Cellular Immune responses during latent tuberculosis : immunodiagnosis and correlates of protection Leyten, E.M.S.

(1)

immunodiagnosis and correlates of protection

Leyten, E.M.S.

Citation

Leyten, E. M. S. (2008, October 8). Cellular Immune responses during latent tuberculosis : immunodiagnosis and correlates of protection. Retrieved from https://hdl.handle.net/1887/13137

Version: Corrected Publisher’s Version

License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden

Downloaded from: https://hdl.handle.net/1887/13137

Note: To cite this publication please use the final published version (if applicable).

(2)

Chapter 10

Summary and general discussion

(3)

(4)

Chapter 10 SUMMARY

The studies in this thesis focus on cellular immunity against mycobacteria during latency with the aim to contribute to improved immunodiagnosis of latent TB and to gain insight into immune responses which play a role in controlling latent infection. Besides ongoing community transmission of TB, many new cases of active TB arise from the reservoir of latently infected individuals, which is currently estimated to consist of 2 billion persons worldwide. The current HIV pandemic further increases number of TB cases which arise from this huge reservoir, as the risk of reactivation in HIV infected individuals is high;

8-10% per year. Latent M. tuberculosis infections are an important reason for the relative lack of success of eradication programmes, which have thus far been based mainly on the detection and treatment of active TB. Accurate diagnosis of latent M. tuberculosis infection and within these identification of those individuals at risk to develop active TB, making targeted treatment of latent TB possible, will be important determinants to be able to achieve control and finally eradication of TB.

The first part of this thesis focused on the development and evaluation of new immu- nodiagnostic assays for accurate diagnosis of latent M. tuberculosis infection. The second part concentrated on the search for antigens that are specifically targeted by the immune system during latency, with the ultimate aim to identify latency associated antigens that are correlated with protection.

Part I. IMMUNODIAGNOSIS

Discovery of new M. tuberculosis-specific antigens for immunodiagnosis

The M. tuberculosis-specific proteins ESAT-6 and CFP-10 are potent T-cell antigens which are recognized by genetically heterogeneous human populations (1). Immunodiagnostic assays based on these antigens were found to be highly specific for the detection of M.

tuberculosis infection. The first two studies, described in chapters 2 and 3, focus on the identification of additional M. tuberculosis antigens or peptides pools to improve the sensitivity of immunodiagnostic assays for M. tuberculosis infections.

In chapter 2, we evaluated the diagnostic potential of four recently characterized antigens, Rv3873 (TB37.6), Rv3878, Rv2653 and Rv2654 (TB7.7) of M. tuberculosis.

Although these proteins are encoded by RD1 and RD11, both genomic regions that are deleted in M. bovis BCG, the results showed that several epitopes in Rv3873, Rv3878, Rv2653 were recognized by PBMC from BCG-vaccinated individuals who were not likely to be exposed to M. tuberculosis. By removing the cross-reactive parts of the molecules, we were able to design four peptide mixtures (Rv2654, Rv3873 (p2-6), Rv3878 (p3-9), and Rv3878 (p11-15)), that were recognized with high specificity, with no more than

(5)

one out of the 29 BCG-vaccinated persons responding. The peptide cocktails of Rv2654 and Rv3873 (p2-6) were most frequently recognized, by 41% and 43% of the TB patients, respectively. Subsequently, it was demonstrated that, by combining these novel peptides, a cocktail of peptides could be composed that is highly specific (95%) and reasonably sensitive (57%) for the detection of M. tuberculosis infection. Further, the data indicated that combining this newly identified cocktail of M. tuberculosis-specific peptides with the antigens ESAT-6 and CFP-10, can increase the sensitivity from 86% (ESAT-6+CFP-10 alone) to 90%, while retaining a high specificity of 95%.

An ideal diagnostic assay for latent M. tuberculosis infection, which could be used for the selection of individuals eligible for chemoprophylaxis, should not only be highly specific, but should also have an optimal sensitivity, as individuals with a false negative test result will be at risk for progression to active TB. Although sensitivity of the individual M. tuberculosis specific peptides pools that were identified in chapter 2 was not very high, they could improved the sensitivity of the assay based on the currently used antigens ESAT-6 and CFP-10. In the most recently marketed IGRA, QuantiFERON TB Gold in- tube (QFT-GIT), a peptide of TB 7.7 was added to ESAT-6 and CFP-10 with exactly this aim. A study from South Africa confirmed improved sensitivity of QFT-GIT over QFT- G which only includes ESAT-6 and CFP-10 (2). Further, new M. tuberculosis-specific peptide pools could become of particular importance in case future TB vaccines would contain ESAT-6, which would limit the usefulness of this antigen in diagnostic assays.

In chapter 3 we investigated the usefulness of an IFN-γ-ELISPOT using the two most promising peptide cocktails identified in the previous study, (TB7.7 and TB37.6 (p2-6)) in addition to ESAT-6 and CFP-10, to screen personnel of a microbiological laboratory who were accidentally exposed to M. tuberculosis. The tuberculin skin test (TST) was of limited value, as many employees were BCG vaccinated or were already TST positive due to previous exposure to M. tuberculosis. Using the ELISPOT, two of the nine exposed laboratory technicians were found to respond strongly to ESAT-6 and CFP-10, precisely those with the highest level of exposure during the accident, and in one of whom a TST conversion was observed. The peptide mix of TB37.6 (p2-6) was exclusively recognized by these two recently exposed individuals and not by controls with a history of TST conversion or cured TB who did respond to ESAT-6 and CFP-10. Moreover, follow-up with ELISPOT one year after the accident revealed that responses to TB37.6 had become undetectable, while responses to ESAT-6, CFP-10 had remained unchanged. These data suggest that a positive ELISPOT response to TB37.6 peptides could be a marker for recent infection. In this study TB7.7 was poorly recognized. In conclusion, this study illustrates the value of an IFN-γ-ELISPOT based on ESAT-6, CFP-10 and peptides 2-6 of TB37.6 for the detection of recent latent infection, in particular in a setting with a background of BCG vaccination and possible previous exposure to mycobacteria.

(6)

Chapter 10 Further development of immunodiagnostic assays which are able to specifically iden-

tify recently infected individuals is of importance as such assays could reduce the number of individuals needed to treat to prevent one case of active TB.

Does prior tuberculin skin testing affect the outcome of IFN-γ release assays?

In 2006, the UK guidelines issued by the National Institute for Health and Clinical Excel- lence included both QFT-G and T-SPOT.TB (both at that time approved for use in Eu- rope) and recommended a two-stage strategy of TST screening followed by an IGRA to confirm a positive TST result. However, there were no studies demonstrating the validity of this two-stage approach. From the two-step TST it is known that a TST can boost subsequent immune responses to tuberculin (or PPD), the preparation used for TST. Since PPD is known to contain fragments of M. tuberculosis-specific antigens, such as ESAT-6, it is conceivable that TST could also boost in vitro immune responses to these antigens and influence the outcome of IGRA.

The study in chapter 4 was designed with the aim to evaluate the effect of TST adminis- tration on the result of QFT-GIT. QFT-GIT was performed on the day of TST administration and the day of reading (day 3) in 66 individuals, including TST-negative persons, TST-positive persons with known exposure to M. tuberculosis and 5 cured TB patients.

