A Novel Physiological Glycosaminoglycan-Deficient Splice Variant of Neuropilin-1 Is Anti-Tumorigenic In Vitro and In Vivo

Neuropilin-1 (NRP1) is a transmembrane protein acting as a co-receptor for several growth factors and interacting with other proteins such as integrins and plexins/semaphorins. It is involved in axonal development, angiogenesis and cancer progression. Its primary mRNA is subjected to alternative splicing mechanisms generating different isoforms, some of which lack the transmembrane domain and display antagonist properties to NRP1 full size (FS). NRP1 is further post-translationally modified by the addition of glycosaminoglycans (GAG) side chains through an O-glycosylation site at serine612. Here, we characterized a novel splice variant which has never been investigated, NRP1-Δ7, differing from the NRP1-FS by a deletion of 7 amino acids occurring two residues downstream of the O-glycosylation site. This short sequence contains two aspartic residues critical for efficient glycosylation. As expected, the high molecular weight products appearing as a smear in SDS-PAGE and reflecting the presence of GAG in NRP1-FS were undetectable in the NRP1-Δ7 protein. NRP1-Δ7 mRNA was found expressed at an appreciable level, between 10 and 30% of the total NRP1, by various cells lines and tissues from human and murine origin. To investigate the biological properties of this isoform, we generated prostatic (PC3) and breast (MDA-MB-231) cancer cells able to express recombinant NRP1-FS or NRP1-Δ7 in a doxycycline-inducible manner. Cells with increased expression of NRP1-Δ7 were characterized in vitro by a significant reduction of proliferation, migration and anchorage-independent growth, while NRP1-FS had the expected opposite “pro-tumoral” effects. Upon VEGF-A165 treatment, a lower internalization rate was observed for NRP1-Δ7 than for NRP1-FS. Finally, we showed that NRP1-Δ7 inhibited growth of prostatic tumors and their vascularization in vivo. This report identifies NRP1-Δ7 as a splice variant displaying anti-tumorigenic properties in vitro and in vivo, emphasizing the need to consider this isoform in future studies.


Introduction
The neuropilin-1 (NRP1) gene is located on chromosome 10p12 and composed of 17 exons [1]. It encodes a transmembrane protein involved in axonal development, angiogenesis and cancer progression [2]. It acts as a co-receptor for several growth factors including some variants of VEGF, EGF, HGF, FGF, TGF-β and PDGF [3][4][5][6][7], and it also interacts with other transmembrane proteins such as integrins and plexins/semaphorins [8,9]. NRP1 can be considered as a proteoglycan since it can be modified by the addition of glycosaminoglycan side chains [10][11][12]. Serine 612 has been identified as the site of O-glycosylation where long chains of oligosaccharides assemble on the protein. The nature of the chains attached on NRP1 depends on the cell type. In smooth muscle cells, NRP1 is predominantly glycosylated with chondroitin sulfate while it is derivatized with heparan sulfate in endothelial cells [12]. The number and the nature of the glycosaminoglycans linked to a protein strongly modulate its function. This has also been observed for NRP1. Mutation of Ser 612 into Ala prevents O-linked glycosylation and markedly affects the biological functions of NRP1 [10,13]. The suppression of the glycosaminoglycan side chains by enzymatic digestion similarly results in altered NRP1-driven cell invasion [10].
The full size NRP1 (NRP1-FS, 923 aa) is known as a pro-angiogenic and pro-tumorigenic factor [13,14]. Alternative splicing mechanisms occurring on its primary transcript generate different protein isoforms with sizes varying from 551 to 907 amino acids [1] (Fig 1). Four variants (NRP1-s11, s12, sIII and sIV) lacking the transmembrane domain are soluble and able to bind VEGF-A 165 but exert anti-angiogenic and anti-tumorigenic effects by sequestration of NRP1 binding growth factors and competition with their signaling pathways [15,16]. NRP1-ΔE16, resulting from the skipping of exon 16 and its replacement by the "AAG" Arg triplet, does not display any functional difference as compared to the NRP1-FS [17].
In the present work, we have identified and studied a so far uncharacterized splice variant of NRP1. Although this variant is quite abundant (10 to 30% of the total amount of NRP1), it was never specifically reported before, most probably because it differs from the NRP1-FS variant by the absence of only 7 amino acids, which prevents its identification on SDS-PAGE. This deletion is located two residues downstream of the O-glycosylation site on serine 612 and contains two aspartic residues that are critical for efficient glycosylation [18]. We have therefore determined if the modification of the consensus sequence around Ser 612 affected the glycosylation status and the functions of this new NRP1 variant.

