**Abstract**

Bladder cancer is a biologically and clinically heterogeneous disease. Traditional classification systems, based on pathologic grade, stage and clinical prognosis fail to fully explain how tumors with similar pathology exhibit diverse biological behavior. The introduction of transcriptomics technology has allowed us to catalog all of the mRNA expression patterns and DNA alterations in a given tumor thus expanding our understanding of human cancers. Molecular subtype profiling was attempted only recently in bladder cancer, with the earliest attempts dating back to 2010. Several different molecular classification systems have emerged since. Some of these systems address early bladder cancer, while others focus exclusively on the life-threatening muscle invasive tumors. These molecular subtypes have distinct morphological and clinical characteristics with different therapeutic and prognostic implications, particularly in the era of targeted therapies and immunotherapy. However, molecular subtyping is not without its limitations. Despite the rapidly expanding evidence for important clinical implications, much work is still needed to establish the utility (or lack thereof) of molecular subtyping, and its application in daily practice.

**Keywords:** molecular classification, muscle invasive bladder cancer, taxonomy, transcriptomics, targeted therapy

### **1. Introduction**

Bladder cancer is the most common malignant tumor involving the urinary tract. Histologically, bladder cancers can be urothelial carcinomas, squamous cell carcinomas and adenocarcinomas, out of which urothelial carcinomas constitute over 90% [1]. Urothelial carcinoma is derived from the specialized epithelia of the bladder wall. Using a well-established differentiation program, basal stem cells at the stromal interface self-renew and generate intermediate and superficial urothelial cells to maintain and regenerate the urothelium in response to daily wear and tear. Bladder cancer results from the deregulation of this program. Conventionally, bladder cancer can be divided into non-muscle invasive bladder cancer (NMIBCs) and muscle invasive bladder cancers (MIBCs), based on the invasion of the muscularis propria. The low grade superficial NMIBCs and high grade MIBCs (HG MIBCs) develop along divergent molecular pathways of tumorigenesis and also show diverse biological behavior and molecular profile. Low grade NMIBCs constitute the majority of newly diagnosed bladder cancer cases, typically have a long protracted clinical course characterized by multiple recurrences, and require life-long monitoring, significantly contributing to bladder cancer morbidity. On the other hand, a large proportion of the MIBCs eventually metastasize, contributing to the bulk of bladder cancer mortality [2, 3].

Low grade papillary urothelial carcinomas occur due to fibroblast growth factor receptor 3 (FGFR3)/RAS/RAF pathway alterations, while the HG MIBCs develop along the TP53/RB1 mutation pathway [4, 5]. Chromosome 9 deletion occurs in the early phase of bladder cancer tumorigenesis. FGFR3/HRAS mutations frequently occur during the development of hyperplasia and low grade (Ta) carcinoma. Hyperplasia develops into high grade urothelial carcinoma (Ta) through the acquisition of CDKN2A alterations, which may progress to become T1 carcinoma after additional TP53/RB1 inactivation. TP53 mutations frequently occur during the development of urothelial dysplasia. These may develop into carcinoma in situ (Tis) after RB1 inactivation, which then progresses through non-muscle invasive infiltrating urothelial carcinomas (T1) to muscle invasive (T2) carcinoma.

Traditional classifications for bladder cancer are mainly based on pathological features and tumor stage. However, even with similar pathological staging and grading, recurrence and progression of bladder cancer shows marked heterogeneity, and directly affects optimal monitoring and treatment response. Only a proportion of cases of bladder cancer of a given grade and stage will progress to a higher stage. Thus the same treatment, such as, transurethral resection alone, or the administration of BCG or neoadjuvant chemotherapy may not be adequate for others. Some tumors are less likely to metastasise and need only local resection, while others are highly invasive and need radical cystectomy and/or other treatments. Currently, there is still no effective means to distinguish between the two. The pathological features of the tumor cannot fully reflect the "intrinsic characteristics" of bladder cancer.

With the rapid development of sequencing, mass spectrometry and other techniques, studies based on the '-omics' technology has transformed our understanding of human cancers. The basic method involves cataloging the entire mRNA expression pattern and DNA alteration profile by sequencing, microarray and other technologies and then performing a cluster analysis of the different genes based on gene expression levels and genes involved in a given biological process. After performing hierarchical cluster analysis on mRNA expression profiling data, clusters were validated by DNA PCR and/or immunohistochemistry (IHC), DNA methylation profiling, miRNA or lncRNA analysis. Among the earliest applications of this approach was to define intrinsic molecular subtypes in human breast cancer by Perou et al., [6]. Subsequently, over the last decade, molecular subtypes with distinct clinical behavior, histology and response to treatment were identified in other malignancies e.g. colon cancer, gliomas, acute leukemias and so on. Molecular subtype profiling was applied only recently in bladder cancer, with the earliest attempts dating back to 2010. Several different molecular classification systems have emerged since, with four standing out in the MIBCs, developed by Lund University group [7–10], The Cancer Genome Atlas (TCGA) consortium [11, 12], MD Anderson Cancer Centre [13] and the University of North Carolina [14]. Several other studies have followed up, in an attempt to unify [15] and reach a consensus [16] between the classification systems. Although each group defended the existence of a different numbers of subtypes (n = 2–6), there was remarkable overall concordance among the groups (**Table 1**). At the highest level, all classification systems recognized the existence of intrinsic luminal and non-luminal (basal) subtypes, which resembled normal luminal/intermediate and basal urothelial cells in gene expression profile.

For the two major types of bladder cancer, NMIBC and MIBC, molecular subtyping of bladder cancer can be divided into early subtyping (which included both NMIBC and MIBC), NMIBC subtyping and MIBC subtyping (**Table 2**).


#### **Table 1.**

*Table showing overlap between the subtypes in the initial classification systems and the cardinal properties of these subtypes, adapted from McConkey et al. [17], CAF-cancer associated fibroblasts, EMT- epithelialmesenchymal transition.*


#### **Table 2.**

*Summary of major molecular subtyping systems, adapted from Zhu et al., 2020 [18].*
