Effect of adipose tissue insulin resistance on metabolic parameters and liver histology in obese patients with nonalcoholic fatty liver disease


  • Potential conflict of interest: Nothing to report.


The role of adipose tissue insulin resistance in the pathogenesis of nonalcoholic fatty liver disease (NAFLD) remains unclear. To evaluate this, we measured in 207 patients with NAFLD (age = 51 ± 1, body mass index = 34.1 ± 0.3 kg/m2) and 22 controls without NAFLD (no NAFLD) adipose tissue insulin resistance by means of a validated index (Adipo-IRi = plasma free fatty acids [FFA] x insulin [FPI] concentration) and as the suppression of plasma FFA during an oral glucose tolerance test and by a low-dose insulin infusion. We also explored the relationship between adipose tissue insulin resistance with metabolic and histological parameters by dividing them based on quartiles of adipose tissue insulin resistance (Adipo-IRi quartiles: Q1 = more sensitive; Q4 = more insulin resistant). Hepatic insulin resistance, measured as an index derived from endogenous glucose production x FPI (HIRi), and muscle insulin sensitivity, were assessed during a euglycemic insulin clamp with 3-[3H] glucose. Liver fat was measured by magnetic resonance imaging and spectroscopy, and a liver biopsy was performed to assess liver histology. Compared to patients without steatosis, patients with NAFLD were insulin resistant at the level of adipose tissue, liver, and skeletal muscle and had higher plasma aspartate aminotransferase and alanine aminotransferase, triglycerides, and lower high-density lipoprotein cholesterol and adiponectin levels (all P < 0.01). Metabolic parameters, hepatic insulin resistance, and liver fibrosis (but not necroinflammation) deteriorated as quartiles of adipose tissue insulin resistance worsened (all P < 0.01). Conclusion: Adipose tissue insulin resistance plays a key role in the development of metabolic and histological abnormalities of obese patients with NAFLD. Treatment strategies targeting adipose tissue insulin resistance (e.g., weight loss and thiazolidinediones) may be of value in this population. (HEPATOLOGY 2012)

Insulin resistance (IR) plays a key role in the development of hepatic steatosis in nonalcoholic fatty liver disease (NAFLD). Previous studies have reported that patients with NAFLD are insulin resistant at the level of the liver and muscle.1-8 Although fat is another important target of insulin action, only a few studies have reported on the role of adipose tissue in patients with NAFLD.2, 8, 9 Moreover, it remains unclear whether the degree of steatohepatitis depends more upon total adiposity per se (i.e., body mass index; BMI) or the severity of adipose tissue dysfunction and resistance to insulin action (i.e., liver exposure to elevated plasma free fatty acids; FFA).

Adipose tissue is important under normal living conditions, being the primary source of FFA (∼70%) for hepatic triglyceride (TG) synthesis.10 Excess release of FFA plays a key role in the development of hepatic “lipotoxicity” in NAFLD.7, 11-13 Failure of insulin to inhibit TG lipolysis in insulin-resistant states leads to the oversupply of FFA to the liver, excess hepatic TG synthesis, and intracellular accumulation of toxic lipid products that impair insulin signaling and activate inflammatory pathways (i.e., nuclear factor of kappa light polypeptide gene enhancer in B-cell inhibitor/nuclear factor kappa light-chain enhancer of activated B cells and Jun N-terminal kinase pathways, Toll-like receptor 4, and others). Adaption to this metabolic stress involves hepatic IR, dyslipidemia and steatohepatitis with mitochondrial dysfunction, endoplasmic reticulum stress, release of reactive oxygen species, and hepatocellular damage.14 Whether the degree of hepatic IR and steatohepatitis is proportional to the magnitude of adipose tissue IR has not been carefully examined. Only one study has reported that the severity of adipose tissue IR is closely correlated with histological damage in patients with nonalcoholic steatohepatitis (NASH) as well as its improvement with a thiazolidinedione (TZD) to the reversal of dysfunctional fat.9

To better understand the relationship between the degree of adipose tissue IR, hepatic steatosis, and NASH, we studied in depth the metabolic and histological profiles of patients with and without NAFLD. If liver disease mirrors adipose tissue dysfunction, future therapies aimed at dysregulated adipose tissue may hold particular promise in NAFLD.


