Proc. Nadl. Acad. Sci. USA Vol. 88, pp. 1706-1710, March 1991 Biochemistry
Energetics of repacking a protein interior (gene V protein/protein stability/hydrophobic effect/protein engineering)
WARREN S. SANDBERG AND THOMAS C. TERWILLIGER Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, IL 60637
Communicated by David Eisenberg, November 20, 1990 (received for review September 5, 1990)
ABSTRACT To test whether interactions in the hydrophobic core of a protein can be adequately modeled based on the properties of a liquid hydrocarbon, we measured the unfolding free energies of the wild-type bacteriophage fl gene V protein and 29 mutants with apolar substitutions at positions 35 and 47. Stability changes arising from identical mutations at these two buried sites are quite different, suggesting that one site is more rigid than the other. Reversals of residues at positions 35 and 47 confirm that their environments are distinct. Mutants containing weakly polar residues at these two sites suggest that the protein interior is more polar than a liquid hydrocarbon. Interactions between residues at the two sites appear to be minimal. These observations are compatible with a view of protein interiors that incorporates properties of liquid hydrocarbons but also includes polar interactions and a sitedependent "packing energy" associated with changes in internal structure.
polar side chains, and from the weakly polar side chains such as phenylalanine and methionine (4, 7, 16). Considering the tightness of packing and variations in polarity, organization, and density found inside proteins, the apolar liquid model may not be sufficient to describe the effects of interior amino acid substitutions on protein stability. For example, amino acid substitutions that increase interior hydrophobicity do not necessarily increase protein stability (17-19), and the deleterious effects of cavities created when large residues are replaced by small ones have been noted (11, 12). Our present goal is to study the effects of mutations in the hydrophobic core of the bacteriophage fl gene V protein in order to make an assessment of the values of simple models of a protein interior, to evaluate the energetics of changing the arrangement of apolar groups inside a protein, and to judge the polarity of a protein hydrophobic core.
The structures and stabilities of proteins are thought to be strongly dependent on the arrangements and compositions of their hydrophobic cores (1-3). Interior side chains must efficiently fill the space between strands of the polypeptide (4), and shielding of nonpolar atoms from solvent water contributes significantly to the stabilization of folded proteins through the hydrophobic effect (1, 5). Beyond these general observations, however, our understanding of the properties of protein interiors is quite limited. For example, many combinations of nonpolar side chains can be accommodated in the interior of a protein without severe loss of function (6), but prediction of the stabilities of such mutants is beyond our grasp, the penalty in terms of stability for distortion of the protein interior is unknown, and the magnitudes of polar and weakly polar interactions in protein interiors are only approximately known (7). The contribution of interior amino acid side chains to protein stability is often assessed by treating the protein interior as an apolar liquid (8-10). Preferences of apolar side chains for the protein interior are estimated from their relative solubilities in water and various apolar liquids. In experiments on T4 lysozyme, barnase, and tryptophan synthetase (11-14), successive reductions in the size of apolar residues at buried sites lead to reductions in stability that are correlated with the corresponding reductions in residue hydrophobicity. However, protein interiors differ from apolar liquids in many ways. Interior residues are more tightly packed than molecules in apolar liquids, with densities approaching those of amino acid crystals (5, 15). Many atoms in proteins, such as those in the main chain of the polypeptide, are held relatively rigidly, and others, such as those of many side chains, have more flexibility. Additionally, local variations in polarity result because protein interiors contain polar groups from the polypeptide backbone, from buried
