Metabolically Protective Cytokines Adiponectin and Fibroblast Growth Factor-21 Are Increased by Acute Overfeeding in Healthy Humans

Context Circulating levels of metabolically protective and adverse cytokines are altered in obese humans and rodent models. However, it is not clear whether these cytokines are altered rapidly in response to over-nutrition, or as a later consequence of the obese state. Methods Forty sedentary healthy individuals were examined prior to and at 3 and 28 days of high fat overfeeding (+1250 kCal/day, 45% fat). Insulin sensitivity (hyperinsulinaemic-euglycaemic clamp), adiposity, serum levels of adiponectin and fibroblast growth factor-21 (FGF21), fatty acid binding protein-4 (FABP4), lipocalin-2 and plasminogen activator factor-1 (PAI1) were assessed. Statistics were performed by repeated measures ANOVA. Results Overfeeding increased weight, body fat and liver fat, fasting glucose, insulin and reduced insulin sensitivity by clamp (all P <0.05). Metabolically protective cytokines, adiponectin and FGF21 were increased at day 3 of overfeeding (P ≤0.001) and adiponectin was also elevated at day 28 (P=0.001). FABP4, lipocalin-2 and PAI-1 were not changed by overfeeding at either time point. Conclusion Metabolically protective cytokines, adiponectin and FGF-21, were increased by over nutrition and weight gain in healthy humans, despite increases in insulin resistance. We speculate that this was in attempt to maintain glucose homeostasis in a state of nutritional excess. PAI-I, FABP4 and lipocalin 2 were not altered by overfeeding suggesting that changes in these cytokines may be a later consequence of the obese state. Clinical trial registration: www.clinicaltrials.gov (NCT00562393)


Introduction
Obesity is associated with increased secretion of metabolically adverse adipokines such as fatty acid binding protein-4 (FABP4) and reduced secretion of metabolically protective adipokines such as adiponectin [1]. Obesity is also associated with ectopic lipid deposition in the liver [2] and altered secretion of cytokines from the liver. Whilst the exact target and function of many novel cytokines remain to be elucidated, there is increasing evidence to suggest direct involvement in the pathogenesis of insulin resistance [3].
Adiponectin is an abundantly produced adipokine that is inversely related to adiposity, and exerts protective antiinflammatory and anti-apoptotic effects that improve insulin sensitivity and reduce oxidative stress [4]. FABP4 is a lipid binding chaperone that interacts with hormone sensitive lipase in adipose tissue to facilitate lipolysis [5]. FABP4 is also released from adipocytes [6], although the mechanism, target and function of the secreted form of FABP4 is not clear. Circulating levels of FABP4 are positively correlated with obesity and non-alcoholic fatty liver [7] and predict the development of metabolic syndrome and type 2 diabetes [8]. Moreover, FABP4 disruption decreases insulin resistance, dyslipidemia and liver steatosis in obese or fat fed animals [9]. Lipocalin-2 was originally identified in human neutrophils, but is also secreted from adipose tissue, macrophages, liver, lung, and kidney and is involved in the inflammatory response [10]. Lipocalin-2 is elevated in obesity, diabetes and non-alcoholic fatty liver disease [7,11]. Plasminogen activator inhibitor-1 (PAI-1) is expressed in many tissues including visceral and subcutaneous adipose tissue depots, although adipose tissueresident macrophages may be the primary source of this cytokine in these depots [12]. PAI-1 is the prime regulator of fibrinolysis and subjects with increased levels of PAI-1 are predisposed to atherothrombotic events [12]. Obese patients with and without type 2 diabetes have elevated levels of PAI-1 as compared to lean individuals and this is closely linked with insulin resistance [13].
Like adiponectin, the recently discovered and predominately liver derived cytokine, fibroblast growth factor-21 (FGF21) is considered to be metabolically protective. Over-expression or treatment with recombinant FGF21 protects mice from development of obesity and fatty liver and improves insulin sensitivity [3] and comparable metabolic benefits are also observed in rhesus monkeys [14]. Paradoxically, FGF21 levels are elevated in obesity [15] and obese mice treated with recombinant FGF21 have a blunted metabolic response to FGF21, which suggests that resistance to the actions of FGF21 may develop in obesity [16].
Short term overfeeding provides a model to examine whether cytokines are altered rapidly in response to over-nutrition. In this cohort, we have previously reported that short term overfeeding does not alter macrophage infiltration of adipose tissue, although increases in C-reactive protein (CRP) and macrophage chemo-attractant protein-1 (MCP1) were identified [17]. Here, we examined the relationships between adiposity and insulin resistance and circulating levels of adiponectin, FABP4, lipocalin-2, PAI-1 and FGF21 before and after high fat overfeeding for 28 days.

