The RECK tumor-suppressor protein binds and stabilizes ADAMTS10

RECK was initially isolated as a transformation suppressor gene against the v-K-ras oncogene (Takahashi et al., 1998); subsequent studies implicated RECK in suppression of tumor growth, angiogenesis, invasion, metastasis and recurrence (Hill et al., 2011; Noda and Takahashi, 2007; Oh et al., 2001; Yoshida et al., 2012). RECK is conserved from insects to mammals as a single gene. Human RECK encodes a glycosylphosphatidylinositol (GPI)-anchored glycoprotein of ∼125 kDa with a weak matrix metalloproteinase (MMPs) inhibitory activity (Miki et al., 2007; Oh et al., 2001; Omura et al., 2009; Takahashi et al., 1998). Reck-deficient mice die around embryonic day 10.5 (E10.5) with reduced tissue integrity, arrested vascular development (Oh et al., 2001) and precocious neuronal differentiation (Muraguchi et al., 2007). Some of these phenotypes have been attributed to increased proteolysis (Oh et al., 2001), attenuated Notch-signaling (Muraguchi et al., 2007) and reduced WNT7-signaling (Noda et al., 2016; Vanhollebeke et al., 2015); however, it remains unclear to what extent these mechanisms explain the Reck mutant phenotypes. Tissue inhibitors of metalloproteinases (TIMPs) are known as a prototypic family of endogenous MMP inhibitors which are structurally unrelated to RECK. Mice lacking all TIMPs (quadruple mutant) are viable (Shimoda et al., 2014), demonstrating distinct physiological roles for TIMPs and RECK.

ADAMTS (a disintegrin-like and metalloprotease with thrombospondin motifs) superfamily of secreted proteins consists of 19 zinc metalloproteinases and 7 ADAMTS-like proteins lacking catalytic activity. A hallmark of ADAMTS proteases is an ancillary domain containing one or more thrombospondin type 1 repeats (TSR1). Known functions of ADAMTS proteases include maturation of procollagen (ADAMTS2, 3 and 14) and von Willebrand factor (ADAMTS13) as well as extracellular matrix (ECM) cleavage involved in morphogenesis (nematode ADAMTS proteins), angiogenesis (ADAMTS1 and 9), ovulation (ADAMTS1), cancer (ADAMTS4, 5, 9 and 15) and arthritis (ADAMTS5) (Apte, 2009; Hubmacher and Apte, 2015). The first biochemical study of ADAMTS10 by Somerville et al. (2004) indicated that the gene is widely expressed in mouse embryos particularly in mesenchymal cells, that zymogen as well as processed forms of ADAMTS10 are secreted from the cells, and that substantial amounts of these protein species are retained in the pericellular region by interaction with either the cell surface or pericellular matrix components. Most ADAMTS proteases contain a consensus furin recognition and cleavage sequence (R-X-R/K-R↓) at the junction of the propeptide and catalytic domain, but ADAMTS10 diverges at this site (i.e. GLKR²³³) and lacks a sequence required for efficient cleavage by furin and other members of the subtilisin/kexin-like proprotein convertase (PC) family (Duckert et al., 2004), making it likely that ADAMTS10 is resistant to activation by all PCs and activated by other means. The discoveries that a rare connective tissue disorder, Weill-Marchesani syndrome, was genetically linked with recessive mutations of ADAMTS10 (WMS1; OMIM #277600) (Dagoneau et al., 2004) as well as dominant mutations of fibrillin-1 (WMS2; OMIM #608328) (Faivre et al., 2003) suggested that ADAMTS10 and fibrillin-1 (FBN1) might interact and/or act cooperatively in a common pathway.

In the present study, we performed yeast two-hybrid screening to find clues to the additional molecular functions of RECK, and this led to the detection of ADAMTS10 as a RECK-interactor. Experiments with purified recombinant proteins and in cultured cells also yielded evidence indicating physical and functional interactions between these molecules of known, but distinct, clinical importance.

