Comprehensive molecular characterization of urothelial bladder carcinoma

Author

The Cancer Genome Atlas Research Network

Doi

10.1038/nature12965

PMID: 24476821 · DOI: 10.1038/nature12965 · Journal: Nature (2014)

TL;DR

The TCGA Research Network performed an integrated multi-platform molecular characterization of 131 chemotherapy-naive, muscle-invasive, high-grade urothelial bladder carcinomas (T2–T4a). Whole-exome sequencing, low-pass whole-genome sequencing, RNA-seq, miRNA-seq, DNA methylation, RPPA, and SNP6.0 arrays identified 32 significantly mutated genes (including 9 not previously reported as significantly mutated in any cancer), recurrent in-frame activating FGFR3–TACC3 fusions, viral integrations (HPV16, BK polyomavirus, CMV), and four mRNA expression subtypes (papillary-like and basal/squamous-like among them). The study established BLCA as the cancer with the highest frequency of chromatin-regulator mutations among epithelial cancers studied to date and identified potential therapeutic targets in 69% of tumours.

Cohort & data

  • n=131 chemotherapy-naive, muscle-invasive, high-grade urothelial bladder carcinomas (stages T2–T4a, Nx, Mx), with matched peripheral blood (n=118) and/or tumour-adjacent normal bladder tissue (n=23). Samples from 19 tissue source sites.
  • Tumour nuclei ≥60%, necrosis ≤20%, variant histology (squamous or small cell) ≤50%.
  • Cancer type: BLCA (urothelial carcinoma of the bladder).
  • Dataset / cohort: blca_tcga_pub (The Cancer Genome Atlas — Bladder Urothelial Carcinoma, 2014 publication).
  • Assays: whole-exome-seq of 130 tumour/normal pairs (186,260 exons across 18,091 genes; mean coverage 100×, ≥82% target bases covered ≥30×); low-pass paired-end whole-genome-seq of 114 tumours (6–8× coverage); rna-seq of all tumours (n=129 for mRNA clustering); miRNA-seq; affymetrix-snp6 for somatic CNAs; hm450-methylation-array for DNA methylation; rppa for protein/phosphoprotein expression.
  • Analytic tools: mutect for somatic point mutations, mutsig (MutSig 1.5) for significantly mutated genes, gistic (GISTIC 2.0) for focal CNAs.

