Oxidized α₂-Macroglobulin (α₂M) Differentially Regulates Receptor Binding by Cytokines/Growth Factors: Implications for Tissue Injury and Repair Mechanisms in Inflammation¹

The 720-kDa homotetrameric broad-spectrum proteinase inhibitor α₂M is found in the plasma at high (micromolar) concentrations (for a review, see 1 . Interaction between α₂M and proteinases in plasma and extracellular fluids involves a unique trapping mechanism by which the proteinase is incorporated covalently into the α₂M molecule via a unique β-cysteinyl-γ-glutamyl thioester bond. Following scission of the thioester bond, the conformation of α₂M changes to a more compacted structure, exposing its receptor recognition sites for binding to both the low density lipoprotein receptor-related protein (LRP)³ (2, 3) and the α₂M signaling receptor (4, 5). This conformational change can be generated chemically by reacting α₂M with small primary amine nucleophiles such as methylamine, forming an α₂M/methylamine complex (hereafter designated α₂M*) that behaves in many ways identically with a proteinase/α₂M complex (6).

Although α₂M has traditionally been viewed as a plasma and inflammatory fluid proteinase scavenger, evidence has accumulated in recent years suggesting that in vivo α₂M can bind to cytokines and growth factors such as TNF-α, IL-1β, IL-2, IL-6, bFGF, β-NGF, PDGF, and TGF-β (for a review, see Refs. 7 and 8). Binding to α₂M abolishes the ability of most cytokines/growth factors to regulate cell functions while enhancing the ability of a few others (9, 10, 11, 12, 13, 14, 15, 16). Cytokines bind to α₂M with affinities that vary from micromolar K_d values for early inflammatory mediators such as TNF-α, to nanomolar K_d values for late inflammatory mediators such as bFGF, β-NGF, PDGF, and TGF-β (10, 17, 18, 19, 20). It has been shown that very little free TGF-β and PDGF are found in the circulation, since 85 to 90% of them are bound to α₂M (10, 13, 14, 21). Yet, to date no mechanism has been demonstrated in vivo that inhibits the binding of growth factors to α₂M, raising the question of how growth factors are able to function when α₂M is present in high concentrations in the plasma and inflammatory fluids. Given that bFGF, β-NGF, PDGF, and TGF-β have been implicated in tissue injury repair mechanisms such as angiogenesis, fibroblast proliferation, smooth muscle cell proliferation, collagen deposition, and neuronal regeneration (22, 23, 24, 25), it appears that a mechanism must exist that inhibits the binding of α₂M to these growth factors and/or allows these growth factors to be released from α₂M to regulate cell functions.

We and others are investigating the role of oxidants in abolishing the ability of α₂M to inhibit proteinases. It is well known that reactive oxygen species such as superoxide anion, hydrogen peroxide, hydroxyl radical, and hypochlorite play an important role during acute and chronic inflammation (26, 27, 28, 29). In addition to neutralizing bacteria, these neutrophil-derived oxidants accelerate tissue destruction by acting either directly on cells, causing apoptosis and tissue necrosis, or indirectly by altering the proteinase-proteinase inhibitor balance (30). Hypochlorite, produced by the neutrophil H₂O₂-myeloperoxidase-Cl⁻ system, but not H₂O₂ or hydroxyl radical from metal-catalyzed oxidation, can abolish the ability of α₂M to inhibit proteinases at low micromolar concentrations (31, 32). The biologic concentration of hypochlorite during inflammation can be as high as millimolar (30). Reactions of hypochlorite with α₂M occur predominantly at methionine and tryptophan residues, although we have recently shown that lysine is a susceptible target of oxidation as well (32, 33). Hypochlorite oxidation of α₂M results in fragmentation of α₂M tetramers into dimers, whereas the effect of oxidation on α₂M* is currently unknown.

Since neutrophil-derived oxidants are presumed to be released concomitant with cytokines/growth factors, and increased concentrations of α₂M in tissue fluids have been demonstrated in a number of inflammatory diseases such as rheumatoid arthritis (RA) (34), pulmonary emphysema (35), pneumonia (36), and periodontitis (37), we hypothesize that hypochlorite oxidation may serve as a physiologically relevant mechanism that regulates the binding of cytokines and growth factors to α₂M.

In this study, we found that hypochlorite oxidation decreases the binding of α₂M and α₂M* to tissue repair growth factors such as β-NGF, PDGF-BB, TGF-β1, and TGF-β2. On the other hand, we found that hypochlorite oxidation enhances the binding of α₂M, but not that of α₂M*, to acute phase cytokines such as TNF-α, IL-2, and IL-6. Additional experiments using α₂M purified from human RA synovial fluid (RASF) indicate that this protein is significantly oxidized and that its binding to TGF-β is decreased. Given these findings, we propose that α₂M oxidation is a switch mechanism that reverses the cytokine/growth factor binding profile of α₂M, thus facilitating the transition from the early phase of inflammation, when tissue injury and destruction predominate, to the late phase, when tissue repair and remodeling are required.

