Bait Region Involvement in the Dimer-Dimer Interface of Human a2-Macroglobulin and in Mediating Gross Conformational Change EVIDENCE FROM CYSTEINE VARIANTS THAT FORM INTERDIMER DISULFIDES*

We have characterized four human a2-macroglobulin (a2M) bait region variants (G679C, M690C, V700C, and T705C) to test the hypothesis that the bait regions are involved in the interface between noncovalently associated dimers. All four variants folded correctly as judged by many normal properties. However, the presence of a cysteine resulted in disulfide formation between otherwise noncovalently associated dimers in all four variants. The extent of disulfide cross-linking varied with the location of the cysteine and gave a mixture of species that probably contained two, one, or zero interdimer disulfides in the tetramer. This was reflected in heterogeneity of conformational change upon thiol ester cleavage by methylamine, with the presence of crosslinks correlating with blockage of conformational change. The stoichiometry of trypsin inhibition was less in all cases than for wild-type a2M. The M690C variant also showed evidence of some species with an intramolecular disulfide between bait regions of monomers within the same dimer. Taken together, the results are consistent with a location of the four bait regions in contact with, or in very close proximity to, one another. This suggests that they form all or part of the “cavity body” seen in the low resolution x-ray structure of transformed a2M.

Inhibition of proteinases by human ␣ 2 -macroglobulin (␣ 2 M) 1 results from a series of conformational changes that are initiated by the proteinase and that result in the physical sequestration of the proteinase within the closed cage-like structure of the conformationally altered ␣ 2 M (1, 2). The initiating event is a proteolytic cleavage by the attacking proteinase near the center of the ␣ 2 M polypeptide in a region termed the bait region (3,4). Cleavage anywhere within the bait region also results in activation of an internal thiol ester toward cleavage by nucleophiles. From studies on the kinetics of cleavage of the thiol ester by nucleophiles and of the subsequent conformational change within the ␣ 2 M, it has been shown that the conformational change occurs cooperatively after both thiol esters within one half of the ␣ 2 M tetramer have been cleaved (5,6).
Knowledge of the structural relationship between the thiol ester and the bait region and hence of the details of the activation mechanism is limited by the absence of a high resolution x-ray structure of either native or conformationally altered human ␣ 2 M. From fluorescence resonance energy transfer measurements, it was shown that the four cysteines that form the four thiol esters of the human ␣ 2 M tetramer are centrally located and are about 35 Å apart (7). This was subsequently confirmed by a low resolution (ϳ10 Å) x-ray structure of conformationally altered ␣ 2 M (8). NMR measurements on ␣ 2 M showed that the bait region of each monomer was unusually flexible (9,10) and that it lies close (10 -17 Å) to the cysteine of the thiol ester in the transformed protein (11). However, the location of the four bait regions is still not known. Studies from this laboratory on ␣ 2 M variants in which large portions of the bait region had been removed showed that bait region truncation prevented the disulfide-linked ␣ 2 M dimers from associating noncovalently to give the functional tetramer (12) and that, despite normal reactivity of the thiol esters toward nucleophiles, there was no cooperativity of conformational change between subunits. This implicated the bait region in formation of the ␣ 2 M tetramer and raised the possibility that the bait regions may lie in the central cavity of ␣ 2 M in contact with one another. This is made more plausible by the presence of a symmetrical body of electron density in the central cavity of methylamine-treated ␣ 2 M, termed the "cavity body," that is absent from proteolytically cleaved ␣ 2 M (8).
