Matrix metalloproteinase‐13 is fully activated by neutrophil elastase and inactivates its serpin inhibitor, alpha‐1 antitrypsin: Implications for osteoarthritis

Matrix metalloproteinase‐13 (MMP‐13) is a uniquely important collagenase that promotes the irreversible destruction of cartilage collagen in osteoarthritis (OA). Collagenase activation is a key control point for cartilage breakdown to occur, yet our understanding of the proteinases involved in this process is limited. Neutrophil elastase (NE) is a well‐described proteoglycan‐degrading enzyme which is historically associated with inflammatory arthritis, but more recent evidence suggests a potential role in OA. In this study, we investigated the effect of neutrophil elastase on OA cartilage collagen destruction and collagenase activation. Neutrophil elastase induced significant collagen destruction from human OA cartilage ex vivo, in an MMP‐dependent manner. In vitro, neutrophil elastase directly and robustly activated pro‐MMP‐13, and N‐terminal sequencing identified cleavage close to the cysteine switch at 72MKKPR, ultimately resulting in the fully active form with the neo‐N terminus of 85YNVFP. Mole‐per‐mole, activation was more potent than by MMP‐3, a classical collagenase activator. Elastase was detectable in human OA synovial fluid and OA synovia which displayed histologically graded evidence of synovitis. Bioinformatic analyses demonstrated that, compared with other tissues, control cartilage exhibited remarkably high transcript levels of the major elastase inhibitor, (AAT) alpha‐1 antitrypsin (gene name SERPINA1), but these were reduced in OA. AAT was located predominantly in superficial cartilage zones, and staining enhanced in regions of cartilage damage. Finally, active MMP‐13 specifically inactivated AAT by removal of the serine proteinase cleavage/inhibition site. Taken together, this study identifies elastase as a novel activator of pro‐MMP‐13 that has relevance for cartilage collagen destruction in OA patients with synovitis.

Matrix metalloproteinase-13 (MMP-13) is a uniquely important collagenase that promotes the irreversible destruction of cartilage collagen in osteoarthritis (OA). Collagenase activation is a key control point for cartilage breakdown to occur, yet our understanding of the proteinases involved in this process is limited. Neutrophil elastase (NE) is a well-described proteoglycandegrading enzyme which is historically associated with inflammatory arthritis, but more recent evidence suggests a potential role in OA. In this study, we investigated the effect of neutrophil elastase on OA cartilage collagen destruction and collagenase activation. Neutrophil elastase induced significant collagen destruction from human OA cartilage ex vivo, in an MMPdependent manner. In vitro, neutrophil elastase directly and robustly activated pro-MMP-13, and N-terminal sequencing identified cleavage close to the cysteine switch at 72 MKKPR, ultimately resulting in the fully active form with the neo-N terminus of 85 YNVFP. Mole-per-mole, activation was more potent than by MMP-3, a classical collagenase activator. Elastase was detectable in human OA synovial fluid and OA synovia which displayed histologically graded evidence of synovitis. Bioinformatic analyses demonstrated that, compared with other tissues, control cartilage exhibited remarkably high transcript levels of the major elastase inhibitor, (AAT) alpha-1 antitrypsin (gene name SERPINA1), but these were reduced in OA. AAT was located predominantly in superficial cartilage zones, and staining enhanced in regions of cartilage damage. Finally, active MMP-13 specifically inactivated AAT by removal of the serine proteinase cleavage/inhibition site. Taken together, this study identifies elastase as a novel activator of pro-MMP-13 that has relevance for cartilage collagen destruction in OA patients with synovitis.

