Structural and Positional Effects of Peptoid Residues on Triple Helix Stability

Abstract

The collagen triple helix, composed of three protein strands with a repeating GlyXaaYaa sequence, achieves maximum stability with proline (Pro) and hydroxyproline (Hyp) at the Xaa and Yaa positions. Previously, we demonstrated that peptoid residues (N-substituted glycines, N-Glys) at the Xaa position significantly enhance triple helix stability. Here, we show that N-Glys at the Yaa position also stabilize the triple helix, with stability influenced by the N-Gly’s position and side chain structure. Over 22 N-Glys with various natural and unnatural side chains were investigated using CD spectroscopy, X-ray crystallography, and computational simulations. The results show that N-Glys at the Yaa position support triple helical folding and are more stabilizing than their corresponding amino acids but less so than Hyp due to suboptimal ϕ-ψ angles. Metadynamics simulations revealed that N-Glys at the Yaa position have a broader conformational space due to minimal steric hindrance from neighboring glycine residues. N-Glys’ side chain bulkiness influenced stability only at the Xaa position and not the Yaa position. Additionally, only S-isomers of chiral N^α−C^β branched N-Glys were compatible with triple helical folding, presenting different backbone conformations and accessible rotamers based on their position. At the Yaa position, (S)-N-(1-phenylethyl)-Gly (Nspe) stabilized the triple helix better than all other residues except Hyp through a newly discovered intrachain CH•••π interactions. This research deepens our understanding of triple helical folding and introduces new strategies for designing stable collagen mimetic peptides with diverse side chains, expanding potential applications in biomedicine and biomaterials.

INTRODUCTION

Collagen is the major constituent of the extracellular matrix (ECM) in animals and provides structural and functional support for tissues.^{1, 2} There are 28 types of human collagens comprised of 45 different polypeptide chains^{3, 4} all of which contain a triple helix, which is the hallmark of the collagen structure. Three left-handed polypeptides, each with a polyproline-II (PPII) helical conformation, intertwine to form a right-handed triple-helical structure. To fold into the triple helix, each polypeptide chain must have a repetitive GlyXaaYaa sequence, where Gly is present at every third residue. Xaa and Yaa can be any amino acids, but Pro and Hyp [Hyp: (2S,4R)-hydroxyproline] frequently populate the Xaa and Yaa position, respectively.⁵ Glys buried in the interior of the triple helix are solvent-inaccessible and their amide N−H hydrogens form interchain hydrogen bonds with the C=O groups of the Xaa residues in another chain, which is the major driving force for triple helical folding.³ Gly mutations can lead to diseases such as osteogenesis imperfecta.⁶ Among natrual amino acids, Pro and Hyp, at the Xaa and Yaa, respectively, provide the highest triple helical stability.⁷ Here, the pyrrolidine ring helps preorganize the PP-II helix, and the (4R)-hydroxyl group on Hyp in the Yaa position exerts a stereoelectronic effect that favors PP-II³ as well as the formation of hydration networks⁸ that may stabilize the triple helix. Charged natural amino acids can also influence triple helical folding by inter-strand charge-pairing interactions, which can be used to produce heterotrimeric collagen like peptides with predetermined chain compositions and registers.^9–14

Since natural collagens are large and complex, simple collagen mimeic peptides (CMPs) with repeats of GlyXaaYaa triplets have been utilized to study the structure and folding of collagen triple helix.^{15, 16} Using CMPs, researchers have investigated triple helical propensities of not only canonical amino acids^{5, 7, 17} but also artificial amino acids, which demonstrated that electronegative substituents, such as fluorine, chlorine, and methoxy, with (R)-configuration on the C^γ of Pro in the Yaa position, stabilize CMP triple helices via the stereoelectronic effect^18–24 and that substitution of Gly with aza-Gly results in hyperstable CMPs due to extra inter-strand hydrogen bonds.^25–29 Egli and coworkers showed that the attachment of a hydrophobic group to (4R)-amidoproline at the Xaa position accelerated folding and enhanced the triple helical stability of CMPs by promoting cis-trans isomerization and the formation of a molten globule-like intermediate.^{30, 31} Considering that numerous bioactive domains of collagen molecules are triple helical, the folding principles of triple helix as applied to non-canonical amino acids are not only scientifically interesting but could also be useful in developing bioactive collagen mimetics with properties superior to natural collagen sequence. Unfortunately, our understanding of various factors responsible for triple helical stabilization applies mostly to α-amino acids and a small number of non-canonical amino acids, which are difficult to synthesize.

Peptoids are oligomers of N-substituted Glys (N-Glys, Figure 1A) that have side chains of α-amino acids moved from the C^α position to the N^α position. The submonomer method offers facile synthesis of a large library of peptoids.³² Early promising studies of peptoid oligomers suggested the possibility of the discovery of a potent new class of therapeutics, owing to their chemical diversity and improved biostability in comparison to conventional peptides.^{33, 34} Concerted efforts in the peptoid community to find peptoid sequences with strong folding propensities have produced a growing understanding of side chain – backbone interactions and other stabilizing factors^{35, 36}, but progress in developing a robust set of functional secondary structures has been slow. Although N-Glys are structurally similar to Pro and Pro residues are the most critical constituent of collagen protein sequences, to date, there has not been a systematic investigation of incorporating N-Glys into the GlyXaaYaa triple helical collagen sequence. This is a missed opportunity in the study of peptidomimetics, as the collagen triple helix is one of the most stable secondary (also known as super-secondary) protein structures and a superb scaffold for displaying N-substituted side chains. In addition, triple helices displaying natural and unnatural side chains offer potential for the discovery of new therapeutics targeting collagen-cellular receptor interfaces as well as collagen-ECM interactions. Previously, it was hypothesized that hydrophobic interactions of N-Glys, especially those at the Xaa position, could stabilize collagen triple helices.^37–40 However, by investigating CMPs substituted with a series of N-Glys at the Xaa position (N^xaaa-CMPs), we previously showed that the bulkiness of peptoid residues at the Xaa position correlated with triple helical stability rather than hydrophobic interactions. We determined that bulky side chains restrict the rotation of the C^α−N^α bond and preorganized each CMP strand into the PP-II structure, thus increasing triple helical stability.⁴¹ Although this investigation demonstrated for the first time that N-Gly is conducive to forming a stable triple helix when incorporated into the Xaa position of the GlyXaaYaa sequence, there is no knowledge of such for Yaa position, which has local spatio-chemical environment different from the Xaa position. In fact, factors stabilizing at the Xaa position can be destabilizing at the Yaa position.^{16, 42}

Figure 1.

