Voluntary exercise in mice fed an obesogenic diet alters the hepatic immune phenotype and improves metabolic parameters – an animal model of life style intervention in NAFLD

Reproducible animal models to recapitulate the pathophysiology of non-alcoholic fatty liver disease (NAFLD) are urgently required to improve the understanding of the mechanisms of liver injury and to explore novel therapeutic options. Current guidelines recommend life-style interventions as first-line therapy for NAFLD and these types of intervention are considered standard-of-care. The current study establishes a reproducible mouse model of a life-style intervention in NAFLD using voluntary wheel running (VWR). Male C57BL/6J mice were fed a high-fat, high-carbohydrate diet (HFD) to induce NAFLD or a corresponding control diet for 12 weeks. Starting at week 9 of the obesogenic NAFLD diet, mice were randomized to either free access to a running wheel or being single caged resembling a sedentary (SED) life-style. VWR induced a transient weight reduction in HFD-fed mice up until week 10. In contrast to the SED mice, VWR mice exhibited normal ALT at the end of the intervention, while the metabolic alterations including elevated fasting glucose, insulin, triglyceride, and total cholesterol levels remained almost unchanged. Additionally, VWR prevented HFD-induced hepatic steatosis by alterations in key liver metabolic processes including the induction of fatty acid β-oxidation and lipogenesis inhibition following increased AMP-activated protein kinase (AMPK)-α activity. Phosphorylation of the serine kinase Akt in hepatic tissue was enhanced following VWR. Furthermore, VWR mice were protected from HFD-induced expression of pro-inflammatory cytokines, chemokines and liver macrophage infiltration. The SED/HFD group exhibited increasing activity of hepatic nuclear factor (NF)-κB p65, which was absent following exercise in the VWR/HFD group. In summary, in an obesogenic mouse model of NAFLD physical exercise improves fatty acid and glucose homeostasis and protects from macrophage-associated hepatic inflammation.

Scientific RepoRts | (2019) 9:4007 | https://doi.org/10.1038/s41598-018-38321-9 disease or stroke 3 . Early treatment of NAFLD is crucial in order to prevent disease progression and associated liver-related and cardio-vascular mortality. Current US and EU-guidelines recommend weight loss through dietary changes and physical exercise 4,5 . This is supported by a series of clinical studies in humans showing a reversal of steatohepatitis and regression of fibrosis with more than 7% of weight reduction [6][7][8] . Interestingly, a reduction of hepatic fat content from exercise occurs even in the absence of weight loss as summarized in a recent review 9 . On the other hand, a subgroup of patients does not improve liver histology, despite weight loss 10 , indicating, that body composition is likely of importance in addition to reducing body weight. A number of animal studies have explored the beneficial effects of exercise. Using a high-fat dietary model, forced exercise was shown to decrease inflammation in the adipose and hepatic tissue 11,12 . Additionally, exercise was shown to reduce hepatic fatty acid synthesis by decreasing lipogenic enzymes and AMP-activated protein kinase (AMPK) 13,14 . Also, mitochondrial inner membrane integrity and fatty acid oxidation are improved from exercise in mice 15 . Only few studies have explored the effect of exercise training on hepatic inflammation and fibrosis. In foz/foz mice, which develop an obese phenotype from hyperphagia, adipose tissue inflammation, hepatic inflammation, fibrosis and muscle insulin sensitivity improved with exercise 16 . In contrast to the most commonly used forced exercise models, the effectiveness of voluntary wheel running (VWR) on the hepatic immunophenotype is less clear. We have previously assessed this type of exercise during acute liver injury 17 , but data relating to obesity-induced NAFLD is currently not available. Therefore, we explored the applicability and effectiveness of VWR compared to sedentary life-style on changes in the hepatic metabolism and immune phenotype in a well-established obesogenic high-fat, high-carbohydrate diet (HFD) model of NAFLD 18 . The aim was to provide a reproducible, well-characterized and applicable mouse model to reflect the current standard of care for NAFLD and which could be used in the context of studying the additional effects of therapeutic drug compounds in addition to life-style.

