Folding in Place: Design of β-Strap Motifs to Stabilize the Folding of Hairpins with Long Loops

Abstract

Despite their pivotal role in defining antibody affinity and protein function, β-hairpins harboring long noncanonical loops remain synthetically challenging because of the large entropic penalty associated with their conformational folding. Little is known about the contribution and impact of stabilizing motifs on the folding of β-hairpins with loops of variable length and plasticity. Here, we report a design of minimalist β-straps (strap = strand + cap) that offset the entropic cost of long-loop folding. The judicious positioning of noncovalent interactions (hydrophobic cluster and salt-bridge) within the novel 8-mer β-strap design RW(V/H)W⋯WVWE stabilizes hairpins with up to 10-residue loops of varying degrees of plasticity (T_m up to 52 °C; 88 ± 1% folded at 18 °C). This “hyper” thermostable β-strap outperforms the previous gold-standard technology of β-strand-β-cap (16-mer) and provides a foundation for producing new classes of long hairpins as a viable and practical alternative to macrocyclic peptides.

Graphical Abstract

INTRODUCTION

In the past 2 decades, the mimicry of proteins’ secondary structures into peptide foldamers has emerged as a powerful modality to build novel materials¹ and bioactive macromolecules.² The identification of stand-alone foldamers such as β-hairpins and their development into macrocyclic mimics have proved to be a prolific strategy for the development of protein–protein interaction (PPI) inhibitors.³ Although pivotal advances in the mimicry of helices, turns, and β-sheets have been made,⁴ the general design principles guiding the folding of β-hairpins with long loops are still in their infancy.⁵ In fact, the folding of β-hairpin peptides is essentially limited to short turn-induced β-sheets, in which the turn-locus forces the secondary structure nucleation through chain reversal and the formation of an antiparallel strand register.^3,6 In stark contrast, most β-hairpins typically found in the proteome are “turnless” with long and unstructured loops (β-arches) stabilized by β-pleated sheets within secondary and tertiary structures.⁷ It is proposed that the folding of these turnless long β-hairpins arises not only from loop nucleation⁸ but also from pairing-residue interactions⁹ and hydrophobic clusters formation¹⁰ along β-sheets.¹¹ From a fundamental standpoint, the current frontier in hairpin structures mimicking closely the native folds found in protein remains largely limited to short [2:2]/[2:4]-, [3:5]-, [4:4]/[4:6] hairpins.^12,13 These synthetic hairpins are typically nucleated around a 4-residue loop having an innate turn propensity (e.g., DDATKT and NPATGK sequences, Figure 1A).^8c,14

Figure 1.

(A) Strand alignment and H-bonding pattern in folded [4:6]-hairpins based on DDATKT and NPATGK sequences. (B) Schematic representation and nomenclature of a [10:12]-hairpin displaying the 10-mer CDR-H3 loop excised from the pembrolizumab antibody target.

Different challenges are indeed associated with length: short loops in [2:2]/[2:4]- and [3:5]-hairpins are usually centered around specific β-turns and β-bulges,¹⁵ while the foldings of [4:4]/[4:6]- and larger hairpins are exceedingly more difficult to control due to the conformational space accessible by random sequences of α-amino acids and the innate conformational strain held by their structures. Except for some rare examples of long hairpins (≥10-mer loop) championed by Andersen¹⁶ by developing a β-cap–β-strand stabilizing motif of 16 amino acids length, the significant entropic penalty inherent to the nucleation of long loops and their vulnerability to fraying have basically hindered their synthetic development. For these reasons, most structural studies have only evaluated the formation of long loops built from simpler and highly flexible polyglycine chains.^16,17 Other studies have shown that associating β-strands and β-caps of 10–12 residues is the minimal length required to stabilize hairpins with relatively short loops (2–4 residues).¹⁸ A net improvement of design is needed to create long and well-folded β-hairpins at physiological temperatures and obtain a viable and practical synthetic alternative to macrocyclic peptides of a large size (>10-mer).¹⁹

Despite the need for novel protein and antibody loop mimics, the lack of technology has hampered the synthesis of large hairpin modalities suitable to this space. Therefore, we were drawn to revisit this issue by designing some minimalist β-straps amenable to crafting loops of conceptually any length, sequence, and conformational strain (Figure 2).²⁰ To this aim, we targeted the synthesis of a [10:12]-hairpin, which mimics a highly constrained loop (RDYRFDMGFD sequence) excised from the heavy chain complementary determining region 3 (CDR-H3) of the marketed antibody drug pembrolizumab, a programmed cell-death 1 inhibitor (Figure 1B).²¹

Figure 2.

Design of short stabilizing β-strap motifs for the hairpin fold.

We hypothesized that combining several types of noncovalent interactions²² at non-hydrogen-bonded (NHB) and hydrogen-bonded (HB) sites would enable long constructs of typical β-cap–strand–loop–strand–β-cap architecture to be truncated into their simplest forms β-strap–loop–β-strap and still retain a high efficacy (Figure 2). Here, we report a side-by-side comparison between three strap designs 1–3, which demonstrates that a judicious combination of Coulombic interactions at the hairpin termini HB-site (persisting at longer distances) with van der Walls and hydrophobic contacts at NHB-sites can overcome the loop entropy to strengthen hairpin structures (Table 1).

Table 1.

Truncated β-strap optimization inspired by known GB1-analog sequences and previously reported hairpin stabilities^a-g

Entry	Name	Peptide sequencea	Tm (°C)b,c	Loop %-fold χF at 291-298Kd,e	β-Strap %-fold χF at 291Kf	% Yieldg
126	Loop	Ac-NPATGK-NH2	n.d.	6 ± 7%d	n.d.	-
226	Tr-HP7	Ac-W-NPATGK-W-NH2	3-7	49 ± 1%d	n.d.	-
326	HP7	KTW-NPATGK-WTE	66	89 ± 1%d	n.d.	-
423,27	GB1m3	KKWTY-NPATGK-FTVQE	60	94d-97e ± 6%	n.d.	-
525d	Trpzip9	GEWVW-DDATKT-WVWTE-NH2	92	n.d.	n.d.	-
Novel peptides from this study--- [4:6]-& [8:10]-Hairpins
6	1a	Ac-W-NPATGK-WTG-NH2	<0	42 ± 4% (70%)e	28 ± 1%	57
7	1b	Ac-W-DDATKT-WTG-NH2	20.0 ± 2.5	n.a.	54 ± 7%	32
8	1c	Ac-WTF-NPATGK-FKWTG-NH2	30.3 ± 0.1	91 ± 2% (151%)e	63 ± 4%	-
9	2a	Ac-WVW-NPATGK-WVW-NH2	101.1 ± 0.1	93 ± 2% (154%)e	99 ± 1%	15
10	2b	Ac-WVW-GPGTGK-WVW-NH2	23.9 ± 2.5	6 ± 33% (9%)e	56 ± 1%	15
11	3a	RWVW-NPATGK-WVWE	97.9 ± 0.1	93 ± 2% (154%)e	89 ± 14%	-
12	3b	RWHW-NPATGK-WVWE	80.7 ± 0.1	>99 ± 2% (167%)e	95 ± 2%	-
13	3c	RW-NPATGK-WE	61.6 ± 0.1	84 ± 2% (139%)e	84 ± 11%	-
14	3d	RWVW-G4KKG4-WVWE	51.8 ± 0.1	>99 ± 2%	88 ± 1%	70
15	3e	RWVW-RDYRFDMGFD-WVWE	25.3 ± 0.1	22 ± 8%	55 ± 1%	30
16	3f	RWVW-RDYRGDMGFD-WVWE	46.7 ± 0.1	55 ± 3%	82 ± 1%	35

