Diagnostic and Prognostic Implications of FGFR3high/Ki67high Papillary Bladder Cancers

Prognostic/therapeutic stratification of papillary urothelial cancers is solely based upon histology, despite activated FGFR3-signaling was found to be associated with low grade tumors and favorable outcome. However, there are FGFR3-overexpressing tumors showing high proliferation—a paradox of coexisting favorable and adverse features. Therefore, our study aimed to decipher the relevance of FGFR3-overexpression/proliferation for histopathological grading and risk stratification. N = 142 (n = 82 pTa, n = 42 pT1, n = 18 pT2-4) morphologically G1–G3 tumors were analyzed for immunohistochemical expression of FGFR3 and Ki67. Mutation analysis of FGFR3 and TP53 and FISH for FGFR3 amplification and rearrangement was performed. SPSS 23.0 was used for statistical analysis. Overall FGFR3high/Ki67high status (n = 58) resulted in a reduced ∆mean progression-free survival (PFS) (p < 0.01) of 63.92 months, and shorter progression-free survival (p < 0.01; mean PFS: 55.89 months) in pTa tumors (n = 50). FGFR3mut/TP53mut double mutations led to a reduced ∆mean PFS (p < 0.01) of 80.30 months in all tumors, and FGFR3mut/TP53mut pTa tumors presented a dramatically reduced PFS (p < 0.001; mean PFS: 5.00 months). Our results identified FGFR3high/Ki67high papillary pTa tumors as a subgroup with poor prognosis and encourage histological grading as high grade tumors. Tumor grading should possibly be augmented by immunohistochemical stainings and suitable clinical surveillance by endoscopy should be performed.


Introduction
Bladder cancer is the second most common genitourinary malignancy [1]. At primary diagnosis, most of the tumors are papillary non-invasive cancers (pTa) which are mostly well differentiated but show a high rate of recurrence. Those tumors are characterized by certain molecular alterations as for example FGFR3 activation [2][3][4][5]. Up to 30% of all patients have invasive disease at diagnosis. These tumors frequently derive from flat carcinoma in situ (CIS) of the urothelium (a high grade lesion, often TP53-mutated) and quickly develop muscle-invasion and metastasis [6,7]. Current prognostic and therapeutic stratification in urothelial cancers is therefore based on tumor staging and grading at histological examination. Staging criteria is the depth of invasion defined by the tumor node metastasis Int. J. Mol. Sci. 2018, 19, 2548 2 of 14 (TNM)-classification of the Union Internationale Contre le Cancer (UICC) [8]. The tumor grading is based upon architectural order and nuclear shape features, which have been thoroughly defined as diagnostic criteria in the 2004 WHO classification of bladder cancer in order to achieve reproducible and comparable diagnoses worldwide. Low grade (LG) tumors show uniform, slightly enlarged nuclei in an orderly, polarized architecture, sometimes with a prominent palisading of the basal layer. Mitotic figures are infrequent [9,10]. High grade (HG) tumors show more pleomorphic nuclei with multiple mitotic features and various extent of architectural disarray [10]. Based on previous genetic analyses and clinical observations, it has been proposed that the histological appearance (grading) of tumors correlates with the underlying genetic alterations, and low grade tumors were regarded genetically stable, whereas high grade tumors, harboring a high number of genetic alterations, were considered genetically "unstable" [7]. Proposed prognostic markers in papillary non-invasive tumors have been the Ki67 labeling index (marker for cell proliferation) and keratin 20 expression (marker for cell differentiation). Tumors with Ki67 ≥ 15% were regarded as highly proliferative [11][12][13] and aberrant expression of keratin 20 was linked to disease recurrence in pTa tumors [14]. Lately, Hurst et al. conducted a comprehensive molecular study on n = 141 papillary non-invasive bladder cancers (low grade, G1 and G2 according to WHO 1973) and found lower overall mutation rates, but more mutations in chromatin modifying genes than in muscle-invasive bladder cancer, and two distinct genomic subgroups of tumors (genomic subtype 1 and 2). The majority of tumors with genomic subtype 1 showed no or only few copy-number alterations. Genomic subtype 2 was characterized by loss of 9q (including the mTORC1 regulator TSC1), increased Ki67 labeling index, upregulated mTORC1 signaling (comprising the overrepresentation of genes in processes that are involved in the unfolded protein response, glycolysis, and cholesterol homeostasis) as well as enrichment for DNA repair and cell-cycle genes [15]. FGFR3 mutations were not found to be significantly different in both subgroups (72% vs. 89%) and TP53 mutations were absent [15]. The authors did not show a correlation of molecular profiles with specific histological features.
However, in routine histological diagnostics, pathologists often see papillary non-invasive tumors with quite uniform, relatively small nuclei, which give a "crowded" impression, but seem to be of "low nuclear grade". Interestingly, Ki67 labeling in these tumors is often enhanced and from this point of view a reconsideration of a possible "high grade"-biology is implicated. Opposite to the negative predictive impact of a high Ki67 index, these tumors often show a strong expression of FGFR3, which indicates an activation of the signaling pathway resulting in cellular proliferation, but is generally associated with a benign course of disease with higher recurrence rates but less progression [7]. Being aware of this diagnostic-biological "dilemma", we delineated in this study the immunohistochemical and genetic basis of such FGFR3 high /Ki67 high papillary bladder cancers in order to reveal their prognostic impact.

