Identification of Tuberculosis Susceptibility Genes with Human Macrophage Gene Expression Profiles

Although host genetics influences susceptibility to tuberculosis (TB), few genes determining disease outcome have been identified. We hypothesized that macrophages from individuals with different clinical manifestations of Mycobacterium tuberculosis (Mtb) infection would have distinct gene expression profiles and that polymorphisms in these genes may also be associated with susceptibility to TB. We measured gene expression levels of >38,500 genes from ex vivo Mtb-stimulated macrophages in 12 subjects with 3 clinical phenotypes: latent, pulmonary, and meningeal TB (n = 4 per group). After identifying differentially expressed genes, we confirmed these results in 34 additional subjects by real-time PCR. We also used a case-control study design to examine whether polymorphisms in differentially regulated genes were associated with susceptibility to these different clinical forms of TB. We compared gene expression profiles in Mtb-stimulated and unstimulated macrophages and identified 1,608 and 199 genes that were differentially expressed by >2- and >5-fold, respectively. In an independent sample set of 34 individuals and a subset of highly regulated genes, 90% of the microarray results were confirmed by RT-PCR, including expression levels of CCL1, which distinguished the 3 clinical groups. Furthermore, 6 single nucleotide polymorphisms (SNPs) in CCL1 were found to be associated with TB in a case-control genetic association study with 273 TB cases and 188 controls. To our knowledge, this is the first identification of CCL1 as a gene involved in host susceptibility to TB and the first study to combine microarray and DNA polymorphism studies to identify genes associated with TB susceptibility. These results suggest that genome-wide studies can provide an unbiased method to identify critical macrophage response genes that are associated with different clinical outcomes and that variation in innate immune response genes regulate susceptibility to TB.


Introduction
TB, a leading cause of death worldwide, is characterized by different clinical forms including latent TB (LTB), localized pulmonary infection, and various forms of extrapulmonary TB including TBM. 90% of people infected with Mtb have latent infection with no symptoms and an immune response that contains the bacilli. In 10% of infected individuals, symptoms develop and most commonly manifest as pulmonary disease, which accounts for 80% of all forms of TB disease [1]. TBM develops in around 1% of all cases of active TB [1] and is the most severe form with mortality rates of 20-25% and high rates of neurological sequelae in many of those who survive [2,3]. Although only 10% of individuals who are infected with Mtb develop active disease, it is not known which immune responses are associated with susceptibility or resistance. In addition, it is not known why some individuals have disseminated TB that spreads to the meninges and central nervous system, while most people have localized disease in the lungs. Although environmental exposures, pathogen virulence traits, and host genetics have the potential to influence the different clinical manifestations of TB, it is not currently understood which factors are the most important [4].
Several lines of evidence, including twin and genome-wide linkage studies, suggest that host genetics strongly influences susceptibility to TB [5][6][7][8][9]. Candidate gene association studies have implicated common polymorphisms in genes that may influence the development of TB [10,11]. Although there is potential for candidate gene study designs to be successful when sample sizes are sufficient and case and control groups are accurately defined, candidate genes are usually selected from lists of genes with known functions. A fundamental problem with this strategy is an inherent selection bias dominated by well-characterized genes. Further-more, many genes are selected based on phenotypes identified from in vivo murine studies. Although mouse studies have provided powerful methods to dissect TB immunopathogenesis, the murine system models primary, progressive disease, which is only one of several phenotypes observed in humans. There are no wellestablished murine models of latent infection or the various types of disseminated disease, including TBM. Mtb intrathecal infection of rabbits recapitulates some of the inflammatory pathology but does not provide insight into the steps in immunopathogenesis involved in dissemination and invasion of the central nervous system [12]. To identify genes involved in TBM pathogenesis and to avoid gene selection bias, we chose to directly examine humans with different clinical types of TB with an array-based method to identify candidate genes.
Macrophages mediate the host innate immune response to Mtb through pathogen recognition and activation of an inflammatory response. Mtb resides in the macrophage phagolysosome, where it evades the immune response in the majority of infected individuals. Successful containment of Mtb replication results in LTB with no clinical symptoms, which depends on stimulation of innate and adaptive immune responses that lead to macrophage activation, formation of granulomas and elimination of the bacilli. In contrast, failure to contain bacilli replication is associated with active pulmonary disease and/or the development of disseminated disease. We hypothesized that different macrophage responses to Mtb are associated with distinct clinical outcomes that are genetically regulated.
Expression microarrays have been previously used to examine gene expression profiles in the immune response to TB [13][14][15][16][17]. None of these studies attempted to distinguish different clinical forms of active TB such as pulmonary and meningeal disease. In addition, the sample sizes were generally small and the findings were often not validated in independent sample sets. Finally, these previous approaches were not coupled with human genetic studies to examine the clinical significance associated with variation in the identified genes. In this manuscript, we examined ex vivo Mtbstimulated monocyte-derived macrophages (MDMs) from subjects with pulmonary, meningeal and latent infection. We attempted to find unique gene expression profiles to determine whether clinical phenotypes in TB are associated with distinct early macrophage responses to Mtb stimulation. We then used a case-control genetic association study to examine whether genetic variation of these selected genes was associated with susceptibility to Mtb.

