Introduction

Since selenium (Se) was shown to prevent liver necrosis in rats [34] and exudative diathesis in chicks [28], it was regarded as an essential trace element. In recent years, most attention was paid to Se in the context of cancer incidence reduction, which was demonstrated in animal studies (recently reviewed by Whanger) [37] and human clinical trials in relation to prostate, colon, lung and liver cancers [6, 39]. The mechanism by which Se suppresses tumour development has not as yet been elucidated. It is supposed that some of the anticancer properties of Se may be associated with specific proteins, which contain this element in the form of selenocysteine, the 21st naturally occurring amino acid [13]. Presently, 25 human selenoproteins are known, including 11 enzymes (four glutathione peroxidases, three iodothyronine deiodinases, three thioredoxin reductases and selenophosphate synthetase 2) and proteins: 15 kDa, H, I, K, M, N, S, O, P, R, S, T, V and W [19].

One of the selenoproteins of greater interest, due to its relevance to cancer risk, is 15 kDa selenoprotein (Sep15) [31]. This protein was first identified by Kacklösch et al. in the epithelial cells of rat prostate and by Gladyshev et al. in human Jurkat T-cells [10, 15]. Sep15 exhibits redox activity [21, 38] and belongs to the thiol-disulfide oxidoreductase family of selenoproteins [7]. Compared to other selenoproteins, Sep15 is unique as it is located in endoplasmic reticulum (ER) and forms a complex with another protein, UDP-glucose: glycoprotein glucosyltransferase (UGGT), which is involved in the quality control of protein folding [18, 22]. As Sep15 contains a thioredoxin-like structure and an unusual CXU motif (one residue (X) between cysteine (C) and selenocysteine (U)), which is similar to CXXC motif characteristic for thiol-disulfide oxidoreductases, it was suggested that this selenoprotein may take part in the process of correct disulfide bond formation, which has been confirmed recently by Ferguson et al. [7].

The cDNA library search shows that Sep15 is expressed in a wide range of tissues, with the highest levels in thyroid, parathyroid and prostate [10]. With the use of northern blot technique, high levels of Sep15 mRNA were also detected in prostate, liver, brain, kidney and testis [20]. Interestingly, the expression of this gene is dysregulated in certain human and animal tumours. Apostolou et al. [2] found that Sep15 was downregulated in approximately 60% of human malignant mesothelioma (MM) cell lines and tumours. Another study showed a lack of expression of 15 kDa selenoprotein in murine adenocarcinoma prostate and liver tumours, compared to the high levels of this protein exhibited by normal mouse liver and prostate [20]. These observations, together with the fact that Sep15 genetic locus (1p31) is commonly mutated or deleted in different human tumours [3, 20, 25] including lung cancer [17], imply that 15 kDa protein may be involved in the process of carcinogenesis. The link between Se and lung cancer has been shown in human clinical and epidemiological studies. Clark et al. [6], in their randomized controlled trial showed that Se supplementation decreased lung cancer incidence (46% reduction) and lung cancer mortality (53% reduction) in the Se-treated group, in comparison with the placebo group. Some epidemiological studies indicate also that high nutritional levels of Se are associated with a decreased incidence of lung cancer in humans [16]. Since Sep15 expression is regulated by dietary Se [27], it can be assumed that this protein is involved in the development of lung cancer. Notably, Sep15 compared to other selenoproteins, was preferentially synthesized in lung of Se-deficient rats after Se administration, which indicates the importance of this protein in this organ [4].

It is possible that Sep15 polymorphism is responsible for the link between Sep15 and cancer development. Two polymorphic sites located at 811 (C/T) and 1125 (G/A) within human Sep15 3′ untranslated region (3′-UTR) were found [10]. Further analyses revealed that allele C811 was exclusively associated with allele G1125, whereas allele T811 was exclusively associated with allele A1125, so there can be only two possible haplotypes existing within human Sep15 gene: C811/G1125 and T811/A1152 [14, 20]; 1125 (G/A) is located in a specific stem-loop structure within 3′-UTR that is called Sec insertion sequence (SECIS), an element responsible for Sec incorporation in selenoprotein during translation [20]; 811 (C/T) was also found at a SECIS-like structure, but this element is not functional [20]. The functional meaning is attributed to site 1125. It was observed that allele containing A instead of G is responsible for a lower responsiveness of the cultured cells to Se [2, 14, 20]. The authors suggest that dietary Se may be less beneficial in chemoprevention for individuals carrying 1125A allele in Sep15 gene [2, 14]. So far, an association between Sep15 1125G/A polymorphism in breast and head and neck cancers among African Americans has been found [14].