Substituting the middle Hyp residue of (GlyProHyp)₇ with N-Glys produces stable triple helices. (A) Chemical structure of host CMP with the sequence of Ac-(GlyProHyp)₃-GlyProYaa-(GlyProHyp)₃-NH₂, where the central Yaa position was substituted with various α-amino acids and N-Glys. The carbon atom next to the carbony group is designated as C^α. The N atom connected to the C^α is the N^α. In a peptoid monomer, the side chain is moved from the C^α to the N^α. (B) Representative CD spectra (upper plot) and melting profiles (lower plot) of aaa^y-CMPs and N^yaaa-CMP. N^yphe-CMP produced CD spectral characteristics similar to those of Hyp^y-CMP and Phe^y-CMP. A two-state transition is also observed in the thermal melting experiment. These results suggest that N^yphe-CMP folded into a triple helix. Although N^yphe-CMP was less stable than Hyp^y-CMP, moving the Phe side chain from the C^α to the N^α (Phe^y-CMP vs. N^yphe-CMP) increased the thermal stability of the host peptide. (C) Comparison of triple helical stability of aaa^y-CMPs and N^yaaa-Glys with natural side chains. All Ν-Glys except four had higher T_m values than their corresponding α-amino acids with the same side chains. However, none of the Ν-Glys produced triple helices more stable than Hyp. The ΔT_m values are the comparisons between N-Glys and their amino acid counterparts. *The thermal denaturation experiments were performed in 1× PBS (pH 7.4). **The T_m value is from Ref. 41. ***N^yhcys-CMP is not exactly equivalent to Cys^y-CMP but structurally similar. The T_m values were averaged from two independent scans and rounded to the nearest integers. The T_m difference between the scans was less than ± 0.5 °C.

The positional effect of residues within the GlyXaaYaa motif (Xaa vs. Yaa) has been documented, particularly for Pro analogs. In making a stable triple helix, Pro analogs in the Xaa position prefer the C^γ-endo ring pucker, whereas those in the Yaa position prefer the C^γ-exo conformation.^{3, 16} For example, (2S, 4S)-fluoroproline, which has a C^γ-endo ring pucker, stabilizes the triple helix only at the Xaa position and not at the Yaa position, whereas (2S, 4R)-fluoroproline, which has a C^γ-exo ring pucker, is stabilizing at the Yaa position and not at the Xaa position.²¹ Similarly, Hyp which has C^γ-exo ring pucker is stabilizing at the Y position but generally not the X position.^{3, 43, 44} Although the roles of Xaa and Yaa positions in stabilizing the triple helix are well established for cyclic Pro derivatives, there is limited understanding of these effects for noncyclic Pro analogs such as N-Glys.

In this study, we systematically investigated the triple helical stabilization effect of N-Glys at the Yaa position, results of which were compared to N-Glys at the Xaa position, revealing N-Glys’ positional effects on triple helix stabilization. The middle Hyp (Yaa position) of Ac-(GlyProHyp)₇-NH₂ was replaced with N-Glys featuring side chains of natural and unnatural amino acids (designated as N^yaaa-CMP). CD spectroscopy, X-ray crystallography, and computational modeling including MD simulation, were used to study the molecular structure and triple helical stability of the N^yaaa-CMPs. The results show that even at the Yaa position, N-Glys are generally more stabilizing than the corresponding amino acids. Although this trend is similar to what was observed for N-Gly at the Xaa position, the stabilization effect at the Yaa position was less pronounced than at the Xaa position, and unlike at the Xaa position, exhibited little preference for the side chain chemical structures. Our work also revealed stereochemical stabilization mediated by N^α−C^β substituent of the N-Gly side chain, which led to the discovery of peptoid monomers that specifically produce large stabilization of the triple helix at the Yaa position.

MATERIALS AND METHODS

All chemicals were used as received from commercial suppliers without further purification. The Fmoc-protected amino acids, primary amines, solvents, resin, and reagents for synthesizing and purifying collagen mimetic peptides and peptoids are listed in Supporting Information (SI) Table S1.

Host-Guest Peptide synthesis

An intermediate peptide, Fmoc-[GlyProHyp(tBu)]₃, was synthesized at a 0.15 mmol scale on TentaGel^® R RAM Resin (90 μm) Rink-type (loading density: 0.2 meq/g) via the standard Fmoc-mediated solid-phase peptide synthesis using a Focus XC Solid Phase Peptide Synthesizer (AAPPTec, Louisville, KY, USA). Amino acid coupling was achieved using the HBTU/HOAt chemistry. The resulting intermediate peptide was split to synthesize various host-guest peptides. Each host-guest peptide was manually synthesized. The manual synthesis, cleavage, and HPLC methods for the host-guest CMPs are described in the SI (Tables S2 to S6). The purified products were verified using a Waters MALDI Micro MX MALDI-TOF mass spectrometer (Waters Corporation, Milford, MA, USA) (Table S7).

Circular dichroism (CD) spectroscopy experiments

After lyophilization, the purified peptides were dissolved in DI water to obtain stock solutions. The concentrations of the host-guest peptide stock solutions were determined by UV-Vis spectroscopy using a SpectraMax M2e microplate reader (Molecular Devices, San Jose, CA, USA). The absorbance of the peptide/peptoid solutions was measured at 214 nm (extinction coefficient: 2200 M⁻¹ cm⁻¹ per peptide bond ¹⁷) or 280 nm for sequences containing a tyrosine side chain (extinction coefficient: 1490 M⁻¹ cm⁻¹ per residue) using a quartz cuvette with a path length of 1 cm. Before the CD experiments, peptide stock solutions were heated at 80 °C for 10 min and stored at 4 °C for at least 48 h. Then, the peptide solutions were diluted to 150 μM (total peptide strands) in 1× PBS (pH 7.4). Peptide sequences containing acidic or basic side chains were prepared at 150 μM (total peptide strands) in HCl (0.1 M) or NaOH (0.1 M) solutions. Dilution was performed using cold diluents. For PP-II host-guest sequences, peptide solutions were prepared at 100 μM in 5 mM phosphate buffer (pH 7.0). All CD measurements of the peptide solutions were conducted on a JASCO J-1500 CD spectrometer (JASCO, Easton, MD, USA) using a 1-mm quartz cuvette.

