Tunable CH/π Interactions within a Tryptophan Zipper Motif to Stabilize the Fold of Long β-Hairpin Peptides

Abstract

The tryptophan zipper (Trpzip) is an iconic folding motif of β-hairpin peptides capitalizing on two pairs of cross-strand tryptophans, each stabilized by an aromatic-aromatic stacking in an edge-to-face (EtF) geometry. Yet, the origins and the contribution of this EtF packing to the unique Trpzip stability remain poorly understood. To address this question of structure–stability relationship, a library of Trpzip hairpins was developed by incorporating readily accessible non-proteinogenic tryptophans of varying electron density. We found that each EtF geometry was in fact stabilized by an intricate combination of XH/π interactions. By tuning up the π-electron density of Trp^face rings, CH/π interactions are strengthened to gain additional stability. On the contrary, our DFT calculations support the notion that Trp^edge modulations are challenging due to their simultaneous paradoxical engagement as H-bond donor in CH/π and acceptor in NH/π interactions respectively.

Graphical Abstract

Introduction.

Attractive non-covalent interactions (NCIs) play a pivotal role not only in the thermodynamic stabilization of supramolecular assemblies and protein–drug complexes^[1–4], but also in kinetic rate accelerations in organocatalysis.^[5–7] In nature, such forces are also responsible for the folding of proteins and participate in numerous protein–protein interaction mechanisms.^[8–10] In the context of protein folds, the side chains of aromatic amino acids (W, F, Y, H) participate in a plethora of π NCIs such as CH– π , π–π, and cation/anion–π interactions.^[11–13] Despite its scarcity in proteins (abundance ~1.5%), tryptophan (W) is a unique residue in the space of π-aromatic interactions, due to an innate hydrophobicity and a substantial π-electronic density offering a large van der Waals contact surface.^[14] Strikingly, tryptophans are far superior π-donor arenes than any other aromatic amino acids, therefore contributing to greater stabilization effects across all π NCIs.^[15–19] Given these features, W residues have attracted a great deal of attention for studying biological systems and miniaturized proteins such as the tryptophan cages.^[20,21] Tryptophans have also been incorporated in stand-alone β-hairpin peptides excised from their protein context.^[22,23] The seminal studies by Robinson and Cochran highlighted the exceptional stabilizing effect of cross strand interactions generated by W/W pairs positioned at non-hydrogen bonded (NHB) sites.^[24,25] These efforts were later championed by Cochran who reported that the intercalation of two W/W pairs at successive NHB positions could yield a hyperstable mimic of the GB1 N-terminal hairpin known as the tryptophan zipper or “Trpzip” scaffold (Figure 1A).^[26,27] Even if our understanding of the Trpzip paradigm has made significant advances in terms of kinetics^[28,29] and thermodynamic features^[30–33], the original folding performance of aryl–aryl W/W π–stacking has never been overthrown.^[31,34–35] Yet, to explore the potential of β-hairpins as an alternative chemical space to large macro- and bicyclic peptides, significant folding improvements are needed. To maximize folding and achieve a physiologically relevant meso- or thermostability, proteins and foldamers alike should display melting temperatures (T_m) superior to 50 °C.^[36] Our group recently reported two short and highly stabilizing β-strap motifs built around a Trpzip motif, yet the folding of hairpins bearing long loop segments (≥10-residues) remained challenging (T_m varying from 25 to 51 °C).^[37] As shown in Figure 1B, probing NOESY correlations within each W/W pair and significant side-chain chemical shift deviations (CSDs) recorded in Trpzip 1a suggested that CH/π interactions could be responsible for the unique edge-to-face (EtF) arene stabilization. To optimize this strategy, we now report the development of a hairpin platform that investigates the relationship between aromatic–aromatic interaction of tryptophans within the Trpzip. Subtle stereoelectronic effects could be observed in our series of hairpins 1-2 and the electronic densities of face tryptophans (pyrrole and benzene rings: ^1MW and ^7OW using the nomenclature shown in Figure 1C) were successfully tuned to enhance the global hairpin stability in water over a larger range of temperature. In addition, our density-functional theory (DFT) calculations confirmed that three distinct CH/π interactions contribute to each EtF assembly and further suggested that additional NH/π interactions must be accounted for in the overall Trpzip stabilization. These results provide a more comprehensive picture of each non-covalent interaction contributing to the structure–stability relationship of Trpzip hairpins.