The results showed no systematic boosting of QFT-GIT responses from negative to positive 3 days after TST administration. In only a few persons with pre-existing immune responses did the administration of a TST enhance QFT-GIT responses. Interestingly, a clear rise in IFN-γ production was seen in two TST-positive study subjects when the QFT-GIT was repeated 10 and 11 days after TST administration indicating that boosting of in vitro responses might occur at a later time point than at the day of TST reading. A surprising observation of this study was that QFT-GIT results were positive in only 37% of TST-positive individuals who had a mean TST of 17 mm. In particular among those who likely had been exposed to M. tuberculosis more than 2 year previously, the sensitivity of QFT-GIT was low, with only 6/23 (26%) positive results. This rather unexpected finding was further studied in chapter 6 of this thesis. In conclusion, this study demonstrated that the specificity of QFT-GIT was not jeopardized by prior TST administration when QFT-GIT was performed on the day of TST reading, which is the most practical moment for IGRA in a two-step approach.

Until the precise role of IGRA for screening of latent TB infection has been clarified, IGRA will often be used in association with a TST. Therefore it is important that our study now indicates that the result of QFT-GIT was not importantly influenced by the administration of tuberculin when performed on the day of TST reading, This observation was later confirmed by others for QFT-G as well as T-SPOT in a small group of healthy subjects (3). However, an observation in two of our study participants indicated that boosting of QFT-GIT responses might occur in case the interval between TST and

(7)

IGRA is longer. Naseer et al. recently showed that boosting of the QFT-G response by prior TST can occur after 6 weeks, as 3 out of 9 subjects with an initial negative QFT- G, became QFT-G positive (3). It has been suggested that repeated TST could work as a micro-vaccination, and could thereby induce new M. tuberculosis-specific immune responses (4). However, repeated TST did not affect T-SPOT.TB when tested after an interval of 6 months, which argues against the idea of priming of immune responses by TST administration. An alternative explanation is that the TST, by intracutaneous injection of PPD, which contains fragments of M. tuberculosis-specific antigens that are also used in IGRA, can boost pre-existing immune responses to these peptides which is in concurrence with the boosting phenomenon of repeated TST (5).

IGRA to detect latent M. tuberculosis infection during contact investigation Recently, two M. tuberculosis-specific IGRA have been marketed, the QuantiFERON®- TB Gold (QFT-G), which is a whole-blood assay using ELISA for detection of IFN-γ responses, and T-SPOT™.TB, which is based on ELISPOT and detects the number of IFN-γ producing cells. Both assays are based on peptides of ESAT-6, CFP-10, while in the latest in tube version of QFT-G (QFT-GIT) p4 of TB7.7 has been added. The study described in chapter 5 evaluated the performance of these two commercial IGRA for the diagnosis of latent TB during a large scale contact investigation. The setting of this study was unique as it allowed for the correlation of IGRA and TST results to the level of exposure to M. tuberculosis by calculating the total number of shopping hours for each customer during the period that an employee with highly contagious smear positive TB has been working in the supermarket. A TST was administered to 15,515 customers, excluding persons born before 1945, BCG vaccinated persons and those with a history of TB or a known positive TST. QFT-GIT and T-SPOT.TB were performed in 469 subjects who where randomly selected on the day of TST administration and in 316 subjects with a TST result > 0 mm as selected on the day of TST reading.

Among the 785 study participants, positive TST results were associated with higher age, while a positive IGRA result, most markedly for QFT-GIT, was not associated with age but was significantly associated with the cumulative shopping time as a quantitative marker of exposure at the supermarket. Of note, when evaluating all persons included in the large scale contact investigation a weak correlation between TST and overall shopping time could be observed (personal communication with Ben Koster, Municipal Health Authority, the Netherlands). Interestingly, the sensitivity of QFT-GIT and T-SPOT.TB was only 42.2% and 51.3%, respectively, among BCG-unvaccinated persons with a TST of

≥ 15 mm which reliably indicates latent TB infection in this setting. The overall agreement between both IGRA was high at 89.6% (κ = 0.59) using the cut-off values as indicated by the manufacturer. By varying the cut-off values of both IGRA, the agreement could be further improved and was optimal at 93.6% (κ =0.71) when using a lower cut-off of 0.20

(8)

Chapter 10 IU/ml instead of 0.35 IU/ml for QFT-GIT and a higher cut-off of 13 spots instead of 6

spots for T-SPOT.TB.

Two important conclusions can be drawn from this large contact investigation. First, the study showed that the two commercial IGRA showed a high inter-assay agreement and were both found to correlate better with the level of exposure than did the TST.

These observations are in line with several other publications A fairly good overall agreement between these IGRA was also found in other studies among TB patients and recent contacts, with T-SPOT.TB being slightly more sensitive and QFT-G(IT) seeming slightly more specific (6-8). Further, a recent meta-analysis of studies evaluating these IGRA in a low TB-endemic setting, showed that IGRA correlated better to the level of exposure, as a surrogate marker for infection, than did the TST (8-15). While most of these studies were performed in predominantly BCG-vaccinated populations, we demonstrated that this was also the case in a BCG-unvaccinated population in which false-positive TST results are unlikely to occur. A small group of supermarket customers had negative TST results but were positive in IGRA. A follow-up study of 29 of these untreated persons by Franken et all., showed that half of them developed a positive TST one year later (16).

This would suggest that in these individuals IGRA was more sensitive than TST in diag- nosing recently infected contacts.

The second, more surprising finding of this study was the high number of negative IGRA results among persons with a TST of ≥ 15 mm. In accordance with this observation were the results from our study in chapter 4, in which QFT-GIT results were negative in more than half of the TST+ individuals with documented past exposure to M. tuberculo- sis. The cause of this discrepancy between TST and both short-culture IGRA was unclear at the time of that study. A cut-off value of 15 mm has a very high specificity for latent M. tuberculosis infection in the Dutch, BCG-unvaccinated population while, at that time, the sensitivity of IGRA was thought to be very high as based on patients with TB disease and close contacts who were examined several weeks after exposure. We hypothesized that the observed discrepancy could be related to the fact that the TST is able to detect infections which have been acquired at any time in the past, as the TST, once positive, as a rule remains positive during a lifetime, while QFT-GIT and T-SPOT.TB mainly detect recent M. tuberculosis infection.

Short versus prolonged incubation IGRA

With the aim to further investigate the above hypothesis that the commercial IGRA are sensitive for the detection of recent but not so much for detection of past infection, we compared the performance of two short-incubation IGRA, QFT-GIT and an in-house ELISPOT, with a “classic” 6-day lymphocyte stimulation test (LST) and the TST for the diagnosis of latent M. tuberculosis infection. As we aimed to assess the effect of vary- ing IGRA formats and in vitro incubation periods on test outcome, the same set of M.

(9)

tuberculosis-specific peptides (ESAT-6, CFP-10 and TB7.7) were used in all three in vitro assays. In this study, which was described in chapter 6, we simultaneously performed the three different IGRA and TST in 27 TST-positive persons with a documented history of exposure to M. tuberculosis, 4 cured TB patients and 9 TST-negative controls.

