ABSTRACT
Neuroendocrine prostate cancer (NEPC) is a highly aggressive and treatment‐resistant subtype of castration‐resistant prostate cancer (CRPC) that often emerges during progression under androgen‐receptor (AR) pathway inhibition. While lineage plasticity in cancer cells has been recognized as a key mechanism of resistance, the role of the tumor microenvironment in driving this transition remains unclear. Among its cellular components, vascular endothelial cells can undergo endothelial‐mesenchymal transition (EndoMT), a phenotypic shift associated with tumor progression and fibrosis. Here, we investigated whether EndoMT contributes to NEPC development. Human umbilical vein endothelial cells (HUVEC) were induced to undergo EndoMT using IL‐1β and TGF‐β2, and are hereafter referred to as EndoMTed HUVEC. EndoMTed HUVEC promoted neuroendocrine features and functional changes in LNCaP cells. Transcriptome analysis revealed marked upregulation of granulocyte‐macrophage colony‐stimulating factor (GM‐CSF) in EndoMTed HUVEC. Neutralization of GM‐CSF signaling using mavrilimumab, a monoclonal antibody targeting the GM‐CSF receptor alpha (CSF2RA), and siRNA‐mediated CSF2RA knockdown both suppressed the neuroendocrine phenotype and STAT3 signaling of LNCaP cells. Conversely, GM‐CSF stimulation alone reproduced these changes. Enzalutamide‐treated LNCaP cells secreted IL‐1β and TGF‐β2, which in turn triggered EndoMT, suggesting a reciprocal loop. These findings indicate that anti‐androgen therapy may inadvertently promote NEPC through a paracrine loop involving tumor‐derived cytokines and endothelial GM‐CSF secretion, highlighting EndoMT as a microenvironmental driver of treatment resistance.
Keywords: androgen deprivation therapy, endothelial‐mesenchymal transition, granulocyte‐macrophage colony‐stimulating factor, neuroendocrine differentiation, prostate cancer
Endothelial‐mesenchymal transition (EndoMT) contributes to neuroendocrine differentiation (NED) in prostate cancer. Using in vitro co‐culture models, we show that EndoMTed endothelial cells promote NED in prostate cancer cells via GM‐CSF signaling. These findings reveal a previously unrecognized paracrine mechanism by which the tumor microenvironment drives therapy‐induced NED.
Abbreviations
- αSMA
Alpha smooth muscle actin
- AR
Androgen receptor
- CAF
Cancer‐associated fibroblast
- CD31
Platelet endothelial cell adhesion molecule‐1 (PECAM‐1)
- CgA
Chromogranin A
- CRPC
Castration‐resistant prostate cancer
- CSF2
Colony‐stimulating factor 2 (gene encoding GM‐CSF)
- CSF2RA
Colony‐stimulating factor 2 receptor alpha subunit
- EMT
Epithelial‐mesenchymal transition
- EndoMT
Endothelial‐mesenchymal transition
- GM‐CSF
Granulocyte‐macrophage colony‐stimulating Factor
- HUVEC
Human umbilical vein endothelial cell
- NE
Neuroendocrine
- NED
Neuroendocrine differentiation
- NEPC
Neuroendocrine prostate cancer
- PSA
Prostate‐specific antigen
- TME
Tumor microenvironment
1. Introduction
Tumor cells exist within a complex and dynamic microenvironment composed of stromal, immune, and endothelial cells. The tumor microenvironment (TME) is now recognized as essential in cancer progression, supporting tumor growth, metastatic dissemination, and therapy resistance [1]. Endothelial cells, besides forming blood vessels, exhibit phenotypic plasticity and can transdifferentiate into mesenchymal‐like cells via endothelial‐mesenchymal transition (EndoMT) [2].
Initially described in the context of cardiac development [3], EndoMT has since been observed in various cancers, where it contributes to tumor progression, metastasis, and therapeutic resistance [4]. Through EndoMT, endothelial cells can become a source of cancer‐associated fibroblasts (CAFs) [5], thereby promoting fibrosis and remodeling the TME to favor malignancy.
Prostate cancer exemplifies how interactions with the TME can drive tumor cell plasticity. Advanced prostate adenocarcinomas under therapeutic pressure can undergo lineage plasticity to acquire neuroendocrine (NE) features [6]. This neuroendocrine differentiation (NED) is characterized by loss of androgen receptor (AR) signaling and epithelial markers, gain of NE markers such as chromogranin A (CgA) and synaptophysin (SYP). This differentiation often leads to an aggressive clinical course [7, 8]. Neuroendocrine prostate cancer (NEPC) typically emerges as a treatment‐induced adaptation rather than a de novo subtype [9, 10]. Patients with treatment‐related NEPC often present with visceral metastases instead of bone metastases, implicating the TME in facilitating this lineage transition [11]. Recent studies highlight the role of inflammatory cytokines in the TME in promoting such plasticity. For example, interleukin‐6 (IL‐6) activates the signal transducer and activator of transcription 3 (STAT3) signaling pathway, which in turn induces NE phenotypes in prostate tumor cells [12]. Additionally, interleukin‐8 (IL‐8) has been reported to suppress AR activity and promote NE‐like features in prostate cancer cells [13].
Given that EndoMT can confer a pro‐inflammatory, fibroblast‐like secretory profile on endothelial cells [14], we hypothesized that EndoMT within the TME might drive NED in adjacent prostate cancer cells via cytokine‐mediated signaling.
Specifically, we proposed that EndoMT‐derived cytokines could activate pathways, such as the JAK/STAT pathway, analogous to IL‐6‐mediated mechanisms [15]. To investigate this, we used an in vitro model of prostate cancer cells co‐cultured with EndoMT human endothelial cells. This study aimed to determine whether the EndoMT‐conditioned microenvironment induces NED in prostate cancer cells and to identify the key molecular mediators involved in this process.
2. Materials and Methods
2.1. Human Samples
Prostate tissue specimens used in this study were carefully selected from surgical specimens of patients with NEPC with complete clinicopathological data. The prostate cancer tissues were acquired via prostate tumor biopsy or transurethral resection. Samples were embedded in paraffin and subjected to immunofluorescence staining.
