We previously reported that the – 2518 MCP-1 genotype GG increases the likelihood of developing tuberculosis (TB) in non-BCG-vaccinated Mexicans and Koreans. Here, we tested the hypothesis that this genotype, alone or together with the – 1607 MMP-1 functional polymorphism, increases the likelihood of developing TB in BCG-vaccinated individuals. We conducted population-based case-control studies of BCG-vaccinated individuals in Mexico and Peru that included 193 TB cases and 243 healthy tuberculin-positive controls from Mexico and 701 TB cases and 796 controls from Peru. We also performed immunohistochemistry (IHC) analysis of lymph nodes from carriers of relevant two-locus genotypes and in vitro studies to determine how these variants may operate to increase the risk of developing active disease. We report that a joint effect between the – 2518 MCP-1 genotype GG and the – 1607 MMP-1 genotype 2G/2G consistently increases the odds of developing TB 3.59-fold in Mexicans and 3.9-fold in Peruvians. IHC analysis of lymph nodes indicated that carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G express the highest levels of both MCP-1 and MMP-1. Carriers of these susceptibility genotypes might be at increased risk of developing TB because they produce high levels of MCP-1, which enhances the induction of MMP-1 production by M. tuberculosis-sonicate antigens to higher levels than in carriers of the other two-locus MCP-1 MMP-1 genotypes studied. This notion was supported by in vitro experiments and luciferase based promoter activity assay. MMP-1 may destabilize granuloma formation and promote tissue damage and disease progression early in the infection. Our findings may foster the development of new and personalized therapeutic approaches targeting MCP-1 and/or MMP-1.
Citation: Ganachari M, Ruiz-Morales JA, Gomez de la Torre Pretell JC, Dinh J, Granados J, et al. (2010) Joint Effect of MCP-1 Genotype GG and MMP-1 Genotype 2G/2G Increases the Likelihood of Developing Pulmonary Tuberculosis in BCG-Vaccinated Individuals. PLoS ONE 5(1): e8881. doi:10.1371/journal.pone.0008881
Editor: Carlos Penha-Goncalves, Instituto Gulbenkian de Ciência, Portugal
Received: September 28, 2009; Accepted: January 1, 2010; Published: January 25, 2010
Copyright: © 2010 Ganachari et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by the National Institutes of Health grant R01 HL084347 and by recruitment funds from the Methodist Hospital Research Institute (TMHRI), Houston, Texas. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
We previously reported an association between the -2518 functional promoter polymorphism (-2518A>G) of the Monocyte Chemoattractant Protein (MCP)-1 gene and susceptibility to pulmonary tuberculosis (TB) in non BCG-vaccinated Mexicans and Koreans . The chromosome region 17q11.2, where the MCP-1 gene is located, was identified as a TB susceptibility region by linkage analysis of multi-case families . Fine mapping of this region using linkage disequilibrium association mapping and potentially functional SNPs revealed association of the -2518 MCP-1 SNP the with disease . Given that most of individuals living in endemic countries are vaccinated with BCG, we wondered whether the – 2518 MCP-1 susceptibility allele G and GG genotype also increase the odds of developing disease in BCG-vaccinated individuals.
The Matrix Metalloproteinase (MMP)-1 gene, which encodes a potent collagenase, is an excellent candidate gene for influencing the expression of pulmonary TB since inflammation-mediated tissue damage may contribute to the spread of M. tuberculosis infection , . Persistent degradation of the extracellular matrix (ECM) during inflammation leads to tissue damage , . Collagen is an essential component of lung ECM and of M. tuberculosis-induced granulomas ,  and provides support to macrophages and T-cells, which play a role in the organization of these immune-mediated granulomas , . In the context of granulomatous inflammation, the collagenase activity of MMP-1 is involved in inflammation-induced tissue damage in several lung diseases  and may contribute to the destabilization of granuloma formation and the damage of adjacent tissues in TB , . As granulomas in TB prevent the spread of infection, MMP-1 may contribute to the dissemination of infection and to disease progression . A functional polymorphism located in the promoter region of MMP-1 enhances expression of the gene . This polymorphism consists of the insertion of a guanine at position – 1607 (-1607_1608insG) and creates an Ets-1 transcription factor binding site . Furthermore, a genome wide analysis showed that the chromosome region 11q14–11q23 may contribute to the regulation of TNF-alpha in TB, and the MMP-1 gene is located within that region at 11q22.2 .
Given that both the -2518 MCP-1 allele G and the -1607 MMP-1 allele 2G increase expression of the genes, and MCP-1 contributes to ECM remodeling through the induction of MMP-1 expression in bronchial epithelial cells , , , we tested the hypothesis that these biologically functional polymorphisms, alone or in combination, might influence the expression of phenotypes prone to disease progression.
We selected Mexicans and Peruvians for this study because they have similar ethnic backgrounds but are exposed to different environmental conditions. Moreover, the TB burden is considered to be low in Mexico (0 to 24 estimated new TB cases per 100,000 population), but moderate to high in Peru (100 to 299 estimated new TB cases per 100,000 population) .
Clinical Features of Cases and Controls
All study subjects were vaccinated with BCG at birth. We studied 193 and 701 sputum smear-positive cases of TB from Mexico and Peru, respectively (Table 1). None of the patients had history of any preconditions that may increase the chance of developing active TB, including preconditions affecting immune function, malnutrition and other lung diseases. Their body mass indexes (BMI) were ≥18.5 kg/m2 before disease onset (Table 1). In addition, all cases were new TB cases: none had a history of previous TB or had received previous treatment for TB. Thus, these TB patients were very unlikely to be cases of re-infection or recidivating TB. The patients developed evident symptoms of disease less than 1 year after exposure to a TB case. All patients had active TB disease of at least three months of evolution with chest radiographs showing alveolar infiltrates in the lungs. Hilar adenopathy was present in most Mexican (98%) and Peruvian cases (96%). None of the patients had multi-drug resistant (MDR)-TB.
Table 1. Demographic and clinical features of tuberculosis cases and controls A.doi:10.1371/journal.pone.0008881.t001
We also studied 243 and 796 healthy tuberculin-positive (PPD+) control individuals in Mexico and Peru, respectively (Table 1). Approximately 91% of Mexican and 98% of Peruvian controls were household contacts of a sputum smear-positive TB case, the rest were close contacts of community cases. All controls showed a PPD response >15 mm of induration and remained healthy for a period of more than 2 years after exposure, making it likely that they were latently infected individuals with capacity to control the spread of infection during that critical period –.
All cases and controls were negative in serologic tests for HIV infection. TB cases and controls were similar in demographics, BMI (prior to development of disease in the case of TB patients), and consumption of cigarettes and alcohol, both within and between populations (Table 1). We did not have a significant proportion of individuals with these habits among our cases or controls in either of the populations studied. Cases and controls were of similar socioeconomic status within nationalities and all individuals recruited were of low socioeconomic status. We attained successful genotyping rates in our cases and controls from Mexico and Peru.
Mexican and Peruvian cases and controls were Mestizo to the third generation. Mestizos from Mexico and Peru are mainly an admixture of Spanish, Amerindians and Africans. None of our cases or controls were pure Amerindians, neither pure Caucasians nor pure non-Caucasian Africans. Thus, the presence of sub-populations in our samples of cases or controls from Peru and Mexico is unlikely. Mexican Mestizos from Mexico D.F. (the Federal District) have an admixture of Spanish (50.03±4.11%), Amerindian (49.03±3.76%), and African (0.94±1.27%) . The genetic background of Peruvian Mestizo is close to that of Mexican Mestizo ; however, since subtle differences in the admixture of cases and controls may introduce bias we applied a methodology developed by Reich and Golstein to adjust for differences in the genetic admixture of cases and controls as explained in Materials and Methods . The results of these corrections are provided in Table 2 and discussed in the next section.
Table 2. Univariate analysis of MCP-1 and MMP-1 alleles and progression to pulmonary tuberculosis A.doi:10.1371/journal.pone.0008881.t002
Based on the frequency of alleles at the MCP-1 and MMP-1 loci in both populations, we estimated that given the size of both samples we had 90% power to detect an odds ratio (OR) of 1.5, which was significant at an α of 0.025. Considering all the characteristics of our samples and our genotyping success rate, it is unlikely that we obtained our results because of selection bias (including poor ascertainment of the phenotypes) or information bias (including genotyping errors), or due to an unadjusted known confounder or type I and II errors.
Univariate Analysis of the -2518 MCP-1 and -1607 MMP-1 Promoter Polymorphisms and Susceptibility to TB in BCG-Vaccinated Individuals
The distribution of MCP-1 and MMP-1 genotypes in Mexican and Peruvian controls were in Hardy-Weinberg equilibrium. The genomic control SNPs were all in Hardy-Weinberg equilibrium (Table S1) . We observed a significant association of the MCP-1 susceptibility allele G with disease progression in Both Mexicans (OR = 1.75; 95% CI 1.3–2.3) and Peruvians (OR = 1.29; 95% CI 1.1–1.5) (Table 2). This association remained significant after correction for genetic admixture and for the number of comparisons in both Mexicans and Peruvians (Table 2). The MCP-1 genotype GG was significantly associated with TB in Mexicans (OR = 2.66, 95% CI 1.47–4.79, p = 0.001, Table 3) and marginally associated with disease in Peruvians (OR = 1.43, 95% CI 1.02–2.0, p = 0.036, Table 4).
