ERG rearrangements (most commonly transmembrane protease, serine 2 [TMPRSS2]:ERG [T2:ERG] gene fusions) have been identified in approximately 50% of prostate cancers . Quantification of T2:ERG in postdigital rectal examination urine, in combination with PCA3, improves the performance of serum prostate-specific antigen for prostate cancer prediction on biopsy. Here we compared urine T2:ERG and PCA3 scores with ERG+ (determined with immunohistochemical analysis) and total prostate cancer burden in 41 mapped prostatectomies. Prostatectomies had a median of 3 tumor foci (range, 1-15) and 2.6 cm of summed linear tumor dimension (range, 0.6-7.1 cm). Urine T2:ERG score correlated most with summed linear ERG+ tumor dimension and number of ERG+ foci (rs = 0.68 and 0.67, respectively, both P < .001). Urine PCA3 score showed weaker correlation with both number of tumor foci (rs = 0.34, P = .03) and summed linear tumor dimension (rs = 0.26, P = .10). In summary, we demonstrate a strong correlation between urine T2:ERG score and total ERG+ prostate cancer burden at prostatectomy, consistent with high tumor specificity.
Recently discovered chromosomal rearrangement in prostate carcinoma, resulting in fusion of the 5′ translated region of the androgen-regulated gene transmembrane protease serine 2 (TMPRSS2) with members of the ETS family of transcription factors, including ERG or ETV1,1 are promising new biomarkers to aid in the detection of prostate cancer.2-6 ETS fusions have been reported in approximately 50% of prostate-specific antigen (PSA)–screened prostate cancers, and fusions between TMPRSS2 and ERG represent 90% of all ETS fusions; using fluorescence in situ hybridization (FISH) or reverse transcriptase–polymerase chain reaction, ERG rearrangements (as a surrogate for TMPRSS2:ERG fusion) are very specific for prostate cancer or high-grade prostatic intraepithelial neoplasia (HGPIN) immediately adjacent to cancer; hence, the ability to detect this fusion potentially can be used for the detection of prostate carcinoma.2-10 More recently, monoclonal antibodies against ERG have been developed which detect the truncated ERG protein product of TMPRSS2:ERG fusions.11,12 Using immunohistochemistry (IHC), these antibodies are strongly correlated with ERG rearrangement as detected with FISH, and stain approximately 50% of prostate cancers and approximately 15% of HGPIN (only those immediately adjacent to ERG+ cancer), with exceptionally rare staining in nonneoplastic prostate tissue.3,11-16
Recently, our group evaluated a clinical grade, transcription- mediated amplification (TMA) assay that quantifies TMPRSS2:ERG (T2:ERG) messenger RNA (mRNA) in postdigital rectal examination (DRE) urine.17 This assay is based on the same technology as the PROGENSA PCA3 assay (Gen-Probe Inc, San Diego, CA), a urine-based assay for the quantification of the noncoding transcript PCA3,18,19 which is approved by the Food and Drug Administration (FDA) for predicting prostate cancer on rebiopsy. In a prospective study of more than 1,300 men, we showed that urine T2:ERG score, used in combination with urine PCA3 score, enhances the usefulness of serum PSA to predict prostate cancer risk on biopsy; urine T2:ERG score was also significantly correlated with the number of involved cores, maximum percentage of core involvement at biopsy, and maximum index tumor dimension at prostatectomy.17
PCA3 encodes a nontranslated transcript overexpressed in more than 95% of all prostate cancers with high prostate specificity.20-22 The PROGENSA PCA3 assay has demonstrated usefulness for predicting prostate biopsy outcome, and urine PCA3 score has shown an association with tumor volume23-29 and multifocality29 in prostatectomy cohorts. However, PCA3 encodes a noncoding transcript, precluding IHC-based detection, and only a single in situ–based evaluation of PCA3 expression has been reported, which described PCA3 expression in the majority of prostate cancers and HGPIN lesions.30
FISH and IHC studies have shown that ERG rearrangements and protein expression are exceptionally rare in benign prostate tissue or cancer mimickers. Therefore we hypothesized that urine T2:ERG should be strongly correlated with the total ERG+ prostate cancer burden in a given patient. Likewise, published studies indicate that urine PCA3 score is correlated with overall tumor burden. Thus, here we compared urine T2:ERG and PCA3 scores with ERG+ and overall cancer burden at prostatectomy to assess the cancer specificity of these urine biomarkers.
