TOAC spin-labeled peptides tailored for DNP-NMR studies in lipid membrane environments

Abstract

The benefit of combining in-cell solid-state dynamic nuclear polarization (DNP) NMR and cryogenic temperatures is providing sufficient signal/noise and preservation of bacterial integrity via cryoprotection to enable in situ biophysical studies of antimicrobial peptides. The radical source required for DNP was delivered into cells by adding a nitroxide-tagged peptide based on the antimicrobial peptide maculatin 1.1 (Mac1). In this study, the structure, localization, and signal enhancement properties of a single (T-MacW) and double (T-T-MacW) TOAC (2,2,6,6-tetramethylpiperidine-N-oxyl-4-amino-4-carboxylic acid) spin-labeled Mac1 analogs were determined within micelles or lipid vesicles. The solution NMR and circular dichroism results showed that the spin-labeled peptides adopted helical structures in contact with micelles. The peptides behaved as an isolated radical source in the presence of multilamellar vesicles, and the electron paramagnetic resonance (EPR) electron-electron distance for the doubly spin-labeled peptide was ∼1 nm. The strongest paramagnetic relaxation enhancement (PRE) was observed for the lipid NMR signals near the glycerol-carbonyl backbone and was stronger for the doubly spin-labeled peptide. Molecular dynamics simulation of the T-T-MacW radical source in phospholipid bilayers supported the EPR and PRE observations while providing further structural insights. Overall, the T-T-MacW peptide achieved better ¹³C and ¹⁵N signal NMR enhancements and ¹H spin-lattice T₁ relaxation than T-MacW.

Significance

Antimicrobial peptides (AMPs) are potential antibiotic alternatives to conventional treatments because of their multiple attack mechanisms toward bacteria. In an attempt to further understand the mode of action of AMPs in intact bacteria, we designed mono- and biradical spin-labeled peptides, TOAC-MacW (T-MacW) and TOAC-TOAC-MacW (T-T-MacW), to allow implementation of in-cell dynamic nuclear polarization solid-state NMR experiments. Here, we characterized the structure, localization, and signal enhancement properties of spin-labeled AMPs within lipid vesicles and demonstrated the potential of the spin-labeled peptides for determination of the structure-function relationship of AMPs in cells.

Introduction

Structural study of biomolecules in their native environments is a difficult but a fundamental requirement for understanding their molecular interactions in physiological conditions (1). In particular, studies of membrane proteins and membrane-active peptides benefit from appropriate, yet extremely complex, cell membrane environments, which in turn impose practical complications such as sample lifetime. Solid-state NMR has been at the forefront of structural studies of membrane proteins and membrane-active peptides in lipid environments (2,3) despite severe limitation in sensitivity. The emergence of dynamic nuclear polarization (DNP) has provided a tremendous boost in signal that now allows investigations of intact cellular systems, including mammalian cells (4, 5, 6). DNP-NMR also benefits from the use of cryogenic temperatures, which, combined with the use of cryoprotectants, considerably extends the lifetime of cells (7,8), although at the cost of retaining the physiological fluidity of the lipid bilayer. A recent in-cell DNP solid-state NMR application demonstrates the feasibility of studying antimicrobial peptides (AMPs) in intact Escherichia coli bacteria (9). As bacteria develop increasing resistance to traditional antibiotics, AMPs are seen as attractive alternatives to conventional treatments (10, 11, 12, 13), and thus, because peptide structure is dependent on environment, in situ studies will play a pivotal role in better understanding their mode of action.

DNP-NMR relies upon excitation of unpaired electrons by microwave irradiation and subsequent dipolar transfer of the spin polarization to surrounding nuclear spins, thereby providing a potential signal enhancement for protons up to a factor of 658 (7,14, 15, 16). The presence of a radical source as polarizing agents (PAs) is essential, and pioneering work has shown that radical properties significantly modulate the experimental signal enhancement. The spin polarization transfer in DNP occurs through a variety of mechanisms, and the most common are via the solid effect (SE) and the cross effect (CE) (17). Briefly, the SE occurs when a transition involves a single electron and a nuclear spin, whereas the CE, the most efficient mechanism at moderate magnetic field, originates from two coupled electrons whose difference in Larmor frequency matches the nuclear spin frequency (18,19). The latter complex mechanism of electron-to-nuclear spin magnetization transfer occurs in three so-called rotor events that are sensitive to strong electron dipolar coupling (20,21), moderate exchange interactions (22), and deleterious nuclear depolarization under magic angle spinning (MAS) (23,24). The molecular scaffold of PAs has been extensively studied and their sensitivity to local environments demonstrated, in particular their unfavorable depolarization and paramagnetic quenching, which can result in severe decrease of signal enhancement (19,25,26). Nitroxides are commonly employed because of their stability in biological systems. The five-membered piperidine ring tends to be more stable than six membered rings typically used in DNP applications (7). The best enhancements in aqueous environments are often reported for hydrophilic biradicals displaying a relatively rigid linker between two nitroxide moieties, e.g., AMUPol or TOTAPol (27) for moderate magnetic fields and TinyPol (28) or mixed radicals such as TEMtriPols (29) for high fields. In the case of binitroxides, enhancements are improved using a rigid linker that constrains the g tensors in a perpendicular orientation (30). Indeed, the relative orientation between the two tensors, denoted by the Euler angle Ω, controls the polarization and depolarization via the CE (31,32), which can be assessed by monitoring the MAS-DNP field profile (15,33). Furthermore, a homogenous radical distribution in the frozen matrix is also critical for obtaining higher signal enhancement, although targeted DNP is emerging as a valuable tool by localizing the PA in a biomolecular compartment, especially for in-cell studies (34). Hydrophilic PAs have been mixed in the so-called DNP “juice” composed of glycerol, D₂O, and H₂O, which form a glassy matrix that protects the radical from aggregating (35) but prevents further distribution inside cells and occupies a large fraction of the sample volume. Thus, in situations in which the PAs have an affinity for the molecular target, matrix-free approaches have been preferred (36, 37, 38). An additional challenge for applying DNP strategies to cells and cell-derived samples is the fast reduction of radicals by native antioxidants in cellular environments (25).

