Engineered Tissue Inhibitor of Metalloproteinases-3 Variants Resistant to Endocytosis Have Prolonged Chondroprotective Activity*

Tissue inhibitor of metalloproteinases-3 (TIMP-3) is a central inhibitor of matrix-degrading and sheddase families of metalloproteinases. Extracellular levels of the inhibitor are regulated by the balance between its retention on the extracellular matrix and its endocytic clearance by the scavenger receptor low density lipoprotein receptor-related protein 1 (LRP1). Here, we used molecular modeling to predict TIMP-3 residues potentially involved in binding to LRP1 based on the proposed LRP1 binding motif of 2 lysine residues separated by about 21 Å and mutated the candidate lysine residues to alanine individually and in pairs. Of the 22 mutants generated, 13 displayed a reduced rate of uptake by HTB94 chondrosarcoma cells. The two mutants (TIMP-3 K26A/K45A and K42A/K110A) with lowest rates of uptake were further evaluated and found to display reduced binding to LRP1 and unaltered inhibitory activity against prototypic metalloproteinases. TIMP-3 K26A/K45A retained higher affinity for sulfated glycosaminoglycans than K42A/K110A and exhibited increased affinity for ADAMTS-5 in the presence of heparin. Both mutants inhibited metalloproteinase-mediated degradation of cartilage at lower concentrations and for longer than wild-type TIMP-3, indicating that their increased half-lives improved their ability to protect cartilage. These mutants may be useful in treating connective tissue diseases associated with increased metalloproteinase activity.

Tissue inhibitor of metalloproteinases-3 (TIMP-3) 2 is an endogenous inhibitor of metalloproteinase catalytic activity and hence an important regulator of connective tissue turnover by these enzymes (1)(2)(3). TIMP-3 is unique among the four mammalian TIMPs in that it can inhibit not only matrix metalloproteinases (MMPs), but also the related adamalysins (a disintegrin and metalloproteinase (ADAMs)) and adamalysins with thrombospondin motifs (ADAMTSs).
Various pathophysiological states are associated with disrupted extracellular matrix (ECM) turnover. For example, osteoarthritis is characterized by degradation of type II collagen and aggrecan by collagenolytic MMPs and aggrecan-degrading ADAMTSs, respectively. Timp3-null mice show increased collagenase and aggrecanase activity in cartilage and increased spontaneous osteoarthritis with age (4). TIMP-3 levels are reduced in human osteoarthritic cartilage (5). Treatment with recombinant TIMP-3 has been shown to block cartilage degradation in vitro (6) and in in vivo models of osteoarthritis (7), further illustrating the chondroprotective activity of TIMP-3.
The half-life of TIMP-3 in tissue is positively regulated by its binding to the ECM (8,9) and negatively regulated by its endocytosis and subsequent lysosomal degradation via low density lipoprotein receptor-related protein 1 (LRP1) (10,11). We postulated that a mutant of TIMP-3 with reduced affinity for LRP1 would have a longer half-life in tissue and an increased ability to block metalloproteinase activities compared with wild-type TIMP-3.
To engineer LRP1-resistant mutants of TIMP-3, we analyzed a model of the three-dimensional structure of full-length TIMP-3 and identified pairs of lysine residues potentially separated by 21 Å. These lysine residues were mutated to alanine singly and in pairs, and the endocytosis resistance, LRP1 binding, and chondroprotective ability of the mutants were evaluated. Two of the engineered mutants, TIMP-3 K26A/K45A and K42A/K110A, exhibited substantial endocytosis resistance and protected cartilage better than wild-type TIMP-3. We thus show that targeting the TIMP-3 endocytosis pathway is a potential strategy for inhibiting excess metalloproteinase activity in pathological settings.

