Polymorphisms of cytokine genes and tuberculosis in two independent studies

Cytokine gene single nucleotide polymorphisms (SNPs) can influence cytokine levels, which may be associated with tuberculosis (TB) susceptibility. There is evidence that interleukin 1B (IL1B), tumor necrosis factor-alpha (TNF-alpha), and IL6 may be involved in the progression of TB. Using a self-validating case-control design, we selected eleven functional SNPs in IL1B, TNF and IL6 to detect their association with TB in Chinese Han and Tibetan populations. The associations between SNPs and TB were estimated by computing the odds ratios (ORs) and 95% confidence intervals (95% CI) using logistic regression analyses. We found that the IL1B rs16944 polymorphism was associated with decreased risk of TB in the two studies. The G allele at rs2069837 of IL6 was significantly more common in controls than in TB patients in the Han population. Moreover, TNF rs1799964 and rs1800630 were risk factors for susceptibility to TB, which were validated in the Chinese Tibetan population. In addition, TNF rs1799724 and rs1800629 were associated with TB, but only in the Tibetan population. In conclusion, SNPs of the IL1B and TNF gene were associated with TB susceptibility in Chinese Han and Tibetan populations. IL6 polymorphism may be considered as a protective factor for TB in the Chinese Han population, but not the Tibetan population.


Results
Demographics of the study population. In the initial cohort, we recruited 636 TB cases (50.9% males) and 608 controls (49.7% males) from the Chinese Han population. The mean (±SD) age was 36.8 (±15.7) years for cases and 37.1 (±15.7) years for controls. There were no significant differences in sex and age between groups ( Table 1). The percentage of smoking is lower in cases than in controls (P = 0.003) ( Table 1). In the validation cohort, 613 TB patients (53.3% males; mean (±SD) age was 34.5 (±14.5) years) and 603 healthy subjects (55.2% males; mean (±SD) age was 34.6 (±13.8) years) were enrolled from the Tibetan population. Cases and controls in this group were also well-matched by age and gender (Table 1). However, data of smoking status among the Tibetan subjects was not collected because smoking rates were very low in this population for cultural reasons.
We observed that the three gene polymorphisms were associated with TB (Tables 3 and 4). In IL1B, at the rs1143634 polymorphic site, the AA + GA genotypes compared with GG were protective factors against TB (P = 0.042). In IL6, the rs2069837 G allele (P = 0.046) and the G carriers (GG + GA) (P = 0.044) were associated with decreased risk of TB. In TNF, two SNPs were significantly associated with TB under the dominant model, i.e., rs1799964 (P = 0.017) and rs1800630 (P = 0.002). We also determined that CT heterozygotes and the C allele of rs1799964 were risk factors for susceptibility to TB. In addition, CA heterozygotes (P = 0.024) and the A allele (P = 0.020) of rs1800630 occurred at a higher frequency in the TB cases than in the control group. The LD between the SNPs is shown in Fig. 1. We found that the TNF frequency of haplotype CACGG was significantly different between TB and control groups (Table 5).
To validate the aforementioned results, we performed an independent study in a Chinese Tibetan population (Tables 3 and 4). Among the three SNPs analyzed in IL1B, the G allele (P = 0.023) and GG genotype (P = 0.028) of rs16944 were more prevalent in controls than in TB patients. No significant difference was identified in the distributions of IL6 genotypes in TB and control groups. In TNF, the rs1799964 C allele (P = 0.011) and CT genotype (P = 0.004) as well as the rs1800630 A allele (P = 0.005) and CA genotype (P = 0.002) increased risk of TB. SNPs in TNF were also associated with TB under other genetic models: rs1799964 (dominant: P = 0.004), rs1800630 (dominant: P = 0.017), rs1799724 (dominant: P < 0.001; recessive: P = 0.003) and rs1800629 (dominant: Examination of smoking-specific effects with TB. We also determined whether genetic determinants of TB were smoking-dependent in the Chinese Han population by performing an allele-smoking interaction analysis. As shown in Table 6, TNF rs1800630 (P = 0.026) was associated with TB in the non-smoking group, but not in the smoking group. In addition, 5 SNPs (IL1B, rs16944: P < 0.001; IL6, rs1800795: P < 0.001; rs2069837: P < 0.001; TNF, rs1800629: P = 0.001; rs361525: P < 0.001) within the three genes were associated with smoking TB.

