Location of the protease-inhibitory region of secretory leukocyte protease inhibitor.

Secretory leukocyte protease inhibitor (SLPI) is a two-domain protein that inhibits a wide range of proteases including chymotrypsin, leukocyte elastase, and trypsin. Based on its homology to other protease inhibitors and on x-ray crystallography of an SLPI-chymotrypsin complex it has been proposed that the elastase and chymotrypsin-inhibitory site is in the COOH-terminal domain and that the trypsin-inhibitory site is in the NH2-terminal domain. We have prepared muteins of SLPI by site-directed mutagenesis of a synthetic gene for the protein, followed by expression in Escherichia coli. The protease-inhibitory activities of these muteins indicate that leucine 72 in the COOH-terminal domain is at the inhibitory site for elastase and chymotrypsin. Unexpectedly, our measurements indicate that the trypsin-inhibitory site is not in the NH2-terminal domain. Instead they suggest that leucine 72 is also the inhibitory site for trypsin, even though the amino acid residues at the inhibitory sites of other trypsin inhibitors are almost always either lysine or arginine.


From Synergen
Inc., Boulder, Colorado 80301 Secretory leukocyte protease inhibitor (SLPI) is a two-domain protein that inhibits a wide range of proteases including chymotrypsin, leukocyte elastase, and trypsin.
Based on its homology to other protease inhibitors and on x-ray crystallography of an SLPI-chymotrypsin complex it has been proposed that the elastase and chymotrypsin-inhibitory site is in the COOHterminal domain and that the trypsin-inhibitory site is in the NH*-terminal domain. We have prepared muteins of SLPI by site-directed mutagenesis of a synthetic gene for the protein, followed by expression in Escherichia coli. The proteaseinhibitory activities of these muteins indicate that leutine 72 in the COOH-terminal domain is at the inhibitory site for elastase and chymotrypsin.
Unexpectedly, our measurements indicate that the trypsin-inhibitory site is not in the NH&erminal domain. Instead they suggest that leucine 72 is also the inhibitory site for trypsin, even though the amino acid residues at the inhibitory sites of other trypsin inhibitors are almost always either lysine or arginine.
Two groups have recently reported the primary structure of a novel human protease inhibitor, secretory leukocyte protease inhibitor (SLPI)' (1) and of a two-chain form of this inhibitor, human seminal plasma inhibitor 1 (2). This protease inhibitor is found in various secretory fluids including parotid secretions, bronchial, nasal, and cervical mucus, and seminal fluid (for reviews on this protein see Refs. 2,3). The inhibitor forms complexes with a variety of proteolytic enzymes including the neutrophil proteases elastase and cathepsin G and the pancreatic proteases chymotrypsin and trypsin. The protein is of interest because it appears to be an important component of the antiprotease defense of tissues bathed by the secretory fluids. SLPI could be useful as a therapeutic in degenerative and inflammatory diseases that lead to proteolytic damage to these and other tissues.
The amino acid sequence of SLPI led to the prediction that the protein consists of two highly homologous domains of 53 and 54 amino acids (1, 4). This prediction was supported by the gene structure showing that the two domains are encoded on separate exons (5), and has recently been confirmed by xray crystallographic structure determination of recombinant SLPI (6). On the basis of a weak homology with the Kazal class of protease inhibitors (7), two regions of the protein were identified as being likely sites of interaction with pro-* 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.
teases (1,4,6). One of these, present in the COOH-terminal domain of SLPI, was proposed to be the site interacting with chymotrypsin-like enzymes and elastase. This prediction of the chymotrypsin inhibitory site of SLPI has recently been confirmed by Grutter and colleagues (6) who have determined the structure of a crystal of the SLPI-chymotrypsin complex. The other region, present in the homologous position of the NH&erminal domain, has been proposed as the site of interaction with trypsin.
Numerous investigations of the interactions of proteases and their inhibitors have shown that these inhibitors bind to proteases in the same manner as substrates (7). As a consequence, proteases generally show a similar specificity for inhibitors and substrates, particularly in terms of the amino acid residue Pl of inhibitors/substrates that binds to subsite Sl of the enzyme. Based on this understanding, and a knowledge of the key amino acid residues that interact with the protease, it has been possible by elegant protein chemistry, and by site-directed mutagenesis techniques, to produce nonfunctional inhibitors and to target inhibitors to new enzymes (8-11). We reasoned that the unknown amino acid residues of SLPI that interact with elastase and trypsin could be identified by determining which residues, when changed, change the specificity of the inhibitor.
