VEGF-A/NRP1 stimulates GIPC1 and Syx complex formation to promote RhoA activation and proliferation in skin cancer cells

ABSTRACT Neuropilin-1 (NRP1) has been identified as a VEGF-A receptor. DJM-1, a human skin cancer cell line, expresses endogenous VEGF-A and NRP1. In the present study, the RNA interference of VEGF-A or NRP1 suppressed DJM-1 cell proliferation. Furthermore, the overexpression of the NRP1 wild type restored shNRP1-treated DJM-1 cell proliferation, whereas NRP1 cytoplasmic deletion mutants did not. A co-immunoprecipitation analysis revealed that VEGF-A induced interactions between NRP1 and GIPC1, a scaffold protein, and complex formation between GIPC1 and Syx, a RhoGEF. The knockdown of GIPC1 or Syx reduced active RhoA and DJM-1 cell proliferation without affecting the MAPK or Akt pathway. C3 exoenzyme or Y27632 inhibited the VEGF-A-induced proliferation of DJM-1 cells. Conversely, the overexpression of the constitutively active form of RhoA restored the proliferation of siVEGF-A-treated DJM-1 cells. Furthermore, the inhibition of VEGF-A/NRP1 signaling upregulated p27, a CDK inhibitor. A cell-penetrating oligopeptide that targeted GIPC1/Syx complex formation inhibited the VEGF-A-induced activation of RhoA and suppressed DJM-1 cell proliferation. In conclusion, this new signaling pathway of VEGF-A/NRP1 induced cancer cell proliferation by forming a GIPC1/Syx complex that activated RhoA to degrade the p27 protein.