¹²⁵I-labeled Bolten Hunter reagent and [methyl-³H]thymidine were purchased from New England Nuclear Life Science Products (Boston, MA). RPMI 1640, DMEM, l-glutamine, penicillin/streptomycin, HBSS, and FBS were purchased from Life Technologies (Gaithersburg, MD). HEPES, sodium hypochlorite, l-methionine, l-glycine, 5,5′-dithio-bis(2-nitrobenzoic acid), and EDTA were purchased from Sigma (St. Louis, MO). All other reagents were of the highest quality commercially available.

Human native α₂M was purified from plasma according to a previously described protocol (2). α₂M was at least 90% active against proteinases as determined by thioester titration using the 5,5′-dithio-bis(2-nitrobenzoic acid) assay (38). α₂M* was prepared as previously described (39). Both α₂M and α₂M* were free of endotoxin as determined by the Limulus amebocyte lysate test purchased from the Associates of Cap Cod (Falmouth, MA) performed according to the manufacturer’s suggested protocol (n = 2). Recombinant carrier-free human TGF-β1, TGF-β2, PDGF-BB, IL-6, β-NGF, and bFGF were purchased from R&D Systems (Minneapolis, MN). Recombinant carrier-free human TNF-α, IL-1β, and IL-2 were purchased from Genzyme Diagnostics (Cambridge, MA). All ¹²⁵I-labeled cytokines/growth factors were either labeled with [¹²⁵I]Bolten Hunter reagent according to the manufacturer’s recommended protocol or purchased from New England Nuclear Life Science Products. The specific radioactivity of the labeled ligands ranged from 500 Ci/mmol ([¹²⁵I]TNF-α) to 4000 Ci/mmol ([¹²⁵I]TGF-β1). Differences in specific radioactivity between commercially purchased proteins and our own preparations were estimated to be <20% (n = 4). Both labeled and unlabeled proteins were reconstituted with 0.1% BSA in PBS, pH 7.4, and stored at −20°C. ¹²⁵I-labeled cytokines/growth factors were used within 2 wk of labeling, and unlabeled proteins were stored at −20°C and used within 3 mo.

Oxidation of α₂M and α₂M* was performed as previously described with minor modifications (33). In brief, α₂M and α₂M* were incubated with sodium hypochlorite (0–100 μM) for 15 min at 37°C in PBS, pH 7.4. The sodium hypochlorite concentration was determined spectrophotometrically at 292.5 nm with an extinction coefficient of ε = 206 M⁻¹ cm⁻¹ at pH 7.4 (40). At the end of the incubation, 200 μM l-methionine was added to the mixture to quench residual oxidants.

¹²⁵I-labeled cytokines/growth factors (0.3–0.5 ng) were added to α₂M and α₂M* that were oxidized with 0 to 100 μM of hypochlorite and incubated for 2 h at 37°C. Following incubation, the mixture was loaded onto either native pore-limit gels or a nonreducing SDS gels to separate bound from unbound ¹²⁵I-labeled ligands. Following electrophoresis, gels were stained with Coomassie brilliant blue to verify equal loading of α₂M into each lane. Gels were subsequently dried and exposed to a PhosphorImager (Molecular Dynamics Sunnyvale, CA) plate for 16 h before the plate was developed and the bands were quantified. To eliminate the possible overexposure of the PhosphorImager plate, we performed control experiments to verify that the relative intensity of the bands on the PhosphorImager plate corresponded to the radioactivity detected by gamma counting of each band (n = 4). Nonspecific binding was determined by the binding of [¹²⁵I]ligand in the presence of a 1000-fold excess of unlabeled ligand. Specific noncovalent binding of [¹²⁵I]cytokines/growth factors to α₂M was determined by subtracting radioactivity associated with α₂M in nonreducing SDS gels from radioactivity associated with α₂M in native pore-limit gels. Nonspecific binding was approximately 0 to 30% and varied with each growth factor. At least three independent binding assays for each cytokine/growth factor and oxidized α₂M or oxidized α₂M* pairs were performed.

Native pore-limit gel electrophoresis was performed as described previously (2). In brief, 5 to 15% gradient polyacrylamide gels in 8.9 mM Tris, 8.9 mM boric acid, and 0.2 mM EDTA, pH 8.8, were made immediately before use. α₂M and oxidized α₂M alone or incubated with ¹²⁵I-labeled cytokines/growth factors (25 μl) in nonreducing, nondenaturing sample buffer were added to each lane and run for 3 h at 150 V. Subsequently, gels were stained with Coomassie brilliant blue and destained for 4 h in methanol/acetic acid. To ensure that α₂M-bound cytokines/growth factors did not dissociate during the destaining procedure, identical gels were autoradiographed without staining/destaining in some experiments. No difference was detected between gels that were stained and gels that were not stained. Nonreducing SDS-gel electrophoresis was performed as previously described (33).