Materials and Methods
A prostatectomy cohort was identified from among 301 men referred for prostate needle biopsy at the University of Michigan Health System (UMHS), who were all assessed using TMA for urine T2:ERG and urine PCA3 scores as described herein. Forty-one men who subsequently underwent prostatectomy at our institution between 2008 and 2011 were included in the study. None of the patients received preoperative radiation or androgen deprivation therapy. Clinicopathologic characteristics including age of patient, ultrasound volume at biopsy, prebiopsy PSA levels, PSA density and biopsy details (total number of biopsy cores, number of positive cores, and percentage of cores positive) were obtained from our clinical database. All biopsy and prostatectomy cases and urine specimens were obtained with institutional review board approval.
Urine T2:ERG and PCA3 Assays
Urine T2:ERG and PCA3 were assessed as described earlier.17 Urine specimens were obtained immediately after attentive DRE, refrigerated, and processed within 4 hours by mixing with an equal volume of urine transport medium. and stored below −70°C until analysis. Amounts of T2:ERG and PSA mRNA were determined with TMA. To generate a T2:ERG score, the amount of T2:ERG mRNA is normalized to the amount of PSA mRNA, which is calculated using the following formula: (100,000 × average urine TMPRSS2:ERG copies/mL)/(average urine PSA copies/mL). Samples with average PSA values of more than 20,000 copies/mL were considered informative. Patients in the current study were assessed with a third-generation, final clinical TMA assay, which is highly correlated to first- (Spearman correlation [rs] = 0.86, P < .001) and second-generation assays as described elsewhere.17 The PROGENSA PCA3 assay similarly quantitates PCA3 and PSA mRNA in post-DRE urine. The PCA3 score was calculated with the following formula: 1,000 × (average urine PCA3 copies/mL)/(average urine PSA copies/mL). Samples with average PSA values of more than 7,000 copies/mL were considered informative. Identical primers for quantifying PSA are used in the PROGENSA PCA3 assay and T2:ERG assay.
Fresh prostate samples removed after surgery were weighed, measured, inked, and fixed in 10% neutral formalin. Seminal vesicles, apex, and base were amputated, and the remaining portion of prostate specimen was serially sectioned at 4-mm to 5-mm intervals perpendicular to the long axis of the gland from the base to apex and quartered. All prostatectomy specimens were reviewed by the study pathologists. Tumor maps were generated by tracking each section and reconstructing them as a whole-mount section. A cancer focus was considered spatially separate or multifocal if it was 3 mm or more from the closest cancer in any single section or 4 mm or more from the closest cancer on the adjacent section above or below, as described previously.31 The largest tumor focus was designated as the index tumor and additional smaller tumors were labeled as multifocal tumors. For each prostatectomy, the total number of tumor foci, linear dimension and Gleason score of the index focus, and linear tumor dimension and Gleason score of all tumor foci was documented. Because the greatest linear dimension of the index focus rather than index focus volume is reported clinically at UMHS, we used the greatest linear dimension of all foci as a cancer volume measurement, which has been validated previously.32 IHC for ERG (as described later) was performed on sections representing all index and multifocal foci from each case. ERG staining was uniformly nuclear, strong, and diffuse except as noted, so we assigned all tumor foci as ERG+ or ERG−, and tumor foci with heterogeneous ERG staining were considered ERG+. The index tumor focus showed the highest Gleason score in the majority of cases. In the rare cases where a smaller multifocal focus had a higher Gleason score compared with the index tumor, the smaller multifocal tumor focus with the highest Gleason score was considered as the index tumor. The summed linear tumor dimension was calculated by summing the largest dimension of the index focus and the largest dimension of all multifocal tumor foci. Likewise, the summed ERG+ linear tumor dimension was calculated by summing the largest dimension of all ERG+ tumor foci, including the ERG+ index tumor.
IHC on unstained formalin-fixed, paraffin-embedded levels of all tumor foci from the prostatectomy specimen blocks was performed using a ready-to-use monoclonal antibody against ERG, clone EPR 3864 (Ventana Medical Systems, Tucson, AZ), using the automated Discovery XT staining platform (Ventana Medical Systems) as described.12,33 ERG staining was evaluated by the study pathologists. Staining of vessels was used as a positive control and slides without staining of vessels were excluded from further analysis. All immunostains were reviewed by study pathologists.