A1c, test for glycated hemoglobin; Adipo-IR, Adipo-IRi, adipose tissue insulin resistance index; ALT, alanine aminotransferase; ANOVA, analysis of variance; AST, aspartate aminotransferase; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; EGP, endogenous glucose production; FFA, free fatty acids; FPI, free plasma insulin; HDL-C, high-density lipoprotein cholesterol; HIRi, hepatic insulin resistance index; HSCs, hepatic stellate cells; IR, insulin resistance; LDL-C, low-density lipoprotein cholesterol; MetS, metabolic syndrome; MHO, metabolically healthy obese; MRI, magnetic resonance imaging; MRS, magnetic resonance imaging and spectroscopy; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score; NASH, nonalcoholic steatohepatitis; OGTT, oral glucose tolerance test; Rd, insulin-stimulated glucose disposal; T2DM, type 2 diabetes mellitus; TG, triglyceride; TZD, thiazolidinedione.

Patients and Methods


We recruited a total of 229 subjects. This included (1) lean subjects without NAFLD, as the gold-standard reference group for metabolic variables, and (2) “metabolically healthy obese” (MHO) subjects with normal adipose tissue insulin sensitivity (as defined below) and without NAFLD, as a control group versus obese patients with NAFLD. Body weight (± 2%) and physical activity were stable for at least 3 months before the study, as assessed by validated questionnaires (food intake: Questionnaire-2005 Food Frequency Questionnaire; physical activity: Energy Expenditure Survey-Adults; Nutritionquest, Berkeley, CA). All subjects underwent a medical history, physical examination, routine chemistries, and electrocardiography. Volunteers were excluded if they had a history of alcohol abuse (≥20 g/day) or liver, renal, pulmonary, and/or heart disease. Baseline data from 47 patients have been previously reported.4 Informed written consent was obtained from each patient before participation.

Study Design.

Metabolic measurements at our research unit included the following: (1) adipose tissue IR (fasting plasma insulin [FPI] x free fatty acids [FFA]); (2) euglycemic hyperinsulinemic clamp with 3-[3H]-glucose to measure endogenous (primarily hepatic) glucose production (EGP) and whole body (largely muscle) insulin-stimulated glucose disposal (Rd); (3) fasting plasma glucose, insulin, and FFA levels every 30 minutes during a 2-hour oral glucose tolerance test (OGTT); (4) liver fat by magnetic resonance imaging (MRI) and spectroscopy (MRS); (5) whole body fat by dual-energy X-ray absorptiometry (DXA) (Hologic, Inc., Waltham, MA); and (6) liver biopsy for the diagnosis, grading, and staging of NASH. Endogenous glucose production and Rd were calculated as previously reported.15, 16 Indices of hepatic IR (HIRi = EGP × FPI) and of adipose tissue IR (Adipo-IRi = FFA × FPI) were calculated based on the linear relationship between the rise in the FPI level and inhibition of the rate of basal (i.e., fasting) EGP and of plasma FFA, respectively, in healthy subjects.17 The higher the rate of fasting EGP and of plasma FFA levels, the greater the severity of hepatic and adipose tissue IR, respectively. Experimental validation for both indices has been published previously by our group.8, 9, 18 Additionally, to investigate the relationship between adipose tissue IR with metabolic and histological parameters, we divided patients with NAFLD based on quartiles of Adipo-IRi (Q1 = more sensitive and Q4 = more insulin-resistant adipose tissue).

Visceral, liver and whole body fat content

MRS was used to measure visceral fat, as reported before.5 Hepatic fat content was measured by a Siemens TIM TRIO 3.0T MRI whole body scanner (Siemens Healthcare Diagnostics, Deerfield, IL), as previously described,4, 5, 8 and was considered diagnostic of NAFLD if >5.5%.

Euglycemic hyperinsulinemic clamp19

After an overnight fast, subjects were studied at the research unit, as described before,8, 15, 16, 18 with the infusion of 3-[3H]-glucose to measure glucose turnover. After the basal equilibration period, insulin was administered as a primed-continuous infusion at 10 μU/m2·min for 120 minutes to assess the suppression of EGP (hepatic), followed by an insulin infusion rate of 80 μU/m2·min for 120 minutes to assess Rd. A variable 20% glucose infusion maintained plasma glucose at ∼90-100 mg/dL.

