Common variation in BRCA1 may play a role in progression to lethal prostate cancer after radiation treatment

Abstract

Purpose

To evaluate if single-nucleotide polymorphisms (SNPs) reflecting common variation in the tumor suppressor BRCA1 affects prostate cancer (CaP) outcomes. Because radiation therapy (RT) induces DNA damage, we hypothesized that common variation in BRCA1 plays a role in progression to lethal CaP, particularly in patients receiving RT.

Methods

We followed 802 men diagnosed with localized CaP (cT1-T3/N0/M0) who were treated with RT in the US Health Professionals Follow-up Study (HPFS) and Physicians’ Health Study (PHS), for progression to lethal CaP. Six single-nucleotide polymorphisms (rs3737559, rs1799950, rs799923, rs915945, rs4474733, and rs8176305) were genotyped in HPFS to capture common variation across BRCA1. rs4474733 and rs8176305 were also evaluated in the PHS cohort. Cox proportional hazards models were used to estimate per-allele hazard ratios (HR) and 95% confidence intervals (CI) stratified by primary treatment.

Results

In the RT group (n=802) 71 men progressed to lethal disease during a mean follow-up of 12 years. We found that two SNPs, rs4473733 (HR: 0.65; 95%CI 0.42-0.99) and rs8176305 (HR: 2.03; 95%CI 1.33-3.10), were associated with lethal CaP in men receiving RT.

Conclusions

Common variation in BRCA1 may influence clinical outcomes in patients receiving RT for localized CaP by modifying the response to RT. Our findings merit further follow-up studies to validate these SNPs and better understand their functional and biological significance.

Introduction

Radiation therapy is a frequently used and often-effective treatment for localized prostate cancer with comparable oncologic outcomes to radical prostatectomy(1, 2); however, current nomograms based on clinical information to direct treatment choices are suboptimal. Furthermore, tumors may have heterogeneity with respect to radiation sensitivity that could affect radiotherapy efficacy. As such, there is a pressing need to discover biomarkers that can predict treatment-specific response.

Recently, genome-wide association studies in prostate cancer patients have identified single-nucleotide polymorphisms (SNPs), in candidate genes that may help predict toxicity to radiation therapy.(3-7) Furthermore, SNPs in genes responsible for DNA damage repair and cell cycle control may play an important role in affecting cell and tissue response to radiation therapy.(8) Among these, BRCA1 is of particular interest given its major role in DNA damage repair and cell cycle control with possible implications for treatment response in prostate cancer patients receiving radiation therapy.(9-11) Furthermore, our group previously demonstrated that patients with higher tumor BRCA1 protein expression had an increased risk of prostate cancer mortality.(12)

Men with rare germline mutations in BRCA1 have a higher risk of prostate cancer and poorer survival outcomes.(13) However, few studies have examined the association of common genetic variation (e.g. SNPs with minor allele frequency (MAF) ≥ 0.05) in these DNA repair genes and prostate cancer outcomes after radiation therapy. One study in a cohort of high-risk prostate cancer families demonstrated that the minor allele of rs1799950, conferring a Gln³⁵⁶Arg amino acid change, was associated with an increased risk of developing prostate cancer (OR 2.25, 95%CI 1.21-4.20). (14) It remains unclear whether common variation in BRCA1 is associated with prostate cancer progression and response to radiation therapy.

We hypothesized that due to the relevance of the DNA repair pathway and cell cycle control with radiosensitivity, inherited common variation in BRCA1 could affect prostate cancer outcomes in men who were treated with radiation therapy.

Materials and methods

Study population

This study includes men with prostate cancer in two prospective male cohorts: the Health Professionals Follow-Up Study (HPFS) and the Physicians’ Health Study (PHS). The HPFS prospectively enrolled 51,529 male medical professionals in 1986 ages 40 to 75 years to investigate the causes of cancer and heart disease. Participants filled out baseline questionnaires with information on age, marital status, height and weight, ancestry, medications, disease history, physical activity, and diet which has been described previously.(15) Men were subsequently followed with biennial questionnaires to collect information on medical diagnoses and life style factors. Between 1993 and 1995, 18,018 HPFS participants provided a blood specimen. PHS began as a randomized, double-blind, placebo-controlled trial of aspirin and beta-carotene for the primary prevention of cardiovascular disease and cancer; in 1982, 22,071 physicians aged 40-84 were enrolled and 14,916 men provided blood samples.(16) Men were subsequently followed with annual questionnaires to collect information on medical diagnoses. The institutional review boards approved these studies.