Material and Methods
Animal model. All animals were bred and held at the animal facility of the University Medical Center Mainz, according to the criteria outlined by the "Guide for the Care and Use of Laboratory Animals". The study was conducted following approval by the committee for experimental animal research (Landesuntersuchungsamt Rheinland-Pfalz) and all experiments were performed in accordance with relevant guidelines and regulations. We employed an obesogenic diet in 8-10-week-old, male C57BL/6J mice by feeding a high-fat diet (35.5% w/w crude fat (58 kJ%), 22.8 MJ/kg = 5.45 kcal/g) and fructose (55% w/v) and glucose (45% w/v) enriched drinking water. Gender-and age-matched controls received a matched control diet (CD; 5.4% w/w crude fat (13 kJ%), 15.7 MJ/ kg = 3.74 kcal/g) and plain water 18 . The composition and energy density of the diets (ssniff Spezialdiäten GmbH, Soest, Germany) are listed in Suppl. Table 1. After 8 weeks of dietary feeding, mice were randomly assigned to a voluntary wheel running (VWR) group or a sedentary (SED) group. The VWR mice (n = 8 mice on HFD and n = 7 mice on CD) were individually housed in cages (size 43 cm in length × 25 cm width and 28 cm height) and outfitted with a 11.5 cm diameter running wheel (Suppl. Fig. 1). Wheel running activity was continuously recorded using a usual bicycle tachometer (Ciclosport, Krailling, Germany). The SED mice (n = 20 mice on HFD and n = 11 mice on CD) were housed in single in corresponding cages without running wheel. The experimental setup is shown in Suppl. Fig. 2. For the duration of the study, all mice were kept on a 12-h light/dark cycle at constant temperature (22 ± 2 °C) and humidity (55 ± 10%) and with free access to the experimental diets and water. Biometric data including body weight and food consumption were measured weekly. After the 12-week experimental period, all mice were kept sedentary and were fasted overnight before sacrifice for collection of blood and liver tissue. serological analysis. Serum  Quantitative real-time pCR. Isolation of total RNA, cDNA synthesis and quantitative real-time PCR (qRT-PCR) were performed as previously described 20 . All samples were performed in duplicate. Roche LightCycler software (LightCycler 480 Software Release 1.5.0, Roche, Mannheim, Germany) was used to perform advanced analysis relative quantification using the 2 (−ΔΔC(T)) method. Expression data were normalized to the housekeeping gene Gapdh (primers from Qiagen, Hilden, Germany) and the mean of SED/CD mice was considered 1. Primer sequences (all primers obtained from Eurofins Genomics, Ebersberg, Germany) are listed in Suppl. Immunoblotting and determination of nuclear factor (NF)-κB p65 DNA binding activity.
Proteins were isolated and separated as previously described 19 . Primary antibodies included: AMPK-α, phospho-AMPK-α (Thr172), Akt, phospho-Akt (Ser473), NF-κB p65, and phospho-NF-κB p65 (Ser536) (all obtained from Cell Signaling Technology, Danvers, MA, USA). Membranes were exposed to anti-rabbit secondary antibodies conjugated with horseradish peroxidase (Santa Cruz Biotechnology, Dallas, TX, USA). PaperPort Professional software v14.0 (Nuance Communications Germany, München, Germany) was used for image acquisition and the Adobe Acrobat Professional software program (Adobe Systems Incorporated, San Jose, CA, USA) was used to cut immunoblot images to size. No post-processing of images was performed and original uncut images blots are provided in Suppl. Figs. 3 and 4. Densitometry was performed using the National Institutes of Health ImageJ software. NF-κB p65 activity was measured in duplicate in whole liver tissue using the TransAM NF-κB Family Kit, according the manufacturer's instructions (Active Motif, Carlsbad, CA, USA) and the mean of SED/CD mice was considered 1.

