Effects of side-chain characteristics on stability and oligomerization state of a de Novo-designed model coiled-coil: 20 amino acid substitutions in position “d”1

doi:10.1006/jmbi.2000.3866 available online at http://www.idealibrary.com on

J. Mol. Biol. (2000) 300, 377±402

Effects of Side-chain Characteristics on Stability and Oligomerization State of a de Novo-designed Model Coiled-coil: 20 Amino Acid Substitutions in Position ``d'' Brian Tripet1, Kurt Wagschal1, Pierre Lavigne2, Colin T. Mant1 and Robert S. Hodges1* 1

Department of Biochemistry and the Medical Research Council Group in Protein Structure and Function University of Alberta Edmonton, Alberta, Canada T6G 2H7 2 Protein Engineering Network of Centres of Excellence University of Alberta Edmonton, Alberta, Canada T6G 2S2

We describe the de novo design and biophysical characterization of a model coiled-coil protein in which we have systematically substituted 20 different amino acid residues in the central ``d'' position. The model protein consists of two identical 38 residue polypeptide chains covalently linked at their N termini via a disul®de bridge. The hydrophobic core contained Val and Ile residues at positions ``a'' and Leu residues at positions ``d''. This core allowed for the formation of both two-stranded and three-stranded coiledcoils in benign buffer, depending on the substitution at position ``d''. The structure of each analog was analyzed by CD spectroscopy and their relative stability determined by chemical denaturation using GdnHCI (all analogs denatured from the two-stranded state). The oligomeric state(s) was determined by high-performance size-exclusion chromatography and sedimentation equilibrium analysis in benign medium. Our results showed a thermodynamic stability order (in order of decreasing stability) of: Leu, Met, Ile, Tyr, Phe, Val, Gln, Ala, Trp, Asn, His, Thr, Lys, Ser, Asp, Glu, Arg, Orn, and Gly. The Pro analog prevented coiled-coil formation. The overall stability range was 7.4 kcal/mol from the lowest to the highest analog, indicating the importance of the hydrophobic core and the dramatic effect a single substitution in the core can have upon the stability of the protein fold. In general, the side-chain contribution to the level of stability correlated with side-chain hydrophobicity. Molecular modelling studies, however, showed that packing effects could explain deviations from a direct correlation. In regards to oligomerization state, eight analogs demonstrated the ability to populate exclusively one oligomerization state in benign buffer (0.1 M KCl, 0.05 M K2PO4(pH 7)). Ile and Val (the b-branched residues) induced the three-stranded oligomerization state, whereas Tyr, Lys, Arg, Orn, Glu and Asp induced the two-stranded state. Asn, Gln, Ser, Ala, Gly, Phe, Leu, Met and Trp analogs were indiscriminate and populated twostranded and three-stranded states. Comparison of these results with similar substitutions in position ``a'' highlights the positional effects of individual residues in de®ning the stability and numbers of polypeptide chains occurring in a coiled-coil structure. Overall, these results in conjunction with other work now generate a relative thermodynamic stability scale for 19 naturally occurring amino acid residues in either an ``a'' or ``d'' position of a two-stranded coiled-coil. Thus, these results will aid in the de novo design of new coiled-coil structures, a better understanding of their structure/function relationships and the design of algorithms to predict the presence of coiled-coils within native protein sequences. # 2000 Academic Press

*Corresponding author

Keywords: a-helical coiled-coil; hydrophobic core substitutions; protein stability; oligomerization states; de novo design

Abbreviations used: RP-HPLC, reversed-phase high-performance liquid chromatography; SE, sedimentation equilibrium; GdnHCl, guanidine hydrochloride; TFE, tri¯uoroethanol; HPSEC, high-performance size-exclusion chromatography. E-mail address of the corresponding author: [email protected] 0022-2836/00/020377±26 $35.00/0

# 2000 Academic Press

378

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Introduction The a-helical coiled-coil is a common assembly motif found in a diverse array of structural and regulatory proteins in nature. Coiled-coils are comprised of two, three or four right-handed a-helices which wrap around each other to form a slight left-handed superhelical twist. The sequence of all coiled-coils is characterized by a seven-residue (heptad) repeat commonly denoted (abcdefg)n (McLachlan & Stewart, 1975), where positions ``a'' and ``d'' are primarily occupied by hydrophobic residues, positions ``e'' and ``g'' by charged residues, and positions ``b'', ``c'' and ``f'' by polar or charged residues. The periodic 3-4 repeat of the hydrophobic residues occurring at positions ``a'' and ``d'' (Hodges et al., 1972) creates a non-polar face along the a-helix, which in aqueous solution, drives the association of the a-helices via burial of their hydrophobic surface areas into the subunit interface. Additionally, residues in positions ``e'' and ``g'' which ¯ank the hydrophobic core (Figure 1), can form interhelical ion pairs, when oppositely charged (i to i0 ‡ 5: g to e0 ), and thus may also contribute to stability, as well as directing chain registry (Kohn et al., 1997, 1998 and references therein; Yu et al., 1996a,b). Extensive structural studies using de novodesigned model coiled-coil and native coiled-coil sequences have clearly demonstrated the importance of the hydrophobic core ``a'' and ``d'' positions for the folding and stability of coiled-coil structures (Hodges et al., 1981, 1990; Hu et al., 1990; Lau et al., 1984; Moitra et al., 1996; Zhou et al., 1992a,b; Zhu et al., 1992, 1993). It is interesting that these studies also revealed that not only is the type of residue important (i.e. hydrophobic versus polar or charged), but also its placement in either an ``a'' or ``d'' position. For example, it was observed that the b-branched hydrophobic residues of Ile and Val provide signi®cantly greater stability to the coiled-coil than that of the similarly sized hydrophobic residue, Leu, when placed at position ``a'' (Zhou et al., 1992a,b; Zhu et al., 1993), whereas Leu appeared to provide more stability at position ``d'' (Zhu et al., 1993). Subsequent high-resolution X-ray crystallographic analyses of several coiled-coil proteins have since shown that this positional preference arises from the structural differences in the ``a'' and ``d'' positions. Because of the parallel and in-register nature of the a-helices, and their ``knobs into holes'' packing arrangement (Crick, 1953), side-chains in an ``a'' position show a structural geometry in which the Ca-Cb side-chain vector extends parallel to the Ca-Ca helix vector (directing the side-chain away from the hydrophobic core), and hence is termed ``parallel packing'' while in position ``d'', the Ca-Cb side-chain vector extends perpendicular to the Ca-Ca helix vector (directing the side-chain into the hydrophobic core) and hence is termed ``perpendicular packing'' (O'Shea et al., 1991). As a result of these geometric differences, the contribution that a particular amino acid

residue side-chain can make to the stability of the hydrophobic core of a coiled-coil is dependent not only on its hydrophobic character but also on its ability to pack ef®ciently within its local geometrical context. In addition to the signi®cant contribution the interfacial ``a'' and ``d'' residues can make to mediation of coiled-coil stability, it has become apparent that the type and placement of these residues can also in¯uence the associative state of the structure. For example, mutational studies in the protein GCN4 have shown that the presence of a central Asn residue in an ``a'' position imparts a strong tendency for the coiled-coil structure to remain exclusively in the two-stranded oligomeric state (Gonzalez et al., 1996a,b,c; Lumb & Kim, 1995; O'Shea et al., 1993). Monera et al. (1996) have shown that the placement of alanine residues in the central position of a model coiled-coil can regulate whether it assumes a two or four-stranded structure, while Harbury et al. (1993), using synthetic analogs of GCN4, have shown that subtle modulations in the type of hydrophobic core residues can result in selective formation of dimer, trimer and tetramer structures. Thus, although subtle and varied, the contribution of the ``a'' and ``d'' positions to the ®nal oligomerization state can also be quite signi®cant. We have carried out the complete biophysical characterization of a model coiled-coil protein termed X19a in which 20 different amino acid residues were substituted at position ``a'' to determine their effect on protein stability. As observed by Wagschal et al. (1999a,b), stability differences and, hence, amino acid residue preferences were observed between the 20 different analogs. For example, the b-branched residues of Ile and Val indeed produced the most stable coiled-coil structure, this observation being in excellent accord with the high frequency of occurrence of such residues found within native protein coiled-coil sequences, as well as their predicted preferential packing in such a position. Polar and charged residues resulted in much less stable coiled-coils as expected from their lower hydrophobicity. Interestingly, several residues showed signi®cant deviations from that of their expected contributions to stability based on hydrophobicity, which molecular modelling studies showed was directly due to their ability to pack either more or less ef®ciently within the geometric context of position ``a''. Additionally, because the model coiled-coil protein could adopt both a monomeric twostranded and a trimeric three-stranded state, the oligomeric preferences of each individual amino acid residue were also observed. For example, Asn in position ``a'' directed only the twostranded state similar to that observed for the GCN4 protein (Gonzalez et al., 1996a,b,c; Lumb & Kim, 1995; O'Shea et al., 1991, 1993), while Leu in position ``a'' induced the three-stranded state under benign conditions, in accord with the results of Harbury (1993). Thus, both a rela-

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

379

Figure 1. The model ``d'' coiled-coil. Top; amino acid sequence of one strand of the model ``d'' coiled-coil protein used in this study. Residues involved in forming the 3-4 hydrophobic repeat at position ``a'' and ``d'' are underlined. The substitution site within the sequence is denoted by X and boxed (i.e. residue 22). Isoleucine residues, 19 and 26, which directly ¯ank the site of substitution in the folded state are bolded. Ac, Na-acetyl, amide, Ca-amide; the cysteine residue used for disul®de bond formation is located at position 1. The ®ve heptads forming the 35-residue coiled-coil are numbered 1 through 5. The positions of the heptad are labelled ``gabcdef''. Middle; end-view of the model d protein starting from the NH2 terminus at residue 4. The direction of propagation of the helices is into the page from NH2 to COOH terminus with the polypeptide chains parallel and in-register. Residues in the ®rst two helical turns are circled. Heptad positions are labeled a-g, with the prime indicating corresponding positions on the opposing helix. Arrows depict the hydrophobic interactions that occur between residues in the ``a'' and ``d'' positions (a to a0 and d to d0 ). Solid arrows represent i to i0 ‡ 5 electrostatic attractions which can occur between the ``g'' position of one helical strand and the ``e'' position of the adjacent helical strand (e.g. positions 4 to 90 , 11 to 160 , and 32 to 370 ). Bottom; side-view. Potential electrostatic interactions occurring across the hydrophobic face are indicated by solid arrows. The site of substitution X (position 22d) within the hydrophobic core as well as the hydrophobic residues at position a which directly ¯ank the site are colored.

tive stability scale and a relative oligomeric preference scale of the 20 amino acid residues in the a position were acquired. Although these results are directly amenable to further interpretation of the speci®c roles of various amino acid residues occurring in position ``a'' of native coiled-coils, they are not directly transferable to those in the hydrophobic core ``d'' position, due in part to the latter's structurally unique packing environment. Thus, the goal here was to repeat the same 20 amino acid substitutions, but this time in a hydrophobic core ``d'' position. In this regard, we re-designed the original model X19a coiled-coil sequence to

contain the substitution site at a central ``d'' position. In addition, we substituted the surrounding ``e'' and ``g'' residues to generate a similar micro-environment to that of model peptide X19a; thus these changes should allow for as direct as possible a comparison between these two protein data sets. Furthermore, by keeping the remainder of the sequence identical to that of the previous model, the new model ``d'' coiled-coil protein should similarly undergo two and three-stranded oligomeric switching, and hence the oligomeric preference as well as the stability data could be obtained.

