Regulatory properties of vitronectin and its glycosylation in collagen fibril formation and collagen-degrading enzyme cathepsin K activity

Vitronectin (VN) is a glycoprotein found in extracellular matrix and blood. Collagen, a major extracellular matrix component in mammals, is degraded by cathepsin K (CatK), which is essential for bone resorption under acidic conditions. The relationship between VN and cathepsins has been unclear. We discovered that VN promoted collagen fibril formation and inhibited CatK activity, and observed its activation in vitro. VN accelerated collagen fibril formation at neutral pH. Collagen fibers formed with VN were in close contact with each other and appeared as scattered flat masses in scanning electron microscopy images. VN formed collagen fibers with high acid solubility and significantly inhibited CatK; the IC50 was 8.1–16.6 nM and competitive, almost the same as those of human and porcine VNs. VN inhibited the autoprocessing of inactive pro-CatK from active CatK. DeN-glycosylation of VN attenuated the inhibitory effects of CatK and its autoprocessing by VN, but had little effect on acid solubilization of collagen and VN degradation via CatK. CatK inhibition is an attractive treatment approach for osteoporosis and osteoarthritis. These findings suggest that glycosylated VN is a potential biological candidate for CatK inhibition and may help to understand the molecular mechanisms of tissue re-modeling.


Results
VN affects collagen fibril formation. Type I collagen accounts for approximately 90% of the whole protein in the organic bone matrix and forms fibrils in bone 15 . First, to clarify the effects of VN on collagen fibril formation, the turbidities of collagen solutions in the presence or absence of VN were measured. Although turbidity does not directly reflect fibril formation, increased turbidity indicates changes in collagen fibril and fibril formation. Therefore, this assay is widely used as a simple and effective method to identify factors influencing fibril formation in solution 33,34 . The turbidity measurement of the collagen solution generates a sigmoid curve composed of a lag phase, a subsequent growth period, and a plateau that indicates complete gelation due to warming 33,34 .
Human VN (hVN) induced rapid fibril formation in porcine and bovine collagen (Fig. 1a). Moreover, porcine VN (pVN) accelerated fibril formation in porcine collagen but inhibited it in bovine collagen. Interestingly, hVN enhanced porcine collagen fibril formation in a dose-dependent manner (Fig. 1b). Both hVN and pVN prolonged T lag, increased turbidity at 60 min, and increased plateau turbidity in Fig. 1c., suggest that VN increased the collagen fibril formation rate and the molecular weight or thickness of the aggregate 33,34 . Collagen fibrils formed in the absence or presence of hVN was observed via scanning electron microscopy (SEM) at both low magnification (7000×) and high magnification (20,000×). Collagen fibers formed in the presence of VN were in closer contact with each other than those formed in the absence of VN, and clumps were randomly present (Fig. 2). Pore areas of the collagen fibers of control and hVN were 5.62 ± 2.57μm 2 and 1.02 ± 0.47μm 2 , respectively. These results show that hVN accelerates collagen fibril formation and changes the morphology of collagen fibers, but pVN acts differently depending on the collagen species.
