Correlation of Mycobacterium Tuberculosis Specific and Non-Specific Quantitative Th1 T-Cell Responses with Bacillary Load in a High Burden Setting

Background Measures of bacillary load in patients with tuberculosis (TB) may be useful for predicting and monitoring response to treatment. The relationship between quantitative T-cell responses and mycobacterial load remains unclear. We hypothesised that, in a HIV-prevalent high burden setting, the magnitude of mycobacterial antigen-specific and non-specific T-cell IFN-γ responses would correlate with (a) bacterial load and (b) culture conversion in patients undergoing treatment. Methods We compared baseline (n = 147), 2 (n = 35) and 6 month (n = 13) purified-protein-derivative (PPD) and RD1-specific (TSPOT.TB and QFT-GIT) blood RD1-specific (TSPOT.TB; QFT-GIT) responses with associates of sputum bacillary load in patients with culture-confirmed TB in Cape Town, South Africa. Results IFN-γ responses were not associated with liquid culture time-to-positivity, smear-grade, Xpert MTB/RIF-generated cycle threshold values or the presence of cavities on the chest radiograph in patients with culture-confirmed TB and irrespective of HIV-status. 2-month IGRA conversion rates (positive-to-negative) were negligible [<11% for TSPOT.TB (3/28) and QFT-GIT (1/29)] and lower compared to culture [60% (21/35); p<0.01]. Conclusions In a high burden HIV-prevalent setting T-cell IFN-γ responses to M. tuberculosis-specific and non-specific antigens do not correlate with bacillary load, including Xpert MTB/RIF-generated CT values, and are therefore poorly suited for monitoring treatment and prognostication.


Introduction
Measures of bacillary load at diagnosis or after the intensive phase of anti-TB treatment are crudely associated with relapse and treatment failure in tuberculosis (TB) patients, yet better and more accurate prognostic and treatment monitoring tools are needed [1]. Such tools may identify high risk patients and facilitate the more rapid evaluation of new therapeutic interventions for TB. It has been suggested that immunodiagnostic biomarkers such as interferon-c (IFN-c) release assays (IGRAs) may serve as surrogate markers of bacillary load and treatment response [2]. Although the diagnostic accuracy of standardised IGRAs [TSPOT.TB (Oxford Immunotec, UK) and QuantiFERON TB Gold-in-tube (QFT-GIT; Cellestis, Australia)] for latent and active TB have been extensively studied [3][4][5][6][7][8][9], there are hardly any data about their relationship with bacterial load [10]. The relationship between bacillary burden and the magnitude of IFN-c in response to RD-1 specific antigens thus remains uncertain [11]. Their role as tools of prognostication and disease monitoring have also been little studied, particularly in high burden settings.
A previous study performed in a small number of HIV-infected TB patients demonstrated a crude relationship between antigenspecific T-cell responses and sputum bacillary load [10]. This finding is supported by the longitudinal increase in quantitative IFN-c responses in TB contacts prior to diagnosis with active TB [12,13], as well as declining IFN-c responses seen in TB patients in low burden settings undergoing treatment [14][15][16]. This latter aspect remains inconsistent, however, as several studies have reported persistently increased positive responses during and after treatment [17][18][19][20][21][22][23][24][25]. Thus, the role of IGRAs as tools of prognostication and disease monitoring continues to remain unclear, especially with regard to their utility in high burden settings. Evaluating the relationship between quantitative T-cell responses using both standardised IGRAs (TSPOT.TB and QFT-GIT) and bacillary load at diagnosis or during treatment has hitherto not been undertaken. These aspects are also largely unexplored with regard to non-specific antigens such as M. tuberculosis-derived purified protein derivative (PPD) and newer measures of bacillary load [such as the World Health Organisation-recommended Xpert MTB/RIF assay [26]].
To address these gaps in our knowledge we comparatively evaluated the magnitude of Th1-based M. tuberculosis-specific and non-specific quantitative T cell responses with measures of sputum bacillary load, some of which are known to be associated with treatment outcomes, prior to and during treatment in TB patients from Cape Town, South Africa.