380

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Results De novo design of the model ``d'' coiled-coil sequence Figure 1 shows the amino acid sequence of one strand of the model ``d'' coiled-coil protein. Residues constituting the hydrophobic core ``a'' and ``d'' positions are underlined and the guest site at which the 20 different amino acid residues were substituted is boxed and denoted by an X. The model sequence consists of 38 residues, 35 of which are directly involved in forming the a-helical coiled-coil structure. A Cys-Gly-Gly linker was incorporated at the N terminus to allow formation of an intermolecular disul®de bridge between two polypeptide chains (to form a two-stranded coiledcoil structure), while the two glycine residues (at the N terminus) added conformational ¯exibility and distance between the disul®de bridge and coiled-coil region. The hydrophobic core of the model sequence was designed to contain valine and leucine residues in the ``a'' and ``d'' positions, respectively (Ile residues at positions 19 and 26 are discussed below). The choice for these residues was based in part on the earlier results of Wagschal et al. (1999a,b), which showed that a hydrophobic core containing such residues can generate a suf®ciently stable coiled-coil to carry out the desired substitutions, yet be structurally heterogeneous such that a core containing these residues can also adopt either a two or threestranded oligomeric state. Thus, the ability of the substituted guest site residue to in¯uence both the stability and the association state can be measured. Positions ``b'' and ``c'' of the peptide sequence contained Gly and Ala residues, respectively, while position ``f'' contained either Ala (two positions) or Lys (three positions) residues. The three Lys residues were introduced to increase solubility, discourage aggregation and maintain a net-positive charge on each of the coiled-coil strands, regardless of the amino acid residue substituted at the guest site. Position 22d, in the 3rd heptad, was chosen as the site of substitution, since perturbing the hydrophobic core in this region has been shown to have the greatest effect on coiled-coil stability and oligomerization state (Zhou et al., 1992a; Harbury et al., 1993; Wagschal, et al., 1999a,b). In addition, the central location allows for the assessment of stability changes independent of end effects, any interactions of the various residues (particularly the charged ones) with the macro helix dipole, and minimizes the helix-capping effects of the various substituted residues. Figure 1, middle and bottom, show the helical wheel and side-view diagrams of the two-stranded model ``d'' coiled-coil protein, respectively. The two-headed open arrows in the hydrophobic core depict van der Waals interactions between ``a'' and ``d'' residues on adjacent chains (a to a0 and d to d0 ). The two-headed solid arrows show the location

of electrostatic salt-bridges between Lys and Glu residues at positions ``e'' and ``g'' of adjacent ahelices (i to i0 ‡ 5 or g to e0 ) designed to stabilize the coiled-coil structure further, as well as to orient the polypeptide chains in an in-register and parallel manner in conjunction with the N-terminal disul®de bond. As indicated in Figure 1, bottom, the hydrophobic interfacial environment surrounding the substitution site is de®ned by Ile19 and Ile26. Ile residues were incorporated at the adjacent ``a'' positions (as opposed to Val residues) in order to match isosterically the same type of hydrophobic environment (Leu residues) previously designed in the X19a model coiled-coil protein (Wagschal et al., 1999a) as well as to mimic the most frequent structural environment observed in native coiled-coil sequences (Cohen & Parry, 1990; Lupas et al., 1991). In addition, Gln residues were strategically placed at positions 18g, 23e, 25g, and 30e to provide a moderately polar yet uncharged microenvironment located directly adjacent to the guest site, thus precluding the formation of interhelical e-g0 (or e0 -g) salt bridges around the substitution site as well as avoiding potential charge-charge electrostatic interactions between the side-chains of these residues with the side-chains of the guest site residues. Finally, the N and C termini of the individual polypeptide chains were acetylated and amidated in order to eliminate both charge-charge repulsions at the termini as well as possible unfavourable electrostatic interactions with the a-helix dipoles (Shoemaker et al., 1987).

Secondary structure analysis To determine the effect of each guest site amino acid residue on the secondary structure of the model ``d'' coiled-coil protein, the analogs were analyzed by far-ultraviolet CD spectroscopy (Table 1) in benign buffer conditions and 50 % (v/ v) TFE. Each analog, with the exception of proline, showed a CD spectrum characteristic of a-helical proteins (Chen et al., 1974), i.e. a maximum at 192 nm and a double minima at 208 and 222 nm (data not shown). The mean residue molar ellipticity average for all analogs was ÿ30,800 deg. cm2 dmolÿ1, indicating essentially all analogs were >95 % in the fully folded state as compared to a theoretical value of ÿ34,700 degrees for a 100 % ahelical 35 residue peptide (Chen et al. 1974). Moreover, the ratio of molar ellipticity of 222 nm/ 208 nm for each analog, often used as an indication of association, showed values greater than unity (with the exception of the Arg, Orn and Pro analogs), indicative of fully folded coiled-coil structures (Engle et al., 1991; Lau et al., 1984; Monera et al., 1993; Zhou et al., 1992a,b). Thus, the observation of high molar ellipticity values and similar 222/208 ratios suggests that all of the residues (with Pro being the only exception) could be incorporated into the model ``d'' coiled-coil scaffold

381

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil Table 1. Biophysical characterization of the model ``d'' analogs [y]222 Substituted amino acida Leu Met Ile Tyr Phe Val Gln Ala Trp Asn His Thr Lys Ser Asp Glu Arg Gly Orn Pro

b

Benign 50% TFE (degrees
cm2
dmolÿ1) ÿ32,530 ÿ33,050 ÿ32,520 ÿ34,350 ÿ32,300 ÿ32,770 ÿ34,150 ÿ30,610 ÿ30,350 ÿ31,540 ÿ30,020 ÿ33,240 ÿ30,660 ÿ31,730 ÿ33,070 ÿ24,830 ÿ21,880 ÿ27,990 ÿ16,560 ÿ4592

ÿ28, 820 ÿ29,380 ÿ28,800 ÿ29,020 ÿ28,190 ÿ24,720 ÿ30,700 ÿ23,780 ÿ26,840 ÿ20,760 ÿ25,010 ÿ28,240 ÿ26,030 ÿ31,260 ÿ23,150 ÿ23,400 ÿ23,870 ÿ22,890 ÿ25,360 ÿ17,510

[y]222/208 1.04 1.08 1.06 1.04 1.06 1.06 1.06 1.06 1.05 1.09 1.09 1.06 1.01 1.06 1.05 1.01 0.98 1.06 0.88 0.49

c

[GdnHCl]1/2 (M) 3.76 3.32 3.25 2.31 2.21 2.20 1.90 1.64 1.58 1.38 1.26 1.17 0.89 0.82 0.78 0.32 0.35 0.31 0.13

d

me 1.54 1.71 1.69 1.98 1.98 1.74 1.94 2.04 2.08 2.50 2.33 2.94 2.65 2.64 2.15 2.10 2.43 3.34 2.02



Gu(Ala)f (kcal molÿ1) 3.8 3.2 3.0 1.4 1.2 1.1 0.5 0.0 ÿ0.1 ÿ0.6 ÿ0.8 ÿ1.2 ÿ1.8 ÿ1.8 ÿ1.8 ÿ2.7 ÿ2.9 ÿ3.6 ÿ3.1

tR g (min) 67.4 63.9 67.5 58.8 65.4 64.5 52.3 59.6 64.7 51.4 53.3 57.1 53.3 54.0 48.3 47.9 53.1 53.0 51.7 52.4


tR(Ala)h (min)

tR i

7.8 4.3 7.9 ÿ0.8 5.8 4.9 ÿ7.3 0.0 5.1 ÿ8.2 ÿ6.3 ÿ2.5 ÿ6.3 ÿ5.6 ÿ11.3 ÿ11.7 ÿ6.5 ÿ6.6 ÿ7.9 ÿ7.2

2 6 1 8 3 5 15 7 4 18 11 9 12 10 19 20 14 13 17 16

a Amino acid sequence for the model ``d'' protein is shown in Figure 1. Position 22 was substituted with each of the listed amino acid residues. b The mean residue molar ellipticities at 222 nm were measured at 25  C in benign buffer (0.1 M KCI, 0.05 M PO4 (pH 7)). For samples containing trifuoroethanol (TFE), the above buffer was diluted 1:1 (v/v) with TFE. Peptide concentrations were 100 mM. c The ratio [y]222/208 was calculated by dividing the observed molar ellipticity value at 222 nm ([y]222) by the observed molar ellipticity value at 208 nm ([y]208) in benign buffer. d [GdnHCI]1/2 is the transition midpoint, the concentration of guanidine hydrochloride (M) required to produce a 50 % decrease in molar ellipticity at 222 nm between the folded and unfolded states. e m is the slope in the equation
Gu ˆ
GuH2 O ÿm[GdnHCl] with units of kcal molÿ1 Mÿ1. f

Gu(Ala) is the difference in free energy of unfolding relative to the Ala-substituted peptide.

Gu(Ala) was calculated using the equation:

Gu ˆ ([denaturant]1/2,X-[denaturant]1/2,Ala)(mX ‡ mAla)/2 (Serano & Fersht, 1989), where X is the substituted amino acid. Positive values denote that the analog with the substituted residue is more stable than that of the Ala analog; negative values denote that the analog with the substituted residue is less stable than that of the Ala analog. The calculated
GuH2 O for the Ala substituted peptide was 3.03 kcal/mol. g Observed RP-HPLC retention time for each carboxamidomethylated model d peptide analog at pH 6.5; see Material and Methods for details. h
tR(Ala) is the difference in RP-HPLC retention time of each single-stranded analog relative to the Ala-substituted analog. Positive values indicate an increase in retention time relative to the Ala analog. i Ranked order of hydrophobicity of the 20 amino acid analogs based on RP-HPLC; 1 denotes most hydrophobic and 20 denotes least hydrophobic.

sequence with minimal perturbation to the native helical structure. In the case of proline, its presence within the guest site essentially caused a complete loss in the a-helical content of the coiled-coil, indicating proline residues are completely unacceptable within a central ``d'' position. Even in the presence of 50 % TFE, only 50 % of the helical content could be restored. Hence, the absence of ahelical content in benign buffer conditions precluded the ability of the Pro analog to be compared further with any of the other analogs, and thus is not mentioned hereafter. Despite their apparent similarities, analysis of the molar ellipticities of the folded structures do reveal subtle and interesting differences (Table 1). For example, in examining the positively charged side-chain residues of Lys, Arg and Orn, only the Lys analog displayed essentially a fully helical structure. Movement of the positive charge closer to the hydrophobic core by one methylene group, as occurs with the Arg and Orn side-

chains, resulted in signi®cant decreases in the level of helicity of these analogs (69 % and 52 % of the Lys analog, respectively), suggesting the presence of either unfavourable packing interactions with the hydrophobic core or a higher energetic cost of their desolvation near the core. Interestingly, however, this trend was not apparent with the negatively charged residues. The Glu substituted analog contained 75 % of the predicted theoretical helical content, while movement of the charged carboxyl group one methylene group closer to the core, i.e. the Asp analog, resulted in a fully helical state. Thus, the fully folded state of the Asp analog compared to the Glu analog suggests the presence of favorable structural interactions for the Asp analog, which offsets the closeness of the charged carboxyl group to the hydrophobic core. In the presence of TFE, all of the model ``d'' analogs, with the exception of the Arg and Orn analogs, showed a decrease in the 222 nm value

382

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

and a decrease in the 222/208 nm ratio to 0.8, indicating all of the structures are making a transition from a two or three-stranded folded state in benign buffer to a single-stranded a-helical state (Cooper & Woody, 1990; Lau et al., 1984; SoÈnnichsen et al., 1992). The signi®cant variation in the ellipticity values observed in benign versus TFE containing medium, e.g. the ellipticity values of the Asn and Asp analogs were 10,000  less in the presence of TFE compared to benign medium, indicates that these residues are clearly more dependent on the packing of the helices to remain in a helical conformation than when in a single-stranded state.

Guanidine-HCl denaturation studies of the model ``d'' analogs To determine the effect of each guest site residue on the stability of the coiled-coil structure, each model ``d'' coiled-coil protein was denatured using GdnHCl and the ellipticity at 222 nm (re¯ective of the a-helical content of the structure) monitored by CD spectroscopy. Figure 2, (a) and (b), show overlay plots of the GdnHCl denaturation pro®les for 18 of the model ``d'' analogs. All of the analogs displayed a sigmoidal denaturation pro®le indicative of a two-state, single transition, cooperative unfolding process. Comparison of the GdnHCl midpoints (see Table 1) for each transition showed

Figure 2. (a) and (b) GdnHCI denaturation pro®les of the model ``d'' analogs. Denaturations were carried out at 25  C in 0.1 M KCl, 0.05 M PO4 (pH 7) buffer with GdnHCl. The fraction folded ( ff) of each peptide was calculated as described in Materials and Methods. Each peptide was analyzed at 100 mM concentration.

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

that, although the mean residue molar ellipticity values were similar under benign buffer conditions for the 19 analogs, the substituted residue caused a great variation in the apparent stability of the model ``d'' coiled-coil. In general, analogs containing hydrophobic amino acid substitutions (Leu, Met, Ile, Tyr, Phe and Val) showed the greatest GdnHCl midpoints, followed by the analogs with polar amino acid substitutions (Gln, Asn, Thr and Ser) and, ®nally, the analogs containing charged residue substitutions (Lys, His, Asp, Glu and Arg). Two analogs exhibited substitutions notably different than expected: the Trp analog clearly demonstrated a lower than expected GdnHCl midpoint for its size and hydrophobic character (i.e. it behaved in a similar manner to the analog substituted with the small hydrophobic side-chain Ala); and the Asp analog displayed a higher GdnHCl midpoint than the Glu analog despite the Glu residue having slightly greater hydrophobicity due to an extra methylene group in its side-chain. Hydrophobicity The stability of coiled-coils has classically been related to the hydrophobicity of the side-chains existing in the hydrophobic core ``a'' and ``d'' positions. To compare and contrast differences between the observed stability of a side-chain and its level of hydrophobicity in the corresponding ``d'' position, we determined the hydrophobicity of the single strand of each model ``d'' peptide analog using reversed-phase high-performance liquid chromatography (RP-HPLC). The cysteine residue on each peptide was carboxamidomethylated to ensure the peptides were single-stranded during chromatography. A summary of the observed elution times (tR), their ``relative elution times''

383

compared to that of the Ala analog (
tR(Ala)), and their ranked order from most hydrophobic to least hydrophobic (t#R), is shown in Table 1. The RPHPLC method conditions employed (see Materials and Methods for details) resulted in all 20 analogs being well resolved with a 19.5 minute difference in retention time between the most hydrophilic analog (Glu tR ˆ 47.9 minutes) and the most hydrophobic analog (Leu tR ˆ 67.4 minutes). The relative hydrophobicity values of the 20 amino acid side-chains, as expressed by the RP-HPLC retention times of the 20 analogs relative to the Ala analog, appear to be essentially the same as those reported using octanol water partitioning of acetyl amino acid amides (Fauchere & Pliska, 1983; Eisenberg & McLachlan, 1986) as well as other model systems including our own X19a model peptide sequence (Wagschal et al., 1999b and references therein). For example, a plot between the observed
tR(Ala) times of this study and those obtained in the previous study (
tR(Ala) ``a'' position) for the X19a model sequence showed a correlation of r ˆ 0.99 (Figure 3). Thus, the hydrophobicity values derived for each side-chain are largely an intrinsic parameter of the side-chain itself and not model or position dependent. Oligomerization To assess whether the substituted guest site residue in¯uences the association state of the model ``d'' coiled-coil protein in a manner similar to that previously observed for the X19a model coiled-coil protein (Wagschal et al., 1999a,b), each of the 20 model ``d'' analogs were analyzed by highperformance size-exclusion chromatography (HPSEC). We had showed that HPSEC offers a reliable and quick approach for establishing the

Figure 3. A plot of the relative RP-HPLC retention times of the model ``d'' analogs versus those obtained from the X19a model sequence (Wagschal et al., 1999b). The cysteine residue side-chain (position 1) was alkylated with iodoacetamide to prevent disul®de bridge formation. The change in retention time (
tR) relative to the alanine analog is plotted for each residue at each position. Least squares ®t analysis showed a correlation of R ˆ 0.99. For RP-HPLC conditions see Materials and Methods.

