The Association Between Hepatic Fat Content and Liver Injury in Obese Children and Adolescents: Effects of ethnicity, insulin resistance, and common gene variants

In an effort to understand the role of NAFLD/NASH in the pathophysiology of youth-onset type 2 diabetes, we began in 2008 to form a multiethnic cohort of obese children and adolescents. As of the time of writing this article, the cohort consists of 229 obese children and adolescents (87 Caucasians, 61 African Americans, and 81 Hispanics; mean age, 12.8 ± 2.9 years; mean BMI, 31.4 ± 7.4) from the New Haven area (New Haven, CT) recruited through the Yale Pediatric Obesity Clinic. Caucasians (15.1 ± 4.0 years) and African Americans (15.1 ± 3.3 years) tended to be older than Hispanics (13.5 ± 2.7 years; P = 0.005), whereas the BMI was similar among ethnicities (P = 0.16). Seventeen Caucasians (13 girls), 17 African Americans (13 girls), and 30 Hispanics (14 girls) showed impaired glucose tolerance, whereas 1 Caucasian (girl), 1 African American (girl), and 2 Hispanics (1 girl) showed type 2 diabetes (P = 0.15). The cohort was carefully phenotyped with respect to quantification of hepatic fat content and abdominal fat distribution using MRI, systemic biomarkers of apoptosis such as noninvasive indicators of NASH, fasting lipid and lipoprotein profiles, glucose homeostasis, and genetic markers of NAFLD and NASH.

The study subjects underwent metabolic and imaging studies, biochemical analyses, and genotyping. Detailed information about these studies is provided as Supplementary Material.

The study was approved by the Yale University Human Investigation Committee. Written parental informed consent and written child assent were obtained from all participants.

Liver biopsy. The liver biopsy was performed in 12 subjects (4 girls) because of persistent elevation in alanine transaminase (ALT) (mean ALT, 133.0 ± 64.4). Biopsy specimens were formalin-fixed, paraffin-embedded, stained with hematoxylin and eosin and trichrome, and underwent Gordon reticulin techniques. All biopsy samples were 2 cm or more in length and were reviewed by a pediatric pathologist according to the Brunt approach (19). The specimens were analyzed and steatosis, ballooning, inflammation, and fibrosis were scored according to Kleiner et al. (20). The NAFLD activity score was calculated by adding the scores of steatosis, inflammation, and ballooning, whereas the stage of fibrosis was determined by using a 4-point scale (20).

Statistics. Before analyzing the data, all the variables were tested for normality, with non-normally distributed variables log-transformed to be better approximated by normality, except for hepatic fat content (HFF%), for which a square-root transformation was used. All continuous variables were compared among the groups using the ANOVA. Adjusted comparisons were performed using a general linear model, adjusting for age, sex, and percent of total body fat. Prevalence among groups was compared using the χ² statistic. A Pearson correlation was used to test correlations between CK-18 and HFF%, whole body insulin sensitivity index (WBISI), or adiponectin in each ethnic group. A Spearman correlation was used to test the correlation between the CK-18 and the NAFLD activity score and liver fibrosis. To evaluate the interaction between ethnicity and HFF% or WBISI, a general linear model including the single terms and an interaction term (e.g., HFF% × ethnicity or WBISI × ethnicity) was performed. To evaluate the interaction between ethnicity and HFF%, age, sex, total body fat, and WBISI were used as covariates; to evaluate the interaction between ethnicity and WBISI, the HFF% was included in the model.

Within each ethnic group, the association between the genotypes and quantitative traits was evaluated by coding the genotype with an additive model of inheritance, i.e., the genotype is coded with 0, 1, or 2, corresponding to the number of minor alleles carried by each individual; age, sex, and total body fat were used as covariates when appropriate. The partial correlation coefficients (r²) were used to evaluate the degree of variance of CK-18 explained by the genotype. The χ² test was used to assess whether the genotypes were in Hardy-Weinberg equilibrium and to test differences in genotype distribution among different ethnic groups. Unless otherwise specified, for all the data raw means and SD are shown.