2.2. Antibodies
The antibodies used in this study for western blotting and immunofluorescence staining, along with their sources, catalog numbers, RRIDs, and working concentration, are detailed in Table S1.
2.3. Cell Sources and Cell Culture
LNCaP, C4‐2, PC3, and DU145 prostate cancer cell lines were obtained from ATCC (Manassas, VA, USA). HUVEC were purchased from Lonza Biosciences (Lonza, C2519A, Basel, Switzerland). Cell culture conditions are detailed in Doc S1.
2.4. Induction of Endothelial‐Mesenchymal Transition
EndoMT was induced in HUVEC by interleukin‐1 beta (IL‐1β) and human transforming growth factor beta 2 (TGF‐β2) as previously described [16]. The treated cells are hereafter termed EndoMTed HUVEC. Detailed protocols are provided in Doc S2.
2.5. Co‐Culture of Prostate Cancer Cell Lines With EndoMTed HUVEC
Prostate cancer cells were co‐cultured with untreated or EndoMTed HUVEC in a two‐chamber system under androgen‐deprived conditions. HUVEC were seeded in inserts and induced to undergo EndoMT prior to co‐culture, as detailed in Doc S3.
2.6. Cell Quantification
Cells were trypsinized and collected to quantify viable cells using an automated cell counter (AMQAX1000, Thermo Fisher Scientific [Thermo], Waltham, MA, USA) with trypan blue exclusion, according to the manufacturer's instructions.
2.7. Cell Migration Assay
Transwell migration assays (12‐well, 8.0 μm pores) were used to assess LNCaP cells' migration toward conditioned media from EndoMTed or control HUVEC, or granulocyte‐macrophage colony‐stimulating factor (GM‐CSF)‐supplemented medium, as detailed in Doc S4.
2.8. Immunofluorescence Staining
Cells were fixed with 4% paraformaldehyde, permeabilized, and blocked before incubation with primary antibodies. Antigen retrieval was performed for paraffin‐embedded sections. Detailed protocols are provided in Doc S5.
2.9. RNA Extraction, cDNA Preparation, Quantitative PCR, and TaqMan Custom PCR Array Plates Analysis
Total RNA was extracted (RNeasy Mini Kit, Qiagen, 74104, Venlo, Netherlands) and reverse‐transcribed (High‐Capacity RNA‐to‐cDNA Kit, Thermo, 4387406). Quantitative PCR (qPCR) was performed using TaqMan assays on a QuantStudio 1 Real‐Time PCR system (Thermo), with expression levels normalized to GAPDH or 18S rRNA. Relative expression levels were calculated using the ΔΔCt method from three independent experiments. Protocol and assay details are provided in Docs S6 and S7.
2.10. ELISA
Aliquots of conditioned medium were analyzed by ELISA (Quantikine kits: IL‐1β [DLB50], TGF‐β2 [DB250], GM‐CSF [DGM00]; Bio‐Techne, Minneapolis, MN, USA) to quantify protein levels per the manufacturer's instructions.
2.11. Western Blot Analysis
Cells were lysed in RIPA buffer with protease and phosphatase inhibitors. Proteins (30 μg) were separated by SDS‐PAGE and transferred to PVDF membranes. Protein bands were visualized using enhanced chemiluminescence and quantified by densitometry using Image (RRID: SCR_003070) [17]. Detailed protocols are provided in Doc S8.
2.12. RNA Sequencing and Transcriptome Analysis
Libraries were sequenced on a NovaSeq 6000 platform (Illumina, San Diego, CA, USA). Reads were processed using Trimmomatic, STAR, featureCounts, and edgeR, and differentially expressed genes were identified with a false discovery rate (FDR) < 0.01. The resulting data were analyzed using RStudio (RRID: SCR_000432). Detailed protocols are provided in Doc S9.
2.13. GM‐CSF Blockade Using Mavrilimumab
To evaluate the effect of GM‐CSF signaling inhibition on NED, mavrilimumab, a human monoclonal antibody targeting the GM‐CSF receptor α subunit (CSF2RA) (Selleck Chemicals, A2532, Houston, TX, USA), was added to the co‐culture at 10 μg/mL under androgen‐deprived conditions. Methods are detailed in Doc S10.
2.14. siRNA‐Mediated Knockdown of CSF2RA
To genetically suppress GM‐CSF receptor signaling, LNCaP cells were transfected with siRNAs targeting CSF2RA (Silencer Select siRNA, Thermo, s3600 and s3601) using Lipofectamine RNAiMAX Transfection Reagent (Thermo, 13778030). Detailed procedures are described in Doc S11.
2.15. Stimulation of LNCaP Cells With Recombinant GM‐CSF
To evaluate the effect of GM‐CSF on NED, LNCaP cells were cultured under androgen‐deprived conditions and treated with recombinant human GM‐CSF (PeproTech, 300–03, Cranbury, NJ, USA) at a final concentration of 1 nM. After 96 h of stimulation, cells were harvested for gene and protein expression analysis. Protocols are detailed in Doc S12.
2.16. Colocalization Analysis
Colocalization analysis was performed to evaluate the spatial relationship between NE cells (CgA‐positive) and vascular endothelial cells (CD31‐positive) in the immunofluorescence‐stained sections. Image processing and analysis were conducted using Fiji (RRID: SCR_002285) [18]. The Coloc2 plugin was used for colocalization analysis, and Pearson's correlation coefficient (PCC) was calculated to assess the degree of colocalization between green (CgA) and red (CD31) fluorescence signals [19]. Detailed protocols are provided in Doc S13.
2.17. Statistical Analysis
All experiments were performed in triplicate. Data are shown as mean ± SD. Group differences were assessed by Student's t‐test or one‐way ANOVA, with significance set at p ≤ 0.05. Analyses were performed using SPSS v29.0 (RRID: SCR_002865).
