Prostate cancer is the most common noncutaneous male cancer with >250,000 diagnoses and 30,000 deaths each year in North America alone. Current prognostic variables including T category (of the TNM staging system), the absolute or kinetics of pretreatment prostatic-specific antigen (PSA), and the pathologic Gleason score (GS) are used to place men in low-, intermediate-, and high-risk prostate cancer risk groupings.1-3 However, despite attempts at optimal risk stratification, patients within the intermediate-risk group display significant heterogeneity with respect to clinical response, as 20% to 40% of patients will fail radical prostatectomy or external beam radiotherapy.2 Over the past decade, the Radiotherapy Oncology Group has conducted retrospective analyses of immunohistochemical (IHC) biomarkers associated with decreased prognosis after radiotherapy with or without hormone therapy. These include altered expression of MDM2, P16-INK4A, P53, KI-67, and mutations in the androgen receptor.4 The identification of novel gene or protein prognosticators in this particular group could individualize treatment to local treatment alone with or without adjuvant systemic treatment (eg, hormone therapy, chemotherapy, or molecular-targeted agents).
In an attempt to genetically place patients into local versus systemic risk categories, Lapointe et al used array-based comparative genomic hybridization (aCGH) to investigate prostate tissues and lymph node metastases after radical prostatectomy to define 3 subtypes of prostate cancer: 1 (subtype 1) was linked to clinically indolent behavior, whereas the 2 remaining subtypes were linked with nonindolent but localized and more aggressive metastatic prostate cancers (ie, subtypes 2 and 3, respectively).5 Specifically, subtype 1 contained prostate cancers with deletions on 5q and 6q and was associated with clinically indolent disease.5 In subtype 3, losses on 10q (PTEN) and gains on 8q (c-MYC) were associated with locally aggressive and metastatic disease. In other work, van Duin et al demonstrated in archival primary and metastatic prostate adenocarcinomas that 8q gain, containing the c-MYC oncogene, was associated with tumor progression.6 Similarly, Sato et al used fluorescent in situ hybridization (FISH) with a region-specific probe for c-MYC on 8q in 50 patients who had undergone radical prostatectomy. They reported that c-MYC gene amplification was strongly associated with higher GS, early disease progression, and prostate-cancer specific deaths.7 These and other studies link c-MYC allelic gain or protein overexpression to disease aggression and poor outcome after surgery.8, 9
Recently, we published data based on high-resolution aCGH from frozen biopsies in a pilot group of 20 men with intermediate-risk prostate cancer.10 We showed that this risk group contained previously reported copy number alterations associated with high-risk disease, including losses at 21q (TMPRSS2:ERG), 13q (RB1), 10q (PTEN), and 8p (NKX3.1), and gains at 8q21 (c-MYC), along with 6 novel microdeletions (eg, noted in >20% of the cohort at 1q, 5q, 20q, and 22q). As such, the genetic alterations found within this intermediate-risk group had overlap with genetic alterations previously documented in high-risk and metastatic disease.5, 10-12 If these copy number alterations correlate with outcome, the heterogeneous risk of occult metastases could explain the heterogeneity in biochemical outcome in this risk category.
Herein, we describe for the first time the role of PTEN loss and c-MYC gain using global high-resolution aCGH in a cohort of intermediate-risk prostate cancer patients after modern era, image-guided radiotherapy (IGRT). We were particularly interested in these loci because preclinical studies support aberrant signaling and DNA repair in tumor cells with changes in PTEN and c-MYC expression that lead to differential cellular radiosensitivity, chemosensitivity, or even synthetic lethality.13-18 Therefore, documenting copy number alterations in these alleles in intermediate-risk prostate cancer may provide a rationale for the use of molecular-targeted agents (eg, inhibitors of the MYC, PI3K-PTEN-AKT, or PARP proteins) in combination with IGRT.
