Polyproline II helical structure in protein unfolded states: Lysine peptides revisited

Abstract

The left-handed polyproline II (PPII) helix gives rise to a circular dichroism spectrum that is remarkably similar to that of unfolded proteins. This similarity has been used as the basis for the hypothesis that unfolded proteins possess considerable PPII helical content. It has long been known that homopolymers of lysine adopt the PPII helical conformation at neutral pH, presumably a result of electrostatic repulsion between side chains. It is shown here that a seven-residue lysine peptide also adopts the PPII conformation. In contrast with homopolymers of lysine, this short peptide is shown to retain PPII helical character under conditions in which side-chain charges are heavily screened or even neutralized. The most plausible explanation for these observations is that the peptide backbone favors the PPII conformation to maximize favorable interactions with solvent. These data are evidence that unfolded proteins do indeed possess PPII content, indicating that the ensemble of unfolded states is significantly smaller than is commonly assumed.

Protein folding is the process by which a polypeptide chain makes the transition from some unfolded or denatured state to its native fold (Anfinsen 1973). Much is known about the end state of protein folding, yet little is known about the starting state. However, understanding how a polypeptide chain negotiates its way into its final active structure requires some quantitative understanding of the state in which the process starts (Dill and Shortle 1991). It generally is believed that this state is actually a vast ensemble of random conformations, although there is a growing body of evidence indicating that this is an erroneous belief (Lattman et al. 1994; Lumb and Kim 1994; Wilson et al. 1996; Shortle and Ackerman 2001).

More than 30 years ago, Tiffany and Krimm (1968b) hypothesized that some protein unfolded states consist of short stretches of left-handed polyproline II (PPII) helix-like conformations interspersed with bends. This hypothesis was based in part on the similarities between circular dichroism (CD) spectra for denatured proteins and the CD spectrum for a homopolymer of proline, which is known to adopt the PPII conformation. Although some of Tiffany and Krimm's arguments have since been discounted or remain controversial, there remains strong evidence in favor of this hypothesis (Woody 1992). Their hypothesis has been examined by several groups over the years, including Drake et al. (1988), Dukor and Keiderling (1991), Wilson et al. (1996), and Park et al. (1997). Most recently, Z. Shi, C.A. Olson, G.D. Rose, R.L. Baldwin, and N.R. Kallenbach (pers. comm.) have shown unequivocally that short alanine peptides exist predominantly in the PPII helical conformation in aqueous solution, a finding that will lead to a greater understanding of the nature of protein unfolded states.

An ideal left-handed PPII helix has backbone dihedral angles (φ, ξ) = (−75°, +145°), resulting in precisely three residues per turn (Adzhubei and Sternberg 1993; Williamson 1994; Creamer 1998). This conformation is adopted by homopolymers of proline in aqueous solution (Isemura et al. 1968; Tiffany and Krimm 1968a; Helbecque and Loucheux-Lefebvre 1982; Bhatnagar and Gough 1996). The right-handed polyproline I helical conformation has similar backbone dihedral angles but has all peptide bonds in the cis conformer. This conformation is observed for homopolymers of proline in organic solvents and is disfavored in aqueous solution (Mutter et al. 1999).

It has been known for many years that homopolymers of lysine adopt the PPII helical conformation at low or neutral pH (Tiffany and Krimm 1968b, 1972; Drake et al. 1988; Makarov et al. 1994). This structure can be disrupted somewhat by the addition of sodium chloride (Drake et al. 1988), although the CD spectrum still possesses PPII character. Neutralization of side-chain charges by raising the pH induces α-helix formation, indicating that electrostatic interactions play a role in stabilizing the PPII helical structure of poly(lysine) (Arunkumar et al. 1997). Subsequent heating of the pH-induced α-helical state results in β-sheet formation (Shibata et al. 1992). Homopolymers of aspartate and glutamate also are known to possess CD spectra consistent with PPII helix (Woody 1992), providing further support for the role of electrostatic interactions.

