Identification of an ADAMTS-4 Cleavage Motif Using Phage Display Leads to the Development of Fluorogenic Peptide Substrates and Reveals Matrilin-3 as a Novel Substrate*

ADAMTS-4 and ADAMTS-5 are aggrecanases responsible for the breakdown of cartilage aggrecan in osteoarthritis. Multiple ADAMTS-4 cleavage sites have been described in several matrix proteins including aggrecan, versican, and brevican, but no concise predictive cleavage motif has been identified for this protease. By screening a 13-mer peptide library with a diversity of 108, we have identified the ADAMTS-4 cleavage motif E-(AFVLMY)-X(0,1)-(RK)-X(2,3)-(ST)-(VYIFWMLA), with Glu representing P1. Several 13-mer peptides containing this motif, including DVQEFRGVTAVIR and HNEFRQRETYMVF, were shown to be substrates for ADAMTS-4. These peptides were found to be specific substrates for ADAMTS-4 as they were not cleaved by ADAMTS-5. Modification of these peptides with donor (6-FAM) and acceptor (QSY-9) molecules resulted in the development of fluorescence-based substrates with a Km of ∼35 μm. Furthermore, the role of Glu at P1 and Phe at P1′ in binding and catalysis was studied by exploring substitution of these amino acids with the d-isomeric forms. Substitution of P1 with dGlu was tolerable for binding, but not catalysis, whereas substitution of P1′ with dPhe precluded both binding and catalysis. Similarly, replacement of Glu with Asp at P1 abolished recognition and cleavage of the peptide. Finally, BLAST results of the ADAMTS-4 cleavage motif identified matrilin-3 as a new substrate for ADAMTS-4. When tested, recombinant ADAMTS-4 effectively cleaved intact matrilin-3 at the predicted motif at Glu435/Ala436 generating two species of 45 and 5 kDa.

The ADAMTS family (A Disintegrin and Metalloprotease with Thrombospondin Motifs) 2 is comprised of 19 members with varied substrate preferences (reviewed in Ref. 1). ADAMTS-2, -3, and -14 have been identified as procollagen N-proteinases. ADAMTS-13 maintains hemostasis through the proteolysis of von Willebrand factor following platelet binding. ADAMTS-7 and ADAMTS-12 have been shown to cleave cartilage oligomeric matrix protein. ADAMTS-1, -4, and -5 have been identified as proteoglycanases, cleaving aggrecan, versican, and brevican. ADAMTS-8, -9, -15, -16, and -18 also have been grouped in this family, but their activity toward aggrecan is substantially less compared with the other three members (2). The substrate preferences for the remaining family members have not been conclusively determined.
Alignment of the known sequences flanking the ADAMTS-4 cleavage sites in the proteoglycan substrates, aggrecan, versican, and brevican, led to the proposal of a 24-amino acid consensus motif (9). Not surprisingly, a glutamic acid residue occupied P1 (100% conserved) with P2Ј occupied by the basic amino acids, Arg or Lys. The authors speculated that activity of ADAMTS family members toward proteoglycan substrates was primarily dictated by an extended 23-amino acid motif N-terminal to the scissile bond, and a short 3-amino acid motif downstream of the site of cleavage. However, unlike the scissile bonds in the aggregating proteoglycans, the site of ADAMTS-4 proteolysis in ␣ 2 -macroglobulin (␣ 2 M) is Met 690 /Gly 691 , with no requirement for Glu at P1 (10). Yet, P1Ј to P3Ј in ␣ 2 M, Gly-Arg-Gly, is remarkably similar to downstream sequences in aggrecan and brevican, implying that PЈ amino acids may be more important in recognition and catalysis than sequences upstream of the scissile bond. ADAMTS-4 has also been shown to undergo autocatalytic C-terminal truncation, producing 40and 53-kDa species. Similar to ␣ 2 M, Glu at P1 is not required for recognition and catalysis at these sites, with P1 comprised of Lys or Thr and P1Ј and P2Ј being Phe and Arg, respectively (11). The site of cleavage in cartilage oligomeric matrix protein has not been determined, except that a 110-kDa fragment is released upon treatment with ADAMTS-4 (12).
