Collagenolytic matrix metalloproteinases antagonize proteinase-activated receptor-2 activation, providing insights into extracellular matrix turnover

The collagenase subfamily of matrix metalloproteinases (MMPs) have important roles in the remodeling of collagenous matrices. The proteinase-activated receptor (PAR) family has a unique mechanism of activation requiring proteolysis of an extracellular domain forming a neo-N terminus that acts as a tethered ligand, a process that has been associated with the development of arthritis. Canonical PAR2 activation typically occurs via a serine proteinase at Arg36-Ser37, but other proteinases can cleave PARs downstream of the tethered ligand and “disarm” the receptor. To identify additional cleavage sites within PAR2, we synthesized a 42-amino-acid peptide corresponding to the extracellular region. We observed that all three soluble MMP collagenases, MMP-1, MMP-8, and MMP-13, cleave PAR2 and discovered a novel cleavage site (Ser37-Leu38). Metalloproteinases from resorbing bovine nasal cartilage and recombinant human collagenases could cleave a quenched fluorescent peptide mimicking the canonical PAR2 activation region, and kinetic constants were determined. In PAR2-overexpressing SW1353 chondrocytes, we demonstrated that the activator peptide SLIGKV-NH2 induces rapid calcium flux, inflammatory gene expression (including MMP1 and MMP13), and the phosphorylation of extracellular signal-regulated kinase (ERK) and p38 kinase. The corresponding MMP cleavage-derived peptide (LIGKVD-NH2) exhibited no canonical activation; however, we observed phosphorylation of ERK, providing evidence of biased agonism. Importantly, we demonstrated that preincubation with active MMP-1 reduced downstream PAR2 activation by a canonical activator, matriptase, but not SLIGKV-NH2. These results support a role for collagenases as proteinases capable of disarming PAR2, revealing a mechanism that suppresses PAR2-mediated inflammatory responses.

