NGF-mediated tumor–stroma crosstalk promotes prostate cancer aggressiveness

Abstract

Background

Nerve growth factor (NGF) is involved in prostate cancer (PC) pathogenesis and progression. Nevertheless, its source and its role in tumor microenvironment remain still unclear as well as its contribution to tumor–stroma interactions.

Methods

Single-cell transcriptomics analysis from multiple PC datasets was performed to identify the NGF-expressing cell populations. Primary cancer-associated fibroblasts (CAFs) isolated from PC patient’s specimens were analyzed for NGF secretion and expression of the NGF receptor, TrkA. 3D tumor-stroma co-cultures were used to investigate the NGF-mediated interactions between CAFs and epithelial PC cells. Migration, invasion, and perineural invasion (PNI) assays were done to evaluate the functional outcomes, alongside the pharmacological and antibody-based inhibition of NGF/TrkA axis.

Results

High NGF expression correlates with reduced disease-free survival in PC patients. Single-cell analyses reveal that stromal fibroblasts and myofibroblasts are the main sources of NGF within the PC microenvironment. Primary CAFs secrete NGF and express TrkA, supporting the existence of an autocrine signaling loop. In 3D co-cultures, NGF promotes CAF spatial organization around tumor spheroids, increases spheroid size, and induces EMT-like changes in epithelial PC cells. CAF-derived NGF further stimulates directional migration and invasion in both, CAFs and tumor cells. Blockade of NGF by neutralizing antibodies or pharmacological TrkA inhibition impairs such effects. In a PNI-like assay, CAF-derived NGF facilitates tumor cell migration across neuronal layers, underscoring its role in perineural invasion.

Conclusions

The present findings identify fibroblasts as the major source of NGF within the PC microenvironment, demonstrating the existence of a reciprocal NGF–TrkA signaling loop between CAFs and PC-derived cells. This axis promotes tumor growth, epithelial plasticity, and perineural invasion, identifying NGF as a key mediator of tumor–stroma crosstalk and a potential therapeutic target in prostate cancer.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07915-7.

Background

Prostate cancer (PC) still represents one of the most diagnosed malignancies in men. Despite the advancements in early detection, it remains one of the major causes of cancer-related death in male population. Active surveillance is often recommended in low- or intermediate-risk patients [1], while management of localized disease might be achieved through urologic surgery, radiotherapy or brachytherapy, alone or in combination with the androgen depletion therapy (ADT). However, the disease often progresses and acquires resistance to ADT. At this stage, PC becomes castration-resistant (CRPC), which can be metastatic or not. Patients then experience a poor clinical outcome as well as the adverse effects induced by therapeutic interventions, such as lifelong ADT with abiraterone or androgen receptor (AR) pathway inhibitors [2, 3], poly (adenosine diphosphate ribose) polymerase inhibitors, chemotherapy, radioligand-based approaches or even their combination [4, 5]. To date, PC progression and therapy-resistance still represent a huge challenge.

Beyond the intrinsic genetic and epigenetic alterations, tumor microenvironment (TME) critically contributes to PC evolution [6, 7]. Among stromal components, CAFs play a central role by modulating tumor growth, metabolism, immune evasion, ECM remodeling, metastatic dissemination, and lineage plasticity [8, 9]. By this way, CAFs communicate with PC epithelial cells. They also contribute to epithelial–mesenchyme transition (EMT) and perineural invasion (PNI), which are strictly associated with PC aggressiveness and poor prognosis [7, 10, 11].

Nerve growth factor (NGF) was initially described for its essential role in neuronal survival and differentiation [12, 13]. It is now largely recognized as a pleiotropic regulator acting far beyond the nervous system. NGF influences the behavior of immune cells, stromal populations, and cancer cells across tumor types [14–16]. Beyond its role in central nervous system, NGF also acts in a variety of non-neuronal tissues [14], cancer cells [15] and the surrounding microenvironment [16]. These effects are mediated by the two cognate NGF receptors, the neurotrophin receptor p75NTR (also called NGF receptor; NGFR) and the high-affinity neurotrophin tyrosine kinase receptor (Trk) A [17].

NGF has been implicated in promoting proliferation, migration, neuroendocrine differentiation and ADT resistance of PC, pointing to its role in shaping tumor aggressiveness [18, 19]. Nevertheless, the cellular source of NGF within the PC-associated microenvironment, its contribution to stromal–epithelial crosstalk, ECM remodeling and neural engagement remain incompletely understood.

In this study, we integrate single-cell transcriptomics with functional analyses in advanced 3D co-culture systems to dissect the origin and role of NGF within the prostate tumor microenvironment.

We show that CAFs represent the major reservoir of NGF in tumor stroma, regardless of Gleason’s grade. CAFs-derived NGF acts in autocrine and paracrine fashion to sustain PC-derived spheroids increase, promote EMT, enhance CAFs recruitment and potentiate PNI. TrkA activation mediates these effects, thus revealing a previously unappreciated reciprocal signaling loop between stromal and epithelial compartments.

Together with clinical analyses showing that high NGF expression correlates with poor patient prognosis, our findings identify the NGF/TrkA circuit as a central driver of tumor–stroma–nerve crosstalk in PC. Targeting this axis may offer new opportunities for therapeutic intervention, particularly in aggressive and treatment-resistant disease.