Twelve subjects (2 Caucasians and 10 Hispanics) had NASH proven by liver biopsy. When compared with the entire cohort, these subjects showed higher HFF% (P = 0.0008), higher CK-18 (P < 0.0001), ALT (P < 0.0001), and triglyceride levels (P = 0.02), and lower WBISI (P = 0.02). Consistent with previous reports (12) in this subgroup, the CK-18 showed a strong correlation with the NAFLD activity score (r = 0.70; P = 0.01) and fibrosis (r = 0.68; P = 0.03).

Characteristics of the studied population irrespective of degree of HFF% and presence of NASH are shown in Supplementary Table 1. Hispanics tended to be younger than African Americans and Caucasians (P = 0.006), but there was no difference among ethnicities in terms of sex prevalence, BMI, total body fat, and glucose tolerance. Whereas Caucasians and African Americans did not differ in terms of insulin resistance (P = 0.44), Hispanics showed a WBISI lower than the other two ethnic groups, independent of age, sex, and total body fat (P = 0.001). As expected, African Americans showed lower hepatic fat content than Caucasians and Hispanics, independent of age, sex, total body fat (P < 0.001), and visceral fat (P = 0.001); similarly, they also showed lower triglycerides (P = 0.0002) and ALT (P = 0.05) independent of age, sex, total body fat, and HFF%. Of particular note, CK-18 levels were different between Caucasians and African Americans (P = 0.004) and between Hispanics and African Americans (P < 0.0001), but not between Caucasians and Hispanics, although Hispanics tended to show higher levels than Caucasians (P = 0.09).

Clinical and laboratory characteristics of patients according to presence of steatosis and NASH are shown in Table 1. Among both Caucasians and Hispanics, subjects with biopsy-proven NASH showed higher BMI (P = 0.03 and P < 0.001, respectively), total body fat (P = 0.001 and P < 0.001, respectively), higher HFF% (both P < 0.001), and ALT levels (both P < 0.001) than subjects with and without steatosis (Table 1). CK-18 levels were higher in subjects with biopsy-proven NASH in both Caucasians and Hispanics (Fig. 1); furthermore, in Caucasians and Hispanics, subjects with and without hepatic steatosis differed for the CK-18 levels of ∼20% (both P = 0.04), whereas in African Americans the CK-18 levels were similar between subjects with and without hepatic steatosis (P = 0.50) (Fig. 1).

To answer our first question and to assess whether the relationship between CK-18 and HFF% is modulated by ethnicity, we performed a regression model including an interaction term between ethnicity and HFF% as an independent variable and CK-18 as the dependent variable. We observed an interaction between ethnicity and HFF% (P = 0.022) that remained significant even after adjusting for age, sex, WBISI, and total body fat (P = 0.011). In fact, although the CK-18 was associated with HFF% in Caucasians and Hispanics, there was no association in African Americans (Fig. 2A). CK-18 levels were directly correlated with HFF% in Caucasians (r = 0.43; P < 0.0001) and Hispanics (r = 0.34; P = 0.0018), but not in African Americans (r = −0.08; P = 0.50).

We also asked whether the degree of insulin resistance might be associated to the degree of liver damage. To answer this question we performed a regression model including an interaction term between ethnicity and WIBISI as an independent variable and CK-18 as a dependent variable. Interestingly, also in this case, we observed an interaction between ethnicity and WBISI (P = 0.004), independent of age, sex, total body fat (P = 0.018), and HFF% (P = 0.002) (Fig. 2B). CK-18 levels, in fact, were inversely correlated with WBISI in Caucasians (r = −0.40; P < 0.0001) and Hispanics (r = −0.24; P = 0.024), but not in African Americans (r = 0.08; P = 0.52). Similarly, we observed the same trend for an interaction between ethnicity and adiponectin (P = 0.05), independent of age, sex, total body fat (P = 0.03), and HFF% (P = 0.06) (Fig. 2C). CK-18 levels were inversely correlated with adiponectin in Caucasians (r = −0.26; P = 0.01) and Hispanics (r = −0.27; P = 0.01), whereas there was no correlation in African Americans (r = 0.12; P = 0.35).

