Two new parameters, I and C, are introduced for the quantitative evaluation of functional chimeras: I (impact) and C (context dependence) are the free energy difference and sum, respectively, of the effects on a given property measured in forward and retro chimeras. The forward chimera is made by substitution of a part “a” from ensemble A into the analogous position of homologous ensemble B (SB → A). The C value is a measure of the interaction of the interrogated position with its surroundings, whereas I is an expression of the quantitative importance of the probed position. Both I and C vary with the evaluated property, for example, kinetics, binding, thermostability, and so forth. The retro chimera is the reverse substitution of the analogous part “b” from B into A, SA → B. The I and C values derived from original data for forward and retro mutations in aspartate and tyrosine aminotransferase, from literature data for quasi domain exchange in oncomodulin and for the interaction of Tat with bovine and human TAR are evaluated. The most salient derived conclusions are, first, that Thr 109 (AATase) or Ser 109 (TATase) is an important discriminator for dicarboxylic acid selectivity by these two enzymes (I < −2.9 kcal/mol). The T109S mutation in AATase produces a nearly equal and opposite effect to S109T in TATase (C < 0.4 kcal/mol). Second, an I value of 5.5 kcal/mol describes the effects of mirror mutations D94S (site 1) and S55D (site 2) in the Ca2+ binding sites of oncomodulin on Ca2+ affinity. The second mirror set, G98D (site 1) and D59G (site 2), yields a smaller impact (I = −3.4 kcal/mol) on Ca2+ binding; however, the effect is significantly more nearly context independent (C = −0.6 versus C = −2.7 kcal/mol). Third, the stem and loop regions of HIV and BIV TAR are predominantly responsible for the species specific interaction with BIV Tat65–81 (I = −1.5 to −1.6 kcal/mol), whereas I = 0.1 kcal/mol for bulge TAR chimeras. The C values are from −0.3 to −1.2 kcal/mol. The analysis described should have important applications to protein design.
Chimeras, in which parts of one biological macromolecule are grafted onto another, are widely applied in structure/function analysis. The substituted pieces may vary from single amino acids to large domains. It has gradually been recognized that the quantitative effects that a given mutation evokes depend on the framework in which it is introduced. For example, the phenotype of a mutation X → Y, may depend on whether a neighboring residue is W or Z; that is, on the context of the background in which the mutation is introduced. Wells (1990) noted that the cumulative effect of mutations is generally additive where the two sites of substitution are well separated but frequently not so where they are spatially adjacent. The free energy of interaction between independent mutations can be addressed quantitatively by the double-mutant cycle method of Horovitz and Fersht (1990), who introduced Equation (1) to describe the free energy of interaction between two sites of mutation, A and B:
This method, while theoretically powerful, requires either very precise experimental data or that the interrogated ΔΔGint values be large (Pons et al. 1999).
The rapidly expanding accumulation of protein sequences has revealed a vast number of homologous proteins with substantial sequence identity and varying functions that are controlled by relatively few differences. The question addressed in this manuscript is: What is the relationship between the effect of a substitution SA → B in framework A and that of SB → A in a homologous framework B (Scheme 1)? That is, to what extent are the replacements context independent? Theory is presented for a quantitative evaluation of the context dependence of such sets of exchanged substitutions. The treatment is general and includes single mutations as well as domain swaps and, in principle, may encompass whole-molecule replacements in macromolecular assemblies. This novel approach is illustrated by analyses of three sets of data covering a wide range of chimera sizes: original data for single and double amino acid swaps in the active sites of aspartate (AATase)4 and tyrosine aminotransferases (TATase) that partially dictate dicarboxylic versus aromatic ligand specificity; literature data (Henzl et al. 1998) on quasi domain switching of two Ca+2 binding sites in oncomodulin; and a set of Tat-TAR interactions (Smith et al. 1998) that extend the treatment to macromolecular interaction.
