Additivity of free energy changes on mutations

Five double mutants and one triple mutant were constructed in order to evaluate quantitatively the thermodynamic stability of mutations most commonly found in cancer that also contained a second- and/or third-site suppressor mutation (Table ​(TableI).I). All multiple mutants were destabilized relative to wild-type p53 except the double mutant G245S/N239Y. The incorporation of the secondary mutation into the cancer mutants V143A and G245S gave double mutants (V143A/N268D and G245S/N239Y) that were 1.09 and 1.22 kcal/mol, respectively, more stable than the single cancer mutants, but introduction of the suppressor mutation T123A into the cancer mutant R249S had no appreciable effect on stability. The mutants R249S/T123A and R249S/T123A/H168R were 0.22 and 0.40 kcal/mol, respectively, less stable than R249S, which was destabilized by 1.69 kcal/mol, relative to wild-type p53. The most interesting of all suppressor mutants was H168R. The stability of the double mutant R249S/H168R was nearly identical to that of the cancer mutant R249S (destabilized by 1.72 and 1.69 kcal/mol, respectively), yet the stability of the single suppressor H168R was 1.06 kcal/mol lower than that of the cancer mutant.

The sums of the individual changes in free energy on mutation are listed in Table ​TableI.I. The change for a multiple mutant should be equal to the sum of energy changes of the individual parental single mutations if the mutations do not interact with each other (Carter et al., 1984). This has been frequently observed for residues that are far apart in the structure and do not interact directly, or interact indirectly via structural reorganization. If, however, the residues interact directly or indirectly via structural changes, then the effects of mutations should not be additive. Of the six multiple mutants subjected to this analysis, four (R249S/T123A, V143A/N268D, T123A/H168R and G245S/N239Y) exhibited near additive effects and two (R249S/H168R and R249S/T123A/H168R) had non-additive effects. Most of the mutants are far apart, but H168R and R249S are nearly in contact (Figure ​(Figure22).

DNA binding

In order to evaluate the restoration of wild-type p53 function to cancer-associated mutants, we measured the binding of a specific DNA sequence to all mutant p53 proteins. A biotinylated double-stranded gadd45 30mer oligonucleotide was immobilized into a streptavidin sensor chip. A fixed volume of p53 protein was injected at a constant flow rate and the binding was monitored by the BIAcore instrument (Table ​(TableII).II). The concentration of wild-type p53 core domain for 50% binding was estimated to be 100 nM at 20°C in 50 mM Tris–HCl pH 7.2 and 5 mM dithiothreitol (DTT).

The cancer mutants have reduced affinity for binding of the gadd45 promoter, with the structural mutant R249S having no detectable binding activity at 20°C (Table ​(TableII).II). The two cancer mutants (V143A and G245S) had concentrations for 50% binding of 150 and 160 nM, respectively. The suppressor mutants as a group also exhibited reduced DNA binding to gadd45 promoter. The concentrations for 50% binding ranged from 160 (N268D) to 190 nM (H168R). It should be noted that H168R was the only mutation found in cancers in the group of four suppressors tested. The other two mutants (T123A and N239Y) had values of 180 and 190 nM, respectively, for concentrations for 50% binding. The affinities of the multiple mutations ranged from no detectable binding (R249S/T123A) to 170 nM (R249S/T123A/H168R). The most remarkable observation was that of the double mutant R249S/H168R, which manifested the most dramatic rescue effect in terms of DNA binding relative to the two single mutants R249S and H168R, both of which were found in human cancers. While R249S has no detectable binding, the introduction of suppressor mutant H168R into R249S rescued the DNA-binding affinity to a concentration of 170 nM for 50% binding of R249S/H168R. On the other hand, the suppressor mutant T123A had no apparent effect on the DNA binding of R249S (Table ​(TableII).II). The double mutant R249S/T123A and R249S had virtually no affinity for gadd45 promoter. The triple mutant R249S/T123A/H168R and the double mutant R249S/H168R had near identical DNA binding to gadd45 promoter (Table ​(TableII).II). The effect of the suppressor mutation N268D on the DNA binding of V143A was negligible. Incorporation of the suppressor mutant N239Y into the cancer mutation G245S had a beneficial effect and resulted in a small increase of DNA binding of the double mutant G245S/N239Y.

