Gain-of-Function Genetic Screens Using Barcoded Libraries of Human Open Reading Frames Identify Regulators of Proliferation and Cancer Drivers

Abstract

The identification of genetic events that cause tumorigenesis is a key goal of cancer research. While recent sequencing efforts have provided unprecedented illumination of cancer genomes, the inherent genomic instability of tumors results in vast numbers of somatic mutations and copy number alterations (CNAs), hampering identification of causative driver events. Functional genetic screens provide a complementary approach to genomics in the search for cancer drivers. While many studies have focused on loss-of-function studies in cancer cells using RNAi technology, we present a platform for gain-of-function screening using sequence-verified human open reading frames (ORFs). We have paired genome-scale ORF collections with unique DNA barcodes of uniform length, facilitating quantitative readout by next generation sequencing. ORFs are expressed from an inducible promoter, allowing precise control of screening conditions. We have used our libraries to identify regulators of proliferation in non-transformed human mammary epithelial cells (HMECs), seeking to uncover early driving events in tumorigenesis. We rediscovered many known oncogenes and observe significant enrichment of our pro-proliferative gene set within regions of recurrent focal amplification in human cancers, indicating the efficacy of our method for cancer gene driver discovery. We have identified a novel family of pro-proliferative genes, the keratin associated proteins, which strongly promote proliferation in HMECs and may contribute to human cancers. Additionally, we report several common sources of error that contribute to noise in pooled genetic screens in mammalian cells. We have observed that library plasmid DNA present in viral supernatants can contaminate screen samples resulting in inaccurate reference measurements of the abundance of library elements. A further artifact of this contamination is a perceived bias towards enrichment of library elements with low GC content. Additionally, libraries containing multiple unique elements carried on a single retroviral genome are subject to recombination during reverse transcription. Our inducible ORF libraries circumvent these problems, allowing for highly accurate, reproducible screen results. As CNAs are considered to be key initiating events in tumorigenesis, and many oncogenes are activated by amplification, we believe that modeling increased gene expression in untransformed human cells will provide critical functional data facilitating the discovery of bona fide cancer drivers.