FGFR2 genotype and risk of radiation-associated breast cancer in Hodgkin lymphoma

Radiotherapy and combination chemotherapy have dramatically improved the prognosis of Hodgkin lymphoma (HL), but at the price of raised risks of secondary malignancy. For women treated with supradiaphragmatic radiotherapy at young ages, breast cancer is a particular hazard^1-3 

It has long been postulated that genetically determined responses may affect the severity of late complications after radiotherapy. Certain rare inherited cancer syndromes are typified by radiosensitivity⁴  and cancer patients and some of their first-degree relatives show greater in vitro radiosensitivity than healthy individuals.⁵ 

Genome-wide association studies (GWAS) have identified 14 independent common single nucleotide polymorphisms (SNPs) that show consistent associations with female breast cancer risk.^6-13  To examine whether these variants contribute to breast cancer risk after radiotherapy, we analyzed blood samples from 2 case-control series. The discovery phase comprised 449 women with HL treated with supradiaphragmatic radiotherapy in England and Wales during 1963-2003 at ages < 36 years: 140 had breast cancer after HL treatment (the “cases”) and 309 had had no solid cancer after HL (the “controls”). The replication series was 244 female Dutch HL patients treated with supradiaphragmatic radiotherapy during 1965-1997 at ages < 41 years: 92 cases and 152 controls. Details are given in supplemental Methods and supplemental Table 1 (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

DNA was extracted from EDTA-venous blood samples using conventional methodologies and quantified using PicoGreen (Invitrogen). To evaluate the impact of variation at the 14 breast cancer risk loci (detailed in supplemental Methods) on breast cancer risk in HL patients, we derived genotypes at these loci for the discovery phase patients using previously generated HL GWAS data¹⁴  using Illumina Infinium HD Human660-Quad BeadChips. Replication genotyping was performed using KasPar allele-specific PCR (KBioscience) and analyzed in an Applied Biosystem ABI7900HT system.

The relative risk of breast cancer associated with SNP genotype was estimated by calculating odds ratios (ORs). All statistical tests were 2-sided. After analyzing the United Kingdom and Dutch datasets separately, we pooled them to increase power. Meta-analysis was performed under a fixed-effects model, estimating the Cochran Q statistic to test for heterogeneity and the I² statistic to quantify variation between studies.

SNP genotypes were obtained for all 14 SNPs for > 95% of patients in the discovery phase (supplemental Table 2). Minor allele frequencies of the 14 SNPs in the HL patients were similar to those in 2 published population control sets (supplemental Table 2), providing no evidence for an association between genetic variation at any of the 14 loci and HL risk (ie, P > .05).

rs1219648 genotype frequencies were significantly different between cases and controls. Specifically, in the discovery series there was an overrepresentation of the minor, G allele, in HL patients with breast cancer (OR = 1.73; P = 000273, Table 1). This association was dose-dependent, with highest risks in patients homozygous for the G allele. The replication series provided additional support for this association, albeit nonsignificantly (Table 1), and in pooled data from both series there was a highly significant relation (per-allele OR = 1.59, 95% confidence interval [CI]: 1.26-2.02, P = .000111, P_het = 0.35, I² = 0%; Table 1), remaining almost unchanged in analyses restricted to invasive breast cancers (ie, excluding cases of ductal carcinoma in situ: n = 28) and remaining significant after Bonferroni correction for multiple SNP testing. The OR is larger than that for breast cancer in relation to FGFR2 genotype in the general population (1.26 per allele¹² ). The effect was somewhat greater in patients first treated before age 20 (OR = 1.70 [1.16-2.50]) than those treated at older ages (OR = 1.48 [1.09-2.00]), and in patients who had not received alkylating chemotherapy or ≥ 5-Gy pelvic radiotherapy (OR = 1.79 [1.25-2.57]), than those who had received either of these (OR = 1.45 [1.00-2.11]; not in table).

rs1219648 maps to a 25-kb region of linkage disequilibrium within the second intron and containing exon 2 of the FGFR2 gene. FGFR2 is overexpressed in estrogen receptor–positive tumors,¹⁵ FGFR2 signaling having oncogenic effects.¹⁶  The functional basis of the rs1219648 association with breast cancer appears to be through allele-specific up-regulation of FGFR2, thereby increasing the propensity for tumor formation.¹⁷  While one study has shown radiosensitizing effects of differential FGFR2 expression in a human prostate cancer cell line,¹⁸  the effect in breast cancer awaits study.

Because of the large risks of breast cancer occurring after supradiaphragmatic radiotherapy for HL (cumulative risks of 25% or more at 25-30 years follow-up³ ), specific follow-up clinics have been established to counsel patients.^19,20  Much is known about radiotherapy-related factors that increase risks, and ovarian-toxic treatments that decrease them,^1,3  but it is unknown whether genetic factors affect these risks further. It has been suggested that radiotherapy might constitute a particular hazard for ataxia-telangiectasia carriers,²¹  but studies have been small and not shown any clear association.^22,23  Family history of breast or ovarian cancer has not been found associated with risk, based on small numbers.^24,25 

Our data provide evidence that genetic variation in FGFR2 influences breast cancer risk in HL patients treated with radiotherapy, especially those treated at young ages and those not treated with ovarian-toxic agents: the groups with the greatest treatment-related risks. This information should improve individualization of advice to HL patients on their risks, and decisions on prophylactic mastectomy and screening regimens.

The United Kingdom case-control study thanks the study participants, study staff, and clinicians who participated in the study as listed. We are indebted to the patients and physicians who participated in the Dutch data and sample collection.

This work was supported by Leukaemia & Lymphoma Research, Cancer Research UK, and the Bobby Moore Fund (C1298/A8362, R.S.H.) for genetic analyses; Breakthrough Breast Cancer and the European Union for United Kingdom sample and data acquisition; and the Dutch Cancer Society (NKI 2001-2425 and NKI 2004-3068, F.E.v.L.) for Dutch sample and data acquisition.

Contribution: Y.P.M., R.S.H., and A.J.S. drafted the manuscript; Y.P.M. and V.E.-M. performed statistical and bioinformatic analyses; F.E.v.L. designed the Dutch National Cancer Institute (NKI) study and obtained financial support; R.C., C.J., A.A., and A.J.S. provided samples and data from a study conducted at the Institute of Cancer Research; A.B. coordinated collection and preparation of NKI samples; B.O. and A.L. performed genotyping; P.B. was responsible for sample coordination and laboratory analyses; N.S.R. and C.J. were involved in identification and inclusion of Dutch cases, study design, review board approval, and clinical implementation; R.S.H. designed the study and obtained financial support; and all authors contributed to the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Prof Anthony Swerdlow, DSc, Section of Epidemiology, Sir Richard Doll Building, Institute of Cancer Research, 15 Cotswold Road, Sutton, Surrey SM2 5NG, United Kingdom; e-mail: anthony.swerdlow@icr.ac.uk.