The serine proteinase hepsin is an activator of pro-matrix metalloproteinases: molecular mechanisms and implications for extracellular matrix turnover

Increasing evidence implicates serine proteinases in the proteolytic cascades leading to the pathological destruction of extracellular matrices such as cartilage in osteoarthritis (OA). We have previously demonstrated that the type II transmembrane serine proteinase (TTSP) matriptase acts as a novel initiator of cartilage destruction via the induction and activation of matrix metalloproteinases (MMPs). Hepsin is another TTSP expressed in OA cartilage such that we hypothesized this proteinase may also contribute to matrix turnover. Herein, we demonstrate that addition of hepsin to OA cartilage in explant culture induced significant collagen and aggrecan release and activated proMMP-1 and proMMP-3. Furthermore, hepsin directly cleaved the aggrecan core protein at a novel cleavage site within the interglobular domain. Hepsin expression correlated with synovitis as well as tumour necrosis factor α expression, and was induced in cartilage by a pro-inflammatory stimulus. However, a major difference compared to matriptase was that hepsin demonstrated markedly reduced capacity to activate proteinase-activated receptor-2. Overall, our data suggest that hepsin, like matriptase, induces potent destruction of the extracellular matrix whilst displaying distinct efficiencies for the cleavage of specific substrates.


Results
Hepsin induces ex vivo ECM degradation. To test whether hepsin was capable of inducing cartilage matrix destruction, recombinant enzyme was incubated with human OA cartilage explant cultures. Hepsin reproducibly induced collagen release from this tissue (Fig. 1A) although this was consistently significantly lower than that induced by equimolar amounts of matriptase. Hepsin also induced potent aggrecan degradation but the levels of release, compared to matriptase, demonstrated variability between patient samples. As seen previously for matriptase 5,6 , inclusion of the broad-spectrum metalloproteinase inhibitor, GM6001, reproducibly blocked the observed hepsin-induced collagen release (Fig. 1B).
Hepsin is a proMMP activator. To assess whether hepsin can activate proMMPs we used cytokine-stimulated bovine cartilage explant cultures. In this ex vivo model of cartilage breakdown, addition of proMMP activators induces early collagen release 5,19 . As seen previously, low levels of collagen release were observed at day 7 following stimulation with IL-1+OSM 5 . The addition of hepsin led to a significant increase in collagenolysis which was observed at both day 7 and 10 of culture ( Fig. 2A, left panel). These data correlated with the levels of active collagenase detected in the media of these cultures which were similar for both hepsin-and matriptase-treated cartilages ( Fig. 2A, right panel). Hepsin processed proMMP-1 and -3 into proteins corresponding to the molecular masses of the mature enzymes (approximately 43 kDa and 46 kDa, respectively; Fig. 2B, left panel), whilst MMP activity assays confirmed this processing generated active MMP-1 and MMP-3 (Fig. 2B, right panel). Although gelatin zymography indicated proMMP-9 activation following hepsin addition (Fig. 2C), this was most probably due to proMMP-3 activation since MMP activity assays failed to demonstrate hepsin-mediated activation of proMMP-9 as well as proMMP-8 and -13 (Supplementary Figure 1).
Hepsin is an aggrecan-degrading enzyme. The variable level of aggrecan release induced by hepsin ( Fig. 1A) led us to assess whether hepsin was capable of direct aggrecan cleavage. Interestingly, hepsin caused proteolysis of the aggrecan core protein, as demonstrated by anti-G1 immunoblotting (Fig. 3A, left panel). To determine whether hepsin cleaved within the pathologically-relevant aggrecan interglobular domain (IGD), incubation with recombinant G1-IGD-G2 (corresponding to amino acids Val 20 -Gly 675 of the aggrecan core protein) confirmed cleavage within this region (Fig. 3A, right panel) at sites distinct from known metalloproteinase cleavage sites (Fig. 3B, left panel). Matriptase also processed this protein, albeit with markedly weaker efficiency. N-terminal sequencing of the hepsin-mediated cleavage product (Fig. 3B, right panel) indicated a novel cleavage site at Arg 375 ↓Gly 376 in close proximity to the known aggrecanase cleavage site (Fig. 3C).