Prostate cancer identification and follow-up

Men with incident prostate cancer were initially identified based on self-report through questionnaires, followed by review of the medical records and pathology reports to confirm the cancer diagnosis. Men with prostate cancer were followed prospectively for treatment, disease progression, and mortality. Cases in HPFS were identified from 1986 through 2012 and followed for progression through 2012. Cases in PHS were identified in 1982 through 2011 and followed for progression through 2011. For the primary analysis, we included men with an available blood specimen who were diagnosed with incident clinically localized (cT1-3N0/M0) prostate cancer and received radiation therapy (n=802). Furthermore, to determine whether the association was specific to those treated with radiation therapy, we also evaluated the association in men with incident localized prostate cancer who received radical prostatectomy (n=1111) as their primary treatment as a comparison group. To reduce the possibility of population stratification we restricted the analysis to men of European descent.

Outcomes measured

The primary outcome of this study was progression to lethal prostate cancer, defined as development of distant metastases or prostate cancer-specific mortality. For men in the HPFS cohort, we also evaluated the secondary outcome of biochemical recurrence. Due to a lack of complete PSA data in the PHS cohort, we did not evaluate this outcome in that cohort. Biochemical recurrence was defined differently for each treatment modality (e.g., radical prostatectomy: PSA higher than 2 mg/mL after surgery for at least two consecutive measures; radiation: an increase of 2 or more ng/mL higher than the nadir PSA; brachytherapy: an increase of 1 or more ng/mL higher than the nadir PSA for at least 2 consecutive measures).(17-19) For the primary outcome and analysis, HPFS and PHS patient cohorts were pooled and the results presented in the study reflect the pooled cohort unless otherwise stated.

Genotype assessment

Six haplotype-tagging SNPs (rs3737559, rs1799950, rs799923, rs915945, rs4474733, and rs8176305) were identified and initially genotyped in the HPFS cohort to provide coverage of common variation (minor allele frequency (MAF) ≥ 5%) with R² > 0.80 across BRCA1 plus areas 10kb up and down-stream (based on HapMap Phase II CEU samples). Of the six SNPs selected, three (rs3737559, rs1799950, and rs799923) were previously studied in the literature (Supplemental Table 1). Genotyping was performed in a blinded fashion after extracting germline DNA from blood using standard QIAamp kit (QIAGEN Inc., Chatsworth, CA, USA) protocol. Four SNPs were genotyped using the Open Array SNP Genotyping Platform (Applied Biosystems, Carlsbad, CA, USA) and 2 were genotyped using TaqMan (Applied Biosystems, Carlsbad, CA USA).

After initial analysis of the 6 SNPs genotyped in the HPFS cohort, we also genotyped two SNPs (rs4474733 and rs8176305) in the PHS cohort based on the findings in the HPFS analysis to improve the study's power. The PHS samples were genotyped using TaqMan assay (Applied Biosystems). All SNPs were in HWE (p>0.0001), had greater than 97% completion, and concordance was 100% for blinded quality control samples.

Statistical analysis

In this case-only survival analysis, we used Cox proportional hazards models and assumed an additive genetic model to calculate per-allele hazard ratios (HR) and 95% confidence intervals (CI) for the association of each SNP (reference group is the major allele) and progression to lethal prostate cancer. Our primary model adjusted for age at diagnosis. We considered clinical factors such as Gleason grade and stage as potential intermediates along the pathway and therefore did not adjust for them in the primary model. However, we did conduct additional adjustment of Gleason score (≤ 6, 7, ≥ 8), PSA at diagnosis (< 10, 10-20, > 20), clinical stage (cT1/T2 versus cT3) in secondary models. For HPFS we could not adjust for clinical stage due to few lethal events in clinical stage T3. We also adjusted for primary treatment and radiation therapy type when relevant in the secondary models. To assess whether the association of the SNPs and prostate cancer progression was specific to radiation therapy, we conducted analyses stratified by primary treatment (radiation therapy vs. prostatectomy). We assessed statistical interaction by treatment using the Wald test by adding a multiplicative interaction term (treatment*SNP) into our models. Among men in the radiation group we also performed an exploratory analysis in which men were stratified by type of radiation therapy (brachytherapy vs. external beam radiation therapy) and receipt of androgen deprivation therapy to assess whether these factors modified the associations between SNPs and prostate cancer progression. We present nominal two-sided p-values without adjusting for multiple testing; however, a p-value significant threshold of 0.008 would control the experiment-wide Type 1 error rate at 0.05. SAS version 9.3 (SAS institute Inc, Cary, North Carolina) was used for all analysis.

