Optimisation and application of a T-cell antigen-specific activation assay as diagnostic and treatment-monitoring tool for tuberculosis

Abstract

Background

Mycobacteria-specific T-cell activation is a robust biomarker of recent Mycobacterium tuberculosis infection, disease progression and tuberculosis (TB) disease. We evaluated this promising biomarker, termed TB-TASA, as a treatment monitoring tool in the context of a treatment-shortening study.

Methods

Using a panel of mycobacterial antigens, we observed higher TB-TASA scores in TB patients compared to interferon-γ (IFN-γ) release assay positive (IGRA⁺) controls, regardless of mycobacterial antigen specificity, consistent with previous studies. We derived a TB-TASA positivity threshold of 10% HLA-DR⁺ mycobacteria-specific T-cells using a computational analysis pipeline developed with the OpenCyto R package. This threshold was robust across three previously published case–control studies that used different sample types (i.e., whole blood or peripheral blood mononuclear cells), varying duration of antigen stimulation and different flow cytometry antibody panels, all of which achieved sensitivity and specificity >90% and >70%, respectively.

Results

In a prospective randomised clinical trial assessing treatment shortening in TB patients with less severe disease, higher TB-TASA scores were observed at treatment completion in patients who relapsed or failed treatment during follow-up compared to those successfully treated with a receiver operating characteristic area under the curve (ROC AUC) of 0.89 (95% CI 0.69–1), supporting TB-TASA as a treatment monitoring tool

Discussion

We demonstrated that TB-TASA could also be reliably measured in capillary blood collected via fingerprick from IGRA⁺ controls and TB patients, achieving an ROC AUC of 0.96 (95% CI 0.9–1) and sensitivity and specificity of 95% and 69%, respectively. Together, these results support the continued development of TB-TASA as a potential tool for diagnosing TB and monitoring treatment responses.

Shareable abstract

Mycobacterium-tuberculosis-specific T-cell activation is a robust biomarker for tuberculosis disease and can be measured in blood collected by finger prick https://bit.ly/4mhsRx2

Introduction

Robust non-sputum-based diagnostic tests (e.g., blood, urine or oral swabs) are needed to improve the diagnosis of tuberculosis (TB) among people unable to produce sputum, with paucibacillary TB, or with extrapulmonary TB. A diagnostic test that can also monitor treatment response and predict treatment failure would also be advantageous. Furthermore, discrimination between Mycobacterium tuberculosis (M.tb)-exposed persons who have cleared or controlled infection and those recently infected with M.tb or persons with asymptomatic disease would allow targeted TB preventive treatment (TPT).

Activation of M.tb-specific T-cells, measured by the expression of HLA-DR, CD38 or Ki-67, is a robust candidate biomarker for TB. Many studies, conducted in populations from distinct geographies, demonstrate that regardless of the type of TB disease (pulmonary versus extrapulmonary), M.tb-specific T-cell activation distinguishes persons with clinical TB from healthy controls with evidence of immune sensitisation to M.tb, measured by interferon-γ (IFN-γ) release assays (IGRA) [1–9]. More importantly, levels of HLA-DR on M.tb-specific CD4⁺ T-cells discriminated symptomatic TB patients from those who had other respiratory diseases, with receiver operating characteristic (ROC) area under the curve (AUC) of 0.89 (95% CI 0.83–0.95) [1]. Additionally, HIV infection does not appear to impact the diagnostic performance of this biomarker, despite the fact that bulk T-cells are more activated in persons living with HIV, especially if viral replication is not controlled, compared to HIV-uninfected individuals [4, 8, 9]. We have shown that this biomarker may be useful to identify persons recently infected (i.e., converted to IGRA⁺ within 6 months) [7]. We have also shown that IGRA⁺ adolescents who progressed to TB have elevated activation of antigen-specific T-cells >1 year before TB diagnosis compared to healthy controls, yielding an AUC of 0.75 (95% CI 0.63–0.87) [7]. During antibiotic treatment, M.tb-specific T-cell activation decreases and may therefore allow monitoring of treatment response [3, 6, 8]. We have termed the method to measure this biomarker the TB T-cell Antigen-Specific Activation (TB-TASA) assay.