The aminotransferase mutants were designed to elucidate the individual contributions of residue changes in an AATase hexamutant (HEX) that were successful in converting 75% of the specificity of AATase for dicarboxylic substrates to that of TATase for aromatic compounds (Onuffer and Kirsch 1995). The loci of the six changes in AATase to the corresponding residue in TATase are clustered. Four of them decrease the polarity of the active site and are not the subject of this communication. The other two, T109S and N297S, are of amino acids that contact PLP (Scheme 2). The context dependencies of these mutations are explored in AATase with a set comprising T109S, N297S, and T109S/N297S and with the retro TATase mutations S109T, S297N, and S109T/S297N. According to the crystal structure of an aromatic inhibitor bound to HEX (Malashkevich et al. 1995), the mutation T109S is likely responsible for a rearrangement of a hydrogen bond network that enables reorientation of Trp 140 to accommodate larger hydrophobic compounds. The N297S mutation allows binding of the phenylalanine analog, hydrocinnamate (Hca), without the otherwise necessary expulsion of an active site H2O molecule that usually forms hydrogen bonds to dicarboxylic substrates.
Dissociation constants of maleate and hydrocinnamate (Hca) complexes formed with wild-type and mutant AATases and TATases
The single AATase mutation, T109S, very sharply reduces the affinity of the enzyme for the dicarboxylic acid, maleate, and somewhat enhances that for the nonpolar Hca (Fig. 1; Table 1). The effect of N297S is manifested entirely in a reduced KD for Hca. The double mutant combines the phenotypes of the two singles, having the negligible affinity for maleate that the single T109S exhibits and the lower KD (Hca) of the N297S mutant. The reverse mutation S109T in TATase results in a sharply decreased value of KD (maleate), while there is no measurable effect on KD (Hca; Fig. 2; Table 1). The major effect of the S297N mutation is a threefold increase in KD (Hca). The combined mutant S109T/S297N is characterized by a KD (maleate) that is fourfold greater than that of S109T and a KD (Hca) that is between the two values for the single mutants and only twofold greater than that of the wild-type (WT) complexes.
Kinetics of the reactions with L-Asp and L-Phe
The construct, T109S, exhibits a 10-fold decrease in kcat/KM (Asp), exclusively because of an increase in KM (Fig. 3; Table 1). N297S AATase displays a threefold decrease in kcat/KM (L-Asp) relative to WT. Addition of N297S into the T109S context yields an enzyme (T109S/N297S TATase) that exhibits a slight decrease in KM from T109S alone and no significant difference in kcat. Neither the single nor the double mutants exhibits saturation kinetics with [L-Phe] ≤ 40 mM. T109S and T109S/N297S AATase both have kcat/KM (Phe) values that are threefold greater than that of WT AATase, whereas that for N297S is similar to the WT. S109T made in the TATase framework has no effect on kcat/KM (Asp) relative to WT TATase (Table 1). This constant value is maintained through a reduced kcat value with a correspondingly equal decrease in KM for the single mutant. Both the S297N and S109T/S297N TATases have roughly the same kcat/KM value for aspartate, ∼25% that of WT TATase. The values of kcat/KM (Phe) decrease in the order of S109T, S297N, and S109T/S297N TATase.
Scheme 1 diagrams a formalism for the quantitative analysis of the context dependence of substitutions in macromolecules and in macromolecular assemblies. ΔΔGSA → B and ΔΔGSB → A′, the free energy differences effected by substitutions in the respective WT framework, shown in Scheme 1, are defined in Equations (2) and (3):
where “param” is a measurable kinetic or equilibrium constant. The substitutions are not limited to single amino acid replacements. Species A and B must share sufficient homology that the chimeras are functional, and the SB → A replacement must be the reverse of the SA → B change. The symbol SA → B means that a region of ensemble A is replaced by the corresponding part of B and SB → A is the retro substitution. Illustrative ΔΔG values are shown as the abscissas of Figures 4–7, Fig. 5., Fig. 6., Fig. 7..
The context dependence of a substitution of interest is defined by Equation (4):
The value of C approaches 0 when the effect of a mutation is context independent. That is, substitution A → B results in a nearly equal and opposite effect as the mutation B → A in their respective frameworks. This implies that the environment does not significantly influence the quantitative effect of the replacement. Nonzero C values arise when the effect of such a perturbation is context dependent.