Altered Molecular FGFR3/TP53 Status Predicts Worse Patients' Survival
Since we hypothesized that FGFR3-overexpression and high cell proliferation might indicate a higher risk for progression in papillary non-invasive tumors, we further investigated the molecular status of our cohort by studying both mutations for FGFR3 as papillary and TP53 as invasive markers (for detailed mutation data, see Table S2). In total, 48 out of 99 (48.5%) analyzed patients harbored mutations within the FGFR3 gene ( Figure 4A).
Since we hypothesized that FGFR3-overexpression and high cell proliferation might indicate a higher risk for progression in papillary non-invasive tumors, we further investigated the molecular status of our cohort by studying both mutations for FGFR3 as papillary and TP53 as invasive markers (for detailed mutation data, see Table S2). In total, 48 out of 99 (48.5%) analyzed patients harbored mutations within the FGFR3 gene ( Figure 4A).    Finally, we assessed the clinical impact of immunohistochemical and mutational status as a combined approach. Survival analysis displayed that FGFR3 mut /TP53 mut double mutated tumors were significantly associated with worse patients' outcome only in FGFR3 and Ki67 overexpressing tumors: reduced PFS in FGFR3 high /Ki67 high double mutated tumors compared with all other combinations of molecular status of TP53 and FGFR3 (FGFR3 wt /TP53 wt p = 0.001; FGFR3 mut /TP53 wt p < 0.001; FGFR3 wt /TP53 mut p = 0.116) (data not shown). However, it has to be noted that the number of double mutations is very low, and, hence, statistical validity should be enhanced in future studies. Finally, we assessed the clinical impact of immunohistochemical and mutational status as a combined approach. Survival analysis displayed that FGFR3 mut /TP53 mut double mutated tumors were significantly associated with worse patients' outcome only in FGFR3 and Ki67 overexpressing tumors: reduced PFS in FGFR3 high /Ki67 high double mutated tumors compared with all other combinations of molecular status of TP53 and FGFR3 (FGFR3 wt /TP53 wt p = 0.001; FGFR3 mut /TP53 wt p < 0.001; FGFR3 wt /TP53 mut p = 0.116) (data not shown). However, it has to be noted that the number of double mutations is very low, and, hence, statistical validity should be enhanced in future studies.

