The Roles of Substrate Thermal Stability and P2 and P1′ Subsite Identity on Matrix Metalloproteinase Triple-helical Peptidase Activity and Collagen Specificity*

The hydrolysis of collagen (collagenolysis) is one of the committed steps in extracellular matrix turnover. Within the matrix metalloproteinase (MMP) family distinct preferences for collagen types are seen. The substrate determinants that may guide these specificities are unknown. In this study, we have utilized 12 triple-helical substrates in combination with 10 MMPs to better define the contributions of substrate sequence and thermal stability toward triple helicase activity and collagen specificity. In general, MMP-13 was found to be distinct from MMP-8 and MT1-MMP(Δ279–523), in that enhanced substrate thermal stability has only a modest effect on activity, regardless of sequence. This result correlates to the unique collagen specificity of MMP-13 compared with MMP-8 and MT1-MMP, in that MMP-13 hydrolyzes type II collagen efficiently, whereas MMP-8 and MT1-MMP are similar in their preference for type I collagen. In turn, MMP-1 was the least efficient of the collagenolytic MMPs at processing increasingly thermal stable triple helices and thus favors type III collagen, which has a relatively flexible cleavage site. Gelatinases (MMP-2 and MMP-9(Δ444–707)) appear incapable of processing more stable helices and are thus mechanistically distinct from collagenolytic MMPs. The collagen specificity of MMPs appears to be based on a combination of substrate sequence and thermal stability. Analysis of the hydrolysis of triple-helical peptides by an MMP mutant indicated that Tyr210 functions in triple helix binding and hydrolysis, but not in processing triple helices of increasing thermal stabilities. Further exploration of MMP active sites and exosites, in combination with substrate conformation, may prove valuable for additional dissection of collagenolysis and yield information useful in the design of more selective MMP inhibitors.

The hydrolysis of collagen (collagenolysis) is one of the committed steps in extracellular matrix turnover. Within the matrix metalloproteinase (MMP) family distinct preferences for collagen types are seen. The substrate determinants that may guide these specificities are unknown. In this study, we have utilized 12 triple-helical substrates in combination with 10 MMPs to better define the contributions of substrate sequence and thermal stability toward triple helicase activity and collagen specificity. In general, MMP-13 was found to be distinct from MMP-8 and MT1-MMP(⌬279 -523), in that enhanced substrate thermal stability has only a modest effect on activity, regardless of sequence. This result correlates to the unique collagen specificity of MMP-13 compared with MMP-8 and MT1-MMP, in that MMP-13 hydrolyzes type II collagen efficiently, whereas MMP-8 and MT1-MMP are similar in their preference for type I collagen. In turn, MMP-1 was the least efficient of the collagenolytic MMPs at processing increasingly thermal stable triple helices and thus favors type III collagen, which has a relatively flexible cleavage site. Gelatinases (MMP-2 and MMP-9(⌬444 -707)) appear incapable of processing more stable helices and are thus mechanistically distinct from collagenolytic MMPs. The collagen specificity of MMPs appears to be based on a combination of substrate sequence and thermal stability. Analysis of the hydrolysis of triple-helical peptides by an MMP mutant indicated that Tyr 210 functions in triple helix binding and hydrolysis, but not in processing triple helices of increasing thermal stabilities. Further exploration of MMP active sites and exosites, in combination with substrate conformation, may prove valuable for additional dissection of collagenolysis and yield information useful in the design of more selective MMP inhibitors.
Current studies identify at least 25 different collagen types, each with a specific role in the extracellular matrix (1,2). The hydrolysis of collagen (collagenolysis) is one of the committed steps in extracellular matrix turnover (3). The triple-helical structure of collagen renders it resistant to most proteases. In vertebrates, enzymes capable of cleaving the triple-helical structure include cathepsin K and collagenolytic matrix metalloproteinase (MMP) 2 family members. One or more of the interstitial collagens (types I-III) are hydrolyzed within their triple-helical domain by MMP-1, MMP-2, MMP-8, MMP-13, MMP-18, MT1-MMP (MMP- 14), and MT2-MMP (MMP-15) (4, 5). MMP-9 cleaves the triple helix of types V and XI collagen (6) but not of types I-III (7).
