MYBL2 Drives Prostate Cancer Plasticity: Inhibiting Its Transcriptional Target CDK2 for RB1-Deficient Neuroendocrine Prostate Cancer

Abstract

Phenotypic plasticity is a recognized mechanism driving therapeutic resistance in patients with prostate cancer. Although underlying molecular causations driving phenotypic plasticity have been identified, therapeutic success is yet to be achieved. To identify putative master regulator transcription factors (MR-TF) driving phenotypic plasticity in prostate cancer, this work utilized a multiomic approach using genetically engineered mouse models of prostate cancer combined with patient data to identify MYB proto-oncogene like 2 (MYBL2) as a significantly enriched transcription factor in prostate cancer exhibiting phenotypic plasticity. Genetic inhibition of Mybl2 using independent murine prostate cancer cell lines representing phenotypic plasticity demonstrated Mybl2 loss significantly decreased in vivo growth as well as cell fitness and repressed gene expression signatures involved in pluripotency and stemness. Because MYBL2 is currently not druggable, a MYBL2 gene signature was employed to identify cyclin-dependent kinase-2 (CDK2) as a potential therapeutic target. CDK2 inhibition phenocopied genetic loss of Mybl2 and significantly decreased in vivo tumor growth associated with enrichment of DNA damage. Together, this work demonstrates MYBL2 as an important MR-TF driving phenotypic plasticity in prostate cancer. Furthermore, high MYBL2 activity identifies prostate cancer that would be responsive to CDK2 inhibition.

Significance: Prostate cancers that escape therapy targeting the androgen receptor signaling pathways via phenotypic plasticity are currently untreatable. Our study identifies MYBL2 as a MR-TF in phenotypic plastic prostate cancer and implicates CDK2 inhibition as a novel therapeutic target for this most lethal subtype of prostate cancer.

Figure 1

MYBL2 is a MR-TF in murine prostate cancer deficient for Rb1. A, Schematic representation of the Coltron algorithm procedures for the construction of transcriptional regulatory networks. B, Candidate MR-TFs in SKO and DKO based on regulatory CESs (range 0–1, two sided t test P value <0.05) and nucleotide motif logos for top candidate factors relative size of each base indicates the relative frequency at that position. Foxa2, Sox1, Sox2, Myb, Ascl1, and Insm1 motifs with the more significant P value for more abundant motifs within H3K27ac-bound chromatin were the most enriched in the DKO sample. C, ChIP-seq tracks of H3K27ac signal at SE regions for Foxa1, Myb, Foxa2, Sox1, Sox2, Ascl1, and Insm1 in SKO (blue) and DKO (red) samples. D, Heatmaps representing the expression profile of 128 TFs overexpressed in SKO samples (FDR </= 1-% log₂ fold change >2; clustering method = “complete;” clustering distance = “Euclidean”) and 139 overexpressed DKO samples (FDR </= 1-% log₂ fold change >2; clustering method = “complete;” clustering distance = “Euclidean”).

Figure 2

MYBL2 expression and activity is enriched in human NEPC and NE-like mouse models. A, Venn diagram integrating the differential gene expression (DEG) list from PBCre4:Ptenf/f (SKO) vs. PBCre4:Ptenf/f:Rb1f/f (DKO) mouse models, human NEPC vs. adenocarcinoma samples, and patient-derived xenografts of LuCaP obtained from human patients with NEPC vs. patients with adenocarcinoma shows that the three independent datasets shared 1,516 genes (left). GSEA of the 1,516 shared genes indicates that the fischer_dream_targets genes may be involved in the regulation of CRPC-AI development (right). All the pathways listed are statistically significant with P < 0.05. B, Bar and box plot summarizing the MYBL2 expression level in human patients with prostate cancer at different stages of the disease progression. C, Correlation analysis mouse and (D) human samples indicate that enrichment of MYBL2 transcriptional function is highly associated with NE and ESC gene signatures in NEPC models and patient samples compared with their adenocarcinoma counterparts. WT, wild type.

