Fluctuating β‐Sheet Secondary Structure in DS119 Explains the Small Effects of Backbone N‐Amination on Thermal Stability

TL;DR

This study shows that backbone N-amination in a miniprotein's β-sheet doesn't increase thermal stability due to the region's high flexibility.

Contribution

The study reveals that β-sheet fluctuations limit the impact of backbone N-amination on protein stability.

Findings

N-amination of Trp9 or Phe33 in DS119 did not increase thermal stability as predicted.

MD simulations showed the β-sheet region is highly fluctuating with low hydrogen bond populations.

Altered conformational preferences from N-amination have minimal effects in disordered β-sheets.

Abstract

The miniprotein DS119 has been used as a model system to probe the effects of introducing residues with backbone N‐amination into a region of parallel β‐sheet secondary structure. Derivatives featuring backbone N‐methylation have also been synthesized for comparison. As expected, N‐methylation of Trp9 or Phe33, positioned on the outer edge of the β‐strands, led to a reduction in the thermal stability of the protein. However, contrary to predictions, the N‐amination of Trp9 or Phe33 did not lead to an increase in thermal stability of DS119. Refinement of the DS119 structure, using NOE restrained molecular dynamics simulations, shows that the β‐sheet region is highly fluctuating in nature. Key interstrand hydrogen bonds have populations of 24%–77%, whereas others have populations of less than 10%. In this disordered β‐sheet region, the altered conformational preferences arising from backbone N‐amination therefore have only minimal effects. This study demonstrates how MD simulation refinement can identify important dynamical features in a protein structure that might be overlooked in standard protein structure determination protocols.

Residues with backbone N‐amination have been introduced into a region of parallel β‐sheet secondary structure. Contrary to predictions, the N‐amination of Trp9 or Phe33 did not give an increase in the thermal stability of DS119. MD simulations suggest this is due to the highly fluctuating nature of the β‐sheet region.

Figures

Backbone trace of structure 1 (most representative structure) of the reported NMR solution structure of DS119 (PDB 2KI0). The residues selected for backbone N‐amination are shown in pink (Trp9, Phe33, Trp34).

Synthesis of N‐amino dipeptide building blocks and incorporation into modified DS119 miniprotein.

Representative plot of the fraction folded as a function of temperature determined by CD thermal melts for wild type DS119 (1) and the backbone amide substituted variants 2–6 at 0.2 mg/mL in 20 mM aq sodium phosphate, pH 7.3, 25°C. The mean residue ellipticity was monitored at 220 nm, and fitted curves were derived from nonlinear regression.

Full CD wavescans (upper panel) and representative plot of fraction folded as a function of temperature (lower panel) determined by monitoring the mean residue ellipticity at 220 nm for wild type DS119 (1) and the variants with N‐amination of Trp9 (2) and Phe33 (4) at 0.2 mg/mL in 20 mM aq sodium phosphate and 1 M Gdn‐HCl, pH 7.3, 25°C. Fitted curves were derived from nonlinear regression.

Secondary structure elements as a function of time calculated for the IRNOE_54A7 (upper panel) and TANOE_54A7 (lower panel) simulations of DS119. Red: α‐helix; blue: β‐strand or β‐bridge; brown: turn.

Tables

TABLE 1. Backbone‐modified DS119 miniprotein analogues.

Entry	X a	Sequence	% yield
1	—	4 QVRTIWVGGTPEELKKLKEEAKKANIRVTFWGD 36	9
2	aTrp9	4 QVRTIXVGGTPEELKKLKEEAKKANIRVTFWGD 36	6
3	MeTrp9	4 QVRTIXVGGTPEELKKLKEEAKKANIRVTFWGD 36	14
4	aPhe33	4 QVRTIWVGGTPEELKKLKEEAKKANIRVTXWGD 36	10
5	MePhe33	4 QVRTIWVGGTPEELKKLKEEAKKANIRVTXWGD 36	21
6	aTrp34	4 QVRTIWVGGTPEELKKLKEEAKKANIRVTFXGD 36	4

TABLE 2. Number of NOE distance bound violations in the simulations of DS119 (total number of NOE distance bounds: 597).

