Main

Prostate cancer is the most frequent malignancy in males and responsible for >250 000 deaths per year worldwide.1 These tumors typically show marked histological and molecular heterogeneity, which can be problematic if a potential biomarker or therapeutic target is found, as only minute core needle biopsies are available for the initial diagnosis. However, comparing the extension of multiple heterogeneous molecular alterations within one cancer can also serve to answer biological questions such as the sequel in which these changes occurred. A prerequisite for such an extensive heterogeneity analysis is that the entire cancer bulk is assessable for molecular analysis. To facilitate such whole tumor analyses, we have previously developed a prostate cancer heterogeneity tissue microarray (tissue microarray) platform.2 This tissue microarray platform contains samples from 10 distant tumor areas each of 189 large prostate cancers, thus enabling a high-throughput mapping of molecular features across the entire tumors. Our initial analysis of this tissue microarray revealed that the prostate cancer-specific TMPRSS2:ERG fusions is heterogeneous in >70% of ERG-positive prostate cancers.2

Inactivation of the phosphatase and tensin homolog (PTEN) gene by genomic deletion or rearrangement, including intragenic breakage and translocation, is another key molecular event in prostate cancer. PTEN deletion or rearrangement has been reported in 20–30% of prostate carcinomas, and is linked to particularly aggressive cancers.3, 4, 5 Mouse models of prostate cancer suggest that PTEN inactivation and ERG fusion may cooperate in triggering development and progression of prostate cancer.6, 7 Studies comparing PTEN and ERG fusion status have demonstrated that these alterations frequently coexist.3, 4, 8, 9, 10, 11, 12 PTEN deletions are about three times more frequent in fusion-positive prostate cancers than in fusion-negative cancers.10 The mechanisms explaining the frequent coexistence of PTEN deletions and TMPRSS2:ERG fusions in prostate cancer are unknown. In principle, it is possible that TMPRSS2:ERG fusions facilitate development of PTEN deletions, that PTEN deletions facilitate development of TMPRSS2:ERG fusions or that a specific molecular background facilitates development of both alterations simultaneously.

In order to distinguish between these possibilities, we extended our heterogeneity tissue microarray analysis for PTEN alterations, including deletions and breakage, and compared the PTEN and ERG status in the same tumor areas. The data from this project show high heterogeneity for PTEN aberrations in prostate cancer and suggest that presence of ERG activation facilitates development of PTEN aberrations, while PTEN altered cancers do not show an increased risk for developing additional TMPRSS2:ERG fusions.

Materials and methods

Patient Samples and Tissue Microarray Construction

Usage of the TMA for this study has been approved by the local ethics committee (AZ: WF-049/09; § 12 HmbKHG). The prostate cancer heterogeneity tissue microarray utilized in this study has been described in detail elsewhere.13 A total of 189 formalin-fixed prostatectomy specimens with tumor in at least 10 different tissue blocks were selected for tissue microarray construction. For each cancer, the number of independent tumor foci was determined according to Wise et al.14 In brief, tumor areas were defined as part of a single focus if they were within 3 mm of each other in any section or within 4 mm on adjacent sections. This method identified 1–6 independent tumor foci in our prostate cancers. Seventy-six prostates had 1 tumor focus, 48 prostates had 2 tumor foci, 28 prostates had 3 tumor foci and 37 prostates had 4 or more tumor foci. The latter group also included nine prostates that contained multiple small and very small tumor foci rather than one or several clearly distinguishable tumor masses. One 0.6 mm tumor tissue core was removed from each of the 10 tumor blocks per patient. The 10 tissue cores were distributed across 10 different tissue microarray blocks, so that the complete heterogeneity tissue microarray consists of 10 tissue microarray blocks, each containing one tissue sample of all 189 patients. The localization of each arrayed tumor sample inside the prostate was recorded in order to distinguish between different tumor foci. Presence of cancer was histologically confirmed in each tissue spot. Normal prostate glands were immunohistochemically identified using the antibody 34BE12 (clone MA903, 1:12.5, pH7.8 DAKO, Glostrup, Denmark) for basal cell detection.