INTRODUCTION
Malignant tumors express vascular endothelial growth factor A (VEGF-A), a glycoprotein that recruits blood vessels, thereby supplying tumors with the oxygen and nutrients that promote tumor cell migration, proliferation, survival, permeability, and metastasis (Potente et al., 2011). VEGF-A signaling involves via two tyrosine kinase receptors, VEGFR1 and VEGFR2. A previous study demonstrated that the blockade of VEGF-A by Avastin, an antibody, or blockade of VEGFR2 with a specific kinase inhibitor, such as Sutent, suppressed tumor angiogenesis (Potente et al., 2011).
Avastin, in combination with chemotherapy, has exhibited some efficacy in clinical trials for metastatic colorectal cancer, non-small cell lung cancer, renal cell carcinoma, and metastatic breast cancer (Herbst and Sandler, 2008;Mountzios et al., 2014). However, its impact on overall survival is not well documented.
Neuropilin-1 (NRP1) is a 130 kDa transmembrane protein that has been identified as a novel VEGF-A receptor (Soker et al., 1998). NRP1 is expressed by endothelial cells and functions as a coreceptor of VEGFR2, enhancing VEGF-A binding to its receptor and promoting downstream signaling, e.g. the MAPK pathway (Kawamura et al., 2008). NRP1 is associated with tumor progression; it is strongly expressed in lung, brain, colon, ovarian, and prostate cancers with poor patient prognoses (Geretti et al., 2008). A Phase III study to evaluate the combined effects of Avastin and chemotherapy in patients with advanced gastric cancer reported that overall survival was worse in patient groups that strongly expressed tumor NRP1 than in patients with low baseline expression levels (Van Cutsem et al., 2012), suggesting that NRP1 is tumorigenic.
Structurally, NRP1 has two extracellular domains, a1a2 and b1b2, that bind SEMA3s and VEGF, respectively, in addition to a dimerization domain, transmembrane domain, and short cytoplasmic region (Geretti et al., 2008). Since NRP1 lacks kinase activity, there has been a concerted effort to elucidate the mechanisms underlying NRP1 signaling. NRP1 possesses a short cytoplasmic region of 44 amino acids that is involved in signaling. To date, the expression of NRP1 by tumor cells has been shown to contribute to proliferative signal transduction from VEGF-A. In renal cell carcinoma, the VEGF-A/NRP1 signal was found to activate Ras and promote tumor growth in vivo (Cao et al., 2012), while VEGF-A/NRP1 signals induced the phosphorylation of Akt leading to breast cancer cell survival (Bachelder et al., 2001). However, the precise mechanisms responsible for molecular interactions with the NRP1 cytoplasmic region remain unknown.
NRP1 lacking the C-terminus three amino acids [Ser-Gln-Ala (ΔSEA)] led to impaired vasculogenesis in zebrafish (Wang et al., 2006) and abnormal vascular remodeling during retinal development in mice (Fantin et al., 2011). A previous study showed that NRP1ΔSEA did not induce medulloblastoma tumorigenesis (Snuderl et al., 2013). NRP1 appears to signal via the SEA region. GIPC1 (GAIP interacting protein C terminus), a scaffold protein, is the first molecule that was shown to interact with the NRP1 cytoplasmic region (Cai and Reed, 1999;Wang et al., 2010). It has a PDZ domain that binds to the SEA of NRP1 (Ballmer-Hofer et al., 2011;De Vries et al., 1998). GIPC1 is overexpressed in breast and pancreatic tumors and promotes tumor proliferation, survival, and metastasis (Chittenden et al., 2010;Muders et al., 2009;Wu et al., 2010); however, its functions have yet to be determined in detail (Muders, 2011). Syx was identified as a GIPC1 binding protein by a yeast two-hybrid system (Gao et al., 2000;Garnaas et al., 2008).
Syx was found to bind to the GIPC1 PDZ domain via its C-terminus amino acids (Liu and Horowitz, 2006). It has a RhoGEF domain and activates a Rho family GTPase, specifically, RhoA. Previous studies demonstrated that Syx was expressed in vascular endothelial cells, neuronal cells, and some tumors, such as glioma cells (De Toledo et al., 2001;Liu and Horowitz, 2006;Nessling et al., 2005). RhoA drives the cell cycle into the S-phase (Croucher et al., 2010). RhoA has been implicated in virtually all stages of cancer progression. It may play a role during tumor cell proliferation and survival; for example, in vitro, constitutively active RhoA was shown to stimulate transformation (Vega and Ridley, 2008). The activation of RhoA is known to induce the protein degradation of p27 kip1 , a cyclindependent kinase inhibitor (CDI), in the G1 phase, which progresses the cell cycle, resulting in proliferation (Hu et al., 1999;Mammoto et al., 2004).
In the present study, we showed that VEGF-A promoted tumor cell proliferation via the NRP1 signaling pathway. The NRP1 cytoplasmic region was found to be essential for the transduction of VEGF-A signaling, which enhanced the interaction with GIPC1. GIPC1 subsequently formed a complex with Syx. This complex formation activated the RhoGEF activity of Syx, which led to the activation of RhoA. The downstream signaling of RhoA induced p27 protein degradation, leading to S phase entry of the cell cycle, resulting in cancer cell proliferation. A treatment with a cellpenetrating peptide designed to inhibit interactions between GIPC1 and Syx suppressed the activation of RhoA as well as cancer cell proliferation.
In summary, we proposed a novel signal transduction pathway of VEGF-A/NRP1 that induced cancer cell proliferation by forming a GIPC1/Syx complex that activated RhoA and degraded p27.

