Three High Molecular Weight Protease Inhibitors of Rat Plasma ISOLATION, CHARACTERIZATION, AND ACUTE PHASE CHANGES*

Rat blood plasma contains three high molecular weight thiol ester-containing proteinase inhibitors, al-macroglobulin (alM), al-inhibitor I11 (al13), and a2- macroglobulin (a2M). Rat serums have been analyzed using a two-dimensional gel electrophoretic technique which optimizes recovery of high molecular weight proteins. alM, an (afl)4-tetramer in native solution, separated in the second sodium dodecyl electrophoretic dimension as a disulfide-linked (~xfl)~-dimer with an approximate M, of 360 kDa. al13 separated in the gels as a single 190-kDa polypeptide. It is also a monomer in native solution by ultracentrifugation criteria. Native rat a2M is a tetramer, but it separates in the gels as a disulfide-linked dimer with an M, of approximately 360 kDa. The kinetics of changes in concentration of these proteins during the of polyarthritis was also measured by quan- In rats adjuvant-induced the of al13 matically and a2M appears and continues to in a biphasic for The

Rat blood plasma contains three high molecular weight thiol ester-containing proteinase inhibitors, almacroglobulin (alM), al-inhibitor I11 (al13), and a2macroglobulin (a2M). Rat serums have been analyzed using a two-dimensional gel electrophoretic technique which optimizes recovery of high molecular weight proteins. alM, an (afl)4-tetramer in native solution, separated in the second sodium dodecyl sulfate-containing electrophoretic dimension as a disulfide-linked (~xfl)~-dimer with an approximate M, of 360 kDa. al13 separated in the gels as a single 190-kDa polypeptide. It is also a monomer in native solution by ultracentrifugation criteria. Native rat a2M is a tetramer, but it separates in the gels as a disulfide-linked dimer with an M, of approximately 360 kDa. The kinetics of changes in concentration of these proteins during the induction of polyarthritis was also measured by quantitative immunoelectrophoresis. In rats with adjuvantinduced polyarthritis, the concentration of al13 dramatically decreases and a2M appears and continues to increase in a biphasic manner for 2 weeks. The alM concentration remains relatively constant.
All three macroglobulins were purified utilizing modern rapid chromatographic techniques, and parallel comparisons of their native physicochemical properties were carried out. The N-terminal sequence of the a-chain of rat a1M was also shown to share sequence homology with that of a2M. In agreement, Esnard et d. (Esnard, F., Gutman, N., El Moujahed, A., and Gauthier, F. (1985) FEBS Lett. 182,[125][126][127][128][129] recently reported that alIs also contains a thiol ester bond, as do alM and a2M, since it reacts covalently with [14C]methylamine and is cleaved autolytically at 80 "C. We have examined negatively stained preparations of native, trypsin-treated, and methylamine-treated human a2M, rat a2M, and rat a l M in the electron microscope. Trypsin appears to convert globular ring-shaped native molecules to rectangular box-like structures, in agreement with the conclusions of a recent report on human aaM (Tapon-Bretaudiere, J., Bros The acute phase changes in rat blood following induction of adjuvant arthritis (1) were examined by a "low resolution" two-dimensional gel electrophoresis method (2) because it has been known for some time that major changes in serum * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ll To whom correspondence should be addressed Marshfield Medical Foundation, 510 N. St. Joseph Ave., Marshfield, WI 54449.
composition coincide with development of this disease (3)(4)(5). This method reveals most of the plasma proteins with concentrations greater than 0.05 mg/ml, including those with high molecular weight. The concentration of aZM,l which increases by a factor of a hundred or more in injured rats (6)(7)(8)(9), also increases dramatically in adjuvant arthritis (3,4). The rat has two related macroglobulins, alM and azM. Both are protease inhibitors (7,8), both appear to contain labile thiol esters, and both have been biochemically characterized (8, 10,11). Rat alM undergoes only a small increase in concentration during inflammation and, like human a2M, it is not classified as an acute phase reactant. However, rat a2M is antigenically (12) and structurally (11,13) more closely related to human azM than is rat aIM.
A very prominent feature of two-dimensional gel patterns was a striking decrease in concentration of the most abundant single globulin species in healthy rat blood during development of arthritis. The existence of this 2 x lo5 M, globulin in rat blood was first reported by Gauthier and Ohlsson (14). It was characterized as alIa (15, 16), and recently evidence has also been presented indicating that it contains a thiol ester, like azM and alM (17).
