Chiral Inversion of Amino Acids in Anti-parallel β-sheets at Interfaces Probed by Vibrational Sum Frequency Generation Spectroscopy

Abstract

Parallel study of protein variants with all (L-), all (D-) or mixed (L-)/(D-) amino acids can be used to assess how backbone architecture versus sidechain identity determines protein structure. Here, we investigate the secondary structure and sidechain orientation dynamics of the anti-parallel β-sheet peptide LK₇β (Ac-Leu-Lys-Leu-Lys-Leu-Lys-Leu-NH₂) composed of all (L-), all (D-), or alternating (L-Leu)/(D-Lys) amino acids. Using interface-selective vibrational sum frequency generation spectroscopy (VSFG), we observe that the alternating (L-)/(D-) peptide lacks a resonant C-H stretching mode compared to the (L-) and (D-) variants, and does not form anti-parallel β-sheets. We rationalize our observations based on density functional theory calculations and molecular dynamics (MD) simulations of LK₇β at the air-water interface. Irrespective of the handedness of the amino acids, leucine sidechains prefer to orient toward the hydrophobic air phase while lysine sidechains prefer the hydrophilic water phase. These preferences dictate the backbone configuration of LK₇β and thereby the folding of the peptide. Our MD simulations show that the preferred sidechain orientations can force the backbone of a single strand of (L-) LK₇β at the air-water interface to adopt β-sheet Ramachandran angles. However, denaturation of the β-sheets at pH = 2 results in negligible chiral VSFG amide I response. The combined computational and experimental results lend critical support to theory that chiral VSFG response requires macroscopic chirality, such as in β-sheets. Our results can guide expectations about the VSFG optical responses of proteins, and should improve understanding of how amino acid chirality modulates the structure and function of natural and de novo proteins at biological interfaces.

Graphical Abstract

Introduction

The enantiomeric enrichment of (L-) chirality amino acids placed significant restrictions on the chemical diversity of proteins in the pre-biotic world. However, (L-) and (D-) amino acids are occasionally incorporated together in natural systems. For example, the antibiotic peptide gramicidin A consists of fifteen alternating (L-) and (D-) amino acids. The peptide adopts the uncommon β-helix-type secondary structure with all sidechains on one face of the strand, allowing gramicidin A to function as a membrane cation channel.^1–2 The incorporation of both (L-) and (D-) amino acids in natural proteins requires a complex biosynthetic machinery acting along non-ribosomal pathways.^3–4 This raises questions concerning the unique biophysical properties of heterochirality that would justify expanding “Nature’s alphabet” to include (D-) amino acids.⁵

(D-) amino acids have been used in the design and synthesis of unnatural proteins.^6–10 Some of the resulting proteins with (D-) amino acids show promise as anti-viral and anti-cancer therapeutic agents, since those unnatural proteins are more resistant to proteolysis, may exchange their chiral recognition in biomolecular interactions, and modulate the biophysical properties of cellular structures.^{9, 11} For example, the antimicrobial peptide sapesin B composed of only (D-) leucine and lysine has higher antimicrobial activity than all (L-) sapesin B.¹² The improved therapeutic potential has been attributed to distinct interactions between (D-) sapesin B and the cell wall of Staphylococcus bacteria. At the molecular level, recent computational work employing all-atom molecular dynamics has systematically studied the impact of mixed (L-) and (D-) chirality on a-helices.¹³ The results showed that the (L-) to (D-) chiral inversion of even a single amino acid can locally break α-helix secondary structure. An outstanding challenge is to understand the perturbations induced by chiral inversion leading to disruption of protein secondary structure.

Vibrational sum frequency generation spectroscopy (VSFG) is a second-order (χ⁽²⁾) vibrational spectroscopy that can distinguish between protein secondary structures.^14–23 VSFG is a three-wave mixing process that irradiates the sample with pulses of visible and IR light with frequencies ω_vis and ω_IR, respectively, and detects the vibrationally-resonant sum frequency response at frequency ω_SFG = ω_vis + ω_IR. Due to non-linear effects of the three-wave mixing process, the VSFG technique is highly sensitive and selective for vibrational probes in anisotropic environments, such as interfaces, and suppresses the background signal from the bulk.^{14, 18, 24–28} When combined with computational modeling, VSFG can provide information on molecular orientation at interfaces.^27–33 The SFG, visible, and IR beams can each be polarized parallel (p) or perpendicular (s) to the incident plane, yielding eight possible polarization combinations. Comparison of SFG intensities using at least two different polarization combinations can, for example, be used to assess interfacial orientation of molecules, such as lipid tails and alkyl chains when probing the CH₃ stretching bands.^{29, 34–38}

In this paper, we combine achiral and chiral VSFG spectroscopy to probe the perturbations of amino acid chiral inversions on the overall chiral secondary structure of anti-parallel β-sheets at interfaces. The experimental data is interpreted through density functional theory (DFT) calculations of the VSFG spectra and protein backbone geometry and bond orientations in molecular dynamic (MD) simulations of the anti-parallel β-sheet at the air-water interface. We find that the VSFG technique can distinguish all (L-) or all (D-) anti-parallel β-sheets from a variant that contains alternating (L-) and (D-) chirality along the length of the peptide backbone, proving the power of VSFG as a vibrational chiroptical spectroscopy. At the molecular level, we find that the presence of an amphiphilic interface induces a preferential orientation of hydrophobic and hydrophilic amino acid sidechains to be exposed to the air or water phases, respectively. These preferred orientations in the peptide sequence of LK₇β confine the Ramachandran angles to β-sheet-like configuration even for a single strand of (L-) LK₇β, providing an invaluable molecular system to critically examine the chiral SFG theory.^39–42 The mixed (L-)/(D-) chirality conforms to the preferences of sidechain orientation and thereby exerts steric constraints that prevent the peptide backbone from forming anti-parallel β-sheets. Revealing this novel aspect of how chirality and chemical properties of side chains can modulate the secondary structure and orientation of proteins at interfaces exemplifies the advantages of employing both experiments and theoretical analyses for interpreting VSFG optical response of peptides and proteins. Moreover, our findings should offer insights into de novo design of unnatural proteins with both (L-) and (D-) amino acids with potential applications for engineering interfaces of biomaterials.

