Mutation pattern analysis reveals polygenic mini-drivers associated with relapse after surgery in lung adenocarcinoma

Laura Bennett,Chris Smowton,Crispin J Miller,Matthew Howell,Danish Memon,Cong Zhou

doi:10.1038/s41598-018-33276-3

Abstract

The genomic lesions found in malignant tumours exhibit a striking degree of heterogeneity. Many tumours lack a known driver mutation, and their genetic basis is unclear. By mapping the somatic mutations identified in primary lung adenocarcinomas onto an independent coexpression network derived from normal tissue, we identify a critical gene network enriched for metastasis-associated genes. While individual genes within this module were rarely mutated, a significant accumulation of mutations within this geneset was predictive of relapse in lung cancer patients that have undergone surgery. Since it is the density of mutations within this module that is informative, rather than the status of any individual gene, these data are in keeping with a ‘mini-driver’ model of tumorigenesis in which multiple mutations, each with a weak effect, combine to form a polygenic driver with sufficient power to significantly alter cell behaviour and ultimately patient outcome. These polygenic mini-drivers therefore provide a means by which heterogeneous mutation patterns can generate the consistent hallmark changes in phenotype observed across tumours.

Highlights

Large-scale cancer sequencing projects have led to a rapid expansion in the catalogue of putative driver mutations and revealed striking genetic heterogeneity across the sets of mutations identified in different patients[1]
This model of the disease is in keeping with findings from large scale Genome Wide Association Studies (GWAS) that have shown, for typical traits, that even the most significant loci explain only a fraction of the predicted genetic variation[9]; the remaining “missing heritability” is provided by the contribution of multiple Single Nucleotide Polymorphisms (SNPs) that each fall below the threshold for genome-wide significance[9]
By building network models representing the expression profiles of normal samples, and mapping the mutation spectra of individual lung cancer patients onto these networks, we were able to place somatic mutations into the context of lung tissue before the system-wide reorganization of gene expression that occurs during oncogenesis

Summary

Introduction

Large-scale cancer sequencing projects have led to a rapid expansion in the catalogue of putative driver mutations and revealed striking genetic heterogeneity across the sets of mutations identified in different patients[1]. As hypothesized by Castro-Giner et al.[7], is that in tumours lacking a clear major driver, oncogenic transformation may instead occur through the collaborative action of a set of ‘mini-drivers’, each with a weak individual effect In this model, these mini-driver mutations accumulate to form a polygenic driver with sufficient power to lead to tumour promoting effects and phenotypic changes[7,8]. Since the background rate is dependent on multiple factors including genomic position, the mutation load in an individual patient, and gene size, these factors are represented as additional covariates in the numerical model[1,2] This tends to raise the bar for long genes and/or those located in regions with higher mutation frequencies. This results in a modular structure in which functionally related genes congregate as sets of densely interconnected nodes within the network[9,16]

Methods

Results

Conclusion