METHODS Isolation of Mutant Gene V Proteins. Derivatives ofplasmid p1T18 (20) encoding mutant gene V proteins were constructed as described (21). Escherichia coli, strain K561, was transformed with these pT118 derivatives, and cell growth and protein purification will be described elsewhere (34). We use the one-letter code to describe mutants; the replacement of valine at position 35 with leucine is denoted by V35L. Estimation of Protein Stability. Guanidine hydrochloride (Gdn HCl) induces reversible, cooperative denaturation of the gene V protein that can be monitored using the gene V protein circular dichroism at 229 nm (34). Stabilities are expressed as free energy changes upon unfolding in the absence of denaturant (AG'UH2O) and in the presence of 2.0 M Gdn HCl (AG0U 2M). Cm is the concentration (in M, mol/ liter) of denaturant at which half the polypeptide chains are unfolded, and m is the dependence of AG0u on (Gdn HCI]. Stabilities are estimated by fitting the unfolding data to a two-state model (22) as described (19) except that the dependence of circular dichroism on [Gdn HCI] at nondenaturing concentrations of denaturant is taken to be identical to that of the wild type (WT) for proteins with Cm 95% buried in the crystallographic model and appear to be in direct contact with each other (26). Replacements of Val-35 and Ile-47 allow comparison of the properties of the protein interior at neighboring sites, and mutants containing simultaneous substitutions at the two sites allow examination of the interaction between the two residues (27). Val-35 and Ile-47 were replaced with all combinations of glycine, alanine, valine, methionine, leucine, isoleucine, phenylalanine, and tryptophan. Additionally, the three mutants containing cysteine at one or both sites were conTable 1. Stabilities of gene V proteins measured in vitro M Residue AG U,2M, AAG0u,2M, Cm, kcal/ AG0U.H20, M mol'M kcal/mol kcal/mol kcal/mol 35 47 V -4.5 4.6 -4.4 A 1.50 13.6 -4.5 15.2 6.1 A L 1.85 -2.9 6.9 -2.2 I A 2.01 -3.9 14.6 A M 1.47 -4.2 12.9 4.6 -4.4 F -4.4 14.2 5.4 A 1.69 -3.6 -4.9 11.6 1.9 -7.1 C C 0.99 I 15.8 7.7 -1.4 C 2.21 -4.1 V A 0.87 -4.3 10.5 2.0 -7.0 V -4.5 C 1.33 12.8 3.8 -5.2 V V 6.5 1.91 -3.9 14.3 -2.5 V L -3.7 15.9 8.4 2.43 -0.6 I V 2.61 -3.6 16.3 9.0 V M 2.02 -3.9 14.6 6.9 -2.1 V F -4.0 15.1 7.1 2.07 -1.9 L V -4.2 12.3 1.32 4.0 -5.0 L L 1.70 14.1 5.5 -4.3 -3.5 I L 1.90 -4.5 15.3 6.4 -2.7 L M 1.37 -5.0 13.7 3.7 -5.3 L F 1.60 -4.3 13.7 5.1 -3.9 I V -4.1 1.80 14.2 6.0 -3.0 I L -4.2 2.27 16.2 7.9 -1.1 I I 2.44 15.7 -3.6 8.4 -0.6 I M 1.87 -4.2 14.7 6.3 -2.8 I F 2.05 -4.4 15.8 7.0 -2.0 M L 2.14 -4.1 15.5 7.4 -1.6 I M 2.33 15.2 -3.6 8.0 -1.0 M M 1.65 -3.8 13.1 5.5 -3.5 M F 1.99 -4.7 16.1 6.8 -2.3 F L 1.51 12.7 -3.9 4.9 -4.1 I F 1.76 -3.8 13.6 5.9 -3.1 Amino acids at positions 35 and 47 are given by their one-letter codes and values of Cm and m are listed (see text). The WT is denoted by bold type. Free energy changes upon unfolding in the absence of denaturant (AG0U H o) and in the presence of 2.0 M Gdn-HCl (AG0U,2M) and the difrerences in AG'U.2M between the mutants and WT (AAG0U,2M) are reported in kcal/mol of dimeric protein. Destabilized mutants have negative values of AAG0u,2M, and the magnitude corresponds to making the same change twice (once for each monomer). Stabilities of the WT and the mutants V351, 147V, and V351-147V have been reported previously (19).
