Next-generation characterization of the Cancer Cell Line Encyclopedia.

Large panels of comprehensively characterized human cancer models, including the Cancer Cell Line Encyclopedia (CCLE), have provided a rigorous framework with which to study genetic variants, candidate targets, and small-molecule and biological therapeutics and to identify new marker-driven cancer dependencies. To improve our understanding of the molecular features that contribute to cancer phenotypes, including drug responses, here we have expanded the characterizations of cancer cell lines to include genetic, RNA splicing, DNA methylation, histone H3 modification, microRNA expression and reverse-phase protein array data for 1,072 cell lines from individuals of various lineages and ethnicities. Integration of these data with functional characterizations such as drug-sensitivity, short hairpin RNA knockdown and CRISPR-Cas9 knockout data reveals potential targets for cancer drugs and associated biomarkers. Together, this dataset and an accompanying public data portal provide a resource for the acceleration of cancer research using model cancer cell lines.

Fully unsupervised deep mode of action learning for phenotyping high-content cellular images.

MOTIVATION: The identification and discovery of phenotypes from high content screening images is a challenging task. Earlier works use image analysis pipelines to extract biological features, supervised training methods or generate features with neural networks pretrained on non-cellular images. We introduce a novel unsupervised deep learning algorithm to cluster cellular images with similar Mode-of-Action (MOA) together using only the images' pixel intensity values as input. It corrects for batch effect during training. Importantly, our method does not require the extraction of cell candidates and works from the entire images directly. RESULTS: The method achieves competitive results on the labeled subset of the BBBC021 dataset with an accuracy of 97.09% for correctly classifying the MOA by nearest neighbors matching. Importantly, we can train our approach on unannotated datasets. Therefore, our method can discover novel MOAs and annotate unlabeled compounds. The ability to train end-to-end on the full resolution images makes our method easy to apply and allows it to further distinguish treatments by their effect on proliferation. AVAILABILITY AND IMPLEMENTATION: Our code is available at https://github.com/Novartis/UMM-Discovery. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Contribution and clinical relevance of germline variation to the cancer transcriptome.

BACKGROUND: Somatic alterations in the cancer genome, some of which are associated with changes in gene expression, have been characterized in multiple studies across diverse cancer types. However, less is known about germline variants that influence tumor biology by shaping the cancer transcriptome. METHODS: We performed expression quantitative trait loci (eQTL) analyses using multi-dimensional data from The Cancer Genome Atlas to explore the role of germline variation in mediating the cancer transcriptome. After accounting for associations between somatic alterations and gene expression, we determined the contribution of inherited variants to the cancer transcriptome relative to that of somatic variants. Finally, we performed an interaction analysis using estimates of tumor cellularity to identify cell type-restricted eQTLs. RESULTS: The proportion of genes with at least one eQTL varied between cancer types, ranging between 0.8% in melanoma to 28.5% in thyroid cancer and was correlated more strongly with intratumor heterogeneity than with somatic alteration rates. Although contributions to variance in gene expression was low for most genes, some eQTLs accounted for more than 30% of expression of proximal genes. We identified cell type-restricted eQTLs in genes known to be cancer drivers including LPP and EZH2 that were associated with disease-specific mortality in TCGA but not associated with disease risk in published GWAS. Together, our results highlight the need to consider germline variation in interpreting cancer biology beyond risk prediction.

Resistance mechanisms to TP53-MDM2 inhibition identified by in vivo piggyBac transposon mutagenesis screen in an Arf-/- mouse model.

