Breaking the Backbone: Central Arginine Residues Induce Membrane Exit and Helix Distortions Within a Dynamic Membrane Peptide

Abstract

Transmembrane domains of membrane proteins sometimes contain conserved charged or ionizable residues which may be essential for protein function and regulation. This work examines the molecular interactions of single Arg residues within a highly dynamic transmembrane peptide helix. To this end, we have modified the GW^4,20ALP23 (acetyl-GGAW⁴(AL)₇AW²⁰AGA-amide) model peptide framework to incorporate Arg residues near the center of the peptide. Peptide helix formation, orientation and dynamics were analyzed by means of solid-state NMR spectroscopy to monitor specific ²H- or ¹⁵N-labeled residues. GW^4,20ALP23 itself adopts a tilted orientation within lipid bilayer membranes. Nevertheless, the GW^4,20ALP23 helix exhibits moderate to high dynamic averaging of NMR observables, such as ²H quadrupolar splittings or ^l5N-¹H dipolar couplings, due to competition between the interfacial Trp residues on opposing helix faces. Here we examine how the helix dynamics are impacted by the introduction of a single Arg residue at position 12 or 14. Residue R14 restricts the helix to low dynamic averaging and a well-defined tilt that varies inversely with the lipid bilayer thickness. To compensate for the dominance of R14, the competing Trp residues cause partial unwinding of the helix at the C-terminal. By contrast, R¹²GW^4,20ALP23 exits the DOPC bilayer to an interfacial surface-bound location. Interestingly, multiple orientations are exhibited by a single residue, Ala-9. Quadrupolar splittings generated by ²H-labeled residues A3, A5, A7 and A9 do not fit to the α-helical quadrupolar wave plot defined by residues A11, A13, A15, A17, A19 and A21. The discontinuity at residue A9 implicates a helical swivel distortion and an apparent 3₁₀-helix involving the N-terminal residues preceding A11. These molecular features suggest that while arginine residues are prominent factors controlling transmembrane helix dynamics, the influence of interfacial tryptophan residues cannot be ignored.

Graphical Abstract

INTRODUCTION

Lipid-bilayer membranes define regions of high dielectric gradient. Indeed, by contrast with a lipid membrane surface, the nonpolar interior environment created by the acyl chains of a lipid bilayer is not readily hospitable for ionizable amino acid residues in membrane proteins. Nevertheless, the transmembrane domains of proteins sometimes contain noteworthy polar or charged residues which may be conserved and essential for protein function and regulation of cellular activity. For example, arginine residues play major roles in membrane voltage sensing domains of voltage-gated ion channels.^1-2 A “snorkeling” of positively charged arginine or lysine side chains to “reach” and interact with the lipid/membrane interface would help to keep the remaining nonpolar transmembrane domain within the hydrophobic membrane.^3-4 Direct evidence for snorkeling comes from NMR measurements of the ¹³C-³¹P distances for the Arg C_α and C_ζ carbons.^5-6 While notable snorkeling is observed for arginines in the voltage-sensing S4 helix of potassium channels,⁶ it is not necessarily observed for some antimicrobial peptides.⁵ Nevertheless, in this regard, mutations that introduce polar and charged residues within transmembrane domains may have serious consequences affecting protein structure, function, and stability.