Methods

Sample Preparation

The LK₇β (Acetyl-Leu-Lys-Leu-Lys-Leu-Lys-Leu-NH₂, MW: 896.24 g/mol) with all (L-) amino acids was synthesized by GL Biochem Ltd. (Shanghai). The LK₇β with all (D-) or (L-)/(D-) amino acids (i.e., Acetyl-(L-)Leu-(D-)Lys-(L-)Leu-(D-)Lys-(L-)Leu-(D-)Lys-(L-)Leu-NH₂) were synthesized by AnaSpec, Inc. (Fremont, CA, USA). The lyophilized powder samples were dissolved in deionized H₂O (pH =7 or pH = 2, HCl adjusted) at 1 mM. The solutions were prepared as 100 uL aliquots, frozen in liquid nitrogen, and stored at −80 °C until use.

For VSFG experiments at the air-glass interface, plain glass microscope slides (Thermo Scientific, Portsmouth, NH, USA; Cat No. 420–004T, 101616–3) were plasma-cleaned on “high” for 3 minutes in a plasma cleaner (Harrick Plasma, Ithaca, NY, USA; PDC-32G). One aliquot of the LK₇β was thawed and pipetted onto the glass slides. The samples dried overnight at room temperature, forming hydrated thin-films at the air-glass interface.

For VSFG experiments at the air-water interface, 200 μL of 1mM LK₇β solution was applied with a glass syringe (Hamilton Company, Reno, NV, USA; Model No. 1705N) at the surface of 3.8 mL of deionized water, giving a final concentration of LK₇β of 50 μM.

VSFG Measurements

The VSFG spectra were collected with a home-built setup previously described.⁴³ For achiral VSFG measurements, we centered the broad-band IR beam at 3300 nm and used the ssp polarization configuration (s-polarized sum frequency, s-polarized visible, and p-polarized IR). For chiral amide I/amide II VSFG measurements, we centered the broad-band IR beam at 6100 nm and used the psp polarization configuration. Spectra taken at the air-glass interface were collected with acquisition times of 10 minutes (unless otherwise noted). Spectra taken at the air-water interface were collected with acquisition times of 20 minutes.

The reported IR frequencies are recovered by calibrating the VSFG response to a polystyrene standard (Buck Scientific, East Norwalk, CT, USA; 0.05 mm film). Background spectra were collected by shuttering the IR beam. The background spectra were then subtracted from the VSFG intensity spectra. For VSFG spectra at the air-glass and air-water interfaces, the background-subtracted spectra were normalized to the IR power envelope by dividing by the VSFG response of a GaAs crystal. The VSFG spectra were manually cleaned of cosmic ray intensities, which are unrelated to the VSFG response.

Finally, the intensity spectra were fit to a Lorentzian function that describes the resonant vibrational mode being probed and a contributing non-resonant term:

$${\mathit{I}}_{\mathit{S}\mathit{F}\mathit{G}}\propto {\left|{\mathit{\chi }}_{\mathit{N}\mathit{R}}^{(\mathbf{2})}\mathbf{+}{\sum }_{\mathit{q}}\frac{{\mathit{A}}_{\mathit{q}}}{{\mathit{\omega }}_{\mathit{I}\mathit{R}}\mathbf{-}{\mathit{\omega }}_{\mathit{q}}+\mathit{i}{\mathbf{\text{Γ}}}_{\mathit{q}}}\right|}^2$$ (1)

where I_SFG is the intensity of the sum frequency generation, ${\mathit{\chi }}_{\mathit{N}\mathit{R}}^{(\mathbf{2})}$ is the non-resonant second order susceptibility, ω_IR is the frequency of the IR beam, A_q is the amplitude of the q^th resonant vibrational mode, ω_q is the frequency, and Г_q is the half-width half maximum.

Molecular Dynamics

Molecular dynamics simulations of all (L-), all (D-) and (L-Leu)/(D-Lys) LK₇β on the water/air interface were performed using the all-atoms CHARMM36/CMAP force field⁴⁴ and the TIP3P water model potential.⁴⁵ The force field parameters for the (D-) amino acids were taken from the residue topology ‘toppar_all36_prot_d_aminoacids.str’ that includes inverted CMAP tables from the corresponding (L-) amino acids. All simulations were performed using the NAMD software package⁴⁶ in the NVT ensemble using a Langevin thermostat with a time constant of 1 ps. Hydrogen bonds were kept fixed using the SETTLE algorithm,⁴⁷ which allows the use of a 2 fs time step. Periodic boundaries conditions in three dimensions were applied using a cutoff of 12 Å for the real-space Coulombic and van der Waals interactions, whereas long-range interactions were handle using the particle-mesh Ewald method (PME).

The initial LK₇ polypeptides strands were built in the β-conformation by tacking backbone dihedral angles (ϕ,ψ) of (−120,113) for L-amino acids and (120,−113) for D-amino acids, using the Molefacture plugin in VMD package.⁴⁸ The sidechains of the lysine residues were taken to be charged. The N-terminal and C-terminal of the polypeptide chains were capped with −NH₂COCH₃ (acetylated N-terminus) and −CONH₂ (amidated C-terminus) groups respectively. From these polypeptide strands, β sheets were assembled in antiparallel conformation using 2 strands.