MATERIALS AND METHODS
Patient Cohort and Radiotherapy Planning and Delivery
Two hundred forty-seven men with histologically confirmed adenocarcinoma of the prostate underwent a prospective clinical study, approved by the University Health Network Research Ethics Board (NCT00160979) and in accordance with the criteria outlined by the International Committee of Medical Journal Editors. From 1996-2001, pretreatment biopsies were derived from patients undergoing radical radiotherapy as primary treatment (REB#00-0443-C) at Princess Margaret Hospital. Before radiotherapy, patients underwent transultrasound-guided insertion of 3 intraprostatic gold fiducial markers for radiotherapy planning.10 At the same time, 3 research biopsies (2 for formalin fixation and 1 flash frozen10) were collected. Staging computed tomography (CT) and bone scans were not routinely performed on low- and intermediate-risk disease patients. Sufficient tumor to permit manual microdissection was identified in biopsies of 142 patients; 126 of those patients met intermediate-risk criteria as defined by D'Amico (T1-T2 disease, GS <8, and PSA <20 ng/mL)1 and had information on long-term biochemical outcome. Therefore, the final study included 126 patients. The primary outcome was time to biochemical failure, measured from the date treatment ended until a PSA rise of at least 2 ng/mL above postradiation nadir value as defined by Roach et al, or the last recorded date of PSA follow-up.19
The clinical target volume (CTV) encompassed the prostate gland alone. The planning target volume was defined by a 10-mm margin around the CTV except posteriorly, where the margin was 7 mm. All patients were treated with 6-field conformal or intensity-modulated radiotherapy. The radiotherapy dose was escalated over the period of accrual in a series of separate phase 1-2 studies. As summarized in Table 1, 33 patients (26%) received a dose of 75.6 grays (Gy) in 1.8-Gy daily fractions, and 78 (62%) received a dose of 78 to 79.8 Gy in 1.8- to 2-Gy daily fractions. Fifteen patients (12%) participated in a study of hypofractionated radiotherapy and received 60 to 66 Gy in 3-Gy daily fractions. Neoadjuvant and concurrent hormonal therapy was used in 33 patients (26%) as bicalutamide 150 mg orally once daily for 3 months before and 2 month concurrent with radiotherapy. None of the patients received adjuvant hormonal treatment. Patients were followed at 6-month intervals after completing treatment with clinical examination and PSA. Additional tests and the management of patients with recurrent disease were at the discretion of the treating physician. The median follow-up of surviving patients was 6.7 years from the end of treatment.
Table 1. Multivariate Analysis of the Association of T Category, Gleason Score, and Pretreatment PSA With the Time to Recurrence in the 126 Intermediate-Risk Patients: Clinical Model
Fresh needle biopsies were embedded in OCT, frozen at −80°C, and cut into 10-μm sections. Hematoxylin and eosin-stained sections were reviewed and marked by an experienced urologic pathologist (T.v.d.K.), who marked areas of tumor suitable for manual microdissection. Sections with <70% tumor were not microdissected. Manual microdissection was performed using sterile scalpel blades. Postdissection slides were stained with hematoxylin and scanned using Aperio (Vista, Calif) digital imaging software to confirm correct areas of microdissection. DNA was extracted according to standard procedures and quantified, and the quality was assessed by gel electrophoresis. Hybridizations were performed using the same amount of input sample DNA. All other reagents used were identical in quantity.
DNA labeling and hybridization were performed essentially as described previously.10 Briefly, 300 ng of sample and reference DNA were labeled in a random priming reaction with cyanine 3-deoxycytidine triphosphate (dCTP) and cyanine 5-dCTP in the dark at 37°C for 16 to 18 hours. DNA samples were mixed with 100 μg of human Cot-1 DNA. Unincorporated nucleotides were removed with YM-30 columns. The mixture was applied onto arrays containing 26,819 bacterial artificial chromosome (BAC)-derived amplified fragment pools spotted in duplicate on aldehyde-coated glass slides (SMRT v.2, BC Cancer Research Centre Array Facility, Vancouver, British Columbia, Canada). Slides were scanned using a dual laser array scanner (Axon Instruments, Union City, Calif), and spot signal intensities were determined using SoftWoRx Tracker Spot Analysis software (Applied Precision, Issaquah, Wash). The log2 ratios of the cyanine 3 to cyanine 5 intensities for each spot were assessed. Data were filtered based on both standard deviations of replicate spots (data points with >0.075 standard deviation were removed) and signal to noise ratio. The resulting data set was normalized using a stepwise normalization procedure.20 Areas of aberrant copy number were identified using a Hidden Markov Model21 and classified as either loss, neutral, or gain for all clones processed. Preprocessed/postprocessed log2 ratios of intensities for clone regions were observed using SeeGH software.22, 23 The genomic positions of clones are mapped to the National Center for Biotechnology Information's Genome Build 36.1, released in March 2006. Percentage genome alteration used for analysis was defined as the cumulative size of the genetic alterations found in each patient DNA sample divided by the total size of the human genome.