Materials and methods

Cell cultures

PC-derived LNCaP and PC3 cell lines (from ATCC) were cultured and growth arrested as reported [20]. Stable GFP-expressing LNCaP cells were generated using the pEGFP-C1 vector (Addgene, Watertown, MA, USA), according to the same report. PC12 cells were cultured in Corning plates, as described [21]. Cells were made quiescent using phenol red–free DMEM (Life Technologies), containing 0.5% charcoal-treated FCS, antibiotics, L-glutamine (at 2 mM; Life Technologies) and then used. Primary prostate CAFs were isolated from biopsy specimens collected from PC patients. Written informed consent was obtained from all participants. The study was approved by the institutional ethics committee (PC-StAR Project, University “L. Vanvitelli”, protocol B.684; 18/12/2017). Tumor grade was determined by histopathological examination of stained frozen sections and classified using the Gleason’s scoring system. Briefly, fresh tissue fragments adjacent to the tumor area were minced and subjected to enzymatic digestion to separate stromal and epithelial components. CAFs derived from 8 patients (Table 1) were isolated following digestion and seeded into six-well plates. Cells were cultured in DMEM supplemented with 10% FBS, penicillin (10 U/ml), streptomycin (100 µg/ml), glutamine (2 mM), insulin (5 µg/ml), transferrin (5 µg/ml), and amphotericin B (2 µg/ml). Cultures were maintained at 37 °C in a humidified 5% CO₂ atmosphere. Once confluent, CAFs were expanded and used for experiments within 4–5 passages. To verify culture purity, CAFs were screened using anti-vimentin, anti-αSMA and anti-FAP1 antibodies. In addition, anti-E cadherin antibodies were used to exclude epithelial contamination. A summary of CAF characteristics derived from PC patients with different Gleason scores is provided in Table 1.

Table 1.

CAFs from PC patients

Patient	Gleason’s score	TrkA	E-cadherin	vimentin	α-SMA	FAP
c1	6	+	−	+	+	+
c2	8	+	−	+	+	+
c3	7	+	−	+	+	+
c4	6	+	−	+	+	+
c5	6	+	−	+	+	+
c6	6	+	−	+	+	+
c7	7	+	−	+	+	+
c8	7	+	−	+	+	+

Chemicals

Recombinant human nerve growth factor (NGF; from Millipore, Burlington, MA, United States) was used at 100 ng/ml. The selective TrkA inhibitor, GW441756 from Selleckchem (Munich, Germany) was dissolved in dimethyl sulfoxide (DMSO) according to the manufacturer’s instructions. It was used at 1µM, with equivalent volumes of DMSO included as vehicle controls. NGF neutralization experiments were done using a specific anti-NGF blocking antibody (1,600 pg/ml; ab6199; Abcam, Cambridge, United Kingdom).

Data mining

Publicly available single-cell RNA-seq datasets from human prostate cancer and related TME were obtained from the Gene Expression Omnibus (GEO) under the accession numbers GSE137829, GSE141445, and GSE172301. Processed data, including UMAP embeddings, cell-type annotations, and normalized expression matrices, were obtained from the Tumor Immune Single-Cell Hub (TISCH v2) (https://tisch.comp-genomics.org/), which provides uniformly processed scRNA-seq datasets with standardized clustering and lineage assignment. Raw gene–cell count matrices from the GEO repositories were imported into R (v4.3) and analyzed using the Seurat package (v4.3.0). NGF expression was assessed across cell populations using Seurat’s FeaturePlot and VlnPlot functions.

For each dataset, NGF expression values were extracted, log-normalized, and compared across annotated clusters. Only clusters consistently observed across at least two datasets were considered biologically robust. Survival analysis related to NGF was derived from Gene Expression Profiling Interactive Analysis (GEPIA 2; http://gepia2.cancer-pku.cn/#index).

In addition, stromal enrichment was evaluated in TCGA prostate adenocarcinoma (PRAD) samples, using the ESTIMATE algorithm (Estimation of STromal and Immune cells in MAlignant Tumors using Expression data). Stromal scores, which quantify the presence of stromal cell signatures within bulk tumor transcriptomes, were downloaded from the UCSC Xena Pan-Cancer Atlas Hub (https://pancanatlas.xenahubs.net), selecting the dataset “PRAD RNAseqV2 ESTIMATE scores”.

Clinical annotations, including survival endpoints, were obtained from the curated Pan-Cancer Atlas clinical dataset (“Survival_SupplementalTable_S1_20171025_xena_sp”). Samples were filtered to include only prostate adenocarcinoma (PRAD) cases. Patient identifiers were harmonized between ESTIMATE and clinical datasets using the first 12 characters of the TCGA barcode to ensure correct matching. Patients were then stratified into stromal-high and stromal-low groups using the median stromal score as cutoff. Progression-free interval (PFI) was selected as the primary survival endpoint. Kaplan–Meier survival curves were generated using the “survival” R package, and differences between groups were evaluated using the log-rank test. A two-sided p-value < 0.05 was considered statistically significant.

Enzyme-linked immunosorbent assay (ELISA)

It was performed to quantify β-NGF, using ELISA kit (EHNGF; Thermo Fisher Scientific). CAFs (10 × 10⁴ cells per well) isolated from different patients were plated in 12-well plates and made quiescent. After 72 h, conditioned media were collected and used for β-NGF measurement according to the manufacturer’s protocol. NGF secretion was also quantified in conditioned media (CM) collected under the same experimental conditions used for PNI assays. PC3 and PC12 cells were cultured for 24 h, whereas CAFs were maintained for 5 days to allow continuous accumulation of secreted factors. NGF concentrations were measured in the collected culture volumes and subsequently normalized to 1 mL to enable direct comparison among different cell types. β-NGF levels were assayed by ELISA as before reported and resulting data were analyzed with GraphPad Prism software using curve-fitting methods.

