Features of an altered AMPK metabolic pathway in Gilbert’s Syndrome, and its role in metabolic health

Energy metabolism, involving the ATP-dependent AMPK-PgC-Ppar pathway impacts metabolic health immensely, in that its impairment can lead to obesity, giving rise to disease. Based on observations that individuals with Gilbert’s syndrome (GS; UGT1A1*28 promoter mutation) are generally lighter, leaner and healthier than controls, specific inter-group differences in the AMPK pathway regulation were explored. Therefore, a case-control study involving 120 fasted, healthy, age- and gender matched subjects with/without GS, was conducted. By utilising intra-cellular flow cytometry (next to assessing AMPKα1 gene expression), levels of functioning proteins (phospho-AMPK α1/α2, PgC 1 α, Ppar α and γ) were measured in PBMCs (peripheral blood mononucleated cells). In GS individuals, rates of phospho-AMPK α1/α2, -Ppar α/γ and of PgC 1α were significantly higher, attesting to a boosted fasting response in this condition. In line with this finding, AMPKα1 gene expression was equal between the groups, possibly stressing the post-translational importance of boosted fasting effects in GS. In reflection of an apparently improved health status, GS individuals had significantly lower BMI, glucose, insulin, C-peptide and triglyceride levels. Herewith, we propose a new theory to explain why individuals having GS are leaner and healthier, and are therefore less likely to contract metabolic diseases or die prematurely thereof.

in the condition of GS. Representing one of the most important energetic controllers and bottleneck of all energy consuming cellular processes, AMPK α1/α2 catalytic activity together with subsequent downstream metabolic effectors (PgC 1α, peroxisome proliferator-activated receptor gamma coactivator 1-alpha; Ppar α and γ, peroxisome proliferator-activated receptors α and γ; Sirt-1, sirtuin-1; FGF-21, fibroblast growth factor 21) were explored in terms of inter-group (GS and controls) differences in (activated) protein levels.
The core regulatory unit studied, AMPK, is a member of a metabolite sensing protein kinase family, that is present in all eukaryotes 9 , and retained in all cell types for regulating energy turnover. It is allosterically activated by increasing levels of ADP and AMP, and therefore considered primarily as a "fuelling-gauge" recognizing ATP depletion (as in fasting), limiting further energy consumption 10 . Together with a decline in ATP, upstream kinase activity determine AMPK's activity through its phosphorylation status 11,12 . Active AMPK subsequently inactivates enzymes responsible for cholesterol-, fatty acid synthesis and gluconeogenesis. For years, this very mechanism has been exploited to routinely treat DM II, by using the anti-diabetic drug Metformin to increase AMPK phosphorylation, and ultimately improve glucose metabolism 13,14 .
Another important effect of AMPK activation includes the post-translational phosphorylation of PgC 1α, which is a positive regulator of energy consuming events such as oxidative processes (including mitogenesis and browning of adipose tissue), and adaptive thermogenesis. Its activity is furthermore fuelled in conditions of physical stress 15 , and is enhanced by the enzyme Sirt-1, another determinant of energy homeostasis 16 .
In immediate response to active PgC 1α and Sirt-1 16 , Ppars, an isotypic group of three (α, β/δ and γ 17 ) nuclear receptor phospho-proteins 18 and transcription factors 19 , are expressed. Means of their activation include phosphorylation through AMPK and ligand binding including fatty acids 20 . Ppars are specifically abundant in certain tissues including the liver, brain, muscle and cells of the immune system 21 . They occur ubiquitously in all cells 22 , as they control the expression of genes involved in adipogenesis and lipid metabolism. Therefore, Ppars are considered as crucial networkers of energy-and nutrient-catabolism [23][24][25] , which is why they are strongly implicated in the development and treatment of the metabolic syndrome 26,27 . Further associated with metabolic regulations, and upon Ppar α signaling, expression of FGF-21 takes place in the liver and adipose tissue 28 , from where it reaches the circulatory system. In animal studies, administration of this factor has been shown to positively impact metabolism in obese mice 29 and diabetic monkeys 30 . Furthermore, serum FGF-21 has been suggested as a potential cardio-metabolic biomarker for humans 28 .
