Metastability in the Inhibitory Mechanism of Human α1-Antitrypsin*

Metastability of the native form of proteins has been recognized as a mechanism of biological regulation. The energy-loaded structure of the fusion protein of influenza virus and the strained native structure of serpins (serine protease inhibitors) are typical examples. To understand the structural basis and functional role of the native metastability of inhibitory serpins, we characterized stabilizing mutations of α1-antitrypsin in a region presumably involved in complex formation with a target protease. We found various unfavorable interactions such as overpacking of side chains, polar-nonpolar interactions, and cavities as the structural basis of the native metastability. For several stabilizing mutations, there was a concomitant decrease in the inhibitory activity. Remarkably, some substitutions at Lys-335 increased the stability over 6 kcal mol−1 with simultaneous loss of activity over 30% toward porcine pancreatic elastase. Considering the location and energetic cost of Lys-335, we propose that this lysine plays a pivotal role in conformational switch during complex formation. Our current results are quite contradictory to those of previously reported hydrophobic core mutations, which increased the stability up to 9 kcal mol−1 without any significant loss of activity. It appears that the local strain of inhibitory serpins is critical for the inhibitory activity.

Metastability of the native form of proteins has been recognized as a mechanism of biological regulation. The energy-loaded structure of the fusion protein of influenza virus and the strained native structure of serpins (serine protease inhibitors) are typical examples. To understand the structural basis and functional role of the native metastability of inhibitory serpins, we characterized stabilizing mutations of ␣ 1 -antitrypsin in a region presumably involved in complex formation with a target protease. We found various unfavorable interactions such as overpacking of side chains, polar-nonpolar interactions, and cavities as the structural basis of the native metastability. For several stabilizing mutations, there was a concomitant decrease in the inhibitory activity. Remarkably, some substitutions at Lys-335 increased the stability over 6 kcal mol ؊1 with simultaneous loss of activity over 30% toward porcine pancreatic elastase. Considering the location and energetic cost of Lys-335, we propose that this lysine plays a pivotal role in conformational switch during complex formation. Our current results are quite contradictory to those of previously reported hydrophobic core mutations, which increased the stability up to 9 kcal mol ؊1 without any significant loss of activity. It appears that the local strain of inhibitory serpins is critical for the inhibitory activity.
Facile conversion of the metastable native structure of proteins into an alternative more stable form, accompanying the execution of their functions, has been recognized as a mechanism of biological regulation. The energy-loaded structure of the fusion protein of influenza virus (1), the strained native structure of plasma serpins (serine protease inhibitors) (2), and possibly the surface glycoprotein of human immunodeficiency virus (HIV) 1 (3) are typical examples. The native strain of serpins is considered to be crucial to their physiological functions, such as plasma protease inhibition (2,4), hormone delivery (5), Alzheimer filament assembly (6,7), and extracellular matrix remodeling (8). The inhibitory serpins, which include ␣ 1 -antitrypsin (␣ 1 AT), antithrombin III, ␣ 1 -antichymotrypsin, and plasminogen activator inhibitor-1, serve as a good model system to study the native metastability; several crystal structures of both the strained native (9 -13) and the relaxed cleaved forms (14 -16) are available. In addition, the inhibitory activity that presumably is related to the native metastability is easy to assay.
The serpin structure is composed of three ␤-sheets and several ␣-helices (Fig. 1). Upon binding a target protease, the reactive center loop of inhibitory serpins is thought to be inserted into the major ␤-sheet, A sheet, to form a very stable complex between the inhibitor and the protease (17,18). Various biochemical (19,20) and structural (21)(22)(23) studies suggest that the loop insertion is necessary for the formation of a stable complex but not sufficient to confer inhibitory activity. Instead, the rate of loop insertion is thought to be critical for inhibitory function. The inhibition process of serpins can be described as a suicide substrate mechanism (17), in which serpins, upon binding with proteases, partition between cleaved serpins and stable serpin-enzyme complexes in a ratio represented by the stoichiometry of inhibition (SI, number of moles of inhibitors required to completely inhibit 1 mol of a target protease). The SI values of most inhibitory serpins are close to one for cognate target proteases. Retardation of the loop insertion, however, would alter the partitioning between the inhibitory and substrate pathways in such a way that the SI value would increase since the SI value is defined by 1 ϩ k substrate /k inhibition (17,18,20). It is conceivable that the energy loaded in the strained native structure of serpins is utilized for the facile loop insertion.
