Researchers identified blood-based DNA methylation signatures that distinguished high-risk prediabetes clusters, according to a study published in Biomarker Research. The findings indicate that peripheral blood epigenetic profiling may support metabolic risk stratification without extensive clinical testing.
Prediabetes is heterogeneous, with variable risks of progression to type 2 diabetes (T2D) and related complications. Previous work defined six prediabetes clusters based on metabolic and imaging parameters, including three high-risk groups: low beta cell function (cluster 3), high insulin resistance (cluster 5), and high insulin secretion (cluster 6). However, cluster assignment required oral glucose tolerance testing, insulin measurements, and MRI-based fat quantification, noted Amandeep Singh, of the Department of Experimental Diabetology, German Institute of Human Nutrition Potsdam-Rehbruecke, and colleagues.
In the current study, investigators profiled genome-wide DNA methylation in peripheral blood from a discovery cohort of 187 individuals and a replication cohort of 146 individuals with prediabetes. Methylation was measured using Infinium MethylationEPIC arrays, followed by normalization and correction for technical and biological confounders.
More than 120,000 differentially methylated CpG sites were initially identified across clusters. To reduce dimensionality, the researchers applied an elastic net machine learning workflow with repeated cross-validation. Through this approach, 1,557 CpG sites were selected as stable predictors of cluster membership.
In the discovery cohort, classification accuracy exceeded 95% across clusters. In the independent replication cohort, methylation-based partitioning reproduced high-risk cluster identities with 92% accuracy. Randomly selected CpG sets did not achieve comparable performance.
Between 300 and 349 CpG sites were specific to each cluster. Cluster 3–specific markers were enriched in genes involved in TGF-β receptor and calcium signaling. Cluster 5 markers were associated with MAPK signaling and extracellular matrix organization. Cluster 6 markers were linked to Wnt and SMAD signaling pathways. These pathway associations were consistent with the metabolic features of each cluster.
Of the 1,557 CpG sites, 1,512 were significantly correlated with at least one anthropometric or metabolic trait. The strongest correlations were observed with insulin sensitivity (Matsuda index), insulin concentrations, and glucose levels during oral glucose tolerance testing. Associations were also identified with triglycerides, HDL cholesterol, visceral adipose tissue, liver fat, and C-reactive protein.
Comparison with published epigenome-wide association data showed that 373 CpG sites overlapped with loci previously linked to T2D incidence, inflammation, cardiovascular disease, or kidney disease.
"Clinically, the identification of blood-borne biomarkers for the prediabetes offers the advantage of eliminating the need for time- and resource-intensive tests, such as the oral glucose tolerance test. This would enable the application of the risk stratification approach to broader populations," wrote investigators.
They noted that the study population was limited to individuals of Central European ancestry and that longitudinal validation is needed to determine predictive value for incident T2D or complications.
The investigators reported having no conflicts of interest.