Liver biopsy

Liver biopsies were not performed in MHO subjects, because there was no clinical indication in the absence of abnormal liver aminotransferases or liver steatosis by MRS. A total of 207 subjects were diagnosed with NAFLD by MRS and all were asked to have an ultrasound-guided liver biopsy. Of these, 66% (n = 136) agreed and 71 refused. The percent of patients accepting to have a liver biopsy was similar among quartiles Q1 through Q3 (55%, 54%, and 65%, respectively) and was only higher for Q4 because of the higher liver aminotransferases, which made a more compelling case for the patient to accept the procedure. There were no differences within each quartile between those that received or did not receive a liver biopsy regarding age, gender, BMI, whole body fat, liver fat by MRS, fasting glucose, test for glycated hemoglobin (A1c), lipids, adiponectin, Adipo-IRi, HIRi, and suppression of EGP or FFA by insulin or visceral fat, and it was the metabolic characteristics what defined each quartile (not a histological one), and data were analyzed within each quartile together. An experienced pathologist unaware of the subjects' identity or clinical information evaluated biopsies based on standardized criteria.20 Definite NASH was diagnosed in 61% of patients in Q1, 52% in Q2, 63% in Q3, and 68% in Q4. The intraobserver agreement between readings was good to excellent (weighted kappa coefficient: 0.84 for steatosis, 0.69 for necroinflammation, and 0.82 for fibrosis).4

Analytical Methods.

Plasma insulin was measured by radioimmunoassay, FFA by standard colorimetric methods, and adiponectin by Luminex beads (Millipore Corp., St. Charles, MO). Plasma glucose radioactivity was measured from deproteinized plasma samples precipitated from barium hydroxide/zinc sulfate.

Statistical Analysis.

All values are reported as the mean ± standard error of the mean (SEM) for continuous variables and the number (i.e., percent) for categorical variables. Comparison of between groups was performed using analysis of variance (ANOVA) or the Kruskal-Wallis test for continuous variables or Pearson's chi-square or Fisher's exact test for categorical variables. Adjusted P values were calculated using fixed-effect models. A P value of <0.05 was considered statistically significant. All statistical calculations were performed using JMP software (version 8.0.2 [8.0]; SAS Institute, Inc., Cary, NC).


Subject Characteristics

Comparison of obese patients with and without NAFLD

Patient characteristics are summarized in Tables 1 and 2. As expected, lean patients without NAFLD had a more favorable metabolic profile versus obese with or without NAFLD (Table 1).

Table 1. Patient Characteristics
 Lean Group without NAFLDObese Group without NAFLDNAFLD Group
  • P values were derived from Pearson's chi-square test for categorical variables (Fisher's exact test when expected numbers were small), from ANOVA for age, BMI, total body, and liver fat, and from the Kruskal-Wallis test for laboratory measurements.

  • Abbreviations: M, male; F, female.

  • *

    Diagnosis was based on NCEP guidelines (as per ATPIII criteria).

  • Direct LDL-C was measured when triglycerides >400 mg/dL.

  • P values were calculated with the Wilcoxon signed-rank test.

  • §

    P < 0.001; ∥P < 0.01; ¶P < 0.05 versus subjects with NAFLD.

  • #

    P < 0.01 versus MHO and patients with NAFLD.

  • **

    P < 0.01 to P < 0.05 versus patients with NAFLD.

