Tuberculosis (TB) remains a leading cause of death in the world, responsible for high morbidity and mortality worldwide with about 10 million new cases in 2020 and 1.5 million deaths (1). Diagnosis and hence appropriate treatment are often delayed due to the wide spectrum of clinical manifestations of active TB (aTB) and to the non-availability of sensitive and specific tests providing a rapid and accurate diagnosis (2, 3). In addition to classical clinical manifestations of aTB, this disease may occur as subclinical TB without suggestive symptoms, or as extrapulmonary TB often pauci-symptomatic in immunocompromised individuals, so that diagnosis strictly based on clinical signs or symptoms is illusive (2). The classical diagnostic method remains the identification of M. tuberculosis by direct smear microscopy or by culture that are both low in sensitivity and/or slow. Early diagnosis of infectious cases by sputum microscopy is only possible in approximately 50% of cases. Moreover, people with subclinical TB may likely be missed if TB culture is not performed and this is often the case in asymptomatic individuals. Molecular techniques such as GeneXpert are more sensitive but they are costly and often unavailable in primary-care settings.

Immuno-diagnosis was therefore identified as a promising approach for diagnosis of aTB. However, the commercially available tests, the interferon-γ-release assays (IGRA), based on the release of IFN-γ by blood cells in response to their in vitro stimulation with mycobacterial peptides corresponding to antigens encoded in the genomic region of difference (RD)-1 (the early-secreted-antigenic-target-6 (ESAT-6), and the culture-filtrate-protein-10 (CFP-10)), have a relatively high false negative rate in patients with aTB (4–6). In addition, these IGRAs, initially developed to diagnose latent TB infection (LTBI), are positive both in LTBI and in patients with aTB (7). The differential diagnosis of LTBI and aTB is therefore not possible with commercial IGRAs. New generation IGRAs were developped to improve the diagnosis of aTB. They are based either on recent studies indicating that the M. tuberculosis specific CD8⁺ T cell responses are positively correlated with the bacterial load and recent exposure to M. tuberculosis for the QuantiFERON-TB Gold Plus, or by inclusion of L-alanine dehydrogenase as an additional antigen for the LIOFeron TB/LTBI (8, 9). The added value of these improvements for the diagnosis of aTB and its differential diagnosis with LTBI was however not confirmed until now (8).

Therefore, several studies aimed to identify other proteins from M. tuberculosis as potential candidates to distinguish aTB from LTBI and they often also extend the cytokine measurements beyond IFN-γ to improve differential diagnosis. These studies generally first identify markers of M. tuberculosis infection, before applying one or several biomarkers to differentiate aTB from LTBI. Among the numerous proteins evaluated, the mycobacterial heparin-binding haemagglutinin (HBHA, Rv0475) appears as one of the most promising antigens to differentiate LTBI from aTB, but only few studies assessed HBHA-induced cytokines other than IFN-γ (10–12). In contrast, the potential added value of various cytokines induced by the peptides of the commercial IGRAs has been investigated, but with conflicting outcomes, as highlighted by a recent meta-analysis (13).

We previously reported on a combined HBHA- and ESAT-6-IGRA performed on either peripheral blood mononuclear cells (PBMC) or whole blood (WB) to diagnose LTBI and to partially differentiate LTBI from aTB (14–16). Even though this combined IGRA provided a better discrimination between LTBI and aTB than any other available in vitro test, it remained imperfect. In this study, we therefore measured cytokines/chemokines other than IFN-γ released in response to HBHA and to ESAT-6 and evaluated the potential added value of a combined analysis of cytokine/chemokine secretion for the differential diagnosis and especially for the diagnosis of aTB.