The CD spectra of each peptide were recorded from 200 to 250 nm at 4 °C. The CD spectral scanning used the following parameters: data pitch, 0.1 nm; digital integration time, 8 sec; bandwidth, 1 nm; scanning speed, 20 nm/min; scanning mode, continuous. The raw CD spectra were smoothed using a mean movement method with a convolution number of 25. The molar ellipticity ([θ], 10³ deg cm² dmol⁻¹) values were converted from the recorded ellipticity using JASCO Spectra Manager Version 2 software (Version 2.15.01). All the spectra are the averages of two independent measurements.

Thermal denaturation of each sequence was performed by monitoring the change in ellipticity at 225 nm from 4−80 °C at a heating rate of 0.5 °C/min. Data were processed using the JASCO Spectra Manager Version 2 software (Version 2.15.01). The denaturation curves and first-derivative plots were smoothed using a means-movement method with a convolution number of 25. The first derivative was calculated using a subtraction method. The reported T_m values were the averages of two independent thermal denaturation scans rounded to the nearest integers. The differences in the T_ms between two measurements for each peptide were within ± 0.5 °C.

X-ray crystallography

Purified and lyophilized N^yphe-CMP and N^ynbz-CMP were dissolved in water at concentrations of 5 and 10 mg/mL, respectively. A Crystal Gryphon LCP robot (Art Robbins Instrument, Sunnyvale, CA, USA) was used for sitting-drop vapor-diffusion crystallization using commercially available crystal screen reagents. Equal amounts (0.25 μL each) of peptides and screen conditions were mixed and sealed in chambers where the reservoirs contained 0.5 mL of screen solutions. N^yphe-CMP crystals were grown at 21 °C with Wizard Cryo 1 & 2 HT-96 (Rigaku, Bainbridge Island, WA, USA) condition D12, corresponding to 40% (v/v) PEG600 containing 100 mM imidazole/hydrochloric acid (pH 8.0) and 200 mM zinc acetate. N^ynbz-CMP crystals were grown at 4 °C in condition H4 of the JBScreen Classic HTS I screen (Jena Bioscience, Thuringia, Germany), corresponding to 30% (w/v) PEG8000 containing 200 mM Ammonium sulfate. In preparation for data collection, crystals were briefly immersed (10–20 sec) in a cryoprotection solution consisting of a crystallization buffer supplemented with 25% glycerol. The cryoprotected crystals were then harvested using a small rayon loop attached to a mounting pin and quickly plunged into a liquid nitrogen reservoir. Diffraction data from the N^ynbz-CMP crystal were collected with a Pilatus-6M detector on beamline 9–2 at the Stanford Synchrotron Radiation Lightsource (SSRL, National Accelerator Laboratory, Menlo Park, CA). The SSRL resources, including Remote Access⁴⁵, Blu-Ice⁴⁶, Automated Sample Mounting System^{47, 48}, and AUTOXDS script⁴⁹, were used for data collection and analysis. Diffraction data from N^yphe-CMP were collected with an Eiger-16M detector on the NECAT beamline 24-ID-E at the Advanced Photon Source (APS, Argonne National Laboratory, Lemont, IL). The diffraction data were processed, integrated, and scaled with XDS⁵⁰ and AIMLESS.⁵¹ The structure of N^ynbz-CMP was determined by molecular replacement using Phaser in the PHENIX software package⁵², where model 1G9W was used as the search model. The initial phases provided electron density maps that allowed us to build a complete model of N^ynbz-CMP using COOT.⁵³ The model was refined with phenix.refine.⁵⁴ The model of N^ynbz-CMP was used as a search model to determine the structure of N^yphe-CMP by molecular replacement. The structure of N^yphe-CMP was determined and refined similarly to N^ynbz-CMP. The structural figures of N^yphe-CMP and N^ynbz-CMP were rendered using UCSF Chimera.⁵⁵ The crystallographic data and refinement statistics are summarized in Table S8.

Metadynamics calculations and simulations

Well-tempered metadynamics simulations were applied to study the conformational stability of Gly, Hyp, Nnbz, Nala, Nleu, Nchx, Nlys, Nphe, and Nval residues within the Yaa-CMP host. We used NAMD and the PLUMED package^56–58 to perform metadynamics simulations for acquiring the free energy landscape as a function of ϕ and ψ angles of the guest residues. Events of energy deposition took place every 200 fs in a Gaussian form with an energy height of 0.01 kcal/mol and a dihedral angle increment of 0.35 rad. We summed 200,000 deposited events to generate the energy landscapes for each selected Yaa-CMP. This Gaussian number yielded the convergence of the energy landscapes. The molecular models and atomic interactions of the selected Yaa-CMP were built from a combination of DFT calculation and CHARMM general force field (CHARMM GenFF).⁵⁹ We used HyperChem software (Hypercube, Inc., Gainesville, FL, USA) for the DFT calculation. We also updated the standard CHARMM27 force field⁶⁰ to enable the modeling and simulation of our peptide-peptoid hybrids. Model optimization was performed to identify the energy favorite coordinates of all atoms using geometric optimization with the conjugate gradient method. Molecular parameters, including charge distribution, van der Waals parameters, bond angles, and dihedral angles, were defined from the optimized molecular models that were made to an automatic analogy with CHARMM GenFF. The penalty scores for all parameters were less than 10, indicating that the analogy was fair.⁵⁹ The intrinsic coordinates of all the tested hybrids were also defined according to the optimized structure to ensure that the initial geometry was at the optimum energy level. All structures were simulated under the same conditions as those used for the Xaa-CMP system reported previously.⁴¹

Molecular dynamics (MD) simulations

Molecular dynamics simulations were performed for five residue variants. In each simulation, three collagen strands were modeled, each containing a single peptoid residue. Molecular models of capped peptoid residues (“dipeptoid”) were prepared using the ETDG method⁶¹ to generate approximate starting conformations, which were subsequently refined by quantum mechanical geometry optimization at the semi-empirical AM1 level of theory. Partial charges were assigned using the AM1-BCC method⁶² to ensure compatibility with the Generalized Amber Force Field.⁶³

Initial CMP structures were derived from the X-ray crystal structure of N^ynbz-CMP by truncating the N-substituted side chains at the Cβ (i.e., the second carbon from the backbone amide nitrogen) position and rebuilding the side chains using coordinates from the corresponding dipeptoid models. Each collagen strand contained a single residue mutation introducing a peptoid side chain. For X-position substitutions, residue 11 was replaced, whereas for Y-position substitutions, residue 12 was replaced.

