Alzheimer’s Peptide Aβ1–42 Binds to Two β-Sheets of α1-Antichymotrypsin and Transforms It from Inhibitor to Substrate*

The serpin α1-antichymotrypsin is a major component of brain amyloid plaques in Alzheimer’s disease. In vitroα1-antichymotrypsin interacts with the Alzheimer’s amyloid peptide Aβ1–42 and stimulates both formation and disruption of neurotoxic Aβ1–42 fibrils in a concentration-dependent manner. We have constructed a new hybrid model of the complex between Aβ1–42 and α1-antichymotrypsin in which both amino and carboxyl sequences of Aβ1–42 insert into two different β-sheets of α1-antichymotrypsin. We have tested this model and shown experimentally that full-length and amino-terminal segments of Aβ1–42 bind to α1-antichymotrypsin as predicted. We also show that Aβ1–42 forms both intra- and intermolecular SDS-stable complexes with α1-antichymotrypsin and that the binding of Aβ1–42 to α1-antichymotrypsin abolishes the inhibitory activity of the latter and its ability to form stable complex with chymotrypsin. The existence of both inter- as well as intramolecular complexes of Aβ1–42 explains the nonlinear concentration-dependent effects of α1-antichymotrypsin on Aβ1–42 fibril formation, which we have reinvestigated here over a broad range of Aβ1–42:α1-antichymotrypsin ratios. These data suggest a molecular basis for the distinction between amorphous and fibrillar Aβ1–42 in vivo. The reciprocal effects of Aβ1–42 and α1-antichymotrypsin could play a role in the etiology of Alzheimer’s disease.

␣ 1 -antichymotrypsin (ACT) 1 is a serpin serine proteinase inhibitor with specificity for cathepsin and chymotrypsin-like enzymes (1). Its physical properties (2), mechanism (3), and structure (4,5) conform to the general model (6,7) that has been established for serpins. These inhibitors present a flexible reactive site loop, which is cleaved at a susceptible residue by target proteinase (8,9) to form a highly stable, covalent complex (10,11). As in other serpins (12), the reactive site loop in the cleaved form of ACT is inserted into a preexisting, flexible ␤-sheet (sA) (4), and this cleaved form is stabilized to denaturation relative to the uncleaved native form (2). There is evidence (13,14) consistent with proposals (10,11) that insertion of the reactive site loop into sA is a prerequisite for stable proteinase-serpin complex formation. In further support of this, it has been shown that exogenous peptides of restricted sequence can insert into sA of the uncleaved serpin (15)(16)(17). These binary peptide⅐serpin complexes are similarly increased in stability to denaturation but have lost inhibitory activity, becoming substrates instead that do not form a stable complex with target proteinase.
A possible link between ACT and Alzheimer's disease was established by the observation that ACT occurs in the senile plaques characteristic of this disease (18). Subsequently, it was shown in vitro that ACT can either stimulate formation of the neurotoxic fibrillar form of Alzheimer's peptide A␤   (19,20) or destabilize preformed A␤ 1-40 fibrils (21,22), depending on the stoichiometry of ACT to A␤. Two models for the binding of A␤  to ACT (20,23), based on the known structure and properties of the latter, suggested how amino and carboxyl segments of A␤ 1-42 might bind into different ␤-sheets of ACT. These two models are not mutually exclusive, and we have synthesized them into a new model presented here in which A␤ 1-42 binds to ACT through insertion of amino-and carboxylterminal segments of A␤ 1-42 into different ␤-sheets of ACT. We have experimentally tested this model and its implications to understand the nature of ACT⅐A␤ 1-42 complexes and their possible relationship to the polymerization of A␤ 1-42 , which is closely associated with Alzheimer's disease.