Simulation	Size of NOE distance bound violation (nm)
	0.05–0.10	0.10–0.15	0.15–0.20	0.20–0.25	0.25–0.30	> 0.3
MD_54A7	6	8	1	5	2	13
MD_54A8	13	6	9	3	1	31
SD_54B7	8	6	3	2	2	9
IRNOE_54A7	1	0	0	0	0	0
IRNOE_54A8	1	0	0	0	0	0
IRNOE_54B7	0	0	0	0	0	0
TANOE_54A7	0	0	0	0	0	0
TANOE_54A8	0	0	0	0	0	0
TANOE_54B7	0	0	0	0	0	0

TABLE 3. Population of main chain hydrogen bonds in the β‐sheet region of DS119. The hydrogen bonds that define a regular parallel β‐sheet are shown in bold. For the NMR ensemble, the number of structures with the hydrogen bond in the ensemble of 20 structures is given. For the simulations, the populations are given as percentages through the simulations.

Unrestrained simulations and ensemble of NMR structures
Hydrogen bond	Simulation			NMR ensemble
	MD_54A7	MD_54A8	SD_54B7
8NH‐30O	82.7	/	75.0	0/20
10NH‐32O	25.4	/	30.2	0/20
12NH‐34O	/	26.2	18.5	0/20
12NH‐35O	/	1.0	70.1	3/20
30NH‐6O	56.9	/	87.3	7/20
32NH‐8O	31.6	/	9.0	0/20
34NH‐10O	14.8	47.8	45.9	17/20
35NH‐10O	/	3.7	8.2	1/20
36NH‐11O	/	/	/	2/20

Full text

Introduction

1

Backbone N‐amination provides a way to introduce new functionalities into a polypeptide chain [1, 2, 3]. The modified amino acids alter hydrogen bonding patterns, main chain conformational preferences, and stability to proteolysis while maintaining the diversity of native side chains [2, 4]. This approach shows great promise for the rational design of new proteomimetics [3]. For example, peptides including α‐hydrazino acid residues have already been shown to block amyloidogenic aggregation of Aβ_42_ and tau protein [5, 6]. In addition, the introduction of N‐aminated residues into gramicidin S has been found to enhance broad‐spectrum antimicrobial activity [7].

We previously demonstrated that Cα‐substituted α‐hydrazino acid residues increase the folded population of antiparallel β‐hairpin peptides [2, 3]. N‐amination of outer edge amides in these model systems stabilizes extended backbone dihedral angles through cooperative noncovalent interactions [3, 4]. In this work, we tested whether these strand‐stabilizing effects translate to parallel β‐sheet structure using the miniprotein DS119 as a model tertiary fold. DS119 is a de novo‐designed miniprotein that has been explored for use in biomedical research and protein engineering [8, 9, 10]. For example, the protein has been used as a scaffold for the design of a novel tumor necrosis factor alpha binding protein [11]. A metal‐binding site has also been introduced into DS119 [12]. In addition, the protein has been used as a model system for understanding the mechanism of protein structure‐stabilizing osmolytes [13, 14] and the folding pathways of designed proteins [15, 16, 17, 18, 19, 20].

The design of DS119 initially used secondary structure preferences and geometrical restrictions to identify a sequence predicted to adopt a βαβ tertiary fold, consisting of an amphipathic helix linked to two parallel β‐strands formed from the N‐ and C‐terminal regions [8]. Inclusion of leucine, valine, and isoleucine residues gave a hydrophobic core between the secondary structure units, which was subsequently refined using computational methods [8]. A pair of tryptophan residues was introduced into the parallel β‐strands to form a tryptophan zipper [21], and charged residues were added to the protein surface to discourage aggregation.

Experimental studies of DS119 showed that it has high thermal stability, a circular dichroism (CD) spectrum characteristic of an α/β fold, and two‐state cooperative unfolding [8]. NMR structure determination for DS119 [8] reported short‐ and medium‐range NOEs characteristic of an α‐helical conformation for residues 15–26, strong αH‐NH(i,i + 1) NOEs characteristic of an extended β‐strand conformation for the parallel β‐sheet region, and long‐range NOEs consistent with the designed hydrophobic core. In addition, Trp9 and Trp34, introduced to form the tryptophan zipper WW interaction, were found to pack together. However, the upfield‐shifted aromatic NMR resonances normally associated with an edge‐to‐face interaction were not observed [8].

