PlexinD1 is a driver and a therapeutic target in advanced prostate cancer

Abstract

Aggressive prostate cancer (PCa) variants associated with androgen receptor signaling inhibitor (ARSI) resistance and metastasis remain poorly understood. Here, we identify the axon guidance semaphorin receptor PlexinD1 as a crucial driver of cancer aggressiveness in metastatic castration-resistant prostate cancer (CRPC). High PlexinD1 expression in human PCa is correlated with adverse clinical outcomes. PlexinD1 critically maintains CRPC aggressive behaviors in vitro and in vivo, and confers stemness and cellular plasticity to promote multilineage differentiation including a neuroendocrine-like phenotype for ARSI resistance. Mechanistically, PlexinD1 is upregulated upon relief of AR-mediated transcriptional repression of PlexinD1 under ARSI treatment, and subsdquently transactivates ErbB3 and cMet via direct interaction, which triggers the ERK/AKT pathways to induce noncanonical Gli1-dictated Hedgehog signaling, facilitating the growth and plasticity of PCa cells. Blockade of PlexinD1 by the protein inhibitor D1SP restricted CRPC growth in multiple preclinical models. Collectively, these findings characterize PlexinD1's contribution to PCa progression and offer a potential PlexinD1-targeted therapy for advanced PCa.

Keywords: Androgen Receptor Signaling inhibitors; Castration-resistant Prostate Cancer; Cellular Plasticity; Neuroendocrine Differentiation; PlexinD1.

Figure 1. PlexinD1 and associated Sema ligands are upregulated in CRPC cells.

(A) GSEA of multiple cellular event and plasticity related gene sets enriched in C4-2B^ENZR vs. LNCaP cells. (B) Dot plot depicting the axon guidance pathway among KEGG pathways highly enriched in C4-2B^ENZR vs. LNCaP cells. (C) Volcano plot of axon guidance regulator genes detected in C4-2B^ENZR vs. LNCaP cells (n = 2 biological replicates). (D) GSEA plots of two semaphorin/plexin signaling-related gene sets enriched in C4-2B^ENZR vs. LNCaP cells. (E) Western blot of PlexinD1, Sema3E, Sema3C, and NE and AR signaling markers in LNCaP and C4-2B^ENZR cells. (F) Representative PLA images and quantification of PlexinD1-Sema3E/Sema3C interaction by per-cell cytoplasmic fluorescence intensity in LNCaP and C4-2B^ENZR cells. PlexinD1 antibody incubation alone served as negative controls. The n of each group indicates the number of cells examined for quantification. Scale bars: 5 µm. (G) Representative Western blotting images and quantification of PlexinD1, Sema3E, and Sema3C protein expression in a panel of cell lines as indicated, with the averaged individual protein levels after normalization to β-actin in RWPE-1 cells from three independent experiments set as 1. Note that unequal amounts of total cell lysates from different cell lines were loaded intentionally due to the large discrepancy in target protein levels among different cell lines. Data information: In (F), data are presented as mean ± SEM. In (A, D), P values were determined by permutation test. In (C), P values were determined by unpaired two-tailed Student’s t-test. In (F), P values were determined by one‐way ANOVA with Dunnett’s multiple comparisons test. **P < 0.01. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 2. Elevated PlexinD1 expression levels are correlated with worse clinical outcomes in PCa.

(A) Representative images and quantification of PlexinD1 IHC staining in primary PCa vs. normal prostate tissue from a US Biomax TMA. Scale bars: 100 μm. (B) Representative images and quantification of PlexinD1 IHC staining in primary PCa categorized by low Gleason score (GS) (GS 4–6) and high GS (GS 7–10) from the same TMA as in (A). Scale bars: 100 μm. (C) Comparisons of PLXND1 mRNA levels in PCa vs. normal prostates from GSE3325 and GSE3933. (D) Representative images and quantification of PlexinD1 IHC staining in PCa tissue from patients post- vs. pre-hormone therapy from the NYU TMA. Scale bars: 100 μm. (E) Comparisons of PLXND1 mRNA levels in recurrent vs. non-recurrent and hormone-refractory vs. -sensitive PCa from GSE21032 and GSE6099, respectively. (F) Representative images and quantification of PlexinD1 IHC staining in bone metastatic vs. primary PCa. Scale bars: 100 μm. (G) Comparisons of PLXND1 mRNA levels in metastatic vs. primary PCa from GSE21032 and GSE35988. Data information: In (A–G), data are presented as mean ± SEM with the n of each group indicating the number of patient tumor samples included for comparisons. P values were determined by unpaired two-tailed Student’s t-test. *P < 0.05, **P < 0.01. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 3. Androgen signaling negatively regulates PlexinD1 expression in PCa.

