Neuropilin1-dependent paracrine signaling of cancer cells mediated by miRNA exosomal cargo

Abstract

Background

Neuropilin-1 (NRP1) is a transmembrane protein involved in surface receptor complexes for a variety of extracellular signals. NRP1 expression in human cancers is associated with prominent angiogenesis and advanced progression stage. However, the molecular mechanisms underlying NRP1 activity in the tumor microenvironment remain unclear. Notably, diffusible forms of NRP1 in the extracellular space have been reported, but their functional role is poorly understood.

Methods

Extracellular vesicles (EV) were isolated from conditioned media of diverse cancer cells. The quality of exosome-enriched preparations was validated by the presence of specific markers in western blotting, as well as by light scattering and nanoparticle tracking analysis. Wound healing, transwell, and digital real-time migration assays were carried out to assess the activity of cancer cell-derived exosomes in the regulation of endothelial cells. RNA interference was applied to obtain NRP1 knock-down, and cDNA transfer to achieve its overexpression, in exosome-releasing cells. The micro-RNA profile carried by exosomes was investigated by Next Generation Sequencing. miRNA-Scope in situ hybridization was used to assess the transfer of miRNA exosome cargo to target cells, and immunofluorescence analysis revealed expression regulation of targeted proteins. miRNA activity was blocked by the use of specific antago-miRs.

Results

In this study, we show that diverse human cancer cells release NRP1 embedded in exosome-like small extracellular vesicles, which mediate a previously unknown NRP1-dependent paracrine signaling mechanism regulating endothelial cell migration. By transcriptomic analysis of the cargo of NRP1-loaded exosomes, we found a significant enrichment of miR-210-3p, known to promote tumor angiogenesis. Gene knock-down and overexpression experiments demonstrated that the loading of miR-210-3p into exosomes is dependent on NRP1. Data furthermore indicate that the exosomes released through this NRP1-driven mechanism effectively transfer miR-210-3p to human endothelial cells, causing paracrine downregulation of the regulatory cue ephrin-A3 and promotion of cell migration. The mechanistic involvement of miR-210-3p in this pathway was confirmed by applying a specific antago-miR.

Conclusions

In sum, we unveiled a previously unknown NRP1-dependent paracrine signaling mechanism, mediated by the loading of pro-angiogenic miR-210-3p in exosomes released by cancer cells, which underscores the relevance of NRP1 in controlling the tumor microenvironment.

Background

Cancer cell behavior, beyond the intrinsic malignant potential, is significantly influenced by signals from the microenvironment, which emphasizes the importance of the crosstalk between neoplastic cells and the surrounding stromal cells [1, 2]. A continuous supply of oxygen and nutrients, essential for cancer progression, is primarily reliant on angiogenesis [3,4,5,6]. While vascular endothelial growth factor (VEGF) is recognized as the principal regulator of this process, substantial advancements have been made in identifying other signals that also govern angiogenesis [7]. Interestingly, the transmembrane proteins Neuropilins (NRP1 and NRP2) act as co-receptors for extracellular cues controlling angiogenesis, including Semaphorins, VEGF-family members, and adhesion molecules [8, 9]. Moreover, neuropilins are involved in the response to additional growth factors relevant in the tumor microenvironment (TME), such as Hepatocyte Growth Factor, Fibroblast Growth Factor, Transforming Growth Factor beta, and Galectin [10]. Notably, several studies have shown that an elevated expression of NRPs correlates with advanced tumor stage, resistance to therapies, and poor patient prognosis [11, 12].

NRP1 and NRP2 share a similar protein structure, featuring a short intracellular tail ends in a consensus sequence capable of interacting with PDZ (PSD-95/Dlg/ZO-1 homology) domains. However, their capacity for intracellular signaling remains poorly understood [12,13,14]. Of note, it has been reported the finding of soluble extracellular forms of NRP1, due to alternative mRNA splicing or proteolytic shedding, but their functional relevance and mechanism of action have not been elucidated [15,16,17,18,19,20].

Extracellular vesicles (EVs) are gaining growing importance for cell–cell communication within the TME. In particular, exosomes (a subtype of EVs with diameters ranging from 30 to 150 nm) have been implicated in the regulation of tumor angiogenesis [21,22,23,24]. Thus, tumor-derived exosomes can stimulate endothelial cells (EC) proliferation, migration and tubulogenesis [25,26,27,28]. The underlying mechanisms have not been fully elucidated and are likely to be varied, depending on EV cargoes; however, accumulating “omic” studies have highlighted some of the pivotal proteins and RNAs involved in this activity [22]. In particular, microRNAs (miRNAs), a class of small non-coding RNAs interfering with mRNA stability and translation, can be selectively packaged into EVs and hence delivered to recipient cells, where they can hamper the expression of target genes and affect proliferation, apoptosis, differentiation, migration, etc. [29, 30]. Exosomal miRNAs have been crucially implicated in cell-to-cell communication in the TME, controlling EC and playing a critical role in angiogenesis [31, 32].

In the present study, we show for the first time that different cancer cells consistently release diffusible NRP1 carried by exosomes into the extracellular environment. Our data further reveal that these NRP1-loaded exosomes are endowed with paracrine signaling capacity, promoting EC migration. Notably, we found that NRP1 expression in the producer cells controls the miRNA cargo composition of exosomes, particularly enriching miR-210-3p, a well-established regulator of ECs and promoter of angiogenesis [33,34,35]. Furthermore, we demonstrated that exosomal miR-210-3p is delivered to ECs, significantly conditioning the expression of the targeted EC-regulatory factor ephrin-A3 [36,37,38,39,40]. In sum, this study reveals that the expression of NRP1 in cancer cells elicits a novel paracrine signaling mechanism controlling ECs through miR-210-3p loaded exosomes.