Study Population
Sputa and blood were collected at the time of diagnosis from 513 consecutively recruited ambulatory patients with suspected TB ($18 years of age) between February 2007 and April 2010 at two primary care clinics in Cape Town, South Africa ( Figure 1). Informed written consent was obtained from all participants and the study was approved by the University of Cape Town, Faculty of Health Sciences Research Ethics Committee. Detailed patient and laboratory-specific information were recorded on a standardized case record form and captured using double data entry.

Patient Flow and Diagnostic Tests
All patients had at least one concentrated fluorescent smear and one BACTEC MGIT 960 (BD Diagnostics, USA) liquid culture performed on spot or morning sputum ( Figure 1). Culture positivity for M. tuberculosis served as a reference standard for definite TB. An HIV test was performed after appropriate counselling. When available, a sputum sample was archived and later used for Xpert MTB/RIF (Cepheid, Sunnyvale, USA) testing [27]. Blood was collected at recruitment and used for the TSPOT.TB, QFT-GIT and PPD enzyme-linked immunosorbent spot (ELISPOT) assays. Where possible, a chest X-ray [CXR; read by 2 readers blinded to clinical information using the validated CXR scoring and recording system (CRRS) [28,29]] was performed at recruitment. 35 randomly selected TB patients received, in addition to routine follow-up smear microscopy and culture, a TSPOT.TB, QFT-GIT and PPD ELISPOT test at 2 months post-treatment initiation. 13 of these patients received these tests again at 6 months. The accuracy of TSPOT.TB, QFT-GIT, Xpert MTB/RIF and CRRS for the diagnosis of TB in this cohort has previously been published [8,27,30,31].

Interferon-c Release Assays
All tests were performed according to the manufacturers' instructions on blood by laboratory technicians blinded to patient information. PBMCs were isolated from the blood and each IGRA performed within 8 hours of venepuncture. TSPOT.TB, QFT-GIT and PPD ELISPOT results were reported and interpreted as described previously [8,32].

Data Analysis
Culture positivity for M. tuberculosis from at least a single sample served as a reference standard for definite TB. Statistical analyses were performed using Graphpad Prism (version 5.0; GraphPad Software, San Diego, USA, www.graphpad.com) and OpenEpi (version 2.3.1) [33]. Statistical tests on longitudinal changes in IFN-c levels were done on a per patient basis. TSPOT.TB and QFT-GIT results defined as indeterminate according to the manufacturers' definitions were excluded. All data was also analysed stratified by HIV status. Table 1 and Figure 1 show the baseline characteristics of the cohort. Of the 513 participants, 63% (354/513) were male and 26% (133/445) were HIV-infected. 29% (147/513) of participants had definite TB, 67% (98/147) of whom were smear-positive. TB patients were significantly more likely to be younger (p,0.01), of black (p,0.01) or mixed ancestry (p,0.01), have a history of previous TB (p = 0.02), HIV-infected (p,0.05) and TSPOT.TB-(p,0.01) or QFT-GIT-positive (p,0.01).   and for each IGRA] was non-significant, irrespective of HIVstatus (data not shown).

Smear-and Culture-status of Patients Undergoing Treatment
The status of the 35 culture-positive patients who were available for IGRA-based follow-up is shown in Table 2. After 2 months of anti-TB therapy, 60% (21/35) of these individuals had cultureconverted and were identified as responders to therapy.

Longitudinal Changes in Quantitative Interferon-c Responses
Quantitative IFN-c responses to RD-1 specific antigens or PPD did not change on a per patient basis between 0 vs.   (Table 3)