Key findings

  • Mutation burden. Mean and median somatic mutation rates of 7.7 and 5.5 mutations per Mb. Average 302 exonic mutations, 204 segmental copy-number alterations, and 22 genomic rearrangements per sample — slightly lower than lung cancer and melanoma, but higher than most other adult malignancies studied by TCGA at the time.
  • 32 significantly mutated genes by MutSig 1.5 (FDR < 0.1), including 9 not previously reported as significantly mutated in any cancer: CDKN1A, ERCC2, RXRA, ELF3, KLF5, FOXQ1, RHOB, PAIP1, and BTG2. MutSig also flagged three genes via COSMIC restriction: ERBB2, ATM, and CTNNB1.
  • Top mutated genes (frequency). TP53 49%, ARID1A 25%, KMT2D (MLL2) 27%, KDM6A 24%, PIK3CA 20%, EP300 15%, CDKN1A 14%, RB1 13%, ERCC2 12%, FGFR3 12%, STAG2 11%, ERBB3 11%, FBXW7 10%, RXRA 9%, TSC1 8%, NFE2L2 8%, KLF5 8%, TXNIP 7%, HRAS 5%.
  • TP53 pathway inactivated in 76% of tumours. TP53 mutations (49%) were mutually exclusive with MDM2 amplification (9%) and overexpression (29%).
  • Focal CNAs (GISTIC2.0). Most common focal deletion (47% of samples) contained CDKN2A (9p21.3) and correlated with reduced expression. Other focal deletions targeted RB1, CREBBP, and others. Focal amplifications affected PPARG, E2F3, EGFR, CCND1, MDM2, ERBB2, BCL2L1, ZNF703, and NECTIN4 (PVRL4).
  • Recurrent FGFR3–TACC3 fusions. In-frame activating fusions detected in 3 of 114 WGS tumours; breakpoints in FGFR3 intron 16 (n=2) or exon 17 (n=1) fused to TACC3 intron 10. Predicted fusion protein contains the N-terminal 758 aa of FGFR3 fused to the C-terminal 191 aa of TACC3 and is predicted to auto-dimerise.
  • ERBB2 rearrangements in 4 tumours with different fusion partners, including a DIP2B–ERBB2 promoter swap that amplified ERBB2.
  • Viral integration. Viral DNAs detected in 7 of 122 tumours (6%); viral transcripts in 5 of 122 (4%). One HPV16-positive tumour had the virus integrated into BCL2L1, which was amplified (~6×) and overexpressed (~10× median). One BK polyomavirus tumour had two BK genes integrated into GRB14. Three tumours expressed CMV transcripts (no integration evidence).
  • APOBEC-mediated mutagenesis. 51% of mutations overall were Tp*C → (T/G) — a signature recently attributed to APOBEC cytosine deaminases. APOBEC3B was highly expressed in all tumours, indicating a major mutagenic role.
  • Chromatin-regulator mutations dominate. 99 of 131 tumours (76%) had an inactivating mutation in at least one chromatin regulatory gene; 53 (41%) had at least two. Four epigenetic regulators were MutSig-significant (KMT2D, ARID1A, KDM6A, EP300). KMT2D and KDM6A mutations were mutually exclusive. Non-silent mutations in chromatin regulators were significantly enriched in bladder cancer compared with all other epithelial cancers studied by TCGA at the time (Supplementary Table 2.10).
  • Three mutation/CNA clusters by NMF (n=125). Group A (focally amplified, enriched in KMT2D mutations); Group B (“papillary CDKN2A-deficient FGFR3-mutant”, enriched in papillary histology); Group C (“TP53/cell-cycle-mutant”, with near-universal TP53 mutation, RB1 mutation and E2F3/CCNE1 amplification).
  • Four mRNA expression clusters (n=129). Cluster I (“papillary-like”) enriched in papillary morphology (P=0.0002), FGFR3 mutation (P=0.0007, q=0.02), FGFR3 copy gain (P=0.04), elevated FGFR3 expression (P<0.0001), and harboured all three FGFR3–TACC3 fusions. Cluster III (“basal/squamous-like”) expressed epithelial lineage genes KRT14, KRT5, KRT6A, and EGFR, resembling basal-like breast and squamous head-and-neck/lung cancers. Clusters I and II resembled luminal A breast cancer with high GATA3 and FOXA1, high uroplakin (UPK3A), elevated HER2/ERBB2 protein, and elevated ESR2 signalling.
  • DNA methylation. 34% of tumours formed a CIMP subgroup; smoking pack-years was the only significant predictor of CIMP phenotype on multivariate regression (age, sex, stage as covariates).
  • Splicing. Average 3% PKM1 / 97% PKM2 transcripts across tumours, consistent with the Warburg shift.
  • Pathway alterations. Cell cycle 93%; chromatin remodelling 89% (histone-modifying genes + 64% SWI/SNF complex); kinase/PI(3)K signalling 72%. PI(3)K/AKT/mTOR pathway altered in 42% (38% by gene-set frequency in Fig. 5): PIK3CA 17%, TSC1/TSC2 9%, AKT3 overexpression 10%. RTK/RAS pathway altered in 44% (39%): FGFR3 17% (mutation/amplification/fusion), EGFR amplification 9%, ERBB2 mutation or amplification 9%, ERBB3 mutation 6%.
  • Smoking and mutations. 72% of patients had current or past smoking history, but unlike lung cancer there was no statistically significant association between smoking status and the mutational spectrum, mutation frequency in any significantly mutated gene, focal CNAs, or expression subtype. Never-smokers had slightly higher fraction of C→G mutations (28.5% vs 23.8%, P=0.032).