The third aim of this study was to explore whether three gene variants previously associated with HFF% (the rs1260326 in the GCKR and the rs738409 in the PNPLA3) (15–17) or with NAFLD activity score (the rs2645424 in the FDFT1) (18) might be associated with the degree of liver damage in obese youths as determined by a noninvasive biomarker.

The GCKR SNP rs1260326 minor allele (T) frequency was 0.366 in Caucasians, 0.152 in African Americans, and 0.390 in Hispanics. The frequency of the PNPLA3 rs738409 minor allele (G) was 0.305 in Caucasians, 0.186 in African Americans, and 0.460 in Hispanics. The frequency of the FDFT1 rs2645424 A allele was 0.458 in Caucasians, 0.466 in African Americans, and 0.350 in Hispanics. Within each ethnic group there was no evidence against the null hypothesis that the genotype distribution was in Hardy-Weinberg equilibrium for all of the variants (all P > 0.05).

Patient features data according to the genotype are provided in Supplementary Table 1 (Caucasians), Supplementary Table 2 (African Americans), and Supplementary Table 3 (Hispanics). Whereas the FDFT1 rs2645424 was not associated with the HFF% (Supplementary Tables 1, 2, and 3), we observed an association between the FDFT1 rs2645424 and CK-18 in Caucasians and Hispanics, independent of age, sex, and total body fat (P = 0.05 and P = 0.02, respectively); although not statistically significant, the same trend was observed in African Americans (P = 0.07) (Fig. 3A). The association between FDFT1 rs2645424 and CK-18 was independent of HFF% (Caucasians, P = 0.01; Hispanics, P = 0.02). We also observed an association between the CK-18 and PNPLA3 rs738409 in Caucasians and Hispanics, independent of age, sex, and total body fat (P = 0.02 and P = 0.05, respectively), whereas we were not able to detect the association in African Americans (P = 0.26) (Fig. 3B). This association was independent of HFF% in Caucasians (P = 0.01) and Hispanics (P = 0.05). The GCKR rs1260326 was not associated with the CK-18 levels in any ethnicities (Fig. 3C).

The main finding of this study relates to the relationship between hepatic fat accumulation and liver injury in obese youth and the factors that influence this relationship. Our results demonstrate that the association between hepatic fat content as measured by fast MRI and liver injury as determined by levels of CK-18 fragments is strongly affected by ethnicity. In particular, we observed that for the same degree of HFF%, African American obese youths show lower levels of CK-18 than Caucasians and Hispanics, meaning that African Americans tend to experience a lower degree of hepatic damage irrespective of the degree of hepatic fat accumulation.

In addition, we found that the CK-18 levels were associated with insulin resistance in Caucasians and Hispanics, but not in African Americans, meaning that in this latter population the effect of insulin resistance in the development of NASH is probably marginal.

Although previously some authors have attributed the lower propensity to development of NAFLD or NASH in African Americans to the lower degree of insulin resistance (21), a report from the Dallas Heart Study clearly has excluded the different rate of insulin resistance as the major cause for the lower rate of NAFLD in African Americans and suggested that such a diversity is more likely to reflect biological and genetic differences in lipid metabolism rather than differences in insulin resistance, obesity, or alcohol intake (6,7). In agreement with these studies, our observations suggest that obese African American youth tend to show lower rates of hepatic fat accretion and a lower degree of hepatic fat damage, independent of the degree of insulin resistance. In fact, whereas insulin resistance is associated with liver damage in Caucasians and Hispanics, this association was lacking in African Americans, meaning that although insulin resistance may play a role in the progression of NAFLD in Caucasians and Hispanics, it may not be associated with the progression of the disease in African Americans.

Thus, two key questions arise. Why is there no association between hepatic steatosis or insulin resistance and liver damage in African Americans? What protects African Americans from development of NASH, even in presence of severe obesity and insulin resistance?

