Reduced Hepatic Aquaporin-9 and Glycerol Permeability Are Related to Insulin Resistance in Non-Alcoholic Fatty Liver Disease

ORIGINAL ARTICLE Reduced hepatic aquaporin-9 and glycerol permeability are related to insulin resistance in non-alcoholic fatty liver disease

A Rodrı´guez1,2, P Gena3,LMe´ ndez-Gime´ nez1, A Rosito3, V Valentı´2,4, F Rotellar2,4, I Sola5, R Moncada6, C Silva7, M Svelto3,8, J Salvador7, G Calamita3,8,9 and G Fru¨ hbeck1,2,7,9

BACKGROUND/OBJECTIVES: Glycerol represents an important metabolite for the control of lipid accumulation and hepatic gluconeogenesis. We investigated whether hepatic expression and functionality of aquaporin-9 (AQP9), a channel mediating glycerol influx into hepatocytes, is impaired in non-alcoholic fatty liver disease (NAFLD) and steatohepatitis (NASH) in the context of insulin resistance. SUBJECTS/METHODS: Liver biopsies were obtained from 66 morbid obese patients undergoing bariatric surgery (66% women, mean body mass index (BMI) 46.1±1.0 kg m À 2) with available liver echography and pathology analysis of the biopsies in this cross-sectional study. Subjects were classified according to normoglycemia (NG), impaired glucose tolerance (IGT) or type 2 diabetes (T2D). Hepatic expression of AQP9 was analyzed by real-time PCR, western blotting and immunohistochemistry, while glycerol permeability (Pgly) was measured by stopped-flow light scattering. RESULTS: AQP9 was the most abundantly (Po0.0001) expressed aquaglyceroporin in human liver (AQP9444AQP34AQP74AQP10). Obese patients with T2D showed increased plasma glycerol as well as lower Pgly and hepatic AQP9 expression. The prevalence of NAFLD and NASH in T2D patients was 100 and 65%, respectively. Interestingly, AQP9 expression was decreased in patients with NAFLD and NASH as compared with those without hepatosteatosis, in direct relation to the degree of steatosis and lobular inflammation, being further reduced in insulin-resistant individuals. The association of AQP9 with insulin sensitivity was independent of BMI and age. Consistent with these data, fasting insulin and C-reactive protein contributed independently to 33.1% of the hepatic AQP9 mRNA expression variance after controlling for the effects of age and BMI. CONCLUSIONS: AQP9 downregulation together with the subsequent reduction in hepatic glycerol permeability in insulin-resistant states emerges as a compensatory mechanism whereby the liver counteracts further triacylglycerol accumulation within its parenchyma as well as reduces hepatic gluconeogenesis in patients with NAFLD.

International Journal of Obesity (2014) 38, 1213–1220; doi:10.1038/ijo.2013.234 Keywords: aquaporins; type 2 diabetes; non-alcoholic fatty liver disease

INTRODUCTION molecular mechanisms linking NAFLD and insulin resistance are The diagnosis of non-alcoholic fatty liver disease (NAFLD) is still poorly understood. performed via the evidence of hepatic steatosis by imaging Glycerol constitutes a direct source of glycerol-3-phosphate for techniques corroborated by histology, and should not be TG synthesis and an important substrate for hepatic gluconeo- 6–8 associated with secondary triacyglycerols (TG) accumulation, such genesis during fasting. The liver has a central role in glycerol as significant alcohol consumption, use of steatogenic medication metabolism, since it is responsible for 70–90% of the whole-body or genetic disorders.1 NAFLD ranges from simple steatosis through glycerol metabolism. Aquaglyceroporins (AQP3, AQP7, AQP9 and fibrosis, non-alcoholic steatohepatitis (NASH), to cryptogenic AQP10) encompass a subfamily of aquaporins (AQPs) that allow cirrhosis.2,3 Ectopic fat accumulation in the liver is strongly the movement of water and glycerol across biological mem- associated with insulin-resistant states, such as obesity, meta- branes.8,9 AQP9 represents the primary route for glycerol uptake bolic syndrome or type 2 diabetes (T2D). The prevalence of NAFLD in hepatocytes, the substrate used for de novo glucose synthesis and NASH in the general population is B20 and 3%, respectively,1 in the murine liver.10–12 In this regard, deletion of Aqp9 in obese, while in morbid obese patients it has been estimated to diabetic db/db mice reportedly reduces plasma glucose concen- increase to 75% and 25–70%, respectively.4,5 Compared with trations between 10 and 40%.13 In rodents, insulin represses normoglycemic subjects, individuals with T2D show an increased hepatic Aqp9 gene expression through the negative insulin risk of developing NAFLD and certainly have a higher risk of response element in its gene promoter.14 Previous studies have developing fibrosis and cirrhosis.1,3 Nevertheless, the underlying shown an increase in hepatic AQP9 in insulin-resistant animal

