Epigenetic Derepression of PROX1 Promotes Neuroendocrine Prostate Cancer Progression

Abstract

Histologic transformation of prostate cancer from adenocarcinoma to neuroendocrine prostate cancer (NEPC) is associated with aggressive disease and poor prognosis. This lineage transition is accompanied by polycomb complex 2-mediated epigenetic derepression of cell fate-determining transcription factors, including prospero homeobox 1 (PROX1). In this study, we sought to functionally characterize the role of PROX1 in NEPC. An unbiased CRISPR screen in two NEPC patient-derived organoid models demonstrated high cellular dependency for PROX1. Knockout of PROX1 impeded tumor growth in NEPC models, and overexpression of PROX1 promoted tumor growth and spontaneous metastasis in prostate adenocarcinoma. Transcriptomic and cistromic analyses across castration-resistant adenocarcinoma and neuroendocrine models pointed to PROX1-mediated regulation of neuroendocrine-lineage transcriptional programs. Immunoprecipitation followed by mass spectrometry identified three phosphorylated sites in the DNA-binding domain of PROX1 that are critical for its stability and function. CHEK1 and CDK2 were predicted to be upstream kinases that phosphorylate PROX1, and treatment with a CHEK1 or CDK2 inhibitor reduced NEPC viability. Together, these results substantiate the role of PROX1 in NEPC and identify PROX1 phosphorylation in the DNA-binding domain, which might represent a therapeutic target in NEPC.

Significance: PROX1 mediates lineage reprogramming, tumor growth, and metastasis in neuroendocrine prostate cancer and represents a cellular dependency that can be exploited for targeted treatment strategies.

Figure 1:. Epigenetic de-repression of PROX1 in NEPC

(a) Scatter plot with differentially enriched 10kb genome-wide bins of H3K27me3 log₂FPM values in NEPC (n=9) and castration resistant PRAD (n=9) samples on the x-axis (19) and fold change of Log₂RPKM RNA-seq values of castration resistant PRAD (n=31) and NEPC (n=22) (20); pale pink, luminal-lineage transcription factors upregulated in PRAD are labelled; burgundy, NE-lineage transcription factors with highly de-repressed candidates labelled; r value corresponding p-value calculated from Pearson’s product-moment correlation analysis; gray line, linear regression; gray band, standard error of regression. (b) mRNA levels of PROX1 in benign (n=34), localized PRAD (n=68), castration resistant PRAD (crPRAD, n=31), and NEPC (n=22) represented as Log₂(RPKM+1) scaled to mean (20); sky blue, benign; blue, localized PRAD; pale pink, castration resistant PRAD; burgundy, NEPC. (c) mRNA levels of PROX1 in PRAD cell lines (LNCaP, LNCaP-abl, and 22Rv1, n=3/model) and NEPC patient derived organoids (WCM154, WCM155, WCM1262, and WCM1078, n=3/model) represented as Log₂(FPKM+1) (19); black line, mean Log2(FPKM+1); pale pink, PRAD models; burgundy, NEPC models. (d) Western blot analysis in PRAD cell lines (LNCaP, LNCaP-abl, and 22Rv1) and NEPC patient derived organoids (WCM154, WCM155, WCM1262, and WCM1078) confirming the PROX1 expression at the protein level corresponding to mRNA levels in (C). Blots were reprobed for β-actin as a loading control. Representative western blot from two independent experiments is shown. (e) Screenshots from Integrative Genomics Viewer of CUT&RUN H3K27me3 or H3K4me3 or IgG levels in NEPC patient derived organoids (WCM154, WCM155, WCM1262, and WCM1078) at the genomic locus of PROX1 confirming bivalency; gray, IgG; red, H3K27me3; blue, H3K4me3. (f) WCM154 control or EZH2 knockout lines were processed for RNA-Seq (19). mRNA levels of PROX1 represented as CPM values normalized to library sizes. Columns, means of values obtained from 3 independent biological replicates; gray, WCM154-sgINC; orange, WCM154-sgEZH2 (#1 and #2); bars, SEM values; *p-value < 0.05, all statistical analyses used Wilcoxon tests. (g) WCM154 control or EZH2 knockout lines or rescue with a dTAG-HA-EZH2 treated with vehicle or with 500nM dTAGv-1 were processed for RNA-Seq (19). mRNA levels of PROX1 represented as CPM values normalized to library sizes. Columns, means of values obtained from 3 independent biological replicates; gray, WCM154-sgINC; orange, WCM154-sgEZH2 #1; light purple, WCM154-sgEZH2_rescue_dTAG-HA-EZH2 treated with vehicle; dark purple, WCM154-sgEZH2_rescue_dTAG-HA-EZH2 treated with 500nM dTAGv1 for 9 days; bars, SEM values; –, p-value > 0.05; *, p-value < 0.05, all statistical analyses used Wilcoxon tests. (h) Screenshots from Integrative Genomics Viewer of CUT&RUN H3K27ac levels in NEPC patient derived organoids (WCM154, WCM155, WCM1262, and WCM1078) at the genomic loci of PROX1 and 190kb upstream containing potential enhancers; green, H3K27ac; red dashed line, H3K27ac peaks indicating potential enhancer peaks. (i) WCM154 control or ASCL1 knockout lines were processed for RNA-Seq (42). mRNA levels of PROX1 represented as CPM values normalized to library sizes. Columns, means of values obtained from 3 independent biological replicates; gray, WCM154-sgGFP; magenta, WCM154-sgASCL1; bars, SEM values; *p-value < 0.05, all statistical analyses used Wilcoxon tests.

