Collaborative study from the Bladder Cancer Advocacy Network for the genomic analysis of metastatic urothelial cancer

Authors

Damrauer JS

Beckabir W

Klomp J

Zhou M

Plimack ER

Galsky MD

Grivas P

Hahn NM

O’Donnell PH

Iyer G

Quinn DI

Vincent BG

Quale DZ

Wobker SE

Hoadley KA

Kim WY

Milowsky MI

Doi

10.1038/s41467-022-33980-9

PMID: 36333289 · DOI: 10.1038/s41467-022-33980-9 · Journal: Nature Communications (2022)

TL;DR

The UC-GENOME study (NCT02643043) performed integrated DNA and RNA sequencing on 218 patients with metastatic urothelial carcinoma (mUC) collected across eight academic centers. While 69.3% of patients had NGS-identified treatment options, only 5.0% received targeted therapy based on results. The study found an enrichment of Stroma-rich molecular subtypes and TP53 E285K hotspot mutations in metastatic versus non-metastatic cohorts, confirmed APOBEC mutational signatures as prognostically important, and developed an elastic net model integrating clinical and immunogenomic features that predicted immune checkpoint inhibitor (ICI) response with AUC 0.84 in an independent validation set, outperforming TMB alone.

Cohort & data

  • 218 patients with metastatic urothelial carcinoma (BLCA) enrolled at 8 academic centers (UNC, Fox Chase, Mount Sinai, UW/Fred Hutch, JHU, U Chicago, MSK, USC), coordinated by HCRN.
  • Dataset deposited in cBioPortal as blca_bcan_hcrn_2022.
  • DNA sequencing of 591 genes (Caris MI TumorSeek 592-Gene NGS Panel) on 191 patients; total RNA-seq on 176 patients; 147 samples with complete molecular profiling plus clinical data.
  • Assays: targeted DNA sequencing (Agilent SureSelect XT on Illumina NextSeq), RNA-seq (Illumina TruSeq RiboZero Gold on HiSeq 4000), Archer FusionPlex Solid Tumor Panel for fusions.
  • Comparator cohorts: TCGA-BLCA (non-metastatic), IMvigor210 (metastatic, atezolizumab-treated), Kamoun et al. consensus subtypes, and UNC-108.

Key findings

  • Molecular subtype distribution differs in metastatic disease. Stroma-rich tumors were significantly enriched in metastatic cohorts (UC-GENOME, IMvigor210) compared to non-metastatic cohorts (TCGA, Kamoun) (Mantel-Haenszel chi-squared p = 1.86e-10). This enrichment persisted when restricted to primary-site specimens only.
  • TP53 E285K enrichment in metastatic UC. The TP53 E285K hotspot variant was the most frequent TP53 mutation in UC-GENOME (11% of TP53 mutations) and was observed at higher frequency in metastatic cohorts (UC-GENOME 10.8%, IMvigor210 6.9%) than in primary TCGA-BLCA (5.9%). Bladder cancers accounted for 39% (15/38) of all TP53 E285K mutations across the pan-cancer TCGA cohort despite representing only 5.6% of TP53 mutations overall.
  • APOBEC signatures predict treatment outcome. Low cosine similarity to APOBEC signature SBS13 was associated with decreased survival from time of treatment initiation for both chemotherapy (HR 2.50, 95% CI 1.2-5.2, adjusted p = 0.013) and ICI (HR 2.19, 95% CI 1.2-4.0, adjusted p = 0.011). Overall survival from diagnosis was not significantly different, suggesting SBS13 is predictive rather than merely prognostic.
  • Ba/Sq subtype is T cell inflamed; luminal subtypes enriched for B cells. Ba/Sq tumors had significantly higher T cell inflamed and IFNG immune gene signature scores. Luminal tumors (LumP in particular) were enriched for plasma cells, memory B cells, and activated dendritic cells – cell types associated with tertiary lymphoid structures.
  • Inflamed immune phenotype enriched in Ba/Sq. IHC-based CD8 immune phenotyping (n = 155) showed 60% Excluded, 39.4% Inflamed, 0.6% Desert. Ba/Sq tumors were significantly enriched for Inflamed phenotype (vs Stroma-rich p = 0.005, vs LumP p = 0.01, vs LumNS p = 0.0009). Patients without ICI clinical benefit were significantly enriched for the Excluded phenotype (p = 0.0126).
  • ERCC2 mutations predict chemotherapy response in the metastatic setting. ERCC2 mutations were associated with significantly higher chemotherapy response rate (p = 0.0134) but not ICI response, validating this biomarker in metastatic disease.
  • ATM/FANCC/RB1 mutations and ICI survival. Patients with RB1, ATM, or FANCC mutations who received ICI showed improved overall survival (p = 0.045) and increased clinical benefit, but not response.
  • Elastic net ICI predictor outperforms TMB alone. A 25-predictor elastic net model trained on IMvigor210 achieved AUC = 0.84 (IMvigor210 validation), 0.82 (UNC-108), and 0.65 (UC-GENOME), significantly outperforming TMB alone (AUC 0.84 vs 0.68, p = 0.038). The model was not predictive of chemotherapy response (AUC 0.536), confirming ICI specificity.

