Identification of Combinatorial Genomic Abnormalities Associated with Prostate Cancer Early Recurrence

Abstract

Multiple biomarkers are needed to distinguish aggressive from indolent prostate cancer. We tested the prognostic utility of a three-marker fluorescent in situ hybridization (FISH) panel (TMPRSS2/ERG rearrangements, AR gain, and PTEN deletion) in a retrospective cohort (n = 210; median follow-up, 5.7 years). PTEN deletion was associated with an increased risk of biochemical recurrence (BcR; hazard ratio, 3.58; 95% CI, 1.39–9.22; P < 0.01) by multivariable Cox regression analyses and earlier BcR (P < 0.02) by Kaplan-Meier analysis. AR gain coexisted with X-chromosome gain and was associated with advanced tumor stage. When this panel was applied, two categories of combinatorial abnormalities proved clinically important. First, PTEN deletion without TMPRSS2/ERG rearrangement was enriched in pT3/4 tumors (70% versus 48%) and tumors with Gleason grades of 8 to 9 (60% versus 17%) compared with the entire cohort. These patients had earlier BcR than patients with normal FISH panel results (P < 0.01). In contrast, patients with PTEN deletion and ERG rearrangement had a BcR rate similar to patients who tested normal for all three markers (P > 0.1). Second, AR gain and concurrent trisomy 10 without TMPRSS2/ERG rearrangement were enriched in pT3/4 tumors and tumors with Gleason grades of 8 to 9. The three-marker FISH panel demonstrated prognostic utility and identified genomic aberrations associated with advanced disease state and early BcR in prostate cancer.

The heterogeneous disease outcomes of newly diagnosed prostate cancer demand effective prognostic markers to separate aggressive from indolent disease. Genomic alterations, such as chromosome rearrangements that involve ETS transcription factor genes, deletions of PTEN and RB1, and gene amplifications of AR and MYC, are recognized as recurrent events in prostate cancer.¹ The prognostic potential of these and other molecular alterations has yet to be established but could provide important information for guiding treatment recommendations. The most common gene rearrangement in prostate cancer, TMPRSS2:ERG fusion, was determined to be associated with adverse tumor behavior in several studies but not in others.2, 3, 4, 5 PTEN deletion alone also correlates with more advanced disease at time of surgery and earlier biochemical recurrence (BcR) in several reports.5, 6, 7, 8 The interaction among different genomic aberrations is also being explored. Reid et al⁵ found that PTEN deletion in the absence of ERG/ETV1 gene rearrangements correlated with worse cancer-specific survival.

We have previously reported that a three-marker fluorescent in situ hybridization (FISH) panel, including AR gene amplification, PTEN deletion, and TMPRSS2/ERG rearrangements, detected higher prevalence of aberrations in castration-resistant or metastatic prostate cancers than in primary prostate tumors.9, 10 In the current study, we evaluated the prognostic significance of this three-marker FISH panel in a retrospective cohort of primary prostate cancer samples obtained at the time of radical prostatectomy. Abnormalities detected by the panel were correlated with disease stage at the time of radical prostatectomy and for associations with BcR after initial treatment.

Materials and Methods

Institutional Review Board Approval

This study was approved by the institutional review boards of the Fred Hutchinson Cancer Research Center and the University of Washington Medical Center. The institutional review boards waived the need for written consent for this study because only de-identified materials were used.

Patient Samples

De-identified, archived, treated primary prostate cancer specimens (n = 210) were obtained from the Virginia Mason Medical Center (Seattle, WA). Patients were all treated by urologists with definitive therapy, namely, radical prostatectomy and pelvic lymphadenectomy. None received neoadjuvant therapy, and they were all followed up regularly with clinical history, physical examination, laboratories, and imaging if necessary to monitor any sign of disease recurrence. BcR was defined as a prostate-specific antigen level higher than 0.2 ng/mL after surgery. Complete clinical information was available for 206 patients, which was collected until the last date of follow-up or date of recurrence. Formalin-fixed, paraffin-embedded tissue blocks from radical prostatectomy specimens were used for tissue microarray generation, using the method described previously.⁴ Each patient is represented by a total of four tissue cores on the tissue microarray, including three with tumor and one with normal prostate epithelia. Of these 206 patients, analyzable FISH data of TMPRSS2/ERG, AR, and PTEN were generated from 154, 167, and 160 patients, respectively. A total of 147 patients had results from all three markers.

