Cation–π interactions studied in a model coiled-coil peptide

Abstract

Cation–π interactions between aromatic amino acids and the positively charged residues lysine and arginine have been proposed to play an important role in stabilizing protein structure. We have used a peptide that adopts a coiled coil structure as a model system to evaluate the energetic contribution of cation–π interactions to protein folding. Peptides were designed in which phenylalanine, tyrosine, and tryptophan were placed at a solvent-exposed position of the helix, one turn removed from an arginine residue that could provide a favorable cation–π interaction. Only the arginine–phenylalanine pairing provided significant stabilization of the peptide structure and it appears that hydrophobic packing, rather than the cation–π effect, is more likely to be responsible for the stability of this peptide. We conclude that any stabilizing effect of cation–π interactions in these peptides is much smaller than that predicted from computational studies.

The cation–π effect arises from favorable electrostatic interactions between the electron-rich π system of an aromatic molecule and a positively charged species such as a metal ion or quaternary amine (Ma and Dougherty 1997; Waters 2002). Interactions between aromatic amino acid residues and positively charged side chains, attributed to the cation–π effect, are commonly observed in proteins (Burley and Petsko 1985; Dougherty 1996). They have attracted interest as they represent a novel type of noncovalent interaction between hydrophobic and hydrophilic side chains that has the potential to confer both stability and specificity in protein folding. Similar interactions also appear to be important in the binding of positively charged substrates in enzymes such as acetylcholine esterase and trimethylamine dehydrogenase, where the active sites contain numerous aromatic residues but no negatively charged residues that could formally neutralize the positive charge of the substrate (Wilson et al. 1995; Scrutton and Raine 1996; Beene et al. 2002; Zacharias and Dougherty 2002).

The strength of the cation–π interaction has been investigated in theoretical studies, and experimentally in gas-phase experiments, in small molecule host–guest model systems in solution, and in proteins (Sunner et al. 1981; Deakyne and Meotner 1985; Dougherty 1996; Ma and Dougherty 1997; Gallivan and Dougherty 2000). A concise and critical summary of much of this previous work is provided by Kallenbach (Shi et al. 2002b). In particular, theoretical studies by Gallivan and Dougherty concluded that cation–π interactions in proteins may attain a strength of up to 4 kcal/mole and are potentially more stabilizing than salt bridges (Gallivan and Dougherty 1999, 2000). However, experimentally the magnitude of cation–π effects in proteins remains poorly defined, because of the difficulty of dissecting out this energetic term from other noncovalent interactions, in particular hydrophobic effects, which contribute to protein folding. In two examples where potential cation–π interactions have been studied, an interaction between a histidine–tyrosine pair in flavodoxin was estimated to contribute only 0.5 kcal/mole in stability (Fernandez-Recio et al. 1999) whereas a cation–π interaction between tryptophan and serotonin in the 5-HT_3AR receptor was estimated to contribute 4 kcal/mole towards ligand binding (Beene et al. 2002).

Peptides provide attractive model systems to investigate the energetics of amino acid side-chain interactions because they can be studied in a well-defined environment where the complicating effects of tertiary interactions can be minimized. Recently, several studies have investigated interactions between positively charged and aromatic side chains spaced one turn apart (in an i, i + 4 relationship) in short, monomeric, alanine-based α-helical peptides (Fernandez-Recio et al. 1997; Olson et al. 2001; Andrew et al. 2002; Shi et al. 2002a; Tsou et al. 2002). The free energies measured range from −1 to 0 kcal/mole, depending on the aromatic–cation pair being considered. Interestingly, the strength of these interactions also appeared to depend on the relative positions of the aromatic and cationic partners: Thus a Trp–Arg_{(i, i + 4)} interaction was observed to be stabilizing by 0.4 kcal/mole whereas the reverse Arg–Trp_{(i, i + 4)} arrangement provided no stability (Shi et al. 2002b). Some studies have concluded that stabilizing interactions are primarily hydrophobic in nature (Andrew et al. 2001, 2002), whereas in other cases it appears that electrostatic interactions provide some contribution to stability, as predicted for a true cation–π interaction (Tsou et al. 2002). Most recently, the cation–π interaction has also been investigated in the context of a β-hairpin peptide. In this case the interaction was concluded to contribute between −0.2 and −0.48 kcal/mole of stability, depending on the residues involved in the interaction (Tatko and Waters 2003). One finding shared by all these studies is that the experimentally measured free energies for this interaction are between 5- and 10-fold smaller than those predicted by computational studies on either small molecule systems or crystallographically defined interactions in proteins.

