Synthetic Control of Tertiary Helical Structures in Short Peptides

Abstract

Helical secondary and tertiary motifs are commonly observed as binding epitopes in natural and engineered protein scaffolds. While several strategies have been described to constrain α-helices or reproduce their binding attributes in synthetic mimics, general strategies to mimic tertiary helical motifs remain in their infancy. We recently described a synthetic strategy to develop helical dimers (J. Am. Chem. Soc. 2015, 137, 11618–11621). We found that replacement of an interhelical salt bridge with a covalent bond can stabilize antiparallel motifs in short sequences. Here we show that the approach can be generalized to obtain antiparallel and parallel dimers as well as trimer motifs. Helical stabilization requires judiciously designed crosslinkers as well as optimal interhelical hydrophobic packing. We anticipate that these mimics would afford new classes of modulators of biological function.

Graphical Abstract

INTRODUCTION

α-Helical coiled coils are ubiquitous components of protein structure and function. Canonical coiled coils are stabilized by a series of hydrophobic knobs-into-holes¹ packing interactions along with inter- and intra-strand electrostatic contacts.^2–4 Exquisite design rules have been outlined for de novo design of coiled coil assemblies;^5–13 however, application of these rules is limited to long sequences (>3–4 heptads or 21–28 residues) where an adequate number of weak interhelical interactions can be designed to stabilize a desired construct.^14,15 Short coiled coils are inherently unstable owing to infrequent access to helical conformations in short peptide sequences, and a lack of sufficient interhelical contacts to mediate coiled coil formation.¹⁶ Thus, the entropic penalty is insurmountable for the formation of both a single helix and helical assemblies in the absence of requisite electrostatic and hydrophobic interactions and inter- and intrahelical hydrogen bonds. However, there exists a disconnect between the length of coiled coils required for conformational stability versus the size of the interaction surface needed for function. Despite their stability and occurrence, long coiled coils are often not required for biological function. A survey of helix dimers in protein-protein interactions suggests that critical binding residues are localized to 1–2 heptads in roughly half of the complexes mediated by this tertiary protein structure.¹⁷ Access to defined protein folds that satisfy functional contact surface area while foregoing length-dependent conformational stability would lead to new therapeutic modalities.

The large entropic penalty in coiled coil formation can be overcome by judicious incorporation of covalent bridges as surrogates for non-covalent contacts. Several classes of covalently-constrained helices have been described.¹⁸ Figure 1a illustrates two examples of helical assemblies in template-assembled synthetic proteins (TASP) and covalently-linked cores as miniproteins (CovCore). In TASPs, a synthetic scaffold, often a macrocycle¹⁹, is linked to peptide termini to reduce the entropic barrier to their assembly.^20,21 CovCore miniproteins utilize a multivalent crosslinker to restrain computationally designed bundles.²² A large effort to stabilize helical assemblies using metal-mediated interactions has also been reported.^23–25 These covalent efforts build on the elegant rules to non-covalently assemble coiled coils.^{5–13,26–30}

Figure 1.

(a) Covalent strategies for stabilizing helical assemblies. The classical template-assembled synthetic proteins (TASP) approach provides a platform for clustering peptide sequences. CovCore links helices to afford a synthetic miniprotein while the crosslinked helices are developed by replacing interhelical salt bridges with covalent bonds. (b) Here we demonstrate that the salt-bridge surrogate strategy can be generalized to precisely control the orientation and the number of helices in the assembly. Placement of a crosslink at the e/e’ or g/g’ positions of designed sequences captures an antiparallel conformation while constraining e/g positions leads to parallel helix dimers. Placement of two crosslinks at appropriate positions can stabilize a helical trimer. N and C designations indicate the amino and carboxy termini in the helix models.