(A) qPCR of PLXND1, SEMA3C, SEMA3E, KLK3, and NCAM1 in LNCaP, LAPC4, and VCaP cells upon culture under CSS condition for 24 h followed by R1881 stimulation (10 nM, 24 h) (n = 3 biological replicates). (B, C) Western blot of PlexinD1, AR, and PSA in LNCaP and LAPC4 cells upon treatment with 10 nM R1881 (B) or 10 µM ENZ (C) for the indicated times. (D) Genomic browser representation of AR binding at PLXND1 promoter encompassing three AREs, with the nucleotides identical to the canonical ARE highlighted in red, by interrogating AR ChIP-seq dataset GSE125245. (E) ChIP-qPCR of AR and H3K9ac occupancy at an ARE-centric PLXND1 promoter sequence as well as an AR-bound KLK3 promoter region upon R1881 stimulation (10 nM, 6 h) in LNCaP cells. Data represent the percent of input (n = 3 technical replicates). (F) Schematic diagrams of WT and mutated forms of individual AREs in PLXND1 ARE-Luc constructs. (G) Determination of WT and mutated PLXND1 ARE-Luc activities upon R1881 stimulation (10 nM, 6 h) in LNCaP cells (n = 3 biological replicates). (H) Pearson correlation analysis of PLXND1 mRNA expression with AR score in the indicated datasets from cBioPortal. (I) Pearson correlation analysis of mRNA co-expression between PLXND1 vs. different AR target genes in the indicated datasets from cBioPortal. Data information: In (A, E, G), data are presented as mean ± SEM. In (A), P values were determined by one-way ANOVA with Tukey’s multiple comparisons test. In (E), P values were determined by two-way ANOVA with Tukey’s multiple comparisons test. In (G), P values were determined by unpaired two-tailed Student’s t-test. In (H), P values were determined by Pearson’s correlation coefficient t-test. *P < 0.05, **P < 0.01; ns, not significant. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 4. PlexinD1 promotes PCa growth in vitro and in vivo.

(A) Western blot of PlexinD1 in control and PlexinD1-manipulated PCa cells. (B) Cell proliferation assays of control and PlexinD1-manipulated PCa cells. Data represent the fold changes of cell proliferation during a 7-day observation period (n = 4 biological replicates). Fold change on the day of cell seeding (day 0) in each group was set as 1. (C) Representative images and quantification of colonies formed by control and PlexinD1-manipulated PCa cells (n = 3 biological replicates). (D) GSEA of Hallmark gene sets most negatively enriched in PlexinD1-knockdown vs. control C4-2B^ENZR cells. (E) Kaplan-Meier tumor-free curves of mice inoculated with control and PlexinD1-knockdown C4-2B^ENZR cells (n = 14 tumor inoculations). (F–H) Tumor growth curves of mice bearing C4-2B^ENZR tumors (F), endpoint tumor weights (G), and anatomic tumor images (H) for the experiment described in (E) (n = 12, 3 and 1 tumor for shCon, shPlexinD1#1 and shPlexinD1#2 groups, respectively). (I–K) Tumor growth curves (I), endpoint tumor weights (J), and anatomic tumor images (K) of mice inoculated with control and PlexinD1-knockdown 22Rv1 tumors (n = 6 tumors). (L) Representative IHC staining of PlexinD1 and Ki-67 and quantification of % of Ki-67+ cells in 22Rv1 tumor samples (n = 3 tumors). Scale bars: 100 μm. Data information: In (B, C, F, G, I, J, L), data are presented as mean ± SEM. In (B, C), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test (for C4-2B^ENZR and 22Rv1) and unpaired two-tailed Student’s t-test (for LNCaP). In (D), P values were determined by permutation test. In (E), P value was determined by log-rank test. In (F, G, I, J, L), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test. *P < 0.05, **P < 0.01. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 5. PlexinD1 drives EMT, migration, invasion, and metastasis in PCa.

