Host transcriptional responses following ex vivo re-challenge with Mycobacterium tuberculosis vary with disease status

The identification of immune correlates that are predictive of disease outcome for tuberculosis remains an ongoing challenge. To address this issue, we evaluated gene expression profiles from peripheral blood mononuclear cells following ex vivo challenge with Mycobacterium tuberculosis, among participants with active TB disease (ATBD, n = 10), latent TB infection (LTBI, n = 10), and previous active TB disease (after successful treatment; PTBD, n = 10), relative to controls (n = 10). Differential gene expression profiles were assessed by suppression-subtractive hybridization, dot blot, real-time polymerase chain reaction, and the comparative cycle threshold methods. Comparing ATBD to control samples, greater fold-increases of gene expression were observed for a number of chemotactic factors (CXCL1, CXCL3, IL8, MCP1, MIP1α). ATBD was also associated with higher IL1B gene expression, relative to controls. Among LTBI samples, gene expression of several chemotactic factors (CXCL2, CXCL3, IL8) was similarly elevated, compared to individuals with PTBD. Our results demonstrated that samples from participants with ATBD and LTBI have distinct gene expression profiles in response to ex vivo M. tuberculosis infection. These findings indicate the value in further characterizing the peripheral responses to M. tuberculosis challenge as a route to defining immune correlates of disease status or outcome.


Introduction
Globally, 10.4 million incident cases of active tuberculosis (TB) disease (ATBD) and 1.4 million TB-related deaths were reported in 2015 [1]. Over the past century, anti-TB drugs, bacille Calmette-Guérin (BCG) vaccination, and public health strategies such as directly observed treatment, have contributed to a reduction in TB-related mortality [1]. However, latent TB infection (LTBI) and recurrent ATBD remain critical issues for global control [1,2], particularly given that previous anti-TB treatment is an established risk factor for drug resistance [3,4]. Determining the contribution of Mycobacterium tuberculosis re-infection, as opposed to reactivation, is a challenge for numerous TB control programs, especially in endemic areas [5]. M. tuberculosis re-infection can occur at any time during ATBD and LTBI, and is independent of relapse [5]. Moreover, the inability of vaccination to protect against M. tuberculosis re-infection and reactivation represent significant gaps in research and therapeutics [6][7][8]. The sole approved vaccine BCG and vaccines in the development pipeline are largely protective against ATBD [9], and not LTBI or exogenous M. tuberculosis re-infection [8,10]. Currently there is no approved and effective vaccine for individuals with LTBI [8,10].
The assessment of vaccine efficacy hinges on identifying biomarkers predictive of disease progression and outcomes [8,11,12]. Given the numerous constraints for TB vaccine development, it has been hypothesized that identification of peripheral correlates of protective immunity against M. tuberculosis may be more realistic, compared to a single biomarker [13][14][15]. A unique transcriptional signature has been identified for ATBD; this whole-blood transcript signature was associated with disease severity and observed to resolve after treatment [16]. The extent to which this signature is predictive rather than diagnostic still needs to be determined. Previous studies have argued that differential gene expression was associated with TB disease recurrence, susceptibility, and host control [17][18][19]. Little is known regarding transcriptional biomarkers of post-primary M. tuberculosis re-infection or even enhanced exposure due to reactivation, despite the public health significance.
Distinguishing gene expression patterns following ex vivo challenge with M. tuberculosis among individuals with ATBD, LTBI, previous active TB disease (PTBD; after successful treatment) has significance in identifying diagnostic and predictive biomarkers, which are required for development of vaccines and therapeutics [20,21]. Our study objective involved delineating patient responses to the ex vivo challenge of M. tuberculosis through analysis of the relative gene expression profiles between study participants with ATBD, LTBI, PTBD, compared to controls.

Ethical conduct of research
The Institutional Ethics Review Committees at St. John's Medical College and Hospital (St. John's National Academy of Health Sciences; Bangalore, Karnataka, India) and Arogyavaram Medical Centre (Arogyavaram, Andhra Pradesh, India) approved the study protocol. Study participants provided voluntary informed consent prior to data collection.
Definitions of TB status included: ATBD, LTBI, and PTBD. ATBD status was based on acid-fast bacilli (AFB) sputum smear microscopy, which was performed by standard protocol (Ziehl-Neelsen staining), or Xpert MTB/RIF (Cepheid, Sunnyvale, California, United States). Patients with ATBD were newly diagnosed and had not initiated anti-TB treatment (or received <1 week of treatment). LTBI was diagnosed by QuantiFERON-TB Gold In-Tube (QFT-G; Cellestis Limited, Carnegie, Victoria, Australia). Blood samples were collected in QFT-G tubes, assayed according to manufacturer protocol and by enzyme-linked immunosorbent assay. Study enrollment was based on a positive QFT-G diagnostic result. PTBD was defined as having previous ATBD (pulmonary), anti-TB treatment, and post-treatment sputum conversion. PTBD patients received 6 months of anti-TB treatment for ATBD; study enrollment occurred between 2-371 days after completing treatment. Controls were considered individuals with a negative QFT-G diagnostic result.

Sample collection and peripheral blood mononuclear cell isolation
Venous blood samples (8 mL) were collected in mononuclear cell preparation tubes (CPT™; Becton Dickinson Vacutainer Systems; Franklin Lakes, New Jersey, United States), and processed per manufacturer's instructions. Peripheral blood mononuclear cells (PBMCs) were isolated from blood samples, based on manufacturer instructions (CPT™; Becton Dickinson Vacutainer Systems; Franklin Lakes, New Jersey, United States). PBMCs were suspended at 5 x 10 6 cells/mL in cryopreservation medium (45% RPMI 1640, 45% fetal bovine serum, 10% dimethyl sulfoxide), and incubated overnight at -80˚C (in Mr. Frosty™ freezing container; Nalgene, Rochester, New York, United States). PBMCs were stored in liquid nitrogen prior to analyses.