^aAll HB-sites from the β-strand/strap including the (S±1) positions are underlined, and residues from hydrophobic clusters and β-straps are bolded in the sequences.

^bT_m are estimates of melting transition temperatures obtained from the CD thermal denaturation curves ([θ(T)]₂₂₈), based on the W/W pairs exciton couplet at 228 ± 1 nm.

^cFor all novel hairpins 1-3, T_m values were obtained from the best-fitted curves of [θ(T)]₂₂₈ to the Gibbs-Helmholtz equation. The loop %-folding in the NPATGK loops were previously determined at 298K from either

^dan average of H_N CSD values at T(L3), K(S+1) and from the G(L4) ΔδH_α/α’, assuming K1R-HP7 as the fully-folded state, see reference [26]

^esolely from ΔδH_α/α’G(L4), assuming a value of 0.33 ppm as the fully-folded state from a macrocycle, see reference [27]; Loop %-folding from this study were calculated from the same ΔδH_α/α’G(L4) model (reported in parentheses), and further corrected assuming hairpin 3b as fully-folded (0.55 ppm). Standard errors for the folded populations were estimated by assuming a 0.01 ppm confidence on the chemical shift values assigned by ¹H NMR.

^fLocal β-Strap %-folding calculated from the best-fitted melting curves of [θ(T)]₂₂₈.

^gIsolated yields from either crude or purified peptides (see Table 2).

RESULTS AND DISCUSSION

We set our aim as to construct monomeric β-hairpins soluble in water and stable at physiological temperatures. To understand the structural requirements of stabilizing motifs in the straps, we began by designing peptide constructs by including the DDATKT^14a and NPATGK,^8c d loops of known [4:6] hairpins as reversing turn loci inspired from GB1, and the B1 domain of Streptococcal protein G (Table 1).²³ The underlying idea was to compare peptides 1a–c harboring an Ac-W⋯WTG-NH₂ β-strap, inspired by a capping created by Andersen (Table 1, entries 6–8)^15,24 with constructs 2a–b having an Ac-WVW⋯WVW-NH₂ β-strap (Table 1, entries 9 and 10) truncated mimic of the tryptophan zipper motif (Trpzip) developed by Cochran²⁵ such as the Trpzip9, and some of most stable stand-alone hairpins known to date HP7 and GB1m3 (60 ≤ T_m ≤ 92 °C, entries 3–5).^26,27 Second, we evaluated the stability of a novel RW(V/HW)⋯(WV)WE β-strap, which combined an ionic pair at the HB-site termini in addition to the Trpzip hydrophobic core of constructs 3a–c (entries 11 – 13) in comparison to the Coulombic capping of β-hairpin HP7 previously positioned at an NHB-site (entry 3). Here, we demonstrated the superiority of β-strap 3 and further evaluated its ability to structure longer loops (e.g., hairpins 3d–f, entries 14–16). All peptides 1–3 were synthesized on resins using a typical Fmoc-based solid-phase peptide synthesis (SPPS) in 15–70% yields with relatively high crude purity after cleavage (≥78% by RP-HPLC), which was improved by semipreparative RP-HPLC purification when needed.

The final deprotection/cleavage step proved to be challenging for constructs 2 due to the steric hindrance imparted by the resin on the nearby C-terminal Trp residue leading to side reactions and irreversible reattachment of the peptide to the resin.²⁸

Secondary Structures and Thermal Stability Analysis Determined by Circular Dichroism (CD) Spectroscopy.

Changes in molar ellipticity intensity as a function of temperature [θ(T)] can be exploited at specific wavelengths of CD spectra to examine the thermodynamic folding/unfolding transition of secondary structures.²⁹ For example, the cross-strand interaction between W/W pairs positioned at NHB-sites is well known to produce a distinct and intense exciton couplet at 212 and 228 ± 1 nm (negative/positive maxima) from a π─π* transition, which can be used as a local spectroscopic probe of β-hairpin folding.^30,31 Indeed, the edge-to-face (EtF) geometry between indole rings of facing Trp residues^30,32 responsible for the π─π* transition is also coupled with a significant electrostatic stabilization created by multiple CH─π interactions.³³ As a result, the W/W EtF interaction was widely exploited to measure the level of backbone folding in hairpins.^30b The decrease in amplitude of the prototypical 228 nm exciton intensity upon heating was therefore monitored by CD spectroscopy to evaluate the thermal denaturation of β-straps 1–3 and determine the thermodynamic parameters characterizing the transition between the folded and unfolded states of hairpins (Figure 3).

Figure 3.

(A) Thermal denaturation of hairpins 1–3: fraction folded (χ_F) curves as a function of temperature^a. (B) Thermal stability curves for [4:6]-and [10:12]-hairpins. (C) Thermodynamic parameters of folding obtained for hairpins 1–3 from a two-state equilibrium model^{b a} Plots of folded fractions were obtained from VT-CD melting curves at 228 ± 1 nm for peptides 1–3 at 22–85 μM concentrations (accurately measured by UV-absorbance) in a phosphate buffer (15 mM, pH 6.5).^b Standard thermodynamic parameters calculated at 298 K with energies reported in kcal/mol and heat capacities in cal·mol⁻¹·deg⁻¹. Thermodynamic data are given ± standard error calculated from the standard deviations of the best-fitted CD-melt curves of [θ(T)]₂₂₈ and propagated from the initial uncertainties on measured θ.

