Abstract
Distinguishing aggressive prostate cancer from indolent disease improves personalized treatment. Although only few genetic variants are known to predispose to aggressive prostate cancer, synergistic interactions of HOXB13 G84E high-risk prostate cancer susceptibility mutation with other genetic loci remain unknown. The purpose of this study was to examine the interplay of HOXB13 rs138213197 (G84E) and CIP2A rs2278911 (R229Q) germline variants on prostate cancer risk.
Genotyping was done in Finnish discovery cohort (n = 2,738) and validated in Swedish (n = 3,132) and independent Finnish (n = 1,155) prostate cancer cohorts. Expression pattern analysis was followed by functional studies in prostate cancer cell models.
Interplay of HOXB13 (G84E) and CIP2A (R229Q) variants results in highest observed inherited prostate cancer risk (OR, 21.1; P = 0.000024). In addition, this synergism indicates a significant association of HOXB13 T and CIP2A T dual carriers with elevated risk for high Gleason score (OR, 2.3; P = 0.025) and worse prostate cancer–specific life expectancy (HR, 3.9; P = 0.048), and it is linked with high PSA at diagnosis (OR, 3.30; P = 0.028). Furthermore, combined high expression of HOXB13-CIP2A correlates with earlier biochemical recurrence. Finally, functional experiments showed that ectopic expression of variants stimulates prostate cancer cell growth and migration. In addition, we observed strong chromatin binding of HOXB13 at CIP2A locus and revealed that HOXB13 functionally promotes CIP2A transcription. The study is limited to retrospective Nordic cohorts.
Simultaneous presence of HOXB13 T and CIP2A T alleles confers for high prostate cancer risk and aggressiveness of disease, earlier biochemical relapse, and lower disease-specific life expectancy. HOXB13 protein binds to CIP2A gene and functionally promotes CIP2A transcription.
Synergistic genetic interaction of HOXB13 rs138213197 (G84E) germline mutation with a common variant CIP2A rs2278911 (R229Q), first demonstrated here, confers risk for aggressive prostate cancer. Our unique findings suggest exceptional clinical potential of HOXB13-CIP2A as novel synergistic genetic markers in aggressive prostate cancer. Combined high expression of HOXB13-CIP2A outperforms each gene alone in prediction of the time to biochemical relapse, showing also prognostic potential. Genetic biomarkers such as HOXB13/CIP2A and their germline genetic testing may bring new opportunities for precision oncology in prostate cancer. Understanding of the clinical relevance of the molecular subclassification may have a critical role in the developments of patient selection strategies and new therapeutic approaches.
Introduction
Prostate cancer is the second most common cancer in men, with an estimated 1.1 million men diagnosed worldwide, accounting for 15% of the cancers diagnosed in men. Prostate cancer represents the fifth leading cause of male cancer-related death with over 300,000 annual deaths (http://globocan.iarc.fr, GLOBOCAN 2012). The Nordic Twin Study of Cancer, the world's largest population-based twin cohort, recently estimated genetic factors to account for 57% of the prostate cancer risk [95% confidence interval (CI), 51%–63%; ref. 1]. The twin-based findings collaborate results of genome-wide association studies (GWAS) that have discovered about 105 susceptibility loci, each low-penetrance common variant individually modestly increasing the risk (average per allele OR, 1.1–1.3; ref. 2). These SNPs are estimated to explain about 30% of the inherited risk of prostate cancer (3, 4). Despite these findings, the polygenic and very heterogeneous nature of prostate cancer has not yet been dissected.
One of the strongest prostate cancer risk predictors is the rare recurrent and highly penetrant missense mutation rs138213197/G84E in HOXB13, which codes a homeobox transcription factor that is important in prostate development (5). There are few additional HOXB13 variants (e.g., Y88D, L144P, G216C, R229G), identified in U.S. Caucasians (5). However, the G84E is a Finnish founder mutation and was recently found to be linked to androgen receptor (AR) cistrome reprogramming through cooperating with the pioneer factor FOXA1 in human prostate tumorigenesis (6). Sharing common haplotype in mutation carriers suggested a founder effect, which was estimated to occur in 1792 in Finland (7). It is present in 8.4% of Finnish familial prostate cancer cases (OR, 8.8) and represents significantly increased prostate cancer risk also in unselected cases (OR, 3.6; ref. 8). In unselected prostate cancer cases, the G84E variant was also associated with a considerable risk of prostate cancer in Swedish and UK populations (OR, 3.5 and 2.93, respectively; refs. 9, 10). G84E mutation has been connected with early-onset, familial cases worldwide (OR, 7.9; ref. 11) and explains partly the hereditary component of prostate cancer (5). Although overexpression of HOXB13 has been linked to several prognostic predictors of prostate cancer (12) and prostate tumorigenesis in men, the associations of G84E with prostate cancer clinical outcomes are not fully known (8).
Protein Phosphatase 2A (PP2A) has recently emerged as a potential prostate cancer tumor suppressor (13). Tumor-suppressor activity of PP2A is inhibited in many cancers by overexpression of endogenous inhibitor proteins, such as SET and CIP2A (14). Importantly, expression of SET and CIP2A is associated with high Gleason scores and the presence of metastatic disease (15). CIP2A protein expression is increased in a variety of cancers, including prostate cancer (16), where it has been associated with castration-resistant prostate cancer (CRPC; refs. 16, 17). Recently, depletion of CIP2A in CRPC cell lines was shown to sensitize the cells to therapeutic agents (18). However, even though CIP2A can be considered functionally as a cancer driver protein, genetic evidence for CIP2A mutations in human cancers is very rare. The only genetic study to date of CIP2A in human cancer found that it exerted a synergetic effect with the exonic missense mutation (rs2278911, C/T, exon 6, p. R229Q) on the risk of hepatocellular carcinoma (HCC) in hepatitis B and C virus infection in a Han Chinese population (19). The high prevalence of C carriers (50%), together with observation that the allele is not associated with HCC risk, but only has a role in the context of hepatitis infection, suggests that CIP2A rs2278911 variant might constitute a cooperating oncogene. However, neither the prevalence of rs2278911 in other populations or disease groups, or the interaction with other genes, nor the potential functional role of this mutation has been studied thus far.
