Susceptibility loci in lung cancer and COPD: association of IREB2 and FAM13A with pulmonary diseases

Genome-wide association studies have identified loci at 15q25 (IREB2) and 4q22 (FAM13A), associated with lung cancer (LC) and chronic obstructive pulmonary disease (COPD). The aim of our research was to determine the association of IREB2 and FAM13A SNPs with LC and severe/very severe COPD patients. We examined IREB2 variants (rs2568494, rs2656069, rs10851906, rs13180) and FAM13A (rs1903003, rs7671167, rs2869967) among 1.141 participants (468 LC, 149 COPD, 524 smoking controls). The frequency of the minor IREB2 rs2568494 AA genotype, was higher in LC vs controls (P = 0.0081, OR = 1.682). The FAM13A rs2869967 was associated with COPD (minor CC genotype: P = 0.0007, OR = 2.414). The rs1903003, rs7671167 FAM13A variants confer a protective effect on COPD (both P < 0.002, OR < 0.405). Haplotype-based tests identified an association of the IREB2 AAAT haplotype with LC (P = 0.0021, OR = 1.513) and FAM13A TTC with COPD (P = 0.0013, OR = 1.822). Cumulative genetic risk score analyses (CGRS), derived by adding risk alleles, revealed that the risk for COPD increased with the growing number of the FAM13A risk alleles. OR (95% CI) for carriers of ≥5 risk alleles reached 2.998 (1.8 to 4.97) compared to the controls. This study confirms that the IREB2 variants contribute to an increased risk of LC, whereas FAM13A predisposes to increased susceptibility to COPD.


Results
Genotype and allele distribution among patients and controls. The genotype/allele frequencies are summarized in Tables 1-3. The observed genotype frequencies of these 7 polymorphisms were all in agreement with the Hardy-Weinberg equilibrium (HWE) in the control subjects (for IREB2 HWE, the p-values were as follows: 0.674, 0.167, 0.149, 0.794; for FAM13A HWE p-values were: 0.986, 0.844, 0.766). We found that the IREB2 variants demonstrated an association with LC cases, while the FAM13A SNPs were associated with COPD.
For the IREB2 rs2568494 polymorphism, the AA genotype frequency, compared with the proportion of the GG genotype, was significantly higher in LC cases (P = 0.0081) than in the controls. Similarly, the presence of GA+ AA genotypes (P = 0.0129) and the A allele (P = 0.0043) were more common among LC patients as compared to controls in single polymorphism analyses (Table 1). No association of IREB2 rs2568494 genotypes with the results of the pulmonary function tests: FEV1 and FEV1/VC (FEV1forced expiratory volume in 1 second; VC-vital capacity; FEV1/VC-ratio) and smoking exposure in LC cases (Table 4) was noticed. We observed that frequencies of the G allele and GG genotype at rs2568494, G allele and GG genotype at rs10851906 and CC genotypes at rs13180 SNPs, were higher in controls compared to LC cases (Tables 1 and 3). The occurrence of the GG genotype of rs10851906 was differed greatly between LC cases and controls (P = 0.0027). When the IREB2 association was examined in the lung cancer cases without COPD (N = 369) compared to controls the results were comparable to those received for general lung cancer group (N = 468). In case of rs2568494 SNP the association was found but was not significant after Bonferroni correction (AA genotype: P = 0.0206, OR = 1.62 [1.08-2.44] and GA+ GG: P = 0.0457, OR = 1.32 [1-1.73 The interactions of selected genotypes of IREB2 (rs2568494, rs10851906, rs13180) with age that influenced the LC risk were observed (Fig. 1 [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15]. No significant associations were found in case of sex variable (P > 0.05) (Fig. 1).
