Mannose-binding lectin (MBL) mutants are susceptible to matrix metalloproteinase proteolysis: potential role in human MBL deficiency.

Mannose-binding lectin (MBL) plays a critical role in innate immunity. Point mutations in the collagen-like domain (R32C, G34D, or G37E) of MBL cause a serum deficiency, predisposing patients to infections and diseases such as rheumatoid arthritis. We examined whether MBL mutants show enhanced susceptibility to proteolysis by matrix metalloproteinases (MMPs), which are important mediators in inflammatory tissue destruction. Human and rat MBL were resistant to proteolysis in the native state but were cleaved selectively within the collagen-like domain by multiple MMPs after heat denaturation. In contrast, rat MBL with mutations homologous to those of the human variants (R23C, G25D, or G28E) was cleaved efficiently without denaturation in the collagen-like domain by MMP-2 and MMP-9 (gelatinases A and B) and MMP-14 (membrane type-1 MMP), as well as by MMP-1 (collagenase-1), MMP-8 (neutrophil collagenase), MMP-3 (stromelysin-1), neutrophil elastase, and bacterial collagenase. Sites and order of cleavage of the rat MBL mutants for MMP-2 and MMP-9 were: Gly(45)-Lys(46) --> Gly(51)-Ser(52) --> Gly(63)-Gln(64) --> Asn(80)-Met(81) which differed from that of MMP-14, Gly(39)-Leu(40) --> Asn(80)-Met(81), revealing that the MMPs were not functionally interchangeable. These sites were homologous to those cleaved in denatured human MBL. Hence, perturbation of the collagen-like structure of MBL by natural mutations or by denaturation renders MBL susceptible to MMP cleavage. MMPs are likely to contribute to MBL deficiency in individuals with variant alleles and may also be involved in clearance of MBL and modulation of the host response in normal individuals.

Mannose-binding lectin (MBL) 1 plays a critical role in innate host defense, recognizing microorganisms and mediating their destruction by complement activation or phagocytosis (1). MBL binds mannose and N-acetylglucosamine residues on microbial surfaces, whereupon MBL-associated serine proteinases (MASPs) initiate the complement cascade by the lectin pathway resulting in cell lysis (2). Opsonization of microorganisms by MBL can also result in phagocytosis mediated by cell surface receptors on polymorphonuclear cells or macrophages (3).
MBL is a member of the collagen-containing lectin family of proteins, termed collectins (4,5). MBL polypeptides consist of a short NH 2 -terminal cysteine-rich domain, a collagen-like domain of 18 -20 repeats of Gly-Xaa-Yaa containing an interruption (Gly-Gln-Gly) that induces a kink in the extended structure of the molecule, and a COOH-terminal carbohydrate recognition, or lectin, domain (6,7). MBL subunits are comprised of three identical 31-kDa polypeptides held together by disulfide bonds and a triple helix formed by the collagenlike domains (8). Three to five subunits are disulfide cross-linked to form MBL oligomers that are required for complement activation (8,9).
MBL serum levels are affected by promoter polymorphisms (10 -12) and three mutations in the collagen-like domain which are known to occur in the human population with a combined frequency of ϳ0.20 (13): R32C, G34D, and G37E (1). These patients have low levels of MBL in their serum and suffer unusual or severe infections early in life before adaptive immunity is fully functional, as well as in adult life (14 -16) or when immunocompromised, for example after chemotherapy (17,18) or when suffering from AIDS (19). Although not firmly established, a link between decreased MBL concentrations and homozygous mutant alleles and the disease severity of rheumatoid arthritis (12,13), cystic fibrosis (20,21), and systemic lupus erythematosus (22,23) has been proposed, implicating MBL in the pathogenesis of these diseases.
Variant human MBL and recombinant rat MBL with homologous mutations show decreased formation of the higher order oligomers (15,24,25). Replacing glycine with relatively bulky acidic side chains or introducing a cysteine reduces hydroxylation and glycosylation of proximal proline and lysine residues. This disrupts the disulfide bond arrangement of the NH 2 -terminal cysteine-rich domain, which perturbs the local structure of the collagen triple helix (24 -26). These structural defects prevent MASPs from binding to the collagen-like domain (27,28), and together with defective oligomerization, result in impaired complement activation (24,25). It is unclear whether MBL serum deficiency is the result of decreased MBL secretion (24 -26), enhanced turnover, or a combination of these. We speculated that variant MBL with its perturbed collagen triple helix may be more susceptible to proteolysis than wild-type protein, but this has yet to be addressed.
