Dr K. Cusi, Division of Diabetes, The University of Texas Health Science Center at San Antonio, Room 3.380S, 7703 Floyd Curl Drive, San Antonio, TX 78284-3900, USA. E-mail: firstname.lastname@example.org
Background Plasma adiponectin is decreased in NASH patients and the mechanism(s) for histological improvement during thiazolidinedione treatment remain(s) poorly understood.
Aim To evaluate the relationship between changes in plasma adiponectin following pioglitazone treatment and metabolic/histological improvement.
Methods We measured in 47 NASH patients and 20 controls: (i) fasting glucose, insulin, FFA and adiponectin concentrations; (ii) hepatic fat content by magnetic resonance spectroscopy; and (iii) peripheral/hepatic insulin sensitivity (by double-tracer oral glucose tolerance test). Patients were then treated with pioglitazone (45 mg/day) or placebo and all measurements were repeated after 6 months.
Results Patients with NASH had decreased plasma adiponectin levels independent of the presence of obesity. Pioglitazone increased 2.3-fold plasma adiponectin and improved insulin resistance, glucose tolerance and glucose clearance, steatosis and necroinflammation (all P < 0.01–0.001 vs. placebo). In the pioglitazone group, plasma adiponectin was significantly associated (r = 0.52, P = 0.0001) with hepatic insulin sensitivity and with the change in both variables (r = 0.44, P = 0.03). Increase in adiponectin concentration was related also to histological improvement, in particular, to hepatic steatosis (r = −0.46, P = 0006) and necroinflammation (r = −0.56, P < 0.0001) but importantly also to fibrosis (r = −0.29, P = 0.03).
Conclusions Adiponectin exerts an important metabolic role at the level of the liver, and its increase during pioglitazone treatment is critical to reverse insulin resistance and improve liver histology in NASH patients.
Insulin resistance and low plasma adiponectin concentration are primary characteristics of NAFLD/NASH patients.1–5 However, the role of adiponectin in NASH remains unclear. Adiponectin effects, although not completely defined, involve the activation of peroxisome proliferator-activated receptors (PPAR) and AMP-activated protein kinase (AMPK), which lead to an increase in fatty acid oxidation in muscle and liver and decreased tissue content of triglycerides (TG). Studies in mice have shown that infusion of adiponectin reduces hepatic steatosis and necroinflammation.6 Moreover, high adiponectin decreases plasma TG levels by increasing skeletal muscle lipoprotein lipase and VLDL receptor expression and consequently, VLDL-TG catabolism.7 In animal models of steatohepatitis, adiponectin has been shown to ameliorate liver damage and block fibrogenesis.6, 8 Adiponectin knockout mice develop a more extensive liver fibrosis compared with wild-type animals, whereas adenovirus-mediated overexpression of adiponectin ameliorate liver damage in wild-type mice.8 Adiponectin inhibits expression of several proinflammatory cytokines9 and treatment with adiponectin suppresses hepatic TNF-α mRNA expression and diminishes plasma levels of TNF-α, indicating that the protective role of adiponectin might be explained, at least partially, by its antagonistic effect on TNF-α.6
In patients with NASH, hepatic adiponectin receptors are diminished,10 although it remains controversial if plasma adiponectin levels predict the severity of steatohepatitis.11–13 Pioglitazone treatment increases adiponectin concentrations, improves hepatic insulin resistance and liver histology in NASH.14–16 Thus, it has been suggested that adiponectin may play an important role in mediating the beneficial effects of pioglitazone in NASH patients, although it is likely that its metabolic effects are not associated with changes in adiponectin receptor expression.12 The goal of this study was to evaluate, in NASH patients, how changes in adiponectin following pioglitazone treatment are related to metabolic and histological improvement.
The protocol has been previously described.14 To the original 10 lean controls with normal glucose tolerance and without a fatty liver in our prior report, we included an additional 10 healthy subjects (n = 20 controls). The obese subjects recruited were otherwise healthy and chosen because they lacked the typical risk factors for NASH (such as elevated LFTs, hypertension, insulin resistance, type 2 diabetes). They probably had a fatty liver, but not NASH. A liver biopsy could not be ethically justified to confirm this. Participants were recruited from the University of Texas H.S.C. at San Antonio, the Audie L. Murphy Veteran Administration Medical Center and Brooke Army Medical Center, San Antonio, TX. After excluding other aetiologies for liver disease, the diagnosis of NASH was confirmed by a liver biopsy. The protocol was approved by the medical school Institutional Review Board and all subjects gave written informed consent prior to participation.