Results

Choice of model system

Isolated, monomeric α-helices are rarely found in nature; instead helices are generally associated with tertiary structure and are thus in an anisotropic environment with nonuniform solvent exposure. Therefore, to better approximate the situation in natural proteins, we have chosen to investigate aromatic–cation side-chain interactions in the context of a de novo designed peptide that adopts a parallel coiled-coil or helical bundle structure. The coiled coil is a motif found in many natural proteins, and one that is very well understood through numerous studies on de novo designed helical bundles (Cohen and Parry 1994; Betz et al. 1995; Munson et al. 1996; DeGrado et al. 1999; Oakley and Hollenbeck 2001). Because a survey of the protein structure database has implicated 70% of arginine residues in possible cation–π interactions (Gallivan and Dougherty 1999), our studies have focused on the interactions of this residue with aromatic side chains.

The peptides described here were based on the sequence of a helical peptide originally designed by Pecoraro and coworkers as a model system for studying heavy metal binding by proteins (Dieckmann et al. 1998; Farrer et al. 2000; Farrer and Pecoraro 2002; Matzapetakis et al. 2002). Here we call this peptide KE, referring to the residues at positions 10 and 14, respectively, that are varied in our experiments (in the original work this peptide was referred to as L16C). KE incorporates leucine residues at the internal “a” and “d” positions that form the hydrophobic core of the folded protein, with the exception of Leu 16, which is substituted by cysteine to provide a metal-binding ligand. The reduced form of KE has been extensively characterized and at neutral pH adopts a trimeric, parallel coiled-coil structure, here designated KE₃, whereas at low pH it is a dimer (Dieckmann et al. 1998). The inclusion of the cysteine residue at an “a” position means that the peptide may be oxidized to form a covalently cross-linked dimer, referred to as KE_S-SKE, and this forces the peptide to adopt a dimeric coiled coil even at neutral pH. This ability to control the oligomerization state of the peptide (trimer versus dimer) has allowed us to investigate cation–π interactions in the context of two different tertiary structures. We included KE as a control in all the experiments we describe here.

To adapt this peptide scaffold to investigate potential cation–π interactions, we initially remodeled an intrahelix salt bridge between Lys 10 and Glu 14 by incorporating arginine at position 10 to produce a control peptide, designated RE, against which we could compare the effects of introducing aromatic residues at position 14. Peptides were synthesized that contained arginine at position 10 and either tryptophan, tyrosine, or phenylalanine at position 14; these peptides are designated RW, RY, and RF, respectively. Finally, as a control, a peptide containing glutamate and tyrosine at positions 10 and 14, respectively (designated EY), was designed. Glutamate was chosen because it has a similar hydrophilicity and helical propensity to arginine, but obviously will not form a cation–π interaction with tyrosine. The sequences of all these peptides are shown in Table 1.

Table 1.

Sequences of the peptides used in this study

	Sequence
Peptide		abcdefg	abcedfg	abcdefg	abcdefg
KE	Ac-G	LKALEEK	LKALEEK	CKALEEK	LKALEEK	G-CONH2
RE	Ac-G	LKALEEK	LRALEEK	CKALEEK	LKALEEK	G-CONH2
RF	Ac-G	LKALEEK	LRALEFK	CKALEEK	LKALEEK	G-CONH2
RW	Ac-G	LKALEEK	LRALEWK	CKALEEK	LKALEEK	G-CONH2
RY	Ac-G	LKALEEK	LRALEYK	CKALEEK	LKALEEK	G-CONH2
EY	Ac-G	LKALEEK	LEALEYK	CKALEEK	LKALEEK	G-CONH2

Residues that vary from the parent KE peptide (residues 10 and 14) are in bold.

