Fatty acid–binding proteins (FABPs) 4 and 5 play coordinated roles in rodent models of inflammation, insulin resistance, and atherosclerosis, but little is known of their role in human disease. The aim of this study was to examine the hypothesis that plasma adipocyte and macrophage FABP4 and FABP5 levels would provide additive value in the association with metabolic and inflammatory risk factors for cardiovascular disease as well as subclinical atherosclerosis. Using the Penn Diabetes Heart Study (PDHS; n = 806), cross-sectional analysis of FABP4 and FABP5 levels with metabolic and inflammatory parameters and with coronary artery calcium, a measure of subclinical coronary atherosclerosis, was performed. FABP4 and FABP5 levels had strong independent associations with the metabolic syndrome (for a 1-SD change in FABP levels, odds ratio [OR] 1.85, 95% confidence interval [CI] 1.43 to 2.23, and OR 1.66, 95% CI 1.41 to 1.95, respectively) but had differential associations with metabolic syndrome components. FABP4 and FABP5 were also independently associated with C-reactive protein and interleukin-6 levels. FABP4 (OR 1.26, 95% CI 1.05 to 1.52) but not FABP5 (OR 1.13, 95% CI 0.97 to 1.32) was associated with the presence of coronary artery calcium. An integrated score combining FABP4 and FABP5 quartile data had even stronger associations with the metabolic syndrome, C-reactive protein, interleukin-6, and coronary artery calcium compared to either FABP alone. In conclusion, this study provides evidence for an additive relation of FABP4 and FABP5 with the metabolic syndrome, inflammatory cardiovascular disease risk factors, and coronary atherosclerosis in type 2 diabetes mellitus. These findings suggest that FABP4 and FABP5 may represent mediators of and biomarkers for metabolic and cardiovascular disease in type 2 diabetes mellitus.
Although fatty acid–binding proteins (FABPs) 4 and 5 have been shown to be important in mouse studies of inflammation, insulin resistance, and atherosclerosis, only limited information is available for their role in human disease. In Chinese population-based studies, circulating FABP4 predicted incident metabolic syndrome and type 2 diabetes mellitus (T2DM), independently of adiposity and insulin resistance. Aside from the Nurses’ Health Study and the Health Professionals Follow-Up Study, in which a genetic variant of FABP4 was associated with metabolic dyslipidemia, T2DM, and coronary artery disease, data are lacking for plasma FABP4’s relation to clinical atherosclerotic cardiovascular disease (CVD). FABP5’s role in human disease remains to be studied. Furthermore, a cooperative role of FABP4 and FABP5 in metabolic perturbation, inflammation, and atherosclerosis in humans has not been studied. In this study, we examined the individual relations of plasma levels of FABP4 and FABP5, and their combined association, with metabolic and inflammatory parameters as well as with coronary artery calcium (CAC), a measure of subclinical coronary atherosclerosis, using the Penn Diabetes Heart Study (PDHS) cohort of adult patients with T2DM.
Methods
Details of the PDHS have been reported previously. In brief, the PDHS is an ongoing single-center, cross-sectional, community-based study of subjects with T2DM without clinical evidence of CVD (defined as myocardial infarction, coronary revascularization, angiographic disease, or positive stress test results) or overt chronic kidney disease recruited at the University of Pennsylvania. Inclusion criteria were age 35 to 75 years, diagnosis of T2DM, and negative pregnancy test result if female. Exclusions were clinical CVD, clinical diagnosis of type 1 diabetes mellitus (insulin use before 35 years of age), serum creatinine >2.5 mg/dl, and weight >300 lb (136 kg). In this report, we studied 806 patients with T2DM in whom FABP4 and FABP5 were measured.
Participants were evaluated at the Clinical and Translational Research Center at the University of Pennsylvania Medical Center after we obtained informed consent from patients and approval from our institutional review board. After a 12-hour overnight fast, clinical parameters, including blood pressure, waist circumference, and laboratory values, were assessed as previously reported. Plasma lipids were measured enzymatically (Hitachi 912 AutoAnalyzer; Roche Diagnostics GmbH, Basel, Switzerland) in lipoprotein fractions after ultracentrifugation (β-quantification technique) at the University of Pennsylvania’s Centers for Disease Control and Prevention–certified lipid laboratory. Framingham risk scores, using total cholesterol, were calculated as described by Wilson et al. Subjects were classified as having the metabolic syndrome using the definition of the National Cholesterol Education Program. All subjects were classified as having National Cholesterol Education Program metabolic syndrome glucose criteria. Global CAC scores were quantified as described according to the method of Agatston et al using electron-beam tomography.