Results
Design of TIMP-3 Mutants-Because no crystal structure of full-length TIMP-3 is available, we constructed a homology model of the TIMP-3 structure using the available TIMP-2 (Protein Data Bank code 1BR9) (22) structure. We then compared the N-terminal domain of TIMP-3 in our model with the available crystal structure of the N-terminal domain of TIMP-3 in complex with ADAM17 (Protein Data Bank code 3CKI) (23) and observed good agreement between the two structures. The most C-terminal lysine residue (Lys-180) is unresolved in the model. The remaining 16 lysine residues of TIMP-3 are predicted to be located on the surface of the protein. We measured the distance between ␣-carbon residues of pairs of lysine residues and identified 10 pairs of lysine residues predicted to be separated by 21 Ϯ 5 Å. (Fig. 1). With the exception of Lys-157, all lysine residues identified were located on the N-terminal inhibitory domain of TIMP-3. Using site-directed mutagenesis, we generated 10 mutants of TIMP-3 in which both lysine residues of the potential LRP1-binding dilysine motif were mutated to alanine as well as 12 mutants in which the individual lysine residues identified were singly mutated to alanine (Table 1). A FLAG tag was included at the C terminus of all mutants for detection and purification, as described previously for wildtype TIMP-3 (24).
Screening of Mutants-We carried out an initial screen of the mutants using immunoblotting to evaluate their rate of disappearance from the medium of HTB94 chondrosarcoma cells. These cells secrete a minimal amount of TIMP-3-binding ECM, and disappearance of TIMP-3 from the medium reflects its endocytosis into cells via LRP1 (11).
HEK-293-EBNA cells were transiently transfected with the mutants and conditioned medium dialyzed against TBS. We confirmed that wild-type TIMP-3 in such dialyzed crude conditioned medium was endocytosed by HTB94 with indistin-FIGURE 1. Identification of potential LRP1-binding residues in TIMP-3. A model of TIMP-3 was generated using the available crystal structure of TIMP-2. The position of Lys-180 was unresolved in the model, but the remaining 16 lysine residues of TIMP-3 were all predicted to be located on the surface of the molecule. Lysine residues located on the N-terminal domain of TIMP-3 are indicated in purple, and those on the C-terminal domain are shown in blue. The distance between ␣-carbon residues of pairs of lysine residues was measured, and 10 pairs of lysine residues (listed on the right) were predicted to be separated by 21 Ϯ 5 Å. This figure was made with PyMOL.

TABLE 1
Candidate LRP1-resistant mutants of TIMP-3 generated guishable kinetics from purified recombinant TIMP-3, validating this method as a screening tool. Clearance was inhibited by RAP (Fig. 2, A and B), confirming our previous findings that TIMP-3 is endocytosed in an LRP1-dependent manner (10,11).
Wild-type TIMP-3 was lost from the medium with a half-life of 3.6 Ϯ 0.3 h ( Fig. 2A). TIMP-3 K42A, K71A, and K157A exhib-ited significantly increased half-lives of 8.8 Ϯ 1.6, 12.9 Ϯ 0.4, and 7.8 Ϯ 2.5 h, respectively (Fig. 2C). The remainder of the single mutants showed no statistically significant change in half-life, and Ͼ75% of all of the single mutants were lost from the medium within 24 h.
We then analyzed the half-lives of TIMP-3 mutants in which both lysine residues of the potential LRP1-binding dilysine motif were mutated to alanine. All of the double mutants remained in the medium for markedly longer than the single lysine mutants (Fig. 3). More than 50% of each remained in the medium after 24 h, so half-lives could not be calculated, and we compared the mutants based on the percentage remaining in the medium at 24 h. TIMP-3 K42A/K110A and K26A/K45A stayed in the medium at the highest concentration, with 92 Ϯ 3% and 69 Ϯ 6% of these mutants remaining in the medium after 24 h, respectively. These mutants were purified for further characterization.
