Ionizable P 1 Residues in Serine Proteinase Inhibitors Undergo Large p K Shifts on Complex Formation*

The burial of charged residues in proteins is rare as it is thermodynamically strongly disfavored. However, in “standard mechanism” protein inhibitors of serine pro- teinases, the P 1 residue, which is highly exposed, becomes buried in the S 1 specificity pocket of the enzyme. In many enzymes, such as Streptomyces griseus proteinase B (SGPB) the S 1 pocket is hydrophobic. We meas- ured the pH dependence of the association equilibrium constant for the interaction of SGPB with turkey ovo- mucoid third domain P 1 mutants, Glu 18 OMTKY3 and His 18 OMTKY3. In order to eliminate the effects of other ionizable groups on the enzyme and the inhibitor, we divided these pH dependences by the pH dependence of the association equilibrium constant for the Gln 18 OMTKY3 mutant. This yielded for Glu 18 , p K f (free inhib- itor) of 4.46 (cid:54) 0.05 and p K c (complex) of 8.74 (cid:54) 0.06. For His 18 the values are p K f 6.63 (cid:54) 0.08 and p K c 4.31 (cid:54) 0.07. At low pH values Glu 18 variant is a relatively good inhibitor for SGPB. This may be biologically relevant. Transferring a charge from a high dielectric medium, such as water, to a low dielectric medium, such as organic solvents or interiors of proteins, is not thermodynamically Therefore, ionizable side chains buried in proteins should ex-perience large p K shifts. Expression of Third Domain Variants— The ovomucoid third domain variants are routinely expressed in our laboratory in the periplasmic space of Escherichia coli as fusion proteins with two Z domains of protein A. The fusion protein is purified by affinity chromatography on IgG-Sepharose (Pharmacia Biotech Inc.), which specifically binds the protein A domains. The third domain proper (Fig. 1, residues 6–56) is released by CNBr cleavage (there is only one Met residue in the fusion protein). The released domain is purified by size exclusion and ion exchange chromatography. It is extensively characterized by amino acid analysis, high resolution mass spectrometry, sequencing beyond the point of replacement (the P 1 residue), and analytical ion exchange chromatography. Determination of Association Equilibrium Constant— Association equilibrium constants of third domain variants with SGPB were determined by the procedure of Green and Work (22) extensively modified in this laboratory (21, 23–25). All measurements were performed at 22 °C in buffers of different pH values containing 0.05 M CaCl 2 and 0.005% Triton X-100. The buffers used were: Gly-HCl (pH 2.5–3.5); NaOAc-HOAc (pH 4–5.5); bis-tris (pH 6–7); Tris-HCl (pH 7.5–9); and Gly-HCl or CAPS (pH 9.5–10.1). The procedure consisted of mixing the enzyme with different amounts of inhibitor for an appropriate length of time (depending upon K a ). The residual enzyme activity was measured on an automated HP 8450 diode array spectrophotometer using peptide sub- strates with a p -nitroanilide leaving group. The maximal errors in K a are (cid:54) 20%.

The burial of charged residues in proteins is rare as it is thermodynamically strongly disfavored. However, in "standard mechanism" protein inhibitors of serine proteinases, the P 1 residue, which is highly exposed, becomes buried in the S 1 specificity pocket of the enzyme. In many enzymes, such as Streptomyces griseus proteinase B (SGPB) the S 1 pocket is hydrophobic. We measured the pH dependence of the association equilibrium constant for the interaction of SGPB with turkey ovomucoid third domain P 1 mutants, Glu 18 OMTKY3 and His 18 OMTKY3. In order to eliminate the effects of other ionizable groups on the enzyme and the inhibitor, we divided these pH dependences by the pH dependence of the association equilibrium constant for the Gln 18 OMTKY3 mutant. This yielded for Glu 18 , pK f (free inhibitor) of 4.46 ؎ 0.05 and pK c (complex) of 8.74 ؎ 0.06. For His 18 the values are pK f 6.63 ؎ 0.08 and pK c 4.31 ؎ 0.07. At low pH values Glu 18 variant is a relatively good inhibitor for SGPB. This may be biologically relevant.
Transferring a charge from a high dielectric medium, such as water, to a low dielectric medium, such as organic solvents or interiors of proteins, is not thermodynamically favored (1). Therefore, ionizable side chains buried in proteins should experience large pK shifts. Such shifts are occasionally seen (2)(3)(4)(5), but they are rare. While ionizable residues are fairly often buried in their uncharged form, acquisition of charge is most often associated with a conformational change (3,4) or with dissociation of the ligand-receptor complex. We describe two cases where the complexes persist with the charged form of a buried side chain. Many more such cases can be generated by examining the comparative pH dependence of the association of ionizable P 1 mutants of protein inhibitors of serine proteinases.
