Peptide Conformational Determination With Vibrational Circular Dichroism and Isotope Labeling: Focused Review of 3₁₀‐Helical Peptide Spectroscopic Structural Studies

ABSTRACT

Optical spectroscopic techniques have long been used to determine average secondary structure in proteins and peptides, but these are primarily limited to determination of relative sheet and helix content, with the remainder often designated as “other.” Several decades ago, in cooperation with Claudio Toniolo and co‐workers, our group at UIC was able to show that vibrational circular dichroism (VCD) could be used to distinguish between formation of 3₁₀‐helices and α‐helices in small peptides, as well as from other uniform structures. Conventional electronic CD (ECD) or infrared (IR) spectroscopy did not give unambiguous distinctions between these two right‐handed, internally H‐bonded, helical forms. Theoretical modeling of the VCD spectra showed that the differences between helical forms could not be attributed to just use of Aib residues or of organic solvents, both used to stabilize the 3₁₀ conformation. Later we showed that, with IR and VCD enhanced by isotopic labeling, it was possible to differentiate site‐specific structures using vibrational spectra interpreted with DFT‐level theoretical modeling. These advances are briefly reviewed here and results of applying them to 3₁₀‐helical structures are discussed.

Distinctive amide I and II VCD band shapes computed for a 3₁₀‐helical octapeptide (red unlabeled, green, labeled with ¹³C=O on 4,5 positions) showing distinctive weak amide I couplet and strong negative amide II for the unlabeled peptide, and an additional couplet side band for the labeled one.

1. Introduction

Peptides form fundamental building blocks of proteins and have long been used as models to establish spectroscopic signatures that correlate to secondary structures. These correlations have often been used in turn to gain insight into protein structure. Optical activity methods, and particularly electronic circular dichroism (ECD) band shapes for amide n‐π* and π‐π* transitions, have had an important role in identifying chiral aspects of biopolymers, particularly in terms of average secondary structure [1, 2]. While amino acids have chiral configurations, in general, that aspect does not need to be determined for peptides, since in most cases their configuration is known. However, conformation and folding of the peptide chain are very important for protein function, and these structural aspects give rise to unique spectroscopic responses. Coupling of locally achiral components, such as the virtually planar peptide linkages, in the polymeric sequence leads to a distinct chiral response when the chain conformations are restricted to one sense of twist. This gives rise to characteristic ECD band shapes, which are more distinct for helical components, but their low‐resolution band profiles limit interpretations to assessment of average structural contributions. Vibrational CD (VCD), measured in the infrared (IR), provides many resolved transitions for structural studies, which can lead to enhanced spectra‐structure correlations for peptides and other biopolymers [3, 4, 5]. VCD resolution is still insufficient for determining site‐specific conformational insight, but by incorporating selected isotopic substitutions and combining VCD and IR measurements with theoretical modeling, site‐ and regio‐specific conformations can be determined [6, 7, 8, 9, 10, 11, 12, 13].

For proteins and peptides, α‐helices have broad, overlapped, characteristic ECD couplet band shapes, with negative lobes at ~222 and ~207 nm, and an intense positive feature below 200 nm. While β‐sheets contribute weaker spectrally shifted couplets that essentially overlap the helical pattern, the frequencies are different and, if their vacuum UV ECD is measured, a distinctive pattern can be obtained [14]. Disordered structures have a distinctly different ECD spectral band shape with roughly the opposite sign variations. When chiral 3₁₀‐helices were prepared and shown to be stable under certain conditions, their ECD had bands at about the same wavelengths as for α‐helices, but with a somewhat different intensity pattern that at times has been promoted as a means of distinguishing them [15]. Other workers felt the difference in the ECD patterns was too small to distinguish between these two related right‐handed helical conformers [16]. This confusion may arise from isolated α‐helices being stable for only relatively long segment lengths, particularly in aqueous solution, while 3₁₀‐helical formation is favored for short segments, which are often stabilized by incorporation of α‐alkylated residues. Furthermore, there is a stability issue in that short α‐helices can convert to 3₁₀‐helices, and longer ones often terminate with 3₁₀‐helical turns, which allows formation of an extra H‐bond [17, 18, 19, 20]. Consequently, deconvolving such spectral contributions is even more difficult. Beyond this, the sequences for which 3₁₀‐helices were stable and could be studied in isolation tended to require non‐polar solvents [21]. In general, ECD interpretations based on band shapes and sign pattern analyses work well if the peptide or protein has a dominant secondary structure type, but for mixed structures, spectrally overlapped transitions limit the accuracy of structural component determinations. While isolated 3₁₀‐helical models often give ECD patterns that fit the weak 222 nm expectation, when very similar structures are mixed in a sequence, deconvolving them with ECD is very challenging.