3. Results
3.1. Treatment With IL‐1β and TGF‐β2 Induces Endothelial‐Mesenchymal Transition in Human Vascular Cells
As previously described [20], we assessed whether co‐stimulation with IL‐1β and TGF‐β2 could induce synergistic induction of EndoMT in HUVEC. As shown in Figure S1A, HUVEC changed their morphology from a typical cobblestone‐like appearance to a spindle‐shaped, fibroblast‐like phenotype with reduced cell–cell adhesion and more prominent stress fibers, characteristic of mesenchymal cells by IL‐1β and TGF‐β2 stimulation, which are the hallmarks of EndoMT. Immunofluorescence staining showed decreased CD31 expression and increased αSMA expression (Figure S1A). mRNA expression analysis of EndoMTed HUVEC showed a decrease in the expression of endothelial markers (CD31, CDH5) and an increase in the expression of mesenchymal markers (ACTA2, TAGLN) and CAF markers (FAP, PDGFR), indicating a potential transition toward a CAF‐like phenotype (Figure S1B). Thus, we confirmed that co‐stimulating with IL‐1β and TGF‐β2 caused the synergistic induction of HUVEC to a mesenchymal phenotype called EndoMTed HUVEC.
3.2. Co‐Culture With EndoMTed HUVEC Induces Neuroendocrine‐Like Differentiation of Androgen‐Dependent Prostate Cancer Cells, LNCaP Cells
As tumor plasticity is modulated by the TME, we evaluated whether EndoMTed HUVEC could induce NED of prostate cancer cells. As shown in Figure 1A, LNCaP cells co‐cultured with EndoMTed HUVEC showed an increased expression of chromogranin A (CgA, CHGA), a NE marker. Western blotting further demonstrated elevated CgA levels and significantly decreased levels of AR in the co‐cultured LNCaP cells (Figure 1C). LNCaP cells co‐cultured with EndoMTed HUVEC showed significantly increased cell proliferation compared to those cultured alone or with untreated HUVEC (Figure 2A,B). Consistently, a Transwell migration assay using conditioned medium from EndoMTed HUVEC also demonstrated a significant increase in LNCaP cell motility (Figure 2C,D).
FIGURE 1.
Neuroendocrine differentiation‐like changes of LNCaP cells after co‐cultured with EndoMTed HUVEC. (A) Immunofluorescence staining of LNCaP cells co‐cultured with EndoMTed HUVEC (w/EndoMTed HUVEC), untreated HUVEC (w/HUVEC), or cultured alone, showing chromogranin A (CgA) expression. Scale bar: 200 μM. (B) CHGA mRNA expression in LNCaP cells under the same conditions, assessed by qPCR. (C) Western blotting of CgA and androgen receptor (AR) in LNCaP cells co‐cultured with EndoMTed HUVEC: β‐actin serves as a loading control. Representative of three independent experiments.
FIGURE 2.
Phenotypical changes in androgen‐dependent prostate cancer cells, LNCaP, co‐cultured with EndoMTed HUVEC. (A) Brightfield microscope images of LNCaP cells alone, co‐cultured with untreated HUVEC (w/HUVEC), or with EndoMTed HUVEC (w/EndoMTed HUVEC) 24 h after seeding (start of co‐culture) and 96 h after co‐culturing. Scale bar 200 μM. (B) Cell proliferation rates at 96 h relative to 24 h under each condition. (C, D) Transwell migration assay using conditioned medium (CM) from HUVEC and EndoMTed HUVEC. Migrated cells were counted in five random fields per insert (20×). The average number of migrated cells per field was calculated. *p ≤ 0.05. (E) Heatmap of the cDNA microarray expression profiles in LNCaP cells under the same three conditions. (F) Western blotting of AR, CgA, FOXA2, FOXA1, and PSA. β‐actin was used as a loading control. Representative image shown. (G) Western blotting and qPCR of PSA and CHGA following 96 h co‐culture with EndoMTed HUVEC, and recovery culture in normal (androgen‐containing) medium. The medium was refreshed once at 48 h. AR, androgen receptor; CgA, chromogranin A; FOXA1, forkhead box A1; FOXA2, forkhead box A2; PSA, prostate‐specific antigen.
To further confirm this transition, gene expression analysis using cDNA microarray revealed upregulation of NE‐related transcriptional programs, including CHGA, SYP, FOXA2, AURKA, and MYCN (Figure 2E). At the protein level, western blotting demonstrated increased expression of CgA and FOXA2 (Forkhead box A2), and decreased expression of AR, PSA, and FOXA1 (Forkhead box A1) in LNCaP cells co‐cultured with EndoMTed HUVEC (Figure 2F). These findings indicate that EndoMTed HUVEC promotes NED and acquisition of malignant phenotypes in androgen‐dependent prostate cancer cells.
We also checked whether the phenomenon is reversible or not. We observed that if LNCaP cells co‐cultured with EndoMTed HUVEC were returned to androgen‐containing, standard RPMI‐1640 medium with 10% FBS relatively early (after 96 h of co‐culturing), they began to re‐express PSA to some extent, and their CHGA levels declined (Figure 2G). This suggests that the NE phenotype was not fully fixed, at least in the intermediate stage.
To assess whether the NED induced by EndoMTed HUVEC is a generalizable phenomenon among prostate cancer cell lines, we performed co‐culture experiments using additional cell lines. In AR‐positive C4‐2 cells, co‐culture with EndoMTed HUVEC led to increased expression of CHGA, similar to what was observed in LNCaP cells (Figure S2). In contrast, AR‐negative prostate cancer cell lines PC3 and DU145 showed no significant increase in the expression of the NE marker CHGA (Figure S2). Based on these results, we selected LNCaP cells as a representative AR‐positive prostate cancer model for subsequent experiments evaluating the mechanistic basis of EndoMT‐induced NED.