Methods

Mammalian cell cultures

PC3 (prostate cancer cells), A549 (lung cancer), OVCAR3 (ovarian cancer) cell lines were grown in RPMI medium (PanBiotech); MDA-MB-231, HT29, and HCT116 cell lines were routinely propagated in DMEM medium (PanBiotech); both culture media were supplemented with 10% Fetal Bovine Serum (FBS), 1% penicillin/streptomycin (Corning), 1% L-Glutamine (Corning) and 1% Hepes (Corning). Human umbilical vein endothelial cells (HUVECs) were obtained from Lonza Sales Ldt (Switzerland) and were cultured in EGM-2 Endothelial Cell Growth Medium-2 (Endothelial Basal Medium EBM-2 + EGM-2 BulletKit, Lonza, which contains 2% FBS) and used at passage number 2–6. Human embryonic kidney cells (HEK293T) were grown in Iscove’s medium (PanBiotech) supplemented with 10% Fetal Bovine Serum (FBS), 1% penicillin/streptomycin (Corning), 1% L-Glutamine (Corning) and 1% HEPES (Corning). Cell cultures were maintained at 37 °C in a 5% CO₂ humidified atmosphere.

Antibodies and other reagents

Anti-NRP1 (ab81321) and anti-Calnexin (ab22595) were from Abcam. The following antibodies were from Santa Cruz Biotechnology: anti-Alix (sc-60913), anti-CD9 (sc-59140), anti-GAPDH (sc-32233) and anti-ephrin-A3 (sc-393727). Anti CD81 (PA5-13,582) was from Thermo Fisher. Sigma-Aldrich provided: anti-Argonaute2 (A308366) and anti-Vinculin (V9131). Anti-CD31 (PECAM-1) (89C2) and anti-Flotilin (3253) were from Cell Signaling. Moreover, miRIDIAN hairpin inhibitor hsa-miR-210-3p was from Dharmacon™ (IH-300565–05-0005).

Isolation of exosome-enriched small EVs

Small EVs enriched in exosomes were isolated from conditioned supernatant according to the International Society of EVs recommendations [41]. Approximately 8 × 10⁶ exosome-releasing cells were seeded in 150 mm culture dishes, grown for 48 h in complete medium and then switched in serum-free medium. The cell-conditioned medium was eventually collected after 24/48 h and centrifuged at 2000 g for 30 min to eliminate cellular debris. The supernatant was concentrated using 100 kDa molecular weight Amicon Ultra-15 Centrifugal Filter (Merck Millipore, Billerica, MA, USA) and Total Exosome Isolation kit (Invitrogen, MA, USA) was used to isolate exosomes, following manufacturer’s instructions. Precipitated exosomes were recovered by centrifugation at 11600 g for 60 min, and the pellets were finally resuspended in PBS and stored at -80 °C.

Characterization of isolated exosomes

In initial experiments, EVs have been characterized by dynamic light scattering (DLS) with Zetasizer Nano-ZS (Malvern Panalytical Ltd, UK). Subsequent, size distribution profile and concentration of purified exosomes were assessed by nanoparticle tracking analysis (NTA) with NanoSight NS300 (Malvern Panalytical Ltd, UK). Five repeat captures, each of 60 s, were acquired at controlled temperature of 25 °C. Detection threshold was set to 5 to reduce noise. Data were processed using NTA 3.2 software (Malvern Panalytical Ltd). D10, D50 and D90 values, mode, mean, and particle concentration were reported. A video of 60-s duration was taken (with a frame rate of 30 frames/sec), and particle movement was analyzed using NTA software. Exosome protein quantification was done by resuspending purified vesicles in a lysis buffer containing TritonX100 1%, followed by Bradford Protein Assay (Bio-Rad Laboratories Inc, Hercules, CA, USA), according to the manufacturer’s instructions.

Western blotting

For immunoblotting analysis, cells were washed in PBS, harvested and lysed by incubation for 30 min at 4 °C in Cell Lysis Buffer (Cell Signaling, Danvers, MA, USA) containing 1 mM PMSF (Cell Signaling, Danvers, MA, USA) and a complete protease/phosphatase inhibitor cocktail (Cell Signaling, Danvers, MA, USA). Then the cells were scraped, and total lysates were centrifuged for 15 min at 14,000 × RPM in a refrigerated microfuge. Protein concentration was determined by Bradford Protein Assay (Bio-Rad Laboratories Inc, Hercules, CA, USA), according to the manufacturer’s instructions. Equal amounts of proteins were eventually solubilized in Laemmli buffer, separated by SDS/PAGE (Mini-PROTEAN® TGX™ Precast Protein Gels, or Mini-PROTEAN TGX stain-free precast PAGE gels, Bio-Rad Laboratories Inc.), and transferred to a nitrocellulose membrane (GE Healthcare, Piscataway, NJ, USA). Membranes were blocked with Every Blot Blocking Buffer (#12,010,020) for 5 min at RT and incubated with the primary antibody overnight + 4 C°. Blots were then incubated with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling) for 1 h at RT. Signals were captured by ChemiDocTM Imaging System (Bio-Rad Laboratories Inc., Hercules, CA, USA) using the Clarity Western ECL enhanced chemiluminescence system (Bio-Rad Laboratories #170–5061), and densitometric analysis was performed with Image LabTM Touch Software (Bio-Rad Laboratories).

Gene expression knockdown by RNA-interference

To achieve stable NRP1 knockdown, we transduced cancer cells with the puromycin-selectable lentiviral construct TRCN0000323055 (Sigma-Aldrich), expressing NRP1-targeted short hairpin RNA (shRNA); pLKO empty vector was used as mock control. Lentiviral-mediated gene transfer in mammalian cells was accomplished as previously described [42]. Briefly, non-replicating viral particles containing transfer plasmids were produced in HEK-293T packaging cells by cotransfection with the calcium phosphate precipitation method. Target cells were then incubated with the conditioned media deriving from transfected 293T cells, in the presence of polybrene 8ug/ml, for 8–12 h. Transduced cells were selected by culturing in presence of 1 µg/ml puromycin.

Cell transfection with cDNA constructs and antagomiR

HEK-293T cells were transiently transfected with cDNA expression plasmids using the calcium phosphate precipitation method. Briefly, two hours before transfection, the cells were incubated with fresh ISCOVE’s medium 10% FBS. Then, 10-20 μg of plasmid cDNA were dissolved in 0.1 × TE buffer with 250 mM CaCl₂, and incubated 5’ RT; then, an equal volume of 2 × HEPES-Buffered Saline was added dropwise to the tube, while swirling on a vortex (all solutions had been prepared using ultrapure water, endotoxin free). Finally, we transferred the transfection mix into cell culture medium. In particular, to produce engineered EVs we transiently transfected HEK-293T cells with previously described lentiviral transfer plasmids encoding human NRP1 or NRP2 (in the pLVX backbone) [43, 44]; as controls, we transfected the cells with mock pLVX or pLKO empty vectors, in multiple experiments. The day after transfection, we switched the cells to serum free medium, and 24 h later we harvested the supernatant for EV isolation.

