Disease extent and anti‐tubercular treatment response correlates with Mycobacterium tuberculosis‐specific CD4 T‐cell phenotype regardless of HIV‐1 status

Abstract Objectives The development of non‐sputum‐based assays for tuberculosis (TB) diagnosis and treatment monitoring is a key priority. Recent data indicate that whole blood‐based assays to assess the phenotype of Mycobacterium tuberculosis (Mtb)‐specific CD4 T cells hold promise for this purpose and require further investigation in well‐characterised TB cohorts. In this study, we investigated the relationship between the phenotypic signature of Mtb‐specific CD4 responses, TB disease extent and treatment response. Methods Using flow cytometry, we measured the expression of phenotypic and functional markers (HLA‐DR, CD27, CD153, KLRG1, IL‐2, MIP‐1β, TNF‐α and IFN‐γ) on Mtb‐specific CD4 T‐cells in whole blood from 161 participants of varying TB and HIV status. TB disease extent was graded as a continuum using the Xpertct value, C‐reactive protein, Timika radiographic score and monocyte/lymphocyte ratio. Results The phenotypic profile of Mtb‐specific CD4 T cells pre‐anti‐tubercular treatment (ATT) strongly correlated with disease extent, irrespective of HIV status. ATT associated with major changes in the phenotype of Mtb‐specific CD4 T cells, with decreased expression of HLA‐DR and increased CD27 and CD153 expression. Principal component analysis showed an almost complete separation between latent TB infection (LTBI) and active TB (aTB) pre‐ATT groups, whereas the profile of the aTB post‐ATT group overlapped with the LTBI group. However, in patients experiencing treatment failure or relapse, no significant changes were observed in Mtb‐specific CD4 T‐cell phenotype pre‐ and post‐ATT. Conclusion Whole blood‐based assays of Mtb‐specific CD4 T‐cell activation and maturation markers can be used as non‐sputum‐based biomarkers of disease extent and treatment monitoring in TB, regardless of HIV‐1 status.


INTRODUCTION
Globally, 10 million people were diagnosed with tuberculosis (TB) and 1.45 million people died as a result of this infection in 2018. 1 TB remains the leading cause of death in HIV-1-infected people where sputum smear-negative, extra-pulmonary disease is frequent and represents a particular challenge in diagnosis and treatment monitoring. 2 Although significant strides have recently been made in TB vaccine development, 3,4 there is still no licensed vaccine to prevent TB in adults. The treatment for drug susceptible TB consists of a 6month long regimen of drugs that can cause adverse effects and drug-drug interactions, particularly in the context of concomitant antiretroviral therapy (ART). 5 Thus, the development of new shortened TB regimens has been defined as a major goal by WHO 6 to improve patient adherence, reduce transmission and reduce costs to patients and health systems. Different approaches are being assessed in ongoing clinical trials including increased doses of existing drugs, adjunctive host-directed therapies and the use of new and repurposed drugs in anti-TB treatment (ATT) regimens. 7,8 Sputum culture for Mycobacterium tuberculosis (Mtb), the current standard measure of monitoring treatment response, has a long turnaround time taking up to 42 days to yield results, and PCR-based methods such as GeneXpert are not recommended to monitor treatment response owing to detection of DNA from dead bacteria. 9 Thus, there is an urgent need for novel biomarkers that assess clinical treatment response both accurately and more rapidly than conventional methods such as sputum culture to aid clinical decision-making in the programmatic setting and assist with the monitoring of ATT response in clinical trials.