An in situ radical source has been shown to generate decent signal enhancements while avoiding the pitfalls mentioned above. Nitroxides can be introduced into target proteins or peptides by site-directed spin labeling, usually taking advantage of cross-linking capabilities between the nitroxide S-(1-oxyl-2,2,5,5-tetramethyl-2,5-dihydro-1H-pyrrol-3-yl)methyl methanesulfonothioate and cysteine residues (39). Another possibility for peptides is the readily available 2,2,6,6-tetramethylpiperidine-N-oxyl-4-amino-4-carboxylic acid (TOAC), which combines an amino acid backbone and a nitroxide side chain (25,40) and can be conveniently introduced in solid-phase peptide synthesis. Site-directed spin labeling using TOAC residues in combination with DNP solid state NMR spectroscopy thus offers a unique means to design spin-labeled peptides that will locate specifically in cell membranes. Previous studies have introduced spin labels into proteins (25,41,42), peptides (43, 44, 45), and lipids (46), which demonstrates the possibility of localizing enhancement by DNP.

Here, we have determined the structure, locality, and NMR signal enhancement of a single (T-MacW) (Fig. 1) and doubly (T-T-MacW) spin-labeled peptide in lipid membranes using a combination of circular dichroism (CD), NMR, and electron paramagnetic resonance (EPR) spectroscopy. Phosphatidylcholine bilayers were used to avoid complications because of the heterogeneity of E. coli membranes, thereby providing higher resolution of localization of the spin label within the lipid bilayer while maintaining typical cell membrane properties. The spin-labeled peptides are based on the maculatin 1.1. (Mac1) peptide sequence, an AMP isolated from the skin glands of the Australian tree frog Litoria genimaculata (12,13) that has demonstrated membrane-anchoring properties in lipid and bacterial membranes (47,48).

Figure 1.

Structure of the spin-labeled peptides T-MacW and T-T-MacW used in this study and expected location of the radicals within the lipid bilayer for targeted intramembrane DNP enhancements. To see the figure in color, go online.

Materials and methods

Materials

Perdeuterated-acyl chain 1,2-dimyristoyl-sn-glycero-3-phosphocholine (d₅₄-DMPC, molecular weight (M_W) 732.27) was purchased from Avanti Polar Lipids (Alabaster, AL). Sodium dodecyl-d₂₅ sulfate (d₂₅-SDS) was purchased from Sigma-Aldrich (Sydney, Australia). ¹⁵N-labeled L-alanine-N-FMOC (9-Fluorenylmethoxycarbonyl) and ¹³C₁ L-valine-N-FMOC were purchased from Cambridge Isotope Laboratories (Andover, MA). 1,2,3-Propanetriol-d₈, deuterated glycerol (glycerol-d₈), and D₂O were purchased from Sigma-Aldrich. FMOC-2,2,6,6-tetramethylpiperidine-N-oxyl-4-amino-4-carboxylic acid (Fmoc-TOAC-OH) was purchased from Chem-Impex (Wood Dale, IL). MacW (GLWGVLAKVAAHVVPAIAEHF-NH₂; M_W 2184.40), T-MacW (TOAC-MacW; M_W 2382.40), T-T-MacW (TOAC-TOAC-MacW; M_W 2580.52; see Fig. 1), and the [¹³C₁-V14, ¹⁵N_H-A18]-labeled Mac1 (GLFGVLAKVAAHVVPAIAEHF-NH₂; M_W 2147.25) were synthesized manually by solid-phase peptide synthesis and purified by HPLC (high-performance liquid chromatography) to a purity >95% using HCl instead of trifluoroacetic acid salt (49).

