NADPH Oxidase Restrains the Matrix Metalloproteinase Activity of Macrophages*

Matrix metalloproteinases (MMPs) regulate numerous functions in normal and disease processes; thus, irreversibly blocking their activity is a key step in regulating MMP catalysis. We previously showed in vitro that oxidizing intermediates generated by phagocytes inactivate MMPs by modifying specific amino acids. To assess whether this mechanism operates in vivo, we focused on MMP-12, a macrophage-specific MMP known to mediate emphysema in mouse models. We found that mice lacking gp91phox, a phagocyte-specific component of the NADPH oxidase, developed extensive, spontaneous emphysematous destruction of their peripheral air spaces, whereas mice deficient in both NADPH oxidase and MMP-12 were protected from spontaneous emphysema. Although gp91phox-null and wild-type macrophages produced equivalent levels of MMP-12 protein, the oxidant-deficient cells had greater MMP-12 activity than wild-type macrophages. These findings indicate that reactive intermediates provide a physiological mechanism to protect tissues from excessive macrophage-mediated damage during inflammation.

Matrix metalloproteinases (MMPs) 1 comprise a family of 25 distinct vertebrate gene products, of which 24 are found in mammals (1). As well as being involved in the turnover of extracellular matrix, MMPs regulate the activities of numerous effector proteins involved in inflammation and immunity (1,2). The net activity of MMPs is governed, to a large extent, by modulation of gene expression, zymogen activation, and enzyme inactivation. If the balance between enzyme activation (the sum of synthesis and zymogen activation) and inactivation is perturbed, excessive proteolysis may occur. In such situations, natural substrates can be exhausted, leading to misguided and potentially detrimental proteolysis of other proteins. Indeed, overexpression of specific MMPs or an excess of MMP-producing cells in mice and humans associates with a wide range of destructive diseases (1,(3)(4)(5). In these inflammatory disorders, the overproduction of MMPs is attributed to macrophages and other phagocytic white blood cells. Thus, understanding how leukocytes govern MMP-mediated proteolysis should identify fundamental regulatory mechanisms of inflammation.
Inhibition of MMP activity has generally been attributed to tissue inhibitors of metalloproteinases (TIMPs), ␣ 2 -macroglobulin, and other proteins (1, 6, 7), but we have remarkably little understanding of how MMP activity is silenced in vivo. For example, although they are effective inhibitors in vitro, the most clearly demonstrated in vivo functions of TIMPs include activation of MMP activity (8,9) and other actions unrelated to MMP inhibition (10 -12). Analogous to serpins, protein inhibitors of serine proteinases, TIMPs may neutralize active MMPs that have escaped the pericellular environment. To fine-tune cell-directed proteolysis, leukocytes may rely on intrinsic inhibitory mechanisms.
Macrophages use a membrane-associated NADPH oxidase to generate an array of oxidizing intermediates (13). In previous studies, we demonstrated that oxidants potently and efficiently inactivate matrilysin (MMP-7) by cross-linking adjacent tryptophan-glycine residues within the catalytic domain of the enzyme (14,15). These in vitro observations suggest that MMP inactivation can occur on or near phagocytes that produce both MMPs and reactive intermediates. In the absence of reactive intermediates, unrestrained proteolytic activity might lead to detrimental tissue damage. Indeed, inherited deficiency of gp91 phox , a phagocyte-specific component of the NADPH oxidase required for oxidant production, and targeted deletion of its mouse homologue result in granuloma formation and excessive tissue destruction (16 -18). Aberrant regulation of MMP activity may contribute to the damage that occurs when phagocytes are unable to generate oxidants.
To determine whether oxidants regulate MMP activity in vivo, we focused on the role of macrophages in inflammatory lung disease. Although human macrophages express a variety of MMPs, mouse macrophages predominately produce metalloelastase (MMP-12), along with relatively lower levels of gelatinase-B (MMP-9) (19). Moreover, the activity of MMP-12, which is produced only by macrophages, is required for the progression of lung damage in a murine model of cigarette smoke-induced emphysema (20). We therefore hypothesized that phagocyte-derived oxidizing intermediates might regulate the activity of MMP-12 in vivo. In the current studies, we have demonstrated that oxidants do indeed inactivate MMP-12. Surprisingly, we also discovered that gp91 phox mutant mice develop spontaneous, progressive emphysema, equal to that seen in smoke-exposed wild-type animals. Consistent with our idea that oxidants play a physiological role in controlling MMP proteolysis, we saw no spontaneous lung damage in mice deficient in both gp91 phox and MMP-12. Our findings indicate that oxidants may function to govern, rather than promote, inflammatory processes.