Materials and Methods
Human subjects TBM subjects were recruited as part of a larger clinical study at the Hospital for Tropical Diseases, in Ho Chi Minh City (HCMC), Vietnam [18]. All subjects were .14 years of age and HIV-negative. TBM patients were described as having clinical meningitis (defined as nuchal rigidity and abnormal cerebrospinal fluid parameters) in addition to having a positive Ziehl-Neelsen stain for acid-fast bacilli and/or Mtb cultured from the cerebrospinal fluid. Subjects were treated for TBM and were clinically well (recovered for .3 years) when samples for this study were taken. For PTB subjects, samples were taken from individuals who had been previously treated and had recovered from uncomplicated PTB (no evidence of miliary or extrapulmonary TB). LTB subjects were defined as highly exposed individuals who had no history of active TB disease. LTB subjects were healthy nursing staff members who had worked at Pham Ngoc Thach Hospital for Tuberculosis and Lung Disease, HCMC, Vietnam for more than 20 years. They were tested for Mtb exposure using an ESAT-6 and CFP-10-specific IFN-c ELISPOT assay using a previously described method [19].
For the initial microarray study, twelve subjects were enrolled with three clinical forms of TB; TBM (n = 4), PTB (n = 4) and LTB (n = 4). All of the LTB subjects tested positive in the ESAT-6 and/ or CFP-10-specific IFN-c ELISPOT assay, suggesting previous or current infection with Mtb. An extended sample set containing 34 subjects with TBM (n = 10), PTB (n = 12) and LTB (n = 12) was used in validation experiments. Of the 12 LTB subjects, 10 were IFN-c ELISPOT positive according to our defined cut-off [at least 10 spot forming units (SFU) more than the negative PBS control and at least twice as many SFU as the negative PBS control]. The 2 IFN-c ELISPOT indeterminate LTB subjects had borderline responses (6.7 SFU with a ratio of 2 and 6 SFU with a ratio of 2.5) which were considerably higher than an unexposed population (average of 22.8 SFU with a ratio of 0.8).
For the case-control genetic association study the cohort of TBM (N = 114) and PTB (N = 159) patients, and population controls (cord blood; N = 188) has been previously described [20].
All samples came from unrelated individuals who were ethnic Vietnamese Kinh, as assessed by questionnaire. Written informed consent was obtained from each patient. Protocols were approved by human subjects review committees at the Hospital for Tropical Diseases and Pham Ngoc Thach Hospital for Tuberculosis and Lung Disease, Ho Chi Minh City, Vietnam. Ethical approval was also granted by the Oxford Tropical Research Ethics Committee, UK (OXTREC), The University of Washington Human Subjects Committee (USA) and the Western Institutional Review Board (USA).