Collagen fibers formed with VN are easily acid-solubilized. Collagen fibers are formed at neural pH and dissolve under acidic conditions. Collagen is degraded by CatK under acidic conditions. Therefore, the collagen in bone is likely solubilized by large amounts of acid secretion from osteoclasts during and/or before degradation by CatK 35 . To determine the effects of VN on the acid solubilization of collagen fibers, the decrease in turbidity was measured using an in vitro assay. Figure 3a shows that the turbidity of porcine collagen fibers formed in neutral pH significantly decreased to 10% and 25% at pH 4.0 and 4.5, respectively, in the control, which did not contain hVN. In contrast, the turbidity did not change at pH ≥ 5.0 compared to the turbidity at 0 min before the pH change. In the presence of hVN, the turbidity of the collagen fibers decreased to approximately 10% at pH 4.0 and 4.5, and to approximately 19% at pH 5.0. Moreover, the pH dependency of pVN and hVN were similar. The turbidity of collagen fibrils formed in the presence of hVN or pVN decreased in a concentration-dependent manner, and pVN reduced turbidity at slightly lower concentrations than hVN did (Fig. 3b). The molecular weight of hVN is 75 kDa, that of nicked product is 65 kDa, and that of pVN is 58 kDa. Thus, the molarities of hVN and pVN at 0.1 mg/mL were 1.3 μM and 1.7 μM, respectively. Notably, hVN increased collagen solubilization and decreased the content of collagen fibers at pH 4.5 (Fig. 3c). Moreover, hVN was mainly found in the solubilized collagen fraction and had the same effect on bovine collagen fibers as on the porcine collagen fibers; pVN also showed effects similar to those of hVN (Supplementary Figure S1). These results showed that the collagen fibers formed in the presence of VN were easily solubilized in acid. This prop- (c) Dose-dependence of hVN and pVN for T lag, turbidity at 60 min, and plateau of porcine collagen fibril formations. Blue bars are hVN; pink bars are pVN. *p < 0.05; **p < 0.01; ***; < 0.001 compared with 0 μg/mL VN as control by one-way ANOVA with Dunnett's post-hoc test. n = 4. Values represent the mean, and error bars represent the SE. Experiments were performed independently 3-4 times. erty was common to both hVN and pVN, which exerted similar effects on porcine-derived and bovine-derived collagen fibers. Since both collagen and VN are ECM components and collagen is a well-known ligand for VN, VN inhibits CatK activity. The potential suppressive effects of VNs on the activity of CatK, which degrades type I collagen under acidic conditions, were examined using a fluorescence-labeled substrate. CatK activities at pH 4.0-7.2 were measured and found to be the highest at pH 5.5 in the absence of hVN (−) (Fig. 4a). The optimal pH for CatK is 5.5, which is consistent with our results. Surprisingly, hVN (+) strongly inhibited CatK activity at all pH values. The inhibition of CatK with 2 μM hVN was especially strong at pH 4.5-7.2 (Fig. 4b). Moreover, pVN strongly inhibited CatK activity at pH 4.5-5.5, similar to hVN (Fig. 4c). Both hVN and pVN dosedependently inhibited CatK activity at pH 5.5. At 2000 nM, hVN and pVN inhibited CatK activity by 6.05% and 9.93%, and their IC 50 values were 16.57 ± 46.33 nM and 8.13 ± 11.47 nM, respectively (Fig. 4d). The IC 50 value of odanacatib, a CatK inhibitor, was 1.24 ± 0.46 nM. In the Dixon plot, CatK inhibition by hVN and pVN resulted in a line crossing into the negative values on the X-axis (Fig. 4e,f). These observations provide evidence that VNs can competitively inhibit CatK activity at a concentration approximately 240-640 times lower than that in blood, despite having a weaker inhibitory effect than odanacatib. It was surprising that VNs inhibited CatK activity, as there is no report of the relationship between VNs and cathepsins.

VN inhibits autoprocessing of pro-CatK.
CatK is activated from a precursor, pro-CatK, via an intermediate under acidic conditions by N-terminal autoprocessing. We evaluated the effects of VNs on the autoprocessing of pro-CatK by reducing the molecular weight of pro-CatK (35.3 kDa) to CatK (~ 26 kDa) using SDS-PAGE.
