A disintegrin-like and metalloproteinase domain with thrombospondin type 1 motif 9 (ADAMTS9) regulates fibronectin fibrillogenesis and turnover

The secreted metalloprotease ADAMTS9 has dual roles in extracellular matrix (ECM) turnover and biogenesis of the primary cilium during mouse embryogenesis. Its gene locus is associated with several human traits and disorders, but ADAMTS9 has few known interacting partners or confirmed substrates. Here, using a yeast two-hybrid screen for proteins interacting with its C-terminal Gon1 domain, we identified three putative ADAMTS9-binding regions in the ECM glycoprotein fibronectin. Using solid-phase binding assays and surface plasmon resonance experiments with purified proteins, we demonstrate that ADAMTS9 and fibronectin interact. ADAMTS9 constructs, including those lacking Gon1, co-localized with fibronectin fibrils formed by cultured fibroblasts lacking fibrillin-1, which co-localizes with fibronectin and binds several ADAMTSs. We observed no fibrillar ADAMTS9 staining after blockade of fibroblast fibronectin fibrillogenesis with a peptide based on the functional upstream domain of a Staphylococcus aureus adhesin. These findings indicate that ADAMTS9 binds fibronectin dimers and fibrils directly through multiple sites in both molecules. Proteolytically active ADAMTS9, but not a catalytically inactive variant, disrupted fibronectin fibril networks formed by fibroblasts in vitro, and ADAMTS9-deficient RPE1 cells assembled a robust fibronectin fibril network, unlike WT cells. Targeted LC-MS analysis of fibronectin digested by ADAMTS9-expressing cells identified a semitryptic peptide arising from cleavage at Gly2196–Leu2197. We noted that this scissile bond is in the linker between fibronectin modules III17 and I10, a region targeted also by other proteases. These findings, along with stronger fibronectin staining previously observed in Adamts9 mutant embryos, suggest that ADAMTS9 contributes to fibronectin turnover during ECM remodeling.

An extracellular matrix (ECM) 3 is a complex network of proteins, glycoproteins, proteoglycans, and glycosaminoglycans that ensures tissue structural integrity. It also provides positional information and mechanical signals to cells and sequesters growth factors and cytokines (1,2). Because nearly all signals perceived by cells are potentially modifiable by the ECM, a highly regulated ECM composition and organization are necessary for correct cell and organ function. ECM is continuously remodeled by coordinated biosynthesis and proteolysis of its components (3). Among several enzymes participating in ECM proteolysis are ADAMTS proteases, a family of 19 secreted zinc proteases with diverse and essential roles in mammalian development and homeostatic processes (reviewed in Ref. 4 -6). Naturally occurring mutations affecting ADAMTS proteases cause a variety of Mendelian disorders in humans and other species (4), and ADAMTS proteolytic activity may drive the pathology of acquired disorders such as osteoarthritis and coronary artery disease (7,8).
