Tuberculosis (TB) is the leading infectious disease, ranking above HIV/AIDS. In 2016, there were an estimated 1.7 million TB deaths and 6.3 million new cases of TB were reported, equivalent to 61% of the estimated incidence of 10.4 million.1 Worldwide, China is the country with the second largest number of tuberculosis new cases (after India) and carries an estimated 8.6% of the global burden of new cases and an estimated 12.1% of multidrug-resistant TB cases.1

After infection with the tubercle bacillus, activated CD4+ T-cells induce Th1 cytokines by the stimulation of macrophages and change innate immune responses to adaptive responses.2 Therefore, the immunological diagnostic tool, tuberculin skin test (TST), can detect Mycobacterium tuberculosis infection based on the immune response. As the antigens contained in the TST is crude and poorly refined supernatant mixture of mycobacteria, the widely or sometimes mandatorily used BCG vaccination and the exposure to environmental mycobacteria weaken the specificity of the TST.3

Interferon gamma (IFN-γ) release assays (IGRAs), including the commercially available kits QuantiFERON-TB Gold In-Tube (QFT-GIT; Qiagen, Germany) and T-SPOT.TB® (Oxford Immunotech, Abingdon, UK) have been recommended for the immunological diagnosis of M. tuberculosis infection. A meta-analysis comprising 27 studies showed that the pooled sensitivity for the diagnosis of active TB was 80% (95% CI 75-84%) for QFT-GIT and 81% (95% CI 78-84%) for T-SPOT.TB® in blood fluids. The pooled specificity was 79% (95% CI 75-82%) and 59% (95% CI 56-62%), respectively.4 Another meta-analysis focusing on mid- and low-income countries reported similar results: pooled sensitivity in HIV-uninfected persons of 88% (95% CI, 81–95%) for T-SPOT.TB and 84% (95% CI, 78–91%) for QFT-GIT. The pooled specificity was low for both assays, namely, 52% (95% CI 41–62%) and 61% (95% CI, 40–79%), respectively.5 Although both IGRAs have higher diagnostic sensitivities than TST, they were still not high enough to justify using IGRA as a test to rule out TB. The specificity limitation in diagnosing active TB could not be easily solved, which was commonly due to inability to discriminate between latent and active tuberculosis infection, in addition to cross-activation with other environmental mycobacteria. However, the sensitivity could be improved by adding more diagnostic antigens into the assays, mainly antigens encoded by the region of deletion 2 (RD2) of MTB.6–12 Peptide 4 from TB7.7 (Rv2654, encoded by RD13) was incorporated into the QFT-GIT to improve sensitivity.13–16

The RD2 of MTB encodes 11 ORF (from Rv1978-Rv1989c) which contribute to bacterial virulence.17 Previously, we demonstrated that Rv1978, Rv1981c, and Rv1985c from RD2 were T-cell antigens and could be used to discriminate TB-infected individuals from healthy BCG-vaccinated controls.12,18,19 T-cell responses against peptide pools from other proteins in RD2, including Rv1983, Rv1986, Rv1987, and Rv1989c have been tested in M. bovis-infected and BCG-vaccinated cattle.8,20 The percentages of positive IFN-γ releasing responders ranged from 41% to 62% in M. bovis-infected cattle, which was higher than the percentages in BCG-vaccinated cattle, which ranged from 0% to 50%. Diagnostic performance and T-cell epitopes may be different in cattle compared to humans; it is therefore necessary to identify and verify these promising T-cell epitopes in individuals. Peptides of Rv1989c have been verified, in 49 TB patients and 38 healthy individuals, but with unsatisfying results. Only 16–23% TB patients had a positive response.7 Cellular epitopes from other proteins remained unknown except for the epitopes of IL-2 responses against Rv1986.21

In this study, we initially screened overlapping synthetic peptides encoded by Rv1983p257-558, Rv1986, Rv1987, Rv1988, and Rv1989c for TB diagnostic purposes, and subsequently pooled six candidate epitopes and evaluated their joint diagnostic performance for active TB.