Residues 10 and 14 occupy the “b” and “f” positions in the helical heptad repeat. These are the most solvent-exposed positions and were therefore chosen to minimize the possibility that either residue, the aromatic one in particular, might interact with hydrophobic side chains on the partner helices. Molecular modeling of these residues into the backbone of a generic coiled-coil peptide, using the program Insight II, demonstrated that the flexible arginine side chain could adopt a conformation in which the guanidino group stacked in a face-to-face manner with its aromatic partner, as shown in Figure 1 ▶. Indeed, the conformation modeled for the RY cation–π pair is very similar to that which has been determined crystallographically for a naturally occurring Arg–Tyr interaction in the protein VP39 (Hodel et al. 1996). Examples of similar, crystallographically characterized stacking interactions involving Arg–Trp and Arg–Phe pairs include the tyrosine kinase domain of the human insulin receptor (Arg–Trp; Hubbard et al. 1994) and the extracellular domain of the receptor for the growth hormone, erythropoietin (Arg–Phe; Livnah et al. 1996).

Figure 1.

Models showing the cation–π stacking interaction between Arg 10 and, from left to right, Phe 14, Tyr 14, or Trp 14 in the RF₃, RY₃, and RW₃ peptides, respectively.

Characterization of folded peptides

Circular dichroism (CD) spectra were measured for both the reduced and disulfide cross-linked peptides in 5 mM phosphate buffer (pH 7.0) at 25°C. All the peptides exhibited characteristics of extensively helical peptides with minima at 208 and 222 nm. The only exception appeared to be the EY₃ peptide that, as discussed later, may not be completely helical under these conditions. The oligomerization state of the reduced peptides was also investigated by sedimentation equilibrium analytical ultracentrifugation. The average molecular weight measured for the sedimenting species was in each case consistent with the peptides adopting a trimeric structure, as found in previous studies on KE peptide (Dieckmann et al. 1997). Some small deviation from ideality was evident in fits to the traces (see accompanying supplementary material), which may be due to the presence of a small amount of monomeric peptide, even at the high loading concentrations, ~1 mM, used in these experiments.

Unfolding experiments

Initially we studied the guanidinium hydrochloride (GuHCl)-induced unfolding of the disulfide cross-linked peptides. Unfolding was followed by monitoring the ellipticity of the peptides at 222 nm as a function of increasing GuHCl concentration. The data from these experiments are shown in Figure 2A ▶ and appear well fitted by a two-state equilibrium between folded and unfolded peptides. Surprisingly, substitution of Lys 10 by Arg in RE_S-SRE results in a peptide that unfolds at higher GuHCl concentrations, although the transition is somewhat broader. The net result is that the RE_S-SRE is less stable than the KE_S-SKE peptide (Table 2), even though Arg retains the ability to form a salt bridge with Glu. Of the aromatic-containing peptides, RF_S-SRF has a fairly sharp unfolding transition. In contrast, RW_S-SRW exhibits a somewhat broader transition, and both tyrosine-containing peptides exhibit very broad unfolding transitions.

Figure 2.

GuHCl-induced unfolding of disulfide-cross-linked dimers and reduced peptide trimers followed by circular dichroism. (A) Unfolding transitions of KE_S-SKE (open circles), RE_S-SRE (filled triangles), RF_S-SRF (filled circles), RW_S-SRW (open squares), RY_S-SRY (filled squares), and EY_S-SEY (open triangles). The data are fit to a two-state equilibrium between folded and unfolded peptides. (B) Unfolding transitions of KE₃ (open circles), RE (filled triangles), RF₃ (filled circles), RW₃ (open squares), RY₃ (filled squares), and EY₃ (open triangles). The data are fit to a two-state equilibrium between a folded trimeric bundle and unfolded monomeric peptide.

Table 2.

Thermodynamic data for folding of peptides calculated from GuHCl denaturation curves

Peptide	Δ G°fa (kcal mole−1)	mfb (kcal mole−1 M−1)	ΔΔG°f corrc (kcal mole−1)
KE3	−8.2 ± 0.1	1.8 ± 0.1	−3.0
RE3	−5.9 ± 0.1	1.4 ± 0.1	−0.7
RF3	−7.7 ± 0.1	−2.4 ± 0.1	−2.3
RW3	−5.7 ± 0.1	−0.85 ± 0.03	−0.3
RY3	−5.1d	−0.59 ± 0.02	0.0
EY3	−4.7d	−0.56 ± 0.03	—
KES-SKE	−3.2 ± 0.2	−1.6 ± 0.1	−1.8
RES-SRE	−2.5 ± 0.2	−0.8 ± 0.1	−0.9
RFS-SRF	−2.7 ± 0.2	−1.3 ± 0.2	−1.0
RWS-SRW	−1.7 ± 0.1	−0.86 ± 0.06	0.1
RYS-SRY	−1.2 ± 0.2	−0.6 ± 0.1	0.3
EYS-SEY	−1.1 ± 0.3	−0.68 ± 0.07	—

^a ΔG°_f is the free energy change for folding expressed as a per-strand value for either the reduced trimeric or disulfide-bridged dimeric peptide.

