Patients with a high polygenic risk of breast cancer do not have an increased risk of radiotherapy toxicity

Abstract

Purpose

It has been hypothesized that increased predisposition to breast cancer may correlate with radiosensitivity and thus increased risk of toxicity following breast irradiation. This study investigated the relationship between common breast cancer risk variants and radiotherapy toxicity.

Experimental Design

Single nucleotide polymorphism genotypes were determined in female breast cancer patients from the RAPPER (Radiogenomics: Assessment of Polymorphisms for Predicting the Effects of Radiotherapy) study using the Illumina CytoSNP12 genome-wide array. A further 15,582,449 genotypes were imputed using the 1000 Genomes Project reference panel. Patient (n=1160) polygenic risk scores were generated by summing risk-allele dosages, both un-weighted and weighted by published effect sizes for breast cancer risk. Regression models were used to test associations of individual variants and polygenic risk scores with acute and late toxicity phenotypes (telangiectasia, breast edema, photographically assessed shrinkage, induration, pigmentation, breast pain, breast sensitivity, overall toxicity).

Results

Genotypes of 90 confirmed breast cancer risk variants were accurately determined and polygenic risk scores were approximately normally distributed. Variant rs6964587 was associated with increased breast edema five years following radiotherapy (beta=0.22; 95% CI=(0.09-0.34); p-value=7 × 10⁻⁴). No other associations were found between individual variants or the un-weighted (p>0.17) or weighted (p>0.13) polygenic risk score and radiotherapy toxicity. This study had >87% power to detect an association between the polygenic risk score (relative risk>1.1) and toxicity.

Conclusions

Cancer patients with a high polygenic predisposition to breast cancer do not have an increased risk of radiotherapy toxicity up to five years following radiotherapy but individual variants may increase risk.

Introduction

To date, 94 common breast cancer susceptibility loci have been confirmed (p-value<5 × 10⁻⁸) via meta-analyses of 11 genome wide association studies (GWAS; 15,748 cases and 18,084 controls) and 41 studies in the Breast Cancer Association Consortium (BCAC; 46,785 cases and 42,892 controls) (1–3). The effect sizes of each locus are generally modest with odds ratios (ORs) ≤1.34. It is estimated that 16% of the familial risk of breast cancer can be explained by common genetic variation (3). In a recent study, single nucleotide polymorphism (SNP) markers of 77 confirmed breast cancer susceptibility loci were combined to create a polygenic risk score for breast cancer (4). Women in the highest 1% of the risk score were 3.6 times more likely to develop breast cancer than women in the middle quintile (4).

Radiation exposure is a known risk factor for the development of breast cancer (5) and individuals with the disease have been reported to have increased radiosensitivity (6) postulated as due to inherited defects in cellular responses to DNA damage from either endogenous or exogenous sources (7). It has also been suggested that elevated radiosensitivity may be a marker of low penetrance predisposition to cancer (8). Since radiosensitive individuals are more likely to experience toxicity following radiotherapy (9), it has been hypothesized that genetic variants that predispose to breast cancer may also increase the risk of radiation side-effects. This hypothesis was based on the assumption that gene variants involved in cancer pre-disposition may have roles in both tumor formation and in the response to radiation-induced DNA damage.

We have tested this hypothesis using data from the UK RAPPER (Radiogenomics: Assessment of Polymorphisms for Predicting the Effects of Radiotherapy) study (10). Common genetic variants individually have small effects on risk of breast cancer and are likely to have similarly small effects on radiotherapy toxicity. Thus, in order to search for evidence of associations with radiosensitivity, we tested the currently known breast cancer risk variants, both individually and within a more statistically powerful polygenic risk score.

