Isolation and properties of recombinant DNA produced variants of human alpha 1-proteinase inhibitor.

Using the glyceraldehyde-3-phosphate dehydrogenase promoter, nonglycosylated human alpha 1-proteinase inhibitor, representing 10% of the soluble cell protein, has been synthesized in yeast. Two forms of this protein were isolated with one being analogous to the human plasma protein and the other having the amino acid valine replacing methionine at position 358 (the P1 position). Both proteins were more sensitive to heat inactivation than the plasma form, and both had shorter half-lives in rabbits. These differences were presumably due to the absence of carbohydrate. Each protein could bind neutrophil elastase at a rate only slightly slower than that of human plasma alpha 1-proteinase inhibitor. However, the valine variant was stable to oxidation, while the P1 methionine-containing protein was readily inactivated. The specificity of alpha 1-proteinase inhibitor (methionine) was identical to that of the plasma form; however, the valine form could only effectively bind to neutrophil or pancreatic elastase, "trypsin-like" serine proteinases not being inactivated at all. These data indicate the potential importance of mutant forms of proteinase inhibitors, produced by recombinant DNA technology, as therapeutic agents for the inactivation of excess proteinases of a specific type in tissues.

induced deficiency (4) or by oxidative inactivation due to conversion of the reactive site methionine to its sulfoxide derivative ( 5 ) , enzymatic degradation of lung connective tissue may occur and, ultimately, pulmonary emphysema (6).
In order to decrease the proteolytic burden on the lung, attempts have been made to either synthesize effective low molecular weight inhibitors for therapeutic use (7, 8) or to develop techniques for the isolation of large amounts of the natural inhibitor for supplementation (9). Recently (IO), cDNA clones of normal al-PI have been constructed in plasmids and the protein expressed, thereby providing a potential source for the isolation of large quantities of inhibitor. However, such a protein would still be susceptible to oxidative inactivation, thereby reducing its potential effectiveness. Because neutrophil elastase cleaves peptide bonds preferentially after valyl residues (2), a mutated form of al-PI with valine in the Pl reactive site position (11) would seem a good potential choice as an inhibitor with therapeutic value. Previously (lo), it had been shown that such a protein could be produced in yeast by mutagenesis of the cDNA sequence coding for the reactive site sequence. Furthermore, extracts of yeast expressing this protein inactivated neutrophil elastase both before and after oxidation by N-chlorosuccinimide, while those containing the normal al-PI sequence lost all activity after this treatment. In this report we describe the isolation and properties of both the normal (al-PI yeast methionine) and the mutated form (al-PI yeast valine) of this inhibitor.

Materials
Human neutrophil elastase (12) and human neutrophil cathepsin G (13) were obtained as previously shown. Porcine pancreatic trypsin and porcine pancreatic elastase were from Sigma. Human plasmin and human Factor Xa were gifts of Dr. John Fenton and Dr. Robert Jordan, respectively. Normal human al-PI and the Pittsburgh mutant form of al-PI were isolated as described elsewhere (14, 15). The S and Z variants of a,-PI were prepared by the method of Jeppsson et al. (16). Rabbit anti-human al-PI was obtained from Behringwerke. Restriction enzymes and the Klenow fragment of DNA polymerase I were obtained from New England Biolabs.

Construction of Yeast Expression
Plasmids-DNA manipulations were carried out as previously described (10,17). Plasmid PAT (Met), containing an al-PI cDNA extending from Asp-2 to the C terminus (IO), was digested with BamHI, made flush-ended using the Klenow fragment of DNA polymerase I , digested with Sal1 and the 1200-bp al-PI fragment isolated from a 1% agarose gel.