Because a relatively concise consensus motif for ADAMTS-4 catalysis has not been established, the assay of ADAMTS-4 activity has relied on the use of natural or recombinant forms of aggrecan, or peptides spanning the known sites of proteolysis in aggrecan of no less than 27 amino acids in length (13). Detection of the proteolytic products then must rely on HPLC or the use of anti-neoepitope antibodies such as the monoclonal antibody BC-3, which recognizes the neoepitope ARGS located on the C-terminal fragment arising from proteolysis at Glu 373 / Ala 374 in the interglobular domain of aggrecan (14). Due to the size of the substrates and the time period of incubation (0.5 to 2 h), kinetic analysis of protease activity is cumbersome at best.
Peptide phage display has been demonstrated to be a valid means of determining the consensus motifs for a number of proteases including MMP-11 (15), human kallikrein 2 (35), rat mast cell protease 4 (16), and outer membrane protein T (17). Motifs derived from such an analysis have been shown to be physiologically relevant and used to determine preferred as well as possible alternate protease substrates. The peptides identified have also found utility as substrates suitable for kinetic analysis of protease activity.
In this report we, 1) describe the screening of a 10 8 random 13-mer peptide phage library, resulting in the determination of a 7-amino acid cleavage motif for ADAMTS-4; 2) demonstrate that this motif is specific for ADAMTS-4 and peptides based on this motif are poor substrates for ADAMTS-5; 3) confirm the importance of the P1 glutamic acid in catalysis and substrate recognition; 4) describe the development of a novel fluorogenic peptide substrate and kinetic assay amenable to high throughput screening of ADAMTS-4 inhibitors; and 5) report the results of a BLAST search of the cleavage motif, revealing matrilin-3 as a substrate of ADAMTS-4.

Expression and Purification of ADAMTS-4-Full-length
recombinant human ADAMTS-4 and ADAMTS-5 were cloned and expressed in Drosophila Sf9 cells as previously described (18,19) and the conditioned media purified by heparin-Sepharose chromatography. The purified ADAMTS-4 preparation was composed almost solely of the 70-kDa active form. Activity was assessed using an enzyme-linked immunosorbent activity assay employing a 36-mer peptide, spanning the site of proteolysis in the interglobular domain of aggrecan, and labeled at the C-terminal with biotin. Upon cleavage of the peptide substrate, the C-terminal proteolytic fragment was detected by BC-3 monoclonal antibody, recognizing the sequence ARGS (13). Purified ADAMTS-4 and ADAMTS-5 were formulated at a final concentration of 20 nM in 100 mM Tris-HCl, 100 mM NaCl, 10 M CaCl 2 , 0.05% Brij, pH 7.5 (MMP buffer), and stored at Ϫ80°C.
Peptide Phage Display-A substrate phage library containing 10 8 polypeptide sequences displaying 13 amino acids of randomized sequence equimolar at each position for all amino acids except cysteine was constructed at Dyax Corp. (Cambridge, MA) (36). N-terminal to each displayed randomized sequence were two streptavidin-binding peptide sequences. If the associated randomized peptide sequence was a substrate for ADAMTS-4, these streptavidin-binding epitopes would be cleaved from the phage particle. Non-substrate phage clones that retained the streptavidin-binding epitopes would then be captured by streptavidin-coated surfaces. Initially the library was buffer exchanged in Reaction Buffer by incubating a total of 10 12 phage in 500 l of TBS (50 mM Tris, 150 mM NaCl, pH 7.4) plus 1ϫ Roche Complete protease inhibitors minus EDTA (Roche) with 2 mg of streptavidin-coated magnetic beads (Dynal M280, Invitrogen) for 1 h at room temperature in a total volume of 200 l. The beads were washed 7 times with TBST (TBS, 10% Triton X-100, 10% Tween) and once with TNT (5 mM Tris-HCl, 50 mM NaCl, 10% Triton X-100, pH 7.5). Phage were then eluted from the beads with 200 l of 15 M biotin in TNT for 1 h at room temperature, followed by precipitation with 50 l of 20% PEG8000, 3.5 M NH 4 OAc for 1 h on ice. The phage were then resuspended with 500 l of Reaction Buffer containing 1ϫ protease inhibitors and incubated with 1.1 nM ADAMTS-4 for 1 h at room temperature. Uncleaved phage were removed using 5 mg of streptavidin-coated magnetic beads. Cleaved phage released into the supernatant, and not captured by the beads, were amplified and purified to provide 10 12 phage for further processing through four rounds of selection.