Interphase FISH was applied to formalin-fixed paraffin-embedded prostate cancer biopsies to validate the genomic imbalances associated with c-MYC gain and PTEN, 5q, and 6q deletions in tumors. For chromosome 8q, LSI c-MYC (8q24.12-q24.13, SpectrumOrange) and CEP8 (D8Z2, SpectrumGreen; Abbott Molecular, Inter Medico, Markham, Ontario, Canada) were used. BAC clones and commercially available probes were used for 5q22.2, 6q15, and Pten deletions. BACs on chromosome 5 were: 1) Apc located at 5q22.2: RP11-124K18 and RP11-159K7; 2) 5q14.1 (control region): RP11-91I17 and RP11-1H14. BACs on chromosome 6 were: 1) Casp8ap2 located at 6q15: RP11-63K6 and RP11-876H4; 2) commercially available CEP6 (D6Z1) SpectrumAqua probe (Abbott Molecular). BACs on chromosome 10 were: 1) commercially available CEP10 SpectrumAqua probe (Abbott Molecular); 2) Bmpr1a located at 10q23.2: RP11-141D8 and RP11-52G13; 3) Pten located at 10q23.31: RP11-846G17; and 4) Fas located at 10q23.31: RP11-399O19 and RP11-360H20. The correct chromosome locations of the BAC clones was verified by hybridization to metaphase spreads from normal peripheral lymphocytes. BAC DNA was labeled by nick-translation with SpectrumGreen-deoxyuridine triphosphate (dUTP), SpectrumOrange-dUTP, or SpectrumRed-dUTP. The establishment of gene copy number status was defined by considering the adjacent or control probes used for the truncation artifacts, aneusomy, nuclear size, and chromatin condensation. Apc (5q22.2), Casp8ap2 (6q15), and Pten (10q23.31) genomic deletions were evaluated for each probe by counting spots in 100 nonoverlapped, intact interphase nuclei per tumor tissue.
Five-year biochemical relapse-free rates were calculated using the Kaplan-Meier method. The log-rank test was used to evaluate the association of PTEN status (loss or normal), 5q status (loss or normal), 6q status (loss or normal), and c-MYC status (gain or normal) with biochemical relapse-free rate. Multivariate PTEN and c-MYC were also tested adjusting for clinical factors (T category, pre-PSA, and GS) using the Cox proportional hazards model. A concordance index (c index) was also calculated to measure how well the model predicts for biochemical relapse-free rate with and without the addition of information pertaining to genetic factors. Differences in percentage genome alteration among patients with genetic aberrations were examined using the Mann-Whitney-Wilcoxon test. A P value of .05 was used to assess statistical significance.
Characteristics of Patient Radiotherapy Cohort and Clinical Prognostic Factors
A total of 126 men with prostate cancer received radical radiotherapy at Princess Margaret Hospital between 1996 and 2001. All patients were in the intermediate-risk group with T1 or T2 and GSs of 6 and 7. The median pretreatment PSA of the group was 7.8 (range, 0.90-19.00). The majority of patients (88.1%) received doses of ≥75 Gy, with a mean biological equivalent dose of 76.4 Gy in 2-Gy fractions.
We first tested T category (T1 vs T2), pretreatment PSA (dichotomized or as a continuous variable), and GS (6 vs 7) as prognostic factors for biochemical outcome in our intermediate-risk cohort. In a multivariate analysis, only pretreatment PSA tested as a continuous variable predicted for relapse, with a hazard ratio (HR) of 1.11 (P = .003; Table 1). These results are consistent with the patients being in a tight prognostic category based on T category and GS, where the dynamic variable within the analysis is the range of pretreatment PSA values.
