Antibodies binding the ADAM10 substrate recognition domain inhibit Eph function

Proteolytic release, or ‘shedding’, of cell surface-bound proteins acts as an important post-translational switch that regulates protein function and activity. The ADAM (a disintegrin and metalloprotease) family of transmembrane proteases are the most prominent shedding enzymes for membrane-anchored proteins. ADAMs contain multiple extracellular domains, including a distal metalloprotease (MP) domain, followed by disintegrin (D)- and cysteine-rich (C) domains involved in substrate interaction, as well as transmembrane and variable cytoplasmic sequences (Blobel, 2005). They are important in regulating inflammatory and growth factor signalling, cell migration, and cell adhesion: in particular, two closely related, atypical ADAMs, ADAM10 (CD156C, MADM, Kuzbanian) and 17 [CD156B, TACE (TNFα-converting enzyme)], shed ligands and/or receptors regulating key cytokine, chemokine and growth factor signalling pathways important in disease. These include erbB/EGF receptor family ligands and receptors, Notch receptors and ligands, TNFα and TNFRI and II, CX3CL1, IL-6R, as well as cadherins and various cellular adhesion molecules (CAMs), and the amyloid precursor protein (APP) (Murphy, 2008; Saftig and Reiss, 2011). ADAM10 and 17 are also overexpressed in a variety of cancers (Murphy, 2008; Saftig and Reiss, 2011; Sanderson et al., 2006). Together this implies their important involvement in diseases such as Alzheimer's, chronic inflammatory and heart diseases, and cancer.

ADAM10 also cleaves ligands for Eph receptors, the largest family of receptor tyrosine kinases, which together with their membrane-bound ephrin ligands, control cell migration and positioning during normal and oncogenic development (Nievergall et al., 2012; Pasquale, 2010). In this context ADAM10 association with A-type Eph receptors is promoted by binding to their ephrin-A ligands on interacting cells (Janes et al., 2005; Salaita et al., 2010), whereupon ADAM10 cleaves ephrin, disrupting the Eph–ephrin tether between cells to allow de-adhesion, or retraction (Hattori et al., 2000; Janes et al., 2005). This function of ADAM10 is further regulated by kinase activity (Blobel, 2005; Hattori et al., 2000), which we found to be mediated through conformational changes in the Eph cytoplasmic domain (Janes et al., 2009), such that ADAM10 acts as a switch between cell-cell adhesion and segregation in response to Eph phosphorylation levels. This switch is thought to be important for Eph-dependent oncogenesis, where aberrant Eph receptor expression and/or mutation contributes to tumour development by promoting neo-angiogenesis, invasion and metastasis (Nievergall et al., 2012; Pasquale, 2010). Interestingly, while EphB/ephrin-B cell contacts were reported to be attenuated through protease-independent trans-endocytosis (Marston et al., 2003; Zimmer et al., 2003), ADAM10 was also recently found to be required for EphB/ephrin-B-dependent cell sorting, where EphB2 activation triggers ADAM10-mediated shedding also of E-cadherin (Solanas et al., 2011).

Despite considerable efforts to develop ADAM metalloprotease inhibitors, to date clinical trials based on compounds blocking the protease catalytic site have failed due to lack of efficacy and specificity (DasGupta et al., 2009; Moss et al., 2001; Saftig and Reiss, 2011). To a large extent, this reflects similarity of the MP active site to matrix metalloproteases (MMPs) (Maskos et al., 1998), and the mechanism of ADAM substrate specificity, which does not rely on a typical cleavage signature recognised by the protease domain, but on non-catalytic interactions between the substrate and the ADAM C domain (Reddy et al., 2000; Smith et al., 2002; White, 2003). We have previously used structure/function studies to identify a substrate-binding pocket within the ADAM10 C domain, which specifically recognises the Eph/ephrin complex and thereby specifies cleavage of Eph-bound ephrin (Janes et al., 2005). We therefore set out to raise monoclonal antibodies (mAbs) against this region and assess their ability to block substrate cleavage. We now describe mAbs specific for the ADAM10 substrate-binding pocket, which inhibit ADAM10-mediated ephrin cleavage, Eph activity and Eph-dependent cell behaviour.