In the absence of VN, pro-CatK was completely transformed to mature-CatK after incubation for 10 min at pH 4.0, while a small amount of pro-CatK remained after incubation for 10 min at pH 4.5 (Fig. 5a). In the presence of VN, pro-CatK and intermediate-CatK were detected after incubation for 10 min at pH 4.0 and pH 4.5. About half of pro-CatK was activated to the mature type and the other half remained as pro-CatK, and there was no significant difference between incubation for 30 min at pH 5.0 and 5.5, with and without hVN. Thus, hVN inhibited the autoprocessing of pro-CatK at pH 4.0 and 4.5. The time course of hVN-mediated effects on pro-CatK activation at pH 4.0 was examined (Fig. 5b). At 10 min after incubation, pro-CatK was completely activated to mature-CatK in the absence of VN, but some of pro-CatK remained inactivated when hVN was present. At 30 and 60 min, the activated mature-CatK content was higher in the presence of VN than in the absence of VN. VN increased the levels of pro-CatK, intermediate-CatK, and mature-CatK (Fig. 5c). These results suggest that VN not only inhibits pro-CatK activation but also inhibits degradation of the mature type.
Binding of VN to collagen and CatK. VN's binding to collagen and CatK at pH 4.0-7.2 was analyzed by ELISA. The binding of VN to collagen showed a U-shaped curve, with a maximum binding pH of 7.2 and low binding pH of 4.5-5.5 (Fig. 6a). Conversely, VN's binding to collagen showed an inverted U-shaped curve, with a pH of 5.0 for maximum binding and a pH of 7.2 for low binding (Fig. 6b).  www.nature.com/scientificreports/ Additionally, VN specifically binds to collagen 5 , but the interaction between VN and CatK has not been reported. To determine whether VNs can specifically bind to CatK, real-time binding analysis by Surface plasmon resonance (SPR) was performed using immobilized CatK on the chip and different dilutions of hVN, pVN, and bovine serum albumin (BSA) (Fig. 6c). Both VNs showed affinities to CatK but not to BSA, suggesting that VN's binding to CatK is specific. The dissociation constants (K D ) of hVN and pVN to CatK were 1.99 ± 0.093 nM and 4.20 ± 0.087 nM, respectively. These results suggest that VN preferentially binds to collagen under neutral conditions and that VN binding targets under acidic conditions switch from collagen to CatK.
Effects of VN glycosylation. hVN has three N-glycosylation sites, and the structures of their carbohydrate chains mainly comprise two or three bi-antennary sialylated complex-type N-glycans, as shown in Fig. 7a 36 . pVN has similar structures of N-glycans 37 . We prepared deN-gly-VN and deNeu-VN by incubation with N-glycosidase and neuraminidase, respectively. Deglycosylation of VN was confirmed by reduced molecular weight, as described previously (Fig. 7b) 5,[9][10][11] . We investigated the effects of hVN glycosylation on collagen fibril formation, collagen acid solubilization, CatK activity, and autoprocessing of pro-CatK.
The increase in turbidity due to collagen fibril formation was accelerated by all VNs, and deNeu-VN caused slightly higher increase than the control-and deN-gly-VNs (Fig. 7c,d). On the other hand, acid-induced www.nature.com/scientificreports/ reduction in relative turbidity was enhanced for all VNs, including the control, deN-gly-VN, and deNeu-VN, although differences between the three VNs were not significant (Fig. 7e,f). Control-, deN-gly-, and deNeu-VN inhibited CatK activity by 10.4%, 21.3%, and 19.5%, respectively, compared to that without VN (Fig. 7g). This inhibition of CatK activity by control-VN was stronger than that by deN-gly-and deNeu-VNs. Figure 7h shows the result of autoprocessing of pro-CatK with SDS-PAGE, and the remaining pro-CatK content was evaluated using ImageJ. In the absence of hVN, the remaining pro-CatK content was reduced from 61.2 to 0% after incubation for 10 min. In contrast, the remaining pro-CatK contents were 37.4%, 7.5%, and 14.2% after 10 min incubation with control-, deN-gly-, and