The WHO 2004 classification defines new histological and molecular variants of urothelial carcinoma. However, there are limited data available on the clinicopathological characteristics or prognosis of these variants. We present histopathological, molecular and clinical data of 32 plasmacytoid carcinomas of the bladder (PUC) showing that PUC is a high-grade tumor with molecular features of aggressive urothelial carcinoma, usually diagnosed in advanced pathological stage (64% pT3, 23% pT4) showing metastases in 60% of the patients. Average survival of our cohort of PUC treated with radical cystectomy and adjuvant chemotherapy was lower than what is typically seen for comparable conventional urothelial carcinomas. Eighty-seven percent of the PUCs showed a negative or strongly reduced membranous staining of E-cadherin. β-Catenin staining was negative in 22.5%, and 16.7% of the remaining tumors showed nuclear accumulation. Aberrant CK20 expression (negative or >10% of cells stained) and negative CK7 staining was found in 100% and 22.6%, respectively. Ninety-seven percent revealed positive staining for PAN-CK. CD138 was positive in 78%, whereas MUM-1 expression was negative in all cases. Multitarget fluorescence in situ hybridization showed all PUCs to be highly aneuploid and polysomic. Deletions on chromosome 9p21 seem to play an important role in this variant. FGFR3 and PIK3CA mutation analyses yielded no mutations in any of the PUCs analyzed. TP53 mutation analysis showed mutations in 29%. In summary, PUC is an aggressive variant of bladder cancer with molecular features of advanced bladder cancer and evidence of WNT pathway activation in some of the cases.
Urothelial carcinoma is one of the most common malignancies with a total of 47,500 new cases per year in Western Europe.1 The plasmacytoid urothelial carcinoma of the bladder (PUC) represents one of several rare variants of urothelial carcinoma, which are included in the WHO classification since 2004.2 Beyond the plasmacytoid variant, several other rare variants such as micropapillary or nested type urothelial carcinomas have gained attention as the diagnosis of these urothelial subtypes may have prognostic and therapeutic consequences.3
Mai et al.4 reported on an incidence of 2.7% of PUC in a series of muscle invasive urothelial carcinoma, which was confirmed by our own series with four cases in 130 muscle invasive urothelial carcinomas.5 Gaafar et al.6 observed an incidence of approximately 1% in their series of 720 high-grade urothelial cancers. Morphologically, PUCs exhibit round to ovoid tumor cells with abundant eosinophilic cytoplasm, eccentrically located nuclei and indistinct nucleoli that show a discohesive single cell growth pattern similar to lobular breast carcinoma or the diffuse type of gastric cancer (Fig. 1).
Till now, only limited data about clinicopathological and molecular characteristics of PUC are available. The largest series reported to date included 17 cases with a limited immunohistochemical panel and different treatment modalities.7 Profound experiences about pathological diagnosis and suitable primary and adjuvant treatment modalities for this rare variant are still lacking. Herein, we present the largest series of PUCs with 32 cases characterized by a broad immunohistochemical panel, SNaPshot mutation analyses of the fibroblast growth factor receptor 3 (FGFR3) and phosphoinositide-3-kinase catalytic alpha polypeptide (PIK3CA) genes as well as TP53 mutation analysis using direct sequencing and multitarget fluorescence in situ hybridization (FISH).