In an attempt to find novel binding partners for RECK, we used the yeast two-hybrid assay, with near-full-length RECK as bait, to screen a mouse embryo cDNA library. Among the 14 positive clones isolated and sequenced (Table S1), two independent clones (clones 2 and 10) were found to encode overlapping, COOH-terminal portions (129 and 251 amino acid residues, respectively) of ADAMTS10 (see the bottom of Fig. 1B for their positions).

To map the domains responsible for this interaction, we constructed a series of deletion mutants of both RECK (Fig. 1A, Rd1-9) and ADAMTS10 (Fig. 1B, Ad1-6) and tested their interactions again using the yeast two-hybrid assay. When the near full-length ADAMTS10 (AF: residues 25-1104) was used as bait (Fig. 1A), the RECK fragment Rd7 (residues 349-629) was strongly positive, but full length RECK (RF) and the other RECK fragments were negative under these stringent conditions: namely, LTAH (−)+10 mM 3AT (see Materials and Methods for details). When the near full-length RECK (RF: residues 27-942) was used as bait (Fig. 1B), ADAMTS10 fragment Ad4 (residues 823-1104) was strongly positive and Ad3 (residues 691-1104) was also clearly positive. While in the settings shown in Fig. 1A and B, the interaction between AF and RF was undetectable (Fig. 1A,B, top row in the tables), under less stringent conditions (condition III in Fig. 1C) their interaction was readily detected when we used RF as bait and AF as prey (Fig. 1C, row 2). Likewise, the interaction between two small fragments, Ad4 (bait) and Rd7 (prey), could be detected (Fig. 1C, row 6). These results implicate a central region of RECK and a C-terminal region of ADAMTS10 in this interaction (Fig. 1D).

To verify the interaction more directly, we expressed a full-length ADAMTS10 protein tagged with C-terminal tandem MYC and His₆ epitopes in the mammalian cultured cell line HEK293 and purified this protein, termed ATS10-MH, from the culture supernatant (Fig. S1A,C). We also raised antibodies (termed pAb-867) against an epitope (residues 867-879) in the second TSR1 (Fig. S1A, arrow); they recognize both mouse and human ADAMTS10. Using these reagents, together with a recombinant RECK protein previously described (termed RECK-His; Fig. S1B,D) (Omura et al., 2009), we could detect the direct binding between ATS10-MH and RECK-His in a solid-phase binding assay in the presence of three cations: Ca²⁺, Mg²⁺ and Zn²⁺ (Fig. 1E). Surface plasmon resonance assay was also used to confirm this binding and to reveal that the binding requires Zn²⁺ (Fig. S1E,F).

Totally separate subcellular localization may preclude the physical interaction between RECK and ADAMTS10 at the cellular level. To test this possibility, we performed two types of double-labeling experiments. First, immunofluorescence staining of endogenous RECK (green) and ADAMTS10 (red) in mouse dermal fibroblasts (MDFs), derived from wild-type mouse at postnatal day 3 (P3), yielded yellow signals in the merged images (suggesting co-localization) that were relatively rich in the peripheral and peri-nuclear regions (Fig. 2A, panel 4; Movie S1); this pattern is reminiscent of the RECK subcellular localization previously observed in NIH3T3 cells (Morioka et al., 2009). We also noted substantial heterogeneity in the levels of RECK expression among these MDFs (Fig. 2A, panel 2; note the green signals abundant in two of the several cells in this field). This approach, however, depends on the specificity of antibodies used (see Materials and Methods). We therefore took the second approach of co-expressing two proteins tagged with different fluorescent proteins, mRFP-RECK (red) and GFP-ADAMTS10 (green), in a human osteosarcoma cell line (MG-63) followed by confocal microscopy. This approach again yielded yellow signals rich in the peri-nuclear region (Fig. 2B; Movie S2).