Table 3. Univariate analysis of MCP-1 and MMP-1 genotypes and progression to pulmonary tuberculosis in Mexicans.doi:10.1371/journal.pone.0008881.t003
Table 4. Univariate analysis of MCP-1 and MMP-1 genotypes and progression to pulmonary tuberculosis in Peruvians.doi:10.1371/journal.pone.0008881.t004
We noticed that our current point estimates of the -2518 MCP-1 allele G and genotype GG associations with disease were lower than those previously published for individuals non-vaccinated with the BCG . Thus, we pooled the information from Mexican cases and controls recruited for the previous and present studies. We observed an almost 2-fold drop in the OR for the association of the -2518 MCP-1 genotype GG with disease progression in BCG-vaccinated individuals (OR = 2.66) when compared with the total Mexican sample (OR = 4.2) or non-vaccinated individuals (OR = 5.4) (Figure S1). In addition, the dose effect of the allele G observed in non-vaccinated was lost in BCG-vaccinated individuals and we did not see a significant association of the MCP-1 genotype AG with disease in this stratum. Given that the odds of developing active disease (reflected by the measured OR) were not homogeneous in strata formed by BCG vaccination status (Figure S1), we concluded that BCG vaccination is a modifier of the effect of the MCP-1 susceptibility allele G and genotype GG on progression to disease .
Our results from analysis of the -1607 MMP-1 polymorphism revealed that this polymorphism influences the expression of active disease in both, Mexican and Peruvian populations. Although the point estimates of main effects (ORs) were limited in strength, we observed an association of the -1607 MMP-1 allele 2G with susceptibility to developing TB in both Mexicans and Peruvians (Table 2). The proportion of carriers of the MMP-1 allele 2G was increased among Mexican (OR = 1.56. 95% CI 1.14–2.16, Table 2) and Peruvian (OR = 1.4, 95% CI 1.19–1.67, Table 2) TB cases. This association remained significant after corrections for population stratification and number of comparisons. Carriers of the MMP-1 genotype 2G/2G were significantly over-represented in Mexican TB cases (OR = 2.71, 95% CI 1.15–6.3, p = 0.02, Table 3) and in Peruvian TB cases (OR = 1.6, 95% CI 1.05–2.4, p = 0.027, Table 4).
The population admixture coefficients (λ) calculated for Mexicans (1.332) and Peruvians (1.2315) were low and did not significantly modify our estimates of associations of alleles with disease (Table S1 and Table 2) . We did not see major differences in the admixture background of our cases and controls, or these differences were not significant enough to bias our analysis of associations .
Multivariate Analysis of Main and Joint Effects of the -2518 MCP-1 and -1607 MMP-1 Genotypes in Susceptibility to Developing Active Pulmonary TB
Based on the results obtained in the univariate analysis of associations we collapsed MCP-1 and MMP-1 genotypes that were not found to be associated with TB in both populations. Thus, the MCP-1 genotypes AA and AG and MMP-1 genotypes 1G/1G and 1G/2G were collapsed into MCP-1 genotype A/- and MMP-1 genotype 1G/-, respectively. Collapsing non-significant genotypes at each locus allowed us to test the hypothesis that the joint effects of susceptibility genotypes -2518 MCP-1 GG and -1607 MMP-1 2G/2G are greater than that of the independent (main) effects of these two genotypes. We obtained a significant and consistent joint effect of the two-locus genotype MCP-1 GG and MMP-1 2G/2G in both Mexicans and Peruvians. The ORs observed in carriers of the two-locus genotype MCP-1 GG and MMP-1 2G/2G were 3.59 (95% CI 1.54–8.33, p = 0.003, Table 5) in Mexicans and 3.9 (95% CI 2.56–5.95, p = 0.0001, Table 6) in Peruvians. In Peruvians, the logistic regression adjusted OR estimated for the carriers of the MCP-1 genotype GG that did not carry the MMP-1 genotype 2G/2G was 0.69 (p = 0.02). This was because most of the Peruvian cases carrying the MCP-1 genotype GG also carried the MMP-1 genotype 2G/2G (Table 6).
Table 5. Multivariate analysis of main and joint effects of MCP-1 and MMP-1 tuberculosis susceptibility genotypes in Mexicans.doi:10.1371/journal.pone.0008881.t005
Table 6. Multivariate analysis of main and joint effects of MCP-1 and MMP-1 tuberculosis susceptibility genotypes in Peruvians.doi:10.1371/journal.pone.0008881.t006
We used the Pearson goodness-of-fit x2 and the Hosmer-Lemeshow (HL) x2 tests to estimate the magnitude of the differences between the event rate predicted by the logistic models and that observed in the cases and controls to which the models were applied , . The null hypothesis for these tests states that the expected and observed event rates predicted by the model are not significantly different. A p-value >0.1 indicates that the model provides a good fit for the data since at this level of significance the null hypothesis cannot be rejected , . The full model (main and interaction terms) and the interaction model alone fitted the observed data. Indeed, the statistic estimates (x2 and p values) were the same for the full and the interaction models in Mexicans and Peruvians (full model: Pearson x2 = 0, p>0.1 and HL x2 = 0, p>0.1; interaction model: Pearson x2 = 0, p>0.1 and HL x2 = 0, p>0.1). These results clearly indicate that the interaction model alone was sufficient and provides a better fit for the observed data. Consistent with this observations, a model with the main effects alone was not sufficient to explain the data observed and the null hypothesis was rejected in Mexicans (Pearson x2 = 8.95, p = 0.0028; HL x2 = 8.95, p = 0.0114) or Peruvians (Pearson x2 = 40.56, p<0.00001; HL x2 = 17.83, p<0.00001). In addition, the ORs obtained for the joint effects were greater than the expected assuming additivity in both populations. Indeed, the ORs calculated for the additive model were 1.21 and 0.55 in Mexicans and Peruvians, respectively (Tables 5 and 6, Materials and Methods section) . Thus, we concluded that the MCP-1 and MMP-1 genotypes GG and 2G/2G jointly increase the odds of developing TB in Mexicans and Peruvians, and that these joint effects are multiplicative .
Immunohistochemistry (IHC) Analysis of Lymph Nodes from Tuberculosis Cases Carriers of Relevant MCP-1 and MMP-1 Genotypes
Early events are relevant for progression from infection to disease , . Consequently, we analyzed lymph nodes from untreated Peruvian TB cases that were experiencing a local (node) inflammatory process with less than 15 days of evolution. We assessed whether carriers of the MCP-1 genotype GG and MMP-1 genotype 2G/2G produce increased levels of MCP-1 and MMP-1 at an early stage of the inflammatory process , . We used immunohistochemistry (IHC) since the available specimens were archived samples fixed in paraformaldehyde and embedded in paraffin, and in our experience protein epitopes are better preserved than mRNA in such samples. Since MMP-1 functions are tightly regulated by tissue inhibitors of metalloproteinases (TIMPS) in tissues undergoing inflammation –, we used a polyclonal antibody that detects only free MMP-1 (pro-MMP-1 and mature activated MMP-1).
All nodes tested showed mainly paucicellular and fibrotic old granulomas, and a great deal of tissue damage as reflected by the loss of normal lymph node histology (Figure 1 and 2). After examining the slides, two trained observers arrived at the consensus that carriers of the genotypes MCP-1 GG and MMP-1 2G2G have larger extensions of tissue containing cells expressing the highest amounts of MCP-1 and MMP-1 (dark red) than carriers of any other two-locus genotypes investigated (Kappa coefficient of agreement = 0.67, Figure 1). Increased production of MCP-1 and MMP-1 might increase the chances of detecting positive cells, which could result in a perceived increased in the number of cells expressing these factors (a threshold effect). For this reason, we also examined whether differences in the proportion of cells expressing high levels of MCP-1 and MMP-1 could be detected between tissue samples from carriers of the relevant two-locus genotypes under study (Table 7). The proportion of cells expressing these two factors (number of positive cells divided by the total number of cells counted) was approximately 4- to 6-fold higher for MCP-1 and approximately 6- to 13-fold higher for MMP-1 in carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G than in carriers of any other relevant two-locus genotypes studied (Table 7). We found significant differences between carriers of the relevant two-locus genotypes studied in the mean number of cells stained positive for MCP-1 (ANOVA F = 232.68, p = 0.00001) and in the mean number of cells stained positive for MMP-1 (ANOVA F = 244.43, p = 0.00001). The Bonferroni pair-wise comparison indicates that the number of cells expressing MCP-1 was significantly greater in carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G than in carriers of two-locus genotypes MCP-1 A/- MMP-1 1G/- (p<0.001), MCP-1 A/- MMP-1 2G/2G (p<0.001), or MCP-1 GG MMP-1 1G/- (p<0.001). No other pair-wise comparisons were significant. Likewise, the number of cells expressing MMP-1 in carriers of the genotypes MCP-1 GG MMP1 2G/2G was significantly greater than in carriers of genotypes MCP-1 A/- MMP-1 1G/- (p<0.001), MCP-1 A/- MMP-1 2G/2G (p<0.001), or MCP-1 GG MMP-1 1G/- (p<0.001). The Bonferroni pair-wise comparison of carriers of genotypes MCP-1 A /- MMP-1G/- and MCP-1 GG MMP-1 1G/- also showed significant differences in the number of MMP-1 positive cells (p = 0.019), while other comparisons did not show significant differences. Thus, carriers of the two-locus MCP-1 MMP1 genotype GG 2G/2G have the highest numbers of cells expressing both MCP-1 and MMP-1.
Figure 1. Carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G express more MMP-1 than carriers of any other two-locus genotypes tested.
