CDK12 loss drives prostate cancer progression, transcription-replication conflicts, and synthetic lethality with paralog CDK13

Abstract

Biallelic loss of cyclin-dependent kinase 12 (CDK12) defines a metastatic castration-resistant prostate cancer (mCRPC) subtype. It remains unclear, however, whether CDK12 loss drives prostate cancer (PCa) development or uncovers pharmacologic vulnerabilities. Here, we show Cdk12 ablation in murine prostate epithelium is sufficient to induce preneoplastic lesions with lymphocytic infiltration. In allograft-based CRISPR screening, Cdk12 loss associates positively with Trp53 inactivation but negatively with Pten inactivation. Moreover, concurrent Cdk12/Trp53 ablation promotes proliferation of prostate-derived organoids, while Cdk12 knockout in Pten-null mice abrogates prostate tumor growth. In syngeneic systems, Cdk12/Trp53-null allografts exhibit luminal morphology and immune checkpoint blockade sensitivity. Mechanistically, Cdk12 inactivation mediates genomic instability by inducing transcription-replication conflicts. Strikingly, CDK12-mutant organoids and patient-derived xenografts are sensitive to inhibition or degradation of the paralog kinase, CDK13. We therein establish CDK12 as a bona fide tumor suppressor, mechanistically define how CDK12 inactivation causes genomic instability, and advance a therapeutic strategy for CDK12-mutant mCRPC.

Keywords: CDK12; CDK13; Cdk12 knockout; R-loops; paralog-based synthetic lethality; prostate cancer; transcription-replication conflicts.

Graphical abstract

Figure 1

Cdk12 ablation in the prostate epithelium induces neoplasia (A) Prostate epithelial Cdk12 ablation scheme. (B) H&E staining and CDK12 immunohistochemistry in representative prostate samples from 52-week-old Cdk12^pc−/− mice (pure C57 background) and WT controls. Left panel scale bars, 100 μm. Other scale bars, 50 μm. (C) Bar graphs indicate percent cross-sectional area occupied by histologically abnormal tissue. AP, anterior prostate; DP, dorsal prostate; VP, ventral prostate; LP, lateral prostate. (n = 8/group). (D) Pathological scoring (Path score) of prostate from the same animals. Numerical scores assigned to normal tissue (0), hyperplasia (1), focal HGPIN (2), and AIP (3) (indicated by respective images). Scale bars, 50 μm. Bar graph shows percentage prostate cross-sectional area occupied by tissue of each path score (scores 2 and 3 added together). (n = 7–8/group). Statistical analysis with Mann-Whitney test. (E) Immunofluorescent staining of cytokeratin-8 (K8) and p63. 52-week-old Cdk12^pc−/− image shows an area of focal HGPIN with expansion of p63(+) BCs. Scale bars in left (100 μm) and right images (50 μm). (F) Ki67 immunohistochemistry in Cdk12^pc−/− or WT mice. Scale bars, 50 μm. Bar graph indicates percentage Ki67(+) cells per high-powered field (n = 9 images from 3 mice/group). Data represented as mean ± SEM. (G) Immunohistochemistry for immune cell markers, indicating T cell-predominant infiltrate surrounding lesions in Cdk12^pc−/− animals. Scale bars, 100 μm. Bar graph data indicate number of each cell type per high-powered field (n = 3–5 images from 3 mice/group). Box indicates standard deviation. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Student’s t test used in (C), (F), and (G). See also Figure S1.