Infrared (IR) spectra have also been used to try to assign 3₁₀‐helical components in peptides by suggesting that higher frequency modes, such as might be seen for turns, indicate 3₁₀‐helical formation [22]. However, spectral frequency characteristics are quite sensitive to solvation and length effects, leading to misinterpretations, as well as being affected by α‐Me substitutions commonly used to stabilize 3₁₀‐helices. Nonetheless, 2D‐IR cross‐peak patterns, which depend on structurally specific coupling of amide groups in the peptide, are different for α‐ and 3₁₀‐helical structures, and have additionally indicated that the normal linear IR results were inconclusive in a model study [23].

In IR and VCD spectra, the amide I (~1650 cm⁻¹, C=O stretch) and amide II and III (~1550 cm⁻¹ and ~1250 cm⁻¹, both composed of N‐H and C_α‐H deformations plus C‐N stretch but coupled with opposite phases) modes are resolved and give characteristic responses for α‐helices, β‐sheets, and disordered conformations [3, 4, 12]. The amide I is the more structurally useful mode, providing an intense positive VCD couplet (positive then negative to higher frequency) for α‐helices and a weak, oppositely (but mostly negative) signed pattern for β‐sheets. The amide II is characteristically negative and shifted away from the IR maximum for helices, but has a couplet shape for sheets. The amide III modes for both are very weak, but they do have different patterns for helices and sheets. In contrast to the helix and sheet patterns, disordered structures (historically called “random coils”) give a somewhat broadened amide I negative couplet. This band shape is virtually the same as for poly‐L‐proline II (PPII), a left‐handed, 3₂‐helix, but the PPII structure has more intense VCD [24, 25]. The difference is due to PPII having a more extended and uniform structure, while the disordered peptides have only local left‐handed turn character but no long‐range order [6].

The Toniolo lab has promoted various methods for syntheses and structural characterizations of small peptides over several decades. One major aspect of their work was the preparation of modestly sized 3₁₀‐helical peptides, most often stabilized by incorporation of a large fraction of achiral amino‐isobutyric acid (Aib, or α,α‐dimethyl‐glycine) residues [21, 26, 27, 28, 29]. Other doubly C_α‐substituted residues have also been employed in their extensive studies [30, 31]. While Aib_n homopeptides are achiral, forming equal populations of left‐ and right‐handed helices, they have been used to establish structures (with X‐ray crystallography) as well as IR and other spectral responses for 3₁₀‐helices. By incorporating a few L‐amino acid residues in a modestly long Aib_n chain, chiral variants were prepared in which the right‐handed helical form was stabilized and for which ECD and VCD could be measured.

2. VCD of 3₁₀ Helices

The canonical α‐helical VCD bandshape, strong positive couplet amide I and weaker, broad negative amide II offset to the low frequency side of the IR band, was first measured for poly‐benzyl‐L‐glutamate and since has been reproduced in a variety of peptides as well as having been shown to be the dominant spectral response in proteins with high helical content [3, 4, 9, 32, 33, 34, 35, 36]. This shape occurs in both non‐aqueous and aqueous solvents but is broadened in the latter. The H₂O bending mode interferes with amide I spectra for aqueous samples, but this can be avoided by instead dissolving them in D₂O, for which the band is denoted amide I′. However, in D₂O, the α‐helical amide I′ VCD gains an extra negative lobe at low frequency, while the amide II' is shifted to ~1450 cm⁻¹ and its VCD is weakened.