To test the hypothesis that the bait region is involved in the noncovalent association of ␣ 2 M dimers and in the mediation of cooperative subunit-subunit interactions that are involved in the gross trapping conformational changes of ␣ 2 M, we used a scanning mutagenesis approach in which four single residue substitution variants of human ␣ 2 M containing a cysteine at distal, central, or proximal ends of the bait region have been prepared and characterized (G679C, M690C, V700C, and T705C) (Fig. 1). The variants folded normally, as judged by several properties, including the formation of tetrameric ␣ 2 M, the presence of thiol esters with normal reactivity toward nucleophiles, and the ability to undergo a proteinase-induced conformational change. However, the extent and rate of conformational change induced by methylamine cleavage of the thiol esters was compromised for all of the variants. Although all of the variants could still inhibit proteinase, the stoichiometries were reduced compared with wild-type recombinant ␣ 2 M or plasma ␣ 2 M. These perturbed properties correlated with the presence of new interdimer disulfide bonds formed between the new cysteine residues in two otherwise noncovalently associated disulfide-linked dimers. Such cross-linking resulted in covalent tetramers with either one or two new disulfides between the dimers. The extent of formation of such new disulfides was a function of the position of the cysteine within the bait region. This suggests that the bait regions are involved both in dimer-dimer association and in mediation of transdimer conformational changes and that there is a requirement for unrestrained reorientation of the bait region interface. In addition, the M690C variant showed evidence of a new intradimer disulfide, suggesting that bait regions within a dimer are also close to one another. These findings suggest that all four of the bait regions are at the center of the ␣ 2 M cavity and may constitute all or part of the cavity body seen in the low resolution x-ray structure of transformed ␣ 2 M.

Site-directed Mutagenesis and Expression of Recombinant ␣ 2 Ms-For
circumstantial reasons, the four mutations were not all created by the same procedure. For the variants G679C and V700C, site-directed mutagenesis was performed in a single-stranded M13mp18 construct containing the 1690-base pair BamHI to BsiWI fragment from plasmid p1167 (13), which covers the coding region for the bait region. Mutagenesis was carried out using the Amersham Corp. in vitro Mutagenesis System 2.1 according to the manufacturer's protocol. The mutations were confirmed by sequencing, after which the mutated BamHI to BsiWI fragment was excised and ligated back into the p1167 expression vector. The M690C variant was created by mutagenesis using the Altered Sites II Mutagenesis System (Promega) on the 2669-base pair BamHI to ClaI fragment from p1167 subcloned into pAlter (Promega). The 28-base oligonucleotide 5Ј-T GAG TCA GAT GTA TGT GGA AGA GGC CAT-3Ј was used for the mutagenesis. After confirmation of the mutation in pAlter, the mutated BamHI to ClaI fragment was excised and religated into p1167. The T705C variant was created in two stages. In the first step, an XmaI site was introduced at base 3572 of p1167 using the Amersham Corp. in vitro Mutagenesis System 2.1 protocol. The mutation at residue 705 was then introduced by replacing the XmaI to BsiWI fragment of p1167 with a synthetic duplex containing the desired change. This duplex was formed by annealing oligonucleotides 5Ј-CC GGG TGT GAG ACC-3Ј and 5Ј-C TAC CGT CTC ACA C-3Ј. The mutation was confirmed by sequencing in p1167. Stably transfected BHK cell lines expressing each of the four ␣ 2 M variants were established by the co-transfection and drug resistance selection procedures previously described (14,15).
Purification of Recombinant and Plasma ␣ 2 Ms-Plasma ␣ 2 M was isolated from outdated human plasma obtained from Rush University Hospital Blood Bank. The ␣ 2 M was precipitated in a 4 -12% cut with PEG8000 and collected as a pellet by centrifugation. The pellet was resuspended in 0.1 M sodium phosphate buffer, pH 6.5, containing 0.8 M sodium chloride and chromatographed on a zinc chelate matrix, as described previously (16). Recombinant ␣ 2 Ms were obtained from the growth medium of the stably transfected BHK cells that were grown to confluence in roller bottles and cycled between serum-containing and serum-free medium, as described previously (15). Purification was by zinc chelate chromatography, as for plasma ␣ 2 M, but without the PEG8000 precipitation step.
Polyacrylamide Gel Electrophoresis-Polyacrylamide gels run under denaturing and reducing conditions (8% acrylamide) or denaturing and nonreducing conditions (3.5% acrylamide cast on Pharmacia Gel Bond film) were run according to the procedure of Laemmli (17). Nondenaturing gels (5% acrylamide) were run as described previously (18). Urea polyacrylamide gels (5% acrylamide) contained 5 M urea and were run in the same way as nondenaturing gels.