Whereas the pathogenetic mechanisms responsible for the progression from simple steatosis to NASH are largely still unknown, it has been clearly shown that the excessive release of free fatty acids plays a key role in the development of hepatic “lipotoxicity” in NAFLD (22–24). Because the first source of free fatty acids for hepatic triglycerides synthesis (∼70%) is the adipose tissue (25), and because African Americans tend to show low plasma triglycerides and hepatic triglycerides content even in presence of severe obesity and insulin resistance, the possibility exists that they may physiologically have a lower release of free fatty acids from adipose tissue. This hypothesis would explain both the lower rate of NAFLD as well as the lower propensity for development of NASH.

It also can be hypothesized that in African Americans, adiponectin still can play a protective role against liver injury even in presence of hepatic fat accumulation; in fact, whereas in Caucasians and Hispanics low levels of adiponectin were associated with an increase of CK-18 levels, we did not observe any association between adiponectin levels and CK-18 in African Americans. These data would be consistent with recent animal and human studies showing that adiponectin may have a hepato-protective effect (10,26,27).

Nevertheless, it is likely that these differences are genetically driven and that different variants in genes involved in several pathways (such as lipid metabolism in the liver and adipose cells, inflammation, and others) might be responsible for it.

In an effort to unravel some of the genetic determinants responsible for the hepatic damage in obese children and adolescents, we studied three gene variants, the rs738409 in the PNPLA3 gene and the rs1260326 in the GCKR gene, both previously associated with hepatic steatosis (15–17), and the rs2645424 in the FDFT1, previously associated with the NAFLD activity score (18).

Here, we show that the rs1260326 variant in the GCKR is not associated with CK-18 levels. We previously have shown that it is associated with hepatic fat content, triglycerides levels, and large VLDL levels (17), and we have suggested, along with other investigators (28), that the mechanism by which it leads to hepatic steatosis is via an increased rate of hepatic de novo lipogenesis. Nevertheless, this variant seems to predispose to development of intrahepatic fat accumulation, without contributing to the progression of NAFLD.

We observed that the rs738409 in the PNPLA3 gene is associated with CK-18; this observation is consistent with previous reports showing that the PNPLA3 rs738409 variant plays a role not only in predisposing to liver fat accumulation but also in the progression to NASH (29,30). Recent studies have shown that the rs738409 variant in the PNPLA3 leads to hepatic steatosis and steatohepatitis by enhancing the lipogenic activity and by impairing the lipolytic activity of the PNPLA3 in the liver (31,32).

Here, we also show for the first time an association between the rs2645424 in the FDFT1 gene and CK-18 levels in our multiethnic cohort of obese youths. This variant, in fact, has been found to be associated with NAFLD activity score in a cohort of adult women by a genome-wide association study performed in sample of 236 non-Hispanic white women with NAFLD (15). The FDFT1 gene, located on chromosome 8, is a key regulator of cholesterol biosynthesis (33,34). It encodes the squalene synthase, an enzyme involved in sterol synthesis; in particular, it converts two molecules of farnesyl pyrophosphate into squalene, which is a precursor to cholesterol. Because the rs2645424 is an intronic variant, it is difficult to speculate on how it may affect the enzyme activity. It is possible that this SNP is in linkage disequilibrium with a variant in the promoter so that enhancing its expression leads to an increased activation of the squalene synthase and to the intrahepatic accumulation of cholesterol. Animal studies have, in fact, shown that transient overexpression of the FDFT1 gene in the liver of both wild-type and LDLR knockout mice resulted in increased de novo cholesterol biosynthesis, oversecretion of cholesterol-rich LDL, higher cholesterol levels, and a 37% increase in liver weight compared with controls attributable to hepatocyte proliferation (35). This hypothesis also would be consistent with recent studies showing the role of intrahepatic cholesterol accumulation in the pathogenesis of NASH (36).

