Andrew A. Davis, Amir Behdad, Kayla Viets Layng, Firas Wehbe, Lorenzo Gerratana, Elizabeth Mauer, Alex Barrett, Ami N Shah, Paolo D’Amico, Lisa Flaum, William J. Gradishar, Leonidas C. Platanias, Massimo Cristofanilli
Background: ILC is the second most common type of breast cancer and accounts for approximately 10% of all invasive breast cancers. A hallmark of ILC is the lack of E-cadherin (CDH1) expression, which is frequently used to discriminate between lesions with borderline ductal and lobular histologies. While the genomic landscape of primary ILCs has been well described, less is known about patients (pts) with metastatic-ILC (mILC). Better characterization of the genomic landscape associated with mILC is critical for identifying biomarkers that may provide new insight into ILC tumor biology and ultimately improve long-term outcomes in pts with mILC.
Methods: We retrospectively analyzed de-identified next-generation sequencing (NGS) data from 150 advanced/metastatic pts with ILC and 51 with mixed lobular/ductal histology, defined using the histology of the sequenced biopsy. Diagnoses were abstracted from pathology reports submitted at the time of sequencing. We used the stage documented closest in time to biopsy collection, and samples were excluded if the staging date was unknown or exceeded 180 days after the biopsy date. Our dataset consisted of samples that were molecularly profiled using the Tempus xT solid tumor assay (DNA-seq of 595-648 genes at 500x coverage, full-transcriptome RNA-seq). The mutations identified for this study include somatic single-nucleotide variants and insertions/deletions. Furthermore, we examined the comutational landscape of CDH1-mutant disease and investigated transcript-level expression variation.
Results: Mutations in CDH1 occurred in 65.3% of all mILC samples (98/150). CDH1 expression was similar between CDH1-mutant and WT mILC samples (Wilcoxon rank sum test, p=0.8). The median tumor mutational burden (TMB) score was significantly higher in CDH1-mutant samples (Wilcoxon ranksum test, p=0.010). CDH1-mutant samples were more likely to have a high TMB (≥10 mutations/MB) when compared with the wild-type CDH1 cohort (10% vs. 6.2%), but this difference was not statistically significant (Fisher’s Exact test, p=0.5). Additionally, we observed that the ER+ subtype was more frequent in CDH1-mutant samples, although this difference was not statistically significant (97% vs 88%; Fisher’s exact test, p=0.063). PIK3CA mutations were enriched in CDH1-mutant mILC (Table 1). TBX3 and NCOR1 mutations were also mildly enriched in CDH1-mutant mILC, but these results were not significant when correcting for multiple testing (Table 1). CDH1-mutant mixed histology pts had lower CDH1 expression than WT pts (p<0.001, Wilcoxon rank sum exact test). PIK3CA mutations were enriched in CDH1-mutant mixed histology pts, but this difference was not statistically significant (50% vs. 31%; p=0.3, Fisher’s exact test). Log10 CDH1 expression across all mILC pts was lower than in mixed histology pts (3.01 vs 3.53; p<0.001, Wilcoxon rank-sum test).
Conclusions: Our real-world dataset illustrates that the molecular landscape of CDH1-mutant mILC pts is distinct from CDH1-WT mILC pts. Additionally, mILC differs from mixed histology at a transcriptional level, with lower CDH1 expression regardless of CDH1 mutational status. Our findings suggest a use for CDH1 RNA expression levels in reclassifying mixed histology samples as mILC. Additionally, therapies targeting PIK3CA may be further investigated for their actionability in CDH1-mutant mILC cases.
VIEW THE PRESENTATION
VIEW THE POSTER
VIEW THE PUBLICATION