NMR line-broadening and transferred NOESY as a medicinal chemistry tool for studying inhibitors of the hepatitis C virus NS3 protease domain.

This work describes the use of NMR as a medicinal chemistry tool for better understanding the binding characteristics of inhibitors of the HCV NS3 protease. The protease-bound structure of a tetrapeptide-like inhibitor that has an acid C-terminus, a norvaline at P1 and a naphthylmethoxy proline at P2 is described. Conformational comparisons are made with a similar compound having a 1-amino-cyclopropylcarboxylic acid at P1 and with a hexapeptide inhibitor. Differences between the free and bound states are identified. 19F NMR also helped in determining that a single complex is observed when an inhibitor is added to the protease at a 1:1 ratio.

Human cytomegalovirus protease complexes its substrate recognition sequences in an extended peptide conformation.

Substrate hydrolysis by human cytomegalovirus (HCMV) protease is essential to viral capsid assembly. The interaction of HCMV protease and the N-terminal cleavage products of the hydrolysis of R- and M-site oligopeptide substrate mimics (R and M, respectively, which span the P9-P1 positions) was studied by NMR methods. Protease-induced differential line broadening indicated that ligand binding is mediated by the P4-P1 amino acid residues of the peptides. A well-defined extended conformation of R from P1 through P4 when complexed to HCMV protease was evidenced by numerous transferred nuclear Overhauser effect (NOE) correlations for the peptide upon addition of the enzyme. NOE cross-peaks between the P4 and P5 side chains placing these two groups in proximity indicated a deviation from the extended conformation starting at P5. Similar studies carried out for the M peptide also indicated an extended peptide structure very similar to that of R, although the conformation of the P5 glycine could not be established. No obvious variation in structure between bound R and M (notably at P4, where the tyrosine of the R-site has been suggested to play a key role in ligand binding) could be discerned that might explain the observed differences in processing rates between R- and M-sequences. Kinetic studies, utilizing R- and M-site peptide substrates for which the P5 and P4 residues were separately exchanged, revealed that these positions had essentially no influence on the specificity constants (kcat/KM). In sharp contrast, substitution of the P2 residue of an M-site peptide changed its specificity constant to that of an R-site peptide substrate, and vice versa.

Autocatalytic processing of recombinant human procathepsin L. Contribution of both intermolecular and unimolecular events in the processing of procathepsin L in vitro.

The autocatalytic processing of procathepsin L was investigated in vitro using purified recombinant proenzyme expressed in Pichia pastoris. Pure intermolecular processing was studied by incubating the mutant procathepsin L (C25S), which cannot autoactivate with a small amount of mature active cathepsin L. The results clearly establish that, contrary to recent reports, intermolecular processing of procathepsin L is possible. The main cleavage sites are located at or near the N terminus of the mature enzyme, in an accessible portion of the proregion, which contains sequences corresponding to the known substrate specificity of cathepsin L. Contrary to procathepsins B, K, and S, autocatalytic processing of procathepsin L can generate the natural mature form of the enzyme. A continuous assay using the substrate benzyloxycarbonyl-Phe-Arg 4-methylcoumarinyl-7-amide hydrochloride has also been used to obtain information on the nature of the steps involved in the autocatalytic processing of wild-type procathepsin L. Processing is initiated by decreasing the pH from 8.0 to 5.3. The influence of proenzyme concentration on the rate of processing indicates the existence of both unimolecular and bimolecular steps in the mechanism of processing. The nature of the unimolecular event that triggers processing remains elusive. Circular dichroism and fluorescence measurements indicate the absence of large scale conformational change in the structure of procathepsin L on reduction of pH. However, the bimolecular reaction can be attributed to intermolecular processing of the zymogen.

Design of fluorogenic peptide substrates for human cytomegalovirus protease based on structure-activity relationship studies.