Solution NMR experiments

Solution NMR samples were prepared by dissolving 3 mM MacW in phosphate buffer (50 mM NaCl, 20 mM phosphate buffer (pH 4.74), 0.5 mM NaN₃, 0.03 mM DSS (sodium trimethylsilylpropanesulfonate), 10% v/v D₂O) containing 300 mM d₂₅-SDS micelles or 1 mM radical inactivated T-MacW in phosphate buffer containing 150 mM d₂₅-SDS micelles.

The MacW in d₂₅-SDS micelles sample was investigated at 310 K on an 800 MHz Bruker Advance II spectrometer (Bruker Biospin, Rheinstetten, Germany). ¹H homonuclear TOCSY (mixing time τ_mix = 80 ms) and NOESY (τ_mix = 200 ms) were acquired with 1024 points and 2048 points in the F1 dimension, respectively, and 8192 points in the F2 dimension. 16 transients were accumulated with a 1.5 s recycle delay, respectively. The data were multiplied with a squared sine bell function shifted by 90°. The ¹H spectral window was set to 9615 Hz. ¹³C-¹H HSQC experiments were performed with 256 points in the F1 dimension and 2048 points in the F2 dimensions. 64 transients were accumulated with a 2 s recycle delay. The ¹³C spectral window was set to 33,200 Hz. Nonuniform sampling ¹⁵N-¹H HSQC experiments were performed with 25% of 128 points in the F1 dimension and 4096 points in the F2 dimension. 2048 Transients were accumulated with a 1.5 s recycle delay. The ¹⁵N spectral window was set to 3240 Hz.

The radical inactivated T-MacW in d₂₅-SDS micelles sample was investigated at 310 K on an 800 MHz Bruker Advance II spectrometer (Bruker Biospin). ¹H homonuclear TOCSY (mixing time τ_mix = 80 ms) and NOESY (τ_mix = 200 ms) were acquired with 512 points and 1024 points in the F1 dimension, respectively, and 4096 points in the F2 dimension. 32 and 24 transients were accumulated with a 1.5 s recycle delay. The data were multiplied with a squared sine bell function shifted by 90°. The ¹H spectral window was set to 9615 Hz.

¹H chemical shifts were referenced to DSS at 0 ppm. Data were processed in TopSpin 4.0.5 (Bruker) and analyzed using CCPNmr Analysis 2.4.2 (50). Full backbone and side chains assignments were made using ¹H homonuclear TOCSY, NOESY, ¹⁵N-HSQC, and ¹³C-HSQC spectra.

Structure calculation

The NOESY crosspeak assignments were subsequently used to generate distance restraints for structure determination. The NOE distance restraints were supplemented with dihedral angle restraints derived from DANGLE prediction (51). Structures were calculated using a standard CNS 1.1-based protocol implemented in the ARIA 2.2 interface (52). The final 10 lowest energy structures were refined in a water shell and assessed with MolProbity (53).

CD

CD measurements were acquired on a Jasco J-815 spectropolarimeter (Jasco, Tokyo, Japan). The sample was scanned from 190 to 260 nm in a triplicate with 1 nm step size, 1 nm bandwidth, 1 s time per point, and 1 mm quartz cell (Starna, Hainault, UK). The signal was recorded as millidegree and zeroed at 260 nm and normalized to give units of mean-residue ellipticity (MRE) according to the following equation:

$$\text{MRE}=\frac{\theta }{C\times l\times N},$$ (1)

where θ is the recorded ellipticity in millidegrees, C is the peptide concentration in decimole per liter, l is the cell pathlength in centimeters, and N is the number of residues per peptide. The secondary structure fraction was processed using CDPro Software Package with SELCON3 Algorithm and SMP56 basis set (54,55).

Solid-state NMR sample preparation

The spin-labeled peptides were co-solubilized with labeled Mac1 in water first and then added to DMPC lipid to reach a lipid/peptide ratio of ∼65:1. The suspension was frozen and freeze dried overnight and resuspended into D₂O/H₂O (3:1, v/v) or glycerol-d₈/D₂O/H₂O (6:3:1, v/v/v) for NMR measurements performed at 303 or 108 K, respectively. Four freeze-thaw cycles were performed before packing the multilamellar liposomes (MLVs) into a 4 mm zirconia or 3.2 mm sapphire rotor for measurements performed at 303 or 108 K, respectively.

Solid-state NMR experiments

NMR experiments were conducted on a 9.4 T wide-bore Bruker Avance-III HD NMR spectrometer (Wissembourg, France) equipped with a 4 mm triple-resonance MAS probe. The temperature of all experiments was maintained at 303 K, and the spinning frequency was set to 8 kHz.