Each of the three FAM13A SNPs showed a significant difference in genotype/allele frequency between COPD cases and controls (Tables 2, 3). The rs2869967 CC genotype was closely associated with COPD (P = 0.0007), also when the COPD with LC+ COPD patients was analyzed (combined COPD with LC+ COPD; P < 0.0001). Similarly, the frequency of the C allele was significantly higher in these two groups (P = 0.0009 and 0.0001, respectively) compared with the controls. We found no association of FAM13A rs2869967 genotypes with lung function measurements FEV1 and FEV1/FVC (FVC-forced vital capacity; FEV1/FVC -ratio) and smoking exposure in the COPD ( Table 4). The TC, CC and TC+ CC genotypes of rs7671167 SNP were more common in the control group as compared to COPD (P = 0.022, < 0.0001, 0.0015) and COPD with LC+ COPD groups (P = 0.036, 0.0004, 0.003) ( Table 2). These results were consistent with the difference in C allele frequency between these two groups of patients (P = 0.0002, 0.0003) (Table 3). Likewise, in case of rs1903003 SNP, the presence of the CT, CC and CT+ CC genotypes was significantly more frequent among the controls compared to COPD (P = 0.024, 0.0014, 0.0028) and COPD with LC+ COPD groups (P = 0.028, 0.0002, 0.0019) ( Table 2).
These observations were compatible with C allele distribution in these two analyzed groups of COPD patients analyzed (P = 0.0009, 0.0002) ( Table 3). The interactions of FAM13A (rs1903003, rs7671167, rs2869967) genotypes with age that influenced the COPD risk were detected (Fig. 1 Frequency of IREB2 and FAM13A haplotypes. To analyze the combined effect of these four investigated IREB2 SNPs and 3 SNPs of FAM13, we the generated the haplotypes on the basis of the genotyping data observed. The construction of haplotypes revealed the presence of seven IREB2 haplotypes and 6 FAM13A haplotypes in COPD/LC patients and controls. Haplotypes with a frequency lower than 1% in all groups were not considered for further analysis. Finally, we analyzed the eight most common haplotypes (four for each gene; Table 5). The most frequent haplotypes in patients and controls were: AAAT (36%), GAAT (27%), GGGC (21%), GAAC (14%) for IREB2 and CCT (43%), TTC (41%), TTT Cumulative genetic risk score. All study variants in both genes indicated a significant association with LC or COPD were included in the cumulative genetic risk score analysis (CGRS). The CGRS of two IREB2 SNPs (rs2568494, rs13180) among LC did not a reveal significant increase in OR value: for carriers of ≥ 3 risk alleles OR reached 1.5957 (1.1717 to 2.1730). In FAM13A gene the risk alleles were defined as T for rs13180, rs7671167 and A for rs2568494. The average (± SD) of cumulative risk scores among COPD patients (3.54 ± 1.97) and mixed COPD with LC+ COPD group (3.459 ± 1.96) were significantly higher than the controls (2.857 ± 1.83) with a P value < 0.0001. We also observed a significant difference in the average (± SD) of weighted cumulative risk scores comparing COPD and COPD with LC+ COPD to the control group (3.54 ± 1.98 and 3.46 ± 1.96 vs. 2.86 ± 1.83, respectively, P < 0.0001). Individuals were stratified according to the number of risk alleles into group carrying ≤ 2 (reference group), 3, 4 and ≥ 5 alleles. The risk of COPD increased with the growing number of the FAM13A risk alleles. OR (95% CI) for COPD carriers of ≥ 5 risk alleles reached 2.998 (1.8092 to 4.9678) for unweighted CGRS (Fig. 3). In COPD with LC+ COPD group, odds ratio for carriers of ≥ 5 risk alleles amounted to 2.542 (1.6705 to 3.8693) for unweighted CGRS (Fig. 3). The odds ratios calculated in unweighted and weighted CGRS analysis were similar, which is probably connected with the low number of risk alleles and small sample size. Statistical power analysis. The post-hoc analysis revealed that the statistical power of our study for analyses of the differences in genotype distribution between subgroups of patients and controls ranged between 80-97%. The statistical power of the study to detect the IREB2 alleles associated with LC (OR:  Table 3. The allele frequency distribution and logistic regression analysis of the IREB2 and FAM13A SNPs in COPD and LC. COPD-chronic obstructive pulmonary disease; LC-lung cancer; COPD with LC+ COPD: the associations with COPD were studied twice: using COPD group and COPD with 40% admixture of subjects with concomitant LC. 1 Reference category; OR (95% CI), odds ratio (95% confidence interval). 2 Result statistically significant (P < 0.0125, after Bonferroni correction). 0.7 to 1.4) for the three SNPs with a frequency of 0.2 to 0.38 (in the additive model) was 80-97%. The analysis of rs10851906 alleles in LC group, in the recessive model, revealed that the power of study is 80%. In case of FAM13A alleles associated with COPD and combined COPD with LC+ COPD group (OR: 0.5 to 1.7) with a frequency of 0.3 to 0.5 (in the additive model) the statistical power was 91-97%. This means that the study had sufficient power to detect an association of the IREB2 gene in LC and FAM13A gene in COPD and combined COPD with LC+ COPD group, in the case-control analysis.