The β sheets were placed on the surface of a preequilibrated TIP3P water slab with the hydrophobic leucine residues facing the air. The water slab consists of 4855 to 4893 water molecules arranged in a ~35 Å thick slab of 70×70 Ų surface area, with ~35 Å of vacuum in the direction perpendicular to the slab (defined as the z direction). In order to maintain charge neutrality, random waters were replaced by Cl⁻ atoms. The combined protein-water system was allowed to relax using conjugate gradient algorithm for 5000 steps to eliminate any close contact interaction between atoms, followed by an 820 ps simulation where the system was heated from 30 K to 298 K. An additional 400 ps simulation at 298 K was carried out to equilibrate the system at the target temperature. For all (L-) and all (D-) LK₇β, a 20 ns production simulation was used to extract dihedral angles and angular distribution using snapshots every 1 ps. For (L-Leu)/(D-Lys) LK₇β, longer simulations of 60 ns were performed. Results for the all (L-) and all (D-) LK₇β correspond to averages over three independent simulations with different initial configurations, whereas results for the (L-Leu)/(D-Lys) LK₇β model correspond to averages over five simulations.

Calculation of VSFG Spectra

VSFG spectra were computed for an antiparallel β-sheet model composed of two LK₃ peptide, with the primary sequence Leu-Lys-Leu. This peptide represents a minimal model with the characteristics of the LK₇β peptide. Geometry optimizations and harmonic frequency analysis at the DFT level were performed with the Gaussian 2009 software package,⁴⁹ using the B3LYP⁵⁰ hybrid functional and the 6–31G(d) basis set.⁵¹ An “ultrafine” integration grid (99 radial shells and 590 angular points per shell) was used for the frequency calculations to obtain accurate results.

Our approach for simulating SFG spectra has been previously described.^{17, 20, 52–58} Here, we describe it only briefly. The calculations involves the determination of the second-order molecular hyperpolarizabilities ${\beta }_{ijk,q}^{(2)}\propto \frac{\partial {\alpha }_{ij}}{\partial Q_q}\cdot \frac{\partial {\mu }_k}{\partial Q_q}$ where α_ij and μ_k (with i, j, k = a, b, c) are elements of the polarizability and dipole moment in the molecular frame, respectively, and Q_q is the normal coordinate of the q-th vibrational mode. The hyperpolarizability ${\beta }_{ijk,q}^{(2)}$ is then rotated to the laboratory frame (x, y, z) and averaged over the azimuthal angle ϕ in 5° increments to obtain the second-order susceptibility ${\chi }_{IJK,q}^{(2)}={\sum }_{ijk}\langle R_{Ii}\cdot R_{Jj}\cdot R_{Kk}\rangle \cdot {\beta }_{ijk,q}^{(2)}$ where R_Ii represent elements of the ZYZ Euler rotation matrix.¹⁸ The susceptibilities ${\chi }_{IJK,q}^{(2)}$ are then used to compute the effective susceptibilities, which for the ssp and psp polarization combinations take the form^{18, 27, 29}

$${\chi }_{ssp,q}^{(2)}=L_{yyz}{\chi }_{yyz,q}^{(2)}$$ (2)

$${\chi }_{psp,q}^{(2)}=L_{zyx}{\chi }_{zyx,q}^{(2)}-L_{xyz}{\chi }_{xyz,q}^{(2)}$$ (3)

where L_yyz are Fresnel factors that depend on the refractive index of the interface as well as the incident angle of the lights.^{18, 27, 29} The Fresnel factors used in this study are listed in Table S2. From these susceptibilities, the homodyne SFG spectra is given by

$$I_{SFG}\left({\omega }_{IR}\right)\propto {\left|{\chi }^{(2)}\right|}^2={\left|A_{NR}{e}^{i\varphi }+k{\sum }_q\frac{{\chi }_q^{(2)}}{{\omega }_{IR}-{\omega }_q+i{\text{Γ}}_q}\right|}^2$$ (4)

where k is a multiplicative constant. All harmonic frequencies were scaled by 0.943 to facilitate comparisons with experiments whereas Γ_q and A_NR were set to 7.5 cm⁻¹ and 0, respectively.

Results

Achiral and Chiral VSFG of (L-), (D-), and (L-Leu)/(D-Lys) LK₇β.

We previously demonstrated by VSFG amide I response that the (L-) LK₇β peptide spontaneously forms anti-parallel β-sheets at interfaces.⁵⁴ The LK₇β peptide localizes to the air-water interface due to alternating hydrophobic leucine and hydrophilic lysine residues.⁵⁹ Figure 1 shows the achiral (ssp) VSFG spectra of (L-), (D-) and (L-Leu)/(D-Lys) LK₇β variants at the air-water interface. The (L-) LK₇β shows three peaks in the C-H stretching region (Figure 1a) in agreement with the spectrum reported by Somorjai and co-workers.⁶⁰ The peaks are centered at 2875 cm⁻¹, 2913 cm⁻¹ and 2938 cm⁻¹. Their assignments are aided by computational studies as discussed in the next section (Computations of VSFG Spectra). The (L-) LK₇β and (D-) LK₇β spectra are indistinguishable (Figure 1b, green and gray). However, the achiral VSFG spectrum for (L-Leu)/(D-Lys) LK₇β is distinguished by significant suppression of the peak at 2913 cm⁻¹ (Figure 1b, blue). Suppression of this peak is also observed at the air-glass interface (Figures 2a and 2b). In contrast, linear vibrational spectroscopy such as attenuated total reflectance-Fourier transform infrared does not distinguish between the three LK₇β species in the C-H stretching region (Figure S1, Section S1 in Supporting Information).