Suppressor mutant N268D rescues cancer mutant V143A by stabilizing the global fold of the p53 core

Val143 is located in strand S3 of the β–sandwich (Table III). This residue is deeply buried and participates in the packing of the hydrophobic core of the fold (Cho et al., 1994). This variant represents an example of the group of structural mutants that is temperature sensitive. DNA-binding and transcriptional activity of V143A were not detected at physiological temperature but were observed at lower temperatures (Zhang et al., 1994). Truncation of two methyl groups from Val to Ala143 dramatically destabilized the fold by 3.34 kcal/mol (Table ​(TableI).I). The V143A mutation introduces structural changes in almost all residues in the β–sandwich and hence it is likely to affect the packing of the hydrophobic core (Wong et al., 1999). Similar substitutions in other monomeric proteins have been reported to cause destabilization by 0.3–4.7 kcal/mol, depending upon the structural contexts and the packing density in the vicinity of the mutated residue (Shortle et al., 1990; Sauer and Lim, 1992; Serrano et al., 1992; Buckle et al., 1993; Jackson et al., 1993). The decreased stability of V143A causes its temperature sensitivity. At 37°C the V143A mutant is estimated to be ~80% denatured (A.N.Bullock and A.R.Fersht, unpublished data). At lower temperatures the V143A variant may retain a sufficient population of folded species, allowing the protein to be at least partially functional. The V143A protein gained 1.1 kcal/mol in stability and retained the DNA binding by having the second-site suppressor N268D (Tables ​(TablesII and ​andII;II; Figure ​Figure1).1). The enhanced stability of V143A/N268D favourably shifts the equilibrium towards the native state relative to V143A, thus providing a mechanism for restoring the function of p53 mutant by second-site suppressor mutation.

The suppressor, N268D, was 1.21 kcal/mol more stable than wild-type p53 (Table ​(TableI).I). Residue Asn268 is located in strand S10 of the β-sandwich (Figure ​(Figure2;2; Table III) and its substitution to Asp was suggested to participate in a hydrogen-bonding network that might keep the two sheets of the β-sandwich together and, therefore, result in a more stable fold (Brachmann et al., 1998).

Suppressor N239Y restores wild-type stability to G245S

The cancer-associated hot spot mutation G245S is destabilized by 1.22 kcal/mol relative to wild-type p53, and has free energy of unfolding in the absence of denaturant of 8.55 kcal/mol (Table ​(TableI).I). This residue is located in the L3 loop of the β-sandwich and is partially buried in the core domain (Figure ​(Figure2;2; Table III). In the wild-type structure, it adopts an unusual backbone conformation, which facilitates the formation of two hydrogen bonds between the backbone amide of Gly245 and the backbone carbonyl of Cys242 (one of the three Cys residues involved in zinc binding), and between the backbone of Gly245 and the guanidinium group of Arg249 (Cho et al., 1994). Structural changes are found extensively in the L2 and L3 loops, suggesting distortion of the conformation required for DNA binding (Wong et al., 1999), consistent with the decreased DNA binding observed in this study (Table ​(TableII).II). The mechanism of inactivation of p53 by G245S is more complex. It is probably caused by a combination of decreased stability of the fold and distortion of the conformation critical for DNA binding. The N239Y mutation recovered the stability of G245S back to wild-type level and improved the DNA binding, suggesting a plausible explanation for how N239Y acts as a suppressor.

Suppressor mutation H168R may recover the DNA binding of R249S by partially restoring wild-type conformation in loops L2 and L3