To generate mAbs that selectively bind the substrate recognition pocket within the C domain of the native ADAM10 extracellular domain, we sequentially immunised and boosted mice with ADAM10/EphA3^+ve human embryonic kidney (HEK) 293 cells and recombinant ADAM10 extracellular domain (ECD) fragments, respectively. In particular, we used a protein fragment spanning residues 214–646 of recombinant bovine ADAM10 ECD (Janes et al., 2005), in keeping with the notion that the lower homology to mouse sequences within the C domain (92.7%, compared to 94.8% homology for human; Fig. 1A), may increase its immunogenicity and bias the mouse immune response against this domain. Screening of hybridoma cell lines generated from these mice, using ELISA and Plasmon Resonance analysis to test binding to recombinant bovine ADAM10 ECD, and immunoprecipitation of endogenous human ADAM10 from human embryonic kidney (HEK) 293 cell lysates, yielded three monoclonal antibodies: 3A8 and 8C7, recognising bovine and human ADAM10 (Fig. 1B,C), as well as 4G1, binding only to bovine ADAM10 (Fig. 1B; supplementary material Fig. S1). 8C7 also bound to mouse ADAM10 present in lysates of wild-type (Wt) mouse embryonic fibroblasts (MEFs), but not in immunoprecipitates from ADAM−/− MEFs (Hartmann et al., 2002), indicating its specificity for ADAM10 (Fig. 1D).

We previously reported that association of EphA3 and ADAM10 is increased by ligand-induced receptor clustering (Janes et al., 2005), a finding since confirmed for other Eph receptors (Salaita et al., 2010; Solanas et al., 2011). To investigate if our mAbs would recognise cell surface ADAM10 in the context of associated EphA3, we analysed EphA3/HEK293 cells expressing abundant endogenous ADAM10 (above) by confocal microscopy. Staining of the cells with Alexa⁶⁴⁷-conjugated 8C7 (Alexa⁶⁴⁷–8C7), when performed on ice to prevent endocytosis, was faint (Fig. 2A, top row), consistent with the reported tightly controlled levels of cell surface ADAM10 (Marcello et al., 2010). Interestingly, incubation with 8C7 at 37°C resulted in prominent punctuate staining and notable internalisation of stained complexes (Fig. 2A, 60 min, 2nd row), suggesting antibody-induced clustering and its endocytosis together with ADAM10-containing protein complexes.

We next sought to test if the 8C7 antibody also recognises ADAM10 that is associated with EphA3 during activation, by incubating cells with 8C7 and with ephrin-A5-Fc, which when cross-linked with anti (α)-human Fc antibodies causes EphA3 activation and internalisation (Lawrenson et al., 2002; Vearing et al., 2005). Again, we added the Alexa⁶⁴⁷–8C7 mAb and Alexa⁴⁸⁸–ephrin-A5-Fc to cells on ice, in order to prevent endocytosis and internalisation of ADAM10-bound 8C7 prior to stimulation. Cells that had been fixed immediately (0 min, ephrin-A5), and those treated for indicated times with α-human IgG at 37°C to cross-link bound ephrin-A5-Fc, displayed closely matching 8C7 and ephrin-A5-staining patterns, suggesting that 8C7 effectively binds to ADAM10 that is associated with Eph/ephrin signalling complexes (Fig. 2B). This is particularly evident after (cross-linked) ephrin-A5-Fc-mediated EphA3 clustering and endocytosis, revealing pronounced uptake of 8C7 into EphA3-containing endocytic vesicles (Fig. 2B, 15, 60 min).

To assess the epitope specificity of our ADAM10 mAbs, we tested their ability to recognise isolated human (hu)-ADAM10 protein fragments, corresponding either to the full length ECD (lacking the Pro domain), or the disintegrin and cysteine-rich domains, expressed together or separately. Western blot analysis revealed that both 3A8 and 8C7 recognise recombinant protein fragments containing the cysteine-rich domain, but not the isolated disintegrin domain alone (supplementary material Fig. S2A). In comparison, 8C7 does not bind the isolated D+C domains from either ADAM17 or ADAM19 (supplementary material Fig. S2B), confirming its specificity. Importantly, BIAcore analysis verified binding of the isolated ADAM10 C domain to 8C7 conjugated to the sensor surface, while in comparison the R&D α-ADAM10 mAb (clone 1427) recognises only the full length ECD, but not the isolated C domain (Table 1). Interestingly, 8C7 seems to bind with six- to seven-fold higher affinity to the isolated C domain (∼14 nM) as compared to the entire ECD (∼93 nM); this is mainly due to a substantially higher association rate, suggesting this interaction might be conformation dependent, with the epitope partially masked in the full length molecule.

