Introduction
Breast cancer is the most common female cancer among western societies and its incidence increases constantly. Hormonal factors like early age at menarche, later age at menopause, later age at first full term pregnancy, and hormone replacement therapy are known to be the main risk factors for sporadic breast cancer [
1,
2]. Also, alcohol appears to contribute to the increased risk for this malignancy, whereas the results concerning smoking are inconsistent [
3‐
6]. The inconsistencies might be due to several factors. For instance, cigarette smoke increases the production of reactive oxygen species (ROS) and contains chemical carcinogens capable of forming DNA adducts [
7], both implicated in carcinogenesis. On the other hand, tobacco smoke has been suggested to have an anti-estrogenic and, therefore, anti-carcinogenic effect [
8]. It has also been suggested that the genetic background might modify the association between tobacco smoke and breast cancer [
4].
Unrepaired or misrepaired DNA damage can lead to gene mutations, chromosomal alterations, and genomic instability, known to have a role in cancer initiation. Accordingly, individuals with reduced DNA repair capacities might have an altered cancer risk [
9‐
11]. During recent years, several polymorphisms in the genes encoding the DNA repair enzymes have been found and studied in relation to cancer proneness [
12,
13].
XRCC1 (X-ray repair cross-complementing group 1) is known to participate in base excision repair (BER) of small lesions such as oxidized or reduced bases, fragmented or nonbulky adducts, and lesions caused by methylating agents [
14]. The XRCC1 is a multidomain protein that has no known catalytic activity itself but it recruits DNA polymerase β, DNA ligase III and poly(ADP-ribose) polymerase (PARP) that are needed at the site of DNA damage [
15].
Three gene polymorphisms resulting in non-conservative amino acid substitutions (Arg194Trp, Arg280His and Arg399Gln) have been identified in the
XRCC1 gene [
12]. Two of these (Arg280His and Arg399Gln) were recently predicted to be likely to affect the function of the protein based on the conservation of the amino acids among protein family members [
16]. The
XRCC1 Arg280His polymorphism lies in between the DNA polymerase β and PARP binding areas [
17]. The
280His variant allele has been suggested to confer increased mutagen sensitivity [
18] and breast cancer risk [
19], although contrasting results exist [
20‐
22].
The
XRCC1 Arg399Gln polymorphism is located in the area coding for a PARP binding site. PARP is a zinc-finger containing enzyme that detects DNA strand breaks [
23]. Carriers of the
XRCC1-399 Gln variant allele have been shown to have higher levels of DNA adducts [
20] and to be at greater risk for ionizing radiation sensitivity [
24] and tobacco-related DNA damage [
25‐
27]. A positive association between the
399Gln variant allele and breast cancer risk has been seen in some studies [
21,
22,
28,
29] while no overall association has been seen in others [
19,
29‐
36].
The nucleotide excision repair (NER) pathway repairs a wide variety of DNA damage, including cross-links, oxidative damage and bulky adducts (such as polycyclic aromatic hydrocarbon (PAH)-DNA adducts). The
XPD (XP complementation group D) gene encodes a helicase involved in transcription and in the NER pathway by unwinding of double helix at the site of deleterious DNA lesions [
37]. Several single nucleotide polymorphisms (SNPs) have been described in the
XPD gene. Two of these SNPs lead to amino acid change, Asp312Asn in exon 10 and Lys751Gln in exon 23, and are in strong linkage disequilibrium with each other [
12,
38]. The
XPD-751 Gln variant allele has been associated with increased DNA adduct levels [
39,
40] and suboptimal DNA repair [
11,
38], but contrasting results also exist [
41]. It has recently been associated with increased risk of smoking-related cancers, such as lung cancer [
42] and squamous cell carcinoma of head and neck [
43], and recently a significant effect was also seen for breast cancer [
44]. Contrasting results also exist [
11,
40,
45‐
47].
In this study, we evaluated the role of XRCC1 Arg280His, XRCC1 Arg399Gln, and XPD Lys751Gln polymorphisms in breast cancer susceptibility in our Finnish Caucasian study population. We especially aimed at examining the genotype effects in relation to smoking, which is known to cause DNA damage repaired by DNA repair enzymes of the BER and NER pathways. The role of alcohol was also evaluated in this context.