The parameter C alone is insufficient to complete the analysis because a small C value may result from any set of equal and opposite ΔΔG values irrespective of their magnitudes, including the trivial case of ΔΔGi→j = 0. This is addressed by the additional parameter, I. The impact, I, of a mutation is the difference between the two ΔΔG values as defined in Equation (5):
If the substitution has a significant impact, I is large, and it is small if the effect is negligible. The latter case can result from forward and reverse substitutions that have similar effects in both magnitude and direction so that ΔΔGSA → B and ΔΔGSB → A cancel. The sign of the I value is arbitrarily dependent on the choice of reference chimera, and in most cases, it is the absolute value of I, |I|, that is important. Generally, ΔΔGSA → B will be opposite in sign from ΔΔGSB → A, realizing the expected effect of the importation of the targeted property into the receiving partner. However, there will be exceptions; for example, moving an amino acid from the AATase framework into that of TATase may make the recipient more TATase-like for a given property. This situation is readily apparent from an examination of the ΔΔG values and from the sign of the I value. For this reason, the vectorial values of I and C must be given. The I value is needed in addition to C because the free energy change associated with context dependence is not correlated with the magnitude of the impact. The relationship between the three parameters is given in the following algebraic rearrangements:
Interpretations of the possible combinations of I and C values are summarized in Table 2.
The I and C values, like their component ΔΔG values, will generally be expected to be additive where the substitutions are spatially remote (see preceding text). In these cases,
where the single subscripts, S1 and S2, denote the sites of the individual chimeric replacement sets, and the double subscript S1S2 represents the combined set.
The forward and retro mutations made to interconvert AATase and TATase provide an experimental data set to illustrate the application of the above described parameters to enzymatic catalysis. It will be shown that I and C values are a function of the interrogated parameter.
The dicarboxylate inhibitor maleate complexes of WT AATase and WT TATase exhibit KD values of 19 and 140 mM, respectively (Table 1). The largest absolute I value for maleate binding from the position 109 and 297 mutation sets is >2.9 kcal/mol for position 109 versus >0.6 kcal/mol for position 297 and >1.7 kcal/mol for the double mutants (Fig. 4). These results emphasize the dominant role of residue 109 in the discrimination exhibited by the two enzymes in the recognition of dicarboxylic ligands. The inhibition of position 109 TATase and AATase mutants by maleate is roughly context independent (C < 0.4 kcal/mol).
The amino acid in position 109 has less influence on the affinity for the aromatic ligand, hydrocinnamate, than does that at position 297 (I(109) > 0.2 kcal/mol, I(297) > 1.2 kcal/mol). Thus, the AATase position 297 is the more significant determinant of affinity for nonpolar side chains. S297N TATase and N297S AATase yield roughly context independent changes in affinity (C ≈ 0.1 kcal/mol). The changes in the dissociation constants of the double-mutant complexes with both Hca and maleate resemble the phenotype of the more dominant substitution.
In principle, conversion of a TATase residue to that found in AATase should endow that mutant with more AATase-like catalytic activity or substrate specificity, but S297N and S109T/S297N TATase are unexpectedly less active (kcat/KM) toward aspartate than is WT TATase, whereas the S109T construct exhibits a negligible effect on this property (Table 1; Fig. 5). It should be noted; however, that the kcat/KM (Asp) values for WT AATase and TATase differ by only 2.5-fold. Therefore, these observations report perturbations over a rather small range of free energies of activation. The T109S/N297S and T109S/N297S mutations do shift kcat/KM (Asp) values in the expected direction (Fig. 1; Table 1), but strong context dependence combined with the small variation in kcat/KM (Asp) makes a molecular interpretation difficult.
However, all of these mutants in TATase do suppress phenylalanine aminotransferase activity as expected. Both enzymes are coupled to the same αKG/L-Glu component of the aminotransferase reaction and, thus, must catalyze dicarboxylic amino and keto acid reactions with comparable efficiency, but the enzymes differ substantially in the ability to interconvert the aromatic substrates. The ratio of kcat/KM (Phe) for WT TATase versus WT AATase is, thus, large (8000-fold). While the C values of the chimeric constructs are not significantly different, as measured by the effects on kcat/KM (Asp) and kcat/KM (Phe), the I values for the latter are much larger and are in the predicted direction (Fig. 5). Interestingly, the mutations made in TATase have a more pronounced effect than the retro mutations made in AATase. Thus, mutations at position 109 and 297 are sufficient to compromise aromatic aminotransferase activity in WT TATase, but they do not suffice to introduce significant such activity in WT AATase. This context dependence suggests that a strong degree of cooperativity (nonadditivity) with the other four positions of the HEX construct (Onuffer and Kirsch 1995) is required to induce aromatic aminotransferase activity in AATase.
