Evaluation of Contemporary Prostate and Urothelial Lineage Biomarkers in a Consecutive Cohort of Poorly Differentiated Bladder Neck Carcinomas

Abstract

Objectives:

New immunohistochemical (IHC) markers of urothelial carcinoma (UCa) and prostatic adenocarcinoma (PCa) have emerged in recent years, yet comparative studies to establish markers remain lacking. We aimed to identify an effective but parsimonious approach for poorly differentiated bladder neck lesions, to establish a best practice panel approach in a setting simulating prospective use.

Methods:

We tested the performance of a panel of IHC markers on whole sections of a consecutive cohort of transurethral resection specimens of poorly differentiated, challenging bladder neck resections (n = 36).

Results:

In the setting of poorly differentiated bladder neck carcinomas, biomarker sensitivities for UCa were as follows: GATA3, 100%; S100P, 88%; p63, 75%; and cytokeratin (CK) 5/6, 56%; specificities of each were 100%. CK7 and CK20 showed sensitivities of 75% and 63%, though these were only 85% and 80% specific. For PCa markers, NKX3.1, p501S, prostate-specific membrane antigen, and androgen receptor (AR) each showed 100% sensitivity, outperforming ERG (35%) and prostate-specific antigen (PSA; 25%). All the prostate histogenesis markers were 100% specific, except for AR, which was positive in 13% of the UCa cases.

Conclusions:

Novel IHC markers show improved diagnostic performance that enables positive and negative support for identifying histogenesis with the use of as few as two markers for this critical therapeutic distinction. PSA underperforms newer markers.

The presence of a poorly differentiated high-grade carcinoma in transurethral resection specimens, particularly from the bladder neck, often raises the differential diagnosis between a urothelial carcinoma (UCa) and prostatic adenocarcinoma (PCa). Traditionally, a constellation of histomorphologic features are used to differentiate UCa from PCa, and well-differentiated cases do not present a diagnostic conundrum. The presence of surface neoplasia, nested growth pattern, marked nuclear pleomorphism, glassy eosinophilic cytoplasm, and brisk mitotic activity suggests a urothelial origin. In addition, foci of squamous differentiation strongly favor the diagnosis of UCa. In contrast, cribriform or focal acinar architecture, minimal nuclear pleomorphism, nucleolar prominence, foamy and pale cytoplasm, and low mitotic activity favor PCa. However, in practice, determining the origin of high-grade, poorly differentiated carcinomas involving the bladder neck is difficult, especially in the context of the multiple procedure-related artifacts of transurethral resection specimens. High-grade UCa may take on widely varied patterns of high-grade or variant differentiation, whereas high-grade PCa (Gleason patterns 4 and especially 5) may show considerable nuclear pleomorphism and other overlapping features with UCa. In a small, but real, subset of cases, UCa and PCa may even coexist in the same resection specimen Image 1.

Prostate-specific antigen (PSA) immunohistochemistry (IHC) has traditionally been used to resolve this diagnostic dilemma; however, a number of reports have shown an inverse correlation between PSA stain and Gleason score such that high-grade PCas are often negative for PSA.^1–3 Also, PSA may stain several normal structures and benign lesions such as periurethral glands, anal glands in men, cystitis cystica glandularis, and nephrogenic adenoma of the prostatic urethra,⁴ all of which could be sampled in transurethral specimens.