Copy Number Alterations in PTEN and c-MYC are Common in Intermediate-Risk Prostate Cancer
We used the aCGH results of the 126 patients to test the Lapointe categories of indolent localized disease versus metastatic disease.5 In our cohort, there were 44 patients who had losses of 5q or 6q and therefore were in category subtype 1. In subtype 3, the potentially most unfavorable category, 44 patients of the cohort were placed based on amplification of 8q24 (c-MYC) or a loss at 10q (PTEN; Fig. 1B, C). Although other deletions or gains were noted by Lapointe et al5 for subtyping, we focused on these main changes as those that identified the most disparate subtypes to be useful in clinical pathologic and outcome analyses as outlined in Figure 1A.
We had previously verified that aCGH detected copy number alterations at loci for c-MYC and PTEN based on confirmatory FISH for their specific loci on 8p24 and 10q.10 In Figure 2, we show FISH analyses of an audit of aCGH results on tissues from 3 patients in this study who had a loss of PTEN, gain of c-MYC, and/or loss of 5q/6q on aCGH analysis. Despite the biopsies for FISH and aCGH analyses being taken from different cores, there appears to be good correlation between these results within each patient.
Copy Number Alterations in 5q, 6q, 8q, and 10q Are Associated With Increased Genomic Instability
We first tested for associations between subtype 1 (5q, 6q deletions) and subtype 3 (10q-PTEN deletions and 8q- c-MYC gain) copy number alterations versus percentage genome alteration. Figure 3A shows that patients with PTEN and c-MYC alterations have significantly increased percentage genome alteration values compared with those with normal PTEN and c-MYC status (P < .0001). Of interest, in our cohort the patients with 5q and 6q deletions (Lapointe subgroup 1) also had increased percentage genome alteration values compared with patients without these alterations (P < .0001; Fig. 3B). We then tested whether percentage genome alteration was associated with GS within an aCGH category in Figure 4. There was a trend toward percentage genome alteration associated with GS in patients with 8q-c-MYC gains or 5q or 6q deletions, but not with 10q-PTEN deletions. Patients with no alterations in 10q-PTEN, 8q-c-MYC, 5q, or 6q had the lowest percentage genome alterations. Percentage genome alteration did not segregate with GS.
Patients With Alterations in 10q-PTEN, 8q-c-MYC, 5q, and 6q Are More Likely to Relapse After IGRT
Finally, we then tested the Lapointe subtypes 1 and 35 as potential predictors of radiotherapy response given the hypothesis that these 2 subtypes may reflect indolent local disease versus metastatic disease, respectively. The univariate plots for subtype 3 are shown in Figure 5A to C, in which Pten deletions did not predict for relapse.
We also tested whether Lapointe subtype 1 (5q or 6q deletion)5 would predict for a favorable response to radiotherapy if this subtype was locally indolent and/or lacked the propensity for metastases. However, the analysis shows that patients with 5q and 6q deletions and no PTEN/c-MYC aberrations do not have a decreased relapse rate compared with patients without these deletions on univariate analysis, as shown in Figure 5D. Furthermore, in an exploratory fashion, we tested a mixture of subtypes 1 and 3 in univariate analysis and observed that PTEN + c-MYC patients who had 5q and 6q deletions did not have a significantly worse outcome after radiotherapy than PTEN + c-MYC patients without 5q and 6q deletions (Fig. 5E). Finally, in sequential multivariate analyses, copy number alterations in c-MYC alone, or combined PTEN + c-MYC alterations, predicted for relapse on multivariate analyses (HRs, 2.58 and 3.21; P = .005 and .0004; Tables 2-4). A Cox proportional hazards regression model using only standard prognostic factors (PSA, GS, and T category) had a c index (a measure of how well the model predicts for biochemical relapse-free rate) of 0.640 and overall score test P value of .023. When a variable representing at least 1 alteration of c-MYC or PTEN was added to the model, the c index increased to 0.698, and overall score test was P = .0001. Therefore, further classification of patients based on PTEN allelic loss and/or c-MYC gain improves on the ability to predict radiotherapy failure in this cohort over clinical factors, alone.