1Metabolic Research Laboratory, Clıńica Universidad de Navarra, Pamplona, Spain; 2CIBER Fisiopatologıá de la Obesidad y Nutricio´ n (CIBERobn), Instituto de Salud Carlos III, Pamplona, Spain; 3Department of Biosciences, Biotechnologies and Biopharmaceutics, University of Bari Aldo Moro, Bari, Italy; 4Department of Surgery, Clıńica Universidad de Navarra, Pamplona, Spain; 5Department of Pathology, Clıńica Universidad de Navarra, Pamplona, Spain; 6Department of Anesthesia, Clıńica Universidad de Navarra, Pamplona, Spain; 7Department of Endocrinology and Nutrition, Clıńica Universidad de Navarra, Pamplona, Spain and 8Centro di Eccellenza di Genomica in campo Biomedico ed Agrario (CEGBA), Bari, Italy. Correspondence: Dr G Calamita, Department of Biosciences, Biotechnologies and Biopharmaceutics, University of Bari Aldo Moro, Via Amendola, 165/A, Bari 70126, Italy or Dr G Fru¨ hbeck, Department of Endocrinology and Nutrition, Clıńica Universidad de Navarra, Avda. Pıó XII, 36, Pamplona 31008, Spain. E-mail: [email protected] or [email protected] 9These authors contributed equally to this work. Received 1 October 2013; revised 20 November 2013; accepted 4 December 2013; accepted article preview online 13 December 2013; advance online publication, 14 January 2014 Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1214 models, such as streptozotocin-induced diabetic rats or genetically RNA extraction and real-time PCR 14,15 obese db/db mice. Nevertheless, the regulation and RNA isolation and purification from human liver samples were performed expression of aquaglyceroporins in the human liver appears to as described earlier.16,17 For first-strand cDNA synthesis equal amounts be different from the control that takes place in mice. In this (2 mg) of total RNA were reverse transcribed using 400 units of M-MLV regard, insulin upregulates the expression of aquaglyceroporins in reverse transcriptase (Invitrogen, Paisley, UK) and random primers (Roche human HepG2 hepatocytes.16 Obese patients with T2D show a Molecular Biochemicals, Mannheim, Germany). The transcript levels for downregulation of hepatic AQP9, compatible with a compen- aquaglyceroporins AQP3, AQP7, AQP9 and AQP10 as well as the gluconeogenic factors GK, PCK1, G6PC and GLUT2 were quantified by satory mechanism aimed at further reducing glycerol uptake real-time PCR (7300 Real-Time PCR System; Applied Biosystems, Foster City, into hepatocytes as well as enhancing the development 16–18 CA, USA). Primers and probes (Supplementary Table 1) were designed of hyperglycemia. Until now, information regarding aqua- using the software Primer Express 2.0 (Applied Biosystems). glyceroporins in human liver is scarce and limited to mRNA and protein determinations without functional studies. Therefore, the aim of the present study was to examine the hepatocyte Western blot studies basolateral membrane glycerol permeability at the same time Samples (30 mg) were run out in 10% SDS–PAGE, subsequently transferred as analyzing the expression and potential participation of onto nitrocellulose membranes and blocked in Tris-buffered saline with aquaglyceroporins AQP3, AQP7, AQP9 and AQP10 in the liver 0.05% Tween-20 containing 5% non-fat dry milk for 1 h at room 1 of obese patients with NAFLD and NASH with varying degree of temperature. Blots were incubated overnight at 4 C with goat polyclonal anti-AQP3, rabbit polyclonal anti-AQP7, goat polyclonal anti-AQP9 (Santa insulin resistance. Cruz Biotechnology, Inc., Santa Cruz, CA, USA), and rabbit polyclonal anti- AQP10 (Sigma) (diluted 1:1000 in blocking solution) or mouse monoclonal anti-b-actin (Sigma) antibody (diluted 1:5000 in blocking solution). The antigen–antibody complexes were visualized using horseradish peroxi- MATERIALS AND METHODS dase-conjugated secondary antibodies for 1 h at room temperature diluted Patients in blocking solution (1:5000) and the enhanced chemiluminescence ECL Sixty-six Caucasian subjects undergoing Roux-en-Y gastric bypass at the Plus detection system (Amersham Biosciences, Buckinghamshire, UK). Clıńica Universidad de Navarra participated in this study. Body mass index (BMI) was calculated as weight in kilograms divided by the square of Immunohistochemistry of AQP9 height in meters, and the percentage of body fat was estimated by air Immunohistochemical detection of AQP9 was carried out as described displacement plethysmography (Bod-Pod; Life Measurements, Concord, 16 CA, USA). Obesity was classified according to both a BMI of X30 kg m À 2 before. Sections were incubated overnight at 4 1C with rabbit anti-AQP9 and a body fat of X25% in males and body fat X35% in females. Obese polyclonal IgG (AQP91-A; Alpha Diagnostic International, San Antonio, TX, patients were classified into three groups (normoglycemia (NG), impaired USA) antibody diluted 1:100 in Tris-buffered saline. After washing three glucose tolerance (IGT) or T2D) following the criteria of the Expert times with Tris-buffered saline (5 min each), slides were incubated with Committee on the Diagnosis and Classification of Diabetes.19 Subjects with Dako Real EnVision horseradish peroxidase-conjugated anti-rabbit/mouse T2D were not on insulin therapy or medication likely to influence (K5007; Dako, Glostrup, Denmark) for 1 h at room temperature. endogenous insulin levels, but eight T2D patients were on metformin After washing in Tris-buffered saline, peroxidase reaction was visualized 0 with a 3,3 -diaminobenzidine (Amersham Biosciences)/H2O2 solution treatment. Patients included in our obese group of T2D did not have a long À 1 À 1 diabetes history (o2–3 years or even de novo diagnosis as evidenced from (0.5 mg ml diaminobenzidine, 0.03% H2O2 diluted in 50 mmol l their anamnesis and biochemical determinations). Tris–HCl, pH 7.36) as a chromogen and Harris hematoxylin solution An intraoperative liver biopsy was performed in the obese patients (Sigma) as a counterstain. during bariatric surgery to establish a histological diagnosis of the hepatic state, but this procedure is not clinically justified in lean subjects. All Stopped-flow light-scattering measurements of glycerol reported investigations were carried out in accordance with the principles permeability of the Declaration of Helsinki, as revised in 2008. The experimental design was approved, from an ethical and scientific standpoint, by the Hospital’s Liver biopsies were homogenized in ice-cold isolation medium consisting of 220 mmol l À 1 mannitol, 70 mmol l À 1 sucrose, 20 mmol l À 1 Tris–HCl, Ethical Committee responsible for research (028/2009), and the written À 1 À 1 informed consent from all volunteers was obtained. 1 mmol l EDTA, and 5 mmol l EGTA, pH 7.40, and protease inhibitors (1 mmol l À 1 phenylmethylsulfonyl fluoride, 1 mmol l À 1 leupeptin, 1 mmol l À 1 pepstatin A), and vesicles of basolateral plasma membranes were prepared by differential centrifugation as previously described.11 The Evaluation of liver histology size of vesicles was determined with an N5 Submicron Particle Size Liver biopsies were analyzed by an expert pathologist. The diagnosis of Analyzer (Beckman Coulter, Palo Alto, CA, USA). The time course of vesicle NAFLD and NASH was determined according to the criteria of Brunt.20 volume changes was monitored by the light scattering at 530 nm NASH was characterized by steatosis, mixed inflammation, balloon wavelength using a BioLogic MPS-200 stopped-flow reaction analyzer 11 degeneration, variable degrees of fibrosis and Mallory body formation.21 (BioLogic, Claix, France) at 20 1C as previously described. Glycerol À 1 Briefly, steatosis was scored as an estimate of the percentage of permeability (Pgly;cms ) was computed using the equation: parenchyma replaced by fat: (1) up to 5%, (2) 6–33%, (3) 34–66% or Pgly ¼ 1/((S/V)t), where S/V is a surface-to-volume ratio and t (1/Ki) is the (4) 466%. Lobular and portal inflammation was graded on a scale of three exponential time constant fitted to the vesicle-swelling phase of light- categories: (0) none, (1) mild or (2) moderate-marked (Brunt grades 2 and 3). scattering time course corresponding to glycerol entry. In some Hepatocellular ballooning, pericellular/perisinusoidal fibrosis, portal fibrosis experiments, vesicles were pre-incubated with the AQP9 inhibitor À 1 and cirrhosis were scored as (0) absent or (1) present. phloretin (0.7 mmol l ) for 10 min before being submitted to the light- scattering analysis.

