Reparameterization of PAM50 expression identifies novel breast tumor dimensions and leads to discovery of a breast cancer susceptibility locus at 12q15

Abstract

Conflict of Interest StatementPhil Bernard is an inventor of the PAM50 gene expression signature and a stakeholder in BioClassifier LLC, a company that licensed the PAM50 know-how to Nanostring Inc for the commercialization of Prosigna.peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/133397 doi: bioRxiv preprint first posted online May. 9, 2017; 2 AbstractIt is well-known that breast tumors exhibit different expression patterns that can be used to assign intrinsic subtypes -the PAM50 assay, for example, categorizes tumors into: Luminal A, Luminal B, HER2-enriched and Basal-like -yet tumors are often more complex than categorization can describe. We used 911 sporadic breast tumors to reparameterize expression from the PAM50 genes to five orthogonal tumor dimensions using principal components (PC). Three dimensions captured intrinsic subtype, two dimensions were novel, and all replicated in 945 TCGA tumors. By definition dimensions are independent, an important attribute for inclusion in downstream studies exploring effects of tumor diversity. One application where tumor subtyping has failed to provide impact is susceptibility genetics. Germline genetic heterogeneity reduces power for gene-finding. The identification of heritable tumor characteristics has potential to increase homogeneity. We compared 238 breast tumors from high-risk pedigrees not attributable to BRCA1 or BRCA2 to 911 sporadic breast tumors. Two PC dimensions were significantly enriched in the pedigrees (intrinsic subtypes were not). We performed proof-of-concept gene-mapping in one enriched pedigree and identified a 0.5 Mb genomewide significant region at 12q15 that segregated to the 8 breast cancer cases with the most extreme PC tumors through 32 meioses (p=2.6×10 -8 ).In conclusion, our study: suggests a new approach to describe tumor diversity; supports the hypothesis that tumor characteristics are heritable providing new avenues for germline studies; and proposes a new breast cancer locus. Reparameterization of expression patterns may similarly inform other studies attempting to model the effects of tumor heterogeneity.

Shared genomic segments in high‐risk multigenerational pedigrees with gastroschisis

“…Shared genomic segment (SGS) analysis is a novel approach recently developed to harness the power of high‐risk multigenerational pedigrees to identify regions harboring disease susceptibility loci (Coon et al, ; Knight et al, ; Madsen et al, ). A single pedigree can have sufficient power to identify shared regions that may harbor susceptibility genes (Knight et al, ).…”

“…Furthermore, multigenerational pedigrees are too large for exact linkage methods, and require sampling techniques, such as Monte Carlo Markov chain methods. Extension of the SGS approach to optimize over subsets to improve robustness to genetic heterogeneity was recently introduced, and used to identify a genome‐wide significant regions in high‐risk pedigrees with multiple myeloma (Waller et al, ), breast cancer (Madsen et al, ), and suicide (Coon et al, ).…”

Re-interpretation of PAM50 gene expression as quantitative tumor dimensions shows utility for clinical trials: application to prognosis and response to paclitaxel in breast cancer

“…Five breast tumor dimensions, PC1–PC5, were previously defined [ 1 ] using PAM50 gene expression data from the LACE [ 18 ] and Pathways [ 19 ] (LACE/PW) population-based breast cancer cohorts. Briefly, a principal components analysis was performed on the PAM50 gene expression matrix from LACE/PW samples.…”

“…The pre-analytical (RNA extraction from FFPE punches), analytical (RT-qPCR on the LC480), and post-analytical (reference control and housekeeper normalization) methodologies were identical. We used gene coefficients previously derived from the LACE/PW data [ 1 ] (Table S1) to determine the tumor dimensions PC1–PC5 in the GEICAM/9906 samples.…”

“…We previously derived five tumor dimensions for breast cancer as a new representation of tumor diversity using the PAM50 gene set. Quantitative dimensions were superior to categorical intrinsic subtypes for identifying tumor characteristics that clustered in high-risk pedigrees and identified a novel genome-wide significant breast cancer susceptibility locus [ 1 ]. The increased power to differentiate tumors from high-risk pedigrees indicates the potential to distinguish important genetic diversity, providing new opportunities for precision genomics.…”