In this study, we established and validated a TB-TASA positivity threshold to distinguish persons with TB from healthy controls and determine if T-cells specific for different M.tb antigens exhibit different activation profiles, such that TB-TASA results are influenced by the mycobacterial antigen used for in vitro stimulation. We also assessed the performance of TB-TASA as a TB treatment monitoring tool in the context of a treatment-shortening clinical trial and evaluated whether TB-TASA could predict the risk of TB relapse. Lastly, we investigated the performance of TB-TASA using capillary blood samples collected via fingerprick, a more scalable and cheaper sample collection method.

Methods

Study design and study participants

This study included a subset of adults diagnosed with drug-sensitive TB who were participants in a large treatment-shortening trial (ClinicalTrials.gov: NCT02821832, PredictTB), enrolled at the Worcester study site in South Africa (figure 1). At enrolment, participants eligible for treatment shortening were identified based on baseline positron emission tomography/computed tomography criteria, or Xpert MTB/Rif Ct value at week 16 [10]. Participants not eligible for treatment shortening at baseline were enrolled into a standard-of-care arm that completed 24 weeks of treatment (Arm A). Participants eligible for treatment shortening completed the intensive phase of standard-of-care treatment and 8 weeks of continuation phase and were then randomised at week 16 of treatment into a standard-of-care arm that completed 24 weeks of treatment (Arm B) or concluded treatment at week 16 (Arm C). All participants were followed up for 18 months from enrolment to assess for potential TB recurrence. To distinguish recurrence due to relapse or re-infection, DNA strain typing was performed to determine whether the same or a different strain was obtained at recurrence [10]. Whole blood samples for the TB-TASA from TB patients enrolled in the PredictTB study at the Worcester site were collected using a convenience sampling approach. No formal power calculation was performed; however, we observed an effect size of 2.85 between TB patients and healthy IGRA controls from a prior dataset [3]. Therefore, we were confident that with a sample size of 10 or greater, the study would be sufficiently powered (<0.95) to detect a significant difference between healthy controls and TB patients and before and after TB treatment.

FIGURE 1.

PredictTB TB-TASA sub-study design. Substudy of tuberculosis (TB) patients enrolled at the South African Tuberculosis Vaccine Initiative (SATVI) Worcester field site into the Predict TB trial [10], from whom whole blood was collected and processed to measure antigen-specific T-cell activation. Participants not eligible for treatment shortening were enrolled into Arm A and received the standard 6-month treatment. Participants who met the treatment-shortening criteria at baseline, week 4 and week 16 were randomised into a standard 6-month treatment arm (Arm B) or an early treatment completion arm (Arm C) at week 16 of treatment. Participants who failed to meet the criteria were allocated to Arm A. The number of TB-TASA performed is indicated in bold; note that not all participants had a TB-TASA performed. M.tb: Mycobacterium tuberculosis; Rx: prescription; FDG: 2-deoxy-2-[¹⁸F]-fluoroglucose; ALT: alanine transaminase; PET: positron emission tomography; CT: computed tomography.

The PredictTB protocol was approved by the South African Health Products Regulatory Authority, University of Cape Town (UCT) Human Research Ethics Committee (HREC 645/2016), the IRB/ethics committees of the NIAID, Stellenbosch University, South African Health Products Regulatory Authority (SAHPRA), Henan Provincial Chest Hospital and the Henan Center for Disease Control. Lab samples from an independent group of healthy IGRA-positive adults (adults aged ≥18 years) from Worcester, South Africa, under a healthy donor protocol approved by the UCT HREC (126/2006), were included as a control group. Matched capillary fingerprick and venous whole blood samples were collected from a separate case–control cohort of healthy IGRA-negative and -positive adult controls, and adult TB patients (UCT HREC 248/2022). All healthy controls were recruited from the community and did not display clinical symptoms of TB. Participants with any acute or chronic disease at the time of screening were excluded. CXR or induced sputum were not performed on healthy participants. All participants provided written, informed consent.