Results
Characteristics of the study population are shown in Table
1. Parity was associated with decreased risk of breast cancer while higher waist-to-hip ratio, first degree family history of breast cancer and history of benign breast disease were associated with increased risk. No significant associations were seen between breast cancer risk and smoking habits. Neither were any statistically significant differences seen in the mean number of cigarettes smoked per day (mean 11.2, SD 9.2, and mean 9.5, SD 6.7, for cases and controls, respectively,
p = 0.10), or in smoking years (16.1, SD 10.8, and 14.5, SD 10.8, for cases and controls, respectively,
p = 0.25).
Table 1
Selected characteristics of the study subjects
Age at menarche | | |
≤12 | 98/101 | 1.0 |
13–14 | 219/251 | 0.82 (0.59–1.16) |
≥15 | 150/127 | 0.99 (0.68–1.46) |
Age at first full-term pregnancy | | |
Nulliparous | 102/57 | 1.0 |
≤25 | 237/263 | 0.55 (0.38–0.81) |
26–30 | 94/122 | 0.44 (0.29–0.69) |
≥31 | 47/40 | 0.64 (0.36–1.12) |
Number of full-term pregnancies | | |
Nulliparous | 102/57 | 1.0 |
1 | 68/64 | 0.59 (0.36–0.98) |
2 | 141/181 | 0.50 (0.33–0.76) |
3+ | 171/180 | 0.54 (0.36–0.80) |
Waist-to-hip ratio | | |
≤0.91 | 187/236 | 1.0 |
>0.91 | 291/243 | 1.38 (1.06–1.81) |
First-degree family history of breast cancer | | |
No | 424/459 | 1.0 |
Yes | 54/22 | 2.53 (1.48–4.31) |
History of benign breast disease | | |
No | 296/313 | 1.0 |
Yes | 180/167 | 1.33 (1.01–1.75) |
Current alcohol intake | | |
Never | 271/206 | 1.0 |
Once a month or less | 134/187 | 0.74 (0.54–1.01) |
Daily-weekly | 75/89 | 0.87 (0.59–1.29) |
Smoking habits | | |
Never active or passive | 210/182 | 1.0 |
Passive only | 153/169 | 0.78 (0.58–1.06) |
Ever active | 112/130 | 0.91 (0.65–1.28) |
<5 pack-years | 45/63 | 0.75 (0.48–1.16) |
≥5 pack-years | 67/67 | 1.08 (0.72–1.62) |
The frequencies of the
XRCC1-280 His,
XRCC1-399 Gln, and
XPD-751 Gln variant alleles in controls were 0.08, 0.27, and 0.43, respectively. All the genotype distributions in the control population conformed to Hardy-Weinberg equilibrium (
p = 0.993,
p = 0.657, and
p = 0.876 for
XRCC1-280,
XRCC1-399, and
XPD-751 locuses, respectively). No statistically significant differences were seen in the frequency of these genotypes between cases and controls (Table
2). Neither was any significant difference seen for the polymorphisms when stratified by menopausal status or age (data not shown).
Table 2
Association between XRCC1 and XPD polymorphisms and breast cancer risk
XRCC1-280
|
Arg/Arg | 399 (83.1) | 406 (84.8) | 1.0 |
Arg/His | 78 (16.3) | 70 (14.6) | |
His/His | 3 (0.6) | 3 (0.6) | 1.15 (0.80–1.66)b |
XRCC1-399
|
Arg/Arg | 237 (49.5) | 256 (53.6) | 1.0 |
Arg/Gln | 196 (40.9) | 185 (38.7) | 1.24 (0.93–1.65) |
Gln/Gln | 46 (9.6) | 37 (7.7) | 1.39 (0.84–2.29)c |
Arg/Gln + Gln/Gln | 242 (50.5) | 222 (46.4) | 1.26 (0.96–1.66) |
XPD-751
|
Lys/Lys | 147 (30.6) | 155 (32.3) | 1.0 |
Lys/Gln | 238 (49.5) | 237 (49.4) | 1.03 (0.76–1.40) |
Gln/Gln | 96 (20.0) | 88 (18.3) | 1.10 (0.74–1.63) |
Lys/Lys+Lys/Gln | 385 (80.0) | 325 (81.7) | 1.0 |
Gln/Gln | 96 (20.0) | 88 (18.3) | 1.08 (0.77–1.53) |
When subjects were studied by the stage of the disease at the time of diagnosis, a significant increase for advanced stage (III or IV) breast cancer was seen for women with the XRCC1-399 Gln/Gln genotype (OR 2.86, 95% CI 1.05–7.81) compared to those with the Arg/Arg genotype. No increase in risk was seen for lower stage (I or II) breast cancer (OR 1.33, 95% CI 0.79–2.26). A similar tendency of increased risk for advanced stage cancer was seen for subjects with the XRCC1-280 Arg/His or His/His genotype (OR 1.99, 95% CI 0.90–4.42) compared to subjects with the Arg/Arg genotype. Similarly, women with the XRCC1-399 Gln/Gln genotype presented with a significantly increased risk for grade II and grade III tumours (OR 1.80, 95% CI 1.06–3.07) when compared to those with the Arg/Arg genotype, while no increase was seen for grade I tumours (OR 1.10, 95% CI 0.75–1.62). Moreover, women who carried the XRCC1-399 Gln/Gln genotype and were diagnosed with early stage (I or II) breast cancer had tumours of higher grade (II or III) marginally (p = 0.065, one-sided) more often than those with the Arg/Arg genotype, 86.1% (31/36) versus 72.5 % (137/189), respectively.