Discussion
In this research we examined 7 SNPs among patients with LC, COPD and in healthy controls from Poland (n = 1141), in an attempt to evaluate the association of IREB2 and FAM13A loci with these diseases. In our study we observed that IREB2 variants appeared to be associated with LC, whereas the FAM13A was linked with COPD. The IREB2 variants have been previously linked to COPD as well as lung cancer. Three independent groups performing GWA studies have identified 15q25 locus, comprising the IREB2 gene, as associated with lung cancer 6,8,9 . In our study, three out of four selected SNPs showed a significant association with the LC, whereas none of them reached the level of significance among COPD cases ( Table 2). The genotype AA of IREB2 rs2568494 was significantly more frequent among LC group (P = 0.0081; OR = 1.682), which suggested that this variant might increase the risk of lung cancer. This association was independent of lung function among cases (FEV1 and FEV1/VC) and smoking intensity (Table 6). However, the effect of the abovementioned variant and age of lung cancer development was observed (Fig. 1). The prevalence of rs10851906 GG and rs13180 CC genotypes was significantly higher among the controls compared with LC cases, which suggests that minor alleles of these SNPs may have an apparently protective effect (OR = 0.379, 0.595, respectively). Our results are difficult to interpret because of the low number of reports on the protective effect of single IREB2 gene polymorphisms. Among 4 IREB2 haplotypes, only one was associated with an increased risk of LC. In the current study, the IREB2 locus was also examined in the lung cancer cases without COPD and the results just reflect the overall case-control relationship. However, it is worth noting that the lung cancer patients selected for this study, were only those considered for surgery. This may explain the low rate of COPD patients in this group (21%), whereas the prevalence of COPD in several lung cancer cases is reported to between 50-70% 18,29 . An alternative approach to estimate an overall genetic risk score, by the multivariate analysis that include genotypes and clinical variables which are independent predictors of lung cancer, were proposed by R. Young et al. In this report, the SNPs data were used to compile the risk model, which includes many risk genotypes from genes implicated in both COPD and lung cancer. These multifactorial approach may be helpful in identifying genes underlying the development of lung cancer 20 .
In contrast to COPD, the potential role of the IREB2 gene in human cancer is not well documented. Recently, studies on mouse models showed that overexpression of IREB2 promoted the growth of tumor xenografts in nude mice. In addition, the authors have noticed that the unique conserved insert of 73 amino acids of IREB2 mammalian protein, contributes to IREB2 pro-oncogenic potential and is essential for accelerated tumor growth 30 . However, a quantitative analysis of IREB2 mRNA levels in paired normal lung and lung adenocarcinoma tissue demonstrated the increased expression of CHRNA5, but no change   in IREB2 31 . In another study, using gene overexpression and/or gene knockdown and apoptosis analyses, the authors have failed to reveal that IREB2 mediates effects on lung cancer cell growth in vitro 32 . Further studies are needed to prove the link between IREB2 protein and cancer biology. DeMeo et al. (2009) have demonstrated that IREB2 may be a COPD susceptibility gene, identified through the integration of lung microarray expression data and replicated genetic association in many Caucasian cohorts 7,33,34 , including the Polish population 35 . Nevertheless, we were not able to confirm these findings in our group of Polish patients with COPD. IREB2 is located in the region near the CHRNA3/CHRNA5/CHRNAB4 genes cluster with a possible role of lung diseases development. The significant LD in this region a is source of difficulties in association analyses. M. Hardin et al. (2012) have demonstrated a significant association of IREB2 rs13180 with 315 severe COPD cases from Poland (P = 3.4 × 10 −3 ; OR = 0.69) 35 . However, after adjustment for the presence of the CHRNA3/5 SNP, this association was no longer significant, indicating that the positive results may be driven by LD with CHRNA3/5 polymorphism.
