Whole Blood vs Serum-Derived Exosomes for Host and Pathogen-Specific Tuberculosis Biomarker Identification: RNA-Seq-Based Machine-Learning Approach

Abstract

Mycobacterium tuberculosis (Mtb) remains a leading infectious disease responsible for millions of deaths. RNA sequencing is a rapidly growing technique and a powerful approach to understanding host and pathogen cross-talks via transcriptional responses. However, its application is limited due to the high costs involved.This study is a preliminary attempt to understand host-pathogen cross-talk during TB infection in different TB clinical cohorts using two biological fluids: Whole blood and serum exosomes (EXO). We conducted an RNA-sequencing machine-learning approach using 20 active TB (ATB), 11 latent TB (LTB), three healthy control (HC) whole blood datasets, and two ATB, LTB, and HC serum EXO datasets. During the study, host-derived differentially expressed genes (DEGs) were identified in both whole blood and EXOs, while EXOs were successful in identifying pathogen-derived DEGs only in LTB. The majority of the DEGs in whole blood were up-regulated between ATB and HC, and ATB and LTB, while down-regulated between LTB and HC, which was vice versa for the EXOs, indicating different mechanisms in response to different states of TB infection across the two different biological samples. The pathway analysis revealed that whole blood gene signatures were mainly involved in host immune responses, whereas exosomal gene signatures were involved in manipulating the host's cellular responses and supporting Mtb survival. Overall, identifying both host and pathogen-derived gene signatures in different biological samples for intracellular pathogens like Mtb is vital to decipher the complex interplay between the host and the pathogen, ultimately leading to more successful future interventions.

Keywords: Biomarkers; Exosomes; Machine learning; RNA sequencing; Tuberculosis; Whole blood.