We show here that full-length A␤ 1-42 and also its aminoterminal segments (A␤ 2-9 and A␤ 1-11 ) bind to ACT. This supports our bimodal model for the binding of A␤  to ACT in which the amino-terminal segment of A␤ 1-42 inserts into sC of ACT and the carboxyl-terminal segment into sA. The binding of A␤  to ACT results in loss of ACT inhibitor activity and its transformation from an inhibitor into a proteinase substrate, an observation consistent with the effects of the insertion of other peptides into sA of serpins. The bimodal model for the intramolecular 1:1 ACT⅐A␤ 1-42 complex also suggested that intermolecular complexes of higher molecular weight between these molecules may form. We present experimental evidence for the occurrence of complexes of higher multiplicity as well as the 1:1 intramolecular complex. This model for independent, bimodal insertion of the amino-and carboxyl-terminal sequences of A␤ 1-42 into ACT also offers a structural basis for the distinct effects of low and high concentrations of ACT on A␤ 1-42 fibril formation and stability. We have tested this hypothesis with measurements of the polymerization of A␤ 1-42 at a broader range of concentrations of ACT than has been examined before, and the results are described here.  (Saveen, Denmark), A␤ 2-9 (Commonwealth Biotechnologies), and A␤ 1-11 (Bachem) were Ͼ98% pure. 125 I-A␤ 1-42 and 125 I-A␤ [1][2][3][4][5][6][7][8][9][10][11] were made by the method of Thorell and Johansson (24) using chloramine T, purified on a Sephadex PD-10 column (G25M), and stored in phosphate-buffered saline, pH 7.4, containing 0.02% sodium azide.

Complexes of A␤ with ACT (Recombinant)-A␤
Recombinant ACT was purified as described (25). A␤ 1-42 was solubilized in trifluoroacetic acid and lyophilized before use. An aliquot of lyophilized A␤ 1-42 equal to 100-or 200-fold molar excess over ACT was dissolved in sterile 0.015 M Tris, pH 7.4, 0.15 M NaCl and combined with ACT (29.7 M). This mixture was incubated under sterile conditions for various time periods from overnight to 3 weeks. In competition experiments with A␤ 2-9 , ACT was first incubated overnight or up to 3 days with a 200-fold excess of A␤ 2-9 followed by addition of a 100-fold or 200-fold excess of A␤ 1-42 over ACT, and the mixture incubated further. 125 I-labeled A␤ 1-11 was used to observe the complex of the aminoterminal peptide with ACT. Labeled A␤ 1-11 was mixed with unlabeled peptide to a final specific activity of approximately 2 mCi/mmol and incubated with ACT at a molar ratio of 200:1 (peptide:ACT) in phosphate-buffered saline for 24 and 48 h at room temperature. To control for the possibility of oxidative damage to ACT by free radicals generated by A␤ peptides, we tested the effect of prolonged incubation with 10 M peroxide on its inhibitor activity and found no loss. ACT is not susceptible to oxidative damage, probably reflecting its function as an acute phase inflammatory reactant that must survive the presence of reactive oxygen species in vivo.
SDS and Agarose Gel Electrophoresis, Autoradiographs, and Western Blot-The mixture of ACT with 125 I-A␤ 1-11 was run on a 1% agarose gel in Tris barbiturate buffer at pH 8.6. Samples of ACT incubated with A␤ 1-42 in the presence and absence of A␤ 2-9 were run on 12% SDSpolyacrylamide gels. Samples were mixed with SDS sample buffer (0.5 M Tris-HCL, 10% glycerol, 10% SDS, 0.05% (w/v) bromphenol blue) and heated at 90°C for 5 min before running. Duplicate samples of the longest incubation were run identically in the presence of ␤-mercaptoethanol, with no change in pattern.
Autoradiographs were made by exposure of high performance autoradiography film (Hyperfilm MP, Amersham Pharmacia Biotech) to the gel overnight at Ϫ70°C. Western blot transfers to Immobilon-P membrane (Amersham Pharmacia Biotech) were incubated with rabbit anti-ACT (DAKO, Denmark) followed by peroxidase-labeled pig anti-rabbit antibody and developed with 3,3-diaminobenzidine tetrahydrochloride (Sigma). Complex formation was detected by coincidence of the autoradiograph peptide band with the ACT band from the Western blot.
Proteinase Digestion of ACT and ACT⅐A␤  Complexes-After incubation, the A␤ 1-42 -ACT mixture and an ACT control were digested with chymotrypsin at a 1:1.2 molar ratio to ACT for 15 min. Digestion was stopped by adding SDS sample buffer, and samples were run on 12% SDS-polyacrylamide gel without reducing agent and stained with Coomassie Blue. Control samples of cleaved, uncomplexed ACT were obtained by digestion with collagenase (Cooper Biomedical) at a molar ratio of 500:1 (ACT ϩ protease) for 2 h at 37°C.