Introduction
Osteoarthritis is a highly prevalent, degenerative joint disease which causes significant pain and disability. The progressive and irreversible destruction of the articular cartilage is the central characteristic of the disease. Cartilage is composed of two major components, the highly charged proteoglycan aggregate aggrecan and type II collagen, which is the major structural component of cartilage exhibiting a tightly interlocking triple helix, making it particularly resistant to proteolysis. Due to slow rates of synthesis, the loss of collagen is an essentially irreversible process [1], making it a critical therapeutic intervention point to prevent further cartilage breakdown. The crucial effector proteinases in this destruction are matrix metalloproteinases (MMPs)specifically a subgroup known as the classical collagenases (MMP-1, MMP-8 and MMP-13). A wealth of evidence implicates MMP-13 (collagenase-3) with OA development and progression [2]. Indeed, this collagenase has the highest efficiency for cleavage of type II collagen [3], and MMP-13deficient mice are protected from cartilage destruction in experimental OA [4]. However, MMP inhibition has failed clinically, with patients presenting with a range of off-target effects including arthralgiareferred to as 'musculoskeletal syndrome' [5]. This failure is attributed to a lack of inhibitor selectivity, which is a major hurdle, due to a common active-site architecture amongst different MMPs, many of which are considered 'antitargets'proteins to be avoided for therapeutic inhibition [6]. Although new generations of MMP-13 inhibitors exhibit markedly improved selectivity [7], none have been clinically successful thus far.
An essential control point in the degradation of cartilage collagen is the activation of pro-MMPs [8][9][10][11] and may represent a novel method to reduce or block cartilage breakdown. Although some MMPs, such as MMP-3, are capable of collagenase activation [3], serine proteinases have long been postulated as putative in vivo activators [12,13] and are more amenable for drug development [14,15]. These enzymes are often assumed to be cartilage-derived, but the synovium can also be a source of proteinases [16]. Despite historically being referred to as a 'noninflammatory' arthritis, recent evidence suggests a significant inflammatory component to OA progression, at least within certain subgroups of patients [17]. Several studies have recently reported that activated macrophages and neutrophils within the synovium contribute to the severity of knee OA and that the serine proteinase neutrophil elastase may act as a functional biomarker for OA progression in this joint [18][19][20][21][22].
Neutrophil elastase is located primarily in azurophilic granules of polymorphonuclear leukocytes and released upon cell degranulation [23]. The proteinase cleaves preferentially after small hydrophobic residues and has a range of substrates and well-established roles such as pathogen digestion and cellular extravasation [24,25]. The major inhibitor of elastase is alpha-1 antitrypsin (AAT), a member of the serine proteinase inhibitor (serpin) superfamily, encoded by the SER-PINA1 gene. This is perhaps most evident in patients with AAT deficiency, who often present with earlyonset emphysema, thought to be driven by uncontrolled neutrophil elastase activity in the lung [26]. This inflammatory proteinase has been proposed as a drug target in numerous disorders, and a commercially available inhibitor, Sivelestat, has been approved for acute respiratory distress syndrome (ARDS), but only in specific jurisdictions [27,28].
The potent, direct cleavage of cartilage proteoglycan by neutrophil elastase has been demonstrated previously, both in vitro and in vivo [29,30]. Indeed, a role for neutrophil elastase in cartilage destruction in rheumatoid arthritis (RA) has long been postulated [31][32][33]. More recently, it has been demonstrated that neutrophil elastase inhibition reduces chronic pain in mice that have undergone monoiodoacetate injection to induce OA [34], and reduces structural changes in rats following induction of post-traumatic knee OA [35]. Considering these observations, we investigated the effect of neutrophil elastase in OA cartilage collagenolysis, and in particular explored the biochemical interplay with the major collagenase in OA, MMP-13.

Results
Neutrophil elastase induces collagen release from human OA cartilage in explant culture in an MMP-dependent manner To investigate the effect of neutrophil elastase on human OA knee cartilage, explant culture experiments were performed. Neutrophil elastase potently and reproducibly induced the destruction of cartilage proteoglycan and collagen. To investigate whether this destruction was direct or indirect, a broad-spectrum MMP inhibitor, GM6001 was utilised. Elastaseinduced collagen release was markedly reduced in the presence of this inhibitor (Fig. 1A) which correlated with reduced MMP activity in the conditioned medium (Fig. 1B), whereas proteoglycan release was unaffected (Fig. 1C). GM6001 does not affect the activity of neutrophil elastase directly (Fig. 1D), and its vehicle control (DMSO) had no effect on cartilage matrix destruction. When the serpin neutrophil elastase inhibitor AAT was also included, ECM degradation was abrogated, demonstrating the requirement for elastolytic activity to induce cartilage degradation, both directly (proteoglycan release) and indirectly (collagen release).
