Introduction

Prostate cancer (PCa) is still the second cause of cancer-related male deaths in highly developed countries [1]. A significant fraction of PCa patients arrives at diagnosis with advanced forms, while others retain indolent tumors which will never progress into aggressive stages [2, 3]. Therefore, an accurate, early diagnosis is likely to improve the outcome and the quality of life of PCa patients while reducing the over-treatment [4].

Extracellular vesicles (EVs) are membrane-enclosed bodies in the nano- to micro-meter scale that are secreted by nearly all cells and shuttle their biological content as a means of cell-to-cell communication [5, 6]. Tumor cells are now recognized to release more EVs than their normal counterpart and tumor-derived EVs can be easily isolated from bodily fluids [7,8,9,10], thus offering an exquisite source in terms of biomarkers and, mechanistically, of cancer treatment strategies [11,12,13]. The EV sub-population in the range of 30–150 nm in diameter is referred to as exosomes and has been shown to actively transport DNA, proteins, long and small RNAs [11, 14] as well as small peptides, such as prions [15]. Different from other vesicles, which are generated by random shedding mechanisms or from dying cells by discharge, exosomes drive intra- and inter-tissue cross-talk [16,17,18], are involved in physiological tissue homeostasis and immune system regulation [11] and in processes [12, 19, 20] that are often aberrant in tumors [7]. In this regard, PCa is characterized by multiple genomic lesions [Luminex assay

100 μg of EVs were lysed in 50 μl of standard RIPA buffer [(20 mM Tris-HCl pH7.2150 mM NaCl;1% NP40 (Igepal CA-630); Distilled water to volume; Proteases-inhibitors)] and diluted 1:4 in PBS for the analysis. 100 μg of parental EVs were left, non-lysed (SN), in 50 μl in the buffer (PBS) of the last step of Ultracentrifugation and was directly analyzed by Luminex. Cytokine/chemokine quantification in EV extracts and in EV(SN) was achieved by xMAP technology through a Luminex platform (Bio-Rad Laboratories, Hercules, CA, USA) equipped with a magnetic washer workstation according to the manufacturer’s protocol. RIPA (dilute 1:4 in PBS) and PBS buffer were used as background controls. Samples were analyzed using a human magnetic Luminex assay (R&D Systems, Minneapolis, MN, USA). Brain-Derived Neutrophil Factor (BDNF), CCL11, Fibroblast Growth Factor 13 (FGF-13), IL-5, IL-4, IL-23, IL-6, MMP-2 (membrane-matrix-metalloprotease-2), beta-Nerve Growth Factor (beta-NGF), N-regulin-1 beta1/NRG-1, Tumor Necrosis Factor alpha (TNF-α), Interferon gamma-induced protein 10 (CXCL10), Interferon gamma (IFN-γ), IL-2, IL-8/CXCL8, IL-17/IL-17A, CCL-2/MCP-1 and Vascular Endothelial Growth Factor (VEGF) were analyzed. The quantification was carried out with a Bio-Plex array reader (Bio-Plex 200 System) and Bio-Plex Manager (Version 6.1 Bio-Rad Laboratories, Hercules, CA, USA) software.

Statistical analysis and data representation

Protein analyses in EV samples

Student’s t or non-parametric Wilcoxon rank-sum tests were used for continuous variables to analyze the differences between groups. A p-value ≤ 0.05 was considered statistically significant. Furthermore, the receiver operating characteristic (ROC) method was used in order to find possible optimal cut-offs of the biomarkers capable of splitting patients into groups with different outcomes probabilities. Statistical analyses were conducted independently by means of SPSS® (v21.0) and MedCalc® (v10.0.1) or ‘R’ [88]. Data standardization (scaling), followed by two-way hierarchical clustering (Euclidean distance and Ward’s method was used if not specified elsewhere), were performed by means of JMP v11 (SAS Institute, Cary, NC) or ‘R’ [88] and RStudio [89]. Principal component analysis (PCA) as well as most data represented throughout the manuscript was independently reproduced by means of ‘R’ using the following packages: base, methods, utils, stats, graphics, grDevices, tcltk, openxlsx, tidyverse [90], data.table, RColorBrewer, reshape2, reshape, readxl, FactoMineR, factoextra, grid, gridExtra, circlize, cluster, dendextend and ComplexHeatmap [91].

Analysis of publicly available datasets

Messenger RNA results from Taylor’s tissue dataset (NCBI GEO accession code GSE21032) have been accessed through the Prostate Cancer Genomics Data Portal (http://cbioportal/) and combined with reported clinical data. Wilcoxon/Kruskal Wallis was used to analyze the differences between groups. GraphPad Prism v4 and JMP v11 (SAS Institute, Cary, NC) were used to perform statistical analyses.

Cell lines, PCa-derived cells and gene expression profiling (Affymetrix)

Cell cultures

All cell lines were obtained by ATCC. All cells were used as precocious (six passages) frozen stocks after arrival. They are routinely tested for Mycoplasma contamination (“PCR mycoplasma test kit”, product no. A3744, PanReac AppliChem) before EV preparations. H1299, HT1975, HT29 cells were cultivated as recommended protocols. SW480 line was maintained in RPMI and 10% of Fetal Bovine Serum (FBS). A431 and 293T cells were cultivated in DMEM (Dulbecco’s Modified Eagle’s Medium) with 10% of FBS, Glutamine (Gln) and Penicillin–Streptomycin (P/S) at standard doses. 293T cells were stable transduced with TWEEN vector [92] empty (Control) or with PD-L1 gene. Sequence-verified cDNA encoding for human PD-L1 was purchased from Dharmacon, cut with XbaI and EcoRV and inserted into TWEEN vector. It was kindly provided by Dr.Valeria Coppola. (PD-L1 Human-MGC Human CD274 Sequence-Verified cDNA-Clone ID: 30915301-Catalog Number: MHS6278-202856825)

(Lentiviral manipulation authorized by Ministry of Health rules. RM/IC/Op2/17/002.notifica I.5.i.s/2017/15 - Biotecnologie. D.L.vo 206/2001). PC20 cancer activated fibroblast were obtained by tumor primary tissue cultures as by previously published [93].

PCa-derived cells

Primary PCa cultures were derived from freshly-explanted tissue specimens (PCa-derived ex vivo model) following immortalization and phenotypic characterization. Clinical data and outcome of patients were collected for 15 years [66]. Briefly, poor prognosis group of donor patients with clinically localized PCa was defined by the presence of biochemical/local recurrence, metastasis, or disease-specific mortality, while the good prognosis group was defined by complete remission after surgery alone. Prognostic signature “Bad versus Good Prognosis profiling” was obtained in PCa cells by Affymetrix array (Human U133A Gene ChIP platform) using PCa cells derived from patients with different progression of disease (recurrent versus non-recurrent disease). Regulated biological processes were identified by the GOAL Web-based application and Gene Ontology (GO) terms with p < 0.01 considered differentially regulated (false discovery rate = 0.013).

Affymetrix data analysis

Affymetrix Gene Chip scanning was analyzed by customized R language-based script [88] using Bioconductor (http://www.bioconductor.org) for quality-control analysis, data normalization, hierarchical clustering, and identification of differentially expressed transcripts. Biological processes and molecular functions involved were identified by the GOAL Web-based application and the Unigene Build 154 according to the Gene Ontology (GO; http://www.geneontology.org) Consortium classification. Genes reported with p < 0.01 were considered differentially regulated (false discovery rate = 0.013 [65, 66]).