Immunohistochemical (IHC) analysis of MMP-1 expression in paraffin-embedded lymph-nodes from four representative Peruvian TB cases. Negative controls are shown in Figure S3. Representative lymph-nodes from each of the following two-locus genotypes are shown: A, MCP-1 A- MMP-1 1G/-; B, MCP-1 A/- MMP-1 2g/2G; C, MCP-1 GG MMP-1 1G/-; D, MCP-1 GG MMP-1 2G/2G. Images were acquired at 100× total magnification. Carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G showed larger extensions of tissue and more cells that stained dark red than tissue sections from carriers of any other combination of genotypes.doi:10.1371/journal.pone.0008881.g001
Figure 2. Carriers of the two-locus genotype MCP-1 G/G MMP-1 2G/2G express more MCP-1 than carriers of other two-locus genotypes tested.
Immunohistochemical (IHC) analysis of MCP-1 expression in paraffin-embedded lymph-nodes from four representative Peruvian TB cases. Negative controls are shown in Figure S3. Representative lymph-node from each of the following two-locus genotypes is shown: A, MCP-1 A- MMP-1 1G/-; B, MCP-1 A/- MMP-1 2g/2G; C, MCP-1 GG MMP-1 1G/-; D, MCP-1 GG MMP-1 2G/2G. Images were acquired at 100× total magnification. Carriers of two-locus genotype MCP-1 GG MMP-1 2G/2G showed larger extensions of tissue and more cells that stained dark red than tissue sections from carriers of any other combination of genotypes.doi:10.1371/journal.pone.0008881.g002
Table 7. Proportion of cells expressing high levels of MCP-1 and MMP-1 in lymph nodes from cases of tuberculosis carriers of -2518 MCP-1 and -1607 MMP-1 two-locus genotypes A.doi:10.1371/journal.pone.0008881.t007
Cells producing large amounts of MMP-1 and MCP-1 were mainly located in areas surrounding old granulomas (Figures 1 and 2). Cells producing copious amounts of MCP-1 and MMP-1 (dark red) were also observed adjacent to or surrounding the necrotic areas (Figure S2). Monocytes produce MCP-1 and MMP-1 in response to M. tuberculosis antigens and their presence in M. tuberculosis-induced granulomas is a hallmark of TB infection (1, 28–30). Consistent with this, we observed that cells of the monocytic lineage were among the most important sources of both, MCP-1 and MMP-1 (Figure S2). A panel of appropriate negative controls is shown in Figure S3.
Analysis of MCP-1 and MMP-1 Biological Interactions
We developed an in vitro model to study whether MCP-1 interacts with MMP-1at the biological level. To test this in a meaningful way we used cells of the monocytic lineage, which are important constituents of granulomatous lesions in TB, and stimulated them with M. tuberculosis-sonicate antigens, human rMCP-1, or both. THP-1 monocytic cells were first stimulated with increasing amounts of M. tuberculosis-sonicate antigens or human rMCP-1 and expression of MMP-1 was monitored by real-time PCR. Stimulation with M. tuberculosis-sonicate antigens induced MMP-1 expression in a dose-dependent manner whereas human rMCP-1 alone did not (Figure 3A and 3B). As a positive control, human rMCP-1 alone induced the expression of CD11b by THP-1 cells, in a dose-dependent manner (data not shown). Stimulation of cells with M. tuberculosis-sonicate antigens alone significantly increased the expression of MMP-1 (Z = −3.85, p = 0.0001, Figure 3C). Remarkably, addition of exogenous human rMCP-1 to THP-1 cells stimulated with M. tuberculosis-sonicate increased the expression of MMP-1 mRNA in a dose-dependent fashion and to higher levels than in cells stimulated with antigen alone (Kruskal-Wallis x2 = 27.13, p = 0.0001, Figure 3C). We observed significant differences in the levels of MMP-1 mRNA when comparing cells stimulated with M. tuberculosis antigen alone versus cells stimulated with antigen plus 1000 pg/ml of rMCP-1 (Z = −3.62, p = 0.0003), 2000 pg/ml of rMCP-1 (Z = −2.817, p = 0.0049), or 4000 pg/ml of rMCP-1 (Z = −2.012, p = 0.0442) (Figure 3C). The levels of MMP-1 mRNA expression were significantly greater in cells stimulated with M. tuberculosis-sonicate plus the various concentrations of rMCP-1 than in cells stimulated with M. tuberculosis-sonicate antigens alone, with the exception of cells stimulated with M. tuberculosis-sonicate antigens plus 500 pg/ml of rMCP-1 (Figure 3C). Addition of rMCP-1 alone at various concentrations did not induce MMP-1 expression.
Figure 3. MCP-1 increases MMP-1 expression inTHP-1 monocytic cells stimulated by M. tuberculosis-sonicate antigens.
We measured the relative changes in MMP-1 gene expression by real-time PCR. Data are presented as the fold change in gene expression normalized to the endogenous reference gene PDHB and relative to untreated controls (RQ values). A, MMP-1 expression following 24 hours in vitro stimulation of THP-1 cells with the indicated amounts of M. tuberculosis-sonicate antigens. The effective dose 50 (ED50) of the M. tuberculosis-sonicate is indicated. B, MMP-1 expression data following 24 hours in vitro stimulations of THP-1 cells with the indicated amounts of human recombinant MCP-1 (rMCP-1). Graph A and B show data from one of three experiments. C, MMP-1 expression data for non-stimulated (−) and M. tuberculosis-sonicate antigen stimulated (+) THP-1 cells that were cultured for 24 hours with the indicated amounts of human rMCP-1. The results presented are from three independent experiments showing minimum and maximum RQ values. D, MMP-1 secretion levels for non-stimulated (−) and M. tuberculosis-sonicate antigen stimulated (+) THP-1 cells that were cultured for 48 hours in serum-free media with or without increasing amounts of human rMCP-1. Results presented are from three independent experiments and the bars indicate the standard deviation from the mean. The ED50 of M. tuberculosis-sonicate was used to stimulate the cells for the experiments shown in B, C, and D. The p-values from Wilcoxon-Mann-Whitney U-tests of comparisons of two independent variables are shown in C and D. Antigen = M. tuberculosis-sonicate 1000 ng/ml.doi:10.1371/journal.pone.0008881.g003
Accumulation of free MMP-1 protein (not neutralized by TIMPs or alpha-2-macroglobulins) was preserved in this in vitro system (Figure 3D) . The data reported in Figure 3D were obtained using a functional assay that assessed levels of active protein in the culture media of stimulated cells (see Materials and Methods). Although it is evident that rMCP-1 induces a dose-dependent incremental accumulation of MMP-1 in culture media, testing for differences in the mean MMP-1 concentrations between variables was not significant using the Kruskal-Wallis test (Figure 3D). However, the accumulation of MMP-1 was greater in cells stimulated with M. tuberculosis-sonicate antigens than that in non-stimulated cells (Z = −3.127, p = 0.0018). Moreover, the accumulation of MMP-1 in culture media of cells stimulated with M. tuberculosis-sonicate antigens plus rMCP-1 was always significantly greater than in that culture media from cells stimulated with antigen alone. The Z and p-values obtained using the Wilcoxon-Mann-Whitney U-test to compare two independent samples were always significant and did not change in magnitude over all comparisons performed (Z = −2.966, p = 0.03, Figure 3D) because there were no statistically significant differences in the levels of MMP-1 accumulation in culture media of cells stimulated with antigen plus various doses of rMCP-1. Nevertheless, the results from these experiments support the idea that MCP-1 potentiates the induction of MMP-1 production following stimulation of cells of the monocytic lineage by M. tuberculosis-sonicate antigens. Non-stimulated control cells did not show significant MMP-1 secretion (Figures 3C and 3D). Hence, we concluded that MCP-1 potentiates MMP-1 mRNA expression and MMP-1 secretion by cells of the monocytic lineage following stimulation by M. tuberculosis-sonicate antigens.
Analysis of the Effects of M. tuberculosis-Sonicate Antigens and MCP-1 on the Activation of -1607 MMP-1 Promoter Variants 1G and 2G
Our in vitro model also allowed us to assess whether the -1607 MMP-1 allele 2G induces a higher promoter activity than the allele 1G in cells of the monocytic lineage that are simulated with M. tuberculosis-sonicate antigens alone or in the presence of human rMCP-1. We used a dual luciferase assay to control for the efficiency of transfections . The results of the luciferase assays were in agreement with those obtained from the ex vivo IHC analysis of lymph nodes and in our in vitro experiments outlined above. M. tuberculosis-antigen stimulated cells transfected with vector containing the -1607 MMP-1 allele 1G expressed higher luciferase activity than non-stimulated cells transfected with the same vector, but the difference was marginally significant (t = 2.48, p = 0.048) (Figure 4). In contrast, antigen stimulated cells transfected with vector containing the allele 2G expressed significantly higher luciferase activity than non-stimulated cells transfected with the same vector (t = 6.99, p = 0.0004, Figure 4). Addition of human rMCP-1 to antigen stimulated cells increased the levels of luciferase activity in a dose-dependent manner beyond those observed in cells stimulated with antigen alone (Figure 4). The increments in luciferase activity produced by the addition of increasing amounts of human rMCP-1 to antigen stimulated cells were significant for the allele 1G (ANOVA F = 10.23, p = 0.005) and highly significant for the allele 2G (ANOVA F = 172.3, p<0.00001). Remarkably, cells transfected with vector containing the MMP-1 allele 2G expressed significantly higher levels of luciferase activity than cells transfected with vector containing the MMP-1 allele 1G when stimulated with M. tuberculosis antigen alone (t = 18.38, p = 0.0001), with antigen plus 500 pg/ml of human rMCP-1 (t = 13.32, p = 0.0001), or with antigen plus 2000 pg/ml human rMCP-1 (t = 11.99, p = 0.001) (Figure 4).