Figure 2:. PROX1 promotes tumor progression.

(a) CRISPR screen performed using WCM1078 NEPC patient-derived organoid model. Dependency score was calculated upon normalization with gene expression; burgundy, NE-lineage transcription factors. (b) Western blot analysis for PROX1 showing efficiency of knockout of PROX1 in WCM154 using two independent sgRNAs targeting PROX1. Blots were reprobed for β-actin as loading control. Representative western blot from three independent experiments is shown. (c) WCM154 control (sgINC) or PROX1 knockout (sgPROX1 #1 and #2) cell lines were implanted subcutaneously in castrated NSG male mice. Day 0 marks the point at which the respective individual tumors reached an approximate volume of 100 mm³. spaghetti line, tumor growth of independent biological replicates within a group; thick line, non-linear regression; band, standard error of regression, gray, WCM154 sgINC (control, n=5); green, WCM154 sgPROX1 #1 (n=5); olive, WCM154 sgPROX1 #2 (n=5). (d) Western blot analysis showing efficiency of knockout of PROX1 and rescue using recombinant dTAG-HA-PROX1 in WCM154. Impact of dTAGv-1 treatment at 8 hours demonstrates robust PROX1 degradation. Blots were reprobed for β-actin as loading control. Representative western blot from two independent experiments is shown. (e) PROX1 rescue WCM154 model described (d) was treated with vehicle (Veh; DMSO) or 500 nM dTAGv-1. CellTiter-Glo^® luminescent cell viability assay was performed at day 0, 3, or 6 upon treatment. Connected dots, means of values (n=5/condition); gray, vehicle treatment; red, dTAGv-1 treatment; bars, SEM values; –, p-value > 0.05 (not significant); *, p-value < 0.05, all statistical analyses used Wilcoxon two-sided tests. (f) Western blot analysis for indicated proteins showing efficiency of PROX1 over-expression in PRAD models (C4–2 and 22Rv1) along with control lines expressing empty vector. Blots were reprobed for β-actin as loading control. Representative western blot from two independent experiments is shown. (g) C4–2 control (Ctrl) expressing empty vector or PROX1 over-expression (PROX1) cell lines were implanted subcutaneously in castrated NSG male mice. Day 0 marks the point at which the respective individual tumors reached an approximate volume of 100 mm³. spaghetti line, tumor growth of independent biological replicates within a group; thick line, non-linear regression; band, standard error of regression, gray, C4–2 Ctrl (control, n=5); dark cyan, C4–2 PROX1 overexpression (n=5). (h) 22Rv1 control (Ctrl) expressing empty vector or PROX1 over-expression (PROX1) cell lines were implanted subcutaneously in castrated NSG male mice. Day 0 marks the point at which the respective individual tumors reached an approximate volume of 100 mm³. spaghetti line, tumor growth of independent biological replicates within a group; thick line, non-linear regression; band, standard error of regression, gray, 22Rv1 Ctrl (control, n=4); dark cyan, 22Rv1 PROX1 overexpression (n=5). (i) 22Rv1 PROX1 overexpression model developed strong liver metastasis (n=4/5 mice). Representative H&E staining distinguishing histology of tumors (Met) in the liver. Scale bar = 3 mm. (j) Western blot analysis for indicated proteins showing high levels of chromogranin A (CHGA), a neuroendocrine marker in tumor samples of 22Rv1 Ctrl of 22Rv1 PROX1 over-expression suncutaneous growth or 22Rv1 PROX1 over-expression liver metastasis. Blots were reprobed for β-actin as loading control. Representative western blot from two independent experiments is shown. (k) Representative immunohistochemistry of tumor samples from (h-j) confirming upregulation of chromogranin A in 22Rv1 PROX1 over-expression. Scale bar = 100 μm. (l) Pathologist’s assessment of chromogranin A levels from 22Rv1 tumor samples (ctrl, n=4; PROX1 over-expression subcutaneous tumors, n=5; PROX1 overexpression liver metastases, n=5).