Protein extraction and western blot analysis

Protein extracts were prepared by lysing cells in a cold buffer containing 50 mM Tris-HCl (pH 7.4), 1 mM EDTA, 1% Triton X-100, 150 mM NaCl, 5 mM MgCl₂, and 1 mM EGTA, supplemented with protease and phosphatase inhibitors, including 1 mM PMSF (phenylmethylsulfonyl fluoride), a protease inhibitor cocktail (LAP), 1 mM sodium orthovanadate (Na₃VO₄), and 100 µg/mL aprotinin. Protein concentrations were determined using the Bio-Rad protein assay kit (Bio-Rad Laboratories, Hercules, CA, USA). Aliquots of 20–30 µg proteins were resolved on 10–15% SDS-PAGE gels and transferred to nitrocellulose membranes (Amersham Protran Premium 0.45 μm, GE Healthcare, Chicago, IL, USA). Membranes were blocked in 5% non-fat milk and incubated overnight at 4 °C with primary antibodies. After washing, membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies. Protein signals were visualized using enhanced chemiluminescence (ECL) reagents (GE Healthcare, Chicago, IL, USA). The following primary antibodies were used: mouse monoclonal antibodies against α-tubulin (#E-AB-20036, Elabsciences, Houston, TX, USA), GAPDH (#E-AB-20079, Elabsciences), α-SMA (CGA7, Santa Cruz Biotechnology, Dallas, TX, USA), FAP (SS-13, Santa Cruz Biotechnology), and E-cadherin (Clone 36, RUO); rabbit polyclonal antibodies against vimentin (H-84, Santa Cruz Biotechnology, Dallas, TX, USA) and TrkA (06-574, Millipore, Burlington, MA, USA). Protein expression levels were quantified by densitometric analysis using ImageJ software (NIH). Digital images of immunoblots were acquired within the linear range of detection to avoid signal saturation. Band intensities were measured by defining rectangular ROIs corresponding to each band. Protein expression was normalized to the corresponding loading control (tubulin). Relative expression levels were calculated as the ratio of target protein intensity to loading control intensity and expressed as fold change relative to control conditions.

Densitometric values represent the mean ± standard deviation (SD) of at least three independent biological experiments.

Full, uncropped Western blot images are provided in the Supplemental Materials.

IF analysis and 3D models in ECM

CAFs cultured on coverslips were made quiescent and 72 h later they were washed with phosphate-buffered saline (PBS). Cells were fixed for 10 min with 4% paraformaldehyde (w/v in PBS; Merck, Saint Louis, MO, USA) and permeabilized for 10 min with 0.05% Tween-20 (v/v in PBS; Bio-Rad, Hercules, CA, USA). Non-specific binding sites were blocked by incubating the cells for 1 h in PBS containing 1% foetal bovine serum (FBS, vol/vol). Coverslips were then incubated overnight at 4 °C with anti-NGF antibody (1:85, ab6199; Abcam, Cambridge, UK). Following extensive PBS washes, cells were incubated for 1 h at 37 °C with a fluorescein-conjugated AffiniPure anti-rabbit IgG secondary antibody (1:200 in PBS with 0.01% BSA; Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Nuclei were counterstained with Hoechst 33,258 (1 µg/mL; Merck) for 5 min. Fluorescence images were acquired using a Leica DMIRB inverted microscope equipped with an HCX PL Fluotar ×63 objective and captured with a Leica DFC 450 C camera. Fluorescence intensity was quantified using ImageJ software (NIH, USA). Acquired images were processed under identical exposure settings for all samples. Images were first split into individual channels, and background subtraction was performed on the NGF (green) channel.

For each field of view, mean fluorescence intensity (MFI) was measured using the “Measure” function after selecting the entire image area. The following parameters were recorded: mean gray value and area. Identical thresholding and analysis parameters were applied across all samples to ensure consistency. To account for variations in cell density, fluorescence intensity was normalized to the number of cells per field. DAPI-positive nuclei were quantified by thresholding the DAPI channel followed by particle analysis. Normalized fluorescence intensity was calculated as:

For each CAF isolate, NGF expression was expressed as fold increase relative to the corresponding secondary antibody (II Ab) control:

At least three independent fields per CAF isolate were analyzed under identical acquisition conditions.

Three-dimensional (3D) cultures were established as described [20]. In monocultures, 1 × 10⁴ LNCaP or PC3 cells were resuspended in 66 µL of phenol red-free, growth factor-reduced Matrigel (10 mg/mL; BD Biosciences, San Jose, CA, USA) and 133 µL of organoid plating medium per well. In co-culture experiments, 1 × 10⁴ GFP-expressing LNCaP cells were combined with 2.5 × 10⁴ CAFs in 100 µL of ECM and 100 µL of organoid plating medium per well. In all conditions, the ECM consisted of a 1:4 mixture of rat tail type I collagen (4.6 mg/mL; BD Biosciences, San Jose, CA, USA) and phenol red-free, growth factor-reduced Matrigel (10 mg/mL; BD Biosciences). The cell–ECM mixture was plated in 24-well plates and allowed to polymerize at 37 °C for 30 min before adding 250 µL of organoid plating medium. Organoid plating medium was prepared as reported [22], using phenol red-free DMEM/F12 supplemented with 7% charcoal-stripped serum (CSS), 100 U/mL penicillin, 100 µg/mL streptomycin, 1× GlutaMAX (Gibco), 10 mM HEPES, B27 supplement (50× stock; Thermo Fisher Scientific), 1 M nicotinamide (Sigma-Aldrich), 500 mM N-acetylcysteine (Sigma-Aldrich), and 10 µM Y-27,632 (Millipore). After 3 days, the medium was replaced with the same formulation lacking N-acetylcysteine and Y-27,632. On day 4, spheroids were either left untreated or exposed to the indicated compounds. The medium was changed every 3 days. Spheroid growth was monitored using a Leica DMIRB inverted microscope equipped with C-Plan ×40 or HCX PL Fluotar ×63 objectives. Phase-contrast and IF images were captured with a Leica DFC 450 C camera. Merged images were generated using Leica Application Suite software. Spheroid size measurements were performed using the same software. Spheroids were fixed in 4% paraformaldehyde in PBS for 30 min at room temperature on a shaker (20 rpm). To reduce matrix autofluorescence, they were incubated twice for 10 min each in PBS containing 0.1 M glycine, which quenches residual aldehyde groups generated during fixation. Samples were then permeabilized with 0.5% Triton X-100 in PBS for 45 min, followed by blocking in PBS supplemented with 1% FBS and 0.01% sodium azide for 3 h at room temperature. Spheroids were incubated overnight at 4 °C with the anti-vimentin primary antibody (1:50, H-84; Santa Cruz Biotechnology, Dallas, TX, USA) in IF buffer (PBS containing 1% BSA, 0.1% Tween-20, 0.01% sodium azide). After extensive PBS washes, samples were incubated for 1 h at 37 °C with Texas Red-conjugated secondary antibody (1:200 in PBS with 0.01% BSA; Jackson Immuno-Research, West Grove, PA, USA). Fluorescence images were acquired using a Leica DMIRB inverted microscope equipped with ×40 or ×63 objectives and captured with a Leica DFC 450 C camera.