Inhibitors of double minute 2 protein (MDM2)-tumor protein 53 (TP53) interaction are predicted to be effective in tumors in which the TP53 gene is wild type, by preventing TP53 protein degradation. One such setting is represented by the frequent CDKN2A deletion in human cancer that, through inactivation of p14ARF, activates MDM2 protein, which in turn degrades TP53 tumor suppressor. Here we used piggyBac (PB) transposon insertional mutagenesis to anticipate resistance mechanisms occurring during treatment with the MDM2-TP53 inhibitor HDM201. Constitutive PB mutagenesis in Arf-/- mice provided a collection of spontaneous tumors with characterized insertional genetic landscapes. Tumors were allografted in large cohorts of mice to assess the pharmacologic effects of HDM201. Sixteen out of 21 allograft models were sensitive to HDM201 but ultimately relapsed under treatment. A comparison of tumors with acquired resistance to HDM201 and untreated tumors identified 87 genes that were differentially and significantly targeted by the PB transposon. Resistant tumors displayed a complex clonality pattern suggesting the emergence of several resistant subclones. Among the most frequent alterations conferring resistance, we observed somatic and insertional loss-of-function mutations in transformation-related protein 53 (Trp53) in 54% of tumors and transposon-mediated gain-of-function alterations in B-cell lymphoma-extra large (Bcl-xL), Mdm4, and two TP53 family members, resulting in expression of the TP53 dominant negative truncations ΔNTrp63 and ΔNTrp73. Enhanced BCL-xL and MDM4 protein expression was confirmed in resistant tumors, as well as in HDM201-resistant patient-derived tumor xenografts. Interestingly, concomitant inhibition of MDM2 and BCL-xL demonstrated significant synergy in p53 wild-type cell lines in vitro. Collectively, our findings identify several potential mechanisms by which TP53 wild-type tumors may escape MDM2-targeted therapy.

Correction of copy number induced false positives in CRISPR screens.

Cell autonomous cancer dependencies are now routinely identified using CRISPR loss-of-function viability screens. However, a bias exists that makes it difficult to assess the true essentiality of genes located in amplicons, since the entire amplified region can exhibit lethal scores. These false-positive hits can either be discarded from further analysis, which in cancer models can represent a significant number of hits, or methods can be developed to rescue the true-positives within amplified regions. We propose two methods to rescue true positive hits in amplified regions by correcting for this copy number artefact. The Local Drop Out (LDO) method uses the relative lethality scores within genomic regions to assess true essentiality and does not require additional orthogonal data (e.g. copy number value). LDO is meant to be used in screens covering a dense region of the genome (e.g. a whole chromosome or the whole genome). The General Additive Model (GAM) method models the screening data as a function of the known copy number values and removes the systematic effect from the measured lethality. GAM does not require the same density as LDO, but does require prior knowledge of the copy number values. Both methods have been developed with single sample experiments in mind so that the correction can be applied even in smaller screens. Here we demonstrate the efficacy of both methods at removing the copy number effect and rescuing hits from some of the amplified regions. We estimate a 70-80% decrease of false positive hits with either method in regions of high copy number compared to no correction.

Gene expression signature predicting high-grade prostate cancer responses to oxaliplatin.

Prostate cancer is one of the leading causes of cancer-related deaths among men. Several prognostic factors allow differentiation of low-grade tumors from high-grade tumors with high metastatic potential. High-grade tumors are currently treated with hormone therapy, to which taxanes are added when the tumors become resistant to castration. Clinical trials with other anticancer agents did not take into account the genetic backgrounds of the tumors, and most trials demonstrated low response rates. Here we used an in silico approach to screen for drug candidates that might be used as alternatives to taxanes, on the basis of a published expression signature involving 86 genes that could distinguish high-grade and low-grade tumors (Proc Natl Acad Sci USA 103:10991-10996, 2006). We explored the National Cancer Institute databases, which include data on the gene expression profiles of 60 human tumor cell lines and the in vitro sensitivities of the cell lines to anticancer drugs, and we identified several genes in the signature for which expression levels were correlated with chemosensitivity. As an example of the validation of this in silico approach, we identified a set of six genes for which expression levels could predict cell sensitivity to oxaliplatin but not cisplatin. This signature was validated in vitro through silencing of the genes in DU145, LNCaP, and C4-2B prostate cancer cells, which was accompanied by changes in oxaliplatin but not cisplatin cytotoxicity. These results demonstrate the relevance of our approach for the identification of both alternative treatments for high-grade prostate cancers and new biomarkers to predict clinical tumor responses.

Exquisite Sensitivity to Dual BRG1/BRM ATPase Inhibitors Reveals Broad SWI/SNF Dependencies in Acute Myeloid Leukemia.