Material and Methods
Thirty-two cases of PUC were identified by the Departments of Pathology of the Universities Erlangen and Regensburg and 17 cooperating institutions: Department of Pathology, Tenon Hospital, APHP, Paris, France; Department of Pathology, University Cordoba, Spain; Department of Pathology, University Bonn, Germany; Department of Pathology, University Münster, Germany; Department of Pathology, Helios Klinikum, Wuppertal, Germany; Department of Pathology, Klinikum Lüdenscheid, Germany; Department of Pathology Klinikum Fulda, Germany; Department of Pathology, Helios Klinikum Schwerin, Germany; Department of Pathology, Marien Krankenhaus Hamburg, Germany; Department of Pathology, University Schleswig Holstein, Germany; Department of Pathology, University Mainz, Germany; Department of Pathology, University Dresden, Germany; Städtisches Klinikum, Department of Pathology, Karlsruhe, Germany; Department of Pathology, Charité University Berlin, Germany; Department of Pathology, University Homburg, Germany; John Hopkins University-Pathology, Baltimore, USA. Clinicopathological data were extracted from patients' medical records. Patients' survival data were extracted from a database containing 327 patients with locally advanced or metastasized bladder cancer, who had been treated with adjuvant chemotherapy after radical cystectomy within a randomized multicenter phase III trial (AUO –AB 05/95) that was recently published.8
Tissue microarray and immunohistochemistry
A tissue microarray was constructed from 32 formalin-fixed, paraffin-embedded primary plasmacytoid urothelial carcinomas as described previously.9 Hematoxylin and eosin (H&E) stained slides were evaluated by surgical pathologists of the collaborating institutions and were reviewed by a single surgical pathologist subspecialized in urogenital pathology (A. H.). One core (diameter: 1.5 mm) was taken from the tissue block for TMA construction from a representative area of the tumor. Tumor stage and grade were assigned according to UICC and WHO criteria.2 All tumors presented at least 50% plasmacytoid differentiation and showed at least focally a conventional urothelial differentiation.
Immunohistochemical staining was performed using an avidin-biotin peroxidase method with diaminobenzidine (DAB) chromatogen using antibodies against Cytokeratin 7 (CK7, Clone OVTL; BioGenex Laboratories Inc., San Ramon), CK20 (Clone Ks 20.8; DakoCytomation, Glostrup, Denmark), PAN-CK (Clone KL-1; Immunotech, Marseilles, France), CD138 (Clone MI 15; DakoCytomation, Glostrup, Denmark), E-cadherin (Clone 36/E-cadherin; BD Bioscience, Franklin Lakes, USA), β-catenin (Clone 14/β-catenin; BD Biosiences, Franklin Lakes), TP53 (Clone DO-7; DakoCytomation, Glostrup, Denmark), Ki67 (Clone MIB-1; DakoCytomation, Glostrup, Denmark) and MUM-1 (Clone MUM1p; DakoCytomation, Glostrup, Denmark).
For analysis of the E-cadherin staining, we used an immunoreactive score resulting from a multiplication of intensity (0–3) and percentage of stained cells ranging from 0 to 12. Tumors with an expression score less than 9 were considered to show a reduced staining of E-cadherin. A score of 0 was classified as negative (Fig. 2).
UroVysion® probe set (Abbott Laboratories, Abbott Park, Illinois) was used to investigate numerical aberrations using centromer enumeration probes (CEP) for chromosomes 3, 7 and 17 as well as a locus-specific indicator (LSI) for 9p21. Deparaffinized tissue sections (6 μm) were treated with proteinase K for 15 min at 37°C, followed by washing and ethanol dehydration (75, 80 and 100%). Slides were dried at 37°C and then denatured (73°C for 15 min). For hybridization, 14 μl of original probe mix were used per slide. The slides were covered by a cover slip and rubber cement (Fixogum; Marabuwerke, Tamm, Germany) and heated for 96°C for 9 min. This procedure was followed by incubation at 37°C for at least 12 h. After resolving the cover glass using 2× sodium saline citrate /0.3% Nonidet (NP40) solution, 4,6-diamidino-2-phenylindole (DAPI) nuclear counter staining was carried out according to the manufacturer's instruction. For control purposes formalin-fixed and paraffin-embedded samples of the two cell lines UROtsa and J82 were used as described recently.10
Scoring of FISH signals
For each tissue sample, 50 tumor cell nuclei were selected for scoring according to morphological criteria using DAPI staining. Only nonoverlapping intact nuclei of urothelial cells were scored. Nuclei showing only one or no FISH signal for the centromer were excluded from analysis. Each cell was simultaneously analyzed for the centromer signals of chromosomes 3, 7 and 17, and the LSI of 9p21. A cell was considered aberrant if there was a gain of at least one of the centromeric signals or if at least one copy of 9p21 was deleted as described previously.10 Tetraploid cells were considered as normal since umbrella cells have been described as being tetraploid without evidence of malignancy.11 Interpretation of 9p21 signals was carried out as described by Qian et al.12 Briefly, for every nucleus, the numbers of CEP and LSI specific signal spots were counted. A relative deletion was defined if the signal number of 9p21 was more than one unit lower than the mean value of centromeric signals. A sample was considered to carry a relative deletion on 9p21 if more than 7 out of 50 cells did show relative deletion on 9p21 (>14%). In a similar way, we defined a homozygous deletion on 9p21 if more than seven out of 50 nuclei harbored a homozygous deletion of 9p21.
