Dorota Kula M.D., Department of Nuclear Medicine and Endocrine Oncology, Maria Skłodowska-Curie Memorial Cancer Centre and Institute of Oncology, Gliwice Branch, Wybrzeże Armii Krajowej 15, 44–101 Gliwice, Poland, phone: +48 32 278 9720, fax: +48 32 278 9457, e-mail: dorota.kula@io.gliwice.pl

Introduction

Papillary thyroid cancer (PTC) is the most frequently diagnosed differentiated thyroid cancer (DTC) [1, 2]. Genetic predisposition to PTC is the highest of all cancers not displaying Mendelian inheritance [3–6]. This predisposition is expected to be multigenetic, with interactions among genes and environmental factors determining individual susceptibility [2, 4]; however, the genes responsible for PTC are poorly known. To date, genetic studies have identified some single nucleotide polymorphisms (SNPs) associated with DTCs or PTC, although only some of them have been validated [1, 7]. In 2009 Gudmundsson et al. identified SNPs associated with PTC: rs965513, located upstream of the FOXE1 gene on chromosome 9q22.33; and rs944289, located upstream of the NKX-2 gene on chromosome 14q13.3 [5]. At the same time, Landa et al. demonstrated an association between PTC and another SNP located in the 5-untranslated region of the FOXE1 – rs1867277 [4]. However, to date, only a few studies have been performed to assess the association of these SNPs with some clinical features of PTC [5, 8, 9]. To our knowledge, no study has analysed the mutual interactions between identified SNPs and their association with patient-related factors, like gender and age at PTC diagnosis.

In the present study, we first analysed the association of SNPs in the FOXE1 locus (rs965513, rs1867277, rs1443434) and near the NKX2-1 gene (rs944289) with PTC. Next, we evaluated the interactions between these SNPs and their associations with gender and age at PTC diagnosis.

Material and methods

Samples and subjects

A total of 2243 DNA samples derived from PTC patients and 1160 samples from controls were analysed. All subjects were Caucasians of Polish origin. There were 1925 (85.8%) women in the PTC group and 949 (81.8%) women in the control group. The median age at diagnosis of PTC was 48 years (first–third quartile: 37–57 years). The median age for the control group was 45 years (first–third quartile: 33–54 years). The number of samples analysed differed depending on the SNP and DNA availability (Table I). Only samples with all SNPs evaluated were included in the linkage disequilibrium (LD), haplotype frequency, and logistic regression analyses (Table II, III, IV, and VI, respectively).

PTC cases were recruited among thyroid cancer patients from all over Poland, who were referred for treatment to the Maria Skłodowska-Curie Memorial Cancer Centre and Institute of Oncology, Gliwice Branch. Data on thyroid cancer histopathology, patient gender, and age at diagnosis were extracted from medical records. Control subjects were recruited among consenting volunteers (from the population in general, from workers of the Institute of Oncology, and the persons accompanying the patients attending the Institute, not related to the patients) and were derived from the same populations as the cases. In all control subjects, thyroid cancer was excluded by anamnesis and thyroid ultrasound. The study was performed with approval from the Local Bioethics Committee, and written, informed consent was obtained from all participants.

DNA extraction and SNP analysis

DNA was isolated from the whole blood. Two main DNA extraction methods were used, i.e., the salting-out method and the anion exchange membrane column separation method (Genomic Maxi AX Blood; A&A Biotechnology, Gdynia, Poland), performed according to the manufacturer’s protocols. Initial screening of DNA purity was determined based on evaluation of the optical density ratio at 260/280 nm.

The SNP analysis with the allelic discrimination technique was performed on 384-well plates (Applied Biosystems, Foster City, CA, USA) using the 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). Reactions were performed in a final volume 5.1 μl per sample, containing 0.125 μl of 40X TaqMan® SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA), 2.5 μl of 2X TaqMan Genotyping Master Mix (Applied Biosystems, Foster City, CA, USA), and 2.5 μl of DNA (at concentrations of 3 ng/μl).

