DiagnosticsExisting blood transcriptional classifiers accurately discriminate active tuberculosis from latent infection in individuals from south India
Introduction
Tuberculosis (TB), an infectious disease caused by Mycobacterium tuberculosis (Mtb), is a global health concern with 10.4 million new cases estimated in 2015 [1]. Untreated TB has a high mortality rate, estimated at a 70% 10-year case fatality rate in smear-positive pulmonary tuberculosis [2]. The WHO estimates that depending on context, 15–50% of TB cases are unreported or undiagnosed [1]. Further, the limitations of bacterial diagnosis in paucibacillary TB that is pediatric, extrapulmonary, or smear-negative pulmonary TB, often lead to empirical treatment. This approach exposes some individuals unnecessarily to side effects of the TB treatment while delaying effective therapy of the actual cause of the disease. Further, patients with TB may be denied appropriate treatment if bacteriologic testing is negative.
Currently, though nucleic acid amplification tests approach the sensitivity of Mtb culture [3], more sensitive diagnostics may need to be based on host biomarkers. These tests are inherently nonspecific, and it is unclear how they will perform in different populations. Several transcriptomic studies of TB cases and individuals infected with Mtb have been performed to characterize systemic gene expression (reviewed in Ref. [4]). Jacobsen et al. (2007) performed a microarray analysis and identified a minimal set of three genes in peripheral blood mononuclear cells that allowed distinction between TB disease and healthy individuals with latent TB infection (LTBI) [5]. The Berry et al. (2010) microarray studies demonstrated a 393-gene whole blood signature that discriminated subjects with active TB disease from those with LTBI, as well as an 86-gene set discriminating TB disease from other inflammatory and infectious diseases [6]. In this study, they also noted that this active TB signature was extinguished in patients following anti-TB treatment [6]. Kaforou et al. (2013) proposed a 27-gene whole blood signature that could distinguish TB and latent TB, regardless of HIV infection status [7]. Zak et al. (2016) recently identified a 16-gene signature for predicting TB disease risk through sequencing analysis of whole blood PAXgene samples from a prospective cohort [8]. Given this recent proliferation of transcriptional studies in the field of TB, Sweeney et al. (2016) performed a meta-analysis of 14 publicly available TB transcriptomic datasets to identify a 3-gene set that they determined was robustly diagnostic for active TB disease [9].
Strikingly, although India accounts for more than one-quarter of the world's TB cases and deaths, transcriptomic studies in an Indian Mtb-infected populations have been limited [1,[10], [11], [12]]. TB is hyperendemic in India with an incidence of 217 per 100,000 and a mortality rate of 36 per 100,000 individuals as estimated in 2015 [1]. Furthermore, there is a high prevalence of other non-communicable confounding conditions and risk factors complicating TB disease cases in India, including diabetes, smoking, and alcohol consumption [[13], [14], [15]]. Presence of such comorbidities in patients with TB can impact their disease courses and treatment responses [[16], [17], [18], [19]]. Thus, while two studies have proposed TB disease gene signatures (Maertzdorf et al. 4-gene signature and Sambarey et al. 10-gene signature) that were evaluated in Indian populations, additional studies are required [10,11]. Thus, the overall goal of this study was to assess how well the published gene signatures of active TB disease classified subjects in patients from South India.
Section snippets
Study subjects and inclusion criteria
Subjects were recruited into the current cross-sectional sub-study from an ongoing observational household contact study being conducted at Jawarharlal Institute of Postgraduate Medical Education and Research (JIPMER). TB cases were recruited through the Revised National TB Control Programme (RNTCP) network of clinics in the city of Pondicherry (Puducherry union territory) and the districts of Villupuram and Cuddalore (Tamil Nadu state) in South India. Through the RNTCP, symptomatic individuals
Differentially expressed genes in Indian subjects with TB versus LTBI overlap with existing TB signatures
The DEGs were identified between patients with active TB disease and asymptomatic individuals with latent TB infection (LTBI) in the Indian study. More than 1200 genes were differentially expressed between TB and LTBI individuals at a 0.00001 adjusted p-value (FDR) cutoff. For visualization and discussion purposes of a shorter DEG list, a more stringent adjusted p-value (FDR) cutoff of <10−11 was used resulting in a list of 76 DEGs (Fig. 1, Supplementary Table S2). Many DEGs identified from
Discussion
The results of this study suggest that gene expression of existing transcriptional signatures can accurately classify TB and LTBI individuals within the South Indian subjects in our study. Distinct patterns of gene expression were observed in active TB patients versus LTBI individuals for each of the eight published gene sets evaluated. Whereas seven of the tested signatures were designed to characterize an active TB profile, the Zak et al. (2016) 16-gene signature was described as a predictor
Funding
This publication is based on work supported by Award No. USB1-31150-XX-13 of the U.S. Civilian Research & Development Foundation (CRDF Global), Department of Biotechnology BT/PR8941/MED/15/108/2013 and by the National Science Foundation under Cooperative Agreement No. OISE-9531011.
Conflicts of interest
None.