These enzymes exhibit broad substrate specificity and in addition to extracellular matrix components, process a wide range of bioactive molecules including cell surface receptors, growth factors, and chemokines (31). Because MBL has a collagen-like domain, we evaluated the susceptibility of human serum MBL to MMP collagenolytic activity and compared this with recombinant rat MBL with mutations homologous to those found in human MBL (25). Our studies revealed that denaturation or mutations in MBL which perturb the collagen-like structure render MBL susceptible to cleavage by multiple MMPs. These findings support the hypothesis that MMPs could contribute to low serum levels of MBL in diseases marked by MBL deficiency.

EXPERIMENTAL PROCEDURES
Purification of Human MBL-Human MBL was purified as described with modification (32). Human serum was obtained from volunteer donors who had provided written informed consent. The study was performed under an approval for human subjects research obtained from the University of British Columbia. A crude preparation of MBL was obtained by passage of human serum through mannose-Sepharose and maltose-Sepharose columns. MBL was purified further using Sephacryl S-300, equilibrated in acetate buffer (100 mM sodium acetate, 100 mM NaCl, 5 mM EDTA, pH 5.0), to remove MASPs. ␣ 2 -Macroglobulin (which can inhibit MMPs) was removed using an anti-␣ 2 -macroglobulin column: ␣ 2 -macroglobulin IgG (Enzyme Research, South Bend, IN) was coupled to cyanogen bromide-activated Sepharose 4B (Amersham Biosciences), and unreacted sites were blocked with glycine. The resin was equilibrated in TBS (50 mM Tris-HCl, 150 mM NaCl, pH 7.4). MBL was recovered in the column flow-through, and purity was assessed by SDS-PAGE and silver staining.
Production of Rat MBL-Wild-type rat MBL (serum form, MBP-A) and MBL containing mutations homologous to those occurring in human MBL deficiency, R23C (R23C-MBL), G25D (G25D-MBL), and G28E (G28E-MBL), were generated and expressed in Chinese hamster ovary cells and purified as described (8,25). CaCl 2 was added to a final concentration of 5 mM before use.
Cleavage of MBL-MBL was denatured by incubation at 60°C for 20 min to disrupt the collagen triple helix and then cooled on ice. Native or heat-denatured MBL was incubated in the presence of MMP, neutrophil elastase, or bacterial collagenase in assay buffer (100 mM Tris, pH 8.0, 30 mM CaCl 2 ). Proteinase inhibitors were included in some incubations as described. For competition assays, increasing amounts of MMP domains were incubated in the reaction. Samples were electrophoresed reduced or nonreduced on 12.5% Tris-Tricine gels and silver stained. Human MBL was Western blotted using an ␣-MBL monoclonal antibody against the carbohydrate recognition domain of human MBL (clone 131-1, a gift from S. Thiel, Aarhus University, Denmark) followed by goat anti-mouse horseradish peroxidase-conjugated secondary anti-body (Bio-Rad). Immunoreactive protein was detected by enhanced chemiluminescence (Amersham Biosciences).
NH 2 -terminal Sequencing-Protease-digested MBL samples (nonreduced) were electrophoresed on 12.5% Tris-Tricine gels and blotted to polyvinylidene fluoride membrane. Membranes were stained with Coomassie Brilliant Blue R-250, destained with 50% methanol, and bands of interest were excised. NH 2 -terminal sequences were determined by Edman Chemistry using an Applied Biosystems Procise Sequencer. Standard derivatives were not detected for proline and lysine residues, which have been shown to be modified to hydroxyproline or glucosylgalactosylhydroxylysine in rat MBL (8).

