Adipocyte fatty acid binding protein: a novel adipokine involved in the pathogenesis of metabolic and vascular disease?
Adipocyte fatty acid binding protein (AFABP, also known as aP2 and FABP4) has recently been introduced as a novel fat-derived circulating protein. AFABP serum concentrations are positively correlated with markers of the metabolic syndrome and vascular disease in various cross-sectional and interventional studies. Furthermore, a small set of prospective studies indicates that high AFABP serum levels at baseline predict the risk for metabolic and vascular morbidity and mortality. Studies in Afabp (also known as Fabp4) knockout mice and AFABP inhibitor-treated animals suggest that total AFABP promotes insulin resistance, hypertriacylglycerolaemia and atherosclerosis by ligand/ligand delivery, as well as ligand-independent mechanisms. In contrast, the pathophysiological significance of circulating AFABP and the mechanisms leading to its release remain to be established. The current review summarises recent findings on the regulation and potential role of AFABP in metabolic and vascular disease.
KeywordsAdipose tissue AFABP Cardiovascular disease Insulin resistance Metabolic syndrome Obesity Review
Adipocyte fatty acid binding protein
Coronary artery disease
Epidermal-type fatty acid binding protein
Jun N-terminal kinase
Monocyte chemoattractant protein
Peroxisome proliferator-activated receptor
Vascular endothelial growth factor
Adipocytes produce and secrete a variety of bioactive peptides, termed adipokines, which influence body weight, insulin sensitivity, lipid metabolism and vascular function. Adipocyte fatty acid binding protein (AFABP), also known as fatty acid binding protein-4 and aP2, has recently been suggested as a third adipokine, in addition to leptin  and adiponectin , that is preferentially produced in and released from adipocytes. AFABP accounts for up to 6% of total cytosolic protein in cultured differentiated fat cells . In addition to production in adipocytes, AFABP is also produced in significant amounts in macrophages  and endothelial cells .
This review summarises findings on the production, structure and pathophysiological significance of total AFABP. Recent data on regulation and the potential role of circulating AFABP in metabolic and vascular disease are presented and discussed.
Production and structure of AFABP
The gene coding for AFABP has been identified from a cDNA library constructed from poly(A)+ RNA isolated from differentiated 3T3–L1 adipocytes . The gene, AFABP (also known as FABP4), encodes a polypeptide with 132 amino acids and a molecular mass of 14.6 kDa (GenBank accession number NM_024406). Afabp mRNA and protein production are significantly induced during adipocyte differentiation . AFABP has frequently been used as an adipocyte differentiation marker [8, 9, 10], based on these observations. The DNA-binding protein CCAAT/enhancer-binding protein interacts with the AFABP promoter and elevates expression of the AFABP gene . Furthermore, cyclic adenosine monophosphate induces AFABP gene expression during adipocyte differentiation by relieving the inhibitory effect of a negative regulatory element in the AFABP promoter . Moreover, insulin-sensitising and pro-adipogenic peroxisome proliferator-activated receptor (PPAR)γ agonists induce AFABP production and secretion [12, 13]. In addition, AFABP mRNA and protein production are elevated in subcutaneous adipose tissue compared with the visceral depot in lean and obese humans .
Taken together, published evidence suggests that AFABP is produced in fat cells, regulates transport of NEFA and PPARγ agonists via a ligand/ligand delivery mechanism, and interacts with proteins influencing lipid metabolism and insulin sensitivity, including HSL and PPARγ.
The role of AFABP in body-weight control, glucose metabolism, lipid metabolism and beta cell function
Animal studies on AFABP function
Beta cell function
↓ Lipolysis-induced IS
↔ Beta cell size
↓ Lipolysis-induced IS
↔ Beta cell size
↓ Foam cell development
↓ Lesion area
↓ Foam cell development
↓ Lesion area
↓ Lesion area
↓ Glucose (f)
↓ Lesion area
↔ Glucose (m)
↑ Endothelial function
The influence of AFABP deficiency on beta cell function has been assessed in various studies. Thus, the lipolytic response to β-adrenergic stimulation is significantly reduced in ob/ob–Afabp−/− animals . These findings are in agreement with results from two independent studies obtained in Afabp−/− mice [28, 34]. Reduced β-adrenoceptor-stimulated lipolysis translates into significantly reduced lipolysis-associated insulin secretion from beta cells in both ob/ob–Afabp−/− mice  and Afabp−/− mice . However, loss of AFABP does not generally impair beta cell function. Thus, glucose-stimulated insulin secretion is significantly improved in ob/ob–Afabp−/− mice compared with controls . Furthermore, other insulin secretagogues such as arginine and glibenclamide (known as glyburide in the USA and Canada) are as effective in stimulating an insulin secretory response in Afabp−/− mice as in controls . No apparent morphological differences in the structure or size of the pancreatic islets are detectable between ob/ob–Afabp−/− animals and ob/ob controls .