3.3. Granulocyte‐Macrophage Colony‐Stimulating Factor Might Participate in the Transition of Prostate Cancer Cells to Neuroendocrine Cells
We comprehensively analyzed gene expressions in HUVEC and EndoMTed HUVEC using RNA‐Seq to investigate the factors excreted by EndoMTed HUVEC that might induce NED of LNCaP cells. Several genes were significantly upregulated in EndoMTed HUVEC (Figure 3A, S3A). Among these factors, we focused on the role of GM‐CSF (CSF2), which showed the most significant differences between the two cell types (Figure 3B). The upregulation of CSF2 in EndoMTed HUVEC was validated by qPCR (Figure S3B). ELISA also demonstrated increased GM‐CSF secretion in EndoMTed HUVEC culture medium (Figure S3C). GM‐CSF is a pro‐inflammatory cytokine that can act on various cells in the TME [21]. We hypothesized that GM‐CSF from EndoMT cells might lead to NED.
FIGURE 3.
Expression analysis between HUVEC and EndoMTed HUVEC. (A) Heatmap of RNA‐seq analysis showing upregulated genes in EndoMTed HUVEC (n = 3). (B) Scatter plot highlighting differentially expressed genes (p‐value < 0.05, FDR < 0.01, Fold Change > 1.5). FDR, False Discovery Rate.
To determine the functional relevance of GM‐CSF signaling in NED, we first examined the activation of downstream effectors in LNCaP cells co‐cultured with EndoMTed HUVEC. Western blotting showed an increased phosphorylation of STAT3 (Y705) and upregulation of MYC (c‐MYC) (Figure 4A). This molecular shift coincided with the elevated expression of NE markers, such as CgA and FOXA2, and the downregulation of AR, PSA, and FOXA1 (Figure 2F).
FIGURE 4.
GM‐CSF–STAT3–MYC axis mediates neuroendocrine differentiation in LNCaP cells. (A) Western blotting of STAT3/MYC signaling pathway in LNCaP cells co‐cultured with HUVEC (w/HUVEC) or EndoMTed HUVEC (w/EndoMTed HUVEC). (B) Western blotting of NE markers and STAT3–MYC (c‐MYC) signaling in LNCaP cells co‐cultured with EndoMTed HUVEC and treated with mavrilimumab. (C) Western blotting of NE markers and STAT3–MYC signaling after siRNA‐mediated CSF2RA knockdown in LNCaP cells. (D) Western blotting of NE markers in LNCaP cells stimulated with recombinant GM‐CSF under androgen‐deprived conditions. (E) Transwell migration assay of LNCaP culture stimulated with GM‐CSF under androgen‐deprived conditions. (F) Quantification of migrated LNCaP cells; five random fields per insert were counted under a light microscope at 20× magnification. *p ≤ 0.05. CgA, chromogranin A; CSF2RA, colony‐stimulating factor 2 receptor alpha subunit; GM‐CSF, granulocyte‐macrophage colony‐stimulating factor; NE, neuroendocrine.
We next performed two independent intervention experiments to validate whether GM‐CSF plays the role through the GM‐CSF receptor α chain (CSF2RA). First, treatment with mavrilimumab, a monoclonal antibody targeting CSF2RA, significantly reduced STAT3 phosphorylation, c‐MYC expression, and suppressed NE marker induced in LNCaP cells co‐cultured with EndoMTed HUVEC (Figure 4B). Second, siRNA‐mediated knockdown of CSF2RA in LNCaP cells prior to exposure to co‐culturation from EndoMTed HUVEC also attenuated NE marker expression and STAT3 phosphorylation (Figure 4C). These results indicate that CSF2RA is essential for transducing GM‐CSF signals and might participate in the activation of the STAT3–c‐MYC pathway.
To confirm that GM‐CSF alone is sufficient to recapitulate these phenotypic changes, we directly treated LNCaP cells with recombinant GM‐CSF under androgen‐deprived conditions. This stimulation led to increased expression of CgA and FOXA2 while suppressing FOXA1 expression—molecular patterns consistent with those observed in the co‐culture system (Figure 4D). In addition, GM‐CSF‐treated LNCaP cells exhibited enhanced proliferation and migratory capacity (Figure 4E,F, S4), further supporting the notion that GM‐CSF secreted by EndoMTed HUVEC might act as a key paracrine effector, inducing NED of prostate cancer cells and activating the STAT3‐c‐MYC signaling cascade.
3.4. Androgen‐Receptor Blockade Induces IL‐1β and TGF‐β2 Excretion From LNCaP Cells and Promotes EndoMT of HUVEC
We further investigated whether IL‐1β and TGF‐β2 would be secreted by prostate cancer cells through AR blockade. Treatment with enzalutamide led to elevated secretion of cytokines, IL‐1β and TGF‐β2, in LNCaP cells (Figure 5A). Furthermore, conditioned medium from enzalutamide‐treated LNCaP cells induced EndoMT in HUVEC (Figure 5B). These results suggested that anti‐androgen therapy might promote endothelial cells to EndoMT partly through the secretion of IL‐1β and TGF‐β2 by prostate cancer cells.
FIGURE 5.
Elevated secretion of IL‐1β and TGF‐β2 in LNCaP cells, and EndoMT induction of HUVEC by LNCaP conditioned medium. (A) ELISA of IL‐1β and TGF‐β2 in culture media from LNCaP cells treated with or without enzalutamide (5 nM for 96 h, Selleck Chemicals, S1250). (B) Immunofluorescence staining of endothelial marker CD31 and mesenchymal marker αSMA in HUVEC exposed to conditioned media from untreated or enzalutamide‐treated LNCaP cells for 96 h, with one medium change at 48 h. *p ≤ 0.05. CD31, platelet endothelial cell adhesion molecule‐1; αSMA, alpha smooth muscle actin.
3.5. Vascular Endothelial Cells Closely Reside With Neuroendocrine‐Differentiated Prostate Cancer Cells in Human Castration‐Resistant Prostate Cancer Tissues
Finally, we examined whether prostate cancer cells and vascular endothelial cells closely reside in human CRPC tissues. We selected three CRPC tissue samples with NED. We found that immunofluorescence staining revealed the proximity of vascular endothelial and NE cells (Figure 6). Colocalization analysis using ImageJ software quantified the spatial relationship between the two cell types, with an average PCC of 0.47, indicating a moderate level of colocalization. This quantitative evaluation provided valuable insights into the interactions between NE and vascular endothelial cells within the TME.
