Single Amino Acid Substitutions of a1-Antitrypsin That Confer Enhancement in Thermal Stability*

-- Arecombinant a,-antitrypsin variant which increased thermal stability was obtained from random mutagenesis followed by screening. The clone was identified as having a single mutation of Phe6' -+ Cys. Heat deactivation of purified recombinant a,-antitrypsin produced in Escherichia coli revealed that the mutation slowed down the deactivation rate 10-fold at 57 "C, increasing thermal stability of recombinant protein to almost that of natural glycosylated plasma form. The mutant pro- tein also exhibited increased stability against denatur-ant. The urea-induced unfolding monitored by the changes in fluorescence intensity at 360 nm showed that the mutation shifted midpoint of the transition from 1.9 M to 2.8 M. The mutation site is particularly interesting in that some genetic variants mapped at adjacent positions were shown previously to cause aggregation of the polypeptides, while the PheS1 + Cys mutation decreased aggregation rate significantly during heat deactivation. The association rate constant with porcine pancreatic elastase revealed that the mutation did not affect inhibi- tory activity significantly. The site identified may be critical for regulating stability of a,-antitrypsin. Char- acterization of various single amino acid substitutions at position 51 suggests that volume and flexibility of hydrophobic side chain at the site are critical factors for enhancing the stability of a,-antitrypsin.

a,-Antitrypsin (a,AT)' is a member of the serine protease inhibitor (serpin) family, which includes antithrombin, a,-antichymotrypsin, C1 inhibitor, ovalbumin, angiotensinogen, and hormone-binding globulins (1). Serpins share a common tertiary structure composed of three P-sheets and several a-helices that connect the strands in the sheets (1). Inhibitory serpins have a native strained ( S ) conformation in which the reactive center loop is open to proteolytic cleavage. The cleavage accompanies an irreversible transition to a very stable relaxed ( R ) form where the newly created N-terminal portion of the cleaved loop is completely inserted as a strand of sheet A ( 2 4 ) . In ovalbumin, a noninhibitory but more stable serpin, the reactive center loop arches over the molecule and forms an a-helix (5), and cleavage of the reactive center loop does not induce a conformational switch as in &,AT (2). A few other conformations of serpins have been identified. The active form of plasminogen activator inhibitor-1 (PAI-1) can be spontaneously converted * This work was supported by Grant G70550 from the Korean Ministry of Science and Technology. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
into a more stable latent form (6). Recent structural determination (7) showed that intact reactive center loop in latent PAI-1 is inserted into the major P-sheet (sheet A) forming a strand as in cleaved a,AT. The rest of the loop stretches over the edge of the protein and joins a strand within the top P-sheet (sheet C). A stable locked conformation of antithrombin and other serpins could be induced from the native form under mild denaturing conditions, in which intact reactive center loop is presumably inserted into the sheet A as in latent form of PAI-1 (8). The existence of a more stable latent structure suggests that the native structure of serpins is not the thermodynamically most stable form. In case of a,AT it was also shown that oligomers are formed on prolonged incubation under mild denaturing conditions or during heat denaturation via a process called loop-sheet polymerization, in which reactive center loop of one molecule is inserted into sheet A of another (9, 10).
Structural comparisons of various forms of serpins and thermostability studies of serpins complexed with synthetic peptides carrying sequences of the reactive center loop (8, 11) support a notion that the enhancement in stability appears to be mainly due to insertion of the cleaved loop into the sheet A with concomitant increase in the number of strands in the sheet and buried surface area. However, latent PAI-1, which is more stable than the active form, is less so than the cleaved form (61, showing that insertion of the loop was not sufficient to confer maximum stability exhibited by cleaved form. In the case of peptide-annealed antithrombin, it was shown that the complex was as stable toward heat deactivation as cleaved antithrombin (8). All the stable forms of serpins so far identified, including ovalbumin, are not active as an inhibitor. The reactive center loop structure of active serpins is considered to be mobile and partially inserted into sheet A of the molecule (8, 12,13). The presence of such a mobile loop is presumably critical for the inhibitory function of serpins. There has been an indication that the stress of native serpin is not limited to the reactive center loop but may be distributed throughout the molecule.

Studies of 'H-'H exchange and Fourier transform infrared
spectroscopy with the intact and cleaved form of a,AT showed that stability enhancement in cleaved form is related to increase in contents of stable secondary structure in many parts of the molecule without much variation in tertiary interaction (14). In the present study we identified single amino acid substitutions of native human a,AT that confer increased thermal stability but maintain inhibitory activity. The mutation also increased stability toward urea denaturation. Characterization of stable variants will help define the relationship between strain and the tertiary fold of the native a,AT structure. Such mutants might be also of practical use because the thermal stability of recombinant a,AT was shown to relate to the biological turnover rate of the protein (15).

