EURURO Vol. 72 No. 3

358 476

biomarkers characteristic of immune and other stromal
cells.
Analysis of a large cohort of 460 tumors recently
identified three distinct subtypes (Classes 1–3) in NMIBC
[16_TD$DIFF] The study validated previously observed subgroups in
NMIBC and identified a new subtype (Class 3) that showed
basal-like characteristics and pronounced expression of
lncRNAs. Mutations were called from RNA-seq data, which
have potential limitations. Frequent mutations were
observed in genes encoding proteins involved in chromatin
organization and cytoskeletal functions. Furthermore, high-
risk tumors (Class 2) were enriched for mutations in, for
example, TP53 and ERBB2, and for APOBEC-related muta-
tions. The observation of enrichment for APOBEC-related
mutations in high-risk NMIBCs indicates that APOBEC may
drive disease progression in NMIBC.
Other groups performed independent studies to identify
molecular subtypes in cohorts of MIBC [56_TD$DIFF] A group led
by one of us (W.K., University of North Carolina [UNC], Chapel
Hill, NC, USA) assembled a meta-dataset of 262 high-grade
tumors from four previously published cohorts for discovery
and created a new dataset from 49 tumors collected at
Memorial Sloan Kettering Cancer Center for validation. Using
consensus clustering, they identified two molecular subtypes
of MIBC in both datasets [57_TD$DIFF] The genes that distinguished
the two clusters had previously been implicated in urothelial
differentiation and overlapped substantially with the genes
that distinguished the basal-like and luminal intrinsic
subtypes of breast cancer, leading the UNC investigators to
name their MIBC subtypes ‘‘basal like’’ and ‘‘luminal’’ [57_TD$DIFF] In
parallel, another of our groups (W.C. and D.J.M., University of
Texas MD Anderson Cancer Center, Houston, TX USA) created
two whole transcriptome datasets (n = 73 and n = 57 tumors)
and used unsupervised hierarchical clustering to identify
three candidate subtypes [58_TD$DIFF] . The MD Anderson group also
noted that the genes that characterized two of the subtypes
were similar to the ones that distinguished basal-like and
luminal breast cancers, prompting them to term them ‘‘basal’’
and ‘‘luminal’’ [58_TD$DIFF] The group’s third subtype was distin-
guished from the other two by stromal biomarkers and an
active p53 gene expression signature, so they termed it ‘‘p53-
like’’ [58_TD$DIFF] Finally, TCGA used a combination of different
approaches to identify four molecular subtypes in an RNA-
seq dataset generated from 129 tumors [15_TD$DIFF] Although TCGA
discussed whether or not to name them ‘‘basal’’ and
‘‘luminal’’ in the first marker paper, they settled on a more
conservative approach and termed them ‘‘clusters I–IV’’ but
discussed their similarities to the intrinsic subtypes of breast
cancer [15_TD$DIFF] In addition, TCGA noted that cluster I was
enriched with tumors having papillary features and cluster III
with tumors having squamous features, inspiring the labels
‘‘papillary’’ and ‘‘squamous,’’ respectively [15_TD$DIFF] As introduced
above, a group based at the Broad Institute then repeated the
exercise on an interim expanded cohort of 238 tumors and
largely reproduced their original data [5_TD$DIFF] They renamed the
clusters ‘‘basal,’’ ‘‘immune undifferentiated,’’ ‘‘luminal im-
mune,’’ and ‘‘luminal’’ to reflect the dominant features they
found in the gene expression signatures that defined each
subtype [5_TD$DIFF]
Parallel efforts attempted to reconcile the molecular
subtypes identified by the groups using shared whole
transcriptome datasets [59_TD$DIFF] [60_TD$DIFF] The results of this work
revealed a significant overlap among the subtype calls. The
most upstream division occurred at the level of the basal
versus luminal bifurcation proposed by the group at UNC
[59_TD$DIFF] [60_TD$DIFF] The other subtypes appeared to mostly represent
subdivisions of these subtypes. The MD Anderson p53-like
subtype, which shared similarities with the Lund infiltrated
subtype [53_TD$DIFF] , TCGA’s original cluster II [15_TD$DIFF] , and the updated
TCGA ‘‘immune undifferentiated’’ and ‘‘luminal immune’’
subtypes [5_TD$DIFF] , consisted of a mixture of basal and luminal
tumors that were heavily infiltrated with stromal cells, and
the Lund uroA and GU tumors largely corresponded to UNC
luminal tumors. Although the uroB tumors were assigned to
the basal subtypes identified by other groups, the Lund
group recognized that the uroB tumors contained FGFR3
pathway gene expression signatures and were enriched
with activating FGFR3 mutations, and concluded that they
probably corresponded to progressed uroA tumors [53_TD$DIFF] .
Subtype membership had important implications for
clinical outcomes. Similar to basal-like breast cancers the
Lund SCCL and uroB tumors and the squamous/basal tumors
identified by the groups at UNC, MD Anderson Cancer Center,
and TCGA were aggressive, and associated with advanced
stage and metastatic disease at presentation, squamous
histopathological features, and shorter survival in the
absence of neoadjuvant cisplatin-based combination che-
motherapy [61_TD$DIFF] However, about half of basal
tumors were downstaged by NAC [58_TD$DIFF] , and early preliminary
data suggested that the overall benefit provided by NAC
might be greatest in patients whose tumors belonged to the
basal molecular subtype [62_TD$DIFF] . If these preliminary data are
confirmed in larger and more mature clinical datasets, the
results would be reminiscent of past experience in breast
cancer, where NAC has produced the greatest benefit in
patients with highly proliferative basal-like (and HER2-
enriched) tumors [10_TD$DIFF] Although the molecular mecha-
nisms that underlie the benefit produced by chemotherapy
in basal tumors are still under investigation, basal human
bladder cancer cell lines are more sensitive to cisplatin-
induced apoptosis than are luminal cell lines (A. Ochoa, D.J.
McConkey, unpublished observations). Conversely, NAC
produced less clinical benefit in patients whose tumors
belong to the infiltrated/p53-like subtype in the clinical trials
that have been performed to date [63_TD$DIFF]
The variable levels of immune cell infiltration observed
in the bladder cancer subtypes [5_TD$DIFF] suggested that patients
with these tumors might derive different amounts of
benefit from immunotherapies [64_TD$DIFF] Consistent with these
predictions, early results suggest that TCGA subtype
membership may be an independent predictor of benefit
from therapy with the anti-PDL1 antibody atezolizumab
[65_TD$DIFF] . In the phase II trial that led to Food and Drug
Administration approval of the drug, patients whose tumors
belonged to TCGA cluster II obtained somewhat more
benefit than patients whose tumors belonged to the other
subtypes, and patients with ‘‘papillary’’ (cluster I) tumors
derived little benefit, if at all [65_TD$DIFF] . Importantly, immune
E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 3 5 4 – 3 6 5 358