The collagenase subfamily of matrix metalloproteinases (MMPs) have important roles in the remodeling of collagenous matrices. The proteinase-activated receptor (PAR) family has a unique mechanism of activation requiring proteolysis of an extracellular domain forming a neo-N terminus that acts as a tethered ligand, a process that has been associated with the development of arthritis. Canonical PAR2 activation typically occurs via a serine proteinase at Arg 36 -Ser 37 , but other proteinases can cleave PARs downstream of the tethered ligand and "disarm" the receptor. To identify additional cleavage sites within PAR2, we synthesized a 42-amino-acid peptide corresponding to the extracellular region. We observed that all three soluble MMP collagenases, MMP-1, MMP-8, and MMP-13, cleave PAR2 and discovered a novel cleavage site (Ser 37 -Leu 38 ). Metalloproteinases from resorbing bovine nasal cartilage and recombinant human collagenases could cleave a quenched fluorescent peptide mimicking the canonical PAR2 activation region, and kinetic constants were determined. In PAR2-overexpressing SW1353 chondrocytes, we demonstrated that the activator peptide SLIGKV-NH 2 induces rapid calcium flux, inflammatory gene expression (including MMP1 and MMP13), and the phosphorylation of extracellular signal-regulated kinase (ERK) and p38 kinase. The corresponding MMP cleavage-derived peptide (LIGKVD-NH 2 ) exhibited no canonical activation; however, we observed phosphorylation of ERK, providing evidence of biased agonism. Importantly, we demonstrated that preincubation with active MMP-1 reduced downstream PAR2 activation by a canonical activator, matriptase, but not SLIGKV-NH 2 . These results support a role for collagenases as proteinases capable of disarming PAR2, revealing a mechanism that suppresses PAR2-mediated inflammatory responses.
Proteinases are enzymes that effect proteolysis of intact proteins via the hydrolysis of peptide bonds (1) and are essential in a myriad of situations in both health and disease such as extracellular matrix (ECM) 3 remodeling during development or pathological destruction. In humans, almost two-thirds of the proteinases that comprise the degradome (2) exert their activities within the extracellular space, making them prime candidates in events that entail remodeling and/or degradation of the ECM. These can be subdivided into metallo-and serine proteinases, and in this context these proteinases have until recently been viewed as mediating the proteolytic degradation of ECM components as occurs for example in tumor invasion or arthritis. However, it is now clear that proteinases also perform precise and specific functions in both homeostasis and disease. The cleavage of triple-helical collagens such as type II collagen (the major structural component present in cartilage) by the soluble collagenase subset of matrix metalloproteinases (MMP-1, MMP-8, and MMP-13) represents an example of highly specific proteolysis whereby only a single peptide bond is cleavable under normal physiological conditions (3). This initial cleavage then leads to collagen denaturation, concomitant with increased susceptibility to other less specific proteinases and disassembly of the collagen network, which is a requirement as cartilage is remodeled into bone during development, but also occurs during cartilage breakdown in degenerative scenarios such as rheumatoid arthritis and osteoarthritis (OA).
Proteinase-activated receptors (PARs) represent one mechanism by which proteinases can delicately regulate cellular responses and are likely to be involved in many disease pro- cro ARTICLE cesses. The importance of PAR2 in the pathogenesis of both inflammatory arthritis and OA has recently emerged, and PAR2 deficiency provides protection in experimental arthritis models (4 -9). We have demonstrated that an activator of PAR2, matriptase, induces cartilage destruction in a PAR2-and metalloproteinase-dependent manner (10,11) and that a related serine proteinase, hepsin, can also activate PAR2, albeit in a markedly less efficient manner (12). Indeed, in complex disease environments (such as arthritis) multiple proteinases could be present to act on PAR2, making our understanding of how PAR2 is activated and regulated important (for a review, see Ref. 13).
PAR2 exhibits receptor plasticity with biased signaling, different proteinases eliciting different downstream responses following cleavage of the extracellular domain, which may be of particular importance in modulating inflammatory responses (see Refs. 14 and 15). Here, we investigated the effect of collagenolytic MMPs, known to be expressed during ECM remodeling events (16), on PAR2. We identified novel cleavages by MMP-1, -8, and -13, and a primary cleavage site common to all was identified. Collagenase cleavage resulted in antagonism of PAR2 activation, suggesting an important negative feedback mechanism whereby canonical PAR2 activation induces MMP expression, and MMP activity can subsequently antagonize PAR2.

MMP collagenases cleave the PAR2 extracellular domain
To determine which collagenolytic MMPs can cleave the extracellular PAR2 sequence, we synthesized a 42-amino-acid peptide, corresponding to the N-terminal extracellular region of PAR2 (PAR2 31-72 peptide) where previous PAR2 cleavage sites have been identified (17)(18)(19) (Fig. 1A). After confirming the peptide was cleaved by known PAR2-cleaving proteinases (Fig. 1B), the peptide was incubated with APMA-activated MMP-1, MMP-8, and MMP-13. Silver-stained SDS-PAGE gels of the digested products demonstrated that all three collagenases were capable of cleaving this sequence in a dose-dependent manner (Fig. 1C). Cleavage was subsequently confirmed by reverse-phase HPLC (Fig. 2, A-C), and peaks were collected and subjected to electrospray mass spectrometry (MS). Exact mass identifications revealed cleavage sites for each collagenase (Fig. 2D), but interestingly, time-course experiments identified a primary cleavage common to all, following Ser 37 (Fig. 3). This novel site is adjacent to the canonical cleavage site described for trypsinlike serine proteinases, Arg 36 .

Collagenases cleave a PAR2 synthetic peptide substrate
To further explore and determine kinetic constants for collagenase cleavage of PAR2, a quenched fluorescent peptide

