Key points

Introduction

Muscle-invasive bladder cancers (MIBCs) are heterogeneous tumors with variable progression patterns and responses to frontline therapies. Although approximately half of patients will be cured of their disease with definitive surgery with or without perioperative cisplatin-based chemotherapy, the others experience rapid progression and uniform mortality. There has been essentially no substantive progress in systemic therapy since the current cisplatin-based regimens were introduced more than 30 years ago. Many patients receive clear clinical benefit from chemotherapy, but no strategies have been developed to prospectively identify them; it has been estimated that their overall impact on disease-specific survival is only 5% to 15%. The perception that perioperative chemotherapy provides only a modest benefit has resulted in its underutilization in patients with potentially lethal tumors, while at the same time our inability to prospectively identify chemoresistant tumors has resulted in the treatment of patients who received no clinical benefit. Finally, no biologically targeted agents have been approved to date, and no real clinical signal has emerged from the trials that have been performed with most of the traditional biological targets.

There are many indications that a dramatic change in this dismal picture is about to occur. The first phase of The Cancer Genome Atlas’ (TCGA’s) bladder cancer project and several parallel private efforts have provided the first deep view of bladder cancer heterogeneity and have stimulated renewed enthusiasm for developing targeted agents in bladder cancer. Whole genome sequencing also provided the research community with its first targeted agent success story in the identification of TSC1 and NF2 mutations in a patient with bladder cancer whose tumor displayed a remarkable durable complete response to everolimus. In addition, a community-wide effort to exploit the scientific opportunities afforded by the use of neoadjuvant therapy has produced a very-high-profile clinical trial (the Southwest Oncology Group’s CoXEN (coexpression extrapolation) trial) that will allow for the prospective evaluation of multiple methods to predict sensitivity and resistance to chemotherapy, and a similar clinical trial that will assign patients to receive multiple targeted agents based on their tumor genomic profiles will open soon. Finally, exciting preliminary results from a phase I-II clinical trial of a blocking anti-PDL1 antibody suggest that immune checkpoint blockade is clinically active in a significant fraction of bladder cancers. Together, these developments have prompted aggressive efforts to design and launch clinical trials that exploit our newly developed understanding of the biological properties of MIBCs stemming from these genomics efforts. However, past experience in other cancer types suggests that taking advantage of this information may not be entirely straightforward. Rather, it may be necessary to obtain a profound understanding of the biological mechanisms that explain why the alterations and particular patterns of gene expression are observed in certain tumors (and not others) and to determine exactly what they do for the cancer cells before the full potential of the genomics discoveries can be exploited.

It is now well established that cancer heterogeneity is controlled by both genetic and epigenetic mechanisms. One powerful strategy that has been used to visualize important epigenetic aspects of this heterogeneity is to use whole genome RNA expression profiling and unsupervised analyses to group tumors into intrinsic subtypes without considering their clinical properties (or sometimes even their tissues of origin). This strategy has had a major impact on breast cancer, whereby knowledge of a tumor intrinsic subtype is used for clinical management. Patients with luminal breast cancers are managed with surgery plus adjuvant endocrine therapy, whereas patients with basal-like or HER2-enriched tumors receive no benefit from endocrine therapy. Conversely, patients with basal-like or HER2-enriched tumors obtain substantial clinical benefit from perioperative chemotherapy, whereas most patients with luminal tumors do not. These patterns of response and resistance are not obviously associated with tumor mutation profiles but rather seem to be more closely related to the tumors’ cells of origin. It is important that we consider and learn from this experience as the bladder cancer research community tries to fully exploit the emerging genomics information.