Flow cytometric analysis of intrahepatic immune cells. Isolation of intrahepatic immune cells and
their flow cytometric analysis were performed as previously described 21 . Quantification of the macrophage population was performed by gating on living CD45 + F4/80 + cells (all antibodies obtained from BioLegend, San Diego, CA, USA). statistical analysis. All statistical analyses were performed using GraphPad Prism 7 software (GraphPad Software, La Jolla, CA, USA). All results were initially submitted to Shapiro-Wilk normality test for normality and to Levene's test for homogeneity of variance. Comparisons between experimental groups were carried out using the unpaired two-tailed Student's t test or the Mann-Whitney U test to determine statistical significance of differences. Results with a p value of <0.05 were considered to be significant. The significance-level α was adjusted using Holm's sequential Bonferroni adjustment in analyses involving multiple comparisons. All data are shown as mean ± standard error of mean (SEM) to determine the precision and differences of means and statistically significant values were assumed with * /$ p < 0.05, ** /$$ p < 0.01, *** /$$$ p < 0.001.
Activity was monitored using a calibrated cycling computer connected to the exercise wheel (see Suppl.  Fig. 1G). In relation to body weight, the running distance was lower in the VWR/HFD group (VWR/HFD (n = 8): 3.3 ± 1.6 km/g BW vs. VWR/CD (n = 7): 11.6 ± 1.4 km/g BW, p < 0.01, Fig. 1H). At an individual level, a large variety of running activity was detected ranging from 9.3 to 454.4 km in the VWR/HFD and 274.9 to 371.7 km in the VWR/CD group over 4 weeks. Distance was independent of starting body weight. 191.3 ± 17.4 mg/dl fasting glucose, p < 0.001; 197.8 ± 6.6 mg/dl vs. 123.6 ± 4.9 mg/dl total cholesterol, p < 0.001) but not triglycerides (64.4 ± 3.6 mg/dl vs. 73.1 ± 9.3 mg/dl, p = ns). Physical exercise did not correct hyperglycemia or total cholesterol and there were no statistically significant differences in these metabolic parameters between the VWR/HFD and SED/HFD group (Fig. 2B). Moreover, physical exercise did not correct fasting   Fig. 3C). While hepatocyte ballooning and inflammation were not significantly reduced following  Fig. 3C). These data suggest that the 4-week VWR-based exercise training was an effective intervention to improve the histological hallmarks -predominantly hepatic steatosis -of NAFLD.
Physical activity altered the expression of the gluconeogenic enzymes phosphoenolpyruvate carboxykinase (PEPCK) and glucose-6 phosphatase (G6Pase) -encoded by Pck1 and G6pc -irrespective of the underlying diet (Fig. 4C), which both are regulators of insulin sensitivity 22 . The differences between VWR and diet-matched SED groups were not significant (Fig. 4C). Fructose-1,6-bisphosphatase 1 (FBP1) -encoded by Fbp1 -was decreased in expression from exercise following HFD-feeding but not CD-feeding (VWR/CD (n = 7): 0. The serine-threonine protein kinase B (PKB/Akt) is activated by serine phosphorylation and controls hepatic glucose metabolism 23 . Densitometry of immunoblots showed increased levels of Ser473-phosphorylation in mice of the exercise groups (Fig. 4D), while no difference was detectable when comparing VWR/HFD and VWR/CD mice (p = ns). Levels of total Akt protein were unchanged (Fig. 4D).