^bm_f is defined as the change in ΔG°_f per unit concentration of GuHCl.

^c ΔΔG°_{f corr} is defined as the difference between the per-strand free energies of folding of each peptide and that of EY₃ (for reduced peptides) or EY_S-SEY (for disulfide cross-linked peptides) after correcting for the differences in helical propensity of the side chains.

^d ΔG°_f for these peptides was estimated by extrapolation of the fit.

We next studied the unfolding of the reduced peptides by GuHCl, shown in Figure 2B ▶. In this case the folded peptides form a triple-stranded coiled coil, and the unfolding transition involves a monomer–trimer equilibrium. This leads to sigmoidal denaturation curves that have an asymmetric appearance, with the curvature of the low GuHCl concentration limb being less sharp than that of the high GuHCl concentration limb. Importantly, if the data were fitted assuming a monomer–dimer equilibrium, significantly worse fits were obtained that exhibited systematic deviation of the data from the curve. This provided further confirmation that the peptides were associating into trimeric coiled coils. The KE₃ RE₃ and RF₃ peptides each show quite sharp denaturation curves that are well fitted by a two-state equilibrium model between folded trimer and unfolded monomer, with KE₃ appearing markedly more stable than either RE₃ or RF₃.

Both the tyrosine-containing peptides, RY₃ and EY₃, exhibit much broader unfolding transitions, and curve fitting indicates that these peptides are not fully folded even in the absence of GuHCl. We note here that whereas GuHCl titrations were performed with peptides at 33 μM concentration, the centrifugation studies that confirmed the trimeric nature of these peptides were performed at the much higher concentrations of peptide (~1 mM) needed to follow sedimentation. Under those conditions all the peptides appear to be associated as trimers.

The free energies of folding, ΔG°_f, and the proportionality constant, m_f, which describes the sensitivity of the peptide to unfolding by GuHCl, were calculated for each peptide; the data are shown in Table 2. For RY₃ and EY₃ peptides ΔG°_f is an estimate because the baseline for the unfolding transition at low denaturant concentrations could not be measured. For all the peptides the trimeric coiled coil adopted by the reduced peptide is more stable, on a per-strand basis, than the corresponding dimeric coiled coil formed by the disulfide cross-linked peptide. It is also evident that replacing the Lys 10–Glu 14 salt bridge in KE by an arginine–glutamate salt bridge results in a less stably folded peptide. As anticipated, the negative control peptides, EY₃ and EY_S-SEY, were the least stable, but surprisingly, perhaps, of those peptides potentially able to form cation–π interactions RF₃ and RF_S-SRF were the most stable. If the electrostatic component of the cation–π interaction was the major cause of the differences in folding energies observed, one would have expected RW₃ and RW_S-SRW peptides to be the most stable (Mecozzi et al. 1996).

Helical propensity correction

It is well established that different amino acid side chains are more or less stabilizing in α-helices, independent of the other specific interresidue interactions. Thus to obtain a better estimate of the interaction energy between the two side chains, a correction must be introduced to account for the intrinsic helix-stabilizing propensities of the residues. Several methods for doing this have been described based on either Lifson–Roig or Zimm–Bragg algorithms (Zimm and Bragg 1959; Lifson and Roig 1961), in particular the AGADIR and SCINT2 programs (Rohl et al. 1996; Munoz and Serrano 1997) have been used previously to correct for helix propensities in studies of cation–π interactions in monomeric helices (Andrew et al. 2002; Tsou et al. 2002). Here, however, we have chosen to use the helical propensity scale developed by O’Neil and DeGrado (1990) as being most appropriate, because this was developed in the context of a coiled-coil peptide system very similar to that which we have used in this study.