We next sought to test if our mAbs selectively target the substrate-binding pocket within the ADAM10 C domain. We previously identified three glutamic acid residues at positions 573, 578 and 579 which contribute to the negative charge of the pocket and are essential for efficient binding of the ephrin/Eph complex, as shown by Glu/Ala substitution at these sites (ADAM10 3E-A mutant) (Janes et al., 2005). We therefore compared binding of αADAM10 mAbs to Wt huADAM10–GFP and the ADAM10 3E-A mutant, by immunoprecipitation from lysates of transfected ADAM10^−/− MEFs. When compared to the R&D α-ADAM10 mAb, which binds equally well to Wt and mutant ADAM10 (supplementary material Fig. S3A), both 8C7 and 3A8 showed significantly attenuated recognition of the substrate-binding pocket mutant, indicating their specificity for the Wt form of this region (Fig. 3B). Similarly, Ala substitutions at residues 617 and 618, which are part of the lip of the binding pocket (Fig. 3A), also decreased binding of our mAbs, while 8C7 binding to an unrelated, catalytic domain ADAM10 mutant with a Glu384 to Ala substitution was equivalent to the Wt protein (supplementary material Fig. S3B).

These data, suggesting that our mAbs may specifically bind, and thereby block access to the substrate recognition motif (ephrin/Eph-binding pocket) of ADAM10, prompted us to test their effect on ADAM-mediated ephrin cleavage, and its uptake into Eph-expressing HEK293 cells. For this we monitored cleavage of ephrin from ephrin-A5-Fc-coated tissue culture plates by HEK293 cells stably transfected with either EphA3 or with EphB2, which also binds and is activated by ephrin-A5 (Himanen et al., 2004). Cells that had been treated with increasing doses of 8C7 were plated onto wells pre-coated with Alexa⁵⁹⁴-labelled ephrin-A5-Fc. After 30 min the cells were recovered and analysed for internalised Alexa⁵⁹⁴–ephrin-A5 by microscopy (Fig. 4A). In similar experiments we analysed by flow cytometry ephrin shedding and uptake by cells treated with 8C7 alone, or with anti-mouse IgG cross-linked 8C7, with enhanced binding avidity (Fig. 4B). Both image and flow analysis showed robust uptake of ephrin-A5 by the cells, and its significant, dose-dependent inhibition by pre-incubation of cells with 8C7 mAb, which was most pronounced with cross-linked 8C7.

To test this inhibition of ADAM10 in a physiologically more relevant model, we assessed ephrin cleavage and uptake in co-cultures of EphA3/HEK293 cells and GFP–ephrin-A5/HEK293 cells (Janes et al., 2005), by monitoring co-localised GFP–ephrin within Cell-Tracker-Red-labelled EphA3 cells (Fig. 4C, cont., arrows). Pre-incubation of cell cultures with 8C7 significantly inhibited ephrin uptake. Despite a more modest effect, likely reflecting the lesser sensitivity of this assay, inhibition by 8C7 was significantly stronger than that achieved by treatment with the broad MMP/ADAM metalloprotease inhibitors GM6001 or TAPI1 (Fig. 4C, graph). Inhibition of cleavage was particularly evident in images in which the GFP-tagged ephrin persisted at cell-cell junctions, rather than being cleaved and internalised by the EphA3-expressing cells (Fig. 4C, 8C7, ⁠).

Receptor endocytosis is an important modulator of signalling, not only in signal attenuation, but importantly by facilitating cytosolic signalling on endosomes acting as signalling platforms (Dobrowolski and De Robertis, 2012). We therefore tested the effect of 8C7 treatment on Eph receptor tyrosine phosphorylation in co-cultures of EphA3- and ephrin-A5-expressing HEK293 cells. Pre-incubation of the cells with increasing concentrations of 8C7 caused a reproducible, dose-dependent decrease in EphA3 phosphorylation (Fig. 5), particularly at later time points (>10 min), consistent with its likely effects on receptor endocytosis. In contrast, 8C7 treatment had no effect on activation of EphA3 by soluble clustered ephrin-A5-Fc (Fig. 5C), which causes EphA3 activation and internalisation independent of ADAM activity, in keeping with 8C7 acting via inhibition of ephrin cleavage.