To analyze these secondary structures in solution, several synthetic “random coils” SI-RC1–4 arguably similar to the corresponding β-strap 1–3 in length, primary sequence, and compactness were synthesized and fully characterized by variable-temperature circular dichroism (VT-CD) spectra and NMR experiments.^14d,34 The unfolded nature of these constructs SI-RC1–4 was confirmed (see linear behavior, Figure 3A) and exploited to calibrate both CD unfolded baselines and NMR chemical shift deviations (CSDs).^25b,26 The far-UV CD spectra of peptides 1–3 were recorded in phosphate buffer (with up to 20% MeOH added in rare cases, see the Experimental Section, Table 2) upon a gradual increase of temperature (0–95 °C). The reversibility of the thermal denaturations was ascertained by obtaining a quasi-identical recovery of spectral features for all constructs 1–3 upon cooling the temperature back to 0 °C. It is well known that the molar ellipticity intensities of the 228 nm exciton vary considerably even for fully folded hairpins ([θ]₂₂₈ +290 000 to +560 000°) due to minor differences in perpendicularity between indole rings,^30b,31 thus raw melting curves from [θ(T)]₂₂₈ could not be utilized to determine fractions of folding precisely. Instead, raw data were corrected by subtracting the corresponding random coil ellipticity values obtained at 95 °C [θ]_RC(95) (unfolded baseline SI-RC1–4) and normalized (per Trp pair) for direct comparison.³⁵ The corrected and normalized raw ellipticity intensities [θ]₂₂₈ for hairpins 1–3 were then fitted to a theoretical [θ(T)]₂₂₈ melt curve as a function of temperature (eq S1) using the thermodynamic Gibbs–Helmholtz equation describing a two-state model (folded/unfolded states).²⁹ The best-fitted melting curves of unfolding transition were computed for hairpins 1–3 by allowing ΔH_m, T_m, and ΔC_pF° to freely vary using a nonlinear least square fitting routine,²⁹ while using the approximation that the heat capacity is constant over the range of temperature evaluated: ΔC_pF° < 0 (Figure S2).^34,36 Plots of %-folding (χ_F) for each hairpin 1–3 as a function of temperature were obtained (Figure 3A), along with the thermal stability curves (Figure 3B), melting temperatures (T_m),³⁷ and thermodynamic parameters characterizing the unfolding transitions (Figure 3C). As shown by the goodness of the fit determined on [θ(T)]₂₂₈ (avg. R² of 0.98 and RMSD of 5,179 deg·cm²·dmol⁻¹), these melting curves indicated a temperature-induced unfolding typical from a hairpin/coil transition in agreement with a two-state model via a single transition barrier resulting from the cooperativity of fracture between the strap and loop segments. Even though the midpoint inflection of the melt curvature (at T_m) cannot be consistently observed due to the finite window of experimental temperatures (Figure 3A), both the cooperativity of the transitions and isodichroic points can be estimated from the extrapolated melts and the raw CD stacks presented in the Supporting Information. In addition to the thermal denaturation, the two-state model approximation was further validated by isothermal urea-induced denaturation titrations (0–15 M) on hairpins 2a and 3b–d (Figure S18).³⁴

Table 2.

Summary of Peptide Characterization

			m/z [M + H]+		m/z [M + Na]+
peptides	yielda (%)	HPLC %-purityb (net charge)	calcd	foundc	calcd	foundc	solubility (mM)d	$[\theta {]}_{\max }^{278\mathrm{K}}$ e	NMR samples preparationf
SI-RC1	81(a2)	88 (1)	1046.15	1046.08	1068.14	1067.94	20.0	n.a.	A1
1a	57(a2)	97 (1)	1159.27	1158.91	1181.26	1181.28	2.9	120 × 103	A1
1b	32(a2)	78 (−1)	1221.53	1221.59	1244.28	1244.11	35.3	77 × 103	B
1c		98 (2)	1681.84	1682.12	1703.83	1704.11	22.3	328 × 103	A1
SI-RC2	55(a1)	94 (1)	1370.70	1370.52	1392.68	1392.62	78.0	n.a.	A1
2a	15(a1)	96 (1)	1570.80	1570.85	1592.77	1592.92	1.2	105 × 103	C1
2b	15(a1)	95 (1)	1499.76	1500.54	1521.74	1522.54	0.8	262 × 103 (e1)	C2
SI-RC3a	91(a2)	98 (1)	927.49	927.91	949.47	949.85	25.0	n.a.	A1
SI-RC3b	83(a2)	82 (0)	905.45	905.94	927.43	927.94	10.0
3a		90 (1)	1815.08	1815.65	1836.89	1837.54	27.1	667 × 103	A1
3b		98 (2)	1852.90	1853.64			56.3	522 × 103	A1
3c		>99 (1)	1244.62	1244.62			39.5	301 × 103	A1
SI-RC4	45(a1)	71 (2)	2273.04	2273.50	2295.03	2295.45	88.0	n.a.	A1
3d	70(a1)	88 (2)	1958.97	1959.63	1980.96	1980.76	5.8	406 × 103	A2
3e	30(a1)	93 (−1)	2549.16	2549.60			0.6	158 × 103 (e2)	D
3f	35(a1)	99 (−1)	2459.11	2459.39	2481.09	2481.35	8.2	124 × 103	E

^aIsolated yields: ^(a1) of peptides obtained after semipreparative RP-HPLC purification; ^(a2) of crude peptides corrected based on purity determined by analytical HPLC.

^bDetermined by analytical RP-HPLC from peak integration detected by UV-absorbance at 220 nm from samples of lyophilized peptides.

^cDetermined by MALDI-TOF mass spectrometry.

^dKinetic solubility measured in a phosphate buffer (PB, 50 mM) at pH 7.4.

^eMaximum molar ellipticity (deg·cm²·dmol⁻¹) recorded for peptide solutions prepared in PB (15 mM, pH 6.5) unless otherwise stated: ^(e1) PB (15 mM)/CH₃OH mixture (90:10 v/v) ^(e2) (80:20 v/v).

^fNMR data set typically recorded in A1 PB (50 mM)/D₂O (90:10 v/v) or A2 PB (15 mM)/D₂O (90:10 v/v), or B PB (50 mM)/ammonium bicarbonate (100 mM)/D₂O (84:6:10 v/v), C1 PB (50 mM)/CD₃CO₂D/D₂O (85:05:10 v/v), C2 (80:10:10 v/v), D PB (50 mM)/DMSO-d⁶ (70:30 v/v), E H₂O/DMSO-d⁶ (85:15 v/v).