The functional consequences of IREB2 polymorphism are unclear. The protein product of the IREB2 gene is involved in maintaining human cellular iron metabolism. It has been demonstrated that smokers had higher concentrations of iron in the lungs 36 . Iron imbalance may lead to oxidative injury, whereas an excess of iron has an impact upon regional inflammation in the lungs, which may be relevant to the pathogenesis of COPD and lung cancer. It is supposed that IREB2 variants may further affect COPD and lung cancer when coupled with the increased levels of iron accumulated through exposure to cigarette smoke 7 .
We confirmed previous observations, which suggest that FAM13A variants may play a role in COPD susceptibility for all three SNPs selected for this study. The statistically significant values were obtained in almost each genotype with a minor allele in COPD patients and in the COPD group merged with  Table 6. Characteristics of patients with lung cancer (LC, n = 468), chronic obstructive pulmonary disease (COPD, n = 149) and controls (n = 524). NSCLC-non-small cell lung cancer; SCLC-small cell lung cancer; "-"-no information. 1 Severity of COPD was classified according to GOLD standard: IImoderate, III-severe, IV-very severe. 2 Information available for 454 participants.
patients suffering from concomitant LC and COPD (Table 3). It is not surprising, that the results for all SNPs were similar, due to the high level of linkage disequilibrium between these three SNP (CEU HapMap, NCBI B36 assembly) 18 . Among three FAM13A variants, only rs2869967 showed an association with an increased risk of COPD, whereas rs7671167 and rs1903003 were significantly more frequent among the controls (Table 3). A positive effect of these SNPs on age at COPD occurrence was observed (Fig. 1). The rs2869967 CC genotype was significantly more frequent among the COPD group (P = 0.0007; OR = 2.414) and in the COPD group combined with LC+ COPD (P < 0.0001; OR = 2.358). This variant has been mainly investigated in large Chinese populations and its CC genotype frequency seemed to be increased 37,38 . Wang et al. showed that rs7671167 and rs2869967 were associated with the FEV1/FVC ratio among all subjects, whereas rs2869967 revealed only a borderline association (P = 0.05) with COPD cases after adjusting for age, gender, body mass index (BMI), pack-years of smoking and current smoking status 38 . Further investigations including functional analyses are warranted to confirm the impact of CC rs2869967 variant in the modulation of the COPD risk.