Taking these findings in AFABP-deficient animals into consideration, AFABP does not contribute to weight gain but promotes insulin resistance and hypertriacylglycerolaemia. Furthermore, AFABP deficiency increases glucose-stimulated insulin secretion from beta cells, whereas it decreases lipolysis-induced insulin secretion. Mechanistically, stimulation of TNF-α production in adipose tissue and induction of lipolysis by AFABP have been suggested as mediators of these effects.
Studies on AFABP function have not only been performed with genetically altered animal models, but also with specific inhibitors. Thus, the small-molecule and orally available synthetic AFABP inhibitor BMS309403 significantly improves insulin sensitivity in ob/ob mice . These findings are in accordance with results in Afabp−/− animals [27, 29, 31, 32]. Mechanistically, macrophage infiltration in adipose tissue is less evident in BMS309403- compared with vehicle-treated animals . Reduced macrophage infiltration in AFABP inhibitor-treated mice is accompanied by diminished abundance of proinflammatory adipokines in adipose tissue, including monocyte chemoattractant protein (MCP)-1, IL-1β, IL-6 and TNF-α . Concerning beta cell function, islet and pancreas morphology are not affected by the AFABP inhibitor . BMS309403 has also been studied in DIO mice . In agreement with findings in ob/ob mice , the AFABP inhibitor reduces plasma triacylglycerol (TG) levels . However, insulin and glucose tolerance are not improved by AFABP inhibition in DIO mice , in contrast to findings in ob/ob animals . In agreement with these findings, knockdown of AFABP mediated by RNA interference technology does not significantly influence glucose homeostasis in DIO mice, but these mice significantly increase in body weight compared with vehicle-treated controls . Interestingly, no compensatory increase in adipocyte Efabp expression is detectable in AFABP knockdown mice  in contrast to previous studies in Afabp−/− animals [27, 28, 29].
Taking the findings in AFABP inhibitor-treated animals into consideration, AFABP promotes hypertriacylglycerolaemia similar to findings in Afabp−/− mice. Whereas AFABP inhibition improves insulin sensitivity in genetically obese mice, this effect is not seen in DIO animals. These studies indicate that the influence of AFABP on insulin sensitivity might depend on the extent of AFABP loss and the mode of body weight gain (DIO vs genetically determined obesity).
The role of AFABP in the pathogenesis of atherosclerosis
Afabp–ApoE double-knockout (Afabp−/−–ApoE−/−) animals show protection from atherosclerosis in the absence of significant differences in serum lipids and insulin sensitivity compared with control ApoE−/− animals (Table 1) . In more detail, quantitative analysis of the proximal aorta reveals that the mean atherosclerotic lesion area in Afabp−/−–ApoE−/− male mice is reduced by 66% compared with control mice . Interestingly, ApoE−/− mice with AFABP deficiency specifically restricted to macrophages show a comparable reduction in atherosclerotic lesion area compared with animals with whole-body AFABP deficiency . In accordance with these results, Afabp−/−–ApoE−/− mice have reduced atherosclerotic lesion formation after 14 weeks of Western diet . In a second paper, Boord et al investigated Afabp–Efabp–ApoE triple knockout (Afabp−/−–Efabp−/−–ApoE−/−) mice, with similar results . Interestingly, lipid metabolism in both sexes and survival in males were also improved in Afabp−/−–Efabp−/−–ApoE−/−mice. In accordance with these findings, the AFABP inhibitor BMS309403 markedly reduces atherosclerotic lesion area in the en face aorta by more than 50% in ApoE−/− mice . Similar results have been obtained in an independent study using the same AFABP inhibitor .