The aim of this study was to evaluate whether the genetic polymorphism of Sep15 in combination with Se status is associated with lung cancer risk.

Materials and methods

Study population

The study was a part of a case–control study conducted by Reszka et al. [33] and was approved by the local Biomedical Ethics Committee. All the subjects were Polish. A total of 325 cases of histologically confirmed lung cancer, aged 30–78 years (mean 58.8 ± 8.8), and 287 controls of similar age (33–74, mean 59.2 ± 8.9 years) were analyzed. All the cases were the patients treated in Lodz Hospitals in 1998–2002, and the controls were selected from among hospital patients, excluding those with other cancers or pulmonary diseases. After obtaining written informed consent to take part in the study, blood samples and personal data concerning the age, gender and the smoking status were collected. No data concerning the use of Se supplements were available. Non-smoking individuals were not included due to the too small sample size and the possible misclassification of the smoking status. The cases were categorized into two histological types: small cell lung cancer (SCC) and non-small cell lung cancer (NSCC). 232 Cases were classified as NSCC, accounting for the majority of cases (71.4%), 91 cases were SCC (28%) and 2 cases were classified generally as lung cancer (0.6%). The characteristics of the study population are displayed in Table 1.

Table 1 Study population characteristics
Full size table

Selenium status determination

The Se concentration was determined in the plasma of 325 lung cancer cases (100% of samples) and 276 controls (96% of samples), using graphite furnace atomic absorption spectrometry (GFAAS) [26]. The method was validated using reference material (lyophilized human reference serum samples of Seronorm™ from Nycomed Pharma AS, Oslo, Norway) and through participation in interlaboratory comparison trials [11].

Genotype analysis

DNA was isolated from whole blood using QIAamp DNA Blood Mini Kit (Qiagen) according to the manufacturer’s instructions. Genotype analysis was performed using polymerase chain reaction-restriction fragment length polymorhism assay (PCR-RFLP) and electrophoresis techniques. Oligonucleotide sequences for PCR primers were: forward 5′-CAG ACT TGC GGT TAA TTA TG-3′ and reverse 5′-GCC AAG TAT GTA TCT GAT CC-3′ [14]. The amplicons were digested with DraI (Deinococcus radiophilus, Fermentas) and FspBI (Flavobacterium species RFL1, Fermentas) and the digestion products were analyzed in 1.5% agarose gel.

Statistical analysis

Hardy–Weinberg equilibrium was assessed using chi square test. Odds ratios (ORs) adjusted for age, sex and smoking index were calculated as the estimates of lung cancer risk using unconditional logistic regression. Smoothing splines were used for continuous variables. For the smoking index, we used log transformation and for the age, the linear trend as a good approximation suggested by a preliminary analysis with splines regression. For plasma Se concentration, spline term was used. Nonlinearity in the trend for Se was assessed comparing the model with the linear trend to the model with the spline term with a fixed smoothness parameter. For testing significance, the likelihood ratio and confidence intervals (CI) based on likelihood profile were used. Mean plasma Se concentrations and smoking index in the cases and controls were compared using Student’s t test (assuming equal variances). Standard level of significance of 0.05 was adopted for statistical inference. All the statistical tests were two sided and data analysis was performed using R statistical environment. MGCV package, version 1.3–23, was used for logistic regression with splines.