Figure 1.

A. Trpzip NMR structure showing the W/W pair geometry in EtF stacking. B. Incorporation of non-proteinogenic tryptophans within the β-strap of Trpzip 1a to enhance the strength of CH/π interactions in W/W pairs. C. List of unnatural building blocks evaluated in this study.

Results and Discussion.

To our knowledge, all earlier attempts to enhance the stability of the W/W packing found in the original Trpzip by swapping tryptophans with other proteinogenic or unnatural aromatic α-amino acids (e.g., Y, F and analogs) have failed.^{[34–35,38–43]} Along these lines, a number of studies support the notion that indoles are some of the best π-donor aromatic heterocycles to achieve CH/π interactions, given their larger and more polarizable π-surface.^{[15,16,44–48]} A double-mutant cycle calculation reported by Kelly suggested that a W/W pair in EtF contact accounts for a stabilization of ~ 0.55 kcal/mol.^[49] As such, we hypothesized that the uniqueness of the tryptophan EtF packing could originate from a combination of CH/π interactions with alkyl (β-methylene, H_β2) and aryl (indole C4(H_ε3) and C5(H_ζ3)) H-bond donors on edge tryptophans interacting with the electron-rich π-system of face tryptophans. Guided by several NMR solution structures of W/W stacking in EtF geometry (PDB codes: 1LE0, 1LE3, 1GIG), we envisioned that both H_β2→benzene nucleus and H_ε3→pyrrole nucleus interactions could be tuned separately to achieve a greater overall stabilization (Figure 1A). To test this hypothesis, we synthesized a library of hairpins inspired by our benchmark Trpzip 1a using a typical Fmoc-chemistry on solid support. A selection of commercially available or readily accessible unnatural tryptophans (Figure 1C) were inserted at either edge or face positions (2/15 and 4/17 respectively) in hairpin scaffolds 1 or within both pairs simultaneously in scaffolds 2 (Figure 2). Whereas the nucleation of short loops (e.g. β-turns) is expected to thermodynamically stabilize the hairpin fold^[50–52], random loop sequences of varying length and conformational strain can be detrimental to the hairpin folding path. To avoid any local perturbation induced by biased loop turns^[53,54], we selected a relatively long and highly flexible polyglycinyl –G₄KKG₄– loop sequence having no intrinsic conformational bias, limited side-chains interactions, and no innate structuring intramolecular hydrogen bonds.^[55–57] As shown by the sharp thermal unfolding transition for our benchmark Trpzip 1a in Figure 2, the flexibility of the polyglycinyl loop is likely reducing the difference in entropy between folded and unfolded states, thus affording a reliable two-state approximation model for the hairpin folding free energy (ΔG_F^o) calculations.^[58–60] At first, we obtained several direct evidences by NMR spectroscopy of the EtF packing in 1a with a set of cross-strand NOESY correlations between the W side chains (Figure 1B) and some pronounced upfield chemical shift deviations (CSDs) of −0.98 and −1.10 ppm for aliphatic H_β3, and of −0.77 / −0.65 ppm and −2.28 / −2.44 ppm for both aromatic H_ζ3 /H_ε3 at the respective W4 and W17 edge positions (inset bar graphs in Figure 2).^[61] The NMR and circular dichroism (CD) data, vide infra, revealed that hairpin 1a is indeed well structured with an overall 72 ±1 % folding at physiological temperature (310 K).

Figure 2.