A hybrid force-field parameter set was constructed, with Amber ff19SB parameters⁶⁴ used for peptide residues and GAFF2 parameters used for peptoid substitution. GAFF2 was chosen due to its broad chemical coverage and suitability for parameterizing complex noncanonical residues. The resulting models were minimized using the limited-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) method. All simulations were performed using the Generalized Born/surface area (GB/SA) implicit solvation model as implemented by Onufriev, Bashford, and Case.^{65, 66} Systems were heated for 200 ps at 310 K, a temperature commonly used to model collagen system under near-physicological condition,⁶⁷ and equilibrated for 800 ps, followed by subsequent simulation for 20 ns. All Root Mean Square Deviation (RMSD) values were stable and remained constant throughout the production trajectories (Figure S13), supporting maintenance of the folded triple-helical state during the simulations. The 20 ns timescale was sufficient for stable sampling of the local backbone and peptoid side-chain conformation used in our mechanistic analysis without reproducing bulk or higher-order properties of the collagen triple helix.

RESULTS AND DISCUSSION

N-Glys at the Yaa position stabilize the triple helix.

We examined the triple helical propensity of N-Glys (Naaa) at the Yaa position using the host-guest peptide Ac-(GlyProHyp)₃GlyProYaa(GlyProHyp)₃-NH₂ (aaa^y-CMP or N^yaaa-CMP), where the central Hyp of the (GlyProHyp)₇ host sequence was replaced with peptoid guest residues (Figure 1A). We first synthesized N^yaaa-CMPs with side chains covering the natural side chains and conducted CD spectroscopy at room temperature as well as CD melting experiments (Figure 1B), results of which were compared to those of the corresponding peptides with natural amino acids in Figure 1C (Figures S1 to S3). The results clearly showed that N^yaaa-CMPs folded into a stable triple helix, as evidenced by the characteristic CD spectrum and sigmoidal shape of the melting transition.³⁷ Similar to N-Glys at the Xaa position, all N-Glys with natural side chains except four (N^yarg, N^ymet, N^ygln, and N^yala), improved the triple helix stability compared to their α-amino acid versions, although none were more stable than the host peptide with Hyp at the guest position. Although Nala was reported to destabilize triple helix at the X position due side chain being too small to lock-in optimal backbone dihedral angles,⁴¹ further study is needed to understand why these four residues at the Y position produce more unstable triple helix compared to their natural amino acid analogs. Interestingly, N^yphe-CMP and N^ytyr-CMP showed the greatest stabilization with a T_m elevation of 18 °C (Figure 1C). These results demonstrate that moving side chains from the C^α position to the N^α position generally improves the triple helical stability for amino acids not only for the Xaa but also for the Yaa position.

As mentioned above, all N^yaaa-CMPs had T_ms lower than that of the host peptide (Hyp^y-CMP), indicating that substituting Hyp with N-Glys destabilized the triple helix. The T_m reduction ranged from 5 − 18 °C, which is in stark contrast to the substitution at the Xaa position, where some N-Glys improved triple helical stability (Table 1) compared to Pro^x-CMP. Even when compared to Pro^y-CMP (T_m = 48 °C), the T_m values were generally lower, except for N^yval-CMP (Figure 1C). These results suggest that, at the Yaa position, N-Glys are less stabilizing than Pro to varying degrees and that hydroxylation of Pro at the Yaa position results in exceptional stabilization for the triple helix beyond what could be achieved by the N-Glys.

Table 1.

Change in triple helical stability (as represented by ΔT_m) of N^xaaa-CMPs and N^yaaa-CMPs compared to the host sequence Ac-(GlyProHyp)₇-NH₂ measured in 1× PBS (pH 7.4).

Residue	ΔTm for GlyXaaHyp (°C)*	ΔTm for GlyProYaa (°C)**
Nrpe	No triple helix	Perturbed triple helix
Nval	−19	−6
Nspe	−7	−5
Nasp	−5	−18
Nala	−5	−16
NEt	−3	−11
Nhcys	−1	−11
Nmet	0	−11
Nasn	0	−15
Nlys	0	−10
Nleu	+3	−12
Ndxn	+4	−11
Ntyr	+6	−9
Nphe	+7	−9
Nnbz	+7	−17
Nchx	+8	−11

^*From Ref 41

^**Compared with Ac-(GlyProHyp)₇−NH₂, which has a T_m of 55 °C.⁴¹

Effects of side chain structures on the triple helical stability

Next, we examined the effect of the side chain structures on the stability of the triple helix. We first investigated the effects of hydrophobic side chains, since Goodman and coworkers suggested that hydrophobic interactions could be the primary driver for the triple helical stabilization of N-Glys with alkyl (e.g. N^xleu) and aryl side chains at the Xaa position.³⁷ Using the same host-guest peptide system, we determined the CD melting temperatures of N^yaaa-CMPs incorporating a series of N-Glys with natural (Figure 2A) and unnatural side chains with varying levels of hydrophobicity (Figure 2B). There was no correlation between the N-Gly’s side chain hydrophobicity and the T_m of the N^yaaa-CMP. Similar to our previous study at the Xaa position⁴¹, hydrophobic interactions are not a major factor contributing to the triple helical stability. This is most evident in the similar T_m values between N^ychx-CMP and N^ydxn-CMP which has contrasting hydrophobicity.

Figure 2.