Deep Sequencing
MCF7 were isolated and prepared for next-generation sequencing analysis, as described in a previous publication [20]. The average number of reads approached or exceeded 20 millions.
Alignment of transcripts to the genome indicated that 16733 genes were expressed in the sample. Detailed protocol and sequencing data are available in the ArrayExpress database (www. ebi.ac.uk/arrayexpress) under accession number E-MTAB-5104.

Immunoprecipitation
Cells were chilled on ice and lysed in buffer containing 1% Igepal, 150 mM NaCl, 20 mM Tris/ HCl pH 7.4, 0.1mM AEBSF and 4 μg/ml aprotinin. Lysates were centrifuged for 10 min at 16000g at 4°C. Supernatants were incubated with 1 μg/ml rabbit anti-NRP1 or anti-FLAG antibodies. After 2h of incubation at 4°C, 30 μl of protein A-Sepharose beads (GE Healthcare, UK) were added for 1h to capture immune complexes. Beads were washed four times in ice-cold lysis buffer and boiled in 60 μl of SDS-PAGE lysis buffer. The samples were subjected to western blotting as described above.

Scratch Assay
A suspension of 50 000 PC3 or 35 000 MDA-MB-231, each inducible for NRP1-FS or -Δ7, was seeded into culture-inserts (#80209, Ibidi, Germany) with or without 200 ng/ml doxycycline in the medium. This type of insert allows the creation of a homogeneous 500 μm width cell-free space in the middle of a confluent cell monolayer. Twenty-four hours after seeding, the inserts were removed and pictures by phase-contrast microscopy of the resulting scratches were taken at T0 (2 pictures/insert, 6 inserts/condition and experiment for each tested clone was performed at least twice). PC3 and MDA-MB-231 were cultured for a further 24h and 8h, respectively. Cell migration was evaluated by measuring the cell-free surface at the beginning of the experiment (T0) and the surface covered by the cells at the end of the experiment by using ImageJ Software (National Institute of Health, USA). Results are expressed as the percentage of wound closure in each condition.

Proliferation Assay
Cells were seeded in 24 wells costar plates (7500 cells/well) and allowed to attach overnight (PC3) or for 2h (MDA-MB-231). Half of the cultures were supplemented with doxycycline (day 0) and cells were collected after 1, 3 and 5 days (PC3) or after 1, 2 and 3 days (MDA-MB-231). DNA content was assayed by a fluorimetric technique using bisbenzimide and a microplate Gemini EM spectrofluorimeter (Molecular Devices, USA).

Cell Apoptosis
PC3/NRP1-FS and PC3/NRP1-Δ7 cells were cultured in absence and presence of doxycycline and apoptosis was evaluated by FACS after annexin V-FITC and propidium iodide staining, as previously described [19] using a FACSCanto II double LASER flow cytometer (UV, 488 nm, 633 nm) (BD Biosciences, Belgium). Data were analyzed using FACSDiva Software (BD Biosciences).

Soft Agar Assay
Anchorage-independent growth was determined by soft agar assay [21]. A total of 2 × 10 3 PC3 cells or of 5 × 10 3 MDA-MB-231 cells overexpressing NRP1-FS or NRP1-Δ7 in a doxycyclinedependent way were plated in 60-mm dishes in growth medium containing 0.3% agar, on top of a 0.6% agar gel. The overexpression of the transgene was induced by addition of 1 μg/ml doxycycline. After 21 days, colonies were counted in the whole dishes using an inverted phasecontrast microscope. Results are expressed as the mean number ± SD of two independent experiments performed in quadruplicate.