The helix propensity of an amino acid side chain is calculated based on its stabilizing or destabilizing effect on the helix relative to glycine in the absence of any specific interresidue interactions. The propensity-corrected free energies were obtained by subtracting the appropriate energy term for each residue at positions 10 and 14, and the resulting free energies of folding were compared with the propensity-corrected folding energy of the reference peptide EY. The residual difference in free energy relative to the reference peptide may be ascribed to the interaction between the two residues. This approach is analogous to the “double mutant” cycles employed to analyze interresidue interactions in larger proteins. The EY peptide was chosen as a reference because the glutamate–tyrosine pairing provides a relatively conservative substitution in which inter-residue interactions should be minimal.

Correcting for helical propensity does not significantly change the order of stability for the peptides (Table 2). Comparing the stabilities of the RF₃ and RF_S-SRF peptides with the control peptides EY₃ and EY_S-SEY, the stability imparted by the Arg–Phe interaction is −2.3 kcal/mole in RF₃ and ~ −1.0 kcal/mole in RF_S-SRF.

Discussion

Whereas most previous investigations of the cation–π effect have focused on monomeric alanine-based helices, we have chosen to investigate this effect in the context of the coiled-coil motif. This motif is commonly observed in proteins, either as a discrete domain or as a subdomain embedded within the structure of a larger protein. We therefore considered that our system would provide a more proteinlike model in which to evaluate this interaction.

Experimentally, the coiled-coil system has several advantages over monomeric helical peptides for assessing the effect of amino acid side chains on stability. These peptides are sufficiently stable that the unfolding transitions may be followed, and in general exhibit two-state melting behavior that can be fitted to an appropriate folding model. In contrast, monomeric helices are marginally stable, and measurements of stability are based on single point determination of “fraction folded” by CD spectroscopy. The free energies extracted from such measurements depend critically on knowing the peptide concentration very accurately and, equally importantly, on being able to accurately calculate the mean residue molar ellipticity of the fully folded peptide, because this is not determined experimentally. As noted by Doig and coworkers, this may be problematic as the presence of tyrosine, for example, perturbs the CD spectrum, giving rise to an incorrect estimation of helicity (Andrew et al. 2002).

As expected, the control peptide, EY, was the least stable, either as a dimeric or trimeric coiled coil, but surprisingly the substitution of arginine for glutamate, which should have provided a good cation–π pairing, results in little, if any, stabilization of the trimeric coiled coil, and may be marginally destabilizing in the dimeric coiled coil. Only the RF peptide, either as RF₃ or RF_S-SRF, proved to be significantly more stable than either the EY or RE control peptides. In particular, the difference in stability between RF₃ and EY₃, ΔΔG°_f = −2.3 kcal/mole, approaches that predicted for the Arg–aromatic interaction, ΔG° ~ −2.9 kcal/ mole, from computational studies of apparent cation–π interactions observed in protein crystal structures (Gallivan and Dougherty 1999). However, it is also significantly larger than those reported in previous investigations of the cation–π effect in monomeric peptides in which ΔΔG°_f in the range 0 to −1 kcal/mole have been measured (Fernandez-Recio et al. 1997; Andrew et al. 2002; Shi et al. 2002b; Tsou et al. 2002). The question therefore arises as to whether this is truly a cation–π effect or whether other physicochemical effects are primarily responsible for this stabilizing interaction.

The magnitude of m_f provides further useful information on the nature of the folding process. For proteins containing the same number of residues, m_f is a measure of the relative change in solvent-exposed area in going from the folded to the unfolded state (Fersht 1999). The sharp transitions (large m_f) exhibited by KE and RF suggest a large decrease in solvent-exposed area on folding whereas the very broad transitions (small m_f) exhibited by the Tyr-containing peptides indicate that the folded state retains greater solvent exposure in these peptides. These observations seem better explained by a model in which desolvation of hydrophobic surfaces is the driving force for folding, and indeed the Tyr side chain is much more hydrophilic than Phe or Trp.

Furthermore, one would expect that the strongest cation–π interaction would be between Arg and Trp as the aromatic ring of Trp is much more electron rich than that of Phe because of electron donation by the indole nitrogen (Mecozzi et al. 1996). In fact, Trp appears only modestly stabilizing compared with the EY control peptides. Also, the Arg–Tyr cation–π interaction is expected to be similar in strength to an Arg–Phe interaction, but after correcting for helix propensity, no stabilizing effect relative to the control EY peptides is evident. The fact that the peptides do not follow the expected trend in stability is perhaps the strongest argument against these differences being due to the cation–π effect.