FISH

TMPRSS2/ERG rearrangement was assessed using a four-color FISH assay as described previously,⁹ as were AR and PTEN FISH analyses.¹⁰ Representative FISH images from the current cohort are presented in Figure 1. TMPRSS2:ERG fusion refers to the presence of fusion of the two genes, between 5′-TMPRSS2 and 3′-ERG with the interstitial region on chromosome 21 either deleted (typical fusion) or translocated elsewhere (atypical fusion). Occasionally, the TMPRSS2:ERG fusion coexisted with another rearrangement of TMPRSS2, ERG, or both on the other allele, referred to as complex fusion. Duplication of the fusion allele was not observed in this cohort. Rearranged 3′-ERG refers to the separation of the 5′ and the 3′ boundaries of the ERG locus without changes in the TMPRSS2 locus. Alternative rearrangement refers to all other rearrangements of the TMPRSS2 and/or ERG loci.

Figure 1.

Representative fluorescent in situ hybridization (FISH) images. A–E: The four-color FISH detects rearrangements of TMPRSS2 and/or ERG: normal (A); typical fusion with the deletion of the interstitial region containing 3′-TMPRSS2 and 5′-ERG (B); rearrangement of 5′-TMPRSS2 (C); rearrangement of 3′-ERG (D); and complex fusion revealing the simultaneous presence of the fusion signal and signals representing additional rearrangements (E). An illustration of the signal pattern is inserted at the bottom left corner of each image. F and G:AR FISH: normal AR signal pattern with one copy each of the AR (orange) and X-chromosome centromere (CEPX, green) signal per nucleus (F), and AR gain detected in the current study revealing concurrent gain of AR and CEPX (G). H–K:PTEN FISH: normal PTEN signal pattern revealing two signals each for the centromere of chromosome 10 (CEP10, green) and the PTEN gene (orange) per nucleus (H); trisomy 10 revealing three copies each of the CEP10 and the PTEN signals (I); heterozygous PTEN deletion indicated by the presence of two CEP10 signals but only one copy of the PTEN signal (J); and homozygous PTEN deletion demonstrated by the presence of two CEP10 signals but none of the PTEN signal (K). White and orange arrows highlight normal and abnormal cells, respectively.

AR gain was defined as a mean copy number of AR per nuclei equal or higher than 2. Samples with PTEN heterozygous deletion had a ratio of the total number of PTEN signals divided by the total number of CEP10 signals equal or below 0.75. A PTEN/CEP10 ratio equal or below 0.2 was considered an homozygous PTEN deletion. Two patients categorized into the PTEN deletion group had monosomy 10.

All samples were analyzed by at least two readers independently. For each tissue core, a qualitative evaluation was performed initially on the entire core and then by documentation of the signal patterns of at least 25 nuclei. To highlight the abnormality of the tissue core, these nuclei comprise multiple areas of the core or a focal area with the abnormality. A tissue core was considered abnormal as long as the abnormality could be confidently identified by both readers, regardless of whether it was observed across the entire tissue core or from a focal area. The abnormality was typically reflected in at least 30% of the cells of which signal patterns were documented. A patient was to have an abnormality for a given marker if at least one of the tumor cores had an abnormal result. All normal cores had normal results by all three markers. FISH results were deposited in the Stanford Tissue Microarray Database (https://tma.im/tma_portal/SCCA-FISH-2015-01, last accessed April 2, 2015).

Statistical Analysis

Fisher's exact test was performed to calculate P values with contingency tables (GraphPad Quick Calcs, http://graphpad.com/quickcalcs/contingency1, last accessed October 6, 2015; Statistical calculator version 3.0, http://www.danielsoper.com/statcalc3/calc.aspx?id=58, last accessed October 6, 2015). The binary analysis for BcR at 5 years was performed with MedCalc (http://www.medcalc.org/calc/odds_ratio.php, last accessed October 6, 2015). Multivariable Cox regression analysis was performed to calculate the hazard ratios (HRs) with adjustment for age, tumor stage, preoperative prostate-specific antigen, and Gleason grade at radical prostatectomy. Kaplan-Meier plots were used to estimate the BcR-free survival.

Results

Patient Characteristics

Table 1 summarizes the patient characteristics of the cohort. The median presurgery prostate-specific antigen level was 6.5 ng/mL (range, 1.2 to 40.5 ng/mL). Most tumors were confined in the prostate [T2, n = 115 (55.8%)], and median follow-up was 70.5 months (range, 1 to 200). BcR occurred in 60 patients.

