Genome-wide profiling of genetic synthetic lethality identifies CDK12 as a novel determinant of PARP1/2 inhibitor sensitivity

Abstract

Small-molecule inhibitors of PARP1/2, such as olaparib, have been proposed to serve as a synthetic lethal therapy for cancers that harbor BRCA1 or BRCA2 mutations. Indeed, in clinical trials, PARP1/2 inhibitors elicit sustained antitumor responses in patients with germline BRCA gene mutations. In hypothesizing that additional genetic determinants might direct use of these drugs, we conducted a genome-wide synthetic lethal screen for candidate olaparib sensitivity genes. In support of this hypothesis, the set of identified genes included known determinants of olaparib sensitivity, such as BRCA1, RAD51, and Fanconi's anemia susceptibility genes. In addition, the set included genes implicated in established networks of DNA repair, DNA cohesion, and chromatin remodeling, none of which were known previously to confer sensitivity to PARP1/2 inhibition. Notably, integration of the list of candidate sensitivity genes with data from tumor DNA sequencing studies identified CDK12 deficiency as a clinically relevant biomarker of PARP1/2 inhibitor sensitivity. In models of high-grade serous ovarian cancer (HGS-OVCa), CDK12 attenuation was sufficient to confer sensitivity to PARP1/2 inhibition, suppression of DNA repair via homologous recombination, and reduced expression of BRCA1. As one of only nine genes known to be significantly mutated in HGS-OVCa, CDK12 has properties that should confirm interest in its use as a biomarker, particularly in ongoing clinical trials of PARP1/2 inhibitors and other agents that trigger replication fork arrest.

Figure 1. Genome-wide shRNA screen for the detection of olaparib sensitization effects

A, Heatmap showing the clustering of Drug Effect (DE) and Viability (V) Z scores from replica olaparib sensitization shRNA screens. The heatmap depicts DE and V Z scores for two replica screens (e.g. DE1 and V1 for replica 1 and DE2 and V2 for replica screen 2 and the average DE Z score). B, Plot of average olaparib DE Z scores for each of the shRNA constructs in the library. Each shRNA is ranked by its average DE Z score. DE Z score thresholds of −1.96 and 1.96 are shown as broken lines and BRCA1 shRNAs are highlighted in red. C, Histogram of average DE Z scores for each shRNA construct in the library. D, E and F, Molecular networks causing PARP1/2 inhibitor sensitivity. D, DNA repair genes, E, HR-associated genes and F, Cohesin and chromatin remodelling associated genes. Each gene in each network is represented by at least one shRNA with DE Z score <−1.96. Blue lines represent known physical or functional interactions between genes/proteins; dark blue lines represent high confidence interactions, with light blue lines representing less well-established interactions. Network diagrams were created from data in Supplementary Table 2, using STRING (41).

Figure 2. Mutations in CDK12

A, Schematic diagram of CDK12 annotated with protein alterations of CDK12 mutations in HGS-OVCa (31). Arginine/serine-rich (RS), proline-rich (PRM), kinase domain (KD) domains are indicated by blue, orange, green, respectively. Numbers below the schemes indicate the amino acid (AA) position. Coding alterations are colored as either black (missense), red (frameshift), green (indel) and blue (nonsense). Mutations in other cancer histologies are described in Supplementary Table S5. B, Heatmap representation of tumor-associated somatic mutations found in BRCA1, BRCA2 and CDK12 in 316 patients with HGS-OVCa, using data from (13). Each tumor is represented by a bar, with each green bar indicating a mutant tumor. Not all 316 tumors in the study cohort are shown. Frequency of gene mutations in this cohort is shown.