TNS Fluorescence Measurements-Both kinetic and wavelength scan fluorescence measurements were made on an SLM8000 spectrofluorometer. Kinetic measurements used excitation at 316 nm and observation of emission at 410 nm, with slits of 4 and 16 nm for excitation and emission, respectively. Wavelength scans used excitation at 316 nm and emission monitored in 2-nm steps from 360 to 600 nm. Slits of 2 nm for both excitation and emission were used.
Quantitation of Thiol Esters and Kinetics of Cleavage-Determination of both the number and kinetics of cleavage of thiol esters present in the variants were carried out by continuous monitoring of the free SH (sulfhydryl) released from the thiol ester by reaction with methylamine, by spectrophotometric measurement of the TNB Ϫ released by reaction of the free SH with DTNB. TNB Ϫ was quantitated from the change in absorbance at 412 nm using an extinction coefficient of 13,600 M Ϫ1 cm Ϫ1 (19). For kinetic measurements, the ␣ 2 M (final concentration, 1.0 M) was preincubated in assay buffer containing 100 M DTNB. The reaction was initiated by addition of amine hydrochloride stock solution (5 M, pH 8.0) to give the desired final concentration. Measurements were made in a dual beam Shimadzu UV2101PC spectrophotometer; the blank contained all components except ␣ 2 M. Data were fitted by nonlinear least squares analysis to a monoexponential using Scientist (MicroMath, Salt Lake City, UT).
Trypsin Inhibition Assay-Stoichiometry of trypsin trapping was determined by measuring the soybean trypsin inhibitor-resistant trypsin activity (20,21) at different trypsin:␣ 2 M ratios. ␣ 2 M (120 nM) was incubated with different molar ratios of trypsin for 5 min at 25°C. The reaction was diluted 1:1 with a solution of soybean trypsin inhibitor (final concentration, 1.25 mM). After 1 min of incubation, residual trypsin activity, representing ␣ 2 M-trapped trypsin, was assayed by addition of 10 l of the reaction mixture to 190 l of assay buffer containing 200 M (final concentration) N-benzoyl-L-isoleucyl-L-glutamylglycyl-L-arginine-p-nitroanilide (S-2222, Chromogenix) in a microcuvette. Change in absorbance at 410 nm, relative to a reference cell containing only substrate, was monitored using a dual beam Shimadzu 2101PC spectrophotometer. Trypsin activities were determined from the slope of the resulting time courses, fitted by least squares linear fit using the program Origin 1.16 (MicroCal Inc., Amherst, MA). All reactions and assays were carried out in 20 mM sodium phosphate buffer, pH 7.4, containing 0.1 M sodium chloride, 0.1 mM EDTA, and 0.1% PEG8000.

Electrophoretic Mobility of Variants on
Nondenaturing PAGE-Following reaction with methylamine or proteinase, human plasma ␣ 2 M undergoes a major change in conformation that results in an increase in electrophoretic mobility on polyacrylamide gels run under nondenaturing conditions (22). This "slow to fast" conversion is frequently used as a diagnostic for this major conformational change, with the result that ␣ 2 M conformations are often referred to as the slow form or fast form. In addition, there is an intermediate mobility species in which conformational change has occurred in only two of the four subunits (6,23). We tested the ability of the four ␣ 2 M variants to undergo this slow to intermediate or slow to fast form interconversion upon reaction with either methylamine or proteinase. All four variants had the expected slow mobility prior to reaction (Fig. 2) that is characteristic of tetrameric ␣ 2 M in the native state. Reaction with a small excess of proteinase (trypsin) for 10 min was sufficient to completely convert each of the variants from slow to fast conformations (Fig. 2). However, methylamine treatment overnight produced only partial con- The hypervariable region that defines the bait region stretches from residue 667 to residue 705 in human ␣ 2 M. Beyond these residues, toward the N and C termini, respectively, there is very high sequence homology between different members of the ␣-macroglobulin family. Within the bait region, there is very high variation both in length and in composition (4).
one-third to one-half fast form and a smaller amount of intermediate mobility species. Recombinant wild-type ␣ 2 M was completely converted to the fast form by methylamine treat-ment under the same conditions (Fig. 2), as was plasma ␣ 2 M (not shown). The failure of the variants to be completely converted to the fast form was not due to incomplete cleavage of the thiol esters by methylamine (see below).