Genes & alterations

  • TP53 – most frequently mutated gene; increased in Ba/Sq and LumU subtypes; E285K hotspot enriched in metastatic cohorts (10.8% in UC-GENOME vs 5.9% in TCGA-BLCA); APOBEC-attributable hotspot mutation.
  • RB1 – frequently mutated; mutations associated with improved ICI survival when combined with ATM/FANCC alterations.
  • KMT2D, ARID1A, KDM6A, KMT2C, EP300, KMT2A, CREBBP, SMARCA4 – chromatin-modifying genes with high prevalence of non-silent variants, consistent with TCGA-BLCA.
  • BRCA2, PRKDC, ATM, ERCC2, FANCA, FANCC – DNA damage repair genes; ERCC2 mutations enriched in chemotherapy responders; ATM/FANCC/RB1 mutations associated with improved ICI survival.
  • FGFR3 – S249C hotspot; mutations enriched in LumP tumors (OR 7.2, p = 6.1e-5); FGFR3-TACC3 fusions found in 4% of evaluable samples.
  • PIK3CA – E545K hotspot variant observed.
  • TACC3 – fusion partner with FGFR3 (4% of evaluable samples).
  • CCND1 – amplification in 10% of samples.
  • MDM2 – amplification in 9% of samples.
  • ERBB2 – amplification in 7% of samples.
  • CCNE1 – amplification in 2% of samples.
  • EGFR – amplification in 3% of samples.

Clinical implications

  • NGS utility gap. Despite 69.3% of patients having NGS-identified treatment options, only 5.0% received targeted therapy based on results, highlighting a gap between molecular findings and clinical action in metastatic UC.
  • ERCC2 as a platinum biomarker in metastatic disease. ERCC2 mutations were significantly associated with chemotherapy response, validating this biomarker beyond the neoadjuvant setting where it was originally described.
  • Integrated ICI prediction. The elastic net model incorporating TMB, ECOG status, molecular subtype, and immune gene signatures outperformed TMB alone for predicting ICI response, suggesting multivariate models may better guide immunotherapy selection.
  • Stroma as an ICI resistance marker. The Excluded immune phenotype and stromal signatures (FTBRS, EMT_Stroma) were associated with ICI non-benefit, reinforcing the role of tumor stroma in immune exclusion.
  • APOBEC activity and outcomes. Low APOBEC signature (SBS13) was predictive of worse outcomes on both chemotherapy and ICI, suggesting APOBEC activity level may inform treatment decisions.

Limitations & open questions

  • Targeted DNA sequencing (591 genes) rather than whole-exome or whole-genome sequencing was used, potentially missing relevant mutations outside the panel.
  • Response was assessed by investigator assessment without formal response criteria (e.g., RECIST), limiting the precision of response data.
  • The elastic net model performed less well in UC-GENOME (AUC 0.65) than in IMvigor210/UNC-108 (AUC 0.82-0.84), possibly due to differences in RNA-seq methodology (total RNA-seq vs capture-based RNA-seq).
  • Analyses are correlative; functional validation is needed to establish causality for the TP53 E285K enrichment in metastatic disease and the putative APOBEC-driven progression model.
  • The EN model is specific to bladder cancer cohorts as it incorporates UC-specific molecular subtypes (e.g., Consensus Stroma-rich).
  • Only 5.0% of patients received targeted therapy based on NGS, limiting the power to evaluate clinical impact of genomic-guided treatment decisions.

Citations from this paper used in the wiki

  • “69.3% of subjects had potential treatment options, however only 5.0% received therapy based on NGS” (Abstract).
  • “The most frequent TP53 variant was E285K (11%)… The E285 variant was observed at a higher frequency in the metastatic cohorts (UC-GENOME, 10.8%, and IMvigor210, 6.9%) than in the primary TCGA cohort (5.9%)” (Results, p. 4-5).
  • “FGFR3 mutations were enriched in LumP tumors (OR 7.2, p-val 6.1e-5)” (Results, p. 4).
  • “Having a low CS to SBS13, but not SBS2, was associated with decreased survival… for both chemotherapy (p = 0.013) and ICIs (p = 0.011)” (Results, p. 4).
  • “ERCC2 mutations were associated with a significantly higher response to chemotherapy, validating this biomarker in a metastatic setting” (Results, p. 7).
  • “The final EN model consisted of 25 predictors… AUC = 0.84 [IMvigor210 validation] and AUC = 0.82 [UNC-108]… significantly better at predicting ICI response than a model using TMB alone (AUC = 0.84 vs. 0.68, p = 0.038)” (Results, p. 7).
  • “A significant increase in proportion of stroma-rich tumors was observed in the metastatic datasets, IMvigor210 and UC-GENOME, compared to non-metastatic cohorts, TCGA and Kamoun” (Results, p. 2-3).

This page was processed by crosslinker on 2026-05-06.