Hepsin expression correlates with synovial inflammation. Many proteinase genes are induced by pro-inflammatory mediators, including various cytokine combinations such as IL-1+OSM, in chondrocytes and many other cell-types 20,21 . In this study, we found this cytokine combination also markedly induced hepsin expression (Fig. 4A) in cartilage but, interestingly, not matriptase (Supplementary Figure 2). Since inflammation is recognized to be a contributing factor in OA, especially TNFα (reviewed in 22 ), we assessed hepsin expression in OA synovia and found it to be significantly elevated compared to normal (neck of femur; NOF) synovial tissues (Fig. 4B). Notably, there appeared to be two distinct groups in the OA samples: one with very little hepsin expression as per the normal group, whilst the second had markedly higher expression. Interestingly, overall hepsin expression correlated with TNFα expression levels (r = 0.57, p = 0.0034; Supplementary Figure 3). Furthermore, immunohistochemistry confirmed the presence of hepsin in OA synovial samples, which correlated with histological grade of synovitis (Fig. 4C).
Hepsin is a weak PAR2 activator. Since we have previously demonstrated that matriptase directly induced MMP expression from cartilage via PAR2 activation 5 , we investigated the ability of hepsin to activate PAR2. Following topical administration of hepsin to murine joints, as reported previously for matriptase 5 , hepsin induced a rapid and reproducible increase in synovial perfusion in wild type PAR2 +/+ mice compared to PAR2 −/− mice (Fig. 5A). However, although the perfusion assay confirmed hepsin can activate PAR2, this assay requires non-physiological amounts of enzyme. Therefore, we next quantified receptor activation using SW1353 chondrocytes transfected to constitutively express PAR2. Addition of hepsin to these cells reproducibly induced the expression of the PAR2-dependent gene IL8; however, an equimolar amount of matriptase stimulated significantly more IL8 expression (approx. 6-7 times at 50-100 nM) in the PAR2-expressing cells (Fig. 5B). Non-transfected SW1353 cells failed to respond (Supplementary Figure 4). To assess PAR2 activation kinetics, a quenched fluorescent peptide mimicking the PAR2 activation sequence confirmed hepsin cleaved with significantly reduced catalytic efficiency compared to matriptase (Fig. 5C, left panel) albeit with normal Michaelis-Menten kinetics (Fig. 5C, right panel). Finally, we assessed whether subsite specificity and structural differences within the active site architecture of matriptase and hepsin could explain the significantly different levels of PAR2 activation displayed by these proteinases. Substrate specificity comparisons 23 highlighted a marked preference of matriptase for the amino acids in the P2 and P2′ positions of the PAR2 activation sequence (SKGR↓SLIG). To further explore these observations, we conducted molecular docking of the SKGRSL sequence in the active sites of both matriptase and hepsin which indicated that the bulky hydrophobic P2′ Leu sidechain is (77 y, F) were incubated for 3 days whilst donor #3 (66 y, F) was incubated for 7 days. (B) Hepsin or matriptase were added to human OA cartilage in the presence or absence of the metalloproteinase inhibitor GM6001 (10 µM). Collagen and aggrecan release were measured by hydroxyproline and DMMB assays, respectively. Statistical analyses were performed using one-way ANOVA with Bonferroni post-hoc test where: *p < 0.05, ***p < 0.001 versus hepsin; and † † p < 0.01, † † † p < 0.001 versus control; ### p < 0.001 versus hepsin or matriptase only. Data are plotted as mean ± SD and representative of at least 3 independent experiments (n = 4).

Figure 2.