Knockdown of endogenous VEGF-A expression decreased human skin cancer cell proliferation in vitro
The DJM-1 cell line was established from a human skin cancer obtained from a patient who died from metastases to the axillary lymph nodes and lungs. DJM-1 cells were orthotopically inoculated into the backs of mice. After 2 weeks, mice were sacrificed and the tumors were isolated. Tumor sections were stained with an anti-CD31 antibody (Arrow: bv) and hematoxylin. The tumors and peritumoral area were highly vascularized (Fig. 1A). The amount of VEGF-A secreted into the DJM-1 cell conditioned media (CM) was 8 ng/ml, as measured by ELISA, while that secreted into the siControl-treated DJM-1 cell CM was 7.5 ng/ml (Fig. 1B). VEGF-A was suppressed by the knockdown using 3 different siRNAs, with siVEGF-A #1 being the most effective (siVEGF-A #1: 90% inhibition, siVEGF-A #2: 88% inhibition, siVEGF-A #3: 65% inhibition, respectively) (Fig. 1B). VEGF-A secreted by DJM-1 cells stimulated the migration of HUVEC. The knockdown of VEGF-A expression suppressed the migration of HUVEC (siVEGF-A #1: 38% and siVEGF-A#2: 48% of siControl, respectively) (Fig. 1C). Colony formation in soft agar indicated cancer proliferation under anchorageindependent conditions. The knockdown of VEGF-A expression suppressed the anchorage-independent proliferation (52% of siControl) of DJM-1 cells themselves (Fig. 1D,E). The addition of exogenous VEGF-A (1 µg/ml) restored the proliferation of siVEGF-A-treated DJM-1 cells to a similar level to that of the siControl-treated cells (siVEGF-A#1+1 µg/ml VEGF-A, 92% of siControl). These results suggested that the endogenous expression of VEGF-A stimulated the proliferation of DJM-1 cells in an autocrine manner.
VEGF-A binding to NRP1 induced the interaction between GIPC1 and Syx, thereby promoting DJM-1 proliferation GIPC1 (RGS-GAIP-interacting protein C-terminus) has a PDZ domain that interacts with the NRP1 C-terminal three amino acid residues, SEA (Wang et al., 2006). Syx has been shown to bind to the GIPC1 PDZ domain via its C-terminus (Gao et al., 2000;Garnaas et al., 2008). NRP1, GIPC1, and Syx proteins were overexpressed in HEK293T cells, which did not express VEGFR1 or VEGFR2 (supplementary material Fig. S2). A coimmunoprecipitation analysis (Co-IP) with GIPC1 (HA) showed that NRP1/GIPC1 and Syx/GIPC1 complexes were more prominent in the presence of VEGF-A (+) than in its absence (−) (Fig. 4Da, asterisks). On the other hand, Co-IP with Syx showed that the GIPC1/Syx complex was increased; however, the NRP1/Syx complex was less prominent in the presence of VEGF-A (+) than in its absence (−) (Fig. 4Db, asterisks). These results suggested that VEGF-A/NRP1 induced GIPC1 binding to NRP1 and formation of the GIPC1/Syx complex, which appeared to be released from NRP1. In order to determine whether GIPC1 and Syx mediated the VEGF-A/NRP1 signal in DJM-1 cells, we treated DJM-1 cells with siGIPC1 or siSyx and analyzed proliferation in the presence of exogenous VEGF-A. The siGIPC1 and siSyx treatments both reduced the expression of GIPC1 and Syx (Fig. 4E) and inhibited the proliferation of DJM-1 cells in the absence of exogenous VEGF-A (Fig. 4F, black columns, siControl: 100%, siNRP1: 43%, siGIPC1: 15%, siSyx: 3%, respectively). When exogenous VEGF-A was added, it increased the proliferation of siControl-treated DJM-1 cells (white columns, siControl: 162%).

