Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 9 (1), 4275

CRISPR Knockout Screening Identifies Combinatorial Drug Targets in Pancreatic Cancer and Models Cellular Drug Response

Affiliations

CRISPR Knockout Screening Identifies Combinatorial Drug Targets in Pancreatic Cancer and Models Cellular Drug Response

Karol Szlachta et al. Nat Commun.

Abstract

Predicting the response and identifying additional targets that will improve the efficacy of chemotherapy is a major goal in cancer research. Through large-scale in vivo and in vitro CRISPR knockout screens in pancreatic ductal adenocarcinoma cells, we identified genes whose genetic deletion or pharmacologic inhibition synergistically increase the cytotoxicity of MEK signaling inhibitors. Furthermore, we show that CRISPR viability scores combined with basal gene expression levels could model global cellular responses to the drug treatment. We develop drug response evaluation by in vivo CRISPR screening (DREBIC) method and validated its efficacy using large-scale experimental data from independent experiments. Comparative analyses demonstrate that DREBIC predicts drug response in cancer cells from a wide range of tissues with high accuracy and identifies therapeutic vulnerabilities of cancer-causing mutations to MEK inhibitors in various cancer types.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
In vivo CRISPR screening identifies novel drug targets. a The experimental outline of in vivo genetic knockout screening in PDX model of PDAC. b The heat map represents log-fold change of the top 2000 enriched and depleted sgRNAs in three separate trametinib-treated tumors relative to the mean of control-treated tumors. c Plot showing the distribution of normalized CRISPR viability scores (treatment vs control) for genes targeted by the sgRNA library. Shown are all targeted genes. Red dots indicate significantly depleted kinetochore genes as assessed by one-sided Kolmogorov–Smirnov test. d The plot showing the significance score (treatment vs control) of all gene depletion (minus log-transformed p values) as determined by Kolmogorov–Smirnov test. Kinetochore genes are labeled with red dots. e Gene ontology (GO) analysis depicts the significantly enriched GO terms for the top enriched and depleted gene sets (n = 100) from in vivo screening. Log-transformed p values are represented by bar plots
Fig. 2
Fig. 2
The validation of CRISPR screening hits. a, b Kaplan–Meier plots showing the survival analysis of TCGA data for pancreatic cancer patients expressing high levels of CENPE and RRM1 genes. High expression: >1 standard deviation (SD), and higher expression: >2 SD above the population mean. The number of patients in each group is indicated with n. c Box plots show fold change distribution of sgRNAs targeting CENPE (n = 10), RRM1 (n = 10), and control sgRNAs (n = 100) in tumors from control-treated and trametinib-treated mice. Bounds of the box spans from 25 to 75% percentile, center line represents median, and whiskers visualize 5 and 95% of the data points. Significance was assessed by Kolmogorov–Smirnov test. d MTT assay-based relative viability/growth inhibition levels are shown in PDX366 and mPanc96 cells expressing WT Cas9 with control sgRNA or two separate sgRNAs targeting CENPE (top) and RRM1 genes (bottom). e Relative viability/growth inhibition levels are shown for three different cell lines, PDX366, MPANC-96, and BxPC3 cells, treated with control, MEK inhibitor (trametinib), CENPE inhibitor (GSK923295), and combined MEK inhibitor and CENPE inhibitor. f RRM1 inhibitor (COH29) and trametinib were tested together in PDX366 and MPANC-96 cells. g Bar plots show relative growth inhibition of MPANC-96 cells treated with inhibitors of Aurora A/B (ZM447439) and PLK1 inhibitor (BI-2536) alone or in combination with trametinib. Each experiment was repeated at least three times and error bars represent the standard error of the mean value. CI cytotoxicity index, Comb combination of two drugs
Fig. 3
Fig. 3
Combinatorial MEK and CENPE inhibition results in synergistic cell death. a Each frame represents movie stills from time-lapse longer-term live-cell imaging as cells undergo mitosis starting from nuclear envelope break down (NEBD) to exit from mitosis. b Each bar graph show total time duration of each of the individually tracked cells spent in mitosis. The NEBD was taken as the beginning of mitosis. Individual cells (n = ~20/treatment group) were manually tracked from lime-lapse movies until they exit mitosis (anaphase) or died in mitosis for each of the treatment groups. c Flow cytometry profiles of DNA content in PDX366 PDAC cells treated with control, single-agent, or combinatorial MEK and CENPE inhibitors are shown. d Bar graphs represent treatment-mediated percent changes in cells with 2n (G1), 4n (G2), and >4n (polyploidy) DNA. e Immunofluorescent images of PDX366 PDAC cells treated with control and combinatorial (trametinib and CENPE inhibitors) are shown after DAPI and tubulin staining in the first two left columns respectively. f Representative hematoxylin and eosin (H&E)-stained tumor sections are shown. Tumors were harvested from mice that had undergone 72 h of treatment with 125 mg/kg CENPE inhibitor alone or in combination with 0.3 mg/kg trametinib. Arrows indicate mitotic figures. g Dot-plot showing the number of mitotic cells quantified from 10 different high-power imaging fields (HPFs) per tumor section. h Results show MRI-measured effects of single and combination of drugs on tumor formation. At 1 week after orthotopic implant, mice received control, MEK inhibitor (trametinib), CENPE inhibitor (GSK923295), or combined MEK and CENPE inhibitor. Results show MRI-measured tumor volumes after 4 weeks of treatment. i MRI-measured tumor volumes are shown in mice where tumors were allowed forming for 4 weeks and the treatments (as in e) were started. Beginning of treatment is marked with an arrow. Bounds of the box spans from 25 to 75% percentile, center line represents median, and whiskers visualize 5 and 95% of the data points. Individual data points are marked with circles or dots. Error bars represent the standard error of the mean value of at least three independent experiments. Symbols * and *** denote P < 0.05 and P < 0.001, respectively