TASA sample processing

Whole blood was collected in sodium heparin tubes and processed according to a previously published 12-h whole blood intracellular cytokine staining (ICS) assay protocol [11]. Blood was stimulated with PBS (negative control) or a pool of 15mer peptides overlapping by 10 amino acids spanning either Rv0288 (TB10.4), Rv1813c (hypothetical protein Rv1813c), Rv1886c (Ag85b), Rv3615c (EspC), Rv3619c (EsxV), Rv3620c (EsxW), Rv3874 (CFP-10), Rv3875 (ESAT-6) (2 μg·mL⁻¹/peptide, GenScript), M.tb-lysate (10 μg·mL⁻¹, BEI Resources), or Phytohaemagglutinin (PHA, positive control; 5 μg·mL⁻¹, Thermo Scientific) for 12 h at 37 °C. Brefeldin A (10 μg·mL⁻¹, Sigma Aldrich) was added to all samples for the final 5 h.

Stimulation of venous or capillary whole blood collected by fingerprick was performed using a modified 6-h whole blood ICS assay [11]. 450 μL of heparinised venous blood or 100 μL of heparinised capillary blood was stimulated with PBS, a pooled 15mer peptides overlapping by 10 amino acids spanning ESAT-6, CFP-10 and EspC (2 μg·mL⁻¹/peptide, GenScript), or PHA (5 μg·mL⁻¹, Thermo Scientific) for 6 h at 37 °C in the presence of Brefeldin A (10 μg·mL⁻¹, Sigma Aldrich) and the co-stimulatory antibodies anti-CD49 (1 μg·mL⁻¹, BD) and anti-CD28 (1 μg·mL⁻¹, BD). Following stimulation, BD FACS lysing solution (BD BioSciences) was added to lyse red blood cells and fix leukocytes, which were cryopreserved in a cryosolution (50% RPMI, 40% fetal bovine serum and 10% dimethyl sulfoxide) and stored at −80 °C for subsequent flow cytometry analysis.

Flow cytometry acquisition

Cells were thawed in PBS and permeabilised using BD Perm/Wash, and stained with the antibodies in supplementary table S1 (12-h ICS) and supplementary table S2 (6-h ICS). Stained cells and compensation controls, prepared for each acquisition run, were acquired on a Becton, Dickinson, and Company (BD) LSR Fortessa flow cytometer.

Statistical analyses

Gating of flow cytometry data was performed using an automated analysis pipeline developed with openCyto in R (4.1.1) [12]. A compensation matrix was generated using compensation controls in FlowJo, and the matrix was extracted in R from the corresponding .wsp file. The pipeline script to perform automated gating of the TB-TASA can be found at https://github.com/SATVILab/DataTidyMusvosviPredictTBTASA. Gating strategies were applied using openCyto gating templates (supplementary tables S3–S7). Data were reviewed and gates were manually adjusted if needed. The Fisher's exact test (p<0.01) was used to define responders. A 2×2 table was constructed using the number of IFN-γ⁺ tumour necrosis factor (TNF)⁺ T-cells and the number of T-cells not co-expressing IFN-γ and TNF from the stimulated condition and the negative control (PBS).

Results

Choice of mycobacterial antigen appears not to impact performance of the TB-TASA assay

We enrolled 78 TB patients; however, five participants were withdrawn before the baseline blood sampling (figure 1). The TB-TASA test was not performed on all participants because the main study, PredictTB, had already commenced before TB-TASA was implemented and due to budgetary constraints. In addition to the TB patients, we enrolled a total of 9 IGRA⁺ healthy controls. We first determined whether stimulation with different mycobacterial antigens, to identify M.tb-specific T-cells based on cytokine expression, influenced the performance of the TB-TASA assay. To achieve this, we selected eight proteins that are likely expressed at different levels by M.tb, based on mRNA transcript abundance in the lungs of M.tb-infected mice [13], and are antigens in subunit TB vaccines or routinely used in IGRA (supplementary figure S1). Antigen-specific T-cells were defined as CD3⁺ co-expressing IFN-γ and TNF following antigen stimulation. Frequencies of antigen-specific T-cells were comparable between healthy IGRA⁺ controls and TB patients for six of the eight antigens; frequencies of ESAT-6- and EspC-specific cells were higher in TB patients than controls (figure 2). A prerequisite to determine a TB-TASA score is the presence of sufficient antigen-specific T-cells to reliably measure expression levels of HLA-DR. We defined the TB-TASA positivity threshold using M.tb-lysate-specific T-cells rather than individual antigen-specific T-cells, because M.tb lysate stimulated cytokine-expressing T-cell responses in almost all participants (figures 2c and 3a), providing the most informative dataset. ROC analysis showed excellent discrimination between TB patients and controls, with an AUC of 0.99 (95% CI 0.98–1; figure 3b). The cutpointr function was used to determine a threshold that provides the maximum number of correctly classified samples. A threshold of 11.53% provided maximum accuracy; however, for practical implementation and simplicity the threshold was rounded down to 10%.

