Host Immune Responses Differ between M. africanum- and M. tuberculosis-Infected Patients following Standard Anti-tuberculosis Treatment

Epidemiological differences exist between Mycobacterium africanum (Maf)- and Mycobacterium tuberculosis (Mtb)-infected patients, but to date, contributing host factors have not been characterised. We analysed clinical outcomes, as well as soluble markers and gene expression profiles in unstimulated, and ESAT6/CFP-10-, whole-Maf- and Mtb-stimulated blood samples of 26 Maf- and 49 Mtb-HIV-negative tuberculosis patients before, and after 2 and 6 months of anti-tuberculosis therapy. Before treatment, both groups had similar clinical parameters, but differed in few cytokines concentration and gene expression profiles. Following treatment the body mass index, skinfold thickness and chest X-ray scores showed greater improvement in the Mtb- compared to Maf-infected patients, after adjusting for age, sex and ethnicity (p = 0.02; 0.04 and 0.007, respectively). In addition, in unstimulated blood, IL-12p70, IL12A and TLR9 were significantly higher in Maf-infected patients, while IL-15, IL-8 and MIP-1α were higher in Mtb-infected patients. Overnight stimulation with ESAT-6/CFP-10 induced significantly higher levels of IFN-γ and TNF-α production, as well as gene expression of CCL4, IL1B and TLR4 in Mtb- compared to Maf-infected patients. Our study confirms differences in clinical features and immune genes expression and concentration of proteins associated with inflammatory processes between Mtb- and Maf-infected patients following anti-tuberculosis treatment These findings have public health implications for treatment regimens, and biomarkers for tuberculosis diagnosis and susceptibility.


Introduction
Mycobacterium africanum (Maf) is an ancient lineage of the Mycobacterium tuberculosis (Mtb) Complex (MTBC), mostly found in West Africa where it causes up to half of all tuberculosis (TB) cases [1]. Apart from descriptions of the epidemiological differences between Maf and Mtb infection in the human population, differences in underlying immune responses, clinical course and outcome of TB therapy have not been described [2]. Other authors have recently attempted to define biomarkers that are able to predict treatment outcome and if validated, these biomarkers could significantly shorten trials of new TB regimens [3][4][5]. Ultimately, the performance of such biomarkers might be influenced by the infecting mycobacterial lineage. Previous studies that have assessed whether the rate of response to treatment differs between infecting MTBC lineages obtained conflicting results [6][7][8][9][10][11][12][13], but data from our own laboratory and others [6,12,13] indicate that their responses to treatment are heterogeneous. Different MTBC lineages may have been responsible for the heterogeneous response to the shorter TB treatment regimen containing Gatifloxacin recently tested in West Africa [14].
We have previously shown that although the proportion of activated T cells were similar in Maf-and Mtb-infected patients pre-treatment, they decreased significantly in Mtb-infected patients, while those of Maf-infected patients were persistently high but consisted of poorly functional T cells post-treatment [15]. In addition, the transcriptomic and metabolic profiles of Maf-and Mtb-infected patients while similar at baseline significantly differed by lineage posttreatment mainly due to changes in Mtb-infected but not in Maf-infected patients [13]. These results suggest that intrinsic host factors determine the immune response to TB and/or differential effect of the standard anti-TB treatment on the two lineages.
This study was conducted to investigate the changes in the host immune response and clinical outcomes following treatment in a larger cohort of Maf-and Mtb-infected tuberculosis patients before, during and after standard anti-TB treatment. Although we found no differences in the clinical parameters measured and found differences in only few cytokines concentration and gene expression profiles between Maf-and Mtb-infected patients pre-treatment, many of these showed significant differences post-treatment suggesting either intrinsic lineagespecific difference in response to standard anti-TB therapy and/or in the underlying host immunity.

Ethics statement
Ethical approval was obtained from The Joint Gambian Government/Medical Research Council (MRC) Ethics Committee in The Gambia and the London School of Hygiene & Tropical Medicine Ethics Committee. All patients provided written informed consent.

