A genome‐scale screen reveals context‐dependent ovarian cancer sensitivity to mi RNA overexpression
2015; Springer Nature; Volume: 11; Issue: 12 Linguagem: Inglês
10.15252/msb.20156308
ISSN1744-4292
AutoresBenjamin B. Shields, Chad V. Pecot, Hua Gao, Elizabeth A. McMillan, Malia B. Potts, Christa Nagel, Scott C. Purinton, Ying Wang, Cristina Ivan, Hyun Seok Kim, Robert Borkowski, Shaheen Khan, Cristian Rodriguez‐Aguayo, Gabriel López-Berestein, Jayanthi Lea, Adi F. Gazdar, Keith Baggerly, Anil K. Sood, Michael A. White,
Tópico(s)MicroRNA in disease regulation
ResumoArticle11 December 2015Open Access Source Data A genome-scale screen reveals context-dependent ovarian cancer sensitivity to miRNA overexpression Benjamin B Shields Benjamin B Shields Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Chad V Pecot Chad V Pecot Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Hua Gao Hua Gao Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Elizabeth McMillan Elizabeth McMillan Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Malia Potts Malia Potts Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Christa Nagel Christa Nagel Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Scott Purinton Scott Purinton Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Ying Wang Ying Wang Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Cristina Ivan Cristina Ivan Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Hyun Seok Kim Hyun Seok Kim Severance Biomedical Science Institute, Yonsei University College of Medicine, Seoul, Korea Search for more papers by this author Robert J Borkowski Robert J Borkowski Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Shaheen Khan Shaheen Khan Immunology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Cristian Rodriguez-Aguayo Cristian Rodriguez-Aguayo Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Gabriel Lopez-Berestein Gabriel Lopez-Berestein Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Jayanthi Lea Jayanthi Lea Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Adi Gazdar Adi Gazdar Pathology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Keith A Baggerly Keith A Baggerly Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Anil K Sood Anil K Sood Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Michael A White Corresponding Author Michael A White Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Benjamin B Shields Benjamin B Shields Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Chad V Pecot Chad V Pecot Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Hua Gao Hua Gao Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Elizabeth McMillan Elizabeth McMillan Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Malia Potts Malia Potts Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Christa Nagel Christa Nagel Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Scott Purinton Scott Purinton Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Ying Wang Ying Wang Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Cristina Ivan Cristina Ivan Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Hyun Seok Kim Hyun Seok Kim Severance Biomedical Science Institute, Yonsei University College of Medicine, Seoul, Korea Search for more papers by this author Robert J Borkowski Robert J Borkowski Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Shaheen Khan Shaheen Khan Immunology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Cristian Rodriguez-Aguayo Cristian Rodriguez-Aguayo Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Gabriel Lopez-Berestein Gabriel Lopez-Berestein Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Jayanthi Lea Jayanthi Lea Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Adi Gazdar Adi Gazdar Pathology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Keith A Baggerly Keith A Baggerly Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Anil K Sood Anil K Sood Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA Search for more papers by this author Michael A White Corresponding Author Michael A White Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA Search for more papers by this author Author Information Benjamin B Shields1, Chad V Pecot2, Hua Gao1, Elizabeth McMillan1, Malia Potts1, Christa Nagel3, Scott Purinton3, Ying Wang4, Cristina Ivan4, Hyun Seok Kim5, Robert J Borkowski1, Shaheen Khan6, Cristian Rodriguez-Aguayo2, Gabriel Lopez-Berestein2, Jayanthi Lea3, Adi Gazdar7, Keith A Baggerly4, Anil K Sood2 and Michael A White 1 1Departments of Cell Biology, University of Texas Southwestern