Effects of side chains in helix nucleation differ from helix propagation

Significance

Complete description of the kinetics and thermodynamics of α-helix formation is fundamental to the understanding of protein folding because α-helices are the most abundant class of secondary structures and are implicated in the earliest steps of the folding process. Kinetic models of protein folding suggest that helix folding is rate-limited by formation of a nucleus followed by rapid propagation. The influence of individual residues on propagation has been evaluated in numerous model peptides and proteins. Here, we describe a synthetic model that enables experimental assessment of the role of individual residues in helix nucleation. Our results suggest that amino acids contribute differently to nucleation than to propagation.

Abstract

Helix–coil transition theory connects observable properties of the α-helix to an ensemble of microstates and provides a foundation for analyzing secondary structure formation in proteins. Classical models account for cooperative helix formation in terms of an energetically demanding nucleation event (described by the σ constant) followed by a more facile propagation reaction, with corresponding s constants that are sequence dependent. Extensive studies of folding and unfolding in model peptides have led to the determination of the propagation constants for amino acids. However, the role of individual side chains in helix nucleation has not been separately accessible, so the σ constant is treated as independent of sequence. We describe here a synthetic model that allows the assessment of the role of individual amino acids in helix nucleation. Studies with this model lead to the surprising conclusion that widely accepted scales of helical propensity are not predictive of helix nucleation. Residues known to be helix stabilizers or breakers in propagation have only a tenuous relationship to residues that favor or disfavor helix nucleation.

The α-helix is the most prevalent secondary structure in proteins and can form extremely rapidly. Helix formation is thus crucial in early steps of protein folding, and a complete description of the kinetics and thermodynamics of α-helix formation is fundamental for understanding protein folding (1). Theoretical models of the helix–coil transition consider helix formation to proceed in two steps: initial nucleation of a helical sequence, denoted by the parameter σ in the Zimm–Bragg notation (2), followed by more favorable helix propagation, denoted by s parameters. This distinction identifies the energetically unfavorable organization of three consecutive amino acid residues to form a helical nucleus as the slow step in helix formation (2–4). Helix propagation, in contrast, refers to the addition of the next hydrogen bond to a preformed helix (Fig. 1A). Zimm–Bragg or equivalent theories posit that formation of short peptide helices in water is unfavorable because the large decrease in entropy required for nucleation is not adequately compensated by the enthalpic gain from forming a small number of hydrogen bonds.

Fig. 1.

Preorganization of three residues into an α-turn conformation is the energy-demanding step in helix formation. (A) Models of the helix–coil transition consider helix formation to proceed in two steps consisting of nucleation and propagation steps. (B) In hydrogen bond surrogate (HBS) α-helices, the nucleus is organized by replacement of a main chain i to i + 4 hydrogen bond with a covalent bond. The dsHBS helices feature a reversible disulfide linkage. (C) Possible intermediates for the conversion of bisthiol I to dsHBS IV include bisthiol II in α-turn conformation and disulfide III as the constrained helix nucleus. Rates of conversion of bisthiol to disulfide were measured as a function of variable residue Λ.

The ability of different peptide sequences to adopt helical conformations has been rigorously investigated, and the stability of α-helices can be estimated from widely used scales of helical propensity (5). These scales are important for our understanding of protein structure, folding, and function. The classical studies for determination of helix propensities have used peptides, coiled-coils, and protein models for host–guest studies in which a guest residue is systematically substituted at a site in a host structure (6–12). Helix–coil transition theory then relates changes in the conformational stability to microscopic helix propensities or s constants for individual residues. The results provide a quantitative basis for evaluating the behavior of peptide and protein helices; for example, the s constant for alanine is strongly helix stabilizing, whereas those of valine and other β-branched residues are destabilizing (11, 13). However, even the most detailed investigations of the stability and dynamics of helix formation treat the energetic contribution of helix nucleation as independent of local sequence effects (14). The ability to differentiate individual σ values from individual s values could provide a deeper understanding of the impact of individual side chains on helix formation than provided by current models (6, 12).