We have, therefore, studied the properties of a number of variants of SLPI prepared by site-directed mutagenesis and expression of the gene for SLPI in Escherichia coli.
The results we have obtained provide independent confirmation of Grutter's assignment of the chymotrypsin-inhibitory site of SLPI to residue Leu7* of the COOH-terminal domain. They also indicate that the same residue, Leu7', is the inhibitory site for leukocyte elastase. However, the results clearly indicate that Arg", the residue homologous to Leu7* in the NH,-terminal domain, is not the residue interacting with the Sl subsite of trypsin. Instead, they indicate that the trypsin-inhibitory site of SLPI is also residue Leu7', even though Leu is not a residue commonly thought to bind strongly to the Sl subsite of this enzyme.

Expression Vector
The vector for expression of SLPI, pCJXI-2, was derived from pKK223-3 (Pharmacia LKB Biotechnology Inc.) by replacing the partial tetracycline-resistance gene with the complete pBR322-derived gene' and by inserting an XhoI linker (New England Biolabs, catalog number 1030) into the unique PuuII site, and cloning the lacl  (34). Amino acid residues R*', L7', M7", and L7', whose codons were subjected to site-directed mutagenesis, are indicated.
to a density of 2 x 10' cells/ml. At this point, the culture was induced with 1 mM isopropyl P-D-thiogalactopyranoside, and after 3 h of continued incubation the culture was cooled and the cells were pelleted, washed with 0.05 M Tris-HCI (pH 7), repelleted, and frozen. To prepare SLPI and its variants the cells were treated as follows. The pellet was thawed and suspended in 100 mM Tris-HCI (pH 7.5) containing 8 M urea, 4 mM Na2EDTA, 50 mM 2-mercaptoethanol and passed twice through a French press at 12,000 p.s.i. to lyse the cells and shear the DNA. The lysate was then adsorbed to Sephadex SP-C25 (Pharmacia LKB Biotechnology Inc.) and was eluted with 0.4 M guanidine HCl, 100 mM Tris-HCl (pH 7.5), and 50 mM 2-mercaptoethanol.
An equal volume of 6 M guanidine HCl was added to the SP-C25 eluant followed by sequential additions of dithiothreitol to 3.5 mM and oxidized glutathione to 13.5 mM. The mixture was then diluted with 4.5 volumes of 50 mM Tris-HCl (pH 10.7) and cysteine was added to a final concentration of 3 mM. After an overnight incubation at 25 "C the SLPI-refolding mixture was adjusted to pH 6.0 and diluted with 20 mM MES (pH 6.0) to lower the guanidine HCl concentration to below 150 mM. It was applied to a Mono S HR 5/5 column (Pharmacia LKB Biotechnology Inc.) equilibrated with 20 mM MES (pH 6.0) and was eluted with a l%/min NaCl gradient in the same buffer.

Preparation of SLPI and SLPI Variants
SLPI protein is conveniently expressed as part of a twocistron operon (see "Materials and Methods") such that translation of the SLPI gene is coupled to translation of an efficiently expressed upstream gene, in this case the first 10 codons of the E. coli ompA gene. Transcription of this operon from the Tat promoter is very tightly regulated due to the presence of high levels of lac repressor expressed from the la0 gene found on the F factor and the wild-type lacl gene carried on the plasmid. Thus, very low levels of SLPI are present in the uninduced state. However, when the culture is induced with isopropyl P-D-thiogalactopyranoside, SLPI is normally expressed to a level of 510% of total cellular protein, making purification, activation, and assaying the protein relatively straightforward.
We have shown, in the case of the parent protein, that the NH2-terminal methionine residue is removed by processing systems in E. coli, and that the recombinant protein has a sequence identical to that of human SLPI.
Site-specific mutagenesis of the SLPI gene was performed using standard methods. A list of oligonucleotides and the corresponding amino acid changes is shown in Table I. The efficiency of mutagenesis was usually between 30 and 50%. SLPI and SLPI variants can be readily purified from lysed cells by cation-exchange chromatography and can be induced to fold to an active conformation by reduction and denaturation followed by oxidation and dilution of the denaturant, and a Targeted amino acid residue (see Fig. 1) and its respective change. * Oligonucleotides were synthesized as described under "Materials and Methods." The anticodon of the residue of interest is underlined.
the addition of cysteine (25). The active protein can then be purified to near homogeneity by a second cation-exchange chromatography step. Recombinant SLPI subjected to this treatment has physical and chemical properties that are indistinguishable from those of the human protein. The SLPI variants studied here differ from natural SLPI in the regions initially identified from the protein sequence as being good candidates for the sites interacting with proteases. Thus the DNA coding for the Leu-Met-Leu sequence at residues 72, 73, and 74, thought to be the region of the protein interacting with chymotrypsin and elastase, has been heavily mutagenized.