Plasma levels of FABP4 and FABP5 were measured by enzyme-linked immunosorbent assay according to the manufacturer’s instructions (Biovendor Laboratory Medicine, Prague, Czech Republic). All samples were assayed in duplicate, and pooled human plasma samples were included to assess variability; intra- and interassay coefficients of variation were 5.6% and 14% for FABP4 and 8.8% and 17% for FABP5. Plasma levels of adiponectin and leptin (Linco, St. Charles, Missouri), as well as interleukin-6 (IL-6; ultrasensitive; R&D Systems, Minneapolis, Minnesota), were measured by enzyme-linked immunosorbent assay, and high-sensitivity C-reactive protein (CRP) was measured by immunoturbidimetric assay (Wako, Ltd., Osaka, Japan). The intra- and interassay coefficients of variation for pooled human plasma were 5.65% and 9.9% for adiponectin, 5.5% and 12.4% for leptin, 8.7% and 10.9% for IL-6, and 8.0% and 8.3% for CRP. Laboratory test results were generated by personnel blinded to the clinical characteristics and CAC scores of research subjects.
Data are reported as median (interquartile range) or as mean ± SD for continuous variables and as proportions for categorical variables. The crude association of FABP data with quantitative lipid, metabolic, and inflammatory parameters was examined using Spearman’s correlation analysis, and the Kruskal-Wallis test was used to test for associations of the FABPs with the metabolic syndrome and its binary components. To assess the combined effect of FABP4 and FABP5, we generated gender-specific quartiles of FABP4 (coded from 0 to 3) and FABP5 (coded from 0 to 3), and combined these quartile data in an additive fashion into an ordinal 7-level score (0 to 6) termed the FABP score. Multivariate associations of FABP data (effect of 1 SD of the log of FABP4, FABP5, or the FABP score) with the metabolic syndrome and log-transformed CRP data were examined in logistic and linear regression models, respectively, with incremental adjustment for confounding risk factors. For CAC scores, we performed logistic regression of the presence of CAC (score >0), because this cut point has been shown to predict CVD events, as well as exploratory analysis using tobit conditional regression of natural log (CAC + 1), which is suited to analysis of the many zero scores but marked right skew of CAC data. Data are presented as a combined sample of men and women because there was no interaction by gender on each of the outcomes studied. Statistical analyses were performed using Stata version 10.0 (StataCorp LP, College Station, Texas).
Results
Table 1 summarizes study sample characteristics. As expected, obesity was seen in about 1/3 of men and women, and the National Cholesterol Education Program–defined metabolic syndrome was present in >2/3. Plasma FABP levels were higher in women than in men. As expected, men had higher CAC scores than women, and CAC prevalence was consistent with greater subclinical atherosclerosis in asymptomatic T2DM : >1/3 of men and almost 30% of women had CAC scores >75th percentile of age and gender adjusted in the general population.