The prevalence of CC genotypes of rs7671167 and rs1903003 was significantly elevated among the controls compared with COPD cases, which suggests the protective effect of these alleles (Table 3). Here, we confirm the findings of M. Cho et al. among non-Hispanic white, African-American and Norwegian cases, followed by R. Young et al. in Caucasians. In the first study, conducted among 2,940 COPD cases and 1,380 controls with normal lung function, the locus most highly associated with COPD included two SNPs: rs1903003 and rs7671167 in high linkage disequilibrium (r 2 = 0.85). Subsequently, they genotyped these two SNPs in 502 COPD cases and 504 controls and in two large family-based cohorts including 3,808 cases and confirmed the associations. Similarly, the study by R. Young et al. demonstrated that the C allele of rs7671167 is associated with a reduced risk of COPD 18 . In addition, they have indicated its protective effect in COPD combined with lung cancer patients with pre-existing COPD, which is in agreement with our observations in COPD combined with LC+ COPD. However, these authors also showed, that this variant was associated with reduced risk of lung cancer (OR = 0.75, P = 0.002). This relationship was absent in our study, because no significant difference was detected in the frequencies of genotypes/alleles of FAM13A among lung cancer patients. Another study 35 , carried out among 305 Polish patients with severe and very severe COPD, did not demonstrate a significant association between the rs7671167 SNP and COPD. However, they indicated that the C allele was associated with improved lung function, both in COPD cases and control population 35 . The association of FAM13A variants with lung function was replicated in several studies 11,12 . D. Hancock et al. meta-analyzed GWA studies, including 20.890 participants 11 . A phenotype, described by two important lung-function parameters: FEV1 and FEV1/FVC, is inheritable and provides the basis for evaluation of lung function. Hancock et al. identified eight loci associated with FEV1/FVC, including also variants in the FAM13A gene, namely rs2869967, which was analysed in our study. In current study, we did not identify an association between this SNP and FEV1 and FEV1/FVC parameters, as well as smoking intensity (Table 6). Among the 4 FAM13A haplotypes selected for this study, the prevalence of TTC was significantly higher in COPD patients than in controls.
We also observed a cumulative effect of three SNPs in the FAM13A locus. The risk of COPD escalated with an increasing number of risk alleles. Although the tendency for increasing OR for individuals with four risk alleles was moderate, probably due to the small number of carriers in this group, among carriers of ≥ 5 risk alleles OR rose significantly to 2.998. These results might suggest an additive effect of the variants being studied.
Whether any of these three SNPs is functional remains unknown, although given their location, an effect on splice variants would be possible. The most statistically significant result in our study was obtained for rs7671167 SNP, which lies in intron 4 of the FAM13A gene and was identified by Cho et al. It localizes downstream of the most important part of the FAM13A molecule: the Rho-GAP domain. Although little is known of the FAM13A function, the Rho GTPase activating role suggests both anti-inflammatory and tumor suppressor activity for FAM13A protein. FAM13A gene expression analyses in cell lines from several tissues (not comprising the lung) have shown a consistent increase in its expression in response to hypoxia 39 . Differences in the expression of this gene have also been noted in respiratory epithelial cells during differentiation into pulmonary type II alveolar cells in vitro model and among individuals with mild as compared to severe cystic fibrosis 40 . Recently, Z. Jin, J. Chung et al. showed that depletion of FAM13A in human lung cancer cells causes a reduction in Wnt signaling activity which provides evidence that Fam13a may contribute to human lung diseases 26 .
The effects of studied SNPs in impact in LC occurrence, COPD, and combined COPD with LC+ COPD were analysed using multiple testing of SNP-SNP interactions. In LC, IREB2 rs252568494 revealed the strongest association. In COPD, FAM13A rs7671167 was the most significantly associated, whereas in combined COPD with LC+ COPD, the FAM13A rs2869967 was the most relevant. After adjustment of the influence of the other SNPs in each gene, these associations were still significant (P < 0.05). Despite the loss of significance after Bonferroni correction, the direction of SNP effect remained the same. It is possible that the strong LD between certain SNP may cause the decrease of P-value level. These SNPs may be independent predictors of phenotype in pulmonary diseases and should be considered in future studies concerning the development of gene-based prognostic scores for LC and COPD. There are some limitations to the current study that need to be addressed. The number of COPD patients is limited, but the group was very accurately selected and the history of the disease and also exposition to noxious substances were collected in detail. 75% of cases had severe or very severe airflow limitation in the course of the disease (stage III and IV in GOLD classification), which represents more extreme phenotype of this disease. Although limited by our sample size, we will still able to demonstrate several significant associations in case of both genes in all groups of cases vs controls. With a Bonferroni correction for multiple testing, which is very conservative approach, the associations between IREB2 and lung cancer and FAM13A and COPD remain significant. Although, many data had provided evidence for a role of IREB2 gene in COPD, we were not able to demonstrated this associations. It is possible that association in IREB2 locus may be driven by LD with CHRNA3/5 SNPs only, but these variants were not examined in current study. It is also possible that IREB2 SNPs have not an impact in severe/very severe COPD occurrence. We did not demonstrated significant associations, however small effect was found in rs2568494 AA genotype in the combined COPD with LC+ COPD group. Although the cohort of COPD cases were pooled restrictively, it seems to be underpowered to detect an effect at the IREB2 locus. However, the statistical power for our analyses ranged between 80-97%, which may suggest that the study had sufficient power to detect an association.