Human cytomegalovirus (HCMV) protease is a slow-processing enzyme in vitro and its characterization would be facilitated if more efficiently cleaved substrates were available. Here we describe the development of improved fluorogenic peptide substrates for this protease and demonstrate that its indolent nature can be overcome by appropriate modifications within existing substrates. Prior structure-activity studies have indicated that replacement of the Val-Val-Asn sequence corresponding to the P4-P2 residues of the maturation site of the enzyme by the optimized Tbg-Tbg-Asn(NMe2) sequence conferred significant binding to inhibitors (Tbg, t-butylglycine). Incorporation of this improved sequence in a variety of substrates invariably led to improved kinetic parameters compared to homologues containing the natural sequence only. For example, the substrate o-aminobenzoyl-Tbg-Tbg-Asn (NMe2)-Ala decreases Ser-Ser-Arg-Leu-Tyr(3-NO2)Arg-OH (2) displayed a kcat/K(m) value of 15,940 M-1 s-1 i.e., more than 60-fold greater than that of the equivalent, nonoptimized substrate 1 under identical conditions. This improved sequence also permitted the development of a sensitive 7-amino-4-methylcoumarin fluorogenic substrate 3 which represents the shortest HCMV protease substrate to date. The kinetic and photometric advantages of these various substrates are discussed along with specific applications.

Peptidomimetic inhibitors of the human cytomegalovirus protease.

The development of peptidomimetic inhibitors of the human cytomegalovirus (HCMV) protease showing sub-micromolar potency in an enzymatic assay is described. Selective substitution of the amino acid residues of these inhibitors led to the identification of tripeptide inhibitors showing improvements in inhibitor potency of 27-fold relative to inhibitor 39 based upon the natural tetrapeptide sequence. Small side chains at P1 were well tolerated by this enzyme, a fact consistent with previous observations. The S2 binding pocket of HCMV protease was very permissive, tolerating lipophilic and basic residues. The substitutions tried at P3 indicated that a small increase in inhibitor potency could be realized by the substitution of a tert-leucine residue for valine. Substitutions of the N-terminal capping group did not significantly affect inhibitor potency. Pentafluoroethyl ketones, alpha,alpha-difluoro-beta-keto amides, phosphonates and alpha-keto amides were all effective substitutions for the activated carbonyl component and gave inhibitors which were selective for HCMV protease. A slight increase in potency was observed by lengthening the P1' residue of the alpha-keto amide series of inhibitors. This position also tolerated a variety of groups making this a potential site for future modifications which could modulate the physicochemical properties of these molecules.

Delineating functionally important regions and residues in the cathepsin B propeptide for inhibitory activity.

Synthetic peptides derived from the proregion of rat cathepsin B were used to identify functionally important regions and residues for cathepsin B inhibition. Successive 5 amino acid deletions of a 56 amino acid propeptide from both the N- and C-termini has allowed the identification of two regions important for inhibitory activity: the NTTWQ (residues 21p-25p) and CGTVL (42p-46p) regions. Alanine scanning of residues within these two regions indicates that Trp-24p and Cys-42p contribute strongly to inhibition, their replacement by Ala resulting in 160- and 140-fold increases in Ki, respectively.

Potency and selectivity of the cathepsin L propeptide as an inhibitor of cysteine proteases.

The cathepsin L propeptide (phcl-2) was expressed in Saccharomyces cerevisiae using a human procathepsin L/alpha-factor fusion construct containing a stop codon at position -1 (the C-terminal amino acid of the proregion). Since the yield after purification was very low, the cathepsin L propeptide was also obtained by an alternate procedure through controlled processing of an inactive mutant of procathepsin L (Cys25Ser/Thrl10Ala) expressed in Pichia pastoris, by small amounts of cathepsin L. The peptide resulting from the cleavage of the proenzyme (phcl-1) was then purified by HPLC. The purified propeptides were characterized by N-terminal sequencing and mass spectrometry and correspond to incomplete forms of the proregion (87 and 81 aa for phcl-1 and phcl-2 respectively, compared to 96 aa for the complete cathepsin L propeptide). The two peptides were found to be potent and selective inhibitors of cathepsin L at pH 5.5, with Ki values of 0.088 nM for phcl-1 and 0.66 nM for phcl-2. The Ki for inhibition of cathepsin S was much higher (44.6 nM with phcl-1), and no inhibition of cathepsin B or papain could be detected at up to 1 microM of the propeptide. The inhibitory activity was also found to be strongly pH-dependent. Two synthetic peptides of 75 and 44 aa corresponding to N-terminal truncated versions of the propeptide were also prepared by solid phase synthesis and displayed Ki values of 11 nM and 2900 nM, respectively, against cathepsin L. The data obtained for the 4 propeptide derivatives of various lengths indicate that the first 20 residues in the N-terminal region of the propeptide are more important for inhibition than the C-terminal region which contributes little to the overall inhibitory activity.