Both R249S and H168R are cancer-associated mutants of p53. In the wild-type structure, Arg249 is a well-buried residue (Figure ​(Figure2;2; Table III) whose replacement with a number of other amino acids results in a complete loss of function (Cho et al., 1994). The mutation of R249S destabilized the protein by 1.69 kcal/mol and abolished the DNA binding (Tables ​(TablesII and ​andII).II). The suppressor mutation H168R destabilized wild-type protein by 2.75 kcal/mol and decreased the DNA binding (Tables ​(TablesII and ​andII).II). These effects are probably caused by the introduction of a guanidinium group that may affect the packing interactions between the L2 and L3 loops. However, the introduction of H168R to the cancer ‘hot spot’ R249S resulted in a double mutant with dramatically rescued DNA binding for gadd45 promoter. The insertion of H168R into R249S did not further destabilize the mutant (Table ​(TableI).I). To understand better how H168R salvages the DNA binding of R249S cancer mutant, the structure of R249S/H168R was compared with that of R249S and wild-type p53 core by NMR spectroscopy. Structural changes in R249S involved L2–L3 loops and terminal residues of strands 4, 9 and 10 (Wong et al., 1999). Inter-proton distances and backbone dynamics analysis revealed that the L3 loop is locally unfolded (B.DeDecker and A.R.Fersht, unpublished data), suggesting that the replacement of Arg249 by Ser disrupts the conformation of L3 required for DNA binding. Here, we have shown that in the R249S/H168R mutant the second mutation H168R reverted some of the chemical shift changes in L2 and L3 introduced by R249S. Taken together, our data strongly suggest that H168R restores, at least partially, the wild-type conformation of L2 and L3 that is necessary for DNA binding. A partially recovered structure of R249S/H168R mutant may either (i) adopt a conformation that resembles a hybrid between the wild-type and R249S structure or, more likely, (ii) represent an ensemble of structures that sample active (wild-type-like) conformations and inactive (R249S-like) conformations. The guanidinium group of Arg249 in the wild-type structure plays an important role in stabilizing the structure of L2–L3 loops by salt bridging to the carboxylate group of E171 in L2 and by hydrogen bonding to backbone oxygens of G245 and M246. The side chains of 168 and 249 are in close proximity. It is likely that the mutation of H168R introduces a guanidinium group that mimics the role of Arg249 in wild type and hence restores the conformation of L2–L3 that is required for DNA binding.

T123A does not act as a suppressor for stability of R249S core domain

Mutation of T123A does not affect the stability of the core domain significantly (Table ​(TableI).I). On the other hand, it decreases the DNA-binding affinities (Table ​(TableII).II). T123A is located in the L1 loop of the loop–sheet–helix motif (Figure ​(Figure2).2). In the native structure Thr123 is 46% solvent exposed (Table III). The substitution to Ala resulted in a protein that is only marginally less stable (0.13 kcal/mol) than wild-type p53. Structural changes induced by T123A are localized in the loop–sheet–helix motif, as shown by NMR spectroscopy (Figure ​(Figure4B).4B). Local structural changes in the loop–sheet–helix motif probably compromise the DNA-binding activities of p53 core domain, consistent with the impaired DNA binding (Table ​(TableII).II). In contrast to the H168R mutation, T123A does not introduce structural changes in L2 and L3 loops. These findings, along with the observed lack of DNA binding for R249S/T123A (Table ​(TableII),II), suggest that T123A does not restore the conformation required for DNA binding. Using a full-length p53 construct, Brachmann et al. (1998) have shown that both T123A and H168R mutations were required to suppress R249S and yield a functional p53 molecule in a yeast-based assay. From the present studies of p53 core domain, it is unclear how T123A plays a role as a suppressor for the R249S mutation. Modelling studies of two p53 core domain molecules bound to DNA suggest that Thr140, Glu198, Gly199 and Glu224 participate in the dimer interface formation (Cho et al., 1994). T123A co-localizes with these residues constituting the protein–protein interface. It might be possible that T123A mediates its ‘suppressor’ effect in full-length p53 via domain–domain interactions.

Compensatory and additivity effects of multiple mutations on the stability of p53 core domain

During molecular evolution, compensatory neutral mutations have evolved as a salvage mechanism against deleterious changes in the functional properties of proteins (Kimura, 1991; Jucovic and Poteete, 1998). These are defined as a pair of substitutions at different sites in a protein that have a detrimental effect individually but restore normal function in combination. This type of effect is reported to occur between amino acid residues that are in close proximity to each other in the native (folded) state of a protein (Kimura, 1991). Multiple mutations involving two or more substitutions may also have a cumulative effect, the net result of which equals the sum of the individual effects of the mutations, i.e. additivity effect (Wells, 1990; Serrano et al., 1993). In the former case, while two or more individual amino acid replacements may have mutually compensatory effects that revert the abnormal effect of the individual mutations, the effect on global stability or function may remain non-additive (Wells, 1990; Mateu and Fersht, 1999).