FIGURE 2.

Quantification of antigen-specific T-cell frequencies in healthy IGRA⁺ individuals and patients with bacteriologically confirmed tuberculosis (TB) disease. a) Representative flow cytometry dot plots depicting the gating strategy used to identify antigen-specific T-cells using openCyto. First, a boundary gate was used to exclude FSC-A and FSC-H outliers, followed by a singlet gate to exclude doublets. Side scatter high and side scatter low cells were then gated. The lymphocyte population was gated from the side scatter low population. The threshold to define HLA-DR⁺ cells was then determined based on HLA-DR expression differences between different lymphocyte populations (i.e., T-cells versus B-cells; B-cell markers not included). CD3⁺ cells were then gated from the lymphocyte population and CD3⁺ T-cells co-expressing interferon-γ (IFN-γ) and tumour necrosis factor (TNF) were identified. The data shown are from a negative control (i.e., PBS stimulation) sample. b) Representative flow cytometry dot plots showing the CD3⁺ T-cells co-expressing IFN-γ and TNF in whole blood following stimulation with the different antigens. c) Box-and-whisker plot depicting the background subtracted frequencies of antigen-specific T-cells in healthy IGRA⁺ adults and TB patients at the time of diagnosis. The horizontal lines represent medians, the bounds of the boxes indicate the 25th and 75th percentiles and the whiskers represent the minimum and maximum. Below is a stacked bar plot depicting the number of study participants with (red bar) or without (blue bar) a T-cell response to the corresponding mycobacterial antigen. The Fishers’ exact test was used to define responders (p<0.01). The Mann–Whitney U-test was used to compare frequencies of antigen-specific T-cells in healthy IGRA⁺ adults and TB patients. M.tb: Mycobacterium tuberculosis.

FIGURE 3.

A TB-TASA threshold of 10% proportion of HLA-DR⁺ cells among cytokine⁺ CD4 T-cells is a robust threshold to distinguish IGRA⁺ from tuberculosis (TB) patients. a) Box-and-whisker plot showing the proportions of HLA-DR⁺ Mycobacterium tuberculosis (M.tb) whole cell lysate-specific T-cells in healthy IGRA⁺ adults and TB patients at diagnosis. Medians are indicated by horizontal lines, with boxes representing the interquartile range (25th and 75th percentiles) and whiskers showing the full data range. The Fisher's exact test (p<0.01) was used to define responders, and the dashed horizontal line marks the 10% TB-TASA positivity threshold. The Mann–Whitney U-test was used to compare groups. b) Receiver operating characteristic (ROC) curve depicting the performance of the TB-TASA assay. The dot indicates the sensitivity and specificity at the 10% threshold. c–e) Box-and-whisker plots of HLA-DR⁺ antigen-specific T-cells in IGRA⁺ adults and TB patients from published studies (re-analysed using an automated gating pipeline with the openCyto R package). The 10% TB-TASA threshold was applied to assess sensitivity and specificity. Statistical details and markers are consistent with those described in panel a. f) ROC curve showing TB-TASA assay performance across the three studies, with the dot indicating sensitivity and specificity at the 10% threshold. AUC: area under the ROC curve. The shaded area depicts AUC 95% confidence intervals. The dotted box in the top left corner of AUC plots depicts the minimum sensitivity (90%) and specificity (70%) criteria for a TB triage test defined in the World Health Organization (WHO) target product profile [22]. IGRA: interferon-γ release assay.