EXPERIMENTAL PROCEDURES Plasmid Construction and Protein
Expression-Construction of an expression plasmid, pEAT8, encoding the a,AT cDNA under control of the phage T7 promoter (16) has been described (17). The plasmid encodes an a,AT polypeptide in which the first residue of the authentic a,AT, glutamate, is substituted by methionine. A transformant of Escherichia coli BL21(DE3) was grown on MSZB medium (per liter: 1 g of NH,CI, 3 g of KH,PO,, 6 g of Na,HPO,, 4 g of glucose, 0.48 g of MgSO,, 10 g of tryptone, 5 g of NaC1) containing 50 pg/ml ampicillin at 37 "C.
When the A, , reached 0.8, isopropyl-P-D-thiogalactopyranoside was added to a final concentration of 0.4 mM and growth was prolonged for 3 h at a designated temperature. The growth temperature affects the solubility of expressed a,AT encoded by the pEAT8 plasmid (17). For the purpose of screening mutants, a,AT was expressed a t 37 "C a s a soluble form. However, for the purpose of purification, the protein was expressed at 40 "C as inclusion body.
Mutagenesis--Random mutagenesis was performed by a modified polymerase chain reaction (PCR) a s described by Eckert and Kunkel (18). The 1260-base pair restriction fragment of a,AT cDNA was cloned into phage M13mp18. Single-stranded template DNA was isolated, and two M13 primers (Sigma) of reciprocal orientations were used for PCR. The reaction mixture was dATP-limiting condition (0.1 mM dATP, 1.0 mM each of three other dNTPs) in 10 m~ MgCI,. After 25 cycles, BcZI-BstXI fragment (770-base pair encoding amino acid positions 17-273) was isolated and exchanged with an equivalent fragment of pEAT8. Site-specific mutagenesis was performed by the method of Kunkel (19). Multiple amino acid substitutions were carried out by a mutagenic oligonucleotides (30-mer) containing NN(G/C) (N is equimolar mixture of four dNTPs) at the target position. Nucleotide sequencing was performed with Sequenase 2.0 (U. S. Biochemical Corp.).
Screening of Thermostable afiT-The colonies resulting from the mutagenesis were inoculated into 0.1 ml of supplemented minimal medium (per liter: 1 g of NH,Cl, 3 g of KH,PO,, 6 g of Na,HPO,, 2 g of glucose, 0.2 g of yeast extract, 3 g of casamino acids) containing 1 mM isopropyl-P-D-thiogalactopyranoside and 50 pg/ml ampicillin on 96-well microtiter plates. The plates were incubated at 37 "C overnight on a shaker. The cultures were lysed by adding 25 pl of lysis buffer containing 250 mM Tris-CI, pH 8.0,25 m~ EDTA, 0.25% Triton X-100,0.5 mg/ml lysozyme. The plates were incubated a t room temperature for 1 h with continuous shaking. Following lysis the plates were incubated at 60 "C for 1 h to inactivate the wild type a,AT activity. Mutant a,AT that survived the above heat deactivation was identified by a chromogenic assay developed for dithiothreitol-sensitive mutants of bovine pancreatic trypsin inhibitor (20). Porcine pancreatic elastase (Sigma) was used as a protease a t a 1.5 nM final concentration and N-succinyl-Ala-Ala-Ala-p-nitroanilide (Sigma) was used a s a substrate a t a 0.3 mM final concentration.
