Interplay among side chain sequence, backbone composition, and residue rigidification in polypeptide folding and assembly

Abstract

The extent to which polypeptide conformation depends on side-chain composition and sequence has been widely studied, but less is known about the importance of maintaining an α-amino acid backbone. Here, we examine a series of peptides with backbones that feature different repeating patterns of α- and β-amino acid residues but an invariant side-chain sequence. In the pure α-backbone, this sequence corresponds to the previously studied peptide GCN4-pLI, which forms a very stable four-helix bundle quaternary structure. Physical characterization in solution and crystallographic structure determination show that a variety of α/β-peptide backbones can adopt sequence-encoded quaternary structures similar to that of the α prototype. There is a loss in helix bundle stability upon β-residue incorporation; however, stability of the quaternary structure is not a simple function of β-residue content. We find that cyclically constrained β-amino acid residues can stabilize the folds of α/β-peptide GCN4-pLI analogues and restore quaternary structure formation to backbones that are predominantly unfolded in the absence of cyclic residues. Our results show a surprising degree of plasticity in terms of the backbone compositions that can manifest the structural information encoded in a sequence of amino acid side chains. These findings offer a framework for the design of nonnatural oligomers that mimic the structural and functional properties of proteins.

The remarkable relationships among subunit sequence, folding pattern, and function in proteins and nucleic acids are fundamental to the existence of life, and the development of a detailed understanding of the interplay among these molecular properties remains a significant challenge (1, 2). The complex behavior of proteins and nucleic acids has motivated a wide range of research, which has broadened in recent years to encompass unnatural oligomers that display biopolymer-like conformational behavior and even biopolymer-like functions. Such efforts have included the study of RNA analogues based on pyranose rather than furanose (3), peptides based on β- or γ-amino acids rather than α-amino acids (4, 5), and systems that diverge more profoundly from biological precedents (6, 7). As the folding propensities of new backbone elements have been elucidated, the first steps have been taken to blend natural and unnatural subunits and thereby create folded oligomers with heterogeneous backbones and distinctive behavior (8–10).

The design of synthetic oligomers that display specific structures and functions requires an understanding of sequence → structure → function relationships among unnatural backbones. Past work on the folding behavior of peptides and proteins indicates that limited backbone modification in a given sequence, including replacement of one or two α-residues with β-residues, can be accomplished without significantly altering native structure or function (11–13). More recently, we have shown that a particular type of nonnatural backbone, composed of a mixture of α- and β-amino acid residues in an ααβαααβ pattern, can partially mimic the folding and self-assembly behavior of parent α-peptide side-chain sequences (14, 15). This observation raises an interesting question: How general is the ability of nonnatural peptide backbones to manifest the higher-order structure encoded in a particular sequence of amino acid side chains? This question has previously been explored conceptually. In pioneering work on the development of the simple exact lattice (“beads on a string”) model of protein folding, Dill et al. (16) hypothesized that nonprotein oligomers with appropriately designed sequences should be able to manifest protein-like properties.

Here, we probe the relationship between backbone composition and sequence-encoded folding in the context of different α/β amino acid combinations. We compare the folding and self-assembly behavior, in solution and crystals, of several 33-residue peptides that contain a side-chain sequence known to induce four-helix bundle formation when projected from a purely α-residue backbone. These “α/β-peptides” contain various regular combinations of α- and β-amino acid residues. Our results show that the ability of heterogeneous α/β backbones to form a helix bundle quaternary structure, as dictated by a specific sequence of side chains, is a general feature of diverse α/β backbone patterns, with intriguing variations in assembly mode and propensity among particular backbones. In addition, we find that quaternary structure stability can be enhanced based on fundamental understanding of β-residue conformational propensities, by taking advantage of useful avenues for conformational preorganization that are available for these unnatural subunits.

Results and Discussion

Experimental Design.

