Elastin Degradation by Matrix Metalloproteinases

Insoluble elastin was used as a substrate to characterize the peptide bond specificities of human (HME) and mouse macrophage elastase (MME) and to compare these enzymes with other mammalian metalloproteinases and serine elastases. New amino termini detected by protein sequence analysis in insoluble elastin following proteolytic digestion reveal the P′1 residues in the carboxyl-terminal direction from the scissile bond. The relative proportion of each amino acid in this position reflects the proteolytic preference of the elastolytic enzyme. The predominant amino acids detected by protein sequence analysis following cleavage of insoluble elastin with HME, MME, and 92-kDa gelatinase were Leu, Ile, Ala, Gly, and Val. HME and MME were similar in their substrate specificity and showed a stronger preference for Leu/Ile than did the 92-kDa enzyme. Fibroblast collagenase showed no activity toward elastin. The amino acid residues detected in insoluble elastin following hydrolysis with porcine pancreatic elastase and human neutrophil elastase were predominantly Gly and Ala, with lesser amounts of Val, Phe, Ile, and Leu. There were interesting specificity differences between the two enzymes, however. For both the serine and matrix metalloproteinases, catalysis of peptide bond cleavage in insoluble elastin was characterized by temperature effects and water requirements typical of common enzyme-catalyzed reactions, even those involving soluble substrates. In contrast to what has been observed for collagen, the energy requirements for elastolysis were not extraordinary, consistent with cleavage sites in elastin being readily accessible to enzymatic attack.

Insoluble elastin was used as a substrate to characterize the peptide bond specificities of human (HME) and mouse macrophage elastase (MME) and to compare these enzymes with other mammalian metalloproteinases and serine elastases. New amino termini detected by protein sequence analysis in insoluble elastin following proteolytic digestion reveal the P 1 residues in the carboxyl-terminal direction from the scissile bond. The relative proportion of each amino acid in this position reflects the proteolytic preference of the elastolytic enzyme. The predominant amino acids detected by protein sequence analysis following cleavage of insoluble elastin with HME, MME, and 92-kDa gelatinase were Leu, Ile, Ala, Gly, and Val. HME and MME were similar in their substrate specificity and showed a stronger preference for Leu/Ile than did the 92-kDa enzyme. Fibroblast collagenase showed no activity toward elastin. The amino acid residues detected in insoluble elastin following hydrolysis with porcine pancreatic elastase and human neutrophil elastase were predominantly Gly and Ala, with lesser amounts of Val, Phe, Ile, and Leu. There were interesting specificity differences between the two enzymes, however. For both the serine and matrix metalloproteinases, catalysis of peptide bond cleavage in insoluble elastin was characterized by temperature effects and water requirements typical of common enzyme-catalyzed reactions, even those involving soluble substrates. In contrast to what has been observed for collagen, the energy requirements for elastolysis were not extraordinary, consistent with cleavage sites in elastin being readily accessible to enzymatic attack.
Elastin is the extracellular matrix protein that imparts elastic recoil to tissues. Its cross-linked nature and extreme hydrophobicity make it one of the most stable proteins in the body (1)(2)(3). A contributing factor to elastin's longevity is its relative resistance to proteolysis by all but a limited number of proteinases that are capable of degrading the mature, insoluble protein under physiological conditions. These enzymes, referred to as elastases, have a wide distribution in nature and are found in animals as well as in plants and bacteria (4,5). Elastases are heterogeneous with differing substrate specificities and catalytic mechanisms. In fact, enzymes with elastolytic activity can be found in most of the major proteolytic families, including serine, thiol, aspartic, and metallo enzymes (4). Despite the differences in catalytic mechanisms, all of these elastases share a common specificity for cleaving peptide bonds associated with hydrophobic or aromatic amino acids.
The most widely studied elastases, human neutrophil elastase (HLE) 1 and pancreatic elastase (PPE), belong to the serine proteinase family of enzymes. Neutrophil elastase is found in the azurophil granules of polymorphonuclear leukocytes and is essential for phagocytosis and defense against infection. Pancreatic elastase is stored as an inactive zymogen in the pancreas and is secreted into the intestine where it is activated by other digestive enzymes. Both enzymes cleave peptide bonds on the carboxyl-terminal side of amino acids with a small alkyl side chain, although HLE has a preference for amino acids with longer aliphatic chains at this position (6).