Results

The control group was in Hardy–Weinberg equilibrium for Sep15 genotype. In all DNA samples, allele C811 was exclusively associated with G1125 and allele T811 was exclusively associated with allele A1125. Allele frequencies and odds ratios with 95% confidence intervals for Sep15 1125G/A are shown in Table 2. The comparison of allele frequency in the cases and controls showed no statistically significant association between Sep15 polymorphism and lung cancer risk. Compared to the individuals with the wild type variant (GG), the risk estimates for lung cancer were: 0.91 (95% CI: 0.64–1.32) for the heterozygous variant (GA) and 0.80 (95% CI: 0.39–1.65) for the homozygous variant (AA) (Table 2). No significant differences were also observed in SCC (OR = 1.01, 95% CI: 0.60–1.70 for variant GA, and OR = 0.50, 95% CI: 0.11–1.62 for variant AA; Table 3) and NSCC (OR = 0.91, 95% CI: 0.61–1.35 for variant GA, and OR = 0.92, 95% CI: 0.43–1.96 for variant AA; Table 4). The distribution of Sep15 alleles in lung cancer differed between the males and females with SCC. In this type of cancer, allele A was less frequent in men (39.3%) than in women (54.1%), whereas in the controls, the frequency was higher for males (44.7%) than for females (40.0%). Odds ratios for genotype AA were: OR = 0.22 (95% CI: 0.12–1.14) for men and OR = 1.78 (95% CI: 0.17–18.25) for women. The difference in the risk (OR) for the males and females was not statistically significant (p = 0.261; Table 3).

Table 2 Sep15 genotype frequency in lung cancer cases and cancer-free individuals
Full size table
Table 3 Sep15 genotype frequency in small cell lung cancer (SCC) cases and cancer-free individuals
Full size table
Table 4 Sep15 genotype frequency in non-small cell lung cancer (NSCC) cases and cancer-free individuals
Full size table

Plasma Se concentration was significantly lower among smoking individuals with lung cancer compared to smoking cancer-free individuals (49.4 ± 17.4 vs. 53.3 ± 14.0 ng/ml; p < 0.002). Although these results suggest a decrease in cancer risk for higher Se levels, we observed a decreased risk only for Se concentrations of up to 70 ng/ml and an increased risk (OR = 10.32; p = 0.025; Table 5) for levels above 90 ng/ml. Lower plasma Se levels (not statistically significant) were found in females compared to males, both in the cases (48.0 ± 14.1 vs. 49.9 ± 18.4 ng/ml; p = 0.407) and controls (51.8 ± 12.6 vs. 53.5 ± 14.3 ng/ml; p = 0.458). In the group with Sep15 1125AA genotype, Se concentration was lower in females compared to males only among the cases (38.9 ± 11.8 vs. 45.5 ± 13.2; p = 0.382), whereas in the controls, a reverse situation was found (64.4 ± 4.0 vs. 54.8 ± 17.2; p = 0.459).

Table 5 Lung cancer risk associated with plasma selenium concentration
Full size table

The results for cancer risk model with interaction for Sep15 polymorphism and plasma Se concentration are presented in Fig. 1. There was a general significant association between Se concentration and lung cancer risk for all three Sep15 genotypes. For the GG, GA and AA genotypes, the p values for the trends in risk were: p = 0.038, 0.035, and 0.030, respectively (Fig. 1). Statistically significant differences were also found in the trends between AA and GG genotypes (p = 0.049) and between AA and GA genotypes (p = 0.025; Fig. 1). For Se concentration close to the population average (53 ng/ml), lung cancer risk was similar for all individuals, but for lower Se concentrations, the risk in the group with AA genotype was increasing more sharply than in those with the other two genotypes (GG and GA). Above this concentration, the risk for GG and GA genotypes stopped decreasing and then for the concentrations higher than 80 ng/ml, it started to increase again. This trend is consistent with the results shown in Table 5 presenting association between Se concentration and lung cancer risk in the whole study population (regardless the Sep15 genotype). In the group with the AA genotype, lung cancer risk was still decreasing even at concentrations above 50 ng/ml. This trend is supported by our data up to the concentration of 80 ng/ml (the highest concentration in this group).

Fig. 1
figure 1

Joint effect of plasma selenium concentration and Sep15 1125G/A polymorphism for lung cancer development. Test for trends in Sep15 genotypes: p = 0.038 for GG, p = 0.035 for GA, p = 0.030 for AA. Test for trend differences: AA vs. GG: p = 0.049, AA vs. GA: p = 0.025

Full size image

We estimated an average risk for two Se concentration intervals, with the cut-off point of 50 ng/ml, as this concentration was close to the population average. Among individuals with a lower Se status (below 50 ng/ml), the risk was higher for those with the AA genotype compared to those with the GG genotype (OR = 1.42, 95% CI: 0.43–4.42; Table 6). On the contrary, among individuals with a higher Se status (above 50 ng/ml), the risk was lower for those with the AA genotype compared to those with the GG genotype (OR = 0.37, 95% CI: 0.12–1.17; Table 6).