Screening was performed using 96-well plates coated with streptavidin (Pierce) and blocked with 3% bovine serum albumin in phosphate-buffered saline. Individual phage clones were incubated with ADAMTS-4 in Reaction Buffer for 1 h at 37°C. Treated and untreated phage were then transferred to streptavidin plates and incubated for 1 h at room temperature, washed 5 times with PBST (phosphate-buffered saline, 10% Tween), and bound phage were assayed by incubation with an antiphage antibody (anti-M13-horseradish peroxidase, GE Healthcare) and developed with TMB peroxidase substrate (KPL, Gaithersburg, MD).
Peptide Sequence Analysis and Motif Generation-A multiple alignment of the peptide sequences was produced with the ClustalW program (20) using the BLOSUM-30 matrix (21). The Pratt motif discovery program was used to identify the conserved patterns in the aligned epitope sequences (22). Peptide sequence logo figures (23) were generated by the local implementation of the WebLogo package (24).
Assay of Proteolytic Activity-Peptides and peptide arrays were either synthesized in house or purchased from New England Peptide (Waltham, MA), or Jerini AG (Berlin, Germany) and were not less than 95% pure. Proteolysis and identification of the scissile bond of substrate peptides was determined by incubating a 10 M solution of the peptide with 400 pM ADAMTS-4 or ADAMTS-5 for 18 h at 37°C. 5 l of each digest was analyzed by reverse phase high performance liquid chromatography (RP-HPLC) on a 75 m ϫ 150-mm Pepmap column (LC Packings, San Francisco, CA) coupled to a quadrupole time-of-flight mass spectrometer (Micromass, Beverly, MA). No prior knowledge of peptide mass was needed. Spectra were acquired in MS mode and MS/MS mode at 1 s/scan. Where MS was not required, 90 l of each digest was analyzed by RP-HPLC on a 4.6 ϫ 150-mm Vydac C-18 column (The Separations Group, Hesperia, CA) with detection at 214 nm.
Direct kinetic fluorescent analysis of peptide turnover was determined using 6-Fam/QSY-9 (6-carboxyfluorescein, QSY-9-maleimide, Molecular Probes, Eugene, OR) fluorescently quenched peptides detailed in Tables 3 and 4. Briefly, a 1 M solution of peptide in MMP buffer was incubated with 400 pM ADAMTS-4 or ADAMTS-5 at 37°C in a total volume of 100 l. Fluorescence at 519 nm was monitored over a 60-min period with excitation of 495 nm.
Proteolysis of Matrilin-3 by ADAMTS-4-Recombinant human matrilin-3 was purchased from R&D Systems (Minneapolis, MN) and reconstituted in MMP buffer at a concentration of 0.5 g/ml. One g of matrilin-3 (750 nM) was incubated with 1.5 nM ADAMTS-4 in a total volume of 25 l at 37°C. N-terminal protein sequencing was performed by automated Edman degradation on an Applied Biosystems model 494 Procise sequencer. The digests were fractionated by SDS-PAGE (NuPAGE), transferred onto a polyvinylidene difluoride membrane, and stained with Coomassie Blue. The bands of interest were excised and placed into the blot cartridge of the sequencer and the samples were run using blot cycles. Model 610A version 2.1 software was employed for data acquisition and processing. SC81956 (s)-2-dimethylamino-N-hydroxy-3,3-dimethyl-4-[(4phenoxyphenyl)sulfonyl]butanamide) was synthesized at Pfizer and is a potent nanomolar inhibitor of both ADAMTS-4 and ADAMTS-5.

