Synthetic Model Proteins POSITIONAL EFFECTS OF INTERCHAIN HYDROPHOBIC INTERACTIONS ON STABILITY OF TWO-STRANDED a-HELICAL COILED-COILS*

We have designed a model protein that consists of two identical 35-residue polypeptide chains, parallel and in-register arranged in a two-stranded a-helical coiled-coil structure. This structure is stabilized by interchain hydrophobic interactions between leucine residues at positions “a” and “d” of a repeating heptad sequence. To determine the positional effects of inter- chain hydrophobic interactions on the stability of the coiled-coil, a single leucine residue in each chain at position “a” (9, 16, 23, 30) and Ud“ (5, 12, 19, 26, 33) was systematically replaced by an alanine. All these proteins formed two-stranded a-helical coiled-coils in benign conditions (0.05 M phosphate, 0.1 M KCl, pH 7). The stability of each mutant protein was determined by guanidine hydrochloride denaturation experiments, where the decrease in ellipticity at 220 nm was monitored by circular dichroism. The single alanine replacements of a leucine residue at hydrophobic posi- tions a and/or d are all shown to destabilize the coiled-coil structure. The non-equivalent hydrophobic posi- tions a and d make an equivalent contribution to protein stability along the majority of the coiled-coil structure (positions 9-30). The small decrease in coiled-coil stability caused by Leu + Ala substitution at either ends of the coiled-coil suggested that the Leu-Leu hy- drophobic interactions are less correct primary ion molecular weights confirmed by time of flight spectroscopy on a BIOION-20 Nordic Circular dichroism spectra were recorded at 20 "C on a Jasco 500C spectropolarimeter attached a Jasco DPdOON data processor and a Lauda (model RMS) water bath used to the temperature of the cell. The instrument was routinely calibrated with an aqueous solution of recrystallized D-10-camphorsulfonic acid. Ellipticity is reported as mean residue ellipticity 0 and the limits of error of measurements at 220 nm were k500". The guanidine hydrochloride denaturation studies were carried out by preparing mixtures of a stock of peptide in KCl, 50 PO4, and a in the ratios and 6 GdnHCl appropriate final GdnHCl for the CD Pep- tide concentrations of stock solutions were determined by amino acid analysis. Energy minimization and molecular dynamics were carried out on a Silicon Graphics Personal Iris with the Insight I1 and Discover program (Biosym Technologies Inc., San Diego, CA) (20). The con- sistent valence force field and a step size of 1 fs were used in all molecular dynamics and energy minimizations. All side-chain inter- actions have been optimized in the coiled-coil structure that was chosen as the starting point for the molecular dynamics simulations. The molecular dynamics simulations were run 5 ps at a temperature of 300 K. The first 3 ps were used to allow the system to equilibrate. Four structures from each of four different molecular dynamic runs were selected in the last 2 ps at 0.5-ps intervals. The 16 structures were superimposed to minimize the root mean square deviation of backbone atoms of residues 15-21 in both chains.

Edmonton, Alberta T6G 2H7, Canad; We have designed a model protein that consists of two identical 35-residue polypeptide chains, parallel and in-register arranged in a two-stranded a-helical coiled-coil structure. This structure is stabilized by interchain hydrophobic interactions between leucine residues at positions "a" and "d" of a repeating heptad sequence. To determine the positional effects of interchain hydrophobic interactions on the stability of the coiled-coil, a single leucine residue in each chain at position "a" (9, 16, 23, 30) and Ud" (5, 12, 19, 26, 33) was systematically replaced by an alanine. All these proteins formed two-stranded a-helical coiled-coils in benign conditions (0.05 M phosphate, 0.1 M KCl, pH 7).