Various subunits of mammalian SWI/SNF chromatin remodeling complexes display loss-of-function mutations characteristic of tumor suppressors in different cancers, but an additional role for SWI/SNF supporting cell survival in distinct cancer contexts is emerging. In particular, genetic dependence on the catalytic subunit BRG1/SMARCA4 has been observed in acute myelogenous leukemia (AML), yet the feasibility of direct therapeutic targeting of SWI/SNF catalytic activity in leukemia remains unknown. Here, we evaluated the activity of dual BRG1/BRM ATPase inhibitors across a genetically diverse panel of cancer cell lines and observed that hematopoietic cancer cell lines were among the most sensitive compared with other lineages. This result was striking in comparison with data from pooled short hairpin RNA screens, which showed that only a subset of leukemia cell lines display sensitivity to BRG1 knockdown. We demonstrate that combined genetic knockdown of BRG1 and BRM is required to recapitulate the effects of dual inhibitors, suggesting that SWI/SNF dependency in human leukemia extends beyond a predominantly BRG1-driven mechanism. Through gene expression and chromatin accessibility studies, we show that the dual inhibitors act at genomic loci associated with oncogenic transcription factors, and observe a downregulation of leukemic pathway genes, including MYC, a well-established target of BRG1 activity in AML. Overall, small-molecule inhibition of BRG1/BRM induced common transcriptional responses across leukemia models resulting in a spectrum of cellular phenotypes. IMPLICATIONS: Our studies reveal the breadth of SWI/SNF dependency in leukemia and support targeting SWI/SNF catalytic function as a potential therapeutic strategy in AML.

Microarray data quality control improves the detection of differentially expressed genes.

Microarrays have become a routine tool for biomedical research. Data quality assessment is an essential part of the analysis, but it is still not easy to perform objectively or in an automated manner, and as a result it is often neglected. Here, we compared two strategies of array-level quality control using five publicly available microarray experiments: outlier removal and array weights. We also compared them against no outlier removal and random array removal. We find that removing outlier arrays can improve the signal-to-noise ratio and thus strengthen the power of detecting differentially expressed genes. Using array weights is similarly effective, but its applicability is more limited. The quality metrics presented here are implemented in the Bioconductor package arrayQualityMetrics.

Systematic dissection of transcriptional regulatory networks by genome-scale and single-cell CRISPR screens.

Millions of putative transcriptional regulatory elements (TREs) have been cataloged in the human genome, yet their functional relevance in specific pathophysiological settings remains to be determined. This is critical to understand how oncogenic transcription factors (TFs) engage specific TREs to impose transcriptional programs underlying malignant phenotypes. Here, we combine cutting edge CRISPR screens and epigenomic profiling to functionally survey ≈15,000 TREs engaged by estrogen receptor (ER). We show that ER exerts its oncogenic role in breast cancer by engaging TREs enriched in GATA3, TFAP2C, and H3K27Ac signal. These TREs control critical downstream TFs, among which TFAP2C plays an essential role in ER-driven cell proliferation. Together, our work reveals novel insights into a critical oncogenic transcription program and provides a framework to map regulatory networks, enabling to dissect the function of the noncoding genome of cancer cells.

arrayQualityMetrics--a bioconductor package for quality assessment of microarray data.

SUMMARY: The assessment of data quality is a major concern in microarray analysis. arrayQualityMetrics is a Bioconductor package that provides a report with diagnostic plots for one or two colour microarray data. The quality metrics assess reproducibility, identify apparent outlier arrays and compute measures of signal-to-noise ratio. The tool handles most current microarray technologies and is amenable to use in automated analysis pipelines or for automatic report generation, as well as for use by individuals. The diagnosis of quality remains, in principle, a context-dependent judgement, but our tool provides powerful, automated, objective and comprehensive instruments on which to base a decision. AVAILABILITY: arrayQualityMetrics is a free and open source package, under LGPL license, available from the Bioconductor project at www.bioconductor.org. A users guide and examples are provided with the package. Some examples of HTML reports generated by arrayQualityMetrics can be found at http://www.microarray-quality.org

Importing ArrayExpress datasets into R/Bioconductor.