Figure 4. Luciferase activity in response to stimulation with M. tuberculosis-sonicate antigens and human rMCP-1 is greater in THP-1 cells transfected with the -1607 MMP-1 2G variant than in those transfected with the MMP-1 1G promoter variant.
The activities of promoters carrying the -1607 alleles 1G and 2G were compared using a dual-luciferase assay system (see Materials and Methods for details). Cells were stimulated overnight with or without M. tuberculosis-sonicate antigens and with the indicated amounts of human recombinant MCP-1. The data are presented as the ratio of Firefly Luciferase signal to Renilla Luciferase (control) signal. The error bars represent the standard deviation of means obtained from four independent experiments. Blue bars indicate cells transfected with plasmids containing the -1607 MMP-1 1G variant and red bars show data from cells transfected with plasmids containing the 2G variant. Addition of human rMCP-1 to cells stimulated with M. tuberculosis-sonicate antigens increased specific luciferase activity in a dose-dependent manner and to a level greater than that induced by M. tuberculosis antigens alone. The p-values from student t-tests are shown. In blue we show the p-values from comparisons of non-stimulated cells and cells stimulated with M. tuberculosis-sonicate antigens alone. Antigen = M. tuberculosis-sonicate 10 ng/ml.doi:10.1371/journal.pone.0008881.g004
This study of BCG-vaccinated Mexicans and Peruvians reveals a significant and consistent joint effect of the -2518 MCP-1 genotype GG and the -1607 MMP-1 genotype 2G/2G. The two-locus genotype MCP-1 GG MMP-1 2G/2G increased the likelihood of developing TB 3.59-fold in Mexicans and 3.9-fold in Peruvians. To build a conceptual frame upon which to understand the significance of our main finding we will first discuss our results from the univariate analysis of associations.
The current point estimates of independent associations of the susceptibility -2518 MCP-1 genotype GG are significantly lower than those previously observed in non BCG-vaccinated Mexicans (OR = 5.4) and Koreans (OR = 6.9) . The odd ratios (OR) obtained in the present study were 2.66 in Mexicans and 1.43 in Peruvians. In addition, we did not see a dose effect of the MCP-1 susceptibility allele G, and consequently did not see an association of MCP-1 genotype AG with disease in BCG-vaccinated Mexicans or Peruvians. Comparison between our current and previous data obtained in Mexican cases and controls suggests that BCG vaccination may have modified the effect of the -2518 MCP-1 allele G and genotype GG on disease progression although other factors may have contributed. For example, differences in the virulence of circulating bacilli between the periods when we recruited the previous and the current Mexican samples. However, it should be noted that both Mexican samples are from the same geographical area, although they were obtained at slightly different periods of time. Moreover, the demographics and clinical features of Mexican cases and controls recruited for both studies were similar, with the only exception being the BCG vaccination status. We could not confirm this finding in Peruvians since it was not feasible to recruit unvaccinated cases or controls in Peru. Although it is likely that we would find non BCG-vaccinated individuals in remote villages in Peru, these villages are difficult to access. However, it is worth mentioning that clinicians have observed that BCG vaccination modifies the rate of progression from infection to active TB . This finding highlights the need to pay special attention to the BCG vaccination status of cases and controls to avoid misclassification of the controls and bias when searching for genetic determinants of susceptibility to progression from infection to active tuberculosis.
It is interesting that the point estimates of MCP-1 association with disease in Mexicans were greater in strength that those observed in Peruvians in the present study. Differences in the genomic structure and antigenic composition of the BCG sub-strains used for vaccination in Peruvians and Mexicans may in part explain these results , . In Mexico, a BCG sub-strain derived from the BCG Copenhagen (Denmark) 1331 sub-strain has been in use since 1927 . In Peru, the original BCG strain or the BCG sub-strain Moscow was used from 1924 to 1990 –. Since our youngest cases and controls from Peru were born before 1991, it is very likely that most of them were immunized with the original BCG or the BCG Moscow sub-strain.
In animal models, diverse BCG sub-strains exhibit varied protective efficacy against experimental infection with M. tuberculosis and induced qualitatively different immune responses . Thus, certain sub-strains of BCG might induce a modest protective immunity that could slightly modify the effect of certain susceptibility genotypes. This might be the case with the BCG Mexico sub-strain. Alternatively, certain BCG sub-strains may induce an overwhelmingly intense pro-inflammatory recalled response to mycobacterial antigens that in turn may promote tissue damage independent of the effect of certain susceptibility genotypes. This might be the case for the original BCG or BCG Moscow used in Peru. It is worth mentioning the results of a clinical trial conducted in Mexican children to assess the expression profiles induced by different BCG sub-strains  that showed that the sub-strain Japan - which is structurally close to the Moscow sub-strain used in Peru - induces a more prominent recalled pro-inflammatory response than the BCG Copenhagen 1331 sub-strain used in Mexico . The intense inflammatory response induced by certain BCG sub-strains may over-ride the protective effect of certain genotypes.
Other prominent factors may have also contributed to differences in the point estimates of MCP-1 genotype GG association with disease in Mexicans and Peruvians, such as a more intense exposure to M. tuberculosis bacilli or the lower socioeconomic conditions in Peru. While Peruvians live in a moderate to high TB burdened country, the TB burden is lower in Mexico and Mexicans enjoy a relatively better socioeconomic status than Peruvians , , . Poor living conditions may partially over-ride the independent effect of some loci, including the effect of genotypes that confer resistance to disease progression. In addition, a more pathogenic M. tuberculosis strain might be circulating in Peru. For example, in a study of M. tuberculosis samples from various Latin-American countries, 5.9% of the isolates from 185 TB cases belonged to the Beijing family . Although there are no data on the incidence of this strain in Mexico, the same study reported that the incidence of this strain was markedly lower in samples from Argentinean TB cases (1% of 512 isolates), Brazilian TB cases (0.8% of 252 isolates), and Paraguayan TB cases (0.6% of 166 isolates), than in the Peruvian samples .
An association between the – 2518 MCP-1 allele G and resistance to TB was observed in Ghanaians . This association was attributed to the -362 MCP-1 allele C inherited in linkage disequilibrium with the - 2518 MCP-1 allele G . The same study found no association of MCP-1 with tuberculosis in a sample from Russia . In a Chinese population from Hong Kong, no associations of MCP-1 with tuberculosis was found , . Notably, association of the MCP-1 genotype GG with susceptibility to developing TB was found in a more ethnically homogeneous population of Han Chinese TB cases and controls . Differences in study design may explain these disparate results, particularly with respect to the criteria for selection of controls, including the adjustment for BCG vaccination and nutritional status, the stringency of the methodology used for the correct ascertainment of exposure and “latent” infection status, and the utilization of methods to correct for genetic admixture, or in the criteria to detect and correct genotyping or sampling errors reflected in departure from Hardy-Weinberg equilibrium , , , , –.
In this study we carefully established a cut-off at which we could increase the specificity of the PPD tests to detect latently infected controls among BCG-vaccinated individuals. To do this, we used the QuantiFERON-TB in-tube test, which discriminates between immunity induced by BCG vaccination and that induced by M. tuberculosis infection (see the Materials and Methods section) . This test is highly sensitive and specific in detecting M. tuberculosis infected individuals . Hence, all the controls in the present study had a PPD response >15 mm and were very likely “latently” infected with M. tuberculosis bacilli –. Since they were “latently” infected and might progress to disease, we ensured that they remained healthy for a period of at least two years after exposure , . This provided a control population to identify genetic determinants of the progression from M. tuberculosis infection to active TB disease . The criteria for the ascertainment of tuberculosis cases in the present study were very stringent and based on clinical features (signs and symptoms of TB), chest X-ray findings, and positive sputum-smear tests. However, these were not culture proven cases of M. tuberculosis infection and this might be seen as a source of bias since there is a remote possibility that they were infected with non-tuberculous mycobacterias (NTM) instead of M. tuberculosis. We think that this constitutes a minor limitation because the incidence of pulmonary TB caused by NTM is extremely low compare with that caused by M. tuberculosis , . Moreover, NTM infection produces pulmonary disease when host immunity is impaired and/or in patients suffering from other chronic lung diseases , , which was not the case in our tuberculosis patients since we excluded individuals who were affected by immuno-deficiencies, receiving treatments that compromise immunity, or suffering from other lung diseases. In addition, if our cases were infected with NTM we would have needed extended periods of multidrug therapy to attain sterilization of the sputum (sputum negativity) , , and all our TB cases attained sputum negativity in the first 4 months of treatment. Moreover, recidivating disease is very common among cases of NTM-caused pulmonary disease , , and, we excluded TB cases suffering from recidivating TB. Hence, it is very unlikely that any of our cases were infected with NTM rather than M. tuberculosis. None of the studies reporting discrepant findings adopted our stringent criteria for the selection of controls.
The frequency of the susceptibility -2518 MCP-1 allele G is markedly higher in Mexicans (0.45, Table 2) and Peruvians (0.64, Table 2) than in Caucasians (0.15 to 0.242) and non-Caucasian Africans (0.05 to 0. 175) . Consequently, ours discrepant results may also reflect ethnic-specific characteristics resulting from unique environments and selective evolutionary pressures , as demonstrated for the natural resistance-associated macrophage protein (NRAMP)-1 gene , . Notably, we have experimentally observed that the -362 MCP-1 polymorphism is not functional (Table S2) and do not think that the findings in Ghanaians are biologically relevant . The contribution of the -1607 MMP-1 genotype 2G/2G to the expression of active pulmonary TB might also be population-specific since the frequency of the allele 2G in Caucasians (0.433) and non-Caucasian Africans (0.375) is lower than in Mexicans (0.73) and Peruvians (0.71) (Table 2) .
