**4. Molecular subtyping of MIBC**

Much of the work on molecular subtyping of bladder cancer has been undertaken with MIBCs in consideration and an increasing number of classification systems have emerged with four of them standing out (**Figure 1**). Although the subtypes within these systems are largely similar, they differ in clinically and biologically meaningful ways.

The UNC classification proposed by Damrauer et al. [14], used K2 consensus clustering, to divide tumors into basal (KRT5/6 and CD44) vs. luminal (PPARG, GATA3, KRT20, and UPK2) subtypes utilizing a 47 gene classifier, BASE47. The basal subtype showed similarities with the basal subtype of breast cancers, as demonstrated by applying the PAM50 signature to their dataset. In addition, like in breast cancers, a claudin-low subgroup was identified among basal tumors. The claudin-low subgroup had outcomes similar to basal tumors and was rich in epithelial-mesenchymal transition (EMT) signatures and tumor initiating cell markers. A significant enrichment in genes related to inflammatory cell infiltration and immune checkpoint was also seen in the basal subgroup and, more specifically, among the claudin-low tumors. There was no significant difference in TP53 pathway alterations in the subtypes. The basal subtype, which was more frequent among females, had a high rate of RB pathway gene alterations, while the luminal subtype was rich in FGFR3 and TSC1mutations.

#### **Figure 1.**

*Figure showing the similarities between the molecular subtypes in the different classification systems. Luminal and basal subtypes exist in all classification systems (figure adapted from Zhu et al. [18]).*

The MDA group [13] also classified BC into basal, p53-like and luminal tumors similar to breast cancer. Their luminal and p53-like subtypes had similar mRNA profiles but wild-type p53 gene was significantly activated in the p53-like subtype. All three subtypes had similar frequency of p53 mutations, but the p53-ness contributed by increased wild-type p53 expression was thought to contribute to chemo-resistance in the p53-subtype. The basal subtype was enriched in squamous cell differentiation markers and activated p63 and was more invasive with poorer clinical prognosis.

The TCGA 2014 [12] subtyped MIBCs into four clusters numbered I to IV utilizing an integrated genomic analysis of chemotherapy-naive, invasive UCs by analyzing for somatic mutations, DNA copy number alterations, mRNA and microRNA expression profile, as well as protein analysis, and DNA methylation studies. Cluster I predominantly exhibited papilloma phenotype and was enriched in FGFR3 mutations. Both cluster I and II expressed GATA3, FOXA1, UPK3A transcription factors and uroplakin family of genes and were enriched in RBB2 mutations and ER beta. Cluster III expressed squamoid phenotype and its associated keratin expression. Subsequently in 2017, TCGA expanded their classification [11] into five distinct subtypes, diving luminal tumors into three subtypes, luminal papillary, luminal and luminal infiltrated. They also included a neural subtype in addition to the earlier described basal subtype.

#### **4.1 Intrinsic molecular subtypes and intra-tumor heterogeneity**

Intrinsic molecular subtype, a term which first used in breast carcinomas [6] refers to subtypes which reflect an intrinsic property of the tumor. The luminal and basal intrinsic subtypes reflect the property of the tumor cells to show urothelial or basal stem-cell-like differentiation signatures [23].

However, transcriptomic studies identified the entire genetic signature of a tumor, which in the case of invasive malignancies, included variable components of stromal and immune signatures. Thus, in addition to the two intrinsic subtypes, some of the subtypes defined in the various study groups were based on the

#### *Molecular Classification of Bladder Cancer DOI: http://dx.doi.org/10.5772/intechopen.97393*

characteristics of the tumor-stroma or tumor infiltrating inflammatory cells such as the p53-like'subtype [13], infiltrated subtype [8], luminal-infiltrated subtype [11] as examples. These non-intrinsic subtypes could be recognized as a property of the stroma or inflammatory cells when studies were performed later which localized the gene expression patterns in situ, by immunohistochemistry [10, 23, 24]. Many of these non-intrinsic subtypes could be resolved into luminal and basal subtypes based on tumor phenotype on immunohistochemistry, while others continued to express non-luminal, non-basal phenotype, like the double negative subtype of Dadhania et al., [24].

## *4.1.1 Intrinsic luminal subtype*

At the top of the hierarchical level, MIBCs were divided into luminal-like and nonluminal-like classes based on the presence or absence of bimodally expressed urothelial differentiation signature.