FIGURE 6.
Histopathological features of human castration‐resistant prostate cancer tissues. Representative images of hematoxylin and eosin (HE) and immunofluorescence (IF) staining from three castration‐resistant prostate cancer (CRPC) cases with neuroendocrine differentiation (A–C). Colocalization analysis using ImageJ software quantitatively assessed the spatial relationship between vascular endothelial cells (CD31‐positive) and neuroendocrine cells (CgA‐positive), with Pearson's correlation coefficients (PCCs) of 0.39, 0.53, and 0.40 for patients A, B, and C, respectively (average PCC: 0.47). Green: CgA; Red: CD31; Blue: Nuclei. Scale bar: 50 μm. Clinical characteristics of each patient are summarized in Table S2. CgA, chromogranin A; CD31, platelet endothelial cell adhesion molecule‐1 (PECAM‐1).
4. Discussion
In this study, we demonstrate that EndoMT within the prostate TME can directly promote NED of cancer cells via a paracrine mechanism. Endothelial cells induced to undergo EndoMT secreted factors that drove adjacent prostate cancer cells toward an NE‐like phenotype, characterized by the expression of NE markers. Among these factors, GM‐CSF was a candidate that activated STAT3 signaling in tumor cells and upregulated c‐MYC, ultimately facilitating NED and cell proliferation. These findings reveal an interesting microenvironment‐driven mechanism of tumor cell plasticity with significant implications for prostate cancer progression and therapy.
Our results reveal an interaction between tumor endothelial cells and cancer cells. EndoMT is increasingly recognized as an adaptive process in cancerous tissue, yielding mesenchymal‐like cells (often akin to CAFs) that can promote tumor growth, invasiveness, and therapeutic resistance [22]. By generating cells that resemble CAFs, EndoMT supplies the tumor with pro‐tumorigenic stroma [23]. Our co‐culture experiments with EndoMT‐conditioned endothelial cells demonstrate that, beyond generating CAFs that remodel the extracellular matrix and promote angiogenesis, EndoMT also influences cancer cells directly by promoting phenotypic changes. These findings suggest that EndoMT is not a passive component of the TME but an active driver of tumor cell plasticity and heterogeneity.
One of the most clinically relevant aspects of NED in prostate cancer is its association with treatment resistance and poor outcomes [24]. This differentiation entails loss of AR signaling and gain of neuronal properties, effectively allowing tumor cells to bypass dependence on hormonal growth signals [11]. Our results provide experimental support to clinical observations that factors in the microenvironment facilitate NED. Prior work has implicated inflammatory cytokines in this process—for instance, IL‐6 produced by infiltrating immune cells or stroma can induce NE characteristics in prostate cancer via STAT3 activation [6]. In line with this, we found that an EndoMT‐derived cytokine, GM‐CSF, can substitute as a key stimulus for NED. GM‐CSF has been traditionally studied in the context of immune cell regulation and as an immunotherapy adjuvant (e.g., in the prostate cancer vaccine Sipuleucel‐T) [25], but our findings suggest it may also act directly on tumor cells to alter their phenotype [26].
A major finding of this study is the identification of the GM‐CSF secreted by EndoMTed HUVEC through CSF2RA as a contributing pathway for NED in prostate cancer. Additionally, this pathway upregulated the STAT3–c‐MYC pathway, which might contribute to the cell proliferation of prostate cancer cells. Neutralizing GM‐CSF signaling with mavrilimumab (an antibody against the GM‐CSF receptor‐α) or silencing its receptor (CSF2RA) suppressed NE marker expression, diminished STAT3 activation, and reduced c‐MYC expression. These findings suggest that the GM‐CSF/CSF2RA pathway might be a key component of the inflammatory network driving lineage plasticity and be a promising target of treatment‐induced NEPC.
From a translational perspective, our finding that factors secreted from EndoMTed endothelial cells promote NED in prostate cancer suggests several potential therapeutic opportunities. One approach may involve targeting EndoMT itself: preventing or reversing this process in tumor‐associated endothelial cells could attenuate the paracrine signaling that promotes NEPC emergence. Given that EndoMT is primarily driven by TGF‐β and other cytokines [27], the inhibition of TGF‐β signaling could simultaneously suppress stromal EndoMT and tumor cell transitions such as epithelial‐mesenchymal transition (EMT) and NED [28]. Therapies targeting GM‐CSF have already shown clinical efficacy in inflammatory diseases such as rheumatoid arthritis and giant cell arteritis [29, 30], and have also been explored in conditions like COVID‐19‐associated lung injury [31]. While their application in oncology remains limited, our findings suggest that GM‐CSF blockade could be repurposed as a novel strategy for treating CRPC, particularly in the context of NEPC transformation.
Notably, hormone therapy with enzalutamide, intended to suppress prostate cancer progression, increased the secretion of cytokines like IL‐1β and TGF‐β2 in LNCaP cells. The IL‐1β and TGF‐β2 further promoted the induction of HUVEC to the mesenchymal phenotype, EndoMTed HUVEC. EndoMTed HUVEC subsequently induced NED of LNCaP cells. Although new‐generation anti‐androgens, such as enzalutamide, may more effectively suppress AR signaling and reduce AR‐positive prostate cancer cells, they promote the enrichment of AR‐negative NE cells, potentially worsening treatment‐resistant forms such as NEPC [32]. This paradox may explain the increased incidence of aggressive NEPC variants following novel hormonal treatments [33].
Although the number of clinical cases evaluated in this study was small, we found that vascular endothelial cells and NE cells in human CRPC tissue were closely located. However, we could not differentiate between EndoMTed endothelial cells and other CAFs by immunofluorescence staining. Therefore, we alternatively examined the relative distances between vascular endothelial cells and CgA‐positive prostate cancer cells, representing neuroendocrine‐differentiated cells. Although our results do not directly translate into an interaction between neuroendocrine‐differentiated prostate cancer cells and EndoMTed endothelial cells, we propose that our in vitro phenomenon might also partly occur in human CRPC tissues.