Analytical procedures Statistical analysis Biochemical and hormonal assays were performed as previously Data are expressed as mean±s.e.m. Statistical differences between mean described.16,17 Non-esterified fatty acids (NEFAs) were determined by values were determined using one-way ANOVA, followed by Scheffe´’s tests enzymatic spectrophotometric methods (WAKO Chemicals, GmbH, Neuss, or Kruskal–Wallis test followed by U Mann–Whitney’s pairwise comparisons, Germany). Plasma glycerol levels were analyzed using a quantitative where appropriate, for quantitative variables, and assessment of chi-square enzymatic determination assay (Sigma, St Louis, MO, USA); intra- and inter- (w2) distributions for categorical variables. Pearson’s correlation coefficients assay coefficients of variation were 3.3 and 4.2%, respectively. The (r) and stepwise multiple linear regression analysis were used to analyze the adipocyte insulin resistance (Adipo-IR) index, as a surrogate of adipocyte association between variables. The calculations were performed using the dysfunction, was calculated as fasting NEFA (mmol l À 1) Â fasting insulin SPSS/Windows version 15.0 statistical package (SPSS, Chicago, IL, USA). (pmol l À 1).22 A value of Po0.05 was considered as statistically significant.

International Journal of Obesity (2014) 1213 – 1220 & 2014 Macmillan Publishers Limited Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1215 RESULTS expressed aquaglyceroporin in human liver, AQP3 gene expression Study population was assumed to be 1 and the mRNA expression of AQP7, AQP9 and The clinical and biochemical characteristics of the patients are AQP10 was related to this aquaglyceroporin. Transcript levels of all summarized in Table 1. As expected, obese subjects with T2D the studied aquaglyceroporins were found in human liver, with showed higher (Po0.0001) basal glucose levels, increased AQP9 being the most abundantly (Po0.0001) expressed glycerol ± ± glycemia (Po0.0001) and insulinemia (Po0.05) 2 h after the oral channel (30.146 4.140 AU) followed by AQP3 (1.000 0.000 AU), glucose tolerance test and higher HOMA (homeostasis model AQP7 (0.142±0.012 AU) and AQP10 (0.001±0.000 AU). Women ± assessment) index (Po0.05) compared with obese normoglycemic showed increased hepatic transcript levels of AQP3 (2.16 0.48 vs individuals. Given the important role that adipose tissue insulin 1.13±0.12 AU, P ¼ 0.048), but no sexual dimorphism (P40.05) resistance plays in the development of NAFLD, the circulating in the hepatic gene expression of AQP7, AQP9 and AQP10 concentrations of NEFA and glycerol were analyzed and the was observed. Adipo-IR was calculated. Plasma NEFA and glycerol concentrations as well as the Adipo-IR index were increased (Po0.05) in obese Insulin-resistant obese patients show a decrease in hepatic AQP9 patients with T2D. The liver of obese patients with T2D also expression and glycerol permeability showed increased (Po0.05) expression of the gluconeogenic Insulin resistance was associated with a decrease (Po0.05) in genes GK, PCK1 and G6PC, while a decrease (Po0.05) in the hepatic mRNA and protein levels of AQP9 (Figures 1a and b), expression levels of the glucose transporter GLUT2 was evident. without signiﬁcant changes in the expression of the other As regards the hepatic function, T2D was associated with an aquaglyceroporins (P40.05) (Supplementary Figure 1). In this increase (Po0.05) in serum transaminase alanine aminotransferase regard, univariate analyses showed that AQP9 transcript levels (ALT) and alkaline phosphatase. were negatively correlated with glycemia and the Adipo-IR index, while they were positively associated with the quantitative insulin sensitivity check index (QUICKI) and hepatic PCK1, G6PC and Human liver expresses all the aquaglyceroporins GLUT2 transcript levels (Figures 1c and d; Supplementary Table 2). We next analyzed by real-time PCR the gene expression of glycerol In our cohort, none of the patients were on thiazolidinediones channels AQP3, AQP7, AQP9 and AQP10 in liver biopsies of the medication, but eight patients with T2D were on metformin patients. To gain further insight into the most abundantly treatment. To gain further insight into the potential effect of this