The DNA coding for Argo, a candidate for the trypsin-binding site, has also been mutagenized. The proteins produced by each of these mutant genes have been purified in the manner described for the natural protein and have been assayed for their effect on the three proteases, human leukocyte elastase, bovine a-chymotrypsin, and bovine trypsin.

Protease-inhibitory Properties of SLPI Variants
Leukocyte Elastase and Chymotrypsin-Peptide bonds COOH-terminal to Met and Leu are often good substrates of chymotrypsin and leukocyte elastase, whereas analogous peptide bonds COOH-terminal to Gly are very poor substrates of these enzymes (26,27). Interestingly, when Leu7' of SLPI is replaced by Gly the protein loses most of its ability to inhibit chymotrypsin and some of its ability to inhibit elastase. In contrast, when Met73 or Leu74 of SLPI is replaced by Gly the protein retains some of its activity against chymotrypsin and has close to full activity against elastase (Table II).
Although these findings are consistent with Leu" being the residue interacting with the Sl subsites of elastase and chymotrypsin, they are also consistent with the alternative hypothesis that the Gly7' variant of SLPI is unable to fold to an active form under the same conditions as the natural protein.
To explore this possibility we compared other properties of the mutant and natural proteins. By several tests SLPI-Gly7' and the natural protein appear to have quite similar conformations.
We first explored the chromatographic properties of SLPI and SLPI-Gly7' on reversed-phase and cation-exchange HPLC. In other experiments we have shown that these properties are quite sensitive to the conformation of SLPI. Active and unfolded forms of SLPI elute at 25 and 30% acetonitrile, respectively, on the C!, column used for reversed-phase chromatography and at 350 mM and 370 mM NaCl, respectively, on the Mono S column used for cation-exchange chromatography. Despite the sensitivity of these techniques to conformation, we were unable to demonstrate any chromatographic differences between SLPI and SLPI-Gly7' on either Cs or Mono S columns.
Another remarkable feature of the folded form of SLPI is its resistance to proteolysis by a number of proteases that are not inhibited by this protein. Since denatured SLPI is an excellent substrate for these proteases the resistance to proteolysis of the native form of this protein is likely to be a measure of its conformational integrity. We find that SLPI-Gly7* shares the marked resistance of SLPI to digestion with pepsin and with thermolysin (3). Of the many pepsin-cleavable bonds of SLPI, only one (Leu72-Met73) is subject to pepsin cleavage in its native conformation since NH*-terminal sequencing of the digested protein shows the presence of two sequences starting at Ser' and Met73. This same resistance to proteolysis is seen with SLPI-Gly7*, with the added twist that even the bond between residues 72 and 73 is now resistant to pepsin, doubtless because Gly-X bonds are not substrates for this enzyme. Similarly, of the many thermolysin-cleavable bonds in SLPI, only two (Cys's-Leu" and MetT3-LeuT4) are subject to thermolysin cleavage in the native protein. SLPI-GlyT2 shares this resistance to proteolysis by thermolysin in that only the Cys"-Leulg bond is cleaved by this enzyme. The absence of one cleavage, characteristic of the natural molecule, is probably attributable to the amino acid substitution (Gly7*) close to the prospective site of cleavage. However, the absence of hydrolysis at all other sites is likely attributable, as with the native molecule, to a characteristic tertiary structure that restricts access of the proteases to what would otherwise be susceptible bonds. We conclude that the variant SLPI-Gly7* has a similar structure to natural SLPI. Its reduced ability to inhibit chymotrypsin and elastase is likely, therefore, a consequence of the fact that Leu7' of the natural inhibitor binds to the Sl subsite of these proteases and contributes strongly to the binding energy of the inhibitor to the enzyme. To confirm the hypothesis that residue Leu7' binds to the active site of leukocyte elastase we made a further variant of SLPI with Phe substituted for Leu7*. It is known that leukocyte elastase cannot hydrolyze peptide bonds COOH-terminal to Phe but that these bonds can be hydrolyzed by chymotrypsin (28). In accord with the idea that Leu7' of SLPI binds to the Sl subsite of elastase we observed that SLPI-Phe7' is a weak inhibitor of this enzyme. However, SLPI-Phe72 does inhibit chymotrypsin, indicating that the protein has adopted the correct information.