Materials and Methods

Patients

The UK RAPPER (UKCRN1471) study recruits patients from clinical trials and observational studies involving radiotherapy given with curative intent. This study comprised 1160 female breast cancer patients who were included in a GWAS described elsewhere, which provides details on the cohorts included (11): 937 patients from the Cambridge Breast IMRT Trial, 167 from the RACE (Radiation Complications and Epidemiology) study and 56 from a prospective study at the Christie Hospital, Manchester. Briefly, all patients underwent conservative surgery followed by adjuvant radiotherapy with doses of: 40 Gy in 15 fractions over three weeks (Cambridge Breast IMRT trial, Christie Hospital); 50 Gy in 25 fractions or 39.0 Gy/42.9 Gy in 13 fractions over five weeks (RACE study). An alpha/beta ratio of 3.4 (12) was used to calculate equivalent dose in 2Gy fraction for these schedules for inclusion in multivariable analyses. Patient characteristics and treatment factors are summarized in Table 1.

Table 1. Demographic and treatment-related factors of patients in the RAPPER cohort.

Characteristic		RAPPER (N=1160)
Age	N; mean (SD)	1160; 58 (9.2)
	Missing (%)	0
Diabetesa	Yes (%)	42 (3.6)
	No (%)	925 (79.8)
	Missing (%)	193 (16.6)
Smokera	Yes (%)	138 (11.9)
	No (%)	845 (72.8)
	Missing (%)	177 (15.3)
Tumour size (mm)a	N; mean (SD)	968; 15.4 (7.6)
	Missing (%)	192 (16.6)
Adjuvant chemotherapy	Yes (%)	198 (17.1)
	No (%)	935 (80.6)
	Missing (%)	27 (2.3)
Tamoxifen use	Yes (%)	785 (67.7)
	No (%)	334 (28.8)
	Missing (%)	41 (3.5)
Post-operative breast infectiona	Yes (%)	182 (15.7)
	No (%)	789 (68.0)
	Missing (%)	189 (16.3)
Post-operative hematomaa	Yes (%)	69 (6.0)
	No (%)	759 (65.4)
	Missing (%)	332 (28.6)
Breast boost	Yes (%)	735 (63.4)
	No (%)	424 (36.5)
	Missing (%)	1 (0.1)
Actual volume of breast receiving >107% of radiation dosea	N; mean (SD)	932; 19.1 (51.5)
	Missing (%)	228 (19.7)

^aNot available in RACE

Assessment of Radiotherapy Toxicity

Late toxicity data were collected prospectively using standardized scoring systems (Supplementary Table 1). This study used toxicity data collected at two specific time points: two years and five years following radiotherapy. Two-year telangiectasia, breast edema, photographically assessed shrinkage, induration, pigmentation, breast pain and breast sensitivity was recorded for 1134 patients. Five-year telangiectasia, breast edema, photographically assessed shrinkage and induration was collected for 594 patients in the Cambridge Breast IMRT trial. Scale-independent Standardized Total Average Toxicity (STAT) scores were derived from the individual endpoints for overall measures of two-year and five-year toxicity (13). Acute toxicity was recorded for 920 patients in the Cambridge Breast IMRT Trial at 3 weeks following radiotherapy, using the Radiation Therapy Oncology Group (RTOG) scoring system (see Supplementary Table 2).

Genotyping, Quality Control and Imputation

Genotyping and imputation across the whole genome was performed for the RAPPER GWAS. Samples were genotyped using the Illumina CytoSNP12 (Illumina, San Diego, CA, USA). Standard quality control procedures were applied to remove variants that were missing in >5% of samples; had a minor allele frequency (MAF) < 1%; or had MAF<5% and were also missing in >1% samples. Variants were also removed if their genotype frequencies deviated from those expected under Hardy-Weinberg equilibrium (p-value < 10⁻⁶). Samples were removed that had >3% of all variants missing. Principle components analysis (PCA) was used to identify and exclude individuals with non-European ancestry. Genome coverage was increased by imputation using SHAPEIT (14) and IMPUTE2 (15) with the 1000 Genomes Phase III reference panel (16). In this study, expected dosages of the alleles of the 94 breast cancer risk variants were extracted from the imputed data.