Introduction

Muscle-invasive bladder cancers (MIBCs) are heterogeneous tumors with variable progression patterns and responses to frontline therapies. Although approximately half of patients will be cured of their disease with definitive surgery with or without perioperative cisplatin-based chemotherapy, the others experience rapid progression and uniform mortality. There has been essentially no substantive progress in systemic therapy since the current cisplatin-based regimens were introduced more than 30 years ago. Many patients receive clear clinical benefit from chemotherapy, but no strategies have been developed to prospectively identify them; it has been estimated that their overall impact on disease-specific survival is only 5% to 15%. The perception that perioperative chemotherapy provides only a modest benefit has resulted in its underutilization in patients with potentially lethal tumors, while at the same time our inability to prospectively identify chemoresistant tumors has resulted in the treatment of patients who received no clinical benefit. Finally, no biologically targeted agents have been approved to date, and no real clinical signal has emerged from the trials that have been performed with most of the traditional biological targets.

There are many indications that a dramatic change in this dismal picture is about to occur. The first phase of The Cancer Genome Atlas’ (TCGA’s) bladder cancer project and several parallel private efforts have provided the first deep view of bladder cancer heterogeneity and have stimulated renewed enthusiasm for developing targeted agents in bladder cancer. Whole genome sequencing also provided the research community with its first targeted agent success story in the identification of TSC1 and NF2 mutations in a patient with bladder cancer whose tumor displayed a remarkable durable complete response to everolimus. In addition, a community-wide effort to exploit the scientific opportunities afforded by the use of neoadjuvant therapy has produced a very-high-profile clinical trial (the Southwest Oncology Group’s CoXEN (coexpression extrapolation) trial) that will allow for the prospective evaluation of multiple methods to predict sensitivity and resistance to chemotherapy, and a similar clinical trial that will assign patients to receive multiple targeted agents based on their tumor genomic profiles will open soon. Finally, exciting preliminary results from a phase I-II clinical trial of a blocking anti-PDL1 antibody suggest that immune checkpoint blockade is clinically active in a significant fraction of bladder cancers. Together, these developments have prompted aggressive efforts to design and launch clinical trials that exploit our newly developed understanding of the biological properties of MIBCs stemming from these genomics efforts. However, past experience in other cancer types suggests that taking advantage of this information may not be entirely straightforward. Rather, it may be necessary to obtain a profound understanding of the biological mechanisms that explain why the alterations and particular patterns of gene expression are observed in certain tumors (and not others) and to determine exactly what they do for the cancer cells before the full potential of the genomics discoveries can be exploited.

It is now well established that cancer heterogeneity is controlled by both genetic and epigenetic mechanisms. One powerful strategy that has been used to visualize important epigenetic aspects of this heterogeneity is to use whole genome RNA expression profiling and unsupervised analyses to group tumors into intrinsic subtypes without considering their clinical properties (or sometimes even their tissues of origin). This strategy has had a major impact on breast cancer, whereby knowledge of a tumor intrinsic subtype is used for clinical management. Patients with luminal breast cancers are managed with surgery plus adjuvant endocrine therapy, whereas patients with basal-like or HER2-enriched tumors receive no benefit from endocrine therapy. Conversely, patients with basal-like or HER2-enriched tumors obtain substantial clinical benefit from perioperative chemotherapy, whereas most patients with luminal tumors do not. These patterns of response and resistance are not obviously associated with tumor mutation profiles but rather seem to be more closely related to the tumors’ cells of origin. It is important that we consider and learn from this experience as the bladder cancer research community tries to fully exploit the emerging genomics information.