By contrast, peptides, such as Z‐Aib₅‐L‐Leu‐Aib₂‐OMe, were shown to be both 3₁₀‐helical and chiral, and their VCD patterns in CDCl₃ had distinct differences from the α‐helical one if both the amide I and II were compared (Figure 1) [37, 38]. These 3₁₀‐helical peptides gave a weak positive couplet amide I VCD, having the same sign but a lower magnitude than for α‐helices, and additionally showed a strong, sharp negative amide II, centered on the IR maximum. So in comparison, these two right‐handed helical conformations do have the same VCD sign patterns (positive couplet amide I and negative monosignate amide II), but their relative amide I and II intensities are opposite (weak‐strong for 3₁₀‐ and strong‐weak for α‐helices). They also have somewhat different band shapes and correlations to their respective IR spectra. While this variation in VCD patterns may be seen as analogous to the intensity ratios proposed for 3₁₀‐helical ECD to distinguish it from α‐helical ECD, the differences in their VCD are sharper and more distinct, as aided by the clear separation of the amide I and II bands. It can be noted that while VCD for Aib‐rich, 3₁₀‐helical models were often measured in CDCl₃ for solubility considerations, ECD could be measured in alcohols, more dilute, providing a better spectral range. Solvation has some impact on the spectra; however, it has been tested on some samples, and consistent but broader VCD has been obtained in TFE. Our initial observations noted that the 3₁₀‐helical amide I band was more symmetrical than the α‐helical one, which tended to have a negative bias. We now feel that this small asymmetry is probably not significant and may depend on the instrument, resolution, and helical stability. Importantly, these signs and relative magnitude patterns qualitatively persist for a number of peptides of modest lengths having variations in their L‐amino acid substitutions and C_α substitutions, which suggests that the α,α‐disubstitution is not a critical factor in the changes seen in the VCD band shape [36, 37, 38, 39, 40].

FIGURE 1.

Canonical 3₁₀‐ and α‐helical FT‐IR (left) and FT‐VCD (right) for moderate‐sized peptides in organic solvents, showing the effect of α‐methylated residues in changing the favored helical conformation from α‐helical to 3₁₀‐helical. Top (b,b’): 3₁₀‐helical Z‐Aib₅‐L‐Leu‐Aib₂‐OMe; bottom (a,a’): α‐helical (Met₂‐Leu)₆ in CDCl₃. Absorbance (on left) is normalized so the amide I has A = 1, VCD (on right) is scaled by the same factor, and its plot is expanded ×10⁵. Figure from ref. [36]: R. A. G.D. Silva, S. C. Yasui, J. Kubelka, F. Formaggio, M. Crisma, C. Toniolo, T. A. Keiderling, Biopolymers, 65, 299–243 (2002).

As these peptides were increased in length and the fraction of α‐methylated residues decreased, the equilibrium between α‐helical and 3₁₀‐helical conformations gradually shifted [36, 41, 42]. In a length and solvent dependence study, we were able to show a gradual shift in the band shape between these two forms using (Aib‐Ala)_n alternating sequence peptides for n = 2–6, or up to dodecamer lengths (Figure 2). While the band shape changes are not dramatic, the octapeptide and decapeptide, n = 4 and 5, in CDCl₃ have band shapes most like what is seen for Aib‐rich 3₁₀‐helical peptides (Figure 1). This profile broadens somewhat for n = 6, suggesting less uniformity. The amide I for n = 6 (dodecamer) is relatively more intense than for the shorter peptides, and the amide II seems to have shifted a bit to the high‐frequency side of the absorbance. These changes could be due to partial α‐helical formation (e.g., in the center residues) or may just be a length‐dependent or ratio of Aib to Ala residues effect, but the overall pattern in all suggests dominant 3₁₀‐helical structures. Using alternate Ala‐(Aib‐Ala)_n sequences, with a higher relative fraction of Ala residues and different N‐termination, the 3₁₀‐helical pattern is dominant for all the longer lengths (including the undecamer). However, with more Ala residues, the sharpness and intensity of the amide II bands are less pronounced than seen in Figure 1, presumably due to partial loss of 3₁₀‐stability introduced by the Aib residues. With TFE as a solvent, the spectra are broader yet, suggesting less uniform helical formation, but the amide II still has a stronger negative VCD than the amide I couplet. Both series show ECD shape changes for the longest members, with an increase in the relative intensity of the 222 nm component. Contrasting these ECD results to the relatively small VCD pattern changes implies that using the ECD shape to determine 3₁₀‐helical structure for longer peptides may be problematic. Helical ECD is known to have a significant length dependence [31], while that of VCD, due to its weaker dipole coupling, is less so [43, 44]. It should be noted that ECD are measured at lower concentrations than VCD, which may also affect helical stability.