We also performed in situ proximity ligation assays (PLA), which can detect close (<40 nm) co-localization between two molecules (Söderberg et al., 2006). PLA demonstrated close co-localization between endogenous RECK and ADAMTS10 in a mouse fibroblast-derived cell line, NIH3T3 (Fig. 2C,D). These results support the idea that RECK and ADAMTS10 may have chances to encounter and interact with each other under physiological conditions.

ADAMTS10 is a protease (Apte, 2009; Hubmacher and Apte, 2015); RECK has been studied as a protease inhibitor (Miki et al., 2007; Oh et al., 2001; Omura et al., 2009; Takahashi et al., 1998). We therefore performed simple mix-and-incubation experiments (Fig. 3A) to ask two obvious questions: (i) could ADAMTS10 cleave RECK? (ii) Could RECK inhibit ADAMTS10? In this experiment, we tried to activate pro-ADAMTS10 using p-aminophenyl mercuric acetate (APMA), a chemical known to activate several other members of the metazincin family.

When RECK-His (125 kDa) was incubated alone or in the presence of different amounts of ATS10-MH in buffer containing APMA for 21 h and then visualized by immunoblot assay, a dose-dependent increase in the density of ∼30 kDa fragment of RECK (Fig. 3B, arrowheads) was detected (Fig. 3D), although the decrease in full-length RECK band was not so prominent, suggesting that ADAMTS10 has a weak activity to cleave RECK.

A previous study by Kutz and colleagues (Kutz et al., 2011) also demonstrated that recombinant human ADAMTS10 gave rise to multiple bands in immunoblot assays and this phenomenon required its protease catalytic site (E393), suggesting that the phenomenon represented autocatalytic fragmentation. ATS10-MH (derived from mouse cDNA) shows two groups of prominent bands: one corresponding to its pro-form and the other close to the size of the mature form (I and II in Fig. 3C). When the amount of ATS10-MH was increased, the overall density of immunoblot bands increased up to a certain input protein concentration but did not increase at a higher concentration (Fig. 3C, compare lanes 3, 5 and 7; densitometric data in Fig. 3E, bars 3, 5 and 7). At the highest ATS10-MH concentration, the density of bands I and II decreased (Fig. 3E, bar 7 versus 5 in red and green) while the density of the smaller fragments increased (Fig. 3E, bars 7 versus 5 in purple), suggesting that ATS10-MH underwent dose-dependent autocatalytic fragmentation (illustrated by the downward arrow in Fig. 3F). In the presence of RECK-His, on the other hand, the density of ATS10-MH bands increased (Fig. 3C, compare lanes 2, 4 and 6 to lanes 3, 5 and 7; Fig. 3E, bars 2, 4 and 6 versus bars 3, 5 and 7), suggesting that RECK-His stabilizes ATS10-MH under these conditions (illustrated by the upward arrow in Fig. 3F). Similar results were obtained in another independent experiment (Fig. S3). The decrease in ATS10-MH band density at higher concentrations and the effect of RECK-His to yield higher ATS10-MH band density were clear only when APMA was added (Fig. S4B, lane 6 versus 3, lane 5 versus 6) and undetectable at an earlier time point (1.5 h; Fig. S4C, lane 5 versus 7, lanes 4 and 6 versus 5 and 7).

Taken together, these results indicate that ATS10-MH has a weak activity to cleave RECK-His and support the idea that RECK-His inhibits the autocatalytic fragmentation of ATS10-MH (illustrated in Fig. 3G).