VEGF Treatment of PC3/FLAG-NRP1-FS and PC3/FLAG-NRP1-Δ7
Cells were serum starved and treated with doxycycline during 24h. Recombinant VEGF-A (isoform 165 at 37 ng/ml) [4] was added in some cultures for 10 minutes. After fixation with 4% paraformaldehyde in PBS for 15 min and permeabilization with 0.25% Triton X-100 in PBS for 10 min, the samples were blocked with 10% bovine serum albumin in PBS for 30 min before incubating for 2h with a monoclonal mouse anti-FLAG (1:500 in PBS). After three washings, the samples were incubated for 1h with a goat anti-mouse Alexa Fluor 555-conjugated antibody (1:500 in PBS) and counterstained with DAPI (2.5 μg/ml). Fluorescent labeling was assessed by using an epifluorescent Nikon TiS microscope. For each cell, a Z stack was performed by using NIS Elements software. By using ImageJ software, the fluorescence was measured in the whole cell (membrane-associated and intracellular vesicles) and in a zone free of vesicles. The maximum pixels value in this vesicle-free zone was taken as the threshold beyond which fluorescence was considered to be associated with intracellular vesicles. The pixels values above the threshold measured in the whole cell were summed and expressed in % of the total fluorescence.

Tumor Xenograft
Experimental protocol was approved by the local ethical committee (University of Liège, approval document n°1293). Nude mice (6 week-old, Charles River Laboratories, USA) received an intraperitoneal injection of 60 mg/kg Nembutal before to be subcutaneously injected in each flank with 2 x 10 6 PC3 conditionally expressing NRP1-Δ7. Control mice had access to regular water while doxycycline (2 mg/ml) was added in drinking water of induced mice. The general condition of mice was monitored daily by visual examination during the entire experiment and tumor volume was evaluated ((length x width x thickness) x π/6, as determined using a caliper). Signs of pain were never observed and the calculated tumor volumes remained always largely below 1 cm 3 which is considered as the upper acceptable limit for subcutaneous tumors. After 7 weeks, mice were sacrificed by cervical dislocation followed by decapitation and the collected tumors were photographed, dissected, measured and weighed. Fragments of tumors were then used for immunohistological analyses and RNA extraction.

Immunohistological Analysis of Tumors
Antigen retrieval was performed on 5 μm sections of PC3 tumors by heating at 126°C for 11 minutes in Target Retrieval Citrate buffer (pH 6, Dako, Denmark). After blocking endogenous peroxydases and non-specific antigenic sites (Universal Blocking Reagent, Biogenex, USA), sections were incubated with primary anti-CD31 antibodies (diluted 1:150 in BSA/PBS 1%) for 1h, followed by biotinylated secondary antibodies (diluted 1:300) for 30 minutes and with HRP-conjugated streptavidin (diluted 1:500) for 30 minutes. Signal was amplified by an incubation in Biotinylated Tyramide (diluted 1:100, Perkin Elmer, USA) for 10 minutes, followed by additional 30 minutes in HRP-conjugated streptavidin. Staining was performed with AEC + High Sensitivity Chromogen (Dako, Denmark) followed by a counterstaining with hematoxylin. Stained sections were scanned with a NanoZoomer slide scanner (Hamamatsu Photonics, Japan) and quantification was performed using ImageJ software (National Institute of Health, USA). Tumor vascularization was quantified both as the number of blood vessels per unit surface and as the % of surface covered by blood vessels.

Statistical Analyses
Statistical analyses were performed using U Mann-Whiney test or by an Anova test (GraphPad Prism software, GraphPad Sofware, La Jolla, CA, USA) with p 0.05 considered as significant.