ADAMTS proteases comprise an N-terminal pro-protease domain upstream of a series of canonical modules constituting an ancillary domain, whose hallmark is the presence of one or several thrombospondin type 1 repeats (TSRs) (9). ADAMTS9 and its homolog ADAMTS20 each contain 15 TSRs and a unique C-terminal module named Gon1 after their Caenorhabditis elegans ortholog Gon1 (10,11). They have high domain and sequence homology with Gon1, which is necessary for gonadal morphogenesis (10), as well as with the Drosophila melanogaster metalloprotease ADAMTS-A, which is essential for tracheal development (12). Thus, these proteases are evolutionarily conserved and likely to have significant functions in mammals. Indeed, genetically engineered mouse mutants have shown that ADAMTS9 is indispensable for several aspects of mammalian embryogenesis. Adamts9-null mouse embryos fail to undergo gastrulation, a crucial early morphogenetic event during which the definitive germ layers are formed, and die by 7 days of gestation (13). This mouse allele, and others, including a hypomorphic Adamts9 allele (Adamts9 Gt ), chemically induced point mutations, and conditional Adamts9 deletion have identified crucial functions in craniofacial, neural, vascular, cardiac, eye, pigment, and limb development (14 -19). In adult mice, ADAMTS9 has a role in parturition via regulation of ECM control of uterine smooth muscle cell differentiation (20). ADAMTS9 and ADAMTS20 were recently found to be essential for biogenesis of the primary cilium, providing an unexpected role for these secreted proteases in formation of a cellular organelle (21, 33). Genome-wide analysis has linked the ADAMTS9 locus to diverse genetic traits, disease susceptibility markers, and common disorders of complex etiology, including type 2 diabetes mellitus, obesity, and age-related macular degeneration (22)(23)(24)(25)(26). ADAMTS9 was identified as a tumorsuppressor gene in gastric, nasopharyngeal, and esophageal squamous cell carcinoma, where it may act via suppression of angiogenesis (27)(28)(29). Despite this considerable biological and disease relevance, the only known ADAMTS9 substrates to date are the ECM proteoglycans aggrecan and versican (11).
Previous work has demonstrated that ADAMTS9 enters the secretory pathway, where it undergoes obligate glycosylation, binds therein to several chaperones, and arrives at the cell surface, where its propeptide is cleaved by furin (11, 13, 14, 30 -32). Recent work shows that furin-processed, catalytically active ADAMTS9 is endocytosed and trafficked to Rab11 vesicles that form a girdle around the basal body of the primary cilium (33). There, through an as-yet unknown mechanism, ADAMTS9 mediates ciliary growth. Here, adding to the growing understanding of this important protease, we report identification of the ECM and circulating glycoprotein fibronectin as a novel ADAMTS9-binding partner and show that its assembly and abundance are reduced in the presence of ADAMTS9. These findings are consistent with stronger fibronectin staining previously observed in mice heterozygous for Adamts9 and mouse embryos homozygous for the loss-of-function Adamts9 Gt mutation (14,19,33).

