The potency and specificity of the interaction between the IA3 inhibitor and its target aspartic proteinase from Saccharomyces cerevisiae.

The yeast IA3 polypeptide consists of only 68 residues, and the free inhibitor has little intrinsic secondary structure. IA3 showed subnanomolar potency toward its target, proteinase A from Saccharomyces cerevisiae, and did not inhibit any of a large number of aspartic proteinases with similar sequences/structures from a wide variety of other species. Systematic truncation and mutagenesis of the IA3 polypeptide revealed that the inhibitory activity is located in the N-terminal half of the sequence. Crystal structures of different forms of IA3 complexed with proteinase A showed that residues in the N-terminal half of the IA3 sequence became ordered and formed an almost perfect alpha-helix in the active site of the enzyme. This potent, specific interaction was directed primarily by hydrophobic interactions made by three key features in the inhibitory sequence. Whereas IA3 was cut as a substrate by the nontarget aspartic proteinases, it was not cleaved by proteinase A. The random coil IA3 polypeptide escapes cleavage by being stabilized in a helical conformation upon interaction with the active site of proteinase A. This results, paradoxically, in potent selective inhibition of the target enzyme.

The yeast IA 3 polypeptide consists of only 68 residues, and the free inhibitor has little intrinsic secondary structure. IA 3 showed subnanomolar potency toward its target, proteinase A from Saccharomyces cerevisiae, and did not inhibit any of a large number of aspartic proteinases with similar sequences/structures from a wide variety of other species. Systematic truncation and mutagenesis of the IA 3 polypeptide revealed that the inhibitory activity is located in the N-terminal half of the sequence. Crystal structures of different forms of IA 3 complexed with proteinase A showed that residues in the N-terminal half of the IA 3 sequence became ordered and formed an almost perfect ␣-helix in the active site of the enzyme. This potent, specific interaction was directed primarily by hydrophobic interactions made by three key features in the inhibitory sequence. Whereas IA 3 was cut as a substrate by the nontarget aspartic proteinases, it was not cleaved by proteinase A. The random coil IA 3 polypeptide escapes cleavage by being stabilized in a helical conformation upon interaction with the active site of proteinase A. This results, paradoxically, in potent selective inhibition of the target enzyme.
Aspartic proteinases participate in a variety of physiological processes, and the onset of pathological conditions such as hypertension, gastric ulcers, and neoplastic diseases may be related to changes in the levels of their activity. Members of this proteinase family, e.g. renin, pepsin, cathepsin D, and human immunodeficiency virus-proteinase are generally typecast on the basis of their susceptibility to inhibition by naturally occurring, small molecule inhibitors such as the acylated pentapeptides, isovaleryl-and acetyl-pepstatin. However, the two most recently identified human aspartic proteinases, ␤-site Alzheimer's precursor protein cleavage enzyme and ␤-site Alzheimer's precursor protein cleavage enzyme 2 (1, 2), are not inhibited by this classical type of inhibitor of this family of enzymes. Pepstatins are metabolic products produced by various species of actinomycetes and, as such, are not themselves gene-encoded. Protein inhibitors of aspartic proteinases are relatively uncommon and are found in only a few specialized locations (3). Examples include renin-binding protein in mammalian kidneys which intriguingly has now itself been identified to be the enzyme, N-acetyl-D-glucosamine-2-epimerase (4); a 17-kDa inhibitor of pepsin and cathepsin E from the parasite, Ascaris lumbricoides (5); proteins from plants such as potato, tomato, and squash (6,7), and a pluripotent inhibitor from sea anemone of cysteine proteinases as well as cathepsin D (8).
The IA 3 polypeptide in yeast is an 8-kDa inhibitor of the vacuolar aspartic proteinase (proteinase A or saccharopepsin) that was initially described by Holzer and co-workers (9). The complete sequence of this 68-residue inhibitor has been elucidated (10,11) and the inhibitory activity of IA 3 has been shown to reside within the N-terminal half of the molecule (10,12). We have recently solved the structure of the IA 3 -proteinase A complex (12), demonstrating that whereas free IA 3 has little intrinsic secondary structure, residues 2-32 of the inhibitor, upon contact with proteinase A, become ordered and adopt a near perfect ␣-helical conformation occupying the active site cleft of the enzyme. This was the first crystal structure to be determined for a gene-encoded aspartic proteinase inhibitor complexed with its target enzyme. It was thus considered important to investigate further the role of the proteinase as a folding template and to attempt to establish the molecular features that enable this unprecedented mode of inhibitorproteinase interaction to occur.