Intrinsic subtypes of breast cancer

Perou and colleagues first observed the intrinsic molecular subtypes of breast cancer in a relatively small cohort of 65 tumors from 42 different patients. They used an in-house cDNA microarray that contained a total of 8102 different probes, filtered out the genes that displayed the greatest variations in expression across the dataset, and then used the filtered genes to perform an unbiased (unsupervised hierarchical clustering) analysis of the relationships among the gene expression patterns they observed in the tumors. They discovered that the tumors formed 4 distinct groups, and they recognized that 2 of them contained basal-like and luminal gene expression patterns that were similar to the ones found in cells at different stages of differentiation in the normal mammary epithelium. They also appreciated that one subtype contained a signature characteristic of ERBB2 ( HER2 ) amplification. The duplicated tumors within their discovery cohort consisted of samples taken before and 16 weeks after treatment with neoadjuvant chemotherapy as well as several matched primary tumor–lymph node metastasis pairs. In almost all cases the clustering analyses revealed that these matched tumors were more similar to each other than they were to any of the other tumors in the dataset; in particular, the matched specimens always segregated to the same gene expression clusters. Therefore, they suggested that the gene expression profiles that dictated subtype membership were intrinsic to the tumor (ie, stable) and did not depend on the site or timing of biopsy or even prior exposure to DNA-damaging chemotherapeutic agents. Since then, the breast cancer intrinsic subtypes have been consistently observed whenever cohorts of breast cancers are subjected to unsupervised hierarchical clustering.

The TCGA’s breast cancer project afforded the opportunity to comprehensively characterize the genomic alterations associated with each intrinsic subtype. Comparisons of tumor subtype membership as determined by unsupervised hierarchical clustering versus the use of an intrinsic subtype classifier (the PAM50) revealed strong concordance between the two, supporting the original conclusion that the intrinsic subtypes are associated with stable biological properties. Overall, the basal-like and HER2-enriched tumors exhibited the highest overall mutation rates consistent with greater genomic instability; but the luminal tumors contained a broader spectrum of different mutations. Luminal tumors contained more activating mutations in PIK3CA and inactivating mutations in the differentiation-associated transcription factor GATA3 , whereas the basal-like and HER2-enriched cancers had more p53 mutations, and the types of p53 mutations observed in the basal-like cancers were more damaging (nonsense and truncating). One possible explanation for these differences in p53 mutation patterns could be that mutant p53 plays an oncogenic (gain-of-function) role in the luminal cancers, whereas complete loss of function is more important in the basal and HER2-enriched tumors. Copy number variations (CNVs) also correlated with the tumor intrinsic subtypes, with the greatest frequencies observed in the basal-like and HER2-enriched tumors.

The clinical characteristics of tumors within each breast cancer intrinsic subtype are also distinct. Basal-like and HER2-enriched tumors are associated with shorter disease-specific and overall survival, particularly in the absence of neoadjuvant or adjuvant chemotherapy. However, most basal-like and HER2-enriched tumors respond to neoadjuvant chemotherapy ; patients whose tumors achieve a complete pathologic response have excellent outcomes. Patients with luminal A tumors have the longest disease-specific and overall survival, and most patients with luminal B tumors also have excellent clinical outcomes. These patients are treated with surgery and adjuvant hormonal therapy. Although luminal cancers are associated with excellent 5-year overall survival, they are also more often associated with late relapse, usually characterized by bone metastasis. Together, the compiled genomic and clinical results support the original suggestions made by Perou and colleagues that knowledge of a given tumor’s intrinsic subtype carries important clinical information and that each subtype should be treated as a distinct disease entity.

Intrinsic subtypes of bladder cancer

One of the most important objectives in cancer research is to develop molecular classifiers that can inform prognostication and predict therapeutic efficacy. It might seem that the most straightforward way to do this would be to assemble cohorts of tumors from patients who experienced extreme clinical outcomes (ie, very short vs very long disease-specific survival or complete pathologic response vs progression with neoadjuvant or adjuvant chemotherapy), perform deep genomic profiling, and correlate the outcome with the genomic features observed. This strategy has been used successfully to develop prognostic classifiers, such as Genomic Health’s Oncotype Dx platforms (Genomic Health, Inc, Redwood City, CA) or GenomeDx’s Decipher classifier (GenomeDx Biosciences Inc, Vancouver, BC, Canada and San Diego, CA). Several academic groups have used this strategy to identify biomarker panels that correlate with clinical characteristics in bladder cancers, including recurrence, disease-specific survival, and/or metastasis; but the reproducibility of these signatures has been questioned. From a biological perspective, the signatures do share general similarities across disease types (usually biomarkers associated with proliferation and tissue-specific differentiation). Indeed, high levels of Ki-67 and other cell cycle/proliferation biomarkers are consistently associated with poor prognosis in MIBC, just as they are in breast or prostate cancers. Past difficulties in reproducing specific signatures may be related to the use of different genomic platforms and tumor heterogeneity. Alternatively, it is possible that better classifiers could be developed by focusing on the gene expression differences associated with extreme clinical outcomes within a given intrinsic subtype.