In other experiments, A549 cells were transiently transfected with 20 nM of antagomiR hsa-miR-210-3p using Lipofectamine 3000 (Life Technologies) according to manufacturer’s instructions.

Transwell migration assay

In order to quantify EC migration, we applied Transwell inserts (Corning), carrying a semipermeable membrane with 8 µm pore diameter. To allow haptotactic cell migration, the lower side of the filter was coated with 10 μg/ml fibronectin. HUVECs were seeded in the upper chamber (25,000/well) in EGM-2 medium (containing 2% FBS) and allowed to migrate through the pores of the membrane into the lower compartment, containing the same medium. Equal amounts (10 μg) of purified exosomes were added in the upper compartment of the inserts, together with ECs. After a minimum of 6 h, we evaluated HUVEC migration. After removing non-migrated cells from the upper side of the membrane with a cotton swab, the cells that had passed the membrane were fixed with PFA 4% for 15 min, then stained with crystal violet for quantification. The migrated cells were imaged under a microscope and quantified by dissolving the dye in 10% of acetic acid, followed by measuring the absorbance at 595 nm using Varioskan Lux microplate reader.

Wound-healing assay

HUVEC (40,000 cells/well) were seeded in 24-well dishes, previously coated with gelatin 0.1%, allowing them to grow until forming a confluent monolayer. A scratch was then made with a 200 µl pipette tip, followed by gentle washes to remove detached cells; the wounded monolayer was initially imaged by phase contrast microscopy at T₀. The culture was then replenished with fresh EGM-2 medium (containing 2% FBS), and the cells were incubated (or not) with equal amounts of different exosome preparations at 37 °C for up to 24 h, allowing them to migrate to fill the gap. Eventually, the HUVEC monolayer was rinsed with PBS, fixed with 4% paraformaldehyde for 15 min, and finally stained with 1% crystal violet for 30 min. Then, the images were acquired by phase contrast microscopy and analyzed quantitatively by ImageJ software. The same wounded area imaged at T_0, for each condition was analyzed at the end of the experiment to quantitatively assess the percentage of wound closure.

Real-time digital assessment of endothelial cell migration

We monitored the directional migration of HUVECs in real-time by using the CIM-Plate 16 of the xCELLigence RTCA DP instrument, as previously described [45]. Briefly, the bottom side of the upper chamber (facing the lower chamber) of the CIM-Plate was coated with 30 μl of fibronectin (3 μg/ml) for 30 min inside the tissue culture hood. Each lower chamber well was first filled with 160 μl of serum-free medium and then assembled to the upper chamber. The assembled plate was incubated at 37 °C for one hour to equilibrate (30 μl of serum-free medium was added to each well of the upper chamber). ECs were detached by means of trypsin–EDTA and resuspended at a final concentration of 30000 cells/100 μl. The BLANK step was started to measure the background impedance of cell culture medium, which was then used as reference impedance for calculating CI values. 100 μl of cell suspension (30,000 cells) with or without exosomes, were then added to each well of the upper chamber. The CIM-Plate 16 was placed in the RTCA DP Instrument equilibrated in a CO2 incubator. ECs migration was continuously monitored using the RTCA DP Instrument. The mean, standard deviation, and p values were calculated on the CI data exported from RTCA instrument, considering the replicates of each experimental condition, normalized to respective controls (put as 100% at the endpoint). The statistical significance of differences between time course series of periodic measurements was analyzed by GraphPad Prism 8.0.0 software, applying two-way ANOVA with Bonferroni correction.

Paracrine signal exchange co-culture assay

Transwell inserts were exploited to assay paracrine signaling between A549 cancer cells seeded in the upper chamber (20,000/each) and separated by an 8 µm nanopore semipermeable membrane from HUVEC (12,000 cells/well), grown onto coverslips placed at the bottom of the well. After 24 h incubation, the insert was removed, endothelial cells were fixed for 15 min in 4% paraformaldehyde, and further subjected to immunofluorescence analysis.

Exosome-borne miRNA isolation and analysis

ExoRNeasy serum/plasma kit (QIAGEN) was used to isolate EV-associated miRNAs from cell-conditioned media, as described in the manufacturer´s handbook. RNA yield and size distribution were verified using Agilent 2100 Bioanalyzer with small RNA kit (Agilent Technologies, Foster City, CA, USA). 5 µl of RNA derived from each condition was used to build a library using QIAseq miRNA Library Kit (Qiagen), according to the manufacturer’s protocol. Next, small RNA sequencing was conducted by NextSeq 550Dx Instrument (Illumina) + NextSeq 500/550 Mid Output Kit v2.5 (150 Cycles).

Bioinformatic analysis of sequencing data

FastQ files have been analyzed with a custom pipeline based on BCGSC miRNA Profiling Pipeline (https://doi.org/https://doi.org/10.1093/nar/gkv808). The Adapter sequences have been removed by the FastQ files using cutadapt software. The reads have been aligned to the human genome (hg19 assembly) using BWA aligner in aln mode and the output has been converted into SAM file using samtools. The miRNA database (miRbase version 20) required to run the BCGSC pipeline has been downloaded, installed locally, and used to annotate the Sam files to produce a list of identified precursor (mir) miRNAs, the corresponding mature MIMAT codes, and the number of reads identified. The Mimat code has been converted into the mature miRNA nomenclature (miR) that was used for subsequent analyses. The number of reads for each miRNA have been normalized assuming 100.000 total reads per individual sample. All subsequent analyses have been performed based on normalized values. For each test/control sample pair, the fold-change of miRNA levels have been expressed as the log(2) ratio of normalized reads.