Refolding a n d Purification of Recombinant afiT-Recombinant a,AT was purified from inclusion bodies after refolding and ion-exchange chromatography. Details of the purification method will be published elsewhere, but a brief description is as follows. The pellet of inclusion body was washed twice with bufferA(50 m~ Tris-HC1, pH 8.0, 50 mM NaCI, 1 m~ EDTA, 1 mM P-mercaptoethanol) containing 0.5% Triton X-100 and was solubilized in 8 M urea in buffer A at a protein concentration of 2 mg/ml. Refolding was carried out by direct 10-fold dilution with buffer B (10 mM phosphate buffer, pH 6.5, 1 m~ EDTA, 1 mM a-mercaptoethanol), which was followed by dialysis against buffer B. a,AT was further purified by chromatography on a DEAE-Sephacel column and Mono Q HR5/5 column of fast protein liquid chromatography (FPLC) system (Pharmacia LKB Biotechnology Inc.). Fractions containing a,AT activity were pooled and stored at 4 "C for the subsequent studies. Purity of the preparation was confirmed by SDS-polyacrylamide gel electrophoresis and capillary zonal electrophoresis (Beckman P/ACE System 2000). The integrity of the N terminus including the first methionine was confirmed by protein sequencing (Applied Biosystem model 477A Sequencer). Human plasma a,AT (Sigma) was further purified by chromatography on an Aff%Gel Blue column (Bio-Rad) and Mono Q column as described (21).
Quantitation of afiTActiuity and Determination of Association Rate Constants-a,AT activity was measured as residual porcine pancreatic elastase activity employing 1 m~ N-succinyl-Ala-Ala-Ala-p-nitroanilide a s a chromogenic substrate (21). The association rate constant for the interaction of a,AT with porcine pancreatic elastase was measured under second order conditions in a reaction mixture containing equimolar concentration (8 nM) of enzyme and inhibitor (22). Concentrations of a,AT were determined in 6 M guanidine hydrochloride using a value of Ai:,,, = 4.3 at 280 nm, calculated from the tyrosine and tryptophan content of the protein (23) and based upon M , = 44,250.
Thermal Deactiuation-Thermal stability of a,AT variants was measured by following the kinetics of inactivation a t a designated temperature. Aliquots of samples were taken along the time course of in- cubation, and the amount of remaining a,AT activity was determined as described above. Determination of Aggregation State-Superose 12 analytical FPLC column was equilibrated with 10 m~ phosphate buffer, pH 6.5, containing 50 m~ NaCI, 1 m~ EDTA, 1 mM 6-mercaptoethanol, and 0.02% NaN,. The flow rate was 0.5 mYmin.
Urea-induced Unfolding Dansition-Equilibrium unfolding a s a function of urea was monitored by fluorescence spectroscopy. Native protein was incubated in 10 m~ potassium phosphate, 50 m~ NaCl, 1 m~ EDTA, 1 mM P-mercaptoethanol, and various concentrations of urea (final pH, 6.5) at 23 "C. Samples were allowed to equilibrate for 8 h. The tryptophan fluorescence was measured for each sample (Shimadzu RF-5000 fluorescence spectrophotometer) with a n excitation at 280 nm and a n emission at 360 nm. The protein concentration for unfolding transition was 50, 30, or 6 pg/ml. Samples for a reversibility test were prepared as follows. The protein, which was unfolded in 7 M urea (150 pg/ml) for 7 min at room temperature, was refolded a t various urea concentrations with a final protein concentration of 10 pg/ml and equilibrated for 8 h at 23 "C. Experimental data was fitted to a two-state unfolding model (Equation 11, assuming a linear relationship between free energy of unfolding and the concentration of urea (24)