Table 6 Lung cancer risk associated with Sep15 polymorphism in two groups of individuals: with lower and higher plasma selenium concentration
Full size table

The smoking index, expressed as cigarettes per day multiplied by the years of smoking, was significantly higher in lung cancer cases than in the controls (765 ± 331 vs. 552 ± 354, p < 0.0002). Unlike the Se status, smoking did not modify the risk of lung cancer due to Sep15 polymorphism (p = 0.210, results not shown).

Discussion

In our study, we investigated an association between lung cancer risk and Sep15 polymorphism combined with Se status in the smoking individuals of Polish origin.

Both the polymorphic sites in Sep15 (811 and 1125 within 3′-UTR) were analyzed in order to confirm the presence of two exclusive haplotypes (C811/G1125 and T811/A1125) in the Polish population. As expected, genotype analysis yielded only two haplotypes: CG or TA, and this exclusiveness is consistent with the results obtained by Hu et al. [14] and the analysis of the EST database. Also the distribution of alleles was similar to that found by Hu et al. in the Caucasian population [14].

In our further analysis, we are referring only to the polymorphic site at 1125, as this polymorphism is supposed to be functionally related to 15 kDa selenoprotein. It was found that the SECIS element with A at 1125 was more efficient in stimulating Sec insertion as compared to the SECIS element containing G at 1125. However, 1125A allele was also associated with a lower response to added Se in cultured cells [20] and these findings were confirmed by other studies [2, 14]. The authors suggest that Sep15 1125 G/A polymorphism may affect individual response to dietary Se and that Se supplementation may be less beneficial in chemoprevention for individuals carrying allele 1125A [2, 14]. This hypothesis is partially consistent with our findings. In the context of lung cancer risk, our study confirms the association between individual response to Se and Sep15 polymorphism. Depending on plasma Se concentration, the risk of lung cancer in the study population was significantly different in the individuals with AA genotype compared to the individuals with GG and GA genotypes. As regards individuals with AA genotype, an increasing plasma Se concentration was associated with a significantly decreased risk of lung cancer, whereas GG and GA genotypes were associated with decreased risk of lung cancer only among individuals with lower Se status. These results suggest that the individuals carrying AA genotype, compared to individuals with GG or GA genotypes, appear to benefit most from increasing dietary Se, which is not consistent with the findings of in vitro studies [2, 14].

Interestingly, in individuals carrying the GG or GA genotype, the increase in plasma Se concentration above 50 ng/ml was not associated with a decreased risk of lung cancer, and what is even more interesting, the highest plasma Se concentration (above 80 ng/ml) was associated with an increase in lung cancer risk. Some human epidemiological studies show an inverse correlation between nutritional levels of Se and the incidence of lung cancer [16, 35]. We also found that among smoking individuals, the Se status was lower in lung cancer cases than in controls (49.4 ± 17.4 vs. 53.3 ± 14.0 ng/ml, p < 0.002; Table 5) and these results were similar to those obtained in a case–control study by Reszka et al. [33], in which both smoking and non-smoking individuals were included (49.7 ± 17.1 ng/ml in the cases vs. 54.3 ± 14.3 ng/ml in the controls, p < 0.001). However, a large meta-analysis of epidemiological data showed that Se might have a protective function against lung cancer only in the populations with low Se levels [40]. Also the reanalysis of clinical trial conducted by Clark et al. [6] showed that Se supplementation decreased lung cancer incidence only among individuals with low baseline Se concentration [32]. As indicated by our findings, not only the low but also the high Se status may be associated with lung cancer risk, this depending on one’s genetic background. To our knowledge, this is the first epidemiological study demonstrating a significant association between elevated levels of Se and an increased risk of lung cancer. A similar association, although not significant, was observed in the prospective Nurses’ Health Study conducted by Garland et al. [8] who showed an increased smoking-adjusted risk for lung cancer (RR  = 4.33, 95% CI: 0.54–34.60) in the highest tertile of toenail Se concentration. The explanation for the increased risk of cancer associated with a higher Se intake, may lie in the prooxidant properties of this micronutrient. Se, when not incorporated into proteins, is present in the form of different Se compounds. In vitro studies indicate that inorganic Se compounds exert prooxidant activity, and are able to induce apoptosis in cancer cells, which is a beneficial effect. On the other hand, some of these prooxidant compounds, when present at high concentrations, are able to cause oxidative DNA damage in normal cells and some of them also have an ability to inactivate the DNA repair processes. According to these observations, it is suggested that Se, depending on the dose and metabolic activity, may be carcinogenic [23]. Some authors postulate that an increased intake of Se may promote tumour formation in a certain genetic background. Therefore, it is very important to distinguish between individuals who may benefit from Se supplementation and those who may not, before the supplementation begins [27]. Our results fully support this hypothesis.