Neutrophil elastase activates pro-MMP-13 As collagen release was MMP-dependent, we hypothesised that neutrophil elastase could activate pro-MMPs that promote collagenolysis. To determine whether elastase can directly activate procollagenases, in vitro incubation experiments were performed using recombinant enzymes followed by an MMP-specific fluorogenic peptide substrate. In this system, neutrophil elastase induced very modest activation of pro-MMP-1 ( Fig. 2A) and no activation of pro-MMP-8 (Fig. 2B). However, when incubated with pro-MMP-13 it induced robust activation of the collagenase, to a similar level as the chemical activator 4-aminophenylmercuric acetate (APMA; Fig. 2C). Remarkably, neutrophil elastase induced approximately threefold higher MMP-13 activity than an equimolar amount of active MMP-3, while matriptase and hepsin failed to significantly activate pro-MMP-13 (Fig. 2D), in line with previous observations [36,37].
Neutrophil elastase fully activates MMP-13, generating the 85 YNVFP neo-N terminus To demonstrate that pro-MMP-13 was processed into its active form, an activation time course was performed. Neutrophil elastase rapidly converted pro-MMP-13 into a 45 kDa species which corresponds to the fully active form as demonstrated by treatment with APMA (Fig. 3A,B). APMA also generated an additional band of a lower molecular weight, likely due to continued autoproteolysis of MMP-13. A doseresponse experiment was conducted to assess the kinetics of processing in more detail. We observed optimum activation at a 1:5 enzyme to substrate ratio, but processing and increased MMP activity were observed at as low as 1:100. A 1:1 ratio resulted in slightly lower MMP activity, consistent with a reduced level of active MMP-13, suggesting a degree of nonspecific proteolysis at this higher concentration (Fig. 3C,D). The inclusion of the metalloproteinase inhibitor GM6001 led to the formation of an intermediate species (Fig. 3E). Nterminal sequencing was performed to identify cleavage sites. Neutrophil elastase alone generated the 85 YNVFP neo-N terminus, while the predominant form in the presence of GM6001 was 72 MKKPR (Fig. 3F). To demonstrate the activation of pro-MMP-13 following neutrophil degranulation, blood-purified neutrophils were primed with GM-CSF then subsequently stimulated with a degranulation cocktail of A B C D Fig. 1. Neutrophil elastase-induced collagen release is MMP-dependent. Neutrophil elastase (NE) was added to human OA cartilage for a period of 7 days in the presence or absence of the MMP inhibitor GM6001 (10 lM), DMSO vehicle control (0.02% v/v) or AAT protein (100 lgÁmL À1 ), after which the medium was removed and remaining cartilage digested with papain as described previously [51]. (A) Collagen breakdown was determined by hydroxyproline (OHP) assay. (B) MMP activity was determined in conditioned medium using an MMP-specific quenched-fluorescent peptide. (C) Proteoglycan release was determined by measurement of sulphated GAG using a DMMB assay. (D) GM6001 or its vehicle DMSO had no effect on neutrophil elastase activity, as determined in vitro using the elastase-specific quenched-fluorescent peptide, MeOSuc-AAPV-AMC. Matrix breakdown is expressed as a percentage of the total, which is determined by level of GAG or OHP in the conditioned medium and the remaining papain-digested cartilage. Cartilage matrix degradation by elastase is representative of at least three independent experiments, while inclusion of the metalloproteinase inhibitor, GM6001, was conducted twice. Data are presented as mean AE SD (n = 4) with statistical analyses performed using a one-way ANOVA with Tukey's post hoc test where *P < 0.05, **P < 0.01 and ***P < 0.001; ns, not significant. Neutrophil elastase is detectable in inflamed OA synovium and OA synovial fluid Neutrophil elastase is poorly expressed in cartilage [38,39], so to determine whether neutrophil elastase could be detected in the OA joint, an