ACT Inhibitor Activity Measurement-ACT (1.9 M) in 50 mM Tris, pH 8.0, alone or preincubated as above with A␤ 2-9 and/or A␤ 1-42 was then incubated with chymotrypsin (1.3 M) for 5 min at room temperature. Substrate succinyl-Ala-Ala-Phe-p-nitroanilide (Sigma) was added immediately to 0.18 mM and residual activity of chymotrypsin measured at 405 nm for 3 min. The mean of three separate measurements is expressed as the percent of full ACT inhibitor activity (with indicated standard error) in the absence of added peptide.
Model of ACT⅐A␤  Complex-The structure of cleaved native ACT was used as a starting point, because only in the cleaved serpin is the complete insertion of strand s4A observed. Our assumptions were that ACT strands s1C and s4A were completely displaced by A␤ 1-42 residues Gly 9 -Glu 3 and Lys 28 -Ala 42 , respectively. There are no constraints in this model for the structure of the strand s4A reactive loop strand and strand s1C of ACT and for residues 10 -27 of the bound A␤ 1-42 in the complex, both of which are depicted by a dotted line, as shown in Fig. 3, with unknown structure.

RESULTS
Direct demonstration of complex formation between ACT and A␤ 1-42 was obtained from SDS gels of 1:100 (ACT:A␤ 1-42 ) mixtures incubated for 3 days. A slightly higher molecular weight (lower mobility) band is observed with the expected molecular weight of an SDS-stable 1:1 complex of A␤ 1-42 with ␣ 1 -antichymotrypsin (Fig. 1A, lanes a and b).
Our earlier study showed that the carboxyl-terminal segment of A␤ 1-42 binds into sA of ACT (20) and that full-length A␤ 1-42 is necessary for this binding. This implied that the amino-terminal segment of A␤ 1-42 contributes to complex stability and model building showed how residues 2-9 of A␤ 1-42 could insert into sC of ACT (23). Confirmation of the binding of A␤ 2-9 was obtained from competition experiments in which A␤ 2-9 was shown to block binding of full-length A␤ 1-42 peptide. Preincubation of ACT with A␤ 2-9 (Fig. 1A, lane f) followed by incubation with A␤  showed no formation of the higher molecular weight (lower mobility) ACT⅐A␤ 1-42 band. If ACT is incubated first with A␤ 1-42 followed by A␤ 2-9 , the higher molecular weight band is observed (Fig. 1A, lane e), indicating that the ACT⅐A␤ 1-42 complex is stable once formed in prolonged incubation. The low molecular weight (high mobility) band in lane e is unreactive with ACT antibodies in the Western blot and is an A␤ species. Control incubation of ACT with A␤ 2-9 did not detectably change the mobility of ACT from that of the native (Fig. 1A, lane d), and the possibility of confounding interactions between A␤ 2-9 and A␤ 1-42 was eliminated by showing that a preincubated (overnight) equimolar mixture of these two peptides did not alter the effects of A␤ 1-42 on ACT (data not shown).
The complex between the amino-terminal segment of A␤ 1-42 and ACT inferred from the above competition experiments was directly confirmed by separation of the complex of radioiodinated A␤ 1-11 with ACT on agarose gel. Fig. 1B shows the autoradiogram of a 1% agarose gel in which the incubated mixture of ACT with 125 I-labeled A␤ 1-11 was run. The band for A␤ 1-11 detected by autoradiography is coincident with that for ACT. We conclude from these experiments that the aminoterminal segment as well as the carboxyl-terminal segment of A␤ 1-42 binds to ACT and that both are essential for formation of stable complex.