So far, there is no literature available on a possible role for the UGT1A1 * 28 polymorphism and/or elevated levels of UCB in these complex regulations. As yet, only one recent study 31 points out a connection between circulating bilirubin (BR) and Ppar α, using an in vitro approach. This very report expands the field of known physiological bilirubin functions and activities, including antioxidant 32 , immune-modulating 33 and signalling effects, the latter of which have been investigated in terms of Ppar activities in BR-treated mice. Results attest to an insulin-signalling effect of BR, ultimately modulating body weight, that appears to be in parts mediated through Ppar γ 26 .
Against this background, an observational case control study involving 120 healthy age-and gender-matched male and female subjects with and without GS, was conducted. The main aim was to further explain these striking metabolic differences, mainly reflecting in beneficial body composition, glucose-and lipid profile, as well as in apparently altered energetic regulations in response to fasting. To further explore particularities of metabolic regulation in GS, a molecular approach was used focusing on the AMPK pathway.  . Levels of (phosphorylated) proteins (AMPK α1/α2, Ppar α and γ, PgC 1α) were analysed using the method of flow cytometry. Data are expressed as relative fluorescence units [rfU], and compared between subjects with Gilbert's syndrome (GS; UGT1A1*28 promoter mutation), and controls (C). * Indicates significant differences between groups (p ≤ 0.05). Medians can be found in Table 2. Abbreviations: pAMPK α1/α2: Phosphorylated 5′-AMP activated kinase; pPpar α: Phosphorylated peroxisome proliferator activated receptor alpha; pPpar γ: Phosphorylated peroxisome proliferator activated receptor gamma; PgC 1α: Peroxisome proliferator-activated receptor c coactivator 1. In the subsequent paragraphs, features of metabolic health are described, summarized for the entire study population, as well as split into gender groups.
Biomarkers associated with energy metabolism in GS-versus C subjects. All subjects. Median phosphorylation/protein expression of AMPK α1/α2 and of associated downstream transcription factors (pPpar α, pPpar γ, PgC 1α; measured in PBMCs) was significantly higher in GS subjects versus controls ( Fig. 1; p = 0.000). No group difference, however, was found in terms of AMPKα1 gene expression (Table 2).
A trend towards higher FGF-21 serum concentration was found in GS relative to controls. No statistical difference was stated concerning Sirt-1 levels between groups, although values were slightly higher in GS as compared to controls.
Data on body composition differed between the groups, in that BMI levels were significantly lower in GS subjects relative to controls (p = 0.001), and lean body mass (LBM) was higher in GS, however, did not reach statistical significance.
Anthropometric measures were significantly different only in terms of BMI (p = 0.023), which was lower in the GS group. For males, no significant results were obtained for LBM. Female subjects. Phosphorylation of AMPK α1/α2 and its downstream effectors (pPpar α, pPpar γ, PgC 1α) was significantly higher in female GS versus C subjects (p = 0.037, p = 0.000). In summary, this result is retained throughout the gender groups. For AMPKα1 gene expression, again no significant results were found (Table 4). In terms of body composition, both BMI and LBM differed significantly between the groups (p = 0.017, p = 0.011), in that BMI was lower and LBM was higher in female GS versus C.

Biomarkers associated with carbohydrate metabolism in GS-versus C subjects. All subjects.
When considering the entire study population, fasting plasma glucose levels as well as concentrations of insulin and C-peptide, were significantly lower in the GS group, as compared to controls (p = 0.004, p = 0.001) ( Table 2).
