Background
Breast cancer is the most commonly occurring cancer among women worldwide [
1]. In Poland it accounts for over 20% of all malignant tumors and is the second most frequent cause of cancer-related death [
2]. Although the majority of breast cancer cases are sporadic, a noticeable portion results from highly penetrating inherited mutation in susceptibility genes and family history remains the best predictor of their individual risk [
3]. Among known predisposition genes, deleterious mutations in
BRCA1 and
BRCA2 confer the strongest effect on disease susceptibility and are associated with a lifetime risk of breast cancer of up to 85% for such mutation carriers [
4,
5].
Even though the impact of high-risk gene mutations is noticeable, they account for only about 25% of the familial risk and less than 5% of total breast cancer predisposition, as their frequencies in general population are very low [
6]. It is suggested that remaining risk may result from a combination of multiple common variants, each conferring a small effect on breast cancer risk, with odds ratio (OR) usually between 1.2 and 1.5 [
7,
8]. According to the polygenic model, a large number of low-penetrance variants may have cumulative effect on both the overall risk of disease [
9] and an early disease onset [
10,
11].
A number of common single-nucleotide polymorphisms (SNPs) associated with slightly modified risk of different cancers have been identified through genome-wide association studies (GWAS). By far, at least 22 GWAS were conducted for breast cancer on different populations revealing over 36 susceptibility
loci[
12]. Fifteen of them were consistently confirmed in other GWAS or large replication studies and meta-analyses [
11,
13‐
24].
Conducting analyses on different populations increases the chance for generalization of conclusions and identification of causal variants [
25]. For its apparently high level of genetic homogeneity [
26,
27], the Polish population seems to be relevant for determining risk variants with relatively small, although significant, effect on cancer prevalence. In this study we focus on verification of selected associations with breast cancer risk among Polish women. Eleven SNPs were chosen for replication as commonly reported in different studies. Additional five variants were selected for evaluation based on data provided by Genetic Counseling of Cancer Center-Institute of Oncology in Warsaw as frequently observed in patients treated in Cancer Center. To our knowledge, by now only two of these SNPs were investigated for association with breast cancer susceptibility in Poland.
Results
The vast majority of 1424 women with breast cancer included in this study were non high-risk
BRCA1 or
BRCA2 mutation carriers: only 243 (17.1%) patients had one of
BRCA1 mutations indicated in Table
2 and none of them had any of 11 genotyped mutations in
BRCA2. The number of patients with family cancer history was similar to the number of sporadic tumor cases (785
vs. 639), as shown in Table
1. The median age at diagnosis was 44, ranging from 17 to 85.
The differences (
p-value ≤ 0.0314) were observed for seven out of 16 SNPs when allele frequencies between all cases and control subjects were assessed with the chi
2-test and four of them (rs2736098, rs13281615, rs1219648, rs2981582) remained significantly associated after multiple testing adjustment (
p-value
cor ≤ 0.0197) (Table
4 and Additional file
3: Table S3 for all results). The same four SNPs show significant association (
p-value
cor ≤ 0.0116) with breast cancer susceptibility in group of patients tested negative for the high-risk mutations. The strongest association was observed for rs1219648 and rs2981582 (
p-value
cor of 1.62E-05 and 1.46E-05, respectively). Both are located in intron 2 of
FGFR2 encoding the fibroblast growth factor receptor 2 protein. The minor allele of rs2736098 located in
TERT gene was associated (
p-value
cor of 9.30E-04) with a decreased risk of breast cancer. Fourth SNP (rs13281615) was located in 8q24 locus, known as multicancer susceptibility region [
38]. None of the 16 SNPs showed association after Bonferroni correction in the group of
BRCA1 mutation carriers.
