Metalloproteinase-mediated Shedding of Integrin β2 Promotes Macrophage Efflux from Inflammatory Sites*

Background: Efflux of macrophages limits inflammation. Results: Macrophage integrin αMβ2 is cleaved during exiting from inflammatory sites, released αMβ2 retains ligand binding capabilities, and inhibition of its metalloproteinase-mediated cleavage impairs macrophage efflux. Conclusion: Metalloproteinase-mediated proteolysis of integrin β2 promotes macrophage efflux from inflammatory sites. Significance: Regulated proteolysis of integrin β2 during inflammation demonstrates the potential of this mechanism to contribute to resolution of inflammation. Macrophage exiting from inflammatory sites is critical to limit the local innate immune response. With tissue insult, resident tissue macrophages rapidly efflux to lymph nodes where they modulate the adaptive immune response, and inflammatory macrophages attracted to the site of injury then exit during the resolution phase. However, the mechanisms that regulate macrophage efflux are poorly understood. This study has investigated soluble forms of integrin β2 whose levels are elevated in experimental peritonitis at times when macrophages are exiting the peritoneum, suggesting that its proteolytic shedding may be involved in macrophage efflux. Both constitutive and inducible metalloproteinase-dependent shedding of integrin β2 from mouse macrophages are demonstrated. Soluble integrin β2 is primarily released as a heterodimeric complex with αM that retains its ability to bind its ligands intracellular adhesion molecule-1, fibrin, and collagen and thus may serve as a soluble antagonist. In a model of accelerated exiting, administration of a metalloproteinase inhibitor prevents macrophage efflux by 50% and impedes loss of macrophage integrin β2 from the cell surface. Exiting of peritoneal macrophages in mice lacking integrin β2 is accelerated, and antibody disruption of integrin β2-substrate interactions can reverse 50% of the metalloprotease inhibitor blockade of macrophage exiting. Thus, our study demonstrates the ability of metalloproteinase-mediated shedding of integrin β2 to promote macrophage efflux from inflammatory sites, and the release of soluble integrin heterodimers may also limit local inflammation.

Macrophage exiting from inflammatory sites is critical to limit the local innate immune response. With tissue insult, resident tissue macrophages rapidly efflux to lymph nodes where they modulate the adaptive immune response, and inflammatory macrophages attracted to the site of injury then exit during the resolution phase. However, the mechanisms that regulate macrophage efflux are poorly understood. This study has investigated soluble forms of integrin ␤2 whose levels are elevated in experimental peritonitis at times when macrophages are exiting the peritoneum, suggesting that its proteolytic shedding may be involved in macrophage efflux. Both constitutive and inducible metalloproteinase-dependent shedding of integrin ␤2 from mouse macrophages are demonstrated. Soluble integrin ␤2 is primarily released as a heterodimeric complex with ␣M that retains its ability to bind its ligands intracellular adhesion molecule-1, fibrin, and collagen and thus may serve as a soluble antagonist. In a model of accelerated exiting, administration of a metalloproteinase inhibitor prevents macrophage efflux by 50% and impedes loss of macrophage integrin ␤2 from the cell surface. Exiting of peritoneal macrophages in mice lacking integrin ␤2 is accelerated, and antibody disruption of integrin ␤2-substrate interactions can reverse 50% of the metalloprotease inhibitor blockade of macrophage exiting. Thus, our study demonstrates the ability of metalloproteinase-mediated shedding of integrin ␤2 to promote macrophage efflux from inflammatory sites, and the release of soluble integrin heterodimers may also limit local inflammation.
Integrin ␤2 has been shown to play a critical role in leukocyte recruitment to sites of inflammation. In particular, the requirement for integrin ␤2 to induce rapid arrest of leukocytes on the vessel wall at sites of inflammation has highlighted the importance of ␤2 activation of integrin avidity in this process (1,2). Avidity, or the overall strength of cellular adhesiveness, is controlled by the intrinsic affinity of the individual receptor-ligand bonds and the number of these bonds (valency), and avidity can be altered by integrin density, geometric conformation, and lateral mobility in the cell membrane (3). Leukocyte recruitment also requires the integrin-ligand bond to be dynamic to allow de-adhesion. Generation of constitutively active integrin ␣L␤2 (lymphocyte function-associated antigen 1 (LFA-1)) 2 dramatically impaired deactivation and de-adhesion from its ligand intercellular adhesion molecule (ICAM)-1 (4). Thus, both activation and deactivation of leukocyte integrin heterodimers are needed to promote firm adhesion and de-adhesion during a normal inflammatory response.