Bacterial elastases belonging to the family of metalloproteinases have been known for many years (5). These proteinases require Zn 2ϩ atoms for activity and in contrast to the serine proteinases cleave peptide bonds on the amino-terminal side of the amino acid that determines specificity. Elastases from Bacillus thermoproteolyticus (thermolysin) and Streptomyces fradiae are four and eight times more active than PPE, respectively, making them some of the most potent elastolytic proteinases reported (5). Recently, metalloproteinases secreted by mammalian cells have been shown to have elastolytic activity (7)(8)(9). Like their bacterial counterparts, these proteinases contain Zn 2ϩ atoms and express primary substrate specificity through the S 1 Ј-P 1 Ј interaction. 2 Members of the matrix metalloproteinase (MMP) family can be grouped into six categories based on similarities in their domain organization and protein substrate specificities. These groups include 1) the collagenases, 2) the gelatinases or "type IV collagenases," which include two distinct enzymes of 72 kDa (MMP 2) and 92 kDa (MMP 9), 3) the stromelysins, 4) matrilysin, 5) metalloelastase, and 6) the mT-MMPs containing a membrane-spanning segment in their hemopexin-like domains conferring cell surface localization. Of the known MMPs, the 92-and 72-kDa gelatinases, mouse and human macrophage metalloelastases (HME and MME), and matrilysin degrade insoluble elastin. Matrilysin and the autoprocessed macrophage elastases possess only a catalytic domain, which must also convey binding and catalytic specificity for the elastin substrate. Binding specificity of 92-and 72-kDa gelatinases for elastin, however, resides in the three fibronectin type II-like repeats inserted in tandem in their catalytic domains (10).
To learn more about the mechanism of elastin degradation by MMPs, we developed a novel assay to detect protease cleavage sites in insoluble elastin. Traditionally, elastolytic enzymes have been detected and characterized through their ability to release soluble peptides from insoluble elastin or to cleave synthetic peptides of known sequence. Several assay methods focusing on the solubilized peptides have been described (4,11), but these lack the ability to provide detailed information about the specificity of peptide bond hydrolysis. By detecting cleavage sites directly in the insoluble protein, we have been able to determine peptide bond and subsite preferences of the elastolytic enzymes and detect differences in specificity that would not be evident using assay techniques involving synthetic substrates or solubilized forms of the protein. Our results demonstrate that HME and MME have an overall preference for amino acids with large aliphatic side chains in the P 1 Ј position. There are also interesting differences between these enzymes and the 92-kDa gelatinase and serine elastases.

MATERIALS AND METHODS
Preparation of Elastin and Blocking with DNFB-Insoluble elastin was purified from bovine ligamentum nuchae using the hot alkali technique of Lansing et al. (12) and was shown by amino acid analysis to be free of microfibrillar protein and other contaminants. Radiolabeled elastin was prepared using sodium [ 3 H]borohydride (13). To block amino groups generated by random peptide bond cleavage during purification, insoluble elastin was washed twice using microcentrifugation with 50 mM sodium phosphate buffer, pH 8.1, containing 0.02% Brij. The pellet was then resuspended in 5 ml of phosphate buffer, and 400 l of DNFB was added. After vortexing, the tube was covered with foil and left on a rotator table for 1.5 h at room temperature. Insoluble elastin in the sample was pelleted by centrifugation, the supernatant was discarded, and the pellet was incubated with a second 400-l aliquot of DNFB as above. Elastin was removed from the reaction mixture by centrifugation and washed three times with acetone and three times with HEPES buffer, pH 8.0, containing 0.02% Brij. It was then washed twice with water, once with 100% trifluoroacetic acid, and twice with acetonitrile and then dried in 25-g aliquots.