Thermal denaturation of Trpzip foldons with one or two W^face/W^edge pair modifications (1-2) in comparison to the benchmark Trpzip 1a. Best-fitted CD-melts of the fraction folded (χ_F) as a function of temperature were used to determine thermodynamic stability parameters (inset Tables). Direct evidence of alkyl and aryl CH/π interaction strength by CSDs amplitude (right panel). Error on chemical shift is estimated to be ±0.01 ppm. A. Combining edge and face modifications. B. Edge ring modulations of H-bond donors. C. Face ring modulations of polarizable π-cloud.

Tuning the W^edge/W^face electronic properties to modulate the EtF packing of Trpzip systems: a CD-study.

Using the Trpzip 1a as our benchmark, we initially tested a HOMO/LUMO approach that we believed could strengthen the CH/π interactions within the two W2/W17 and W4/W15 pairs (Figure 2A). Far-UV CD spectra were obtained for assessing the folded hairpin conformations in solution. Facing W/W residues within an EtF packing generate a characteristic and intense π–π* exciton couplet with negative and positive maxima at 214 and 228 ± 1 nm respectively (Figure 2).^[35] The decay in amplitude of the molar ellipticity at 228 nm ([θ(T)]₂₂₈) upon heating was used to measure each hairpin unfolding transition. These thermal denaturations were performed in an aqueous buffer (0 to 95 °C), and the unfolding transitions were fitted to the Gibbs-Helmholtz equation for a two-state model.^[61] The thermal-CD analysis clearly showed that all the hairpins studied herein unfolded reversibly. The best-fitted melting curves were computed to obtain plots of %-folding as a function of temperature and determine the corresponding melting temperatures and folding thermodynamic parameters with accuracy (goodness of the fits, R² ≥ 0.98; Supporting Information). As shown in Figure 2A (inset Table), the introduction of a ^5OW at position 2 slightly destabilized the fold of hairpin 1b. In contrast, the introduction of a small but highly electronegative fluorine of ^5FW at position 17 created an unexpectedly larger destabilization in 1c.^[62] This effect was attributed to an electronic repulsion, as the fluorine substituent being in close proximity to the cross-strand pyrrole nitrogen of W2 (see increased internal entropy in Table S3, Supporting Information). Then we combined these two modifications to increase the non-covalent π interactions between ^5OW/^5FW pairs. The results for 1d demonstrated that a single modified pair was destabilizing (ΔΔG_F^o of + 0.83 kcal/mol to 1a), while two pairs fully unfolded peptide 2d to afford a random coil-like structure. In addition, NMR CSDs in 1d revealed more insights about the EtF interactions (inset Figure 2A). While the aliphatic CH_β2/π interactions appeared to be slightly enhanced (H_β2 upfield shift of −1.20 and −1.19 ppm), the aromatic CH_ε3/π interactions were significantly reduced, suggesting that the indole rings are on average further apart on the NMR time-scale. These unexpected results —that were later explained by the simultaneous destabilization of NH/π contacts—prompted us to evaluate each CH/π interaction more systematically by modifying edge and face tryptophans separately (Figure 2B–C).