One study in human participants supports the hypothesis that AFABP might also contribute to human atherosclerotic disease. Thus, a functionally significant genetic variation (T-87C polymorphism) at the AFABP locus in humans not only results in reduced adipose tissue AFABP expression, but is also associated with lower serum TG levels as well as a lower risk for coronary heart disease and type 2 diabetes mellitus . However, the relationship between AFABP genetic variants and type 2 diabetes mellitus has not been confirmed recently (e.g. in the Women’s Health Initiative observational study) . Here, none of the AFABP single-nucleotide polymorphisms analysed showed a significant association with type 2 diabetes mellitus risk.
The studies summarised above convincingly demonstrate that AFABP promotes atherosclerotic disease. The impact of AFABP on atherosclerosis appears to be related exclusively to its action in macrophages . Thus, AFABP-deficient macrophages display reduced inhibitor of κB kinase and nuclear factor κB activity, resulting in decreased levels of the proinflammatory cytokines TNF-α, IL-6 and MCP-1 . In agreement with these findings, treatment with the AFABP inhibitor BMS309403 significantly decreases TNF-α, IL-6 and MCP-1 production in THP-1 macrophages . In addition, AFABP-deficient mice are resistant to several inflammatory disorders, including allergic airway inflammation  and experimental autoimmune encephalomyelitis/multiple sclerosis . Several proinflammatory stimuli induce AFABP production, including phorbol myristate acetate, oxidised (ox)LDL and toll-like receptor (TLR) agonists [4, 35, 47, 48]. Among these, oxLDL upregulates AFABP during the transformation of macrophages into foam cells . This AFABP elevation involves activation of PPARγ by components of oxLDL that function as PPARγ ligands . Lipopolysaccharides (LPS) are potent TLR4 agonists that induce AFABP production in macrophages . LPS stimulates AFABP transcription through Jun N-terminal kinase (JNK), which in turn induces c-Jun recruitment to a highly conserved activator protein-1 recognition site within the proximal region of the AFABP promoter . In addition, LPS-induced JNK phosphorylation, activator protein-1 stimulation and production of proinflammatory cytokines are significantly attenuated by pharmacological or genetic AFABP suppression in macrophages . Besides TLR4-agonistic LPS, the TLR2 ligand zymosan and the TLR3 ligand polyinosine:polycytidylic acid significantly upregulate AFABP in murine macrophages . Furthermore, saturated fatty acids including palmitate induce AFABP production in macrophages . Most interestingly, AFABP is an obligatory mediator coupling toxic lipids (i.e. saturated fatty acids) to endoplasmic reticulum (ER) stress in macrophages in vitro and in vivo . The impact of AFABP on macrophage ER stress is mediated by upregulation of key lipogenic enzymes by the liver X receptor . It is interesting to note from a clinical standpoint that mitigation of ER stress with a chemical chaperone not only prevents macrophage AFABP production, but also leads to marked protection against macrophage ER stress, cell death and atherosclerosis .
Upregulation of AFABP in foam cells regulates cholesterol and TG accumulation in macrophages. Thus, adenovirus- or lentivirus-mediated overproduction of AFABP enhances the accumulation of cholesteryl esters in macrophage foam cells . In accordance with these results, human THP-1 macrophages treated with an AFABP inhibitor exhibit 44% reduced cholesteryl ester accumulation and significantly increased cholesterol efflux relative to vehicle-treated cells . Moreover, upregulation of AFABP gene expression by advanced glycation end-products has also been shown to induce cholesterol and TG accumulation in THP-1 macrophages in vitro in an independent study .
It is interesting to note in a clinical context that upregulation of AFABP is also found in vivo in macrophage/foam cells from human atherosclerotic plaques . Furthermore, independent studies suggest that increased AFABP synthesis in atherosclerotic plaques is associated with disease severity [53, 54, 55, 56]. Various studies agree that AFABP synthesis is localised in macrophages in atherosclerotic plaques [54, 55, 56]. Moreover, atorvastatin, which has antiatherogenic properties, potently reduces Afabp mRNA and AFABP protein level in oxLDL-treated human macrophages .