The stability of each mutant protein was determined by guanidine hydrochloride denaturation experiments, where the decrease in ellipticity at 220 nm was monitored by circular dichroism. The single alanine replacements of a leucine residue at hydrophobic positions a and/or d are all shown to destabilize the coiledcoil structure. The non-equivalent hydrophobic positions a and d make an equivalent contribution to protein stability along the majority of the coiled-coil structure (positions 9-30). The small decrease in coiled-coil stability caused by Leu + Ala substitution at either ends of the coiled-coil suggested that the Leu-Leu hydrophobic interactions are less important at the ends of the coiled-coil and the ends of the coiled-coil are more flexible. Analysis of the difference between the ellipticity in benign buffer and in 50% trifluoroethanol (ABzz0) and the slope term from a plot of the free energy of unfolding versus guanidine hydrochloride concentration also supported the conclusion that the leucine residues at the ends of the coiled-coil are much less buried than in the middle section of the coiled-coil.
The de nouo design of model proteins is an important endeavor that not only tests our understanding of protein folding and structure, but also lays the groundwork for the design of novel proteins with the desired biological/immunological activities. The purpose of our research is to design a small and unique protein molecule with defined secondary, * This project is an integral part of the Protein Engineering: 3D-Structure, Function and Design Network of Centres of Excellence Program supported by the Government of Canada. A postdoctoral fellowship stipend (to N. E. Z.) and research allowance was provided by the Alberta Heritage Foundation for Medical Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "uduertisernent" in accordance with 18 U. S tertiary, and quaternary structure, and then to modify systematically the structure to delineate the contributions that various amino acid side chains make to control folding, conformation, and stability (1-5). It is generally assumed that protein folding is not a random process and that elements of secondary structure such as @-turns or a-helices are the most likely candidates for nucleus formation in the unfolded protein (6-8). It seemed appropriate to design the model protein with only one type of secondary structure, Le. the a-helix. An a-helical protein was chosen because it shows the least conformational variability among hydrogen-bonded secondary structures found in globular proteins. Also, the a-helical structure is the easiest type of secondary structure to monitor in aqueous solution at neutral pH using such techniques as circular dichroism and NMR spectroscopy. To introduce tertiary and quaternary structure into a molecule, a minimum of two interacting a-helices is required, We concluded that the ideal model protein could be a two-stranded a-helical coiledcoil.
The coiled-coil consists of two right-handed a-helical polypeptide chains that are parallel, in-register, and coil about one another. The main advantage of studying coiled-coils in preference to single-stranded a-helices is that coiled-coils are stabilized by both intrachain and interchain interactions. Interchain interactions are the main feature responsible for the folding and stabilization of the three-dimensional structure of proteins and the stabilization of a-helices in aqueous solution by side chain interactions characterizes the globular proteins as well as a-helical coiled-coils (1, 5). Model coiledcoils can be used to investigate all the noncovalent interactions involved in maintaining the three-dimensional structure of proteins ( i e . hydrogen bonds, ionic interactions, van der Waals bonds, hydrophobic interactions, helix dipole, charge compensation effects, and side chain packing effects). The hydrophobic interactions between nonpolar residues are thought to be among the most important in determining the three-dimensional structure of a protein (9-11). In this study, we have focused on the hydrophobic interactions and their positional effects on the stability of the coiled-coil protein.
The coiled-coils are biochemically significant in their own right and are involved in muscle regulation (12-14), DNA binding properties (15-lB), and other functions (19). Thus, a detailed understanding of coiled-coil structure may delineate the reasons for a wide range of hydrophobic residues in the hydrophobic core and how a conformational change is transmitted along the coiled-coil. 430A as described previously (4). The peptide was cleaved from the resin by reaction with hydrogen fluoride (20 ml/g resin) containing 10% anisole and 2% 1,2-ethanedithiol for 1 h at -5 "C. The crude peptides were purified by reversed-phase HPLC' on a Synchropak RP-P semipreparative ClS column (250 X 10-mm inner diameter, 6.5-pm particle size, 300-A pore size) (SynChrom, Lafayette, IN), with a linear AB gradient (ranging from 0.2 to 1.0% B/min, depending on the peptide) at a flow-rate of 2 ml/min, where solvent A was 0.05% aqueous trifluoroacetic acid and solvent B was 0.05% trifluoroacetic acid in acetonitrile. Peptide purity was verified by analytical reversedphase HPLC and amino acid analysis. Amino acid analysis was performed on a Beckman model 6300 amino acid analyzer (Beckman). The correct primary ion molecular weights were confirmed by time of flight mass spectroscopy on a BIOION-20 Nordic (Uppsala, Sweden).