Hepsin is an activator of proMMP-1 and proMMP-3. (A) Bovine nasal cartilage explants were stimulated with IL-1 (1 ng/ml) and OSM (10 ng/ml) in the presence and absence of matriptase or hepsin (100 nM). Media were removed and the cartilage re-stimulated at day 7. The experiment was terminated as described in the experimental procedures, and the percentage collagen release determined by hydroxyproline assay. Left panel: Data are presented as day 7 collagen release and total collagen release (day 7 + 10). Right panel: Collagenase activity in Day 7 media was monitored by a 3 H collagen bioassay. Data are plotted as mean ± SD (n = 6) and representative of 3 independent experiments. Statistical significance was measured by one-way ANOVA with Bonferroni post-hoc test where ***p < 0.001 vs IL-1+OSM. (B) Recombinant full-length proMMP-1 or proMMP-3 were incubated with recombinant hepsin in a 1:5 serine proteinase:MMP molar ratio at 37 °C for up to 24 h. Left panel: Products were separated by 10% SDS-PAGE and gels stained with silver. Right panel: For activity assays, proMMP-1 or proMMP-3 were incubated with hepsin or matriptase for 4 h at 37 °C. Enzyme to substrate ratios were 1:5 for proMMP-1 and 1:10 for proMMP-3. MMP activity was assessed with the MMP-specific substrate FS-6. ProMMP alone or in the presence of APMA were used as negative and positive controls, respectively. Hepsin and matriptase without MMP did not cleave the substrate significantly. Hepsin and matriptase were also pre-treated with the irreversible serine proteinase inhibitor di-isopropyl fluorophosphate (DFP), to demonstrate the involvement of serine proteinase activity in MMP activation. Data are presented as a percentage of APMA-activated proMMP (mean ± SD) following subtraction of the blank measurement. Statistical significance was measured by a one-way ANOVA with Bonferroni post-hoc test where ***p < 0.001 compared to proMMP alone. (C) Day 7 culture media from stimulated bovine cartilage in A were subjected to gelatin zymography. Where gel images have been cropped for clarity, full length images are presented in supplementary information. better accommodated in the deep hydrophobic channel of matriptase compared to the more shallow and polar S2′ pocket offered by hepsin. Furthermore, the Gly residue in the P2 position of the PAR2 peptide has better complementarity to the receptor surface in matriptase, engaging in Van der Waals contacts with a Phe residue that is absent from the more open S2 pocket of hepsin (Fig. 6). These structural data are consistent with the notion that hydrolysis of the PAR2 activation sequence (SKGR↓SLIG) by hepsin is considerably less favourable than that provided by matriptase. Hepsin expression is associated with inflammation. (A) Bovine articular cartilage was excised from metacarpophalangeal joints and stimulated with IL-1 (5 ng/ml) and OSM (10 ng/ml) for 4 days. RNA was isolated and reverse transcribed prior to real-time PCR. Data were normalized to the level of 18 S rRNA and plotted as mean ± SD fold change (n = 4). Significant differences were determined using a t-test where ***p < 0.001. (B) Gene expression levels in hip synovium from patients with OA (closed circles; n = 12) or normal controls (NOF) (open squares; n = 12) of HPN were determined as described in the Experimental procedures and normalized to the level of glyceraldehyde 3-phosphate dehydrogenase (GAPDH). HPN expression was undetected in 8/12 NOF samples and 2/12 OA samples, and significant differences between the normal and OA groups were determined using a two-sided Mann-Whitney U test, where **p < 0.01. (C) Human OA synovia samples were fixed, embedded in paraffin, sectioned and stained with haematoxylin and eosin. Synovitis was scored blinded by two independent scorers based on levels of inflammation, stromal activation and hyperplasia. Summed scores were totaled and described as no synovitis (0-1), slight synovitis (2-3), and moderate synovitis (4-6). Representative samples from within this scoring scale are presented where hepsin was detected by immunohistochemistry using a rabbit anti-hepsin polyclonal antibody. A control consisting of no primary antibody (NPA) was also performed. The scale bars represent 100 µm (except 50 µm for the insets).

Discussion
Despite our understanding of the importance of metalloproteinases in ECM destruction, and the requirement for proMMP activation [24][25][26] , the interplay between serine and metalloproteinases is still an emerging paradigm 4 . Moreover, human OA cartilage is particularly resistant to resorption in vitro, and we have previously observed that although addition of potent cytokine stimuli such as IL-1 + OSM produce high proMMP levels, significant collagen release is rarely observed 6,20 . It is now believed that serine and metalloproteinase cascades are interconnected during cartilage catabolism, and such interplay likely functions in other connective tissue pathologies. The identity of specific serine proteinases, however, has remained elusive. Recent genomic and proteomic advances helped discover entirely novel families of proteinases, including TTSPs 16,18 . Of this family, we previously reported matriptase expression to be elevated in end-stage OA, identifying it to have a putative role as an initiator of cartilage collagen destruction by inducing the expression and, crucially, activation of proMMPs 5 . Herein, we now describe a second TTSP, hepsin, can also activate proMMPs and thus promote ECM turnover. Moreover, as shown previously for matriptase which induced early collagen release 5 , hepsin also increased collagenase activity to comparable levels suggesting similar capabilities in activating relevant proMMPs that promote collagenolysis. In vitro proMMP activation experiments further confirmed hepsin and matriptase display similar propensities for the activation of both proMMP-1 and proMMP-3. Indeed, hepsin activity is important for prostate and ovarian cancer cell invasion, a process linked to metalloproteinase activity in many circumstances [27][28][29][30] .