deNeu-VNs, respectively. These results indicate that control-VN inhibited pro-CatK autoprocessing more effectively than deN-gly-VN and deNeu-VN. VNs' binding to collagen or CatK was assayed by ELISA. Since glycosylation of VNs may affect immunoreactivity, the reactivity of glycosidase-treated VNs was examined by dot-blotting before binding assay. It was confirmed that all VNs had almost the same immunoreactivity ( Fig. 8a,b). By ELISA binding assay, VNs bound to collagen in the order of deN-gly-VN > deNeu-VN > Control-VN (Fig. 8c). In effects of VN glycosylation, VNs' binding to CatK was similar to VNs' collagen binding (Fig. 8d). that VNs disappeared after incubation with pro-CatK, indicating that they can be degraded by CatK. In this study, VN degradation by CatK was demonstrated by reduced VN molecular weight, as determined using SDS-PAGE (Fig. 9a). The molecular weights of control hVN and its degraded form were 75 and 65 kDa, respectively, and those of deN-gly and deNeu-VNs were reduced due to cleavage of N-glycosidase F and neuraminidase, respectively (Fig. 7b). The control-, deN-gly-, and deNeu-VNs were time-dependently degraded by 10 nM CatK. There was no significant difference in the contents of undegraded VN among the three VNs due to incubation with 10 nM CatK (Fig. 9b). VN was not degraded after 120 min of treatment with 1 nM CatK but was completely decomposed within 10 min of treatment with 100 nM CatK. VNs were degraded by 10, 30, and 100 nM CatK, and no significant difference was found between the control, deN-gly, and deNeu-VNs (Supplementary Figure S2). These results suggest that N-glycosylation of VNs is dispensable, but enhances their inhibitory effects on CatK activity and pro-CatK activation, and that sialylation of VN is also not essential but enhances collagen fiber formation.

Discussion
This is the first report of VNs interacting with CatK, which also demonstrates CatK inhibition by these biomolecules. The results of this study and the inferences from the results are summarized in a schematic (Fig. 10). i) VN accelerates collagen fibril formation under neutral pH conditions (Fig. 1). ii) Under acidic conditions, collagen fibers formed with VNs were readily acid soluble (Fig. 3) and (iii-iv) VNs inhibited autoprocessing of pro-CatK (Fig. 5a) as well as CatK activity at concentrations 240-640 times lower than blood levels (Fig. 4). VN, bound to collagen at neutral pH, switches binding to CatK to inhibit its activity around pH 5.0 (Fig. 6). Although VN glycosylations are not essential for the regulation of these activities, VN desialylation is effective for collagen fibril formation, and its N-glycosylation is effective for inhibition of both pro-CatK autoprocessing and CatK activity (Fig. 7). The promotion of collagen fibril formation by VNs was expected, because of the adhesive properties of VNs. However, it was surprising that the collagen fibers formed with VNs were easily acid-dissolved (Fig. 3a,b). Most VNs were in the fraction of solubilized collagen (Fig. 3c). VN's collagen-binding activity is high at neutral pH and low around pH 5.5, which is CatK optimum pH (Fig. 6a). In contrast, the VN's CatK binding activity is high around pH 5.0 and low at neutral (Fig. 6b). These results suggest that VN crosslinks collagen fibers by strongly binding collagen and promoting the formation of collagen fibers at neutral pH. And that the change to acidic pH weakened the crosslinking of collagen fibers by VN and promoted solubilization of collagen fibers. Then, VN switches binding to CatK to inhibit its activity (Fig. 10, In the left four frames). www.nature.com/scientificreports/ As shown in Fig. 4e,f, Dixon plots demonstrate that the CatK inhibition by both hVN and pVN is competitive. Figure 9 show that VN was degraded by the high concentration of CatK, suggesting that VN is a substrate for CatK; this is consistent with the results of Fig. 4e,f that CatK inhibition by VN is competitive inhibition.
