Of the 149 SNPs passing quality control as described in materials and methods, allelic and genotypic association analysis identified four SNPs in the TLR8 gene with nominal p-values below 0.05 (Table 1 and Table S1). We also observed two rare variants, within TOLLIP and TLR9 (GeneID:54106), with significant p-values that were not followed up in this study due to their very low allele frequencies. Three of the associated TLR8 variants, rs3764879, rs3788935 and rs3761624 localize in the putative regulatory regions, within five kilobases upstream of the gene (Figure 1). The fourth associated polymorphism was a missense variant, rs3764880 (Met1Val), which would ablate the putative start codon in one of the transcripts encoded by this gene. Given that TLR8 is located on the X chromosome, we performed separate tests for males and females (Table 2). We found a strong allelic association with the minor allele A of the putatively functional polymorphism, rs3764880, with susceptibility to pulmonary TB in males [OR (95% c.i.) = 1.8 (1.2–2.7), P = 0.007]. Very similar and significant association values were found in the three promoter variants, attributable to perfect linkage disequilibrium (r² = 1) between all four polymorphisms (Figure 2).

In order to address the significance of our findings, a permutation analysis of the allelic p-values was carried out (Table 1). One of the polymorphisms passed the permutation test (N = 10,000), with its p-value remaining statistically significant at an adjusted P<0.05. The same analysis was applied separately by gender. In this case, all four SNPs maintained statistical significance at an adjusted P<0.05 in males (Table 2).

Analysis of genotypes for polymorphisms located on Chromosome X was done using a likelihood ratio test. The same four variants on TLR8 were found to be more frequent in cases than controls, indicating susceptibility to disease for carriers of the minor allele. Due to the fact that males carry only one copy of each allele, the genotype association outcome was expected to be the same as for the previous allele association result. Thus, we analyzed genotypes of female subjects (Table S2). The observed number of homozygotes for the associated missense polymorphism, rs3764880 (AA), may have been too low to detect an effect (14 affected vs. 9 controls). Nevertheless there was an apparent trend towards the same outcome observed in the overall sample, with affected females showing an increase of homozygotes for the minor allele, compared to the control group.

Investigation of the gene structure of TLR8 showed two distinct haplotype blocks (Figure 2) in our population. As expected from our initial results, all four associated polymorphisms appeared in the same haplotype block (Block1). Performing separate association analysis of haplotypes in males and females confirmed the genetic association in males (Table S3). The minor haplotype (H2) harboring allele A of rs3764880 showed a pronounced risk effect of disease among male carriers [OR (95% c.i.) = 1.8 (1.2–2.7)]. The population attributable risk of the associated haplotype in males was 4% [15].

In order to confirm these results, genotyping of the four associated polymorphisms was carried out in a follow-up cohort from Russia (1,873 tuberculosis cases, 1,779 controls). Importantly, the minor allele frequency of rs3764880 in the Indonesian population was strikingly different compared to frequencies in the Russian cohort, which was concordant with results obtained for the populations of European and Asian origin from the HapMap project [16]. Around 30% of the Indonesian subjects carried the A allele (Met) associated with risk to TB, whereas this same allele was present in 78% of Russians. Despite these obvious differences in frequencies, the genetic association with pulmonary TB was replicated in the Russian males for allele A of rs3764880 OR (95% c.i.) = 1.2 (1.02–1.48) P = 0.03 (Table 3). Combined evidence for association in males from both Indonesian and Russian populations was P = 1.2×10⁻³–6×10⁻⁴. Analysis of haplotype blocks showed the same trend of association as observed in the Indonesian cohort. In this case, however, the most common haplotype was the one harboring rs3764880A, and displayed a risk effect among male carriers [OR (95% c.i.) = 1.22 (1.01–1.47)] (Table S3).

A subset of 23 patients with active pulmonary TB was selected from the cohort recruited from the outpatient clinic in Jakarta, Indonesia. After informed consent, blood samples were taken from each of the patients during their initial diagnosis with active pulmonary TB disease (at the time of admission), and again following resolution of disease (6 months after completion of a standard anti-tuberculosis multi-drug chemotherapy). Real-time reverse transcription PCR was used to study the mRNA levels of 18 genes (Table S4). We observed that both TLR8 transcripts, TLR8v1 (p = 9.36×10⁻⁵), TLR8v2 (p = 5.29×10⁻⁵) and MYD88 (NM_002468.3) (p = 4.09×10⁻⁵) were the most significantly upregulated in TB patients during active disease, relative to their convalescence. Both variants of TLR8 showed a greater than two-fold increase in expression, whereas MYD88 increased by 1.9-fold (Table 4). Transcript levels of TLR7 (NM_016562.3) and CD14 (NM_001040021.1) were also significantly increased (p = 2.5×10⁻³; 3×10⁻² respectively), whereas TIRAP (NM_148910.2) showed a downregulation of mRNA expression, fold change = 0.2, but without statistical significance (Table 4).

Expression of TLR8 over time was assessed in human THP1 macrophages after infection by M. bovis BCG (Figure 3). Uninfected control cells showed no change of TLR8 protein levels over time (Figure 3A, B). However, infected macrophages showed a marked raise of TLR8 expression at 20 hours post-infection (Figure 3D). It is noteworthy that even cells that didn't take up the whole bacteria displayed an important increase on TLR8 (Figure 3D).

Genomic DNA was extracted from whole blood following a protocol described elsewhere [34]. After genotyping, 74 samples were excluded from the Indonesian cohort because of sample duplication and/or familial relationships not originally reported, but identified by RelPair [35].

RNA was successfully extracted using an RNeasy Mini Kit (Qiagen, Germany) from peripheral blood mononuclear cells (PBMCs) of a subset of 23 patients.