Table 1. Clinical characteristics of the subjects included in the study

Obese NG Obese IGT Obese T2D P

n 23 22 21 — Gender (male/female) 7/16 4/18 11/10 0.138 Age (years) 38±244±247±2 0.054 Height (m) 1.65±0.02 1.63±0.02 1.68±0.02 0.193 Weight (kg) 127±7 121±4 132±5 0.393 BMI (kg m À 2) 46.4±2.3 44.7±0.9 47.0±1.7 0.641 Body fat (%) 51.0±1.4 53.7±1.2 51.1±1.7 0.347 Waist circumference (cm) 124±4 124±3 135±4 0.071 Glucose (mmol l À 1) 5.2±0.1 5.9±0.1 8.3±0.8a,b o0.00001 Glucose 2-h OGTT (mmol l À 1) 6.7±0.4 8.2±0.7a 16.1±1.4a,b o0.00001 Insulin (pmol l À 1) 109±16 121±19 146±18 0.445 Insulin 2-h OGTT (pmol l À 1) 517±68 824±104a 622±175 0.035 HOMA 4.5±0.7 5.9±1.0 9.4±2.9a 0.047 QUICKI 0.32±0.01 0.31±0.01 0.29±0.01 0.085 NEFA (mmol l À 1) 0.50±0.03 0.56±0.03 0.66±0.06a 0.031 Glycerol (mmol l À 1) 0.74±0.10 0.84±0.10 1.32±0.23a 0.041 Adipo-IR index 64.3±10.8 75.6±13.9 125.0±26.5a 0.039 Triacylglycerols (mmol l À 1) 1.3±0.1 1.6±0.1 2.0±0.3 0.186 Total cholesterol (mmol l À 1) 5.1±0.2 5.5±0.2 4.9±0.1b 0.034 LDL cholesterol (mmol l À 1) 3.2±0.2 3.5±0.2 3.1±0.2 0.170 HDL cholesterol (mmol l À 1) 1.3±0.1 1.3±0.1 1.2±0.1 0.203 CRP (mg l À 1) 7.7±1.4 10.7±1.8 9.4±2.4 0.530 Uric acid (mmol l À 1) 345±12 339±18 345±18 0.972 Leptin (ng ml À 1) 42.5±4.4 56.3±5.6 44.0±5.5 0.126 AST (IU l À 1)15±116±117±2 0.515 ALT (IU l À 1)17±223±3a 26±4a 0.043 Alkaline phosphatase (IU l À 1)58±875±965±6a 0.037 g-GT (IU l À 1)20±324±431±7 0.349 Hepatic GK gene expression (AU) 1.10±0.22 1.41±0.26 2.29±0.42a 0.027 Hepatic PCK1 gene expression (AU) 0.15±0.05 0.17±0.05 0.47±0.13a 0.032 Hepatic G6PC gene expression (AU) 0.11±0.03 0.15±0.03 0.41±0.10a,b 0.006 Hepatic GLUT2 gene expression (AU) 1.21±0.04 1.00±0.10 0.74±0.02a 0.046 Abbreviations: Adipo-IR, adipocyte insulin resistance; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CRP, high-sensitivity C-reactive protein; G6PC, glucose-6-phosphatase; GK, glycerol kinase; GLUT2, glucose transporter 2; HDL, high-density lipoprotein; HOMA, homeostasis model assessment; IGT, impaired glucose tolerance; LDL, low-density lipoprotein; NEFA, non-esteriﬁed fatty acid; NG, normoglycemia; OGTT, oral glucose tolerance test; PCK1, phosphoenolpyruvate carboxykinase variant 1; QUICKI, quantitative insulin sensitivity check index; T2D,type2 diabetes; g-GT, g-glutamyltransferase. Bold values denote statistically signiﬁcant P-values. Differences between groups were analyzed by one-way ANOVA followed by Scheffe´ ’s test or w2 test, where appropriate. aPo0.05 vs obese NG subjects. bPo0.05 vs obese IGT patients.

& 2014 Macmillan Publishers Limited International Journal of Obesity (2014) 1213 – 1220 Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1216

Figure 1. Impact of insulin resistance on AQP9 expression in human liver. Bar graphs show AQP9 mRNA (a)(n ¼ 20–22 per group) and protein (b)(n ¼ 6–10 per group) expression in liver samples obtained from obese patients with NG, IGT and T2D. The gene and protein expression in obese NG subjects was assumed to be 1. Representative blots are shown at the top of the figure. Correlations between hepatic AQP9 expression and QUICKI (c) and Adipo-IR (d) indices are shown. The Pearson’s coefficient of variation (r) and P-value are indicated. (e) Immunohistochemical detection of AQP9 protein (right panels) in hepatic sections obtained from obese patients (magnification, Â 400; scale bar ¼ 100 mm). No immunoreactivity was found without primary antibody (left panels). (f) Glycerol permeability (Pgly) of hepatocyte basolateral membrane was evaluated by stopped-flow light scattering (n ¼ 5–8 per group) in the absence/presence of AQP9 inhibitor phloretin (0.7 mmol l À 1). Values are the mean±s.e.m. Differences between groups were analyzed by one-way ANOVA followed by Scheffe´’s test or Kruskal–Wallis test followed by U Mann–Whitney’s pairwise comparisons, where appropriate. *P 0.05, **P 0.01 vs obese NG subjects; www o o Po0.001 vs the corresponding group of glycemia in the absence of phloretin.

antidiabetic drug on hepatic AQP9 expression, the transcript and Since multiple cell types (that is, cholangiocytes, sinusoidal protein levels of this aquaglyceroporin were compared between endothelial cells and Kupffer cells) may represent a source of metformin-treated vs non-treated patients. Metformin did not AQP9 expression, the presence of this aquaglyceroporin in change the expression of AQP9 mRNA (treated 0.41±0.13 AU vs hepatocytes was conﬁrmed by immunohistochemistry (Figure 1e). non-treated 0.65±0.19 AU, P ¼ 0.283) and protein (treated Liver sections obtained from obese patients showed a strong 0.40±0.19 AU vs non-treated 0.44±0.08 AU, P ¼ 0.909), which is immunoreactivity for AQP9 in the basolateral membrane of in accordance with the data obtained by Asai et al.,23 showing that hepatocytes. AQP9 staining was strongest around the central thiazolidinediones troglitazone and tolbutamide reduced AQP3 vein (perivenous zone), whereas the sinusoids around the portal expression, but the biguanides metformin and buformin did not vein (periportal zone) were weakly stained. In line with results induce such suppression. obtained from gene and protein analyses, the immunoreactivity