Figure 1.

VSFG of (L-), (D-), and (L-Leu)/(D-Lys) LK₇β at the air-water interface in the C-H stretching region. The spectra of (L-) and (D-) LK₇β show identical profiles whereas the (L-Leu)/(D-Lys) variant shows strong suppression of the peak at 2913 cm⁻¹. The spectra are measured using the ssp (achiral) polarization combination. Solid colored lines are fits to the spectra based on equation 1. Solid black lines are the component peaks of the fits.

Figure 2.

VSFG of (L-), (D-), and (L-Leu)/(D-Lys) LK₇β at the air-glass interface in the C-H stretching region. Similar to the air-water interface, the peak at 2911 cm⁻¹ is suppressed in the (L-Leu)/(D-Lys) variant. The spectra in (a) and (b) are measured using the ssp (achiral) polarization combination. The spectra in (c) and (d) are measured with the psp (chiral) polarization combination. Solid colored lines are fits to the spectra based on equation 1. Solid black lines are the component peaks of the fits.

We also probed LK₇β variants at the air-glass interface using chiral (psp) VSFG, which is sensitive to anisotropic orientation of local chiral moieties as well as macroscopic chirality.^{23, 41–42} The chiral VSFG response can report on formation of higher-order chiral structures of biological macromolecules, for example, α-helix or β-sheet secondary structure of proteins and hybridization of double-helix DNA.^{19, 23, 61} The (L-) LK₇β peptide shows at least four peaks in the C-H stretching region (Figure 2c), in agreement with our prior report of chiral VSFG spectra of LK₇β at the air-water interface.⁶² The peak positions, centered at 2869 cm⁻¹, 2918 cm⁻¹, 2965 cm⁻¹ and 2988 cm⁻¹ were previously assigned to the CH₃ symmetric stretch, CH₂ asymmetric stretch, CH₃ asymmetric stretch, and C_α-H stretching modes, respectively.⁶² The (D-) LK₇β spectrum is very similar and overall indistinguishable (Figure 2d, green and gray). In contrast, chiral response for (L-Leu)/(D-Lys) LK₇β was not observed (Figure 2d, blue). These observations suggest that (L-Leu)/(D-Lys) LK₇β at the air-glass interface neither assumes a configuration in which local chiral moieties produce a coherent VSFG response, nor adopts a chiral secondary structure.

We have also analyzed the effect of pH by preparing (L-) LK₇β at the air-glass interface at neutral pH = 7 and denaturing pH = 2. We find that the peak at 2913 cm⁻¹ is clearly not suppressed at pH = 2 (Figure 3a). Furthermore, we measured the chiral amide I/amide II response of (L-) LK₇β at the air-glass interface at both pH = 7 and pH = 2. We have previously shown that the chiral amide I response is a useful probe of the protein secondary structure at interfaces.^{16, 18} For LK₇β, the peaks at 1563 cm⁻¹, 1618 cm⁻¹, and 1687 cm⁻¹ are assigned to the amide II, amide I (B₂), and amide I (B₁) vibrational modes associated with anti-parallel β-sheets, respectively (Figure 3b, left).^{18, 54} However, the chiral amide I/amide II spectrum measured with the psp polarization is almost completely abolished at pH = 2 (Figure 3b, right). This suggests that the anti-parallel β-sheet character of the peptide is mostly lost under denaturing conditions. Altogether, these results show that the loss of anti-parallel β-sheet structure is not correlated with suppression of the peak at 2913 cm⁻¹ as observed for the (L-Leu)/(D-Lys) LK₇β spectrum (Figures 1b and 2b). Finally, Figure 3c shows the chiral amide I/amide II spectrum measured for the different LK₇β variants at pH = 7. The (L-) and (D-) LK₇β spectra show strong chiral response, while no VSFG signal is observed for (L-Leu)/(D-Lys) LK₇β.

Figure 3.

Achiral and chiral VSFG of LK₇β at the air-glass interface prepared using native pH = 7 and denaturing pH = 2 conditions. (a,b) Loss of anti-parallel β-sheet structure at pH = 2 induces loss of chiral amide I/amide II signal, which is not correlated with suppression of the C-H stretching peak at 2913 cm⁻¹. (c) (L-Leu)/(D-Lys) LK₇β does not show chiral VSFG signal in the amide I/amide II region. The spectra in (a) are measured using the ssp (achiral) polarization combination. The pH = 2 spectra in (a) was scaled by 2x. The spectra in (b) and (c) are measured with the psp (chiral) polarization combination.

The suppression of the VSFG response of the resonant vibrational mode at 2913 cm⁻¹ for (L-Leu)/(D-Lys) LK₇β (Figures 1 and 2) and our finding that the suppression of this mode is not correlated with loss of anti-parallel β-sheet structure (Figure 3) prompts some interesting questions. What is the identity of the vibrational normal mode at 2913 cm⁻¹? Why is this mode suppressed in heterochiral versus homochiral LK₇β variants? How do chiral inversions inhibit or promote protein secondary structure at interfaces? To address these research questions in molecular detail, we used DFT to recapitulate the VSFG response of (L-) LK₇β. We also performed MD simulations to understand the effects of chiral inversions on protein secondary structure and sidechain dynamics of (L-), (D-) and (L-Leu)/(D-Lys) LK₇β at the air-water interface.