Due to numerous experimental challenges with large membrane proteins, simplified model systems can be useful for understanding the physical chemistry of the lipid interactions of ionizable protein side chains. The membrane-spanning peptide GW^5,19ALP23 (acetyl-GGALW⁵(LA)₆LW¹⁹LAGA-amide)⁷, for example, has been useful for defining the titration properties of membrane-imbedded ionizable Arg, Lys, His and Glu residues.^8-10 GW^5,19ALP23 is advantageous because its robust helix adopts a well-defined orientation within lipid bilayers, wherein the helix tilt is dependent on membrane acyl chain length.¹¹ The chain length dependence is likely a result of helix adaptation to the lipid-peptide hydrophobic mismatch.^11-13 Within this framework, specific arginine substitutions introduce new and interesting interactions between the peptide helix and its lipid environment.¹⁴ Arg at position 14, for example, changes the tilt, forces a large helix rotation and prefers to “snorkel,” moving its side chain guanidium group into the bilayer lipid head-group region. Molecular dynamics simulations additionally predict that the helix movement is accompanied by local membrane thinning.¹⁴ By contrast, Arg placed at the dead-center of the GW^5,19ALP23 helix, at position 12, causes the helix to adopt multiple states. In DOPC membranes, both coarse-grained molecular simulations and solid-state NMR experiments predict the presence of three major states.¹⁴ Two states retain a transmembrane helix with the arginine side chain “snorkeling” either “up” or “down” toward the polar lipid head groups. The third state is entirely different, with the complete helix exiting the lipid bilayer to adopt an interfacial orientation, perpendicular to the membrane normal. Modest amounts of cholesterol, about 10 mol %, in the DOPC membrane force essentially the entire R¹²GW^5,19ALP23 population into this interfacial state.¹⁵ The multi-state features and membrane-exit property of R¹²GW^5,19ALP23 are likely dictated by a Trp “cage” surrounding the R12 guanidium group, which effectively is trapped between the two aromatic side chains and restricted from favorable interactions with the bilayer head groups. Moving the Trp residues outward, to positions 3 and 21, increases the size of this cage and effectively frees the Arg side chain (at both positions 12 and 14) to permit a stable transmembrane orientation.¹⁶

Single transmembrane model helices reveal significant biophysical features for larger membrane proteins and are especially important for understanding (a) single-span membrane proteins ^17-19, and (b) polar or charged functional groups that are in direct contact with the lipids.^5,8,14,20 Limitations arise when the inter-helix protein-protein structural interactions in multi-helix bundles^21-22 are not represented by the individual helices considered here. Nevertheless, the molecular features that underlie protein-lipid interactions, individual helix stability and dynamics are well represented.^23-24

Solid-state NMR methods are useful for examining transmembrane helices. For example, the pattern for the ²H quadrupolar splittings of labeled Ala side chains around a helix can reveal the helix orientation and dynamics in a bilayer membrane.^25-28 In parallel fashion, the ¹H-¹⁵N dipolar coupling and ¹⁵N chemical shifts from labeled peptide backbone groups, observed in separated local field experiments,^29-30 also reveal the helix orientation and dynamics. The separated local field methods, in particular, have been applied to membrane proteins,^31-32 including the Vpu domain from HIV-1³³, the chemokine receptor CXCR1³⁴, phospholamban,³⁵ cytochrome P450 reductase,³⁶ cytochrome b5³⁷ and cytochrome P450,³⁸ individually and in complex,^18,39 the influenza A M2 proton channel⁴⁰ and cell division regulatory protein CrgA,⁴¹ among others.

Recently, we developed a highly dynamic peptide framework by relocating the Trp residues of GW^5,19ALP23 outward by only one sequence position on each side. With the Trp residues in positions 4 and 20, the large indole side chains then reside on opposite faces (Figure 1) of the α-helix of acetyl-GGAW⁴(AL)₇AW²⁰AGA-amide.⁴² The GW^4,20ALP23 helix experiences moderate to high motional averaging of solid-state NMR observables such as ²H quadrupolar splittings, ¹H-¹⁵N dipolar couplings and ¹⁵N chemical shifts. The excess dynamic averaging, much more than observed for GW^5,19ALP23, is caused primarily by additional rotation about the helix axis. The observed azimuthal slippage is attributed tentatively to a competition between the two Trp indole rings for preferential locations at the bilayer interface. An added benefit of the GW^4,20ALP23 sequence (Table 1) is the availability of two additional Ala residues for deuterium labeling, such that more ²H are available for analysis of the α-helical structural perturbations. We take advantage of these features to examine the influence of single arginine substitutions in GW^4,20ALP23 (Table 1).

Figure 1.

GWALP-like Peptide Models. From left to right: GW^4,20ALP23, R¹²GW^4,20ALP23, R¹⁴G^4,20ALP23, GW^5,19ALP23. See Table 1 for the amino acid sequences. While not depicted here, the helix terminals tend to fray.

TABLE 1.