There is evidence from studies using PET-CT imaging that some TB patients may exhibit ongoing lung inflammation, suggestive of active TB (aTB) disease, accompanied by detectable Mtb mRNA in sputum and bronchoalveolar lavage samples after completing 6 months of ATT and up to 1 year of follow-up. 10 Thus, while may be possible to shorten ATT regimens for some, it may be necessary to prolong ATT in others to prevent relapse or recurrence of drug sensitive TB. A recent large population-based cohort in South Africa showed high TB recurrence rates, 11 in line with previous reports from high prevalence settings. 12,13 The cause of high rates of TB recurrence is likely to be multifactorial; 11 however, strategies are required to identify and track those at highest risk. It has been shown that CD4 T-cell activation status denoted by the expression of the MHC class II cell surface receptor HLA-DR is indicative of TB risk and could potentially be investigated for this purpose. 14 Furthermore, several studies have highlighted the potential of blood-based assays as diagnostic tools; and while measuring the magnitude of Mtb-specific T-cell response does not offer much value for TB diagnostic 15,16 or treatment monitoring, 17 assessing the phenotypic profile of these cells showed promising results. Indeed, the activation, memory differentiation or functional profile of Mtb-specific CD4 T cells has been shown to relate to TB disease activity 15,[18][19][20][21][22][23][24][25][26][27] and evaluate treatment response. 18,19,28,29 However, very few of these studies have included HIV-infected participants. Thus, in this study, we set out to better understand the relationship between the phenotypic attributes of Mtb-specific CD4 T cells, TB disease severity and treatment response. We comprehensively investigated the expression of multiple promising phenotypic markers on Mtb-specific CD4 T cells using flow cytometry measured in whole blood samples from a well-characterised cohort of HIV-infected and HIVuninfected persons with latent TB infection (LTBI) and aTB disease before and after ATT completion.

Study population
The clinical characteristics of participants are presented in Table 1. Participants (n = 161) were classified into four groups according to their HIV-1 and TB status: LTBI/HIV À (n = 35), LTBI/HIV + (n = 31), aTB/HIV À (n = 41), and aTB/HIV + (n = 54). Median age was comparable between the four groups. IGRA values were comparable between the HIV-uninfected and HIV-infected LTBI groups (P = 0.14). The Xpert MTB/RIF cycle threshold (Xpert ct ) was higher in HIV-infected than in HIVuninfected patients (median: 23.3 vs 18.5, respectively, P = 0.001). HIV infection status did not significantly affect the level of plasma C-reactive protein (CRP) in aTB patients (medians: 72 µg mL À1 in HIV + and 100 µg mL À1 in HIV À , P = 0.99). Finally, HIV-infected participants with LTBI had a significantly lower plasma HIV-1 viral load (VL) and higher absolute CD4 count than the HIV-infected aTB group (median VL: < 20 vs 43 415 copies mL À1 , P = 0.0003 and median CD4: 490 vs 206 cells mm À3 , P < 0.0001, respectively). These differences are due to higher ART usage in the LTBI group than in the aTB group (80.6% vs 38.9%, respectively, P = 0.0002). Follow-up blood samples were available for 28 HIV-uninfected participants and 33 for HIV-infected patients after the completion of standard ATT regimen at week 24.

Measures of TB disease extent
Five different measures of TB disease extent were assessed at the time of ATT initiation: sputum Xpert ct values, sputum Mtb culture time to positivity (TTP), Timika radiographic severity score, serum CRP levels and peripheral blood monocyte to lymphocyte ratio (M/L ratio). Xpert ct values and culture TTP were significantly lower in HIVuninfected patients than in those infected with HIV (P = 0.001 and P = 0.003, respectively), while the Timika score was significantly higher in HIVuninfected participants (P = 0.011, Supplementary figure 1a). This can be ascribed to the fact that cavitatory disease is more frequent in HIVuninfected persons. In the aTB group, no differences were observed in plasma CRP concentration or the M/L ratio between HIVinfected and HIV-uninfected participants. However, CRP and the M/L ratio were higher in those with aTB than those with LTBI, regardless of HIV status (P < 0.0001 for all sub-group comparisons, Supplementary figure 1a). All measures of disease extent were assessed for the strength of their correlation with each other by Spearman correlation rank (r) values and further stratified by HIV status (Supplementary figure 1a and c). There was significant correlation between all the measures regardless of HIV status, with Xpert ct value vs culture TTP, Xpert ct value vs Timika score and CRP vs M/L ratio showing the strongest correlations overall (P < 0.0001).