Statistical Analysis

Polygenic risk scores were created to quantify the patients’ genetic risk of breast cancer. For each patient, genotype dosages for the 94 breast cancer risk-increasing alleles were calculated and then summed across all the variants. Two types of risk score were calculated:

$$\begin{array}{cc}\text{Un-weighted, for patient}i:\hfill & \mathit{risk}{\mathit{score}}_i={\sum }_1^j{G}_{\mathit{ij}}\hfill \\ \hfill & \hfill \\ \text{weighted, for patient}i:\hfill & \mathit{weighted}\mathit{risk}{\mathit{score}}_i={\sum }_1^j{\beta }_j{G}_{\mathit{ij}}\hfill \end{array}$$

where j = variants 1..94

β_j = the per-allele log-odds ratio for risk of breast cancer associated with variant j

G = risk allele dosage

The log-odds ratios used to weight the risk score were taken directly from the report by Mavaddat et al and were estimated by testing the association of each variant with breast cancer risk while adjusting for the effect of other variants (4). Seventeen variants were identified after Mavaddat et al performed their analysis, so log-odds ratios used for these variants were estimates from BCAC (3).

For univariable analysis, ordinal logistic regression was used to test the association between the acute or individual late toxicity endpoints and polygenic risk score. Linear regression was used to test the association between STAT and risk score. Each genetic variant was also tested individually. For multivariable analysis, a residual analysis was performed (17). Residuals from multivariable analyses of the two-year toxicity endpoints with non-genetic covariates, produced as part of the original RAPPER GWAS of late toxicity (11), were used to analyze two-year toxicity. These residuals estimate the risk of toxicity not explained by available patient- and treatment- related factors including baseline symptoms that resulted from surgery and were present before radiotherapy. Residuals for acute toxicity and five-year toxicity were calculated from multivariable regression models including significant non-genetic factors as part of this study. Linear regression was used to evaluate the association between the residuals and both the genetic risk scores and the individual variants. The association between polygenic risk and baseline symptoms was assessed separately to confirm that adjustment for baseline symptoms would not substantially change the results. For the primary analysis, the polygenic risk scores were considered statistically significant if p-value<0.05. To reduce the chance of a false-positive association due to multiple testing, the individual genetic variants were deemed statistically significant if p-value<5 × 10⁻⁴.

The study was well powered to detect significant associations between breast cancer polygenic risk score and many radiotherapy toxicity endpoints. Assuming a 20% prevalence of moderate/severe toxicity (grade≥2) in breast cancer patients, the power to detect an association between toxicity (grade≥2) and polygenic risk score at p-value<0.05 would be 87% for a small difference in mean risk score of 0.1. This difference in mean risk would be equivalent to a relative risk of toxicity of 1.1 for the subset of patients with a higher mean polygenic risk of breast cancer. All analyses were performed using statistical software package STATA version 13.1. STrengthening the Reporting Of Genetic Association studies in Radiogenomics (STROGAR) guidelines were followed throughout (18).

Results

Out of the 94 genetic variants known to increase the risk of breast cancer, 90 were genotyped or imputed successfully (r2>0.66) in the 1160 patients in the RAPPER cohort – details in Table 2. Figure 1 shows the distributions of the un-weighted and weighted polygenic risk scores among all patients with late toxicity data available (N=1134), which approximate to a normal distribution in these samples.

Table 2. Genetic variants known to influence risk of breast cancer.