¹H MAS experiments were performed with a single 68 kHz excitation pulse, and 20,480 complex points and 40 scans were collected with 4 s recycle delay. The free induction decays were zero filled to 65,000 points without line broadening. T₁ experiments were performed using an inversion-recovery pulse sequence (56). The intensities were plotted against each time delay and were fitted using TopSpin 4.0.6 (Bruker) to a single exponential component.

¹³C cross-polarization MAS (CPMAS) experiments were performed with 68 kHz proton excitation pulse followed by 4 ms Hartmann-Hahn contact with a 30% linear ramped amplitude (RAMP) with 4 s recycle delay; 4096 scans and 4096 complex points were acquired under 70 kHz SPINAL64 decoupling. The FIDs were zero filled to 16,000 points, and 10 Hz line broadening was applied. The spectra were indirectly referenced using adamantane (29.46 ppm). The direct polarization experiments were performed using a single 50 kHz excitation pulse with 5 s recycle delay; 4096 scans and 4096 complex points were acquired under 70 kHz SPINAL64 decoupling. T₁ experiments were performed using an inversion-recovery pulse sequence. The intensities were plotted against each time delay and fitted using TopSpin 4.0.6 (Bruker) to a single exponential component.

³¹P MAS and static experiments were performed with a single 60 kHz excitation pulse, 55 kHz SPINAL decoupling, and 5 s recycle delay; 4096 scans and 5120 complex points were acquired. The FIDs were zero filled to 16,000 points.

DNP-enhanced solid-state NMR experiments

DNP MAS NMR experiments were conducted on a 9.4 T wide-bore Bruker Avance-III HD DNP-NMR Spectrometer equipped with a 3.2 mm low-temperature triple-resonance MAS probe and a 263.33 GHz gyrotron source. The temperature of all experiments was maintained at 108 K, the spinning frequency was set to 8 kHz, and the spectrometer frequency was initially set so that microwave irradiation of the gyrotron occurred at the maximal DNP enhancement of the binitroxide AMUPol.

¹³C CPMAS experiments were performed with 102 kHz proton excitation pulse followed by 1 ms Hartmann-Hahn contact with a 10% RAMP CP sequence; 3.5 s recycle delay and 2048 complex points were acquired under 106 kHz SPINAL64 decoupling. 64 scans were collected and the FIDs zero filled to 4000 points, and 10 Hz line broadening was applied. The spectra were referenced using the internal silicon plug ¹³C signal set at 0 ppm.

¹⁵N CPMAS experiments were performed with 100 kHz proton excitation pulse followed by 2 ms Hartmann-Hahn contact with a 30% RAMP and 3.5 s recycle delay; 128 scans and 1024 complex points were acquired under 104 kHz SPINAL decoupling. FIDs were zero filled to 4000 points, and 70 Hz line broadening was applied.

³¹P CPMAS experiments were performed with 105 kHz proton excitation pulse followed by 1 ms Hartmann-Hahn contact with a 10% RAMP and 3.5 s recycle delay; 32 scans and 4096 complex points were acquired under 110 kHz SPINAL decoupling. FIDs were zero filled to 4000 points, and 30 Hz line broadening was applied.

The DNP field profile was performed by recording on-resonance ³¹P CPMAS spectra while sweeping the magnetic field at fixed 125 mA microwave power. 16 scans were averaged, and the enhancement ε_on/ε_off was reported versus the on-resonance frequency. The MAS frequency dependance was determined by recording the ¹H ε_on/ε_off while adjusting the gas flow rates so that the temperature would vary less than ∼5 K.

EPR spectroscopy

For continuous wave (CW) EPR experiments, the single and double TOAC-labeled peptides were resuspended in milli-Q water to a spin concentration of ∼400 μM. Peptides were also added to DMPC by incubating at 35°C, followed by three to four freeze-thaw cycles until clearly homogenized. Glycerol was added to each peptide or peptide/DMPC sample (60:1 lipid to peptide molar ratio) in a ratio of 4:6 (water/glycerol). 100 μL samples (∼100 μM TOAC) were transferred into 4 mm outer diameter quartz tubes and flash frozen in liquid nitrogen before being inserted into an ELEXSYS E500 (Bruker) spectrometer. CW spectra were collected at 165 K with a 9.4 GHz (x-band) resonator using a microwave power of 0.05 mW, modulation amplitude of 2 G, and scan range of 160 G.

Molecular dynamics simulation

The starting conformation of MacW was generated from NMR data of the peptide in SDS micelles. The CHARMM-GUI membrane builder (57) was then used to prepare the DMPC bilayer system with 100 DMPC molecules per leaflet within a rectangular box containing KCl to enforce neutrality and a 12.5 Å layer of water. The force field for the TOAC residue was adapted from (58), and the ff14SB protein force field was used (59). All histidine residues were singly protonated to model the ionization state expected at pH 7.4, and the peptide C-terminus was amidated to match the experimental conditions. The simulations were performed using the AMBER force fields, and the minimization, equilibration, and production runs were performed with the AMBER CUDA package (60) on a desktop machine fitting a GPU GeForce GTX 1080 titanium and a CPU with 12 cores.