Confocal Microscopy Imaging of Endocytosis-We added purified wild-type TIMP-3 to HTB94 cells and observed co-localization of the protein with early endosomal antigen 1 (EEA1) in endosomal vesicles (Fig. 4). Preincubation of cells with RAP ablated this co-localization, confirming that TIMP-3 endocytosis is LRP1-dependent. K26A/K45A and K42A/K110A were not visible within cells and did not co-localize with EEA1, confirming that they are not endocytosed.
Uptake by LRP1-null Cells-We compared uptake of wildtype and mutant TIMP-3 by LRP-1 wild-type and null cells to evaluate whether the altered uptake of the mutants was dependent on LRP1. In LRP1 wild-type PEA10 cells, TIMP-3 was rapidly taken up from the medium with a half-life of 1.9 h, whereas K26A/K45A and K42A/K110A remained in the medium (Fig. 5A). Conversely, all three proteins remained at the same level in the medium of LRP1-null PEA13 cells (Fig. 5B).
Binding to LRP1-Binding of the mutants to full-length LRP1 was evaluated by ELISA. Wild-type TIMP-3 bound efficiently to LRP1, whereas K26A/K45A bound with markedly reduced affinity, and K42A/K110A displayed minimal binding (Fig. 6A).

TIMP-3 Mutants Resistant to LRP1 Have Increased Half-life
Characterization of Mutant Affinity for Prototypic Metalloproteinases-We compared the inhibition kinetics of the K26A/K45A and K42A/K110A mutants with the wild-type inhibitor. As shown in Table 2 (mean Ϯ S.D., n ϭ 3-5), K i (app) values for wild-type TIMP-3 were in agreement with previous findings (24 -26), and TIMP-3 K26A/K45A and K42A/K110A both displayed unaltered affinities for the catalytic domains of MMP-1, MMP-2, and MMP-3, as well as ADAMTS-4 lacking the C-terminal spacer domain, ADAMTS-5 lacking the C-terminal thrombospondin domain, and the ectodomain of ADAM17. LRP1 shedding, which is primarily caused by metalloproteinases, was also unaltered by the mutants (not shown).
Affinity for ADAMTS-5 in the Presence of Heparin-We have previously shown that sulfated glycosaminoglycans, such as heparin and heparan sulfate, increase affinity between wildtype TIMP-3 and glycosaminoglyan-binding enzymes, such as ADAMTS-5, by mediating formation of high affinity trimolecular complexes (27). As shown in Fig. 6D (representative experiment) and Table 2 (mean Ϯ S.D., n ϭ 3-5), heparin increased TIMP-3 affinity for ADAMTS-5 from a K i (app) value of 1.27 Ϯ 0.46 nM to a value too low to be calculated. Heparin similarly increased K26A/K45A affinity for ADAMTS-5 from a K i (app) value of 0.95 Ϯ 0.55 nM to a value too low to be calculated. In contrast, affinity of K42A/K110A for ADAMTS-5 was not affected by heparin, with K i (app) values of 1.12 Ϯ 0.94 and 1.20 Ϯ 0.42 nM in the absence and presence of heparin, respectively.
Inhibition of Cartilage Degradation-The ability of the mutants to inhibit IL-1-stimulated aggrecan degradation in porcine cartilage explants was evaluated. In a 48-h assay, glycosaminoglycan release into the conditioned medium was significantly inhibited by wild-type TIMP-3 at 100 nM but not at 10 or 50 nM (Fig. 7A). Comparable inhibition was seen using the DMMB dye-binding assay and immunoblotting with the anti-AGEG neoepitope antibody, which detects ADAMTS-mediated cleavage at the TAQE 1771 2A 1772 GEG site in the aggrecan C-terminal chondroitin sulfate-rich region. K26A/K45A and K42A/K110A inhibited aggrecan degradation at lower concentrations, with significant inhibition observed at 10, 50, and 100 nM concentrations of each mutant (Fig. 7, B and C). No cell toxicity was observed for any of the treatment conditions (not shown).

