OmpA can form folded and unfolded oligomers.

The monomeric outer membrane protein OmpA from Escherichia coli has long served as a model protein for studying the folding and membrane insertion of ß-barrel membrane proteins. Here we report that when OmpA is refolded in limiting amounts of surfactant (close to the cmc), it has a high propensity to form folded and unfolded oligomers. The oligomers exist both in a folded and (partially) unfolded form which both dissociate under denaturing conditions. Oligomerization does not require the involvement of the periplasmic domain and is not strongly affected by ionic strength. The folded dimers can be isolated and show native-like secondary structure; they are resistant to proteolytic attack and do not dissociate in high surfactant concentrations, indicating high kinetic stability once formed. Remarkably, OmpA also forms significant amounts of higher order structures when refolding in the presence of lipid vesicles. We suggest that oligomerization occurs by domain swapping favored by the high local concentration of OmpA molecules congregating on the same micelle or vesicle. In this model, the unfolded oligomer is stabilized by a small number of intermolecular ß-strand contacts and subsequently folds to a more stable state where these intermolecular contacts are consolidated in a native-like fashion by contacts between complementary ß-strands from different molecules. Our model is supported by the ability of complementary fragments to associate with each other in vitro. Oligomerization is probably avoided in the cell by the presence of cellular chaperones which maintain the protein in a monomeric state.

pH regulation of the kinetic stability of the lipase from Thermomyces lanuginosus.

Thermomyces lanuginosus lipase (TlL) is a kinetically stable protein, resistant toward both denaturation and refolding in the presence of the ionic surfactant sodium dodecyl sulfate (SDS) and the nonionic surfactant decyl maltoside (DecM). We investigate the pH dependence of this kinetic stability. At pH 8, TlL remains folded and enzymatically active at multimillimolar surfactant concentrations but fails to refold from the acid urea-denatured state at submillimolar concentrations of SDS and DecM, indicating a broad concentration range of kinetic trapping or hysteresis. At pH 8, very few SDS molecules bind to TlL. The hysteresis SDS concentration range shrinks when moving to pH 4-6; in this pH range, SDS binds as micellelike clusters. Although hysteresis can be eliminated by reducing disulfide bonds, destabilizing the native state, and lowering the unfolding activation barrier, SDS sensitivity is not directly linked to intrinsic kinetic stability [its resistance to the general chemical denaturant guanidinium chloride (GdmCl)], because TlL unfolds more slowly in GdmCl at pH 6.0 than at pH 8.0. However, the estimated net charge drops from approximately -12 to approximately -5 between pH 8 and 6. SDS denatures TlL at pH 6.0 by nucleating via a critical number of bound SDS molecules on the surface of native TlL to form clusters. These results imply that SDS sensitivity is connected to the availability of appropriately charged regions on the protein. We suggest that conformational rigidity is a necessary but not sufficient feature of SDS resistance, because this has to be combined with sufficient negative electrostatic potential to avoid extensive SDS binding.

Using micellar mole fractions to assess membrane protein stability in mixed micelles.

The increased focus on the structural and physical properties of membrane proteins has made it critical to develop methods that provide a reliable estimate of membrane protein stability. A simple approach is to monitor the protein's conformational changes in mixed detergent systems, typically consisting of an anionic (denaturing) and non-ionic (non-denaturing) component. Linear correlations between, e.g., the melting temperature and the bulk mole fraction of the anionic component have been observed. However, a potential complication is that the bulk mole fraction is not identical to the mole fraction in the mixed micelle, which is the local environment experienced by the membrane protein. Here, we present an extensive analysis of the thermal stability of the membrane-integrated domain of the outer membrane protein AIDA in the presence of different mixed micelles. In the micelle system SDS-octyl-polyoxyethylene, the melting temperature in the absence of SDS extrapolates to 113 degrees C using bulk mole fractions. However, for mixed micelles involving short-chain detergents or phospholipids, the melting temperature calculated using bulk mole fractions reaches values up to several hundred degrees higher than 113 degrees C and can only be obtained by extrapolation over a narrow mole fraction interval. Furthermore, there is a non-linear relationship between the melting temperature and bulk mole fractions for mixed micelle systems involving cationic detergents (also denaturing). We show that if we instead use the micellar mole fraction as a parameter for denaturing detergent strength, we obtain linear correlations which extrapolate to more or less the same value of the melting temperature. There remains some scatter in the extrapolated values of the melting temperature in different binary systems, which suggest that additional micellar interactions may play a role. Nevertheless, in general terms, the mixed micellar composition is a good parameter to describe the membrane protein's microenvironment. Note, however, that for the mixed micelle system involving SDS and dodecyl maltoside, which has been used by several research groups to determine membrane protein stability, the estimate provided by bulk mole fraction leads to similar values as that of micellar mole fractions.

