IL12RB2 and ABCA1 Genes Are Associated with Susceptibility to Radiation Dermatitis

Breast-conserving therapy, consisting of breast-conserving surgery and prophylactic breast irradiation, is the standard therapy for patients with early breast cancer. In most institutions, a total dose of 45 to 50 Gy is delivered to the whole breast with a daily fraction of 1.8 to 2 Gy; this dose fractionation is regarded to be effective and safe, considering both the excellent local control rate and the low probability of severe radiation-related toxicity (1).

Mild acute radiation dermatitis is commonly observed during or shortly after the completion of radiotherapy. However, large interindividual variations in the severity of dermatitis exist even when the patients have been uniformly irradiated. Approximately 5% to 10% of patients develop moderate to severe acute radiation dermatitis following whole-breast radiotherapy (2, 3).

Variations in the severity of radiation dermatitis are influenced by both treatment-related and patient-related factors. Treatment-related factors include radiotherapy variables (e.g., beam energy, radiation dose, dose fractionation, overall treatment time, heterogeneity in dose distribution, and concurrent chemotherapy). Hotspots produced by radiation dose heterogeneity often result in focally enhanced dermatitis. Patient-related factors include patient age, menopausal state, physique, and coexistent diabetes mellitus or collagen disease. In addition, several genetic syndromes such as ataxia telangiectasia (4, 5), Fanconi anemia (6, 7), and Nijmegen breakage syndrome (8, 9) have been reported to account for a small, but prominent, percentage of the hyperradiosensitive population.

In the majority of patients with moderate to severe radiation dermatitis, however, the cause of the radiosensitivity is unknown, implying the existence of undetermined intrinsic factors (10). Establishing a system for predicting patients with intrinsic radiosensitivity before subjecting them to conventional radiotherapy would be clinically useful for the individualization of radiotherapy and, consequently, for improving the treatment outcome.

The aim of the present study was to identify radiation dermatitis-related single-nucleotide polymorphisms (SNP) using peripheral lymphocytes from patients who developed radiation dermatitis after whole-breast irradiation.

Patient selection and study design. A multi-institutional, case-control study comprehensively analyzing SNPs and comparing alleles between control patients and patients who were considered to have intrinsic radiosensitivity was conducted. This study was approved by the institutional review board of each institution involved in the study. Candidate patients were selected from a pool of more than 3,000 female patients who had undergone whole-breast radiotherapy following breast-conserving surgery for the treatment of early breast cancer since June 1993 at nine institutions. The eligibility criteria were as follows: (a) patients who had undergone a quadrantectomy or wide excision for unilateral early breast cancer, (b) patients who had received traditional tangential whole-breast irradiation at a total radiation dose of 44 to 52 Gy at a daily fraction of 1.8 to 2.2 Gy over a period of less than 8 wk using a cobalt-60 source or a linear accelerator generating 4 to 6 MV X-rays, (c) patients whose radiation dose distribution was available, (d) patients who did not receive systemic chemotherapy or hormonal therapy except for oral fluorouracil or tamoxifen during the radiotherapy period, and (e) patients whose skin reactions were followed-up for at least 6 mo after the completion of radiotherapy.

The severity of dermatitis was graded according to the National Cancer Institute Common Toxicity Criteria. The dermatitis group in the present study was regarded to represent patients with intrinsic hypersensitivity. Clinical records, photographs of the skin before and after irradiation, and the radiation dose distribution charts were carefully reviewed to exclude confounding external factors. To exclude patients with treatment-related dermatitis, patients with focal dermatitis that could be explained by comparing the dose distribution and the dermatitis distribution to reveal focal hotspots or focal boost irradiation were strictly excluded from the present study. For example, localized dermatitis of grade ≥2 that was limited to the axilla, the submammary fold, or the nipple was regarded as indicating ineligibility for inclusion in the dermatitis group. Only patients with acute radiation dermatitis of grade ≥2 distributed evenly over the irradiated skin were regarded as having intrinsic radiosensitivity and were included in the dermatitis group.

