Abstract

Background: Chronic Obstructive Pulmonary Disease (COPD) is a highly heterogeneous condition. However, the biological mechanisms (endotypes) underlying this heterogeneity are still largely undefined.

Methods: In this study for the first time we integrate three omics levels (mRNA, miRNA and DNA methylation) determined in 135 lung tissue samples of ex-smokers with COPD. A patient network, identifying the molecular similarities between individuals, was built for each omic level resulting into a multi-layer network. Finally, communities of patients were detected within the multi-layer framework.

Results: Four stable multi-omics communities were identified with significant differences in relevant clinical features, including prebronchodilator FEV1 % ref. (p-value= 0.0031), Body Mass Index (p-value= 0.022) and blood eosinophils (p-value= 0.015).Specifically, in one unique subtype enriched with GOLD3-4 patients we saw differences at all the layers. It had an hypomethylation of DNA regions that overlapped with genes that participate in the immune response and the same inflammatory pathways were upregulated at the mRNA level. These individuals also had a strong cilia dysfunction as observed in transcriptomics data, which was mediated by the downregulation of miR-34/449 family (required in ciliogenesis) at miRNA level. We could finally validate the mRNA and miRNA features of this endotype in an independent cohort from the Lung Genomics Research Consortium.

Conclusion: We show for the first time that the use of a multi-layer network based on the lung tissue patient similarities, uncovered communities that differed in clinical COPD characteristics and shed light on the biological mechanisms underlying the heterogeneity of the disease.