Fibronectin is a novel binding partner of the ADAMTS9 Gon1 domain
A yeast two-hybrid screen of a human placenta library of Gal4 fusion proteins was undertaken using the human ADAMTS9 Gon1 module (residues 1734 -1935, SwissProt accession no. Q9P2N4) (Fig. 1A) as the "bait," fused in-frame and downstream of the Gal4 DNA-binding domain. In addition to several other secreted or cell-surface proteins, which could be putative binding partners, eight independent ADAMTS9 Gon1-interacting cDNA clones, each encoding part of fibronectin (Fig. 1B), were identified in-frame with the Gal4 DNA-activating domain. The selected interacting domains (SIDs), of fibronectin, which are minimum interacting regions determined by the overlap between these clones (34), are shown in Fig. 1B. The three SIDs, SID1, SID2, and SID3, spanned three nonoverlapping regions of fibronectin comprising both type I and type III repeats (Fig. 1B).
Yeast two-hybrid interactions occur in a reducing environment between proteins that are synthesized without posttranslational modifications typical for secreted mammalian proteins. Such modifications include introduction of disulfide bonds and N-glycosylation. Two disulfide bonds stabilize fibronectin type I modules (35). The human ADAMTS9 Gon1 module contains 10 cysteines (11), which are conserved in other Gon1 modules (http://pfam.xfam.org/family/PF08685) 4 and very likely form disulfides, and two sites of potential N-glycosylation. Therefore, experiments were performed in follow-up of the yeast two-hybrid results asking whether the ADAMTS9fibronectin interactions could be replicated with proteins that had been processed and undergone modification in a eukaryotic secretion pathway. Because we were unable to purify fulllength ADAMTS9, myc-His 6 -tagged ADAMTS9 Gon1 module and longer C-terminal ADAMTS9 constructs with exogenous signal peptides were expressed and purified from the medium of mammalian cells (Fig. 1, A and C). Recombinant ADAMTS9 Gon1 bound to plasma fibronectin, a disulfide-linked dimer of 220-kDa subunits in a solid-phase binding assay ( Fig. 2A), as well as to N-and C-terminally truncated fibronectin constructs N-1F(III) and 1F(III)-C, respectively (Fig. 2B). These overlapping constructs were expressed as secreted proteins using recombinant baculoviruses to ensure proper disulfide bonds and represent approximately the N-terminal one-third and the C-terminal two-thirds of fibronectin, respectively (Fig. 1B).
In solid-phase binding assays, ADAMTS9 Gon1 showed stronger binding to fibronectin than ADAMTS9 TSR9 -15 ( Fig.  2C), whereas ADAMTS9 TSR2-8 did not bind (data not shown). Purified ADAMTS9 TSR9-Gon1 construct containing both binding regions was unstable and therefore could not be used in the binding assays. ADAMTS9 Gon1 and ADAMTS9 TSR9 -15 binding to fibronectin was validated and compared using surface plasmon resonance (Fig. 2D). These constructs bound fibronectin with an estimated K D of 2.8 nM ( 2 /R max ϭ 4%) and 104 nM ( 2 /R max ϭ 6%), respectively. Thus, binding of the Gon1 module to different parts of fibronectin as identified in the yeast-two-hybrid screen was replicated using proteins purified after secretion from eukaryotic cells, and the TSR9 -15 region of ADAMTS9 was also found to contain one or more binding sites for fibronectin.