Here, we describe an approach that estimates the population of N-terminal tripeptide sequences organized in an α-helix nucleus as a function of individual guest residues. The key difference between this investigation and classical measures of propensity is that our approach assesses the ability of a given residue to favor or disfavor nucleation; literature propensities are largely derived by measuring how the stability of a preformed helix responds to different residues in a peptide sequence. Nucleation and propagation effects are intrinsically convoluted. The current approach offers a direct measure of helix nucleation and allows nucleation effects to be interrogated independently of propagation. Although previous studies have implemented correction factors to account for structural effects on nucleation in calculations of helix propensities, such as N-caps and C-caps (15), no independent and general measure of nucleation has been available (6, 12, 16, 17). The results of our study are in fact surprising, as they suggest that steric effects in alkyl side chains, such as those from β-branching, do not have a strong impact on helix nucleation. This result stands in contrast to the observed negative effects of β-branched residues on helix propagation (13, 18, 19).

Results and Discussion

Our approach to assess helix nucleation uses hydrogen bond surrogate (HBS) α-helices in which the N-terminal main-chain hydrogen bond is replaced with a covalent bond (Fig. 1B) (20). In earlier reports, we described HBS helices in which the hydrogen bond is replaced with a carbon–carbon bond (21). Characterization of these synthetic helices by CD and 2D-NMR spectroscopy, as well as single crystal X-ray diffraction analysis, reveals that they faithfully mimic the conformation of canonical α-helices (21, 22). Importantly, HBS helices are able to inhibit intracellular protein–protein interactions mediated by α-helical domains, supporting the hypothesis that these compounds reproduce the structure and function of protein α-helices (23, 24). To analyze the effect of different residues on α-helix formation, we prepared an HBS analog with a disulfide linkage whose rates of formation could be monitored under aqueous conditions (25). We conjectured that the rates of formation of this linkage would provide a unique probe for examining biophysical parameters related to helix formation. Because the disulfide–HBS (dsHBS) linkage mimics the intramolecular hydrogen bond in α-helices, we hypothesized that the bisthiol (bt) to disulfide (ds) oxidation would provide a direct measure of helix nucleation (Fig. 1C). Rates of helix formation for model peptides have been measured by ultrafast spectroscopies and fall in the nanosecond range (14, 26–29). The rates of disulfide formation are much slower (on the order of minutes) under our reaction conditions (30). The dramatically slower timescale for disulfide formation suggests that sequence-dependent differences in the rate of disulfide formation reflect preequilibrium populations: residues with higher nucleating propensity favor disulfide formation.

Disulfide-linked HBS helices can be derived from oxidation of the parent bisthiol peptides allowing a facile approach to control the conformation of the peptide (31–34). We based our initial peptide design on a short segment from the p53 activation domain: XQEG*FSDLWKLLS (1), where X and G* represent the dsHBS-constrained residues (35). The p53 sequence was chosen because it has relatively few side-chain contacts such as ionic bridges, which may bias the results. We have previously characterized a number of p53 HBS analogs by CD and NMR spectroscopies allowing us to predict potential aggregation issues that can affect stabilized constructs (36–38). The bisthiol p53 peptide is weakly helical in aqueous buffers at 295 K. CD spectroscopy (Fig. 2A) shows that the p53 dsHBS 1 is highly helical compared with the parent bt-peptide (25). We selected A, G, D, I, K, L, P, and V as the guest residues for this investigation. This selection includes representative aliphatic straight chain and β-branched residues along with positively and negatively charged side chains.

Fig. 2.

Conformational analysis and synthesis of unconstrained peptides. (A) CD spectroscopy suggests that the disulfide-linked (ds) HBS peptide is highly helical compared with the parent bisthiol (bt). The CD studies were performed with 50 μM peptides in 1 mM PBS. (B) NMR-derived structures of ds-2A. Views of 20 lowest energy structures are shown with carbon, nitrogen, and oxygen atoms in gray, blue, and red, respectively. The disulfide linkage is shown in yellow color. (C) Schematic for the oxidation reaction. The conversion of bt-2Λ → ds-2Λ was affected under mild oxidative conditions; the unreacted bisthiol was quenched with maleimide 3.