Human MBL Is Cleaved Differentially by MMPs-Native
and denatured collagen (gelatin) are cleaved differentially by members of the MMP family. To determine whether collagenolytic MMPs have the potential to regulate MBL levels in vivo by cleavage in the collagen-like domain, we tested the susceptibility of native or denatured human serum MBL to proteolysis.
Because MMP-2, MMP-9, and MMP-14 were the most efficient at cleaving denatured human MBL, these enzymes were analyzed further. On nonreducing SDS-PAGE, MBL oligomers migrate near the top of the gel, whereas cleavage fragments, no longer bound to the oligomer, separate in the gel and are distinguished readily. After overnight incubation, MMP-2 generated two fragments from denatured human MBL of 24.8 and 22.2 kDa ( Fig. 2A). The 22.2-kDa fragment was also produced by MMP-9, whereas a fragment of 26.3 kDa was released by MMP-14. To determine the sites of MMP cleavage, fragments were subjected to NH 2 -terminal sequencing (Table I). Cleavage sites are designated hP 1 -P 1 Ј, where h indicates human MBL and P 1 and P 1 Ј are the residues either side of the scissile bond. Specificity was indicated by unique sites of cleavage in denatured human MBL by MMP-2, hGly 54 -Lys 55 , and MMP-14, hGly 48 -Leu 49 , and by a common cleavage site, hGly 72 -Gln 73 for MMP-2 and MMP-9. All three cleavage sites are located within the collagen-like domain of MBL, COOH-terminal to the GQG interruption, such that the stable products consist of the carbohydrate recognition domain, neck, and 2-10 Gly-Xaa-Yaa repeats (Fig. 3). This was confirmed by Western blotting using an antibody raised against the human MBL carbohydrate recognition domain (Fig. 2B). Bacterial collagenase released similar sized fragments of 20.4 and 18.3 kDa (Fig. 2B).
To confirm MMP-specific cleavage, BB94 (a synthetic MMP hydroxamate inhibitor), TIMP-2 (a specific proteinaceous inhibitor of MMPs), or phenylmethylsulfonyl fluoride (a serine proteinase inhibitor) was included in the proteolytic assays with MMP-2 ( Fig. 2C). 2.5 mM phenylmethylsulfonyl fluoride had no effect on the cleavage of denatured MBL by MMP-2, whereas both 250 nM BB94 and 482 nM TIMP-2 completely inhibited cleavage at hGly 72 -Gln 73 and reduced the intensity of the hGly 54 -Lys 55 fragment. Incomplete inhibition indicates the rapidity of the generation of these products before inhibition occurs. This confirms that the processing of human MBL was caused by the added MMPs and not by any MASPs remaining in the preparation.
Because neutrophils are present at sites of infection and inflammation where MBL is localized, we examined the ability of neutrophil elastase to cleave human MBL. Unlike neutrophil-specific MMP-8 ( Fig. 1C), neutrophil elastase partially cleaved native MBL to fragments of 26.2 and 22.4 kDa (Fig. 4A, ϪDTT), visible as a band of 27.7 kDa under reducing conditions (Fig. 4B, ϩDTT). The intensity of these bands was enhanced after denaturation of MBL, indicating that although elastase could cleave native MBL, proteolysis was limited by the triple helical fold of the collagen-like domain. No synergistic effect on MBL proteolysis was detected using combinations of neutrophil elastase and MMPs (data not shown).
Proteases often localize to substrates via specific binding  a X represents a cycle where no derivative was obtained. These residues were assigned as either hydroxyproline (O) or glucosylgalactosylhydroxylysine (K*) according to the rat MBL sequence (8), except for the first residue of hG 48 -L 49 , which was deduced to be leucine by comparison with the human MBL sequence. sites, termed exosites, which also increase catalytic activity (40). Gelatinases bind to collagen via a domain inserted into the MMP catalytic domain, the CBD (37), comprised of three fibronectin type II modules, and to the chemokines monocyte chemoattractant protein-3 (41) and stromal cell-derived factor-1␣ (33) via the hemopexin C domain. To determine the role of the CBD and hemopexin C domain in the cleavage of MBL, recombinant domains were used in competition assays. The addition of Ն 500:1 molar ratio of CBD:MMP-2 resulted in an increase in the 24.8 and 22.2 kDa bands (Fig. 6A, lower panel).