Syx RhoGEF activity was important for signaling DJM-1 cell proliferation
The MAPK and PI3K pathways are responsible for tumor malignancy and poor patient prognoses. The phosphorylation of MAPK (ERK) and Akt has been shown to contribute to cell proliferation and survival (Owonikoko and Khuri, 2013;Santarpia et al., 2012). However, siVEGF-A and siNRP1 did not significantly change the phosphorylation levels of either MAPK or Akt in DJM-1 cells from those in siControl cells (Fig. 5A). These results suggest that MAPK and Akt were not involved in VEGF-A/NRP1-induced DJM-1 cell proliferation.
RhoA is a regulator of cell proliferation and drives the cell cycle into the S phase (Hu et al., 1999). In siControl-treated RhoGEF is an activator of RhoA. The Syx RhoGEF domain is located from 423 to 612 in its amino acid sequence (Marx et al., 2005) (Fig. 5D). The amino acid residue Leu, located at 571, is important for the binding and activation of RhoA. In order to elucidate whether Syx RhoGEF activity was important for VEGF-A/NRP1-induced DJM-1 cancer cell proliferation, we constructed a lentivirus vector encoding Syx WT or a Syx mutant with a point mutation at the position of 571 Leu replaced to Glu in order to lose binding and the activation of RhoA (Marx et al., 2005). The lentiviruses of Syx WT and the Syx mutant both induced protein expression in DJM-1 cells. The infection amounts among the viruses with the different titers for protein expression were adjusted for equal expression levels in the RhoA activity assay and colony formation assay (Fig. 5E). The lentiviral overexpression of the Syx mutant interfered with the VEGF-A-induced activation of RhoA in DJM-1 cells (Fig. 5F) and inhibited DJM-1 cell proliferation (no addition: 100%, Syx WT: 129%, Syx MT: 45%) (Fig. 5G). These results suggested that Syx, the RhoGEF of RhoA, was an essential and key signaling molecule for mediating the VEGF-A-induced signal transduction that activates RhoA, leading to DJM-1 cell proliferation. RhoA was activated by VEGF-A/NRP1, GIPC1, and Syx to promote cancer cell proliferation In order to determine whether the activation of RhoA promoted DJM-1 cell proliferation, DJM-1 cells were treated with C3 exoenzyme, a specific inhibitor of RhoA. C3 exoenzyme completely suppressed DJM-1 cell proliferation in the absence and presence of exogenous VEGF-A (2% and 1% of siControl, respectively) (Fig. 6A). Y27632, a ROCK inhibitor that is a downstream effector of RhoA, suppressed DJM-1 cell proliferation (no addition: 100%, 10 µM: 51%, 20 µM: 50%, respectively) ( Fig. 6B). Proliferation was recovered (31% to 82%) when RhoA constitutively active form (RhoA CA) was overexpressed in siVEGF-A-treated DJM-1 cells (Fig. 6C,D). p27 was degraded by the activation of RhoA, thereby leading to cell proliferation The Syx/GIPC1 interaction was increased in the presence of VEGF-A (asterisk). On the other hand, the NRP1/Syx interaction was decreased (asterisk). A 10% input as the loading control of NRP1, GIPC1, and Syx co-expressed in HEK293T cell lysates are shown in the right panels. Percentages from each blotted protein amount relative to "VEGF-A (−)" are shown below each lane. (E) Confirmation of the siRNA effects for GIPC1 or Syx. GIPC1 or Syx was overexpressed in HEK293T cells treated with 20 nM siControl, siGIPC1, or siSyx. The inhibitory efficiency of each siRNA on the expression of GIPC1 or Syx was compared to the siControl. The inhibitory percentages relative to the siControl are shown below each lane. (F) Colony formation assay in siNRP1, siGIPC1, or siSyx treated-DJM-1 cells in the presence or absence of exogenous VEGF-A (1 µg/ml) indicated as white columns (+) or black columns (−), respectively. Percentages from each mean relative to the siControl are shown below the graph. These data represent the means±s.d. N.S., not significant; **P<0.005; ***P<0.001.
(S-phase entry). p27 is an inhibitor of G1 cyclin-dependent kinase and regulates cell proliferation downstream of RhoA. Under anchorage-independent conditions, the accumulation of p27 was greater in siVEGF-A-and siNRP1-treated DJM-1 cells than in siControl-treated DJM-1 cells (Fig. 6E). Taken together, these results demonstrated that VEGF-A/NRP1 signaling activated RhoA activity via a GIPC1/Syx complex to inhibit the accumulation of p27.