Plasmid pPGAP, whose properties will be described in detail elsewhere: is a derivative of pBR322 (18) and was prepared as follows. pBR322 was digested with EcoRI and SalI, these sites converted to BamHI sites by the use of a synthetic DNA linker (5'GGATCCGGATCC), and the molecule then ligated to produce a deleted derivative (about 3.8 kb) with a single BamHI site. A 1300bp BarnHI fragment containing the yeast glyceraldehyde 3-phosphate 49 gene (19) promoter and transcription terminator was inserted at this BamHI site. This fragment consists of about 400 bp corresponding to the 5' untranslated region of the glyceraldehyde-3-phosphate dehydrogenase and contains the promoter as well as 900 bp corresponding to the entire 3' untranslated region extending from the natural SalI site which is about 60 bp from the glyceraldehyde-3phosphate dehydrogenase translation termination codon to the BamHl site. These two fragments were linked using a 35-bp synthetic DNA fragment c o n~i n i n g an Ne01 site (B'CCATGG) at the gtyceraldehyde-3-phosphate dehydrogenase initiator methionine at one end and a Sat1 site at the other for joining to the natural Sal1 site of the glyceraldehyde-3-phosphate dehydrogenase gene as described above.
Plasmid pPGAP was prepared as shown in Fig. 1 and used to clone the 1200-bp al-PI coding sequence described above. The resulting plasmid was partially digested with RanHI, and a 2.5-kb fragment containing the (ul-PI cDNA flanked by the yeast glyceraldehyde-3phosphate dehydrogenase promoter and transcription terminator was isolated. This fragment was inserted into the yeast shuttle vector pC1/1 (10,20) at its BarnHI site and a recombinant clone containing the 2.5-kb fragment in the orientation shown was obtained.
An analogous plasmid to pCl/lGAPal-PI with Met-358 changed to Val-358 was prepared as follows. A 1030-bp BamHI-BstE2 fragment containing the glyceraldehyde-3-phosphate dehydrogenase promoter and N-terminal half of the nl-PI cDNA was isolated from plasmid p C l / l G A P~~-P~. A BstE2-BamHI fragment of 1440 bp, containing the C-terminal half of the nl-PI cDNA and the gfyceraldehyde-3-phosphate dehydrogenase terminator where the Met-358 to Val-358 mutation had been engineered by site-specific mutagenesis, was isolated from pCl/lPH05AT (Val) (10). These fragments were ligated with the pC1/1 vector which had been treated with BamHI and alkaline phosphatase, and a plasmid with the same structure as pCl/lGAP al-PI with the Met-358 to Val-358 substitution was obtained.
Isolation of Yeast-derived al-Proteinase Inhibitor Variants-Preliminary experiments had indicated that extracts containing q -P I rapidly lost inhibitory activity, presumably due to the presence of yeast proteinases. Therefore, extraction buffers utilized in the purification of al-PI routinely contained 1 mM EDTA and 1 mM PMSF. Both inhibitors were purified by the procedures described below and behaved identically during the isolation.
Extraction-Typically, 14 g (wet weight) of yeast containing the <?,-PI mutant were extracted with 40 ml of 0.05 M Tris-HCI, 0.05 M NaC1, pH 8.0. The cells were broken by addition of 10 g of washed glass beads (Corning), followed by vortex swirling for 10 min at room temperature. Further extractions yielded only trace amounts of inhibitor. The yellow extract was centrifuged at 24,000 X g for 10 min at 4 "C, and the s u~r n a t a n t was retained (35 ml). C h r o m Q~o g r Q~h~ on C~~a c~~ Blue Sepharose-The extract was directly applied to a column of Cibacron Blue Sepharose (3.0 X 40 cm), equilibrated with the extraction buffer. Proteins were then eluted in two major peaks by further washing of the column with buffer. The inhibitory activity towards neutrophil elastase was associated with the leading fractions of the second peak.
Chromatography on DEAE-cellulose-Fractions containing inhibitory activity (50 ml) were applied to a column of DEAE-cellulose (3.0 X 50 cm), equilibrated with 0.05 M Tris-HC1, 0.05 M NaC1, pH 8.0. The column was washed until the A2mnm was less than 0.010, and a linear gradient (500 ml total) to 0.2 M NaCl was then applied. Gel Electrophoresis-The purification and characterization of the CYI-PI variants was followed by two different gel electrophoretic tech-P. Tekamp-Olson, S. Rosenberg, Q. L. Choo, and D. Coit, manuscript in preparation.