Circular dichroism spectra were recorded at 20 "C on a Jasco J-500C spectropolarimeter (Jasco, Easton, MD) attached to a Jasco DPdOON data processor and a Lauda (model RMS) water bath (Brinkmann Instruments, Rexdale, Ontario, Canada) used to control the temperature of the cell. The instrument was routinely calibrated with an aqueous solution of recrystallized D-10-camphorsulfonic acid. Ellipticity is reported as mean residue ellipticity 0 and the limits of error of measurements at 220 nm were k500". The guanidine hydrochloride denaturation studies were carried out by preparing mixtures of a stock solution of peptide in buffer (0.1 M KCl, 50 mM PO4, pH 7), buffer alone, and a solution of 6 M GdnHCl in buffer where the ratios of buffer and 6 M GdnHCl solutions varied to give the appropriate final GdnHCl concentrations for the CD measurements. Peptide concentrations of stock solutions were determined by amino acid analysis.
Energy minimization and molecular dynamics were carried out on a Silicon Graphics Personal Iris with the Insight I1 and Discover program (Biosym Technologies Inc., San Diego, CA) (20). The consistent valence force field and a step size of 1 fs were used in all molecular dynamics and energy minimizations. All side-chain interactions have been optimized in the coiled-coil structure that was chosen as the starting point for the molecular dynamics simulations. The molecular dynamics simulations were run 5 ps at a temperature of 300 K. The first 3 ps were used to allow the system to equilibrate. Four structures from each of four different molecular dynamic runs were selected in the last 2 ps at 0.5-ps intervals. The 16 structures were superimposed to minimize the root mean square deviation of backbone atoms of residues 15-21 in both chains.

RESULTS
The amino acid sequence of the 70-residue synthetic twostranded a-helical coiled-coil (two identical 35-residue polypeptide chains) is shown in Fig. 1. In choosing the amino acid sequence for a particular protein, one must select a sequence that specifies one particular secondary, tertiary, and quaternary structure and disfavors alternative arrangements. In this regard, our choice of sequence was based upon the sequence of tropomyosin, the first two stranded a-helical coiled-coil to be sequenced (21)(22)(23)(24). The most striking feature observed in the amino acid sequence is a repeating pattern of hydrophobic residues which extends throughout the entire sequence of the 284-residue polypeptide chain. Hodges and co-workers (21,25,26) first identified this repeat and proved that the hydrophobic residues at positions "a" and "d" (Fig. 1 and right side of top panel in Fig. 2) were responsible for stabilizing this structure. In addition, the positions "e" and "g" contain glutamic acid and lysine residues, respectively, which strongly favor the formation of interchain ionic interactions (e-g' and e'-g) (left side of top panel in Fig. 2) that would stabilize a parallel and in-register arrangement of the polypeptide The abbreviations used are: HPLC, high performance liquid chromatography; TFE, trifluoroethanol; GdnHC1, guanidine hydrochloride; AG,, free energy of unfolding at each individual concentration of denaturant; AGUHZ0, free energy of unfolding in the absence of guanidine hydrochloride (estimated by extrapolating AG, values to zero, assuming that they are linearly related (AG, = AGUH2Om[GdnHCl])).

FIG. 1.