Figure 3:. PROX1 regulates a network of NE-lineage transcription factors.

(a) Volcano plot of differentially expressed genes in 22Rv1 PROX1 over-expression (PROX1-OE) vs control cells expressing empty vector (n=3/condition); blue, significantly downregulated genes upon PROX1 overexpression (log₂FoldChange < 0; p-value < 0.05); red, significantly upregulated genes upon upon PROX1 overexpression (log₂FoldChange > 0; p-value < 0.05); pale pink, luminal-lineage or AR-target genes downregulated upon PROX1 overexpression; burgundy, NE-lineage transcription factors upregulated upon PROX1 overexpression. (b) Gene Set Enrichment Analysis of androgen response hallmark pathway in 22Rv1 PROX1 over-expression (PROX1-OE) vs control cells expressing empty vector; NES, normalized enrichment score; FDR, false-discovery rate (c) dTAG-HA-PROX1 22Rv1 model was treated with vehicle (Veh; DMSO) or 500 nM dTAGv-1 or 20 μM enzalutamide or combination of dTAGv-1 and enzalutamide for 6 days and CellTiter-Glo^® luminescent cell viability assay was performed. Columns, means of values (n=5/condition); gray, vehicle treatment; red, enzalutamide treatment; dark cyan, dTAGv-1 treatment; green, combination treatment; bars, SEM values; *, p-value < 0.05, all statistical analyses used Wilcoxon two-sided tests. (d) Volcano plot of differentially expressed genes in WCM154 PROX1 knockout (sgPROX1 #1) vs control (sgINC) organoids (n=3/condition); blue, significantly downregulated genes upon PROX1 knockout (log₂FoldChange < 0; p-value < 0.05); red, significantly upregulated genes upon PROX1 knockout (log₂FoldChange > 0; p-value < 0.05); pale pink, luminal-lineage or AR-target genes upregulated upon PROX1 knockout; burgundy, NE-lineage transcription factors downregulated upon PROX1 knockout. (e) Gene Set Enrichment Analysis of androgen response hallmark pathway in WCM154 PROX1 knockout (sgPROX1 #1) vs control (sgINC) organoids; NES, normalized enrichment score; FDR, false-discovery rate (f) Schematic of PROX1 rescue and strategy to identify direct transcriptional targets. Created in BioRender. Venkadakrishnan, V. (2025) https://BioRender.com/iv2bjk4 (g) Volcano plot of differentially expressed genes in WCM154 rescue dTAG-HA-PROX1 (rescue) vs PROX1 knockout (sgPROX1 #1) (n=3/condition); blue, significantly downregulated genes upon PROX1 rescue (log₂FoldChange < 0; p-value < 0.05); red, significantly upregulated genes upon PROX1 rescue (log₂FoldChange > 0; p-value < 0.05); orange, PROX1 expression labelled as control, pale pink, luminal-lineage or AR-target genes downregulated upon PROX1 rescue; burgundy, NE-lineage transcription factors upregulated upon PROX1 rescue. (h-i) Volcano plot of differentially expressed genes in WCM154 rescue dTAG-HA-PROX1 treated with vehicle (veh) vs treatment with dTAGv-1 for 8 hours (h) or after 48 hours (i) (n=3/condition); blue, significantly downregulated genes upon dTAGv-1 treatment (log₂FoldChange < 0; p-value < 0.05); red, significantly upregulated genes upon dTAGv-1 treatment (log₂FoldChange > 0; p-value < 0.05); pale pink, luminal-lineage or AR-target genes upregulated upon dTAGv-1 treatment; burgundy, NE-lineage transcription factors downregulated upon dTAGv-1 treatment. (j) IgG (gray) or PROX1 (dark cyan) or H3K27ac (green) CUT&RUN from WCM154 was analyzed and plots show profiles and heatmaps are shown of normalized peaks at ±6kb from the center of peaks found within the promoter regions. (k) Scatter plot of differentially expressed genes in WCM154 rescue dTAG-HA-PROX1 treated with vehicle (veh) vs treatment with dTAGv-1 for 8 hours (n=3/condition); blue, significantly downregulated genes upon dTAGv-1 treatment (150 genes, log₂FoldChange < 0; q-value < 0.05); red, significantly upregulated genes upon dTAGv-1 treatment (343 genes, log₂FoldChange > 0; q-value < 0.05). (l) Curated C2 oncogenic signature analysis of genes bound by PROX1 and downregulated upon dTAGv-1 treatment at 8 hours with p-adj < 0.05 and Log₂FoldChange < 0; size, −log₁₀FDR; orange-black gradient, gene ratio. (m) Screenshots from Integrative Genomics Viewer of CUT&RUN PROX1 or IgG levels in WCM154 NEPC patient-derived organoids at the genomic loci of indicated genes confirming PROX1 binding; gray, IgG; dark cyan, PROX1.