Migration, invasion and perineural invasion assays

Migration assay was done using quiescent CAFs (3 × 10⁴ cells per well in 150 µL of culture medium) seeded in collagen-coated Boyden chambers with 8 μm polycarbonate membranes (Corning, Corning, NY, USA). The indicated stimuli were added to the upper chamber, and cytosine arabinoside (50 µM; Sigma-Aldrich, St. Louis, MO, USA) was included to inhibit proliferation. Cells were allowed to migrate for 6 h as reported [20]. Invasion assay was done using quiescent CAFs (5 × 10⁴ cells per well in 150 µL of culture medium) in Boyden chambers with 8 μm polycarbonate membranes pre-coated with growth factor-reduced, phenol red-free Matrigel (0.6 mg/mL; Corning). Stimuli and cytosine arabinoside (50 µM) were added, and cells were allowed to invade for 18 h as reported [20]. Perineural invasion was assessed using a modified Trans-well co-culture system mimicking tumor–neuron interactions. Briefly, Boyden chambers equipped with 8-µm pore size polycarbonate membranes (Corning, Corning, NY, USA) were coated on the upper surface with growth factor–reduced, phenol red–free Matrigel (0.5 mg/ml Corning) to generate a permissive extracellular matrix layer. Neuronal-like PC12 cells (2 × 10⁴) were seeded onto the Matrigel-coated upper surface of the membrane and allowed to adhere and form a continuous monolayer, serving as a neuronal barrier. GFP-expressing PC3 cells (3.5 × 10⁴) were then plated on top of the PC12 monolayer. CAFs (5 × 10 ⁴) were seeded in the lower compartment of the Trans-well system five days before the co-culture to allow continuous accumulation of CAF-derived conditioned medium. Stimuli, inhibitors and antibodies were added to both the compartments as indicated in the corresponding Figure. GFP- PC3 cells were allowed to migrate for 24 h at 37 °C in a humidified incubator with 5% CO₂. At the end of the assay, non-migrated cells were removed from the upper surface of the membrane, while GFP-PC3 cells that had migrated across the PC12 neuronal layer and the membrane were fixed and analyzed. Migrated cells were visualized by fluorescence microscopy and quantified by counting GFP-positive cells in multiple randomly selected fields per membrane. Data were expressed as relative invasion compared to control conditions.

Statistical analysis

Unless otherwise stated, experiments were performed in at least three independent replicates. Results are shown as mean ± standard deviation (SD). Statistical significance was assessed using either one-way or two-way ANOVA with Bonferroni’s multiple comparison test, or unpaired two-tailed Student’s t-tests, as appropriate for the dataset. Analyses were conducted with GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA), and a p-value of less than 0.05 was considered significant.

Results

High NGF expression in prostate cancer correlates with poor prognosis and is predominantly of stromal origin

To investigate the clinical relevance of NGF in prostate cancer (PC), we first interrogated the GEPIA2 database. Patients (n = 492) were stratified into high (n = 246) and low (n = 246) NGF expression groups. Kaplan–Meier analysis revealed that elevated NGF levels were significantly associated with worse clinical outcome, as reflected by reduced disease-free survival (p = 0.0012; Fig. 1A).

Fig. 1.

NGF expression is prognostically unfavorable and predominantly localized in tumor-infiltrating fibroblasts. (A) Kaplan–Meier analysis of disease-free survival (DFS) stratified according to NGF expression (high, red line vs. low, blu line). Patients with elevated NGF levels show significantly shorter DFS, indicating an unfavorable prognostic impact of NGF in prostate cancer. Number of patients (n) 246 for each sub-group. p = 0.0012. (B–D) UMAP visualizations of single-cell RNA-seq data from 3 independent prostate cancer datasets (B: GSE137829; C: GSE141445; D: GSE172301). Each dot represents a single cell, colored by cluster identity. The color scale indicates the NGF expression levels (normalized TPM). Across all datasets, NGF expression is consistently enriched in fibroblast and myofibroblast clusters, while epithelial, immune, endothelial and smooth muscle populations display low or undetectable expression. The term “Plasma” refers to plasma cells, as annotated in the original dataset metadata. Abbreviations: B, B cells; CD8T, CD8⁺ T cells; CD8Tex, exhausted CD8⁺ T cells; Endothelial, endothelial cells; Epithelial, epithelial cells; Fibroblasts, stromal fibroblasts; Malignant, tumor epithelial cells; Mast, mast cells; Mono/Macro, monocytes/macrophages; Plasma, plasma cells; SMC, smooth muscle cells.These findings identify tumor-infiltrating fibroblasts as the major source of NGF within the prostate tumor microenvironment

We next sought to define the cellular source of NGF within the PC tumor microenvironment. Single-cell transcriptomic profiling was done by considering three independent scRNA-seq datasets (GSE137829, GSE141445, and GSE172301). NGF transcripts were consistently enriched in stromal fibroblast and myofibroblast populations, whereas epithelial and immune compartments showed minimal or negligible expression in all datasets (Fig. 1B–D). These findings indicate that CAFs represent the predominant source of NGF within the PC microenvironment.

To further explore the clinical impact of stromal enrichment per se, Kaplan–Meier analysis of TCGA-PRAD patients stratified according to stromal score (median split, ESTIMATE algorithm) was done. No statistically significant difference in progression-free interval (PFI) was observed between stromal-high and stromal-low tumors (log-rank p = 0.20; Fig. 1S). A total of 20 progression events were recorded (8 in stromal-high and 12 in stromal-low cases), suggesting limited statistical power.