Table 4
The significant SNP associations with breast cancer considering allelic and Cochran-Armitage trend tests
rs10941679 | 5p12 | | 0.24 | C vs N | 1.14 (1.01-1.28) | 2.97E-02 | 4.75E-01 | 2.71E-02 | 4.34E-01 |
| | | | C noMut vs N | 1.15 (1.02-1.30) | 2.29E-02 | 3.66E-01 | 2.04E-02 | 3.27E-01 |
| | | | S vs N | 1.17 (1.01-1.35) | 4.25E-02 | 6.80E-01 | 3.94E-02 | 6.31E-01 |
| | | | S noMut vs N | 1.20 (1.03-1.40) | 2.06E-02 | 3.29E-01 | 1.87E-02 | 2.99E-01 |
rs2736098 | 5p15.33 |
TERT
| 0.36 | C vs N | 0.77 (0.68-0.88) | 5.81E-05 |
9.30E-04
| 5.51E-05 |
8.82E-04
|
| | | | C noMut vs N | 0.78 (0.69-0.89) | 2.37E-04 |
3.78E-03
| 2.25E-04 |
3.60E-03
|
| | | | C with Mut vs N | 0.74 (0.57-0.94) | 1.70E-02 | 2.73E-01 | 1.55E-02 | 2.48E-01 |
| | | | F vs N | 0.81 (0.69-0.93) | 4.68E-03 | 7.49E-02 | 4.56E-03 | 7.30E-02 |
| | | | F noMut vs N | 0.80 (0.68-0.94) | 8.63E-03 | 1.38E-01 | 8.25E-03 | 1.32E-01 |
| | | | S vs N | 0.74 (0.63-0.87) | 2.61E-04 |
4.17E-03
| 2.38E-04 |
3.80E-03
|
| | | | S noMut vs N | 0.76 (0.64-0.90) | 1.46E-03 |
2.34E-02
| 1.36E-03 |
2.18E-02
|
rs13281615 | 8q24.21 | | 0.45 | C vs N | 1.19 (1.07-1.32) | 1.23E-03 |
1.97E-02
| 1.17E-03 |
1.88E-02
|
| | | | C noMut vs N | 1.21 (1.08-1.35) | 7.23E-04 |
1.16E-02
| 6.77E-04 |
1.08E-02
|
| | | | F vs N | 1.20 (1.06-1.35) | 5.19E-03 | 8.31E-02 | 5.24E-03 | 8.39E-02 |
| | | | F noMut vs N | 1.22 (1.06-1.39) | 4.60E-03 | 7.36E-02 | 4.53E-03 | 7.25E-02 |
| | | | S vs N | 1.18 (1.03-1.35) | 1.61E-02 | 2.57E-01 | 1.42E-02 | 2.27E-01 |
| | | | S noMut vs N | 1.20 (1.04-1.38) | 1.09E-02 | 1.74E-01 | 9.61E-03 | 1.54E-01 |
rs1219648 | 10q26 |
FGFR2
| 0.41 | C vs N | 1.30 (1.17-1.45) | 1.01E-06 |
1.62E-05
| 1.13E-06 |
1.81E-05
|
| | | | C noMut vs N | 1.36 (1.22-1.52) | 7.20E-08 |
1.15E-06
| 7.95E-08 |
1.27E-06
|
| | | | F vs N | 1.26 (1.11-1.43) | 4.21E-04 |
6.73E-03
| 3.76E-04 |
6.01E-03
|
| | | | F noMut vs N | 1.33 (1.16-1.53) | 4.02E-05 |
6.43E-04
| 3.24E-05 |
5.18E-04
|
| | | | S vs N | 1.36 (1.19-1.56) | 7.32E-06 |
1.17E-04
| 8.53E-06 |
1.36E-04
|
| | | | S noMut vs N | 1.39 (1.20-1.59) | 5.58E-06 |
8.92E-05
| 6.76E-06 |
1.08E-04
|
rs2981582 | 10q26 |
FGFR2
| 0.41 | C vs N | 1.31 (1.17-1.45) | 9.10E-07 |
1.46E-05
| 1.17E-06 |
1.88E-05
|
| | | | C noMut vs N | 1.35 (1.21-1.51) | 1.20E-07 |
1.91E-06
| 1.54E-07 |
2.46E-06
|
| | | | F vs N | 1.26 (1.11-1.43) | 4.05E-04 |
6.48E-03
| 4.09E-04 |
6.54E-03
|
| | | | F noMut vs N | 1.32 (1.15-1.51) | 6.49E-05 |
1.04E-03
| 6.11E-05 |
9.77E-04
|
| | | | S vs N | 1.37 (1.19-1.56) | 5.70E-06 |
9.12E-05
| 7.67E-06 |
1.23E-04
|
| | | | S noMut vs N | 1.38 (1.20-1.59) | 5.69E-06 |
9.11E-05
| 7.97E-06 |
1.27E-04
|
rs3817198 | 11p15.5 |
LSP1
| 0.34 | F vs N | 1.16 (1.02-1.32) | 2.45E-02 | 3.92E-01 | 2.36E-02 | 3.78E-01 |
| | | | F noMut vs N | 1.16 (1.00-1.33) | 4.46E-02 | 7.14E-01 | 4.34E-02 | 6.94E-01 |
rs3803662 | 16q12.1 |
TOX3
| 0.30 | C vs N | 1.13 (1.01-1.27) | 3.14E-02 | 5.02E-01 | 3.22E-02 | 5.15E-01 |
| | | | C noMut vs N | 1.16 (1.03-1.31) | 1.30E-02 | 2.08E-01 | 1.35E-02 | 2.16E-01 |
| | | | F noMut vs N | 1.16 (1.00-1.34) | 4.45E-02 | 7.11E-01 | 4.53E-02 | 7.25E-01 |
| | | | S noMut vs N | 1.16 (1.00-1.35) | 4.75E-02 | 7.61E-01 | 4.82E-02 | 7.71E-01 |
SNPs rs2981582 and rs766173 are in the same linkage disequilibrium (LD) blocks with rs1219648 (
r
2
= 0,967) and rs1799944 (
r
2
= 1), respectively [
39]. Consistent with expectations, both pairs indicate similar associations.