Among TST-positive persons, the LST was more frequently positive (92%; P<.01) than either QFT-GIT (33%) or ELISPOT (46%). Both short-incubation IGRA showed a poor agreement with the TST. In contrast, agreement between LST and TST was excellent (91%, κ=0.78). While good agreement was observed between QFT-GIT and ELISPOT (86%, κ=0.71), agreement between these short-incubation IGRA and 6-day LST, based on same M. tuberculosis-specific peptides, was low at 60% (κ=0.31) and 59% (κ=0.27), respec- tively. Individual discordant results consisted mainly of positive TST and LST responses in association with negative QFT-GIT and ELISPOT responses. This marked discrepancy between prolonged- versus short-incubation IGRA could not be explained by arbitrary differences in IFN-γ cut-off level, as a substantial number of persons with negative QFT- GIT and ELISPOT responses had high IFN-γ responses in LST. Furthermore, in the small subgroup of INH-treated TST-positive persons, 6-day LST was positive in nearly all (6/7), while only few were positive in the short-culture IGRA, which was most marked for QFT- GIT with only 1/8 positive.

The limited data that are available in the literature with direct comparison of IGRA with short- versus prolonged culture are in accordance with our findings. Among TB contacts in the Gambia, 22/33 (67%) were QFT-GIT positive while after incubating the diluted whole blood for 6 days as many as 31/33 (93%) had a positive test result (17). Another recent study showed that among BCG vaccinated persons a PPD based prolonged-culture assay was more often positive than PPD based short culture assays(18). Further, in a small group of M. tuberculosis infected individuals they also observed that TST was best correlated with prolonged culture IGRA (19).

In conclusion, we found a remarkable discrepancy between the outcome of the two short-incubation IGRA, i.e., QFT-GIT and ELISPOT, on the one hand and the LST with a prolonged incubation and the TST on the other. It is known that a positive TST due to infection with M. tuberculosis as a rule remains positive thereafter (5). The readout of the TST is induration, reflecting influx of cells in the skin rather than IFN-γ production per se, and is read after 48-72 h allowing ample time for effector as well as memory cells to respond. The kinetics of the TST therefore resemble more closely that of the 6-day LST which could explain the high level of interassay agreement between TST and LST and the disagreement between these two assays and the short-term culture assays. Based on the above findings we hypothesize that short-incubation IGRA mainly detect the presence of effector T cells mostly reflecting recent or ongoing infection with M. tuberculosis, while prolonged-incubation IGRA, as well as TST, are also able to detect central memory T cells

(10)

Chapter 10 and are therefore more sensitive for detection of past latent infection as well. This concept

will be further referred to in the general discussion.

Can IGRA contribute to diagnosis of infection with M. tuberculosis in immunocompromised persons?

Anti-TNF-α treatment is known to increase the risk to progress from latent M. tubercu- losis infection to active TB disease (20-22). Therefore, accurate diagnosis of latent M. tu- berculosis infection in individuals eligible for this treatment is of utmost importance. In addition, the risk of immediate progression to TB disease following exposure during anti- TNF treatment can be expected to be extremely high (23). The TST is of limited value for diagnosis of active or latent M. tuberculosis infection in this population because of its reduced sensitivity during immune suppression. Although short-culture IGRA are most likely not suitable to exclude remotely acquired latent M. tuberculosis infections, as has been discussed in the previous chapter, IGRA might contribute to improved diagnosis of recent infection with M. tuberculosis in this setting.

The case presented in chapter 7 illustrates several important issues related to M. tuber- culosis infection in individuals on anti-TNF-α treatment. A 24-y-old man with Crohn’s disease, who was treated with infliximab for several months, was exposed to a case of smear-positive TB. Soon after exposure he complained of malaise, dry cough and weight loss. Despite a normal chest radiography and negative TST, TB disease was considered the most likely diagnosis. Awaiting other laboratory results, a strongly positive QFT-GIT result contributed to the decision for empirical treatment of TB. Acid-fast staining and PCR on broncho-alveolar lavage fluid were negative but M. tuberculosis was later isolated.

After TB treatment was initiated and infliximab was discontinued, the patient developed a distinct immune reconstitution syndrome.

This report demonstrated several diagnostic pitfalls of M. tuberculosis infection in patients on anti-TNF-α treatment. First, the clinical presentation can be atypical with common clinical signs and symptoms of TB being masked, as was illustrated by the normal chest radiograph in this patient with ZN-negative, culture positive pulmonary TB.

Secondly, IGRA can contribute to the diagnosis of recent M. tuberculosis infection in immunocompromised individuals, which was demonstrated by the positive QFT-GIT despite of a negative TST and a negative result of acid-fast staining and PCR on broncho- alveolar lavage fluid. Thirdly, the case clearly demonstrated that TNF-α antagonists increase the risk of early progression to primary TB after de novo exposure to a case of contagious TB. And finally, discontinuation of TNF-α antagonists in patients with TB can cause an immune reconstitution syndrome which is comparable to the phenomenon in HIV positive TB patients after starting highly active anti-retroviral treatment.

In accordance with the observation in this case report that IGRA results may be less influenced by immune suppression than the TST, are a number of studies in HIV infected

(11)

individuals. Both QFT-G(IT) and ELISPOT performed well in HIV infected individuals (24-28). The sensitivity of IGRA in HIV-positive persons was superior to TST for detection of active TB as well as latent M. tuberculosis infection. Only in case the CD4 counts was below 200 the sensitivity of IGRA was somewhat reduced, which was more pronounced for QFT-GIT than ELISPOT (27). The latter observation was most likely related to intrinsic differences in IGRA format, as ELISPOT uses a standardized number of PBMC while whole blood IGRA use a standard volume of blood, not correcting for the potentially low concentration of CD4 T cells in HIV patients.

Part II. CELLULAR IMMUNITY DURING LATENCY

From the studies described in part 1 and from numerous previous studies it is clear that IFN-γ responses to ESAT-6 and CFP-10 are observed both in TB patients and in latently infected individuals and T cell responses to these antigens therefore do not discriminate between active disease and latent infection. ESAT-6 and CFP-10 are present in abundance in the culture filtrate of in vitro M. tuberculosis cultures and immune responses to these antigens are present early in the course of infection. However, data from the studies in part 1 (chapter 4-6) suggest that in part of persons with past latent infection direct effector-T cells to these antigens are no longer circulating, as indicated by negative short-culture IGRA. Recently, accumulating evidence from in vitro models of latency indicate that M. tuberculosis changes its metabolism in order to adapt to conditions of nutrient and oxygen deprivation (29-31). During such unfavourable conditions in which M. tuber- culosis is not replicating, different proteins are expressed of which a 16kDa-α-crystallin protein (Rv2031c) was most strongly upregulated (30;32;33). In the study described in Chapter 8 we compared cellular immune responses to Rv2031c (α-crystallin) with re- sponses to ESAT-6 in TB patients, in contacts with well-documented recent exposure to TB, and in healthy individuals with and without evidence of prior infection with M.

tuberculosis. Study participants originated from two high TB-endemic regions and from one low-endemic region. The Ethiopian cohort consisted of 44 active TB patients (TB), 82 healthy household contacts who were recently exposed (HHC) and 20 community controls (CC). The Gambian cohort included 12 TB patients, 32 HHC and 20 CC. The Netherlands cohort consisted of 19 TB patients (9 active TB and 10 cured TB), 23 mostly untreated latently infected (TST-positive) individuals (Dutch HHC), and 13 healthy, non- BCG-vaccinated, TST-negative controls (Dutch CC). IFN-γ production in responses to the antigens was measured by ELISPOT (Gambia and the Netherlands) or by lymphocyte stimulation assay and ELISA (Ethiopia). ESAT-6 was recognized by most TB patients, African HHC, Dutch HHC and African CC. Since the African CC group was selected as much as possible to avoid recent TB exposure, the responses to ESAT-6 in this group were

(12)

Chapter 10 most likely explained by more remote latent infection (CC/LTBI). Although α-crystallin

was also frequently recognized in all groups, responses were significantly stronger in the African CC/LTBI groups, than in TB patients or their recently exposed HHC. Even in the subgroup of TB patients with high ESAT-6 responses, the response to Rv2031c was significantly lower than responses observed in the CC group with a high-ESAT-6 response.