FGFR3 high /Ki67 high Tumors Define a Subset of pTa Tumors Including Lesions with Molecular FGFR3 Pathway Activation
Prognostic stratification of bladder cancer patients in routine histopathological diagnostics claims simple and cost-effective means, therefore, we evaluated the concordance of immunohistochemical staining results and molecular status.
The majority of FGFR3 high /Ki67 high tumors was characterized by conjunct FGFR3 mutations with a significant correlation only in papillary non-invasive tumors (pTa n = 19/27 (70.4%), p < 0.001) but not in pT1 and pT2-4 tumors (Table S5). There was no significant correlation between FGFR3 high /Ki67 high tumors and TP53 mutations independently of the given tumor stage (data not shown).
To evaluate the diagnostic potential of immunohistochemical markers covering the molecular FGFR3 pathway, we performed ROC (Receiver operating characteristics) curve statistics to calculate sensitivity and specificity. Accordingly, both immunohistochemical markers detect FGFR3 mutations with 90.5% sensitivity and 61.9% specificity (area under curve (AUC): 0.776, p = 0.004, positive predictive value (PPV): 70.4%, negative predictive value (NPV): 85.7%). These data show that FGFR3 high /Ki67 high tumors include papillary lesions with mutation-based altered FGFR3 signaling, but also tumors without molecular alterations (pTa n = 8/27 (29.6%)). Hence, our data give evidence that FGFR3 high /Ki67 high tumors define a subset of pTa tumors associated with poor prognosis potentially decoupled from the described protective effect of FGFR3 activation [7].

Discussion
In our study, we systematically analyzed papillary non-invasive and invasive tumors for distinct prognostic immunohistochemical and molecular markers. We focused on a subgroup of immunohistochemically FGFR3 high /Ki67 high tumors in order to reveal their prognostic impact on patient survival and re-evaluate their histological classification/grading.
Although, according to nuclear and architectural criteria, these papillary tumors appear to be orderly and more "nuclear low grade", we found them associated with worse PFS compared with FGFR3 high /Ki67 low tumors. This was especially evident in pTa tumors, where mean progression-free survival was reduced to 55 instead of 113 months. Therefore, we asked whether these tumors harbor a special molecular phenotype turning them into aggressive ones. In literature FGFR3 and TP53 mutations were initially thought to be mutually exclusive as FGFR3 mutations were associated with pTa and LG tumors ("papillary pathway"), whereas the TP53 mutations were often found in invasive and HG carcinomas ("CIS/invasive pathway") [17,18]. Notwithstanding, Hernandez et al. reported FGFR3 and TP53 mutations to be independently distributed in a large series of pT1G3 tumors, that were consequently interpreted as a particular group of bladder tumors that could not be classified into either one pathway or the other [4]. In our study, we saw a similar trend with well-known inverse relationships between FGFR3 and TP53 mutations for both stage and grade, while mutations in FGFR3 and TP53 revealed an independent but not mutually exclusive assignment (six tumors with double mutations). Biologically activated FGFR3 signaling promotes cell proliferation and tumor growth, however interestingly, highest numbers of FGFR3-alterations are found in benign papillary or low grade papillary tumors with usually low proliferation (Ki67) index [19][20][21]. TP53 inactivation results in reduced cellular apoptosis and thus maintains tumor growth via reduced cell death [22][23][24][25]. We hypothesized that a FGFR3 high /Ki67 high phenotype might be resulting from inactivated p53, however we found no sufficient molecular evidence for this theory in our cohort. Recent comprehensive sequencing data of papillary non-invasive bladder tumors revealed a genomic subtype 2, which is characterized by loss of 9q (including TSC1), increased Ki67 labeling index, upregulated mTORC1 signaling, glycolysis, features of the unfolded protein response, altered cholesterol homeostasis and DNA repair [15]. Therefore, high proliferation might be explained by mutations in DNA repair genes or the deletion/mutation of TSC1, which consequently leads towards an upregulation of mTORC1 and PIK3CA mutations. Further analyses to strengthen this theory have to be performed in the future.
Comprehensive molecular data of bladder cancer has been gained in the recent years [15,26,27], however, complex multigene analysis and RNA expression analysis are costly and laborious, and therefore cost-effective simple analyses for routine histological examination are needed. In our study, we analyzed whether fast and simple immunohistochemical analyses are suitable to detect a more aggressive molecular subtype. We found a highly significant correlation between strong FGFR3/Ki67 immunohistochemical staining and FGFR3 mutation status, which indicates that FGFR3 protein expression is more frequent than mutational activation [16]. Moreover, our FGFR3 high /Ki67 high subgroup also comprises those neoplasms without any molecular (FGFR3 and/or TP53) alterations defining in this combination a subset of pTa tumors with poor prognosis, i.e., FGFR3 overexpression was associated with unfavorable outcome as previously shown, for instance, for invasive bladder tumors treated with adjuvant chemotherapy [28]. Thus, our data support the proposed clinical significance of these two immunohistochemical markers for diagnostic and prognostic stratification of more aggressive papillary non-invasive bladder tumors.
Taken together, we found immunohistochemically FGFR3 high /Ki67 high pTa tumors associated with worse prognosis/survival, despite appearing histologically of "lower nuclear grade"/G2. Even if these tumors appear to be "low grade" (according to the 2004 WHO classification), we recommend classifying them as "high grade" pTa tumors. In light of our findings, we suggest immunohistochemical staining for FGFR3 and Ki67 in order to gain evidence for this more aggressive molecular subgroup with worse prognosis. These patients probably could profit from close endoscopic follow-up, as especially urine cytology might also be challenging/less sensitive due to their minimal nuclear changes.