Characterization of a Novel Variant of NRP1: NRP1-Δ7
During an exploratory experiment aiming at determining the expression of the different splice variants of NRP1, several primer pairs were designed and used, including primers at the exon10/exon11 (forward) and exon12/exon13 (reverse) junctions (Fig 1). In addition of the expected 233bp product, a smaller fragment was reproducibly and significantly co-amplified. It was cloned and sequenced, showing that it results from the splicing of intron 11 at an alternative donor site (AG-gtgat) in exon 11 located 21 bases upstream of the usual splice site (AGgtgca). This deletion does not change the reading frame of the mRNA and is responsible, at the protein level, for the absence of a 7 amino acids long sequence explaining the name given to this variant (NRP1-Δ7). The actual existence of this variant of NRP1 was further confirmed by different independent observations. Its presence was first evaluated in various cell lines and tissues of human or mouse origin by regular RT-PCR followed by gel electrophoresis. The relative proportion of NRP1-Δ7 was also determined by RT-qPCR using variant-specific primers. The ratio Δ7/FS varies between 0.13 and 0.43 (Fig 2). As an additional approach, the existence and presence of NRP1-Δ7 was also evaluated in "Deep Sequencing" data previously obtained using MCF7 cells mRNA [20]. From such data, the abundance of alternative splicing variants can be calculated by comparing their FPKM (Fragments Per Kilobase Of Exon Per Million Fragments Mapped). Values of 29.13 and 12.18 were obtained for NRP1-FS and NRP1-Δ7, which correspond to a ratio of 29%. Finally, the existence of NRP1-Δ7 is further assessed by its identification by the German Genome Project (clone DKFZp686E15115; http://www.ncbi.nlm.nih.gov/ nuccore/BX648250.1) by entries in Genbank under the accession numbers NM_001244973 (NRP1 variant 5) and NP_001231902 (NRP1 isoform E). Based on the exon content of the previously described variants of NRP1, NRP1-s12 and NRP1-ΔE16 should be also affected by the 21 bases deletion. It was confirmed for NRP1-s12 (NRP1-s12-Δ7), while it was not further investigated for NRP1-ΔE16 which is barely detectable in our cell models (S1 Fig).

NRP1-Δ7 Displays Reduced O-Linked Glycosylation
Recombinant proteins were produced to identify potential functional consequences of the 7 aa deletion. HEK293 cells were transiently transfected with an expression vector for NRP1-FS or NRP1-Δ7, while PC3 cells were transfected with a vector expressing FLAG-tagged NRP1-FS or NRP1-Δ7 to allow discriminating exogenous and endogenous NRP1. They were immune-precipitated, respectively, with anti-NRP1 and anti-FLAG antibodies and subjected to western blot analysis revealed by anti-NRP1. As illustrated in Fig 3, a substantial fraction of NRP1-FS migrated as high molecular weight products forming a smear corresponding to proteoglycanmodified NRP1 [10,12]. Remarkably, these high molecular weight products were far less abundant in NRP1-Δ7 overexpressing cells, both in HEK and PC3 cells. These results suggest that the 7 aa deletion close to Ser 612 affects its glycosylation because it changes the consensus sequence required for optimal O-glycosylation. These results also show that our cell models are adequate for investigating the biological functions of NRP1-Δ7. To improve the significance of the results of our biological models and avoid potential off-target effects linked to the use of

Generation of Inducible PC3 and MDA-MB-231 Clones Overexpressing NRP1-FS or NRP1-Δ7
PC3 and MDA-MB-231 cancer cells are known to constitutively express a significant level of NRP1 and to respond to NRP1-mediated signaling [22][23][24]. In order to benefit from a robust experimental model, we decided to create inducible clones expressing recombinant NRP1 (either FS or Δ7) only upon stimulation by doxycycline added in the culture medium. The advantage of this model is that each clone can be used as its own control by comparing its phenotype in absence and presence of inducer, an experimental setting much more reliable and biologically relevant than the comparison of different cell lines. We generated doxycyclineinducible PC3 and MDA-MB-231 clones overexpressing NRP1-FS (PC3/TR/NRP1-FS and MDA/TR/NRP1-FS) or NRP1-Δ7 (PC3/TR/NRP1-Δ7 and MDA/TR/NRP1-Δ7). Two clones of each type for PC3 and one clone of each type for MDA-MB-231 were isolated and amplified for further investigations. In control conditions without doxycycline the ratio Δ7/FS represents roughly 0.17 to 0.33 (Fig 4 and S3 Fig). Upon stimulation of expression of recombinant NRP1 Neuropilin-1-Δ7 Is Anti-Tumorigenic by doxycycline, this ratio drops to 0.10 in NRP1-FS clones and increases up to 0.60-0.90 in NRP1-Δ7 clones, as measured by RT-qPCR. Western blotting analysis, which cannot discriminate the two isoforms, indicates a maximum of 2-fold increase in total NRP1 protein expression.