The control group was determined by selecting patients from the pool of patients who developed grade 0 to 1 dermatitis. Patient-to-patient matching between the dermatitis group and the control group was used to ensure that the two groups would be well balanced in terms of their major characteristics and therapeutic variables. The matched factors included the patients' ages, menopausal states, use of concurrent oral fluorouracil, institution where the radiotherapy was done, radiation dose fractionation, radiation quality, and the year of radiotherapy. Two typical cases from the dermatitis group and a case from the control group are shown in Fig. 1.

Between September 2001 and September 2003, 77 patients in the dermatitis group and 79 patients in the control group were interviewed to confirm their eligibility for enrollment in the study; 15 mL of peripheral venous blood were then drawn after obtaining the patient's written informed consent.

Single-nucleotide polymorphisms. To identify polymorphisms associated with radiation dermatitis, 494 genes were selected for analysis (Table 1). These genes included DNA repair genes, apoptosis-related genes, inflammation-related genes, angiogenesis-related genes, and transporter genes. We then searched for SNPs located within or close to these genes using the JSNP database (11). A total of 3,144 SNPs were selected for genotyping. A complete list of the SNPs used in the present study is available on our Internet web page (URL to be updated).

Genotyping. After informed consent was obtained, 15 mL of peripheral blood were obtained from each patient. DNA was extracted from mononuclear cells using standard methods. Genotyping was done according to the high-throughput SNP typing method developed at Riken (12). Fluorescent signals were detected using Tecan Ultra (Tecan Group Ltd.). The genotypes were determined using automated genotyping software (13).

Statistical analysis. The associations between the SNPs and radiation dermatitis were examined using a test for independency with a 2 × 3 contingency table, in which the two distributions corresponded to patients with and those without radiation dermatitis. As we had already selected candidate genes that we suspected to be associated with radiation dermatitis, the SNPs were searched for in a manner that minimized the chance of missing any SNP associated with dermatitis. For this reason, the cutoff point was set at P < 0.01. We selected a locus when multiple SNPs mapped to that region were associated with radiation dermatitis at a P value smaller than the cutoff value. After screening the loci, a Fisher's exact text for a 2 × 2 contingency table was done on all possible SNPs mapped to and around the loci, assuming allele frequency, dominant, or recessive models.

Estimation of probability of radiation dermatitis. The probability of radiation dermatitis was calculated for each combination of the two SNPs selected during the second-step screening process. To compute the probability, a logistic regression model with four indicator variables for the two different SNPs was adopted. The allele frequencies of each genotype combination were then calculated based on the allele frequency of individual SNPs obtained from our genotyping data of 1,594 unrelated Japanese individuals.

Patient characteristics. The major patient characteristics and therapeutic variables are summarized in Table 2. About the patient-related factors, age, menopausal state, and the pathologic T and N classifications were similar in the radiation dermatitis and control groups. The two groups were generally well balanced with regard to the treatment-related factors such as the type of surgery, period between surgery and the commencement of radiotherapy, the presence of concurrent chemotherapy with oral fluorouracil, the radiation source, the total radiation dose to the whole breast, the dose per fraction, the overall treatment time for whole-breast radiation, and the accumulative dose per week. A significant difference was found for only one factor, the presence of concurrent tamoxifen chemotherapy; 51% of the radiation dermatitis group and 68% of the control group (P = 0.024) had received tamoxifen during the radiotherapy period.

Genotyping of SNPs. The genotypes of several loci were determined using an Invader assay in 156 patients with breast cancer who received radiation therapy and 1,438 unrelated subjects. Consequently, the genotypes of 3,144 SNPs covering 494 genes were determined. The genes investigated in this study are listed in Table 1. To obtain accurate results in this case-control study, SNPs that departed from Hardy-Weinberg equilivbrium or that had a low minor allele frequency were excluded from further examination. Consequently, 2,651 SNPs were chosen for further analysis. The accuracy of the genotyping was tested by comparing the genotypes determined using the Invader assay with those determined using a RFLP analysis. More than 1,000 genotypes were assayed in three randomly chosen SNP loci. No genotypes showed any discrepancies between the two methods, indicating that the Invader assay had produced highly accurate genotypes (data not shown).

