Figures and data
Patterns of CDK12 genomic alterations and tandem duplication phenotype
(a) The number of samples harboring a CDK12 genomic alteration are shown for four CRPC cohorts. CDK12 alterations that lead to bi-allelic loss (CDK12BAL): homozygous deletion, 2 or more (2+) small mutations (SNV/INDEL), or 1 mutation plus a loss of heterozygosity (LOH) event. CDK12 alterations that lead to monoallelic loss (CDK12MAL): 1 small mutation or hemizygous deletion leading to LOH. Deletion was determined based on copy loss relative to the tumor ploidy. The number of CDK12-altered samples with concomitant tandem duplicator phenotype (TDP) is shown in colors; grey indicates no TDP was observed; white indicates sample was unevaluable for TDP (see Methods).The total number and percentage of samples with CDK12 alterations are shown in parentheses. Samples with less than 20% estimated tumor cellularity are excluded. UW, University of Washington Rapid Autopsy; SU2C-I, Stand-Up-2-Cancer International Dream Team; SU2C-WC, Stand-Up-2-Cancer West Coast Dream Team; HMF, Hartwig Medical Foundation.
(b) Nonsynonymous mutations in CDK12. Counts of missense (green) and truncating (black) mutations are shown on top and bottom, respectively. Mutations included are from UW, SU2C-I, SU2CWC, HMF cohorts (n=832) as well as panel based genomic testing (n=46). A total of 135 SNVs/INDELs in CDK12 are present in 89 patient samples.
(c) LOH Score for UW, SU2C-I, SU2C-WC, and HMF cohorts comparing tumors with biallelic loss of BRCA2, CDK12, TP53, or no alterations in these genes (and in BRCA1, CHD1, PALB2) for both alleles (All Intact). The LOH Score was defined as the proportion of the genome altered by copy number segments having zero minor copy number due to a deletion event but excluding aneuploidy involving whole or arm chromosome events. For each grouping, samples included are those with biallelic loss status of the specified gene only with absence of mutations from the other groups, or other recurrently mutated HR genes: BRCA1, CHD1, PALB2 or genes of other groupings. ‘All Intact’ group consists of samples that do not have biallelic loss in any of BRCA2, CDK12, TP53, BRCA1, CHD1 and PALB2. Wilcoxon rank-sum test p-value shown.
(d) Mutational signature 3 (SBS3) that is associated with homologous recombination deficiency (COSMIC v3.4) is shown for UW, SU2C-I, SU2C-WC and HMF cohorts comparing between biallelic loss of BRCA2, CDK12, TP53, or no alterations in these genes (and in BRCA1, CHD1, PALB2) for both alleles (All Intact). Presence of SBS3 requires greater than 0.05 proportion of SNVs assigned. Mann-Whitney U test p-value shown. The samples included in the groups were selected based on the same criteria as in (c).
(e) The proportion of CDK12 alterations for 46 samples determined to exhibit a TDP across four CRPC cohorts. Definitions for biallelic and monoallelic loss of CDK12 are same as in (a); intact refers to wildtype for both alleles. The number of samples harboring each CDK12 alteration category is indicated.
(f) Frequency of alteration by tandem duplications (TDs) in the 46 TDP cases. TDs were defined as simple duplication events that had flanking regions of lower copy number and length less than 10 mega base-pairs (see Methods). (Top) For each 50 kb genomic bins, the proportion of samples overlapping TDs are shown for TDP+ (grey) and TDP-cases (black). Regions with significant enrichment of TD overlap by ξ2-test of independence (red, Bonferroni adjusted p < 0.001) spanning a total of 35.8 Mb across the genome. (Middle) Cancer Gene Census oncogenes and/or fusions whereby the gene, based on its start and end coordinates, is fully contained within a TD event. Genes with significant enrichment for being altered by TDs in TDP+ samples relative to TDP-samples by Fisher’s exact test (Bonferroni adjusted p-value < 0.01) are shown in red. (Bottom) Cancer Gene Census tumor suppressor genes and/or fusions whereby the gene is transected (broken) by either side of the TD boundary or both. Genes with significant enrichment for being altered by TDs in TDP+ samples relative to TDP-samples by Fisher’s exact test (Bonferroni adjusted p-value < 0.01) are shown in blue.