This indicates that the lower α-crystallin responses observed in the TB patients cannot be attributed to a general reduced T cell responsiveness, as can occur during active TB disease. Results from Dutch study groups showed that ELISPOT responses to ESAT-6 tended to be higher than that to α-crystallin in TB patients, while in the group of HHC the median responses to ESAT-6 and α-crystallin did not differ significantly. When the ESAT-6/α-crystallin response ratio was analyzed, a consistent gradient was observed in the study groups from all 3 cohorts, with a high ratio in TB patients, and intermediate ratio in African HHC with recent exposure and the lowest ratio in Dutch HHC and African CC (mostly consisting of persons with past latent infection).

In conclusion, this study that included persons from high- and low-TB endemic regions showed that α-crystallin, which is known be strongly expressed during non- replicating persistence of M. tuberculosis in vitro, was especially well recognized by T cells from latently infected individuals, lending support to the notion that α-crystallin is indeed expressed by M. tuberculosis during latent infection in humans. Data from all three different cohorts in our study indicated a preferential recognition of ESAT-6 by TB patients, while in latently infected individuals the immune response was targeted more towards α-crystallin. This finding suggests that the profile of antigen-specific responses is correlated with the outcome of infection and this could be a rational starting point for the design of improved TB vaccines.

By studying DNA expression profiles of M. tuberculosis, a regulon of 48 genes was found to be upregulated during nitric oxide- or hypoxia-induced growth arrest (31). This “dormancy regulon” (or DosR regulon) includes α-crystallin, precisely the antigen that was well recognized by latently infected individuals as described in the previous chapter. The last study, described in Chapter 9, aimed to evaluate the immunogenicity of 25 proteins encoded by the most strongly upregulated genes of the DosR regulon, which are further referred to as latency antigens. IFN-γ production in response to these 25 recombinant latency antigens was measured first using M. tuberculosis specific T cell lines and sub- sequently by using PBMC from 20 TB patients, 23 latently infected individuals and 21 healthy controls. The results showed that 18 latency antigens were recognized by M.

tuberculosis infected individuals, indicating expression of these antigens during natural infection. Differential analysis showed that TST+ individuals recognized more latency antigens and with a stronger cumulative IFN-γ response than TB patients, while the opposite profile was found for culture filtrate protein-10, as a representative antigen that is expressed during the growth phase of M. tuberculosis. Rv1733c, Rv2029c, Rv2627c and

(13)

Rv2628 induced particularly strong IFN-γ responses in TST+ individuals, with 61%, 61%, 52% and 35% responding, respectively. Interestingly, several latency antigens were also recognized by healthy controls, but these responses coincided with strong responses to M. tuberculosis lysate, indicating that they were most likely induced by cross-reactivity to environmental mycobacteria.

In summary, several new M. tuberculosis antigens within the DosR-regulon were identi- fied. These latency antigens were preferentially recognized by latently infected individuals suggesting that T cell responses to latency antigens may contribute to protection against reactivation of TB, warranting their further study as vaccine candidates.

(14)

Chapter 10 GENERAL DISCUSSION

Despite its shortcomings, the TST had been the only available tool for diagnosis of latent M. tuberculosis infection during the previous century. The recent discovery and develop- ment of in vitro M. tuberculosis-specific immunodiagnostic assays for detection of latent TB has therefore been regarded as an important step forward in the fight against TB.

The studies described in this thesis helped to contribute to further improvement and understanding of IGRA. The study in chapter 2 describes the development of a highly specific M. tuberculosis peptide mixture from several new M. tuberculosis antigens. One of the peptides identified in this study (TB7.7 p4) has now been added to ESAT-6 and CFP-10 in the latest version of a commercial IGRA (QFT-GIT), which further optimized the sensitivity of this assay (2). Results from the study among laboratory technicians who were accidentally exposed to M. tuberculosis suggested that ELISPOT responses to TB37.6 peptides could be a marker for recent infection. As yet, there have been no other reports of studies addressing this hypothesis but if it holds true TB37.6 responses could be of practical value to discriminate between past and recent infection and thereby to the as- sessment of the need of prophylactic treatment. The studies in chapter 4 to 6 concentrated on the performance of short-incubation IGRA for the detection of latent M. tuberculosis infection. We showed that the specificity of QFT-GIT was not jeopardized by prior TST administration when QFT-GIT is performed on the day of TST reading. Further, in a large scale contact investigation among BCG-unvaccinated individuals QFT-GIT and T-SPOT.

TB were found to correlate better to the level of exposure than did TST. An unexpected observation from our studies was the large number of negative short-incubation IGRA results in individuals in whom latent TB infection was very likely. Finally, chapter 6 provided a plausible explanation for the unresolved issue of frequent discordance between TST and short-incubation IGRA, which is discussed in detail below.

In recent years several other studies also evaluated the performance of IGRA and TST for the detection of latent M. tuberculosis infection under different circumstances; in high and low TB endemic regions, among BCG vaccinated and unvaccinated individuals, in contact investigations, screening of high risk groups such as health care workers and pris- oners (9;12-15;34-41). And yet, no straightforward answer can be given to the question which assay is best. The reason for this is that it is not clear how the sensitivity of IGRA for diagnosis of latent M. tuberculosis infection should be best estimated, as discussed in detail below. In contrast, there is not much debate about how the specificity of IGRA should be estimated and IGRA are generally agreed to be highly specific for M. tuberculo- sis infection, i.e. they are almost always negative in the absence of such infection. A recent meta-analysis by Menzies et al. found a pooled specificity of 97% for QFT-G (n=711) and of 92% for ELISPOT (n=229), based on studies performed in populations at very low risk

(15)

for latent M. tuberculosis infection (10). The high specificity of IGRA was not compro- mised by BCG vaccination and was clearly superior to the pooled specificity of TST in a BCG vaccinated population (56%) (10). Although IGRA are very specific for detecting M. tuberculosis infection, they cannot discriminate between active disease and latent in- fection. This substantially reduces the specificity of IGRA when used for the diagnosis of active TB disease (42;43), in particular in high endemic settings because positive results due to latent TB infection could be regarded as false-positive in that setting. Thus, when sensitivity and specificity of an assay are discussed, it is important to precisely indicate which condition must be detected.