A simple way to measure protein refolding rates in water.

Refolding of proteins is traditionally carried out either by diluting the denaturant-unfolded protein into buffer (GdmCl-jump) or by mixing the acid-denatured protein with strong buffer (pH-jump). The first method does not allow direct measurement of folding rates in water since the GdmCl cannot be infinitely diluted, and the second method suffers from the limitation that many proteins cannot be pH-denatured. Further, some proteins do not refold reversibly from low pH where they get trapped as aggregation prone intermediates. Here, we present an alternative approach for direct measurement of refolding rates in water, which does not rely on extrapolation. The protein is denatured in SDS, and is then mixed with alpha-cyclodextrin, which rapidly strips SDS molecules from the protein, leaving the naked unfolded protein to refold.

The rate of isomerisation of peptidyl-proline bonds as a probe for interactions in the physiological denatured state of chymotrypsin inhibitor 2.

There are four peptidyl-proline bonds in the 64-residue protein chymotrypsin inhibitor 2 (CI2), all of which are in the trans conformation in the native structure. The isomerisation of one or more of these peptidyl-proline bonds to the cis conformation in the denatured state gives rise to heterogeneity, leading to both fast and slow-folding species. The refolding of the fast-folding species, which has all trans peptidyl-proline bonds, is much faster than that of the slow-folding species, which have one or more cis peptidyl-proline bonds. In CI2, the slow-folding species can be classified into two groups by their rates of refolding, temperature-dependence, pH-dependence and [GdmCl]-dependence of the rate constants and the effect of peptidyl-prolyl isomerase on the rate constants. The replacement of Pro6 by Ala removes one of the slow refolding phases, suggesting that the cis peptidyl-Pro6 conformation is solely responsible for one of the slow-folding species. Pro6 is located in a region of the protein where non-random interactions have been found in a series of N-terminal fragments of CI2 (residues 1 to 13, 1 to 25, 1 to 28 and 1 to 40). In addition, NMR studies on a mutant fragment, (1-40)T3A, have confirmed that this non-native interaction is associated with the bulky side-chain of Trp5. The atypical rate of cis to trans isomerisation of the peptidyl-Pro bond is indicative of the presence of a similar hydrophobic cluster in the physiological denatured state of intact CI2.

The structure of the transition state for folding of chymotrypsin inhibitor 2 analysed by protein engineering methods: evidence for a nucleation-condensation mechanism for protein folding.

The 64-residue protein chymotrypsin inhibitor 2 (CI2) is a single module of structure. It folds and unfolds as a single co-operative unit by simple two-state kinetics via a single rate determining transition state. This transition state has been characterized at the level of individual residues by analysis of the rates and equilibria of folding of some 100 mutants strategically distributed at 45 sites throughout the protein. Only one residue, a helical residue (Ala16) buried in the hydrophobic core, has its full native interaction energy in the transition state. The only region of structure which is well developed in the transition state is the alpha-helix (residues 12 to 24). But, the interactions within it are weakened, especially at the C-terminal region. The rest of the protein has varying degrees of weakly formed structure. Thus, secondary and tertiary interactions appear to form concurrently. These data, reinforced by studies on the structures of peptide fragments, fit a "nucleation-condensation" model in which the overall structure condenses around an element of structure, the nucleus, that itself consolidates during the condensation. The high energy transition state is composed of the whole of the molecule making a variety of weak interactions, the nucleus being those residues that make the strongest interactions. The nucleus here is part of the alpha-helix and some distant residues in the sequence with which it makes contacts. The remainder of the protein has to be sufficiently ordered that it provides the necessary interactions to stabilize the nucleus. The nucleus is only weakly formed in the denatured state but develops in the transition state. The onrush of stability as the nucleus consolidates its local and long range interactions is so rapid that it is not yet fully formed in the transition state. The formation of the nucleus is thus coupled with the condensation. These results are consistent with a recent simulation of the folding of a computer model protein on a lattice which is found to proceed by a nucleation-growth mechanism. We suggest that the mechanism of folding of CI2 may be a common theme in protein folding whereby fundamental folding units of larger proteins, which are modelled by the folding of CI2, form by nucleation-condensation events and coalesce, perhaps in a hierarchical manner.