Male subjects. Again, as stated for the entire study group, fasting glucose levels were significantly lower in male GS versus C, as were insulin and C-peptide concentrations (p = 0.016, p = 0.009, p = 0.001) ( Table 3).
Female subjects. Differences in parameters of glucose metabolism did not reach statistical significance between the study groups (Table 4).
Biomarkers associated with lipid metabolism in GS-versus C subjects. All subjects. Plasma TG levels were significantly lower in GS subjects (p = 0.045), LPA2 by trend was higher in GS. The remaining lipid parameters (as listed in Table 2) did not differ significantly between the groups ( Table 2).

MALES Variable
Mean° (±sd)/median^ (IQR)  Table 3. Energy-, carbohydrate and lipid metabolic biomarkers of male subjects of the BiliHealth study. (Table 3 summarises and compares (GS versus C) biomarkers of energy-, carbohydrate and lipid metabolism of male subjects of the BiliHealth study. Values are specified as applies according to distribution of data. For parametric variables, means° (±sd) are shown, for non-parametric data, medians^ (50 th percentiles) and interquartile range (IQR) are displayed. Comparison of means (for parametric data) or of ranks (for non-parametric data) was completed using independent samples t-test or Mann-Whitney-U-test (*p-value: significant on a 5% level of significance; T p-value: trend on a 10% level of trend). Abbreviations: GS: Gilbert's syndrome; C: Controls; pAMPK α1/α2: Phosphorylated 5′-AMP activated kinase; pPpar α: Phosphorylated peroxisome proliferator activated receptor alpha; pPpar γ: Phosphorylated peroxisome proliferator activated receptor gamma; PgC 1α: Peroxisome proliferator-activated receptor c coactivator 1; AMPK α1 expr.: AMPK gene expression as relative quantification, RQ to cDNA pool; Sirt-1: Male subjects. Those significant results reported above for both genders in terms of plasma TG and LPA2, were in retained in male subjects. LPA2 was again higher and TG levels were significantly lower in male GS (p = 0.048, p = 0.023). For the remaining parameters (as listed in Table 3), no significant results were obtained (Table 3).
Female subjects. As for the plasma lipid fractions, results for females are different to those presented for males. In female GS, LDL was lower by trend relative to C, as was the LDL/HDL ratio. Also Apo B trended to be lower in female GS as compared to controls (Table 4).

Inter-variable connection of metabolic parameters and the AMPK-pathway, is merely established via UCB and the underlying UGT1A1 genotype (-TA repeats). For the reason of statistical validity
and power, the entire dataset was analysed for inter-variable connections. A graphical summary of the correlations found is presented in Fig. 2, generating an idea as to how the parameters analysed could be networking on a physiological level. A detailed list of significant inter-variable associations, expressed as correlation coefficients (R) with corresponding p-values (in brackets), is provided in the figure.
To begin with UCB as the crucial determinant of the study design, and main feature of GS, strong correlations were found with pAMPK α1/α2, (R = 0.507, p = 0.000), and with downstream transcription factors (Fig. 2). A similar association was found when looking at the UGT1A1 genotype, representing the underlying molecular background of GS (R = 0.412, p = 0.000). Gender-specific associations between UCB/UGT1A1 and AMPK regulations, are visualized in Fig. 3. Interestingly, stronger correlations were generally found in women than in men.  Not surprisingly, UCB levels and UGT1A1 genotype have been found to be strongly correlated (R = 0.731, p = 0.000). For these two specific features of GS (UCB and UGT1A1 genotype), negative correlations were observed for a series of important lipid-and glucose biomarkers (as are listed in Fig. 2), emphasizing an improved metabolic state as UCB levels/TA-repeats increase.

