The Combination of DNA Ploidy Status and PTEN/6q15 Deletions Provides Strong and Independent Prognostic Information in Prostate Cancer

Prostate cancer is the most prevalent cancer in men in Western societies (1). While most cancers have a rather indolent clinical course, prostate cancer still represents the third most common cause of cancer-related death in men. Despite recent advances, the only established pretherapeutic prognostic parameters currently include Gleason grade and tumor extent on biopsies, preoperative prostate-specific antigen (PSA), and clinical stage. Because these data are statistically powerful but not sufficient for optimal individual treatment decisions, there is a tremendous need for better prognostic parameters enabling a more reliable prediction of the aggressiveness of individual prostate cancers.

Measurement of the cellular DNA content by either flow cytometry or static image cytometry has been extensively discussed as a potential prognostic tool in many cancer types in the 1980s and early 1990s of the last century (2–6). At that time, ploidy measurement represented the best possible method to globally assess tumor DNA content. It was postulated that a non-diploid DNA status would indicate genomic instability, which in principle should be linked to an aggressive cancer phenotype. Various studies have demonstrated that tetraploidy and aneuploidy are associated with unfavorable disease outcome in many cancer types, such as cancers of the colon, breast, oral cavity, urinary bladder and lungs (2–6). However, cytometry has not become a routine tool in any of these cancers, mostly because the prognostic information was clinically not relevant or because better predictors were identified. Moreover, current methods, for example fluorescence in situ hybridization (FISH) or—more recently—next-generation sequencing (NGS), enable detection of much more subtle genomic alterations and have thus recently gained more attention by researchers.

Studies evaluating DNA ploidy in prostate cancer have come to varying conclusions. Some studies on 45–850 (mean 139) cancers have suggested a relevant prognostic impact of ploidy (7–20), while other studies on 33–447 (mean 122) cancers have failed to find significant associations (7, 20–24). Deletions affecting various chromosomal loci are a hallmark of prostate cancer. In particular, deletion of PTEN and 6q15 has been linked to adverse tumor phenotype and early biochemical recurrence in several studies, including our own work (25–28). In this study, we took advantage of our large Hamburg prostate cancer cohort to evaluate the potential impact of DNA cytometry data in a large consecutive series of cancers. Moreover, we evaluated the combined impact of measuring both a generally disturbed DNA content (ploidy status) and presence of subtle genomic alterations—such as small chromosomal deletions with known prognostic relevance, including PTEN and 6q15 (25, 28). The data demonstrate that quantitative DNA measurement in combination with deletion measurement is a powerful predictor of prognosis in prostate cancer.

Radical prostatectomy specimens were available from 3,845 patients for which both PTEN and 6q15 deletion status had been successfully determined in earlier studies (25, 28). All patients underwent surgery between 1992 and 2012 at the Department of Urology and the Martini Clinics at the University Medical Center Hamburg-Eppendorf. Follow-up data were available from a total of 3,845 patients with a median follow-up of 45.1 months (range, 0–240.4 months). The clinical and pathological parameters of the prostate cancers are given in Table 1. PSA values were measured following surgery and PSA recurrence was defined as a postoperative PSA of 0.2 ng/mL and increasing at subsequent measurements. In addition, patients were grouped into 4 subsets defined by clinically and biologically important features: Group 1 “organ confined tumor growth” included 444 patients with organ-confined tumors and no evidence of local or systemic dissemination [no histological sign of extraprostatic extension and no biochemical relapse (BCR) in long-term follow up (≥3 years after surgery)], group 2 “local invasive tumor growth” included 172 patients with histological proof of extraprostatic tumor growth (pT3a or pT3b), but no BCR in long-term follow-up, or pT3 tumors with BCR but permanent response to salvage radiation, group 3 “occult systemic dissemination” included 307 cancers with BCR after two local therapies (radical prostatectomy and secondary—adjuvant or salvage—radiation), but without development of overt distant metastases in long-term follow-up, and group 4 “metastatic tumor growth” included 119 patients with development of distant (bone and/or visceral) or regional (lymph node) metastases. The remaining 2,803 cancers had not sufficient clinical information to allow for unequivocal allocation to these four groups. All prostate specimens were analyzed according to a standard procedure, including a complete embedding of the entire prostate for histological analysis (29).