It has recently come to our attention that Saito and Sinohara (18) have also described a rat "murinoglobulin" trypsin inhibitor containing a thiol ester group and which is probably identical to a113. Our antisera to al13 was also found to react with the protein isolated by their procedure.' The three rat protease inhibitors, alM, azM, and al13 appear to be related proteins on the basis of size, evidence for the presence of a thiol ester, and a requirement for proteolytic cleavage to activate the thiol ester. Included in the same group are macroglobulins isolated from fish (19, ZO), reptiles (21,22), birds (23,24), a number of mammals (11,(25)(26)(27)(28), and possibly even invertebrates (29,30). The complement proteins C3 and C4 also belong in this superfamily because they are activated by proteolytic cleavage, contain thiol esters, and share sequence homologies with human aZM (31). Although a salient feature of members of the azM family is the presence of a thiol ester in each subunit, it is noteworthy that C5, which has none (32), should also be included in this family since preliminary evidence suggests that its primary structure resembles that of C3 and C4 (33). L. Sottrup-Jensen, personal communication.
than 7 X IO5. Early electron microscopic studies described ultrastructural changes in human aZM which supported the "trap" hypothesis. In this hypothesis a conformational change follows proteolytic cleavage of a susceptible "bait region"; contraction of four arm-like projections produces a box-like structure which entraps the protease (34-36). Rat ~z M and rabbit alM were also reported to have arm-like projections which contracted following treatment with trypsin (37). However, recently published evidence suggests that the native form of azM is more globular (38). Another report on the ultrastructure of a crocodilian azM also shows its native form to be globular (21).
This report presents a parallel comparison of the changes occurring during inflammation, new isolation procedures, the subunit composition, and the ultrastructure of human aZM, rat aZM, alM, and (~113. (See "Methods," "Other Rat Proteins" for details on identification.) Fig. 2 shows two-dimensional gel analysis of serum samples from an individual rat at several time intervals following intraperitoneal injection of 0.2 ml of complete adjuvant. Acute phase proteins are indicated in the 0-h frame. An important feature of this analysis is that major proteins with denatured M , of up to 400 kDa are recovered virtually completely in the gels. Careful examination of two-dimensional gels showed that not all changes were in synchrony. For example, spot 6 is more pronounced than spot 2 at 24 h. Spot 1 (aZM) also decreases to near normal levels by 72 days, while spots 2, 3, and 6 remain elevated at that time (data not shown).

Fig
The kinetics of change in spots 1, 11, and 18 ( Fig. 2) in a rat blood sample during the development of adjuvant induced arthritis are shown in Fig. 3. The proteins were quantitated by rocket immunoelectrophoresis of azM, alM, and a113. There are reciprocal changes in azM and ~1113, while alM remains relatively constant, increasing by a factor of less than 2. Similar patterns were found with other individual rats (not illustrated). There are two phases of the acute phase reaction following injection with adjuvant. The first peaks by day 3 and this is followed by a period of partial recovery. This is then followed, after day 10, by increased levels of acute phase reactants and the development of polyarthritis as evidenced by swelling of joints in all extremities. Partial recovery may be observed in surviving animals after a few months. In animals injected with bacterial endotoxin (25 pg/kg intrave-Portions of this paper (including "Materials and Methods," Table   I  nously or 250 pg/kg intraperitoneally), maximum acute phase changes are seen at day 2, and these resemble changes seen 2 days following adjuvant; however, recovery is almost complete within a week (data not shown).
Rat azM was purified from the plasma of rats with adjuvantinduced polyarthritis, and alM and al13 were purified from healthy rat plasma (see "Methods"). Rabbit antisera prepared against the purified proteins showed no cross-reactivity between the three rat macroglobulins (data not shown). In contrast to the report by Saito and Sinohara (18), we found that rabbit antisera to a113 did not cross-react with alM, providing that the immunogen was highly purified.
Rat aZM and human azM showed partial identity when compared by Ouchterlony immunodiffusion using antibody to either human or rat aZM (not shown). Cross-reactivity of human and rat aZM has also been reported (12), as has lack of cross-reactivity of rat aZM and alM (6). Antibody to rat aZM also does not react with normal serum containing aIM (9, 10). Rabbit antisera to a113 did not react with human plasma proteins, but two rabbits immunized with purified human inter-a-inhibitor ( I d ) showed very faint precipitation bands with purified rat ~~113, but only when using undiluted serum (not shown). Those faint bands were too indistinct to confirm a pattern of partial immunological identity. Thus, the early observation by Gauthier and Ohlsson (14) of a possible immunological cross-reactivity between these proteins is unresolved and deserves further study. We also confirmed by peptide mapping that the three rat proteins each have a unique primary structure (see "Methods") (data not illustrated). Saito and Sinohara (18) reported similar results. Fig. 4 shows the SDS-PAGE separation of the subunits of purified human and rat azM, rat alM, and culIa. Each protein has a single-chain subunit, except alM which has an a-chain and a @-chain. The M , of these were estimated repeatedly from similar separations to average 167 kDa (human a2M), 159 kDa (rat azM), 143 and 42 kDa (rat alM), and 183 kDa (rat a113). However, there is internal evidence from the sum of the molecular weights of cleavage fragments that our electrophoretic gels systematically underestimated the molecular weights of proteins larger than 150 kDa, and we propose "model" molecular weights several percent higher? Table I1 K.   . . .