(A) Representative images and quantification of transwell-based cell migration and invasion by control and PlexinD1-knockdown C4-2B^ENZR and 22Rv1 cells (n = 3 biological replicates). Scale bars: 400 μm. (B) Representative images and quantitation of transwell-based cell migration and 3D spheroid cell invasion by control and PlexinD1-overexpressing LNCaP cells (n = 3 biological replicates). Scale bars: 400 μm. (C) Western blot of select EMT markers in control and PlexinD1-manipulated PCa cells. (D) GSEA plot of a Hallmark EMT gene set enriched in PlexinD1-overexpressing vs. control LNCaP cells. (E, F) Bioluminescence (BLI)-based growth curves (E) and endpoint BLI images (F) of Luc/RFP dually tagged control and PlexinD1-knockdown 22Rv1 tumors developed in an intracardiac xenograft model (n = 5 mice). (G) Representative fluorescence images and quantification of tumor metastasis developed at indicated organ sites by control and PlexinD1-knockdown 22Rv1 cells (n = 5 metastatic tumors). (H) Representative images of H&E and IHC staining of PlexinD1 and Ki-67 and quantification of % of Ki-67+ cells in control and PlexinD1-knockdown 22Rv1 tumors grown in mouse liver and adrenal glands (n = 3 metastatic tumors). Scale bars: 100 μm. Data information: In (A, B, E, G, H), data are presented as mean ± SEM. In (A), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test. In (B, E, G, H), P values were determined by unpaired two-tailed Student’s t-test. In (D), P values were determined by permutation test. *P < 0.05, **P < 0.01. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 6. PlexinD1 promotes cellular plasticity, stemness, and NE differentiation in PCa.

(A) GSEA of cell stemness and lineage-related gene sets enriched in PlexinD1-overexpressing vs. control LNCaP cells. (B) Representative images and quantification of number and size of tumorspheres formed by control or PlexinD1-manipulated PCa cells (n = 3 biological replicates). Scale bars: 100 µm. (C) Flow cytometric analysis of % of ALDH+ cells in control and PlexinD1-manipulated PCa cells, with DEAB-treated groups as negative controls (n = 3 biological replicates). (D) Western blot of stemness, luminal, basal, and NE markers as indicated in control and PlexinD1-manipulated PCa cells. (E) Representative images of control and PlexinD1-knockdown C4-2B^ENZR cell morphology and quantification of per-cell number of neurites and neurite length in each group (n = 50 cells per group). A representative of 3 independent experiments is shown. Scale bars: 20 µm. (F) Cell proliferation assays of control and PlexinD1-overexpressing LNCaP cells upon ENZ treatment (20 μM, 5 days). Data represent the fold changes of cell proliferation on day 5 relative to cell seeding day (day 0) (n = 4 biological replicates), with fold changes in non-treated groups set as 1 for normalization of paired treated groups. (G) Representative PlexinD1 and SYP IHC staining in serial sections of tumor tissue from a CRPC patient cohort and corresponding Pearson correlation analysis of protein co-expression. Scale bars: 100 µm. (H) Comparisons of PLXND1 mRNA levels in NEPC vs. Adeno or CRPC Adeno patient samples from multiple PCa clinical datasets as indicated. The n of each group indicates the number of patient tumor samples included for comparisons. (I) GSEA plots of Hallmark Androgen Response and NEPC gene sets enriched in PLXND1-high vs. -low PCa patient samples from TCGA and SU2C/PCF 2019 cohorts in cBioPortal. (J) GSEA plots of PlexinD1 targets and semaphorin/plexin signaling-related gene sets enriched in NE score-high vs. -low cells of a CRPC patient tumor as in the scRNA-seq dataset GSE137829. Data information: In (B, C, E, F, H), data are represented as mean ± SEM. In (A, I, J), P values were determined by permutation test. In (B, E), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test [for C4-2B^ENZR and 22Rv1 in (B)] and unpaired two-tailed Student’s t-test [for LNCaP in (B)]. In (C, H), P values were determined by unpaired two-tailed Student’s t-test. In (F), P values were determined by one-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01; ns, not significant. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 7. PlexinD1 activates ErbB3 and cMet in PCa.

