Categories
Prostate cancer

Activation of neural lineage networks and ARHGEF2 in enzalutamide-resistant and neuroendocrine prostate cancer and association with patient outcomes

Neural lineage network enriched in enzalutamide-resistant prostate cancer cells

It has been well established that the alteration of the neural-associated molecular landscape in castration-resistant tumors may contribute to the androgen and antiandrogens indifferences and the neuroendocrine progression after the AR-targeted treatment5,7,13. However, the overall neural lineage network underlying the end-stage emergences of these neuroendocrine markers remains unknown. We have previously generated an enzalutamide-resistant cell subline named C4-2B MDVR from C4-2B cell line through a long-term culture of C4-2B cells in the presence of increasing doses of enzalutamide22. Recent studies have shown multiple characteristics presented in C4-2B MDVR cells including overexpression of AR/AR-V7, Wnt signaling activation such as Wnt5a and WLS, and increased expression of markers of neuroendocrine such as NSE and CHGA23,24, suggesting that C4-2B MDVR cells may represent multiple cellular lineages including neuroendocrine.

To dissect the molecular changes associated with neuroendocrine differentiation in C4-2B MDVR cells, we first analyzed several well-known NED gene signatures in the transcriptome of C4-2B MDVR cells. Genes encoding several of the classic NE markers such as ENO2, CHGA, and SYP were upregulated in C4-2B MDVR cells compared to the parental C4-2B cells (Fig. 1a). Since neuroendocrine cells may derive from neural crest cells and embryonic stem cells, we analyzed the transcriptome of C4-2B MDVR cells for the pathways relative to neural stem cell proliferation and neuron differentiation and projection pathways (Supplementary Table 1) in order to determine if neural lineage genes are enriched in C4-2B MDVR cells. Among these pathways, we found that eight pathways are significantly enriched (NES <−1.4, FDR p value <0.05) in the transcriptome of C4-2B MDVR cells compared to that of parental C4-2B cells (Fig. 1b, c). The eight enriched pathways fall into four major categories, including neural stem cell differentiation, neural precursor cell proliferation, embryonic stem cell pluripotency, and neuron differentiation/projection (Fig. 1b, c). Our data suggest that neural lineages may emerge from enzalutamide-resistant cells, which may adopt a cell plasticity of early or intermediate stage of neuroendocrine differentiation.

Fig. 1: Neural lineage pathways enriched in enzalutamide-resistant prostate cancer C4-2B MDVR cells.
figure 1

a Gene and protein expression of the classical NE markers in C4-2B MDVR cells compared to the parental C4-2B cells. b Summarized GSEA analysis of enriched pathway or gene sets in C4-2B MDVR cells compared to the parental cells. The altered gene sets from Gene Ontology and PathCards were generated by GSEA software. c Enrichment plots of GSEA analyses for the neural lineage pathways in C4-2B MDVR cells compared with C4-2B parental cells. NES, showing normalized enrichment score (NES) and corresponding significance for the eight pathways. *FDR q value <0.05, **FDR q value <0.01, ***FDR q value <0.001. NE neuroendocrine, GSEA Gene Set Enrichment Analysis, FDR false discovery rate.

Neural lineage pathways and gene signatures in clinical neuroendocrine prostate cancer

To examine if the neural lineage pathways identified in C4-2B MDVR cells are presented in patients with NEPC, we performed GSEA pathway enrichment analyses on small-cell and NE prostate cancer cohorts from two multi-institutional prospective studies, mCRPC13, and treatment-emergent small-cell neuroendocrine prostate cancer (t-SCNC)5. The analyses were based on a total of 49 mCRPC cases, including 34 CRPC and 15 NEPC samples in Beltran’s study, and a total of 119 samples including 15 pure small cells, 6 mixed small cells, and 98 adenocarcinoma samples from Aggarwal’s study. We found that the eight neural lineage pathways enriched in C4-2B MDVR cells were also enriched in these NEPC patient datasets, including Neural Stem Cell Differentiation Pathways and Lineage-Specific Markers, Neural Precursor Cell Proliferation, Central Nervous System Neuron Differentiation, and Neuron Projection Development from Gene Ontology (Fig. 2a, b and Supplementary Fig. 1). The results from patient datasets suggest that the neural lineage pathways identified in C4-2B MDVR cells are also enriched in NEPC patients.