The amino acid sequence of the 70-residue synthetic two-stranded a-helical coiled-coil. The a-amino group of each chain is acetylated and the COOH-terminal carboxyl group is amidated. Each 35-residue polypeptide chain consists of five heptads. The two series of hydrophobes (denoted as a and d ) which repeat at 7-residue intervals along the polypeptide chain (also referred to as 3-4 or 4-3 hydrophobic repeat) are responsible for the formation and stabilization of the coiled-coil. The solid lines represent the interchain hydrophobic interactions that occur between leucine residues in chain 1 and chain 2. A leucine residue was systematically substituted in each chain at the same position by an alanine residue to create a series of coiled-coil analogs with a single Ala-Ala pair replacing one of nine Leu-Leu pairs at positions 5,9, 12, 16, 19, 23, 26, 30, and 33. The coiled-coil analog in which an alanine substitution is made at position 5 in chain 1 and chain 2 is denoted as A5 and similarly A9, A12, A16, A19, A23, A26, A30, or A33. chains. The selection of the amino acid sequence of the basic heptapeptide repeat, Lys-Leu-Glu-Ala-Leu-Glu-Gly, and the chain length of 35 residues have been described in detail (2, 3).
In this study, a leucine residue was systematically replaced in each chain by an alanine residue to create a series of coiledcoil analogs with a single alanine replacing 1 of 9 leucine residues at positions 5, 9, 12, 16, 19, 23, 26, 30, and 33. The coiled-coil analogs are denoted as A5 through A33, respectively, with each chain of an analog containing 8 leucine The side chains are colored as follows. The hydrophobic Leu residues are brown (in the white polypeptide chain) and green (in the yellow one). Ala residues are purple, Glu (position e) red, Glu (position c) pink, Lys blue. This theoretical coiled-coil model was obtained by molecular dynamics and energy minimization with torsion angles (@, $) and interchain Leu-Leu residue backbone constraints using the Insight11 and Discover program on a Silicon Graphics personal Iris computer. All side chain interactions have been optimized in the coiled-coil structure and the two a-helices are identical. Left side of top panel is the representation of the two-stranded a-helical coiled-coil. Of importance are the 10 pairs of ionic interactions between Lys (blue) and Glu (red) residues. The right side of the toppanel is a display of the Leu-Leu interactions in the coiled-coil. All the side chains have been removed from the coiled-coil, except the Leu residues in the hydrophobic 3-4 repeat positions of both chains. In the bottom panel, only Leu side chains in the hydrophobic 3-4 repeat positions of the coiled-coil are shown to illustrate that a Leu side chain of one a-helix interacts with the equivalent Leu in the opposing a-helix (such as 16-16') and 2 Leu residues in the alternative hydrophobic positions in the opposing ahelix (such as 16 residues and 1 alanine residue in a and/or d positions. Thus, the A5 analog contains alanine at position 5 in chain 1 and position 5' in chain 2 or 1 alanine pair and 8 leucine pairs in the coiled-coil. The native coiled-coil model, in which all nine a and d positions are occupied by leucine residues results in 9 leucine pairs and is denoted as L. The questions to be answered in this study were as follows. First, do the hydrophobic (Leu-Leu) interactions along the length of coiled-coil contribute equally to the stability of the coiled-coil? Second, do the nonequivalent hydrophobic positions of 4-3 repeat (positions a and d) contribute equally to stability?
Characterization of the a-Helical Coiled-coils-The criteria used to establish that the synthetic peptides have formed a coiled-coil structure are as follows. 1) Size-exclusion chromatographic studies of these peptides compared to the 1 4 , 21-, 28-, and 35-residue a-helical peptide standards using methodology described previously (5, 30, 53) demonstrated once again that the peptides are dimeric in benign buffer (0.1 M KC1, 50 mM PO4, pH 7) and monomeric in 50% TFE.

2)
The ellipiticity at 220 n m was dependent on peptide concentration in benign medium (inset of Fig. 3). As the concentration of peptide is increased, the monomer * dimer equilibrium is shifted toward the formation of coiled-coil dimer which increases the a-helical content of the peptide. At peptide concentrations greater than 200 pM, the ellipiticity of the peptide was independent of peptide concentration indicating the complete association of monomers to coiled-coil dimer. On the other hand the ellipiticity at 220 nm was independent of peptide concentration in 50% TFE (6-615 p~, inset of For samples containing TFE, the above buffer was diluted 1:1 (v/v) with TFE. Temperature was 20 "C for ellipticity measurements. Inset: the concentration dependence of the mean residue molar ellipiticity at 220 nm for peptide A5. 0 and W denote peptide A5 in benign buffer and in 50% TFE, respectively.