(g-h) Genes enriched for being altered by TDs in TDP+ (n=46) versus TDP-(n=775) tumors. Shown are 433 COSMIC Cancer Gene Census oncogenes and fusions (g) and 390 tumor suppressor genes and fusions (h). Genes with -log10(p-value) > 5.5 and log odds ratio > 2 are shown in red for oncogenes and blue for tumor suppressors. Higher log odds ratio indicates that the proportion of the TD events altering a gene higher in TDP+ cases. Definitions of gene and TD event overlap same as in (f).
(i) AR absolute copy number distribution between cases with TDP+ (n=39) and TDP- (n=730) tumors. Mann-Whitney U test p-value is shown.
(j) The proportions of BRCA2, TP53, and PTEN genomic alteration status (MAL, BAL or intact), for the cases with CDK12BAL (n=39). Fisher’s exact test was used to compare the proportions in CDK12BAL with CDK12MAL or intact cases.
The transcriptional phenotype of mCRPCs with CDK12 loss
(a) Copy number profiles for UW rapid autopsy patient 05-217 that exhibits the TDP and has CDK12BAL. Copy number alteration events of deletions (1 copy; green), copy neutral (2 copies; blue), copy gains (3 copies; dark red), and amplifications (4+ copies; red) were predicted using TITAN. Copy number values (y-axis) is shown after tumor purity and ploidy correction.
(b) Relationship between metastatic tumors from patient 05-217 based on co-occurrence of tandem duplications (TDs). Red arrows indicate unique or shared TDs between samples. The display and orientation of the four metastases were determined by phylogenetic analysis using neighbor joining tree estimation (ape_R_package) for the presence or absence of TD events using Manhattan distance and rooted tree configuration.
(c) Intra-patient concordance in AR active prostate cancer phenotype classification in two metastases with CDK12BAL and a tandem duplication genome. GSVA signature scores and log2 FPKM values are colored according to scales shown on plot. AR: androgen receptor, NE neuroendocrine, EMT: epithelial-mesenchymal transition, CCP: cell cycle progression.
(d) RNAseq based comparison of transcript abundance differences in mCRPC tumors with and without CDK12BAL (log2 FC and p<0.05 across 2 or more cohorts). The top 20 up and down regulated gene symbols are listed
(e-h) RNAseq based quantitation of AR and NE (neuroendocrine) pathway activity, cell cycle progression score (CCP) and levels of transcripts encoding the ARv7 splice variant comparing mCRPCs with versus without CDK12BAL (GSVA scores; Wilcoxon rank test p-values shown).
(i) Quantitation of expressed gene fusions in mCRPCs without vs with CDK12BAL. (Wilcoxon rank test p-values shown).
(j) Assessment of recurrence of CDK12BAL-specific expressed gene fusions within and across mCRPC cohorts.
(k) Expression of COSMIC oncogenes and tumor suppressor genes contained within tandem duplications. * = Up-regulated or ^ down-regulated with p<0.05, FC>abs(2) in at least 1 cohort.
(l) Hallmark Pathway enrichment in mCRPC without vs with CDK12BAL (pathways with FDR<0.05 in at least 1 cohort shown.)
(m) KEGG Pathway enrichment in mCRPC without vs with CDK12BAL (pathways with FDR<0.05 in at least 1 cohort shown.)
Assessment of transcript characteristics in the context of CDK12 loss
(a) Analysis of transcript abundance levels determined by RNAseq comparing expression in tumors with CDK12BAL vs tumors with intact CDK12 based on gene length (genome span). The graph represents the distribution of up-regulated or unchanged genes (red) and genes with lower transcript abundance (P<0.05; FC<-2) (blue) in CDK12BAL vs tumors with intact CDK12. The human gene length determination were assessed and compared using Ensemble genes with the ShinyGO tool (t-test p-value shown on plot). The y-axis shows density and the x-axis is gene length base pairs (bp).
(b) Association of increased/unchanged or decreased (p<0.05; FC<-2) transcript abundance levels in relation to gene size features in tumors without or with CDK12BAL in the SU2C-I cohort as assessed in (a).