Age (years)35 ± 4#50 ± 351 ± 1
Gender (M/F, %)56/4485/1562/38
BMI (kg/m2)22.2 ± 1.6**30.0 ± 1.3∥34.1 ± 0.3
Whole body fat by DXA (%)22.1 ± 5.2**26.1 ± 2.8∥34 ± 0.6
Liver fat content by MRI (%)1.2 ± 0.4**1.7 ± 0.4§24.7 ± 0.9
A1c (%)5.1 ± 0.3**5.6 ± 0.3∥6.3 ± 0.1
Fasting plasma glucose (mg/dL)93 ± 1**111 ± 8120 ± 2
Presence of MetS (%)*0#23§92
AST (IU/L)28 ± 1026 ± 3¶42 ± 2
ALT (IU/L)21 ± 7**24 ± 4∥59 ± 2
Normal aminotransferases (%)100**90¶48
Plasma cholesterol (mg/dL)168 ± 6176 ± 8184 ± 2
Plasma LDL-C (mg/dL)103 ± 8107 ± 6112 ± 3
Plasma HDL-C (mg/dL)50 ± 4**50 ± 3§38 ± 1
Plasma triglycerides (mg/dL)79 ± 13**92 ± 10∥173 ± 8
Fasting plasma insulin (μU/L)3.2 ± 0.4**3.8 ± 0.5§14.3 ± 0.7
Fasting plasma FFA (μmol/L)406 ± 47**324 ± 25§603 ± 16
Suppression of plasma FFA by insulin (%)92 ± 4**68 ± 8¶49 ± 2
Adiponectin (μg/mL)19.9 ± 2.9#12.9 ± 1.3§8.0 ± 0.3
Adipo-IRi (mmol/L·μU/mL)1.2 ± 0.1**1.2 ± 0.1§8.0 ± 0.5
HIRi (mg·kg−1·min−1·μU/mL)9.4 ± 0.87.3 ± 0.8¶22.4 ± 1.6
Suppression of EGP by insulin (%)68 ± 159 ± 2¶43 ± 2
Table 2. Clinical Characteristics by Quartiles Based on Adipose Tissue Insulin Resistance in Patients With and Without NAFLD
 No NAFLDPatients With NAFLD
Adipose tissue IRNormalQ1 (less)Q2Q3Q4 (Worse)
  • Abbreviations: M, male; F, female; VAT, visceral abdominal tissue; SAT, subcutaneous abdominal tissue.

  • *

    P < 0.001 to P < 0.05 versus NAFLD+;

  • P < 0.01 to P < 0.05 versus all groups;

  • P < 0.05 versus Q4.

Age (years)50 ± 347 ± 252 ± 151 ± 151 ± 1
Gender (M/F) (%)85/1558/4268/3260/4059/41
BMI (kg/m2)30.0 ± 1.3*32.8 ± 0.833.6 ± 0.633.8 ± 0.635.7 ± 0.6
Whole body fat (%)26.1 ± 2.8*33.8 ± 1.534.1 ± 1.034.0 ± 1.034.9 ± 1.0
Liver fat (%)1.7 ± 0.4*22.2 ± 2.120.7 ± 1.724.1 ± 1.628.9 ± 1.6
Visceral fat (VAT) (cm2)123 ± 22171 ± 16173 ± 11175 ± 12192 ± 12
VAT/SAT0.4 ± 0.10.5 ± 0.10.5 ± 0.10.6 ± 0.10.5 ± 0.1
A1c (%)5.6 ± 0.36.2 ± 0.26.3 ± 0.16.2 ± 0.16.7 ± 0.1
Plasma adiponectin (μg/mL)12.9 ± 2.3*8.0 ± 0.88.8 ± 0.67.9 ± 0.67.2 ± 0.6
HOMA (mg/dL·μU/mL/405)1.0 ± 0.11.7 ± 0.42.8 ± 0.33.8 ± 0.37.9 ± 0.3

The use of the Adipo-IRi allowed us to separate the obese patients based on this index into MHO (n = 13) with normal adipose tissue insulin sensitivity and without NAFLD by MRS versus obese subjects with adipose tissue IR (elevated Adipo-IRi) and NAFLD by MRS. The MHO group had a similar BMI compared to NAFLD patients, with this group being just slightly less obese by DXA (P < 0.05). Despite the similar BMI, liver fat content in patients with NAFLD was much higher versus MHO subjects (24.7% ± 0.9% versus 1.7% ± 0.4%, respectively; P < 0.0001) and a higher prevalence of metabolic syndrome (MetS) (92% vs. 23%; P < 0.001), higher aspartate aminotransferase (AST) and alanine aminotransferase (ALT), and a much worse lipid profile (i.e., triglycerides [TGs] and high-density lipoprotein cholesterol [HDL-C]; all P < 0.05 to P < 0.001). Plasma adiponectin was >50% higher in lean, compared to obese, patients without NAFLD, suggesting an early defect in adipose tissue regulation, whereas obese NAFLD patients had a further decrease, compared to MHO subjects (Table 1). This was consistent with severe adipose tissue IR and worse Adipo-IRi in patients with NAFLD versus MHO patients (P < 0.001).