Exact molecular mechanisms underlying the initiation and progression of prostate cancer still remain largely unknown. One potential reason for this is that the interaction of genetic variants using large population-based cohort has not been widely studied. In British men, no evidence of interaction between the HOXB13 G84E variant and polygenic risk score was found (10). The Finnish cohort is particularly suited for thorough further search for the interactive genetic variants in clinically differentiated prostate cancer cases. Therefore, considering the overexpression of both HOXB13 and CIP2A in prostate cancer, our aim was to elucidate the role of CIP2A in prostate cancer genetic risk and investigate the possible interaction of HOXB13 rs138213197 and CIP2A rs2278911 in prostate cancer susceptibility, and in relation to several disease progression and clinical outcome parameters.
Materials and Methods
In the discovery study, the genotyped cancer patients and controls were of Finnish origin. The study was conducted in accordance with the ethical guidelines of Helsinki Declaration (1975). Written-informed consent was obtained from each study subject. The human investigations were performed after approval of the study protocol by the research ethics committee at Pirkanmaa Hospital District (Tampere, Finland) and by the National Supervisory Authority for Welfare and Health (VALVIRA). For HOXB13 mutation 2,669 and for CIP2A variant 2,738 unselected nonfamilial eligible prostate cancer cases were analyzed, respectively. The genotyping was partly carried out by the PRACTICAL (Prostate Cancer Association group to Investigate Cancer Associated Alterations in the Genome) consortium. Of cases, 2,281 were clinically diagnosed cases from the Pirkanmaa Hospital District, confirmed from medical records. Another set of subjects consisted of 457 Finnish screen-detected cancer cases recruited by the Finnish arm of The European Randomized Study of Screening for Prostate Cancer (ERSPC; ref. 20). Clinical characteristics of genotyped Finnish prostate cancer patients are summarized in Supplementary Table S1. Prostate cancer control subjects (HOXB13 G84E, n = 2,423; CIP2A R229Q, n = 2,427) belonging to the screening trial control group were derived from the Finnish arm of the ERSPC (20). Control subjects were population-matched healthy individuals of ages between 70 and 86 years who had undergone PSA screening. Their disease status is annually evaluated from the records of the Finnish Cancer Registry.
In the validation phase of the study, the Swedish Stockholm2 (STHM2) cohort (prostate cancer patients, n = 3,132; controls, n = 1,429) and the TAMPERE2 cohort (prostate cancer, n = 1,155; controls, n = 1,184) were studied. High PSA (>20 ng/mL) was present in 8.9% (n = 278) of STHM2 and in 18.3% (n = 211) of TAMPERE2 patients. Similarly, a high Gleason score ≥ 8 was observed for 10.5% (n = 329) of STHM2 patients and for 5.1% of TAMPERE2 patients. Written-informed consent was obtained from each study subject.
Clinical and pathologic subclassification of prostate cancer patients can be found in Supplementary Material including the combined modality staging according to ERSPC classification (Supplementary Table S2).
SNP genotyping and sequencing details are given in Supplementary File.
In statistical analyses, the Hardy–Weinberg equilibrium equation was used to determine whether the proportion of each genotype obtained was in agreement with the expected values as calculated from the allele frequencies.
Statistical analyses were performed with IBM SPSS version 22 (SPSS Inc.) unless otherwise specified. As the number of parameters was relatively small compared with the number of subjects, unconditional logistic regression analyses were used to measure the association between HOXB13 and CIP2A variants and the risk of prostate cancer by estimation of OR and its 95% CI. P values were two-sided, and P < 0.05 was considered to indicate a statistically significant result.
To evaluate the relative effects of HOXB13 and CIP2A variants on prostate cancer development, binary stepwise logistic regression with backward elimination method was used as described by ref. 21. By fitting statistical models with main effects, we employed a test with few degrees of freedom that is likely to be powerful for detecting primary etiologic determinants. This approach is applicable to case–control study modeled via unconditional logistic regression. By testing two polymorphisms, small number of degrees of freedom will be present. In this procedure, we started with all candidate variables and tested if the deletion of the variable improves the model the most by being deleted, and repeating this process until no further improvement is possible.
For the time-to-event analyses, the Cox regression method was used.
We conducted binary logistic regression analyses to evaluate the impact of the mutations on tumor phenotype and the following selected clinical features: age at onset, PSA at diagnosis, Gleason score, tumor stage (T), the presence of nodal (N) or distant metastases (M), general progression, PSA progression, local and distant progression, age at progression, T2:ERG fusion status [transmembrane protease, serine 2 (TMPRSS2):v-ets erythroblastosis virus E26 oncogene homolog (avian; ERG)], other cancer, clinical, and screen detection, CRPC development, benign prostatic hyperplasia (BPH) status and prostate cancer development, and general and disease-specific death.
In analyses of the association between HOXB13 and CIP2A variants and the combined modality stage (ERSPC classification) of prostate cancer patients, we applied case–case multinomial logistic regression analyses.
In order to explore the role of HOXB13 and CIP2A in overall and prostate cancer–specific survival, we applied Kaplan–Meier survival analyses. Survival time (years) was compared between carriers and noncarriers. Follow-up characteristics, defined follow-up periods (birth-death, diagnosis-progression, progression-death, diagnosis-death), and follow-up periods used in survival analyses of Finnish prostate cancer patients are summarized in Supplementary Table S3.
The pathogenicity prediction of G84E and R229Q variants was investigated by using in silico tool for missense variants (see Supplementary Material).
Assessment of the effect of combined HOXB13-CIP2A expression on prediction of prostate cancer biochemical recurrence was based on The Cancer Genome Atlas (TCGA) dataset (22) containing 333 patients with primary prostate cancer obtained from cBioPortal for Cancer Genomics (23, 24). The mRNA expression levels of HOXB13 and CIP2A (KIAA1524) were clustered by the k-means method from the “amap” package in R (25). Parameters were set as “n = 2, nstart = 25, method = pearson,” which aims to partition the patients into two groups whose members share some measure of similarity in their expression pattern according to the expression levels of their genes. The Kaplan–Meier analysis from the R “Survival” package was used to estimate the biochemical recurrence-free survival of patients after which they were partitioned into two groups across the TCGA datasets.
Methodologic details of functional studies are detailed in Supplementary File, including plasmids and site-directed mutagenesis, cell culture protocol and Western blot analyses, the cell viability, proliferation and wound-healing assays, chromatin immunoprecipitation followed by quantitative PCR, lentivirus production and infection, shRNA- and siRNA-mediated knockdown of HOXB13, RNA extraction and quantitative RT-PCR (Supplementary Table S4), RNA-seq, and data analysis.