CD melts for hairpins 1a–c, built around an Ac-W⋯WTG-NH₂ β-strap motif, provided thermodynamic parameters representative of a relatively poor folding (ΔG_F° ~ 0). In contrast to the initial report by Andersen,^16a our results suggest that this strap is not extremely stabilizing for longer loops. Even so, the innate nucleation induced by the DDATKT and NPATGK loops was shown to affect the overall hairpin structures (1a vs 1b, ΔΔG_F° = −0.51 kcal/mol), suggesting that a DDATKT sequence might be a more favorable loop. Adding a pair of aromatic residues F(S ± 2) to the β-strap flanking the loop in 1c resulted in the most stable hairpin of this series despite a modest folding χ_F(291) of 63 ± 4% (Table 1, entry 8). In contrast, hairpin 2a stabilized by the Ac-WVW⋯WVW-NH₂ strap and derived from a truncated Trpzip9 (Table 1, entry 5 vs entry 9) revealed some remarkable folding properties (T_m (2a) > 100 °C; ΔG_F° = −2.36 kcal/mol). This strap was further evaluated by crafting a more flexible GPGTGK loop sequence in hairpin 2b. Surprisingly, the N4G and A6G substitutions largely abrogated the fold of hairpin 2b (Figure 3B). The important entropic penalty associated with this flexible loop (T·ΔΔS_F°(2b/2a) = + 3.25 kcal/mol) could not be compensated by the noncovalent interactions within the strap. Another drawback of this short β-strap is its high hydrophobicity in comparison to the longer original Trpzip sequence,^25d thus creating peptides poorly soluble in both water and organic solvents (e.g., MeOH, TFE, or DMSO).

To counterbalance this hydrophobic effect, improve aqueous solubility, and increase amphipathicity, an ionic pair was inserted at the terminal HB-site resulting in the RW(V/HW)⋯(WV)WE strap design 3. Hairpins 3a–c demonstrated some of the most interesting thermodynamic properties with high melting temperatures (62 ≤ T_m ≤ 98 °C) and sizable stability (−1.63 ≤ ΔG_F° ≤ −0.87 kcal/mol). As shown in Figure 3A, melt curves of hairpins 3a vs 3b highlighted some subtle differences in stability. Given the large heat capacity ΔC_pF° of −158 ± 3 cal·mol⁻¹·deg⁻¹) typically associated with the burial of nonpolar moieties, the V3H substitution in 3b appeared to enhance the overall hairpin stability, although being less resistant to the effect of temperature T_m(3b) ≪ T_m(3a). A truncated version of this strap in hairpin 3c also demonstrated significant stability with T_m ~ 62 °C and a ΔG_F° = −0.87 kcal/mol. Overall, hairpin 3c presented the similar stability to the slightly longer HP7 construct (Table 1, entry 3), which led us to speculate that an ionic pair at the terminal HB-site might provide a new framework for building valuable β-straps. In contrast to the folding of 2a, which is associated with an extended hydrophobic packing (highly enthalpically favorable, ΔH_F° ~ −6 kcal/mol), hairpin 3a carries a smaller enthalpic contribution (ΔH_F° ~ −4 kcal/mol), likely to be compensated by an additional conformational flexibility from the salt-bridge.³⁷ Indeed, such Coulombic interaction persistent over long distances could conceivably tame the overall hairpin fraying and afford the appropriate flexibility to achieve an optimal geometry within the neighboring EtF W/W aromatic interaction.³⁸ Taken together, our thermodynamic results imply that the RWVW⋯WVWE strap sequence of 3a was optimal in stabilizing short β-hairpins.

This strap was further evaluated on hairpins 3d–f with long loops of variable plasticity. The CD signature from the W/W exciton maxima at 228 ± 1 nm was also observed for all constructs 3d–f despite in lower intensity. As expected, the entropic cost associated with the folding of a polyglycinyl loop in 3d is relatively large in comparison to 3e (T·ΔΔS_F°(3d/3e) = −3.53 kcal/mol). Even so, hairpin 3d is extremely well folded with a T_m of ~ 52 °C and a χ_F of 88% at 18 °C (291 K). In contrast, hairpin 3e bearing a conformationally constrained loop of “random” sequence (pembrolizumab H3 loop model, see Figure 1B) proved less stable, as shown by the CD-melt in Figure 3A. To reduce both hydrophobicity and rigidity in the loop, an F9G substitution was evaluated at the L4 position of 3f. Given that the heat capacities calculated for the transitions of hairpins 3e and 3f were comparable, the release of loop strain in 3f likely contributed to enhance its folding. Overall, the CD melts revealed that short hairpins 2a, 3a, and 3b are “hyperstable” with a long plateau of thermal stability from 0 to ~50 °C. Below 50 °C, such foldings exist in a dynamic equilibrium of stable conformational states (>80% folded, vide infra NMR data), which affect the cooperativity of the transitions (continuum of discrete steps).^25b On the other hand, both hairpins bearing flexible loops 2b and 3d have a more ideal all-or-none transition than their constrained counterparts 2a and 3e. This result might be explained by the lesser change of entropy in flexible loops between folded and unfolded states.