FIGURE 2.

FT‐IR and FT‐VCD length dependence study of a series of alternating, Z‐(Aib‐Ala)_n‐OMe, peptides. Spectral changes with increasing lengths show a gradual change from a disordered structure for the tetrapeptide (n = 2), to a dominant 3₁₀‐helical form for the octa‐ and deca‐peptides (n = 4,5), and what may be a mixed but mostly 3₁₀‐helical conformation for n = 6 (dodecapeptide). Amide I IR were normalized to A = 1, and that factor was used to scale the respective VCD intensities and show the changes of the amide I with respect to the amide II. Figure from ref. [36]: R. A. G.D. Silva, S. C. Yasui, J. Kubelka, F. Formaggio, M. Crisma, C. Toniolo, T. A. Keiderling, Biopolymers, 65, 299–243 (2002).

We used quantum mechanical ab initio‐based, harmonic VCD simulations to model the spectra for idealized α‐helical and pure 3₁₀‐helical forms to probe the source of their different spectral profiles [41, 45]. These earlier calculations simulated tripeptide and octapeptide spectra using only Ala‐based sequences, which both gave results qualitatively consistent with experiment for the α‐helix conformer. However, for the 3₁₀‐helix, only the octapeptide computed VCD resembled the experimental pattern, suggesting a significant end effect distortion. By calculating the whole decapeptide and optimizing its geometry in vacuum, the α‐helical structure reverted to a 3₁₀‐helical structure, but the α‐helix could be stabilized using a simulated solvent. These results are consistent with short 3₁₀‐helices being more stable, while length and solvation can favor the α‐helical form.

To test the impact of variations in sequence, we redid the calculations for idealized dodecapetides [18, 42]. These simulations were done at a higher DFT‐based level, first for smaller peptides, Ac‐Ala₄‐NH₃, Ac‐(Aib‐Ala)₂‐NH₃, and Ac‐Aib₄‐NH₃, constrained to the 3₁₀‐helical structure, and Ac‐Ala₆‐NH₃, Ac‐(Aib‐Ala)₃‐NH₃ for the α‐helical structure, so that, in both cases, the center amide is fully H‐bonded forward and back in the sequence. The computed VCD for the five‐amide‐containing 3₁₀‐helical tetrapeptide models did show a change in amide I band shape with increasing Ala composition. But there was a negligible effect of Aib content on the seven amide α‐helical hexapeptide spectra.

These results were then transferred [46] onto longer sequences, Ac‐Ala₁₂‐NH₃, Ac‐(Aib‐Ala)₆‐NH₃, for the α‐helical structures and Ac‐Ala₈‐NH₃, Ac‐(Aib‐Ala)₄‐NH₃, and Ac‐Aib₈‐NH₃, for the 3₁₀‐helical ones, with both the shorter and longer sequences constrained to the same ideal conformations [42]. The transferred force field and computed APT and AAT intensity‐component tensors were then used to compute the IR and VCD spectra for comparison to the experimental spectra, as indicated in Figure 3. [Calculations were actually done for various structures, and the computed 3₁₀‐helical dodecamer and octamer spectra were virtually identical, and in fact did not change for lengths as long as 21 residues, but those in Figure 3 are representative choices for comparison to the experimental spectra shown. The frequencies are not corrected for the standard errors found with vacuum DFT calculations of amides.] The α‐helical results show very good band shape and intensity agreement with experiment, with a strong positive couplet amide I and weak amide II offset to lower frequency from the absorbance, and evidence of virtually no effect of Aib substitution. However, for the 3₁₀‐helical structure, there is a change in the amide I shape and intensity with an increase in Aib content, leading to a broadening of the IR (not shown), which in turn results in a weaker VCD. The sequence dependence between Ac‐(Aib‐Ala)₄‐NH₃ and Ac‐Aib₈‐NH₃ seen experimentally for the amide I VCD is reflected in these theoretical model calculations and presumably originates from the lack of a C_αH mode contribution for those peptide linkages derived from the Aib residues. While the Ac‐Ala₁₂‐NH₃ 3₁₀‐helical simulation shows a larger amide I, its pattern is still distinct from that of the Ac‐Ala₁₂‐NH₃ α‐helical VCD, reinforcing that the computed α‐helical VCD agrees much better with the α‐helical experimental VCD and does not fit the 3₁₀‐helical experimental results, no matter the sequence. These comparisons provide added demonstration that the VCD discrimination between the two helical types is not dependent on the nature of the residues, nor even the peptide lengths. It can be noted that, to stabilize the α‐helical structure experimentally, longer peptides are generally needed, and that computed spectra for longer α‐helical peptides yield sharper and more intense amide I IR and VCD, but changes beyond the dodecamer are minimal [7, 41, 42, 47, 48].