International Journal of Obesity (2014) 1213 – 1220 & 2014 Macmillan Publishers Limited Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1217 for AQP9 was weaker in hepatic sections of patients with IGT or normal liver histology. The pathology analysis revealed that 35.0% T2D compared with normoglycemic subjects. of obese patients with T2D showed steatosis, while 65.0% À 1 To analyze glycerol membrane permeability (Pgly;cms ), exhibited different degrees of NASH. Compared with the group stopped-flow light scattering was carried out using the hepatocyte with T2D, the prevalence of NASH was significantly lower basolateral membrane vesicles prepared from hepatic biopsies (P ¼ 0.024) in NG (43.5%) and IGT (52.4%) patients. Liver biopsies obtained from the obese patients. The Pgly value of the vesicles of obese subjects with T2D showed higher (Po0.05) steatosis and from obese men was significantly higher than that of obese portal inflammation than obese normoglycemic individuals. No women (5.52±0.44 vs 3.22±0.14 Â 106 cm s À 1, Po0.0001). This patient had evidence of established cirrhosis. increase in Pgly may be related to higher hepatic AQP9 protein Since one of the best known hepatic derangements associated levels in men, although differences did not reach statistical with obesity-associated insulin resistance is NAFLD, we aimed to significance (men 0.60±0.08 AU vs women 0.46±0.06 AU, investigate AQP9 expression in this condition. Compared with P ¼ 0.178). The basolateral membrane Pgly values of liver samples patients without steatosis, a significant (Po0.05) decrease in AQP9 were dramatically decreased (Po0.0001) upon treatment with mRNA and protein expression was observed in liver samples 0.7 mmol l À 1 phloretin (Figure 1f), a well-known inhibitor of obtained from patients with NAFLD and NASH (Figures 2a and b). AQP9-mediated glycerol diffusion,24 suggesting that glycerol Consistent with these findings and the above-mentioned biophy- permeability is mediated by AQP9 in human liver. Moreover, a sical studies, AQP9 was downregulated according to the degree of significantly (Po0.05) reduced hepatic glycerol permeability was steatosis as well as by the intensity of lobular inflammation observed in samples obtained from insulin-resistant (IGT and T2D) (Figures 2c and d). Furthermore, AQP9 transcript levels were patients compared with insulin-sensitive individuals (Figure 1f). negatively correlated with aspartate aminotransferase, ALT and g-glutamyltransferase (Supplementary Table 2). Interestingly, lower AQP9 mRNA levels were found in insulin-resistant patients NAFLD and NASH are associated with a decreased hepatic AQP9 with NAFLD (0.82±0.12 vs 0.51±0.06 AU, Po0.05) or NASH expression (0.71±0.11 vs 0.47±0.08 AU, Po0.05) as compared with insulin- All obese patients with T2D exhibited NAFLD as evidenced by sensitive subjects with similar hepatic alterations. echography, whereas the prevalence in normoglycemic and IGT groups was 69.6 and 90.9%, respectively (P ¼ 0.010). The histological findings to assess the severity of NAFLD and NASH Relationship between hepatic AQP9 expression and markers of in liver biopsies of patients undergoing bariatric surgery are insulin sensitivity and inflammation illustrated in Table 2. All patients that evidenced NAFLD by In the multiple lineal regression analysis (Table 3), insulin echography showed steatosis in the histological analysis. None of sensitivity, as evidenced by the QUICKI, contributed indepen- the studied subjects in the obese group with T2D exhibited a dently to 16.8% (Po0.05) of hepatic AQP9 expression after controlling for the effects of age and BMI. C-reactive protein (Po0.05) accounted for 22.5% (Po0.05) of AQP9 transcript levels Table 2. Pathology analysis of NAFLD and NASH in hepatic biopsies in the liver after controlling for age, BMI and white blood cell from subjects included in the study count. Otherwise, fasting insulin (Po0.05) and C-reactive protein (Po0.05) contributed independently to 33.1% (Po0.01) of hepatic Pathological characteristics Obese Obese Obese P AQP9 mRNA expression variance after controlling for the effects NG IGT T2D of age and BMI. (%) (%) (%)

Steatosis 0.039 0(o5%) 23 22 8 DISCUSSION 1 (5–33%) 10 10 14 The adipose tissue is the most important source of plasma 2 (34–66%) 0 0 5 glycerol. Under lipolytic conditions, TG are hydrolyzed in NEFA 4 3( 66%) 3 2 3 and glycerol, which are released into the bloodstream.25,26 AQP3 Lobular inflammation 0.155 and AQP7 constitute the major glycerol gateways in human 0 (None) 28 23 20 9,16 1 (Mild) 8 9 8 adipocytes. In the present study, T2D was related to higher 2 (Moderate-Marked) 0 0 4 circulating NEFA and glycerol as well as Adipo-IR index, suggesting Portal inflammation 0.042 an increased lipolytic activity. In this regard, T2D is associated with 0 (None) 13 9 5 augmented AQP3 and AQP7 expression in human visceral fat,16,17 1 (Mild) 23 22 23 providing further evidence that the glycerol efflux is increased in 2 (Moderate-Marked) 0 2 3 the insulin-resistant state. Hepatocyte ballooning 0.861 The liver has a pivotal role in glycerol uptake, since it is 0 (Absence) 27 23 24 responsible for 70–90% of whole-body glycerol metabolism.6 1 (Presence) 8 10 8 In rodents, AQP9 constitutes the main gateway for glycerol uptake Mallory bodies 1.000 10–12,25 0 (Absence) 36 32 32 in hepatocytes. Our results confirm the previously reported 1 (Presence) 0 0 0 presence of AQP3, AQP7 and AQP9 in human liver by our group Pericellular/perisinusoidal fibrosis 0.521 and others,16,17,27,28 and provide evidence, for the first time, of low 0 (Absence) 34 28 28 hepatic AQP10 expression. AQP9 was the most abundantly 1 (Presence) 2 3 5 expressed aquaglyceroporin with this glycerol channel being Portal fibrosis 0.959 localized at the sinusoidal plasma membrane that faces the portal 0 (Absence) 30 26 28 vein, which is consistent with studies performed by other authors 1 (Presence) 6 5 5 in mouse and rat liver.29,30 Moreover, glycerol permeability was Abbreviations: IGT, impaired glucose tolerance; NAFLD, non-alcoholic fatty strongly inhibited by the AQP9 blocker, phloretin, suggesting liver disease; NASH, non-alcoholic steatohepatitis; NG, normoglycemia; that AQP9 also constitutes the major contributor to glycerol T2D, type 2 diabetes. Bold values denote statistically significant P-values. transport over the basolateral plasma membrane of human Data are shown as the percentage of samples with the specific clinical 2 hepatocytes. Nevertheless, since phloretin also acts as an feature (%). Differences between groups were analyzed by w test. inhibitor of other AQPs and does not fully block AQP9-mediated

& 2014 Macmillan Publishers Limited International Journal of Obesity (2014) 1213 – 1220 Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1218

Figure 2. Impact of obesity-associated NAFLD and NASH on the expression of AQP9 in human liver. Bar graphs show AQP9 mRNA (n ¼ 20–22) (a) and protein (n ¼ 6–10) (b) expression in liver samples obtained from obese subjects classified according to the histological diagnosis as normal liver, NAFLD or NASH. The gene and protein expression in obese patients with normal liver was assumed to be 1. Representative blots are shown at the top of the figure. Comparison of the AQP9 transcript levels according to the degree of hepatic steatosis (c) and lobular inflammation (d). Values are the mean±s.e.m. Differences between groups were analyzed by one-way ANOVA followed by Scheffe´ ’s test or Kruskal–Wallis test followed by U Mann–Whitney’s pairwise comparisons, where appropriate. *Po0.05, **Po0.01 vs obese subjects with normal liver.

