Toll-like receptor chaperone HSP90B1 and the immune response to Mycobacteria

Rationale HSP90B1, also known as gp96, is a chaperone for multiple Toll-like receptors (TLRs) and is necessary for TLR-mediated inflammatory responses in murine myeloid cells. The molecule is also expressed in T-cells though its specific role is unknown. We hypothesized that human HSP90B1 regulates monocyte and T-cell responses to Mycobacterium tuberculosis (Mtb) and bacilli Calmette-Guerin (BCG) and that its variants are associated with susceptibility to TB disease. Methods We screened 17 haplotype-tagging SNPs in the HSP90B1 gene region for association with BCG-induced T-cell cytokine responses using both an ex-vivo whole blood assay (N = 295) and an intracellular cytokine staining assay (N = 180) on samples collected 10 weeks after birth. Using a case-control study design, we evaluated the same SNPs for association with TB disease in a South African pediatric cohort (N = 217 cases, 604 controls). A subset of these SNPs was evaluated for association with HSP90B1 expression in human monocytes, monocyte-derived dendritic cells, and T-cells using RT-PCR. Lastly, we used CRISPR/Cas9 gene editing to knock down HSP90B1 expression in a human monocyte cell line (U937). Knockdown and control cell lines were tested for TLR surface expression and control of Mtb replication. Results We identified three SNPs, rs10507172, rs10507173 and rs1920413, that were associated with BCG-induced IL-2 secretion (p = 0.017 for rs10507172 and p = 0.03 for rs10507173 and rs1920413, Mann-Whitney, dominant model). SNPs rs10507172 and rs10507173 were associated with TB disease in an unadjusted analysis (p = 0.036 and 0.025, respectively, dominant model) that strengthened with sensitivity analysis of the definite TB cases, which included only those patients with microbiologically confirmed Mtb (p = 0.007 and 0.012, respectively). Knockdowns of HSP90B1 in monocyte cell lines with CRISPR did not alter TLR2 surface expression nor influence Mtb replication relative to controls. Conclusion Among infants, an HSP90B1 gene-region variant is associated with BCG-induced IL-2 production and may be associated with protection from TB disease. HSP90B1 knockdown in human monocyte-like cell lines did not influence TLR2 surface localization nor Mtb replication. Together, these data suggest that HSP90B1 regulates T-cell, but not monocyte, responses to mycobacteria in humans.


Introduction
Lat interactome, which is responsible for the amplification and diversification of T-cell receptor signaling [22], though the role for the chaperone in this complex has not been determined. Conditional knockout of Hsp90b1 in Treg cells resulted in an uncontrolled autoimmune inflammatory response in mice and a loss of the Treg phenotype [23]. Used as a vaccine adjuvant to a BCG vaccine, HSP90B1 was recently shown to enhance T-cell responses and protection from M. bovis in mice [24]. We do not know if HSP90B1 influences T-cell responses indirectly as a chaperone for innate immune molecules or directly as a chaperone of T-cell signaling molecules.
In this paper, we examine the role of HSP90B1 genetic variants on BCG-induced T-cell cytokine responses and the development of TB disease in a South African pediatric cohort. We hypothesized that genetic variants would influence both monocyte and T-cell mediated immune responses to Mycobacteria and protection from TB disease.