Sequences of Arginine Containing GW^4,20ALP23 and GW^5,19ALP Peptides

Name	Sequence	Reference
GW4,20ALP23	acetyl-GGAW4ALALALALALALALAW20AGA-amide	(42)
R12GW4,20ALP23	acetyl-GGAW4ALALALARALALALAW20AGA-amide	This work
R14GW4,20ALP23	acetyl-GGAW4ALALALALARALALAW20AGA-amide	This work
GW5,19ALP23	acetyl-GGALW5LALALALALALALW19LAGA-amide	(7)
R12GW5,19ALP23	acetyl-GGALW5LALALARALALALW19LAGA-amide	(14)
R14GW5,19ALP23	acetyl-GGALW5LALALALARALALW19LAGA-amide	(14)

In this paper, we present the unique outcomes that results from introducing a central Arg residue into the dynamic GW^4,20ALP23 framework at either position 12 or 14. Arginine at either position is situated on a different helix face from those of either Trp residue (Figures 1-2). Thus, neither R12 nor R14 in GW^4,20ALP23 is likely to fall within a Trp “cage.” Nevertheless, the results will indicate that the R12 and R14 substitutions confer quite different modulations of the helix properties, with R12 not only driving the helix to the surface of DOPC membranes but also distorting the structure of the surface-bound helix. The results are significant for considering the complexity of arginine-guided rearrangements in membrane proteins, for example in response to specific mutations,^43-44 transmembrane voltages^44-46 or defects in arginine transport.⁴⁷

Figure 2.

Helical wheel plots for GWALP-like peptides highlighting the Trp (W) locations with respect to residues 12 and 14. Top: GW^5,19ALP23; bottom: GW^4,20ALP23. Ala residues used for ²H-labeling are depicted in gray. Trp residues are shown in yellow. Positions Z12 and Z14 (blue) are either Leu or one of them is substituted with Arg.

METHODS

Materials

Fmoc-amino acids with protected side chains were purchased from Novabiochem (San Diego, CA). The tryptophan and arginine side chains were protected with t-butoxycarbonyl and 2,2,4,6,7-pentamethyldihydrobenzofuran-5-sulfonyl protecting groups, respectively. Commercial fmoc-L-alanine-¹⁵N and fmoc-l-leucine-¹⁵N were purchased from Cambridge Isotope Laboratories (Andover, MA). Commercial l-alanine-d₄ was also purchased from Cambridge Isotope Laboratories (Andover, MA) and was modified with an Fmoc group on a 100 mg scale as described.⁴⁸ Fmoc-l-alanine-d₄ was recrystallized from 20 mL ethyl acetate:hexane 80:20 and successful synthesis was confirmed using ¹H NMR. DLPC, DMPC, and DOPC lipids were purchased from Avanti Polar Lipids (Alabaster, Alabama). Other solvents and chemicals were the highest grade available.

Peptide Synthesis

Peptides R¹²GW^4,20ALP23 and R¹⁴GW^4,20ALP23 (see Table 1) were synthesized on a 0.1 mmol scale using solid-phase FastMoc® chemistry⁹ on a model 433A Applied Biosystems peptide synthesizer (Life Technologies, Foster City, CA). In most cases, two ala-d₄ residues were incorporated into each peptide in different isotopic abundances. In certain cases where spectral assignments remained ambiguous, a single ²H-labeled alanine residue was used instead. ¹⁵N-labeled peptides were also synthesized, containing two or three ¹⁵N-labeled Ala or Leu residues. The peptides were purified on a Zorbax 300SB-C3 column (9.4 × 250 mm, 5-μm particle size; Agilent Technologies, Santa Clara, CA) with a gradient of 86-90% methanol (with 0.1% trifluoroacetic acid) over 11 min (R¹²GW^4,20ALP23) or 13 min (R¹⁴GW^4,20ALP23). MALDI mass spectrometry and analytical HPLC were used to confirm peptide identity and peptide purity of at least 95% (see Figure S1). Minor impurities were not characterized and will not affect the NMR spectra.

Circular Dichroism Experiments

Circular dichroism (CD) spectroscopy samples were prepared to analyze peptide helicity by using a 1:60 peptide:lipid mixture (62.5 nmol peptide and 3.75 μmol lipid). To prepare lipid vesicles, the samples were sonicated at 22 °C. The samples were analyzed using a Jasco (Easton, MD) J-1500 CD/Fluorescence spectropolarimeter with a 1 mm cell path, 1.0 nm bandwidth, 0.1 mm slit, and a scan speed of 20 nm/min, with averaging of 10 scans. CD-based estimates of the overall helix content were obtained from the CONTIN-LL and K2D programs implemented on the DICHROWEB online server.^49-50

