β-Sheet ¹³C Structuring Shifts Appear only at the H-bonded Sites of Hairpins

Abstract

The ¹³C chemical shifts measured for designed β hairpins indicate that the structuring shifts (upfield for Cα and C′, downfield for Cβ) previously reported as diagnostic for β structuring in protein appear only at the H-bonded strand residues. The resulting periodicity of structuring shift magnitudes is not, however, a consequence of H-bonding status; rather, it reflects a previously unrecognized alternation in the backbone torsion angles of β strands. This feature of hairpins is also likely to be present in proteins. The study provides reference values for the expectation shifts for ¹³C sites in β structures that should prove useful in the characterization of the folding equilibria of β sheet models.

Over the last fifteen years, β hairpins have been developed as models of β-sheet structures in proteins.^1-5 Until very recently,^6-9 the fold stabilities of designed hairpins were marginal: the fold populations in water at ambient temperature rarely exceeded 80%. With hairpin dynamics in the 1 – 50 μs range,^10,11 which leads to population-weighted chemical shifts, chemical shift deviations (CSDs) from random coil values have emerged as the primary tool for determining hairpin fold populations. Since 1999, we have advocated the use of the larger CSDs (circa 1 ppm) of the cross-strand directed Hα sites in non-H-bonded strand sites and those of the H_Ns of the H-bonded sites for estimating hairpin fold populations.¹² Detailed studies of hairpins and correlations with shifts in the edge strands of protein β sheets have established that the CSDs of backbone NHs and Hα's do, indeed, display alternating magnitudes along β strands.^13,14 Although the ¹³C CSDs for residues in protein β sheets are even larger; there do not appear to have been any reports suggesting a similar differentiation of ¹³C CSDs in β structures. The present study of hairpins was undertaken to look for differences in structuring shifts for H-bonded (HB) versus non-H-bonded (NHB) sites in hairpins.

The directions of the structuring shifts associated with β structuring, upfield for ¹³C=O and ¹³Cα versus downfield for ¹³Cβ, have been known at least since 1991,^15,16 and these have been widely used to assign protein secondary structure. Computational studies suggest that these reflect predominantly the distinct φ / ψ torsion angles in helical versus β structures.^17,18 In 2008, Vila and Scheraga¹⁹ classified nuclei in proteins in order of the usefulness their CSDs in elucidating secondary structure; for β structure definition: the following sequence was given: ¹Hα > ¹³Cβ > ¹H_N ∼ ¹³Cα ∼ ¹³C=O ∼ ¹⁵N. Given the larger structuring shifts associated with ¹³C nuclei, and the general expectation that ring current shifts should be less significant for ¹³C CSDs,^20,21 this statistical observation is somewhat surprising and was another impetus for the studies reported herein. The present study indicates that the diagnostic ¹³C=O, Cα and ¹³Cβ CSDs for β-structuring are associated, almost exclusively, with the HB residues; and, in the case of ¹³C=O shifts, run counter the prior expectation based on both experimental studies of H-bonding effects²² and density functional theory (DFT) calculations¹⁸ at the B3LYP/6-31G** basis set level for β models.

The dominating downfield shifts of ¹³Cβ, and upfield shifts of ¹³Cα, at HB sites of designed hairpins are illustrated in Figure 1. We selected the well-studied MrH hairpin^3,12 series (KKLTV-IXGK-KITVSA analogs, for which proton CSD methods for assessing the extent of folding have been validated^14,23,24) to illustrate the trend in ¹³C CSDs along the strands. The X = Asn to D-Pro mutation increases hairpin fold stability and this is reflected by comparable magnitude increases in the CSDs for both the Cα and Cβ CSDs at the HB sites. An (S6W,K11W)-double-mutation results in additional fold stabilization and further increases in the CSD magnitudes (Figure 1).

Figure 1.

¹³C CSD values along the sequence of a series of hairpin analogs demonstrating a fold population associated increase in magnitude for the HB sites (L3, V5, S7, K10, I12 and V14). The asterisks indicate CSDs that, in part, reflect ring current effects associated with the added Trp residues: upfield for the Cα within a Trp residue in a β configuration, and (lower panel) upfield at Cβ of the edge Trp in an edge-to-face cross-strand indole/indole interaction geometry²⁴.

