Nonalcoholic fatty liver disease risk and histologic severity are associated with genetic polymorphisms in children

Abstract

Background and aims: NAFLD is the most common chronic liver disease in children. Large pediatric studies identifying single nucleotide polymorphisms (SNPs) associated with risk and histologic severity of NAFLD are limited. Study aims included investigating SNPs associated with risk for NAFLD using family trios and association of candidate alleles with histologic severity.

Approach and results: Children with biopsy-confirmed NAFLD were enrolled from the NASH Clinical Research Network. The Expert Pathology Committee reviewed liver histology. Genotyping was conducted with allele-specific primers for 60 candidate SNPs. Parents were enrolled for trio analysis. To assess risk for NAFLD, the transmission disequilibrium test was conducted in trios. Among cases, regression analysis assessed associations with histologic severity. A total of 822 children with NAFLD had mean age 13.2 years (SD 2.7) and mean ALT 101 U/L (SD 90). PNPLA3 (rs738409) demonstrated the strongest risk ( p = 2.24 × 10 -14 ) for NAFLD. Among children with NAFLD, stratifying by PNPLA3 s738409 genotype, the variant genotype associated with steatosis ( p = 0.005), lobular ( p = 0.03) and portal inflammation ( p = 0.002). Steatosis grade associated with TM6SF2 ( p = 0.0009), GCKR ( p = 0.0032), PNPLA3 rs738409 ( p = 0.0053), and MTTP ( p = 0.0051). Fibrosis stage associated with PARVB rs6006473 ( p = 0.0001), NR1I2 ( p = 0.0021), ADIPOR2 ( p = 0.0038), and OXTR ( p = 0.0065). PNPLA3 rs738409 ( p = 0.0002) associated with borderline zone 1 NASH.

Conclusions: This study demonstrated disease-associated SNPs in children with NAFLD. In particular, rs6006473 was highly associated with severity of fibrosis. These hypothesis-generating results support future mechanistic studies of development of adverse outcomes such as fibrosis and generation of therapeutic targets for NAFLD in children.

FIGURE 1

Summary of results from regression analysis. (A) Heatmap shows where there are nominally significant associations between response variables and SNPs of interest (non-white squares). (B) Heatmap shows the absolute value of effect sizes for nominally significant associations (non-white squares). Regression models adjust for Age, Sex, Height, Weight, and Ethnicity. The results from each response variable tested are shown in the heatmap columns, whereas the results from each SNP tested are shown in the rows. Rows and columns are ordered by hierarchical clustering. (C,D) Results from the lasso regression on nominally significant associations, showing relative feature importance, for fibrosis (C), and steatosis (D).

FIGURE 2

Summary of results from regression analysis. (A) Heatmap shows where there are nominally significant associations between response variables and SNPs of interest (non-white squares). (B) Heatmap shows the absolute value of effect sizes for nominally significant associations (non-white squares). Regression models adjust for Age, Sex, Height, Weight, and Ethnicity. The results from each response variable tested are shown in the heatmap columns, whereas the results from each SNP tested are shown in the rows. Rows and columns are ordered by hierarchical clustering. (C–E) Results from the lasso regression on nominally significant associations, showing relative feature importance, for borderline zone 3 NASH (C), borderline zone 1 NASH (D), and definite NASH (E).

FIGURE 3

A multi-SNP predictive model constructed using all SNPs that were nominally significant in the fibrosis model. The model discriminates cases with absent-to-mild fibrosis from those with moderate-to-severe fibrosis. All nominally significant individual SNPs were included along with ALT and genetics and control variables of age, sex, height, weight, and ethnicity. The mean ± 1 SD are shown for the area under the curve.