During the 4-week run-in period, subjects were instructed not to change the caloric content of their diet or their physical activity level. Subjects were started on placebo and compliance assessed by pill-count on follow-up visits. Baseline metabolic measurements were performed between weeks 3 and 4 and repeated at the end of the study. They included (i) fasting plasma glucose, A1c, lipid profile, insulin, free fatty acid (FFA) and adiponectin concentrations; (ii) whole body fat by dual-energy X-ray absorptiometry (DXA); (iii) hepatic fat content by magnetic resonance spectroscopy; (iv) double-tracer oral glucose tolerance test (OGTT) for the determination of peripheral and hepatic insulin sensitivity.
After baseline metabolic measurements, participants were randomized to either oral placebo or pioglitazone (ACTOS; Takeda Pharmaceuticals Inc., Deerfield, IL) 30 mg/day, titrated after 2 months to 45 mg/day until the end of the study (6 months). All patients were instructed at baseline by a research dietician to reduce their intake by −500 kcal/day. This instruction continued during follow-up visits in all participants. Every 2 weeks, subjects were followed up by the investigators who recorded vital signs, performed a physical examination, reviewed home glucose monitoring results if diabetics, assessed study drug compliance (confirmed by pill-count) and adverse events. Blood sample was drawn for liver transaminases and metabolic measurements. After 6 months of treatment, subjects repeated all baseline measurements and the study was considered completed.
Subjects were admitted to the research unit (06:00 hours) after an overnight fast. A primed (25 μCi/min × FPG/100)-continuous (0.25 μCi/min) [3-3H] glucose infusion was started and continued until study end to measure hepatic glucose production and peripheral glucose clearance. After allowing a 3-h isotopic equilibration, four blood samples were drawn every 10 min for plasma 3H-glucose radioactivity and glucose, FFA, insulin concentrations. At 09:30 hours, patients received a 75-g oral glucose load containing 1-14C-glucose (100 μCi). Blood sample was drawn every 15–30 min for the next 4 h.
Body fat content measurements
Total body fat content and fat-free mass (FFM) were measured by DXA (Hologic Inc., Waltham, MA, USA). For the determination of hepatic fat content, localized 1H NMR spectra of the liver were acquired on a 1.9T MRI scanner (Elscint Prestige Ltd, Haifa, Israel) as previously described.14
Plasma glucose concentration was determined by the glucose oxidase method (Beckman Instruments Inc., Fullerton, CA, USA), plasma insulin levels by radioimmunoassay (Diagnostics Products, Los Angeles, CA, USA) and plasma FFA concentration spectrophotometrically (Wako Chemicals Gmbh, Neuss, Germany). Plasma glucose radioactivity was measured on barium hydroxide/zinc sulphate-precipitated plasma extracts. Plasma adiponectin levels were measured by RIA (Linco Research, St Charles, MO, USA).
In the basal state, endogenous glucose production (EGP) was calculated by dividing the tracer infusion rate by the mean basal plasma 3H glucose specific activity.14 After glucose load, glucose rate of disappearance (Rd) was calculated using Steele’s equation.14 The metabolic clearance rate of glucose or MCR was calculated as mean Rd (0–240 min) divided by the mean plasma glucose and normalized by mean insulin concentration. Hepatic insulin sensitivity under basal conditions was determined as the inverse of the product of EGP and fasting plasma insulin (μmol/min kg nmol/L)−1, as previously described.14, 17, 18
Data are given as the mean ± S.E. if normally distributed. Variables with skewed distribution are expressed as median and interquartile range (in square brackets). Group differences were analysed by Student t-test, Mann–Whitney U-test and χ2-test, for normally distributed, non-normally distributed and noncontinuous variables respectively. Univariate and multivariate analysis were used to estimate associations among continuous variables in the whole dataset. A two-tailed P < 0.05 was considered statistically significant. To study the impact of pioglitazone changes in adiponectin vs. variables of interest (Figures 2 & 3) we performed a regression analysis using both pre- and post-values in the subgroup of subjects treated with pioglitazone. Correlation coefficient and P-values were reported in the graphs as well as in the text.
We studied 47 patients with NASH and 20 normal controls.14 Each group of subjects was subdivided into obese and non-obese individuals (BMI < 30 kg/m2). Patients with NASH were more insulin resistant independent of their obesity or glucose tolerance as shown by reduced glucose clearance during OGTT (i.e. peripheral insulin resistance) and decreased hepatic insulin sensitivity index (Table 1). Plasma adiponectin concentration was reduced by ∼50% in patients with NASH, regardless of obesity or glucose tolerance status (Figure 1).