An alternative mechanism that could functionally act in a manner similar to integrin deactivation, and could also induce an immediate change in integrin density, is proteolytic shedding of leukocyte integrins. Although integrin proteolysis has not been shown to regulate leukocyte trafficking, cleavage of another adhesion molecule, L-selectin, selectively limits early neutrophil influx (5). Shedding of other adhesion molecules can also elicit rapid changes in cellular responses, instantly lower adhesion molecule density, and release soluble forms of their extracellular domains that can act as antagonists (6,7). In an effort to detect novel proteins proteolytically cleaved from the surface of inflammatory macrophages, we identified the extracellular domain of integrin ␤2 in two proteomic screens using different inflammatory stimuli (8). 3  Previous studies have also demonstrated soluble forms of integrin ␤2 associated with the loss of integrin ␤2, or one of its ␣ subunits, from the leukocyte cell surface during inflammation (9 -12). Time-dependent loss of integrin ␣M␤2 (macrophage-1 antigen or Mac-1) from infiltrating neutrophils in the reperfused canine myocardium was shown to be associated with the appearance in lymph of soluble heterodimers, which had lower molecular weights than cellular heterodimers (9). In human blister fluid, truncated ectodomains of ␣L␤2 (LFA-1) were detected, and neutrophils and monocytes showed a loss of cell surface ␣L␤2 (10). More recently, shedding of integrin ␤2 complexed with different ␣ subunits was shown to be a part of synovial inflammation in rheumatoid arthritis and spondyloarthritis but not in osteoarthritis (12). These authors further demonstrated the ability of soluble integrin ␤2 to inhibit monocytic cell adhesion to ICAM-1. Thus, soluble integrin ␤2 with the ability to buffer leukocyte adhesion has been identified in physiological fluids from multiple inflammatory states in vivo, but the functional significance of in vivo shedding from the cell surface has not been investigated. Analysis of molecular mechanisms involved in the resolution of acute inflammation has also identified the loss of macrophage surface ␣M (CD11b) in a distinct macrophage subpopulation that may represent specialized "pro-resolving" macrophages (13,14). Together these studies raise the possibility that leukocyte shedding of integrin ␤2 heterodimers could play a role in leukocyte recruitment to inflammatory sites and/or the resolution of inflammation.
This study biochemically documents proteolytic shedding of the integrin ␤2 ectodomain from the surface of mouse macrophages, and it identifies zinc-dependent metalloproteinases as major regulators of this cleavage. Soluble integrin ␤2 retains its ability to bind its substrates, and thus may serve as a soluble antagonist. Analysis of peritoneal lavage fluid following administration of the sterile irritant thioglycollate shows that in vivo release of soluble integrin ␤2 is most marked at times when macrophages are exiting the peritoneal cavity. Furthermore, interference with proteolytic shedding is sufficient to significantly reduce macrophage exiting in a model of accelerated macrophage exiting from the peritoneal cavity. Diminished exiting can be partially rescued by disruption of integrin ␤2-substrate interactions, and peritoneal macrophage exiting in mice lacking integrin ␤2 is accelerated. Together, these data establish a functional role for integrin ␤2 heterodimer shedding in macrophage efflux from sites of inflammation.

EXPERIMENTAL PROCEDURES
Mice-C57BL/6J mice were from The Jackson Laboratory. Integrin ␤2 null mice were previously described (15) and were backcrossed 10 times onto the C57BL/6J background as were the Mmp9 null mice obtained from Robert Senior (Washington University). Hematopoietic chimeras lacking Adam17 were recently described (5), and floxed Adam10 mice (16) were provided by Howard Crawford and were crossed with LysM-Cre transgenic mice (The Jackson Laboratory). Animals were housed in a pathogen-free facility, and procedures were approved by the University of Washington Institutional Animal Care and Use Committee.
Peritonitis Model-Thioglycollate-induced peritonitis was initiated by intraperitoneal injection of 1 ml of 4% sterile thioglycollate (BD Biosciences). Peritoneal cells were collected by injection of 5 ml of PBS, 5 mM EDTA, and the lavage fluid was saved following cell removal by centrifugation.