Genomic DNA was isolated from serial sections (5 μm) of the paraffin blocks. Tumor cells were microdissected under microscopic control to achieve a purity of at least 85%. DNA isolation was performed with the use of the High Pure PCR Template Preparation Kit (Roche, Penzberg, Germany) according to the manufacturer's instructions.
Mutation analysis of FGFR3, PIK3CA and TP53
FGFR3 mutation analysis was performed on 30 PUCs using the SNaPshot method as described previously.13 Analyses covered the following codons: S373,G372, S249, A393, Y375, R248, K652, G382. PIK3CA mutation analysis was performed on 31 PUCs using the SNaPshot method for the simultaneous detection of four frequent PIK3CA hotspot mutations: E542K, E545G, E545K and H1047R, as described by Hurst et al.14 Exons 5–9 of the TP53 tumor suppressor gene were directly sequenced in both directions. The methodology has been described in detail in Ref. 15. All mutations were confirmed in a second PCR reaction.
Tumor samples from 32 patients with PUC were included in our study. Survival data were available from 16 patients, who received adjuvant chemotherapy following radical cystectomy within the above mentioned clinical AUO –AB 05/95 trial, and could be compared to 269 patients with conventional invasive bladder cancer within the same trial. All patients within this trial had high risk locally advanced tumors with either invasion beyond the muscularis propria (≥pT3) or lymph node metastasis (pN1). The median age of these 16 patients at the time of radical cystectomy was 56.9 years. The male to female ratio was 13:3. For comparison, the median age of patients with conventional muscle invasive urothelial bladder carcinoma was 62.9 years, which was significantly higher than that of the PUC patients (p = 0.025). Reanalysis of histopathology according to the WHO 2004 classification showed that PUC is a high-grade tumor in 100% of the cases and three of 32 tumors (9.4%) showed a concomitant carcinoma in situ. All PUC patients presented with locally advanced disease: pT3pN0 (n = 4), pT4pN0b (n = 1), pT2pN+ (n = 2), pT3pN+ (n = 7), pT4pN+ (n = 2). Median overall survival of our PUC patients cohort was lower to what is typically seen for comparable conventional urothelial carcinomas (23.4 months).
A reduced staining of E-cadherin was found in 87.1% of the PUCs analyzed. Of these, 74% were even completely negative for membranous E-cadherin. In 12.9% of the tumors, a strong staining of E-cadherin with an immunoreactive score ranging from 9 to 12 was detected. The expression of E-cadherin was used to stratify our series of PUCs into an E-cadherin positive and an E-cadherin negative subgroup, representing 11 and 20 tumors, respectively. In six cases of E-cadherin negative PUC, we additionally analyzed the E-cadherin expression of accompanying TCC, which showed a loss of membranous E-cadherin expression in all cases, too.
β-Catenin staining was completely (membranous, cytoplasmic, nuclear) negative in 22.5% of the cases and 16.7% of the remaining tumors showed nuclear accumulation (Fig. 3). Remarkably, all β-catenin negative tumors or tumors with nuclear accumulation of β-catenin also showed loss of E-cadherin.
CK20 was negative or aberrantly expressed (>10%) in all PUCs and furthermore 22.6% showed a negative staining for CK7 (Table 1). In contrast, using a PAN-CK antibody all but one of PUCs showed a positive staining pattern.