In this work, we investigated DS119 as a model system for characterizing the effects of backbone N‐amination on parallel β‐sheet structure. This system also provides the opportunity to evaluate these effects in a more complex tertiary structure where the β‐sheet strands are not directly linked by a turn or loop region. In addition to experimental studies, MD and SD simulations of DS119 were performed. These give a refined structure of the protein that provides a framework for interpreting the experimental data.

Materials and Methods

2

Solution‐Phase Dipeptide Synthesis

2.1

Solvents and reagents were purchased from commercial suppliers and used as received unless otherwise specified. 2‐(t‐Butyl)‐3,3‐diethyl‐1,2‐oxaziridine‐2,3,3‐tricarboxylate (TBDOT) was synthesized according to our previously reported procedure, which is a modification to the original protocol from Armstrong and coworkers [1, 22]. Thin‐layer chromatography (TLC) was performed using Merck 60 F254 silica gel precoated glass‐backed plates (0.25 mm). Flash chromatography was performed using F60 silica (40–63 μm, 230–400 mesh). Reaction progress was monitored by TLC analysis (single spot/two solvent systems) using a UV lamp, bromocresol green, CAM (ceric ammonium molybdate), ninhydrin, or basic KMnO_4_ stain(s) for detection purposes. NMR spectra were recorded on a Bruker 400 or 500 MHz spectrometer. Proton chemical shifts are reported as δ values relative to residual signals from deuterated solvents (CDCl_3_ or DMSO‐d 6). Analytical HPLC chromatograms were acquired with a reversed‐phase column (C12, 150 mm × 4.6 mm, 4 μm, 90 Å) using linear gradients of MeCN in H_2_O (mobile phases modified with 0.1% formic acid) over 20 min, and spectra are provided for λ = 220 nm. HRMS spectra were acquired using a Bruker Impact II ESI‐QTOF.

The α‐amino acid benzyl ester (H‐Phe‐OBn⋅HCl or H‐Trp(Boc)‐OBn⋅HCl, 1.1 equiv) was suspended in a 1:1 biphasic mixture of THF and sat aq NaHCO_3_ and was vigorously stirred at room temperature. After 10 min, the mixture was treated with 1.0 equiv of TBDOT [1], and the reaction was stirred at room temperature. After 3 h, the reaction was treated with ethylenediamine (5 equiv). After 5 min, the reaction was diluted with EtOAc, and the aqueous layer was separated. The organic layer was washed with water and 1 M aq HCl, then dried over anhydrous Na_2_SO_4_, filtered, and concentrated to yield the Boc‐protected α‐hydrazino benzyl ester. A solution of Fmoc‐Ile‐OH, Fmoc‐Phe‐OH, or Fmoc‐Thr(tBu)‐OH (1.2 equiv) in DCM was treated with Ghosez's reagent and stirred for 10 min. The solution was then transferred to a flask containing a mixture of the Boc‐protected α‐hydrazino benzyl ester and NaHCO_3_ (10 equiv) in DCM. The reaction was stirred for 3 h and quenched with water. The organic layer was collected, and the aqueous layer was extracted with DCM. The combined organic layers were dried over anhydrous Na_2_SO_4_, filtered, and concentrated to yield the protected α‐hydrazino dipeptide benzyl ester. This benzyl ester was dissolved in 1:1 EtOAc/2‐propanol and treated with 10% Pd/C (200 mg/mmol). The reaction was placed under H_2_ atmosphere (balloon) and stirred at room temperature until TLC monitoring indicated consumption of the benzyl ester starting material (3 h). The reaction was quenched with MeOH, filtered through a Celite pad, and concentrated. The crude materials were purified by flash chromatography (EtOAc/hexanes) to afford the protected α‐hydrazino dipeptide acid as a crystalline solid.