Discussion
Understanding the genetic factors underlying TB has attracted increasing attention. In this study, we analyzed the association between polymorphisms in three cytokine genes and TB in two independent studies. We found the IL1B rs16944 was associated with TB susceptibility in the two studies. Significant associations were also found for both rs1799964 and rs1800630 in TNF with TB, which were validated in the Tibetan population. IL1B is a member of the IL1 cytokine family, which has an important role in the initiation and propagation of immune and inflammatory reactions. IL1B is a typical proinflammatory cytokine which can promote both local and systemic responses 41 . It was shown that alveolar macrophages from active TB subjects secreted high levels of IL1B 42 . Additionally, the production of this cytokine was associated with severity of human TB 43 . In addition, IL1B was reported to play an important role in the pathogenesis of TB in mice and human subjects 44,45 . rs1143634 and rs16944 are functional SNPs in IL1B. rs16944 has been associated with some diseases such as esophageal cancer 46 , inflammatory bowel disease 47 and schizophrenia. However, rs16944 was variably associated with TB among different populations. One report in a small cohort from China suggested no association between the SNP and TB disease 48 . A similar study conducted in Kazakhstan did not show a significant association between the risk of TB and rs16944 49 . Awomoyi et al. reported that rs16944 both heterozygosity and homozygosity were a protective factor for TB in Gambian 50 , in accordance with our results in the Chinese Han population. However, in our validation cohort, we found that rs16944 G allele and GG genotype instead of AA/GA were risk factors for TB. It was reported that rs16944 regulates IL1B levels and the homozygosity TT genotype had the highest expression of IL-1β 34,38 . Higher IL-1β levels could also lead to the progressive immunopathology during M. TB infection, which may result in TB progression in human 45 . Combined with the information above, we speculate that rs16944 SNP may influence immune reaction caused by M. TB infection by increasing IL1B levels and RNA expression. Two other SNPs (rs1143634 and rs1143623) were not associated with TB in the two cohorts. This result was the same as previous studies that were conducted in various populations 16,51-53 . IL6 is involved in the immune response, inflammation, coagulation, cell differentiation and tumorigenesis. High IL6 levels were related to inflammatory diseases such as rheumatoid arthritis 54 . In local inflammatory reactions, IL6 induces chemotaxis to mononuclear phagocytes 55 . When infected with M. TB, impaired bronchial epithelial cells activate mononuclear cells and lymphocytes to produce IL6, which then produces numerous antigen immune responses 56 . IL6 promotes IFNγ induction in early inflammatory responses and the important role of IL6 in TB susceptibility has been identified by IL6 knock-out mice 24 . A recent study in the Chinese Tibetan population showed an association between the rs2069837 G allele and TB 57   www.nature.com/scientificreports www.nature.com/scientificreports/ cohort were derived from the Aba Tibetan Autonomous Prefecture, and their subjects were recruited from the Tibet Autonomous Region. We also demonstrated that the other two SNPs (rs17147230 and rs1800795) were not associated with TB in the Han and Tibetan populations. More studies are warranted to validate these results in the future.
The response to M. TB infection is characterized by a strong immunocyte-mediated immune response, with the induction of TNF gene expression. TNF-alpha is important in the control of M. TB infection, and latent infection subjects with blocked TNF production can rapidly reactivate 58 . TNF-alpha not only plays a critical role in the immune response to TB but also participates in granuloma formation 59 . In our initial study, we demonstrated that rs1799964 (allele C and heterozygote CT) and rs1800630 (allele A and heterozygote CA) were protective factors against TB. The same results were also detected in our validation cohort, which strongly suggested that TNF was a causal gene for TB susceptibility. However, the r 2 between rs1799964 and rs1800630 was 0.87 and 0.81, respectively in the Han and Tibetan cohort, signifying that the association of TB with rs1800630 may be due to its LD with rs1799964 or vice versa. Studies have demonstrated that the rs1799964 and rs1800630 could increase the TNF RNA expression and secretion of the TNF-alpha protein 31,36,39 . In the present study, the same genotypes were associated with TB disease. Therefore, we speculate that these two SNPs may control the progression of TB disease by increasing the RNA expression and cytokine levels of TNF. In addition, overexpression of TNF-alpha has been implicated in the immunopathology of TB, such as tissue necrotising reactions which are important for the transmission of M. TB 60,61 . However, previous study proposed that other genes polymorphisms were also found to control the production of the cytokines included in our study 62 . Thus, a limitation of our study was that we did not perform functional validation of the significant SNPs.
