Comprehensive molecular portraits of invasive lobular breast cancer

Authors

Giovanni Ciriello

Michael L. Gatza

Andrew H. Beck

Matthew D. Wilkerson

Suhn K. Rhie

Alessandro Pastore

Hailei Zhang

Michael McLellan

Christina Yau

Cyriac Kandoth

Reanne Bowlby

Hui Shen

Sikander Hayat

Robert Fieldhouse

Susan C. Lester

Gary M.K. Tse

Rachel E. Factor

Laura C. Collins

Kimberly H. Allison

Yunn-Yi Chen

Kristen Jensen

Nicole B. Johnson

Steffi Oesterreich

Gordon B. Mills

Andrew D. Cherniack

Gordon Robertson

Christopher Benz

Chris Sander

Peter W. Laird

Katherine A. Hoadley

Tari A. King

TCGA Research Network

Charles M. Perou

Doi

10.1016/j.cell.2015.09.033

PMID: 26451490 · DOI: 10.1016/j.cell.2015.09.033 · Journal: Cell (2015)

TL;DR

Ciriello and colleagues report the TCGA Network’s comprehensive multi-platform molecular characterization of 817 breast tumors — 127 invasive lobular carcinoma (ILC), 490 invasive ductal carcinoma (IDC), 88 mixed IDC/ILC, and 112 other histologies — profiling DNA, RNA, miRNA, methylation, copy number, and (for 633 samples) reverse-phase protein arrays. Beyond confirming E-cadherin (CDH1) loss as the ILC hallmark, the study identifies PTEN, TBX3, and FOXA1 mutations as ILC-enriched features, reports the highest pAKT levels of any breast subtype in ILC, defines three ILC transcriptional subtypes (reactive-like, immune-related, proliferative) with survival differences, and shows that mixed IDC/ILC tumors molecularly resolve into ILC-like or IDC-like classes rather than a third hybrid entity.

Cohort & data

Samples: 817 primary breast tumors from TCGA — 127 ILC, 490 IDC, 88 mixed IDC/ILC (MIXED), and 112 other histologies, with intrinsic BRCA PAM50 subtyping.
Dataset: brca_tcga_pub2015. Validation in the METABRIC cohort (brca_metabric, Curtis et al. 2012).
Platforms: whole-exome-seq, rna-seq, mirna-seq, affymetrix-snp6 for copy number, hm27-methylation-array and hm450-methylation-array, and rppa (n=633). Whole-genome bisulfite sequencing was run on 5 samples to validate methylation calls.
ILC clinical composition: Predominantly Luminal A (LumA = 83% of ILC vs 41% of IDC); 94% (113/120) ER+ by IHC (PMID:26451490).