Discussion
The phenotype of Timp3-null mice demonstrates that TIMP-3 is a central inhibitor of metalloproteinase-mediated ECM degradation and that it serves to regulate homeostatic tissue turnover (1-3) and limit pathological tissue remodeling (4). We have previously demonstrated that extracellular levels of TIMP-3 are primarily regulated at the post-translational level, by the balance between its binding to sulfated glycosaminoglycans of the ECM and its cellular uptake by the LRP1 endocytic receptor (28). We sought to engineer a mutant of TIMP-3 with reduced affinity for LRP1 and increased half-life in the tissue, to establish whether tissue remodeling could be modulated by targeting the TIMP-3 endocytic pathway.
Following previous studies that had identified an apparent LRP1-binding motif (13-15), we used a model of the threedimensional structure of TIMP-3 and identified pairs of lysine Control wells were incubated without TIMP-3, and LRP dependence was evaluated by incubating wells with RAP (500 nM) for 1 h before the addition of TIMP-3. After fixing and permeabilizing the cells, TIMP-3 was detected with M2 anti-FLAG antibody (green), and endocytic vesicles were visualized with an antibody against EEA1 (red). Nuclei were visualized by DAPI staining (blue), and the cytoskeleton was visualized by staining with phalloidin (white). Images were gathered by confocal microscopy on a PerkinElmer Life Sciences spinning disc confocal microscope equipped with a Nikon TE-2000 CCD camera (ϫ40 objective lens).
residues whose ␣-carbon atoms are predicted to be separated by 21 Ϯ 5 Å. In agreement with previous findings (13,14), our initial screening procedure indicated that none of the 12 lysine residues assessed is solely responsible for LRP1 binding. Only three of the single mutants (K42A, K71A, and K157A) displayed significantly longer half-lives. Of these, K71A was the most effective, with a 3.6-fold increased half-life. However, all 10 of the double mutants, in which pairs of lysine residues were mutated to alanine, showed substantially longer half-lives, with 50 -90% of the initial concentration remaining in medium after 24 h. The increased half-lives of all double mutants assessed argues against LRP1 binding being governed by a single pair of lysine residues. This is in agreement with the recent findings of van den Biggelaar et al. (29), who found that binding of factor VIII to LRP1 was mediated by multiple lysine residues within an extended basic area on the surface of the ligand, leading them to suggest an additive binding model. The crystal structure of RAP in complex with a fragment of the LDL receptor shows that 2 lysine residues of the ligand interact with adjacent acidic pockets on the receptor (15). The extended binding model suggests that this interaction is supported by the surrounding charge landscape and that the interacting lysine residues do not act in isolation. This model may also apply to other LRP1 ligands. For example, double lysine mutants other than RAP K256A/K270A also display reduced affinity for LRP1 (13,14).
We compared the most endocytosis-resistant of the double mutants, namely K26A/K45A and K45A/K110A, with wildtype TIMP-3 in more detail. Consistent with the lysine residues being distal from the metalloproteinase-binding region of TIMP-3, the mutants showed no change in their ability to inhibit target prototypic metalloproteinases, aggrecanases, and ADAM17. We assessed their biological activity using a cartilage degradation model in which aggrecan degradation is stimulated by treating porcine cartilage explants with the proinflammatory cytokine IL-1. Aggrecan degradation was inhibited by 100 nM wild-type TIMP-3, in agreement with what has been observed previously for the inhibitory N-terminal domain of the inhibitor (6,30). TIMP-3 K26A/K45A and K45A/K110A inhibited aggrecan degradation more effectively, with significant inhibition observed at lower concentrations and for longer than wildtype TIMP-3. This confirms our hypothesis that LRP1-resistant mutants of TIMP-3 would have an improved ability to inhibit metalloproteinase-mediated extracellular matrix degradation.
TIMP-3 has been reported to induce apoptosis in various cell types (31)(32)(33), and we explored the possibility that endocytosisresistant mutants of TIMP-3 may induce higher levels of apoptosis due to their ability to be retained at higher levels in the extracellular environment. We found that cell viability was not affected by treating chondrosarcoma cells or primary chondrocytes with 100 nM wild-type or mutant TIMP-3 for 2 days, in agreement with previous studies using wild-type N-TIMP-3 (6).
