An evaluation of the challenges to developing tumor BRCA1 and BRCA2 testing methodologies for clinical practice.

Ovarian cancer patients with germline or somatic pathogenic variants benefit from treatment with poly ADP ribose polymerase (PARP) inhibitors. Tumor BRCA1/2 testing is more challenging than germline testing as the majority of samples are formalin-fixed paraffin embedded (FFPE), the tumor genome is complex, and the allelic fraction of somatic variants can be low. We collaborated with 10 laboratories testing BRCA1/2 in tumors to compare different approaches to identify clinically important variants within FFPE tumor DNA samples. This was not a proficiency study but an inter-laboratory comparison to identify common issues. Each laboratory received the same tumor DNA samples ranging in genotype, quantity, quality, and variant allele frequency (VAF). Each laboratory performed their preferred next-generation sequencing method to report on the variants. No false positive results were reported in this small study and the majority of methods detected the low VAF variants. A number of variants were not detected due to the bioinformatics analysis, variant classification, or insufficient DNA. The use of hybridization capture or short amplicon methods are recommended based on a bioinformatic assessment of the data. The study highlights the importance of establishing standards and standardization for tBRCA testing particularly when the test results dictate clinical decisions regarding life extending therapies.

VarDict: a novel and versatile variant caller for next-generation sequencing in cancer research.

Accurate variant calling in next generation sequencing (NGS) is critical to understand cancer genomes better. Here we present VarDict, a novel and versatile variant caller for both DNA- and RNA-sequencing data. VarDict simultaneously calls SNV, MNV, InDels, complex and structural variants, expanding the detected genetic driver landscape of tumors. It performs local realignments on the fly for more accurate allele frequency estimation. VarDict performance scales linearly to sequencing depth, enabling ultra-deep sequencing used to explore tumor evolution or detect tumor DNA circulating in blood. In addition, VarDict performs amplicon aware variant calling for polymerase chain reaction (PCR)-based targeted sequencing often used in diagnostic settings, and is able to detect PCR artifacts. Finally, VarDict also detects differences in somatic and loss of heterozygosity variants between paired samples. VarDict reprocessing of The Cancer Genome Atlas (TCGA) Lung Adenocarcinoma dataset called known driver mutations in KRAS, EGFR, BRAF, PIK3CA and MET in 16% more patients than previously published variant calls. We believe VarDict will greatly facilitate application of NGS in clinical cancer research.

A reliable method for the detection of BRCA1 and BRCA2 mutations in fixed tumour tissue utilising multiplex PCR-based targeted next generation sequencing.

BACKGROUND: Germline mutations in BRCA1 or BRCA2 lead to a high lifetime probability of developing ovarian or breast cancer. These genes can also be involved in the development of non-hereditary tumours as somatic BRCA1/2 pathogenic variants are found in some of these cancers. Since patients with somatic BRCA pathogenic variants may benefit from treatment with poly ADP ribose polymerase inhibitors, it is important to be able to test for somatic changes in routinely available tumour samples. Such samples are typically formalin-fixed paraffin-embedded (FFPE) tissue, where the extracted DNA tends to be highly fragmented and of limited quantity, making analysis of large genes such as BRCA1 and BRCA2 challenging. This is made more difficult as somatic changes may be evident in only part of the sample, due to the presence of normal tissue. METHODS: We examined the feasibility of analysing DNA extracted from FFPE ovarian and breast tumour tissue to identify significant DNA variants in BRCA1/ BRCA2 using next generation sequencing methods that were sensitive enough to detect low level mutations, multiplexed to reduce the amount of DNA required and had short amplicon design. The utility of two GeneRead DNAseq Targeted Exon Enrichment Panels with different designs targeting only BRCA1/2 exons, and the Ion AmpliSeq BRCA community panel, followed by library preparation and adaptor ligation using the TruSeq DNA PCR-Free HT Sample Preparation Kit and NGS analysis on the MiSeq were investigated. RESULTS: Using the GeneRead method, we successfully analysed over 76% of samples, with >95% coverage of BRCA1/2 coding regions and a mean average read depth of >1000-fold. All mutations identified were confirmed where possible by Sanger sequencing or replication to eliminate the risk of false positive results due to artefacts within FFPE material. Admixture experiments demonstrated that BRCA1/2 variants could be detected if present in >10% of the sample. A sample subset was evaluated using the Ion AmpliSeq BRCA panel, achieving >99% coverage and sufficient read depth for a proportion of the samples. CONCLUSIONS: Detection of BRCA1/2 variants in fixed tissue is feasible, and could be performed prospectively to facilitate optimum treatment decisions for ovarian or breast cancer patients.

