Journal list menu

Volume 284, Issue 11 p. 1657-1671
Original Article
Free Access

Haemorrhagic snake venom metalloproteases and human ADAMs cleave LRP5/6, which disrupts cell–cell adhesions in vitro and induces haemorrhage in vivo

Tadahiko Seo

Tadahiko Seo

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Taketo Sakon

Taketo Sakon

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Shiori Nakazawa

Shiori Nakazawa

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Asuka Nishioka

Asuka Nishioka

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Kohei Watanabe

Kohei Watanabe

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Kaori Matsumoto

Kaori Matsumoto

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Mari Akasaka

Mari Akasaka

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Narumi Shioi

Narumi Shioi

Department of Chemistry, Faculty of Science, Fukuoka University, Japan

Search for more papers by this author
Hitoshi Sawada

Hitoshi Sawada

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Search for more papers by this author
Satohiko Araki

Corresponding Author

Satohiko Araki

Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Japan

Correspondence

S. Araki, Sugashima Marine Biological Laboratory, Graduate School of Science, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi 464-8602, Japan

Fax: +81 52 789 4204

Tel: +81 52 789 2514

E-mail: [email protected]

Search for more papers by this author
First published: 20 April 2017
Citations: 17

Abstract

Snake venom metalloproteases (SVMPs) are members of the a disintegrin and metalloprotease (ADAM) family of proteins, as they possess similar domains. SVMPs are known to elicit snake venom-induced haemorrhage; however, the target proteins and cleavage sites are not known. In this work, we identified a target protein of vascular apoptosis-inducing protein 1 (VAP1), an SVMP, relevant to its ability to induce haemorrhage. VAP1 disrupted cell–cell adhesions by relocating VE-cadherin and γ-catenin from the cell–cell junction to the cytosol, without inducing proteolysis of VE-cadherin. The Wnt receptors low-density lipoprotein receptor-related proteins 5 and 6 (LRP5/6) are known to promote catenin relocation, and are rendered constitutively active in Wnt signalling by truncation. Thus, we examined whether VAP1 cleaves LRP5/6 to induce catenin relocation. Indeed, we found that VAP1 cleaved the extracellular region of LRP6 and LRP5. This cleavage removes four inhibitory β-propeller structures, resulting in activation of LRP5/6. Recombinant human ADAM8 and ADAM12 also cleaved LRP6 at the same site. An antibody against a peptide including the LRP6-cleavage site inhibited VAP1-induced VE-cadherin relocation and disruption of cell–cell adhesions in cultured cells, and blocked haemorrhage in mice in vivo. Intriguingly, animals resistant to the effects of haemorrhagic snake venom express variants of LRP5/6 that lack the VAP1-cleavage site, or low-density lipoprotein receptor domain class A domains involved in formation of the constitutively active form. The results validate LRP5/6 as physiological targets of ADAMs. Furthermore, they indicate that SVMP-induced cleavage of LRP5/6 causes disruption of cell–cell adhesion and haemorrhage, potentially opening new avenues for the treatment of snake bites.

Abbreviations

  • ADAM
  • a disintegrin and metalloprotease
  • ECM
  • extracellular matrix
  • EGF
  • epidermal growth factor
  • HUVEC
  • human umbilical vein endothelial cell
  • LDLa
  • low-density lipoprotein receptor class A domain
  • LRP
  • low-density lipoprotein receptor-related protein
  • PVDF
  • poly(vinylidene difluoride)
  • SVMP
  • snake venom metalloprotease
  • VAP
  • vascular apoptosis-inducing protein
  • Introduction

    Snake venom metalloproteases (SVMPs) in haemorrhagic snake venom are known to be responsible for haemorrhage, and they belong to the a disintegrin and metalloprotease (ADAM) family of proteins having similar domains [1, 2]. However, the target proteins and causative cleavage sites are not known. Although various haemorrhagic metalloproteases appear to cleave extracellular matrix (ECM) proteins such as collagen and laminin, the cleavage and haemorrhagic activity are not always correlated [3-5]. Except for the toxins, ADAM-type metalloproteases are mainly involved in shedding cell membrane proteins, while matrix metalloproteases are mainly involved in lysing ECM proteins. These features led us to propose that the target proteins of SVMPs may be cell membrane proteins. Although several membrane proteins such as integrin α2β1 and annexin V have been reported to be potential substrates of SVMPs, the cell membrane protein(s) responsible for haemorrhage elicited by SVMPs remains unknown [6, 7].

    Several mammalian ADAMs including ADAM8, 9, 12, 15, 19, 28 and 33, which are homologues of SVMPs, have been reported to be involved in leukocyte and cancer cell invasion through vascular endothelial cells [8-10]. However, the main substrates of these ADAMs for invasion are also unclear as is the case for SVMPs. The mechanisms of the tissue invasion itself, especially the mechanisms of cell barrier opening, are also unclear. Since both SVMPs and mammalian ADAMs are involved in opening of the vascular barrier, endothelial receptors of haemorrhagic SVMPs appear to be closely related to the target proteins of mammalian ADAMs.

    We have reported that several haemorrhagic SVMPs, including vascular apoptosis-inducing proteins (VAPs), induce characteristic cellular responses in cultured vascular endothelial cells [11, 12]. VAPs induce disruption of cell–cell adhesion in cultured vascular endothelial cells and also cause severe cell fragmentation in apoptotic cells [13, 14]. However, they showed little or no appreciable cleavage of collagens or laminins [15]. In this context, we focused on the cell surface receptor proteins of SVMPs on endothelial cells using VAP1, which induces the same characteristic cellular responses.