EXPERIMENTAL PROCEDURES
Mice-Both gp91 phoxϪ/Ϫ mice (21) (Jackson Laboratories, Bar Harbor, ME) and mmp12 Ϫ/Ϫ mice (22) were in a C57Bl/6J background and had been outcrossed and then intercrossed for three generations to generate animals deficient in both genes. In preliminary studies, mice were exposed to the smoke of two research-grade cigarettes per day, 6 days/week for 1-6 months as described (20). Mice were killed with CO 2 , and their lungs were fixed and processed as described (23). Sections were stained for MMP-12 (20) or with a Mac3 antibody (BD Biosciences) using a Vectastain ABC Elite kit (Vector Laboratories, Inc., Burlingame, CA). Mac3-postive macrophages were counted in 10 randomly chosen areas of sections from each of 3 different animals under ϫ100 magnification and with a 10 ϫ 10 grid superimposed on the image.
Morphometry-We examined hematoxylin/eosin-stained sections in a blinded manner at ϫ50 corrected magnification. For each time point, 10 areas from sections of 6 -12 mice were selected, and the digital images were analyzed using Scion Image and a lung morphometry protocol generated by Dr. Steven Boukedes (Brigham and Women's Hospital, Boston, MA). Data were averaged to provide mean chord length and mean alveolar area (20,24). Average airspace area was calculated by dividing the total airspace by the total number of alveoli using Meta-Morph software (Universal Imaging Corp., Downingtown, PA).
Invasion Assay-Macrophages were harvested by peritoneal lavage 5 days after intraperitoneal injection of 4% thioglycolate (21). Cells were collected by centrifugation, and red blood cells were removed by hypotonic lysis. Cells (Ͼ95% macrophages) were plated on Matrigel-coated inserts (BD Biosciences) in Dulbecco's modified Eagle's medium/10% fetal calf serum and 0.1 g/ml GM-CSF (22). The inserts were placed in tissue culture wells with or without 5 ng/ml of recombinant mouse JE (MCP-1/CCL2; R&D Systems, Minneapolis, MN) in the bottom well and with or without 25 M GM6001 (Chemicon, Temecula, CA) in the upper well. After 72 h, noninvading cells were scraped from the upper surface of the filters, which were then fixed and stained. The number of macrophages that had invaded to the bottom surface of the filters was counted in 10 high power fields (ϫ40 magnification).
The Mca peptide was also used in an MMP-12 fluorescent assay kit (BioMol) to measure MMP activity in isolated macrophage lysates. Peritoneal macrophages (5 ϫ 10 6 ) were lysed in 500 NL of 10 mM Tris buffer containing 1% Nonidet P-40, 150 mM NaCl, 10 mM EDTA, 10 mM NaN 3 , 50 M ␤-hydroxytoluene, and complete protease inhibitor mixture (Roche Diagnostics). The mixture was vortexed and centrifuged to remove cell debris. MMP activity was measured as above, diluting the sample to remove effects of protease inhibitors. Protein levels were estimated by immunoblotting. Total protein was separated through a 4 -12% polyacrylamide gel using MOPS running buffer (50 mM MOPS, 50 mM Tris, 3.5 mM SDS, 1 mM EDTA, pH 7.7) and transferred to a polyvinylidene difluoride membrane (Invitrogen). MMP-12 was detected by immunoblotting using an antibody from Santa Cruz Biotechnology (Santa Cruz, CA). Peritoneal macrophages were also used for RNA isolation and Northern hybridization or were cultured in Dulbecco's modified Eagle's medium supplemented/10% calf serum for 24 h. MMP-12 activity in the conditioned medium was detected by immunoblotting and by substrate zymography (4 -16% prestained Blue Casein Zymogram gel; Invitrogen) (22).
Cytokine Measurements-Lungs were frozen in liquid nitrogen and homogenized in water supplemented with protease inhibitors. Following the addition of cytokine lysis buffer (0.5% Triton X-100, 150 mM NaCl, 15 mM Tris, 1 mM CaCl 2 , and 1 mM MgCl 2 , pH 7.4), homogenates were incubated for 30 min at 4°C and centrifuged at 2000 ϫ g for 20 min at 4°C. Cytokine levels in the supernatants were quantified using the Luminex 100 protein analysis system (Luminex Corp., Austin, TX) (27).