Hydrogen exchange in chymotrypsin inhibitor 2 probed by mutagenesis.

Two-dimensional NMR spectroscopy has been used to monitor hydrogen-deuterium exchange in chymotrypsin inhibitor 2. Application of two independent tests has shown that at pH 5.3 to 6.8 and 33 to 37 degrees C, exchange occurs via an EX2 limit. Comparison of the exchange rates of a number of mutants of CI2 with those of wild-type identifies the pathway of exchange, whether by local breathing, global unfolding or a mixture of the two pathways. For a large number of residues, the exchange rates were unaffected by mutations which destabilized the protein by up to 1.9 kcal mol(-1), indicating that exchange is occurring through local fluctuations of the native state. A small number of residues were found for which the mutations had the same effect on the rate constants for exchange as on the equilibrium constant for unfolding, indicating that these residues exchange by global unfolding. These are residues that have the slowest exchange rates in the wild-type protein. We see no correspondence between these residues and residues involved in the nucleation site for the folding reaction identified by protein engineering studies. Rather, the exchange behaviour of CI2 is determined by the native structure: the most protected amide protons are located in regions of hydrogen bonding, specifically the C terminus of the alpha-helix and the centre of the beta-sheet. A number of the most slowly exchanging residues are in the hydrophobic core of the protein.

Folding of circular permutants with decreased contact order: general trend balanced by protein stability.

To examine the influence of contact order and stability on the refolding rate constant for two-state proteins, we have analysed the folding kinetics of the small beta-alpha-beta protein S6 and two of its circular permutants with relative contact orders of 0.19, 0.15 and 0.12. Data reveal a small but significant increase of the refolding rate constant (log k(f)) with decreasing contact order. At the same time, the decreased contact order is correlated to losses in global stability and alterations of the folding nucleus. When the differences in stability are accounted for by addition of Na2SO4 or by comparison of the folding kinetics at the transition mid-point, the dependence between log k(f) and contact order becomes stronger and follows the general correlation for two-state proteins. The observation emphasizes the combined action of topology and stability in controlling the rate constant of protein folding.

Structure of the transition state for folding of a protein derived from experiment and simulation.

Independent experimental and theoretical studies of the unfolding of barley chymotrypsin inhibitor 2 (CI2) are compared in an attempt to derive plausible three-dimensional structural models of the transition states. A very simple structure index is calculated along the sequence for the molecular dynamics-generated transition state models to facilitate comparison with the phi F values. The two are in good agreement overall (correlation coefficient = 0.87), which suggests that the theoretical models should provide a structural framework for interpretation of the phi F values. Both experiment and simulation indicate that the transition state is a distorted form of the native state in which the alpha-helix is weakened but partially intact and the beta-sheet is quite disrupted. As inferred from the phi f values and observed directly in the simulations, the unfolding of CI2 is cooperative and there is a "folding core" comprising a patch on the alpha-helix and a portion of the beta-sheet, nucleated by interactions between Ala16, Ile49 and other neighbouring residues. The protein becomes less structured radiating away from this core. Overall the data indicate that CI2 folds by a nucleation-collapse mechanism. In the absence of experimental information, we have little confidence that the molecular dynamics simulations are correct, especially when only one or a few simulations are performed. On the other hand, even though the experimentally derived phi values may reflect the extent of overall structure formation, they do not provide an actual atomic-resolution three dimensional structure of the transition state. By combining the two approaches, however, we have a framework for interpreting phi F values and can hopefully arrive at a more trustworthy model of the transition state. The process is in some ways similar to the combination of molecular dynamics and NMR data to solve the tertiary structure of proteins.

Conformational pathway of the polypeptide chain of chymotrypsin inhibitor-2 growing from its N terminus in vitro. Parallels with the protein folding pathway.