Accurate distinction between UCa and PCa is essential from both prognostic and therapeutic viewpoints. The 5- and 10-year survival rates for high-grade, muscle-invasive UCa involving the bladder neck (at least stage pT2b) range from 70% to 83% and 69% to 78%, respectively, whereas, for a high-grade PCa involving the bladder neck (at least stage pT3a), the 5- and 10-year survival rates range from 95% to 100% and 81% to 93%, respectively.⁵ More importantly, the distinction of UCa and PCa implies marked differences in management and therapeutic strategy. The standard therapeutic approach for UCa is cystectomy or cystoprostatectomy and chemotherapy, whereas an advanced PCa involving the bladder neck (at least stage pT3a) is often managed with hormonal therapy and radiation.⁶

In recent years, various emerging urothelial-associated and prostate-associated markers, including GATA3, S100P (placental S100), p501S, NKX3.1, and PSMA have become available to help identify urothelial^7–10 or prostatic^11,12 origin of high-grade, poorly differentiated carcinomas. Heretofore, studies have generally used tissue microarrays to test the performance of these markers.^13,14 Although such studies have the advantage of testing a large number of cases in parallel under identical conditions and limiting use of reagents, they do not validate the usefulness of these biomarkers in a cohort simulating the clinical setting in which such markers would be used prospectively.¹⁵ For that matter, a comprehensive comparison of all these markers, as well as more established markers, has not been performed. Thus, we undertook the current study to evaluate the usefulness of each marker in a comprehensive immunohistochemical panel for the UCa/PCa differential, using whole sections of a retrospective consecutive cohort of cases of high-grade carcinomas involving the bladder neck to simulate a prospective validation study.

Materials and Methods

With institutional review board approval, the Cedars-Sinai Medical Center (Los Angeles, CA) anatomic pathology database was searched for high-grade, poorly differentiated carcinomas in transurethral resection bladder neck specimens from 2009 to 2013. These cases were diagnosed as urothelial or prostatic in origin based on well-characterized clinical history, imaging, histology, and/or available IHC markers. For UCa, inclusion criteria were muscle invasive disease and lack of variant morphology characteristic for UCa, whereas for PCa, inclusion criteria were high-grade PCa of Gleason score 9 or 10. As an additional criterion, the study pathologists (S.K.M., S.C.S., M.B.A.), on review of recut tissue sections, excluded any cases that were felt to be discernible with reasonable certainty as UCa or PCa on morphologic grounds.

Single, representative whole tissue serial sections from all of the cases were subjected to a panel of 12 established and recently introduced IHC markers, with reactions performed at the Cedars-Sinai Medical Center Clinical Immunohistochemical Core Facility using the clones, sources, and reaction parameters detailed in Table 1. (The marker NKX3.1 was initially validated in a pilot series of cases at Phenopath Laboratories [Seattle, WA]). The urothelial markers tested were GATA3, S100P, p63, CK5/6, CK7, and CK20, and the prostate-associated markers were PSA, prostate specific membrane antigen (PSMA), NKX3.1, P501S, androgen receptor (AR), and ETS transcription factor ERG (avian v-ets erythroblastosis virus E26 oncogene homolog).

The pattern of IHC stain was interpreted semiquantitatively by intensity (0, negative; 1, weak; 2, moderate; 3, strong) and proportion (0, negative; 1, focal; 2, multifocal; 3, diffuse) by two pathologists (S.K.M., S.C.S.) with discordances resolved by consultation with a senior genitourinary pathologist (M.B.A.). The minimal threshold for focal positive stain was arbitrarily considered the least amount that would be considered focally positive if encountered prospectively in a difficult case. For GATA3, S100P, p63, NKX3.1, AR, and ERG, only nuclear staining was considered as positive. P501S perinuclear punctate cytoplasmic staining was evaluated. For CK5/6, CK7, CK20, PSA, and PSMA cytoplasmic and/or membranous stain was assessed. Statistical analysis was performed with Prism Version 5.0 (GraphPad Software, La Jolla, CA), using the Friedman test with the Dunn post-test to compare paired distributions of ordinal IHC scores.