Genes & alterations

  • TP53 — mutation in 49%; inactivation (mutation + MDM2 amp/overexpression) in 76%. Mutually exclusive with MDM2 amplification.
  • RB1 — mostly inactivating mutations (13%); associated with reduced mRNA; mutually exclusive with CDKN2A deletions.
  • CDKN2A — focal homozygous deletion in 47% of tumours.
  • FGFR3 — kinase-activating mutations (12%) at canonical hotspots; recurrent in-frame FGFR3–TACC3 fusions (n=3); copy-number gain. Enriched in mRNA cluster I (papillary-like).
  • PIK3CA — mutations in 20%, clustered in the helical domain near E545.
  • TSC1 — truncating mutations in 8%; six homozygous.
  • KMT2D (MLL2) — mutated in 27%; truncating enriched; encodes H3K4 methyltransferase; novel SMG in bladder cancer; mutually exclusive with KDM6A.
  • KDM6A — mutated in 24%; H3K27 demethylase; mutually exclusive with KMT2D.
  • ARID1A — mutated in 25%; truncating enriched; SWI/SNF component.
  • EP300, CREBBP — significantly mutated histone acetyltransferases (15% and lower frequencies).
  • CDKN1A — mutated in 14%; predominantly null/truncating, indicating loss of function; encodes p21CIP1 cyclin-dependent kinase inhibitor.
  • ERCC2 — mutated in 12%; 15 of 16 mutations were deleterious missense (dominant-negative); ERCC2-mutant tumours had significantly fewer C→G mutations and trended toward higher overall mutation rate.
  • STAG2 — predominantly inactivating mutations (11%); X-linked cohesin complex component.
  • ERBB3 — mutated in 11% (6% in pathway analysis), potentially sensitive to ERBB kinase inhibitors.
  • ERBB2 — mutation or amplification in 9% (significant by COSMIC restriction); ERBB2 alteration frequency comparable to TCGA breast cancer but with fewer amplifications and more mutations.
  • FBXW7 — mutated in 10%.
  • RXRA — mutations in 9%; 7 of 12 at hotspot S427 (5 S427F, 2 S427Y) in the ligand-binding domain; mutant tumours showed elevated adipogenesis/lipid-metabolism gene expression, suggesting constitutive activation.
  • ELF3 — mutated in 8%; novel SMG.
  • NFE2L2 — deleterious missense in the Neh2 domain in 11 tumours (8%); mutually exclusive with TXNIP mutations; mutant tumours upregulated genotoxic-metabolism and ROS-response genes.
  • TXNIP — mutated in 7% (5 inactivating); mutually exclusive with NFE2L2.
  • KLF5, FOXQ1, PAIP1, ZFP36L1, CCND3, FOXA1, HRAS — additional SMGs at ≤8%.
  • ATM, CTNNB1 — significant by COSMIC restriction.
  • TACC3 — fusion partner of FGFR3 (n=3).
  • APOBEC3B — highly expressed across all tumours; likely driver of TpC mutational signature representing 51% of mutations.
  • Focal amplifications: PPARG (17%), E2F3 (20% — peak with SOX4), CCND1 (10%), EGFR (11%), MDM2 (9%), ERBB2 (7%), CCNE1 (12%), BCL2L1 (11%), ZNF703 (10%), NECTIN4 (PVRL4, 19%).

Clinical implications

  • Potential therapeutic targets in 69% of tumours. 42% had targets in the PI(3)K/AKT/mTOR pathway; 45% had targets (including ERBB2) in the RTK/MAPK pathway.
  • FGFR3-targeted therapy. Tumours with FGFR3 mutations, copy-number gain, FGFR3–TACC3 fusions, or that share the cluster I expression profile may respond to FGFR inhibitors or downstream pathway inhibitors.
  • ERBB2/HER2-targeted therapy. ERBB2 amplifications and recurrent mutations support ongoing ERBB2-targeted trials (paper cites NCT01353222 as an example trial of an immunotherapy approach in HER2-overexpressing bladder cancer); HER2 protein levels in a subset of tumours are comparable to those in TCGA HER2-positive breast cancers.
  • PI(3)K/AKT/mTOR axis. Activating PIK3CA mutations (potentially PI(3)K-inhibitor responsive), TSC1/TSC2 loss (potentially mTOR-inhibitor responsive), and AKT3 overexpression (potentially AKT-inhibitor responsive).
  • Chromatin-targeting therapy. Bromodomain inhibitors and other agents targeting chromatin modifications could be useful for the 76% of tumours with chromatin regulator mutations.
  • HSP90 inhibitors identified as a potential target by network analysis (HSP90AA1 as a critical signalling hub).
  • Metabolism / Warburg effect. Near-uniform PKM2-dominant splicing suggests potential sensitivity to glycolysis inhibitors.
  • Hormone-receptor signalling. mRNA clusters I and II showed elevated ESR2 (oestrogen receptor beta) signalling, suggesting potential targeting with tamoxifen or raloxifene.
  • Baseline. The authors note that no molecularly targeted agent had been approved for bladder cancer as of publication; median survival for recurrent/metastatic disease was 14–15 months with cisplatin-based chemotherapy.