We also examined the effect of adipose tissue IR across other target tissues. Patients with NAFLD had severe hepatic IR, compared to MHO patients, either measured as the HIRi or the suppression of EGP (hepatic) by low-dose insulin infusion (both P = 0.05) (Table 1). Patients with NAFLD also had severe muscle IR, compared to MHO patients, with ∼50% reduction in Rd (5.8 ± 0.3 versus 12.1 ± 0.8 mg·kgLBM−1·min−1; P < 0.0001).

Comparison of MHO to patients with NAFLD divided by quartiles of adipose tissue IR

To further investigate the relationship between adipose tissue IR with metabolic and histological parameters, we divided NAFLD patients based on quartiles of Adipo-IRi (Q1 = more sensitive; Q4 = more insulin-resistant adipose tissue). The four groups of patients with NAFLD were well matched for age, gender, body fat, visceral fat, and A1c (Table 2). Adipose tissue IR was associated with a threshold effect in relation to liver fat content between patients without, compared to those with, NAFLD, increasing rapidly in the presence of dysfunctional fat (Q1) and remaining rather constant between quartiles 1 to 3. However, there was a progressive stepwise increase in the homeostasis model assessment (HOMA), an indirect measure of hepatic IR. This was consistent with worsening IR at the level of the liver when directly assessed by means of the HIRi from Q1 to Q4 (see below). Similarly, plasma adiponectin decreased markedly by 39% and abruptly at the least severe quartile (Q1) versus MHO patients (P < 0.001), but did not deteriorate further from Q2 to Q4.

Effect of adipose tissue IR on postprandial insulin and FFA concentration

There was no significant difference in fasting (3.2 ± 0.4 versus 3.8 ± 0.5 μU/L, respectively; P = 0.82) or postprandial (28.7 ± 6.2 versus 27.8 ± 3.6 μU/L, respectively; P = 0.98) plasma insulin concentration between lean and MHO patients (Fig. 1A). In contrast, fasting plasma insulin increased by 1.5-, 2.5-, 3.4-, and 6.2-fold from Q1 to Q4 (from 6.4 ± 0.5 to 24.7 ± 1.4 μU/L; Q4 versus MHO; P < 0.0001). Similarly, compared to MHO patients, the postprandial insulin increased by 2-fold in patients with mildly abnormal fat (Q1 versus MHO; P = 0.08) to 7-fold in those with the worse adipose tissue IR (Q4 versus MHO; P < 0.0001).

Figure 1.

Fasting (white) and postprandial (black) plasma insulin (A) and FFA (B) concentrations during a 2-hour OGTT in lean and obese patients without NAFLD and in obese subjects with worsening adipose tissue IR (Q1-Q4). †P < 0.0001 versus all groups; *P < 0.01 to P < 0.05 versus lean and obese controls; ‡P < 0.05 versus obese controls; §P < 0.01 to P < 0.05 versus Q1 and Q2. Results are mean ± SEM.

A similar pattern emerged for plasma FFA concentration (Fig. 1B). Fasting FFA levels were comparable between lean and MHO patients (406 ± 47 versus 324 ± 25 μmol/L, respectively; P = 0.36). Despite increasing plasma insulin concentration, there was a progressive increase in fasting plasma FFA concentration from Q1 to Q4 (436 ± 28 μmol/L [36% increase] to 718 ± 29 μmol/L [220% increase]; Q4 versus MHO; P < 0.0001). Postprandial FFA suppression during the OGTT was only slightly and nonsignificantly lower (i.e., worse) in MHO versus lean subjects (84% ± 2% versus 74% ± 5%, respectively; nonsignificant). Consistent with the fasting state, resistance to insulin's inhibitory effect on lipolysis was also evident in the postprandial state in Q1-Q4, being only 62% ± 3% in Q4 versus 74% ± 5% in MHO patients (P < 0.0001).

Effect of adipose tissue IR on plasma aminotransferase concentration

Plasma AST (Fig. 2A) and ALT (Fig. 2B) were similar among lean and MHO patients, but significantly higher in patients with NAFLD. They rapidly increased in Q1 by ∼1.5- to 2.0-fold (P < 0.05 versus lean and MHO). The percentage of patients with normal (arbitrarily <40 IU/L) aminotransferases decreased with worsening adipose tissue IR. Though all lean and MHO patients had normal AST/ALT, patients with normal AST/ALT decreased from 81/47% in Q1 to 51/16% in Q4 (Q4; P < 0.0001 versus MHO).