Types I-III collagen are all fibrillar interstitial collagens, but differences in their sequences, glycosylation patterns, and tissue distribution have long been documented (1, 8 -10). For example, type I collagen has a low level of glycosylation and is found in skin, bone, cornea, and tendon, whereas type II collagen has much higher levels of glycosylation and is found in cartilage (1,9). In a similar fashion to type I collagen, type III collagen has low levels of glycosylation but has a much more restricted tissue distribution (skin, aorta, uterus, and intestine) than type I (1,9). Types V and XI collagen are also fibrillar collagens (1,10). Within the MMP family distinct preferences for collagen types are seen (11). MMP-1 hydrolyzes type III collagen more rapidly than type I, whereas MMP-8 and MT1-MMP show a slight preference for type I collagen compared with type III (11,12). Neither MMP-1 nor MMP-8 hydrolyze type II collagen efficiently (11). Conversely, MMP-13 prefers type II collagen and hydrolyzes this collagen much more rapidly than MMP-1 or MMP-8 (11). Type V collagen is hydrolyzed by MMP-2 and MMP-9 but not MMP-1, MMP-8, or MMP-13, * This work was supported by National Institutes of Health Grants AR 40994 (to K. B.), CA 98799, and EB 000289 (to G. B. F.), the Wellcome Trust Grant 057508 (to H. N.), and the Florida Atlantic University Center of Excellence in Biomedical and Marine Biotechnology. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
whereas type XI collagen is cleaved by MMP-1, MMP-2, and MMP-9 but not MMP-13 (11). Traditionally, substrate sequence is viewed as the guiding principle for protease specificity. However, it has been demonstrated that the collagen specificity of MMPs is not guided by sequence alone (13,14). Thus, the substrate features that may guide these specificities, such as local triple helix instability, possibly in combination with subtle variations in sequence, are unknown.
Whatever the collagen specificity determinants are, they must reflect differences in MMP collagen-binding and active sites. Although complex, several steps in collagenolysis have been defined. Collagen is first bound by the MMP via secondary substrate-binding sites, termed exosites (5,15). The initial binding event is responsible for proper orientation and destabilization of the collagen molecule (16 -18). The collagen triple helix is unwound, and individual ␣1 chains are bound by the active site and subsequently cleaved in rapid succession (17). Collagenolysis-related exosites have been identified within the MMP-1 catalytic domain (prepro-MMP-1 residue numbers 202-210) (19) and the active site cleft (20) and hypothesized in analogous regions in MMP-2, MMP-8, MMP-9, and MMP-13 (21). Collagen-binding exosites have also been proposed in the MMP linker and/or hemopexin-like domains (5,17,19,(21)(22)(23)(24). Unique interactions between the catalytic and hemopexin-like domains are found in pro-MMP-1 compared with MMP-1 (25), suggesting that binding of substrate may result in considerable conformational changes within MMPs. Ultimately, dissecting out collagen specificity elements requires identification of specific MMP-binding sites and their roles in the steps of collagenolysis.
To correlate collagen selectivity with MMP structural features, we need to examine the interactions occurring between MMPs and collagen on a molecular level. Triple-helical peptide (THP) substrate models of collagen have allowed for significant advancement in the understanding of collagenolysis. For example, we recently demonstrated that MT1-MMP is much more active than MMP-1 in cleaving triple-helical collagen models (26). The same study indicated that increased thermal stability of substrates translated into decreased ability of enzymes to cleave such substrates. However, MMP-1 and MT1-MMP were affected to a different degree, as MT1-MMP was better able to hydrolyze more thermally stable substrates. Additionally, substitution of Leu by Cys(Mob) in the P 1 Ј subsite of a consensus MMP substrate led to increased specificity toward MT1-MMP. These observations lead us to believe that by modifying substrate sequence and thermal stability, we can better rationalize the collagen specificities of MMPs. In this study we explore the effects of substitutions in the P 2 and P 1 Ј substrate subsites on MMP triple-helical specificity. Activity has also been examined herein as a function of triple-helical thermal stability. This has required the use of "peptide amphiphiles," in which the thermal stability of the triple helix is modulated by pseudo-lipids attached to the N terminus of the peptide (26,27). We have utilized fluorogenic triple-helical substrates ( Fig. 1) to compare the triple-helical peptidase activities of 10 MMPs. In tandem, MMP-8 mutant triple-helical peptidase activities have been quantified to evaluate the role of Tyr 210 on substrate selectivity, as this residue in MMP-1 and MMP-8 had been shown previously to be an important participant in collagenolysis (19,28).
Peptide Purification-RP-HPLC purification was performed on a Rainin autoprep system with a Vydac 218TP152022 C 18 column (15-20-m particle size, 300 Å pore size, 250 ϫ 22 mm) at a flow rate of 10.0 ml/min. Eluants were 0.1% trifluoroacetic acid in water (A) and 0.1% trifluoroacetic acid in acetonitrile (B). The elution gradient was adjusted as required. Detection was by absorbance at ϭ 220 nm. Analytical RP-HPLC and MALDI-TOF MS (see below) were used to identify fractions of homogeneous product.