immunohistochemical examination of human OA synovial tissues was performed. In total, 17 human OA synovia were scored histologically using a previously described synovitis scoring system [37,40]. Neutrophil elastase was virtually undetectable in tissues with no synovitis but was detectable in invading leukocytes and in blood vessels of samples categorised as having 'slight' or 'moderate' synovitis. MMP-13, however, was detectable at all levels of synovitis, although staining was observed predominantly in fibroblasts in no/low synovitis samples and often in invading leukocytes in higher grades (Fig. 5A). To determine the presence of neutrophil elastase and MMP-13 in OA synovial fluid, ELISAs were performed on fluid which was taken at the time of joint arthroplasty. Neutrophil elastase was detectable in all OA synovial fluids tested although this varied significantly between patients (mean AE standard deviation; SD): 6374 (AE 5745) pgÁmL À1 , n = 8; Fig. 5B). MMP-13 was detectable in all except three OA synovial fluid ΔF·s -1  To determine relative serpin expression in human cartilage, we interrogated our previously published RNAseq data (GSE111358; [39]). In non-OA control human hip cartilage, several serpins were highly abundant, with 10 serpins having transcripts per million (TPM) counts of 10 or more. Of these, the SERPINA1 gene, encoding AAT, the major neutrophil elastase inhibitor, was particularly abundant with a TPM count exceeding 7000. Indeed, SERPINA1 was the 9 th most abundant transcript within this dataset (Fig. 6A). Using the human protein atlas, cross-tissue expression levels were determined. Interestingly, SERPINA1 expression was markedly enriched in cartilage, second only to expression within the liver (Fig. 6B). When comparing macroscopically intact hip OA cartilage to control non-OA (NOF fracture) hip cartilage, levels of SERPINA1 transcript were markedly downregulated (P adj = 0.004; Fig. 6C).
To determine specific localisation of the AAT protein within human cartilage, immunohistochemistry revealed that the staining was particularly strong in the upper regions of the cartilage matrix, consistent with expression predominantly in the superficial cartilage layer. In undamaged cartilage, the staining was principally localised to the chondrocyte, but interestingly, staining intensity was markedly increased in areas of damaged ECM, with intense AAT staining in the most damaged areas of OA cartilage (Fig. 6D).

MMP-13 inactivates AAT at two sites, removing the reactive-centre loop
Matrix metalloproteinases have previously been demonstrated to inactivate AAT by removing the reactive-centre loop required for irreversible inhibition of target proteinases [41]. Here, we demonstrate that in vitro incubation of AAT with active MMP-13 results in the generation of a single protein bandeven after a 16-h incubationwith a lower molecular size than that of native AAT (Fig. 7A), and AAT is no longer able to inhibit neutrophil elastase (Fig. 7B). A time course experiment was performed, and products separated using tris-glycine (protein) or tris-tricine (peptide) SDS/PAGE (Fig. 7C). N-terminal sequencing revealed that MMP-13 cleaved AAT at two sites, 377 LEAIP and 382 MSIPP (Fig. 7D), upstream of the canonical neutrophil elastase cleavage site. Thus, active MMP-13 induces a specific inactivation of AAT, removing the inhibitory bait region of the serpin.

Discussion
In this study, we demonstrate that neutrophil elastase is a potent activator of the major OA collagenase, MMP-13. Historically described as noninflammatory arthritis, it is becoming clear that OA is a disease of the whole joint and that inflammation contributes to the disease processes, at least in a significant subgroup of patients [16]. The importance of the synovium in OA is becoming increasingly apparent, making factors often thought of having a role only in inflammatory arthritis progressionincluding proteinasesof renewed interest for study in OA.