    The Wnt signalling pathway is involved in development, differentiation and oncogenic transformation [16, 17]. This pathway appears to be responsible for proliferation signalling and epithelial–mesenchymal transition, by which epithelial cells acquire mesenchymal fibroblast-like properties, resulting in reduction of cell–cell adhesion ability and increase in motility. In canonical Wnt pathway activation, β-catenin becomes unsusceptible to degradation and accumulates in the cytosol, enabling β-catenin-dependent transcriptional activation [18]. On epithelial and endothelial cells with well-developed cell–cell junctions, β-catenin associates with cadherin beneath the cell membranes. However, the role of cadherin-interacting catenins in Wnt signalling had remained elusive. Recently, it has been reported that β-catenin and γ-catenin, a paralogue of β-catenin, dissociate from the membrane-bound cadherin and are recruited into the cytosol by Wnt signalling in epithelial cells with cell–cell junctions [19-21]. It is known that low-density lipoprotein receptor-related proteins 5 and 6 (LRP5/6), co-receptors of Wnt with frizzled (Fzd), exist as a complex with cadherin in the cell membrane and that LRP5/6 are involved in the release of catenin from cadherin during Wnt signalling [22, 23]. It is notable that LRP5/6 mutants lacking extracellular β-propeller domains exhibited constitutive activity of Wnt signalling regardless of the absence of Wnt and Fzd [24, 25]. It has also been reported that truncated LRP6 is capable of inducing β-catenin release from E-cadherin without Wnt signalling [26].

    To elucidate the mechanism of vascular barrier opening induced by haemorrhagic SVMPs and ADAMs, we investigated whether VAP1 affects proteins located in the cell–cell junctions such as cadherin and catenin. We found that LRP5/6 are the promising candidates for the targets of SVMPs and ADAMs that are responsible for disruption of cell–cell adhesions and haemorrhage induced by ADAM toxins.

    Results

    VAP1 induces cadherin and catenin relocation

    We previously reported that a haemorrhagic SVMP, VAP1, induces disruption of cell–cell adhesions in cultured vascular endothelial cells [13]. To elucidate the mechanism of the disruption of cell–cell adhesions, we investigated whether VAP1 affects membrane proteins in the cell–cell junctions. Immunocytochemistry showed that treatment with VAP1 resulted in a notable change in localization of VE-cadherin from the cell membrane region to the cytosol (Fig. 1A). Then, in order to investigate whether VAP1 cleaves VE-cadherin, we compared the amounts of VE-cadherin in cells with and without VAP1 treatment by western blotting. However, no appreciable cleavage was observed (Fig. 1B). The same result was obtained using a recombinant protein of human VE-cadherin, i.e. no apparent degradation by VAP1 was observed (Fig. 1C). Therefore, VE-cadherin was not considered to be a target of VAP1. Next, we investigated whether the amounts of catenins in vascular endothelial cells are changed by the influence of VAP1. The results showed that γ-catenin, but not β-catenin, is located at the region of the cell–cell junctions before VAP1 treatment (Fig. 1D,E). In contrast, after VAP1 treatment, γ-catenin disappeared from the cell membrane region in the cell–cell junctions, and fluorescence owing to γ-catenin was exclusively observed in the cytosol (Fig. 1E). The results suggest that VAP1 promoted relocation of VE-cadherin and γ-catenin from the cell membrane to the cytosol in cultured cells without degradation of VE-cadherin.

    Details are in the caption following the image
    VAP1 induces cadherin and catenin relocation. (A) HUVECs were incubated with or without 140 ng·mL−1 VAP1 in serum-free medium for 1 h at 37 °C. After being fixed, the cells were stained with anti-VE-cadherin antibody. Scale bar, 50 μm. (B) HUVECs were incubated with or without 140 ng·mL−1 VAP1 in the medium for 1 h at 37 °C. The cells (3 × 106 cells) were harvested with a lysis buffer (5% 2-mercaptoethanol, 2% SDS, 5% sucrose, 0.005% bromophenol blue, 63 mm Tris/HCl, pH 6.8) and were subjected to SDS/PAGE (10% separating gel). After blotting, the blotted membrane was treated with anti-VE-cadherin antibody and secondary horseradish peroxidase-conjugated antibody. Bands were detected by chemiluminescence. (C) Human ectodomain recombinants of VE-cadherin at 23 μg·mL−1 were incubated with 0.03, 0.1 and 1 μg·mL−1 VAP1 in PBS for 1 h at 37 °C. Samples (12 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining. (D) HUVECs were incubated with or without 70 ng·mL−1 VAP1 in serum-free medium for 1 h at 37 °C. After being fixed, the cells were stained with anti-β-catenin antibody (green) and Hoechst 33342 (blue). Arrows indicate β-catenin in the perinuclear region of the cytoplasm. Scale bar, 50 μm. (E) The same experiment as that in (D) was carried out except for the use of anti-γ-catenin antibody. Arrows indicate γ-catenin at the cell–cell junctions. Scale bar, 50 μm.