Thermostability curves are usually applied in the analysis of protein stability to estimate both entropic and enthalpic contributions induced by mutations to the proteins’ folding free energy.⁴⁰ To rationalize the differences between β-straps 1–3 and inform their rational design, such plots of temperature dependence on unfolding free energy ΔG_F(T) were analyzed (Figure 3B).⁴⁰ The thermodynamic parameters calculated from the two-state model approximation (Figure 3C) were applied to plot these thermostability curves, which can be extrapolated outside the range of experimental temperatures. Thermostability curves have typical parabolic shapes (if ΔC_pF° ≠ 0) that intercept the temperature axis ΔG_F(T) = 0) at two different temperatures characteristic of cold and thermal denaturations (T_C and T_m). At the parabolic maxima, the temperature T* of maximal stability and free energy ΔG_F* can be determined for each hairpin (Figure 3B,C).⁴¹ Indeed, at T*, ΔS_F* = 0 resulting in ΔG_F* = ΔH_F*. The differences in enthalpy ΔΔH_F* between hairpins of the identical loop sequence (i.e., NPATGK) were therefore readily compared to estimate the net contribution of straps on the overall secondary structure’s thermostability. As shown in Figure 3C, the RWVW⋯WVWE strap performed well in hairpins 3a–c with ΔΔH_F* of −1.19 to −1.76 kcal/mol, while the Ac-WVW⋯WVW-NH₂ strap in 2a was the most stabilizing motif with a ΔΔH_F* of −2.42 kcal/mol. Another significant phenomenon observed from the thermostability curves is the parabolas’ curvatures.⁴⁹ Even if 2a and 3b are the most stable hairpins ΔG_F* of −2.50 and −1.84 kcal/mol), the narrower parabolic curvature of their thermostability curves implies that these hairpins have a small range of thermal stability. On the other hand, the wider curvatures observed in the case of 3c and 3a suggest that these hairpins absorb more efficiently heat energy for the same increase in temperature (thermostable). This phenomenon can be explained by the fact that hairpins 3a and 3c have the smallest ΔC_pF° of −53 and −106 cal·deg⁻¹·mol⁻¹, respectively, which ultimately is a consequence of having a smaller hydrophobic surface exposed to water. On the contrary, even if the V3H mutation created additional contact interactions in hairpin 3b (i.e. higher fraction folded), the narrower curvature of the thermal stability parabola confirms that the enthalpic contribution of such electrostatic interactions weakened rapidly upon the effect of heat.⁴² The aforementioned entropy held by the loop also affects the thermal stability of longer hairpins and the cooperativity of the unfolding transitions. The large entropic energy contained in the polyglycinyl loop of hairpin 3d likely contributed to its sharp thermal denaturation thus resulting in a narrow range of thermal stability. Conversely, the wider curvature of the ΔG_F* plot for 3e–f (shift of T*) demonstrated how compactness and local side-chain interactions in more conformationally constrained loops can preclude the rapid thermal denaturation. This effect culminated in the stability of hairpin 3f (slightly less rigid and hydrophobic than 3e) with the largest ΔG_F* of −3.18 kcal/mol and T_m of 46.7 ± 0.1 °C.

Structural and Conformational Analysis by NMR Spectroscopy.

To confirm the secondary structures suggested by the CD study, a careful structural assignment and conformational analysis were secured by NMR spectroscopy.⁴³ Detailed NMR studies of hairpins 1–3 at 18 °C (291 K) were carried in a mixture of phosphate buffer (pH 6.5) with D₂O, with small proportions of additives (DMSO-d₆, or CD₃OH) in rare cases, at concentrations ca. 1–20 mM with DSS as the internal standard for chemical shifts. Signals assignments were obtained on the basis of a set of ¹H and TOCSY spectra (for intraresidue connectivities), NOESY spectra (for vicinal and interstrand backbone connectivities), and HSQC spectra (for H_α to C_α connectivities and W side-chains assignments).³⁴ First, chemical shift deviations (CSDs) were calculated for all of the H_N and H_α proton resonances in each hairpin 1–3 (Figure 4).⁴⁴ Average CSDs obtained for the synthetic unfolded constructs SI-RC1–4 were in each case close to zero (avg. ∣ΔδH∣ < 0.2 ppm) indicating the absence of the secondary structure as expected for coil-like structures.^34,45 These results further validated our unfolded models used to calibrate the unfolded baselines in the CD-melt analysis. For constructs 1a,b, the overall chemical shift dispersion of H_N and H_α signals is small, and CSDs characteristic of β-hairpins (upfield in the loop and downfield in strands) are not observed (Figure 4A). Furthermore, the W(S − 2)/W(S + 2) pair in 1a was characterized by upfield shifts of 0.58 and 0.45 ppm for the H_ε3 and H_β3 signals of the “edge” W(S − 2) residue indicative of a reverse EtF arrangement characteristic of misfolding (instead of FtE interaction). In hairpin 1c, the addition of an F/F pair at the S ± 2 positions flanking the loop enabled the β-strap to actually fold⁴⁶ in a similar fashion to the [2:2]-hairpins reported by Andersen. An FtE W(S − 4)/W(S + 4) interaction was plainly observed in this case, with CSDs for W(S + 4) ΔδH_β3 and ΔδH_ε3 of −0.45 and −1.62 ppm, respectively, as well as a distinctively large upfield G14H_N proton shift (δ 6.33 pm corresponding to a CSD of −1.78 ppm from its random coil value). Overall, the backbone H_N and H_α CSDs from the Ac-W⋯WTG-NH₂ strap motif in 1c revealed a pattern consistent with a β-sheet register. The folded conformation of hairpin 1c was further supported by a set of cross-strand NOESY correlations (Figure S19).

Figure 4.

Histograms of backbone H_N, H_α, and Trp side-chain CSD values obtained at 291 K along the hairpin primary sequences. (A) In the series of strapped β-hairpins 1a–c and 2a–b. (B) In the series of strapped β-hairpins 3a–c; the arrow on the histogram points to the unusually deshielded H_N(G4) loop signal, suggesting the presence of an intramolecular H-bond as expected in a [4:6]-hairpin register. (C) In the series of strapped β-hairpins 3d–e.

Earlier studies on [2:2]-hairpins established that the extent of chemical shift splitting between glycine methylene signals in ¹H NMR spectra (ΔδH_α/α’(Gly)) can be exploited to quantify the %-folding of β-hairpins.^43a,45c,47 These two H_α/H_α’ diastereotopic protons experience a different chemical environment relative to the ratio of equilibrating conformers on the NMR time scale. Thus, a greater separation is indicative of a larger folded population. This local folding probe was notably applied by Horne and Del Valle to [4:4]/[4:6]-hairpins featuring the NPATGK reverse-turn locus, by measuring the ΔδH_α/α’(Gly) separation of G(L4).^27,48 In a series of macrocyclic peptides, the largest ΔδH_α/α’G(L4) value of 0.33 ppm was attributed to 100%-folding and used as a benchmark for the “fully” folded hairpin. Strikingly, several synthetic hairpins from our study (e.g., 1c, 2a, 3a–c) have shown even larger shifts 0.46–0.55 ppm (see Table 1). We therefore re-estimated the folded population for the loops of hairpins 1–3 based on a 0.55 ppm splitting for the G(L4) H_α/_α’ resonances as the >99%-folded ref 49. ΔδH_α/α’G(L4) values of 0.23 and 0.50 ± 0.01 ppm observed at 291 K for 1a and 1c correspond to a local folded population estimate of 42 ± 4 and 91 ± 2%, respectively. Although the poor solubility of constructs 2a–b in water limited our NMR interpretations, the folding of 2b was found minimal. Indeed, H_N and H_α CSDs in construct 2b were mostly insignificant throughout, and the local loop folding was characterized by a ΔδH_α/α’G(L4) of 0.03 ppm corresponding to 6 ± 33% of folding at 291 K (Table 1, entry 10). On the contrary, hairpin 2a was 93 ± 2%-folded and characterized by high CSDs for the tryptophans side-chain signals due to two FtE W(S ± 4) and Etf W(S ± 2) interstrand interactions. For the turn-flanking edge W(S − 2) residue, H_ε3, H_β3, and H_δ1 experienced upfield ring current shifts as large as 1.22, 0.95, and 0.60 ppm, due to aromatic shielding, whereas the edge C-terminal W(S + 4) residue displayed upfield shifts of 1.25, 0.99, and 0.44 ppm for the same proton signals (Figure 4, bottom panel). These NMR structuring shifts are similar to those reported for the HP7 and Trpzip4 hairpins, thus confirming the efficiency of the truncated β-strap.⁵⁰ The folded nature of hairpin 2a was further supported by various cross-strand and long-range NOESY correlations (Figure S20). Collectively, NMR and CD data obtained at 291 K revealed that hairpin 2a is well folded with χ_F > 90%.