Fig. 2: Neural lineage pathways both enriched in clinical neuroendocrine prostate cancer patients and enzalutamide-resistant prostate cancer cell model.
figure 2

a Parallel comparison of neural lineage network in clinical NEPC patients and enzalutamide refractory prostate cancer cell model C4-2B MDVR. b GSEA enrichment plots of the neural lineage gene sets between the indicated groups showing NES scores and corresponding significance in the C4-2B MDVR vs. C4-2B parental cells (green), NEPC patients vs. CRPC patients (yellow), and t-SCNC patients vs. prostate adenocarcinoma (orange). *FDR q value <0.05, **FDR q value <0.01, ***FDR q value <0.001. c Venn diagram showing the number of upregulated genes from neural lineage pathways among Beltran, Aggarwal’s patient datasets (Wilcox rank-sum test p < 0.05), and C4-2B MDVR cells (|log 2-fold change|>1, FPKM value >1). NEPC neuroendocrine prostate cancer, GSEA Gene Set Enrichment Analysis, t-SCNC treatment-emergent small-cell neuroendocrine prostate cancer.

Having demonstrated that neural lineage programs are enriched in NEPC, we next analyzed the signature panel of neural lineage genes that are upregulated in the neural lineage programs. Based on the enriched neural lineage pathways aforementioned, we performed the Wilcox rank-sum test analyses and found that 239 genes were differentially upregulated in t-SCNC groups in the Aggarwal study and 165 genes in CRPC-NE cohorts in the Beltran study (p < 0.05). In parallel, we found 1060 genes upregulated in C4-2B MDVR cells (|log 2-fold change|>1, FPKM value >1). Venn diagram (Fig. 2c, displayed the commonly upregulated genes from the two patient datasets and C4-2B MDVR cells, and a collective 95 unique genes were identified as the gene panel of NLS (Fig. 2c and Supplementary Data 1). The neural lineage 95-gene panel was comprised of commonly upregulated genes in both Aggarwal and Beltran study (57 genes) and shared 7 genes and upregulated genes in C4-2B MDVR overlapped with either of the patient cohorts (13 and 18 genes for overlapped in MDVR vs. Aggarwal and MDVR vs. Beltran). Unsupervised hierarchical cluster analysis of 95 differentially expressed NLS genes (95 NLS genes) in Beltran CRPC-NE and Aggarwal t-SCNC patients were presented in Fig. 3a, b, which markedly clustered in the small-cell neuroendocrine groups. A correlation plot revealed that 95 neural lineage genes were positively correlated with these 29 NE markers and inversely correlated with AR and classical AR-targeted genes in both Beltran CRPC-NE and Aggarwal t-SCNC patients (Supplementary Fig. 2a, b). GSEA analysis also showed a significant enrichment of 95-gene NLSs in NEPC groups of these two patient datasets (Fig. 3c, d), which is consistent with the enrichment of 29 genes from the defined NEPC classifier13. Collectively, our analyses suggested that the 95 NLS genes (95 NLS) provide a molecular background giving rise to neuroendocrine differentiation in enzalutamide-resistant prostate cancer.

Fig. 3: Identified neural lineage signature genes upregulated in CRPC-NE patients.
figure 3

a Unsupervised hierarchical cluster of 95 differentially genes expressed in Beltran CRPC-NE cohorts. Pathology classification (CRPC vs. CRPC-NE) is indicated in the annotated track next to the heatmap. b GSEA enrichment plots with NES scores and FDR q values shown. *FDR q value <0.05, **FDR q value <0.01, ***FDR q value <0.001. c Hierarchical cluster analysis of 95 differentially expressed neural lineage genes in Aggarwal t-SCNC cohorts. Pathology classification (Pure/Mixed small cell/Adenocarcinoma) and molecular cluster of small-cell and non-small-cell groups is indicated in the annotation track next to the heatmap. d Enrichment plots for neural lineage signature in Aggarwal t-SCNC cohorts with NES scores and FDR q values shown. *FDR q value <0.05, **FDR q value <0.01, ***FDR q value <0.001. CRPC castration-resistant prostate cancer, CRPC-NE castration-resistant prostate cancer with neuroendocrine features, GSEA Gene Set Enrichment Analysis, NES normalized enrichment score.