Identification of SNPs associated with radiation dermatitis. As a result of the screening steps described above, two loci were found to be associated with an adverse effect: the IL12RB2 gene locus and the ABCA1 gene locus. At the IL12RB2 locus, 13 SNPs were examined (Supplementary Table S1). Among them, five SNPs were associated with radiation dermatitis (Table 3A). The strongest association was observed at SNP-K (rs3790568) in recessive model (P = 0.0065). To test the possibility that known coding SNPs mapped to the IL12RB2 gene were associated with radiation dermatitis, we searched a dbSNP database and found six coding SNPs (rs2307146, rs7526769, rs2307148, rs2307145, rs2307153, and rs2307154). The genotypes of these coding SNPs were determined, but all the patients were wild-type (data not shown). At the ABCA1 locus, 50 SNPs were analyzed (Supplementary Table S2); 8 SNPs were associated with radiation dermatitis, and the lowest P value of which was 0.0013 for SNP-17 (rs2253304) in allele frequency model (Table 3B). No other locus was associated with radiation dermatitis.

Haplotype block structure of IL12RB2 and ABCA1 gene loci. To narrow down the loci associated with radiation dermatitis, the haplotype block structures of both genes were estimated. The haplotypes and diplotypes of 13 SNPs at the IL12RB2 locus were estimated in 1,594 unrelated individuals. Using the diplotypes of these individuals, the haplotype block structure of the IL12RB2 locus was estimated using AdBlock software. Consequently, the 13 SNPs at the IL12RB2 locus were subdivided into three haplotype blocks. All the SNPs associated with an adverse effect were located in the second haplotype block (Supplementary Table S1). As for the ABCA1 gene, haplotype block analysis of the ABCA1 gene revealed that the locus was subdivided into seven haplotype blocks. The SNPs associated with an adverse effect were located in the third and fifth haplotype blocks (Supplementary Table S2).

Estimation of the probability of radiation dermatitis for combined genotypes. To assess the probability of radiation dermatitis precisely, the genotypes of the two loci were used in a logistic regression model to estimate the probability of radiation dermatitis. Model selection using Akaike information criterion selected the best pair of SNPs, which consisted of SNP-15 in ABCA1 and SNP-K in IL12RB2. These analyses revealed that the genotype combination of a G homozygote at both SNP-15 and SNP-K exhibited the highest probability (0.75) of developing radiation-induced dermatitis (Table 4). This result indicates that 75% of patients carrying the combination of genotype have a chance to develop radiation-induced dermatitis. The genotype frequencies of these two SNPs were estimated based on genotyping data from 1,594 unrelated individuals. Based on these results, the frequency of the genotype with the highest probability of radiation dermatitis was estimated to be 0.147 in the Japanese population (Table 4).

In this study, we showed associations between polymorphisms in the IL12RB2 and ABCA1 loci and an adverse effect in patients who received radiation therapy after undergoing surgery for breast cancer. SNPs associated with radiation dermatitis were observed in one or two of the haplotype blocks of the IL12RB2 or ABCA1 genes, respectively. These observations indicated that the associations between the SNPs and the adverse effect were nonrandom events.

Interleukin-12 (IL-12) receptor is a heterodimer protein consisting of two subunits: IL12RB1 and IL12RB2 (14). Several studies have shown that IL12RB2 may regulate IL-12 function (15). On the other hand, the cytokine IL-12 shows a wide variety of biological activities both in vitro and in vivo. Several studies have shown the antitumor activity of IL-12 through an effect on the host immune system. Schwarz et al. (16) reported that IL-12 suppresses UV radiation-induced apoptosis. IL-12 also prevents immunosuppression caused by UV irradiation. They concluded that IL-12 induces the expression of genes involved in nucleotide excision repair and that the activated nucleotide excision repair system may protect UV-irradiated cells from undergoing apoptosis. In their report, IL-12 was unable to suppress apoptosis caused by γ-irradiation in vitro. However, the effect of IL-12 on γ-irradiated cells in vivo remains to be observed. Our results may suggest that changes in the genotype of SNPs in IL12RB2 may regulate the biological function of IL-12, causing differences in the biological response to ionizing radiation.