The association of "standard mechanism" (6) canonical (7) protein inhibitors of serine proteinases is the most studied and best understood of protein-protein interactions. In these inhibitors, each inhibitory domain has a single reactive site peptide bond, which connects residues designated as P 1 and P 1 Ј (8). This bond serves as substrate for the cognate proteinase. In all free inhibitors whose structure has been determined, the P 1 residue is highly exposed to solvent (7,9). In complexes, the reactive site peptide bond remains intact and planar, but the carbonyl carbon atom of the P 1 residue is in exceptionally close contact with the OG atom of the catalytic Ser residue of the enzyme (9). Most importantly, the P 1 side chain becomes imbedded in the S 1 pocket (7,9) of the enzyme. This pocket is often called the primary specificity pocket of serine proteinases. While the interaction of P 1 with the enzyme is energetically the most important, about a dozen of the inhibitors' residues interact both by main chain-main chain hydrogen bonds and by specific side chain-side chain interactions. Collectively, the interactions of all the contact residues, other than P 1 , are energetically dominant. They ensure that the same peptide bond in the inhibitory domain acts as the reactive site for numerous different cognate enzymes (10). More importantly, they ensure that, in inhibitor mutants, substitutions of the P 1 residue do not shift the reactive site peptide bond and that the mutant P 1 side chain is still imbedded in the S 1 cavity of the enzyme, even if that interaction alone is energetically adverse rather than favorable 1 (11).
Turkey ovomucoid third domain, OMTKY3, 2 ( Fig. 1) is a widely studied "standard mechanism," canonical inhibitor of serine proteinases. With its P 1 Leu residue, it is a powerful inhibitor of many chymotrypsins, elastases, subtilisins, and two of the many Pronase components, Streptomyces griseus proteinase A (SGPA) and proteinase B (SGPB) (10). All of these enzymes have predominantly hydrophobic S 1 pockets and prefer substrates and inhibitors with hydrophobic P 1 residues. We have recently acquired 25 different P 1 variants (all 20 coded and 5 noncoded) of X 18 OMTKY3 and measured their association constants, K a , at pH 8.3 with six enzymes, all of which are members of the set listed above (18,19). In a parallel study, Huang and James 1 undertook to obtain high resolution threedimensional structures of all 20 coded X 18 variants of OMTKY3 complexed with SGPB. Many of these structures, which crystallized near pH 6.5, especially Glu 18 OMTKY3, are already in hand. 1 The binding of Glu 18 to the S 1 specificity pocket of SGPB is similar to the binding of Leu 18 (11,14,15). The interpretation of the K a (at pH 8.3) and the x-ray data (at pH 6.5) requires knowing whether the Glu 18 residue is protonated (Glu o ) or deprotonated (Glu Ϫ ) in complex at these values. When appraised of our results that the pH 6.5 structure was of the protonated form, Glu o , Huang and James 1 raised the pH of their crystals to 10.7. The global structure of the complex was unaltered, but clear evidence was obtained for Glu Ϫ rather than Glu 0 .

EXPERIMENTAL PROCEDURES
Proteinase-S. griseus proteinase B, SGPB, was purified from Pronase (Boehringer Mannheim) by the procedure of Jurasek et al. (20) modified in this laboratory (21). The amino acid composition of the isolated enzyme is consistent with the published sequence.

Expression of Third Domain Variants-
The ovomucoid third domain variants are routinely expressed in our laboratory in the periplasmic space of Escherichia coli as fusion proteins with two Z domains of protein A. The fusion protein is purified by affinity chromatography on IgG-Sepharose (Pharmacia Biotech Inc.), which specifically binds the protein A domains. The third domain proper (Fig. 1, residues 6 -56) is released by CNBr cleavage (there is only one Met residue in the fusion protein). The released domain is purified by size exclusion and ion exchange chromatography. It is extensively characterized by amino acid analysis, high resolution mass spectrometry, sequencing beyond the point of replacement (the P 1 residue), and analytical ion exchange chromatography.
Determination of Association Equilibrium Constant-Association equilibrium constants of third domain variants with SGPB were determined by the procedure of Green and Work (22) extensively modified in this laboratory (21,(23)(24)(25). All measurements were performed at 22°C in buffers of different pH values containing 0.05 M CaCl 2 and 0.005% Triton X-100. The buffers used were: Gly-HCl (pH 2.5-3.5); NaOAc-HOAc (pH 4 -5.5); bis-tris (pH 6 -7); Tris-HCl (pH 7.5-9); and Gly-HCl or CAPS (pH 9.5-10.1). The procedure consisted of mixing the enzyme with different amounts of inhibitor for an appropriate length of time (depending upon K a ). The residual enzyme activity was measured on an automated HP 8450 diode array spectrophotometer using peptide substrates with a p-nitroanilide leaving group. The maximal errors in K a are Ϯ20%.