A group at Lund University was the first to intentionally search for intrinsic subtypes in bladder cancers. They performed whole genome mRNA expression profiling on a mixed cohort of more than 300 non-muscle invasive bladder cancers (NMIBCs) and MIBCs. They concluded that the tumors could be easily separated into 2 major subtypes, one that consisted almost entirely of grade 1 and 2 tumors (MS1) and another that contained mostly high-grade (G3) tumors (MS2). Parallel analyses of mutations and focal genomic amplifications (FGAs) revealed that the MS1 tumors were enriched with activating mutations in FGFR3 and PIK3CA and low levels of genomic instability, whereas MS2 tumors were enriched with mutations in TP53 , amplification of E2F3 , RB deletions, and large numbers of FGAs consistent with genomic instability. In a subsequent study they used unsupervised hierarchical clustering to further define subtypes within the MS1 and MS2 tumors. They concluded that the MS1 tumors could be subdivided into 2 subtypes and that the MS2 tumors formed up to 5 distinct subtypes. Importantly, they recognized that 2 of the MS2 subtypes (urobasal B and squamous cell carcinoma (SCC)-like) were enriched with high-molecular-weight basal cytokeratins, whereas 2 others (grouped together and termed genomically unstable ) were enriched with low-molecular-weight luminal cytokeratins and uroplakins, which are urothelial terminal differentiation markers. Finally, one of the subtypes seemed to be infiltrated with stromal cells because it was enriched with extracellular matrix, immune cell, and fibroblast biomarkers; tumors in this infiltrated subtype also expressed low levels of genes involved in the late cell cycle. They consistently observed these subtypes in multiple independent data sets and concluded that they corresponded to the intrinsic subtypes of bladder cancer. Parallel work by another group used immunohistochemistry to show that MIBCs expressed basal and terminal differentiation-associated cytokeratins in a mutually exclusive fashion and concluded that cancers could be separated into subtypes resembling different states of differentiation within the normal urothelium.

Parallel efforts by the authors’ group and 2 others confirmed and extended these initial observations. Using whole genome mRNA expression profiling and unsupervised analyses, the authors’ group concluded that MIBCs could be grouped into 3 distinct subtypes characterized by differential expression of breast basal and luminal biomarkers. In parallel, another group used a very similar approach on a large metadata set and concluded that the tumors formed 2 distinct basal and luminal clusters. They also demonstrated that there was a subset of basal MIBCs that were very similar to the so-called claudin-low breast cancers. Finally, using RNA sequencing (RNAseq) data from 128 MIBCs, the TCGA concluded that the tumors could be grouped into 4 mRNA expression-based subtypes and recognized that they were enriched with breast basal and luminal biomarkers. Direct head-to-head comparisons of the 3 groups’ calls in the TCGA data set revealed that the intrinsic subtypes identified by the 3 groups were extremely similar. The differences in the number of clusters were explained by the fact that 2 of the groups subdivided the luminal tumors into 2 subsets, and the TCGA also subdivided the basal tumors into 2 subsets ( Fig. 1 ). The significance of these divisions in the basal and luminal MIBCs is discussed later.