Model selection for prognostic time-to-event gene signature discovery with applications in early breast cancer data.

Model selection between competing models is a key consideration in the discovery of prognostic multigene signatures. The use of appropriate statistical performance measures as well as verification of biological significance of the signatures is imperative to maximise the chance of external validation of the generated signatures. Current approaches in time-to-event studies often use only a single measure of performance in model selection, such as logrank test p-values, or dichotomise the follow-up times at some phase of the study to facilitate signature discovery. In this study we improve the prognostic signature discovery process through the application of the multivariate partial Cox model combined with the concordance index, hazard ratio of predictions, independence from available clinical covariates and biological enrichment as measures of signature performance. The proposed framework was applied to discover prognostic multigene signatures from early breast cancer data. The partial Cox model combined with the multiple performance measures were used in both guiding the selection of the optimal panel of prognostic genes and prediction of risk within cross validation without dichotomising the follow-up times at any stage. The signatures were successfully externally cross validated in independent breast cancer datasets, yielding a hazard ratio of 2.55 [1.44, 4.51] for the top ranking signature.

Genome-wide CRISPR screens identify the YAP/TEAD axis as a driver of persister cells in EGFR mutant lung cancer.

Most lung cancer patients with metastatic cancer eventually relapse with drug-resistant disease following treatment and EGFR mutant lung cancer is no exception. Genome-wide CRISPR screens, to either knock out or overexpress all protein-coding genes in cancer cell lines, revealed the landscape of pathways that cause resistance to the EGFR inhibitors osimertinib or gefitinib in EGFR mutant lung cancer. Among the most recurrent resistance genes were those that regulate the Hippo pathway. Following osimertinib treatment a subpopulation of cancer cells are able to survive and over time develop stable resistance. These 'persister' cells can exploit non-genetic (transcriptional) programs that enable cancer cells to survive drug treatment. Using genetic and pharmacologic tools we identified Hippo signalling as an important non-genetic mechanism of cell survival following osimertinib treatment. Further, we show that combinatorial targeting of the Hippo pathway and EGFR is highly effective in EGFR mutant lung cancer cells and patient-derived organoids, suggesting a new therapeutic strategy for EGFR mutant lung cancer patients.

Knowledge graph-based recommendation framework identifies drivers of resistance in EGFR mutant non-small cell lung cancer.

Resistance to EGFR inhibitors (EGFRi) presents a major obstacle in treating non-small cell lung cancer (NSCLC). One of the most exciting new ways to find potential resistance markers involves running functional genetic screens, such as CRISPR, followed by manual triage of significantly enriched genes. This triage process to identify 'high value' hits resulting from the CRISPR screen involves manual curation that requires specialized knowledge and can take even experts several months to comprehensively complete. To find key drivers of resistance faster we build a recommendation system on top of a heterogeneous biomedical knowledge graph integrating pre-clinical, clinical, and literature evidence. The recommender system ranks genes based on trade-offs between diverse types of evidence linking them to potential mechanisms of EGFRi resistance. This unbiased approach identifies 57 resistance markers from >3,000 genes, reducing hit identification time from months to minutes. In addition to reproducing known resistance markers, our method identifies previously unexplored resistance mechanisms that we prospectively validate.

Drug-gene Interaction Screens Coupled to Tumor Data Analyses Identify the Most Clinically Relevant Cancer Vulnerabilities Driving Sensitivity to PARP Inhibition.

