Stephen J. Eyles is in the Department of Polymer Science & Engineering, University of Massachusetts, 120 Governors Drive, Amherst, Massachusetts 01003,USA.eyles@polysci.umass.edu
A new study suggests that isomerization of nonprolyl peptide bonds may also contribute to slow kinetic phases observed in protein folding.
What are the determinants of protein folding? This question has baffled scientists for the past several decades, but gradually general rules are emerging. One of the most widely accepted concepts is that cis-trans isomerization of peptide bonds immediately preceding proline residues (Xaa−Pro) can cause slow refolding phases1. Proteins generally fold on a time scale of microseconds to seconds. If a population of the folding polypeptide chains contains an incorrect Xaa−Pro isomer, however, then it may take several seconds for the protein to reorganize itself such that folding can continue to completion. Thus the appearance of slow folding kinetics has often been attributed to proline isomerization with little further investigation. A provocative article by Pappenberger et al.2 on page 452 of this issue of Nature Structural Biology forces us to reconsider this conclusion and to entertain the possibility that proline may not be the only culprit in slow kinetic phases. In fact there may be a contribution from isomerization of some or indeed all of the other peptide bonds in a folding protein.
Proline isomerization As an imino acid, proline is unique among the 20 naturally occurring building blocks of proteins. The tertiary nature of the amide nitrogen in proline means that the preceding peptide bond in a protein can adopt either the cis or trans conformation almost isoenergetically (Fig. 1). The activation energy barrier for cis-trans isomerization of Xaa−Pro peptide bonds has been well characterized, with values ranging from 80 to 100 kJ mol-1 (20−25 Kcal mol-1) in model peptides and proteins. This activation barrier in turn gives rise to kinetic refolding phases with time constants () of 10−20 s. For the amino acids the amide hydrogen provides significantly less steric bulk than the peptide backbone, thus distinctly favoring the trans conformation. Calculations3 and experiments4 suggest that a cis nonprolyl peptide bond can destabilize a native protein by up to 20 kJ mol-1 (5 Kcal mol-1). In unfolded polypeptides, nevertheless, the equilibrium constant may be such as to allow population of the cis conformation of nonprolyl peptides to 0.1%.
Figure 1. Time constraints and schematic free energies for cis-trans isomerization of peptide bonds.
a, For Xaa−Pro bonds the cis and trans forms are almost isoenergetic with an activation energy barrier in the range 80−100 kJ mol-1 (20−25 Kcal mol-1) yielding rates of 0.1 s-1. b, In the case of nonprolyl peptide bonds the trans form is more stable by 20 kJ mol-1 (5 Kcal mol-1). Pappenberger et al.2 calculate the rates of interconversion from their data to be 2 s-1 for cis-trans and 0.075 s-1 for trans-cis isomerization.
Tendamistat folding Tendamistat is a 74-residue -sheet protein that conforms well to a two-state folding mechanism. Of the unfolded population, 80% refolds to the native structure in a rapid reaction with a time constant of 10 ms. Interestingly, this was initially thought to occur without formation of a collapsed state5. More recent evidence suggests that this protein does in fact fold via a classic hydrophobically collapsed state but does not become populated to a detectable level6. The remaining 20% of the population follows a parallel pathway, which comprises two kinetic phases. These phases were initially assigned to isomerization at one or more of the three native proline residues. However, the new study by Pappenberger et al.2 shows that only the slowest of these phases (10 s) involves proline isomerization since it is abolished in a variant in which all three prolines are replaced2. Intriguingly, the medium ( = 400 ms) phase persists even in this proline-free tendamistat, with amplitude 5% of the total change in fluorescence. The task then was to determine what other factor could be affecting the folding of this small protein to give rise to parallel folding trajectories. Using interrupted refolding experiments, in which the protein is allowed to refold for varying times and then unfolded by rapid dilution into high concentration of denaturant, the authors determined that all of the unfolding kinetics originated from native protein — there was no evidence for kinetics arising from a transient intermediate state. In a double jump experiment the reverse process is monitored: protein is unfolded for varying times and then diluted to conditions that favor refolding. If the unfolding time is short relative to isomerization processes in the unfolded state, the refolding molecules would be in a homogeneous population, and the kinetic phases arising from these processes should be absent. Indeed this is the case for tendamistat; the = 400 ms folding phase only emerges after extended incubation under unfolding conditions which permit cis-trans equilibration processes in the denatured state.