Variant	Nearest Gene	Chromf	Position (build 37g)	Breast cancer risk allele	Published odds ratioa	RAPPER risk allele frequency	RAPPER imputation r2c
rs616488	PEX14	1	10566215	A	1.06	0.67	1
rs11552449	PTPN22-BCL2L15-AP4B1-DCLRE1B-HIPK1	1	114448389	T	1.08	0.20	0.98
rs11249433	None	1	121280613	G	1.10	0.38	0.67
rs12405132	RNF115	1	145644984	C	1.05b	0.64	1
rs12048493	OTUD7B	1	149927034	C	1.07b	0.39	0.66
rs6678914	LGR6	1	202187176	G	1.01	0.59	0.99
rs4245739	MDM4	1	204518842	C	1.03	0.28	0.99
rs72755295	EXO1	1	242034263	G	1.15b	0.03	0.69
rs12710696	OSR1	2	19320803	A	1.04	0.38	0.99
rs4849887	INHBB	2	121245122	C	1.09	0.90	0.99
rs2016394	METAP1D-DLX1-DLX2	2	172972971	G	1.05	0.54	1
rs1550623	CDCA7	2	174212894	A	1.06	0.83	0.98
rs1045485	CASP8 d	2	202149589	G	1.04	0.86	0.98
rs13387042	IGFBP5d	2	217905832	A	1.14	0.55	1
rs16857609	DIRC3	2	218296508	T	1.07	0.29	0.98
rs6762644	ITPR1-EGOT	3	4742276	G	1.07	0.40	0.99
rs4973768	SLC4A7	3	27416013	T	1.09	0.50	1
rs12493607	TGFBR2	3	30682939	C	1.05	0.33	0.99
rs6796502	PRSS42	3	46866866	G	1.09b	0.90	0.96
rs1053338	ATXN7	3	63967900	G	1.08b	0.15	0.98
rs9790517	TET2	4	106084778	T	1.05	0.20	0.99
rs6828523	ADAM29	4	175846426	C	1.10	0.89	0.99
rs10069690	TERT d	5	1279790	T	1.02	0.27	1
rs7726159	TERT d	5	1282319	A	1.04	0.33	0.85
rs2736108	TERT d	5	1297488	C	1.07	0.72	0.90
rs13162653	MARCH11	5	16187528	G	1.05b	0.55	0.89
rs2012709	SUB1	5	32567732	T	1.05b	0.49	1
rs10941679	None	5	44706498	G	1.12	0.30	0.98
rs889312	MAP3K1 d	5	56031884	C	1.12	0.31	0.99
rs10472076	RAB3C	5	58184061	C	1.04	0.39	0.96
rs1353747	PDE4D	5	58337481	T	1.09	0.90	0.98
rs7707921	ATG10	5	81538046	A	1.08b	0.76	0.99
rs1432679	EBF1	5	158244083	G	1.07	0.46	0.99
rs11242675	FOXQ1	6	1318878	T	1.06	0.63	0.94
rs204247	RANBP9	6	13722523	G	1.05	0.45	0.99
rs9257408	None	6	28926220	C	1.05b	0.39	0.99
rs17529111	None	6	82128386	G	1.05	0.21	0.99
rs12662670	ESR1 e	6	151918856	G	1.14	0.09	0.99
rs2046210	ESR1 e	6	151948366	A	1.05	0.37	0.99
rs6964587	AKAP9	7	91630620	T	1.05b	0.40	0.98
rs4593472	LINC-PINT	7	130667121	C	1.05b	0.66	0.92
rs720475	ARHGEF5-NOBOX	7	144074929	G	1.06	0.74	0.95
rs9693444	None	8	29509616	A	1.07	0.34	0.99
rs13365225	KCNU1	8	36858483	A	1.05b	0.85	0.98
rs6472903	CASC9	8	76230301	T	1.10	0.84	0.94
rs2943559	HNF4G	8	76417937	G	1.13	0.09	0.98