We are motivated to develop a strategy that provides inhibitors for protein-protein interactions (PPIs) mediated by helical assemblies.¹⁷ A requirement for molecular recognition is that at least one face of the assembly needs to be programmable for binding. Therefore, the designs must avoid a strict adherence to a conserved set of residues that guide noncovalent or covalent assemblies. Several successful classes of coiled coil mimics have been shown to inhibit protein-protein interactions from classical examples that inhibit gp41 assembly and HIV entry,³¹ to more recent examples of Affibodies,³² Alphabodies³³ and α/β peptide foldamers.³⁴

We recently discovered a general approach for stabilizing short dimeric antiparallel coiled-coils.³⁵ To arrive at the optimal method, we tested various approaches to constrain minimal coiled coils based on the following key hypotheses: (a) stabilization of individual helices will enhance stability of the dimeric assembly.^26,36,37 (b) Linking of the dimeric scaffold would aid interpeptide contacts and helix formation.³⁶ (c) Noncovalent interhelical contacts can be strengthened by substitution with covalent bonds.^38–40 These studies revealed that replacement of an interhelical salt-bridge with a covalent bond (i.e. a salt-bridge surrogate) provides a general and versatile approach for the stabilization of short helix dimers (Figure 1a).^35,41 We extensively characterized the crosslinked helix dimers (CHDs) by circular dichroism (CD) and 2D NMR spectroscopies. We then applied the CHD design to the modulation of a protein-protein interaction involved in leukaemogenesis, where complex formation depends on a coiled-coil assembly.³⁵ Results showed that a designed CHD consisting of 20 residues could bind the target with higher affinity than the native tetramer coiled coil consiting of 276 residues, thus validating our hypothesis that long coiled coils are not necessary for enhanced binding if short binding epitopes display could be appropriately stabilized.

The antiparallel coiled coils were stabilized by bis-triazole linkers formed via a copper catalyzed azide-alkyne cycloaddition reaction^42,43 to constrain two peptides at their putative e/e’ positions. Bis-triazole bridges of varying lengths resulting from azidoalanine, azidohomoalanine, and azidolysine residues were tested to deduce that azidolysine crosslinking provided highest stability. In these studies, we also learned that optimal hydrophobic packing between individual helices is needed, in addition to the correct crosslinker, to stabilize helical dimers.³⁵

Here we apply the CHD strategy to access short peptides of various orientation and oligomerization states: parallel and antiparallel dimeric and antiparallel trimeric coiled coils. We sought to determine if a salt-bridge surrogate along with the optimal knob-in-hole packing interactions provides a general approach for programming short peptides into dimeric and trimeric coiled coils with the desired geometries (Figure 1b). A salient feature of these minimal coiled coil mimics is that they contain a single knob-into-hole interaction per peptide, thus providing a rigorous model to test the optimal pairing for this critical component of coiled coils. Accordingly, the design effort required us to critically evaluate knob-into-hole packing to stabilize dimers and trimers. [Note: canonical coiled coils are characterized by a helical supercoil and shorter pitch than helix dimers. The constructs described here are too short to supercoil and should be considered to be minimal examples of helix dimers.]

ASSESSMENT OF CROSSLINKERS FOR MINIMAL ANTIPARALLEL COILED COILS

We began design of minimal dimeric and trimeric coiled coils by testing the requirement for a bis-triazole crosslinker. The advantage of the copper-catalyzed azide-alkyne derived triazole linkers is that they can be readily synthesized from bioorthogonal functionality. A disadvantage is that nonnatural side chains are required. To make the CHD synthetic strategy less reliant on non-natural residues and therefore broaden applications, i.e. modification of proteins that contain short helical segments, we sought linkers that could directly alkylate nucleophilic cysteine residues. We compared the stability of an antiparallel helix dimer with different linkers in a generic model sequence (Figure 2). The designed sequence incorporates hydrophobic residues at a/d positions, and inter- and intrahelical electrostatic contacts to enhance the stability of dimers as previously described.³⁵ Critically, the model sequence features an optimal vertical triad that has been judged to offer enhanced stability to two-strand coiled-coils in studies by Gellman and Woolfson.^7,44,45 We designed and tested three crosslinks: bis-triazole (CHD-1), flexible ethylene glycol (CHD-2), and dibenzyl ether (CHD-3). The bis-triazole crosslinker consists of two aromatic groups. We incorporated a glycol linker in CHD-2 to determine if an easily accessible flexible linker would suffice. In order to directly mimic the aromatic nature of the CHD-1 linker, we also synthesized the dibenzyl ether linker, CHD-3. Synthesis of CHD2/3 via cysteine alkylation was performed by leveraging reactive cysteine side chains with halide containing cross linkers in basic aqueous buffer (Supporting Information Figure S1). Synthesis of CHD-1 was performed as previously described for CHDs.³⁵