PARP inhibitors (PARPi) are currently indicated for the treatment of ovarian, breast, pancreatic, and prostate cancers harboring mutations in the tumor suppressor genes BRCA1 or BRCA2. In the case of ovarian and prostate cancers, their classification as homologous recombination repair (HRR) deficient (HRD) or mutated also makes PARPi an available treatment option beyond BRCA1 or BRCA2 mutational status. However, identification of the most relevant genetic alterations driving the HRD phenotype has proven difficult and recent data have shown that other genetic alterations not affecting HRR are also capable of driving PARPi responses. To gain insight into the genetics driving PARPi sensitivity, we performed CRISPR-Cas9 loss-of-function screens in six PARPi-insensitive cell lines and combined the output with published PARPi datasets from eight additional cell lines. Ensuing exploration of the data identified 110 genes whose inactivation is strongly linked to sensitivity to PARPi. Parallel cell line generation of isogenic gene knockouts in ovarian and prostate cancer cell lines identified that inactivation of core HRR factors is required for driving in vitro PARPi responses comparable with the ones observed for BRCA1 or BRCA2 mutations. Moreover, pan-cancer genetic, transcriptomic, and epigenetic data analyses of these 110 genes highlight the ones most frequently inactivated in tumors, making this study a valuable resource for prospective identification of potential PARPi-responsive patient populations. Importantly, our investigations uncover XRCC3 gene silencing as a potential new prognostic biomarker of PARPi sensitivity in prostate cancer. Significance: This study identifies tumor genetic backgrounds where to expand the use of PARPis beyond mutations in BRCA1 or BRCA2. This is achieved by combining the output of unbiased genome-wide loss-of-function CRISPR-Cas9 genetic screens with bioinformatics analysis of biallelic losses of the identified genes in public tumor datasets, unveiling loss of the DNA repair gene XRCC3 as a potential biomarker of PARPi sensitivity in prostate cancer.

Hydroxymethylation profile of cell-free DNA is a biomarker for early colorectal cancer.

Early detection of cancer will improve survival rates. The blood biomarker 5-hydroxymethylcytosine has been shown to discriminate cancer. In a large covariate-controlled study of over two thousand individual blood samples, we created, tested and explored the properties of a 5-hydroxymethylcytosine-based classifier to detect colorectal cancer (CRC). In an independent validation sample set, the classifier discriminated CRC samples from controls with an area under the receiver operating characteristic curve (AUC) of 90% (95% CI [87, 93]). Sensitivity was 55% at 95% specificity. Performance was similar for early stage 1 (AUC 89%; 95% CI [83, 94]) and late stage 4 CRC (AUC 94%; 95% CI [89, 98]). The classifier could detect CRC even when the proportion of tumor DNA in blood was undetectable by other methods. Expanding the classifier to include information about cell-free DNA fragment size and abundance across the genome led to gains in sensitivity (63% at 95% specificity), with similar overall performance (AUC 91%; 95% CI [89, 94]). We confirm that 5-hydroxymethylcytosine can be used to detect CRC, even in early-stage disease. Therefore, the inclusion of 5-hydroxymethylcytosine in multianalyte testing could improve sensitivity for the detection of early-stage cancer.

Robust regression for periodicity detection in non-uniformly sampled time-course gene expression data.

BACKGROUND: In practice many biological time series measurements, including gene microarrays, are conducted at time points that seem to be interesting in the biologist's opinion and not necessarily at fixed time intervals. In many circumstances we are interested in finding targets that are expressed periodically. To tackle the problems of uneven sampling and unknown type of noise in periodicity detection, we propose to use robust regression. METHODS: The aim of this paper is to develop a general framework for robust periodicity detection and review and rank different approaches by means of simulations. We also show the results for some real measurement data. RESULTS: The simulation results clearly show that when the sampling of time series gets more and more uneven, the methods that assume even sampling become unusable. We find that M-estimation provides a good compromise between robustness and computational efficiency. CONCLUSION: Since uneven sampling occurs often in biological measurements, the robust methods developed in this paper are expected to have many uses. The regression based formulation of the periodicity detection problem easily adapts to non-uniform sampling. Using robust regression helps to reject inconsistently behaving data points. AVAILABILITY: The implementations are currently available for Matlab and will be made available for the users of R as well. More information can be found in the web-supplement 1.

Altered regulation and expression of genes by BET family of proteins in COPD patients.

BACKGROUND: BET proteins (BRD2, BRD3, BRDT and BRD4) belong to the family of bromodomain containing proteins, which form a class of transcriptional co-regulators. BET proteins bind to acetylated lysine residues in the histones of nucleosomal chromatin and function either as co-activators or co-repressors of gene expression. An imbalance between HAT and HDAC activities resulting in hyperacetylation of histones has been identified in COPD. We hypothesized that pan-BET inhibitor (JQ1) treatment of BET protein interactions with hyperacetylated sites in the chromatin will regulate excessive activation of pro-inflammatory genes in key inflammatory drivers of alveolar macrophages (AM) in COPD. METHODS AND FINDINGS: Transcriptome analysis of AM from COPD patients indicated up-regulation of macrophage M1 type genes upon LPS stimulation. Pan-BET inhibitor JQ1 treatment attenuated expression of multiple genes, including pro-inflammatory cytokines and regulators of innate and adaptive immune cells. We demonstrated for the first time that JQ1 differentially modulated LPS-induced cytokine release from AM or peripheral blood mononuclear cells (PBMC) of COPD patients compared to PBMC of healthy controls. Using the BET regulated gene signature, we identified a subset of COPD patients, which we propose to benefit from BET inhibition. CONCLUSIONS: This work demonstrates that the effects of pan-BET inhibition through JQ1 treatment of inflammatory cells differs between COPD patients and healthy controls, and the expression of BET protein regulated genes is altered in COPD. These findings provide evidence of histone hyperacetylation as a mechanism driving chronic inflammatory changes in COPD.