We next asked whether ATS10-MH affects the known function of RECK to inhibit MMPs by mix-and-incubation experiments with four purified factors (see Fig. 4A) in the absence of APMA. Furin-activated MT1-MMP cleaves fibronectin (FN) into smaller fragments as expected (Sternlicht and Werb, 2001) (Fig. 4B, lane 3 versus 1), and this reaction is inhibited by RECK-His (Fig. 4B, compare the bracketed region of lanes 6 and 3; Fig. 4C, 6 versus 3). In this system, ATS10-MH suppresses the effect of RECK-His (Fig. 4B, compare the bracketed region of lanes 5 and 6; Fig. 4C, lane 5 versus 6). One possibility is that ATS10-MH cleaves and reduces the level of intact RECK-His, allowing MT1-MMP to freely work again; however, RECK-His seems to be largely intact under these conditions (Fig. 4F, lane 5 versus 6). A more feasible model is that ATS10-MH binds RECK-His thereby interfering with the interaction between MT-MMP and RECK-His. This experiment also demonstrated the activity of MT1-MMP to efficiently cleave ATS10-MH (Fig. 4D, lane 4 versus 2). These findings are summarized in Fig. 4G.

In an attempt to confirm the biological relevance of the above findings using recombinant proteins, we prepared MDFs from mice with four different RECK expression levels: 100% (+/+), ∼70% (Low/+), ∼50% (+/Δ) and ∼20% (Low/Δ) (see Fig. S5A,B). Note that among these mice, conspicuous developmental abnormalities are found only in Low/Δ mice (Yamamoto et al., 2012) but more subtle phenotypes, especially after damage, have been found in +/Δ mice (Gutierrez et al., 2015; Wang et al., 2010). Confluent cultures of these cells (three animals per genotype) were lysed and subjected to immunoblot assay to detect RECK, ADAMTS10 and α-tubulin (Fig. 5A). The level of RECK in the cell lysates decreased in agreement with their genotypes (Fig. 5A, panel 2; densitometry data in Fig. 5B, black bars). When the level of cellular RECK protein was low (≤50%: +/Δ and Low/Δ), the level of ADAMTS10 associated with the cells was significantly reduced (Fig. 5B, white bars). Reduction of ADAMTS10 associated with MDFs derived from Low/Δ mice (P3 and P5) could also be detected by immunofluorescence staining (Fig. 5C,D). The level of ADAMTS10 associated with BE6, a RECK-deficient MG-63 clone, was significantly lower and diffuse compared to the parental line (Fig. 5E,F).

Taken together, our data indicate that RECK interacts with ADAMTS10 and increases the level of ADAMTS10 associated with the cells, possibly by protecting ADAMTS10 from degradation (Fig. 5G).

Compelling evidence from experimental studies as well as analyses of clinical specimens implicate RECK in tumor suppression (Chung et al., 2012; Fakhry et al., 2016; Hill et al., 2011; Noda and Takahashi, 2007; Yoshida et al., 2012). Previous biochemical evidence led to the idea that competitive inhibition of MMPs may be the major molecular basis of RECK's biological functions. The affinity of RECK for MMPs (Ki∼10⁻⁸) (Miki et al., 2007; Oh et al., 2001; Omura et al., 2009; Takahashi et al., 1998), however, is lower than that of the prototypic endogenous MMP inhibitors, TIMPs (Ki: 10⁻⁹-10⁻¹⁶ M) (Bourboulia and Stetler-Stevenson, 2010; Butler et al., 1999; Murphy, 2011), raising the possibility that RECK may have additional molecular functions. Our attempts to find novel RECK-interactors using the unbiased approach of yeast two-hybrid screening led to the detection of another new target, ADAMTS10, a member of a metalloproteinase family distinct from MMPs.