Proc. Natl. Acad. Sci. USA 88 (1991)
1707
structed. We purified 30 of the most active proteins (29 mutants plus WT) as judged by their ability to support phage growth (28). The estimated free energy change upon unfolding in 2 M Gdn HCI (AGCu,2M) for each mutant is listed in Table 1 along with the difference in stability (AAG0U,2M) between the WT and each mutant (19, 22). Properties of the Protein Interior. Fig. 1 compares the stability difference (AAG0u,2M) between each of the 29 mutant proteins and the WT with the hydrophobicity changes expected for these substitutions based on free energies of transfer from cyclohexane to water (10). The correlation between the hydrophobicity changes and the observed stability changes is very poor. Mutants with core hydrophobicity differences of >10 kcal/mol (per dimer) have nearly identical stabilities. Scales using octanol, ethanol, or air as the apolar phase (8, 9, 29) or those based on the tendencies of apolar side chains to be buried in the protein interior (30) show even lower correlations with the observed stability changes. These results indicate that liquid hydrocarbons or alcohols do not adequately model the effects of interior amino acid substitutions on gene V protein stability. In particular, the gene V protein interior differs from apolar liquids by not behaving like a uniform phase. If the protein interior has uniform density, flexibility, and polarity then the stabilities of pairs of mutants with identical amino acid compositions, but with residues at positions 35 and 47 reversed, should be very similar. Fig. 2 shows that the stabilities of such pairs are quite different despite their identical core hydrophobicities. For example, the stabilities of the proteins V35L-147M and V35M-I47L differ by 3.7 kcal/mol. The contribution of an amino acid to protein stability evidently depends strongly on its position. To analyze the effect of an interior substitution on protein stability, at least two factors must be considered beyond the change in hydrophobicity. The first is the effect of placing residues with different sizes and shapes at an interior site. Because this is the energetic effect of altering the arrangement of groups inside the protein, we call it the interior packing energy (19). The second factor is the local polarity of the protein interior-that is, the potential for specific polar or weakly polar interactions between core atoms and atoms of the substituting residues (7). We examined the interior packing energetics at positions 35 and 47 by measuring the stabilities of proteins containing aliphatic residues at these 5.0 0
E
as U
8 8
0.0
.W
o
000 o
-W
n
-5.0
=
0
0
0
0
4)
0 0 0
-10.0
4)
co) -15.0
-10.0
-5.0
0.0
5.0
Change in Core Hydrophobicity (kcal/mol) FIG. 1. Comparison of stability changes of homodimeric gene V protein measured at 2.0 M Gdn HCI (AAG'U.2M) with changes in core hydrophobicity (AAGtr). The hydrophobicity changes are the differences in the free energy of transfer of 2 mol of amino acid side chains, relative to the WT side chains, from cyclohexane to water (10).
1708
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Sandberg and Terwilliger
Position 47 8.0
r --
II1 .L
6.0
204.0
Ala
Val
Leu
lie
Cys
Met
Phe
Position 35 8.0
-o -8.0
-6.0
-4.0
-2.0
0.0
AAI\Gou, 2M (kcal/mol) FIG. 2. Comparison of stabilities, relative to WT (AAG',2M), of pairs of proteins with identical core composition but with residues at positions 35 and 47 reversed. The WT protein is indicated by large type.