Ex vivo generation and stimulation of MDMs
Peripheral blood mononuclear cells (PBMCs) were separated from heparinized whole blood by Lymphoprep (Asix-Shield, Norway) gradient centrifugation according to the manufacturer's protocol. From 20 ml of blood we obtained approximately 1-1.5610 7 PBMCs. To derive monocytes, PBMCs were plated in Nunclon Suface 6-well plates (Nunc, Denmark) containing RPMI-1640 (Sigma, Germany) with 10% heat-inactivated fetal calf serum

Author Summary
Although TB is a leading cause of death worldwide, the vast majority of infected individuals are asymptomatic and contains the bacillus in a latent form. Among those with active disease, 80% have localized pulmonary disease and 20% have disseminated forms. TB meningitis (TBM) is the most severe form of TB with 20-25% of sufferers dying, and of the survivors, many have disability. We currently do not understand the host factors that regulate this diverse spectrum of clinical outcomes. We hypothesized that variation in innate immune gene function is an important regulator of TB clinical outcomes. We measured the mRNA expression levels of .38,500 genes in macrophages taken from people with a history of latent, pulmonary, or meningeal TB and found genes with unique activation patterns among the clinical groups. Furthermore, we studied one of these genes further and found that CCL1 polymorphisms were associated with pulmonary TB (PTB) but not other types of TB disease. To our knowledge, this is the first study to combine mRNA expression studies with genetic studies to discover a novel gene that is associated with different clinical outcomes in TB. We speculate that this approach can be used to discover novel strategies for modulating immune function to prevent adverse outcomes in TB.
(FCS; Sigma, Germany), 2 mM L-glutamine and 100 units of penicillin for 2 hours at 37uC. Non-adhered cells were removed by washing with phosphate buffered saline (PBS) containing 3% FCS and adhered cells were incubated for 5 days at 37uC, 5% CO 2 to obtain MDMs. Cells were subsequently stimulated with PBS or 5 mg/ml of an irradiated, soluble, whole cell lysate of Mtb H37Rv [obtained from the Mycobacteria Research laboratories at Colorado State University, USA (http://www.cvmbs.colostate.edu/ microbiology/tb/top.htm)] for 4 hours before RNA extraction. Pilot studies indicated that 5 mg/ml was an optimal dose for stimulating TNF-a production.

RNA preparation and microarray hybridization
RNA was extracted from macrophages using Trizol according to the manufacturer's protocol (Invitrogen, USA), dissolved in RNase-free water and stored at 270uC until use. Total RNA (100 ng) was reverse transcribed to cDNA, amplified, labeled, and hybridized to the Human Genome U133 Plus 2.0 Array (Affymetrix, USA), according to the manufacturer's instructions. This array contains probe sets to measure the expression level of 47,000 transcripts, including 38,500 well-characterized human genes. Twelve Mtb-stimulated (TBM n = 4, PTB n = 4, and LTB n = 4) and 12 PBS-stimulated (hereafter called unstimulated) samples were hybridized to the array. The microarray data is publicly available at ArrayExpress, EMBL-EBI (Submission in progress, awaiting Accession number; http://www.ebi.ac.uk/ microarray-as/aer/?#ae-main[0]).

Microarray data processing and statistics
After normalization of the expression values, the data from 12 Mtb-stimulated samples were compared with the 12 unstimulated samples. Data were considered significant when (1) the false discovery rate (FDR) from the Significance Analysis of Microarray (SAM) analysis for the comparison of stimulated and unstimulated expression values was ,0.05, and (2) the P value of the comparison between stimulated versus unstimulated expression values by Student's t-test was ,0.05. In order to focus on highly regulated genes, we also restricted the majority of the analysis to genes with changes in expression levels of at least 2-fold. To compare gene expression levels among the three different clinical types of TB, we first calculated the fold stimulation of each gene for each individual by dividing the Mtb-stimulated value by the unstimulated control values. The averages of the 4 samples in each clinical group were calculated and then compared to the other groups by calculating the ratios of expression levels. The pair-wise comparisons included TBM vs. PTB, TBM vs. LTB, and PTB vs. LTB. SAM [20] was used to derive the FDR for microarray data, which is the proportion of genes likely to have been identified as significant by chance. Student's t-test and analysis of variance (ANOVA) were used to compare mean expression levels. To analyze expression patterns in multiple genes simultaneously we used Hierarchical Clustering [21]. Analyses were performed using MultiExperiment Viewer (MeV version 4.0, USA) [22] and SPSS (version 14.0, USA).