High or low levels of Eph kinase activity correlate with cell-cell segregation or adhesion, respectively (Holmberg et al., 2000; Janes et al., 2011; Nievergall et al., 2012; Poliakov et al., 2008). Also, ADAM10-mediated cleavage of ephrin (Hattori et al., 2000; Janes et al., 2005) and cadherins (Solanas et al., 2011) during Eph/ephrin-mediated cell-cell interactions suggests essential roles in severing the bond between cells that allows cell segregation between the ephrin- and Eph-expressing populations to occur. We analysed effects of our ADAM10 mAbs specifically on cleavage of ephrins during Eph/ephrin-mediated cell repulsion, using a modified stripe assay (Knöll et al., 2007; Walter et al., 1987), in which Eph-expressing HEK293 cells are plated onto coverslips pre-coated with stripes of fluorescently labelled ephrin, in our case Alexa⁵⁹⁴–ephrinA5-Fc and EphB2/HEK293 cells. As a positive control for Eph-dependent cell adhesion, we tested cells expressing a truncated, signalling defective, dominant negative mutant EphB2 (ΔICD), which cannot segregate in response to ephrin (Janes et al., 2011). EphB2/293 cells were found to actively avoid stripes of high ephrin density, with only 20% remaining on stripes after overnight incubation (Fig. 6A, top row). In comparison, EphB2 ΔICD cells displayed marked adhesion to the stripes (Fig. 6A, bottom row), consistent with the Eph cytoplasmic domain being required for Eph-mediated cell repulsion (Janes et al., 2011; Mellitzer et al., 1999) and validating the assay. Interestingly, pre-treatment of EphB2/293 cells with increasing concentrations of 8C7 caused significant, dose-dependent inhibition of repulsion from ephrinA5-Fc stripes, leading to a more random distribution, similar to the effect of the metalloprotease inhibitor GM6001 (Fig. 6A,B), confirming specific ADAM10 inhibition as effective in blocking Eph-mediated repulsion. We confirmed that, underlying this inhibition of cell retraction, 8C7 inhibits EphB2 phosphorylation that is induced with cell-expressed ephrin-A5 to stimulate its kinase activity: as expected and consistent with its effects on ephrin-A5/EphB2-mediated repulsion, we observed marked, dose-dependent inhibition of EphB2 phosphorylation by 8C7 (Fig. 6C).

We also tested effects of 8C7-inhibited ADAM10 activity on the segregation, or sorting, between Eph- and ephrin-expressing cells, a process previously shown to be dependent on Eph signalling (Poliakov et al., 2008) and on ADAM10 (Solanas et al., 2011). In co-cultures EphB2/293 cells, co-expressing membrane-targeted GFP, were found to segregate from ephrin-A5/293 cells, such that the GFP-labelled EphB2 cells formed tight clusters of increased fluorescence intensity compared to mono-cultures, as previously observed when co-cultured with ephrin-B1/293 cells (Janes et al., 2011; Poliakov et al., 2004). Co-incubation with 8C7 alone, or cross-linked with anti-mouse antibody, clearly affected this segregation process, reducing the number of bright segregated colonies, as determined by image analysis of intensity-thresholded images (Fig. 7A–C). Inhibition was most pronounced using cross-linked 8C7, consistent with its greater inhibition of ephrin cleavage (Fig. 4), and comparable to effects of the less selective inhibitor TAPI1. In separate experiments, cross-linked 8C7 treatment also dose-dependently inhibited segregation of EphB2- and ephrin-B1-293 cells (Fig. 7D), and segregation of U251 glioma and ephrin-A-expressing HEK293 cells (Fig. 7E), a co-culture model we previously used as a cancer relevant model of Eph/ephrin-A-dependent cell sorting (Nievergall et al., 2010).

