Abstract

Background: The patatin-like phospholipase domain-containing protein 3 (PNPLA3) gene is strongly implicated in the development of metabolic dysfunction-associated steatohepatitis (MASH). Gene silencing approaches are being considered for treatment of MASH but there are limited blood-based biomarkers of PNPLA3 homozygosity. There’s an unmet need to develop a serum-based biomarker panel that is specific for PNPLA3 homozygosity and could then be used to assess the pharmacodynamics response to PNPLA3 targeted therapies.

Methods:

We examined an extensive lipidomic and metabolomic dataset derived from the UK Biobank (Access number 71300). First, we selected the 4018 homozygous carriers of PNPLA3 rs738409_G. We then performed a 1:1 propensity score matching of non-carriers based on age, sex and BMI. Liver fat content was evaluated by proton density fat fraction on MRI in a subset of UKB participants (n=323 homozygous and 347 non-carriers). The resulting dataset was split into an 85% training set and a 15% test set. After benchmarking, a random forest machine learning algorithm was chosen as the predictive model,

Results: The mean liver fat content of PNPLA3 homozygotes versus non-carriers was 8.1% versus 4.3%. The PNPLA3 rs738409_G homozygotes exhibited a distinct metabolic profile compared to non-carriers (Figure 1). In the training cohort, a panel consisting of 9 metabolites and 1 routine hepatic screening parameter including saturated fatty acids to total fatty acids percentage, phospholipids to total lipids in IDL percentage, glutamine, glycine, phospholipids to total lipids in chylomicrons and extremely large VLDL percentage, average diameter for LDL particles, tyrosine, phospholipids to total lipids in very small VLDL percentage, cholesteryl esters to total lipids in small LDL percentage and alanine aminotransferase (ALT) was able to differentiate between PNPLA3 homozygotes versus non-carriers with an AUROC of .90 (95%CI: 0.89-0.90, p-value < 0.001), while ALT alone only received an AUROC of .71 (95%CI: 0.71-0.71, p-value < 0.001). The results remained consistent in the validation cohort with a statistically significant AUROC of .63 (95%CI: 0.63-0.63, p-value < 0.001). Among these top 9 differentially expressed metabolites, we examined which increased with liver fat content among participants with PNPLA3 homozygosity and observed that phospholipids to total lipids in IDL percentage, glutamine, average diameter for LDL particles and glycine decreased as the liver fat content increased, while all others increased.

Conclusion: Here, we demonstrate a novel metabolomics signature of PNPLA3 homozygosity, and a set of metabolites that change with changes in liver fat content among those with PNPLA3 homozygosity. These data have major clinical implications in clinical drug development and companion diagnostics for PNPLA3 targeted therapies.

Figure 1: Metabolic profile of PNPLA3 rs738409_G homozygotes compared to non-carriers.