Quasi domain exchange—the Ca2+ sites of oncomodulin
Oncomodulin contains two Ca2+ binding sites, termed CD and EF. Each is found within a helix–loop–helix motif (Henzl et al. 1998; Fig. 6). Most of the interactions involved in the binding of Ca2+ are conserved between the CD and EF sites, which have KD values of 0.80 μM and 0.045 μM, respectively. Two significant exceptions are found in the side chains at positions 55 and 59 in the CD site, which differ from the ones at analogous positions, 94 and 98, in the EF site. The substitutions, S55D and D59G, convert the amino acids found in the CD to those found in the EF site (Henzl et al. 1998). The double mutant, S55D/D59G, in the CD site thus replaces the two most obvious differences to mimic a quasi EF site. The retro mutations, D94S, G98D, and D94S/G98D, were also made in the EF site to the corresponding amino acids in the CD site to elaborate further how the minimal differences in residues that directly contact Ca2+ account for the differential binding of this ligand.
The quantitative importance of the mirror mutations, D94S and S55D, which occupy the analogous positions in their respective CD and EF sites, is made apparent from ΔΔG, C, and I values. These mutations produce a large impact value of 5.5 kcal/mol that is context dependent (C = −2.7 kcal/mol; Fig. 6). Thus, substitution of Ser for Asp in the CD site decreases affinity for Ca2+ much more than the retro substitution increases the affinity in the EF site. Mutation at D59G and its mirror, G98D, results in a more context independent phenotype (C = −0.6 kcal/mol). The negative sign of the I value (−3.4 kcal/mol) highlights the fact that the D59G, G98D mutations, which were designed respectively to introduce EF affinity into the CD site and vice versa, have the opposite outcome. That is, D59G increases KD from 0.8 μM (WT) to 22 μM, and G98D decreases KD (Ca2+) from 0.045 μM to 0.004 μM. The latter value is lower than exhibited by either WT CD or WT EF complexes. Henzl et al. (1998) noted the overwhelming importance of the negative charge in the interactions with Ca2+. The C and I values of the set of double mutants most resemble that of the single-mutant set, D94S, S55D, in terms of the quantitative values of I and C. The ΔΔG values are nearly additive in these chimeras, that is, the D94S + G98D and D59G + S55D ΔΔG values each sum to those observed in the corresponding double mutant. In other words, each of the liganding side chains contributes independent of the other.
Tat/TAR-A peptide/RNA interaction
The formalism for analysis of context dependence can be extended to macromolecular interaction. The Tat protein of HIV and BIV enhances transcription by associating with a specific 5′ end of nascent RNA, termed the transactivation response region (TAR), rather than to DNA. The Tat-RNA complex interacts with additional transcriptional activators in vivo and increases the rate of transcription by several hundred–fold. Smith et al. (1998) determined the quantitative effects on transcriptional activation produced by the association of WT and chimeric (HIV/BIV) TARs with a BIV Tat peptide consisting of residues 68–81. The constructs are shown schematically at the bottom of Figure 7. Some of the chimeras were inexact because of insertions or deletions, as discussed in Smith et al. (1998). The values of ΔΔG, I, and C derived from their results are presented in Figure 7.
Association of WT HIV TAR with BIV Tat65–81 activates transcription only 7% relative to that observed for WT BIV TAR with BIV Tat65–81 (Smith et al. 1998). The latter activation is standardized as 100%. Large similar I values of −1.6 kcal/mol are calculated here for L1 (BIV TAR with HIV loop) and L2 (HIV TAR with BIV loop), and for S1 (BIV TAR with HIV stem) and S2 (HIV TAR with BIV stem), TAR constructs with replacements in the loop and stem, respectively (Fig. 7). Mutations L6 (HIV TAR with two BIV neck bases) and L9 (BIV TAR with two HIV neck bases) at the neck of the TAR loop result in a similarly large I value of −1.5 kcal/mol and, thus, also contribute to the species specificity of the transcriptional activator for the RNA binding site. Smith et al. (1998) noted that these regions are important for specificity in the interactions of the TARs with their cognate Tats. L1 and L2 along with L6 and L9 are sets of mutations that have large impacts in the predicted directions but differ in degree of context dependence. The former set (L1 and L2) exhibits a high degree of context dependence (C = −1.2 kcal/mol); whereas the effect of the latter set of replacements is more nearly context independent (C = −0.3 kcal/mol). The large C combined with the large I value for the L1 and L2 pair means that a simple qualitative statement about the importance of this region in association can not be made. Replacement of the BIV TAR loop with the corresponding human sequence does indeed grossly compromise the extent of association (L1), but replacement of HIV TAR loop with the BIV TAR sequence (L2) has almost no effect (Fig. 7). However, the L6/L9 pair has a small C value because the forward and retro chimeric constructs yield nearly equal and opposite effects on the ΔΔGassoc. Therefore, the phenotype of the mutation in the TAR loop is species specific and involves local interactions of the microenvironment. Transcriptional activation by BIV Tat is also sensitive to mutation of the bases that clamp the neck of the TAR loop but does not rely as much on interactions with other regions for this large impact to be observed, that is, there is little context dependence. Bulge hybrids, B1 and B2, result in the lowest I value for the sets of converse mutants studied, quantitating the conclusion of Smith et al. (1998) that the bulge does not play a large role in this RNA–peptide interaction.
