The influence of metformin treatment on the circulating proteome

Abstract

Background: Metformin is one of the most used drugs worldwide. Given the increasing use of proteomics in trials, bioresources, and clinics, it is crucial to understand the influence of metformin on the levels of the circulating proteome.

Methods: We analysed a combined longitudinal proteomics dataset from the IMPOCT, RAMP and S3WP-T2D clinical trials in 98 participants before and after metformin exposure. This discovery analysis contained 372 proteins measured by proximity extension assays (Olink). We followed up experiment-wise statistically significant findings in two cross-sectional cohorts of people with type 2 diabetes comparing metformin treated and untreated individuals: IMI-DIRECT (784 participants, 372 proteins, Olink) and IMI-RHAPSODY (1175 participants, 1195 proteins, SomaLogic).

Findings: Overall, 23 protein analytes were robustly associated with exposure to metformin in the discovery and replication. This includes 11 protein-metformin associations that replicated in both replication sets and platforms (REG4, GDF15, REG1A, t-PA, TFF3, CDH5, CNTN1, OMD, NOTCH3, THBS4 and CD93), with the remaining 12 protein-metformin associations replicated using the Olink platform (EPCAM, SPINK1, SAA-4, COMP, ITGB2, ADGRG2, FAM3C, MERTK, COL1A1, HAOX1, VCAN, TIMD4) but not measured on the SomaLogic platform. Gene-set enrichment analysis revealed that the metformin exposure was associated with intestinal associated proteins.

Interpretation: These data highlight the need to account for exposure to metformin, and potentially other drugs, in proteomic studies and where protein biomarkers are used for clinical care.

Funding: Innovative Medicines Initiative Joint Undertaking 2, under grant agreement no. 115881 (RHAPSODY) and the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115317 (DIRECT), resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA companies in kind contribution as well as the Swiss State Secretariat for Education Research' and Innovation (SERI), under contract no. 16.0097 (RHAPSODY).

Keywords: Biomarker; Metformin; Proteomics; Type 2 diabetes.

Fig. 1

Flowchart showing the study design which identifies many proteins that are significantly changed following exposure to metformin.

Fig. 2

Effect of metformin on plasma protein levels in the discovery analysis. Volcano plot showing if protein concentrations are significantly increased or decreased following metformin treatment in the longitudinal Olink analysis (n = 98). Estimate (beta coefficient) is plotted on the x axis and -log10 of the unadjusted p value (calculated from the linear mixed-model) is plotted on the y axis. Proteins with an adjusted p-value (Bonferroni method) of less than 0.05 are represented by a yellow dot and all other non-significantly changed proteins are represented by a grey dot. Proteins which have an increased concentration following metformin treatment have a positive effect size whereas proteins which have a decreased concentration following metformin treatment have a negative effect size.

Fig. 3

Comparison of effect size across the three studies showing the 11 protein-metformin associations that were replicated in both replication studies (all P ≤ 0.05). Data represent mean ± SE. X-axis, effect size; y-axis, protein. In dark blue the main result for each cohort is shown. Discovery (n = 98), Replication Olink (n = 784) and replication SomaLogic (n = 1175). Green colours represent the results of analysis stratified by diabetes status (dark green = controls, n = 50; light green = T2D, n = 48). Results for the replication cohorts aren't shown as these consist of persons with T2D only. Results of sex stratified analyses where shown as dark red = men and lighter red = women. Discovery (men, n = 52; women, n = 46), Replication Olink (men, n = 449; women, n = 335) and replication SomaLogic (men, n = 679; women, n = 496).