Another limitations of the current study, is connected with smoking controls selection. Controls were matched by sex with both groups of patients (~70% of males in all groups). However, in the case of age distribution considerable deviation was observed between COPD and lung cancer cases and controls (median age 66, 64 and 55 years, respectively), which may lead to bias being introduced. The smoking status of controls was not matched exactly, because information about smoking exposure measures (like cigarettes/day or pack-years) was not available for 40% of controls. In this subgroup, we were not able to calculate pack years, because no information about all smoking exposure measures were pooled.
The control group included only cancer-free cases with normal pulmonary function, which allowed us to simultaneously compare COPD and lung cancer groups with the matched controls. This study design enabled us to determine the possible link between COPD and lung cancer patients related to IREB2 and FAM13A variants. To our knowledge, this is the first report in which IREB2 and FAM13A genes were analyzed in parallel among COPD and lung cancer cases, with cumulative genetic risk score analyses. This approach provided an assessment of the potential role of IREB2 and

Conclusion
We confirmed the association between IREB2 gene and lung cancer and between FAM13A gene and COPD in Polish patients. Further studies are required to elucidate the functional role of these variants, which may have an impact on lung disease susceptibility.

Methods
All experiments were performed in accordance with relevant guidelines and regulations approved by the Ethics Committee of the Poznan University of Medical Sciences (decision no. 802/10). All subjects gave their written informed consent to participate.

Materials.
All subjects were of Caucasian ancestry, recruited from Wielkopolska region in Poland. The detailed characteristics of patients and controls are displayed in Table 6. COPD with LC+COPD Patients. In 99 (21%) out of 468 LC patients, COPD was diagnosed as the concomitant disease, which allowed us to distinguish the subgroup of LC patients with coexisting COPD (LC+ COPD, N = 248). This subgroup was combined with COPD patients and analyzed separately as the COPD subjects with 40% admixture of subjects with concomitant LC (COPD with LC+ COPD).

Controls.
Control subjects (N = 524) demonstrated normal lung function with no evidence of airflow limitation (FEV 1 /FVC > 0.7). This group consisted of individuals for screening check-up in hospital or healthy blood donors with negative history of medical illnesses recruited at the Regional Blood Transfusion Centre. Efforts were undertaken to frequency-match cases by ethnicity, age and sex. Most of them were males (72%), with median age of 55 years (range 35-80 years). All controls were treated as smokers: current (29%), former (71%). Among former smokers were two groups: quitting smoking less than year (23%) and more than year (48%).
Methods. DNA samples from COPD patients, LC patients and controls were isolated from peripheral blood lymphocytes by Gentra Puregene Blood Kit (Qiagen, Hilden, Germany). DNA purity and concentration was confirmed using NanoDrop ND-1000 spectrophotometer.
We selected the SNPs previously associated with lung function and/or COPD in GWAs. We chose genomic regions based on review of the literature, and used the most significant reported SNPs which were analyzed in relatively large groups of cases. For IREB2 gene we selected the SNPs described by D. De Meo et al. 7 , for FAM13A gene we chose the SNPs which were identified by M. Cho et al. 10 . All polymorphisms selected for this study had minor allele frequencies > 0.4 in order to achieve sufficient statistical power. Altogether four SNPs in IREB2 (rs2568494, rs2656069, rs10851906, rs13180) and three variants of FAM13A (rs1903003, rs7671167, rs2869967) were analysed. Statistical analysis. The Hardy-Weinberg equilibrium (p 2 + 2pq + q 2 = 1, where p is the frequency of the variant allele and q = 1− p) was tested by chi-square test to compare the observed and expected genotype frequencies in healthy controls. The statistical differences in genotype and allele frequencies between the studied groups and controls were also evaluated by the chi-square test. For all polymorphisms, the ancestral or major allele was considered wild-type and the homozygous wild-type genotype was considered the referent genotype. Odds ratios (ORs) and 95% confidence intervals (CIs) for association between the IREB2 or FAM13A genotypes and alleles and studied diseases were estimated using logistic regression analysis. Each patients group (COPD and LC) was examined as an independent case series. An extended COPD cohort, including COPD cases and LC patients with coexisting COPD (COPD with LC+ COPD) was additionally considered in a separate case-control analysis. Statistical calculations were performed using GraphPad PRISM 5, module Statistical Analysis, Contingency Tables (GraphPad Software Inc.San Diego, CA).