Results
Germline risk of prostate cancer defined by dual T alleles of HOXB13 G84E and CIP2A R229Q
In the Finnish discovery cohort, the overall minor T allele frequency of HOXB13 G84E variant was 1.8% and for the CIP2A R229Q was 13.8%. The G84E mutation was more frequently observed in patients (n = 167, carrier frequency 6.24%) than in healthy controls (n = 20, carrier frequency 0.80%). The carrier frequency of R229Q variant was similar in cases (24.7%, n = 675) and in controls (26.5%, n = 642; Table 1). HOXB13 G84E T allele carriers had an 8.0-fold increased risk of prostate cancer (95% CI, 5.0–12.8; P = 7.69E–25) in unselected nonfamilial cases in the analyzed Finnish samples. Similarly, CT heterozygotes also showed significantly elevated risk for prostate cancer (OR, 7.9; 95% CI, 4.9–12.7; P = 1.22E–24). CIP2A T allele had no significant effect on prostate cancer risk. However, most importantly HOXB13 T and CIP2A T dual carriers showed striking 21.1-fold increased odds for prostate cancer (95% CI, 5.2–87.5; P = 0.000024).
Locus genotype . | . | . | . | . |
---|---|---|---|---|
Discovery cohort (TAMPERE) . | Case, n (%) . | Control, n (%) . | OR (95% CI) . | P value . |
HOXB13 rs138213197 | ||||
CC | 2,502 (93.7) | 2,403 (99.2) | 1.0 | |
CT | 166 (6.20) | 20 (0.80) | 7.9 (4.9–12.7) | 1.22E-24 |
TTa | 1 (0.10) | 0 (0.00) | — | — |
T carriers | 167 (6.24) | 20 (0.80) | 8.0 (5.0–12.8) | 7.69E-25 |
CIP2A rs2278911 | ||||
CC | 2,063 (75.3) | 1,785 (73.5) | 1.0 | |
CT | 625 (22.8) | 588 (24.2) | 0.9 (0.8–1.1) | 0.236 |
TT | 50 (1.80) | 54 (2.20) | 0.8 (0.6–1.2) | 0.308 |
T carriers | 675 (24.7) | 642 (26.5) | 0.9 (0.8–1.0) | 0.139 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 2,620 (98.2) | 2,369 (97.8) | 1.0 | |
HOXB13 T&CIP2A T | 46 (1.70) | 2 (0.10) | 21.2 (5.2–87.5) | 0.000024 |
Validation cohort (STOCKHOLM) | ||||
HOXB13 rs138213197 | ||||
CC | 3,022 (96.5) | 1,418 (99.2) | 1.0 | |
CT | 108 (3.44) | 11 (0.77) | 4.6 (2.5–8.6) | 1.57E-6 |
TTa | 2 (0.06) | 0 (0.00) | — | — |
T carriers | 110 (3.51) | 11 (0.77) | 4.7 (2.5–8.7) | 1.15E-6 |
CIP2A rs2278911 | ||||
CC | 2,455 (78.4) | 1,103 (77.2) | 1.0 | |
CT | 626 (20.0) | 303 (21.2) | 0.9 (0.8–1.1) | 0.345 |
TT | 51 (1.63) | 23 (1.61) | 1.0 (0.6–1.6) | 0.988 |
T carriers | 677 (21.6) | 326 (22.8) | 0.9 (0.8–1.1) | 0.365 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 3,104 (99.1) | 1,427 (99.9) | 1.0 | |
HOXB13 T&CIP2A T | 28 (0.89) | 2 (0.14) | 6.4 (1.5–27.0) | 0.011 |
Validation cohort (TAMPERE2) | ||||
HOXB13 rs138213197 | ||||
CC | 1,067 (92.4) | 1,178 (99.5) | 1.0 | |
CT | 88 (7.6) | 6 (0.50) | 16.2 (7.1–37.2) | 5.15E–11 |
TTa | 0 (0.00) | 0 (0.00) | — | — |
T carriers | 88 (7.6) | 6 (0.77) | 16.2 (7.1–37.2) | 5.15E–11 |
CIP2A rs2278911 | ||||
CC | 857 (74.2) | 871 (73.6) | 1.0 | |
CT | 272 (23.5) | 295 (24.9) | 0.9 (0.8–1.1) | 0.441 |
TT | 26 (2.3) | 18 (1.5) | 1.5 (0.8–2.7) | 0.196 |
T carriers | 298 (25.8) | 313 (26.4) | 0.9 (0.8–1.2) | 0.727 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 1,129 (97.7) | 1,166 (98.5) | 1.0 | |
HOXB13 T&CIP2A T | 23 (2.0) | 1 (0.10) | 24.0 (3.2–178.3) | 0.002 |
Locus genotype . | . | . | . | . |
---|---|---|---|---|
Discovery cohort (TAMPERE) . | Case, n (%) . | Control, n (%) . | OR (95% CI) . | P value . |
HOXB13 rs138213197 | ||||
CC | 2,502 (93.7) | 2,403 (99.2) | 1.0 | |
CT | 166 (6.20) | 20 (0.80) | 7.9 (4.9–12.7) | 1.22E-24 |
TTa | 1 (0.10) | 0 (0.00) | — | — |
T carriers | 167 (6.24) | 20 (0.80) | 8.0 (5.0–12.8) | 7.69E-25 |
CIP2A rs2278911 | ||||
CC | 2,063 (75.3) | 1,785 (73.5) | 1.0 | |
CT | 625 (22.8) | 588 (24.2) | 0.9 (0.8–1.1) | 0.236 |
TT | 50 (1.80) | 54 (2.20) | 0.8 (0.6–1.2) | 0.308 |
T carriers | 675 (24.7) | 642 (26.5) | 0.9 (0.8–1.0) | 0.139 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 2,620 (98.2) | 2,369 (97.8) | 1.0 | |
HOXB13 T&CIP2A T | 46 (1.70) | 2 (0.10) | 21.2 (5.2–87.5) | 0.000024 |
Validation cohort (STOCKHOLM) | ||||
HOXB13 rs138213197 | ||||
CC | 3,022 (96.5) | 1,418 (99.2) | 1.0 | |
CT | 108 (3.44) | 11 (0.77) | 4.6 (2.5–8.6) | 1.57E-6 |
TTa | 2 (0.06) | 0 (0.00) | — | — |
T carriers | 110 (3.51) | 11 (0.77) | 4.7 (2.5–8.7) | 1.15E-6 |
CIP2A rs2278911 | ||||
CC | 2,455 (78.4) | 1,103 (77.2) | 1.0 | |
CT | 626 (20.0) | 303 (21.2) | 0.9 (0.8–1.1) | 0.345 |
TT | 51 (1.63) | 23 (1.61) | 1.0 (0.6–1.6) | 0.988 |
T carriers | 677 (21.6) | 326 (22.8) | 0.9 (0.8–1.1) | 0.365 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 3,104 (99.1) | 1,427 (99.9) | 1.0 | |
HOXB13 T&CIP2A T | 28 (0.89) | 2 (0.14) | 6.4 (1.5–27.0) | 0.011 |
Validation cohort (TAMPERE2) | ||||
HOXB13 rs138213197 | ||||
CC | 1,067 (92.4) | 1,178 (99.5) | 1.0 | |
CT | 88 (7.6) | 6 (0.50) | 16.2 (7.1–37.2) | 5.15E–11 |
TTa | 0 (0.00) | 0 (0.00) | — | — |
T carriers | 88 (7.6) | 6 (0.77) | 16.2 (7.1–37.2) | 5.15E–11 |
CIP2A rs2278911 | ||||
CC | 857 (74.2) | 871 (73.6) | 1.0 | |
CT | 272 (23.5) | 295 (24.9) | 0.9 (0.8–1.1) | 0.441 |
TT | 26 (2.3) | 18 (1.5) | 1.5 (0.8–2.7) | 0.196 |
T carriers | 298 (25.8) | 313 (26.4) | 0.9 (0.8–1.2) | 0.727 |
Dual carriers of HOXB13 and CIP2A | ||||
HOXB13 C&CIP2A C | 1,129 (97.7) | 1,166 (98.5) | 1.0 | |
HOXB13 T&CIP2A T | 23 (2.0) | 1 (0.10) | 24.0 (3.2–178.3) | 0.002 |
NOTE: Case–control logistic regression analyses. Results are in bold if the 95% CI excluded 1 and the association significant at P < 0.05 vs. controls.