We also followed the rate of conversion of the variants from slow to intermediate or fast conformations as a function of time of reaction with methylamine at pH 8 by nondenaturing PAGE. The appearance of the intermediate and fast mobility species was complete within 2 h, after which no further changes were observed up to the longest time examined of 24 h (data not shown). Therefore, although the variants initially have uniform slow mobility, they are heterogeneous with respect to the ability to undergo change from slow to intermediate or fast mobility species.
Quantitation of Free Cysteine and of Intact Thiol Esters and Determination of the Kinetics of Thiol Ester Cleavage-Because of the limited quantities available for some of the variants, the number of free thiols present initially in each of the variants was estimated by reaction with 5-iodoacetamido fluorescein rather than by DTNB assay because this gave higher sensitivity and allowed fluorescence quantitation. When allowance was made for a small amount of nonspecific labeling, which was also present in recombinant wild-type ␣ 2 M, there was little evidence of a significant fraction of the new cysteine residues being in unconjugated form ( Table I), suggesting that the new cysteines introduced into the bait region were all involved in disulfide linkages, either to other cysteines within the tetrameric protein or to small thiols, such as glutathione. The curves represent the nonlinear least squares fits to the data to a monoexponential. The rate constants representing these curves are given in Table I. The number of intact thiol esters present was also determined by DTNB assay, which involved reaction of the variants with methylamine and quantitation of the free SH generated (Fig. 3). From analysis of the time dependence of the change in absorbance at 412 nm, both the total number of methylaminecleavable thiol esters and the kinetics of their cleavage were determined (Table I). By both criteria, all four of the variants were similar to both plasma and recombinant wild-type ␣ 2 Ms, having between two and four intact thiol esters that could be cleaved at rates that were 63-79% that of recombinant wildtype ␣ 2 M. These results indicated that the mutations had not greatly altered either the ability of the ␣ 2 M subunits to form thiol esters or the subsequent reactivity of the thiol esters toward small nucleophiles. The lower stoichiometry of thiol esters for the V700C and T705C variants correlated with the presence of a small amount of intermediate mobility material prior to reaction (Fig. 2). This form is likely to have only two intact thiol esters. However, the major species had slow electrophoretic mobility and probably contained three or four intact thiol esters.
Proteinase-and Methylamine-induced Change in Conformation Detected by TNS Fluorescence-The conformational changes that accompany reaction with proteinase were complete within the 10 min used for the reaction, so the TNS emission spectrum was stable after this time. Comparison of the TNS emission spectra for unreacted and proteinase-reacted variants with those for plasma ␣ 2 M showed significant differences for both the starting and the finishing conformations (Fig. 4). TNS in the presence of plasma ␣ 2 M gave an emission spectrum with max at about 440 nm. Upon reaction with proteinase, the TNS fluorescence intensity increased more than 2-fold, and the wavelength maximum blue shifted from 440 to about 410 nm (Fig. 4). The same behavior was seen for wildtype recombinant ␣ 2 M (Fig. 4). In contrast, TNS in the presence of any of the four variants already showed enhanced intensity, and wavelength maxima between those of native and trypsin-treated wild-type or plasma ␣ 2 Ms (Fig. 4). Reaction of the variants with either chymotrypsin or trypsin produced very little additional increase for the M690C and T705C variants, and a reduction in intensity for the G679C and V700C variants. Almost no change in emission maximum was seen for the M690C and T705C variants after reaction with either trypsin or chymotrypsin. Although small blue shifts were observed upon proteinase treatment of the G679C and V700C variants, the wavelength maxima of the proteinase-treated variant was still not as much blue shifted as wild-type or plasma ␣ 2 M. Methylamine treatment resulted in almost no change in wavelength maximum for any of the variants and much smaller percentages of enhancement for any of the variants, particularly T705C, than for recombinant wild-type or plasma ␣ 2 Ms.