FIGURE 3.

Comparison of experimental VCD spectra (center) for 3₁₀‐helical peptides [Z‐Aib₅‐Leu‐Aib₂‐OMe, Z‐(Aib‐Ala)₄‐OMe] and α‐helical (Met₂‐Leu)₆ in CDCl₃ (red) and TFE (black) with DFT‐based simulated spectra of dodecamer α‐helices (left) and octamer 3₁₀‐helices (right), where the different computed lengths were chosen for comparison with the experimental data in the center panel. The vertical lines (black, left, and blue, right) indicate relative rotational strengths, R, for individual modes, and the black band shapes represent their summed contributions to the spectrum. Figure partially recreated from data in ref. [42]: J. Kubelka, R. A. G.D. Silva, T. A. Keiderling, J. Am. Chem. Soc. 2002, 124, 5325–5332.

3. Isotopic Substitution and Site‐Specific Interaction

In the past couple of decades, many other groups and we have incorporated isotopically labeled residues into specific sites in synthesized peptides to determine local structure and changes in folding‐unfolding processes [4, 12, 49]. Placing ¹³C=O substituted residues at a specific point in the sequence leads to an amide I sideband shifted down by ~40 cm⁻¹ in frequency, which is usually detectable in the IR spectrum. However, the shift alone does not identify the structure or local conformation. By placing multiple labeled residues in the sequence, their coupling can provide structural insight. For β‐sheet and hairpin structures, this is best done using IR frequency and intensity patterns to probe cross‐strand coupling, which can be used to detect hairpin folding and dynamics [11, 13, 50, 51, 52, 53, 54]. Additionally, in aggregate or fibril formation, such labeling can distinguish parallel vs. anti‐parallel forms [10, 55, 56, 57, 58, 59, 60]. However, VCD is less useful for sheet structures, although there are a few hairpin studies [61, 62]. In a different vein, a series of fibril studies based on enhancement of VCD intensities have appeared [10, 55, 63, 64, 65, 66]. By contrast, isotopic labeling of helical peptides has proven quite useful for determination of coupling between specific sites [6, 7, 8], determination of local conformation and monitoring of folding [9], and distinguishing local secondary structure types [6, 10, 67, 68].

To extend our joint studies characterizing 3₁₀‐helices, the Toniolo group prepared several selectively isotopically labeled Aib‐containing peptides, substituted with ¹³C=O on the amide linkage, that were motivated as a test of the coupling interactions between residues that explicitly give rise to the 3₁₀‐helical VCD band shape [8]. For 3₁₀‐helical hexapeptides with ¹³C=O substitutions incorporated on the 4 or 5 positions (A4H, iPrCO‐Aib‐Ala‐Aib‐Ala*‐Ala‐Aib‐NHiPr and A5H, iPrCO‐Aib‐Ala‐Aib‐Ala‐Ala*‐Aib‐NHiPr, respectively), the IR and Raman amide I spectra showed broadening to low frequency as compared with the unlabeled hexapeptide, UH. The A4H and A5H VCDs were a bit weaker than for UH, but their shapes were virtually identical, indicating that VCD is not very useful for detecting helical conformations with only one labeled residue. On the other hand, the VCD and IR spectral patterns for the hexapeptide with two sequentially labeled residues (A4A5H, iPrCO‐Aib‐Ala‐Aib‐Ala*‐Ala*‐Aib‐NHiPr) and a similarly labeled octapeptide (A4A5O, Ac‐Aib‐Ala‐Aib‐Ala*‐Ala*‐Aib‐Aib‐Aib‐OMe) had more intense IR and Raman sidebands. These corresponded to easily detected VCD couplets originating from the ¹³C=O labels and were accompanied by significant changes to the ¹²C=O VCD band shape (Figure 4). The dramatic change in VCD from single‐ to double‐label substitution indicates the strong VCD dependence on the coupling of similar dipole oscillators. This coupling is conformationally sensitive, which makes VCD of the labeled residues particularly useful.