Figure 3

Mybl2 supports tumoral proliferation and promotes self-renewal in prostate cancer cell line models. A, Normalized gene expression of MYBL2 in human prostate cancer cell lines. DEMETER gene dependency of MYB family members across human prostate cancer cell lines indicating MYBL2 as a significant dependency compared with other family members. B, Schematic representation of the DKO Ctl and DKO Mybl2 KO and PPKO Ctl and PPKO Mybl2 KO cell line generation using the two-step CRISPR/Cas9 method. Western blot indicating Mybl2 protein levels in the DKO Ctl, DKO Mybl2 KO, PPKO Ctl, and PPKO Mybl2 KO generated cell. C, Growth kinetics representation using relative luminescence intensity of DKO Clt vs. DKO Mybl2 KO spheroids and PPKO Ctl vs. PPKO Mybl2 KO spheroids cultured from day 0 (D0) to day 5 (D5) in a low-attachment round-bottom 96-well plate. D, Spheroid diameter and pictures of the DKO vs. DKO Mybl2 KO and PPKO vs. PPKO Mybl2 KO cells for 21 days postseeding of 500 or 350 cells/well, respectively. E, Flow cytometry analysis of the EdU-positive cells in DKO vs. DKO Mybl2 KO and PPKO vs. PPKO Mybl2 KO cells after 2 hours of EdU labeling in vitro.

Figure 4

Increased Mybl2 expression supports stem-related gene signatures in murine and human prostate cancer. A, Two-dimensional principal component analysis visualization of bulk RNA-seq analysis performed in DKO Ctl vs. DKO Mybl2 KO cells (left) and PPKO Ctl vs. PPKO Mybl2 KO cells. B, Venn diagram representation of down- and upregulated genes from the DEG analysis between DKO Mybl2 KO vs. DKO Ctl cells and PPKO Mybl2 KO vs. PPKO Ctl cells showed that both cell lines shared 197 down and 127 upregulated genes. The P adjusted value used to calculate the DEG was <0.05 and the LFC was < or > 2. C, Bar chart visualization of the top 10 enriched terms and their P values using ENCODE and ChIP-x enrichment analysis (ChEA), ChEA 2022, and Hallmarks 2020 databases using the 197 downregulated genes shared between DKO Mybl2 KO and PPKO Mybl2 KO cells. All pathways listed were statistically significant with P < 0.05. D, Normalized gene expression values for TCGA-PRAD and (E) SU2C samples from patient with CRPC show high MYBL2 expression (top 25% of samples) is associated with significantly increased expression of gene signatures for NEPC (NE), ASC, and ESC signatures when compared with patient samples with lowest MYBL2 expression (bottom 25% of samples).

Figure 5

Mybl2 is a genetic dependency associated with CDK2 expression in NEPC. A, Tumoral growth curve of murine DKO Ctl and DKO Mybl2 KO cells and (B) PPKO Ctl cells and PPKO Mybl2 KO cells in NOD SCID mice. Tumor growth was monitored by serial caliper measurements every second day for 16 days (D16). Tumor weight was measured at D16 after dissection (n = 5 mice per treatment group (±1 SEM). C, Top responses from the cancer therapeutics response portal analysis using a MYBL2 gene signature. DEMETER gene dependency of CDK family members 2, 5, and 7 across human prostate cancer cell lines indicating as significant, as outlined in Fig. 3A. D, Box plot graphs summarizing the RNA-seq analysis from published prostate cancer GEMMs (first and second graphs, including wild-type, SKO, DKO, NP:Nkx3.1^CreERT2:Ptenf/f and NPp53:Nkx3.1^CreERT2:Ptenf/f:Trp53f/f), and human samples (third graph, including CRPC and NEPC samples) indicate CDK2 upregulation in NEPC mouse models and human patients. E, Immunofluorescent IHC staining for CDK2 in DKO and (F) PPKO murine tumors indicates a significant loss of CDK2 expression following MYBL2 KO.

Figure 6

Inhibition of CDK2 represents a novel therapeutic target for NEPC. A, Tumoral growth curve of murine DKO cells in C57BL/6N and (B) PPKO cells in C57BL/6N mice and treated with either vehicle control or the CDK2 PROTAC—CPS2. Mice were treated every 7 days (D7 and D14) with CPS2 (400 mg/kg) by intraperitoneal injection. Tumor growth was monitored every second day for 16 days (D16) by serial caliper measurements. Tumoral weight was measured at D16 after dissection (n = 5 mice per treatment group (±1 SEM). C, Immunofluorescent IHC staining for CDK2 in DKO and (D) PPKO tumors show that CPS2 inhibition significantly reduces CDK2 expression. E and F, Example pictures of hematoxylin and eosin, Ki67, and p-γH2AX IHC staining in murine DKO and PPKO tumors from the in vivo study and corresponding quantification of the percentage of positive nuclei from total cells (n = 5 mice per treatment group, ±1 SEM).