Table 1.

Patient Characteristics

Characteristic	All patients studied (n = 206)	Patients with TMPRSS2/ERG data (n = 154)	Patients with AR data (n = 167)	Patients with PTEN data (n = 160)	Complete panel data (n = 147∗)
Age, mean (median, range), years	64 (64, 44–77)	63 (64, 44–77)	63 (64, 44–77)	63 (64, 44–77)	63 (64, 45–77)
PSA, mean (median, range), ng/mL	9.2 (6.5, 1.2–40.5)	9.2 (6.7, 1.2–40.5)	8.9 (6.5, 1.2–40.5)	9.1 (6.6, 1.2–40.5)	9.3 (6.7, 1.2–40.5)
Tumor grade
6	118	79	86	80	74
7	55	48	52	51	48
8–10	33	27	29	29	25
Tumor stage (PT)
T2	115	81	91	85	76
T3	86	69	72	72	68
T4	5	4	4	3	3
N stage (NX)
0	122	95	101	98	91
1	6	6	5	5	4
X	66	45	54	50	45
NA	12	8	7	7	7
M stage
0	197	147	160	153	140
1	9	7	7	7	7
Biochemical recurrence	60	48	51	50	46
Follow-up, mean (median, range), months	75.4 (70.5, 1–200)	70.9 (64.5, 1–188)	72.8 (68, 1–179)	70.8 (64.5, 1–179)	69.8 (62.0, 1–179)

PSA, prostate-specific antigen.

^∗Compared with the remaining patients, those with complete panel data had shorter median follow-up time (62 versus 64 months, P = 0.003). Hence, the tissue blocks were newer. The age of the patients was not different between the two groups.

Prevalence of Abnormalities Detected by the Three-Marker FISH Panel

TMPRSS2/ERG FISH (Figures 1, A–E, and 2A) revealed normal results in 72 (46%) of 158 patients (Figure 1A). The most prevalent abnormality by TMPRSS2/ERG FISH was fusion of the two genes, detected in 58 patients (36%), including 37 (23%) with typical fusion (Figure 1B) and 21 (13%) with an atypical or complex fusion (Figure 1E). The design of the four-color FISH assay also enabled the detection of multiple forms of gene rearrangements, including various forms of non-TMPRSS2:ERG fusion-alternative rearrangements in seven (4%) (Figure 1C) and rearranged 3′-ERG without TMPRSS2 fusion in nine patients (6%) (Figure 1D). Twelve patients (8%) had copy number increase (CNI) at both TMPRSS2 and ERG loci without any rearrangements (image not shown).

Figure 2.

Prevalence of aberrations detected by the fluorescent in situ hybridization (FISH) panel and Kaplan-Meier curves of biochemical recurrence (BcR)–free survival in patients grouped based on FISH findings. A, C, E, and G: The prevalence of different FISH abnormalities detected by TMPRSS2/ERG, PTEN/CEP10, AR/CEPX, and the three-marker panel, respectively. B, D, F, and H: The Kaplan-Meier BcR-free survival in patients grouped based on FISH findings. All curves are marked at censoring times. Normal is denoted by white in all pie charts and by black lines in all Kaplan-Meier curves. A and B: Solid red, dashed red, solid blue, dashed blue, and gray denote typical fusion, atypical or complex fusion, alternative rearrangement, rearranged 3′-ERG without fusion with TMPRSS2, and copy number increase (CNI), respectively. C and D: Red and blue denote PTEN deletion and trisomy 10, respectively. E and F: Red denotes AR gain. G and H: Blue, red, green, dashed red, and gray denote patients with TMPRSS2:ERG fusion only, PTEN deletion only, AR gain plus trisomy 10, PTEN deletion plus ERG rearrangement (including TMPRSS2:ERG fusion), and other combinatorial abnormalities, respectively. Log-rank P values in D and H were calculated between the patients with abnormal and normal results.

PTEN deletion was seen in 28 (17%) (Figure 2C) of 164 patients with readable PTEN FISH results, including 19 with heterozygous deletion (Figure 1J), seven with homozygous deletion (Figure 1K), and two with monosomy 10 (image not shown). AR gain was seen in eight (5%) (Figure 2E) of 171 patients, all with two AR and two X-chromosome signals per cell in at least 40% of the tumor cells evaluated (Figure 1G). Examples of normal PTEN and AR FISH signal patterns are presented in Figure 1, F and H, respectively.