CCL64 mink lung epithelial cells were obtained from American Type Culture Collection (Manassas, VA) and cultured in 150-cm² flasks in RPMI 1640 medium supplemented with 10% FBS, 2 mM l-glutamine, 25 mM HEPES, and 15 ml of 10,000 U/ml penicillin/streptomycin. Cell surface receptor ligand binding assays were performed as previously described (39). Cells were seeded into 24-well plates at 500,000 cells/well and allowed to adhere and grow until confluence (∼1–2 days) in a 5% CO₂ humidified incubator at 37°C. [¹²⁵I]TGF-β1 (0.5 ng) was then added into each well in the presence or the absence of the indicated concentrations of α₂M or oxidized α₂M in HBSS containing 5% BSA and 25 mM HEPES, pH 7.4, and incubated at 4°C for 16 h. In some experiments, receptor-associated protein (RAP; 6.65 μM), which inhibits binding of all ligands to LRP (41), was also added to determine the amount of oxidized α₂M/[¹²⁵I]TGF-β1 complex binding to the cell surface via the scavenger receptor LRP. Following incubation, unbound ligand was washed twice with ice-cold HBSS buffer and solubilized with 0.5 M NaOH/0.1% SDS for 4 h at 25°C before gamma counting using CliniGamma 1272 from LKB-Wallac (Turku, Finland). Total binding was determined by measuring the binding of [¹²⁵I]TGF-β1 in the absence of α₂M. Nonspecific binding was determined by measuring the binding in the presence of a 1000-fold molar excess of unlabeled TGF-β1 and range from 20 to 30% of total binding.

For TGF-β studies, we used CCL64 cells that were cultured as described above, except that 2% FBS was used during the experiment to minimize the interference by bovine macroglobulins to the assay. Cell proliferation assays were performed according to a standard protocol (14). In brief, 1 day before experimentation, cells were trypsinized and plated into 96-well tissue culture plates from Costar (Cambridge, MA) at 5000 cells/well and incubated overnight. On the day of the experiment, TGF-β at the indicated concentrations was added alone or in the presence of 1 mg/ml α₂M, oxidized α₂M, α₂M*, or oxidized α₂M* and incubated for 16 h at 37°C in a 5% humidified CO₂ incubator. Following incubation [methyl-³H]thymidine (0.5 μCi) was added to each well, and the plates were incubated for an additional 5 h. Cells were then trypsinized and harvested using a Skatron (Sterling, VA) cell harvester, and the cell-associated radioactivity was counted in a MINAXIβ liquid scintillation counter from Packard Instruments (Downers Grove, IL). As controls, CCL64 cells were incubated with α₂M, α₂M*, oxidized α₂M, and oxidized α₂M* in the absence of TGF-β and harvested in parallel with the experimental wells.

For bFGF assays, fetal bovine heart endothelial cells obtained from American Type Culture Collection were cultured in 75-cm² flasks in DMEM supplemented with 10% FBS, 50 ng/ml bFGF, 2 mM l-glutamine, 25 mM HEPES, and 15 ml of 10,000 U/ml penicillin/streptomycin until 80% confluent and then transferred to bFGF-deficient medium for 48 h to reach quiescence. One day before experimentation, quiescent cells were trypsinized from flasks, plated into 96-well plates at 2000 cells/well in bFGF-deficient medium, and allowed to adhere. On the day of experimentation, cell medium was replaced with α₂M (either oxidized or nonoxidized) alone (1 mg/ml) or α₂M and various concentrations of bFGF-containing medium and allowed to incubate for 48 h at 37°C. Following incubation, cells were pulsed with [methyl-³H]thymidine (0.5 μCi) for an additional 5 h and then harvested for scintillation counting.

For TNF-α assays, murine fibrosarcoma (WEHI 13VAR) cells that are highly sensitive to TNF-α-induced cell death were used. These cells were cultured under the same conditions as CCL64 cells described above. Assays for TNF-α-induced cell death was performed as described previously (42) with modifications. Cells were cultured in 75-cm² flasks and plated into 96-well plates 2 days before experimentation at 10,000 cells/well. On the day of experimentation, α₂M alone (1 mg/ml) or α₂M and various concentrations of TNF-α were added to each well in the presence of 10 μg/ml of cycloheximide and allowed to incubate for 24 h at 37°C. Following incubation, cell viability was measured using Celltiter 96 (Promega, Madison, WI) according to the manufacturer’s suggested protocol and verified by cell counting.

Synovial fluids from six patients fulfilling the American College of Rheumatology’s revised criteria for the classification of RA (43) and five patients fulfilling criteria for OA were obtained from the Rheumatology Clinic of Duke University Medical Center (Durham, NC). Informed consent was obtained in each case for the use of these fluids. Synovial fluids were aspirated as a standard procedure to drain inflamed joint effusions. Fluids were anticoagulated with 5 mM EDTA or 10 U/ml heparin and frozen immediately at −70°C. Before analysis synovial fluids were thawed and treated with a mixture of proteinase inhibitors to give a final concentration of 2 mM PMSF, 2 mM 3,4-dichloroisocoumarin, 5 mM 1,10-phenanthroline, and 2 μM E-64. Cell debris was removed by centrifugation.