Fluorescence In Situ Hybridization (FISH)

FISH was used to detect genomic PTEN deletions and translocations. For deletion analysis, a dual-color FISH probe was constructed from two Spectrum Orange-labeled BAC clones (RP11-380G05, RP11-813O03; Source Bioscience, Nottingham, UK) and a commercial Spectrum Green-labeled centromere 10 (CEP10) reference probe (Abbott Molecular, Wiesbaden, Germany). For PTEN translocation analysis, a dual-color FISH break-apart probe consisting of two Spectrum Orange-labeled BACs (5′PTEN: RP11-659F22, RP11-79A15) and two Spectrum Green-labeled BACs (3′PTEN: RP11-765C10, RP11-813O03) flanking the PTEN gene were used. The localization of all probes and representative FISH images are shown in Figures 1a and b, respectively. Freshly cut 4 μm tissue microarray sections were deparaffinized and proteolytically pretreated using a commercial kit (paraffin pretreatment reagent kit; Abbott Molecular), followed by dehydration in 70, 80 and 96% ethanol, air-drying and denaturation for 10 min at 72 °C in 70% formamide-2x SSC solution. Hybridization was done overnight at 37 °C in a humidified chamber; slides were then washed and counterstained with 0.2 μmol/l 4’-6-diamidino-2-phenylindole in an antifade solution.

Figure 1
figure 1

(a) Localization of the FISH probes used for detection PTEN deletion and intragenic breaks. For detection of PTEN deletion, a PTEN deletion probe covering the entire PTEN gene and 3′ adjacent DNA, and a spectrum green-labeled centromere 10 control probe (data not shown), were co-hybridized. For detection of PTEN breakage, two differentially labeled probes corresponding to the 5′ and 3′ edges of PTEN were co-hybridized. (b) Examples of PTEN aberrations as detected by FISH. (I–III) Examples of the deletion probe: (I) normal copy number with two red PTEN signals and two green centromere 10 signals, (II) heterozygous deletion of PTEN showing loss of one red signal, but two green centromere signals, (III) homozygous deletion of PTEN lacking PTEN signals but showing two green centromere 10 signals. (IV–VI) Examples of the break-apart probe: (IV) intact PTEN as indicated by presence of two adjacent red (3′ PTEN) and green (5′ PTEN) signals, (V) translocation of PTEN showing one normal, adjacent red/green signal and a split of the second one into clearly separated green and red signals, (VI) heterozygous deletion of one allele and translocation of the remaining allele.

Scoring of FISH

The stained slides were visually inspected under an epifluorescence microscope. At least 30 different tumor cell nuclei were inspected for the FISH signal numbers per tissue spot. PTEN deletion was defined as follows: homozygous PTEN deletion was assumed if PTEN deletion probe signals were completely lacking in ≥60% of tumor nuclei, while PTEN FISH signals were present in adjacent normal cells. Heterozygous deletion of PTEN was defined as presence of less PTEN signals than centromere 10 probe signals of ≥60% tumor nuclei. For structural PTEN rearrangement analysis using the break-apart probe, tumors were defined as ‘normal’ when two pairs of overlapping orange and green signals were seen per cell nucleus. A PTEN rearrangement was assumed if at least one split signal consisting of a separate orange and green signal was observed per cell nucleus in ≥60% of the tumor cell nuclei (indicating balanced translocations) or if individual orange and green signals from the overlapping orange/green signal were lost (indicating deletions with breakpoint inside the PTEN gene or imbalanced translocations). Presence of only one overlapping orange/green signal in >60% of tumor cells was considered a heterozygous deletion. Tumors with complete lack of overlapping orange/green signals were regarded as homozygous deletions provided that FISH signals were present in adjacent normal cells. A final score was assigned to each tissue spot that was analyzable for both the PTEN deletion and the break-apart probe according to the following criteria: normal (no deletion and no rearrangement), heterozygous inactivation (heterozygous deletion or rearrangement of one allele), homozygous inactivation (homozygous deletion or heterozygous deletion combined with rearrangement of the remaining allele or rearrangement of both alleles). In case of polyploidy, presence of normal PTEN FISH signals and split signals was rated as heterozygous inactivation. A total of 103 tissue spots that had interpretable results with only one of the two probes were scored based on the respective FISH finding only.