When the association between
XRCC1-399 genotypes and breast cancer risk was studied according to smoking habits, increased breast cancer risk with dose-response was seen among women who had ever smoked actively and carried either one (OR 2.14, 95% CI 1.15–3.97) or two (OR 3.27, 95% CI 1.25–8.58,
p for trend = 0.003)
XRCC1-399 Gln variant alleles compared to those carrying the
Arg/Arg genotype (
p for interaction between smoking habits and
XRCC1-399 genotype 0.025) (Table
3). A similar increase in risk was seen for ever smoking women with the
XPD-751 Gln/Gln genotype compared to ever smoking women without this genotype (OR 2.52, 95% CI 1.27–5.03,
p for interaction 0.011).
Table 3
Association between XRCC1 and XPD genotypes and breast cancer risk according to smoking habits
XRCC1-280
| | | | | | | | | |
Arg/Arg | 177 (84.7) | 154 (84.6) | 1.0 | 120 (79.5) | 140 (83.8) | 1.0 | 94 (83.9) | 111 (86.0) | 1.0 |
Arg/His+His/His | 32 (15.3) | 28 (15.4) | 1.09 (0.60–1.99) | 31 (20.5) | 27 (16.2) | 1.11 (0.59–2.08) | 18 (16.1) | 18 (14.0) | 1.41 (0.65–3.08) |
XRCC1-399
| | | | | | | | | |
Arg/Arg | 118 (56.5) | 89 (48.9) | 1.0 | 72 (47.7) | 91 (54.5) | 1.0 | 45 (40.5) | 75 (58.6) | 1.0 |
Arg/Gln | 76 (36.4) | 76 (41.8) | 0.83 (0.53–1.31) | 67 (44.4) | 66 (39.5) | 1.40 (0.84–2.32) | 49 (44.2) | 43 (33.6) | 2.14 (1.15–3.97) |
Gln/Gln | 15 (7.2) | 17 (9.3) | 0.73 (0.33–1.64) | 12 (7.9) | 10 (6.0) | 1.61 (0.61–4.23) | 17 (15.3) | 10 (7.8) | 3.27 (1.25–8.58)b |
Arg/Gln+Gln/Gln | 91 (43.5) | 93 (51.1) | 0.81 (0.53–1.25) | 79 (52.3) | 76 (45.5) | 1.42 (0.87–2.32) | 66 (59.5) | 53 (41.4) | 2.33 (1.30–4.19)c |
XPD-751
| | | | | | | | | |
Lys/Lys | 66 (31.6) | 58 (31.9) | 1.0 | 40 (26.3) | 56 (33.3) | 1.0 | 40 (35.7) | 40 (31.0) | 1.0 |
Lys/Gln | 109 (52.2) | 91 (50.0) | 1.03 (0.63–1.66) | 83 (54.6) | 77 (45.8) | 1.35 (0.78–2.34) | 40 (35.7) | 77 (45.8) | 0.68 (0.35–1.33) |
Gln/Gln | 34 (16.3) | 33 (18.1) | 0.78 (0.40–1.49) | 29 (19.1) | 35 (20.8) | 0.84 (0.41–1.69) | 32 (28.6) | 35 (20.8) | 1.96 (0.89–4.32) |
Lys/Lys+Lys/Gln | 175 (83.7) | 149 (81.9) | 1.0 | 123 (80.9) | 133 (79.2) | 1.0 | 80 (71.4) | 109 (84.5) | 1.0 |
Gln/Gln | 34 (16.3) | 33 (18.1) | 0.77 (0.43–1.39) | 29 (19.1) | 35 (20.8) | 0.70 (0.38–1.29) | 32 (28.6) | 20 (15.5) | 2.52 (1.27–5.03)d |
When ever smoking women were further stratified by pack-years smoked (<5, ≥5 pack-years), the increase in risk was seen to be confined to those who had smoked over five pack-years and carried at least one
XRCC1-399 Gln allele (OR 4.14, 95% CI 1.66–10.3), or the