Polyclonal Abs against human α₂M were made in New Zealand White rabbits and isolated using α₂M-Sepharose prepared by coupling α₂M to cyanogen bromide-activated Sepharose purchased from Pharmacia (Uppsala, Sweden). Purified Ab was then coupled to cyanogen bromide-activated Sepharose according to the manufacturer’s suggested protocol and incubated for 2 h with RA and OA synovial fluid prepared as described above and subsequently diluted threefold with PBS, pH 7.4. Following incubation, α₂M bound to anti-α₂M IgG-Sepharose was eluted with 0.1 M Tris/0.5 M NaCl, pH 10.8, and immediately readjusted to pH 7.4. As controls, α₂M from healthy human donor plasma (n = 6) was isolated using the same procedure as that used for synovial fluid α₂M. The α₂M protein concentration was determined both spectrophotometrically using A₂₈₀ (1%; 1 cm) = 8.93 (44) and with bicinchoninic acid protein assay (Pierce, Rockford, IL), and the purity of the protein was verified by gel electrophoresis and Western blotting.

Measurement of the extent of oxidation in synovial fluid α₂M compared with plasma α₂M was performed using 2,4-dinitrophenylhydrazine (DNPH) derivatization of protein carbonyls as previously described (45) with modifications. One hundred micrograms of protein in 800 μl of PBS was added to 200 μl of 10 mM DNPH in 2 M HCl and incubated at 25°C for 1 h. Following incubation, proteins were precipitated with 150 μl 70% TCA and placed on ice for 10 min. Proteins were then centrifuged at 800 × g for 10 min. Protein pellets were washed with ethyl acetate/ethanol (1/1, v/v), and the centrifugation/washing process was repeated two more times before final solubilization in 6 M guanidine-HCl, pH 7.4. Each protein sample was then scanned from 200 to 500 nm using a Beckman DU-640 spectrophotometer (Arlington Heights, IL), and the quantity of protein carbonyl was calculated using an extinction coefficient for dinitropheynlhydrazone of 22,000 M⁻¹ cm⁻¹ (45). To control for the possible loss of proteins during washing steps, all spectrophotometric readings were adjusted to an identical A₂₈₀. Background absorption was determined by experiments performed in the absence of DNPH, and this value was used as the blank for each reading.

In studies of [¹²⁵I]cytokine/growth factor binding to oxidized α₂M and oxidized α₂M*, the K_d values were determined by least squares curve fitting using the SYSTAT program (version 5.04, Systat, Evanston, IL). We chose to determine the K_d values using this method because it gave more consistent data (r² >0.95 for all calculations) than those derived from Scatchard plots. The K_d values determined from Scatchard analysis, however, were within the SE. In cell proliferation assays the EC₅₀ was also determined using SYSTAT. The growth factor effects (percentages) on cell proliferation was determined as was previously described (46): % effect = EC₅₀ (control)/EC₅₀ (treatment) × 100, where EC₅₀ (control) represents half-maximum growth factor effects in the absence of α₂M, and EC₅₀ (treatment) represents half-maximum growth factor effects in the presence of α₂M or oxidized α₂M.

The effects of various physiologically relevant concentrations of hypochlorite on the structures of α₂M and α₂M* were determined. As shown in Figure 1, α₂M oxidation at >50 μM hypochlorite concentrations resulted in significantly faster electrophoretic migration. The faster migratory position corresponded to α₂M dimers as previously described (32) and not α₂M*, which migrated only slightly faster (by 2 mm) than α₂M tetramers in the same gel system (data not shown). This effect was absent with oxidation of α₂M*, which remained as intact tetramers at the concentrations shown. A slight decrease in staining by Coomassie brilliant blue was evident in 100 μM hypochlorite-oxidized α₂M and α₂M*. Decreased staining of chlorinated α₂M has been previously reported (31, 33). Additional verifications using Western blotting analysis and autoradiography of radiolabeled proteins showed equivalent protein quantity in all lanes (data not shown).

To determine whether oxidation affects the binding of α₂M to cytokines/growth factors, we performed in vitro binding experiments using ¹²⁵I-labeled TNF-α, IL-2, IL-6, bFGF, β-NGF, PDGF-BB, TGF-β1, and TGF-β2. These labeled ligands were incubated for 2 h with 20 μl of 0.25 mg/ml α₂M oxidized at the indicated concentrations. Unbound ligands were separated by native pore-limit electrophoresis. Figure 2 shows the α₂M-associated radioactivity for each labeled cytokine/growth factor binding assay with α₂M. As can be seen, the affinity of α₂M for TNF-α, IL-2, IL-6, and bFGF increases with oxidation. It appears that the increase in cytokine/growth factor binding is not dependent on the tetramer to dimer transition, since some ligands, such as IL-2 and IL-6, show increased binding to α₂M even while it is still in the tetrameric state. In contrast, the binding of oxidized α₂M to β-NGF, TGF-β1, TGF-β2, and, to a lesser extent, PDGF-BB decreases with increasing oxidant concentrations. The decrease in binding is more dramatic for TGF-β2 than β1, possibly reflecting the higher affinity of α₂M binding to TGF-β2 than TGF-β1 (17). It is important to point out that the specific radioactivity is different for each cytokine, and therefore the binding intensities cannot be compared between cytokines. Since methionine was added to quench residual oxidants, control experiments were performed to determine whether methionine alone or methionine sulfoxide, the product of the reaction between methionine and hypochlorite, can affect the binding of cytokine/growth factors to α₂M. ¹²⁵I-labeled ligands were incubated with α₂M in the presence or the absence of methionine or methionine/hypochlorite mixture. In three independent experiments with [¹²⁵I]TGF-β1, PDGF-BB, and TNF-α, there was no difference in the amount of cytokine/growth factor bound to α₂M (data not shown).