ERG Immunohistochemistry

ERG immunohistochemistry data were available from a previous study.15

Large Section Analysis

To validate the PTEN tissue microarray findings, corresponding tumor blocks were selected from 15 tumor foci showing either a PTEN alteration in only one tissue core (n=10) or in all but one tissue core (n=5).

Statistics

Statistical calculations were performed using JMP 9.0 statistical software (SAS Institute, Cary, NC, USA). Contingency tables were calculated with the χ2 test to investigate the relationship between the degree of PTEN and ERG heterogeneity.

Results

Technical Issues

A total of 1890 tissue spots were included in this study. A total of 887 and 907 tissue spots were interpretable with the PTEN deletion probe and with the PTEN break-apart probe, respectively. As all tumors with deletions according to the deletion probe also showed FISH signal losses with the break-apart probe, the deletion status was assigned solely based on the result of the break-apart probe in 125 tissue spots lacking results with the deletion probe. In additional 103 tissue spots that had a result with the deletion probe but were not interpretable with the break-apart probe, the PTEN status was assigned based on the deletion probe only, and no breakage was assumed in these cases. This assumption was based on the low overall rate of PTEN breaks (1%) in tumors without simultaneous deletion.16 Accordingly, the PTEN FISH status was available from 1008 (53%) tissue spots. Reasons for analysis failure included missing tissue spots on the tissue microarray sections (n=111), lack of unequivocal tumor cells in the tissue spots (n=501) or insufficient hybridization for both FISH probe sets (n=270). The 1008 interpretable tissue spots belonged to 173 different tumor foci obtained from 136 prostatectomy specimens. As different tumor foci within the same prostate represent individual cancers, and because some tumor foci had only few interpretable tissue spots, we refer to different tumor foci in our analysis (rather than to different patients) and excluded all foci with <3 interpretable tissue cores. In summary, 821 tissue spots originating from 123 tumor foci with at least three interpretable tissue spots obtained from 118 patients were included into all subsequent analyses.

Heterogeneity of PTEN Alterations

PTEN alterations (including deletion and rearrangement) were found in 48 (39%) of the 123 individual tumor foci belonging to 47 patients. We observed a high degree of intratumoral heterogeneity: 44 (92%) of the foci were heterogeneously, but only 4 (8%) were homogeneously PTEN altered. In contrast, the inactivation status (heterozygous or homozygous) was usually identical in all PTEN altered tissues of the same focus, with homozygous inactivation being more frequent (25/48, 52%) than heterozygous inactivation (17/48, 35%). Only six (13%) foci showed a mixture of tissue spots with homo- and heterozygous inactivation. All results are summarized in Figure 2.

Figure 2
figure 2

Summary of all PTEN FISH results in 48 tumor foci with PTEN alterations. ‘Homozygously’ and ‘heterozygously affected’ indicates foci with PTEN alterations in all analyzable tissue spots and foci with an admixture of tissue samples with and without PTEN alterations. Circles represent the individual tissue cores belonging to the same tumor foci. The number of interpretable tissue cores per foci ranged between 3 and 10.

Sequel of PTEN Rearrangement and Deletion

Of the 48 PTEN altered foci, 40 foci had deletions only, 7 foci had structural PTEN rearrangements in addition to deletions and 1 focus had a PTEN rearrangement as the sole detectable PTEN alteration. In the 47 foci with deletion and/or structural rearrangements, deletions and rearrangements of PTEN were strongly linked. Rearrangements were found in 7/47 tumor foci with PTEN deletion, but only in 1/76 foci with normal PTEN copy numbers (P=0.003). The presence of tissue spots with both deletions and rearrangements of PTEN within the same cancer focus enabled us to study the sequel of PTEN breakage and deletion. Of these seven foci, six had ≥2 interpretable tissue spots, some of which showed structural PTEN rearrangements, while others had partial and/or complete deletions of PTEN. This pattern of alterations is consistent with a specific sequel of molecular events starting with PTEN breakage and followed by deletion of DNA sequences flanking the breakpoint (Figure 3), including multiple 5′ adjacent genes (MINPP1, PAPSS2, ATAD1 and KLLN) covered by our break-apart FISH probe (Figure 1).