Estimating the sensitivity of IGRA for latent M. tuberculosis infection is complicated for two reasons. First, reliable calculation of the sensitivity is hampered by the absence of a gold standard. Secondly, how do we define latent M. tuberculosis infection and what do we want the assay to detect? Should it detect all persons ever infected with M. tuberculosis, or only recently infected persons, or only the presence of viable M. tuberculosis or, finally, should it predict those individuals at risk to develop active TB? And in the latter case it becomes relevant what kind of persons are screened as in immunocompetent persons the detection of recent infection is most important, while in immunocompromised persons it is essential to also detect more remotely acquired infection.

Latent M. tuberculosis infection is best defined as the presence of viable M. tuberculosis in an individual without signs and symptoms of disease. The ideal diagnostic assay would therefore be able to detect M. tuberculosis in a healthy individual. However, direct detec- tion of M. tuberculosis is hampered by the fact that in an asymptomatic person it is not apparent where the bacteria reside. Another complicating factor is that acid-fast staining, a classical way to detect the mycobacteria during active disease, often does not reveal the micro-organism during latency (44).

In 1909 Von Pirquet first introduced the term latent TB, when he observed TST reac- tions of ≥ 5mm in children who did not manifest TB disease (45). As some of those children later progressed to develop active TB, a positive TST was thought to be indica- tive of the presence of viable M. tuberculosis. Thus, latent TB was defined as the presence of a positive TST in an individual without signs and symptoms of disease. However, the TST does not directly detect the pathogen but instead measures the presence of a cellular immune response to M. tuberculosis. From studies of other pathogens it is well known that specific memory T cells remain present long after clearance of the pathogen. Ninety percent of individuals who become TST positive after exposure to a case of contagious TB will never develop TB disease. Although those persons usually remain TST positive, it is unknown whether they harbour viable M. tuberculosis for the rest of their lives or whether part of those individuals are able to clear the pathogen. In successfully treated TB patients the TST is known to remain positive, even though, most likely, mycobacteria are no longer present. This latter assumption is supported by the fact that no case of

(16)

Chapter 10 reactivation of TB has been described in INH-treated persons receiving infliximab, while

this TNF-α blocker is known to substantially increase the risk of reactivation in latently infected persons (21). Thus, assays which detect a M. tuberculosis-specific immune re- sponse do not necessarily indicate the presence of viable mycobacteria.

Until it becomes possible to detect viable M. tuberculosis during latency, there is no gold standard for the diagnosis of latent TB which remains therefore dependent on indirect assays detecting the presence of a specific cellular immune response to M. tuberculosis.

This brings us back to an important problem which is encountered while evaluating the performance of IGRA, i.e. how can we estimate sensitivity of IGRA for latent infection in the absence of a gold standard? To this purpose, the most informative and useful approach most likely consists of longitudinal studies determining the incidence of progression to active TB in untreated persons with positive and negative IGRA results. At the same time, however, this is also the most difficult and demanding approach as it requires large, long-term observational studies, in which people should not be treated for presumed latent infection and which will not be ethical in a low-endemic setting. No such studies have been published as yet.

A more practical and often used approach is to use the sensitivity for active TB as a surrogate marker for latent infection, which is conceptually debatable. A meta-analysis pool- ing such studies calculated a pooled sensitivity for active TB of 76% (n=544) for QFT-G and of 88% (n= 557) for ELISPOT (10). The limited sensitivity of IGRA observed in these studies could in part be attributed to T-cell anergy during active disease, as is known from the TST, which could lead to false negative test results. Besides the possibility of generally reduced T cell responsiveness (anergy) during extensive TB disease, it has been suggested that also homing of the antigen-specific effector cells at the site of infection could play a roll. The later hypothesis could lead to a positive response to PHA (a strong unspecific T cell stimulator) but a negative M. tuberculosis-antigen specific response, resulting in a false negative IGRA outcome in patients with active TB. This may explain relatively low sensitivity of IGRA for detection of TB disease when used in daily practice including patients with all degrees of illness (42;43). Thus, using the sensitivity in active TB as a surrogate for latent infection could underestimate the true sensitivity of IGRA for detection of latent infection. On the other hand, such an approach could also overestimate the sensitivity, as results from the study in chapter 6 indicated that short-incubation IGRA mainly detect recent or ongoing infection with M. tuberculosis but are relatively insensi- tive for past latent infection. The sensitivity of IGRA for detection of latent TB infection therefore strongly depends on the precise study population and on the time elapsed since infection and therefore cannot be expressed in a singly percentage that fits all settings.

Another possible way to evaluate the sensitivity of IGRA is to correlate test outcome to the level of exposure to M. tuberculosis, as a surrogate marker for infection. Using this approach, our study and those of others, showed that IGRA correlates well to the

(17)

level of exposure, and in low-endemic setting was found to correlate better than TST. In contrast to the findings in low TB endemic settings, one large contact investigation in a high-endemic setting showed the opposite, with TST results correlating better to the level of exposure (41). A possible explanation for the suggestion that IGRA performs less well in a high-endemic, African setting will be discussed here below.

Finally, the performance of IGRA could be evaluated by comparing the level of agreement between TST and IGRA. Many different studies followed this approach. Unfortu- nately, this approach raised more questions than it answered, as the level of agreement was found to vary widely between studies. This is probably related to difference in study- population, each with its own prevalence of prior BCG vaccination, expected prevalence of M. tuberculosis infection in the tested population, and the timing of exposure (recent, past or recurrent). In a low TB endemic setting with a high rate of BCG vaccination the agreement tended to be low (14;15), with a superior performance of IGRA compared to TST. However, in a high endemic setting the advantage of IGRA over TST was less clear, despite the often high BCG coverage in these areas. The varying performance of IGRA in different study settings is partly due to the fact that the superior specificity of IGRA compared to TST will be most pronounced in a study-population with a high BCG vaccination rate and is in part related to the prevalence of M. tuberculosis. Thus, while sensitivity and specificity are test characteristics, the positive and negative predictive value (PPV and NPV) of a test are affected by the prevalence of the infection in the tested population. For example, the PPV will be more influenced by the specificity of a test in a low-endemic setting than in a high-endemic setting. This could explain why IGRA, which were designed to be more specific than TST, have a superior PPV over TST in a low endemic setting, while the improved specificity of IGRA is hardly of influence on the PPV in a high endemic setting. Furthermore, in a population with a high prevalence of the infection, a test with a lower sensitivity will strongly reduce the NPV of the test, while in low-prevalence setting changes in sensitivity have little effect on the NPV. Therefore, the suggestion that IGRA perform less well when tested in a population with a high preva- lence of M. tuberculosis, than when tested in population with a low prevalence, support the notion that IGRA are less sensitive than TST.

Although the level of agreement between TST and IGRA tended to vary between studies, as discussed above, most studies observed a considerable discordance, mostly consisting of negative IGRA results in TST positive individuals. This finding was often attributed to false positive TST results in BCG vaccinated individuals. However, some studies already indicated that this was not likely to be the only explanation, as the low rate of positive IGRA results in their studies was not in accordance with the expected prevalence of latent TB infection in the tested population. In two cross-sectional studies in South Africa, ap- proximately one third of adults with a TST > 15 mm had a negative QFT-GIT (2;27) and

(18)

Chapter 10 38% a negative T-spot.TB (27). Another study found that, in a mostly BCG-vaccinated

Korean control population, 51% was TST positive and only 4% were QFT-G positive, while the expected prevalence of latent M. tuberculosis infection was 33% (35). From these data it could be concluded that IGRA detect only a subgroup of latently infected individuals and could be less sensitive for detection of past latent infection. In this regard it is interesting that a good agreement was observed between TST and QFT-GIT in health care workers in India, despite a high prevalence of BCG vaccination (34). Recurrent M.

tuberculosis exposure leading to persistent presence of effector T-cells could be a plausible explanation for the good performance of QFT-GIT in these health-care workers.