Preparation of Proteolytic Enzymes-Matrilysin was expressed in baculovirus, purified over SP-Sepharose and activated with 1 mM 4-aminophenylmercuric acetate at 37°C, and attainment of full catalytic activity was verified by thiopeptolide assay as described previously (10). 92-kDa gelatinase was obtained using HT-1080 cells co-transfected with E1A to suppress endogenous MMP production and 92-kDa gelatinase cDNA (14). 92-kDa gelatinase free of associated tissue inhibitor of metalloproteinase was purified by gelatin-agarose chromatography; the enzyme was activated with small amounts of stromelysin, and full catalytic activity was verified by thiopeptolide assay (15). MME and HME were expressed in Escherichia coli and purified as described (16,17).
Enzyme Kinetics against Elastin-Values for activation energy (E A ) and deuterium solvent kinetic isotope effect (k H /k D ) for matrilysin, the 92-kDa gelatinase, and neutrophil elastase were determined with 3 H-labeled insoluble elastin as a substrate (18). For determination of E A , initial velocities were calculated at the various temperatures indicated. Reactions were measured over at least five time points for each temperature studied. Reaction velocity was linear for at least the initial three time points. For determination of k H /k D , all reactions were performed in buffer containing a final concentration of 100% H 2 O versus 90% D 2 O. Again, initial velocities were calculated.
Conditions for Enzyme Reactions-All enzyme reactions were conducted in 0.05 M Tris buffer, pH 7.5, containing 10 mM CaCl 2 at 37°C. Catalysis was terminated by the addition of a final concentration of 1 mM o-phenanthroline.
Sequencing Reactions-Sequencing of insoluble, digested elastin was performed using an Applied Biosystems model 473A protein sequencer equipped with a model 610A data analysis program. Digested insoluble elastin was pelleted by centrifugation, the supernatant was discarded, and the pellet washed twice with water and twice with acetonitrile. The pellet was then suspended and dried onto a bioprene-treated glass fiber filter and sequenced using the normal liquid phase sequencing cycles.
Nearest Neighbor Analysis-A nearest neighbor table was constructed by examining the sequence of bovine tropoelastin (19) as dipeptides (e.g. residues 1 and 2, residues 2 and 3, residues 3 and 4, etc.). Each entry in the table lists the number of occurrences of each dipeptide pair and is predictive for n Ϫ 1 and n ϩ 1 relationships.

Energy Requirements and Deuterium Effects Characterizing Elastin Degradation by Metalloproteinases versus Neutrophil
Elastase-To gain insights into biochemical mechanisms of elastolysis by metalloproteinases, we compared several catalytic parameters with those exhibited by the well characterized serine proteinase, neutrophil elastase. Insoluble elastin is hydrophobic, extensively cross-linked, and highly resistant to proteolytic degradation, properties similar to another extracellular matrix molecule, fibrillar collagen. Collagenases are responsible for the degradation of fibrillar collagen, and we have previously studied the kinetics of this biologic process for collagenase-1 (18,20). As summarized in Table I, the degradation of type I collagen fibrils is temperature-dependent, with an activation energy of ϳ100,000 cal/mol (18). However, the energy requirements for collagen degradation are reduced by a factor of two if the native collagen is in solution form (monomeric) and by a factor of ten if the collagen is denatured (gelatin). Likewise, the deuterium isotope effect, a measure of the dependence of catalysis upon proton transfer, also varies strongly with the state of substrate organization, with k H /k D values ranging from 9.0 for fibrillar collagen to 1.0 for denatured gelatin ␣ chains.