NRP1-FS and NRP1-Δ7 Differentially Regulate Cellular Functions In Vitro
It is known that NRP1 can affect various cellular functions, such as migration or proliferation, which are relevant to cancer progression [25]. The migration of PC3 cells conditionally overexpressing NRP1-FS or NRP1-Δ7 was assessed in a scratch wound healing assay ( Fig 5). As expected, overexpression of NRP1-FS significantly enhanced the migration rate of PC3 cells whereas it was significantly inhibited in cells overexpressing NRP1-Δ7. Similar results were obtained with a second PC3 clone of each type (NRP1-FS or NRP1-Δ7) and confirmed in MDA-MB-231 clones, showing that this regulation is not limited to a single cell type (S5 Fig). The proliferation rate of PC3 and MDA-MB-231 cells was also increased upon overexpression of NRP1-FS, while being reduced in cells overexpressing NRP1-Δ7 (Fig 6). These effects were not related to alterations of apoptotic processes which were not influenced by the overexpression of either variant as shown by flow cytometry analyses (S6 Fig). The consequences of the overexpression of NRP1-FS and NRP1-Δ7 on the in vitro tumorigenic properties of PC3 and MDA-MB-231 cells were further evaluated in an anchorage-independent growth model. Cells were seeded in soft agar in absence and presence of doxycycline and the total number of colonies formed after 3 weeks was counted. The overexpression of NRP1-FS in both cell types significantly increased the number of colonies while overexpression of NRP1-Δ7 had the opposite effect (Fig 7). It has to be noticed that doxycycline by itself did not affect the phenotype of PC3/ TR cells transfected with an empty pcDNA4/TO or of MDA-MB-231/TR in the three experimental models, confirming that the observed effects are specifically related to the transgene overexpression ( S4 Fig). Altogether, these data demonstrate an opposite effect of NRP1-FS and NRP1-Δ7 on the in vitro tumorigenic properties of PC3 and MDA-MB-231 cells.

NRP1-FS and NRP1-Δ7 Are Differentially Relocalized upon VEGF-A 165 Stimulation
NRP1 is a well-known co-receptor for VEGF. PC3 inducible clones expressing the FLAGtagged NRP1-FS or NRP1-Δ7 were cultured in presence of doxycycline and exposed to VEGF-A 165 for 10 min. Immunofluorescence labeling was performed with an anti-FLAG antibody and the fluorescence associated with the membrane and intracellular vesicles was quantified on Z stacked sections as detailed in Materials and Methods. As illustrated in Fig 8, VEGF-A 165 treatment induced a significant relocalization of NRP1-FS in intracellular vesicles with a concomitant reduction of membrane-associated fluorescence. No significant difference in the compartmentalization of the NRP1-Δ7 was observed after VEGF-A 165 treatment.
analyses or lysed for western blotting visualization using an anti-NRP1 antibody. PC3 and MDA-MB-231 cells express endogenous NRP1-FS and NRP1-Δ7, which can be identified as two RT-PCR products in absence of dox. The ΔCt of each isoform were measured by SYBR green RT-qPCR and the relative expression of NRP1-FS and NRP1-Δ7 were calculated as described in Material and Methods. Induced expression of recombinant NRP1 by dox increases the relative proportion of NRP1-FS (A) or NRP1-Δ7 (B) in the corresponding clones. Due to the small difference in size (7 amino acids), NRP1-FS and NRP1-Δ7 cannot be discriminated by western blotting. Increases in total NRP1 protein expression is however observed in presence of dox. 28S rRNA and ERK1/2 were used as control for, respectively, RT-PCR and western blotting analysis. doi:10.1371/journal.pone.0165153.g004 Neuropilin-1-Δ7 Is Anti-Tumorigenic