Figure 4:. Phosphorylation of PROX1 is critical for its function.

(a) Immunoprecipitation (IP) of PROX1 followed by in-gel digestion and mass-spectrometry. Immunoprecipation was done using non-targeting IgG antibody or an antibody targeting PROX1 using WCM154 NEPC organoids. Western blotting was done using PROX1 antibody. Mass spectrometry (LC-MS/MS) identified three novel phosphorylation sites in the homeo-prospero DNA binding domain depicted in the schematic. Low, low exposure; high, high exposure. Created in BioRender. Venkadakrishnan, V. (2025) https://BioRender.com/zkvu1v7 (b) Schematic representation of phospho-null (serine mutated to alanine) and phospho-mimetic mutation (serine mutated to aspartic acid) made using site-directed mutagenesis. Created in BioRender. Venkadakrishnan, V. (2025) https://BioRender.com/52xnko6 (c) Phospho-mimetic mutations increase PROX1 expression and S719A phospho-null mutation decrease its expression. Western blot analysis for indicated proteins showing transient expression of control empty vector (Ctrl) or PROX1 wild-type (WT) or indicated mutants of PROX1 impacting expression/stability in C4–2 PRAD model. Blots were reprobed for β-actin as loading control. Representative western blot from two independent experiments is shown. (d) Mutations of PROX1 phosphorylation site S719 impacts its downstream function. ASCL1 expression upon PROX1 WT or mutant overexpression in C4–2 PRAD model as compared to control cells expressing empty vector. Target gene (ASCL1) mRNA levels were normalized to GAPDH expression and are represented as relative expression using one of the values obtained from control conditions as 1. Columns, means of values (n=3); gray, control (Ctrl); dark cyan, PROX1 overexpression; olive, PROX1 S719A phospho-null mutant overexpression; red, PROX1 S719D phospho-mimetic mutant overexpression; bars, SEM values; *p-value < 0.05, all statistical analyses used Wilcoxon two-sided tests. (e) Analysis of the phosphorylated peptides using the kinase library (45) and prediction of upstream kinases potentially mediating phosphorylation of corresponding serine residues. (f) Western blot analysis for indicated proteins showing decreased levels of chromogranin A (CHGA), a neuroendocrine marker and PROX1 in WCM154 NEPC organoids treated with 5 nM prexasertib (CHK1 inhibitor) or 500 nM tagtociclib (CDK2 inhibitor) or vehicle (Veh; DMSO) for 6 days. Blots were reprobed for β-actin as loading control. Representative western blot from two independent experiments is shown. (g) WCM154 treated with vehicle (Veh; DMSO) or 5 nM prexasertib (CHK1 inhibitor) or 500 nM tagtociclib (CDK2 inhibitor) for 6 days and CellTiter-Glo^® luminescent cell viability assay was performed. Columns, means of values (n=5/condition); gray, vehicle treatment; purple, tagtociclib treatment; indigo, prexasertib treatment; bars, SEM values; *, p-value < 0.05, all statistical analyses used Wilcoxon two-sided tests.