Enzyme Assays-The inhibitory activities of the two n,-PI variants were followed by measuring the loss of enzyme activity upon incubation of an excess of proteinase with inhibitor. After 10-min incubation in 0.03 M sodium phosphate buffer, 0.16 M NaCI, pH 7.4, residual enzyme activity was determined by the addition of substrates specific for that given enzyme and the difference was measured in comparison to the activity of control enzyme solutions containing no inhibitor. Determination of Second Order Association Rate Constants-The rates of association of -,-PI variants with a number of serine proteinases were determined as previously described (3), except that peptide thiobenzyl ester substrates were utilized in order to determine association rates in very dilute solution for rapidly reacting enzymeinhibitor systems (e.g. @,-PI variants and neutrophil elastase). Briefly, n1-H solutions were standardized against active site-titrated porcine pancreatic trypsin, and these solutions were used to titrate the activity of the enzymes being tested. Because the q -P I valine mutant did not inhibit pancreatic trypsin (Table I), the activity of this inhibitor was measured by titration against a secondary standard, neutrophil elastase. For the actual determination of association rates, equal molar activities of inhibitor and enzyme were mixed and incubated in 0.03 M sodium phosphate buffer, 0.16 M NaCI, pH 7.4, at room temperature for specific time periods, substrate was added, and the residual enzyme activity was measured. The final molarity of substrates utilized in these measurements was 2 mM for those used in conjunction with neutrophil elastase, cathepsin G , and human plasmin assays, 0.5 mM for porcine pancreatic trypsin, human factor Xa, and human thrombin assays, and 1.2 mM for porcine pancreatic elastase. In order to determine the effect of oxidation of the two variants on their individual rates of complex formation with various proteinases, each was also subjected to treatment with N-chlorosuccinimide (1 mg of inhibitor to 1 pg of oxidant), as previously described (5). Ail of the association rates given are average values with deviations of up to 20%.
Heat Stability Measurements-The heat stability of both the methionine and valine mutants was tested using a method based on that of Lieberman (24). The purified mutant was diluted to a concentration of 0.2 mg/ml with 0.075 M Tris, 0.075 M glycine, 0.075 M sodium phosphate buffer, pH 7.48. Samples were incubated at 56 "C and aliquots were removed at 20-min intervals over a 4-h period, Protein that remained soluble, i.e. heat-stable protein, was determined by electroimmunoassay uersus rabbit anti-human (uI-PI (25).
Measurement of Bio~gical Turnover of cut-Proteinase Inhibitor Mutants-The turnover rates of human plasma n,-PI and yeast al-P1 variants were compared in the rabbit. Approximately 0.5 mg of each variant was labeled with *' ' I by the chloramine-T method of Hunter injected into a New Zealand White rabbit (weight, 1 kg) which was (26). In each case, a volume containing 3.0 X 10" counts/ml/min was on an iodine supplemented diet. Sequential blood samples were taken and counted to give the disappearance curves shown in Fig. 4.
Sequence Analysis of al-PI Variants-Amino-terminal sequence analysis of the individual variants of cul-PI was performed with a Beckman 890C Sequencer, as described previously (5).

Characterization of a,-Proteinase ~n h i~~t o r
Variants from Yeast-The c o n s t~c t i o n of high efficiency expression plasmids for the production of the (uI-PI variants in yeast using the glyceraldehyde 3-phosphate promoter is shown in Fig. 1 and also detailed under "Experimental Procedures." After growth in either selective or nonselective media the al-PI variants represented an average of 10% of the soluble cell protein (from assays of 6 independent cultures) as judged by elastase inhibition assays (10). Both the methionine and valine al-PI variants obtained from yeast were prominent bands in the yeast crude extract and were purified to homogeneity in two steps after cell extraction (Fig. 2), with mobilities on agarose gel electrophoresis slightly slower than that of normal plasma aI-PI. Each yielded approximately 28 mg of protein from 14 g (wet weight) of cells, assuming an E (1%, 280 nm) of 5.1 (111, which is that reported for the plasma cY,-Proteinase Inhibitor Variants h m n 1 EcoRl form of this inhibitor. However, this will probably differ in individual yeast preparations since expression of the protein may be variable. Both variants migrated as single bands after NaDodS0,-gel electrophoresis, with molecular weights near 46,000, relative to 52,000, for that of the native plasma protein (Fig. 3). Since plasma nl-PI contains approximately 13% carbohydrate, while the yeast-produced proteins have none, the lower molecular weight is in agreement with that for a nonglycosylated form of either inhibitor.