Medical Center, Dallas, TX, USA 2Center for RNA interference and Non-Coding RNA, MD Anderson Cancer Center, Houston, TX, USA 3Obstetrics and Gynecology, University of Texas Southwestern Medical Center, Dallas, TX, USA 4Department of Bioinformatics and Computational Biology, MD Anderson Cancer Center, Houston, TX, USA 5Severance Biomedical Science Institute, Yonsei University College of Medicine, Seoul, Korea 6Immunology, University of Texas Southwestern Medical Center, Dallas, TX, USA 7Pathology, University of Texas Southwestern Medical Center, Dallas, TX, USA *Corresponding author. Tel: +1 214 648 4212; Fax: +1 214 648 5814; E-mail: [email protected] Molecular Systems Biology (2015)11:842https://doi.org/10.15252/msb.20156308 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Figures & Info Abstract Large-scale molecular annotation of epithelial ovarian cancer (EOC) indicates remarkable heterogeneity in the etiology of that disease. This diversity presents a significant obstacle against intervention target discovery. However, inactivation of miRNA biogenesis is commonly associated with advanced disease. Thus, restoration of miRNA activity may represent a common vulnerability among diverse EOC oncogenotypes. To test this, we employed genome-scale, gain-of-function, miRNA mimic toxicity screens in a large, diverse spectrum of EOC cell lines. We found that all cell lines responded to at least some miRNA mimics, but that the nature of the miRNA mimics provoking a response was highly selective within the panel. These selective toxicity profiles were leveraged to define modes of action and molecular response indicators for miRNA mimics with tumor-suppressive characteristics in vivo. A mechanistic principle emerging from this analysis was sensitivity of EOC to miRNA-mediated release of cell fate specification programs, loss of which may be a prerequisite for development of this disease. Synopsis A genomewide miRNA mimic toxicity screen indicates common and selective vulnerabilities of epithelial ovarian cancer cells. Follow-up analyses offer mechanistic insights into the selective sensitivity of ovarian cancer cells to select miRNAs. Screening the sensitivity of 16 ovarian cancer cell lines to 400 miRNA mimics reveals miRNAs with broad and selective effects. miR-181 and miR-155 are selectively toxic in chemoresistant ovarian cancer cells through dual modulation of TGFβ and AKT signaling. miR-517a targets a common vulnerability, primarily via its target ARCN1. miR-124 is selectively toxic, mainly by inducing terminal cell differentiation via its target SIX4. Introduction Epithelial ovarian cancer (EOC) is the most lethal gynecologic malignancy in the United States (Siegel et al, 2012). Recent advances in treatment of this disease have been limited to empirical optimization of chemotherapeutic agents and improved delivery of drugs (Armstrong et al, 2006). While these have yielded measurable improvements in overall survival of ovarian cancer patients, there is an urgent need for novel treatment modalities. A greater understanding of the linchpin biology of this disease would likely help provide inroads toward the development of new therapies. Multiple public and private efforts have focused on large-scale annotation of the landscape of genomic alterations associated with EOC. These studies have detected over 60 tumor-acquired mutations per patient. Though mutation of the p53 tumor suppressor has been identified as an almost universal characteristic (95% of ovarian tumors), all other somatic mutations have been found to occur in 3–6% of tumors or less (Kan et al, 2010; Cancer Genome Atlas Research, 2011). In combination with this diversity of somatic nucleotide variation, pervasive and recurrent copy number variation has been detected (Etemadmoghadam et al, 2009; Cancer Genome Atlas Research N, 2011), giving rise to the notion that ovarian tumor progression is driven by a "turbulent genome". The seemingly enormous diversity of molecular etiology of EOC poses a significant challenge to intervention target discovery and is fueling efforts to identify common biological vulnerabilities that occupy the nexus of diverse EOC genomes. A compelling candidate is defective miRNA biogenesis and function. When measured by quantitative PCR, Dicer and Drosha, the RNases required for miRNA processing, show decreased expression in over half of ovarian tumors sampled, and high Dicer expression correlates with remarkable patient survival (Merritt et al, 2008). However, we note that microarray-based tools have failed to uncover this