Associations among urine T2:ERG score, urine PCA3 score, and clinicopathologic data were assessed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA). The number of ERG+ and ERG–foci and summed ERG+ and total tumor dimension per case were compared using paired t tests. Correlations between urine T2:ERG and PCA3 scores and continuous and categorical clinicopathologic variables were assessed with Spearman ρ (rs) and the Wilcoxon rank-sum test, respectively. Linear regression analysis was also performed to assess the association between urine biomarkers and between urine biomarkers and ERG+ and total tumor volume. Urine T2:ERG scores were log transformed (log[T2:ERG+1]) to minimize the effect of outliers, which resulted in increased R2 values for all associations compared with nontransformed T2:ERG scores. Two-tailed tests were used for all comparisons and P values less than .05 were considered statistically significant.
The 41 prostatectomies included in the study had a median of 3 tumor foci (range, 1-15) and summed linear tumor dimension of 2.6 cm (range, 0.6-7.1 cm). The index focus showed the highest Gleason score in 39 (95%) of 41 cases. In 2 cases (cases 12 and 38), a smaller multifocal focus showed higher Gleason grade than the larger index focus (4+3 and 3+4, respectively, in the smaller multifocal focus vs 3+3 index focus) and was considered as the index focus for analysis. The vast majority of cases in this study were confined to the prostate (pT2, 37/41, 90%), with index tumor Gleason scores of 7 (31/41, 76%). Pathologic data for all cases are shown in Table 1.
A total of 159 tumor foci were evaluated for ERG staining (including index foci), of which 78 tumor foci (49%) were ERG+. Tumor foci, when positive, showed strong nuclear staining with ERG in all cancerous glands within the tumor focus, except for 3 foci in which the index tumor showed heterogeneous ERG expression (moderate to strong staining, considered ERG+ for analysis). ERG was expressed in cancerous glands in 32 (78%) of 41 cases within at least 1 tumor focus, while the remaining 9 (22%) of 41 cases lacked ERG expression in all tumor foci. ERG expression in the index tumor was noted in 24 (59%) of the 41 cases. Representative ERG+ and ERG–foci are shown in Image 1. The pathologic data, ERG IHC, and urine T2:ERG and PCA3 scores are summarized in Table 1.
Mapping ERG+ and ERG− tumor foci in prostatectomy specimens. Prostatectomy specimens (n = 41) were mapped, and the index focus and all multifocal foci were identified (see Materials and Methods). Immunohistochemistry for ERG was performed on sections representing all index and multifocal foci from each case, and each focus was classified as ERG+ or ERG−. Staining of vessels was used as a positive control and sections without staining of vessels were excluded and staining repeated. Sections of an ERG+ index focus (case 31) stained with H&E (A) and ERG (B). Areas of benign glands, high-grade prostatic intraepithelial neoplasia, and carcinoma are indicated by red arrows, arrowheads, and black arrows, respectively. Sections of an ERG− multifocal focus (case 6) stained with H&E (C) and ERG (D) . Endothelial cells and lymphocytes serve as an internal positive control (green arrows). (×10)
The median summed linear dimension of ERG+ cancer per case was 1.2 cm (range, 0-5.0 cm). No significant difference was found between the number (mean, 1.9 vs 2.0; paired t test, P = .89) or summed linear tumor dimension (mean, 1.6 cm vs 1.3 cm; paired t test, P = .52) of ERG+ and ERG–foci per case. Given the low frequency of disease higher than stage T2 and Gleason scores less than or higher than 7, association of index focus ERG status and Gleason grade and stage were not assessed.
Overall, across 169 total sections, ERG staining was extremely specific for prostatic adenocarcinoma. Although vessels and lymphocytes stain with ERG, this represents expression of wild-type ERG and does not represent ERG rearrangements leading to TMPRSS2:ERG overexpression.12 When positive for ERG staining, HGPIN was always present adjacent to ERG+ cancer, with the exception of 1 ERG+ HGPIN focus not immediately adjacent to ERG+ cancer (this HGPIN gland was 0.4 cm from ERG+ cancer). ERG positivity in benign glands was extremely rare, with approximately 35 ERG+ glands noted in 8 cases across 169 total sections. Only 1 focus composed of 2 benign acini was greater than 0.4 cm from ERG+ cancer. Using the estimated number of benign glands per prostatectomy section reported by Furusato et al,34 the overall specificity of ERG staining for prostate cancer is more than 99.99%. Representative ERG+ HGPIN and benign glands not immediately adjacent to ERG+ carcinoma are shown in Image 2.