Wild type and mutant forms of IA3 were subcloned into the NdeI-XhoI sites of pET-22b (Novagen, Cambridge, United Kingdom), thus introducing a C-terminal Leu-Glu-His 6 tag. Escherichia coli strain BL21DE3(pLysS) was transformed with wild type or mutant clones, then grown at 37°C in LB medium to an A 600 of ϳ0.6 before induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside. Each soluble recombinant protein was loaded onto a nickel-chelate affinity column in 0.05 M sodium phosphate buffer, pH 8.0, containing 0.3 M NaCl, washed with the same buffer adjusted to pH 6.0 containing 10% glycerol and each protein form of IA 3 was eluted with the glycerol containing buffer at pH 4.0. Appropriate fractions containing IA 3 were then heated twice at 100°C for 5 min.
Analytical Measurements-N-terminal sequencing of wild-type and mutant forms of IA 3 was performed by automated Edman degradation on protein bands that were electroblotted onto polyvinylidene difluoride membrane following SDS-polyacrylamide gel electrophoresis on 20% gels. Samples for amino acid analysis were hydrolyzed for 16 h at 105°C in 6 M HCl before being loaded onto a Biochrom 20 amino acid analyzer (Amersham Pharmacia Biotech, Cambridge, UK). MALDI-TOF mass spectrometry was carried out using a PE Biosystems Voyager Elite XL instrument incorporating a UV laser and delayed extraction. Samples at ϳ10 pmol/l were added to a matrix solution consisting of ferulic acid (55 mg/ml in ethanol containing 0.1% trifluoroacetic acid) in a sample/matrix ratio of 3:1.
Kinetic Measurements and Peptide Cleavage-Inhibition assays were conducted at pH 3.1 and 4.7 as described previously (12,14) using Lys-Pro-Ile-Glu-Phe*NitroPhe-Arg-Leu (where the asterisk indicates the scissile peptide bond) as substrate for all enzymes except yapsin 1 (15). For this enzyme, the substrate used was ACTH (residues 1-39). This was incubated at 37°C in 50 mM sodium acetate buffer, pH 5.3, with 10 units of purified yapsin 1 (15) for 30 min in the absence and presence of peptide 1 at final concentrations up to 20 M. The proteolytic product from ACTH (residues 1-15) was detected and quantified by reverse phase high performance liquid chromatography.
The susceptibility of peptide/protein forms of IA 3 to proteolytic cleavage was examined by incubation at 37°C for various lengths of time (commonly 16 or 72 h) with the appropriate enzyme in 100 mM sodium formate/acetate buffers at pH 3.1/4.7, respectively, each containing 300 mM NaCl. Each synthetic peptide was incubated initially with proteinase A for 16 h at a molar ratio of 40:1 and, if no cleavage was detected under these circumstances, then the incubation was repeated at a peptide:proteinase ratio of 10:1 for 72 h. Peptide 1 was incubated with nontarget proteinases such as human pepsin at both pH 3 and 5 at a molar ratio of 1,000:1 for only 1 h. Each digest was separated by reverse-phase fast protein liquid chromatography using a Pep-RPC column (Amersham Pharmacia Biotech, Bucks, United Kingdom) and the fractions containing any cleavage products were collected and subjected to acid hydrolysis followed by amino acid analysis. Protein samples were frozen for storage prior to subsequent mixing with the matrix solution for application to a stainless steel target and analysis by MALDI-TOF mass spectrometry.