After initial experiments, peptide 1 was modified by replacing the native glutamic acid and phenylalanine residues with lysine and alanine, respectively, to obtain 2Λ: XΛKG*ASDLWKLLS. The new sequence was designed to avoid potential side-chain interactions between the guest residues (Λ) and the corresponding i + 3 position; lysine was incorporated to increase water solubility of the host. The second position was chosen for the introduction of guests because the conformation of this residue is implicated in helix induction near the N terminus (39, 40). The helical conformation of ds-2A: XAKG*ASDLWKLLS was assessed using a combination of 1D NMR, total correlation spectroscopy, and rotating-frame nuclear Overhauser effect correlation spectroscopy (ROESY) studies in aqueous solutions (SI Text, Table S1, and Fig. S1). The ROESY spectra revealed nuclear Overhauser effect (NOE) cross-peaks indicative of a helical structure, including sequential NH-NH (i and i + 1) and several longer range NOEs (between i and i + 3/i + 4) (SI Text, Table S2). Importantly, all backbone ϕ dihedral angles, calculated from ³J_NHCHα coupling constants, are in the expected range for helical conformations (SI Text, Table S1) (41, 42). The coupling constants observed for the residues within the macrocycle do not differ from the rest of the chain, indicating that the disulfide constraint is faithfully inducing an α-helical conformation. The solution structure of ds-2A was determined from 48 ROESY cross-peaks and 11 calculated ϕ angles using Monte Carlo conformational search in Macromodel 2014. An overlay of the 20 lowest energy structures for ds-2A is shown in Fig. 2B. Notably, the disulfide-constrained macrocycle does not perturb the N terminus of the α-helix. Overall, together with CD spectroscopy, the NMR studies confirm stabilization of a major α-helical conformation in dsHBS-constrained peptides.

The bisthiol and HBS peptides were synthesized as shown in SI Text, Fig. S2. We used mild oxidation conditions consisting of 2% (vol/vol) DMSO to oxidize the bisthiols (Fig. 2C) (25, 30). To separate the closely related bisthiol and dsHBS peptides on HPLC and to quench the thiol groups, we treated the reaction mixture with a maleimide derivative (43). Adequate separation of the bisthiol and disulfide peptides on HPLC allowed us to quantify the rates of disulfide formation for each sequence from analysis of HPLC peak areas (Fig. 3A and Fig. S3). The initial rates of disulfide formation are plotted in Fig. 3B, and the tabulated data are presented in Table 1.

Fig. 3.

Analysis of bisthiol to disulfide conversion. (A) Rates of the bt-to-ds conversion were analyzed by HPLC, with tryptophan as an internal control. Representative HPLC results for peptide 2 with alanine (Λ = A) as the guest residue are shown. (B) Plots of bt-to-ds conversion for different guest residues. (C) CD spectra of sequences with different guest residues illustrate that the helical content of most sequences, with the exception of sequences with Λ = G, P, and D, is identical. The CD studies were performed with 50 μM peptides in 5 mM KF, pH 7.3.

Table 1.

Comparison of dsHBS rate of formation to literature helix propensity values

Peptide†	Rate of bt → ds conversion, μM/min‡	Rates relative to 2G	∆∆G from propensity, kcal/mol§
XAKG*ASDLWKLLS (2A)	0.20 ± 0.02	2.86	−1.88
-–D––––––––––––––– (2D)	0.06 ± 0.01	0.86	−1.00
-–G––––––––––––––– (2G)	0.07 ± 0.005	1.00	0.00
-–I–––––––––––––––– (2I)	0.15 ± 0.01	2.14	−1.18
-–K––––––––––––––– (2K)	0.08 ± 0.01	1.14	−1.52
-–L––––––––––––––– (2L)	0.19 ± 0.01	2.71	−1.60
-–P––––––––––––––– (2P)	––	––	>5.00
-–V––––––––––––––– (2V)	0.20 ± 0.01	2.86	−0.83

^†X and G* represent the dsHBS constraint.

^‡Calculated from HPLC analysis.

^§From ref. 19.

We find that the nucleation propensity of alanine is stronger than that of glycine, whereas the disulfide bond formation with proline as guest is too slow to measure accurately. These results are consistent with known relationships: glycine is a highly flexible residue, whereas proline aids helicity only as an N-cap residue and not at other positions in the helix. These results also indicate that conformation of the macrocycle influences rates of the disulfide formation. If the rates of the bond formation were independent of the macrocycle conformation, substitution of a single glycine in place of an alanine would not cause a significant effect. It is likely that the peptide chain beyond the putative macrocycle is influencing the preequilibrium conformation but because the bisthiol peptide is largely non–α-helical, we expect this effect to be subtle and equivalent for every mutant.