Each system was first minimized for 1000 steps using the steepest-descent method, followed by 1000 steps of the conjugate gradient method with a 12 Å nonbonded interaction cutoff. The peptides and lipids were restrained with a 10 and 2.5 kcal ⋅ mol⁻¹ potential, respectively. Then, 1.6 ns equilibration molecular dynamics (MD) simulations were run at 310 K, using decreasing positional restraints to maintain the peptide backbone and the lipid atom positions. All covalent bonds involving hydrogen atoms were constrained using the SHAKE algorithm (61), and the rigid internal geometry for TIP3P water molecules was constrained with the SETTLE algorithm (62). The system temperature was maintained at 310 K using a Langevin thermostat (63) with a 3 ps⁻¹ collision frequency. The system pressure was controlled at 1 bar using a semi-isotropic Monte-Carlo barostat with an xy surface tension. Each restart was performed with a random seed. 100 production runs of 5 ns each were performed with a 5 kcal ⋅ mol⁻¹ distance restraint on the two nitroxide oxygens set at 1.00 ± 0.25 nm.

The MD trajectories were visualized and analyzed using VMD (64) with custom scripts and the CPPTRAJ (65) software, and fitting procedures and plots were created in Gnuplot.

Results and discussion

F3W mutation and N-terminal TOAC addition did not perturb the peptide helical structure

The CD spectra of Mac1, MacW, T-MacW, and T-T-MacW peptides in buffer were typical of random coil structures with a single minimum at 198 nm. The presence of SDS micelles induced a transition from random coil to helical structures, as indicated by two minima in the vicinity of 222 and 209 nm and a maximum at 195 nm (Fig. 2 A) with similar amplitudes of MRE for all peptides. Spectral deconvolution based on a protein structure database (see Experimental section: solid-state NMR experiments) confirmed that there is no significant structural difference between the peptides, with helical content of 55% (Mac1), 57% (MacW), 55% (T-MacW), and 56% (T-T-MacW). Because CD is not able to provide atomic details on the peptide structures, solution NMR was used to determine the high-resolution structures of the peptides in SDS micelles (55,66).

Figure 2.

(A) CD spectra of Mac1 (black circles), MacW (red squares), T-MacW (orange stars), and T-T-MacW (blue triangles) in phosphate buffer (open symbols) and SDS micelles (solid symbols). (B) The 10 lowest energy NMR structures of MacW in SDS micelles. Chemical shift perturbation of T-MacW (C) from random coil and (D) from MacW in SDS micelles is shown. All experiments were performed at pH 5.0 and 37°C. To see the figure in color, go online.

The MacW structure in SDS micelles was determined by using homonuclear ¹H-¹H (NOESY, TOCSY) and heteronuclear ¹³C-¹H HSQC and ¹⁵N-¹H HSQC NMR spectroscopy. The residue classes were first determined using the TOCSY peak patterns, and the sequential assignment walk was performed using the NOESY spectra (67) (Figs. S1–S3). Notably, amino protons of Gly1 residues were not observed because of rapid hydrogen-deuterium exchange (68,69). The presence of strong d_{NN(i, i+1)}, d_{αN(i, i+1)}, and d_{βN(i, i+1)} NOEs were complemented by weaker d_{NN(i, i+2)} and d_{αN(i, i+2)} NOEs, and a series of long-range NOEs—d_{αN(i, i+3)}, d_{αβ(i, i+3}), and d_{αN(i, i+4)}—support that the peptides were predominantly in a helical conformation (Fig. S4). Structural calculations of MacW were based on a total of 211 NOE distance restraints comprising 129 intraresidue, 82 inter-residue, and 61 sequential restraints. The ¹⁵N and ¹³C chemical shifts of MacW were obtained from the ¹⁵N-¹H HSQC and ¹³C-¹H HSQC, respectively, and used to restrict further the dihedral angles restraints for structure calculation. The quality and statistical analysis of the structures showed no dihedral violation and no distance violation within 0.4 Å. The Cα root mean-square deviations of the mean structures was 0.47 Å. The MolProbity statistics indicated only minor clashes with no Ramachandran outliers or Cβ deviations and no bad bonds or angles (Table S1). Ensembles of 10 MacW lowest energy structures refined in a water shell displayed a single helical stretch (Fig. 2 B). The peptide exhibited a bend in the helical pitch, likely induced by the Pro¹⁵ residue, as previously observed (70, 71, 72). The chemical shift perturbations can be used to probe the propensity for secondary structures, e.g., α-helices, β-strand, or extended conformations. A positive difference of ¹³Cα and negative difference of ¹³Cβ, Hα, and ¹⁵N chemical shifts from the corresponding random coil shifts is an indication of α-helical propensity for a given segment of residues (73). The chemical shifts perturbation from Leu2 to Phe21 for MacW confirmed a long helical stretch propensity (Fig. S5). The addition of TOAC (inactivated) onto MacW produced a similar chemical shift perturbation, which suggested that the spin-labeled peptide adopted a similar helical stretch (Figs. 2, C and D, and S6), which is in agreement with the CD results. The main difference was an upfield shift of Trp3 Hα, but overall, the addition of TOAC to the MacW N-terminus did not perturb the peptide structure and its interaction with the micelles.