Research/Study participant recruitment
Study participants were enrolled by the South African Tuberculosis Vaccine Initiative (SATVI) at field sites in Worcester, South Africa [11,[25][26][27][28]. This region has one of the highest incidences of pediatric TB disease in the world [29]. The cohort used in the current study is part of a larger BCG vaccination correlates-of-risk project with 11,680 infants and has been described in detail previously [25,28,29]. Enrolled infants were vaccinated with BCG at birth, as is standard practice in South Africa. A nested genetics case-control study was performed to identify cases and controls during a 2-year prospective observation period as previously described [30]. The cases and controls included those from Cape Mixed Ancestry (CMA) and Black African descent. The CMA ethnicity represents the genetic admixture of Khoisan, Black African, European and both east and south Asian populations that has existed for over 350 years [31,32]. Several terms have been used to describe CMA including South African Mixed Ancestry and "Coloured", a collective term for people of mixed ancestry in southern Africa, which is an officially recognized census term in South Africa and routinely used for self-classification.
The criteria for the case and controls definitions have been described previously [28,29] and are summarized below. Community-wide passive surveillance systems identified patients with TB disease and children with symptoms suggestive of TB disease. Briefly, all children with symptoms consistent with TB disease or who had contact with an adult with TB disease were admitted to a dedicated research ward for examination, chest imaging, tuberculin skin testing, two early-morning gastric aspirates, and two induced sputa for Mtb smear and culture. Subjects were described as "definite TB" if they had a positive Mtb culture, smear, or PCR from one of their samples. Subjects were described as "probable TB" if they had a chest radiograph consistent with or suggestive of TB in addition to one or more laboratory or clinical features (smear negative, cough >2 weeks, PPD skin test � 15mm, failure to thrive, and recent weight loss). Subjects diagnosed with TB by the treating physician and with 2 or more clinical features suggestive of TB but without consistent chest radiography were described as "possible TB". Subjects with a symptom consistent with TB but with no other supporting evidence were labeled "unlikely TB". All others were described as "not TB". Two types of controls were enrolled. Household controls were children without TB disease over a 2 year follow up period who were living with an individual with active TB disease. Community controls had no history of TB disease in first two years of life. Demographic information for the cohort is described in the supporting information (S1 Table).
Exclusion criteria included HIV positive infant or mother, BCG vaccine not administered within 24 hours of birth, significant perinatal complications in the infant, any pre-existing acute or chronic disease in the infant, or clinically apparent anemia. Parents or legal guardians of study participants were informed of the risks and benefits of study participation and signed informed consent prior to enrollment. The protocol was approved by the University of Cape Town Research Ethics Committee, the Aeras Global TB Vaccine Foundation, and the University of Washington Institutional Review Board. Two audits were done during the enrollment period, on behalf of the sponsor, by an independent contract research organization, Triclinium, Johannesburg, South Africa [29].

SNP selection
We identified haplotype-tagging SNPs from the International HapMap Project (http://www. hapmap.org) and other public databases within the Genome Variation Server (http://www. ncbi.nlm.nih.gov.offcampus.lib.washington.edu/SNP/ and www.innateimmunity.net). We searched a region on chromosome twelve ten kilobases upstream and downstream of the HSP90B1 gene as well as within the gene itself for tagged SNPs in the Yoruba in Ibadan population (YRI) using an R 2 cutoff of 0.8 for linkage disequilibrium and a minor allele frequency cut-off of 5%. We used STATA version 14 and the software program PWLD (StataCorp, College Station, TX) to calculate R 2 and D' as measurements of linkage disequilibrium between polymorphisms. Using the above criteria, we identified 19 SNPs for analysis (Fig 1). Hardy-Weinberg equilibrium was tested for each SNP in the CMA control population. Two of these SNPs, rs17034939 and rs7313381, were in violation of Hardy-Weinberg equilibrium (x 2 p<0.001) and were not analyzed further. Chromosome location and genotypic and allelic frequencies for each SNP in the case and control populations are described in the supporting information (S2 Table) as are linkage disequilibrium R 2 values from the CMA control group (S3 Table). For reference, genotype and allele frequencies for the YRI population are described the supporting information (S4 Table) as are linkage disequilibrium R 2 values for select SNPs in the YRI population (S5 Table). For the case-control study we also included a set of 96 Ancestry Informative Markers (AIMS) for the CMA population. Further details of this analysis are provided in the Statistical Analysis section of the Materials and Methods section and in the supporting information (S6 Table).