Mean° (±sd)/median^ (IQR)
The same negative connection applies to UCB/UGT1A1 genotype and the anthropometric measure of BMI (R = −0.274, p = 0.002), whereas conversely, a positive association was found with LBM (R = 0.217, p = 0.019). Pursuing the interplay of anthropometric measures and other parameters, revealed connections of BMI and LBM with markers of lipid and carbohydrate metabolism (Fig. 2). Furthermore, an association of LBM with Sirt-1 was found (R = 0.242, p = 0.017), the latter of which statistically looping back to carbohydrate metabolism (HbA1c; R = −0.235, p = 0.019). HbA1c was also negatively associated with Ppar γ (R = −0.204, p = 0.028), one of the downstream effectors of the AMPK pathway, emphasizing its direct connection to energy-and carbohydrate metabolism (Fig. 2).
In summary, those correlations found point to close connections between characterising features of GS, body composition and an altered metabolic state in this condition, altogether having strong implications for macronutrient metabolism (carbohydrate and lipid), and importantly for energy turnover.
Heat maps visualizing correlated variables with respect to energy-, glucose-and lipid metabolism can be found on the online supplementary information (supplementary Figures S2 a, b, c). It all starts with UCB: bilirubin has statistical explanatory power for AMPK pathway regulations and body composition. To further pursue inter-variable connections, and to explore possibilities as to how those entities studied could explain each other, stepwise linear regression models were generated. A graphical abstract of the most important findings, can be found in Fig. 4, where percentages (based on corrected R 2 regression coefficients) specifying inter-variable explanatory power, are presented. Furthermore, tables summarising all relevant correlations that were found, are provided (Tables 5 and 6).
Most compelling, and in line with the findings from bivariate correlation analysis, UCB has noteworthy explanatory power for AMPK phosphorylation (pAMPK, 19%), and, although not as pronounced, for related pathway characteristics (pPpar α, 2.8%). Importantly, UCB furthermore connects the AMPK-pathway with body composition, by in parts explaining the variable BMI (7.2%), and possibly providing an important explanation for the significantly lower BMI stated in GS subjects, relative to controls (Table 1).
Interestingly, however not surprisingly, measures of body composition (BMI, LBM) appear to be merely linked to parameters of lipid metabolism, thereby likely explaining the improved lipid status determined in GS versus control-subjects in this (Tables 2-4), and previous studies 6,7 .
An entirely new inter-variable dependence was found, in that LBM had some explanatory power for Sirt-1 (6.9%), an important controller of metabolism with respect to ageing.
The measure of LBM was furthermore interlinked with PgC 1α, being the immediate activator of Ppar α and γ. This is clearly emphasized by its substantial explanatory power for the latter two (74.3% and 49.6%, respectively). Interestingly, there seems to be also an inverse correlation between these variables, suggesting a feedback-loop from Ppar α/γ to PgC 1α (78.1%).
As is known from the literature 37,38 , and newly reported here for the matrix of PBMCs, the AMPK pathway via its effectors Ppar α and γ, is ultimately linked to glucose metabolism, thereby likely explaining the relatively improved glucose metabolism generally determined in GS subjects (Tables 2-4). On a larger scale, this result provides a mechanism for the known low prevalence of type II diabetes among subjects having GS 8 .
Measures of lifestyle (physical activity, frequency of consuming specific foods) did not have significant further influence on variables of the AMPK pathway, as has been confirmed using regression analysis.

Discussion
The aim of this study was to establish a theory to explain the compelling differences repeatedly found between GS and control subjects, concerning body composition and overall metabolic health. The AMPK pathway was  the investigated model of choice, elegantly networking and determining features of energy-and macronutrient metabolism on a cellular and whole-body level. Specific findings of the study at hand are discussed and summarized in the following paragraphs. The obtained results provide robust evidence, that the energy-and macronutrient metabolic response to fasting are clearly boosted in GS. Accordingly, even though all subjects were metabolically healthy and within the reference ranges concerning their blood biochemistry parameters, several inter-group differences were found, confirming the improved metabolic health status of GS individuals. The relative extent to which these metabolic shifts are due to a direct effect of elevated UCB levels or based on a more complex genetic association with the UGT1A1*28 promoter mutation, remains to be clarified.