T-MacW and T-T-MacW act via the CE mechanism in DMPC membranes

The coupling of free electrons is a critical feature modulating the efficiency of the DNP enhancements. Mono- and biradical spins interact via two different dipolar mechanisms with the nuclear spins: the SE and the CE, respectively (18,20,33). The CE is usually an order of magnitude greater than the SE (26,74,75) at the moderate magnetic field and spinning frequency used in this study. Although the proximity of two electrons is constrained for the two TOAC residues linked by a peptide bond in the biradical T-T-MacW peptide, the monoradical T-MacW peptide may perhaps be isolated within the lipid bilayer, which would imply an SE mechanism. CW EPR experiments were performed to assess the proximity between the TOAC paramagnetic centers of the spin-labeled peptides in glycerol solution and in the presence of cryo-protected DMPC bilayers (Fig. 3). The EPR spectra of T-T-MacW with DMPC bilayers displayed significant broadening, whereas no broadening was observed for T-MacW, in comparison to the peptides in the water-glycerol matrix. As described by Mathies et al. (29), extracting the intra-electron-electron distance distribution accurately would require a multifrequency EPR analysis or structural modeling using, e.g., Density Functional Theory calculations (32).

Figure 3.

CW-EPR spectra at 165 K of (top) T-T-MacW and (bottom) T-MacW in (blue line) water-glycerol matrix or (black line) reconstituted in DMPC vesicles (1:60 peptide to lipid molar ratio). The EPR spectra of T-T-MacW are offset for clarity. a.u., arbitrary unit. To see the figure in color, go online.

Because the EPR data were not sufficient to determine whether T-MacW generates an SE or a CE mechanism in DMPC bilayers, a DNP field profile was performed (Fig. 4) (20). The ³¹P CPMAS DNP field profile spanned ∼60 mT without a plateau between a negative enhancement maximum at 9377 mT and a positive enhancement maximum at 9402 mT (Fig. 4 a). Furthermore, the ¹H enhancement modulation by the MAS rate (Fig. 4 b) displayed a sharp rise till 3.5 kHz, followed by a plateau with a maximum reached at 7 kHz. These features are typical of a PA enhancing the ³¹P NMR signal via the CE (26,32,74), which support a close proximity of the TOAC moieties via interpeptide self-assembly.

Figure 4.

(a) Normalized ³¹P CPMAS DNP-enhancement field profile and (b) MAS frequency dependance of ¹H enhancement of DMPC bilayers in the presence of T-MacW peptides (L/P 40:1). The experiments were performed at 9.4 T, T = 107 K, and 8 kHz MAS for (a) and at T = 106–110 K for (b). The displayed lines serve to guide the eye. To see the figure in color, go online.

TOAC-labeled peptides location in lipid membranes

The parent peptide Mac1 is known to form pores within lipid membranes (10), and thus, the small modification of the peptide N-terminus (F to W mutation, which also allows better estimate of peptide quantity) was not expected to significantly alter MacW peptide-lipid interactions. Furthermore, the addition of the TOAC residue at the N-terminus did not significantly change the secondary structure of MacW (Fig. 2), so the spin-labeled peptides are likely tightly bound to the lipid bilayers. This is supported by a recent study showing that T-MacW and T-T-MacW peptides were tightly bound to E. coli cells and the AMPs were not present in the supernatant after centrifugation (9).

To localize the spin-labeled peptides within a lipid bilayer, the paramagnetic relaxation enhancement (PRE) effects on ¹H, ³¹P, and ¹³C signals from DMPC MLVs incubated with the peptides were assessed (Figs. 5 and S5) at 303 K. The intensity and linewidth changes allowed qualitatively mapping of the distance-dependent PRE effects along the DMPC lipid molecules. The strongest PRE effects were observed near the glycerol-carbonyl lipid backbone for both T-MacW and T-T-MacW, the latter inducing greater signal loss in the lipids. The phospholipid headgroup and the terminal methylene were not strongly affected by the presence of the radicals, indicating that the TOAC residues were deeper below the phosphorous in the membrane. At the typical lipid/peptide molar ratio used for in-cell DNP-NMR studies, it is likely that many lipids are not in proximity to the spin-labeled peptides, even if the peptides are homogeneously distributed. Thus, a decent proportion of the signals should not sense any PRE effects. However, lipid and peptide diffusion could modulate the PRE effective area as observed by the almost complete loss of some signals, in particular for C2 and C3 atoms and to a lesser extent for G1-3 atoms (Fig. 5).