Human Monocyte Cultures-Human peripheral blood mononuclear cells were isolated from citrated blood by Ficoll-Paque Plus (Amersham Biosciences) separation and enriched for monocytes by negative selection with monocyte isolation kit II (Miltenyi Biotech). Monocytes were resuspended in Opti-MEM at 4 ϫ 10 6 /ml in 15-ml polypropylene tubes and incubated with activating CD18 MEM-48 antibody (10 g/ml) or control IgG or in the presence of ultrapure LPS for 4 h at 37°C. Monocytes were evaluated by FACS and conditioned media by ELISA.
Soluble Integrin ␤2 Binding to Its Ligands-Polysorb microtiter plates were coated with different amounts of purified human fibrinogen (American Diagnostica), bovine collagen type I (Cohesion), and fibronectin (Invitrogen). For fibrinogen, bovine thrombin (Innovative Research; 6 l of solution at 1.2 mg/ml, activity 1900 NIH units/mg) was added at 4°C and left at RT for 30 min to allow fibrin formation. Collagen gels were allowed to form (1 h at 37°C) after neutralization. For rICAM-1, plates were coated with goat anti-human IgG (Fcspecific) at 0.5 g/well in 50 mM sodium bicarbonate buffer (pH 9.2) for 16 h at 4°C, and nonspecific binding was blocked with 0.5% (w/v) BSA. Thereafter, 50 l of different concentrations of recombinant ICAM-1/Fc (R&D Systems) in PBS containing 0.1% (w/v) BSA was added to the plate and incubated overnight at 4°C. Lavage or conditioned media were added to the extracellular matrix-or ICAM-1-coated wells and incubated for 4 h at 37°C and assayed for depletion of soluble integrin ␤2 levels by ELISA.
Macrophage Binding to Integrin ␤2 Ligands-Plates were coated with extracellular matrix proteins or ICAM-1 as described above, and 4-day elicited peritoneal macrophages were added to the plate at 10 ϫ 10 6 /ml, 100 l/well, and incubated for 4 h at 37°C. Media were assayed for soluble integrin ␤2 by ELISA.
Model of Accelerated Macrophage Exiting-As published previously, LPS (1 g of LPS in 200 l of PBS) (17) or TNF-␣ (0.5 g in 200 l of PBS) was given by intraperitoneal injection 3 days after thioglycollate injection, and peritoneal cells were harvested 4 h later as described for the peritonitis model. For mice that received GM6001, it was dissolved in 4% carboxymethylcellulose in PBS to give a final concentration of 1.5 mg/ml, and peritoneal injection of the GM6001 (200 l/mouse, or 0.3 mg/mouse) was 20 min before LPS injection. For blocking integrin ␤2 antibody experiments, 50 g of 2E6 Fab fragments or control hamster IgG was administered 20 min before LPS or TNF-␣ together with GM6001 or with carboxymethylcellulose control. Fab fragments of 2E6 were generated following purification of the IgG from hybridoma supernatants using protein G-Sepharose (Pierce) and Fab preparation kit (Pierce). Control hamster IgG was purchased from eBioscience.
Protein Analysis and Immunoprecipitation of Soluble Integrin ␤2 Complexes-Cell lysates and conditioned media were separated by 4 -12% gradient gel SDS-PAGE and analyzed by Western blot. For immunoprecipitation, peritoneal lavage was incubated in 96-well plates coated with antibodies to integrin ␤2, integrin ␣M, integrin ␣L, and integrin ␣X overnight at 4°C and eluted with SDS sample buffer. Immunoblots were probed with rat monoclonal anti-integrin ␤2 (MAB2618) or goat polyclonal anti-integrin ␤2 (AF2619) from R&D Systems.
Statistics-Statistical analysis was performed using a twotailed Student's t test using InStat (GraphPad software). All values are presented as the means Ϯ S.E. with p Ͻ 0.05 considered significant unless otherwise noted.