As shown by the trend of thermodynamic stability observed in the CD melts, the RW(VW)⋯(WV)WE β-strap proved to be superior (e.g., 3a–c) throughout the NMR study (Figure 4B). Indeed, ¹H NMR signals for both H_N at HB sites and H_α at NHB-sites within the β-strap were highly dispersed and experienced significant downfield shifts (~1.5 ppm on average).^45a The strong FtE and EtF W/W interactions at the S ± 4 and S ± 2 positions were further rationalized by the formation of a more compact hydrophobic cluster. Distinctive NMR shifts with CSDs of −1.49 and −1.16 ppm for H_β3, −0.62 and −0.32 ppm for H_δ1, and −1.21 and −1.19 for H_ε3 observed at the respective S − 2 and S + 4 edge sites of hairpin 3a denoted that the indole rings of these tryptophans experienced a significant ring current shift. A similar range of shifts was observed for the W(S − 2) of the shorter hairpin 3c. As shown in Figure 5, strong NOESY correlations between the H_N protons of A(L2), T(L3), G(L4), and K(S + 1) secure the overall loop conformations in both 3a and 3c. Correlations between the W(S−2)H_ζ2 and both G(L4) H_α and H_N were also consistent in both hairpin spectra. Both S ± 4 and S ± 2 W/W interactions were also confirmed by a combination of long-range side-chain to side-chain NOESY correlations. NOESY correlations between the W(S + 4) H_ε3 and H_ζ3 with the R(S − 5) H_α proton suggest that the C-/N-termini fraying for this hairpin is minimal at 291 K. More importantly, we found several intense H_N─H_N NOESY correlations between N(S − 1) and K(S + 1) residues in 3a and 3c and an H_N─H_α interaction N(S − 1) to W(S + 2) in 3a supporting the formation of an interstrand hydrogen bond network in the β-strap. Finally, the NOESY correlation between the S ± 3 positions VH_N in 3a definitely established the presence of a β-sheet register within a [4:6]-hairpin (Figure 5). Overall, folded populations of 84, 93, and >99% (±2%) were estimated from the ΔδH_α/α’G(L4) values in the loops of hairpins 3c, 3a, and 3b, respectively.

Figure 5.

Structural analysis of hairpins 3a, 3c–d, and 3f by NOESY correlations establishing side-chain to side-chain interactions, the loop orientation, and the hydrogen-bonding register within β-straps^{a,b a}Selected key NOESY correlations recorded at 291 K, with edge arenes involved in EtF arene–arene interactions highlighted in light orange. ^b Hydrogens circled (in pink) in the NOESY diagrams participate in a network of intramolecular hydrogen bonds characteristic of a hairpin register.

[10:12]-hairpins 3d–f bearing the longest loops were also extensively characterized by combinations of NMR experiments. Excitingly, in both examples of flexible and rigid loops (3d and 3f), a large positive CSD pattern across the β-strap residues unambiguously established the foundation of a β-hairpin register (Figure 4C). In the case of hairpin 3d experiencing the most flexibility, both H_N and H_α of the strap residues resonated downfield (CSD of 0.30–2.15 ppm), while hydrogens of the loop residues witnessed small CSDs (<0.10 ppm). Except for Gly residues flanking the strap edges, ΔδH_α/α’G of 0 ppm was observed in the loop due to its innate flexibility and a lack of asymmetric environment. By analogy to the shorter loops, folding of 3d was estimated to be >99 ± 2% at 291 K given a ΔδH_α/α’G(S + 1) of 0.55 ppm. CSDs of −0.94 and −1.07 ppm for H_β3, −0.44 and −0.31 ppm for H_δ1, and −1.25 and −1.15 ppm for H_ε3 observed at the respective W(S − 2) and W(S + 4) of 3d were comparable to the ring current shifts observed in hairpins 3a–c, which confirmed that both W/W pairs were creating FtE arene interactions. As shown in Figure 5, NOESY correlations between R1H_α with W17H_ε3 and W2H_ε3 with W15H_β in hairpin 3d are stronger than in 3a, and the long-range interstrand correlation of W2H_α with W17H_N further supported a hairpin fold. These results suggested that strap 3 afforded enough stabilization to offset the flexibility of the polyglycinyl loop. Seemingly, even if a high plasticity typically results in slower kinetics of nucleation,⁵¹ the thermodynamics of such systems (i.e., 3d vs 3e) are more favorable to a hairpin structuration due to a more flexible strand alignment. In regards to the more strained hairpin 3e, ¹H NMR spectra at 291 K displayed overlapping and broader signals in the H_N region with small CSD values (avg. of 0.30 ppm) indicative of a rapid sampling of multiple conformer populations, which is not surprising considering that 298 K was the melting temperature of 3e determined by CD.³⁴ In stark contrast, the rigid hairpin 3f (one glycine mutation in the loop) was remarkably well folded with downfield deviations for H_N signals within the strap (CSD of 0.47–1.97 ppm) strikingly similar to 3d suggesting a fine alignment. Large CSDs of −0.99 and −1.28 ppm for H_β3, −0.67 and −0.21 ppm for H_δ1, and −1.01 ppm for both H_ε3 were also observed at the respective W(S − 2) and W(S + 4) positions. In addition, a number of medium to strong NOESY correlations within the strap were observed as exemplified by the V3H_N and V16H_N long-range interaction. Taken together, these results conclusively demon-strate that 3f is a well-structured hairpin with the stabilizing edge tryptophan side chains at residues W(S − 2) and W(S + 4) in a zipper-like motif. Along those lines, the loop %-folding in 3f was estimated to be ~55 ± 2% at 291 K based on the glycine shifts at the G(L7) position (ΔδH_α/α’ of 0.30 ppm). Collectively, structural NMR data demonstrated that the RWVW⋯WVWE strap 3 was the most efficient hairpin-folding manifold for short and long loops of varying plasticity.