Serine proteinases have been shown to be important in the proteolytic cascades leading to cartilage destruction previously [8,9,12,36,37,42], although identifying the specific proteinases involved is challenging. Indeed, the conversion of pro-MMP-13 into its active form is an essential, yet often overlooked, control point for the breakdown of cartilage collagen in OA. To our knowledge, the only described direct serine proteinase activators of MMP-13 are trypsin [3] and plasmin [43], with the former lacking physiological relevance to the joint, while the latter has a complex and debated role in arthritis [44]. This is the first study to describe the direct and full activation of pro-MMP-13 by neutrophil elastase. This serine proteinase has often been linked to the pathophysiology of RA and has been previously shown to be present in patient synovial fluid [45][46][47]. Indeed, systemic administration of human plasma AAT, reduced arthritis scores and joint destruction in murine collagen-induced arthritis [48], and mice deficient for neutrophil elastase and the related proteinase cathepsin-G, are also protected in this model [32]. As a relatively broad-spectrum serine proteinase, the role of neutrophil elastase has been considered one of matrix destruction, and it has been shown to directly degrade proteoglycans [30,49]. The data in our study are consistent with this, as neutrophil elastase-induced proteoglycan release in human OA cartilage was insensitive to metalloproteinase inhibition, and therefore direct. Elastase has previously been shown to directly cleave type III collagen in vitro, but was unable to cleave type I collagen [50]. The hydroxyproline (OHP) assay is a sensitive and specific way to measure collagen breakdown from cartilage in explant culture [37,42,51]. Cartilage is composed overwhelmingly of type II collagen (> 90%; [52]), but does include other minor collagens such as type IX and type XI, which elastase has been shown to degrade directly, at least in vitro [53]. In our study, we observed up to 15% collagen release from human OA cartilage, which was mediated by metalloproteinase activity, not the direct activity of elastase. The unique structure of triple-helical collagen makes it susceptible to significant proteolysis from only a very restricted number of collagenases, likely due to important  interactions of exosites required for unfolding of the triple-helix [54,55]. Although proteoglycan breakdown by neutrophil elastase is well established, the finding that this proteinase can induce MMP-dependent collagen degradation is important, as this is an irreversible step in the process of joint destruction [1]. Indeed, explant data in this study indicate that a pool of latent collagenases exists within diseased OA cartilage, something which has been observed in other studies through the addition of MMP activators to the diseased tissue [37,56]. Correlation between elastase activity and collagen neo-epitopes generated by a collagenase has been observed in the synovial fluids of patients with noninfectious knee joint synovitis previously [57]. Although neutrophil elastase has been shown to be an activator of other MMPs [13,58], here we demonstrate that it is a direct activator of the collagenase most commonly associated with OA, MMP-13. [12]. Under the described experimental conditions, neutrophil elastase was unable to significantly activate pro-MMP-8 while providing only modest activation of pro-MMP-1, further supporting a role for MMP-13 in neutrophil elastase-induced collagen destruction. Degranulation experiments further demonstrated a rapid activation of pro-MMP-13, which could be blocked in the presence of AAT. Neutrophil degranulation products have previously been shown to induce cartilage breakdown in porcine cartilage explants, with neutrophil elastase implicated in the process [59]. Neutrophil elastase was detectable in all synovial fluid samples tested. While these levels are likely to be significantly lower than concentrations observed in RA [45], it is important to recognise that OA is disease of slow progression. Furthermore, the timing of sample collection will likely be of critical importance. For example, levels of neutrophil elastase have been shown to be high immediately after sports injury, but then gradually reduce over a 90-day period [60]. Periodic, localised inflammation could contribute to temporary release of neutrophil elastase into the OA joint, which could contribute to collagen degradation through MMP-13 activation. The serpins are a superfamily of structurally similar proteins, the majority of which inhibit serine proteinases, and we have recently described their emerging role in cartilage [61]. In this study, we have identified remarkably high transcript expression levels of SERPINA1 gene in human cartilage. Indeed, others have demonstrated high levels of the AAT protein within synovial fluid previously [62,63]. This serpin is a serum protein with levels in the blood of approximately 0.9-2 mgÁmL À1 [64], the vast majority of which is produced in the liver [65]. Here, we show that levels of SERPINA1 transcript in human cartilage are second only to those in the liver. An interesting study which used microdissection and iTRAQ proteomics to investigate the protein composition of human cartilage layers found AAT to be mainly located in the upper levels of cartilage [66]. Indeed, this is something we also observed in our current study. It is important to recognise that AAT has multiple proteinase targets, although neutrophil elastase is by far the most kinetically favourable (K a = 6.5 9 10 7 M À1 s À1 ; [24,67]). It is plausible that AAT may act as a 'first line of defence' against damaging immune cell-derived serine proteinases following degranulation. We also observed that AAT localises to damaged cartilage regions which requires further investigation. Possible reasons could include a localised damage response or that greater levels of AAT from the synovial fluid are able to penetrate the cartilage following significant degradation of the proteoglycan matrix (as evidenced by loss of safranin-O staining in these areas). 