Peptide Synthesis

2.2

Automated SPPS was carried out on a CEM Liberty Blue or PurePep Chorus peptide synthesizer using Fmoc‐Asp(OtBu)‐Wang resin (0.23 mmol g^−1^ loading). The following derivatives suitable for Fmoc SPPS were used: Fmoc‐Ala‐OH, Fmoc‐Arg(Pbf)‐OH, Fmoc‐Asn(Trt)‐OH, Fmoc‐Glu(tBu)‐OH, Fmoc‐Gln(Trt)‐OH, Fmoc‐Gly‐OH, Fmoc‐Ile‐OH, Fmoc‐Leu‐OH, Fmoc‐Lys(Boc)‐OH, Fmoc‐Phe‐OH, Fmoc‐Pro‐OH, Fmoc‐Thr(tBu)‐OH, Fmoc‐Trp(Boc)‐OH, Fmoc‐Val‐OH, Fmoc‐(N‐Me)Phe‐OH, Fmoc‐(N‐Me)Trp(Boc)‐OH, Fmoc‐Ile‐(Boc)aTrp(Boc)‐OH, Fmoc‐Thr(tBu)‐(Boc)aPhe‐OH, and Fmoc‐Phe‐(Boc)aTrp(Boc)‐OH. Fmoc deprotection steps were carried out by treating the resin with a solution of 20% (v/v) piperidine/DMF at rt (5 min), then at 50°C (2 min). After Fmoc deprotection, the resin was washed with DMF (4×). Coupling of the Fmoc‐protected building blocks was achieved using 5 equiv HCTU (0.25 M in DMF), 10 equiv NMM (1 M in DMF), and 5 equiv Fmoc‐protected amino acid (0.2 M in DMF) at 50°C (10 min). Deprotection and coupling steps were repeated until peptide synthesis was complete, and then a final Fmoc deprotection of the N‐terminus was performed. The resin was then transferred to a suitable vessel, washed with DCM (3×) and MeOH (3×), and dried under vacuum. Cleavage from solid support and global deprotection was effected by incubating the dried resin in 5 mL of TFA/H_2_O/TIPS:acetone (37:1:1:1) for 3 h. The resin was filtered, and the filtrate was collected in a 50‐mL centrifuge tube. The resin was washed with TFA (5 mL) and filtered, and the combined filtrate was evaporated under a stream of N_2_. The concentrated filtrate was precipitated by the addition of cold Et_2_O (45 mL, −20°C). The mixture was centrifuged, and the supernatant was decanted. The pellet was dried thoroughly under vacuum. All peptides were dissolved in H_2_O purified by preparative HPLC with a reversed‐phase column (C12, 250 mm × 21.2 mm, 5 μM, 90 Å) using a linear gradient of MeCN in H_2_O (mobile phases modified with 0.1% formic acid) over 30 min and then lyophilized to afford the peptides as white powders.

CD

2.3

All peptides were prepared by dissolving lyophilized powder in 20‐mM sodium phosphate buffer (pH 7.3) at a concentration of 0.2 mg/mL. CD spectra were acquired using a JASCO J‐1500 CD spectrometer in a 1‐mm pathlength quartz cell with 2‐s digital integration time, 1‐nm band width, 0.1‐nm data pitch, and a scan speed of 50 nm/min at 25°C. All ellipticity values are reported as a mean across three scans. Mean residue ellipticity (MRE) at a given wavelength (deg cm^2^ dmol^−1^ residue^−1^) was calculated based on the equation MRE = 𝜃/(10 × 𝑏 × 𝐶 × 𝑛), where 𝜃 is ellipticity (mdeg), 𝑏 is pathlength (cm), 𝐶 is the concentration of peptide (mol L^−1^), and 𝑛 is the total number of residues. Temperature‐dependent CD spectra were acquired using the same parameters from 25°C to 95°C in 1°C increments with a ramp rate of 1°C per minute and monitoring the MRE at 220 nm. Nonlinear regression curves and melting temperatures were calculated according to the method described by Shortle and coworkers [23]. Melting temperatures are reported as the mean ± standard deviation across two independent denaturation experiments.

MD and SD Simulations

2.4

All simulations were carried out with the GROMOS biomolecular simulation software (http://www.gromos.net) [24, 25]. The GROMOS 54A7 and GROMOS 54A8 force‐field parameter sets were used for the MD simulations in explicit solvent [26, 27, 28]. The GROMOS 54B7 force field with the SASA implicit‐solvation term and the set of solvation parameters given in Pechlaner et al. [29] was used for the SD simulations in implicit solvent [30, 31]. All simulations were started from Model 1 in the DS119 NMR structure (PDB code 2KI0) [8]. To model the pH 6 conditions from the NMR structure determination, all Asp and Glu side chains and the protein C‐terminus were unprotonated, and Lys and Arg side chains and N‐terminus were protonated. For the explicit solvent simulations, the protein structure was solvated in a rectangular box, and minimum image periodic boundary conditions were applied. The minimum solute‐box wall distance was set to 1.2 nm, giving 6140 simple point charge (SPC) water molecules [32]. Two chloride ions were added to achieve overall neutrality of the system.

