Clonal Evolution of Chemotherapy-resistant Urothelial Carcinoma

Authors

Bishoy M. Faltas

Davide Prandi

Scott T. Tagawa

Ana M. Molina

David M. Nanus

Cora Sternberg

Jonathan Rosenberg

Juan Miguel Mosquera

Brian Robinson

Olivier Elemento

Andrea Sboner

Himisha Beltran

Francesca Demichelis

Mark A. Rubin

Doi

10.1038/ng.3692

PMID: 27749842 · DOI: 10.1038/ng.3692 · Journal: Nature Genetics (2016)

TL;DR

Faltas and colleagues at Weill Cornell performed whole-exome sequencing (WES) of 72 urothelial carcinoma (UC) tumors from 32 patients — including 16 matched pre/post-chemotherapy primary–metastasis sets and two rapid autopsies — to map how platinum-based chemotherapy reshapes UC clonal architecture (PMID:27749842). Only ~28% of mutations were shared between matched pre- and post-chemotherapy samples; phylogenetic reconstruction showed early branching evolution and early metastatic seeding. Post-chemotherapy tumors were clonally enriched for mutations in L1-cell adhesion molecule (L1CAM) and integrin-signaling pathways and showed an APOBEC3A-dominant mutational signature, suggesting chemotherapy continues to actively edit the UC genome and that L1CAM/integrin signaling is a candidate resistance axis.

Cohort & data

  • n = 72 urothelial tumors from 32 patients, prospectively collected at Weill Cornell Medicine; cohort designed to enrich for advanced disease (28/32 patients = 88% with metastatic disease) (PMID:27749842).
  • 16 matched primary + advanced/post-chemotherapy sets plus 2 rapid autopsy cases (WCM117 with 12 samples across 8 anatomical sites; WCM259) (PMID:27749842).
  • Cancer type: bladder urothelial carcinoma (BLCA). Dataset: blca_cornell_2016. Comparator: TCGA bladder cohort (blca_tcga_pub).
  • Assays: whole-exome sequencing using the New York State–approved CLIA-grade EXaCT-1 assay (Agilent HaloPlex Exome, 21,522 genes, mean coverage 85x); orthogonal validation by an N250 targeted panel of 250 cancer genes (SeqCap EZ Choice, mean coverage 400x; Pearson r=0.93 with WES VAFs, P<10⁻¹⁷¹) (PMID:27749842).
  • Analysis: CLONET for tumor purity/ploidy and clonality estimation; SNV calling with MuTect and SNVseeqer; annotation with Oncotator; GSEA over REACTOME pathways; mutational signature decomposition vs Sanger COSMIC signatures; phylogeny via parsimony Ratchet; copy-number validation by FISH. Of 72 samples, 53 passed CLONET QC for CN analysis and 44 (25 patients) had reliable purity/ploidy estimates for clonality work (PMID:27749842).
  • BAMs deposited in dbGaP phs001087.v1.p1 (PMID:27749842).