FIGURE 4.

IR (bottom) and VCD (top) for the amide I and II modes of sequentially isotopic substituted peptides on the 4 and 5 positions (indicated by *) for the octapeptide [A4A5O, Ac‐Aib‐Ala‐Aib‐Ala*‐Ala*‐Aib‐Aib‐Aib‐OMe, red dash‐dot] and hexapeptide [A4A5H, iPrCO‐Aib‐Ala‐Aib‐Ala*‐Ala*‐Aib‐NHiPr, blue dash] as compared with the unlabeled hexapeptide sequence [UH, black]. The spectra were measured in TFE:CDCl₃ (1:3) and normalized to a fixed amide I IR area; this factor was then used to scale the VCD. The improved S/N for the VCD is due to measurement on a newer dispersive VCD instrument with improved optics optimized for peptide studies. From ref. [8]: A. Lakhani, A. Roy, M. De Poli, M. Nakaema, F. Formaggio, C. Toniolo, T. A. Keiderling, J. Phys. Chem. B 2011, 115, 6252–6264.

While the isotope‐shifted VCD couplets indicated somewhat weaker couplings than we had previously determined to be characteristic of α‐helical peptides, they had the same sign pattern [7, 8, 9]. For sequential labeling, the ¹³C=O VCD amide I band shape in the octapeptide reflected that of the ¹²C=O modes, characterized by an IR sideband shifted down in frequency correlated to VCD with a positive couplet shape. That α‐ and 3₁₀‐helices have the same sign for sequential coupling is expected since their right‐handed helical structures both have the C=O groups oriented roughly along the axis direction (with those in the 3₁₀‐helix having more tilt). Additionally, this could be assumed from their amide I VCD having the same pattern for similar structures, since the VCD sign pattern reflects the sign of the coupling. While the labeling in the octapeptide is more central, in that the residues following are internally H‐bonded, the hexapeptide had the same sort of shape for the ¹³C=O VCD amide I band, but the IR frequencies were different, suggesting the short segments, in sequence before and after the labeled positions, may have altered coupling. There is also a minor frequency shift for the hexapeptide amide II mode, which is probably insignificant. The similarity of the VCD patterns for the octapeptide and hexapeptide in Figure 4 should not be overinterpreted since in the hexapeptide the A4 position is different from the A5, because the latter has no internal H‐bond and is subject to end effects. Similarly, the octapeptide has proportionally less ¹³C=O, so the intensity balance with the ¹²C=O band should differ from that of the hexapeptide, as it does in both the IR and VCD. The IR, Raman, and VCD did apparently have some frequency shifts between the isotope‐correlated bands after band shape fitting, but since these different vibrational spectra were measured on three different instruments with three different resolutions, we did not pursue correlations further. It might be noted that the improved S/N seen in these Figure 4 VCD spectra as compared with the previous figures in this review is a result of our using a newer dispersive VCD instrument optimized for peptide studies that was developed in the UIC lab during the time as these labeled 3₁₀‐helical peptide studies were carried out [62, 69].