ADAMTS-4 Fails to Cleave a Commercially Available Peptide Substrate
Array-A commercial peptide array comprised of 360 fluorescently quenched peptides of 8 amino acids in length, bearing known proteolytic consensus motifs (Jerini AG) was screened for ADAMTS-4 activity. No ADAMTS-4 specific cleavage of the 360 peptides was detected, demonstrating narrow substrate specificity for ADAMTS-4 (data not shown).
Peptide Phage Display Yields 50 Unique Sequences Cleaved by ADAMTS-4-In that a small peptide substrate for ADAMTS-4 had not been identified using existing commercial peptide arrays, and that truncation of longer peptides based on sequences spanning the sites of proteolysis found in aggrecan failed to identify a shorter peptide amenable for using quench/ fluorescent-based substrates for kinetic analysis of proteolytic activity, it was reasoned that a larger peptide library composed of random sequences could identify short peptides cleaved by ADAMTS-4.
ADAMTS-4 was screened against a peptide phage library comprised of ϳ10 8 random 13-amino acid peptides equal-molar for all 20 amino acids except cysteine. This library was screened for peptides that could be cleaved by ADAMTS-4 after binding streptavidin immobilized on polystyrene magnetic beads. Non-cleaved phage was removed by serial incubation with streptavidin beads and the cleaved phage amplified for the next round of selection. After the third and fourth rounds of selection, 96 randomly picked clones were screened for peptide cleavage by phage enzyme-linked immunosorbent assay. The phage clones cleaved in the shortest time of incubation were considered preferred substrates.
Fifty unique substrate phage clones that gave the best cleavage were identified and aligned using the ClustalW algorithm (20). This multiple alignment was then analyzed by the Pratt motif discovery program to identify possible amino acid sequence motifs. After identification of the major motif, the sequences were manually realigned by introducing another gap into some members to maximize the multiple alignments and increase the sensitivity of detecting other important contextual amino acids (Table 1). After manual re-alignment and re-analysis it was determined that E-(AFVLMY)-X(0,1)-(RK)-X(2,3)-(ST)-(VYIFWMLA) was the predominant motif. As expected, Glu is the most important amino acid in the ADAMTS-4 consensus motif and if we assume that Glu is P1 (based on the fact that Glu is P1 in the 5 cleavage sites in aggrecan), then each amino acid can be assigned a position. After Glu, the next most important amino acid for recognition is Arg or Lys (basic amino acids) at P2Ј/P3Ј followed by Thr or Ser (alcohol containing amino acids) at P5Ј/P6Ј/P7Ј.
Peptides Identified from Phage Library Are Specific for ADAMTS-4 and Amenable to Kinetic Analysis of Activity-All 50 peptides identified from the peptide phage screen were synthesized and analyzed for ADAMTS-4-mediated cleavage by RP-HPLC. Two peptides that displayed the greatest turnover were peptides B05 (HNEFRQRETYMVF) and B06 (DVQE-FRGVTAVIR) ( Table 1). LC/MS/MS analysis of peptide B06 was carried out following digestion with ADAMTS-4. Fig. 1 shows the results for peptide B06 with two peptide peaks being identified. After ADAMTS-4 cleavage, one peak with a retention time of 12.151 min was identified as non-cleaved B06, whereas the peak eluting at 11.158 min was identified as the C-terminal fragment of B06 arising from cleavage at Glu/Phe (Fig. 1). The N-terminal fragment was not resolved. Incubation of this same peptide with ADAMTS-5 failed to produce a cleaved product, and only full-length B06 was detected by LC/MS/MS (data not shown). Similarly, no cleavage products were detected for the remaining 48 peptides when assayed with ADAMTS-5, as determined by RP-HPLC. Of the 50 peptides identified from the phage screen, only a subset of the synthesized peptides were analyzed by LC/MS/MS to determine the scissile bond, namely peptides B05, B06, B07, and A08 as well as a peptide based solely on the identified cleavage motif where X-residues were replaced with Ala (data not shown).