384

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

oligomerization state as well as respective ratios between monomer and higher-order oligomeric states, if present (Wagschal et al., 1999a). This is due in part to the observation that such association states exchange relatively slowly in time and hence allow for their respective separations during the chromatographic run. Thus, HPSEC runs were carried out by injecting each analog in benign buffer from either a 500 mM stock solution or a tenfold diluted sample (50 mM, a similar concentration to that used for CD measurements). Figure 4(a) shows representative HPSEC chromatograms for the three model ``d'' analogs Val, Glu and Ser. In general, the model ``d'' analogs were eluted from the size-exclusion column as either single peaks with apparent molecular masses corresponding to the two-stranded disul®de bridged monomer (denoted M; 7000 Da), the three-stranded trimer (denoted T; 21,000 Da, where each strand is

made up of two polypeptide chains joined by a disul®de-bridge (Wagschal, 1999a) or mixtures of monomeric and trimeric species (e.g. Glu, Val and Ser analog chromatograms, respectively; Figure 4(a)). No peaks greater or intermediate in apparent molecular mass (i.e. tetramers or dimers, respectively) were observed. The observation of monomeric and/or trimeric states depending on the substituted amino acid residue indicates, ®rst, that the model ``d'' coiled-coil protein is indeed similar to that of the model X19a coiled-coil protein in that a hydrophobic core consisting of three Val and two Ile residues in the ``a'' position and four Leu residues in the ``d'' position is structurally heterogeneous and can adopt both a two-stranded monomeric or a three-stranded trimeric state; and second, that a single residue substituted at a ``d'' position can in¯uence the ®nal association state. Thus, the effects of each of the 20 amino acid side-

Figure 4. (a) Representative HPLC size-exclusion chromatograms for the model d peptide analogs Glu, Ser and Val. Each chromatogram was acquired in benign buffer (0.1 M KCl, 0.05 M PO4 (pH 7)) at a ¯ow rate of 0.2 ml/minute on a Superdex 75 column. M and T denote monomeric and trimeric molecular mass species, respectively. Each run consisted of a 100 ml injection of a 0.25 mg/ml (65 mM) solution of the corresponding) analog. (b) Representative HPLC size-exclusion chromatograms for the model ``d'' analog Ser in the presence of increasing GdnHCI concentrations. Each chromatogram was acquired in benign buffer (0.1 M KCl, 0.05 M PO4 (pH 7) containing 0 to 2.0 M GdnHCl) as denoted in the Figure. Tf denotes folded trimer; Mf and Mu denote the two-stranded disul®de-bridged folded and unfolded monomer, respectively. Each run consisted of a 100 ml injection of a 0.25 mg/ml (65 mM) solution of the Ser analog in the above running buffer. All runs were carried out at ambient temperature. S1 and S2 are random coil peptide standards (see Material and Methods).

385

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

depending on their concentration. It is noteworthy that the Asn analog, which clearly adopts only the two-stranded monomeric state in the ``a'' position, is completely devoid of this effect in the ``d'' position. To reinforce the accuracy of the HPSEC data, several of the model ``d'' analogs were analyzed by sedimentation equilibrium (SE) ultracentrifugation. Representative curve ®ts corresponding to the Val, Ser and Lys analogs are shown in Figure 5. As observed by HPSEC, the Val analog data were best ®t to a single species ®t corresponding in molecular mass to a three-stranded coiled-coil; the Lys data were best ®t to a single species ®t corresponding in molecular mass to a two-stranded monomeric model; while the Ser data were best ®t to a monomer to trimer equilibrium ®t. Thus, the de®nition of oligomeric states and their relative variation for these peptides by sedimentation equilibrium analysis appear to be in excellent agreement with the results observed by HPSEC. Moreover, SE data generated for several of the other analogs (data not shown) were also in excellent agreement with that observed by HPSEC, thus strongly reinforcing the HPSEC data.

chains in their ability (or inability) to de®ne a single oligomeric state can be assessed. Table 2 shows a comparison of the oligomerization data derived from the integrated areas of each eluted peak as a percentage of total area generated by each protein observed by HPSEC (Figure 4(a)) with those previously reported for the same substitutions in the X19a model coiled-coil protein (Wagschal et al., 1999a,b). As observed, all of the charged residues at position ``d'' with the exception of His clearly de®ned only the monomeric two-stranded state. Even at a tenfold higher injected concentration (data not shown), no higherorder oligomeric states were observed for any of these analogs, suggesting that the presence of the charged group clearly destabilizes the threestranded state. In contrast, the analogs containing b-branched residues of Ile, Val and Thr at position ``d'' demonstrated a strong tendency to populate exclusively the three-stranded state, suggesting the greater burial of surface area in an ``acute'' packing arrangement (rather than perpendicular) is apparently more favorable for these residues in the three-stranded state compared to the two-stranded monomeric state. It is interesting that several analogs (i.e. those containing His, Asn, Gln, Ser, Trp, Phe, Gly, Ala, Met or Leu at position ``d''), were indiscriminate in their ability to de®ne clearly a single oligomeric state, and thus adopted various ratios of the monomeric and trimeric states

Trimer to monomer switch A critical issue in being able to compare directly the stability data obtained for each of the model

Table 2. Oligomerization data Position ab Substituteda amino acid Leu Ile Val Met Ala Gly Phe Tyr Trp Thr Ser Gln Asn Lys Orn Arg His Glu Asp Pro

Position dc

% Trimerd

% Monomerd

% Trimerd

% Monomerd

100 39 57 90 38 35 65 100 0 76 25 100 0 0 0 0 100 46 0 -

0 61 43 10 62 65 35 0 100 24 75 0 100 100 100 100 0 54 100* -

31 100 100 80 39 50 46 0 22 90 42 39 72 0 0 0 45 0 0 -

69 0 0 20 61 50 54 100 78 10 58 61 28 100 100 100 55 100 100 -

a Amino acid residue substituted at position 22d of the model ``d'' protein (Figure 1) or position 19a of the X19a model protein (Wagschal et al., 1999a,b). b Observed percent oligomerization states of the model ``a'' peptide analogs at 50 mM peptide concentration in benign buffer (0.1 M KCl, 0.05 M PO4(pH 7) using HPSEC (Wagschal et al., 1999a,b). c Observed percent oligomerization states of the model ``d'' peptide analogs at 50 mM peptide concentration as observed by HPSEC chromatography in this study (see Materials and Methods for details). d Monomer represents disul®de-bridged two-stranded a-helical coiled-coil (7000 Da) and trimer represents a three-stranded coiled-coil consisting of three disul®de-bridged proteins (21,000 Da) as depicted in Wagschal et al. (1999a,b). *The Asp analog was 100 % folded ([y]222 ˆ ÿ 32,300) and monomeric at 5  C. At room temperature, the analog was only 20 % folded ([y]222 ˆ ÿ 6200).

386

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Figure 5. A plot of concentration versus radial distance from sedimentation equilibrium analysis of the model ``d'' analogs Val, Ser and Lys peptides performed at 18,000 and 26,000 rpm. The Val data was best ®t to a single species model with a trimeric molecular mass (21,000 Da). The Ser data was best ®t to an associating monomer to trimer equilibrium model. The Lys data was best ®t to a single species model with a monomeric molecular mass (7000 Da). Plots of the residuals from each curve ®t are shown above.

``d'' analogs is that each of the analogs is required to be denatured from the same oligomeric state. We had showed that in the case of the model X19a analogs (Wagschal et al., 1999a,b), the variation in associative states observed initially in benign buffer medium was eliminated upon addition of denaturant. That is, all peptides in the trimer state or mixtures of trimers/monomer states changed to fully folded two-stranded monomeric states prior to unfolding. To establish that this is also the case for the model ``d'' analogs, several of the heterogeneous and trimeric analogs where subjected to HPSEC in the presence of increasing concentrations of GdnHCl prior to, and through, their denaturation transition areas. A representative series of elution pro®les for the Ser analog is shown in Figure 4(b). Similar to that observed for the model

X19a analogs, the Ser analog (initially occurring as 50 % trimer (Tf):monomer (Mf)) completely reverts to the two-stranded monomeric state prior to a concentration of 0.8 M GdnHCl. As the GdnHCl concentration is further increased to 2.0 M, the retention time of the Ser analog peak gradually decreases until the retention time of the unfolded monomer state (denoted Mu) is reached. A plot of the change in retention time of the monomeric peak versus GdnHCl concentration (data not shown) showed a transition midpoint value of 0.8 M, in excellent agreement with that observed by CD spectroscopy. Similar results to these were also observed for several of the other model ``d'' analogs which populated various amounts of the trimeric state. Thus, HPSEC suggests that all of the

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

heterogeneous or trimeric analogs are being denatured from the monomeric state. It is worth mentioning that, although the primary basis of the above investigation was to verify that the higher oligomeric species reverted back to their monomeric state, several of the monomeric analogs were also tested to verify that they did not adopt a higher order oligomeric state upon denaturation; no such change was observed. Concentration independence Since size-exclusion chromatography cannot totally rule out the possibility that some of the trimeric analogs could unfold directly to the unfolded monomer as a function of GdnHCl concentration, we therefore also tested the concentration dependence of the GdnHCl midpoint of several of the analogs. That is, one of the hallmark features of higher order associating systems is that their denaturant midpoint values are dependent on concentration. Thus, several analogs showing monomer-trimer or trimeric oligomeric states were diluted 100-fold (1 mM) further from that analyzed initially by CD (100 mM) and denatured with GdnHCl once again. In each case, the GdnHCl denaturation pro®le (data not shown) was superimposable with that previously observed at a higher concentration, indicating the absence of a concentration dependence and, hence, con®rming that unfolding is indeed occurring from a singlespecies state (native coiled-coil monomer). Therefore, the unfolding pathway for the trimer state is tri-mer to native coiled-coil monomer to denatured monomer. Calculation of the relative stability difference between model ``d'' analogs and their comparison to hydrophobicity Having established that each of the model ``d'' analogs is denatured from the fully folded monomeric state irrespective of its initial oligomerization state in benign buffer conditions, we were now able to treat all of the GdnHCl denaturation data similarly, and hence calculate the difference in free energy of unfolding for each of the analogs using an equation which is concentration independent (see Materials and Methods for details). Table 1 shows the calculated relative stability order in terms of the difference in free energy of unfolding of each of the model ``d'' analogs compared to that of the substituted alanine analog (i.e.