2) was embedded and sectioned. Sections of different depths were stained with haematoxylin and eosin and the level of synovitis scored according to a previously described scoring system [40] where samples were categorised as having no synovitis (0-1), slight synovitis (2-3), moderate synovitis (4-6) or strong synovitis (7)(8)(9). Immunohistochemistry was performed using antibodies to neutrophil elastase and MMP-13. Sections where no primary antibody was included, as well as an isotype control antibody at the same concentration as primary antibody, served as negative controls. NE was absent from samples categorised as having 'no synovitis' (synovitis score < 1; 0/3), but present in the majority of samples with a synovitis score > 1 (11/14). MMP-13 was detectable in all synovial samples regardless of synovitis grade. Synovial fibroblasts are indicated with red arrows, invading leukocytes with green arrows and blood vessels with orange arrows. NE mainly stained positive in leukocytes and blood vessels. MMP-13 was detectable in fibroblasts and invading leukocytes. Images taken at 40x magnification and scale bars represent 50 µm.   High SERPINA1 expression in cartilage raises important questions about the physiological relevance of neutrophil elastase deposition from an inflamed synovium. Indeed, using the total protein ELISAs undertaken in this study, it is not possible to determine the level of active neutrophil elastase or the level of functional AAT protein. However, it has been previously demonstrated that cartilage-bound neutrophil elastase is protected from the activity of proteinase inhibitors [68]. Moreover, serpins can be inactivated by other proteinases, as well as by oxidation, both of which have been demonstrated in the RA joint previously [69][70][71]. These mechanisms suggest total serpin levels may not be representative of the effective inhibitor concentrations within the joint. In this study, we describe the inactivation of AAT by MMP-13. While this finding is novel, it is likely that in vivo, several MMPs may contribute to the inactivation of this inhibitor. Indeed, this serpin has been shown to be inactivated by MMP-1, MMP-2 and MMP-3 previously and the inactivated form has been observed in the synovial fluid of arthritis patients [70,71]. Here, we have identified cleavage sites generated by MMP-13 which correspond to regions N-terminal of the canonical cleavage site of neutrophil elastase (M 382 is the neo-N terminus generated following this cleavage/inhibition). Both sites have been identified as cleavage sites for other MMPs [71]. Future studies could involve the generation of neo-epitope antibodies which specifically recognise MMP-cleaved AAT. Such antibodies could be used to determine the level of AAT inactivation within the synovial joint of OA patients and the likely proteolytic source. Alternatively, the relative levels of different AAT (cleaved/uncleaved or complexed) could be determined by mass spectrometry approaches. The disarming of cognate inhibitor(s) will make joints more susceptible to the damaging effects of serine proteinases which present during periods of localised inflammation. A working hypothesis regarding the interplay between MMP-mediated AAT inactivation and pro-MMP-13 activation by neutrophil elastase is outlined in Fig. 7E. A corollary of this hypothesis is the crucial role of functional AAT in mediating joint protection. To our knowledge, studies investigating the susceptibility of cartilage for degradation in OA (or indeed other arthritic diseases) in patients with AAT deficiency have not been thoroughly investigated previously. This too should be the focus of future endeavours. The high level of SERPINA1 expression within cartilage suggests mutations within the SERPINA1 gene (rendering a weaker/nonfunctional AAT) may have significant consequences for the integrity of the tissue during periods of inflammation.
There are limitations to this study. Although our observations regarding cleavages (and cleavage sites) are novel and important, we must be candid about extrapolating biochemical observations to the OA joint. To determine pathophysiological importance of neutrophil elastase on MMP-13 activation in OA, future studies would need to: (a) determine the level of active neutrophil elastase in the OA joint; (b) assess how this compares to the level of pro-MMP-13 in the cartilage; (c) establish the degree of AAT inactivation and therefore the viable inhibitor pool in the OA joint; and (d) how these change with stage and severity of OA (and the presence or absence of synovitis). Larger cohorts of human clinical samples and the use of different preclinical models of OA will prove important in this regard. Recently, Kaneva et al [72] demonstrated that AAT protected cartilage in two animal models of inflammatory arthritis, including one induced by neutrophil elastase. Our study both supports and compliments these observations and provides novel insights into the proteolytic interplay between AAT, neutrophil elastase and MMP-13, with potential consequences for the destruction of OA cartilage. AAT can have a role beyond proteinase inhibition [61] and the authors conclude that AAT was not only chondroprotective but also anti-inflammatory and has a significant effect on pain [72]. This study and our own highlight the need to also investigate the effects of AAT administration using in vivo models of OA, with direct quantification of cartilage degradation, pain and inflammation.