Key findings

ILC-enriched mutations vs IDC (all-comers): CDH1 63% vs 2% (q=3.94E-53, mostly truncating); TBX3 9% vs 2% (q=0.003); RUNX1 10% vs 3% (q=0.008); PIK3CA 48% vs 33% (q=0.02); FOXA1 7% vs 2% (q=0.08). Conversely, ILC was depleted for TP53 mutations (8% vs 44%, q=1.9E-14), focal MYC amplification (6% vs 27%, q=7.42E-7), and CCNE1 amplification (0% vs 7%, q=0.01) (PMID:26451490).
LumA-restricted comparison (n=106 ILC vs n=201 IDC): CDH1 (q=1.4E-30), TBX3 (q=0.05), and FOXA1 (q=0.065) remained ILC-enriched; GATA3 mutations were IDC-enriched (5% ILC vs 20% IDC, q=0.003). Combined PTEN inactivating alterations (homozygous loss + mutation) reached 14% in LumA ILC vs 3% in LumA IDC (p=9E-4) (PMID:26451490).
E-cadherin loss is genetic, not epigenetic: 108 CDH1 mutations in 107/817 patients (80 in ILC); 83% predicted truncating. Mutations co-occurred with heterozygous 16q loss in 89% of ILC. Combining DNA/RNA/protein, CDH1 alterations were detected in 120/127 (95%) ILC cases and in all 79 cases with full data. CDH1 promoter DNA hypermethylation was NOT observed in any tumor, contradicting prior MSP-based reports (PMID:26451490).
FOXA1 mutation structural hotspot (MSH): 33 FOXA1 mutations in 30/817 tumors (3.7%); 11/127 (7%) ILC. All ILC FOXA1 mutations fell in the fork-head (FK) domain, clustered in the W2 wing loop with recurrent I176 (n=4) and D226 (n=3) residues that are spatially close (5–10 Å) to W2 in 3D space. 22/25 FK mutations land in this MSH; 8/127 ILC vs 4/490 IDC (p=6E-4). FOXA1 mutations were positively associated with FOXA1 mRNA expression (p=0.002) and with low methylation at FOXA1 binding sites, suggesting gain-of-function rather than loss (PMID:26451490).
GATA3/FOXA1 differential ER modulation: GATA3 mutations were mutually exclusive with FOXA1 events; LumA ILC had lower GATA3 mRNA (p=0.007) and protein (p=2E-4) than LumA IDC. ER total protein (p=0.005) and phospho-ER (p=2E-5) were also lower in LumA ILC (PMID:26451490).
AKT pathway is the most activated of any breast subtype in ILC: Lower PTEN protein (p=4E-4), increased pAKT-S473 (p=0.004) and pAKT-T308 (p=7E-5), higher total EGFR (p=1E-4) and pEGFR Y1068 (p=0.005)/Y1173 (p=0.007), pSTAT3 Y705 (p=7E-4), pp27-T157 (p=0.002), p70S6K-T389 (p=1E-4). Phospho-AKT levels in LumA ILC matched those in HER2+ and Basal-like IDC. Despite high PIK3CA mutation frequency, PIK3CA mutations were NOT associated with elevated pAKT in this dataset; instead, upstream events (including ERBB2 amplifications and mutations) converged on AKT/mTOR signaling in 45% of samples via MEMo analysis, with 40% of ILC showing upstream pathway alterations (PMID:26451490).
Three ILC transcriptional subtypes (LumA ILC, n=106; validated in METABRIC): Reactive-like (high stromal/epithelial signaling — keratins, kallikreins, EGFR, MET, PDGFRA, KIT; low tumor purity); immune-related (ILs, chemokines, MHC, TNFs, IDO1, IFNG; macrophage-dominated by CD68/MacCSF/MacTH1/TCR signatures); proliferative (high cell-cycle and DNA-repair proteins — Cyclin E1, FoxM1, PCNA, pChk1-S345, Rad50, Rad51, XRCC1, BRCA2). 1,277 SAM-differentiating genes (q=0). Reactive-like ILC patients had better DSS (p=0.038, HR=0.47) and OS (p=0.023, HR=0.50) than proliferative ILC in METABRIC; proliferative-score-high ILC had worse DSS (p=0.025, HR=2.0) (PMID:26451490).
Mixed IDC/ILC tumors are NOT a third entity: Three orthogonal classifiers (ISOpure, OncoSign-adapted, ElasticNet) agreed that 24/88 mixed cases are ILC-like and 64 are IDC-like by majority vote. All CDH1-mutated mixed cases were classified ILC-like; CDH1 status was the dominant feature (PMID:26451490).

Genes & alterations

CDH1 — biallelic loss (truncating mutation + 16q loss) in 95% of ILC; the defining lesion. No promoter hypermethylation detected.
PTEN — homozygous deletion (6%) and somatic mutation (7%) in ILC, mutually exclusive with PIK3CA; combined PTEN inactivation 14% in LumA ILC vs 3% LumA IDC (p=9E-4).
PIK3CA — 48% mutation rate in ILC (vs 33% IDC); not associated with pAKT levels in this dataset.
FOXA1 — recurrent fork-head W2-region mutations (I176, D226 spatial hotspot); ILC-enriched (7%); associated with increased FOXA1 mRNA expression, suggesting activating rather than loss-of-function.
GATA3 — IDC-enriched mutations (20% LumA IDC vs 5% LumA ILC); mutually exclusive with FOXA1; LumA ILC has lower GATA3 mRNA and protein.
TBX3 — ILC-enriched mutations (9% ILC vs 2% IDC).
RUNX1 — ILC-enriched mutations (10% ILC vs 3% IDC, all-comers).
TP53 — IDC-enriched (8% ILC vs 44% IDC), reflecting underrepresentation of Basal-like in ILC.
MYC, CCNE1 — focal amplifications IDC-enriched.
ERBB2 — amplification and mutations identified among upstream drivers of AKT activation in ILC; previously reported in relapsed ILC (Ross et al., 2013).
EGFR — overexpressed in ILC at total and phospho levels (Y1068, Y1173); high in reactive-like ILC subtype.
ESR1 — ILC and IDC differentially modulate ER activity via FOXA1 vs GATA3; LumA ILC has lower total and phospho-ER.
EP300 — hypothesized to acetylate FOXA1 K264/K267/K270 in W2; FOXA1 W2 mutations may alter EP300-dependent acetylation (mechanistic hypothesis, not directly tested).
BRCA2, MAPK3, RB1, JAK2, TYK2, MET, KIT, PDGFRA, IDO1, IFNG — markers of the three ILC mRNA subtypes (proliferative; reactive-like; immune-related).