    VAP1 cleaves LRP6 in the activation region

    It is known that Wnt signalling induces catenin relocation [19-21]. It has also been reported that β-propeller domain-deleted LRP6 functions as a constitutively active Wnt-signalling activator, which allows catenin relocation [24, 25]. These results led us to investigate whether VAP1 carries out limited proteolysis of LRP6 that results in activation of Wnt signalling. The results showed that LRP6 was degraded by VAP1 in HeLa cells, which have a large amount of LRP6 (Fig. 2A). Furthermore, VAP1 cleaved a recombinant protein of human LRP6 extracellular domains, producing 140-kDa and 60-kDa fragments (Fig. 2B). VAP1 cleaved LRP6 faster than it did fibrinogen α-chain (Fig. 2B), which is known to be the most sensitive substrate for most ADAM-type SVMPs [27]. Although gelatin and fibronectin are also sensitive substrates for many SVMPs, gelatin is insensitive to VAP1 [15] and fibronectin was a less sensitive substrate than LRP6 for VAP1 (Fig. 2C). The 140-kDa and 60-kDa fragments of LRP6 separated by SDS/PAGE were subjected to in-gel digestion with trypsin, reduction and alkylation, and liquid chromatography–tandem mass spectrometry analysis. Since trypsin-digested fragments contain arginine and lysine residues in their C termini, we can identify a cleavage site by VAP1 by searching for a peptide missing a C-terminal basic residue. We detected a peptide, IAQLSDIHAVKE, from the 140-kDa fragment and also a peptide, LNLQEYR, from the 60-kDa fragment of LRP6. The results indicated that VAP1 cleaved LRP6 at Glu1196–Leu1197 (Fig. 2D). This site corresponds to the C terminus of the fourth β-propeller domain [not including the fourth epidermal growth factor (EGF) domain; schematic representation in Fig. 2D]. The molecular mass deduced from the sequences of N-terminal Glu1196 and C-terminal Leu1197 fragments was 133 kDa and 45 kDa, respectively, which coincided with 140 kDa and 60 kDa, respectively, estimated by SDS/PAGE in consideration of carbohydrate chains. The P1′ residue of the cleavage site was a hydrophobic amino acid, leucine, which would be easily accommodated by a hydrophobic pocket of the S1′ site on the basis of the 3D structure of VAP1 [28]. Our results indicate that VAP1 is able to remove all four inhibitory β-propeller structures from LRP6, leaving the C-terminal side activation-promotive low-density lipoprotein receptor class A (LDLa) domains (Fig. 2D) [24, 25, 29]. In other words, it is notable that VAP1, a haemorrhagic SVMP as an ADAM-type protease, is capable of cleaving LRP6 at the site presumed to generate the constitutively active form. Similarly, 0.1 μg·mL−1 VAP1 cleaved recombinant mouse LRP6, producing a 140-kDa fragment (Fig. 2E).

    Details are in the caption following the image
    VAP1 cleaves LRP6 at the activation region. (A) HeLa cells were incubated with or without 0.4 μg·mL−1 VAP1 in serum-free medium for 2 h at 37 °C. The cells (1.5 × 107 cells) were harvested with a lysis buffer (25 mm Tris/HCl, pH 7.4, 300 mm NaCl, 1.5 mm MgCl2, 0.5% Triton X-100, 1 mm PMSF, 4 mm EDTA). Samples were immunoprecipitated with an anti-LRP6 antibody. The precipitates were subjected to SDS/PAGE (10% separating gel) and western blotting by using an anti-LRP6 antibody as a primary antibody. (B) Ectodomain recombinants of human LRP6 and mouse LRP5 at 17 μg·mL−1 and 50 μg·mL−1 purified bovine fibrinogen were incubated with 0.03 μg·mL−1 VAP1 in PBS for 0, 1 and 3 h at 37 °C. Samples (12 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining. (C) Human fibronectin at 15 μg·mL−1 was incubated with each dose of VAP1 in PBS for 1 h at 37 °C. Samples (12 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining. (D) The VAP1-cleavage point is shown by a black arrow in the schematic representation of LRP6. Fragment sequences of 140-kDa and 60-kDa bands of LRP6 detected by mass spectroscopic analysis are shown by a double underline and dashed underline, respectively. (E) A mouse ectodomain recombinant of LRP6 at 10 μg·mL−1 was incubated with 0.1 μg·mL−1 VAP1 in PBS for 1 h at 37 °C. Samples (15 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining.

    VAP1 also cleaves LRP5 at a position identical to the cleavage site of LRP6

    Low-density lipoprotein receptor-related proteins 5, a paralogue of LRP6, also plays a key role in Wnt signalling, together with LRP6. In addition, it has been reported that LRP5 mutants lacking extracellular domains are constitutively active in Wnt signalling [30]. Therefore, we next investigated whether ADAM toxins are able to cleave LRP5, as well as LRP6. VAP1 cleaved mouse recombinant LRP5, though a higher concentration of VAP1 (0.3 μg·mL−1) was necessary to cleave LRP5 than to cleave LRP6 (compare Figs 2B and 3A). VAP2, another haemorrhagic ADAM toxin, also cleaved LRP5 (Fig. 3B). As was the case for LRP6, a 140-kDa fragment of LRP5 was produced by VAP1 (Fig. 3C). A peptide, VAHLTGIHAVEE, was detected by mass spectrometric analysis of the 140-kDa fragment, suggesting that the sessile bond by VAP1 is Glu1206–Val1207, which is the same position in the domain structure as the site of cleavage of LRP6 by VAP1 (Fig. 3D,E). P1′ and P3′ amino acid residues of both LRP5 and LRP6 were hydrophobic (Fig. 3E). Since truncated LRP5 is also a constitutive activator of the Wnt signalling pathway [30], it is likely that VAP1 activates not only LRP6 but also LRP5 via limited proteolysis, although the cleaving efficiency for LRP5 must be lower than that for LRP6.