These findings indicate that NGF expression is associated with adverse clinical outcome in PC and that fibroblasts represent its principal cellular source. However, overall stromal abundance alone does not appear sufficient to predict prognosis, supporting the concept that the functional activation state of NGF-expressing stromal subsets, rather than stromal quantity per se, might drive disease aggressiveness.

CAFs derived from prostate tumors secrete NGF and express the high affinity receptor TrkA

To corroborate the evidence collected by single-cell transcriptomic analyses, we next quantified by ELISA assay the NGF released by a panel of primary CAFs isolated from PC patient’s specimens spanning a range of Gleason’s scores (6–8; see Table 1). Despite the heterogeneity in tumor grade, nearly all CAF cultures released appreciable and biologically relevant amounts of NGF into the conditioned medium (CM; Fig. 2A), indicating that NGF production is a conserved feature of prostate CAFs.

Fig. 2.

PC-derived CAFs release NGF and express mesenchymal/myofibroblast markers as well as TrkA. (A) Quantification of NGF secretion by ELISA in conditioned media collected from primary CAFs lines (C1–C8) isolated from independent PC patients at Gleason’s scores 6–8. Data are expressed as pg/ml (mean ± SD) from three independent experiments performed in duplicate. (B) Representative IF images of NGF expression in primary CAFs (C1). NGF is shown in green and nuclei in blue. Images are representative of three independent experiments. Scale bar, 10 µM. (C) Quantitative analysis of NGF fluorescence intensity in primary CAF cultures (C1–C8). Mean fluorescence intensity (MFI) was normalized to the number of nuclei and expressed as fold increase relative to the fluorescence from the secondary antibody, used as control (II Ab). Data are presented as mean ± SD from several independent fields analyzed for each culture. (D) Western blot analysis of TrkA and mesenchymal/myofibroblast markers in lysates from primary prostate CAFs (C1–C8). Vimentin was used as a mesenchymal marker, while α-SMA and FAP-1 were used as myofibroblast-associated markers. Tubulin served as loading control. LNCaP cells, positive for the expression of E-cadherin, were included as epithelial control

To further validate these findings at the single-cell level, we done immunofluorescence (IF) analysis. Representative images in Fig. 2B show a robust NGF staining within CAF cultures, as compared with images from secondary antibody alone, used as control (II Ab). Consistent with the ELISA results, quantitative analysis of mean fluorescence intensity (MFI), normalized to cell number, confirmed a significant expression of NGF across independent CAF lines (Fig. 2C).

Protein expression profiling of CAFs showed that all primary CAF cultures (Fig. 2D) exhibited significant levels of vimentin, α-SMA, and FAP-1, while lacking E-cadherin, thus excluding any epithelial contamination and validating their stromal identity. Notably, all CAF lysates displayed detectable amounts of TrkA. These latter findings support the existence of an autocrine/paracrine feedback mechanism, whereby CAFs-derived NGF may directly act on the stromal compartment, contributing to CAFs activation and functional responsiveness.

CAF-derived CM induces EMT-like changes of PC spheroids in a 3D-dependent manner

We next analyzed the impact of CAFs-derived CM on EMT features in PC cell-derived spheroids. Prolonged exposure to CM markedly increased the size of LNCaP spheroids after 15 days in culture, as shown by representative images and quantitative analysis (Fig. 3A–B). Similar effects were seen in PC3-derived spheroids (Fig. 3F-G).

Fig. 3.

Soluble factors derived from CAFs promote spheroid growth and induce EMT-like changes in PC-derived spheroids cultured in 3D. (A) Representative phase-contrast and immunofluorescence (IF) images showing vimentin expression (red) in LNCaP-derived spheroids cultured for 15 days in control medium or CAF-derived conditioned medium (CAF-CM). (B) Quantification of LNCaP spheroid size expressed as fold increase relative to control conditions after 15 days of treatment with ctrl medium or CAF-derived CM (n = 3). Bar, 100 µM. Data are presented as mean ± SD. **p < 0.01. (C) Western blot analysis of EMT markers (E-cadherin, vimentin) in LNCaP cells cultured in 2D or 3D conditions in the absence or presence of CAF-CM and anti-NGF neutralizing Ab (anti-NGF Ab). (D) Densitometric quantification of vimentin/tubulin and (E) E-cadherin/tubulin ratios from independent experiments shown in (C), including conditions with anti-NGF neutralizing antibody (anti-NGF Ab) where indicated. (F) Representative phase-contrast and immunofluorescence images showing vimentin expression (red) in PC3-derived spheroids cultured for 15 days in control medium or CAF-CM. (G) Quantification of PC3 spheroid size, expressed as fold increase relative to control conditions after 15 days of treatment (n = 3). Data are presented as mean ± SD. **p < 0.01. Bar 100 µM. (H) Western blot analysis of EMT markers (E-cadherin and vimentin) in PC3 cells cultured under 2D or 3D conditions. (I-L) Densitometric analysis of vimentin/tubulin and E-cadherin/tubulin ratios from experiments shown in (H)

IF analysis revealed that CM treatment clearly induced de novo appearance of vimentin in LNCaP spheroids (Fig. 3A), marking a shift toward a mesenchymal-like state. The Western blot analysis consistently confirmed the appearance of vimentin in 3D cultures exposed to CM, which was completely undetectable in untreated spheroids. Simultaneously, a marked reduction in E-cadherin levels was observed in 3D cultures (Fig. 3C). These EMT-associated changes were not observed in 2D cultures exposed to CM for 3 days, as vimentin remained undetectable and E-cadherin levels were largely unchanged (Fig. 3C–E). These findings underscore the requirement for a 3D architecture and prolonged exposure to microenvironment-derived factors for the acquisition of EMT-like features.