rs13267382	None	8	117209548	A	1.05b	0.36	0.96
rs13281615	MYC d	8	128355618	G	1.10	0.44	0.99
rs11780156	MYC d	8	129194641	T	1.07	0.17	0.99
rs1011970	CDKN2A/B	9	22062134	T	1.05	0.18	1
rs10759243	KLF4 d	9	110306115	A	1.05	0.29	0.75
rs865686	KLF4 d	9	110888478	T	1.11	0.62	0.99
rs2380205	ANKRD16	10	5886734	C	1.02	0.59	0.99
rs7072776	MLLT10-DNAJC1	10	22032942	A	1.06	0.27	0.91
rs11814448	DNAJC1	10	22315843	C	1.22	0.02	0.84
rs10995190	NRBF2 d	10	64278682	G	1.17	0.87	0.99
rs704010	ZMIZ1	10	80841148	T	1.07	0.43	0.97
rs7904519	TCF7L2	10	114773927	G	1.06	0.48	0.99
rs11199914	FGFR2 d	10	123093901	C	1.06	0.70	0.73
rs2981579	FGFR2 d	10	123337335	A	1.25	0.47	0.99
rs3817198	LSP1	11	1909006	C	1.07	0.33	0.92
rs3903072	DKFZp761e198-OVOLI-SNX32-CFL1-MUS81	11	65583066	G	1.06	0.55	0.99
rs78540526	CCND1 d	11	69331418	T	1.18	0.08	0.95
rs554219	CCND1 d	11	69331642	G	1.12	-	-
rs75915166	CCND1 d	11	69379161	A	1.024	0.06	0.86
rs11820646	BARX2	11	129461171	C	1.05	0.60	0.98
rs12422552	None	12	14413931	C	1.03	-	-
rs10771399	PTHLH	12	28155080	A	1.16	-	-
rs17356907	NTN4	12	96027759	A	1.10	-	-
rs1292011	None	12	115836522	A	1.08	0.59	1
rs11571833	BRCA2-N4BP2LI-N4BP2L2	13	32972626	T	1.26	0.01	0.87
rs2236007	PAX9-SLO25A21	14	37132769	G	1.09	0.82	1
rs2588809	RAD51L1	14	68660428	T	1.07	0.17	0.97
rs999737	RAD51L1	14	69034682	C	1.08	0.77	0.97
rs941764	CCDC88C	14	91841069	G	1.06	0.36	0.95
rs11627032	RIN3	14	93104072	T	1.06b	0.74	0.93
rs3803662	TOX3	16	52586341	A	1.23	0.29	1
rs17817449	MIRI972-2-FTO	16	53813367	T	1.08	0.60	0.99
rs11075995	FTO	16	53855291	T	1.04	0.76	0.99
rs13329835	CDYL2	16	80650805	G	1.08	0.24	0.98
rs146699004	TEFM	17	29230520	GGT	1.08b	0.74	0.99
rs6504950	COX11 e	17	53056471	G	1.07	0.72	0.97
rs745570	CBX8	17	77781725	A	1.05b	0.50	0.97
rs527616	None	18	24337424	G	1.04	0.66	0.98
rs1436904	CHST9	18	24570667	T	1.06	0.63	0.99
rs6507583	SETBP1	18	42399590	A	1.10b	0.93	0.99
rs8170	ABHD8/ANKLE1 e	19	17389704	A	1.03	0.20	0.99
rs2363956	ABHD8/ANKLE1 e	19	17394124	T	1.03	0.51	0.98
rs4808801	SSBP4-ISYNA1-ELL	19	18571141	A	1.07	0.67	0.99
rs3760982	C19orf61-KCNN4-LYPD5-ZNF283	19	44286513	A	1.06	0.49	1
rs2823093	NRIP1	21	16520832	G	1.08	0.75	0.96
rs17879961	CHEK2	22	29121087	G	1.36	0.001	0.69
rs132390	EMID1-RHBDD3-EWSR1	22	29621477	C	1.11	0.04	1
rs6001930	MKL1	22	40876234	C	1.13	0.11	0.99