As stated in the materials and methods section, all 120 subjects were required to fast on the day before blood sampling (400 kcal restriction), as well as overnight for 16 (±1) hours. This is important to mention since serum UCB is known to rise in response to fasting, and was found to be the main determining factor of post-translational AMPK α1/α2 activation, as well as of BMI. This important feature of body composition was significantly lower in GS individuals.

AMPK -it all starts with physical stress.
The tight association of UCB and AMPK α1/α2 activity is not surprising and was expected, since both effectors are triggered by fasting leading to a drop in ATP, and by other physical stressors including exercise. On a molecular level, these common influencing variables emphasize a potential molecular cross-talk, thereby connecting GS genetics with AMPK α1/α2 activation. The importance of an altered fasting response specifically in GS, is further highlighted by the fact that levels of AMPKα1 gene expression were equal between the groups. With reference to the influencing factor of physical activity, control subjects (self-reportedly) more frequently engaged in physical activity. In theory this study group (C) should have thus benefitted from the various health-benefits known to arise from exercising, and AMPK activity should have been higher. However, the obtained results reported an opposing trend, in that AMPK phosphorylation was significantly higher in GS versus controls. Therefore AMPK α1/α2 was more re-/active, with significantly improved parameters of metabolic health, including glucose, C-peptide, insulin and TG in GS subjects. Interestingly the parameter of TG in fact was found to be the "universal" statistical connector and determinant of the pathways explored, in that it had explanatory powers for C-peptide, insulin, and all fractions of cholesterol (TChol, HDL, LDL).

AMPK from a different angle -metabolic interplay in GS.
With reference to the observed phenomenon of lower plasma lipid fractions, it is well established that Ppar α and γ trigger pathways that are involved in lipogenesis and lipid storage 39,40 . When activated through ligand-binding or phosphorylation (e. g. through AMPK), these pathways lead to body-wide redistribution of fat, consequently lowering TG levels, and thereby improving insulin sensitivity. This effect was clearly present in male GS individuals, expressed through significantly lower TG levels, along with slightly lower TChol, and an improved glucose metabolism. This confirms the above role for Ppars as ultimate regulators of lipid metabolism (ultimately influencing that of glucose), and implies an even more pronounced metabolic effect in GS individuals, likely based on their more re-/active AMPK pathway. These important results readily connect to the low prevalence of metabolic diseases previously stated for GS individuals 6,7 . The significantly lower TChol levels previously stated for GS individuals 5 , and the slightly lower TChol levels reported for GS individuals in this study, could be predicated once more on this group's increased AMPK activity. AMPK has a known post-translational deactivating effect on HMG-CoA reductase (3-hydroxy-3methyl-glutaryl-CoA reductase; not measured in this study), which is the rate-limiting enzyme in cholesterol synthesis 41 . This may be hypothesised to contribute to the slightly lower TChol levels in the GS group.
With reference to aspects of glucose regulation, a statistical association of Ppar γ activity (together with glucose), and the long term glucose parameter HbA1c was found. This is in agreement with results from animal studies in which an indirect BR-mediated insulin effect through the direct Ppar γ agonist HO-1 26 , and an insulin-sensitizing activity of BR based on anti-inflammation 42 have been reported. The overall effects and proposed underlying mechanisms were similarly found in the present study, in that GS individuals (like BR-treated mice in the above studies), were generally lighter, leaner and as mentioned, had improved glucose parameters relative to controls.