Limitations & open questions

  • Cohort is exclusively chemotherapy-naive, muscle-invasive, high-grade urothelial carcinoma — findings may not generalise to non-muscle-invasive or low-grade disease.
  • Variant histology (squamous or small cell) capped at ≤50%, so heavily variant tumours are under-represented.
  • Low-pass WGS (6–8×) limits sensitivity for some structural variants.
  • Although smoking status was associated with CIMP phenotype, no significant association was found between smoking and mutational spectrum, SMG frequency, CNAs, or expression subtype — leaving the mechanistic link between smoking and bladder carcinogenesis unresolved.
  • Network analyses linking chromatin-regulator mutations to transcription-factor activity (TieDIE/PARADIGM) were statistically significant by proximity permutation tests, but the authors call out that further studies will be needed to assess biological relevance.
  • Biological significance of two of the four ERBB2 fusion products (other than the DIP2B–ERBB2 promoter swap) is unknown.
  • Whether the cluster III (basal/squamous-like) “stem-cell” signature (KRT14, KRT5) translates to therapeutic targetability remains to be tested.

Citations from this paper used in the wiki

  • “We report here an integrated analysis of 131 urothelial carcinomas to provide a comprehensive landscape of molecular alterations. There were statistically significant recurrent mutations in 32 genes, including multiple genes involved in cell-cycle regulation, chromatin regulation, and kinase signalling pathways, as well as 9 genes not previously reported as significantly mutated in any cancer.” (Abstract)
  • “Our analyses identified potential therapeutic targets in 69% of the tumours, including 42% with targets in the phosphatidylinositol-3-OH kinase/AKT/mTOR pathway and 45% with targets (including ERBB2) in the RTK/MAPK pathway.” (Abstract)
  • “Chromatin regulatory genes were more frequently mutated in urothelial carcinoma than in any other common cancer studied so far, indicating the future possibility of targeted therapy for chromatin abnormalities.” (Abstract)
  • “Nearly half (49%) of the samples had TP53 mutations… hence, TP53 function was inactivated in 76% of samples.” (p. 316)
  • “The most common recurrent focal deletion, seen in 47% of samples, contained CDKN2A (9p21.3) and correlated with reduced expression.” (p. 316)
  • “We found several recurrent translocations of probable pathogenic significance, including an intra-chromosomal translocation on chromosome 4 involving FGFR3 and TACC3 (n=3).” (p. 317)
  • “We identified viral DNAs in 7 of 122 tumours (6%), and viral transcripts in 5 of 122 (4%)… In the HPV-16-expressing case, the virus integrated into BCL2L1, an apoptosis-regulating gene.” (p. 317)
  • “Fifty-one per cent of mutations overall were Tp*C → (T/G)… APOBEC3B was expressed at high levels in all of the tumours, suggesting a major role for APOBEC-mediated mutagenesis in bladder carcinogenesis.” (pp. 317–318)
  • “Ninety-nine (76%) of the tumours analysed here had an inactivating mutation in one or more of the chromatin regulatory genes, and 53 (41%) had at least two such mutations.” (p. 320)
  • “Median survival for patients with recurrent or metastatic bladder cancer remains 14–15 months with cisplatin-based chemotherapy, and there is no widely recognized second-line therapy.” (p. 320)

This page was processed by crosslinker on 2026-05-09.