Peptide Analyses-Analytical RP-HPLC was performed on a Hewlett-Packard 1100 liquid chromatograph equipped with a Vydac 218TP5415 C 18 RP column (5-m particle size, 300 Å pore size, 150 ϫ 4.6 mm). Eluants were 0.1% trifluoroacetic acid in water (A) and 0.1% trifluoroacetic acid in acetonitrile (B). The elution gradient was 0 -100% B in 20 min with a flow rate of 1.0 ml/min. Detection was at ϭ 220, 324, and 363 nm. MALDI-TOF-MS was performed on a Applied Biosystems Voyager DE-STR MALDI-TOF mass spectrometer using ␣-cyano-4-hydroxycinnamic acid matrix (35 Circular Dichroism Spectroscopy-CD spectra were recorded over the range ϭ 190 -250 nm with a Jasco J-810 spectropo-larimeter using a 1.0-cm path-length quartz cell. The sample concentration was 2 M in EAB (see below). Thermal transition curves were obtained by recording the [molar] ellipticity ([⍜]) at ϭ 225 nm, whereas the temperature was continuously increased in the range of 5-95°C at a rate of 0.2°C/min. Temperature was controlled using a Jasco PFD-425S temperature control unit. For samples exhibiting sigmoidal melting curves, the inflection point in the transition region (first derivative) is defined as the melting temperature (T M ).
fTHP Assay-Substrate stock solutions were prepared at various concentrations with 0.25-1.0% Me 2 SO in EAB buffer (50 mM Tricine, 50 mM NaCl, 10 mM CaCl 2 , 0.05% Brij-35, pH 7.5). MMP assays were conducted in EAB buffer by incubating a range of substrate concentrations (1.9 -11.4 M) with 10 nM enzyme at 30°C. Fluorescence was measured on a Molecular Devices SPECTRAmax Gemini EM dual scan microplate spectrofluorometer using excitation ϭ 324 nm and emission ϭ 393 nm. Rates of hydrolysis were obtained from plots of fluorescence versus time, using data points from only the linear portion of the hydrolysis curve. The slope from these plots was divided by the fluorescence change corresponding to complete hydrolysis and then multiplied by the substrate concentration to obtain rates of hydrolysis in units of micromolar/s. Kinetic parameters were evaluated by Lineweaver-Burk, Eadie-Hofstee, and Hanes-Woolf analyses. In a few cases, K M values were found to be slightly greater than the maximum substrate concentrations utilized, because of limitations of substrate solubility. For these circumstances, data were additionally analyzed using nonlinear regression, one-site hyperbolic binding model with GraphPad Prism 4 software. The relationship between the rate of hydrolysis and substrate concentration (below K M ) for MMP/fTHP pairs for which individual kinetic parameters were determined was found to be linear, with an overall R 2 value of 0.9703 Ϯ 0.0221. MMP substrate cleavage sites were established by MALDI-TOF MS.
Inhibition Assay-Stock solutions of C 10 -fTHP-10 and fTHP-4 were prepared as above. The following concentrations of C 10 -fTHP-10 were prepared in EAB buffer: 5.0, 2.5, 1.0, 0.5, 0.25, and 0.10 M. Into a 384-well plate, 36 l of 10 nM MMP-8 was added to 36 l of each C 10 -fTHP-10 concentration and incubation proceeded for 1 h at 37°C. After incubation, 8 l of 40 nM fTHP-4 was added to the enzyme-inhibitor complex, and fluorescence was recorded for 15 min as described above. Rates of hydrolysis were obtained as described above. K i was determined using Dixon analysis for competitive inhibition (48).
Collagen Assay-DQ TM bovine skin type I collagen (fluorescein-conjugated) was purchased from Molecular Probes (Eugene, OR). A stock solution was prepared in EAB buffer. MMP assays were conducted in EAB buffer by incubating a range of substrate concentrations (0.028 -0.290 M) with 0.0125 nM enzyme at 37°C. Fluorescence was measured using excitation ϭ 485 nm and emission ϭ 535 nm. Rates of hydrolysis and kinetic parameters were obtained as described above.

Design of Fluorogenic Triple-helical (fTHP) Substrates-
The general considerations for the design and synthesis of fTHP substrates, including Mca as a fluorophore and Dnp as a quencher, were reported previously (26,31,49,50). For this study, THPs were designed to explore the roles of substrate as follows: (a) subsite substitutions and (b) thermal stabilities on collagenolytic activity. The triple-helical specificity of MMPs was examined via substitutions in the P 2 and P 1 Ј substrate subsites. Prior experimental and computational studies indicated that these subsites may be exploited to obtain selectivity between collagenolytic MMPs (51,52). The template substrate, fTHP-4 ( Fig. 1), was reported previously (26,31,53). fTHP-4 incorporates a consensus sequence derived from the collagenolytic MMP cleavage sites in human types I-III collagen (11). To serve as the FRET fluorophorequencher pair, Mca and Dnp are linked to Lys side chains in the P 5 and P 5 Ј subsites, respectively. The consensus sequence was subsequently modified in the P 1 Ј subsite, where Leu was replaced with Cys(Mob), to create fTHP-9 ( Fig. 1). Collagenolytic MMPs differ in the depth of their S 1 Ј subsite pocket (52), and computational studies indicate that large, bulky hydrophobic groups in the substrate P 1 Ј position could result in substrate discrimination within collagenolytic MMPs (51). The Cys(Mob) residue in the P 1 Ј subsite was found previously to enhance MT1-MMP activity and specificity in linear (54) and triple-helical (26) contexts. Both fTHP-4 and fTHP-9 were substituted in position P 2 (Gln to Orn), creating fTHP-10 and fTHP-11, respectively (Fig. 1). This substitution was based on a computational study where it was hypothesized that a residue with a small, positively charged side chain would interact favorably with MT1-MMP because of the presence of Glu 210 in the MT1-MMP S 2 -binding pocket (51). Such a substitution could enhance selectivity for MT1-MMP because of the structural differences within the S 2 pockets of collagenases (51). All substrates were acylated with decanoic (C 10 ) or palmitic (C 16 ) acids to confer differential thermal stability. Our prior study had demonstrated that MMP triple-helical peptidase activity is dependent on the substrate thermal stability (26). The purified, lipidated substrates exhibited decreased solubility in assay buffer compared with the nonlipidated substrates. Me 2 SO (0.25-1.0% in stock solutions) was utilized for improved substrate solubility, but the upper limit of substrate concentrations was nonetheless limited to a maximum of 12 M.