AFABP is also detected in endothelial cells as well as in adipocytes  and macrophages . Afabp mRNA expression and AFABP protein synthesis are significantly induced in cultured endothelial cells by vascular endothelial growth factor (VEGF). Interestingly, knockdown of AFABP has been reported to dramatically reduce proliferation of endothelial cells both under baseline conditions and in response to VEGF in vitro . The same group has identified AFABP as a mediator of angiogenesis . Thus, AFABP-deficient human umbilical vein endothelial cells demonstrate a markedly increased susceptibility to apoptosis, as well as decreased migration and capillary network formation . Furthermore, aortic rings from Afabp−/− mice show decreased angiogenic sprouting, which is reversible upon Afabp re-expression .
Taken together, these studies indicate that AFABP is upregulated by proatherogenic stimuli and induces atherosclerosis by (1) promoting cholesterol and TG accumulation in and inhibiting cholesterol efflux from macrophages; and (2) inducing proinflammatory and proatherogenic cytokines in macrophages. Furthermore, AFABP might have a role in the proliferation of endothelial cells, as well as in angiogenesis.
Circulating AFABP—a potential novel adipokine
The first description of AFABP as a circulating protein was published in 2006 . Since this publication, several cross-sectional, interventional and prospective studies have investigated the regulation and predictive value of circulating AFABP in humans. Furthermore, potential mechanisms leading to the release of AFABP, as well as potential effects of circulating AFABP, have been elucidated recently and are summarised below.
Potential mechanisms by which AFABP reaches the circulation
While AFABP has been established as a circulating protein, the mechanisms by which the adipokine reaches the circulation are unclear. So far, there is no solid evidence that AFABP is, in fact, secreted or that any regulated release of AFABP from adipocytes or macrophages occurs. Thus, there is no secretory signal sequence in the primary and three-dimensional structure of AFABP [60, 61]. However, proteins lacking exocytosis-targeting signal sequences can cross the plasma membrane by different, unconventional, secretion mechanisms . As an example, fibroblast growth factor-1 is exported through an unconventional release pathway involving the formation of a specific multiprotein complex . Furthermore, murine adipocytes release various proteins lacking a secretory signal sequence by microvesicles and, most interestingly, AFABP is detected in microvesicles . These findings are supported by an independent study that convincingly demonstrated that the microvesicle fraction prepared by ultracentrifugation of human adipocyte supernatant fractions contains AFABP . Taking these findings into consideration, AFABP might get into the circulation at least in part by adipocyte-derived microvesicles. Nevertheless, it is also possible that AFABP reaches the circulation from lysis of large adipocytes. In agreement with this hypothesis, most studies on circulating AFABP have shown a positive correlation with measures of adiposity, including BMI and percentage body fat. However, whereas both β-tubulin and AFABP are present in the cell lysate of cultured 3T3–L1 adipocytes, only AFABP is present in the culture medium, suggesting the secretion of AFABP from cultured adipocytes . Clearly, more work is needed to define the mechanisms by which AFABP reaches the circulation.
Potential effects of circulating AFABP
Only one study to date has elucidated the potential role of circulating AFABP in vitro . In an elegant study, Lamounier-Zepter et al identified AFABP as a cardiodepressant factor. Recombinant AFABP in a concentration of 100 nmol/l decreased shortening amplitude by more than 50% and intracellular systolic peak Ca2+ by more than 20% in isolated adult rat cardiomyocytes. The N-terminal amino acids 1–20 of AFABP were the most effective cardiodepressive structure, showing approximately 40% suppression of rat cardiomyocyte contraction. Interestingly, adipocyte-secreted microvesicles contained up to 8% of the total amount of AFABP detected in the supernatant fraction, and the microvesicle fraction from human adipocytes had a low but significant cardiodepressant effects on rat cardiomyocytes (−28% decrease in shortening amplitude). These findings suggest that ligand-independent mechanisms might also contribute to actions of AFABP in addition to ligand/ligand delivery mechanisms. However, these potential ligand-independent mechanisms need to be confirmed and further characterised in additional studies. Furthermore, future experiments should aim to determine whether AFABP exists in its apo- and/or holo-form in human serum.