Relationships between the intrinsic subtypes identified by the groups at the Lund University, MD Anderson Cancer Center (MDACC), the University of North Carolina (UNC), and the TCGA. Subtype calls made by the groups using the same tumor cohorts were compared. For the comparisons between the MDACC calls and those made by UNC or TCGA, the TCGA RNAseq dataset served as the reference. For the comparisons between the MDACC and Lund calls, the MDACC and Lund Illumina data sets were used. For the actual comparisons, see Choi and colleagues and McConkey and colleagues.

The authors also performed a head-to-head comparison of their subtype calls to the calls made by the group at the University of Lund. Not unexpectedly, the results revealed strong similarities in the subtype calls made by the two groups (see Fig. 1 ). The Lund SCC-like tumors corresponded to the basal subtype; the Lund genomically unstable tumors corresponded to the luminal subtype; and the Lund infiltrated subtype corresponded to the MD Anderson p53-like subtype and the TCGA’s cluster II tumors (see Fig. 1 ). Therefore, 4 independent groups using different cohorts of MIBCs discovered very similar subtypes; there can be very little doubt that they represent the intrinsic subtypes of MIBC.

Like their breast cancer counterparts, each of the intrinsic subtypes of bladder cancer possesses unique clinical characteristics. Patients with basal MIBCs tend to have more advanced stage and metastatic disease at presentation, whereas patients with luminal MIBCs tend to have better clinical outcomes. Basal MIBCs are relatively more prevalent in women than in men and are enriched with squamous histopathologic features. On the other hand, luminal MIBCs were enriched with papillary histopathologic features, and micropapillary MIBCs seem to be mostly luminal in origin (Choi W, Czerniak B, unpublished observations 2014). Finally, luminal MIBCs are enriched with activating FGFR3 mutations, which, as discussed earlier, are present in most low-grade NMIBCs. The presence of papillary features and FGFR3 mutations suggests that luminal MIBCs may represent NMIBCs that have progressed to become muscle invasive. Recent lineage-tracing studies in a mouse model of bladder cancer support this idea.

Biological properties of the intrinsic subtypes

Bioinformatic analyses of the gene expression signatures that characterized the MIBC subtypes revealed roles for transcription factors that had also been implicated in breast cancer. The basal epithelial/stem cell transcription factors ΔNp63α and STAT3 were both identified as central regulators of basal MIBC gene expression, and ΔNp63 has also been implicated in basal breast cancers. Basal expression of active STAT3 promoted carcinogen-induced tumorigenesis in a mouse model ; high-level expression of ΔNp63 in MIBCs was associated with poor clinical outcomes, consistent with their central roles in basal MIBC biology. Aside from STAT3 and ΔNp63, basal cancers displayed gene expression features consistent with active NFκB, c-Myc, and hypoxia-induced factor (HIF) signaling. ΔNp63 knockdown resulted in downregulation of Myc and the HIF signature (data not shown), indicating that both pathways are directly or indirectly connected to ΔNp63.

Luminal MIBCs were enriched with gene expression patterns that were similar to the ones found in luminal breast cancers, including active estrogen receptor (ER), TRIM24, FOXA1, and GATA3 signatures. Luminal MIBCs also contained strong peroxisome proliferator activator receptor-gamma (PPARγ) signatures, which are not recognized as an important feature of luminal breast cancers. (Perou and colleagues’ initial work actually identified PPARγ-associated biomarkers in basal-like cancers.) The TCGA’s analyses of MIBCs demonstrated that PPARG is amplified in a fairly large fraction of tumors (16%), and this PPARG amplification is enriched in the luminal subtypes (TCGA clusters I and II) ( Fig. 2 ).

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Bladder Cancer Nonurothelial Bladder Cancer and Rare Variant Histologies First-line Treatment of Metastatic Disease Diagnosis and Staging of Bladder Cancer Perioperative Therapy for Muscle Invasive Bladder Cancer First-Line Treatment and Prognostic Factors of Metastatic Bladder Cancer for Platinum-Eligible Patients

Stay updated, free articles. Join our Telegram channel

Join