Statistical Procedures-Results represent means Ϯ S.E. Statistical comparisons were made using the unpaired Student's t test, assuming unequal variance, and one-way analysis of variance, using either Tukey's or Dunnett's multiple comparison post-test (Prism 4; GraphPad Software, San Diego, CA). p values Ͻ0.05 were considered significant.

RESULTS
NADPH oxidase reduces molecular oxygen to superoxide, which dismutates to yield hydrogen peroxide. This oxidant is made more reactive by myeloperoxidase, a heme protein secreted by phagocytes that converts peroxide to hypochlorous acid and reactive nitrogen species (28). To assess whether any of these reactive intermediates can directly affect MMP-12 activity, we incubated purified, active enzyme with xanthine and xanthine oxidase, a model system that produces superoxide and hydrogen peroxide (25). This system inhibited the proteolytic activity of MMP-12 by ϳ40% (Fig. 1). The peroxide scavenger catalase and the superoxide scavenger superoxide dismutase blocked inactivation of the proteinase. MMP-12 was inhibited ϳ90 and ϳ60% when myeloperoxidase and chloride or myeloperoxidase and nitrite were included in the system. The heme poison azide blocked inhibition by the myeloperoxidase system, indicating that an active peroxidase was required to inactivate the MMP. These observations demonstrate that reactive intermediates derived from the NADPH oxidase can inactivate MMP-12 in vitro.
Based on these in vitro studies, we hypothesized that oxidizing intermediates act on MMP-12 to inhibit its activity. Thus, in the absence of NADPH oxidase and, in turn, the resulting oxidants, the proteolytic activity of MMP-12 should be greater than in oxidant-producing macrophages. To test this idea, we assessed whether net pericellular proteolysis is enhanced in gp91 phoxϪ/Ϫ phagocytes by measuring the ability of peritoneal macrophages to migrate through Matrigel. Macrophages from gp91 phoxϪ/Ϫ mice were significantly more invasive than macrophages from wild-type mice (Fig. 2a). Moreover, GM6001, a broad-acting inhibitor of metalloproteinase activity, reduced the invasiveness of gp91 phoxϪ/Ϫ macrophages to wild-type levels (Fig. 2a). Thus, the enhanced invasiveness of gp91 phoxϪ/Ϫ macrophages requires the activity of a metalloproteinase.
Because macrophages use MMP-12 to migrate through Matrigel, (22) we used cells from double knock-out (gp91 phoxϪ/Ϫ ,mp12 Ϫ/Ϫ ) mice to determine whether the enhanced invasiveness of gp91 phoxϪ/Ϫ macrophages was due to increased MMP-12 activity. Macrophages from double mutants were markedly less invasive than macrophages from mice deficient in only NADPH oxidase (gp91 phoxϪ/Ϫ ,mp12 ϩ/ϩ ) (Fig. 2a). Furthermore, the invasive activity of double mutant macrophages was essentially the same as for wild-type or gp91 phoxϪ/Ϫ single knock-out cells in the presence of GM6001.
The increased invasive activity of gp91 phoxϪ/Ϫ macrophages could be due to higher MMP-12 expression per cell or greater Catalytically active recombinant human MMP-12 was exposed to superoxide and hydrogen peroxide (H 2 O 2 ) generated by xanthine oxidase and xanthine (25). Where indicated, the reaction mixture was supplemented with myeloperoxidase (MPO) plus chloride (Cl Ϫ ), myeloperoxidase plus nitrite (NO 2 Ϫ ), catalase (CAT) and superoxide dismutase (SOD), or myeloperoxidase plus chloride and azide. *, p Ͻ 0.01. Activity was measured using a fluorogenic substrate. net MMP-12 activity with no relative change in protein level. To distinguish between these possibilities, we measured MMP-12 mRNA and protein levels in isolated macrophages. Relative levels of MMP-12 protein, normalized to cell number, in conditioned medium (data not shown) or cell lysates (Fig. 2b) and steady-state mRNA levels (data not shown) were essentially the same in macrophages from wild-type or gp91 phoxϪ/Ϫ mice. In contrast, MMP-12 activity in conditioned medium (data not shown) and cell lysates was 3-fold higher in macrophages from gp91 phoxϪ/Ϫ mice than in wild-type (gp91 phoxϩ/ϩ ) macrophages (Fig. 2c). These findings suggest that increased MMP-12 activity, not production, led to the enhanced ability of gp91 phoxϪ/Ϫ macrophages to migrate through a Matrigel barrier.