Figure 2.

Plasma AST (A) and ALT (B) concentrations in lean and obese patients without NAFLD and in obese subjects with worsening adipose tissue IR (Q1-Q4). *P < 0.01 to P < 0.05 versus lean and obese controls; ‡P < 0.01 versus obese controls; §P < 0.01 versus Q1 and Q2; †P < 0.05 versus Q3. Results are mean ± SEM.

Effect of Adipo-IR on dyslipidemia

Lean and obese insulin-sensitive subjects had a similar plasma lipid profile (Table 1; Fig. 3). Dysfunctional adipose tissue had no effect on total cholesterol (Fig. 3A) or LDL-C (Fig. 3B). However, HDL-C (Fig. 3C) decreased significantly by 20% in Q1 versus MHO patients (P < 0.01) and was most pronounced at Q4: 34 ± 1 (P < 0.001 versus MHO). Plasma TG increased in a similar pattern, even with mild adipose tissue IR (Q1 versus MHO: 92 ± 10 versus 158 ± 19; P = 0.05), and paralleled the worsening of adipose tissue IR (P < 0.001 versus MHO).

Figure 3.

Plasma total cholesterol (A), LDL-C (B), HDL-C (C), and TG (D) in lean and obese patients without NAFLD and in obese subjects with worsening adipose tissue IR (Q1-Q4). *P < 0.01 versus lean and obese controls; §P < 0.01 to P < 0.05 versus Q1 and Q2. Results are mean ± SEM.

Effect of adipose tissue IR on hepatic and skeletal muscle insulin sensitivity

MHO versus lean subjects showed a trend for decreased liver (Fig. 4A) and muscle (Fig. 4B) insulin sensitivity, although this difference did not reach statistical significance (Table 1). There was ∼40%-50% worsening of HIRi between lean and MHO subjects versus Q1 and Q2 (P = 0.11), suggesting that hepatic IR develops even with a mild (Q1) to moderate (Q2) deterioration in adipose tissue insulin sensitivity (Fig. 4A). This was even more evident for Q3 and Q4, although liver fat remained constant (Q3) or was only slightly higher (Q4). As for skeletal muscle (Fig. 4B), there was an abrupt early-on decline in insulin action (Q1-Q3: −40%-50%; P < 0.001), with a further reduction to 62% in Q4 patients (P < 0.0001 versus MHO). There was a close relationship between adipose tissue, liver, and skeletal muscle IR. The liver had the strongest correlation with adipose tissue IR (r = 0.59, P < 0.0001; Fig. 5A), indicative of the deleterious effect of dysfunctional fat on hepatic metabolism. Skeletal muscle was also significantly affected (Fig. 5B), although the correlation was not as strong (r = 0.44, P < 0.0001).

Figure 4.

Liver (A) and skeletal muscle (B) insulin sensitivity in lean and obese patients without NAFLD and in obese subjects with worsening adipose tissue IR (Q1-Q4) *P < 0.01 to P < 0.05 versus lean and obese controls; §P < 0.001 versus Q1 and Q2; ‡P < 0.05 versus Q4; †P < 0.05 versus Q1-Q3. Results are mean ± SEM.

Figure 5.

Relationship between adipose tissue IR with liver and skeletal muscle IR.

Effect of adipose tissue IR on liver histology

Worsening adipose tissue IR was not associated with worsening hepatic steatosis (Q1: 2.1 ± 0.2; Q2: 1.8 ± 0.2; Q3: 2.1 ± 0.1; Q4: 2.1 ± 0.1; all nonsignificant), consistent with the nonsignificant increase in liver fat by MRS (Table 2). Similarly, necroinflammation was also present, but not different, between Q1 versus Q4, even as patients had more dysfunctional fat (Q1: 2.4 ± 0.2; Q2: 2.8 ± 0.2; Q3: 2.8 ± 0.1; Q4: 2.8 ± 0.1; all nonsignificant). The NAFLD activity score (NAS) was similar across Q1-Q4 groups (Fig. 6A). In contrast, adipose tissue IR played an important role on the severity of liver fibrosis, as suggested when comparing Q3 and Q4 versus Q1 and Q2 (Fig. 6B; P < 0.05). A fibrosis stage 2 or 3 was present in 18% of subjects in Q1 (3 of 17) and Q2 (5 of 29), compared to stages 2-4 occurring in 30% of Q3 and Q4 patients (P < 0.05).