Cell nuclei were extracted from formalin-fixed paraffin-embedded tissues and stained for flow cytometry analysis using a modified standard protocol (16, 21, 30). Two punches (diameter 0.6 mm) per tissue block were taken with a hollow needle from 3,845 prostate cancers included in this study. Special emphasis was placed on taking the punches from the same tumor area as used for PTEN/6q15 deletion analysis before. For preparation of cell suspensions, punches were re-embedded in paraffin, and cut into 50-μm sections using a microtome. The sections were placed in a nylon bag with a mesh of 50 μm, dewaxed in xylene, and rehydrated in a descending series of ethanol (100%, 96%, 70%, water) prior to digestion in 5 mg/mL pepsin (Serva #31820.02), and resolved in 0.07 mol/L HCl in a 37°C water bath for 30 minutes. Five milliliter cold phosphate buffered saline was added to stop the reaction. The suspension was centrifuged at 2,500 rpm to release cell nuclei from the nylon bags, resuspended to a volume of 450 μL and transferred to 5 mL FACS tubes. RNAse A (Sigma #R4875) was added to a final concentration of 0.05 mg/mL (adjusted to pH 7.4) before incubation at 37°C for 30 minutes. Nuclei were then stained by adding 100 μL propidium iodide solution (1 mg/mL) (Sigma, #P4864) for 5 minutes at 4°C in the dark. The DNA content was measured in at least 1,000 stained nuclei using a FACS Canto II using the blue laser (488 nm) with the filter configuration longpass 556 and bandpass 585/42.

DNA flow histograms were interpreted following standard criteria as described elsewhere (16, 17, 22). In brief, the first analyzable peak, indicating the fraction of cells with the lowest DNA content, was considered DNA diploid and was given a DNA index of 1.0. DNA indices for the following peaks were calculated relative to the 1.0 peak. Samples were considered DNA aneuploid when one or more unequivocal peaks between DNA index 1.0 and 1.8 or >2.2 was present. Tetraploidy was defined as the presence of one unequivocal peak between DNA index 1.8 and 2.2 with more than 10% of all cells. Otherwise, the sample was considered DNA diploid.

The CAPRA-S score was calculated for the entire set of 3,845 prostate cancers as described before (31).

Statistical calculations were performed using JMP 11 software (SAS Institute Inc.). Contingency tables and the χ² test were performed to search for associations between molecular parameters and tumor phenotype. Analysis of variance (ANOVA) analysis was performed to compare the preoperative PSA levels with molecular and cytometry data. Survival curves were plotted according to Kaplan–Meier. The log-rank test was applied to detect significant differences between groups. Cox proportional hazards regression analysis was performed to test the statistical independence and significance between pathological, molecular, and clinical variables. Gleason grade (in biopsies and radical prostatectomies), clinical (cT) and pathological tumor stage (pT), pathological nodal stage (pN), resection margin stage, and preoperative PSA levels were selected for multivariate analysis because they represent the clinically most relevant variables. Logistic regression was used to quantify the area under the receiver-operator curve (ROC).

Between 1,043 and 85,049 cell nuclei were measured per tumor (mean 7,282 ± 2,317). The average coefficient of variation was 8.03 ±2.05 for the first peak. DNA cytometry revealed a diploid DNA content in 2,605 of 3,845 (67.8%), a tetraploid DNA content in 984 (25.6%), and an aneuploid DNA content in 256 (6.7%) of interpretable cancers. Aberrant ploidy was significantly associated with pT category, Gleason grade, clinical T stage, R status, pN category, and clinical endpoints (P < 0.0001 each; Table 1). Of note, among 218 patients with a Gleason score of ≥4+4 = 8; 60.1% had an abnormal DNA content, but there were still 39.9% of these highly aggressive cancers with a normal diploid DNA content. Abnormal DNA content was also tightly linked to early PSA recurrence (P < 0.0001; Fig. 1A).

Deletions of PTEN and 6q15 were earlier shown to be linked to unfavorable tumor phenotype and PSA recurrence in our patients (25, 28). The respective data for the 3,845 cancers with interpretable cytometry profiles are shown in Table 1. As expected, significant associations were seen with phenotypical parameters and PSA recurrence (P < 0.0001) for both parameters.