We then assessed the robustness of the 10% TB-TASA positivity threshold in multiple independent TB-TASA datasets. We adapted the automated hierarchical gating pipeline to re-analyse data from previously published studies that assessed HLA-DR⁺ antigen-specific T-cells in whole blood or peripheral blood mononuclear cells from TB patients and healthy controls [3, 7, 8]. We observed sensitivities and specificities of 91% and 86% for the Musvosvi et al. dataset [3] (figure 3c), 93% and 84% for the Riou et al. dataset [8] (figure 3d), and 89% and 83% for the Mpande et al. dataset [7] (figure 3e) in samples with detectable antigen-specific T-cell responses (figure 3f). It is important to note that a detectable antigen-specific T-cell response, defined by a Fisher's exact test (p<0.01), is required to measure this biomarker; therefore measurement of the TB-TASA score was not possible in up to 12% of samples in some datasets (figure 3c).

Next, we determined if the TB-TASA assay performance was robust to in vitro stimulation with different M.tb antigens. Significantly higher proportions of HLA-DR⁺ antigen-specific T-cells were observed from TB patients relative to healthy IGRA⁺ adults, irrespective of antigen (figure 4a). High AUC values (>0.91) were consistently obtained across the eight antigens, suggesting that HLA-DR expression is a consistent T-cell activation biomarker regardless of M.tb antigen (figure 4b). This trend was even observed for T-cells targeting the hypothetical protein Rv1813c, which were detectable in only one control and six TB patients.

FIGURE 4.

Effect of in vitro stimulation with different mycobacterial antigens on the performance of the TB-TASA assay. a) Box-and-whisker plot depicting the proportions of HLA-DR⁺ cells among those with positive antigen-specific T-cell responses (responders) in healthy IGRA⁺ adults and bacteriologically confirmed tuberculosis (TB) patients at the time of diagnosis. The number of responders is indicated below each plot. The horizontal lines represent medians, the bounds of the boxes indicate the 25th and 75th percentiles and the whiskers represent the minimum and maximum. The Mann–Whitney U-test was used to compare proportions of HLA-DR⁺ antigen-specific T-cells in healthy IGRA⁺ adults and TB patients. The dotted line indicates the TB-TASA threshold of 10%. b) Receiver operating characteristic (ROC) curves showing the performance of the TB-TASA assay for the different antigen-specific T-cell populations. The shaded area depicts area under the curve (AUC) 95% confidence intervals. The dot indicates the sensitivity and specificity achieved using a TB-TASA threshold of 10%. The dotted box in the top left corner of AUC plots depicts the minimum sensitivity (90%) and specificity (70%) criteria for a TB triage test defined in the World Health Organization (WHO) target product profile [22]. IGRA: interferon-γ release assay.

TB-TASA levels are similar in patients with severe and less severe TB

Since assignment to Arms B and C was performed at week 16, TB-TASA scores measured at TB diagnosis, before treatment initiation, were compared in TB patients who were either allocated to Arm A or Arms B/C. TB-TASA scores in TB patients with severe (Arm A) or less severe TB (Arms B/C) were not significantly different before TB treatment (figure 5a) [14].

FIGURE 5.

The TB-TASA identifies tuberculosis (TB) patients with poor treatment outcomes at the end of treatment. a) Box-and-whisker plots showing the proportions of HLA-DR⁺ M.tb whole cell lysate-specific T-cells in healthy IGRA⁺ adults and TB patients in the Predict TB trial with more severe disease at TB diagnosis (Arm A) or less severe disease at TB diagnosis (Arm B/C). The horizontal lines represent medians, the bounds of the boxes indicate the 25th and 75th percentiles and the whiskers represent the minimum and maximum. The Fisher's exact test was used to define responders (p<0.01). The dashed line represents the threshold of a positive TB-TASA assay. The Mann–Whitney U-test was used to compare the proportion of HLA-DR⁺ antigen-specific T-cells in healthy IGRA⁺ adults, TB patients in Arm A and TB patients in Arm B/C. Plots showing longitudinal trajectories of proportions of HLA-DR⁺ M.tb-whole cell lysate-specific T-cells during TB treatment in PredictTB participants enrolled into b) Arm A, c) Arm B or d) Arm C. Patients in Arm A and B completed treatment at week 24 and patients in Arm C completed treatment at week 16. Samples collected at the end of treatment or during study follow-up have been shaded grey. The Wilcoxon matched-pairs test was used to compare the proportions of HLA-DR⁺ M.tb-whole cell lysate-specific T-cells during treatment. e) A box-and-whisker plot showing the proportions of HLA-DR⁺ M.tb whole cell lysate-specific T-cells at the end of treatment timepoint in TB patients allocated into Arm A (week 24), Arm B (week 24) or Arm C (week 16) of the PredictTB trial, stratified by outcome. f) ROC curves showing the performance of the TB-TASA at the end of treatment (week 16) to distinguish patients in Arm C with unfavourable outcomes from those patients with successful treatment. The shaded area depicts AUC 95% confidence intervals. IGRA: interferon-γ release assay; Tx: Treatment