In each MD simulation, an initial equilibration scheme comprising five 20‐ps simulations at temperatures of 60, 120, 180, 240, and 298 K was used. During the first 80 ps of this equilibration, the solute atoms were harmonically restrained to their positions in the starting structure with force constants of 25,000, 2500, 250, and 25 kJ mol^−1^ nm^−2^, respectively. Following equilibration, each simulation was run at 298 K and at a constant pressure of 1 atm for 100 ns (unrestrained simulations: MD_54A7 and MD_54A8) or 50 ns (NOE restrained simulations: IRNOE_54A7, IRNOE_54A8, TANOE_54A7, and TANOE_54A8), the temperature and pressure being maintained using the weak coupling algorithm [33], with relaxation times of τ _ T _ = 0.1 ps and τ _ p _ = 0.5 ps and an isothermal compressibility of 4.575 × 10^−4^ (kJ mol^−1^ nm^−3^)^−1^. Protein and solvent were separately coupled to the heat bath. The SHAKE algorithm [34] was used to constrain bond lengths and the geometry of the water molecules, with a relative geometric tolerance of 10^−4^, allowing for an integration time step of 2 fs. The center of mass motion was removed every 1000 time steps. Nonbonded interactions were calculated using a triple‐range cutoff scheme [35] with cutoff radii of 0.8 and 1.4 nm. Interactions within 0.8 nm were evaluated every time step, and intermediate‐range interactions were updated every fifth time step. To account for the influence of the dielectric medium outside the cutoff sphere, a reaction‐field force [36] with a relative dielectric permittivity ε of 61 [37] was used.

For the SD simulations (unrestrained SD_54B7 and NOE restrained TANOE_54B7 and IRNOE_54B7), the protein was slowly heated using the same protocol as was used for the protein in explicit solvent. After equilibration, 200‐ns SD simulations were performed with a reference temperature of 298 K, maintained by the Langevin equations or thermostat and by weak coupling to a heat bath (τ _ Τ _ = 0.1 ps), the latter to control the temperature of atoms that have a friction coefficient equal to zero, whose temperature is thus not controlled by the Langevin thermostat [29].

For the simulations with NOE restraints [38], the NOE list was taken from the NMR data deposited with the DS119 structure in the Protein Data Bank (PDB code 2KI0) [8]. Only the unique, nonambiguous NOE restraints were used (597 restraints). Pseudo‐atom NOE distance bound corrections were used [39]. Either instantaneous NOE restraining was used with a force constant of 2000 kJ mol^−1^ nm^−2^ (IRNOE_54A7, IRNOE_54A8, and IRNOE_54B7) or time‐averaged NOE restraining (TANOE_54A7, TANOE_54A8, and TANOE_54B7) was used with a force constant of 6000 kJ mol^−1^ nm^−2^ and a memory relaxation time of 20 ps.

Analysis of all the simulations was performed with the GROMOS++ suite of analysis programs [40], using trajectories written to disk every 5 ps. The secondary structure assignment was done with the program DSSP, based on the Kabsch–Sander rules [41]. Hydrogen bonds were identified according to a geometric criterion: A hydrogen bond was assumed to exist if the hydrogen–acceptor distance was smaller than 0.25 nm and the donor–hydrogen–acceptor angle was larger than 135°.

Results

3

Introduction of α‐Hydrazino Acid Residues Into DS119

3.1

The initial sites selected for modification were Trp9, Phe33, and Trp34 near the Trp zipper motif (Figure 1). We hypothesized that installation of N‐amino substituents on the outer edges of the parallel β‐sheet in DS119 (Trp9, Phe33) would enhance thermal stability. In contrast, N‐amination of the inward‐facing hydrogen‐bonded amide of Trp34 would likely destabilize the protein fold.

Backbone trace of structure 1 (most representative structure) of the reported NMR solution structure of DS119 (PDB 2KI0). The residues selected for backbone N‐amination are shown in pink (Trp9, Phe33, Trp34).

