Integrated genomic characterization of endometrial carcinoma

Author

The Cancer Genome Atlas Research Network

Doi

10.1038/nature12113

PMID: 23636398 · DOI: 10.1038/nature12113 · Journal: Nature (2013)

TL;DR

The Cancer Genome Atlas performed an integrated genomic, transcriptomic, and proteomic characterization of 373 endometrial carcinomas (307 endometrioid, 53 serous, 13 mixed histology) and proposed a four-category molecular classification: (1) a POLE ultramutated group (~7%) with hotspot exonuclease-domain POLE mutations and very high mutation rates (232 × 10⁻⁶/Mb); (2) an MSI hypermutated group (~28%) dominated by MLH1 promoter methylation; (3) a copy-number low endometrioid group with frequent CTNNB1 mutations; and (4) a copy-number high serous-like group with TP53 mutations, extensive SCNAs, and molecular similarities to high-grade serous ovarian and basal-like breast carcinomas. The classification refines histology-based subtyping and may direct post-surgical adjuvant therapy.

Cohort & data

  • 373 endometrial carcinomas with matched germline DNA, including 307 endometrioid, 53 serous, and 13 mixed-histology cases; median follow-up 32 months (range 1–195 months); 21% recurred and 11% died at time of analysis. Cohort: ucec_tcga_pub (cancer types: UCEC, UEC, USC).
  • Platforms. SCNAs on 363 samples by Affymetrix SNP 6.0 (affymetrix-snp6) with GISTIC (gistic) recurrent-event calling. Whole-exome sequencing (whole-exome-seq) on 248 tumour/normal pairs (≥20× depth). Low-pass paired-end whole-genome sequencing on 106–107 tumours (mean depth 6×). mRNA expression by RNA-Seq (rna-seq) on 333 tumours, microRNA on 367, RPPA protein expression on 293, and DNA methylation by Illumina Infinium 450K (450k-methylation-array) on 373. MSI testing on all samples using seven repeat loci.
  • Integrated analyses. MEMo (mutual exclusivity), iCluster, PARADIGM pathway clustering, and a new SuperCluster algorithm; consensus clustering for mRNA/miRNA/RPPA/methylation.

Key findings

  • Four integrated subgroups derived from somatic substitutions, MSI, and SCNAs: POLE ultramutated (n=17, ~7%; 232 × 10⁻⁶ mut/Mb), MSI hypermutated (n=65; 18 × 10⁻⁶ mut/Mb), copy-number low endometrioid (n=90; 2.9 × 10⁻⁶ mut/Mb), and copy-number high serous-like (n=60; 2.3 × 10⁻⁶ mut/Mb).
  • MSI burden. MSI detected in 40% of endometrioid tumours and 2% of serous tumours; MSI endometrioid tumours had a mutation frequency ~10× greater than MSS endometrioid tumours.
  • POLE hotspots. Novel hotspot mutations in the POLE exonuclease domain at Pro286Arg and Val411Leu in 13 of 17 (76%) ultramutated samples; significantly better progression-free survival in this group (log-rank P=0.02). No germline POLE Leu424Val or POLD1 Ser478Asn variants found in endometrial normals.
  • SCNA clustering. Unsupervised hierarchical clustering of SCNAs yielded four clusters; cluster 4 (serous-like) was characterised by recurrent focal amplifications of MYC (8q24.12), ERBB2 (17q12), CCNE1 (19q12), and newly reported amplifications containing FGFR3 (4p16.3) and SOX17 (8q11.23), plus frequent TP53 mutations (90%) and low PTEN mutation rate (11%). Cluster 4 had significantly worse progression-free survival than endometrioid clusters (log-rank P=0.003).
  • Subgroup-specific mutation frequencies. Across the four groups, significantly mutated genes (Q<0.05) differed: in POLE ultramutated tumours — PTEN (94%, Q=0), PIK3R1 (65%, Q=8.3×10⁻⁷), PIK3CA (71%, Q=9.1×10⁻⁵), FBXW7 (82%, Q=1.4×10⁻⁴), KRAS (53%, Q=9.2×10⁻⁴), POLE (100%, Q=4.2×10⁻²³). MSS copy-number-low endometrioid tumours had unusually high CTNNB1 mutation frequency (52%). Copy-number high tumours had frequent TP53 (90%), FBXW7 (22%, Q=0) and PPP2R1A (22%, Q=1.7×10⁻¹⁶) mutations.
  • ARID5B, a SWI/SNF-related ARID-family chromatin remodelling gene, was newly identified as recurrently mutated and differentially mutated: 23.1% MSI vs 5.6% MSS endometrioid vs 0% high-SCNA serous tumours.
  • RPL22 frameshift indels near a Lys15 homopolymer were almost exclusively in the MSI group (36.9%).
  • TP53 / PTEN co-occurrence. 50% (10/20) of endometrioid tumours with non-silent TP53 mutations also had non-silent PTEN mutations, vs only 1/39 (2.6%) of serous tumours — suggesting a distinct endometrioid tumorigenic mechanism even when TP53 is mutated.
  • Mutual exclusivity in WNT axis. MEMo identified a strongly mutually exclusive module of CTNNB1, KRAS, and SOX17 in copy-number-low tumours. Novel recurrent SOX17 mutations at Ala96Gly and Ser403Ile (8% in copy-number-low subgroup) — SOX17 mediates β-catenin proteasomal degradation.
  • PI(3)K pathway. 93% of endometrioid tumours harboured PI(3)K/AKT pathway alterations; PIK3CA and PIK3R1 showed mutual exclusivity in all subgroups, but co-occurred with PTEN mutations in the MSI and copy-number-low subgroups (unlike other tumour types).
  • Structural translocations. Low-pass WGS on 106 tumours found recurrent translocations involving BCL family members (BCL2, BCL7A, BCL9, BCL2L11) in 5/106 cases; all predicted to result in in-frame fusions and increased BCL family expression.
  • Cross-tumour similarity. Serous-like endometrial tumours share focal SCNA patterns, MC3 minimal-methylation profile, and TP53 mutation frequency (uterine serous 91%, HGSOC 96%, basal-like breast 84%) with high-grade serous ovarian carcinoma and basal-like breast carcinoma; however, PIK3CA, FBXW7, PPP2R1A, and ARID1A mutations are >30% more frequent in uterine serous than in basal-like breast or HGSOC.
  • Histologic reclassification. A single serous case lacking TP53 mutation and extensive SCNAs but harbouring a KRAS mutation and high mutation rate was molecularly consistent with grade-3 endometrioid on pathology re-review.