The Wheeler–Houk model of benzene dimers in EtF stacking suggest that substituent effects on the edge ring are primarily electrostatic in nature.^[63] These electrostatic interactions originate from the quadrupole moment of the aromatic ring which creates an electron-poor edge. One would therefore expect that the introduction of a fluorine at C6 (^6FW), meta to the C–H_ε3 bond donor should result in a constructive stabilization (Hammett correlation), yet the fold of 1e was strongly destabilized (ΔΔG_F^o of + 0.56 kcal/mol to 1a, Figure 2B). As surprising to us was the destabilization created by one or two ^7AW residues in hairpins 1f and 2f. In principle, the introduction of electron-withdrawing groups at the para-position 7 should have considerably enhanced the electrostatic component of the CH/π by increasing the partial positive charge of the edge H_ε3 hydrogen. Previous studies reported that ^7AW residues are prone to form water bridges via a H-bond network between the N1 and N7 positions, which we felt could be associated with an entropic penalty.^[64,65] However, the incorporation of two methylated aza-residues ^MaW in 2g turned out to produce a fairly similar effect (see CD-melts and T_m inset Table, Figure 2B). Having a strongly electron-withdrawing nitrile at C5 (^5CW) proved also extremely detrimental to the hairpin fold of 2h as shown by a ΔG_F^o of + 0.48 kcal/mol resulting in a complete unfolding at 310 K. We then tested a reversed strategy using an electron-donating group at position 7 that could enhance the negative electrostatic potential above and below the indole ring to enhance the edge C–H_ε3 bond polarization . This electronic adjustment was achieved through a 7-methoxy donating group (^7OW) leading to the most stable edge-modified hairpins 1i and 2i. Yet the NMR data on 2i revealed that the all the local CH/π interactions remained weaker than in the benchmark Trpzip 1a (upfield shifts of −2.21 and −2.44 ppm). Collectively, the results of edge tryptophans modulations (W4 and W17) with electron withdrawing groups at the C5, C6, or C7 ring positions were largely unexpected (Figure 2B). Conversely, the stabilization by an electron-rich ^7OW tryptophan (at C7) tend to suggest that additional interactions might contribute to the local stabilization of edge tryptophans (vide infra DFT calculations section).

We then turned our attention towards modifications of face tryptophans W2 and W15 which play the role of π-electron donors in EtF interactions (Figure 2C). It is well documented that for such π-donor systems, contributions to the global interaction energy from dispersion and electrostatic forces vary with the arene substituents.^[66–70] To maximize the overall interaction energy and tune the CH/π at these positions, electron-donating substituents were therefore investigated. Face substitutions at C5 are likely to generate steric clashes with the hairpin strand, and 2b (e.g. ^5OW) is no exception to this empirical observation. To probe the effect of electron-rich substituents on those indole rings while minimizing any potential steric hindrance, positions N1 and C7 were selected. The introduction of ^7OW residues in 1j and 2j afforded a substantial stabilization throughout a larger range of temperatures while the %-folding at physiological temperature did not exceed the one of our benchmark Trpzip 1a. These results were further confirmed by the electron-donating effect of the N-methyl tryptophan (^MeW) in peptides 1k and 2k. Indeed, by tuning the electron density of the pyrrole ring, the largest net gain in stability was observed in these Trpzip systems. Hairpin 2k was characterized by a significantly greater stability over the range of temperature studied, with a large T_m upshift of 6 °C relative to the benchmark 1a. This stabilization gain in 2k was further quantified through the largest CSDs recorded for both alkyl and aryl CH/π components (CSDs for H_β2 up by 0.20 ppm, and up by 0.20–0.30 ppm for H_ε3 ~15%). The magnitude of these anisotropic effects for both edge tryptophans relative to all other analogs indicates that the electronically enriched ^MeW/W pairs are in close proximity forming a strong EtF packing. The steepest slopes observed on CD-melts (at T_M) for hairpin 1a, 1j, and 1k highlight a greater cooperativity during the unfolding transition between strap and loop segments, in comparison to analogs 2j and 2k (see Supporting Information, Figure S7). According to the Zimm-Bragg theory, these results strongly suggest that the unnatural tryptophans are enhancing local interactions by transiently populating conformational states throughout the global unfolding. Collectively, these results suggest that by augmenting the π-electron density of both face tryptophans (with inductive and π-resonance donors), a conclusive enhancement of stability can be obtained.

NMR study reveals that a combination of CH/π and NH/π interactions are combining forces.

While the tertiary structures of hairpins 1–2 can be readily assessed by CD-melts, the structure-stability relationship induced by XH/π interactions of varying strength can only be examined at atomic level by NMR experiments (Figure 3).^[71] Indeed, the strongest EtF W/W interactions can be definitely recognized in the TOCSY fingerprint aromatic region (4.80–5.50 ppm) due to the magnitude of ring currents imparted by face indole-rings onto the edge-tryptophans W4 and W17 (Figure 3B). The anomalous chemical shifts experienced by edge H_ε3 resonances of any new Trpzip system in an EtF geometry can therefore be easily compared by spectral overlay to the benchmark hairpin 1a. For example, the strongest upfield-shifted H_ε3 resonances along with a larger split of W4/W17 CSDs are to be expected for the most-folded Trpzip structure 2k, because the two indoles should be in a relatively static interacting conformation. Conversely, minor W4/W17 CSDs for edge tryptophans are characteristic of a more disordered structure (e.g. 2i). The strength of CH_ε3/π interactions can therefore be qualitatively ranked such as 2k >> 2j >1a > 2i.