The importance of synovial inflammation in OA progression is contested, and the central role of dysregulated chondrocyte biology in OA cartilage destruction is clear. However, we and others have demonstrated a variable degree of synovitis in OA patients (reviewed in [73]), and in these instances, it is credible that neutrophil elastasegiven its activation potency and ability to generate a fully active MMP-13will contribute to OA cartilage collagen destruction through direct activation of latent MMP-13. The search for in vivo activators of MMP-13 is important, particularly as this represents a key control point for cartilage destruction and may represent a novel approach for therapy. The rapid, full activation of MMP-13 by neutrophil elastase may have significance in a subset of OA patients where an inflamed synovium contributes to the progression and exacerbation of the disease process.

Reagents
Unless stated otherwise all reagents were of the highest purity and purchased from Sigma-Aldrich (Gillingham, UK). Human neutrophil elastase (EC3.4.21.37) purified from human sputum was purchased from Elastin Products (SE663; Owensville, MO, USA). AAT protein was purified as described previously [74]. Human recombinant pro-MMP-1 (E.C.3.4.24.7) and pro-MMP-13 (E.C.3.4.24.B4) were expressed using an insect cell expression system (and purified) as previously described [11,37]. ). The following day, explants were stimulated with neutrophil elastase for the time period and concentrations indicated. At the end of the experiment, medium was harvested and remaining cartilage digested with papain as previously described [51]. For both medium and digested cartilage, OHP assays were performed to assess percentage collagen destruction, while proteoglycan destruction was assessed by measuring glycosaminoglycan (GAG) levels using dimethyl methylene blue (DMMB) assay (see [51] and references therein). were subsequently incubated with neutrophil elastase to determine inhibitory capacity (see below for assay details). For protein gels, digestion products (10 µL for AAT inactivation and 20 µL for pro-MMP-13 activation experiments) were mixed with 59 sample buffer (625 mM Tris/HCl pH 6.8, 10% (w/v) SDS, 40% (v/v) glycerol, 0.5% (w/v) bromophenol blue, 3% (v/v) ß-mercaptoethanol) and separated by 10% SDS/PAGE as described previously [37]. Gels were subsequently stained using a PlusONE silver staining kit according to the manufacturer's instructions (GE healthcare). For low molecular mass products of AAT digests, peptides were detected by separating reaction products by Tris-Tricine SDS/PAGE as described previously [76]. For N-terminal sequencing (both activation and inactivation experiments), gels were blotted onto PVDF and stained with Coomassie Brilliant Blue for 2 min. Bands were excised with a clean scalpel blade and sent for N-terminal sequencing (Alphalyse, Odense, Denmark).