Genotyping
Genomic DNA was prepared from peripheral blood using the QIAmp DNA Blood or Blood and Tissue Kit (Qiagen). Genotyping was performed using an Illumina GoldenGate genotyping platform (Illumina, Inc) or a Fluidigm array platform (Fluidigm Corporation) and a subset of calls for one SNP (rs10507173) were genotyped using the TaqMan SNP Genotyping Assay (Applied Biosystems, Step One Plus instrument) to confirm genotyping accuracy.

Cytokine assays
Whole blood cytokine levels and intracellular cytokine staining were conducted on heparinized blood drawn from 10-week old South African infants. 10 milliliters of blood were taken and used both for the cytokine assays and genotyping. Samples were stimulated ex-vivo with media or BCG (strain SSI, 1.2 x 10 6 CFU's/ml) for 7 hours at 37˚C. For whole-blood assays, IL-2 and IFNγ were measured by multiplex bead array technology according to manufacturer's instructions (Bio-Rad, Hercules, CA, USA) and read on a Luminex luminometer (Luminex, Austin, TX, USA). Basal cytokine levels measured in plasma harvested from unstimulated blood were subtracted from values obtained from BCG-stimulated blood [11].

HSP90B1 gene expression in leukocytes
Peripheral blood mononuclear cells (PBMC's) were isolated from whole blood using Ficoll gradient separation. Monocytes, dendritic cells (DC's), and T-cells were isolated as described below for each cell type. Monocyte samples were derived from PBMC's from South African donors. DC's and T-cells were derived from healthy donors local to Seattle, WA. Sixty milliliters of blood were taken from the local, healthy cohort for genotyping and cell isolations. Following cell isolation, RNA was collected using the RNeasy kit (Qiagen). Single-stranded cDNA was generated from RNA samples using Multiscribe reverse transcriptase (Invitrogen). Real-time PCR was performed on an Applied Biosystems One Step Plus Real Time PCR System (ABI) using HSP90B1 probe Hs.PT.53a.4251075 (Integrated DNA Technologies) and GAPDH-JOE (ABI 402869).
Monocytes. PBMCs were thawed and plated to tissue culture plates overnight in the presence of 50ng/ml M-CSF with RPMI supplemented with 10% FCS. RNA was isolated from plate-adherent monocytes using RNeasy kits (QIAGEN, Inc). Nonadherent cells were used to collect genomic DNA using peripheral blood DNA collection kits (QIAGEN). RNA concentrations were determined using Nanodrop; cDNA was amplified for long-term storage at -20C. qPCR studies were performed on an ABI Step One Plus (ABI, Inc). For gene expression assays following LPS stimulation, monocytes were exposed to media for four hours. Cells were subsequently harvested for RNA collection and RT-PCR as described above.
Dendritic cells. Dendritic cells were generated from monocytes and confirmed by flow cytometry as previously described [34]. Briefly, monocytes were isolated from PBMCs by positive selection using CD14 microbeads (Miltenyi Biotec) and magnetic column separation. Monocytes were then incubated in RPMI + 10% FCS (Life Technologies) supplemented with L-glutamine and GM-CSF (100ng/ml) plus IL-4 (20ng/ml) (PeproTech) for 3 days.
T-cells. T-cells were isolated from PBMCs by positive selection using CD3 microbeads (Miltenyi Biotec) and magnetic column cell separation. T-cell survival and specificity was confirmed by flow cytometry with live/dead stain aqua (Life Technologies) and CD3-ECD (Beckman-Coulter).

TLR2 surface expression
Wild type and HSP90B1 knockdown U937 cell lines were incubated with either mouse anti human TLR2 conjugated to APC (eBioscience 17-9922) or isotype control. Flow cytometry was performed at the University of Washington Flow Cytometry Core Facility on an LSRII 5-laser flow cytometer (Becton Dickenson, Inc).