Crystallography and Molecular Modeling-Crystals of a complex of proteinase A with the K24M mutant protein form of IA 3 (Table I) were grown by vapor diffusion under the conditions described previously for other IA 3 -proteinase A complexes (12). The initial solution was prepared at a molar ratio of inhibitor:proteinase of 5:1, and after separation from the excess of inhibitor by gel filtration on Sephadex G-50, was concentrated to 5 mg/ml by ultrafiltration. The mother liquor contained 30% PEG1500, 0.14 M ammonium sulfate in 0.1 M MES buffer, pH 6.0. Data extending to 1.9 Å were collected at 100 K on beamline X9B at NSLS, Brookhaven National Laboratory, Upton, NY, using an ADSC 4K CCD detector. Data were processed with HKL2000 (16). The initial data set consisted of 217,446 reflections that could be scaled with R sym of 8.7% (last shell 34.6%) to yield 41,718 unique measurements. The completeness was 92.8% for the whole data and 75.5% for the final shell. The structure of proteinase A complexed with peptide 1 (12) was used as the initial model with replacement of Lys 24 by Met. The structure was refined with CNS 1.0 (17) utilizing data extending to 2.0-Å resolution. The first two rounds of refinement included positional and B factor refinements and model adjustment, while solvent molecules were added in the third round. The final model contained the enzyme, residues 2-31 of the inhibitor and 243 water molecules. The final R factor was 19.84% and R free was 23.1%. The root mean square deviations for bond lengths and bond angles from ideality was 0.012 and 1.59 Å, respectively.
Modeling calculations were made on a Silicon Graphics Octane with a single R12000 processor using the Moloc modeling package. Individual amino acid side chains (K18M/D22L) in IA 3 were changed with a built-in function in Moloc. Side chain conformations were adapted manually and a subsequent round of optimization, maintaining proteinase A and the remainder of IA 3 fixed, resulted in low energy conformations for the newly introduced side chains. The new protein-inhibitor complex was checked for attractive and repulsive interactions, and allowed conformations, respectively.

Interaction of Protein/Peptide Forms of IA 3 with Target and
Nontarget Enzymes-The nucleotide sequence encoding all 68 residues of IA 3 was amplified by PCR and introduced into the pET-22b vector for expression in E. coli as described under "Experimental Procedures." The recombinant protein that accumulated in E. coli was soluble and was purified to homogeneity from cell lysates by taking advantage of the His 6 tag introduced from the pET-22b vector at the C terminus of the IA 3 polypeptide chain (see "Experimental Procedures"). N-terminal analysis of one batch of the homogeneous wild-type inhibitor through 10 cycles of Edman degradation gave the sequence Asn-Thr-Asp-Gln-Gln-Lys-Val-Ser-Glu-Ile, which is exactly coincident with that predicted by the DNA sequence for residues 2-11 (Table I) (11) and indicates that the initiator Met 1 residue had been removed during the accumulation of this batch of recombinant protein in E. coli. Analysis of a separate batch of recombinant wild-type inhibitor by MALDI-TOF mass spectometry gave a mass of 8772 Da, identical with that predicted (8772 Da) for the IA 3 sequence plus the C-terminal extension of ϳLeu-Glu-His 6 introduced from the pET-vector.
The K i value determined at pH 3.1 for the inhibition of yeast proteinase A by this C-terminal tagged, wild-type recombinant protein (wild-type in Table I) was comparable to that reported previously at the same pH for the naturally occurring protein purified from S. cerevisiae (14). It is readily apparent then that the introduction of the extra ϳLeu-Glu-His 6 residues at the C terminus of the recombinant inhibitor did not have any significant detrimental effect on inhibitory potency. Since the target proteinase is unlikely ever to encounter a pH as low as 3.1 in its cellular environment, attempts were also made to determine inhibition constants at higher pH values such as 4.7 and 6.0. In both cases, the interaction with proteinase A was so tight that the K i values lay at or beyond the limits of accurate determination using the available assay methodology and so were estimated to be Ͻ0.1 nM.