Given that oxidation appears to alter cytokine/growth factor binding to α₂M and that α₂M-proteinase complexes may represent a significant portion of the α₂M in inflammatory fluids (30, 47, 48), we performed α₂M* binding experiments using the same ¹²⁵I-labeled ligands as those in Figure 2. Figure 3 shows the changes in α₂M*-associated ¹²⁵I-labeled cytokine/growth factor binding with increasing oxidant concentrations. In contrast to binding to oxidized α₂M, all the labeled ligands showed either a decrease in affinity for oxidized α₂M* or no effect.

Of the eight cytokines/growth factors studied, the most significant changes in binding were observed with TNF-α, PDGF-BB, β-NGF, and TGF-β. To further quantify the oxidation-induced changes in the binding affinity of α₂M for these cytokines/growth factors, we performed concentration-dependent binding assays. Figure 4,A shows the results of [¹²⁵I]TNF-α binding to oxidized α₂M and oxidized α₂M*. Figure 4,B shows the results of [¹²⁵I]PDGF-BB binding. Similar experiments were performed for [¹²⁵I]TGF-β1 (Fig. 4,C) and for [¹²⁵I]β-NGF (Fig. 4,D). A summary of the binding K_d values for these experiments is presented in Table I. As shown in the table, the affinity of α₂M for binding to TNF-α increased by 5.3-fold with oxidation. The affinities for TGF-β1 and β-NGF decreased by 3.1- and 5.9-fold, respectively. With oxidized α₂M*, an 8.5- and 13.1-fold decreases in affinity were observed for binding to PDGF-BB and TGF-β1, respectively. These binding K_d values are in close agreement with the results of previous studies using a combined protein cross-linking/electrophoresis assay (17) despite the report that 10 to 20% dissociation of radiolabeled ligand from α₂M is possible during 2-h gel electrophoresis in the absence of cross-linking. We have attempted the combined protein cross-linking/electrophoresis method for our studies but have found a large decrease in the cross-linking efficiency for oxidized proteins compared with nonoxidized proteins, possibly because the lysine residues that are necessary for cross-linking using bis-(sulfosuccinimidyl) suberate have been modified by oxidation (33). Since the binding K_d values between our studies and the previous studies are similar, we assume that the amount of radioligand dissociation during gel electrophoresis does not significantly alter the measurement of K_d values.

The binding of cytokines/growth factors to cell surface receptors decreases significantly in the presence of α₂M or α₂M* (10, 13, 14). This effect is directly related to the affinities of α₂M and α₂M* for binding to cytokines/growth factors. Given that oxidation alters the binding affinities of α₂M and α₂M* to these growth factors, we postulated that this may result in altered growth factor binding to cells. We chose [¹²⁵I]TGF-β as the model ligand for these binding experiments because TGF-β has high affinities for binding to α₂M and α₂M*, and these interactions may play significant roles in inflammation in vivo (7, 8). Figure 5,A shows the results of [¹²⁵I]TGF-β1 binding to CCL64 cells in the presence of α₂M or oxidized α₂M. As shown in Figure 5, a significant decrease in [¹²⁵I]TGF-β1 binding was observed in the presence of α₂M (IC₅₀ = 130 nM). This effect was reduced by 4.1-fold in the presence of oxidized α₂M (IC₅₀ = 530 nM). Similar results were obtained with α₂M* and oxidized α₂M*, where the IC₅₀ increased from approximately 10 to 90 nM (Fig. 5 B).

The binding of TGF-β1 to cell surface receptors is higher in the presence of oxidized α₂M compared with nonoxidized protein (Fig. 5,A). This is most likely due to the decreased binding affinity between α₂M and TGF-β as a result of oxidation. There is another explanation, however, that must be considered. We have recently shown that while unmodified α₂M does not bind to LRP, oxidation results in the exposure of its receptor recognition sites for binding to LRP (33). It is possible that the increase in TGF-β binding to the cell surface in the presence of oxidized α₂M is the result of the formation of TGF-β-oxidized α₂M/LRP complexes on the cell surface in addition to TGF-β/TGF-β receptor complexes. This would be consistent with the hypothesis that receptor-recognized forms of α₂M may serve as a vehicle that carries growth factors to the cell surface for delivery to growth factor receptors (9, 16). To investigate whether LRP may be involved in binding to oxidized α₂M/TGF-β complexes, we incubated CCL64 cells with oxidized α₂M and [¹²⁵I]TGF-β1 in the presence or the absence of a 50-fold molar excess of RAP, which competes for the binding of all ligands to LRP (41) (Fig. 6). As shown in Figure 6, RAP does not significantly alter the cell surface binding of [¹²⁵I]TGF-β1 in the presence of oxidized α₂M. RAP alone also has no effect on [¹²⁵I]TGF-β1 binding.