aOR could not be calculated because of the limitations of log regression method.
In the validation phase of this study, we examined the possible risk of prostate cancer as defined by studied polymorphisms in Swedish (prostate cancer patients, n = 3,132; controls, n = 1,429) and in independent Finnish (prostate cancer, n = 1,155; Controls, n = 1,184) cohorts (Table 1). In the Swedish validation cohort, the same trends of associations were observed, supporting the original findings. HOXB13 G84E T allele carriers showed a 4.7-fold increased risk of prostate cancer (95% CI, 2.5–8.7; P = 1.15E–6). Although the frequency of dual T allele carriers was lower in the Swedish cohort (0.89% in prostate cancer patients vs. 0.14% in controls) than in the Finnish cohort, we observed a significant 6.4-fold increased risk of prostate cancer in these Swedish carriers (95% CI, 1.5–27.0; P = 0.011). In addition, the results of the Finnish validation study fully support the synergistic effect of studied combination variants on prostate cancer risk (OR, 24.0; 95% CI, 3.2–178.3; P < 0.003). The significantly high prostate cancer risk (OR, 16.2; 95% CI, 7.1–37.2; P = 5.15E–11) caused by the HOXB13 G84E T allele was also replicated successfully. Important to state, that the Finnish validation cohort study resulted in similar carrier frequencies of the studied variants in general as it was observed in discovery cohort of Finland.
The increases in prostate cancer risk observed for carriers of only one of the HOXB13 T and CIP2A T alleles were statistically less augmented than the increase observed for dual T carriers (Supplementary Table S5), suggesting that the simultaneous presence of variant T alleles at both loci might play an explicit biological role in prostate cancer oncogenesis. Statistical dissection of the effect of HOXB13 G84E and CIP2A R229Q variants can be found in Supplementary Result.
Clinical features of prostate cancer associated with dual carriers of HOXB13 G84E and CIP2A R229Q
Statistically significant associations of HOXB13 G84E carriers and dual carriers of HOXB13 G84E and CIP2A R229Q (vs. noncarriers) with clinical characteristics are presented in Table 2. Similar to the findings regarding disease risk, no evidence of an association between CIP2A R229Q T allele carrier status and any clinical feature of prostate cancer was found in any of the studied cohorts (data not shown). Significant association was observed between the HOXB13 mutation status and high PSA at diagnosis (PSA > 20 ng/mL; OR, 1.5; 95% CI, 1.1–2.3; P = 0.012). No significant association was observed between HOXB13 mutation status and Gleason score (Gleason score ≤ 6 vs. ≥ 8, P = 0.093). Importantly, dual HOXB13 T and CIP2A T carriers showed a significant association with Gleason score (Gleason ≥ 8; OR, 2.3; 95% CI, 1.1–4.8; P = 0.025). This score is a commonly used clinical parameter to define aggressive prostate cancer and poor prognosis when predicting disease. We also examined the effect of HOXB13 and CIP2A mutation status on age at diagnosis, age at progression, PSA progression, local or distant progression, tumor–node–metastasis stage, and T2:ERG fusion status, but no significant associations were found (data not shown). Uniquely, in the Swedish prostate cancer validation cohort, dual T carrier status was associated with high PSA at diagnosis (OR, 3.30; 95% CI, 1.1–9.6; P = 0.028) but not with elevated Gleason score. In the Finnish validation cohort, we successfully replicated the significant association between the HOXB13 T allele and high PSA at diagnosis (OR, 1.86; 95% CI, 1.1–3.1; P = 0.01). However, possibly due to different selection criteria of patients for the validation cohorts, such as clinical versus PSA screening cases, and smaller sample size, statistically significant association between dual T carriers and high Gleason score of prostate cancer biopsy was not seen neither in Finnish nor in Swedish validation cohorts.