miRNA expression analysis by qRT-PCR

miRNAs purified from exosomes or cells, as described above, were subjected to reverse transcription by TaqMan™ Advanced miRNA cDNA synthesis kit, according to the manufacturer’s instructions. The quantification was done with TaqMan™ Advanced miRNA Assays, according to the manufacturer’s instructions, applying the following: Assay ID-477970_mir (for hsa-miR-210-3p), Assay ID-477985_mir (for hsa-miR-22-3p), and Assay ID-477864_mir (for hsa-miR-103a-2). The qPCR reactions were run in triplicate by StepOne thermal cycler (Applied biosystems; 48-well format).

miRNAscope in situ hybridization coupled with immunofluorescence on adherent cultured cells

miRNAscope™ (Advanced Cell Diagnostic, USA) on HUVECs was performed according and adapting to the manufacturer’s protocol as in ref [46]. Briefly, ECs seeded on glass coverslips were fixed with 4% PFA for 15 min and dehydrated through passages in 50%, 70% and 100% ethanol, for 1 min each, at RT. Each coverslip was then immobilized on a glass slide by adding a small drop of mounting medium, and rehydration and cell permeabilization with 0.1% Tween for 10 min at RT was achieved, followed by overnight incubation with a primary antibody (anti-CD31, 89C2) at + 4 °C. After fixation with 10% neutral buffered formalin for 30 min, and protease III (322337) treatment (30 min at 40 °C), cells were incubated for 2 h at 40 °C with the miR-210-3p probe (728551-s1 miRNAScope tm probe-sr-has-miR210-3p-s1), according to the manufacturer’s protocol. The cells were rinsed and then incubated with the secondary antibody Alexa Fluor-488 donkey anti-rabbit IgG (1:200; Molecular Probes, A31572), counterstained with DAPI (Sigma, D9542) and a covering slide were glued with Fluoromount (Sigma, F4680). Samples were examined under a confocal laser scanning microscope (CLSM; Leica SP5, Germany) equipped with four laser lines: violet diode emitting at 405 nm for DAPI, argon emitting at 488 nm and helium/neon emitting at 543 nm and 633 nm. Densitometric analysis of detection signal was done on confocal images acquired through the 60 × objective at 0.5 zoom factor, applying constant confocal settings. Images were analyzed with Fiji software (http://imagej.net/software/fiji). Specifically, for each sample, different independent fields were randomly acquired (n ≥ 3 for sample). For each image, the miRNA-associated signal (optical density O.D.) was quantified in the total area (1024 × 1024 pixels). All quantitative analyses were conducted blind to the experimental group assignment. Data were normalized over controls.

Immunofluorescence and confocal analysis

Cells seeded on glass coverslips were fixed for 15 min in 4% paraformaldehyde, permeabilized with 0.1% Triton/phosphate-buffered saline for 5 min at RT, and blocked by incubation with 5% normal donkey serum for 1 h. After incubation with anti-ephrin-A3 (sc-393727) primary antibodies overnight, and subsequent rinses, fluorochrome-conjugated secondary anti-mouse were added for 1 h at room temperature, in presence of 4,6-diamidino-2-phenylindole (DAPI) to reveal cell nuclei. The coverslips were finally washed and mounted on slides. Images were acquired with a confocal laser scanning microscope equipped with a 63X. Confocal settings were maintained constant throughout the acquisition of slides of the different experimental groups. Images were analyzed off-line with Fiji software (http://imagej.net/software/fiji).

Results

Cancer cells release exosome-borne NRP1 in the extracellular environment

NRP1 expression in cancer cells has been associated with tumor angiogenesis [47]; however, the underlying mechanisms are poorly understood. It has been suggested that NRP1-induced paracrine signals in TME regulate ECs, but these mechanisms are still elusive. Interestingly, NRP1 is also present in a soluble form within extracellular fluids, though its functional significance remains unclear [19, 48].

Accumulating evidence underscores that cell–cell communication within the TME is mediated not only by soluble molecules, but also by extracellular vesicles (EVs), such as exosomes, which transfer signaling cargoes from producer to recipient cells [49]. This led us to investigate whether transmembrane NRP1 is transported by cancer cell-derived exosomes and to explore the potential relevance of this mechanism in the regulation of ECs. To address this, we first analyzed the presence of NRP1 in EVs released by diverse cancer cells representative of common tumor types. In particular, small EVs enriched in exosomes were purified from the conditioned medium (CM) of human cancer cell lines derived from lung adenocarcinoma (A549), colorectal carcinoma (HCT116 and HT29), breast carcinoma (MDA-MB-231), ovarian carcinoma (OVCAR3), and prostate carcinoma (PC3). We found that NRP1 is present at varied levels in exosomes released by cancer cells (Fig. 1), partially reflecting the protein expression in cell lysates (Suppl. Figure 1A). The identity of purified extracellular vesicles was confirmed through Dynamic Light Scattering (DLS) and Nanosight analysis (Suppl. Figure 1B-C), as well as by the presence of specific EV markers Alix and CD81 (Fig. 1 and Suppl. Figure 1A), while the intracellular vesicle marker calnexin was not detected in this fraction (Fig. 1 and Suppl. Figure 1A).

Fig. 1

NRP1 expression in tumor-derived exosomes. Western blotting analysis revealing NRP1 protein (130 kDa) in representative batches of purified exosomes derived from the conditioned medium of the indicated cancer cell lines. 10 µg of EV proteins were loaded in each lane. The detection of the exosomal marker Alix (95 kDa) and the absence of endoplasmic reticulum marker Calnexin (90 kDa) provided a control for EV purification. Further EVs characterization is shown in Suppl. Figure 1

Notably, higher levels of NRP1 were observed in the exosomes derived from A549 lung carcinoma cells and PC3 prostate cancer cells. Consequently, subsequent experiments focused on these two cancer cell models. Importantly, by analyzing NRP1 distribution within a specific volume of cancer cell-derived CM and comparing it to the exosomes isolated from the same volume, we revealed for the first time that the majority of NRP1 released by cancer cells is found in the exosomal fraction (Suppl. Figure 2).

Cancer cell-derived exosomes elicit EC migration in NRP1-dependent manner

We evaluated the influence of cancer cell-derived exosomes on EC activity using HUVEC scratch wound healing assays, which serve as a model for EC migration. Notably, we observed a significant increase in HUVEC migration following incubation with exosomes derived from A549 cancer cells (Fig. 2A), indicating that these cells release exosome-borne pro-angiogenic signals. Consistent results were obtained using exosomes derived from PC3 cells, whereas incubation with the same concentration of vesicles released by HCT116 or HT29 cells, which were found to carry limited NRP1 protein, was ineffective (Fig. 2A).