relationship in larger cohorts (Gyorffy et al, 2013; Madden et al, 2014). Dicer is a haploinsufficient tumor suppressor in mice, and compound deletion of Dicer and the PTEN tumor suppressor is sufficient to induce spontaneous epithelial ovarian cancer (Kumar et al, 2009; Kim et al, 2012). Together, these observations indicate miRNA production may be generally deleterious to ovarian tumor initiation and/or progression, perhaps through translational suppression of tumor promoting gene products. Of note, the 3′-untranslated regions (3′-UTRs) of many mRNAs are clipped in some cancer cell types, which can release oncogenes from miRNA regulation (Mayr & Bartel, 2009). Thus, sensitivity to restoration of miRNA activity may represent a common vulnerability among diverse EOC oncogenotypes. To test the commonality of sensitivity to miRNA activity in EOC, we examined the consequence of introducing each of 400 miRNA mimics on the viability of each of 16 ovarian cancer cell lines, telomerase-immortalized ovarian surface epithelial cells, and hepatocytes. Selectively toxic mimics were recovered across the panel, the majority of which displayed highly individualized activity. Clinical correlations and mechanistic follow-up in multiple disease lineages indicated that the idiosyncratic miRNA mimic toxicity profiles were a consequence of fractional representation of biologically relevant ovarian cancer subtype/miRNA relationships that are more commonly encountered in other disease sites. Rare mimics targeting common vulnerabilities in the EOC panel corresponded to miR-517a and miR-124. miR-517a toxicity was primarily accounted for by its target ARCN1, a component of the COPI complex, and effectively impaired xenograft tumor growth when administered in vivo. miR-124 toxicity was primarily accounted for by its target SIX4, a homeobox transcription factor, which was also validated in vivo. A convergent mechanistic principle derived from this analysis was the common vulnerability of EOC to miRNA-mediated release of aberrant cell differentiation programs, loss of which may be a prerequisite for development of disease. Results A genomewide screen for miRNAs with antineoplastic potential in ovarian cancer For broad-scale interrogation of the selective consequences of gain-of-function microRNA activity, in ovarian cancer cell regulatory contexts, we combined a genome-scale synthetic miRNA collection with a panel of ovarian cancer cell lines representative of the genomic diversity found in this disease. The miRNA mimic collection corresponded to 400 unique human miRNA annotated in miRBase8-10 (Dataset EV1). As a test-bed within which to assess selective inhibition of cancer cell viability, we collected a panel of 16 ovarian tumor-derived cell lines together with non-tumorigenic telomerase-immortalized ovarian surface epithelial cells and spontaneously immortalized hepatocytes. This panel included commonly employed laboratory lines, a matched pair of chemoresponsive and chemoresistant lines from the same patient (PEO1, PEO4), and newly derived low-passage non-clonal cultures isolated from the malignant peritoneal effusions of 3 patients with high-grade serous papillary adenocarcinoma of the ovary (HCC5012, HCC5019, HCC5030, Table EV1). Each miRNA mimic was introduced into each cell line using optimized transient transfection protocols (Fig EV1A, Table EV1), and consequent effects on cell viability were measured 120 h later from biological triplicates. Standard deviation distributions indicated high reproducibility among biological triplicates across the cell line panel (Fig EV1B, black curve), and high phenotypic correlation among miRNA seed family members (Fig EV1B, red curve) relative to the total phenotypic variation (Fig EV1B, blue curve). Mean viability scores were normalized against position and batch effects and converted to z-scores to facilitate inter-line comparisons (Ho et al, 2012; Ward et al, 2012; Singh et al, 2013) (Dataset EV1). Affinity propagation clustering (APC) (Frey & Dueck, 2007; Witkiewicz et al, 2015) was used to delineate deterministic patterns of commonality among the miRNA mimic phenotypes across the cell line panel (Fig EV1C and Dataset EV11) and among the cell line responses to the miRNA mimic library (Fig EV1D). At least 50 phenotypic miRNA clusters were recovered which corresponded to five distinct cell line clusters (Fig EV1). APC of available whole-genome transcript profiles (Barretina et al, 2012) suggested at least 4 expression subtypes are present within the cell panel (Fig EV1E). However, these