MMPs antagonize PAR2 signaling
mimic of PAR2 was employed (2-Abz-SKGRSLIG-Y(NO 2 )). To demonstrate cleavage by native MMPs, the peptide was incubated with conditioned medium from 14-day IL-1 ϩ OSMstimulated bovine nasal cartilage, a well-established model of proteinase-driven cartilage breakdown (10,20,21). Cleavage was observed; however, this was completely abolished by the addition of the MMP inhibitor GM6001 but not by serine or cysteine proteinase inhibition (Fig. 4A). Subsequently, recombinant APMA-activated MMP-1, -8, and -13 were shown to cleave the peptide, with the effect dependent on active-site accessibility (Fig. 4B). Velocity was calculated from progress curves at variable concentrations of substrate, and the resultant plots are consistent with each collagenase exhibiting Michaelis-Menten kinetics (Fig. 4C). APMA-activated MMP-1, MMP-8, and MMP-13 exhibited similar K m values for this substrate, 36 -53 M, which are similar to that of matriptase (46.6 M). MMP-13 exhibited the highest turnover number (k cat ) of all three collagenases and therefore the highest catalytic efficiency (k cat /K m ), although this was markedly lower than that of matriptase (Fig. 4D).

Canonical PAR2 activation induces collagenolytic MMP expression
To generate a model of chondrocytes expressing higher levels of PAR2 (as is the case in OA (22, 23)), we overexpressed PAR2 using a lentivirus transduction system. SW1353 cells transduced with a PAR2-expressing lentivirus (hereafter SW1353-PAR2; dependence of PAR2 expression validated in Fig. S1) were stimulated with the PAR2 activator peptide SLIGKV-NH 2 , which resulted in calcium mobilization (Fig.  5A). Furthermore, stimulation for varying durations resulted in ATF3, MMP1, and MMP13 gene expression (Fig. 5, B and C). ATF3 is a transcription factor we recently identified as regulating MMP13 expression following cytokine stimulation (24). Secreted MMP-1 and MMP-13 were subsequently detected in the culture medium 48 h poststimulation (Fig. 5D).

The N terminus generated by collagenase cleavage does not activate canonical PAR2 signaling
Due to the proximity of the primary collagenase cleavage site to the canonical PAR2 activation sequence, the effect of the  Table S1. mAU, milli-absorbance units.

Collagenase cleavage antagonizes PAR2
To determine the effect of collagenase cleavage on the potential for PAR2 to be subsequently activated by canonical PAR2 activators, we performed preincubation experiments. When recombinant active MMP-1 was preincubated with SW1353-PAR2 cells, matriptase-induced PAR2 activation was abrogated, but importantly, activation by SLIGKV-NH 2 was not (Fig. 7A). There was no evidence of ATF3 induction by MMP-1 on either SW1353 control cells or SW1353-PAR2 cells (Fig. 7B). SW1353-PAR2 cells preincubated with LIG-