In recent years, there has been great progress in developing new approaches for the immune-diagnosis of tuberculosis, which includes the measurement of alternative biomarkers secreted as part of cell-mediated immune responses,24,25 the identification of novel antigens,12,26,27 monitoring of T-cell subgroup changes 28,29 and the detection of host biomarkers of immune response.27,30 Using a peptide-based approach to identify new antigens and epitopes from the old immune-dominant proteins can be an effective way to screen and improve the performance of the current methods.

In our screening study, four epitopes from Rv1986, and two epitopes each from Rv1987, Rv1988, and Rv1983 were recognized by more than 10% of TB patients (Figure 3A). Rv1986c, a probable LysE family protein involved in the transport of lysine, may contribute to bacillus persistence, as it is up-regulated at later time points (4-7 days) when mycobacteria is treated with hypoxia, developing what is known as the enduring hypoxic response (EHR).31 Rv1986c, together with other EHR-induced antigens, can stimulate an immune response that leads to the secretion of both IFN-γ and IL-2 in TB-infected cattle.32 One identified epitope region, P15 and P16 (amino acid 157-187) was identified previously,21 but the other region P9 and P10 (amino acid 91-120) was not. The most dominant epitopes were Rv1987-P8 and Rv1987-P11, with median intensity reached at 38Ypg/mL and 12Ypg/mL, respectively, in the TB1 group. Despite the high intensity, they were unsuitable for diagnostic purposes due to high cross-reactivity in the control population. Rv1987 encodes a possible chitinase family protein that is distributed widely among bacteria, fungi and even animals, and which may underlie the high cross-reactivity of Rv1987. The recognition in healthy individuals was in agreement with a previous study with M. bovis-infected cattle: 57% recognition in infected cattle and 50% in BCG-vaccinated cattle.8 Rv1983 (PE_PGRS35), which belongs to the PE/PPE protein family of Mycobacterium, is another antigen that is definitively recognized in both infected and BCG-vaccinated cattle.8 In order to avoid cross-reactivity maximally, the region of PE_PGRS35p1–p256, which had common motifs shared among the PE/PPE family members, was not chosen for epitope screening. In the screening experiments, 56% (9/16) of TB patients had a positive response to at least one peptide of Rv1983. The percentage that was recognized was equivalent to that in the cattle study (59%).8 TB patients preferentially recognized Rv1983-P4, P6-P13 (amino acid 312-408), and P24, but the percentage that was recognized was low for any single peptide. Among them, two were demonstrated to be immune-dominant epitopes and one, Rv1983-P4 (P290-309), had potential diagnostic value. Rv1988 encodes an intrinsic macrolide-resistant gene named erm,37,33 probably through its methyltransferase function. In this study, Rv1988 was recognized by 52% of TB patients and identified as a T-cell antigen. Two dominant epitopes, Rv1988-P4 and P11, had relatively high recognition in TB patients but less or none in NTP1 or healthy controls, which makes them potentially useful for the diagnosis of TB. To our knowledge, this is the first time they have been studied. It is less common for an antibiotic-resistant gene to be identified as a diagnostic antigen with high specificity; however, it has been found that a high similarity homolog of the erm family is mainly restricted to the tuberculosis complex and limited species of Mycobacterium, including Mycobacterium abscessus and Mycobacterium bolletii.34

In this study, T-cell responses were measured by using a diluted whole blood assay and the incubation period was lengthened to 5 days instead of overnight to ensure the production of a sufficient amount of secreted IFN-γ.35 Although the antigen-specific cytokines were reported to progressively decrease when whole blood was diluted, the assay required much less blood than expected, considering that screening included 87 peptide subgroups for each patient. The data showed that the amount of cytokine secreted from 20μL of blood was sufficient to determine the T-cell epitopes. All but three of the 117 NTP1 individuals had positive IFN-γ responses to the positive control antigen, PHA, and the results ranged from 10 to>400Ypg/mL, with the median value exceeding 400Ypg/mL. Similarly, all but one participant had a positive T-cell response to PHA in an undiluted parallel IGRA test. The patient who had negative responses to PHA had a positive response to TB antigen, suggesting that most patients were immune-competent at a cellular level. However, we noted that the 5-day incubation was not suitable for clinical diagnosis. Therefore, overnight incubation and an undiluted whole blood assay were performed in the assessment study. Prolonged incubation can stimulate more memory T-cells, but short-time incubation mainly stimulates effector T-cells. In addition, a longer stimulation duration (6 days) was shown to be necessary for the optimal induction of IFN-γ and TNF-β, and was practically convenient for the detection of IL-10, IL-1-β, TNF-α, and IL-6.36