RESULTS AND DISCUSSION
The pH dependence of log K a for three ovomucoid third domain variants, namely ionizable Glu 18 OMTKY3 and His 18 OMTKY3 and a nonionizable Gln 18 OMTKY3, as a reference, is shown in Fig. 2. The behavior of log K a for the last case is extremely simple. It can be rationalized by the pK of His 57 in the catalytic triad of SGPB being 6.8 in the free enzyme and shifting to a very low value in the enzyme inhibitor complex (26,27). It is worth noting that while the assignment of the pH dependence of Gln 18 OMTKY3-SGPB association is likely correct and interesting, it is not essential to this paper. All that is required is that the division of K a (Glu 18 ) and (His 18 ) by K a (Gln 18 ) eliminates the effect of all residues other than the P 1 residue. For Glu 18 and His 18 variants, the pH dependence is more complex, involving a combination of the effects seen in the Gln 18 variant and the pK perturbation of the P 1 residue by complex formation. To separate the effects, consider, where K a is at arbitrary pH, K a 0 is at very low pH where all the groups in the complex, inhibitor, and enzyme are fully protonated. Q c , Q I , and Q E are the protonation state partition functions, often called binding polynomials (28), but, in this case, probably better referred to as proton release polynomials (29 -31). Equation 1 is a general equation for the pH dependence. However, in the free inhibitor, the P 1 residue is fully exposed and does not interact with other residues (6,7,9). Therefore, for Glu 18 OMTKY3 (or His 18 OMTKY3) variants we can write, where K ion,f is the ionization constant of P 1 in the free inhibitor and Q I * is the proton release polynomial for all the other ionizable residues of the inhibitor. In this approximation, Q I * is obtained from a nonionizable P 1 variant such as Gln 18 OMTKY3. Similarly, since the P 1 side chain in the complex with SGPB is insulated in a hydrophobic S 1 cavity 1 (11), it seems plausible to write, where K ion,c is the ionization constant of the P 1 residue in the complex. Simple division now yields as follows.
The derivation of this relation is based on a simple assumption that the ionizable P 1 residue is not interacting with any other ionizable group in the free inhibitor or in complex. This assumption is supported by the published x-ray structure of ovomucoid third domains (7) and by the unpublished structure of Glu 18 OMTKY3 in complex with SGPB. 1 For plotting and fitting of the data the logarithmic form of this equation has been used, log R ϭ log R o ϩ log͑1 ϩ 10 ͑pHϪpKc͒ ͒ Ϫ log͑1 ϩ 10 ͑pHϪpKf͒ ͒ (Eq. 5) where R o is the equilibrium constant ratio of the fully protonated to the nonionizable residue and pK f and pK c are the pK values in the free inhibitor and in the complex with SGPB. The plots of log R versus pH are shown in Fig. 3. The excellent leveling off for the His 18 and Glu 18 curves, both on the low and high pH side, clearly indicates that, around pH 3, the P 1 side FIG. 1. Amino acid sequence of OMTKY3 (12). For historical reasons the numbering starts from residue 6 rather than 1. The 12 consensus contact residues (13), which are in contact with SGPB (14,15), chymotrypsin (16), and human leukocyte elastase (17), are highlighted in black. The reactive site peptide bond between the P 1 residue Leu 18 and P 1 Ј residue is indicated by an arrow. It is the Leu 18 residue that was replaced in the Glu 18 , His 18 , and Gln 18 variants. Large pK Shifts on Burial of Residues 27420 chain in complex and in the free inhibitor is predominantly in the His ϩ and Glu o forms and, near pH 10, it is predominantly in His o and Glu Ϫ forms, both in the free inhibitor and in complex. The plots in Fig. 3 are fitted by nonlinear least squaring (32) using pK f , pK c , and R o as fit parameters. For Glu 18 pK f is 4.46 Ϯ 0.05, pK c is 8.74 Ϯ 0.06, and R o is 13.9 Ϯ 1.3. The large 4.3-unit pK shift shows that (a) both in the crystal structures and in measurements at pH 8.3, the side chain of Glu 18 is predominantly in the Glu o form, (b) binding of Glu o is 16,000 times stronger than binding of Glu Ϫ , and (c) binding of Glu o is 14 times better than binding of Gln. A simple illustration of the pK shift of Glu in Glu 18 variant upon complex formation with SGPB and the associated changes in K a are shown diagrammatically in Fig. 4. For His 18 , pK f is 6.63 Ϯ 0.08, pK c is 4.31 Ϯ 0.07, and R o is 0.027 Ϯ 0.003. The downward shift of the His 18 pK is consistent with avoidance of a charged form in the S 1 cavity, albeit now only by a factor of 160. The binding of His 18 , in its neutral form, is 4.5 times better than that of Gln 18 ; this was already known from the pH 8.3 measurements (18,19). Note the striking and unforced agreement between pK f of both side chains with the expected or model values. The excellent fit of the curves to Equation 5 and the clear evidence that measurable association still takes place at the low pH limit of Fig. 3 for the His 18 variant, and at the high pH limit for the Glu 18 variant, argue for the validity of the model. The correctness of this treatment is also strongly validated by the ratio of two nonionizable side chains, i.e. (K a (Leu 18 ))/(K a (Gln 18 )) being independent of pH in the pH range 4 -8.3 (Fig. 3).
