Multiomics study of nonalcoholic fatty liver disease

Abstract

Nonalcoholic fatty liver (NAFL) and its sequelae are growing health problems. We performed a genome-wide association study of NAFL, cirrhosis and hepatocellular carcinoma, and integrated the findings with expression and proteomic data. For NAFL, we utilized 9,491 clinical cases and proton density fat fraction extracted from 36,116 liver magnetic resonance images. We identified 18 sequence variants associated with NAFL and 4 with cirrhosis, and found rare, protective, predicted loss-of-function variants in MTARC1 and GPAM, underscoring them as potential drug targets. We leveraged messenger RNA expression, splicing and predicted coding effects to identify 16 putative causal genes, of which many are implicated in lipid metabolism. We analyzed levels of 4,907 plasma proteins in 35,559 Icelanders and 1,459 proteins in 47,151 UK Biobank participants, identifying multiple proteins involved in disease pathogenesis. We show that proteomics can discriminate between NAFL and cirrhosis. The present study provides insights into the development of noninvasive evaluation of NAFL and new therapeutic options.

Fig. 1. The distribution of PDFF measurements.

a, Histogram of PDFF in the IDEAL cohort. b, Histogram of PDFF in the GRE cohort. c, PDFF plotted against BMI in the UKB IDEAL cohort (n = 27,668). d, PDFF plotted against BMI in the UKB GRE cohort (n = 8,448). The vertical line is at BMI = 25 kg m⁻². The horizontal line is at PDFF = 5%.

Fig. 2. Effects of sequence variants on PDFF compared with effects on liver disease, liver enzymes and lipids.

(a–i) Effects of sequence variants on PDFF (n = 36,116) compared with their effects on ICD-10 code-diagnosed NAFL (n_cases = 5,921) (a), cirrhosis (n_cases = 2,301) (b), HCC (n_cases = 374) (c) and measures of AST (n = 389,272) (d), ALT (n = 390,519) (e), γ-glutamyl transferase (n = 390,457) (f), total cholesterol (n = 390,652) (g), triglycerides (n = 390,346) (h) and low-density lipoprotein (n = 389,974) (i) in the UKB. The effects (box of error bars) and their 95% confidence intervals (CIs) (error bars) are shown for the allele that increases PDFF and either in s.d. for quantitative traits or as log(OR) for binary traits.

Fig. 3. Pleiotropic effects of NAFLD variants.

The effects of the identified variants with 51 phenotypes, including liver enzymes and lipid levels. The effect (colored from red to blue) on each phenotype is shown for the allele that increases PDFF and risk of NAFL or cirrhosis and is scaled to the range of [−1:1] for binary and quantitative phenotypes separately. Effects are shown only for the significant associations after an FDR correction. Apo, Apolipoprotein; HbA1c, glycated hemoglobin.

Fig. 4. Association pattern of missense and loss-of-function variants in GPAM and MTARC1.

(a,b) Plots show a similar association pattern of common missense variants in MTARC1 (a) and GPAM (b) with selected phenotypes. On the x axes, the effects of the minor alleles (box of the error bars) and their corresponding 95% CIs (error bars) are shown (log(OR) for binary traits). (c,d) Plots show the association pattern of rare predicted loss-of-function variants identified in the Icelandic population with the corresponding phenotypes in Iceland for MTARC1 (c) and GPAM (d). The OR of p.Thr189GlyfsTer5 for cirrhosis is 0, which is outside the plotting region. The n measures are shown for quantitative traits and n cases for binary traits. The data used to generate these plots are presented in Supplementary Tables 4 and 5.

Fig. 5. ROC for models trained to discriminate between NAFL and cirrhosis.

a–f, ROC AUCs for models trained to discriminate between NAFL and population (a,b), cirrhosis and population (c,d), NAFL and cirrhosis (e,f). a,c,e Plots show results for the SomaScan data using the Icelandic population. b,d,f Plots show results for the Olink data using the UKB population. The SomaScan analysis was performed on 181 individuals with NAFL and 73 with cirrhosis, and the Olink analysis was performed on 610 individuals with NAFLD and 262 with cirrhosis. Models trained to discriminate between the presence of disease diagnosis and population were trained, respectively, on an additional set of 20,619 individuals (SomaScan) and 38,018 individuals (Olink) without cirrhosis and NAFL. FPR, false-positive rate; TPR, true-positive rate.