| Variable | Women (n = 313) | Men (n = 493) |
|---|---|---|
| Age (years) | 58 (52–65) | 60 (54–68) |
| White | 50.5% | 70.2% |
| Black | 46.3% | 24.1% |
| Other | 3.2% | 5.7% |
| Blood pressure (mm Hg) | ||
| Systolic | 129.5 (120.5–140) | 130.5 (121.5–140) |
| Diastolic | 74 (70–79) | 76.5 (71–82) |
| Body mass index (kg/m 2 ) | 33.5 (29.0–37.8) | 31.1 (28.0–34.2) |
| Total cholesterol (mg/dl) | 180 (158–207) | 171 (148–196) |
| LDL cholesterol (mg/dl) | 102 (82–122) | 94 (78–116) |
| HDL cholesterol (mg/dl) | 53 (45–64) | 44 (38–51) |
| Triglycerides (mg/dl) | 104 (75–147) | 123 (83–186) |
| Metabolic syndrome | 78.9% | 69.6% |
| Estimated GFR (ml/min) | 110 (85–140) | 104 (85–132) |
| Duration of diabetes (years) | 5 (2–10) | 5 (2–10) |
| Glycosylated hemoglobin (%) | 6.9 (6.3–7.7) | 6.8 (6.2–7.8) |
| Current smokers | 7.6% | 11.5% |
| Medications | ||
| Statins | 43.9% | 61.0% |
| Fibrates | 2.0% | 9.2% |
| Aspirin | 40.7% | 47.3% |
| ACE inhibitors | 56.9% | 63.7% |
| Metformin | 63.0% | 64.2% |
| Sulfonylureas | 31.0% | 44.4% |
| Thiazolidinediones | 23.6% | 31.0% |
| Insulin | 22.0% | 16.2% |
| FABP4 (ng/ml) | 37.1 (23.0–55.2) | 20.9 (14.0–30.9) |
| FABP5 (ng/ml) | 1.5 (1.2–1.93) | 1.4 (1.1–1.8) |
| Adiponectin (μg/ml) | 10.4 (7.0–16.2) | 7.9 (5.3–12.3) |
| Leptin (ng/ml) | 27.1 (17.5–36.5) | 8.5 (5.3–13.7) |
| CRP (mg/l) | 2.9 (1.5–7.0) | 1.5 (0.7–3.0) |
| IL-6 (pg/ml) | 1.6 (0.9–2.5) | 1.3 (0.8–2.0) |
| CAC score | ||
| Mean ± SD | 109 ± 289 | 448 ± 801 |
| Median (IQR) | 0 (0–74) | 108 (4–538) |
| >75th percentile | 29.6% | 33.3% |
Spearman’s correlations revealed modest associations of FABP4 and FABP5 with each other in men and women and with metabolic and inflammatory parameters but not with total or low-density lipoprotein cholesterol ( Table 2 ). The 2 FABPs were strongly correlated with waist circumference and body mass index. Overall, FABP4 tended to have stronger associations with measures of adiposity than FABP5, but the 2 were equally related to the inflammatory markers CRP and IL-6.
| Risk Factor | FABP4 | FABP5 | ||
|---|---|---|---|---|
| Women | Men | Women | Men | |
| Age | 0.13 ⁎ | 0.02 | 0.05 | 0.1 ⁎ |
| Body mass index | 0.46 ‡ | 0.38 ‡ | 0.34 ‡ | 0.16 ‡ |
| Waist circumference | 0.46 ‡ | 0.41 ‡ | 0.38 ‡ | 0.20 ‡ |
| Leptin | 0.48 ‡ | 0.51 ‡ | 0.40 ‡ | 0.30 ‡ |
| Adiponectin | 0.08 | 0.22 ‡ | −0.17 † | −0.08 |
| Blood pressure | ||||
| Systolic | 0.4 | 0.08 | 0.06 | 0.13 † |
| Diastolic | −0.01 | 0.02 | 0.03 | 0.01 |
| Total cholesterol | −0.04 | −0.06 | −0.04 | −0.01 |
| LDL cholesterol | −0.03 | −0.09 | −0.04 | −0.03 |
| HDL cholesterol | −0.23 ‡ | −0.11 ⁎ | −0.30 ‡ | −0.17 ‡ |
| Triglycerides | 0.25 ‡ | 0.14 † | 0.25 ‡ | 0.12 † |
| Estimated GFR | 0.13 ⁎ | 0.08 | 0.06 | −0.07 |
| CRP | 0.18 † | 0.08 | 0.18 † | 0.14 † |
| IL-6 | 0.18 † | 0.26 ‡ | 0.15 † | 0.27 ‡ |
| FABP 4 | — | — | 0.45 ‡ | 0.38 ‡ |
In unadjusted analysis, plasma levels of FABP4 and FABP5 were higher in participants with the metabolic syndrome ( Table 3 ), but they had a distinct pattern of association with metabolic syndrome components; FABP4 had stronger associations with waist criteria, whereas FABP5 had greater relations with high-density lipoprotein cholesterol, triglycerides, and blood pressure ( Table 3 ). Notably, levels of FABP4 (median 34.4 ng/ml [interquartile range 22.9 to 55.0] vs 22.7 ng/ml [interquartile range 15.0 to 35.9], p <0.001), but not FABP5 (p = 0.88), were higher in patients receiving thiazolidinedione therapy.