This appeared to be the result of stimulation of cleavage at hGly 54 -Lys 55 and hGly 72 -Gln 73 because the oligomeric forms of MBL decreased with increasing CBD (Fig. 6A, upper panel). Neither CBD nor MMP-2 was detected by the ␣-MBL monoclonal antibody (data not shown). The addition of exogenous MMP-2 hemopexin C domain also resulted in an increase of the 24.8-kDa product at Ն 100:1 C domain:MMP-2 (Fig. 6B). However, in this case there was a concomitant decrease in the  The hemopexin C domain of MT1-MMP with (Fig. 6C) or with-out (Fig. 6D) the "linker" that connects this domain to the catalytic domain had the same effect on cleavage of MBL as the MMP-2 hemopexin C domain at a similar concentration. However, cleavage of MBL by MT1-MMP was not affected significantly by the addition of the MMP-2 or MT1-MMP hemopexin C domain (data not shown). Hence, native human MBL was resistant to MMP activity, but upon denaturation it was cleaved efficiently in the collagen-like domain by MMPs. In the case of MMP-2, the mechanism involves both the CBD and hemopexin C domains, whereas MT1-MMP appears less dependent on the hemopexin C domain for MBL cleavage.
Rat MBL Containing Naturally Occurring Human Mutations Is Cleaved Readily by MMPs-Naturally occurring mutations in MBL are known to perturb the helical structure of the collagen-like domain. Because denatured human MBL was cleaved efficiently by MMPs, we hypothesized that proteolysis of variant MBL proteins might contribute to low serum levels in these patients. To address this, we examined the MMP susceptibility of recombinant rat MBL containing point mutations corresponding to natural human mutations, R23C-MBL, G25D-MBL, and G28E-MBL (Fig. 3). Nonreduced recombinant wild-type rat MBL electrophoresed at the top of gels consistent with higher order oligomers (Fig. 7, first lane from left). In contrast, because of defective oligomerization (25), R23C-MBL, G25D-MBL, and G28E-MBL electrophoresed as multiple bands ranging from 27 to 157 kDa (ϪMMP lanes, Fig. 7), representing single polypeptides and complexes of up to five or six polypeptides. Wild-type rat MBL was resistant to cleavage by MMPs (Fig. 7, A-C, first two lanes) unless heat-denatured (data not shown), whereas the nondenatured mutant MBL proteins were processed efficiently by MMP-2, MMP-9, and MMP-14 (Fig. 7, A-C). Virtually all polypeptide forms of R23C-MBL, G25D-MBL, and G28E-MBL were degraded, although the 100 -200-kDa complexes of R23C-MBL were more resistant than those of the other two mutants. NH 2 -terminal sequencing of major fragments was performed to determine the sites of cleavage, designated rP 1 -P 1 Ј where r signifies rat MBL (Table II). After incubation, MMP-2 degraded the three mutants to a doublet with a major component of 19.5 kDa and a second product of 19 kDa (Fig. 7A). Degradation of R23C-MBL by MMP-2 resulted in an additional minor product of 17.2 kDa. The 19.5-kDa product for all three mutants resulted from cleavage at rAsn 80 -Met 81 (Table II and Fig. 3). Cleavage by MMP-2 was inhibited by the MMP inhibitors BB94 and TIMP-2, but not by phenylmethylsulfonyl fluoride, ruling out the involvement of MASPs (data not shown). Incubation with MMP-9 resulted in major products of 23 and 19.5 kDa with minor products of 24.4 and 17.8 kDa for all three mutants (Fig. 7B). The 19.5-kDa product was identical to that generated by MMP-2, rAsn 80 -Met 81 , and the 23-kDa fragment resulted from cleavage at rGly 63 -Gln 64 (Table II and Fig. 3). MMP-14 cleaved the three mutants to a unique 27.5-kDa fragment by cleavage at rGly 39 -Leu 40 and the 19.5-kDa product by cleavage at rAsn 80 -Met 81 (Fig. 7C and Table II).
