Methods

4.3.1: Study design

Participant enrolment into TRuTH!

Participants from the START and SAPiT trials who satisfied the following criteria were enrolled into the TRuTH study.

Inclusion criteria

• Adults at least 18 years old

• Previously enrolled in the SAPiT and START trials

• Previous history of active TB disease with HIV co-infection

• Consent to participate and contribute specimens to the KwaZulu-Natal Research Institute for TB and HIV (K-RITH) repository for future investigations

Exclusion criteria

• Refusal to consent

• Diagnosed with XDR TB

This study was reviewed and approved by the University of Kwazulu-Natal biomedical research ethics committee (BF051/09). All participants gave informed consent and consented to have samples stored for future studies.

Participants who met the above criteria were enrolled and longitudinally followed up for up to three years. Recurrent TB disease was detected by chest x-rays, sputum smear and culture using induced sputa, at study entry (baseline), at monthly intervals for the first three months and at three-monthly intervals thereafter. Additional cultures were performed when participants presented with TB symptoms and/or abnormal

confirmed TB were included in our biomarker sub-study. In addition to TB investigations, serum and PBMC were collected at every visit.

Of the 82 microbiologically confirmed progressors, 43 had PBMC samples available and were all included in our sub-study to determine the prognostic value of the ACS 11-gene and rPSVM.1 signatures. Each progressor was assigned to two non- progressors, who did not develop recurrent TB during follow up. Non-progressors were matched to progressors based on time on ART and gender at TRuTH baseline.

Due to the confounding effects of CD4 counts on risk of TB disease and viral loads on IFN responses, we compared these variables between the 43 progressors and 86 non-progressors at first TB episode diagnosis, end of TB treatment, and at TRuTH baseline and confirmed that they were not different. At least three PBMC samples were selected for analysis at three to six monthly intervals from baseline to recurrent TB diagnosis, where possible. Samples collected around the same study day (visit) were selected for progressors and matched non-progressor controls.

Because we could not predict how well the two signatures would perform, we prepared an analysis plan to allow the possible re-parameterisation of the signatures should they perform poorly. Thus, progressors and non-progressors were split into a training and test set. Ten different iterations, randomly selecting 29/14 progressors per split were done and a split with suitable distributions of two-thirds of samples in the training set and a third of samples in the test set was nominated.

We also sought to determine the ability of the signatures to differentiate subclinical TB from active TB disease or Mtb infection in HIV-infected persons in a cross-

sectional cohort of ART naïve HIV-infected individuals (Esmail cohort). We hypothesized that the signatures would differentiate subclinical TB from Mtb infection but not from active TB disease. HIV-infected persons with no previous history of active TB disease were screened from an outpatient clinic just outside Cape Town (Esmail et al., 2016, 2018). This study was reviewed and approved by the research ethics committees of the Universities of Cape Town (013/2011) and Stellenbosch (N12/11/079). Informed consent was obtained from all participants before screening and study entry. Mtb-infected (QFT-positive) persons, with no evidence of active TB on chest x-ray and CD4 counts greater than 350/mm³ were recruited to undergo a 2- deoxy-2 [¹⁸F] fluoro-D-glucose positron emission and computed tomography (PET- CT) scan. Included in this analysis were participants with no features of disease activity on PET-CT, classified as latently (Mtb)-infected persons. Those with features of disease activity on PET-CT, but who were asymptomatic with initial negative sputum culture, were classified as subclinical TB. In addition, culture or Xpert MTB- RIF positive symptomatic active TB disease cases were included (Esmail et al., 2016, 2018).

4.3.2: RNA extraction and quantification !

RNA from the TRuTH cohort was extracted from cryopreserved PBMC as per manufacturer’s instructions, as described in section 2.2.2. Extracted RNA was quantified using the Nanodrop ND 2000 spectrophotometer as described in section 2.2.3. Already extracted RNA samples from whole blood from participants enrolled into the Esmail cohort were received from our collaborators and these were not quantified.

4.3.3: Gene expression measurements !

cDNA was synthesized, reverse transcribed and pre-amplified using TaqMan primer- probes (described in section 2.3.1). Gene expression levels of the signature genes were measured by microfluidic qRT-PCR on the Biomark HD instrument (described in section 2.3.2).

4.3.4: Data analysis

!

Data generated were quality controlled and analysed as described in sections 2.4 and 2.5. Blind predictions were applied to both the training and test sets simultaneously before unblinding of participants’ progressor/non-progressor status in the training set. Analyses on the test set were only carried out after analyses of the training set were complete. Time to diagnosis of recurrent TB (days) was calculated for all progressors by subtracting the sample collection date from the date of recurrent TB diagnosis. For non-progressors, a “time to recurrent TB” was calculated by subtracting sample collection date from the recurrent TB diagnosis date of assigned progressors. Statistical analyses were done in GraphPad Prism (v7.0a) and R.