Study participants
Sputum smear and culture positive TB patients were recruited at the TB Clinic, MRC Unit, Fajara, The Gambia. On recruitment, we recorded clinical symptoms using a questionnaire that included duration of cough, weight lost, night sweats, and fever; routine clinical assessment including anthropometry (weight, height, skinfold thickness (SFK) and body mass index (BMI)), and tuberculin skin test (TST), as previously reported [16]. Sputum was sent for TB smear and culture. The genotypes of the infecting bacilli in sputum were determined by spoligotyping analysis and assessing the presence or absence of lineage defining Large Sequence Polymorphisms (LSP) RD702 and TbD1 as previously described [15,17]. All patients were HIV-negative with no history of previous TB disease and were enrolled before anti-TB treatment. All patients received conventional therapy of 2 months intensive treatment with Isoniazid, Rifampicin, Pyrazinamide, Ethambutol, followed by a second phase of four months with only Isoniazid and Rifampicin (2HRZE/4HR) [18]. They were actively followed-up at 2 and 6 months of treatment, during which chest x-ray (CXR), haematological and sputum smear examination, and anthropometric measurements were done, and heparinized blood samples collected. All patients were confirmed sputum smear negative at the end of the 6 months treatment.
Data were analysed using GeneMapper 4.0 software package (Applied Biosystems, Warrington, UK) and peak areas were exported to a Microsoft Excel file for downstream analysis. Data were subsequently normalized to GAPDH housekeeping gene and signals below the threshold value for noise cut-off (peak area #200) were assigned threshold value for noise cut-off. A positive control that encompassed the complement reverse sequence of the combined target-specific sequences of the left and right hand half-probes was used for all runs.

Statistical analysis
Demographic and clinical characteristics were compared between Maf-and Mtb-infected patients using Mann-Whitney test for continuous variables and Fisher's exact test for categorical variables. BMI, skinfold thickness and haematology parameters were logarithmically transformed. No transformation was needed for the chest X-ray score [16], and all were analysed using a random intercept model based on restricted maximum likelihood (REML) estimation.
The cytokine responses were positively skewed and contained zero as values. We first added a constant 0.5 to all cytokine responses as suggested by Yamamura [22], then used a base-2 logarithmic transformation to reduce skewness. Three-level random-intercept model (time points nested in cytokines nested in patients) was fitted to account for the dependence of the cytokine responses within subject and between time points. The model included triple interaction terms between cytokines, lineages, stimulants and time points to estimate the difference in infecting lineages effect on cytokine production in blood incubated with medium alone, as well as the incremental difference in infecting lineages effect induced by each stimulant at each treatment time points. This approach did not require any background subtraction. Contrast analysis was used to estimate differences in infecting lineages effect on cytokine production with Sidak multiple comparison correction [23].
Gene expression data were available for each subject at the same time points as for the cytokines. At each time point, the effect of the four culture conditions (Medium, ESAT-6/CFP-10, live Maf and live Mtb) on the expression of 85 selected immune-related genes were assessed. Thirty (30) genes showed expression data above the cut-off value (200) for 1% of the patients and were discarded from the analysis to avoid inflation of the cut-off value. The remaining 55 genes were log2 transformed and analysed as described for cytokines' data.
Predicted values of clinical outcomes, cytokines production and genes expression from the constructed model were used to study kinetics following anti-TB treatment for each group and plotted using R software. We tested for interaction between lineages and treatment time points using the Wald test. All analyses were adjusted for age, gender and ethnicity and performed using STATA 12.1 (StataCorp, USA). Statistical significance was considered at p-value 0.05.

Study participants
Seventy-five HIV-negative TB patients were enrolled in this study, 26 were infected with Maflineage 6 and 49 with Mtb-lineage 4. Before treatment, Maf-and Mtb-infected patients had similar clinical symptoms, age, sex, ethnicity, sputum smear microscopy grade and TST results ( Table 1). The BMI, skinfold thickness and CXR scores were also similar in both groups at enrolment, but following treatment these were more significantly improved in Mtb-compared to Maf-infected patients post-treatment after adjusting for age, sex and ethnic group (p = 0.02, p = 0.04 and p = 0.007 respectively; Table 1). The BMI and CXR scores were significantly affected by the infecting lineages following treatment (interaction p = 0.006 and p = 0.02 respectively; Fig 1). Mean corpuscular volume (MCV) was significantly higher in Mtb-compared to Maf-infected patients before and post-treatment after adjusting for the mentioned potential confounders (p = 0.02 and p = 0.03; respectively), while all other measured haematology parameters were similar between the groups ( Table 1). The monocytes/lymphocytes (M:L) ratio was similar before treatment but higher in Maf-compared to Mtb-infected patients posttreatment after adjusting for confounders (p = 0.05 respectively; Table 1).
Kinetics of specific cytokines production following treatment of Mtb-and Maf-infected patients The kinetics of differentially produced cytokines between the groups was assessed in order to understand the effect of treatment on cytokine responses in each group. In the unstimulated  samples, the patterns of IL-12p70 and IL-15 were similar at 2 months, but differed significantly at 6 months of treatment, whereas those of IL-8, MIP-1α and PDGF-ββ differed already at 2 months of treatment between the groups (interaction p<0.001; Fig 3A). In response to ESAT-6/CFP-10 stimulation, the patterns of IFN-γ and GM-CSF were significantly different between the groups at 2 months of treatment (interaction p<0.01; Fig 3B).  Table).