Table 2. Multivariate Analysis of the Association of PTEN, c-MYC, T Category, Gleason Score, and Pretreatment PSA With the Time to Recurrence in the 126 Intermediate-Risk Patients: Effect of PTEN When Adjusting for the Clinical Factors
Table 3. Multivariate Analysis of the Association of PTEN, c-MYC, T Category, Gleason Score, and Pretreatment PSA With the Time to Recurrence in the 126 Intermediate-Risk Patients: Effect of c-MYC When Adjusting for the Clinical Factors
Table 4. Multivariate Analysis of the Association of PTEN, c-MYC, T Category, Gleason Score, and Pretreatment PSA With the Time to Recurrence in the 126 Intermediate-Risk Patients: Effect of Overall PTEN/c-MYC (at Least 1 Aberrant) When Adjusting for the Clinical Factors
Validation: Patients With PTEN Allelic Loss Are Not Statistically More Likely to Relapse After Surgery
We next interrogated the publically available Memorial Sloan Kettering Prostate Cancer database (http://www.cbioportal.org/cgx/index.do?cancer_type_id = pca) and built a validation cohort consisting of 133 intermediate-risk patients who subsequently underwent surgical therapy. Univariate analysis showed that patients whose tumors displayed c-MYC gain did not have a statistically significant difference in failure at 5 years as compared with patients with normal c-MYC status (biochemical relapse-free rate 83% vs 74%; P = .36; Fig. 6A). The HR associated with c-MYC gain alone was 1.65 (95% confidence interval [CI], 0.56-4.82). Patients whose tumors displayed PTEN loss also did not have a statistically significant difference in failure at 5 years as compared with patients with normal PTEN status (biochemical relapse-free rate 85% vs 61%; P = .19; Fig. 6B). The HR associated with PTEN loss alone was 1.9 (95% CI, 0.71-5.07). Furthermore, patients whose tumors displayed PTEN loss and c-MYC gain did not have a statistically significant difference in failure at 5 years as compared with patients with normal loci status (biochemical relapse-free rate 86% vs 69%; P = .18; Fig. 6C). The HR associated with PTEN loss and c-MYC gain was 1.78 (95% CI, 0.77-4.12). In summary, in the surgical cohort there appears to be a trend toward failure; however, this was not statistically significant.
To our knowledge, this is the first report associating PTEN loss and c-MYC gain with biochemical failure after IGRT; the latter planning technique reduces geometric miss and therefore allows for better biological assessment of novel prognostic factors in prostate cancer radiotherapy. Our data support the study by Lapointe and colleagues5 and other groups,9, 24-27 in which a combination of c-MYC gain and PTEN loss is prognostic for poor prognosis after local therapy. We investigated a cohort of men who had intermediate-risk prostate cancer based on classic prognostic variables of pretreatment PSA, T category, and GS, yet still show clinical heterogeneity in treatment response. As expected in a tight prognostic cohort, only increasing pretreatment PSA as a clinical factor was associated with biochemical failure in a multivariate analysis, with a significant HR of 1.1. However, patients could be further subclassified into a more aggressive subgroup on the basis of PTEN allelic loss and c-MYC gain with an HR of 3.2, whereby our analysis shows that the genetic information adds to the clinical factors in delineating prognosis in the intermediate-risk group. This approach supports the use of genetic biomarkers, which are independent of the current prognostic variables, as they could further subclassify intermediate-risk patients into those patients with suspected occult metastases or radioresistant disease. Preclinical models in which the same genetic loci are altered may help define whether these patients would benefit from an improvement in local and/or systemic control using novel molecular-targeted therapies in combination with IGRT.