Positional Effects on
Coiled-coil Stability 2667 ence optical system at a speed of 30,000 rpm and at an initial peptide concentration of 344 p~ in the benign medium containing 2 mM dithiothreitol revealed the complete absence of monomer in the system. These results are in agreement with previous sedimentation equilibrium studies (3) which showed that the four or five heptad repeats of the sequence Ac-[Lys-Leu-Glu-Ala-Leu-Glu-Glyl,-Lys-amide formed a homogeneous, two-stranded structure in benign media. 4) The CD spectra for all nine analogs (A5, A9, A12, A16, A19, A23, A26, A30, and A33) are very similar to each other and the native coiled-coil protein (L). All showed two mimima, one near 220 nm and another at 207 nm in benign conditions (0.1 M KC1, 50 mM PO4, pH 7) with high ellipticity values indicating high helical content. The peptides do not show any increase in helicity upon addition of the a-helix inducing solvent TFE, as measured by molar ellipticity at 220 nm (the ellipticities in 50% TFE were less than in benign buffer, Table I). The n-a* transition (220-nm CD band) is responsive to the a-helical content. The PA* excitation band at 205 nm polarizes parallel to the helix axis and is sensitive to whether the a-helix is single-stranded or is an interacting helix as in the case of the two-stranded coiled-coils (27). The decrease in parallel band intensity, coupled with the red shift in the parallel band maximum, corresponds to the conversion of a rigid singlestranded a-helix to an a-helical coiled-coil structure (27). The maximum ellipticity at 207 nm in benign medium shifts to 205 nm in 50% TFE and the ratio of 0220/0207 changes from 1.03 (f0.03) in benign medium to 0.86 (f0.03) in 50% TFE (Fig. 3). These results indicated that all analogs form the twostranded a-helical coiled-coil structure in benign medium and single-stranded a-helices in 50% TFE. These results are in good agreement with our previous studies on coiled-coil peptides (3-5). A 1:l (v/v) mixture of benign buffer with TFE. Temperature was 20 "C for ellipticity measurements.
[GdnHCljl/Z is the transition midpoint, the concentration of guanidine hydrochloride (M) required to give a 50% decrease in ellipticity. e AGUHz0 is the free energy of unfolding in the absence of guanidine hydrochloride and is estimated by extrapolating the free energy of unfolding a t each individual concentration of guanidine hydrochloride (AG,) to zero concentration assuming that they are linearly related (Pace, 1986;Shortle, 1989  Helicity of the Model Peptides-The molar ellipticities of the peptide analogs (Table I) were measured in benign conditions and in the presence of TFE, a solvent that induces helicity in a single-stranded potentially a-helical polypeptide (28)(29). Lau et al. (30) have shown previously that 50% TFE disrupts the quaternary structure of a-helices, i.e. TFE is a denaturant of tertiary and quaternary structure stabilized by hydrophobic interactions. While there are small differences in the ellipiticities at 220 nm of some of the individual analogs relative to each other, all of the analogs (A5 + A33) in 50% TFE have values of about -20,600 f 2,000", implying that a replacement of leucine residue by alanine in the NH2 terminus, in the middle or in the COOH terminus of the peptide, has a similar effect on the single-stranded a-helical structure. These results are in agreement with the study of Merutka and Stellwagen (31). They substituted 1 alanine residue by serine or methionine at position 4, 9, or 14 in a 17-residue a-helical peptide and demonstrated that the effect of substitution on the stability of the a-helix is independent of the substitution position.