ADAMTS9 constructs bind fibronectin fibrils
Many cell types that secrete fibronectin or are provided exogenous fibronectin assemble it into fibrils, starting within a few hours of attachment if plated at sufficiently high density (36). Fibronectin fibrils provide a template for assembly of fibrillin-1 microfibrils, which appear soon after initial fibronectin assembly and co-localize extensively with them (37,38). Furthermore fibrillin-1, the major fibrillin produced by most cells, is a binding partner for several ADAMTS proteins (39). Therefore, to eliminate fibrillin-1 as a potential confounder in the analysis, we used Fbn1 Ϫ/Ϫ mouse embryo fibroblasts. In the short-term cultures used for these experiments, Fbn1 Ϫ/Ϫ mouse embryo fibroblasts do not assemble fibrillin-2, an ADAMTS ligand and substrate (40 -42).
To test whether the ADAMTS9 Gon1 module bound fibronectin fibrils, Fbn1 Ϫ/Ϫ fibroblast monolayer cultures were co-stained with anti-myc (for the exogenous myc-His 6 -tagged Gon1 domain) and an anti-fibronectin antibody. Despite interacting with full-length fibronectin dimer and fragments of fibronectin monomer (Fig. 2, A, C, and D), recombinant ADAMTS9 Gon1 module did not co-localize with fibronectin fibrils (data not shown). ADAMTS9 constructs containing upstream regions of the ancillary domain were subsequently tested by adding conditioned medium containing them to the fibroblast cultures. ADAMTS9 TSR9-Gon1 bound to fibronectin fibrils, indicated by co-staining of fibrillar structures by antimyc and fibronectin antibodies (Fig. 3A). We similarly tested binding of an ADAMTS9 TSR9 -15 construct to ask whether the TSRs could have a role in fibronectin fibril binding indepen-dent of the Gon1 module. This construct co-localized with fibrils in a comparable manner as ADAMTS9 TSR9-Gon1 (Fig.  3A). Because these co-staining experiments were done after addition of exogenous ADAMTS9, i.e. via conditioned medium, they suggest that the ADAMTS9 constructs were incorporated into forming fibrils or bound to the surface of fibronectin fibrils. The experimental approach thus indicated that it was not necessary for ADAMTS9 and fibronectin to undergo initial association in the secretory pathway. Therefore, fibronectin fibrils assembled by one cell or previously deposited in the ECM could bind to ADAMTS9 produced by another cell. Specificity of the ADAMTS9 binding for fibronectin fibrils was demonstrated by the absence of fibrillar myc staining in cultures treated with 100 nM functional upstream domain (FUD) (43) which specifically prevents fibronectin fibril assembly (Fig. 3B). Elimination of the possibility of fibrillin assembly by these MEFs provided additional evidence that ADAMTS9 constructs bound specifically to fibronectin fibrils and not to fibrillin-1 and -2 microfibrils.