Here, we show that a short lysine peptide also adopts a PPII helical structure in neutral aqueous solution. Unlike long homopolymers of lysine, our peptide consisting of seven lysines does not form an α-helix at high pH but rather retains significant PPII helical structure. Titration with sodium chloride up to 4.0 M results in a decrease in PPII content but does not abolish the characteristic PPII CD signal. We believe that these data show that short lysine peptides adopt PPII helical structure as a result of the nature of the backbone rather than as a consequence of electrostatic interactions between side chains. That backbone favors this conformation was previously suggested by Krimm and Tiffany (1974) based on the CD spectrum of poly(Ala-Gly-Gly).

Results and Discussion

CD spectra for two peptides, Ac-(Lys)₇-Gly-Tyr-NH₂ (K7) and Ac-(Pro)₇-Gly-Tyr-NH₂ (P7), collected at pH 7.0 and a temperature of 20°C, are given in Figure 1 ▶. The P7 peptide previously has been shown to be a PPII helix via a combination of CD and NMR spectroscopy (Kelly et al. 2001). The K7 peptide possesses a CD spectrum similar to that of P7, with positive and negative bands characteristic of a PPII helical conformation. These bands are moved to lower wavelengths (218 and 195 nm, respectively) compared with those for P7 (228 and 205 nm) as a result of differences in transition energies between tertiary, secondary, and primary amides (Woody 1992). A temperature titration of K7 from 5 to 90°C results in a loss of structure, as adjudged by the weakening of the positive band in the CD spectra at 218 nm (data not shown). The behavior observed is identical to that reported first by Tiffany and Krimm (1972) and later by Makarov et al. (1994) for a homopolymer of lysine.

Fig. 1.

Circular dichroism spectra for the K7 (black spectrum) and P7 (gray spectrum) peptides in pH 7.0 solutions at 20°C.

The CD spectrum for K7 (Fig. 1 ▶) is identical in shape to those obtained by previous workers for long homopolymers of lysine under similar conditions (Tiffany and Krimm 1972; Drake et al. 1988; Shibata et al. 1992; Arunkumar et al. 1997). The per residue molar ellipticities for the two characteristic bands are somewhat weaker for K7 than for lysine polymers, indicating a lower per residue PPII helical content. Nonetheless, it is clear that K7 possesses significant PPII helical content. If lysine polymers and peptides adopt the PPII helical conformation as a result of electrostatic repulsion between the positively charged side chains, it seems remarkable that a seven-residue lysine peptide would possess such a strong PPII helix signal. The short lysine peptide used here could readily undergo fraying at the ends of a PPII helix. The terminal lysine side chains can point away from the bulk of the peptide, allowing their respective backbones to adopt conformations other than PPII helix. Furthermore, the two lysines adjacent to those at the termini also will be somewhat capable of pointing away from the center of the peptide, potentially allowing those backbones to adopt non-PPII conformations. One would expect electrostatic repulsion between side chains to promote such fraying in our short peptide, lowering the PPII helix content and weakening the subsequent CD signal. Despite this, the PPII signal is strong.

Assuming that K7 is adopting a PPII helical conformation as a result of electrostatics, it would be reasonable to expect that the addition of high concentrations of salt would result in significant screening of the side-chain charges and significant weakening, or even abolishment, of the PPII CD signal. Figure 2 ▶ is a plot of the CD spectra for the K7 and P7 peptides in 0 and 4 M sodium chloride solutions at 20°C. The spectra could not be collected below a wavelength of 200 nm because of the absorbance properties of the salt solution. Clearly, there is a decrease in PPII helix content for the K7 peptide as adjudged by the weakening of the positive band at 218 nm. The difference between the CD spectrum without salt and with 4 M sodium chloride is shown in the inset in Figure 2 ▶. It would appear that the K7 peptide is losing PPII helical character, but not gaining other structure (e.g., α-helical structure). Taken by themselves, these spectra for K7 would support the argument that a lysine peptide adopts the PPII helical conformation as a result of electrostatic interactions. However, the P7 peptide also undergos a loss of PPII helical character in 4 M sodium chloride (Fig. 2 ▶). This observation is consistent with the findings of Tiffany and Krimm (1968a), Mattice and Mandelkern (1970), and Dukor and Keiderling (1991) who all found that the PPII helical content of proline homopolymers or peptides decreased with increasing salt concentration. Clearly, in the case of P7, this is not a case of screening charges because this peptide does not possess any formal charges. More likely this is a result of the chaotropic action of concentrated NaCl. The K7 peptide must undergo similar effects in addition to electrostatic screening. The loss of PPII structure is certainly greater for K7 than for P7, which could be expected given the preceding arguments. However, K7 clearly possesses significant PPII helical character even in 4 M sodium chloride, casting some doubt as to the relative importance of electrostatic interactions in determining this structure for this particular peptide.