A synthetic peptide which spanned residues 2-34 of the IA 3 sequence (Peptide 1 in Table II) was found to have inhibitory potency against yeast proteinase A at pH 3.1 and 4.7 compa-rable to those of the naturally occurring and wild-type recombinant protein forms of IA 3 , described previously (12,14). In contrast, peptide 1, at a concentration of 2 M, had no significant ability to inhibit any one of a number of other aspartic proteinases from a wide range of other species (Table III). These have considerable sequence and structural similarities to yeast proteinase A and included yapsin 1 (a membraneattached aspartic proteinase also from S. cerevisiae (15) and other enzymes of fungal, mammalian, parasite (plasmepsin II from Plasmodium falciparum (18) and plant (cyprosin from Cynara cardunculus (19)) origin. Thus, IA 3 is a potent specific inhibitor directed solely against its target enzyme, yeast proteinase A. Since peptide 1 had no effect on the nontarget enzymes listed in Table III, reciprocally, the effect of a number of these (including e.g. human pepsin, cathepsin D, and cathepsin E) on peptide 1 was examined. With human pepsin, for example, using a molar ratio of peptide 1:pepsin of 1,000:1 at pH 3.1 and pH 5, peptide 1 was cleaved rapidly at the ϳGlu 10 * Ile 11 ϳ and ϳAla 29 *Phe 30 ϳ bonds, as revealed by amino acid analysis of the collected peptide fragments (data not shown). Identical results were obtained for the other enzymes so it would appear that peptide 1 is unable to inhibit the nontarget aspartic proteinases such as pepsin, cathepsins D and E (and the other enzymes listed in Table III) because these enzymes cut the polypeptide effectively as a substrate. Consequently, the ϳAla 29 *Phe 30 ϳ and ϳGlu 10 *Ile 11 ϳ sites that were cleaved by the nontarget proteinases were changed singly and in concert to introduce residues at the P 1 position which were known from extensive previous studies (e.g. Refs. 20 and 21) to be refractory to cleavage by such enzymes as human pepsin and cathepsin D. The resultant peptides (2 and 3, Table II) were just as potent as peptide 1 as inhibitors of proteinase A but still showed no significant ability to inhibit, e.g. human pepsin or cathepsin D at concentrations as high as 5 M. Although cleavage of peptides 2 and 3 between residues ϳVal 29 -Phe 30 ϳ and

NTD QQKVS EIFQS S
No inhibition a ND, not determined. b NI, no inhibition at a final concentration of 2 M.

TABLE III
Aspartic proteinases unaffected by the IA 3

inhibitor
The activity of each enzyme at the appropriate, indicated pH value was not affected significantly by inclusion of peptide 1 at a final concentration of 2 M in each assay.   3 Recombinant protein forms of IA3 consisting of 68 residues plus a ϳLeu-Glu-(His) 6 tag at their C terminus were produced in E. coli and purified to homogeneity. The sequence of residues 2-32 (only) of each protein is given. Yeast IA 3 Inhibitor ϳLys 10 -Ile 11 ϳ plus ϳVal 29 -Phe 30 ϳ, respectively, was no longer evident, nevertheless, cleavage at other locations now became apparent. For example, after incubation of peptide 3 with human cathepsin D at pH 3.1, the digest was analyzed by MALDI-TOF mass spectrometry and the large product that was detected (3238 Da observed; 3237 Da predicted) indicated that cleavage had occurred between residues ϳGln 5 *Gln 6 ϳ to generate the fragment spanning residues 6 -34. Thus, peptides 1, 2, and 3 appeared to be cleaved by the nontarget proteinases including human pepsin and cathepsin D at whatever peptide bonds were accessible and which met the sub-site specificity requirements of each enzyme. In contrast, peptides 1, 2, and 3 were not cleaved by proteinase A (at pH 3.1 or 4.7), even upon prolonged incubation for 3 days at 37°C at molar ratios of peptide:proteinase A as low as 5:1. Similarly, the recombinant, wild-type protein form of IA 3 (Table I) was not cleaved by yeast proteinase A since the mass ion (8772 Da) observed by MALDI-TOF mass spectrometry after prolonged incubation was identical to that of the starting material. However, the wild-type protein was digested by pepsin and cathepsin D, e.g. analysis of the material incubated at pH 3.1 with human cathepsin D indicated that one large product had accumulated which was consistent in size (5575 Da observed; 5578 predicted) with that of a fragment spanning residues Phe 30 -Glu 68 plus the ϳLeu-Glu-(His) 6 tag. Thus the full-length protein form of IA 3 also appears to be degraded by nontarget proteinases with cleavage taking place, at least, at one of the bonds (ϳAla 29 *Phe 30 ϳ) that was identified earlier for peptide 1 (no attempts were made to detect by mass spectrometry any small products from the N terminus of IA 3 in the cathepsin D digest). This ready susceptibility of IA 3 in both peptide and protein forms to proteolytic cleavage by nontarget enzymes provides further substantiation to our conclusion described previously (12) that the free IA 3 polypeptide has little intrinsic secondary structure.