Figure 5.

(A) DMPC molecule, (B) ¹H MAS, and (C) ¹³C CPMAS spectra of DMPC MLVs (top black) and in the presence of T-MacW (middle red) and T-T-MacW (bottom blue). (D) PRE effect on DMPC NMR signals in the presence of T-MacW or T-T-MacW with a rough scale indicating loss of intensity from center of bilayer to headgroup region. To see the figure in color, go online.

There are several benefits for having a radical source inserted deep within lipid bilayers. As recently shown, aqueous radical sources can be degraded by cellular activity (76). Membrane insertion protected T-MacW and T-T-MacW peptides from significant degradation during incubation with E. coli cells (9). Membrane insertion of the spin-labeled peptides is also beneficial for enhancing signals within complex cell membrane architecture, where aqueous radicals and/or hyperpolarized protons may struggle to diffuse.

TOAC-TOAC configuration from molecular dynamics simulations

The 500 ns MD simulation of T-T-MacW in a bilayer made of DMPC lipids supported the location of the TOAC residues below the lipid phosphate groups and near the glycerol backbone (Figs. 6 and 7). Interestingly, a long-lived hydrogen bond (80% of the simulation time) between a phosphate group and the amino proton of TOAC1, hopping to the TOAC2 amino proton, was observed (see snapshot, Fig. 6). TOAC1 and TOAC2 were slightly exposed to water, especially in the first 100 ns of the simulation because of the initial configuration, but as the simulation progressed, TOAC2 was mainly shielded within the hydrophobic core of the bilayer, whereas TOAC1 maintained some water in proximity.

Figure 6.

Depth-cued snapshot of T-T-MacW (orange ribbon) in DMPC bilayers (phosphorus atoms represented by gold spheres) at 500 ns showing a hydrogen bond (purple line) between the phosphate group of a DMPC lipid and the amide protons of TOAC1 but hopping to TOAC2 across the simulation. Water (blue surface) contact within 5 Å of NO (depicted as blue and red sticks, respectively) slightly penetrates into the bilayer. To see the figure in color, go online.

Figure 7.

(A) Analysis of the trajectories obtained from 500 ns simulation of T-T-MacW in DMPC bilayers. The density versus depth of the phosphorous atoms (orange line), TOAC1 (black line), TOAC2 (blue line), and water molecules (purple line) from the center of the bilayer was fitted with a Gaussian function. (B) Water exposure within 5 Å of TOAC1 (orange circle) and TOAC2 (blue square). The line represents a linear fit of the number of contacts. (C) Nitroxide-to-nitroxide distance across the 500 ns simulation. The line represents a linear fit of the distance values. (D) Nitroxide dipole to nitroxide dipole angle across the 500 ns simulation. The line represents a linear fit of the angle values. To see the figure in color, go online. To see the figure in color, go online.

The strength of the DNP CE is modulated by the distance and angle between the two nitroxide NO^• dipoles. During the simulation, the distance fluctuated around 1.1 nm, showing some dynamics that are not favorable for reaching the highest NMR signal enhancement. Furthermore, the angle between the two dipoles (the vector along the NO^• bond) also fluctuated around 109°. However, the MD simulation was not able to be performed at 100 K, at which motions would be severely restricted. Therefore, the fluctuating angle and distance of the two nitroxides is likely suppressed during the DNP-NMR experiments.

A recent MD study of the bi-nitroxide radical TinyPol gave similar electron-electron distance and angle distribution (∼1.04 nm and 110°, respectively) (28). The dipolar and exchange coupling values calculated from these parameters were favorable to DNP via CE, albeit that the radical was designed for higher magnetic field and faster MAS speed. The local geometry of the biradical source is critical to the performance of CE (77), and at 9.4 T and moderate MAS speed, the TOAC-TOAC scaffold is within the optimal configuration for providing significant NMR signal enhancement.

TOAC-labeled peptide as radical source significantly enhanced DNP-NMR signals

The use of spin-labeled peptides was intended to deliver a localized radical source able to produce significant NMR signal enhancement of other membrane-located biomolecules via DNP. Although the spin-labeled MacW peptides are located within lipid bilayers of model membranes and E. coli bacteria, the signal enhancements observed for the lipids and for a [¹³C-V14, ¹⁵N-A18]-labeled AMP Mac1 were relatively high for the unoptimized radical moiety in such complex systems.