We acknowledge that our study has several limitations. First, we employed an in vitro co‐culture and conditioned medium model using HUVEC to investigate the influence of EndoMT on prostate cancer cells. While this reductionist system allowed us to dissect specific stromal–epithelial interactions, it does not fully recapitulate the complexity of the in vivo TME, which includes immune cells, fibroblasts, and various soluble factors. For instance, other cytokines such as IL‐6 and IL‐8 are also known to promote NED and may act synergistically with GM‐CSF in vivo [34].
Second, although we observed spatial proximity between CD31‐positive endothelial cells and CgA‐positive tumor cells in human CRPC tissue samples, these data are correlative and do not provide causal evidence for the contribution of EndoMT to the development of t‐NEPC. To directly address this question, in vivo studies are warranted. A possible strategy would be to establish a subcutaneous xenograft model by implanting spheroid models constructed by co‐culturing LNCaP cells and HUVEC in a 3D culture into immunocompromised mice, as demonstrated in similar multicellular tumor spheroid models for non‐small cell lung cancer [35]. We are also considering using Bcl‐2 transduced HUVEC, which may support the viability of HUVEC and support the spheroid models [36]. Such models allow the reconstruction of reciprocal epithelial–endothelial interactions in a 3D architecture. By combining this co‐xenograft model with androgen deprivation (surgical castration), we can examine whether ADT itself promotes EndoMT in vivo, and whether this process alters the TME to favor NED. Therapeutic relevance could be evaluated by systemic administration of mavrilimumab to test whether inhibition of this pathway suppresses tumor growth, NED, and downstream STAT3–c‐MYC signaling. We anticipate that this in vivo strategy will demonstrate, within a physiologically relevant TME, whether the ADT‐induced EndoMT–GM‐CSF signaling axis causally contributes to t‐NEPC progression.
Third, although the GM‐CSF/CSF2RA pathway might participate in prostate cancer cell proliferation through the STAT3–c‐MYC pathway, we did not have any results of molecular pathways contributing to prostate cancer cell migration or invasion. There are several hypotheses as follows: (1) GM‐CSF has been reported to upregulate matrix metalloproteinases (MMPs) in tumor cells and associated stromal cells [37], which could facilitate extracellular matrix degradation and invasion. (2) By inducing EndoMT in endothelial cells and possibly partial EMT in tumor cells, GM‐CSF might increase the motility of both vascular and cancer cells, making the TME more conducive to invasion [26]. (3) GM‐CSF can mobilize and activate myeloid cells that assist in angiogenesis, and these aberrant vasculatures could provide pathways for tumor cells to escape (though GM‐CSF's role in angiogenesis is complex) [22].
In conclusion, this study identifies EndoMT in the TME as a key driver of NED in prostate cancer (Figure 7 ). We show that EndoMTed endothelial cells promote NED of LNCaP cells via secretion of GM‐CSF through CSF2RA, activating the STAT3–c‐MYC axis in cancer cells. These findings reveal a novel stromal–epithelial interaction that contributes to therapy resistance under androgen pathway suppression.
FIGURE 7.
LNCaP‐derived IL‐1β and TGF‐β2 induce HUVEC endothelial‐mesenchymal transition under androgen deprivation conditions, and the EndoMTed HUVEC, in turn, promotes LNCaP castration resistance.
Targeting the EndoMT–GM‐CSF signaling axis may represent a new therapeutic approach to prevent the emergence of treatment‐induced NEPC. Our results underscore the importance of tumor microenvironmental cues in shaping cancer cell plasticity and support more integrative strategies for managing advanced prostate cancer.
Author Contributions
Takumi Kageyama: data curation, formal analysis, investigation, visualization, writing – original draft. Manabu Kato: conceptualization, methodology, supervision. Shiori Miyachi: resources. Xin Bao: resources. Sho Sekito: data curation, software. Yusuke Sugino: formal analysis. Shinichiro Higashi: resources. Takeshi Sasaki: funding acquisition, supervision. Kouhei Nishikawa: supervision. Yasuhiro Murakawa: software, supervision. Masatoshi Watanabe: supervision. Takahiro Inoue: funding acquisition, project administration, writing – review and editing.
Disclosure
Registry and the Registration No. of the Study/Trial: Not applicable.
Animal Studies: All animal experiments were approved by the Animal Research Committee of Mie University (approval number: 2019–32).
Ethics Statement
Approval of the research protocol by an Institutional Review Board: This study was approved by the Institutional Review Board of the Mie University Hospital (approval number: 2425) and adhered to the highest ethical standards.
Consent
Written informed consent was obtained from all patients after full disclosure of the purpose and nature of all procedures.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Table S1. Antibodies.
Table S2. Patient background.
Doc S1. Cell sources and cell culture.
Doc S2. Induction of endothelial‐mesenchymal transition.
Doc S3. Co‐culture of prostate epithelial cell lines with EndoMTed HUVEC.
Doc S4. Transwell migration assay.
Doc S5. Immunofluorescence staining.
Doc S6. RNA extraction, cDNA preparation, and qPCR analysis.
Doc S7. Gene expression analysis using TaqMan custom array plates.
Doc S8. Western blot analysis.
Doc S9. RNA sequencing and transcriptome analysis.
Doc S10. GM‐CSF blockade using mavrilimumab.
Doc S11. siRNA‐mediated knockdown of CSF2RA.
Doc S12. Stimulation of LNCaP cells with recombinant GM‐CSF.
Doc S13. Colocalization analysis of chromogranin A and CD31 in NEPC.
Figure S1. Transient IL‐1β and TGF‐β2 treatment of HUVEC induces EndoMT.
Figure S2. Co‐culture with EndoMTed HUVEC of prostate cancer cells.
Figure S3. Correlation analysis of gene expression profiles in HUVEC and EndoMTed HUVEC samples and the validation of GM‐CSF expression and secretion by qPCR and ELISA.
Figure S4. GM‐CSF stimulation enhances the proliferation of LNCaP cells.
Acknowledgments
We would like to thank DNA Chip Research Inc. (Kanagawa, Japan) for their support with RNA‐Seq analysis and data acquisition. We also thank Editage (Mumbai, India) for professional English language editing and graphical abstract preparation.