Side chain chemistry of N-Gly does not have a strong correlation with triple helical stability. (A) The T_m values for a series of hydrophobic side chains are shown in order of increasing hydrophobicity, measured by reverse-phase HPLC using a neutral mobile phase (pH 7 at 25 °C), as reported by Kovacs and coworkers.⁶⁸ (B) The T_m values of the noncanonical side chains are similar except N^ynbz- and N^ytol-CMPs that have substituents at the ortho position. Comparing the side chains in (A) and (B), the triple helical stability has little dependency on the hydrophobicity or the bulkiness of the side chain. (C) In contrast to amino acids, the ionization state of Ν^y-Glys had little effect on triple helical stability except for N^yasp, which bears a charge closest to the peptide backbone. The experiments were performed in 1× PBS (pH 7.4), 0.1 M HCl (pH 1.3) for acidic residues, and 0.1 M NaOH (pH 13.0) for basic residues. The T_m values were averaged from two independent scans and rounded to the nearest integers. The T_m difference between the scans was less than ± 0.5 °C.

N^yphe-CMP and N^ytyr-CMP have similar T_m values, although their aromatic rings have different electron densities. Surprisingly, N^ynbz, an electron-deficient N^yphe derivative, produced a much less stable triple helix than N^yphe and N^ytyr (Figure 2B), although it produced CMP with a high T_m when incorporated at the Xaa position.⁴¹ We hypothesized that at the Yaa position, the nitro group in the ortho position is causing steric clashes that destabilize the triple helix. To test this hypothesis, we tested N^ytol, which has a structure similar to that of N^ynbz but is more electron-rich. The T_m of N^ytol-CMP was similar to that of N^ynbz-CMP (Figure 2B) and markedly lower than those of N^yphe- and N^ytyr-CMPs. In addition, deprotonation of N^ytyr (with 0.01 M NaOH, pH 11.9) which further increases the electron density of the aromatic ring had little effect on the T_m of N^ytyr-CMP (47 °C vs. 46 °C, Figures S4 and S5). Therefore, the position of the ring substituent, not the electronic properties of the aromatic ring, seems to affect the triple helical stability of N^yaaa-CMP containing aromatic side chains.

We also examined the effects of side-chain ionization. First, although Arg is one of the most stabilizing amino acids at the Yaa position, N^yarg was only moderately stabilizing because the T_m of N^yarg-CMP belonged to the middle of the T_m range for natural side chains (Figure 1C). Unlike their α-amino acid versions, the side-chain ionization of N^ylys and N^yglu had little effect on their T_ms (Figure 2C). Only N^yasp-CMP exhibited T_m differences between the protonated and deprotonated forms. The results suggest that, in the case of N-Glys at the Yaa position, the side chain charges can influence the triple helical stability only when charged groups are close to the CMP backbone.

Finally, the CD melting data indicated that the thermal stability of the N^yaaa-CMPs was indifferent to the bulkiness of the side chains, which contrasts with the trend observed for the N^xaaa-CMPs, where an increase in side chain bulkiness clearly elevated the T_m. T_m values of N^yEt-CMP, N^yleu-CMP, and N^ychx-CMP were essentially the same (T_m 43 – 44 °C), despite their drastic difference in side-chain bulkiness. In comparison, N^yval-CMP exhibited a relatively high T_m (49 °C). We speculate that N^yval may orient one of its methyl groups to occupy the space between two CMP strands, which limits random conformation sampling, resulting in reduced conformational entropy and, consequently, increased overall triple-helical stability (see subsequent sections for studies on Nα-Cβ branched N-Glys). Based on this observation, we conclude that the bulkiness of the side chains is not a major factor of triple helical stability in the case of N-Glys at the Yaa position. As explained below, when compared to Xaa position, Yaa position provides more space to accommodate the bulky N-Gly side chains, which diminishes their effect on the triple helical stability.

Crystal structures of N^yphe-CMP and N^ynbz-CMP

We chose N^yphe-CMP and N^ynbz-CMP for X-ray crystallographic analysis because they have a large difference in thermal stability despite their similarity in side chain structures. Both CMPs crystallized with a single triple helical molecule in the asymmetric unit. The triple helical structure of N^yphe-CMP and N^ynbz-CMP were determined to a resolution of 1.30 Å (Table S8). When models for both structures were refined (Table S8), they contained all peptide atoms, except the N-terminal acetyl group and Gly₁ residue of the leading strand in the N^yphe-CMP structure, which have high flexibility. The refined structures of N^yphe-CMP and N^ynbz-CMP are shown in Figure 3A.The average (ϕ, ψ) angles of Gly, Pro, and Hyp from both structures were (−71.3° ± 3.8°, 174.6° ± 3.0°), (−71.1° ± 5.3°, 163.9° ± 5.8°), and (−58.6° ± 4.6°, 148.6° ± 4.3°) respectively, which are consistent with those of N^xaaa-CMPs reported previously.⁴¹ The average (ϕ, ψ) angles of the N^yphe and N^ynbz residues were (−66.7° ± 2.9°, 159.1° ± 5.0°) and (−74.3° ± 1.2°, 162.2° ± 1.7°) respectively. Angles shifted from Hyp were more pronounced in N^ynbz than N^yphe suggesting that the triple helix becomes more unstable as the (ϕ, ψ) angles move away (in the direction of wider angles) from those of Hyp. On average, the angles shifted from Hyp by Δϕ = −11.9°, Δψ = +12.1°. Despite these substantial shifts from the typical collagen peptide geometry at the substituted position, no obvious distortions were observed in either crystal structure. The Pro-peptoid peptide bonds were in the trans configuration, similar to other peptide bonds in the triple helix, suggesting that N-Glys did not alter the backbone structure.

Figure 3.

N-Glys do not disrupt the supersecondary structure of the host CMPs. (A) Crystal structures of N^yphe-CMP and N^ynbz-CMP show a characteristic conformation of the collagen triple helix without noticeable structural distortion. The N^yphe and N^ynbz residues are labeled in green and magenta, respectively. Both residues adopt the trans-peptide bond. (B) The top panel is the Ramachandran plot showing (ϕ, ψ) angles of all residues in the crystal structures of N^yphe-CMP and N^ynbz-CMP. The (ϕ, ψ) angles of most residues are found in the PPII region of the Ramachandran plot. The general N-Gly structure with the backbone dihedral angles in the upper right quadrant is provided for reference. The lower panel shows the magnified Ramachandran plot of N^yphe-CMP and N^ynbz-CMP. The (ϕ, ψ) angle coordinates of the N^yphe and N^ynbz residues are near the Pro region.