The new parameters, the I and C values defined in Equations (4) and (5), allow for quantitative analysis of the thermodynamic and kinetic effects introduced by chimeric constructs of an arbitrary degree of complexity. They enable precise comparisons of related systems. The I and C values obtained for the tyrosine and aspartate aminotransferase frameworks demonstrate clearly and quantitatively that these parameters are functions not only of the structure of the construct but, in addition, of the kinetic or thermodynamic property addressed. The context sensitivity of different regions within a single framework is analyzed by I and C values for oncomodulin variants that were engineered to interconvert critical residues of two comparable Ca2+ binding sites. The evaluation of Tat interactions with chimeras of HIV and BIV TAR's illustrates the use of I and C values as informative indices of the context dependence of mutational effects in macromolecular recognition. The evaluation of I and C parameters should prove to be generally useful in rational protein design based on homology modeling.
Materials and methods
AATase and TATase mutants were constructed by PCR-based site-directed mutagenesis. PCR reactions were performed in 100 μL volumes and rotated through 25 consecutive cycles of the set temperatures: 95°C, 55°C, and 75°C. Mutagenic fragments, whose ends were digested by restriction enzymes were ligated to plasmids, pUC 118 (AATase) or pUC 119 (TATase). The vectors were cut by the same enzymes. Ligation mixtures were transformed into DH5α cells, and DNA was extracted with Promega Wizard Preps. Silent restriction sites were introduced into DNA constructs to screen for the mutagenic inserts, which were subsequently verified by automated sequencing.
AATase and TATase were overexpressed in Escherichia coli MG204 cells and purified according to the protocol of Herold and Kirschner (1990), with modifications by Onuffer and Kirsch (1995). HO-HxoDH was purified as described in Luong and Kirsch (1997).
L-aspartate transamination was monitored by MDH coupled assay in the presence of saturating concentrations of the cosubstrate, α-ketoglutarate (αKG). L-phenylalanine aminotransferase activity was followed by coupling the reaction to the HO-HxoDH catalyzed oxidation of NADH, which was monitored as a decrease in absorbance at 340 nm in a Molecular Devices SPECTRAmax 340 and SPECTRAmax 250 spectrophotometer. See Table 1 for details. Background rates in the presence of coupling enzymes were subtracted from those recorded after addition of the aminotransferase.
Absorbance changes at 430 nm were followed as a function of [maleate] or [Hca]. Samples were preincubated at 25°C before the addition of enzyme. The spectra of the samples in 200-μL volumes in 96 well plates were read in a Molecular Devices SPECTRAmax 340 or a SPECTRAmax 250 spectrophotometer.
Values of KD were calculated from Equation (1), where A, Ao, and A∞ are the absorbance monitored, the initial absorbance in the absence of inhibitor, and the final absorbance at saturating concentration of ligand [L], respectively:
Errors for kcat/KM were determined from Equation (2), a transformation of the Michaelis Menten equation:
(L. Feng and J.F. Kirsch, unpubl.).
Limits for those parameters that could not be determined accurately were obtained by comparing the theoretical curves calculated for these limiting values with the experimental data (panels A and B in Figs. 1,2).
We thank Andrew Eliot, Steven Rothman, Wendy Shaffer, and James A. Wells for their insightful discussions; Elizabeth Kirsch and Joseph Hou for assistance in mutagenesis, protein purification, and characterization; and J.N. Jansonius for providing financial support and encouragement for part of this work. This work was supported by NIH Grant GM-35393. T.N.L. was an University of California undergraduate McNair Scholar and was supported, in part, by the Howard Hughes Medical Institute funded Biology Fellows Program.
The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.