Figure 2.

(a) Helical wheel diagram of antiparallel CHD and the designed model sequences. (b) Three crosslinking strategies were employed to stabilize model sequences. (c) Circular dichroism spectra of the four constructs. Legend: equimolar mixture of N and C peptides (unconstrained; X = Cys), bis-triazole (CHD-1), glycol (CHD-2), and dibenzyl ether (CHD-3) crosslinkers. (d) Overlay of 20 lowest conformations derived from CHD-3 NMR analysis. The CD studies were performed in 50 mM aq potassium fluoride (pH=7.4) at 20 μM peptide concentrations. The NMR studies were performed in aqueous trifluoroacetate buffer, pH 4.5, supplemented with 10% D₂O.

CD analysis reveals that linker rigidity directly impacts dimer helicity (Figure 2b). For example, dimer containing the flexible glycol linker, CHD-2, is considerably less stable than the more rigid analogs, CHD-1 and CHD-3. The CD studies were performed in 50 mM aq potassium fluoride (pH=7.4) at 20 μM peptide concentrations.

We further analyzed CHD-3 by NMR spectroscopy. Combination of 1D NMR, total correlation spectroscopy (TOCSY), and nuclear Overhauser effect spectroscopy (NOESY) analyses were performed on 1.5 mM CHD-3 in aqueous buffer (pH 4.5) with 10% D₂O and 0.1% TFA. The NMR studies confirm a robust helical signature of the coiled coil mimic when stabilized by the dibenzyl ether crosslinker similar to that observed for the previously reported bis-triazole linked CHD.³⁵ The helical signature of CHD-3 is supported by sequential NOE cross peaks, d_NN (i, i+1), and several medium/strong range NOEs (Figure S3). Backbone dihedral angles (Φ) were calculated based on coupling constants, ³J_NHCHα; the Φ values fall in the range expected for canonical helices. A structural model of CHD-3 was obtained using 74 NOESY crosspeaks and 16 Φ values. An overlay of 20 lowest conformations is depicted in Figure 2d. (The geometry of the structure was optimized with Schrodinger’s MacroModel software using experimental NOE and coupling constant restraints.⁴⁶) Size exclusion assays, in combination with the NMR studies, provide strong evidence that the crosslinked dimers do not aggregate in aqueous buffers into higher order assemblies (Figure S2).

The stability of a dibenzyl ether-linked CHD is similar to that of the previously characterized bis-triazole derived construct as judged by CD and NMR studies.³⁵ To evaluate the stabilization provided by both crosslinks, we performed a CD titration experiment with organic solvents that favor formation of helical assemblies. The CD signature of peptide helices often improves dramatically in aqueous solutions containing trifluoroethanol (TFE) and peptides reach their maximum helicity between 30–40% TFE. The TFE tiration studies provide experimental conditions to probe sequence-dependent percent helicity.⁴⁷ We observed little change in the CD spectra of CHDs upon addition of 10% and 20% TFE, which potentially suggests that the crosslinked CHDs are fully helical under aqueous conditions (Figure S4).