To understand the structural basis of the loaded energy in the native structure, we have been characterizing stabilizing amino acid substitutions of ␣ 1 AT, a prototype inhibitory serpin. We previously reported that decrease in the size of side chains at the hydrophobic core of ␣ 1 AT (Fig. 1, blue) confers increased stability (24). That was quite unusual, as a decrease in size inside the hydrophobic core usually yields a cavity that causes loss of stability (25). We proposed that side chain locking in the native ␣ 1 AT prohibits rearrangement of the side chains for maximal packing (24). Such structural defects are likely to be the basis of the native strain in the inhibitory serpins. If the energy loaded in the strained native structure of serpins is utilized for the inhibitory function, mutations that decrease the loaded energy should also decrease the activity. In the present study, we tested this concept of the native strain. We searched for mutations that increased the stability and simultaneously affected the inhibitory activity. We characterized stabilizing amino acid substitutions of ␣ 1 AT in a region that is presumably involved in the conformational change for the insertion of the reactive center loop during the inhibitory complex formation: strands 3 and 5 of A sheet (s3A and s5A), helix F (hF), and the connecting loop (Fig. 1, purple). Characterization of the mutant proteins and structural examination of the mutation sites revealed various unusual interactions as the structural basis of the native metastability. Functional analyses provide direct evidence that the energy loaded in the native inhibitory serpins is utilized for the biological activity.

EXPERIMENTAL PROCEDURES
Recombinant ␣ 1 AT Proteins-Plasmids for ␣ 1 AT expression in Escherichia coli, and the refolding and purification of recombinant ␣ 1 AT protein were described previously (26). Concentrations of ␣ 1 AT were determined in 6 M guanidine hydrochloride using a value of A 1 cm 1% ϭ 4.3 at 280 nm, calculated from the tyrosine and tryptophan content of the ␣ 1 AT protein (27) and based upon M r ϭ 44,250.
Chemicals-Ultrapure urea was purchased from ICN Biochemicals. Porcine pancreatic elastase and N-succinyl-(Ala) 3 -p-nitroanilide were purchased from Sigma. All other chemicals were reagent grade.
Mutagenesis and Screening of Thermostable Mutants-Random mutagenesis at the target region was performed as described previously using degenerative oligonucleotides (24), and thermostable mutations of ␣ 1 AT were screened after heat treatment of cell lysates at 60°C for 1 h as described previously (26). Substitutions at specific sites were generated by oligonucleotide-directed mutageneses.
Urea-induced Equilibrium Unfolding Transition-Equilibrium unfolding as a function of urea was monitored by fluorescence spectroscopy ( ex ϭ 280 nm and em ϭ 360 nm, excitation and emission slit widths ϭ 5 nm for both), details of which were described previously (24,26). The buffer was 10 mM phosphate, 50 mM NaCl, 1 mM EDTA, and 1 mM ␤-mercaptoethanol, pH 6.5, and the protein concentration was 10 g/ml. The native protein was incubated in the buffer containing various concentrations of urea at 25°C. Experimental data of the fluorescence measurement were fitted to a two-state unfolding model.
Determination of the Stoichiometry of Inhibition-The stoichiometry of inhibition was determined by titration reactions as described (28). The active concentration of porcine pancreatic elastase was determined by measuring the initial rates of hydrolysis of 1 mM N-succinyl-(Ala) 3p-nitroanilide. Various amounts of purified recombinant wild-type or mutant ␣ 1 AT proteins were incubated in 50 l of assay buffer (30 mM phosphate, 160 mM NaCl, 0.1% PEG 6000, and 0.1% Triton X-100, pH 7.4) with 100 nM porcine pancreatic elastase at designated molar ratios of ␣ 1 AT to protease. After incubation with protease for 10 min at 37°C, the reaction mixture was diluted 10-fold with the assay buffer and the residual enzyme activity was determined using 1 mM N-succinyl-(Ala) 3p-nitroanilide as a substrate (29).