The oligopeptide that inhibited GIPC1 and Syx interactions suppressed RhoA activation and DJM-1 proliferation We designed a membrane-penetrating peptide targeted to inhibit complex formation between GIPC1 and Syx (Fig. 7A). The 30 kDa targeted peptide consisted of TAT, a cell penetrating sequence of the HIV virus, EGFP, and eight amino acid residues that included the Syx C-terminal amino acid sequence (STLTASEV). The Syx C-terminal amino acid sequence was important for recognizing the GIPC1 PDZ domain in the GIPC1/Syx interaction; therefore, the targeted peptide acted as a competitive inhibitor. The incorporation of these peptides into DJM-1 cells was confirmed through the detection of a green fluorescent protein linked to the peptides after a 1 h treatment (Fig. 7B). In order to establish whether the Targeted peptide interacted with GIPC1, HA-tagged GIPC1 was overexpressed in HEK293T cells and the cell lysate was incubated with either the Scrambled peptide or Targeted peptide. Binding of the Targeted peptide with GIPC1 was 3fold greater than that with the Scrambled peptide (Fig. 7C). In order to evaluate whether the Targeted peptide inhibited the interaction between GIPC1 and Syx, NRP1, GIPC1, and Syx vectors were transfected and expressed in HEK293T cells and these cells were then treated with the Targeted or Scrambled peptide for 16 h. After a 10 min stimulation with (+) or without (−) VEGF-A (100 ng/ml), the cells were lysed and the indicated proteins in the cell lysates were co-immunoprecipitated with V5-tagged Syx (left panels).
VEGF-A/NRP1 induced GIPC1/Syx complex formation in the presence of the Scrambled peptide (Fig. 7D, asterisk). On the other hand, the Targeted peptide abrogated the VEGF-A/NRP1 signalinduced GIPC1/Syx interaction (Fig. 7D, asterisk). In addition, the Targeted peptide more strongly prevented the activation of RhoA than the Scrambled peptide in the absence and presence of VEGF-A (Fig. 7E). Additionally, DJM-1 cell proliferation was inhibited by the Targeted peptide (Scramble peptide: 99%, Targeted peptide: 43%) (Fig. 7F). These results demonstrated that, in the VEGF-A/NRP1 signaling pathway, the GIPC1 and Syx interaction was necessary for the activation of RhoA in order to promote the proliferation of cancer cells. (glioblastoma cell) in an anchorage-independent manner via the NRP1 signaling pathway (Fig. 8). The knockdown of VEGF-A or NRP1 abrogated the proliferation of these cancer cells. We selected skin cancer-derived DJM-1 cells, which only express NRP1 as the VEGF-A receptor and grow faster than other cancer cells under anchorage-independent conditions. The NRP1 structure governs its bioactivity. The A domain mostly binds semaphorins, whereas the B domain binds VEGF-A (Geretti  , 2008). The treatment of cancer cells with the soluble NRP1 B domain or siNRP1 inhibited proliferation in an anchorageindependent manner. Stable shNRP1-DJM-1 clones also decreased proliferation. Taken together, these results indicated that endogenous NRP1 transduced a VEGF-A proliferative signal. PlGF, a member of the VEGF-A family, activated the MAPK pathway via NRP1 in medulloblastoma (Snuderl et al., 2013). NRP2, an isoform of NRP1, was previously shown to transduce the activation of Akt in pancreatic cancer cells (Dallas et al., 2008). However, in the present study, these pathways were not involved in VEGF-A/NRP1 signaling in DJM-1 cells because ERK and Akt phosphorylation levels did not change even when the expression of VEGF-A or NRP1 was decreased by siRNA. The NRP1 C-terminal 3 amino acids (SEA), which contribute to recognition of the PDZ domain of GIPC1, were of particular interest in the present study. The SEA motif was previously shown to be critical for binding GIPC1 (Wang et al., 2006). In the present study, lentiviral infection of the two NRP1 cytoplasmic deletion mutants (NRP1ΔSEA or NRP1ΔCyto) into the shNRP-DJM-1 clone failed to induce anchorage-independent growth in response to VEGF-A. VEGF-A increased the GIPC1 interaction with NRP1, indicating that the NRP1/GIPC1 interaction is necessary for stimulating DJM-1 cell proliferation. GIPC1 has been suggested to play an important role in cancer cell proliferation. GIPC1 was shown to bind to IGF-1R via its PDZ domain in order to promote pancreatic cancer cell proliferation (Muders et al., 2009). By binding VEGF-A to NRP1, GIPC1 mediates the interaction between NRP1 and ABL1, which activates tyrosine kinase activity and associates with integrins, leading to induce tumor growth (Goel and Mercurio, 2013). Syx is a RhoGEF that stimulates RhoA activity (Garnaas et al., 2008). The molecular interaction between GIPC1 and Syx was identified using two-yeast hybridization (Gao et al., 2000). Syx has been implicated in tumorigenesis, brain tumors, and neuroblastoma (De Toledo et al., 2001). siSyx inhibited the proliferation of DJM-1 cells, indicating that Syx is involved in a signaling pathway that promotes cell proliferation.