Notably, NGF neutralization fully abrogated CM-induced vimentin upregulation and restored E-cadherin expression in LNCaP spheroids (Fig. 3C–E), indicating that NGF acts as a key mediator of the CM-driven EMT program. Densitometric quantification of the Western blot data confirmed the statistical significance of these changes (Fig. 3D–E).

In PC3 cells, which already exhibit a mesenchymal-like phenotype, CM exposure further increased vimentin expression in both 2D and 3D setting. However, CM induced a complete loss of E-cadherin only in 3D cultures (Fig. 3H). Consistent with our previous findings [23], PC3 cells in 2D already displayed a partial reduction in E-cadherin levels. Densitometric quantification of the Western blot data confirmed the statistical significance of these changes (Fig. 3I-L).

This set of experiments indicates that factors released by CAFs, with NGF playing a central role, promote an EMT-like phenotype in PC cells. These responses appear particularly evident in a 3D- and context-dependent manner, with markedly stronger effects arising over extended culture periods.

NGF drives CAF reorganization and enhances prostate tumor spheroid growth in 3D

To investigate the role of NGF in shaping tumor–stroma interactions, we next established 3D co-culture models using GFP-LNCaP-derived spheroids [20], in combination with a pool of primary CAFs. Challenge with NGF promoted the reorganization of CAFs into dense, bundle-like structures that tightly enveloped the spheroids (Fig. 4A). Simultaneously, spheroid size was significantly increased, as compared with untreated controls (Fig. 4A and B).

Fig. 4.

NGF promotes CAF organization around PC spheroids and enhances spheroid’s growth in 3D co-cultures. (A) Representative images of 3D co-cultures from LNCaP-derived spheroids (green) with CAFs, untreated or treated with exogenous NGF (100 ng/mL), CAF-derived conditioned medium (CM), or CM supplemented with the anti NGF-neutralizing antibody. Bar, 100 µM. (B) Quantification of spheroid size (represented as fold increase over the basal level at 3rd day, black bar) across treatments. In B and F, n represents the number of experiments. Means ± SD are shown. **p < 0.01

Albeit at a lesser extent, treatment with CAF-derived CM, containing approximately 2 ng/mL NGF, produced similar effects (Fig. 4A–B). Notably, neutralization of NGF by a specific blocking antibody abolished both the CAF reorganization and spheroid enlargement, confirming that NGF is a central mediator of these processes. These data demonstrate that NGF not only orchestrates stromal cell architecture, but also promotes tumor spheroid growth in a 3D microenvironment.

NGF–TrkA axis promotes CAF migration and directional recruitment by tumor-derived factors

We then analyzed the NGF-induced invasion of different CAFs, as indicated in Fig. 5. NGF stimulation significantly increased the invasion of CAFs derived from three different patients. TrkA inhibitor, GW441756 abolished this effect (panels in A), indicating a role for TrkA activation in this response. Expectedly, an increase in CAFs invasiveness was detected upon treatment with CM derived from LNCaP cells, as these cells release NGF [20]. Addition of the anti-NGF neutralizing antibody blocked such effect, confirming that NGF promotes the recruitment of CAFs towards PC cells (panels in B).

Fig. 5.

NGF–TrkA signaling drives reciprocal migration of CAFs and prostate cancer cells. (A) Transwell invasion assays performed with 3 independent primary CAF lines (C2, C4, and C7), each derived from distinct PC patients. Treatment with recombinant NGF (100 ng/mL) significantly increased CAF invasiveness compared to untreated controls (Unt.). GW441756 (GW) abolished NGF-induced invasion. Data are expressed as fold increase over basal level (Untr.). (B) Transwell invasion assays of CAF lines (C2, C4, and C7) exposed to conditioned medium (CM) derived from LNCaP cells. LNCaP-CM significantly enhanced CAF invasiveness compared to control medium. The pro-invasive effect of LNCaP-CM was blocked by a neutralizing anti-NGF antibody (anti-NGF Ab). (C) Perineural invasion (PNI) assay assessing GFP-PC3 cells migration across a differentiated PC12 monolayer in the presence of a pool of CAFs derived from C2, C4 and C7 lines plated in the lower chamber. CAFs significantly enhanced tumor cell migration, which was reduced by anti-NGF Ab or TrkA inhibition (GW). Data are expressed as fold increase over basal level. In panels (A–C), schematic illustrations summarize the corresponding Transwell experimental setups. Data represent mean ± SD of three independent experiments (n = 3). **p < 0.01; ***p < 0.001

NGF promotes perineural invasion through stromal and epithelial crosstalk

At last, we investigated whether NGF contributes to perineural invasion (PNI), a hallmark of aggressive PC [18, 19]. GFP-PC3 cells were, hence, seeded on a monolayer of neuronal-like PC12 cells coated with Matrigel [24]. Albeit at different extent, exogenous NGF and CAFs-derived CM (with CAFs plated in the bottom chamber) both enhanced the invasion of PC3 cells across the neuronal layer. In both cases, GW441756 abrogated the invasive response (Fig. 5C). Similarly, NGF neutralization inhibited PC3 invasion. While PC12 and PC3 cells may release NGF during the timeframe (24 h) used in our setting, CAFs had been cultured for 5 days prior to the co-culture experiments (see also Table 2 indicating the levels of NGF released by CAFs), thus providing a continuous source of NGF. Thus, CAFs likely represent the predominant source of NGF-driving perineural invasion in our experimental setting.

Table 2.

NGF secretion by PC3 cells, PC12 cells and CAFs under PNI-related experimental conditions

Cell type	Cell number	Culture duration	Collection volume	NGF measured in collected volume (pg)	NGF normalized to 1 mL (pg/mL)
PC3	3.0 × 10⁴	24 h	125 µL	~ 25	~ 200
PC12	2.0 × 10⁴	24 h	200 µL	~ 10	~ 50
CAFs	5.0 × 10⁴	5 days	500 µL	~ 460	~ 920

In summary, our findings reveal a reciprocal NGF-mediated loop in PC microenvironment. Epithelial PC cells release NGF to recruit CAFs, which in turn secrete NGF to enhance tumor growth, EMT, and perineural invasion. This bidirectional signaling establishes a pro-invasive, pro-proliferative niche and identifies NGF as a key mediator of tumor–stroma crosstalk in PC.

