Do genetic factors protect for early onset lung cancer? A case control study before the age of 50 years

We carried out a frequency matched case-control study.

Caucasian patients with newly diagnosed and histologically or cytologically confirmed primary lung cancer with age 50 years or younger were recruited in 21 major lung clinics across Germany. They completed an interviewer-administered questionnaire in which detailed information on personal history, history of lung diseases, family history of cancer and smoking habits were assessed. Blood samples were taken from all patients. A DNA bank was established. From July 2000 to April 2003 blood samples and case report forms were obtained from 247 young lung cancer patients. One of the patients was excluded because the parents were Vietnamese.

Cancer free control individuals are a random sample (frequency matched by 5-year age categories and sex to cases) drawn from the participants of a population based survey (KORA – Cooperative Health Research in the Region of Augsburg [50], survey S4). KORA, a continuation of the WHO MONICA study, provides a platform for research in epidemiology, health economics and genetics, where data and blood samples can made available. Since 1984/85 four representative surveys have been performed, including approximately 4000 – 5000 adults each.

After excluding one control individual because the preparation of the blood sample for genotyping failed, a total of 246 cases were subsequently compared with 223 control individuals. With respect to yet genotyped markers neither major population stratification between the whole KORA sample (southwest Germany) or two other cohorts from Northern Germany [51] nor deviations in the Minor Allele Frequencies (MAFs) to those of the HapMap Ceu Population could be detected.

The study was approved by the ethics committee of the Bayerische Landesärztekammer München and all necessary local ethic committees of the involved recruitment clinics. All participants signed an informed consent form.

The selection of DNA variants in candidate genes for this study was based on two criteria: published significant association together with plausible biological relevance of a polymorphism to lung cancer. We searched MEDLINE for reviews about the genetics of lung cancer published between 1995 and 2002. (Search term: "Lung Neoplasms/genetics" [Mesh] AND ("molecular" [TI] OR "gene" [TI] OR "genetic" [TI]) Limits: Publication Date from 1995/1/1 to 2003/1/1, Humans, Review, English, German). From 138 hits, we selected 35 by screening for promising titles or abstracts. All mentioned genes and DNA variants with significant association in these reviews and in a wide ranging selection of original study reports were listed. Two experts in the molecular biology of cancer rated these DNA variants for plausible biological relevance to lung cancer. The final selection furthermore also needed to meet limited financial constraints.

We investigated the following 11 SNPs: CYP1A1(Val462Ile) (Cytochrom-P450 Enzyme, rs1048943), EPHX1(His113Tyr) and EPHX1(Arg139His) (microsomal Expoide Hydrolase, rs1051740, rs2234922), GSTP1(A-193C) (Glutathione S-Transferase, rs947895), NAT2(Thr114Ile) and NAT2(Gln197Arg) (N-Acetyl-Transferase 2, rs1801280, rs1799930), GPX1(Pro200Leu) (Glutathione Peroxidase 1, rs1050450), p53(Arg72Pro) (tumor suppressor gene, rs1042522), XRCC1(Arg280His) and XRCC1(Arg399Gln) (X-ray Repair Cross-Complementing group1, rs25489, rs25487) and finally XPD(Asp312Asn) (Xeroderma Pigmentosum group D, rs1799793).

Blood samples (4 × 9 ml) were taken by clinicians and sent to the study center (GSF-National Research Centre for Environment and Health) within 24 hours. Immortalized cell lines were prepared and stored in liquid nitrogen. DNA was isolated from fresh or frozen blood using the DNA isolation kit of Gentra, Minneapolis, and stored at -80°C. Genotyping of SNPs was performed by matrix-assisted laser desorption/ionization time-of flight (MALDI-TOF MS, Sequenom) according to Weidinger et al. [50]. Standard genotyping quality control including 10% duplicate samples, checking for Hardy-Weinberg equilibrium as well as negative samples revealed no major errors. In none of the duplicate samples a deviating genotype could be determined.