Computations of VSFG spectra

To approximate the VSFG response of (L-) LK₇β and assign the peak at 2913 cm⁻¹, we performed DFT calculations on an antiparallel β-sheet model composed of two (L-) LK₃ peptides with primary sequence Ac-Leu-Lys-Leu-NH₂ (Figure 4a). The model peptide represents a minimal ‘building block’ with the characteristics of (L-) LK₇β amenable to ab initio calculations. While treatments based on second-order perturbation theory (VPT2)^63–66 or vibrational self-consistent field (VSCF)^67–69 methods are available to include overtones and combination bands in the description of the spectra, the dimension of the system under consideration makes such calculations infeasible. Nevertheless, we find that an anharmonic analysis comprising Fermi resonances (FRs) on single leucine and lysine residues including couplings between modes^70–72 is sufficient for understanding the C-H stretch spectral region (Section S2 in Supporting Information).

Figure 4.

(a) Molecular representation of the antiparallel β-sheet all (L-) (LK₃)₂ model peptide, defined in the molecular frame a, b, c. DFT-based (b) ssp and (c) psp SFG spectra. Shaded areas correspond to symmetric CH₂ (purple), symmetric CH₃ (red), asymmetric CH₂ (green) and asymmetric CH₃ (cyan) stretching modes.

Figure 4b shows the calculated ssp VSFG spectrum of the antiparallel β-sheet model for a molecular orientation with the c-axis perpendicular to the surface plane (axis coordinates are shown in Figure 4a). This orientation exposes the hydrophobic leucine sidechains to the hydrophobic phase and the charged lysine sidechains to the hydrophilic phase. This is the dominant structure of the anti-parallel β-sheet at the air-water interface (see next section). The spectrum is characterized by a broad peak centered around 2875 cm⁻¹ attributed to the symmetric CH₃ stretching (normal mode analyses are presented in Section 3 in Supporting Information). This peak has shoulders at both sides due to symmetric and asymmetric CH₂ stretching (note that a shoulder possibly appears around 2850 cm⁻¹ in the experimental spectrum in Figure 1). Based on this analysis, the peak at 2875 cm⁻¹ in the experimental spectra in Figures 1 and 2 can be assigned to symmetric CH₃ stretch. Allowing for the contributions of FRs of the symmetric CH₃ and CH₂ stretching of leucine and lysine sidechains (Supporting Information) and given the strong intensity of the symmetric CH₃ and CH₂ peaks, from which intensity of the FR is borrowed, we assign the peak at 2913 cm⁻¹ to FR of the symmetric CH₂ stretching and the peak at 2938 cm⁻¹ to CH₃ FR. The lysine Cα-H is coupled to the CH₂ FRs and possibly also contributes in the experimental spectrum. This assignment, based on DFT calculations, is in agreement with previous studies of the LK₇β polypeptide as well as individual leucine and lysine residues.^{22, 33, 60, 73–75} It should be noted that the symmetric CH₃ stretching band of the capping group appears at ~2893 cm⁻¹ and might also contribute to the spectral signals. We cannot presently discard the contribution of this capping group to the experimental spectra. However, the reasonable agreement in peak position and relative intensity between the calculated and experimental spectra, as well as previous SFG studies of LK₇β and related systems such as LK₁₄α, leucine, and lysine amino acids,^{22, 33, 60, 73–77} suggest that the contribution of the capping to the SFG signal should be small.

In Figure 4c we present the chiral spectra for the model peptide. The simulated and experimental spectra (Figure 2c) agree very well. The spectra consist of four bands, including symmetric CH₃, asymmetric CH₂, asymmetric CH₃ and Cα-H stretching modes.⁶² As noted above, the asymmetric CH₂ stretching modes that appear in the region 2900–2930 cm⁻¹ are highly coupled to the chiral Cα-H of the lysine residues. The higher frequency peaks comprise the stretching of the leucine chiral Cα-H moieties (with some possible overlap with the asymmetric CH₃ stretching of the acetyl capping group). For the spectrum in Figure 4c, it is difficult to assess the effects of FRs in the chiral SFG signal. Further studies are required.

Molecular Dynamics and Orientation Analysis of (L-), (D-), and (L-Leu)/(D-Lys) LK₇β.

To obtain molecular detail on the interfacial behavior of the LK₇β system, we complement the VSFG experiments and calculations with MD simulations. The simulated systems consist of a water slab with (L-), (D-), or (L-Leu)/(D-Lys) LK₇β placed at the air-water interface (Figure 5).

Figure 5.

Typical snapshots of the molecular dynamics simulations of LK₇β at the air-water interface. (a,b) Side view and top view of (L-) LK₇β. (c,d) Side view and top view of (L-Leu)/(D-Lys) LK₇β.

We first characterize the backbone conformation of polypeptides with different chirality by computing free energy profiles along the backbone dihedral angles ϕ and ψ. In Figure 6a we present the free energy profile for (L-) LK₇β composed of two strands in an antiparallel β-sheet conformation at the air-water interface. The free energy landscape presents a well spanning the region −160°<ϕ<−80° and 100°<ψ<150°, characteristic of β-sheet conformations.⁷⁸ An analysis of the free energy profile per residue (Figure S5) also corroborates the presence of β-sheet conformations. In this conformation all the hydrophilic lysine residues are placed on one side of the sheet and are solvated by the water phase; all the hydrophobic leucine residues point into the air (Figure 5a). In fact, a simulation test in which the LK₇β is started with the leucine residues pointing into the water and the lysine residues into the air, shows that the system naturally evolves to invert the orientations of the residues (Figure S7).

Figure 6.

Free energy (k_BT units) as a function of the dihedral angles ϕ and ψ for (a) (L-) (two-strands), (b) (D-) (two-strands), (c) (L-Leu)/(D-Lys) (one-strand) and (d) (L-) (one-strand) and LK₇β.