Genes & alterations

  • POLE — exonuclease-domain hotspots Pro286Arg and Val411Leu define the ultramutated subgroup (100% mutated; Q=4.2×10⁻²³); associated with C→A transversion-rich spectrum and improved PFS.
  • TP53 — mutated in ~90% of copy-number-high (serous-like) tumours; only 11.4% of endometrioid; truncations (not missense) implicated by PARADIGM as the loss-of-function class.
  • PTEN — mutated in 84% of MSS endometrioid tumours and 94% of POLE ultramutated; only 11% in serous-like; co-occurs with TP53 in endometrioid (50%) but rarely in serous (2.6%).
  • PIK3CA — frequent across endometrioid subgroups (71% in POLE-ultramutated; 42% in uterine serous); mutually exclusive with PIK3R1.
  • PIK3R1 — 65% in POLE ultramutated; mutually exclusive with PIK3CA.
  • KRAS — 53% in POLE ultramutated; activating mutations also implicated in alternative WNT/β-catenin activation; mutually exclusive with CTNNB1 and SOX17.
  • CTNNB1 — 52% in copy-number-low MSS endometrioid; defines this subgroup; mutually exclusive with KRAS and SOX17.
  • ARID1A — high mutation rate in non-serous groups; decreased protein expression by RPPA consistent with inactivating mutations; differential frequencies across subgroups.
  • ARID5B — newly identified recurrently mutated SWI/SNF/ARID-family gene; 23.1% MSI vs 5.6% MSS endometrioid vs 0% serous.
  • FBXW7 — 82% in POLE ultramutated; 22% in copy-number-high; mutually exclusive with FGFR2/ERBB2 amplification in WNT module analysis.
  • PPP2R1A — 22% in copy-number-high serous-like (Q=1.7×10⁻¹⁶); previously noted as serous-specific.
  • MLH1 — promoter hypermethylation in most MSI tumours, with decreased MLH1 mRNA expression; drives MSI/CIMP (MC1) phenotype.
  • RPL22 — Lys15 homopolymer frameshift indels in 36.9% of MSI tumours; nearly absent in other subgroups.
  • SOX17 — recurrent novel missense Ala96Gly and Ser403Ile in 8% of copy-number-low tumours; mediates β-catenin degradation.
  • ERBB2 — focally amplified with protein overexpression in 25% of serous/serous-like tumours; amplification reported in 27% of uterine serous carcinomas — basis for HER2-targeted therapy hypothesis.
  • MYC — focally amplified at 8q24.12 in cluster-4 (serous-like).
  • CCNE1 — focally amplified at 19q12 in cluster-4.
  • FGFR1, FGFR3 — amplified in cluster 4; FGFR3 amplification at 4p16.3 newly reported in endometrial cancer.
  • FGFR2 — included in WNT/RTK MEMo module of mutual exclusivity with FBXW7 and ERBB2.
  • IGF1R — within cluster-2 15q26.2 focal amplification; potential therapeutic target.
  • LRP1B — deletion in cluster-4; recently associated with liposomal doxorubicin resistance in serous ovarian cancer.
  • CDKN2A — increased expression in copy-number-high cluster; cited as feature distinguishing serous from endometrioid carcinoma.
  • AURKA, CCNB1, ASNS — high RPPA expression in POLE-ultramutated/cell-cycle deregulated subgroups.
  • PGR — increased in copy-number-low cluster; consistent with hormonal-therapy responsiveness.
  • FOXA1 — low oestrogen receptor/FOXA1 signalling in PARADIGM cluster 3 (copy-number high).
  • RAD50 — raised RPPA expression in copy-number-low subgroup; suggests DNA repair capacity differences.
  • AKT3, CTCF — recurrently altered across subtypes (see Fig. 2d/Fig. 5b mutation panels).
  • POLD1 — checked for germline susceptibility (Ser478Asn) — not found in this endometrial cohort.
  • BCL2, BCL7A, BCL9, BCL2L11 — recurrent translocations involving these BCL-family members in 5/106 tumours by low-pass WGS; predicted in-frame fusions with increased BCL family expression.