A dipeptide building block approach was employed to synthesize the N‐aminated DS119 variants. The modified building blocks were prepared by electrophilic amination of the respective α‐amino benzyl esters, coupling to Fmoc‐protected amino acid chlorides, and hydrogenolysis to afford the Fmoc‐protected N‐amino dipeptides (Scheme 1) [15]. Although DS119 was previously prepared via heterologous expression, here, the parent sequence (1) and backbone‐modified variants (2, 4, and 6) were synthesized using Fmoc SPPS on Wang resin (Scheme 1). Analogues 3 and 5 were also prepared to compare the effect of outer‐edge N‐methylation to that of N‐amination (Table 1).

Synthesis of N‐amino dipeptide building blocks and incorporation into modified DS119 miniprotein.

First, the overall secondary structure of parent peptide 1 was characterized by far‐UV CD in sodium phosphate buffer. The CD signature of 1 showed close agreement with the reported spectrum obtained from recombinant wild‐type DS119 (Figure S1), with a positive maximum at 195 nm and a broad negative minimum between 208 and 222 nm. This indicates that chemically synthesized DS119 retains the βαβ structure of the recombinantly expressed protein. Backbone amide substitutions at Trp9 (2 and 3) showed slightly altered conformations. The N‐amino substituted aTrp9 analogue 2 showed a more intense positive band and emergence of more distinct minima at 208 and 225 nm. The N‐methylated MeTrp9 analogue 3 exhibited decreased intensity at every extremum, indicating a gradual transition to a random coil structure. Incorporation of aPhe or MePhe at position 33 (4 and 5) resulted in only minimal perturbation to the wild‐type CD signature. As expected, substitution of Trp34 for aTrp34 in 6 resulted in decreased CD band intensities relative to wild‐type DS119, suggesting a destabilized fold due to modification of a hydrogen‐bonded amide NH.

The thermal stability of each analogue was then measured by monitoring the MRE at 220 nm from 25°C to 95°C (Figure 2). In agreement with previous reports, unmodified DS119 (1) exhibited high thermal stability, with no clear unfolding transition below 95°C [1]. N‐Aminated analogues 2 and 4 also did not show a cooperative unfolding transition within the measured temperature range. In contrast, the N‐aminated aTrp34 variant 6 exhibited a T _ m _ of 82.5°C, consistent with disruption of a cross‐strand hydrogen‐bond between Tyr34 and Val10 in the wild‐type structure. N‐Methylated analogues 3 and 5 were the least thermally stable of the synthesized miniproteins (T _ m _ = 74.4°C and 78.7°C, respectively). This indicates that N‐methylation of outer edge amides in DS119 is significantly more disruptive to parallel β‐sheet structure relative to N‐amination.

Representative plot of the fraction folded as a function of temperature determined by CD thermal melts for wild type DS119 (1) and the backbone amide substituted variants 2–6 at 0.2 mg/mL in 20 mM aq sodium phosphate, pH 7.3, 25°C. The mean residue ellipticity was monitored at 220 nm, and fitted curves were derived from nonlinear regression.

Coupled thermal/chemical denaturation experiments were then performed in 1 M guanidinium hydrochloride (Gdn‐HCl) to more accurately quantify stability differences between 1, 2, and 4. Full wavescans of each miniprotein were initially acquired to confirm that the addition of 1 M Gdn‐HCl does not significantly disrupt the CD signature (Figure 3A). The only major artefact from the additional chemical denaturant was some signal interference below 210‐nm wavelengths. The MRE at 220 nm was then monitored between 25°C and 95°C for 1, 2, and 4. Under these conditions, each miniprotein exhibited two‐state cooperative unfolding transitions and a quantifiable melting temperature well within the measured range (Figure 3B). Miniprotein 2, with amination on the first β‐strand at Trp9, exhibited similar thermal stability to unmodified 1 (T _ m _ = 61.0°C ± 0.6°C and 62.0°C ± 0.1°C, respectively). Miniprotein 4, featuring amination at Phe33, was less thermally stable (T _ m _ = 51.0°C ± 1.1°C).

Full CD wavescans (upper panel) and representative plot of fraction folded as a function of temperature (lower panel) determined by monitoring the mean residue ellipticity at 220 nm for wild type DS119 (1) and the variants with N‐amination of Trp9 (2) and Phe33 (4) at 0.2 mg/mL in 20 mM aq sodium phosphate and 1 M Gdn‐HCl, pH 7.3, 25°C. Fitted curves were derived from nonlinear regression.