Figure 6b shows the Ramachandran free energy profile for a two-strand (D-) LK₇β system. The free energy landscape obtained for this system is a mirror image of the one for the (L-) LK₇β (Figure 6a), and represents the β-sheet conformational region for (D-) amino acids.⁷⁸

We found that a system of two strands of (L-Leu)/(D-Lys) LK₇β does not form stable anti-parallel β-sheets as the strands break apart over the course of the simulation (Figure S8, Section 6 in Supporting Information). Indeed, previous experimental and computational studies showed that inversion of chirality at a particular placement along a peptide chain leads to a consequent break in secondary structure.^{13, 78–83} Our MD results also suggest that the (L-Leu)/(D-Lys) LK₇β do not form any stable secondary structure, consistent with the VSFG silent chiral amide I/amide II spectrum (Figure 3c). Hence, for the (L-Leu)/(D-Lys) LK₇β we only present results obtained with one polypeptide strand (Figure 5c,d). Figure 6c shows the free energy profile for the one-strand (L-Leu)/(D-Lys) LK₇β system. The free energy landscape is richer and more complicated than for the (L-) or (D-) LK₇β variants and demonstrates a more complex behavior of the polypeptide, which dynamically samples various configurations over the course of the simulation (Figure S6). The periodic alternation of chirality in the (L-Leu)/(D-Lys) LK₇β system represents a huge perturbation in the “natural” behavior of the polypeptide, not allowing for the formation of local secondary structure motifs.

To further highlight the effect of chiral inversion on the dynamics and structure of peptides, we also performed simulations of a single strand of (L-) LK₇β on the air-water interface. By comparing this system with the (L-Leu)/(D-Lys) LK₇β and the two strands of (L-) LK₇β it is possible to disentangle the effect of chiral inversion from the absence of a β-sheet. The free energy profile of the backbone for this system is presented in Figure 6d. The conformation adopted by the single strand of (L-) LK₇β does not present such a complex landscape as the one adopted by the (L-Leu)/(D-Lys) LK₇β (see also Figure S6), demonstrating that the inversion in chirality has a more profound effect than just disrupting the secondary structure of the peptide. Quite remarkably, while the isolated one strand of (L-) LK₇β displays a richer landscape than for the system with two strands (Figure 6a), the single strand is predominantly characterized at the air-water interface by a stable backbone geometry consistent with β-strand conformation. Since it is not possible to form inter-strand hydrogen bonds, the single strand must be stabilized by the interface itself. Indeed, the hydrophobic leucine sidechains continue to point towards the air phase, while the hydrophilic lysine residues point towards the water phase (Figures S9 and S10).

The Ramachandran free energy plots presented in Figure 6 are informative with respect to the effect of chirality on the conformation of the peptide backbone but do not give information of the orientation of the sidechains with respect to the interface, which is the information gathered from achiral VSFG spectra.¹⁸ To characterize the behavior of the sidechains we computed the tilt angle (θ) of certain bonds with respect to the surface normal. A schematic representation of the representative selected bonds is presented in Figure 7.

Figure 7.

(Left) Schematic representation of leucine (top) and lysine (bottom) residues with colored definition of bond vectors used to describe the orientation with respect to the surface normal. Angular distributions of selected bond vectors of leucine (a) and lysine (b) for (L-) LK₇β. Angular distributions of selected bond vectors of leucine (c) and lysine (d) for (L-Leu)/(D-Lys) LK₇β. Shaded areas correspond to standard deviation of the distributions. Dashed lines correspond to an isotropic distribution of the angles.

In Figure 7a we present the tilt angle distributions over all the leucine residues for two strands of (L-) LK₇β. Average values and standard deviations for tilt angles obtained from these distributions are presented in Table S1. The angle θ of the C_α-C_β bonds (blue) presents a peaked distribution with a maximum near 0° (cosθ = 1), which demonstrate the extreme preference of leucine hydrophobic sidechains to point towards the air. The tilt orientation of Cα-H bonds (black) present a broad distribution centered around θ = 90° (cosθ ≈ 0) and is consistent with the formation of anti-parallel β-sheets with hydrogen bonds between C=O---H-N lying on the plane of the water surface (Figure S11). The isopropyl group of leucine residues, characterized by the Cγ-H bond (red), presents a broad distribution centered around 120° (cosθ ≈ −0.52), implying a strong ordering at the interface as suggested by previous studies.^{19, 33, 60, 74, 84–86} Note that since the principal eaks at 2875 cm⁻¹ and 2938 cm⁻¹ of the ssp SFG spectra in the C-H stretch region are ascribed to the methyl groups of the leucine residue (symmetric stretching and FR, respectively), the peaked distribution of Cγ-H supports the presence of an SFG-active ordered interface. The θ distributions for (D-) LK₇β are essentially the same compared to the results for (L-) LK₇β (Figure S9).

The distributions of the tilt angles over all the lysine residues in (L-) LK₇β are presented in Figure 7b. The lysine sidechains preferentially face the water phase, as can be appreciated from the peaked C_α-C_β distribution (blue) at θ = 180°. The distribution of C_α-H angles (black) is centered at θ = 90° consistent with the formation of anti-parallel β-sheets. The C_βH₂ moiety (green) is ordered. Note that the distribution of the C_γH₂ moieties (red) is bimodal, with peaks at θ ≈ 120° (cosθ ≈ −0.47) and θ ≈ 65° (cosθ ≈ 0.40). This implies that the alkyl chains are not fully ordered in perfect trans conformation and present some gauche defects, consistent with their VSFG optical activity.³⁵ Again, the θ distributions for (D-) LK₇β are unchanged (Figures S10 and S11).