Inhibitors of ADAM10 and its closest relative ADAM17 (TACE) have been actively sought over the last decade. While inhibition of TNFα shedding by TACE was the initial target for treatment of inflammatory diseases, including rheumatoid arthritis, the role of both ADAM10 and 17 in shedding a range of disease relevant targets has broadened interest in raising ADAM-targeted therapeutics. This is particularly true in cancer, in which growth factor/RTK signalling is frequently de-regulated, and where ADAM-mediated shedding of ligands for the EGF receptor (erbB, HER) family underlies autocrine feedback and transactivation of EGFR signalling by stress and G-protein-coupled receptor signalling (Blobel, 2005; Fischer et al., 2003). ADAM10- and -17-mediated cleavage of HER2 and HER4, respectively, also contribute to their signalling and undermine receptor-targeted therapies (Liu et al., 2006; Rio et al., 2000; Sardi et al., 2006). In addition, ADAM10-mediated shedding regulates cell-cell interactions important for tumour development and invasion, in particular via targeting cadherins (Solanas et al., 2011) and ephrins (Hattori et al., 2000), which modulate cell-cell adhesion and segregation during multiple stages of cancer development (Nievergall et al., 2012; Pasquale, 2010).

Despite this interest, and the validation of ADAM inhibition in vitro, no inhibitors targeting the metalloprotease catalytic site have yet successfully passed phase II clinical trials, due to lack of specificity, stemming from the structural similarity between the proteolytic cleft of ADAMs and matrix metalloproteases (MMPs) (DasGupta et al., 2009; Saftig and Reiss, 2011). However, the specificity of ADAM function is mediated via enzyme-substrate interactions outside the protease domain, within the Cys-rich domain, and we previously elucidated the structure of the ADAM10 Cys-rich domain and identified a binding pocket that confers selectivity for the EphA/ephrin-A complex formed at cell-cell junctions (Janes et al., 2005). We thus surmised that antibodies raised against this pocket might potentially block substrate binding and more specifically inhibit ADAM10-mediated proteolysis.

We now describe α-ADAM10 mAbs, which recognise the C domain and appear to specifically target the substrate-binding pocket, as suggested by loss of binding following mutagenesis of key exposed residues within this region. Interestingly, considerably higher affinity of these mAbs for the isolated C domain compared to the full-length extracellular domain, principally due to a higher association rate, suggests the binding epitope may be masked in much of the full-length protein, and thus dependent on the conformation of the ADAM10 ECD. Although the structure of the full ADAM10 ECD is yet to be solved, functional studies of the close relative ADAM17 suggest activity-dependent conformation changes (Wang et al., 2009; Willems et al., 2010), and structures of related snake venom metalloproteases also suggest distinct open and closed conformations that are suggested to regulate substrate access to the C domain (Guan et al., 2010), a notion which would be consistent with the binding behaviour of our mAbs to this region.

Our α-ADAM10 antibodies also show specific binding to ADAM10 on cells, where staining with 8C7 colocalises with ephrin-A5-Fc-bound EphA3, consistent with the previous reported interactions of ADAM10 with EphA receptors (Janes et al., 2005; Salaita et al., 2010; Solanas et al., 2011). Moreover, 8C7 is effectively co-internalised following EphA3 receptor stimulation with soluble ephrin, suggesting a close functional association of ADAM10 and EphA3 in cells and the effective targeting of this complex by 8C7. Importantly, assays with recombinant or cell-expressed ephrin show that 8C7 blocks cleavage and internalisation of ephrin-/Eph complexes, but also inhibits receptor phosphorylation and ensuing biological responses. Thus, repulsion of EphB2-expressing HEK293 cells from stripes of immobilised ephrin-A5-Fc was inhibited, as was segregation of cells expressing various combinations of A- and B-type Ephs and ephrins. The inhibitory effect of 8C7 on EphB2-mediated ephrin shedding, cell repulsion and cell-cell segregation in our assays indicates that in addition to EphA/ephrin-A clusters (Janes et al., 2005; Salaita et al., 2010), ADAM10 also effectively targets ephrins bound to EphB2: It confirms the recently-reported interaction between ADAM10 and EphBs (Solanas et al., 2011), and demonstrates that 8C7 is effective in blocking the ADAM10-facilitated function of both Eph subtypes.

