Characterising metabolomic signatures of lipid-modifying therapies through drug target Mendelian randomisation

• Objective: This study employed drug target Mendelian randomization (MR) to emulate the effects of five different classes of lipid-modifying therapies and characterize their resulting causal impact on 229 circulating metabolic traits.
• Key Findings: The analysis successfully generated distinct metabolomic signatures for each drug class, confirming known effects (e.g., Statins reducing LDL-C) while providing novel insights into their differential impact on lipoprotein subclasses; for instance, PCSK9 inhibitors showed limited effect on VLDL/triglycerides compared to Statins.
• Implication: The findings validate drug target MR as a valuable tool for predicting the pleiotropic effects of therapies on the metabolome, which can inform drug development, safety assessment, and personalized clinical choice.

PubMed: 35213538 DOI: 10.1371/journal.pbio.3001547 Overview generated by: Gemini 2.5 Flash, 28/11/2025

Key Findings: Metabolomic Signatures of Lipid-Lowering Drugs

This study utilized drug target Mendelian randomization (MR) to emulate the effects of five different classes of lipid-modifying therapies and characterize their resulting causal impact on the circulating metabolome (229 metabolic traits).

• Drug Class Signatures Identified: The study successfully characterized the distinct causal metabolomic signatures of the five drug classes, leveraging genetic variants that proxy the function of the drug target gene:
  1. HMGCR inhibitors (Statins): Genetically proxied HMGCR inhibition caused reductions in LDL-C and triglycerides, and importantly, showed significant causal effects on numerous intermediate density lipoprotein (IDL) and very low-density lipoprotein (VLDL) traits.
  2. PCSK9 inhibitors (PCSK9i): Genetically proxied PCSK9 inhibition primarily reduced LDL-C and total cholesterol but had no significant effect on triglycerides or VLDL traits, highlighting a key distinction from statins.
  3. NPC1L1 inhibitors (Ezetimibe): Similar to PCSK9i, NPC1L1 inhibition predominantly reduced LDL-C and showed limited effects on triglycerides or VLDL.
  4. LPL activators (Fibrates/Gemfibrozil): Genetically proxied LPL activation showed the largest causal effect on reducing triglycerides and VLDL traits.
  5. CETP inhibitors: Genetically proxied CETP inhibition primarily increased HDL-C and total cholesterol.
• Causal Impact on Metabolism: The analysis confirmed the established effects of these drug classes on primary lipid fractions (e.g., LDL-C, triglycerides) but provided novel insights into their distinct effects on various metabolite sub-fractions, particularly within the lipoprotein subclasses and amino acid metabolism (e.g., changes in the ratio of polyunsaturated fatty acids to total fatty acids by LPL activation).
• Validation of MR for Drug Targets: The study validated the use of MR to predict the pleiotropic effects of drug targets on a high-dimensional panel of metabolic traits, making it a valuable tool for drug development and safety assessment.

Study Design and Methods

Study Design

This was a two-sample drug target Mendelian randomization (MR) study. The design used genetic variants located near or within the genes encoding the drug targets as instrumental variables (IVs) to proxy the lifelong effect of the drug on the metabolome.

Exposure and Outcome Data

• Exposures (Drug Target Proxies): Genetic variants (SNPs) associated with changes in the concentration of the primary lipid fraction targeted by the drug were used as instrumental variables, effectively mimicking the biological effect of the drug:
  • HMGCR: Proxied Statins (effect on LDL-C).
  • PCSK9: Proxied PCSK9i (effect on LDL-C).
  • NPC1L1: Proxied Ezetimibe (effect on LDL-C).
  • LPL: Proxied Fibrates/Gemfibrozil (effect on Triglycerides).
  • CETP: Proxied CETP inhibitors (effect on HDL-C).
  • Exposure GWAS data for the primary lipids were sourced from large-scale consortia.
• Outcomes (Metabolomic Traits): Summary statistics were used for 229 circulating metabolic traits measured using the high-throughput Nuclear Magnetic Resonance (NMR) spectroscopy platform, predominantly sourced from the UK Biobank (up to \(n=115,076\)). These traits included lipoprotein subclasses, amino acids, fatty acids, and glycolysis-related metabolites.

Statistical Analysis

• MR Methods: The Inverse-Variance Weighted (IVW) method was used as the primary MR approach.
• Sensitivity Analyses: Robust sensitivity methods were performed to ensure the validity of the causal estimates by addressing potential pleiotropy:
  • MR-Egger regression.
  • Weighted Median and Weighted Mode estimators.
  • Multivariable MR (MVMR): Used to confirm that the observed effects were not confounded by the genetic effect on other lipid fractions (e.g., confirming the effect of \(HMGCR\) on metabolites was independent of its effect on \(HDL-C\)).
• FDR Correction: False Discovery Rate (FDR) correction was applied to adjust for multiple testing across the 229 metabolic outcomes for each drug target.

Conclusions and Recommendations

The study concludes that drug target MR is a powerful and efficient method for predicting the comprehensive metabolomic signature and potential pleiotropic effects of new and existing therapies.

• Clinical Utility: The distinct metabolic profiles characterized (e.g., the difference between statin and PCSK9i effects on VLDL/triglycerides) can inform the rational choice of therapy for patients with specific metabolic risk profiles.
• Drug Development: This approach allows for the early assessment of the unintended (pleiotropic) effects of drug targets on a wide range of biological processes, potentially accelerating the identification of both beneficial and adverse effects before large-scale clinical trials.
• Future Work: The authors recommend applying this drug target MR approach to other classes of cardiovascular drugs and integrating even larger metabolomic datasets to increase the statistical power for detecting smaller, more subtle causal effects.