To Örebro University

Background: MiRNAs represent a class of small non-coding RNAs which are involved in regulation of protein-coding gene expression. Being implicated in various processes such as development and regu-latory circuits of cells, miRNAs also play an important role in the etiology of a variety of diseases. Imbalance of the regulatory pro-cesses within immune system development and response may lead to disturbed production of pro-inflammatory cytokines and over-reactivity of the immune cells, thus causing relapsing inflamma-tion, a characteristic feature of inflammatory bowel disease (IBD). Recent studies of colonic miRNAs employed NGS for the distinction between CD, UC and healthy controls, or among different CD sub-types. However, NGS-based profiles of blood-circulating miRNAs have thus far not been investigated in the context of IBD together with other immune-mediated diseases, including ankylosing spon-dylitis, psoriasis, systemic lupus erythematosus, rheumatoid arthritis and sarcoidosis, as well as non-immune hemolytic-uremic syndrome.

Methods: Study participants were recruited in Germany and Sweden, where peripheral blood samples (PAXgene) as well as phenotypical and clinical information (such as treatment status, dis-ease activity and location) was collected. Small RNA transcriptomes of 680 individuals (Figure 1) were sequenced using Illumina NGS platform. Small RNA-seq data preprocessing and quantification were performed using cutadapt and miraligner (ref. miRBase v22), respectively. Differential expression analysis (DESeq2) and correla-tion (Spearman) analysis have been performed to identify disease activity-, trait- and treatment-specific miRNA signatures. These sig-natures were then utilized in a machine-learning approach to build classification models for IBD diagnostics.

Results: The results of multiple pairwise differential expression anal-yses among different immune-mediated inflammatory conditions and healthy controls revealed inflammation-specific as well and dis-ease-specific deregulation of miRNAs. Correlation analysis identified miRNAs positively and negatively correlated with IBD activity. The preliminary results of machine learning classifiers based on miRNA profiles showed that median Matthews correlation coefficient for all model types showed remarkable predictive performance estimated as being 1.00 (median over main diagnoses), as well as ranging from 0.68 to 0.76 (median over CD location) and from 0.69 to 0.77 (median over UC extent).

Conclusions: Immune-mediated inflammatory diseases share com-mon and distinct differentially expressed miRNAs, which have a potential to be used in the diagnostics of IBD, including the evalua-tion of the disease activity.