By comparing the difference in ellipticities of all the coiledcoil analogs in 50% TFE and in benign buffer (AOzz0), we can conclude that the hydrophobic interactions between the two a-helices in the coiled-coil are better than TFE in inducing a-helical structure. In addition, the observation that the value of peptide L is greater than its analogs (A5 + A33) suggests that the hydrophobicity of the hydrophobic face has a major impact on induction of a-helical structure through the interchain hydrophobic interactions which stabilize the coiled-coil and consequently stabilize the a-helices. Another interesting result is that the decrease in ellipticity (AOsz0) for the analogs with Ala substitution at the ends of the coiledcoil is less than for analogs with substitution in the central region (compare the value of A5(-900), A9(-1,650), and A33 (-2,650) with A12-A30(-4,000 + -6,400)). These results indicated that the interchain hydrophobic interactions are less important at the ends of the coiled-coil, suggesting a more flexible structure at the ends compared to the center.
Denaturation of Coiled-coil Analogs with Guanidine Hydrochloride-The denaturation curves determined by monitoring the ellipticities of the peptides at 220 nm as a'function of guanidine hydrochloride (GdnHCl) concentration at 20 "C are displayed in Figs. 4A and 5A. The transition midpoint of GdnHCl concentration, [GdnHC1]1/2, at which the helical content is 50%, depends on peptide concentration (Fig. 44). The [GdnHC1]1/2 was 2.5 M for the higher peptide concentration (1347 p~) and 1.1 M for the lower peptide concentration (48 p~) . The concentration dependence observed for the unfolding curves shows that chain dissociation is a normal feature in the denaturation of the coiled-coils. In contrast, the disulfide bridged coiled-coil (a disulfide bond formed between 2 cysteine residues at position 2 in each chain) did not show any concentration dependence in the GdnHCl denaturation experiments (data not shown).
If we assume that the single-stranded a-helix (folded monomer) is not present at signficiant concentrations in equilibrium, then only the transition between the two-stranded ahelical coiled-coil and the single-stranded random coil is monitored by the change in ellipticity at 220 nm. Therefore the unfolding reaction follows a two-state model of folded dimers in equilibrium with unfolded monomers    (Fig. U ) , the AGUHs0 is essentially the same for different peptide concentrations (the average AGUHz0 = 8.30 (k0.2) (kcal/mol)). The G, values for all peptides are listed in Table I.
The Positional Effects of Hydrophobic Interactions on the Coiled-coil Stability-In order to compare the stability of coiled-coil analogs, the GdnHCl denaturation for all these coiled-coils was carried out at similar peptide concentrations (-200 p~) to rule out the possible concentration effects on stability. If the stability of a two-stranded coiled-coil is due to hydrophobic interactions in the a and d positions (Fig. l), one would expect to find a decrease in coiled-coil stability with a substitution of any leucine residue with a less hydrophobic alanine residue at these positions. All alanine-substituted coiled-coils are less stable than the native coiled-coil as shown by the transition midpoints and AGUHz0 values ( Fig. 5 and Table I). These results are in agreement with previous studies on coiled-coils (4, 5). However, the relative contribution of these hydrophobic interactions at different positions in the coiled-coil are not identical. A5 and A33 coiled-coils are the most stable of the analogs as indicated by the same denaturation curve (Fig. 5A). A30 shows a further decrease in stability compared to A5 and A33. All remaining analogs have essentially identical stability and are the least stable of the analogs. Fig. 5B plots the transition midpoint of GdnHCl concentration versus the position of the alanine substitution. From Fig. 5B, it is clearly seen that hydrophobic interactions at positions 9, 12, 16, 19, 23, and 26 are very important and contribute equally to the coiled-coil stability. Substitution of a leucine residue with an alanine at each of positions 9-26 greatly reduces protein stability, while substitution of a leucine residue with an alanine at position 5 or 33 only slightly decreases the coiled-coil stability. The small decrease in stability found for Leu + Ala substitution at either end of the coiled-coil suggests that Leu-Leu interactions at the ends of the coiled-coil do not contribute significantly to the coiledcoil stability, and leucine residues at the ends of the coiledcoil are much less buried than in the middle section of the coiled-coil. An explanation based on the theoretical model of the coiled-coil structure is that, in a parallel coiled-coil, a leucine side chain in one helix makes inter-chain contact with the equivalent leucine in the opposing a-helix (such as 16-16') and also interacts extensively with 