VN has several functional domains: the somatomedin-B, cell attachement site (the R-G-D motif), a connecting region, and hemopexin consist of four blades 38 (Fig. 11a). Hemopexin blade four is interrupted by heparin-binding domain, responsible for VN's binding to several other ligands, including collagen 39 , sulfated glycosaminoglycans (GAGs) 40,41 , and integrin α V β 5 42 (not integrin αvβ3). hVN and pVN have 67-71% amino acid sequence identity with respect to the common domain organization and ligand-binding activities 43 . pVN has an unusually low molecular weight among mammalian VNs including hVN. This is ascribed to a lack of 22 amino acids in the connecting region and a lack of an 80 amino acid fragment from the C-terminal end compared to hVN, but pVN has all four functional domains 43 . In this report, we show that pVN and hVN are comparable in CatK inhibition. It is suggested that the CatK binding site of VN is sufficient for the sequence contained in pVN.
hVN's three-dimensional structure has been determined partially, its full length has not been solved, suggesting the existence of an intrinsically disordered region. It is generally known that the intrinsically disordered region interacts with other molecules undergoing a disorder-to-order transition upon binding 44 . hVN's hemopexin domain has disordered regions where collagen binds 39 and Yersinia pestis outer membrane protein Ail interacts 38 . To find the CatK interaction sites in hVN, we predicted intrinsically disordered regions using the following four methods' consensus: PONDR 45 44 . We used a website built by the developer for PONDR, DisEMBL, and PrDOS. For DisEMBL, a region predicted as Hot-loop was employed in this analysis. For FoldIndex, the in-house program was used. The disorder values were calculated along the amino acid sequence in the sliding window with size 19, and a residue with a value less than zero was defined as disordered residue. Furthermore, multiple sequence alignment was performed by ClustalW (ver. 2.1, https:// www. genome. jp/ tools-bin/ clust alw) using the protein sequence of hVN and pVN; both VNs inhibit CatK activity. As shown in Fig. 11a,b, red bars (aa 100-130, 292-296, 364-397) in hVN are the predicted intrinsically disordered regions, and the three regions were conserved in pVN. From these computational analyses, we conjectured that CatK binds to the disordered region in the hemopexin domain.
CatK, an attractive therapeutic target for diseases with excessive bone resorption such as osteoporosis and osteoarthritis [26][27][28][29][30][31][32] , is a major lysosomal protease that accounts for 98% of total cysteine protease activity in osteoclasts and is responsible for the degradation of bone matrix proteins 49 . CatK inhibition circumvents the bone coupling mechanism and has excellent effects on osteoporosis, thus suppressing bone resorption without reducing bone formation 50,51 .
Findings of this study regarding CatK activity could be relevant to treating human osteoporosis, particularly as we demonstrated the importance of the VN-CatK interaction between purified molecules in vitro. Notably, because chemical compounds such as odanacatib 27 , ONO-5334 28,29 , and MIV-711 31,32 , which show CatK inhibitory effects in the same in vitro assay used in this study, effectively alleviated osteoporosis or osteoarthritis in clinical trials. Other characteristics of VN, which include promoting collagen fibril formation as found in this study-and enhancement of osteoblast adhesion 52 , might be beneficial in bone formation. Unfortunately, odanacatib development has been discontinued, due to the reports of increased stroke risk, although it has been shown to be effective in clinical trials 26,30 . MIV-711 has beneficial effects on bone and cartilage with a reassuring safety profile in humans in clinical trials 31,32 . The present study identified VN as an alternative biomolecule with CatK inhibitory activity. VN may be a key molecule in the treatment of osteoporosis and osteoarthritis, although further research is required in future.
