Are Molecular Markers for Bladder Cancer Ready for Prime Time?

Article Highlights

• There has been an exponential increase in our understanding of the molecular and genetic basis for the development and progression of urothelial carcinoma (UC), including recognition of histologic subtypes and tumor variants, next-generation sequencing, gene-expression profiling, and personalized treatment selection.
• Therapy response for muscle-invasive bladder cancer and metastatic UC has been evaluated based on molecular classification of tumors, but data highlight a need to better understand molecular signatures associated with immune checkpoint inhibitor therapy response.
• In non–muscle invasive bladder cancer, studies are ongoing, but clinical intervention trials are needed before we can determine clinical benefit of the biomarker-based risk stratification.

In recent years, there has been an exponential increase in our understanding of the molecular and genetic basis for, and the mechanisms which underlie, the development and progression of malignancies. Urothelial carcinoma (UC) is no exception: The treatment of bladder tumors, long based on histologic analysis of the nuclear grade and depth of invasion, now routinely incorporates nuances in recognition of histologic subtypes and variants of tumors. The development of next-generation sequencing and gene-expression profiling is emerging as an important part of risk stratification of UC, and rapid advances are being pursued in using these data in personalized treatment selection for our patients.

Herein, we focus on select areas in non–muscle invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC)/metastatic UC (mUC) with the goal of addressing this question: Are molecular and genetic markers ready for prime time use in clinical practice?

Non–Muscle Invasive Bladder Cancer

Cystoscopy with cytologic analysis of cells in the urine remains the gold standard for detection of NMIBC. Currently, the following urine-based tests are available for clinical use (e.g., Urovysion FISH, ImmunoCyt, NMP22, the bladder tumor antigen [BTA stat] test, and Cxbladder) have a narrow clinical utility and suffer from poor sensitivity at the cost of specificity, or vice versa. The preponderance of evidence suggests a role for FISH testing (Urovysion) in predicting response to intravesical immunotherapy and bacillus Calmette-Guérin (BCG) treatment failure.¹ Urovysion FISH detects aneuploidy of chromosomes 3, 7, and 17 as well as deletions of chromosomal locus 9p21. A prospective study by Lotan et al. validated this finding and demonstrated that a positive Urovysion FISH result within 3 months of intravesical therapy was associated with a 3.3-fold increased recurrence risk, thus identifying a possible tool for trial risk stratification.² In addition, a 12-gene real-time polymerase chain reaction (PCR) assay was developed and evaluated prospectively in a series of 750 patients by the UROMOL consortium.³ This 12-gene progression score showed independent prognostic power beyond clinical and histopathologic risk factors in predicting disease progression to MIBC. Furthermore, in an additional study of the same patient cohort, addition of FGFR3 mutation status and GATA2 methylation status to the European Association of Urology clinical risk-stratification system was shown to increase the accuracy of the prediction of progression to MIBC.⁴ We caution, however, that both these modalities should be incorporated into clinical risk calculators rather than be used in isolation.

The largest analysis of gene expression (RNA-sequencing) data from 460 prospectively enrolled patients (by UROMOL) demonstrated that NMIBC could be divided into three distinct molecular classes with different biological features and different risk of progression to MIBC. Molecular classes were not associated with BCG response.⁵ UC has a high mutational burden when compared with other solid malignancies. A common theme is the recognition of the high frequency of DNA damage response (DDR) gene defects and high mutational burden often associated with APOBEC-related mutagenesis activity. In the UROMOL study, a high frequency of mutations in an APOBEC context (APOBEC signature) in NMIBC was found to be associated with molecular high risk (class 2) and disease progression to MIBC. In addition, molecular high-risk tumors were significantly associated with mutations in the genes TP53 and ERBB2. The high mutational burden in UC, caused in part by the propensity of DDR and APOBEC-related mutations, might obscure delineation and analysis of disease development and targetable pathways, as mutations might arise and associate with multiple, redundant pathways.

Targeted sequencing of a prespecified gene panel was performed for 105 index tumors in treatment-naive patients with adequate clinicopathologic correlates and NMIBC. FGFR3 and STAG2 mutations were significantly associated with low-grade tumors, and TP53 and DDR pathway mutations were significantly associated with high-grade tumors Not unexpectedly, the overall mutational burden was significantly higher in high-grade tumors, especially in those with DDR mutations. Novel potential biomarkers for prognosticating BCG therapy response (ARID1A), as well as potential targets for pharmacotherapy, were also noted.⁶ In addition, this study identified a high prevalence of TERT promotor mutations across NMIBC, suggesting a possible marker for early detection of recurrence. This confirms findings from a prior retrospective analysis, which demonstrated TERT promotor mutations in over 70% of UC and developed an assay for mutation detection, demonstrating promising specificity and sensitivity dependent on the clinical scenario.⁷ Unfortunately, in all the aforementioned studies, follow-up has been limited, progression events are few, and the proportion of patients with carcinoma in situ (which is increasing in prevalence, especially in patients who have tumors after BCG therapy) is limited or has not been evaluated. Thus, clinical intervention trials are needed before we can determine clinical benefit of the biomarker-based risk stratification for NMIBC.