In the assessment study, it is inspiring that a pool of six peptides from RD2 could detect more than half of all active TB patients, although this is lower than the number detected by the T-SPOT.TB assay. A total of 57% of the patients also had positive results for the peptides in the PTB group. Patients in this group all had symptoms, radiographic signs or histological evidence that suggested a high possibility of TB infection, while 60% (25/42) of them received empirical anti-tuberculosis therapy and most showed clinical improvement. Positive detection in this group is also meaningful and instructive as microbiological investigation failed to provide conclusive results. In the preliminary screen, only 10 peptides were recognized by more than 10% of TB patients. The peptides with less frequent recognition would be useful for diagnostic purposes and would have potential to increase sensitivity, if the recognition pattern to these peptides were complementary to the more frequently recognized peptides. However, considering the budget limitation and work load, the recognition value in TB1 group was set to 8%, which included 22 peptides (25%) in the latter screen. Although the sensitivity is not high enough to allow them to be used alone, the peptides have shown potential diagnostic value, and to be adjunctive antigens to the current immunological diagnostic methods.

In the NTP2 group and the ID group, in which TB was less likely, the positive results for the R6 assay were lower than in the highly likely TB groups, as expected. Interestingly, patients diagnosed with cancer, especially lung cancer, were the major reasons for false positive results in the R6 assay in the NTP2 group. In total, 10 of 16 patients with false positive results by the R6 assay had cancer. After excluding those patients whose condition was complicated by cancer, the specificity could be improved to 90.3% (56/62). Lung cancer may cause an autoimmune disorder that results in frequent positive IGRA results37; therefore, one should be very careful when interpreting IGRA positive results from lung cancer patients. For the remainder, false positive results in the NTP2 group and latent TB infection could be the most feasible explanation. Peptide Rv1988-P4, which was recognized by control individuals in screening assay (Figure 4G), could be another reason for reducing the specificity of the peptide pool. As all the patients are from China, which is a high TB epidemic country, the 10-20% positive IGRA results in the NTP group are reasonable and close to the previously reported positive rate in healthy individuals in China.38 Although we assumed that latent TB infection could be the reason for false positivity in the study population, as these antigens are not specifically expressed in latent infection, we did not design a group to test whether latent TB could be recognized by the R6 assay or any different recognition between active and latent TB individuals, which is worthy investigating in the future study. Four patients were culture-positive but confirmed to have NTM disease by selective PNB culture media. None had any positive results for the R6 assay, suggesting that the peptides may also be suitable for differentiating NTM from TB. Despite this, more cases need to be evaluated.

For the parallel T-SPOT.TB assay, the sensitivity and specificity were 76% (37/49) and 75% (68/91), respectively, for the two groups. Unlike the assessments in healthy controls or in some retrospective studies, the unsatisfactory results reflect the real-life diagnostic challenge in highly epidemic regions. The diagnosing performance in active TB was close to that summarized in meta-analysis studies,4,5 and reinforces the conclusion that the sensitivity is not high enough to rule out TB. Like WBIGRA for the R6 peptides, more than half of the false positive results (13/23) in the NTP2 group could be attributed to the diagnosis of cancer. If patients diagnosed with cancer were excluded, the specificity improved to 83.9% (52/62).

In this study, the diagnostic performance of the six promising epitopes, Rv1986-P9, P15, P16, Rv1988-P4, P11, and Rv1983-P4, was assessed on a panel of 233 patients suspected of having TB for the first time. Our data suggest that these epitopes from RD2 of M. tuberculosis may have potential diagnostic value in the immunological diagnosis of TB.

The authors have applied a patent for five peptides, including Rv1986-P9, Rv1986-P15, Rv1986-P16, Rv1988-P4, Rv1988-P11, in TB diagnosis.