A T G K P * Y V V L V P S E L Y A G V P . . .
gives the apparent M, and model M, for each chain.
The N-terminal sequence of the a-chain of alM was also determined and compared with the known sequences of the N termini of human and rat azM (Table 111). Selected regions of the a-chain of alM showed 77% homology (residues 3-15) and 65% homology (residues 1-17) to human and rat aZM, respectively.
The  (Table 11). The frictional ratio was also calculated to be 1.5 from the sedimentation coefficient (51). This indicated an axial ratio of 10 for a prolate elipsoid and may explain why we obtained a high M, by gel filtration. In agreement with Esnard and Gauthier (15), we conclude that purified native al13 is probably a monomer. The three rat proteins share the ability to react covalently with the nucleophile methylamine. This is illustrated in Fig.  5, where aliquots of sera were reacted with excess [14C]methylamine and analyzed by two-dimensional gel electrophoresis. There was approximately 0.5% incorporation into trichloroacetic acid-precipitable counts, and a significant portion of this was specifically incorporated into aIM and al13 in normal rat serum (Fig. 5 B ) or into a2M and alM in arthritic serum ( Fig. 5 D ) . Presumably there was also incorporation into C3, but C3 was degraded to C3b and other products under the conditions of incubation. A small amount of nonspecific labeling of albumin is also discernible in Panel B. Comparison with the Coomassie Blue-stained wet gels (Fig. 5 , A and C) shows that the differences in incorporation parallel the differences in relative concentration of alM and a2M, and aIL.
Methylamine treatment prevents autolytic cleavage of a113 as illustrated in Fig. 6. Similar kinetics of autolytic cleavage and prevention by prior reaction with methylamine were found for human a2M, rat azM, and rat alM (not illustrated). Human IaI did not undergo autolytic cleavage under these same conditions (data not shown).
The macroglobulins may be visualized in the electron microscope. Interpretation of the images is aided by the fact that they are tetramers, containing repetitive substructure.
Difficulty was experienced in getting the native purified macroglobulins to adhere reproducibly to the carbon-coated Formvar grids used for negative staining, and this was not corrected by use of uncoated grids, or a-particle or ultraviolet lightirradiated grids and/or substitution of uranyl acetate for phosphotungstic acid negative stain. Likewise, prior treatment of grids with glutaraldehyde or pre-exposure of grids to glutaraldehyde vapor or exposure of grids wetted with samples to glutaraldehyde vapor did not improve the reproducibility of adherence. In general, adequate adherence was found in only about 1 grid in 12. We do not know of a specific cure for this problem, which may reflect the hydrophilic properties of the glycoproteins. Whenever purified native macroglobulin adhered to the grids, they appeared as globular or ring-like structures, as seen for human and rat a,M and rat alM, in Fig. 7A.
In contrasts to the native macroglobulins, trypsin-treated samples adhered virtually every time. However, trypsintreated samples showed some structural variation which could not be explained by either the length of time between trypsin treatment and staining (0 to 48 h at 4 "C) or the molar ratio of trypsin to inhibitor (1, 2, or 4). Some of the trypsin complexes resembled the Cyrillic letter x, but usually they resembled tighter box-like structures. The box-like trypsin complexes predominate in the fields shown in Fig. 7

B .
Ultrastructural changes, following treatment of the macroglobulins with methylamine, were not uniform. Human azM and rat alM were converted to forms which resembled the trypsin complexes and which adhered well to the grids; however, rat azM was unchanged. Addition of trypsin to methylamine-treated rat apM converted it to structures indistinguishable from those of complex between trypsin and native aZM (not illustrated).

DISCUSSION
This report examines the kinetic behavior of alM, a2M, and a113 plasma protease inhibitors during severe inflammation and places these changes within the context of changes in other major rat plasma proteins. This is the first detailed comparison of the behavior of these three proteins and clearly shows a reciprocal relationship between the concentrations of al13 and azM. We have also isolated these proteins and run parallel comparisons on some of their properties and provided evidence supporting the contention that they all contain the unusual thiol ester bond. The electron microscope was also used to confirm the structural relationship between two of the rat proteins (alM and a2M) and human a2M.
The low resolution two-dimensional gel electrophoresis method we have used to study the acute phase changes has the important advantage that most plasma proteins including the high molecular weight polypeptides are recovered virtually quantitatively: This method showed a113 to be the most abundant single globulin in the healthy rat and also the major negative acute phase protein (Fig. 2). Rat azM was revealed as a major positive acute phase protein, together with a number of other globulins including haptoglobin, ceruloplasmin, and the proteins of spots 2 and 6 which we have not characterized. It is impossible that our spot 2 corresponds to the rat a,-cysteine protease inhibitor of Esnard and Gauthier (55), based upon electrophoretic mobility in agarose.