α-Helical coiled-coil assemblies constitute one of the most intensely studied families of protein quaternary structure and have proven excellent model systems for recapitulating the folding complexity of proteins in synthetically tractable peptides (17, 18). We selected a well studied sequence as the basis for our studies. GCN4-pLI (1, Fig. 1A) is a 33-residue α-peptide derived from the dimerization domain of the yeast transcriptional regulator GCN4 (19–21). GCN4-pLI forms a very stable four-helix bundle (16 kDa) and features a canonical heptad residue repeat pattern (abcdefg), with hydrophobic Leu and Ile residues occupying most a and d positions, respectively (Fig. 1B). The a and d residue positions align along one face of the α-helix and define the hydrophobic core of the helix bundle assembly. We previously described α/β-peptide 2, an analogue of GCN4-pLI with an ααβαααβ residue repeat (Fig. 1A) (14). Each α → β substitution in 2 consists of a β³-amino acid bearing the side chain of the replaced α-residue in the GCN4-pLI sequence (Fig. 1C). This substitution strategy leads to an oligomer with the same sequence of side chains as 1 but a different backbone. α/β-Peptide 2 adopts a parallel four-helix bundle fold in the crystalline state, analogous to that of GCN4-pLI α-peptide 1; however, the behaviors of 1 and 2 differ in solution, with the α-peptide self-assembling to a tetramer and the α/β-peptide forming a trimer (14). These variations in stoichiometry complicate the interpretation of observed differences in helix bundle stability and led us to wonder whether the loss in specificity for the sequence-encoded oligomerization state was a general consequence of extensive backbone modification. We therefore prepared and tested a series of derivatives with other regular α → β³ replacement patterns that maintain the GCN4-pLI side-chain sequence.

Fig. 1.

Chemical structures and helical wheel diagrams of 1–9. (A) Primary sequences of α-peptide 1 and α/β-peptides 2–9, sorted according to α/β backbone pattern. Bold letters indicate hydrophobic core residues in the GCN4-pLI sequence. Colored circles indicate sequence positions occupied by β-residues, cyan for β³-residues and orange for cyclic β-residues. (B) Helical wheel diagram of 1 with hydrophobic core residues indicated. (C) Structures of an α-amino acid, a β³-amino acid, and the cyclic β-amino acids ACPC (X) and APC (Z). (D) Helical wheel diagrams of the α/β residue patterns of 2–9 based on a heptad repeat. Each circle represents a residue and is colored by residue type, yellow for α-residues, cyan for β³-residues, and orange for cyclic β-residues. Bold circles indicate hydrophobic core positions.

Previously uncharacterized α/β-peptides 3–6 each bear the side-chain sequence of GCN4-pLI with different patterns of α → β³ replacement (Fig. 1): αααβ, ααβ, ααβαβαβ, and αβ, respectively. Among these oligomers, only 5 features a β substitution pattern (ααβαβαβ) that follows the heptad repeat of the α-helix. α/β-Peptide 2 also has a substitution pattern that follows the heptad repeat, which in this case confines the β-residues to the periphery of the helix bundle quaternary structure (14). For 3, 4, and 6, formation of an α-helix-like conformation, presumably a prerequisite for self-association, would cause the β-residues to spiral around the helix axis (Fig. 1D). α/β-Peptides 3–6 all have multiple β-residue substitutions at hydrophobic core positions in the GCN4-pLI sequence in contrast to α/β-peptide 2, which has no core substitutions.

We obtained crystals of 3 and 4 by hanging-drop vapor diffusion, and diffraction data were collected and solved to 2.0 Å and 1.8 Å, respectively [supporting information (SI) Table S1]. Each α/β-peptide adopts a parallel four-helix bundle quaternary structure that is quite similar to those previously observed for GCN4-pLI (1) and α/β derivative 2 (Fig. 2A). Individual helices formed by the different α/β backbones of 3 and 4 show significant homology to an α-helix (Fig. 2B, Table S2). Helical folds adopted by both the αααβ backbone of 3 and the ααβ backbone of 4 display the i → i+4 hydrogen bonding pattern of an α-helix, and C_α atoms of α-residues in each of the two new structures overlay well with the corresponding α-residues in GCN4-pLI (1). The local conformations adopted by α- and β-residues in the helices do not vary over different α/β backbone compositions; backbone dihedral angles for α-residues in 2, 3, and 4 are tightly clustered in the α-helical region of the Ramachandran plot, and the β³-residues show significant internal consistency in their backbone torsional preferences (Table S3). Side-chain rotamer preferences among β-residues in 2–4 do not appear to be significantly affected by the local change in backbone chemical composition (Fig. S1).

Fig. 2.