Subjects and Study Design
The study was approved by the Human Research and Ethics Committee of St Vincent's Hospital and registered as a Clinical Trial at clinicaltrials.gov, registration number NCT00562393, all participants signed informed consent prior to the study. The study protocol and CONSORT diagram is described previously [18,19] and this analysis includes the entire cohort as previously reported.
The cohort consisted of 20 women (5 post-menopausal) and 20 men, mean age 37 years (range 21-59 years); 17 subjects reported a family history of type 2 diabetes. All subjects were healthy, sedentary and not taking any medications. All foods were provided for 3-days before metabolic testing at baseline, day 3 and study end as described previously [18]. At baseline, this was at calculated energy requirements, with a nutrient composition of 30% fat, 15% protein and 55% carbohydrate. During the first and last 3 days of overfeeding, all foods were provided at 1250 kcal/day above calculated energy requirements, with a nutrient composition of 45% fat, 15% protein and 40% carbohydrate. During the remainder of the overfeeding phase (days 3 -25), subjects were asked to consume their regular diet with high fat snacks provided to increase energy intake by 1250 kcal/day. Average (± SEM) reported energy intakes during the self-selecting phase increased from 1980 ± 110 to 3100 ± 140 kcal/d and the contribution of fat from 34 ± 1 at baseline to 45 ± 1 %. Weight gain and compliance were monitored weekly by the research nurse and dietician.

Metabolic Testing at Baseline and after 28 Days of Overfeeding
All metabolic tests were conducted at the Clinical Research Facility at the Garvan Institute of Medical Research after a 12h overnight fast. Weight was measured in a hospital gown after voiding. Height, blood pressure, waist and hip circumference were also measured. Fasting blood samples were drawn and insulin sensitivity was assessed by a 2-h hyperinsulinaemiceuglycaemic clamp (60 mU/m 2 /min), as described [18]. Steady state glucose infusion rate (GIR) between 90 and 120 min of the clamp procedure was averaged and normalized for body fat free mass (FFM). Resting metabolic rate and respiratory quotient were assessed by True One metabolic cart following 30 minutes of supine rest (ParvoMedics, UT, USA). Body composition was measured by dual energy x-ray absorptiometry (DXA; Lunar DPX GE Lunar, Lunar Corp., Madison, WI). Three cross-sectional computed tomography (CT) scans (Phillips Gemini GXL), 1 cm-width, centered on the L2-L3 and L4-L5 disc space, and the T12-L1 disc space were also performed to assess abdominal adipose tissue distribution and hepatic fat content as described [18]. An additional blood sample for fasting levels of glucose, insulin and cytokines was taken at 3 days of overfeeding.

Statistical Analysis
Data are presented as mean ± SEM unless otherwise stated. FABP4 and insulin data were log 10 -transformed prior to statistical analysis. Since no differences were detected for any markers of inflammation at baseline or in response to overfeeding between individuals with and without a family history of type 2 diabetes, these groups were combined. The effects of overfeeding were tested using repeated measures-ANOVA with gender as the between-subject factor. In case of significance a paried t test was used to evaluate differences between time-points. Pearson's correlations were performed and partial correlations were used to adjust for gender. Sample size requirements were calculated for primary endpoint ∆ (glucose infusion rate/kg FFM) using 2 sided within subject contrasts with α<0.05 and statistical power 1-β>0.8. Statistical Package for Social Sciences (SPSS) version 21 (Chicago, IL) was used, without adjustment for multiplicity.