XPD-751 Gln/Gln genotype (OR 4.41, 95 % CI 1.62–12.0) compared to similarly smoking women without these genotypes (Table
4). Similar effects were seen for the
XRCC1 Arg399Gln genotypes when smokers were stratified by daily tobacco consumption (<10, ≥10 cigarettes/day) or by smoking years (<15, ≥15 years). The ORs were 5.32 (95% CI 1.97–14.4) for women smoking ≥10 cigarettes/day and carrying at least one
XRCC1-399 Gln allele, and 4.03 (95% CI 1.40–11.6) for women who had smoked ≥15 years and carried at least one
XRCC1-399 Gln allele, compared to women with similar smoking habits but with the
XRCC1-399 Arg/Arg genotype. For the carriers of the
XPD-751 Gln/Gln genotype, a similar increase in the risk of breast cancer was seen for women smoking ≥10 cigarettes/day (OR 4.78, 95% CI 1.50–15.2) while no statistically significant increase was seen by smoking years (OR 2.25, 95% CI 0.85–5.96 for women smoking ≥15 years), compared to women smoking the same amount, but not carrying the homozygous variant
XPD-751 Gln/Gln genotype.
Table 4
The association between XRCC1 and XPD genotypes and breast cancer risk according to pack-years smoked
XRCC1-280
| | | | | | |
Arg/Arg | 37 (82.2) | 51 (82.3) | 1.0 | 57 (85.1) | 60 (89.6) | 1.0 |
Arg/His+His/His | 8 (17.8) | 11 (17.7) | 0.91 (0.28–2.95) | 10 (14.9) | 7 (10.4) | 1.99 (0.57–7.02) |
XRCC1-399
| | | | | | |
Arg/Arg | 20 (44.4) | 35 (57.4) | 1.0 | 25 (37.9) | 40 (59.7) | 1.0 |
Arg/Gln | 18 (40.0) | 22 (36.1) | 1.21 (0.46–3.18) | 31 (47.0) | 21 (31.3) | 4.31 (1.66–11.2) |
Gln/Gln | 7 (15.6) | 4 (6.6) | 3.99 (0.82–19.4) | 10 (15.2) | 6 (9.0) | 3.55 (0.81–15.6) |
Arg/Gln+Gln/Gln | 25 (55.6) | 26 (42.6) | 1.61 (0.64–4.06) | 41 (62.1) | 27 (40.3) | 4.14 (1.66–10.3) |
XPD-751
| | | | | | |
Lys/Lys | 18 (40.0) | 20 (32.3) | 1.0 | 22 (32.8) | 20 (29.9) | 1.0 |
Lys/Gln | 18 (40.0) | 31 (50.0) | 1.01 (0.37–2.77) | 22 (32.8) | 38 (56.7) | 0.40 (0.14–1.12) |
Gln/Gln | 9 (20.0) | 11 (17.7) | 1.59 (0.44–5.74) | 23 (34.3) | 9 (13.4) | 2.77 (0.90–8.59) |
Lys/Lys+Lys/Gln | 36 (80.0) | 51 (82.3) | 1.0 | 44 (65.7) | 58 (86.6) | 1.0 |
Gln/Gln | 9 (20.0) | 11 (17.7) | 1.45 (0.46–4.56) | 23 (34.3) | 9 (13.4) | 4.41 (1.62–12.0) |
When stratified by current use of alcohol, women who reported using alcohol weekly to daily and carried the XPD-751 Gln/Gln genotype were at 3.18-fold (95% CI 1.34–7.57) increased risk of breast cancer compared to similarly drinking women carrying the other genotypes (p for interaction 0.026). No interaction was found between XRCC1-280 or XRCC1-399 genotypes and current use of alcohol (data not shown).