Substrate Selectivity Testing-Initially, 10 MMPs were screened for triple helicase activity. MMP-1, -8, and -13 and MT1-MMP were chosen based on their ability to cleave types I-III collagen (12, 59 -63). Of the gelatinase family members, MMP-2 is also known to cleave types I and III collagen (64,65), in contrast to MMP-9, which does not process types I-III collagens (66). It is also of interest to compare the capabilities of gelatinases and collagenases to process more thermally stable substrates, and thus both gelatinases were included in these studies. MT2-MMP was included because, although the full range of activity is unknown, it has been reported to cleave type I collagen (67,68). MMP-3 was chosen to provide a negative control for secreted MMPs, because it was previously shown to have poor or no triple-helical peptidase activities (49). MT5-MMP and MT6-MMP were used as noncollagenolytic controls for transmembrane type and glycosylphosphatidylinositol-anchored MMPs, respectively. The 10 MMPs  ences in triple-helical peptidase activity were only slight (40). One may assume that MT2-MMP behaves in a similar fashion as MT1-MMP in terms of the role of the C-terminal domain. The C-terminal domain of MMP-9 does not modulate proteolytic activities of the enzyme (70,71). As discussed previously (26), the C-terminal hemopexin-like domain may ultimately be involved in the following: (i) movement of the MMP along collagen to the site of hydrolysis and (ii) distortion of the triple helix to facilitate hydrolysis. The C-terminal domain is not critical for triple-helical peptidase activity, as noted previously (23,26,31,40,(72)(73)(74).
Rates of hydrolysis were next compared for fTHP-9 (Gln in P 2 , Cys(Mob) in P 1 Ј) (Fig. 2B). Substitution of Leu in P 1 Ј by Cys(Mob) was detrimental for MMP-1, MMP-2, and MMP-9(⌬444 -707) activities, but the rest of the collagenases and MT2-MMP(⌬268 -628) showed higher activity compared with fTHP-4 (Gln in P 2 and Leu in P 1 Ј). MMP-8 showed the most dramatic increase in activity as compared with fTHP-4 (more The substitution of Gln in the P 2 subsite position by Orn was compared using fTHP-10 (Orn in P 2 and Leu in P 1 Ј) or fTHP-11 (Orn in P 2 and Cys(Mob) in P 1 Ј) (Fig. 2, C and D). In general, substitution by Orn in combination with Leu (fTHP-10) resulted in lower activities for most MMPs compared with fTHP-4 (Fig. 2, A and C). In relative terms, much lower activities were seen by MMP-8 and MT1-MMP(⌬279 -523). Conversely, MMP-13 showed higher activity with Orn in P 2 and Leu in P 1 Ј. The ability of MMP-8 to hydrolyze this substrate was limited to the least thermally stable variant, whereas the Orn substitution enhanced MT2-MMP(⌬268 -628) activity toward a more thermally stable substrate. MMP-13 and MT1-MMP(⌬279 -523) showed similar relative activity levels for both variants of this substrate.
Substitution by Orn in combination with Cys(Mob) resulted in an inability of MMP-1 to cleave the substrate (Fig. 2D). The other collagenases and MT2-MMP(⌬268 -628) showed very good rates of hydrolysis. Addition of C 10 and C 16 drops MMP activity significantly, as generally seen with other substrates. In the case of the C 16 variant, MT1-MMP(⌬279 -523) shows complete loss of activity despite being the most active enzyme toward the "no tail" and C 10 variants of this substrate.
