Rhodocetin-αβ–induced Neuropilin-1–cMet Association Triggers Restructuring of Matrix Contacts in Endothelial Cells
Objective—The snake venom component rhodocetin-αβ (RCαβ) stimulates endothelial cell motility in an α2β1 integrin–independent manner. We aimed to elucidate its cellular and molecular mechanisms.
Methods and Results—We identified neuropilin-1 (Nrp1) as a novel target of RCαβ by protein-chemical methods. RCαβ and vascular endothelial growth factor (VEGF)-A avidly bind to Nrp1. Instead of acting as VEGF receptor 2 coreceptor, Nrp1 associates upon RCαβ treatment with cMet. Furthermore, cell-based ELISAs and kinase inhibitor studies showed that RCαβ induces phosphorylation of tyrosines 1234/1235 and thus activation of cMet. Consequently, paxillin is phosphorylated at Y31, which is redistributed from streak-like focal adhesions to spot-like focal contacts at the cell perimeter, along with α2β1 integrin, thereby regulating cell–matrix interactions. Cortactin is abundant in the cell perimeter, where it is involved in the branching of the cortical actin network of lamellipodia, whereas tensile force–bearing actin stress fibers radiating from focal adhesions disappear together with zyxin, a focal adhesion marker, on RCαβ treatment.
Conclusion—Our data demonstrate that (1) Nrp1 is a novel target for venom components, such as RCαβ; (2) Nrp1 coupled to cMet regulates the type of cell–matrix interactions in a manner involving paxillin phosphorylation; and (3) altered cell–matrix interactions determine endothelial cell migration and cellular force management.
Endothelial cells (ECs) are directly attacked by snake venoms, which after injection circulate throughout the victim’s body. Although the plethora of snake venom components and their individual modes of action are not yet fully understood, some of them alter the EC anchorage to the subjacent basement membranes, resulting in endothelial desheathing of the vessel wall.
Integrins are cell adhesion molecules that mediate cell anchorage and force transmission.1 On binding to extracellular matrix proteins, integrins trigger the formation of distinct supramolecular adhesome organelles: streak-like force-transmitting focal adhesions and spot-like focal contacts in nascent lamellipodia.2,3
Integrin-targeting toxins belong to the 2 protein families of disintegrins and C-type lectin-like proteins. The latter specifically recognize the collagen-binding α2β1 integrin on ECs and platelets.4 Rhodocetin (RC) is the prototypic C-type lectin-like protein, which selectively and effectively blocks α2β1 integrin with its γδ subunit.5 After integrin binding, the heterotetrameric RC dissociates and releases its RCαβ subunit,6 which does not interact with α2β1 integrin but inhibits the von Willebrand factor receptor glycoprotein Ib on platelets.7 Although ECs lack any glycoprotein Ib, purified RCαβ increases the motility of ECs using neuropilin-1 (Nrp1) in a novel molecular mechanism, which we disclose in this study.
Snake venoms also contain proteins that are highly homologous to vascular endothelial growth factors (VEGFs), hence assigned as the VEGF-F subfamily.8 Because they differ in their C terminus from other VEGF subfamilies, VEGF-Fs bind the VEGF receptors (VEGFRs), VEGFR1 and VEGFR2, but fail to bind Nrp1.8 Nrp19 is a 923-aa type I transmembrane protein. Its extracellular domain consists of tandem domains a1a2 and b1b2, followed by a homodimerizing c domain.10 The semaphorin 3A–binding a1a2 domain associates with plexins and is crucial for neuronal guidance. Highly relevant for angiogenesis and cardiovascular development, the tandem domain b1b2 captures the C terminus of VEGF-A165,8,11–13 thus complexing Nrp1 with VEGFR2 and allowing full VEGF signaling.14,15 VEGF-A165 or other peptides with a C-terminal arginine, preferentially located within the consensus sequence R/KXXR/KXXR, interact with Nrp1, according to the C-end rule.16
In addition, Nrp1 is involved in hepatocyte growth factor (HGF) signaling in ECs.17 After autophosphorylation, the HGF receptor cMet activates several effector proteins, such as cortactin and paxillin,18,19 and thus elicits migration and other pleiotropic effects.20 Cortactin organizes the cortical actin network via Arp2/3 in lamellipodia formation.19,21 Paxillin, a key component of adhesomes, integrates signals from both integrins and growth factor receptors via several phosphorylation and protein interaction sites.22 On integrin-mediated adhesion, cells initially form spot-like focal contacts, in which paxillin is highly phosphorylated.23 Its dephosphorylation releases paxillin from focal contacts, unless mechanical forces across the cell–matrix contact retain it and additional proteins, such as zyxin, are recruited.24 Thus, focal adhesions are established that serve as docking sites for actin stress fibers bearing tensile forces.3
In this study, we provide strong evidence for an essential role of the complex of Nrp1 with cMet in regulating the structure of integrin-containing cell–matrix contacts, which are of vital importance for the endothelial lining of blood vessels. We demonstrate that RCαβ binds to Nrp1, a hitherto unknown target of snake venom components, thereby inducing the complex formation of Nrp1 with cMet. Consequently, paxillin phosphorylation converts focal adhesions into focal contacts, along with destruction of actin stress fibers and intense formation of a branched cortical actin network in new lamellipodia. This induces EC motility.
Materials and Methods
Additional methods can be found in the online-only Data Supplement.
Human umbilical vein ECs (HUVECs; PromoCell), cultured in 2% FCS/EC medium with penicillin and streptomycin under 5% CO2, were used from passages 2 to 6. Isolation of RCαβ and the corresponding antibodies is described previously.6 Antibodies were purchased against the intracellular domain of Nrp1 (C19, Santa Cruz Biotechnology) and its ectodomain (AF3870), VEGFR2 (MAB3573) and cMet (AF276; the 2 latter antibodies from R&D Systems), paxillin (murine mAb 5H11) and cortactin (both from Abcam), goat and rabbit antibodies against paxillin and pY31-paxillin, respectively (both from Santa Cruz Biotechnology), vinculin (hVin, Sigma), and zyxin (Epitomics). The monoclonal antibodies (mAbs) 16B4 and JA218 against the integrin α2 subunit were gifts from M.J. Humphries and D.S. Tuckwell (both at University of Manchester, England, United Kingdom). The rat mAb 9EG7 recognizing activated β1 integrin was provided by D. Vestweber (Max Planck Institute for Molecular Biomedicine, Münster). Heparin (Sigma), recombinant human VEGF-A165 (termed VEGF-A) and HGF (both from PeproTech), phorbol-12-myristate-13-acetate (PMA; Sigma), the peptide KGRPARPAR (Thermo Scientific), VEGFR2 inhibitor Ki8751 (Calbiochem), and cMet inhibitor SU11274 (Selleckchem) are commercially available.
Affinity Chromatography of Placenta Extract on an RCαβ Resin
RCαβ was coupled to Affigel 15 (BioRad) according to the manufacturer’s instructions. Placental tissue was extracted with a 10-fold volume of extraction buffer (1% reduced Triton X-100, TBS (50 mmol/l TRIS/HCI, pH 7.4, 150 mmol/l NaCI), pH 7.4, 2 mmol/L MgCl2, 1 mmol/L PMSF, 2 µg/mL of each aprotinin, leupeptin, and pepstatin) under stirring at 4°C overnight. After centrifugation and addition of 1 µmol/L CaCl2, the supernatant was loaded onto the RCαβ column. The resin was washed with 0.1% reduced Triton X-100 in TBS, pH 7.4, 150 mmol/L NaCl, 1 mmol/L of each MgCl2 and CaCl2, with an intermittent washing step with 400 mmol/L NaCl in the same buffer. After elution with a pH shift with 0.1 mol/L acetic acid, 150 mmol/L NaCl, and 0.1% reduced Triton X-100, peak fractions were analyzed by SDS-PAGE and silver staining or Western blot analysis. Gel bands were excised for mass spectrometric sequence analysis.
Binding and Inhibition ELISAs With Recombinant Nrp1 b1b2 Tandem Domain
The recombinant Nrp1 b1b2 tandem domain (aa 271–58325; see Methods available in the online-only Data Supplement), termed Nrp1b1b2, was immobilized to microtiter plates at 3 µg/mL in TBS (50 mmol/l TRIS/HCI, pH 7.4, 150 mmol/l NaCI), pH 7.4, 2 mmol/L MgCl2, at 4°C overnight. After washing and blocking with 1% BSA, wells were incubated with VEGF-A or RCαβ. For inhibition assays, constant concentrations of RCαβ (10 and 100 nmol/L for inhibition with C-end rule peptide and mAbs, respectively) and VEGF-A (0.4 nmol/L) were challenged with increasing concentrations of inhibitors. Unbound ligand was washed off the plate, and the bound ligands, VEGF-A or RCαβ, were detected with biotinylated anti–VEGF-A antibodies (diluted 1:3000) or a rabbit antiserum against RC (diluted 1:3000), respectively, followed by alkaline phosphatase–conjugated extravidin and anti-rabbit IgG antibodies (both from Sigma), diluted at 1:2000. Conversion of para-nitrophenyl phosphate was measured at 405 nm.
Impedance-Based Measurement of HUVEC Migration
HUVEC migration was monitored as impedance change in an xCelligence cell invasion and migration manufacturer’s instructions. The filters were coated with 10 µg/mL collagen I in 5 mmol/L acetic acid from their bottom faces overnight at 4°C. After rinsing with water and PBS, pH 7.4, the filter wells were placed onto the bottom compartments containing chemoattractants (RCαβ, VEGF-A, and HGF) in the serum-free HUVEC medium supplemented with 25 mmol/L HEPES, pH 7.4. HUVECs (1×105) were plated into the top compartments, whereas inhibitors were added to both compartments. Impedances were recorded every 5 minutes in the incubator. The RTCA software (version 1.2) calculated cell migration as cell index change versus time.
Immunofluorescence Detection of Nrp1 and cMet Association
To test for association of Nrp1 with potential signaling partners (cMet, VEGFR2), proximity ligation analysis was performed (Duolink II technology of Olink Bioscience). Antibody AF3870 against the Nrp1 ectodomain (R&D Systems) was labeled with the Duolink II Plus-oligonucleotide probe. Antibodies against VEGFR2 (MAB 3573) and against the cMet ectodomain (R&D Systems) were labeled with Duolink II Minus-oligonucleotide probe according to the manufacturer’s instructions. HUVECs (5×104) were seeded onto collagen I–coated (10 µg/mL) and blocked (0.1% BSA, PBS, pH 7.4) permanox chamber slides (Nunc) for 2 hours in a serum-free medium, before medium without and with 200 nmol/L RCαβ, VEGF-A (20 ng/mL), or HGF (200 ng/mL) was added for 30 to 40 minutes. After fixation with 2% formaldehyde in PBS, pH 7.4, and permeabilization with saponin buffer (0.5% saponin, 0.1% BSA, 2% normal horse serum in PBS, pH 7.4) for 1 hour, cells were stained with a combination of Plus-probe– and Minus-probe–labeled antibodies according to the instructions of DuoLink II Probemaker and Detection kit (Olink Bioscience). Cells were counterstained with biotinylated concanavalin A (1.5 µg/mL) and Oregon Green–labeled neutravidin (1:3000) during the ligation and polymerase reaction steps, respectively. Finally, cells were washed and embedded in 50 ng/mL 4′,6-diamidino-2-phenylindole-containing medium (Molecular Probes).
For immunofluorescence analysis, 6×104 HUVECs per chamber were plated onto collagen I–precoated slides for 1 hour. After treatment with or without 200 nmol/L RCαβ or 200 ng/mL HGF for 30 to 40 minutes, cells were fixed with 2% formaldehyde in PBS, pH 7.4, for 10 minutes. SU11274 was added at 10 µmol/L to the medium before and during incubation with RCαβ, where indicated. After blocking and permeabilization with saponin buffer for 1 hour, the specimens were stained with antibodies directed against paxillin (goat, 1:100), pY31-paxillin (rabbit, 1:100), 16B4 (murine mAb, 10 µg/mL), cortactin (rabbit, 10 µg/mL), vinculin (murine mAb hVin, 1:200), and zyxin (rabbit, 1:150), followed by incubation with species-specific antibodies labeled with AlexaFluor488 or AlexaFluor568. Actin was labeled with Alexa568-conjugated phalloidin (1:40; Molecular Probes). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole.
RCαβ Subunit Stimulates the Motility of ECs
In 2 independent assays, RCαβ enhanced the motility of HUVECs. In an open space invasion assay, HUVECs migrated into a pristine collagen-coated area significantly faster in the presence of 200 nmol/L RCαβ (Figure 1A and 1B). In addition, in an impedance-based filter transmigration assay, RCαβ stimulated single-cell migration on type I collagen chemotactically and dose dependently with an ED50 value of 28.0±8.7 nmol/L (Figure 1C). This promigratory effect of RCαβ on HUVECs was not restricted to type I collagen, because it was observed on other matrix ligands of β1 integrins, yet at considerably lower migration rates (Figure 1D). At the molecular level, RCαβ did not interfere directly with α2β1 integrin activity because titration curves of collagen with α2β1 integrin did not differ significantly in the absence and presence of RCαβ (Figure IA in the online-only Data Supplement). Neither at the cellular level did we observe any indirect activation of β1 integrins by RCαβ, as the number of activation-dependent epitope of the β1 integrin–directed mAb 9EG7 was not augmented on HUVECs after treatment with RCαβ, but with phorbol-12-myristate-13-acetate and Mn2+ ions (Figure IB in the online-only Data Supplement).
Nrp1 Is the Target of the RCαβ Subunit
To elucidate the mechanism of these α2β1 integrin–independent cellular effects of RCαβ, the molecular target of RCαβ was identified by affinity chromatography of a placenta extract on resin-immobilized RCαβ. After loading the affinity resin and after stringent washing, a 135-kDa protein was eluted by an acidic pH shift together with RCαβ (16 and 14 kDa; Figure 1E). Mass spectrometric sequencing and immunoblotting with polyclonal antibodies against the Nrp1 ectodomain (Figure 1F) identified the 135-kDa band as Nrp1. Furthermore, the absence of Nrp1 in the flow-through showed that Nrp1 had been quantitatively removed from the placenta extract, indicating an avid binding to immobilized RCαβ.
Nrp1 b1b2 Tandem Domain Binds Both VEGF-A and RCαβ
Nrp1 binds the C terminus of VEGF-A via its b1b2 tandem domain. We recombinantly expressed this tandem domain, shortly termed Nrp1b1b2, with an N-terminal oligo-His-tag to facilitate its isolation by a metal-chelating resin (Figure 1G). Subsequently, Nrp1b1b2 was purified on a heparin column. Two peaks were eluted at low- and high-salt concentrations. In SDS-PAGE (low- and high-salt eluates in Figure 1G), both peaks contained the 38-kDa band of Nrp1b1b2, whereas the high-salt eluate additionally contained few contaminating bands or proteolytic fragments. Both peak fractions showed similar biochemical and immunologic activities. Nrp1b1b2 avidly bound VEGF-A (Figure 1H) with a Kd of 64.5±3.5 pmol/L, which was almost unaffected by heparin (Kd=79.9±2.6 pmol/L).
RCαβ avidly recognized Nrp1b1b2 with a Kd value of 1.8±0.1 nmol/L (Figure 2A). The binding values fitted well for a 1:1 stoichiometric complex (χ2=1.78%). Among the mAbs raised against RC tetramer,6 3 mAbs affected the molecular interaction of RCαβ with Nrp1. Although the mAb IIIB6 slightly increased the affinity of RCαβ to Nrp1b1b2 (Kd=0.48±0.12 nmol/L; Figure 2A), the 2 mAbs, VIIG2 and VD10, inhibited binding with IC50 values of 787 and 204 nmol/L, respectively, albeit only the latter one blocked to completion (Figure 2B). To test whether RCαβ as a novel ligand of Nrp1 also obeys the C-end rule,16 we measured the inhibition of VEGF-A and RCαβ binding to immobilized Nrp1b1b2 with the peptide KGRPARPAR. It entirely inhibited the binding of both Nrp1 ligands, VEGF-A and RCαβ, in a dose-dependent manner with IC50 values of 245 and 12 nmol/L, respectively (Figure 2C), in line with the ≈25 times higher Nrp1b1b2 affinity for VEGF-A compared with RCαβ.
Mutual Interference of RCαβ and VEGF-A in Nrp1 Binding
To elucidate whether the 2 Nrp1 ligands, VEGF-A and RCαβ, affect each other in their binding to Nrp1, VEGF-A binding to Nrp1b1b2 was assessed in the presence of increasing concentrations of RCαβ (Figure 2D). Up to 78 nmol/L RCαβ (vertical dashed line in Figure 2D), the VEGF-A binding signals to immobilized Nrp1b1b2 were unaffected. Above 78 nmol/L RCαβ, the VEGF-A binding signals decreased in a dose-dependent manner with an IC50 of 1425 nmol/L. However, at these concentrations, the Nrp1b1b2 domain was already >89% saturated with RCαβ (Figure 2D, right y axis), making a competitive inhibition unlikely. Consistently, in reciprocal interference assays, VEGF-A did not influence RCαβ binding to Nrp1b1b2 (Figure 2E), although VEGF-A binds to Nrp1b1b2 with an ≈25-fold higher affinity. Therefore, both RCαβ and VEGF-A can interact simultaneously with Nrp1 in a mutually nonexclusive manner.
RCαβ Functions as a Matchmaker of Nrp1 With cMet in HUVECs
Nrp1 is involved in the motility-stimulating effect of RCαβ, because soluble Nrp1b1b2 as decoy receptor (Figure 2F), as well as affinity-purified antibodies against the Nrp1 ectodomain (AF3870) and against Nrp1b1b2 (Figure 2G), were able to inhibit RCαβ-induced migration of HUVECs on type I collagen. Furthermore, siRNA silencing of Nrp1 expression significantly reduced HUVEC migration compared with HUVECs transfected with mock siRNA (Figure 2H).
RCαβ induced complex formation of Nrp1 with cMet, but not with VEGFR2, in HUVECs. In the proximity ligation assay, only at the coincidence, where 2 partners are associated in close proximity, corresponding antibodies conjugated with Plus- and Minus-oligonucleotides form a circular DNA template, which subsequently is detected by red fluorescence labeling. When probed with antibodies against Nrp1 and VEGFR2, HUVECs showed red fluorescent spots, irrespective of the absence or presence of RCαβ, similar to the positive control in the presence of VEGF-A (Figure 3, left panels). In contrast, the combination of antibodies against Nrp1 and cMet gave the most prominent signal with significantly more red spots only if HUVECs were treated with RCαβ compared with an HGF-treated sample or an RCαβ-free control, but also when compared with an Nrp1(+)/VEGFR2(−)-probed RCαβ-treated sample (Figure 3, double arrows in the bottom right panel). Virtually, no red dots occurred when the samples were stained with single proximity ligation assay probes only (Figure II in the online-only Data Supplement). The red spots indicated a close physical interaction of Nrp1 and cMet forming a signaling complex coincidentally only in the presence of RCαβ.
The RCαβ-induced complex formation of Nrp1 with cMet was independently demonstrated by coimmunoprecipitation (Figure 4A). Nrp1 was dragged down together with cMet from lyzed HUVECs that had been treated with RCαβ or HGF.
RCαβ Signals via cMet But Not via VEGFR2 to Stimulate HUVEC Migration
HGF and RCαβ in combination did not enhance HUVEC migration additively, whereas VEGF-A and RCαβ together stimulated HUVEC migration more than the individual factors (Figure IIIA and IIIB in the online-only Data Supplement). In addition, the RCαβ-induced complex formation of Nrp1 with cMet suggested that both RCαβ and HGF signal via cMet into the cell. We also tested binding of HGF to Nrp1 at the molecular level. HGF binds only weakly to recombinant Nrp1b1b2 with an affinity constant of >900 nmol/L. Despite this low affinity, RCαβ failed to inhibit HGF binding to Nrp1b1b2 effectively (Figure IIIC and IIID in the online-only Data Supplement), suggesting independent binding sites for HGF and RCαβ within Nrp1.
To analyze the signaling pathway of RCαβ at the cellular level, we challenged the chemotactic migration of HUVECs along an RCαβ gradient with inhibitors of receptor tyrosine kinases. Although the VEGFR2 inhibitor Ki8751 inhibited HUVEC migration dose-dependently in a VEGF-A gradient (IC50=13.6 nmol/L), it was entirely ineffective in an RCαβ gradient (Figure 4B). In contrast, the cMet inhibitor SU11274 reduced the migration of HUVECs in gradients of HGF and RCαβ with indistinguishable inhibition characteristics (IC50: 2.66 and 1.89 nmol/L, respectively), suggesting that RCαβ in complex with Nrp1 signals via the cMet receptor, like its canonical ligand HGF (Figure 4C). To further prove the hypothesis of an RCαβ-stimulated activation of cMet via Nrp1, we analyzed the effect of RCαβ and HGF on the phosphorylation of the tyrosine residues 1234 and 1235 of cMet with a cell-based ELISA (Figure 4D). With the same time course as HGF, RCαβ elicited phosphorylation of these 2 tyrosines within the activating loop of the cMet kinase domain within a few minutes, thus indicating cMet activation. After the first and major activation peak at 15 minutes, cMet activation again gradually increased after 60 minutes in a second slower wave, which might be because of cMet signaling after endocytosis.
Paxillin Is Phosphorylated as a Consequence of RCαβ-Triggered Nrp1–cMet Complex Formation
Paxillin, a potential target of cMet, was analyzed for its phosphorylation at tyrosine 31 with a sandwich ELISA. It was captured with immobilized mAb 5H11 from lysates of HUVECs treated with RCαβ for different time intervals. Quantification of phosphorylated paxillin (Figure 5A) revealed that paxillin was phosphorylated at Y31 with an initial peak at 30 to 40 minutes, similar to the time course of cMet phosphorylation, and in an SU111274-inhibitable manner (Figure 5A, right panel). HGF caused a similar phosphorylation of paxillin within the first 60 minutes. In a second, much slower wave, phosphorylation of paxillin slowly increased again after 60 minutes. Along with phosphorylation, localization of paxillin within the cell changed remarkably (Figure 5B). Whereas in nonstimulated HUVECs nonphosphorylated paxillin was characteristically found in streak-like focal adhesions underneath the cell soma, it was located after phosphorylation within dot-like focal contacts along the cell perimeter. Evaluation of fluorescence of paxillin and phosphopaxillin along a focal adhesion or the cell perimeter of an unstimulated and an RCαβ-stimulated HUVEC, respectively (left and right bottom panels in Figure 5B), demonstrated that phosphorylation of paxillin in focal adhesions hardly exceeded background values, whereas it was distinctly abundant in focal contacts after 40 minutes of RCαβ treatment, in agreement with previous observations.23 Because dephosporylation of paxillin has been correlated with maturation of focal contacts into focal adhesions, RCαβ reverted this process. Similar to RCαβ, HGF also induced phosphorylation and cellular relocation of paxillin into focal contacts along the cell perimeter (Figure 5A and Figure IV in the online-only Data Supplement). In addition, both processes were inhibited by SU11274, underlining the role of cMet in RCαβ’s mode of action (Figure 5A, right panel, and Figure IV in the online-only Data Supplement).
Using phosphopaxillin as a marker for focal contacts, RCαβ treatment of HUVECs induced redistribution of α2β1 integrin from focal adhesions into the newly formed focal contacts in lamellipodia (Figure 5C).
Restructuring of Cell–Matrix Contacts Is Accompanied by Cytoskeletal Changes
Vinculin, a general marker of cell–matrix contacts, was relocated from large focal adhesions into focal contacts within lamellipodia on RCαβ treatment (Figure 6A, top panels). Cortactin and zyxin are specific marker proteins for lamellipodia and focal adhesions, respectively. RCαβ induced high concentrations of cortactin in numerously formed lamellipodia, indicative of a prominent cortical actin network. In contrast, the focal adhesion marker zyxin, which was clearly localized to focal adhesions before RCαβ treatment, was not detected in any well-defined structures within RCαβ-treated HUVECs. Because focal adhesions are the docking sites for actin stress fibers, their loss was accompanied by the destruction of actin stress fibers that span the cell body and transmit tensile forces (Figure 6A, bottom panels). Statistical evaluations (ANOVA) underlined these observations (Figure IV in the online-only Data Supplement). SU11274 reverted the effects of RCαβ on HUVECS, most prominently the relocation of vinculin and the percentage of cortactin-positive lamellipodia. HGF induced similar effects to RCαβ, additionally suggesting that RCαβ acts via cMet signaling (Figure 6A and Figure IVB in the online-only Data Supplement).
Several growth factors have been reported to interact with Nrp1. RCαβ is the first snake venom component to be described as a ligand for Nrp1. The C-end rule postulates that a C-terminal arginine residue within the consensus sequence R/KXXR/KXXR should mediate the interaction with Nrp1.16 In contrast to VEGF-A165, the C-terminal sequences of fibroblast growth factor 2 and platelet-derived growth factor-BB, although reported as other Nrp1 ligands, do not comply with this C-end rule. Platelet-derived growth factor-BB has been reported to interact with the glycosaminoglycan chain, which is only present in the Nrp1 isoform expressed in vascular smooth muscle cells.10,26 Although the C-end rule is valid for VEGF-A165, RCαβ behaves differently, because the C termini of the α and β subunits end with KNVFMCKFQLPR and KNAFLCKFPKPH, respectively. Nevertheless, the synthetic C-end rule peptide inhibited RCαβ binding to Nrp1b1b2.
mAbs against RC proved the specific interaction of RCαβ with Nrp1b1b2 and, furthermore, delineated the contact of RCαβ with Nrp1. Two antibodies, VD10 and VIIG2, directed against sequence epitopes within the RC β subunit inhibited interaction of RCαβ with Nrp1, indicating that the β subunit is involved in this interaction. In contrast, mAb IIIB6 directed against a conformational epitope within the RCαβ heterodimer slightly activated the interaction. One well-known ligand of Nrp1 is VEGF-A165. Its binding site within Nrp111,27 does not seem to overlap with the one of RCαβ, because the binding of the 2 Nrp1 ligands are not mutually exclusive unless RCαβ was used at saturating concentrations. We cannot rule out that RCαβ induces conformational changes within Nrp1, which may affect VEGF-A165 binding.
An exclusive relationship has been postulated for HGF to be the only cMet ligand,28 as opposed to the rather promiscuous Nrp1. This idea is supported by the unique structure of HGF that does not resemble any other growth factor, but is similar only to plasminogen-like serine proteases. To activate cMet, HGF must be proteolytically cleaved into its 2-chain form, which interacts via different domains with distinct noncontiguous binding sites within cMet.28 Our findings add a new, yet pharmacological, agonist for cMet, which together with internalin B, a bacterial cell-adhesive virulence protein,29 mitigates the postulated exclusive relationship of cMet with HGF as the only endogenous ligand. Despite acting on the same target, HGF, internalin B, and RCαβ do not share any homologous 3-dimensional structure. RCαβ belongs to the versatile family of C-type lectin-like proteins.5,30
Our study introduces C-type lectin-like proteins of venoms from snakes as potential effectors for Nrp1, a hitherto unobserved target in toxinology.4 So far, only galectin, an endogenous C-type lectin expressed by tumor vessel ECs, seems to interact with Nrp1, thereby activating VEGFR2 signaling.31 Functionally, RCαβ acts as an agonist of cMet via Nrp1 binding. Most reports on HGF signaling via cMet do not mention Nrp1 as a coreceptor.28 However, in several tumor cells, expression of Nrp1 does not only correlate with cMet expression32 but also plays a role in cMet signaling.33 To our knowledge, the only report of an interaction of Nrp1 with cMet on ECs17 demonstrates that the N-terminal domain of HGF, but none of the kringle domains, binds to Nrp1 and that Nrp1 thus serves as a coreceptor for cMet signaling in ECs. Corroboratively, we furthermore show a direct interaction of Nrp1 and cMet on HUVECs by proximity ligation analysis and by coimmunoprecipitation, only after addition of RCαβ to the cells. Furthermore, in the absence of RCαβ, Nrp1 preferentially associates with VEGFR2. RCαβ acts as matchmaker to bring Nrp1 in physical contact with cMet. As a consequence, cMet is phosphorylated at tyrosine residues 1234 and 1235 within the kinase activation loop, similar to the effect of the authentic HGF ligand. This mode of action is further corroborated by the inhibitory potential of the cMet kinase inhibitor SU11274. Furthermore, VEGFR2 is not involved in this signaling pathway because the VEGFR2 kinase inhibitor Ki8751 did not prevent cell migration in an RCαβ gradient.
HGF elicits various cellular reactions via its cMet receptor.20,28 In this study, we focused on potential cMet-regulated proteins that play a key role in cell–matrix contacts, using cortactin and paxillin as examples.18,19,21 Cortactin is an essential regulator of the actin cytoskeleton and is regulated by kinases. Interestingly, RCαβ led to cortactin accumulation in the cell periphery. The morphology of RCαβ-treated HUVECs also indicates enhanced formation and persistence of lamellipodia and increased branching of the cortical actin network. Furthermore, activated cortactin could also support formation of focal complexes,34 in line with our observations.
Similarly, we could reveal an RCαβ-induced modulation of paxillin. RCαβ triggered the phosphorylation of paxillin, which led to its relocation from focal adhesions to focal contacts. Dephosphorylation of phosphopaxillin usually occurs during maturation of focal contacts to focal adhesions.23 This process could be reverted by RCαβ-triggered cMet activation. Figure 6B summarizes our model of RCαβ’s mode of action: whereas VEGF-A leads to the formation of an Nrp1–VEGFR2 complex that stabilizes focal adhesions,18 RCαβ binds to Nrp1 and induces its association with cMet. Acting similar to HGF, RCαβ shifts the balance between VEGF-A and HGF in favor of the latter. Thus, paxillin is phosphorylated and relocates to focal contacts along the cell perimeter, whereas the force-transmitting focal adhesions vanish together with the actin stress fibers. This might dramatically affect the EC lining of blood vessels that have to withstand high shear rates. Together with phosphopaxillin, cortactin supports the formation of the cortical actin network and of lamellipodia, helping cell motility. Our study underlines the role of paxillin as cytoskeletal protein that integrates signals from the extracellular matrix via integrins and from various growth factors via their cognate receptors.22 Furthermore, it discloses the molecular mechanism by which the snake venom component RCαβ influences the motility and force management of ECs.
Although the αβ subunit of RC has been the focus of this study, its combination with the γδ subunit within the tetrameric RC may have an additional effect in envenomation. Injected into the blood, tetrameric RC is likely to block the most accessible α2β1 integrin molecules of platelets, thereby releasing RCαβ, which either blocks glycoprotein Ib on platelets or acts on the endothelial lining of blood vessels via Nrp1. The ECs consequently restructure their cytoskeleton and their integrin-containing cell–matrix contacts. Although this does not detach the ECs from the basement membrane, it increases the motility on different matrix proteins. Although RCαβ, like other venom components, may have further effects locally and systemically, such as on neurons or other Nrp1-expressing cells, the explanation of its motility-stimulating effect on ECs at the molecular level provides insights into the pathophysiology of envenomation. ECs are a direct target of RCαβ that increases their motility. This implies that their cell–matrix contacts are repetitively established and released at an increased frequency compared with static adhesion of quiescent ECs. In vivo, such restructuring of cell–matrix contacts may facilitate the access of integrin-blocking venom components to their target molecules. Interestingly, the RC-γδ subunit inactivates α2β1 integrin that has the potential to transmit major mechanical forces.35 Consequently, this may accelerate detachment/desheathing of ECs from the vessel wall. In fact, this is often observed during envenomation and is under vivid investigation in toxinology36 because a desheathed vessel wall becomes an immediate target for snake venom proteinases, which degrade the basement membrane and cause hemorrhage.
The combination of the 2 RC entities, αβ and γδ, provides an evolutionary advantage as tetrameric RC affects integrin-mediated cell adhesion in 2 independent ways: (1) directly by extracellularly abolishing the binding of α2β1 integrin to collagen with its γδ subunit, and (2) indirectly by intracellularly restructuring integrin-containing adhesion structures and affecting the cellular force management by means of its αβ subunit.
We thank A. Schmidt-Hederich for her excellent technical assistance.
Sources of Funding
This study was supported by the Deutsche Forschungsgemeinschaft (DFG) (SFB/TR23, project A8 to J.A. Eble).
The online-only Data Supplement is available with this article at http://atvb.ahajournals.org/lookup/suppl/doi:10.1161/ATVBAHA.112.00006/-/DC1.
- Received July 2, 2012.
- Accepted December 18, 2012.
- © 2013 American Heart Association, Inc.
- Worth DC,
- Parsons M
- Eble JA,
- Niland S,
- Bracht T,
- Mormann M,
- Peter-Katalinic J,
- Pohlentz G,
- Stetefeld J
- Bracht T,
- Figueiredo de Rezende F,
- Stetefeld J,
- Sorokin LM,
- Eble JA
- Yamazaki Y,
- Matsunaga Y,
- Tokunaga Y,
- Obayashi S,
- Saito M,
- Morita T
- Zachary IC
- Pellet-Many C,
- Frankel P,
- Jia H,
- Zachary I
- Parker MW,
- Xu P,
- Li X,
- Vander Kooi CW
- Mamluk R,
- Gechtman Z,
- Kutcher ME,
- Gasiunas N,
- Gallagher J,
- Klagsbrun M
- Koch S,
- Tugues S,
- Li X,
- Gualandi L,
- Claesson-Welsh L
- Teesalu T,
- Sugahara KN,
- Kotamraju VR,
- Ruoslahti E
- Sulpice E,
- Plouët J,
- Bergé M,
- Allanic D,
- Tobelem G,
- Merkulova-Rainon T
- Deakin NO,
- Turner CE
- Zaidel-Bar R,
- Milo R,
- Kam Z,
- Geiger B
- Hirata H,
- Tatsumi H,
- Sokabe M
- Ebbes M,
- Bleymüller WM,
- Cernescu M,
- Nölker R,
- Brutschy B,
- Niemann HH
- Matsushita A,
- Götze T,
- Korc M