glycerol permeability,24 the existence of other operative glycerol decreased hepatic glycerol permeability leads to lower channels in human hepatocytes cannot be discarded. availability of glycerol as a gluconeogenic substrate and, AQP9 has an important role in hepatic glycerol and glucose hence, prevents a further aggravation of hyperglycemia in metabolism, since it mediates glycerol influx to hepatocytes and patients with T2D (Figure 3). glycerol represents a gluconeogenic substrate during fast- AQP9 has been proposed as a pharmacological therapy for ing.11,13,25,31 Our study shows that obese patients with T2D NAFLD and NASH.37,38 NAFLD is characterized by an abnormal exhibit lower hepatic expression of AQP9, a finding consistent lipid metabolism and the overaccumulation of TG stored in lipid with previous observations by our group and others.16–18 droplets.1 Under lipogenic conditions, such as obesity or insulin We deemed necessary to ascertain whether changes in hepatic resistance, hepatocytes synthesize TG from the esterification of AQP9 abundance were sufficient to cause a biologically relevant free fatty acids (FFA) with glycerol-3-phosphate.25 The potential difference in glycerol permeation. A gender-specific regulation of sources of FFA contributing to hepatic steatosis include: (1) dietary hepatic handling of glycerol during starvation has been fatty acids (mainly through the uptake of intestine-derived demonstrated in rodents, with female rats showing lower chylomicron remnants); (2) increased lipolysis of fat that enters expression of AQP9 and glycerol permeability than male the liver as plasma FFA; and (3) FFA from de novo lipogenesis in counterparts.30,32 Thereisevidenceindicatingthatsteroid the liver.39 The other metabolite required for TG biosynthesis, hormones regulate the expression of AQP9 rat liver, epidydimis glycerol-3-phosphate, derives from three metabolic sources: and epithelium of Fallopian tubes, where cyclical changes in (1) glucose, since glycerol-3-phosphate constitutes a secondary estrogen and/or progesterone can induce marked fluctuations in metabolite of glycolysis; (2) lipolysis-derived glycerol, which is the expression of AQP9 in these target tissues.33–36 The sexual phosphorylated by glycerol kinase; and (3) AQP9-mediated dimorphism in AQP9 regulation in rat liver has been ascribed to glycerol uptake.9,11 In the present study, the expression of AQP9 the ability of 17b-estradiol to suppress hepatic AQP9 was decreased in obese patients with NAFLD in parallel to the expression.32,34 In the present study, all females were in a pre- degree of hepatic steatosis. These findings lead to the notion that menopausal state and pregnant females or during the lactation lower intrahepatocellular glycerol due to a decreased AQP9 period were excluded. Obese women showed lower glycerol expression may represent a compensatory mechanism to reduce permeability in the hepatocyte basolateral membranes, but we the de novo TG synthesis in the liver of patients with NAFLD did not find differences attributable to gender in AQP9 (Figure 3). expression levels, which is in agreement with a previous study A dysregulation of AQP9 has been observed in several hepatic performed in morbid obese patients.18 Consequently, the derangements in rats, such as extrahepatic cholestasis40 or downregulation of hepatic AQP9 levels together with the alcoholic steatohepatitis.41 Defective AQP9 expression and

International Journal of Obesity (2014) 1213 – 1220 & 2014 Macmillan Publishers Limited Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1219 negatively correlated with circulating aspartate aminotransferase, Table 3. Multiple linear regression analyses with hepatic AQP9 gene ALT or g-glutamyltransferase in obese patients. Consistent with expression as dependent variable for all subjects in the cross-sectional these findings, AQP9 protein expression was reduced in patients study with NASH according to the degree of lobular inflammation. AQP9 AQP9 mRNA expression b P downregulation does not seem to be related to any major modification of the AQP9 transcript, which is in agreement with Model I previous observations of hepatic AQP8 expression in rat Age À 0.008 0.165 cholestasis.44 Thus, the downregulation of AQP9 in patients with BMI 0.020 0.070 NASH may reflect the altered hepatic function. Novel drugs have Fasting insulin À 0.011 o0.05 recently been identified that specifically inhibit AQP9,10 thereby CRP 0.204 o0.05 opening an interesting research field targeting this glycerol WBC À 0.066 0.083 channel for therapeutical purposes. Adjusted R2 0.331 o0.01 Model II In summary, our findings show that low AQP9 levels are found Age À 0.007 0.215 in obese patients with NAFLD and NASH, with this downregulation BMI 0.016 0.185 being further aggravated in the context of insulin resistance. CRP 0.250 0.013 According to multiple regression analyses, AQP9 expression was WBC À 0.076 0.061 associated with insulin resistance even after controlling for BMI Adjusted R2 0.225 o0.05 and inflammatory markers, suggesting a role of this aquaglycero- Model III porin in the regulation of hepatic gluconeogenesis and hepatic Age À 0.002 0.650 function. AQP9 downregulation and the reduction in hepatic BMI 0.012 0.104 glycerol permeability contribute to lower glycerol availability as a QUICKI 4.239 0.021 2 gluconeogenic and lipogenic substrate and, hence, may constitute Adjusted R 0.168 o0.05 a defensive mechanism to prevent hyperglycemia and steatosis in Abbreviations: AQP9, aquaporin-9; BMI, body mass index; CRP, C-reactive patients with NAFLD. protein; QUICKI, quantitative insulin sensitive check index; WBC, white blood cells. b is the regression coefficient, which allows evaluating the relative significance of each independent variable in multiple linear regression analyses. Adjusted R2 expresses the percentage of the variance CONFLICT OF INTEREST explained by the independent variables in the different models (that is, The authors declare no conflict of interest. 0.168 is 16.8%). Statistical significant values are in bold.

ACKNOWLEDGEMENTS We gratefully acknowledge the valuable collaboration of all members of the Departments of Surgery, Anesthesia and the Nutrition Unit of the Clıńica Universidad de Navarra for their technical support. We are also thankful to Prof. Piero Portincasa, Gabriella Garruti, Domenico Ferri, Giuseppa E Liquori and Maria Mastrodonato for valuable suggestions and stimulating discussions. This work was funded by the Instituto de Salud Carlos III and fondos FEDER (FIS PI10/01677 to AR and PI12/00515 to GF) and by grants from the Ministero dell’Istruzione, dell’Universita` e della Ricerca (MIUR; PRIN20089SRS2X_003, 2009–2012) and Fondazione Cassa di Risparmio di Puglia (Ricerca Scientifica e Tecnologica, 2010–2012) to GC and Plan de Investigacioń de la Universidad de Navarra (PIUNA) (2011–2013) to AR. CIBER de Fisiopatologıádela Obesidad y Nutricioń (CIBERobn) is an initiative of the Instituto de Salud Carlos III, Spain.