Our results from the genetic analysis of joint effects support the notion that susceptibility to disease progression is a complex trait. Notably, 38% of Mexicans and 35% of Peruvians TB cases carry the two-locus susceptibility genotype MCP-1 GG MMP-1 2G/2G. However, 13% and 17% of latently infected healthy controls in Mexico and Peru also carry that susceptibility genotype and have not developed disease to date, indicating that other factors are involved in determining the fate of infection in these populations, including other genetic loci . Our analysis of lymph nodes from TB cases and the results of our in vitro experiments support the hypothesis that carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G are at increased risk for progression to active TB because they express a unique phenotype characterized by high levels of both MCP-1 and free MMP-1. Our results from the dual-luciferase experimental model further support to this notion. Our in vitro data indicate that MCP-1 potentiates M. tuberculosis-antigens induction of MMP-1 expression in cells of the monocytic lineage. Consequently, we observed numerous cells expressing large amounts of MCP-1 and MMP-1 adjacent to or surrounding necrotic areas in lymph nodes of cases with an ongoing inflammatory process. This support the notion that increased production of MMP-1 may destabilize the organization of new granulomas, thus contributing to the dissemination of infected cells and disease progression. Hence, we have brought together results from human population genetic and in vitro studies into a coherent hypothesis. Moreover, TB cases that are carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G may benefit from treatment to neutralize the deleterious effect of MMP-1.
In summary, our results indicate that BCG vaccination modifies the effect of the MCP-1 susceptibility allele G. More studies are required to understand the mechanisms underlying this interesting effect. Most important, we show that the – 2518 MCP-1 susceptibility genotype GG and the -1607 MMP-1 genotype 2G/2G jointly increase the likelihood of progression to active TB in a sub-group of BCG-vaccinated Mexicans and Peruvians. The strength and consistency of this association in both populations supports the notion that this joint effect is more important than the independent effects of either gene. Given that the two promoter polymorphisms are functional and both result in increased gene expression, we propose that increased levels of MCP-1, and consequently MMP-1, contribute to disease progression in BCG-vaccinated carriers of the two-locus genotype MCP-1 GG MMP-1 2G/2G. Increased MMP-1 availability and activity early in M. tuberculosis infection may destabilize granuloma formation, in particular the correct organization of new granulomas, promoting spread of the infection and progression to active TB. Animal models are needed to better understand how the two-locus genotype MCP-1 GG MMP-1 2G/2G may induce tissue damage in the lung during early stages of infection and to assess whether carriers of that two-locus genotype may benefit from treatment to neutralize the deleterious effect of excess production of MCP-1 and MMP-1.
Materials and Methods
All subjects provided informed consent, under protocols approved by the institutional review boards of the University of Texas Health Center at Tyler (the principal investigator spent 2 years as a faculty member in this institution), The Methodist Hospital (TMH) in Houston (Texas, USA), the Ministry of Health in Peru, and the Mexican National Institute of Medicine and Nutrition “Salvador Zubiran”.
We conducted case-control studies in Mexico and Peru. TB patients and controls were adults of Mestizo ethnicity aged 18 to 50 years, who were recruited in the cities of Mexico and in Lima and Callao in Peru, as part of the World Health Organization's DOTS community surveillance program for early detection of new TB cases . Samples from Mexico were from the same geographical area as the cases and controls of our previous study . From April 2006 through May 2007, 220 TB cases and 243 healthy controls were enrolled in Mexico, and from November 2005 through June 2009, 737 TB cases and 819 controls were enrolled in Peru. None of the potential participants declined to participate in the study. We could not obtain genotyping data for 27 TB cases from Mexico, and 36 TB cases and 23 controls from Peru. Thus, our sample size was 193 TB cases and 243 controls from Mexico, and 701 TB cases and 796 controls from Peru.
We applied most of the inclusion and exclusion criteria used in our previous study , with the only exception being BCG vaccination status. Thus, all cases and controls recruited in the present study were vaccinated with BCG at birth, had negative serologic tests for HIV infection, were of similar socioeconomic status within nationalities, and were unrelated to the third generation, as determined by a questionnaire. BCG vaccination status of cases and controls was confirmed by the presence of a scar in the left shoulder and vaccination records in the clinical charts. We assessed the nutritional status of our cases and controls using BMI as the main criterion , , . The BMI for each subject was determined, based on self-reported weight, before disease in the case of TB patients, and height, measured by a nurse . Individuals with a BMI <18.5 kg/m2 were considered to be malnourished ,  and were excluded from the study because this condition compromises immunity and predisposes to disease progression .
All TB cases exhibited symptoms (weight loss >10 kg, cough, fever, night sweats for more than 1 month, or cervical or axillary lymphadenopathy) and chest radiographic findings consistent with pulmonary TB, a positive sputum acid-fast smear, and a history of close contact to a TB patient. Only new TB cases were included in the study. All cases developed disease less than 1 year after recorded exposure to an index case.
The controls were healthy latently infected individuals with a history of close contact to a sputum-smear positive TB patient for a period of more than one month. A tuberculin skin test was administered to all controls, using the Mantoux method to deliver 5 tuberculin units of purified protein derivative (Tubersol; Sanofi Pasteur Inc., Swifwater, PA) intradermally –. The diameter of induration was measured 48 hours after inoculation. We recruited controls that had a tuberculin reaction >15 mm of induration: by using this cut-off, we increased the specificity of the test , . This strongly positive tuberculin reaction is much more likely to be caused by M. tuberculosis infection than by the BCG vaccination , .
To find the optimal PPD cut-off to distinguish infected from non-infected individuals, we performed a preliminary study in Peru involving 200 randomly selected and heavily exposed persons (household contacts of a sputum-smear positive TB case) who remained healthy for at least 2 years after exposure. Subjects were stratified into those with PPD >15 and those with PPD >10 mm but ≤15 mm. In each group, 100 randomly selected individuals were tested to determine whether they were infected with M. tuberculosis using the QuantiFERON (QFT)-TB Gold in-tube test (Cellestis, Inc., Valencia, CA). A positive QTF-TB test is an unequivocal sign of positive M. tuberculosis infection . Among those with a PPD >15 mm, 91% were QFT-TB positive compared with only 48% of those with a PPD >10 mm but ≤15 mm. Hence, we decided to use a PPD >15 mm of induration as the cut-off for selection of controls.
To further ascertain that these control individuals were resistant to disease progression, we ensured that a period of at least 2 years since exposure to a TB patient had passed at the time of recruitment. The first 2 years is a critical period during which most of M. tuberculosis infected individuals who are susceptible to disease progression develop active TB . All control individuals had negative sputum smears for acid-fast bacilli tests and normal chest radiographs. None of these healthy controls had received isoniazid, consistent with standard medical practice in Mexico and Peru. All potential controls whose tuberculin skin tests showed ≤15 mm of induration were excluded from the study.
Blood Samples and Single Nucleotide Polymorphisms (SNPs) Analysis
Genomic DNA was isolated from blood cell pellets using DNA extraction kits (Qiagen, Valencia, CA). We genotyped the -2518A>G SNP in MCP-1 (rs1024611), the -1607_1608insG variant in MMP-1 (rs1799750), and 42 genomic control SNPs listed in Table S1. The rs1024611 and rs1799750 SNPs were genotyped in duplicate using the tetra-arms technique . We did not find discrepancies in the typing results for MCP-1 and MMP-1. We will make primer sequences available and procedures available upon request. To confirm our results for the MCP-1 and MMP-1 gene analysis, we sequenced 10 randomly selected Mexican and Peruvian cases and controls. Previously published genotyping results of genomic controls for Mexicans were applied to our sample of Mexicans in the present study . The 42 SNPs used as genomic controls in Peruvians were genotyped using an Illumina platform. Given that Africans are the founders of all human races and that the -2518 MCP-1 allele A and the -1607 MMP-1 allele 1G are the most frequent in this population , , we used them as reference alleles.
Ex Vivo Analysis of Lymph-Nodes and In Vitro Studies
For immunohistochemistry (IHC), 35 axillary or cervical lymph nodes fixed in paraformaldehyde and embedded in paraffin were obtained from archived samples at the Hospital Maria Auxiliadora in Lima, Peru. All of these were from TB patients that met the inclusion/exclusion criteria described above. We then selected 16 TB cases that were carriers of the genotype combinations under study and had a lymph node inflammatory process of no more than 15 days of evolution. The lymph nodes analyzed were all from males with no significant differences in age, either within or between groups. We extracted genomic DNA from four core sections (80 µm thick) per sample, using the Recoverall Total Nucleic Acid Isolation kit (Applied Biosystems, Foster City, CA) and deparaffination and extraction protocols provided with the kit. The genomic DNA was genotyped as described above. For IHC we used heat-induced epitope retrieval in citrate buffer (Thermo/Fisher Scientific Inc., Waltham, MA), anti-human MMP-1/Collagenase-1 rabbit purified total polyclonal IgG antibody Ab-6 (Thermo/Fisher Scientific Inc) or anti-human MCP-1/CCL-2 mouse IgG2B monoclonal antibody from clone 23002 (R&D Systems, Minneapolis, MN), and the biotin-free HRP-polymer AEC-based detection system (UltraVision LP Detection System, Lab Vision Products, Fremont, CA). We immunostained 5-µm thick sections in duplicate for each case, using contrast Blue for counterstaining (KPL Inc., Gaithersburg, ME). Negative controls were sections incubated with irrelevant normal rabbit IgG (KPL Inc., Gaithersburg, ME) or irrelevant normal mouse IgG2B antibody (R&D Systems). We used 1X Automation buffer pH 7.5 (Biomeda Inc., Albuquerque, NM) in all wash steps. SuperMount permanent aqueous mounting media (BioGenex, San Ramon, CA) was used for mounting immunostained tissue sections. Immunostained tissue sections were first assessed at 100× or 200× total magnification and 0.8 numerical aperture of the objective lenses. The tissue sections were assessed by two trained observers who were blind to the clinical and genotyping information. Image acquisition was performed using a computerized analysis system comprising: a BX41 microscope with a U-TVIX-2 and a U-CMAD3 tube and adapter attached for on-screen viewing, a C3040 4.1 megapixel digital camera, and Magnafire-SP software (Olympus America Inc.). The number of cells expressing high levels of MCP-1 and MMP-1 was assessed at 400× magnification with the aid of a 10×10 mm ocular grid with 100 individual squares of 1 mm square (Microscope World, Carlsbad, CA). Five randomly selected fields per slide (one slide per individual with duplicate samples) were evaluated.