The effects of VN on CatK activity and autoprocessing using osteoclasts, which predominantly express CatK, are particularly interesting. For both cell-and animal-based assays, it is necessary to identify the active site of VN showing CatK inhibitory activity. VN is an adhesive glycoprotein and enhances cell adhesion as well as spreading of almost all cells, including osteoclasts and osteoblasts, in cell-based assays 12,53 . The adhesive properties of VN are likely attributable to the RGD-motif that binds integrin, especially osteoclasts highly expressing the VN receptor α v β 3 integrin 12 . In the study using osteoclasts and osteoblasts, it is presumed that the true effect of CatK cannot be evaluated, because the RGD-motif in VN largely contributes to the observed effects. Therefore, we aimed to determine the active sites of VN with CatK inhibitory and collagen fibril formation promoting activities. In the near future, we plan to investigate how (sugar) fragments or (sugar) peptides affect osteoclasts and osteoblasts, and an osteoporosis-or osteoarthritis-based mouse model to determine whether the CatK inhibitory effects of VN can be applied in treating osteoporosis and/or osteoarthritis.
Recently, it was reported that VN-derived peptides that do not contain RGD-motif reversed ovariectomyinduced bone loss via the regulation of osteoblast and osteoclast differentiation 14 . The VN-derived peptide was identified by screening for agents promoting the attachment and spread of osteogenic cells but restraining osteoclast differentiation and function in vivo. We suspect that the restraining effects of the VN-derived peptide on osteoclasts may be mediated by CatK, although there is no mention of CatK in the concerned report.
In this study, the measurement of collagen fibril formation based on turbidity showed that VNs shortened the T lag and increased turbidity for 60 min (Fig. 1c), suggesting that they promote the collagen fibrosis rate. In addition, VNs increased the plateau value (Fig. 1c). A previous report has shown that turbidity mostly increases during collagen self-assembly due to increasing fibril thickness 33 and increased molecular weight of the aggregates 34 . Analysis of collagen morphology using SEM showed that VN brought the collagen fibers into close contact with each other to increase the molecular weight and the thickness, consistent with previously reported interpretations.
Bovine and skin collagen from both porcine and bovine are widely utilized for industrial purposes. Since porcine collagen resembles human collagen, it does not cause a significant allergic response in humans, whereas ~ 3% www.nature.com/scientificreports/ of the population is allergic to bovine collagen [54][55][56] . Therefore, it is likely that pVN accelerates collagen fibril formation in humans. Human, rabbit, and mouse VNs have completely conserved potential N-glycosylation sites, as deduced from protein sequences [57][58][59] . pVN and hVN have two and three potential N-glycosylation sites, respectively. Two potential glycosylation sites in pVN and hVN correspond to the same positions; however, pVN lacks one site on the C-terminal side that correspond to the third glycosylation site present in hVN 60 . The major oligosaccharides in pVN and hVN are common structures belonging to the biantennary complex N-glycan containing 1-3 mol sialic acid, and pVN has more core fucosylation than hVN 36,37 . In this study, hVN glycosylation is primarily important for enhancing the effects on CatK activity and autoprocessing of pro-CatK, whereas there is little effect of hVN glycosylation on VN degradation by CatK and acid solubilization (Figs. 7, 9). These results suggested that the site and/or mode of hVN that inhibits CatK activity and its activation is different from the ones collagen-binding of hVN.
VN's binding to collagen and CatK was both deN-gly-VN > deNeu-VN > Control-VN in order of strength (Fig. 8). In our previous study, VN is multimerized, and its molecular weight is larger in the order of deN-gly-VN > deNeu-VN > Control-VN 10 . It is considered that de-glycosylated VN enhances CatK binding activity due to its multivalent effect. Then, VN's larger multimerization by deglycosylation may prevent VN from binding to the substrate binding site of CatK. In the interaction between VN and collagen, the slight enhancement of collagen fibril formation by deNeu-VN (Fig. 7d) may be due to the shift of VN pI to alkali after neuraminidase treatment 9 . Because the molecular organization within collagen fibrils is determined by numerous non-covalent interactions between collagen molecules, particularly charge-charge interactions 61 .