When investigating potential modifications for lung cancer risk by marker genotypes, we considered sex, age and smoking habits as covariates. Patient's age was defined by age at first diagnosis, while for KORAS4 controls age at recruitment was recorded. Cumulative smoking exposure of former and current smokers was measured as packyears (PY). Cases and controls were grouped according to their smoking exposure level (SEL) into never and light smokers (≤1 PY), moderate (1-<20 PY) and heavy smokers (20 and more PY).

For cases we collected the smoking history in detail, as recommended [52]. For controls PY had to be approximated from the last amount of cigarette consumption per day and the duration of smoking. Preliminarily we classified all cases into the upper mentioned grouped by both concepts. We found these classifications to agree for 79% of cases. A similar agreement had been found by Bernaards et al. [53], when comparing retrospectively calculated PY with prospectively calculated PY. They concluded, that misclassification error in categorizing PY is smaller than quantitative error on continuous retrospectively PY calculation. Hence, we assume the use of PY groups to be at least as reliable as the collected retrospectively PY.

Using exact tests the distribution of histological subtypes was compared to a published German collection of 251 lung cancer patients with age of onset before the age of 46 years [20].

Hardy Weinberg equilibrium (HWE) was tested in controls using a likelihood ratio test [54].

As exact tests of genetic association we performed a BWS-Test (Baumgartner-Weiss-Schindler-Test [55]). These tests were also carried out in subsamples according to sex, age (grouped in age of onset ≤ 45 years and age of onset = 46–50 years, which almost splits the sample in two equally sized groups), smoking status (never, former and current smoker) as well as SEL (never and light, moderate, heavy smokers) and histological tumor subtype (small-cell, SCC and adenocarcinoma). For markers showing any significant association we performed two logistic regression models including age and sex as covariables. In model I smoking exposure level (SEL) was incorporated as the main effect while the genotype was nested within the SEL groups. Thus the genetic association was investigated nested within the SEL groups. The relative chance for lung cancer is estimated compared to genetic protected of the same smoking exposure. We also test for modification of the genetic effect by smoking by testing the contrast between the estimated parameters of model I for never or light smokers versus for moderate or heavy smokers.

In model II SEL-genotype interaction was directly included. Here all effects are given in comparison to wildtype-never and lightsmokers. The relative chance for lung cancer is estimated compared to genetic protected never or light smokers. Similar models were fitted with smoking status instead of SEL.

When appropriate, only subgroups of patients of a particular histological subtype were included. Motivated by single marker results on two genes we defined a genetic protection score (gPS) as the count of protective alleles, which are the T-allele of GPX1(Pro200Leu) and the C-allele of EPHX1(His113Tyr). Logistic regression including gPS was performed as described above.

We also carried out a sensitivity analysis for missing data. The level of significance was set to 5% for all tests. To take multiple comparisons into account, the p-values of BWS tests were interpreted at the familywise significance level of 5%/11 = 0.445%.

As expected, the leading histological subtypes were squamous-cell carcinoma (SCC) in men (30%) and adenomacarcinoma in women (37%). The gender specific distributions of histological subtypes within cases were similar to an other German study [20]. For more details see Table 1.

Nearly all patients (97% of men, 87% of women) were ever smokers (current or former) with high tobacco consumption (current smokers: mean 32.4 PY, former smokers: mean 28.8 PY). According to the cumulative smoking dose 2 of 3 patients (men: 73%, women: 63%) were classified as highly exposed to tobacco (≥20 pack years), while in controls this were 14% and 5%, respectively. Furthermore, the frequency of 55% female smokers among patients clearly exceeded the nationwide percentage given by the micro-census 1998 (31% at 15 to 50 years of age) [56].

The call rates of genotyping were on average 91% across markers.

Estimated allele frequencies are given in Table 2. Significant departures from HWE were not found in controls and only for p53 (Arg72Pro) in patients (p = 0.0384). Results of BWS-tests for a genetic association, an estimator of a main genetic effect within the total study population, age groups, male and female and current smokers are given in Table 3.

The estimated odds ratios for lung cancer were OR = 6.6 (95% CI: 3.4–12.8) for moderate and OR = 22.7 (95% CI: 11.9–43.3) for heavy smokers without taking any genetic marker information into account.