Gu(Ala)). As shown, the Leu analog is the most stable, with a

Gu(Ala) of 3.8 kcal/mol (or 1.9 kcal/mol/Leu). This is followed by analogs substituted by Met, Ile, the aromatic residues Tyr and Phe, and Val and Gln. Analogs which are of lower stability than the Ala analog include (in order of decreasing stability) those substituted by Trp, Asn, His, Thr, Lys, Ser, Asp, Glu, Arg, Orn and Gly. The variation in stability includes a free energy change of 7.4 kcal/mol from the most

387

stable to the least stable analog, and thus clearly demonstrates the importance of the ``d'' position in modulating the overall stability of the protein. In general, the observed order appears to demonstrate the trend that the analogs containing hydrophobic residues are more stable than those containing polar residues, which in turn are more stable than those substituted by charged residues, and thus ®ts the general proposal that hydrophobicity is a major factor in stabilizing proteins. It is interesting to note that despite the large energetic cost due to desolvation when bringing and packing a charged side-chain close to the hydrophobic core interface, the absence of a side-chain, as in the case of Gly, caused the largest destabilization to the coiled-coil structure, i.e. the Gly analog is the least stable next to that of the proline analog, which did not fold at all. In addition to the wide variation in stability values, Table 1 also shows that each of the model ``d'' analogs also exhibit signi®cantly different slope or m-values derived from their plot of
G versus denaturant concentration. In general, the least stable analogs exhibited the greatest m values (2.5-3.3 kcal molÿ1 Mÿ1), while the more stable mutants exhibited the lowest values (1.5-2.0 kcal molÿ1 Mÿ1). The signi®cance of these variations in terms of the folded and unfolded states are discussed below. Stability versus hydrophobicity As described above, it is generally accepted that the hydrophobic effect represents the major force in stabilizing the folded structure of proteins. To compare the relationship between observed stability and hydrophobicity for each of the residues analyzed, a scatter plot was generated. As indicated in Figure 6(a), there is a general trend, but not an excellent correlation between these two parameters (correlation coef®cient of 0.82). In particular, three analogs showed signi®cant deviations from their expected hydrophobicity quadrant (see legend for more details): the Trp, Tyr and Gln analogs. The Trp analog, which is signi®cantly more hydrophobic than the Ala analog, was less stable than the Ala analog or the rest of the analogs substituted by hydrophobic residues. The Tyr analog, which displayed lower hydrophobicity than the Ala analog, is approximately equivalent in stability to Phe and Val analogs of the hydrophobic group, and the Gln analog, a member of the polar group, is clearly more stable than any other member of its group. Such deviations likely re¯ect ``packing effects'' of the side-chain when in the folded coiled-coil structure, which do not exist when the side-chain interacts with the RP-HPLC matrix. For example, the Trp analog likely represents ``poor packing'' in the folded state where its bulky hydrophobic side-chain is not buried within the hydrophobic core, thus lowering its level of stability relative to its inherent hydrophobicity (Figure 7). The Tyr analog, on the other hand, likely re¯ects

388

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Figure 6. Comparison of the relative stability of each substituted amino acid with its relative hydrophobicity for the ``d'' position analogs. (a) A scatter plot of the relative stability versus relative hydrophobicity. The upper righthand quadrant (shaded) represents residues which are of greater hydrophobicity and stability than that of Ala (i.e. the non-polar residues). The lower left quadrant (shaded) represent residues which are of lesser hydrophobicity than that of Ala (i.e. the polar/charged residues). Residues which deviate from either group due to differences in either their hydrophobicity or stability are shown with ®lled points. (b) A histogram displaying the relative stability and hydrophobicity of each analog. Amino acid residues indicated along the x-axis are listed in order of their descending stability. Stability and hydrophobicity values are normalized where Leu ˆ 100 and Ala ˆ 0. The three model ``d'' analogs: Gln, Tyr and Trp which appeared to deviate the most in plot A are indicated with an asterisk.

``good packing'', where the polar hydroxyl group remains solvent accessible, while its aromatic ring is buried in the hydrophobic core, thus making its stability similar to that of the Phe analog despite its relative hydrophobicity being lower. The Gln analog likely re¯ects ``good packing due to hydrogen bonding'', which increases its apparent stability relative to its moderate hydrophobic character. To highlight more the differences between relative hydrophobicity and stability for all of the analogs, the respective data were also plotted in a histogram format where the Leu analog represents a value score of 100 and the Ala analog represents a value score of 0 (Figure 6(b)). As shown, the Met analog, exhibiting the second most stable coiledcoil (Table 1), is clearly more stable than its level of hydrophobicity would have indicated, suggesting favourable packing effects are clearly playing a role in de®ning its position in the relative stability

order. The Ile and Val substituted analogs are less stable than expected based on their hydrophobicity, likely re¯ecting the unfavorability of b-branched residues in the ``d'' position due to their less than optimal van der Waals packing. The Tyr analog (as mentioned above) is more stable than its hydrophobicity would suggest, and interestingly is approximately equivalent to that of the Phe analog, suggesting that packing of the aromatic rings may be similar, or at least independent of, the para-hydroxyl group. The Trp analog is signi®cantly less stable than expected from Trp's hydrophobicity, while the Gln and Asn analogs are clearly more stable than expected from the hydrophobicity of these residues. It is interesting to note that the Gln analog is more stable than the Asn analog in the ``d'' position, whereas this order is reversed in the ``a'' position (Wagschal et al., 1999a,b; Table 3). Finally, the analogs substituted

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

389

Figure 7. Molecular models depicting the side-chain orientations of Met, Trp and Asp residues in position ``d'' of a two-stranded parallel coiled-coil. Side-chains are displayed in stick format with the carbon atoms green and the sulfur atoms yellow; backbone atoms are white with oxygen atoms in red and nitrogen atoms in blue. Van der Waals surface area (dotted spheres) are displayed in yellow and white. Also shown is the predicted packing orientation of the Gln residue side-chains which exist in the adjacent helix e0 positions. Potential hydrogen bonds between the oxygen atoms of the Asp carboxyl group and the hydrogen atom of the Gln amide side-chain group are denoted by broken white lines.

390

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Table 3. Summary of stability and statistical occurrence data Position a Amino acida Leu Met Ile Tyr Phe Val Gln Ala Trp Asn His Thr Lys Asp Ser Glu Arg Orn Gly Pro



Gu(Ala)b 3.5 3.4 3.9 2.2 3.0 4.1 ÿ0.1 0.0 0.8 0.9 ÿ1.2 0.2 ÿ0.4 ÿ1.3 ÿ2.0 ÿ0.8 ÿ1.9 ÿ2.5 -

Position d

Normalizedc stability

Normalizedd occurrence

100 98 105 74 89 108 41 43 55 56 28 50 37 23 10 31 10 0 -

100 70 82 45 17 53 6 41 8 26 11 5 43 1 12 8 21 1 0



Gu(Ala)b 3.8 3.2 3.0 1.4 1.2 1.1 0.5 0.0 ÿ0.1 ÿ0.6 ÿ0.8 ÿ1.2 ÿ1.8 ÿ1.8 ÿ1.8 ÿ2.7 ÿ2.9 ÿ3.1 ÿ3.6 -

Normalizedc stability

Normalizedd occurrence

100 91 89 67 64 63 56 49 47 41 37 33 25 24 24 12 9 7 0 -

100 36 23 43 17 24 16 67 12 1 10 17 5 6 11 26 1 6 0

a Amino acid residue substituted at position 19a of the X19a model protein (Wagschal et al., 1999a,b) or position 22 in the model ``d'' protein (this study). b

Gu(Ala) is the difference in the free energy of unfolding (
Gu) relative to the Ala-substituted analog. c Normalized stability (NS) represents the stability of each substituted analog relative to Gly ˆ 0 and Leu ˆ 100. Values were calculated from the equation: NS ˆ (

Gu(Ala)X ‡

Gu(Ala)Gly/

Gu(Ala)Leu ‡

Gu(Ala)Gly)  100, where X denotes the stability of any analog. d Normalized occurrence was calculated as (fx/fLeu)  100, where fX is the relative frequency of each amino acid occurring in either an ``a'' or ``d'' position of the coiled-coil repeat, and fLeu is the relative frequency of Leu occurring in that same position. Relative frequency represents the frequency of occurrence of each amino acid in that position divided by the overall frequency of occurrence of each amino acid in the GenBank (Lupas et al., 1991).

by charged residues (His, Lys, Asp and Glu), with the exception of Arg, are all signi®cantly more stable than their hydrophobicity would have suggested. Particularly notable within this group is the observation that the Asp analog is more stable than that of the Glu analog, despite Glu containing one more methylene group (as mentioned above). In general, the lack of a clear correlation between stability and hydrophobicity for many of the ``d'' analogs clearly underscores the importance of the stability data obtained above for future prediction and analysis of coiled-coils as opposed to hydrophobicity data. Packing analysis of Met, Trp, and Asp To investigate the structural features which give rise to the apparent deviations between hydrophobicity and stability of the model ``d'' analogs, molecular modelling studies of each of the 20 analogs were carried out. A complete description of these results are to be presented elsewhere (P.L. & R.S.H., unpublished results), but for the purpose of this study, molecular modelling analysis of three residues: Met, Trp, and Asp which clearly demonstrate ``good'', ``poor'' and ``hydrogen bonding'' packing effects are reported. In the case of methionine, which was observed to impart a greater stability than would be expected from its observed hydrophobicity, molecular modelling analysis (Figure 7) showed that

the methionine residue side-chain can be packed at a ``d'' position with dihedral angles which, although not identical to leucine, can easily emulate it in creating excellent geometric complementarity and van der Waals packing interactions between the adjacent d0 , a, a0 , e0 and g residues. Furthermore, the side-chain packing arrangement also brings the sulfur atoms of the side-chains close to the core. Because sulfur has a greater level of polarizability than that of carbon, it has been proposed that this could lead to greater van der Waals interactive strengths between the side-chain and the core (Nelson & Chazin (1998), and references therein). In addition, analysis of three crystal structures in which methionine residues are observed in a ``d'' position of a homodimeric parallel two-stranded coiled-coil (PDB# 3GAP, 1HLO and 1CGP) showed that the side-chain can adopt two different stereochemical packing arrangements, thus suggesting the side-chain can possibly ¯ip between these two arrangements and hence maintain a signi®cant degree of conformational entropy. Thus, the higher than expected stability of the Met analog appears to be due to its unique ability to adopt excellent geometric packing and van der Waals interactions as well as maintain a certain degree of conformational freedom. Molecular modelling of Trp, on the other hand, shows quite the opposite. As opposed to being buried within the hydrophobic core, the side-chain of Trp extends directly out of the core (Figure 7).

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Although this side-chain conformation does not necessarily create a cavity in the core, it is clearly unable to pack its largest hydrophobic part, the indole ring, into the core, thus leaving it exposed to the solvent, and hence unable to be utilized. The modelling of Trp is further reinforced by the high resolution X-ray crystallographic structure of dimeric rat brain kinesin, which shows two Trp residues in ``d'' positions of the homo twostranded coiled-coil neck region (PDB# 3kin). In each case, the Trp residue adopts either a parallel or perpendicular orientation relative to the helical axis of the coiled-coil but, in both of these arrangements, the large indole ring is still largely solvent exposed. Thus, the lower than expected stability of the Trp analog relative to its hydrophobicity is clearly a result of its poor packing in a hydrophobic core ``d'' position. Molecular modelling studies of the Asp analog show that the side-chain can be packed entirely within the hydrophobic core environment, with the buried carboxyl groups making intermolecular hydrogen bonds with the adjacent g0 position Gln residue (Figure 7). Thus, the high cost of desolvation of the carboxyl groups appears to be compensated in part due to the formation of new intermolecular hydrogen bonds with the adjacent ``g'' positions. Interestingly, the packing arrangement of the Asp side-chain is almost completely isosteric with that of the packing arrangement of Leu, with the exception that the d methyl group is replaced by a carbonyl group. Thus, Asp's unique contribution to stability (i.e. greater than Glu) to a ``d'' position can be ascribed to its unique ability to maximize the use of its methylene component in the hydrophobic core and its hydrophilic component through hydrogen bonding adjacent to the core. Protein stability versus natural occurrence at position ``d'' Having derived a relative stability scale for the 20 amino acid residues, we then compared their stability magnitude and order to the apparent frequency of occurrence for each amino acid residue occurring in a ``d'' position in naturally occurring coiled-coils (Cohen & Parry, 1990; Lupas et al., 1991). As observed in Figure 8 and Table 3, the Leu analog, which represents the most stable analog, also occurs with the greatest statistical frequency in the ``d'' position of native coiled-coils. Similarly, the Arg and Gly analogs, which were the least stable analogs, occurred least often. Comparison of the 17 other residues, however, indicates that, in general, there is no direct relationship between stability and frequency of occurrence. For example, Ala, which was only intermediate in its contribution to stability, occurred with the second greatest frequency. Owing that Ala does not discriminate a particular oligomeric state (Table 2), its high frequency of occurrence in comparison to that of the other residues may be due in part to its

391

ability to modulate coiled-coil stability without affecting the choice of two or three-stranded coiled coils. The signi®cant deviations in occurrence of the other residues (i.e. lower occurrence than may have been predicted from their contribution to coiled-coil stability) may be due in part to the high metabolic cost associated with their biosynthesis (e.g. Trp) or their ability, or inability, to control the association state. In general, the lack of a clear correlation between stability and frequency of occurrence for many of the analogs clearly underscores the importance of the stability data obtained above for de novo design applications, the creation of algorithms designed to predict the occurrence and stability of coiled-coils in naturally occurring peptides and proteins, and our ability to understand better the function/structural role of different residues in coiled-coils. Comparison of the ``d'' position versus ``a'' position stability data Comparison of the crystal structures of coiledcoils has revealed two distinct types of packing arrangements at the ``a'' and ``d'' positions of twostranded coiled-coils. In the case of the ``a'' position, the Ca-Cb vector projects out of the interior of the core roughly parallel to the Ca-Ca vector, and thus is termed ``parallel packing''. In contrast, at a ``d'' position, the Ca-Cb vector projects roughly perpendicular to the Ca-Ca vector and thus is termed ``perpendicular packing''. As a result of these steric differences, differences in the frequency of occurrence and contribution to coiledcoil stability of certain residues has been ascribed to their ability to pack better in each of these positions. For example, Harbury et al. (1993) reported that parallel packing arrangements in the ``a'' position are more favorable for the b-branched residues (e.g. Ile, Val and Thr), where the methyl group projecting from the b-carbon atom is able to pack back and into the centre of the hydrophobic core. In contrast, the perpendicular packing arrangement observed in a ``d'' position is more ideal for leucine residues, where, although the g-carbon atom moves out of the core, one of the d-carbon atoms can pack back into the hydrophobic core, resulting in excellent geometric complementarity and packing between the symmetry-related side-chains. To compare and contrast the differences between the

Gu(Ala) stability values obtained herein with those obtained from the model ``a'' study (Wagschal et al., 1999b), and thus highlight more clearly the thermodynamic preferences of the individual residues in either an ``a'' or ``d'' position, a histogram plot depicting the relative stability values for each data set was plotted. As shown in Figure 9(a), large differences in stability for the individual residues in the two positions are observed, and thus allude to positional stability preferences. For example, the b-branched residues of Ile, Val and Thr are clearly more favored in an

392

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Figure 8. Histograms comparing the measured relative stability and relative statistical occurrence (Lupas et al., 1991) of each amino acid residue in a ``d'' position. Stability and statistical occurrence data have been normalized such that Leu equals 100, and Gly equals 0 for the stability data (see Table 2 and footnotes for more details).