Metabolic consequences of overfeeding
Anthropometric and metabolic measures at baseline and after 28 days of overfeeding are summarized in Table 2. Overfeeding induced significant weight gain (2.7 ± 0.3 kg, P <0.001). Similarly, BMI, percent body fat, abdominal subcutaneous and visceral adipose tissue and liver fat were significantly increased by overfeeding ( Table 2, all P <0.01). Fasting glucose and insulin increased at both 3 and 28 days (P <0.01) and fasting NEFA decreased transiently at day 3 (P<0.001) and returned to baseline levels at day 28 (P=0.4). Insulin sensitivity by hyperinsulinemic clamp and HOMA-IR was reduced (P=0.03 and P<0.001, respectively). Resting metabolic rate (RMR) increased significantly (P≤0.003) and the respiratory quotient (RQ) increased significantly at day 3 (P<0.001) and returned to baseline levels on day 28 (P=0.2). Overfeeding transiently increased serum levels of FGF21 at day 3 (P=0.001, Figure 2A) and increased adiponectin levels at day 3 and day 28 (P ≤0.001, Figure 2B). The responses in serum adiponectin were independent of gender (P >0.2 for both), but FGF21 was increased to a greater extent in men at day 3 (P=0.04). No statistically significant changes were detected in lipocalin-2 (P=0.1, Figure 2C), FABP4 (P=0.7, Figure 2D) and PAI-1 (P=0.4, Figure 2E). Of the cytokines examined, only the change in adiponectin was correlated with weight gain and change in BMI at day 28 ( Figure 3A and B, respectively). The change in FGF21 was positively related to the change in resting energy expenditure ( Figure 3C) and the change in FABP4 correlated with the change in NEFA ( Figure 3D). The change in PAI-1 and lipocalin-2 were not correlated with changes in any metabolic variable measured in response to overfeeding.

Discussion
The secretion profile of pro and anti-inflammatory cytokines is altered in obesity and may directly impact insulin sensitivity. However, it is not clear if changes occur as an early event in response to overfeeding and insulin resistance, or as a later consequence of the obese state. In this study, we show that overfeeding resulted in weight gain, increased lipid deposition in adipose tissue and liver and reduced insulin sensitivity. In parallel, increases in metabolically protective cytokines adiponectin and FGF21 were observed. No changes were detected in the metabolically adverse cytokines FABP4, PAI-1 and lipocalin-2, suggesting that these may be altered as a later consequence of the obese state.
Adiponectin is primarily produced by the adipocyte and acts as a metabolically protective cytokine, with beneficial actions not only on insulin sensitivity, but also on the cardiovascular system [4]. Circulating adiponectin concentrations are not changed in response to a single high fat meal [24]. However, weight loss significantly increases circulating adiponectin [25]. Here, we observed that total adiponectin levels were increased in response to short-term overfeeding and weight gain. This is also previously reported after 5 days of high fat overfeeding in lean young men [26], thus we confirm and extend these findings to show elevations are sustained at least to 28 days in healthy humans. Moreover, we observed that the increase in adiponectin was correlated with the amount of weight gained during overfeeding.
Whilst this response appears counterintuitive, since insulin sensitivity was reduced, it may be the metabolically healthy response in an attempt to maintain an anti-inflammatory profile and glucose homeostasis in liver and muscle. However, it should be noted that adiponectin circulates in multimeric forms [4], with the high molecular weight (HMW) being the biologically active form, which was not assessed in this study.
FGF21 is also postulated to be a metabolically protective cytokine since treatment with FGF21 mimetics protects against obesity, fatty liver and insulin resistance in mouse and primates [3,14]. Weight loss does not alter FGF21 levels in obese humans [27], but FGF21 is increased in response to 24-hours of fasting in mice and stimulates adipose tissue lipolysis and lipid metabolism [28]. In humans, fasting for between 1-3 days did not alter FGF21 [29,30], but fasting for 7 days significantly increased FGF21 [31]. FGF21 was also increased by lipidheparin infusion which concurrently increased NEFA above 1.5