Complex Formation with a Protease-Complex formation of mutant ␣ 1 AT with porcine pancreatic elastase was examined by monitoring the SDS-resistant ␣ 1 AT-proteinase complex (28). The active concentration of porcine pancreatic elastase was determined as described above. Purified ␣ 1 AT was incubated in assay buffer with the protease at designated molar ratios of ␣ 1 AT to protease. Samples were incubated at 37°C for 10 min and were analyzed by 10% SDS-polyacrylamide gel electrophoresis. The protein bands were visualized by Coomassie Brilliant Blue.

Molecular Properties of Mutant ␣ 1 AT
We found various stabilizing single amino acid substitutions at over 10 sites in the target region (Fig. 1). The conformational stability of representative mutant ␣ 1 AT was measured by equilibrium unfolding in the presence of urea in which changes in intrinsic tryptophan fluorescence intensity were monitored. The changes in free energy of stabilization (⌬⌬G) of the mutant proteins are summarized in Table I. For many of the sites, additional substitutions were made by introducing alanine, among which stabilizing mutations were further characterized (Table I). One remarkable result is that K335A and K335G mutations showed a profound effect on stability ( Fig. 2A), shifting the midpoint of the unfolding transition from 1.8 M to 3.9 M and 4.0 M urea, respectively, which resulted in the increase of stability over 6 kcal mol Ϫ1 . This is the biggest increase in stability of ␣ 1 AT by a single amino acid substitution studied so far. The effect of the mutations in ␣ 1 AT on the inhibitory activity was examined by determining the SI values against a target protease. The SI values of K335A and K335G ␣ 1 AT, which showed dramatic increase in stability ( Fig. 2A), were increased substantially (Fig. 2B). The SI value of K168A, which showed little increase in stability (Fig. 2B), was also increased (Fig. 2B). The changes in SI value were due to the alteration of the partitioning between the inhibitory and substrate path- The region subjected to the present studies is colored purple. The designation of secondary structures are described in the text. The identified mutation sites are marked by beads, and the amino acid residues are shown by one-letter codes. The colors of the beads are the same as in Fig. 4. The hydrophobic core region which was targeted in the previous study (24) is represented in blue. The figure was prepared using MOLSCRIPT program. ways. When the mutational effects on the formation of protease-inhibitor complex were analyzed on SDS-polycrylamide gel, the ratios of cleaved proteins over complexes were increased by K335A and K335G mutations compared with wildtype ␣ 1 AT (Fig. 3), indicating that there was a shift in partitioning toward substrate pathway in mutant proteins. The effects of other stabilizing mutations on inhibitory activity were determined and the results are summarized in Table I.

Correlation between Loaded Energy and Inhibitory Activity
If the loaded energy in the native serpins is utilized for inhibitory complex formation, there should be a correlation between stability increase and loss of inhibitory activity. The activity change was plotted as the function of stability increase for each mutant protein (Fig. 4), and the mutations were grouped into three classes according to their properties (Table  I). For group I mutations (top group in Table I), there seems to be a correlation between the increase in conformational stability and the loss in inhibitory activity (Fig. 4, circles; Fig. 1, yellow beads). Remarkably, the substitutions at Lys-335, K335A and K335G, increased the stability over 6 kcal mol Ϫ1 with concomitant losses of activity over 30% toward porcine pancreatic elastase. These results suggest that the loaded energy is related to the inhibitory function at Lys-335, and possibly at other sites. For the group II substitutions (middle group in Table I), K168A, K168I, or I169V barely increased the conformational stability but affected inhibitory activity sub-  4. The relation between stability increase and activity loss of the mutant ␣ 1 AT obtained in this study. Activity loss against porcine pancreatic elastase by each mutation was calculated as: 1 Ϫ relative activity of the mutant protein toward porcine pancreatic elastase. E, mutations with a tendency to correlate stability increase and activity loss (Table I, top group); ‚, mutations exhibiting a significant activity loss without much increase in stability (Table I, middle  group); Ⅺ, mutations showing less activity loss than expected from the stability increase (Table I, bottom group). Multi-7 mutation that increased the stability by 9.3 kcal mol Ϫ1 without affecting inhibitory activity significantly (28) is also shown for comparison. The