Figure 3
figure 3

Schematic representation of the FISH findings in six cases with at least two tissue spots affected by PTEN rearrangements. Black dots correspond to the centromere 10 probe, yellow and green dots represent the corresponding PTEN break-apart FISH signals. All six cases showed translocations of all present alleles, subsequent loss of the break-apart signals mostly starting at the green 3′ sequence, followed by the orange 5′ sequence and finally loss of all FISH signals. For reference, the spot pattern of a normal case (no deletion, no rearrangement) is indicated in the bottom left corner.

Association to ERG

ERG data were available from a previous study in 786 of 821 (96%) tissue spots with informative PTEN deletion and break-apart status. A positive ERG status was strongly linked to presence of PTEN alterations in these samples. PTEN aberrations were seen in 120 (36%) of 332 ERG positive, but only in 26 (6%) of 454 ERG-negative tissue spots (P<0.0001; Figure 4).

Figure 4
figure 4

Association between immunohistochemical ERG expression and PTEN alterations evaluated in a spot by spot analysis.

Sequel of PTEN Alterations and ERG Fusion

To address the question whether alterations of ERG and PTEN occur in a specific order, we searched for homogeneously ERG-positive foci with focal PTEN inactivation and large PTEN inactivated tumor areas containing focal ERG positivity. Figure 5 shows the ERG and PTEN findings in all 29 tumor foci that were suitable for such an analysis. Sixteen of 19 foci with homogenous ERG positivity had focal PTEN alterations but none of the 10 foci with homogeneous or heterogeneous PTEN alterations had focal ERG positivity (P<0.0001).

Figure 5
figure 5

Frequent (16/19 foci) presence of focal areas of PTEN alterations in ERG-positive tumors but lack (0/10 foci) of focal areas of ERG positivity in PTEN altered tumors indicate that ERG fusion precedes PTEN alterations but not vice versa.

Large Section Validation

Large section analysis of 15 tumor foci with intrafocal PTEN heterogeneity confirmed presence of small areas with genomic PTEN alterations in large ERG-positive tumor foci. Heterogeneous findings occurred not only between different areas within one tumor focus but also within the area represented by a single tissue block. Examples of two cases with heterogeneous FISH findings are shown in Figure 6.

Figure 6
figure 6

Examples of large section validation analysis of two cases with heterogeneous PTEN alterations. (a) Case with PTEN breakage detected with the PTEN break-apart probe. Breakage with loss of 3′ PTEN sequences was found in the tissue punch area taken for tissue microarray construction. An adjacent area labeled by (I) contained tumor cells with an intact PTEN locus, while another area (II) showed homozygous breakage as indicated by the loss of green signals. (b) Case with PTEN deletion detected with the deletion probe. Heterozygous PTEN deletion was found in the tissue punch area taken for tissue microarray construction. An adjacent area labeled by (III) contained tumor cells with normal PTEN locus, while another area (IV) showed heterozygous deletion indicated by loss of one red signal.

Discussion

The results of our study show maximal heterogeneity for PTEN alterations in prostate cancer. Almost all (92%) tumors with PTEN alterations also contained areas without PTEN alterations in the same tumor focus. In about half of the tumors, PTEN alterations were only found in one single tumor area, suggesting that PTEN alterations often occur in late stages of tumor progression. Large section analysis and validation of the tumor cell content in reference tissue microarray sections excluded technical errors or misinterpretation of individual tumor spots as a possible reason for artificial heterogeneity in cases with a single PTEN altered tumor area. Our findings are also supported by previous studies reporting heterogeneity of PTEN alterations, including deletions and mutations, within a single prostate cancer tumor focus11, 17 and in different metastasis of a tumor.18 PTEN alterations were strongly linked to ERG fusion-positive tumor foci in our tumor set, which is in concordance with early studies by others8, 9, 19 and ourselves.10