Overall, the dibenzyl ether linker expands the choice of synthetic crosslinks that may be employed for stabilizing helical assemblies. The bis-triazole and the dibenzyl ether linker chemistries offer different yet complementary advantages for the stabilization of short peptides. While azidolysine constructs can be assembled bioorthogonally, the cysteine alkylation chemistry can be implemented with natural amino acid residues.³⁵

APPLICATION OF THE CROSSLINKING STRATEGY TO STABILIZE MINIMAL PARALLEL COILED COILS

The above studies illustrate that antiparallel helix dimers can be reproducibly accessed in short peptides with two different salt bridge surrogates. We next evaluated if the general CHD strategy similarly provides access to stable parallel helix dimers. Similar to the antiparallel helix dimer, the sequence was designed to have hydrophobic residues at a/d positions, and inter- and intrahelical electrostatic contact to improve the stability of dimers and a single helix. For parallel assembly, possible positions for interhelical ionic interactions are e/g’ and g/e’. Here, cysteine residues were incorporated at e/g’ positions and alkylated with dibenzyl ether linker to obtain a parallel dimer CHD-4 (Figure 3a).

Figure 3.

(a) Helical wheel diagram of parallel dimer CHD-4. (b)CD spectra of parallel and antiparallel CHD dimers in 50 mM aqueous KF, pH= 7.4. (c-d) Overlay of 20 lowest conformations derived from parallel dimer NMR analysis and the lowest energy structure. (e) Size exclusion chromatography to assess the oligomerization of the designed antiparallel (CHD3) and parallel (CHD4). Lysozyme, aprotinin, and vitamin B12 were used as molecular weight controls.

CD analysis based on intensity at 222 nm minimum and 222/208 nm ratio indicates that the parallel helix dimer CHD-4 maintains comparable helicity to the antiparallel dimer in aqueous buffers (Figure 3b). We performed thermal denaturation studies on both systems to further gauge their conformational stability. These studies also suggest that both parallel and antiparallel CHDs are highly stable (Figure S5).³⁵ Both CHDs display a broad denaturation curve and retain roughly half of their maximum helicity at 80 °C. Extensive NMR analysis, which included 1D, TOCSY, and NOESY studies, confirmed that the crosslinked parallel CHD is helical and dimeric. These experiments were conducted in 10% d₃-CH₃CN and 10% D2O in H₂O with 0.1% TFA (pH 4.5). The organic cosolvent was necessary to enhance the solubility of the peptide to achieve the desired concentrations (0.5 mM) for NMR. Sequential NOE cross peaks of d_NN (i, i+1) and several medium/strong range NOEs denoting helical signature of parallel CHD were observed (Figure S6). Coupling constants, ³J_NHCHα, calculated from Φ values also fall into the range expected of canonical helices. 68 NOESY crosspeaks and 18 Φ were used to obtain a structural model of parallel CHD. An overlay of 20 lowest conformations is depicted in Figure 3c. The lowest energy NMR structure is shown in Figure 3d. The NMR structures were calculated using Schrodinger’s MacroModel software from experimental NOE and coupling constant restraints. Size exclusion assays (Figure 3e) were also performed on the parallel CHD and, in combination with the NMR studies, provide strong evidence that the crosslinked dimers do not aggregate in aqueous buffers into higher order assemblies.

The above studies establish the versatility of the CHD strategy to provide parallel and antiparallel minimal coiled coils. We found that a combination of appropriate crosslinks and hydrophobic packing is needed for the requisite conformational stability. As both parallel and antiparallel dimeric coiled coils are intimately engaged in mediating biomolecular recognition, we expect that our approach will yield new classes of inhibitors for difficult interactions. In published³⁵ and ongoing studies, we have successfully demonstrated the potential of the approach to target chosen protein-protein interactions.