Energetic differences between dimeric and trimeric coiled coils

For each peptide we have measured the folding energy in two different folded structures: a disulfide cross-linked two-stranded coiled coil and the noncovalently associated three-stranded coiled coil. This has the advantage that ΔΔG°_f between the same two sequences can be compared in two different tertiary structures, rather than simply comparing Δ ΔG°_f between two sequences, something previous model studies have been unable to do. If the interaction being evaluated is “robust”, that is, independent of the context in which it is measured, then ΔΔG°_f should be closely similar.

We found that the two-stranded coiled coil is significantly less stably folded than the three-stranded structure, both in overall terms and on a per-strand basis. A three-stranded coiled coil has a larger per-strand buried surface area and would therefore be expected to be more stable if a trimeric structure allows more hydrophobic surface area to be buried. The two-stranded coiled coil may also be less stable if the interstrand disulfide bridge causes the α-helices to be distorted from ideal coiled-coil geometry.

It is noteworthy that the ΔΔG°_{f corr} values (Table 2) associated with the Arg–aromatic interactions are uniformly less energetically favorable for the two-stranded versus the three-stranded coiled coil. In the ideal case where the interaction between two residues was completely context independent, the per-residue ΔΔG°_{f corr} calculated for the interaction should be the same in both the two-stranded and three-stranded cases, even though the absolute stabilities are significantly different. The Arg residue, which occupies a “b” position in the heptad, is slightly less solvent exposed in the three-stranded coiled coil, and might form favorable interhelical interactions that are not possible in the two-stranded coiled coil structure. However, it is not obvious from examining models why, for example, the Arg–Phe interaction should be weaker in one context than the other.

This suggests that the differences in stability being measured may arise from small differences in many weak interactions caused by the change in sequence, rather than a specific interaction between two functional groups. It is possible that a similar phenomenon may explain the small energetic differences attributed to specific interresidue interactions in monomeric helical peptides as well.

One impetus for investigating the Arg–Tyr interaction was that this was found to be the most common apparent cation–π interaction in a survey of protein structures (Gallivan and Dougherty 1999), and occurs at many surface-exposed positions in proteins. One example in particular, that between Arg 79 and Tyr 83 in the vaccinia virus protein VP39 (Hodel et al. 1996) has several features in common with our model peptide system. These residues occur with the same i, i + 4 spacing and in the context of a parallel two-stranded coiled coil, although in VP39 the coiled coil is imbedded within the more extensive tertiary structure of the protein. Also, the conformations of the side chains observed in the protein can be readily modeled as low energy conformers in our peptide system. The Arg 79–Tyr 83 cation–π interaction in VP39 was estimated to be −4.2 kcal/mole (Gallivan and Dougherty 1999); thus it is perhaps surprising that we observe no stabilizing effect in RY₃ or RY_S-SRY peptides, even allowing for the approximate nature of the calculation and our peptide model.

Closer inspection of the VP39 structure reveals that whereas Tyr 83 is quite solvent exposed, Arg 79 is conformationally restricted by the Leu 106 residue that buttresses the opposite face of the guanidinium group, which, in turn, is further immobilized by a salt bridge with Asp 108, as shown in Figure 3 ▶. The salt bridge should reduce the positive charge on the arginine, making its interaction with Tyr 83 more hydrophobic in nature. Small molecule model studies have demonstrated a synergistic effect of a salt-bridging carboxylate in facilitating the stacking of arginine or lysine against an aromatic ring, and have also noted that hydrophobic effects make a significant contribution to stacking (Thompson and Smithrud 2001).

Figure 3.

A “strong” cation–π interaction between Arg 79 and Tyr 83 observed in the crystal structure of the viral protein VP39. The salt bridge between Arg 79 and Asp 108 and the van der Waals contacts between Arg 79 and Leu 106 that serve to orient the guanidinium group of Arg 79 toward stacking with Tyr 83 are illustrated.

In conclusion, although our studies demonstrate that an i, i + 4 Arg–Phe pairing appears to specifically stabilize both two-stranded and three-stranded coiled coils, overall we have not found convincing evidence for stabilization by the cation–π interactions. This interaction, if present in these peptides, would appear to be significantly weaker than that predicted by computation. It is possible that without other structure-organizing elements, the energetic penalty associated with the loss of conformational freedom and especially desolvation outweigh favorable electrostatic and van der Waals interactions associated with cation–π stacking.