Our experiments using recombinant proteins (i.e. ATS10-MH) demonstrated dose-dependent auto-degradation of ADAMTS10 (see Fig. 3). We speculate that the large variability in its RECK-binding activity observed at higher dosages in a solid-phase binding assay (e.g. Fig. 1E, 0.5 unit) might reflect such auto-degradation. In immunoblot assays, our preparations of ATS10-MH consistently yielded two sets of major bands of ∼138 kDa (band I) and ∼123 kDa (band II) (see Fig. 3C), and band II is likely to contain activated ADAMTS10 (Somerville et al., 2004). Of note, RECK-His not only increases the intensity of ATS10-MH but also increases the relative intensity of band II to band I (see Fig. 3C; e.g. compare lane 6 with lane 7), raising the possibility that RECK may help generate and selectively stabilize the active form of ADAMTS10. Such cooperative interaction is in sharp contrast to the effect of RECK on MMPs (i.e. negative regulation) and is consistent with our observation that RECK and ADAMTS10 promote fibrillin microfibril formation in a cooperative manner in cultured cells (T. Matsuzaki et al., unpublished).

Yeast two-hybrid assays indicated that RECK has the potential to bind to a C-terminal portion, rather than the catalytic domain, of ADAMTS10 (Fig. 1B–D). Ihara and Nishiwaki reported that a nematode ADAMTS protein, MIG-17, undergoes autocatalytic activation which requires its C-terminal PLAC domain and suggested that this domain might interact with the metalloproteinase domain to promote autocatalytic activation (Ihara and Nishiwaki, 2007). Of note, the PLAC domain of MIG-17 shows the highest similarity to that of mammalian ADAMTS10 among all family members (Ihara and Nishiwaki, 2007). We therefore speculate that RECK may facilitate this interaction while inhibiting the auto-degradation of once activated ADAMTS10. The reason why this interaction requires Zn²⁺ is presently unclear. Metalloproteinases are known to coordinate Zn²⁺ in their catalytic domain, and therefore Zn²⁺ may well have allosteric effects on other domains. Alternatively, Zn²⁺ may be required for this particular protein-protein interaction more directly.

Some of our data demonstrated a weak, APMA-dependent activity of ADAMTS10 to cleave RECK in a dose-dependent manner (Fig. 3B). An activity of ADAMTS10 to suppress the MMP-inhibiting activity of RECK was also found (Fig. 4B,C). In these respects, ADAMTS10 may be considered as a regulator of RECK's functions.

Previous studies by Apte and colleagues (Kutz et al., 2008; Somerville et al., 2004) raised several important questions about the molecular nature of ADAMTS10, which include (1) why ADAMTS10 undergoes extensive fragmentation, (2) why a large fraction of ADAMTS10 forms aggregates in the culture supernatant, and (3) why its pro-domain cannot be efficiently cleaved off by furin. These questions are based mainly on the observations made using the epithelial cell line HEK293F which expresses a minimal level of RECK protein. Our findings that MT1-MMP efficiently cleaves pro-ADAMTS10 yielding a fragment of about the size of its active form (see Fig. 4D) and that RECK influences the fragmentation of activated ADAMTS10 (see Fig. 3C,E–G) shed some new light on these questions. In mouse embryos, RECK expression is relatively high in mesenchymal cells in which ADAMTS10 is also expressed abundantly (Somerville et al., 2004). The fate of ADAMTS10 co-produced with RECK by mesenchymal cells is likely to differ from that of ADAMTS10 produced by HEK293F cells. Whether such difference has to do with the reported differences in the modes of microfibril formation between mesenchymal and epithelial cells (Baldwin et al., 2014; Cain et al., 2016) is an interesting issue to be addressed in future studies.

In summary, this study has uncovered the novel ability of RECK to interact with ADAMTS10, an enzyme known to be critical for connective tissue development, and to increase the amount of ADAMTS10 associated with the cell. This interaction may also affect the function of RECK itself, a tumor suppressor downregulated in a wide variety of cancers.