The failure of patients after prostate IGRT could be associated with loss of PTEN and gain of c-MYC through altered radiation-induced intracellular signaling or DNA damage-repair response pathways. These alterations are potentially amenable to the individual use of molecular targeted therapies.14 Radioresistance in vitro has been associated with loss of PTEN and was attributed to up-regulated pAKT signaling and other survival pathways.28, 29 In terms of DNA damage response, the MRE11 protein (involved in the repair of DNA double-strand breaks [DSBs]) can drive the local activation of p-AKT at DSBs.30 A loss of PTEN could result in elevated pAKT levels and altered sensing or repair of DSBs during clinical radiotherapy. Other reports have demonstrated loss of RAD51 expression in PTEN-null cells31; however, these observations have recently been challenged.32 A recent study from our group13 challenged whether aberrant c-MYC expression and transcription could alter DSB and radiosensitivity in vitro, but we have not confirmed this data using in vivo radioresponse assays. As the prostate is still in situ when the determination of PSA failure is preformed, we cannot rule out the hypothesis that failure in the PTEN/c-MYC patients is because of regrowth of local radioresistant clones soon after therapy. In a future prospective study, one could partially address this by acquiring postradiation biopsies at 2.5 to 3 years after radiotherapy for all patients to confirm whether there is local recurrence at the time of biochemical failure.2
Our clinical failure data in Figure 5 suggest that PTEN/c-MYC patients fail early after completion of radiotherapy, and we speculate that this may be in part because of occult metastatic disease pre-existing at the time of therapy. There are at least 3 potential implications of our clinical findings for patients with these copy number alterations at PTEN/c-MYC loci: 1) such patients should be considered for trials that combine IGRT and androgen deprivation, in addition to radiotherapy, to offset the poor prognosis associated with radiotherapy alone; 2) such patients should be considered for novel agents molecularly targeted directly at abnormal PTEN-mTOR-AKT or MYC signaling axes in combination with IGRT; and 3) such patients may be candidates for trials that test the utility of bone scans and CT scans in the intermediate-risk group who are uniquely prone to metastases. The sensitivity of these tests might be improved in subgroups of patients who are at greater risk for suboccult disease at the time of starting treatment. However, a subanalysis of our patients with c-MYC allelic gain who did or did not receive hormone therapy in addition to radiotherapy did not show a benefit for combined modality therapy (data not shown). Although this would have to be validated in a prospective predictive study, other systemic agents might have increased efficacy when directed toward the individual patient based on an idiosyncratic tumor genetic signature for a given signaling pathway.
Although future studies will use an unbiased approach to our aCGH data to attempt to derive novel aCGH prognostic signatures, this initial study was hypothesis-based toward the prognostic role of PTEN, c-MYC, and 5q21/6q15 in tumors. This group of chromosomal changes was suggested by Lapointe et al as a classifier of increasing GS and the presence of metastasis, allowing for the triaging of patients into indolent versus aggressive disease and potentially metastatic disease.5 However, they did not study relative long-term outcome in the patients. In further support of their model, we found that the loss of 5q21 or loss of 6q15 was not associated with increased failure on a univariate analysis. We note that alterations in all 4 loci were confirmed in an audit of selected cases in situ using FISH analyses. In these more sensitive studies, both biallelic and monoallelic losses were detected by FISH within the same specimens; in certain cases, these regions were in close proximity. Creating validation-based tissue microarrays in future radiotherapy or surgery cohorts with localized disease could test the utility of high-throughput FISH assays or protein IHC assays to rapidly characterize the PTEN and/or c-MYC status in patients undergoing radical radiotherapy and personalize treatment. The study of prostate genomic signatures that relate to outcome could be further enhanced by the use of next generation deep sequencing methodology, and such studies are ongoing within the Canadian Prostate Cancer Genome Network and the International Cancer Genome Consortium (www.icgc.org).
In summary, our results confirm and expand the findings by Lapointe et al by suggesting that within the intermediate-risk group category, information regarding PTEN and c-MYC loci may improve individual prognostication. Other loci that may be of interest include whole-gene mutational analyses of the tumor suppressor gene, p5333 or TMPRSS2-ERG fusions,34 as these loci have been linked to poor prognosis after radical radiotherapy or surgery. Importantly, our data support the concept that pretreatment biopsies can be used for genetic studies. If validated, this information has the potential to determine individual risk for biochemical failure after IGRT and the requirement for targeted systemic treatments.
We thank C. Sanders and N. Schultze for helpful comments on negotiating the Memorial Sloan-Kettering Cancer Center cBio portal for our validation work.
Supported by a grant from the Ontario Institute for Cancer Research and Canadian Foundation for Innovation grant to the STTARR Innovation Facility. This research was funded in part by the Ontario Ministry of Health and Long Term Care. The views expressed do not necessarily reflect those of the Ontario Ministry of Health and Long Term Care. R.G.B. is a Canadian Cancer Society Research Scientist. J.A.L. is in the Comprehensive Research Experience for Medical Students program at the University of Toronto.