Among the 11 markers investigated, only the marker for the GPX1(Pro200Leu)gene showed a significant difference in the distribution of genotypes between all cases and controls (p_BWS-Test = 0.002). The variant T-allele was associated with a lower risk for lung cancer and showed a frequency of 22% (95% CI: 20%–24%) in cases, compared to 31% (95% CI: 29%–33%) in controls. Significant association could also be observed within men (p_BWS-Test = 0.011) and women (p_BWS-Test = 0.020), current (p_BWS-Test = 0.034), former smokers (p_BWS-Test = 0.001) and heavy smokers (p_BWS-Test = 0.003), in patients younger than 46 years of age (p_BWS-Test = 0.004), in patients with adenocarcinoma (p_BWS-Test = 0.024) and in patients with SCC (p_BWS-Test = 0.007).

Carriers of the T-allele showed a risk reduction of OR = 0.3 (95% CI:0.1–0.8, p = 0.012) within heavy smokers whereas for moderate smokers (OR = 0.6, 95% CI:0.2–1.6, p = 0.278) and for never and light smokers (OR = 0.9, 95% CI:0.3–3.0, p = 0.825) significance was not reached. Because of the small number of never or light smokers beyond cases no significance (p = 0.9012) was achieved when testing for modification or the genetic effect by SEL groups. Even if the T-allele seems to have some protective effect, the risk for lung cancer for moderate and heavy smokers in the presence of a T-allele is clearly increased compared to 'genetically unprotected' never and light smokers (moderate: OR = 10.1, heavy: OR = 22.1). For more details see Table 4.

The SNP within exon 3 of the EPHX1 gene showed significant associations within current smokers (p_BWS-Test = 0.015). Within current smokers the variant C-allele was associated with a lower risk for lung cancer and showed a frequency of 25% (95% CI: 21%–28%) for cases, compared to 36% (95% CI: 32%–40%) for controls.

We could find a risk decreasing genetic effect for C-carriers only for moderate smokers (OR = 0.2, 95% CI:0.1–0.7, p = 0.012). Please note the small number of TT-carriers within cases and controls, which lowers the evidence – not the significance – of this finding. No such significant effect was found for heavy smokers (OR = 0.8, 95% CI:0.3–1.9, p = 0.593), where the 95%-confidence interval for OR does not cover the point estimate of OR for moderate smokers. For more details see Table 4. Because of the small number of never or light smokers beyond cases no significance (p = 0.1898) was achieved when testing for modification or the genetic effect by SEL groups.

As a combined effect of these two polymorphisms, one might look at the count of protective alleles (T for GPX1(Pro200Leu), C for EPHX1(His113Tyr)) as a genetic prediction score (gPS). Such a score – modeled as continuous variables – yields an estimated OR per predisposing allele of 1.03 (95% CI: 0.5–2.0, p = 0.943) for never and light smokers, OR = 0.48 (95% CI: 0.2–1.1, p = 0.067) for moderate and OR = 0.55 (95% CI: 0.3–0.9, p = 0.020) for heavy smokers (over all SEL levels: p = 0.033). The difference between moderate and heavy smokers was not found significant (p = 0.777). Because of the small number of never or light smokers beyond cases no significance (p = 0.1770) was achieved when testing for modification or the effect of gPS by SEL groups.

In the presence of one protective allele only (gPS = 1) no differences in the decrease of lung cancer risk compared to gPS = 0 could be found between moderate (OR = 0.5, 95% CI:0.1–3.0), heavy (OR = 0.4, 95%: 0.1–1.3) and never and light smokers (OR = 0.4, 95% CI:0.1–2.0) (see Figure 1 and Table 5). In the presence of two or more protective alleles (gPS > 1) the risk for lung cancer further decreases for moderate and heavy smokers (OR = 0.2, 95% CI:0.1–0.6). For never and light smokers no further risk reduction was observed. However, the subsample of never and light smokers is too small to gain statistical evidence for such a conclusion (only 2 never and light smoking cases have a gPS > 2).