Rabbit polyclonal anti-ADAMTS10 antibodies (pAb-867) were raised against a synthetic oligopeptide (SGHSKLPKRQRAC; in the second TSR domain). Other primary antibodies: mouse monoclonal antibodies against RECK (5B11D12) (Takahashi et al., 1998), FN (610078, BD), MT1-MMP (ab51074; Abcam) and α-tubulin (DM1A; Calbiochem/Merck Millipore, Burlington, USA). Secondary antibodies: anti-rabbit IgG-HRP (ab6721; Abcam), anti-mouse IgG-HRP (A4416; Sigma-Aldrich), anti-mouse IgG-CF488 (20018; Biotium, Fremont, USA), anti-mouse IgG-CF405M (20182, Biotium) and anti-rabbit IgG-CF647 (20282: Biotium). Antibody dilution: primary 1:1000 and secondary 1:20000 for immunoblot assays, primary 1:300 and secondary 1:1000 for immunofluorescence staining. Specificity of pAb-867 and 5B11D12 in immunofluorescence staining was assessed in MG-63 cells genetically manipulated to suppress (ADAMTS10) or inactivate (RECK) the expression of endogenous target genes (see Fig. S2).

The Reck mutant allele, Δ (also known as Reck^tm2.2Noda) and Low (also known as Reck^tm1.2Noda), have been described (de Almeida et al., 2015; Chandana et al., 2010; Yamamoto et al., 2012) and were on the Mus musculus C57BL/6 background. The animal experiments were approved by the Animal Experimentation Committee, Kyoto University and conducted in accordance with its regulations.

A mouse embryo (E12, C57BL/6) cDNA library (2×10⁶ clones) was generated using the pGADT7 vector. Lambda Zap II-based cDNA libraries were also generated using the same cDNA pool and adult mouse brain cDNAs. A fragment of mouse Reck cDNA (codons 27-942; termed RF) was amplified by PCR (Table S2A), digested with EcoRI and SalI and inserted into pGBKT7 to obtain pGBKT7-mRECK. A full-length Adamts10 cDNA was re-screened from a Lambda Zap II-based library; its upstream sequence (gcccctATGgct) was modified to make it better fit with the Kozak sequence (gaattcgccATGgct), and it was inserted between the EcoRI and XbaI sites of the pEF6/Myc-His A vector to obtain pEF6A-mAdamts10. To generate the vector expressing ATS10-MH, pEF6A-mAdamts10 was modified so that the termination codon of Adamts10 was replaced in frame with the sequence encoding the MYC-His₆ epitope tag present in the pEF6/Myc-His A vector. For protein production, the same coding sequence was inserted into pTet-splice (Tet-off) to obtain pTet-mAdamts10MycHis. For the yeast two-hybrid assay, an Adamts10 cDNA (codons 25-1104; termed AF) was amplified by PCR (Table S2A), digested with EcoR1 and Sal I and inserted into pGBKT7 to obtain pGBKT7-mAdamts10. To generate cDNAs encoding deletion mutants of RECK or ADAMTS10, the target sequences were amplified by PCR and inserted into pGADT7 (Rd1-9 and Ad1-6). Sources of vectors and hosts: pGBKT7 and pGADT7 (Clontech, Palo Alto, USA), pEF6/Myc-His A (Invitrogen), lambda Zap II (Stratagene, La Jolla, USA), pTet-splice (GIBCO) and the yeast strain AH109 (Clontech). pLXSB-hRECK has been described (Morioka et al., 2009; Sasahara et al., 1999). To express EGFP-tagged ADAMTS10 protein in mammalian cells, the Ad1 cDNA was inserted into pEGFP-C1 (Clontech) containing the RECK signal peptide sequence (Morioka et al., 2009) to obtain p-ss-EGFP-Adamts10. pmRFP-RECK was also described previously (Morioka et al., 2009).