Results

Cohort

A total of 36 consecutive cases of transurethral resections of poorly differentiated, high-grade carcinomas involving the bladder neck and showing morphology indeterminate for UCa vs PCa were identified in retrospective database searches and subsequent slide reviews. These included 16 cases of UCa, with patient ages ranging from 45 to 97 years (median, 73.5 years; mean, 74.5 years) and a male-to-female ratio of 4:3. None of the patients with UCa were treated with neoadjuvant hormonal, radiation, or chemotherapy. Bacillus Calmette-Guérin treatment was given in three cases. A total of 20 cases of poorly differentiated high-grade prostate cancer with Gleason scores of 9 (n = 10) and 10 (n = 10) were identified, with patients ranging in age from 46 to 95 years (median, 80.5 years; mean, 78.4 years). Of the 20 high-grade PCa cases, seven were diagnosed as PCa on transurethral resection for obstructive uropathy, seven were known cases of PCa after radiation and androgen deprivation therapy, three cases were of PCa with docetaxel and external beam radiation, and one case each of PCa after radical prostatectomy, radical prostatectomy and androgen deprivation therapy, and brachytherapy. Representative archival blocks for each case were retrieved, and serial whole sections submitted for IHC using the 12 IHC markers of histogenesis.

Immunohistochemical Markers of UCa

Diffuse and strong nuclear immunoreactivity for GATA3 was observed in all cases of UCa (100% sensitivity), whereas 88% of UCa showed diffuse and strong nuclear (and, generally, cytoplasmic) positivity for S100P Image 2. Both of these markers were 100% specific for UCa, because all PCa cases were negative. Of other markers showing 100% specificity to UCa, p63 showed sensitivity of 75% and CK5/6 showed sensitivity of 56% Image 3. These markers differed significantly overall in scores for proportion of cells staining (P < .0001) and in distribution of intensity (P = .02). In particular, both GATA3 and S100P showed significantly higher scores for proportion (but not intensity) of cells staining than CK5/6 (both P < .05) but not p63. In contrast, CK7 showed a sensitivity of 75% and CK20 showed a sensitivity of 63% for UCa (50% double positive, at least focally); however, these markers showed a lack of specificity relative to the newer markers (CK7, 85%; CK20, 80%), with staining noted in subsets of PCa. Complete data for the urothelial markers are presented in Table 2.

Immunohistochemical Markers of PCa

In our cohort, the sensitivity for PSA in the setting of high-grade PCa of the bladder neck was only 25% Image 4A and Image 4B. In contrast, all cases of PCa studied were positive for p501S, NKX3.1, PSMA, and AR Image 4C, Image 4D, Image 4E, and Image 4F. For NKX3.1, PSMA, and AR, all cases showed diffuse (3+) intense (3+) staining, including the 75% of cases negative for PSA, consistent with a significant difference in proportion and intensity between any of these markers and PSA (P < .0001). For p501S, three cases (15%) showed focal intense positivity. In terms of specificity, NKX3.1, PSMA, p501S, and PSA were all 100% specific to PCa, with no staining in any case of UCa. In contrast, AR showed expression in two of 16 UCa cases (13%) Image 5. ERG was positive in 35% of PCa cases and 100% specific to PCa vs UCa Image 6 and to PCa compared with benign prostatic glandular tissue. Complete data for the prostatic markers are presented in Table 3.

Discussion

Designating poorly differentiated carcinomas of the bladder neck as UCa or PCa remains a challenge in surgical pathology. Difficulties stemming from sampling and procedural artifacts further compound interpretation in this diagnostic setting. The therapeutic stakes, in terms of differences in therapeutic approach for UCa vs PCa, are manifest. Indeed, recent advances in treatment of PCa in the metastatic or locally advanced/inoperable setting (as with extensive bladder neck invasion), which include a number of targeted agents and high potency antiandrogen drugs,¹⁶ make distinguishing these cases all the more important.

Most cases of UCa and PCa remain readily distinguishable on morphologic grounds: the characteristic morphology of “garden variety” PCa, consisting of small, infiltrative or crowded malignant glands with nucleolar atypia, is not imitable by UCa except in the most unusual of circumstances. Cases of UCa with papillary or solid invasive growth and usual cytomorphologic features also do not generally resemble PCa, nor do many of the variant morphologies of UCa. Routine cases may be signed out on morphology alone and are not the subject of this study.