The three-marker FISH panel identified different combinations of abnormalities. Among the 147 patients with data available from all three markers (Figure 2G), 57 (39%) had normal results in all tests (category 1). Because PTEN deletion in the absence of ERG or ETV1 rearrangement was reported to correlate with worse cancer-specific survival,⁵ we defined three separate categories based on these two aberrations: category 2 includes patients with TMPRSS2:ERG fusion alone (without PTEN deletion) [n = 44 (30%)]; category 3 includes patients with PTEN deletion only (without ERG rearrangement) [n = 10 (7%)]; and category 5 includes patients with PTEN deletion plus ERG rearrangement [n = 15 (10%)]. CEP10 and CEPX are centromere probes that detect aneuploidy, which is often a sign of genomic instability in cancer. We therefore designated category 4 [n = 7 (5%)] for patients with CNI of both CEP10 and CEPX. The remaining patients were grouped together as category 6 [n = 14 (10%)].

Intratumor Heterogeneity

The three-marker FISH panel had heterogeneous results from multiple tumor cores of the same patients. When the same abnormality was seen in all tumor cores from the same patient, it was considered a homogeneous result. When a given type of abnormality was seen in only one or two of the three tumor cores, it was considered heterogeneous (Table 2). Heterogeneity involving the PTEN locus was more frequent than heterogeneity in TMPRSS2/ERG (P = 0.0225). Among all patients with readable TMPRSS2/ERG FISH results, 80 had results from at least two tumor cores, and at least one was abnormal. Among these, 64 (80%) had the same abnormal results among different tumor cores. Among 36 samples deemed abnormal by PTEN FISH where data were available from at least two cores, 15 (42%) had heterogeneous results: 10 patients had tumor cores with normal PTEN and PTEN deletion, four had both normal PTEN and trisomy 10 (Figure 1I), and one patient had PTEN deletion and trisomy 10. The risk and the timing of BcR did not differ between patients with homogenous abnormal PTEN results and those with heterogeneous abnormal PTEN results. Variable AR results were seen in two patients. Patients with heterogeneous TMPRSS2/ERG and those with heterogeneous PTEN results did not overlap.

Table 2.

Heterogeneity among Tumor Cores from the Same Patients

Heterogeneity	No. (%) of patients		P∗
TMPRSS2/ERG	PTEN/CEP10
Homogeneous results among tumor cores	64 (80)	21 (58)	0.0225
Heterogeneous results among tumor cores	16 (20)	15 (42)

^∗P value calculated using Fisher's exact test.

Association of FISH Abnormalities with Pathologic Features of Prostate Cancer

Chromosomal aberrations that involve PTEN, TMPRSS2/ERG, and AR were associated with advanced disease based on pathological features (Table 3). These features included aberrations by individual markers (CNI detected by TMPRSS2/ERG FISH, AR gain, PTEN deletion, and trisomy 10) and combinatorial abnormalities identified by the panel (PTEN deletion without TMPRSS2/ERG rearrangement, AR gain plus trisomy 10). Among aberrations detected by single markers, CNI of TMPRSS2 and ERG, AR gain, PTEN deletion, and trisomy 10 were associated with higher Gleason grade (P < 0.01). AR gain and trisomy 10 were also associated with higher tumor stage (P < 0.05). Among combinatorial abnormalities detected by the panel, PTEN deletion without TMPRSS2/ERG rearrangement was associated with higher Gleason grade (P < 0.001); AR gain plus trisomy 10 was associated with higher tumor stage (P = 0.04) and Gleason grade (P < 0.001).

Table 3.

Correlation of FISH Aberrations and Tumor Stage, Gleason Grade, and Biochemical Recurrence