NRP1-Δ7 Antagonizes Tumor Growth and Vascularization In Vivo
For evaluating the role of NRP1-Δ7 in in vivo tumorigenesis, PC3/TR/NRP1-Δ7 cells were injected in the flank of nude mice. Control mice had access to regular water while the water of NRP1-Δ7 mice was supplemented with doxycycline. After 7 weeks, mice were sacrificed and the tumors weighed (Fig 9A-9C). Expression of NRP1 in tumors was determined by RT-PCR. In control tumors, NRP1-Δ7 represents roughly 20% of the total amount of NRP1, while this proportion rises to 50% upon doxycycline treatment, as also observed in in vitro experiments. A significant inhibition of growth was observed in tumors formed by NRP1-Δ7 overexpressing cells as compared to the tumors formed by the same pool of cells in absence of doxycycline. Tumors vascularization was evaluated by labeling with anti-CD31 antibodies (Fig 9D-9F). A significantly decreased number of intratumoral blood vessels per unit surface as well as a Neuropilin-1-Δ7 Is Anti-Tumorigenic Neuropilin-1-Δ7 Is Anti-Tumorigenic reduction of the surface covered by blood vessels was observed in tumors overexpressing NRP1-Δ7. Neuropilin-1-Δ7 Is Anti-Tumorigenic

