Protein Engineering, Vol. 12, No. 1, 41-45,
January 1999
© 1999 Oxford University Press
Analysis of protein–protein interactions by mutagenesis: direct versus indirect effects
MRC Unit for Protein Function & Design, Cambridge IRC for Protein Engineering, University Chemical Laboratory, Lensfield Road, Cambridge CB2 1EW, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
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.
Keywords: chymotrypsin inhibitor 2/inhibitory activity/loop flexibility/protein–protein interactions/stability/subtilisin BPN'
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Protease inhibitors bind very tightly to their cognate proteases, usually via an extended loop that docks onto the protease (Bode et al., 1992). Binding is tight for two main reasons: the inhibitors form numerous specific side chain and backbone interactions with the active site of the protease (even when cleaved) and the loop is generally very rigid, reducing the unfavourable decrease in entropy that accompanies complex formation (Blow, 1974
). The structure of an inhibitor–protease complex will generally reveal the intermolecular contacts that contribute most to the strength of association, but this analysis needs to be complemented by functional studies, as is the case for all protein–protein interactions. For instance, only a quarter of the side chains buried at the interface between human growth hormone and its receptor contribute significantly to binding energy (Cunningham et al., 1993). Conversely, several electrostatic intermolecular interactions in the barnase–barstar complex are stronger than might be expected at first sight of the structure (Schreiber et al., 1995). The energetic coupling between two residues in a protein–protein complex is best analysed using double mutant cycles, where the two residues are mutated singly and together and the effect on binding is analysed (Carter et al., 1984; Schreiber et al., 1995). In the absence of a direct or indirect coupling, mutation of one residue will be independent of the mutation of the other residue. This approach has also been used to map the architecture of the complex between the Shaker potassium channel and a high-affinity peptide inhibitor (Hidalgo et al., 1995) where the structures of the individual components, but not the complex, are available. Arg24 on the inhibitor was deduced to be in intimate contact with Asp431 in the K+-channel, since a detailed mutagenic analysis of the residues surrounding Arg24 and Asp431 ruled out an unspecific through-space electrostatic attraction (Hidalgo et al., 1995).
In the absence of available structural information, interactions between two proteins are often mapped out by mutational scanning. When deletion of a given side chain results in decreased binding, that side chain is assumed to be situated at the intermolecular interface. However, we show that it is necessary to distinguish between the direct and indirect effects of deletion of a side chain. As a model system we use chymotrypsin inhibitor 2 (CI2), which inhibits subtilisin BPN' with a dissociation constant Ki of about 10–12 M (Longstaff et al., 1990). It consists of a single domain, in which one face of a six-stranded ß-sheet packs against an 
-helix to form the hydrophobic core. From the opposite face of the ß-sheet, between strands 3 and 4, projects an 11-residue extended loop (Gly35–Asp45) which contains the reactive site bond between Met40 and Glu41 (Harpaz et al., 1994; McPhalen et al., 1987). In contrast to many other inhibitors, CI2 does not have disulfide bonds in the reactive site loop; nonetheless, a backbone dynamics study using 15N relaxation data showed that the loop is almost as constrained as the rest of the protein (Shaw et al., 1995), because of a number of noncovalent side-chain interactions. In complex with subtilisin BPN', residues 36–42 in the reactive site loop are responsible for virtually all the contacts <4 Å with the protease (McPhalen et al., 1985).