clusters had unimpressive correspondence to miRNA viability phenotype-based clusters (Fig EV1F) indicating global gene expression phenotypes, considered as a whole, did not specify selective response to the miRNA mimic library. Click here to expand this figure. Figure EV1. Primary screen data analytics Transfection conditions were optimized using normalized cell viability values upon transfection of mimic negative control and a pan-toxic siRNA, siUBB. Values plotted as the mean of three separate experiments with the bars representing the range. For each microRNA mimic, a standard deviation value was calculated for mean viability across the panel of 16 cell lines (blue curve). In addition, a within-miRNA seed family standard deviation value was calculated for mean viability across the cell panel (red curve). Lastly, a within-replicate standard deviation value was calculated (black curve). A kernel density estimation was fit to each of the three standard deviation distributions and plotted. The indicated microRNA mimics were clustered, using hierarchical APC, based on their viability z-scores across 16 cell lines. A Euclidean distance metric was used. Node colors indicate cluster membership. A high-resolution searchable and zoomable pdf has been provided as Dataset EV11 to facilitate visualization of cluster entities. The indicated ovarian cancer cell lines were clustered, using hierarchical APC, according to miR mimic viability z-scores (400 miRs). Node colors indicate cluster membership. RNASeq data for the indicated cell lines (acquired from the CCLE or internal analysis (this study)) was first filtered to select the top 20% of the most highly variant genes (2,686 genes total). The cell lines were then clustered using hierarchical APC based on a Euclidean distance metric. Node colors indicate cluster membership. Nodes in the gene expression-based APC from (E) were relabeled according to their cluster membership in the miR mimic response-based clusters from (D). Download figure Download PowerPoint A total of 108 miRNA mimics, corresponding to 94 unique mature microRNA sequences, reduced cell viability two standard deviations below the mean (z-score ≤ −2) in at least one cell line screened (Dataset EV1). Activity profiles, as visualized by two-way unsupervised hierarchical clustering, indicated a wide variation of selectivity patterns and potencies (Fig 1A). Of note, the most common miRNA mimic phenotype was idiosyncratic activity within the panel. About 80% of the miRNA mimics recovered in the screen significantly reduced the viability of only 1 or 2 cell lines, and no mimic reduced viability in more than 9 cell lines (Fig 1B). We considered the possibility that the selective activity profiles may be a consequence of fractional representation of biologically relevant ovarian cancer subtype/miRNA relationships within the test-bed, a consequence of artificial diversity from clonal genetic divergence in vitro, or a combination of the two. To help evaluate this, we first queried patient outcome data for the presence of significant clinical correlations to miRNAs with selective activity against the most resistant (and therefore least prone to noise from multiplicity of testing) cell line screened, SKOV3 (Fig EV2A). The expression of miRNAs corresponding to the top 5% of miRNA mimics with selective toxicity in SKOV3 (Dataset EV1) was evaluated in tumors from two independent ovarian cancer patient cohorts. We found that patients with higher expression of miR-146a and miR-505 have significantly increased overall survival with median overall survival times of 17.1 months and 10.4 months, respectively (Figs 1C and EV2D and E). The selective activity of miR-146a and miR-505 in SKOV3 was independent of transfection efficiencies or endogenous miR expression (Fig EV2B and C). Notably, recent studies indicate that both miR-146a and miR-505 have antitumorigenic activities in cell models of breast and lung cancer (Verduci et al, 2010; Yamamoto et al, 2011; Chen et al, 2013a). Next, we focused on miRNA mimics that selectively inhibited viability of the non-clonal short-term ascites-derived cultures (HCC5030, HCC5012, HCC5019), which should be least prone to in vitro genetic drift. We found significant enrichment (P < 0.05 by hypergeometric density distribution) of the miRNAs corresponding to these mimics among those miRNAs demonstrated to be downregulated in human serous ovarian tumors (Iorio et al, 2007) (Fig 1D). The clinical correlations of idiosyncratic hits, with patient prognosis or molecular features from