Due to the highly insoluble nature of elastin, we were interested in whether the degradation of elastin by matrix metalloproteinases showed marked dependence upon energy and water transfer like that found with type I collagen. Arrhenius plots describing the degradation of insoluble elastin by neutrophil elastase, the 92-kDa gelatinase, and matrilysin are shown in Fig. 1. All three enzymes showed a linear dependence on temperature, with activation energies in the range of 8,412-  (Table I), values again typical for most enzyme-catalyzed reactions. Therefore, although elastin is highly insoluble and extensively cross-linked, we found that the activation energies involved in elastolysis by matrilysin and 92-kDa gelatinase were much lower than the energies previously observed for degradation of triple helical collagen by interstitial collagenase-1. Similar results regarding elastin degradation were found with neutrophil elastase, a serine proteinase. Deuterium isotope effects were also minimal for elastin degradation compared with those of collagen. It is apparent that the catalysis of insoluble elastin is characterized by temperature effects and water requirements typical of most enzyme-catalyzed reactions, even those involving soluble substrates. The degradation of elastin therefore differs from the degradation of fibrillar collagen by interstitial collagenase, which involves enormous dependence upon energy and proton transfer. In this regard, we previously hypothesized that the extremely high values of E A and k H /k D in collagenolysis reflect the extent to which triple helical structure and the aggregation of collagen molecules into the fibril presents a barrier to the transport of the water required for peptide bond hydrolysis. Further implicating such a scenario is the observation that the Gly 725 -Ile 726 and Gly 725 -Leu 726 peptide bonds cleaved in the ␣1 and ␣2 chains of native type I collagen are oriented directly into the center of the molecule's hydrophobic triple helical core. Taken as a whole, our data suggest that peptide bonds cleaved in insoluble elastin by both metalloproteinases and serine proteinases are more accessible to these enzymes, with water transport much less of a barrier than for collagen degradation by interstitial collagenase. These results also demonstrate that the energy requirements for elastolysis are not extraordinary, consistent with the idea that the cleavage sites in elastin are readily accessible to enzymatic attack.
Despite the seemingly dichotomous thermodynamics of collagen versus elastin degradation, our assays may not measure precisely equivalent catalytic events. Solubilization of highly cross-linked elastin requires at least two proteolytic cleavages, except when peptides are liberated from the amino or carboxyl terminus of tropoelastin chains. The first cleavage would free a polypeptide chain and the second, solubilizing cleavage would release the peptide and thus occur on a less constrained substrate. In contrast, collagen is solubilized by a single 3 ⁄4: 1 ⁄4 cleavage through all three constituent ␣ polypeptide chains. Our data do not eliminate the possibility that the initial cleavages in elastin (occurring before any solubilized peptides are released), especially in the vicinity of the desmosine and isodesmosine bridges on cross-linked tropoelastin molecules, would show a large activation energy and marked deuterium isotope effect.
Identification of Proteinase Cleavage Sites in Elastin Using Protein Sequence Analysis-Previous approaches to characterizing the peptide bond specificity of elastolytic proteinases involved the identification of cleavage sites in soluble proteins or oligopeptides of known sequence. Although much has been learned about the catalytic properties of elastolytic enzymes using this approach, the sites of peptide bond hydrolysis in insoluble elastin may not be predicted from these types of studies because of differences in available sequences on the insoluble protein recognized by the proteinase's catalytic domain or influences of remote site contacts between enzyme and substrate. To more precisely determine proteinase cleavage sites in elastin, we used protein sequencing to identify new amino termini generated by proteinase-induced peptide bond cleavage in the insoluble protein as shown schematically in Fig.  2. The amino termini in the insoluble residue exposed as the result of proteolytic digestion represent the P 1 Ј residues in the carboxyl-terminal direction from the scissile bond. 2 The relative proportion of each amino acid in this position reflects the proteolytic preference of the elastolytic enzyme. In some instances, comparison of results from a few rounds of sequencing with the known primary sequence of elastin can indicate the precise peptide bonds that were cleaved by the enzyme.
The insolubility of elastin makes it ideally suited for sequence analysis using gas phase or pulse liquid sequencers. In contrast to soluble proteins that gradually wash off the immobilizing membrane during the sequencing reaction, small particles of elastin can be placed on the membrane and sequenced without losses due to washout. Because elastin is a cross-linked polymer of tropoelastin molecules, sequence analysis of the insoluble protein should theoretically show one sequence, GGVP . . . , beginning at the amino terminus of each component monomeric chain. However, several amino acids (Gly, Ala, Val, Pro, and Leu being the most prevalent) are evident in this first sequencing cycle. This was not unexpected because the elastin used in this study was purified by hot NaOH. Hot NaOH treatment produces random cleavage of a small number of peptide bonds throughout the polymer. As a result, each hydrolyzed peptide bond produces a new amino terminus for sequencing.