In this study, we have mutated residues from two different environments (Figure 1). (i) A residue (Tyr42 in the P2' site) which forms intermolecular contacts with subtilisin: Tyr42 is very solvent-exposed in free CI2, and stacks against Phe189 of subtilisin BPN' in a hydrophobic interaction. [Nomenclature of Schechter and Berger (1967). Amino acid residues of the inhibitor are numbered P1, P2, P3, etc. towards the N-terminus and P1', P2', P3', etc. towards the C-terminus from the reactive-site bond (the peptide bond in the inhibitor cleaved by the enzyme).] The mutation YA42 removes this stacking interaction, while YG42 additionally increases the conformational freedom of the reactive loop. (ii) Two residues (Arg46 and Arg48) which are not in contact with subtilisin but whose side chains contribute to the rigidity of the loop by hydrogen bonding and electrostatic interactions with Thr39 and Glu41 (residues P1 and P1'). RC48 removes the guanidinium group while retaining a mimic of the hydrophobic linear part of the side chain; the mutants RA46, RA48 and RA46/RA48 remove both charged and hydrophobic moieties from the Arg side chain. The mutants TA39 and EA41 have dissociation constants increased 80- and 140-fold relative to wild-type CI2 (Jackson et al., 1994), and in the case of EA41, inhibition is reversed over longer timescales. We show how both sets of residues contribute substantially to the binding energy in the CI2–subtilisin complex, but the effect is direct in the case of Tyr42 and only indirect in the case of Arg46 and Arg48.
|
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Materials
All materials were obtained as described (Jackson et al., 1994).
Mutagenesis
Mutants of CI2 were prepared, expressed and purified as described (Jackson et al., 1993). For all mutants of CI2, except YA42 and YG42, concentrations were determined spectroscopically using an extinction coefficient 
= 6965 cm–1 M–1, as determined by the method of Gill and von Hippel (1989). The loss of the Tyr side chain reduced the extinction coefficient of CI2 to = 5780 cm–1 M–1 for YA42 and YG42.
Equilibrium denaturation
Equilibrium unfolding of CI2 mutants in guanidinium chloride (GdmCl) was monitored by fluorescence spectroscopy and the data analysed as described (Clarke et al., 1993; Jackson et al., 1993). The destabilization relative to wild-type CI2 (
GU-F) is calculated as follows:
|
|
where <m> is the average dependence of the free energy of unfolding on denaturant concentration [the value of 1.93 kcal mol–1 M–1 is based on measurements of a large number of CI2 mutants (Itzhaki et al., 1995)], and [GdmCl]50% is the denaturant concentration at which the protein is 50% unfolded. Equilibrium denaturation and inhibition studies on RC48 were performed in 5 mM DTT to avoid possible dimer formation.
Inhibition studies
CI2 is a slow tight-binding inhibitor of subtilisin BPN' with a dissociation constant of 2.910–12 M (Longstaff et al., 1990). The inhibition parameters are determined by following the proteolytic activity of subtilisin BPN' towards the substrate succinyl-Ala-Ala-Pro-Phe-p-nitroaniline (sAAPFpNA) (measured as the increase in absorbance at 412 nm) in 10 mM Tris–HCl pH 8.6 at 25°C in the presence of different concentrations of CI2 (where [CI2] 
10[subtilisin BPN']) (Cha, 1975; Jackson et al., 1994). The concentration of subtilisin BPN' was between 100 and 1000 pM, while the range of inhibitor concentration depended on the dissociation constant. For strong inhibitors (wild-type CI2 and the mutants YA42, RA46, RA48 and RC48), the concentration was 10–30 times that of the enzyme. For the weak inhibitors, the concentration was 10–2000 times that of the enzyme. At these concentrations, the association rate (kon[CI2]) is so fast that steady-state conditions are reached within the deadtime of the experiment (30 s). However, the association between the weak inhibitors and subtilisin BPN' could be measured directly by stopped flow fluorescence spectroscopy using an Applied Photophysics Stopped-Flow spectrophotometer model SF 17MV. Excitation was at 280 nm and emission was monitored at wavelengths above 315 nm using a glass filter. To obtain sufficiently large fluorescence signals, the enzyme concentration was 0.5 µM, while the inhibitor concentration was 5–20 µM.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Stability data
Both wild-type CI2 and the six mutants unfold according to a cooperative two-state mechanism (Figure 2). Stability data are shown in Table I. Despite its solvent exposure, the phenol ring of Tyr42 makes hydrophobic contacts with the side chains of Met40, Glu41, Arg43, Ile44 and Val63. Based on the correlation between loss of hydrophobic contacts within 6 Å and loss of stability in a three-residue hydrophobic patch in the reactive loop (Otzen et al., 1995), we would predict that YA42 and YG42 are destabilized by 2.25 and 2.46 kcal/mol, which is close to the actual values of 1.93 and 2.21 kcal/mol. Therefore, although the insertion of Gly in position 42 is likely to increase the flexibility of the loop, the destabilization of YG42 relative to YA42 can be attributed mainly to the loss of packing interactions by deletion of the Cß methyl group.
|
|
Arg48 is only able to form a hydrogen bond with Thr39, whereas Arg46 is involved in electrostatic interactions with both Glu41 and the C-terminus. RA48 is destabilized by only 1.2 kcal/mol and RA46 by 2.9 kcal/mol.