Other MMPs were less efficient at cleavage. MMP-3 cleaved the monomers of R23C-MBL, G25D-MBL, and G28E-MBL to products of 19.5 and 17.8 kDa, but it was selective in its cleavage of the 100 -200 kDa bands (data not shown), suggesting that oligomerization inhibited cleavage by this enzyme. There was a small amount of cleavage of the mutants to low molecular mass forms by MMP-1 (18.9 and 16.7 kDa) and an enhancement of the 27.5 kDa band by MMP-8, but no cleavage by MMP-7 or MMP-13 (data not shown).
Although the human collagenases (MMP-1, MMP-8, and MMP-13) were ineffective against wild-type rat MBL and only weakly active against the MBL mutants, native wild-type rat MBL was partially degraded by bacterial collagenase to fragments of 19.4, 18.2, and 17.3 kDa (Fig. 8A). The three MBL mutants were less resistant to cleavage, with bacterial collagenase generating high levels of these fragments together with an additional product of 16.7 kDa. The latter was most likely generated from the 17.3-kDa fragment, which was depleted. Neutrophil elastase also partially cleaved native wild-type rat MBL to a prominent product at 16.9 kDa and a faint band at 29.3 kDa (Fig. 8B). The mutants were cleaved by neutrophil elastase to a doublet of 17.7 and 16.9 kDa, and R23C-MBL was processed to an additional product of 29.3 kDa.
Efficacy of MBL Proteolysis-To determine the MMP potency and cleavage pattern of MBL, MBL mutants were incubated with MMP-2, MMP-9, or MMP-14 for various periods of time and at different enzyme concentrations (Fig. 9). A series of intermediates was observed over time when G25D-MBL was incubated with MMP-2, MMP-9, or MMP-14 at a constant enzyme:substrate ratio of 1:12 (Fig. 9A). Cleavage by MMP-2 was highly efficient (Fig. 9Ai). A 23.7-kDa fragment appeared rapidly after a 1-min incubation at 37°C. This initial fragment resulted from cleavage at rGly 51 -Ser 52 (Table II and Fig. 3). After 10 min of incubation, an intermediate band of 23 kDa was produced by cleavage at rGly 63 -Gln 64 . The rAsn 80 -Met 81 fragment was produced after 10 min and was the major stable product that accumulated after overnight incubation together with a minor product at 19 kDa as found previously (Fig. 7A). The oligomers were largely degraded after a 10-min incubation, which was also the case with MMP-9 (Fig. 9Aii). With this enzyme, the initial 24.4-kDa product, rGly 45 -Lys 46 , was prominent with the rGly 63 -Gln 64 fragment visible after a 1-min incubation, becoming a major band by 10 min. After overnight incubation, the rGly 63 -Gln 64 fragment and rAsn 80 -Met 81 fragment were the predominant stable products as before (Fig. 7B). A secondary signal in the NH 2 -terminal sequencing of rGly 45 -Lys 46 gave the sequence SVGAXG, suggesting some cleavage at rGly 51 -Ser 52 as for MMP-2. Conversely, the minor band above the MMP-2-derived rGly 51 -Ser 52 23.7-kDa product (Fig. 9Ai,  1-60 min) could also be the result of cleavage at rGly 45 Table I. Parentheses indicate minor cleavage products for the MMP indicated.