Kinetics of gene expression in Maf-and Mtb-infected patients following treatment
In unstimulated whole blood, the patterns of TLR9 and IL12A expression were significantly different between Maf and Mtb-infected patients at 2 months of treatment (interaction p = 0.004 and p = 0.0005 respectively; Fig 5A). In ESAT-6/CFP-10 stimulated whole blood, the kinetics of CXCL10 expression differed between the groups at 2 months of treatment (interaction p = 0.04; Fig 5B), while following live Maf stimulation, the kinetics of NLRC4 expression was significantly different at 2 month of treatment between the groups (interaction p = 0.0007; Fig 5C).

Maf-and Mtb-infected patients
The direct relationship among the 23 cytokines and genes differentially expressed between the Maf and Mtb-infected patients was analysed using the Ingenuity Pathway Analysis (IPA, QIA-GEN Redwood City, USA, www.qiagen.com/ingenuity). Ingenuity canonical pathway analysis identified a high enrichment in the communication between innate and adaptive immune cells. Analysis of the upstream regulators revealed a very high enrichment (p = 6.6 x 10 −21 ) of the nuclear factor NF-kappa-B p65 subunit (RELA) that regulates more than half (n = 12) of the differentially expressed cytokines and genes between Maf-and Mtb-infected patients posttreatment. Interestingly, all the cytokines fitted into 2 network functions with the top one containing 13 of the 23 markers and centred on the NFκB complex (Fig 6).