In the present study, VEGF-
Previous studies reported that NRPs required an interaction with co-receptors such as VEGFR1 or VEGFR2 to induce signals for cell survival and migration (Kawamura et al., 2008). Favier et al. demonstrated that NRPs associated with VEGFR2 in liganddependent and ligand-independent manners in HEK293T cells that transiently overexpressed NRPs and VEGFR2, thereby inducing VEGFR2 phosphorylation (Favier et al., 2006).
In the present study, we showed that the VEGF-A/NRP1 signaling pathway promoted cancer cell proliferation even though cancer cells do not express VEGFR1 or VEGFR2 (Fig. 8). We used HEK293T cells that expressed NRP1, but not VEGFR2 in coimmunoprecipitation to analyze interactions between NRP1/GIPC1 or GIPC1/Syx (supplementary material Fig. S2). As a consequence, VEGF-A induced NRP1/GIPC1 and GIPC1/Syx complex formation without the expression of VEGFR1 or VEGFR2. These results indicated that VEGF-A signals via NRP1 in a VEGFR1-or VEGFR2independent manner, resulting in cancer cell proliferation.
RhoA is a small GTPase that drives the cell cycle into the Sphase with the degradation of p27. p27 protein levels were previously reported to be reduced in many organ sites in the majority of human malignancies (Hu et al., 1999). The activation of RhoA is known to induce the protein degradation of p27 Kip1 . The "dominant negative" effect of the Syx mutant (Syx DN) on RhoA suggests that Syx is involved in the VEGF-A/NRP1 signaling pathway. RhoA has been implicated in virtually all stages of cancer progression. The treatment of DJM-1 cells with the RhoA-specific inhibitor, C3 exoenzyme or ROCK inhibitor (Y27632) suppressed DJM-1 cell proliferation. The knockdown of VEGF-A or NRP1 upregulated the expression of the p27 protein. In addition, the overexpression of constitutively active RhoA in siVEGF-A-treated cells rescued the inhibition of proliferation, indicating that endogenous VEGF-A-binding NRP1 activated Syx RhoGEF in order to stimulate the activation of RhoA, thereby leading to the degradation of p27. Taken together, these results suggested that endogenous VEGF-A/NRP1 signaling in DJM-1 cells induced the degradation of p27 in order to constitutively stimulate progress into the S-phase.
The importance of the molecular mechanism by which the GIPC1 interaction with Syx activates Syx GEF activity was demonstrated in the present study. As a novel tactic, we herein generated a peptide that contained the HIV TAT sequence, which enabled the peptide to penetrate the cell membrane and inhibited complex formation by GIPC1 and Syx (Fig. 8). This peptide consisted of eight amino acids corresponding to the sequence of Syx C-terminus (STLTASEV). The C-terminal five or six amino acids of a binding partner of GIPC1 were shown to lead to sufficient affinity for binding for the PDZ domain (Muders et al., 2009). This study showed that the peptide abrogated the complex formation necessary for Syx RhoGEF activation. Syx is known to include an auto-inhibitory domain in the C-terminal region. A hypothetical mechanism has been proposed by which GIPC1 binding to the auto-inhibitory domain in the C-terminal region of Syx triggers RhoGEF activity.
In conclusion, this study provided evidence for a new pathway of VEGF-A/NRP1 signaling leading to the proliferation of cancer cells. Furthermore, we showed that the molecular function of GIPC1 and its interaction with Syx played a key role in the activation of RhoA, which induced the degradation of p27. The inhibition of VEGF-A/NRP1 signaling may represent a new strategy against cancer and be applied in the design of new cancer drugs.