Data from an OA model now indicate that ECM remodeling events are phasic, with significant changes occurring throughout the disease course primarily as a result of changes to the joint microenvironment 31 ; in reality, these genotypic/phenotypic fluctuations are likely in many pathological matrix remodelling scenarios. Moreover, it is now widely accepted that inflammation (typically cytokine-mediated) is a contributing factor that drives and perpetuates ECM turnover in OA 32,33 as well as being a prime effector of matrix remodeling in many inflammatory pathologies. Our finding that hepsin is markedly up-regulated in chondrocytes in response to the catabolic stimulus IL-1 + OSM 20,21 suggests hepsin expression is dynamic and highly regulated. Furthermore, this expression was consistent with the concomitant up-regulation of MMP substrates such as proMMP-1 and proMMP-3 to effect ECM turnover 21,34 .
The importance of proMMP-3 activators in the destruction of ECMs cannot be understated since this proteinase is capable of activating many other MMPs including MMP-13 35 . Although hepsin activation of proMMP-1 was relatively weak in vitro, proteolytic attack of proMMP-1 by a serine proteinase typically produces a partially active MMP which only becomes fully active following further cleavage by active MMP-3 36,37 . It is well established, for example, that cartilage ECM destruction involves multiple interdependent proteolytic cascades, the corollary of which is that some proteinases must function as upstream initiators. Interestingly, matriptase and hepsin undergo auto-activation, a unique phenomenon consistent with TTSPs functioning in this role 38 although hepsin has also been reported to activate pro-matriptase 17 . With the initial stages of cartilage destruction localised to the pericellular space, somewhat analogously to an 'invasive front' , membrane-anchored serine proteinases are ideally positioned for proMMP activation and the subsequent destruction of the surrounding pericellular matrix. To our knowledge, this is the first report demonstrating the ability of hepsin to activate proMMPs, and may have significant implications in settings where metalloproteinase-dependent matrix turnover is prevalent.
An intriguing finding was that the aggrecan IGD is a substrate for hepsin, cleavage occurring very close to the reported aggrecanase processing site. Our previously reported P1-P1′ specificity data for hepsin 39 are consistent with the observed cleavage at Arg 375 ↓Gly 376 . Although it is likely that the majority of catabolic events in cartilage are effected by metalloproteinases, our data suggest hepsin has the potential to generate an aggrecan peptide (Phe 342 -Arg 375 ) very similar to the bioactive Phe 342 -Glu 373 peptide (generated by combined MMP and aggrecanase activities) known to contribute to cartilage breakdown 40 in the absence of aggrecanase activity. Thus, hepsin acting as a proMMP activator may further exacerbate ECM turnover in aggrecan-rich ECMs. It has also been shown to cleave the structurally ubiquitous basement membrane component laminin-332, as indeed have several MMPs, which helps maintain epithelial-mesenchymal cohesion in tissues exposed to external forces; laminin-332 cleavage may generate pro-migratory signals to cells (see 41 and references therein). Combined with its ability to activate pro-hepatocyte growth factor 42 , hepsin is therefore capable of promoting invasion and metastasis.
Since we have previously demonstrated that matriptase-induced breakdown of OA cartilage is PAR2-dependent 5,6 and that PAR2 antagonism is protective 8 , we investigated the propensity for hepsin to activate PAR2 in comparison to matriptase. We demonstrated that hepsin cleaved and activated PAR2 in vivo, albeit in a somewhat non-physiological setting. Kinetic assessment of PAR2 cleavage however highlighted a significantly reduced capability of hepsin to cleave PAR2 compared to matriptase. Consistent with these findings, hepsin-deficient mice subjected to destabilisation of the medial meniscus (DMM) surgery to induce OA were not   Figure 5). We have previously demonstrated the importance of the matriptase/PAR2 axis in the DMM model 6,8,9 whereby PAR2 activation is a significant contributor to the observed cartilage damage. We hypothesise little or no activation of PAR2 by hepsin in cartilage explains the lack of protection in the DMM-treated hepsin-deficient mice and accounts for the lower levels of hepsin-induced collagen release from human OA cartilage explants compared to matriptase, whilst the similar efficiencies of MMP activation for both TTSPs further support this assumption. Indeed, weak activation of PAR2 by hepsin (relative to matriptase) has been reported previously in a study which indicated that hepsin-mediated PAR2 activation was primarily matriptase-dependent due to its activation by hepsin 17 .
