The daily food intake was virtually identical among the four groups of rats (Fig. 1); however, the high-fat diet caused significant body weight gain compared with the control diet and the body weight of the CH (fat liver) group was significantly greater than that of the CS (control) group at week 12 (p < 0.01) (Table 1). Injection of rAAV2/1-CMV-GH1 resulted in a small but not significant increase in body weight in the GS group compared with the CS group (Table 1). In contrast, in the high-fat diet fed rats, exogenous GH levels reduced the body weight gain compared with the control group (GH vs. CH, p < 0.05) (Table 1), even though food intake was marginally increased in the GH group.

As expected, the high-fat diet increased visceral fat weight and visceral fat percentage in the CH group, compared with the CS group (both, p < 0.001). Meanwhile, chronic exogenous GH improved the body composition and decreased visceral fat (VF) weight and VF percentage (VF%) in the GH group compared with the CH group (p < 0.001 and p < 0.05, respectively). Hepatomegaly, which is common in NAFLD, is determined by liver wet weight (LWW) and the HI. LWW and HI were significantly increased in the CH group than in the CS group (both, p < 0.001). Chronic exogenous GH levels prevented hepatomegaly, because LWW and HI decreased in the GH group compared with the CH group (both, p < 0.01) (Table 1). However, HI was higher in the GS group than in the CS group, although not significantly (Table 1).

GH1 gene expression can be sustained for at least 6 months after the injection of rAAV2/1-CMV-GH1, as we have already reported 20 . Serum hGH was detected in the GS and GH groups. As a result, serum IGF-1 levels were higher in the GS and GH groups than in the CS and CH groups. Interestingly, the high-fat diet decreased the serum IGF-1 levels in the CH group, which was reversed by the injection of rAAV2/1-CMV-GH1 in the GH group compared with the CH group (Fig. 2).

The high-fat diet significantly increased the serum levels of total cholesterol (TC) (p < 0.001), triglyceride (TG) (p < 0.001) and low-density lipoprotein cholesterol (LDL-C) (p < 0.001) and decreased the serum level of high-density lipoprotein cholesterol (HDL-C) in the CH group compared with the CS group. Chronic exogenous GH markedly reduced the serum TC (p < 0.01), TG (p < 0.001) and LDL-C levels (p < 0.01) and increased, albeit not significantly, the HDL-C levels (p > 0.05) in the GH group compared with the CH group, indicating that exogenous GH levels reversed the dyslipidaemia induced by the high-fat diet (Table 2). Similar improvements in lipid profiles were also noted in the GS group compared with the CS group. However, chronic exogenous GH also increased the serum glucose (p < 0.05) and insulin levels, although not significantly, in the GS group relative to the CS groups, and may increase the risk of IR, which was reflected by changes, although not significant, in the homeostasis model assessment of IR (HOMA-IR).

Photomicrographs of hepatic specimens stained with H&E are shown in Figure 3. Lipid accumulation was not observed in the CS group (Fig. 3A). The injection of rAAV2/1-CMV-GH1 viral particles did not cause obvious hepatic changes in the GS group (Fig. 3B). As would be expected, the high-fat diet induced hepatic lipid accumulation in the CH group (Fig. 3C), with a mean grade of 2 (Table 3). The livers from the CH group also showed microvesicular or macrovesicular steatosis around the periportal zone, necrosis, and inflammation, along with enlarged hepatocytes (Fig. 3C). Fat deposition in this group was classified as mixed. Similar degenerative changes were noted in the GH group (Fig. 3D), but to a much lesser extent; the grade of lipid accumulation was 1, which was significantly lower than that in the CH group (p < 0.05). Fat deposition in the GH group was classified as microvesicular. Inter-observer agreement was 0.86.

Serum TNF-alpha levels were remarkably elevated in the CH group compared with the CS group (p < 0.001). Although exogenous GH levels did not affect the serum TNF-alpha level in the GS group, it did partly prevent the high-fat diet-induced increase in serum TNF-alpha level in the GH group, as the serum TNF-alpha level was significantly lower in the GH group than in the CH (p < 0.001) (Table 4). The hepatic MDA level was also significantly higher in the CH group compared with the GH group (p < 0.001), as well as the CS and GS groups (Table 3). The hepatic levels of superoxide dismutase (SOD) (p < 0.01), glutathione peroxidase (GSH) (p < 0.001) and catalase (CAT) (p < 0.001) were significantly greater in the GH group compared with the CH group. However, hepatic nitric oxide synthase (NOS) (p < 0.001) was significantly lower in the GH group than in the CH group. Therefore, chronic exogenous GH significantly decreased the effects of stress oxidative in the liver (Table 4).