Table 1. IHC characteristics of PUC stratified according to the presence of E-cadherin
Expression patterns of cytokeratines were different between E-cadherin positive and negative tumors. E-cadherin positive PUCs predominately showed a coexpression of CK20 and CK7 (54.5%), whereas the E-cadherin negative subgroup was frequently diagnosed to be CK20 negative and CK7 positive (45%). This CK20−/CK7+ pattern was less frequent in the E-cadherin positive subgroup (27.3%). The differences in the CK expression patterns do not reach statistical significance between the E-cadherin positive and negative subgroups (p = 0.44).
CD138 is a transmembrane proteoglycan typically expressed in normal and malignant plasma cells and lymphoplasmacytoid cells. It is commonly used for differential diagnostic exclusion of plasmacytomas. We found that the majority (78.1%) of PUC were also CD138 positive. There was no difference in the CD138 expression between E-cadherin positive and negative PUCs.
MUM-1 expression as an additional differential diagnostic marker for plasmacytoma was negative in all cases (Fig. 4).
For the assessment of the proliferative capabilities of PUC, we utilized the Ki67 labeling index. A tumor sample was classified as highly proliferative if more than 15% of the cells showed Ki67 staining. Sixty-nine percent of the tumor samples were found to display a highly proliferative phenotype.
Regarding p53 expression, we observed a positive staining in 53.3% of the cases with a high variability in the nuclear detection rate ranging from 10 to 90% of the cells per case. E-cadherin negative PUCs showed 68.4% TP53 accumulation, whereas the E-cadherin positive ones showed only 27.2%. This difference just failed to reach significance (p = 0.056).
A comparison of the immunohistochemical staining patterns between the E-cadherin positive and negative PUC subgroups is given in Table 1.
Multitarget FISH Analyses
We analyzed 31 cases of PUC by FISH using the UroVysion® probe set (Fig. 5). The mean probe copy number for CEP3 was 4.1 ± 0.79, for CEP7 3.3 ± 0.76 and for CEP17 3.1 ± 0.88. In all PUCs analyzed we detected an increase of at least one of the CEP indicating a high frequency of polysomy and frequent aneuploidy in this rare variant of urothelial carcinoma (Fig. 5). The mean probe copy number of the LSI 9p21 was 2.27 ± 1.08. Homozygous 9p21 deletion was present in only 15% of the cases. However, relative deletion of 9p21 was detected in 70% of PUC (Fig. 5). Comparative analysis of E-cadherin positive and negative PUCs did not reveal any significant difference in the multitarget FISH signal pattern.
FGFR3, PIK3Ca and TP53 mutation analyses
SNaPshot analyses of the nine FGFR3 gene mutations detected in bladder cancer so far and the four hotspot codon mutations of the PIK3CA gene did not reveal any mutation in neither the FGFR3 nor the PIK3CA gene. The tumor samples all showed a wildtype sequence for FGFR3 (n = 30) and PIK3CA (n = 31). TP53 analysis of at least three of five exons could be performed in 30 cases, in one case only two exons could be investigated. A TP53 mutation was found in nine of 31 (29%) PUC. The following mutations were identified: c.780delC; p.Met160Ile; 2x p.Glu285Lys; p.Thr230Ile; p.Glu221Gly; c.687_688insG; p.Arg249Ser; p.Cys238Phe. Detailed information about the detected mutations is given in Supporting Information. There was a good correlation between the results of TP53 immunohistochemistry staining and the TP53 mutation status of the tumors: eight of 15 (53%) cases with nuclear P53 accumulation showed a mutation and 13 of 14 (93%) cases without a detectable TP53 accumulation showed wildtype sequence. There were neither correlations between TP53 mutation status and the clinical and the histopathological characteristics nor the immunohistochemistry data.