Mtb replication assay
U937 cell lines containing the HSP90B1 knockdown or the empty vector or wild type controls were plated in sextuplet at a density of 75,000 cells/well and differentiated with PMA at 50ng/ ml overnight, washed x 2 with HBSS, and rested for 24 hours. A frozen aliquot of Mtb Erdman strain expressing lux (Mtb-lux, a gift from Jeffrey Cox, University of California, San Francisco) was thawed, centrifuged, and resuspended in RPMI with 10% FBS. Luminescence (relative light units, RLU's) are a proxy for colony count using this assay [30]. To act as a control for background luminescence, a separate set of wells had Mtb-Lux added without human cells. Relative light units were measured daily for 1 week following infection on a plate reader (Synergy H4 reader, BioTek Instruments). Means of daily luminescence values, standard deviations, and two-tailed Student's t tests were performed and graphed using GraphPad Prism Version 7.0.

Statistical analysis
We used STATA 14 (StataCorp) to perform Pearson's X 2 testing and logistic regression on the case-control cohort and general linearized modeling or analysis of variance for cytokine assays and gene-expression assays, as specified in the text for each experiment. We used Prism 7 (GraphPad Software, Inc) for analysis of normality and nonparametric testing for cytokine assays. Results were not adjusted for multiple comparisons. For the case-control study, to control for potential genetic heterogeneity in the CMA ethnicity, we included a set of 96 Ancestry Informative Markers (AIMS) for the complex five-way admixed South African Coloured population as described previously [28,30,37]; see S6 Table for genotype and allele frequencies of AIMs SNPs in the case and control populations. In CMA individuals, there were no significant differences in genotype frequencies of the AIMS between cases and controls. The AIMS were used to calculate a coefficient incorporating the first five principal components of the AIMS data, which accounted for over 60% of the variation in the dataset. We then used this data to create a regression coefficient for adjusting the primary case-control data for ethnicity by converting it into an ethnicity principal component coefficient using the "pca" command in STATA 14. This provided an alternative means of regressing for ethnicity within the CMA populations.

Association of HSP90B1 gene region SNPs with BCG-induced cytokine responses in South Africa
To examine the role of HSP90B1 in TB pathogenesis, we assessed whether genetic variants are associated with innate and adaptive immune responses as well as clinical disease. Following BCG vaccination at birth in South Africa, infant blood was drawn at 10 weeks of age and restimulated with BCG. We measured secretion of IL-2 and IFN-γ [29]. To assess T-cell responses, we first examined whether 17 haplotype-tagging HSP90B1 gene region SNPs were associated with whole blood secreted cytokine responses to ex-vivo BCG stimulation (N = 295) (Figs 1 and 2, S7 Table).
SNPs rs10507172 and rs10507173 were associated with IL-2 production (p = 0.002) while two other SNPs (rs1305392 and rs1920413) showed a trend towards association with IL-2 production (p = 0.055 and 0.080, respectively) (general linearized model, Fig 2A, 2C, 2E and 2G). No other SNPs were associated with IL-2 secretion nor with secretion of IFN-γ (S7 Table). Due to low numbers of minor-allele homozygotes for rs10507172 and 105017173, we examined a dominant model which showed that rs10507172, rs10507173 and rs1920413 were associated with IL-2 production (p = 0.017, 0.03, and 0.03, respectively, Mann-Whitney). In each case the minor allele was associated with an increase in IL-2 production. rs10507172 and rs10507173 were in moderate linkage disequilibrium (R 2 = 0.59, D' = 0.97, Fig 1). Together, these data suggest that HSP90B1 genetic variants may be associated with BCG-specific T-cell responses. To determine a cellular origin of HSP90B1-associated T cell responses, we performed intracellular cytokine staining with flow cytometry to evaluate for BCG-specific IFN-γ or IL-2 CD4 + or CD8+ T cell memory responses in an independent set of samples (N = 180). Of the four SNPs with significant or borderline significant IL-2 associations in the whole blood assay, only rs10507173 was associated with CD4+ IL-2 production (p<0.0001, general linearized model), though the association was driven by a single minor allele homozygote outlier (2H) and was not significant under a dominant model (p = 0.95). None of the other SNPs were associated with IL-2 or with IFN-γ in CD4+ cells, nor with the same cytokines in CD8+ cells (S8 Table). Together, these data are inconclusive but suggest that rs10507173 may be associated with BCG-induced IL-2 produced by CD4+ T cells.