Previous studies have shown that complex N-glycan is important for bone formation and bone diseases 62 . The congenital disorder of glycosylation type IIa (CDG-IIa) has a deficiency in complex N-glycans, which is commonly observed in several diseases including osteopenia and kyphoscoliosis 62,63 . Mice lacking the Mgat2 gene were deficient in GlcNAcT-II glycosyltransferase activity and complex N-glycans, which resulted in severe gastrointestinal, hematologic, and osteogenic abnormalities 43 . Previously, there have been no published reports regarding the effects of VN glycosylations on bone diseases. In this study, we should show for the first time the importance of VN glycosylation in relation to bone diseases.
This study also revealed that VN is degraded by CatK (Fig. 9). Glycosylation of proteins is generally protease resistant; however, deN-gly-VN and deNeu-VN were found to be more likely to be degraded by CatK than control-VN, although the difference was not statistically significant. Furthermore, we confirmed that VN was not degraded by CatK (0.5 nM) in the CatK activity assessment experiment.
In summary, we showed that VN, a glycoprotein in the extracellular matrix and plasma, enhanced collagen fibril formation and inhibited both CatK activity and autoprocessing of pro-CatK in vitro. Enzymatic deglycosylation of VN attenuated the inhibitory effects of CatK. These data suggest that glycosylated VN may prevent bone destruction through CatK inhibition and that VN could serve as a therapeutic agent for osteoporosis or osteoarthritis. Furthermore, these findings may provide clues to elucidate the mechanism of bone formation and dissolution, and may help to understand the side effects caused by currently used CatK inhibitors 30 .
Collagen fibril formation assay. Collagen (2 mg/mL, 50 μL) and VN (indicated concentration, 50 μL) in TBS were combined in a well of an Iwaki 96-well assay plate. After mixing, the collagen solution with or without VN was incubated at 37 °C for 2 h to form fibrils. The turbidity of each well was measured at 350 nm every 5 min using a Cytation3 reader (BioTek) 33,35 . All raw datasets had buffer controls subtracted. The lag phase time (T lag) is defined as the time when the turbidity increased to 0.01, and the plateau value was calculated using Prism 9.0 (GraphPad Software, San Diego, CA, USA). After dehydration, the samples were continuously dehydrated with 100% ethanol at room temperature overnight. The samples were processed once in 50% tert-butyl alcohol/ethanol for 1 h and three times in 100% tertbutyl alcohol for 1 h at room temperature. After dehydration, the samples were frozen at 4 °C and vacuum dried. The samples were coated with a thin layer (30 nm) of osmium using an osmium plasma coater (NL-OPC80A; Nippon Laser & Electronics Laboratory, Nagoya, Japan). The samples were observed via SEM (JSM-7500F; JEOL Ltd., Tokyo, Japan) at an acceleration voltage of 3.0 kV. The pore size area (μm 2 ) in SEM images was measured by NIH imaging.  www.nature.com/scientificreports/ 30 min, it was diluted to 4 nM with 5 mM EDTA/DTT. One hundred microliters of the reaction mixture (VN, 10 μM Z-Phe-Arg-AMC, and 0.5 nM CatK in 25 mM pH buffer; acetate buffer, pH 4.0-6.0, or Tris-HCl buffer, pH 7.2, containing 2.5 mM EDTA/DTT) was added to a 96-well assay plate and incubated at 37 °C. Fluorescence emissions were measured at 460 nm after excitation at 380 nm using a Cytation3 reader.

SEM.