Variations in time of incubation and stoichiometry of A␤ 1-42:ACT result in different forms of the complexes between these two molecules. A stable high molecular weight (low mobility) band having the approximate molecular weight of an ACT dimer (Fig. 1C, lane c) is observed on prolonged (3 weeks) incubation with A␤ 1-42 . This dimer is stable to reduction by ␤-mercaptoethanol and is blocked by preincubation with A␤ 2-9 (Fig. 1C, lane f). The identity of this complex was confirmed in the gels by autoradiography, which detected 125 I-A␤ 1-42 complexes with ACT (Fig. 1D, band 2 ϭ dimer), and by Western blot using anti-ACT antibody (data not shown) to detect coincident ACT. This autoradiograph also shows that smaller amounts of higher polymers of the ACT⅐A␤ 1-42 complex are formed (Fig. 1D, band 3). We conclude that A␤ 1-42 can link two molecules of ACT into a highly stable closed dimer through intermolecular bimodal binding to the two different ␤-sheets, sA and sC, in different molecules and that this is a slow process. The blocking of ACT⅐A␤ 1-42 complex formation by A␤ 2-9 confirms the requirement for complex formation of binding of the amino terminus of A␤  . This experiment parallels the diminished neurotoxicity and A␤ 1-42 fibril formation of ACT⅐A␤ 1-42 observed in the presence of A␤ 2-9 (27), suggesting that the complex we have characterized here is relevant to the neurotoxic and fibrillogenic properties observed by others toward cells in cultures. Shorter incubation time (24 h) of A␤  with ACT is insufficient to produce SDS stable complexes (Fig.  1E, lane c versus Fig. 1A, lane b). We infer that formation of the full bimodal complex is necessary to observe SDS-stable complex and that the kinetics for full complex formation are multistep and slow.
The mode of binding of A␤ 1-42 to ACT proposed here leads to the prediction that ACT in complex with A␤ 1-42 will lose its proteinase inhibitor activity. Inhibition of serine proteinase activity by serpins occurs with formation of an irreversible proteinase-serpin complex whose structure is stabilized by insertion of the proteinase-linked reactive site loop into sA. If A␤ 1-42 is inserting into the same site in sA as the reactive site loop, we predict that ACT in complex with A␤ 1-42 will lose its proteinase inhibitor activity. Fig. 1E shows that overnight pre-incubation of ACT with A␤ 1-42 abolishes formation of stable complex between ACT and chymotrypsin. Fig. 1E, lane b, shows the higher molecular weight inhibitory complex formed on incubation of ACT with chymotrypsin (band 1), whereas lane d shows that, when ACT is preincubated overnight with a 50-fold molar excess of A␤ 1-42 prior to incubation with chymotrypsin, no ACT⅐chymotrypsin complex is formed and only intact (band 2) and cleaved ACT (band 3), along with the cleaved low molecular weight carboxyl-terminal peptide band, are observed. This is consistent with our observations (see below) and earlier ones of other investigators (27) that incubation of ACT with A␤ 1-42 results in loss of ACT inhibitor activity. It also confirms our earlier conclusion that the stabilization of intact ACT to denaturation after incubation with A␤ 1-42 is because of insertion of A␤ 1-42 into sA as a substitute for s4A, the reactive site loop strand.
Direct assay of proteinase inhibitor activity of the ACT⅐A␤ 1-42 complex lent further weight to the interpretation of the above experiments. Preincubation of ACT with A␤ 1-42 results in a loss of 20% and 50% of ACT inhibitor activity versus chymotrypsin at molar concentration ratios of 1:100 and 1:200, respectively (ACT:A␤ 1-42 ) (Fig. 2, sample 3, light and dark gray bar, respectively). This effect parallels the loss of inhibitor activity of the serpins ␣ 1 -antitrypsin (15), antithrombin (16), and plasminogen activator inhibitor 1 (17) on incubation with exogenous peptides that can insert into sA. We exploited this assay system to further test whether the amino-terminal A␤ [2][3][4][5][6][7][8][9] binds in the complex between A␤ 1-42 and ACT. If A␤ 2-9 is essential for the binding of full-length A␤ 1-42 to ACT, then A␤ 2-9 should reverse the loss of inhibitor activity that occurs when A␤ 1-42 binds to ACT. Parallel assays of ACT inhibitor activity in the presence of A␤ 1-42 with and without overnight preincubation with A␤ 2-9 show such a reversal in the presence of A␤ [2][3][4][5][6][7][8][9] . When ACT is preincubated with A␤ 2-9 at a 200:1 molar ratio (A␤ 2-9 :ACT), followed by incubation with A␤ 1-42 at 100:1 and 200:1 (A␤ 1-42 :ACT) ratios, the loss of inhibitor activity is reduced to 5 and 20% from 20 and 50% (sample 4, light and dark gray). These data further support the model for interaction of A␤ 1-42 with ACT in which the amino-terminal A␤ 2-9 segment as well as the carboxyl-terminal segment of A␤ 1-42 bind to ACT.