Identification of a Thermostable Mutation,
Phe5' + Cys (F5lC)"Initial screening and heat deactivation assays with lysates acquired one clone that expressed substantially more stable a,AT activity than the wild type. Nucleotide sequencing identified that the 51th codon, TTC (Phe), was changed into TGC (Cys). Heat deactivation was performed at 57 "C with purified a,AT proteins. As shown in Fig. 1, F51C mutant recombinant protein was indeed more thermostable than the wild type recombinant protein. The stability of mutant a,AT was comparable with that of the glycosylated plasma a,AT. In order to determine the aggregation rate during heat deactivation, gel exclusion chromatography was performed with the proteins  ( a and c) and 3 h ( b and d ) .
incubated at 55 "C. As shown in Fig. 2, aggregates were formed for both wild type and F51C mutant proteins during heat deactivation, but the mutation decreased the aggregation rate significantly. Association rate constants of F51C and the wild type recombinant a,AT with porcine pancreatic elastase were 1.4 x lo6 M-, s-' and 1.6 x lo6 M" s-', respectively, which was comparable to the value (1.7 x lo6 M-' s-') of human plasma a,AT. The results confirmed that F51C mutation caused enhancement in thermal stability without altering inhibitory activity significantly.

Effect of the Phe5' + Cys Mutation on the Urea-induced
Unfolding Pansition of aflT-In order to examine the effect of F51C mutation on the conformational stability of a,AT, equilibrium unfolding in urea was carried out. Unfolding was monitored by the changes in intrinsic tryptophan fluorescence intensity at 360 nm. The fractions of unfolded molecules a t various urea concentration were determined by fitting the experimental data to a two-state unfolding model (Fig. 3a). The midpoints of transition (C,) of the wild type and the mutant protein were 1.9 (-. 0.1) and 2.8 (e 0.1) M, respectively. Both transitions of the wild type and F51C appeared to be concentration-independent between 6 and 50 pg/ml (Fig. 3a). The transitions were fully reversible, and almost all the signals were regained in the refolding from 8 M urea, as shown in Fig.  3b. The C , values of the wild type protein were 1.9 and 2.0 M for unfolding and refolding, respectively, while the identical value of 2.7 M was obtained with the mutant protein for both unfolding and refolding. The results in Fig. 3b also showed that the native fluorescence intensity of the mutant protein was lower than that of the wild type for the same concentration (10 pg/ml as measured from the absorbance at 280 nm in 6 M guanidine hydrochloride), while the unfolded values were similar.
Various Single Amino Acid Substitutions at Position 51-In order to identify factors at the position 51 influencing the thermal stability of a,AT, various single amino acid substitutions were introduced. All 18 substitutions except lysine were obtained. Initial screening with lysates from cells grown at 37 "C revealed that substitutions with Pro, Trp, and charged residues (including His) did not yield any soluble activity of a,AT although polypeptides are synthesized normally. For the other substitutions soluble activity of each lysate divided them into two groups: (i) substitutions with Leu, Val, Ile, Cys, Ala, and Gly belonged to one group with substantial inhibitory activity and (ii) substitutions with Asn, Ser, Gln, Met, Thr, and Tyr belonged to the other group with decreased activity when compared to the wild type lysate. Some of a,AT protein carrying mutations belonging to the first group were purified, and heat stability and inhibitory activity were determined. Results are summarized in Table I. None of the mutations shown in Table  I affected the function of a,AT significantly. The heat deactivation rate indicates that amino acid residues carrying hydrophobic aliphatic side chains at position 51 increased the stability, and among them side chains with branched C, (Val, Ile) were less effective than those with linear ones (Cys, Leu).

DISCUSSION
Stability of Native Serpins-One of the intriguing aspects of serpins is that the native conformation is not the thermody-  namically most stable conformation. Strain of the active form of serpins appears to be very critical to inhibitory function and conformational switch that occurs after cleavage. The question is then how much of the molecule is strained and the nature of the structural mechanism of maintaining the stress. A genetic approach was adopted to examine these problems. In order to obtain a stable variant of alAT, a method of random mutagenesis and screening was adopted because (i) the X-ray crystal structure of intact native form of &,AT is not yet available and (ii) the effect of mutation on the stability of a protein is hard to predict because the stability is a collective property contributed by the whole molecule. The F51C mutation of a,AT increased thermal stability without loss of inhibitory activity. It also increased conformational stability toward urea denaturation. The existence of the F51C mutation, which maps far apart from the reactive center loop, supports the notion that the stress of native form of a,AT may be distributed throughout the molecule. This is consistent with a recent finding that the intact native a,AT in comparison with the cleaved form has weakened hydrogen bonds in many parts of the molecule (14). Identification of more of stable mutants will help understanding factors regulating the stability of the intact native structure of serpins.
Mechanism of F51C Mutation of afiT-The position of F51C is particularly interesting in that some genetic variants, Mmalton (Phe5' or Phe5' deleted) (25) and Siiyama ( S e P + Phe) (26), that cause aggregation of the a,AT polypeptides are located at adjacent positions. In case of the best characterized genetic defect, 2 mutation ( G~u~~' + Lys), aggregation is formed by loop-sheet polymerization (9, 10). It was shown that Siiysma mutation also enhanced loop-sheet polymerization (27). The F51C mutation decreased heat-induced aggregation (Fig.  2), and it may well be that the mutation diminished loop-sheet polymerization. Thermolability of the Z mutation was solely attributed to loop-sheet polymerization (10). Since the Z mutation is located at the tip of strand 5 of sheet A (s5A), which is likely to be a critical site for partial insertion of loop into the sheet A, a local perturbation by the mutation may directly lead to loop-sheet polymerization without significant change in the rest of the molecule. The mutations at 51 or 53, however, may have a distinct mechanism to influence loop-sheet polymerization. Residues 51 and 53 are located in the strand 6 of the sheet B (s6B) and at the beginning of the helix B, respectively. The site is part of the hydrophobic core of the molecule which is composed of the helix B and the strands s4B, s5B, and s6B (2). Analysis of crystallographic structure of the cleaved form of a,AT and ovalbumin suggested that sheet A slides upon helix B when a strand insertion occurs (28). It was predicted that the Siiyama mutation of Ser53 to Phe would perturb the sliding movement and lock the conformation of the sheet A in an open form, leading to loop-sheet polymerization (27). The simplest explanation for the effect of the F51C mutation is then just the