. | HOXB13 rs138213197 . | HOXB13 rs138213197 and CIP2A rs2278911 . | |||||||
---|---|---|---|---|---|---|---|---|---|
Clinical parameter . | T carrier % (n) . | T noncarrier % (n) . | OR (95% CI) . | P value . | Dual T carrier % (n) . | Dual T noncarrier % (n) . | OR (95% CI) . | P value . | |
Discovery cohort (TAMPERE) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 5.7 (117) | 94.3 (1924) | 1.0 | 1.7 (35) | 98.3 (2,006) | 1.0 | |||
>20 ng/mL | 8.9 (42) | 91.1 (432) | 1.50 (1.11–2.31) | 0.012 | 2.1 (10) | 97.9 (464) | 1.24 (0.61–2.51) | 0.560 | |
Gleason score | |||||||||
≤6 | 5.9 (76) | 94.7 (1,207) | 1.0 | 1.5 (19) | 98.5 (1,264) | 1.0 | |||
≥8 | 8.4 (30) | 91.6 (327) | 1.46 (0.94–2.26) | 0.093 | 3.4 (12) | 96.6 (345) | 2.31 (1.11–4.81) | 0.025 | |
Validation cohort (STOCKHOLM) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 3.2 (64) | 96.8 (1,930) | 1.0 | 0.6 (11) | 99.4 (1,983) | 1.0 | |||
>20 ng/mL | 3.6 (10) | 96.4 (268) | 1.21 (0.64–2.28) | 0.556 | 1.8 (5) | 98.2 (273) | 3.30 (1.14–9.57) | 0.028 | |
Gleason score | |||||||||
≤6 | 4.1 (65) | 95.9 (1,515) | 1.0 | 1.1 (17) | 98.9 (1,562) | 1.0 | |||
≥8 | 3.0 (10) | 97.0 (319) | 0.72 (0.37–1.41) | 0.346 | 1.2 (4) | 98.8 (325) | 1.13 (0.38–2.28) | 0.825 | |
Validation cohort (TAMPERE2) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 6.7 (58) | 93.3 (803) | 1.0 | 1.7 (15) | 98.3 (846) | 1.0 | |||
>20 ng/mL | 11.8 (25) | 88.2 (186) | 1.86 (1.13–3.05) | 0.01 | 3.8 (8) | 96.2 (203) | 2.22 (0.93–5.31) | 0.07 | |
Gleason score | |||||||||
≤6 | 6.3 (26) | 93.7 (384) | 1.0 | 0.7 (3) | 99.3 (407) | 1.0 | |||
≥8 | 7.5 (12) | 92.5 (149) | 1.19 (0.59–2.42) | 0.63 | 1.9 (3) | 98.1 (158) | 2.58 (0.51–12.90) | 0.25 |
. | HOXB13 rs138213197 . | HOXB13 rs138213197 and CIP2A rs2278911 . | |||||||
---|---|---|---|---|---|---|---|---|---|
Clinical parameter . | T carrier % (n) . | T noncarrier % (n) . | OR (95% CI) . | P value . | Dual T carrier % (n) . | Dual T noncarrier % (n) . | OR (95% CI) . | P value . | |
Discovery cohort (TAMPERE) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 5.7 (117) | 94.3 (1924) | 1.0 | 1.7 (35) | 98.3 (2,006) | 1.0 | |||
>20 ng/mL | 8.9 (42) | 91.1 (432) | 1.50 (1.11–2.31) | 0.012 | 2.1 (10) | 97.9 (464) | 1.24 (0.61–2.51) | 0.560 | |
Gleason score | |||||||||
≤6 | 5.9 (76) | 94.7 (1,207) | 1.0 | 1.5 (19) | 98.5 (1,264) | 1.0 | |||
≥8 | 8.4 (30) | 91.6 (327) | 1.46 (0.94–2.26) | 0.093 | 3.4 (12) | 96.6 (345) | 2.31 (1.11–4.81) | 0.025 | |
Validation cohort (STOCKHOLM) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 3.2 (64) | 96.8 (1,930) | 1.0 | 0.6 (11) | 99.4 (1,983) | 1.0 | |||
>20 ng/mL | 3.6 (10) | 96.4 (268) | 1.21 (0.64–2.28) | 0.556 | 1.8 (5) | 98.2 (273) | 3.30 (1.14–9.57) | 0.028 | |
Gleason score | |||||||||
≤6 | 4.1 (65) | 95.9 (1,515) | 1.0 | 1.1 (17) | 98.9 (1,562) | 1.0 | |||
≥8 | 3.0 (10) | 97.0 (319) | 0.72 (0.37–1.41) | 0.346 | 1.2 (4) | 98.8 (325) | 1.13 (0.38–2.28) | 0.825 | |
Validation cohort (TAMPERE2) | |||||||||
PSA at diagnosis | |||||||||
≤20 ng/mL | 6.7 (58) | 93.3 (803) | 1.0 | 1.7 (15) | 98.3 (846) | 1.0 | |||
>20 ng/mL | 11.8 (25) | 88.2 (186) | 1.86 (1.13–3.05) | 0.01 | 3.8 (8) | 96.2 (203) | 2.22 (0.93–5.31) | 0.07 | |
Gleason score | |||||||||
≤6 | 6.3 (26) | 93.7 (384) | 1.0 | 0.7 (3) | 99.3 (407) | 1.0 | |||
≥8 | 7.5 (12) | 92.5 (149) | 1.19 (0.59–2.42) | 0.63 | 1.9 (3) | 98.1 (158) | 2.58 (0.51–12.90) | 0.25 |
NOTE: Case–case logistic regression analyses. Results are in bold, if the 95% CI excluded 1 and the association significant at P < 0.05 vs. noncarriers.
aSummary of statistically significant associations.
Additional associations with clinical features of prostate cancer are detailed in Supplementary Results. In short, we show that HOXB13 and CIP2A dual T carriers develop prostate cancer 7.2 months earlier (HR, 2.1; P = 4.52E–7; Supplementary Table S6). Dual T carriers have remarkably high risk to be detected both through clinical symptoms (OR, 23.4; P = 0.000013) and through screening (OR, 10.7; P = 0.006) compared with controls (Supplementary Table S7). This study also shows that G84E variant confers for a 12.6-fold risk of developing prostate cancer in BPH cases compared with BPH controls (95% CI, 2.8–56.8; P = 0.001; Supplementary Table S7). In comprehensive evaluation of the clinical factors, using the combined modality stage based on the ERSPC classification, we could not reveal reasonable association with dual T carriers (Supplementary Table S8).
Life-span survival of prostate cancer patients is shaped by dual T alleles of HOXB13 and CIP2A
Kaplan–Meier survival analyses were applied to assess the overall survival and prostate cancer–specific survival differences between HOXB13 and CIP2A carriers and prostate cancer patients without the mutations. We subdivided the follow-up period between birth and death into periods of diagnosis-death, diagnoses-progression, and progression-death.
Based on the data, our survival analyses suggested no worse prostate cancer–specific prognosis for HOXB13 T carriers after the disease progressed (case-only, Breslow test; HR 1.4; 95% CI, 15.6–16.9; P = 0.240). No association was found between CIP2A T carriers and the survival of prostate cancer patients. Exploring the longest follow-up time (birth-death), dual T carriers of HOXB13 and CIP2A have significantly worse prostate cancer–related life expectancy and die earlier of prostate cancer (Breslow test; HR, 3.9; 95% CI, 91.6–94.5; P = 0.048) versus nondual T carriers.