Variable	No. (%) of patients
Tumor stage			Gleason grade				Biochemical recurrence
T2	T3/4	P∗	6	7	8–10	P∗	No	Yes	HR (95% CI)†	P†
TMPRSS2/ERG‡ (n = 154)
Total	81 (100)	73 (100)		79 (100)	48 (100)	27 (100)		106 (100)	48 (100)
Normal	39 (48)	31 (42)		36 (46)	23 (48)	11 (41)		43 (41)	27 (56)
Typical fusion	18 (22)	18 (25)	0.68	22 (28)	9 (19)	5 (19)	0.66	26 (25)	10 (21)	0.57 (0.22–1.46)	0.24
Atypical or complex fusion	13 (16)	8 (11)	0.80	13 (16)	4 (8)	4 (15)	0.49	15 (14)	6 (13)	0.69 (0.22–2.13)	0.52
Alternative rearrangement	2 (2)	5 (7)	0.24	2 (3)	5 (10)	0 (0)	0.16	5 (5)	2 (4)	0.76 (0.10–5.58)	0.78
Rearranged 3′-ERG	6 (7)	2 (3)	0.46	5 (6)	2 (4)	1 (4)	0.88	7 (7)	1 (2)	0.18 (0.02–1.63)	0.13
CNI	3 (4)	9 (12)	0.06	1 (1)	5 (10)	6 (22)	0.004	10 (9)	2 (4)	0.14 (0.02–0.80)	0.027
AR (n = 167)
Total	91 (100)	76 (100)		86 (100)	52 (100)	29 (100)		116 (100)	51 (100)
Normal	90 (99)	69 (91)		86 (100)	51 (98)	22 (76)		111 (96)	48 (94)
AR gain	1 (1)	7 (9)	0.02	0 (0)	1 (2)	7 (24)	<0.001	5 (4)	3 (6)	0.54 (0.10–2.75)	0.46
PTEN (n = 160)
Total	85 (100)	75 (100)		80 (100)	51 (100)	29 (100)		110 (100)	50 (100)
Normal	68 (80)	50 (67)		63 (79)	44 (86)	11 (38)		88 (80)	30 (60)
PTEN deletion	13 (15)	14 (19)	0.40	14 (18)	4 (8)	9 (31)	0.003	13 (12)	14 (28)	3.58 (1.39–9.22)	0.008
Trisomy 10	4 (5)	11 (15)	0.03	3 (4)	3 (6)	9 (31)	<0.001	9 (8)	6 (12)	1.57 (0.44–5.52)	0.49
Three-marker panel (n = 147)§
Total	76 (100)	71 (100)		74 (100)	48 (100)	25 (100)		101 (100)	46 (100)
Category 1	34 (45)	23 (32)		31 (42)	22 (46)	4 (16)		40 (40)	17 (37)
Category 2	26 (34)	18 (25)	1.00	27 (36)	12 (25)	5 (20)	0.49	35 (35)	9 (20)	0.62 (0.23–1.65)	0.34
Category 3	3 (4)	7 (10)	0.10	3 (4)	1 (2)	6 (24)	<0.001	3 (3)	7 (15)	4.79 (0.89–25.79)	0.068
Category 4	1 (1)	6 (8)	0.04	0 (0)	1 (2)	6 (24)	<0.001	4 (4)	3 (7)	1.04 (0.16–6.82)	0.97
Category 5	8 (11)	7 (10)	0.77	10 (14)	3 (6)	2 (8)	0.29	8 (8)	7 (15)	2.42 (0.70–8.33)	0.16
Category 6	4 (5)	10 (14)	0.07	3 (4)	9 (19)	2 (8)	0.06	11 (11)	3 (7)	0.37 (0.08–1.70)	0.2

CNI, copy number increase; FISH, fluorescent in situ hybridization; HR, hazard ratio.

^∗P value denotes the statistical significance compared with normal calculated using Fisher's exact test.

^†Multivariable Cox regression analysis was performed with the adjustment for age, tumor stage, preoperative prostate-specific antigen, and Gleason grade at radical prostatectomy.

^‡Fusion between 5′-TMPRSS2 and 3′-ERG is classified as typical fusion, meaning the fusion of the two genes with deletion of the interstitial region on chromosome 21; atypical fusion, meaning the insertion or translocation, instead of deletion, of the interstitial 3′-TMPRSS2 and 5′-ERG probes to a different genomic location; and complex fusion, meaning the fusion coexists with another rearrangement of TMPRSS2, ERG, or both on the other allele.⁹

^§Category 1, normal in all three markers; category 2, TMPRSS2:ERG fusion only; category 3, PTEN deletion only; category 4, AR gain plus trisomy 10; category 5, PTEN deletion plus ERG rearrangement; category 6, other.