Discussion
The neuropilin-1 pre-mRNA is alternatively spliced in several mature mRNA encoding proteins with various, sometimes opposite, properties. The full size NRP1 is anchored in the cell membrane, functions as a co-receptor for multiple ligands such as VEGF-A 165 , TGF-β and semaphorinA, and forms complexes with their signaling receptors. NRP1 is also involved in endocytosis and initiates intracellular trafficking of a number of membrane receptors. The soluble isoforms of NRP1 (NRP1-s11 and -s12) act in several settings as antagonists of neuropilin-1 signaling, mainly by competitive binding to the ligands [16]. Here we report the characterization of a novel variant (NRP1-Δ7) which differs from the full size form by the absence of a 7 amino acids sequence (GDDFQLT) located 2 amino acids downstream of Ser 612 , the main O-glycosylation site on NRP1. We have shown that NRP1-Δ7 is significantly and differently expressed in various cell types of human and mouse origin, varying between 10 and 30% of total NRP1, which suggests a specific physiological function. The ability of cells to add glycosaminoglycan chains strongly depends upon the amino acids sequence around the glycosylation site [26]. For instance, Zhang and colleagues showed that a serine-glycine peptide flanked by acidic residues favored the addition of heparan sulfate [18]. In NRP1-FS, the sequence around Ser 612 (S 612 GTGDDF) contains two "D" acidic residues, while the "GDDFQLT" deletion in NRP1-Δ7 results in a more hydrophobic consensus sequence (S 612 GTGGTT) which is likely responsible for the defective O-glycosylation of NRP1-Δ7 reported here. Alterations of NRP1 glycosylation, induced enzymatically or by mutations of Ser 612, profoundly alter cancer cells phenotype [10,13]. The role of the naturally occurring NRP1-Δ7 was therefore evaluated in functional assays relevant to cancer biology. For this purpose, prostate (PC3) and breast (MDA-MB-231) adenocarcinoma cells were used as models since they express a significant level of endogenous neuropilin-1 and are known to respond to activation of neuropilin-1-dependent pathways [22][23][24]. They were modified to conditionally express either recombinant NRP1-FS or NRP1-Δ7 upon doxycycline treatment. As a consequence, the relative proportion of NRP1-Δ7, which is around 20% in control cells, is decreased to about 10% in doxycycline-induced NRP1-FS while it reaches almost 50% in doxycyclinetreated NRP1-Δ7 clones. It has to be pointed out that, in our inducible model, the total amount of expressed NRP1 is never increased by more than about two fold, which is important to prevent non-relevant effects that could be induced by an excessive overexpression of recombinant proteins.
NRP1-FS is known to regulate cell functions involved in cancer progression such as proliferation, migration or anchorage-independent growth [9,13,27]. However, all the previous studies related to endogenous NRP1 did not consider separately NRP1-FS and NRP1-Δ7. The observed effects could thus reflect the net results of a balance between two NRP1 isoforms with opposite functions. By specifically evaluating the functions of NRP1-Δ7, we demonstrated here that its overexpression to the same level as endogenous NRP1-FS was sufficient to significantly decrease the proliferation rate, the migration and the anchorage-independent growth of prostatic and breast cancer cells, pointing to an anti-tumorigenic potential of NRP1-Δ7. It is well known that NRP1 forms clusters with various growth factor receptors like VEGFR and c-Met, or plexins, and serves as a guide during their intracellular recycling [28]. It was also shown that . The animals were separated in two groups. The expression of the transgene was induced in one group by adding doxycycline (2 mg/ml) in drinking water. After 7 weeks, mice were sacrificed. The tumors were macroscopically observed (A), measured and weighed (B) (n = 15 tumors in control condition and n = 14 tumors in doxycycline-induced mice) and processed for total RNA extraction and measurement of both variants by RT-PCR (C). The relative proportion of NRP1-Δ7 increases from 20 to 50% in presence of dox. (D) Immunological analysis of tumor vascularization by CD31 immunostaining. (E) The number of blood vessels per unit surface was determined and (F) the surface covered by blood vessels was calculated and expressed in %. Results are expressed as the mean ± SD of n = 13 tumors in control condition and n = 10 tumors in doxycyclineinduced mice (*: p<0.05, Mann-Whitney U test)-Scale bar = 100 μm. doi:10.1371/journal.pone.0165153.g009 Neuropilin-1-Δ7 Is Anti-Tumorigenic VEGF-A and TGF-β can markedly affect the phenotype and aggressiveness of PC3 cells [24]. By analyzing the subcellular dynamics of NRP1-FS and NRP1-Δ7, we showed that NRP1-Δ7 is less internalized than NRP1-FS upon VEGF-A binding. These results suggest that the association of NRP1-Δ7 with transmembrane growth factor receptors could affect their intracellular recycling which might inhibit their downstream signaling pathways. Alternatively, NRP1-Δ7 could exert its anti-tumorigenic activity by competing with NRP1-FS for binding to ligands as reported for the soluble forms of NRP1. NRP1-Δ7 could also mediate its anti-tumorigenic effects via cell adhesion receptors like integrins. NRP1 can indeed synergize with these extracellular matrix receptors to promote cancer cell proliferation, migration or colony formation in soft agar in vitro [9,13,27] and to stimulate in vivo tumor growth and vascularization [29]. The potential anti-tumorigenic properties of NRP1-Δ7 were therefore evaluated in vivo by using inducible PC3/TR/NRP1-Δ7 in a xenograft model. As observed in vitro, the addition of doxycycline in drinking water allowed continuous expression of NRP1-Δ7 in the xenografts. Using this model, we clearly demonstrated that increased expression of NRP1-Δ7 inhibits tumor growth and intratumoral angiogenesis.
This report is the first study characterizing the specific functions of NRP1-Δ7, a glycosaminoglycan-defective neuropilin-1 splice variant that displays anti-tumorigenic properties in vitro and in vivo. Although the precise mechanisms of regulation depending on NRP1-Δ7 were not identified, our observations emphasize the critical importance of considering specifically the different NRP1 variants, especially NRP1-Δ7, instead of analyzing the whole NRP1 expression.