Although the deviations of the N-Glys’ average (ϕ, ψ) angles did not distort the triple helical structure, these shifts imply destabilization of the triple helix conformation. The average ϕ and ψ values of up-puckered Pro^y from a quantum mechanical/molecular mechanical study of triple helical (ProProGly)₁₀ were −51.3° ± 1.1° and 149.6° ± 6.2°, respectively.⁴⁴ Furthermore, the preferred ψ angle at the Yaa position of the collagen triple helix for efficient n → π* interactions is 151° ± 4°.⁸ The N-Glys at the Yaa position in our crystal structures adopted (ϕ, ψ) angles between the Pro and Hyp regions, deviating from the average ϕ and ψ angles of Yaa residues in other CMP structures available in the PDB database (ϕ = −60.3° ± 7.6°, ψ = 143.0° ± 39.4°) by Δϕ = −11.9°, Δψ = +12.1° (Figures 3B and S6; Table S9). This suggests that N-Glys negatively impact the efficiency of the n → π* interactions resulting in triple helices that are comparatively more unstable than (GlyProHyp)₇. This is in contrast to N-Glys at the Xaa position, where the backbone dihedral angles are within the range of down-puckered Pro^x in triple helices.^{41, 43, 44} Furthermore, when Hyp is in the Yaa position, its hydroxyl group participates in hydration networks that contribute to triple helical stability.⁶⁹ Such hydration networks were absent in the crystal structures of N^yphe- and N^ynbz-CMPs (Figures S7 and S8), although some studies suggested that hydration network may not be important for triple helical stabilization.^{69, 70} Because the above data were extracted from static crystal structures and do not reflect all possible structural conformations in solution, we performed computational simulations to further dissect the factors affecting the triple helical stability contributed by the N-Glys.

Metadynamics simulations of N-Glys in the Yaa position

The metadynamics calculations were performed for the triple helical structures of N^yaaa-CMPs. We selected residues covering different side chains, including small aliphatic, branched aliphatic, alicyclic, aromatic, and hydrophilic side chains. We also conducted simulations of the host CMP, (GlyProHyp)₇, and Gly^y-CMP for comparison. Figures 4A and S9-a show the free-energy landscapes of the selected residues, and Figure S9-b shows the convergence of the simulations. We assigned regions to the energy landscapes following the Ramachandran nomenclature proposed by Hollingsworth and Karplus.⁷¹ Although both Hyp^y and Gly^y had the lowest-energy valleys distributed within the PP-II (P_II) region and slightly higher-energy valleys in the inverse γ-turn (γ') and bridge (δ) regions, due to additional flexibility, Gly^y also accessed an additional (ϕ, ψ) angle space encompassing the extended chains (ε) as well as mirror images of the PP-II helix (P_II') and bridge (δ') regions. N-Glys skewed the energy landscapes of Gly^y towards that of Hyp^y. We believe that N-Glys restrict conformational flexibility via steric hindrance between their side chains and carbonyl groups (Figure 4B), resulting in limited access to all allowed (ϕ, ψ) dihedral angles observed for Gly^y. The energy landscape of N^yala resembled that of Gly^y because its small side chain exerts less steric hindrance than the other peptoid residues. None of the N-Glys completely diminished the δ'- and P_II'-regions down to Hyp^y, suggesting that the N-Glys could sample conformers that did not have the lowest energy levels, thus suppressing global triple helical stability. These metadynamics simulation results are in agreement with our CD melting data that substituting Hyp with the peptoid residues lowers T_m for all peptoid residues. Nevertheless, we believe that, regardless of side chain bulkiness beyond a single carbon, N-Gly restricts conformational flexibility to a similar degree, though less effectively than Hyp, promoting triple helical stabilization.

Figure 4.

(A) Free energy landscapes (kcal/mol) of the guest residues Gly, Nala, Nleu, and Nchx, in the CMPs are compared with those of the native residue (Pro for Xaa position and Hyp for Yaa position). All residues had the lowest-energy valleys localized within the PP-II basin. However, those at the Yaa position also had additional energy wells in other regions. These results suggest that residues in the Yaa position could access more (ϕ, ψ) space than those in the Xaa position. The free energy landscapes of the residues in the Xaa position were adapted with permission from ref. 41, Copyright © 2021 American Chemical Society. (B) Newman projections of Ν-Glys and Gly in the Yaa position suggest that the presence of a side chain on the N^α atom restricts the rotation around the N^α−C^α bond via the steric interaction with the C=O of residue i+1 (Gly). This results in reduced conformational flexibility compared to that of Gly^y. (C) Residues in the Xaa and Yaa positions experience different molecular environments. The Xaa residue can have steric interactions with the H_Cδ of the succeeding Hyp, resulting in less accessible ψ angles. In contrast, the Yaa residue precedes Gly, which has no side chain. The amide H of Gly is buried in the helical core and hydrogen bonded to the C=O of Pro in an adjacent strand, which avoids potential steric interaction with the Yaa residue’s side chain. These factors may allow the Yaa residue to access more ψ angles than the Xaa residue.

In our previous work, metadynamics calculations showed that the energy distributions of N-Glys in the Xaa position and Pro^x were highly confined within the P_II region and had no energy wells in other regions of the Ramachandran plot, except for N^xala-CMP, as shown in Figure 4A.⁴¹ Interestingly, the energy landscapes of the same N-Glys at the Yaa position were not limited to the P_II region. Notably, the Hyp^y residue also had some conformers accessing other regions, although it has a pyrrolidine ring that restricts backbone conformations, similar to Pro^x. A plausible explanation for this discrepancy is the difference in steric interactions with the residues following the N-Gly. The residue at the Xaa position precedes Hyp. As shown in Figure 4C, the ψ angle of the Pro^x_i residue could be restricted by steric interactions between its C^β−Η and the C^δ−Η of the Hyp^y_i+1. A bulky N-Gly side chain at the Xaa position may also exhibit similar steric interactions, which limits the accessible ψ angles. On the other hand, Gly succeeds the residue at the Yaa position. Its N^α−Η forms a hydrogen bond with a Pro^x residue in an adjacent CMP strand and is buried within the triple helix. Therefore, potential steric interactions between Gly_i+1 residue and C^β−H of Hyp^y_i or an N-Gly side chain at the Yaa position are unlikely to occur. This allows Hyp^y_i and N-Glys in the Yaa position to access more conformations than those in the Xaa position.

