A proposed workflow to robustly analyze bacterial transcripts in RNAseq data from extracellular vesicles.

INTRODUCTION: The microbiota has been unequivocally linked to various diseases, yet the mechanisms underlying these associations remain incompletely understood. One potential contributor to this relationship is the extracellular vesicles produced by bacteria (bEVs). However, the detection of these bEVs is challenging. Therefore, we propose a novel workflow to identify bacterial RNA present in circulating extracellular vesicles using Total EV RNA-seq data. As a proof of concept, we applied this workflow to a dataset from individuals with multiple sclerosis (MS). METHODS: We analyzed total EV RNA-seq data from blood samples of healthy controls and individuals with MS, encompassing both the Relapsing-Remitting (RR) and Secondary Progressive (SP) phases of the disease. Our workflow incorporates multiple reference mapping steps against the host genome, followed by a consensus selection of bacterial genera based on various taxonomic profiling tools. This consensus approach utilizes a flagging system to exclude genera with low abundance across profilers. Additionally, we included EVs derived from two cultured species that serve as biological controls, as well as artificially generated reads from 60 species as a technical control, to validate the specificity of this workflow. RESULTS: Our findings demonstrate that bacterial RNA can indeed be detected in total EV RNA-seq from blood samples, suggesting that this workflow can be a powerful tool for reanalyzing RNA-seq data from EV studies. Additionally, we identified promising bacterial candidates with differential expression between the RR and SP phases of MS. DISCUSSION: This approach provides valuable insights into the potential role of bEVs in the microbiota-host communication. Finally, this approach is translatable to other experiments using total RNA, where the lack of a robust pipeline can lead to an increased false positive detection of microbial genera. The workflow and instructions on how to use it are available at the following repository: https://github.com/NanoNeuro/EV_taxprofiling.