Key players in body composition -UCB, Ppars and anti-inflammation. Excess body mass is a known
crucial regulator of glucose homeostasis, lipid metabolism 43,44 and inflammation 45 . An increased fat mass triggers inflammation, involving the production and release of TNFα and IL-6 45 . Inversely, the comparably higher LBM that was determined in female GS individuals, as well as the lower BMI reported for all GS individuals, both would have an easing impact on inflammation. In the present study, it is evident that entities of metabolic health are apparently influenced to a large extent by BMI. This marker was in part explained by UCB levels, thereby connecting AMPK activity with body composition (Fig. 4). C-reactive protein (CRP), as well as TNFα and IL-6 were in fact (significantly) lower in GS individuals as compared to controls (results presented in ref. 36). This emphasizes the above hypothesis of a possible involvement of BR (UCB in the present study), in lowering inflammation through an increase in energy turnover (AMPK phosphorylation), and provides a link to improved body composition. Considering the fact that all study participants were free of (inflammatory) diseases, this result is particularly remarkable. With reference to body composition another interesting observation was made in that LBM had explanatory power for Sirt-1, a known controller of metabolism with specific relevance to ageing. This result could further bridge the gap towards explaining the epidemiological evidence for longevity in GS, which has been experimentally explored recently 36 .
More detailed statistical analyses into body composition and its connection to energy metabolism revealed an interesting gender-specific effect, which to date cannot be definitively explained. It is, however, possibly based on the gender-specific difference in oestrogen levels, that are known to influence energy metabolic pathways 46,47 . As mentioned, LBM was significantly higher in GS individuals (relative to controls) only in females, and the beneficial difference in BMI between the two female groups was more pronounced as compared to that between the male groups (GS versus C). Ultimately connecting these results to energy turnover, they are readily confirmed by the generally stronger correlations between the AMPK pathway and increasing -TA repeats and UCB levels, found in women as compared to men (Fig. 3a,b). These observations are particularly remarkable in view of the relatively smaller female versus male group sizes.

Summary
In conclusion, the AMPK pathway not only is a master regulator of (energy) metabolism and main crossroad of various pathways, it furthermore seems to be a powerful switch that in GS more readily reacts to fasting, possibly leading to an increased energy turnover in this condition.
In this study, (i) not only those beneficial metabolic features were confirmed that had been established previously for GS individuals, but (ii) also the new finding of an apparently boosted AMPK pathway in GS in response to fasting, was presented for the first time. To this end, it cannot be estimated to which extent a potential increase in energy turnover in GS individuals is based on adaptive thermo-and mitogenesis, and possibly adipose tissue browning.  Table 6. Stepwise linear regression analysis with reference to metabolic pathways (Fig. 2). Corrected However, our findings expand and complement data previously obtained from animal studies 26,42 , and propose a precise connection point to pursue future investigations into the molecular background and particularities of metabolic regulation in GS. Specific approaches could include immune-precipitation analyses to assess potential direct binding of BR as Ppar agonist, or microarray/genome wide association studies (GWAS) to screen for SNP-SNP associations or for SNP-interactions with certain phenotypic characteristics and aspects of metabolism in GS.

Materials and Methods
Subjects and study design. This study (abbreviated "BiliHealth") was designed as an observational case-control study, at a single centre in Vienna, Austria. The study was performed at the Department of Clinical Pharmacology at Vienna General Hospital, and subjects were recruited between June 2014 and January 2015, by direct advertising (bulletin boards, posters and flyers) and from the department's subject database.
One hundred twenty-eight (128) healthy subjects between 20 and 80 years of age were initially recruited from the general Austrian population. Eight thereof, had to be excluded for medical reasons. Exclusion criteria included smoking, excess drinking, routine intake of medications and nutritional supplements, pregnancy, acute and chronic (inflammatory/metabolic) diseases, liver diseases, present or past neoplasia and organ transplants. After providing their signed written consent form, each subject completed an initial health check-up (fasting blood biochemistry including levels of unconjugated bilirubin (UCB) and liver enzymes, blood pressure, body weight/-height, questionnaires).