REFERENCES 1 Chalasani N, Younossi Z, Lavine JE, Diehl AM, Brunt EM, Cusi K et al. The diagnosis and management of non-alcoholic fatty liver disease: practice Guideline by the American Association for the Study of Liver Diseases, American College of Gastroenterology, and the American Gastroenterological Association. Hepatology Figure 3. Proposed mechanism for the role of AQP9 in hepatic 2012; 55: 2005–2023. gluconeogenesis and steatosis in patients with NAFLD. G6PC, 2 Utzschneider KM, Kahn SE. Review: the role of insulin resistance in nonalcoholic glucose-6-phosphatase; GK, glycerol kinase; GLUT, glucose transpor- fatty liver disease. J Clin Endocrinol Metab 2006; 91: 4753–4761. ter; PEP, phosphoenolpyruvate; PCK1, phosphoenolpyruvate carbox- 3 Roden M. Mechanisms of disease: hepatic steatosis in type 2 diabetes--patho- ykinase variant 1. genesis and clinical relevance. Nat Clin Pract Endocrinol Metab 2006; 2: 335–348. 4 Machado M, Marques-Vidal P, Cortez-Pinto H. Hepatic histology in obese patients undergoing bariatric surgery. J Hepatol 2006; 45: 600–606. reduced basolateral osmotic water permeability are associated 5 Boza C, Riquelme A, Ibanez L, Duarte I, Norero E, Viviani P et al. Predictors of with secretory dysfunction of rat hepatocytes during obstructive nonalcoholic steatohepatitis (NASH) in obese patients undergoing gastric bypass. extrahepatic cholestasis, contributing to bile ﬂow dysfunction of Obes Surg 2005; 15: 1148–1153. cholestatic hepatocytes.40,42 Hepatocellular ballooning, a histological 6 Reshef L, Olswang Y, Cassuto H, Blum B, Croniger CM, Kalhan SC et al. hallmark of alcoholic steatohepatitis and, to a lesser extent, of Glyceroneogenesis and the triglyceride/fatty acid cycle. J Biol Chem 2003; 278: NASH, is produced due to protein and water retention by 30413–30416. hepatocytes. Acetaldehyde, an oxidation product derived from 7 Fru¨hbeck G. Obesity: aquaporin enters the picture. Nature 2005; 438: 436–437. ´ ´ ´ ethanol by hepatic alcohol dehydrogenase, induces a rapid 8 Fru¨hbeck G, CatalanV,Gomez-Ambrosi J, Rodrıguez A. Aquaporin-7 and glycerol permeability as novel obesity drug-target pathways. Trends Pharmacol Sci 2006; increase in glycerol uptake and cell size in rat hepatocytes 27: 345–347. via AQP9, but prolonged exposure to acetaldehyde reduces 9 Rodrı´guez A, Catala´nV,Go´mez-Ambrosi J, Fru¨hbeck G. Aquaglyceroporins serve 9,43 AQP9 expression, suggesting a negative feedback regulation. as metabolic gateways in adiposity and insulin resistance control. Cell Cycle 2011; In the present study, hepatic AQP9 transcript levels were 10: 1548–1556.