Bild vergrößern

Please note, even if both DNA-variants independently showed some protective effect for smoking exposed individuals, the risk for lung cancer in the double protected ever smokers (gPS ≥ 3) was significantly higher (OR = 4.8, 95% CI: 1.8–20, p = 0.028) compared to genetically unprotected never and light smokers (gPS = 0).

For the marker NAT2(Gln197Arg)we did not find an overall association (p_BWS = 0.252, crude OR_{per A-allele} = 1.0, 95% CI: 0.7–1.2). Within the investigated subgroups the BWS-test yielded p-values between 0.163 and 0.495. The only exception were patients with adenocarcinoma (p_BWS = 0.014).

For GSTP1(A-193C)we did not find an overall genetic association (p_BWS = 0.213, crude OR_{per G-allele} = 1.1, 95% CI: 0.8–1.4). In the investigated subgroups the BWS-test yielded p-values between 0.101 and 0.490, with two exceptions. For patients with SCC we achieved a p_BWS = 0.01. For heavy smokers significance was slightly missed (p_BWS-Test = 0.089).

For none of the other markers any significant genetic association was observed.

Antioxidant enzymes like glutathione peroxidase (GPX) are thought to be the primary cellular defence mechanism against reactive oxygen species. The lung epithelium is in particular endangered by exogenous NOx that causes epoxides, aldehydes and peroxides. They react to superoxidradical anion and hydrogen peroxide and in the presence of transition metal ions these continue to react to the aggressive OH radical. These reactive oxygen species (ROS) have the ability to cause massive injury to the cell. They are involved in inflammation processes, peroxidation of membranes which influences their permeability, binding on SH-groups of several enzymes which interferes its activity. Extracellular and intracellular antioxidative defence systems protect the cells from this damage. Extracellular defence is mainly done by small molecular particles like vitamins and small molecular proteins. Intracellular antioxidative defence mostly consists of the anti oxidative enzymes from the glutathion redox cycle (glutathion reductase and glutathion peroxidase). GPX is a tetramerical enzyme with four selenium-atoms bound as selenocystein in the active centre. It is important in the cellular defence against cytotoxic lipid peroxidation products [62]. The catalytic activity of GPX depends on the availability of reduced glutathione as coenzyme and on several endogenous and exogenous influences like genotype and nutrition. Smokers need more protection against these ROS. There is some evidence pointing to a lung cancer protecting role of vitamins C and E. However drinking alcohol is also found to increase the daily demand for vitamins [63, 64].

Ratnasinghe et al. [65] found evidence for association with lung cancer for a Pro→Leu (C→T) polymorphism at the amino acid position 200 of the GPX1-gene in a sample of Caucasian men aged 50–69 years which currently smoked at least five cigarettes per day. The odds ratio for heterozygotes (CT) was 1.8 (95% CI:1.2–2.8) and 2.3 (95% CI:1.3–3.8) for homozygotes (TT) compared to wildtype (CC) individuals. In a recently published investigation, Raaschou-Nielsen et al. [66] found in general a protective effect of the T-allele, which seems to be stronger in those being diagnosed in the age of 50–60 years (OR_{TT vs. CC} = 0.35, 95% CI: 0.2–0.8) than in older patients (OR_{TT vs. CC} = 0.8, 95% CI: 0.4–1.6). However, within those smoking more than 20 g tobacco a day the T-allele was found to carry a risk for lung cancer (OR_{CT vs. CC} = 1.95*, 95% CI: 1.4–2.6; OR_{TT vs. CC} = 1.7*, 95% CI: 1.1–2.7; *per 5 g tobacco a day). Hence, the T-allele increases the risk for lung cancer in smokers. Yang et al. [60] reported for never smokers a non-significant decreased risk for CC-carriers in younger (<50 years, OR = 0.6, 95% CI: 0.3–1.2) and an even stronger and significant decreased risk in older (>80 years, OR = 0.12, 95% CI: 0.02–0.7) patients. However, in older smokers the risk for lung cancer was increased for CC-carriers (OR = 3.3, 95% CI: 1.3–8.4), which is contradictory to the findings of Ratnasinghe et al. and Raaschou-Nielsen et al. The T-allele acts protective against lung cancer in smokers.