²H Solid-State NMR Experiments

Mechanically aligned solid-state NMR samples were prepared by adaptations of previous procedures.^9,25,51-52 A peptide:lipid mixture having a ratio of 1:60 (1.33 μmol /80 μmol) was dispersed over about 40 thin glass plates and hydrated to 45% hydration (w/w) using deuterium-depleted water from Cambridge Isotope Laboratories (Andover, MA). Bilayer alignment in the sealed samples was confirmed using ³¹P NMR at 121.5 MHz on a Bruker (Billerica, MA) Avance 300 spectrometer with broadband ¹H decoupling (4.2 kHz) for samples oriented at both β = 90° (bilayer normal perpendicular to the magnetic field) and β = 0° (see Figure S2). Solid-state ²H NMR experiments were performed with a quadrupole-echo pulse sequence, with full phase cycling,⁵³ at 50 °C using a Bruker Avance 300 spectrometer at both sample orientations. The pulse sequence included a pulse time of 3.0 μs, an echo delay of 105 μs, and a recycle delay of 120 ms. Each ²H NMR spectrum was acquired using 0.8 to 1.5 million scans. Based on measurements of two or more duplicate samples, an experimental uncertainty of ±1 kHz was estimated for the ²H quadrupolar splittings.

¹⁵N-¹H Dipolar Coupling/¹⁵N Shift Solid-State NMR Experiments

Static solid-state NMR samples were prepared as described previously⁹ using glass slides with dimensions measuring 5.7 × 10 mm NO. 000 purchased from Matsunami Glass (Bellingham, WA) and glass cells with dimensions measuring 5.4 × 7.4 × 18 mm purchased from New Era Enterprises (Vineland, NJ). Samples contained 1.33 μmol labeled peptide and 80 μmol lipid (1:60) and were hydrated to 45% (w/w) using deuterium depleted water.

Using a selective averaging “magic” polarization index (“SAMPI4”) method,⁵⁴ separated local field spectra were recorded on a Bruker Avance NEO 600 MHz NMR spectrometer with Larmor frequencies of 600.13 and 60.81 MHz for ¹H and ¹⁵N, respectively, using a low electrical field static 1H-X probe with a flat-coil configuration⁵⁵ with 1600 scans, 32 t1 increments, and a recycle delay of 4.0 s at 50 °C. The t1 evolution was preceded by Cross-Polarization with Mismatch-Optimized IS Transfer⁵⁶ with a contact time of 810 μs during which the radio-frequency (RF) spin-lock amplitude of 50 kHz was used in both ¹H and ¹⁵N RF channels. The SPINAL-64¹⁷ decoupling sequence with the ¹H RF amplitude of 62.5 kHz was applied for ¹H heteronuclear decoupling during the 15 ms acquisition time.⁵⁷ The ¹H carrier frequency of 9 ppm was used and is optimal for transmembrane helices in oriented bilayers. The ¹⁵N NMR data were processed and displayed using NMRPipe/NMRDraw⁵⁸ and Sparky⁵⁹ and adjusted using the theoretical dipolar coupling scaling factor for SAMPI4 evolution.⁵⁴ The ¹⁵N chemical shifts were externally referenced to ¹⁵N-labeled solid ammonium sulfate, set to 26.8 ppm, corresponding to the signal from liquid ammonia at 0 ppm.⁶⁰

Data Analysis

Analysis of the ²H NMR spectra was performed using the semi-static geometric analysis of labeled alanines (GALA) as described previously by van der Wel et al 2002.²⁵ The deuterium quadrupolar splittings of the alanine methyl groups (Δν_q) are dependent on the macroscopic sample orientation as seen in eq 1 below.

$$\Delta v_q=\mathrm{QCC}\times S_{ZZ}\times \left[\frac{1}{2}(3{\mathit{cos}}^2\theta -1)\right]\times \left[\frac{1}{2}(3{\mathit{cos}}^2\beta -1)\right]\times \langle \frac{1}{2}(3{\mathit{cos}}^2\gamma -1)\rangle$$ (1)

Known constants within this equation include the angle between the membrane normal and the applied magnetic field (β) = 90° or 0°, the quadrupolar coupling constant for an aliphatic C-D bond (QCC) = 168 kHz, and the tetrahedral bond angle of the CD₃ group (γ) = 109.5° which results in a 1/3 reduction of the coupling constant due to the fast rotation of the methyl group. This leaves the principal order parameter S_zz, which serves as an estimate for peptide motion, and θ, the angle between the magnetic field and the C_α-C_β bond of the alanine side chain, as variables. The θ angle, in turn, is dependent on the average peptide orientation as seen in eq 2 below.

