Boolean network modeling for systematic identification of anticancer drug resistance in colorectal cancer

Abstract

Targeted inhibition of aberrantly activated signaling pathways has proven effective for treatment of many solid tumors such as non-small cell lung cancer (NSCLC), colorectal cancer (CRC) and breast cancer. Unfortunately cancer cells adapt to such monotherapies and eventually develop clinical resistance [1][2]. The drug resistance originates from genetic or epigenetic alterations within cancer cells that enable activation of other growth and survival pathways [3]. Systematic identification and mechanism of drug resistance pathways is desirable for enhancing the efficacy and durability of anticancer drugs. Additionally a comprehensive understanding of drug effecting and resisting pathways can help in the design of multicomponent combination therapies. In this work we employ Boolean network modeling of signaling pathways to systematically analyze drug resistance in colorectal cancer. Colorectal cancer progresses from adenoma to invasive CRC via a number of genetic alterations including those in APC, KRAS, SMAD4 and BRAF. Crosstalk among components of MAPK, Wnt, TGF-β and NF-kβ pathways play critical role in CRCs evading detection and sustaining tumor phenotype. We first derive a unified Boolean signaling network of all major KEGG signaling pathways. Network reduction is performed retaining all hub genes and other seed genes implicated for colorectal cancer in literature. All uncontrollable and unobservable nodes and edges are removed from the network. In addition other unmarked (non-hub/seed) nodes with single input or single output are eliminated from analysis. The unified and compressed Boolean signaling network is then enriched with CRC RPPA protein expression data from The Cancer Genome Atlas (TCGA) to mark active nodes and edges in CRC tumors. Clustering is employed to identify different molecular subtypes in colorectal cancer based on mRNA gene expression data. The enrichment of compressed Boolean network is carried out for each tumor class separately. Drug resistance to anticancer therapies manifests itself in three possible ways. The originally deregulated pathway inhibited by therapy can be reactivated via mutations in the inhibited target protein or via a bypass activation mechanism due to genomic and epigenetic alterations. Drug resistance is also seen when cancer cells dedifferentiate due to activation of independent pathways such as Notch. A Boolean simulation engine and line justification algorithm work together to analyze possible source of drug resistance in CRC. For each tumor class we record the status of deregulated pathways and then simulate the effect of mono-therapy (such as PI3K inhibitor) to understand the therapeutic effects on transcription factors and pathway nodes. Thereafter we justify the previously recorded status of originally deregulated pathways in the presence of mono-therapy using a line justification algorithm. The line justification algorithm treats the original effects as stuck-at faults and tries to justify it at the receptors in the presence of a mono-inhibitor. Our research shows that KRAS-mutant CRCs develop resistance to PI3K inhibitors due to increased expression of EGFR. They also show BRAF-mutant CRCs developing resistance to RAF inhibitors via deregulation of NF-kB pathway.