A recently reported series of hairpins, stabilized by a capping interaction at the hairpin termini,⁹ provides a confirmation of the HB site specificity of β structuring shifts: Ac-WITVT-IHGK-K I RVWTG-NH₂ displays Cα CSDs of −1.5±0.6 and Cβ CSDs of +2.8±0.5 ppm at the HB sites (underlined) with the adjacent NHB site Cα/Cβ CSDs not significantly different from zero (−0.2±0.5 ppm). Further details and examples appear in the Supporting Information. In follow up studies of additional hairpins, these trends continue to be observed; for example, in the current set of Cβ CSDs, corrected to 100%-folding as needed (162 Cβ sites, nearly equally distributed between HB and NHB) the mean and standard errors are: +2.64±0.93 (HB sites) versus +0.42±0.65 (NHB sites). As a result, we recommend using Cβ CSDs of HB sites as an alternative measure of hairpin formation. In this regard, we note that Santiveri et al.^21,25 have provided guidelines for using ¹³Cβ and ¹³Cα CSDs for hairpin fold stability estimation (based on +1.95 and −1.55 ppm shifts, respectively, for 100% folding with averaging over all non-terminal strand sites).

Backbone carbonyl shifts are more difficult to access, but the isotopic substitutions required²⁶ for such studies appeared to be worth it to extend our study to the ¹³C site that might be expected to display the largest H-bonding effect²⁷. Our initial expectation was that a distinct HB versus NHB shift difference would be found, with the cross-strand H-bonding interaction producing a deshielding effect which might nullify the ¹³C=O shielding associated with the φ / ψ torsion angles in β strands. Thus we expected the NHB positions to display the larger upfield shifts. A series of MrH analogs with two Val ¹³C=O labels incorporated into each hairpin, one on each β strand, selectively at cross-strand HB and NHB positions, were prepared (the complete structures and chemical shifts as well as the control peptides for calculating CSD values appear in the Supporting Information). In addition, ¹³C′-Val and -Ala were incorporated in β-capped hairpins at HB and NHB sites (see Supporting Information). The HB sites were further upfield and also the only ones displaying a CSD magnitude melt as the fold population decreases on warming. The mean and standard error for ¹³C′ at NHB sites in well-folded hairpins was 0.09 ± 0.34, not significantly different than zero.

The hairpin fold populations of MrH peptides can be altered by both mutations (insertion alanine in a strand position or altering the turn locus) and fluoroalcohol co-solvent addition. With the additional observations made available from these mutational and media studies as well as at higher temperatures, ¹³C′ CSD data spanning the full range of hairpin fold populations was available for species with a pair of ¹³C′-Val units at HB sites. The correlation between hairpin fold population (χ_f, derived from Hα CSDs as previously described^14,24), and the average ¹³C′ CSD appears in Figure 2.

Figure 2.

¹³C=O CSDs display a linear correlation with the extent of hairpin folding. The complete set of CSDs and fold estimates at 280K are tabulated in the Supporting Information.

The correlation between ¹³C′ CSDs for HB sites and the extent of folding is linear with no systematic deviations for possible solvent or temperature (280 – 340K) effects. There do appear to be small, but significant, differences in the 100%-folded CSD value depending on the specific placement of the ¹³C′-Val probes in the hairpin (see Supporting Information). It is, however, apparent that any of the HB sites examined can serve as a dependable probe of the extent of hairpin formation. There was no correlation between the extent of folding and the magnitude of the smaller CSDs observed for the NHB ¹³C=O sites.

The chemical shift data presented indicates that the C′, Cα and Cβ chemical shift diagnostics previously associated with being in β-structured regions of proteins appear exclusively at the HB sites of hairpins. These site specific resonances can be added to the cross-strand directed Hα and HN sites as probes of folding and melting in β sheet models, a field where downhill folding scenarios are under consideration as alternatives to two-state folding.²⁸ We have adopted CSD values of +3, −2.6, and −2.2 for Cβ, C′ and Cα, respectively, as default 100%-folded standards in our continuing effort to assess the extent of folding of β sheet models. These values apply only for the HB strand sites. While the smaller structuring shifts at Cα and Cβ of NHB sites also melt out on thermal unfolding they do not display diagnositc values; a more detailed analysis indicates that Coulombic and ring current effects are often the largest contributors to these CSDs. The greater effect of electrostatics on ¹³C shifts has been noted previously.²²

In the designed hairpins of the present study, the alternating magnitude of ¹³C CSDs along β strands is most distinct for ¹³Cβ and somewhat less dramatic for ¹³Cα. This observation appears to be consistent with Scheraga's finding¹⁹ that, of the ¹³C shifts, Cβ provides the best definition of β structuring. Preliminary surveys of protein ¹³C shifts (work in progress) suggest that the shift trends observed for hairpins also apply to antiparallel edge strands of protein β sheets.