Results

Table 1 describes the clinical characteristics of prostate cancer cases in our analysis by treatment status and cohort. In total, 802 men received radiation therapy as their primary treatment (447 in HPFS and 355 in PHS). In total, 1111 men had a radical prostatectomy as their primary treatment (543 in HPFS and 568 in PHS). The majority of participants in the RT and RP cohorts had a PSA < 10 at diagnosis, clinical stage T1/T2 disease, and Gleason ≤ 7 disease. Patients who received radiation therapy were more likely to be older in age, have Gleason ≤ 6, and receive androgen deprivation therapy compared to those treated with radical prostatectomy. During follow-up, 71 (9%) men progressed to lethal prostate cancer after radiation therapy and 70 (6%) men progressed after radical prostatectomy.

Table 1.

Patient characteristics at diagnosis in the HPFS, PHS, and pooled cohorts by primary treatment.

	Rad iation Therapy (RT)			Radical Prostatectomy (RP)
Characteristics at presentation	HPFS (n=447)	PHS (n=335)	Pooled (n=802)	HPFS (n=543)	PHS (n=568)	Pooled (n=1111)
Age (yrs), mean (SD)	71.9 (5.9)	72.4 (6.0)	72.1 (5.9)	65.1 (6.3)	66.0 (5.9)	65.6 (6.1)
Follow-up time (yrs), mean (SD)	12.0 (3.3)	11.1 (4.6)	11.6 (3.9)	12.9 (3.6)	13.7 (5.0)	13.3 (4.4)
Clinical stage, %*
T1/T2	98	92	95	98	97	98
T3	2	8	5	2	3	2
Gleason score, %*
≤ 6	67	60	64	40	47	43
7	23	27	25	50	41	45
≥ 8	10	13	11	10	12	11
PSA (ng/mL), %***
< 10	71	69	70	81	72	77
10 to 20	22	20	21	15	19	17
> 20	6	11	9	4	8	6
Neoadjuvant/adjuvant ADT, %	42	22	33	11	8	9

Percentages may not add up to 100% due to rounding

^*82 patients in PHS had missing clinical stage

** 8 patients in HPFS and 58 patients in PHS had missing Gleason score

^***51 patients in HPFS and 133 patients in PHS had missing PSA (includes those diagnosed i n the pre-PSA era)

Lethal = distant metastases and death

Abbreviations: RT = radiation therapy, RP = radical prostatectomy, and ADT = androgen deprivation therapy

^# difference between PHS and HPFS cohorts significant (p<0.05)

Table 2 shows the treatment-stratified associations of each SNP and progression to lethal prostate cancer. Two SNPs, rs4474733 and rs8176305 were nominally associated (p<0.05) with lethal prostate cancer in men who received radiation therapy. The minor allele of rs4474733 was associated with a decreased risk of lethal prostate cancer (HR: 0.65; 95%CI 0.42-0.99; p = 0.05) in men who received radiation therapy while there was no statistically significant association in men who received radical prostatectomy (HR 0.95; 95%CI 0.64-1.40; p=0.79); p-interaction=0.17. The minor allele of rs81776305 was associated with a higher risk of lethal prostate cancer in patients receiving radiation therapy (HR: 2.03; 95%CI 1.33-3.10; p=0.001) while there was no significant association in patients receiving radical prostatectomy (HR 0.82; 95%CI 0.44-1.53; p=0.53); the p-interaction=0.02. Adjusting for Gleason grade, PSA at diagnosis, and clinical stage resulted in some attenuation in point estimates suggesting that these clinical characteristics could be intermediates along the SNP-prostate cancer progression pathway (Supplemental Table 2). Results for biochemical recurrence in the HPFS cohort were consistent with our findings for lethal prostate cancer (Supplemental Table 3). Among men receiving radiation therapy, the minor allele of rs8176305 was associated with an increased risk of biochemical recurrence (HR: 1.58; 95%CI 1.06-2.38; p=0.03) and the minor allele of rs44474733 was associated with a decreased risk of biochemical recurrence (HR: 0.68; 95%CI 0.48-0.95; p=0.02). No association was seen in the prostatectomy cohort (Supplemental Table 3). Complete biochemical recurrence data were unavailable for the PHS cohort.