The plasmacytoid urothelial cancer of the bladder represents a rare variant of high grade urothelial carcinoma, which is usually diagnosed in an advanced pathologic stage. In this, to our knowledge, the largest series published so far on PUC, we did not have any case of superficial pTa or pT1 tumor. In 9.4% of PUC, we could diagnose a concomitant carcinoma in situ, but this analysis was limited as we could only analyze one tissue block per specimen because of the multi-institutional tissue sampling. In most published studies, the frequency of concomitant CIS in invasive bladder cancer varies between 10 and 46%16–18 suggesting also a higher frequency in PUC. Therefore, our result might be impaired by the limited number of tissue blocks available. The aggressive behavior of the PUCs was further emphasized by survival data showing a tendency toward a shorter overall survival of the patients with PUC compared to those with conventional invasive urothelial cancer of the bladder treated with cystectomy and adjuvant chemotherapy within the clinical AUO –AB 05/95 trial. We therefore set out to investigate whether PUC represents a distinct molecular subgroup of invasive bladder cancers with features differing from those of conventional bladder tumors.
The CK profile of CK7 and CK20 did not show any significant differences between E-cadherin negative or positive PUC. CK20 is normally expressed by umbrella cells of the urothelium and has been shown to predict recurrence in urothelial pTa/pT1 cancers.19 Interestingly, all PUCs of our series display an abnormal CK20 expression as a sign of a poorly differentiated, high-grade cancer.
As there are cases negative for both markers PAN CK is recommended for diagnosis. However, one case of our series was negative for PAN CK, but did show the typical morphology of PUC together with accompanying PAN CK-positive conventional urothelial carcinoma. Lobular breast carcinoma and plasmacytoma was excluded clinically.
A positive CD 138 (Syndecan-1) expression was found in 78% of our cases. Herewith we confirm results of a molecular feature previously described.20 Because of this high frequency in PUC, CD138 can not be recommended as a differential diagnostic marker to plasmacytomas.
Shimada et al.20 postulated in their study that CD138 might contribute to cell survival and progression in urothelial carcinoma as it is highly expressed in high grade carcinomas but not in low-grade TCC. This finding would fit in the molecular profile of PUC gathering different molecular signs of poor prognosis.
As an additional differential diagnostic marker for plasmacytoma we used MUM-1 expression, which was negative in all cases. Overexpression of MUM-1, a member of the interferon regulatory factor family, results from the translocation t(6;14)(p25;q32). Its expression is frequently found in lymphoid malignancies like in multiple myelomas or lymphoplasmatic lymphomas.21–23
The E-cadherin protein is expressed in all epithelial tissues and mediates cell–cell adhesion.24 Abnormal E-cadherin staining in urothelial carcinomas has been shown to be associated with a poor prognosis, advanced pathologic stage and higher frequency of lymph node metastases.25 In contrast to mere reduction of E-cadherin expression in conventional urothelial cancer,26 we show that the membranous E-cadherin expression is completely lost in 64.5% of our PUC cases allowing for a separation of this rare variant into E-cadherin positive and E-cadherin negative PUCs. As within these tumors membraneous E-cadherin expression is also lost in accompanying conventional TCC. This molecular feature seems to be characteristic of the whole tumor and not only of the plasmacytoid component.
To date, only seven cases of PUC analyzed for E-cadherin expression have been published.4, 5, 27 All cases showed a complete expression loss, which is in line with our findings. Till now, there are no precise data about E-cadherin loss in conventional invasive TCC. Recently published papers differ rather in reduced or heterogenous E-cadherin expression (including cases with E-cadherin loss) than in E-cadherin positive and negative tumors.28–32 Therefore, we suggest that the complete loss of E-cadherin expression is a prominent feature of PUC.
Interestingly, E-cadherin positive and negative PUC also differ in other molecular features in our cohort. Fifty-eight percent of the E-cadherin negative PUCs either show complete loss of β-catenin or its nuclear accumulation. This contrasts to the E-cadherin positive PUCs, which show membranous or cytoplasmic β-catenin staining in all cases. Our findings of aberrant ß-catenin staining either as a complete loss or nuclear accumulation may provide evidence for two different mechanisms in E-cadherin negative PUCs. On the one side, loss of E-cadherin may result in a rapid degradation of β-catenin by lacking interaction with the cytoplasmic domain of E-cadherin. Tumors with nuclear accumulation of ß-catenin accounting for 23% of all cases, activation of the WNT-pathway might to be involved in PUC development. It is still unclear whether WNT target genes are activated in PUCs and whether these genes contribute to the aggressive behavior of this rare variant.