Prioritisation of structural variant calls in cancer genomes.

Sensitivity of short read DNA-sequencing for gene fusion detection is improving, but is hampered by the significant amount of noise composed of uninteresting or false positive hits in the data. In this paper we describe a tiered prioritisation approach to extract high impact gene fusion events from existing structural variant calls. Using cell line and patient DNA sequence data we improve the annotation and interpretation of structural variant calls to best highlight likely cancer driving fusions. We also considerably improve on the automated visualisation of the high impact structural variants to highlight the effects of the variants on the resulting transcripts. The resulting framework greatly improves on readily detecting clinically actionable structural variants.

Correction: Altered regulation and expression of genes by BET family of proteins in COPD patients.

[This corrects the article DOI: 10.1371/journal.pone.0173115.].

Rational Selection of Syngeneic Preclinical Tumor Models for Immunotherapeutic Drug Discovery.

Murine syngeneic tumor models are critical to novel immuno-based therapy development, but the molecular and immunologic features of these models are still not clearly defined. The translational relevance of differences between the models is not fully understood, impeding appropriate preclinical model selection for target validation, and ultimately hindering drug development. Across a panel of commonly used murine syngeneic tumor models, we showed variable responsiveness to immunotherapies. We used array comparative genomic hybridization, whole-exome sequencing, exon microarray analysis, and flow cytometry to extensively characterize these models, which revealed striking differences that may underlie these contrasting response profiles. We identified strong differential gene expression in immune-related pathways and changes in immune cell-specific genes that suggested differences in tumor immune infiltrates between models. Further investigation using flow cytometry showed differences in both the composition and magnitude of the tumor immune infiltrates, identifying models that harbor "inflamed" and "non-inflamed" tumor immune infiltrate phenotypes. We also found that immunosuppressive cell types predominated in syngeneic mouse tumor models that did not respond to immune-checkpoint blockade, whereas cytotoxic effector immune cells were enriched in responsive models. A cytotoxic cell-rich tumor immune infiltrate has been correlated with increased efficacy of immunotherapies in the clinic, and these differences could underlie the varying response profiles to immunotherapy between the syngeneic models. This characterization highlighted the importance of extensive profiling and will enable investigators to select appropriate models to interrogate the activity of immunotherapies as well as combinations with targeted therapies in vivo Cancer Immunol Res; 5(1); 29-41. ©2016 AACR.

Simulation of microarray data with realistic characteristics.

BACKGROUND: Microarray technologies have become common tools in biological research. As a result, a need for effective computational methods for data analysis has emerged. Numerous different algorithms have been proposed for analyzing the data. However, an objective evaluation of the proposed algorithms is not possible due to the lack of biological ground truth information. To overcome this fundamental problem, the use of simulated microarray data for algorithm validation has been proposed. RESULTS: We present a microarray simulation model which can be used to validate different kinds of data analysis algorithms. The proposed model is unique in the sense that it includes all the steps that affect the quality of real microarray data. These steps include the simulation of biological ground truth data, applying biological and measurement technology specific error models, and finally simulating the microarray slide manufacturing and hybridization. After all these steps are taken into account, the simulated data has realistic biological and statistical characteristics. The applicability of the proposed model is demonstrated by several examples. CONCLUSION: The proposed microarray simulation model is modular and can be used in different kinds of applications. It includes several error models that have been proposed earlier and it can be used with different types of input data. The model can be used to simulate both spotted two-channel and oligonucleotide based single-channel microarrays. All this makes the model a valuable tool for example in validation of data analysis algorithms.

Disambiguate: An open-source application for disambiguating two species in next generation sequencing data from grafted samples.