Fluorescently quenched versions of B06 and B05 were synthesized using 6-Fam and QSY-9. To increase the solubility of each peptide, the N-and C-terminals were not capped, and amino acids KGK were added to the C terminus, K(6-Fam)-DVQEFRGVTAVIRC(Qsy-9)-KGK (Table 2, peptide 6) and K(6-Fam)-HNEFRQRETYMVFC(Qsy-9)-KGK (Table 3, peptide 6), respectively. Both peptides were found to give a maximal rate of fluorescence increase at a concentration of 1 M with an ADAMTS-4 concentration of ϳ400 pM. Peptide concentrations above 1 M exhibited an inner filter effect, and the fluorescent yield dropped proportionately with increasing peptide concentration. Following a 10-min equilibration of peptide B06 to allow the sample to warm to 37°C, a linear increase in fluorescence was detected for the first 40 min upon addition of ADAMTS-4 (Fig. 2). A slight increase in substrate conversion was noted from 40 to 60 min, indicated by the slight upward curvature of the activity trace possibly due to enhanced solubility of the peptide over time at this temperature. ADAMTS-5 failed to cleave peptide B06, and therefore did not generate a fluorescent signal above background. Similar data were obtained with peptide B05 for both ADAMTS-4 and ADAMTS-5 (data not shown).
K m values for the 6-Fam/QSY-9 derivatives of B05 and B06 could not readily be determined. Although these peptides were significantly more soluble than the non-fluorescent versions because of the addition of the KGK tails, solubility above 100 M was not achievable, and thus meaningful substrate curves could not be generated. K m values could be approximated using the non-fluorescent versions in competition with the 6-Fam/QSY-9 derivatives. The IC 50 of the "cold" peptides, Ac-NEFRQRETYMVF-NH 2 ( Table 3, peptide 1) for B05, and Ac-DVQEFRGVTAVIR-NH 2 (Table 2, peptide 1) for B06, were calculated to be ϳ35 M (Fig. 3). As to whether this value represents the true K m for each peptide, it must be stressed that the competing peptides were significantly different in that they lacked the 6-Fam and QSY-9 labels, as well as the C-terminal KGK extension to aid solubility; therefore, the true K m could be higher.
A 1 M solution of fluorescently quenched peptide B06 was incubated with 400 pM ADAMTS-4 for 1 h at 37°C. Approximately 31% of the peptide was hydrolyzed in this 1-h period, yielding a velocity of 310 nM/h. The turnover rate of ADAMTS-4 for this peptide substrate was calculated to be 0.22 s Ϫ1 . Because the peptide concentration was significantly less than the estimated K m for this peptide, k cat /K m could be calculated directly and was found to be 215,000 M Ϫ1 s Ϫ1 . Similar analysis of fluorescently quenched peptide B05 yielded a k cat /K m value of 160,000 M Ϫ1 s Ϫ1 for ADAMTS-4 (data not shown).

Multiple sequence alignment of the ADAMTS-4 epitope sequences with corresponding average percentage of phagemid cleavage and the derived ADAMTS-4 cleavage motif
Predominant amino acids found at a frequency of greater than 40% in a particular position are illustrated with a black background, whereas related amino acids are shown with a grey background. APRIL (Fig. 4). The modified peptide containing dPhe at P1Ј (Table 2, peptide 4) was shown to be unfavorable, in that competition with the fluorescent substrate was not detected, indicating that the peptide does not bind the active site of ADAMTS-4. Similarly, substitution of P1 and P1Ј with dGlu and dPhe, respectively (Table 2, peptide 5), was also unfavorable, in that this peptide was unable to compete and inhibit cleavage of the fluorescent substrate.