Author contributions
PS, JJE, NSH, SS and CRH contributed to cohort design; SL and YZ performed the experiments and analyzed the data; PS and WEJ designed experiment and interpreted the data; NJ, SS, JP, NSH, DH, SL, GR contributed to subject recruitment and sample collection, storage and shipping, and data management; SL, YZ, PS and WEJ wrote the manuscript; PS, JJE and WEJ contributed to discussion and conclusions of the study.
Acknowledgements
We thank all of the participants who made the study possible. We also acknowledge the dedication and hard work of the field staff and study team. We thank Rachel Kubiak and Astrid Loomans for help with data management, and Laura White for help with statistics. This work was performed under the auspices of the RePORT India Program. We wish to acknowledge the support of Sudha Srinivasan and Peter Kim from DAIDS/NIAID/NIH and Jyoti M. Logani from Department of Biotechnology, Ministry of Science
References (29)
- et al.
Advances in tuberculosis diagnostics: the Xpert MTB/RIF assay and future prospects for a point-of-care test
Lancet Infect Dis
(2013) - et al.
Genome-wide expression for diagnosis of pulmonary tuberculosis: a multicohort analysis
Lancet Resp Med.
(2016) - et al.
Unbiased identification of blood-based biomarkers for pulmonary tuberculosis by modeling and mining molecular interaction networks
EBioMed
(2017) - et al.
Non-communicable disease comorbidities and risk factors among tuberculosis patients, Meghalaya, India
Indian J Tubercul
(2016) - et al.
Improving tuberculosis prevention and care through addressing the global diabetes epidemic: from evidence to policy and practice
Lancet Diabet Endocrinol
(2014) - World Health Organization. reportGlobal tuberculosis report...
- et al.
Natural history of tuberculosis: duration and fatality of untreated pulmonary tuberculosis in HIV negative patients: a systematic review
PLoS One
(2011) - et al.
Tricks to translating TB transcriptomics
Ann Transl Med
(2015) - et al.
Candidate biomarkers for discrimination between infection and disease caused by Mycobacterium tuberculosis
J Mol Med (Berl)
(2007) - et al.
An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis
Nature
(2010)
Detection of tuberculosis in HIV-infected and -uninfected African adults using whole blood RNA expression signatures: a case-control study
PLoS Med
A blood RNA signature for tuberculosis disease risk: a prospective cohort study
Lancet
Concise gene signature for point-of-care classification of tuberculosis
EMBO Mol Med
Systems immunology of diabetes-tuberculosis comorbidity reveals signatures of disease complications
Sci Rep
Cited by (49)
A protein signature associated with active tuberculosis identified by plasma profiling and network-based analysis
2022, iScienceCitation Excerpt :The final signature consisted of the proteins interferon (IFN)-gamma, interleukin (IL)-6, CUB-domain containing protein 1 (CDCP1), CXCL9, matrix metalloproteinase (MMP)-1, monocyte chemotactic protein (MCP)-3, CCL19, CD40, vascular endothelial growth factor A (VEGFA), IL-7, IL-12B, and programmed cell death ligand 1 (PD-L1). To assess the significance of the 12-marker signature in independent TB cohorts, we performed ssGSEA on published transcriptomic TB datasets28,29,30,31,32,33,34,35,36 from the curated TBData R package (Table S1). We observed higher enrichment scores in the active TB group, compared with either latent TB or healthy controls in all the transcriptomic TB datasets except one (GEO: GSE19444) (Figure 4A).
A 10-gene biosignature of tuberculosis treatment monitoring and treatment outcome prediction
2021, TuberculosisCitation Excerpt :For TB treatment outcome prediction, the ability of the biosignature to predict the cured versus relapse patients (GES67589) and failure patients (GSE89403) were assessed. Moreover, the predictive capacity of obtained biosignatures in different TB diagnostic situations was investigated, such as TB versus non-TB comprising other respiratory diseases and healthy controls (E-MTAB-8290), versus latent TB infection (GSE101705 [21], GSE107991 [22]), and healthy control subjects (GSE107991). Additionally, the data set GSE94438 [16] was utilized to evaluate the performance of the biosignature to predict the risk of TB progression by separating progressors and non-progressors.
Food for thought: addressing undernutrition to end tuberculosis
2021, The Lancet Infectious DiseasesCitation Excerpt :Consequently, current diagnostic tests for tuberculosis infection, such as interferon-γ release assays and tuberculin skin testing, might have reduced sensitivity in severely undernourished individuals, although conflicting data exist.33,34 Blood transcriptomic biomarkers are being evaluated as surrogate predictors of infection and disease.35 On the basis of unpublished data from the consortium Regional Prospective Observational Research in Tuberculosis (RePORT) in India, presented at the 50th Union World Conference on Lung Health in 2019, undernutrition does not alter the most validated tuberculosis gene signatures in distinguishing active tuberculosis from latent infection.36
An empirical approach towards detection of tuberculosis using deep convolutional neural network
2024, International Journal of Data Mining, Modelling and Management
- 1
Equal contribution.