The above experiments show that there are multiple complexes between ACT and A␤ 1-42 , whose stability is a function of incubation time. Formation of SDS-stable ACT⅐A␤ 1-42 complex is slow, like that of binary complex formation between serpins and exogenous peptides that bind in sA (15,16). Slow rates of complex formation may also reflect slow structural isomerizations that are known to occur in A␤ peptides converting from their soluble form to the ␤-rich fibrillogenic form (28), or the conversion of the ACT structure from the uncleaved, native conformation to the more stable A␤ 1-42 strand-inserted conformation, which occurs in cleaved and proteinase-complexed serpins.
The model for bimodal insertion of A␤ 1-42 into ACT (Fig. 3), which was the hypothetical basis for the above experiments, is validated by these results. Our earlier observation that A␤  interaction with ACT stabilizes ACT to denaturation implied that A␤ 1-42 is inserting into sA of ACT, a property already observed in several serpins. Our results here further confirm this interpretation by showing that ACT loses its inhibitor activity as a result of A␤ 1-42 binding to ACT, another property previously associated with the insertion of exogenous peptides into sA of inhibitory serpins. Our earlier observation that fulllength A␤ 1-42 is necessary for stabilization of ACT to denaturation is now also explained by the proposed model incorporating the second binding site for the A␤ 2-9 segment of A␤ 1-42 to sC of ACT.
Bimodal insertion of A␤ 1-42 into ACT suggested a possible explanation for the concentration-dependent effects of ACT on A␤ 1-42 fibril formation and stability. ACT stimulates A␤ 1-42 fibril formation (19,20) at relatively low molar ratios of ACT to A␤ 1-40 or A␤ 1-42 , but at higher concentrations of ACT, preformed A␤ 1-40 fibrils are destabilized and show signs of disintegration (21,22). We revisited this phenomenon by determining the dependence of A␤ 1-42 polymerization on a broader range of ACT concentration than had been explored before. Using the thioflavin T fluorescence assay, we found that higher concentrations of ACT do not stimulate A␤ 1-42 polymerization and that there appears to be a threshold ratio of ACT to A␤  below which aggregate formation is stimulated (Fig. 4). Our earlier work and that of others (27) showed by electron microscopy that these aggregates are overwhelmingly fibrils. This threshold effect may be a consequence of a switch between intramolecular and dimeric and higher multiplicity intermolecular complexes of ACT with A␤ 1-42 . The intramolecular complex consisting of a single molecule of A␤ 1-42 bound through its carboxyl and amino terminus in sA and sC, respectively, of a single molecule of ACT (Fig. 3) is more likely to occur at lower ACT concentrations. The closed ACT dimer circularly linked in complex by two molecules of A␤ 1-42 would not aggregate, whereas open, noncircular, head-to-tail complexes formed through binding of different A␤ 1-42 molecules to multiple different ACT molecules, which is more likely at high relative ACT concentration, could immobilize A␤ 1-42 in aggregates. DISCUSSION We have shown here that A␤ 1-42 binds into two different ␤-sheets of ACT. Binding to one of these ␤-sheets, sA, abolishes the inhibitor activity of ACT, transforming it into a substrate. We speculate that the promotion of A␤ 1-42 fibril formation by ACT in vitro is a result of the imposition of ␤-sheet conformation on both amino and carboxyl-terminal sequences of A␤ 1-42 as a result of its binding to ACT. Fibrillar A␤ 1-42 is predominantly in ␤ conformation, but soluble, nonfibrillogenic A␤  at low concentrations appears to require a conformational isomerization to a more ␤-rich conformation to become fibrillogenic (28 -30). The conformational transition leading to polymerization may be the result of cooperative interactions among A␤ 1-42 molecules at high concentration (30) or could be pro-vided by a chaperone, such as ACT, whose complex with A␤ 1-42 we have modeled here.
Our observation of a nonlinear dependence of A␤ 1-42 fibril formation on ACT suggests that the relative amounts of A␤ 1-42 deposited in amorphous plaque or fibrils in Alzheimer's diseased brains may depend on fluctuating ratios of ACT and A␤  . Elevated A␤ 1-42 levels, which occur as a result of specific mutations in the amyloid precursor protein or of raised gene dosage as in Down's syndrome, correlate with early onset Alzheimer's disease (31). In the case of late onset Alzheimer's disease, ACT levels, which rise during inflammation, may also be an important determinant of the neurotoxic fibril load in the brain and therefore be an important parameter in the etiology of Alzheimer's disease.