``a'' position. Leu is more stable in the ``d'' position. In the case of the charged residues, Lys and Arg are clearly more destabilizing in the ``d'' position compared to that in an ``a'' position. This is interesting, since both Lys and Arg can initially pack their aliphatic portions to emulate Leu. Thus, their signi®cantly greater destabilization in a ``d'' position must re¯ect the closer distance of the charged group to the hydrophobic core. Other notable differences include Asn and Gln residues, where Asn clearly confers a greater level of stability than Gln in the ``a'' position, while this difference is completely reversed in the ``d'' position. Previously, it has been noted that an Asn residue can potentially form an extra hydrogen bond within the core in an ``a'' position, thus likely adding to its stability preference in the ``a'' position (O'Shea et al., 1991, 1993; Gonzalez et al., 1996a,b,c; Lumb & Kim 1995). Molecular modelling studies indicate this is not possible in the ``d'' position and thus the extra methylene group of Gln and the greater distance of the amide group from the core is likely responsible for the greater stability of Gln at the ``d'' position. To highlight further the difference between the stability of the amino acid residues in an ``a'' and ``d'' position, the relative stability of each residue in a ``d'' position was subtracted from the corresponding relative stability in an ``a'' position. As shown in Figure 9(b)) and Table 4, almost all residues, with the exception of Leu, Gln, His and Asp, contributed more to stability in the ``a'' position. Thus, these results are in excellent agreement with proposals which stated that the packing arrangement at an ``a'' position, because it directs the side-

chain away from the hydrophobic core, is much more permissive than that in the ``d'' position, and hence most residue side-chains are not as destabilizing (Hu et al., 1990). Residues showing a signi®cant stability preference for the ``a'' position (>1 kcal/mol) were Phe, Val, Asn, Thr, Lys, and Arg. At the ``d'' position, Leu and Gln are favored, along with Asp which is highly destabilizing in the ``a'' position at 25  C. To determine if the observed stability preferences (between positions ``a'' and ``d'') are also in accord with differences in frequency of occurrence differences observed for each amino acid residue in an ``a'' or ``d'' position of naturally occurring coiled-coils, the plots shown in Figure 10 were made. In contrast to the stability data, each of the core positions now showed a more even distribution between the number of residues which statistically preferred either position. However, similarities between stability and frequency of occurrence do exist, i.e. residues which showed high frequency of occurrence for the ``a'' position (e.g. Met, Ile, Val, Asn, Lys, and Arg; Figure 10(b)) were also observed to contribute favorably to coiled-coil stability in this position in the model analogs (Figure 9(b)) (e.g. Ile, Val, Asn, Lys and Arg). Similarly, Leu, which was more frequent in the ``d'' position, was also observed to confer more stability to the coiled-coil when in a ``d'' position. Thus, taken together, the stability positional preferences for particular amino acid residues in either an ``a'' or ``d'' position appear to be in good agreement with those observed by statistical frequency occurrence in native protein coiled-coil domains

Table 4. Estimation of packing effects for amino acid side-chains in the ``a'' and ``d'' positions of a two-stranded a-helical coiled-coil Stability contributiona per residue Amino acid Leu Ile Val Met Phe Tyr Trp Ala Thr Ser Gln Asn His Arg Lys Glu Asp Gly a

``a'' 1.75 1.95 2.05 1.70 1.50 1.09 0.40 0.00 0.10 ÿ0.65 ÿ0.05 0.45 ÿ0.60 ÿ0.40 ÿ0.20 ÿ1.00 ÿ1.25

(kcal/mol)

``d'' 1.90 1.50 0.55 1.60 0.60 0.70 ÿ0.05 0.00 ÿ0.60 ÿ0.90 0.25 ÿ0.30 ÿ0.40 ÿ1.45 ÿ0.90 ÿ1.35 ÿ0.90 ÿ1.80

Positional differenceb ``a''-``d'' (kcal/mol) ÿ0.15 0.45 1.50 0.10 0.90 0.39 0.45 0.00 0.70 0.25 ÿ0.30 0.75 ÿ0.20 1.05 0.70 0.35 0.55

Packing effecte Helix propensityc (kcal/mol) ÿ0.15 ÿ0.37 ÿ0.54 ÿ0.29 ÿ0.48 ÿ0.53 ÿ0.47 0.00 ÿ0.68 ÿ0.63 ÿ0.35 ÿ0.63 ÿ0.63 ÿ0.06 ÿ0.26 ÿ0.64 ÿ0.75 ÿ0.96

Hydrophobicityd (kcal/mol) 1.90 2.04 1.24 1.26 2.02 0.89 2.65 0.00 ÿ0.07 ÿ0.47 ÿ0.72 ÿ1.24 ÿ0.24 1.79 ÿ1.77 ÿ1.29 ÿ1.47 0.42

``a'' 0.00 0.28 1.35 0.73 ÿ0.04 0.74 ÿ1.78 0.00 0.85 0.45 1.02 2.32 0.27 1.45 1.83 0.93 0.13

(kcal/mol)

``d'' 0.15 ÿ0.17 ÿ0.15 0.63 ÿ0.94 0.34 ÿ2.23 0.00 0.15 0.20 1.32 1.57 0.47 0.40 1.13 0.58 1.32 ÿ0.42

Stability contribution per residue is (

GAla)/2, the difference in free energy of unfolding of a single amino acid residue relative to alanine in a two-stranded coiled-coil at an ``a'' or ``d'' position. b Positional difference is

Gu, the difference in free energy of unfolding between positions ``a'' and ``d''. c Helix propensity is

GAla, the difference in free energy for a single mutation in a solvent exposed position of a monomeric a-helix relative to alanine as determined from CD studies (Zhou et al., 1994). d Hydrophobicity is

GAla, the change in solvent transfer free energy relative to alanine (Eisenberg & McLachlan, 1986). e Packing effect is

Gu, the difference in the free energy of unfolding between the stability contribution per residue (

GAla/2) (columns 2 and 3) and the sum of

GAla (helix propensity) and

GAla (hydrophobicity).

394

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Figure 9. (a) Comparson of the relative stability contribution each amino acid makes to the hydrophobic core of a two-stranded coiled-coil in either an ``a'' or ``d'' position. Relative stability values are as those calculated in Table 1. The stability of the Asp ``a'' position analog (denoted with an asterisk) is estimated to be greater than ÿ4 kcal/mol. (b) Difference in stability between the ``a'' and ``d'' positions. Positive values represent greater stability in an ``a'' position relative to that in its ``d'' position. Correspondingly, negative values represent greater stability in a ``d'' position relative to that observed in an ``a'' position. The Asp ``d'' position analog (denoted with an asterisk) is estimated to be 3 kcal/mol or greater in stability compared to the ``a'' position analog.

(Cohen & Parry, 1990; Hu et al., 1990; Lupas et al., 1991).

Discussion Secondary structure, hydrophobicity and packing Here, we have determined the effects of substituting 20 different amino acid residues upon the secondary structure, stability and oligomerization state of a de novo-designed model two-stranded parallel coiled-coil protein. Our results show that all residues, with the exception of proline, can be incorporated into the helical coiled-coil structure. Proline's distinct inability to allow folding appears to be completely independent of the hydrophobic

core position, as our previous studies also showed that proline would not allow folding when substituted in an ``a'' position (Wagschal et al., 1999b). Thus, the destabilizing effect of proline is primarily ascribed to its sterically constrained cyclic sidechain and strong a-helical disrupting nature. It is important to note that although proline cannot be incorporated at either of the hydrophobic core positions, several crystal structures displaying proline residues at exterior positions of a-helices of globular and coiled-coil containing proteins do exist. The observation that all of the other 18 naturally occurring amino acid residues can be incorporated into the coiled-coil appears to be in good agreement with previously acquired statistical occurrence data for such residues existing in long ®lamentous coiled-coil proteins (Cohen & Parry, 1990; Lupas

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

395

Figure 10. Top; comparison of the statistical occurrence of each amino acid residue occurring in an ``a'' or ``d'' position of the heptad repeat in native coiled-coil sequences (Lupus et al., 1991). Bottom; difference in frequency of occurrence between the two positions. Positive values indicate greater occurrence in an ``a'' position relative to the ``d'' position; correspondingly, negative values represent greater occurrence in a ``d'' position relative to an ``a'' position.

et al., 1991; Hu et al., 1990). It is interesting that the occurrence of several residues, in particular the polar and charged residues, is signi®cantly less, if observed at all, in short coiled-coil (4-5 heptads)/ leucine zipper-containing proteins. As observed here, their lower-than-expected occurrence does not re¯ect their inability to be incorporated into such a structure, but rather can be ascribed to their ability to decrease signi®cantly the stability of such a structure. For example, several residues (Glu, Gly and Arg) almost completely eliminated the overall stability of the model coiled-coil molecule, despite the presence of an otherwise optimally packed hydrophobic core (Table 1). In regards to the level of hydrophobicity, we observed the general trend that greater stability of the coiled-coil protein correlated with a net increase in hydrophobicity of the substituted ``d'' interfacial residue. Thus, this result is consistent

with previous studies by Hodges et al. (1990) and Zhou et al. (1992a,b) who demonstrated, using replacements of core Ala residues with different hydrophobic residues, that net increases in stability of the coiled-coil protein were observed with greater hydrophobic character. Moreover, these results are also in accord with the systematic substitution studies in the hydrophobic cores of globular proteins (Pace et al., 1996. Although it is generally accepted that the hydrophobic effect is the major driving force involved in folding and stabilizing the three-dimensional structures of proteins, both hydrophobicity and packing of the residue side-chains in the hydrophobic core of a protein can affect protein stability (Kellis et al., 1988, 1989; Matsumura et al., 1988; Lim & Sauer 1989; Sandberg & Terwilliger, 1989; Dill & Shortel, 1991; Eriksson et al., 1992; Mendel et al., 1992). The exact contribution that packing effects can play in

396

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

protein folding and stability, however, is still the subject of much debate (Behe et al., 1991). Thus, it has become important to determine the relative magnitude of such effects in the hydrophobic interiors of proteins particularly within different folds. Because we have carried out a systematic study of substituting 20 different amino acid residues in the hydrophobic core ``d'' position of a model coiled-coil as well as in an ``a'' position (Wagschal et al., 1999b), we can now calculate the magnitude of the packing effect term for each amino acid residue independent of its hydrophobicity at two sterically different micro-environments, as well as relative to its hydrophobicity term. Table 4 (fourth column) shows the stability difference for each residue between an ``a'' and ``d'' position of a two-stranded coiled coil. As observed, the packing effect term can be as great as 1.5 kcal/ mol between the two geometrically different coiled-coil positions (e.g. Val analog). As the difference in stability between the greatest and least stable analogs was 3.3 and 3.7 kcal/mol per substitution in an ``a'' and ``d'' position, respectively, the ``positional'' packing effect term thus can be almost as great as 1/2 of the net stability. In addition, calculation of the magnitude of the packing effect for each position (based on the premise that the factors contributing to stability are helical propensity, hydrophobicity and packing; Table 4, ®nal two columns) shows that the packing effect term can be as great or greater in net magnitude than the hydrophobicity term itself. For example Trp, which is very hydrophobic (with a hydrophobicity contribution of 2.65 kcal/mol), produces no apparent net gain in stability due to an almost equal negative packing effect energy of 1.78-2.23 kcal/mol. Val, which packs well in an ``a'' position generates a positive packing effect of 1.35 kcal/mol, slightly greater in net magnitude than its hydrophobicity contribution of 1.24 kcal/ mol. In a similar way, the packing effect of Asn (a polar residue) in either an ``a'' or ``d'' positions is very large compared to its hydrophobicity contribution (2.32 kcal/mol and 1.57 kcal/mol versus ÿ1.24 kcal/mol, respectively) which is attributed to the ability of the Asn side-chain to make hydrogen bonds as well as maintain conformational entropy when packed. Thus, packing effects are indeed important and in some cases very important relative to the hydrophobic effect. Comparison of the stability values A comparison of the stability values observed here with those from Vinson's group (Moitra et al., 1997), which recently derived a partial list of seven residues in the ®fth ``d'' position of the bZIP coiled-coil protein, shows excellent agreement for four of the ®ve comparable residues. For example, Met, Ile, Val, and Ser analogs showed