(c-d) APAlyzer analysis for up- and down-regulated intronic alternative polyadenylation usage (APA) using RNA-seq from data sets comparing CDK12BALcases vs CDK12-intact controls: University of Washington Autopsy study (UW), and (d) StandUp2Cancer (SU2C-WC). RED = relative expression difference; each point is a different APA site
(e) Comparison of APA usage for two different IPA sites in the ATM gene in UW and SU2C-I cohorts- without vs with CDK12BAL (Wilcoxon rank test p-values shown).
(f) Transcript pile-up of reads mapping to exons demonstrating increased transcript reads corresponding to an IPA in the ATM gene in SU2C-I mCRPC tumors with CDK12BAL versus tumors with intact CDK12 and diminished transcripts mapping to the distal 3’ exon.
(g) RNAseq based transcript abundance measurements of genes involved in DNA repair, comparing mCRPCs with versus without CDK12BAL. * = Down-regulated with p<0.05, FC>abs(2) in at least 1 cohort.
(h-l) RNAseq based transcript abundance levels for specific genes involved in the HR DNA repair pathway (Wilcoxon rank test p-values shown).
Acute CDK12 inhibition increases alternate polyadenylation usage and downregulation of DNA repair genes
(a) Genes downregulated by acute CDK12 inhibition skew longer. Distribution by gene length (bp) of downregulated genes by RNA-seq (<-2 fold, FDR <0.05, ‘n’ depicted on plot) in LNCaP and LuCAP35_CL prostate cancer cell lines following six hours of exposure to vehicle (DMSO), CDK4/6 inhibitor palbociclib (Palbo, 10uM), broad RNA Pol-II inhibitor actinomycin D (ActD, 5ug/mL), or CDK12/13 inhibitor SR4835 (200nM) (n=3). Plots were made with ShinyGO 0.80 (87) and show significance (t-test) of downregulated vs upregulated (Up) or no change (NC) in expression genes.
(b) Acute CDK12 inhibition increases intronic alternate polyadenylation site usage. APAlyzer analysis of the RNA-seq data showing treatment effect on intronic APA usage (compared to vehicle). Colored dots show significant (p<0.05) up and downregulated relative expression difference of APA sites. Numbers inside plots indicate the number of significant up (red) or down (blue) APA sites.
(c) Treatment with the CDK12/13 inhibitor SR4835 alters gene expression. Heatmap showing log2(fold) color-coded expression and genes with >2 fold change up or down (FDR<0.05) upon SR4835 treatment with top and bottom 20 labeled.
(d) Multiple HR genes are downregulated with SR4835 treatment. Heatmap as in (c) showing expression of HR-related genes upon acute CDK12 inhibition. The * indicates genes with any negative FC at FDR<0.05 and ^ indicates FC<-2 at FDR<0.05.
(e) Some, but not all, HR gene downregulation with SR4825 is due to cell arrest. Example genes from the LNCaP RNA-seq showing the magnitude of Palbo vs SR4835 effects on RNA expression (log2(fold) RPKM inhibitor vs DMSO). Significance determined by two-way ANOVA with Dunnett multiple testing correction.
(f) Acute CDK12 inhibition decreases BRCA2 protein and induces apoptosis after 24h. Western blot of 22Rv1 cells treated 6, 24, or 48 hours with vehicle (DMSO), SR4835 (200nM), or SR4835 plus pancaspase inhibitor Z-VAD (50µM). Lysates were probed for key DNA repair genes and markers of DNA damage (γH2A.X) and apoptosis (total/cleaved PARP1, cleaved caspase 3). Figure is a composite from two gels run with identical lysates (3-8% tris-acetate for BRCA2, ATM, ATR, Vinculin and 4-12% bistris for RAD51, γH2A.X, PARP1, and CC3).
(g) Long genes show 5’/3’ transcript imbalance upon acute CDK12 inhibition. qPCR with same RNA samples as above (a-e) using two sets of primers for each target to show selective loss of 3’ transcript in long DNA repair-associated genes plus Vinculin/VCL, a long gene not involved in DNA repair. Plots show mean log2(fold) vs DMSO -/+stdev (n=3) with significance between 5’/3’ primers determined by two-way ANOVA.
