Background: The molecular mechanisms underlying the beneficial effects of sodium-glucose co-transporter-2 inhibition (SGLT2i) on cardiovascular outcomes are largely unknown. To better understand the (patho)mechanistic effects of SGLT2i, we investigated the effects of the SGLT2i empagliflozin on circulating proteins in patients with type 2 diabetes mellitus (T2DM) and diastolic dysfunction.

Methods: In the EmDia study (NCT02932436), 538 plasma proteins were quantified at baseline and after 12 weeks of treatment in 132 participants with T2DM and elevated left ventricular end-diastolic pressure (estimated using echocardiographic E/e’ ratio) who were randomized 1:1 to empagliflozin vs. placebo. Proteins were measured using immuno-qPCR based on proximity extension assay technology (Olink, Uppsala, Sweden). Elastic net regression of the 12-week change values of the 538 proteins was used to identify the protein signature associated with 12 weeks of empagliflozin vs. placebo. Pathway enrichment analysis was performed to aid interpretation of results. Linear regressions were used to investigate whether 12-week changes in identified proteins were associated with clinical traits altered by empagliflozin. To validate the proteomic results, a LASSO regression based dimension-reduced protein score for empagliflozin treatment was generated and assessed for its prognostic value for all-cause mortality in N=3,188 participants with and without T2DM from an independent heart failure (HF) cohort (MyoVasc, NCT04064450).

Results: Elastic net regression selected a signature of 67 proteins mapping to empagliflozin treatment (10-fold internal cross-validated AUC=0.78). Approximately 75% (N=50) of these proteins were associated with at least one clinical trait altered by empagliflozin over 12 weeks of intervention. Pathway enrichment analyses showed that the identified proteins are particularly involved in inflammatory processes, immune responses, extracellular matrix, angiogenesis and metabolism. The protein score showed prognostic value in terms of all-cause mortality in the independent HF cohort (hazard ratio (HR) per -SD 1.15, [95% confidence interval (95%CI) 1.03,1.28], p=0.016) independent of age, sex, cardiovascular risk factors and comorbidities. The protein score predicted all-cause mortality even in individuals with pre-HF or symptomatic HF (stage B-D) without T2DM (HR per -SD 1.32, [95%CI 1.05,1.65], p=0.017).

Conclusions: This investigation provided insight into putative molecular mechanisms of action of empagliflozin treatment in patients with T2DM in relation to cardiovascular, haematological, metabolic and neurological clinical traits. The importance of the protein signature was underlined by its prognostic effect on all-cause mortality independently of diabetes status and cardiovascular risk factors in the independent HF cohort.