Hierarchical cluster analysis of markers of disease extent revealed two major subgroups of aTB patients: cluster 1 being associated with a phenotype associated with evidence of greater pulmonary and systemic disease extent (lower Xpert ct values and higher CRP, monocyte to lymphocyte ratio and Timika score) and dominated by HIV-uninfected patients (57%), whereas the other phenotype (cluster 2), characterised by lower sputum bacterial load, systemic inflammation and radiographic disease extent (higher Xpert ct values and lower CRP, M/L ratio and Timika score), was dominated by HIVinfected patients (63%) (Figure 1a). Furthermore, principal component analysis showed that the principal component 1 (PC1) accounted for 52.6% of the observed variance in measures of disease severity in the aTB cohort ( Figure 1b), with significantly higher PC1 loading scores being observed in HIV-uninfected than in HIV-infected participants (P = 0.018, Figure 1c). Sputum culture TTP, a variable not included in the hierarchical cluster and principal component analysis (PCA) to avoid collinearity with sputum Xpert ct values, progressively prolonged with duration of ATT regardless of HIV-1 status, with most participants showing culture conversion by week 8 of treatment (Figure 1d and e). In keeping with the decline in sputum bacterial load, there was a significant decline in plasma CRP concentration, M/L ratio and Timika score with increasing duration of ATT in both HIV-infected and HIVuninfected patients (Supplementary figure 2). Those with higher PC1 loading scores (e.g. greater pulmonary and systemic disease extent) at baseline tended to show slower bacterial clearance with later conversion time of their sputum culture during the course of ATT (≥ 8 weeks) compared to patients exhibiting low PC1 loading scores (Figure 1f and g).
The impact of ATT on the magnitude and polyfunctional profile of the Mtb-specific CD4 T-cell response First, we compared the frequency of the CD4 Tcell response to a pool of 300 Mtb-derived peptides (Mtb300) in the aTB group at baseline and at the end of ATT (week 24). At baseline, the   overall frequency of Mtb-specific CD4 T cells was comparable in HIV-uninfected and HIV-infected patients (medians: 0.47% and 0.57%, respectively, P = 0.15). By week 24 of ATT, there was a median reduction of 36% in the overall frequency of the Mtb-specific CD4 T-cell response in those who were HIV-uninfected, and a median reduction of 59% in HIV-infected persons (P < 0.0001, Figure 2a and b). Due to concomitant ART usage in HIV-infected patients, we took into account the variation in absolute CD4 + T-cell counts between baseline and W24, and the absolute number of Mtb-specific CD4 + T cells was compared between the two time points. Despite a significant increase in CD4 count in HIV-infected patients initiating ART (medians: 226 cells mm À3 at BL vs 335 cells mm À3 at W24, P = 0.014), the absolute number of Mtb-specific CD4 T cells remained approximately twofold lower at W24 compared to BL (median: 1.39 cells mm À3 vs 0.75 cells mm À3 , respectively; P = 0.001, Supplementary figure 3a and b). This contraction of Mtb-specific CD4 T cells could be related to the elimination of the activated effector antigen-specific T cells upon pathogen clearance. Moreover, in the context of HIV infection, suboptimal immune reconstitution, despite long-term ART and viral suppression, has been reported. 30 Thus, incomplete restoration of pathogen-specific T cells could also contribute to the more rapid contraction of Mtb-specific CD4 T cells observed in the HIV-infected group after ATT.
Furthermore, we quantified the polyfunctional potential of Mtb300-specific CD4 T cells based on their capacity to co-express IL-2, IFN-c, TNF-a or MIP-1b ( Figure 2c). While no significant difference was observed in the polyfunctional profile of Mtb300-specific CD4 T cells at baseline between HIV-uninfected and HIV-infected patients, completion of ATT associated with a significant change in the overall functional capacity of Mtbresponding CD4 cells in both groups (P = 0.001 and P = 0.0005, respectively). This was characterised by the contraction of IFN-c, TNF-a double-positive cells and IFN-c, TNF-a, MIP1b coproducing cells and counterbalanced by the expansion of IL-2 + TNFa + CD4 cells. Overall, ATT resulted in similar changes in the HIV-uninfected and HIV-infected groups at week 24, characterised by the expansion of IL-2 producing CD4 T cells, suggesting that ATT could improve cell proliferation capacity as previously described. 31 Phenotypic signature of Mtb-specific IFNc + CD4 T cells in LTBI and aTB at BL and week 24 To better understand the phenotypic attributes of Mtb-specific CD4 T cells that associate with TB disease extent, we compared the expression of HLA-DR (activation marker), CD27 (memory marker), IL-2 (functional potential), CD153 (TNF superfamily member, implicated in Mtb protection), KLRG1 (co-inhibitory molecule) and MIP-1b (chemokine, CCR5 agonist) in the LTBI group and the aTB cohort (at baseline and week 24; Figure 3a and b). The frequency of HLA-DR expressing Mtb300-specific IFNc + CD4 T cells significantly declined between baseline and week 24 (P < 0.0001), although frequencies at week 24 remained significantly higher than that observed in the LTBI groups (P < 0.0001), regardless of HIV status. Conversely, frequencies of CD27, IL-2 and CD153 expressing Mtb300-specific IFNc + CD4 T cells significantly increased after ATT in both HIVinfected and HIV-uninfected persons. In the HIVuninfected aTB group, CD27 expression after ATT was comparable to that of the LTBI group. Interestingly, this was not the case for HIVinfected individuals, where significantly lower expression of CD27 was seen at week 24 than for the LTBI group (P = 0.0015). This could be indicative of failure of the HIV co-infected subgroup to reconstitute the pool of central memory CD4 T cells after ATT. CD153 expression on Mtb300-specific IFNc + CD4 T cells was also only partially restored after ATT in HIV-uninfected persons, with a similar trend seen in HIV-infected patients. Additionally, no significant changes were observed in the expression of KLRG1 and MIP-1b in Mtb300-specific IFNc + CD4 T cells pre-and post-ATT in both groups. Overall, these results concur with previous reports showing that the phenotypic profile of Mtb-specific IFNc + CD4 T cells in aTB disease is associated with a highly differentiated and activated profile. 15,18,29 However, the Mtb-specific IFNc + CD4 T-cell phenotypic profile was only partially restored after 24 weeks of ATT.