    Details are in the caption following the image
    VAP1 cleaves LRP5 at the same site as that of LRP6. (A,B) A mouse ectodomain recombinant of LRP5 at 26 μg·mL−1 was incubated with each dose of VAP1 (A) and VAP2 (B) in PBS for 1 h at 37 °C. Samples (15 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining. (C) Mouse LRP5 ectodomain recombinant at 70 μg·mL−1 was incubated with 7 μg·mL−1 VAP1 in PBS for 1 h at 37 °C. Samples (60 μL) were subjected to SDS/PAGE (5–20% precast SDS/PAGE gel) and stained with Coomassie Brilliant Blue R250. (D) The VAP1-cleavage point is shown by a black arrow in the schematic representation of LRP5. The fragment sequences of the 140-kDa band of LRP5 detected by mass spectroscopic analysis is shown by a double underline. (E) Amino acid residues from cleavage site position P5 to P5′ of LRP6 and LRP5 are aligned. The VAP1-cleavage point is shown by a black arrow.

    Substrates and cleavage points of VAP1

    Hydrophobic residues were also preferred at the P1′ position of the VAP1-cleavage site of fibrinogen α chain and fibronectin (Fig. 4A), although these proteins were less sensitive to VAP1 than LRP6 (Fig. 2B,C). Similarly hydrophobic residues were preferred of the VAP1-cleavage site of perlecan and recombinant domains of fibronectin (Fig. 4A). Docking simulation by using a protein–protein docking prediction program demonstrated that the C-terminal portion from the cleaved bond of LRP6 (ball-and-stick model in Fig. 4B, inset) fits well into the catalytic cleft of VAP1 (space-filling model in Fig. 4B, inset). The portion is the short linker region between the fourth β-propeller domain and the fourth EGF domain of LRP6. The preference of hydrophobic residues at P1′ and P3′ is not at variance with the fact that the S1′ pocket and the potential S3′ pocket, which are newly made by dimerization of VAP1, consist of a hydrophobic residue, Leu363, near the VAP1 catalytic site (Fig. 4B).

    Details are in the caption following the image
    Substrates and cleavage points of VAP1. (A) Substrates cleaved by VAP1 are indicated. The cleaved fragments were subjected to N-terminal sequence analysis. P5 to P5′ amino acid residues of the cleavage points are shown. (B) Docking model of VAP1 and the cleavage site moiety of LRP6. The inset indicates the substrate-binding cleft. VAP1 is shown by a white space-filling model. The cleavage site moiety of LRP6 is shown by a red space-filling model (in the upper figure) and by a ball-and-stick model (in the lower figure). Zn2+ and Glu336A, both of which are involved in catalysis, and Leu363A are shown by grey spheres. S1′ and S3′ pocket regions of VAP1 are shown by white dotted circles.

    LRP6 cleavage by VAP1 is responsible for disruption of cell–cell junctions and haemorrhage

    In order to investigate whether the cleavage of LRP6 is involved in VAP1-induced cellular response, we made an antibody against a peptide corresponding to the VAP1-induced LRP6 cleavage site. The anti-LRP6 cleavage site antibody significantly inhibited the cleavage of recombinant LRP6 by VAP1 and production of the 140-kDa fragment of LRP6 (Fig. 5A,B). The LRP6 cleavage inhibitory antibody suppressed the cell–cell detachment by VAP1 on the endothelial cell layer as revealed by phase contrast microscopy (Fig. 5C). VAP1-induced elimination of VE-cadherin at the cell–cell junctions was also inhibited by this antibody (Fig. 5C). The results indicate that LRP6 cleavage by VAP1 is involved in disruption of cell–cell adhesions. Next, we investigated whether LRP6 cleavage is involved in VAP1-induced haemorrhage in vivo. The above cleavage-inhibitory antiserum, but not control serum, significantly inhibited intracutaneous haemorrhage induced by VAP1 injection (Fig. 5D,E). These results strongly support the idea that VAP1-induced haemorrhage is triggered by VAP1-mediated LRP6 cleavage in vivo.

    Details are in the caption following the image
    LRP6 cleavage by VAP1 is involved in disruption of cell–cell junctions and haemorrhage. (A,B) A human ectodomain recombinant of LRP6 at 30 μg·mL−1 was incubated with 30 ng·mL−1 VAP1 and with 2 mg·mL−1 LRP-cleavage site antibody or control antibody in PBS for 1 h at 37 °C. Samples (15 μL) were subjected to SDS/PAGE and silver staining (A). Cleavage inhibition is shown by the 140-kDa fragment density score (B). The data (n = 5; error bars correspond to standard errors) were compared by Student's t test. (C) HUVECs were incubated with 140 ng·mL−1 VAP1 and with 1.35 mg·mL−1 LRP6-cleavage site antibody or control antibody in a medium for 1 h at 37 °C. After being fixed, the cells were stained with anti-VE-cadherin antibody. Arrows indicate the remaining membrane VE-cadherin. Scale bar, 50 μm. (D,E) VAP1 with LRP6 cleavage site antiserum or control serum was intradermally injected into mice bisymmetrically. One hour later, haemorrhagic plaques on the inner surface of the skin were observed (D). Scale bar, 10 mm. Densitometry scores of VAP1-induced haemorrhage with the antiserum or control serum in each individual were compared by paired t test (E).