Before we could use this elastin to detect new sequencing FIG. 1. Arrhenius plots of elastin degradation by neutrophil elastase, 92-kDa gelatinase, and full-length matrilysin. The degradation of insoluble [ 3 H]elastin was measured over the indicated temperature ranges as described under "Materials and Methods." Initial velocity was calculated for each temperature. The relationship of velocity and temperature (K) was plotted according to Arrhenius. sites produced by proteinase cleavage, it was necessary to eliminate the background signal produced during the purification procedure. This was accomplished by incubating purified insoluble elastin with DNFB. DNFB reacts covalently with both primary and secondary amino groups (21) and prevents their subsequent reaction with the sequencing reagent phenylisothiocyanate. Treatment of purified elastin with DNFB effectively blocked all free amino groups such that no significant signal was detected through multiple sequencing cycles.
Time Course of Digestion-To determine the time course for elastolysis, radiolabeled counts released into the supernatant were monitored during the reaction, and the insoluble pellet was taken for sequencing when approximately 1, 2.5, 5, 10, and 50% of the total counts were released. The rationale for selecting these data points was that early digestion times should reflect the primary and preferential cleavage sites in the molecule. Gly, Ala, Val, Ile, and Leu were the predominant amino acids detected at each time point, and there was little change in the relative ratio of these amino acids in the first cycle of sequencing up to 10% hydrolysis. Only at 50% hydrolysis were significant differences noted (not shown).
Cleavage Sites in Elastin for Different Proteinases-Figs. 3 and 4 show the amino-terminal amino acids found in insoluble elastin following incubation with two serine elastases (PPE and HLE) and three metalloproteinases (MME, HME, and 92-kDa gelatinase). In all cases, the digestion reaction was allowed to proceed until 10% of the total radiolabeled counts were solubilized from the insoluble elastin substrate. The data are presented as a molar ratio normalized to the number of leucine residues detected in each sequencing reaction. To interpret the substrate specificity of the proteinases, it is important to understand that the amino acid residue being detected in the first cycle of amino-terminal sequencing is the amino acid on the carboxyl-terminal side of the scissile bond or the P 1 Ј amino acid. The second round of sequencing would contain amino acids in the P 2 Ј position, the third cycle would contain the P 3 Ј amino acids, and so forth.
Although it is not possible to directly determine the P 1 amino acid using our sequencing approach, it is possible to make predictions using nearest neighbor analysis. Fig. 5 shows nearest neighbor relationships for bovine tropoelastin indicating the frequency with which amino acid pairs occur in the molecule. Because of the unusual sequence properties of elastin, if the P 1 Ј amino acid is known, then it is possible to use this table to predict the amino acids most likely to be in the P 1 position.
Porcine Pancreatic and Human Neutrophil Elastase-The amino acid residues detected in insoluble elastin following hydrolysis with PPE and HLE are shown in Fig. 3. In both cases, Gly and Ala were the predominant amino acids with lesser amounts of Val, Phe, Ile, and Leu. There were some interesting differences between the two enzymes, however. HLE digestion resulted in elevated levels of Gly over Ala, whereas approximately equivalent levels of these amino acids were present in elastin degraded by PPE. Valine levels were higher in elastin digested with PPE compared with that digested with HLE.
Past studies have shown that the primary substrate specificity of PPE and HLE mostly resides with the P 1 residue (4). Thus, the P 1 Ј amino acids determined in our study do not directly indicate the proteolytic preferences for peptide bond cleavage in elastin by these two enzymes. Nearest neighbor analysis, however, indicates that approximately 50% of the Gly residues in elastin are preceded by another Gly, Ala, or Val residue. Likewise, Ala is preceded by Gly 27% of the time and by another Ala 45% of the time. These findings predict that the bonds being cleaved by the elastases are predominantly (Gly/ Ala/Val)-Gly and (Gly/Ala)-Ala sequences, in agreement with the known specificity of these proteinases for preferential cleavage of the nonbulky amino acids glycine and alanine in the P 1 position.