The association between radiation-induced dermatitis and polymorphisms in the ABCA1 gene is rather unexpected because the ABCA1 gene plays an essential role in the efflux of cholesterol to high-density lipoprotein. However, the ABCA1 gene has also been shown to be involved in the engulfment of apoptotic cells. Hamon et al. (17) showed that the loss of ABCA1 function impaired the engulfment of cell corpses generated by apoptosis. This result also suggests that the functions of the ABCA1 gene product are linked to apoptosis. In our study, a coding SNP (SNP-15; rs2230806), in which the nucleotide polymorphisms cause an amino acid substitution from Lys to Arg at the 219th amino acid, was associated with radiation-induced dermatitis. Considering the link between the ABCA1 gene and apoptosis, this amino acid substitution may influence apoptotic reactions caused by ionizing radiation, thereby influencing the degree of dermatitis that occurs.

Data interpretation. The patient selection policy of the present study was developed to minimize the problems that a multi-institutional case-control study is liable to suffer. This policy was characterized by the patient-to-patient matching selection to avoid intergroup bias, the photograph-based confirmation for grading dermatitis, and the dose distribution-based exclusion of radiotherapy-related confounding factors.

In spite of patient-to-patient matching for the major prognostic factors, the proportion of patients treated with concurrent tamoxifen therapy were significantly larger in the control group than in the radiation dermatitis group (P = 0.024). This difference was the only intergroup bias observed out of all the patient- and treatment-related factors. The clinical significance of concurrent tamoxifen therapy in radiotherapy-related toxicity is controversial. Several reports have shown an increased risk of pulmonary and breast fibrosis after concurrent treatment with tamoxifen therapy and radiotherapy (18, 19), whereas other reports did not show any risk (20). Furthermore, an in vitro experimental study showed that the radiosensitivity of hormone-dependent breast cancer cells might be reduced through the cytostatic effect of tamoxifen (21). However, no experimental or clinical studies have suggested a radioprotective effect of tamoxifen on hormone-independent normal cells, including lymphocytes or fibroblasts. Therefore, we suspect that this intergroup bias did not seriously affect the interpretation of the SNP data.

Clinical relevance. Many ex vivo prediction systems have been developed and evaluated in which various numbers of patients treated with radiotherapy were studied prospectively or retrospectively using cellular or molecular ex vivo assays of lymphocytes or fibroblasts, with acute or late adverse effects as the clinical end points. The results of studies comparing clinical and cellular radiosensitivity using colony assays, micronucleus assays, and comet assays were often suggested to be promising. However, as reviewed by Twardella and Chang-Claude (22), the clinical relevance of these predictive assays should be carefully evaluated, considering any possible bias that may arise from inappropriate settings of patient cohorts, poorly defined clinical end points, or contamination by confounding external factors.

Recently, several studies have been undertaken to evaluate the correlation between clinical radiosensitivity and SNPs (23–26). Mostly, SNPs in selected genes whose functions may modify radiation-related toxicity, such as ATM, TGFB1, and XRCC1, have been analyzed. For instance, Andreassen et al. (23) analyzed seven SNPs in five genes (TGFB1, SOD2, XRCC1, XRCC3, and APEX) using cultured fibroblasts obtained from 41 patients who had been treated with uniform postmastectomy radiotherapy. Their data indicated that five of the seven SNPs were significantly correlated with the occurrence of grade 3 s.c. fibrosis or grade 2 to 3 telangiectasia in the irradiated skin. They concluded that SNP analysis may have the potential to predict clinical radiosensitivity, particularly when multiple SNPs are analyzed.

By logistic regression analysis, probabilities of radiation-induced dermatitis were estimated on each combination of genotypes of two SNPs (rs3790568 and rs2230806). These analyses revealed that three patients in four patients carrying a specific combination of genotypes, G homozygote in two SNPs, develop radiation-induced dermatitis. Genotype frequency of combination of the SNPs was estimated as 0.147 in the Japanese population. Therefore, ∼11% of Japanese are predicted to develop radiation-induced dermatitis. Considering incidence of radiation-induced dermatitis (5-10%), this number seems to be reasonable.

The present study featured a much more comprehensive SNP analysis than any other previous study. At least a considerable part of clinically observed hyperradiosensitivity may be regulated by the integration of multiple minor alterations in genetic function. Therefore, a comprehensive analysis, as was done in the present study, is most suitable for identifying radiosensitive populations of patients. These results may serve as preliminary data for the construction of personalized radiation therapy.

No potential conflicts of interest were disclosed.

We thank the patients for participating in the study and Yusaku Wada, Yuko Kanto, and Kiyoko Ogawa for their technical assistance.