Comparison of the GCN4-pLI side chain sequence on four different backbone patterns. (A) Helix bundle quaternary structures of each derivative with yellow and blue indicating α- and β³-residues, respectively. (B) Helical secondary structures of each α/β-peptide 2–4 (red) compared with that of α-peptide 1 (yellow); the overlays and accompanying RMSD values are based on C_α atoms in shared α-residues. The coordinates for 1 (PDB: 1GCL) (21) and 2 (PDB: 2OXK) (14) were obtained from previously published structures.

The packing of hydrophobic core residues in GCN4-pLI is a key component in specifying the structure, oligomerization state, and stability of the helix bundle (21, 22). The occurrence of β³-residues in the hydrophobic cores of tetramers formed by 3 and 4 does not substantially alter the helix bundle quaternary structure relative to α-peptide prototype 1. For some core β-residues (β³-hLeu₂₃ of 3 and β³-hLeu₁₆ of 4), the packing is comparable with that of the α-residue at the equivalent position in GCN4-pLI (Fig. S2). At other positions (β³-hIle₁₉ of both 3 and 4), the packing is less efficient, leading to a small cavity in the core of the helix bundle (Fig. S2). This imperfect core packing may contribute to the reduced stability of 3 and 4 relative to α-peptide 1 and α/β-peptide 2 as determined by the measurements detailed below.

We compared the folding/assembly behavior of 1–6 in aqueous solution (Fig. 3A and Table 1). Analytical ultracentrifugation (AU) suggests that 3 and 4 form discrete tetrameric assemblies (200 μM; 25°C), which is consistent with the self-association stoichiometry of each α/β-peptide in the crystalline state. In contrast, α/β-peptide 5, which has higher β-residue content than 3 or 4, does not appear to self-associate in a well defined stoichiometry at 200 μM. In light of the self-association established for 2–4 by AU, CD data suggest that α-helix-like secondary structure in these α/β-peptides gives rise to an intense CD minimum at ≈206 nm (Fig. 3A). This CD signature is similar to reported spectra for helical α/β-peptides with a 1:1 backbone alternation of α- and β-residues (23). α/β-Peptide 6 appears to be poorly folded, as indicated by a low-intensity CD signature that is independent of concentration and temperature. CD spectra of 6, along with high-temperature CD data obtained during the course of thermal denaturations of 2–5, suggest that unfolded α/β-peptides are generally characterized by a less-intense CD minimum ([θ] ∼6,000 to -10,000 deg cm² dmol⁻¹ residue⁻¹) that is slightly blue shifted (λ_min ≈ 202 nm) with respect to the folded state. α/β-Peptides 3 and 4 both display cooperative thermal unfolding transitions (Fig. S3) with midpoints (T_m) at ≈77° and 82°C, respectively, at 100 μM concentration. The T_m values vary with α/β-peptide concentration, as expected if the helicity detected by CD develops concomitantly with self-assembly (Table 1). Coupling of helix formation and self-assembly is well known for 1 and other α-peptides that form helix bundles (19, 21).

Fig. 3.

CD spectra of 1–9. (A) CD spectra for GCN4-pLI α-peptide 1 and α/β-peptides 2–6 generated from simple α → β³ substitution. (B) CD spectra for GCN4-pLI α/β-peptide derivatives bearing cyclic β-residues. The colors of the spectra in B match the corresponding acyclic β-residue derivatives in A. All spectra were acquired for 100 μM α-peptide or α/β-peptide in 10 mM NaOAc (pH 4.6).

Table 1.

Summary of CD and AU data for 1-9

Peptide	Backbone pattern	Cyclic β-residues	[θ]min, deg cm2 dmol−1 res−1*	Tm, °C†		Nassoc‡
				100 μM	25 μM
1	(α)	−	−30,000	>100	n.d.	4
2	(ααβαααβ)	−	−39,000	>100	>100	3
3	(αααβ)	−	−34,000	77	63	4
4	(ααβ)	−	−42,000	82	73	4
5	(ααβαβαβ)	−	−22,000	44	–	(1,3)
6	(αβ)	−	−5,000	–	–	n.d.
7	(ααβ)	+	−41,000	>100	>100	4
8	(ααβαβαβ)	+	−32,000	>100	>100	4§
9	(αβ)	+	−30,000	80	70	(1,5)

*Minimum molar ellipticity observed in the CD spectrum of a 100 μM sample in 10 mM NaOAc (pH 4.6) at 25°C.