A published study in the Chinese Uygur population observed that rs1799964 was associated with TB drug-resistance whereas it was not related to TB susceptibility 63 . They also found the rs1800630 AA + CA genotype was a risk factor for TB. The significant association of rs1799724 and rs1800629 with TB in our Tibetan cohort was the same as a previous study [63][64][65] . A published meta-analysis suggested that rs361525 was not associated with TB 66 and our study in both the Han and Tibetan populations also reports the same result. Since some of our significant findings were only detected in dominant/recessive models calculated by logistic regressions, correction for multiple comparisons such as Bonferroni may be appropriate to reduce a type I error in the data analysis. However, limiting the type I error may simultaneously increase the type II error 67 . In addition, the primary finding in this study was the significant association between TNF polymorphisms and TB disease in allele/ genotype models and the association did not change in dominant/recessive models. What's more, this association was validated in the Tibetan population, which suggested that this finding was not due to chance. Moreover, the rs1800630 showed the same effect on TB disease in previous studies as ours (Table S2). Therefore, Bonferroni correction may have limited utility in our study.  www.nature.com/scientificreports www.nature.com/scientificreports/ As mentioned above, all of the three cytokine genes were associated with TB in the initial study. However, only TNF polymorphisms rs1799964 and rs1800630 were fully validated under the allele/genotype model in the Tibetan group. Thus, we speculated that TNF has a more critical role in TB risk than other two cytokines. Meanwhile, the contradictory results in our study are also worth consideration. These discrepant results may be due to the differences in genetic and environmental factors between the Han and Tibetan populations. Firstly, the MAF and genetic background between the two cohorts were different, which may have caused the different genetic association results. Secondly, many lifestyle factors were different between the two populations. Most Tibetans live in the plateau area and their staple foods are barley, beef and butter. However, the effect of the environment on our association results is unknown.

Gene/SNPs
In the Tibetan cohort, the HWE test for rs1800630 was P = 0.024 in the control group. This deviation was not found in the case group. To detect if there were any genotyping errors, we repeated genotyping in 5% of the samples, and the results remained the same. The deviation from HWE could be explained by natural selection. In 1949, geneticist JBS Haldane proposed that infectious diseases have been the primary means of natural selection during the past 5000 years 68 .
Stratified analyses were performed on the 11 SNPs based on smoking status at TB onset in the Chinese Han population (subjects were divided into smoking and non-smoking). Interestingly, we found that the associations of IL1B and IL6 polymorphisms were more pronounced among smokers, which was similar to a previous study of TB 69 . Besides, SNPs within TNF revealed significant differences in allele frequencies between TB and controls in both smoking and non-smoking groups. Smoking is likely to be a risk factor for TB progression 70 Table 4. Association between genotype of cytokine genes and TB in the two populations. Abbreviation: SNP, single nucleotide polymorphism; CI, confidence interval; OR, odds ratio. # Adjusted by sex and age.
www.nature.com/scientificreports www.nature.com/scientificreports/ further underline the critical role of smoking at TB onset as an important factor to consider in future TB association studies to reduce TB phenotypic heterogeneity.
In our study, a total of three haplotypes in IL1B, two in IL6 and five in TNF were strongly associated with TB in the Tibetan population. In addition, one haplotype in TNF showed significant differences between cases and control in the Han population. However, the results of haplotypes in the two cohorts were inconsistent. Since   www.nature.com/scientificreports www.nature.com/scientificreports/ haplotype-based methods were based on the association between a polymorphism and the ancestral haplotype 71 , we speculate the discrepant results are likely attributable to differences in demographic history 72 .