Discussion
It is becoming increasingly clear that proteolysis not only mediates catabolic events but can also act to precisely regulate cellular processes. One such example, the PARs, represent a way in which controlled cleavage of a substrate by different proteinases leads to different downstream events and outcomes. In this study, we outline such regulation of PAR2 by a family of proteinases that, to our knowledge, have not been investigated before in this context. We demonstrated that the collagenases MMP-1, -8, and -13 can cleave the PAR2 extracellular region, with MMP-1 yielding a single product resulting from cleavage at Ser 37 -Leu 38 with an additional cleavage at Val 68 -Leu 69 observed following MMP-8 and -13 incubations. These cleavage sites are in agreement with the described substrate specificities of the collagenases, which show a preference for a hydrophobic residue in the P1Ј position, a basic or hydrophobic amino acid in the P2Ј position, and a small amino acid in the P3Ј position (28 -30). Thus, the Ser 37 -Leu 38 site fits this preference with leucine, isoleucine, and glycine in P1Ј-P3Ј, respectively, whereas the Val 68 -Leu 69 site fits at the P1Ј and P3Ј positions (leucine and glycine, respectively). Taken together with time-course experiments, it is likely that Ser 37 -Leu 38 is the primary collagenase cleavage site on PAR2. It remains possible, however, that both cleavages could hold functional relevance. For example, following canonical PAR2 activation by mast cell tryptase, signaling has been shown to be attenuated by a secondary downstream cleavage (31). ; presented graphs show combined means (each with n ϭ 2 technical replicates) of three independent experiments) were generated using TIMP-1-titrated APMA-activated recombinant pro-MMP-1, -8, and -13 (C). The hydrolysis of substrate was quantified (nM⅐s Ϫ1 ) using a standard curve determined by total substrate hydrolysis, and nonlinear regression analysis was performed to generate kinetic constants K m and V max . k cat was subsequently calculated from V max and active enzyme concentration. Tabulated kinetic constants (mean Ϯ S.D.) are from three independent experiments. Matriptase kinetic parameters are included for comparison (D). Selected statistical comparisons were performed using Student's two-tailed unpaired t tests where *** indicates p Ͻ 0.001. All error bars represent S.D.
The kinetic profiles of PAR2 cleavage by the collagenases were explored and compared with matriptase, a potent activator of PAR2 (12). Unsurprisingly, all three collagenases were less efficient at cleaving the substrate compared with matriptase (which exhibited similar kinetics to previously published findings (32,33)), primarily as a result of lower k cat values, with broadly similar K m values. When considering the enzyme kinetics of the collagenases, it is important to note that APMA-activated pro-MMPs were utilized; they are not considered to be fully active with 10 -25% of maximal activity typically observed when compared with MMP-3 activated pro-MMP (34); thus, the collagenase kinetics for PAR2 may be underrepresented compared with in vivo cleavage. Furthermore, the abundance of MMPs is an important consideration as tissues undergoing active remodeling, such as during cartilage destruction in arthritis, have high expression levels of MMPs (35)(36)(37). Taken together, these data suggest that the collagenases and matriptase can bind PAR2 with broadly similar affinities, but the collagenases are less efficient at turning over the substrate.
Data presented within the present study demonstrate that the addition of active MMP-1 to PAR2-expressing cells followed by the activation of PAR2 by matriptase leads to an attenuated level of activation, importantly a result not observed following SLIGKV-NH 2 activation. This observation can be interpreted as cleavage of PAR2 by MMP-1 in a classical disarming mechanism similar to cathepsin G or proteinase 3 disarming of PAR2 (17,38). This would entail the cleavage of PAR2 at Ser 37 -Leu 38 , thus removing the canonical activation site to leave a disarmed receptor on the surface (as evidenced by full SLIGKV-NH 2 activation).
Cleavage of cell-surface receptors by collagenases has previously been described, with both MMP-1 (39) and -13 (40) shown to be able to activate the PAR1 receptor by cleavage at two amino acids upstream or a single amino acid downstream of the canonical site, respectively. Furthermore, low levels of trypsin-activated MMP-1 have been previously described to up-regulate monocyte chemoattractant protein-1 in A549 cells (which endogenously express PAR2); however, no mechanism or cleavage site was described (41).
Both PAR2 (23) and matriptase (10) are expressed at significantly higher levels in OA chondrocytes than in normal chondrocytes, which may contribute to the elevated levels of MMP-1 and MMP-13 expression observed in the disease (22). We have previously shown matriptase to induce MMP expression in explant OA cartilage (10), and herein we show that PAR2 activation in chondrocytes leads to the expression and secretion of both these collagenases. To our knowledge, the expression of MMP-1 and MMP-13 in chondrocytes following PAR2 activa-
Biased agonism following activator peptide stimulation of PAR2 has previously been explored. The peptide SLAAAA-NH 2 has been shown to induce ERK1/2 phosphorylation in the absence of calcium mobilization (45) as observed with LIG-KVD-NH 2 in the present study. In rat PAR2, ALIGRL-NH 2 peptide (based on the rat canonical activation sequence, SLI-GRL) has been shown to result in detectable canonical PAR2 activation (as measured by calcium mobilization) with no detectable activation when substituting the P2Ј leucine for alanine (SAIGRL-NH 2 ), identifying a key role for the P2Ј leucine in canonical activation (46). In the present study, this P2Ј leucine becomes the P1Ј leucine, and no calcium mobilization was detected. Furthermore, the presence of a basic amino acid in P5Ј has been postulated as a requirement for calcium mobilization (47), which is lost in LIGKVD compared with SLIGKV (underlined). Data presented within the present study suggest competitive binding in the extracellular region of PAR2 between the two similar sequences, SLIGKV and LIGKVD,

MMPs antagonize PAR2 signaling
despite their differing downstream effects. The functional consequences of ERK1/2 phosphorylation induced by the collagenase-derived peptide requires further investigation, although this does not involve the activation of canonical PAR2 signaling pathways or downstream induced genes such as ATF3 and MMP1. The data presented herein support a mechanism by which collagenolytic MMPs, the expression of which is induced by canonical PAR2 activation, negatively regulate PAR2 by abrogating its canonical activation (Fig. 8).