Interestingly, 8C7 also inhibited Eph receptor phosphorylation in EphA3- and EphB2-expressing HEK293 cells in response to ephrin-A5/HEK293 cells, but did not affect EphA3 stimulation with soluble, pre-clustered ephrin-A5-Fc. We argue that rather than directly affecting ephrin-induced Eph activation, 8C7 inhibition of ADAM10 prevents ephrin cleavage and thus endocytosis of ligated Ephs, resulting in persisting Eph/ephrin clusters at cell-cell junctions (Fig. 4B), highly reminiscent of those observed in cells overexpressing dominant-negative ADAM10 (Nievergall et al., 2010). However, the increased residency at the cell membrane does not result in a prolonged activation of EphA3, but in agreement with the lower Eph activity commonly observed under conditions of Eph-mediated cell-cell adhesion (Holmberg et al., 2000; Janes et al., 2011; Nievergall et al., 2012; Poliakov et al., 2008) leads to an overall reduced phosphorylation. In agreement, in other studies we have also noted that inhibition of EphA3 endocytosis by overexpression of the clathrin coat-associated protein AP180 decreases ligand-induced receptor phosphorylation (C. Stegmeyer, P.W.J., M.L., unpublished data). The underlying relationship between reduced EphA3 endocytosis and phosphorylation is also consistent with the inhibition of EphA3 endocytosis observed upon overexpression of the protein tyrosine phosphatase (PTP) 1B, which regulates EphA3 function and trafficking by controlling its phosphorylation and kinase activity (Nievergall et al., 2010). Reduction of EphA3 phosphorylation by 8C7 was particularly evident at later time points (after 10 min) coincident with EphA3 endocytosis (Wimmer-Kleikamp et al., 2004) which would imply it has reduced tyrosine-dependent signalling from endosomes. This is in keeping with the notion that endosomes, rather than merely downregulating RTK signalling, compartmentalise and thus facilitate signals that are prevented at the plasma membrane (Dobrowolski and De Robertis, 2012; Scita and Di Fiore, 2010). A prominent mechanism, known to regulate endocytic signalling of several RTKs including receptors for EGF, PDGF and Insulin, is via RTK-initiated, localised production of reactive oxygen species (ROS), which accumulate in endosomes (Ushio-Fukai, 2009), and transiently inactivate regulatory PTPs to enhance RTK phosphorylation and signalling activity (Oakley et al., 2009; Rhee, 2006; Tiganis, 2011). While such a mechanism remains to be confirmed for Ephs, similarities in the regulation of their signalling activity by regulatory PTPs and endocytosis (Nievergall et al., 2012; Pasquale, 2010) would suggest a similar concept controlling Eph signalling.

In conclusion, our novel monoclonal antibodies against ADAM10, by targeting the substrate recognition pocket, effectively inhibit its ability to cleave ephrin from cell surfaces and thus facilitate Eph function. Considering that ADAM10 has a variety of substrates in addition to ephrins, many of which are critically implicated in a range of chronic diseases, including inflammatory, heart and neurodegenerative diseases as well as cancer, it will be of considerable interest to examine the efficacy of our antibodies to inhibit shedding of these other ADAM10 substrates in cell and animal-based disease models. Importantly, the high specificity of our mAbs for ADAM10 will likely provide a significant advantage over current, less selective, protease-targeted inhibitors. We expect that highly specific inhibition of ADAM10 activity in pro-inflammatory and/or oncogenic signalling pathways will provide a basis for the development of our antibodies as potential therapeutics.

The sequences encoding the extracellular domain [amino acids (AA) 214–646, lacking the prodomain] from bovine (Howard et al., 1996) and human (OriGene) ADAM10, as well as isolated disintegrin (AA 455–550) and cysteine-rich domains (AA 551–646), or D+C domains combined (AA 455–646), were subcloned as Fc fusions into pcDNA3.1 vector (Invitrogen), including an N-terminal prolactin signal sequence and a C-terminal thrombin cleavage site followed by the Fc domain of human IgG, for expression in human embryonic kidney 293 (HEK293) cells. Following initial purification on Protein-A–Sepharose (Amersham), the Fc tag was removed by thrombin cleavage, and ADAM10 proteins were further purified to homogeneity by gel filtration chromatography.

Point mutations to human ADAM10-turboGFP (OriGene) were introduced by site-directed mutagenesis (Quickchange XL, Stratagene). Human ADAM10 3EA and 617AA mutants were generated by introducing Alanines at G_573,578,579, or at RH_617,618, respectively.