Grafting of cell lines and primary tumours is a crucial step in the drug development process between cell line studies and clinical trials. Disambiguate is a program for computationally separating the sequencing reads of two species derived from grafted samples. Disambiguate operates on DNA or RNA-seq alignments to the two species and separates the components at very high sensitivity and specificity as illustrated in artificially mixed human-mouse samples. This allows for maximum recovery of data from target tumours for more accurate variant calling and gene expression quantification. Given that no general use open source algorithm accessible to the bioinformatics community exists for the purposes of separating the two species data, the proposed Disambiguate tool presents a novel approach and improvement to performing sequence analysis of grafted samples. Both Python and C++ implementations are available and they are integrated into several open and closed source pipelines. Disambiguate is open source and is freely available at https://github.com/AstraZeneca-NGS/disambiguate.

AZD5153: A Novel Bivalent BET Bromodomain Inhibitor Highly Active against Hematologic Malignancies.

The bromodomain and extraterminal (BET) protein BRD4 regulates gene expression via recruitment of transcriptional regulatory complexes to acetylated chromatin. Pharmacological targeting of BRD4 bromodomains by small molecule inhibitors has proven to be an effective means to disrupt aberrant transcriptional programs critical for tumor growth and/or survival. Herein, we report AZD5153, a potent, selective, and orally available BET/BRD4 bromodomain inhibitor possessing a bivalent binding mode. Unlike previously described monovalent inhibitors, AZD5153 ligates two bromodomains in BRD4 simultaneously. The enhanced avidity afforded through bivalent binding translates into increased cellular and antitumor activity in preclinical hematologic tumor models. In vivo administration of AZD5153 led to tumor stasis or regression in multiple xenograft models of acute myeloid leukemia, multiple myeloma, and diffuse large B-cell lymphoma. The relationship between AZD5153 exposure and efficacy suggests that prolonged BRD4 target coverage is a primary efficacy driver. AZD5153 treatment markedly affects transcriptional programs of MYC, E2F, and mTOR. Of note, mTOR pathway modulation is associated with cell line sensitivity to AZD5153. Transcriptional modulation of MYC and HEXIM1 was confirmed in AZD5153-treated human whole blood, thus supporting their use as clinical pharmacodynamic biomarkers. This study establishes AZD5153 as a highly potent, orally available BET/BRD4 inhibitor and provides a rationale for clinical development in hematologic malignancies. Mol Cancer Ther; 15(11); 2563-74. ©2016 AACR.

Robust detection of periodic time series measured from biological systems.

BACKGROUND: Periodic phenomena are widespread in biology. The problem of finding periodicity in biological time series can be viewed as a multiple hypothesis testing of the spectral content of a given time series. The exact noise characteristics are unknown in many bioinformatics applications. Furthermore, the observed time series can exhibit other non-idealities, such as outliers, short length and distortion from the original wave form. Hence, the computational methods should preferably be robust against such anomalies in the data. RESULTS: We propose a general-purpose robust testing procedure for finding periodic sequences in multiple time series data. The proposed method is based on a robust spectral estimator which is incorporated into the hypothesis testing framework using a so-called g-statistic together with correction for multiple testing. This results in a robust testing procedure which is insensitive to heavy contamination of outliers, missing-values, short time series, nonlinear distortions, and is completely insensitive to any monotone nonlinear distortions. The performance of the methods is evaluated by performing extensive simulations. In addition, we compare the proposed method with another recent statistical signal detection estimator that uses Fisher's test, based on the Gaussian noise assumption. The results demonstrate that the proposed robust method provides remarkably better robustness properties. Moreover, the performance of the proposed method is preferable also in the standard Gaussian case. We validate the performance of the proposed method on real data on which the method performs very favorably. CONCLUSION: As the time series measured from biological systems are usually short and prone to contain different kinds of non-idealities, we are very optimistic about the multitude of possible applications for our proposed robust statistical periodicity detection method.

The calcineurin-NFAT pathway controls activity-dependent circadian gene expression in slow skeletal muscle.