The interaction energy between Arg46 and Arg48 was measured by a double mutant cycle. The interaction energy Gint between two residues X and Y is calculated from
|
|
where GU-F
XA,YA is the change in stability (relative to wild type) of the mutant where both X and Y are mutated, and GU-FXA and GU-FYA are the changes in stability of the mutants where X and Y, respectively, are mutated (Fersht et al., 1985, 1992; Horovitz et al., 1990). The interaction energy between Arg46 and Arg48 is 1.10±0.10 kcal/mol, that is, Arg46 and Arg48 interact favourably in the native protein, relative to its denatured state. Evidently the interactions between the hydrophobic groups in the Arg side chains outweigh the unfavourable charge–charge interactions, although the guanine groups are within 4–6 Å of each other.
Inhibition data
Representative inhibition profiles are shown in Figure 3. These plots yield the observed rate constant of binding kobs, the initial rate of hydrolysis vo and the steady-state rate of hydrolysis vs (Figure 4), from which we calculate the inhibition parameters (Table II). Dissociation constants are converted into binding energies relative to wild-type CI2 as follows:
|
|
|
|
|
There is no correlation between changes in stability and changes in inhibitory activity. YA42 and YG42 are almost equally destabilized but YG42 has an inhibitory constant ~70 times higher than YA42. Conversely, RA46, RA48 and RC48 have very different stabilities but very similar Ki values. In all cases, the increase in Ki is predominantly due to an increase in the dissociation rate koff.
Tyr42 is strongly conserved in the potato inhibitor I family of serine protease inhibitors, either as Tyr, Phe or Leu (Svendsen et al., 1982). The phenol ring is offset-stacked against Phe189 in subtilisin BPN'. This parallel-ring aromatic–aromatic interaction is not so favourable as the perpendicular arrangement, where there is an electrostatic attraction. In the offset stacking arrangement, the main contribution to binding is from van der Waals interactions (Jorgensen et al., 1990). From the number of hydrophobic contacts between Tyr42's phenol ring and subtilin within 6 Å (42), we predict a reduction in binding energy (Gbinding) of 2.3 kcal/mol on removal of the phenol ring, which is identical within error to the actual value (2.2 kcal/mol). In this case, loss of binding energy can be attributed entirely to loss of packing interactions. However, the further removal of the Cß methylene group (YG42) has an effect on binding energy (Gbinding = 4.7 kcal/mol) which is far beyond that predicted from losses in packing interactions. This may be explained either by the increase in conformational freedom of an otherwise rigid reactive site loop, resulting in an increase in the loss of entropy on binding, or by structural rearrangements in the area of the mutation.
The side chains of both Arg46 and Arg48 are almost completely buried; only N
1 and N2 of Arg48, and N1 of Arg46, are slightly exposed. Mutations of either Arg46 or Arg48 to Ala decreases kon 3–5-fold and increase koff 50-fold. In contrast to Tyr42, mutations of Arg46 and Arg48 affect the stability of the complex indirectly, since their side chains make no contacts to subtilisin in the crystal structure of the CI2–subtilisin BPN' complex (McPhalen et al., 1985).