Real-time quantitative PCR
Taqman real time PCR was used to validate microarray gene expression results. cDNA was synthesized from total RNA samples using reverse transcription with Superscript II following the manufacturer's protocol (Invitrogen, USA). A commercial Low Density Array (LDA) format with Taqman probes and primers was then used for PCR validation (Applied Biosystems, USA). Expression levels in 88 genes [86 selected genes and 2 controls (GAPDH; Hs00237184_m1 and Hs00266705_g1)] were exam-ined in each sample according to the manufacturer's instructions. CCL1 gene expressions on human and mice were examined by using Taqman probes and primers (Applied Biosystems, USA). Samples were normalized to GAPDH and analyzed by using either Applied Biosystems SDS 2.1 Relative Quantification software or an Excel spreadsheet to perform relative quantification analysis.
CCL1 chemokine assay PBMC were isolated from whole blood and cytokine assays were prepared by plating 10 5 cell per well with RPMI (Life Technologies) in a 96-well dish, stimulating for 24 hours, and then harvesting supernatants.

Case-control genetic association study and statistics
SNPs in the CCL1 and CCR8 genes were genotyped in patients with TBM (N = 114), PTB (N = 159), and in Vietnamese Kinh population controls (N = 188). This genotyping was performed as part of a larger genome-wide genetic association study of TB using the Affymetrix 250K NspI Chip (unpublished). The whole genome SNP genotyping was performed according to the manufacturer's specifications and the data obtained was analyzed following rigorous quality control. Briefly, data quality control was performed using DM, BRLMM, RELPAIR, and manual viewing of cluster plots prior to statistical analysis. STRUCTURE and Eigentstrat were also used to analyse the population structure of the sample set. Genomic DNA quality was first assessed with 50 control SNPs and only samples with a call rate of greater than 93% were studied further. For each polymorphism in the full dataset, filter criteria were applied that included ,5% missing values and HWE P value.10 25 . Power for this study was calculated by using Power Calculator for Genetic Studies, CaTS version 0.0.2 (http://www.sph.umich.edu/csg/abecasis/CaTS). With a sample size of controls = 188 and PTB = 159 we have 82% power to detect an effect with an odds ratio of 2 for SNPs with an allele frequency of 10% and significance level of 0.01. With a sample size of controls = 188 and TBM = 114, we have a power of 71% to detect the same effects.
Genotyping was also carried out on selected CCL1 SNPs using a larger sample set TBM (N = 162), PTB (N = 175), and in Vietnamese Kinh population controls (N = 380). This was performed by a MassARRAY TM technique (Sequenom, San Diego, USA) using a chip-based matrix-assisted laser desorption/ ionization time-of-flight mass spectrometer as previously described [18]. All of the CCL1 SNPs genotyped by Sequenom were in Hardy Weinberg Equilibrium (HWE) (P.0.05) in population controls.
Univariate analysis was performed for categorical variables with a Chi-Square test. Two-sided testing was used to evaluate statistical significance.