In line with the marked heterogeneity, the overall frequency of PTEN deletions detected in our present study (48/123, 39%) analyzing 3–10 different tissue spots per tumor was significantly higher than in our previous study analyzing only a single 0.6 mm tissue microarray spot per cancer (457/2266, 20%; P<0.0001).10 This discrepancy is explained by the higher likelihood for detecting deletions in heterogeneous tumors if multiple samples are analyzed. Along with others,3, 4, 20, 21 we10 have previously shown that PTEN loss or mutation is strongly linked to aggressive and rapidly progressing prostate cancers and early tumor recurrence, making PTEN a potentially promising prognostic marker. However, the high degree of heterogeneity detected in our study indicates a significant problem for potential prognostic tests.

Next-generation sequencing studies performed by us16 and others22 revealed that PTEN can be disrupted by chromosomal breakage, resulting in partial deletions, inversions or translocations. We have previously used a break-apart FISH probe to demonstrate presence of PTEN breakage in 4% (including cases with simultaneous deletion of the remaining allele) of prostate cancers using a large prostate cancer prognosis tissue microarray containing a single tissue microarray core per donor tumor.16 The results of our current heterogeneity study suggest that PTEN breakage may be much more frequent (18%), but may be a transient event during the development of complete PTEN loss. This assumption is based on the observation that all tumor foci with PTEN rearrangements also had tissue spots with partial and/or complete deletion of the broken PTEN allele, which can only be explained by initial PTEN breakage and subsequent deletion of chromosomal material flanking the breakpoint. Importantly, such additional deletions occurred even in cases where PTEN was already completely inactivated, for example, by deletion of one allele and breakage of the other. This finding suggests additional selection pressure acting beyond sole inactivation of PTEN. It is possible, that the nearby Killin gene (KLLN), encoding a p53-dependent inducer of apoptosis23 drives these additional deletions. The small (537 bp) KLLN locus is located directly adjacent to the 5′ edge of the PTEN locus23 and, therefore, included in 10q23 deletions detected by ‘typical’ PTEN deletion probes including ours (Figure 1a).10 It is tempting to speculate that tumors with co-deletion of PTEN and KLLN may escape the p53-dependent fail–safe mechanism resulting in growth arrest and senescence in PTEN-deficient cells.24 Such a co-deletion model would also be compatible with earlier reports demonstrating accumulation of tumor cells with homozygous deletion during tumor progression.21

Although cancer heterogeneity limits the applicability of diagnostic, prognostic and predictive test assays, it can be useful to monitor the sequel of two or more molecular events occurring during tumor progression. Depending on the sequel of development of PTEN alterations and ERG fusion, one can expect a small area of cancer having both alterations within a larger area having only one (the earlier) of these changes. We found small PTEN altered areas in 16 of 19 homogeneously ERG-positive tumors, but we failed to see any ERG-positive areas within 10 PTEN-deficient cancer foci, suggesting that ERG fusion typically precedes alterations of PTEN. These findings would be consistent with ERG fusions facilitating the development of PTEN alterations rather than a sole cooperative effect that had been earlier suggested from mouse models with prostate-targeted ERG expression and PTEN depletion,6, 7 In a recent study, Rickman et al25 demonstrated that ERG expression is associated with broad changes in chromatin topology, giving rise to specific genomic rearrangements driven by a particularly high density of ERG-binding sites located close to the breakpoints. It may be speculated that also the PTEN locus may be prone to such ERG-driven rearrangements.

In summary, the heterogeneity tissue microarray approach used in this study did not only allow estimating the degree of heterogeneity of individual markers, but also provided insights into the sequel of PTEN and ERG alterations in prostate cancer. The results of our study show a marked degree of heterogeneity of PTEN alterations, including deletions and gene breakage, and outline a stepwise molecular progression of PTEN alterations from heterozygous breakage to homozygous deletion. In addition, our data strongly suggest a driving role of ERG fusion to promote development of PTEN alterations.