Both proteins had the amino-terminal sequence beginning NH2-Met-Asp-Pro-Gln-Gly-, after five Fxlman cycles. This agrees with the DNA sequence of the expression vector (Fig.   1).
Stability-The methionine yeast mutant was found to be quite labile, rapidly losing inhibitory activity on standing a t 4 "C. The protein still migrated as a single band; however, it could no longer inhibit neutrophil elastase. Activity could be regained (up to 80%) upon dialysis uersu. 9   0.001 M 2-mercaptoethanol to maintain stability. Presumably, the inactivation in the absence of reducing agents and the protection afforded in their presence was due to oxidation, especially since the a,-PI valine yeast mutant was quite stable, only very slowly losing activity on standing a t 4 "C. In terms of heat stability, both the methionine and valine yeast mutants had identical heat stability patterns, being significantly less stable than the M, S, or Z plasma forms of aI-PI (Fig. 4).
Biological Turnover-The disappearance rates of the al-PI variants injected into rabbits are shown in Fig. 5. The methionine mutant was tested in two different animals in order to confirm the short half-lives obtained. The validity of the rabbit as a model for plasma protein turnover in the human is supported by previous studies with other proteins and the observation that the half-life of human al-PI in the rabbit (2.2 days) is the expected value in terms of the size of the animal and the known half-lives of other proteins in the rabbit (27). We have also found3 that the turnover of the S variant of human a,-PI in the rabbit is increased with the '' R. Carrell, unpublished observations. same proportionality as in the human (28). Fig. 5 also shows the turnover of fresh, normal human al-PI and a stored sample ofthe Pittsburgh al-PI variant in which the PI residue (number 358) had mutated from methionine to arginine (15). Taking into account the age of the samples, these have, as expected, identical half-lives of just over two days, i.e. the mutation at the reactive site has not significantly altered the plasma turnover rate. However, there is a considerable increase in the rate of disappearance of the yeast a,-PI, both the methionine and the valine forms having half-lives of 8.5 h.
Effect of Oxidation of a l -P I Yeast Variants on Complex Formation with Neutrophil Elastase-As shown in Fig. 3, lanes  3 and 6, both of the yeast-produced mutant forms of al-PI formed NaDodS0,-stable complexes with neutrophil elastase, molecular weights of approximately 70,000 being obtained. However, there was some breakdown of the yeast al-PI methionine form. Based on the known molecular weights of each of the individual proteins being complexed, the data support the fact that these mutants form 1:l complexes with elastase. After oxidation of either inhibitor with N-chlorosuccinimide (5) there was a distinct effect on the methionine mutant. This protein (Fig. 3, lane 7) was unable to form complexes with neutrophil elastase, instead being degraded to low molecular weight peptides, as shown by the heavy staining material running near the bottom of the gel. The oxidized valine mutant continued to form complexes with neutrophil elastase, although minor degradation products could be detected (Fig.  3, lane 4).
Rates of Association of cul-Proteinase Inhibitor Yeast Variants with Proteinases-As shown in Table I the rates of interaction of either aI-PI variant (yeast methionine or yeast valine) with human neutrophil elastase were similar, but slightly less than those originally found for the native plasma inhibitor (3). However, although the methionine mutant was still able to slowly inactivate the other proteinases tested (rates not measured) there were significant differences with the valine mutant. It was able to inactivate pancreatic elastase a t a far more rapid rate than that of either the yeast methionine mutant or the plasma form of al-PI, but it was unable to inactivate trypsin, plasmin, thrombin, or Factor Xa, and only weakly inhibited cathepsin G. Of utmost importance was the fact that it was unaffected by oxidation and continued to rapidly form complexes with both neutrophil and pancreatic elastase, whereas the oxidized methionine mutant had no detectable inhibitory activity to any of the enzymes tested. In fact, it is possible that many of the interactions in which we could not find detectable inhibitory activity may actually represent rapid turnover of either the oxidized yeast methionine inhibitor or the yeast valine inhibitor by those specific enzymes.

