Classification of a2-Macroglobulin-Cytokine Interactions Based on Affinity of Noncovalent Association in Solution under Apparent Equilibrium Conditions*

solutions which have been used in other investigations to stop B S 3 reactions (0.3 M Tris-HC1, pH 7.4 or pH 8.5; 0.3 M lysine, 60 mM Tris-HC1, pH 7.4 or 8.5; 0.3 M glycine, 60 mM Tris-HC1, pH 7.4 or 8.5) were not instantaneous inhibitors of the cross-linking reaction. Analysis of the Results of BS3 Cross-linking Experiments-Since BS3 was added at high concentration, azM-cytokine complex was cross-linked according to pseudo-first order kinetics; the fraction of noncovalent a2M-cytokine complex cross-linked (2) was independent of the concentration of azM-cytokine complex. Under these conditions, experimentally detected (BS3-stabilized) noncovalent azM-cy- tokine complex (AC,) is related to AC by the following relationship.

Human a2-macroglobulin (a2M)l is a tetrameric glycoprotein (MI-718,000) and an inhibitor of proteinases from all of the major classes (1)(2)(3). The four identical subunits in azM form two proteinase binding sites. Reaction of a2M with proteinase is initiated when the proteinase cleaves 1 of approximately 12 sensitive peptide bonds in the a2M bait region, which is located near the center of each subunit (1,4). Bait region cleavage initiates a major conformational change in a2M, which efficiently and irreversibly traps the reacting proteinase without requiring covalent bond formation (5)(6)(7)(8)(9).
After reaction with proteinase or amine, a2M is recognized by the cellular receptor, a2-macroglobulin receptor/low density lipoprotein receptor-related protein (LRP) (14,15). LRP is present on the surfaces of many different cell types (16) and responsible for the rapid plasma clearance of conformationally transformed a2M (17,18). The native (unreacted) conformation of azM demonstrates no affinity for LRP (17).
1533 Binding to a2-Macroglobulin enhances cytokine binding (22,31,46). In cell culture, azM may inhibit (22,25) or promote cytokine activity (29,48). For example, the mitogenic effect of TGF-P1 in smooth muscle cell cultures is enhanced up to 20-fold by the simultaneous addition of azM-methylamine (48). Cytokines that are bound to conformation~ly transformed azM are targeted to cells expressing LRP (21); this process may mediate cytokine clearance or regulate cytokine activity.
At least three different mechanisms are recognized whereby a2M binds cytokines (46,47). First, cytokines may bind noncovalently and reversibly to azM. This type of binding predominates in the extensively studied TGF-Pl/a*M interaction. Second, cytokines may bind covalently to azM by thioldisulfide exchange. This reaction principally involves the transformed conformation of aZM, since free Cys residues are generated by thiol ester cleavage. The third mechanism is somewhat more complex, since it requires a temporal relationship between cuzM binding of proteinase and cytokine. When azM is cleaved by an attacking proteinase, the a2M thiol esters become available for reaction with nucleophilic groups in other proteins (in addition to the proteinase) for a short period of time (13). Cytokine Lys residues can function as nucleophiles and cleave the thiol esters, forming covalent linkages with thiol ester glutamyl residues (42). The cytokine must be present at the instant of proteinase reaction with aZM, since the thiol esters are rapidly hydrolyzed by water in the absence of alternate nucleophiles (13).
Due to the diverse assortment of techniques used previously to study cytokine binding to aZM, it is difficult to assess the significance of each reported interaction. Many previous investigations have included experiments which only detected covalent binding. In these studies, the great majority of the azM-cytokine complex may have been unrecognized. In other studies, noncovalent and covalent binding of cytokines to azM was demon st rat^; however, the experiments did not provide quantitative results which could be interpreted in terms of affinity. For example, gel filtration chromatography and nondenaturing PAGE have been commonly used to study cytokine binding to a2M; results obtained with these techniques may be significantly affected by the rate of dissociation of the azM-cytokine complex.