The biologic activities of various cytokines/growth factors are decreased in the presence of α₂M and α₂M* (13, 14). To determine whether the altered binding interaction between oxidized α₂M and α₂M* with these cytokines/growth factors can translate into altered cytokine/growth factor bioactivity, we performed in vitro bioassays to test the activities of these cytokines/growth factors in the presence of oxidized and nonoxidized α₂M and α₂M*. Figure 7,A shows that in the presence of oxidized α₂M, TNF-α activity is inhibited by 66% compared with nonoxidized α₂M, which had no effect. In the presence of α₂M*, however, no significant difference in inhibition was observed between oxidized and nonoxidized protein, consistent with the in vitro binding data. Figure 7,B shows that the biologic activity of TGF-β is significantly decreased in the presence of α₂M (62%) and α₂M* (82%). In the presence of oxidized proteins, however, this inhibition is abolished. Figure 7,C shows the effects of α₂M and oxidized α₂M on bFGF bioactivity. Interestingly, there is no difference in inhibition between α₂M and oxidized α₂M despite the apparent increase in oxidized α₂M binding to bFGF in vitro (Fig. 2). For α₂M*, however, a significant decrease in inhibition of bFGF bioactivity is observed with the oxidized protein compared with the nonoxidized protein.

Given that oxidation regulates α₂M-cytokine/growth factor binding in vitro, we investigated the in vivo relevance of this process by asking whether oxidized α₂M is present in the tissue fluids of patients with acute inflammatory diseases such as RA. This hypothesis seems likely given that oxidized proteins have been demonstrated in RASF (49), and increased levels of α₂M are present in this fluid during inflammation (34, 50). α₂M was isolated from the knee joint synovial fluid of six patients with active RA and five patients with OA by affinity chromatography using rabbit anti-human α₂M polyclonal Ab, and the results were verified by Western blotting. As controls, α₂M was also isolated from the plasma of six healthy volunteers by the same method. To determine whether the rheumatoid synovial fluid α₂M is oxidized, we performed DNPH derivatization, which measures the carbonyl content of proteins due to oxidation. Figure 8,A shows the mean protein carbonyl content of plasma α₂M, plasma α₂M oxidized in vitro with 100 μM hypochlorite, RASF α₂M, and osteoarthritis synovial fluid (OASF) α₂M. As shown in Figure 8, the level of protein carbonyl was approximately sevenfold higher for RASF α₂M compared with plasma α₂M (p < 0.005). This level is comparable to the level of protein carbonyl generated by oxidizing plasma α₂M in vitro with 100 μM hypochlorite.

Having shown that RASF α₂M is significantly more oxidized compared with plasma and OASF α₂M, we determined whether this protein has decreased binding to TGF-β as well. Figure 8 B shows the binding of [¹²⁵I]TGF-β to plasma α₂M, plasma α₂M*, plasma α₂M oxidized in vitro with 100 μM hypochlorite, RASF α₂M, and OASF α₂M. As shown in this figure, a 26% decrease in binding was observed for RASF α₂M compared with plasma α₂M (p < 0.05). This decrease is even more significant considering that a large portion of the RASF α₂M is actually proteinase bound and is expected to bind TGF-β as well as α₂M* (i.e., compared with plasma α₂M*, RASF α₂M shows a 55% decrease in TGF-β binding (p < 0.01)). To verify that RASF α₂M does not carry natural TGF-β, which may explain the decrease in its ability to bind to [¹²⁵I]TGF-β, additional TGF-β bioassays were performed with acidified RASF α₂M, since acidification releases noncovalently bound TGF-β from α₂M (21). In two independent experiments, we found that acidified RASF α₂M has no TGF-β activity (data not shown).

In the present study, we have identified a physiologic oxidant that can potentially regulate the interaction between α₂M and inflammatory cytokines/growth factors in vivo. The reaction between α₂M and hypochlorite, a powerful oxidant released by the neutrophil H₂O₂-myeloperoxidase-Cl⁻ system, results in fragmentation of the α₂M tetramer into dimers with an enhanced binding capacity toward acute inflammatory cytokines such as TNF-α, IL-2, and IL-6. Binding of β-NGF, PDGF-BB, TGF-β1, and TGF-β2 to oxidized α₂M, on the other hand, is significantly decreased. We investigated the potential biologic relevance of this change by demonstrating that the ability of α₂M and α₂M* to inhibit the bioactivity of TGF-β and bFGF is abolished by oxidation. To determine whether α₂M in inflammatory tissue is oxidized in vivo, we isolated α₂M from the synovial fluid of patients with active RA and found that this protein is significantly more oxidized than plasma and OASF α₂M. The binding of TGF-β to this protein is decreased as well. Taken together, it appears that oxidation may play an important role in regulating inflammatory cytokine/growth factor functions by altering the affinity of extracellular binding proteins such as α₂M for these signaling molecules.