Fig. 2.

Circular dichroism spectra for K7 in low- (solid black line) and high- (4 M sodium chloride; dashed black line) salt concentrations, and for P7 at low- (solid gray line) and high- (dashed gray line) salt concentrations. All spectra were collected at pH 7.0 and 20°C. (inset) Differences between the K7 (black line) and P7 (gray line) low- and high-salt spectra.

CD spectra for the K7 peptide in neutral solution (pH 7.0) and at pH 12.0, both at 20°C, are shown in Figure 3 ▶. The CD spectrum for the P7 peptide at pH 12.0 is essentially identical to that acquired at pH 7.0 (data not shown for reasons of clarity). There is a definite loss of PPII character for K7 at pH 12.0. The characteristic positive band moved to 215 nm, and the negative band moved to 197 nm, and a negative band at ∼226 nm appeared. This additional negative band is probably the result of the superposition of the PPII and unordered CD spectra (Krimm and Tiffany 1974). Notably, however, the K7 spectrum is still PPII helix-like in character, in contrast with homopolymers of lysine which, at pH 12.0, adopt the α-helical conformation (Arunkumar et al. 1997). The spectrum taken at pH 12.0 is remarkably similar to that observed by Z. Shi, C.A. Olson, G.D. Rose, R.L. Baldwin, and N.R. Kallenbach (pers. comm.) for an alanine peptide at pH 7.0 and 15°C. These workers have confirmed via NMR that this alanine peptide is indeed predominantly a PPII helix under these conditions. The difference between the K7 spectra at pH 7.0 and pH 12.0 is shown in the inset to Figure 3 ▶. This difference spectrum resembles a distorted PPII helix spectrum, or, if inverted, is vaguely β-sheet-like in shape. The difference spectrum then would indicate that K7 has lost some PPII helical character at high pH, with an increase in the population of disordered states and/or the appearance of some small amount of β-sheet. What is clear is that K7 still adopts the PPII helical conformation even at a pH at which the lysine side chains are predominantly uncharged.

Fig. 3.

Circular dichroism spectra for the K7 peptide at pH 7.0 (solid black line) and at pH 12.0 (dashed black line) at 20°C. (inset) Difference between the pH 7.0 and pH 12.0 spectra.

The data presented in Figures 2 and 3 ▶ ▶ indicate that K7 adopts the PPII helical structure in part for reasons other than electrostatic repulsion between the side chains. The K7 peptide still possesses PPII character when the side-chain charges are heavily screened, or even neutralized. Electrostatic repulsion between side chains clearly is not the dominant determinant of the PPII helical conformation adopted by K7. One is led to ask what other factors may contribute. One possibility is hydrophobic interactions between the side chains. Lysine side chains possess significant hydrophobic character in their hydrocarbon portions, and it is possible that these interact, burying hydrophobic surface area, in situations in which the electrostatic interactions have been significantly weakened. However, side chains on one side of an ideal PPII helix are spaced ∼9 Å apart (Cβ to Cβ), a distance that would preclude the kind of significant and systematic interactions that would be required to stabilize the PPII helical conformation.