Statistical analysis

STATISTICA 6 software was used to evaluate the genotypes frequency. The odds ratios (ORs) were calculated with their 95% confidence intervals (CI). The deviation from Hardy-Weinberg equilibrium for each polymorphism was verified. Due to multiple testing, the Bonferroni correction was applied to p-values resulting from χ2 tests. Haplotype frequencies were estimated with the use of the EM algorithm. Logistic regression analysis was performed to model the effects of analysed SNPs and FOXE1 haplotypes, corrected for age at diagnosis and gender. For the logistic regression, the Akaike’s Information Criterion-corrected (AICc) value was used to evaluate models. A p-value less than 0.05 was considered to be statistically significant.

Results

Association of SNPs with risk of PTC

All analysed SPNs were in Hardy-Weinberg equilibrium both in the case and control groups. The genotype distributions, allele frequencies, and ORs for all SNPs are presented in Table 1. A significant association for all analysed SNPs was observed regarding all models of inheritance: genotypic, allelic, dominant, and recessive (data not shown). Only OR values for the dominant model are included in Table I: 1.72, 1.59, 1.53, 1.52 for rs965513, rs1867277, rs1443434, and rs944289, respectively; all of them significant after Bonferroni correction.

SNPs located in the FOXE1 locus (rs965513, rs1867277, rs1443434) were in LD (pairwise D’ ranged from 0.5371 to 0.9971 with p values < 10-6) (Table II). AAG FOXE1 haplotype was more frequent in PTC cases than in controls (40.5% vs. 32.5%), and its presence was significantly associated with PTC risk (OR= 1.42, p = 1.1 × 10-5) (Table III). By contrast, there was a lower frequency of GGT haplotype in PTC patients than in controls (43.3% versus 53.7%), and this haplotype was significantly associated with a reduced risk of PTC (OR = 0.66, p = 8 × 10-8) (Table IV).

Association of SNPs with age at diagnosis and gender

The PTC patients were divided into two subgroups with the reference to the age at diagnosis: ≤ 45 years, and > 45 years. (Table V). Differences were observed only for rs965513, but they were not significant after Bonferroni correction. We did not observe any difference in genotype distribution between women and men in the univariate analysis (data not shown).

Logistic regression

In the first step of the analysis, evaluation of individual effects of age and gender was performed - both of them were significantly associated with PTC risk (p = 0.0032 and 0.0375; AICc = 1907.5 and 1911.9, respectively). Next, models of inheritance were evaluated. For all analysed SNPs and FOXE1 locus haplotypes, the additive model emerged with the lowest AICc value. Finally, based on the AICc, rs1867277, rs944289, the GGT protective haplotype, age at diagnosis, and gender were included into the final regression analysis (Table VI), which showed an increased risk of PTC for rs944289 (OR = 2.65, p = 1.47 × 10-6) and a decreased risk for the GGT haplotype (OR = 0.54, p = 3 × 10-4). Moreover, analysis of interactions revealed an enhanced risk for rs944289 per year of age (OR = 1.014, p = 6 × 10-4) and an enhanced protective effect of the GGT haplotype in men (OR = 0.699, p = 0.012).

Discussion

We observed the association of the rs965513, rs1867277, rs1443434, and rs944289 SNPs with PTC, and two of them (rs1867277 and rs1443434) have not yet been analysed in a Polish population. The association of rs1867277 with PTC was first described by Landa et al., based on a case-control study [4], while rs1443434 b was first described by Gudmundsson et al. in a GWAS [5] – the results were similar to those obtained in our study. The association of rs965513 and rs944289 with PTC was analysed by Liyanarachchi et al. in cohorts from Poland and the United States [2], with results that are in accordance with our study.

The association of rs965513, rs1867277, rs944289, and rs1443434 with PTC has been confirmed in different populations in recent years [8–17]. Moreover, four meta-analyses concerning the roles of rs965513, rs1867277, and rs944289 in PTC have been published: the association of rs965513 has been confirmed among both Caucasians and Asians, rs1867277 SNP was analysed and confirmed in Caucasians [18–20], whereas rs944289 was associated with PTC in Asians, Europeans, and Americans [21].