^aadjusted odds ratios from Mavaddat et al. 2015

^bunadjusted odds ratios from Michailidou et al. 2013 or Michailidou et al. 2015

^cimputation r² calculated by the mean info score from IMPUTE2. The info score represents the certainty with which the SNP has been imputed and lies between 0 (no certainty) and 1 (high certainty; r²=1 for genotyped SNPs)

^dpublished target gene

^eknown target gene, not yet published

^fchromosome

^ghuman genome build 37 (GRCh37) by the Genome Reference Consortium, released 3 March 2009. Available from: http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/

Figure 1.

Distribution of polygenic risk scores in the 1134 RAPPER patients for whom late radiotherapy toxicity data were available

A) Un-weighted polygenic risk score

B) Weighted polygenic risk score

The distributions of the nine different acute and late-arising toxicity endpoints (acute toxicity, telangiectasia, breast edema, photographically assessed shrinkage, induration, pigmentation, breast pain, breast sensitivity and a standardized total average toxicity (STAT) score) measured in this study are shown in Table 3. Baseline symptoms that resulted from surgery and were present before radiotherapy were not associated with any of the genetic variants either individually or in the polygenic risk score. The significant non-genetic factors associated with acute toxicity and the five-year toxicity endpoints are shown in Supplementary Table 3.

Table 3. Distributions of toxicity endpoints in RAPPER patient sample.

	Number of patients (%)
Acute toxicity	Grade 0	Grade 1	Grade 2a	Grade 2b	Grade 3
RTOG acute toxicity	30 (3.3)	549 (59.7)	328 (35.7)	12 (1.3)	1 (0.1)
2 year toxicity	Grade 0	Grade 1	Grade 2	Grade 3	Grade 4
Telangiectasia	738 (79.7)	116 (12.5)	53 (5.7)	18 (1.9)	1 (0.1)
Breast edema	522 (56.4)	261 (28.2)	110 (11.9)	33 (3.6)	0
Shrinkage assessed photographically	NA	593 (63.5)	285 (30.5)	56 (6.0)	NA
Induration	177 (18.3)	444 (45.8)	276 (28.5)	70 (7.2)	3 (0.3)
Pigmentation	629 (83.0)	62 (8.2)	67 (8.8)	0	0
Breast pain	NA	416 (49.9)	362 (43.4)	45 (5.4)	11 (1.3)
Breast sensitivity	NA	507 (60.7)	278 (33.3)	41 (4.9)	10 (1.2)
Overall toxicity (STATa)	N=1134, mean (standard deviation) = 0.003 (0.66)
5 year toxicity	Grade 0	Grade 1	Grade 2	Grade 3	Grade 4
Telangiectasia	493 (83.1)	48 (8.1)	30 (5.1)	22 (3.7)	0
Breastedema	487 (82.1)	73 (12.3)	27 (4.6)	6 (1.1)	0
Shrinkage assessed photographically	NA	337 (58.7)	201 (35.0)	36 (6.3)	NA
Induration	174 (29.3)	267 (45.0)	130 (21.9)	22 (3.7)	0
Overall toxicity (STATa)	N=593, mean (standard deviation) = 0.01 (0.64)

^aStandardised total average toxicity

None of the individual genetic variants were associated, at the pre-specified significance level of p<5 × 10⁻⁴, with acute toxicity or two-year toxicity in either univariable or multivariable analyses (Supplementary Table 4, Supplementary Figures 1-9). Univariable analysis of five-year toxicity identified one SNP, rs6964587, to be significantly associated with increased risk of breast edema five years following radiotherapy (OR=1.75, 95%CI=(1.30-2.49), p-value=4 × 10⁻⁵). The association remained strong after adjustment for non-genetic factors (residual beta=0.22, 95% CI=(0.09-0.34), p-value=7 × 10⁻⁴). No other individual variants were associated with five-year toxicity on univariable or multivariable analysis (Supplementary Table 5).