RESULTS
Proteolytic Release of Integrin ␤2 by Mouse Resident and Elicited Peritoneal Macrophages-We have proposed that proteolytic shedding of cell surface proteins contributes to regulation of leukocyte recruitment to inflammatory stimuli (6). To characterize properties of cleaved integrin ␤2 identified in macrophage proteomics screens, elicited mouse peritoneal macrophages (stimulated to emigrate into the peritoneal cavity by injection of the sterile irritant thioglycollate) were harvested 4 days after thioglycollate and activated in vitro with different stimuli. Without stimulation, soluble integrin ␤2 is detected in conditioned medium, and both LPS and PMA stimulation lead to a 2-fold increase in integrin ␤2 in the medium (Fig. 1A). Analysis of the ratio of shed integrin ␤2 in conditioned medium to lysate integrin ␤2 shows ϳ1-4% is shed in 4 -6-h conditioned medium and 5-20% after 24 h. These percentages are even more significant considering that up to 90% of total cellular integrin ␤2 is intracellular, as shown by comparing flow cytometry of fixed versus fixed and permeabilized macrophages and consistent with previous reports of significant intracellular stores of ␣M␤2 (18). Adhesion of peritoneal macrophages to increasing amounts of integrin ␤2 physiological ligands fibrin and collagen, but not another matrix protein fibronectin, also leads to a dose-dependent increase in soluble integrin ␤2 (Fig.  1B). As compared with macrophage lysates, soluble integrin ␤2 shows a small molecular mass reduction to ϳ90 kDa under all conditions (Fig. 1, A, C, and E), and this small shift is consistent with cleavage within the ectodomain.
Because LPS stimulation and adhesion to integrin ligands can both lead to integrin activation, we asked whether integrin ␤2 activation alone could induce its proteolytic release. We utilized well characterized antibodies to human integrins that can induce integrin activation and also detect the activated epitope. As shown in supplemental Fig. 1, human monocytes release soluble integrin ␤2 during a 4-h incubation in suspension, and levels released are increased 3-fold by incubation with LPS. However, addition of integrin ␤2-activating antibody MEM-48 did not increase soluble integrin ␤2 (supplemental Fig. 1A) despite a 4-fold increase in the activated epitope (supplemental Fig. 1B). Thus, integrin ␤2 activation is not sufficient to enhance its shedding.
To evaluate protease families involved in integrin ␤2 cleavage, the broad spectrum metalloproteinase inhibitor GM6001 and a mixture of metalloproteinase-sparing protease inhibitors were added to adherent cultures of peritoneal macrophages (Fig. 1C). GM6001 almost completely inhibits shedding of integrin ␤2 with soluble levels reduced by 80 -90% (Fig. 1D), suggesting a matrix metalloproteinase (MMP) or ADAM (A Disintegrin And Metalloproteinase) protease. In contrast, the matrix metalloproteinase-sparing protease inhibitors have no effect on soluble integrin ␤2 (Fig. 1C), ruling out serine, cysteine, aspartic acid, and aminopeptidases as major contributors.
Soluble integrin ␤2 is detected in peritoneal fluid collected in vivo with resident (no stimulation) and elicited macrophages (4 days following thioglycollate injection) with higher levels of soluble integrin ␤2 following thioglycollate injection (Fig. 1D, right  panel). In vitro culture of the peritoneal macrophages also leads to release of soluble integrin ␤2, which is significantly reduced in the presence of GM6001 (Fig. 1D, left panel). The decrease in molecular weight of soluble integrin ␤2 as compared with cell lysates is seen in both plasma and peritoneal lavage fluid (Fig.  1E), suggesting proteolytic cleavage of integrin ␤2 in vivo. No signals are detected by Western blot in fluids from ␤2 knockout mice (Fig. 1E), establishing the specificity of the antibodybased data. Together, these experiments establish that integrin ␤2 is cleaved in vitro and in vivo in a metalloproteinase-dependent manner, and cleavage leads to the release of a soluble form of integrin ␤2.
Because we previously identified soluble integrin ␤2 in a proteomics screen of medium from macrophages overexpressing active MMP-9 (8), we asked whether MMP-9 is required for soluble integrin ␤2 release. As shown in Fig. 1D (left panel), MMP-9 null peritoneal macrophages in vitro release soluble integrin ␤2 at levels comparable with wild type peritoneal macrophages. However, GM6001 blocks integrin ␤2 release from MMP-9 null macrophages less efficiently than for wild type (Fig. 1D). Table 1 shows that levels of soluble integrin ␤2 are also comparable in peritoneal fluid from MMP-9 null and wild type mice. Both the pro-and active forms of MMP-9 have previously been demonstrated in peritoneal fluid following thioglycollate injection (19). Peritoneal fluid from mice deficient in myeloid ADAM10 and ADAM17, two enzymes that have been shown to shed many cell surface proteins (6,20), was also tested (Table 1), and their deletion also did not significantly alter shedding of integrin ␤2. The decreased efficiency of GM6001-mediated blockade of soluble integrin ␤2 shedding from MMP-9 null macrophages suggests that increased levels of nonmetalloproteinase enzymes may compensate for the loss of MMP-9. Thus, although MMP-9 is capable of cleaving integrin ␤2 (8), neither this MMP nor ADAM10 and ADAM17 are required, suggesting that additional enzymes participate in shedding of integrin ␤2.