HOXB13 G84E and CIP2A R229Q variants are predicted to be pathogenic (in silico)
We applied CADD (Combined Annotation Dependent Depletion) framework for functional effect prediction analysis (26). HOXB13 G84E was predicted to be deleterious with a scaled C-score of 29.6, which indicates it to be a potentially causative mutation. CIP2A R229Q variant has a scaled C-score of 23.9, denoting similarly deleterious variant as well. In addition, we applied M-CAP classifier and identified the pathogenicity score of 0.110 (T allele, possibly pathogenic) with high clinical sensitivity for HOXB13 G84E rare missense variant (recommended threshold > 0.025; ref. 27).
High expression pattern of HOXB13-CIP2A associates with earlier biochemical recurrence
We next assessed the effect of combined HOXB13-CIP2A expression pattern on clinical features of prostate cancer patients (see Patients and Methods). We observed a significant association in a TCGA cohort of prostate cancer (ref. 22; Supplementary Fig. S1). The time to PSA relapse was marginally significantly shorter in the patient group with higher CIP2A expression (HR, 1.718; 95% CI, 0.993–2.975; P = 0.0505), but not with HOXB13. Remarkably, the combination of HOXB13-CIP2A expression data is somewhat better than CIP2A alone in prediction of the risk of biochemical recurrence of these prostate cancer patients (HR, 1.86; 95% CI, 1.09–3.20; P = 0.0219), suggesting an obvious additive value of two genes in disease prognosis.
Gene-level allele-specific expression of CIP2A R229Q
Gene-level allele-specific expression (ASE) analyses using RNA-seq gene-expression data and genome-wide high-density genotypes from 471 samples of benign primary prostate tissue (28) were conducted to evaluate the effect of studied variant alleles on gene expression. This analysis identified a significant association of CIP2A R229Q T allele with CIP2A upregulation (P = 0.0187), suggesting that the CIP2A R229Q T allele is associated with increased expression of CIP2A. Due to the rare frequency of HOXB13, G84E mutation analysis of ASE was not feasible.
Ectopic expression of HOXB13 G84E and CIP2A R229Q variants stimulates prostate cell growth and migration
We next sought to examine the functional impact of HOXB13 G84E and CIP2A R229Q variants on cellular behavior of prostate normal and cancer cells. Thus, we overexpressed wild-type alleles and their variants in multiple human prostate cell models to monitor cell growth using cell proliferation assays (Fig. 1). We firstly confirmed the overexpression of indicated V5-tagged genes in the tested immortalized benign prostate cell line RWPE1 and prostate cancer cell models, 22Rv1and PC3 (Fig. 1A, C, and E). Although the HOXB13 G84E variant was reported to be associated with an increased risk of prostate cancer in previous work (5), we found that overexpression of HOXB13 other than HOXB13 G84E promoted cell growth in RWPE1 (Fig. 1B) and 22Rv1 (Fig. 1D). Overexpression of HOXB13 G84E somehow reduced cell growth of RWPE1 and PC3 (Fig. 1B and F). RWPE1 and PC3 cells overexpressing CIP2A or CIP2A R229Q showed no difference in proliferation from control cells, whereas overexpression of CIP2A or CIP2A R229Q in 22Rv1 cells slightly promoted cell growth (Supplementary Table S9). In contrast, all the tested cell lines with cooverexpression of HOXB13 G84E and CIP2A R229Q displayed elevated proliferation in comparison with the cells coexpressing both wild-type HOXB13 and CIP2A, consistent with our genetic finding of a synergistic interaction between the two variants.
In agreement with these results, our wound-healing assays showed that the cooverexpression of HOXB13 G84E and CIP2A R229Q increased the wound closure rate in a time-dependent manner in RWPE1 and 22Rv1 cells (Fig. 2; Supplementary Fig. S2; Supplementary Table S9), suggesting synergistic roles of these variants in cell migration. Moreover, similar to the results of cell proliferation assays, overexpression of HOXB13, HOXB13 G84E, CIP2A, or CIP2A R229Q in different cell line has different impact on cell migration, which might be due to the different genetic background of these prostate cell lines or their representation of different stage of prostate cancer.
To further investigate the molecular alterations in prostate cancer cells coexpressing HOXB13 G84E and CIP2A R229Q, we sought to identify their potential downstream target genes. We thus performed RNA-seq analysis of 22Rv1 cells cooverexpressing HOXB13 G84E and CIP2A R229Q and of 22Rv1 cells coexpressing HOXB13 and CIP2A. We found that overexpression of HOXB13 G84E and CIP2A R229Q upregulated known oncogenic drivers of prostate cancer, such as NAMPT (29) and MALAT-1 (30), and downregulated tumor suppressors, including AHNAK (31) and HUWE1 (ref. 32; Supplementary Table S9), thereby potentially promoting prostate cancer cell proliferation and migration. Together, these functional results further support the synergistic roles of these variants in prostate cancer predisposition.
HOXB13 transcriptionally regulates CIP2A expression
HOXB13 is a prostate-lineage–specific transcription factor and known to be important for prostate normal development and tumorigenesis. CIP2A is often highly expressed in prostate cancer samples and cell lines. We thus sought to examine whether HOXB13 potentially regulates the transcriptional expression of CIP2A. We have previously mapped genome-wide binding sites for HOXB13 using chromatin immunoprecipitation coupled with massively parallel sequencing (ChIP-seq; ref. 33). Interestingly, query from this HOXB13 ChIP-seq data set, we observed a strong chromatin binding of HOXB13 at CIP2A locus in the prostate cancer cell line VCaP (Fig. 3A). Consistently, HOXB13 binding at CIP2A was also observed using additional ChIP-seq data derived from both LNCaP cells and prostate cancer tissue samples (Fig. 3A). We next performed HOXB13 ChIP-qPCR assays and confirmed HOXB13 chromatin binding at CIP2A regulatory regions in both the LNCaP and 22Rv1 prostate cancer cell lines (Fig. 3B and for VCaP cell line in Supplementary Fig. S3). To assess whether the HOXB13 variant influences the transcriptional regulation of CIP2A, we performed ectopic expression of V5-tagged HOXB13 or HOXB13 G84E in 22Rv1, RWPE1, and LNCaP cell lines (Fig. 1A and C; Supplementary Fig. S4) followed by ChIP-qPCR with V5 antibody. Intriguingly, HOXB13 indicated stronger binding at CIP2A regulatory regions than HOXB13 G84E in both LNCaP and 22Rv1 prostate cancer cell lines (Fig. 3C and D). In contrast, the binding affinity of HOXB13 G84E at CIP2A regulatory regions is higher than that of HOXB13 in the human immortalized prostatic epithelial RWPE1 cells (Fig. 3E), suggesting different roles of HOXB13 and HOXB13 G84E in gene regulation at different stages of prostate cancer tumorigenesis. We next examined if HOXB13 functionally regulates the expression of CIP2A and thus performed RNAi-mediated knockdown of HOXB13 in prostate cancer LNCaP and 22Rv1 cell lines, respectively. Knockdown efficiency was confirmed at both mRNA and protein levels (Fig. 3F and G). The results showed that depletion of HOXB13 led to greatly diminished mRNA levels of CIP2A (Fig. 3F and G). Taken together, these results demonstrate that CIP2A is a direct regulatory target of HOXB13.