$$\theta ={\epsilon }_{\Vert }[\mathit{cos}{\tau }_0-\mathit{sin}{\tau }_0\times \mathit{cos}({\rho }_0+{\epsilon }_{\perp }+\phi )\times \mathit{tan}{\epsilon }_{\Vert }]$$ (2)

The angles defining the local side chain orientation with respect to the C_α-C_β bond (ε_∥) and a plane perpendicular to the helix direction (ε_⊥) are fixed to 59.4° and −43.3° respectively, as determined previously.²⁵ The angle φ is the angle between a reference point (here C_α of Gly¹)²⁵ and C_α of the deuterium labeled residue in the peptide.

The GALA method, therefore, considers an α-helical geometry and three adjustable parameters: the apparent tilt τ₀ of the helix axis with respect to the bilayer normal, the helix azimuthal rotation ρ₀ and the order parameter S_zz (eq1 and eq 2).²⁵ The GALA calculation finds the lowest RMSD fit between the experimental Δν_q’s and those calculated for the peptide helix as it samples available orientations with respect to the bilayer normal and the reference point for φ.

Deviations from α-helical geometry result in either a uniformly high global RMSD or individual data points deviating from the GALA quadrupolar wave.¹⁰ For our analysis, we searched ranges of τ₀ (0 - 90°) and ρ₀(0 - 359°), each incremented by steps of 1°, and S_zz (0 - 1.0) by steps of 0.1.

To modify the GALA calculation to match the ²H NMR data to a 3₁₀ helix instead of an α-helix, the calculation for the angle φ was changed by modifying η, the helical wheel separation between the residue in question (n) and the previous residue (n – 1) in equation 3 below. For a canonical α-helix η = 100°, and for a canonical 3₁₀-helix η = 120°.

$$\phi =(n-1)\times \eta +360$$ (3)

A canonical 3₁₀-helix with η = 120° would result in every third Ala residue generating, predictably, the same ²H-methyl quadrupolar splitting, which was not observed in this study. The value of η = 120° also is unreasonable because of side chain repulsion between residue n and residue (n+3). Instead, average 3₁₀-helix torsion angles (−71°, −18° for ϕ, ψ) found in nature were used resulting in η = 112.5°.⁶¹ This value of η also results in ε_∥ of 61° and ε_⊥ of −38° required for equation 2 above. The angles ε_∥ and ε_⊥ were calculated using a combination of UCSF Chimera and Blender (open-source 3D graphics software).^62-63

The availability of eight core alanine residues in GW^4,20ALP23 is especially useful when studying peptides that exhibit increased dynamic motion.^42,64 The large data pool allows for a more demanding Gaussian analysis to be utilized over ²H quadrupolar splittings individually or combined with ¹⁵N chemical shifts and ¹⁵N/¹H dipolar couplings, as derived from model 6 of Strandberg et al, 2009.⁶⁵ This model of helix dynamics considers a Gaussian distribution of helix tilt τ and rotation ρ angles centered at the angles τ₀ and ρ₀ with widths as the oscillations about them, σ_τ (helix wobble) and σ_ρ (rotational slippage) respectively. The calculation performs a grid search over the variables τ₀, ρ₀ and their standard deviations σ_τ and σ_ρ, while fixing a principal order parameter S_zz to either 1.0 (no isotropic motion) or 0.88 as an estimate for the isotropic internal motion of a transmembrane peptide, or another value. Our analysis was performed by varying, in 1° increments, τ₀ from 0° to 90°, ρ₀ from 0° to 359°, σ_τ from 0° to 30° and σ_ρ from 0° to 200°. In cases of a limited availability of data points, a modified Gaussian calculation was used instead by restraining σ_τ to a small finite value and varying the remaining three parameters as above. The in-house program estimates the helical geometry based on a polyalanine α-helix model. Modifying the torsion angles within the model allows the data to be fit to a tighter 3₁₀-helix. The α-helix torsion angles for ϕ and ψ (−64° and −40°)⁶⁶ were modified to −71° and −18°, respectively, to represent a 3₁₀ helix, based on the mean torsion angles found within naturally occurring 3₁₀-helices⁶¹.