Fig. 2

ECs migration is induced by cancer cell-derived exosomes in NRP1-dependent manner. A Representative images and quantification of wound healing assays, scoring the migration of HUVECs, either untreated (NT) or in the presence of 10 µg exosomes derived from the indicated cancer cells. The yellow boxes mark the scratch area imaged at T₀, which was analyzed to calculate the fraction of migrated cells after 24 h. Box plots show wound closure efficiency in different conditions, summarizing the results of multiple experiments (n ≥ 3, each including duplicate wells). In the plots, the box always extends between 25 and 75th percentile values; the line in the middle of the box is plotted at the median, while the whiskers span from minimum to maximum values. Statistical analysis was performed by one-way ANOVA followed by Tukey’s test and asterisks indicate p-values: **p < 0.01. B Wound healing assays comparing HUVEC migration in the presence of 10 µg exosomes derived from either NRP1-depleted (shNRP1) cancer cells or mock controls (transduced with the empty vector pLKO). The fraction of migrated cells after 24 h was calculated as above. Box plots show wound closure efficiency under different conditions, summarizing the results of multiple experiments (n ≥ 3, each with duplicate wells). Statistical analysis was performed by one-way ANOVA followed by Tukey’s test; p-values: ****p < 0.0001; **p < 0.01. C The graph shows a digital real-time analysis of HUVEC migration in the presence of 10 µg exosomes derived from either NRP1-depleted (shNRP1) A549 cancer cells or pLKO-mock controls (same as in panel B), assessed by the xCELLigence system (see Methods for details). The timelines, based on continuous measurements (approx. every 10 min), represent average values of multiple separate wells (n = 7 mock, n = 6 shNRP1) analyzed over three independent experiments; error bars are shown at periodic time points. Statistical analysis was done by two-way ANOVA with Bonferroni correction; p-values: ****p < 0.0001; *p < 0.05

To determine the functional importance of NRP1 in this paracrine signaling mechanism, exosomes were purified from the CM of NRP1-depleted A549 and PC3 cancer cells, which had undergone targeted shRNA transfer, as well as from mock control cells transduced with the empty vector pLKO (see Suppl. Figure 3). Notably, by treating HUVECs with comparable amounts of exosomes, we found that NRP1 depletion severely impaired the capacity of EVs to induce cell migration (Fig. 2B). To achieve further confirmation of this effect, we performed real-time analyses of cell migration with xCELLigence system, based on digital automated measurements of electric impedance due to cells passing through a semipermeable membrane with 8 µm pores. The results, shown in Fig. 2C, confirmed our previous conclusions that cancer cell-derived exosomes promote EC migration in NRP1-dependent manner. Altogether, these findings underscore the crucial role of NRP1 in this paracrine signaling pathway.

NRP1-loaded exosomes enhance endothelial cell migration

To further demonstrate the specific role of NRP1 in exosome-mediated signaling, we undertook a gain-of-function approach, by overexpressing this protein (or its homologue NRP2 for comparison) in immortalized HEK293T normal kidney cells, bearing low levels of endogenous neuropilins. These cells are known to release abundant exosomes [50]; thus, vesicles retrieved from the CM of transfected cells were isolated and characterized, confirming that neuropilin loading was reflected in EV cargoes, reaching a comparable level as found in cancer cells (see Suppl. Figure 4). We then investigated the effect of these engineered exosomes on HUVEC migration, using both wound healing and Transwell™ assays. Notably, unlike those derived from cancer cells, a similar amount of exosomes released from empty vector mock-transfected non-tumoral HEK293T cells did not induce EC migration in either migration assay (compared to untreated conditions). In contrast, we observed a significant increase in HUVEC migration upon treatment with NRP1-loaded exosomes (exoOverNRP1), compared to controls (exo-Mock) (Fig. 3A, B). Interestingly, NRP2-loaded exosomes induced only a slight non-significant increase in EC migration, indicating a specific role for NRP1 in mediating this paracrine signaling mechanism.

Fig. 3

Endothelial cell migration in response to NRP-loaded exosomes. A Representative images and quantification of Transwell assays assessing HUVEC migration upon treatment with 10 µg of either ExoOverNRP1, ExoOverNRP2, or Exo-Mock control exosomes (released from HEK293T cells transfected with the respective constructs, or with empty vectors pLKO or pLVX, in multiple experiments). Box plots show HUVEC migration rates under different conditions, summarizing the results of multiple experiments (n = 4, each with duplicate wells). Values were normalized to the mean of controls. In the graphs, the box extends between 25 and 75th percentile values; the line in the middle of the box is plotted at the median, while the whiskers span from minimum to maximum values. Statistical analysis was performed by one-way ANOVA followed by Tukey’s test and asterisks represent p-values: ***p < 0.001; **p < 0.01. B Representative images and quantification of wound healing assays of HUVEC similarly treated with 10 µg of the same exosomes above. Box plots show migration rates in different conditions (as above), summarizing the results of multiple experiments (n = 4, each with duplicate wells). Statistical analysis performed by one-way ANOVA didn’t show significant differences comparing all the experimental groups; however, Student’s t-test revealed clearly significant differences between ExoNRP1 and Exo-Mock p-value: *p < 0.05. C The graph shows a digital real-time analysis of HUVEC migration in the presence of 10 µg exosomes derived from either NRP1-overespressing (ExoOverNRP1) HEK293T cells or pLVX-mock controls (same as in the above panels), assessed by the xCELLigence system (see Methods for details). The timelines, based on continuous measurements (approx. every 10 min), represent average values of multiple separate wells (n = 6 mock, n = 6 OverNRP1) analyzed over two independent experiments; error bars are shown at periodic time points. Statistical analysis was done by two-way ANOVA with Bonferroni correction; p-values: ****p < 0.0001; ***p < 0.001

Also in this case, we performed real-time analyses of directional EC motility with xCELLigence system, confirming that NRP1 expression in cancer cells promotes exosome-borne signals eliciting endothelial cell migration (Fig. 3C).