To assess whether reactive intermediates regulate MMP activity in vivo, we planned to examine the extent of cigarette smoke-induced emphysema in gp91 phoxϪ/Ϫ mice. Much to our surprise, we found that the peripheral air spaces of 3-6-monthold air-breathing control gp91 phoxϪ/Ϫ mice had enlarged spontaneously (Fig. 3, a and b). Unlike other gene-targeted mice that develop early onset, spontaneous alveolar enlargement (29,30), defective organogenesis did not lead to the emphysema in gp91 phoxϪ/Ϫ mice. When they reached 3 months of age, their lung architecture was indistinguishable from that of wild-type mice (Fig. 3c). By 6 months of age, however, the gp91 phoxϪ/Ϫ mice began to lose alveoli and alveolar ducts, and these changes progressed with age. By 12 months, the air spaces in all gp91 phoxϪ/Ϫ mice examined were markedly enlarged and clearly emphysematous (Fig. 3, b and c). In fact, we could accurately deduce the genotype of these otherwise apparently normal mice from their lung phenotype. At 12 months, the average air space in gp91 phoxϪ/Ϫ lungs was significantly greater (p ϭ 0.0095) than in wild-type lungs (Fig. 3c). The nearly 2-fold difference was similar, if not slightly greater than, the emphysematous damage seen in wild-type mice exposed to cigarette smoke for 6 months (20).
The delayed appearance of emphysema in the gp91 phoxϪ/Ϫ mice indicated that a postnatal inflammatory process contributed to tissue injury. Compared with wild-type mice, lungs of gp91 phoxϪ/Ϫ mice had more leukocytes (Fig. 3b). As determined by staining with Mac3, most of these cells were macrophages. In 3-month-old gp91 phoxϪ/Ϫ mice, the number of macrophages in lungs was only 1.4-fold higher than in wild-type mice, but at 12 months there was an ϳ3-fold increase (Fig. 3d). However, lungs of gp91 phoxϪ/Ϫ mice showed no histological evidence of infection, such as neutrophil infiltration or abscess formation. Thus, the increased number of macrophages likely resulted from an impairment of cell-mediated processes rather than from an opportunistic infection. We cannot exclude the possibility that the increase in macrophages may have been due to a persistent, low-level infection, but the lack of neutrophils argues against this suggestion.
MMP-12, a macrophage-specific MMP, mediates the development of smoke-induced emphysema in mice (20). To assess FIG. 2. Enhanced invasion and MMP-12 activity of macrophages from NADPH oxidase-deficient mice. a, peritoneal macrophages were plated on Matrigel-coated membranes in medium with or without GM6001, a broad-acting metalloproteinase inhibitor, and the number of cells that invaded the matrix barrier was quantified. Results were normalized to the number of invaded gp91 phoxϩ/ϩ macrophages and represent the mean cell number Ϯ S.E. of two to five individual experiments with two to three determinations per analysis. *, p Ͻ0.05; **, p Ͻ0.001. b, peritoneal macrophages from gp91 phoxϩ/ϩ were lysed, and total protein was quantified. Equal amounts of protein (260 or 520 g) were resolved by electrophoresis, and MMP-12 (arrow) was detected by immunoblotting. c, total MMP activity was quantified using a fluorogenic substrate. Data shown are the mean Ϯ S.E. for assays from four different macrophage preparations. whether MMP-12 was responsible for the lung destruction seen in mice lacking gp91 phox , we generated mice deficient in both NADPH oxidase and MMP-12 (gp91 phoxϪ/Ϫ ,mmp12 Ϫ/Ϫ ). Like either of the single knock-outs, the double mutants were healthy and reproduced normally. Also, their peripheral airways (Fig. 4, e and f) were identical to those in age-matched wild-type mice at 6 and 12 months of age (Fig. 2, a and b). In contrast, the alveolar spaces were markedly enlarged in mice deficient in NADPH oxidase alone (gp91 phoxϪ/Ϫ ,mmp12 ϩ/ϩ ) (Fig. 4, c and d, and Fig. 5, a and b).
Immunostaining for Mac3 and MMP-12 indicated that more macrophages and, in turn, more immunoreactive proteinase were present in lungs of gp91 phoxϪ/Ϫ mice (Fig. 4, c and d) than in wild-type (gp91 phoxϩ/ϩ ) mice (Fig. 4, a and b). Although lungs of gp91 phoxϪ/Ϫ ,mmp12 Ϫ/Ϫ mice also contained more macrophages than age-matched wild-type mice, they had only 33% fewer cells than the lungs of NADPH oxidase-deficient (gp91 phoxϪ/Ϫ ,mmp12 ϩ/ϩ ) mice (Figs. 4 and 5c). Despite this slight reduction in macrophage influx, double mutants showed marked protection against the development of emphysema, with morphometric values that did not differ significantly from wild-type values (Fig. 5, a and c). These data indicate that the damage seen in gp91 phoxϪ/Ϫ mice was more dependent on the presence of MMP-12 than on macrophage number.