sites, and we examined the interior polarity by studying proteins containing the weakly polar residues cysteine, methionine, and phenylalanine (7). Repacking the Protein Interior. To evaluate the packing energetics of the gene V protein interior, we first defined the mutants V35A or I47A as reference, "minimal" proteins. Then we compared the stability of the protein containing alanine at position 35 or 47 with the stabilities of proteins containing the aliphatic amino acids valine, isoleucine, and leucine (Fig. 3 A and B). The portions of the observed stability changes arising from altered hydrophobicity were estimated from free energies of transfer (AGtr) of aliphatic side chains from cyclohexane to water (10). AGtr is the free energy of removing aliphatic groups from water and placing them in a uniform liquid phase capable of van der Waals interactions. We use cyclohexane rather than more polar liquids-e.g., alcohols-for this calculation so that AGtr will not include effects such as disruption of hydrogen bonding in the nonaqueous phase (32). The difference between the measured stability change and the change in hydrophobicity is an estimate of the packing energy difference between the protein interior at these sites and cyclohexane. Fig. 3A shows that aliphatic replacements of alanine at position 47 lead to changes in stability (AAG0u,2M) that are nearly equal to the changes in side chain hydrophobicity (AAGtr). Valine, isoleucine (the WT residue), and leucine at position 47 stabilize the protein by 4.5, 7.0, and 6.4 kcal/mol of dimers relative to alanine, respectively. The WT residue yields the most stable protein, but leucine provides similar stabilization, suggesting that position 47 accommodates either residue efficiently. In contrast, Fig. 3B shows that aliphatic replacements of alanine at position 35 always lead to smaller increases in stability than the corresponding changes in hydrophobicity. Again, the WT residue is most stabilizing, indicating that when all factors are considered, the site at 35 is best designed to be occupied by valine. However, the protein with valine at position 35 is stabilized by only about 2 kcal/mol relative to alanine, whereas valine has a preference for an apolar environment that is about 4 kcal (for 2 mol of side chains) greater than that of alanine (10). Similarly, leucine actually destabilizes the protein relative to alanine at position 35, and isoleucine stabilizes the protein somewhat
0
1 fl0
20i -2.0
Ala
Val Leu
lie
Cys
Met
Phe
FIG. 3. Changes in stability (AAG°U.2M) when alanine at position 47 or position 35 in the gene V protein interior is replaced by aliphatic residues (A and B) or weakly polar residues (C and D). Stability changes are indicated by black bars; shaded bars show the change in free energy of transfer (AAGtr), relative to alanine, from cyclohexane to water for 2 mol of each side chain (10). Free energies of transfer from cyclohexane to 2.0 M Gdn-HCl differ from these values by
Energetics of repacking a protein interior (gene V protein/protein stability/hydrophobic effect/protein engineering)
WARREN S. SANDBERG AND THOMAS C. TERWILLIGER Department of Biochemistry and Molecular Biology, The University of Chicago, Chicago, IL 60637
Communicated by David Eisenberg, November 20, 1990 (received for review September 5, 1990)
ABSTRACT To test whether interactions in the hydrophobic core of a protein can be adequately modeled based on the properties of a liquid hydrocarbon, we measured the unfolding free energies of the wild-type bacteriophage fl gene V protein and 29 mutants with apolar substitutions at positions 35 and 47. Stability changes arising from identical mutations at these two buried sites are quite different, suggesting that one site is more rigid than the other. Reversals of residues at positions 35 and 47 confirm that their environments are distinct. Mutants containing weakly polar residues at these two sites suggest that the protein interior is more polar than a liquid hydrocarbon. Interactions between residues at the two sites appear to be minimal. These observations are compatible with a view of protein interiors that incorporates properties of liquid hydrocarbons but also includes polar interactions and a sitedependent "packing energy" associated with changes in internal structure.
polar side chains, and from the weakly polar side chains such as phenylalanine and methionine (4, 7, 16). Considering the tightness of packing and variations in polarity, organization, and density found inside proteins, the apolar liquid model may not be sufficient to describe the effects of interior amino acid substitutions on protein stability. For example, amino acid substitutions that increase interior hydrophobicity do not necessarily increase protein stability (17-19), and the deleterious effects of cavities created when large residues are replaced by small ones have been noted (11, 12). Our present goal is to study the effects of mutations in the hydrophobic core of the bacteriophage fl gene V protein in order to make an assessment of the values of simple models of a protein interior, to evaluate the energetics of changing the arrangement of apolar groups inside a protein, and to judge the polarity of a protein hydrophobic core.