OBJECTIVE: Physical activity and circadian rhythms are well-established determinants of human health and disease, but the relationship between muscle activity and the circadian regulation of muscle genes is a relatively new area of research. It is unknown whether muscle activity and muscle clock rhythms are coupled together, nor whether activity rhythms can drive circadian gene expression in skeletal muscle. METHODS: We compared the circadian transcriptomes of two mouse hindlimb muscles with vastly different circadian activity patterns, the continuously active slow soleus and the sporadically active fast tibialis anterior, in the presence or absence of a functional skeletal muscle clock (skeletal muscle-specific Bmal1 KO). In addition, we compared the effect of denervation on muscle circadian gene expression. RESULTS: We found that different skeletal muscles exhibit major differences in their circadian transcriptomes, yet core clock gene oscillations were essentially identical in fast and slow muscles. Furthermore, denervation caused relatively minor changes in circadian expression of most core clock genes, yet major differences in expression level, phase and amplitude of many muscle circadian genes. CONCLUSIONS: We report that activity controls the oscillation of around 15% of skeletal muscle circadian genes independently of the core muscle clock, and we have identified the Ca(2+)-dependent calcineurin-NFAT pathway as an important mediator of activity-dependent circadian gene expression, showing that circadian locomotor activity rhythms drive circadian rhythms of NFAT nuclear translocation and target gene expression.

PIM kinase inhibitor AZD1208 for treatment of MYC-driven prostate cancer.

BACKGROUND: PIM1 kinase is coexpressed with c-MYC in human prostate cancers (PCs) and dramatically enhances c-MYC-induced tumorigenicity. Here we examine the effects of a novel oral PIM inhibitor, AZD1208, on prostate tumorigenesis and recurrence. METHODS: A mouse c-MYC/Pim1-transduced tissue recombination PC model, Myc-CaP allografts, and human PC xenografts were treated with AZD1208 (n = 5-11 per group). Androgen-sensitive and castrate-resistant prostate cancer (CRPC) models were studied as well as the effects of hypoxia and radiation. RNA sequencing was used to analyze drug-induced gene expression changes. Results were analyzed with χ(2) test. Student's t test and nonparametric Mann-Whitney rank sum U Test. All statistical tests were two-sided. RESULTS: AZD1208 inhibited tumorigenesis in tissue recombinants, Myc-CaP, and human PC xenograft models. PIM inhibition decreased c-MYC/Pim1 graft growth by 54.3 ± 39% (P < .001), decreased cellular proliferation by 46 ± 14% (P = .016), and increased apoptosis by 326 ± 170% (P = .039). AZD1208 suppressed multiple protumorigenic pathways, including the MYC gene program. However, it also downregulated the p53 pathway. Hypoxia and radiation induced PIM1 in prostate cancer cells, and AZD1208 functioned as a radiation sensitizer. Recurrent tumors postcastration responded transiently to either AZD1208 or radiation treatment, and combination treatment resulted in more sustained inhibition of tumor growth. Cell lines established from recurrent, AZD1208-resistant tumors again revealed downregulation of the p53 pathway. Irradiated AZD1208-treated tumors robustly upregulated p53, providing a possible mechanistic explanation for the effectiveness of combination therapy. Finally, an AZD1208-resistant gene signature was found to be associated with biochemical recurrence in PC patients. CONCLUSIONS: PIM inhibition is a potential treatment for MYC-driven prostate cancers including CRPC, and its effectiveness may be enhanced by activators of the p53 pathway, such as radiation.

Acquired Resistance to the Mutant-Selective EGFR Inhibitor AZD9291 Is Associated with Increased Dependence on RAS Signaling in Preclinical Models.

Resistance to targeted EGFR inhibitors is likely to develop in EGFR-mutant lung cancers. Early identification of innate or acquired resistance mechanisms to these agents is essential to direct development of future therapies. We describe the detection of heterogeneous mechanisms of resistance within populations of EGFR-mutant cells (PC9 and/or NCI-H1975) with acquired resistance to current and newly developed EGFR tyrosine kinase inhibitors, including AZD9291. We report the detection of NRAS mutations, including a novel E63K mutation, and a gain of copy number of WT NRAS or WT KRAS in cell populations resistant to gefitinib, afatinib, WZ4002, or AZD9291. Compared with parental cells, a number of resistant cell populations were more sensitive to inhibition by the MEK inhibitor selumetinib (AZD6244; ARRY-142886) when treated in combination with the originating EGFR inhibitor. In vitro, a combination of AZD9291 with selumetinib prevented emergence of resistance in PC9 cells and delayed resistance in NCI-H1975 cells. In vivo, concomitant dosing of AZD9291 with selumetinib caused regression of AZD9291-resistant tumors in an EGFRm/T790M transgenic model. Our data support the use of a combination of AZD9291 with a MEK inhibitor to delay or prevent resistance to AZD9291 in EGFRm and/or EGFRm/T790M tumors. Furthermore, these findings suggest that NRAS modifications in tumor samples from patients who have progressed on current or EGFR inhibitors in development may support subsequent treatment with a combination of EGFR and MEK inhibition.