In cases of helix distortion, the peptide kink angle (κ) can be calculated using equation 4 below with the rotation (ρ) and tilt (τ) relative to either the C-terminal or N-terminal segment of the peptide helix.⁶⁷

$$\cos \kappa =\sin ({\tau }_N)\sin ({\tau }_C)\cos (\Delta \rho )+\cos ({\tau }_N)\cos ({\tau }_C)$$ (4)

RESULTS

The parent peptide helix of GW^4,20ALP23 exhibits extensive dynamic averaging in lipid-bilayer membranes.⁴² The present experiments reveal the influence of introducing a single charged Arg residue into this highly dynamic helical framework at position 14 or position 12 of the 23-residue sequence.

R¹⁴GW^4,20ALP23

Solid-state NMR techniques have proven useful for determining the extent of a transmembrane peptide’s helicity²⁴ and furthermore the helix average orientation and dynamics within its lipid environment. A key result is that the presence of R14 lowers dramatically the extent of motional averaging. The reduced motion is evident from the wide range of ²H quadrupolar splittings, from 1 kHz to 33 kHz, observed for the collection of Ala methyl side chains in R¹⁴GW^4,20ALP23 (Figure 3, Figure S3, Table 2), compared to the narrow range of 1 kHz to 16 kHz when R14 is absent in GW^4,20ALP23.⁴² For reference, a moderately dynamic peptide such as GW^5,19ALP23 produces core alanine ²H quadrupolar splittings that span a range of about 1 kHz to 27 kHz.¹¹ Therefore, the presence of R14 indeed decreases the high extent of motional averaging of the ²H NMR signals from the host peptide helix.

Figure 3.

²H NMR spectra for labeled alanines of R¹⁴G^4,20ALP23 in mechanically aligned DLPC, DMPC and DOPC bilayers. Samples consisted of 1.3 μmol peptide with 80 μmol lipid oriented at β = 90°, temperature 50°C. Between 0.8-1.5 million scans were acquired during 25-50 hr. The sequence numbers of the ²H-labeled alanines are from top to bottom (5^100%, 7^50%); (9, 11); (13, 15); (17, 19); (3, 21). The first alanine of each pair is 100% deuterated while the second is 50% deuterated.

TABLE 2.

Quadrupolar Splitting Magnitudes (∣Δν_q∣, in kHz) for Labeled Alanine CD₃ Groups in R¹⁴GW^4,20ALP23 and R¹²GW^4,20ALP23^a

Lipid	R14GW4,20ALP23 Alanine CD3 Position/ Δνq (kHz)
	3	5	7	9	11	13	15	17	19	21
DLPC	30.0	20.2	25.8	0.1	11.2	20.0	13.2	27.0	33.0	24.0
DMPC	29.4	14.2	23.2	2.6	8.8	19.7	12.5	28.0	28.0	24.0
DOPC	30.2	8.2	20.1	7.3	4.1	19.3	10.8	28.1	22.4	14.3
Lipid	R12GW4,20ALP23 Alanine CD3 Position/ Δνq (kHz)
	3	5	7	9(a)/(b)	11	13	15	17	19	21
DOPC	18.9	0.5	23.2	27.4/11.4	21.0	0.7	13.2	37.1	25.0	31.0

^aβ = 0° sample orientation. The measurement uncertainty is ±1 kHz.

^bResidue A9 gives two signals in DOPC, designated as 9a (27.4 kHz) and 9b (11.4 kHz).

The helix orientations can be compared using the semi-static GALA method to analyze the patterns of alanine methyl ²H quadrupolar splittings.²⁵ Indeed, the GALA quadrupolar wave-plots (Figure 4A) reveal a rather constant helix azimuthal rotation ρ₀ of ~224° (Table 3) for R¹⁴GW^4,20ALP23 in DLPC, DMPC and DOPC bilayer membranes. An essentially constant preferred value of ρ₀, independent of the bilayer thickness, is indicative and even diagnostic of low dynamic averaging.^11,28 Hand in hand with the low dynamic averaging, the tilt τ₀ of the R¹⁴GW^4,20ALP23 helix decreases systematically as the lipid bilayer thickness increases (τ₀ being 21°, 17°, 14° in DLPC, DMPC, DOPC; respectively). This behavior contrasts sharply with that of GW^4,20ALP23, which shows extensive variation for ρ₀ and no correlation for τ₀ with the lipid environment (Figure 4C). The arginine residue R14 has served to stabilize the membrane-incorporated helix in a preferred orientation.