Funding: This work was supported by Japan Society for the Promotion of Science, 24K02576, 24K12434.
Data Availability Statement
The RNA‐seq data from this study are available online from NCBI Sequence Read Archive (RRID: SCR_00489, accession number: PRJNA1181033) and Gene Expression Omnibus (RRID: SCR_005012, accession number: GSE281995). All the other raw data generated in this study are available upon request from the corresponding author.
References
- 1. Maman S. and Witz I. P., “A History of Exploring Cancer in Context,” Nature Reviews Cancer 18, no. 6 (2018): 359–376. [DOI] [PubMed] [Google Scholar]
- 2. Quail D. F. and Joyce J. A., “Microenvironmental Regulation of Tumor Progression and Metastasis,” Nature Medicine 19, no. 11 (2013): 1423–1437. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Kovacic J. C., Dimmeler S., Harvey R. P., et al., “Endothelial to Mesenchymal Transition in Cardiovascular Disease: JACC State‐Of‐The‐Art Review,” Journal of the American College of Cardiology 73, no. 2 (2019): 190–209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Zeisberg E. M., Potenta S., Xie L., Zeisberg M., and Kalluri R., “Discovery of Endothelial to Mesenchymal Transition as a Source for Carcinoma‐Associated Fibroblasts,” Cancer Research 67, no. 21 (2007): 10123–10128. [DOI] [PubMed] [Google Scholar]
- 5. Yoshimatsu Y., Wakabayashi I., Kimuro S., et al., “TNF‐α Enhances TGF‐β‐Induced Endothelial‐To‐Mesenchymal Transition via TGF‐β Signal Augmentation,” Cancer Science 111, no. 7 (2020): 2385–2399. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Yamada Y. and Beltran H., “Clinical and Biological Features of Neuroendocrine Prostate Cancer,” Current Oncology Reports 23, no. 2 (2021): 15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Belkahla S., Nahvi I., Biswas S., Nahvi I., and Ben Amor N., “Advances and Development of Prostate Cancer, Treatment, and Strategies: A Systemic Review,” Frontiers in Cell and Developmental Biology 10 (2022): 991330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Owa S., Sasaki T., Ikadai R., et al., “Psoas Mass Index at the Level of the Third Lumbar Vertebra on Computed Tomography Is a Prognostic Predictor for Metastatic Castration‐Sensitive Prostate Cancer,” International Journal of Clinical Oncology 29, no. 6 (2024): 840–846. [DOI] [PubMed] [Google Scholar]
- 9. Akamatsu S., Inoue T., Ogawa O., and Gleave M. E., “Clinical and Molecular Features of Treatment‐Related Neuroendocrine Prostate Cancer,” International Journal of Urology : Official Journal of the Japanese Urological Association 25, no. 4 (2018): 345–351. [DOI] [PubMed] [Google Scholar]
- 10. Lee A. R., Gan Y., Xie N., Ramnarine V. R., Lovnicki J. M., and Dong X., “Alternative RNA Splicing of the GIT1 Gene Is Associated With Neuroendocrine Prostate Cancer,” Cancer Science 110, no. 1 (2019): 245–255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Romero R., Chu T., González Robles T. J., et al., “The Neuroendocrine Transition in Prostate Cancer Is Dynamic and Dependent on ASCL1,” Nature Cancer 5, no. 11 (2024): 1641–1659. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Sun F., Zhang Z. W., Tan E. M., et al., “Icaritin Suppresses Development of Neuroendocrine Differentiation of Prostate Cancer Through Inhibition of IL‐6/STAT3 and Aurora Kinase A Pathways in TRAMP Mice,” Carcinogenesis 37, no. 7 (2016): 701–711. [DOI] [PubMed] [Google Scholar]
- 13. Araki S., Omori Y., Lyn D., et al., “Interleukin‐8 Is a Molecular Determinant of Androgen Independence and Progression in Prostate Cancer,” Cancer Research 67, no. 14 (2007): 6854–6862. [DOI] [PubMed] [Google Scholar]
- 14. Platel V., Faure S., Corre I., and Clere N., “Endothelial‐To‐Mesenchymal Transition (EndoMT): Roles in Tumorigenesis, Metastatic Extravasation and Therapy Resistance,” Journal of Oncology 2019 (2019): 8361913–8361945. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Johnson D. E., O'Keefe R. A., and Grandis J. R., “Targeting the IL‐6/JAK/STAT3 Signalling Axis in Cancer,” Nature Reviews Clinical Oncology 15, no. 4 (2018): 234–248. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Liguori T. T. A., Liguori G. R., Moreira L. F. P., and Harmsen M. C., “Adipose Tissue‐Derived Stromal Cells' Conditioned Medium Modulates Endothelial‐Mesenchymal Transition Induced by IL‐1β/TGF‐β2 but Does Not Restore Endothelial Function,” Cell Proliferation 52, no. 6 (2019): e12629. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Inoue T., Yoshida T., Shimizu Y., et al., “Requirement of Androgen‐Dependent Activation of Protein Kinase Czeta for Androgen‐Dependent Cell Proliferation in LNCaP Cells and Its Roles in Transition to Androgen‐Independent Cells,” Molecular Endocrinology (Baltimore, md) 20, no. 12 (2006): 3053–3069. [DOI] [PubMed] [Google Scholar]
- 18. Schindelin J., Arganda‐Carreras I., Frise E., et al., “Fiji: An Open‐Source Platform for Biological‐Image Analysis,” Nature Methods 9, no. 7 (2012): 676–682. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Dunn K. W., Kamocka M. M., and McDonald J. H., “A practical guide to evaluating colocalization in biological microscopy,” American Journal of Physiology‐Cell Physiology 300, no. 4 (2011): C723–C742. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Maleszewska M., Moonen J. R., Huijkman N., van de Sluis B., Krenning G., and Harmsen M. C., “IL‐1β and TGFβ2 Synergistically Induce Endothelial to Mesenchymal Transition in an NFκB‐Dependent Manner,” Immunobiology 218, no. 4 (2013): 443–454. [DOI] [PubMed] [Google Scholar]