Soluble Mouse Integrin ␤2 Is Primarily Released as a Complex with Integrin aM and the Soluble Complex Retains Ligandbinding Capacity-Leukocyte ␤2 integrins form heterodimers with one of four ␣ subunits as follows: ␣L (LFA-1, CD18/ CD11a), ␣M (Mac-1/CR3, CD18/CD11b), ␣X (CD18/CD11c), and ␣D (CD18/CD11d). Because many integrin ␤2 functions depend on the presence of the ␣ subunit (21), we asked whether peritoneal lavage fluid contained integrin ␤2 complexed with one or more of its potential ␣ subunits. Immunoprecipitation shows that integrin ␤2 is primarily complexed with integrin ␣M (Fig. 2A). Because we were unable to detect the different ␣ subunits by Western analysis using multiple commercially available antibodies to mouse integrins, and because the ␤2 ELISA is

Levels of soluble integrin ␤2 in the peritoneal fluid are not altered by the absence of MMP-9 or by ADAM10 or ADAM17 myeloid deficiency
Integrin ␤2 levels in the peritoneal fluid (units/ml) collected 3 days following thioglycollate injection or 3 days following thioglycollate and an additional 4-h post-injection of LPS were measured by ELISA. The means Ϯ S.E. are shown for the indicated number of mice. ND indicates not determined. a more sensitive method for detection of soluble form than immunoblotting, we performed immunoprecipitation followed by ELISA to quantify the contribution of different ␣ subunits. Immunodepletion of peritoneal lavage fluid (Fig. 2B) with antibodies either to integrin ␤2 or combined antibodies to ␣L, ␣M, and ␣X remove Ͼ70% (70.0 Ϯ 11.1 and 72.5 Ϯ 7.8%, respectively, in three independent experiments) of the total soluble integrin ␤2, whereas antibodies to ␣M and ␣X are able to deplete by 40.0 Ϯ 5.3 and 26.5 Ϯ 0.2%, respectively. To assess whether the shed integrin ␤2 complex still retains biological function, its ability to bind to its in vivo ligands, ICAM-1, fibrin, and collagen, was evaluated. Recombinant mouse ICAM-1-coated plates (Fig. 2C), fibrin (Fig. 2D), and collagen (Fig. 2E) dose-dependently deplete integrin ␤2. However, fibronectin, which is not an integrin ␤2 ligand, failed to remove any significant amount of soluble integrin ␤2 (Fig. 2D).

Soluble integrin ␤2 in 3-day post-thioglycollate also with 4-h LPS peritoneal fluid
Thus, soluble integrin ␤2 is released as a complex primarily with integrin ␣M, and this complex readily binds its ligands.
Metalloproteinase Inhibition Blocks LPS-induced Macrophage Exit from the Peritoneum and Increases ␤2 Surface Levels Specifically on Retained Macrophages-Sterile peritonitis models have been a valuable approach to identify specific mediators and cell types involved in regulating the acute inflammatory response because the models phenocopy macrophage exiting, inflammatory neutrophil and monocyte infiltration, and subsequent self-resolving phases of the response (14). A single peritoneal injection of the sterile irritant thioglycollate leads to the exit of almost all of the resident peritoneal macrophages within 4 h and their migration to the draining lymph nodes. Following the exit of resident peritoneal macrophages, inflammatory monocytes infiltrate to levels that peak at 3-4 days, followed by a slow exiting of recruited monocyte/macrophages over 4 -14 days to the omentum and draining lymph nodes (22,23). Evaluation of peritoneal lavage fluid collected at different times using this model demonstrates that ␤2 integrin levels were highest at times of peritoneal macrophage exiting ( Table 2). This led us to hypothesize that macrophages may shed integrin ␤2 to leave the peritoneal cavity.