ADAMTS-4 Cleavage Motif
To confirm that the dGlu containing peptide is a competitive, non-cleavable peptide inhibitor of ADAMTS-4, reverse phase chromatographic analysis of peptide Ac-DVQ(e)FRGVTAVIR-NH 2 was performed (Table 2, peptide 3). Following incubation with ADAMTS-4 for 18 h, no cleavage of the scissile bond was detected, demonstrating that the orientation of the Glu at P1 is critical for catalysis (data not shown). Similar results were obtained for the P1 dGlu variant of peptide B05 (Table 3, peptide 4); the D-isomeric peptide was able to compete for substrate binding with an IC 50 of 8 M, but was not cleaved by ADAMTS-4 (data not shown). Like the B06 variant, the dPhe and dGlu/dPhe variants of B05 failed to compete and were not cleaved by ADAMTS-4 (data not shown).
P1 Glutamic Acid Is Critical for Substrate Binding-To determine the role of P1 Glu in binding and catalysis, P1 was substituted with aspartic acid in peptides B05, KHNDFRQRE-TYMVFKGK (Table 3, peptide 3), and B06, Ac-DVQD-FRGVTAVIR-NH 2 (Table 2, peptide 2). These peptides were assayed in competition with their respective 6-Fam/QSY-9 derivatives containing Glu at P1. Although Asp is a conservative substitution, competition was not observed for either peptide, even at concentrations above 100 M (Fig. 5), demonstrating that Glu at P1 is needed for cleavage and recognition by ADAMTS-4. To verify that the lack of competition observed in      the kinetic assay was real and not due to preferential cleavage of the 6-Fam/QSY-9-modified peptides by ADAMTS-4 (due to the presence of the 6-Fam and QSY-9 groups that may lower the K m ), a competition assay was performed with non-fluorescently quenched native peptide B06 ( Table 2, peptide 1) and P1 Aspsubstituted B06 (Table 2, peptide 2). Cleavage was then monitored by LC. Similar to the fluorescent kinetic data, little or no inhibition of cleavage of the native peptide was observed, even at concentrations above 100 M of the P1 Asp competing peptide (data not shown). Longer incubation (24 h) of the P1 Aspsubstituted peptide B06 with ADAMTS-4 failed to produce proteolytic products as determined by LC (data not shown). Collectively, the data suggest that Glu at P1 is critical for recognition and cleavage.
Incubation of ADAMTS-4 with recombinant human matrilin-3 (50 kDa) produced three bands after 30 min at 37°C, as visualized by SDS-PAGE (Fig. 7). N-terminal sequencing of each band identified the 50-kDa species as mature, intact, nonproteolyzed matrilin-3; the 45-kDa species (with the native N-terminal of matrilin-3) results from truncation at the C terminus, and a 5-kDa protein bearing the N-terminal sequence, 436 ARRLVS, corresponds to cleavage of matrilin-3 at one of the predicted sites of cleavage at Glu 435 /Ala 436 . Interestingly, no other proteolytic fragments were detected even though the BLAST results identified a potential site of cleavage in the von Willebrand factor A (VWF A) domain of matrilin-3 at Glu 98 / Phe 99 . To verify that the proteolysis of matrilin-3 was ADAMTS-4 mediated and not due to a contaminating protease, the small molecule aggrecanase inhibitor, SC81956, was included in the digest and no cleavage of matrilin-3 was observed after 24 h (Fig. 7, lane 9). Additionally, recombinant ADAMTS-5, expressed and purified in an analogous manner as ADAMTS-4, failed to process matrilin-3, producing no detectable proteolytic fragments (data not shown).