A characteristic feature of TIMP-3 is its ability to bind to sulfated glycosaminoglycans of the extracellular matrix (8,9). Cell surface HS proteoglycans are also likely to affect the trafficking of TIMP-3 in the extracellular environment. Lee et al. (34) demonstrated that this binding was mediated by an extended area of basic residues on the opposite side of TIMP-3  OCTOBER 14, 2016 • VOLUME 291 • NUMBER 42

TIMP-3 Mutants Resistant to LRP1 Have Increased Half-life
to its inhibitory ridge. The lysine residues in K26A/K45A and K42A/K110A are situated within or adjacent to this basic area (Fig. 9). The mutants retain detectable, albeit reduced, binding to heparin and heparan sulfate, consistent with the requirement to mutate 6 basic residues to ablate ECM binding (34).
Binding of TIMP-3 to sulfated glycosaminoglycans both inhibits its clearance by LRP1 and increases its affinity for glycosaminoglycans-binding target enzymes through formation of high affinity TIMP-3⅐metalloproteinase⅐glycosaminoglycan complexes (28). In the context of osteoarthritis, both the collagenase MMP-13 and the aggrecanases ADAMTS-4 and ADAMTS-5 are susceptible to such ECM-mediated increase in TIMP-3 potency (10). Comparing the affinities of the mutants for sulfated glycosaminoglycans and LRP1 gives insight into the relative importance of matrix binding and LRP1-mediated clearance for the biological activity of TIMP-3. TIMP-3 K42A/ K110A had greater resistance to endocytosis and lower affinity for LRP1 than was observed for K26A/K45A but also had lower affinity for sulfated glycosaminoglycans and did not display increased affinity for ADAMTS-5 in the presence of heparin. Both mutants protected cartilage more effectively than wildtype TIMP-3, but K26A/K45A showed greater inhibition after extended incubation on cartilage. This suggests that interaction with matrix glycosaminoglyans supports the protective effect of TIMP-3 in the complex, dynamic tissue environment.
Extracellular levels of other LRP1 ligands may be similarly regulated by the equilibrium between matrix binding and

TABLE 2 Affinity of wild-type and mutant TIMP-3 for prototypic target metalloproteinases
Wild-type TIMP-3 and its mutants K26A/K45A and K42A/K110A (0.25-25 nM) were incubated with target metalloproteinases (1 h, 37°C), and residual activity against synthetic fluorescent substrates was determined. Equilibrium velocities were used to calculate the apparent inhibition constant K i (app) (mean Ϯ S.D., n ϭ 3-5 technical replicates) using the tight binding equation. ND, too low to be determined.

TIMP-3 Mutants Resistant to LRP1 Have Increased Half-life
LRP1-mediated uptake that we have observed for TIMP-3. For example, heparin inhibits LRP1 binding of factor IXa (35), apolipoprotein A-V (36), C4b-binding protein (37), and PAI-1 (38,39), and several other LRP1 ligands (e.g. ADAMTS-4, ADAMTS-5, MMP-13, TGF␤, CTGF, midkine, thrombospondin) are known to bind to heparin or heparan sulfate (40 -46). The bioavailability of these proteins is thus likely to be modulated by factors governing their retention on heparan sulfate proteoglycans, such as the pattern and degree of sulfation on heparan sulfate. Matrix turnover may thus be regulated by the dynamic alterations in sulfation that are observed, for example, with age (47) and inflammation (48).