correlation curve was drawn only with the first group mutations (dotted line; the correlation coefficient is 0.93), and the relationship between stability increase and activity change by Multi-7 is also drawn (dashed line). stantially (Fig. 4, triangles; Fig. 1, orange beads). It seems, however, that for these mutations the loss of activity is not due to an increase in conformational stability of the native form per se but to a decrease in the stability of the serpin-protease complex. Such mutations can be obtained in our initial screening for thermostable variants, because the screening is based on enhanced kinetic stability against aggregation (26). It has been suggested that the mechanism of heat-induced aggregation of ␣ 1 AT is loop-sheet polymerization in which the reactive center loop of one molecule is inserted into A sheet of another (30). For the group III mutations (bottom group in Table I), K163T, G164V, or A183V, the activity of ␣ 1 AT was less affected than expected from the increase in stability (Fig. 4, green squares; Fig. 1, green beads). Unlike group I mutations, which decreased the size of side chains, G164V and A183V increased the size of side chains. The hydrophobic core mutation reported previously, Multi-7, did not show any significant activity change even with a stability increase up to 9 kcal mol Ϫ1 (28). It is also included in the plot for comparison, and the relationship between stability increase and activity change is also indicated (Fig. 4, blue square; dashed line).

Structural Basis of Metastability
In the crystal structure of the native ␣ 1 AT (12), the stabilizing mutation sites are composed of many unusual interactions, which appear to be the basis of the metastability. These interactions are likely to be mobilized during the complex formation with a target protease. From the mutational analysis, the following modes were revealed as structural basis of the native metastability of ␣ 1 AT.
Overpacking of Side Chains-Lys-335 interacts with hydrophobic residues, Ile-169 and Leu-172 (Fig. 5A). The N atom of Lys-335 side chain may be paired with the backbone carbonyl oxygen of Ile-188 (2.87 Å) and O␥ atom of Thr-165 (3.72 Å), but the aliphatic part of Lys-335 is squeezed in hydrophobic interactions provided by Ile-169 (3.45 Å) and Leu-172 (3.24 Å). It appears that K335G or K335A mutation relieved such strain by decreasing the size of the side chain. Remarkably, many stabilizing substitutions such as T165S, I169V, L172V, and L172A are the ones occurring at the residues interacting with Lys-335 (Fig. 5A). Furthermore, the substitution pattern showed size reduction as a common theme. Leu-172 is unusually close to Lys-335 (3.24 Å). L172V as well as L172A, which increased the stability even more (Table I), might have eliminated such an overpacking of the side chains. Stabilizing mutations such as F189I and F189V also showed the size reduction pattern. Size reduction to alanine at 189, however, resulted in the loss of stability. We previously reported that decrease in the size of side chains at the hydrophobic core of ␣ 1 AT (Fig. 1, blue) confers increased stability (24). Structural determination of ␣ 1 AT variants carrying stabilizing substitutions at the hydrophobic core revealed better packing at the mutation sites (12,13,31). The substitution pattern observed in the present study (Table I) revealed that size reduction in the present target region (Fig. 1, purple) also increases the stability of the molecule. It appears that overpacking of the side chains is one mechanism of the native strain of ␣ 1 AT.
Polar-Nonpolar Interactions-The present studies revealed various polar-nonpolar interactions as an additional mode of destabilization in the target region. The side chain of Lys-335 is not only squeezed but also surrounded by hydrophobic residues. It is unusual to find charged residues in the hydrophobic environment. Studies on other mutant proteins revealed that the energetic cost of burying a charge in a hydrophobic environment is estimated be 3-9 kcal mol Ϫ1 (32). Mutations at Lys-335 of ␣ 1 AT increased the stabilizing energy by 3-7 kcal mol Ϫ1 (Table I). Phe-189 is on a surface pocket (Fig. 5A), interacting unfavorably with the backbone carbonyl oxygen of Gly-164 (2.89 Å). Thr-165 appears to stabilize the end of helix F and the following turn by providing hydrogen bonding to the backbone carbonyl oxygen of Val-161 and backbone nitrogen of Ile-169. The C␥2 atom of Thr-165, however, has close interactions with the polar atoms of the backbone, while the hydroxyl  (Table I) possibly by providing more favorable interactions with the surrounding residues. It is likely that these unfavorable polar-nonpolar interactions are another basis of the native metastability of ␣ 1 AT.