HSP90B1 gene region SNPs are not associated with gene expression levels in monocytes, dendritic cells, or T-cells
SNPs rs10507172, rs10507173, rs1920413, and rs1305392 fall in noncoding sequence either within intronic regions (rs1920413, rs1305392) or in the 3' downstream region of HSP90B1 (rs10507172, rs10507173) (Fig 1). To explore the mechanisms by which these SNPs may influence IL-2 cytokine production and pediatric TB disease, we evaluated whether any polymorphisms were expression quantitative trait loci (eQTLs).
We examined whether HSP90B1 SNPs were associated with HSP90B1 gene expression in monocytes from the same South African cohort (Fig 3A, 3B and 3C). The rs10507173 rare variant was associated with increased expression (p = 0.007, ANOVA). However, a single outlier homozygous for the rare variant contributed to the association and, in the absence of the outlier, an association was not seen. We did not find an association between rs10507172, rs1920419 or rs1305392 genotype and gene expression in monocytes (graphs for rs1305392 shown in S1 Fig). We did not identify an association between genotype in rs10507172, rs10507173, rs1920419, or rs13105392 and HSP90B1 expression in either monocyte-derived DCs or T-cells. (Fig 3D-3I). Together, these data suggest that HSP90B1 variants are not eQTLs in monocytes, dendritic cells, or T cells.

HSP90B1 knockdown did not influence TLR2 surface expression or Mtb replication in a monocyte cell line
We next examined whether HSP90B1 mediated antimicrobial effects in Mtb-infected macrophages. Using CRISPR/Cas-9 technology we targeted early exonic regions of the HSP90B1 gene for knockout in the monocyte-like U937 cell line (Fig 4A). Western blot confirmed at least partial knockdown of HSP90B1, although an HSP90B1 band at~100 kDa remained after CRISPR/Cas9 targeting and suggested possible partial expression of full length HSP90B1. Since previous studies have shown that Hsp90b1 is necessary for surface TLR expression and TLR-mediated signaling in mice, we tested the knockdown monocyte line for surface expression of TLR2 but did not observe a difference between KO and control cell lines (Fig 4B). We also did not see a difference in Mtb intracellular replication between the HSP90B1 knockdown and the control lines ( Fig 4C). Together, these data suggest that HSP90B1 does not regulate TLR2 surface expression or bacterial growth in monocyte cell lines infected with Mtb.