The latter result provides a compelling argument that the substitution of Phe for Leu has specifically altered the ability of the protein to inhibit elastase by changing a residue that is critical for the interaction with that enzyme.
Trypsin-Peptide bonds COOH-terminal to Arg are generally excellent substrates of trypsin whereas analogous peptide bonds COOH-terminal to Gly are very poor substrates for this enzyme. However, when Ar$O of SLPI is replaced with Gly the resulting protein is still active as an inhibitor of trypsin and the Kd of the complex of trypsin with SLPI-Gly2' is almost indistinguishable from that of natural SLPI (Table  II). That this variant still inhibits trypsin argues strongly against the view that Arg20 is the residue of SLPI binding to the Sl subsite of trypsin. Also inconsistent with this view are our findings that replacement of Arg*' by Val or Met gives SLPI variants that have activity against trypsin. We conclude that Arg2" is not involved in interaction between SLPI and trypsin.
Earlier results from our laboratory have shown that the COOH-terminal domain of SLPI expressed from a partial gene in yeast has full activity against elastase and chymotrypsin but has little or no activity against trypsin (29), leading us to discount the possibility that the residues in this domain could play a significant role in the antitrypsin activity of the inhibitor.
More recently, we have split the two domains of SLPI by treatment with formic acid and again found undetectable antitrypsin activity in the COOH-terminal fragment. However, when we tested the antitrypsin activity of the NH2terminal domain produced by formic acid cleavage we found that it was also inactive against trypsin, indicating that the antitrypsin activity of SLPI is a property of the whole molecule, and that the antitrypsin site could well reside in the COOH-terminal domain. We therefore turned our attention to the variants in the COOH-terminal domain in our search for the trypsin-inhibitory site. Neither SLPI-G~Y~~ or SLPI-G~Y~~ inhibit trypsin (Table  II) indicating that residue Leu72 or Met7" might occupy the Sl subsite of trypsin in SLPI-trypsin complexes. This would be surprising because neither of these residues have the basic side chain expected of the Pl residues of normal trypsin substrates and inhibitors.
To investigate further the possibility that either Leu7* or MetT3 of SLPI may be the Pl residue in SLPI-trypsin complexes we changed residue 72 to Val, since it had been noted earlier that changing the Pl residue of oil protease inhibitor to Val reduced the ability of this protein to inhibit trypsin (30). SLPI-Va172 is a poor trypsin inhibitor but appears to have the correct conformation since it is an excellent elastase inhibitor (Table II). This suggests that Leu7' is the trypsin-inhibitory active site of SLPI. To obtain further evidence for this possibility we substituted a Lys residue and an Arg residue for the Leu at position 72. Since peptides with Pl Lys or Arg residues bind much more strongly to trypsin than peptides without a basic group at this position, we would expect these variants to be considerably stronger inhibitors of trypsin than the natural molecule if this residue binds to the Sl subsite of trypsin. We found that the dissociation constants of both SLPI-Lys7'-trypsin and SLPI-Arg7'-trypsin complexes are at least 1000-fold lower than that of the natural SLPI-trypsin complex (Table II). We conclude that either Leu7* of SLPI is the trypsin-inhibitory active site, or that some residue other than Leu72 could be the original trypsin-inhibitory site and changing Leu7' to Lys or Arg has created a new antitrypsin site which has masked the original.
It is possible to distinguish between these two possibilities by succinylating SLPI-Lys7*. Succinylation of the Pl lysine residue of a protease inhibitor should reduce its ability to bind to the trypsin-active site. As a control for this experiment we showed that the activity of bovine pancreatic trypsin inhibitor, whose active site is known to be Lys18 (31), is destroyed by succinylation (Table III). The antitrypsin activity of wild type SLPI is only slightly affected by succinylation; therefore, we reasoned that if SLPI-Lys7' has a completely novel antitrypsin site succinylation should inactivate this site and reveal the continued presence of the old antitrypsin site. If, on the other hand, the substitution of Lys for Leu72 has simply improved the old antitrypsin site, succinylation should abolish the antitrypsin activity of SLPI-LYS~~ altogether. The data in Table III show that succinylated SLPI-Lys7' has no significant trypsin-inhibitory activity, thus favoring the hypothesis that residue 72 of SLPI is the trypsin-inhibitory active site.

DISCUSSION
The specificities of proteases for substrates and inhibitors are known to show extensive similarities.
At the structural level, this result is explicable in terms of the similar binding modes of substrates and inhibitors.