Gene expression profiles in Mtb-stimulated and unstimulated MDMs
We hypothesized that macrophages from individuals with different TB clinical phenotypes have distinct gene expression profiles in response to Mtb stimulation. All subjects with pulmonary and meningeal disease had been treated and were free of symptoms at the time of venipuncture. Gene expression of MDMs from subjects with three clinical forms of TB including LTB, PTB, and TBM (n = 4 in each group) was examined by microarray. MDMs were stimulated either with a whole cell lysate of Mtb H37Rv or PBS for 4 hours. RNA expression was analyzed using a Human Genome U133 Plus 2.0 Array (Affymetrix, USA) which contains probe sets for 47,000 transcripts including 38,500 well-characterized human genes. We compared RNA transcription levels in Mtb-stimulated (n = 12) versus PBS-stimulated (n = 12) MDMs. 1,608 genes with a FDR of ,5% and a P value of ,0.05 by Students's t-test were differentially expressed by greater than 2-fold (Table 1). Of these genes, 1,260 were upregulated and 348 genes were down-regulated. A list of the 1,608 genes that were differentially expressed in the two groups (n = 24) with their mean expression intensities, FDR and P values are presented in Table S1. 74 genes were up-regulated more than 10fold, whereas only one gene was down-regulated by greater than 10-fold (Table 1). We used PANTHER (Protein Analysis Through Evolutionary Relationships; http://www.pantherdb.org/) to analyze the molecular functions and biological processes of genes induced and repressed in Mtb-stimulated MDMs. The changes in gene expression induced after stimulation contained 144 (8.4%) immunity and defense genes, including cytokines, chemokines, and receptors. Thirty six of these genes (25%) were up-regulated more than 10-fold. In contrast, no immunity and defense genes were repressed more than 10-fold. Other categories included; development (6.7%), protein and nucleic metabolism (19.2%) and signal transduction (11.9%). By comparison to the entire human genome, the proportion of immunity and defense genes is 5.2%. Percentages of other categories include: development (8.5%), protein and nucleic metabolism (25.1%) and signal transduction (13.4%).

Gene expression in different clinical phenotypes of TB (TBM, PTB, LTB)
To examine whether individuals with different clinical forms of TB have distinct gene expression profiles, we calculated the fold stimulation of each gene for each individual (dividing Mtb stimulated value by the unstimulated value) and then calculated the ratios of gene expression levels in each pair of TB forms. Six pair-wise comparisons in Table 1 show the change of gene expression between disease types (in fold change). 33 genes were differentially expressed between disease types with a ratio .10 and 228 genes had a ratio from 5 to 10.

Validation results
We used real-time PCR using a TaqMan Low Density Array technique to confirm microarray results in 86 genes in an extended sample set which included 12 LTB, 12 PTB, and 10 TBM individuals. Fifty-eight of the 86 genes were selected from the microarray data based on high levels of induction (.15 fold) or repression (.5 fold) following Mtb stimulation. Forty six genes were selected based on array expression differences among the 3 clinical groups (.5 fold). We first assessed whether the expression patterns of the 58 up and down-regulated genes were replicated in the independent sample set using RT-PCR. In total, 90% (52/58) of the microarray results were confirmed by RT-PCR when assessing Mtb and PBS-stimulated expression values in the validation sample set (Table 3 and Table S2). The RT-PCR results showed that 5/58 genes (IFIT1, CXCL6, MERTK, CD36, and MS4A6A) were not significantly induced or repressed by Mtb stimulation (n = 34; P.0.05 by t-test) and the expression pattern of one gene, CCR2, was reversed (Table 3). In addition, the majority of the genes in the validation group (n = 34) had a higher induction level in comparison to the microarray group (n = 12; Table 3).
We next compared gene expression levels in the 3 clinical groups in the validation sample set. The RT-PCR results showed that 2/46 genes (CCL1 and HS3ST3B1) were differentially expressed in groups with different TB phenotype (P,0.05 by ttest; Table 4). CCL1 was up-regulated in PTB when compared to LTB in both the RT-PCR LDA validation samples (P = 0.02 by ttest; 1.9-fold) and the initial microarray analysis (12.8-fold; Table 4 and Table S3). HS3ST3B1 was down regulated in LTB when compared to TBM in the RT-PCR LDA validation samples (P = 0.008 by t-test; ratio = 0.4) but this pattern of expression was reversed in the initial microarray analysis (ratio = 12.8) ( Table 4). Scatter plots of CCL1 and HS3ST3B1 are shown in Figure 1 along with 3 other representative genes. Seven other genes (INHBA, TSLP, LY6K, IL12B, MMP1, CCL20 and HAS1) had a greater than 2-fold change in expression ratios of the validation  samples in each pair-wise comparison, but these differences did not reach statistical significance (P.0.05; Table 4). These results suggest that the different TB clinical phenotypes cannot easily be distinguished by examining expression levels of single genes.