Scatter plots comparing each variant’s effect on breast cancer risk with their effects on STAT (overall toxicity score) at two years and five years are shown in Figure 2. Scatter plots for all other toxicity endpoints can be found in Supplementary Figures 10-17. Points in the top right hand quarters of the plots represent alleles associated with both increased breast cancer and radiation toxicity risk. Conversely for points in the bottom right hand quarter, the allelic effects are opposite for breast cancer and radiation toxicity risk. The plots reveal no correlations between these confirmed breast cancer risk alleles and increased radiotherapy toxicity risk in this RAPPER cohort.

Figure 2.

Scatter plot of individual breast cancer risk variants comparing their effect on breast cancer risk with their effect on STAT (Standardized Total Average Toxicity) score

A) STAT at two years

B) STAT at five years

When the more statistically powerful polygenic risk scores were tested, there were no associations with acute toxicity (weighted p-values>0.30; un-weighted p-values>0.39) or any of the two-year toxicity endpoints (weighted p-values ≥0.13; un-weighted p-values≥0.17) or five-year endpoints (weighted p-values≥0.19; un-weighted p-values≥0.39) (Table 4).

Table 4. Association of risk scores with toxicity endpoints from residual analyses.

Toxicity	betaa (95% confidence interval) & p-value
	Un-weighted risk score	Weighted risk score
Acute toxicity
RTOG grade	−0.01 (−0.02-0.01) p=0.39	−0.08 (−0.23-0.07) p=0.30
2 year toxicity
Telangiectasia	−0.01 (−0.02-0.01) p=0.40	−0.05 (−0.20-0.10) p=0.54
Breast edema	0.001 (−0.01-0.01) p=0.80	0.02 (−0.13-0.17) p=0.80
Shrinkage assessed photographically	−0.01 (−0.02-0.01) p=0.41	−0.08 (−0.22-0.07) p=0.30
Induration	−0.003 (−0.01-0.01) p=0.64	−0.05 (−0.20-0.10) p=0.51
Pigmentation	0.01 (−0.004-0.02) p=0.17	0.13 (−0.04-0.29) p=0.13
Breast pain	0.003 (−0.01-0.01) p=0.67	0.01 (−0.14-0.17) p=0.85
Oversensitivity	−0.001 (−0.01-0.01) p=0.85	−0.05 (−0.21-0.10) p=0.49
Overall toxicity (STATb)	0.002 (−0.01-0.01) p=0.72	0.02 (−0.13-0.16) p=0.81
5 year toxicity
Telangiectasia	−0.0004 (−0.02-0.01) p=0.96	−0.05 (−0.24-0.14) p=0.61
Breast edema	−0.003 (−0.02-0.01) p=0.72	−0.05 (−0.23-0.13) p=0.58
Shrinkage assessed photographically	0.001 (−0.01-0.02) p=0.92	0.05 (−0.14-0.25) p=0.58
Induration	−0.01 (−0.03-0.01) p=0.19	−0.09 (−0.28-0.11) p=0.39
Overall toxicity (STATb)	−0.01 (−0.02-0.01) p=0.47	−0.07 (−0.26-0.12) p=0.48

^abeta estimates are for the association between polygenic risk score and change in residual toxicity; a positive beta represents an increased risk of toxicity and a negative beta represents reduced risk of toxicity

^bStandardised total average toxicity

Discussion

The hypothesis behind this study was that common genetic variants known to increase the risk of breast cancer may also increase the likelihood of developing toxicity following radiotherapy for breast cancer. We found no conclusive evidence supporting this hypothesis. One SNP, rs6964587, was identified as associated with increased breast edema at 5 years (OR=1.75; 95% CI=(1.30-2.49)). This SNP is a missense SNP lying in a coding region of the AKAP9 gene, which encodes the A kinase (PRKA) anchor protein 9 - a key component of signal transduction that contributes to carcinogenesis (2,19,20). The SNP is predicted to generate a M643I substitution in the AKAP9 protein sequence. Of note, a study by Han et al suggested that breast tumor cells secrete the AKAP9 protein in response to ionizing radiation (20). This SNP association merits further investigation in independent radiotherapy toxicity cohorts.