One also can rule out side-chain-to-side-chain hydrogen bonds because charged lysine side chains can act as hydrogen bond donors but not as acceptors. A hydrogen bond between the lysine side chain and the backbone carbonyl oxygen of the next residue in sequence is a possibility. Such a hydrogen bond has been hypothesized to be a determinant of the remarkably high propensity of glutamine to adopt the PPII helical conformation (Kelly et al. 2001). However, one must question whether such a hydrogen bond would be particularly favorable in 55 M water. In addition, a search of 275 high-resolution protein structures found relatively few such hydrogen bonds: 30 occurrences out of 7419 lysines in the data set, with only 12 of the 30 lysines being found close to the PPII helix region of (φ.ξ) space (data not shown). It seems unlikely then that side-chain-to-backbone hydrogen bonds are a major determinant of the propensity for K7 to adopt PPII helical structure.

A final possibility involving the side chains would be side-chain conformational entropy. The PPII helical conformation is certainly very extended. The side chains are relatively far apart and pointing perpendicularly away from the helical axis, allowing for significant freedom. However, in models for protein unfolded states the lysine side chain also has significant freedom, with entropies estimated in the range TS = 1.8–2.1 kcal/mole (T = 298 K; Lee et al. 1994^; Creamer 2000). It seems unlikely, then, that there would be any significant gain in side-chain entropy when K7 is in the PPII helical conformation. Moreover, there must be a significant decrease in backbone conformational entropy moving from disordered conformations to the PPII helix.

Given that the major determinants of the PPII conformation as adopted by K7 do not arise from the side chains, other than electrostatic repulsion, one is left to conclude that the backbone somehow favors adoption of this structure. This is in keeping with the observations of Z. Shi, C.A. Olson, G.D. Rose, R.L. Baldwin, and N.R. Kallenbach (pers. comm.), who have shown conclusively that a short alanine peptide adopts the PPII conformation. In addition, it has been known for some time that poly(glycine) can adopt a PPII-like structure (Bhatnagar and Gough 1996). Furthermore, both alanine and glycine possess high propensities to adopt the PPII helical conformation in poly(proline)-based host peptides (Kelly et al. 2001). One can readily argue that alanine, and particularly glycine, are all backbone in composition. Backbone entropy loss would oppose adoption of any dominant structure, intrapeptide backbone-to-backbone hydrogen bonds are not possible in such an elongated structure, and stabilizing side-chain-to-backbone hydrogen bonds appear to be an unlikely major determinant, as discussed above. This leaves two possibilities: interpeptide interactions and backbone interactions with solvent. Interpeptide interactions seem unlikely at pH 7.0 and low-salt concentrations because a PPII backbone conformation is incompatible with significant interactions between peptides (e.g., hydrogen bonding). We may be detecting some β-sheet formation by the K7 peptide at high pH (Fig. 3 ▶), but this results in a weaker PPII helix signal rather than a strengthening of the conformation.

This leaves backbone solvation as the one remaining major possibility, as indicated by the molecular dynamics computer simulations of Sreerama and Woody (1999). Kelly et al. (2001) also hypothesized that backbone solvation is a major determinant of the PPII helical propensities measured for a variety of residues in a poly(proline)-based host peptide system. The reasoning is that short bulky side chains partially occlude their own backbone amide hydrogen and carbonyl oxygen atoms when the peptide is in the PPII conformation, preventing some level of hydrogen bonding with solvent, and disfavoring this structure. Residues with little or no side chain (alanine and glycine), and residues with long flexible side chains (glutamine, methionine, and leucine) are less prone to occlude backbone from solvent, allowing for multiple backbone-to-solvent hydrogen bonds in this conformation. Other interactions can modulate the propensities somewhat, such as potential side-chain-to-backbone hydrogen bonds in the case of glutamine, but the underlying determinant is thought to be interactions between the backbone and solvent. Such reasoning applies readily to the lysine residues in K7. Side-chain electrostatic interactions, and perhaps side-chain-to-backbone hydrogen bonds, contribute to the overall level of PPII helical structure observed (see Figs. 2 and 3 ▶ ▶); however, it is backbone interactions with solvent that drive formation of this structure. The PPII conformation is very extended, with the peptide units being well exposed to solvent. The amide hydrogens and carbonyl oxygens point perpendicularly out from the helical axis, readily allowing for interactions with solvent molecules (Creamer 1998; Sreerama and Woody 1999; Stapley and Creamer 1999; Kelly et al. 2001).