^†Thermal unfolding transitions observed by CD of peptide at the indicated concentration in 10 mM NaOAc (pH 4.6).

^‡Apparent association state determined by AU of a 200 μM sample in 10 mM NaOAc (pH 4.6), 150 mM NaCl; numbers in parentheses indicate a mixture of two species.

^§Data acquired for a 100 μM sample in 10 mM NaOAc (pH 4.6); higher concentration and the addition of 150 mM NaCl led to aggregation.

Overall, the crystal structures of 2–4 and the solution data for 2–6 demonstrate that a variety of different α/β backbone patterns can manifest the helix bundle quaternary structure encoded in the GCN4-pLI side-chain sequence. The ααβαααβ, αααβ, and ααβ patterns of 2, 3 and 4, respectively, each exhibit a folded structure reminiscent of that displayed by α-peptide 1. Comparisons among 1–6 in solution show a general decline in quaternary structure stability upon β-residue incorporation. Quaternary structure stability is not, however, a simple function of β-residue content: 4, with 11 β-residues, forms a four-helix bundle that has a higher T_m than does the four-helix bundle formed by 3, with 8 β-residues. The peptides containing the largest number of β³-residues, 5 and 6, failed to adopt stable folds in solution. One possible explanation for the poor conformational stability of the peptides with high β-residue content is that the helices formed by these backbones are too different from an α-helix to exhibit the quaternary structure encoded by the GCN4-pLI side-chain sequence in a pure α-residue context. Given the remarkable structural homology observed among 1–4, however, we favor an alternative hypothesis. We propose that the differences in quaternary structural stability among 1–6 are strongly influenced by an entropic penalty associated with the larger number of torsional restrictions necessary for helix formation by a β³-residue relative to an α-residue. This proposal is consistent with differences noted for formation of isolated helices by pure α and pure β³ backbones (24). We tested the residue flexibility hypothesis with the experiments described below.

β³ → Cyclic β-Residue Substitutions.

If the loss of quaternary structural stability that accompanies α → β³-residue replacements among GCN4-pLI analogues 2–6 arises at least in part from increased torsional freedom in the backbone, then it should be possible to recover stability by incorporation of appropriately rigidified β-residues. Previous work with β-peptides (25, 26) and 1:1 α/β-peptides (27, 28) has shown that cyclic β-residues can enhance helicity. The cyclic constraint (e.g., residues X and Z, Fig. 1C) restricts rotation about the C_α–C_β backbone bond of a β-residue. Theoretical studies of protein-folding energetics suggest that decreasing residue configurational entropy should be an effective general strategy for improving conformational stability in any natural or nonnatural oligomer backbone (29). With the above considerations in mind, we prepared and characterized α/β-peptides 7, 8, and 9, derivatives of 4, 5, and 6 in which a subset of β³-residues is replaced with cyclically constrained analogues (Fig. 1).

We selected β-amino acids with a five-membered ring constraint, such as (S,S)-trans-2-aminocyclopentane carboxylic acid (ACPC; X in Fig. 1), for incorporation into the GCN4-pLI α/β-peptides. This particular cyclic constraint has been shown to promote helical folding in peptides with a 1:1 α/β pattern (27, 28, 30). The backbone dihedral angles adopted by ACPC residues in β- and α/β-peptide crystal structures (26, 27, 30, 31) are similar to those seen for the β³-residues in the crystal structures of 2, 3, and 4. A pyrrolidine-based analogue of ACPC (Z in Fig. 1) was incorporated into 7–9 at sequence positions that originally (i.e., in 4–6) contained β³-residues with a cationic side chain in an effort to minimize deviation from the charge distribution of the original side-chain sequence. Visual inspection of the structures of 3 and 4 suggested that core packing of the four-helix bundle may be improved by replacement of β³-hIle₁₉ with ACPC.