The combination of crude DNA alterations (ploidy) and subtle specific changes (deletions) leads to substantial improvements of associations with phenotype and prognosis. We create a “3-level DNA score” for the combination according to the following criteria: negative, DNA diploid and no deletion; intermediate, DNA diploid with deletion (of at least one of PTEN and 6q) or DNA tetraploid without deletion; high, DNA tetraploid with deletion or DNA aneuploidy. As for deletions and ploidy alone, the combined approach revealed a tight link with unfavorable phenotype (Table 1). It is remarkable that the fraction of cancers with Gleason score ≥4+4 with a diploid DNA content and no deletion decreased to 21.1% in this combinatorial approach. The relevant effect of deletions over ploidy alone is best demonstrated in the PSA recurrence data (Fig. 1H–J). Here, the presence of deletions resulted in a massive deterioration of outcome in the subgroups of diploid, tetraploid, and aneuploidy cancers. The combination of ploidy and deletions was also tightly linked to our alternative clinical endpoints (P < 0.0001; Fig. 2A).

Separate analyses in subgroups of cancers with Gleason score ≤3+3, 3+4 (with and without tertiary Gleason 5), 4+3 (with and without tertiary Gleason 5), and Gleason ≥4+4 revealed that the assessment of DNA alterations was particularly relevant in Gleason 3+4 cancers (Fig. 1C), while the effect was less relevant and failed to reach statistical significance in the other subgroups. Also the analyses of our alternative clinical endpoints show that the effect was only relevant in the Gleason 3+4 subset, too (Fig. 2B).

Four multivariate analyses were performed evaluating the clinical relevance of our “3-level DNA score” in different scenarios (Table 2). Scenario 1 was utilizing all postoperatively available parameters, including pathological tumor stage, pathological lymph node status (pN), surgical margin status, preoperative PSA value, and pathological Gleason grade obtained after the morphological evaluation of the entire resected prostate. Scenario 2 was utilizing all postoperatively available parameters with exception of the nodal status. The rationale for this approach was that the indication and extent of lymph node dissection is not standardized in the surgical therapy of prostate cancer and that excluding pN in multivariate analysis can markedly increase case numbers. Two additional scenarios had the purpose to model the preoperative situation. Scenario 3 included the “3-level DNA score,” preoperative PSA, clinical tumor stage (cT stage), and Gleason grade obtained on the prostatectomy specimen. Because postoperative determination of a tumor's Gleason score is more accurate than the preoperatively determined Gleason grade (subjected to sampling errors and consequently under-grading in more than one third of cases), another multivariate analysis was added. In scenario 4, the preoperative Gleason grade obtained on the original biopsy was combined with preoperative PSA, cT stage, and the “3-level DNA score.” In all these analyses, the “3-level DNA score” emerged as an independent predictor of prognosis with high statistical significance. ROC analyses were then performed to estimate whether the “3-level DNA score” can improve the predictive power in the 4 different clinical scenarios in subsets of 2,392 to 3,780 cancers for which data on these parameters and ploidy/deletion were available. These analyses revealed a gain of the predictive accuracy by 0.6% in the strongest scenario 1, 1.3% in scenario 2, 1.6% in scenario 3, and 2.6% in scenario 4 containing the preoperative prognostic parameters (Gleason at biopsy, clinical stage, preoperative PSA; Table 3).

High scores of the established clinical CAPRA-S nomogram were strongly linked to early biochemical recurrence in our cohort (P < 0.0001; Supplementary Fig. S1a). In order to estimate if our “3-level DNA score” can improve the predictive power beyond the nomogram, Kaplan–Meier plots of the 3-level DNA score were prepared in subsets of cancers with an identical CAPRA-S score. This analysis revealed that the assessment of DNA alterations provided additional prognostic information beyond the nomogram. The “3-level DNA score” added significant prognostic information in subsets with CAPRA-S scores 0–1 (P = 0.0007; Supplementary Fig. S1b), 2–3 (P = 0.0016; Supplementary Fig. S1c), 4–5 (P = 0.0278; Supplementary Fig. S1d), and 6–8 (P = 0.0010; Supplementary Fig. S1e). For the highest score (9–12), there was at least nonsignificant trend toward a poor prognosis for cancers with an intermediate or high 3-level DNA score (P = 0.4273; Supplementary Fig. S1f).