The results for the tilt angle for one strand of (L-Leu)/(D-Lys) LK₇β peptide are presented in Figures 7c and 7d. The leucine and lysine sidechains still point up and down, respectively, maintaining the tendency of hydrophobic or hydrophilic residues to point into the air or water phase (blue curves). The biggest contrasts with (L-) LK₇β are the distributions of lysine C_γH₂ and C_α-H (Figure 7d, red and black curves). Both distributions become nearly isotropic, indicating that these modes should not be observed in a VSFG experiment.

Discussion

To the best of our knowledge, there have been no reports of VSFG being applied to study the effects of mixed (L-)/(D-) chirality on protein secondary structure. Here, we used achiral and chiral VSFG in combination with computational studies of protein dynamics to investigate biophysical properties of (L-Leu)/(D-Lys) LK₇β at interfaces. We now provide an explanation for the achiral and chiral VSFG optical responses of (L-), (D-), and (L-Leu)/(D-Lys) LK₇β variants, and relate chiral inversions of the peptide backbone with protein secondary structure and sidechain dynamics.

Identical Achiral VSFG Spectra of (L-) and (D-) LK₇β.

The achiral VSFG spectra of (L-) and (D-) LK₇β show strong resonant responses in the C-H stretching region around 2875 cm⁻¹, 2913 cm⁻¹, and 2938 cm⁻¹ (Figures 1a and 2a). Our harmonic DFT frequency calculations of a model peptide, supplemented with anharmonic calculations of single amino acids, suggest assigning these peaks to leucine CH₃ symmetric stretch, Fermi resonance modes arising from CH₂ symmetric stretching (with possible contributions from lysine C_α-H stretching and CH₃ symmetric stretch of the N-terminal acetyl cap), and leucine symmetric CH₃ Fermi resonance, respectively, in agreement with previous studies.^{22, 33, 60, 73–75} The spectral response of (L-) and (D-) LK₇β in the C-H stretching region are identical, suggesting that a total inversion of the chirality of the amino acids does not affect the properties of the LK₇β at the interface. These results are confirmed by molecular dynamics simulations demonstrating that both the distribution of backbone dihedral angles (Figure 6) as well as the orientations of sidechains (Figures S7 and S8) are identical for (L-) and (D-) LK₇β peptide at the air-water interface.

A Silent CH₂ Fermi Resonance Vibration in VSFG of (L-Leu)/(D-Lys) LK₇β.

When the LK₇β peptide is formed by an alternation of (L-Leu) and (D-Lys) amino acid residues, the vibrational mode at 2913 cm⁻¹ in the achiral VSFG spectrum is silenced, although the vibrational modes at 2875 cm⁻¹ and 2938 cm⁻¹ still show (Figures 1b and 2b). This is especially intriguing given that linear, non-surface-selective spectroscopy such as ATR-FTIR (Figure S1) hardly distinguishes between the (L-), (D-), and (L-Leu)/(D-Lys) LK₇β variants in the C-H stretching region.

In general, three possible conditions can lead to a silent achiral VSFG signal (Figure 8). (i) The molecular system adopts an ordered orientation that is VSFG inactive due to the polarization combination used in the experiment and, consequently, the selection rules over χ⁽²⁾. For example, the dipole transition moment might be perpendicular to the applied electric field of the linearly-polarized IR beam. (ii) The molecular system adopts an ordered centrosymmetric orientation around the interface resulting in a (destructive) interference between vibrational modes and a null mean second-order polarizability,${\overline{P}}^{(2)}=0$. (iii) The molecular system adopts a disordered, isotropic orientation around the interface, resulting in ${\overline{P}}^{(2)}=0$. In order to understand why the lysine CH₂ 2913 cm⁻¹ peak is absent in the achiral VSFG spectrum of (L-Leu)/(D-Lys) LK₇β, it is necessary to determine which of the three conditions is responsible for the silent VSFG signal. This required a combination of experimental and computational research, including DFT calculations of VSFG spectra and MD simulations.

Figure 8.

Conditions for an inactive VSFG signal. For the purpose of example in the cases outlined, it is assumed that the permanent dipole moment points parallel to the bond.

Scenario (i): Reorientation.

Our Ramachandran analyses (Figure 6c and S6) show that (L-Leu)/(D-Lys) LK₇β does not form stable β-sheets, but dynamically samples different conformations. On the other hand, the analysis of the orientation of CH₂ groups of the lysine sidechain (Figure 7d) shows no apparent net orientation for this group. These results allow us to rule out scenario (i), namely that (L-Leu)/(D-Lys) LK₇β at the interface adopts an ordered conformation that is differently oriented and thus VSFG inactive.

Scenario (ii): Centrosymmetric Order.

A similar analysis of our simulations also let us exclude scenario (ii), a centrosymmetric ordering at the interface. The Ramachandran free energy plot of (L-Leu)/(D-Lys) LK₇β certainly demonstrates that the peptide backbone lacks secondary structure at the interface (Figure 6c). Moreover, the analysis of the orientation of lysine sidechains does not show an alternate orientation of CH₂ groups that might lead to a destructive interference of vibrational modes. However, achiral VSFG reports not only on the absence or presence of protein secondary structure, but is convoluted with the vibrational modes of local chemical groups. This means that the free energy landscape of the peptide backbone structure is not enough to rule out all kinds of conformational order at the interface that could contribute (or silence) the VSFG response. Therefore, when analyzing achiral VSFG data it should be necessary to understand local bond tilt angle distributions as we have done in this work (Figure 7).

Scenario (iii): Isotropic Dynamics.