MD and SD Simulations of DS119

3.2

As the N‐amination of Trp9 and Phe33 in the parallel β‐sheet region of DS119 did not result in the hypothesized enhancement of the thermal stability, an NOE restrained MD and SD refinement of the protein structure was performed to understand the structure in more detail. MD simulations of DS119 were performed using the GROMOS explicit solvent force fields 54A7 [27] and 54A8 [28]. To compare the results of our simulations with those reported in the original de novo design study of DS119 [8], we also carried out SD simulations in implicit solvent using the force field 54B7 [27] with an added solvent‐accessible‐surface‐area implicit solvent term [30, 31]. With each force field, an unrestrained simulation and two NOE‐restrained simulations were run using instantaneous and time‐averaged NOE restraining, respectively [38]. All simulations were initiated from the published NMR structure of DS119 (PDB code 2KI0) using Model 1 of the ensemble, which is identified as the most representative conformer [8].

The unrestrained MD simulations in explicit solvent using the 54A7 and 54A8 force fields (MD_54A7 and MD_54A8, respectively) showed a significant level of unfolding. The center of the α‐helix unfolded in the MD_54A7 trajectory, and the β‐sheet was completely lost in the MD_54A8 trajectory. The secondary structure is maintained in the implicit solvent unrestrained SD simulation with the 54B7 force field (SD_54B7). However, comparison of all the unrestrained simulations with the experimental NOE data showed a considerable number of very large NOE violations, greater than 0.3 nm (13, 31, and 9 violations > 0.3 nm for the simulations MD_54A7, MD_54A8, and SD_54B7, respectively; Table 2). The majority of these large NOE violations involve the aromatic side chains of Trp9, Phe33, and Trp34, suggesting that the aromatic side chain packing in the starting structure may not be sufficiently well‐defined to stabilize the protein conformation throughout the simulations.

We therefore ran simulations of DS119 including NOE restraints to produce a refined structure of the protein. Simulations were run with all three force fields with both instantaneous and time‐averaged NOE restraining (instantaneous NOE restraining simulations IRNOE_54A7, IRNOE_54A8, and IRNOE_54B7 and time‐averaged NOE restraining simulations TANOE_54A7, TANOE_54A8, and TANOE_54B7 for simulations run with the 54A7, 54A8, and 54B7 force fields, respectively). The experimental NOEs were very well satisfied in these simulations, with no NOE violations greater than 0.05 ns in the TANOE_54A7, TANOE_54A8, IRNOE_54B7, and TANOE_54B7 simulations and only one violation greater than 0.05 nm (15NH‐13NH) in the IRNOE_54A7 and IRNOE_54A8 simulations (Table 2).

In all NOE‐restrained simulations in explicit solvent, residues 14–26 form an α‐helix for at least 90% of the simulation duration. The persistent α‐helix is slightly shorter in the implicit solvent simulations running from residues 17–26 for at least 90% of the simulations, due to lower populations of the 17NH‐13O, 18NH‐14O, and 19NH‐15O hydrogen bonds. However, the population of the β‐sheet secondary structure for residues 6–10 and 30–34 is low in all NOE‐restrained simulations ranging from 0% to 25%. Although most of these residues adopt extended conformations for most of the trajectories, the populations of interstrand hydrogen bonds are low. The time series of the secondary structure present through the simulations show the transient nature of the β‐sheet throughout the simulations (Figure 4 and Supporting Information).

Secondary structure elements as a function of time calculated for the IRNOE_54A7 (upper panel) and TANOE_54A7 (lower panel) simulations of DS119. Red: α‐helix; blue: β‐strand or β‐bridge; brown: turn.

Table 3 gives the main chain hydrogen bonds present in the β‐sheet region of the original ensemble of NMR structures and in each MD simulation. Interestingly, in most of the 20 structures in the NMR ensemble, there is only one or two interstrand hydrogen bonds present, 30NH‐6O and 34NH‐10O, rather than the six expected for a canonical parallel β‐sheet. In a few of the NMR structures, there are additional hydrogen bonds in this region, namely 12NH‐35O, 35NH‐10O, and 36NH‐11O. However, these are not expected for a regular β‐sheet. In the NOE‐restrained MD simulations in explicit solvent, the six hydrogen bonds expected for a parallel β‐sheet are observed. Hydrogen bonds 8NH‐30O, 30NH‐6O, and 34NH‐10O have the highest populations (24%–77%), whereas the other three have populations of 1%–8%. The patterns of hydrogen bonds seen in the implicit solvent simulations are similar, but with significant populations (33%–56%) of the additional hydrogen bonds 12NH‐35O and 35NH‐10O, which would not be expected in a typical parallel β‐sheet. Thus, the NOE‐refined structures of DS119 have given a more regular parallel β‐sheet for residues 6–10 and 30–34 than was observed in the original NMR structures, but with a highly fluctuating nature.