Reactive oxygen species produced by activated neutrophils and macrophages have been recognized as a hallmark of inflammation (26, 27, 30). Interactions between oxidants and proteins lead to modification of amino acid residues such as cysteine, methionine, histidine, tryptophan, tyrosine, and lysine and increase the carbonyl content and negative surface charge of proteins (26, 27). A large body of evidence has supported the role of oxidants in disease pathogenesis. Oxidation of low density lipoprotein leads to foam cell formation and atherosclerosis, while cigarette smoking generates inhaled oxidants that induce emphysema (28, 51). Despite a long standing interest in the role of oxidants in disease, only recently has there been in-depth investigation of the molecular mechanisms involved in oxidant-mediated tissue injury. Oxidants can interact directly with cells, causing activation of endogenous oxidative stress-mediated signaling pathways involving extracellular-receptor mediated kinase, c-jun N-terminal-activated kinase/stress-activated protein kinase, and/or NF-κβ, resulting in either cell proliferation or apoptosis depending on the cell type (52, 53, 54). Activation of the NF-κβ pathway induces the production of TNF-α, IL-1β, IL-2, and IL-6 (54), which can further activate the inflammatory cascade. Interaction of oxidants with proteinase inhibitors such as α₁-proteinase inhibitor, secretory leukocyte proteinase inhibitor, and α₂M destroys the inhibitory activity of these antiproteinases while enhancing the proteolytic activity of latent collagenase (30). This may contribute to tissue destruction in adult respiratory distress syndrome, RA, pulmonary emphysema, and glomerulonephritis. On the other hand, oxidants may also be involved in tissue repair mechanisms by stimulating cell proliferation and collagen deposition either alone or in conjunction with growth factors such as PDGF, TGF-β, bFGF, and β-NGF (54).

Evidence has accumulated in recent years that α₂M, in addition to its ability to inhibit proteinases, can bind to various cytokines and growth factors with high affinity. Binding of α₂M to these molecules generally results in neutralization of their activities on various different cell types (6, 7). It has been hypothesized that the physiologic role of α₂M in binding to these growth factors is to down-regulate the effects of these extremely potent growth factors by inhibiting their interactions with cell surface receptors and by internalizing bound growth factors via LRP. However, the available data suggest that a mechanism must exist that allows growth factors to be released from α₂M during inflammation. It seems paradoxical that during inflammation when the concentration of α₂M in tissue fluid increases dramatically due to increased vascular permeability and local synthesis, the concentration of growth factors increases as well. In addition, α₂M is present in inflammatory fluids at micromolar concentrations, whereas the affinity of α₂M for growth factors is on the order of nanomolar (10, 17, 18, 19, 20). At this concentration most, if not all, growth factors should be bound to α₂M, yet a large increase in growth factor level and activity has been detected in inflammatory fluids (24, 55, 56). Moreover, despite the finding that TGF-β and PDGF are both predominantly bound to α₂M in the plasma, to our knowledge no data has demonstrated the isolation of TGF-β or PDGF-BB/α₂M complexes from inflammatory lesions. We hypothesize that hypochlorite oxidation, which selectively and potently inactivates α₂M inhibition of proteinases in vitro, might serve as a mechanism that abolishes the binding of growth factors to α₂M in vivo.

Our initial survey of the binding of eight different cytokines and growth factors to oxidized α₂M reveals an interesting and unexpected finding. The binding of acute inflammatory mediators such as TNF-α, IL-2, and IL-6 to α₂M appears to increase significantly with oxidation. In the absence of oxidation, these cytokines all bind to α₂M with micromolar affinity; however, with oxidation, the affinity increases approximately fivefold to nanomolar affinity. The binding K_d value observed between oxidized α₂M and TNF-α (340 nM) is similar to the binding K_d values of soluble growth factor receptors to growth factors (57), highlighting the potential importance of this interaction.

The mechanism responsible for this increase is unknown. Binding of cytokines/growth factors to α₂M involves a number of mechanisms including, but not limited to, noncovalent binding, trapping within the α₂M cage, covalent incorporation via the glutamine side chain of the thioester, or disulfide cross-linking (7). Interesting differences in cytokine binding to α₂M and various α₂M/proteinase complexes have been demonstrated (9, 18, 58, 59). Our cytokine binding assay measures only noncovalent association with α₂M and oxidized α₂M. It is possible that changes in other modes of binding can influence the overall effects observed in comparing α₂M and oxidized α₂M. We did not, however, observe any difference in covalent association with oxidized α₂M relative to α₂M, and it appears unlikely that noncovalent trapping could be responsible for this increase given that oxidation fragments α₂M tetramers into dimers. Since the treatment of α₂M with methylamine or plasmin results in increased binding to TNF-α (9), it is possible that oxidation can induce a similar structural change; however, neither earlier studies of hypochlorite oxidation of α₂M (31, 32) nor our own studies (33) have demonstrated cleavage of the thioester bond and the major conformational change that resembles α₂M* in oxidized α₂M. Mechanisms that may be responsible for the increased TNF-α, IL-2, and IL-6 binding to oxidized α₂M include increased electrostatic interaction, exposure of a previously buried cytokine binding site, increased affinity to an existing cytokine binding site, and increased access to this site. Given that very little is currently known about the exact mechanism of noncovalent cytokine/growth factor binding to α₂M, additional studies will be necessary to determine which of these mechanisms is likely to play a role.