Clinical implications

PI3K/AKT/mTOR inhibition as an ILC strategy: Because ILC has the highest average pAKT activity among breast cancer subtypes — comparable to HER2+ and Basal-like — selective PI3K/AKT pathway inhibition is proposed as a particularly attractive therapeutic avenue for this otherwise-Luminal A disease (PMID:26451490).
Endocrine therapy choice: Lower GATA3 and reduced total/phospho-ER in LumA ILC are consistent with prior reports of improved letrozole vs tamoxifen response in ILC (Metzger et al., 2012; Sikora et al., 2014); authors note this warrants further investigation but do not claim a biomarker.
Prognostic stratification within ILC: The three mRNA subtypes (reactive-like, immune-related, proliferative) carry survival differences (reactive-like better DSS/OS; proliferative worse), suggesting future trial stratification potential.
Mixed-histology classification: Molecular ILC-like vs IDC-like calls (driven primarily by CDH1 status) may inform metastatic surveillance — ILC characteristically metastasizes to GI tract, peritoneum, and resists detection on PET — and treatment selection for the ~3–11% of breast cancers with mixed histology.
Diagnostic workup for ILC: Authors recommend anatomical scanning (CT) rather than PET for ILC given low metabolic activity, and note limited response to primary chemotherapy for classic-type ILC.

Limitations & open questions

Short TCGA follow-up: TCGA median follow-up was <2 years vs METABRIC’s 7.2 years; survival analyses primarily rely on METABRIC for adequate event counts.
Functional validation absent: FOXA1 W2 hotspot mutations are hypothesized to act as gain-of-function via altered EP300 acetylation, but no in vitro or in vivo experiments test this mechanism here.
PIK3CA–pAKT discordance: The lack of association between PIK3CA mutation and pAKT in this dataset is unexplained; authors invoke E-cadherin loss as a permissive context and upstream RTK alterations but acknowledge the mechanism is incomplete.
Methylation discrepancy: Contradicts older MSP-based reports of CDH1 promoter hypermethylation; authors attribute to assay sensitivity differences but call for further investigation.
mRNA subtypes need validation: Three ILC subtypes were derived from n=106 LumA ILC and replicated in METABRIC, but authors flag that additional independent validation is required.
Mixed tumors lack uniform pathology criteria: The 11% mixed category aggregates “mixed ductal/lobular” and “IDC with lobular features” — a non-standardized clinical designation.
No actionable drug claims: The paper proposes PI3K/AKT inhibition as a hypothesis-generating direction rather than reporting clinical-trial outcomes; no specific drug is named for ILC.

Citations from this paper used in the wiki

“we comprehensively profiled 817 breast tumors, including 127 ILC, 490 ductal (IDC), and 88 mixed IDC/ILC” — Summary
“ILC cases were significantly enriched for CDH1 mutations, (63% in ILC vs. 2% in IDC, q=3.94E-53), most of them truncating, and mutations affecting TBX3 (9% vs. 2%, q=0.003), RUNX1 (10% vs. 3%, q=0.008), PIK3CA (48% vs. 33%, q=0.02) and FOXA1 (7% vs. 2%, q=0.08)” — Results, Genetic determinants
“Collectively, PTEN inactivating alterations were identified in 14% of LumA ILC versus 3% of LumA IDC (p=9E-4)” — Results
“22 out of 25 FK-mutations in our dataset fall into a restricted 3D space or ‘mutation structural hotspot’ (MSH) … 8/127 ILC cases have FOXA1 mutations within this MSH compared to 4/490 IDC cases (p=6E-4)” — FOXA1 mutations in breast cancer
“ILC tumors also showed significantly increased Akt phosphorylation at both S473 (p=0.004) and T308 (p=7E-5)” — Akt signaling
“reactive-like ILC patients had a significantly better disease-specific (DSS) (p=0.038, HR: 0.47) and overall survival (OS) (p=0.023, HR: 0.50) compared to proliferative ILC patients in the METABRIC dataset” — ILC mRNA subtypes
“24/88 cases (18/57 LumA cases) being called ILC-like by at least two approaches, and 64 being called IDC-like” — Tumors with mixed ILC and IDC histology
“ILC has on average the highest levels of Akt activation, measured by phospho-Akt and PI3K/Akt signaling among all breast cancer subtypes (comparable to IDC Basal-like), making selective inhibition of this pathway in ILC a particularly attractive strategy” — Discussion

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