Circulating AFABP, the metabolic syndrome and atherosclerosis—cross-sectional studies
Main findings on circulating AFABP in cross-sectional studies in humans
Main finding concerning circulating AFABP
↑ in MS
↑ in CVD
↑ in obesity and MS
+ BP, fasting insulin, HOMA-IR, LDL-cholesterol, TG, waist circumference
− Adiponectin, HDL-cholesterol
↑ in MS
12 MS+/LD–, 32 MS–/LD+, 17 MS+/LD+, 122 controls
↑ in MS and LD
+ BMI, BP, fasting insulin, TG
↑ in FCHL
+ BMI, fasting insulin, HOMA-IR, waist circumference (FCHL)
Predicts inflammation and fibrosis in NAFLD
Associated with obesity
+ BMI, waist circumference
Determined by fat distribution, renal function and sex
+ C-reactive protein, TG, WHtR
− GFR, Stumvoll index
Associated with atherogenic dyslipidaemia in T2DM
+ LDL-cholesterol, TG (T2DM)
− HDL-cholesterol (T2DM)
Associated with MS
479 with carotid IMT
Associated with IMT in women
+ Age, BP, carotid IMT (in women)
234 with coronary angiograms
Associated with CAD
↑ in CAD
↑ in CAD
+ Plaque volume (CAD)
↑ in CAD
+ Coronary atherosclerosis index
↑ in CVD
Associated with vascular inflammation
+ Vascular inflammation
306 acute ischaemic stroke
↑ in ischaemic stroke
The first description of the association between circulating AFABP and atherosclerosis was published in 2007 . This study reported that circulating AFABP levels are significantly higher in women with atherosclerotic plaques compared with those without plaques. Furthermore, AFABP concentrations significantly increase with the number of stenotic coronary arteries . In accordance with these findings, three independent studies have reported that serum AFABP is significantly elevated in patients with coronary artery disease (CAD) [76, 77, 78]. Very recently, a small case–control study detected significantly elevated serum AFABP levels in patients with cerebrovascular disease, whereas no difference was found in patients with CAD . Similarly, AFABP plasma concentrations are increased in patients with carotid atherosclerosis compared with sex- and age-matched healthy controls . Furthermore, the adipokine positively correlates with vascular inflammation measured by 18F-fluorodeoxyglucose positron emission tomography .
Taken together, these studies provide evidence that circulating AFABP is increased in patients with the metabolic syndrome and atherosclerosis, and might be a marker for these disease states. Whether AFABP directly contributes to these disorders needs to be determined in future studies.
Interventional studies on circulating AFABP
Main findings on circulating AFABP in interventional studies in humans
Type of intervention
Main finding concerning circulating AFABP
30 obese women
↓ after weight loss
77 morbidly obese women
↓ after weight loss
73 prepubertal obese children
Diet and lifestyle intervention
↓ after weight loss
189 obese participants
↑ after weight loss
43 participants with atherosclerosis
↓ after atorvastatin treatment
In summary, interventional studies suggest that circulating AFABP decreases after weight loss and statin therapy, and might contribute to the beneficial effects of these interventions.
Prospective studies on circulating AFABP
Elevated concentrations of circulating AFABP independently predict the risk of developing a metabolic syndrome . In agreement with these findings, baseline serum AFABP concentrations are significantly higher in children who develop a metabolic syndrome compared with controls . High baseline AFABP serum levels are also a biomarker predicting the development of type 2 diabetes mellitus independent of obesity, insulin resistance and glycaemic indices . Furthermore, circulating AFABP predicts prognosis in patients with end-stage renal disease, a high-risk population for atherosclerosis-associated morbidity and mortality . Moreover, serum AFABP is independently and positively associated with stroke and 3-month mortality in ischaemic stroke patients . Similar findings have been described in two independent studies [54, 56]. Very recently, it was shown that circulating AFABP concentrations are independently associated with cardiovascular morbidity and mortality in patients with CAD .
Main findings on circulating AFABP in prospective studies in humans
Follow-up period (years)
Main finding concerning high circulating AFABP
Variables besides AFABP associated with endpoint
202 with acute ischaemic stroke (60 died)
Predicts death after ischaemic stroke
Age, troponin T, stroke localisation
284 (70 reached primary endpoint)
Predicts cardiovascular events
HR 1.7 (p = 0.07)
376 without MS (50 developed MS)
Predicts development of MS
159 without MS (10 developed MS)
Predicts development of MS
544 without T2DM (96 developed T2DM)
Predicts development of T2DM
Sex, 2 h post-OGTT glucose
61 with ESRD (13 died)
Predicts cardiovascular mortality in ESRD
366 with acute ischaemic stroke (60 died)
Predicts early death after ischaemic stroke
Age, NIH Stroke Scale score
1,069 with CAD (204 reach endpoint)
Predicts cardiovascular morbidity and mortality
Summary and concluding remarks
Convincing evidence has been presented that AFABP is an adipocyte- and macrophage-produced protein that promotes insulin resistance, hypertriacylglycerolaemia and atherosclerosis by ligand/ligand delivery, as well as ligand-independent, mechanisms (Table 1). Results from animal experiments suggest that AFABP production in adipocytes leads to insulin resistance, NEFA release and proinflammatory gene expression. Furthermore, AFABP synthesis in macrophages is crucial for its effects on atherosclerosis. Thus, AFABP overproduction leads to foam cell development by increasing cholesterol and TG accumulation, as well as by inducing expression of proinflammatory genes. Furthermore, macrophage infiltration into adipose tissue is stimulated by AFABP. These data indicate that AFABP might be an interesting therapeutic target for metabolic and vascular disease.