Interestingly, none of the studied SNPs was associated with insulin resistance (Supplementary Tables 1, 2, and 3). Dissociation between insulin resistance and fatty liver in subjects carrying the PNPLA3 rs738409 variant ahs been previously observed (16,37). In particular, as first suggested by Romeo et al. (16), the dissociation between fatty liver and insulin resistance in subjects carrying the rs738409 risk allele was demonstrated by Stefan et al. (37) using a hyperinsulinemic–euglycemic clamp. The authors suggested that the PNPLA3 variant might be involved in the generation of a metabolically benign fatty liver (37). Taken together with our findings, those data corroborate the hypothesis that the PNPLA3 rs738409 may contribute to the dissociation between NAFLD and insulin resistance observed in some individuals (38).

We did not observe any association between insulin resistance and FDFT1 variant; although this may suggest that this gene variant predisposes to liver damage independent of insulin resistance, studies using state-of-the-art techniques are needed to highlight this point.

This study has several strengths, such as the following: the young age of the patients; the absence of risk factors linked to alcohol consumption and aging; the use of MRI measurement to assess hepatic fat content; and the use of a strongly validated marker of NASH (CK-18).

We acknowledge that the major limitation of this study is the lack of liver biopsies performed in the entire cohort of subjects. Although this is a weakness, it should be pointed out that these studies can be performed only in cohorts showing a wide spectrum of the disease, and thus only enrolling subjects with and without NASH, and that liver biopsy performed in healthy children is unethical. We also acknowledge that CK-18 might not be specific for NASH because it can be elevated in a number of other conditions and thus sometimes may give spurious results (39). In fact, it has been shown that one possible limitation of the use of CK-18 for the prediction of NASH is its intrinsic inability to discriminate between NASH and other chronic diseases that involve apoptosis, such as cholangitis and cholestasis, chronic hepatitis, cancer, and trauma (39).

Another limitation of the study is the relatively small sample size of the overall population. Although this is not a large sample, to our best knowledge, so far this is the largest study dealing with liver injury in a multiethnic cohort of obese children and adolescents carefully phenotyped for liver (e.g., liver fat accumulation, CK-18) and metabolic parameters (oral glucose tolerance test–derived measures).

This study provides the evidence that African American obese children and adolescents show a lower degree of liver damage than Caucasians and Hispanics, independent of the degree of hepatic fat accumulation and insulin resistance. These data suggest that African Americans are protected from hepatic damage even in presence of a high degree of hepatic fat accumulation and insulin resistance. Our observation of a different interethnic propensity to development of liver injury is clinically relevant. This suggests that Hispanic and Caucasian obese children and adolescents require closer follow-up of their liver metabolic status. Also, the observation that genes involved in the lipid metabolism (de novo lipogenesis for PNPLA3, GCKR, and cholesterol synthesis for FDFT1) are associated with liver injury provide indirect evidence that common gene variants in genes involved in this pathway may play a major role in the pathogenesis of NAFLD and NASH, and that studying functionally relevant gene variants of genes modulating lipid metabolism may provide more insights in the pathophysiology and the treatment of fatty liver disease in obese children and adolescents.

This work was supported by the American Heart Association (AHA) (11CRP5620013 to N.S.) and the National Institutes of Health (NIH) (grants R01-HD-40787, R01-HD-28016, and K24-HD-01464 to S.C.; grants DK076852 and DK082451 to A.E.F.; Yale Center for Clinical Investigation scholar award to N.S.). This publication also was made possible by Clinical and Translational Science Award grant UL1 RR024139 from the National Center for Advancing Translational Science (NCATS), a component of the National NIH, and NIH roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NIH.

No potential conflicts of interest relevant to this article were reported.

N.S. analyzed the data and wrote the manuscript. A.E.F., E.E., B.P., R.K., and G.K. researched the data. S.C. reviewed the data and wrote and edited the manuscript. S.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors are grateful to the patients and their families, as well as to the Yale Center for Genome Analyses (YCGA) and Yale Center for Clinical Investigation and Hospital Research Unit (HRU) personnel.