Cavities-Structural examination of the target region also revealed the existence of surface cavities. Two of the substitutions identified in the present study, A183V and G164V, do not fit the common theme of size reduction for stabilization of the native ␣ 1 AT. In both cases, size increase rather than size reduction caused the stability increase. In the native structure (12), the region near Gly-164 and Ala-183 is not well packed, leaving an empty patch on the surface (Fig. 6). The surface cavity near Gly-164 is surrounded by the side chain atoms of Phe-189 and Lys-335. Likewise, the side chain of Ala-183 does not show much interaction with other side chains. Surrounding hydrophobic residues such as Phe-147, Ile-157, Val-185, and Leu-172 are at least 5 Å apart. Cavities are very likely to be the source of energetic cost in conformational stability (33). The stabilizing substitutions, G164V and A183V, appear to provide better packing by filling the nearby cavities.

Favorable Interactions in the Complex Form
In the cleaved relaxed form, the identified mutation sites have favorable interactions. The crystal structure of the serpinprotease complex is not available yet. However, it has been proposed that accompanying the structural transition to the relaxed complex form, the reactive center loop is at least halfway (up to P9 position, the 9th residue amino-terminal to the scissile peptide bond) inserted into A sheet (34 -36), as in the relaxed cleaved structure. If this is the case, one can refer to the cleaved structure (14) to understand the complex. The side chain of Lys-335, which is buried in the native state (Fig. 5A), is exposed in the cleaved structure and makes a salt bridge with Asp-171 (Fig. 5B). Likewise, Lys-168 is paired with Glu-346 (P13 position) of the inserted reactive center loop (Fig. 5B). The side chain of Ile-169, which is exposed on the surface in the native form (Fig. 5A), interacts in the cleaved form with the side chain of Val-337 of s5A (Fig. 5B) and C␣ atom of Gly-349 (P10 position) of the inserted reactive center loop. Residues at other mutation sites also are engaged in favorable interactions in the cleaved structure. It is very likely that the conversion from the native unfavorable interactions to more favorable interactions in the complex is the driving force for the confor-mational switch needed for the complex formation.
Mutations at Lys-168 and Ile-169, which decreased the inhibitory activity without increasing the stability (Fig. 4, triangles) suggest that the salt bridge between Lys-168 and Glu-346 shown in the cleaved structure (Fig. 5B) plays an important role in the complex stability. Likewise, Ile-169 appears to play a role in stabilizing the loop-inserted structure. It is not clear, however, if the salt bridge between Lys-335 and Asp-171 also plays a significant role in the complex formation, because D171N mutation did not affect the inhibitory activity significantly. 2 It may be that Lys-335 mainly contributes to destabilization of the native form during the complex formation. It appears that each identified residue in the present target region contribute distinctively to the complex formation; some play a critical role in destabilizing the native form, some in the complex stability, and others in serving the dual role of destabilizing the native structure and stabilizing the complex. Many of these residues are conserved among inhibitory serpin sequences (2). The crystal structure of a serpin-protease complex will confirm the contribution of individual residues.