GAla values of 2.0, 1.65, 1.1 and ÿ0.25 kcal/mol, respectively, compared to 1.6, 1.5, 0.55, and ÿ0.9 observed here (Table 4). Vinson's group observed that a

Leu-to-Ala substitution caused a free energy change of 4.6 kcal/mol, signi®cantly greater than the 1.9 kcal/mol change observed here. It is not entirely clear why this should be the case; however, it should be noted that a second determination, by the same group, of a Leu-to-Ala substitution near the termini of the coiled-coil was 2.0 kcal/mol, the latter result being within experimental error of our result. In general, the observation of similar stability values despite completely different contexts, i.e. Vinson's substitution site was located between two valine residues at the a positions, and the e and g0 positions contained Glu and Arg residues compared to that described here (Figure 1), suggests that the observed stability values here may be approaching an intrinsic value for each amino acid residue at a ``d'' position. This is not to say that the stability of such residues cannot be in¯uenced by the adjacent residues, but rather that, here we have designed our model coiled-coil to derive such values in the absence of such effects, with the exception of the Asp analog, where molecular modelling studies have clearly shown the potential for hydrogen bonding between the side-chain carboxyl group and the adjacent e and g0 Gln residues. Previously, the Hodges group have also measured the energetic contribution for several residues substituted in a ``d'' position of a model two-stranded coiled-coil (Hodges et al., 1990; Zhou et al., 1992a,b; Zhu et al., 1993). Relevant to this discussion was the observation that the free energy difference for a Leu-to-Ala substitution in which the disul®de bond was located in position 2a was 1.6 to 1.8 kcal/mol per substitution in an ``a'' position and 0.8 to 1.05 kcal/mol per substitution in a ``d'' position. In the reduced state, both positions demonstrated a free energy difference of 1.5 kcal/ mol per substitution. This indicated that, in the reduced state, the individual strands of coiled-coils can easily rearrange, making the ``a'' and ``d'' sites similar, as opposed to different when they are locked in a conformational state by disul®de bridge formation in the core position. A comparison of a Leu-to-Ala substitution observed here with Hodges' values shows the stability difference to be in good agreement with that observed for the reduced coiled-coil model (1.9 kcal/mol versus 1.5 kcal/mol, respectively). In a similar way, a Leu-to-Ala substitution in the X19a model coiledcoil (Wagschal et al., 1999a,b) also demonstrated a free energy change of 1.75 kcal/mol per substitution. These results suggest that the location of the disul®de bridge interspaced by two glycine residues from the coiled-coil region appears to allow full ¯exibility of each helical strand to rearrange during substitution. Therefore, the values recorded here are likely to re¯ect the rearrangement of the two helices minimizing any cavity formation. It is interesting to note that the energetic difference for a Leu-to-Ala substitution in the hydro-

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

phobic core of the globular protein T4 lysozyme when the cavity size is extrapolated to zero is estimated to be 1.9 kcal/mol per substitution (Erikson et al., 1992), exactly as that observed in this study. In general, however, mutations in the hydrophobic core of globular proteins are observed to be much greater than those observed here. For example, Pace et al. (1992) reported values for Ile, Leu, Val, Met, and Phe of 3.8(0.7), 3.5(0.1.1), 2.5(0.9), 3.0(0.9) and 3.8(0.5) kcal/mol, respectively, which are approximately twice the values reported here (Table 4). This difference is most likely explained by the simplicity and ease of the rearrangement of the polypeptide chains in a coiled-coil to eliminate cavity formation, as compared to the larger and more energetically costly rearrangement of more complex folds in globular proteins. Thus, the present results can be taken as the relative free energy term for each substitution in a globular fold in the absence of the second energy term, speci®ed by Eriksson et al. (1992), which depends on the size of the cavity created by the substitution. Effect of substitutions on the m-values There has been considerable discussion regarding the interpretation of the m value obtained from plots of
G versus denaturant concentration (for review see Bowler et al., 1993; Green et al., 1992; Shortel & Meeker, 1986, 1989; Shortle et al., 1990; Shortel, 1992, 1993; Myers et al., 1995). In brief, it has been suggested that the m value re¯ects the difference in the amount of denaturant bound to the denatured versus the fully folded protein state, and that the differences in the m values within a series of mutants re¯ects differences in the solvent accessible hydrophobic surface area (
ASA) that is exposed upon unfolding (Schellman, 1978). As observed here and other similar host guest studies, a wide variation in the m values was observed for the 20 substitutions (1.54-3.34 kcal molÿ1 Mÿ1; Table 1). Attempts to correlate these values with the calculated
ASA (data not shown), however, demonstrated a poor level of correlation between the two values. This was also the case for plots between the m values and the individual components, i.e. m versus the difference in non-polar surface area and m versus the difference in polar surface area. Additionally, plots of the m values of this study and those from the model ``a'' study also showed little correlation, further suggesting that the m values are not an intrinsic property of residues but rather are dependent upon their local context. Since the 19 analogs appeared to start from the same folded state (based upon their similar ellipticity values) and demonstrated single cooperative denaturation transitions indicative of the absence of intermediates, we would have to conclude that the variation in m values is re¯ective of the level of unfolding in the unfolded state. This conclusion is in good agreement with the recent studies by Smith et al. (1996) who showed that the

397

only physical feature difference between the high and low slope mutants of Protein G was in their unfolded states. It is rather interesting that despite the similarities in the sequences between the ``a'' and ``d'' model peptides, the small contextual difference (i.e. the substitution site between two Leu residues in the ``a'' model and between two Ile residues in the ``d'' model) is suf®cient to result in large differences in the m values. Oligomerization The second aim of the present study was to investigate the ability of each side-chain in a ``d'' position to discriminate between a particular oligomerization state (i.e. the ability to form a twostranded versus a three-stranded coiled-coil structure). As noted above, structural uniqueness arises as a result of a distinct stability difference between two alternative states, and in the case of coiledcoils this has been directly attributed to more favorable packing effects and burial of hydrophobic surface area observed in one or the other associative states. As observed here, Leu only marginally favors the two-stranded state (69 % versus 31 % threestranded), indicating that the packing and burial of hydrophobic surface area in a ``d'' position in a two-stranded versus a three-stranded oligomeric state is not signi®cantly different. This is in good agreement with the high frequency of occurrence of Leu residues found in ``d'' positions of native coiled-coils occurring in either two, three or fourstranded oligomeric states. The indiscriminate ability of Leu in a ``d'' position, however, is in direct contrast to that when it is placed in an ``a'' position, which we have shown induces only trimer formation (Wagschal et al., 1999). Thus, in an ``a'' position, the greater burial of hydrophobic surface area with acute packing versus parallel packing of Leu in the trimeric state creates a clear stability difference, thus driving the exclusive trimeric oligomeric state. As expected, based on the packing constraints placed on b-branched residues when substituted in the ``d'' position, Ile, Val and Thr residues clearly promoted trimer formation (Table 2). Therefore, the ``acute'' packing arrangement in the trimer state is signi®cantly more favorable for these residues than the destabilized two-stranded state (``acute'' packing refers to geometries intermediate to those described as parallel and perpendicular). Thus, these results appear to be in good agreement with native trimer coiled-coil sequences, which show a higher frequency of occurrence of Ile and Val residues in the ``d'' position of three-stranded coiled coils, and the mutant GCN4 construct which contained only Ile residues at its hydrophobic core positions which only formed a trimeric state (Harbury et al., 1993). It is interesting that Asn, which has been noted to play a signi®cant role in de®ning the two stranded oligomerization state of leucine zippers in

398

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

the a position (Gonzales et al., 1996a,b,c), appears to demonstrate almost the opposite effect in the ``d'' position (e.g. 72 %:28 %, three-stranded:twostranded). Thus, as with Thr (above), polarity per se is not the sole determinant in de®ning and/ or promoting a speci®c oligomeric state. Further comparison of the ``a'' and ``d'' oligomerization data also reveals how subtle differences such as a single hydroxyl group on the benzyl side-chain of Tyr can impart a signi®cant effect on the association state. For example, Tyr in the ``a'' position demonstrated a clear preference for the three-stranded state; in contrast, in the ``d'' position, it de®ned only the two-stranded state. Molecular modelling studies (P.L. & R.S.H., unpublished results) have shown that this unique discriminatory ability, compared to Phe, arises from the ability of Tyr to hydrogen bond via its hydroxyl group to the polypeptide backbone in the ``a'' position trimer, thus increasing its stability; whereas, the hydroxyl group sterically clashes in the trimer in the ``d'' position, reducing its stability compared to that of the two-stranded state. Interestingly, Tyr occurs twice as often in both the ``a'' or ``d'' positions compared to Phe, perhaps due to its ability to help de®ne an unique oligomerization state. Of the remaining residues, four (Gly, Ala, Phe, Ser) exhibit the distinct feature of being largely indiscriminate in de®ning a speci®c oligomerization state in both the ``a'' and ``d'' positions. This observation suggests that these residues neither stabilize nor destabilize either oligomeric state to any signi®cant degree and thus can be accommodated equally well in either state. This may be an important feature for these residues, as it allows nature to modulate overall protein stability without imparting any signi®cant effect on the oligomerization choice. Finally, the charged residues can be classi®ed as promoting preferentially only the two-stranded state. This observation, independent of the ``a'' or ``d'' position geometry, likely re¯ects the greater desolvation cost associated with burying a charged group closer to the hydrophobic core in the three-stranded state versus the two-stranded state. Overall, these results further support earlier observations which show that the type and placement of residues within the hydrophobic core ``a'' and ``d'' positions can indeed signi®cantly in¯uence the ®nal associative state of a coiled-coil structure.

Conclusion In summary, we have investigated the effects of substituting 20 different amino acid residues in a ``d'' position of a de novo-designed two/threestranded model coiled-coil. Our results indicate that, although hydrophobicity is a dominant factor in stabilizing these structures, complementary sidechain packing is also a signi®cant factor. The abil-

ity of the model sequence to adopt both a twostranded and three-stranded structure has also allowed us to determine which residues within a ``d'' position can impart a preferential effect on the ®nal oligomerization state. The above results clearly show that the b-branched residues of Ile, Val and Thr promote trimer formation, while the charged residues promote the two-stranded state. Finally, summation of both data sets, i.e, substitution in both the ``a'' and ``d'' positions, allows us to generate a table of relative stability values for each residue in each position, which can thus be used to aid in the future de novo design of new coiled-coil structures, investigate further the structure-function relationships of native coiled-coils, as well as investigate and improve our understanding of protein folding issues.

Materials and Methods Peptide synthesis, cleavage and purification Synthetic model d analogs were prepared manually by stepwise solid-phase synthesis methodology using conventional Na-t-butyloxycarbonyl (t-Boc) chemistry. Brie¯y, 4  1 g of copoly(styrene, 1 % divinylbenzene)4methylbenzhydrylamine-HCl resin, 100-200 mesh (substitution 0.52 mmol amino groups/g) was weighed into four 25 ml polypropylene solid-phase extraction reservoirs equipped with 8 mm polypropylene frits and the resin washed several times with dichloromethane (DCM), dimethylformamide (DMF) and DCM (5 ml/ wash). For amino acid residue couplings, 2 mmol of the desired amino acid was reacted with 4 ml (1.8 mmol) of a solution of 0.45 M O-benzotriazol-1-yl-N,N,N0 ,N0 -tetramethyluronium hexa¯uorophosphate (HBTU) in DMF and 500 ml (6 mmol) of diisopropylethylamine (DIEA) for ®ve minutes. The activated residues were added into each reaction reservoir and the vessel agitated (shaken) for 30 minutes. Following coupling, excess amino acid was removed by three washes with DMF and DCM (5 ml/wash). The ef®ciency of the coupling was determined by ninhydrin monitoring, and coupling of the amino acid was repeated if the desired level of completeness had not been achieved. The t-Boc a-amino protecting group was removed by agitation with a solution of tri¯uoroacetic acid (TFA):DCM (90:10, v/v) for ®ve minutes, followed by neutralization by washing three times with a solution of DIEA:DCM (10:90, v/v), and three washes with DMF, and DCM. Coupling of the four resins was carried out until residue 21, at which time the ``core resins'' were sub-divided into ®ve and the substituted amino acid coupled. Each resin was then extended further in a similar manner to that described above with the exception that the activated amino acid was reduced to 1 mmol. After ®nal coupling of the last amino acid, the N terminus was deprotected and N-terminally acetylated using acetic anhydride (2.5 equivalents) and DIEA (2.5 equivalents) in DCM for 15 minutes. Peptide resins were washed with DCM, methanol, and acetone and air dried. Peptides were cleaved from the resin by reaction with hydrogen ¯uoride (20 ml/g resin) containing 10 % (v/v) anisole and 2 % (w/w) 1,2-ethanedithiol for 1.5 hours at 4  C. Following cleavage, crude peptides were washed with cold ether several times, extracted from the resin with glacial acetic acid and lyophilized. Puri®cation of each peptide was performed by reversed-phase HPLC

399

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil (RP-HPLC) on a SynChropak semi-preparative C8 colÊ umn (250 mm  10 mm I.D., 6.5 mm particle size, 300 A pore size; SynChrom, Lafayette, IN, USA) with a linear AB gradient (ranging from 0.2 % to 1.0 % B/minute) at a ¯ow rate of 2 ml/minute, where solvent A is aqueous 0.05 % (w/w) tri¯uoroacetic acid (TFA) and solvent B is 0.05 % (w/w) TFA in acetonitrile. The homogeneity of the puri®ed peptides was veri®ed by analytical RPHPLC, amino acid analysis and electrospray quadrapole mass spectrometry.