Canonical PAR2 activation occurs at..SKGR↓SLIG.. and experimental substrate preference data 23,39 indicate that matriptase prefers small residues at P2 (Gly/Ala) whilst hepsin is more tolerant and generally prefers larger side chains (Ala/Gln/Pro/Leu). Thus, although Gly at this position in the PAR2 sequence is accepted by hepsin, it is a much more preferred requirement in matriptase. Moreover, structural modelling provides some rationale for the observed substrate difference; Asn 254 in the S2 pocket of hepsin provides much more space but less favourable surface complementarity than the corresponding Phe 708 in matriptase. The other key difference in substrate preference is at the P2′ position where the Leu residue in the PAR2 sequence is well tolerated in matriptase but much less so in hepsin.
As hepsin gene expression is responsive to pro-inflammatory cytokines, and correlates with synovitis from human OA patients, we conclude that hepsin may have a role in cartilage destruction that has a more inflammatory phenotype. Indeed, increasing evidence supports a role for inflammation in OA progression amongst a subset of patients 33 which mirrored our findings, whilst the DMM model is known to have little inflammatory involvement 43 . Thus, herein we demonstrate that hepsin is a proteinase that directly and indirectly promotes pathological ECM alterations (see Fig. 7). Indeed, we propose that its ability to function as an activator of proM-MPs and other proteinase zymogens may have implications in not only arthritis but many other matrix remodeling scenarios such as cancer metastasis and invasion.

Experimental procedures
Reagents and antibodies. All chemicals of the highest purity available were obtained from Sigma Chemical Co (Poole, UK) unless otherwise stated. All cytokines and proteins used were recombinant human unless stated otherwise. IL-1α was a generous gift from Dr Keith Ray (GlaxoSmithKline, Stevenage, UK). Oncostatin M (OSM), proMMP-1, proMMP-13, matriptase (aa596-895) and hepsin (aa45-417) were produced in-house (see 6,12 and references therein). Purified serine proteinases were active site titrated with the burst titrant 4-methylumbelliferyl-p-guanidinobenzoate. ProMMP-3 and proMMP-8 were purchased from R&D Systems (Abingdon, UK). GM6001 was purchased from Calbiochem (Nottingham, UK). The anti-G1 antibody was kindly Figure 7. Schematic of the direct and indirect contributions of hepsin to alter ECM integrity. Firstly, hepsin activity is known to modulate cell motility via laminin 332 52 . Hepsin activates the uPA/plasminogen/plasmin cascade and also processes latent hepatocyte growth factor (HGF) leading to HGF/cMet signaling; together, these promote metastasis and altered epithelial cohesion 30,53 . Moreover, hepsin can activate pro-matriptase 17 which leads to PAR2 activation and subsequent expression of proMMP-3 and proMMP-1 5 as well as inflammatory mediators including IL-8 54 leading to chemotaxis of inflammatory cells such as neutrophils and mast cells. Plasmin 55 , as well as hepsin, generate active MMP-3 which is a prime activator of many proMMPs including proMMP-1 and proMMP-13, key collagen-degrading proteinases. Inflammation induces the expression of prohepsin which auto-activates 17,38 and can directly activate proMMP-1 as well as cleave within the aggrecan IGD to promote both aggrecanolysis and collagenolysis to further impair the structural integrity of tissues.