Performance of TB-TASA to predict unfavourable treatment outcomes

We then assessed the kinetics of mycobacteria-specific T-cell activation during TB treatment and determined if the TB-TASA score could predict TB unfavourable outcomes (i.e. relapse or treatment failure). TB-TASA scores decreased significantly during treatment, regardless of the study arm (figure 5c–e). We then assessed whether persons who experienced relapse during study follow-up could be identified based on the TB-TASA score at the end of treatment. A total of five participants had confirmed TB relapse in our sub-study cohort, one in Arm A and four in Arm C. Of the four participants who had confirmed TB relapse in Arm C, three had whole blood samples collected at treatment completion (i.e., week 16). There were two participants in Arm C who failed treatment. Therefore, a total of five participants from Arm C with unfavourable outcomes had TB-TASA samples at treatment completion. We observed that participants who had unfavourable outcomes (i.e., relapse or treatment failure) in Arm C had higher TB-TASA scores at the end of treatment (week 16) compared to Arm C participants successfully treated (p=0.03; ROC AUC: 0.89, 95% CI: 0.69–1; figure 5e–f). We did not observe differences in TB-TASA scores at the end of treatment in participants successfully treated between Arms A, B and C (figure 5f). Whole blood samples were not collected from the participant in Arm A who experienced TB relapse, as this participant was enrolled after the TB-TASA sub-study had been completed.

Developing fingerprick-based TB-TASA

Lastly, we wished to determine if TB-TASA could be performed on capillary blood collected by fingerprick and whether the antigen stimulation time could be reduced. Whole blood samples collected via venepuncture or fingerprick from an independent cohort of healthy IGRA⁻ or IGRA⁺ persons and bacteriologically confirmed TB patients were stimulated for 6 h with a pool of 15-mer peptides spanning CFP-10, ESAT-6 and EspC. The frequencies of antigen-specific T-cells in venous and capillary blood were highly correlated with a concordance correlation coefficient of 0.87 (95% CI: 0.84–0.89) (supplementary figure S2). IGRA⁻ individuals had lower frequencies of CFP-10, ESAT-6 and EspC-specific T-cells compared to IGRA⁺ or TB patients (figure 6a–b). TB patients had higher frequencies of CFP-10, ESAT-6 and EspC-specific T-cells compared to healthy IGRA⁺ persons. The TB-TASA scores were significantly higher in TB patients compared to healthy IGRA⁺ persons and AUC values were >0.90, regardless of blood sample source or volume (figure 6c–f). Using the 10% TB-TASA positivity threshold, the venous blood-based assay achieved a sensitivity of 91% (95% CI: 71–99) and a specificity of 87% (95% CI: 66–96), while the capillary blood sample-based assay achieved a sensitivity of 95% (95% CI: 77–100) and a specificity of 69% (95% CI: 32–77) (figure 6c–d). Lastly, the 100 μL venous blood sample-based assay achieved a sensitivity of 91% (95% CI: 76–100) and a specificity of 80% (95% CI: 41–87) (figure 6f).

FIGURE 6.