Culture Conditions, RT-PCR and Dual-Luciferase Assay
For the assessment of MMP-1 expression we used the 7500 Fast Real-Time PCR System. We used assay-on-demand primers for MMP-1 (ID Hs00233958), CD11b (Hs00167304), and the house keeping gene PDHB (Pyruvate dehydrogenase beta, ID Hs00168650) (Applied Biosystems). To calculate the relative quantity (RQ) of MMP-1, we used the 2 -ΔΔCT method implemented in the software . The data are presented as the fold change in gene expression normalized to the endogenous reference gene PDHB and relative to the untreated controls (RQ values). THP-1 cells (1×106 cells/ml) were stimulated with the indicated amounts of M. tuberculosis-lysate obtained after sonication (M. tuberculosis-sonicate) as described previously , or the indicated amounts of human rMCP-1 (R&D Systems), or both. Cells were cultured for 24 and 48 hours in complete RPMI (10% FCS) or serum-free media (Macrophage SFM medium, Invitrogen, Carlsbad, CA) as indicated in the figure legends. Cells were then harvested and total RNA was extracted using TRIzol (Invitrogen). Complementary DNA (cDNA) was obtained from 3 µg of total RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Approximately 100 ng cDNA was used to determine the expression levels of the housekeeping gene PDHB and MMP-1. Supernatants from the same cultures were harvested and kept at −20°C until analyzed. MMP-1 activity in the supernatants was tested using Fluorokine E, Human Active MMP-1 kits (R&D Systems) and the Multi-mode Microplate Reader Synergy 2 (BioTek Instruments Inc., Winooski, VT). This assay uses a monoclonal antibody to capture free MMP-1 and assesses its levels by measuring the amount of cleaved collagen peptide by monitoring fluorescence. Since the antibody captures only free MMP-1 and the assay is a functional assay, detection of MMP-1 that is complexed and neutralized by alpha-2-macroglobulin or TIMPs is excluded.
We compared the activity of promoters carrying the -1607 MMP-1 alleles 1G and 2G using the dual-luciferase assay system and Glomax 96-microplate luminometer (Promega, Madison, WI). The MMP-1 promoter region spanning positions −1350 to −1766 was cloned into the Firefly Luciferase vector pGL4.10[luc2] (Promega), expanded in DH5α E. coli cells (Invitrogen) and purified using an endotoxin free plasmid purification system (Qiagen). Alleles containing promoter fragments were obtained by PCR from genomic DNA obtained from -1607 MMP-1 homozygous 1G/1G and 2G/2G. THP-1 cells were co-transfected with pGL4.10[luc2] vectors containing the allele 1G or the allele 2G and control pGL4 Renilla Luciferase vector pGL4.73[hRluc/SV40] (Promega) in a 50:1 ratio using Effectene transfection reagents from Qiagen. Cells were cultured overnight in complete RPMI (10% FCS) and then washed three times in RPMI, re-suspended in complete RPMI, and stimulated overnight with or without M. tuberculosis-sonicate antigens and human rMCP-1. Protocols provided by the dual-luciferase assay system were followed for transfection of the cells, preparation of cell lysates, and measurements of the dual-luciferase signals.
Statistical analysis was performed using Intercooled STATA10 software (Stata Corporation, College Station, TX). Based on preliminary data for frequencies of alleles observed at the two loci in both populations, we calculated that our sample size would provide 90% power to detect an OR of 1.5 with a two-sided α of 0.025. Sample size calculations were performed as explained elsewhere . Hardy-Weinberg equilibrium was calculated using chi-square (x2) tests for n(n+1)/2 degrees of freedom, where n is the number of alleles in the polymorphism tested , . Expected genotype proportions were obtained using allele frequencies observed in the controls and the binomial equation .
Univariate analysis of categorical data.
Association of alleles with disease was analyzed using 2×2 contingency tables, and two-sided x2 tests . The x2 values were corrected for population stratification, primarily to control for differences in the level of admixture between cases and controls, by dividing the x2 values by an estimated value of λ . We used our previously calculated λ values for the Mexican sample . For Peruvians λ was calculated as the mean of x2 from comparison of the allele frequencies of 42 SNPs located across the genome in a randomly selected sample of 96 cases and 96 controls. SNPs selected as genomic controls were not in LD (Table S1) , . The x2-values obtained from the analysis of allele associations with disease were divided by λ and the p-values calculated based on the adjusted x2-values. The resulting p-values were further adjusted according to the number of comparisons, using the Bonferroni correction (*p-values). We used the method of Reich and Goldstein to control for genetic admixture because this method provides a precise coefficient to correct for stratification while testing for associations . This practical approach is unmatched by other proposed methods . For example, methods that rely on principal component analysis to deal with stratification might not be practical since they impose the stratified analysis of sub-groups exhibiting similar characteristics, increasing dimensionality and the sample size required .
We used multiple 2×2 tables, with genotypes arranged in an ordinal scale, and x2 Mantel-Haenszel statistics to test each of the three genotypes produced by the diallelic locus for association with disease, primarily to obtain valid estimates for association for each stratum, and to assess dose effects and homogeneity of the odd ratios (OR) across strata . To test for homogeneity of OR we used the Breslow and Day's test provided in the SATA10 output. Before routinely collapsing information across strata we considered it important to obtain estimates of association for each stratum using x2 Mantel-Haenszel statistics to make sure that they could be meaningfully pooled .
Multivariate analysis of categorical data.
To test main and joint (interaction) effects, we used multiple logistic regression analysis. Based on the results from univariate analysis of genotypes we collapsed MCP-1 and MMP-1 genotypes that were not found to be associated with disease and ran a hypothesis driven analysis of joint effects. We declared a significant interaction effect if the observed joint OR was greater than the expected OR, and the Z statistics and p values were significant , . We performed Pearson and Hosmer-Lemeshow goodness-of-fit x2 tests to determine whether the observed models differed from the predicted . The desirable outcome of non-significance (p>0.1) means that the predicted model does not differ from the observed . The ORs for the additive model were obtained using STATA10 output and the following formula: (ORGG + OR2G/2G) – 1; were ORGG is the adjusted OR for the MCP-1 genotype GG in the absence of MMP-1 genotype 2G/2G and OR2G/2G is the adjusted OR for the MMP-1 genotype 2G/2G in the absence of MCP-1 genotype GG .
Other statistic methods.
To analyze IHC results we used the Kappa statistic and a modified Rietveld and van Houd scale . According to this scale the interpretation of Kappa coefficients is as follows: <0 = Less than chance agreement, 0.01–0.2 = Slight agreement, 0.21–0.40 = Fair agreement, 0.41–0.60 = Moderate agreement, 0.61–0.8 = Substantial agreement, 0.81–0.99 = Almost perfect agreement . We performed one-way analysis of variance (ANOVA) to determine whether differences in the means number of positive cells between groups were significant . Twenty observations for each genotype group (4 patients per group ×5 randomly selected fields per patient) were tested. ANOVA tests were followed by Bonferroni least significant difference pair-wise comparisons . For the analysis of some in vitro experiments non-parametric Kruskal-Wallis or one-way ANOVA were used to compare mean values between groups, and Wilcoxon-Mann-Whitney U-test or student t-tests for independent variables were used to compare two groups mean values . Before running ANOVA we confirmed normal distribution of the data and homocedasticity using Shapiro-Wilk and Barlett's tests, respectively . If these assumptions were violated non-parametric tests were run.
Figure S1 shows the BCG-modifier effect in Mexicans comparing the strength of OR and CI in the pool and in samples used in the previous and present studies. Figure S2 shows IHC of lymph nodes from TB cases in which cells expressing MCP-1 and MMP-1 producing surround necrotic areas. Figure S3 shows typical examples of negative controls results from IHC analysis. Table S1 lists SNPs typed as genomic controls, SNP name, chromosome location and position, allele characteristics, call rates (CR) for each SNP in cases and controls, genotype frequencies, Hardy-Weinberg equilibrium assessment (HWE), minor allele frequencies (MAF), and chi-square values resulting from comparison of allele frequencies. Table S2 shows results of the assessment of the effect of the -362 G>C MCP-1 SNP (rs2857656) on MCP-1 promoter activity.