Building on a number of promising studies demonstrating improved specificity of several new immunohistochemical biomarkers to UCa or PCa, we aimed to test the performance of these markers in a setting targeted to emulate that of their eventual application: using serial, full sections of transurethral resection tissues from a consecutive cohort of cases of high-grade carcinomas of the bladder neck where UCa vs PCa had been the diagnostic consideration. Our hope was that we would verify the performance of these markers in this key clinical setting and generate data to guide improved resource utilization and recommend best practices.¹⁷

For UCa, we studied two recently introduced markers GATA3 and S100P. These markers, identified as potentially useful and specific for the urothelium in a microarray study, have shown some promise as urothelial markers.^7,13,18–20 GATA3 is a nuclear transcription factor that has also been used as a marker for adenocarcinoma of the breast.^21,22 Recent studies have identified a broader array of positive carcinomas than initially reported,²³ including squamous cell carcinomas, basal cell carcinomas, pancreaticobiliary carcinomas, and germ cell tumors (trophoblastic and yolk sac). Of particular interest to the differential diagnosis of epithelioid lesions of the bladder neck is the recent observation of GATA3 positivity in more than 80% of paragangliomas,²⁴ as well as generally less intense staining in squamous lesions of the anus (<10%) and uterine cervix (<20%).²⁵ Most relevant to our study of the bladder neck is a prior tissue microarray–based study that described positivity for GATA3 in 28 (80%) of 35 UCa and 0 of 38 PCa cases.²⁵ Our results identified diffuse, intense staining of GATA3 in 100% of UCa and 0% of PCa (all lesions in poorly differentiated/challenging cases); this finding validates prior findings and supports the use of this marker as a first-line and sensitive urothelial marker in the differential between UCa and PCa, taking into consideration squamous and paraganglioma lesions. In fact, given the punctate nuclear stain of GATA3, future studies should address whether this marker could be usefully combined in a “urothelial cocktail” with the (non-nuclear) specific urothelial lineage markers uroplakin II and uroplakin III.²⁶

In tandem, we found a high degree of sensitivity and specificity (88% and 100%, respectively) of placental S100 (S100P), a member of the S100 family, in the same cohort of cases of diagnostically challenging bladder neck lesions. Although fewer studies have tested the specificity of this marker to urothelium and recent reports highlight positivity for S100P in pancreaticobiliary lesions,^27–29 we propose that this marker may be a useful first-line or confirmatory marker for UCa in the bladder neck. Although no primary squamous lesions were encountered in this consecutive cohort (in as much as squamous differentiation argues against PCa and such cases were excluded), we recently reported that S100P shows very infrequent staining in squamous cases.⁷ Thus, this marker may help confirm the impression of urothelial differentiation in a case with GATA3 positivity but has somewhat limited value in determining UCa vs PCa.

Lastly, though we demonstrate superior sensitivity of GATA3 and S100P, our results confirm the usefulness of established markers of the urothelium, most especially p63 but also basal-type CKs (CK5/6 or high-molecular-weight cytokeratin). A number of studies have proposed these markers for this very diagnostic setting^13,30; our sensitivity for p63 for UCa was somewhat lower than that reported by Kunju et al³⁰ and Chuang et al,¹³ which may be because we tested only poorly differentiated cases. In contrast, GATA3 and S100P showed statistically significantly greater proportion of UCa cells staining positive compared with CK5/6. Lastly, our results for CK7 and CK20 mirror those of Kunju et al³⁰ as well as those of prior, large studies across multiple cancer types.³¹ These latter keratins cannot be recommended as specific markers for urothelial differentiation.