Key findings

  • Mutational discordance between matched pre- and post-chemotherapy tumors: only 28.4% of mutations were shared on average (range 0.2%–76.4%); effect was consistent for primary–primary and primary–metastasis pairs (p=0.17, Wilcoxon test) (PMID:27749842).
  • Even canonical UC drivers (PIK3CA, KMT2D/MLL2, ATM, TP53) were inconsistently shared between matched samples. Example: patient WCM077’s primary and pelvic LN metastasis shared TP53 p.Y234C, while a post-chemotherapy lung metastasis carried a private TP53 p.G266V — demonstrating parallel, distinct hits in the same gene (PMID:27749842).
  • Early branching evolution: phylogenetic analysis of 21 matched sets (parsimony Ratchet) consistently positioned the primary tumor as a branch (not the trunk) of the tree, indicating multiple cell populations diverged from an ancestral clone in parallel early in tumor life (PMID:27749842).
  • WCM117 rapid autopsy case (12 samples, 16 months): untreated TURBT primary harbored 138 private mutations (many in cisplatin-response genes including POLD2 and FOXP1) that disappeared after gemcitabine–cisplatin chemotherapy. Sub-clonal heterozygous CDKN2A deletion in the primary evolved into a homozygous clonal CDKN2A deletion in distant lymph-node and liver metastases (FISH-confirmed), implying selection under chemotherapy pressure. A non-silent TSPAN8 mutation marked the transition from primary to metastatic state at divergence node 5. At least 5 waves of clonal expansion had already occurred from the lowest common ancestor by the time of initial diagnosis (PMID:27749842).
  • Clonal enrichment after chemotherapy: post-chemotherapy samples showed a significant increase in clonal mutations across the cohort (P=0.0134, Fisher’s exact test). GSEA on REACTOME pathways highlighted clonal enrichment in trans-membrane transport of small molecules (OR=1.9, FDR=0.002), L1CAM signaling (OR=1.9, FDR=0.12), and integrin signaling (OR=2.8, FDR=0.02). Most L1CAM (83%) and integrin (90%) hits were missense — consistent with potential gain-of-function (PMID:27749842).
  • Copy-number landscape is more stable than the SNV landscape: hierarchical clustering of 44 tumors yielded two stable clusters — Cluster A (9p21 CDKN2A/CDKN2B/MTAP deletions in a euploid background) and Cluster B (1q21.1 SETDB1/MLLT11 amplifications, P=0.0002; 6p22.3 E2F3 amplifications, P=0.001; enriched for TP53 mutations, P=0.0001) — with no significant differential enrichment for chemotherapy treatment or metastatic status. Median Hamming distance between intra-patient pairs (HD=0.20) was significantly lower than between inter-patient pairs (HD=0.53; P=3×10⁻⁸, Wilcoxon test) — i.e., per-patient CN landscape is relatively stable across evolution (PMID:27749842).
  • ATM/RB/FANCC chemotherapy-response signature: present in 11/15 (73.3%) pre-chemotherapy tumors vs 11/29 (37.9%) post-chemotherapy tumors (p=0.05), suggesting clones with this signature are eradicated by treatment and superseded by ATM/RB/FANCC wild-type clones that progress to metastatic resistant disease (PMID:27749842).
  • Mutational signatures: chemotherapy-treated tumors were enriched for C>A and C>G substitutions; cisplatin treatment specifically drove a C>A pattern matching the C. elegans cisplatin signature. Four COSMIC-like signatures were identified — APOBEC (Sanger sigs 2/13), age, smoking, and ERCC2-associated. ERCC2 mutations were rare in this chemotherapy-treated cohort, consistent with ERCC2-mutated clones being cisplatin-responsive and selected against during progression (PMID:27749842).
  • APOBEC3A is the dominant cytidine deaminase shaping post-chemotherapy UC: post-chemotherapy tumors showed significant enrichment of APOBEC3A YTCA-context mutations (P=1×10⁻⁵, Fisher’s exact test) and APOBEC3B RTCA-context mutations (P=0.0395), while APOBEC3G CCC-motif mutagenesis decreased. APOBEC clonality also rose post-chemotherapy. APOBEC-induced mutations were enriched in ABC-transporter (OR=2.7, P=0.038) and homologous-recombination DNA-repair (OR=3.8, P=0.033) pathways (PMID:27749842).

Genes & alterations

  • TP53 — recurrent driver mutations; inconsistently shared between matched pre/post-chemo samples (e.g., parallel TP53 p.Y234C / p.G266V hits in patient WCM077). Enriched in CN Cluster B (P=0.0001) (PMID:27749842).
  • PIK3CA, KMT2D, ATM — canonical UC drivers shown to be heterogeneously shared between primary and post-chemotherapy samples (PMID:27749842).
  • TSC1 — among non-truncal driver mutations acquired at clonal divergence nodes in the WCM117 rapid-autopsy phylogeny (PMID:27749842).
  • CDKN2A, CDKN2B, MTAP — 9p21 codeletion defines CN Cluster A; in WCM117, CDKN2A progressed from sub-clonal heterozygous deletion in the primary to clonal homozygous deletion in distant metastases (FISH-confirmed) (PMID:27749842).
  • RB1, FANCC, ATM — combined ATM/RB/FANCC alteration signature was present in 73.3% of pre-chemotherapy tumors but only 37.9% of post-chemotherapy tumors (p=0.05), supporting selective elimination of ATM/RB/FANCC-altered clones by chemotherapy (PMID:27749842).
  • ERCC2 — rare in this cohort, consistent with prior data that ERCC2 mutations mark cisplatin responders and are selected against in chemotherapy-progressing tumors (PMID:27749842).
  • L1CAM — clonally enriched missense mutations (83%) in post-chemotherapy tumors; proposed as a candidate driver of chemotherapy resistance through cell-adhesion-mediated drug resistance (CAM-DR) (PMID:27749842).
  • APOBEC3A, APOBEC3B — APOBEC3A YTCA mutagenesis (P=1×10⁻⁵) and APOBEC3B RTCA mutagenesis (P=0.0395) significantly enriched in post-chemotherapy tumors (PMID:27749842).
  • E2F3, SETDB1, MLLT11 — defining amplifications of CN Cluster B (1q21.1 SETDB1/MLLT11 amp P=0.0002; 6p22.3 E2F3 amp P=0.001) (PMID:27749842).
  • TSPAN8 — non-silent mutation acquired at the primary→metastasis transition (divergence node 5) in patient WCM117; proposed as a candidate metastasis-initiating event (PMID:27749842).
  • RYR2, ANKRD62, NCOA3, LSS — sub-clonal mutations in the WCM117 primary that became enriched in chemotherapy-treated metastatic lesions (early founder signal that survived selection) (PMID:27749842).
  • POLD2, FOXP1, FGFR4, TRRAP, EGFR — present in WCM117 untreated TURBT primary but absent from post-chemotherapy metastases (private to the eradicated clone); POLD2 and FOXP1 are implicated in cellular cisplatin response (PMID:27749842).