Fibronectin proteolysis by ADAMTS9 Fibronectin fibrils are fragmented in the presence of catalytically active ADAMTS9
Fibronectin binding by the C-terminal ADAMTS9 constructs suggested that it could be an ADAMTS9 substrate. We therefore investigated the ability of full-length ADAMTS9 and the N-terminal construct ADAMTS9 N-L2, each containing the protease domain, to bind fibronectin fibrils and specifically, to modify fibronectin dimers or fibronectin fibrils proteolytically. Full-length ADAMTS9 is a large and unstable enzyme (Ͼ200 kDa), which is expressed at low levels by transfected mammalian cells, and no high-production natural sources are known. Despite multiple attempts, we have been unable to obtain substantially pure recombinant full-length ADAMTS9 or ADAMTS9 N-L2. As an alternative approach, we co-cultured mouse NIH-3T3 cells with HEK293T cells expressing ADAMTS9 or an inactive mutant construct in which the catalytic residue, Glu 435 was replaced by Ala. Similar experiments were conducted using the ADAMTS9 N-L2 construct, which excludes the TSR9-Gon1 region (Fig. 1A) and its corresponding inactive mutant (ADAMTS9 N-L2 E435A) and demonstrated that they co-localized with fibronectin fibrils (Fig. 4A).
The fibronectin networks showed considerable discontinuity or fragmentation in the presence of catalytically active ADAMTS9 constructs compared with the inactive mutants (Fig. 4, A and B). These observations suggested impaired fibronectin assembly or fibronectin fibril degradation in the presence of catalytically active ADAMTS9. We also studied the human retinal pigment epithelium-derived cell line hTERT RPE-1, which expresses endogenous ADAMTS9, and an RPE-1 Figure 2. Fibronectin binding to ADAMTS9 constructs. A, solid-phase binding assay of ADAMTS9 Gon1 module to immobilized fibronectin (FN, red line). BSA (blue line) was used as a control for binding. B, ADAMTS9 Gon1 module was tested for binding to N-and C-terminally truncated fibronectin constructs N-1F(III) and 1F(III)-C, respectively. Binding to these constructs was compared with binding to BSA (negative control). ***, p Ͻ 0.0001, n ϭ 4. C, solid-phase binding assay comparing ADAMTS9 Gon1 and ADAMTS9 TSR9 -15 binding to fibronectin. ***, p Ͻ 0.0001, n ϭ 4. D, surface plasmon resonance analysis of plasma fibronectin (analyte, 16.9 -270 mM) binding to ADAMTS9 Gon1 module immobilized on a CM5 chip. The binding K D was 2.8 nM, 2 /R max ϭ 4% for ADAMTS9 Gon1, and 104 nM ( 2 /R max ϭ 6%) for ADAMTS9 TSR9 -15. The purple and green lines represent the sensorgram, and the black line is the fitted curve produced by BIAevaluation software.

Fibronectin proteolysis by ADAMTS9
clone, D12, in which ADAMTS9 was mutated using CRISPR-Cas9 (33). Fibronectin fibril networks formed by these two lines were compared after 48-h culture in the presence of serum or 24-or 48-h culture subsequent to this period without serum. Under both conditions, a more robust fibril network was observed in D12 cultures (Fig. 5, A and B). Quantitative reverse transcriptase (qRT)-PCR showed no significant difference in FN1 mRNA levels between the two lines (Fig. 5C), strongly sug-gesting that ADAMTS9 either hindered the assembly of or proteolytically eliminated fibronectin fibrils in hTERT RPE1 cells.