- 21. Hamilton J. A., “GM‐CSF in Inflammation,” Journal of Experimental Medicine 217, no. 1 (2020): e20190945. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Leone P., Malerba E., Susca N., et al., “Endothelial Cells in Tumor Microenvironment: Insights and Perspectives,” Frontiers in Immunology 15 (2024): 1367875. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Piera‐Velazquez S. and Jimenez S. A., “Endothelial to Mesenchymal Transition: Role in Physiology and in the Pathogenesis of Human Diseases,” Physiological Reviews 99, no. 2 (2019): 1281–1324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Liu S., Alabi B. R., Yin Q., and Stoyanova T., “Molecular Mechanisms Underlying the Development of Neuroendocrine Prostate Cancer,” Seminars in Cancer Biology 86, no. Pt 3 (2022): 57–68. [DOI] [PubMed] [Google Scholar]
- 25. Kumar A., Taghi Khani A., Sanchez Ortiz A., and Swaminathan S., “GM‐CSF: A Double‐Edged Sword in Cancer Immunotherapy,” Frontiers in Immunology 13 (2022): 901277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ansari K. I., Bhan A., Saotome M., et al., “Autocrine GMCSF Signaling Contributes to Growth of HER2(+) Breast Leptomeningeal Carcinomatosis,” Cancer Research 81, no. 18 (2021): 4723–4735. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Qian C., Dong G., Yang C., et al., “Broadening Horizons: Molecular Mechanisms and Disease Implications of Endothelial‐To‐Mesenchymal Transition,” Cell Communication and Signaling : CCS 23, no. 1 (2025): 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Fontana R., Mestre‐Farrera A., and Yang J., “Update on Epithelial‐Mesenchymal Plasticity in Cancer Progression,” Annual Review of Pathology 19 (2024): 133–156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Cid M. C., Unizony S. H., Blockmans D., et al., “Efficacy and Safety of Mavrilimumab in Giant Cell Arteritis: A Phase 2, Randomised, Double‐Blind, Placebo‐Controlled Trial,” Annals of the Rheumatic Diseases 81, no. 5 (2022): 653–661. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Fleischmann R. M., van der Heijde D., Strand V., et al., “Anti‐GM‐CSF Otilimab Versus Tofacitinib or Placebo in Patients With Active Rheumatoid Arthritis and an Inadequate Response to Conventional or Biologic DMARDs: Two Phase 3 Randomised Trials (contRAst 1 and contRAst 2),” Annals of the Rheumatic Diseases 82, no. 12 (2023): 1516–1526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Temesgen Z., Burger C. D., Baker J., et al., “Lenzilumab in Hospitalised Patients With COVID‐19 Pneumonia (LIVE‐AIR): A Phase 3, Randomised, Placebo‐Controlled Trial,” Lancet Respiratory Medicine 10, no. 3 (2022): 237–246. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Butler W. and Huang J., “Neuroendocrine Cells of the Prostate: Histology, Biological Functions, and Molecular Mechanisms,” Precision Clinical Medicine 4, no. 1 (2021): 25–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Beltran H., Prandi D., Mosquera J. M., et al., “Divergent Clonal Evolution of Castration‐Resistant Neuroendocrine Prostate Cancer,” Nature Medicine 22, no. 3 (2016): 298–305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Culig Z., “Response to Androgens and Androgen Receptor Antagonists in the Presence of Cytokines in Prostate Cancer,” Cancers 13, no. 12 (2021): 2944. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Kim S. H., Song Y., and Seo H. R., “GSK‐3β Regulates the Endothelial‐To‐Mesenchymal Transition via Reciprocal Crosstalk Between NSCLC Cells and HUVECs in Multicellular Tumor Spheroid Models,” Journal of Experimental & Clinical Cancer Research 38, no. 1 (2019): 46, 10.1186/s13046-019-1050-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Enis D. R., Shepherd B. R., Wang Y., et al., “Induction, Differentiation, and Remodeling of Blood Vessels After Transplantation of Bcl‐2‐Transduced Endothelial Cells,” Proceedings of the National Academy of Sciences of the United States of America 102, no. 2 (2005): 425–430. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Hong I. S., “Stimulatory Versus Suppressive Effects of GM‐CSF on Tumor Progression in Multiple Cancer Types,” Experimental & Molecular Medicine 48, no. 7 (2016): e242. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Antibodies.
Table S2. Patient background.
Doc S1. Cell sources and cell culture.
Doc S2. Induction of endothelial‐mesenchymal transition.
Doc S3. Co‐culture of prostate epithelial cell lines with EndoMTed HUVEC.
Doc S4. Transwell migration assay.
Doc S5. Immunofluorescence staining.
Doc S6. RNA extraction, cDNA preparation, and qPCR analysis.
Doc S7. Gene expression analysis using TaqMan custom array plates.
Doc S8. Western blot analysis.
Doc S9. RNA sequencing and transcriptome analysis.
Doc S10. GM‐CSF blockade using mavrilimumab.
Doc S11. siRNA‐mediated knockdown of CSF2RA.
Doc S12. Stimulation of LNCaP cells with recombinant GM‐CSF.
Doc S13. Colocalization analysis of chromogranin A and CD31 in NEPC.
Figure S1. Transient IL‐1β and TGF‐β2 treatment of HUVEC induces EndoMT.
Figure S2. Co‐culture with EndoMTed HUVEC of prostate cancer cells.
Figure S3. Correlation analysis of gene expression profiles in HUVEC and EndoMTed HUVEC samples and the validation of GM‐CSF expression and secretion by qPCR and ELISA.
Figure S4. GM‐CSF stimulation enhances the proliferation of LNCaP cells.
Data Availability Statement
The RNA‐seq data from this study are available online from NCBI Sequence Read Archive (RRID: SCR_00489, accession number: PRJNA1181033) and Gene Expression Omnibus (RRID: SCR_005012, accession number: GSE281995). All the other raw data generated in this study are available upon request from the corresponding author.