When the joint effect of the
XRCC1-280,
XRCC1-399, and
XPD-751 genotypes was studied, a statistically significant increase in the risk of breast cancer was seen for subjects carrying two at-risk genotypes of these genes (OR 1.54, 95% CI 1.00–2.37) compared to subjects with the wild-type genotypes for all three polymorphic sites (Table
5). This increase was mainly due to the combined effect of
XRCC1-399 and
XPD-751 genotypes (OR 1.80, 95% CI 1.05–3.08,
p for gene-gene interaction 0.043). A trend of increasing risk with increasing number of at-risk genotypes was seen (
p for trend 0.042). However, this estimate did not reach statistical significance (OR 4.76, 95% CI 0.48–47.8), possibly due to the low number of subjects with all the three at-risk genotypes (four cases and one control). When the combined effects were studied among ever active smokers, women who carried any two at-risk genotypes were at remarkably increased risk of breast cancer; the adjusted OR was 10.7 (95% CI 3.62–31.6) compared to those without these genotypes (Table
5). This effect was mainly confined to combination of
XRCC1-399 and
XPD-751 at-risk genotypes (OR 12.1, 95% CI 3.52–41.5). When the combined effect was calculated for the number of at risk alleles (
XRCC1-280 His,
XRCC1-399 Gln and
XPD-751 Gln), a similar increase in the risk was seen; subjects with three at-risk alleles had an OR of 1.72 (95% CI 1.03–2.87;
p for trend 0.069) among all women, and OR 4.62 (95% CI 1.56–13.7;
p for trend 0.01) among ever actively smoking women compared to women with no at-risk alleles. Only four cases and one control carried simultaneously four at-risk alleles, and none more than four (of the six).
Table 5
Combined XRCC1 and XPD genotypes and breast cancer risk among all and ever smoking women
0 |
Arg/Arg
|
Arg/Arg
|
Lys/Lys+ Lys/Gln
| 145/159 | 1.0 | 1.0 | 23/46 | 1.0 | 1.0 |
1 |
Arg/Arg
|
Arg/Arg
|
Gln/Gln
| 32/44 | 0.80 (0.48–1.33) | 0.72 (0.42–1.25) | 8/12 | 1.33 (0.48–3.72) | 1.20 (0.37–3.86) |
1 |
Arg/Arg
|
Arg/Gln+ Gln/Gln
|
Lys/Lys+ Lys/Gln
| 172/170 | 1.11 (0.81–1.51) | 1.15 (0.83–1.60) | 43/48 | 1.79 (0.94–3.43) | 1.93 (0.93–4.05) |
1 |
Arg/His+His/His
|
Arg/Arg
|
Lys/Lys+ Lys/Gln
| 45/44 | 1.12 (0.70–1.80) | 1.19 (0.72–1.97) | 11/15 | 1.47 (0.58–3.70) | 1.94 (0.69–5.44) |
1 (any one at risk genotype) |
-
|
-
|
-
| 249/258 | 1.06 (0.80–1.41) | 1.09 (0.81–1.47) | 62/75 | 1.65 (0.91–3.02) | 1.80 (0.91–3.56) |
2 |
Arg/Arg
|
Arg/Gln+ Gln/Gln
|
Gln/Gln
| 46/32 | 1.58 (0.95–2.61) | 1.80 (1.05–3.08) | 19/5 | 7.60 (2.52–22.0) | 12.1 (3.52–41.5) |
2 |
Arg/His+ His/His
|
Arg/Gln+ Gln/Gln
|
Lys/Lys+ Lys/Gln
| 19/18 | 1.16 (0.59–2.29) | 1.28 (0.62–2.64) | 3/0 | - | - |
2 |
Arg/His+ His/His
|
Arg/Arg
|
Gln/Gln
| 13/9 | 1.58 (0.66–3.82) | 1.36 (0.52–3.59) | 3/2 | 3.00 (0.47–19.2) | 3.47 (0.44–27.1) |
2 (any two at risk genotypes) |
-
|
-
|
-
| 78/59 | 1.45 (0.97–2.18) | 1.54 (1.00–2.37) | 25/7 | 7.14 (2.69–19.0) | 10.7 (3.62–31.6) |
3 |
Arg/His+His/His
|
Arg/Gln+ Gln/Gln
|
Gln/Gln
| 4/1 | 4.39 (0.49–39.7) | 4.76 (0.48–47.8)b | 1/0 | - | - |
Discussion
In this study, we examined the role of
XRCC1 Arg280His,
XRCC1 Arg399Gln and
XPD Lys751Gln polymorphisms in relation to breast cancer risk in a Finnish study population. As the products of these genes act in BER and NER pathways, and as some evidence exists on the association of these polymorphisms with smoking-related cancers [
13,
42,
43,
52], our special interest was to study the role of these DNA repair enzymes among smoking women. The hypothesis was also supported by a recent finding of an association between the
XPD-751 Gln/Gln genotype and breast cancer risk in smoking women [
44].