Coupling energies
The coupling energy between Arg46 and Arg48 in complex with subtilisin BPN' can be calculated by an equation analogous to the interaction energy in eq. 2 (Jackson et al., 1994):
|
|
(nomenclature as in eq. 2). This gives a value of 0.95±0.18 kcal/mol, which is similar to the interaction between the two Arg side chains in stabilizing free CI2 (1.10±0.10 kcal/mol).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
We have prepared two different sets of mutants in CI2. In the first set (YA42 and YG42), interactions within CI2 as well as interactions with subtilisin are removed, while only intramolecular interactions are removed in the second set (RA46, RA48, RA46/RA48 and RC48). Both sets of mutations have a detrimental effect on the binding to subtilisin, but for different reasons. The loss of binding energy for the mutant YA42 can be ascribed entirely to loss of packing interactions with subtilisin. All the other mutants exert their effect indirectly, either by increasing the flexibility of the loop or by distorting it. In the absence of structural knowledge, the loss of binding energy would most likely be misinterpreted as the removal of an intermolecular contact. For example, the double mutant cycle in this study reveals an apparent coupling energy between Arg46 and Arg48 with regards to binding to subtilisin. In the absence of structural rearrangements, double mutant cycles report on specific interactions between side chains, and the coupling binding energy between Arg46 and Arg48 a priori suggests that the two residues cooperate in binding interactions with subtilisin, but structural data (McPhalen et al., 1985
) rule out such a conclusion. This emphasizes that it is necessary to combine mutagenic analysis with structural insight to reach a valid conclusion. The fact that the coupling energy between Arg46 and Arg48 is the same in the complex as in free CI2 does not in itself demonstrate that the two residues do not contact subtilisin, since the two coupling energies apply to two entirely different processes. For unfolding, it is the coupling energy of the two side chains in the native structure relative to that in the denatured state. For the association constants, it is the coupling between the two side chains in the native state of free CI2 and the native state of CI2 in complex with subtilisin. Any similarity in coupling energy is purely fortuitous; the only time one would expect identical coupling energies is for the process of denatured inhibitor binding to subtilisin to give a native CI2–subtilisin complex.
When a side chain participating in an intermolecular interaction is deleted by mutagenesis, the stage at which the intermolecular interaction is formed may be deduced from the effect on kon and koff. For instance, the removal of non-electrostatic interactions observed in the structure of the barnase–barstar complex predominantly increases koff, while kon is only lowered to a small extent (Schreiber et al., 1995). This is also observed for the complex between human growth hormone and its receptor (Cunningham et al., 1993). This indicates that these interactions are formed not in the transition state of the initial binding step (measured as kon). Therefore, these interactions are also the first to be broken in the dissociation step. A similar conclusion applies to the mutant YA42 in CI2, where koff is increased 40-fold, while kon decreases by less than 50%. Obviously such a deduction cannot be made for the five other mutations where no intermolecular interactions are removed. One can only make the general conclusion that either the rigidity or structure of the extended loop, maintained via conformational restrictions on the polypeptide backbone and via extensive side chain interactions, is crucial not for the association step but for slowing down the release of CI2 from subtilisin BPN'. The situation is analogous to the study of protein folding by 
-value analysis (Fersht et al., 1992). Here the formation of side-chain interactions during folding is probed by selective removal of interacting groups and measuring the effect on the kinetics of folding, but for a meaningful interpretation it is essential that the mutation does not change the folding pathway, i.e. the probe must not perturb the system.
|
Acknowledgments |
|---|
We thank Ben Davis and Ole Kristensen for helpful discussions and graphic analysis. D.E.O. was supported by a predoctoral fellowship from the Danish Natural Science Research Council.
|
Notes |
|---|
1 Present address: Department of Biochemistry, Chemistry Centre, University of Lund, PO Box 124, S- 22100 Lund, Sweden
2 To whom correspondence should be addressed
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Blow,D.M. (1974). In Fritz,H., Tschesche,H., Greene,L.J. and Truscheit,E. (ed.), Proteinase Inhibitors, Bayer Symposium V. Springer-Verlag, Berlin, pp. 677–678.
Bode,W. and Huber,R. (1992) Eur. J. Biochem., 294, 433–451.