& 2014 Macmillan Publishers Limited International Journal of Obesity (2014) 1213 – 1220 Aquaglyceroporins in human NAFLD A Rodrı´guez et al 1220 10 Jelen S, Wacker S, Aponte-Santamaria C, Skott M, Rojek A, Johanson U et al. 27 Ishibashi K, Sasaki S, Saito F, Ikeuchi T, Marumo F. Structure and chromosomal Aquaporin-9 protein is the primary route of hepatocyte glycerol uptake for localization of a human water channel (AQP3) gene. Genomics 1995; 27: 352–354. glycerol gluconeogenesis in mice. J Biol Chem 2011; 286: 44319–44325. 28 Ishibashi K, Kuwahara M, Gu Y, Tanaka Y, Marumo F, Sasaki S. Cloning and 11 Calamita G, Gena P, Ferri D, Rosito A, Rojek A, Nielsen S et al. Biophysical functional expression of a new aquaporin (AQP9) abundantly expressed in the assessment of aquaporin-9 as principal facilitative pathway in mouse liver import peripheral leukocytes permeable to water and urea, but not to glycerol. Biochem of glucogenetic glycerol. Biol Cell 2012; 104: 342–351. Biophys Res Commun 1998; 244: 268–274. 12 Gena P, Mastrodonato M, Portincasa P, Fanelli E, Mentino D, Rodrı´guez A et al. 29 Elkjaer M, Vajda Z, Nejsum LN, Kwon T, Jensen UB, Amiry-Moghaddam M et al. Liver glycerol permeability and aquaporin-9 are dysregulated in a murine model Immunolocalization of AQP9 in liver, epididymis, testis, spleen, and brain. Biochem of non-alcoholic Fatty liver disease. PLoS One 2013; 8: e78139. Biophys Res Commun 2000; 276: 1118–1128. 13 Rojek AM, Skowronski MT, Fuchtbauer EM, Fuchtbauer AC, Fenton RA, Agre P et al. 30 Nicchia GP, Frigeri A, Nico B, Ribatti D, Svelto M. Tissue distribution and Defective glycerol metabolism in aquaporin 9 (AQP9) knockout mice. Proc Natl membrane localization of aquaporin-9 water channel: evidence for sex-linked Acad Sci USA 2007; 104: 3609–3614. differences in liver. J Histochem Cytochem 2001; 49: 1547–1556. 14 Kuriyama H, Shimomura I, Kishida K, Kondo H, Furuyama N, Nishizawa H et al. 31 Rojek A, Fuchtbauer EM, Fuchtbauer A, Jelen S, Malmendal A, Fenton RA et al. Coordinated regulation of fat-specific and liver-specific glycerol channels, Liver-specific aquaporin 11 knockout mice show rapid vacuolization of the rough aquaporin adipose and aquaporin 9. Diabetes 2002; 51: 2915–2921. endoplasmic reticulum in periportal hepatocytes after feeding amino acids. 15 Carbrey JM, Gorelick-Feldman DA, Kozono D, Praetorius J, Nielsen S, Agre P. Am J Physiol Gastrointest Liver Physiol 2013; 304: G501–G515. Aquaglyceroporin AQP9: solute permeation and metabolic control of expression 32 Lebeck J, Gena P, O’Neill H, Skowronski MT, Lund S, Calamita G et al. in liver. Proc Natl Acad Sci USA 2003; 100: 2945–2950. Estrogen prevents increased hepatic aquaporin-9 expression and glycerol 16 Rodrı´guez A, CatalańV,Go´mez-Ambrosi J, Garcıá-Navarro S, Rotellar F, Valentı´ V et al. uptake during starvation. Am J Physiol Gastrointest Liver Physiol 2011; 302: Insulin- and leptin-mediated control of aquaglyceroporins in human adipocytes and G365–G374. hepatocytes is mediated via the PI3K/Akt/mTOR signaling cascade. JClinEndocrinol 33 Branes MC, Morales B, Rios M, Villalon MJ. Regulation of the immunoexpression of Metab 2011; 96: E586–E597. aquaporin 9 by ovarian hormones in the rat oviductal epithelium. Am J Physiol Cell 17 CatalańV,Go´mez-Ambrosi J, Pastor C, Rotellar F, Silva C, Rodrı´guez A et al. Physiol 2005; 288: C1048–C1057. Influence of morbid obesity and insulin resistance on gene expression levels 34 Wellejus A, Jensen HE, Loft S, Jonassen TE. Expression of aquaporin 9 in rat of AQP7 in visceral adipose tissue and AQP9 in liver. Obes Surg 2008; 18: liver and efferent ducts of the male reproductive system after neonatal diethyl- 695–701. stilbestrol exposure. J Histochem Cytochem 2008; 56: 425–432. 18 Miranda M, Ceperuelo-Mallafre´ V, Lecube A, Hernańdez C, Chacoń MR, Fort JM et al. 35 Pastor-Soler NM, Fisher JS, Sharpe R, Hill E, Van Hoek A, Brown D et al. Aquaporin Gene expression of paired abdominal adipose AQP7 and liver AQP9 in patients 9 expression in the developing rat epididymis is modulated by steroid hormones. with morbid obesity: relationship with glucose abnormalities. Metabolism 2009; 58: Reproduction 2010; 139: 613–621. 1762–1768. 36 Ji YF, Chen LY, Xu KH, Yao JF, Shi YF, Shanguan XJ. Reduced expression of 19 Genuth S, Alberti KG, Bennett P, Buse J, Defronzo R, Kahn R et al. Follow-up report aquaporin 9 in tubal ectopic pregnancy. J Mol Histol 2013; 44: 167–173. on the diagnosis of diabetes mellitus. Diabetes Care 2003; 26: 3160–3167. 37 Calamita G, Portincasa P. Present and future therapeutic strategies in 20 Brunt EM. Nonalcoholic steatohepatitis. Semin Liver Dis 2004; 24: 3–20. non-alcoholic fatty liver disease. Expert Opin Ther Targets 2007; 11: 1231–1249. 21 Younossi ZM, Stepanova M, Rafiq N, Makhlouf H, Younoszai Z, Agrawal R et al. 38 Widhalm K, Ghods E. Nonalcoholic fatty liver disease: a challenge for pediatricians. Pathologic criteria for nonalcoholic steatohepatitis: interprotocol agreement and Int J Obes (Lond) 2010; 34: 1451–1467. ability to predict liver-related mortality. Hepatology 2011; 53: 1874–1882. 39 Postic C, Girard J. The role of the lipogenic pathway in the development 22 Bell LN, Wang J, Muralidharan S, Chalasani S, Fullenkamp AM, Wilson LA et al. of hepatic steatosis. Diabetes Metab 2008; 34: 643–648. Relationship between adipose tissue insulin resistance and liver histology in 40 Calamita G, Ferri D, Gena P, Carreras FI, Liquori GE, Portincasa P et al. Altered nonalcoholic steatohepatitis: a pioglitazone versus vitamin E versus placebo expression and distribution of aquaporin-9 in the liver of rat with obstructive for the treatment of nondiabetic patients with nonalcoholic steatohepatitis trial extrahepatic cholestasis. Am J Physiol Gastrointest Liver Physiol 2008; 295: G682–G690. follow-up study. Hepatology 2012; 56: 1311–1318. 41 Potter JJ, Koteish A, Hamilton J, Liu X, Liu K, Agre P et al. Effects of acetaldehyde 23 Asai M, Higuchi S, Kubota M, Iguchi K, Usui S, Hirano K. Regulators for blood on hepatocyte glycerol uptake and cell size: implication of aquaporin 9. Alcohol glucose level affect gene expression of aquaporin 3. Biol Pharm Bull 2006; 29: Clin Exp Res 2011; 35: 939–945. 991–996. 42 Portincasa P, Calamita G. Water channel proteins in bile formation and flow in 24 Tsukaguchi H, Weremowicz S, Morton CC, Hediger MA. Functional and molecular health and disease: when immiscible becomes miscible. Mol Aspects Med 2012; characterization of the human neutral solute channel aquaporin-9. Am J Physiol 33: 651–664. 1999; 277: F685–F696. 43 Potter JJ, Koteish A, Hamilton J, Liu X, Liu K, Agre P et al. Effects of acetaldehyde 25 Rodrı´guez A, CatalańV,Go´mez-Ambrosi J, Fru¨hbeck G. Role of aquaporin-7 in the on hepatocyte glycerol uptake and cell size: implication of aquaporin 9. Alcohol pathophysiological control of fat accumulation in mice. FEBS Lett 2006; 580: Clin Exp Res 2011; 35: 939–945. 4771–4776. 44 Lehmann GL, Carreras FI, Soria LR, Gradilone SA, Marinelli RA. LPS induces 26 Fru¨ hbeck G, Go´mez-Ambrosi J, Salvador J. Leptin-induced lipolysis opposes the the TNF-alpha-mediated downregulation of rat liver aquaporin-8: role in tonic inhibition of endogenous adenosine in white adipocytes. FASEB J 2001; 15: sepsis-associated cholestasis. Am J Physiol Gastrointest Liver Physiol 2008; 294: 333–340. G567–G575.

Supplementary Information accompanies this paper on International Journal of Obesity website (http://www.nature.com/ijo)

International Journal of Obesity (2014) 1213 – 1220 & 2014 Macmillan Publishers Limited