Discussion

NGF-mediated signaling is involved in survival, proliferation, invasion and drug-resistance in multiple cancer types [25]. Association of NGF signalling with tumor progression and metastasis has been studied in melanomas [26], pancreatic cancer [27] and neuroblastomas [28, 29]. Derangement of NGF signaling has been reported in breast and PC, which are classically considered as ‘hormone-dependent cancers’ [23, 30–46].

PC development and progression occur through derangement of genetic and epigenetic programs, in concert with alterations in the surrounding TME [9, 11, 47, 48]. Beyond cancer cell–intrinsic alterations, soluble factors released within the TME shape tumor behavior, influence proliferation, plasticity, invasion and therapeutic resistance. Among these factors, NGF triggers the activation of various signaling cascades, promoting the acquisition of aggressive phenotype (reviewed by [18]).

Emerging findings suggest that NGF signalling within the TME not only affects epithelial PC cells but also modulates immune infiltration and stromal remodelling, promoting polarization of tumor-associated macrophages toward pro-tumorigenic states and enhancing CAF-driven ECM remodelling and stiffening, thereby facilitating tumor cell migration and invasion [49–52]. Moreover, NGF-induced TrkA activation has been shown to activate YAP/TAZ signalling in CAFs, connecting neurotrophins signalling to mechano-transduction and biomechanical regulation of TME [53–56]. Collectively, these findings poise the NGF signalling as a relevant target in advanced PC and neuroendocrine disease variants.

By engaging its two cognate receptors, NGFR or TrkA, NGF might play a dual function in PC. While TrkA activation promotes PC aggressiveness [23, 45], NGFR signalling has been associated with growth-inhibitory effects [32]. NGF signalling has been also involved in ADT-mediated neuroendocrine differentiation of PC (NEPC) through up-regulation of cholinergic receptor muscarinic 4 (CHRM4), thus highlighting its contribution to lineage plasticity and drug resistance [57]. Collectively, these findings position NGF signalling as a relevant target in NEPC [57] and suggest that pharmacologic blockade of this pathway can be envisaged in aggressive disease.

Despite this evidence, however, the role and functions of NGF in PC stromal compartments are still matter of investigations. The present study aims to fill this gap and identifies CAFs as the major source of NGF within PC microenvironment. Using patient-derived CAF cultures and independent single-cell RNA sequencing datasets, we now report that NGF expression is enriched in fibroblasts as well as myofibroblasts and correlates with reduced disease-free survival.

Reactive stroma in PC is predominantly composed of CAFs, which orchestrate ECM remodelling, enhance tumor growth and invasiveness, and contribute to immune evasion and therapy resistance [47, 58–60]. Our findings show that CAFs isolated from PC patients release appreciable amounts of NGF, as assessed by ELISA and IF approaches. Although epithelial PC and immune cells might also release NGF [20, 61, 62], in our experimental setting the CAF-derived NGF represents the dominant contributor to the observed phenotypes. Nevertheless, the relative contribution of distinct cellular sources remains to be fully elucidated in vivo. Notably, NGF signalling has been implicated in tumor-associated immune modulation, as NGF and its receptors are expressed by activated T cells and attenuate CD8⁺ T-cell effector functions, potentially contributing to immune tolerance within the TME [63–66]. Thus, CAF-derived NGF may contribute to PC progression through coordinated effects on tumor cells, stromal architecture, and immune evasion.

Regardless the Gleason’s score, CAFs express TrkA and respond to NGF stimulation by increasing migratory and invasive behaviour. Pharmacological inhibition of TrkA or neutralization of NGF abrogates CAF motility, underscoring the functional relevance of this axis in stromal activation. In parallel, CAF-derived NGF promotes PC spheroid growth and induces EMT-like changes in both LNCaP and PC3 cells, specifically within 3D culture systems [67]. The latter findings are consistent with the recent evidence that ECM composition and tissue architecture critically modulate NGF/TrkA signalling outputs, with matrix stiffness amplifying YAP/TAZ-driven transcriptional programs that foster EMT and invasion [54, 68].

A relevant aspect of this study is the involvement of CAF-derived NGF in PNI, a hallmark of aggressive PC. Using a neuronal barrier model, we now report that NGF released by CAFs promotes directional invasion of PC cells along neuronal-like structures, an effect reversed by TrkA inhibition or NGF neutralization. These results extend previous observations linking CAFs to PNI through modulation of epithelial NGF secretion [55, 69] and support a model in which CAFs actively sustain perineural niches through paracrine NGF signalling. Mechanistically, NGF/TrkA activation might engage the Rho/ROCK pathway to promote axonogenesis within the TME, thus facilitating nerve–tumor interactions and directional migration [18, 70–73].

Notably, the biological relevance of CAF-derived NGF should be interpreted within the broader context of stromal organization and fibroblast plasticity in the prostate. In the normal gland, the stromal compartment constitutes the bulk of the tissue and is essential for epithelial homeostasis through coordinated stromal–epithelial interactions that support organ development and function [74, 75]. While normal prostate stroma is largely composed of smooth muscle cells, tumorigenesis is accompanied by progressive remodelling with partial replacement of smooth muscle elements by activated fibroblasts and myofibroblasts [76]. These stromal populations contribute to ECM deposition, increased tissue stiffness, and altered interactions with nerves, immune infiltrates and microvasculature. These features are collectively named “reactive stroma” [76–78]. Notably, stromal changes can occur early during PC pathogenesis and may be detectable adjacent to pre-neoplastic lesions [76], supporting the concept that stromal evolution actively participates in the disease pathogenesis.