MMP-8 and MT1-MMP(⌬279 -523) showed similar overall patterns of activity, with fTHP-10 (Orn in P 2 and Leu in P 1 Ј) being hydrolyzed the most slowly and fTHP-9 (Gln in P 2 and Cys(Mob) in P 1 Ј) the most rapidly hydrolyzed substrate. However, MMP-8 activity increased 68-fold from fTHP-10 to fTHP-9, compared with 8.4-fold for MT1-MMP(⌬279 -523). For both enzymes, the presence of Cys(Mob) in the P 1 Ј subsite had a beneficial effect on activity (compare fTHP-9 to fTHP-4), and both enzymes were most active toward fTHP-9. The presence of Orn in the P 2 subsite together with Cys(Mob) in P 1 Ј, compared with fTHP-9, had a more negative effect on MMP-8 than on MT1-MMP(⌬279 -523), decreasing MMP-8 activity 4-fold and MT1-MMP(⌬279 -523) activity ϳ30%. When Orn was present without Cys(Mob), both enzymes exhibited their lowest activities. Effects of substitutions on MMP-13 activity were less profound; the most active substrate (fTHP-9) was only hydrolyzed ϳ4 times more rapidly than the least active one (fTHP-4). Another interesting dissimilarity is in the response of MMP-13 to the Orn substitution, as it seemed to be opposite to MMP-8 and MT1-MMP(⌬279 -523). The presence of Orn in fTHP-10 (Orn in P 2 and Leu in P 1 Ј) leads to an increase in MMP-13 activity as compared with fTHP-4 (Gln in P 2 and Leu in P 1 Ј), whereas the substrate that has both Orn and Cys(Mob) was hydrolyzed at a slightly greater rate than the substrate with just Cys(Mob).
Evaluation of MMP Kinetic Parameters-The MMPs exhibiting significant triple-helical peptidase activities in the rates of hydrolysis studies were further compared by determining individual kinetic parameters for fTHP hydrolysis. Of all the MMPs tested, MT1-MMP(⌬279 -523) showed the highest activity toward fTHP-4 (Gln in P 2 and Leu in P 1 Ј). The parameter responsible for this difference is k cat , which is considerably higher for MT1-MMP(⌬279 -523) compared with MMP-1, MMP-8, and MMP-13 (Table 1). An increase in the thermal stability of fTHP-4 most significantly affects MT1-MMP(⌬279 -523), as seen by a 600-fold decrease of k cat , and MMP-1, as seen by a virtual loss of activity ( Fig. 2A). In contrast, MMP-8 and MMP-13 do not show large activity changes with an increase in substrate thermal stability, and both have comparable or better activity toward C 10 -fTHP-4 than fTHP-4. For MMP-13 and MT1-MMP(⌬279 -523), an increase in substrate thermal stability mostly decreases K M values, i.e. the more stable the triple helix, the better the binding by the enzyme. 4 fTHP-9 differs from fTHP-4 by the substitution of Cys(Mob) for Leu in the P 1 Ј subsite. Cys(Mob) has a beneficial effect on MMP-8, MMP-13, and MT1-MMP(⌬279 -523) rates of hydrolysis (Fig. 2B). K M is decreased 3-fold relative to fTHP-4 in the case of MT1-MMP(⌬279 -523), which resulted in a much higher k cat /K M value ( Table 2). This can be explained by additional hydrophobic interactions formed by the side chain of Cys(Mob) residue and residues of the S 1 Ј pocket of this enzyme (51). The lack of a similar effect for MMP-13 can be rationalized by the totally open MMP-13 S 1 Ј pocket and therefore lack of additional interactions with Cys(Mob) (51). For both MMP-8 4 The K M value is assumed to represent MMP binding to THPs, as turnover values (k cat ) are relatively slow compared with k on and k off rates for THP binding by MMPs (23). A previous study found that MMP-1(E200A) bound to a heterotrimeric type I collagen model THP with a K D ϭ 3.7 M (23). This K D value is similar to the K M value for MMP-1 hydrolysis of fTHP-3 (Table 1), and thus the assumption of K M ϳ K D appears valid.  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50

MMP Triple-helical Specificity
and MMP-13, k cat is dramatically improved by the substitution of Cys(Mob) for Leu ( Table 2). An increase in the thermal stability of fTHP-9 yields the same k cat trends as seen with fTHP-4 (Table 1), ultimately resulting in an almost equal loss of activity by all three enzymes. fTHP-10 (Orn in P 2 and Leu in P 1 Ј) differs from fTHP-4 by the substitution of Orn for Gln in the P 2 subsite. Orn had a detrimental effect on the MT1-MMP(⌬279 -523) rate of hydrolysis (Fig. 2C). k cat was decreased compared with fTHP-4 ( Table 1 versus Table 3). MMP-13 showed an increase in k cat and k cat /K M compared with fTHP-4 ( Table 1 versus Table 3 fTHP-11 (Orn in P 2 and Cys(Mob) in P 1 Ј) differs from fTHP-9 (Gln in P 2 and Cys(Mob) in P 1 Ј) by substitution of Orn for Gln in the P 2 subsite, from fTHP-10 (Orn in P 2 and Leu in P 1 Ј) by the substitution of Cys(Mob) for Leu in the P 1 Ј subsite, and from fTHP-4 (Gln in P 2 and Leu in P 1 Ј) by both substitutions. As compared with fTHP-10 (Table 3 versus  Table 4), the presence of Cys(Mob) restores MMP-8 activity. The more thermally stable variants of fTHP-11 have better affinities for MMP-8 but greatly reduced k cat values. It is therefore possible that the inability of MMP-8 to cleave C 10 -fTHP-10 was because of too high a binding affinity. MT1-MMP(⌬279 -523) exhibits a similar trend with the C 10 and C 16 variants of fTHP-11, showing lower K M and k cat values with C 10 and no activity with C 16 . MT2-MMP(⌬268 -628) hydrolyzed fTHP-11 efficiently but showed no ability to hydrolyze more thermally stable variants.    