ERG staining is specific for prostate cancer and immediately adjacent high-grade prostatic intraepithelial neoplasia (HGPIN). ERG staining was evaluated in 169 prostatectomy sections as described in Image 1. Across all 169 sections, only 1 focus of ERG+ HGPIN and approximately 2 ERG+ benign glands not immediately adjacent to ERG+ carcinoma were identified. Sections from an ERG− cancer focus showing no expression in cancerous or benign glands stained with H&E (A) and ERG (B). Sections from case 38 demonstrating ERG+ HGPIN not immediately adjacent to cancer stained with H&E (C) and ERG (D).
Sections from case 35 demonstrating ERG+ morphologically benign glands not immediately adjacent to cancer stained with H&E (E) and ERG (F). Areas of benign glands, HGPIN, and carcinoma are indicated by red arrows, arrowheads, and black arrows, respectively. Staining of vessels was used as a positive control.
All 41 patients had sufficient urine PSA mRNA expression to evaluate the T2:ERG score. The median T2:ERG score was 41 (range, 0-6,032). As shown in Table 2, of patients with summed ERG+ linear tumor dimensions of 0 cm, 0.1 to 1.0 cm, 1.1 to 2.0 cm, and more than 2.0 cm, 1 (11%) of 9, 4 (44%) of 9, 6 (75%) of 8, and 15 (100%) of 15, respectively, had urine T2:ERG scores higher than 30 (T2:ERG scores >30 are associated with an ~70%-75% risk of prostate cancer on biopsy) and 1 (11%) of 9, 7 (78%) of 9, 8 (100%) of 8, and 15 (100%) of 15, respectively, had urine T2:ERG scores higher than 10 (T2:ERG scores <10 are associated with an ~30%- 35% risk of prostate cancer on biopsy).17
All 41 patients had sufficient urine PSA mRNA expression to evaluate the PCA3 score. The median PCA3 score was 40 (range, 3-187). As shown in Table 2, of patients with summed total linear tumor dimensions of 0.1 to 1.0 cm, 1.1 to 2.5 cm, 2.6 to 3.5 cm, and more than 3.5 cm, 3 (60%) of 5, 6 (50%) of 12, 6 (46%) of 13, and 9 (82%) of 11, respectively, had urine PCA3 scores higher than 35, which has been proposed as an optimal cutoff for the detection of cancer on biopsy.35 Of these patients, 3 (60%) of 5, 7 (58%) of 12, 10 (77%) of 13, and 9 (82%) of 11, respectively, had urine PCA3 scores higher than 25, the FDA-approved cutoff for predicting the presence of prostate cancer after initial negative biopsy.
We next addressed our hypothesis that urine T2:ERG should be strongly correlated with total ERG+ prostate cancer burden, and urine PCA3 should be strongly correlated with total prostate cancer burden. As shown in Table 3, comparing correlations of urine T2:ERG with various clinicopathologic parameters, urine T2:ERG most significantly correlated with the summed linear dimension of ERG+ cancer and the number of ERG+ tumor foci (rs = 0.68 and rs = 0.67, respectively, both P < .0001). No significant association was found with other parameters, including summed total linear tumor dimension (rs = 0.24, P = .13) and urine PCA3 score (rs = 0.22, P = .18). Similarly, summed linear dimension of ERG+ cancer was most significantly associated with the total number of ERG+ tumor foci and urine T2:ERG score (rs = 0.86 and rs = 0.68, respectively, both P < .0001), as shown in Table 4. Urine T2:ERG was significantly associated with ERG+ vs ERG–index focus status (median, 130 vs 6.1, Wilcoxon rank-sum test, P = .002), and a trend of association with index Gleason score higher than 6 vs 6 was observed (median, 51 vs 22, P = .11); however, the small number of cases with Gleason score 6 limited this analysis.