Proline residues, which comprise almost 15% of the total amino acid residues in elastin, were remarkably absent in the first two rounds of sequencing of cleavage sites generated by both PPE and HLE. Exclusion of proline at the P 1 Ј subsite is predicted from the three-dimensional structure of PPE (22) and has been confirmed experimentally in previous studies (4). Proline can fit into the S 2 subsite of both enzymes, however, so FIG. 2. Strategy for characterizing proteinase cleavage sites in insoluble elastin. Free amino groups in insoluble elastin resulting from random peptide bond cleavage occurring during purification with hot alkali were blocked using DNFB. DNFB reacts covalently with both primary and secondary amino groups and prevents their subsequent reaction with the sequencing reagent phenylisothiocyanate. The blocked elastin was then incubated with a proteinase, and the new amino termini generated by proteinase-induced peptide bond cleavage were determined by protein sequence analysis. The amino termini exposed as the result of proteolytic digestion represent the P 1 Ј residues in the carboxyl-terminal direction from the scissile bond.

FIG. 3. Amino-terminal amino acids found in insoluble elastin following incubation with PPE or HLE.
Insoluble elastin was sequenced after 10% of the total radiolabeled counts had been released into the medium. Shown is the molar ratio of each amino acid at step 1 of the sequencing reaction normalized so that leucine ϭ 1.
its absence in the P 2 Ј position of cleavage sites in elastin cannot be due to steric factors. Nearest neighbor analysis shows that proline residues are seldom found after glycine or alanine residues, the most prevalent amino acids found at P 1 Ј. Thus the absence of proline in P 2 Ј is simply a statistical circumstance resulting from its sequence distribution in the elastin molecule.
Elastolytic Metalloproteinases-The predominant amino acids found in the P 1 Ј position following cleavage of insoluble elastin by the 92-kDa gelatinase, HME, and MME were Ala, Gly, Val, Leu, and Ile (Fig. 4). HME and MME were similar in their substrate specificity and showed a stronger preference for Leu/Ile in the P 1 Ј position than did the 92-kDa enzyme. In contrast, amino acids found in the first sequencing cycle for the 92-kDa gelatinase somewhat resembled that of serine elastases with high levels of Ala and Val.
Our findings are essentially in agreement with the known specificity of MMPs for hydrophobic, aliphatic residues in subsite P 1 Ј (23)(24)(25). There are, however, some interesting differences between the 92-kDa gelatinase and the two macrophage elastases that warrant comment. HME and MME have a strong preference for Leu or Ile in the P 1 Ј position. In contrast, whereas Leu/Ile remained favored in that position for the 92-kDa enzyme, Ala, Val, and Phe were also found in significant quantities. Using soluble peptides with substitutions covering the P 4 through P 4 Ј subsites, Netzel-Arnett et al. (24) have shown that the 92-kDa enzyme will hydrolyze peptide bonds with Val and Phe in the P 1 Ј position, although with one-third the activity found with P 1 ЈϭLeu (Ala in the P 1 Ј position was not tested). The 92-kDa gelatinase has also been shown to cleave the interglobular domain of aggrecan at the bond between Asn and Phe (26), providing further support for cleavage specificity directed by P 1 Ј Phe.
The known P 1 subsite specificities for the gelatinases are also in agreement with the cleavage sites we found in elastin. Previous studies have shown that the 92-kDa gelatinase tolerates only small amino acids in this position, with Ala being slightly preferred over Gly (24). Consistent with this restriction, nearest neighbor relationships in elastin shown in Fig. 5 confirm that Leu and Ile are preceded almost 75% of the time by a Gly or Ala residue.
In addition to cleavage at Leu/Ile, sequence data shown in Fig. 4 document cleavage by the 92-kDa enzyme at Gly, Ala, Val, and Phe residues in elastin. Analysis of the elastin sequence shows that many of these amino acids are preceded by the consensus sequence Pro-Xaa-(Gly/Ala) and hence are possible cleavage sites. Interestingly, most of these putative cleavage sites occur in the amino and carboxyl regions of elastin. There are, in addition, potential cleavage sites containing Pro-Xaa-Ala-Ala sequences at the beginning of all but two of the FIG. 4. Amino-terminal amino acids found in insoluble elastin following incubation with 92-kDa gelatinase, MME, and HME. Insoluble elastin was sequenced after 10% of the total radiolabeled counts had been released into the medium. Shown is the molar ratio of each amino acid at step 1 of the sequencing reaction normalized so that leucine ϭ 1. alanine-rich cross-linking domains. Cleavage at these sites would explain the observed liberation of cross-link-containing peptides from the insoluble protein and would be consistent with the preponderance of alanine residues found in most steps of the sequence analysis (not shown).