We have obtained a series of fragments growing from the N terminus of the protein chymotrypsin inhibitor-2 (C12) in order to study the development of structure on elongation of the polypeptide in solution. We present an extensive biophysical characterization of ten fragments using different conformational probes. Small fragments up to residue 40 of the 64-residue protein are disordered. Fragment (1-40) has non-native local hydrophobic clusters, but nevertheless does not bind 8-anilinonaphthalene-1-sulphonate (ANS). Hydrophobic regions in longer fragments become gradually more capable of binding ANS as the chain grows to completion, with a tendency to form native structures. Major changes in secondary structure and accessibility to hydrophobic sites occur in parallel, between (1-40) and (1-53), together with changes in hydrodynamic volume and flexibility. NMR studies of (1-53), the first fragment displaying tertiary interactions, show that a subcore is fully formed and the alpha-helix (residues 12 to 24) is of fluctuating structure. Fragments (1-53) and (1-60) share many properties with molten globule-like structures, with varying degrees or order. Fluorescence properties of the native fold are gradually recovered from fragments (1-60) to full-length C12, together with a decrease in hydrophobic exposure. A small degree of co-operativity of formation of structure appears when residue 60 is added, gradually increasing as residue 62 is added, but a full two-state co-operative transition appears only on addition of Arg62 and Val63. We believe this is the result of correct side-chain packing of the hydrophobic core, capping the major elements of secondary structure in C12 at this late stage, which is probed by the complete recovery of the fluorescence of the unique Trp5. The structures that develop as the polypeptide chain increases in length parallel the structural features present in the nucleus for the folding of intact protein, which develops in the transition state. The folding nucleus consists of much of the helix and the interactions made by Ala16 in the helix with residues in the core, especially with Leu49 and Ile57, with the rest of the structure being formed only very weakly in the transition state.

A comparative study of the unfolding of the endoglucanase Cel45 from Humicola insolens in denaturant and surfactant.

Cellulases are increasingly being used for industrial purposes, particularly in washing powders, yet little is known of the factors governing the stability of proteins in detergent solutions. We present a comparative analysis of the behavior of the cellulase Cel45 from Humicola insolens in the presence of the denaturant guanidinium chloride and the anionic detergent C12-LAS. Although Cel45 unfolds in GdmCl according to a simple two-state model under equilibrium conditions, it accumulates a transient intermediate during refolding. The four disulfide bonds do not contribute detectably to the stability of the native state. Cel45 is unfolded by very low concentrations of C12-LAS (1-4 mM). An analysis of 16 mutants of Cel45 shows a very weak correlation between unfolding rates in denaturant and detergent; mutants that have the same unfolding rate in GdmCl (within a factor of 1.5) vary 1,000-fold in their unfolding rates in C12-LAS. The data support a simple model for unfolding by detergent, in which the introduction of positive charges or removal of negative charges greatly increases detergent sensitivity, while interactions with the hydrophobic detergent tail contribute to a smaller extent. This implies that different detergent-mediated unfolding pathways exist, whose accessibilities depend on individual residues. Double-mutant cycles reveal that mutations in two proximal residues lead to repulsion and a destabilization greater than the sum of the individual mutations as measured by GdmCl denaturation, but they also reduce the affinity for LAS and therefore actually stabilize the protein relative to wild-type. Ligands that interact strongly with the denatured state may therefore alter the unfolding process.

Reduction in the amount of 8-hydroxy-2'-deoxyguanosine in the DNA of SV40-transformed human fibroblasts as compared with normal cells in culture.

DNA damage due to oxidative free radicals is considered to be a major cause of ageing and age-related diseases including cancer. Of more than 20 modifications formed in DNA by the action of hydroxyl radicals, 8-hydroxy-2'-deoxyguanosine (oh8dG) is potentially highly mutagenic and is known to occur most frequently. Using HPLC combined with electrochemical (HPLC/EC) detection of oh8dG, fivefold higher levels of oh8dG are detected in the DNA of cultured normal human skin fibroblasts as compared with SV40-transformed human fibroblasts MRC-5V2. In comparison, the levels of oh8dG were similar in the growth medium of both types of cells. Applications of this method range from studies on the genomic stability and instability of normal and cancerous cells to the clinical and laboratory testing of toxic substances and drugs.

Dual hydrolytic role for Pb(II) ions.