stable product (Fig. 9Aiii). The rAsn 80 -Met 81 fragment began to appear after 1 h but was a minor product compared with the MMP-14-specific rGly 39 -Leu 40 fragment. Transient bands at 58, 39.6, and 30 kDa (indicated by arrows in Fig. 9Aiii) began with the sequence SGSQTX, which corresponds to the NH 2 terminus of full-length rat MBL. These likely consist of disulfide cross-linked fragments that are seprated by cleavage at rAsn 80 -Met 81 . The polypeptide complexes persisted up to 4 h of incubation with this enzyme. With all three MMPs, cleavage was concentration-dependent and proceeded through a series of intermediates that corresponded to those generated in the time course experiments (Fig. 9B). After overnight incubation with 0.2 pmol of MMP-2, all of the forms of G28E-MBL were degraded except for the 24.9 kDa band, which remained along with a new band of 23.7 kDa (Fig. 9Bi). This initial cleavage product likely corresponds to the rGly 51  Both human serum MBL and recombinant rat MBL were resistant to MMP activity in vitro unless denatured, indicating that the triple helical collagen-like domain masks proteasesensitive sites on the individual polypeptide chains. However, recombinant rat MBL proteins with mutations homologous to those found in human variant MBL were rapidly and concentration-dependently cleaved at specific sites by the gelatinolytic MMPs, MMP-2, MMP-9, and the MMP-2 activator MMP-14 and less efficiently by MMP-3, MMP-1, MMP-8. Neutrophil elastase and bacterial collagenase also cleaved at specific sites within the collagen-like domain. Cleavage of variant rat MBL occurred at sites homologous to those cleaved in human MBL (Table II), except for rAsn 80 -Met 81 and rGly 51 -Ser 52 which are sites unique to rat MBL. Digestion of denatured wild-type rat MBL by MMP-2, MMP-9, and MMP-14 produced the same fragments as digestion of G28E-MBL (data not shown). Although it is possible that the mutations have generated a sequence that is now recognized in the native state by the hemopexin C domain exosites, this is unlikely because the location of the Gly residues in the core of the triple helix renders them relatively inaccessible to solvent, and the three mutations are different and occur independently. Hence, we suggest that MBL mutations structurally destabilize the collagen-like domain of MBL, permitting cleavage by MMPs through the exposure of cleavage sites and exosite binding regions on individual strands within the perturbed triple helix. The processing of MBL mutants at sites identical to those cleaved in heat-denatured human or wild-type rat MBL supports previous studies reporting that the structure of the collagen-like domain of rat MBL mutants R23C-MBL, G25D-MBL, and G28E-MBL is perturbed (25,26). Cleavage of variant MBL and denatured human MBL was rapid and efficient, occurring within minutes of exposure to MMP-2, MMP-9, and MMP-14 at low enzyme:substrate ratios, suggesting that MMPs are likely to cleave the mutants in vivo. Effective MMP cleavage of mutant MBL but resistance of wild-type MBL provides an explanation for the absence of MBL in homozygotes but persistence of low levels of MBL in heterozygotes (12,13,15).
The MBL binding site for the receptor C1qRp, which is proposed to mediate phagocytosis (3), lies in the collagen-like domain (42) (see Fig. 3) encompassing the sequence G 37 EKGEP which includes Gly 37 and is near the other MBL mutations. Hence, receptor binding may already be disrupted in variant MBL, but cleavage of MBL by MMPs between the receptor binding site and carbohydrate recognition domain (Fig. 3) would separate the recognition and effector functions of the molecule. However, the biological effects of MMP cleavage on MBL function in vitro or in vivo cannot be assessed because both mutant MBL and denatured MBL have a perturbed collagen domain structure that alone disrupts MASP binding and oligomerization, which are requirements for complement activation (25,26,28). This also suggests that MMP inhibitors would be of little therapeutic benefit for patients with variant forms of MBL. Hence it is likely that in addition to any effects of the mutations in the collagen-like domain on secretion, increased turnover of variant MBL by members of the MMP family contributes to the low serum levels of variant MBL in vivo.
With regard to wild-type human MBL, there was no evidence for cleavage of the purified protein in vitro by MMPs without heat denaturation. A role for MMPs in the clearance of MBL cannot be ruled out however, because additional factors may be involved in vivo. MBL is part of a complex with MASPs which, like ligand binding (2), may alter the conformation of MBL rendering it susceptible to cleavage. Oxidation by free radicals derived from activated neutrophils and macrophages, which impairs the function of the structurally homologous molecule C1q (43), may affect the structure of MBL. Finally, other proteolytic events may occur, initiating a cascade resulting in degradation of MBL by MMPs. However, using purified MASPfree MBL in vitro, neither binding to mannose-Sepharose or bacterial cell fragments nor oxidation by hydrogen peroxide increased the susceptibility of native MBL to MMP cleavage. 3 Further studies are required to elucidate the clearance pathway for MBL in vivo and to assess the involvement of MMPs.