Truncation and Mutagenesis of IA 3 -Since the inhibitory activity of IA 3 toward proteinase A appeared to be contained within residues 2-34 (compare K i values for the wild-type protein form (Table I) and the peptide form (peptide 1, Table  II)), the effect on inhibitory potency of further truncation of this sequence was examined using a systematic series of synthetic peptides in which residues were successively deleted from the N and C termini. Removal of Asn 2 had little effect on the K i value measured at pH 3.1 (compare peptide 4 with peptide 1, Table II) but the potency was diminished somewhat at pH 4.7, as a K i value was readily quantified for peptide 4 at this pH value (Table II). Similarly, deletion of residues 3-6 progressively diminished inhibitory potency (peptides 5 and 6, Table  II) and the absence of residues 2-11 (in peptide 7) resulted in almost complete abolition of the inhibitory activity. The activity of IA 3 was totally destroyed when residues 2-15 were lacking (peptide 8, Table II). The residues at the N-terminal end of the 2-34 polypeptide sequence would thus appear to contribute substantially to the potency of inhibition.
To investigate the importance of the individual side chains of these residues, alterations in sequence were introduced into the full-length protein form of IA 3 at its N terminus by PCR mutagenesis as described under "Experimental Procedures." Initially, residues 2-10 of the wild-type IA 3 sequence were replaced with Gly-Gly-His-Asp-Val-Pro-Leu-Thr-Asn. This is the sequence of residues that is present at the N terminus of the target enzyme, yeast proteinase A itself (22) and, as such, was chosen totally arbitrarily. The recombinant, chimera inhibitor was purified to homogeneity and 10 cycles of Edman degradation yielded the sequence Gly-Gly-His-Asp-Val-Pro-Leu-Thr-Asn-Ile, identical to that predicted by the DNA se-quence, and indicating once again that the initiator Met 1 residue had been removed by E. coli proteinases. Consistent with this, analysis by MALDI-TOF mass spectrometry gave a mass of 8502 Da, identical to that (8502 Da) predicted for the sequence of residues 2-68 plus the ϳLeu-Glu-His 6 tag of the chimeric protein. Remarkably, this chimeric protein with 9/34 (26%) of its residues exchanged was still almost as effective as an inhibitor as the wild-type protein, with its interaction at pH 4.7 still being so tight as to lie at or beyond the limits of accurate determination (Table I). This result, together with the deletion experiments described above with the peptide form of IA 3 , suggests that, for effective inhibition, backbone atoms contributed by residues 2-10 are essential but that the (side chain) identity of the individual amino acids in these positions is of lesser importance. On this basis, we replaced residues 2-10 of the natural sequence with nine glycine residues ((Gly) 9 mutant, Table I) and purified the resultant protein to homogeneity. No attempts were made to sequence this protein because of the plethora of glycine residues but MALDI-TOF mass spectrometry gave a mass of 8124 Da, identical to that (8124 Da) predicted by the nucleotide sequence but, once again, lacking the initiator Met 1 residue. The yield of this mutant protein obtained from E. coli was about 5-fold lower than that obtained for the wild-type (and other mutant) sequence(s). This drastic introduction of nine consecutive glycine residues resulted in a poorer inhibitor with a K i of 40 nM at pH 3.1 (Table I). However, at pH 4.7, the (Gly) 9 mutant protein was still a very effective inhibitor, with its potency quantified at around 1 nM (Table I). Thus, the main chain atoms at the N terminus of the 2-34 sequence would appear to be the major contributors to inhibitory potency from this region with only minor influences being introduced by the individual residue side chains. This was substantiated further by replacement of individual residues, for example, the Lys 7 residue was replaced with methionine which is quasi-isosteric with lysine but lacks the ⑀-NH 2 group. This K7M mutant inhibitor was just as potent as the wild-type protein ( Table I).