Figure 2

Organoids derived from the Cdk12^pc−/− prostate are morphologically abnormal, with impaired basal-luminal segregation (A and B) Images of organoids derived from Pb-Cre;Cdk12^f/f;mT/mG prostate BCs (52-week time point). Tom indicates Td-tomato-expressing cells with wild-type Cdk12 (Cdk12^WT). GFP indicates GFP-expressing cells with Cdk12 ablation (Cdk12^KO). Scale bars, 500 μm in (A) and 250 μm in (B). (C) CDK12 immunoblot in Cdk12^WT vs. Cdk12^KO organoids. (Vinculin, loading control). (D) Cdk12^KO organoid morphology: H&E staining, and CDK12 immunohistochemistry. Scale bars 200 μm in top and 50 μm in bottom panels. (E) Immunofluorescence for cytokeratin-8 (K8) and p63 indicating basal-luminal disorganization in Cdk12^KO organoids. Scale bars, 200 μm. (F) Uniform Manifold Approximation and Projection (UMAP) of scRNA-seq from Cdk12^WT organoids (n = 3). The five identified cell states progress from Basal_1, Basal_2, Basal_3, Lum_1, to Lum_2. (G) UMAP of scRNA-seq from Cdk12^KO organoids (n = 3). (H) Cells from Cdk12^KO organoids (n = 3) projected into the UMAP of Cdk12^WT. Pseudocolor indicates presence (yellow) or absence (purple) of Cdk12 transcript. (I) Distributions of different cell states in Cdk12^WT and Cdk12^KO organoids. The most differentiated (Lum_2) population is lost in Cdk12^KO organoids. (J) GSEA for human CDK12-loss signature—shared down-regulated genes from human PCa with CDK12 inactivation and siCDK12 knockdown LNCaP cells (18)— in Cdk12^KO organoids. Heatmap of logFC (Cdk12^KO vs. Cdk12^WT organoids) for genes in signature. Genes contributing to negative enrichment (leading edge) are labeled. See also Figure S2.

Figure 3

Cdk12 and Trp53 inactivating alterations interact to promote PCa (A) Workflow for CRISPR library screening of Cdk12-interacting genes. (B) Snake plot representing log₂ fold change of guide RNAs in sequenced tumor samples described in (A). (n = 3/group in 2 unique experiments). (C) Immunohistochemistry for p53 (left panels) and γH2AX (right panels) in prostates of one-year-old WT and Cdk12^pc−/− mice. Scale bars, 50 μm. Bar graph indicates percent γH2AX(+) cells from average of 3–5 sections from each of 3 mice. Data represented as mean ± SEM. t test used for individual comparisons. (D) Protein expression of p53 and γH2AX in Cdk12^WT and Cdk12^KO organoids (GAPDH, loading control). (E) CDK12-p53 co-staining in Cdk12^WT and Cdk12^KO organoids. Scale bars, 50 μm. (F) CRISPR-mediated Trp53 ablation in Cdk12^WT and Cdk12^KO organoids. sgp53 indicates Trp53-specific guide RNA. sgNT indicates control non-targeting guide RNA. (α-tubulin, loading control). (G) Relative expression (Rel Exp) levels of Trp53 and p53 target genes in samples described in (F). (n = 3/group). Data represented as mean ± SEM. One-way AVOVA used for statistical comparisons. (H) Cell proliferation in organoids from groups indicated in (F) measured by CellTiter-Glo (CTG assay). (n = 3–4/group). Data represented as mean ± SEM. Two-way ANOVA used for statistical comparisons. (I) Kaplan-Meier plots indicating tumor formation during 70 days post-implantation of Cdk12^WT and Cdk12^KO organoids with or without Trp53 ablation (n = 10/group). (J) Immunohistochemistry of AR, p53, CDK12, and γH2AX in Cdk12^KO-sgp53 allografts. Scale bars, 50 μm. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. See also Figure S3 and Table S1.