To define the relationship between Mtb300specific IFNc + CD4 T-cell phenotype and TB disease extent, we next assessed the association of the PC1 loading score (described in Figure 1) with the expression of each measured marker. Strong correlations were observed between the Only response patterns contributing more than 1% to the overall Mtb responses are depicted. The proportion of each response pattern contributing to the total Mtb300-specific CD4 response per individual is shown. The median (black bar) and interquartile ranges (box) are shown. Each response pattern is colour-coded, and data are summarised in the pie charts, where each pie slice represents the median contribution of each response pattern to the total Mtb300 response. A Wilcoxon rank test was used to compare response patterns between groups (***P < 0.001, **P < 0.01, *P < 0.05). Statistical differences between pie charts were defined using a permutation test.
We next analysed the overall phenotypic signature of Mtb300-specific IFNc + CD4 T cells by conducting a hierarchical clustering analysis, where all measured variables were considered simultaneously (Figure 5a). Three main clusters (1, 2 and 4) were identified accounting for 93% of the samples (155/166). Clusters 1 and 2 were characterised by elevated expression of CD153, IL-2 and CD27 and low expression of HLA-DR, while cluster 4 presented an opposite profile with low expression of CD153, IL-2 and CD27 and elevated HLA-DR. Visualising the distribution of the different study groups (LTBI, aTB pre-ATT and aTB post ATT) within each cluster using a constellation plot, we observed that cluster 1 was dominated by those who were LTBI and aTB post-ATT (61% and 35%, respectively), cluster 2 included mostly aTB post-ATT and LTBI (49% and 40%, respectively), whereas cluster 4 was composed of mostly of aTB pre-ATT (77%) and 23% of aTB post-ATT (Figure 5b).
To further evaluate the combined power of the measured parameters to discriminate between the three study groups, a PCA was performed. PC1 accounted for 58.3% and PC2 17.1% of the variance (Figure 5c). The PCA showed an almost complete separation between LTBI and aTB pre-ATT (based on PC1), whereas the profile of aTB post-ATT group mostly overlapped with that observed for the LTBI group. The corresponding loading plot shows that HLA-DR, CD27, CD153 and IL-2 were the main drivers of PC1 variance (Figure 5d). The PC1 scores were then compared between each group stratified according to HIV status. Figure 4e shows that while there was no significant difference between the HIV-uninfected aTB post-ATT and LTBI groups, the PC1 score in the aTB post-ATT group remained significantly lower than that in the LTBI group in HIV-infected participants. This is indicative of incomplete restoration of the phenotypic profile of Mtbspecific IFNc + CD4 T cells in HIV-1 co-infected persons, despite completion of the standard ATT regimen. Previous reports have suggested that activation markers such as HLA-DR and CD38, as well as the expression of the maturation marker CD27 measured individually, may have value in discerning between LTBI, aTB pre-ATT and aTB post-ATT. 29 Receiver operating characteristic curves (Figure 6a, left panel) showed that HLA-DR discerned with high sensitivity and specificity aTB patients from successfully treated TB patients, irrespective of HIV infection (Area under the curve (AUC) = 0.95, P < 0.0001 for HIV-uninfected and AUC = 0.91, P < 0.0001 for HIV-infected). Moreover, specificity/sensitivity crossover plots showed that successful treatment was associated with HLA-DR expression on Mtb-specific CD4 T cell post-ATT lower than 30% for HIV-uninfected patients and lower than 42% for HIV-infected participants (Figure 6a, right panel). CD27, by contrast, displayed much poorer sensitivity and specificity ( Figure 6b). Importantly, analysis of the fold change in HLA-DR expression on IFNc + CD4 T cells between the pre-ATT and post-ATT time points indicated that those who failed treatment or relapsed (n = 5) exhibited a significantly lower fold change in HLA-DR expression than in patients classified as cured (median: 0.84 vs 0.31, respectively, P < 0.0001; Figure 6c).