The above suggestion that discordant test results between TST and IGRA can not be solely attributed to false positive TST resulting from cross-reactivity to BCG, is confirmed by data from our contact investigation, which showed that roughly half of the BCG-unvaccinated persons with a TST of ≥ 15 mm had negative QFT-GIT and ELISPOT results.

The study in chapter 6, together with information from publications concerning other intracellular pathogens and basic immunology, provided a plausible explanation for the above mentioned unresolved issue of discordant test results of TST and IGRA. This study showed that negative QFT-GIT and ELISPOT responses in TST-positive persons were not explained by false positive TST results because the positive results of 6-day LST, using the same M. tuberculosis-specific peptides, unevoqually indicated M. tuberculosis-specific immune responses (46-49). The contradiction in the results of QFT-GIT and ELISPOT on the one hand and 6-day LST (and TST) on the other, should therefore be explained on an immunological basis. Differences in IFN-γ detection level between the assays are not likely as the underlying mechanism. Firstly, ELISPOT technique is specifically designed to be a highly sensitive IGRA, as it is able to detect one IFN-γ releasing cell in a well of 250.000 PBMC. In comparison, the LST measures the IFN-γ production in the supernatant of only 150.000 cells per well. Secondly, we showed that high levels of IFN-γ could be measured by LST in ELISPOT/QFT negative persons, indicating that there was not a problem of arbitrary cut-off levels of the assays. More likely the observed discrepant results were due to intrinsic differences in assay format. In both the studies in chapter 5 and 6, the observed agreement between QFT-G(IT) and ELISPOT was high, even though these assays differ in several respects with the first using whole-blood and measures IFN-γ by ELISA in the supernatant, while the latter uses isolated PBMC and measures numbers of IFN-γ releasing cells. The only common characteristic is the short-incubation period of 16-24 hours and in this aspect they both differ from the LST, which uses PBMC as in ELISPOT and measures IFN-γ by ELISA in the supernatant like QFT, but which uses an incubation period of 6 days. The incubation period was therefore the most likely determinant of the difference in test outcome.

(19)

It has recently been demonstrated that, besides naïve T cells, two different subsets of antigen specific CD4 memory T cell can be found in peripheral blood, consisting of CD4 cells that express chemokine receptor 7 (CCR7+ cells; central-memory T cells) and CCR7 negative cells (effector-memory T cells) (50). Upon stimulation, effector-memory T cells directly started to produce IFN-γ, while the central-memory cells did not produce IFN-γ immediately. When central-memory cells were stimulated and expanded for ten days did they differentiate into CCR7- cells and acquired the capacity to produce IFN-γ (50). This finding suggests that short-incubation IGRA will not be able to detect central-memory T cell as their readout is IFN-γ production within 24 hours. In contrast, both pools of memory cells would be able to contribute to the IFN-γ produced in a 6-day LST. This is in agreement with the findings of a study on hepatitis C in which central memory T cell responses were not detected by overnight-ELISPOT while they contributed to antigen-induced IFN-γ production in prolonged culture-ELISPOT. The discordant test results between short- and prolonged-incubation IGRA could hence be explained by the kinetics of different immune-cell populations in the course of infection. At an early stage of infection, naïve T-cells undergo profound changes when they encounter pathogenic epitopes and differentiate into effector T cells. After clearance of the pathogen most of the expanded pool of effector T cells die, but some will persist as memory cells. Thus, in individuals who were infected with M. tuberculosis in the past, results of short-incubation assay could be negative due to a low number of circulating effector cells.

Nevertheless, some subjects with past infection had positive results in the short incubation assays (chapter 6). This could represent immunological memory at a high level, also including long term survival of effector-memory cells, in persons with initially very vigorous immune responses (51). In agreement with this idea is the observation of persistent strongly positive results of short-incubation IGRA in treated TB patients or latently infected subjects. Also in support of this latter hypothesis is the observation that rever- sion of IGRA results, with or without treatment, mostly occurred in persons with initially low or moderate IGRA responses (52). Alternatively, positive IGRA results in persons who acquired M. tuberculosis in the past may reflect ongoing antigenic stimulation main- taining a sufficient circulating pool of effector T cells. Only limited data are available to directly support this hypothesis in the field of M. tuberculosis research. Nonetheless, studies on hepatitis C and HIV show that overnight-ELISPOT responses correlate well to the viral load (53). Some indirect evidence suggests that the latter correlation also holds true for M. tuberculosis. In a large Gambian study, an association was found between quantitative overnight-ELISPOT responses to PPD and mycobacterial load, using the level of exposure to M. tuberculosis as a surrogate marker (54). Furthermore, three studies indicated a trend towards decreased ELISPOT responses at the end of treatment for latent TB (52;55;56). However, in Indian health-care workers, QFT-GIT remained positive after INH-treatment, which would argue against a correlation between mycobacterial load and

(20)

Chapter 10 QFT responses, but these individuals continued to be exposed to cases of pulmonary TB

and reinfection could have played a role (57). As data are rather inconsistent, further studies are needed to evaluate the kinetics of different IGRA during treatment. Such studies should best be performed in a low TB- endemic setting in order to avoid interference caused by re-infection.

Thus, in case of past M. tuberculosis infections, it still remains unclear whether IGRA outcome correlates to antigenic load currently present in the host or whether it merely reflects the strength of the memory response (including effector-memory cells) which remained after the primary infection. It could be hypothesized that negative results of short-incubation IGRA in persons with a positive TST and 6-day LST are not false negative but instead indicate absence of viable or replicating mycobacteria in the presence of a memory response which remained after clearance of the pathogen. Although it has been proven through DNA fingerprinting that M. tuberculosis can reactivate more than 30 years after primary infection, indicating a long period of latency (58;59), it can not be proven that everybody remains latently infected for life. In line with the course of infection of other intracellular pathogens such as hepatitis B or C, it could well be assumed that some people are able to completely clear M. tuberculosis. The observation that “only”

a part (estimated to be 22%) of the latently infected individuals who are treated with the TNF-α blocker infliximab develop active TB, while the risk of immediate progression to TB following exposure during infliximab treatment is estimated to be much higher (23), could lend support to this assumption. In agreement, only a relatively limited number of untreated HIV-infected immigrants from TB-endemic regions develop active TB, while many of these immigrants are likely to be exposed to M. tuberculosis in the past. Even though it could well be possible that some individuals are capable of clearing M. tubercu- losis, it is not likely that this explains all discordant TST and IGRA results. A recent case illustrated that a negative IGRA does not exclude the presence of viable M. tuberculosis; a liver transplant patient, with a negative IGRA result pre-transplantation, developed active TB soon after transplantation which was suggestive for reactivation of a latent M. tuber- culosis infection (60). A more plausible explanation for the low number of circulating ESAT-6/CFP-10/TB7.7-specific effector cells in individuals with past latent infection could be a change in protein expression profile of M. tuberculosis during latency, resulting in a different repertoire of M. tuberculosis antigens available for T cell recognition, and in turn in a change in antigen-specific effector cells. This hypothesis formed the basis for to the two studies described in part 2 of this thesis.