DISCUSSION
ADAMTS-4 has been shown to play a major role in the initiation and progression of osteoarthritis, in human cartilage, through the proteolysis of aggrecan as well as other proteoglycan and non-proteoglycan substrates, predominantly at sites containing a Glu at the P1 residue of the scissile bond. Alignment of the known ADAMTS-4 cleavage sites has failed to produce a concise motif unique to this protease. We report for the first time the derivation of a 7-amino acid cleavage motif specific for ADAMTS-4 relative to ADAMTS-5: E-(AFVLMY)-X(0,1)-(RK)-
It has been shown that deglycosylation of aggrecan or removal of the thrombospondin domain of ADAMTS-4 impedes proteolysis, indicating that exosite interactions are a strong component of ADAMTS-4-mediated proteolysis of aggrecan as well (25). A question that arises from the data is why are the peptides identified from the phage screen substrates of ADAMTS-4, given that this motif is extremely similar to the scissile bonds in aggrecan, yet aggrecan is not cleaved when the GAG chains are removed? The answer could lie in substrate presentation. The peptide substrates are short 13-mers, likely lacking any rigid secondary structure and could conceivably conform to the active site of ADAMTS-4, contrasting with a protein substrate where the scissile bond is likely held in a more rigid three-dimensional conformation. A K m value for the peptides was not determined due to solubility issues, but a lower limit was estimated to be ϳ35 M, representing modest affinity.
Conformational restraint of the scissile bond coupled with a mediocre affinity for the core protein may then explain why deglycosylated aggrecan is no longer a substrate of ADAMTS-4. Although the K m for the synthetic peptides was average, they did prove valuable as fluorescently quenched substrates for the kinetic analysis of ADAMTS-4 activity (Fig. 2). Solubility of both peptides (B05 and B06)  was enhanced by the inclusion of a Lys at the N-terminal, and Lys-Gly-Lys at the C-terminal of each peptide. Solubility was further enhanced by the addition of the long-range quencher QSY-9 (Molecular Probes). Although a K m could not be measured directly, assay of ADAMTS-4 activity at a concentration of 1 M 6-FAM/QSY-9 peptide B05 or B06, which is significantly lower than the K m (ϳ35 M), allowed for the determination of the specificity constant (k cat /K m ) for each peptide. k cat /K m values for peptides B06 and B05 were calculated to be 215,000 and 160,000 (M Ϫ1 s Ϫ1 ), respectively, and are in agreement with values published for coumarin-labeled PUMP and gelatinase peptide substrates, where k cat /K m ranged from 169,000 to 690,000 (M Ϫ1 s Ϫ1 ), respectively (26), but are an order of magnitude larger than MMP-11 peptide substrates identified by phage display (15), or coumarin-labeled stromelysin 1 peptide substrates (27). Although the fluorescently quenched peptides B05 and B06 are amenable to the high throughput screening of ADAMTS-4 inhibitors, it should be stressed that molecules that disrupt positive exosite interactions between ADAMTS-4 and a natural substrate may not be identified using these short cleavage motif peptides as substrates.

TABLE 4 Blast results of motif E-(AFVLMY)-X(0,1)-(RK)-X(2,3)-(ST) for extracellular matrix-relevant proteins
The importance of the P1 Glu and P1Ј Phe residue was examined in both peptides B05 and B06, either by changing the stereochemistry of the Glu or Phe residues, or by substituting P1 Glu with Asp. Interestingly, substitution with dGlu at P1 in both peptides was favorable, whereas inclusion of dPhe was not tolerated (Fig. 4). The dGlu substitution seemed to enhance binding in that the measured IC 50 for peptides B05 and B06 with P1 dGlu was ϳ8 -10 M compared with 35 M for the native sequences. It can be hypothesized that substitution at P1 with dGlu produces a favorable kink in the peptide chain, allowing for enhanced substrate binding; future homology modeling will address this. Although the dGlu-substituted peptides bound ADAMTS-4, they were not cleaved, suggesting that orientation of the Glu residue is critical for catalysis. Perhaps this is mediated through specific interaction of the Glu residue with the amino acid side chains in the catalytic cleft of ADAMTS-4. In addition, inclusion of dGlu may indirectly disrupt scission of the peptide by displacing favorable P and PЈ interactions, resulting in misalignment of the P1 residue. Substitution at P1Ј with dPhe, on the other hand, produced peptides that were not recognized or cleaved by ADAMTS-4, implying P1Ј binding into the S1Ј pocket of ADAMTS-4 is very stringent with regard to stereochemistry. Examination of Table 1 clearly shows the importance of P1 being occupied by glutamic acid. Conservative replacement of P1 with Asp in peptides B05 and B06 failed to produce a peptide that would compete for binding with the native P1 Glu-substituted peptides (Fig. 5).