The two polymorphisms
Arg280His and
Arg399Gln in the coding region of the
XRCC1 gene were recently predicted to be 'possibly damaging' to
XRCC1 function based on the conservation of the sequences in mammalian orthologues [
16]. In agreement with this, the frequency of the variant
XRCC1-399 Gln allele was somewhat higher among the present cases compared to controls, leading to a tendency of increased breast cancer risk. A similar effect has been reported in studies among Korean [
28], US radiologic technologists [
21], Indian [
22], and African-American [
29] women. No increased risk was found for white American women [
29], in agreement with three other studies performed among American women [
30‐
33]. Moreover, no association was seen in studies among Chinese [
35], French [
19], Canadian [
34], Turkish [
36] and Danish [
46] women.
In contrast to the
XRCC1 Arg399Gln polymorphisms, the
Arg280His polymorphism did not significantly modify breast cancer risk in the present study. Similarly, no association was seen for Indian women [
22] or for US radiologic technologists [
21]. On the other hand, our findings are in contrast to a French study showing a 1.8-fold (95% CI 1.04–3.08) increase in breast cancer risk for the
XRCC1-280 Arg/His genotype [
19]. One reason for this divergence could be the lack of power; the frequency of the
280His allele is low (0.08) among Caucasians, including Finns. Consequently, even though having almost twice the size of the French study, the power of our study to detect an OR of 1.5 at a 0.05 significance level was only 67% for the
XRCC1-280 His allele containing genotypes. As the
XRCC1-194 Trp variant allele is even less frequent among Finns (approximately 0.03) [
45] compared to other Caucasians, and as the amino acid change has not been shown to affect protein function [
16], we decided not to include this polymorphism in our study.
The
XPD Lys751Gln polymorphism has been suggested to be the most important functional polymorphism in the gene due to major change in the electronic configuration of the respective amino acid in an important interaction domain of the protein [
53]. However, no significant overall association with breast cancer was seen in our study for the
XPD Lys751Gln genotypes. This was in agreement with the other five studies on the
XPD Lys751Gln polymorphism and breast cancer risk including one in Finnish [
45], one in Danish [
46], one in German [
47], and two in US Caucasian women [
11,
40]. In contrast, a significant association between the
XPD-751 Gln allele and breast cancer risk was seen in a recent study among American women [
44]. Moreover, the
XPD Asp312Asn polymorphism was recently shown to be associated with breast cancer risk in a German population [
47]. We decided not to analyse the
XPD Asp312Asn polymorphism as it has been shown to be strictly linked with the
Lys751Gln polymorphism [
12,
38].
When the present study subjects were stratified by stage of disease, the XRCC1-399 Gln allele posed an elevated risk for more advanced stage breast cancer. A similar tendency of increased risk for more advanced stage breast cancer was also seen for the Arg280His polymorphism. It can be hypothesized that defective DNA repair leads to more aggressive and, therefore, more advanced tumours at the time of diagnosis. This was also supported by the association of the XRCC1-399 Gln allele with higher grade tumours. However, as earlier studies on breast cancer have not evaluated the genotype effects by the stage of the disease or tumour grade, these findings remain to be confirmed in future studies.