ADAMTS9 activity leads to fibronectin cleavage at a specific site
To identify fibronectin cleavages generated in the presence of ADAMTS9, we coated tissue culture wells with human

Fibronectin proteolysis by ADAMTS9
fibronectin on which were cultured ADAMTS9 N-L2-or ADAMTS9 N-L2 EA-expressing HEK293T cells (Fig. 6A). The rationale for this experimental design was that surface-attached and fibrillar fibronectin may have an extended conformation, which would expose cryptic cleavage sites, in contrast to the globular conformation of fibronectin in solution (35,36). HEK293T cells do not assemble fibronectin fibrils; hence the experiment tested proteolysis of exogenous fibronectin adherent to the plates. The ADAMTS9 N-L2 and ADAMTS9 N-L2 EA constructs were expressed at comparable levels by HEK293T cells (Fig. 6B). Gelatin zymography indicated comparable levels of MMP2 in the medium of the respective HEK293T cells, whereas MMP9 was not detected (Fig. 6C). Conditioned medium from these experiments was obtained after 24 h of culture for targeted analysis of human fibronectin by LC-electrospray ionization-MS/MS. Fibronectin was identified by 44 and 23 tryptic peptides covering 26 and 15% of the fibronectin sequence from the medium of cultures expressing WT and inactive ADAMTS9 N-L2, respectively. The MS 1 spectrum identified a unique peptide ion of 1034 Da in digests of fibronectin with active ADAMTS9 N-L2 but not ADAMTS9 N-L2 EA (Fig. 6D, inset). The sequence of this peptide was determined in the MS 2 spectrum by assignment of fragment ions (Fig. 6D). The C-terminal Arg residue of this peptide was consistent with tryptic cleavage, but its N terminus was preceded by Gly in full-length fibronectin (Fig. 6E); hence the peptide did not arise by tryptic digestion during sample preparation and represents a "semitryptic peptide" indicative of fibronectin

Fibronectin proteolysis by ADAMTS9
proteolysis at the Gly 2196 -Leu 2197 peptide bond (sequence enumeration after fibronectin isoform with UniProt identifier P02751; Fig. 6E). This peptide bond lies within a short linker connecting fibronectin III 17 and I 10 modules that may comprise an unstructured region susceptible to proteolytic processing (Fig. 6E). The semitryptic peptide was not observed in analysis of fibronectin exposed to catalytically inactive ADAMTS9 N-L2. We undertook another targeted LC-electrospray ionization-MS/MS analysis, i.e. selective reaction monitoring, for the semitryptic peptide and other fibronectin ions and confirmed its presence exclusively in digests using the active ADAMTS9 N-L2 (Fig. 6F). Indeed, relative quantification of fibronectin peptides detected by LC-MS/MS of the medium of cells expressing WT or inactive ADAMTS9 N-L2 demonstrated that several tryptic fibronectin peptides were more abundant in the medium of cells expressing WT ADAMTS9 N-L2 than ADAMTS9 N-L2 EA (Fig. 6F). The relative abundance of fibronectin was determined in relation to enolase, detected at comparable levels in the samples, which was thus used as a normalization factor (examples of extracted ion chromatograms are shown in Fig. 6G). Thus, in addition to identification of a unique semitryptic peptide, the greater fibronectin sequence coverage and quantitative fibronectin peptide excess upon digestion of adherent fibronectin with active ADAMTS9 N-L2 suggested an overall greater abundance of fibronectin released in these experiments, also supporting fibronectin proteolysis by ADAMTS9. In the absence of an antibody suitable for recognition of the C terminus of fibronectin, we have not been able to obtain additional evidence of cleavage by a Western blotting.

Discussion
Observations from complementary approaches, including the yeast two-hybrid assay, solid-phase binding and surface plasmon resonance using purified proteins collectively support direct interaction of ADAMTS9 with fibronectin. ADAMTS9 localized specifically to fibronectin fibrils, because all ADAMTS9 fibrillar ECM staining was abolished by FUD, which prevents fibronectin polymerization, and these cells could not assemble fibrillin-1, a known ADAMTS ligand. ADAMTS9 produced by HEK293 cells localizes to their cell surface and is released into medium (11), indicative of binding in trans to fibronectin fibrils formed in the co-cultures and demonstrating that the interaction can occur extracellularly. Thus, fibronectin and ADAMTS9 could be co-assembled during fibril assembly, or ADAMTS9 may interact with preformed fibronectin fibrils. Distinguishing between these possibilities is challenging, because fibronectin assembly in cultured cells is a continuous, rapid, and extremely dynamic process involving binding of integrins to the secreted fibronectin dimer, extension of the bound fibronectin by cell movements, and polymerization via exposed binding sites on fibronectin dimers (36). Although the Gon1 module bound to purified fibronectin dimer in solid-phase assays, it did not co-localize with fibronectin fibrils. In contrast, upstream nonoverlapping ADAMTS9 constructs such as TSR9 -15 and ADAMTS9 N-L2, both lacking the Gon1 module, consistently bound to fibrils. Taken together with three nonoverlapping fibronectin SIDs identified in the yeast two-hybrid screen, the findings suggest multiple points of contact between the two molecules and distinct modes of interaction of different ADAMTS9 regions with fibronectin dimer or fibrils.
The observed binding of noncatalytic regions of ADAMTS9, i.e. Gon1 and TSR9 -15 constructs, logically led to consideration of fibronectin as a potential ADAMTS9 substrate. A semitryptic peptide identified by MS was determined to be specific for immobilized fibronectin digested by active ADAMTS9,