TB control could be significantly improved if simple and rapid triage or rule-out tests were available to first exclude non-infected subjects and then differentiate as much as possible aTB from LTBI. The development of triage tests designed for use by first-contact health care providers as a rule-out test of TB was reported by the World Health Organization as a high priority need for TB control (22). This strategy would limit the number of individuals requiring a confirmatory test. High sensitivity is needed for this approach to avoid missing patients with aTB. In contrast, the specificity of this triage test approach may be lower, as the patients initially selected by the triage test will benefit from more in-depth evaluation of their status. Blood-based immunological tests might be appropriate as triage tests. We previously reported that secretion of IFN-γ by PBMC or WB in response to in vitro stimulation with HBHA is a biomarker for LTBI contrasting with ESAT-6-induced IFN-γ secretion that is more characteristic of aTB (14, 15). Therefore, we suggested that combining IFN-γ secretion in response to HBHA and to ESAT-6 would help to differentiate LTBI from aTB and allows to stratify LTBI subjects in different groups associated with risks of reactivation of the infection (16, 21). In this study, we evaluated combinations of cytokines/chemokines released in response to these two antigens for improved identification of aTB patients, which could then be proposed as a triage test to select patients likely presenting aTB for further investigations to confirm the diagnosis. We propose a two-step approach consisting of first identifying infected subjects and then differentiating aTB from LTBI among infected individuals. Aiming to provide an easy test, we searched for optimal cytokine/chemokine combinations secreted by 24 hours-stimulated WB. However, this approach might not be sensitive enough for patients with lymphopenia or important inflammatory syndrome associated with high levels of plasma proteins that could inhibit cellular immune responses. It might also not be suitable for frozen material. We therefore also evaluated optimal cytokine/chemokine combinations secreted after 24 hours of in vitro-stimulated PBMC.

Among the 13 measured cytokines/chemokines secreted in response to HBHA and ESAT-6 and their various combinations to identify M. tuberculosis-infected patients, the best results were obtained both on WB and on PBMC by combining IFN-γ concentrations secreted in response to HBHA and to ESAT-6, as previously shown (15, 16). This approach allowed us to discard for further analysis most non-infected controls, 82%/71% in the training cohort and 84%/96% in the validation cohort, in the PBMC and WB assays, respectively. Misclassified controls were most often doubtful rather than being clearly positive. In addition, among the 10 misclassified controls in the PBMC and/or WB assay for both cohorts, five were health care workers potentially exposed to aTB patients, and four reported frequent travel to high TB endemic countries. Although these individuals were classified as non-infected, based on a negative TST result as recommended in Belgium (23), we cannot formally exclude that they developed immune responses to mycobacterial antigens as a consequence of exposure to M tuberculosis. They represented a minority of the enrolled controls that should benefit from further investigation (medical visit, chest radiograph, eventually followed by microbiological analysis of sputum samples) to exclude an ongoing M. tuberculosis infection. On the other hand, most infected individuals and especially most aTB patients were correctly identified and were selected for further investigations (100%/91% and 88%/82% in the training cohort in the validation cohort, in the PBMC and WB assays, respectively). Among the five misclassified aTB patients either in the PBMC or the WB assay in both cohorts, two had extra-pulmonary TB (one meningitis and one spondylodiscitis in a pregnant women) with obvious symptoms, while the other three had symptomatic pulmonary TB with severe denutrition and serious abnormalities on chest radiographs ( Tables 2 and 3 ).

As a second step, we searched for the best combination of cytokines/chemokines secreted by the two stage-specific mycobacterial antigens to allow us to identify patients with aTB among infected individuals. Results were slightly different between the PBMC and WB assays. For PBMC stimulation, the best combination was HBHA-induced IFN-γ and ESAT-6-induced IP-10 that allowed us to correctly identify 77% and 71% of aTB patients in the training and validation cohorts, respectively. The interest of ESAT-6-induced IP-10 for diagnosis of M. tuberculosis infection was previously reported but with little added value compared to IFN-γ, except for patients with immune deficiencies (24). ESAT-6-induced IP-10 was not reported to provide differential diagnosis between aTB and LTBI. However, we show here that, when combined with HBHA-induced IFN-γ, it improves the differentiation between aTB and LTBI. This differentiation allowed us to identify ¾ of the patients with aTB, which is an improvement over previous studies. In addition, most misclassified aTB patients were shown to be infected in the first stage analysis and will therefore receive clinical attention.

Globally, eight LTBI subjects were misclassified as aTB. Seven of them were considered in a previous study as being at risk to reactivate their infection, based on their high ESAT-6-induced IFN-γ secretion (21). Latency is considered as a spectrum of different stages of activity of persistent mycobacteria, from totally quiescent to persistently multiplying (17, 18). This latter stage is the highest risk stage and therefore may not be possible to be differentiated from subclinical TB using immunological biomarkers. We may therefore hypothesize that these 8/49 misclassified LTBI subjects might have subclinical TB. Therefore, these individuals should be prioritized for more extensive evaluation for a possible aTB.