To further explore associations with breast cancer, we performed analyses separately in groups of familial and sporadic cases, with additional stratification based on mutations in high-risk genes. Two SNPs in
FGFR2 show significant association with both familial and sporadic cases (
p-value
cor ≤ 6.73E-03) (Table
4). Additionally, rs2736098 (
TERT) was associated with sporadic cancers only (
p-value
cor ≤ 0.0234). Results for both types of breast cancer did not change when carriers of high-risk mutation were excluded. All significant associations obtained in the chi
2-test were confirmed by the Cochran-Armitage trend test analysis (Table
4).
Discussion
Several association studies support the polygenic inheritance model of breast cancer, showing increasing risk of disease when many predisposition variants of low effect size were combined [
11,
40]. However, strong bias of the association results, by highly penetrant genetic determinants, such as deleterious mutation in
BRCA1 or
BRCA2 gene, should be taken into account. Also, significant modification of breast cancer risk in
BRCA1/2 mutation carriers was observed in association with selected low-penetrant risk alleles [
41].
In this replication study, association of selected susceptibility SNPs with both familial and sporadic breast cancers was analyzed. Among studied patients, at least one from nine different
BRCA1 mutations was shown in over 11% of sporadic cases and 21% of familial cancers, which is in agreement with previous findings for women in Poland [
28].
From 16 susceptibility variants selected for analysis, four SNPs, representing three different
loci, significantly associated (
p-value
cor < 0.05) with breast cancer risk, both in group of all cases as in sporadic and familial cancer subgroups, and after exclusion of
BRCA1 mutation carriers (Table
4). Two SNPs (rs1219648 and rs2981582) lie within intron 2 of
FGFR2 gene, encoding a receptor tyrosine kinase which participates in activation of signaling pathways engaged in tumor induction and progression [
42] and mediates breast cancer cell proliferation through D-type cyclins [
43]. Amplification or overexpression of
FGFR2 was observed in 5-10% of breast tumors [
44] and breast cancer cell lines [
45].
The association of
FGFR2 variants with breast cancer risk was reported in several studies and is very well documented, with the strongest association observed for rs2981582 [
8,
30]. Minor allele of rs2981582 was found to correlate with positive family history of breast cancer [
46‐
48] and early-onset of non-familial breast cancer [
47]. Data provided by The Consortium of Investigators of Modifiers of BRCA1/2 (CIMBA) indicated that the
FGFR2 locus associated with breast cancer in
BRCA2 mutation carriers but not in
BRCA1 mutation carriers [
49,
50]. This finding may reflect the differences in the distribution of tumor subtype. Rs2981582 is consistently most strongly associated with the estrogen receptor (ER)-positive/low grade tumors [
23], which are more typical for
BRCA2 mutation carriers and general population not selected for carrier status [
51]. Consistently, our findings indicated higher OR and stronger association of both
FGFR2 variants in groups of patients without
BRCA1 mutations (Table
4). Possible correlation of
FGFR2 risk alleles with gene expression was suggested, as intron 2 contains several putative transcription-factor binding sites [
52]. However, relationship of risk allele rs2981582 with increased expression of
FGFR2 has not been clarified yet [
52,
53].