The structures and stabilities of proteins are thought to be strongly dependent on the arrangements and compositions of their hydrophobic cores (1-3). Interior side chains must efficiently fill the space between strands of the polypeptide (4), and shielding of nonpolar atoms from solvent water contributes significantly to the stabilization of folded proteins through the hydrophobic effect (1, 5). Beyond these general observations, however, our understanding of the properties of protein interiors is quite limited. For example, many combinations of nonpolar side chains can be accommodated in the interior of a protein without severe loss of function (6), but prediction of the stabilities of such mutants is beyond our grasp, the penalty in terms of stability for distortion of the protein interior is unknown, and the magnitudes of polar and weakly polar interactions in protein interiors are only approximately known (7). The contribution of interior amino acid side chains to protein stability is often assessed by treating the protein interior as an apolar liquid (8-10). Preferences of apolar side chains for the protein interior are estimated from their relative solubilities in water and various apolar liquids. In experiments on T4 lysozyme, barnase, and tryptophan synthetase (11-14), successive reductions in the size of apolar residues at buried sites lead to reductions in stability that are correlated with the corresponding reductions in residue hydrophobicity. However, protein interiors differ from apolar liquids in many ways. Interior residues are more tightly packed than molecules in apolar liquids, with densities approaching those of amino acid crystals (5, 15). Many atoms in proteins, such as those in the main chain of the polypeptide, are held relatively rigidly, and others, such as those of many side chains, have more flexibility. Additionally, local variations in polarity result because protein interiors contain polar groups from the polypeptide backbone, from buried
METHODS Isolation of Mutant Gene V Proteins. Derivatives ofplasmid p1T18 (20) encoding mutant gene V proteins were constructed as described (21). Escherichia coli, strain K561, was transformed with these pT118 derivatives, and cell growth and protein purification will be described elsewhere (34). We use the one-letter code to describe mutants; the replacement of valine at position 35 with leucine is denoted by V35L. Estimation of Protein Stability. Guanidine hydrochloride (Gdn HCl) induces reversible, cooperative denaturation of the gene V protein that can be monitored using the gene V protein circular dichroism at 229 nm (34). Stabilities are expressed as free energy changes upon unfolding in the absence of denaturant (AG'UH2O) and in the presence of 2.0 M Gdn HCl (AG0U 2M). Cm is the concentration (in M, mol/ liter) of denaturant at which half the polypeptide chains are unfolded, and m is the dependence of AG0u on (Gdn HCI]. Stabilities are estimated by fitting the unfolding data to a two-state model (22) as described (19) except that the dependence of circular dichroism on [Gdn HCI] at nondenaturing concentrations of denaturant is taken to be identical to that of the wild type (WT) for proteins with Cm 95% buried in the crystallographic model and appear to be in direct contact with each other (26). Replacements of Val-35 and Ile-47 allow comparison of the properties of the protein interior at neighboring sites, and mutants containing simultaneous substitutions at the two sites allow examination of the interaction between the two residues (27). Val-35 and Ile-47 were replaced with all combinations of glycine, alanine, valine, methionine, leucine, isoleucine, phenylalanine, and tryptophan. Additionally, the three mutants containing cysteine at one or both sites were conTable 1. Stabilities of gene V proteins measured in vitro M Residue AG U,2M, AAG0u,2M, Cm, kcal/ AG0U.H20, M mol'M kcal/mol kcal/mol kcal/mol 35 47 V -4.5 4.6 -4.4 A 1.50 13.6 -4.5 15.2 6.1 A L 1.85 -2.9 6.9 -2.2 I A 2.01 -3.9 14.6 A M 1.47 -4.2 12.9 4.6 -4.4 F -4.4 14.2 5.4 A 1.69 -3.6 -4.9 11.6 1.9 -7.1 C C 0.99 I 15.8 7.7 -1.4 C 2.21 -4.1 V A 0.87 -4.3 10.5 2.0 -7.0 V -4.5 C 1.33 12.8 3.8 -5.2 V V 6.5 1.91 -3.9 14.3 -2.5 V L -3.7 15.9 8.4 