Discussion
It has been suggested that IFN-c responses to RD1-specific and non-specific antigens may be useful as prognostic tools, and for monitoring bacterial load and treatment response [2]. However, there is limited information from high-burden settings. Our main aim was to evaluate, at diagnosis, the relationship between IGRAs and measures of bacterial load, which is a prognostic tool associated with treatment failure [1], in patients with TB. We further comparatively evaluated one in-house and two standardised IGRAs longitudinally in relation to 2-month culture conversion, a measure known to be associated with treatmentrelated outcome [1].
The key findings of our study were: (i) T-cell IFN-c responses do not correlate with measures of bacterial load in TB patients at We did not detect any relationship between quantitative IFN-c responses and bacillary load at diagnosis irrespective of HIV status. By contrast, a smaller study in HIV-infected TB patients in similar setting found a significant but weak correlation between quantitative TSPOT.TB responses and smear grade (a relatively crude measure of bacterial load) but none with culture TTP [10]. Cavitation on the radiograph, previously shown to be associated with either significantly elevated sputum bacillary load or RD1specific responses in low burden settings [21,34], was not associated with elevated IFN-c responses in our study. There is, Table 2. Status of TB patients at diagnosis (all culture-positive) and after 2 and 6 months of anti-TB treatment.  The poor relationship between IFN-c response and bacillary load may be a function of T cell compartmentalisation and immunosuppressive mediators including IL-10, TGF-b and regulatory T cells [35,36]. Thus, TB-specific T cells are concentrated in the lung and not in the sampled compartment, and those with the most extensive disease may have the greatest attenuation of RD-1specific Th1 immunity. Whether other antigens such as HBHA behave differently remains unclear [37][38][39][40].
Are serial IFN-c responses useful for monitoring patient-related outcomes in TB patients undergoing treatment? A comparative analysis of the different IGRA formats for treatment monitoring (or correlation of initial bacillary load) has not previously been undertaken. The majority of studies that have examined longitudinal IGRA responses have been performed in a low burden setting [17,19,23,24]. In our study, and similar to findings in The Gambia where an in-house assay was used [25], standardised IFNc responses to both RD1-specific antigens and PPD did not appreciably change longitudinally. Although the sensitivity of the T-SPOT and QFT-GIT assays are known to differ [6] we found no differences in the longitudinal readouts. In contrast to a previous study which found that RD-1, but not PPD responses, selectively diminished during the course of treatment [16] we found that responses to both antigens remained persistently elevated. Of note, however, is that this earlier study used an inhouse experimental assay which, in contrast to the commercially available TSPOT.TB or QFT-GIT tests, utilised two ESAT-6derived peptides. This, rather than setting-specific differences with our study, might account for these discordant results, especially given that the findings of the original study have since been confirmed in India, where a similar assay was used [41].
Given the poor correlation we found between IFN-c responses and culture TTP and cavitation at diagnosis, both of which are known associates of 2-month culture conversion [34,[42][43][44][45]; it is unsurprising that we also detected no association between IFN-c responses and this outcome. This high proportion of persistent T-cell responses may be due to several factors including: a history of repeated or on-going TB exposure in a high burden environment, the slow decay kinetics of T cells [24,46], or greater T cell stimulation with antigen release during the course of treatment. Indeed, history of past-exposure is likely to cause an initially high IFN-c response which, in patients with latent TB, has shown to remain largely unchanged throughout chemotherapy [46].
There are several limitations to this study. Firstly, all patients did not receive the same combination of tests due to the study design. Although this restricted our sample size, our study nonetheless remains the largest that has specifically looked at the issue of IFN-c responses and bacillary load. Second, only a subset of patients underwent 6 month follow-up and, as a result, the conclusiveness of our findings limited. Third, given the relatively high prevalence of TB and HIV co-infection in South Africa, these findings may not be generalizable to other settings. Fourth, the Xpert MTB/RIF assay in this cohort was performed on archived sputum samples, which may influence the C T values generated by the assay, however, this effect has previously not shown to be significant [27,47]. Finally, it is worth noting that culture conversion itself correlates only crudely with unfavourable treatment outcomes [1] but nevertheless remains the best surrogate biomarker of treatment outcome.
In conclusion, our data suggest that T-cell IFN-c responses to either RD1-specific or non-specific antigens do not correlate with bacillary burden or treatment outcome amongst TB patients in a high burden setting. They are hence unlikely to have utility as prognostic tools at initial evaluation or to monitor either disease severity or treatment response.  Table 3. Differences in IFN-c responses between culture converters (positive-to-negative) and non-converters after 2 months of anti-TB treatment.