Timp3-null mice also exhibit increased sensitivity to TNFmediated pathologies, such as impaired liver regeneration (49), indicating the importance of TIMP-3 as a regulator of ADAM17, which sheds membrane-bound TNF to release the soluble form of this pleiotropic proinflammatory cytokine. The effects of K26A/K45A and K42A/K110A on TNF-dependent chronic inflammatory conditions would be of interest. Aggrecanase-selective and ADAM17-selective TIMP-3 mutants with increased half-life could be created by combining the LRP1resistant mutants described here with previously characterized selective TIMP-3 mutants (30,50). These mutants or small molecule inhibitors of TIMP-3 endocytosis may be of use in osteoarthritis as well as in other pathological conditions characterized by excess, dysregulated metalloproteinase activity, such as cancer and chronic diabetic wounds.
Construction of TIMP-3 Model-ICM-Pro from Molsoft was used to create a homology model of TIMP-3 based on its sequence similarity with TIMP-2 (22). Global pairwise sequence alignment between the homologues was used as the basis for initial structural superposition followed by constrained optimization of side chains and loops. The distance between ␣-carbon residues of pairs of lysine residues was measured using PyMOL.
Generation of TIMP-3 Mutants-TIMP-3 mutants were generated using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Cheshire, UK), with C-terminally FLAG-tagged wild-type TIMP-3 in pCEP4 mammalian expression vector used as the template (24). All mutants generated are shown in Table 1. Primers for generation of K22A were 5Ј-TCC GAC ATC GTG ATC CGG GCC GCG GTG GTG GGG AAG  that 100 ng of template was used per reaction, and the extension time per cycle was increased to 150 s/kb (1435 s). PCR products were digested with DpnI and transformed into E. coli XL10-Gold ultracompetent bacteria by heat shock, and cells were plated on carbenicillin agar. Candidate colonies were screened by restriction digestion with XhoI and subsequent DNA sequencing (Eurofins MWG Operon, Ebersberg, Germany).
Screening of Mutants-For initial screening of the mutants, we evaluated their resistance to clearance from conditioned medium.
We evaluated clearance of the TIMP-3 mutants from the medium of HTB94 chondrosarcoma cells. HTB94 cells were seeded overnight (8 ϫ 10 5 cells/well of a 12-well plate) in DMEM with 10% FCS. Wells were washed three times in serum-free DMEM and incubated in DMEM with 0.1% FCS and the dialyzed HEK-293-EBNA conditioned medium containing wild-type or mutant TIMP-3 (1 nM). Samples were harvested at various time points (0 -24 h), and proteins were concentrated by the addition of TCA to a final concentration of 5% (18 h, 4°C) (51) and subsequent centrifugation (13,000 rpm, 10 min, 4°C). Pellets were resuspended in SDS-PAGE sample buffer and analyzed by immunoblotting with anti-FLAG M2 antibody. Pixel volumes within the linear range of detection were quantified using Phoretix 1D densitometry software (TotalLab, Newcastle-upon-Tyne, UK), and half-life or percentage of wild-type or mutant TIMP-3 remaining in the medium (mean Ϯ S.D.) was calculated relative to their own pixel volume at t ϭ 0 h (defined as 100%). Half-lives were calculated by fitting the data to a first order exponential decay model using Prism (GraphPad Software, Inc., La Jolla, CA). Data from three independent experiments were analyzed for statistical significance using

TIMP-3 Mutants Resistant to LRP1 Have Increased Half-life
one-way ANOVA with Dunnet's post hoc test for multiple comparisons with Prism (GraphPad Software).
Uptake of wild-type and mutant TIMP-3 by LRP1 wild-type (PEA10) and LRP1-null (PEA13) cells was compared. Cells were seeded overnight at 3 ϫ 10 5 cells/well of a 12-well plate. Wells were washed three times in serum-free DMEM and incubated with purified wild-type or mutant TIMP-3 (1 nM) in DMEM with 0.1% FCS. Conditioned media were analyzed by immunoblotting as described above for HTB94.