Aziridine analogs of [[trans-(epoxysuccinyl)-L-leucyl]amino]-4-guanidinobutane (E-64) as inhibitors of cysteine proteases.

Aziridine derivatives of E-64 have been synthesized, and their characterization against the cysteine proteases cathepsin B, cathepsin L, and papain is reported. The inhibition was found to be strongly pH-dependent, with maximum activity observed at pH 4, indicating that the protonated aziridinium ion form of the inhibitor is the more reactive form. At low pH, the peptide aziridine HO-(L)Az-Leu-NH-iAm inactivated papain with a second-order rate constant, kinac/Ki, of 7.0 x 10(4) M-1 s-1, a value very close to that observed with E-64 or with the corresponding epoxysuccinyl analog HO-(L)Eps-Leu-NH-iAm. This demonstrates that with the correct peptide sequence, aziridine analogs of E-64 can be good irreversible inhibitors of cysteine proteases. Substitution of the epoxysuccinyl moiety by an aziridine does not affect the specificity of inhibition against the three proteases used in this study. The D-diastereomer is the preferred (by 10-fold) diastereomer for the inhibition of cysteine proteases. The reactivity of both diastereomers of iBuNH-Az-LeuPro-OH against cathepsin B was also found to be much lower than that of iBuNH-(L)Eps-LeuPro-OH, which is a potent selective inhibitor of cathepsin B. These differences are attributed mainly to the presence of the protonated aziridine ring, which can modify the binding mode of aziridine analogs at the active site of cysteine proteases.

Structural and functional roles of asparagine 175 in the cysteine protease papain.

The role of the asparagine residue in the Cys-His-Asn "catalytic triad" of cysteine proteases has been investigated by replacing Asn175 in papain by alanine and glutamine using site-directed mutagenesis. The mutants were expressed in yeast and kinetic parameters determined against the substrate carbobenzoxy-L-phenylalanyl-(7-amino-4-methylcoumarinyl)- L-arginine. At the optimal pH of 6.5, the specificity constant (k(cat)/KM)obs was reduced by factors of 3.4 and 150 for the Asn175-->Gln and Asn175-->Ala mutants, respectively. Most of this effect was the result of a decrease in k(cat), as neither mutation significantly affected KM. Substrate hydrolysis by these mutants is still much faster than the non-catalytic rate, and therefore Asn175 cannot be considered as an essential catalytic residue in the cysteine protease papain. Detailed analyses of the pH activity profiles for both mutants allow the evaluation of the role of the Asn175 side chain on the stability of the active site ion pair and on the intrinsic activity of the enzyme. Alteration of the side chain at position 175 was also found to increase aggregation and proteolytic susceptibility of the proenzyme and to affect the thermal stability of the mature enzyme, reflecting a contribution of the asparagine residue to the structural integrity of papain. The strict conservation of Asn175 in cysteine proteases might therefore result from a combination of functional and structural constraints.

Purification of turnip mosaic potyvirus viral protein genome-linked proteinase expressed in Escherichia coli and development of a quantitative assay for proteolytic activity.

The 49-kDa, nuclear inclusion a-like, viral protein genome-linked proteinase (VPg-Pro) of turnip mosaic potyvirus (TuMV) was expressed in Escherichia coli. The protein was produced in a soluble form at high levels and was active, as demonstrated by intermolecular cleavage of the polymerase capsid protein (Pol-CP) substrate. The VPg-Pro was purified by metal-chelation and ion-exchange chromatographies. Two forms of VPg-Pro, which differed in molecular masses, were obtained during isolation; their identities were confirmed by immunoblot analysis and N-terminal amino acid sequencing. Data indicated that cleavage took place at a site near the C-terminus of VPg-Pro and was the result of the proteolytic activity of the viral protein. The purified proteinase retained enzymic activity on its natural substrate (Pol-CP) and was also capable of hydrolysing the synthetic peptide acyl-Ala-Ala-Val-Tyr-His-Gln-Ala-Ala-NH2, derived from the consensus cleavage site for the TuMV polyprotein. Analysis by mass spectrometry of the two fragments resulting from this reaction indicated that cleavage took place between the Gln and Ala residues, as expected. A fluorogenic derivative of this peptide was hydrolysed by VPg-Pro, affording a convenient quantitative assay for intermolecular proteolytic activity, and was used to determine the pH-activity profile. The availability of large quantities of pure proteinase and of a rapid and sensitive assay will permit detailed kinetic and structural studies which are essential to obtain a better understanding of the mode of action of this and related viral proteinases, such as the 3C proteinase of picornaviruses.