Figure 4

Cdk12/Trp53 double knockout allografts exhibit lymphocytic immune responses and increased sensitivity to ICB therapy (A) Growth of Cdk12^KO-sgp53, Myc-CaP, and TRAMP-C2 allografts in immunocompetent wild-type mice. (n = 10–15/group). (B) Immunohistochemistry of CDK12, T cell markers (CD3, CD4, CD8), and natural killer cell marker granzyme B in Cdk12^KO-sgp53 allografts, Myc-CaP allografts, and TRAMP-C2 allografts, and prostates of Pten^pc−/− PCa mouse model. Scale bars, 50 μm. (C and D) Tumor growth curve and endpoint weights of Cdk12^KO-sgp53 (C) and TRAMP-C2 (D) allografts treated with anti-PD1/CTLA4 cocktail. (n = 8–14/group). (E) Flow cytometry-based quantification of CD4(+) and CD8(+) T cells (total, IFNγ(+), granzyme B(+)) in Cdk12^KO-sgp53 and TRAMP-C2 allograft samples +/− treatment with anti-PD1/CTLA4 cocktail. (n = 7–8/group). (F) Kaplan-Meier plots demonstrating survival of prostate-specific Pten^pc−/− and Pten^pc−/−Cdk12^pc−/−mice. (G) Genitourinary tract weights of Pten^pc−/− and Pten^pc−/−Cdk12^pc−/− mice as well as wild-type mice (52 weeks). (H) Cell proliferation in complete media of epithelial cell organoids derived from Pten^pc−/− and Pten^pc−/−Cdk12^pc−/− mice measured by CTG assay. (n = 4/group). Data represented as mean ± SEM. Data represented as mean ± SEM. One-way ANOVA for multiple comparisons (G), two-way ANOVA for multiple variables (C) and (E), and unpaired t test was used for tumor weight in (C) and (D). ∗∗∗∗p < 0.0001; ns, not significant. See also Figures S4 and S5.

Figure 5

Cdk12 ablation increases AR- and MYC-mediated signaling and promotes TRCs (A) Protein expression of CDK12, AR, and FOXA1 in multiple monoclonal Cdk12^WT and Cdk12^KO organoid lines. (GAPDH, loading control). (B) Gene set enrichment of AR target genes (activated and repressed) in Cdk12^KO organoids compared to Cdk12^WT. (C) Proliferation of Cdk12^WT and Cdk12^KO organoids grown in the absence of epidermal growth factor (EGF) and dihydrotestosterone (DHT) as measured by the CTG assay. (n = 3 replicates per group in 2 unique experiments). (D and E) Morphology and viability quantification of Cdk12^WT and Cdk12^KO organoids subjected to enzalutamide (Enza) treatment. (n = 3 replicates per group in 2 unique experiments). ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; ns, not significant. (F) Protein expression of CDK12, MYC, BRD4, BRD3, and BRD2 in Cdk12^WT and Cdk12^KO organoid lines. (G) Gene set enrichment of MYC target genes in Cdk12^KO organoids compared to Cdk12^WT. (H) Morphology of Cdk12^WT and Cdk12^KO organoid lines treated with JQ1 (1 μM). (n = 3/group in 2 unique experiments). (I) Viability curves and IC₅₀ values for JQ1-treated Cdk12^WT and Cdk12^KO organoid lines. (J) Dot blot analysis quantifying R-loops in Cdk12^WT and Cdk12^KO organoids. RNase H1 treatment serves as a negative control. (K) Immunofluorescence images of R-loop (red) staining of Cdk12^WT and Cdk12^KO organoids (left) and quantification of fluorescence intensity (right). 100–200 cells/group. (L) Experimental workflow for identification of TRCs. Briefly, 2.5 mM of Thymidine was used to synchronize the cells, and 75 μM of DRB was used to inhibit transcription. (M) Representative immunofluorescence images of γH2AX staining in organoids treated as described in (L). (N) Quantification of γH2AX-positive cells in (M); (n = 6/group, 3 unique experiments conducted). (O) Representative immunofluorescence images of γH2AX staining in unsynchronized organoids. (P) Quantification of γH2AX-positive cells in (O); n = 6–8 per group (3 unique experiments conducted). (Q) Detection of TRC by PLA assay. (R) Quantification of PLA foci per nucleus in (Q); 100–400 cells analyzed per group (2 unique experiments conducted). Data represented as mean ± SEM. One-way ANOVA for multiple comparisons, two-way ANOVA for multiple variables.