Effect of N^α−C^β chiral residues on the triple helical stability of CMPs

N^α−C^β chiral side chains introduce chirality into N-Glys and can limit the conformational space accessible by the N-Gly backbone. Previously, we found that two model N^α−C^β chiral peptoid residues, Nspe and Nrpe (Figure 5A), destabilized the triple helix at the Xaa position. N^xspe-CMP (T_m: 48 °C) had a T_m noticeably lower than the unbranched N^xphe-CMP (T_m: 62 °C), and N^xrpe completely nullified the triple helical folding.⁴¹ When the two stereoisomers were tested at the Yaa position (Figure 5A), N^yspe-CMP (T_m: 50 °C) surprisingly formed a more stable triple helix than the N^yphe-CMP (T_m: 46 °C), with its T_m highest among all N-Glys tested at the Yaa position, whereas N^yrpe-CMP showed an unclear transition in its thermal melting profile, suggesting the loss of the triple helical structure. These findings showed that the N^α−C^β chirality of N-Glys influences triple helical folding differently at the Xaa and Yaa positions. Therefore, we conducted a set of experimental and computational investigations to understand this intriguing stereo-positional effect.

Figure 5.

Stereochemistry of N^α−C^β branched N-Glys affects the stability of collagen triple helices. (A) Nrpe and Nspe-CMPs have a methyl branch on their N^α−C^β atoms. The melting curve of N^yrpe-CMP does not show a clear two-state transition. In contrast, N^yspe-CMP folded into a more stable triple helix with a T_m 4 °C higher than that of N^yphe-CMP. The T_m values were averaged from two independent scans and rounded to the nearest integers. The T_m difference between the scans was less than ± 0.5 °C. (B) Molecular models of CMP triple helices containing N^yrpe (gray) and N^yspe (purple) residues. The N^α−C^β methyl group of N^yspe is positioned between the intrastrand Pro_i-1 residue and the Pro_i-1 in an adjacent CMP strand, which may restrict N^α−C^β bond rotation (highlighted in yellow). Potential clashes of intra-residue atoms for N^yrpe are indicated by red arrows and encircled in blue. The third CMP strands in both models were removed for clarity.

We first examined whether the inherent PP-II propensity of Nrpe and Nspe monomers influence the triple helical folding. We incorporated Nspe and Nrpe into the Zaa guest position of another host peptide with the sequence Ac-GlyProProZaaProProGly-NH₂ (N^zaaa-PP5) to compare their PP-II propensities. This host-guest system was previously used to assess the PP-II propensities of canonical amino acids and Ν-Gly peptoid monomers.^{41, 72} Surprisingly, N^zspe-PP5 showed a CD trace similar to that of the random coil, whereas N^zrpe-PP5 exhibited an unusual CD trace similar to a combination of PPI and PP-II helices with regions of negative and positive ellipticity evident around 204 nm and 223 nm, respectively (Figure S10). Nrpe and Nspe normally promote cis amide configuration; however a recent study showed that N-Gly peptoid monomers favoring cis configuration can form stable triple helices when incorporated into a sequence with high triple helical propensity [e.g. (GPO)_x].⁷³ These results indicated that factors other than PP-II propensity of the N-Gly were responsible for the stereochemical effects of the Nrpe and Nspe residues on the triple helical folding.

Three dimensional computational models of N^xr/spe-CMPs and N^yr/spe-CMPs were generated from the crystal structures of N^xphe-CMP (PDB entry 7JX5) and N^yphe-CMP, respectively, using UCSF Chimera.⁵⁵ The models revealed that the van der Waals radii of the α-methyl groups of N^xspe and N^xrpe residues overlapped with the OH group of Hyp_i+1 and the C=O group of Pro_i, respectively, in an adjacent CMP strand (Figure S11). Rotations of the N^α−C^β bond to avoid these clashes in N^xspe-CMP and N^xrpe-CMP led to new inter-strand steric clashes between the phenyl ring and Hyp residues, which explains their lower T_m compared to N^xphe-CMP. In the case of N^yrpe-CMP, the phenyl branch of the N^yrpe residue oriented towards the backbone clashing with its C=O group (Figures 5B and S12). In contrast, the (S)-phenylethyl side chain of the Nspe residue at the Yaa position did not have inter- or intra-strand steric clashes. The spatial arrangement of the N^yspe side chain appeared to restrict the N^α−C^β bond rotation and possibly favor CH•••π interactions between the aromatic ring and the intra-strand Pro_i-1 (Figure 5B) as observed in other peptides and proteins.⁷⁴ This may have led to the stabilization of the triple helix and an increase in the T_m by 4 °C.

To further understand the stereo-positional effects of N^xr/spe and N^yr/spe on triple helical folding (Figure 6), we conducted MD simulations of the N^(x/y)s/rpe-CMPs with focus on not only the backbone structure but also the side chain rotamers. All simulations were run for 20 ns, which was sufficient to achieve convergence of backbone and side-chain conformational sampling relevant to the analyses presented here (Figure S13). All analyses were performed on the equilibrated portion of the trajectories. The MD simulations showed that a number of N^xrpe and N^xspe rotamers adopted backbone ϕ and ψ angles outside the P_II region (Figure 6A), indicating the disruption of the PP-II and triple helices. N^xrpe caused more disruption than N^xspe, which is consistent with our previous experimental data.⁴¹ In contrast, the rotamers of N^yrpe and N^yspe were more localized in the P_II region. These results suggest that substitutions of Nrpe and Nspe at the Yaa position cause less triple helical perturbation than substitutions at the Xaa position. This is consistent with the CD thermal melting results of N^yspe being the most stabilizing residue among the N^xr/spe and N^yr/spe series. Additionally, the (ϕ, ψ) maps of Nrpe in the Xaa and Yaa positons were more spread out than those of Nspe, emphasizing that Nrpe is less compatible with the collagen triple helix than Nspe.

Figure 6.