Table 2.

Age-adjusted per-allele hazard-ratios (HRs) for progression to lethal prostate cancer stratified by treatment.

			Treatment Stratified
			Radiation Therapy			Radical Prostatectomy
	SNP	MAF	n (lethal)	HR (95% CI)	p-value	n (lethal)	HR (95% CI)	p-value	p-interaction*
Cohort
HPFS	rs3737559 (C,T)	0.09	32	1.09 (0.47 - 2.52)	0.84	29	0.95 (0.38 - 2.35)	0.90	0.81
	rs1799950 (T,C)	0.06		1.75 (0.75 - 4.06)	0.19		1.77 (0.69 - 4.52)	0.23	0.97
	rs799923 (G,A)	0.23		1.05 (0.59 - 1.85)	0.88		1.31 (0.76 - 2.28)	0.33	0.54
	rs915945 (C,T)	0.36		1.29 (0.80 - 2.08)	0.29		0.87 (0.52 - 1.48)	0.61	0.28
	rs4474733 (G,T)	0.26		0.65 (0.34 - 1.22)	0.18		1.10 (0.62 - 1.94)	0.75	0.23
	rs8176305 (A,G)	0.09		3.55 (2.04 - 6.17)	< 0.001		0.36 (0.09 - 1.45)	0.15	0.003
PHS	rs4474733 (G,T)	0.23	39	0.67 (0.3 7 - 1.20)	0.18	41	0.87 (0.50 - 1.49)	0.60	0.44
	rs8176305 (A,G)	0.08		1.19 (0.59 - 2.39)	0.62		1.23 (0.60 - 2.51)	0.58	0.94
Pooled	rs4474733 (G,T)	0.25	71	0.65 (0.4 2 - 0.99)	0.05	70	0.95 (0.64 - 1.40)	0.79	0.17
	rs8176305 (A,G)	0.09		2.03 (1.33 - 3.10)	0.001		0.82 (0.44 - 1.53)	0.53	0.02

^*p-interaction between treatment and SNP

Abbreviations: MAF = minor allele frequency, CI = confidence interval

In our exploratory analyses among the radiation cohort, there was no significant evidence of modification of the SNP-lethal prostate cancer relationship by receipt of ADT or specific radiation therapy type; however, estimates were unstable due to the small number of events (Table 3).

Table 3.

Age-adjusted per-allele hazard ratios (HRs) for progression to lethal prostate cancer in patients receiving radiation therapy stratified by type of radiation and ADT status.

		rs4474733 (G,T)		rs8176305 (A,G)
	n	HR (95% CI)	p-interaction**	HR (95% CI)	p-interaction**
All radiation	71	0.65 (0.42 - 0.99)	n/a	2.03 (1.33 - 3.10)	n/a
Radiation type*
Brachytherapy	9	0.33 (0.08 - 1.41)	0.34	4.40 (1.43 - 13.53)	0.09
EBRT	58	0.70 (0.44 - 1.13)		1.69 (1.05 - 2.70)
ADT status^
ADT	13	0.50 (0.15 - 1.67)	0.73	3.02 (1.26 - 7.22)	0.29
No ADT	58	0.65 (0.41 - 1.03)		1.83 (1.12 - 2.98)

Model controls for age and brachytherapy.

^*Number of brachytherapy plus EBRT events do not equal those of radiation therapy due to missing documentation of radiation type

^**Interaction between ADT status and treatment type

^{^}Adjuvant/Neoadjuvant ADT

Abbreviations: EBRT = external beam radiation therapy, ADT = androgen deprivation therapy, n/a = not applicable.

Discussion

Consistent with our hypothesis, we observed nominal associations between two BRCA1 SNPs and prostate cancer progression in patients treated with radiation therapy and no association in those treated with radical prostatectomy. The minor allele of rs4474733 was associated with a 35% decreased risk of lethal prostate cancer after radiation therapy and the minor allele of rs8176305 was associated with a two-fold increased risk of lethal prostate cancer after radiation therapy. Moreover, the findings were consistent with respect to biochemical recurrence. Our study requires validation, but provides novel evidence that common genetic variation in BRCA1 may play a role in prostate cancer progression following radiation therapy.