NRP1-dependent regulation of miRNA exosomal cargo

Previous studies have suggested that NRP1 binding in trans to ECs can trigger VEGFR2 phosphorylation and cell migration [19]. We initially aimed to test this hypothesis in our setting but found no evidence that the treatment with NRP1-loaded exosomes directly activates VEGFR2 in HUVECs at significant levels (Suppl. Figure 5). We then postulated that NRP1 expression might mediate the exosomal loading of other regulatory molecules capable of inducing EC migration. In particular, EVs have often been implicated in the control of angiogenesis through the transfer of microRNAs with specific regulatory functions [22]. Thus, we analyzed by next generation sequencing the miRNome expression profile of three independent preparations of NRP1-loaded and two of NRP2-loaded exosomes, compared to n = 3 corresponding preparations of control exosomes (obtained in parallel from empty vector-transfected HEK-293T cells, as described above). 181 different miRNA sequences yielded NGS reads above the reliable detection threshold in at least one of the samples; but only for 32 of them log(2) fold changes above ± 1.0 were confirmed in at least two experiments. Interestingly, among the most consistently differentially regulated miRNAs, miR-210-3p, well known to regulate ECs and promote angiogenesis [36, 38], was significantly and specifically induced in NRP1-loaded exosomes (Table 1 and Suppl. Table 1).

Table 1 Differentially expressed miRNAs in NRP1-loaded exosomes (vs. controls), showing consistent and concordant results in at least 2 (out of n = 3) independent experiments, with an average log(2) Fold Change beyond ± 1.0. The right column indicates whether miRNA expression changes in NRP1-loaded exosomes are concordant or discordant with those observed in NRP2-loaded vesicles. A more comprehensive set of data from these experiments is provided in Suppl. Table 1

Notably, miR-210 has also been described as an onco-miRNA due to its activity in promoting tumor angiogenesis [33,34,35]. The NRP1-induced enrichment of miR-210-3p in the exosomes was confirmed by qPCR (Fig. 4A); conversely, miR-210-3p levels were profoundly downregulated in exosomes released by NRP1-depleted cancer cells (Fig. 4B). Interestingly, we found that this regulation of exosomal miRNA content was not due to changes in miR-210-3p levels in the cytosolic lysates of producer cells (Fig. 4A, B), suggesting a NRP1-dependent regulation of exosomal miRNA cargo loading. Since Argonaute-2 (Ago2) is a known miRNA-binding protein, we asked whether its loading in exosomes may be controlled by NRP1 levels; however, data shown in Suppl. Figure 6 seem to rule out this hypothesis.

Fig. 4

NRP1-dependent regulation of miR-210-3p exosomal cargo. miR210-3p expression was assessed by qPCR in: (A) exoMock and exoOverNRP1, and respective HEK293T producer cells; (B) exoMock and exo-shNRP1, and respective A549 producer cells. The analysis shows an upregulation of miR210-3p in exoOverNRP1 and -conversely- a downregulation in NRP1-depleted exosomes; on the other hand, miRNA levels in the lysates of producer cells are unaffected by overexpression or silencing of NRP1. Statistical analysis was performed by Student’s t-test and asterisks represent p-value: **p < 0.001; *p < 0.05

Of note, two other miRNAs consistently regulated by NRP1 emerged from the transcriptomic screening of exosome cargo, namely miR-103a-2 and miR-22-3p (see Table 1). The regulation of these additional, and poorly understood, candidates of potential interest was further validated by qPCR analysis (Suppl. Figure 7); their functional role downstream NRP1 will be investigated in future studies.

Exosomal miR-210-3p is efficiently transferred into ECs, downregulating the expression of its target ephrin-A3

To assess the potential functional impact of exosome-borne miR-210-3p on EC, we first aimed to demonstrate the efficient transfer of this miRNA into targeted cells. The intracellular delivery of miR210-3p from exosomes into HUVECs was indeed confirmed by miRNAscope in situ hybridization combined with immunofluorescence (Fig. 5). Our results clearly showed increased miR-210-3p levels in ECs incubated for 10 min with NRP1-loaded exosomes, compared to the basal content in untreated HUVECs or cells incubated with control exosomes.

Fig. 5

miRNAscope analysis of endothelial cells. A Representative images of miR-210-3p miRNAscope analysis combined with anti-CD31 immunofluorescence performed on fixed HUVECs treated with 10 µg of exo-Mock or exoOverNRP1 (or left untreated). Confocal images individually revealing DAPI (blue), CD31 (green) and miR-210-3p (red) labeling, or merged signals. Scale bar = 20 $\mu m$. B Box plot summarizing the results of multiple (n = 3) experiments, by the analysis of multiple independent fields (see Methods for details). Values were normalized to the mean of controls. In the graph, the boxes extend between 25 and 75th percentile values; the line in the middle of the box is plotted at the median, while the whiskers span from minimum to maximum values. Statistical analysis was performed by one-way ANOVA followed by Tukey’s test and asterisks represent p-values: ****p < 0.0001; **p < 0.001

To further corroborate the biological relevance of exosomal miR-210-3p transferred into ECs, we aimed at verifying its effect on targeted transcripts. Bioinformatic analyses have consistently indicated ephrin-A3 as a valid miR-210-3p target; and previous studies have reported that miR-210 regulates angiogenesis, cell migration, and tumor progression through targeting ephrin-A3 expression [36,37,38,39]. Thus, we asked whether miR-210-3p transferred to ECs by NRP1-loaded exosomes could regulate ephrin-A3 levels, indicating the functional impact of this NRP1-dependent paracrine signaling on EC protein expression. HUVECs were therefore incubated with NRP1-loaded (or control) exosomes, and ephrin-A3 expression was assessed by immunofluorescence. Confocal analysis of ephrin-A3 endogenous staining showed that it was weakly but readily detectable in HUVECs treated with exo-Mock control exosomes. In contrast, upon exoOverNRP1 treatment, ephrin-A3 immunostaining became extremely faint and quantitative analysis of the fluorescence signal showed its significant reduction compared to controls, consistent with the activity of miR-210-3p transferred by exosomes (Fig. 6A). Conversely, ephrin-A3 immunostaining increased in HUVECs treated with NRP1-depleted (vs. control) exosomes (Fig. 6B). Intriguingly, we show in Suppl. Figure 8 that NRP1 expression changes did not directly affect ephrin-A3 levels into exosome-producing cells, consistent with the evidence that NRP1 does not impact transcription and intracellular levels of miR-210-3p, but rather its loading in the released exosomes (see Fig. 4, above).