The MMP-8 mutant Y189F (equivalent to MMP-1 position 210 (76)) had been shown to have 3-fold less activity toward collagen than wild type MMP-8, but only slightly (ϳ12%) decreased activity toward Mca-Pro-Leu-Gly-Leu-[3-(2,4-dinitrophenyl)-L-2,3-diaminopropionyl]-Ala-Arg-NH 2 (28). Thus, MMP-8(Y189F) was considered "sensitive" to triple-helical substrates. MMP-8(Y189F) activity toward the best MMP-8 substrate, fTHP-9 (Gln in P 2 and Cys(Mob) in P 1 Ј), was examined and compared with MMP-8 activity. MMP-8(Y189F) hydrolyzed fTHP-9 with a k cat /K M value 30 times lower than MMP-8 (Table 5). K M ϭ 33.0 M and k cat ϭ 0.20 s Ϫ1 for MMP-8(Y189F), whereas wild type MMP-8 exhibited K M and k cat values of 5.40 M and 0.94 s Ϫ1 , respectively. Thus, the Tyr 3 Phe mutation was detrimental to both binding and turnover. C 10 -fTHP-9 and C 16 -fTHP-9 were hydrolyzed 3 and 6.5 times more slowly than fTHP-9. This reduction in activity is comparable with that seen with wild type MMP-8, and thus the mutation did not result in a modified ability for MMP-8 to process more thermally stable substrates. As a control, collagen hydrolysis for MMP-8 and MMP-8(Y189F) was compared and found to differ by 3-fold (Table 5), as reported previously (28).
To evaluate triple-helical peptidase activity in light of collagenolysis, kinetic parameters were determined for MMP-8 and MMP-13 hydrolysis of fluorescein-labeled bovine type I collagen (Table 5) (77,78), comparable with the results reported here. In turn, MMP-13 was found to hydrolyze type I collagen 3.4-fold slower than MMP-8 (79), similar to the 2.3-fold slower rate seen here. Some differences in activities are expected, as prior studies utilized radiolabeled type I collagen, whereas this study utilized fluorescein-labeled collagen. k cat /K M values for MMP-8 and MMP-13 collagenolysis are, depending on sequence, within the same relative range as for fTHPs. However, the individual K M and k cat values are lower for collagenolysis.

DISCUSSION
Collagenolysis is a complex, multistep process. THP substrates are one model system that has been utilized to better understand collagenolysis. The use of THP substrates has allowed for the identification of distinct aspects of collagenoly-sis. For example, deletion of the C-terminal hemopexin-like domain from MMP-1, MMP-8, MMP-13, and MT1-MMP results in virtually a complete loss of collagenolytic activity (12, 40, 80 -83), whereas triple-helical peptidase activity, although reduced, is retained (Table 1) (23,26,31,40,(72)(73)(74). Thus, the hemopexin-like domain is critical for the orientation and destabilization of a large macromolecule such as collagen, but it is not essential for unwinding and hydrolysis of a small triplehelical element. In turn, the behavior of other MMPs is identical for THP substrates and collagen. Some sequences are readily cleaved by MMPs when found in triple-helical conformation (for both THP substrates and collagen) but only slowly hydrolyzed or not cleaved in a single-stranded form (49). The ratio of activation energies for MMP hydrolysis of THP substrates to analogous single-stranded substrates is virtually identical to that of collagen to gelatin (31). The efficiency of triple helix hydrolysis is dependent upon substrate thermal stability for both THPs and collagen (26). Thus, some aspects of collagenolysis are well modeled by THPs, whereas others may be distinct. In this study, the origins of MMP collagen specificity have been examined using THP substrates.
As described earlier, collagenolysis requires an initial binding event whereby the triple helix interacts with MMP exosites, followed by unwinding of the triple helix (17). A second binding event is the interaction of an unwound strand with the MMP active site, followed by hydrolysis. The THPs used in this study were designed to explore the capabilities of MMPs to process differentially thermal substrates and MMP sequence specificities. It can be presumed that certain regions of the substrate (the P 2 through P 2 Ј sites) do not bind to exosites but rather only to the enzyme active site. Thus, the substitutions in the P 2 and P 1 Ј subsites herein would be anticipated to affect binding to the active site. Computational analyses using molecular interaction fields indicated potential discrimination between MMPs based on their S 2 and S 1 Ј specificity pockets (51,52), whereas the P 2 and P 1 Ј subsites of types I-III collagen show subtle variations (Fig. 3).