AH109 cells were transformed with pGBKT7-mRECK and selected on SD plates without tryptophan to obtain the indicator strain AH109-mRECK. To screen RECK-interactors (Table S1), AH109-mRECK cells were transformed with the pGADT7-based cDNA library and selected on SD plates without leucine, tryptophan, adenine, or histidine (-LTAH) and with 10 mM 3-amino-1,2,4-triazole (3AT; an inhibitor of histidine biosynthesis which increases the stringency of selection in a dose-dependent manner) for 11 days. For binding assays, the AH109 cells harboring pGBKT7-mAdamts10 or pGBKT7-Ad4 were transformed with pGADT7-RECK (RF) or its deletion mutants (Rd1-9), and their growth on SD (-LTAH) plates with 3AT at different concentrations (0, 2, 5, 10) was examined after incubation for 4 days (Fig. 1A,C). Similarly, AH109-mRECK cells were transformed with pGADT7-mAdamts10 (AF) or its deletion mutants (Ad1-6), and their growth on SD (-LTAH) plates with or without 3AT was examined.

HEK293, MG-63 (ATCC) and NIH3T3 cells (a gift from Dr Robert H. Bassin) were maintained in growth medium (GM) consisting of DMEM supplemented with 10% fetal bovine serum (FBS), 100 µg/ml streptomycin sulfate and 100 U/ml penicillin G. MDFs were prepared from mouse-back skin (P3-P5) using the cold trypsin method (Freshney, 1987). Transfection was performed using Lipofectamine 2000 (Invitrogen), FuGENE 6 (Roche) or CalPhos Mammalian Transfection Kit (Clontech).

HEK293 Tet-off Advanced (Clontech) was stably co-transfected with pTet-mAdamts10MycHis and pEF6/Myc-His A, selected with 8 µg/ml blasticidin in GM (selection medium), and a clone abundantly producing ATS10-MH was maintained in selection medium containing 2 μg/ml doxycycline. Subconfluent cultures of these cells were washed with PBS and incubated in CD293 medium for 5 days. Culture supernatant was centrifuged to remove cellular debris, dialyzed against PBS, adjusted to 0.5 M NaCl and applied to a His-Trap HP column (GE) that was equilibrated with 20 mM imidazol. After washing with 40 mM imidazole, proteins were eluted using a step-gradient of imidazole. The 60 mM imidazol fraction rich in ATS10-MH was used in the solid-phase binding assay after replacing its solvent with 20 mM Hepes (pH7.2), 150 mM NaCl by ultrafiltration (AmiconUltra 50K; Millipore). The 250 mM imidazole fraction was dialyzed against 0.15 M NaCl/20 mM KPO₄ (pH 8.0) and applied to a cation-exchange column (RESOURCE S, GE). The flow-through fraction was applied to an anion-exchange column (RESOURCE Q, GE); the column was washed with 20 mM KPO₄ (pH 8.0) containing 310 mM NaCl, and the bound proteins were eluted sequentially using buffer containing 400 mM NaCl and 1 M NaCl, respectively. The 400 mM NaCl fraction was used for the protein-mixing assays shown in Fig. 4 and Fig. S4B. This fraction was further applied to a gel filtration column (Hiload 26/60) to eliminate small molecules and used for surface plasmon resonance assay. The 1 M NaCl fraction was dialyzed against 20 mM Hepes (pH 7.0)/180 mM NaCl and applied to a heparin HP column (GE), and the flow-through fraction was used for the protein-mixing assays shown in Fig. 3 and Fig. S4D.

Purified RECK-His protein (Omura et al., 2009) (100 ng/170 µl/well) was immobilized onto 96-well Maxisorp Plates (Nunc/Thermo Fisher Scientific, Waltham, USA) with 15 mM Na₂CO₃, 35 mM NaHCO₃, 6 mM azide. The wells were blocked with 1% bovine serum albumin (200 µl), washed with PBS twice, and then with 20 mM Hepes pH7.4/150 mM NaCl (HBS). ATS10-MH (after His-Trap purification from 400 ml culture supernatant which yielded 3.2 ml buffer-exchanged sample – see the section ‘Recombinant ATS10-MH protein’ for details – 1 unit in Fig. 1E correspond to undiluted sample) serially diluted in HBS containing 2 mM CaCl₂, 2 mM MgCl₂, 2 nM ZnCl₂ was added to the wells (170 µl/well) and incubated overnight at 4°C. After extensive washing with PBS containing 0.0005% Tween 20 (WB), ATS10-MH bound to the well was labeled with anti-ADAMTS10 and then with HRP-conjugated anti-rabbit IgG (Cell Signaling Technology), and detected using chemiluminescent HRP substrate (Immobilon Substrate, WBKLS0500, Millipore) followed by luminometry (MicroLumatPlus, PerkinElmer, Waltham, USA).