    VAP1-cleavage site and/or LDLa domains of LRP5/6 are deleted in many venom-resistant animals

    We showed that cleavage of LRP5/6 and their activation by VAP1 are responsible for haemorrhage. This finding prompted us to examine whether some snake venom-resistant animals have LRP5/6 variants that are not susceptible to SVMPs. Haemorrhage resistance against snake venom has been reported in venomous snakes themselves and in other animals including opossums, mongooses, hedgehogs, ground squirrels, rats and camels [31-34]. In searching the NCBI reference sequence database, regardless of haemorrhage resistance, animals often lacked the first/second β-propeller domains in LRP5/6 (Fig. 6A,B). The first domain is known to be a binding domain of many Wnts and related factors, while the third β-propeller domain is known to be a binding domain of Wnt3, 3a and DKK1 [35, 36]. Thus, the deletion variation may be related to the selectivity of Wnt isoforms. Predicted sequences of LRP5/6 in pythons and green anole lizards, which do not have ADAM toxins, have no deleted isoforms. Also in the Tasmanian devil, which belongs to marsupials together with opossums, deleted isoforms are not known. On the other hand, in snakebite haemorrhage-resistant animals, the king cobra possesses haemorrhagic SVMPs and shows resistance against SVMP-induced haemorrhage [37], although king cobra venom is known to be neurotoxic. Interestingly, king cobra LPR6 had no VAP1-cleavage site (Figs 6A and 7). In the grey short-tail opossum, the extracellular LDLa domains, which contribute to constitutive activation in truncated mutants of LRP6 [24, 25, 29], were deleted in two predicted LRP6 isoforms among three isoforms (Figs 6A and 7). King cobra LRP6 also lacked LDLa domains. Similar characteristic deletion of LRP6 was also observed in goats (Figs 6A and 7), which show moderate resistance to snakebite haemorrhage [33]. Regarding LRP5, either of the LDLa domains of LRP5 was deleted or replaced in rats, dromedary camels and European hedgehogs (Figs 6B and 7). In thirteen-lined ground squirrels, the VAP1-cleavage region of LRP5 was replaced with non-homologous sequences (Figs 6B and 7). Besides the abovementioned haemorrhage-resistant animals, similar characteristic deletions in LRP5/6 were observed in several animals including tigers, bears and orangutans, but it is not known whether these animals are resistant to haemorrhage (Figs 6B and 7). Taking all of these findings into account, we conclude that many snake venom-resistant animals possess characteristic LRP5/6 lacking the VAP1-cleavage site and/or LDLa domains.

    Details are in the caption following the image
    VAP1-cleavage site and/or LDLa domains of LRP5/6 are deleted in many venom-resitent animals. Deleted regions of LRP6 (A) and LRP5 (B) in animals are shown in schematic representation. LRP5/6 of animals associated with tolerance to snake bite-induced haemorrhage are indicated by red boxes. LRP5/6 of animals with moderate resistance and animals with unknown sensitivity to snake bite-induced haemorrhage are shown by orange and green boxes, respectively. The presented data are from NCBI RefSeq (Table S1). Except for humans, mice and rats, the data are all predicted sequences from the genome of each animal. Amino acid sequences of the cleavage site and LDLa region are shown in Fig. 7. Regarding the predicted king cobra LRP6 (ETE69657.1), although the genome sequence of the full-length mRNA region of LRP6 contains several gaps, the indicated deleted region does not contain gaps.
    Details are in the caption following the image
    Amino acid sequences of the cleavage site and LDLa region in animals. LRP5/6 and isoforms of animals are shown. Highlighted area shows deleted (pink) or replaced (green) region compared with LRP5/6 of humans, mice and guinea pigs.

    Recombinant human ADAM8 and ADAM12 cleave LRP6 at the same cleavage site as that of VAP1

    To determine whether LRP6 cleavage activity of VAP1 is a feature common to animal ADAMs, we used recombinant human ADAM8 and ADAM12, which show relatively high homology to VAP1 (40% and 38% identity in the metalloprotease domain, respectively). Both recombinant human ADAM8 and ADAM12 cleaved recombinant LRP6 and produced a 140-kDa fragment, as was the case for VAP1, although the hydrolysing activity was very weak compared with that of VAP1 (Fig. 8A,B). A 60-kDa fragment was also detected by SDS/PAGE after purification and concentration of the C-terminal fragment of ADAM12-cleaved LRP6 (Fig. 8C). By mass spectrometric analysis of the 140-kDa and 60-kDa fragments, the same cleaved-site fragments, IAQLSDIHEVKE and LNLQEYR, were detected in both ADAM8 and ADAM12 and in ADAM12, respectively, as was the case for VAP1 (Fig. 8D). These results indicate that specific cleavage of LRP6 and resulting activation of Wnt signalling is common among haemorrhagic SVMPs and human ADAMs (at least ADAM8 and ADAM12).

    Details are in the caption following the image
    Recombinant human ADAM8 and ADAM12 cleave LRP6 at the same site as that of VAP1. (A) Recombinant human LRP6 at 64 μg·mL−1 was incubated with 20 μg·mL−1 recombinant human ADAM8 for 3 h at 37 °C. Samples (10 μL) were subjected to SDS/PAGE (12% separating gel) and silver staining. (B) LRP6 at 20 μg·mL−1 was incubated with 20 μg·mL−1 recombinant human ADAM12 for 16 h at 37 °C. Samples (15 μL) were subjected to SDS/PAGE (10% separating gel) and silver staining. (C) LRP6 at 20 μg·mL−1 was incubated with 200 μg·mL−1 recombinant human ADAM12 for 18 h at 37 °C. Fragments containing C-terminal Fc peptide (20 μL) were purified with protein A–Sepharose. Samples were subjected to SDS/PAGE (15% separating gel) and silver staining. (D) The LRP6 140-kDa fragments treated with ADAM8 (A) and ADAM12 (B) and the LRP6 60-kDa fragment treated with ADAM12 (C) were analysed by mass spectrometry. Detected fragments by mass spectrometry and the proposed cleaved sites are shown.