Association of HSP90B1 gene region SNPs with TB disease in South African children
To test the clinical significance of the association of HSP90B1 variants with BCG-induced Tcell responses, we used a case-control study design in a South African pediatric cohort to determine if HSP90B1 gene-region SNPs are associated with pediatric TB disease. We genotyped HSP90B1 SNPs in 217 cases of children with TB disease and 604 controls ( Table 1, S1 and S2 Tables). SNPs rs10507172, rs10507173, and rs1920413 showed significant or near-significant associations with TB disease using a dominant model in an unadjusted analysis (i.e. comparing AA vs Aa/aa; p = 0.036, p = 0.025, and p = 0.069, respectively). When adjusting for gender, all three SNPs again showed significance or near-significance (OR = 0.68, p = 0.035 for rs10507172, OR = 0.62, p = 0.023 for rs10507173 and OR = 0.74, p = 0.063 for rs1920413). All three SNPs have an odds ratio less than one, suggesting that the minor allele is protective from the development of pediatric TB disease. SNP rs1305392, which was also borderline associated with IL-2 production, was not significantly associated with TB disease. We analyzed for strength of association in each ethnic population alone (S9 Table) or adjusted for population admixture with ancestry informative markers and found the same directions of effect, though less statistical significance (p = 0.29, OR = 0.78 for rs10507172, p = 0.134, OR = 0.65 for rs10507173 and p = 0.177, OR = 0.75 for rs1920413 under dominant models adjusting for AIMs; S10 Table). Finally, we performed a sensitivity analysis by examining the association based on the strength of the clinical case classifications (N = 68 Definite TB, 111 Probable TB and 38 Possible TB with genotyping data for these SNPs). When only Definite and Probable cases were included, the association was stronger for all three SNPs (p = 0.007, OR = 0.58 for rs10507172, p = 0.012, OR 0.55 for rs10507173 and p = 0.051, OR 0.71 for rs1920413; S11 Table). Together, these data suggest a possible association of three HSP90B1 SNPs with a decreased risk of pediatric TB.