Chemical denaturation measurements Denaturation midpoints, slopes and differences in free energy of unfolding (

Gu) values for the model ``d'' analogs (see Table 1) were determined by following the change in molar ellipticity at 222 nm using a Jasco J-720 spectropolarimeter (as described above). Ellipticity readings were normalized to the fraction of peptide folded ( ff) or unfolded ( fu) using the standard equations (2) and (3): ff ˆ …‰yŠ ÿ ‰yŠu †=…‰yŠn ÿ ‰yŠu †

…2†

fu ˆ …1 ÿ ff †

…3†

Preparation of oxidized peptides Formation of homo-two-stranded disul®de-bridged molecules was carried out by dissolving 5 mg of each peptide in 2 ml of 100 mM NH4HCO3(pH 8) buffer and magnetically stirring the solution in an open vial at 20  C overnight. Completeness of the oxidation was veri®ed by RP-HPLC and mass spectrometry. To purify the oxidized products, each peptide solution was neutralized with dilute acetic acid and then injected onto a semi-preparative reversed-phase C8 column (Zorbax 300SB-C8, Ê pore size) 9.4 mm I.D.  25 cm, 5 mm particle size, 300 A and subsequently eluted at 2 ml/min with a linear AB gradient of 1 % B/minute, where solvent A and B are described above. Circular dichroism spectroscopy Circular dichroism (CD) spectroscopy was performed at 20  C on a Jasco J-720 spectropolarimeter (Jasco Inc., Easton, MD) interfaced to an Epson Equity 386/25 computer running the Jasco DP-500/PS2 system software (version 1.33a). A Lauda model RMS circulating water bath (Lauda, Westbury, NY) was used to maintain temperature control of the optic cell. The instrument was calibrated with an aqueous solution of re-crystallized d-10-(‡)-camphorsulfonic acid at 290.5 nm. Results are expressed as mean residue molar ellipticity [] (deg
cm2
dmolÿ1) calculated from equation (1): ‰Š ˆ …obs  MRM†=…10  l  c†

…1†

where obs is the observed ellipticity expressed in millidegrees, MRM is the mean residue molecular mass (molecular mass of the peptide divided by the number of amino acid residues), l is the optical path length in cm, and c is the ®nal peptide concentration in mg/ml. For wave scans, stock solutions of disul®de-bridged peptides (5 mg/ml) were prepared in benign buffer (50 mM phosphate containing 0.1 M KCl (pH 7), diluted tenfold with either buffer (10 ml stock peptide solution with 90 ml buffer) or buffer containing tri¯uoroethanol (TFE) (10 ml stock peptide solution with 40 ml buffer and 50 ml TFE), scanned from 190-255 nm at 0.05 nm intervals and the average of ten scans reported. For GdnHCl denaturation studies, stock peptide samples (as above) were mixed with various ratios of benign buffer and a solution of 8 M GdnHCl in benign buffer to give a series of solutions containing the same peptide concentration but different GdnHCl concentrations. The ellipticity of each of the peptide solutions were read at 222 nm and the average of ten scans recorded. All peptide concentrations were determined by amino acid analysis on a Beckman model 6300 amino acid analyzer, and the standard deviation of measurements at 222 nm was ‡/ÿ 300 deg. cm2 molÿ1.

where [y]n and [y]u represent the ellipticity values for the fully folded and fully unfolded species, respectively. [y] is the observed ellipticity at 222 nm at any denaturant concentration. The free energy of unfolding (
Gu) at each denaturant concentration was calculated using the equation
Gu ˆ ÿ RTln(Ku), where Ku is the equilibrium constant of the unfolding process (Pace, 1986; Shortle, 1989). In the case of disul®de-bridged peptides, this is simply given by Ku ˆ fu/ff ˆ (1 ÿ ff)/( ff). The free energy 2O of unfolding in the absence of denaturant (
GH ) can u be obtained by linear extrapolation according to the equation
Guˆ
GuH2 O ÿm [denaturant] (Pace, 1986; Shortle, 1989), where m is the slope. However, because small errors in the slope term may lead to large errors in 2O the extrapolated
GH value (Green et al., 1992; Pace, u 1986), the difference in free energy of unfolding between analogs (the

Gu values) was calculated using the equation given by Serrano & Fersht (1989):

Gu ˆ…‰denaturantŠ1=2;A ÿ ‰denaturantŠ1=2;B †…mA ‡ mB †=2

…4†

where [denaturant]1/2,A and [denaturant]1/2,B are the midpoints of unfolding of analog A and B, respectively, derived from the plots of fraction folded versus GdnHCl concentration, and mA is the slope determined from the equation
Gu ˆ
GuH2 O ÿ m [denaturant] (as described above). This approach calculates the value of

Gu at the denaturant concentration half-way between the [denaturant]1/2 values of the peptides being compared. RP-HPLC analysis The relative hydrophobicity of each analog was performed by RP-HPLC on a Zorbax Eclipse XDB-C8 colÊ umn (150 mm  4.6 mm I.D., 5 mm particle size, 300 A pore size); Hewlett-Packard, Little Falls Site, DE, USA) using a Hewlett-Packard 1090 HPLC work station. Brie¯y, 5 mg of each carboxamidomethylated-modi®ed peptide (see below) was run individually, as well as together in appropriate groups to determine the retention time of each analog relative to that of the Ala-substituted analog. Each run consisted of a linear AB gradient of 1.0 % B/minute at a ¯ow-rate of 1 ml/minute and a temperature of 70  C, where solvent A is aqueous 50 mM KH2PO4, 100 mM NaClO4 (pH 7) and solvent B is 50 mM KH2PO4, 100 mM NaClO4 (pH 7) in 50 % acetonitrile. In order to derive retention times that were independent of the stability of the coiled-coils, each oxidized peptide was ®rst reduced with two equivalents of dithiothreitol (DTT) followed by reaction of the sulfhydryl group of the cysteine residue side-chain with four equivalents of iodoacetamide. Excess DTT was then

400

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

added to neutralize any remaining iodoacetamide. The carboxamidomethylation prevented formation of the disul®de bond, thus maintaining the peptides as single stranded during chromatography. Size-exclusion chromatography Determination of the oligomeric state(s) of each analog in benign buffer conditions (50 mM potassium phosphate, 100 mM KCl (pH 7)) and in the presence of various concentrations of GdnHCl denaturant in the buffer was carried out using a SuperdexTM75 HR 10/30 (10 mm I.D.  300 mm) high-performance size-exclusion column (Pharmacia, Uppsala, Sweden) and a Hewlett-Packard 1090 HPLC work station. The column was conditioned in several column volumes at a ¯ow-rate of 0.2 ml/min prior to injection. Each analog was injected either from a 1 mM stock or tenfold diluted 0.1 mM solution and the chromatograms collected at 210 nm. To correct for bed volume changes due to changes in the ionic strength of the various buffers used, two random coil internal peptides standards were co-injected with each peptide analog and the chromatogram corrected to the elution times of the standards: S1 ˆ Ac-(GLGLG)6-amide (30 residues, MM 2534 Da); S2 ˆ Ac-KYGLGGAGGLK-amide (11 residues, MW 1062 Da). Molecular mass were estimated based on their correlation in retention time to several in-house coiled-coil standards of various sizes (Mant et al., 1997). Ultracentrifugation studies Sedimentation equilibrium experiments were performed at 20  C on a Beckman Model Optima XLI analytical ultracentrifuge equipped with both Rayleigh interference optics and absorbance optics. Each peptide was dissolved in apo buffer (50 mM potassium phosphate, 100 mM KCl (pH 7)) at a concentration of 1.5 mg/ml and dialyzed against the same buffer at 4  C overnight. The initial concentration of each sample was determined by fringe counts using an average refractive increment of 4.1 fringes/mg/ml as described (Babul et al., 1969). A 100 ml aliquot of each sample was loaded into a 12 mm double-sector, charcoal-®lled Epon cell and run for 48 hours with a rotor speed of 18,00028,000 rpm (24,000 g-58,000 g). Total peptide concentration was approximately 500 mM. The partial speci®c volume of each peptide was calculated using the program SEQSEE (Wishart et al., 1994; Cohn & Edsall, 1943) 0.7506 ml/g. The density of the solvent was calculated to be 1.009 g/ml. In the case of the model E analytical ultracentrifuge equilibrium, photographs were taken at the end of each run and fringe counts measured using a Nikon Model 6C microcomparator. For experiments using the model XLI analytical ultracentrifuge, data were collected electronically and processed directly. Apparent molecular masses were calculated according to the method described by Chervenka (1969) using the program NONLIN (non-linear least squares analysis) to determine the best associative model. For sedimentation velocity experiments, i.e. determination of the translational frictional ratio ( f/fmin) of the monomeric and trimeric oligomeric states, peptides were dissolved in apo buffer (2.5 mg/ml) and introduced into the cell. The rotor was then accelerated to 48,000 rpm and photographs of the schlieren peak taken regular intervals as the protein sedimented to the bottom of the cell. Calculation of sobs,

s20,w, f/fmin and axial ratio were carried out as described (Wagschal et al., 1999a). Molecular structure analysis Molecular modelling of Met, Trp and Asp was carried out on a Silicon Graphics Crimson Elan work station using the INSIGHT II, BIOPOLYMER and DISCOVER programs (Biosym Technologies Inc. San. Diego, CA, USA). GCN4, PDB code 2zta.pdb (O'Shea et al., 1991) was used initially as the coordinate ®le for the backbone of each helix, and the side-chains replaced to conform with the amino acid sequence of the model sequence. Met, Trp and Asp side-chain torsion angles were rotated manually to achieve the ®nal modelled state.
ASA analysis The difference in surface accessible area (
ASA) between the folded and the unfolded state was calculated using the algorithm DSC (P. Lavigne, unpublished results). The fully folded coiled-coil structure was modelled using the initial coordinates for the parallel twostranded helical coiled-coil obtained from the crystal structure of the GCN4 protein, PDB code 2zta.pdb (O'Shea et al., 1991). Residues in GCN4 were replaced with the amino acid sequence of the model ``d'' sequence (Figure 1) while maintaining similar side-chain conformational angles. The substitution site residues were manually checked to verify acceptable dihedral angles (based on the library of preferred rotamers (Ponders & Richards, 1987)) and no direct steric clashes. The structures were then further re®ned using a quick descent four-step energy minimization procedure. The unfolded coiled-coil state was modelled as a linear polypeptide strand.

Acknowledgments We thank Kim Oikawa for CD spectroscopy and Leslie Hicks for ultracentrifugation analysis. This work was funded by the Government of Canada through the Medical Research Council Group in Protein Structure and Function and a studentship (B.T.) from the Alberta Heritage Foundation for Medical Research, Alberta Canada.

References Babul, J. & Stellwagen, E. (1969). Measurements of protein concentration with interference optics. Annu. Biochem. 28, 216-221. Bowler, B. E., May, K., Zaragoza, T., York, P., Dong, A. & Caughey, W. S. (1993). Destabilizing effects of replacing a surface lysine of cytochrome c with aromatic amino acids: implications for the denatured state. Biochemistry, 32, 183-190. Behe, M. J., Lattman, E. E. & Rose, G. D. (1991). The protein-folding problem: the native fold determines packing, but does packing determine the native fold? Proc. Natl Acad. Sci. USA, 88, 4195-4199. Chen, Y.-H., Yang, J. T. & Chau, K. H. (1974). Determination of the helix and b form of proteins in aqueous solution by circular dichroism. Biochemistry, 13, 3350-3359.