Relationship between the phenotypic signature of Mtb-specific IFNc + CD4 T cells and TB disease extent
Finally, PC1 of the measures of TB disease extent and PC1 of the Mtb-specific IFNc + CD4 T cell phenotype at baseline showed strong correlation (P < 0.0001; r = À0.71), irrespective of HIV infection (P < 0.0001; r = À0.78 for HIV À and P < 0.0001; r = À0.67 for HIV + , data not shown; Figure 7a), indicating that, unlike the magnitude of Mtb-specific CD4 response, 32 the phenotypic characteristics of these cells associate with TB disease activity. Furthermore, only those who showed early sputum culture conversion (e.g. before week 8) had PC1 loading scores for CD4 Tcell phenotypes at week 24 that were comparable to that of the LTBI group (Figure 7b). Thus, the persistence of a skewed CD4 T-cell phenotype in those with late Mtb culture conversion (≥ week 8) suggests that there is a lag (~16 weeks) between Mtb clearance and the full restoration of Mtbspecific CD4 T-cell attributes. Moreover, no significant changes were observed between baseline and treatment completion in the overall phenotypic profile of Mtb-specific IFNc + CD4 T cells of those who experience treatment failure or relapse (P = 0.125, Figure 7b).

DISCUSSION
A better understanding of the relationship between the host immune response, TB disease severity and treatment outcome is key to the development of novel TB diagnostic and treatment monitoring tools. 33 In this study, we defined the profile of Mtb-specific IFNc + CD4 Tcell response in TB patients, analysed its evolution after completion of a standard ATT regimen and evaluated how the host CD4 response associates with clinical TB extent and treatment response. There is great variability in clinical presentation of TB disease. [34][35][36][37] Here, we show that by combining a number of established clinical variables used in TB diagnosis measuring bacterial burden (Xpert ct value), systemic inflammation (plasma CRP and monocyte/lymphocyte ratio) and radiographic evidence of pulmonary disease extent on chest radiograph (Timika score), we were able to define TB disease extent as a continuum with higher scores associating with delayed culture conversion and unfavorable clinical outcomes (treatment failure and relapse). This finding is in accordance with existing evidence showing that sputum smear grade and/or radiological extent of disease at presentation associates with sputum culture conversion time. [38][39][40] We then evaluated whether the Mtb-specific IFNc + CD4 T-cell response could inform clinical TB severity. Our data show that while the magnitude of the response was unrelated to TB disease activity, as previously described 16,21,41 , the overall phenotypic profile of these cells strongly associated TB disease extent, irrespective of patient HIV status. When parameters were analysed individually, our data reveal preferential associations between certain markers and clinical features. The extent of lung involvement (measured by Timika score) showed a strong negative association with IL-2 expression in Mtb- specific IFNc + CD4 T cells. HLA-DR expression on these cells displayed the strongest association with systemic inflammation (defined using plasma CRP concentration) and high bacterial burden (inferred by the Xpert cycle threshold), while CD153 expression on Mtb-specific IFNc + CD4 T cells negatively associated with bacterial burden. Two recent publications have highlighted the potential role of CD153 for Mtb protection in murine and non-human primate Mtb-infection models 42 and in a non-human primate model of intravenous BCG vaccination. 43 However, at this stage, we cannot infer whether the Mtb-specific CD4 T-cell profile is a cause or a consequence of TB disease activity. The mechanisms underlying these observed associations are still unclear, yet our results suggest that the phenotypic attributes of the Mtb-specific response could be independently determined by distinct clinical features. Overall, our results confirm and expand on previous findings, 21 showing that the phenotype of Mtbspecific CD4 T cells associates with TB disease severity.