Discussion
Our primary findings suggest that three HSP90B1 gene region SNPs are associated with BCGspecific IL-2 in T-cells in a South African pediatric cohort and that another SNP had a trend toward association. Two of these SNPs were associated with TB disease in an unadjusted analysis and the association strengthened using the most strict diagnostic criteria. We found no evidence that HSP90B1 regulated monocyte responses to Mtb.
The finding of increased IL-2 and protection from TB disease is consistent with a model in which an enhanced Th1 response is protective [38], though we might also expect an increase in IFN-γ, which we did not see. IL-2 expression by T cells is a feature of central memory Tcells while IFN-γ expression, especially in the absence of IL-2, is a feature of effector or effector memory T-cells [39].
Our study has several limitations and strengths. First, the BCG-specific T-cells associations described above in the discovery sample set, when corrected for multiple comparisons, do not meet a significance threshold of p < 0.05. Using a conservative Bonferroni approach and considering that 17 SNPs were included in the analysis, we would require a threshold of p = 0.0029 (0.05/17) to achieve significance. Second, the TB case-control data became more highly significant when using strict criteria for the diagnosis of TB disease. This observation may be due to misclassification of the less stringent cases of 'probable TB' that were based on symptoms and chest x-ray findings. Third, pediatric TB is paucibacillary and difficult to diagnose. This challenge requires classifying case likelihood by different criteria. Reassuringly, the sensitivity analysis suggested a similar magnitude of effect in the definite TB cases. Despite these challenges, this pediatric cohort is one of the largest available in the world for genetic studies and presents opportunities for genetic insights that may be fundamentally different than adult TB.
Our negative results with the CRISPR knockdown in monocytes are surprising, particularly given the dramatic loss of TLR-mediated signaling seen in mouse conditional knockouts. One limitation to our innate immune molecular findings includes the possibility of an incomplete knockout of the molecule. While western blotting showed clear evidence of a reduction in expressed protein, a band roughly corresponding to the expected molecular weight of HSP90B1 was present in the knockout. In a mouse model showing loss of TLR signaling in the absence of the protein, Hsp90B1 was completely knocked out using CRE/Flox in conditional macrophage lines [20]. It is possible that even a small amount of functioning molecule, for example the approximate 1/10 band density seen in our U937 knockdown line, is sufficient to chaperone TLRs and thus produce no appreciable drop in surface expression. The association of HSP90B1 SNPs with BCG-specific T-cell responses suggests a possible direct regulatory role on T-cell function. HSP90B1 is expressed in T-cells and may play an as-yet undetermined role in the downstream signal amplification of the T-cell receptor [22]. Interestingly, Foxp3-selective conditional knockout of Hsp90B1 in murine T-cells led to a loss of Treg phenotype and an uncontrolled inflammatory response which included increased IL-2 production [23]. These data provide a possible Treg-dependent model for HSP90B1 regulation of BCG-specific T-cell responses. This is supported by evidence that GARP, a molecule that may also be chaperoned by HSP90B1, is a marker of activated FOXP3-expressing Tregs in humans [40].
Both historic and recent evidence describes a potential role for HSP90B1 as a vaccine adjuvant [24,38,41]. These studies have focused on the use of the molecule provided via an exogenous route and also in combination with a vaccine such as BCG. In this paper we studied the endogenous role of the molecule, specifically with regards to its function as a chaperone for TLRs. Whether there is a link between the endogenous function of HSP90B1 and its activity as an adjuvant remains to be seen. One possibility is that client-bound HSP90B1 is exposed to other immune cells following cell lysis. This mechanism might also explain the association that we did see with a gene-region SNP and a T-cell cytokine response. More work is needed to elucidate the complete role of HSP90B1 in the human immune response to infection.  Table. Genotype and allele frequencies from the Yoruba in Ibadan (YRI) population for SNPs that were included in our study. Data were generated from the 1000 Genomes Project phase 3 using Ensembl (www.ensembl.org). Note that for SNP rs11111854 and SNP rs2700509, the minor allele from the YRI population was the major allele in our study. (DOCX) S5 Table. Linkage disequilibrium R 2 values for relevant SNPs in the Yoruba in Ibadan (YRI) population. LD plot was generated using Ensembl (www.ensembl.org) for the YRI population from the 1000 Genomes Project phase 3. Minor allele frequencies are shown at the convergence of the vertical and horizontal cells for each individual SNP (unshaded cells). Note that for two SNPs (rs11111854 and rs2700509) the alternate allele was minor in our control population (S3 Table) relative to the YRI population. (DOCX) S6 Table. Ancestry informative markers (AIMs) for cases and controls. Genotype and allele frequencies for ancestry informative markers (AIMs) are listed. Results of x 2 testing for Hardy-Weinberg equilibrium are shown (HWE-p). No SNPs were in violation of HWE. Results of x 2 testing using a genotypic model between cases and controls are also shown (Gen p). No significant differences between cases and controls were observed in any of the included SNPs.  Table. Genetic association data stratified by ethnicity. Genotypic frequencies and results of the Pearson's x 2 test for the dominant model are shown for selected SNPs based on selfdescribed ethnicity. For individuals who self-identified as Black and for whom more detailed ethnicity data were available,~93% identified as Xhosa and~6% as Sotho. CMA = Cape Mixed African. OR = odds ratio. 95% CI = 95% confidence interval. (DOCX) S10 Table. Genetic association data for Cape Mixed Ancestry adjusted for AIMs. Cape Mixed Ancestry ethnicity genotype frequencies for select SNPs for individuals within for whom ancestry informative marker (AIMs) data were available. AIMs adjusted statistics are provided for cases versus controls under a dominant model. OR = odds ratio. 95% CI = 95% confidence interval. (DOCX) S11 Table. Sensitivity analysis of TB genetic association. Data from sensitivity analysis includes only individuals with a more stringent case or control definition. For the sensitivity analysis, a TB case was defined as individuals with 'Definite' and 'Probable' TB diagnoses, with individuals carrying a 'Possible TB' diagnosis excluded. Controls were defined as individuals with 'Not TB'; individuals carrying the 'Unlikely TB' diagnosis were excluded. Genotypic frequencies and results of the Pearson's x 2 test for the dominant model are shown for selected SNPs. (DOCX) S1 Fig. rs1305392 is not associated with HSP90B1 gene expression. Gene expression was compared to expression of GAPDH in unstimulated monocytes (A), dendritic cells (B), and Tcells (C) and stratified by genotype. Median and interquartile ranges are included and sample sizes are provided in parentheses below each genotype. Results were not significant for any cell type (p>0.05, ANOVA). (DOCX)