Discussion
Shown evidence of HOXB13 G84E in prostate cancer susceptibility (33) is validated here through the significantly larger risk effect compared with common SNPs identified by GWAS with ORs ranging between 1.04 and 2.90 (3, 34, 35). Present study justifies the earlier association (OR, 3.6) in unselected nonfamilial Finnish prostate cancer cases (8) with 8-time higher odds in carriers. Our results are consistent with Laitinen and colleagues, that in the presence of HOXB13 G84E mutation the risk to have more likely >20 ng/mL PSA at diagnosis is significantly higher (8), providing strengthened clinical evidence for combined application of HOXB13 G84E variant and PSA as biomarkers in population screening trials (36). Similar to the findings of other studies, no association was observed between G84E and Gleason score, which is commonly considered as a marker of aggressive disease (8, 10). However, the speculated carcinogenic potential of HOXB13 G84E carriers in the shift between BPH and prostate cancer was successfully revealed in this study, supporting earlier observations (8, 29, 37).
Here, in addition to risk evaluation, we assessed the prognostic impact of HOXB13 G84E mutation on survival. This is the first study to report on longer than 10-year follow-up of prostate cancer patients. Earlier studies analyzed the whole survival time after diagnoses until death (8), or 5- and 10-year follow-up (10), and no association with HOXB13 G84E mutation was found.
Through the applied thorough follow-up approach, we were able to show a suggestive association that patients possessing the HOXB13 G84E had higher chance of dying from prostate cancer when PSA progresses, or when any of local or distant progression is present; though, the difference was not statistically significant.
This study represents the first comprehensive case–control study assessing the prevalence of the CIP2A (rs2278911, R229Q) variant in prostate cancer. Prior to this analysis, the role of CIP2A R229Q genetic variation has been investigated only in Asian hepatocellular carcinoma patients (19). Contrary to the results from Asians, in our analysis, the ancestral C allele is the major and the variant T is the minor allele with a 25.6% carrier frequency in entire sample set versus the 82.3% reported in Asians. As rs2278911 variant did not show risk for prostate cancer, we next focused on potential clinicopathologic, predictive, and prognostic roles of CIP2A variant in unselected prostate cancer samples in Finland. We were able to show no clinical role of the CIP2A alone. This is consistent with findings in PP2A, which showed that it may enhance tumorigenic potential only in combination with other cancer drivers (38).
The biggest tribute of this study was the interactive modeling of HOXB13 G84E and CIP2A R229Q variants on prostate cancer outcome. Exceeding all previous expectations, dual carriers of the HOXB13 T and CIP2A T alleles were at considerably high risk of prostate cancer. The odds for prostate cancer development were 21.1 times higher in dual carriers versus nondual T carriers in the Finnish discovery cohort. Further to this end, synergistic effect of HOXB13 and CIP2A variants enables us to detect the risk of prostate cancer with 3-fold higher odds than the HOXB13 T allele alone. In addition, the earlier time-to-event (7.2 months) reflects the clinical importance of synergistic risk effect of HOXB13 and CIP2A in dual T carriers. A high prostate cancer risk conferred by the combination of HOXB13 T and CIP2A T alleles was verified both in independent Finnish (OR, 24.0) and in Swedish (OR, 6.4) validation cohorts.
To date, only few variant and no synergistic effect have been shown to be associated with clinical features and aggressive disease. In a metaanalysis, the rs11672691 SNP on chromosome 19 showed association with aggressive prostate cancer (Gleason score ≥ 8; OR, 1.12; ref. 39). The rs1571801 (9q33) in the DAB2IP gene was associated with aggressive prostate cancer in European Americans, as measured by a complex set of clinical variables (OR, 1.32; ref. 40). In a Finnish study, a rare SNP (rs200331695) within the EMSY intron in 11q13.5 region was associated with aggressive prostate cancer (PSA ≥ 20 μg/L or Gleason grade ≥ 7) in unselected cases compared with controls (OR, 6.0; ref. 41). Here, we showed for the first time a synergistic effect of two variants on aggressive prostate cancer, obtaining a high OR of 2.3. Significant association between the synergistic variants and high Gleason score, i.e., aggressive disease, was not seen in the Swedish population possibly because of the underlying genetic heterogeneity between the two populations. In addition, no association with aggressive disease in the validation cohorts might be due, at least in part, to patient selection criteria and clinical differences between the discovery and validation cohorts. Higher percentage of nonclinically detected prostate cancer cases was included in the validation cohort (TAMPERE2 48%) compared with the discovery cohort (17%). This is also reflected in clinical criteria: in the discovery cohort, the percentage of high Gleason score patients is higher (21.8%) than in the STHLM2 (17.2%) or in the TAMPERE2 (18.7%).
In addition, dual T carrier status was significantly associated with high PSA at diagnosis, with an observed effect size of OR 3.30 in Swedish cohort, which might be population specific. In addition, other interacting genes may differ in Swedish population compared with the Finnish, and this may modify the effect on PSA.
To the best of our knowledge, this is the first study on synergistic interaction of genetic biomarkers in prostate cancer to report association with already existing biomarker and high-grade, aggressive disease. To take advantage of these findings, employment of both variants in prostate cancer screening would enable the identification of clinically relevant cases already at screening, and at the same time enhance the efficacy of it. Overall, CIP2A T allele seemingly increased the effect of HOXB13 T allele in more than one aspect of prostate cancer clinical characteristics.
Here, our functional experiments provide evidence of a synergistic role of HOXB13 G84E and CIP2A R229Q variants in promoting prostate cancer cell proliferation and migration.