Historically, PSA has been useful in identifying the prostatic origin of a tumor, owing to its relative specificity^3,4,9; however, the sensitivity of PSA decreases in higher-grade lesions.^1–3,9 Another classic marker proven to be useful in identifying prostatic origin is prostate specific acid phosphatase (PSAP),³ which shows similar performance to PSA, with similar decreased sensitivity in higher-grade lesions.³² In this study, we elected not to test the latter, given the reported similarities and reports that PSAP use has become less widespread than PSA.³³ Given the limitations of PSA and PSAP, especially in the frequently challenging higher-grade lesions, a number of improved markers of prostate histogenesis have been introduced. These include the homeodomain transcription factor NKX3.1, the transmembrane Golgi-associated protein, prostein (p501S), and PSMA. All PCa cases in our series were immunoreactive to these markers, with no UCa cases reactive, whereas only 25% of our PCa cases were positive for PSA. This difference between PSA and the newer markers is highly statistically significant. We speculate that our cases may show low prevalence of PSA stain because they were poorly differentiated lesions, as defined by the study inclusion criteria. For that matter, we previously noted that polyclonal PSA is more sensitive than monoclonal PSA for detecting prostatic origin, though we used monoclonal PSA in this study (as we do in routine clinical use) for the sake of ensuring specificity.³⁴ In addition, as many as eight of these cases had a history of androgen deprivation therapy, which could also account for lower PSA expression; studies have found that in the setting of treated disease, PSA expression is markedly variable.³⁵ Thus, in this setting we regard NKX3.1, p501S, and PSMA as diagnostically equivalent, such that any one of these markers may be used to support the diagnosis of prostate cancer. For that matter, though we still have PSA in our laboratory, we regard either NKX3.1 or p501S as unequivocally superior to PSA. One final issue regards the prior observation of PSMA positivity in a subset of UCa³⁶; we did not observe such aberrant staining in the cases we studied.

The additional markers tested as biomarkers of PCa include AR and ERG. AR has potential in this setting, with the caveat that a minor subset of UCas may show positivity, as we observed in two cases. AR positivity in UCa has been observed infrequently before,^37,38 and our results in this particular setting are cautionary in the setting of lesions of the bladder neck. Finally, immunostaining for ERG has been proposed as a biomarker for PCa compared with benign glandular lesions and of prostatic histogenesis among carcinomas,^39,40 based on the observation that ERG fusion with androgen responsive genes is a prevalent and highly specific molecular feature of PCa.⁴¹ We observed a prevalence of 35% of ERG expression in this cohort, and similar to the prior studies, it was 100% specific to PCa vs UCa. In summary, while AR is very sensitive for PCa, its specificity remains questionable. By the same token, while ERG can be an exceptionally useful diagnostic and lineage marker in challenging cases (Image 6)^39–42 (and might have future value in selection for emerging therapeutic options¹⁶) its sensitivity, though admittedly better than our PSA, is lower than the favored markers NKX3.1, p501S, and PSMA.

Although these immunohistochemical markers are discriminatory in this cohort, we reiterate that all ancillary studies must be viewed in the context of the morphology and clinical history. cinclude the full range of morphologic patterns, because some cases may show both tumor types in only one of several blocks, as was apparent with the case shown in Image 1.

In conclusion, we interpret these findings as confirmatory of a broad role for the recently introduced markers, GATA3, S100P, p501S, NKX3.1, and PSMA, in conjunction with the better established marker p63, as diagnostic markers for poorly differentiated carcinoma of the bladder neck. Based on the strong performance of these markers in this sequential cohort, selected retrospectively as the same types of cases as would be encountered prospectively, we recommend an initial panel to include a minimal number of these markers, with the flexibility to expand if necessary and based on the availability and experience of individual laboratories. For individuals with access to a relatively limited range of IHC antibodies, a reasonable first-line panel would still include p63 and PSA, with the understanding of their somewhat limited sensitivities, especially for the latter. For challenging and unresolved cases, we recommend at least two lineage-specific markers each for UCa (GATA3, S100P) and PCa (NKX3.1, p501S, PSMA).

References