Association of single nucleotide polymorphisms in the genes ATM, GSTP1, SOD2, TGFB1, XPD and XRCC1 with risk of severe erythema after breast conserving radiotherapy

The treatment of malignant tumours by radiotherapy (RT) is limited by the need to avoid unacceptable normal tissue toxicity. Despite advances in RT-technique, treatment modalities as well as therapeutic strategies, normal tissue damage is still a limiting factor in radiotherapy. In this context, late complications are especially important because they are generally progressive and appear to be associated with a lifelong risk [1]. In contrast, acute normal tissue toxicity is generally a transient phenomenon, with symptoms settling within months after treatment. However, these effects are not less clinically relevant, especially when accelerated fractionation schedules or adjuvant radiochemotherapy treatment are used [2‐4] with new substances often increasing normal tissue toxicity, e.g. Cetuximab.

Both acute and late normal tissue effects are known to vary considerably, ranging from negligible to severe, even between patients treated with identical schedules. It has been suggested that these variations in clinical radiosensitivity mainly result from differences in genetically determined radiosensitivity, as only 30% of this variation can be attributed to changes in treatment related parameters [5, 6]. In addition to clinical and genetic parameters for radiation response, a number of patient-related confounding factors exist influencing adverse effects definitely, some of which probably have yet to be identified. Based on this background, pronounced scientific interest is currently being directed towards the use of genetic markers such as single nucleotide polymorphisms (SNPs) as parameters for the individual risk of experiencing radiation-induced normal tissue toxicity. Studies on the impact of SNPs are either performed following the candidate gene approach or employing genome-wide association (GWA) analysis.

So far, numerous studies have been performed, but the results obtained are heterogeneous and often conflicting; for reviews, see [7, 8]. Even for the C-509 T polymorphism in the transforming growth factor 1 (TGFB1) gene, which represents one of the most studied SNPs, there are several reports [9‐13] showing that this polymorphism promotes chronic inflammatory and fibrotic reactions, but there are also data suggesting a lack of association or even the opposite effect [14‐17].

In addition to the identification of those SNPs which are relevant for normal tissue toxicity, their functional consequences concerning the involved molecular pathways and their mechanisms of action are currently of great scientific interest. A respective consortium addressing this topic has recently been established [18].

The discrepancies in the currently available data may be attributed to the fact that with a frequency of about one SNP every 160 to 180 bp [19‐21], the vast majority of SNPs are assumed to have no or a small effect on the respective protein or functional pathway. respectively. Therefore individual genetic characteristics must be determined by the combination of several SNPs each one associated with a small effect. This was indicated by previous studies [7, 22] which demonstrated a significant association with normal tissue toxicity only if several - each of one only weakly associated SNPs - were combined to a risk score. Such risk scores can easily be created by adding the number of risk alleles per patients and correlating the resulting numerical value with the severity of the normal tissue toxicity.

It is discussed that a risk profile, based on a combination of SNPs in genes which are involved in relevant pathways may vary for the type of normal tissue toxicity scored [7], i.e. different endpoints are characterized by different mosaic-like displays of certain SNPs.

This concept implies that it might be more informative to analyse a certain combination of SNPs in independent studies rather than to change this combination within different studies.

To date, such analyses have only been performed by Andreassen et al. [10, 14, 23] using certain combinations of SNPs all including SNPs in TGFB1 SOD2 XRCC1 XRCC3 APEX and ATM genes and studying their association with skin fibrosis in post-mastectomy radiotherapy patients. While the agreement between the first two studies analysing two different cohorts was fairly good [10, 14], the third study on a larger cohort of patients [14] failed to confirm the association with radiation induced, indicating that future studies should consider other combination of SNPs [7].

In our studies, we concentrated on six SNPs located in genes (ATM, GSTP1, SOD2, TGFB1, XPD, XRCC1) involved either in the induction or repair of DNA double-strand breaks and therefore considered to be of relevance for individual radiosensitivity. In a first report, the association of these SNPs with the risk of late tissue effects was analysed for breast cancer patients treated with breast conserving radiotherapy [22]. No significant association with risk was obtained for any individual SNP, as indicated by p-values ranging from 0.064 to 0.643. However, when these SNPs were combined into a risk score, a highly significant association (p = 0.0005) was found.

We now investigated in a retrospective study on 83 breast cancer patients the association of these SNPs with the risk of acute tissue toxicity in terms of erythema, with special focus on the relevance of breast size. Erythema grade 2 and more was used as clinical endpoint determined by using the RTOG score. DNA extracted from blood was used to determine the SNP status using either PCR-RFLP or MALDI-TOF. The results imply that fundamental differences exist concerning the risk profile for late and acute tissue toxicity, respectively.

There is great interest in establishing methods which can be used to predict the individual risk of normal tissue effects after radiotherapy. If these risks were known prior to the onset of therapy, the total dose applied could be reduced in the small proportion of highly sensitive patients and, conversely, radiation dose and possibly the chance to cure could be increased for normal and resistant patients [28, 29].