2.43 -0.6 I V 2.61 -3.6 16.3 9.0 V M 2.02 -3.9 14.6 6.9 -2.1 V F -4.0 15.1 7.1 2.07 -1.9 L V -4.2 12.3 1.32 4.0 -5.0 L L 1.70 14.1 5.5 -4.3 -3.5 I L 1.90 -4.5 15.3 6.4 -2.7 L M 1.37 -5.0 13.7 3.7 -5.3 L F 1.60 -4.3 13.7 5.1 -3.9 I V -4.1 1.80 14.2 6.0 -3.0 I L -4.2 2.27 16.2 7.9 -1.1 I I 2.44 15.7 -3.6 8.4 -0.6 I M 1.87 -4.2 14.7 6.3 -2.8 I F 2.05 -4.4 15.8 7.0 -2.0 M L 2.14 -4.1 15.5 7.4 -1.6 I M 2.33 15.2 -3.6 8.0 -1.0 M M 1.65 -3.8 13.1 5.5 -3.5 M F 1.99 -4.7 16.1 6.8 -2.3 F L 1.51 12.7 -3.9 4.9 -4.1 I F 1.76 -3.8 13.6 5.9 -3.1 Amino acids at positions 35 and 47 are given by their one-letter codes and values of Cm and m are listed (see text). The WT is denoted by bold type. Free energy changes upon unfolding in the absence of denaturant (AG0U H o) and in the presence of 2.0 M Gdn-HCl (AG0U,2M) and the difrerences in AG'U.2M between the mutants and WT (AAG0U,2M) are reported in kcal/mol of dimeric protein. Destabilized mutants have negative values of AAG0u,2M, and the magnitude corresponds to making the same change twice (once for each monomer). Stabilities of the WT and the mutants V351, 147V, and V351-147V have been reported previously (19).
Proc. Natl. Acad. Sci. USA 88 (1991)
1707
structed. We purified 30 of the most active proteins (29 mutants plus WT) as judged by their ability to support phage growth (28). The estimated free energy change upon unfolding in 2 M Gdn HCI (AGCu,2M) for each mutant is listed in Table 1 along with the difference in stability (AAG0U,2M) between the WT and each mutant (19, 22). Properties of the Protein Interior. Fig. 1 compares the stability difference (AAG0u,2M) between each of the 29 mutant proteins and the WT with the hydrophobicity changes expected for these substitutions based on free energies of transfer from cyclohexane to water (10). The correlation between the hydrophobicity changes and the observed stability changes is very poor. Mutants with core hydrophobicity differences of >10 kcal/mol (per dimer) have nearly identical stabilities. Scales using octanol, ethanol, or air as the apolar phase (8, 9, 29) or those based on the tendencies of apolar side chains to be buried in the protein interior (30) show even lower correlations with the observed stability changes. These results indicate that liquid hydrocarbons or alcohols do not adequately model the effects of interior amino acid substitutions on gene V protein stability. In particular, the gene V protein interior differs from apolar liquids by not behaving like a uniform phase. If the protein interior has uniform density, flexibility, and polarity then the stabilities of pairs of mutants with identical amino acid compositions, but with residues at positions 35 and 47 reversed, should be very similar. Fig. 2 shows that the stabilities of such pairs are quite different despite their identical core hydrophobicities. For example, the stabilities of the proteins V35L-147M and V35M-I47L differ by 3.7 kcal/mol. The contribution of an amino acid to protein stability evidently depends strongly on its position. To analyze the effect of an interior substitution on protein stability, at least two factors must be considered beyond the change in hydrophobicity. The first is the effect of placing residues with different sizes and shapes at an interior site. Because this is the energetic effect of altering the arrangement of groups inside the protein, we call it the interior packing energy (19). The second factor is the local polarity of the protein interior-that is, the potential for specific polar or weakly polar interactions between core atoms and atoms of the substituting residues (7). We examined the interior packing energetics at positions 35 and 47 by measuring the stabilities of proteins containing aliphatic residues at these 5.0 0
E
as U
8 8
0.0
.W
o
000 o
-W
n
-5.0
=
0
0
0
0
4)
0 0 0
-10.0
4)
co) -15.0
-10.0
-5.0
0.0
5.0
Change in Core Hydrophobicity (kcal/mol) FIG. 1. Comparison of stability changes of homodimeric gene V protein measured at 2.0 M Gdn HCI (AAG'U.2M) with changes in core hydrophobicity (AAGtr). The hydrophobicity changes are the differences in the free energy of transfer of 2 mol of amino acid side chains, relative to the WT side chains, from cyclohexane to water (10).