DNA flow cytometry revealed 6.8% aneuploid and 25.6% tetraploid prostate cancers in this study. All earlier studies analyzing series of at least 30 cases are summarized in Fig. 3. These data show that our rate of non-diploid cancers is in the range of earlier studies analyzing consecutive patient cohorts (9, 11–14, 16–19, 22). Markedly higher rates—ranging from 39.7% to 74.8%—were mainly found in studies enriched for advanced (7, 10, 14, 16, 17, 23, 32) or high grade cancers (22, 33). Some of the studies resulting in particularly low frequencies of non-diploid cancers had preselected for T1a or clinically localized cancers (7, 8, 13, 16, 20, 34).

Non-diploid DNA status was tightly linked to unfavorable cancer features, such as high Gleason grade, advanced tumor stage, and positive nodal stage. Comparable findings had been reported from the vast majority of earlier studies. Most of the larger studies analyzing at least 100 cancers reported associations with Gleason grade (13, 16, 18), other grading systems (7, 13, 16, 20), or tumor stage (13,16,18). Accordingly, DNA ploidy was also strongly associated with PSA recurrence at least in some of these studies (7, 10). Similar findings were also made in earlier and smaller studies involving fewer than 100 patients (7, 12, 17, 19,23, 32, 34). Given the undisputed link of non-diploid DNA with unfavorable tumor phenotype, it is likely that those studies on 33 (23), and 81 (24) patients failing to find associations with PSA recurrence were disturbed by low patient numbers.

Based on this obvious link of non-diploid DNA status and poor tumor phenotype and clinical outcome, ploidy measurement was earlier suggested for routine use in the diagnostic work-up of biopsies (35, 36). It was claimed that ploidy data could compensate for possible mistakes made by the pathologist in his Gleason grading, which is a very strong prognostic feature but notorious for subjectivity and interobserver variability (37–39). That ploidy measurement by itself cannot make Gleason grading “safe” in its capability to identify the “dangerous” cancers is, however, demonstrated by the 39.9% of Gleason ≥4+4 cancers (known to have a dismal prognosis), and the 55.2% of cancers with PSA recurrence, having a diploid DNA content. Given that many subtle DNA changes—such as small chromosomal deletions—are strongly linked to unfavorable prostate cancer prognosis (25, 26, 40), and considering that a loss of a small fragment of a chromosome will not become detectable by DNA flow cytometry, it is not surprising that many aggressive prostate cancers are DNA diploid.

We therefore hypothesized that adding powerful prognostic deletion markers to ploidy analysis would optimize our assay for identifying biologically aggressive prostate cancers. Deletions of PTEN and 6q15 were selected for this purpose because both of them are strongly linked to unfavorable disease outcome and also because these deletions cover relevant molecular subgroups (25–28). PTEN deletions preferably occur in ERG fusion–positive and 6q15 deletions in ERG fusion–negative cancers (25, 28). That the addition of these deletions to pure ploidy data may be advantageous for an approach to identify potentially aggressive cancers is already suggested by the much lower rate of cancers with Gleason ≥4+4 (now 21.1% as compared with 39.9% without considering deletions) and of cancers with PSA recurrence (now 29.0% as compared with 55.2% without considering deletions) when deletion analysis is combined with ploidy analysis. The strong additional benefit of our combined approach is further demonstrated by the striking difference in the risk of PSA recurrence in each individual subgroup defined by DNA ploidy (diploid, tetraploid, aneuploid) depending on whether or not PTEN and/or 6q15 deletions were additionally present.

Limitations of this study include the use of PSA recurrence as a clinical endpoint instead of cancer specific death, which was a rare event in our cohort, and that the analysis was performed on prostatectomy specimens instead of true core needle biopsies. We do not feel that these issues question the message of this study. Specifically, we believe that PSA recurrence is an excellent surrogate marker if it comes to the assessment of prognostic biomarkers. From our experience, all prognostic parameters predicting PSA recurrence in prostatectomy samples are also linked to other endpoints such as time to metastasis, time to cancer-related death, or unfavorable disease course in patients not undergoing potentially curative therapy. Gleason grade represents the best example for this notion. It is well accepted that high Gleason grade is associated with all possible unfavorable clinical endpoints. To our knowledge, there is no molecular or morphologic feature that was shown—in sufficiently large studies—to be robustly linked to PSA recurrence but to be unrelated to metastasis and cancer-related death. Molecular parameters that have been described to be linked to metastasis and cancer-related death in patient cohorts describing at least 100 patients with unfavorable events include p53 (41, 42), bcl-2 (41), microvessel density (41), Ki-67 (42), CD10 (43), cyclin D1 (44), telomere length (45), HER2-neu (46), Stat5a/b (47), AMACR (48), HSP-27 (49), and MDM2 (42). All these parameters have also been linked to PSA recurrence in the respective studies (44, 46–48).