For radiotherapy, the risk of side effects is considered to be determined mostly by treatment-related factors such as dose and dose per fraction, but also to a considerable extent by genetic parameters [5, 30]. In this context, SNPs are considered to be the best genetic markers [8].

Following a candidate gene approach, six polymorphisms were selected in genes (ATM GSTP1 TGFB1 SOD2 XPD XRCC1) that are known to be associated with the induction or repair of DNA-damage [31‐36]. In our first study on breast cancer patients treated by conserving radiotherapy, these SNPs were compared with the risk of severe radiation-induced fibrosis [22]. For each SNP, only a weak association was found, with p-values ranging between 0.064 and 0.643. However, combining all six SNPs into one risk score yielded a highly significant association (p = 0.0005).

In this study, we have now analysed the association of the same SNPs with the risk of acute effects as observed after breast conserving radiotherapy. In order to increase the performance of the study, the inclusion criteria were applied strictly for all treatment parameters; i.e. variation in the total dose was negligible, as 50 or 50.4 Gy were applied using only two fractionation regimes (2.0 or 1.8 Gy). Since total dose is considered to be a crucial factor in studying genetic determinants of radiation response [30], the detriment of a relatively small patient collective was counterbalanced by the extremely homogenously performed treatment.

The association between polymorphisms and acute effects was analyzed using the trend test, which reflects the dosage effect of the number of mutant alleles. Due to deviation from the Hardy-Weinberg-equilibrium, one SNP (SOD2, codon 16) was excluded from the analysis (Table 1). When the analysis was performed with all patients, none of the five SNPs studied was found to show a significant association with erythema (Table 2). This is in line with most other studies in breast cancer patients (for review see [8]) showing no significant association with acute effects for the vast majority of SNPs. An association has only been found for GSTP1 codon 105 [37] - which was, however, not observed in our report (Table 2) - as well as for SNPs in IL12RB2 and in ABCA1[38] as well as CD44, MAD2L2, PTTG1, RAD9A and Lig3[39].

Since a previous study [26] proved breast volume to be of importance for the incidence of radiation induced erythema, and other reports also verify breast size as an important confounding factor in the development of radiation-induced erythema [16, 40‐43] we performed an additional analysis dichomising the patients according to their breast volume. This analysis yields a significant association of TGFB1 C-509 T for patients with smaller breast volumes and XPD codon 751 for patients with larger breast volume (Table 2). Due to relative small sample sizes we abstained from Bonferoni-correction, following other studies on this issue {[8, 44, 45] for Review see [8, 46]. It cannot be excluded that the observed differences in SNP associations between breast size might be the result of random fluctuations rather than being a reflection of a true difference in the radiobiology of large and small breasts, respectively. Therefore verification studies are needed to test the evolving hypothesis that different pathways and with that different SNPs might be relevant for clinical radiosensitivity according to breast size - and by that possibly to body mass index and metabolic parameters.

For the other four SNPs, no significant association was found (Table 2), even when combined into a risk score. These results indicate that probably due to differences between smaller and larger breasts with respect to micro-milieu and/or metabolism, other parameters and with them other genes are of relevance for the formation of acute effects. As a consequence, more significant associations could potentially have been found in other reports if breast volume had been considered as a relevant confounding factor as done here (Table 2). It has also been shown that the total dose applied must be considered as a relevant confounding factor in a radiogenomic analysis [23, 47]. Certainly more of these factors need to be taken into consideration in order to detect the moderate effects caused by a single SNP.

For the five SNPs studied, we also demonstrated that the risk profile obtained for erythema was clearly different from that previously found for fibrosis in breast cancer patients treated by conserving radiotherapy [22]. There are also several reports on breast conserving radiotherapy in which the association between certain SNPs was compared with both acute and late effects. The by far largest study was performed by Chang-Claude and colleagues, analysing the association between polymorphisms in certain DNA repair genes and either acute [48] or late side effects [49]. Although these studies did not always use the same set of polymorphisms, the data obtained confirm that for different clinical endpoints, the risk is determined by different combinations of SNPs. There are numerous reports investigating the impact of mutations in ATM on side effects in breast cancer patients. A clear different influence was seen in one study [50], but not in the others [51‐53].

The results obtained in the present study indicate that the endpoints of acute and late tissue toxicity are determined by different molecular and cellular pathways, respectively. Therefore, analysing the association of SNPs with both acute and late effects will help us not only to identify the SNPs which might be used as markers of the respective risks, but also to unravel the underlying biological mechanisms.

In summary, this study demonstrates for the first time that significant associations between a specific SNP and risk of erythema can be identified if breast cancer patients are grouped by their breast volume. The combination of SNPs using risk alleles according to erythema is substantially different from a risk score previously defined for risk of fibrosis. However, these results need to be replicated in an independent and larger study.

It is hereby confirmed that all authors disclose all financial and personal relationships with other people or organisations that could inappropriately influence (bias) their work.