patient samples, suggest they can reflect germane ovarian tumor biology. Figure 1. Public and private miRNA vulnerabilities among ovarian cancer cell lines Two-way unsupervised hierarchical cluster of z-score distributions by Euclidian distance. Any miRNA mimic with a z-score ≤ −2 in at least one cell line was included. Seed sequences of each miRNA mimic are on the right. Bars correspond to arbitrary cluster boundaries. Seeds in red correspond to miRNAs with decreased expression in serous, clear cell, or endometrioid ovarian cancer relative to normal tissue (Iorio et al, 2007). The histogram indicates the number of non-redundant hits binned according to the number of responsive cell lines. Expression of miRNAs miR-146a and miR-505 correlated with overall survival in ovarian cancer patients. The validation cohort (n = 150 samples) is shown. See Fig EV1 for the training cohort. Intersection of miRNA sensitivities in the non-clonal short-term cultures and miRNAs with reduced expression in serous ovarian cancer (Iorio et al, 2007). P-value from hypergeometric distribution. Download figure Download PowerPoint Click here to expand this figure. Figure EV2. Increased expression of miR-146a and miR-505 correlates with increased patient survival A Tukey plot of the range of z-scores for all mimics screened in each cell line screened. Z-scores are representative of the mean viability of 3 replicates. Quantitative PCR (qPCR) analysis of endogenous miR-146a expression revealed no association between level of expression in a cell line and toxicity in the screen (top panel). To facilitate line-to-line comparison, endogenous expression values were normalized to a reference cell line (Hey cells). qPCR showed similar levels of overexpression among all cell lines (bottom panel). Bars in red display the endogenous expression of miR-146a (normalized to RNU6B using the comparative CT method), while bars in black display the overexpression of miR-146a mimic after transfection (normalized similar to endogenous expression). qPCR analysis of endogenous miR-505 expression revealed no association between level of expression in a cell line and toxicity in the screen (top panel). To facilitate line-to-line comparison, endogenous expression values were normalized to a reference cell line (Hey cells). qPCR showed similar levels of overexpression among all cell lines (bottom panel). Endogenous and overexpression values displayed as in (B). Bars represent the mean of 2 replicates ± SD. Kaplan–Meier plots of training and validation sets of tumors using miRNA expression from both Agilent arrays and Illumina miRseq showed that higher expression of miR-505 correlated with increased patient survival. Kaplan–Meier plots of training and validation sets of tumors using miRNA expression from both Agilent arrays and Illumina miRseq showed that higher expression of miR-146a correlated with increased patient survival. Download figure Download PowerPoint miR-155 and miR-181b sensitivity is specified by intolerance to epithelial/mesenchymal transition To begin to define mechanisms underpinning selective sensitivity to miRNA mimic exposure, we first focused on the patient-matched PEO1 and PEO4 cell lines. Derived later in the patient's treatment, the PEO4 cell line is a model for recurrent, platinum-resistant EOC (Fig EV3A). Cytogenetic analyses indicate these cell lines descended from a common ancestor as opposed to arising through direct linear descent (Wolf et al, 1987; Cooke et al, 2010; Stronach et al, 2011). Thus, these two cell lines provide a unique opportunity to investigate acquired vulnerabilities within a model of a single patient's recurrent disease. A scatter plot of the z-scores of each mimic from the PEO1 and PEO4 viability screens revealed two remarkably distinct tails of activity predominantly corresponding to miRNA-induced inhibition of viability in only one line or the other (Fig EV3B). Click here to expand this figure. Figure EV3. Distinct miRNA sensitivities in patient-matched cell lines from recurrent disease A schematic depicting the derivation of the PEO1 and PEO4 cell lines from a patient with high-grade serous adenocarcinoma of the ovary. A scatter plot of z-scores corresponding to inhibition of viability in PEO1 versus PEO4 in response to each of 400 miRNA mimics. Values were derived from the mean of triplicate experiments. Mimics with a z-score < 2 were considered to have significantly reduced viability in a cell line. 17 mimics reduced PEO1 cell viability, 15 mimics reduced PEO4 viability, and 2 mimics reduced viability