In contrast to the acute inflammatory mediators, our studies show that the binding of growth factors, such as β-NGF, PDGF-BB, TGF-β1, and TGF-β2, to oxidized α₂M is significantly decreased. The affinity of PDGF-BB decreased approximately 9-fold, while the affinity of TGF-β2 decreased 13-fold. The mechanism of binding by these growth factors to α₂M and α₂M* is primarily noncovalent, and cross-competition between PDGF-BB and TGF-β1 has been demonstrated (60), suggesting that these growth factors may bind to the same site on α₂M. Oxidation can alter the overall structure of α₂M in such a way as to destroy the three-dimensional conformation required for growth factor binding. This hypothesis is consistent with our recent finding that oxidized α₂M is partially unfolded (33). It is also possible that oxidation selectively modifies a site on α₂M that is involved in binding to growth factors. Supporting this hypothesis is a recent study demonstrating that a polypeptide corresponding to the N-terminal sequence of α₂M binds to TGF-β (61). Identification of the amino acid residues in this sequence susceptible to oxidation may shed light on the nature of TGF-β binding to α₂M.

The selective regulation of α₂M cytokine/growth factor binding by oxidation offers an intriguing interpretation for the potential physiologic significance of this interaction. During acute inflammation, activated neutrophils release proteinases and oxidants as part of the endogenous defense mechanisms against foreign organisms. As a result, susceptible extracellular proteins such as α₁-proteinase inhibitor and α₂M may also be oxidized (30). Although the effects of oxidation on cytokines/growth factors have not been reported, nor have oxidized cytokines/growth factors been isolated from inflammatory fluids, the possibility exists that oxidation may directly alter cytokine/growth factor functions. The finding that oxidized α₂M binds with greater affinity to acute phase cytokines such as TNF-α, IL-2, and IL-6 suggests that oxidized α₂M may play an anti-inflammatory role by inhibiting the progression of the proinflammatory cascade induced by these molecules. In this regard, the decreased binding of oxidized α₂M and α₂M* to TGF-β, PDGF-BB, and β-NGF, all of which have been considered as inflammatory cytokines/growth factors, argues against the anti-inflammatory role of oxidized α₂M. Inflammation, however, is complex process involving multiple different phases (62). Growth factors such as β-NGF, bFGF, PDGF-BB, and TGF-β have generally been considered mediators of tissue repair processes including neurite out-growth, angiogenesis, fibroblast proliferation, and collagen deposition. It is possible that oxidation of α₂M may play a different role at different stages of the inflammatory process. In this regard, it is worth mentioning that α₂M oxidation, which results in greater in vitro binding affinity to bFGF, has no differential effect on bFGF stimulation of endothelial cell proliferation (Fig. 7C). Oxidation of α₂M*, however, results in decreased binding affinity to bFGF as well as decreased ability to inhibit the biologic activity of bFGF. Taken together, we offer the hypothesis that oxidation may facilitate the transition from tissue damage to repair during inflammation by enhancing the ability of α₂M to bind to proinflammatory molecules while decreasing its ability to bind to tissue repair molecules. This would suggest that in the absence of oxidation (as in patients with chronic granulomatous disease) chronic and persistent inflammation may occur due to poor transition from acute inflammation to resolution. It should be pointed out that this hypothesis may apply only to the inflammatory process once neutrophil activation has occurred, since chemokines, which play important roles in leukocyte chemotaxis and transmigration into inflammatory sites, were not investigated in this study, and HOCl is known to be released by neutrophils only after they have become activated. Additional in vivo studies will be necessary to determine whether this hypothesis is correct. We herein offer some additional evidence that suggests our hypothesis is possible.

Using RA as a model of an inflammatory disease with a prominent tissue repair phase, we isolated synovial fluid α₂M from the knee joints of six patients with this disease. RASF α₂M from these patients was sevenfold more oxidized than control plasma α₂M and threefold more oxidized than α₂M isolated from OASF, which is generally considered a noninflammatory fluid. The binding of RASF α₂M to TGF-β1 was significantly decreased as well. Protein oxidation has been demonstrated in patients with RA, adult respiratory disease syndrome, atherosclerosis, bronchitis, and a number of other inflammatory diseases (49, 63, 64, 65). Oxidation of α₂M has been suggested to be the mechanism involved in enhanced tissue degradation in RA (66). To date, however, no study has isolated and characterized oxidized α₂M in disease tissues and fluids. Our studies with RASF α₂M are the first to demonstrate the presence of oxidized α₂M in human disease tissue. That RASF α₂M, moreover, has significantly decreased ability to bind to TGF-β1 suggests that oxidation may play a major role in regulating the functions of extracellular cytokine/growth factor binding proteins.

We thank Drs. George Cianciolo and Maureane Hoffman for their comments on the manuscript, and Marie Thomas for her assistance with preparation of the manuscript.