Bacterial translocation from mesenteric lymph nodes (MLNs) is considered one of the main events in the pathogenesis of spontaneous bacterial peritonitis and other infections in cirrhosis 21 . TNF-alpha is known to be involved in bacterial translocation in rats with cirrhosis. Therefore, we tested whether bacterial translocation acted as a stimulus for TNF-alpha production in our model. In fact, the MLN cultures were negative in all four experimental groups, which indicates that bacterial translocation was not responsible for the elevated levels of TNF-alpha.

The full-length leptin receptor isoform, ob-rb, contains intracellular motifs required for the activation of the JAK/STAT signal transduction pathway, and is considered to be the functional receptor 22 . The effects of adiponectin are principally mediated through the R1 and R2 adiponectin receptors (adipoR1 and adipoR2), and adipoR2 is predominantly found in the liver 23 . Figure 4 shows the differential expression of ob-rb, resistin, and adipoR2 mRNA among the four groups of rats. The high-fat diet significantly decreased hepatic ob-rb (p < 0.001) and adipoR2 (p < 0.001) mRNA expression and increased resistin mRNA expression (p < 0.001). Meanwhile, exogenous GH levels significantly increased hepatic ob-rb and adipoR2 mRNA expression (both, p < 0.001) and decreased hepatic resistin mRNA expression (p < 0.01) in the GH group compared with the CH group (Fig. 4).

The high-fat diet induced marked changes in the expression of a group of signal transducers and activators in the CH group compared with the CS group, which were reversed by exogenous GH. Figure 5 shows the changes in expression of hepatic signal transducers and activators in the four different groups. Western blotting revealed that the significant inductions in the p-JAK2 (p < 0.01), p-STAT3 (p < 0.001), p-STAT5 (p < 0.01), p-AMPK-alpha (p < 0.001), p-ERK2/1 (p < 0.001) and p-PPAR-alpha (p < 0.01) relative to the expressions of the corresponding total proteins in the GH group compared with that in the CH group. Furthermore, the relative protein expression of p-P38 MAPK (p < 0.001) and p-JNK (p < 0.001) relative to those of P38 MAPK and JNK was significantly lower in the GH group than in the CH group (Fig. 5). Collectively, these results indicate that changes in the expression of signal transducers and activators induced by the high-fat diet could be reversed by exogenous GH levels.

We used a previously described method to construct the rAAV2/1 vector containing GH1 20 23 45 46 . Briefly, GH1 was cloned from a PCR product using 5'-CAGAATTCGCCACCATGGCTACAGGCTCCCGG-3' (forward primer) and 5'-CTGCGTCGACGAAGCCACAGCTGCCCTC-3' (reverse primer) (EcoRI and SalI restriction sites are indicated in bold) from the template of a pUC19 plasmid DNA containing GH1 (Xinxiang Medical University, Henan Province, P. R. China). The GH1 DNA fragment (677 bp, including the 651-bp cds) was digested with SalIand EcoRI and inserted into the SalI and EcoRI sites of the pSNAV2.0 vector (AGTCGene Technology Co. Ltd., P.R. China). rAAV2/1 production and purification were performed as previously described 47 . The viral genome particle titre (1.0 × 10¹² v.g./ml) was determined by a quantitative DNA dot blot method 48 .

Animal experiments were performed in accordance with the guidelines of the National Institutes of Health (Bethesda, MD, USA) and the Chinese People's Liberation Army General Hospital for the humane treatment of laboratory animals. All efforts were made to minimize the number of animals used and their suffering.