Cluster analysis of the 3 clinical phenotypes
We next hypothesized that expression profiles from multiple genes would need to be combined to detect patterns that could distinguish the different clinical disease phenotypes. We selected 1,608 highly induced or repressed genes from the microarray data set (Table S1) and used an unsupervised, hierarchical clustering algorithm [21] of 12 individual samples to attempt to distinguish the profiles of the 3 groups ( Figure S1). These results show that, (1) there was more relatedness between expression levels of samples from the same clinical group, i.e. L1 and L3 are very similar, P1, P2 and P3 are very similar, and M1 and M4 are very similar, and (2) one large cluster containing data from all TBM subjects, all PTB subjects and one LTB subject (L4) is very distinct to data from subjects L2, L1 and L3. Together, these findings suggest that cluster analysis can partially distinguish different clinical forms of TB.

CCL1 SNPs are associated with TB
CCL1 was the only gene whose expression was up-regulated in both the microarray and validation data sets when comparing clinical forms of TB (PTB vs LTB). We next examined whether genetic variants of CCL1 were associated with susceptibility to TB in a case-control study with TBM (N = 114) and PTB patients (N = 159), and population controls (N = 188) by using gene chip mapping assays. Forty nine SNPs were genotyped across a 200 kb region of the chromosome 17 CCL gene family cluster. Eight of the forty nine SNPs were associated with TB. To further locate the region associated with TB, we arbitrarily divided the whole region into four 50 kb sections. The first section containing CCL2 had 1/ 9 associated SNPs, the second containing CCL7 and CCL11 had 1/9 associated SNPs, the third containing CCL8 and CCL13 had 1/7 associated SNPs and the fourth containing CCL1 had 4/23 associated SNPs (Figure 2). To investigate this further we genotyped 10 SNPs nearby and in the coding region of CCL1 using Sequenom. Two more SNPs in the CCL1 gene were significantly associated with TB by genotypic comparison (Table 5). Together these results suggest that polymorphisms near and within the CCL1 genomic region are associated with susceptibility to different TB phenotypes.

Regulation of CCL1 Expression
To further investigate the role of CCL1 in Mtb pathogenesis, we examined regulation of its expression. We found that CCL1 mRNA expression was cell-specific and highly induced in monocytic (THP-1, U937, & PBMCs) cells stimulated with Mtb lysates or TLR ligands (LPS, PAM2, PAM3) ( Figure 3A). In contrast, no expression was found in epithelial cell lines (HeLa & A549, data not shown). We also found that CCL1 protein secretion was induced in THP1 cells and PBMCs by Mtb, including whole cell lysates, cell wall and cytosolic fractions [ Figure 3B and data not shown; PBS vs TB whole cell lysate (TBWCL; P = 0.01), PBS vs TB cell wall (TBCW; P = 0.006) and PBS vs TB cytosol (P = 0.02)]. Finally, we examined CCL1 expression in murine bone-marrow derived macrophages stimulated with PBS, LPS or Mtb from wild-type (WT) and Myd882/2 mice. CCL1 expression was highly induced by LPS and Mtb in WT bone marrow macrophages (BMMs). However, CCL1 expression was decreased in MyD88-deficient BMMs stimulated with LPS (P = 0.03) or Mtb (P = 0.002) ( Figure 3C). Together, these results suggested that CCL1 expression is highly enriched in monocytes and induced by Mtb components in a MyD88-dependent manner.