Consent forms approved by local Institutional Review Boards of the Medical Faculty of University of Indonesia and the Eijkman Institute for Molecular Biology in Jakarta were signed upon recruitment by all participants. Written informed consents from all Russian subjects as well as permission from ethics committees were obtained [32].

Selection of SNPs was carried out using an in-house database, GISSNP, which integrates data from public databases (Ensembl, Celera, dbSNP build 123/126) as well as proprietary data. Polymorphisms with the following characteristics were preferentially chosen: putative functional variants resulting in changes in the protein sequence, minor allele frequencies over 5%, average spacing of one SNP every 1 to 2 kilobases. To screen for possible regulatory elements, flanking regions five kilobases upstream and downstream of the gene were also covered.

Design of a custom Oligo Pool Assay (Illumina) was implemented following the manufacturer's specifications. Genotyping was performed with a BeadStation 500G Genotyping System (Illumina). Genotypes were analyzed with Beadstudio software also from Illumina. Ten SNPs were genotyped with a Sequenom primer extension-based protocol described elsewhere [36],[37]. The genotype concordance among the two systems used for genotyping in this study has been reported to be over 99.5% [37]. Assessment of genotypes was done by laboratory personnel without any prior knowledge of the diagnosis of the subjects.

Genotyping of 247 SNPs from 18 candidate genes was performed in the Indonesian cohort. We found that 75 SNPS (30%) of the genotyped polymorphisms were monomorphic in this population (Table S4). Variants with call-rates below 90% (N = 17) were not considered. Six SNPs showed deviations from Hardy-Weinberg equilibrium (HWE) in the control group, and were removed from further analysis (Table S4). Association analyses were applied to the remaining 149 polymorphic SNPs with reliable genotypes (Table S1).

Hardy-Weinberg equilibrium was calculated in the control group using HelixTree v4.4.1 (GoldenHelix Inc., Bozeman, MT, United States) and Exemplar (Sapio Sciences, LLC, York, PA, United States). Similarly, allelic association analysis was carried out in both software packages. Allelic p-values were calculated by means of a 2×2 chi-square table. A two-sided Fisher Exact test, when counts in any cell fell below five, as well as odds ratios were calculated with Exemplar. Allelic analysis of SNPs located on Chromosome X was performed with Haploview v3.31 [38]. A likelihood ratio test was applied to calculate genotypic associations of SNPs on Chromosome X. Combined p-values were calculated by Fisher's combined probability test which allows pooled information across several tests that share the same null hypothesis [39].

The statistical significance of nominal allelic p-values was assessed by permutation analysis (N = 10,000) with Haploview v3.31.

Haplotype blocks and linkage disequilibrium plots were constructed with Haploview v3.31 using the default algorithm proposed by Gabriel et al [40].

A subset of 23 patients with active pulmonary TB was selected from a cohort recruited from an outpatient clinic in Jakarta, Indonesia. Blood samples were taken from each of the patients at 2 time points: the active phase of pulmonary TB disease (at the time of admission), and the convalescent phase (6 months after admission and standard anti-tuberculosis multi-drug chemotherapy). Real-time reverse transcription PCR was used to study expression levels of 18 genes (Table S4) on peripheral blood mononuclear cells (PBMCs). Briefly, an aliquot of 10 µl RNA was reverse transcribed to cDNA using the high capacity cDNA kit (Applied biosystems Asia Pte Ltd). The obtained cDNA was diluted 1/5 with water and 10 µl was used for amplification.

The gfp cDNA cloned into mycobacterial-E.coli shuttle plasmid pMV206 was a gift from Dr. Alain Baulard (Pasteur Institute, France). The plasmid was incorporated into competent BCG cells by electroporation. GFP BCG was observed by the FACS Calibur flow cytometer (Becton Dickinson).

The monocytic cell line THP-1(ATCC, Rockville, MD) cells were cultured in RPMI 1640 supplemented with 10%FBS, penicillin (100 U/ml) and streptomycin (100 ug/ml) (Invitrogen). Cells were plated at a density of 2×10⁵/ml in 8-well chamber coverglass (LAB-TEK). Monocytes were allowed to adhere and differentiate into macrophages for 48 hours with 5 nM PMA (Sigma Aldrich) at 37°C in a humidified atmosphere of 5% CO₂. Differentiated macrophages were infected with GFP BCG at an MOI of 20∶1 and incubated at 37°C, 5% CO₂. Infected cells at 1 hr post infection were washed twice with RPMI without antibiotics to remove uningested and unadhered bacteria. Cells were then harvested for TLR8 expression as mentioned below. Infected cells at 20 hours post infection were washed to remove uningested and unadhered bacteria at 4 hours post infection to minimize cell death and re-incubated for a further 16 hours. Cells were then fixed with 4% paraformaldehyde (Sigma Aldrich) for 15 minutes at room temperature and cell membrane was disrupted with 1% saponin (Sigma Aldrich) for 20 mins at room temperature. Mouse anti-human TLR8 phycoerytrhin (PE) antibody (Imgenex) was used at 3 ug/ml for 1 hour at room temperature and excess antibody was washed twice with PBS (Sigma Aldrich). To prevent bleaching of the dyes during confocal viewing, anti fade prolong gold with DAPI (Invitrogen) was added to slides so that it formed a protective layer over the cells. Slides were stored at 4°C in the dark until confocal viewing.

The LSM 510 scanhead of the confocal laser-scanning microscope system (Zeiss 5 duo, Germany) was used to detect intracellular fluochrome. Cells were scanned by triple excitation for PE (red), GFP (green) and DAPI (blue) fluorescence. A 63× oil objective with numerical aperture of 1.4 was used and images were captured.