The 95 neural lineage gene signatures stratified NEPC from CRPC

We next determined if the 95 neural lineage gene signatures (NLS) could be used to stratify NEPC from CRPC in two advanced CRPC databases19,25. A hierarchical clustering heatmap demonstrated that NLS genes significantly clustered in the group of small cell or adenocarcinoma with NE features in the Abida-Wassim cohort (Fig. 4a, b), which align with upregulation patterns of the aforementioned Beltran NEPC classifier genes and downregulation of AR related genes (Supplementary Fig. 3a). We also analyzed the NLS in the Labrecque study including refractory metastatic CRPC specimens (AR-positive tumors (ARPC, n = 59), AR-low tumors (ARLPC, n = 9), amphicrine tumors expressing both AR and NE markers (AMPC, n = 11), double-negative tumors (DNPC, n = 7), and tumors with small cell or NE makers without AR features (SCNPC, n = 10))19. The heatmap of NLS genes demonstrated a clear distinction between the subtypes of mCRPC specimens with GSEA enrichment significance (NES = 2.44, FDR q value <0.01) (Fig. 4c, d), which was also consistent with the expression pattern of classic NED markers and AR target genes (Supplementary Fig. 3b). We also examined our NLS genes in an experimental model, TLT331R NEPC tumor established after castration and relapsed after 24–32 weeks26. As shown in Supplementary Fig. 4, top upregulated NLS genes were displayed in the unsupervised hierarchical clustering heatmap, which aligned well with NE markers and negatively correlated AR target genes. GSEA enrichment analyses also showed that NLS genes were significantly enriched in TLT331R NEPC tumor groups compared with prostate adenocarcinoma groups. In summary, these data indicated that these differentially expressed NLS genes could stratify prostate cancer with neuroendocrine differentiation from prostate adenocarcinoma.

Fig. 4: The 95 NLS gene signatures stratify prostate cancer with neuroendocrine differentiation from prostate adenocarcinoma.
figure 4

a Hierarchical cluster of neural lineage signature genes in Wassim small-cell cohorts. Pathology classification (small cell, adenocarcinoma with neuroendocrine (NE) features, adenocarcinoma, not available and inadequate for the diagnosis) are indicated in the annotation track next to the heatmap. b GSEA enrichment plots for neural lineage signature in small cell and adenocarcinoma with NE feature groups with NES and FDR q values shown. c Hierarchical cluster heatmap of neural lineage signature genes in Labrecque cohorts. Molecular characteristic classification (AR high (ARPC), AR low (ARLPC), amphicrine expression of both AR and NE markers (AMPC), double-negative expression of both AR and NE markers (DNPC), and small-cell neuroendocrine SCNPC) are indicated in the annotation track next to the heatmap. d GSEA enrichment plots of neural lineage signature genes in Labrecque cohorts with NES score and FDR q values. NLS neural lineage signature, GSEA Gene Set Enrichment Analysis. *p < 0.05, **p < 0.01, ***p < 0.001.

Higher levels of ARHGEF2, EPHB2, and LHX2 expression correlate with poor survival in castration-resistant prostate cancer

Among the genes from the 95 NLS, we further analyzed 7 of the neural lineage genes (ARHGEF2, EPHB2, LXH2, DPYSL3, EPHB2, FYN, and GNG4) shared among C4-2B MDVR cells, Beltran, Aggarwal, Abida-Wassim and Labrecque datasets5,13,19,25. Our data showed that all the seven genes were upregulated in CRPC-NE/small-cell groups compared to the CRPC-adeno group across the four databases (Fig. 5a). To determine whether the seven NLS genes were associated with survival in prostate cancer patients, we further conducted the Kaplan–Meier survival analysis and log-rank test to determine the correlation of expression of the neural lineage genes to the overall survival in prostate cancer patients. In 75 out of 266 prostate cancer patients with overall survival of the first-line AR-targeted inhibitors treatment in the Abida-Wassim study25, higher expression of ARHGEF2 (p = 0.041), LHX2 (p = 0.0091), and EPHB2 (p = 0.15) showed correlation with shortened overall survival time (Fig. 5b), while DPYSL3, EPHB2, FYN, and GNG4 did not positively correlate with shortened overall survival. We also performed the Kaplan–Meier survival analysis on 148 patients with both disease-free survival information and RNA sequencing data available from the MSKCC study, which included 131 primary tumors and 9 metastases27. The data revealed that higher expression of ARHGEF2 (p = 0.011), LHX2 (p = 0.0091), and EPHB2 (p = 0.0019) correlate with shorter disease-free time, respectively (Fig. 5c), while DPYSL3, EPHB2, FYN, and GNG4 did not reach statistical significance (data not shown). Collectively, these results suggest the potential of ARHGEF2, LHX2, and EPHB2 as indicators of poor survival for advanced prostate cancer.