Truncation of the inhibitory sequence of residues 2-34 at its C terminus also resulted in a progressive loss of inhibitory potency (compare peptide 1 and peptides 9, 10, 11, 12, and 13, Table II). Replacement of Lys 24 by Met (K24M mutant) and Lys 31 ϩ Lys 32 (together) in the double mutant K31M/K32M again had no significant effect on the inhibitory potency of the resultant inhibitors toward proteinase A (Table I). The structure of the K24M mutant complexed with proteinase A was solved at 2-Å resolution and refined to an R factor of 19.84% (see "Experimental Procedures"). Comparison with the structures reported previously (12) for the K31M/K32M protein form and for the peptide 1 form of IA 3 complexed with proteinase A (Protein Data Bank accession codes 1dp5 and 1dpj) revealed that, in all three cases, residues 2-32 of the inhibitor had adopted a near-perfect ␣Ϫhelical conformation in the active FIG. 1. The helical conformation adopted by residues 2-32 of IA 3 upon interaction with the active site of proteinase A. The sequence shown is that of the K31M/K32M mutant, depicting the distribution of selected hydrophilic and hydrophobic residues on opposite faces of the amphipathic helix. site cleft of the enzyme. Electron density was only observed for these residues in all 3 structures and the root mean square deviation between the C␣ coordinates was 0.224 between the protein and peptide form(s) of the inhibitor. The IA 3 helix is amphipathic with the charged residues including Lys 24 located on one face, protruding into solvent (Fig. 1).
The structures all reveal that the main chain carbonyl and amido moieties of residues 2-10 of the IA 3 polypeptide are involved in H-bond formation with one another within the helix of the inhibitor but the side chains of these residues make no significant contacts with the proteinase, with the exception of Val 8 which is involved in hydrophobic interactions (described later). In the sequence of the chimera inhibitor described earlier (Table I), Val 8 was replaced with Leu which may be able to make hydrophobic contacts with the enzyme requiring only minor re-adjustment so that the derived K i value was not significantly perturbed. The binding energy of these contacts is clearly lost in the (Gly) 9 mutant protein but since the side chains of all of the other residues in the 2-10 sequence point largely into solvent, the still considerable inhibitory potency of the (Gly) 9 mutant (Table I) can be readily understood. The predominant requirement in this N-terminal region appears to be for the backbone atoms to satisfy the H-bonding arrangement within the inhibitory helix, thereby stabilizing the helix by providing a somewhat lengthy "cap." A comparable extended "capping" arrangement exists at the C-terminal end to stabilize the inhibitory helix. No electron density was observed for the side chains of any residues beyond Lys 32 in the crystal structure of the K24M mutant complex or in the structure of peptide 1 complexed with proteinase A described previously (12) and the side chain of residue 32 makes no significant contacts with the enzyme. Yet the most potent inhibition (subnanomolar at pH 4.7) was measured when the IA 3 sequence was extended at its C terminus beyond Lys 32 . This was achieved by the inclusion of Nle 33 (a synthetic isostere of the natural Met residue at this location) and Ala 34 in peptide 1 (Table II) or by introducing a C-terminal Lys 32 amide (peptide 9-NH 2 , Table II) to stabilize the part of the inhibitor helix necessary for interaction with the proteinase by dissipation of the negative charge from the macromolecular dipole. The contribution of the main chain amido groups in satisfying the H-bonding arrangements at this end of the helix is substantiated by the increased potencies that were measured when each peptide terminated in a C-terminal amide instead of the free COOH group (compare peptides 9-NH 2 , 10-NH 2 and 11-NH 2 with 9, 10, and 11, respectively, Table I). A peptide that consisted only of residues 2-28 had lost essentially all of its inhibitory potency (peptide 12, Table II) and residues 2-26 (peptide 13) were not active at all as an inhibitor.
Hydrophobic Clusters-Peptide 14 and peptide 8 (Table II) together span the entire sequence of residues 2-34. Neither of these "half-sized" peptides was able to inhibit proteinase A when added singly or in combination with each other at a variety of molar ratios. Thus a contiguous sequence is necessary for IA 3 to inhibit proteinase A by forming the amphipathic helix (Fig. 1). The residues located on the hydrophobic face make extensive hydrophobic contacts with cognate residues in proteinase A. Particularly noticeable in this regard (Fig. 1) is the "cluster" arrangement of Val 8 -X-X-hydrophobic-Phe 12 toward the front end of the helix and Val 26 -X-X-hydrophobic-Phe 30 at the back end of the inhibitory sequence. In the front end cluster, replacement of Val 8 or Ile 11 individually by Ala had minor effects on inhibitory potency (compare peptides 15 and 16 with peptide 9, Table IV). However, deletion of the benzene ring of phenylalanine at position 12 resulted in a considerably larger drop in potency in the resultant Ala-containing peptide (peptide 17, Table IV), emphasizing the importance of van der Waals interactions of this benzene ring with its hydrophobic environment (Fig. 2). The peptide carrying the double replacement of V8A together with F12A (peptide 18, Table IV) had lost much of its inhibitory potency and the double mutant peptide containing ϳAla 11 -Ala 12 ϳ (peptide 19) was virtually ineffective as an inhibitor, even at pH 4.7. The triple mutant in which all three of the front-end cluster residues were changed to ϳAla 8 -X-X-Ala 11 -Ala 12 ϳ (peptide 20, Table IV) was completely inactive as an inhibitor.