Fig. 6

Ephrin-A3 regulation via exosomal miR-210-3p. Representative confocal images showing ephrin-A3 immunostaining (green) in fixed HUVECs treated for 24 h with: (A) Exo-Mock or exoOverNRP1 purified from HEK-293T cells; (B) exo-Mock or exo-shNRP1 purified from A549 cells. Nuclei were stained with DAPI. Scale bar = 10 $\mu m$. Bar graphs in all three panels summarize the quantification of the fold change in ephrin-A3 optical density (mean values ± SD) (n = 2 experiments, each including triplicate coverslips). The statistical analysis was performed by Student’s t-test and asterisks represent p-value: ***p < 0.001; **p < 0.01

We further designed co-culture experiments to confirm this paracrine regulation of EC gene expression, in NRP1 dependent manner. To this end, A549 cancer cells were grown in the upper chamber of a transwell insert, separated by a semipermeable membrane from HUVECs grown on coverslips in the bottom chamber. The results shown in Suppl. Figure 9 demonstrate that cancer cells release paracrine factors downregulating the miR-210 target ephrin-A3 in ECs, in NRP1-dependent manner.

Exosome-borne miR-210 is responsible for NRP1-dependent endothelial cell regulation

As discussed above, our working hypothesis postulated that miR-210 transfer is responsible for EC regulation by exosomes overexpressing NRP1. To verify this, we aimed to interfere with exosome-borne miR-210 by using antagonistic miRNA oligonucleotides (antagomiR-210). We first showed that the transfection of antagomiR-210 strongly reduced miR-210 levels into exosomes released by A549 cells (Fig. 7A). This is consistent with the idea that the intracellular coupling of the miRNA with the antagonistic sequence leads to its degradation or anyway prevents its loading into EVs. We then found that exosomes released from antagomiR-210-treated tumor cells have lost their capacity to decrease ephrin-A3 levels in ECs (Fig. 7B), indicating that miR-210 transfer by exosomes is the driving mechanism of NRP1-dependent paracrine EC regulation by tumor cells.

Fig. 7

Blockade of exosome-borne miR-210 affects endothelial cell migration and gene expression. A Representative qPCR experiment showing the effect of antagomiR-210 on miR210-3p expression levels after its transfection in A549 cells. B Representative confocal images of Ephrin-A3 immunostaining (green) in HUVECs treated with purified exosomes released from A549 cells transfected (or not) with antago-miR-210; nuclei were stained with DAPI. Scale bar = 15 $\mu m$. Bar graphs show the quantification of the fold change of Ephrin-A3 optical density upon exosome treatment (n = 3 independent experiments). Statistical analysis was performed by Student’s t-test and asterisks represent p-value: **** p < 0.0001. C The graph shows a digital real-time analysis of HUVEC migration in the presence of 10 µg purified exosomes released from A549 cells transfected (or not) with antago-miR-210 (same as in panel B), assessed by the xCELLigence system. The timelines (based on continuous measurements) represent the average values from multiple separate wells (n = 5 control, n = 6 antagomiR-treated) analyzed over two independent experiments; error bars are shown at periodic time points. Statistical analysis was done by two-way ANOVA with Bonferroni correction; p-values: ****p < 0.0001; **p < 0.01

We finally assayed the impact of antagomir-210 induced knock-down of exosome-borne miR-210 in EC migration experiments. Thus, we applied the xCELLigence system to digitally measure HUVEC migration in the presence of antagomiR-treated cancer cell-derived exosomes vs. the respective controls. These results, shown in Fig. 7C, clearly indicate that inactivating miR-210 dramatically blunts the pro-migratory activity mediated by exosomes.

Discussion

Research in the last decades has shown that tumor features and malignant behaviour are not exclusively defined by tumor cells [51]. A crucial element contributing to cancer progression and metastasis is the so-called ‘tumor microenvironment’ (TME) [52] with its dynamic network of cell–cell communication pathways [3]. EVs secreted by cancer cells have been well documented to modulate the TME. Indeed, they are highly specialized vehicles of communication carrying several surface proteins and signaling molecules, including oncogene products and non-coding small RNAs, that can be horizontally transferred to target cells in the TME [52]. In particular, exosomes are well recognized as major regulators of pathological angiogenesis, sustaining tumor growth, through the enhancement of signaling pathways in targeted ECs [22, 52]. However, the underlying molecular mechanisms have not been fully elucidated, which calls for a better understanding of regulatory molecules carried by exosomes.

Neuropilins, originally described as VEGF coreceptors in ECs, are now recognized as multifunctional proteins of paramount importance, expressed in multiple cell populations in the TME [12, 53]. Notably, high NRP1 expression in cancer cells has been associated with enhanced tumour angiogenesis [47], although the mechanisms involved are poorly understood. Moreover, NRP1-induced paracrine signals released in the TME to regulate ECs have been postulated but remain elusive. NRP1 itself can be found in soluble form in the extracellular fluids, but its functional significance remains unclear [19, 48]. Two previous studies reported the presence of NRP1 in EVs released by normal cells, including Treg lymphocytes [54, 55]. In the present study, we unveiled for the first time that different cancer cells consistently release NRP1 in the extracellular environment in association with small EVs, which have been bona fide identified as exosomes, rather than as soluble free protein.

Consistent with previous studies, we confirmed that tumor-derived EVs carry a functional activity inducing EC migration. Interestingly, here we show that this motility was actively promoted by exosomes rich in NRP1, whereas vesicles with low-NRP1 were much less effective. It is widely recognized that EV cargoes are heterogeneous and usually reflect the content of the cells of origin, hence their definition as “cellular biopsies”. Therefore, to elucidate the mechanistic relevance of NRP1 in this paracrine signaling mechanism mediated by tumor-derived exosomes, we applied gene knock-down and overexpression approaches, which demonstrated that the exosome activity inducing EC migration is dependent on NRP1 levels.