About half of the MIBCs expressed this signature characterized by the expression of KRT20, UPK1–3 (uroplakin 1, 2 and 3), epithelial biomarkers (E-cadherin/ CDH1 and members of the miR-200 family), along with transcriptional regulators PPARG (peroxisome proliferator activated receptor-γ), GRHL2–3 (grainyhead like transcription factor), ELF3 and TBX2–3. Luminal MIBCs also displayed active Estrogen Receptor/TRIM24 pathway gene expression and were enriched with FOXA1, GATA3, ERBB2 and ERBB3 expressed on superficial (umbrella cells) and intermediate cells of the normal urothelium [7]. There was increased expression of fibroblast growth factor receptor-3 (FGFR3), with activating FGFR3 mutations in the most differentiated luminal tumors.

The luminal tumors were further sub-stratified into urothelial-like (UroA, UroB and UroC) and genomically unstable (GU) subtypes by the Lund University group [8], luminal papillary, luminal and luminal infiltrated subtypes by TCGA [11] and luminal papillary, luminal non-specified, and luminal unstable subtypes in the Consensus classification [16].

The urothelial-like tumors expressed FGFR3 and CCND1, and frequently showed 9p21 (CDKN2A) loss. On immunohistochemistry, only the urothelial-like tumors retained the basal stratification seen in normal urothelium and express CK5 at least focally, particularly at the tumor-stroma interface [9, 10]. Similar to UroA tumors in the Lund classification, the luminal-papillary subtype in TCGA and Consensus classification were also characterized by FGFR3 mutations; by papillary histology; and by low carcinoma-in-situ scores. Such cancers had a low risk for progression, and while preliminary data suggests a low likelihood of response to cisplatin-based NACT [25], they may respond to tyrosine kinase inhibitors of FGFR3 family [26, 27] or to PPARγ-inhibitors [10] or to Estrogen receptor modulators [28].

Genomically unstable subtype (GU) of luminal tumors expressed FOXM1 and absent to low levels of FGFR3, but not KRT5 [9]. They also frequently showed RB1 loss, and had a high rate of TP53 mutations. Highest ERBB2 expression is also seen in GU subtype. Although they showed urothelial differentiation signature, GU tumors were in fact poorly differentiated and frequently high grade on histology [9, 10]. On immunohistochemistry, they did not express CK5, but expressed late cell cycle makers such as p16 [10]. The luminal unstable subtype of Consensus classification showed similar features to the GU subtype described by Lund University group. These tumors may respond to drugs targeting ERBB2 [23].

In terms of prognosis, luminal papillary tumors or UroA tumors had very good prognosis, while the GU subtype showed an intermediate prognosis compared to urothelial-like and basal/SCC-like tumors [10, 12, 16].

#### *4.1.2 Intrinsic non-luminal subtype: Basal and neuronal*

Intrinsic non-luminal MIBC included basal MIBC (Choi et al., 2014) [13] and neuronal or small cell neuroendocrine MIBC. The basal subtype has been renamed basal-squamous in the later classification systems as it is characterized by squamous differentiation [8, 12].

Basal-squamous MIBCs expressed signal biomarkers similar to normal basal cells in the urothelium like high molecular weight cytokeratins KRT1, KRT5, KRT6, KRT14, KRT16, 15 KRT6A, KRT6B, KRT6C and CD44 and CDH3 [8, 11]. However, unlike normal basal urothelial cells which retained urothelial differentiation factors (GATA3 and PPARG), the basal-squamous subtype showed down-regulation of this signature. Interestingly, they had a higher incidence in females unlike all the other subtypes which were male predominant [11].

Basal MIBCs were also characterized by up-regulation of the epidermal growth factor receptor (EGFR) and other ligands of the epidermal differentiation complex such as S100A7 and SPRR1B, similar to basal breast and head and neck squamous cell carcinomas. Cell cycle regulator p63 played a central role in controlling the basal pathway of differentiation, and STAT3, NFκB, and Hypoxia Induced Factor-1α (HIF-1α) were also involved [23].

Without treatment, basal MIBCs had poorer survival [13, 14] but they responded well to neoadjuvant chemotherapy [11]. Because NACT pathological complete response is associated with excellent long-term survival, aggressive early management of basal MIBCs with NACT offers the best chance for improved survival for these patients.