In line with the paradigm of malignant tumors as “wounds that do not heal” [79], fibroblasts within the TME may remain in a chronically activated state, producing ECM and paracrine factors that sustain tumor growth, invasion, and therapy resistance [80]. CAFs are therefore best viewed as a heterogeneous and plastic population rather than a single cell type [81]. In PC, CAF identification typically relies on morphology, anatomical localization, absence of epithelial/leukocyte lineage markers and expression of activation-associated proteins such as α-SMA, FAP1, FSP-1, vimentin, and PDGFRα/β [82–84]. However, none of these markers is CAF-specific and they often overlap with other stromal or immune populations (e.g., FSP-1 in subsets of inflammatory macrophages; FAP expression in non-fibroblast lineages in other tumor contexts) [82]. The lack of unique markers, hence, complicates subtype assignment and contributes to discrepancies in CAF nomenclature across studies. Moreover, CAFs may arise from multiple sources, including activation of resident fibroblasts as well as recruitment/conversion of mesenchymal stem/stromal cells, adipose-derived stromal cells in extracapsular invasion contexts, endothelial-to-mesenchymal transition and senescent fibroblast states [85–89]. These concepts collectively provide a framework to interpret our data: the NGF/TrkA axis likely acts within an already dynamic reactive stroma, potentially amplifying fibroblast activation programs and reinforcing tumor–stroma crosstalk. In this context, our findings suggest that NGF/TrkA signalling may contribute to fibroblast functional polarization within the PC microenvironment. The autocrine NGF/TrkA loop identified in primary CAFs might promote acquisition of myofibroblastic features (including α-SMA induction) and enhanced invasive behaviour in 3D systems, while NGF-mediated signalling may also modulate CAF secretory programs, potentially bridging inflammatory and contractile phenotypes.

Although our study was not designed to resolve CAF lineage trajectories, the NGF-dependent modulation of fibroblast activation observed here supports the hypothesis that neurotrophic signalling participates in shaping stromal heterogeneity and tumor-promoting plasticity. Future studies integrating single-cell transcriptomics and spatial profiling approaches will be instrumental in defining how NGF differentially impacts distinct CAF subsets and contributes to tumor progression and perineural dissemination. Within this reactive and heterogeneous stromal landscape, the identification of an autocrine NGF/TrkA loop in patient-derived CAFs suggests that neurotrophic signalling may represent one of the soluble cues sustaining stromal activation and pro-invasive tumor–stroma interactions.

Conclusions

We are aware that our study may not capture the heterogeneity and complexity of PC. Nevertheless, it highlights NGF as a central mediator of reciprocal epithelial–stromal communication. By integrating clinical data, single-cell analyses, and functional 3D models, our findings support a paradigm in which NGF sustains a self-reinforcing circuit between tumor cells, CAFs and neural elements, fostering tumor aggressiveness. The translational relevance of these findings seems immediate, as they strengthen the rationale for targeting the NGF/TrkA axis in aggressive PC. Emerging therapeutic strategies, including TrkA inhibitors, NGF-neutralizing antibodies and combinatorial approaches with immunotherapy may provide opportunities to disrupt tumor–stroma crosstalk and counteract the disease progression [56, 57, 90, 91]. Future studies in patient’s-derived organoids and in vivo models would validate in the future these concepts and define their clinical applicability.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

NGF

Nerve growth factor

PC

Prostate cancer

TME

Tumor microenvironment

CAFs

Cancer-associated fibroblasts

TrkA

Tropomyosin receptor kinase A

PNI

Perineural invasion

ADT

Androgen depletion therapy

CRPC

Castration-resistant prostate cancer

AR

Androgen receptor

ECM

Extracellular matrix

EMT

Epithelial-mesenchymal transition

α-SMA

Alpha smooth muscle actin

FAP1

Fibroblast activation protein 1

DMSO

Dimethyl sulfoxide

GEO

Gene expression omnibus

UMAP embeddings

Uniform manifold approximation and projection embeddings

CM

Conditioned medium/media

scRNA-seq

Single-cell RNA sequencing

PBS

Phosphate-buffered saline

FBS

Fetal bovine serum

ECL

Enhanced chemiluminescence

EDTA

Ethylenediaminetetraacetic acid

EGTA

Ethylene glycol-bis(β-aminoethyl ether)-N, N,N′,N′-tetraacetic acid

PMSF

Phenylmethylsulfonyl fluoride

SDS-PAGE

Sodium dodecyl sulfate–polyacrylamide gel electrophoresis

YAP/TAZ

Yes-associated protein / transcriptional co-activator with PDZ-binding motif

NGFR

Nerve growth factor receptor (p75^NTR)

CHRM4

Cholinergic receptor muscarinic 4

NEPC

Neuroendocrine prostate cancer

ELISA

Enzyme-linked immunosorbent assay

RHO

Ras homolog family

ROCK

Rho-associated coiled-coil containing protein kinase

Author contributions

M.D.D.: Conceptualization, Data curation, Formal analysis, Validation, Investigation, Methodology, Writing—original draft, Writing—review and editing. P.G.: Methodology, Formal analysis, Investigation. M.D.S.: Patient recruitment, Provision of clinical samples and clinical data. A.M.: Formal analysis, Writing—review and editing. G.C.: Conceptualization, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing. All authors read and approved the final manuscript.

Funding

This work was supported by P.R.I.N.2022-PNRR (P2022BWAJN_002 to M.D.D; P2022T7FXB_002 to G.C; P2022WARR8 to P.G), P.R.I.N. (2022Y79PT4 to A.M; 202243KSPJ to P.G), PNRR- MISSIONE 6 - Progetto PNRR-POC-2023-12377405 -BEN-HuR- Codice Cup I63C24000340006 -to G.C.

Data availability

Data will be made available on reasonable request. Full and uncropped western blots have been uploaded as supplemental materials.

Declarations

Ethics approval and consent to participate

The study was performed in accordance with the Declaration of Helsinki and approved by the ethical Committees (PC-StAR Project; University ‘L. Vanvitelli’ B.684; 18/12/2017). Informed consent was obtained from all subjects for experimentation in human subjects.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interest.