“Clinical intervention trials are needed before we can determine clinical benefit of the biomarker-based risk stratification for NMIBC.” —Dr. Nathan Brooks, Dr. Lars Dyrskjøt, and Dr. Ashish M. Kamat

Muscle-Invasive Bladder Cancer

In 2012, the first molecular taxonomy for bladder cancer was published,⁸ and the field was soon flooded with increasingly informative findings. Generally, these classification systems broadly categorized MIBC subtypes as either basal or luminal, with differing subclassifications within each cohort. To provide a uniform platform, the Bladder Cancer Molecular Taxonomy Group was formed to develop a consensus of six molecular subtypes for MIBC, derived and validated in a total of 1,750 samples.⁹ The consensus system offers a robust framework that will enable identification and validation of predictive biomarkers in future clinical trials. The classifier is constrained by the limitations of the original works: by being retrospective in nature, by lacking complete clinical annotation, and by using data developed across different platforms. The Table summarizes this consensus classification system.

Therapy response for MIBC and mUC has been evaluated based on molecular classification of tumors. Choi et al. first reported that p53-like MIBCs were resistant to cisplatin-based neoadjuvant chemotherapy (NAC) in 73 patients.¹⁰ Seiler et al. evaluated the molecular subtypes of 343 transurethral resection specimens of MIBC where NAC was used, comparing survival outcomes using subtyping based on the University of North Carolina, The University of Texas MD Anderson Cancer Center, The Cancer Genome Atlas (TCGA), and Lund University subtypes.¹¹ Results were compared with outcomes from subtyping in the TCGA dataset, where no NAC was administered but where patients were potentially treated with chemotherapy at the time of metastasis. Overall survival for patients appeared subtype-dependent between the cohorts. Patients with the basal subtype showed better overall survival in the NAC cohort compared with the TCGA cohort. Notably, the subtyping was not shown to be significantly associated with pathologic response, which is of some concern given the implications otherwise stated by the authors. Overall, these results must be interpreted with caution, as comparisons were generated across patient cohorts. Although patient risk stratification and decisions regarding NAC based on molecular markers show promise, validation in prospective studies, ideally in the context of the MIBC consensus classification system, are necessary for diffusion into clinical practice.

The utility of tumor subtypes in predicting response to immunotherapy (checkpoint inhibitors) has also been studied with varying findings. In the IMvigor210 trial, it was retrospectively observed that luminal cluster II tumors exhibited a significantly higher overall response rate (ORR) with atezolizumab, while in CheckMate 275 (with nivolumab), a higher ORR was seen in the TCGA basal I subtype.^12,13 Necchi reported on the prospective correlation of molecular characteristics in the neoadjuvant setting with pembrolizumab in the PURE-01 study. Interim biomarker results from the study emphasized the role of biomarkers in pretherapy transurethral resection of a bladder tumor (TURBT) tissue samples and found them to be associated with pathologic response in small subsets of patients with nonurothelial, variant histology. Additionally, transcriptome profiling from PURE-01, compared with a retrospective multicenter cohort of patients with MIBC treated with NAC and radical cystectomy, revealed that molecular subtypes with higher pre-existing levels of immune infiltration have a favorable clinical response to neoadjuvant pembrolizumab, but not to platinum-based NAC. Different outcomes were also obtained according to the molecular subtype, with the neuroendocrine subtype revealing the worst prognosis regardless of the type of neoadjuvant therapy.^14,15

Analysis of the similar ABACUS trial of neoadjuvant atezolizumab for localized MIBC demonstrated that 64% of tumors changed molecular taxonomy groups (Lund taxonomy system) during the course of treatment without significant alteration of DNA. Tumors responsive to therapy demonstrated consistent upregulation of tissue repair pathways. Tumors that demonstrated increased expression of cell cycle and proliferation genes were more likely to relapse. The findings in the ABACUS trial were contrary to those seen in the mUC setting, suggesting that gene expression profiles of tumors might be dynamic and altered based on the clinical setting and the treatment administered.¹⁶

These results highlight a need to better understand molecular signatures associated with immune checkpoint inhibitor therapy response.

“The evidence supporting tumor-centric biomarkers and subtyping remains too immature for clinical use and should be tested in new cohorts and clinical trials.”—Dr. Nathan Brooks, Dr. Lars Dyrskjøt, and Dr. Ashish M. Kamat

Conclusion

To answer the question of whether molecular markers for urothelial bladder cancer are ready for prime time: No, not yet. The evidence supporting tumor-centric biomarkers and subtyping remains too immature for clinical use and should be tested in new cohorts and clinical trials. We should apply new approaches/techniques to delineate tumor microenvironment interactions before, during, and after treatment. Tumor heterogeneity influences biomarker discovery, and large molecular differences within primary tumors and between primary tumors and metastasis have been found. Biomarker studies including analysis of metastasis need to be performed in larger cohorts. Finally, the tumor-centric analysis should be combined with liquid biopsy analysis to better monitor disease development and treatment response during treatment.¹⁷