Carter,P.J., Winter,G., Wilkinson,A.J. and Fersht,A.R. (1984) Cell, 38, 835–840.[Web of Science][Medline]
Cha,S. (1975) Biochem. Pharmacol., 24, 2177–2185.[Web of Science][Medline]
Clarke,J. and Fersht,A.R. (1993) Biochemistry, 32, 4322–4329.[Medline]
Cunningham,B.C. and Wells,J.A. (1993) J. Mol. Biol., 234, 554–563.[Web of Science][Medline]
Fersht,A.R., Matouschek,A. and Serrano,L. (1992) J. Mol. Biol., 224, 771–782.[Web of Science][Medline]
Fersht,A.R., Shi,J.P., Knill-Jones,J., Lowe,D.M., Wilkinson,A.J., Blow,D.M., Brick,P., Carter,P., Wae,M.M.Y. and Winter,G. (1985) Nature, 314, 235–238.[Medline]
Gill,S.C. and von Hippel,P.H. (1989) Anal. Biochem., 182, 319–326.[Web of Science][Medline]
Harpaz,Y., elMasry,N.F., Fersht,A.R. and Henrick,K. (1994) Proc. Natl. Acad. Sci. USA, 91, 311–315.
Hidalgo,P. and MacKinnon,R. (1995) Science, 268, 307–310.
Horovitz,A. and Fersht,A.R. (1990) J. Mol. Biol., 214, 613–617.[Web of Science][Medline]
Itzhaki,L.S., Otzen,D.E. and Fersht,A.R. (1995) J. Mol. Biol., 254, 260–288.[Web of Science][Medline]
Jackson,S.E. and Fersht,A.R. (1994) Biochemistry, 33, 13880–13887.[Medline]
Jackson,S.E., Moracci,M., elMasry,N., Johnson,C.M. and Fersht,A.R. (1993) Biochemistry, 32, 11259–11269.[Medline]
Jorgensen,W. and Severance,D. (1990) J. Am. Chem. Soc., 112, 4768–4774.
Longstaff,C., Campbell,A. and Fersht,A.R. (1990) Biochemistry, 29, 7339–7347.[Medline]
McPhalen,C.A. and James,M.N.G. (1987) Biochemistry, 26, 261–269.[Medline]
McPhalen,C.A., Svendsen,I., Jonassen,I. and James,M.N.G. (1985) Proc. Natl. Acad. Sci. USA, 82, 7242–7246.
Otzen,D.E., Rheinnecker,M. and Fersht,A.R. (1995) Biochemistry, 34, 13051–13058.[Medline]
Schechter,I. and Berger,A. (1967) Biochim. Biophys. Res. Comm., 27, 157–162.[Web of Science][Medline]
Schreiber,G. and Fersht,A.R. (1995) J. Mol. Biol., 248, 478–486.[Web of Science][Medline]
Shaw,G.L., Davis,B., Keeler,J. and Fersht,A.R. (1995) Biochemistry, 34, 2225–2233.[Medline]
Svendsen,I., Boisen,S. and Højgaard,J. (1982) Carlsberg Res. Comm., 47, 45–53.
Received May 13, 1998; revised September 23, 1998; accepted September 30, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Otzen Antagonism, non-native interactions and non-two-state folding in S6 revealed by double-mutant cycle analysis Protein Eng. Des. Sel., November 1, 2005; 18(11): 547 - 557. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. S. Radisky and D. E. Koshland Jr. From the Cover: A clogged gutter mechanism for protease inhibitors PNAS, August 6, 2002; 99(16): 10316 - 10321. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Germain, K. Merienne, S. Zinn-Justin, J.-C. Boulain, F. Ducancel, and A. Menez Molecular and Structural Basis of the Specificity of a Neutralizing Acetylcholine Receptor-mimicking Antibody, Using Combined Mutational and Molecular Modeling Analyses J. Biol. Chem., July 7, 2000; 275(28): 21578 - 21586. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Antil-Delbeke, C. Gaillard, T. Tamiya, P.-J. Corringer, J.-P. Changeux, D. Servent, and A. Menez Molecular Determinants by Which a Long Chain Toxin from Snake Venom Interacts with the Neuronal alpha 7-Nicotinic Acetylcholine Receptor J. Biol. Chem., September 15, 2000; 275(38): 29594 - 29601. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||