 13) compared with lean (n = 11) subjects (0.420 ± 0.387 vs. 0.228 ± 0.187 ng/mL, P = 0.04) (Fig. 2A). In adipose of obese subjects undergoing bariatric surgery (n = 13), CX3CL1 levels were higher in visceral than subcutaneous adipose (0.736 ± 0.308 vs. 0.420 ± 0.387 ng/mL, P = 0.01) (Fig. 2B); depot levels were, however, correlated with each other (Spearman r = 0.91).

In fractionated subcutaneous adipose, CX3CL1 mRNA was expressed, as expected (8), in the SVF but was also detected in the adipocyte fraction (Fig. 2C); protein levels were also present in both fractions (Fig. 2D). In contrast, mRNA for CX3CR1, the fractalkine receptor, was expressed only in SVF (ΔCt [cycle threshold] = 6.2 relative to β-actin mRNA) but not in adipocytes (Ct levels >40). CX3CR1 protein was present at Western blotting in human monocytes and macrophages, but was undetectable in adipocytes (Supplementary Fig. 1).

The presence of CX3CL1 in the isolated adipocyte fraction of human adipose prompted us to examine CX3CL1 regulation in human adipocytes in vitro. CX3CL1 mRNA (ΔCt 8.6 relative to β-actin mRNA) and secreted protein (0.566 ± 0.345 ng/mL/24 hr) were present in unstimulated cultured human adipocytes. LPS treatment (100 ng/mL) increased CX3CL1 mRNA (45-fold at 4 h, P < 0.0001) (Fig. 3A) and secreted protein (peak 2.587 ± 0.937 ng/mL, P < 0.01 for each) (Fig. 3B). Combined treatment with TNF-α and IFN-γ also markedly increased CX3CL1 mRNA (27.5-fold at 12 h, P = 0.006) and protein (0.565 ± 0.3445 to 1.767 ± 0.483 ng/mL, P = 0.007). Pretreatment of human adipocytes with a PPAR-γ agonist GW347845 10 μmol/L for 24 h completely attenuated TNF-α/IFN-γ induction of CX3CL1 (Fig. 3C and D).

CX3CL1, through its membrane-anchored form, promotes adhesion and retention of leukocytes to endothelial cells (29), but its effect on adipocyte-monocyte interactions has not been reported. CX3CL1 blocking antibody reduced human monocyte adhesion to human adipocytes by almost 50% (75 ± 37 vs. 121 ± 74 cells/high-powered field [HPF], P < 0.0001) (Fig. 4A and C), similar to the effects of MCP-1 blocking antibody (Fig. 4A and D). Combined antibody blockade of CX3CL1 and MCP-1 showed an additive attenuation of monocyte adhesion to adipocytes (58 ± 26 cells/HFP, P < 0.0001 vs. control, P < 0.005 vs. CX3CL1 antibody alone, and P = 0.04 vs. MCP-1 antibody alone) (Fig. 4A and E).

Two common nonsynonymous coding SNPs in CX3CR1 rs3732378 (T280M) and rs3732379 (V249I), which are in moderate linkage disequilibrium (r² = 0.66 in 1000 Genomes Pilot 1 publically available dataset), are related to several disease traits in human (8) including a recent association with obesity in French Canadians (27). McDermott et al. (30) showed that rs3732378 (T280M) has functional impact on human monocyte migration, adhesion, and activation.

Demographic and metabolic characteristics of the PennCAC sample are presented in Supplementary Table 1. Consistent with the findings of Sirois-Gagnon et al., individuals in PennCAC with the A allele at rs3732378 or rs3732379 tended to have greater waist circumference and higher levels of homeostasis model assessment of insulin resistance and leptin but lower levels of adiponectin (Table 1). Further, these alleles tended toward association with both NCEP-defined (31) metabolic syndrome and type 2 diabetes.

In order to further explore the association of CX3CL1 with a clinically important obesity-related disease outcome, we performed a proof-of-concept case-control study of plasma CX3CL1 in type 2 diabetes. Supplementary Table 2 presents baseline clinical characteristics for the age- and sex-matched case-control sample. As expected, there were differences in several clinical and biochemical measures as well as medication use between patients with diabetes and nondiabetics. Diabetic subjects had greater use of statins and ACE-inhibitors and were using insulin, metformin, and thiazolodinediones. Levels of CX3CL1 stratified by demographic, metabolic parameters, and medications are provided in Supplementary Table 3. There were no differences in CX3CL1 levels among these parameters or across strata of diabetic medications (Supplementary Table 4).

Plasma CX3CL1 was significantly higher in patients with diabetes than in nondiabetics (0.506 ± 0.262 ng/mL vs. 0.422 ± 0.210 ng/mL, P < 0.0001) (Fig. 5). In unadjusted analysis, there was more than a twofold increased odds ratio for diabetes for a 1 SD (0.21 ng/mL) increase in plasma CX3CL1 concentration; this remained significant after adjusting for BMI, ethnicity, and multiple other metabolic risk factors (Table 2).

We have identified CX3CL1 as a novel adipose tissue chemokine in humans. First, adipose CX3CL1 is increased in obesity as well as in evoked adipose inflammation. Second CX3CL1, which is secreted by adipocytes and adipose SVF, promotes monocyte adhesion to human adipocytes. Third, the association of genetic variation in CX3CR1 with multiple metabolic traits and the finding of higher levels of CX3CL1 in type 2 diabetes provide support for clinical relevance. Thus, CX3CL1 may modulate recruitment of leukocytes in clinical syndromes of adipose inflammation and represent a novel therapeutic target for obesity and type 2 diabetes.