Fig. 5: Upregulated neural lineage genes correlate with poor overall survival in prostate cancer patients.
figure 5

a Gene expressions of the seven neural lineage signature genes were presented in the indicated groups of Beltran, Aggarwal, Labrecque, and Wassim cohorts. Gene expression from CRPC patients in Beltran study, prostate adenocarcinoma samples from Aggarwal, Labrecque, and Wassim study were presented in blue, while that of CRPC-NE samples from Beltran study, small-cell neuroendocrine prostate cancer from Aggarwal, Labrecque and Wassim studies was presented in purple. CRPC castration-resistant prostate cancer; CRPC-NE castration-resistant prostate cancer with neuroendocrine features; Adeno; non-SC non-small-cell prostate adenocarcinoma; Small-Cell small-cell neuroendocrine prostate cancer; prostate adenocarcinoma; SCNPC small-cell neuroendocrine prostate cancer; SC + NE small-cell neuroendocrine prostate cancer. *p < 0.05, **p < 0.01, ***p < 0.001, using unpaired t-test (median, quartiles and distribution of all data points presented). b Kaplan–Meier analysis of neural lineage signature genes including ARHGEF2, LHX2, and EPHB2 for the overall survival in Wassim cohorts. c Kaplan–Meier analysis of neural lineage signature genes for disease-free survival in MSKCC cohorts using log-rank test.

Downregulation of ARHGF2 expression suppresses viability and neuroendocrine markers of C4-2B MDVR and H660 cells

We analyzed ARHGEF2, LHX2, and EPHB2 gene expression in the RNA sequencing data from GSE154576 (DeLucia et al., 2021). As shown in Fig. 6a, gene expression of ARHGEF2, LHX2, and EPHB2 were significantly upregulated in NEPC MSKCC-EF1 and H660 compared to 22RV1 and LNCaP95 cell lines. Quantification of ARHGEF2, LHX2, and EPHB2 mRNA levels verified that these three genes increased in enzalutamide-resistant C4-2B MDVR compared with C4-2B parental cells, and further increased in H660 cells (Fig. 6b). We next focused on the effect of ARHGEF2 on cell growth and neuroendocrine differentiation by knocking down ARHGEF2 expression using siRNA in C4-2B MDVR and H660 cells. Knocking down of ARHGEF2 expression downregulates CHGA, NSE, and SYP protein expression (Fig. 6c) and inhibits cell viability (Fig. 6d) in both C4-2B MDVR and H660 cells. Furthermore, we analyzed LuCaP49 neuroendocrine PDX tumors and LuCaP35CR castration-resistant PDX tumors for their expression of AR-targeted genes and NED markers and found that LuCaP49 tumors express higher levels of NE markers such as CHGA, NSE, and SYP than LuCaP35CR tumors (Fig. 6e), consistent with the characteristics of NED for LuCaP49 and CRPC for LuCaP35CR28. We also found that ARHGEF2 mRNA levels were much higher in LuCaP49 tumors than LuCaP35CR tumors (Fig. 6e). Knocking down ARHGEF2 expression through siRNA significantly inhibits the viability and growth of organoids derived from LuCaP49 PDX tumors (Fig. 6f). Collectively, these data suggest that ARHGEF2 could serve as a potential therapeutic target for NEPC.

Fig. 6: Downregulating neural lineage marker ARHGEF2 gene expression decreased C4-2B MDVR and H660 cell viability.
figure 6

a FPKM value of ARHGEF2, LHX2 and EPHB2 gene expression in 22RV1, LNCaP95, MSKCC-EF1 and H660 cells from GSE154576 dataset (n = 2). b Relative mRNA level of ARHGEF2, LHX2, and EPHB2 in C4-2B parental, MDVR, and H660 cells (n = 4). c Knocking down ARHGEF2 by siRNA decreased the expression of the markers of neuroendocrine including CHGA, NSE, and SYP in C4-2B MDVR and H660 cells. C4-2B MDVR and H660 cells were transfected with siRNAs for ARHGEF2 for 48 h, total cell lysates were collected and subjected to western blot analysis. d Knocking down ARHGEF2 by siRNA inhibited cell viability of C4-2B MDVR and H660 cells. C4-2B MDVR and H660 cells were transiently transfected with siRNAs targeting ARHGEF2 and viable cells were quantified using CellTiter Glo assay (n = 5). e qPCR analysis of AR targets and neuroendocrine markers and ARHGEF2 in NEPC LuCaP49 patient-derived xenograft (PDX) tumors and prostate adenocarcinoma LuCaP35CR tumors (n = 3). f Downregulation of ARHGEF2 by siRNA inhibits the growth of organoids derived from LuCaP49 PDX tumors. LuCaP49 organoids were seeded in a 96-well plate in the format of 3D Matrigel and then transfected with 50 nM ARHGEF2 siRNA and cultured for 14 days. The viability of the organoids was analyzed by CellTiter Glo and visualized by LIVE/DEAD staining (n = 4). Green = Calcein staining of live cells, Red = Ethidium homodimer-1 staining of dead cells. Scale bar 100 µm. *p < 0.05, **p < 0.01, ***p < 0.001 (mean and SD in a, b, d, e, f) using one-way ANOVA with multiple-comparisons test.