9630
Thermostable Mutations of a,-Antitrypsin reverse of the effect of Siiyama mutation; mutation of Phesl to Cys somehow diminished the tendency of sheet A to convert into an open conformation. Indeed, in the crystal structure of cleaved @,AT (2) Phesl interacts with aromatic residues of s5B and s4B in the hydrophobic core including Phe384 (located in s5B, 3.4 A apart) upon which sheet A packs (28). It also interacts with somg residues in sheet A sue? as Ala347 (s4A, 4.4 A), Ala336 (s5A,

A)
, and Leu33s M A , 3.8 A), which are located at or interact with the newly inserted strand of sheet A (s4A) in cleaved form (28). The F51C mutation may interfere with opening of sheet A through interactions with some of these residues. The effect of F51C mutation on the conformational stability was investigated by equilibrium unfolding in urea. The guanidine-induced equilibrium unfolding of a,AT has been studied previously by various groups (4, 11, 29, 301, and a thermodynamic reversibility of unfolding has been demonstrated in the transition measured by fluorescence emission intensity (29). Our initial studies of the guanidine-induced unfolding monitored by the changes in intrinsic tryptophan fluorescence did not show an obvious effect of the mutation, which was probably due to the fact that both transitions occurred at very low guanidine concentration (data not shown). In order to observe the effect of the mutation on unfolding more clearly, urea-induced unfolding was performed. The results in Fig. 3 showed that the mutation shifted the transition midpoint from 1.9 to 2.8 M urea, increasing the free energy of stabilization (AAG) by 3 kcal/mol. Concentration independence and reversibility, shown in Fig. 3, indicated that the aggregation under the experimental conditions is negligible. It was reported that the recombinant a,AT tends to aggregate during unfolding (29), but a much higher concentration (0.1-0.5 mg/ml) of alAT was used in that experiment. In our refolding experiment, recombinant a,AT was found to recover 50% activity at 0.2 mg/ml but over 90% activity below 30 pg/ml for both wild type and the mutant proteins. The unfolding transitions of a,AT monitored by fluorescence intensity could be fitted to a two-state model very well. However, the transitions monitored by circular dichroism spectroscopy (3,29) or transverse urea gradient gel elctrophoresis (13,29) showed that the unfolding of a,AT is multi-phasic and continues at higher concentrations of denaturant. Therefore, the transitions shown in Fig. 3 reflect only the step of unfolding to which the fluorescence signal is sensitive, and the calculated energy difference is not that of the native and the unfolded states. There are two different ways of increasing the free energy of stabilization: first, lowering the energy level of the native folded state, and second, increasing the energy level of unfolded (partially unfolded in our case) state. These possibilities are to be distinguished to understand the mechanism of F51C mutation in detail.
The transition shown and changed by the mutation in urea denaturation may be the same as the transition from the native to the open conformation described above. Trplg4, the major contributor of the signal among two tryptophans in a,AT (29, 30), is likely to be a sensitive probe for sheet opening because it is located at the top of s3A and is hydrogen-bonded to the side chain of Asp341 (top of s5A) in the cleaved form (2) and also to the backbone of Lys343 in model native a,AT (31). The mutational effect of F51C on aggregation during heat deactivation is likely to be a secondary effect resulted from the change in conformational stability shown in urea denaturation. The site identified in the present study may be critical for regulating stability of a,AT.
Side-chain Contribution at Position 51-In the crystal structure of the cleaved form of alAT, the side chain of Phesl is involved in hydrophobic interaction with several residues including Phe372 and Phe384 (2). This part of the structure is quite similar in the cleaved form of a,AT and the crystal structure of ovalbumin (5), i.e. an intact serpin, and may not be different in the intact form of a,AT. Results of various amino acid substitutions showed that charged and polar residues are not tolerable at the site, consistent with the hydrophobic environment of position 51. The fact that smaller aliphatic hydrophobic side chains at 51 are more effective than the wild type residue, Phe, in enhancement of thermal stability of a,AT and among them linear chains at the P carbon are more efficient than the branched ones (Table I) suggests that volume and flexibility of hydrophobic side chains at 51 are critical factors for the stability of a,AT. Phesl is moderately conserved among various serpins (1). Most of the amino acid substitutions that are shown to increase stability of a,AT against heat deactivation (Ala, Val, Ile, and Leu) are the ones found in sequences of other serpins. All of these residues have high propensity to form P-sheet (32). Mutational effect may be at the level of secondary structure (e.g. increasing the stability of the sheet). Alternatively, the substitutions with smaller and more flexible side chains at position 51 might have improved an overall tertiary packing. Structural studies on these mutant proteins will shed light on understanding the mutational effects and eventually the stability of serpins in general.