Identification of alIB as still another plasma protein containing a putative thiol ester bond is of great interest. Auto-Lonberg-Holm, K., Sandberg lytic cleavage at 80 "C was prevented by prior treatment with methylamine as shown for aI13 in Fig. 6. Esnard et al. (17) also reported autolytic cleavage of al13 and its inhibition by methylamine. This, together with the ability to bind labeled methylamine (Fig. 5 ) , suggests that aI13, like the a-macroglobulins and C3, contains a thiol ester (56).
Both native alM and native azM are tetramers, and our model M , is close to reported values obtained by ultracentrifugation (8, 57). Our values for the sum of the alM a-and pchains are about 9% lower than the sum of the values reported by Nelles and Schnebli (11) and about 6% lower than values reported by Schaeufele and Koo (13). The p-chain in our model constitutes 22.5% of the total as opposed to 18.5% and 21.5% as reported by these authors. Rat a113 is probably a monomer, as proposed by Esnard and Gauthier (15). We found a113 to have a high (380 kDa) M, by gel filtration chromatography, but an M , of about 200 kDa by using two ultracentrifugation methods. The later values are independent of the shape of a113 molecule, while the former may be an artifactual result of the high axial ratio of 10 which we calculated from the ultracentrifugation data. Our 190-kDa model M, for a113 is about 12% lower than the value reported by Esnard and Gauthier (15). The differences between model and various experimentally derived M, are probably less important than errors caused by carbohydrate contents, and establishment of the correct sizes of the chains will require sequencing.
Human and rat azM show patterns of immunological partial identity and are probably structurally closely related proteins, as also supported by N-terminal sequence analysis (58) and by partial internal sequencing (59). the a-chain of aIM is apparently derived from the N-terminal part of a precursor to both chains because it shares significant sequence homology with the N terminus of rat a z M . The P-chains of both C3 and C4 are N-terminal in their uncleaved precursors and also share homologies with the N terminus of a,M (31). The N terminus of a113 also has been found to share significant sequence homology with rat and human azM clearly placing a,13 within the macroglobulin family. 7 Human IaI was once reported to show antigenic crossreactivity with al13 (14). Both proteins have similar mobility and size in two-dimensional gel electrophoresis. However, IaI does not appear to contain a thiol ester. When the sequences of both IaI and a113 are known, it will be of interest to determine if they bear any homology in the same manner that C5, which has no thiol ester, is homologous to the thiol ester containing C4 (33).
Concentrations of alM, a 2 M , and al13 during induction of adjuvant arthritis were measured by rocket immunoelectrophoresis (Fig. 3). The observed biphasic kinetics of induction of aZM is similar to that found by the semiquantitative "titer" method reported by Bogden et al. (4). The high levels obtained in Fig. 3 (8 mg/ml) are about twice those usually found L. Sottrup-Jensen, personal communication to be reported in detail elsewhere.
following injection of turpentine into rats (8, 9,60); however, some rats injured by turpentine attain levels up to 11 mg/ml (7). We found that serum from normal male rats contained less than 50 wg/ml a z M , in agreement with many other estimates. In Fig. 3 the initial level of alM (2 mg/ml) was somewhat lower than the level of 3-4 mg/ml reported by Gordon (8), but in some rats we found about 3 mg/ml. There is a small increase in alM during inflammation.
Rat alIs has a very high normal concentration of 7-9 mg/ ml and it undergoes a very dramatic 15-fold biphasic decrease during induction of arthritis. This is much greater than a decrease to 36% found following injection of turpentine (14) or a decrease to about 60% normal levels following intravenous injection of bacterial endotoxin (data not shown) and reflects the severe and systemic nature of adjuvant arthritis. Saito and Sinohara (18) reported that the normal level of rat murinoglobulin concentration is 14 mg/ml. This protein is probably identical to aJ3, as we have described, and the difference in concentration may reflect either differences in the rats or methodological details.
The conformational changes accompanying activation and cleavage of the thiol ester bonds of human azM and the rat macroglobulins were visualized in the electron microscope by negative staining (Fig. 7). Fig. 8 is a diagrammatic representation of some of the structures observed. The most commonly seen product of trypsin treatment is represented by form B, while a variable amount of form C (the Cyrillic letter x) was also detected, as well as other more compact square forms which are not illustrated. This confirms the recent report on human azM by Tapon-Bretaudier et al. (38), who interpreted B and C as representations of two side views of the Same structure. Treatment with methylamine induced similar conformational changes in human azM and rat alM, while rat a,M remained unaltered. This agrees with other evidence (24) that the rat protein differs from its human homologue in methylamine reactivity.