Functional annotation of rare gene aberration drivers of pancreatic cancer

Abstract

As we enter the era of precision medicine, characterization of cancer genomes will directly influence therapeutic decisions in the clinic. Here we describe a platform enabling functionalization of rare gene mutations through their high-throughput construction, molecular barcoding and delivery to cancer models for in vivo tumour driver screens. We apply these technologies to identify oncogenic drivers of pancreatic ductal adenocarcinoma (PDAC). This approach reveals oncogenic activity for rare gene aberrations in genes including NAD Kinase (NADK), which regulates NADP(H) homeostasis and cellular redox state. We further validate mutant NADK, whose expression provides gain-of-function enzymatic activity leading to a reduction in cellular reactive oxygen species and tumorigenesis, and show that depletion of wild-type NADK in PDAC cell lines attenuates cancer cell growth in vitro and in vivo. These data indicate that annotating rare aberrations can reveal important cancer signalling pathways representing additional therapeutic targets.

Introduction

Large-scale efforts such as The Cancer Genome Atlas and International Cancer Genome Consortium (ICGC) are generating a compendium of genomic aberrations across major cancer types. Such studies are revealing the complexity of cancer genomes, which contain pathogenic driver aberrations and many more biologically neutral passenger events. While most cancers acquire one or more well-studied, high-frequency drivers, much less is known about which and how the abundant low-frequency variants contribute to tumour progression. Experimental assessment of low-frequency aberrations, in both under-characterized genes or among the long ‘tail’ mutations observed in known cancer genes, is difficult, given their large number and the fact that they may either directly drive tumour progression or indirectly influence tumour behaviour through modifying activities of other driver aberrations. Moreover, distinguishing driver events from passengers is further complicated by the fact that driver activity is shaped by the biological context of a given cancer, including its tissue type, microenvironment and other host determinants. Identifying rare driver aberrations therefore requires robust pipelines that allow context-specific functional prioritization of the thousands of potential targets emerging from Next Generation Sequencing (NGS) data.

Here we describe a gain-of-function (GOF) screening platform enabled by a High-Throughput Mutagenesis and Molecular Barcoding (HiTMMoB) method allowing cost-effective, rapid modelling and expression of molecularly barcoded mutant gene variants into open reading frame (ORF) expression clones for pooled functional screens. HiTMMoB permits annotation of mutation-activated oncogenes, whose identification is particularly desirable, given the efficacy of antibody and small molecule inhibitor therapies that may be tailored towards such proteins. Our approach complements gene depletion and knockout strategies by RNA interference (RNAi) and CRISPR/Cas9-based platforms, respectively, which have been tremendously successful for identifying and validating tumour suppressor genes and other genetic dependencies^1,2. HiTMMoB builds on previous over-expression screen-based studies designed to identify wild-type ORFs whose expression promotes tumorigenesis^3,4, drug resistance⁵ and other cancer cell phenotypes such as anchorage-independent growth^6,7,8 among others. Such screens are made possible through efforts by the ORFeome collaboration (http://www.orfeomecollaboration.org/; refs 9, 10, 11) and others¹² who have systematically cloned and sequenced-validated a large majority of expressed human transcripts.

In this study we use HiTMMoB to identify low-frequency mutational events in pancreatic ductal adenocarcinoma (PDAC) that serve as drivers or effectors for recurrent drivers such as KRAS, a proto-oncogene activated in >90% of PDAC cases¹³. We employed HiTMMoB to build PDAC aberration libraries, whose compositions were based on individual PDAC patient genomes and computational selection, to identify rare oncogenic events that, when expressed in a human pancreatic cell model and implanted into mice, promotes in vivo tumour formation. This approach revealed potent activity for aberrations in genes including NAD Kinase (NADK) that regulates NADP(H) homeostasis and cellular redox state. We show that mutant NADK provides GOF enzymatic activity, leading to a reduction in cellular reactive oxygen species (ROS) and tumorigenesis. In addition, we show that depletion of wild-type NADK in PDAC cell lines attenuated cancer cell growth in vitro and in vivo. We conclude that barcode (BC)-assisted, GOF screening has the potential to have an impact on the field of cancer target discovery as a whole, particularly with respect to identifying low-frequency somatic events by complementing structural characterizations of cancer genomes with functional annotation of gene aberrations identified from tumour sequencing programmes.

Results

HiTMMoB method

Successful translation of NGS data requires knowledge on which DNA aberrations represent actionable events, either for development or re-positioning of approved agents to target their activated pathways. To circumvent technical bottlenecks related to scaled construction of aberration expression clones, we developed a cost-effective HiTMMoB method (Fig. 1a) that employs: (1) two common PCR primers (P1/P2) flanking the ORF gene target and a third primer containing the desired mutation (PM), (2) a three-stage PCR reaction, whose design incorporates differential P1/P2 and PM melting temperatures⁵⁴ using ɛ (solute)=2.0; ɛ (solvent)=78.0; probe radius=1.4 Å; [NaCl]=0.15 M and T=310 K. Amber 99SB partial atomic charges and Amber atomic radii were used for the electrostatic calculations. We used the Pymol graphics package for all illustrations.

NADK enzymatic assay

GST-NADK WT and GST-NADK^I90F were cloned into pGEX-6 P-1 and expressed in One Shot BL21 (DE3; Life Technologies) for purification using Glutathione Agarose (Thermo Scientific) following the standard protocol. NADK and NADK^I90F activity was measured spectrophotometrically with a coupled assay that reduces NADP generated by NADK into detectable NADPH using the enzyme glucose-6-phosphate dehydrogenase⁵⁵. Data were analysed using GraphPad Prism 6 by means of Michaelis–Menten kinetics⁵⁶.

NADK knockdown studies

NADK knockdown studies were performed using pLKO-shGFP (NT) or each of the following shRNAs available from the RNAi Consortium (TRC; Broad Institute): TRCN0000037702 (shNADK#4), TRCN0000199808 (shNADK#8), TRCN0000199040 (shNADK#10). shRNA target sequences are as follows: shGFP (NT): 5′-CAAGCTGACCCTGAAGTTCAT-3′; shNADK#4: 5′-GCATCAGCATCACTACCTCAT-3′; shNADK#8: 5′-GATGACATTTCCAATCAGATA-3′; shNADK#10: 5′-GAATAAGGAATTGAGTCCAGA-3′.