(A) GSEA plots of transmembrane RTK-related gene sets enriched in PlexinD1-knockdown vs. control 22Rv1 cells, and PLXND1-high vs. -low PCa patient samples in TCGA cohort. (B) Representative images of a RTK phosphorylation antibody array and quantification of p-ErbB3 levels in control and PlexinD1-knockdown C4-2B^ENZR cells (n = 2 technical replicates). The raw values from measurement of p-ErbB3 spot intensity after background subtraction are presented. (C) Western blot of p-ErbB3, p-ErbB2, p-cMet, p-ERK, p-AKT, and their total protein forms in control and PlexinD1-knockdown C4-2B^ENZR and 22Rv1 cells. (D) qPCR of ErbB3 (NRG1 and NRG2) and cMet (HGF) ligand mRNA levels in control and PlexinD1-knockdown C4-2B^ENZR and 22Rv1 cells (n = 3 biological replicates). (E) Representative PLA images and quantification of PlexinD1-ErbB3/cMet interaction by per-cell cytoplasmic fluorescence intensity in C4-2B^ENZR cells treated with recombinant Sema3E/Sema3C proteins (200–500 ng/ml, 4 h) or SEMA3E/SEMA3C siRNA (10 μM, 48 h). PlexinD1 antibody incubation alone served as negative controls. The n of each group indicates the number of cells examined for quantification. Scale bars: 5 μm. (F) Co-IP assays of PlexinD1-ErbB3/cMet interaction in 22Rv1 cells. IgG was used in the IP step as negative control. Five percent of input was blotted as positive control. (G) Cell proliferation assays of 22Rv1 cells pre-treated with an ErbB3 neutralizing antibody (100 ng/ml, 24 h) or SGX-523 (5 µM, 24 h) and then subjected to PLXND1 siRNA addition followed by a 5-day observation period, and control and PlexinD1-overexpressing LNCaP cells upon treatment with an ErbB3 neutralizing antibody or SGX-523 during a 5-day observation period. Data represent the fold changes of cell proliferation on day 5 relative to siRNA treatment day (day 1) for 22Rv1 cells or treatment day (day 1) for LNCaP cells (n = 3 biological replicates), with fold changes in non-treated groups of control cells set as 1 for normalization of other groups. (H) qPCR of stemness, basal, and NE markers as indicated in 22Rv1 cells receiving an ErbB3 neutralizing antibody (100 ng/ml, 24 h), SGX-523 (5 µM, 24 h), or transient transfection of ErbB3/cMet expression plasmids followed by PLXND1 siRNA treatment for another 48 h (n = 3 biological replicates). (I) qPCR of stemness, basal, NE, and EMT markers as indicated in control and PlexinD1-overexpressing LNCaP cells upon treatment with 100 ng/ml ErbB3 neutralizing antibody or 5 µM SGX-523 for 24 h (n = 3 biological replicates). Data information: In (B, D, E, G, H, I), data are presented as mean ± SEM. In (A), P values were determined by permutation test. In (B, D), P values were determined by unpaired two-tailed Student’s t-test. In (E), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test. In (G, H, I), P values were determined by two-way ANOVA with Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01; ns, not significant. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 8. PlexinD1 induces noncanonical Gli1-dictated Hedgehog signaling in PCa.

(A) IPA upstream regulator analysis of RNA-seq data from PlexinD1-knockdown vs. control C4-2B^ENZR cells, with significantly downregulated hits shown (absolute value of z-score of ≥2, p < 0.05). (B) STRING analysis of Gli1 linked to multiple stemness, NE, and EMT marker and/or driver genes. (C) Western blot of Gli1 in control and PlexinD1-manipulated PCa cells. (D) Determination of Gli-Luc reporter activity in control and PlexinD1-manipulated PCa cells (n = 3 biological replicates). (E) qPCR of Gli1 target genes in control and PlexinD1-knockdown C4-2B^ENZR cells (n = 3 biologial replicates). (F) qPCR of Hh ligands in control and PlexinD1-knockdown C4-2B^ENZR and 22Rv1 cells (n = 3 biological replicates). (G) Determination of Gli-Luc reporter activity in 22Rv1 cells receiving an ErbB3 neutralizing antibody (100 ng/ml, 24 h), SGX-523 (5 µM, 24 h), or transient transfection of ErbB3/cMet expression plasmids followed by PLXND1 siRNA treatment for another 48 h (n = 3 biological replicates), and in control and PlexinD1-overexpressing LNCaP cells upon treatment with 100 ng/ml ErbB3 neutralizing antibody or 5 µM SGX-523 for 24 h (n = 3 biological replicates). (H) Representative images of Gli1 (red color) and α-tublin (a cytoplasm marker, green color) co-IF images and quantification of Gli1+ cells in the nuclei of control and PlexinD1-manipulated 22Rv1 and LNCaP cells under the conditions as in (G) (n = 3 biological replicates). Scale bars: 10 µm. (I) Western blot of Gli1 in nuclear and cytoplasmic fractions of control and PlexinD1-manipulated 22Rv1 and LNCaP cells under the conditions as in (G). (J) Cell proliferation assays of control and PlexinD1-overexpressing LNCaP cells upon treatment with 10 µM cyclopamine or 5 µM GANT61 for 5 days (n = 3 biological replicates). (K) Cell proliferation assays of C4-2B^ENZR and 22Rv1 cells upon treatment with 10 µM cyclopamine or 5 µM GANT61 for 5 days (n = 3 biological replicates). (L) qPCR of stemness, NE and EMT markers in control and PlexinD1-overexpressing LNCaP cells under treatment with GANT61 (5 µM, 24 h) (n = 3 biological replicates). (M) GSEA plots of two Hh signaling-related gene sets enriched in PLXND1-high vs. -low PCa patient samples from TCGA cohort. (N) Chi-square analysis of distribution of select Gli1 target gene mRNA expression (L, low; H, high) in PLXND1-low and -high PCa patient samples from TCGA cohort. Data information: In (D–H, J–L), data are presented as mean ± SEM. In (A), P values were determined by Fisher’s exact test. In (D–F, K), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test [for C4-2B^ENZR and 22Rv1 in (D)] and unpaired two-tailed Student’s t-test [for LNCaP in (D)]. In (G, H, J, L), P values were determined by one-way ANOVA with Tukey’s multiple comparisons test. In (M), P values were determined by permutation test. In (N), P values were determined by chi-square test. *P < 0.05, **P < 0.01; ns, not significant. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.