For WB stimulation, the best discrimination between aTB and LTBI was obtained by combining HBHA-induced IL-2 and GM-CSF with ESAT-6-induced IP-10. This allowed us to identify 90% and 93% of the aTB patients among infected individuals from the training and validation cohorts, respectively. The main advantages of the WB assay are the relatively small blood volume required (feasible on 1 mL WB) and easiness to perform. The good sensitivity should allow us to propose these tests as a triage test for aTB as the recommended sensitivity is to be at least 90% (22), and should help to limit the number of individuals who require confirmatory tests. One third of the LTBI subjects were misclassified and would require the confirmatory test. However, as for LTBI subjects misclassified with the PBMC assay, most of them were considered at risk to reactivate the infection and should therefore be prioritized for further evaluation of their status.

The sensitivities achieved with these combinations are similar to those reported in a few other studies aiming to define combinations of biomarkers that provide diagnostic test accuracy consistent with WHO specifications for a rule-out test for aTB. These studies that were recently summarized (25), are based on combinations of various serum markers, mostly cytokines/chemokines (26–28), sometimes combined with antibodies against a TB antigen to raise the specificity for TB (29). We chose in our study to measure cytokines/chemokines released in response to mycobacterial antigens to increase the specificity for TB. A few studies already addressed this question by measuring chemokines/cytokines induced by the peptides from the QuantiFERON, with different marker selections among the studies (13). Our results are difficult to compare to these studies, as we selected combinations of chemokines/cytokines induced by two different stage-specific mycobacterial antigens, ESAT-6 and HBHA, in order to cover a wide range of the M. tuberculosis metabolism. ESAT-6 is highly expressed during bacterial multiplication, while HBHA is a latency-associated antigen, whose gene is upregulated in hypoxic conditions and in cells harboring M. tuberculosis during latency (30–32). Not surprisingly, the selected combinations to differentiate aTB from LTBI were different between the PBMC and WB assays, since during acute inflammation plasma proteins may modify cellular immune responses.

Strengths of our study are that the biomarkers identified in the training cohort were validated in an independent validation cohort, both for the PBMC and the WB assays, and that these biomarkers are easy to measure in most laboratories. In addition, as blood and not sputum-based biomarkers, they are suitable to identify both pulmonary and extra-pulmonary TB who were both included in our cohorts. A limitation of this study results from the relatively low numbers of individuals in each cohort and from the heterogeneity of the cohorts that comprised both patients with pulmonary and extra-pulmonary TB. Moreover, we did not include in our cohorts TB-like diseases, which may perhaps represent more relevant negative controls for the evaluation of biomarkers to diagnose aTB. Inclusion of TB-like diseases was unfortunately not possible, because no screening for LTBI is performed in these patients in Belgium so that correct classification of these patients would have been difficult.

Based on the results of this pilot study performed in a low TB incidence country, we propose a two-steps algorithm to identify patients who are highly suspected to present aTB and should be further investigated ( Figure 2 ). This algorithm warrants further evaluation on larger cohorts of subjects both in low and high TB endemic countries to confirm its robustness as triage test for aTB. Albeit not as simple as recommended to be a point of care test (22), it has the advantage of being relatively easy to perform on small blood volumes, to provide acceptable identification of aTB patients, and to be less expensive than systems biology approaches that have identified diagnostic signatures to discriminate aTB from LTBI (33, 34).

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethics Committee of ULB-Hôpital Erasme, Brussels, Belgium. The patients/participants provided their written informed consent to participate in this study.

VD designed the study, analyzed the data and wrote a first draft of the manuscript. PC performed the complete statistical analysis of the data. AP performed all the experimental work. MH, ND, and SA included patients, provided clinical data and reviewed the manuscript. JR did the initial statistical analysis. MS prepared ESAT-6 antigens. CL coordinated the preparation of antigens and critically reviewed the manuscript. FM designed the study, interpreted the data, and drafted the final manuscript. VC designed the study, analyzed and interpreted the data, collected clinical data, drafted figures and tables and critically reviewed the manuscript. All the authors contributed to this article and approved the submitted version.

This work was supported by the Fonds National de la Recherche Scientifique (FNRS—PDR T.0147.13), by the European Community within the Seventh Framework Program (FP7) NEWTBVAC (grant HEALTH-2009-2.3.2-2), and within the Horizon2020 program TBVAC2020 (grant agreement 643381), and by Innoviris Brussels.

The author MS is employed by Lionex Diagnostics and Therapeutics, Braunschweig, Germany.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.