2 leucine residues from the opposing helix, one above and one below (such as 16-12' and 16-19'). All of these hydrophobic interactions between leucine residues in chain 1 and chain 2 are shown in Fig. 2 (bottom panel, left side). In the native coiled-coil, a leucine residue at the middle part of the coiled-coil interacts with 3 leucine residues in the opposing helix (one green side chain contacts with three brown side chains or one brown with three greens), while a leucine residue at either end of the coiled-coil (positions 5 and 33) only interacts with 2 leucines in the opposing helix. If one considers a Leu-Leu pair in the central region of the coiled-coil such as 16-16', the total number of interchain Leu-Leu interactions are five (16-16', 16-12', 16-19', 16'-12, and 16'-19). On the other hand, a Leu-Leu pair at the end of the coiled-coil (33-33') has only 3 Leu-Leu interchain interactions (33-33', 33-30', and 33'-30). This analysis would predict that the ends of the coiled-coil would be less stable and more flexible. The molecular dynamics simulation of the model coiled-coil (Fig.  6) is in good agreement with our experimental results. When an alanine substitution is made at the end of the coiled-coil (such as position 33), 3 leucine-leucine hydrophobic interactions are lost (33-30', 33-33', and 33'-30). By comparison, 5 leucineleucine hydrophobic interactions would be lost, when Ala was substituted in the middle part of the coiled-coil (for example, 19- 16', 19-19', 19-23', 19'-16, and 19'-23) in the A19 coiledcoil analog (bottom panel in Fig. 2). One of the mechanisms by which hydrophobic side chains stabilize protein structure is through van der Waals interactions. The presence of an empty cavity in the mutant protein where the wild-type side chain normally resides would destabilize the native state relative to the denatured state because of reduced interactions (35).

DISCUSSION
Previous studies by our laboratory (3-5) have demonstrated that a large part of the stability of the coiled-coils is due to the presence of hydrophobes in the 3-4 repeat positions exactly as predicted. It seems reasonable that the two helices are stabilized primarily by the burial of closely packed, hydrophobic side chains (36), although hydrogen bonds and salt bridges between interacting helices, usually involving nonburied atoms, also occur quite often (37). Mutations or modifications which disturb either the close packing or the hydro- phobicity of a helix pairing site would then destabilize that interaction and consequently the protein (38). Several experiments indicated that helix pairing sites are indeed sensitive to these two types of disruptions (39-42). Such disruptions have also been observed in this model coiled-coil protein ( 5 ) . When the leucine residues a t position 16 and 19 of each chain ( Fig. 1) are replaced with Ile, Val, Ala, Phe, or Tyr, the stability of the protein with regard to the hydrophobic aliphatic amino acid residue substitution correlates with the hydrophobicity of the side chain (order of protein stability is Leu > Ile > Val > Ala). The aromatic amino acids did not stabilize the proteins to the same extent as aliphatic side chains with similar hydrophobicities (5), suggesting that the aliphatic side chains provide much better interchain packing than the aromatic side chains. Similar results have been reported in studies of the globular proteins tryptophan synthase a subunit (43) and bacteriophage T4 lysozyme (44). In this regard, it is important to note that the relative contributions of nonpolar residues to protein stability using our two-stranded a-helical coiled-coil model and the study by Matsumura et al. (44) on T4 lysozyme showed an excellent correlation. In both proteins, the mutations were made in the hydrophobic residues involved in the hydrophobic core responsible for overall protein stability. These results verify the general utility of our model to investigate the contribution of side chains to protein stability. In the present study, the single alanine replacements of a leucine residue at hydrophobic positions a and/or d are all shown to destabilize the coiledcoil structure, implying that all of the hydrophobic interactions between these positions contribute to the coiled-coil stability. However, the positions at the ends of the coiled-coil make a smaller contribution to stability. These variations must be a consequence of the variations in the environment surrounding the different residue positions. For instance, a leucine residue would be expected to make a greater contribution to protein stability if it was fully buried than if it was fully exposed to solvent. Indeed, the reverse hydrophobic effects on protein stability have been observed on XCro protein, whereupon amino acid substitution on the surface of the protein resulted in a mutant protein which showed a good correlation between stability and decreasing side chain hydrophobicity (45).