To date, the gold standard for TB diagnosis relies on detection of Mtb using microscopy, culture or molecular methods. 44,45 However, these tests present substantial limitations as they require either sputum or a specimen from the site of disease in patients with extrapulmonary TB. It may be particularly difficult to obtain representative sputum samples in children or HIVinfected persons, where immunosuppression associates with reduced lung cavity formation. 46 Additionally, in those with extrapulmonary TB, it is frequently difficult or impossible to obtain a specimen from the site of disease as this requires invasive sampling procedures, necessitating heavy reliance on clinical diagnosis and empiric treatment. Our study adds to the mounting evidence that simple whole blood-based assays Figure 6. Performance of HLA-DR and CD27 to discriminate active from treated tuberculosis. (a, b) Receiver operating characteristic curves and specificity/sensitivity crossover plots for HLA-DR and CD27 to discriminate between active from treated tuberculosis (TB) in HIV-uninfected (blue) and HIV-infected (orange) individuals. The diagonal dashed line depicts an AUC of 0.5, representing a random test. The dashed vertical line on the crossover plots represents the optimal threshold to distinguish active and successfully treated individuals. (c, d) Fold change in HLA-DR and CD27 expression in Mtb-specific IFNc + CD4 T cells between baseline and 24 weeks post-treatment initiation. Green dots identify patients considerate as cured based on TB treatment guideline (n = 42), red dots participants who experience treatment failure or relapse (n = 5) and grey dots participants with unknown treatment outcome due to missing sputum culture data (n = 5).
measuring particular Mtb-specific CD4 T-cell maturation and activation markers can be applied not only in the diagnosis of TB, 15,[18][19][20] but also in the assessment of TB disease severity at presentation which is indicative of prognosis, irrespective of HIV co-infection.
Current sputum culture and smear-based assessment of treatment response to standard TB treatment regimens and investigational new drugs also have significant limitations. 47 While sputum culture conversion at 2 months of ATT correlates with clinical cure and lower rates of relapse, 48 it failed as single measure to determine whether shortened, 4-month ATT regimens would be successful with significantly more frequent relapse being observed in those treated for only 4 months compared to conventional 6-month regimens. 49 Moreover, lack of specificity limits the use of sputum Xpert to monitor the response to ATT. 9 We thus investigated whether the phenotypic attributes of Mtb-specific IFNc + CD4 T cells could be relevant to monitor treatment response. Our results showed that, upon ATT completion, the phenotypic profile of Mtb-specific CD4 cells reverts to profiles that were comparable to those of LTBI participants. However, complete normalisation was not observed in all participants, most notably in those co-infected with HIV. Furthermore, all participants with incomplete restoration of their Mtb-specific CD4 T-cell profile post-ATT (e.g. with phenotypic scores significantly lower than that of the LTBI groups) exhibited delayed sputum culture conversion (at or beyond week 8 of ATT). One might hypothesise that those exhibiting phenotypic scores in the ranges of untreated aTB at the end of their ATT regimen may be at particular risk for poor treatment outcomes (treatment failure/relapse) and recurrence, and although further studies would be required to confirm this, we did find that this was indeed the case in the a small number of participants (n = 5) experiencing treatment failure/relapse in our cohort.
Translation of whole blood assay-based approaches into diagnostic and treatment monitoring tools requires use of simplified flow cytometry panels. Upon focusing our analyses solely on HLA-DR expression, our data concur with previous findings from our group and others that a 4-colour panel (including CD3, CD4, IFN-c and HLA-DR) could discriminate latent from aTB 15,[18][19][20] and could also be useful to monitor the response to treatment, as recently suggested by Ahmed et al. 29 Here, we show for the first time that this approach: (1) could identify those at highest risk of poor treatment outcomes, (2) could gauge the extent of TB disease severity and (3) importantly, exhibit comparable performance in HIV-infected persons, who are at highest risk of TB disease. While flow cytometry may appear complex, its usage in clinical tests is becoming more common with the availability of compact flow cytometers. Flow cytometry-based assays, such as CD4 testing, are already in use in programmatic settings globally, including in resource-constrained settings. The translation of a whole blood assay into a point of care product is thus conceivable for clinical reference laboratories. Further development would however be required (e.g. using tubes containing lyophilised reagents for cell stimulation, pre-mixed antibody cocktails for cell staining and automated gating for flow analyses) to ensure commercial scalability.