CONCLUSIONS

Accessing larger macrocyclic peptides is an emerging trend to build mini protein-like structures thermodynamically stable in water.⁵² However, the identification of β-hairpins that closely mimic the native fold of protein and antibody loops remain challenging due to the large entropic penalty associated with their conformational folding. In this study, a de novo design of short β-straps featuring a combination of ionic interaction with a zipper motif provided a benchmark manifold to fold hairpin architectures with long loops of variable plasticity (up to 10-mer). Several novel β-strap motifs of minimalist length were crafted by combining β-strand + β-cap to generate hairpins containing loops of variable length, plasticity, and turn propensity. Although strap Ac-WVW⋯WVW-NH₂ 2 was found to be highly stabilizing, the hydrophobic cluster modeled upon the Trpzip molecules (large ΔC_pF° of −159 ± 17 cal·deg⁻¹·mol⁻¹) leads to a poor and unpractical solubility in water. This motif was further optimized in strap RW(VW)⋯(WV)WE 3 by combining the folding driving force associated with the W/W cluster hydrophobicity to the long-distance stabilizing features of a salt-bridge^38,39 positioned at hairpins’ termini. The CD-exciton signature at 228 nm from two W/W EtF interactions as well as the large NMR upfield CSDs (1.16–1.49 ppm and 1.19–1.21 ppm for H_β3 and H_ε3) and long-range NOESY correlations reported for 3a conclusively demonstrated that well-folded hairpins were formed (89–93% fold).

Finally, the strong enthalpic contribution from β-strap 3 (ΔΔH_F*) was able to remarkably offset the entropic penalty associated with the wealth of conformations generated by longer 10-mer loops.¹⁷ To our knowledge, this is the first observation of a short β-strap (8-mer long) able to structure loops longer than the strap itself (up to 88% fold, T_m of 25–52 °C). Strap 3 largely outperforms the technology previously developed by Andersen for long-loop closure (16-mer β-cap-β-strand).¹⁶ Collectively, these results suggest that the specific site positioning and combination of a terminal salt-bridge with van der Walls and CH─π hydrogen bond interactions from the two EtF W/W pairs within the strap are an efficient hairpin-folding manifold. Our results suggest that β-strap 3 might found useful applications to fold hairpin architectures with long loops of varying plasticity and random sequences. Ultimately, we anticipate that the present study will provide the foundation for the development of long hairpin foldings as an attractive alternative to macrocyclic peptides for the mimicry of proteins and antibodies.^3,53

EXPERIMENTAL SECTION

General Information.

All reagents, Fmoc-amino acids, and resins used in the present paper were purchased from Chemimpex and Millipore Sigma. All bulk solvents were acquired from Fischer Scientific. Peptides 1c, 3a–c were purchased from Peptide 2.0 Inc. Peptides, 1a–b, 2a–b, 3c–d, and SI-RC1–2 were synthesized using a standard automatized Fmoc-SPPS technique (solid-phase peptide synthesis) on a Protein Technologies PS-3 peptide synthesizer. Syntheses were accomplished either on a Rink amide resin of medium loading (0.27 mmol/g) or Fmoc-Glu-Wang resin (0.4 mequiv/g). Random coil model peptides SI-RC3a/b and SI-RC4 were synthesized on a manual peptide synthesizer, using a Fmoc-FPPS (fast-parallel peptide synthesis) technique, on a Rink amide resin of medium loading (0.27 mmol/g) and Fmoc-Glu-Wang resin (0.4 mequiv/g). Our synthetic hairpin peptides have a net charge of −1, +1, or + 2 (from loop sequence) to enhance water solubility and prevent aggregation. Structural assignments were made by a set of NMR spectra including TOCSY, NOESY, HSQC, and HMBC experiments recorded on a Varian Mercury500 (500 MHz) spectrometer and processed using the Vnmrj 4.2 software.

Peptide Synthesis.

General Procedure for Peptide Synthesis Via Fmoc-SPPS.

Syntheses were carried out at room temperature in anhydrous DMF using the resins described above (200–300 mg, 1.0 equiv, as reported below for each peptide) by successive iterations of deprotection/coupling. Vials of each Fmoc-protected α-amino acid (4.0 equiv) were prepared with HBTU (4.4 equiv) and HOBt (4.4 equiv) neat. For each iteration, the deprotection/coupling sequence entails (1) wash of the resin with DMF (3 × 5.0 mL) for 0.5 min each, (2) Fmoc-deprotection run twice using an excess of piperidine in DMF (20% v/v, 5.0 mL) for 5 min each, (3) wash with DMF (6 × 5.0 mL) for 0.5 min each, (4) the cocktail from the entire vial (described above) was dissolved with N-methylmorpholine in DMF (3.0 mL, 4 M) and added for a 40 min coupling, (5) wash with DMF (3 × 5.0 mL) for 0.5 min each, finalized the sequence. After the final N-terminal coupling, a final Fmoc-deprotection was achieved (step 2). The resulting peptides attached to the resin were washed with CH₂Cl₂ (2 × 10 mL), and the resin was dried under vacuum before storage under an argon atmosphere at −78 °C until cleavage.

General Procedure for Peptide Synthesis Via Fmoc-FPPS.

Syntheses were carried out at room temperature in anhydrous DMF using the resins described above (200–300 mg, 1.0 equiv, as reported below for each peptide) by successive iterations of deprotection/coupling. For each iteration, the deprotection/coupling sequence entails (1) resin swelling in DMF (10 mL) for 10 min, (2) Fmoc-deprotection run twice using an excess of piperidine in DMF (20% v/v, 5.0 mL) for 2 min each, (3) wash with DMF (3 × 5.0 mL) for 0.5 min after each deprotection, (4) coupling reaction run twice for 5 min each by adding in order Fmoc-amino acid stock solution (5 equiv in 5 mL DMF), HATU stock solution (5 equiv in 5.0 mL DMF), and DIEA (10 equiv), and (5) wash with DMF (3 × 5.0 mL) for 5 min after each coupling reaction. After the final N-terminal coupling, a final Fmoc-deprotection was achieved (step 2). The resulting peptides attached to the resin were washed with CH₂Cl₂ (2 × 10 mL), and the resin was dried under vacuum before storage under an argon atmosphere at −78 °C until cleavage.

General Procedure for Peptide Deprotection and Cleavage from Resins.