Urine PCA3 score was most correlated with the total number of tumor foci and summed total linear tumor dimension (rs = 0.34, P = .03 and rs = 0.26, P = .10). However, these correlations were substantially weaker than the correlation between urine T2:ERG score and the number of ERG+ tumor foci or summed ERG+ linear tumor dimension (Table 3). No significant correlation was found between urine PCA3 score and urine T2:ERG score or other clinical parameters (Table 3). Likewise, as shown in Table 4, summed total linear tumor dimension was significantly correlated with a number of indicators of total tumor burden (as well as ERG+ tumor burden) at biopsy and prostatectomy. However, no significant correlation was seen with urine PCA3 or T2:ERG scores, which showed nearly equal correlations (0.26 and 0.24, respectively). Lastly, urine PCA3 score was not significantly associated with ERG+ vs ERG− index focus status (median, 42 vs 39, Wilcoxon rank-sum test, P = .78) or index Gleason score higher than 6 vs 6 (median, 37 vs 41, P = .74).
Linear regression analysis also supported the aforementioned associations of urine T2:ERG and PCA3 with ERG+ and total tumor burden. T2:ERG was not significantly associated with summed total linear tumor dimension (R2 = 0.07, P = .10), while PCA3 showed a statistically significant correlation with summed total linear tumor dimension. However, it explained little of the variation in total tumor dimension (R2 = 0.15, P = .01). T2:ERG showed a strong correlation with summed ERG+ linear tumor dimension (R2 = 0.43, P < .0001), while PCA3 was not significantly associated with summed ERG+ linear tumor dimension (R2 = 0.08, P = .08). Lastly, no significant association was seen between T2:ERG and PCA3 scores (R2 = 0.06, P = .12).
In this study, by assessing both urine and prostatectomy tissues, we demonstrated a strong correlation between urine T2:ERG score and total ERG+ tumor burden, thus supporting the very high cancer specificity of this biomarker in urine and tissue. Recurrent T2:ERG fusions, which occur in approximately 50% of PSA-screened prostate cancers, result in overexpression of a truncated ERG protein.4,11,12,16,33
This rearrangement can be confidently detected at the chromosomal level using FISH, as demonstrated in numerous studies.2,4,6 We and others more recently evaluated ERG protein expression in prostatectomy specimens using a monoclonal antibody against ERG (EPR3864), and documented excellent concordance of ERG staining with IHC compared with FISH for ERG rearrangements in more than 1,000 tumors.3,12-16,33,36,37 Similar concordance between another monoclonal antibody against ERG and FISH for ERG rearrangement was reported recently.3,11,16 Likewise, we previously confirmed the high concordance (92%) between the TMA T2:ERG assay and the presence of ERG rearrangements with FISH in prostate needle biopsy cores.17 Importantly on IHC, ERG expression is extremely rare in benign prostatic acini11,12,33 and is nearly 100% specific for prostate cancer or HGPIN immediately adjacent to prostate cancer in tissue studies,3,11-16,33,36,37 which we confirmed here (>99.99% specificity for cancer). Although the protein product of T2:ERG fusion cannot be detected in serum, our group recently evaluated a clinical grade, urine-based TMA assay for quantifying T2:ERG fusion mRNA.17 This assay is based on the same technology as the FDA-approved urinebased PROGENSA PCA3 assay which, in conjunction with serum PSA, has proven useful for prostate cancer detection and was shown to correlate with features of clinically significant disease.18,19,25-29,35,38-41
We previously showed that urine T2:ERG and PCA3 scores correlate moderately, but significantly, with the greatest linear dimension of the index focus at prostatectomy (n = 187, rs = 0.26 and rs = 0.30, respectively, both P < .001).17 However, prostate cancer is frequently multifocal and shows multiple separate tumor foci in addition to the index tumor.31 Heterogeneity among the multifocal tumor foci with respect to both histology and Gleason grade has been well described.42,43 Similar to previous results,42 we found that the majority of our cases (37/41, 90%) had multifocal tumor foci, with 18/37 (49%) cases with multifocal tumors demonstrating heterogeneity in Gleason scores between index and multifocal tumor foci. Recently, multiple groups have confirmed the heterogeneity of ETS gene fusion status (as indicated with FISH for TMPRSS2 or ERG rearrangements) between tumor foci in multifocal prostate cancers.44-46 For example, our group analyzed 93 multifocal prostate cancer foci from 43 radical prostatectomy specimens and found that 70% of cases harbor TMPRSS2 rearrangement, of which 70% were discordant in at least 1 tumor focus, consistent with multifocal prostate cancer arising from multiple, independent clonal expansions.46