Smoking alone did not significantly affect breast cancer risk in the present study. This is in agreement with the majority of epidemiological studies on smoking and breast cancer risk, as well as with a recent report of the Collaborative Group on Hormonal Factors in Breast cancer, which concluded that cigarette smoking has little or no effect on the risk of developing breast cancer [
3]. There are, however, some studies reporting increased risk in special subgroups, such as women who started to smoke at an early age or before first pregnancy, women smoking high intensity or long duration, passively smoking women, and women with specific genotypes (reviewed in [
4]). In our study, a significant interaction was seen between smoking habits and the
XRCC1 Arg399Gln (
p = 0.025) or
XPD Lys751Gln (
p = 0.011) genotypes. Subjects with the variant
Gln/Gln genotypes were at increased risk of developing breast cancer if they had ever smoked. Furthermore, a gene-dosage effect was seen for the
XRCC1 Arg399Gln genotype; the increased risk was higher for subjects carrying two
XRCC1-399 Gln alleles. In contrast, the effect of the
XPD-751 Gln allele seemed to be recessive. When ever smokers were further stratified by the amount of smoking, the risk estimates remained statistically significant among those who reported having smoked more than five pack-years.
The associations found between the
XRCC1 Arg399Gln and/or
XPD Lys751Gln genotypes, smoking, and breast cancer risk are biologically plausible. Cigarette smoke is a rich source of chemical carcinogens and ROS [
7]. ROS can initiate lipid peroxidation, oxidize proteins, and cause base damage and DNA single strand breaks that are repaired through BER. Accordingly, the
XRCC1-399 Gln/Gln genotype has been found to be related to increased sister chromatid exchange frequencies among smokers [
25]. Similarly, a higher frequency of sister chromatid exchange has been reported for carriers of the
399Gln allele who have smoked ≥10 cigarettes/day [
27], and higher levels of tobacco-specific nicotine-derived nitrosamino ketone-induced sister chromatid exchange in cells with the
XRCC1-399
Gln allele containing genotypes compared to cells homozygous for the
Arg allele [
26]. It can thus be hypothesized that the
XRCC1-399 Gln allele is associated with increased risk of smoking-dependent cancers. In agreement with this, a higher risk of lung cancer has been reported for the carriers of the
XRCC1 Gln/Gln genotype [
52], although negative studies also exist [
54,
55]. Breast cancer is not generally regarded as a smoking-dependent cancer. In a recent study on breast cancer, the
XRCC1-399 Gln allele was significantly associated with detectable PAH-adducts only in never smokers [
33]. They speculated that smoking might stimulate DNA repair and thus the effect seen would be due to other sources of PAHs. Only one earlier study on breast cancer risk has reported a positive association between
XRCC1-399 polymorphism and smoking, but in contrast to our study, the highest risk was seen for subjects who were homozygous for the
XRCC1-399 Arg allele [
29]. Unfortunately, the possible association with smoking has not been reported in all studies on the
XRCC1-399 genotypes [
19,
22,
30,
31,
36].
In addition to ROS, tobacco smoke includes chemical carcinogens, such as polycyclic aromatic hydrocarbons, aromatic amines and tobacco-specific nitrosamines, that can produce bulky DNA adducts, which are repaired through the NER pathway. There is evidence that tobacco smoke constituents can reach breast tissue [
7], and higher levels of the tobacco smoke derived 4-aminobiphenyl adducts [
56] and PAH-DNA adducts have been found in breast tissue of breast cancer cases [
57‐
59]. Mammary epithelial cells are capable of metabolising and activating these compounds. Therefore, the hypothesis is that breast cancer cases with suboptimal DNA repair capacity would have less efficient removal of these adducts, thus making them more vulnerable to the hazardous effects of these compounds. Accordingly, the
XPD-751 Gln allele has been associated with increased DNA adduct levels in never smokers [
39] and in tumour tissue from breast cancer cases [
40], as well as with suboptimal DNA repair [
11,
38]. A recent meta-analysis showed an association between the
XPD-751 Gln/Gln genotype and increased lung cancer risk [
42]. This was not, however, confirmed by another recent meta-analysis [
60]. In accordance with our results, a similar increased breast cancer risk was recently seen among currently smoking American women with the
XPD-751 Gln/Gln genotype [
44]. Similarly to our study, no clear pattern with duration of smoking was seen. This could partly be due to low numbers in these strata, or it can be hypothesized that a higher intensity of smoking is needed to see the difference between these genotypes, while in subjects with lower intensity but longer duration the repair system is capable of overcoming the harmful effects.