It has been demonstrated that both HOXB13 (42) and CIP2A (15) are overexpressed in CRPCs. Here, we add that combined high expression of HOXB13-CIP2A outperforms each gene alone in prediction of the time to biochemical relapse, which underpins the prognostic potential of HOXB13-CIP2A in prostate cancer. Notably, synergism at the expression level and its association with the clinical features of prostate cancer has not been described to date.
The suggested dual role of HOXB13 gene, namely to act both as a tumor-suppressor gene and as an oncogene, has been described previously in the literature (5). Recently, Pomerantz and colleagues reported the first demonstration of an oncogenic effect of HOXB13 in combination with pioneer transcription factor FOXA1 involving the reprogramming of genome-wide AR-binding sites (cistrome) during human prostate epithelial transformation (6). Nevertheless, the molecular explanation for oncogenic role of HOXB13 G48E remains largely unknown (37) and therefore requires further investigation. Novel molecular pathways driving prostate cancer in HOXB13 G84E carriers have been suggested. Smith and colleagues identified aberrant molecular features in HOXB13 G84E carriers. These patients had low prevalence of ERG fusion–positive cancers and increased prevalence of SPINK1 overexpression (37). Here, we revealed the molecular basis of synergistic activities of HOXB13 and CIP2A mutants in clinical prostate cancer progression and the promotion of prostate cancer cell proliferation and migration. Our results also demonstrate that HOXB13 functionally regulates CIP2A transcriptional expression. We provided strong evidence supporting CIP2A as a direct target of HOXB13.
Furthermore, if we take into account cancer pathways as a whole, it is easy to see that CIP2A-mediated PP2A inhibition may promote activities of various prostate cancer pathways (43, 44). For example, HOXB13 has been shown to promote prostate cancer metastasis by stimulation of NF-κB signaling (45), and NF-κB is a target for PP2A tumor-suppressor activity in prostate cancer cells (46). Importantly, we also identify here HOXB13 as a positive regulator of CIP2A transcription in prostate cancer cell lines. This result, together with the results of a previous study demonstrating the role of AR in positively regulating CIP2A expression (47), provides a well-defined example of the lineage-specific roles of HOXB13 in reprogramming the AR cistrome and driving gene expression in human prostate tumorigenesis (29). Obviously, we cannot rule out the possibility that HOXB13 might regulate many important genes other than CIP2A in driving prostate carcinogenesis. Collectively, HOXB13 regulation of CIP2A expression and other findings of this study demonstrate the cooperation of HOXB13 and CIP2A at multiple levels in prostate cancer.
In conclusion, the combination of HOXB13 T and CIP2A T alleles appears to have a definite effect on prostate cancer risk, the time to develop the disease, disease aggressiveness, the detection of clinically relevant prostate cancer disease-specific life expectancy, and prostate cancer cell proliferation and migration. Genetic synergism was confirmed through the synergistic findings of higher HOXB13-CIP2A mRNA expression, predicting earlier biochemical relapse of prostate cancer patients. Furthermore, we describe here the molecular basis of the synergism, namely that HOXB13 protein binds to CIP2A gene and functionally promotes CIP2A transcription. Synergistic effects need to be confirmed in other ethnic groups and populations, and in familial background of prostate cancer. Further molecular studies are needed as well.
Disclosure of Potential Conflicts of Interest
A. Auvinen is a consultant/advisory board member for Epid Research Inc. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: C. Sipeky, T.L.J. Tammela, A. Auvinen, G.-H. Wei, J. Schleutker
Development of methodology: Q. Zhang, A. Auvinen, G.-H. Wei
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Sipeky, L. Wang, O. Ettala, K.M. Talala, T.L.J. Tammela, A. Auvinen, F. Wiklund, G.-H. Wei, J. Schleutker
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Sipeky, P. Gao, Q. Zhang, L. Wang, A. Auvinen, F. Wiklund, G.-H. Wei, J. Schleutker
Writing, review, and/or revision of the manuscript: C. Sipeky, P. Gao, K.M. Talala, T.L.J. Tammela, A. Auvinen, F. Wiklund, G.-H. Wei, J. Schleutker
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Sipeky, P. Gao, K.M. Talala, G.-H. Wei
Study supervision: T.L.J. Tammela, G.-H. Wei, J. Schleutker
Acknowledgments
The authors thank the patients who participated in this study. Liisa Määttänen is thanked for assistance related to FinRSPC screening trial samples, Riina Kylätie, Elina Kaikkonen, and Jukka Karhu for help in laboratory work, and Katri Pylkäs and Meeri Otsukka in helping next-generation sequencing. Tero Vahlberg's help is acknowledged for his advice in biostatistics. This work was financially supported by the Academy of Finland grants [#251074, J. Schleutker; and #284618 and #279760, G.-H. Wei), The Finnish Cancer Organisations, and the Sigrid Juselius Foundation.
We thank the members from the Prostate Cancer Association Group to Investigate Cancer Associated Alterations in the Genome (PRACTICAL) consortium who are provided in the Supplement/footnotes. Information of the consortium can be found at http://practical.icr.ac.uk/.
Genotyping of the OncoArray was funded by the US NIH [U19 CA 148537 for ELucidating Loci Involved in Prostate cancer SuscEptibility (ELLIPSE) project and X01HG007492 to the Center for Inherited Disease Research (CIDR) under contract number HHSN268201200008I]. Additional analytic support was provided by NIH NCI U01 CA188392 (principal investigator: Schumacher).
The PRACTICAL consortium was supported by Cancer Research UK Grants C5047/A7357, C1287/A10118, C1287/A16563, C5047/A3354, C5047/A10692, and C16913/A6135, European Commission's Seventh Framework Programme grant agreement no.° 223175 (HEALTH-F2-2009-223175), and The NIH Cancer Post-Cancer GWAS initiative grant no. 1 U19 CA 148537-01 (the GAME-ON initiative).
We would also like to thank the following for funding support: The Institute of Cancer Research and The Everyman Campaign, The Prostate Cancer Research Foundation, Prostate Research Campaign UK (now Prostate Action), The Orchid Cancer Appeal, The National Cancer Research Network UK, and The National Cancer Research Institute (NCRI) UK. We are grateful for support of NIHR funding to the NIHR Biomedical Research Centre at The Institute of Cancer Research and The Royal Marsden NHS Foundation Trust.
This study is also funded by Johanna Schleutker Academy of Finland grant (#251074) and Gong-Hong Wei Academy of Finland grant (#284618 and #279760).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.