Without knowledge of binding affinities, different (YzMcytokine interactions cannot be compared. In addition, it is not possible to predict whether azM might function as a significant cytokine binder in the presence of other macromolecules which demonstrate affinity for the same cytokine, such as cytokine-specific cellular receptors, proteoglycans, and extracellular matrix proteins. To address this problem, our laboratory recently developed an immobilization system to study cytokine binding to aZM under equilibrium conditions (23); however, this system cannot be used to detect certain low-affinity (>0.3 p~) interactions (31). In addition, limitations were identified in the use of immobilized azM to determine equilibrium dissociation constants (23).
In the present investigation, we used a new method to analyze noncovalent binding of seven cytokines to native cvzM and cuzM-methylamine under apparent equilibrium conditions. Solutions containing azM and cytokine were pulseexposed to the cross-linking agent, ~~(sulfosuccinimidyl) suberate (BS3) to stabilize a fraction of the noncovalent ~z Mcytokine complex before performing SDS-PAGE. The efficiency of BS3 cross-linking and the extent of covalent binding were accounted for. Then, using a one-site binding model, equilibrium dissociation constants were determined. In previous studies, it has been suggested that each of the seven cytokines studied here form complexes with azM, which may be physiologically significant. Our results show that the seven cytokines vary in affinity for azM by at least 2 orders of magnitude.

MATERIALS AND METHODS
~e~e~s -M e t h y l a m i n e hydrochloride, chloramine T, 1,4-dithiothreitol, and bovine serum albumin (BSA) were purchased from Sigma. BS3 was purchased from Pierce Chemical Co. Na' "1 was purchased from Amersham Corp. ENZ~MOBEADS were purchased from Bio-Rad. Immulon 2 microtiter plates were from Dynatech Laboratories (~h~t i l l y , VA). Tween 20 (enzyme grade) was from Fisher.
a&"azM was purified from fresh human plasma by the method of Imber and Pizzo (49) and stored at -20 "C in 40% glycerol. The concentration of a2M was determined by absorbance at 280 nm, using an of 8.93 (3). apM-methylamine was prepared by dialyzing apM against 200 mM methylamine HC1 in 50 mM Tris-HC1, pH 8.2, for 12 h at 22 "C and then extensively against 20 mM sodium phosphate, 150 mM NaC1, pH 7.4 (PBS) at 4 "C. Reaction of native apM with methylamine was confirmed by loss of trypsin binding activity (greater than 96%).
Cytokines-Recombinant human TGF-fi1 was purchased from R&D Systems (Minneapolis, MN). TGF-fi1 was also purified from human platelets by the method of Assoian (50). TGF-fi1 was radioiodinated as described previously (21)  A Model of Cytakine Binding to azM-The binding of cytokines to azM in a system which contains no free proteinase (apM not actively undergoing conformational change) can be described as follows.
A is unbound aZM; C is free (unbound) cytokine; AC is reversibly associated (noncovalent) apM-cytokine complex; and AC* is irreversibly associated (covalent) a,M-cytokine complex. If the rate of conversion of AC to AC* (kz) is slow compared with k-' (rapid equilibrium assumption), then the dissociation constant, KD, for noncovalent binding of 1251-cytokine to a2M may be expressed as: Assumptions in this model include the following. 1) Each azM tetramer has one cytokine binding site. 2) Noncovalent association of lZ6I-cytokine with the azM binding site is adequately described by a single KD. If in fact there are two or four independent and equivalent cytokine binding sites per aZM tetramer instead of one (which is plausible due to the symmetry in the azM structure) then the actual KB for the interaction of cytokine with each individual binding site is 2-or 4-fold higher than the values reported here.
~ in HzO) was added to a final concentration of 5 mM. HzO alone was added to identical control incubations. Each tube was incubated for 60 s at 22 "C, unless otherwise specified (see below). To stop the cross-linking reaction, solutions were acidified to pH 2.5-3.0 with HCl (52). Samples were then denatured in 2.0% SDS for 30 min at 37 'C. Finally, 100 mM Tris-HC1 and 10% glycerol (final concentrations) were added to each sample, and SDS-PAGE was performed (no reductant). Dried gels were subjected to autoradiography. The amount of 'z51-cytokine in each band was determined by cutting the gel into sections and determining the radioactivity in each section using a y-counter. Recovery of radioactivity in the gels was typically between 75 and 90%. Free and azM-associated '251-cytokine were recovered in the gels equivalently.
In the absence of BS3, '251-cytokine which was covalently associated with azM comigrated with the high molecular mass azM bands by SDS-PAGE. When BS3 was added, additional '251-cytokine-azM complex was detected. The increase in binding was presumed to represent a fraction of the noncovalent aZM-cytokine complex. The fraction is determined by the cross-linking efficiency of BS3 ( z ) , which is a distinct value for each cytokine, a2M conformation, and BS3 crosslinking time (0 < z 4 ) . When SDS-PAGE was performed using reductant-treated samples (no BS3), binding of cytokines to azM was not observed. Therefore, the covalent binding observed in the absence of BS3 probably represented disulfide-stabilized complexes.
The effectiveness of the procedure for terminating the cross-linking reaction (acidification) was tested by adding 15 mM HCl and BS3 simultaneously to solutions of azM and '251-TGF-@l that had reached apparent equilibrium (see below). Cross-linking was not observed by SDS-PAGE. Analysis of the Results of BS3 Cross-linking Experiments-Since BS3 was added at high concentration, azM-cytokine complex was cross-linked according to pseudo-first order kinetics; the fraction of noncovalent a2M-cytokine complex cross-linked (2) was independent of the concentration of azM-cytokine complex. Under these conditions, experimentally detected (BS3-stabilized) noncovalent azM-cytokine complex (AC,) is related to AC by the following relationship.
AC, was determined from the radioactivity in gel slices and corrected for the presence of AC* by subtracting covalent binding detected in the absence of BS3. Experimentally detected free cytokine (C,) included free cytokine and cytokine that was bound to azM but not cross-linked by BS3.
By substituting the expressions for AC. and C, into Equation 2, a linear relationship in the form y = m X + b was derived.
(Eq. 5) azM-methylamine. The K D and z values were determined for each BS3 exposure time.
TGF-pZ Binding to Immobilized azM-'251-TGF-f12 binding to immobilized aZM was examined according to the method of Webb et al (23). Briefly, azM-methylamine was incubated in each well of a 96well microtiter plate so that 90 fmol of the azM-methylamine became immobilized. The wells were then blocked with Tween 20 (0.1% u/u). '251-TGF-(32 (0.2 nM) in PBS with 15 p~ BSA was equilibrated in the microtiter wells with immobilized a2M-methylamine at 22 "C. Different concentrations of native a2M or azM-methylamine (0.1 nM to 0.25 p~) were included in solution with the '251-TGF-p2. After 1 h, the wells were washed. Surface-associated lZ5I-TGF-P2 was recovered by incubation in 0.1 M NaOH, 2% SDS and quantified in a y counter. When '251-TGF-p2 was incubated in microtiter wells that did not have azM-methylamine, binding was decreased by greater than 98%. Reported results represent the average of three separate experiments with duplicate determinations.
Dissociation of 'z5Z-TGF-p2 from Immobilized azM-The apparent rate of dissociation of '251-TGF-@2 from immobilized azM-methylamine and immobilized native azM was determined according to the method used previously with TGF-Pl (23). a2M-methylamine and native azM were incubated in separate microtiter wells so that 90 fmol of either species was immobilized. '251-TGF-p2 was then incubated in the wells as described above. The wells were washed and equilibrated in fresh buffer (PBS with 15 p~ BSA) containing 0.1 p~ azM-methylamine (to inhibit reassociation of dissociated 9 -T G F -p2) at 22 "C. At various times, the wells were washed and treated with NaOH/SDS; free '251-TGF-P2 and '251-TGF-,92 remaining surface-associated were determined in a y counter.
Other Methods for Analyzing Binding of TGF-02 and TNF-a to ad4-'251-TGF-P2 (0.5 nM) was incubated with native azM or azMmethylamine (2.8 p M ) in PBS containing 8 p~ BSA for 90 min at 37 "C. '"I-TNF-a (1.5 nM) was incubated with azM-methylamine (2.8 p M ) in PBS with 3 pM BSA for 2 h at 37 "C. The reaction mixtures were subjected to FPLC gel filtration on Superose-6 equilibrated in PBS (0.4 ml/min). Binding of '251-TGF-p2 or '251-TNF-a to azM was determined by the radioactivity coeluting with azM (percent of total radioactivity loaded) as described previously (22, 31). Similar incubations were also analyzed by nondenaturing PAGE (21, 22) and by SDS-PAGE.

RESULTS
Demonstration of Apparent Equilibrium-Since noncovalent AC complex may be slowly converted into AC*, methods that measure total aZM-cytokine complex may fail to detect apparent equilibrium for the reversible association step (kl/ k l ) in the reaction. Using BS3, we measured AC, as a function of time for each cytokine and azM (1.0 p~) .
Representative plots with 1251-TGF-~1 and '251-TGF-/32 are shown in Fig. 1 In experiments with native a2M and 1251-TGF-01 or 1251-TGF-/32 (and other cytokines not shown), maximum AC, levels were lower (compared with a2M-methylamine); however, the times required to achieve apparent equilibrium were essentially unchanged. Based on these experiments, all apparent equilibrium measurements were made at 1 h, except when TGF-Bl and bFGF were studied. The later two growth factors formed AC* with a2M-methylamine more rapidly (see below) and therefore were incubated with a2M-methylamine or native a2M for 15 min to limit this conversion. a&i/Cytokine Binding Isotherms-Different concentratipns of a2M-methylamine were incubated with each 1251cytokine. Binding was detected by SDS-PAGE after exposing the solutions to BS3. Fig. 2 shows representative autoradiographs from experiments in which a2M-methylamine was incubated with 1251-TGF-j32 (panel A ) and 1251-NGF-/3 (panel B ) . The low-mobility bands included BS3-stabilized AC complex (AC,) and covalent a2M-methylamine-cytokine complex (AC*). Comparable low-mobility bands were observed in experiments with each of the other cytokines, except 1251-IFNy. We were not able to detect BS3-stabilized a2M-methylamine-12'I-IFN-y complex or native ( u~M -~~~I -I F N -~ complex, even when the a2M concentration was 4.0 p~ or when the time of incubation with BS3 was extended to 7.0 min. In each cytokine binding experiment, AC, was determined by correcting for AC* (detected in the absence of BS3). Ace/ (AC, + C,) varied as a hyperbolic function of a2M-methylamine concentration. A composite of results from five separate experiments with 12'I-TGF-P2 is shown in Fig. 3. Similar plots were generated for the other cytokines that bound a2Mmethylamine or native a2M.
Analysis of Cytokine Binding Experiments-In Fig. 4, results obtained in representative experiments with TGF-82 (panel A), TGF-81 ( p a e l B ) , and NGF-/3 (panel C) were plotted according to Equation 5. Similar graphs were constructed to analyze the results of experiments with PDGF-BB, TNF-a, and bFGF. Apparent equilibrium dissociation constants were determined from the slopes in each graph. These values were then averaged to generate the results presented in Table I. All of the cytokines, with the exception of TGF-82, bound a2M-methylamine with greater affinity than native a2M. Our native a2M preparations did not contain detectable levels of a2M that had undergone conformational  ( l a n e a ) , 10 nM ( l a n e b), 19 nM ( l a n e c), 36 nM (hne d ) , 68 nM ( h e e), 0.13 p M ( l a n e f), 0.25 p M ( l a n e g); 0.47 pM ( l a n e h), 0.89 pM ( l a n e i), 1 ( l a n e k).  Table I.

Cytokine
KD with azM-KD with native methyjarnine

B B cross-linker
The binding of '*'I-TGF-81 to a,M-methylamine was analyzed using different BS3 exposure times. In experiments with TNF-a, binding was insufficient for analysis after a 1.0 min pulse-exposure to BS3. Therefore, we extended the pulse exposure time to 7 min. This longer crosslinking time may have artificially decreased the apparent KD of the TNF-a-azM interaction. To examine this possibility, we analyzed the binding of 1261-TGF-~1 to azM-methylamine using different BS3 pulse exposure times. As shown in Table  11, slightly lower KD values were determined with longer BS3 incubation times. The rank order of affinity for cytokine binding to native a2M was similar to that presented for azM-methylamine. The most significant difference involved the unique ability of TGF-82 to bind native azM and azM-methylamine with equal affinity. As a result, the gap in affinity between TGF-82 and TGF-B1 was much larger (TGF-82 >> TGF-81) for native azM compared with a~M -m e t h y l a~n e . z Vulues and Covalent Binding of Cytokines to a&-The BS3 cross-linking efficiencies (2) were determined from the y intercepts in the plots shown in Fig. 4. Altho~gh many factors may influence the efficiency with which BS3 cross-links a complex, it is interesting that the z values were consistently higher in experiments with atM-methylamine (Table 111). Using the presented z values, we estimated the total amount of noncovalent azM-cytokine complex (AC) for each azM concentration in all of the experiments, according to Equation 3. The percentage of total binding which was covalent (Ai?/ AC + AC*) was then determined (Table 111). Under the conditions of our studies (1-h incubation with each cytokine except for TGF-81 and bFGF which were incubated for 15 min), E@"$% of the azN-cytokine complexes were noncovalent. Greater fractional covalent binding was observed with 82 > TGF-81, NGF-/3 > PDGF-BB 2 bFGF > TNF-a > IFN-  a~M-me~hyIamine, as expected due to the free Cys residues in this s t~~t u r e .

1537
The extent of covalent binding of azMmethylamine to TGF-fi1 and PDGF-BB (shown in Table 111) is consistent with previous studies (22,31).
TGF-82 Binding to ~~~o b i~~e d a&G"hen 1251-TGF-82 was incubated with immobil~z~d a~M-methyi~ine, 2.7 f 0.2 fmol of 1z51-TGF-j32 bound per well ( n = 5). Binding of '%I-TGF-82 to ~mmobilized ~zM-methylamine was inhibited equivalently by native azM and azM-methylamine in solution (Fig. 5). The ICso values were 19 k 3 nM for native azM and 19 & 5 nM for azM-methylamine, These results are markedly different from those obtained previously with TGF-fi1 (23). In the earlier studies, a~~-methylamine in solution was seven times more effective than native azM in inhibiting lz5I-TGF-81 binding to immobilized ~z M -m e t h y l~i n e .
The results of the experiments with immobilized aZM confirm the finding that TGF-82 i s unique in its ability to bind to native azM and azM-methylamine with comparable affinity.

~F -8 2
S~~i~ to azM as ~e~~r~~n e d by FPLC and Nond e~~u r~n g PAGE-Danielpour and Sporn (25) performed nondenaturing PAGE experiments and demonstrated that TGF-82 binds almost exclusively to conformationally transformed a2M. We hypothesized that their results differ from those presented here due to the use of a nonequ~librium system (nondenaturing PAGE:) to analyze TGF-82 binding. To address this hypothesis, we studied '261-TGF-@2 binding to native aZM and azM-methyIamine by FPLC and nondenaturing PAGE. In FPLC experiments, '251-TGF-fi2 binding to azM-methy~amine and native azM was 65 rt 2% and 20 k 0% (of total cytokine), respectively. In native PAGE experi-ments, lZ5I-TGF-/32 binding to a2M varied depending on the time of electrophoresis. After 1.5 h of electrophoresis, 1251-TGF-82 binding to a2M-methylamine and native azM was 85 and 51%, respectively (determined by radioactivity in gel slices). After electrophoresis for 3.5 h, 1251-TGF-P2 binding to azM-methylamine and native a2M was 72 and 32%, respectively (autoradiograph shown in Fig. 6). These results demonstrate that TGF-/32-a2M complexes dissociate during electrophoresis. The extent of dissociation depends on the time of electrophoresis and possibly the a2M conformation.
1251-TGF-P2 dissociation from immobilized native azM and immobilized azM-methylamine was studied at 22 "C. The dissociation curves were nonlinear (Fig. 7), as has been shown previously with TGF-Dl (23). Nonlinearity might represent heterogeneity in the structure of the a2M which is immobilized. 1251-TGF-B2 dissociated more rapidly from immobilized native azM compared with a2M-methylamine. The time required for 50% dissociation was 57 min with native azM and 128 min with azM-methylamine. Although cytokine dissocia- '*61-TGF-f12 (0.2 nM) was incubated in wells with immobilized azMmethylamine (0) and native a2M (0). After washing, the wells were equilibrated in fresh buffer containing azM-methylamine (0.1 p~) . At various times, the solutions and immobilized phases were separated, and the radioactivity in each was determined.
tion from immobilized a2M may differ from the process as it occurs in solution, these results support the hypothesis that different rates of dissociation influence recovery of native azM-cytokine and azM-methylamine-cytokine complexes by FPLC or nondenaturing PAGE.

TNF-a Binding to afl as Determined by FPLC and Non-
denaturing PAGE-Since the KD values determined for the interaction of a2M with TNF-a were unexpectedly low, we performed nondenaturing PAGE and FPLC experiments similar to those performed by Wollenberg et a1 (40). As shown in Fig. 6, lZ5I-TNF-a bound preferentially to a2M-methylamine, as determined by nondenaturing PAGE. In FPLC experiments, 44% of the lZ5I-TNF-a bound to a2M-methylamine. Therefore, the low affinity of the TNF-a/apM interaction does not preclude recovery in nonequilibrium systems.

DISCUSSION
The function of aZM as a cytokine carrier is complicated due to the different conformational states of azM. Only conformationally transformed azM (azM-proteinase, azM-methylamine) is recognized by the azM receptor (LRP) and binding of cytokines, including TGF-Dl, PDGF-BB, and TNF-a, does not inhibit the a2MILRP interaction (21, 24, 40). Therefore, cytokines which are bound to transformed a2M may be targeted to cells expressing LRP, either for catabolism or potentially to induce cellular responses. By contrast, cytokines which are bound to native aZM are stabilized. The plasma clearance of these cytokines is retarded (24,31), and the susceptibility to proteinase cleavage may be decreased (38). Due to LRP, native a2M is present in large molar excess to transformed a2M in the blood and most likely in many extravascular spaces. The form of a2M to which a cytokine binds will depend not only on the affinity of the interaction but also on the concentration of each a2M species.
We demonstrated previously that PDGF-BB and TGF-Bl bind predominantly to native a2M in the blood in vivo (31). This result was not entirely unexpected, since the plasma concentration of a2M is 2-4 PM (17). For the seven cytokines studied here, plasma represents a relatively simple system since there are few plasma proteins which compete with a2M for cytokine binding. By contrast, in cell culture systems, and in the extravascular spaces in uiuo, many macromolecules might compete with a2M for cytokine binding, including cellular receptors, solubilized forms of cellular receptors, proteoglycans, and extracellular matrix proteins. In order to model the function of a2M in complex systems, we must have some understanding of the strength of azM-cytokine interactions. We hypothesized and our results confirmed that the major component of many a2M-cytokine interactions is noncovalent complex formation. Information regarding the affinity of noncovalent azM-cytokine binding was only available in our previous study of TGF-Pl and immobilized aZM (23). Therefore, we focused the present investigation on the noncovalent association step in the a2M/cytokine binding mechanism. We examined cytokines which had previously been shown to associate with azM. In our studies, conversion of noncovalent azM-cytokine complex into disulfide-stabilized complex occurred slowly but progressively with time. This conversion may occur more rapidly and therefore be more important in the a2M regulatory mechanism of some cytokines such as IL-1P (34-37).
In experiments with azM-methylamine, the equilibrium dissociation constants for the seven cytokines varied from 13 nM (TGF-P2) to undetectable (greater than 5 p~ in the BS3 cross-linking system). The K D determined for the binding of TGF-Pl to azM-methylamine in solution (80 nM) was equiv-alent to that determined previously (79 nM) in the immobilized azM system (23). The relatively weak interaction of a~M-methylamine with bFGF, TNF-a, and PDGF-BB suggests that relatively high concentrations of conformationally transformed azM would be necessary to compete with other macromolecules which bind these cytokines (including cytokine-specific cellular receptors). Using the BS3 method, we were not able to demonstrate compiex formation between IFN-7 and azM-methylamine or native azM.