A total of 80 males and 40 females completed the study. This gender distribution is representative of the occurrence of GS in the general population 48 . All subjects were age-and gender-matched, and study group allocation (GS, C) was based on the subjects' respective fasting serum UCB concentrations (</≥17.1 μM) 48 , that had been analysed using HPLC. For the most part, subjects with GS (in contrast to C) showed visible signs of mild jaundice, reflecting in a yellowish pigmentation of the skin and the conjunctival membranes over the sclerae. Liver parameters and parameters of haemolysis were within the normal ranges. Participants were furthermore allocated to age groups (</≥35 years of age). For a graphical summary of the study design refer to supplementary Figure S1.
For the purpose of diagnosing GS, all subjects of both study groups were required to fast on the day before participating in the study, and therefore had to follow a 400 kcal fasting protocol 4,49 . Furthermore, a complete overnight fast of 16 (±1) hours was required, before the day of blood sampling.
Characteristics of the study population, including age distribution, UCB levels and aspects of lifestyle, are summarized in Table 1. Blood biochemistry (whole blood, plasma, serum). For each subject, fasting blood samples were collected on a single occasion (baseline), no longer than two weeks from the entry health check-up. Samples were drawn by venepuncture into EDTA, Li-Heparin and serum tubes (K 2 EDTA, Li-Heparin and Z Serum Sep, respectively). Samples were cooled and protected from light until being analysed or aliquoted. Aliquots were stored at −80 °C until further analysis.

UCB measurement (HPLC) in serum.
For a detailed analysis of UCB (isomers), the method of HPLC was applied (after) 50 , as had been used and published by our group 4,51 and others 52 previously. Briefly, fasting serum samples (stored light-protected in amber vials) were diluted in isocratic mobile phase (methanol, water, n-dioctylamine and acetic acid) and centrifuged. Supernatants were run on a chromatograph (Merck, Hitachi, LaChrom), equipped with a photodiode array detector (PDA, Shimadzu) and a Fortis C18 HPLC column (4.6 × 150 mm, 3 μm), with a Phenomenex C18 HPLC guard column (4 × 3 mm). Sample preparation and analysis followed the previously published protocol 4 . Unconjugated bilirubin (Frontier Scientific Europe, Carnforth, Lancashire, UK) served as an external standard/quality control. As an internal standard, a reference serum sample was run in each analysis. UGT1A1 Genotyping (-TA repeats in UGT1A1*28 promoter region). For UGT1A1 genotyping purposes, DNA was extracted from whole blood, using QIAsymphony SP automated system with QIAsymphony DSP DNA Midi Kit (QIAGEN), as instructed.
Analyses were performed as described elsewhere 53 . Primers and probes were used as 10 μM working solutions. LightCycler FastStart DNA Master HybProbe Mix (Roche) was used on a LightCycler 480 Instrument II (Roche). Alleles were determined according to the melting curves obtained.
Anthropometric measurements. Standing height (subjects without shoes and in relaxed upright position) was measured with a commercial stadiometer, to the nearest 0.5 cm. Body mass (subjects barefooted and lightly dressed) was assessed to the nearest 0.1 kg, using digital scales. The body mass index (BMI) was calculated following the equation BMI = body mass [kg]/(body height [m 2 ]) 2 . To determine body composition, Bioelectric Impedance Analysis (BIA) was used, providing reliable data of body composition 54 , and was performed in the mornings of the study days, using a BIA Analyser 2000-S (Data-Input GmbH, Darmstadt, Germany).
Lifestyle assessment of subjects. All participants were required to answer questions about their lifestyle, including (everyday) activity, exercise/training, drinking and eating habits. For this purpose, food frequency and lifestyle questionnaires were completed by each subject.
Indices of food intake were calculated for the reported weekly frequency of health food, snack food and red meat as well as alcohol consumption, and statistically analysed. "Health foods" included data on foods rich in vitamins, antioxidants, unsaturated fatty acids and fibres; "snack foods" referred to fatty and sugary energy-dense snacks; "red meat" included specifications on intake of red meat and meat products; "alcohol" referred to weekly alcoholic beverage consumption.