Cells adapted to CDK12 loss do not show dramatic HR gene downregulation and retain an intact HR pathway.
(a) CDK12BAL cells have mostly restored 5’/3’ transcript balance. qPCR with CDK12-intact (LuCaP35_CL; LuCaP35CR_CL; LuCaP173.1_CL) and CDK12BAL (LuCaP189.4_CL) prostate cancer cell lines using 5’ and 3’ targeting primers. Plot shows mean-centered log2(fold) -/+stdev (n=3), with significance determined by two-way ANOVA.
(b) Cells with stable CDK12BAL show normal expression of key DNA repair genes. Immunoblot of factors involved in homology directed repair in prostate and ovarian cancer models with intact or absent CDK12. Figure is a composite from two gels run with identical lysates.
(c) 22Rv1-CDK12-KO clones retain ATM 3’ loss but restore BRCA1 and BRCA2 transcript balance. qPCR as in (a) comparing 22Rv1 vs CDK12-KO clones. Mean log2(fold) -/+stdev (n=4) vs parental line with significance determined by two-way ANOVA.
(d) 22Rv1 CDK12 KO clones do not show selective long gene downregulation. Same analysis as in Fig. 4a. Plots show distribution by gene length of downregulated vs upregulated/unchanged genes in the indicated CRISPR KO clone line vs parental line. Significance was determined by t-test and ‘n’ is depicted on plots.
(e) 22Rv1-CDK12-KO clones do not show increased APA usage. APAlyzer analysis (as in Fig. 4b) showing APA usage differences in 22v1-CDK12-KO5 and Skov3-CDK12-KO1 compared to parental cells. APA counts from 22Rv1-CDK12-KO2 and LuCaP189.4-CDK12 lines can be found in Fig. S5a.
(f) CDK12 loss does not impair RAD51 foci formation. 22Rv1 cells with Tet-inducible shBRCA2 were treated -/+ doxycycline (100ng/mL) for 4 days then exposed to 6Gy ionizing radiation (IR) and fixed at 3h post IR. The same was done with 22Rv1 CDK12 KO clones. Immunofluorescence staining was performed for γH2A.X and RAD51 and images were acquired by confocal microscopy. Left: representative images (white: DAPI, green: γH2A.X, purple: RAD51). Right: quantification of images (∼200-500 cells analyzed per treatment). Line is at mean and significance was determined by one-way ANOVA (Kruskal-Wallis).
(g) Alternate quantification of (f) and S5c showing the percent of cells with 5 or more RAD51 foci before and after irradiation (IR).
(h-i) CDK12 loss does not confer platinum or PARPi sensitivity. Dose response curves for prostate cancer cell lines treated 8 days with carboplatin (h) or olaparib (i) (n=3, mean-/+stdev). UWB1.289 is a control BRCA1mut ovarian cancer line. Y-axis shows relative viability vs no drug and x-axis shows concentration (moles/L).
Prostate cancers with CDK12BAL are sensitive to CDK13 loss and therapeutics targeting transcription
(a) CDK12 loss is generally detrimental to cells. Selected control and CDK12-related sgRNA fitness results from DepMap for relevant cell lines and average score across cancer type. RB1 is a control for positive selection, MAPK12 (p38γ) is a control for neutral selection, and EIF3A is a control for negative selection.
(b) LuCaP189.4_CL growth is greatly impaired by sgCDK13. GFP-tagged LuCaP 189.4_CL were transduced with CRISPR/Cas9 vectors containing dual sgRNAs against AAVS1 (neg. control), CDK12, or CDK13 and growth was monitored by GFP imaging starting on day 7 (n=5). Plot shows mean confluence (%GFP -/+stdev, n=5) with significance vs sgAAVS1 determined by two-way ANOVA. Images below graph show example microscopy from days 7, 14, and 21.