Linkage disequilibrium (LD) measures were used to detect associations between the 4 investigated IREB2 polymorphisms and 3 SNPs of FAM13A gene. On the basis of the genotype data of all subjects, pair-wise LD was calculated using 2 standardized LD coefficients: r 2 and Levontin's D'. These associations were also tested for significance by means of the chi-square test. The expectation-maximization (EM) algorithm was applied to generate maximum likelihood estimates of haplotype frequencies. It was assumed that the Hardy-Weinberg equilibrium is applicable to the constructed haplotypes based on observed genotypes of IREB2 and FAM13A genes. Finally, the associations between specific haplotypes and the disease were studied among patients and controls (using chi-square test). To account for false-positive findings, multiple testing Bonferroni correction was applied. Because we tested 2 genes in 2 groups of patients, the result is significant if the corrected P-value is below the cutoff of < 0.0125 [P-value * n (number of genes in test) < 0.05] 42 . Calculations were performed using HAPLOVIEW, Linkage format [http://www.broad.mit.edu/mpg/haplo-view]. In case of two SNPs, which appeared significantly more frequent in patients' groups in relation to controls, frequency of genotypes, lung function measures and smoking exposure were compared by One-way ANOVA, Ordinary test. Association between rs2568494 IREB2 variant and lung function measures (FEV1, FEV1/VC) was analyzed in the lung cancer cases, whereas rs2869967 FAM13A variant and COPD phenotype (FEV1, FEV1/FVC) was compared among COPD patients. Smoking exposure was set as a quantitative feature (pack-years of smoking). A multiple linear regression analysis was used to investigate the correlation of variables, including age, sex, and genotype with disease by Logistic Regression 05.07.20 program (http://statpages.org/logistic.html). Effects of selected IREB2 and FAM13A genotypes on COPD/LC development, according to sex and age were estimated with the use of control group as a reference category. Logistic regression was used to multiple testing of SNP-SNP interactions by STATISTICA version 10.0 software. QUANTO software (version 1.2.4) was employed for sample size calculations.
Cumulative genetic risk score. SNPs in FAM13A showing significant association with COPD were included in the cumulative genetic risk score (CGRS) analysis. Genotypes for each SNP were coded as 0, 1 or 2 indicating the number of COPD risk alleles in the genotype. Unweighted and weighted CGRS (wCGRS) were calculated. In an unweighted approach the number of risk variants carried by an individual at each FAM13A SNP was counted to create cumulative genetic risk score (possible score for three SNPs range 0-6). In a weighted approach, OR estimated for risk allele in current study was used (T vs. C, OR 1.634; 1.565 and C vs. T, OR 1.548 of rs7671167, rs1903003 and rs2869967, respectively). A weighted risk score is the sum of the weighted COPD allele counts, weighted by natural logarithm of odds ratio for each risk allele ln(OR) and scaled by factor 3/w 1 + w 2 + w 3 where w i = ln(OR) for the i th SNP and i = 1 to 3 43 . Only samples without missing genotype were included in the analysis (COPD n = 149, Controls n = 524).
The effect of unweighted and weighted cumulative genetic risk score on COPD was calculated using logistic regression analysis. To compare the average (± SD) of CGRS and wCGRS between COPD patients and controls the unpaired t test was applied.