Thirty-two adult male Sprague-Dawley rats (180 ± 10 g) were obtained from the Institute of Laboratory Animal Sciences, Chinese Academy of Medical Sciences (CAMS) & Peking Union Medical College (PUMC) (Beijing, China). Rats were housed at 23°C with a 12-h light/dark cycle, and were allowed free access to food and water. Half of the rats were randomly selected and intravenously injected with a single dose of 1.25 × 10¹¹ rAAV2/1-CMV-GH1 viral particles into the tail vein. The remaining (control) rats were injected with a single dose of empty rAAV2/1 vectors. Two weeks later, the GH1 gene-treated rats were divided into two groups and fed either a standard (control) diet (4.5 g/100 g fat) (GS group, n = 6) or a high-fat diet to induce fatty liver (GH group, n = 10) consisting of 10% hog fat, 2.5% glucose, 2% cholesterol and 0.25% cholic acid added to normal chow for 10 weeks, as previously described 6 . Similarly, the control rats were also divided into two groups and fed the standard (control) diet (CS group, n = 6) or the high-fat diet to induce fatty liver (CH group, n = 10). Food intake and body weight were recorded during the experimental period.

At week 12 (i.e., 2 weeks after injection followed by 10 weeks of feeding), the rats were fasted overnight and sacrificed by anaesthesia with 10% chloral hydrate solution. Body length (distance from the nose to the anus [N-A distance]) and total body weight (TBW) were measured by an electronic balance (Scout Pro Balance; Ohaus, Pine Brook, NJ, USA) that was calibrated every day and an electronic digital calliper (Control Co., Friendswood, TX, USA). Body mass index (BMI) was calculated using the formula BMI = TBW/(BL)². The perirenal and epididymal fat pads were pooled (visceral fat, VF) and weighed using a precision electronic balance (AV264; Ohaus). The VF weights of each rat were normalized to TBW by calculating the percentage as follows: weight/TBW × 100%. Fresh liver tissues were immediately dissected and some samples were prepared for histological analyses, while others were weighed and snap-frozen between blocks of dry ice and stored at -80°C. The HI was calculated as liver weight/TBW × 100%.

Blood was collected via the aorta and allowed to clot at room temperature for 60 min to form serum. The serum samples were then centrifuged at 3000 × g for 15 min at 4°C, and stored at -80°C until used for biochemical and hormone assays.

The freshly dissected rat livers were immediately fixed in 4% paraformaldehyde, dehydrated, embedded in paraffin, and sectioned. Formalin-fixed, paraffin-embedded sections were cut (5 μm thick) and mounted on glass slides. The sections were deparaffinized in xylene, stained with haematoxylin and eosin (H&E) using standard techniques, and examined by an investigator blind to the treatment group. Biopsies were classified into four grades based on fat accumulation using the classification method devised by Brunt et al 49 , where grade 0 indicates no fat present in the liver, while grades 1 (light), 2 (mild) and 3 (severe) were defined as the presence of fat vacuoles in < 33%, 33-66% or >66% of hepatocytes, respectively. The pattern of fat deposition was classified as macrovesicular, microvesicular or mixed. Two experienced pathologists blinded to the experimental conditions evaluated all samples. Agreement between both pathologists was determined.

Serum human GH (hGH; Roche, Pleasanton, California, USA), TNF-alpha (R&D Systems, Minneapolis, MN, USA), IGF-1 (ADL, Alexandria, VA, USA) and insulin (ADL) was measured using enzyme-linked immunosorbent assay (ELISA) kits. All assay kits included quality controls. Each sample was assayed in duplicate. The serum levels of glucose, alanine aminotransferase (ALT), aspartate aminotransferase (AST), TG, TC, and HDL-C were measured using standard methods. LDL-C level was calculated using Friedwald's formula 50 . IR was assessed using HOMA-IR as blood glucose × blood insulin/22.5 51 52 .

To measure the level of hepatic MDA, 25 mg of liver tissue was added to 250 μl of radioimmunoprecipitation assay (RIPA) buffer containing protease inhibitors. This mixture was sonicated for 15 s at 40 V over ice and centrifuged at 1600 × g for 10 min at 4°C. The supernatant was used for analysis. MDA was quantified using the thiobarbituric acid reaction as described by Ohkawa 21 53 , and measured using a thiobarbituric acid reactive substances (TBARS) assay (Cayman Chemical Co. Inc., MI, USA). The hepatic levels of SOD, CAT, GSH and NOS as oxidant/antioxidant biochemical parameters were also quantified using appropriate assays (Cayman Chemical Co. Inc.).