For details on weekly frequency of bodily activity, indices on overall activity, endurance exercise and resistance exercise were calculated. "Overall activity" included climbing stairs and walking; "endurance exercise" referred to frequency of at least 30 min bouts of endurance training; "resistance training" included reported frequency of resistance exercise (using own body weight and/or weights) per week.
Flow cytometric (FACS) analyses of pAMPK α1/α2, PgC1 alpha, pPpar alpha and -gamma in PBMCs. Active (phosphorylated) intracellular protein concentrations were measured in peripheral blood mononucleated cells (PBMCs). Cells were extracted from EDTA whole blood immediately after blood samplings. Density gradient centrifugation using separation tubes (Leucosep TM , Greiner bio one GmbH, Austria) was applied as instructed. Following isolation, cells were washed twice with ice-cold PBS. Cell count and viability were assessed using the trypan blue exclusion assay on an automated cell counter (Countess TM , Life Technologies). For short-term storage, cells were aliquoted in freezing medium (FBS + 10% DMSO) and gradually cooled (1 °C/min) to −80 °C, using the CoolCell TM system (Biozym).
All FACS analyses were completed on a four-channel FACS Calibur TM flow cytometer (BD, Europe). Signal compensation (using Calibrite TM beads and FACS Comp software, BD) was successfully completed prior to each experimental run.
The synthesis of cDNA from RNA was performed using the High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor 1000 Reactions (Applied Biosystems) on a Biometra thermocycler. Concentration and quality of cDNA first strand was determined by NanoDrop 2000c spectrophotometer.
For qPCR 10ng cDNA samples were used on commercially available TaqMan assays (Life Technologies) as single-plex reactions. The TaqMan assay for PRKAA1 (assay Hs01562315_m1, FAM-MGB-labelled) gene was used according to manufacturer instructions. ACTB (assay Hs99999903_m1) and GAPDH (Hs99999905_m1) were used as endogenous controls.
The assays were performed using TaqMan Universal PCR master mix on a QuantStudio ™ 6 Flex Real-Time PCR System (Thermo Fisher), on a 384-well block. All samples were run on the same 384-well plate in a single run, to avoid inter-plate variations. All samples were run in triplicate, and samples with a standard deviation higher than 0.5 Ct-units were excluded from further analyses. Relative quantification was performed using the RQ (Relative Quantification) feature on the Thermo Fisher Cloud qPCR analysis software with ACTB and GAPDH as endogenous controls and a pooled cDNA sample as a reference sample.
ELISA measurement of FGF-21 and Sirt-1 protein levels in serum. Fibroblast growth factor-21 (FGF-21) as well as Sirtuin-1 (Sirt-1) were measured in serum using respective ELISA kits (FGF-21 Human ELISA Kit, ab125966; Human SIRT1 ELISA Kit SimpleStep, ab171573; both Abcam), and following manufacturer instructions. Plate assays were run in a 96-well plate format, using a BMG FLUOstar OPTIMA microplate reader (BMG LABTECH GmbH), set to absorbance mode (450 nm). All samples and external standards were run in duplicate. Data distribution was checked using Kolmogorov-Smirnov (K-S test) and histograms. For comparison of means (for parametric data), ANOVA was used, for comparing medians or ranks (for non-parametric data), Mann-Whitney-U-test was selected. Data are summarized as is appropriate according to their respective distribution. For parametric data, means ± sd (standard deviation), for non-parametric variables, medians ± IQR (inter-quartile range) are presented. Bivariate correlations were modelled using Spearman correlation (Spearman's rho). Correlation coefficients (R) and p-values are presented. Regressions were calculated by applying the model of stepwise linear regression. Corrected R2 coefficients and p-values are presented. For all statistical measures, the level of significance was set to be 5% (p ≤ 0.05).