SUMMARY: ArrayExpress is one of the largest public repositories of microarray datasets. R/Bioconductor provides a comprehensive suite of microarray analysis and integrative bioinformatics software. However, easy ways for importing datasets from ArrayExpress into R/Bioconductor have been lacking. Here, we present such a tool that is suitable for both interactive and automated use. AVAILABILITY: The ArrayExpress package is available from the Bioconductor project at http://www.bioconductor.org. A users guide and examples are provided with the package.

Genetic heterogeneity and clonal evolution during metastasis in breast cancer patient-derived tumor xenograft models.

Genetic heterogeneity within a tumor arises by clonal evolution, and patients with highly heterogeneous tumors are more likely to be resistant to therapy and have reduced survival. Clonal evolution also occurs when a subset of cells leave the primary tumor to form metastases, which leads to reduced genetic heterogeneity at the metastatic site. Although this process has been observed in human cancer, experimental models which recapitulate this process are lacking. Patient-derived tumor xenografts (PDX) have been shown to recapitulate the patient's original tumor's intra-tumor genetic heterogeneity, as well as its genomics and response to treatment, but whether they can be used to model clonal evolution in the metastatic process is currently unknown. Here, we address this question by following genetic changes in two breast cancer PDX models during metastasis. First, we discovered that mouse stroma can be a confounding factor in assessing intra-tumor heterogeneity by whole exome sequencing, thus we developed a new bioinformatic approach to correct for this. Finally, in a spontaneous, but not experimental (tail-vein) metastasis model we observed a loss of heterogeneity in PDX metastases compared to their orthotopic "primary" tumors, confirming that PDX models can faithfully mimic the clonal evolution process undergone in human patients during metastatic spreading.

The Discovery of SWI/SNF Chromatin Remodeling Activity as a Novel and Targetable Dependency in Uveal Melanoma.

Uveal melanoma is a rare and aggressive cancer that originates in the eye. Currently, there are no approved targeted therapies and very few effective treatments for this cancer. Although activating mutations in the G protein alpha subunits, GNAQ and GNA11, are key genetic drivers of the disease, few additional drug targets have been identified. Recently, studies have identified context-specific roles for the mammalian SWI/SNF chromatin remodeling complexes (also known as BAF/PBAF) in various cancer lineages. Here, we find evidence that the SWI/SNF complex is essential through analysis of functional genomics screens and further validation in a panel of uveal melanoma cell lines using both genetic tools and small-molecule inhibitors of SWI/SNF. In addition, we describe a functional relationship between the SWI/SNF complex and the melanocyte lineage-specific transcription factor Microphthalmia-associated Transcription Factor, suggesting that these two factors cooperate to drive a transcriptional program essential for uveal melanoma cell survival. These studies highlight a critical role for SWI/SNF in uveal melanoma, and demonstrate a novel path toward the treatment of this cancer.

PAX8 activates metabolic genes via enhancer elements in Renal Cell Carcinoma.

Transcription factor networks shape the gene expression programs responsible for normal cell identity and pathogenic state. Using Core Regulatory Circuitry analysis (CRC), we identify PAX8 as a candidate oncogene in Renal Cell Carcinoma (RCC) cells. Validation of large-scale functional genomic screens confirms that PAX8 silencing leads to decreased proliferation of RCC cell lines. Epigenomic analyses of PAX8-dependent cistrome demonstrate that PAX8 largely occupies active enhancer elements controlling genes involved in various metabolic pathways. We selected the ferroxidase Ceruloplasmin (CP) as an exemplary gene to dissect PAX8 molecular functions. PAX8 recruits histone acetylation activity at bound enhancers looping onto the CP promoter. Importantly, CP expression correlates with sensitivity to PAX8 silencing and identifies a subset of RCC cases with poor survival. Our data identifies PAX8 as a candidate oncogene in RCC and provides a potential biomarker to monitor its activity.

FGF401, A First-In-Class Highly Selective and Potent FGFR4 Inhibitor for the Treatment of FGF19-Driven Hepatocellular Cancer.