Clinical implications

  • Treatment selection by molecular subtype. The authors argue ~25% of high-grade endometrioid tumours have a copy-number-high serous-like molecular phenotype and may benefit from chemotherapy (e.g., doxorubicin, cisplatin, paclitaxel) rather than the adjuvant radiotherapy commonly used for endometrioid carcinomas.
  • HER2-targeted therapy hypothesis. Frequent ERBB2 amplification with overexpression in 25% of serous/serous-like tumours supports prospective trials of HER2-targeted inhibitors; the authors note a prior small trastuzumab trial accrued few FISH-amplified serous cases.
  • PI(3)K/AKT pathway inhibition. With 93% of endometrioid tumours harbouring PI(3)K/AKT pathway alterations, this pathway is highlighted as the dominant targetable axis across endometrioid subgroups.
  • POLE ultramutated good prognosis. POLE-mutant tumours have significantly improved PFS — relevant for de-escalation considerations.
  • Hormone therapy. Increased PGR expression in the copy-number-low cluster suggests hormonal-therapy responsiveness in this subgroup.
  • Histology-discordant cases. Molecular profiling can reclassify histologically ambiguous high-grade endometrial tumours (case example reclassified from serous to grade-3 endometrioid based on lack of TP53 mutation, KRAS mutation, and high mutation rate).

Limitations & open questions

  • The POLE subgroup was excluded from pathway integration analyses because of high background mutation rate and small sample size (n=17), limiting subtype-level pathway inference.
  • Translational chemotherapy/HER2-targeted recommendations require prospective trials; the prior trastuzumab trial cited was small and underpowered.
  • The cohort is cross-sectional — temporal evolution, treatment response, and resistance trajectories are not addressed.
  • Functional consequences of novel SOX17 missense mutations (Ala96Gly, Ser403Ile) and ARID5B alterations remain to be characterised experimentally.
  • Distinction between truncating and missense TP53 mutations as functionally distinct classes (PARADIGM observation) is intriguing but not biochemically validated here.
  • Whether the four molecular subgroups generalise to other endometrial cohorts (especially non-TCGA populations and rare histologies) requires external validation.

Citations from this paper used in the wiki

  • “We performed an integrated genomic, transcriptomic and proteomic characterization of 373 endometrial carcinomas” — defines cohort scope (abstract).
  • “POLE ultramutated, microsatellite instability hypermutated, copy-number low, and copy-number high” — establishes the four-subtype classification (abstract).
  • “newly identified hotspot mutations in POLE at Pro286Arg and Val411Leu present in 13 (76%) of the 17 ultramutated samples” — anchors POLE hotspot evidence.
  • “The MSI endometrioid tumours had a mutation frequency approximately tenfold greater than MSS endometrioid tumours” — quantifies hypermutator phenotype.
  • “Cluster 4 tumours were characterized by significantly recurrent previously reported focal amplifications of the oncogenes MYC (8q24.12), ERBB2 (17q12) and CCNE1 (19q12)” — anchors copy-number-high amplifications.
  • “93% of endometrioid tumours had mutations that suggested potential for targeted therapy with PI(3)K/AKT pathway inhibitors” — anchors therapeutic rationale.
  • “Clinicians should carefully consider treating copy-number-altered endometrioid patients with chemotherapy rather than adjuvant radiotherapy” — anchors the treatment-paradigm recommendation.

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