Figure 6.

NAS (A) and fibrosis (B) in lean and obese patients without NAFLD and in obese subjects with worsening adipose tissue IR (Q1-Q4).


The aim of the present study was to understand the role of dysfunctional adipose tissue on metabolic and histological parameters of obese patients with NAFLD. To this end, we performed in each patient an in-depth metabolic assessment coupled with a liver biopsy. This approach allowed an integrated metabolic and histological evaluation of the liver in relation to adipose tissue in NAFLD and led to the following important clinical findings: (1) MHO subjects with normal insulin-sensitive adipose tissue do not usually develop hepatic steatosis and have a near-normal metabolic profile; (2) there is a low threshold for the metabolic effects of dysfunctional adipose tissue. Even modest adipose tissue IR rapidly leads to an elevation of liver aminotransferases, dyslipidemia (i.e., high TG/low HDL-C), reduction in plasma adiponectin, marked liver and muscle IR, hepatic steatosis and NASH; and (3) from an histological perspective, adipose tissue IR triggers the development of hepatic lipotoxicity in NASH (also with a rather low threshold), but appears to play less of a role in determining the severity of necroinflammation. In contrast, fibrosis is susceptible to the severity of adipose tissue IR. Taken together, these observations have significant clinical implications to the prevention and treatment of patients with NAFLD.

There were major differences in the severity of adipose tissue dysfunction in obese subjects with and without NAFLD for similar degrees of adiposity (i.e., similar BMI and whole body fat). Plasma FFA levels were much higher in patients with NAFLD, despite higher insulin levels, which is indicative of a severe defect in the suppression of plasma FFA by insulin (Table 1; Fig. 1). This observation should shift our focus about the metabolic effect of obesity in NAFLD from the severity of adiposity to the magnitude of adipose tissue dysfunction (i.e., from quantity to quality), a concept explored previously in the fields of cardiovascular risk assessment21-23 and type 2 diabetes mellitus (T2DM),24 but never carefully examined in NASH. It also implies that additional factors, either genetic or acquired, may play a greater role than previously appreciated in obese subjects with NAFLD. Several hypotheses have been proposed to explain the etiology of adipose tissue dysfunction in obesity.25-30 A genetic link to adipose tissue IR is suggested by the observation that nonobese subjects with a strong family history of T2DM already have early defects in adipose tissue function,25, 31 although these studies have not focused on the effect of adipose tissue on hepatic steatosis.

Although MHO subjects had a much worse BMI, their metabolic profile was similar to that of lean insulin-sensitive subjects. However, it was not completely normal because there was already a trend toward worsening hepatic insulin sensitivity (Table 1) and a significant reduction in plasma adiponectin, insulin suppression of plasma FFA, and established muscle insulin resistance (Fig. 4B). Nevertheless, this reduction was not as severe as in Q1. Patients in Q1 already had significant signs of metabolic distress with higher AST/ALT (Fig. 2), dyslipidemia (i.e., high TG/low HDL-C) (Fig. 3), liver and muscle IR (Fig. 4), hepatic steatosis (Table 2) and NASH (Fig. 6). Of note, visceral fat was not different across quartiles and failed to explain the metabolic and histological differences. This is consistent with recent work suggesting that hepatic fat is more closely associated with the metabolic abnormalities in NAFLD than visceral fat.32 Though the metabolic disturbances described here cannot be entirely ascribed to dysfunctional adipose tissue, their strong association with dysfunctional fat suggests an important role in the pathogenesis of metabolic/histological defects in NAFLD. It also suggests that lipotoxicity has a low threshold in NAFLD and that its impact varies among target tissues. Skeletal muscle appeared rapidly affected by dysfunctional adipose tissue (Q1-Q3), whereas it was more gradual at the level of the liver (Fig. 4). However, at the extreme of adipose tissue IR (Q4), all metabolic variables (i.e., AST/ALT, TG/HDL-C, and hepatic/muscle IR) further deteriorated, suggesting that target tissues continue to be affected and susceptible to worsening lipotoxicity. This has clinical implications for lipotoxicity in the development and treatment of steatohepatitis and fibrosis. The lack of an association between an exacerbation of adipose tissue IR and steatohepatitis (Fig. 6) does not mean that, upon reversal of adipose tissue IR with a TZD, there cannot be a marked improvement in steatohepatitis, as previously reported.9 Indeed, the low threshold for steatohepatitis (already observed in Q1) would suggest that even modest reversal of adipose tissue IR may be beneficial in NASH. In our hands, reversal of adipose tissue IR by a TZD had the closest correlation with necroinflammation (r = 0.47, P < 0.01), but also was associated with changes in steatosis (r = 0.29; P = 0.049) and, to a lesser degree, fibrosis (0.25; r = 0.09). It should be kept in mind that changes in adipose tissue IR is only one aspect by which TZDs may improve histology in NASH, and that many other systemic/local mechanisms (associated with changes in adipose tissue IR) are likely to play a role.