Table 1. Anthropometric and clinical characteristics
Control non-obese (n = 13)
Control obese (n = 7)
NASH non-obese (n = 11)
NASH obese (n = 36)
BMI, body mass index, FFA, free fatty acid; AST, aspartate aminotransferase; ALT, alanine aminotransferase; OGTT, oral glucose tolerance test.
Data are given as the mean ± S.E. if normally distributed. Variables with skewed distribution are expressed as median and interquartile range (in square brackets).
* P < 0.05 vs. lean control.
† P < 0.05 vs. obese controls.
‡ P < 0.05 lean patients with NASH.
45 ± 3
46 ± 5
54 ± 3
50 ± 1
25.4 ± 0.9
34.8 ± 1.4*
27.8 ± 0.4†
35.2 ± 0.7*‡
Body weight (kg)
71.7 ± 3.8
80.8 ± 7.2
71.8 ± 2.3
97.8 ± 2.2*†‡
Body fat (kg)
23.5 ± 2.3
35.4 ± 3.7
24.1 ± 1.2
35.8 ± 1.5
Liver fat (%)
4.3 ± 1.5
12.3 ± 2.5
14.3 ± 2.7
24.1 ± 2.3
Fasting glucose (mmol/L)
5.4 ± 0.1
5.6 ± 0.2
6.1 ± 0.4
6.3 ± 0.3
Fasting insulin (pmol/L)
18 ± 1
21 ± 2
35 ± 5*
41 ± 2*†
16 ± 1
20 ± 1
38 ± 6*
59 ± 4*†‡
579 ± 56
527 ± 61
770 ± 57*†
695 ± 32†‡
14.1 ± 1.7
12.5 ± 2.1
6.8 ± 1.1*†
6.4 ± 0.5*†
Mean OGTT glucose clearance (mL/min kg)
3.9 ± 0.4
3.1 ± 0.2
2.5 ± 0.3*
2.2 ± 0.1*†
Hepatic insulin sensitivity (μmol/min kg nmol/L)−1
4.5 ± 0.6
3.0 ± 0.3*
1.9 ± 0.5*
1.4 ± 0.1*†
Effect of pioglitazone treatment
NASH patients were randomized to pioglitazone (n = 26) or placebo (n = 21). Pioglitazone treatment increased plasma adiponectin concentration by 2.3-fold (P < 0.001 vs. placebo), whereas no change was observed in the placebo group (Figure 2). We further explored the relationship between the improvement in adiponectin and in hepatic insulin sensitivity with pioglitazone treatment. We observed that in the pioglitazone group, the plasma adiponectin concentration was significantly associated (r = 0.52, P = 0.0001) with hepatic insulin sensitivity (Figure 2). Changes between both variables were significantly correlated (r = 0.44, P = 0.03).
We have previously shown that pioglitazone improved hepatic histology in NASH patients.14 In this analysis, we aimed to examine histological improvements relative to changes in plasma adiponectin concentration. There was a strong inverse relationship between reduction in the mean score of hepatic steatosis (r = −0.46, P = 0.0006), necroinflammation (r = −0.56, P < 0.0001) and overall NASH activity score (r = −0.57, P < 0.0001) observed after pioglitazone treatment. Of interest, there was also a significant inverse relationship between adiponectin and hepatic fibrosis (r = −0.29, P = 0.03; Figure 3).
Patients with NASH have low plasma adiponectin levels, which are inversely related to insulin resistance and hepatic TG content.5, 14, 19 Animal studies have provided solid proof-of-concept that low plasma adiponectin levels are involved in the aetiology of NAFLD/NASH and that adiponectin deficiency is an important factor in the progression of fibrosis.13 However, in humans, this relationship has been difficult to validate as few interventions have a significant impact on adiponectin levels. Here, we provide a comprehensive analysis of the relationship between adiponectin and metabolic and histological changes in patients with NASH.
In this study, subjects with NASH had decreased plasma adiponectin levels, independent of the degree of obesity or glucose tolerance status (Figure 1). A key finding is that in non-obese patients with NASH, plasma adiponectin concentrations were ∼50% lower than in obese controls (Table 1) and similar to those in obese patients with NASH. This finding suggests that it is not the total amount of adipose tissue, but rather its biological function that determines plasma adiponectin release by the adipocyte.