Fibronectin proteolysis by ADAMTS9
because it was not generated in the presence of the catalytically inactive mutant. Detection of higher levels of fibronectin tryptic peptides overall in the presence of active ADAMTS9 suggests additional ADAMTS9 cleavage sites that liberate fibronectin fragments or that cleavage at the Gly 2196 -Leu 2197 bond may promote fibronectin instability and/or depolymerization. One caveat of this study is that it has not been possible to purify ADAMTS9; hence we cannot exclude the possibility that ADAMTS9 is not the terminal protease that degrades fibronectin but instead activates another fibronectinase. Fibronectin

Fibronectin proteolysis by ADAMTS9
was previously identified as a substrate of several proteases, including the metalloproteases ADAM12 (44), MMP2, MMP8, and MMP14 (45,46), which may cleave it at the Ala 1078 -Thr 1079 bond (47). The ADAMTS9 processing site identified here, Gly 2196 -Leu 2197 , was previously identified as a cleavage site for MMP2 and MMP8 (48), but no change in MMP-2 was seen on gelatin zymography in our experimental system. Furthermore, direct binding of ADAMTS9 to both the fibronectin dimer and fibronectin fibrils suggests that it likely cleaves fibronectin directly. The data presented here using either the addition of ADAMTS9 conditioned medium or knockout of endogenous ADAMTS9 each show that fewer fibrils are present in the presence of catalytically active ADAMTS9, supporting its role in turnover of fibronectin fibrils. The observed ADAMTS9 incorporation in trans from conditioned medium or during co-culture experiments suggests that ADAMTS9 is integrated into fibrils as they form, potentially providing an innate mechanism for their physiological turnover, i.e. a "Trojan horse" effect. The cell-surface localization of ADAMTS9 (11,31,33) suggests it may access fibronectin fibrils quite early in the fibril assembly process.
Previously, endocytosis of MMP-cleaved fibronectin bound to integrins was implicated as a major degradation pathway operational in cultured cells (46). Like ADAMTS9, MMP14 also attenuates fibronectin matrix assembled by cultured cells. Among ADAMTS proteases, it was recently shown that ADAMTS2, ADAMTS3, and ADAMTS14, which are evolutionarily distinct from ADAMTS9, cleaved the fibronectin Ala 292 -Val 293 bond within a linker between domains I 5 and I 6 to release an N-terminal 30-kDa fragment that is crucial for fibronectin assembly (49). In addition, this region contains an ADAMTS16 cleavage site just upstream of the ADAMTS2/ 3/14 cleavage site, i.e. at the Ser 285 -Phe 286 peptide (50). Putative ADAMTS7 cleavage sites, Thr 279 -Ser 280 and Ser 281 -Gly 282 in the linker between I 5 and I 6 , were also recently identified (51). Also like ADAMTS9, ADAMTS16 processing at this site led to reduced fibronectin fibril assembly (50), but the effect of ADAMTS2, ADAMTS3, ADAMTS7, and ADAMTS14 on fibronectin assembly and turnover is yet to be investigated. Like ADAMTS9 processing in the III 17-I 10 linker, Thr 279 -Ser 280 , Ser 281 -Gly 282 , Ser 285 -Phe 286 , and Ala 292 -Val 293 are present within a linker that may present an exposed, unstructured region susceptible to proteolysis.
Fibronectin is present in the circulation and in many matrices where it is a critical cell adhesion substrate. ADAMTS9 is constitutively produced by vascular endothelium, suggesting a potential physiological role in fibronectin turnover during wound healing or a potential mechanism of its potent antiangiogenic effect (28). Fibronectin is a foundational matrix molecule that promotes assembly of fibrillin microfibrils, elastic fibers, collagen fibrils, and latent TGF-␤ binding proteins (36) and is an essential component of embryonic extracellular matrix (52). Thus, fibronectin assembly and turnover have considerable physiological relevance to embryogenesis and adult processes. The present work, together with a recent publication (50), emphasizes that fibronectin proteolysis and not only its transcriptional and assembly mechanisms are crucial for controlling its levels in ECM. Most work on fibronectin proteolysis has utilized cultured cells, and thus few relevant fibronectindegrading proteases operating at the tissue level are known. ADAMTS9 and MMP14 are among the few for which combined biochemical, in vitro, and in vivo evidence as fibronectinases is available. ADAMTS9 appears to be essential for embryonic fibronectin turnover as evidenced by stronger fibronectin staining in several tissues of Adamts9 Gt/Gt embryos. In contrast to Adamts9-null embryos, Adamts9 Gt/Gt embryos survive past gastrulation to undertake organogenesis, during which fibronectin-rich ECMs are produced. In previous analysis, Adamts9 Gt/Gt umbilical cord was noted to have stronger fibronectin staining than WT littermate umbilical cord, and immunostaining of Adamts9 Del/ϩ eyes demonstrated stronger fibronectin staining (14,19). More intense fibronectin staining was also seen in the pregnant mouse uterus upon conditional Adamts9 deletion in smooth muscle cells (20). Taken together with these in situ observations, the present data suggest that fibronectin turnover is a widespread morphogenetic role of ADAMTS9.
In contrast, Mmp14 appears not to have as crucial a role in fibronectin turnover during early development, because it is most highly expressed during skeletal development (53), and Mmp14-deficient mice survive to birth without dramatic morphogenetic defects (54,55). However, Mmp14 appears to be necessary for fibronectin turnover during formation of tendons (47), a tissue where it is highly expressed (53), but Adamts9 is not (56). Thus, diverse proteases may regulate fibronectin content of matrix in specific spatial and temporal contexts, with