| Metabolic Syndrome Component | Plasma FABP4 | Plasma FABP5 |
|---|---|---|
| HDL cholesterol | ||
| Absent (n = 527) | 23.9 (15.3–38.1) | 1.3 (1.1–1.7) |
| Present (n = 274) | 29.0 (18.5–44.2) | 1.6 (1.3–2.0) |
| Chi-square | 13.5 (p <0.001) | 31.9 (p <0.001) |
| Triglycerides | ||
| Absent (n = 550) | 24.7 (15.6–40.1) | 1.4 (1.1–1.8) |
| Present (n = 251) | 26.7 (17.2–43.0) | 1.5 (1.2–2.0) |
| Chi-square | 2.7 (p = 0.10) | 13.9 (p <0.001) |
| Blood pressure | ||
| Absent (n = 407) | 26.2 (15.8–41.5) | 1.4 (1.1–1.8) |
| Present (n = 399) | 25.0 (16.5–40.6) | 1.5 (1.2–1.9) |
| Chi-square | 0.08 (p = 0.78) | 7.1 (p <0.01) |
| Waist circumference | ||
| Absent (n = 240) | 17.3 (12.1–26.0) | 1.3 (1.0–1.6) |
| Present (n = 566) | 29.2 (20.0–46.3) | 1.5 (1.2–1.9) |
| Chi-square | 115.9 (p <0.001) | 43.0 (p <0.001) |
| All | ||
| Absent (n = 323) | 22.3 (14.1–32.6) | 1.3 (1.0–1.7) |
| Present (n = 480) | 28.0 (18.1–44.9) | 1.5 (1.2–1.9) |
| Chi-square | 29.8 (p <0.001) | 36.2 (p <0.001) |
In multivariate analysis, the association of 1 SD of plasma FABP4 data ( Table 4 ) and FABP5 ( Table 4 ) with the metabolic syndrome remained significant even after adjusting for individual metabolic syndrome components (except for FABP4 adjusted for waist circumference). Remarkably, relative to either FABP, 1 SD of the combined FABP score had a stronger association with the metabolic syndrome even after adjusting for each individual metabolic syndrome component and individual FABP levels ( Table 4 ). Likelihood ratio testing revealed that the addition of FABP score added value to the association with the metabolic syndrome beyond each individual FABP (e.g., chi-square = 22.1, p <0.001, for score added to FABP4 and chi-square = 22.6, p <0.001 for score added to FABP5 in an age-, gender-, and race-adjusted model). These findings suggest an additive relation of FABP4 and FABP5 levels with atherogenic metabolic disturbance in T2DM.
| Model | FABP4 | FABP5 | FABP Score |
|---|---|---|---|
| Basic model (age, gender, race) | 1.85 (1.53–2.23) | 1.66 (1.41–1.95) | 2.01 (1.68–2.41) |
| Basic model + HDL cholesterol | 1.66 (1.35–2.02) | 1.46 (1.23–1.72) | 1.76 (1.45–2.13) |
| Basic model + triglycerides | 1.65 (1.35–2.02) | 1.56 (1.32–1.86) | 1.84 (1.52–2.23) |
| Basic model + blood pressure | 1.85 (1.53–2.25) | 1.61 (1.36–1.89) | 1.97 (1.64–2.37) |
| Basic model + waist circumference | 1.17 (0.95–1.45) | 1.39 (1.17–1.65) | 1.35 (1.10–1.65) |
| Basic model + FABP4 | — | 1.46 (1.23–1.72) | 1.85 (1.43–2.39) |
| Basic model + FABP5 | 1.58 (1.30–1.93) | — | 1.85 (1.43–2.38) |
In unadjusted analysis, FABP4 and FABP5 levels were correlated with the 2 major inflammatory markers CRP and IL-6 ( Table 5 ). In adjusted analysis, the association of 1 SD of plasma FABP4 and FABP5 ( Table 5 ) with plasma CRP levels was significant after adjusting for multiple CVD risk factors. Again, relative to either FABP, 1 SD of the combined FABP score had a stronger association with plasma CRP, even after adjusting for Framingham risk score, the metabolic syndrome, additional risk factors, and levels of each individual FABP ( Table 5 ). Findings were almost identical for associations with plasma IL-6. These results further support an additive relation of the 2 FABPs with inflammatory atherogenic risk in T2DM.