Thiazolidinediones increase plasma adiponectin levels and this has been shown in both patients with type 2 diabetes15, 19and those with NASH.1, 14, 20 The current data suggest that the increase in plasma adiponectin levels could mediate some of the insulin-sensitizing effects of PPARγ agonists. Pioglitazone more than doubled adiponectin concentrations that became similar to the values observed in control subjects. Moreover, the reduction in hepatic steatosis observed after pioglitazone treatment was inversely correlated with the increase in plasma adiponectin (Figure 3, top panel). How exactly adiponectin reverses hepatic lipid accumulation during TZD therapy is still unclear. Plasma adiponectin levels have been related to both hepatic21, 22 and peripheral insulin sensitivity.22 Secretion of adiponectin by the adipose tissue is known to inhibit hepatic fatty acid synthesis, gluconeogenesis and de novo lipogenesis, via activation of AMPK.23 It also activates PPAR-α with stimulation of fatty acid oxidation.24 For reasons of these well-known effects in vitro and in vivo, we felt compelled to examine the relationship between changes in total plasma adiponectin concentration and changes in insulin resistance in our NASH cohort. Pioglitazone improved insulin sensitivity in muscle and adipose tissue,14 but in particular enhanced liver insulin sensitivity with improved fasting and postprandial (i.e. during the OGTT) glucose metabolism. One may speculate that the insulin-sensitizing effect of adiponectin may have not only reduced hepatic TG synthesis but also played a role in reducing VLDL secretion25 and plasma TG levels, as observed with treatment.14 More importantly, improvement in hepatic insulin sensitivity was proportional to the increase in adiponectin concentration.
Before treatment, both obese and non-obese NASH patients had similarly low plasma adiponectin concentrations and the degree of necroinflammation and fibrosis were not different (data not shown). However, liver fat was 70% higher in the obese patients with NASH, suggesting that plasma adiponectin level and not steatosis per se could be the most important determinant of disease severity. Whether adiponectin is a biomarker of disease severity or plays a mechanistic role remains to be established. Pioglitazone significantly improved not only steatosis but also necroinflammation and fibrosis, in contrast to just a mild improvement in inflammation in the placebo plus diet group.14 Of note, the histological improvement was directly related to the increase in adiponectin (Figure 3). This effect could be indirect through an amelioration of hyperinsulinemia, hyperglycaemia and FFA-induced lipotoxicity, or mediated through direct effects on the hepatic tissue. In addition to the metabolic effects on hepatocytes, adiponectin could act as an anti-inflammatory agent reversing the activation of intracellular pro-inflammatory signalling pathways (NFkB, JNK, etc.) observed in patients with NASH.11, 26, 27
As reported previously,14 liver fibrosis improved within the pioglitazone group (P = 0.002), although falling short of reaching statistical significance against placebo (P = 0.08). However, the improvement in fibrosis within the pioglitazone group also correlated with the increase in plasma adiponectin concentration (Figure 3). This finding is in agreement with studies in vitro and in vivo in animals showing that adiponectin has hepatoprotective and antifibrogenic effects. Administration of adiponectin to mice has been shown to ameliorate liver damage induced by both alcohol and obesity.6 In a diet-induced NASH model, adiponectin-knockout mice have enhanced liver fibrosis, inflammation and tumour formation compared with wild-type animals.28 Of particular relevance to the role of adiponectin in hepatic fibrosis is the observation that adiponectin receptors are present on hepatic stellate cells and administration of adiponectin in vitro inhibits HSC activation and collagen production.29 Taken together, these findings suggest a greater need to examine the mechanisms involved by which adiponectin ameliorates inflammation and fibrosis in NASH.
In conclusion, this work highlights the close interaction between adiponectin, insulin action at the level of the liver and histological changes in patients with NASH. Although much work is needed at the molecular and in translational research to understand this relationship fully, these results are of value in establishing a strong link among them and may imply that pharmacological modulation of plasma adiponectin secretion or sensitivity, ideally without weight gain and other thiazolidinedione-related known side effects, may play an important role in patients with NASH.
The authors thank the study volunteers, the nursing staff, and the nutrition and laboratory staff for their assistance in performing the described studies. Declaration of personal interests: Gastaldelli A, Harrison S, Belfort R, Hardies J, Balas B, Schenker S and Cusi K have no conflicts of interest. Declaration of funding interests: The project described (and Cusi K) was supported by the Veterans Affairs Medical Research Fund and by Award Number UL 1RR025767 from the National Center for Research Resources. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources of the National Institutes of Health. A.G. was supported by the Italian National Research Council. Partial support was received for the original study from Takeda Pharmaceuticals.