Cell culture
SW1353 chondrosarcoma cells were purchased from American Type Culture Collection (catalog number HTB-94; Manassas, VA). Cells were cultured at 37°C in DMEM/F-12 medium supplemented with 2 mM L-glutamine, 100 IU ml Ϫ1 penicillin, 100 g ml Ϫ1 streptomycin, and 10% fetal bovine serum (all Figure 7. MMP-1 is antagonistic to canonical PAR2 activation. SW1353-PAR2 cells were pretreated with either 1000 nM active MMP-1 or serum-free medium for 120 min prior to the addition of 10 nM matriptase or 100 M SLIGKV-NH 2 for an additional 60 min, and RT-qPCR was performed for ATF3 (A). SW1353-PAR2 or empty vector control cells were stimulated with either 1000 nM active MMP-1 or 100 M SLIGKV-NH 2 for 60 min, and RT-qPCR was performed for ATF3 (B). SW1353-PAR2 cells were pretreated with 100 M LIGKVD-NH 2 or serum-free medium for 120 min prior to the addition of 100 M SLIGKV-NH 2 for an additional 60 min, and RT-qPCR was performed for ATF3 (C). SW1353-PAR2 cells were pretreated with 100 M DVKGIL-NH 2 or serum-free medium for 120 min prior to the addition of 100 M SLIGKV-NH 2 for an additional 60 min, and RT-qPCR was performed for ATF3 (D). Data are expressed relative to GAPDH and presented as -fold change compared with basal expression (mean Ϯ S.D., n ϭ 6) and are representative of at least three independent experiments. Selected statistical comparisons were performed using Student's two-tailed unpaired t tests where *** indicates p Ͻ 0.001. All error bars represent S.D.

MMPs antagonize PAR2 signaling
from Thermo Fisher, Paisley, UK). Cells were starved in serumfree medium for at least 6 h prior to stimulation.
SW1353 cells at 50% confluence were transduced for 48 h with lentivirus in serum-containing DMEM/F-12 supplemented as above with the addition of 4 g ml Ϫ1 Polybrene at a 1:200 concentrated virus:medium ratio. Cells were assessed for successful transduction by examining GFP expression using inverted fluorescence microscopy and then serum-starved for a minimum of 6 h prior to use.

Peptide digestion and visualization
A 42-amino-acid peptide corresponding to amino acids 31-72 of human PAR2 (H 2 N-RSSKGRSLIGKVDGTSHVTGK-GVTVETVFSVDEFSASVLTGK-COOH) was synthesized (see supporting methods) (PAR2 31-72 peptide). The peptide was incubated at a final concentration of 10 M with various proteinases at 37°C for varying durations (see supporting methods for full details of incubation buffers).
For visualization, samples were resolved on 20% polyacrylamide Tris-Tricine gels (49), which were subsequently fixed for 1 h in 12% TCA, 30% methanol followed by silver staining using a Plus One silver staining kit (GE Healthcare) according to the manufacturer's instructions. Analytical reversed-phase HPLC and infusion electrospray MS were performed as described previously (50) with some modifications (see supporting methods).

