E-Cadherin/ROS1 Inhibitor Synthetic Lethality in Breast Cancer

Abstract

The cell adhesion glycoprotein E-cadherin (CDH1) is commonly inactivated in breast tumors. Precision medicine approaches that exploit this characteristic are not available. Using perturbation screens in breast tumor cells with CRISPR/Cas9-engineered CDH1 mutations, we identified synthetic lethality between E-cadherin deficiency and inhibition of the tyrosine kinase ROS1. Data from large-scale genetic screens in molecularly diverse breast tumor cell lines established that the E-cadherin/ROS1 synthetic lethality was not only robust in the face of considerable molecular heterogeneity but was also elicited with clinical ROS1 inhibitors, including foretinib and crizotinib. ROS1 inhibitors induced mitotic abnormalities and multinucleation in E-cadherin-defective cells, phenotypes associated with a defect in cytokinesis and aberrant p120 catenin phosphorylation and localization. In vivo, ROS1 inhibitors produced profound antitumor effects in multiple models of E-cadherin-defective breast cancer. These data therefore provide the preclinical rationale for assessing ROS1 inhibitors, such as the licensed drug crizotinib, in appropriately stratified patients.Significance: E-cadherin defects are common in breast cancer but are currently not targeted with a precision medicine approach. Our preclinical data indicate that licensed ROS1 inhibitors, including crizotinib, should be repurposed to target E-cadherin-defective breast cancers, thus providing the rationale for the assessment of these agents in molecularly stratified phase II clinical trials. Cancer Discov; 8(4); 498-515. ©2018 AACR.This article is highlighted in the In This Issue feature, p. 371.

Figure 1. ROS1 inhibition is synthetic lethal with E-cadherin defects in isogenic models.

A, Wild-type E-cadherin protein is lost in the CDH1 CRISPR-Cas9 mutagenised MCF7^A02 clone. Western blot illustrating E-cadherin expression in parental MCF7 cells (MCF7^parental) and the MCF7^A02 clone is shown. Upper band is a glycosylated isoform of E-cadherin, the lower band representing the non-glycosylated form. B, Confocal microscopy images of MCF7^parental and MCF7^A02 cells, illustrating loss of E-cadherin expression (green) in MCF7^A02 cells. Nuclei are imaged with DAPI (blue). C, Light microscopy images of MCF7^parental and MCF7^A02 cells, illustrating reduction in cell-cell contact in MCF7^A02 cells. D, Volcano plot illustrating AUC ratios (MCF7^A02 E-cadherin defective/MCF7 E-cadherin wild type parental cells) of 80 small molecule inhibitors assessed in the high-throughput screen. AUC ratio <1 indicates candidate E-cadherin synthetic lethal effects. Blue dots represent ROS1 inhibitors. E, Volcano plot of data from the siRNA SMARTpool sensitivity screens in cell lines described in (B). Blue dot highlights ROS1 siRNA identified in the screen as selectively targeting E-cadherin defective cells. F, ROS1 expression is upregulated in E-cadherin defective cells. Western blot showing expression of ROS1, MET and ALK in MCF10A CDH1^+/+ and MCF10A CDH1^−/− cells. ACTIN expression is used as loading control. G, Illustrative confocal microscopy images indicating an increase in ROS1 expression in E-cadherin defective MCF10A CDH1^−/− cells compared to MCF10A CDH1^+/+ cells. ROS1 expression is shown in red, E-cadherin in green and DNA in blue. H, Ectopic expression of E-cadherin in MCF7^A02 E-cadherin defective cells reduces ROS1 expression. Western blot illustrating ROS1 expression in E-cadherin defective MCF7^A02 cells transfected with FLAG epitope-CDH1 cDNA. ACTIN is used as loading control.

Figure 2. Validation of ROS1 synthetic lethality in E-cadherin defective isogenic models.