Leucine and alanine sequences are similar in their disulfide formation rates, as might be expected from the literature propensity values (Table 1) (6). These results provide useful metrics to gauge the potential of our synthetic model as a probe of helix nucleation. Surprisingly, our investigations reveal that β-branched residues are not detrimental to α-helix nucleation. Indeed, the rate of the disulfide formation with valine is nearly the same as that of alanine and leucine, whereas isoleucine leads to a slight decrease. Charged residues, lysine and aspartate, slow down the reaction rate to the level of glycine. Because aspartic acid is a known helix breaker whereas lysine is considered to be a helix stabilizing residue (44, 45), these results suggest that charged residues potentially participate in long-range interactions. Although the sequences of this study were designed to minimize context dependence, backbone interactions cannot be completely ruled out. CD spectroscopy shows that five of the eight dsHBS sequences have very similar helical content; substitution of glycine, proline, and aspartic acid lowers the helicity (Fig. 3C). We conjecture that helical stability of the constrained sequence is not a major determinant of disulfide rate formation. Comparison of the lysine and aspartic acid sequences provides support for this hypothesis, because these sequences display similar rates of disulfide formation, whereas the constrained peptides show different helical stabilities. Charged residues at the end of helices can positively or negatively influence helix stability based on their interactions with the helix macrodipole (46). However, such macrodipolar interactions should not influence the process of helix nucleation because the helix is not yet organized.

To obtain theoretical support for our experimental observations, we calculated activation barriers to formation of an i to i + 4 hydrogen bond in peptide sequences using metadynamics simulations (47, 48). Helix–coil transitions have been previously investigated using molecular-dynamics simulations to analyze helix propensity and nucleation timescales, but, to our knowledge, the effect of point mutations on helix nucleation has not been explored systematically (28, 39, 49). We determined the impact of different residues on the formation of an α-turn in an acetylated and methylamidated model peptide sequence, Ac-AΛA-NHMe. We used Schrodinger’s 2012 distribution of Desmond 3.1 (50) to conduct 50-ns metadynamics simulation on the tripeptide using the Amber forcefield ff12SB and associated monovalent ion parameters (51, 52). Two-dimensional free-energy surfaces were determined and activation energies were identified by inspection (Fig. 4). The activation energies stabilize as simulation length increases, suggesting that the simulation converged at 15 ns (Fig. S4). We also performed the simulations in triplicate and measured the average rmsd between the resulting free-energy surfaces (Fig. S4). Analysis reveals that all guest (Λ) residues, except proline, sample dihedral angles corresponding to left- and right-handed α-helix, β-strand, and polyproline II (PPII) regions of the Ramachandran plot (Fig. 4A). The plot of the proline containing tripeptide features only the α and PPII minima (Fig. 4B). The [ϕ,ψ] plots for the other 18 guest residues in Ac-AΛA-NHMe are shown in Fig. S5. The weighted fraction of α, β, and PPII populations for different guest residues is plotted in Fig. 4C. To calculate the barrier for helix formation, we compared the height of the shortest peak between the lowest energy basin corresponding to the PPII or β dihedral angles to the α-helical dihedral for each residue (Fig. 4D). The PPII dihedral angles were found to be the lowest energy conformation for most guest residues in our calculations.

Fig. 4.

Results of metadynamics simulation to evaluate barrier to the formation of an i to i + 4 hydrogen bond in a model peptide (AcAΛA-NHMe) as a function of different guest residues. (A and B) Two-dimensional free-energy surfaces for Λ = alanine and proline are shown with the others included in SI Text. The Ramachandran plots of all residues show low energy basins for the α, β, and polyproline II (PPII) dihedral space with the exception of proline, where the PPII and α spaces dominate. (C) Weighted populations of all low-energy basins corresponding to the α, β, and PPII dihedral angles. (D) The activation energy barrier between the lowest energy region corresponding to the PPII space and α-helical dihedral angles.

The results of these calculations correlate with the experimental data and suggest that the barrier to preorganization of the α-helical conformation does not follow propensity scales. Although the activation energies reflect starting populations and energy profiles of the α and β or PPII conformations, they provide a gauge for estimating the residue-dependent difficulty of adopting an α-helical conformation. The calculated ∆G^‡ values identify guest residues with fast and slow folding rates but suggest that rates of several others are not readily distinguishable. Overall, the calculations support the experimental observations that helical propensity scales are not applicable to helix nucleation.

Conclusions

We have developed a synthetic model to evaluate propensities for helix nucleation. Although the organization of an α-turn leading to an intramolecular hydrogen bond between the C=O of the ith residue and the amide NH of the i + 4th residue is considered the slow step in helix formation, there are no reported experimental attempts to segregate effects of individual side chains on helix nucleation. A large number of studies have focused on the propensity of amino acid residues to propagate a preformed helix in peptide and protein models. To experimentally interrogate the nucleation step, we replaced the intramolecular hydrogen bond at the N-terminus of a short α-helix by a disulfide linkage whose rates of formation from the analogous bisthiol could be measured as a function of individual residues in the turn. We also performed metadynamics simulations on tripeptide sequences to estimate the barrier to formation of a hydrogen bond between the ith and the i + 4th residues.