Modification of the electrostatic environment is tolerated in the oxyanion hole of the cysteine protease papain.

The oxyanion hole in cysteine and serine proteases can be viewed as an arrangement of prealigned dipoles that complements the changes in charge distribution during the enzymatic reaction. Because of the electrostatic nature of the interaction involved in the oxyanion hole, the introduction of charged residues in that region could have a major effect on the catalytic properties of the enzyme. In this study, residue Gln19, which contributes to one of the hydrogen bonds in the oxyanion hole of papain, has been replaced by glutamic acid, histidine, and asparagine residues. These mutations result in 65-315-fold decreases in kcat/KM, supporting our previous finding that the side chain of Gln19 contributes to transition state stabilization in the oxyanion hole of papain (Ménard et al., 1991a). Since papain is active over a wide range of pH values, the influence of side chain ionization on activity could be measured quantitatively with the mutant Gln19Glu. The pH dependency of kcat/KM for Gln19Glu is not of the classical bell-shaped form normally observed for papain, but instead is modulated by ionization of the Glu19 side chain with a pKa of 6.02. The Gln19Glu mutant at low pH, where the Glu19 side chain is neutral, is the enzyme that displays activity closest to that of wild-type enzyme, with a (kcat/KM)1lim value only 20-fold lower than that for papain. As expected, the activity of the Gln19Glu mutant decreases when the Glu19 side chain ionizes. However, introduction of the negatively charged glutamate into the oxyanion hole of papain leads to a further reduction in activity of only 12-fold, and this mutant is still more active than the Gln19Ser enzyme and only 3-fold less active than Gln19Asn.(ABSTRACT TRUNCATED AT 250 WORDS)

E64 [trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane] analogues as inhibitors of cysteine proteinases: investigation of S2 subsite interactions.

A number of epoxysuccinyl amino acid benzyl esters (HO-Eps-AA-OBzl) and benzyl amides (HO-Eps-AA-NHBzl) (where AA represents amino acid) were synthesized as analogues of E64, a naturally occurring inhibitor of cysteine proteinases. These inhibitors were designed to evaluate if selectivity for cathepsin B could be achieved by varying the amino acid on the basis of known substrate specificity. Contrary to the situation with substrates, it was found that variation of the amino acid in the E64 analogues does not lead to major changes in the kinetic parameter kinac./Ki and that the specificity of these analogues does not parallel that observed for substrates. This is particularly true in the case of the benzyl ester derivatives where the deviation from substrate-like behaviour is more important than with the benzyl amide derivatives. The results suggest that the amide proton of the benzyl amide group in HO-Eps-AA-NHBzl interacts in the S2 subsite in both cathepsin B and papain and contributes to increase the potency of these inhibitors. The kinetic data also suggest that differences in the orientation of the C alpha-C beta bond of the side chain in the S2 subsite of the enzyme might explain the differences between substrate and E64 analogue specificities. This hypothesis is supported by the fact that the order of inactivation rates with chloromethane inhibitors (which are believed to be good models of enzyme-substrate interactions) is indeed very similar to that observed with the corresponding amidomethylcoumarin substrates. In conclusion, the information available from S2-P2 interactions with substrates cannot be used to enhance the selectivity of the E64 analogues in a rational manner.

The specificity of the S1' subsite of cysteine proteases.

The specificity of the S1' subsite of the cysteine proteases cathepsin B, L, S and papain has been investigated using a series of intramolecularly quenched fluorogenic substrates (Dansyl-Phe-Arg-AA-Trp-Ala) where the P1' amino acid (AA) has been varied. Taken individually, each enzyme displays a relatively broad S1' subsite specificity and this subsite cannot be considered as a primary site of specificity. Notable differences do exist however between the various proteases. Cathepsin B prefers large hydrophobic residues in the P1' position of a substrate while cathepsin L has an opposite trend, favoring amino acids with small (Ala, Ser) or long but non-branched (Asn, Gln, Lys) side chains. Cathepsin S and papain display a somewhat broader S1' subsite specificity.