In addition to NRP1, the neuropilin family comprises the homologous member NRP2, which is also implicated in tumor progression [13]. The two proteins share a similar structure, but have partly different tissue distribution and signaling mechanisms. Notably, we found that NRP2 too can be transported by exosomes; however, NRP2-loaded vesicles poorly induced ECs migration, indicating a specific role for NRP1 in controlling this paracrine signaling mechanism.

We next asked about the NRP1-dependent mechanism responsible for EC regulation. Previous studies have suggested that surface-exposed NRP1 binding in trans to ECs can trigger VEGFR2 phosphorylation and cell migration [19]. However, in our setting we found no evidence that the treatment with NRP1-loaded exosomes directly activates VEGFR2 in HUVECs at significant levels. Notably, tumor-derived exosomes have been found to regulate tumor angiogenesis by transferring their cargo of non-coding RNAs [56]. Indeed, miRNAs can be selectively packed into EVs and hence delivered to recipient cells, where they can hinder the expression of target genes and affect cell behavior [29, 30]. Exosomal miRNAs are therefore considered as a novel class of signaling molecules for cell-to-cell communication, which have been found to profoundly impact ECs and angiogenesis [31, 32]. Based on our findings, we postulated that NRP1 expression might mediate the exosomal loading of regulatory miRNA molecules that promote EC migration. To address this hypothesis, we applied next generation sequencing to determine the miRNome expression profile in three independent preparations of NRP1-loaded exosomes, compared to NRP1-low control vesicles isolated in parallel. Among the top differentially regulated miRNAs, miR-210-3p, which is well known to regulate ECs and promote tumor angiogenesis [33,34,35,36, 38], was significantly increased in NRP1-loaded exosomes. Further analysis by qPCR confirmed a NRP1-dependent regulation of miR-210-3p content in exosomes, which -intriguingly- was not accountable to changes in the intracellular levels of miR-210-3p in producer cells; this rather suggested a NRP1-dependent regulation of exosomal miRNA cargo loading, paving the ground for future investigation into the underlying molecular mechanisms. Of note, our experimental evidence seems to rule out the involvement of the well-known miRNA-interacting protein Ago2 in controlling miR-210-3p loading in exosomes.

Importantly, by using miRNAscope breakthrough technology coupled with immunofluorescence, we confirmed the efficient transfer of miR-210-3p carried by NRP1-loaded exosomes into ECs, and its regulatory impact on the transcript of the EC-regulatory cue ephrin-A3, known to be targeted by this miRNA [36,37,38,39]. Ephrins are signaling molecules exposed on the cell surface, which are widely implicated in cell–cell communication within the TME, and in EC regulation [40]. Ephrin-A3, in particular, has been shown to restrain EC function and angiogenesis, unless post-transcriptionally targeted by miR-210-3p [36, 57]. This finding established a functional correlation between NRP1 expression in tumor cells and the paracrine regulation of ephrin-A3 levels in ECs, mediated by diffusible exosomes. Intriguingly, and in line with what was reported above for miR-210-3p, NRP1 expression did not affect ephrin-A3 levels in a cell-autonomous manner, but rather through paracrine exosome-borne signals acting in the microenvironment. This also seems to exclude in our case another reported mechanism regulating miRNA loading in exosomes, by the abundance of cytosolic targeted transcripts [58]. The mechanistic role of miR-210-3p in this novel NRP1-dependent paracrine regulation of ECs was demonstrated by experiments applying antagomiR-210. In fact, exosomes released by antagomiR-210-treated cells carry reduced levels of miR-210-3p and have lost their capacity to downregulate ephrin-A3 protein levels in ECs as well as promoting EC migration.

The continuous supply of oxygen and nutrients is a key factor in the progression of solid tumors [3, 4]. Thus, hypoxic conditions activate hypoxia-inducible transcription factors (HIFs), which in turn stimulate the expression of angiogenic cues, promoting the formation of new blood vessels [5, 6]. Intriguingly, miR-210-3p expression and paracrine signaling through EVs is known to be induced in hypoxic cells, and this mechanism has been reported to enhance tumor angiogenesis and metastatic progression [39, 59, 60]. In our study, we demonstrate a novel hypoxia-independent and NRP1-dependent mechanism enhancing miR-210 loading in exosomes, which could act independently or synergistically with hypoxic pathways in the tumor context.

This study unveils a novel mechanism of NRP1-dependent signalling impacting the TME. Although we focused on the paracrine regulation through miR-210-3p, our data indicate that the exosomal load of additional miRNAs is controlled by NRP1. Thus, NRP1-regulated exosomes may be involved to affect further cell populations in the TME. Interestingly, the multifaceted tumor-promoter function of NRP1 has prompted the discovery of a variety of blocking molecules, which have been validated to inhibit cancer progression in preclinical murine models [43, 61, 62]. Although it is hard to discriminate the impact of NRP1 blockade on the surface of exosomes, our unpublished preliminary results indicate that the incubation with a NRP1-interfering drug could specifically blunt the capacity of high-NRP1 exosomes to induce HUVEC migration (not shown). Thus, future drugs capable of eliciting NRP1 downregulation or degradation not only could inhibit cancer cell viability, but also interfere with exosome-driven communication in the TME. Besides, further studies will tell if the detection of NRP1-loaded exosomes in liquid biopsies may feature a significant clinical biomarker of tumor progression.

Intriguingly, NRP1 expression in other cell types, such as neurons, leucocytes, lining epithelia, etc. plays key regulatory roles in a variety of pathological processes [63]. These functions have been usually ascribed to NRP1 intracellular signaling pathways, in association with plexins, tyrosine kinase receptors, L1-CAM, integrins, or others [10, 13]. Our findings indicate a novel extracellular, non-cell autonomous, signaling mode of NRP1, regulating miRNA loading in diffusible exosomes, which could pave the way for relevant discoveries in other fields of investigation.

Conclusion

In conclusion, we have identified a novel NRP1-dependent paracrine signaling mechanism, mediated by miR-210-3p loading in exosomes. Our data provide evidence for a previously unknown function of NRP1, strengthening its relevance as tumor promoter in the TME in association with signaling EVs.