Mice were immunised with ADAM10/EphA3-expressing HEK293 cells, followed by subsequent injections with bovine ADAM10 extracellular domain and isolated D+C fragments. Subsequently, B cells from mouse spleens were fused with myeloma cells in order to generate hybridomas which were selected for monoclonal antibody production. Hybridoma supernatants from pooled fusions and from isolated clones were collected and screened for binding to ADAM10 using enzyme-linked immunosorbant assay (ELISA), followed by IP/western blotting, immunofluorescence and surface plasmon resonance (BIAcore) analysis, as described in the text. Antibodies were purified from supernatants on Protein G.

Analysis of protein interactions by surface plasmon resonance was carried out on a BIAcore 3000 biosensor (BIAcore) as described previously for other antigen antibody interactions (Lackmann et al., 1996). Monoclonal antibodies were immobilized onto BIAcore CM5 sensorchips using NHS chemistry and binding of ADAM10 proteins, diluted between 50 and 0.08 nM into running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% Tween 20), was performed on sensor-chip surfaces derivatised on parallel channels with non-relevant protein. The binding kinetics was derived from the plasmon resonance sensorgrams after subtraction of baseline responses (measured on the control channel) by global analysis using the BIA Evaluation software (version 3.02, BIAcore). The surface of the chip was regenerated after each injection of sample with 3 M MgCl₂, 0.075 M HEPES/NaOH, 25% ethylene glycol, pH 7.2, followed by two washes with running buffer.

HEK293 cells stably expressing, membrane-targeted GFP and EphB2, or EphB2ΔICD, were kindly provided by David Wilkinson and Alexei Poliakov (NIMR, London) and Claus Jorgensen (ICR, London) and were previously described (Janes et al., 2011; Jørgensen et al., 2009; Poliakov et al., 2008). EphA3/HEK293 cells were also previously described (Lawrenson et al., 2002). Mouse embryonic fibroblasts (MEFs) from Wt or ADAM10−/− mice (Hartmann et al., 2002) were kindly provided by Dr Saftig (Kiel). Cell lines were cultured in DMEM supplemented with 10% fetal calf serum and transiently transfected using Fugene HD (Roche). Metalloprotease inhibitors GM6001 and TAPI1 were from Calbiochem.

Cells lysed in buffer containing 1% Triton X-100 and 0.1% SDS (Lawrenson et al., 2002) were immunoprecipitated with antibodies against ADAM10 (R&D systems mAb 1427, or mAbs 3A8 or 8C7) or turboGFP (OriGene) followed by protein A–Sepharose, or EphA3 (mAb IIIA4 (Lackmann et al., 1996) conjugated to Minileak™ beads). Western blots were performed using rabbit polyclonal antibodies against ADAM10 (Abcam pAb 39177), turboGFP (OriGene) or PTyr (Invitrogen), or α-EphA3 sheep pAb (Nievergall et al., 2010) or α-EphB2 mouse mAb (R&D systems). Where indicated cells were stimulated with 1.5 µg/ml ephrin-A5-Fc [produced as described previously (Lawrenson et al., 2002)] pre-clustered with 0.75 µg/ml α-human IgG (Jackson Immunoresearch). Densitometric analysis was with ImageQuant software.

Antibodies or ephrin-A5-Fc proteins were coupled to Alexa dyes (488, 594, 647; Invitrogen) as per the manufacturer's instructions, and incubated with cells for 1 hour or as indicated followed by washing with PBS and fixing with 4% PFA in PBS for 30 mins at room temperature. Samples were mounted in Moviol and coverslipped. Confocal microscopy was on a Leica SP5 TCS SP5 II or Nikon C1 confocal microscope equipped with 405, 488, 561 and 633 nm lasers. Stripe and segregation assays were imaged on a Leica AF6000LX Live Cell Imaging workstation. Images were processed using Leica TCS, ImageJ and Adobe Photoshop and InDesign software.

Alexa⁶⁴⁷–ephrin-A5-Fc was bound to 24-well tissue culture plates (BD Falcon, USA) as follows: to increase binding capacity, plates were pretreated with nitrocellulose dissolved in methanol (Lemmon et al., 1989) under a laminar hood, followed by 100 µg/ml protein A/G (Pierce) treatment for 1 hour at 37°C. Unbound protein A/G was removed by two PBS washes. Alexa⁶⁴⁷–ephrin-A5-Fc was bound to the protein A/G layer at 5 µg/ml for 1 hour at 37°C followed by two PBS washes.