Epoxysuccinyl dipeptides as selective inhibitors of cathepsin B.

Epoxysuccinyl dipeptide analogs of E-64 (R-EpsLeuPro-R') (Figure 1) have been synthesized with the carboxylate group on the epoxide ring either free (R = OH) or converted to an ester or an amide (R = EtO or i-BuNH) and with the C-terminal amino acid proline either blocked (R' = OBzl) or free (R' = OH). These compounds were used to investigate the recently reported selectivity of this type of inhibitor for the lysosomal cysteine protease cathepsin B. It was shown that derivatization of the carboxylate on the epoxide ring confers selectivity for cathepsin B over papain only when it is combined to a dipeptidyl moiety with a free negatively charged C-terminal residue. It is proposed that this selectivity reflects interactions with histidine residues on a loop located in the primed subsites of cathepsin B which provides a positively charged anchor for the C-terminal carboxylate group of the inhibitor. The primed subsite loop of cathepsin B is not found in other cysteine proteases of the papain family and offers a unique template for designing selectivity in cysteine protease inhibitors.

Contribution of the glutamine 19 side chain to transition-state stabilization in the oxyanion hole of papain.

The existence of an oxyanion hole in cysteine proteases able to stabilize a transition-state complex in a manner analogous to that found with serine proteases has been the object of controversy for many years. In papain, the side chain of Gln19 forms one of the hydrogen-bond donors in the putative oxyanion hole, and its contribution to transition-state stabilization has been evaluated by site-directed mutagenesis. Mutation of Gln19 to Ala caused a decrease in kcat/KM for hydrolysis of CBZ-Phe-Arg-MCA, which is 7700 M-1 s-1 in the mutant enzyme as compared to 464,000 M-1 s-1 in wild-type papain. With a Gln19Ser variant, the activity is even lower, with a kcat/KM value of 760 M-1 s-1. The 60- and 600-fold decreases in kcat/KM correspond to changes in free energy of catalysis of 2.4 and 3.8 kcal/mol for Gln19Ala and Gln19Ser, respectively. In both cases, the decrease in activity is in large part attributable to a decrease in kcat, while KM values are only slightly affected. These results indicate that the oxyanion hole is operational in the papain-catalyzed hydrolysis of CBZ-Phe-Arg-MCA and constitute the first direct evidence of a mechanistic requirement for oxyanion stabilization in the transition state of reactions catalyzed by cysteine proteases. The equilibrium constants Ki for inhibition of the papain mutants by the aldehyde Ac-Phe-Gly-CHO have also been determined. Contrary to the results with the substrate, mutation at position 19 of papain has a very small effect on binding of the inhibitor.(ABSTRACT TRUNCATED AT 250 WORDS)

Importance of hydrogen-bonding interactions involving the side chain of Asp158 in the catalytic mechanism of papain.

In a previous study, it was shown that replacing Asp158 in papain by Asn had little effect on activity and that the negatively charged carboxylate of Asp158 does not significantly stabilize the active site thiolate-imidazolium ion pair of papain (Ménard et al., 1990). In this paper, we report the kinetic characterization of three more mutants at this position: Asp158Gly, Asp158Ala, and Asp158Glu. From the pH-activity profiles of these and other mutants of papain, it has been possible to develop a model that enables us to dissect out the contribution of the various mutations toward (i) intrinsic activity, (ii) ion pair stability, and (iii) the electrostatic potential at the active site. Results obtained with mutants that place either Gly or Ala at position 158 indicate that the hydrogen bonds involving the side chain of Asp158 in wild-type papain are indirectly important for enzyme activity. When CBZ-Phe-Arg-MCA is used as a substrate, the (kcat/KM)obs values at pH 6.5 are 3650 and 494 M-1 s-1 for Asp158Gly and Asp158Ala, respectively, as compared to 119,000 M-1 s-1 for papain. Results with the Asp158Glu mutant suggest that the side chain of Glu moves closer to the active site and cannot form hydrogen bonds similar to those involving Asp158 in papain. From the four mutations introduced at position 158 in papain, we can conclude that it is not the charge but the hydrogen-bonding interactions involving the side chain of Asp158 that contribute the most to the stabilization of the thiolate-imidazolium ion pair in papain. However, the charge and the hydrogen bonds of Asp158 both contribute to the intrinsic activity of the enzyme.