DISCUSSION
The control of proteolytic activity by plasma proteinase inhibitors is critical for the maintenance of homeostasis in man. Accordingly, when control is lost either genetically or by chemical or enzymatic inactivation of the inhibitor in question, as exemplified by a,-PI (4, 5,29), tissue damage is likely. Recent advances in recombinant DNA technology have allowed production in bacteria (30) or yeast (10) of quantities of inhibitors sufficient for supplementation of deficient individuals. In addition, improvement of protein function has been made possible by site-directed mutagenesis of critical amino acid residues (10). The levels of recombinant heterologous protein that are expressed here are the highest yet reported using Saccaromyces cerevisiae (31) and demonstrate that yeast is a viable alternative system to bacteria for high level production of mammalian proteins. The data presented herein represent a first approach to modifying the structure of al-PI in order to provide an efficient inhibitor which is also unaffected by oxidizing agents. The choice of a valine replacement was based on three lines of reasoning as follows: (a) there was strong evidence that neutrophil elastase had a specificity directed towards amino acid residues with valine in the Pl-position (2); (6) an amino acid at the P1-position in al-PI was needed which was nonoxidizable and had the specificity for which elastase would be directed; (c) a natural mutation in the P1-position in al-PI Pittsburgh had led to a marked change in inhibitor specificity, supporting the primary importance of this residue (15).
The properties of each of the purified al-PI yeast mutants differed somewhat from those already known for plasma cyl-PI. Neither contained any carbohydrate (data not shown), and their molecular weights were, therefore, those expected for the deglycosylated form of normal plasma al-PI (32, 33). The amino terminal sequence, while differing slightly at the first residue (methionine instead of glutamic acid) was identical in all other respects. This modification, utilized at the DNA level for convenient linkage of al-PI cDNA to the GAPDH promoter is almost assuredly an innocuous alteration since active inhibitor missing the first nine amino acid residues of al-PI has already been isolated (34).
The lack of carbohydrate is most probably responsible for the differences in both the heat stability profiles and the halflives in rabbits of each mutant. The carbohydrate side chains contribute both bulk and negative charge to the molecule and their absence would increase losses through the glomerular membrane of the kidney. Whether the increased turnover is due to increased renal clearance or increased circulatory catabolism has not been determined. However, it can be concluded that the nonglycosylated yeast al-PI variants are likely to have half-lives in the human of close to one day versus the six-day half-life of the normal human plasma al-PI (28).
In the purification of the al-PI methionine yeast variant we noted an increased sensitivity to oxidizing agents, based on the fact that the protein could be stabilized if kept in the presence of reducing agents. This may also be a reflection of some type of protective effect given by the carbohydrate moiety in the normal plasma al-PI. As one might have predicted, the valine mutant was stable to oxidizing agents and continued to form complexes after oxidation. This indicates the unimportance of the Ps methionine residue which is also oxidized during the inactivation of plasma al-PI (5). Furthermore, there was very little change in the second order asso-ciation rate constant after oxidation, suggesting the fact that other oxidizable residues in this protein play a limited role in its function.
Of particular importance in this study was the finding that alteration in the PI residue of al-PI resulted in a major change in inhibitor specificity. While both mutants showed association rates with neutrophil and pancreatic elastase that were slightly slower than that of the normal plasma form of al-PI, presumably because of structural changes occurring in the absence of carbohydrate, the valine mutant also had a considerably altered specificity. It interacted more strongly with pancreatic elastase and more weakly with cathepsin G, but it no longer showed any interactions with the trypsin-like serine proteinases tested. Such results parallel data obtained with the Pittsburgh mutant where the change to an arginine residue in the P1 position had caused a significant increase in the rates of interaction with trypsin-like serine proteinase^.^ Therefore, the production of mutants of al-PI with alterations in the P1 position could lead to a series of inhibitors with targeted specificities towards individual proteinases. For example, an aromatic amino acid or leucine in PI would probably result in an inhibitor with increased activity towards chymotrypsin-like proteinases, such as cathepsin G, especially since it has already been reported that leucine is at the PI position in al-antichymotrypsin (35). The availability of modified inhibitors makes possible a detailed study of their potential use as therapeutic agents. Supplementation, on a day-to-day basis, of individuals genetically deficient in al-PI may require larger quantities of recombinant inhibitor due to decreased half-lives of molecules lacking carbohydrate. Furthermore, it is not known whether such nonglycosylated proteins will be antigenic. However, in acute conditions where the level of active, normal al-PI is compromised, recombinant al-PI and oxidation-resistant derivatives may prove to be extremely beneficial.