The TASA assay performs similarly to distinguish healthy individuals from tuberculosis (TB) patients, using venous or capillary blood samples. Boxplots depict frequencies of M.tb-specific T-cells across three groups: healthy IGRA⁻ participants, healthy IGRA⁺ participants and TB patients, measured using a) 450 μL of venous blood or b) 100 μL of capillary blood collected by fingerprick. The horizontal lines represent median values, with boxes indicating the 25th and 75th percentiles. Frequencies were analysed using the Mann–Whitney U-test. For better visualisation, a log10 scale is used on the y-axes, with values below 0.001% adjusted to 0.001%. Boxplots showing the proportion of HLA-DR⁺ ESAT6/CFP10/EspC-specific T-cells measured in healthy IGRA⁻ participants, healthy IGRA⁺ participants or TB patients using c) 450 μL of venous blood, d) 100 μL of capillary blood collected via fingerprick or e) 100 μL of venous blood. The horizontal lines indicate the median, and the boxes indicate the 25th and 75th percentiles. The Fishers’ exact test was used to define responses to ESAT6/CFP10/EspC (flow responder). The Mann–Whitney U-test was used to compare the frequencies of ESAT6/CFP10/EspC-specific T-cells between study groups. f) A plot of receiver operating characteristic (ROC) curves showing the performance of the TASA biomarker to distinguish healthy IGRA⁺ participants and TB patients. The shaded area depicts area under the curve (AUC) 95% confidence intervals. The dots indicate the sensitivity and specificity achieved using a TASA positive threshold of >10%. The dotted box in the top left corner of AUC plots depicts the minimum sensitivity (90%) and specificity (70%) criteria for a TB triage test defined in the World Health Organization (WHO) target product profile [22]. IGRA: interferon-γ release assay.

Discussion

Mycobacteria-specific T-cell activation, measured in this study with the TB-TASA test, was previously shown to be a robust biomarker of recent M.tb infection [7], disease progression [7] and TB disease [1–9]. In our study, we found that TB-TASA scores differentiate between TB patients and IGRA⁺ controls, regardless of mycobacterial antigen specificity. Using a novel computational flow cytometry analysis pipeline, we found that a TB-TASA positivity threshold of 10% HLA-DR⁺ mycobacteria-specific T-cells was robust across multiple, previously published studies and different sample types, differences in antigen stimulation duration, and flow cytometry antibody panels. We also observed promising utility of TB-TASA as a treatment monitoring tool. Finally, TB-TASA performed well on capillary blood collected via fingerprick. These findings support the continued development of the TB-TASA as a candidate biomarker.

Consistent with previous studies assessing mycobacteria-specific T-cell activation using multiple mycobacterial antigen preparations [3, 6, 7], our results demonstrate that HLA-DR as a biomarker was not markedly affected by the choice of mycobacterial antigen for in vitro stimulation. However, should a participant not have detectable T-cell responses to the M.tb-specific antigens, determining antigen-specific T-cell activation is not possible. Some antigens such as ESAT-6, CFP-10, EspC and TB10.4 were recognised by more individuals compared to other antigens and thus allow flexibility in selecting antigen preparations depending on the context. For example, stimulation of blood with M.tb-specific antigens such as ESAT-6, CFP-10 and EspC may distinguish persons with prior M.tb sensitisation from those vaccinated with Bacille Calmette-Guerin (BCG), a live attenuated vaccine derived from M. bovis, but not exposed to M.tb. Indeed, stimulation of venous blood with a pool of 15mer peptides spanning ESAT-6, CFP-10 and EspC resulted in detectable IFN-γ⁺ TNF⁺ T-cell responses in all TB patients (22 out of 22) compared to 88% of TB patients (22 out of 25) when blood was stimulated with ESAT-6 and CFP-10. It is important to note that the TB patients were from different cohorts, and this was not a head-to-head comparison; however, these data suggest that expanding the M.tb antigens beyond ESAT-6 and CFP-10 would increase the responder rate. Furthermore, antigen-specific T-cell responder rates could be increased by utilising peptide pools containing other mycobacterial antigens, such as the “TB-megapool” [14] or the “ATB116” [15], which comprises many dozens of antigens, or inclusion of the highly immunodominant antigen TB10.4, may be useful to maximise detection of mycobacteria-specific T-cell responses and determining activation, even in individuals without M.tb sensitisation. The choice of antigen used to identify the cognate mycobacteria-specific T-cell population may thus be tailored to the biomarker application. We propose that a peptide pool spanning ESAT-6, CFP-10 and EspC may be ideal to detect prior M.tb infection, while maximising the number of responders.