BCG vaccination modifies the effect of the -2518 MCP-1 susceptibility allele G and genotypes. Odd ratios (OR) and 95% confidence intervals (CI) are shown as measurements of the strength of associations between the -2518 MCP-1 A to G transition and progression from infection to active TB. Diamonds represent the OR and error bars represent the CI. An OR of 1 indicates no association. The further the value of the OR is from 1, in a negative or positive direction, the stronger the association. In the upper part of the figure, the pool of Mexican TB cases and the latently infected PPD+ controls from this and our previous study (1) is shown. The pool consists of 628 TB cases (61 homozygous AA, 245 heterozygous, and 322 homozygous GG) and 577 latently infected healthy PPD+ controls (128 homozygous AA, 288 heterozygous, and 161 homozygous GG). In the lower part of the figure the strata of non-BCG (left) vaccinated cases and controls from our previous study (1) and BCG (right) vaccinated cases and controls from the present study (Table 3) are shown. We used x2 Mantel-Haenszel statistics to test for genotype association with disease progression.
(0.01 MB PDF)
Cells producing MMP-1 and MCP-1 are also located adjacent to necrotic areas. Immunohistochemical (IHC) analysis of MMP-1 and MCP-1 expression in paraffin-embedded lymph-nodes from Peruvian TB cases carriers of the two- locus genotype MCP-1 GG MMP-1 2G/2G. Arrows indicate necrotic areas surrounded by cells expressing large amounts of MMP-1 (A) and MCP-1 (B) in dark red. There are many cells with pycnotic (condensed) nuclei in A and B. Images were acquired at 200× total magnification.
(0.21 MB PDF)
Negative controls. Immunohistochemical (IHC) analysis of MMP-1 and MCP-1 expression in paraffin-embedded lymph-nodes from Peruvian TB cases presented in Figure 1 and 2. In A, negative control incubated with normal rabbit IgG. In B, negative control incubated with normal mouse IgG2B. Images were acquired at 100× total magnification.
(0.45 MB PDF)
Genomic Controls. A These SNPs were genotyped as genomic controls. All SNPs selected are not in linkage disequilibrium and should segregate independently from each other. To control for errors in the estimation of chi-square (x2) values and HWE p-values all the SNPs selected have Minor Allele frequencies (MAF) >0.1 (high MAF). This ensured that we will have at least 10 out of 100 carriers of the rare allele and at least 1 homozygous for the rare allele in the dataset. We list SNP names, chromosome location and position, allele characteristics, call rates (CR) for each SNP in cases and controls, genotype frequencies (in black for controls and in red for cases), Hardy-Weinberg equilibrium assessment (HWE), minor allele frequencies (MAF), and chi-square (x2) values resulting from comparison of allele frequencies. B We included in the analysis those SNPs with call rates (CR) >95% to avoid errors in the estimation of x2 values. We did not need to re-cluster any of the loci tested to obtained call rate values presented in the table. We reanalyzed five loci using Tetra-Arms technique and the concordance rate was 100%. Thus, genotyping results obtained with the Illunina platform are highly reproducible. C Hardy-Weinberg equilibrium (HWE) test was done using the BeadStudio 3.0 Genotyping Module (GT) software from Illumina. The Null hypothesis for the HWE tested is: the distribution of genotypes is not in Hardy-Weinberg equilibrium. All loci tested in the control population were in HWE. D A chi-square (x2) value >3.8 is significant at p<0.05, and a x2 value >6.6 is significant at p<0.01 with 1 degree of freedom. A value of lambda was obtained as follows: λ = Σ x2/number of SNPs = 51.723/42 = 1.2315.
(0.10 MB DOC)
Luciferase activity in response to PMA and M. tuberculosis-sonicate antigens stimulation of THP-1 cells transfected with vectors containing the -362 MCP-1 alleles G or C. A The promoter region spanning positions -537 to position -156 of MCP-1 gene containing the -362 MCP-1 alleles G and C were amplified from - 362 MCP-1 homozygous GG or CC and cloned into the Firely Luciferase vector pGL4.10[luc2] (Promega Corporation, Madison, WI), expanded in DH5α E. coli cells (Invitrogen, Carlsbad, CA) and purified using an endotoxin free plasmid purification system (Qiagen, Valencia, CA). B THP-1 cells were co-transfected with pGL4.10[luc2] vectors containing the -362 MCP-1 allele G or the allele C and control pGL4 Renilla Luciferase vector pGL4.73[hRluc/SV40] (Promega Corporation, Madison, WI) in a 50:1 ratio using Effectene transfection reagents from Qiagen (Valencia, CA). Cells were cultured overnight in complete RPMI (10% FCS) and then washed three times in RPMI, re-suspended in complete RPMI, and stimulated overnight with or without PMA or M. tuberculosis-sonicate antigens. Protocols provided by the dual-luciferase assay system were followed for transfection of cells, preparation of cell lysates, and measurement of dual-luciferase signals. The data are presented as the ratio of Firefly Luciferase and Renilla Luciferase (control) signal. The standard deviation of means obtained from three independent experiments. C We run student t-tests using STATA10 to determine whether the alleles induce statistically different expression levels of luciferase in THP-1 cells. Vectors containing the -362 MCP-1 alleles are available upon request.
(0.03 MB DOC)
We are grateful to all patients and healthy donors for their kind cooperation. We thank Dr. Michael Lieberman and Dr. James Musser at TMHRI for their encouraging support of our work.
Conceived and designed the experiments: MG JARM POFV. Performed the experiments: MG JARM JD POFV. Analyzed the data: MG JARM JCGdlTP JD JG POFV. Contributed reagents/materials/analysis tools: JCGdlTP JG POFV. Wrote the paper: POFV. Helped in designing protocols for the recruitment of TB cases and controls in Peru: JCGdlTP, JG.
- 1. Flores-Villanueva PO, Ruiz-Morales JA, Song Ch-H, Flores LM, Jo E-K, et al. (2005) A functional promoter polymorphism in monocyte chemoattractant protein-1 is associated with increased susceptibility to pulmonary tuberculosis. J Exp Med 12: 1649–1658.
- 2. Jamienson SE, Miller EN, Black JF, Peacock CS, Cordell HJ, et al. (2004) Evidence for a cluster of genes on chromosome 17q11.2-q21 controlling susceptibility to tuberculosis and leprosy in Brazilians. Genes Immun 5: 46–57.
- 3. Monack DM, Muller A, Falkow S (2004) Persistent bacterial infections: the interface of the pathogen and host immune system. Nat Rev Microbiol 9: 747–765.
- 4. Coraux C, Roux J, Jolly T, Birembaut P (2008) Epithelial cell-extracellular matrix interactions and stem cells in airway epithelial regeneration. Proc Am Thorac Soc 5: 689–694.
- 5. Suki B, Stamenovic D, Lutchen KR, Ingenito EP (2005) Biomechanics of the lung parenchyma: critical roles of collagen and mechanical forces. J Appl Physiol 5: 1892–1899.
- 6. Kaarteenacho-Wiik R, Sademies O, Päähö P, Risteli J, Soini Y (2007) Extracellular matrix proteins and myofibroblasts in granulomas of sarcoidosis, atypical mycobacteriosis, and tuberculosis of the lung. Hum Pathol 38: 147–153.
- 7. Ulrichs T, Kaufmann HE (2006) New insights into the function of granulomas in human tuberculosis. J Pathol 208: 261–269.
- 8. O'Reilly PJ, Gaggar A, Blalock JE (2008) Interfering with extracellular matrix degradation to blunt inflammation. Curr Opin Pharmacol 8: 242–248.
- 9. Chang JC, Wysocki A, Tchou-Wong K-M, Moskowitz N, Zhang Y, et al. (1996) Effect of Mycobacterium tuberculosis and its components on macrophages and the release of matrix metalloproteinases. Thorax 51: 306–311.
- 10. Elkington PTG, Nuttfall RK, Boyle JJ, O'Kane CM, Horncastle DE, et al. (2005) Mycobacterium tuberculosis, but not vaccine BCG, specifically upregulates Matrix Metalloproteinase-1. Am J Crit Care Med 172: 1596–1604.
- 11. Reutter JL, Mitchell TI, Buttice G, Meyers J, Gusella JF, et al. (1998) A single nucleotide polymorphism in the matrix metalloproteinase promoter creates an Ets binding site and augments transcription. Cancer Res 58: 5321–5325.
- 12. Stein CM, Zalwango S, Malone LL, Won S, Mayanja-Kissa H, et al. (2008) Genome Scan of M. tuberculosis infection and disease in Ugandans. PLoS One 3: e4094.
- 13. Gueders MM, Foidart J–M, Noel A, Cataldo DD (2006) Matrix metalloproteinases (MMPs) and tissue inhibitors of MMPs in the respiratory tract: Potential implications in asthma and other lung diseases. Eur J Pharmacol 533: 133–144.
- 14. Rovin BH, Saxena R (1999) A novel polymorphism in the MCP-1 gene regulatory region that influences MCP-1 expression. Biochem Biophys Res Commun 259: 344–348.
- 15. World Health Organization (2008) Global tuberculosis control: surveillance, planning, financing: WHO report 2008.
- 16. Fitzgerald D, Haas DW (2005) Mycobacterium tuberculosis. In: Mandell GL, Bennett JE, Dolin R, editors. Mandell, Douglas, and Bennett's Principles and Practice of Infectious Diseases. Pennsylvania: Elsevier. pp. 2852–2886.
- 17. Caminero Luna JA (2004) A tuberculosis Guide for Specialist Physicians. Paris: International Union Against Tuberculosis and Lung Disease..
