A modular transcriptional signature identifies phenotypic heterogeneity of human tuberculosis infection

Abstract

Whole blood transcriptional signatures distinguishing patients with active tuberculosis from asymptomatic latently infected individuals have been described but, no consensus exists for the composition of optimal reduced gene sets as diagnostic biomarkers that also achieve discrimination from other diseases. We have recapitulated a blood transcriptional signature of active tuberculosis using RNA-Seq, previously reported by microarray that discriminates active tuberculosis from latently infected and healthy individuals, also validated in an independent cohort. We show that an advanced modular approach, which preserves and presents a signature of the entire transcriptome, can better discriminate patients with active tuberculosis from both latently infected and acute viral and bacterial infections. We suggest a method of targeted gene selection across modules for constructing diagnostic biomarkers, more representative of the transcriptome that overcomes some limitations of existing techniques. Finally, we utilise the modular approach to demonstrate dynamic heterogeneity in a longitudinal study of recent tuberculosis contacts.Tuberculosis (TB) is the leading cause of global mortality from an infectious disease. In 2016, there were 6.3 million new cases of TB disease and 1.67 million deaths and its diagnosis is problematic 1 . However, clinical disease represents one end of a spectrum of infection states. It is estimated that up to one third of all individuals worldwide have been infected with the causative pathogen, Mycobacterium tuberculosis, but the vast majority remain clinically asymptomatic with no radiological or microbiological evidence for active infection. This is termed latent TB infection (LTBI) and conceptually denotes a state in which M. tuberculosis persists within its host, while maintaining viability with the potential to replicate and cause symptomatic disease. Indeed, LTBI represents the primary reservoir for future incident TB, with 90% of all TB cases estimated to arise from reactivation of existing infection 1,2 . The risk of incident TB arising from existing LTBI is heterogeneous, poorly characterised and modifiable with anti-tuberculous treatment. Modelling studies indicate effective TB prevention to significantly reduce future TB incidence requires policies directed at the identification and treatment of LTBI 3 . However, implementation of mass screening programmes for this purpose are severely constrained by the size of the target population. Transformative advances in diagnostic tools that can effectively stratify TB risk in the LTBI population are therefore implicit to the realisation of systematic screening.The basis for LTBI heterogeneity rests with the limited scope of the tools we have available to identify the state. LTBI is inferred solely through evidence that immune sensitization has occurred, by the tuberculin skin test (TST) or the M. tuberculosis antigen-specific interferon-gamma (IFN-g) release assay (IGRA). Although these tests are both sensitive and specific for identi...

Clinical metagenomics

“…Although RNA seq analyses are commonly performed on whole blood or peripheral blood mononuclear cell (PBMC) samples, any body fluid or tissue type is potentially amenable to these analyses. Classification of genes by expression profiling using RNA seq has been used to characterize several infections, including staphylococcal bacterae mia 67 , Lyme disease 68 , candidiasis 69 , tuberculosis (dis criminating between latent and active disease risk) [70][71][72] and influenza [73][74][75] . Machine learningbased analyses of RNA seq data have been used for cancer classifi cation 76 , and translation of these approaches may be promising for infectious diseases.…”

Interleukin-1 receptor antagonist mediates type I interferon-driven susceptibility to Mycobacterium tuberculosis

“…Since only individuals with active TB readily transmit infection, and as humans are the only natural reservoir of Mtb , a favored strategy to contain the TB epidemic is to identify and treat latently infected individuals likely to progress to active disease 1,3 . Identification of such individuals is challenging, but recent studies have demonstrated that an enhanced type I interferon (IFN) signature correlates with active TB 5,6,8 and can predict progression to active TB up to 18 months prior to diagnosis 3,4 . A partial loss-of-function polymorphism in the type I IFN receptor (IFNAR1) is associated with resistance to TB in humans, suggesting that elevated levels of type I IFNs not only predict but may even be causally linked to TB progression 15 .…”

“…Although ~1.7 billion people are infected with Mtb 2 , most infections are asymptomatic. Progression to active disease occurs in ~10% of infected individuals and is predicted by an elevated type I interferon (IFN) response 3–8 . Type I IFNs are vital for antiviral immunity, but whether or how they mediate susceptibility to Mtb has been difficult to study, in part because the standard C57BL/6 (B6) mouse model does not recapitulate the IFN-driven disease that appears to occur in humans 3–5,8 .…”

Peripheral inflammatory response in human tuberculosis treatment is predicted by a combination of pathogen sterilization and microbiome dysbiosis

“…Specifically, infection with Mtb leads to heightened expression of inflammatory pathways, most notably the Type I and Type II interferon pathways [12][13][14][15] , and this pattern resolves with antibiotic therapy 12,15,16 . A recent meta-analysis combining microarray and RNAseq data from studies aimed at identify active TB transcriptional signatures confirmed the findings about a specific set of peripheral blood transcripts that are biomarkers of active TB disease, relative to healthy individual, or those with latent TB infection (LTBI) 17 .…”

“…B. Comparison of log2 Fold Change in gene expression from17 where the 373 active TB signature was first introduced with the log2 Fold Change of the same genes pre-post HRZE (left) or NTZ (right) from this study. A significant positive correlation (Pearson's p<0.01) is observed for the 151 genes that renormalize with HRZE treatment.All rights reserved.…”