In summary, it can be concluded that short-incubation IGRA are likely to be highly sensitive for detection of recent infections, but are less sensitive for past latent infection, while a positive TST (or prolonged-incubation IGRA) indicates M. tuberculosis infection at any time in the past. In line with this is our observation that IGRA correlated better to the level of exposure, as a surrogate marker for recent infection, than the TST in a BCG-

(21)

unvaccinated population. From the above it could be hypothesized that short-incubation IGRA better predict the risk of reactivation than the TST in immunocompetent persons, as the risk of developing active TB disease is highest in the first two years after infection and declines to a low level during subsequent years. If this hypothesis holds true, the use of short-incubation IGRA the selection of individuals eligible for chemoprophylaxis would substantially diminish the number of persons needed to treat to prevent one case of active TB. However, in case of more remote latent infection short-incubation IGRA can be negative and the risk of reactivation to active TB in these persons remains to be eluci- dated. As a consequence, TST or prolonged-incubation IGRA, and not short-incubation IGRA, would be the assays of choice for screening of persons in whom detection of past latent infection is relevant as well, e.g. in immunocompromised patients or those eligible for immune suppressive drugs or transplantation.

In addition to the assumption that short-incubation IGRA have a reduced sensitivity for the detection of past latent infection, several other possibilities could, at least in part, explanation the occurrence of negative IGRA results in TST+ persons. By definition, IGRA measure M. tuberculosis-specific immunity using IFN-γ as a read-out. However, a nega- tive IFN-γ response does not necessarily excluded the presence of M. tuberculosis-specific T cells, as these cells may produce other cytokines upon antigen specific stimulation, such as TNF, IL2, IL22, IL4δ2, or IL10, IL4. In line with this is the observation that during treatment of active TB patients the cytokines production by CD4 T cells, in response to ESAT-6 and CFP-10, shifted from predominantly IFN-γ to predominantly IL2 production (61). Further, high levels of Il4δ2 were found in healthy latently infected individuals (62-64).

The following suggested explanations probably play a more modest roll. First, the sensitivity could vary with differences in genetic background of the tested population, because difference in HLA types could result in a different ability to recognize the peptides used in IGRA. Even though ESAT-6 and CFP-10 tend to be broadly recognized, one study indicated that the sensitivity of IGRA for detection of active TB was found to differ significantly between ethnical groups (43). The sensitivity of IGRA could hence be improved by increasing the number of antigens or peptides used in these assays to assure optimal recognition in a genetically heterogenic population. Using this approach (chapter 2) we showed that adding new M. tuberculosis-specific peptides to ESAT-6 and CFP-10 can further improve the sensitivity. Secondly, differences in strains of M. tuberculosis complex could negatively affect the sensitivity of IGRA. ELISPOT responses to ESAT-6 were found to be attenuated in patients infected with M. africanum as well as in their household contacts (65). Since M. africanum accounts for up to half of the cases of pulmonary TB in west-Africa, this could substantially affect the performance of IGRA based on ESAT-6 and CFP-10 in that region. Thirdly, IGRA sensitivity could be reduced due to decreased T-cell responsiveness; e.g, in older age, malnutrition and in immunocompromised

(22)

Chapter 10 persons. However, studies in HIV infected individuals indicate that HIV infection did

not influence the performance of IGRA, except in case of very low CD4 counts (25-28).

Finally, suboptimal cut-off levels of IGRA could contribute to its limited sensitivity. For example, the cut-off for QFT-G(IT) was based on a Japanese study among 119 patients with active TB disease and 216 controls (66). For latent TB infection this cut-off might not be optimal. Results of the large contact investigation study described in Chapter 5 showed that lowering the cut off level of QFT-GIT would increase sensitivity without significant loss of specificity. Results from a recent study among TB patients were in accordance with this finding (67).

In conclusion, based on the studies in this thesis and current literature, IGRA can be con- sidered highly specific for detection of M. tuberculosis infections, in particular in a BCG vaccinated population, and shows a better correlation with exposure in a low-endemic setting. At present, IGRA is a valuable new tool for screening and targeted treatment of latent TB in particular in a study population with a high rate of BCG vaccination and in low-endemic settings. However, the varying performance of IGRA in different study settings and the result of several of the studies described in this thesis which suggest a limited sensitivity of short-incubation IGRA for past latent infection are reason for concern. It remains unresolved whether negative IGRA results in persons with past latent infection truly reflect the absence of viable or replicating mycobacteria or, maybe more likely, whether it reflects a low number of circulating ESAT-6/CFP-10-specific effector T cells while dormant yet viable M. tuberculosis is still present. This is particularly relevant for subjects with an increased risk of late reactivation due to impaired immunity, e.g. due to HIV infection or treatment with immunosuppressive drugs. Therefore, longitudinal studies determining the negative and positive predictive value of IGRA for TB risk are ur- gently needed. These studies should compare different IGRA format, such as short- versus prolonged incubation IGRA and should be done in different study settings and populations (eg contact investigations, screening of high risk groups, immunocompromised patients, high and low endemic settings.) Such studies would demonstrate which assay provides the best basis for therapeutic decision making under various circumstances.

Furthermore, future studies should aim to develop IGRA specifically design for detection of latent infection, including M. tuberculosis specific antigens which are upregulated in M. tuberculosis during latency.

The above section relates to the detection of latent TB infection and it was clear that responses to RD-encoded specific antigens such as ESAT-6 and CFP-10 remain detect- able in short term assays only in part of latently infected persons. This suggests that the expression of these antigens decreases in the course of time in some infected persons.

Until recently little was known about how M. tuberculosis behaves during latency and

(23)

how it adapts to the surroundings where it is able to persist. In order to study the possible metabolic and physiological changes of M. tuberculosis during latency, in vitro-models were developed that aimed to accurately mimic latent infection in humans. Although the conditions in which M. tuberculosis persists are poorly defined, oxygen depriva- tion and exposure to nitric oxide (NO), which is produced by activated macrophages, are thought the be the two most important conditions that M. tuberculosis encounters during latency (reviewed in (29;68-70)). When M. tuberculosis was exposed in vitro to gradually decreasing oxygen tensions its replication stopped, a state which is referred to as non-replicating persistence (NRP). Growth resumed when oxygen was reintroduced (the Wayne model of latency;(71)). These conditions were used to study M. tuberculosis response to stress signals. (30;32;33) It was shown that M. tuberculosis displays an altered metabolic state during NRP in vitro. An important M. tuberculosis protein that was pro- duced in abundance during NRP was 16kDa, α-crystallin-like small heat shock protein (Rv2031c) (32;33). The α-crystallin protein has been associated with cell-wall thickening and is thought to operate as a chaperone, playing a role in stabilizing cell structures during long-term survival.