Degradation of MBL by MMPs might be a mechanism to evade MBL-mediated host detection and destruction. Both mi-croorganisms and tumor cells, which express mannose-containing proteins on their surface and are killed or growth-inhibited by MBL (44,45), express collagenases (46 -48) and can induce host cells to secrete MMPs (49 -51). Cleavage of native MBL by bacterial collagenases, some of which are reported to activate MMPs (52), supports the idea that these collagenases could act as virulence factors in this regard. Tumor cells sequester MMPs at the cell surface (for review, see Ref. 53), where they may be poised to degrade MBL bound to mannose-containing cell surface proteins, potentially enabling the tumor to escape host detection.
The mechanism of collagen cleavage by MMPs has yet to be elucidated fully, although a role for exosites on the hemopexin C domain, acting in concert with specific residues in the catalytic domain in the case of MMP-1, is established (40,54). Competition experiments confirmed that exosite domains are important for cleavage of MBL. The apparent stimulation of cleavage of denatured human MBL at hGly 54 -Lys 55 by the CBD suggests that in the full-length enzyme this exosite, which is specific to gelatinases, orients the collagen strands in a conformation that is cleaved effectively by MMP-2 (for review, see Ref. 40). Although the hemopexin C domain of MMP-2 also enhanced the generation of hGly 54 -Lys 55 , the lack of effect on oligomer levels and decreased cleavage at hGly 72 -Gln 73 suggests that binding of exogenous hemopexin C domain masks the second cleavage site and/or prevents processivity of the enzyme driven by the hemopexin C domain, which may locate the catalytic domain in the correct orientation for cleavage at hGly 72 -Gln 73 . The interchangeable effect of the hemopexin C domain of MT1-MMP implies that the hemopexin C domain of this cell membrane-bound MMP is also involved in the definition of substrate cleavage sites. Indeed, the MT1-MMP hemopexin C domain binds native type I collagen. 2 However, the hemopexin C domains of MMP-2 and MT1-MMP had no significant effect on cleavage of denatured human MBL by MT1-MMP at hGly 48 -Leu 49 (data not shown). This implies that either the mode of cleavage of this collagen-like substrate by MMP-2 and MMP-14 is different or that the effect on processivity is the same, except that MMP-14 only cleaves at a single site so the effect is not apparent.
Many of the biologically active proteins processed by MMPs play important regulatory roles in inflammation or innate host defense. For example, MMP-7 activates the bacteriocidal activity of ␣-defensins in the small intestine (for review, see Ref. 31). Interestingly, bacteria induce MMP-7 expression by mucosal epithelia by an unknown mechanism, potentially involving MBL as a bacterial sensor. Our laboratory recently reported that MMP-2 dampens inflammation by cleaving the chemokines monocyte chemoattractant protein-3 (41) and stromal cell-derived factor-1␣ (33). MMPs may have additional roles in modulating the acute inflammatory response by effecting the clearance of MBL to facilitate resolution.
There is a growing number of modular plasma and cell surface proteins recognized with trimeric subunits comprised of collagen-like regions and noncollagenous recognition domains which share considerable structural and functional homology with MBL. These include the other members of the collectin family (55,56): surfactant proteins SP-A and SP-D; bovine proteins conglutinin and collectin-43; and CL-L1 as well as other proteins such as C1q (57), ficolins (58), EMILINs (59), and adiponectin (60). Most of these proteins are believed to function in innate immunity where they specifically bind various microorganisms or allergens, leading to complement activation, opsonization, or enhanced phagocytosis. Some are inhibitors: SP-A suppresses C1q-mediated complement activation (61), and adiponectin negatively regulates hematopoeisis and phagocytosis via C1qRp (60), suggesting involvement in termination of the inflammatory response. A newly discovered group of homologous proteins, the EMILINs, have C1q and collagen-like domains and are found in extracellular matrix where they may mediate elastogenesis and cell adhesion (59). The presence of homologous collagen-like regions suggests that this may be a common target for cleavage of these proteins by MMPs. Indeed pepsin-treated C1q is cleaved in the collagenlike domain by several MMPs (62). 3 New members and functions of this modular protein family are being discovered continually, and it will be interesting to determine whether MMP processing plays a role in their regulation.