The data obtained for 7, 8, and 9 in solution support our hypothesis that replacing β³-residues with cyclic β-residues can improve helix bundle stability (Table 1). Each peptide containing cyclic residues displays an intense CD minimum near 206 nm similar to those observed for the well folded α/β-peptides containing only acyclic (β³) residues (Fig. 3B). AU analysis indicates that 7–9 all self-associate in aqueous solution, and variable-temperature CD analysis suggests that the assemblies formed by 7–9 are significantly more stable than the assemblies formed by more flexible counterparts 4–6 (Table 1). α/β-Peptide 7 forms a discrete tetramer in solution, which parallels the assembly behavior of flexible analogue 4. α/β-Peptide 8 forms a tetramer, in contrast to the less-specific self-assembly behavior of more flexible analogue 5; however, 8 shows nonspecific high-order aggregation under certain experimental conditions (SI Text). The AU data for 9 can be fit to an equilibrium between monomer and 5-mer. Although this result constitutes a departure from the tetrameric assembly of α prototype 1, the folding behavior of 9 nevertheless represents an improvement relative to flexible analogue 6, for which CD data indicate little or no helical structure.

The use of cyclic β-residues in 7–9 necessarily involves departure from the side-chain sequence that is common among 1–6. Having established that cyclic residue incorporation can improve fold stability in α/β-peptide analogues of GCN4-pLI, we sought to determine the structural consequences of the sequence alterations. We collected diffraction data on crystals of 7 and solved the structure to 2.2 Å resolution (Fig. 4 and Table S1). α/β-Peptide 7 has the same ααβ backbone pattern as 4, and these two oligomers differ only in six substitutions of a β³-residue for a cyclic β-residue. The unit cell dimensions and crystal symmetry suggested that crystals of 7 were isomorphic with those of 4; however, the refined structure of 7 showed some subtle but important differences. The overlay of the helix bundles of 4 and 7 is rather poor (backbone RMSD = 1.15 Å for residues 2–30), considering the high sequence homology between these oligomers (Fig. 4B). The divergence between the structures is localized to the C-terminal half of the sequence; the N-terminal segment shows excellent overlap (backbone RMSD = 0.28 Å for residues 2–17).

Fig. 4.

Comparison of the crystal structures of α/β-peptides 4 and 7. (A) Four-helix bundle quaternary structure of α/β-peptide 7. (B) Overlay of the helical backbone of 7 (pink) with that of 4 (blue). (C–E) Comparison of an acyclic β³-residue in 4 to the cyclic β-residue at the same position in 7: β³-hGlu₁₀ in 4 (C); β-ACPC₁₀ in 7 (D); overlay of C and D (E). (F and G) Top view of the hydrophobic core packing at residues 22 and 23 in 4 (F) and 7 (G).

Inspection of the backbone conformations of the three cyclic β-residues in the N-terminal segment of 7 indicates that the rigidified subunits effectively mimic the conformation of the flexible β³-residues in 4 (Fig. 4 C–E). The structural differences between 4 and 7 arise not from changes in helical secondary structure but, rather, from a tighter superhelical winding in the C-terminal region of the helix bundle. The source of this tighter winding appears to be the substitution at residue 22, β³-hGlu → ACPC. Glu₂₂ occupies an interface g position in GCN4-pLI (1), and β³-hGlu₂₂ in 4 adopts a similar position (Fig. 4F). In contrast, ACPC₂₂ in α/β-peptide 7 (Fig. 4G) occupies a position in the hydrophobic core that is occupied by Leu₂₃ in 4 and GCN4-pLI oligomers 1–3. This change leads to a break in the heptad repeat of hydrophobic core residues along the GCN4-pLI sequence (SI Text and Fig. S4). Such discontinuities are well documented in natural coiled-coil structures, and the type seen in the structure of 7 is an example of a “helical stammer.” (32) Here, the stammer appears to result, at least in part, from replacing a negatively charged Glu side chain in the parent GCN4-pLI sequence with a hydrophobic ACPC side chain (SI Text).

Overall, the significant and consistent improvements in fold stability observed for oligomers 7–9 relative to more flexible analogues 4–6 show that proper choice and placement of constrained β-residues can result in helix bundle quaternary structures with stabilities and architectures approaching those of an α-peptide prototype. This trend supports the hypothesis that the increased number of low-energy torsions in a β³-residue relative to an α-residue can lead to diminished conformational stability in α/β-peptides relative to analogous α-peptides but that the rigidification strategies possible with β-residues can allow recovery of that lost stability. Subtle differences between the crystal structures of 7 and 4 as well as deviations from tetrameric association of 8 (under some conditions) and 9, however, underscore the importance of primary sequence in dictating folded structure. Cyclic β-residues can rigidify secondary structure and stabilize quaternary structure, but the loss of sequence information resulting from side-chain modification may affect the ability of an unnatural peptide backbone to reproduce sequence-encoded folding behavior displayed by an α-peptide prototype.