We only had one alternately labeled hexapeptide sequence available, with an Aib between the two labeled Ala residues [A2A4H, iPrCO‐Aib‐Ala*‐Aib‐Ala*‐Ala‐Aib‐NHiPr]. The A2A4H peptide isotope band apparently had the opposite sign pattern, with what appeared to be a weaker negative VCD couplet, which would have been consistent with previous α‐helical results [7]. However, due to overlap with the oppositely signed ¹²C=O modes and probably added terminal end effects, this VCD was more distorted and difficult to resolve and could only be distinguished by comparison to the IR sideband. This sign change pattern is consistent with what we have seen in α‐helical and 3₁‐helical (PPII) peptides, even though it is dependent on detailed band shape analyses and comparison to theoretical modeling [6, 7].

Our initial DFT‐based theoretical modeling of the isotopically substituted peptides used idealized structures and showed the apparent coupling constants for these 3₁₀‐helical peptides to be more dependent on sequence than is evident from the close match of the octa‐ and hexa‐peptide patterns in Figure 4. The octapeptide had a weaker coupling than seen for α‐helices, with the negative VCD component, in‐phase to higher frequency, being more intense in the IR than the weaker positive VCD, out‐of‐phase mode (Figure 5). By contrast, the hexapeptide appears to have a much larger splitting of the labeled positions, but this is not due to coupling, as can be seen from the near‐identical dipole strengths computed for the two transitions. In fact, this splitting arises from end effects so that the lower frequency mode is almost all Ala4 and the next one Ala5. Since the Ala5 C=O has no internal H‐bond, its intrinsic frequency is higher (as we saw in comparison of the A4H and A5H single‐label hexapeptides), and this non‐degeneracy will decrease the effect of coupling. Thus, the close agreement of the A4A5H and A4A5O experimental results in Figure 4 is at least partly accidental, in that while the sign pattern does reflect the helical nature of the hexapeptide residues, it does not reflect their coupling strength. This illustrates that assuming observed couplets reflect inter‐residue coupling can be too simplified, and other sources of splitting must be factored out to get at the true coupling. This comparison emphasizes the need to view observed spectral shapes critically. For detailed band shape interpretations that are possible with isotopic substitution, it is necessary to correlate the results to computed mode frequencies and behaviors, if coupling between sites is to be determined.

FIGURE 5.

Computed IR and VCD amide I spectra for the A4A5H and A4A5O isotopically substituted 3₁₀‐helical hexa‐ and octapeptides, respectively. Calculations used an idealized 3₁₀‐helix constrained to (ϕ,ψ) values of (−60°,‐30°) with all other geometrical coordinates optimized at the DFT/BPW91/6‐31G*/PCM level using Gaussian 03 [70]. Band shapes were simulated by summing Gaussians assigned to each computed frequency and intensity as indicated by the vertical lines. Dashed red lines are fits of two Gaussian band shapes to the absorbance spectra. The high‐frequency mode at ~1755 cm⁻¹ arises from the ester termination of the octapeptide. The lowest hexapeptide mode computed at ~1653 cm⁻¹ correlates to Ala4, and the next at ~1676 cm⁻¹ is Ala5, which has no H‐bond to another residue. The splitting is due to inequivalence rather than coupling. Note that DFT‐computed amide I frequencies are always high and dependent on basis set, functional, and solvation modeling. From ref. [8]: A. Lakhani, A. Roy, M. De Poli, M. Nakaema, F. Formaggio, C. Toniolo, T. A. Keiderling, J. Phys. Chem. B 2011, 115, 6252–6264.

The couplings for the sequential and alternate labeled peptides are opposite in sign, so that for A2A4H the lower frequency negative VCD component is more intense in the IR than the higher positive VCD one. Our simulations showed the qualitative pattern seen experimentally for A2A4H, but the agreement was not as good as for Figures 4 and 5 above due to overlap of the ¹³C=O bands with the broadened ¹²C=O bands.

To seek better agreement between experimental and computed spectra for these isotopically labeled 3₁₀‐helical peptides, we [J.R. Cheeseman, unpublished results] have undertaken a series of higher‐level DFT calculations with full minimization of the exact sequences studied experimentally and inclusion of solvent effects using a PCM correction. Unlike what we found for longer isotopically substituted α‐helical models [48], these calculations have not yet provided significant improvement over our earlier idealized structure computations, which now appears to be due to the limitations of the peptide models we had available, rather than the theory. However, the revised octapeptide calculations are in better agreement with the experimental VCD for the UO (unlabeled octapeptide) and A4A5O sequences, as can be seen in Figure 6. The spectral differences resulting from full minimization, PCM modeling of the solvation with CHCl₃, and a larger basis set [6–311 + G(2d,p)] using a different functional (B3LYP) developed lowered frequencies (seemingly due to the basis set and optimizing the geometries) and a more distinct ¹²C=O band shape than seen in the previous calculations (Figure 5b) but show the same coupling (~7 cm⁻¹) between the Ala4 and Ala5 labeled residues.