Figure 3.

Probing the strength of EtF interactions between W/W pairs in TOCSY spectra. Overlay of the aromatic fingerprint spectral regions for A. ^FaceW and B. ^EdgeW of hairpins 1a, 2i-k

To probe the strength of each CH/π and other H-bond interactions in 1a and 2k, the temperature dependence of amides (H_N) and tryptophan (H_ζ3, H_β2, H_ε3) resonances were recorded over a wide range of temperature (0–70 °C, Fig. S13/S14, Supporting Information). Upon increase in temperature, the resonances of amide protons exposed to bulk water molecules suffer a large upfield shift while hydrogen bonded amides lengthen to a lesser extent resulting in smaller upfield shifts. This effect is characterized by temperature coefficients (ΔδH_N/ΔT) of about −6.0 to −10.0 ppb/K for random-coil like amide protons, and more positive than −4.6 ppb/K for hydrogen bonded amides.^[72] Using this criterion, the ΔδH_N/ΔT of −4.0 and −4.4 ppb/K for the terminal residue E18 and the corrected values of −1.9/−2.0 and −1.7/−1.7 for the V3/V16 pairs, in 1a and 2k respectively, are suggestive of persisting H-bonds within these hairpin structures. Due to the large shielding effect imparted by an indole ring current on X-H bonds, one would expect that H-bond donor hydrogens would experience an opposite downfield chemical shift deviation at higher temperatures characterized by a positive temperature shift coefficient (ΔδH/ΔT ≥ 0 ppb/K). As the energy of hydrogen bonds in XH/π interactions is strongly distance dependent (1/r³), the chemical shift deviations are expected to deviate from linearity upon heating.^[32,73,74] Indeed, this is exactly what we found with the large resonance shifts of H_ε3 and H_β2 at both edge W4/W17 and the H_N of G14 (see Table S9) characterized by a markedly non-linear behaviour at temperatures comparable to the global hairpin denaturation T_m~50 °C obtained in the CD study. H_N and H_ar chemical shifts typically display a linear and negative temperature dependence (Δδ/ΔT ≤ 0, within experimental uncertainty), although it is striking that protons involved in XH/π interactions deviate strongly from linearity into melting curves resembling to a local unfolding (Δδ/ΔT ≥ 0, up to 25 ppb/K at T_m).The temperature-NMR data for hairpin 1a suggested that even after local conformational corrections, CH/π interactions are not well protected from solvent exposure while the N-H bond of G14 was confirmed to be hydrogen bonded (ΔδH_N/ΔT of +3.6 ppb/K). In hairpin 2k, the amide-π interaction remained and all the CH/π interactions appeared more shielded from solvent revealing a stronger hydrogen bond characteristic. The strength of the W17 CH_ε3/π could be explained by an additional local rigidity of the edge W17 induced by a N-terminal backbone ammonium NH/π interaction with R1.^[44,75] Collectively, these results are pointing out that XH/π interactions can be correlated to temperature coefficients within the same order of magnitude than typical hydrogen bonds (Δδ/ΔT ≤ +4.6 ppb/K). It is clear from the temperature coefficient data of individual residues that CH/π and NH/π interactions ruptured at different temperatures and that the hairpin 2k unfolded more gradually (with less cooperativity) than expected by a more simplistic two-state model. Overall the TOCSY fingerprint analysis is a quick label-free technique to determine the presence of CH/π interactions, while the strength of each individual CH/π interaction can be qualitatively measured at the atomic level by temperature experiments. The information of temperature shift coefficients serves as a proxy for the strength of each individual XH/π interaction. Given the importance of XH/π interactions in biological systems^[10,12–13], protein—ligand interactions, and catalyst design^[6], we anticipate that the measurement of positive temperature coefficients could serve as useful quantitative predictors of contact strength.