The study was limited by the fact that it was conducted on ambulant outpatients attending a combined HIV-TB clinic, thus potentially biasing especially the HIV-1 co-infected population towards a less severely diseased phenotype. However, CRP and M/L ratios were comparable between HIV-1-infected and HIV-1-uninfected aTB groups at baseline, indicating comparable levels of systemic inflammation induced by TB infection in these two groups. Sputum culture was limited to pre-determined study visit dates and thus could not be ascertained at more regular intervals. There were missing sputum culture results in a number of patients due to unattended study visits and culture contamination; however, these were in the minority. It also remains to be defined whether Mtb-specific CD4 T-cell phenotyping could be used for early identification of and targeted intervention in those at highest risk for poor treatment outcomes and whether the normalisation of these markers is indicative of relapse-free cure.

Study population
Participants were recruited from the Ubuntu Clinic, Site B, Khayelitsha (Cape Town, South Africa) between March 2017 and December 2018. All participants were adults (age ≥ 18 years) and provided written informed consent. The study was approved by the University of Cape Town Human Research Ethics Committee (HREC 050/2015) and was conducted under DMID protocol no. 15-0047. Those in the aTB group (n = 95) all tested sputum Xpert MTB/RIF (Xpert, Cepheid, Sunnyvale, CA, USA) positive and had clinical symptoms and/or radiographic evidence of TB. All aTB cases were drug sensitive and had received no more than one dose of ATT at the time of baseline blood sampling. The latent TB healthy control group (n = 66) were all asymptomatic, had a positive IFN-c release assay (IGRA, QuantiFERONâ-TB Gold In-Tube, Qiagen, Hilden, Germany), tested sputum Xpert MTB/RIF negative and exhibited no clinical evidence of aTB. Clinical characteristics of the study participants are shown in Table 1. Sputum Xpert MTB/RIF, sputum culture for Mtb, CD4 count, HIV VL, differential full blood count and CRP tests were performed by the South African National Health Laboratory Services. All participants were followed up over the duration of their ATT. Sputum culture was performed at baseline, week 2, week 4 and week 8. Sputum culture was repeated for those where treatment failure or relapse was suspected. Time of sputum culture conversion was defined as the timepoint where the sputum culture result changed from a previous positive result to negative and maintained negative status for the duration of follow-up sampling. Details of those who tested sputum culture negative at baseline are included in Supplementary table 1.

Timika scoring of chest radiographs
Chest radiographs from the enrolment visit of those with aTB were scored by the study clinician using the Timika score. 50 Previously published standardised user guidelines were strictly adhered to. 38 Briefly, posteroanterior chest radiographs were assessed for the total percentage of the lung fields affected by known features of active pulmonary TB. A value of 40 was added to the overall percentage affected lung where at least one cavity ≥ 1 cm could be identified.

Blood collection and whole blood assay
Blood was collected in sodium heparin tubes and processed within 3 h of collection. The whole blood assay was adapted from the protocol described by Hanekom et al. 51 Briefly, 0.5 mL of whole blood was stimulated with a pool of 300 Mtb-derived peptides (Mtb300, 2 µg mL À1 ) 52 at 37°C for 5 h in the presence of the co-stimulatory antibodies, anti-CD28 and anti-CD49d (1 lg mL À1 each; BD Biosciences, San Jose, CA, USA) and Brefeldin-A (10 µg mL À1 ; Sigma-Aldrich, St Louis, MO, USA). Unstimulated cells were incubated with co-stimulatory antibodies and Brefeldin-A only. Red blood cells were then lysed in a 150 mM NH 4 Cl, 10 mM KHCO 3 and 1 mM Na 4 EDTA solution. Cells were stained with a Live/Dead Near-InfraRed dye (Invitrogen, Carlsbad, CA, USA) and then fixed using a Transcription Factor Fixation buffer (eBioscience, San Diego, CA, USA), cryopreserved in freezing media (50% foetal bovine serum, 40% RPMI and 10% dimethyl sulfoxide) and stored in liquid nitrogen until use. . Samples were acquired on a BD LSR-II and analysed using FlowJo (v9.9.6, FlowJo LCC, Ashland, OR, USA). A positive cytokine response was defined as at least twice the background of unstimulated cells. To define the phenotype of Mtb300-specific CD4 T cells, a cut-off of 30 events was used.