Purification of Wild-type and Mutant TIMP-3-Wild-type and mutant TIMP-3 were expressed and purified by modification of the protocol described previously (24). HEK-293-EBNA cells were stably transfected with pCEP4 expression plasmids encoding wild-type or mutant TIMP-3 and selected using hygromycin B (100 g/ml for TIMP-3 mutants, 800 g/ml for wild-type TIMP-3). Stable cell lines were treated with serumfree DMEM containing 30 mM NaClO 3 for 48 h. Conditioned medium was harvested, centrifuged to remove cell debris, and passed over an anti-FLAG M2 affinity resin (Sigma-Aldrich) equilibrated in DMEM. The resin was washed in DMEM containing 1 M NaCl and then extensively washed in DMEM. Bound wild-type or mutant TIMP-3 was eluted with FLAG peptide (200 g/ml) in DMEM and stored at Ϫ80°C in LoBind microcentrifuge tubes (Eppendorf, Stevenage, UK). The purity of isolated TIMPs was confirmed by silver staining (silver staining kit, Pierce), and their active concentrations were determined by titration against a known concentration of MMP-1.

TIMP-3 Mutants Resistant to LRP1 Have Increased Half-life
Binding to Heparin and Heparan Sulfate-Heparin and heparan sulfate (10 g/ml) were coated (overnight, 25°C) onto glyosaminoglycan-binding ELISA plates (BD Biosciences, Oxford, UK) in PBS, and wells were blocked with 0.2% gelatin in PBS (1 h, 37°C). Wells were washed in PBS containing 0.1% Tween 20 after this and every subsequent step. Purified wild-type or mutant TIMP-3 (0.2-25 nM) in blocking solution was applied to wells (3 h, 37°C), and binding was detected with anti-FLAG M2 primary antibody and anti-mouse-HRP-conjugated secondary antibody. 3,3Ј,5,5Ј-Tetramethylbenzidine (Becton Dickinson) substrate was added, the reaction was stopped when appropriate by the addition of 2 N H 2 SO 4 , and absorbance at 450 nm was measured using a FLUOstar OMEGA microplate reader (BMG Labtech). Data (mean Ϯ S.D., n ϭ 3 technical replicates) were analyzed using Prism (GraphPad Software).
Cartilage Explant Cultures-Porcine articular cartilage explants were dissected from metacarpophalangeal joints of 3-9-month-old pigs within 24 h of slaughter using a biopsy punch to ensure uniformity of size. Explants were rested for 2 days in cartilage medium (DMEM containing 100 units/ml penicillin, 100 units/ml streptomycin, 2 mg/ml amphotericin B, 10 mM HEPES) with 10% FCS (500 l of medium, 48-well plates) at 37°C in 5% CO 2 . Cartilage explants were then washed into cartilage medium (250 l) with 0.1% FCS and treated with IL-1 (10 ng/ml, PeproTech, London, UK) and/or wild-type or mutant TIMP-3 (0 -100 nM) for 0 -72 h. IL-1 treatment stimulated catabolic pathways leading to degradation of the cartilage ECM by TIMP-3 target enzymes and release of degraded aggrecan fragments from the cartilage into the conditioned medium (63). Conditioned media were harvested, and aggrecan degradation was quantified (3 technical replicates per condition) using the dimethylmethylene blue (DMMB) dye-binding assay that quantifies chondroitin sulfate-rich fragments of aggrecan released into the medium though the action of ADAMTSs (64). Data from experiments with three biological replicates were analyzed for significance using one-way ANOVA with Dunnet's post hoc test for multiple comparisons.
Aggrecan fragments released into the medium were also analyzed by immunoblotting (2 technical replicates per condition) using a neoepitope antibody that recognizes the new N terminus generated by ADAMTS-dependent cleavage of aggrecan at the TAQE 1771 2A 1772 GEG site (10).
Author Contributions-L. T. and H. N. conceived the study, coordinated the work, and wrote the paper. D. D. generated the TIMP-3 model, R. V. designed the mutants, D. K. S. provided the LRP1-null cells, and C. M. D. generated and analyzed the mutants.