To test this hypothesis, we used a model of accelerated macrophage exiting from the peritoneal cavity stimulated by LPS activation of thioglycollate-elicited macrophages (17,24) shown in Fig. 3A. In this model, LPS promotes macrophage adhesion and migration across the mesothelium into the lymphatics in a fibrin-dependent manner (24). To block ␤2 integrin cleavage, the metalloprotease inhibitor GM6001 was injected 15 min prior to LPS administration, and cells remaining in the peritoneal cavity 4 h after LPS were collected and analyzed (Fig.  3). Macrophages that remain in the peritoneal cavity are used to assess exiting because it is difficult to quantitatively evaluate macrophage content in lymph nodes and omentum around the peritoneal cavity. LPS promotes exiting of 15 million macrophages, although 5 million remain in the peritoneum. However, in the presence of GM6001, macrophage exiting is reduced to 7 million, a reduction of 50% (Fig. 3B). Because TNF-␣ is downstream of LPS stimulation, we also tested whether TNF-␣ was sufficient to induce macrophage exiting and integrin ␤2 shedding. As shown in Fig. 3C, TNF-␣ stimulates macrophage exiting to a similar extent as LPS and was inhibited by GM6001, suggesting that they may both act through a common molecular mechanism.
GM6001 also prevents the release of soluble integrin ␤2 into peritoneal fluid (Fig. 3D). Furthermore, surface expression of ␤2 integrin is specifically elevated 5-fold on macrophages of mice treated with GM6001, but no difference in ␤2 expression FIGURE 2. Soluble integrin ␤2 from peritoneal lavage is primarily complexed with integrin ␣M, and adsorbed ligands deplete integrin ␤2 in a concentration-dependent manner. A, mouse peritoneal lavage fluid from wild type mice was collected 4 days after thioglycollate and was immunocaptured with antibodies to individual and combined integrin ␣ subunits, integrin ␤2 subunit, and control rat IgG. The captured complexes were separated by 4 -12% gradient SDS-PAGE and analyzed by Western blotting with polyclonal anti-mouse integrin ␤2 antibody. B, soluble integrin ␤2 from peritoneal lavage collected from three different mice was subjected to multiple depletions with antibodies against integrin ␤2 and individual and combined antibodies to ␣ subunits. The remaining unbound soluble integrin ␤2 was collected and measured using the mouse ␤2 ELISA. *, p Ͻ 0.002; **, p ϭ 0.0002 relative to isotype control. C-E, quantification by ELISA is shown of soluble integrin ␤2 remaining following 4 h of incubation of peritoneal lavage fluid from four to five different mice with plates coated with increasing amounts of the following. C, recombinant mouse rICAM-1/Fc at concentrations of 0.01-2 g/well (*, p Ͻ 0.001); D, purified human fibrinogen at 0.05-300 g/well treated with thrombin to generate fibrin (*, p ϭ 0.0013), and purified fibronectin at the same concentrations (not significant); E, purified bovine type I collagen at 0.05-300 g/well (*, p Ͻ 0.001). Means Ϯ S.E. is observed on neutrophils, B-cells, or T-cells (Fig. 3E). These data are consistent with our hypothesis that ␤2 integrin, in the absence of GM6001, is shed from macrophages as they exit the peritoneal cavity. However, it is unclear to what extent the cleavage of ␤2 integrin and/or other metalloproteinase substrates are responsible for macrophage exiting.
Integrin ␤2 Deficiency Accelerates Macrophage Exiting and GM6001 Inhibits Exiting That Is Partially Rescued by Disruption of Integrin ␤2 Substrate Interactions-If cleavage of ␤2 integrin is functionally involved in promoting macrophage exiting, we would predict that absence of ␤2 integrin would accelerate their exit. We tested this possibility in mice either ϩ/ϩ or Ϫ/Ϫ for ␤2 integrin (Fig. 4A). As predicted, mice lacking ␤2 integrin showed twice as many macrophages exiting the peritoneal cavity at 2 h after LPS injection.
Another strategy to evaluate the role of proteolytic shedding of integrin ␤2 in peritoneal macrophage exiting is to ask whether an antibody capable of disrupting integrin ␤2 interactions with its ligands, antibody 2E6 (25), can reverse a significant portion of macrophage exiting inhibited by GM6001. To avoid dimer formation and associated signaling from whole IgG, Fab fragments of 2E6 or a hamster isotype control IgG were injected with GM6001 15 min prior to LPS injection. GM6001 with control IgG shows the expected retention of macrophages in the peritoneal cavity (Fig. 4B). However, administration of 2E6 Fab fragments promotes macrophage exiting from the peritoneum despite the presence of GM6001 and is able to rescue ϳ50% of the macrophage exiting blocked by GM6001 following LPS (Fig. 4B) and TNF-␣ stimulation (Fig. 4C). Together these data demonstrate that proteolytic shedding of integrin ␤2 significantly contributes to the regulation of macrophage exiting during the inflammatory responses.