The SNPs rs1867277 and rs965513 in the FOXE1 locus are in moderate pairwise LD in Europeans [12, 13]; however, these two SNPs were found to be more strongly correlated in a Belarusian population [17]. LD between rs965513 and rs1867277 is significantly weaker in populations of African or Asian ancestry [20], which is not surprising, since the LD block structure may be different in individuals of different ancestry [22]. In our study we observed LD between rs965513, rs1867277, and rs1443434; however, that between rs1867277 and rs1443434 was the strongest. Our analysis of rs965513/rs1867277/rs1443434 FOXE1 haplotypes showed an increased risk of PTC for the AAG haplotype and a reduced risk for the GGT haplotype. Similar results were obtained by Jones et al. [13], who showed that carriers of haplotypes consisting of both risk alleles of rs965513 and rs1867277 in a population from the United Kingdom had the highest risk of developing thyroid cancer [13]. A significant association also was observed for a haplotype in 9q22 consisting of a FOXE1 polymorphism: rs965513/rs10759944/rs1867277 in thyroid cancer families [14].

In our study, some differences in the genotype distributions of rs965513 were observed between younger (diagnosed at ≤ 45 years of age) and older (diagnosed at > 45 years of age) patients, although these differences were not significant after Bonferroni correction. We did not observe differences in genotype distribution between women and men in the univariate analysis (data not shown). To integrate the results concerning the genotype distribution of rs965513, rs1867277, rs1443434, and rs944289 and their association with patient-related factors (age at PTC diagnosis and gender), we performed a logistic regression analysis. Multivariate analysis showed an increased risk of PTC for rs944289. A decreased risk was observed for the GGT FOXE1 haplotype. Moreover, analysis of interactions revealed that risk for rs944289 was increased per year of age, and male gender enhanced the protective effect of the FOXE1 GGT haplotype.

Only a few studies have been performed to assess the association of the analysed SNPs with PTC clinical features [5, 8, 9, 16]. To our knowledge, only one previous study has evaluated the association between age at diagnosis and genetic predisposition, which showed a higher frequency of rs965513 A allele carriers among PTC patients diagnosed at a younger age [5]. Furthermore, only one study has analysed the differences between genders, which showed that rs965513 was significantly associated with PTC among both women and men [8].

The female-to-male ratio of thyroid cancer in European patients is 2.2 [23], and these sex differences in thyroid cancer occurrence have been attributed to different endogenous hormone profiles between the genders. The protective hormonal effects on the development of thyroid cancer in females have been reported, but the mechanism remains unclear [24]. However, opposite effects were recently observed in rats. Fortunato et al. showed that a redox imbalance elicited by oestrogen could be involved in the sex differences found in the prevalence of thyroid dysfunctions [25]. They observed sexual dimorphism in the thyrocyte redox balance, which was characterised by increased hydrogen peroxide production due to higher NOX4 and Poldip2 gene expression and weakened enzymatic antioxidant defence in the thyroids of adult female rats compared with male rats [25]. These findings could lead to the hypothesis that the same genetic background has a different effect in males and females because of different hormone profiles, which would explain the observed enhancement of the protective effect of the FOXE1 GGT haplotype in men in our study. Thus, the presence of the protective haplotype of the FOXE1 gene, a thyroid transcription factor gene involved in thyroid peroxidase (TPO) transcription [4, 5], might only reveal its protective effect in males and not females because of the interacting effects of oestrogens. Oestrogens have a general stimulatory effect on thyroid function, at least in rats, by increasing iodide uptake and TPO activity, as well as H2O2 production [25], which, as a reactive oxygen species, causes DNA damage, which could lead to the higher frequency of thyroid cancer development in women.

Conclusions

We have confirmed the association of PTC with SNPs in the FOXE1 locus on chromosome region 9q22.33 (rs965513, rs1867277, rs1443434) and one SNP (rs944289) near the NKX2-1 gene on chromosome 14q13.3 in a Polish population. The importance of FOXE1 haplotypes was demonstrated in our population. Logistic regression revealed interesting new interactions: the risk of PTC in carriers of the rs944289 SNP was increased per year of age, and a stronger protective effect of the GGT FOXE1 haplotype was observed in men. Thus, patient-related factors appear to modify the genetic predisposition to PTC, at least in our population. To our knowledge, this is the first study to evaluate these four SNPs and their associations with patient-related factors in the same patient set.