In chapter 8, we showed that the antigen encoded by Rv2031c was strongly recognized by latently-infected individuals, indicating that Rv2031c is actually expressed by M.

tuberculosis during latent infection in humans. Moreover, the T cell response, as mea- sured by IFN-γ, in persons with active TB or recent M. tuberculosis infection was biased towards ESAT-6, while in persons who had remained latently infected after exposure to M. tuberculosis, the immune response was targeted more towards Rv2031c. Importantly, a gradient in the ESAT-6/Rv2031c response ratio was observed, with a high ratio in TB patients, a low ratio in latently infected individuals and intermediate ratios in the group consisting of recently exposed healthy individuals. Finally, our data indicate that a strong IFN-γ response to Rv2031c correlates with a certain level of protection against TB disease, as Rv2031c was most strongly recognized by individuals with a latent M. tuberculosis infection without a recent exposure to TB. These findings justify further study of Rv2031c as a potentially interesting candidate antigen for post-exposure vaccination.

With the deciphering of the genome of M. tuberculosis and availability of advanced molecular techniques, such as micro-arrays, it has become possible to study changes in genetic transcripts in M. tuberculosis in models of latency, including oxygen depletion (Wayne model). A novel approach was followed by Voskuil et al., who exposed M. tu- berculosis to a low, non-toxic, dose of nitric oxide (NO) by using a diethylenetriamine/

NO adduct which generates nanomolar concentration of NO. NO, which is produced by activated macrophages, possesses antimicrobial properties and can irreversibly damage bacteria. However, NO was also found to act as a reversible inhibitor of aerobe respiration in bacteria, and as such could play a role in the initiation or maintenance of latency during natural infection. One study indicated that NO is indeed present in human granulomas

(24)

Chapter 10 a)

load Antigen

bacterial l n-specific

Mycob c T cells

Primary infection latent infection

Immune response induced dormancy of MTB Immune response induced dormancy of MTB

b)

load Antigen

Mycob c T cells

primary infection latent infection reactivation

c)

load Antigen

Mycob c T cells

primary infection latent infection clearance of mtb Post-exposure vaccination boosting responses to latency antigens

Figure 1. Suggested models of cellular immune responses during M. tuberculosis infection.

a) Model of latent M. tuberculosis infection, b) model of reactivation of TB, c) model for post-exposure vaccination. Abbreviation: MTB, M. tuberculosis.

MTB in growth phase MTB dormant, non replicating Early secreted antigens specific- effector T cells

Dormancy antigens specific- effector and memory T cells effector T cells

central- and effector-memory T cells

(25)

(72). Exposure of M. tuberculosis to low dose of NO in vitro reduced the respiration and replication rate and upon removal of NO both these effects were found to be reversible, a process that is very similar to that observed in response to oxygen depletion.

Using microarray for comparison of DNA expression profiles of M. tuberculosis during log-phase growth and during NO induced growth arrest, Voskuil et al. identified 48 genes that were upregulated under this condition (31). Interestingly, the same set of 48 genes was found to be upregulated by gradual reduction of oxygen. One of these genes was Rv2031c, encoding α-crystallin, which had been identified earlier as the most strongly expressed protein during hypoxia (32). They further showed that the two-component response regulator, dormancy survival regulator (DosR), encoded by Rv3133c, induced the transcription of this gene set. Previously, DosR was already found to be required for induction of 16kDa-α-crystallin by hypoxia (30). Genes of this DosR regulated gene set were also found to be upregulated when M. tuberculosis was grown in activated murine macrophages in vitro (73). Both in the lungs of chronically infected mice and of patients with chronic, active TB the transcription pattern of M. tuberculosis showed characteris- tics of NRP (74;75). In view of the above results this set of 48 genes was designated the

“dormancy regulon”.

The study described in chapter 9 demonstrated specific cellular immune responses directed towards dormancy proteins (or latency antigens) in infected humans and was the first to provide evidence that proteins of the dormancy regulon are not only expressed by M. tuberculosis in vitro but are actually expressed during latent infection in humans.

TST+ individuals recognized more latency antigens and with a stronger cumulative IFN-γ response than TB patients, while the opposite profile was found for culture filtrate protein-10, an antigen which is strongly expressed during log-phase growth of M. tuber- culosis in vitro.

The finding of T cell responses to α-crystallin and several other latency antigens in TB patients may seem contradictory to the antigen being expressed during latent infection, but this can be explained by the fact that TB disease is almost universally preceded by a period of latency such that immune responses to latency-associated antigens can be expected to occur during that phase. Further, another study indicated that the expression of latency antigens may not be limited to true latent infection alone, but can occur during chronic TB disease, as α-crystallin was found to be expressed in the lungs of patients with chronic active TB (75).

Responses to latency antigens were also observed in healthy uninfected controls and our finding suggests these responses could result from cross-reactivity to non-tuberculous mycobacteria. It is interesting to hypothesize that these potential cross-reactive immune responses to latency antigens among M. tuberculosis naïve persons could contribute to the natural protection that develops in 90% of individuals who are infected with M. tuber- culosis and remain free of disease. This is may be an important topic for further studies.

(26)

Chapter 10 The findings of the studies described in chapter 8 and 9 have two important implica-

tions. First, the differential recognition of latency antigens and early secreted antigens (e.g.: ESAT-6, CFP-10 and Ag85B) by TB patients versus latently infected individuals could be of use for improving immunodiagnosis of latent TB. It is conceivable that in latently infected individuals with negative short-incubation IGRA responses to ESAT-6 and CFP-10, latency antigen-specific T cell responses can be found in persons who have developed protective immunity and are at no or very low risk of reactivation TB in the absence of immune suppression. (Figure). Thus, the inclusion of latency antigens in immunodiagnostic assays could increase the sensitivity for the diagnosis of past latent infection and perhaps even be used as a surrogate marker for protection in non-immunocompromised persons. Ideally, such diagnostic latency associated antigens would be M. tuberculosis-specific, as many antigens encoded by the DosR regulon also occur in apathogenic mycobacteria and T cell responses to these antigens can at present not be regarded as proof of actual M. tuberculosis infection.

The second implication of the findings of the last two studies of the thesis concerns the possibility that cellular immune responses to latency antigens could represent correlates of protection (Figure 1). M. tuberculosis exposed individuals who did not develop TB after primary infection but remained latently infected apparently developed a natural protective immune response to M. tuberculosis. This course of events is thought to pre- vent reactivation of TB infection in 90% of all infected persons and the occurrence of progression to TB or reactivation TB could be regarded as an exception rather than the rule. The study of M. tuberculosis specific responses in persons with natural immunity is very important as it could potentially provide correlates of protective immunity against TB and may guide further studies on the development of an improved TB vaccine.

From the long and extensive experience with BCG vaccination it is clear that BCG does not protect adequately against establishment of latent TB nor against reactivation from latent infection to active disease. In this regard it is relevant that BCG, which has a DosR regulon nearly identical to M. tuberculosis, does not induce an adequate cellular immune response to the most abundantly expressed latency antigen Rv2031c, that was the subject of Chapter 8 (76;77). Further, also in mice, BCG vaccination was not able to induce cellular immune responses to Rv2031c, while vaccination with the recombinant protein of Rv2031c was able to induce such responses (76). A similar observation was done for the 8 most immunogenic latency antigens, as in BCG vaccinated uninfected persons immune responses to these antigens were practically absent (78). Furthermore, vaccination of mice with BCG did not induce any immune responses to these latency antigens (78). These data indicate that the expression of latency antigens by BCG is de- ficient, which may contribute to the lack of protective efficacy of BCG in preventing late reactivation of TB. Recently, it was shown that strong cellular Th1 type immune responses can be induced against several of the latency antigens, that were identified in our study