The same protocol of simultaneous resin cleavage and sidechains’ deprotection was applied to both resins. The dried resin was suspended in a cleavage cocktail (TFA/thioanisole/EDT/anisole, 90:5:3:2 v/v, 1.0 mL per 20 mg of resin) and shaken for 1.5 h at RT. The mixture was filtered to remove the resin and the mother liquor was evaporated on a rotary evaporator. The crude peptides (~250 mg) were precipitated in cold ether (40 mL), then centrifuged, and washed with cold ether (3 × 40 mL). The resulting crude peptides were solubilized in water and lyophilized before being stored as dry powders under an argon atmosphere at −78 °C.

All lyophilized crude peptides were analyzed by analytical reverse-phase high-performance liquid chromatography (RP-HPLC) Hitachi L-7000 series equipped with an XBridge BEH C₁₈ column (130 Å, 10 μm, 4.6 mm × 250 mm). HPLC grade acetonitrile and deionized water, each containing 0.1% trifluoracetic acid, were used for analytical and semipreparative RP-HPLC. Each peptide was analyzed using a gradient from 10 to 50% of acetonitrile over 30 min with a flow rate of 1.0 mL/min at room temperature and a detection at 220 nm. Yields refer to chromatographically and spectroscopically pure compounds, unless otherwise stated. If required, peptides were purified with a Hitachi L-7000 series HPLC equipped with a semipreparative XBridge BEH C₁₈ column (130 Å, 10 μm, 10 mm × 250 mm) stationary phase by scaling up the analytical conditions. The pure fraction of peptides was collected and analyzed by analytical RP-HPLC and mass spectrometry using a Microflex LRF matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) instrument.

Nuclear Magnetic Resonance Data Collection.

NMR samples were prepared by dissolving the freeze-dried peptide (~ 1–2 mg) in a mixture of phosphate buffer (50 mM, pH 6.5) and D₂O (9:1, v/v) unless otherwise stated, using 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as the internal standard for chemical shifts (0 ppm). Samples were prepared in a range of 3–10 mM of the peptide for ¹H NMR, TOCSY (mixing time 80 ms), NOESY (mixing time 200 ms), and ¹H─¹³C HSQC experiments and in a range of 10–20 mM to record ¹H─¹⁵N HSQC and HMBC spectra (with the exception of SI-RC1, 1a, and 1b samples, which were prepared in a range of 10–70 mM range). A PRESAT experiment was used to suppress the H₂O solvent signal, to record initial ¹H NMR spectra. All spectra were recorded at 291 K (18 °C) on a Varian Mercury500 (500 MHz) spectrometer and processed using the Vnmrj 4.2 software. Signals assignments were obtained on the basis of a set of ¹H and TOCSY spectra (for intraresidue connectivities), NOESY spectra (for vicinal and interstrand backbone connectivities), and HSQC spectra (for H_α to C_α connectivities and W side-chains assignments).

Chemical Shift Deviation (CSDs).

Chemical shift deviations (or CSD) have been calculated based on a simulated unfolded reference that represents the predicted (calcd.) chemical shift values along the random coil sequence, based on the established work from Poulsen and others.⁴⁴ The calcd. random coils’ chemical shifts for the backbone H_α and H_N protons were obtained from the online software available at: https://spin.niddk.nih.gov/bax/nmrserver/Poulsen_rc_CS/. The random coil calcd. δ values were validated by comparison to the experimental values obtained for SI-RC1–4 (Δ∣CSD∣ ≤ 0.20 ppm). CSDs for each hairpin construct were then calculated according to CSD(H) = δH_exp. −δH_calcd. and reported in tables (Supporting Information Sections 2-4). Furthermore, CSDs related to the side chains of tryptophan residues have been calculated from the list of known chemical shifts reported on the BMRB database (Biological Magnetic Resonance Bank) for Trp in random coils: http://www.bmrb.wisc.edu/ref_info/pentapeptide.tbl. This list of chemical shifts (Trp random coils) has been previously validated by Andersen et al.²⁶

Far-UV Circular Dichroism (CD) Spectroscopy.

Peptide solutions were prepared at the 20–100 μM concentration range in phosphate buffer (15 mM, pH 6.5), with the addition of MeOH (up to 50% v/v) if required to increase solubility. Each sample concentration was determined accurately by measuring the solution UV-absorbance using a JASCO V-670 spectrophotometer based on the combined molar absorptivity of Trp and Tyr residues at 280 nm (ε₂₈₀ = 5580 M⁻¹·cm⁻¹ per Trp, ε₂₈₀ = 1280 M⁻¹·cm⁻¹ per Tyr). CD spectra were recorded on a JASCO J-810 Spectropolarimeter with a temperature controller module JASCO PFD-425S. In brief, raw CD data were recorded in mdeg from 185 to 270 nm every 0.1 nm, and CD spectra of the blank solutions were subtracted for baseline correction. Spectra were smoothed with the baseline set to zero between 260 and 270 nm and the ellipticity scale was converted into molar ellipticity (deg.cm².dmol⁻¹) using SpectraGryph 1.2.⁵⁴ To obtain melting curves representing peptide unfolding transitions, variable-temperature CD experiments were recorded. For SI-RC1, 1a, 1b, and 2b, VT-CD spectra were manually recorded every 5 °C from 5 to 95 °C with a 10 min stabilization between each temperature. CD spectra of the blank solution were recorded similarly and subtracted to the peptide spectra at each temperature. Molar ellipticity values at 228 nm (Trp exciton band) were plotted to obtain experimental melting curves. For SI-RC2–4, 1c, 2a, and 3a–e, the temperature ramp was controlled automatically from 5 to 95 °C with a ramp speed of 0.75 °C/min. Intensity values at 228 nm were recorded by the instrument every 0.1 °C to obtain the desired raw melting curves. Likewise, intensities for the corresponding blank solutions were recorded and subtracted to afford the experimental melting curves’ output. These thermal denaturation curves were reported in molar ellipticity (deg·cm²·dmol⁻¹) and normalized against the number of tryptophan pairs present in the peptide construct ([θ(T)]₂₂₈/pTrp) for comparison (see Figure S2). Finally, a nonlinear least square fitting routine program has been developed in Origin 9.0 (Originlab Corporation) to compute the best-fitted melting curves to the raw [θ(T)]₂₂₈ data by using the thermodynamic Gibbs–Helmholtz equation of free-energy variation for a two-state model (eq S2). A detailed description for obtaining the %-folding best-fitted curves, all calculations, ΔG°, ΔH°, ΔS°, ΔC_p°, and T_m thermodynamic values is given in the Supporting Information Section 1.