Overall, the NMR signals were enhanced almost twofold when using the T-T-MacW radical as source compared to T-MacW (Fig. 8; Table 1). The signal enhancements were also localized as observed for T-MacW and T-T-MacW. Furthermore, the signal enhancements observed within DMPC bilayers were similar to, albeit slightly less than, our recent in-cell studies of live E. coli using similar radical concentrations; the minor discrepancies may be due to the difference in spin-labeled peptide distribution and heterogeneity of the E. coli membranes. To determine the efficiency of the DNP-NMR signal enhancement for membrane-located peptides, a ¹⁵N-labeled antimicrobial peptide, Mac1, was added to the DMPC bilayers in the presence of the spin-labeled peptides (T-MacW or T-T-MacW). Interestingly, DNP-enhanced ¹⁵N CPMAS spectra showed two signals: the Ala18 ¹⁵N-labeled Mac1 peptide amide produced a broad peak at 120 ppm, whereas a weak peak at 48 ppm is likely due to the lipid choline headgroup. The ¹⁵N CPMAS signal enhancements were comparable to the ¹³C signal enhancements (Fig. 8), indicating that the hyperpolarized proton bath was coupled to Mac1 as well as the lipids.

Figure 8.

DNP-enhanced ¹³C CPMAS spectra of [¹³C-V14, ¹⁵N-A18] Mac1 in DMPC bilayers: (A) T-MacW and (B) T-T-MacW. DNP-enhanced ³¹P CPMAS spectra of DMPC bilayers with [¹³C-V14, ¹⁵N-A18] Mac1: (C) T-MacW and (D) T-T-MacW. Experiments recorded without (red) and with (black) microwave irradiation were performed under identical conditions at 8 kHz spinning speed and 108 K. (E) ¹⁵N CPMAS DNP-enhanced spectra of [¹³C-V14, ¹⁵N-A18] Mac1 in DMPC bilayers: T-MacW (red) and T-T-MacW (black) as radical sources. (All microwave off spectra are scaled ×4.) To see the figure in color, go online.

Table 1.

Enhancements and relaxation values obtained from CPMAS experiments at 108 K

	13C εDNP			31P εDNP	13C T1H (s)			31P T1H (s)
	Carbonyl	Glycerol	Aliphatic	Lipids	Carbonyl	Glycerol	Aliphatic	Lipids
TOAC − MacW + Mac1	14.4	13.8	13.7	12	6.9	8.0	7.6	7.7
TOAC − TOAC − MacW + Mac1	27.0	23.9	22.1	19.4	2.2	2.4	2.5	2.6
Mac1	1.0	1.0	1.0	1.0	16.2	17.8	16.4	17.8

ε^DNP = I_on/I_off with microwaves on/off; T₁-values were obtained by fitting the intensities of the signals obtained using inversion-recovery CP experiments to an exponential decay. Carbonyl ¹³C signal range: 178–163 ppm, glycerol ¹³C signal: 77–66 ppm, aliphatic ¹³C signal range: ∼38–18 ppm.

The radical source not only triggers signal enhancement but also provides a beneficial PRE effect, allowing faster CP signal acquisition because of reduced ¹H spin-lattice T₁ relaxation times and, in the CE situation, shorter DNP signal build-up time. Indeed, the required DNP setup (cryogenic temperatures and limited proton bath because of deuteration of the “juice”) is not favorable for fast recycle delay, lowering the gain in signal/noise per unit of time. Significantly, the presence of T-MacW and T-T-MacW decreased the ¹H T₁ values by approximately twofold and eightfold, respectively (Table 1). Note that only T₁ and signal enhancements are reported here to compare the two spin-labeled peptides and that the DNP build-up time has not been quantified, which also could modulate the maximal DNP signal enhancements (78).

Conclusions

Following recent work (9) demonstrating the benefit of using spin-labeled peptides for in-cell DNP-enhanced solid-state NMR, we report the interaction and localization of these spin-labeled peptides within a lipid bilayer. First, the substitution of Phe for Trp at position 3 of the well-characterized AMP Mac1 and the addition of TOAC to the peptide N-terminus induced no significant conformational change, as revealed by CD and high-resolution NMR. Both T-MacW and T-T-MacW performed well at 9.4 T by inducing localized ¹³C signal enhancement up to a factor of 14.4 and 27, respectively. Future work will focus on optimizing the design of the biradical scaffold grafted onto an amino acid backbone to improve the signal enhancement and manipulate localization of the radical source within cellular systems through use of specific peptide sequences. This would facilitate multidimensional experiments that could enable solid state NMR studies of self-assembly processes of membrane-active peptides and lead to targeted in-cell studies of membrane proteins.

Author contributions

M.-A.S. and F.S. designed the research. M.-A.S., S.Z., E.K., and L.J.B. performed the research and analyzed the data. M.-A.S., S.Z., E.K., F.S., and L.J.B. wrote the manuscript. M.-A.S., S.Z., E.K., J.R.H., L.J.B., and F.S. edited the manuscript.

Editor: Timothy Cross.