A similar response was quantified when the residues contributing to the ϳVal 26 -X-X-hydrophobic-Phe 30 ϳ cluster at the back end of the helix were replaced. Substitution of Phe 30 by Ala, Gly, or Lys (compare peptides 21 and 22/23 with peptide 10, Table IV) resulted in a significant loss in potency, again emphasizing the contribution to binding by appropriate positioning of the large benzene ring of the side chain of Phe 30 . In contrast, replacement of Val 26 by Ala did not diminish inhibitory potency. Rather, it appeared to improve the binding interaction at pH 3.1 marginally (peptide 24,  side chain into a hydrophobic environment. However, the K i value measured at pH 3.1 for peptide 25 was tighter than that derived at pH 4.7 (Table IV). Of all the inhibitors listed in Tables I, II, and IV, this was the only occasion when such an effect was observed and most likely is a reflection of the Asp side chain in its protonated and therefore uncharged form being less unfavorable in its contact with the hydrophobic environment offered by the enzyme. The double mutant peptide carrying the V26D/F30K substitutions was completely ineffective as an inhibitor (peptide 26, Table IV), indicating that introduction of two hydrophilic, charged residues was highly unfavorable since there are no H-bond partners available in the enzyme to compensate for desolvation of the two side chain functions. However, amphipathic helices are often stabilized by electrostatic interactions between residues at positions i and i ϩ 4 (23) and, indeed exactly such a salt bridge is present between the Lys 24 and Asp 28 residues on the hydrophilic face of the IA 3 helix when complexed with its target proteinase (Fig. 1). However, the attempt to encourage Asp 26 and Lys 30 to interact with one another to form an additional salt bridge in the V26D/F30K double mutant peptide was clearly not tolerated on the hydrophobic face of the amphipathic helix in the active site cleft of the enzyme. A further mutant was also constructed in which the sequence of ϳSer 27 -Asp 28 -Ala 29 -Phe 30 -Lys 31 -Lys 32 ϳ was shuffled to ϳLys 27 -Ala 28 -Asp 29 -Lys 30 -Phe 31 -Ser 32 ϳ in the protein form of the IA3 inhibitor (Mix in Table I). In this arrange-ment, Val 26 was retained but the salt bridge between Lys 24 and Asp 28 on the hydrophilic face of the helix was disrupted and the crucial Phe 30 residue was replaced by lysine, as in peptide 23. The resultant mutant protein (Mix, in Table I) was purified to homogeneity from E. coli and found to have a K i value at pH 4.7 comparable to that observed for the single F30K mutant peptide (peptide 23, Table IV). This might be interpreted to indicate that the salt bridge interaction between Lys 24 and Asp 28 on the hydrophilic surface of the IA 3 helix is, not unexpectedly, weak.
Central Residues in the Inhibitor Helix-The "centerpiece" of the inhibitory 2-34 residues of IA 3 is the ϳLys 18 -Leu 19 -X-X-Asp 22 ϳ sequence. In the three crystal structures of the K24M and K31M/K32M mutant proteins and peptide 1 forms of IA3 complexed with proteinase A, the ⑀-NH 2 group of Lys 18 of the inhibitor hydrogen bonds to one of the carboxyl oxygens of Asp 32 of the enzyme (Fig. 3). This is one of the two catalytic Asp residues that operate the catalytic mechanism of all aspartic proteinases (24). The ⑀-NH 2 group of Lys 18 also hydrogen bonds to one of the carboxyl oxygens of the side chain of Asp 22 in the IA 3 inhibitory sequence (Fig. 3). The other oxygen of the side chain COOH of Asp 22 hydrogen bonds to the phenolic OH group of Tyr 75 in the enzyme, a residue that is totally conserved in all eukaryotic aspartic proteinases and which is positioned almost at the tip of the ␤-hairpin loop or flap that overlays the active site cleft in these enzymes. A network of interactions thus cross-links these charged residues of IA 3 with the catalytically