The thrust of these experimental and computational studies reveals for the first time (to our knowledge) that helix nucleation differs from propagation in an important aspect: although folding of a chain to create a nucleus is sensitive to the effects of a rigid template (as observed in crystal packing and growth), the process of nucleation minimizes what must be relatively minor enthalpic and entropic penalties due to side-chain packing. The large entropic barrier that must be overcome to organize three consecutive residues predominates in nucleation in marked contrast to the side-chain steric constraints that impact helix propagation (18).

Methods

General.

Commercial-grade reagents and solvents were used without further purification, except as indicated. Peptides were purified using a Beckman Coulter HPLC equipped with a System Gold 168 Diode array detector, equipped with a reversed-phase C-18 column, and buffers containing 0.1% TFA in acetonitrile and 0.1% TFA in water. Purified peptides were analyzed by liquid chromatography (LC)/MS on an Agilent 1100 Series LCMSD system. High-resolution MS data were obtained using an Agilent 6224 Accurate-Mass TOF LC/MS. HPLC analysis for the conversion of bt to ds peptides was performed using an Agilent 1200 series HPLC system with a Thermo Hypersil C-8 column (50 × 2 mm, 3 μm; 5–65% acetonitrile in water over 10 min). CD spectra for dsHBS helices were recorded on an AVIV 202SF CD spectrometer equipped with a temperature controller using a 1-mm length cell and a scan speed of 5 nm/min. Samples were run in 5 mM KF buffer (pH 7.3) or 0.1× PBS buffer (pH 7.3) at a concentration of 50 μM ds-2Λ (concentration determined by Trp residue absorbance, ε = 5,560 cm⁻¹⋅M⁻¹ at 280 nm). The spectra were averaged over 10 scans with the baseline subtracted from analogous conditions to those of the samples. NMR experiments were performed using a Bruker Avance 600-MHz spectrometer. IR analysis was performed using a Thermo Nicolet 6700 FT-IR spectrometer, equipped with Smart ITR. Solids were dissolved in methanol and dried on smart sensor.

Representative Procedure for the Oxidation of bt-2Λ → ds-2Λ.

Disulfide peptides are prone to oxidation and were stored under inert atmosphere as lyophilized powders. Concentrations of purified bt-2Λ peptides were determined under acidic conditions (0.1% aqueous TFA) to minimize oxidation. Lyophilized bisthiol peptides (200 nmol) were dissolved in 40 μL of DMSO in a centrifuge tube, and oxidation was initiated by addition of 2 mL of oxidation buffer (800 mM ammonium bicarbonate, 840 mM acetic acid, pH 6.0). The oxidation buffer was supplemented with 50 μM tryptophan as an internal control for HPLC studies. Fifty-microliter aliquots were removed from the reaction mixture after set intervals (T = 0, 1, 2, 5, 10, 20, and 30 min), quenched with 25 μL of 1.4 mM maleimide 3 in H₂O, and analyzed by HPLC.

Metadynamics Simulations.

Acetylated and methylamidated peptide structures were initialized with tleap (AmberTools12) using the ff12SB force field and monovalent ion parameters from Joung and Cheatham (51, 52). Structures were neutralized by adding chloride or sodium ions and surrounded by a truncated octahedron of TIP3P water with a buffer distance of 8.0 Å. The resulting topology and input coordinates were converted to Desmond cms format for simulations with Desmond 3.1.

The ϕ and ψ dihedral angles of the guest residue were chosen as the reaction coordinates for metadynamics column variables. The standard Desmond relaxation protocol was conducted on each tripeptide: 2,000 steps minimization with 10 steps steepest descent with 50 kcal/mol restraints on peptides, 2,000 steps minimization with 10 steps steepest descent without restraints, 12 ps of Berendsen NVT simulation at 10 K, and finally 24 ps of Berendsen NVT equilibration at 300 K. A 50-ns metadynamics simulation was conducted on each peptide, depositing 5°-wide, 0.01-kcal-high Gaussians every 12 fs. The resulting 2D free-energy surfaces were analyzed manually. The transition state connecting the α-helical conformation to the nonhelical minimum energy conformation was identified and the activation energy was calculated.