Cartilage explant culture. Explants were taken from macroscopically normal areas of cartilage from OA patients, undergoing total knee replacement at the Freeman Hospital, Newcastle. All tissue was obtained following informed consent from patients, and processing and handling carried out in accordance with University guidelines with appropriate ethical approval from NHS Health Authority NRES Committee North East -Newcastle and North Tyneside 1 (REC14/NE/1212). Cartilage was prepared and cultured as described previously 6 , and stimulated with recombinant TTSP for either 3 or 7 days. For bovine cartilage explants, nasal septum (obtained from a local abattoir) was prepared as previously described 20 . Two discs were plated in each well of a 48-well plate and incubated in serum-free medium overnight. Cartilage was stimulated with IL-1 + OSM (1 ng/ml and 10 ng/ ml, respectively) in the presence or absence of matriptase or hepsin (100 nM). Cartilage was re-stimulated at day 7 and the experiment terminated following complete resorption of ≥1 of the wells (typically day 10-12). Media were removed and stored at −20 °C prior to analysis. The remaining cartilage was digested with papain 20 . To analyse cartilage destruction, hydroxyproline levels were used as an estimate for collagen release into conditioned medium, whilst aggrecan release was measured by assessing levels of glycosaminoglycan (GAG) by dimethyl methylene blue (DMMB) assay (see 20 and the references therein). Collagenolytic activity present in culture media from cartilage explants was determined using a diffuse fibril assay with 3 H acetylated collagen 45 . Gelatin zymography was performed as previously described 25 . For cytokine experiments, bovine articular explants were stimulated for a total of 96 hours with IL-1α (5 ng/ml) and oncostatin-M (10 ng/ml). Explants were snap frozen and homogenised in TRIzol ® (Invitrogen, Paisley, UK) in liquid nitrogen using a dismembrator (Sartorious, Epsom, UK), RNA extracted and complementary DNA synthesised using Superscript III reverse transcriptase according to the manufacturer's instructions (Invitrogen). Real-time PCR was performed using a MxPro3000 QPCR system (Agilent Technologies, Stockport, UK) using ReadyMix ™ Taq PCR reaction mix in conjunction with a Taqman ® gene expression assay (Hpn: Bt03260765_m1; ThermoFisher Scientific (Loughborough, UK)). 18 S was used as a reference gene as previously described 46 .
Synovial Tissue and Immunohistochemistry. Human synovia were generously provided by Prof Ian Clark (University of East Anglia, Norwich, UK) from patients undergoing total hip replacement surgery at the Norwich and Norfolk Hospital, Norwich, UK for OA (6 M/6 F; mean age = 74.3 yrs (±3.6 SD), range 68-78 yrs) or NOF fracture (2 M/10 F; mean age = 76 yrs (±10 SD), range 52-89 yrs). Human knee synovia were obtained from OA patients undergoing total knee replacement surgery (3 M/8 F; mean age = 65.6 yrs (±8.4 SD), range 54-86 yrs), at the Freeman Hospital, Newcastle, UK. All tissue was obtained following informed consent from patients, and processing carried out in accordance with University guidelines and with ethical approval from either Norfolk Research Ethics Committee (REC05/Q010/196) or NHS Health Authority NRES Committee North East -Newcastle and North Tyneside 1 (REC14/NE/1212). RNA was isolated from hip synovia and subjected to TaqMan low density array PCR as previously described 5,48 . Knee synovial tissue was fixed overnight in 10% formalin and paraffin embedded as described 6 . Sections were generated at 4 μm intervals and stained with haemotoxylin and eosin as described previously 19 . Synovitis was scored according to a previously described system 49 . For immunohistochemistry, sections were de-waxed in xylene followed by decreasing alcohol gradients and antigen retrieval performed by immersion in hot 10 mM sodium citrate, pH 6.0, for 6 minutes. Endogenous peroxidase activity was quenched in 0.3% (v/v) H 2 O 2 for 30 minutes. A biotin block was performed according to the manufacturer's instructions (Vector Laboratories, CA, USA), followed by blocking in 5% horse serum. Hepsin antibody at 1:100 dilution (Cayman Chemicals, Michigan, USA) was applied for 30 minutes. After washing, detection was performed using an ABC peroxidase staining kit (Thermofisher Scientific, Loughborough, UK) according to the manufacturer's instructions, followed by incubation with 3, 3′-diaminobenzidine tetra-hydrochloride (DAKO, Ely, UK) for 5 minutes. Sections were washed in tap water, counterstained with Mayer's