A model to explain the pH-dependent specificity of cathepsin B-catalysed hydrolyses.

1. Three synthetic substrates of cathepsin B (EC 3.4.22.1) with various amino acid residues at the P2 position (Cbz-Phe-Arg-NH-Mec, Cbz-Arg-Arg-NH-Mec and Cbz-Cit-Arg-NH-Mec, where Cbz represents benzyloxycarbonyl and NH-Mec represents 4-methylcoumarin-7-ylamide) were used to investigate the pH-dependency of cathepsin B-catalysed hydrolyses and to obtain information on the nature of enzyme-substrate interactions. 2. Non-linear-regression analysis of pH-activity profiles for these substrates indicates that at least four ionizable groups on cathepsin B with pKa values of 3.3, 4.55, 5.46 and greater than 7.3 can affect the rate of substrate hydrolysis. 3. Ionization of the residue with a pKa of 5.46 has a strong effect on activity towards the substrate with an arginine in P2 (8.4-fold increase in activity) but has only a moderate effect on the rate of hydrolysis with Cbz-Cit-Arg-NH-Mec (2.3-fold increase in activity) and virtually no effect with Cbz-Phe-Arg-NH-Mec. The kinetic data are consistent with this group being an acid residue with a side chain able to interact with the side chains of an arginine or a citrulline in the P2 position of a substrate. Amino acid sequence alignment and model building with the related enzyme papain (EC 3.4.22.2) suggest that Glu-245 of cathepsin B is a likely candidate. The relative importance of electrostatic and hydrophobic interactions in the S2 subsite of cathepsin B is discussed. 4. For all three substrates, activity appears after ionization of a group with a pKa of 3.3, believed to be the active-site Cys-29 of cathepsin B. The identity of the groups with pKa values of 4.55 and greater than 7.3 remains unknown.

Removal of an inter-domain hydrogen bond through site-directed mutagenesis: role of serine 176 in the mechanism of papain.

A mutant of papain, where an inter-domain hydrogen bond between the side chain hydroxyl group of a serine residue at position 176 and the side chain carbonyl oxygen of a glutamine residue at position 19 has been removed by site-directed mutagenesis, has been produced and characterized kinetically. The mutation of Ser176 to an alanine has only a small effect on the kinetic parameters, the kcat/Km for hydrolysis of CBZ-Phe-Arg-MCA by the Ser176Ala enzyme being of 8.1 x 10(4) /M/s compared with 1.2 x 10(5) /M/s for papain. Serine 176 is therefore not essential for the catalytic functioning of papain, even though this residue is conserved in all cysteine proteases sequenced. The pH-activity profiles were shown to be narrower in the mutant enzyme by up to 1 pH unit at high ionic strength. This result is interpreted to indicate that replacing Ser176 by an alanine destabilizes the thiolate-imidazolium form of the catalytic site Cys25-His159 residues of papain. Possible explanations for that effect are given and the role of a serine residue at position 176 in papain is discussed.

A protein engineering study of the role of aspartate 158 in the catalytic mechanism of papain.

The controversy concerning the various suggested roles for the side chain of Asp158 in the active site of papain has been clarified by using site-directed mutagenesis. Both wild-type papain and an Asp158 Asn variant were produced in a baculovirus-insect cell expression system, purified to homogeneity from the culture, and characterized kinetically. With CBZ-Phe-Arg-MCA as substrate, the kcat/KM and kcat values obtained for the Asp158Asn papain are 20,000 M-1.s-1 and 34 s-1, respectively, as compared with values of 120,000 M-1.s-1 and 51 s-1 obtained for the wild-type papain. In addition, the pH-(kcat/KM) profile for the Asp158Asn enzyme is shifted relative to that for the wild-type enzyme to lower values by approximately 0.3 pH unit. This shows clearly that Asp158 is not, as previously postulated, an essential catalytic residue. In addition, the pH dependency data are interpreted to indicate that, contrary to earlier suggestions, the negatively charged side chain of Asp158 does not significantly stabilize the active-site thiolate-imidazolium ion pair. However, its presence does influence the pKa's associated with ion-pair formation in a manner compatible with electrostatic considerations.