The prior literature does not provide a rationale for the sign of the shielding differences which we observe along a β strand associated with another antiparallel β strand. Hydrogen bonding has been viewed²⁷ as a major contributor to the ¹³C′ chemical shift differentiation associated with secondary structure motifs. In the case of carbonyl carbons there is a large body of experimental data indicating that H-bonding produces downfield shifts, rather than the upfield increment we observed for the H-bonded sites in hairpins: in fact, downfield shifts associated with short H-bonds as large as +10 ppm have been observed.^22,29 An early rationale of the upfield location of β sheet C′ (versus helical C′ sites) took this effect of H-bonding into account; De Dios and Oldfield suggested²⁷ that the relatively upfield location of β sheet C′ do not reflect “direct φ,ψ effects, but rather are dominated by strong peptide carbonyl-HN H-bonding in the α-helices”. With regard to the downfield C′ shifts in helices, experimental data³⁰ indicates that at least half of the 3.5 ppm downfield shift for C′ sites at the central residues of peptide helices is associated with the H-bonded state of these sites. DFT calculations¹⁸ of shifts for helical models also support this conclusion: for an (Ala)₉ model restrained to helical φ/ψ values the central site was 2.2 ppm downfield of the non-H-bonded sites closer to the termini. When the same shift calculations were applied to monomeric, and H-bond-associated dimeric and trimeric models of a (Gly)₇ β-strand, a 0.9 ppm shift difference results for HB versus NHB C′ sites, with the HB sites downfield. The calculations also predict a 0.8 ppm H-bond-associated shift for Cα sites. Both of these calculated H-bonding effects are in the opposite direction of the shifts we observe in β hairpin peptides. The greater upfield shift of the H-bonded ¹³C=O sites in β hairpin strands reported herein can not be rationalized based on any of the reported studies.

In the case of Cα and Cβ structuring shifts, essentially all theoretical studies imply that the φ/ψ values of the residue are the primary determinants of shift deviations from random coil norms. Recent calculation³¹ suggest that β hairpin ¹³C structuring shifts may also reflect χ¹ differences. The compilation of amino acid pairing preferences of Hutchinson and co-workers (http://www.rubic.rdg.ac.uk/betapairprefsparallel/) provides access to χ¹ values at NHB versus HB sites in antiparallel strands of proteins; while there are some χ¹ preferences for certain cross-strand amino acid pairings, there are no differences in the overall χ¹ preferences (g+ > g– > t) for HB versus NHB sites. To our knowledge, there have been no reports noting φ/ψ differences for NH and NHB sites in antiparallel protein β sheets. As a result, we turned to high resolution NMR structures of β hairpins to look for possible correlations: we found eleven structures, 4 from our prior studies^9,24,32 and 7 from the literature³³. The g+ conformation is also preferred at both the NHB (61%) and HB (68%) sites of these hairpins. The relative amounts of the minor t and g– χ¹ conformations is different, but this is judged unlikely to be the source of the ¹³C shift trends observed. A rationale was, however, immediately apparent from the Ramachandran plot for the strand residues of the β hairpin structures (Figure 3).

Figure 3.

A Ramachandran plot for the strand residues of eleven hairpins demonstrates a differentiation in φ/ψ values: with the H-bonded sites close to the center of the β structure basin and the non-HB sites in the polyproline II basin³⁴.

Based on the ¹³C shift dependence of A2 and A3 (in GAAAG) on the φ/ψ values of A3 presented by Case¹⁸, both the downfield shift of Cβ and the upfield shift of Cα are maximal at φ ≈ −130°, with a very rapid decrease in the CSD magnitude predicted as φ moves toward −80°. The calculated shift changes are in near perfect agreement with the CSD magnitude changes we observe for HB versus NHB sites in hairpins. In the case of ¹³C=O, the effect observed can also be rationalized as the result of the φ/ψ variation seen in Figure 3. The ¹³C′ shielding increases associated with the φ/ψ changes at the NHB sites, combined with a ≥1 ppm deshielding at HB sites due to the less negative φ value at the following NHB site, account for the CSD difference observed. We conclude that φ/ψ torsion angles are indeed, the primary basis of variations in the structuring shifts of all three ¹³C sites examined.

With the present findings taken into account, ¹³C CSDs should prove to be excellent probes of hairpin fold stability and unfolding pathways²⁸ and can provide details of β sheet geometry changes in proteins. We expect that ¹³C CSDs will provide more dependable measures of structure than ¹H CSDs. In the case of hairpin folding studies, we suggest that only the ¹³C structuring shifts of the HB sites should be used for fold assessment.