The thermodynamic analysis of multiple mutants of human p53 core domain (Table ​(TableI;I; Figure ​Figure2)2) showed that these mutants fall into two different categories. The single (suppressor) mutation, H168R, with the greatest compensatory effect for functional repair (but not for stability) of R249S, falls in the group that represents the non-additivity effect. The sum of the stability of single mutants (H168R and R249S), 4.46 kcal/mol, was not equal to the stability of the double mutant (R249S/H168R), 1.72 kcal/mol. This non-additivity is in agreement with the NMR structural observations for the two residues in R249S/H168R, which are in very close proximity though not in direct contact (Figures ​(Figures22 and ​and4A).4A). In the wild-type structure, N of His168 is 3.16 Å away from the NH group of Arg249. Similarly, the triple mutant R249S/T123A/H168R was destabilized by 1.91 kcal/mol compared with the expected additivity value of 4.57 kcal/mol. Here T123 and R249 are ~16 Å apart while T123 and H168 are ~24 Å apart. The rest of the double mutants exhibited near additive effects on stability, consistent with the observed distances between the residues concerned.

Equilibrium urea denaturation

Urea-mediated equilibrium unfolding of p53 core domain was performed at 10°C in 50 mM sodium phosphate buffer and 5 mM DTT at pH 7.2 using an Aminco-Bowman luminescence spectrofluorimeter with excitation at 280 nm and emission scans between 300 and 370 nm as described (Bullock et al., 1997). DTT was included in all solutions in order to keep the 10 cysteine residues reduced. For each data point of the denaturation curve, 100 μl of 18 μM protein in the above buffer were added to 800 μl of the corresponding denaturant solution. Samples were incubated overnight at 10°C constant temperature and analysed at 10°C in 1 ml thermostatted cuvettes. The temperature of each sample was monitored throughout the measurements, using a thermocouple. All the proteins unfolded reversibly at 10°C and followed two-state co-operative transitions from native to denatured state. Experimental data were fitted to a two-state transition curve for which the fluorescence of the folded and unfolded states is dependent on the denaturant concentration according to equation (1):

F=(α_N+β_N[D]) + (α_D+β_D[D])exp{m([D] – [D]_50%)/RT}1 + exp{m([D] – [D]_50%)/RT}(1)

where F is the fluorescence at a given concentration of urea, [U], α_N and α_D are the intercepts, and β_N and β_D are the slopes of the baselines at low and high denaturant concentration, respectively (Clarke and Fersht, 1993). The concentration of denaturant at which half of the protein is denatured is denoted as [D]_50%, R is the gas constant, T is the temperature in K, and m is a constant that is proportional to the increase in solvent-accessible surface area between the native and denatured state of the protein (Pace, 1986). The data were fitted to the above equation by non-linear least squares analysis using the KALEIDAGRAPH program (Abelbeck Software, Reading, PA), which gave the values of the linear slope of the transition from native to denatured state, m, and [D]_50%, as well as their standard errors.

DNA-binding studies

The capacity of wild-type p53 DNA-binding domain and the mutant proteins to bind dsDNA was tested using surface plasmon resonance binding experiments at 20°C, using immobilized gadd45 promotor on the surface of a streptavidin-coated chip (SA5) and a BIAcore 2000 Pharmacia Biosensor (AB) as described (Nikolova et al., 1998). The sequence of the gadd45 promoter is:

Gadd45: 5′-biotin-GTACAGAACATGTCTAAGCATGCTGGGGAC-3′

Gadd45: 3′-CATGTCTTGTACAGATTCGTACGACCCCTG-5′

The relative change in resonance units was monitored over a period of 500 s. The changes in resonance units upon DNA binding were plotted as a function of protein concentration. The experimental data were fitted to a two-state transition equation using the KALEIDAGRAPH program. The midpoint of the transition curves was used as an estimate of the affinity of p53 core domain for gadd45 promoter.