CONTROL OF TRIMERIC HELICAL ASSEMBLIES

Trimeric coiled coils have emerged as a new class of therapeutic miniproteins.^32,33 A crosslinked helix trimer (CHT) would offer a straightforward path to miniaturize these miniproteins, potentially reducing off-target effects and the cost of development. Classical studies on peptide coiled coils have intensely investigated the oligomerization preferences of peptides into dimeric, trimeric, and higher order assemblies. In their seminal paper, Harbury et al. showed that geometric preferences of residues at a/d positions dictate the size of the assemblies and that β-branched residues at a/d positions sterically favor coiled coil trimers over dimers.^5,6 Geometric rules for coiled coils have been extended to develop a range of coiled coil assemblies in a predictable manner.^13,48,49

Designed coiled coil antiparallel trimers, including therapeutic Affibodies and Alphabodies,^32,33 have generally featured isoleucine cores to disfavor dimer and tetramer stoichiometries; however, trimer cores can tolerate a distribution of hydrophobic residues.⁴⁹ We began by testing if isoleucine residues at a/d positions in our generic sequence provided stable antiparallel trimers upon crosslinking (Table S2). In our crosslinking strategy, the crosslinks are placed at the e or g positions of the coiled coil heptads. The distance between these positions on neighboring helices in dimers and trimers varies; the Cα to Cα distances decrease from the 10 Å in dimers to 9 Å in trimers. We retested bis-triazole bridges of varying lengths resulting from azidoalanine, azidohomoalanine, and azidolysine residues for optimal helicity in trimers, and found that the bis-triazole bridge from azidohomoalanine provides highest helicity, as probed by circular dichroism spectroscopy, in trimers as compared to azidolysine in dimers (Figure 4a–b). This result matches the expectation from modeling studies and the shorter interhelical distance in trimers.

Figure 4.

(a) Helical wheel diagram depicting antiparallel CHT with interhelical hydrophobic interactions at a/d positions highlighted in gold and synthetic crosslink positions in black. (b) CD spectra demonstrating length dependence of azido amino acid crosslinks on CHT stability. (c) CD spectra indicate that sequences containing optimal vertical triad residues (vertical triad constellations or VTC) at a/d positions are slightly more helical than all isoleucine core.

Although the azidohomoalanine crosslink provides higher stability than the other crosslinks tested, the resulting CHT^ILE-2 remains weakly helical. This result prompted us to investigate if the isoleucine core is truly optimal for a minimal trimeric coiled coil that features a single knob-into-hole interaction. The designed CHT features antiparallel and parallel dimeric pairs (Figure 4a). We initially tested if the vertical triad constellations^7,44 predicted to provide optimal vertical packing for dimeric parallel and antiparallel helices could be imported into the trimer at each correponsding dimer interfaces. Sequences were designed to contain vertical constellations Ile-Leu-Ile (a’-a-a’ and d”-d’-d”) in the antiparallel pair, and Leu-Ile-Leu (d-a”-d) in the parallel pair to obtain CHT^VTC. This trimer proved to be slightly more helical than the all isoleucine CHT^ILE-2 construct (Figure 4c).

Encouraged by this finding that the all isoleucine core can be improved upon, we performed a bioinformatics analysis on natural coiled coil trimers to identify other potential packing residues at the a/d positions. The bioinformatics analysis was performed with the CC+ database. CC+ is a coiled coil structure repository that allows analysis of sequence-structure information within coiled coil assemblies.⁵⁰ Analyses using CC+ have previously revealed a/d preferences within dimeric knob-into-hole packing.⁷ We produced a curated dataset of 3,546 trimeric protein assemblies from CC+ to understand residue composition within a minimal trimeric knob-into-hole motif. We performed the bioinformatics survey for both parallel and antiparallel trimers. The full set of results are included in the Supporting Information. Below we discuss results for the antiparallel motif as they are directly relevant to our synthetic models and commonly engineered three helix bundles.