Geometrically optimized DFT-model of the Trpzip β-strap motifs.

To confirm the presence of XH/π interactions and their relative strength in the EtF packing, DFT calculations were performed on a set of Trpzip molecules^[76] at the M05–2X/6–31G(d) level of theory.^[77] To ease the DFT calculations the –G₄KKG₄– loop of hairpin 1a excluding the glycine unit (G14) (RW²VW⁴–G₄KKG₄–W¹⁵VW¹⁷E) was replaced by a flexible alkyl chain segment of 13 methylenes in to create an hairpin model 1a’ (Figure S17). The geometry of 1a’ was then optimized in a solution phase using a polarizable continuum model (PCM) considering H₂O as solvent.^[78] The strength of individual CH/π interaction present between the W2/W17 and W4/W15 pairs in 1a’ was calculated through natural bond orbital (NBO) analysis and the corresponding hydrogen bond interaction energies (E²) were obtained (see Table S11).^[79] The NBO energies calculated for each individual CH/π interaction were found in excellent agreement with our NMR data (Tables S8–S9) further establishing a trend in strength H_ζ3 < H_β2 < H_ε3 of about two-fold. Importantly, the energy minimized DFT structure 1a’ revealed that two NH/π interactions of significant energy (Figure S16) are critical to close both ends of the Trpzip structure (ammonium NH/π of R1 to W17, and amide NH/π of G14 to W4).^[] As shown in Figure 4 on the DFT-model hairpin 2k’, the presence of strong NH/π interactions flanking both edge tryptophans certainly explains the difficulties in tuning the electronics of the edge tryptophan residues.^[80,81] If electron-withdrawing substituents on W^edge enhance the CH/π strength, they concurrently weaken the NH/π interactions produce by the same ring therefore mitigating the overall energy gain (Figure 2B). To verify this hypothesis, the same DFT calculations and NBO analyses were performed on other Trpzip hairpins 1e’, 1f’, 1i’, and 2k’ having electronic modifications on the W^face (W2/W15) as well as W^edge (W4/W17) residues (Figures S18–S21). All the optimized peptide structures were confirmed as energy minima through their frequency calculations performed at the same level of theory. NBO overlap views for the CH/π and NH/π interactions present in these hairpins are displayed in Figures S22–S26. Overall, the trend of the NBO interaction energies 2k’~1a’>1i’>1f’>1e’ agrees well with the experimental ΔΔG_F⁰ values obtained from thermal denaturations (Table S11), and the measured chemical shift deviations (Table S8). In addition, the calculated stabilization energy in 1e’ was far lower the reason being that NH(R)/π(W17) interaction was largely mitigated by an NH•••F interaction. This new insight supports our experimental data for hairpin 1e, suggesting that electron-withdrawing groups on edge rings (i.e. fluorine, cyano, or aza-ring) are abrogating the NH/π interactions, which in turn destabilizes the entire Trpzip structure. To correlate our results to the Wheeler—Houk model of EtF benzene dimers, the tryptophan/tryptophan interaction was revisited in a simplified system of substituted 3-ethylindole dimers. This model preserved both alkyl and aryl CH/π interactions. Interestingly, the optimized heterodimeric EtF structures presented a similar CH/π pattern than the one found in the β-hairpin peptides (i.e. alkyl C-H_β facing phenyl ring and aryl C-H_ε facing pyrrole ring (see Supporting Information, Figure S27). To evaluate the effect of substituents on the aryl—aryl interactions, the overall interaction energy was decomposed into electrostatic, dispersion, polarization, and exchange-repulsion components (Supporting Information, Tables S17–18).^[82] Our results suggest that substituent effects on the edge ring are dictated by a combination of electrostatic and dispersion in large part generated by direct interactions between the local dipoles of substituents to the other ring (Table S17). This trend is exemplified by the destabilizing effect at position 5 in which repulsion can be severe due to the proximity of the substituents to the other ring. On the face ring (Table S18), substituent effects were generally found to be dominated by dispersion, thus suggesting that positions 1 and 5 (to a lesser extent) can be readily tuned to stabilize the EtF stacking (electron-donating groups providing more favorable interactions). Taken together, the effect of substituents on the EtF interactions in our heterodimeric 3-ethylindole model was found to be in general agreement to those reported by Wheeler and Houk,^[63,66] and Sherill^[82] on benzene-derived dimers. Collectively the DFT model revealed that both edge tryptophans W4/W17 are not only involved in CH/π interactions with W^face residues, but also in critical NH/π interactions with the flanking residues in diagonal positions (H_N of both terminal R1 and loop G14) thus providing an additional stabilization.