Molecular dynamics (MD) simulations of Nrpe and Nspe showing the potential origins of stereo-positional effects of N^xr/spe and N^yr/spe in the triple helical stabilization. (A) The ϕ−ψ angle plots of N^xr/spe and N^yr/spe show the stereochemical and positional differences in backbone rotamer distributions. (B) The χ₁−χ₂ plots of the N^yphe, N^yspe, and N^yrpe side chains from the MD simulations demonstrate that the α-methyl branch restricts the rotation of the N^α−C^β bond, resulting in fewer rotamers adoptable by the phenyl ethyl side chains. Each blue dot in the dihedral angle plots in (A) and (B) represents one rotamer. The dark blue area indicates the highly populated region. The general N-Gly structure with dihedral angles of side chains is provided for reference. (C) The potential CH•••π interactions (as indicated by red dashed lines) between the phenyl ring and intrastrand Pro_i-1 are enabled by the restricted N^α−C^β bond rotation within N^yspe-CMP.

In addition to the backbone dihedral angles, we speculated that side chain conformations of N^yrpe and N^yspe may also contribute to the difference in the triple helical stabilization because their (ϕ, ψ) maps are generally similar. Many experimental results⁷⁵ have established that the Nspe and Nrpe residues favor a cis amide bond, and that peptoid side chains generally orient away from the backbone (χ₁ has the same sign as ϕ). In the context of the collagen triple helix, the amide is in the trans configuration and ϕ is necessarily around −75°, therefore the capacity of the side chain to accommodate these features may determine the stability of the construct. The three key contacts that these side chains can make are with the carbonyl oxygen, the C^β of the preceding Pro_i-1, and the C^δ of a Pro on an adjacent chain of the triple helix.

The MD simulation results for side chain conformations are in agreement with the aforementioned molecular modeling. As shown in Figures 6B and S14-S16, the N^yphe side chain is free to adopt more rotamers than N^yrpe and N^yspe but favors rotamers with χ₁ around −120°, which pack the phenyl ring against C^γ and C^δ of the adjacent chain. Both N^yrpe and N^yspe adopted rotamers with χ₁ values between 60° and 180°, which orient the phenyl ring away from the adjacent chain. These positive χ₁ values are generally preferred by Nrpe in other chemical structures⁷⁵, but they lead to clashes between the ortho carbon atoms of the aromatic ring and the carbonyl oxygen (Figures 5B, S12, and S15) in the context of the collagen triple helix. In N^yspe, however, one set of rotamers (χ₁ ~90°, χ₂ ~60°, ~−120°) avoids all of these clashes and comprises only rotamers populated in the simulations (Figures 6B and S16). Although both of these side chains show an effect of distorting their backbone ψ value away from its ideal value (~170°) to reduce the clashes between the aromatic C and C=O (Figures S15 and S16), ψ distortion is more pronounced in N^yrpe, which may have resulted in the perturbation of the overall helical conformation of CMP, as evidenced by the ill-defined melting transition.

Since computational experiments suggested that the N^yspe residue is conducive to triple helical folding and that a highly populated rotamer show potential CH•••π interactions between the aromatic ring and intra-strand Pro_i-1, we synthesized N^ysch-CMP which has phenyl ring of N^yspe-CMP replaced with cyclohexyl ring. The phenyl to cyclohexyl substitution reduced the T_m of this CMP by a 10 °C (T_m of N^ysch-CMP: 40 °C, Figures S3 and S17), suggesting that the aromaticity of the N^yspe residue plays a major role in stabilizing the triple helix. In contrast, the T_m remained the same for N^yrch-CMP and N^yrpe-CMP (Figures S3 and S17). These results support the implications from molecular modeling that the highly populated rotamers of N^yspe-CMP are stabilized by CH•••π interactions between the aromatic ring and intra-strand Pro_i-1 (Figure 6C). Typically, one CH•••π interaction can provide stabilization by 0.5−1.0 kcal/mol.⁷⁶ Although relatively weak, research showed that CH•••π interactions play a significant role in the stabilization of folded proteins.^{76, 77} N^yspe-CMP has three CH•••π interactions in proximity, which can work cooperatively in the folding process, resulting in enhanced triple helical stability. Thus, our work shows that Nspe stands out as the most stabilizing residue for the triple helix at the Yaa position, surpassing all other canonical and non-canonical amino acids except Hyp.

CONCLUSION

We demonstrated that at both Xaa and Yaa positions of the GlyXaaYaa repeat, moving the amino acid side chain from C^α to the amino group to produce N-Glys results in stabilization of the triple helix. At the Xaa position, some N-Glys were able to stabilize the triple helix more than Pro; however N-Glys at the Yaa position were less stabilizing than Pro and hydroxylation of Pro created exceptional stabilization for the triple helix beyond what could be achieved by natural amino acids or N-Glys. X-ray crystallography confirmed that the N-Glys at both positions are conducive to triple helical folding; however, it also showed clear difference in preferred backbone ϕ and ψ angles, which explained the different level of triple helix stabilization at the Xaa and Yaa position compared to Pro and Hyp, respectively. Metadynamic simulation demonstrated a more dispersed free energy landscape map [wider (ϕ, ψ) space] for N-Glys at the Yaa position than at the Xaa position, likely due to Gly being the residues after the Yaa. Chirality introduced at the N^α−C^β position had distinctively contrasting effects on triple helical folding at the Xaa and Yaa position, evidenced by extensive experimental and MD simulation data including dihedral angle plots of side chain rotamers. Finally, we discovered that chiral Nspe at the Yaa position can stabilize the triple helix much more than any other canonical or non-canonical residues, except for Hyp, via the intra-strand CH•••π interactions. This work expands our understanding of the N-Glys’s influence on triple helical folding at both Xaa and Yaa position and also provides new strategies for designing N-Glys that can not only produce large triple helix stabilization but also selective stabilization at the Xaa or Yaa positions.

Triple-helical collagen peptides have long served as essential tools not only for elucidating the molecular features responsible for collagen’s unique structure and diverse biological functions⁴² but also for the design of self-assembled nanomaterials^78–81, biomaterials^{82, 83}, and bioactive molecules.^{84, 85} The expanded side-chain diversity of the collagen triple helix made possible by our study now enables the unpresedented modulation of helix stability, intermolecular interactions, and biological recognition. This opens new opportunities to engineer collagen-inspired materials with programmable assembly, tunable mechanics, and tailored bioactivity for applications ranging from regenerative medicine to targeted therapeutics.