Analysis of the side chains of Trp9 and Trp34 shows that these two side chains remain parallel to each other, with a distance of around 0.45 nm throughout all NOE‐restrained simulations. Hence, the interactions between the sidechains introduced to form the Trp‐zipper WW interaction are maintained with the same parallel interaction seen in the reported NMR ensemble, rather than the typical edge‐to‐face interaction. This is consistent with the absence of experimental upfield‐shifted aromatic proton NMR resonances [8].

Discussion

4

During the design of DS119, the miniprotein was found to have a molten globule character before the introduction of the tryptophan zipper motif [8]. The results reported here show that although this tryptophan zipper stabilized the folded structure by bringing the N‐ and C‐termini sections of the chain together, it did not result in the formation of a persistent parallel β‐sheet. Instead, although the two regions of the chain remain in proximity, the hydrogen bonds between them are highly fluctuating. Presumably, the favorable conformational entropy from this flexibility compensates for the reduced hydrogen bonding interactions.

We introduced N‐amino substituents onto the outer‐edge backbone amides of the parallel β‐sheet of DS119 (at Trp9 and Phe33), expecting that these would enhance the thermal stability of the protein. However, coupled thermal/chemical denaturation experiments in 1 M guanidinium hydrochloride showed that the N‐aminated DS1119 variant incorporating aTrp9 has similar thermal stability to wild‐type DS119, whereas the aPhe33 DS119 variant has reduced thermal stability. The simulation data serve to shed light on these results. Previous studies have suggested that residues with N‐amination favor β‐sheet‐like φ, ψ torsion angles, in part, through short‐range hydrogen bonds involving the hydrazine NH_2_ group [4]. However, in this case, the N‐aminated residues have been introduced into a region of fluctuating secondary structure where hydrophobic interactions of the aromatic sidechains are likely to be far more significant than any changes in fluctuating main chain hydrogen bonds. Changes in the backbone conformation may have altered the projection of the Phe33 sidechain and hence its interaction with other hydrophobic residues leading to the observed reduction in stability. However, the tryptophan zipper interaction involving Trp9 is clearly maintained as this derivative shows a similar level of stability to the wild‐type protein.

The observed 3D structure of a given protein results from a complex interplay between a wide range of noncovalent interactions, which contribute via both enthalpy and entropy terms to the overall stability of the structure. The network of interactions present can often allow a range of different dynamical processes to occur within folded protein structures ranging from the motions of individual side chains and disordered loop regions to hinge bending and local unfolding events. It is being increasingly recognized that these dynamical properties of proteins can have significant functional roles [42, 43, 44]. However, most protein structure determinations start from the assumption that the protein adopts a single well‐defined conformation and only considers conformational flexibility where there are clear indications in the available experimental data. However, as in the case of DS119, this can result in certain important properties of a protein structure being overlooked. MD simulations provide an approach for refining experimentally determined protein structures. For DS119, NOE‐restrained MD simulations have identified the highly fluctuating nature of the parallel β‐sheet region, highlighting the importance of considering conformational disorder and entropic contributions to the stability of designed proteins.

Funding

L.J.S. thanks St. Hilda's College, University of Oxford, for financial support. J.R.D. thanks the National Science Foundation (CHE2109008) for funding, and S.K.S. thanks the National Institutes of Health (T32 GM145773) for fellowship support.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Far‐UV CD wavescans of the backbone‐substituted DS119 variants at each modified site compared with wild‐type DS119. Figure S2: Secondary structure elements as a function of time calculated for the IRNOE_54A7 and TANOE_54A7 simulations of DS119. Figure S3: Secondary structure elements as a function of time calculated for the IRNOE_54B7 and TANOE_54B7 simulations of DS119.