A, Western blot illustrating ROS1 silencing caused by four different ROS1 siRNAs (, , and 4) compared to two different non-targeting siRNAs (siCONT1, siCONT2). Uncropped western blot images are shown in Supplementary Fig. S20. B, and C, Bar charts illustrating cell inhibition caused by ROS1 siRNAs in MCF7^parental and MCF7^A02 cells (B) and MCF10A CDH1^+/+ and MCF10A CDH1^−/− cells (C). NPI = normalized percentage inhibition (compared to siCONT (NPI=1) and siPLK1 (NPI=0)). Error bars represent standard error of the mean (SEM) from three independent experiments. D, Bar chart illustrating the effect of 1 μM foretinib or crizotinib in MCF7^parental compared to two additional E-cadherin defective clones, MCF7^B04 and MCF7^B05. Error bars represent SEM from three independent experiments. E, Western blot illustrating E-cadherin expression in MCF7^A02 cell line model transfected with a FLAG epitope-CDH1 cDNA expression construct. F, Surviving fraction data from MCF7^A02, MCF7^parental and MCF7^{A02 +FLAG-CDH1} cell lines exposed to 1μM foretinib, crizotinib and TAE684. Cells were transfected with FLAG epitope-CDH1 cDNA and clones selected in G418. Clones expressing FLAG epitope-E-cadherin were exposed to 1 μM foretinib, crizotinib or TAE684 for six continuous days, at which point cell viability was assessed. *p<0.05 Student’s t-test as shown. G, Dot chart illustrating cell survival effects of ROS1/MET/ALK inhibitors in MCF7^parental and MCF7^A02 cells. SF₅₀ = surviving fraction 50 (concentration required to cause 50 % reduction in survival). H, and I, Dose-response survival curves in MCF10A CDH1^+/+ and MCF10A CDH1^−/− cells exposed to foretinib for six days (H) or two weeks (I). Error bars represent SEM from three independent experiments. In each case, ANOVA p value MCF10A CDH1^+/+ vs. MCF10A CDH1^−/− cells < 0.0001. Dose-response in SLC34A2-ROS1 translocation-positive HCC78 cells is shown as a positive control. J, Dose-response survival curves in MCF10A CDH1^+/+ and MCF10A CDH1^−/− cells exposed to ROS1 kinase inhibitor PF-06463922 for six days. Error bars represent SEM from three independent experiments. K, Dose-response survival curves in MCF7^A02 cells transfected with different concentrations of ROS1 siRNA SMARTpool and subsequently exposed to foretinib for six days. Increasing concentration of ROS1 siRNA caused a dose-dependent reduction in foretinib SF₅₀. Error bars represent SEM from three independent experiments. L, Expression of a crizotinib-refractory p.G2032R mutant ROS1 fusion cDNA (18) causes crizotinib resistance in E-cadherin defective MCF10A CDH1-/- cells. Cells were transfected with either a p.G2032R CD74-ROS1 cDNA expression vector or a cDNA expression vector without a CD74-ROS1 insert (“empty”). Twenty-four hours after transfection, cells were exposed to crizotinib for a subsequent six days, at which point cell viability was assessed. Data shows the median effects of three independent experiments. Error bars illustrate the SEM. ***p < 0.001 Student’s t test as shown.

Figure 3. Synthetic lethality of ROS1 inhibition in E-cadherin deficient breast tumour cell lines.

A, Western blot illustrating E-cadherin expression in 37 breast tumour cell lines (including MCF7 as positive control). Cell line names are colour-coded according to the presence of CDH1 gene mutations, gene deletion events or CDH1 promoter hypermethylation events. Uncropped western blot images are shown in Supplementary Fig. S20. B, Schematic illustrating breast tumour cell line cohort sizes used to interrogate siRNA sensitivity screens (19). C, Bar chart illustrating E-cadherin expression of a subset of the breast tumour cell lines shown in (A), determined by mass spectrometry, described in (20). Western blot classification of E-cadherin status from (A) is shown. D, Scatter plot illustrating CDH1 mRNA expression levels in breast tumour cell lines from the CCLE dataset (51). E, Volcano plot of data from the siRNA SMARTpool sensitivity screens described in (B). Blue dot highlights ROS1 siRNA identified in the screen as selectively targeting E-cadherin defective cells. F, REVEALER analysis identifies E-cadherin status as an important determinant of ROS1 siRNA sensitivity. Heatmap illustrates the REVEALER output identifying complementary genomic alterations associated with ROS1 siRNA sensitivity in the 34 breast tumour cell line panel described in (B). IC (information correlation coefficient) scores and nominal p values with respect to the target are shown on the right side of the heatmap. E-cadherin status demonstrated the most profound IC score -0.5 from 23 molecular features examined. G, and H, Box whisker plots illustrating foretinib (G) or crizotinib (H) sensitivity in 12 breast tumour cell lines, defined by Log₂ area under the curve (AUC) values. Individual AUC values are listed in Supplementary Table S8. Log₂ AUC values for SLC34A2-ROS1 translocation-positive HCC78 cells are shown as a positive control. ** p value = 0.003, Student’s t-test, * p value =0.035 Student’s t-test. I, Volcano plot of drug sensitivity effects in ex vivo cultured breast cancer explants (27) annotated according to CDH1 gene copy number. Red dot highlights crizotinib as selectively targeting explants with CDH1 gene copy number loss. Median AUC in explants with CDH1 copy number loss = 0.0713, median AUC in explants without CDH1 copy number loss = 0.138, p value =0.00127. Y-axis shows adjusted p-value (-log₁₀) from a t-test comparing AUC in ex vivo explant models with CDH1 loss vs. models with no CDH1 copy number change. X-axis shows the negative value of the t-statistic describing the difference in AUC means for ex vivo explant models with CDH1 loss vs. models with no CDH1 copy number change. Negative values suggest drug sensitivity in ex vivo explant models with CDH1 loss vs. models with no CDH1 copy number change.