So, if P1 Glu is critical for binding and catalysis, how does ADAMTS-4 cleave ␣ 2 M or undergo autocatalytic processing where P1 is not Glu but Met or [Lys,Thr], respectively? An examination of the 50 peptides identified from the phage screen also reveals the presence of peptide substrates devoid of a Glu residue, ϳ16% ( For ␣ 2 M and autocatalytic processing, the contribution of substrate presentation may be important, even though the cleavage motifs in these proteins may not be optimal compared with the peptide derived motif that we have identified. Further evidence weakens the contribution of amino acids upstream from P1 with respect to substrate recognition, based on preliminary data that show that N-terminal truncation up to the P1 Glu of peptide EFRQRETYMVFKGK (Table 3, peptide 5) is not detrimental to binding/recognition (data not shown). Collectively, these data suggest that PЈ amino acids play a predominant role in substrate recognition, binding, and catalysis. More studies are required to address this subject.
Results from the BLAST search using the motif E-(AFV-LMY)-X(0,1)-(RK)-X(2,3)-(ST) produced over 9000 hits in ϳ5700 proteins. Apart from aggrecan and the proteoglycans versican and brevican, other cartilage proteins were identified including: cartilage homeoprotein 1, cartilage intermediate layer protein-1 and -2 percursors, and matrilins 1-4. The matrilin family is highly conserved and ubiquitously expressed, the distinguishing features between members being the number of VWF A domains and epidermal growth factor repeats. Matrilin-3 is expressed in articular cartilage and has been shown to play a role in matrix stabilization through the formation of covalent and non-covalent interactions with itself, type II collagen, and aggrecan (28). Mutations in the VWF A domain of matrilin-3 have been linked to multiple epiphyseal dysplasias (29), whereas a missense mutation in the epidermal growth factor domain tracks with OA of the hand (30). Recently, it was shown that mice deficient in matrilin-3 spontaneously develop OA (31). Two possible sites of ADAMTS-4-mediated proteolysis of matrilin-3 were identified (Fig. 6), one site in the VWF A domain and the second site in the oligomerization domain. ADAMTS-4 efficiently processed matrilin-3 producing three distinct species (Fig. 7, lane 5). N-terminal sequence analysis of each band identified the 50-and 45-kDa bands as having the native mature N-terminal sequence of matrilin-3, whereas the 5-kDa band bore the N terminus ARRLVS, corresponding to cleavage of matrilin-3 at Glu 435 /Ala 436 . Interestingly, cleavage within the VWF A domain was not detected, even though this site matched the cleavage motif for ADAMTS-4. This site may be cryptic in the intact matrilin-3 molecule, and therefore unavailable for proteolysis, whereas in the matrix this site may be exposed due to the propensity of matrilin-3 to form interactions with other matrix components. The list of ADAMTS-4 substrates continues to grow and now includes matrilin-3. The fact that ADAMTS-5 failed to cleave matrilin-3 suggests that ADAMTS-4 may represent a more matrix destabilizing force in OA. To date ADAMTS-4, but not -5, has been shown to cleave multiple cartilage matrix proteins other than aggrecan, including cartilage oligomeric matrix protein (12), biglycan (32), TIMP-4 (33), matrilin-2 (34), and now matrilin-3.