For immunofluorescence staining, wild-type MDFs were fixed in acetone at −80°C and stained with anti-ADAMTS10 (pAb-867; CF647) (Fig. 5C,D) or double-stained with anti-ADAMTS10 (pAb-867; CF647) and anti-RECK (5B; CF488) (Fig. 2A). For co-expression of fluorescent-protein-tagged proteins (Fig. 2B), MG-63 cells on 2-well chamber slides (SCS-002, Matsunami, Osaka, Japan) were co-transfected with p-ss-EGFP-Adamts10 (1 µg/well) and pmRFP-RECK (32) (1 µg/well) using Lipofectamine2000 and incubated for 2 days. Fluorescence signals were recorded using a confocal microscope (Leica sp8).

Confluent NIH3T3 cells in 8-well chamber slides (8CS; SCS-008, Matsunami) were fixed with IHC Zinc Fixative (BD Biosciences) for 4.5 h and subjected to PLA using Duolink Green (Sigma-Aldrich) with anti-RECK (5B) and/or anti-ADAMTS10 (pAb-867) as primary antibodies and anti-Mouse PLUS and anti-Rabbit MINUS as secondary antibodies. PLA signals (green) and the number of nuclei (DAPI-stained) were quantified using ImageJ (NIH).

Immunoblot assay was performed as described (Oh et al., 2001) using α-tubulin as a control. Cell lysate was prepared using lysis buffer containing 50 mM Tris pH7.5, 0.5 M NaCl, 1% NP40, 0.1% SDS, 0.5% deoxycholate, 5 mM N-ethylmaleimide and 20×protease inhibitor cocktail (03969, Nakarai, Kyoto, Japan). Samples were frozen at −80°C and just before SDS-PAGE, boiled for 10 min in Laemmli sample buffer (1×) containing 83 mM DTT. Images were captured using LAS3000 (Fujifilm, Tokyo, Japan). Densitometry was carried out using MultiGauge v3.2 (Fujifilm) or ImageJ (NIH).

To test the effects of RECK and ADAMTS10 on each other (Fig. 3, Figs S3 and S4), purified RECK-His (0.36 pmol), ATS10-MH (multiple concentrations; ‘1’=0.72 pmol), or both in 20 µl Buffer E [150 mM NaCl, 5 mM CaCl₂, 2.5 mM MgCl₂, 0.5 mM ZnCl₂, 0.05% Brij-35, 20 mM Tris.HCl (pH 7.0), with or without 1 mM APMA] were incubated at 37°C for the indicated period of time. To test the effects on FN (Fig. 4; Fig. S4), mature MT1-MMP (0.31 pmol), RECK-His (0.48 pmol) and/or ATS10-MH (0.43 pmol) were mixed in various combinations with FN (0.87 pmol) and incubated at 37°C for the indicated period of time in 40 µl Buffer E. The reaction mixture was separated by SDS-PAGE under non-reducing or reducing conditions. To prepare mature MT1-MMP, recombinant MT1-MMP (Millipore; CC1043, 230 ng) was pre-treated for 1 h with 20 ng furin in 62 µl Buffer E.

Significance of difference between two groups of data was evaluated using Student's t-test.

We thank Aiko Nishimoto, Hai-Ou Gu, Mika Fujiwara and Kumi Kawade for technical assistance and Aki Miyazaki for administrative assistance.