Complementary deoxyribonucleic acid subtractive libraries
SSH libraries were created with the PCR-select cDNA subtraction kit (Clontech Laboratories, Inc.; Palo Alto, California, United States), per the manufacturer instructions. In forward subtraction (ATBD-LTBI), cDNA from the ATBD group was the tester and LTBI was the driver. Conversely, in reverse subtraction (LTBI-ATBD), LTBI cDNA was the tester and ATBD cDNA was the driver. Real-time PCR (Rotor Gene 6000; Qiagen Inc., Hilden, Germany) amplified the subtracted, unsubtracted, and control cDNA with M13 primers. Following PCR (DNA amplification), products were cloned in the PCR 2.

PCR quantification of gene expression
Gene expression of the identified sequences (from hybridization) were further assessed with real-time PCR (Rotor Gene 6000; Qiagen Inc., Hilden, Germany) in replicate. Primers were designed with NCBI Primer-BLAST (S1 Table). Each PCR reaction included: 0.3 μg cDNA, 200 nM primers, 12.5 μL 2x KAPA SYBR FAST quantitative polymerase chain reaction (qPCR; KAPA Biosystems; Boston, Massachusetts, United States). PCR cycling conditions were: 1 cycle (95˚C for 3 minutes); 35 cycles (95˚C for 30 seconds, 60˚C for 1 minute). Human acidic ribosomal protein (HUPO) was the internal control gene. Cycle thresholds (C T ) were calculated through the Rotor Gene software (version 1.7.87).

Statistical analysis
Relative gene expression was reported as fold-change, based on the comparative cycle threshold (2 -ΔΔCT ) method [22] with HUPO as the internal control gene. Specifically: where 2 À DDCT ¼ 2 À ðð½CT target geneÀ CT HUPO sample AÞÀ ð½CT target gene À CT HUPO sample BÞÞ ð2Þ [22] For each comparison, fold-change was reported as the average of the fold-changes of the two replicates (Table 1). For sociodemographic characteristics, the normality assumption of the age variable was assessed by the Kolmogorov-Smirnov test. Comparisons between the four study participant groups were evaluated by Kruskal-Wallis and Fisher's exact tests. Statistical analysis was conducted with SAS (version 9.4; SAS Institute, Inc., Cary, North Carolina, United States); statistical significance was based on alpha value < 0.05 and two-tailed tests.

Discussion
In summary, the gene expression profiles in ATBD, LTBI, PTBD, and control human study participants exhibit distinct patterns. These data provide a foundation for characterizing biomarker panels as correlates of protective immunity, which would serve as valuable surrogates in future development of TB vaccines and therapeutics.
Additionally, similar to our results other data indicated that the gene expression of chemokines (including MIP1α) may differ during ATBD. Studies have identified a role for microbial lipoproteins in stimulating cytokine production (such as IL-12) in macrophages (via Toll-like receptors) [29]. MIP has also been found to be present in infectious pathogens, such as Chlamydial trachomatis [30].
Broadly, other studies have also highlighted the potential of biomarker panels (including host protein biosignatures) as indicators of the immune response against M. tuberculosis that could have diagnostic and/or predictive value [14][15][16]31]. Furthermore, prior studies contend that differential gene expression was associated with increased risk of TB disease progression and relapse [17][18][19]. One study with whole-blood microarray analysis reported differential gene expression between patients with ATBD and LTBI, and identified gene expression profiles associated with host control of M. tuberculosis (specifically apoptosis and natural killer cell activity) [17]. Two studies indicated that differential gene expression profiles could associate with TB relapse [18,19]. These studies show that biomarker profiles have potential to be more robust than single biomarkers of TB immunity. Interferon (IFN)-gamma, for example, is perhaps the most utilized candidate TB biomarker for a protective immune response against M. tuberculosis [32][33][34][35]; however IFN-gamma has low accuracy and predictive power, especially as a biomarker of protection or disease outcome [33].
Our study has several limitations. Firstly, PBMCs were cryopreserved prior to M. tuberculosis stimulation and RNA extraction, which could affect transcriptional processes. Secondly, the interpretation of gene expression in biological function remains preliminary [36]. Additionally, given the use of 2 -ΔΔCT as a relative gene expression quantitation method [22], the findings need to be confirmed and extended with additional housekeeping controls [37]. Due to the pooling of samples in each group, there was no variance data, which restricted statistical analyses and corrections for multiple hypothesis testing. Further confirmatory studies could alternatively utilize groupwise comparisons.
Future studies are needed to elucidate the role of transcriptional signatures as immune correlates (including via single cell analysis [38,39], prior to any claims that such profiles are predictive of clinical morbidities (ATBD, LTBI) and treatment outcomes. Studies to date have assessed cross-sectional data and therefore have not allowed for causal inference, which we could potentially address in a human cohort. Furthermore, studies need to account for other potential confounding factors, including the heterogeneity of patient immune response [40,41]. Nonetheless, these data provide a useful platform in defining initial immunological themes that allow us to differentiate between human patients that fall into the active (ATBD), latent (LTBI), and active-treated (PTBD) disease classes.
Supporting information S1