The bioinformatics analysis suggests preferred hydrophobic amino acid packing in trimeric assemblies. Key a/d residue compositions for antiparallel trimers are presented in Figure 5a, and the overall analysis of 20 proteinogenic amino acids is presented as a heatmap in Figure S7. The preferred amino acids for a and d positions in antiparallel trimers are clearly evident from the curated dataset. The analysis shows that there is a strong preference for a β-branched residue at the d position of antiparallel coiled-coil trimers, with isoleucine the clear winner (Figure 5a). The residue preference is not as distinct at a positions, although leucine residues are most common. Overall the data is consistent with the earlier analyses that the trimer interface contains a distribution of hydrophobic residues.⁴⁹ This analysis reveals that the highest scoring pair for an antiparallel trimer (Figure 4a) is leucines and isoleucines at all a and d positions, respectively. CHT^CC+ reflects this preference (Table S2)

Figure 5.

(a) Data from the CC+ database indicating amino acid preferences at a/d positions in antiparallel trimers. (b) Buried solvent accessible surface area (SASA) for 512 isoleucine and leucine combinations calculated from MD simulations support experimental biophysical results with CHTs. Relative buried SASA values for the tri-isoleucine construct (CHT^ILE-2), trimer combination derived from consideration of the vertical triad constellation (CHT^VTC), and pairing suggested by the bioinformatics analysis (CHT^CC+) are highlighted on the graph. (c) The helical wheel diagram of CHT^CC+ and surface renderings showing hydrophobic packing for the three trimeric knob-into-holes combinations. (d) CD and T_M studies support CHT^CC+ sequence stability as compared to CHT^VTC and CHT^ILE-2. (e) Size exclusion chromatography to assess the oligomerization of CHTs and CHDs.

Control of the coiled coil oligomerization state between dimer, trimers, and higher order assemblies is a challenge in self-assembling peptides but since crosslinking dictates the oligomerization states in CHD and CHT peptides, these constrained peptides directly report on the conformational stability rendered by hydrophobic packing. We utilized buried solvent accessible surface area (SASA) calculations as a metric to determine optimal pairing from the bioinformatics results. The SASA calculations were performed on trajectories obtained from all-atom MD simulations of 512 combinations of leucine and isoleucine at the trimer a/d positions using the AMBER 16 molecular dynamics package.⁵¹ We restricted these calculations to leucine and isoleucine residues at the a and d positions as these are the most common residues found at trimer interfaces; furthermore, the trimer was restrained (as would be with crosslinkers) in the calculations so the SASA values can be directly correlated with hydrophobic packing. The results of the SASA calculations are depicted in Figure 5b, and show a large degree of variation among the 512 combinations. The relative SASA scores for the all isoleucine core (CHT^ILE-2), residues representing the vertical triad constellation (CHT^VTC), and the sequence predicted from the CC+ database (CHT^CC+) are highlighted in Figure 5c.

The SASA analysis suggests that the hydrophobic packing in CHT^CC+ buries significantly more surface area (620 Ų) to that found in the all isoleucine core (530 Ų). CD studies show that CHT^CC+ is highly helical and stable, confirming our bioinformatics and SASA analyses (Figure 5d). Size exclusion studies support the formation of a trimer (Figure 5e). The SASA hydrophobic packing analysis suggests that the stability of CHT constructs may be further improved; detailed experimental studies to investigate other pairings are underway.

CONCLUSION

Synthetic mimics that capture the topology and binding epitopes of engineered and native proteins have proven to be attractive as probes for the study of protein folding and recognition, and as leads for drug design.^52–55 Here we described a general crosslinking strategy to stabilize helical dimers and trimers, which are often involved in protein-protein interactions.¹⁷ Our and similar efforts in other groups are aimed at capturing these localized binding interactions in minimal motifs.^34,35 The crosslinking strategy described herein provides rapid access to parallel and antiparallel helical dimers and trimers. Helical stability requires optimal crosslinkers as well as interstrand hydrophobic packing. Helical dimers were built with the previously described vertical triad constellations at the heptad a and d residues.^7,44 Stabilization of minimal crosslinked trimer required bioinformatics and computational analyses to identify optimal hydrophobic packing. We expect that these crosslinked helices, featuring single knob-into-hole interactions, will prove to be attractive biophysical tools for identifying optimal hydrophobic interfaces in helical tertiary motifs.