Figure 4.

DFT calculated 3D-structure of hairpin 2k’ with XH-π interactions and H-bonds correlated to the NMR data of 2k. Side chains of R and E residues at hairpin termini are omitted for clarity.

Conclusions.

In summary, a practical platform for studying the XH/π interactions of unnatural tryptophans within Trpzip paradigm was presented. Our results demonstrated for the first time that a strong correlation exist between the global unfolding of hairpins (tertiary structures) and the gradual rupture of local XH/π interactions observed at an atomic level within the Trpzip motif. The thermal denaturations by CD-melts provided a good estimate of the local EtF stabilization between unnatural tryptophans. To our knowledge, the introduction of two unnatural W/^MeW pairs represents the first example of an engineered Trpzip with improved stability over a larger range of temperatures (T_m upshift of 6 °C). Our NMR data definitely established that the Trpzip EtF stacking stabilization stem from a combination of alkyl and aryl CH/π and NH/π interactions. Whereas CH_ζ3/π are relatively weak in this system, the chemical shifts temperature coefficients of H_ε3/H_β2 from edge tryptophans W4/W17 (ΔδH/ΔT ≤ +4.6 ppb/K) are consistent with the magnitude of typical hydrogen bonds. The combination of CH/π and NH/π interactions described herein was found to confer a unique stability to the β-strap motif to minimize its length and ultimately to assemble β-hairpins with challenging long loops of large conformational entropy. While the π-systems of W^face can be fine-tuned by introducing electron-donating substituents at N1 or C7, the modulation of W^edge H-bond donors proved difficult. In fact, our DFT study revealed that in this pseudo-symmetrical Trpzip system, the flanking NH/π interactions act in an opposite interplay to the CH/π interactions thus rendering the edge tryptophans particularly sensitive to subtile electronic changes. Ultimately, we expect that the present study will provide a foundation for accessing challenging β-hairpins with long loops of varying rigidity typically found in proteins and antibodies. We also anticipate that the combination of alkyl/aryl CH/π interactions and the tuning of their relative strength using modified indole rings in EtF stacking will find further applications in designing catalysts and other supramolecular assemblies.

METHODS SECTION

Peptide Synthesis.

The library of peptides was prepared by solid-phase peptide synthesis (SPPS), using Fmoc chemistry as described elsewhere.^[83] Individual peptides were purified by semi-preparative RP-HPLC. Peptide identity was confirmed by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry and NMR spectroscopy. Details of synthesis and characterization for all unnatural tryptophans and the corresponding peptides are in the Supporting Information.

UV Spectrophotometry.

The molar absorptivity of each unnatural tryptophan derivative was obtained at 280 nm and exploited to calculate the peptide concentrations for NMR and CD experiments. In brief, a tryptophan derivative sample was precisely weighted (2–3 mg) and solubilized in methanol to prepare a stock solution (1.0 mg/mL). Three diluted solutions of 100, 50, and 25 μg.ml⁻¹ respectively were prepared from the stock solution and the molar absorptivity (ε₂₈₀) was calculated by measuring the absorbance (Abs) of each solution at 280 nm. ε₂₈₀ was obtained by fitting experimental values to a linear model following Beer’s Law Abs=C*ε₂₈₀*l with C, and l being respectively the α-amino acid concentration and the path length of the cuvette. Standard deviations for the molar absorptivity (ε280) were calculated from four individual measurements. All UV spectra are in the Supporting Information.