Figure 6

Cdk12^KO organoids and CDK12-mutant tumors are preferentially sensitive to a CDK13/12 degrader (A) Snake plot representing data from siRNA screen for CDK12 synthetic lethal effects via 1NM sensitivity in CDK12^as cells. Negative Z scores indicate CDK12 synthetic lethal effects, with CDK13 representing most profound effect. (B) Immunoblot indicating CDK13 gene silencing with two different siRNAs (siCDK13.1 and siCDK13.2). (C) Curve depicting cell survival in 1NM-exposed CDK12^as cells transfected with one of two unique CDK13 siRNAs (siCDK13.1 and siCDK13.2) or control siRNA (siCON). (D) CRISPR-mediated Cdk13 (sgCdk13(1 + 3), or sgCdk13(2 + 4)) knockout in Cdk12^WT and Cdk12^KO organoids harvested on day 5 after lentiviral transduction. Protein expression of CDK12 and CDK13 in organoids (Vinculin, loading control). (E) Bright-field images of organoids described in (D). Scale bars, 200 μm. (F) Relative growth quantification from images in (E). (n = 3/group). (G) CRISPR ablation of Cdk12 (sgCdk12) and Cdk13 (sgCdk13(1 + 3), or sgCdk13(2 + 4)) in Myc-CaP cells. Protein expression of CDK12 and CDK13 in Myc-CaP cells treated with indicated sgRNAs. (H) Colony formation assay showing survival in cells treated with indicated sgRNAs (representative data from 3 unique experiments). (I) Relative growth quantification from images in (H) (analysis of 11 high-powered fields per sample over 2 unique experiments). (J) (Top panel) C4-2B cells subjected to CRISPR-based CDK12 ablation (CDK12KO) or control sgRNA (C4-2B CTRL): percent confluence with siRNA-based CDK13 knockdown (siCDK13) or control siRNA (siNTC). (Bottom panel) C4-2B CDK12KO and C4-2B CTRL cells: percent confluence with siRNA-based CCNK knockdown or control siRNA treatment. (n = 3/group). (K) Images of Cdk12^WT and Cdk12^KO organoids (with or without Trp53 ablation) following treatment with CDK12/13 degrader (YJ9069). sgp53 indicates Trp53 ablation, while sgNT indicates intact Trp53. Scale bars, 1,000 μm. (L) Viability curves and IC₅₀ values for YJ9069 treatment of groups described in (K). (n = 4) (M) Protein expression of p-Ser RNA Pol-II, CDK12, CDK13, and p53 in Cdk12^WT and Cdk12^KO organoids with or without Trp53 ablation subjected to YJ9069 degrader or vehicle treatment. (N) IC₅₀ of organoids derived from PDX lines with WT CDK12 (MDA153, MDA146-12, LuCaP23.1, LuCaP86.2, LuCaP96, PC295) and inactivating CDK12 mutation (LTL706B, MDA117, MDA328). (n = 3 per line). Data represented as mean ± SEM. One-way ANOVA for multiple comparisons, two-way ANOVA for multiple variables, ∗∗p < 0.01, ∗∗∗∗p < 0.0001. See also Figures S6, S7 and Table S2.

Figure 7

CDK13/12 degrader inhibits CDK12-mutant tumor growth in vivo (A) In vivo treatment of Cdk12^KO-sgp53 allografts with YJ9069 or vehicle: line graph indicates tumor volume normalized to baseline. Bar graph indicates tumor weight at endpoint. (n = 9–10 mice, each with 2 tumors, per group) (B) In vivo treatment of TRAMP-C2 allografts with YJ9069 or vehicle: graphs as indicated in (A) (n = 9–10 mice, each with 2 tumors, per group). (C and D) Unmodified (sgNT-treated) Myc-CaP allografts (C) or sgCdk12-treated Myc-CaP allografts (D) subjected to in vivo YJ9069 treatment: line graphs indicate tumor volume. Bar graphs indicate tumor weight at end of treatment time course. (n = 6–9/group). (E–H) CDK12 immunohistochemistry and TUNEL staining of unmodified (sgNT-treated) Myc-CaP allografts (E) and sgCdk12-treated Myc-CaP allografts (F). Bar graphs (G and H) indicate quantification of TUNEL(+) cells per high-powered field (scale bars, 50 μm). (I and J) YJ9069 treatment of subcutaneously implanted PDX lines, LTL706B (CDK12-mutant), and MDA146-12 (intact CDK12). Graphs indicate tumor volume (n = 8–9 mice, each with 2 tumors per group). Two-way ANOVA used for tumor volume in (A), (D), and (I); unpaired t test used for tumor weight in (A–D) and (I and J) and TUNEL staining (G and H). ∗∗∗∗p < 0.0001; ns, not significant. See also Figure S7.