Figure 4. E-cadherin/ROS1 synthetic lethality is independent of endocrine therapy resistance.

A, Dose-response survival curves in MCF7 and MCF7 LTED cells exposed to fulvestrant for five days. Error bars represent SEM from three independent experiments. B, and C, Dose-response survival curves in MCF7 LTED cells transfected with two independent E-cadherin (CDH1) siRNA reagents; siRNA 1 (B) and siRNA 2 (C) and a non-targeting siRNA (siCONT1) for 48 hours and subsequently exposed to foretinib or fulvestrant for five days. Error bars represent SEM from three independent experiments. ANOVA p value MCF7 LTED transfected with two independent E-cadherin siRNA reagents vs. MCF7 LTED transfected with siCONT1 cells p < 0.0001 for foretinib. No statistically significant difference was observed in the same cells exposed to fulvestrant. D, and E, Dose-response survival curves in cells as in (B) and (C) exposed to crizotinib or fulvestrant for five days. Error bars represent SEM from three independent experiments. ANOVA p value MCF7 LTED transfected with two independent E-cadherin siRNA reagents vs. MCF7 LTED transfected with siCONT1 cells p < 0.0001 for crizotinib. No statistically significant difference was observed in the same cells exposed to fulvestrant.

Figure 5. Cell cycle and mitotic defects in E-cadherin defective cells.

A, FACS plots illustrating increase in DNA content in E-cadherin defective MCF7^A02 cells exposed to foretinib. B, Enlarged image from (A) indicating >4n fraction. The fraction of cells with >4n DNA content is shown. C, Box whisker plots indicating fraction of cells with abnormal mitoses (e.g. multinuclear defects) in cells exposed to foretinib. *** p value = 0.0001, ** p value = 0.0054, Student’s t-test. Nuclear defects were visualized by confocal microscopy and a minimum of 200 cells were counted for each cell line. Data is representative of two independent experiments in each case. D, Illustrative confocal microscopy images indicating multinuclear phenotype in E-cadherin defective MCF7^A02 cells exposed to foretinib. E, Illustrative confocal microscopy images indicating multinuclear phenotype in E-cadherin defective SKBR3 and BT549 cells and E-cadherin wild-type T47D and SUM149 cells exposed to 1μM foretinib. F, Box whisker plots indicating fraction of cells with abnormal mitoses (e.g. multinuclear defects) in cells exposed to foretinib. * p-value =0.021, Student’s t-test. Data is representative of two independent experiments in each case. G, Western blot illustrating increased p21 levels in the E-cadherin defective MCF7^A02 cells exposed to 1μM foretinib vs. MCF7^parental cells. H, Bar chart illustrating increased caspase 3/7 activity in E-cadherin defective cells exposed to either crizotinib or foretinib. Median effects from three independent experiments are shown, with error bars representing the SEM. I, Bar chart illustrating increased caspase 3/7 activity in E-cadherin defective cells transfected with a siRNA targeting ROS1. siPLK1 was used as a positive control. p<0.05 between E-cadherin defective vs. wild-type groups, using the Student’s t-test. Median effects from three independent experiments are shown, with error bars representing the SEM. Bar chart illustrating increased caspase 3/7 activity in E-cadherin defective MCF7^A02 J BT549 K and SKBR3 L cells exposed to either crizotinib or foretinib, compared to E-cadherin wild type MCF7^Parental cells. * p < 0.01, ** p < 0.001 and *** p < 0.0001, E-cadherin defective vs. MCF7^Parental cells, Student’s t-test. Median effects from three independent experiments are shown, with error bars representing the SEM.

Figure 6. ROS1 inhibition exacerbates p120 catenin and cytokinesis defects in E-cadherin defective cells (previous page).