Circular Dichroism.

Peptide solutions for circular dichroism (CD) were prepared in a phosphate buffer (15 mM, pH 6.5) at 20–100 μM concentration range. CD spectra were collected on a JASCO J-810 Spectropolarimeter with a temperature controller module JASCO PFD-425S. In brief, both CD spectra and CD-melts (0 to 95 °C) were recorded and the raw spectra were smoothed with SpectraGryph 1.2. A synthetic ‘random’ coil RGVW-G₄KPG₄-WVWE (SI-4) with a primary sequence similar in length and compactness to all the hairpins studied herein, was prepared and characterized by CD to determine the molar ellipticity value [θ₂₂₈]_RC(95) of an unfolded molecule (0%-folded baseline). The elipticity value [θ₂₂₈]_RC(95) of SI-4 was subtracted to the melting curves to normalize the molar ellipticity which was then fitted as a function of temperature [θ(T)]₂₂₈ to the thermodynamic Gibbs-Helmholtz equation for two-state model as previously described using a nonlinear fitting square routine protocol in OriginPro 9.0 (Originlab Corporation, U.S.A.).^[61] CD spectra, full experimental fitting protocols, thermal stability curves, and tabulated thermodynamic data are in the Supporting Information.

NMR Spectroscopy.

NMR samples were prepared by dissolving the freeze-dried peptide (~ 1–5 mg) in D₂O or a mixture of water and D₂O (9:1 v/v) using 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as internal standard for chemical shifts (0 ppm). Residual water peak was suppressed using a PRESAT pulse sequence. All spectra were recorded at 291K (18 °C) either on a Varian Mercury500 (500 MHz) or on a Bruker 400 UltraShield (400 MHz) spectrometer. For temperature NMR experiments, lyophilized hairpins 1a and 2k were reconstituted in H₂O/D₂O (9:1 v/v) using DSS as internal standard. Proton temperature dependence was measured by recording a set of ¹H NMR and TOCSY with PRESAT for water suppression, in a temperature range of 273 to 343 K by intervals of 10 K. The temperature was stabilized for 5 mins before spectral acquisition. Full NMR spectral data, and tabulated data of chemical shift deviations, NOESY schematics, and temperature coefficients (H_N and H_ar) are in the Supporting Information.

DFT Calculations.

To confirm the presence of XH/π interactions and their relative strength in the EtF packing, DFT calculations were performed on our Trpzip model system. An initial 3D-structure generated in Pymol from a known Trpzip system 4 (PDB: 1LE3) was used by replacing the polyglycine loop sequence −G₄KKG₄− of 1a by a flexible chain segment of 13 methylenes in hairpin 1a’ while keeping the last glycine unit of the loop (G14). The geometry of 1a’ was optimized at the M05–2X/6–31G(d) level of theory, in the solution phase considering H₂O as solvent, using the integral equation formalism variant of polarizable continuum model (IEFPCM)^[84] available with the Gaussian 09 package.^[77] All final geometry-optimized structures had zero imaginary frequencies. The strength of individual CH/π interaction present between the W2/W17 and W4/W15 pairs in 1a’ were qualitatively calculated through natural bond orbital (NBO) analysis using NBO 6.0 software.^[85] In addition, a simplified EtF stacking of substituted 3-ethylindole heterodimers (calculated at the same M05–2X/6–31G(d) level of theory) is reported with a decomposition of the interaction energies in the Supporting Information. Optimized structures of 1a’, 1i’, 1f’, 1e’, and 2k’ and their Cartesian coordinates, along with the NBO analysis of CH/π and NH/π interactions are in the Supporting Information.