Figure 4.

GALA quadrupolar wave plots for A) R¹⁴GW^4,20ALP23, B) R¹⁴GW^5,19ALP23 and C) GW^4,20ALP23 in DLPC (red), DMPC (blue) and DOPC (black) bilayers. Data points with white filling were omitted from the analysis. The helix orientations corresponding to the quadrupolar waves are listed in Table 3.

TABLE 3.

GALA and Gaussian Fits Using Ala-CD₃ ∣Δν_q∣ magnitudes of Arg Containing GWALP23 Family Peptides^a

		GALA fit results				Gaussian fit results
Lipid	Peptide	τ0	ρ0	Szz	RMSD	τ0	ρ0	σρ	στ	RMSD	Ref
DLPC	W4,20	6.0°	322°	0.72	0.70	16°	321°	85°	15°	0.49	(42)
	W5,19	20.7°	305°	0.71	0.66	23°	304°	33°	5°b	0.70	(42)
	R14W4,20	21.3°	223°	0.77	0.92	21°	223°	24°	14°	0.74	This work
	R14W5,19	30.0°	259°	0.83	1.58	30°	260°	< 10°	5°b	1.65	(14);This work
DMPC	W4,20	3.3°	349°	0.71	0.85	5°	347°	51°	20°	0.68	(42)
	w5,19	11.7°	311°	0.87	0.90	13°	308°	44°	5°b	1.10	(42)
	R14W4,20	16.7°	228°	0.85	0.75	17°	228°	12°	6°	0.72	Tliis work
	R14W5,19	25.8°	252°	0.81	1.59	26°	252°	28°	5°b	0.97	(14);This work
DOPC	W4,20	1.7°	133°	0.81	0.80	9°	129°	122°	5°	0.78	(42)
	w5,19	6.0°	323°	0.87	0.60	9°	321°	48°	5°b	0.70	(42)
	R14W4,20	13.3°	221°	0.90	0.81	14°	220°	9°	2°	0.87	Tliis work
	R14W5,19	16.1°	246°	0.94	1.29	16°	246°	< 10°	5°b	1.20	(14);This work

^aThe abbreviations refer to peptides based on the locations of selected aromatic residues, W^4,20 in GW^4,20ALP23 and W^5,19 in GW^5,19ALP23, with or without arginine R14, as noted. See also Table 1. The uncertainties in τ₀ and ρ₀ are about 2° and 5°, respectfully. The widths of the distributions around τ₀ and ρ₀ are given by στ and σρ.

^b A modified three variable gaussian treatment^{28, 93} was used to analyze the six core Ala-CD₃ data points constraining στ to 5°.

The quadrupolar wave plots for R¹⁴GW^4,20ALP23 are nicely similar to those of R¹⁴GW^5,19ALP23, illustrating similar helix tilt angles and a difference of only 25° in the azimuthal rotation ρ₀ about the helix axis (Figure 4 A,B). While it is apparent that the location of the Arg residue dominates the tilt in both peptides, the small difference in rotation is due to differences in the locations of the interfacial tryptophans, W^4,20 as opposed to W^5,19. Interestingly, residues A19 and A21 of R¹⁴GW^4,20ALP23 fall off their respective curves for the core helix in all three lipids, whereas A3 fits in all three cases (Figure 4A). These features imply more extensive fraying of the helix at the C-terminal as opposed to the N-terminal.^10,23 The circular dichroism spectra (Figure 5A) show differences in the ε₂₂₂ to ε₂₀₈ ratio in several lipid vesicles, which may be due to small lipid-dependent changes in helix backbone structure. Analysis using the CONTIN-LL program based on ridge regression,^68-69 available on the DICHROWEB online server,^49-50 gave estimates of 66% helix in DLPC, 67% in DMPC and 54% in DOPC. Analysis using an alternative method K2D⁷⁰ gave comparable estimates of 57% helix in DLPC, 69% in DMPC and 77% in DOPC. Notably, the NMR results indicate an incomplete helix from which residues 19-23 are unraveled in each lipid membrane. Differing side-chain torsion angles for the indole rings also could raise uncertainty about the CD estimates of the backbone helicity.⁷¹

Figure 5.