Hepatocellular carcinoma (HCC) is the most common primary malignancy of the liver and it is the third leading cause of cancer-related deaths worldwide. Recently, aberrant signaling through the FGF19/FGFR4 axis has been implicated in HCC. Here, we describe the development of FGF401, a highly potent and selective, first in class, reversible-covalent small-molecule inhibitor of the kinase activity of FGFR4. FGF401 is exquisitely selective for FGFR4 versus the other FGFR paralogues FGFR1, FGFR2, FGFR3, and all other kinases in the kinome. FGF401 has excellent drug-like properties showing a robust pharmacokinetic/pharmacodynamics/efficacy relationship, driven by a fraction of time above the phospho-FGFR4 IC90 value. FGF401 has remarkable antitumor activity in mice bearing HCC tumor xenografts and patient-derived xenograft models that are positive for FGF19, FGFR4, and KLB. FGF401 is the first FGFR4 inhibitor to enter clinical trials, and a phase I/II study is currently ongoing in HCC and other solid malignancies.

Overexpression of DNA repair genes is associated with metastasis: a new hypothesis.

Tumorigenesis is a multistep process, where it is believed that the transformation of normal cells into tumoral cells needs a succession of genetic and epigenetic changes, such as point mutations, chromosomal rearrangements, and changes in gene expression level. All these modifications are supposed to confer a selective advantage and to generate highly malignant cancer cells. Until recently, the same selection procedure of rare cells in the tumour mass was believed to be necessary for the metastatic process. Using gene expression profiling, several recent publications report that a gene expression signature could discriminate between primary tumours with high metastatic potentiality and poor clinical outcome, and primary tumours that are not going to metastasize. Analysis of the biological pathways associated with metastatic potential points to cell adhesion, angiogenesis, cell cycle regulation, initiation of DNA synthesis, and DNA repair. Analysing human primary malignant melanoma and various biological processes, we have shown that the overexpression of DNA repair pathways, particularly those involved in double-stand break repair and surveillance of the DNA replication forks, is associated with metastasis and poor patient survival [V. Winnepenninckx, V. Lazar, S. Michiels, P. Dessen, M. Stas, S.R. Alonso, M.F. Avril, P.L. Ortiz Romero, T. Robert, O. Balacescu, A.M. Eggermont, G. Lenoir, A. Sarasin, T. Tursz, J.J. van den Oord, A. Spatz, Gene expression profiling of primary cutaneous melanoma and clinical outcome, J. Natl. Cancer Inst. 98 (2006) 472-482]. These results, also found by analysing other types of human tumours, such as breast or bladder cancers, would clearly explain the high resistance of metastasis towards chemo- and radiotherapies. Our hypothesis is that genetic instability is absolutely necessary to go from normal cells to tumoral cells, but one needs some type of genetic stabilization, which can be obtained by overexpressing specific DNA repair genes, in order to produce primary tumour cells that are genetically stable enough to be able to invade and give rise to distant metastasis.

Dose and Schedule Determine Distinct Molecular Mechanisms Underlying the Efficacy of the p53-MDM2 Inhibitor HDM201.

Activation of p53 by inhibitors of the p53-MDM2 interaction is being pursued as a therapeutic strategy in p53 wild-type cancers. Here, we report distinct mechanisms by which the novel, potent, and selective inhibitor of the p53-MDM2 interaction HDM201 elicits therapeutic efficacy when applied at various doses and schedules. Continuous exposure of HDM201 led to induction of p21 and delayed accumulation of apoptotic cells. By comparison, high-dose pulses of HDM201 were associated with marked induction of PUMA and a rapid onset of apoptosis. shRNA screens identified PUMA as a mediator of the p53 response specifically in the pulsed regimen. Consistent with this, the single high-dose HDM201 regimen resulted in rapid and marked induction of PUMA expression and apoptosis together with downregulation of Bcl-xL in vivo Knockdown of Bcl-xL was identified as the top sensitizer to HDM201 in vitro, and Bcl-xL was enriched in relapsing tumors from mice treated with intermittent high doses of HDM201. These findings define a regimen-dependent mechanism by which disruption of MDM2-p53 elicits therapeutic efficacy when given with infrequent dosing. In an ongoing HDM201 trial, the observed exposure-response relationship indicates that the molecular mechanism elicited by pulse dosing is likely reproducible in patients. These data support the clinical comparison of daily and intermittent regimens of p53-MDM2 inhibitors.Significance: Pulsed high doses versus sustained low doses of the p53-MDM2 inhibitor HDM201 elicit a proapoptotic response from wild-type p53 cancer cells, offering guidance to current clinical trials with this and other drugs that exploit the activity of p53. Cancer Res; 78(21); 6257-67. ©2018 AACR.