Materials and methods

Reagents

Rink Amide resin, Fmoc-protected amino acids, N-hydroxybenzotriazole (HOBt) and 2-(1H-benzotriazole-1-yl)-1,1,3,3-tetrameth-yluronium hexafluorophosphate (HBTU) were purchased from NovaBiochem. Peptide synthesis grade N-methylpyrollidinone (NMP), N,N-dimethylformamide (DMF), ACS-grade N,N-diisopropylethylamine (DIEA), and piperidine were purchased from Fisher. Guanidinium chloride, 99%+ purity, was obtained from Gibco BRL.

Peptide synthesis and purification

All peptides were synthesized using standard Fmoc protocols on an ABI 433A automated synthesizer. Peptides were cleaved from the resin by stirring for 2 h at room temperature with 10 mL of a mixture of 85% trifluoracetic acid (TFA), 2.5% ethanedithiol, 5% thioanisole, 5% phenol, and 2.5% water. The resin beads were filtered off and rinsed with an additional 1–2 mL TFA. TFA was evaporated from the filtrate under a stream of nitrogen, and 50 mL of cold diethyl ether were added to precipitate the peptide. The crude peptide was collected by filtration on a fritted funnel, dissolved in 10% aqueous acetic acid, and lyophilized. Peptides were redissolved at ~10 mg/mL in 10% aqueous acetic acid and purified by reverse-phase HPLC on a Waters semipreparative C₁₈ column equilibrated in 0.1% TFA and eluted with a linear gradient of 0%–90% acetonitrile containing 0.1% TFA. The peptides were determined to be pure by analytical HPLC and either MALDI-TOF mass spectrometry or electrospray ionization mass spectrometry. The concentration of the peptides was determined by reaction of the single cysteine with Ellman’s reagent, or, for peptides containing tyrosine or tryptophan, by their absorbance at 275 nm or 280 nm, respectively, using an extinction coefficient of 1420 cm⁻¹ M⁻¹ for tyrosine and 5400 cm⁻¹ M⁻¹ for tryptophan.

Formation of disulfide cross-linked peptides

Oxidation of cysteine residues to form disulfide cross-links was performed by bubbling O₂ overnight through a 100 μM solution of peptide in 0.5% aqueous NH₄HCO₃. The extent of oxidation was followed by Ellman’s assay for free thiol groups. After the reaction was complete the peptide was lyophilized to remove the NH₄HCO₃.

Circular dichroism

CD spectra of peptides were recorded with an Aviv 62DS spectropolarimeter at 25°C. Mean residue ellipticities, [θ], were calculated as [θ] = θ_obsd /10lcn, where θ_obsd is the ellipticity measured in millidegrees, c is the molar concentration, l is the cell path length in centimeters, and n is the number of residues in the protein. Titrations with denaturant were performed using an automated dual syringe titrator. Stock solutions were prepared containing 33 μM peptide (concentration of monomer) in 5 mM KHPO₄ (pH 7.0), buffer with and without 8.0 M GuHCl. Samples were allowed to equilibrate for 3 min after each injection of denaturant before the ellipticity at 222 nm was measured.

Curve fitting

The denaturation profiles for the disulfide-linked peptides were analyzed assuming a two-state equilibrium between folded and unfolded forms with no significantly populated intermediates being present (Fersht 1999). For the reduced peptides, the denaturation profiles were analyzed assuming a two-state equilibrium between unfolded monomeric peptide and folded, trimeric bundle, as described previously (Boice et al. 1996). Igor Pro software (Wavemetrics, Inc.) was used to fit the denaturation curves.

Correction for helical propensity

The helix propensity scale of O’Neil and DeGrado (1990) was used to correct for intrinsic side-chain helix stabilizing effects. The Δ Δ G_α values used for various side chains were as follows: Arg, −0.68 kcal/mole; Lys, −0.65 kcal/mole; Trp, −0.45 kcal/mole; Phe, −0.41 kcal/mole; Tyr, −0.17 kcal/mole; Glu, −0.27 kcal/mole.

Electronic supplemental material

CD spectra for the reduced trimeric and disulfide-cross-linked peptides and analytical ultracentrifugation traces for the reduced peptides are provided as supplementary material.