Implications for Protein Folding.

The results here provide direct experimental evidence bearing on hypotheses arising from two distinct theoretical treatments of protein folding. A key tenet of the beads on a string protein folding model described by Dill et al. (16) is that the relationship between side chain sequence and folding is independent of backbone composition, and this model predicts that nonprotein backbones should be capable of manifesting protein-like structure. The results for our α/β-peptide analogues of GCN4-pLI bear out this hypothesis to some extent, but an energetic penalty (presumably entropic) for β-residue incorporation on quaternary structure stability is apparent. Interestingly, the loss of quaternary structure stability accompanying α → β-residue substitution is predicted from a key principle of protein folding, the relationship between fold stability and backbone configurational entropy of the unfolded state. The spin glass theory of Wolynes suggests that per-monomer configurational entropy plays an important role in folding and predicts that backbone rigidification should promote folding and higher-order structure in both natural and nonnatural oligomers (29, 33). The theoretical role of backbone rigidity in protein stability is supported by experimental examples such as a study by Matthews et al. (34) that demonstrated that a single Ala → Pro substitution at a solvent-exposed position in T4 lysozyme improves stability without impacting structure or enzymatic activity. Our findings regarding the consequences of cyclic β-residue incorporation on the fold stability of GCN4-pLI α/β-peptides suggest that, analogous to the case for proteins, controlling backbone configurational entropy may prove to be a general strategy for promoting tertiary and quaternary structure in oligomers with mixed α/β backbones. It is noteworthy that cyclic backbone constraint of the C_α–C_β torsion of β-residues, unlike Xxx → Pro substitution, does not preclude the participation of amide protons in hydrogen bonding.

Conclusions.

We have demonstrated folding and self-assembly in diverse α/β-peptide backbones displaying the primary side-chain sequence of an α-peptide prototype. In three of the oligomer backbones, ααβαααβ, αααβ, and ααβ, systematic α → β³ substitution can lead to faithful mimicry of the quaternary structure encoded by the prototype sequence. In backbone patterns with high β-residue content, ααβ, ααβαβαβ, and αβ, coupled folding and self-assembly can be dramatically enhanced by replacing a subset of acyclic β³-residues with cyclically constrained analogues. Overall, these results reveal a surprising degree of plasticity with respect to the backbone compositions that can recapitulate the quaternary structure encoded by the GCN4-pLI side-chain sequence; however, it is important to note that α → β substitutions can impact both conformational preferences and conformational stability. The relationship between fold stability and backbone composition is complex, and we show here only what is possible to achieve by relatively straightforward modification of a prototype sequence. Our results indicate that the particular side chains selected for substitution with cyclically constrained β-residues are critical in determining how effectively those modifications stabilize a target folded structure. Recent function-oriented studies with α/β-peptides based on a BH3 domain have shown that the properties of α/β-peptides prepared by simple sequence-based design, of the type illustrated here, depend on the frame shift of the backbone repeat within a given sequence (15). Continuing exploration of sequence-based backbone modification will lead to a better understanding of the relationships among backbone composition, side chain sequence, and folding, which should ultimately enable the design of nonnatural oligomers that mimic the structural and functional complexity of proteins. Recent advances in the ribosomal synthesis of peptides with significant unnatural amino acid content are intriguing in this regard (35).

Materials and Methods

Synthesis of α- and α/β-Peptides.

α-Peptide 1 and α/β-Peptides 2–5 were prepared by standard Fmoc automated solid phase synthesis. α/β-Peptides 6–9 were prepared by microwave-assisted Fmoc solid-phase peptide synthesis (36, 37). Detailed synthetic methods can be found in SI Text. Crude peptides were purified to >95% purity by preparative reverse-phase HPLC on a C18 column by using gradients between 0.1% vol/vol TFA in water and 0.1% vol/vol TFA in acetonitrile. The identity and purity of 1–9 were confirmed by MALDI-TOF-MS and analytical HPLC, respectively. Stock solutions for physical characterization were quantified by UV absorbance assuming ε₂₇₆ = 1450 M⁻¹ cm⁻¹ for the Tyr or β³-hTyr residue in each sequence (38).