FIGURE 6.

Higher‐level simulation of the amide I IR and VCD for the octapeptide, both unlabeled (UO, red) and labeled on the Ala4 and Ala5 positions (A4A5O, green), using DFT minimization of the full sequence Ac‐Aib‐Ala‐Aib‐Ala*‐Ala*‐Aib‐Aib‐Aib‐OMe at the B3LYP/6–311+G(2d,p) level with a PCM (CHCl₃) solvent correction. Vertical lines indicate the positions and intensities of individual modes. Comparison with Figure 5 shows the same coupling of the labeled residues, with only small band shape and intensity variations, but the modes are at lower frequencies due to geometry optimization as well as functional and basis set differences. The high‐frequency mode at ~1748 cm⁻¹ is due to the ester linkage of the C‐terminal group. Data from J.R. Cheeseman, unpublished results, with permission.

While these DFT‐optimized VCD band shape profiles (Figure 6) only differ incrementally from the octapeptide intensity patterns in Figures 4 and 5, the revised hexapeptide results (not shown) only qualitatively agree with experiment, presumably due to the lack of a H‐bond for A5. Tests with ideal structures gave qualitatively the same spectral predictions but had more distortions of the ¹²C=O mode band shapes. The difference in the computations is probably a factor of using optimized geometries, as well as inclusion of the terminal groups, both of which can affect specific modes in the hexapeptide more than in the octapeptide. Calculational tests showed the frequencies for singly labeled A4 and A5 in the octapeptide differ by ~2 cm⁻¹, but in the hexapeptide by ~20 cm⁻¹, so the computed separation in the A4A5H peptide is not due to coupling and is further affected by the iPrCO termination, which reduces the separation by shielding the terminal amides from solvation.

Even though octapeptides with alternate isotopic substitution patterns were not experimentally available to us, we tested them computationally. Consistent with what is seen in other helical models, alternately labeled peptides, such as A3A5O, had weaker coupling (4 cm⁻¹) and the opposite sign pattern for the labeled residues, with the more intense IR mode (negative VCD) lower in frequency than the positive (and also more intense) VCD component. These patterns of sequential and alternate labeling, with oppositely signed VCD patterns, are what we have seen in α‐helices and 3₁‐helices as well, and reflect a reversal in chirality, changing from a positive sense of rotation to negative down the helical chain for the inter‐residue coupling as they become more separated.

4. Summary

The 3₁₀‐helical conformation is the third most common structure in proteins and has important roles in transitions between helical segments and other parts of the sequence as well as providing important intermediate folding minima. The termini of α‐helical segments in proteins often have 3₁₀‐helical conformations, or distorted type I β‐turns. As these helices unfold, it is reasonable that 3₁₀‐helical intermediates will occur. Similarly, short helical segments in proteins can gain an extra H‐bond for stability by forming a 3₁₀‐helix. Also, these conformations occur in certain peptide antibiotics (e.g., peptaibols). The similarity to α‐helices has long made detection of 3₁₀‐helices a challenge, and several methods have been proposed and used. The Toniolo lab has made great progress over the decades in creating peptide models for 3₁₀‐helices and in determining their physical properties and structures. We have tried to help this effort by adding a VCD method for the determination of 3₁₀‐helical structure that has advantages for isolated peptides. This review has critically evaluated the sources of this VCD method that arose from UIC‐Padova collaboration, one that has been deeply rewarding, certainly on our part. This peptide application would have never come about without Claudio Toniolo's suggestions for the initial studies, as well as his insatiable curiosity and deep knowledge of the literature. Claudio is certainly a scientist and collaborator whom we all admire, and my friend.

Conflicts of Interest

The author declares no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.