Clinical implications

  • Repeat metastatic biopsy is necessary for actionable targeting: because only ~28% of mutations are shared between matched pre- and post-chemotherapy tumors, clinically actionable targets present in metastatic chemotherapy-resistant disease can be missed if treatment decisions rely solely on the diagnostic untreated-primary biopsy (PMID:27749842).
  • L1CAM and integrin signaling are candidate resistance targets. The authors highlight existing therapeutic modalities — anti-L1CAM antibodies (with xenograft efficacy in cholangiocarcinoma, ovarian, and pancreatic ductal carcinoma) and Focal Adhesion Kinase (FAK) inhibitors targeting integrin signaling (in early-phase trials) — as approaches that warrant evaluation in chemotherapy-resistant UC (PMID:27749842).
  • Platinum-based chemotherapy itself acts as a mutagen, increasing C>A burdens (cisplatin signature) and amplifying APOBEC3A/3B-driven mutagenesis, plausibly via increased ssDNA availability during double-strand-break repair of platin–DNA adducts. Understanding this iatrogenic editing is presented as foundational for designing strategies to prevent or reverse the chemotherapy-resistant state (PMID:27749842).
  • ATM/RB/FANCC alteration status is a candidate biomarker of chemotherapy response; clones bearing the signature are selected against during treatment, consistent with their previously reported neoadjuvant-response association (PMID:27749842).
  • Drugs administered in the cohort: cisplatin and gemcitabine (neoadjuvant/first-line); docetaxel + ramucirumab at later progression in the WCM117 case (PMID:27749842).

Limitations & open questions

  • Small sample size — 32 patients (16 matched pre/post sets); statistical power for cohort-level associations is limited and authors call this out explicitly (PMID:27749842).
  • Functional impact of L1CAM and integrin missense mutations is inferred, not validated — authors note functional studies are needed to confirm gain-of-function and CAM-DR mechanism in UC (PMID:27749842).
  • Unfit clones are missing from the evolutionary record; phylogenies necessarily reconstruct only surviving lineages, so the contribution of eradicated clones to early dynamics may be underestimated (PMID:27749842).
  • Genetic drift vs selection cannot be fully disentangled — authors acknowledge time-dependent drift could partly explain divergence, although they argue chemotherapy is too potent a selective pressure to be ignored (PMID:27749842).
  • Multicentricity and malignant seeding remain unresolved — early branching evolution implies these processes need to be addressed in any model of UC oncogenesis beyond the traditional two-pathway grade/stage classification (PMID:27749842).
  • Mechanism linking platinum chemotherapy to APOBEC3A/3B mutagenesis (proposed: increased ssDNA from HR repair of platin adducts) is hypothesized rather than shown (PMID:27749842).
  • One patient (WCM117, 12 samples = 17%) dominates parts of the analysis; authors restricted WCM117 to the case-study figure (Fig. 4) to avoid statistical bias, but generalizability of single-patient observations (e.g., TSPAN8 as a metastasis driver) is limited (PMID:27749842).

Citations from this paper used in the wiki

  • “On average, only 28.4% (range 0.2%–76.4%) of mutations were shared between pre- and post-chemotherapy samples” (Results, Clonal mutational heterogeneity).
  • “We identified this signature in 11/15 (73.3%) in our pre-chemotherapy tumors and 11/29 (37.9%) (p=0.05) in post-chemotherapy tumors” — ATM/RB/FANCC chemotherapy-response signature (Heterogeneity in copy number alterations).
  • “GSEA demonstrated a significant enrichment in mutations mediating L1-cell adhesion molecule (L1CAM) (odds ratio = 1.9, FDR = 0.12), and integrin signaling pathways (odds ratio = 2.8, FDR = 0.02)” (Results, Clonal enrichment of mutations).
  • “We detected a significant enrichment in APOBEC3A-induced mutations (P=0.00001, Fisher’s exact test), and a similar enrichment of APOBEC3B mutagenesis (P=0.0395, Fisher’s exact test) in post-chemotherapy tumors. In contrast, APOBEC3G mutagenesis was substantially decreased” (Results, Mutagenesis mechanisms).
  • “The transition from the primary to the metastatic state was marked by the acquisition of a non-silent mutation in tetraspanin8 (TSPAN8), a well-recognized pro-metastatic and angiogenesis-promoting gene” (Results, WCM117 rapid autopsy case).
  • “We identified a significant difference between intra-patient tumor pairs (median HD=0.20) and inter-patient pairs (median HD=0.53) (P=0.00000003, Wilcoxon test)” (Results, Heterogeneity in copy number alterations).
  • “All BAM files and associated sample information are deposited in dbGap phs001087.v1.p1.” (URLs / Accession codes).

This page was processed by crosslinker on 2026-05-14.