Pro-CatK autoprocessing assay. Pro-CatK (3 μM) in 100 mM acetate buffer (pH 4.0-5.5), containing 2.5 mM EDTA/DTT, was incubated at 37 °C with or without VN 49,65 . After incubation at 37 °C for 10-120 min, 2 × SDS-PAGE sample buffer was added immediately to the samples, which were then heated at 98 °C for 10 min. The proteins were separated in 15% Tris-glycine gels with 2-mercaptoethanol and stained with EzStain Silver (AE-1360; ATTO Technology, Amherst, NY, USA). The total intensities of both pro-CatK and CatK bands were quantified using ImageJ (NIH, Bethesda, MD, USA). were used, and solution volume was 40% of ELISA for collagen binding of VN. CatK was activated from pro-CatK by the same methods of "CatK activity assay in Methods" and immobilized at 100 nM overnight at 4 °C. After washing with cold TBS-T, the wells were blocked with 3% BSA in cold TBS-T for 3 h at 4 °C. Subsequent operations were performed in the same way as ELISA for collagen binding of VN, except that the room temperature was changed to 4 °C. SPR. VN binding to CatK was determined by SPR multicycle analyses by using BiacoreT-100 (GE Healthcare, USA). All experiments were performed at 20 °C. VN (100 nM) was pre-injected on the CM5 chip and the chip was regenerated by 10 mM NaOH to avoid non-specific binding of VN onto the chip. The CatK activated by same methods of "CatK activity assay in Methods" was diluted to 100 nM with 10 mM sodium acetate buffer pH 5.0 and immobilized on the chip by amine coupling method with a flowrate of 10 mL/min following the manufacturer's instructions. BSA was also immobilized in 10 mM sodium acetate buffer pH 5.5 on another flow cell of the same chip as a reference. Approximately 150 RU of CatK and BSA were immobilized on each flow cell. HBS-EP buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% v/v Surfactant P20) was used as the immobilization running buffer.

ELISA. Collagen and
hVN, pVN, and BSA (0, 625, 12.5, 25, 50, 100 nM) were injected as analytes with a contact time of 60 s, dissociation time of 120 s, and a flowrate of 20 mL/min. The chip was regenerated with 10 mM NaOH for 30 s. Filtrated 10 mM acetate buffer (pH5.5) containing 2.5 mM EDTA/DTT and 0.005% v/v tween 20 was used as the running buffer. The data were determined kinetic constants for the interaction using the 1:1 kinetic binding model in the Biacore T100 evaluation software. Dot-blotting. Three-fold dilution series of VNs (each 100 μL) were dot-blotted onto polyvinylidene fluoride membranes (PVDF) and blocked with 3% BSA in TBS-T, then cut into lanes and reacted with the rabbit anti-VN IgG and goat anti-rabbit IgG Fc HRP conjugated at room temperature for 2 h each. The spots were visualized by ECL reagents (GE Healthcare ECLTM Western Blotting Detection Reagents). Chemiluminescence was detected using an ImageQuant LAS4000 mini imager (GE Healthcare). The plotted signal intensities were calculated using ImageJ (NIH) software.
VN degradation by CatK. CatK was autoactivated from pro-CatK (5 μM) upon incubation in 100 mM acetate buffer (pH 4.0), containing 2.5 mM EDTA/DTT, at 37 °C for 10 min. CatK was diluted with ice-cold 200 mM acetate buffer (pH 5.5), containing 2.5 mM EDTA/DTT, into 2, 20, 60, and 200 nM solutions, and mixed with 4 μM VN in a 1:1 ratio. The mixed solutions were incubated at 37 °C for 0, 10, 30, 60, or 120 min for CatK to degrade VN. After incubation, ice-cold E-64 (inhibitor of proteinases including CatK; final concentration: 1 μM) and 2 × SDS-PAGE sample buffer were added to the solutions. The samples were immediately heated at 98 °C for 10 min. The samples were electrophoresed using 7% Tris-glycine gels with 2-mercaptoethanol and stained with SYPRO Ruby gel stain. The protein bands were detected using ImageQuant LAS 4000 (GE Healthcare, Chicago, IL, USA). ImageJ (NIH) was used to quantify the intensity of the VN bands corresponding to the molecular weight of the undegraded VN at an incubation time of 0 min.

Statistical analyses.
The results are expressed as the mean ± standard error (SE). For statistical analyses of data, unpaired t test or one-way analysis of variance (ANOVA) with a post-hoc test was used as appropriate. All analyses were performed using IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA).