Localization of the Two Protease Binding Sites in Human a2-Macroglobulin*

The distance between the two protease binding sites in human plasma az-macroglobulin has been estimated using singlet-singlet energy transfer experiments. of (a- chymotrypsin)z the surface of the protease calculated to be 4 to 11 A, the equivalence of the

* This work was supported by financial aid from the Institut National de la Sante et de la Recherche Medicale. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' The abbreviations used are : w2M, a?-macroglobulin; CHY-CHY, dimeric a-chymotrypsin; FITC, fluorescein isothiocyanate; DNS, 5-broad spectrum of proteolytic enzymes. It binds irreversibly 2 molecules of elastase (5), trypsin (6, 7), or a-chymotrypsin (8) and only 1 molecule of the higher molecular weight plasmin (8). The azM plasmin complex binds, however, 1 molecule of trypsin (7,9). In addition, the various proteases compete for binding to azM (10). azM-bound proteases are able to react with small substrates or inhibitors at rates comparable with those of free proteases, but the degradation of high molecular weight substrates is markedly inhibited by steric hindrance (2). The localization of the protease binding sites over the ~z M surface might control these properties in part. In the following we propose an approach to this problem, using the transfer of electronic excitation energy by the Forster dipole-dipole resonance interaction (11) between fluorescent labels covalently linked to a2M-bound a-chymotrypsin. The preparation of a dimer of a-chymotrypsin with a spacer arm between the protein moieties proved to be the key to these investigations.
An aliquot of an ethanolic solution of this reagent was added under stirring to a 0.5 mM solution of a-chymotrypsin in 100 m phosphate, 100 mM NaCl buffer, pH 7.5. The final molar concentration of the reagent was 10-fold that of a-chymotrypsin. The mixture was allowed to stand at room temperature (22°C) for 20 min, after which it was filtered on Sephadex G-25 to remove excess reagent. Half of the achymotrypsin derivative was dialyzed against a 100 mM sodium acetate, 100 mM NaCl buffer, pH 4.5, and treated for 20 min at room temperature with 40 parts of dithiothreitol in order to produce the free thiol conjugate. This derivative was then dialyzed against the above phosphate/NaCl buffer and freed of reductant by filtration on Sephadex G-25. The thiol and 2-pyridyl disulfide-containing a-chymotrypsin derivatives were then mixed and reacted for an additional 3 h, yielding a heterogeneous mixture of conjugates. Unreacted achymotrypsin was isolated at 4°C by Sephadex G-50 S chromatography (phosphate/NaCl buffer) and discarded. The heavy fractions were collected, concentrated by ultrafiltration (Amicon YM 10 membrane), and rechromatographed on a column of Ultrogel AcA-54 in order to separate CHY-CHY (Kav = 0.27) from conjugates of higher molecular weight and from residual a-chymotrypsin. The yield of CHY-CHY was 7% assuming € 2~ to be twice as large as that of the monomer. CHY-CHY migrated as a single band on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (10% acrylamide). Labeling of Proteins with Fluorescent Dyes-Two to five milligrams of FITC (isomer I) or of DNS were added under vigorous stirring to 20 mg of a-chymotrypsin or CHY-CHY dissolved in 5 rnl of 10 mM borate, 20 m CaCI2 buffer. The pH of the buffer was between 8.1 and 9.4 depending on the degree of labeling desired (in most cases it was 8.8). After 1 h or less at 0°C. the excess dye was removed by Sephadex G-25 filtration (50 mM Tris-HC1 buffer, pH 7.5). The labeled proteins were rechromatographed on Sephadex G-50 S or on Ultrogel AcA-54 with the same buffer. After measurement of enzymic activity (see below), the solutions were dialyzed against water and lyophilized. The average degrees of labeling ( p ) were determined by absorbance measurements using em = 45 mM-' cm" for a-chymotrypsin (12), €30 = 3.4 rn"' cm-' for DNS (13). and erg:, = 66.8 mM-' for FITC (14). a2 Me Chymotrypsin Complexes-Human aaM was prepared as described previously (15). Molarities of solutions of this protein were determined by absorbance measurements assuming = 590 m"' cm" (1). Complexes were formed by reacting 40 p~ alM with 8 to 100 p~ a-chymotrypsin or CHY-CHY in 50 m Tris-HC1, 100 m NaCl dimethylaminonaphthalene-1-sulfonate; p, average degree of labeling (moles of dye/mol of protein).

cuz-Macroglobulin-cu-Chymotrypsin Interactions
buffer, pH 7.5, for 10 min at room temperature. The excess of protease, if any, was removed by Sephadex G-200 filtration.
Chymotrypsin Activity Measurements-These assays were performed at 25°C by the method of Erlanger et al. (16) using N-glutaryl-L-phenylalanine-p-nitroanilide as the substrate, except that the buffer (100 mM Tris-HC1,lOO m~ NaC1, pH 7.6) did not contain CaC12 which precipitates CHY-CHY.
Singlet Energy Transfer Measurements-These experiments were done by measuring the quenching of the DNS fluorescence (Aeicltation = 360 n m , A. , , , i , , = 470 nm) by FITC with a Farrand Mark I spectrofluorometer equipped with a Hamamatsu R-446 photomultiplier. Controls with FITC-free proteins were run to take aspecific effects into account. The average transfer efficiency d is given by: d = 1 -FD*/FD' where FD" and FD' are the normalized fluorescence intensities of the donor (DNS) in the presence and in the absence of the acceptor (FITC), respectively. The fluorescence quantum yield of DNS conjugates was determined using quinine bisulfate in 0.1 N H2S04. Since the concentration of protein-bound DNS or FITC was lower than 0.5 p~, absorbance values were sufficiently low for inner filter effects to be negligible. for N-glutaryl-L-phenylalanine-p-nitroanilide was the same as that of a-chymotrypsin while its kcat value was lower (0.016 s-l compared to 0.022 s-l for a-chymotrypsin). In practice, not more than 2 molecules of FITC could be bound per molecule of CHY-CHY because the disulfide bridge of the spacer arm of CHY-CHY was partly cleaved upon reaction with this dye. Attempts a t producing CHY-CHY from heavily FITC-labeled a-chymotrypsin resulted in poor yields, due probably to the substitution by the dye of surface lysine residue(s) being essential to reaction with the N-hydroxysuccinimide ester of the coupling reagent. Stoichiometries of the Dye-labeled Chymotrypsin. a2M Complexes-In order to interpret correctly the singlet energy transfer data, it was necessary to know exactly the stoichiometries of the various complexes used for the transfer experiments. For FITC-labeled a-chymotrypsin, the stoichiometry was determined by measuring the absorbance at 280 nm (proteins) and 495 nm (dye) and using the appropriate E values (see "Experimental Procedures"). The absorption of the dye at 280 nm was taken into account. The a-chymotrypsin:azM binding ratio was found to be 2.1:l and 2.0:l using two batches of labeled a-chymotrypsin whose p values were 0.45 and 3.1 (see also Footnote 2).

Properties of Dimeric
The absorption band of DNS being ill-defined in the presence of azM, the binding of DNS-labeled a-chymotrypsin was also monitored by measuring the fluorescence intensity of the dye a t 510 nm (&rcitation = 340 nm) and comparing it to the absorbance at 280 nm. The a-chymotrypsin:azM binding ratio was found to be 2.1:l for pDNs = 1.1 and 1.9:l for pDNS = 2.2.
Labeling with DNS or FITC decreases the activity of a-chymotrypsin. For pDNS = 1.8 and pFITC = 3.1, the residual activity on Nglutaryl-L-phenylalanine-p-nitroanilide is 81% and 51% respectively. If the binding of the labels obeys a Poisson distribution, part of the a-chymotrypsin molecules will be poorly labeled or even unlabeled. Thus, they will have a higher enzymic activity than the more heavily labeled molecules. If asM reacts faster with native (i.e. 100% active) than with labeled a-chymotrypsin, the binding stoichiometry might therefore be underestimated. To rule out this possibility, we have compared the enzymic activities of free and azM-bound labeled and unlabeled a-chymotrypsin. We have found that azM does not change the specific activity of unlabeled a-chymotrypsin. On the other hand, the activity of FITC-labeled a-chymotrypsin (pFtTc' = 3.1) per A495 unit was found to be exactly the same whether the protease is free or complexed to a2M. This rules out the possibility that CYPM binds more than 2 chymotrypsin molecules.  The CHY-CHY dimer was labeled with FITC and its binding stoichiometry was determined in the same way as that of monomeric a-chymotrypsin. The stoichiometry of the CHY-CHY .azM complex was 1:l for pFITC = 0.7 and 0.9:l for pLFITC = 2. Reaction of this complex with FITC-labeled a-chymotrypsin gave a ternary complex formed of 1 mol of azM, 1 mol of CHY-CHY, and 1 mol of a-chymotrypsin.
Experiments were also undertaken to see whether CHY-CHY would be able to cross-link aZM molecules. Measurements were then performed on a donor-and acceptor-labeled azM. (CHY)z complex prepared by reacting equimolar amounts of aZM and DNS-labeled a-chymotrypsin ( p = 1.8); this 1:l a-chymotrypsin. azM complex was then reacted with a %fold molar excess either of a-chymotrypsin or of FITC-labeled a-chymotrypsin ( p = 3.1). Another complex was prepared by reacting equimolar amounts of azM and a-chymotrypsin; this 1:1 complex was subsequently reacted with a %fold molar excess of FITC-labeled a-chymotrypsin ( p = 3.1) ( Table I). In each case, the excess of protease was removed by gel filtration. The two aZM. (CHY)z complexes containing only donor or acceptor labels served as controls for the (donor + acceptor)-labeled complex. Absorption measurements indicated that 1 mol of FITC-labeled a-chymotrypsin was present per mol of azM within the two acceptor-containing complexes. t was measured as before and was found to be 0.125. As is shown below, this result may need to be corrected. If the two binding sites of azM are equivalent or if the occupancy of one site does not affect the binding capacity of the other site, the In order t o c o n f i i this result in a more straightforward manner, we have prepared complexes formed of 1 mol of CHY-CHY and 1 mol of a-chymotrypsin per mol of a2M. One mol of aZM was reacted with either 1 mol of unlabeled CHY-CHY or 1 mol of FITC-labeled CHY-CHY ( p = 0.7 or p = 2.0).4 One mol of the former complex was reacted with 1 mol of DNS-substituted a-chymotrypsin ( p = 2.2) and 1 mol of the latter complex was reacted with either 1 mol of free CHY or of DNS-labeled CHY (p = 2.2). Since the a2M:CHY-CHY stoichiometry is 1:1, each molecule of azM contains 1 molecule of CHY-CHY and 1 molecule of a-chymotrypsin. Hence, k needs no correction. The two complexes containing only donor or acceptor molecules served as controls for the (donor + acceptor)-labeled complex. k was found to be 0.12 for the poorly FITC-labeled complex ( p = 0.7) and 0.27 for the more heavily labeled complex (p = 2.0). Interpretation of these data is a priori more complicated than in the previous case because the quenching of the donor fluorescence is brought about by acceptor molecules present on the two monomers of CHY-CHY. The distance between the a2M-bound a-chymotrypsin and CHY-CHY molecules may however be estimated "by trial and error." For instance, if we assume a distance of 4 8, (ie. the shortest distance found in the preceding experiment) between a-chymotrypsin and one subunit of CHY-CHY, we may calculate the transfer efficiencies between these 2 a2Mbound CHY molecules. These efficiencies  0.075 and 0.18). Hence we may conclude that the two a2M-bound a-chymotrypsin molecules are not distant by more than 4 8, (Fig. 1). DISCUSSION End group sequence determinations using highly purified preparations (4) have demonstrated that all four M, = 185,000 peptide chains in human a2M are very similar or identical5; hence, the noncovalently bound M, = 365,000 subunits of azM are very likely equivalent and the two protease binding sites should be symmetrically localized about some CZ symmetry axis. These binding sites appear surprisingly close to each other from the experiments reported in this paper, namely about 44 8, center to center (ie. a-chymotrypsin molecules complexed to a2M are not far from touching). The binding sites should therefore be located at the interface of the subunits and near the Cz symmetry axis. This model, however, raises some questions about the environment of enzymes in the a2M complex; it suggests that bound proteases are buried in the a2M core whereas the residual activity of azM-bound proteolytic enzymes on medium-sized substrates of physiological importance (2, 20-22) and the effect of natural macromolecular inhibitors on a2M-bound trypsin6 (23, 24), both support binding on the surface of a2M. A broad channel at the interface of the M, = 365,000 halves of a2M would provide for the protease binding sites a means to meet the requirements of accessibility, proximity, and symmetrical localization. On the other hand, an asymmetrical binding of proteases such as a-chymotrypsin in a2M would fit our results but appears unlikely from what is known about the proteolytic cleavage of the M, = 185,000 quarters which parallels the trapping of enzymes by a2M (4). The present findings indicate that the binding of protease to a2M can be controlled by steric hindrance. In fact, a2M appears unable to trap 2 molecules of our synthetic CHY-CHY dimer (as previously observed for plasmin (8)), despite the presence of a -CO-CH2-CH2-S-S-CHz-CH2-CO-spacer arm between the a-chymotrypsin moieties. We would suggest the M, = 185,000 to M, = 85,000 proteolytic cleavage (4, 25,26) does not occur on both halves of the a2M molecule when using sterically hindered proteases such as plasmin or CHY-CHY, thus leaving the second binding site undamaged and ready to bind small endoproteases. The kinetics of protease binding should allow this problem to be elucidated.