Discussion
We report here a comparison of clinical and immunological responses before and following standard anti-TB treatment between TB patients infected by Maf and Mtb lineages of the MTBC in The Gambia. We used an unbiased rigorous statistical approach that accounted for both intra and inter patient's group variation, experimental condition variability, repeated measurement over time and corrected for multiple comparison to identify biomarkers that are associated with MTBC lineage-specific pathogenesis and response to anti-TB treatment. Overall, before treatment there were no differences in clinical parameters and differences in very few cytokine concentrations and gene expression profiles between Maf-and Mtb-infected patients. However, after treatment the BMI, CXR, skinfold thickness and immunological responses were more significantly improved in Mtb-than in Maf-infected patients. The high similarity in the peripheral blood cells populations, clinical presentation, cytokines production and genes expression profiles before treatment between Maf-and Mtb-infected patients further support uniformity in host responses after succumbing to infection, irrespective of the infecting lineage of MTBC as previously reported [13,15]. This result also suggests that Maf and Mtb lineages differences may not affect diagnosis of active TB based on the measurement of soluble cytokines.
The pattern of the changes after treatment are indicative of a quicker recovery from disease in Mtb-infected patients, which might be due to their stronger host immunity and/or a better immediate response to treatment of the Mtb-lineage 4 [10]. In contrast, the poorer improvement in the clinical outcome following treatment of Maf-infected patients could be related to the immune profile. Higher ratio of monocytes/lymphocytes following treatment has been attributed to unresolved infection with on-going inflammation [24]. The high IL-12p70 production in Maf-patients correlates with the higher expression of IL12A gene, which encodes IL-12p35, a subunit of IL-12p70 produced by macrophages to induce production of inflammatory cytokines from T helper-1 (Th1) and Th17 cells [25,26]. In addition, the increased TLR9 expression also induces IL-12 production [27] that will further promote prolonged inflammatory processes, which could lead to poor weight gain [28,29] as seen in our Maf-infected patients.
IL-15 is a pleiotropic cytokine produced mainly by macrophages, activates a broad range of cells including T and NK cells [30], and promotes the survival of BALB/c mice infected with Mtb [31]. Increased MIP-1α levels in unstimulated plasma of HIV-negative TB patients were lineage groups and used Wald test for assessing the significance at each time point. P-values of the differences are shown.
doi:10.1371/journal.pntd.0004701.g002  Gene expression profiles differ between Mafand Mtb-infected patients before and after treatment. dcRT-MLPA was performed on RNA extracted from whole blood incubated overnight with medium associated with favourable treatment outcomes [32]. Therefore, the increased IL-15 and MIP-1α in Mtb-infected patient after 2 months of treatment could reflect a better response to treatment in contrast to the Maf-infected patients.
The significantly higher CCL5 (RANTES) level and IL13 and BPI expression in Maf-compared to Mtb-infected patients before treatment was not seen post-treatment, implying that treatment did not amplify the differences in the markers already detected pre-treatment but rather revealed new markers. Abundance of IL13 has predicted progression to active tuberculosis disease in high-risk groups [33], suggesting higher susceptibility as well as a lower ability to mount adequate immune response to recover from disease in our Maf-infected patients. This is in line with our previous report of immune exhaustion, and higher prevalence among HIVpositive, severely malnourished and older individuals, of Maf-infected patients [1,15,16,34,35], implying that Maf mainly causes diseases in a permissive host environment.
ESAT-6/CFP-10 stimulation induced the greatest differences in cytokines concentration and genes expressed between the groups post-treatment. These were mainly pro-inflammatory markers that have previously been associated with protection, such as IFN-γ, and were all significantly higher in Mtb-than Maf-infected patients. This suggests that the immune system of Mtb-infected patients were more capable of mounting a robust response to ESAT-6/CFP-10 on recovering following anti-TB treatment as previously described [15,36,37]. However, other studies have reported the opposite [38,39]. Lower cytokine response post-treatment in Mafinfected patients may reflect a poor immune recovery due to an ineffective response to treatment in this group, or a pre-TB immune suppressed state. This is further supported by a generally lower cytokine profile in response to all stimulants in this study, as well as corroborates with the previous finding of increased proportion of HIV-infected amongst patients with Mafcaused tuberculosis [34].
The enrichment of NF-κB activation-related genes among the differentially expressed biomarkers between the groups post-treatment is similar to previous reports [40,41]. NF-κB is a key nuclear transcription factor of pro-inflammatory genes activation [41], which corroborates our identified cytokines. We have previously reported high enrichment of HNF4-α, which regulated about 15% of all genes differentially expressed between Maf-and Mtb-infected patients post-treatment [13]. Low HNF4-α expression has been associated with worst prognosis of hepatocellular carcinoma (HCC) through a robust activation of RELA, whereas higher HNF4α expression inhibited NF-κB expression and improved HCC outcome [42]. The enrichment of HNF4-α and NF-κB in our datasets is very interesting. Although we cannot yet establish a direct connection based on our current data, this might suggests that the dampened Mafinfected patient's pro-inflammatory response to antigen stimulation post-treatment occurs through the inhibition of NF-κB pathway.
Our study was nested within a platform that has a commendable record of treatment monitoring and completion. [43]. Although both Mtb and Maf-infected patients had negative smears for AFB by the end of the follow up, differences may not have been observed due to poor sensitivity of the microscopy method used [44]. The more sensitive culture conversion method is not used to monitor treatment in The Gambia and was therefore not used in our only (A), ESAT-6/CFP-10 (B) and live Maf (C). Median gene expression levels (peak areas normalized to GAPDH and log2 transformed) of the indicated genes are shown in box-and-whisker plots. Equal number of samples were analysed at 0 and 6 months of treatment in each group of Maf-infected (n = 20; white boxes) and Mtb-infected (n = 31; grey boxes) patients, respectively. Log-2 transformed gene expression data were compared between the groups using contrast analysis following a random-intercept model adjusted for age, sex and ethnicity, and Sidak multiple comparison correction. Wald test was used to assess the significance at each time point. P-values of significant differences are shown.