The nearest gene to each individual breast cancer risk locus studied is listed in Table 2. For some of these, (marked with ^d or ^e), the target gene and mode of action of the causal variant has been elucidated, but for the majority this process is still on-going. A few putative target breast cancer genes on the list (for example, CHEK2 and RAD51b) have known roles in DNA damage repair, but it is unlikely that the majority will prove to have similar functions. In this study of common low-penetrance breast cancer loci, the rare, more highly penetrant mutations in genes such as ATM, BRCA1, BRCA2 and TP53 were not genotyped and could not be included in this analysis. Investigation of these genes requires complete sequencing of the relevant genes and different statistical methods and is on-going. The link between radiosensitivity and cancer risk has been exemplified by ATM gene mutations, causing ataxia-telangiectasia, a cancer predisposing immunodeficiency syndrome characterized by extreme sensitivity to radiation and life-threatening radiotherapy toxicity (21). Carriers of ATM gene mutations, including relatives of people with ataxia-telangiectasia, have an increased risk of breast cancer(22). An in vitro radiosensitivity assay revealed that around 40% of breast cancer patients tested displayed evidence of increased radiosensitivity (7,8,23). Another study demonstrated the heritability of enhanced chromosomal radiosensitivity in breast cancer families (6).

This RAPPER study had a relatively small sample size. The original studies confirming the associations of the 94 genetic variants with increased breast cancer risk were performed in BCAC with over 100,000 study participants. In contrast, the RAPPER breast cohort comprises just fewer than 1,200 patients and the study had low power to detect a true association between individual variants and radiotherapy toxicity. Despite this, a single variant associated with an ~75% increased odds of breast edema five years following radiotherapy has been detected at a stringent statistical significance threshold. Future larger studies with greater power will be able to identify other individual SNPs that may be associated with radiotherapy toxicity. These studies could also explore whether the effect sizes of individual variants on breast cancer risk differ from those for radiation toxicity risk. Although this study was generally underpowered to detect individual variants, the study was well powered (87%) to detect an association between radiotherapy toxicity and a polygenic risk score. The 90 confirmed breast cancer variants, used in this study, have been estimated to explain 16% of the familial risk of breast cancer (3) and an earlier polygenic risk profile, which included a subset of 77 of these SNPs, demonstrated that women in the highest 1% of the risk score were 3.6 times more likely to develop breast cancer than women in the middle quintile (4). We have extended the polygenic risk profile to include 90 currently assayable breast cancer risk loci and have shown that they have no detectable predictive value for risk of radiation toxicity in women being treated for breast cancer. Our analysis was limited to women of European ancestry only and therefore the findings may not be generalizable to women of other ancestries.

In summary, our study found no evidence to support a suggestion that patients with a high polygenic risk of breast cancer have an increased likelihood of developing acute or late radiotherapy toxicity. The conclusion from this work is that it appears to be safe for breast cancer patients with a high polygenic risk of breast cancer to receive standard radiotherapy for breast cancer.

Translational Statement.

Common germline genetic variation impacts on the development of cancer and 94 common variants have now been identified, which together explain almost 20% of the variance in breast cancer risk. Up to 75% of breast cancer patients are likely to have radiotherapy and, depending on the country of treatment, up to 80% of patients survive breast cancer. For over 20 years clinicians have been concerned that the biological mechanisms underlying genetic predisposition to breast cancer risk may also increase radiosensitivity, such that some patients receiving breast tumor radiotherapy may be at increased risk of normal tissue toxicity. We have tested the breast cancer risk variants, both individually and as a multi-SNP risk profile, for association with acute and late radiotherapy toxicity phenotypes in 1,160 breast cancer patients. Our study showed that those with a high polygenic risk of breast cancer do not have increased risk of radiotherapy toxicity.