Discussion
Physical exercise is an effective intervention and recommended as first line treatment in patients with NAFLD 5 . The underlying molecular mechanisms that contribute to the beneficial effects of physical exercise related to the hepatic phenotype are incompletely understood. In the literature alterations of immune cell populations and the metabolic phenotype have been described 8,11,12 . In the current study we explored the effects of a short, voluntary exercise protocol and combined this with a well-characterized obesogenic diet that produces NAFLD without severe steatohepatitis or fibrosis. This model well reflects the mild hepatic involvement of an obese, pre-diabetic patient that is most commonly encountered in hepatology clinics 24 . Physical exercise over 4 weeks followed by 8 weeks of an obesogenic diet slowed weight gain transiently but did not revert the metabolic syndrome and fasted glucose, insulin, triglyceride and total cholesterol remained elevated. Exercise induced resolution of histological changes -predominantly of hepatic steatosis -but also hepatocyte ballooning and inflammation. Therefore, we examined the effects of physical exercise on hepatic lipid metabolism. Firstly, we addressed changes in hepatic de novo lipogenesis and catabolic mitochondrial fatty acid β-oxidation. Among the key liponeogenic regulators PPAR-γ, which stimulates the expression of genes that control uptake, trafficking, and lipid synthesis of fatty acids in the liver 25 , was induced from HFD feeding and in addition significantly downregulated from exercise. On the catabolic side, CPT1 -a rate limiting enzyme for entry of long-chain fatty acid into the mitochondria and subsequent mitochondrial β-oxidation -was increased following exercise. Furthermore, HFD-feeding suppressed PGC1-α -a transcriptional regulator of PPAR-α and genes encoding enzymes of mitochondrial fatty acid β-oxidation. PGC-1α has previously been identified as a potent anti-steatotic factor induced from exercise and acts through increasing mitochondrial β-oxidation and disposing of potentially injurious lipid species 26 .
Our data showed variability with exercise however no clear trend was observed. Previous findings suggested that exercise-induced PGC-1α acts to repress SREBP-1c resulting in decreased triglyceride synthesis and secretion. Among the PGC-1α mediated effects, increasing activation of the nuclear receptor FXR has been observed. FXR activation through the synthetic ligand obeticholic acid has been shown to exert anti-inflammatory and anti-steatotic effects in NASH, albeit increasing serum LDL levels 22 . In the current study we explored the expression of FXR-α which was unchanged from activity. Therefore, the addition of an FXR agonist to physical exercise could be of particular use to treat patients with NASH and activate all relevant signaling pathways. The metabolic alterations in exercising mice were at least partially mediated through AMPK. AMPK is known to phosphorylate SREBP-1c at Ser372 residues and to inhibit proteolytic cleavage and nuclear translocation of SREBP-1c which represses hepatic lipogenesis 27 . Likewise, liver-specific activation of AMPK completely protects against hepatic steatosis in mice fed a high-fructose diet through inhibition of de novo fatty acid synthesis 28 . Apart from direct phosphorylation of metabolic enzymes, AMPK also has long-term effects at the transcriptional level and acts to adapt gene expression to energy demands. There is considerable data mostly derived from in vitro assays supporting that AMPK activation leads to increased PGC-1α expression and through this modulates the expression of several key players of mitochondrial function and glucose metabolism 29 . PGC-1α regulates gluconeogenic genes directly through coactivation of key transcription factors, including hepatocyte nuclear factor (HNF) 4α and forkhead box protein O1 (FOXO1), and indirectly involving other factors of insulin sensitivity 30 . Another possible explanation was provided by a study published recently by Hughey et al., which showed that AMPK-α promotes hepatic glycogenolysis, but does not influence gluconeogenesis in exercising mice 31 -a Figure 6. Effects of VWR on NF-κB activation in the liver of HFD-fed mice. (A) Phospho-NF-κB p65 (Ser536) and total NF-κB p65 protein expression were determined in liver lysates from VWR mice compared with SED mice and (B) from VWR mice with increasing individual physical activity levels (increasing from left to right) fed the diets for 12 weeks by western blot, and (C) activated p65 was quantified by a specific DNA-binding ELISA. Data are shown as fold change relative to the mean of SED/CD mice, which was considered 1. In (A,B) representative immunoblots with densitometric analysis are shown. Uncropped images of original blots are shown in Suppl. Fig. 4A,B. Data in C are means ± SEM of n = 5 mice/group. *p < 0.05, **p < 0.01 for CD vs. HFD using two-tailed Student's t-test (A-C). finding consistent with the current observations. Our experimental data suggests that mitochondrial fatty acid β-oxidation in exercise trained mice increases -a mechanism by which hepatocytes provide energy supplies during exercise. In addition to liver-generated glucose, ketone bodies, which are synthesized in the liver form acetyl-CoA derived primarily from fatty acid oxidation, are an essential source of energy for extrahepatic tissues during exercise 32 . These metabolic adaptions during exercise contribute to improved hepatic lipid metabolism and glucose homeostasis in obese mice. Second, key regulators of hepatic insulin signaling were examined. The principle downstream effector of insulin signaling in the liver is the serine-threonine protein kinase Akt, also known as protein kinase B (PKB) 33 . Akt activation mediates the inhibitory effects of insulin on gluconeogenesis and glycogenolysis, in addition to hepatic fatty acid oxidation 27,34 . Metabolic and oxidative stress from steatohepatitis deregulate Akt signaling and promote hepatic insulin resistance 35 . These deleterious effects were previously shown to be reversible from swimming 36 and the current study recapitulates these findings using the voluntary running model suggesting, that physical exercise is capable of priming the liver for treatments that address insulin signaling at the level of the hepatocytes.