A, Time lapse microscopy images illustrating cell division in MCF7^Parental (top panel) and E-cadherin defective MCF7^A02 (bottom panel) cells exposed to foretinib, crizotinib or vehicle. MCF7^A02 and MCF7^Parental cells were first transfected with a mCherry-H2B plasmid, FACS sorted for mCherry-H2B to facilitate DNA visualization, and then exposed to foretinib, crizotinib or vehicle for a 24-hour period. Initial formation of the cleavage furrow, followed by formation of a multinuclear cell is highlighted with white arrows. Scale bar, 10 μm. B, Loss of E-cadherin is associated with a reduction in p120 levels and ROS1 inhibition causes a reduction in p120 tyrosine phosphorylation. Western blot showing phospho-p120 (Tyr228) and total p120 catenin levels in MCF10A CDH1^+/+ and MCF10A CDH1^−/− cells, exposed to a range of foretinib concentrations (vehicle, 0.03 μM, 0.1 μM, 0.3 μM or 1 μM) for 16 hours. C, ROS1 interacts with p120. Western blot illustrating co-immunoprecipitation of GFP-p120 and FLAG-ROS1 proteins in MCF7^A02 cells. D, Western blot illustrating p120 catenin silencing caused by two different p120 siRNAs (sip120_1 and _2) compared to a non-targeting siRNA (siCONT1). E, p120 or ROS1 siRNA causes mitotic defects in E-cadherin defective cells. Bar chart indicating fraction of cells with abnormal mitoses in cells transfected with siRNAs targeting p120 or ROS1 siRNAs. *** p value < 0.0001, ** p value < 0.001, Student’s t-test. A minimum of 100 cells were analyzed for each cell line. Data is representative of three replica experiments, error bars represent SEM. F, p120 gene silencing is synthetically lethal with E-cadherin deficiency. Bar chart illustrating cell inhibition caused by two different p120 siRNAs (1 and 2) in E-cadherin wild type MCF7^Parental and E-cadherin defective MCF7^A02 cells. NPI = normalized percentage inhibition (compared to non-targeting siRNA, siCONT (NPI=1) and cytotoxic siRNA targeting PLK1 (NPI=0)). Error bars represent standard error of the mean (SEM) from three independent experiments. *** p value < 0.0001, Student’s t-test. G, Representative images from a colony forming assay illustrating cell inhibition caused by two different p120 siRNAs (1 and 2) or ROS1 siRNA in E-cadherin wild type MCF7^Parental and E-cadherin defective MCF7^A02 cells. A non-targeting siRNA, siCONT and siPLK1 are used as controls.

Figure 7. E-cadherin synthetic lethal effects operate in vivo in E-cadherin defective breast tumours.

A, Therapeutic response to foretinib treatment in mice bearing E-cadherin deficient mammary tumours. Mammary tumour fragments from KEP mice were transplanted into 22 recipient mice; once tumours had established, animals were treated over a 27-day period with either drug vehicle or foretinib (25 or 50 mg/kg every other day, n=8 for vehicle-treated cohorts and n=7 for each drug treatment cohort). Tumour volumes after the initiation of treatment are shown. ANOVA p<0.0001 for both foretinib treatment regimes compared to vehicle-treated mice. B, Kaplan–Meier plot of data from (A) indicating anti-tumour efficacy of 25 mg/kg foretinib treatment. Mice were sacrificed once tumours reached a volume of 1500 mm³. C, Kaplan–Meier plot of data from (A) for vehicle vs. foretinib 50 mg/kg cohort. D, Immunohistochemistry images of tumours extracted from animals from (A) at the end of foretinib treatment are shown. Representative images of H&E, Ki67 and cleaved Caspase 3 are shown (magnification = 20x). Scale bar represents 250μm. E, Therapeutic response to crizotinib treatment in mice bearing KEP tumour allografts as in (A). ANOVA p<0.0001 for both crizotinib treatment regimes compared to vehicle-treated mice. F, Data from (E), plotted to illustrate tumour volume reduction in both crizotinib treated cohorts. G, Kaplan–Meier plot of data from (E), indicating effect of 25 mg/kg crizotinib treatment. H, Kaplan–Meier plot of data from (E) for vehicle vs. crizotinib 50 mg/kg cohort. I, Immunohistochemistry images illustrating lack of E-cadherin expression in a patient-derived xenograft (PDX) model of breast cancer (BCM2665) compared to positive (HCC1954 breast tumour cells) and negative controls (MDAMB231 breast tumour cells). J, Therapeutic response to foretinib treatment in mice bearing BCM2665 PDX. BCM2665 was transplanted into 19 recipient mice; once tumours had established, animals were treated over a 47-day period with either drug vehicle (n=11), or foretinib (25 mg/kg every other day, n= 8) as shown. ANOVA = p<0.0001. K, Kaplan–Meier plot of data from (I). L, Representative images of FFPE tumours from animals in (I) stained with H&E, Ki67 and cleaved caspase 3 are shown (magnification = 20x). Scale bar represents 250μm.