Discussion
In this study we examined macrophage transcriptional profiles in individuals with different clinical forms of TB. The majority of reported TB microarray studies have examined healthy donors,   cell lines or murine cells [13][14][15][16][17]. Only one previous study has compared gene expression profiles of individuals with different clinical forms of TB [24]. Mistry et al obtained whole blood from individuals with active, latent, cured (following 1 disease episode) and recurrent TB (following 2-3 episodes) [24]. Discriminant analysis suggested that 9 genes could distinguish the 4 clinical TB groups [24]. We examined these 9 genes in our data set and found these genes could not differentiate our latent and cured TB groups.
These differences may be attributable to the study design, which was substantially different from the current investigation with regard to cell population (whole blood vs MDMs), stimuli (none vs whole cell Mtb lysate), ethnic background (South African vs Vietnamese) and comparison of different clinical phenotypes. Despite these methodologic differences, both studies suggest that host gene expression profiles uniquely identify groups of individuals with different types of TB. Our study further illustrates     that macrophages, the primary host cell involved in TB pathogenesis, are a key source of the unique transcriptional profile that distinguishes clinical forms of TB. One limitation of our study was the small sample size. Although this is the largest number of individuals ever studied in a TB microarray study, comparable only to the study by Mistry et al [24], the sample size remains small for this statistically challenging question. To overcome some of the limitations of a small sample size for microarrays (n = 12), we included an independent set of samples for validation (n = 34). We also chose to use a whole cell lysate of a standardized Mtb strain rather than live organisms and a relatively short stimulation time (t = 4 hours) to minimize variation in our stimulation conditions and to enhance the detection of early innate immune response genes. We examined these cells in an ex vivo environment to avoid variability that is attributable to complex in vivo environments. For example, we studied individuals after they had been treated for TB to avoid detecting gene expression changes that are attributable to stimulation of in vivo inflammatory pathways from active disease. We also chose to study macrophages rather than whole blood in order to concentrate on a single cell population that is most relevant for Tb pathogenesis. A number of studies have shown that the strain of Mtb induces different immune responses [25,26]. Although the choice of Mtb strain could stimulate different gene expression profiles, we chose to study the commonly used laboratory strain (Mtb H37Rv). Each of these experimental conditions was selected to maximize the opportunities of detecting differences attributable to genetic variation in the macrophage innate immune response to TB. Comparison of gene expression results with alternative experimental conditions (such as different cell types, Mtb strains, Mtb growth conditions, and time points) could further illuminate the role of these genes in Tb pathogenesis.
In addition to comparing expression profiles among people with different types of TB, our study contributes further data on the set of genes that are activated in response to Mtb stimulation of macrophages. Our results demonstrated that 1,608 genes in macrophages were stimulated (up or down-regulated) by Mtb. Furthermore, 90% of a subset of these genes (n = 58 genes induced .15 fold by Mtb stimulation) in a second round validation also showed altered expression. Many genes identified in our study have also been detected in previous studies investigating the host response to Mtb infection [13,16,17]. Ragno     genes significantly induced following 6 hr stimulation in their data set (MIP-1a, MIP-1b, MIP-3a, MPIF-1, PARC, RANTES, IL-8,  GRO-a, GRP-b, GRO-c, CCL1, CCR3, IL-1b, TNFa, and VEGF) [17]. Nau et al studied primary human MDMs stimulated with live Mtb [16]. Eleven genes were highly expressed in both data sets (TNFAIP6, CXCL3, CXCL1, CCL4, PTGS2, SER-PINB2, PTX3, INHBA, TRAF1, JAG1, and SOD2) and 3 genes were inhibited (MERTK, GLUL, and DAB2). These gene lists include cytokines, chemokines and immune receptors, which may be involved in inflammatory responses in the early phases of defense against Mtb. All of the up-regulated genes identified by Nau et al were found in our dataset [16]. In contrast, only 50% (24/50) of highly expressed genes in our dataset were identified by Nau et al, a difference that is likely due to the array sizes that were utilized (38,000 vs. 980 genes). Although these microarray studies have important methodologic differences (e.g primary cells vs cell lines, healthy subjects vs. TB patients, live versus dead Mtb stimulation, stimulation times, arrays and genes analyzed), all of these studies have identified novel genes potentially related to the host macrophage response to Mtb.