Enzyme activity assays
To test MMP activation, pro-MMP-1, pro-MMP-8 and pro-MMP-13 were incubated with elastase as described above, for 4 h. As a negative control, an excess of AAT (1.14 mgÁmL À1 ) was included, which demonstrated the requirement for elastase activity. Incubation products (5 µL) were added to white-walled plates and diluted to 80 µL in MMP assay buffer (reaction buffer with 0.1% (v/ v) polyethylene glycol 6000). MMP fluorogenic substrate (FS-6) was added to start the reaction at a final concentration of 10 µM, to a total well volume of 100 µL. Reaction velocity was calculated by linear regression of initial rates of progress curves. For mole-per-mole comparison experiments, neutrophil elastase was titrated with freshly resuspended AAT protein (Merck Millipore; 178251); MMP-3 was titrated with full-length TIMP-1, while matriptase and hepsin were titrated with 4-methylumbelliferyl 4guanidinobenzoate (MUGB). Pro-MMP-13 (100 nM) was incubated with 20 nM of each enzyme for 4 h in the reaction buffer described above prior to conducting the MMP

Neutrophil isolation, degranulation and MMP-13 activation
Neutrophils (purity typically > 97%, assessed by cytospin) were isolated from heparinised whole blood using HetaSep and Ficoll-Paque as previously described [77] and resuspended at 5 9 10 6 cellsÁmL À1 in 1 mL of RPMI 1640 media supplemented with HEPES (25 mM) and L-glutamine (2 mM). Neutrophils were incubated with GM-CSF (5 ngÁmL À1 ) for 30 min followed by cytochalasin B (6 lgÁmL À1 ) and f-Met-Leu-Phe (fMLP, 1 lM) for a further 30 min. Following incubation, cells were pelleted at 1000 g for 3 min and supernatant decanted into clean tubes. Degranulation products (supernatant; 5 lL) from either treatment were incubated with pro-MMP-13 (100 nM) for 2 h at 37°C. Controls of the degranulation stimulants (cytochalasin/GM-CSF/fMLP mix in the absence of neutrophils) were also incubated with pro-MMP-13. AAT (50 nM) was included as a negative control. After the incubation, 5 lL was added to a white-walled 96-well plate and an MMP and elastase activity assays were undertaken as described above. Data are presented as mean AE SD using three independent donors of neutrophils. For SDS/PAGE, 10 lL of this incubation was mixed with 59 sample buffer (as above), boiled and kept at À20°C until the products were separated by SDS/PAGE and western blotting performed.

Western Blotting
For western blotting, samples were separated by SDS/ PAGE using 10% (w/v) precast acrylamide gels and transferred onto polyvinylidene fluoride (PVDF) using the

Histology and Immunohistochemistry
Human synovial tissues (17 patients, 14 Female/three male, mean age = 65 years AE 9.2) were fixed in 10% formalin and embedded in paraffin, and sections were taken at 4-µm intervals. Haematoxylin and eosin (H&E) staining was performed as previously described [11] and synovitis scored according to a previously described system [37,40]. Immunohistochemistry was performed essentially as previously described [37], except antigen retrieval was performed using a mini autoclave (Aptum 2100 retriever; Aptum, Southampton, UK) for 30 min. Primary antibodies for neutrophil elastase and MMP-13 were used at 1 : 100 dilution (both 10 µgÁmL À1 ) for 1 h, and staining conducted using VECTASTAIN Elite ABC HRP Kit according to the manufacturer's instructions (Vector Laboratories; Burlingame, CA, USA). To visualise AAT human OA cartilage, samples from patients undergoing total hip replacement (mean age 61.3 AE 7.5, 3 9 Female) or from hip replacement following fracture (NOF; mean age 74 AE 8.4, 1 9 male, 1 9 female) were fixed, embedded and sectioned as described above. Serial sections were stained with safranin-O or underwent IHC staining using an antibody against AAT protein (16382-1-AP; ProteinTech, Manchester, UK) at 1 : 150 dilution (3.2 µg mL À1 ) for 2 h, followed by signal detection using VECTASTAIN Elite ABC HRP Kit. In all cases, sections treated in the same manner in the absence of a primary antibody, or with an isotype control antibody (R&D systems; AB-105-C) at the same concentration as the corresponding primary antibody, served as negative controls. Images were taken using a Leica Biosystems DM4000B light microscope.

Bioinformatic analyses and protein visualisation
Transcripts per million values in control cartilage (NOF) samples were generated using Salmon as described [39] while Human Protein Atlas TPMs were retrieved from the Human Protein Atlas ([78]; www.proteinatlas.org). R software was used to generate figures from RNA sequencing data in Fig. 6A

Statistics
Statistical analyses where multiple comparisons are required were performed using a one-way ANOVA with a Tukey's post hoc test. In all cases, *P < 0.05, **P < 0.01, ***P < 0.001 and data are presented as mean AE SD. RNAseq statistical analysis was performed using DESeq2 where P-values were adjusted using the Benjamini-Hochberg method which controls for the false discovery rate (< 0.05). Topography of a 2.0 A structure of alpha1-antitrypsin reveals targets for rational drug design to prevent conformational disease. Protein Sci 9, 1274-1281.