Thirdly, we explored inflammation and hepatocellular injury. Clinical trials examining the effects of physical exercise in patients with NAFLD suggested that endurance exercise reduces hepatic inflammation in obese subjects even in the absence of weight reduction 37 . The obesogenic model of NAFLD used in the current study is relatively mild and does not produce major liver injury or hepatic fibrosis 38 . Nonetheless, levels of ALT were elevated in sedentary mice on HFD. This was paralleled by increased expression of pro-inflammatory cytokines including IL-6, IL-1β, TGF-β and the chemotactic MCP-1/CCL2. The expression of these cytokines has been implicated in the progression to severe steatohepatitis and fibrosis 39,40 . Likewise, these pro-inflammatory cytokines impair hepatic insulin sensitivity by negatively interfering with insulin signaling and promote a vicious cycle by inducing more inflammatory cytokines 41 . Physical activity blunted cytokine expression -to the level seen in non-obese mice -and tuned down NF-κB p65 activity. In the literature there is extensive evidence linking AMPK to inhibition of NF-κB-mediated pro-inflammatory signaling either through direct effects or indirectly through downstream mediators including PGC-1α, which can subsequently repress the expression of inflammatory factors 42 . In the current study we extended these functional findings and explored the hepatic immunophenotype from hepatic steatosis and the role of physical activity in altering immune cell composition. This was suggested by the observation that MCP-1/CCL2 -a key immune-chemokine 43 -is suppressed from exercise. Using FACS analysis, we were able to demonstrate an effect of exercise on intrahepatic CD45 + F4/80 + macrophages which are decreasing following 4 weeks of voluntary exercise.
In summary, the beneficial changes of voluntary exercise -highlighted in Fig. 7 -result from improvement of the hepatic metabolism with decreasing liponeogenesis and gluconeogenesis, as well as increasing insulin sensitivity mediated through AMPK and Akt signaling. The running distance correlated with decreasing hepatic injury and improved metabolism. In addition, the inflammatory, intrahepatic immunophenotype was changed as evident by decreased expression of pro-inflammatory, chemotactic and pro-fibrogenic cytokines, reduced NF-κB signaling and dampened macrophage recruitment. In order to maximize therapeutic effects of developing pharmacological treatment, the current study highlights the effect of physical exercise that could be favorable and compensate for short comings of potential monotherapies or augment the beneficial effect of combination drug regimens.

Conclusions
Overall, this study defines the metabolic and immunological mechanisms in a well characterized, short, voluntary exercise model combined with an obesogenic mouse model of NAFLD. Metabolic effects are in parts mediated through activation of hepatic AMPK-α and Akt. Furthermore, physical activity protects against a pro-inflammatory and pro-fibrogenic liver milieu by inhibiting the recruitment of inflammatory macrophages and thus is potentially capable of slowing the progression to NASH.