Samples of MLNs and portal and peripheral blood were collected under sterile conditions before the rats were killed. The MLNs were homogenized in physiological (0.9%) saline, and 0.1 ml aliquots of the homogenate were cultured in MacConkey agar (Oxoid, Basingstoke, UK), Columbia sheep blood (Oxoid), and Esculin-Bile-Azide agar (Merck, Darmstadt, Germany), and incubated at 37°C for 48 h. Bacterial translocation was defined as a positive culture of MLNs. Systemic infections were defined as a positive culture of any of the other biological samples, as previously described 54 .

Total RNA was extracted from the livers of each group of rats and isolated and purified with TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and NucleoSpin^® RNA clean-up kit (Macherey-Nagel, Duren, Germany). The mRNA analysis was carried out by quantitative real-time RT-PCR using LightCycler technology (Roche Diagnostics) for continuous fluorescence detection. The primers for rat leptin receptor isoform b (ob-rb) 55 56 , adipoR2 23 and resistin 57 were used as previously reported and are summarized in Table 5. The reactions were performed using a LightCycler-FastStart DNA Master SYBR Green I Kit (Roche Applied Science) using 50 ng of total RNA extracted from the liver samples. The PCR reaction conditions and the annealing temperature were as previously described 23 55 56 57 . All samples and standards were amplified in triplicate. The expression level of beta-actin 20 was used as a reference to adjust for an equal amount of sample RNA.

Goat polyclonal IgG antibodies against phospho-signal transducer and activator of Transcription 3 (p-STAT3) (Tyr705), total STAT3 (STAT3), phospho-signal transducer and activator of Transcription 5 (p-STAT5) (Tyr694/Tyr699), total STAT5 (STAT5), phospho-Janus kinase 2 (p-JAK2) (Tyr1007/1008) and total JAK2 (JAK2) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Rabbit polyclonal IgG antibodies against phospho-adenosine monophosphate-activated protein kinase-alpha (p-AMPK-alpha) (Thr172), total AMPK-alpha (AMPK-alpha), phospho-extracellular signal-regulated kinase 1/2 (p-ERK1/2) (Thr202/Tyr204), total ERK1/2 (ERK1/2), phospho-c-Jun NH2-terminal kinase (p-JNK) (Thr183/Tyr185), total JNK (JNK), phospho-p38 MAPK (p-P38 MAPK) (Thr180/Tyr182) and total P38 MAPK (P38 MAPK) were obtained from Cell Signaling Technology (Beverly, CA, USA). Rabbit polyclonal IgG antibodies against phospho-peroxisome proliferator-activated receptor -alpha (p-PPAR-alpha) (p-Ser21) and total PPAR-alpha (PPAR-alpha) were purchased from GenScript USA Incorporated (Piscataway, NJ, USA).

Soluble protein was extracted from rat livers using a protein extraction reagent (Pierce, USA) and protein concentration was measured using a bicinchoninic acid assay kit (Pierce, USA). The extracted proteins were resolved by 6-8% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by electrophoretic transfer to nitrocellulose membranes. The membranes were then incubated in 1.0% non-fat dried milk in 50 mM Tris (pH 8.0) followed by incubation overnight with the primary antibodies at 300-500-fold dilution. The bound primary antibody was detected using biotinylated rabbit anti-goat or rabbit anti-rabbit antibody (Zhongshan Goldenbridge Biotechnology, China) and visualized using 3', 3'-diaminobenzidine tetrahydrochloride. Detection was carried out using an electrochemiluminescence kit (Amersham Pharmacia Biotech, UK). Image-Pro Plus software version 6.0 (Media Cybernetics Incorporated, Silver Spring, MD, USA) was used to determine the mean optical density.

Data are mean ± standard deviation. Statistical analyses were done using SPSS software version 13.0 (SPSS Inc., Chicago, IL, USA). A bifactorial ANOVA (followed by post hoc protected least square difference method) were used for statistical comparison. The two fixed factors were treatment (two levels, control treatment/GH treatment) and diet (two levels, control diet/high-fat diet). Weighted kappa values were calculated to determine inter-observer agreement in pathological evaluation. Values of p < 0.05 were considered significant.