Crystallization and Data Collection.

Crystallizations were performed by the hanging-drop vapor-diffusion method. Peptide stock solutions were ≈15 mg/ml in water and mixed 1:1 with the crystallization buffers indicated before equilibration. α/β-Peptide 3 was crystallized from 0.2 M (NH₄)₂SO₄, 0.1 M Tris (pH 8.5), 30% vol/vol PEG 550 monomethyl ether, and the crystal was frozen after cryoprotection in the above buffer supplemented with 5% vol/vol glycerol. α/β-Peptide 4 was crystallized from 0.1 M Hepes (pH 7.5), 4.3 M NaCl, and the crystal was frozen after cryoprotection in the above buffer supplemented with 25% vol/vol glycerol. α/β-Peptide 7 was crystallized from 0.7 M diammonium tartrate, 0.1 M Tris (pH 8.5) and frozen after cryoprotection in 1 M diammonium tartrate, 0.1 M Tris (pH 8.5), 30% vol/vol glycerol. Crystals were frozen in liquid nitrogen, and diffraction data were collected on a Bruker X8 Proteum Diffractometer by using Cu Kα radiation. Data were indexed, integrated, and scaled with the Bruker Proteum2 software package. The crystal of 3 diffracted to 2.0 Å resolution with P2₁ symmetry and unit cell dimensions of a = 35.2 Å, b = 35.2 Å, c = 47.3 Å, β = 90.9°. Attempts to integrate and scale the data for 3 in higher-symmetry crystal systems led to significantly worse statistics, supporting the assignment of the monoclinic space group. The crystal of 4 diffracted to 1.8 Å resolution with P42₁2 symmetry and unit cell dimensions a = b = 38.4 Å, c = 46.5 Å. The crystal of 7 diffracted to 2.0 Å resolution with P42₁2 symmetry and unit cell dimensions a = b = 35.5 Å, c = 45.8 Å.

Structure Determination.

Structure solution was carried out by using the CCP4 software suite (39). Molecular replacement was carried out with Phaser (40). The search model for 3 was derived form the published structure of α-peptide 1 (21), the model for 4 was derived from the published structure of α/β-peptide 2 (14), and the model for 7 was derived from the refined structure of 3. The same R_free sets were used for 3 and 7 to avoid contamination of the test set used during the refinement of 7. Refinement was accomplished by a combination of Refmac (41) for automated refinement, Coot (42) for manual model building, and ARP/wARP (43) for map improvement by free atom density modification and water building. A Refmac library containing geometric restraints for the β-amino acid residues as well as the α → β and β → α amide linkages was included in the refinement. Helical and superhelical parameters (Table S2) were calculated by using the program TWISTER (44).

CD.

CD measurements were carried out on an Aviv 202SF Circular Dichroism Spectrophotometer with a 1-nm step and 5-sec. averaging time. Thermal melts were carried out in 5°C intervals with an equilibration time of 10 min. after each temperature change. In cases where cooperative folding transitions were observed, T_m values were determined by fitting the data to a simple two-state folding model by using GraphPad Prism. The unfolded baselines for samples that did not show complete unfolding were estimated based on baselines observed for samples that did unfold. Although such estimates may compromise the accurate determination of T_m values, qualitative trends in stability are evident from the raw thermal denaturation data (Fig. S3) and support the relative stability within the series as determined from T_m values.

AU.

AU measurements were carried out on a Beckman–Coulter XL-A Analytical Ultracentrifuge equipped with an An-60 Ti Rotor (Beckman-Coulter part number 361964, Fullerton, CA). All samples were prepared at a final concentration of 200 μM in 10 mM NaOAc (pH 4.6), 150 mM NaCl except for α/β-peptide 8, which was measured at 100 μM in 10 mM NaOAc (pH 4.6). Samples were measured at three or more speeds ranging from 12,000 rpm to 45,000 rpm and monitored at 275 nm (Figs. S5 and S6). Apparent molecular mass for each peptide was determined by global nonlinear fitting of the equilibrium radial absorbance data at all speeds by using the Igor software package (WaveMetrics). Details regarding the fitting procedure can be found in SI Text.