Analysis of the difference between the ellipticity in benign buffer and in 50% TFE (A&,) (see "Results") and the slope term (m value in AG, = AGUH2Om[GdnHCl]) also supported the conclusion that the end of the coiled-coil structure is more flexible and Leu-Leu hydrophobic interactions are less important at the end of the coiled-coil. The chemical interactions by which GdnHCl and other denaturants destabilize protein structure have not been established, with the result that the detailed structural basis of the variations in m value cannot be elucidated clearly. However, as demonstrated by Shortle et al. (35), the values of m for mutant forms of staphylococcal nuclease (in each mutation, single alanine and glycine substitutions were constructed for each of the 11 Leu, 9 Val, 5 Ile, 4 Met, and 3 Phe) are dependent upon the sites of mutations.
Mutations that increased the m value only involved residues that contribute side chains to the major hydrophobic core of the protein, whereas mutations that caused the m value to decrease are located outside of this major hydrophobic core. Upon comparison of the m values of all coiled-coils in this study, it is noted that there is an increase in the m value for alanine substitution in the central region of the coiled-coil. For example, comparing the m values of A12 4 A26 (2.32 + 2.44) with L (2.06) suggests that the hydrophobic residues at positions 12 + 26 form the major hydrophobic core of this coiled-coil structure. The similar m values of A5 (2.10) and A33 (2.13) compared to the native coiled-coil L (2.06) indicate that the hydrophobic positions 5 and 33 are out of the major hydrophobic core.
Integrating the results from this study with other studies on mutants of globular proteins, such as bacterial ribonuclease barnase (46), staphylococcal nuclease (35), and glyceraldehyde-3-phosphate dehyrogenase (47), it would appear that the importance of hydrophobic interactions varies considerably with position within a protein. The mutations which use less hydrophobic residues in place of more hydrophobic ones are generally more deleterious for protein stability at buried than at less buried positions (35,38,(46)(47)(48)(49).
Although hydrophobic residues at positions a and d contribute to the coiled-coil stability, one may ask the question, does a Leu residue at position a make the same contribution to the coiled-coil stability as a Leu at position d? According to the survey by Cohen and Parry (14) on naturally occurring two-stranded a-helical coiled-coils (tropomyosin, myosin, paramyosin, and intermediate filament proteins), leucine is almost equivalently occupied in both positions a and d. The occurrence of leucine is 32.2% in position a and 34.7% in positions d. Our results clearly demonstrate that the a and d hydrophobic positions provide equivalent contributions to stability along the majority of the coiled-coil protein structure (positions 9-26). This result is in good agreement with the distribution of leucine on a and d positions in the naturally occurring coiled-coil proteins.
Analysis of the occurrence of leucine residues at position a and d in the coiled-coil regions (five heptads) of 15 DNAbinding proteins (50)(51), showed 88% leucine in position d, compared to 8% leucine in position a. This difference of sequence requirements between DNA-binding proteins and other coiled-coils may imply that there are differences in structure which may be related to differences in function, even though these two types of proteins are both a-helical coiled-coils. Interestingly, from the study of Hu et al. (52) for DNA-binding coiled-coil proteins, the functional importance of the amino acid side chain at eight positions (a and d) that form the hydrophobic interface of the leucine zipper dimer were determined. It was shown that the d positions were more restricted in tolerance to substitutions than the a positions.