Transcriptomic signatures of recurrent tuberculosis disease and treatment response in HIV-infected individuals

Doctoral Thesis


Permanent link to this Item
Journal Title
Link to Journal
Journal ISSN
Volume Title

University of Cape Town

HIV-infected persons are at particularly high risk of tuberculosis (TB) disease, especially in TB endemic countries where M. tuberculosis transmission is common. Although antiretroviral therapy (ART) reduces risk of TB, it does not return to that of HIV-uninfected persons. In addition, a previous history of TB disease significantly increases the risk of recurrent TB disease. Identification of HIV-infected individuals at greatest risk of recurrent TB for highly targeted therapy before disease manifestation would be a major advance in the fight against TB. This would also allow the provision of TB treatment in persons with subclinical TB. Diagnosis of TB in HIV-infected persons is markedly undermined by the paucibacillary nature of HIV-associated disease. A non-sputum based diagnostic test that is highly sensitive and specific, such as a blood-based RNA signature, would be an important new tool. Such a test may also facilitate monitoring of TB treatment, opening the possibility for customizing the duration of TB treatment to that necessary for cure. We previously discovered and validated a 16-gene transcriptomic signature with promising prognostic and diagnostic utility for TB in HIV-uninfected persons. The transcriptomic signature could predict progression to active TB disease up to a year before diagnosis and was shown to be a useful tool for TB treatment response monitoring in a treatment cohort of HIV-uninfected persons. The signature was reduced to an 11-gene signature to improve throughput with equivalent prognostic and diagnostic performance. In addition, we developed a smaller, 6-gene signature in preparation for translation to point-of-care testing. In this thesis, we aimed to determine (1) the diagnostic performance of these two transcriptomic signatures in HIV-infected persons, (2) whether they could predict recurrent TB disease in HIV infected persons, and (3) their utility to monitor TB treatment response in HIV infected persons. To assess aim 1, we designed a cross-sectional study of HIV-infected (n=40) and uninfected persons (n=60), each comprising equal numbers of active TB cases and QuantiFERON-positive controls. To assess aims 2 and 3, we designed retrospective substudies among participants enrolled into two clinical studies previously completed by our collaborators at CAPRISA, namely the TRuTH and IMPRESS studies. In the TRuTH cohort participants who developed recurrent TB diagnosis, diagnosed by microbiological testing of induced sputum, were assigned as progressors (n=43), while those who remained asymptomatic were assigned as non-progressors (n=86). In the IMPRESS cohort, participants with a new diagnosis of recurrent TB who initiated TB treatment were stratified into early (n=44) and late (n=19) converters based on time to sputum culture conversion from diagnosis. RNA was isolated from cryopreserved PBMC or PAXgene whole blood and gene expression measured by microfluidic qRT-PCR. Signature scores were generated using in-house customised scripts in R and performance of the signatures was measured using receiver operating characteristic area under the curve (ROC AUC), calculated using the pROC and verification packages in R. The 11-gene and 6-gene signatures could diagnose active TB disease in HIV infected persons with good accuracy (AUC = 0.83 and 0.92, respectively), although performances were lower than those observed in HIV-uninfected persons (AUC = 0.97 and 0.96). Signature performance was decreased in HIV-infected persons due to higher signature scores, reflecting high expression of IFN-stimulated genes, especially in HIV-infected controls. In the TRuTH cohort, these signatures could identify those with recurrent TB within 3 months of diagnosis (AUC = 0.77, p = 0.003), suggesting detection of subclinical disease. Scores of both signatures decreased during TB treatment in the IMPRESS cohort, in participants with early or late sputum conversion. Importantly, two months after initiating TB treatment, the ACS 11-gene signature could differentiate early from late converters. Detectable plasma viral load was associated with higher signature scores in both cohorts, leading to a decrease in signature specificities. We show that the 11-gene and 6-gene signatures performed well as blood-based diagnostic tests for active TB disease in HIV-infected persons. The signatures could detect recurrent TB disease during the subclinical phase of disease progression and demonstrated promise as treatment response markers in HIV-infected persons. The signatures performed best in persons with effectively suppressed HIV load, highlighting the importance of ART adherence and integration of HIV and TB care for effective clinical management of TB.