- 18. Molina-Gamboa J, Ponce-de-León-Rosales S, Rivera-Morales I, Romero C, Báez R, et al. (1994) Evaluation of sensitivity of RT-23 purified protein derivative for determining tuberculin reactivity in a group of health care workers. Clin Infect Dis 19: 784–786.
- 19. American Thoracic Society (2000) Diagnostic standards and classification of tuberculosis in adults and children. Am J Respir Crit Care Med 161: 1376–1397.
- 20. Wang L, Turner MO, Elwood RK, Shulzer M, FitzGerald JM (2002) A meta-analysis of the effect of Bacille Calmette Guérin vaccination on tuberculin skin test measurements. Thorax 57: 804–809.
- 21. Cerda-Flores RM, Villalobos-Torres MC, Barrera-Saldana HA, Cortes-Prieto LM, Rivas F, et al. (2002) Genetic admixture in three Mexican Mestizo populations based on D1S80 and HLA-DQA1 loci. Am J Hum Biol 14: 257–263.
- 22. Cervantes J (2002) Genetic differences between mestizo populations of Bolivia and Peru. Bull Inst Fr Etudes Andines 2: 213–222.
- 23. Reich DE, Goldstein DB (2001) Detecting association in a case-control study while correcting for population stratification. Genet Epidemiol 20: 4–16.
- 24. Szklo M, Nieto FJ (1999) Defining and Assessing Heterogeneity of Effects: Interaction. Epidemiology: Beyond the Basics. Massachusetts: Jones and Barlett Publishers. pp. 211–254.
- 25. Dupont WD (2002) Multiple Logistic Regression. Statistical Modeling for Biomedical Researchers. Cambridge: Cambridge University Press. pp. 207–262.
- 26. Davis JM, Ramakrishnan L (2009) The role of the granuloma in expansion and dissemination of early tuberculosis infection. Cell 136: 37–49.
- 27. Cawston TE, Mercer E (1986) Preferential binding of collagenase to alpha-2-macroglobulin in the presence of the tissue inhibitor of metalloproteinases. FEBS Lett 209: 9–12.
- 28. Hoheisel G, Sack U, Hui DS, Huse K, Chan KS, et al. (2001) Occurrence of matrix metalloproteinase and tissue inhibitors of metalloproteinases in tuberculosis pleuritis. Tuberculosis 81: 203–209.
- 29. Rand L, Green JA, Saraiva L, Friedland JS, Elkington PT (2009) Matrix metalloproteinase-1 is regulated in tuberculosis by a p38 MAPK-dependent, p-aminosalicylic acid-sensitive signaling cascade. J Immunol 182: 5865–5872.
- 30. Colotta F, Borre A, Wang JM, Tattanelli M, Maddalena F, et al. (1992) Expression of a monocyte chemotactic cytokine by human mononuclear phagocytes. J Immunol 148: 760–765.
- 31. Sherf BA, Navarro SL (1996) Dual-Luciferase reporter assay: an advanced co-reporter technology integrating Firefly and Renilla luciferase assays. Prom Notes Mag 57: 2–8.
- 32. Behr MA, Wilson MA, Gill WP, Salamon H, Schoolnik GK, et al. (1999) Comparative genomics of BCG vaccines by whole-genome DNA microarray. Science 284: 1520–1523.
- 33. Brosch R, Gordon SV, Garnier T, Eiglmeier K, Frigui W, et al. (2007) Genome plasticity of BCG and impact on vaccine efficacy. Proc Natl Acad Sci U S A 104: 5596–5601.
- 34. Oettinger T, Jørgensen M, Ladefoged A, Hasløv K, Andersen P (1999) Development of the Mycobacterium bovis BCG vaccine: review of the historical and biochemical evidence for a genealogical tree. Tuber Lung Dis 79: 243–250.
- 35. Behr MA, Small PM (1999) A historical and molecular phylogeny of BCG strains. Vaccine 17: 915–922.
- 36. Santiago EM, Lawson E, Gillenwater K, Kalangi S, Lescano AG, et al. (2003) A Prospective study of bacillus Calmette-Guérin scar formation and tuberculin skin test reactivity in infants in Lima, Peru. Pediatrics 112: e298–2302.
- 37. Castillo-Rodal AI, Castañón-Arreola M, Hernádez-Pando R, Calva JJ, Sada-Díaz E, et al. (2006) Mycobacterium bovis BCG substrains confer different levels of protection against Mycobacterium tuberculosis infection in a BALB/c model of progressive pulmonary tuberculosis. Infect Immun 74: 1718–1724.
- 38. Wu B, Huang C, Garcia L, Ponce de Leon A, Sifuentes Osorio J, et al. (2007) Unique gene expression profiles in infants vaccinated with different strains of Mycobacterium bovis Calmette-Guérin. Infect Immun 75: 3658–3664.
- 39. Lönnroth K, Jaramillo E, Williams BG, Dye C, Raviglione M (2009) Drivers of tuberculosis epidemics: The role of risk factors and social determinants. Soc Sci Med 12: 2240–2246.
- 40. Paes de Barros R, Ferreira FHG, Molinas Vega JR, Saavedra Chanduvi J.Measuring Inequality of Opportunities in Latin America and the Caribbean. Washington: The World Bank..
- 41. Ritacco V, Lopez B, Cafrune PI, Ferrazoli L, Suffys PN, et al. (2008) Mycobacterium tuberculosis strains of the Beijing genotype are rarely observed in tuberculosis patients in South America. Mem Inst Oswaldo Cruz 103: 489–492.
- 42. Thye T, Nejentsev S, Intemann CD, Browne EN, Chinbuah MA, et al. (2008) MCP-1 promoter variant -362 associated with protection from pulmonary tuberculosis in Ghana, West Africa. Hum Mol Genet 18: 381–388.
- 43. Chu SF, Tam CM, Wong HS, Kam KM, Lau YL, et al. (2007) Association between RANTES functional polymorphisms in Hong Kong Chinese. Genes Immun 8: 475–479.
- 44. Xu ZE, Xie YY, Chen JJ, Xing LL, Zhang AH, et al. (2009) Monocyte chemotactic protein-1 gene polymorphism and monocyte chemotactic protein expression in Chongqing Han children with tuberculosis. Zhonghua Er Ke Za Zhi 47: 200–203.
- 45. Takiff HE (2007) Host genetics and susceptibility. In: Palomino CP, Cardoso Leão S, Ritacco V, editors. Tuberculosis 2007, from basic science to patient cares. pp. 207–262. TuberculosisTextbook.com.
- 46. Thomas DC (2004) Statistical Methods in Genetic Epidemiology. Testing Candidate Gene Associations. New York: Oxford University Press. pp. 253–282.
- 47. Lynch M, Walsh B (1998) The Hardy-Weinberg Principle. Genetics and Analysis of Quantitative Traits. Massachusetts: Sinauer Associates Inc. pp. 54–56.
- 48. Diel R, Loddenkemper R, Meywald-Walter K, Gottschalk R, Nienhaus A (2009) Comparative performance of tuberculin skin test, QuantiFERON-TB-GOLD in tube assay, and T-spot.TB test in contact investigations for tuberculosis. Chest 135: 1010–1018.
- 49. Field SK, Cowie RL (2006) Lung disease due to the more common nontuberculous Mycobacteria. Chest 129: 1653–1672.
- 50. Johnson MM, Waller EA, Leventhal JP (2008) Nontuberculous mycobacterial pulmonary disease. Curr Opin Pulm Med 14: 203–210.
- 51. http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=1024611.
- 52. Delgado JC, Baena A, Thim S, Goldfeld AE (2002) Ethnic-specific genetic associations with pulmonary tuberculosis. J Infect Dis 186: 1463–1468.
- 53. Li HT, Zhang TT, Zhou YQ, Huang QH, Huang J (2006) SLC11A1 (formerly NRAMP1) gene polymorphisms and tuberculosis susceptibility: a meta-analysis. Int J Tuberc Lung Dis 10: 3–12.
- 54. http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=1799750.
- 55. WHO Working Group (1986) Use and interpretation of anthropometric indicators of nutritional status. Bull World Health Organ 64: 929–931.
- 56. Bailey BE, Ferro-Luzzi A (1995) Use of body mass index of adults in assessing individual and community nutritional status. Bull World Health Organ73: 673–680.
- 57. Schaible UE, Kaufman SHE (2007) Malnutrition and infection: Complex mechanisms and global impacts. Plos Medicine 4: 806–812.
- 58. Ye S, Dhillon S, Ke X, Collins AR, Day NM (2001) An efficient procedure for genotyping single nucleotide polymorphisms. Nucleic Acid Res 29: e88–8.
- 59. Livak KJ, Schmittgen TD (2001) Analysis of relative gene expression data using real-time quantitative PCT and the 2 -ΔΔCT method. Methods 25: 402–4008.
- 60. Hartl DL, Clark AG (1997) Population frequencies and genotypes. Principles of population genetics. Massachusetts: Sinauer Associates Inc. pp. 140–143.
- 61. Pritchard JK, Rosenberg NA (1999) Use of unlinked genetic markers to detect population stratification in association studies. Am J Hum Genet 65: 220–228.
- 62. Paschou P, Ziv E, Burchard EG, Choudhr S, Rodriguez-Cintron W, et al. (2007) PCA-correlated SNPs for structure identification in worldwide human populations. PLoS Genet 3: 1672–1686.
- 63. Viera AJ, Garrett JM (2005) Understanding interobserver agreement: The Kappa Statistic. Fam Med 5: 360–363.
- 64. Dupont WD (2002) Fixed effect analysis of variance. Statistical Modeling for Biomedical Researchers. Cambridge: Cambridge University Press. pp. 207–262.