Fig. 4
Fig. 4
Drug response evaluation by in vivo CRISPR Screening (DREBIC) approach. a The 450 CGP cancer cell lines were ranked according to the expression levels of the indicated genes. Box plots depict the distribution of log IC50 growth inhibition values of MEK inhibitor PD-0325901 in the high- (top quartile) and low-expressing cell lines (bottom quartile). Statistical significance was calculated by Kolmogorov–Smirnov test. Comparable results were obtained for other MEK inhibitors. b DREBIC integrates gene-specific CRISPR viability scores with basal expression levels to model drug response phenotype. c Box plot showing a significant (p < 0.0001, Kolmogorov–Smirnov test) growth inhibition (IC50) difference in response to PD-0325901 MEK inhibitor in ~450 cancer cell lines scoring high (top quartile) or low (bottom quartile) in DREBIC score analysis. d Receiver operation characteristics curves demonstrating true vs. false discovery rate of DREBIC prediction of cellular responses to three separate MEK inhibitors (PD-0325901 red, AZD6244 blue, and RDEA119 green). The random DREBIC score distribution was generated by 10,000 permutations of the same number of genes (Supplementary Figure 12). The average ROC curve of permutation analysis is shown in purple (AUC = 0.5) and the area corresponding to one standard deviation of prediction is marked in gray. e The box plots in the top panel are showing the distribution of tissue-specific DREBIC scores of CGP cell lines as ordered by ascending median DREBIC score. The lower panel box plots are showing log IC50 growth inhibition values of AZD6244 MEK inhibitor drug for the same cancer types. f The PD-0325901 MEK inhibitor drug response curves are shown for hematopoietic (Heme, red) and melanoma cancer cell lines (Skin, blue) from the CCLE data set. g The ROC curve shows the accuracy rate of etoposide-based DREBIC analysis of cellular response to three independent topoisomerase inhibitors from CPG data set. The CRISPR-gene viability scores for etoposide were obtained from a previously published in vitro CRISPR screening by Wang et al.. In the box plots, bounds of the box spans from 25 to 75% percentile, center line represents median, and whiskers visualize 5 and 95% of the data points. Symbol **** denotes P < 0.0001
Fig. 5
Fig. 5
DREBIC reveals genotype-specific MEK inhibitor sensitivity. a Box plot showing DREBIC scores for the responder and non-responder lung cancer cell lines to MEK inhibitor (PD-0325901). b Scatter plot showing a correlation between DREBIC score and log IC50 growth inhibition values of PD-0325901 MEK inhibitor. Cell lines with KRAS mutations are displayed as green dots and cells with RB1 mutations are displayed as magenta dots. c Box plots show the DREBIC score and log IC50 growth inhibition due to PD-0325901 treatment in KRAS WT (gray) and KRAS mutant (green) CGP cancer cell lines. The difference in drug response and DREBIC scores between these two groups has been defined as a difference between medians (∆DREBIC and ∆log IC50). d The same analysis as performed in (c) has been carried out for RB1 WT and RB1 mutant cells. e Bar plot shows the difference in DREBIC score (∆DREBIC) between WT cell lines and mutant cell lines with indicated genes. The significance of the difference was calculated by double-sided t-test and p values are shown with dots on the right side of the panel. Bars with statistically significant differences are marked with magenta and green. f Bar plot shows the difference in median log IC50 scores of PD-0325901 MEK inhibitor drug response (∆Log IC50) in WT cell lines and cells with mutations in the indicated genes. The significance of the difference has been calculated as in (e) and p values are shown with dots on the right side of the panel. Bars with statistically significant differences are marked with magenta for positive difference and green for the negative difference. g Box plots showing the DREBIC score and log IC50 growth inhibition due to PD-0325901 treatment in BRAF WT (gray) and mutant (green) CGP cell lines. Significance assessed by Kolmogorov–Smirnov test for all panels except (b). In the box plots, bounds of the box spans from 25 to 75% percentile, center line represents median, and whiskers visualize 5 and 95% of the data points. Significance is defined as: **p < 0.01, ***p < 0.001

Similar articles

See all similar articles

Cited by 4 PubMed Central articles

References

    1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2016. CA Cancer J. Clin. 2016;66:7–30. doi: 10.3322/caac.21332. - DOI - PubMed
    1. Kamisawa T, Wood LD, Itoi T, Takaori K. Pancreatic cancer. Lancet. 2016;388:73–85. doi: 10.1016/S0140-6736(16)00141-0. - DOI - PubMed
    1. Rahib L, et al. Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States. Cancer Res. 2014;74:2913–2921. doi: 10.1158/0008-5472.CAN-14-0155. - DOI - PubMed
    1. Biankin AV, et al. Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes. Nature. 2012;491:399–405. doi: 10.1038/nature11547. - DOI - PMC - PubMed
    1. Jones S, et al. Core signaling pathways in human pancreatic cancers revealed by global genomic analyses. Science. 2008;321:1801–1806. doi: 10.1126/science.1164368. - DOI - PMC - PubMed

Publication types

MeSH terms

Substances

Feedback