Figure 3. Low CDK12 expression correlates with PARP1/2 inhibitor sensitivity in serous ovarian tumor cell lines

A, 14 day olaparib survival curves of serous ovarian tumor cell lines are shown. Error bars represent SEM from three independent experiments. PEO1, which harbors a hemizygous BRCA2 p.Y1655X mutation, is included as a positive control B, (top panel) log₂ transformed olaparib SF₅₀ values for serous ovarian cancer cell line models and (bottom panel) log₂ transformed CDK12 mRNA expression levels in the same panel of serous ovarian cancer cell lines. Error bars represent SEM from triplicate independent experiments. C, Western blot of CDK12 expression in whole cell lysates from the tumor cell lines described in (B). ACTIN expression is shown as the loading control. D, Box and whiskers plot illustrating the difference in CDK12 mRNA expression between olaparib resistant (log₂ SF₅₀>0 M, n= 4 cell lines) and sensitive (log₂ SF₅₀<0 M, n= 3 cell lines) serous ovarian tumor cell models, PEO1 cell line was excluded from this analysis on account of its BRCA2 mutation. CDK12 mRNA expression is significantly lower in the sensitive cohort (p<0.05, Student’s t-test).

Figure 4. Silencing of CDK12 sensitises serous ovarian tumor cells to olaparib

A, Bar chart indicating CDK12 mRNA levels in OV90 cells expressing different CDK12 shRNA expression constructs (shCDK12_2, 3 and 4) or a control, non-targeting (shNTC) shRNA construct. Error bars represent SEM from three independent measurements. Each CDK12 shRNA significantly suppressed CDK12 mRNA levels compared to shNTC, (p<0.05, Student’s t test). B, Western blot of CDK12 expression in OV90 cells described in (A). The level of ACTIN expression was used as a loading control. C, 14 day olaparib survival curves of cells shown in (A). Error bars represent the SEM from three independent experiments. Each CDK12 shRNA significantly sensitised OV90 cells to olaparib compared to the shNTC population (p<0.05, ANOVA in each case). D, 14 day olaparib survival curves of OV90 cells transfected with a control non-targeting siRNA (siControl) or siRNAs targeting either CDK12 or BRCA1. Error bars represent the SEM from three independent experiments. CDK12 siRNA silenced OV90 cells sensitized to olaparib (p<0.05, ANOVA).

Figure 5. Silencing of CDK12 suppresses homologous recombination

A, Bar chart showing the effect of CDK12, BRCA1, BRCA2 siRNA silencing on HR frequency in HeLa cells harboring a single-copy genomic DR-GFP reporter. Error bars represent SEM from three independent experiments. CDK12 silencing significantly reduced HR frequency (***p<0.0001, Student’s t-test). B, Bar chart illustrating the frequency of nuclear RAD51 foci in OV90 cells expressing CONTROL, BRCA1 or CDK12 shRNA after exposure to ionizing radiation. Silencing of CDK12 significantly reduced RAD51 focus formation, (**p<0.001, Students t-test). Error bars for each individual experiment represent SEM. C, Representative confocal microscopy images of nuclear RAD51 foci (red) of OV90 cells as in (B). D, Bar chart illustrating CDK12 silencing caused a significant reduction in CDK12 and BRCA1 expression at the mRNA level (*p<0.05 and ***p<0.001, Student t-test). Error bars represent SEM from three independent measurements. E, Western blot of BRCA1 expression using whole cell lysates also analysed in Fig 4B. OV90 cells expressing CONTROL, BRCA1 or CDK12 shRNA were western blotted and probed for BRCA1 or ACTIN. F, Surviving fractions at 1μM olaparib in OV90 cells transfected with CONTROL, CDK12 or 53BP1 siRNAs (***p<0.001, Student t-test). Error bars represent SEM from three independent experiments.

Figure 6. CDK12 and therapy response in vivo

A, Kaplan–Meier plot of tumor CDK12 mRNA low vs. high patient groups. Tumor CDK12 mRNA levels from 316 HGS-OVCa ovarian cancer patients in the TCGA dataset treated with platinum therapy (13) were categorized into high or low CDK12 expression groups. p=0.0076 Log-rank Test CDK12 mRNA high vs. low, hazard ratio = 0.55. B, Box and whiskers plot representing the Kaplan–Meier analysis shown in (A) (p=0.0375, Student t-test) C and D, Mice bearing OV90 xenografts expressing either control shRNA (shNTC) or CDK12 shRNA (shCDK12), were treated as indicated. Each data point represents the mean increase in tumor volume after the instigation of treatment and error bars represent SEM, where n for each cohort = 10 animals. (p<0.001, ANOVA for shNTC olaparib or vehicle vs. shCDK12 olaparib and shCDK12 vehicle vs. shCDK12 olaparib).