In our study, responders were defined by the presence of detectable antigen-specific T-cells co-expressing IFN-γ and TNF. We previously demonstrated that defining response on the basis of co-expressing two cytokines resulted in less nonspecific binding [3]. Interestingly, in the current study, we observed that the frequency of ESAT-6/CFP-10/EspC-specific co-expressing IFN-γ and TNF was higher in TB patients compared to healthy IGRA⁺ controls. This was consistent with other studies that observed higher frequencies of antigen-specific CD4 T-cells co-expressing IFN-γ and TNF in TB patients compared to controls, regardless of HIV infection status [16, 17]. It should be noted that the IFN-γ⁺ TNF⁺ subset is associated with higher T-cell differentiation [18]. Therefore, the higher frequency of the IFN-γ⁺ TNF⁺ subset is congruous with an increased in vivo T-cell activation during TB disease.

We also assessed TB-TASA as a TB treatment monitoring test in the PredictTB trial, which assessed treatment shortening in patients with nonsevere TB [10]. The majority of patients achieved relapse-free cure following 4 months of treatment of drug-sensitive TB disease. However, 15–20% of patients can be at risk of relapse [10, 19], underpinning the 6-month duration of standard TB treatment used for decades. In the PredictTB trial, 12.1% of patients with nonsevere TB who received 4 months of therapy had an unfavourable outcome compared to 1.5% of patients with nonsevere TB who completed 6 months of treatment [20]. A biomarker that can stratify persons into low- or high-relapse risk categories following 4 months of treatment would be highly valuable, identifying patients at low risk of relapse and sparing them from unnecessary drug exposure. Previously, Riou et al. [8] demonstrated that TB patients cured after standard 6-month treatment had a larger fold reduction in M.tb-specific T-cell activation compared to patients who were not cured. Here, we observed that in the treatment-shortening study arm, participants who relapsed or failed treatment had higher TB-TASA scores compared to cured patients. We acknowledge that our sample size is small, with only a limited number of relapse cases. Confirmation in larger treatment-shortening studies will be crucial to establish whether the TB-TASA biomarker will be useful for stratifying persons eligible for shorter TB treatment from persons who require standard duration treatment.

We also provided evidence that the TB-TASA could be performed using capillary blood. Future studies are needed to validate these findings, as to our knowledge this is the first time M.tb-specific T-cell activation has been measured using very small volumes of capillary blood collected via fingerprick. In the future, the volume of fingerprick capillary blood should be optimised to minimise outcome variability associated with small volumes. In the current study, the sensitivity (95%) and specificity (69%) we observed were comparable to the sensitivity and specificity observed using the Cepheid 3 Xpert-HR [21]. Fingerprick blood collection is an easier blood collection method compared to venous puncture for use of TB-TASA as a screening test, for example, to identify TB disease versus latent TB in community settings. Additionally, capillary blood may be a more suitable sample type in infants and children.

Lastly, we provide evidence supporting the continued development of the TB-TASA biomarker to diagnose and monitor treatment. While promising, we acknowledge our study had limitations. Healthy IGRA⁺ persons are not the ideal control group to assess the diagnostic performance of the TB-TASA in symptomatic TB patients. While many studies have relied on healthy IGRA⁺ persons as a control group, symptomatic persons with other respiratory diseases would be another important control group, depending upon the application intended. Encouragingly, a study by Halliday et al. [1] revealed lower activation of M.tb-specific T-cells in symptomatic persons with other respiratory diseases compared to TB patients. While our present study did not include persons with HIV, others have demonstrated higher levels of M.tb-specific T-cell activation in TB patients with HIV compared to IGRA⁺ persons with HIV [4, 8, 9].

Future work will be required to further simplify sample collection, sample processing and sample acquisition to move the TB-TASA closer to clinical translation. We are encouraged that our computational pipeline appears robust and that the TB-TASA could be measured using capillary blood collected via fingerprick.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary material

01016-2025.SUPPLEMENT

Supplementary material

01016-2025.SUPPLEMENT

Supplementary material

01016-2025.SUPPLEMENT2

Supplementary material

01016-2025.SUPPLEMENT2

Supplementary material

01016-2025.SUPPLEMENT3

Supplementary material

01016-2025.SUPPLEMENT4