This subtype also had the strongest immune expression signature, including T cell markers, inflammation genes and lymphocytic infiltrates. It is predicted that the basal-squamous subtype may respond to anti-PD-L1, anti-PD-1 and anti-CTLA-4 agents [28, 29]. EGFR-, NFκB, HIF-1α/VEGF, and/or STAT3-targeted agents may also have a role within this subtype [27].

The neuronal subtype showed no histopathological distinction from other types of MIBC in most cases. Nonetheless, they had high levels of TP53 and RB1 mutations, similar to small cell carcinomas in other tissues. It had the worst survival of the mRNA expression subtypes, making it important to recognize [10, 11].

#### *4.1.3 Non-intrinsic subtypes: P53-like, luminal-infiltrated, stromal-like, infiltrated, claudin-low*

P53-like MIBCs showed some overlap in gene expression with luminal and basal subtypes of the UNC classification but were characterized by the expression of an active wild-type p53-associated gene expression signature [13]. P53-like subtype of bladder cancer responded poorly to neo-adjuvant chemotherapy [13, 28, 30]. Wild-type p53-induced reversible senescence and quiescence had been implicated in causing chemo-resistance. However, even though p53-associated expression signatures were present, *TP53* mutation frequencies were found to be similar in luminal, p53-like and basal subtypes defined by Choi et al., (2014). The p53-ness as measured by mRNA expression was found to be a more accurate predictor of de novo and induced MIBC chemo-resistance than analysis of *TP53* mutational status [23].

The luminal-infiltrated subtype reported by the TCGA was characterized by low tumor purity, with high expression of epithelial-mesenchymal transition (EMT) and myofibroblast markers, and of the miR-200 s. It showed medium expression of PD-L1 and CTLA4 immune markers. This subtype had been reported to respond to immune checkpoint inhibitors like Anti-PD-L1 [29].

#### *Molecular Classification of Bladder Cancer DOI: http://dx.doi.org/10.5772/intechopen.97393*

Stromal-like subtype from the Consensus classification, the infiltrated subtype from the Lund classification [10] and claudin-low subtype of the MD Anderson Cancer Centre classification [14] all showed similar features of low-tumor purity, high EMT and stromal related transcripts with increased cancer stem cell-like gene expression profile. Claudin-low tumors described by Damrauer et al. [14], in addition, showed increased expression of claudins-1, 3, and 7 and had a similar expression profile to the claudin-low breast cancer subtype. Dadhania et al., [24] in their meta-analysis of the TCGA, Lund and MD Anderson cohorts also identified a subset of tumors with low urothelial and basal expression signatures, which they termed "double negative", which showed similar expression profile to claudin-low tumors.

With tumor progression, alterations are seen both in the intrinsic characteristics of the tumor cell, as well as in the tumor microenvironment (TME). Early MIBC molecular classification systems mainly focused on the molecular classification of tumor cells themselves. With a deeper understanding of BC cells and their TME, molecular subtyping efforts have begun to focus more on intratumor heterogeneity, stromal-extracellular matrix (ECM) interactions and immune cell infiltration, allowing further refinement of the molecular subtypes. Currently, studies on molecular subtyping are mainly based on whole tumor DNA or RNA studies rather than focusing on a single tumor cell. In this method, intratumoral heterogeneity can greatly affect the accuracy of molecular subtyping. Warrick et al. [31] conducted a pathological examination on 309 bladder cancer markers and found that nearly one fourth of them exhibited intratumoral variation in tissue samples. Out of the 83 specimens subtyped by them with the Lund subtyping system, 39% exhibited molecular heterogeneity. Even among the subtypes, the basal-squamous subtype particularly showed the greatest variability; with approximately 78% of these tumors simultaneously exhibiting the genomically unstable or urothelial-like subtype.

Several immunohistochemistry based algorithms have been developed in an attempt to classify bladder cancer into clinically and prognostically significant molecular subtypes [9, 10, 24, 32]. The use of immunohistochemistry as a surrogate to molecular testing shows promise in making molecular subtyping amenable to widespread use. The use of a simple panel comprising of a luminal urothelial markers like GATA3 and basal keratin marker like KRT5 can help identify a GATA3 positive, KRT5 low luminal subtype and a GATA3 negative, KRT5 high basal subtype. Tumors which are negative for GATA3 and have low keratin may be further tested for mesenchymal or neuroendocrine markers. The luminal tumors may also be further subtyped into uro-like tumors which are p16 negative and genomically unstable tumors which show p16 positivity [33]. The subtypes thus identified have demonstrated significant prognostic and predictive value [24, 32].