Although dysfunctional fat clearly predisposed to hepatic steatosis (Table 1) and necroinflammation (Fig. 6), contrary to what was expected, there was no additional effect of worsening adipose tissue IR on the NAS (Fig. 6). This would be consistent with a low threshold for FFA to trigger lipotoxicity and steatohepatitis, but also that other factors determine the severity of NASH.11, 12, 33, 34 Once FFA triggers intracellular inflammatory pathways, it appears that steatohepatitis would depend less on the magnitude of the FFA/lipotoxicity insult than on other local factors. In contrast, liver fibrosis did show a susceptibility to more severe adipose tissue IR. Because fibrosis is strongly associated with the activation of hepatic stellate cells (HSCs),35 it is possible that the susceptibility to lipotoxicity may be different for hepatocytes, compared to HSCs. Studies in vitro indicate that HSCs are very sensitive to exposure to palmitate and other long-chain fatty acids.36, 37 This has two major clinical implications. First, adipose tissue IR may be an overlooked aspect regarding future risk for cirrhosis. Though obesity is an established risk factor for NASH progression10-13, 38 and cirrhosis,39 no previous studies have directly investigated the role of adipose tissue IR in relation to the natural history of the disease. Second, it may offer a novel target for disease prevention. Adiponectin is important in the regulation of HSC function.40-42 Because plasma adiponectin is decreased in NASH,43 modulation of its levels by peroxisome proliferator-activated receptor gamma agonists44 or by newer, more potent pharmacological agents may reverse fibrogenesis in this population.

A practical aspect of the study is the possible value of a simple index of adipose tissue IR (Adipo-IRi) to establish more accurately the metabolic effect of obesity in patients with NAFLD. The Adipo-IRi is derived from the plasma FFA x insulin concentration, and both measurements are quite simple, widely available, and rather inexpensive. Traditionally, BMI has been used as an indicator of metabolic risk in NAFLD.45, 46 Given the known limitations of BMI measurements,46, 47 we believed that a direct measure of adiposity, such as whole body fat by DXA, would be a more precise, useful guide of metabolic risk. However, neither was particularly helpful to assess metabolic risk associated with obesity. The Adipo-IRi has been proven useful in studying IR in patients with T2DM4 and the response to pioglitazone in patients with NASH.8 Abnormal Adipo-IRi was consistent with an impaired suppression of plasma FFA by insulin (Table 1) and a low plasma adiponectin concentration, which are all indicative of severe adipose tissue dysregulation. Future work will determine the clinical value of the Adipo-IRi in the management of patients with NAFLD.

In summary, the current work expands our understanding about the role of dysfunctional adipose tissue on metabolic and histological parameters of patients with NAFLD. Liver steatosis is rare in MHO subjects with normal insulin-sensitive adipose tissue, highlighting the important role of lipotoxicity in NAFLD. There is a low threshold for the metabolic effects of dysfunctional adipose tissue, including elevation of liver aminotransferases, IR, and steatohepatitis. A similar low threshold for adipose tissue IR triggers the development of NASH in susceptible patients, but appears to be less of a factor in determining the severity of necroinflammation. In contrast, fibrosis may worsen in proportion to the severity of dysfunctional fat. We believe that these findings are a first step toward a better understanding of the role of obesity in NASH and may increase awareness about dysfunctional adipose tissue as a potential target for intervention in these patients.


The authors thank all the volunteers, the Clinical Translational Science Award nursing staff (in particular, Rose Kaminski-Graham and Norma Diaz), and the laboratory and nutritional staff for their assistance in performing the above-described studies.