Animal studies
All animal experiments were performed in accordance with Kyoto Sangyo University's animal experiment guidelines. DJM-1 cells (4×10 6 cells per 100 µl HBSS) were orthotopically inoculated at the right flank of 6-weekold female BALB/C Slc-nu/nu mice (SHIMIZU Laboratory Supplies Co., Ltd., Sakyo-ku Kyoto, Japan). After 2 weeks, mice were sacrificed and tumors were isolated and embedded in OCT compound (SAKURA Tissue Teck, Koto-ku, Tokyo, Japan).

Cell culture and transfection
The human skin cancer line, DJM-1 was kindly provided by Dr H. Katayama (Katayama et al., 1989) (Katayama Clinic, Maebashi, Japan) and cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and glucose (final 4.5 mg/ml). HEK293T cells were purchased from ATCC and cultured in DMEM supplemented with 10% FBS. PC3M cells and U87MG cells were purchased from ATCC and cultured in RPMI-1640 supplemented with 10% FBS for PC3M cells. U87MG cells were cultured in EMEM supplemented with 10% FBS. Human Umbilical Vein Endothelial Cells (HUVEC) were purchased from LONZA (Gampel, Valais, Switzerland) and maintained in endothelial cell growth medium (EGM-2).
The transfection of expression vectors into HEK293T was performed with FuGENE6 (Promega, Madison, WI, USA). siLentFect™ reagents (Bio-Rad, Hercules, CA, USA) were used for all siRNA treatments as directed in the instruction manual.
The secondary antibodies used were: horseradish peroxidase (HRP)conjugated anti rabbit IgG; HRP-conjugated anti goat IgG; a HRPconjugated anti-mouse IgG antibody (Jackson Immuno Research, West Grove, PA, USA). DAPI was purchased from Invitrogen (Life Technologies, Van Allen Way, Carlsbad, CA, USA). A biotin-conjugated rat IgG antibody was purchased from VECTOR (Burlingame, CA, USA).

Plasmids
Human NRP1WT, HA-tagged GIPC1, V5-tagged Syx, HA-tagged constitutively active RhoA, and soluble NRP constructs were inserted using the pcDNA 3.1 TOPO expression vector (Life Technologies). HALOtagged Syx was purchased from the Kazusa DNA Research Institute (KIAA0720) and used as a template to generate the V5-tagged Syx construct. The human NRP1ΔSEA construct was generated by PCR using NRP1WT as a template and primers that introduced a NotI or BamHI restriction site were inserted into the pcDNA 3.1 TOPO expression vector.
PCR products were digested with NotI and BamHI and inserted into the pcDNA 3.1 expression vectors.
NRP1WT, NRP1ΔSEA, and NRP1ΔCyto lentiviruses were based on the NRP1 pcDNA 3.1 construct and generated by PCR using the following primers that introduced NotI and BamHI restriction sites. The same primer for all NRP1 constructs was used as the forward primer and PCR products were subcloned into the pHAGE lentiviral backbone vector as described above .
Syx WT or the DN-V5 tagged lentivirus was based on each pcDNA 3.1 construct and generated by PCR using the following primers that introduced the NotI or BamHI restriction site. PCR products were subcloned into the pHAGE lentiviral backbone vector.

Preparation of Lentivirus vectors of NRP1 WT and mutants
Each NRP1WT, NRP1ΔSEA, and NRP1ΔCyto in the pHAGE lentiviral backbone vector was co-transfected with the helper plasmids (tat, rev, gagpol and VSV-G) to HEK293 cells as described previously . Viral supernatants were assembled and concentrated at 38,000×g for 1.5 h at 4°C. The collected virus was infected with 10 µg/ml polybrene (Millipore) to express NRP1WT and the mutants in DJM-1 cells.

Immunoprecipitation (IP)
HEK293T cells were transfected with NRP1WT, GIPC1, and Syx plasmids with FuGENE6. The cells were stimulated with or without 100 ng/ml VEGF-A for 15 min and were lysed with RIPA buffer (1% NP-40, 0.5% Sodium deoxycholate, 0.1% SDS, NaCl 100 mM, Tris-HCl 50 mM, pH7.4). The cell lysates were incubated with the anti-HA, anti-Syx, or anti-V5-antibody at 4°C overnight. Protein G Sepharose beads (Protein G Sepharose 4 Fast Flow, GE Healthcare) were added and rotated at 4°C for 1.5 h. After washing the beads with cold RIPA buffer, proteins were removed from the beads in 40 μl 2× Laemmlli Sample buffer and analyzed by SDS-PAGE and western blotting. In the input analysis, 1/10 volume of the cell lysate was used.

Western blotting
Cells were lysed with cold RIPA buffer. After running SDS-PAGE, proteins were transferred to a PVDF membrane (Millipore) and blotted with primary antibody-diluted 4% skim milk in TBST at 4°C overnight followed by incubation with a HRP-conjugated secondary antibody. The blots were treated with chemiluminescent substrate solution (Thermo Fisher Scientific, Waltham, MA, USA) and exposed to LAS-4000 mini (Fujifilm Co., Tokyo, Japan) to reveal immunoreactive bands. Percentages from each band on densitometry compared to the control were indicated in the lower lanes in the figures. The western blot analysis was repeated 3 times.