Fibronectin proteolysis by ADAMTS9
this abundant, large dimeric molecule and its supramolecular structures being targeted by a plethora of proteases at different sites.

Yeast two-hybrid screen
A two-hybrid screen of the human placenta_RP5 library (Hybrigenics, Cambridge, MA) was undertaken using the human ADAMTS9 Gon1 module (residues 1734 -1935) fused in-frame and C-terminal of a Gal4 DNA-binding domain in the pB66 vector. The hybrid protein was determined to be neither auto-activating nor toxic in yeast; hence screening was done without inclusion of 3-amino-1,2,4-triazole. 81.4 ϫ 10 6 yeast clones were screened. A proprietary Predicted Biological Score (34) was computed by Hybrigenics to assess interaction reliability of isolated fibronectin clones and suggested that the probability of the interaction being nonspecific was low. Selected interaction domains representing the minimal region corresponding to all the isolated clones were identified using PIM-Rider R (Hybrigenics).

Surface plasmon resonance analysis
Purified ADAMTS9 Gon1 or ADAMTS9 TSR9 -15 in 10 mM acetate, pH 4.0 were immobilized on a BIAcore CM5 sensor chip (research grade) with the amine coupling kit following the manufacturer's instructions (GE Healthcare). 900 and 1600 resonance units of ADAMTS9 Gon1 or ADAMTS9 TSR9 -15, respectively, were coupled to the chip for analysis in a BIAcore 3000 instrument (GE Healthcare). The single cycle kinetic analysis was performed at 25°C in 10 mM Hepes buffer, pH 7.4 with 0.15 M NaCl, 1 mM CaCl 2 , and 0.005% (v/v) surfactant P20 at a flow rate of 30 l/min. Human plasma fibronectin (57) was diluted in the above buffer at different concentrations and injected through an uncoupled control flow cell in series with a flow cell coupled to ADAMTS9 constructs. The sample injection time was 2 min and was followed by a 1-min pause before the next injection. The dissociation time was 15 min. 50 mM glycine, pH 3.0 was used for chip surface regeneration at a flow rate of 100 l/min for 30 -60 s after Gon1-binding analysis, whereas 50 mM NaOH was required for regeneration of the ADAMTS9 TSR9 -15 chip surface. All data were corrected with reference to the background binding in the control flow cell. The kinetic constants were calculated assuming a 1:1 (Langmuir) binding model using the BIAevaluation software (version 4.0.1; GE Healthcare).

Solid-phase binding assay
96-well plates were coated with either fibronectin or BSA (control) at 50 nM/well. After washing and blocking the wells with 2.5% milk, 20 mM Tris-HCl buffer, pH 7.4, 150 mM NaCl, 2 mM CaCl 2 , 0.05% Tween 20 (TBST Ca 2ϩ ), 50 l of purified ADAMTS9 Gon1 in 2.5% milk/TBST Ca 2ϩ at a concentration between 0.3 and 10 M was added to the coated wells and incubated on a shaker (300 Hz) for 2 h at room temperature. Antimyc mouse mAb was used to detect bound Gon1. For fibronectin subconstruct binding experiments, 10 g/ml fibronectin solution was used per well for coating, and 2.5 M ADAMTS9 Gon1 was used as ligand. To compare binding of different ADAMTS9 constructs, 10 mg/ml of fibronectin was used for coating the wells, and 80 nM of purified ADAMTS9 Gon1, TSR9 -15, or TSR2-8 constructs were added to the coated wells. Absorbance was measured at 492 nm.

Proteomics determination of fibronectin processing by ADAMTS9
2 ϫ 10 6 HEK293T cells transiently transfected with ADAMTS9 N-L2 plasmids were washed and suspended in serum-free DMEM and seeded onto wells coated with fibronectin (5 g/well) in 6-well plates. The conditioned medium was collected 24 h later, and proteins were precipitated using TCA, dried in a SpeedVac, and reconstituted with 50 l of Tris-HCl buffer containing 6 M urea. The samples were reduced with DTT, alkylated with iodoacetamide, and further diluted in 100 l of 50 mM ammonium bicarbonate, and trypsin was added at a 1:20 (trypsin:protein) ratio. A first round of digestion was carried out overnight, a second aliquot of trypsin was added, and the sample was incubated for an additional 6 h. The protein digest was desalted using the C18 SPE method and reconstituted in 50 l of 1% acetic acid for LC-MS analysis. The LC-MS system was a Finnigan LTQ-Orbitrap Elite hybrid mass spectrometer system. The HPLC column was a Dionex 15-cm ϫ 75-m inner diameter Acclaim Pepmap C18, 2 m, 100 Å reversed phase capillary chromatography column. 5 l volumes of the tryptic digest were injected. The peptides were eluted from the column by an acetonitrile, 0.1% formic acid gradient at a flow rate of 0.3 l/min for introduction into the source of the mass spectrometer online. The nanoelectrospray ion source was operated at 1.9 kV. The digest was analyzed using the data-dependent multitask capability of the instrument acquiring full scan mass spectra to determine peptide molecular masses and product ion spectra to determine amino acid sequence in successive instrument scans. The data were analyzed by using all CID spectra collected in the experiment to search the human UniProtKB sequence databases using Mascot (false discovery rate, 0.01%).
Additional samples were analyzed using a targeted selective reaction monitoring (SRM) (61) experiment, which involves the fragmentation of specific ions over the course of the LC experiment, including two peptides from enolase and several fibronectin peptides. The relative abundance of the fibronectin peptides was determined by normalizing the fibronectin (FN) peptide peak areas to the peak areas (PA) of the enolase peptides: PA(FN)/PA(enolase).