In an attempt to analyze additional and possibly better clinical endpoints in prostatectomy samples, clinical subgroups were defined that reflect the “biological milestones” of cancer progression, including (i) organ-confined tumors (pT2) without relapse in long-term follow-up, (ii) “local invasive” cancers (pT3) without relapse or with permanent response to local secondary radiation in long-term follow-up, (iii) “occult systemic” disease characterized by failure of two local therapies (radical prostatectomy and secondary radiation), but no evidence of distant metastases in long-term follow-up, and (iv) metastatic disease characterized by presence or development of regional or distant metastases. These subgroups are clinically valuable as only group 1 is potentially suited for an active surveillance strategy. That our “3-level DNA score” was strongly associated with unfavorable alternative clinical endpoints (Fig. 2A) provides additional support for the potential clinical relevance of the combination from DNA ploidy, PTEN, and 6q15 deletions.

It is obvious that studies evaluating prognostic biomarkers should optimally be performed on pretherapeutic tissue samples (i.e., prostate biopsies) rather than on samples obtained after definitive therapy (i.e., prostatectomy). From a practical point of view, such analyses are hardly feasible, however. This is not only because initial biopsies from cancer patients are typically distributed among many diagnostic institutions but also because the precious needle biopsies would be exhausted after only a few studies. To mimic as much as possible the presurgical situation, we have performed multivariate statistical analyses that use parameters available before surgery in addition to statistical models using post-surgical parameters. That strong independent associations of our “3-level DNA score” with early PSA recurrence was found in all analyzed pre- and post-surgical scenarios strongly argues for this combined DNA score representing a biomarker with clinical relevance in prostate cancer. It was therefore no surprise that our DNA score also provided a benefit of prognostic information beyond an established clinical nomogram, i.e., the CAPRA-S score.

In summary, the results of our study demonstrate that measurement of the gross DNA content in combination with analyzing two specific deletions by FISH provide strong and independent prognostic information in prostate cancer. This especially holds true for the subgroup of Gleason 3+4 cancers, which is the most disputed subset with respect to a possible inclusion in active surveillance concepts. In contrast to most previously published prognostic features, e.g., based on RNA or protein expression analyses, FISH and flow cytometry provide highly reproducible yes/no answers. We thus expect that future trials utilizing this combined DNA measurement for prediction of cancer aggressiveness on pretherapeutic prostate cancer biopsies will prove strong utility of our approach.

No potential conflicts of interest were disclosed.

Conception and design: R. Simon, G. Sauter, T. Schlomm

Development of methodology: M. Lennartz, E. Öztürk, R. Shihada, M. Ruge, C. Koop, T. Steuber, R. Simon, T. Schlomm

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Lennartz, S. Brasch, H. Wittmann, L. Paterna, K. Angermeier, E. Öztürk, R. Shihada, M. Ruge, M. Kluth, C. Koop, H. Heinzer, T. Steuber, M. Graefen, A. Haese, T. Schlomm

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Lennartz, S. Brasch, H. Wittmann, L. Paterna, K. Angermeier, E. Öztürk, R. Shihada, M. Ruge, M. Kluth, P. Lebok, C. Wittmer, M. Graefen, R. Simon, T. Schlomm

Writing, review, and/or revision of the manuscript: M. Lennartz, S. Minner, H. Wittmann, E. Öztürk, M. Kluth, W. Wilczak, T. Krech, P. Lebok, H. Heinzer, T. Steuber, M. Graefen, A. Haese, R. Simon, G. Sauter, T. Schlomm

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Lennartz, E. Öztürk, R. Shihada, C. Koop, T. Krech, M. Adam, H. Huland, M. Graefen, A. Haese, R. Simon, G. Sauter, T. Schlomm

Study supervision: H. Huland, T. Schlomm

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.