Figure 1. Relationships between circulating fatty acid binding protein 4 and fasting insulin (A) and percent body fat by DXA (B), and between adiponectin and visceral adipose tissue (C) and fasting serum non-esterified fatty acid (D) in men (empty circles) and women (dark circles).
Depicted are the line of fit and the 95% confidence curves that were obtained from linear regression. Abbreviations: FABP4, fatty acid binding protein 4; AT, adipose tissue; NEFA, non-esterified fatty acid. doi: 10.1371/journal.pone.0078864.g001 mmol/L [32]. These studies suggest that under conditions of increased delivery of fatty acids to the liver, FGF21 is increased and promotes lipid oxidation. Supporting this concept, high fat liquid gavage also increased expression of FGF21 in mouse liver [33]. In this study, we observed a transient increase in circulating FGF21 after 3-days of overfeeding and a similar finding has been reported in healthy men after 5 days of overfeeding [34]. This supports the notion that FGF21 is increased by conditions that likely promoted increased delivery of fatty acids to the liver. However, we do not have evidence of a role for FGF21 in promoting lipid oxidation under these conditions. Rather, FGF21 was increased alongside an increase in glucose oxidation at day 3 and returned to baseline despite an increase in measured lipid content in the liver at day 28. Additionally, we also observed a positive relationship between FGF21 and respiratory quotient at baseline. Whilst this is based on correlational analysis only, there is some evidence to support the concept that FGF21 may be regulated by both fasting and feeding signals, since the promoter region not only has a peroxisome proliferator activated receptor-α (PPARα) element that responds to fatty acids but a carbohydrate response element (ChREBP) that responds to glucose [35]. Of note however, FGF21 is also increased by changes in glucagon [36], which are known to be increased by overfeeding in humans [37]. Further study into the role of FGF21 under conditions of energy excess in humans is required. Serum levels of FABP4 are elevated in obesity and metabolic syndrome and may contribute to development of nonalcoholic steatohepatitis (NASH) [38]. We also observed relationships between serum FABP4 and numerous measures of adiposity and insulin resistance at baseline. However, FABP4 was not changed in response to overfeeding and moderate weight gain, suggesting that this is not nutritionally regulated or responsive to acute gains in liver fat. Our results are in contrast to previous studies that have reported that moderate exercise and weight losses produce small reductions in FABP4 [39], but are in line with reports in mouse that show elevations in FABP4 are aligned with models of NASH rather than fatty liver per se [33]. Lipocalin-2 is elevated in animal models of obesity and in obese humans [11]. However, the results from studies into the phenotype of the lipocalin-2 knockout mouse are mixed, with reports of no phenotype, as well as protection from and promotion of an obesity phenotype and insulin resistance [40][41][42]. Weight loss may reduce lipocalin-2 in humans [43], however this is not observed consistently [39]. In this study, we did not observe any relationship between lipocalin-2 and adiposity or insulin resistance at baseline. Similarly, we did not observe any relationships between obesity and PAI-1. Both may simply reflect the narrow BMI band investigated. However, we also observed no change in PAI-1 or lipocalin-2 in response to overfeeding and insulin resistance. In this cohort, short term overfeeding did not alter macrophage infiltration of adipose tissue [17], and therefore we suggest that PAI-1 and lipocalin-2 may be elevated as a consequence of macrophage infiltration of adipose tissue in the obese state [1].
In summary, over-nutrition increased body weight and markers of insulin resistance in healthy humans. Serum levels of adiponectin and FGF-21 were increased by overfeeding, possibly in an attempt to maintain insulin sensitivity. PAI-I, FABP4 and lipocalin 2 were not altered by overfeeding and moderate weight gain suggesting that changes in these cytokines are a later consequence of the obese state.