Our study compares transcriptional profiles of individuals with TBM with individuals with other forms of TB. We identified genes that were distinctly expressed in macrophages from individuals with a history of TBM. After bacilli invade the host lung within the pulmonary alveolar macrophage, they replicate and disseminate to the regional lymph nodes. During this early stage of infection, before the development of adaptive immunity, the bacteria can spread haematogenously to other organs in the body and cause extrapulmonary disease, such as TBM [27,28]. This step may be determined by the nature and extent of the innate immune response activated by infected macrophages. We found that several macrophage immune response genes (IL1B, IL12B, TNF, TNIP3, CXCL10, CXCL11, CCL12, and CCL1) were upregulated in TBM subjects in comparison to those with PTB and LTB. In addition, some genes, such as MMP1 and HAS1, were found with differing expression in PTB and TBM patients. These genes are involved in degrading the extracellular matrix and could mediate a role in granuloma formation and bacillus containment, which could influence dissemination and development of TBM [23]. Although the relationship between the inflammatory response and TBM pathogenesis is only partially understood, excessive immune activation may be intimately associated with disease severity and outcome.
Case-control genetic association studies of biologically plausible candidate genes have been performed with the hope to identify genes involved in susceptibility to, and clinical outcome of, TB. However it has always been challenging to identify potential candidate genes in an unbiased manner. The expression profiling study we describe here can serve as a hypothesis generating, unbiased methodological approach to identify genes for potential association studies. Despite this advantage, gene regulation is not the only mechanism for genetic resistance or susceptibility and non-synonymous coding region SNPs which alter protein structure and function also play an important role. From the genes that were differentially expressed between TB disease types, as assessed by microarray, we tested 46 genes in a separate, larger sample set by RT-PCR. The expression of only one of these genes, CCL1, remained significantly different between patients with different clinical TB outcomes. To test our selection approach we performed a case-control genetic association study and found that SNPs near CCL1 were associated with susceptibility to PTB. The fact that SNPs near CCL1 were significantly associated with PTB in our study highlights the feasibility of this unbiased selection approach.
Even though the associated SNPs are not within the CCL1 coding region, it is a likely candidate gene due to it's proximity to the cluster of associated SNPs and its functional relevance. CCL1, like other members of the CC chemokine family, is an inflammatory mediator that stimulates the migration of human monocytes [29]. CCL1 is produced by monocytes (as well as other cells) and binds its receptor CCR8, which is present on lymphocytes and monocytes [30]. Interestingly, CCR8 has enriched expression on Th2 and regulatory T cells and may influence the development of Th2 type T cell responses in vivo [31,32]. In addition, CCR8 regulates migration of dendritic cells to lymph nodes [33]. Hoshino et al [34] found that the expression of CCR8 was specifically up-regulated by CCL1 stimulation of peritoneal macrophages, which may lead to cell aggregation at a site of tissue damage. In the lungs, CCL1 expression was upregulated in Mycobacterium bovis purified protein derivative (PPD) induced granulomas [35]. In this study, we found that CCL1 expression was induced by Mtb and TLR ligands in several monocyte/macrophage lineages. Furthermore, we found that its expression was MyD88-dependent when cells were stimulated with LPS or Mtb. Genetic variation leading to the loss or alteration of CCL1 function may influence the ability of T cells, monocytes and dendritic cells to migrate to the site of infection, aggregate into granulomas and develop an effective immune response. This may result in inadequate containment of the bacterium and allow unimpeded bacterial growth leading to pulmonary disease.
With currently available tools, clinicians are unable to identify the subset of latently infected patients who will develop active disease. Furthermore, there are no techniques available to prospectively identify individuals at risk for the devastating consequences of TBM versus more treatable forms of TB such as localized pulmonary disease. Further studies in this area could lead to tests that could alter treatment algorithms with more accurate prognostic information. In addition, such studies may lead to novel molecular insight into TB pathogenesis.   Vuong Obstetric Hospital, and all Vietnamese individuals who were recruited into this study. We also thank the technical assistance and advice received from Ms. Eileen Png and Dr. Annelie Lundin at the Genome Institute of Singapore and Richard Wells at the University of Washington.