293/EphA3 cells and EphB2/GFP 293 cells were treated in suspension with 8C7 at the indicated concentrations, with or without pre-cross-linking with α-mouse IgG (Jackson Immunoresearch) at 4∶1 ratio for 1.5 hours. They were then added to the Alexa⁶⁴⁷–ephrin-A5-Fc coated plates and incubated for 30 min at 37°C. The cells adhered to the bottom of the plate were taken off with two vigorous PBS washes followed by another wash with 10 ml of PBS. Cells were mounted on to slides for analysis by confocal microscopy and fixed with 2% PFA/PBS for 20 min, or fixed in suspension with 2% PFA in FACS buffer (1% FCS, 1 mM EDTA in PBS) for flow cytometry. Flow cytometry analysis was performed using BD LSR II flow cytometer (BD Biosciences) under HeNe laser (18 mW at 633 nm). Further analysis was done on FlowJo 7.6.1 software package (Tree Star Inc., USA). Only single cells were gated for expression of Alexa⁶⁴⁷–ephrin-A5-Fc. Histograms were generated for every sample indicating the shift of fluorescence. Mean of the fluorescence distribution was calculated using FlowJo.

For assaying ephrin cleavage from cells, 293/EphA3 cells were stained with Cell Tracker Red (Invitrogen) as described in the manufacturer's protocol, allowed to adhere onto eight-well chamber slides (BD falcon), and pretreated with vehicle, 8C7, GM6001 or TAPI1 at the indicated concentrations. 293/ephrinA5 GFP cells were then added for 40 min at RT followed by fixing with 4% PFA, and nuclear staining with DAPI. Cells were mounted and imaged using a Nikon upright C1 confocal microscope and 100× oil objective. Ephrin internalisation was estimated from the Mander's coefficient of colocalisation of green (ephrin) with red (EphA3 cells).

A silicon matrix (obtained from Prof. Bastmeyer, Karlsruhe Institute of Technology, Germany) with channels was placed on a 22×22 mm glass coverslip, and protein A/G (Pierce, USA) solution (20 µg/ml) resuspended in Hanks Buffered Salt Solution (HBSS, Invitrogen) was applied to the channels as described (Knöll et al., 2007). Coverslips were then coated with fibronectin (Sigma) at 10 µg/ml in PBS. EphrinA5-Fc labelled with Alexa⁵⁹⁴ (Invitrogen) at 16 µg/ml in HBSS was added to the coverslips for 30 mins at 37°C and washed twice in HBSS. EphB2/GFP 293 cells treated with or without 8C7 mAb antibody or GM6001 at the indicated concentrations, or EphB2ΔICD 293 cells pre-labelled with Cell Tracker Green, were incubated on the coverslips at 37°C overnight before imaging by fluorescence microscopy. To determine the percentage of migrated cells in each image the number of cells on the stripes (counted manually) was taken as a percentage of the total number of cells (estimated using the image analysis software Cell-Profiler, Broad Institute).

EphB2-expressing HEK293 cells co-expressing membrane-targeted GFP (Poliakov et al., 2008), or U251 cells labelled with Cell Tracker Green (Invitrogen) were co-cultured with ephrin-expressing HEK293 cells in 96-well plates seeded with 20,000 cells per well, for 48 h or until confluent. Cells were fixed and nuclei stained with Hoechst before imaging on a Leica AF6000LX fluorescence microscope, using the tile scanning function to image the entire well. The images were then analysed with ImageJ software by thresholding areas of fluorescence intensity above that seen in monocultures. Cell clusters greater than a set area corresponding to roughly 40–50 cells (Solanas et al., 2011) were then counted by particle count of thresholded areas. Each treatment was performed on four replicate wells, and repeated in three or more experiments.

We thank Momchil Kolev (Sloan-Kettering), Chanley Chheang and Julien Robinson (Monash) for technical assistance, Monash Micro Imaging for microscopy assistance, and Kaye Wycherley, Monoclonal Antibody Facility, The Walter and Eliza Hall Institute of Medical Research, for invaluable help with generating monoclonal antibodies.