RNA phosphodiester bonds can be cleaved by metal ions, of which Pb2+ is one of the most effective. It can cleave both generally and site-specifically, depending on the substrate and the conditions. In addition, metal ions are also known to cleave ester bonds between amino acid and the 3'-end of transfer RNA. Here we report that in aminoacylated transfer RNA, Pb2+ ions cleave internucleotide bonds in the 3'-end of tRNA and also cleaves the bond between tRNA and its amino-acid, attached at the 3'-end via an ester bond to the terminal ribose in aminoacyl tRNA. The two reactions proceed at different rates. The rate of deacylation is significantly faster than the rate of cleavage of phosphodiester bonds, with a pH-optimum of 7. This dual hydrolytic role is not seen for other metal ions examined, namely Zn(II), Cd(II) and Mn(II). The rate of the two kinds of hydrolyses by Pb2+ ions is compared with that of other metal-ions. The mechanism of cleavage is investigated further by modification of the 3'-end of tRNA.

Variant of the Thermomyces lanuginosus lipase with improved kinetic stability: a candidate for enzyme replacement therapy.

Lipases with high kinetic stability and enzymatic efficiency in the human gastro-intestinal tract may help against exocrine pancreatic insufficiency. Here we mimic gastric conditions to study how bile salts and pH affect the stability and activity of Thermomyces lanuginosus lipase (TlL) and its stabler variant StL using spectroscopy, calorimetry and gel electrophoresis. Both enzymes resist trypsin digestion with and without bile salts. Bile salts activate native TlL and StL equally well, bind weakly to denatured TlL and StL at lower pH and precipitate native TlL and StL at pH 4. StL refolds more efficiently than TlL from gastric pH in bile salts, regaining activity when refolding from pH as low as 1.8 and above while TlL cannot go below pH 2.6. StL also unfolds 10-40 fold more slowly in the denaturant guanidinium chloride and the anionic surfactant SDS. We ascribe StL's superior performance to general alterations in its electrostatic potential which makes it more acid-resistant. These superior properties make StL a good candidate for pancreatic enzyme replacement therapy.

SAXS models of TGFBIp reveal a trimeric structure and show that the overall shape is not affected by the Arg124His mutation.

Human transforming growth factor ß induced protein (TGFBIp) is composed of 683 residues, including an N-terminal cysteine-rich (EMI) domain, four homologous fasciclin domains, and an Arg-Gly-Asp (RGD) motif near the C-terminus. The protein is of interest because mutations in the TGFBI gene encoding TGFBIp lead to corneal dystrophy (CD), a condition where protein aggregates within the cornea compromise transparency. The complete three-dimensional structure of TGFBIp is not yet available, with the exception of a partial X-ray structure of the archetype FAS1 domain derived from Drosophila fasciclin-1. In this study, small-angle X-ray scattering (SAXS) models of intact wild-type (WT) human TGFBIp and a mutant (R124H) are presented. The mutation R124H leads to a variant of granular CD. The deduced structure of the TGFBIp monomer consists of four FAS1 domains in a simple "beads-on-a-string" arrangement, constructed by the superimposition of four consecutive Drosophila fasciclin domains. The SAXS-based model of the TGFBIp R124H mutant displayed no structural differences from WT. Both WT TGFBIp and the R124H mutant formed trimers at higher protein concentrations. The similar association properties and three-dimensional shape of the two proteins suggest that the mutation does not induce any major structural rearrangements, but points towards the role of other corneal-specific factors in the formation of corneal R124H deposits.

In vitro association of fragments of a beta-sheet membrane protein.

Although the beta-barrel membrane protein OmpA can be produced in a biologically active form in E. coli from co-expressed fragments, the fragments have not been demonstrated to associate in vitro. We have produced 3 complementary fragment pairs of OmpA which can associate to form a folded complex according to the SDS band-shift assay. We are able to convert 25-35% of the fragment populations to non-covalent but SDS-stable complexes. The periplasmic chaperone Skp effectively prevented this association. Two separately expressed and purified overlapping fragments of OmpA can form a protease-resistant complex that undergoes the characteristic band-shift upon heating. Our work demonstrates that although membrane insertion and folding of beta-barrel membrane proteins may be a cooperative process, the fragments can associate in vitro without any additional components. However, the low yield and slow folding rates indicate that partially unfolded or destabilized beta-sheet membrane proteins can potentially engage in many non-native interactions.

The diversity of FtsY-lipid interactions.