Alcohol is a known risk factor for breast cancer; increased estrogen and androgen levels have been observed in alcohol consuming women [
61]. Alcohol consumption is also believed to contribute to oxidative stress [
62]. We therefore analysed the results also stratified by the use of alcohol. A significantly increased three-fold risk of breast cancer was seen for women who carried the
XPD-751 Gln/Gln genotype and used alcohol weekly or daily. No earlier studies on the
XPD genotype and breast cancer have reported an association with the use of alcohol. However, a 2.59-fold (95% CI 1.25–5.34) risk was seen for squamous cell carcinoma of head and neck among current drinkers carrying the
XPD-751 Gln/Gln genotype compared to those drinking the same amount but with the other genotypes [
43].
When the gene-gene interactions were studied, clear combined effects were seen, especially among ever smoking women; smoking women with any two at-risk genotypes were at remarkably increased risk of breast cancer (OR 10.7, 95% CI 3.62–31.6) compared to smoking women without these genotypes. These effects were mainly due to XRCC1-399 and XPD-751 genotypes while the XRCC1-280 genotype was a minor contributor, possibly partly due to the very low frequency of the XRCC1-280His variant alleles. When the combined effect was studied according to the number of at-risk alleles, subjects with any three variant alleles were found to be at increased risk. There was no subject who simultaneously carried more than four at-risk alleles in these three polymorphic sites.
It should be noted that even if our study involved a reasonable number of study subjects, the numbers in some subgroup analyses were small, and the statistical power to detect significant point-estimates is, therefore, low. Furthermore, due to multiple comparisons performed in the study, the possibility of chance findings should be borne in mind. However, in addition to the plausibility of the findings, the dose-response seen by the number of at-risk alleles or genotypes and amounts smoked support the causality of the findings.
In case-control studies, the possibility of selection and recall bias should be considered. However, the detailed data about smoking habits and other life style factors were collected prior to diagnosis, and the exposure to passive smoking, quite often disregarded, was taken into account. It is also unlikely that selection would be related to an individual's genotype. As the statistical analyses were performed within exposure groups and not between them, the possibility of biased estimates is further reduced.
Even though the present findings are biologically plausible, the possibility of the association with another gene polymorphism in the same region should be considered. Accordingly, it was recently reported that three markers,
RAI,
ASE-1 and
ERCC1, in the same chromosomal region 19q13.2-3 define a high-risk haplotype that poses women under 55 years to a significantly increased breast cancer risk [
46]. In that study, no association was seen with the
XRCC1 or
XPD genotypes but the effects were not studied by smoking habits. In our study, no difference in the risk estimates was seen by age.
Finally, it should be noted that not all the known polymorphisms in the BER and NER pathways were analysed in this study. For the
XRCC1 Arg194Trp polymorphism, association has been seen with increased mutagen sensitivity [
63], risk of developing an adverse response to radiotherapy [
19], as well as interaction with high intake of fruits and vegetables [
33], and it could thus be hypothesized to affect breast cancer risk in population where it exists in higher frequency. Similarly, polymorphism has been detected in another gene,
OGG1, encoding an enzyme important in the BER pathway, as well as numerous other genes encoding for possibly important repair enzymes in the NER pathway, namely
XPA,
XPC,
XPF and
XPG.
Authors' contributions
KM was responsible for supervising the laboratory work, for performing statistical analyses, and for manuscript preparation. VK participated in the design and coordination of the study, helped to draft the manuscript and acts as the principal clinician in the study group. PS, PS and LH carried out the genotyping analyses and drafted the manuscript. ME participated in the design and coordination of the study and was clinical coordinator of collection of study patient material. VMK performed histopathological analyses of all cases, and helped to draft the manuscript. MU has been involved in the planning of the study, collection of clinical data, in particular with respect to the genetic part of the study, and he has commented on the manuscript in the writing phase. AH participated in the design and coordination of the molecular epidemiology part of the study, was responsible for acquisition of funding for the molecular epidemiology part, and was involved in the preparation of the manuscript. All authors read and approved the final manuscript.