The substrate specificity of the three enzymes studied here has been established by several groups. Briefly summarized, these investigations indicate that trypsin is highly specific for peptide bonds COOHterminal to arginine and lysine, chymotrypsin is specific for bonds COOH-terminal to large hydrophobic and especially aromatic amino acid residues, and leukocyte elastase is specific for bonds COOH-terminal to amino acids with moderately large hydrophobic side chains. Although these substrate specificities are suggestive in pointing to which residues of SLPI might bind to the Sl subsites of these enzymes, they are insufficient to unambiguously define the region of SLPI interacting with these proteases. To distinguish those regions where the polypeptide chain of SLPI accidentally conforms to the specificity of the protease from those where the match is vital to the function of the inhibitor, we have systematically altered the sequence of the inhibitor, and determined how these changes affect its specificity.
It is possible to envision three extreme results of this kind of investigation.
An amino acid substitution that would greatly reduce the ability of a peptide to be a substrate but that has no effect on the inhibitory ability of SLPI should indicate unambiguously that the residue substituted does not play a vital role in SLPI interaction with the protease. In contrast, an amino acid substitution that would reduce the ability of a peptide to act as a substrate and which also reduces the inhibitory activity of SLPI supports, but does not conclusively prove, that the residue altered plays a role in the enzyme-inhibitor interaction: an amino acid substitution that prevents the protease inhibitor reaching its active conformation will also produce this result. To confirm that the residue changed has some direct role in the enzyme-inhibitor interaction, some independent evidence that the defective inhibitor has reached the "active" conformation is required. Finally, an amino acid substitution that would improve the ability of a peptide to act as a substrate and that improves the activity of SLPI strongly supports the hypothesis that the residue changed plays an important role in the enzyme-inhibitor interaction.
By the criteria described above we believe that we have proved that residue Leu72 of SLPI binds to the Sl subsite of leukocyte elastase. The evidence in favor of this hypothesis is that a change from Leu to Gly or Phe or Lys at this position reduces the protein's ability to act as an elastase inhibitor and that by several other criteria, including in the latter two cases a continued ability to inhibit related proteases (chymotrypsin and trypsin, respectively), the variant inhibitor appears to be in the native conformation.
The change from Leu to Gly or Phe in an elastase substrate is known to abolish the ability of a peptide to act as an elastase substrate (27).
By the above criteria we believe that we have also located the chymotrypsin-active site of SLPI at Leu?'. In favor of this hypothesis is the observation that SLPI-Gly7', SLPI-Lys7', and SLPI-Arg7* are poor chymotrypsin inhibitors, even though their conformations appear to be similar to that of the natural protein. This evidence is particularly strong for SLPI-Lys7* and SLPI-Arg7' because these proteins retain the ability of SLPI to inhibit trypsin. Residues Met73 and Leu74, two other candidates for binding to the chymotrypsin-active site, are unlikely to bind to subsite Sl because SLPI-GlyT3 and SLPI-GlyT4 are still chymotrypsin inhibitors (relative to SLPI-Gly7') although they are poorer than wild type SLPI. The reduced affinity of SLPI-GIY~~ for chymotrypsin is easily explained in terms of its binding to subsite Sl' of chymotrypsin since this subsite also shows some specificity for hydrophobic residues (27). Our work, therefore, confirms the conclusion of Grutter et al. (6) from x-ray crystallography of a complex of SLPI with chymotrypsin.
By the criteria described above we believe that we have conclusively demonstrated that A&' is not the trypsin-inhibitory site of SLPI. The best candidate for this residue appears to be Leu7'. In favor of this assignment we note that SLPI-GlyzO, SLPI-Met", and SLPI-Val*' are all good inhibitors of trypsin, whereas SLPI-GlyT2 and SLPI-Va17' are not good inhibitors of trypsin even though, as shown above, the latter two proteins appear to have the correct conformation.
SLPI-Gly73 is also not a strong trypsin inhibitor suggesting that Met-73 could be the trypsin-inhibitory site. The strongest evidence in favor of the hypothesis that Leu7' is the trypsin-inhibitory site is the fact that SLPI-Lys'* and SLPI-Arg7' are much stronger inhibitors of trypsin than native SLPI. In addition, the trypsin-inhibitory activity of SLPI-Lys7*, unlike that of SLPI itself, is fully inactivated by treatment with succinic anhydride. Finally, we note that Kramps et al. (32), have previously proposed that the antielastase and antitrypsin sites of SLPI might be the same on the basis of their similar susceptibility to oxidative inactivation.
The finding that a nonbasic amino acid residue can occupy subsite Sl of the active site of trypsin, although surprising, is not entirely without precedent. The a-1 protease inhibitor is