Although we previously showed significant correlation between urine T2:ERG and the greatest linear dimension of the index tumor at prostatectomy (rs = 0.26),17 we hypothesized that measuring all multifocal tumor foci and stratifying ERG+ and ERG–tumor foci would demonstrate more significant correlation between urine T2:ERG score and ERG+ tumor burden. Importantly, our study confirms a strong correlation between urine T2:ERG and the summed total dimension and number of ERG+ tumor foci (rs = 0.68 and 0.67). Our present study found no significant correlation between urine T2:ERG score and greatest dimension of the index tumor focus (rs = 0.21, P = .19) or summed total linear tumor dimension (rs = 0.24, P = .13). However, the correlation coefficients are similar to those seen in our previous study on index tumor size,17 suggesting that the smaller size of our current cohort may explain the lack of statistical significance. Linear regression analysis also demonstrated similar findings. Importantly, we recently evaluated ERG protein expression in a full spectrum of lesions encountered in routine diagnostic prostate needle biopsies, including diagnostically challenging biopsies. We found ERG positivity in 44% of prostate cancer and 18% of HGPIN; ERG expression was not observed in benign mimics of cancer such as adenosis and partial atrophy, and was also exceedingly rare in benign glands.33 These results are consistent with those observed in our current prostatectomy cohort and other studies showing that nearly the entire burden of ERG+ prostate tissue (as a surrogate for ERG rearrangements and TMPRSS2:ERG transcript) is carcinoma or HGPIN adjacent to carcinoma.3,11-16,33,36,37 Thus, the total amount of ERG+ protein in a given prostate is nearly entirely ERG+ cancer, which our study demonstrates is strongly correlated with the urine T2:ERG score. Thus, although a limitation of T2:ERG as a biomarker is that it is not expressed in all tumor foci, it is extremely specific for prostate cancer, and there is no known mechanism for markedly elevated T2:ERG in the urine other than prostate cancer.
Interestingly, the correlation between urine PCA3 score and the summed total linear tumor dimension (rs = 0.26, P = .10) is weaker than the correlation between urine T2:ERG score and summed linear ERG+ tumor dimension (rs = 0.68, P < .0001). Although the small size of our current cohort may explain the lack of statistical significance between urine PCA3 score and summed total linear tumor dimension, we observed a similar correlation between urine PCA3 score and the largest dimension of the index tumor at prostatectomy in our previous study (rs = 0.30, P < .001).17 Our current results are consistent with previously published correlations of urine PCA3 scores and total tumor volume at prostatectomy (r and rs= 0.27-0.41).24-26,28 Linear regression analysis also identified a significant correlation between PCA3 and summed total linear tumor dimension, but PCA3 scores accounted for a limited amount of the variation in total tumor dimension (~15%).
PCA3 has been shown to be markedly overexpressed in more than 95% of prostate cancers, and gene expression studies support prostate specificity, suggesting that the lower correlation between urine PCA3 and total cancer burden compared with urine T2:ERG and ERG+ cancer burden is not because of lack of (or variation in) PCA3 expression in some prostate cancer foci. Importantly, given that PCA3 is a noncoding transcript that does not produce a protein product that can be detected by an antibody, there is a lack of studies that have systematically evaluated the specificity of PCA3 across precursor lesions and benign mimics of prostate cancer at the tissue level, unlike ERG. Only 1 reported study has evaluated PCA3 expression in situ in prostatic tissues. Evaluating 24 and 26 cases using chromogenic and radioactive in situ hybridization detection methods, respectively, Popa et al30 showed that although the majority of prostate cancers (92%-96%) showed at least focal cytoplasmic PCA3 expression, the majority of HGPINs (71%-96%) as well as a subset of benign glands (29%-33%) also showed PCA3 expression. Hence, in a single study, tissue expression of PCA3 appears similar to that of alpha-methylacyl-CoA racemase (AMACR), a sensitive and specific prostate cancer marker useful in tissue-based diagnosis, but which is also expressed in the majority of HGPIN foci.47-53 Additional studies will be needed to determine if PCA3, like AMACR, is expressed in a subset of benign mimickers of prostate cancer as well.48,54,55