Ability of Variants to Inhibit Trypsin by Trapping-The ability of each variant to inhibit trypsin by the normal bait region cleavage-induced conformational change was determined by a trapping assay in which increasing amounts of trypsin were incubated with each ␣ 2 M variant, and any nontrapped trypsin was inhibited by addition of a large excess of soybean trypsin inhibitor. The residual trypsin activity was then determined by chromogenic assay. The stoichiometry of trapping was determined from the turnover point of a plot of residual activity against the trypsin:␣ 2 M ratio (not shown). Each variant was able to trap trypsin, but the stoichiometry was greatly reduced compared with plasma and wild-type recombinant ␣ 2 Ms (Table  I), despite complete and specific cleavage of all four of the bait regions (Fig. 5). The activity of trypsin trapped by each of the variants was also reduced from that trapped by wild-type ␣ 2 M (Table I). The largest reductions in specific activity were seen for the M690C and T705C variants.
Origin of Aberrant Properties of Cysteine Variants-From the above characterization of the four variants, it is clear that, whereas they behave normally in forming tetramers with intact thiol esters that are cleavable by methylamine at normal rates and bait regions that are still specifically cleavable by proteinase, their ability to undergo rapid cooperative conformational changes and to efficiently trap 1-2 mol of proteinase has been compromised by the introduction of cysteine into the bait region. We therefore determined whether the aberrant properties arose from the presence of new disulfides within the tetramer.
Plasma ␣ 2 -macroglobulin is a tetramer, made up of two noncovalently associated disulfide-linked dimers (24). Upon treatment with either urea (Fig. 6A) or SDS (Fig. 6, B and C), the noncovalent interactions can be disrupted, and the protein migrates as an ϳ360-kDa dimer. We found that all of the cysteine variants showed evidence of the presence of additional disulfide bonds between the dimers, from the presence of nondissociable 720-kDa tetramers (Fig. 6). Such behavior could result from either one or two additional disulfides between the dimers. In each case, a minority of the protein lacked such additional interdimer disulfides and was still dissociable to dimer. This varied among the variants and was the greatest for the T705C variant (Fig. 6).
Another difference between variants was that the M690C variant showed a different mobility for its dimer compared with the dimers of the other variants when dissociation was carried out in SDS (Fig. 6C), but less so when carried out with urea (Fig. 6A). This may be due to the presence of a new intradimer disulfide in a fraction of the subunits that may constrain the conformation of the SDS-unfolded protein so that it has a different mobility from the equivalently unfolded dimers that do not contain such an additional disulfide. SDS-PAGE run under reducing conditions confirmed that all of the variants were composed of the expected 180-kDa monomers and that the bait regions were all accessible to proteinase and could be cleaved upon exposure to trypsin (Fig. 5).
Kinetics of Fast Phase Conformational Change-The timecourse of change in mobility of the variants on nondenaturing PAGE showed that only a fraction of each variant showed conversion at rates similar to that of wild-type or plasma ␣ 2 M when reacted with methylamine (see above). To test whether there was an equivalent fraction that also showed normal kinetics of conformational change, we followed the time-dependence of the TNS fluorescence change upon reaction with methylamine (Fig. 7) and determined the rate constant for the change. For all four variants it appeared that there was a rapid phase of smaller magnitude than for wild-type ␣ 2 M, that was equivalent to the normal conformational change step of recombinant wild-type and plasma ␣ 2 Ms, because the fluorescence changes were well fitted to a monoexponential, as expected for the conditions of the reaction, and gave rate constants (k c ) no more than about 3-fold slower than those for plasma and wildtype recombinant ␣ 2 Ms (Table I). DISCUSSION We have examined four variants of human ␣ 2 M, each of which contains a single cysteine at a different location within the bait region, to test the involvement of the bait region in dimer-dimer association and in the mechanism of conforma-FIG. 6. Presence of new disulfide cross-links between disulfide-linked dimers from presence of large but variable amounts of nondissociable tetramer on polyacrylamide gels run under different denaturing/dissociating conditions in the absence of reducing agent. A, samples incubated and run in a gel containing 5 M urea. B and C, samples incubated and run in 1% SDS on either 4% total acrylamide (B) or 3.5% total acrylamide (C). Note the different types of dimer that are resolvable in C (indicated by 1 and 2). 