Calcium mobilization assay
Following serum starvation of SW1353 cells transduced in black-walled clear-bottom 96-well plates, cells were washed with 200 l of Hank's balanced salt solution supplemented with 20 mM HEPES, pH 7, and 2 mM CaCl 2 (HHBS/Ca 2ϩ ). To each well, 50 l of a 5 M stock solution of the calcium indicator Rhod-4-AM (Santa Cruz Biotechnology, Heidelberg, Germany) in HHBS/Ca 2ϩ supplemented with 2.5 mM probenecid (Santa Cruz Biotechnology) and 0.02% Pluronic F-127 (Thermo Fisher) was added and incubated at room temperature in the dark for 45 min. Cells were then washed with HHBS/Ca 2ϩ before incubating at room temperature for 20 min in HHBS/ Ca 2ϩ to allow for dye de-esterification.
Assays were performed in a FLUOstar OPTIMA fluorimeter (BMG Labtech) at 37°C during which its injection pumps were each primed with 2 ml of HHBS/Ca 2ϩ before priming with 1 ml of test solutions. The test solution in pump A was always at 2ϫ final concentration, whereas the test solution in pump B was at 3ϫ final concentration. To perform the assay, each well was replaced with 30 l of HHBS/Ca 2ϩ and placed in the fluorimeter programmed to read in single-well mode (readings taken every second at ex 520 nm and em 590 nm). For each assay, a 10-s baseline was established before a 30-l injection from pump A and a 90-s incubation followed by a 30-l injection from pump B.

Active matrix metalloproteinase-1 production
Purified active MMP-1 was produced for cell culture work to avoid addition of APMA on cells. Recombinant human pro-MMP-1 was expressed in Escherichia coli and refolded from chaotrope-solubilized inclusion bodies as described previously (52). The refolded proenzyme was purified using an automated two-step procedure on an ÄKTAxpress system (GE Healthcare): 1) immobilized metal ion affinity chromatography (IMAC) using a Ni 2ϩ -charged 5-ml HiTrap IMAC FF column with 50 mM Tris, pH 7.4, 150 mM NaCl, and 10 mM CaCl 2 (TNC) as binding buffer and an elution buffer of TNC supplemented with 0.5 M imidazole and 2) size-exclusion chromatography (SEC) performed on a HiLoad 26/60 Superdex 75 prep grade column with TNC as running buffer. Following incubation with APMA and MMP-3 (53), the activated MMP-1 was repurified by tandem IMAC-SEC as above. Aliquots of 5 M MMP-1 were lyophilized from TNC buffer supplemented with 1% (w/v) BSA and stored at Ϫ80°C until required.

Assessment of PAR2 receptor antagonism
Lyophilized active MMP-1 was resuspended to 1 M in DMEM/F-12 medium supplemented with 10 mM CaCl 2 and subsequently added to cells for the required time. Matriptase or SLIGKV-NH 2 was added at twice the final concentration, and cells were incubated for the required time at 37°C prior to RNA extraction (see below).

Gene expression analyses
RNA was extracted from cells in a 96-well format using Cells-to-cDNA lysis buffer (Thermo Fisher) according to the manufacturer's instructions prior to cDNA synthesis by Moloney murine leukemia virus reverse transcriptase (Thermo Fisher) according to the manufacturer's instructions. Real-time quantitative (q) PCR was performed using TaqMan Fast Advanced Master Mix with the following cycling conditions (QuantStudio 3, Thermo Fisher): 2 min at 50°C, 20 s at 95°C followed by 40 cycles of 1 s at 95°C, and 20 s at 60°C. Primer and probe sequences for MMP1, MMP13 (54), and ATF3 (24) were as described previously. GAPDH expression was measured using forward primer 5Ј-GTGAACCATGAGAAG-TATGACAAC-3Ј, reverse primer 5Ј-CATGAGTCCTTCCA-CGATACC-3Ј, and probe FAM-CCTCAAGATCATCAG-CAATGCCTCCTG-TAMRA.

Enzyme-linked immunosorbent assay (ELISA)
Total MMP-1 was measured as described previously (55), and total MMP-13 was measured using a Human Total MMP-13 DuoSet ELISA kit (R&D Biosystems) according to the manufacturer's instructions.

Statistical analyses
Statistical differences between sample groups were assessed using the Student's two-tailed unpaired t test where *** indicates p Ͻ 0.001, ** indicates p Ͻ 0.01, and * indicates p Ͻ 0.05. For clarity, only selected comparisons are presented in some figures.