The expression of PCA3 in the majority of HGPIN lesions (based on a single tissue-based study) may contribute to the lower correlation of urine PCA3 scores and total tumor burden compared with urine T2:ERG scores and ERG+ cancer burden. A number of studies have correlated urine PCA3 score with the presence of HGPIN at biopsy, with conflicting results. For example, Deras et al18 found no difference in PCA3 score for HGPIN vs no evidence of abnormal pathology, whereas Haese et al56 found increased PCA3 scores in men with HGPIN; these 2 studies yielded equivalent diagnostic accuracy for biopsy-detectable cancer. However, these studies were based on HGPIN identified at biopsy, and did not assess the entire prostatic HGPIN burden. Nevertheless, urine PCA3 is clearly useful for predicting the presence of prostate cancer at biopsy and is significantly associated with indicators of aggressive disease at prostatectomy; multiplexing urine PCA3 and T2:ERG will likely allow for more complete assessment of prostate cancer risk and evaluation of prostate cancer burden.17,24-26,28,29,35,39-41,57,58
The current study has some limitations. Our series is rather small, it does not include the full spectrum of pathology (ie, Gleason scores and stage) seen at prostatectomy, and it lacks long-term follow-up. Therefore associations with outcome measures are limited and will require additional studies. In addition, although urine specimens were prospectively collected before performing the biopsy, our study does not represent a prospectively defined prostatectomy cohort. Lastly, although ERG expression with IHC has been highly correlated to ERG rearrangement on FISH, and TMPRSS2 is the 5′ partner in the vast majority of ERG-rearranged prostate cancers, other 5′ partners can pair with ERG, such as NDRG1,59 which would not be detected with the urine T2:ERG assay but would result in ERG protein expression. However, a strength of our study was the ability to directly compare urine T2:ERG and PCA3 scores with tissuebased ERG+ and total cancer burden; correlations observed between urine T2:ERG scores and index tumor dimension, and PCA3 and index dimension and total tumor volume, are consistent with previous reports.17,25,26,28 Although the clearly stronger correlation of urine T2:ERG with total ERG+ cancer burden supports the very high specificity of ERG (and urine T2:ERG) for T2:ERG–positive prostate cancer, our findings will need to be validated in larger series.
In summary, by comparing urine T2:ERG and PCA3 scores with ERG+ and total cancer burden at prostatectomy, our results confirm the extraordinary specificity of prostatic tissue ERG expression for prostate cancer (>99.99%) and demonstrate strong concordance of total ERG+ prostate cancer burden with urine T2:ERG score. This strong correlation supports the potential usefulness of T2:ERG in various clinical situations. These situations can now be prospectively addressed, including risk stratifying men with elevated serum PSA concentrations, those with prior negative biopsy findings, and those considering active surveillance (because high urine T2:ERG scores is strongly associated with a large volume of ERG+ prostate cancer). Urine T2:ERG score may also be useful for predicting upgrading based on prostatectomy findings, because a high urine T2:ERG score but low tumor volume on biopsy may indicate undetected ERG+ cancer.
The authors thank Gary Pestano (Ventana Medical Systems) for providing the ERG antibody and IHC reagents, Jack Groskopf (Gen-Probe Inc) for providing T2:ERG and PCA3 assays, and Angela Fullen and Amy Gursky for technical assistance.
Supported in part by the Early Detection Research Network (U01 CA111275 and U01 CA113913) and NIH S.P.O.R.E. (P50 CA69568). Dr Chinnaiyan is supported by the Prostate Cancer Foundation and the Doris Duke Foundation, and is an American Cancer Society Clinical Research Professor and an A. Alfred Taubman Scholar.
The University of Michigan has been issued a patent on the detection of ETS gene fusions in prostate cancer, for which Drs Chinnaiyan and Tomlins are listed as coinventors. The University of Michigan licensed the diagnostic field of use to Gen-Probe Inc (San Diego, CA), which sublicensed rights to Ventana Medical Systems Inc (Tucson, AZ). Neither company played a role in data collection, interpretation, or analysis, and did not participate in the study design or the decision to submit for publication. Dr Palanisamy has served as a consultant for Ventana Medical Systems. Dr Chinnaiyan has served as consultant to Gen-Probe Inc and Ventana Medical Systems. Dr Tomlins has received honoraria from and serves as a consultant to Ventana Medical Systems.