crzM-methylamine has frequently been used as a model of the transformed azM conformation and therefore a2M-proteinase complexes in general (17,46,47). The structure of a~M-methylamine is identical or nearly identical to that of most a~M-proteinase complexes (6, '7,461. crzM"methylamine and a~M-proteinase complexes are recognized equivalently by LRP (17) and by a n t i b~e s which do not recognize native azM (54). By contrast, in cytokine binding studies, a2Mmethylamine has not consistently behaved equivalent~y to a~M-proteinase complexes. For example, Huang et al. (20) and LaMarre et ul. (26) reported that the TGF-Bl binding activity of azM-methylamine is increased compared with native aZMt whereas the TGF-@l binding activity of a2M-trypsin is decreased. Hall et at. (22) subsequently demonstrated that this apparent discrepancy depends upon azM-proteinase binding stoichiometry. Although azM is capable of binding up to 2 mol of p r o~~n a s e / m o~, under most physiologic conditions (azM present in excess to proteinase), almost exclusively 1:1 (binary) azM-proteinase complexes form (8). These binary a~M-proteinase complexes uniformly demonstrate increased TGF-@l binding activity unlike the ternary (21) azM-proteinase complexes which demonstrate decreased TGF-@1 binding activity (22). ~q u~v a l e n t results have now been obtained in studies with PDGF-BB (31) and NGF-@? Therefore, we consider a2M-methylamine an appropriate model of physiologically significant azM-proteinase complexes in cytokine binding experiments.
The interaction of TGF-j32 with azM was unique in a number of ways. First, the affinity of the interaction of TGF-cj2 with azM-methylamine was the highest measured in this series. Second, TGF-BZ was the only cytokine which bound with comparable affinity to azM-methylamine and native aZM. This result was confirmed in two systems that measure apparent equil~brium binding (BS3 cross-linking and immobilized aZM). In experimental systems that are affected by a~M-cytokine diss~iation (FPLC, nondenaturing PAGE), greater TGF-B:! binding to azM-methylamine was observed, This result probably reflected different rates of dissociation of TGF-~-azM"methylamine complex and TGF-@%native azM complex. A second factor which may have been important in earlier studies is covalent binding. In our experiments, only a small percentage of the azM-methylamine-TGF-B2 complex was stabilized by disulfide bonds. If higher levels were achieved in other studies (perhaps due to longer incubation times), then this subpopulation of azM-methylamine-TGF-@2 complex would be resistant to dissociation during nondenaturing PAGE. Since native azM forms far less covalent complex with cytokines, stabilization would not be factor for this ~z M conformation. Our results with TGF-@2 add insight into previous studies comparing the regulation of TGF-@1 and TGF-132 by a2M. Danielpour and Sporn (25) reported that a2M significant~y inhibits the receptor binding and biological activity of TGF-cj2. In the same experimental systems, a2M was only 5% as effective in inhibiting receptor binding of TGF-Pl and had no reported a similar finding based on studies of primary hepatocyte cultures. aZM counteracted the mitoinhibitory activity of TGF-82 but did not affect the activity of TGF-81. The crzM used in these earlier studies (25, 55) was a commercial~y available bovine preparation which consists primarily of the native form. The large difference in results obtained with TGF-B2 versus TGF-Bl (25, 55) probably reflected not only the generally increased TGF-@2 binding activity of all azM species, but also the unique capacity of TGF-82 to bind with equal affinity to native a2M and conformationally transformed a2M.
The Kn values reported in this investigation assume one cytokine binding site per azM. If instead there are two or four independent, noninteracting cytokine binding sites, then the reported KB values are 2-or 4-fold lower than the actual values, respectively. In our ex~riments, azM was always present at large molar excess to 'z51-cytokine. Therefore, our results would not permit an assessment of cooperativity between multiple cytokine binding sites if they exist. Finally, it is possible that radioiodination affected the structure of some cytokine molecules so that a subpopulation of '251-cytokine was incapable of binding to a2M. If this is true, the determined KD values are still accurate for the fraction of '251-cytokine which retains azM binding activity. Variability in the fraction of 1251-cytokine which can bind to a&f would be reflected in the z values.
In a recent study, Liebl and Koo (56) compared the binding of nine cytokines to aZM. Some of the studied cytokines (TGF-8, NGF-6, PDGF, TNF-a) were the same as those surveyed here; however, the other invest~gators performed only electrophoresis experiments, Their results and the conclusions drawn from these results are highly disparate from the results and conclusions presented here. We attribute this discrepancy, at least in part, to the exclusive use (in the other study) of a nonequilibrium system (PAGE) to analyze cytokine binding to a2M. As mentioned above, this approach is problematic due to unpredictable levels of a2M-cytokine dissociation and the absence of a clear correlation with actual azM/cytokine binding affinity.