The bacterial signal recognition particle (SRP) receptor FtsY forms a complex with the SRP Ffh to target nascent polypeptide chains to the bacterial inner membrane. How FtsY interacts with lipids and associates to the membrane is unclear. Here, we show that vesicle binding leads to partial protection against proteolytic degradation and a change in secondary structure, which differs depending on whether the lipids are simple mixtures of zwitterionic and anionic lipids, mimics of Escherichia coli lipids, or lysolipids. Lipid binding alters the stability of FtsY. Thermal unfolding of FtsY in buffer shows two transitions, one occurring at approximately 60 degrees C and the other at approximately 90 degrees C. The thermal intermediate accumulating between 60 and 90 degrees C has structural features in common with the state induced by binding to E. coli lipids. E. coli lipid extract induces a single transition around 70 degrees C, anionic lipids have no effect while cooperative unfolding is completely removed in lysolipids. Thus, the lipid environment profoundly influences the dynamic properties of FtsY, leading to three different kinds of FtsY-lipid interactions with different effects on structure, proteolytic protection, and stability, and is driven both by hydrophobic and electrostatic interactions. Trypsin digestion experiments highlight the central role of the N-domain in lipid contacts, whereas the A- and G-domains appear to play a more minor part.

Alternative membrane protein conformations in alcohols.

Alcohols modulate the oligomerization of membrane proteins in lipid bilayers. This can occur indirectly by redistributing lateral membrane pressure in a manner which correlates with alcohol hydrophobicity. Here we investigate the direct impact of different alcohol-water mixtures on membrane protein stability and solubility, using the two detergent-solubilized alpha-helical membrane proteins DsbB and NhaA. Both proteins precipitate extensively at intermediate concentrations of alcohols, forming states with extensive (40-60%) beta-sheet structure and affinity for the fibril-specific dye thioflavin T, although atomic force microscopy images reveal layer-like and spherical deposits, possibly early stages in a fibrillation process trapped by strong hydrophobic contacts. At higher alcohol concentrations, both DsbB and NhaA are resolubilized and form non-native structures with increased (DsbB) or decreased (NhaA) helicity compared to the native state. The alternative conformational states cannot be returned to the functional native state upon dilution of alcohol. The efficiency of precipitation and the degree to which DsbB is destabilized at low alcohol concentrations show the same correlation with alcohol hydrophobicity. Thus, in addition to their effect on the membrane, alcohols perturb membrane proteins directly by solvating the hydrophobic regions of the protein. At intermediate concentrations, this perturbation exposes hydrophobic segments but does not provide sufficient solvation to avoid intermolecular association. Resolubilization requires a reduction in the relative dielectric constant below 65 in conjunction with specific properties of the individual alcohols. We conclude that alcohols provide access to a diversity of conformations for membrane proteins but are not a priori suitable for solution studies requiring reversible denaturation of monomeric proteins.

Side-chain determinants of beta-sheet stability.

beta-Sheet propensities of different amino acids depend on the context of both secondary and tertiary structure. In an attempt to establish general empirical relationships that determine this context dependence, we have determined the free energy of unfolding of a series of mutants at six positions in the beta-sheet of chymotrypsin inhibitor 2 (CI2). We have generated the series Val-->Ala-->Gly and Val<==>Thr at five positions, as well as the side-chain deletion Ile-->Val at residue 49 and Ala-->Gly at residue 77. In the series Val-->Ala-->Gly, the ranking order in terms of stability is Val > Ala > Gly at all positions. However, the change in free energy on deletion of methylene groups varies greatly. When Val and Thr are interchanged, the wild-type residue is always the more stable, but by a different amount at each position. We have attempted to rationalize the data by relating it to changes in solvent-accessible surface area, packing density, and statistically derived pseudo-energy functions that depend on phi, psi angles. There is no significant correlation of the energies with any of the variables except with the pseudo-energy function, but the deviations from these values are large. We conclude that thermodynamic scales for beta-sheet propensity are currently of insufficient precision for general design purposes, although they may be useful in special cases.

Analysis of protein-protein interactions by mutagenesis: direct versus indirect effects.

Site-directed mutagenesis, including double-mutant cycles, is used routinely for studying protein-protein interactions. We now present a case analysis of chymotrypsin inhibitor 2 (CI2) and subtilisin BPN' using (i) a residue in CI2 that is known to interact directly with subtilisin (Tyr42) and (ii) two CI2 residues that do not have direct contacts with subtilisin (Arg46 and Arg48). We find that there are similar changes in binding energy on mutation of these two sets of residues. It can thus be difficult to interpret mutagenesis data in the absence of structural information.