7. Normal kinetics of conformational change for the fraction of each variant that undergoes rapid conformational change, monitored by change in TNS fluorescence. The solid lines are the best fits to a single exponential decay. The behavior of plasma ␣ 2 M wild type (p-wt) and recombinant wild type (r-wt) are shown for comparison. All ␣ 2 M species were prereacted with methylamine at an amine concentration of 2.5 M at pH 8.0 and 25°C for 30 s. This ensured cleavage of all thiol esters prior to monitoring of fluorescence change. Samples were then diluted into the assay buffer containing TNS, and data acquisition was started after a delay of approximately 20 s. The dilutions were between 40-and 15-fold to give a constant final ␣ 2 M concentration. tional change and proteinase inhibition by determining the tendency of these variants to form new interdimer disulfides and thereby alter the properties of the ␣ 2 M. We found that a major fraction of each of the four variants formed new interdimer cross-links, suggesting a very close approach of bait regions in separate dimers to one another in the tetramer and consequently a location for the bait regions at the interface between these dimers. These new cross-links blocked the conformational change that occurs upon cleavage of the thiol esters by nucleophiles and suggests that the bait regions not only are in contact with one another across the dimer-dimer interface but also mediate the cooperative conformational changes that constitute the slow to fast form interconversion, perhaps through a rearrangement of their contact surfaces. Formation of a new disulfide cross-link between pairs of bait regions across the interface may lock the interface so that conformational changes can only occur very slowly or not at all compared with the unconstrained bait regions of wild-type recombinant or plasma ␣ 2 Ms. Consistent with this is the greatly lowered ability of the variants to trap trypsin, even though trypsin completely and specifically cleaved all bait regions in each of the variants. The finding that cysteine at different positions within the bait region is able to cross-link to a bait region of the other dimer with different efficiencies suggests that the interface is more complex than a parallel alignment of bait regions. Instead, different parts of the bait region must approach the equivalent residue in the other dimer to greater or lesser extents. In addition, the possible formation of a new disulfide between monomers of the same dimer in the M690C variant suggests that the bait regions within a dimer may approach one another at this point. This suggests an organization of the four bait regions of human ␣ 2 M in the central cavity of the tetramer, perhaps as the cavity body of the x-ray structure, with extensive interaction between pairs of bait regions from different dimers and less extensive contact between pairs from the same dimer (Fig. 8).
A possible concern in trying to deduce structural properties of plasma ␣ 2 M from the properties of these variants is that formation of a new disulfide between dimers might have grossly distorted the interface that results in the normal dimerdimer association and thus either precluded tetramer formation or driven the tetramer into a distorted structure that is not representative of native ␣ 2 M. However, by several criteria, this appears not to be the case for these variants, because they possess a number of structurally dependent properties that are very similar to those of plasma ␣ 2 M. Thus, all four variants form tetramer as the nearly exclusive species, with the exception of M690C and V700C, which also form a small amount of dimer. This contrasts with more drastically modified bait region variants that we have previously examined, which had large truncations within the bait region and that consequently formed exclusively dimers that could not associate to tetramer (12). The present variants also form close to the full complement of thiol esters that can be cleaved by nucleophiles at rates very similar to plasma or recombinant wild-type ␣ 2 M (Table I).
They are also capable of undergoing complete slow to fast conformational change upon reaction with proteinase. The presence of new interdimer disulfides is therefore likely to result from the juxtaposition of the new cysteines across the dimer-dimer interface as a consequence of initial dimer-dimer association in the normal manner. This then requires the bait regions to be directly involved in dimer-dimer interactions in plasma ␣ 2 M.