in both cell lines. The scatter plot indicates the expression value of each miRNA in PEO1 (y-axis) and the effect of each miRNA on cell viability in the same cell line as represented by z-score (x-axis). The scatter plot indicates the expression value of each miRNA in PEO4 (y-axis) and the effect of each miRNA on cell viability in the same cell line as represented by z-score (x-axis). The scatter plots indicate the expression values of each miRNA in PEO1 (x-axis) and PEO4 (y-axis) (left panel), and miRNA expression relative to that in HOSE (Normal) cells (right panel). Points are color-coded according to the corresponding miRNA mimic activity in the viability screens. PEO1-specific hit miR-210 continues to show a robust phenotype at even a 10-fold dilution while increasing the dosage 4-fold does not sensitize PEO4 cells, suggesting that selectivity was not due to overt dosage effects. Each data point represents the mean of 3 independent experiments ± SD. Source data are available online for this figure. Download figure Download PowerPoint To examine whether differential miRNA mimic responsiveness corresponded to differential expression of the corresponding endogenous miRNAs, global miRNA expression in PEO1 and PEO4 was quantitated by Illumina array relative to that observed in non-tumorigenic human ovarian surface epithelium (HOSE) cells (GEO Reference GSE67329). miRNA toxicity was not solely a function of its presence or absence in either cell line (Fig EV3C and D). Although numerous significant differences in miRNA expression between PEO1 and PEO4 were identified (Datasets EV2 and EV8), there was no detectable correlation with selective miRNA mimic viability phenotypes (Fig EV3E). This suggests that the specificity of mimic toxicity was not defined by the relative presence or absence of the corresponding endogenous miRNA. Gross miRNA mimic dosage effects were also unlikely to account for specificity, as selective responses to miR-210 were preserved across a 10-fold dose response curve (Fig EV3F). These cumulative observations suggest that distinct miRNA sensitivities are reflecting the presence of distinct acquired molecular vulnerabilities within the PEO1 and PEO4 regulatory frameworks. To help define the nature of these vulnerabilities, we used whole-exome hybridization-capture sequencing (70× average read depth) to estimate cell line-specific somatic mutations and genomic copy number variation together with RNAseq to quantitate relative mRNA expression profiles (Fig EV4A, Datasets EV3, EV4, and EV5, SRA Accession SRP065357). Due to the absence of patient-matched constitutional DNA, we filtered single nucleotide variation (SNV) calls through 16 normal human exomes to quell the detection of common germline polymorphisms to some extent. Gene-level copy number variation (CNV) was defined using exon read depth at each locus relative to a non-tumorigenic reference cell line. While clearly highly related at the genome level (437 shared SNVs and 2,817 shared focal copy number alterations), 879 cell line-specific SNVs were detected (519 in PEO1 and 360 in PEO4) together with extensive differences in copy number that closely correlated with mRNA expression. We next used these molecular annotations to help inform the biology underlying PEO4-specific miRNA vulnerabilities. Click here to expand this figure. Figure EV4. Genomic characterization of patient-matched cell lines from recurrent disease A summary of the PEO1 and PEO4 genomic characterization is displayed by circos plot. The outer track indicates relative expression of genes (the log2 of the ratio of (RPKM+1) values of PEO1 versus PEO4). Red peaks correspond to genes with greater than two-fold overexpression in PEO1 cells, and blue peaks correspond to genes with greater than two-fold overexpression in PEO4 cells. The next three tracks represent SNVs unique to PEO1 in red, unique to PEO4 in blue, and common in black. This is followed by an axis indicating chromosome position. The inner four tracks summarize CNV as indicated (variation is defined by a read depth ratio of ≥ 1.5 (gain) or ≤ 0.5 (loss) as compared to the reference cell line). See Datasets EV2, EV3, and EV4 for detailed annotations. Principal component analysis (PCA) plot of 3 PEO1 (blue squares) and 3 PEO4 (pink squares) replicates based on data from genes with RPKM value of one or greater. EV, Eigen value. Download figure Download PowerPoint miR-155 and miR-181b mimics were the top-ranked reagents that selectively reduced cell viability in PEO4 cells and were largely innocuo
Referência(s)