DISCUSSION
Our study highlights the contribution of metalloproteinasemediated cleavage of integrin ␤2 to macrophage exiting from an inflammatory site, a previously unrecognized mechanism for controlling macrophage trafficking and limiting the local response to injury. Defects in the process of inflammatory cell exiting can promote progression to chronic inflammation and persistent tissue injury. However, because disrupting integrin ␤2-substrate interactions only restored 50% of the LPS-induced macrophage exiting blocked by GM6001 (Fig. 4), cleavage of other macrophage proteins is likely involved in regulation of macrophage exiting. It is also unclear what the specific ligand is that integrin ␣M␤2 binds to within the peritoneal cavity, and which signaling pathways are involved in initiation of shedding and macrophage exiting.
Extravascular fibrin deposition is an early and persistent hallmark of the inflammatory response in the peritonitis model and promotes macrophage cell-cell and cell-fibrin interactions, thus serving as a provisional matrix for recruited macrophages (24,26,27). Deficiency in fibrin(ogen) does not prevent macrophage infiltration in response to thioglycollate, but it does impair macrophage activation (28) and adhesion to the peritoneal cavity mesothelial lining (24). However, our detection of soluble forms of two fibrin(ogen)-binding integrin dimers, ␣M␤2 and ␣X␤2, in peritoneal fluid suggests that macrophage binding to fibrin may promote their shedding. Mac-1 deficient macrophages (mice lacking integrin ␣M) did not exit the peritoneal cavity in the accelerated exiting model (17), although we observed more rapid exiting in the same model with mice lacking integrin ␤2. Outside-in signaling (29) and binding to low density lipoprotein receptor-related protein (LRP)-1, which is required for migration (30), are both impaired in the absence of integrin ␣M. Thus, although Mac-1 null mice still express ␣X␤2 that could mediate fibrin binding, their macrophages would be impaired in "outside-in" signaling from fibrin. It would also be interesting in future experiments to evaluate whether ␣X␤2 is shed in Mac-1-deficient mice and the extent of shedding in wild type and mutant mice of the most recently identified ␤2 integrin for which reagents have been lacking, ␣D␤2, which can also bind fibrinogen (31). In contrast, macrophages lacking integrin ␤2 would not express any of the fibrinbinding integrin heterodimers, and thus they would not need to shed the ␤2 integrins to leave and could utilize integrin ␣4␤1 for migration (32,33). It is interesting to note that the integrin dimer that does not bind fibrin, ␣L␤2 (LFA-1), is expressed on recruited macrophages but is not shed into peritoneal lavage fluid. Thus, we propose that LPS and TNF-␣ may promote integrin activation and fibrin adhesion leading to shedding of integrins ␣M␤2 and ␣X␤2 and enhancement of macrophage efflux (Fig. 5).
Macrophage exiting from the peritoneal cavity is impaired by deletion of components of another protease pathway, the plasminogen activation cascade, and specifically by deletion of tissue plasminogen activator (tPA) (34). tPA is expressed on the macrophage cell surface and specifically interacts with integrin ␣M␤2 (34). Cao et al. (34) further suggested that the endocytic receptor LRP-1 associates with integrin ␣M␤2 through a mechanism in which tPA-plasminogen activator inhibitor-1 complexes bridge LRP-1 with ␣M␤2 and promote the internalization of ␣M␤2, thus facilitating cell detachment. However, an alternative explanation for their data is that rather than being internalized, ␣M␤2 is cleaved from the cell surface, a possibility that is more consistent with the low inhibition of macrophage exiting observed following administration of the LRP-1 inhibitor RAP (34). Fibrinolysis is required for optimal exiting from the peritoneal cavity (27), and it seems likely that fibrinolysis mediated by tPA may contribute to the nonmetalloproteinase component of macrophage efflux (Fig. 5).