(c-e) CDK12 KO is poorly tolerated in LNCaP and C4-2B cells. GFP-tagged LNCaP and C4-2B were transduced with dual sgRNA vectors and monitored as in (b). (c) Example images showing sgRNA effect on cell growth and growth rate plots (mean confluence (%GFP) -/+stdev, n=5) over time for LNCaP (d) and C4-2B (e) with significance vs sgAAVS1 determined by two-way ANOVA.
(f-g) CDK13 sgRNA is detrimental in Skov3 lacking CDK12. Skov3 (f) or Skov3-CDK12-KO1 (g) were transduced with sgRNAs and monitored as in (c-e). Plot shows mean confluence (%GFP) -/+stdev, n=5) with significance vs sgAAVS1 determined by two-way ANOVA.
(h-i) CDK12 loss corresponds with sensitivity to CDK13 inhibition. Plots show dose response curves for 22Rv1 or 22Rv1-CDK12-KO clones and 189.4 empty vector (189.4-vec) or CDK12-transduced (189.4-CDK12) treated four days with SR4835 (h) or THZ531 (i) CDK12/13 inhibitors. Y-axis shows relative viability vs no drug and x-axis shows concentration (moles/L).
(j) In vivo LuCaP189.4 tumors respond to SR4835. PDX LuCaP lines were treated 28 days with vehicle or SR4835 (20mg/kg, 5 days on, 2 days off). Plots show number of mice per group, tumor volume (mean with 95%CI) over time, and final tumor weight. Significance was determined by Kolmogorov-Smirnov test (tumor volume) or unpaired, two-tailed t-test (tumor weight).
(k) Three day SR4835 treatment in vivo increases APA usage, but the effect is absent in 28 day tumors. APAlyzer analysis (as in Fig. 4b) on RNA-seq from the treated PDX tumors harvested in (j) and LuCaP189.4 treated 72h (n=3 tumors per treatment).
(l) Heatmap showing mean log2(fold) expression of genes involved in HR determined by RNAseq from tumors resected after 3d or 28d treatment with SR4835 compared to vehicle, as in (k).
(m) CDK12 loss increases sensitivity to α-amanitin. 22Rv1 and LuCaP189.4_CL were treated four days with α-amanitin to generate dose response curves, as in (h,i). Plots shows mean-/+stdev(n=4) and table shows effective concentration 50% viability (EC50) values and legend.
CDK12 genomic alterations and tandem duplication phenotypes
(a) Tandem duplication (TD) size analysis confirms multiple modes of TD lengths (top) in whole exome (WES; orange) and whole genome sequencing (WGS; blue) data for 46 tumors with a TDP. Fitting mixtures of Gaussian distributions to each of 26 WES and 20 WGS samples provided estimates of Gaussian means. The distribution of these mean values across 26 WES cases and 20 WGS cases are shown in red dotted lines; samples may have multiple mean values, one for each estimated mixture. The mixture means are categorized into six possible TD Groups based on the length ranges (bottom). The proportion of each TD Group are shown for the four CRPC cohorts.
(b) The proportions of CDK12, BRCA2, and TP53 genomic alteration status, including mono-allelic and bi-allelic losses and intact, for the 46 TDP cases compared across TD Groups.
Gene expression alterations in CDK12BAL prostate cancers.
(a) GSEA of CDK12-mut up and down signatures reported in Wu et al., 2018 applied to mCRPC cohorts with versus without CDK12BAL. (Normalized enrichment scores (NES) shown in heatmap on plot, all enrichment false discovery rates (FDR) <0.0001).
(b-c) Overlap of differentially expressed genes up-regulated (c) or down-regulated (d) across mCRPC cohorts with versus without CDK12BAL (significance level shown on plot.)
(d-g) RNAseq based quantitation of AR pathway activity in mCRPC cohorts with versus without CDK12BAL (GSVA scores; Wilcoxon rank test p-values shown).
(h) Differential expression of genes contained within CDK12BAL-specific and recurrent fusions. * = Up-regulated or ^ down-regulated with p<0.05, FC>abs(2) in at least 1 cohort.
(i-l) RNAseq based transcript abundance levels of ATM in mCRPC cohorts with versus without CDK12BAL (Wilcoxon rank test p-values shown).
APA usage in CDK12BAL prostate cancers.