doi:10.1371/journal.pntd.0004701.g004 study. In addition, although the drug resistance profile of the infecting organisms was not tested, The Gambia has a very low prevalence of TB drug resistance determined in the last survey performed in 2003 [45]. It is therefore unlikely that the different immune profiles observed post-treatment between the groups was due to drug resistance. Moreover, previous studies from West Africa showed that Maf is less likely to develop drug resistance compared to Mtb [46][47][48]. However, future studies should include systematic mycobacterial culture to monitor anti-TB treatment response and mycobacterial drug susceptibility testing in order to strengthen this evidence.
There was a striking similarity in the kinetics of cytokines and genes expression between Mtb-and Maf-infected patients during the first two months, but significant differences at 6 months of treatment. These together with the clinical responses provide a coherent and robust evidence of differential responses to treatment in Maf-and Mtb-infected patients. This pattern may indicate that the interval for differences to emerge between the groups by 2 months was short and the difference at 6 month of treatment may reflect a slower rate of clinical and immunological disease resolution of Maf-infected TB patients. The fact that most differences between the two groups only emerge after 2 months of treatment suggest that Maf-infected patient may require longer intensive treatment phase, a longer treatment regimen overall, or reveals the pre-TB disease immune profile of the individuals. These results suggest that Maf leaves a permissive host profiles after treatment with poor lung function and health quality in general, which might promotes the susceptibility to future disease. It becomes important to investigate Maf-infected patients long-term recovery post-treatment and their risk of relapse or re-infection as a permissive host immunity could favour TB disease from Maf, which remains a less virulent bacillus [49] but yet has not been outcompeted by other Mtb lineages according to recent report in West Africa [35,48,[50][51][52][53][54][55][56]. Moreover, Maf causes up to half of all tuberculosis cases in West Africa where the same treatment regimen is given irrespective of the infecting MTBC lineage. New host-directed therapeutic approaches that aim to reduce inflammatory responses associated with immunopathology might be proven useful for these patients [57,58]. Furthermore, these results demonstrate that the identification and evaluation of immunological biomarkers to monitor anti-tuberculosis treatment response in West Africa should consider the diversity of MTBC lineages.
Studies in Maf non-endemic regions have shown heterogeneous response to TB treatment depending on the infecting MTBC lineages [6,10,59]. However, other studies in highly heterogeneous populations found no effect of MTBC lineages, but rather patient ethnicity was a significant determinant for difference in treatment response [9,11]. Clearly, host and bacteria factors are important determinants for treatment response, therefore stratified approaches to TB treatment that account for MTBC lineages is required to improve treatment outcome especially now with clinical trials of shorter drug treatment regimen for TB [14,60].
In conclusion, our data show differences in clinical parameters and immune genes and proteins associated with inflammatory processes recovery between Mtb-and Maf-infected patients following anti-tuberculosis treatment. This profile may be an indication of differences in the resolution of disease or the pre-tuberculosis status of the host immune system. These findings may have public health relevance for therapeutic and biomarker discovery purposes, and warrant further investigation for the use of the identified biomarkers as potential targets for preventive or therapeutic intervention against tuberculosis.
Supporting Information S1 Checklist. STROBE checklist. List of pages and paragraphs containing keys information about this study. The pages and paragraphs numbers responding to the specific questions of the STROBE Checklist are provided to ease the reading of this manuscript and understanding the study. (DOC) S1 Table. Estimated Difference (ED) and Estimated Incremental Difference (EID) of cytokines production between Mtb and Maf-infected patients by stimulants, adjusted for age, gender and ethnicity before anti-TB treatment. Show the Estimated Difference (ED) of cytokines production in blood incubated with Medium only and the Estimated Incremental Difference (EID) of cytokines production above the baseline induced by the respective stimulants between Mtb and Maf-infected patients before treatment. The statistical analyses were done using a random intercept model based on restricted maximum likelihood (REML) and adjusted for age, gender and ethnicity as well as applying Sidak multiple comparison correction. Statistical significant ED and EID are highlighted in bold. (DOCX) S2 Table. Estimated Difference (ED) and Estimated Incremental Difference (EID) of cytokines production between Mtb and Maf-infected patients by stimulants, adjusted for age, gender and ethnicity at 6 months of anti-TB treatment. Show the Estimated Difference (ED) of cytokines production in blood incubated with Medium only and the Estimated Incremental Difference (EID) of cytokines production above the baseline induced by the respective stimulants between Mtb and Maf-infected patients at 6 months of treatment. The statistical analyses were done using a random intercept model based on restricted maximum likelihood (REML) and adjusted for age, gender and ethnicity as well as applying Sidak multiple comparison correction. Statistical significant ED and EID are highlighted in bold. (DOCX) S3 Table. Estimated Difference (ED) and Estimated Incremental Difference (EID) in genes expression between Mtb and Maf-infected patients by stimulants, adjusted for age, gender and ethnicity before anti-TB treatment. Show the Estimated Difference (ED) of genes expression in blood incubated with Medium only and the Estimated Incremental Difference (EID) of genes expression above the baseline induced by the respective stimulants between Mtb and Maf-infected patients before treatment. The statistical analyses were done using a random intercept model based on restricted maximum likelihood (REML) and adjusted for age, gender and ethnicity as well as applying Sidak multiple comparison correction. Statistical significant ED and EID are highlighted in bold. (XLSX) S4 Table. Estimated Difference (ED) and Estimated Incremental Difference (EID) in genes expression between Mtb and Maf-infected patients by stimulants, adjusted for age, gender and ethnicity at 6 months of anti-TB treatment. Show the Estimated Difference (ED) of genes expression in blood incubated with Medium only and the Estimated Incremental Difference (EID) of genes expression above the baseline induced by the respective stimulants between Mtb and Maf-infected patients at 6 months of treatment. The statistical analyses were done using a random intercept model based on restricted maximum likelihood (REML) and adjusted for age, gender and ethnicity as well as applying Sidak multiple comparison correction. Statistical significant ED and EID are highlighted in bold. (XLSX)