Human genetic investigation of TMPRSS6 as a therapeutic target for liver disease: efficacy and safety liabilities

Abstract Background & Aims: Disease-modifying therapies for steatohepatitis and fibrosis remain limited. Matriptase-2 (TMPRSS6), a serine protease suppressing hepcidin to regulate iron homeostasis, is under clinical investigation for iron overload, and preclinical evidence suggests its inhibition may attenuate metabolic liver disease. Whether these effects translate to humans remains unknown. Methods: We proxied TMPRSS6 inhibition using the missense variant rs855791, weighted on circulating hepcidin. Using Mendelian randomization (MR), we assessed effects on (1) magnetic resonance imaging-derived liver phenotypes, including corrected T1 (cT1) and proton density fat fraction (PDFF), liver enzymes and clinical liver outcomes; (2) cardiometabolic endpoints; (3) 2,940 plasma proteins; (4) clinical biomarkers; and (5) 2,272 disease endpoints. Colocalization and mediation analyses were used to characterize mechanisms. Results: Genetically proxied TMPRSS6 inhibition was associated with lower liver iron content (β = -2.50, P<0.001) and higher cT1 (β = 1.59, P<0.001), with colocalization indicating a shared causal signal. No association was observed with PDFF, liver volume, NAFLD, cirrhosis or hepatocellular carcinoma. Genetically proxied TMPRSS6 inhibition was also associated with lower C-reactive protein (β = −0.17, P<0.001), higher LDL cholesterol (β = 0.22, P<0.001) and higher coronary artery disease (CAD) risk (OR = 1.59, P<0.001), with weak colocalization of CAD with hepcidin (PP.H₄ = 0.56) and LDL cholesterol (PP.H₄ = 0.48). Mediation analysis suggested LDL cholesterol may account for approximately 21% (95% CI 9% to 34%) of the CAD association. Conclusions: Human genetic evidence supports hepatic iron lowering by TMPRSS6 inhibition, but not an effect on steatosis. Higher cT1, without concordant associations with clinical liver outcomes, may reflect reduced hepatic iron rather than worsening fibroinflammation. The CAD signal, partly mediated by LDL cholesterol, warrants careful consideration in trial design.