Several inflammatory signaling pathways have been implicated in adipose dysfunction and insulin resistance. In particular, adipose recruitment and activation of macrophages and T-cells play a proximal role in obesity and insulin resistance (2,3,32). However, the specific signals that initiate recruitment of leukocytes to adipose and their activation are unclear, particularly in humans.

We discovered CX3CL1 as a novel adipochemokine induced in human adipose using experimental endotoxemia. Although an acute model, endotoxemia-induced metabolic perturbations (17,18,33,34) are of specific relevance to chronic changes in diet-induced obesity, insulin resistance, and metabolic syndrome (16,35). For example, endotoxemia induces insulin resistance in humans (17,34) with adipose inflammation preceding loss of insulin sensitivity (17). Further, ablation of Toll-like receptor-4 (TLR4), the receptor for endotoxin, in high-fat–diet fed mice attenuates obesity and improves metabolic traits (36–38). Recent work demonstrates that certain dietary and adipocyte fatty acids are endogenous TLR4 ligands (36) and high-fat diets increase plasma LPS and mononuclear cell TLR signaling (39).

Inflammatory induction of CX3CL1 in adipose, therefore, raises the possibility of its involvement in the pathogenesis of insulin resistance and type 2 diabetes. Indeed, emerging data implicates CX3CL1 as a unique chemokine in human disease. The soluble form of CX3CL1, cleaved from the membrane by the enzyme ADAM17 or TACE (TNF-α converting enzyme), may be involved in chemotaxis, while the membrane-anchored form promotes adhesion and retention of leukocytes (40). CX3CL1 modulates monocyte and T-cell recruitment and adherence to endothelium in several chronic inflammatory diseases including atherosclerosis in both rodents (15,41) and humans (30,42). Experimental studies also suggest that therapeutic blockade of CX3CL1-CX3CR1 may produce fewer side effects than other chemokines because CX3CL1-CX3CR1 play a limited role in homeostasis or acute inflammation (43,44).

Whether CX3CL1-CX3CR1 modulates inflammatory monocyte recruitment to adipose is unknown. Recent work, however, is suggestive; adipose tissue macrophages in obese adipose express high levels of CX3CR1 (45). We demonstrate that obese human adipose secreted greater amounts of CX3CL1 than lean adipose. Compared with subcutaneous fat, visceral adipose, known to have greater inflammatory leukocyte infiltration (46), had higher CX3CL1 levels. Consistent with literature demonstrating the endothelium as a source of CX3CL1 (12), the SVF of adipose secreted CX3CL1. The adipocyte fraction, however, also expressed CX3CL1 with marked induction by adipocytokines (TNF-α and IFN-γ), which are upregulated and implicated in obesity. Given the fundamental role of PPAR-γ in adipocyte differentiation and healthy function, suppression of this phenomenon by a PPAR-γ agonist is consistent with a role of CX3CL1 in early adipose inflammation and adipocyte dysregulation in obesity and insulin resistance. Thus, both adipocytes and SVFs may secrete CX3CL1 and modulate leukocyte recruitment to adipose in obesity.

Like others (34), we found that CX3CR1 is highly expressed on primary human monocytes and THP-1 monocytes. Using blocking antibodies, we found that THP-1 monocyte adherence to human adipocytes is CX3CL1-CX3CR1–dependent. Several groups recently showed independent and additive effects of CX3CL1-CX3CR1 and CCR2 in atherosclerosis (12,47). Remarkably, our preliminary data also show additive effects of CX3CR1 and CCR2 pathways in monocyte-adipocyte interactions raising the possibility that these chemokines may have independent, nonredundant actions in leukocyte recruitment to adipose in obesity. Whether this is true in the pathogenesis of insulin resistance and type 2 diabetes in vivo remains to be established.

In humans, variation in CX3CR1 has been related to monocyte chemotactic function (30) and atherosclerotic CVD (48). Recently, Sirois-Gagnon et al. found the association of SNPs in CX3CR1 with obesity. We replicated and extended these findings by showing the association of these SNPs with multiple metabolic traits including metabolic syndrome and type 2 diabetes. Finally, our case-control study found higher CX3CL1 levels in type 2 diabetic subjects compared with control subjects. These data extend our experimental findings by suggesting potential associations of CX3CL1-CX3CR1 with clinically relevant metabolic traits and disease.

CX3CL1-CX3CR1 is a novel adipochemokine system in humans. Its expression in obese adipose, inflammatory upregulation in adipose and adipocytes, modulation of monocyte-adipocyte interactions, and association with human disease traits suggest a role in clinical syndromes of adipose inflammation and its potential as a novel therapeutic target in obesity and type 2 diabetes. Future studies in rodents with gene deficiency and humans using pharmacological modulation are required to define the causal role of CX3CL1-CX3CR1 in adipose inflammation, insulin resistance, and type 2 diabetes.

This work was supported by a Clinical and Translational Science Award (UL1RR024134) from the National Center for Research Resources and a Diabetes and Endocrine Research Center award (P20-DK 019525), both to UPenn. M.P.R. is also supported by RO1 HL-073278 and P50 HL-083799-SCCOR from the National Institutes of Health (NIH). R.S. is supported by the UPenn Clinical and Translational Science Award K12 KL1RR024132 from NIH. N.N.M. is supported by 5K23HL097151 from NIH.

No potential conflicts of interest relevant to this article were reported.

R.S. researched data and wrote the manuscript. C.C.H. researched data and contributed to discussion. J.F.F. researched data and reviewed and edited the manuscript. N.N.M. researched data. M.L. researched data. L.Q. researched data. Y.L. researched data. M.E.P. researched data and reviewed and edited the manuscript. R.S.A. contributed to discussion and reviewed and edited the manuscript. M.P.R. researched data, contributed to discussion, and reviewed and edited the manuscript.