Although our in vitro analysis of shedding with matrix metalloproteinase-sparing inhibitors suggests that Ͼ90% of integrin ␤2 shedding is mediated by MMPs or ADAMs, other proteases may contribute under different conditions. The serine proteases elastase, proteinase-3, and cathepsin G were recently reported to cleave human neutrophil integrins ␣M␤2 (11). In addition to release of the ϳ90-kDa fragment of integrin ␤2 representing the majority of the extracellular domain, Zen et al. (11) also detected a fragment of ϳ50 kDa. However, in our analysis of peritoneal fluid, plasma, and conditioned media, FIGURE 4. In vivo administration of disrupting integrin ␤2 antibody 2E6 promotes macrophage exit from the peritoneum prevented by the metalloproteinase inhibitor GM6001. A, mice either integrin ␤2 ϩ/ϩ or Ϫ/Ϫ received an LPS injection as described in Fig. 3A, and then cells were harvested after 2 h to evaluate exiting (n ϭ 4 -5/group; n.s. ϭ not significant). B and C, experimental protocol to measure macrophage exiting was the same as in Fig. 3A except that the 15-min incubation with GM6001 or vehicle control also included treatment with 150 g of 2E6 Fab fragments or hamster isotype IgG control prior to LPS injection. Macrophage cell number was quantitated by FACS analysis (F4/80ϩ, Ly6GϪ). The numbers of macrophages remaining in the peritoneal cavity with different treatments is shown: B, following LPS administration (n ϭ 3-4/group); C, after TNF-␣ stimulation (n ϭ 3/group). The data are expressed as the number of macrophages that remain in the peritoneal cavity after 2 or 4 h. Means Ϯ S.E. are shown, and the data are representative of two replicate experiments.
only the ϳ90-kDa form of soluble integrin ␤2 has been detected, suggesting that integrin ␤2 cleavage by serine proteases may be neutrophil-specific and/or human-specific.
Based on our identification of soluble integrin ␤2 in a proteomics screen of medium from macrophages overexpressing active MMP-9, this enzyme was a strong candidate. Similarly, a significant number of leukocyte cell surface adhesion proteins are shed by ADAM10 and ADAM17 (6,20). However, analysis of mice lacking MMP-9, or with myeloid deficiency of ADAM10 or ADAM17, suggests that these three enzymes are not major contributors to integrin ␤2 cleavage from the macrophage cell surface under conditions comparable with the model of accelerated macrophage exiting from the peritoneal cavity. Although multiple enzymes appear to cleave the same substrates in vitro, it is becoming increasingly clear that in vivo cleavage is both cell-and context-specific. For example, we recently demonstrated context specificity in ADAM17 cleavage of L-selectin that is required for limiting neutrophil rolling on endothelial cells, but neutrophil L-selectin levels are reduced Ͼ90% during neutrophil extravasation independent of ADAM17 (5). So, despite the almost complete lack of L-selectin shedding in Adam17 null cultured cells in vitro (35), conditions exist in vivo that utilize distinct enzymes not revealed in studies of cultured cells. Thus, additional candidate metalloproteinases need to be evaluated in vivo.
Rapid efflux of resident macrophages from an inflammatory site is an important component of the adaptive immune response (Fig. 5). In studies of accelerated macrophage efflux, pro-inflammatory cytokine release is induced with subsequent neutrophil influx (data not shown) in a manner similar to the initial response to thioglycollate (36). In contrast, the slow and spontaneous exiting of macrophages with the resolution of inflammation induced by thioglycollate or zymosan proceeds over 4 -14 days after the stimulus and is associated with the release of anti-inflammatory cytokines and pro-resolving lipid mediators (14). The pro-resolving lipid mediators lipoxin A4 and resolvin E1 stimulate monocyte shedding of L-selectin in the blood helping to reduce monocyte influx (37). We show that peritoneal levels of soluble integrin ␤2 are elevated both at the time of initial exiting of resident macrophages and during the slow resolution phase. As shown by studies of synovial fluids (12), the release of soluble integrin ␤2 may buffer leukocyte adhesion thus further limiting leukocyte influx and/or activation. In addition, soluble complexes may retain signaling capabilities.
Defects in the initial efflux of resident macrophages can have significant implications for controlling the inflammatory stimuli, whereas impaired exiting of inflammatory macrophages during the resolution phase can promote progression to chronic inflammation and persistent tissue injury. Thus, there is a need to identify pathways that regulate macrophage efflux as therapeutic targets, and our study now adds integrin ␤2 shedding to the list of mediators of macrophage efflux from inflammatory sites.