To explore potential causes for the increased macrophage influx seen in the gp91 phoxϪ/Ϫ lungs, we measured levels of inflammatory chemokines and cytokines (Table I). Macrophage inflammatory protein-2 (MIP-2/CXCL2) was undetectable in wild-type lungs, but high levels were detectable in lungs of gp91 phoxϪ/Ϫ and double mutant (gp91 phoxϪ/Ϫ ,mmp12 Ϫ/Ϫ ) mice. In contrast, interleukin-6 and KC/CXCL1 were detected in lungs of wild-type animals, and levels of both factors were modestly higher in lungs of both mutant strains. Thus, whereas MMP-12 deficiency almost completely protected the NADPH oxidase-deficient mice from emphysema, the lack of this proteinase only partially prevented macrophage accumulation in lungs and did not reduce the higher levels of MIP-2, interleukin-6, or KC associated with gp91 phox deficiency. DISCUSSION Our data indicate that pericellular production of oxidants by phagocytes controls proteolysis by MMPs in vivo. We observed excessive lung destruction in mice deficient in phagocyte NADPH oxidase (gp91 phoxϪ/Ϫ ). Because smoke-induced emphysema in mice requires the activity of MMP-12 and because we observed emphysema in gp91 phoxϪ/Ϫ mice in the absence of an initiating insult (i.e. cigarette smoke), we conclude that macrophage-derived oxidants govern pericellular proteolysis by these cells. The likely mechanism involves direct inactivation of catalytically active MMP-12 by reactive intermediates, as demonstrated in our in vitro experiments.
Macrophages and neutrophils use membrane-associated NADPH oxidase to generate reactive oxygen intermediates (13,18). The initial product of the NADPH oxidase is superoxide, which is converted into oxidizing oxygen and nitrogen species. Because their characteristic end products have been detected in diseases ranging from atherosclerosis to neurodegenerative disorders, reactive oxygen and nitrogen intermediates are thought to contribute to inflammatory tissue injury (31-33). However, humans and animals deficient in phagocyte NADPH oxidase tend to form granulomas and to have excessive tissue destruction and an exuberant inflammatory response, (16,17,34), raising the possibility that oxidants derived from white cells actually govern or suppress inflammation.
One potential mechanism whereby reactive oxygen species can influence inflammation and the associated tissue damage is by regulating the activity of MMPs. In addition to their ability to act on extracellular matrix, MMPs can affect inflammation by directly or indirectly regulating the activity of inflammatory mediators such as chemokines (1,23,(35)(36)(37). Because reactive intermediates effectively inactivate MMPs in vitro, (14, 15, 38 -40) they provide an efficient mechanism for inhibiting unregulated catalysis by these extracellular proteinases, thereby preventing pathological destruction of tissue proteins and exuberant inflammation. Production of reactive intermediates by the phagocyte NADPH oxidase could confine MMP activity in space and time, permitting only bursts of pericellular proteolysis (41).
Another implication of our in vivo findings is that oxidants participate in tissue homeostasis, because the progression of emphysema in the gp91 phoxϪ/Ϫ mice in the absence of an overt exogenous stimulus implies that resident macrophages are engaged in the normal turnover of alveolar extracellular matrix. The turnover of many structural matrix proteins is slow. For example, lung elastin, a key matrix protein targeted in the pathogenesis of emphysema, is stable once lung growth is complete (42). It is likely that matrix turnover is controlled, in part, by mechanisms that restrain proteolysis, and we speculate that oxidants serve this function in normal tissue.
If our idea proves correct, it may help explain why clinical trials of antioxidants have generally yielded disappointing results (43,44) and perhaps why the antioxidant beta-carotene increases the risk for lung cancer in humans who smoke (44,45). The basis for such trials has been that antioxidants should prevent the oxidative modification of proteins, lipids, and nucleic acids by phagocytic cells that is widely believed to inflict tissue injury (31)(32)(33)46). Our observations suggest, however, that oxidants may also play beneficial roles, such as restraining MMPs. Thus, they cast doubt on the widely held view that reactive intermediates are necessarily harmful.