p53 accumulation as detected by IHC has been described to be a specific indicator for mutations of the TP53 tumor suppressor gene33 although this assumption has recently been questioned for breast cancer.34 p53 acts as a cell cycle regulator and inhibits cell cycle progression at G1-S transition by transcriptional activation of p21 and has been proven to be of prognostic value for patients with invasive urothelial carcinomas.35 Stratifying into E-cadherin negative and positive PUCs, p53 nuclear accumulation was detected in 68% and 27%, respectively (p = 0.056). Our assumption that E-cadherin staining is suitable to identify PUCs with aggressive biological behavior is further underlined by our observation that none of the E-cadherin positive cases showed a TP53 mutation.
FGFR3 mutations are the most frequent somatic mutations found in bladder cancer. They are predominately found in superficial low-grade bladder tumors and are associated with a favorable outcome.36 It has been discussed that FGFR3 could be used as a clinical biomarker and a therapeutic target.37 In several large series of muscle invasive bladder tumors, FGFR3 mutations were identified in about 20% of all patients.38 Data about FGFR3 or PIK3CA mutation status in rare bladder variants are still lacking. Interestingly in PUCs we could not detect any mutation of neither the FGFR3 nor the PIK3CA gene, which is often associated to FGFR3 mutations in bladder cancer, again underlining the aggressive nature of this tumor type.39, 40
In contrast to superficial, low-grade urothelial carcinomas, which develop through a different pathway characterized by FGFR3 mutations and homozygous deletions of chromosome 9 and usually show only very few chromosomal aberrations and are frequently diploid, muscle invasive tumors are genetically unstable and often show aneuploidy, which implicates checkpoint dysfunction in cell cycle regulation. Accordingly, the latest WHO classification distinguishes genetically stable carcinomas (low-grade papillary tumors) from those with genomic instability (high-grade, invasive tumors).2 As PUC is known for an aggressive course of disease, the high grade of aneuploidy is typical for its biological behavior. Using the UroVysion® probe set, Kipp et al.41 showed in their analysis that the mean signal number is significantly increased in tumor cells of rarer bladder cancer variants compared with conventional TCC. We show for the first time that in PUC the mean probe number seems to be even higher than in other variants (4.1 CEP3, 3.1 CEP7, 3.01 CEP17, 2.29 for LSI 9p21), which emphasizes the distinct aneuploidy of this tumor variant. In addition the relative deletion of p16 as detected by LSI 9p21 seems to be a widespread phenomenon in PUC. Loss of p16 leads to an overexpression of Cyclin D1, which in turn results in phosphorylation of Rb and subsequent E2F-mediated DNA synthesis.
In conclusion, PUC represents an aggressive high grade subtype of invasive urothelial carcinoma usually diagnosed in advanced pathologic stage. Till now, the histomorphological development of PUC or other rare variants is not clearly understood. However, presence of conventional TCC favors the idea of a progression-route from TCC to this histological type, which seems to harbor a worse prognosis to what is typically seen for comparable conventional urothelial carcinomas. The immunohistochemical and molecular features are typical for invasive bladder cancers and are associated with loss of membranous E-cadherin, which is present in the majority of PUC. We, therefore, suggest that PUCs should be stratified into E-cadherin negative and positive tumors to further eludicate the clinical relevance of this molecular marker.
This study was supported by grant of the ELAN Fonds of the University of Erlangen to BK and RS. We thank Rudi Jung and Anne Pietryga-Krieger for excellent technical assistance. We furthermore thank all pathologists diagnosing patients within the AUO-AB 05/95 trial for providing paraffin blocks for this study (R. Büttner, Bonn; E. Eltze, Münster; S. Störkel and R. Goltz, Wuppertal; J. Friemann, Lüdenscheid; O. Basten, Fulda; R. Hinze, Schwerin; W. Saeger, Hamburg; Schleswig-Holstein, A. Feller, Lübeck, C. Kirkpatrick, Mainz; G. Baretton, Dresden; H. Frenzel, Karlsruhe; M. Dietel, Berlin; K. Remberger, Homburg).