The values of K a for variants with ionizable side chains that were reported (18,19) at pH 8.3 refer to the equilibrium mixtures of the ionized and unionized forms of the P 1 side chain both in the free inhibitor and in the complex. In the specific case of Glu 18 variant interacting with SGPB, the free side chain is overwhelmingly Glu Ϫ , while the bound side chain is largely Glu o . It is of interest to calculate the K a values for Glu o in the free inhibitor binding to form a complex with Glu o and for Glu Ϫ in the free inhibitor forming a complex with Glu Ϫ . The constants are K a (Glu o ) 18 Ϸ 1 ϫ 10 10 M Ϫ1 and K a (Glu Ϫ ) 18 Ϸ 6 ϫ 10 5 M Ϫ1 (see Fig. 4). The K a (Glu o ) is the fourth strongest of K a values (after Leu 18 (5.6 ϫ 10 10 M Ϫ1 ), Met 18 (2.7 ϫ 10 10 M Ϫ1 ), and (Asp o ) 18 (2.0 ϫ 10 10 )) of all 20 coded amino acid residues. At low pH (pH 4.5 or lower), the Glu 18 variant is one of the most effective inhibitors of SGPB. It is possible that the frequently seen P 1 Glu residues in natural protein inhibitors of serine proteinases serve not only to inhibit glutamic and aspartic acid-specific enzymes such as Glu-specific S. griseus proteinase (33) but also at low pH values to inhibit the widely distributed enzymes with hydrophobic S 1 pockets.
While the method is fairly labor intensive, we believe that it could be readily extended to more, possibly all, ionizable side chains (Asp and Lys seem particularly good) and to more enzymes with hydrophobic S 1 pockets, e.g. SGPA and chymotrypsin. Extension to enzymes, such as trypsin or Glu-specific S. griseus proteinase, with ionizable residues in their S 1 pockets may require more complex interpretation.
Ionizable side chains in proteins in their protonated form are either uncharged (Asp, Glu, Cys, Tyr) or cationic (His, Lys, Arg) (34). Ionization of uncharged acids yields charged products. Therefore the large upward pK shift, seen upon addition of organic cosolvents to such acids, is expected (35)(36)(37). On the other hand, ionization of cationic acids involves a transfer of charge from the cationic reactant to the hydronium ion, which is one of the products. Therefore, cosolvent effects on pK are anticipated to be, and are, small and erratic (35)(36)(37). Burial in a protein differs from transfer to an organic solvent, as the hydronium ion resides in an aqueous solution; only the side chain is buried. As a result, opposite effects, about half as great in absolute value as those observed in the transfer of an uncharged acid to an "equivalent" organic solvent, should be expected if the S 1 pockets were purely isotropic. Clearly, they are not, and dipole and charge effects may further affect the observed values.
Since this communication was first submitted, we have determined the pH dependence of the interaction between SGPB and Asp 18 OMTKY3 and found that pK f is 4.40 Ϯ 0.10 and pK c is 9.26 Ϯ 0.10. Therefore, the shift is 4.9 units, somewhat greater than for Glu 18 . Recently (38) it was shown that Asp 26 in thioredoxin has a pK of 7.5 in the oxidized form and greater than 9 in the reduced form, another huge pK shift for an Asp residue. Huang and James 1 completed their high resolution structures of Asn 18 OMTKY3 at pH 6.5 and of Asp 18 , Glu 18 , and Gln 18 variants at pH 6.5 and 10.7. An extended joint manuscript is now readied for publication. Large pK Shifts on Burial of Residues 27421