Our bond tilt angle analyses show that the introduction of chiral inversions at the lysine positions result in lysine C_α-H and C_γH₂ being isotropic (Figure 7d), while the orientation of leucine sidechains is not much affected by chiral inversion (Figure 7c). From this, we conclude that it is the isotropic nature of the coupled lysine C_α-H and CH₂ groups in (L-Leu)/(D-Lys) LK₇β that silences the response at 2913 cm⁻¹, compared to the achiral VSFG spectra of (L-) LK₇β and (D-) LK₇β at interfaces.

Sidechain Preferences Dominate Protein Secondary Structure at Interfaces.

The structural free energy landscape of the peptide backbone of (L-Leu)/(D-Lys) LK₇β shows multiple minima (Figure 6c) that are dynamically sampled by the peptide (Figure S6). This behavior might simply be attributed to the lack of inter-strand hydrogen bonding. However, our simulations show that even a single strand of (L-) LK₇β adopts a conformation that resembles an extended β-strand structure (Figure 6d). Remarkably, the isolated strand of (L-) LK₇β adopts a β-strand-like conformation even in the total absence of stabilizing inter-strand hydrogen bonds.

For all the LK₇β variants studied here, tilt angle analyses show that the leucine sidechains point towards the air and the lysine sidechains point towards the water phase. This strong preference persists strongly even for the mixed chirality (L-Leu)/(D-Lys) system (Figure 7c and 7d). In line with previous studies,⁸⁶ our results demonstrate that peptides adapt their architectures in line with the preferences of hydrophobic and hydrophilic sidechains towards the most stable phase even if these preferences incur large energetic costs by abolishing inter-strand hydrogen bonds. Our results go further, though: they also show that sidechain dynamics at the interface can be independent of chiral inversions along the peptide backbone.

Thus, our results suggest that interfacial (L-) and (D-) β-strand formation (and possibly β-sheet formation) can be driven from the alternating hydrophobicity and hydrophilicity of the peptide sidechains, more so than the disposition of the peptide backbone to form inter-strand hydrogen bonds. It is notable, however, that even this extremely strong alternating hydrophobic/hydrophilic preference can fail to form secondary structure if there is (L-)/(D-) heterochirality. Note that due to the chiral inversion in the lysine residues, the formation of an antiparallel β-sheet secondary structure would require the positioning of both leucine and lysine sidechains on the same side of the sheet. This arrangement of sidechains not only imposes steric hindrances on the β-sheet, due to the presence of bulky neighbor groups, but also prevents the hydrophobic or hydrophilic groups to be directed towards the more stable phase, completely destabilizing the β-sheet secondary structure at the interface. Because of the ability of VSFG to probe both local chemical functionalities and macroscopic chirality, these results can guide researcher expectations about the VSFG optical responses of proteins at interfaces.

Experimental Support that Chiral VSFG Response is Due to Macroscopic Chirality of β-sheets.

The achiral and chiral VSFG results in Figures 3a,b show that acid denaturation of (L-) LK₇β and consequent loss of secondary structure does not silence the CH₂ FR mode at 2913 cm⁻¹. In fact, the achiral C-H stretch spectrum of (L-) LK₇β is almost indistinguishable at pH = 7 or pH = 2 (Figure 3a). By contrast, the chiral amide I/amide II signal, which is indicative of anti-parallel β-sheets in (L-) and (D-) LK₇β, is very strong at pH = 7 and almost completely abolished at pH = 2 (Figure 3b).

Intriguingly, our MD simulations of a single strand of (L-) LK₇β (Figure 6d) offer insight about why the VSFG optical activities of the achiral C-H modes are insensitive to acid denaturation. The orientations of the C-H stretching modes are virtually unchanged when (L-) LK₇β exists as an isolated β-strand (Figures S9 and S10) versus in anti-parallel β-sheets (Figures 7a and 7b). Based on this result and our vibrational mode assignments for the C-H stretching region, one could have expected that the achiral C-H spectrum at denatured pH = 2 versus native pH = 7 would be relatively indistinguishable.

In contrast, our research group previously showed that (L-) LK₇β at the air-water interface shows no chiral C-H response at pH = 2.⁶² However, our MD simulations (Figure 6d) show that even one (L-) LK₇β will form a β-strand due to the preference of its hydrophilic and hydrophobic sidechains to structure at the interface. Therefore, our studies lend critical support for the theory of Simpson^41–42 that suggests the chiral VSFG optical response of LK₇β is due to macroscopic chirality of anti-parallel β-sheets.

Conclusions

We have applied VSFG spectroscopy in combination with DFT calculations and MD simulations to investigate the effects of chiral inversions of amino acids along the full length of a model anti-parallel β-sheet peptide at interfaces. We learned how chiral inversions perturb local structure of the peptide at the interface, and how such perturbations consequently disrupt the peptide secondary structure. The combination of VSFG experiments, DFT calculations of the VSFG optical response, and MD analyses of the peptide at the air-water interface show that, regardless of amino acid chirality, the peptide sidechains exhibit strong preferences for either the hydrophobic or hydrophilic phases. The peptide backbone adopts a conformation that accommodates these preferences within steric constraints. In the case of alternating (L-)/(D-) chirality in LK₇β, for instance, this results in complete loss of the anti-parallel β-sheet secondary structure. Future studies that systematically address the effects of differently patterned (non-alternating) chiral inversions on the protein backbone, or extend these investigations to the study of protein-protein interfaces, will help to define the rules that link protein primary structure to secondary and tertiary structures. Such work should be particularly relevant for de novo design of unnatural protein structures with novel functions. Finally, controllable perturbations to protein molecular structure (e.g., chiral inversions, chemical and thermal denaturation, sidechain modification, etc.) can be employed as critical tests of chiral VSFG theory.