Page 169 - Read Online

Belizario et al. Cancer Drug Resist 2019;2:527-38 I http://dx.doi.org/10.20517/cdr.2018.009 Page 533

are separated from left to right order. The PAM50 scores incorporate most highly enriched clusters of genes
associated with ER signaling (12 genes), growth factor signaling (4 genes), proliferation (21 genes), invasion
(1 gene), basal phenotypes (9 genes) and miscellaneous (3 genes). Cytokeratins KRT5, KRT14 and KRT17,
as well as basal markers SFRP1 and MIA differentiates Basal-like from Luminal B tumors. ERBB2, EGFR,
FGFR4 and GRB7 differentiates HER2-enriched tumors from other subtypes. The PAM50 and Oncotype
DX share the genes MKI67, CCNB1, BIRC5 and MYBL2, which contribute to their proliferation score. The
prognostic value of PAM50 is greater than ER/HER2 immunohistochemistry classification [30,45] . However,
studies have shown that PAM50 score did not improved outcomes of patients who underwent anthracycline-
and taxane-based therapy [30,47] .

Many predictive modeling methodologies for phenotyping breast cancer subtypes have focused on
signaling transduction upon the canonical signaling pathways, including PI3K, mammalian target of
rapamycin, MAPK, transforming growth factor β, Wingless/Integrated, cell cycle, apoptosis, immune
responsiveness, and DNA damage response pathways [3,27,28] . Protein phosphorylation, a reversible post-
translational modification at serine (S), threonine (T) and tyrosine (Y) residues, involves a system of
sequence-specific kinases (writers), phosphatases (erasers) and reader proteins. The advent of reverse RPPA
databases has allowed the phophoproteomic profiling of key cancer-related proteins. The RPPA platform
of the University of Texas MD Anderson Cancer Center (https://bioinformatics.mdanderson.org/main/
MCLP) displays the protein expression in 650 independent cell lines with known genomic, transcriptomic,
and drug-screening data. Analyses in this dataset revealed mutation-induced perturbations of over 200
[48]
proteins and their phosphorylated forms in various cancer types . The representative nodes and modules
of cell network perturbations were used to determine specific clusters for each cancer type. For instance,
the breast cancer cluster was distinct from other cancer types by high expression of proteins of androgen
[48]
hormone receptor canonical pathway and HER2 protein . RPPA-based pathway-activation profiling can
be a powerful tool to predict putative mechanisms underlying sensitivity and resistance to breast tumor
[48]
subtypes for specific therapy . The phospho-HER2 and phospho-EGFR profiles are particularly important
in clinical setting for prescription of EGFR pathway-targeted drugs such as Herceptin, lapatinib and
[40]
pertuzumab .

HER2/ERBB2 is a transmembrane glycoprotein belonging to EGF family of receptors that regulate cell
[40]
growth, proliferation, survival, differentiation, and angiogenesis . Activation of the ERBBs receptors
[40]
by phosphorylation leads to complex signaling pathways . GRB7 is one of the 105 protein coding genes
located in the same amplicon as HER2 on chromosome 17q12. Over-phosphorylation of GRB7 in ERBB2
amplified tumors is involved in resistance to anti-HER2 and antiestrogen therapy. A comprehensive
gene set enrichment analysis revealed perturbed ERBBs signaling in HER2 amplified breast cancer cells
overexpressing dermcidin (DCD), a gene localized at chromosome 12q13 locus and potential oncogene of
[49]
breast epithelial cells . Agreeing with this postulated, we found that DCD is co-expressed with GRB7,
ERBB2 and FGFR4 in various cohorts of breast tumor samples, including METABRIC cohort (data not
shown). More importantly, DCD has been considered as biomarker for cellular resistance of various tumor
cells to the EGFR/ErbB1 tyrosine kinase inhibitors erlotinib and lapatinib .
[50]

Mutations in PIK3CA gene encoding the p110α catalytic subunit of PI3K, class IA, are among the most
[30]
common alterations in human malignancies and contribute to approximately 25% of breast cancers .
Such mutation confers a gain of function to p110α and resistance to HER2-based therapy . A combination
[51]
of three proteins comprising the receptors EGFR, ERBB3/HER3, and the cyclin-dependent kinase
inhibitor p27 (CDKN1B) was found to be a potential biomarker for dependence on PI3K/AKT vs. MAPK/
[52]
ERK signaling for drug resistance in HER2 breast cancers . A computational algorithm called affinity
regression was developed for analysis of distinct dysregulated transcriptional regulators downstream of
[53]
oncogenic somatic alterations . The analyses and validation of the method was done across 12 TCGA