Figure 9. D1SP inhibits ErbB3/cMet-Gli1 signaling and CRPC aggressive behavior in vitro and in vivo.

(A) Graphic depicting D1SP as a recombinant PlexinD1 decoy protein. (B) Western blot of D1SP in whole cell lysate (WCL) and conditioned medium (CM) of D1SP-expressing CHO-K1 cells using a hIgGFc-specific antibody. (C) Flow cytometric analysis of binding of IgG-PE antibody conjugated D1SP at various doses in 22Rv1 cells (n = 3 biological replicates). (D) Representative PLA images and quantification of PlexinD1-Sema3E/Sema3C interaction by per-cell cytoplasmic fluorescence intensity in 22Rv1 cells treated with D1SP (1 µM, 2 h) or PBS as a vehicle. PlexinD1 antibody incubation alone served as negative controls. The n of each group indicates the number of cells examined for quantification. Scale bars: 5 µm. (E) Cell viability assays of C4-2B^ENZR and 22Rv1 cells stimulated with 200 ng/µl recombinant Sema3E protein and then subjected to treatment with 1 µM D1SP for 7 days (n = 3 biological replicates). (F) Transwell-based cell migration and invasion assays of C4-2B^ENZR and 22Rv1 cells under 1 µM D1SP treatment (n = 3 biological replicates). Scale bars: 400 µm. (G) Representative brightfield and fluorescence microscopic images and quantification of LuCaP 147CR, 49 and 173.1 PCa PDX-derived live organoids after incubation with 1 µM D1SP or PBS as a vehicle for 10 days (n = 3 biological replicates). Scale bars: 100 µm. (H–J) Tumor growth curves (H), endpoint tumor weights (I), and anatomic tumor images (J) of s.c. 22Rv1 xenografts grown in male nude mice (n = 6 tumors) and receiving intratumoral injection of D1SP (30 µg/tumor, 2–3 times per week) and saline on the right and left flanks of mice, respectively. (K, L) Representative images (K) and quantification (L) of IHC staining of Ki-67, p-ErbB3, p-ErbB2, and p-cMet (n = 6 tumors) in control and D1SP-treated 22Rv1 tumors. Scale bars: 100 μm. (M) qPCR of select Gli1 target genes in control and D1SP-treated 22Rv1 tumors (n = 6 tumors). Data information: In (C–H), data are presented as mean ± SEM. In (C, D), P values were determined by one-way ANOVA with Dunnett’s multiple comparisons test. In (E), P values were determined by one-way ANOVA with Tukey’s multiple comparisons test. In (F–H), P values were determined by unpaired two-tailed Student’s t-test. In (I, L, M), P values were determined by paired two-tailed Student’s t-test. *P < 0.05, **P < 0.01; ns, not significant. Exact P values are listed in Appendix Table S4. Source data are available online for this figure.