Initial screening of triple-helical peptidase activity exhibited trends consistent with collagenolytic activity, in that collagenolytic MMPs (MMP-1, MMP-8, MMP-13, MT1-MMP, and MT2-MMP) cleaved the type I-III collagen consensus sequence THP, whereas noncollagenolytic MMPs (MMP-3, MT5-MMP, and MT6-MMP) displayed no activity toward this substrate ( Figs. 2A and 4). Activity was also seen for both gelatinases (MMP-2 and MMP-9). MMP-2 had been reported pre-  (64,65), whereas MMP-9 has been shown to possess triple-helical peptidase activity (49) and types V and XI collagenolytic activity (6). We have observed that some noncollagenolytic MMPs possess triple-helical peptidase activity (26,73). This difference has been attributed to an inability of these MMPs to properly orient and/or destabilize large macromolecules such as collagen. In the case of MMP-9, it appears that specific features of types I-III collagen prevent the enzyme from catabolizing it, as MMP-9 is active against other collagen types.
MT1-MMP has favorable interactions with hydrophobic groups because of Met at the bottom of the pocket (51). MMP-2, MMP-3, MMP-9, and MMP-13 have open pockets and thus possess no interaction tendencies (51). Overall, the kinetic changes observed here for Cys(Mob) substitution of Leu (Fig.  4) could be anticipated based upon previously observed MMP S 1 Ј pocket features. For example, hydrolysis became slower for MMP-1, which has a relatively shallow S 1 Ј pocket, but better for MMP-8, MMP-13, MT1-MMP(⌬279 -523) ( Table 1 versus Table 2), and MT2-MMP(⌬268 -628) (Fig. 2, A versus B), which have deep S 1 Ј pockets. Only MMP-2 and MMP-9(⌬444 -707) show anomalous behaviors, as Cys(Mob) in P 1 Ј reduced activity for both enzymes, although their S 1 Ј pockets are deep (38). However, prior single-stranded peptide studies had shown that some aromatic residues, such as Trp or Tyr, may be greatly detrimental for substrate hydrolysis by MMP-2 and MMP-9, while being only slightly disfavored or favored by MMP-8 (88). It is important to note that those studies did not consider the conformational constraints of triple-helical substrates. Nonetheless, there are some yet to be described S 1 Ј pocket features that discriminate between the gelatinases (MMP-2 and MMP-9) and some collagenases (i.e. MMP-8). These features could include fluctuations in the binding pocket, as suggested by molecular dynamics simulations of MMP-2 (86), or the effects of the nearby fibronectin type II repeats.
The THP sequence specificity of MMPs was further examined by substitution of a charged residue in P 2 (Gln 3 Orn). The S 2 subsite is shallow in most MMPs, with the size and polarity partially dependent upon the residue in the position equivalent to MMP-1 227 (76). MMP-2, MT1-MMP, and MT2-MMP should favorably interact with positively charged residues because of Glu at this position (51). However, the shape of the pocket is also determined by the presence of Pro in position 104, which is found for MMP-2, MMP-8, and MMP-13. MMP-2 and MMP-13 have Phe 103 , resulting in a small, hydrophobic pocket (51). MMP-8 has Thr 103 , creating a comparatively large pocket. When position 104 is not Pro (MMP-1, MMP-9, and MT1-MMP), the S 2 pocket is mainly defined by position 186 (51). Pro 186 in MMP-9 leads to a relatively large hydrophobic pocket; Gln 186 in MMP-1 and MMP-8 reduces the size and hydrophobicity, and Phe 186 in MT1-MMP and MT2-MMP leads to a small hydrophobic pocket. The substitution of Orn for Gln in P 2 is disfavored by MMP-1, MMP-2, MMP-5 MMP numbering is based upon prepro-MMP-1 and MMP alignment (76).  9(⌬444 -707), and MT1-MMP(⌬279 -523), while being favored by MMP-13 (Fig. 4). For MMP-1 and MMP-9, Orn was shown previously to be disfavored compared with Gln in P 2 for linear, collagen-model sequences (89). Our results are somewhat surprising for MMP-2 and MT1-MMP(⌬279 -523), although Orn may have too long a side chain to favorably interact with the Glu 227 residue in S 2 , or Phe 103 (in MMP-2) and Phe 186 (in MT1-MMP) render the pocket too hydrophobic. Why the Orn substitution was favored in MMP-13 is unknown, particularly because Phe 103 is found in the S 2 subsite.
Analysis of the double substitution of Orn for Gln in the P 2 subsite and Cys(Mob) for Leu in the P 1 Ј subsite (fTHP-11) indicated nonadditive effects on MMP activity. For example, the activity of MMP-8 was increased 39-fold by the substitution of Cys(Mob) for Leu in the P 1 Ј subsite and remained approximately the same when Orn was substituted for Gln in the P 2 subsite. Therefore, the combined effect of both substitutions should have been an ϳ39-fold increase in activity. However, only a 4.3-fold increase in activity was observed in the substrate with the double substitution. In a similar fashion, the combined effect of both substitutions should have increased MMP-13 activity 192-fold, but only an 8.3-fold increase was observed. Only for MT1-MMP(⌬279 -523) were the theoretical rate increase (1.3-fold) and the actual rate increase (1.2-fold) similar. These results suggest the following: (a) the P 2 and P 1 Ј subsites could have a role in initial exosite binding of the substrate and/or (b) MMP subsite interactions are nonindependent. Nonindependence of the interaction of substrate subsites within MMPs was observed previously using single-stranded substrates (29,89,90). The relative "mobility" of the S 1 Ј pocket (87) may play a role in nonindependence of MMP subsites. However, as alluded to earlier, prior studies defining nonindependence did not consider the conformational constraints of triple-helical substrates.