(a-c) Association of increased/unchanged or decreased (p<0.05; FC<-2) transcript abundance levels in relation to gene size features in tumors without or with CDK12BAL in the HMF, SU2C-WC and UW mCRPC cohorts.
(d-f) APAlyzer analysis for up- and down-regulated APA usage using RNA-seq from mCRPC data sets comparing CDK12BAL cases vs CDK12(intact) controls: (c) TCGA-PRAD, (d) SU2C-I, and (e) HMF. RED = relative expression difference; each point is a different APA site.
(g) Transcript pile-up of reads mapping to exons demonstrating increased transcript reads corresponding to an IPA in the ATM gene in TCGA prostate (PRAD) and ovarian primary tumors with CDK12 alterations versus tumors with intact CDK12 and diminished transcripts mapping to the distal 3’ exon.
Effects of acute CDK12 inhibition.
(a) Treatment with ActD, Palbo, and SR4835 all lead to downregulation of longer transcripts. Distribution by transcript length (nucleotides/nt) of downregulated genes by RNA-seq (<-2 fold, FDR <0.05, ‘n’ depicted on plot) in LNCaP and LuCaP35_CL prostate cancer cell lines following six hours of exposure to vehicle (DMSO), CDK4/6 inhibitor palbociclib (Palbo, 10uM), broad RNA Pol-II inhibitor actinomycin D (ActD, 5ug/mL), or CDK12/13 inhibitor SR4835 (200nM) (n=3). Plots were made with ShinyGO 0.80 (76) and show significance (t-test) of downregulated vs upregulated or unchanged genes.
(b) CDK12 inhibition increases the number and ratio of upregulated APAs. Table with the number of APA sites down (DN), no change (NC), or up (UP) upon treatment vs vehicle and the ratio UP/DOWN, indicating any skew towards the IPA phenotype.
(c-d) Cell arrest and CDK12 inhibition both lead to HR pathway downregulation. (c) Selected KEGG pathway enrichment for DNA-repair related pathways focusing on changes from general G1/S arrest (palbociclib) vs CDK12 inhibition (SR4835). (d) Many SR4835 downregulated pathways overlap with cell cycle arrest. Venn diagram showing high overlap of KEGG pathways downregulated (NES <-1) with acute palbociclib or SR4835 treatment (6h).
(e-f) Effects of pharmacological CDK12/13 inhibition. Similar experiments as in Fig. 4e using LNCaP
(e) and Skov3 (f) cells showing the effect on DNA repair and apoptotic proteins with SR4835 treatment with or without Z-VAD caspase inhibitor.
(g) LuCaP189.4 carries bi-allelic loss of function CDK12 mutations. Diagram showing the two frame- shift mutations carried by the LuCaP189.4 PDX upstream of the key functional kinase domain (yellow).
(h) LuCaP 189.4 does not express CDK12 protein. Immunohistochemical (IHC) staining (with amplification) for CDK12 on FFPE sections from LuCaP PDX tumors or cell spots from LuCaP189.4_CL and LuCaP189.4-CDK12 cell lines. Counterstained with hematoxylin.
(i) LuCaP189.4 shows hallmark TDP genomic pattern. Copy number plot from exome-seq of three LuCaP PDX lines (78CR, 174.1, and 189.4).
(j) CDK12 knockout lines grow slower than parental lines. GFP tagged cells were grown for seven days and monitored by GFP imaging. Graphs show GFP confluence normalized to day 1 for each line with mean-/+stdev (n=5). Significance vs parental line was determined by two way ANOVA.
Assessments of HR competency in cells with CDK12 loss.
(a) Stable CDK12(-) cells show fewer upregulated intronic APAs. Table with the number of APA sites down (DN), no change (NC), or up (UP) in isogenic paired models (CRISPR KO clones vs parental, or 189.4-CDK12 vs 189.4-vec). The ratio UP/DOWN indicates skew towards the IPA phenotype.
(b) Validation of Tet-shRNA lines. Western blot with lysates from Tet-shRNA lines treated four days -/+ 100ng/mL doxycycline.