Association of PTEN, TMPRSS2/ERG, and AR Chromosomal Aberrations and BcR after Radical Prostatectomy

We evaluated whether PTEN, TMPRSS2/ERG, and AR chromosomal aberrations detected by FISH were associated with BcR after radical prostatectomy (Table 3). PTEN deletion regardless of other genetic aberrations was associated with increased risk for BcR at 5 years, supported by both univariate (P = 0.01) and multivariate Cox regression analyses (HR, 3.58; P = 0.008). Among patients who had recurrence, Kaplan-Meier analysis revealed an earlier BcR in patients with PTEN deletion compared with patients with a normal PTEN FISH result (P = 0.002) (Figure 2D). No significant difference in BcR risk was observed between heterozygous and homozygous PTEN deletion. When viewed in the context of other aberrations, the three-marker panel revealed that compared with patients with normal results for all three markers, patients with PTEN deletion in the absence of TMPRSS2/ERG rearrangement was associated with BcR at 5 years. This association was statistically significant by univariate (P = 0.03) but not multivariate Cox regression analyses (P = 0.068). Similarly, Kaplan-Meier analysis revealed earlier BcR in patients with PTEN deletion without TMPRSS2/ERG rearrangement compared with patients with normal results by all markers (P = 0.002) (Figure 2H). Patients who had both PTEN deletion and ERG rearrangement were not different from patients with normal results in terms of BcR (Table 3 and Figure 2H). When TMPRSS2/ERG FISH was evaluated alone, patients with CNI had less BcR compared with patients with normal TMPRSS2/ERG FISH results in a Cox regression analysis (Table 3). Patients classified based on TMPRSS2/ERG or AR FISH alone did not differ significantly in terms of BcR free survival by Kaplan-Meier analysis (Figure 2, B and F).

We next compared outcomes in patients with the typical TMPRSS2:ERG fusion versus other abnormalities that involve TMPRSS2 or ERG detected by the four-color TMPRSS2/ERG FISH assay (Table 4). There was no difference in regard to BcR between patients with typical fusion and patients with other TMPRSS2 and/or ERG abnormalities. Among all other abnormalities, CNI was associated with a high Gleason grade (P = 0.002).

Table 4.

Clinical Outcome Comparison between Patients with TMPRSS2:ERG Fusion and Other Aberrations

TMPRSS2/ERG∗ (all abnormal)	No. (%) of patients
Tumor stage			Gleason grade				Biochemical recurrence
T2	T3/4	P†	6	7	8–10	P†	No	Yes	HR (95% CI)‡	P‡
Total	42 (100)	42 (100)		43 (100)	25 (100)	16 (100)		63 (100)	21 (100)
Typical fusion (reference)	18 (43)	18 (43)		22 (51)	9 (36)	5 (31)		26 (41)	10 (48)
Atypical or complex fusion	13 (31)	8 (19)	0.42	13 (30)	4 (16)	4 (25)	0.8	15 (24)	6 (29)	2.48 (0.39–15.70)	0.33
Alternative rearrangement	2 (5)	5 (12)	0.42	2 (5)	5 (20)	0 (0)	0.08	5 (8)	2 (10)	2.98 (0.42–21.22)	0.28
Rearranged 3′-ERG‡	6 (14)	2 (5)	0.26	5 (12)	2 (8)	1 (6)	1	7 (11)	1 (5)	1.47 (0.13–17.22)	0.76
CNI	3 (7)	9 (21)	0.18	1 (2)	5 (20)	6 (38)	0.002	10 (16)	2 (10)	1.18 (0.76–18.24)	0.91

CNI, copy number increase; HR, hazard ratio.

^∗Fusion between 5′-TMPRSS2 and 3′-ERG is classified as typical fusion, meaning the fusion of the two genes with deletion of the interstitial region on chromosome 21; atypical fusion, meaning the insertion or translocation, instead of deletion, of the interstitial 3′-TMPRSS2, and 5′-ERG probes to a different genomic location; and complex fusion, meaning the fusion coexists with another rearrangement of TMPRSS2, ERG, or both on the other allele.⁹

^†P value denotes the statistical significance compared with normal calculated using Fisher's exact test.

^‡Multivariable Cox regression analysis was performed with the adjustment for age, tumor stage, preoperative prostate-specific antigen, and Gleason grade at radical prostatectomy.