Such involvement of the bait regions of different dimers in the noncovalent interface of ␣ 2 M is not surprising, given the cooperative nature of the conformational changes that result from bait region or thiol ester cleavage. Earlier studies by Roche et al. on chemical cross-linking of subunits of ␣ 2 M (25, 26) had already shown that covalent cross-links could lead to a rigidification of the ␣ 2 M interface such that it no longer responded normally to bait region or thiol ester cleavage. However, unambiguous interpretation of these earlier studies was not possible because numerous cross-links were always present and the sites of cross-linking were unknown. In contrast, it was seen in the present study that the introduction of one or two interdimer cross-links between residues within the bait region leads to a blockage of the thiol ester cleavage-induced conformational change, as evidenced both by incomplete conversion from slow to fast forms after cleavage of the thiol esters by methylamine and by the small fraction of each variant that shows near-normal conformational change reported by changes in TNS fluorescence. Even when conformational change is brought about by proteinase cleavage of the bait region, there is evidence from the greatly lowered efficiency of proteinase trapping (Table I) that the structural rearrangements may have been slowed down. Such changes in efficiency of proteinase trapping as a function of the relative rates of cleavage and conformational change have been documented for plasma ␣ 2 M and trypsin (27,28). This suggests that when the thiol esters alone are cleaved, the conformational change that occurs cooperatively between noncovalently interacting subunits across the dimer-dimer interface involves a reorganization of the bait region-bait region contact area. By introducing a disulfide between such pairs of bait regions, as in any of the four variants considered here, this reorganization is either halted or greatly slowed down. The observation that this blockage of movement is much less when the bait region is cleaved by proteinase, such  (1,29). The enlargement on the right is a stylized depiction of the four bait regions and how they might interact. A1 and A2 are bait regions from the same dimer, as are B1 and B2. Interactions between A1 and B1 and between A2 and B2 represent noncovalent contacts at the dimer-dimer interface. It is envisioned that residues Gly-679, Val-700, and Thr-705 lie in the region of greatest contact (hatched), whereas Met-690 may lie outside this region, in a position (marked with an asterisk) at which it is close both to a bait region from a monomer of the second dimer and to the bait region from the second monomer of the same dimer. that all of the ␣ 2 M is in the fast form in less than 10 min, is quite consistent with and even supportive of such a model, in that the cleavage by proteinase, which is specifically restricted to the bait region, could isolate the disulfide cross-link from the parts of the bait region that need to move relative to one another, thereby freeing them to reorient and thus abrogating the blockage of conformational change.
Our proposed location of the four bait regions at the center of the ␣ 2 M tetramer in the position of the cavity body found in the crystal structure of methylamine-transformed ␣ 2 M is also consistent with the changes in TNS fluorescence caused by proteinase treatment of the variants and of the reduced activity of trypsin in complex with these ␣ 2 M variants. The location and size of the cavity body in the crystal structure of methylaminetransformed ␣ 2 M is such that it would have to be removed in whole or in part to accommodate one or two proteinase molecules in the central cavity. If the bait regions constitute all or part of the cavity body, their cleavage by proteinase and the consequent reorientation of the interface, as suggested by results presented here, could constitute such a restructuring. For the four variants, we found that TNS fluorescence spectra not only were different from either native or transformed plasma ␣ 2 M but also were much less responsive to change upon reaction with proteinase than was plasma ␣ 2 M. This was particularly so for the M690C and T705C variants (Fig. 4). These are the variants that also show the lowest specific activity for trypsin that is trapped ( Table I), suggesting that the trypsin may be very much more constrained in these variants than in plasma ␣ 2 M. The G679C and V700C variants show somewhat higher specific activity for trapped trypsin and also give TNS fluorescence spectra for the proteinase-treated forms that are more like that of proteinase-treated plasma ␣ 2 M. These findings for the four variants are consistent with bait region cleavage giving less complete interface restructuring than for plasma ␣ 2 M as a result of the constraints imposed by the interdimer disulfides, with the consequence that the cavity body is not cleared out of the way as completely as in plasma ␣ 2 M. This could also be the explanation for the lowered trapping ability of the variants, in that there is less volume available to form the trap.
In conclusion, we have provided evidence that all four of the bait regions of human ␣ 2 M are in contact with or are very close to one another, probably in the central cavity of the protein, and that the major conformational change that is required for proteinase trapping involves a reorientation of the bait regionbait region contacts both to bring about the trapping conformational change and to make room in the central cavity to accommodate the proteinase. Introduction of cross-links across this interface, either by disulfide formation, as described here, or by chemical cross-linking, as suggested elsewhere (25,26), slows or prevents this reorientation with consequent reduction in rate of conformational change and in efficiency of proteinase trapping.