1708
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Sandberg and Terwilliger
Position 47 8.0
r --
II1 .L
6.0
204.0
Ala
Val
Leu
lie
Cys
Met
Phe
Position 35 8.0
-o -8.0
-6.0
-4.0
-2.0
0.0
AAI\Gou, 2M (kcal/mol) FIG. 2. Comparison of stabilities, relative to WT (AAG',2M), of pairs of proteins with identical core composition but with residues at positions 35 and 47 reversed. The WT protein is indicated by large type.
sites, and we examined the interior polarity by studying proteins containing the weakly polar residues cysteine, methionine, and phenylalanine (7). Repacking the Protein Interior. To evaluate the packing energetics of the gene V protein interior, we first defined the mutants V35A or I47A as reference, "minimal" proteins. Then we compared the stability of the protein containing alanine at position 35 or 47 with the stabilities of proteins containing the aliphatic amino acids valine, isoleucine, and leucine (Fig. 3 A and B). The portions of the observed stability changes arising from altered hydrophobicity were estimated from free energies of transfer (AGtr) of aliphatic side chains from cyclohexane to water (10). AGtr is the free energy of removing aliphatic groups from water and placing them in a uniform liquid phase capable of van der Waals interactions. We use cyclohexane rather than more polar liquids-e.g., alcohols-for this calculation so that AGtr will not include effects such as disruption of hydrogen bonding in the nonaqueous phase (32). The difference between the measured stability change and the change in hydrophobicity is an estimate of the packing energy difference between the protein interior at these sites and cyclohexane. Fig. 3A shows that aliphatic replacements of alanine at position 47 lead to changes in stability (AAG0u,2M) that are nearly equal to the changes in side chain hydrophobicity (AAGtr). Valine, isoleucine (the WT residue), and leucine at position 47 stabilize the protein by 4.5, 7.0, and 6.4 kcal/mol of dimers relative to alanine, respectively. The WT residue yields the most stable protein, but leucine provides similar stabilization, suggesting that position 47 accommodates either residue efficiently. In contrast, Fig. 3B shows that aliphatic replacements of alanine at position 35 always lead to smaller increases in stability than the corresponding changes in hydrophobicity. Again, the WT residue is most stabilizing, indicating that when all factors are considered, the site at 35 is best designed to be occupied by valine. However, the protein with valine at position 35 is stabilized by only about 2 kcal/mol relative to alanine, whereas valine has a preference for an apolar environment that is about 4 kcal (for 2 mol of side chains) greater than that of alanine (10). Similarly, leucine actually destabilizes the protein relative to alanine at position 35, and isoleucine stabilizes the protein somewhat
0
1 fl0
20i -2.0
Ala
Val Leu
lie
Cys
Met
Phe
FIG. 3. Changes in stability (AAG°U.2M) when alanine at position 47 or position 35 in the gene V protein interior is replaced by aliphatic residues (A and B) or weakly polar residues (C and D). Stability changes are indicated by black bars; shaded bars show the change in free energy of transfer (AAGtr), relative to alanine, from cyclohexane to water for 2 mol of each side chain (10). Free energies of transfer from cyclohexane to 2.0 M Gdn-HCl differ from these values by