Because of the multiple peptide chains that are being sequenced in the insoluble protein, it was not feasible in most cases to assign sequence to verify cleavage at any of the predicted sites. It was possible, however, to document one unique sequence based upon the occurrence of a Gln residue at step two and again at step three of the sequence obtained following cleavage with 92-kDa gelatinase. A single Gln-Gln pair occurs in exon 12 of bovine elastin and follows a potential Gly residue that is at the P 1 Ј position of a predicted cleavage site. Hydrolysis at this site is consistent with the sequence we obtained. HME and MME-HME and MME exhibit a stronger preference for Leu in the P 1 Ј subsite than is shown by the 92-kDa gelatinase. In this respect, the specificity of both metalloelastases resembles the specificity of matrilysin (24,25) and the bacterial enzyme thermolysin (5). We were unable to identify unique cleavage sites in elastin, but our results are in general agreement with the known preference of macrophage elastase for leucine residues in the P 1 Ј position (27,28). Our findings also suggest that Gly, Ala, and Val residues can be accommodated in the P 1 Ј site, with HME showing a stronger preference for Ala compared with MME.
Although sequence data that might define the subsite specificity of HME and MME are limited, examination of sequences amino-terminal to the scissile bonds in insulin and ␣1-proteinase inhibitor showed that the aromatic amino acids Tyr and Phe occupied the P 1 position in two of the four peptides with the hydrophobic amino acids Pro and Ala found in the other two peptides (27,28). The P 2 position was variable in composition, whereas three of the four peptides had Ala at P 3 . The fourth peptide had a Val residue in the P 3 position. These data suggest that aromatic or hydrophobic amino acids are preferred at the P 1 site, with small hydrophobic residues (preferably alanine) occupying P 3 . An examination of the bovine tropoelastin sequence shows that there are many sites that meet this sequence consensus, several of which immediately flank the cross-linking domains, which would explain the ability of these enzymes to liberate cross-linking amino acids from the intact protein.
Collagenase-1-Like HME, MME, and the 92-kDa gelatinase, interstitial collagenase cleaves on the amino-terminal side of hydrophobic residues. Although collagenase classically attacks the interstitial collagens, it is also proteolytically active against fibronectin, proteoglycans, entactin, and ␣1-proteinase inhibitor (7,29,30). It is therefore surprising that interstitial collagenase had no activity against the insoluble elastin substrate.
An important factor that markedly influences the hydrolysis rate of the P 1 -P 1 Ј bond is the sequence of amino acids that accommodate subsites P 4 through P 4 Ј (23). There is a strong preference for Pro in subsite P 3 and for unbranched hydrophobic residues at subsite P 2 . For other sites, Ala is the best residue in subsite P 1 , and the aromatic residues Phe and Trp are best in subsite P 2 Ј. Val in subsite P 1 and Val or Phe in subsite P 1 Ј are particularly bad substitutions. Even though Gly-(Leu/Ile) is a common sequence in elastin, it is easy to understand why elastin is not a substrate for collagenase, because sequences that fit all of the subsite criteria are not found in the protein. Elastin's resistance to proteolysis by collagenase emphasizes the importance the subsites play in sub-strate recognition and degradation by this enzyme.
Insoluble Elastin as a Proteolytic Substrate-The use of Edman sequencing techniques to determine proteolytic cleavage sites in insoluble elastin represents a novel approach to deciphering the peptide bond specificities of elastolytic proteinases. Because the mechanism of elastin solubilization is not well understood, it is difficult to accurately infer the complete peptide bond specificity of an elastolytic proteinase using soluble proteins or synthetic substrates. Although many of our findings are consistent with results previously reported with synthetic model substrates, we have also found new and interesting aspects of the proteolytic susceptibility of elastin in its mature, insoluble state. Being able to detect and characterize peptide bond cleavage in insoluble elastin provides information of greater potential relevance to degradative diseases involving elastin than would otherwise be available.