VEGF-A ELISA
A human VEGF Quantikine ® ELISA kit (R&D Systems) was used. DJM-1 cells were seeded at a density of 2×10 5 cells/well/6-well plate, followed by a treatment with 20 nM siRNA. The medium was changed to DMEM containing 1% BSA, and cells were incubated for 3 days at 37°C in 5% CO 2 . The conditioned media were diluted to ten-fold with serum free-DMEM, and VEGF-A levels were measured using the manufacturer's protocol. The measurement of VEGF-A concentrations from each sample was duplicated and the ELISA experiment was repeated twice.

Soft agar assay
DJM-1, PC3M, or U87MG cells were treated with 20 nM siRNA or infected with a lentivirus before being seeded in agar. Two milliliters of growth medium containing 0.72% agar was prepared in a 35-mm dish and solidified as bottom agar. Cells (DJM-1: 5×10 4 cells, PC3M and U87MG: 1×10 5 cells) were suspended in 2 ml of culture medium containing 0.36% agar and, after the addition of ligands or chemicals, layered on the bottom agar. Two weeks later, viable cells were stained with 300 µg/ml 3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) solution. Colony diameters were analyzed by Image J software and the numbers of colonies larger than 80 µm in diameter were counted per 6 to 7 microscopic fields. The means±s.d of colony numbers are shown. Percentages from each mean compared to the control are indicated below the graphs in the figures. Each experiment was repeated at least two or three times.

HUVEC migration assay
A migration assay was performed for HUVEC using Transwell inserts with a pore size of 8.0 μm (Corning, NY, USA). Membranes were coated with 0.1% gelatin. The conditioned medium of DJM-1 cells was prepared with a siRNA treatment, cultured in 2% FBS-EBM-2 for 72 h, and placed into the bottom chamber. Five thousand HUVEC were suspended in 2% FBS-EBM-2 medium, seeded into the upper compartments, and cultured for 16 h. Migrated cells were stained with Diff-Quick. The stained cells in 6 microscopic fields were counted.

RhoA activity assay
A RhoA activity assay was performed and quantified using the RhoA activation assay kit according to the manufacturer's instructions (Cytoskeleton, Denver, CO, USA). Cells were treated with 20 nM siRNA before use and were seeded in 5 mg/ml polyHEMA ( poly 2-hydroxyethyl methacrylate; Sigma)-coated 100-mm dishes for cultivation under anchorage-independent conditions overnight (Fukazawa et al., 1995). Cells were stimulated with 100 ng/ml VEGF-A 165 for 10 min or the indicated time. Cells were lysed with Lysis buffer. The clarified cell lysate was incubated with Rhotekin-RBD protein agarose beads and rotated at 4°C for 90 min. The beads were washed with wash buffer and denatured. Active RhoA was detected by western blotting.
Immunohistochemistry DJM-1 tumors were frozen and sliced to a thickness of 10 µm using Leica CM3050 S (Leica, Wetzlar, Germany). The frozen sections were rehydrated with TBST and fixed with 100% methanol at −20°C for 20 min. In order to inactivate endogenous peroxidase activity, sections were incubated in 0.3% H 2 O 2 solution containing 1% sodium azide overnight at room temperature. These sections were washed with TBST and blocked with 3% BSA/3% house serum and 1% sodium azide in TBST for 40 min, followed by incubation with an anti-mouse CD31 antibody diluted with blocking buffer (1/200) overnight. The sections were then incubated with biotin-conjugated anti-rat IgG for 1 h and developed with an ABC kit (Vector Labs, Burlingame, CA, USA) according to the manufacturer's instructions. Sections were counterstained with hematoxylin (Wako Chemicals Co., Kyoto, Japan). Photographs were taken with NIS-Elements (Nikon, Tokyo, Japan).

Statistical analysis
Differences in means among treatments in the colony formation assay, migration assay and ELISA were evaluated with the Student's t-test: **P<0.005; ***P<0.001; N.S, not significant.