Maximizing the Efficacy of MAPK-Targeted Treatment in PTENLOF/BRAFMUT Melanoma through PI3K and IGF1R Inhibition.

The introduction of MAPK pathway inhibitors paved the road for significant advancements in the treatment of BRAF-mutant (BRAF(MUT)) melanoma. However, even BRAF/MEK inhibitor combination therapy has failed to offer a curative treatment option, most likely because these pathways constitute a codependent signaling network. Concomitant PTEN loss of function (PTEN(LOF)) occurs in approximately 40% of BRAF(MUT) melanomas. In this study, we sought to identify the nodes of the PTEN/PI3K pathway that would be amenable to combined therapy with MAPK pathway inhibitors for the treatment of PTEN(LOF)/BRAF(MUT) melanoma. Large-scale compound sensitivity profiling revealed that PTEN(LOF) melanoma cell lines were sensitive to PI3Kß inhibitors, albeit only partially. An unbiased shRNA screen (7,500 genes and 20 shRNAs/genes) across 11 cell lines in the presence of a PI3Kß inhibitor identified an adaptive response involving the IGF1R-PI3Kα axis. Combined inhibition of the MAPK pathway, PI3Kß, and PI3Kα or insulin-like growth factor receptor 1 (IGF1R) synergistically sustained pathway blockade, induced apoptosis, and inhibited tumor growth in PTEN(LOF)/BRAF(MUT) melanoma models. Notably, combined treatment with the IGF1R inhibitor, but not the PI3Kα inhibitor, failed to elevate glucose or insulin signaling. Taken together, our findings provide a strong rationale for testing combinations of panPI3K, PI3Kß + IGF1R, and MAPK pathway inhibitors in PTEN(LOF)/BRAF(MUT) melanoma patients to achieve maximal response.

CRISPR Screens Provide a Comprehensive Assessment of Cancer Vulnerabilities but Generate False-Positive Hits for Highly Amplified Genomic Regions.

UNLABELLED: CRISPR/Cas9 has emerged as a powerful new tool to systematically probe gene function. We compared the performance of CRISPR to RNAi-based loss-of-function screens for the identification of cancer dependencies across multiple cancer cell lines. CRISPR dropout screens consistently identified more lethal genes than RNAi, implying that the identification of many cellular dependencies may require full gene inactivation. However, in two aneuploid cancer models, we found that all genes within highly amplified regions, including nonexpressed genes, scored as lethal by CRISPR, revealing an unanticipated class of false-positive hits. In addition, using a CRISPR tiling screen, we found that sgRNAs targeting essential domains generate the strongest lethality phenotypes and thus provide a strategy to rapidly define the protein domains required for cancer dependence. Collectively, these findings not only demonstrate the utility of CRISPR screens in the identification of cancer-essential genes, but also reveal the need to carefully control for false-positive results in chromosomally unstable cancer lines. SIGNIFICANCE: We show in this study that CRISPR-based screens have a significantly lower false-negative rate compared with RNAi-based screens, but have specific liabilities particularly in the interrogation of regions of genome amplification. Therefore, this study provides critical insights for applying CRISPR-based screens toward the systematic identification of new cancer targets. Cancer Discov; 6(8); 900-13. ©2016 AACR.See related commentary by Sheel and Xue, p. 824See related article by Aguirre et al., p. 914This article is highlighted in the In This Issue feature, p. 803.