While the distributions of genotypes of all mentioned investigations are rather similar within controls, the proportion of the CC-wildtype is much smaller in the case-sample of Ratnasinghe et al. (29%) than in any other case-sample (45% to 61%). With the investigation of Yang et al. [60] we share the focus on early-onset patients. Estimates for OR in younger smokers are not reported by them, but the crude odds ratios of the younger study sample (age <50 years, calculated from table III in Yang et al. [60]) are OR_{CT vs. CC} = 0.7 and OR_{TT vs. CC} = 0.6 and within the confidence intervals of our estimations (Table 4). Yang et al. constructed hierarchical trees of risk modifying factors performing recursive partitioning (RPART). By this procedure, the sample size of effects in lower hierarchical steps becomes fairly small. They included in their final conclusion a non-significant effect of GPX1(Pro200Leu) within never-smokers, but didn't show any effect by GPX1(Pro200Leu) within smokers. In contrast to them we could see significant association by GPX1(Pro200Leu) for heavy SEL, as for current and former smokers, with ORs from 0.3 to 0.5. We missed significance for never and light and for moderate smokers possibly owing to low sample size. Nevertheless, the estimated effect was of the same size. Hence, T-carriers seem to have some genetic protection against lung cancer within smokers of age 50 years or less. Sensitivity analysis could demonstrate that even under worst conditions for missing genotypes the findings were qualitatively identical and genotyping errors appeared as missing at random. Finally, findings from our sample of early-onset cases are consistent with the previous reports, when restricting samples to age of 50 years or less [60]. However they are in conflict with findings from non-early-onset samples [63, 66].

Several factors have yet been identified to modify the activity of GPX. The consumption of fruits and vegetables as well as a supplementation with the trace elements selenium in populations with a low rate of daily intake affect the activity of GPX in human erythrocytes [67]. Serum concentrations of selenium and erythrocyte GPx activity were lower in smokers [68]. Additionally it was reported that alcohol induces lipid peroxidation which might lead to a decrease in GPX activity. Ravn-Haren and colleges could recognize a correlation between alcohol consumption and GPX activity to be modified by the GPX(Pro200Leu) genotype [63, 69]. Stronger association between smoking, alcohol intake and lung cancer was seen in carriers of the genotype TT of GPX1(Pro200Leu) than in carriers of genotype CC [11].

Thus the observed association between GPX(Pro200Leu) and the risk for lung cancer might be caused by the complex interplay between smoking, nutrition and GPX activity.

In combining both observed protective alleles of both genes we defined gPS, a genetic protection score. We could observe a positive association between the count of protective alleles and the reduction of tobacco smoke induced risk for lung cancer (OR = 0.20, 95% CI: 0.07–0.60).

Within our control group 53% are T-carriers for the GPX1(Pro200Leu) variant and 50% are C-carriers for the EPHX1 variant. Therefore, we might expect 3 out of 4 individuals of the population to have some genetic protection against lung cancer at younger age. However, the risk raising effect of smoking cigarettes is much stronger. Even under double protection by GPX1(Pro200Leu) and EPHX1(His113Tyr) the risk of current smokers is at least 4.5-times larger than in unprotected never and light smokers.

In conclusion, our study investigates the association of several candidate genes with lung cancer in the young. Only some of the results from this sample of early-onset lung cancer patients are consistent with previously reported age independent findings or suspicions. However, their role in the developing process is different.

We used a candidate gene approach based on the literature lung cancer as a whole. Thus, we can identify no other than previously reported susceptible genes to general age of onset within our young age sample. We also restricted considerations to the most promising marker per gene and did not consider haplotypes within candidate genes. So far, results presented here need to be understood as further investigation of controversial findings. We limited the chance of false positive results by applying a two-step strategy. First we performed two-group-comparisons with BWS-tests, followed by multiple logistic regression modeling for selected markers only.

For four markers the call rate of genotyping is shortly lower than 90%, which results from some suboptimal logistic in the early phase of the study and is not due to genotyping errors.