Reconciling this hypothesis with the conformational behavior of homopolymers of lysine, and even with that of the alanine-rich peptides commonly used to study α-helix formation, is relatively straightforward. The seven-residue peptide used here does not form an α-helix at high pH, unlike long polymers of lysine, simply because it cannot form enough intrapeptide hydrogen bonds to stabilize such a structure. The peptides commonly used to study α-helix formation tend to be of the order of 15 or more residues in length (for examples, see Lyu et al. 1990; Park et al. 1993; Chakrabartty et al. 1994), and the homopolymers of lysine that have been studied are tens or even hundreds of residues in length (Tiffany and Krimm 1968b, 1972; Drake et al. 1988; Shibata et al. 1992; Makarov et al. 1994; Arunkumar et al. 1997). These all form relatively stable helices in solution (lysine polymers at high pH) but tend to have frayed ends. The seven-residue peptide used here is essentially all ends and simply cannot form a stable α-helix and must satisfy its hydrogen bonding requirements in another manner. Because sufficient intrapeptide hydrogen bonds cannot be made, the only other option is to hydrogen-bond extensively with solvent. The PPII helical conformation is an energetically favorable option that results in all backbone polar groups being well solvated and consequently allows for the satisfaction of all potential hydrogen bonds.

That the backbone of a polypeptide has some propensity to adopt the PPII helical conformation is not a new idea. Tiffany and Krimm (1968b) suggested that protein unfolded states consist of short stretches of PPII helix interspersed with bends. This was based on their observation of extensive PPII structure in homopolymers of lysine and glutamic acid and the similarities between the CD spectra of these homopolymers, poly(proline), and of unfolded proteins. Several peptides with little or no proline content also have been shown to have significant PPII helical character (Woody 1992; Toumadje and Johnson 1995). Most recently, Kelly et al. (2001) have shown that each residue possesses its own propensity to adopt the PPII helical conformation. Finally, Pappu et al. (2000) have shown using a short alanine peptide model that conformations corresponding to the general PPII region of (φ,ξ) space are well populated even when just steric interactions are taken into account. Interactions between the backbone and solvent will make this region more favorable.

Evidence that the PPII helical conformation is an important component of protein unfolded states has been steadily accumulating for more than 3 decades now. Coupling this evidence with recent experiments showing stable structure in proteins under highly denaturing conditions (Shortle and Ackerman 2001), it is becoming clear that the widely held belief that unfolded proteins occupy an essentially uncountable number of conformations is wrong. The ensemble of conformations is likely to be far smaller than previously thought and is also likely to possess persistent structural elements. These findings will simplify the task of modeling protein unfolded states and will lead to a more rapid understanding of the pathways by which proteins fold, because it is now becoming possible to understand where protein folding starts.

Materials and methods

Peptides of sequence Ac-(Lys)₇-Gly-Tyr-NH₂ (K7) and Ac-(Pro)₇-Gly-Tyr-NH₂ (P7) were purchased from PeptidoGenic Research and Co. and purified via reverse-phase high pressure liquid chromatography. The identities of the peptides were confirmed using mass spectrometry. Peptides were dissolved in buffer solution containing 5 mM potassium phosphate, 5 mM sodium fluoride, and 0.02% sodium azide, with the pH adjusted to 7.0. Peptide concentration was determined using the method of Brandts and Kaplan (1973). Absorbance was measured using a 1.0-cm pathlength cell in a Hitachi U-2000 spectrophotometer. CD spectra were measured using a Jasco J-710 spectropolarimeter using a 0.1-mm pathlength quartz cuvette. Measurements were conducted at 20°C, 0.5 nm resolution, and a scan rate of 100 nm min⁻¹. Reported spectra are averages of 20 or 30 scans with no smoothing. Spectra were measured with peptide concentrations of ∼100–200 μM and are corrected for solvent/buffer contributions. Errors in molar ellipticities are estimated to be approximately ±3%.