The effects of substrate stability on MMP triple-helical peptidase activity were evaluated by using substrates of identical sequence but varying their thermal stabilities. Increased thermal stability typically results in a decrease in enzymatic activity, but there is also some sequence dependence. In many cases, greater thermal stability results in lower K M values for MMP-13 and MT1-MMP(⌬279 -523). Sequence was found to affect the capability to process more thermally stable substrates. For example, MT1-MMP(⌬279 -523) can hydrolyze the more thermally stable versions of sequences containing either the Orn for Gln substitution in the P 2 subsite or Cys(Mob) for Leu substitution in the P 1 Ј subsite, but not the most thermally stable sequence containing both substitutions. For the most part, MMP-13 hydrolysis appears to be the least affected by an increase in substrate thermal stability, in combination with sequence changes. Gelatinases (MMP-2 and MMP-9(⌬444 -707)) appear incapable of processing more stable helices and are mechanistically distinct from other collagenolytic MMPs, as discussed previously (24,50,86).
MMPs that are capable of processing more thermally stable substrates also favor Cys(Mob) in the P 1 Ј subsite compared with Leu (Fig. 4). One may speculate that a correlation exists between the potential for an MMP to accommodate hydropho-bic interactions deep in the S 1 Ј pocket and to hydrolyze more thermally stable triple-helical structures. It is possible that more thermally stable substrates may be partially destabilized upon binding to the MMP, but subsequent interactions within the active site are required to properly present individual strands for hydrolysis. Additional studies are required to determine whether a deep S 1 Ј pocket is necessary for the cleavage of more thermally stable triple helices.
The collagen specificity of MMP family members may be based on any combination of factors. It was reported previously that the collagen binding modes of MMPs differ, with MT1-MMP distinct from MMP-1 and MMP-8 (24). However, this does not explain collagen specificity, in which MMP-1 is distinct from MMP-8 and MT1-MMP (11,12). The sequence specificities of MMPs also do not explain their collagen specificities (13,88). Our prior study demonstrated that MT1-MMP(⌬279 -523) was more efficient at processing more thermally stable substrates than MMP-1, suggesting that part of the collagen specificity for these two enzymes was based on their different abilities to process triple helices of differing stabilities (26). In this study, this concept has been advanced, as MMP-13 was found to be distinct from MMP-8 and MT1-MMP(⌬279 -523), in that enhanced thermal stability has only a modest effect on activity, regardless of sequence. This may be related as to why MMP-13 has a different collagen specificity than MMP-8 and MT1-MMP, because MMP-13 hydrolyzes type II collagen efficiently and MMP-8 and MT1-MMP are similar in their preference for type I collagen. The ability of MMP-8 to hydrolyze more thermally stable sequences is affected by the P 2 subsite, whereas MT1-MMP(⌬279 -523) is affected by substitutions in both the P 2 and P 1 Ј subsites. Although similar, the MMP cleavage sites in types I-III collagen are not identical (Fig. 3). Differences are found in the P 2 and P 1 Ј subsites, such as Gln versus Leu in P 2 and Ile versus Leu in P 1 Ј. The collagen specificity of MMPs appears to be based on a combination of sequence and thermal stability (Fig. 4).
Triple-helical peptidase activities do not provide insight as to which regions of the MMPs are responsible for the collagen specificity. Such information may be obtained by analyzing THP substrate hydrolysis by MMP mutants. Results with MMP-8(Y189F) demonstrate that the Tyr 210 hydroxyl group is critical for both triple-helical peptidase and collagenolytic activities. Tyr 210 functions in triple helix binding and hydrolysis but not in processing more thermally stable substrates. The putative role of Tyr 210 is either to bind to the hemopexin-like domain or to form a hydrogen bond with the substrate (25,28).
Based on the triple-helical peptidase activities of collagenolytic MMPs, collagen selectivity is not because of the initial binding of collagen by the MMP. This conclusion is consistent with earlier studies reporting very similar K M values for MMP-1 and MMP-8 hydrolysis of types I-III collagen (78,91). Future studies will be directed at assigning regions of MMPs responsible for the discrete steps of collagenolysis. One of the long term goals of understanding collagenolytic MMP mechanisms is to use this information (as presented in Fig. 4) for the design of selective MMP inhibitors. Initial clinical trials with MMP inhibitors were disappointing, with one of the problematic features being a lack of selectivity (92)(93)(94)(95)(96). At present, the only approved MMP inhibitor is Periostat TM (97). Further exploration of MMP active sites and exosites, in combination with substrate conformation, may prove valuable for additional dissection of collagenolysis and yield information useful in the design of more selective MMP inhibitors.