haemotoxylin and slides mounted. Images were taken using a Leica Biosystems DM4000B light microscope. PAR2 assays. All mice were C57BL/6 J background, and housed in standard cages. All procedures were approved and conducted in accordance with UK Home Office legislation (project licence number 70/8790) and institutional regulations. Food and water was available ad libitum and all animals were kept in a thermoneutral environment. In vivo PAR2 activation was assessed by consecutive laser Doppler imaging (Moor Instruments, Axminster, UK) of synovial perfusion following administration of hepsin (5 µg) to PAR2 +/+ and PAR2 −/− mice, as described previously 5 . For cell experiments, full-length PAR2 was amplified from a commercially available shuttle vector (R&D Systems, Abingdon, UK) and cloned into a pcDNA3.1 vector using an Infusion-HD kit (Clontech, CA, USA) according to the manufacturer's protocol. 5′-TACCGAGCTCGGATCGCCACCATGGACTACAAGGATGACGATGACAAGCGAAGTCCTAGTGCT-3′ and 5′-AAACGGGCCCTCTAGTTAATAGGAGGTCTTAACAG-3′ were used as forward and reverse primers, respectively, which provided an N-terminal FLAG tag and kozak sequence. SW1353 chondrosarcoma cells were cultured as described previously 50 . For transfection, 2.75 µg plasmid DNA with 8.25 µl Fugene-HD transfection reagent (Promega, Southampton, UK) were added to cells for 48 hours in antibiotic-free medium followed by selection in medium containing G418 disulphate (neomycin analogue, 400 µg/ml). Cells were incubated with matriptase or hepsin (0-100 nM) for 3 hours. RNA was stabilised in Cells-to-cDNA lysis buffer and converted to cDNA by Moloney-Murine Leukemia Virus reverse transcriptase as directed (Invitrogen). IL-8 (CXCL8) expression was monitored by TaqMan real-time PCR (forward primer: 5′-AGACAGCAGAGCACACAAGC-3′; reverse primer: 5′AGGAAGGCTGCCGAGAG-3′; Probe: #72 from Universal Probe Library (Roche, Welwyn Garden City, UK), and normalised to 18 S ribosomal RNA levels (primers and probe as described in 51 ). The kinetics of PAR2 cleavage by hepsin or matriptase (both 1 nM) were assessed in pre-warmed 100 mM Tris pH 7.5, 150 mM NaCl, 500 µg/ml bovine serum albumin, 0.01% (w/v) Brij 35 before addition of 2-Abz-SKGRSLIG-Y(NO 2 ) at a range of concentrations (up to 200 µM). Progress curves were monitored over 30 minutes at 37 °C, and slopes from the linear portions of each curve plotted against substrate concentration and kinetic constants generated by non-linear regression using GraphPad Prism 7.0 software. Molecular Modelling. All operations were performed within the Schrödinger Molecular Modelling Suite version 2017-1 (NY, USA). Crystal structures of matriptase and hepsin were downloaded from the Protein Data Bank (3PF8 and 1ZG8, respectively) and aligned using the backbone atoms of chain A. The amino acid sequence RCTKSI of the bound peptide (chain I) was extracted from 3PF8 and mutated to the PAR2 cleavage sequence SKGRSL. The termini of the resulting PAR2 peptide were amide capped and restrained minimization performed using Impact. Glide SP was used to obtain a rough positioning of the peptide in the binding sites of matriptase and hepsin (flexible peptide, rigid receptor), followed by refinement using Prime MM-GBSA, allowing simultaneous sampling of peptide and receptor, with full flexibility for residues within 6 Å of the peptide. Statistical analysis. Statistical differences between sample groups were assessed using ANOVA with a post-hoc Bonferroni's multiple comparison test. For comparisons between only two samples, a two-tailed Student's t-test was performed. A Pearson correlation co-efficient calculation was performed for synovial gene expression data. All statistical analyses were performed using Graphpad Prism 7.0 software where ***p < 0.001; **p < 0.01; *p < 0.05. Equipment and Settings. All gel and blot images were taken using Syngene using GeneSnap software.
Western blots were detected after 5 minutes in enhanced chemiluminescent (ECL) substrate. Images were captured under 'no light' conditions with 5 × 30 seconds exposure time. Upper white images were taken to determine the position of the ladder, and composite images generated using GeneSnap. Images of silver stain gels were taken using the 'transilluminator' setting, using a visible light converter screen to produce an evenly illuminated area of white light. Any changes to brightness or contrast were applied uniformly across the whole gel or blot. Data Availability. The datasets generated/analysed during the current study are available from the corresponding author on reasonable request.