    Discussion

    The present study demonstrated that LRP5/6 are cleaved in the Wnt signal-activation region by haemorrhagic SVMPs and human ADAMs. It was also demonstrated that the specific cleavage of LRP6 by ADAM-type toxin VAP1 is responsible for disruption of cell–cell junctions in culture cells and for haemorrhage in vivo.

    Historically, LRP5/6 mutants lacking extracellular β-propeller domains have been reported to be constitutively active in Wnt signalling [24, 25]. However, the reason for this has not been elucidated. Our results convincingly demonstrated that limited proteolysis of LRP5/6 by ADAMs eliminated the four β-propeller domains of LRP5/6, resulting in activation of Wnt signalling. In connection with this, it is notable that the complement factor C1q (with C1r and C1s) cleaves a middle part of the third β-propeller domain (Arg792–Ala793) of LRP6, which in turn activates the Wnt signalling pathway [38, 39]. These phenomena appear to be closely related to human ageing and to tissue functional decline and disease [38, 39]. Activation of protein kinase C and Wnt3a induces LRP6 shedding at the juxtamembrane region, although the responsible protease has not been identified [40]. It should be emphasized that haemorrhagic SVMPs and at least ADAM8 and ADAM12 in human ADAMs are activators of LRP5/6 (Fig. 9). In the present study, we showed that the relocation of γ-catenin and VE-cadherin is induced by LRP6 cleavage. γ-Catenin has been reported to have low transcription activity compared with that of β-catenin [41, 42]. Therefore, γ-catenin relocation induced by ADAM may be involved in disruption of cell–cell adhesions rather than induction of proliferation and epithelial (endothelial)–mesenchymal transition in vascular endothelial cells. Interestingly, it has been reported that γ-catenin rather than β-catenin exists at tricellular junctions, where leukocytes prefer the transvascular pathway [43]. On the other hand, ADAMs and SVMPs may also relocate β-catenin, similar to γ-catenin, because truncated LRP6 has been reported to induce β-catenin relocation [26].

    Details are in the caption following the image
    Hypothetical roles of ADAM and LRP5/6 in haemorrhage and invasion. (1) Cell–cell junctions with cadherin and catenin are normally stable. LRP5/6 may form a complex with cadherin and catenin. (2) Haemorrhagic SVMP, which is an ADAM-type toxin, can cleave LRP5/6 at the activation region. Cleaved LRP5/6 may form a dimer or multimer. (3) Cleaved LRP6 is involved in γ-catenin and VE-cadherin relocation and in disruption of cell–cell adhesions. (4) Cleaved LRP6 mediates haemorrhage. Some animals with snake venom tolerance have deleted LRP5/6. (2′) Invasion-associated ADAMs are located at the tips of invadopodia in invasive cells such as leukocytes and cancer cells. The ADAMs can cleave the same site of LRP6. (4′) Cleaved LRP6 has the potential to cause disruption of cell–cell adhesions and to induce invasion. Summarizing the above, it is thought that ADAMs, as cell barrier openers, cleave novel ADAM receptors, LRP5/6, to induce haemorrhage and potentially invasion.

    Various hypotheses for the mechanism of SVMP-induced haemorrhage, including ECM lysis and direct protein cleavage in cell–cell junctions, have been proposed [5, 44-46]. Haemorrhagic snake venom contains multiple SVMPs possessing a variety of broad or narrow substrate specificity. To induce haemorrhage, a set of ECM barrier openers and cell barrier openers are necessary. Therefore, several ADAM toxins may have broad substrate specificity to lyse ECM proteins. Collagen IV and perlecan are promising candidates for ECM targets of haemorrhagic SVMPs [46]. To disrupt the cell barrier, it has been reported that VE-cadherin is cleaved within several hours along with cell apoptosis directly or indirectly by another snake toxin, graminelysin [45]. On the other hand, VAP1 treatment resulted in disruption of cell–cell adhesions and haemorrhage within 1 h, and we showed that VE-cadherin was not cleaved by VAP1 directly or indirectly within this period. In contrast to the previous hypotheses, our results demonstrated that a novel cell barrier-opening ADAM receptor, which is involved in eventual haemorrhage, is LRP5/6 (Fig. 9). Although the detailed cooperative functions of haemorrhagic and non-haemorrhagic SVMPs in ECM barrier and cell barrier opening remain to be elucidated, the present study provides a new insight to understand the mechanism of snake venom haemorrhage. Regarding the substrate selectivity of SVMPs, disintegrin/cysteine-rich domains, which bind von Willebrand factor A domain and/or collagen, have been proposed to modify toxin localization [47, 48]. It is notable that LRP6 has been reported to form a complex with two receptors on endothelial cells, tumor endothelial marker 8 and capillary morphogenesis protein 2, both of which have a von Willebrand factor A domain [49]. Such proteins may support SVMPs to approach and cleave LRP6.

    In snake venom-resistant animals such as venomous snakes and opossums, various factors involved in venom-resistant mechanisms have been proposed, including serum protease inhibitors [50, 51], toxin-binding serum factors [52] and specialized coagulation factors [53]. In the present study, we showed that many haemorrhage-resistant animals have characteristic LRP5/6 lacking the VAP1-cleavage site and LDLa region. This indirectly indicates the importance of LRP5/6 cleavage and activation in venom-induced haemorrhage. Thus, deletion of LRP5/6 is also thought to be a unique and widely used haemorrhagic resistance strategy. It is also intriguing that the above resistant animals have the deletion in either LRP6 or LRP5 but not in both (Fig. 6A,B). Therefore, the lack of the region in either LRP5 or LRP6 may be sufficient to exhibit haemorrhage resistance. In other words, cleavage and activation of both LRP5 and LRP6, but not either one, may be necessary to induce haemorrhage. In this regard, it is notable that Wnt1, Wnt9b and Wnt10b require both LRP5 and LRP6 for their signalling, whereas Wnt3a requires only LRP6 [54].

    Several ADAMs including ADAM8 and ADAM12 have been reported to be involved in invasion of leukocytes and cancer cells [8-10]. In addition, the ADAMs have been shown to be located at the tips of invasive pseudopodia [55, 56]. However, the precise roles and the substrates of ADAMs in invasion are not known, though some ADAM substrates such as HB-EGF and Delta-like 1 have been reported to activate cells [57, 58]. We showed that ADAM8 and ADAM12 have the ability to cleave LRP6 at the activation site and that the activation is involved in cell barrier disruption. Thus, ADAMs at the tips of invasive pseudopodia of leukocytes and cancer cells may have a role in opening of the cell barrier by cleaving LRP5/6 during invasion (Fig. 8). Different from soluble SVMPs, which are able to function in a wide area of the vascular wall, membrane-anchored ADAMs such as ADAM8 and ADAM12 located at the tips of pseudopodia appear to function only in a restricted area. There are many reports on chemoattraction, cell adhesion and ECM lysis during leukocyte extravasation and cancer cell invasion [59]. However, the mechanisms of disruption of cell–cell adhesions in cell-barrier opening are poorly understood. Studies on the mechanisms of ADAM-mediated LRP5/6 cleavage and cell-barrier opening will provide a novel insight for studies on invasion and metastasis of cancer cells and extravasation of leukocytes.

    In conclusion, this is the first report that LRP5/6 are cleaved in the Wnt signal-activation region by snake ADAM toxins and several human ADAMs. It was also demonstrated that the specific cleavage of LRP6 by ADAM-type toxin VAP1 is responsible for disruption of cell–cell junctions in culture cells and for haemorrhage in vivo. Our results indicate that haemorrhagic SVMPs and at least ADAM8 and ADAM12 of the human ADAMs are novel activators of Wnt receptors LRP5/6. In addition, it should be emphasized that LRP5/6 are novel targets of SVMPs for cell barrier opening in snake-bite haemorrhage. ADAM–LRP5/6 signalling may also be involved in vascular barrier opening in leukocyte extravasation and cancer cell invasion. The results of the present study provide insights into future design of drugs for cancer metastasis, inflammation and snake bite-induced haemorrhage.

    Experimental procedures

    Ethics statement

    The care and use of mice were conducted in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Our experimental protocols were approved by the Committee for the Care and Use of Experimental Animals in Nagoya University, Japan.

    Materials

    Vascular apoptosis-inducing protein 1 and VAP2 were isolated as previously described [12]. In brief, crude venom from Crotalus atrox (Sigma-Aldrich, St Louis, MO, USA) was subjected to hydroxyapatite chromatography and anion ion-exchange chromatography. Recombinant human LRP6 with an Fc-tag, human ADAM8 and ADAM12, VE-cadherin, and mouse LRP6 and LRP5 having a His-tag were purchased from R&D Systems, Inc. (Minneapolis, MN, USA).

    Cells

    Human umbilical vein endothelial cells (HUVECs) were purchased from Becton, Dickinson and Company (Franklin Lakes, NJ, USA). Fibroblast growth factor (FGF) was isolated as described previously [12]. HUVECs were cultured in MCDB105 medium (Sigma-Aldrich) containing 10% fetal bovine serum and FGF at 37 °C. HeLa cells (RIKEN BioResource Center, Tsukuba, Japan) were cultured in MCDB105 medium with 10% fetal bovine serum at 37 °C.

    Mass spectrometry assay

    Cleaved fragments and controls were subjected to SDS/PAGE. The gels were stained by using a SilverQuest™ Silver Staining Kit (Thermo Fisher Scientific, Waltham, MA, USA) and cut into fragment band pieces. The pieces were subjected to in-gel digestion. After treatment with DTT and iodoacetamide, the gels were treated with 0.01 μg·mL−1 trypsin in 50 mm ammonium bicarbonate at 37 °C overnight. Digests were subjected to liquid chromatography–mass spectrometry analysis using an Ultimate3000 liquid chromatogram (Thermo Fisher Scientific) and an LTQ-XL mass spectrometer (Thermo Fisher Scientific). The results were confirmed three times by liquid chromatography–mass spectrometry analysis. A 60-kDa fragment with a C-terminal Fc peptide that had been cleaved by ADAM12 was purified by using protein A–Sepharose beads (Sigma-Aldrich). Proteins that had been adsorbed to protein A–Sepharose beads by incubating for 30 min were washed with PBS and were eluted with lysis buffer [5% 2-mercaptoethanol, 2% SDS, 5% sucrose, 0.005% Bromophenol Blue, 63 mm Tris/HCl (pH 6.8)]. These samples were subjected to SDS/PAGE.