Importance of Local Strain in the Inhibitory Function
The mutational effect on the inhibitory activity observed in the present study is in contrast to that obtained previously with the stabilizing mutations at the hydrophobic core (Fig. 1,  blue). The inhibitory activity of ␣ 1 AT is very sensitive to the stabilization in the target region (Table I; Fig. 4). In contrast, the increase in the stabilization energy up to 9 kcal mol Ϫ1 in the hydrophobic core by combining seven stabilizing single amino acid substitutions (F51L, T59A, T68A, A70G, M374I, S381A, and K387R; the mutant was named Multi-7) did not affect the inhibitory activity of ␣ 1 AT toward target elastases (28). The results imply that local stability of the serpin is critical to inhibitory activity. It may be that some parts of the serpin molecule, especially the region where the reactive center loop is inserted, have to be loosened during the complex formation, whereas other parts like the hydrophobic core need not change as much. In this regard, it is interesting that while side chain locking among nonpolar side chains is the major cause of metastability in the hydrophobic core of ␣ 1 AT (24), at the loop insertion site additional schemes like unfavorable polar-nonpolar interactions are observed. The polar groups may govern structural specificity during the complex formation. Internal polar groups are often found to be destabilizing (37,38), but have evolved to impart structural uniqueness (37,39). It appears that the local strain of inhibitory serpins is critical for relating the loaded energy to functional regulation.  Fig. 1, and the right panel is the view from the right side. The backbone of Gly-164 is colored red, and the side chains of other representative residues are also colored (lysines are yellow and hydrophobic residues are cyan). This figure was prepared using INSIGHT II program (Molecular Simulation Inc.).

Implications for the Inhibitory Mechanism
Results in the present study strongly suggest that nature designed the native form of inhibitory serpins to be poorly folded with the purpose of carrying out a sophisticated regulation of protease inhibition. Our study also revealed that various folding defects such as side chain locking, buried polar groups in unfavorable hydrophobic environments, and cavities are the structural basis of the metastability of ␣ 1 AT. These folding defects appear to be designed to destabilize the interaction of helix F and the following loop with A sheet (Fig. 1). Especially, Lys-335 is in a strategic position for opening A sheet; the side chain of Lys-335 in the uncleaved native form is tightly squeezed by the surrounding hydrophobic residues from helix F and the following loop that covers A sheet (Fig. 5A). Considering the location (Fig. 1) and energetic cost ( Fig. 2; Table I) of Lys-335, we propose that this lysine plays a pivotal role during the complex formation, possibly by lifting helix F and the following loop from A sheet, which is likely to lead to the facile loop insertion during the complex formation. This lysine is conserved among inhibitory members of serpins such as antithrombin-III, ␣ 1 -antichymotrypsin, and plasminogen activator inhibitor-1, and hormone binding globulins of serpin family, but not in non-inhibitory members like ovalbumin and angiotensinogen (2). From our present data, however, it is not yet clear how much the cavities in this region contribute to the conformational transition during the complex formation, because the mutations supposedly filling the cavities of ␣ 1 AT, G164V and A183V did not affect the inhibitory activity as much as expected from the stability increase (Fig. 4, green squares).
It is worth noting that the recently determined crystal structure of gp120, the surface glycoprotein of HIV, complexed with soluble CD4 also revealed unusual interactions such as a large cavity and polar-aromatic interactions (40). Since the structure of free gp120 is not known, it is not yet clear how these unusual interactions contribute to conformational change of gp120 or gp41, the membrane fusion protein of HIV. Interestingly, the native form of hemagglutinin, the fusion protein of influenza virus, also has big cavities especially in the subunit contacts of the trimers (41). Although it was shown (42) that destabilization of the metastable native hemagglutinin yielded a fusion active state (43), structural basis underlying the conversion, including the role of the cavities, has not been understood.
Finally, a relationship between protein stability and function has not been established unequivocally. According to still controversial "stability-function hypothesis," residues contributed to function may not be optimal for stability. This was clearly shown for the active site residues of T4 lysozyme (44) and barnase (45). However, in case of E. coli major cold shock protein, CspA, residues constituting surface aromatic network has evolved for both function and stability (46). In the case of ␣ 1 AT, there is an inverse correlation between protein stability and function in the protease binding site. Still, suboptimal stability of ␣ 1 AT appears to be a prerequisite for functional execution rather than a consequence of functional reconciliation. Interestingly, most of the substituting residues in stabilizing mutations of ␣ 1 AT are the ones already existing in the sequence of ovalbumin (Table I), a non-inhibitory member of the serpin family. Ovalbumin and inhibitory serpins share a common ancestor (47). The molecule might have evolved for better folding and stability in the ovalbumin line, but for acquiring inhibitory function in the inhibitory serpin line.