(c) LuCaP189.4_CL cells are RAD51 competent. Irradiation and immunostaining (same as in Fig 5f). Cells were exposed to 6Gy IR and fixed at 3h. Immunofluorescence staining was performed for γH2A.X and RAD51 and images were acquired by confocal microscopy. Left: representative images (white: DAPI, green: γH2A.X, purple: RAD51). Right: quantification of images (∼200-500 cells analyzed per treatment). Line is at mean and significance was determined by unpaired t-test (Mann-Whitney).
(d-e) CDK12 knockdown does not prevent RAD51 foci. Additional immunostaining (same as in Fig 5f) using LNCaP (d) and Skov3 (e) with Tet-shCDK12 or Tet-shBRCA2. Cells were treated four days -/+ dox. Graphs show mean with significance determined by one-way ANOVA (Kruskal-Wallis).
Drug sensitivities in prostate cancers with CDK12 loss.
(a-b) CDK12 loss does not confer HRd-expected platinum or PARPi sensitivity. Dose response curves for prostate cancer (a) or ovarian cancer (b) cell lines treated 8 days with carboplatin (n=3 for prostate lines, n=4 for ovarian lines). EC50 values are shown on the legend.
(c) Ovarian cancer cells (n=4) were treated 12 days with olaparib.
(d) Prostate cancer lines and UWB1.289 (n=3) were treated 12 days with olaparib.
(e) Prostate cancer lines and UWB1.289 (n=3) were treated 8 days with rucaparib.
(f-g) LuCaP189.4 organoids do not show obvious PARPi sensitivity. LuCaP PDX tumors were dissociated into organoids and treated (at passage 3) with olaparib (f) or rucaparib (g) for 14 days. Plots show mean-/+stdev (n=4).
(h-j) CDK13 sgRNA is detrimental in 22Rv1 lacking CDK12. Experiment as in Fig 6f. GFP tagged 22Rv1 or 22Rv1-CDK12-KO5 were transduced with sgRNAs and monitored by imaging. Example images are in (h). Plots show growth rates for 22Rv1 (i) and 22Rv1-CDK12-KO5 (j) by confluence (%GFP- -/+stdev, n=5) with significance vs sgAAVS1 determined by two-way ANOVA.
(k-m) LuCaP189.4_CL shows sensitivity to CDK13 inhibition. Dose response curves from four prostate cancer lines, including CDK12BAL LuCaP189.4_CL, treated six days with the CDK12/13 inhibitors SR4835 (k) or THZ531 (l) and a table (m) with the calculated EC50 values.
(n) APAlyzer analysis from SR4835 treated PDX tumors. Three tumors of each group from Fig. 6j plus
3 day treated LuCaP189.4 were analyzed by RNA-seq. The table shows the number of intronic APA sites down (DN), no change (NC), or up (UP) in treated vs vehicle comparisons. The ratio UP/DOWN indicates skew towards the IPA phenotype.
Effects of WEE1, ATR and CHEK1 inhibitors toward prostate cancers with CDK12 loss.
(a-c) One CDK12-KO prostate line, 22Rv1-CDK12-KO5, shows increased sensitivity to WEE1 inhibition. Prostate cancer lines were treated with WEE1 inhibitors adavosertib/MK-1775 (a) and PD0166285
(b) for four days. Dose response curves show relative viability vs drug concentration. Plots show mean-/+stdev (n=4). Legend and EC50 values (μM) are shown in (c).
(d-f) CDK12-KO prostate lines do not show sensitivity to ATR inhibitors. Prostate cancer lines were treated with ATR inhibitors berzosertib/VX-970 (d) and elimusertib/BAY-1895344 (e) for four days. Dose response curves show relative viability vs drug concentration. Plots show mean-/+stdev (n=4). Legend and EC50 values (nM) are shown in (f).
(g-i) One CDK12-KO prostate line, 22Rv1-CDK12-KO5, shows increased sensitivity to CHEK1 inhibition. Prostate cancer lines were treated with CHEK1 inhibitors rabusertib (g) and MK-8776 (h) for four days. Dose response curves show relative viability vs drug concentration. Plots show mean-/+stdev (n=4). Legend and EC50 values (μM) are shown in (i).