Patient Samples, Tissue Microarrays and DNA
We retrospectively selected urothelial bladder cancer cases (mutational analysis: n = 42 pTa, n = 39 pT1, n = 18 pT2-4; immunohistochemical analysis: n = 82 pTa, n = 42 pT1, n = 18 pT2-4) from our pathology archive and from the archive of the Institute of Pathology in Erlangen. Formalin-fixed paraffin-embedded (FFPE) surgical specimens were used to construct tissue microarrays (all samples) and extract DNA (n = 99 samples) using Qiagen kits (Qiagen, Hilden, Germany) as previously described [29][30][31]. Patient information was obtained by the Department of Urology and the local ethics committee approved a retrospective, pseudonymized study of archival tissues (RWTH EK 009/12). Histological tumor grade and stage was classified according to WHO 2004 classification [8].

Sanger Sequencing
PCR-amplification of exons 7, 10 and 15 of the FGFR3 gene and exons 5, 6, 7, 8 and 9 of TP53 were carried out using routine protocols. Primers and annealing temperatures are given in Table S6. PCR products were purified by either ExoSAP-IT (Affymetrix, Lahr/Schwarzwald, Germany) or a PCR purification kit (PerkinElmer Chemagen, Baesweiler, Germany) according to the manufacturer's instructions. Sanger sequencing of both strands was run on an ABI PRISM 3500 Genetic Analyzer (Applied Biosystems, Weiterstadt, Germany) using the Big dye Terminator kit (Applied Biosystems), the same primer sets and the seq purification kit (PerkinElmer Chemagen).

Statistical Analysis
Statistical analysis was performed using SPSS (Statistical Package for the Social Sciences) software version 23.0 (SPSS Inc., Chicago, IL, USA). p-values < 0.05 were considered significant. Statistical associations between clinico-pathological and molecular factors were determined by Fisher's exact test. Correlation analysis was performed by calculating a Spearman's rank correlation coefficient. Survival (progression-free survival (PFS)) was calculated using the Kaplan-Meier method with log-rank statistics. Survival was measured from surgery until relapse, death or progression and was censored for patients alive without evidence of event at the last follow-up date. Multivariate Cox-regression analysis was performed to test for an independently prognostic value of FGFR3-Ki67 protein expression and FGFR3-TP53 mutations. Receiver operating characteristics (ROC) curves were calculated to assess biomarker performance of immunohistochemical markers regarding molecular alterations.