Among individuals with lower plasma Se concentration, those with AA genotype seem to have a higher risk for lung cancer, compared to the individuals carrying GG genotype (OR = 1.48; Table 6). The difference was not statistically significant, which may have been due to the too small number of individuals with both the polymorphic variants 1125A in this group (14 cases vs. 5 controls). Hu et al. [14] found that 1125A allele was four times as frequent among the African Americans as among the Caucasian population. Interestingly, African Americans present a higher prevalence and mortality from lung cancer than do the Whites [1] and this observation was also made among cigarette smokers of different ethnicity [12]. Notably, serum Se concentration was shown to be lower in the Blacks than in the Whites [36]. Although our results were not statistically significant and concerned only smoking individuals, an assumption that people with low Se status who carry Sep15 1125AA genotype are more susceptible to developing lung cancer would contribute to explaining the racial disparity. It may also partially account for the gender differences which we found in our study in relation to SCC: the women were at a higher risk of developing this type of tumour. It is notable that the risk associated with the Sep15 1125AA genotype was increased for females (OR = 1.78 CI: 0.20–16.20; Table 3) whereas it was decreased for males (OR = 0.22 CI: 0.03–1.71; Table 3). The difference in the risk was not statistically significant; however, it could be due to the relatively small number of female cases. Interestingly, an association between Sep15 and cancer in females was found by Nasr et al. [25] who analyzed microsatellite markers for Sep15 and found allelic loss at this gene locus in breast cancer. We may assume that Sep15 polymorphism may be a stronger risk factor for women than for men. It is also possible that women who possess Sep15 1125AA genotype are more prone to developing lung cancer than the men with the same Sep15 genotype, because women have a lower Se status. However, the gender difference in Se concentration was not statistically significant in our study population, possibly due to, again, the small number of female cases. Therefore, studies involving a larger female population would be necessary to investigate the above.

Generally, the association between Sep15 and lung cancer merits further investigation. It was shown that the number of different selenoproteins is expressed in human lung cancer cells [5] but Sep15 seems to be particularly important. As indicated by in vivo studies, this protein is one of the selenoproteins preferentially synthesized in the lung under selenium deficiency [4]. Interestingly, the study conducted by Apostolou et al. [2] showed that Sep15 was downregulated in malignant mesothelioma cells that were isolated from the patients’ tumour [2]. Most of these cells displayed loss of heterozygosity (LOH) at 1p31, genetic locus for Sep15. However, LOH did not correlate with Sep15 downregulation. Some authors investigate Sep15 as a tumour suppressor gene referring to the fact that 1p31 is often deleted in human cancer development. LOH at 1p31 in lung cancer was detected by different authors but the frequency of regional loss varied from 53 to 28% and 27 to 14% [9, 24, 29, 30]. It should, however, be remembered that there are more genes mapped to 1p31 that have been implicated in human lung cancer as tumour suppressor genes [17].

As indicated by our study, the genetic polymorphism of Sep15 may modulate the influence of dietary Se on cancer development. Further studies are necessary in order to clarify this novel interaction between gene and nutrition in human lung cancer. The limitation of the present study is the lack of data concerning the use of Se supplements among the cases and controls. It should also be noted that our findings concern only the smoking individuals and smoking as a very strong risk factor for lung cancer may have influence on the modifying effect of Sep15 polymorphism. Therefore, a larger population should be investigated in order to analyze the differences between smoking and non-smoking individuals as well as to elucidate the possible gender differences in susceptibility to lung cancer in relation to Sep15 polymorphism. The controversial finding of an increased risk of lung cancer in association with the high Se status should be interpreted with great caution and needs further research.