Nevertheless, several limitations should be addressed in this study. First, nearly all the OR we observed were <2, which suggested that the power of the study may be inadequate to detect a role of the three gene polymorphisms on TB. However, our sample size was larger than most of the previous genetic association studies (Table S1). Second, functional validation of the meaningful SNPs was not performed. Thus the mechanism underlying the genetic association is unknown.

Conclusion
In summary, our study demonstrated that one SNP (rs16944) in IL1B and two SNPs (rs1799964 and rs1800630) in TNF were associated with susceptibility to or protection against TB development in our two studies. In addition, IL6 rs2069837 was a risk factor for TB in the Chinese population. Furthermore, TNF rs1799724 and rs1800629 were associated with TB in the Tibetan cohort. Our study could enhance the understanding of TB pathogenesis for clinicians and may improve the diagnosis of TB in the future. Further studies in other ethnic groups are needed to fully validate these results to disclose the potential function of these SNPs.   www.nature.com/scientificreports www.nature.com/scientificreports/ Methods study population. In the initial cohort, a total of 636 TB patients and 608 healthy controls were included from the West China Hospital of Sichuan University. An independent validation cohort including 613 TB patients and 603 healthy was enrolled from the Aba Tibetan Autonomous Prefecture. TB cases were diagnosed by two independent experienced respiratory physicians based on sputum smear tests, sputum culture, clinical symptom, radiological, histological pathologic examination and positive response to anti-TB therapy. Healthy controls were who had no TB-related symptoms, previous of history of TB and chest x-ray signs of disease. However, most of the participants lack the detection of Tuberculin Skin Test (TST) and Interferon Gamma Release Assay (IGRA).
Participants with HIV, hepatitis B and C, cancer and immune-related diseases were excluded from this study. All of the study subjects were unrelated. Specialized nurses drew 2-5 mL of blood from each study participant into tubes containing EDTA. Blood sample was then preserved in an −80 °C freezer for further DNA extraction and genotyping. Written informed consent was obtained from each participant. All experimental procedures were done in a BSL2 laboratory. This study was approved by the Ethics Committees of the West China Hospital of Sichuan University and the People's Hospital of the Aba Tibetan Autonomous Prefecture. All experimental procedures were conducted in accordance with the approved guidelines and regulations and the Declaration of Helsinki.
sNp selection and genotyping. Candidate SNPs for this study were selected according to the literature review of previous studies and in silico functional prediction from the NIH FuncPred website (https://snpinfo. niehs.nih.gov/snpinfo/snpfunc.html). SNPs were selected if they were reported to be associated with disease and/ or predicted to have effects on function. Genomic DNA was isolated from 250 μL of blood using a genomic DNA purification kit according to the manufacturer's instructions (Axygen Scientific Inc, Union City, CA, USA). SNPs were genotyped using the improved multiplex ligase detection reaction (iMLDR), with technical support from the Shanghai Genesky Biotechnology Company. A subgroup of 5% of the samples was repeated by iMLDR to confirm reproducibility. statistical analysis. Hardy-Weinberg equilibrium (HWE) proportion tests were used to evaluate the quality of the genotype data. Continuous variables were calculated by Student t-test and categorical variables were assessed by χ 2 -test in our study. The distribution of genotypes/alleles between cases and controls were compared by computing the odds ratio (OR) and 95% confidence intervals (95% CI) using logistic regression analyses. ORs were defined with respect to the case groups i.e. ORs >1 represent increased risk of TB. Linkage disequilibrium (LD) measure r 2 was calculated and haplotype blocks were estimated using the SHEsis online software platform (http://analysis.bio-x.cn). The power of our study design was assessed by using the Power and Sample Size Calculation software 73 . We calculated the power for each SNPs under an allelic model as described previously 74 . The frequency of each SNP in the two cohorts was sufficient to provide reasonable power (>80%) to detect an allele/genotype effect with OR 2.0 or above (Table 7). Data were considered statistically significant if P < 0.05. All analyses were performed utilizing the Statistical Package for the Social Sciences (SPSS, SPSS Inc., Chicago, IL, USA), version 19.0.

Data Availability
The data from this study are available online (https://figshare.com/s/33c072e600546dc1ba14).