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil Chervenka, C. H. (1969). A Manual of Methods for the Analytical Ultracentrifuge, pp. 1-100, Spinco Division of Beckman Instruments, Palo Alto, CA. Cohen, C. & Parry, D. A. D. (1990). a-Helical coiled coils and bundles: how to design an a-helical protein. Proteins: Struct. Funct. Genet. 7, 1-15. Cohn, E. J. & Edsall, J. T. (1943). In Proteins, Amino Acids, and Peptides as Ions and Dipolar Ions, pp. 370381, Reinhold Publishing Corp., New York. Cooper, T. M. & Woody, R. W. (1990). The effect of conformation on the CD of interacting helices: a theoretical study of tropomyosin. Biopolymers, 30, 657-676. Crick, F. H. C. (1953). The packing of a-helices: simple coiled-coils. Acta Crystallog. 6, 689-698. Dill, K. A. & Shortle, D. (1991). Denatured states of proteins. Annu. Rev. Biochem. 60, 795-825. Eisenberg, D. & McLachlan, A. D. (1986). Solvation energy in protein folding and binding. Nature, 319, 199-203. Engel, M., Williams, R. W. & Erickson, B. W. (1991). Designed coiled-coil proteins: synthesis and spectroscopy of two 78 residue alpha-helical dimers. Biochemistry, 30, 3161-3169. Eriksson, A. E., Baase, W. A., Zhang, X.-J., Heinz, D. W., Blaber, M., Baldwin, E. P. & Matthews, B. W. (1992). Response of a protein structure to cavitycreating mutations and its relation to the hydrophobic effect. Science, 255, 178-183. Fauchere, J. L. & Pliska, V. (1983). Hydrophobic parameters p of amino acid side-chains from the partioning of N-acetyl-amino-acid-amides. Eur. J. Med. Chem. 18, 369-375. Gonzalez, L., Jr, Brown, R. A., Richardson, D. & Alber, T. (1996a). Crystal structures of a single coiled-coil peptide in two oligomeric states reveal the basis for structural polymorphism. Nature Struct. Biol. 3, 1002-1009. Gonzalez, L., Jr, Plecs, J. J. & Alber, T. (1996b). An engineered allosteric switch in leucine-zipper oligomerization. Nature Struct. Biol. 3, 510-515. Gonzalez, L., Jr, Woolfson, D. N. & Alber, T. (1996c). Buried polar residues and structural speci®city in the GCN4 leucine zipper. Nature Struct. Biol. 3, 1011-1018. Green, S. M., Meeker, A. K. & Shortle, D. (1992). Contributions of the polar, uncharged amino acids to the stability of staphylococcal nuclease: evidence for mutational effects on the free energy of the denatured state. Biochemistry, 31, 5717-5728. Harbury, P. B., Zhang, T., Kim, P. S. & Alber, T. (1993). A switch between two, three, and four-stranded coiled coils in GCN4 leucine zipper mutants. Science, 262, 1401-1407. Hodges, R. S., Sodek, J., Smillie, L. B. & Jurasek, L. (1972). Tropomyosin: amino acid sequence and coiled-coil structure. Cold Spring Harbor Symp. Quant. Biol. 37, 299-310. Hodges, R. S., Saund, A. K., Chong, P. C., St.-Pierre, S. A. & Reid, R. E. (1981). Synthetic model for twostranded a-helical coiled-coils. Design, synthesis, and characterization of an 86-residue analog of tropomyosin. J. Biol. Chem. 256, 1214-1224. Hodges, R. S., Zhou, N. E., Kay, C. M. & Semchuk, P. D. (1990). Synthetic model proteins: contribution of hydrophobic residues and disul®de bonds to protein stability. Peptide Res. 3, 123-137. Hu, J. C., O'Shea, E. K., Kim, P. S. & Sauer, R. T. (1990). Sequence requirements for coiled-coils: analysis

401

with lambda repressor- GCN4 leucine zipper fusions. Science, 250, 1400-1403. Kellis, J. T., Jr, Nyberg, K. & Fersht, A. R. (1989). Energetics of complementary side-chain packing in a protein hydrophobic core. Biochemistry, 28, 49144922. Kellis, J. T., Jr, Nyberg, K., Sali, D. & Fersht, A. R. (1988). Contribution of hydrophobic interactions to protein stability. Nature, 333, 784-786. Kohn, W. D., Kay, C. M. & Hodges, R. S. (1997). Salt effects on protein stability: two-stranded a-helical coiled-coils containing inter- or intrahelical ion pairs. J. Mol. Biol. 267, 1039-1052. Kohn, W. D., Kay, C. M. & Hodges, R. S. (1998). Orientation, positional, additivity, and oligomerizationstate effects of interhelical ion pairs in a-helical coiled-coils. J. Mol. Biol. 283, 993-1012. Lau, S. Y. M., Taneja, A. K. & Hodges, R. S. (1984). Synthesis of a model protein of de®ned secondary and quaternary structure; effect of chain length on the stabilization and formation of two-stranded a-helical coiled-coils. J. Biol. Chem. 259, 13253-13261. Lim, W. A. & Sauer, T. (1989). Alternative packing arrangements in the hydrophobic core of l repressor. Nature, 339, 31-36. Lumb, K. J. & Kim, P. S. (1995). A buried polar interaction imparts structural uniqueness in a designed heterodimeric coiled coil. Biochemistry, 34, 86428648. Lupas, A. (1997). Predicting coiled-coil regions in proteins. Curr. Opin. Struct. Biol. 7, 388-393. Lupas, A., Van Dyke, M. & Stock, J. (1991). Predicting coiled coils from protein sequences. Science, 252, 1162-1164. Mant, C. T., Chao, H. & Hodges, R. S. (1997). Effect of mobile phase on the oligomerization state of alphahelical coiled-coil peptides during high-performance size-exclusion chromatography. J. Chromatog. sect. A, 791, 85-98. Matsumura, M., Becktel, W. J. & Matthews, B. W. (1988). Hydrophobic stabilization in T4 lysozyme determined directly by multiple substitutions of Ile3. Nature, 334, 406-410. McLachlan, A. D. & Stewart, M. (1975). Tropomyosin coiled-coil interactions: evidence for an unstaggered structure. J. Mol. Biol. 98, 293-304. Mendel, D., Ellman, J. A., Chang, Z., Veenstra, D. L., Kollman, P. A. & Schultz, P. G. (1992). Probing protein stability with unnatural amino acids. Science, 256, 1798-1802. Moitra, J., Szilak, L., Krylov, D. & Vinson, C. (1997). Leucine is the most stabilizing aliphatic amino acid in the d position of a dimeric leucine zipper coiled coil. Biochemistry, 36, 12567-12573. Monera, O. D., SoÈnnichsen, F. D., Hicks, L., Kay, C. M. & Hodges, R. S. (1996). The relative positions of alanine residues in the hydrophobic core control the formation of two-stranded or four-stranded alphahelical coiled-coils. Protein Eng. 9, 353-363. Monera, O. D., Zhou, N. E., Kay, C. M. & Hodges, R. S. (1993). Comparison of anti-parallel and parallel two-stranded alpha-helical coiled-coils. Design, synthesis, and characterization. J. Biol. Chem. 268, 19218-19227. Myers, J. K., Pace, C. N. & Sholtz, J. M. (1995). Denaturant m values and heat capacity changes: relation to changes in accessible surface areas of protein unfolding. Protein Sci. 4, 2138-2148.

402

20 Amino Acid Substitutions in Position ``d'' of a Model Coiled-coil

Nelson, M. R. & Chazin, W. J. (1998). An interactionbased analysis of calcium-induced conformational changes in Ca2‡ sensor proteins. Protein Sci. 7, 270282. O'Shea, E. K., Klemm, J. D., Kim, P. S. & Alber, T. (1991). X-ray structure of the GCN4 leucine zipper, a two-stranded, parallel coiled coil. Science, 254, 539-544. O'Shea, E. K., Lumb, K. J. & Kim, P. S. (1993). Peptide ``Velcro'': Design of a heterodimeric coiled coil. Curr. Biol. 3, 658-667. Pace, C. N. (1986). Determination and analysis of urea and guanidine hydrochloride denaturation curves. Methods Enzymol. 131, 266-280. Pace, C. N., Laurents, D. V. & Erickson, R. E. (1992). Urea denaturation of bamase: pH dependence and characterization of the unfolded state. Biochemistry, 31, 2728-2734. Pace, C. N., Shirley, B. A., McNutt, M. & Gajiwala, K. (1996). Forces contributing to the conformational stability of proteins. FASEB J. 10, 75-83. Ponder, J. W. & Richards, F. M. (1987). Tertiary templates for proteins. Use of packing criteria in the enumeration of allowed sequences for different structural classes. J. Mol. Biol. 193, 775-791. Sandberg, W. S. & Terwilliger, T. C. (1989). In¯uence of interior packing and hydrophobicity on the stability of a protein. Science, 245, 54-57. Schellman, J. A., Lindorfer, M., Hawkes, R. & Grutter, M. (1981). Mutations and protein stability. Biopolymers, 20, 1989-1999. Serrano, L. & Fersht, A. R. (1989). Capping and alphahelix stability. Nature, 342, 296-299. Shoemaker, K. R., Kim, P. S., York, E. J., Stewart, J. M. & Baldwin, R. L. (1987). Tests of the helix dipole model for stabilization of a-helices. Nature, 326, 563567. Shortle, D. (1992). Mutational studies of protein structures and their stabilities. Quart. Rev. Biophys. 25, 205-250. Shortle, D. (1993). Denatured states of proteins and their roles in folding and stability. Curr. Opin. Struct. Biol. 3, 66-74. Shortle, D. (1989). Probing the determinants of protein folding and stability with amino acid substitutions. J. Biol. Chem. 264, 5315-5318. Shortle, D. & Meeker, A. K. (1986). Mutant forms of staphylococcal nuclease with altered patterns of guanidine hydrochloride and urea denaturation. Protein: Struct. Funct. Genet. 1, 81-89. Shortle, D. & Meeker, A. K. (1989). Residual structure in large fragments of staphylococcal nuclease: effects of amino acid substitutions. Biochemistry, 28, 936944. Shortle, D., Stites, W. E. & Meeker, A. K. (1990). Contributions of the large hydrophobic amino acids to the stability of staphylococcal nuclease. Biochemistry, 29, 8033-8041.

Smith, C. K., Bu, Z., Anderson, K. S., Sturtevant, J. M., Engelman, D. M. & Regan, L. (1996). Surface point mutations that signi®cantly alter the structure and stability of a protein's denatured state. Protein Sci. 5, 2009-2019. SoÈnnichsen, F. D., Van Eyk, J. E., Hodges, R. S. & Sykes, B. D. (1992). Effect of tri¯uoroethanol on protein secondary structure: an NMR and CD study using a synthetic actin peptide. Biochemisty, 31, 8790-8798. Wagschal, K., Tripet, B. & Hodges, R. S. (1999a). De novo design of a model peptide sequence to examine the effects of single amino acid substitutions in the hydrophobic core on both stability and oligomerization state of coiled-coils. J. Mol. Biol. 285, 785803. Wagschal, K., Tripet, B., Lavigne, P., Mant, C. & Hodges, R. S. (1999b). The role of position a in determining the stability and oligomerization state of alpha-helical coiled coils: 20 amino acid stability coef®cients in the hydrophobic core of proteins. Protein Sci. 8, 2312-2329. Wishart, D. S., Boyco, R. F., Willard, L., Richards, F. M. & Sykes, B. D. (1994). SEQSEE: a comprehensive program suite for protein sequence analysis. Comp. Appl. Biosci. 10, 121-132. Yu, Y., Monera, O. D., Hodges, R. S. & Privalov, P. L. (1996a). Investigation of electrostatic interactions in two-stranded coiled-coils through residue shuf¯ing. Biophys. Chem. 59, 299-314. Yu, Y., Monera, O. D., Hodges, R. S. & Privalov, P. L. (1996b). Ion pairs signi®cantly stabilize coiled-coils in the absence of electrolyte. J. Mol. Biol. 255, 367372. Zhou, N. E., Kay, C. M. & Hodges, R. S. (1992a). Synthetic model proteins: the relative contribution of leucine residues at the nonequivalent positions of the 3-4 hydrophobic repeat to the stability of the two-stranded a-helical coiled-coil. Biochemistry, 31, 5739-5746. Zhou, N. E., Kay, C. M. & Hodges, R. S. (1992b). Synthetic model proteins. Positional effects of interchain hydrophobic interactions on stability of twostranded a-helical coiled-coils. J. Biol. Chem. 267, 2664-2670. Zhou, N. E., Monera, O. D., Kay, C. M. & Hodges, R. S. (1994). a-Helical propensities of amino acids in the hydrophobic face of an amphipathic a-helix. Protein Peptide Letters, 1, 114-119. Zhu, B.-Y., Zhou, N. E., Semchuk, P. D., Kay, C. M. & Hodges, R. S. (1992). Design, synthesis and structural characterisation of model heterodimeric coiled-coil proteins. Int. J. Peptide Protein Res. 40, 171-179. Zhu, B.-Y., Zhou, N. E., Kay, C. M. & Hodges, R. S. (1993). Packing and hydrophobicity effects on protein folding and stability: effects of beta-branched amino acids, valine and isoleucine, on the formation and stability of two-stranded a-helical coiled coils/ leucine zippers. Protein Sci. 2, 383-394.

Edited by P. E. Wright (Received 4 April 2000; received in revised form 11 May 2000; accepted 11 May 2000)