SNP rs13281615 is located in a ‘gene desert’ on chromosome 8q24, where five independent cancer susceptibility
loci were identified so far [
54]. One
loci, termed ‘region 2’ (stretched from 128.35-128.51 Mb), is specific for breast cancer only and tagged by rs13281615 [
38]. Gene desert at 8q24 is located few hundred kilo bases telomeric to the proto-oncogene
Myc. As multiple enhancer elements were identified in this region, it was suggested that they can regulate the transcription of
Myc. One such element was shown to physically interact with
Myc promoter via Tcf-4 transcription factor binding and this interaction affected c-
Myc expression in an allele specific manner [
55]. Overexpression of c-
Myc was observed in breast cancer tissue [
56]; its reduction inhibited breast tumor cells growth [
57]. In agreement with our findings, SNP rs13281615 was associated with increased risk of breast cancer among people at higher risk (who have positive family cancer history or
BRCA1/BRCA2 mutation) [
17,
48]. It was also shown that the association of rs13281615 was stronger for ER-positive disease, with no evidence of an association for ER-negative disease [
58], although, no association with either
BRCA1 or
BRCA2 carriers was observed [
19].
In recent years, the association of rs2736098 (5p15) with cancer risk at different locations was reported, especially for lung and bladder cancers [
59‐
61]. For breast cancer, the reported findings have been controversial [
31,
60,
62]. Haiman et al. [
62] observed positive association of 5p15 locus with increased risk of breast cancer. In turn, Savage et al. [
31] suggested protective effect of three correlated SNPs in this region, including rs2736098, among Polish women with positive family history. Similarly, in our study, rs2736098 minor allele was associated with reduced overall and sporadic breast cancer risk. For familial cancers, association was also observed, although not statistically significant after Bonferroni adjustment.
Rs2736098 is located in coding sequence of
TERT gene, therefore it has been considered as a putative cancer susceptibility gene.
TERT encodes the catalytic subunit of telomerase, which is crucial in cellular proliferation because counteracts telomere-dependent replicative aging [
63]. In many types of cancer,
TERT shows a high-level of expression, which possibly induces excessive cell growth and carcinogenesis [
64]. Although rs2736098 is a synonymous polymorphism, it has been shown to be correlated with telomere length, however not with
TERT expression [
59]. On the other hand, rs2853669, which is in LD with rs2736098 (
r
2
= 0.79), was shown to be involved in allele specific regulation of telomerase activity in non-small cell lung cancer [
65]. Therefore, rs2736098 might be just a tagging SNP of causal variant.
To our knowledge, this is the first such comprehensive study examining association of several potential low-penetrance breast cancer susceptibility
loci among women in Poland. Beside rs2736098 in
TERT, only the association of rs3731249 in
CDKN2A was analyzed previously and significant correlation was identified for early-onset breast cancers [
37]. Our study do not confirm this association in any of analyzed models. One of possible explanations of this discrepancy is that more invasive and aggressive types of cancers might have been included in previous study. Also, correction of significance for multiple testing was not conducted in that study, comparing with ours. However, lack of this SNP association, similarly like in case of studied
BRCA1 variants, could be also explained by insufficient study sample size to detect such association (Additional file
2: Table S2) or SNP effect size lower than expected. The rs2736098 in
TERT locus shows protective effect in both studies and it seems to be specific for the Polish women, indicating the benefit of studying small, homogenous populations for low-penetrance risk variants associations.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Conceived and designed the experiments: JO and EH. Enrolled the patients and performed the experiments: JKL, NM, DN and AN. Analyzed the data: KG, JKL and EH. Wrote the manuscript: JKL and EH. All authors read and approved the final manuscript.