Association between FISH Results and BcR in Patients with Different Tumor Stage and Grade

Because some of the FISH aberrations were associated with advanced tumor stage, we next investigated how they predict BcR in patients with T2 and T3/4 tumors, respectively. Kaplan-Meier analysis revealed that PTEN deletion was associated with earlier recurrence in T3/4 prostate cancer (P = 0.015) (Figure 3B). In T2 cancer, both trisomy 10 and PTEN deletion revealed trends of earlier BcR than did the normal group, although this finding was not statistically significant (Figure 3A). When the three-marker panel was evaluated (Figures 3, C and D), PTEN deletion without TMPRSS2/ERG rearrangement was observed in only three patients with T2 cancer, one of whom did not have BcR at the last follow-up (70 months) and the other two who had BcR at 1 and 20 months after prostatectomy. AR gain plus trisomy 10 were seen in only one patient with T2 cancer, whose BcR was detected at 9 months after prostatectomy (Figure 3C). Patients with advanced stage tumor (T3/4) as a whole had earlier BcR than those with low-grade tumor. Among patients with T3/4 tumor, those with FISH aberrations did not differ significantly in terms of BcR from those who were normal by FISH, although a trend toward earlier BcR in patients with T3/4 tumors with PTEN deletion without TMPRSS2/ERG rearrangement was observed compared with patients with T3/4 tumors with normal FISH results (P = 0.13).

Figure 3.

Kaplan-Meier curves of biochemical recurrence–free survival in patients with different tumor stages grouped based on fluorescent in situ hybridization. All curves are marked at censoring times. A and C: Patients with prostate cancer T2 at the time of surgery. B and D: Patients with prostate cancer T3/4.

The patients who had early stage/low-grade tumor by pathological analysis but who had FISH abnormalities had higher tumor grade or earlier BcR in this study. There were few patients in this category except for those with PTEN deletion. PTEN deletion independently predicted earlier BcR regardless of grade, as indicated by the multivariate Cox regression analysis described above (HR, 3.58; P = 0.008). AR gain plus trisomy 10 revealed a trend for early BcR in T2 tumors (Figure 3C).

Discussion

Because of the importance of identifying biomarkers capable of discriminating indolent versus aggressive prostate cancer behavior in a clinical setting, we determined whether recurrent genomic aberrations observed in prostate cancers were associated with pathological parameters and patient outcomes after initial curative treatment. We previously found that the three-marker FISH panel identified higher percentages of abnormal results in castration-resistant or metastatic prostate cancers than in primary prostate cancer.¹⁰ We present our findings on the clinical and prognostic utility of individual FISH markers and a three-marker FISH panel. The key points include the following: the poor prognostic indication of PTEN gene deletion attenuated by concurrent TMPRSS2/ERG rearrangement, the lack of prognostic impact of TMPRSS2/ERG alone, and the association of general CNI with higher grade and advanced stage tumors.

In the current study, TMPRSS2/ERG FISH revealed no significant correlation with pathological parameters, such as tumor stage and Gleason grade, or BcR. CNI without rearrangement correlated with higher Gleason grade and trends of more advanced tumor stage (Table 3). These findings are consistent with previous reports.3, 4 We did not observe duplicated ERG rearrangement, which was previously found to occur in 6% of a cohort of 445 patients with clinically localized prostate cancer and that reportedly correlated with poor cause-specific survival.² In the study by Gopalan et al,³ CNI with or without TMPRSS2/ERG rearrangement (11%) was associated with generalized aneuploidy, high Gleason grade, and advanced tumor stage, as well as BcR. Our Cox regression analysis found that CNI alone without rearrangement (8%) was associated with less BcR compared with patients with normal results. Therefore, building on previous publications,3, 4 data from this study further delineated the clinical significance of genomic subgroups such that, when present alone, neither TMPRSS2/ERG rearrangement nor CNI increases the BcR risk; however, typical TMPRSS2:ERG fusion in the background of CNI, resulting in increased copy number of the fusion allele, exerts an unfavorable effect on outcome.

Consistent with previous findings,¹¹ we did not detect true AR gene amplification in this study of clinically localized prostate cancer. All cases with AR copy number gain in the current study also found concurrent gain of the chromosome X centromere, which indicates the aneuploid state of the tumor. These tumors had higher Gleason grade and more advanced tumor stage compared with those without AR gain (Table 3). A similar association between the gain of AR gene and chromosome X and pathological tumor classification was observed previously.¹² Our data also indicated a trend toward earlier BcR in these patients with AR gain, although a larger cohort is needed to assess the significance of this observation in the context of disease recurrence and other adverse events.