To our knowledge, our study is the first to report on treatment specific association of prostate cancer progression and common variation in BRCA1. Previous studies have focused on rare germline mutations in BRCA1 and prostate cancer incidence and progression.(13, 20) Recently, a study of 1302 patients with non-metastatic prostate cancer treated with radiation therapy (n=767) or radical prostatectomy (n=535) found that those with germline BRCA mutations (n=67) had increased disease-specific mortality. No modification was observed by treatment type, but numbers were limited for this analysis; of 67 BRCA mutation carriers, 18 had BRCA1 germline mutations.(21)

One study has evaluated common variation in BRCA1 and prostate cancer incidence in a cohort of patients with a strong family history of prostate cancer and observed linkage on chromosome 17q.(22) Three SNPs (rs3737559, rs1799950, and rs799923) were found to be associated with prostate cancer risk. The strongest linkage was seen for rs1799950 conferring a Gln³⁵⁶Arg missense mutation with unclear clinical significance.(14) This study did not assess prostate cancer mortality.. Few GWAS studies have assessed prostate cancer mortality (23, 24) and no GWAS studies have assessed treatment specific outcomes for prostate cancer.

BRCA1 plays a primary role in DNA damage repair and cell cycle control and has been shown to have an important role in radiosensitivity in prostate cancer cell lines in vitro. (25) Alternatively, SNPs in other DNA-repair genes have also been associated with an increased risk of radiation-induced toxicity in patients treated for prostate cancer.(26) However, few studies have evaluated radiosensitivity and survival outcomes in prostate cancer. The functionality of rs4474733 and rs8176305 is unknown. We identified seven SNPs (rs1824890, rs8176199, rs8176235, rs8176257, rs4793213, rs17599948, and rs11654307) in strong linkage disequilibrium with rs4474733 (R² > 0.8), but none of these SNPs have any documented function. Eleven SNPs (rs8176266, rs8176256, rs4986850, rs76063231, rs118070107, rs58858902, rs112960602, rs113713543, rs74792387, rs113163396, rs74966673) were in linkage disequilibrium (R² > 0.8) with rs8176305. Of particular interest is rs4986850, which has been shown to cause a missense mutation in BRCA1 at position 693 leading to an Asp⁶⁹³Asn amino acid change. Furthermore, this SNP was previously shown to be associated with aggressive forms of breast cancer(27) and others have shown an association with an increased risk of ovarian cancer.(28) However, in silico prediction tools, including SIFT,(29) PolyPhen-2,(30) and I-mutant 3,(31) did not predict whether this mutation is benign or deleterious. Thus, the clinical relevance of these SNPs remains unclear and merit further study.

There are some limitations to consider in interpreting our study findings. Even with a mean follow-up of 12.6 years, and 141 lethal cases we had limited numbers for exploratory analyses (e.g., stratifying the radiation therapy cohort by use of ADT or type of radiation therapy (external beam radiation therapy vs brachytherapy)). The population in our study was limited to Caucasians and it is unknown whether the findings may be generalizable to men of other races who may have a different distribution of allele frequencies (For example, the MAF for rs4474733 is less common among Africans (0.14) and more common in Asians (0.37); for rs8176305, the MAF for Africans is 0.01 and in Asians is monomorphic based on the 1000 genomes population) and different risk of prostate cancer progression (7). We cannot rule out that our results may be due to chance; while the association of rs8176305 was significant even after considering multiple testing (p=0.001), rs4474733 was only nominally significant (p=0.05). In addition, while the point estimates were consistent between cohorts for rs4474733, the results for rs8176305 were mainly driven by the HPFS findings. Lastly, given the attenuation of effect size in the models adjusted for clinical predictors and the relatively modest magnitudes of the associations it is likely that these SNPs alone may not make large differences in risk prediction. However, our findings may yield important clues about the complex biology of treatment-specific progression and deserve further follow-up along with additional assessment of other key proteins involved in DNA repair and cell cycle control. A unique strength of our study was the availability of germline DNA for genotyping on a relatively large number of men with prostate cancer who had long term follow-up for assessment of prostate cancer specific mortality and detailed treatment information which allowed for comparison of treatment responses between men receiving radiation therapy and radical prostatectomy.

In summary, we found suggestive evidence that common genetic variation in the BRCA1 gene may play a role in disease progression for men with early-stage prostate cancer receiving radiation therapy. Additional studies are needed to confirm these associations and further identify the causal loci and functional significance of these findings.