    Substrate specificity analysis

    Human fibronectin (Wako, Osaka, Japan) and heparan sulfate proteoglycan (perlecan) (Sigma-Aldrich) were incubated with 10 μg·mL−1 VAP1 for 3 h at 37 °C. Bovine fibrinogen (Wako) was incubated with 10 μg·mL−1 VAP1 for 12 h at 37 °C. Recombinant proteins of the first and second fibronectin type 3 domains (rFib1, 2) and the second and third domains (rFib2, 3) of human fibronectin fused with glutathione S-transferase were made with Escherichia coli according to the procedures described previously [60]. rFib1, 2 and rFib2, 3 were incubated with 20 μg·mL−1 VAP1 for 4 h at 37 °C. Digested fragments were subjected to SDS/PAGE and transferred to a poly(vinylidene difluoride) (PVDF) membrane. The PVDF membrane was stained with Coomassie Brilliant Blue R250, and the bands that appeared with digestion were cut out and subjected to N-terminal sequence analysis at Hokkaido University Global Facility Center (Sapporo, Japan). Docking simulation was performed by using the software zdock 3.0.2 [61]. Three-dimensional structural data of VAP1 (2ERP) and LRP6 (3S8Z) were obtained from the protein database of NCBI. The docked structure was analysed by using the molecule visualization software icm-browser (Molsoft, San Diego, CA, USA).

    Immnoprecipitation

    Harvested cells were lysed with a lysis buffer (25 mm Tris/HCl, pH 7.4, 300 mm NaCl, 1.5 mm MgCl2, 1 mm PMSF, 4 mm EDTA, 0.5% Triton X-100). Samples were incubated for 30 min on ice and centrifuged at 12 000 g for 30 min at 4 °C. Supernatants were incubated overnight with 2.5 μg anti-human LRP6 polyclonal goat antibody (R&D Sytems, Inc.) at 4 °C. Then the samples were incubated with Protein G–agarose at 4 °C for 1 h and washed with lysis buffer two times. Precipitates were subjected to SDS/PAGE and western blotting.

    Western blotting

    Harvested cells were lysed with each lysis buffer. The samples were subjected to SDS/PAGE. After SDS/PAGE, the samples were transferred onto PVDF membranes. After blocking with 5% skim milk in Tris-buffered saline, 0.1% Tween 20 buffer, the membranes were incubated with the indicated dose of anti-VE-cadherin monoclonal mouse antibody clone TEA1/31 (Beckman Coulter Inc., Brea, CA, USA) or anti-human LRP6 polyclonal goat antibody (R&D Systems, Inc.). The membranes were then washed in Tris-buffered saline, 0.1% Tween 20 and incubated with 1/3000 anti-mouse IgG-HRP (GE Healthcare, Little Chalfont, UK) or 1/30000 anti-goat IgG-HRP (Abcam, Cambridge, UK). Reacted bands were detected by chemiluminescence (ECL prime, GE Healthcare).

    Immunofluorescence analysis

    The cells were fixed with 4% paraformaldehyde, blocked by 1% skim milk, permeated with 0.1% saponin, and incubated overnight with 1/200 anti-VE-cadherin antibody (Beckman Coulter Inc.) or 1/200 anti-γ-catenin antibody (Novus Biologicals, Minneapolis, MN, USA). Then the cells were washed with Tris-buffered saline three times and incubated with Alexa Fluor 488-conjugated anti-mouse IgG antibody (Invitrogen, Carlsbad, CA, USA). Images were obtained by an inverted fluorescence microscope (IX83, Olympus, Tokyo, Japan).

    Neutralization by antibody

    Rabbit antiserum was generated against a peptide around the VAP1-cleavage site of LRP6. Polyclonal antibodies were purified from the antiserum and control rabbit serum by using protein A. HUVECs were incubated with 1.35 μg·mL−1 of each of the antibodies for 1 h. Then the cells were treated with 140 ng·mL−1 VAP1 for 1 h.

    Haemorrhage assay

    Five micrograms VAP1 with 20 μL LRP6 cleavage-site antiserum or control serum was injected intradermally into each of two bisymmetrical points of the shaved dorsal skin of male mice. One hour later, the mice were sacrificed and the skin was carefully removed, without stretching, and the haemorrhagic spot on the inner surface of the skin was measured. The haemorrhage index of the haemorrhage plaque was calculated as the summation of red colour density with a densitometer (CS Analyzer, ATTO, Tokyo, Japan).

    Statistical analysis

    Data (means ± standard error) were analysed using Student's t test. For the haemorrhage assay, analyses of two groups were assessed by the paired t test. P values smaller than 0.05 (P < 0.05) were considered as indicating a significant difference.

    Acknowledgements

    This work was supported in part by Grant-in-aid (No. 22590061) for Scientific Research (C) from JSPS, Japan. We are grateful to Mrs Tomoko Shibuya, Ms Maiko Kitagawa, Mr Kazuma Onizuka, Ms Yui Tajima and Mr Yuto Suzuki for their technical assistance. We are grateful to Prof. Hiroki Shibata, Kyusyu University, for helpful suggestions.

      Conflicts of interest

      The authors declare no conflict of interest.

      Author contributions

      TSe and SA designed the research and analysed data; TSe, TSa, SN, AN, KW, NS, KM, and MA performed the research; TSe, HS and SA wrote the paper.