We determined that PTEN deletion was clearly associated with advanced tumor at the time of prostatectomy and earlier BcR compared with tumors that had no genomic aberrations, consistent with previous studies in clinically localized prostate cancers.7, 8 The prognostic significance of PTEN deletion by FISH was also observed when we applied the same PTEN FISH technique on two population-based case-control studies of primary prostate cancer.¹³ PTEN loss by immunohistochemistry (IHC) was also associated with decreased time to metastasis in high-risk patients surgically treated for prostate cancer and earlier death in low-risk patients.14, 15 PTEN gene deletion, although the most common mechanism for PTEN deactivation, is likely not the only mechanism because 30% to 45% of the prostate cancer samples were PTEN negative by IHC without PTEN gene deletion by FISH.¹⁶ Therefore, there may be patients in our cohort with PTEN loss at the protein level who were not identified using FISH. Our results could potentially be further strengthened with additional IHC. FISH, however, enables us to identify genomic aberrations beyond PTEN normal versus deletion as the readout of IHC. For example, in our study, PTEN FISH identified that gain of chromosome 10 was associated with advanced pathological parameters compared with normal. Trisomy 10 may be a result of genomic instability of the tumor in general, which could explain the advanced features. PTEN copy number increase detected by FISH was associated with prostate cancer death in a conservatively managed prostate cancer cohort.¹⁵

When PTEN, TMPRSS2/ERG, and AR status were combined as a panel, patients with PTEN deletion but no ERG rearrangement had earlier BcR compared with patients who tested normal for all markers (Table 3 and Figure 2H). These results augmented a previous study, which revealed that in clinically localized prostate cancer, PTEN deletion in the absence of ERG/ETV1 gene rearrangements correlated with poorer cancer-specific survival compared with patients with normal FISH results.⁵ In addition, the current study determined that tumors with both AR gain and trisomy 10, indicative of an unstable genome, were associated with advanced pathological features (Table 3 and Figure 3). Patients in this category also had a trend toward earlier BcR. Two limitations of this study are multiple comparisons to a control group and lack of power for many comparisons. Few of the comparisons were statistically significant, so the impact on type I error is likely minimal. On the other hand, small sample sizes in some of the groups may result in low statistical power to determine differences in BcR; therefore, nonsignificant P values should be considered in light of this issue.

The evaluation of multiple cores from the same tumor sample enabled us to investigate the intratumor heterogeneity of each marker. Our previous study revealed significant intratumor heterogeneity in castration-resistant prostate cancer.¹⁰ In the current cohort, prostate cancer tumors had more heterogeneous results in PTEN FISH compared with TMPRSS2/ERG FISH. Similar findings were reported using IHC assessment¹⁷: most ERG-positive radical prostatectomy samples had homogeneous ERG staining, whereas more than half of the PTEN loss samples had heterogeneous staining. Together, these observations support the theory that TMPRSS/ERG rearrangement is an early event of prostate carcinogenesis, whereas PTEN deletion is more likely a later event and hence associated with advanced diseases. When we compared patients with homogeneous abnormal PTEN results to those with heterogeneous abnormal PTEN results, no significant difference was observed regarding the risk for BcR and the timing of BcR.

An important clinical question concerns whether the assessment of these FISH aberrations can effectively identify patients who initially present with low-risk tumor defined by low tumor volume and low Gleason grade but will have earlier than expected BcR. In this context, we identified multiple FISH aberrations that correlated with advanced tumor stage (T3/4) and high tumor grade (Gleason grade ≥8), which are typically associated with earlier BcR. We specifically evaluated the outcomes of patients with T2 tumors and those with low tumor grade but with FISH aberrations indicative of advanced disease and earlier BcR, such as AR gain plus trisomy 10 and PTEN deletion without TMPRSS2/ERG rearrangement. Multivariate Cox regression analysis revealed that PTEN deletion was associated with early BcR, independent of tumor stage and grade. As for the other high-risk FISH abnormalities, there were too few patients in the T2 or low-grade category to draw any meaningful conclusions even though there was a trend toward early BcR (Figure 3). Future studies with a focus on low-stage and/or low-grade tumors are needed to confirm the clinical utility of these high-risk FISH aberrations in this context. Given that BcR may not necessary predict the clinical outcome, such as metastases, in prostate cancer,¹⁸ the association between these high-risk FISH aberrations and time to metastasis, as well as prostate cancer–related death, should be assessed.

In summary, the three-marker FISH panel for assessing PTEN, TMPRSS2/ERG, and AR abnormalities revealed prognostic utility in a retrospective cohort and identified genomic aberrations associated with advanced disease state and early BcR in localized prostate cancer. Results of an independent cohort support these findings.¹³ As such, application of this FISH panel for clinical testing has the merit to identify high-risk patients with prostate cancer prone to early relapse and may help physicians with management decisions regarding whether to subject patients to aggressive therapy or active surveillance.