Thermodynamic parameters for interconversion between the cis and trans isomers in the unfolded ensemble were extracted by measuring the temperature dependence of folding. Proline isomerization shows only a weak dependence, presumably because the cis and trans forms remain isoenergetic7. In contrast, the rate and amplitude of this intermediate folding phase increases linearly with temperature. The activation energy obtained corresponds well with that measured using model peptides9 strongly suggesting that this kinetic phase arises from cis-trans isomerization of peptide bonds that do not involve prolyl residues.
As a further test of this hypothesis, the authors refold from an alternative denaturant — anhydrous trifluoroethanol containing varying concentrations of lithium chloride. This solvent system strongly promotes formation of cis peptide bonds in model compounds: 50% or more for Xaa−Pro8, and here demonstrated up to 0.5% cis population for nonproline in the peptide Ala-Ala-Tyr. An approximately linear relationship was observed between the fraction of slow folding molecules and LiCl concentration, lending further support to the idea that incorrect peptide bond isomers might indeed be the source of heterogeneity in the unfolded protein. The authors thus conclude that a small population of the thermodynamically disfavored cis conformation of peptide bonds in the unfolded state is the cause of the small amplitude kinetic phase observed in tendamistat folding.
A general phenomenon? The question arises as to whether cis peptide bond isomers are a common feature in denatured states of proteins. If there is a similar fractional population at all peptide bonds then the amplitude of the observed refolding kinetic phase should increase linearly with the size of the protein. Most proteins that have been found to fold by a two-state mechanism are <100 residues in length, whereas larger proteins tend to exhibit more complex folding kinetics. Whether this is a result of a higher overall population of cis conformers remains to be seen. On the other hand, the rate of isomerization for nonprolyl peptide bonds seems to increase with peptide length9, so it may be that this effect becomes insignificant beyond a certain protein size.
The authors are unable to say at this time whether the 5% slow phase is caused by isomerization of a specific peptide bond, as in the case of RNase T1 (ref. 4), although they favor the idea that this is a generalized phenomenon with a certain level of occupancy of cis isomer at each peptide bond in a protein. Alternatively there may be a higher propensity for cis peptide bond formation depending on the specific amino acid pairing, as is the case for Xaa−Pro bonds. Only a detailed analysis of other proteins and model peptides would be able to answer this question. A third, perhaps more likely, possibility is that the influence of the cis conformer on folding depends on the structural context in the folding protein. As suggested by Scheraga10, placement of prolines throughout a protein sequence is a useful probe of early folding events, since only those positions that are critical for correct folding will influence the kinetics. Likewise naturally occurring prolyl residues can have very different influences on the folding properties depending on their sequence locale11.
Nonprolyl cis peptide bonds have been postulated as slow folding steps in other proteins. The folding of E. coli dihydrofolate reductase12 and staphylococcal nuclease13 may be complicated by isomerization of one or more nonprolyl peptide bonds. Odefey et al.4 showed that nonprolyl peptide bond isomerization gives rise to an extremely slow phase in refolding of the P39A mutant of ribonuclease T1, and in this case the cis conformation is adopted at the Tyr 38−Ala 39 bond in the native state. Furthermore, cis peptide bonds may be more common in native proteins than first suspected, as suggested by Weiss et al.14; their occurrence may be overlooked by the commonly used structure refinement algorithms.
In the cellular environment enzymes have evolved to deal with the problem of undesirable incorrect Xaa−Pro isomers. Peptidyl prolyl isomerases such as cyclophilin are thought to be specific to proline isomerization but as yet no enzyme has been identified that catalyzes isomerization of nonprolyl peptide bonds. Perhaps a function of the ubiquitous Hsp70 family of chaperones is to maintain peptides not only in an unfolded but also in an all-trans conformation before release to facilitate rapid and efficient folding.
Whether cis-trans isomerization of nonprolyl peptide bonds is a general phenomenon important in defining slow phases of protein folding remains to be seen. Certainly the interpretation of such small amplitude kinetic phases requires painstaking analysis. Care must also be taken that denaturants such as trifluoroethanol do not induce other structures in the unfolded ensemble that could influence the kinetic parameters or otherwise alter folding pathways. Nevertheless, the thought-provoking paper of Pappenberger et al.2 of reminds us there is much more to learn about protein folding.
) of 
10−20 s. For the amino acids the amide hydrogen provides significantly less steric bulk than the peptide backbone, thus distinctly favoring the trans conformation. Calculations
-sheet protein that conforms well to a two-state folding mechanism. Of the unfolded population, 80% refolds to the native structure in a rapid reaction with a time constant of 10 ms. Interestingly, this was initially thought to occur without formation of a collapsed state