Since 2006, data have accumulated that AFABP is also a circulating protein, the levels of which are independently and positively associated with various facets of the metabolic syndrome and vascular disease (Table 2). Furthermore, interventional studies suggest that circulating AFABP decreases after weight loss and statin therapy (Table 3). In addition, a small set of prospective studies indicates that high AFABP serum levels at baseline predict the risk for metabolic and vascular morbidity, as well as mortality (Table 4).
Despite these recent advances in our understanding of AFABP physiology, various important questions remain unanswered, particularly concerning the role of AFABP as an adipokine. These include: (1) Is AFABP secreted from adipocytes and, if yes, what are the mechanisms? (2) Is circulating AFABP only a marker of metabolic and vascular disease or does it directly influence these conditions? (3) In the latter case, does AFABP mediate its effects via ligand-dependent and/or independent mechanisms? Furthermore, what are the target tissues of circulating AFABP?
These questions need to be addressed in future studies and are critical to determining the role of circulating AFABP in metabolic and vascular disease in more detail.
Studies from our group summarised in this review were supported by grants to M. Fasshauer from the Deutsche Forschungsgemeinschaft (DFG, KFO 152 ‘Atherobesity’, FA476/4–1), the Federal Ministry of Education and Research (BMBF), Germany, FKZ: 01EO1001 (IFB Adiposity Diseases, projects K7-3, K7-9, and K7-31), and the Deutsche Hochdruckliga e.V. The authors would like to thank S. Lorenz, J. Richter and J. Schaarschmidt (all Department of Endocrinology and Nephrology, University of Leipzig, Leipzig, Germany) for their help in generating the figures.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
Both authors were responsible for the conception, design and drafting of the manuscript, and approved the final version for publication.
- 22.Smith AJ, Sanders MA, Juhlmann BE, Hertzel AV, Bernlohr DA (2008) Mapping of the hormone-sensitive lipase binding site on the adipocyte fatty acid-binding protein (AFABP) identification of the charge quartet on the AFABP/aP2 helix-turn-helix domain. J Biol Chem 283:33536–33543PubMedCrossRefGoogle Scholar
- 29.Shaughnessy S, Smith ER, Kodukula S, Storch J, Fried SK (2000) Adipocyte metabolism in adipocyte fatty acid binding protein knockout (aP2(−/−)) mice after short-term high-fat feeding—functional compensation by the keritinocyte fatty acid binding protein. Diabetes 49:904–911PubMedCrossRefGoogle Scholar
- 44.Makowski L, Brittingham KC, Reynolds JM, Suttles J, Hotamisligil GS (2005) The fatty acid-binding protein, aP2, coordinates macrophage cholesterol trafficking and inflammatory activity. Macrophage expression of aP2 impacts peroxisome proliferator-activated receptor gamma and IkappaB kinase activities. J Biol Chem 280:12888–12895PubMedCrossRefGoogle Scholar
- 76.Doi M, Miyoshi T, Hirohata S et al (2011) Association of increased plasma adipocyte fatty acid-binding protein with coronary artery disease in non-elderly men. Cardiovasc Diabetol 10:44Google Scholar
- 87.Choi KM, Yannakoulia M, Park MS et al (2011) Serum adipocyte fatty acid-binding protein, retinol-binding protein 4, and adiponectin concentrations in relation to the development of the metabolic syndrome in Korean boys: a 3-y prospective cohort study. Am J Clin Nutr 93:19–26PubMedCrossRefGoogle Scholar