Circular dichroism spectra for A) R¹⁴GW^4,20ALP23 and B) R¹²GW^4,20ALP23 in lipid vesicles. The dotted black lines indicate where the mean residue ellipticity is zero. The DOPC double bond absorbs below 200 nm and is responsible for the distortion shown.

A Gaussian analysis of the ²H quadrupolar splittings ^27-28,65 provides further insight into the dynamics of the R¹⁴GW^4,20ALP23 helix, in particular the lowering of the rotational slippage about the helix axis (σ_ρ). As seen in Table 3, with R14 present, σ_ρ drops from a large value to less than 25° in all three lipids. These values are not only significantly less than those of GW^4,20ALP23, but are also lower than those observed for the parent GW^5,19ALP23 helix. Indeed, a single arginine residue generally dictates a preferred helix rotation ρ₀ and lowers the width of the distribution σ_ρ about ρ₀.⁶⁴ The strikingly low extent of rotational slippage exhibited by R¹⁴GW^4,20ALP23 implies that the mere presence of R14 is enough to restrict the host peptide helix to minimal amounts of motional averaging.

R¹²GW^4,20ALP23

In the framework of GW^5,19ALP23, an introduction of R12 leads to multiple states for the helix.¹⁴ In similar fashion, in the framework of GW^4,20ALP23, at least one residue, A9, exhibits two states (Figure 6; see also Figure S4). Do the two different quadrupolar splittings of 11.4 kHz and 27.4 kHz (Table 2) for residue A9 in R¹²GW^4,20ALP23 reflect global changes for the entire molecule or local structural plasticity near residue A9? We note in this case that local variations are more likely because only residue A9, and possibly A11, exhibits multiple ²H NMR peaks (Figure 6). (Only one state can easily be resolved for A11; see discussion.) This contrasts with the behavior of R¹²GW^5,19ALP23 which shows three states for the global helix in DOPC and generates multiple ²H NMR peaks for every Ala residue.¹⁴ The range of observed ²H-Ala quadrupolar splittings (0.5 to 37 kHz; see Table 2) also is higher for R¹²GW^4,20ALP23 than for R¹²GW^5,19ALP23.

Figure 6.

²H NMR spectra for labeled alanines R¹²GW^4,20ALP23 in mechanically aligned DOPC bilayers. Samples consisted of 1.3 μmol peptide with 80 μmol lipid oriented at β = 90°, temperature 50°C. Between 0.8-1.5 million scans were acquired during 25-50 hr. The label positions and % deuteration are from top to bottom: (A5^100%, A7^50%); (A9^100%, A11^50%); (A13^100%, A15^50%); (A17^100%, A19^50%); (A3^100%, A21^50%).

To fit describe the orientation of the R¹²GW^4,20ALP23 helix, a GALA or Gaussian fit (Table 4) describes a surface-bound α-helix for the C-terminal of the core sequence from residue A9(a) through A19. This α-helical segment is oriented with τ₀ of about 86° with respect to the bilayer normal, rotated such that ρ₀ = 33°, and fits with an RMSD of 1.25 kHz (Table 4) on the surface of a DOPC bilayer. Notably, the alternate configuration for alanine-9 (spectrally indicated by the ²H NMR ∣Δν_q∣ for A9(b); Figure 6 and Table 2) and the sole configurations for alanines 3, 5, and 7 do not fit the α-helix defined by the C-terminal segment (Figure 7A). Instead, the N-terminal segment fits a suitably analyzed 3₁₀-helix on the DOPC membrane surface.

TABLE 4.

R¹²GW^4,20ALP23 Structure, Orientation and Dynamics at the Surface of DOPC Bilayers^a

	C-terminus					N-terminus
α-helix	τ0	ρ0	σρ	στ	RMSD	τ0	ρ0	σρ	στ	RMSD
2H	83°	31°	12°	28°	1.28b			No Fit
15N,1H/15N	87°	42°	25°	5°	0.69	88°	178°	44°	7°	0.89
Combined	84°	34°	22°	5°	0.86			No Fit
310-Helixc
2H			No Fit			85°	353°	27°	30°	0.66d
15N,1H/15N	86°	22°	42°	5°	0.66	88°	348°	48°	7°	0.74
Combined			No Fit			85°	349°	29°	27°	1.22

^aThe N-terminal fits were calculated using ²H resonances for residues A3, A5, A7, A9b, and A11, or ¹H/¹⁵N observables for residues A5, L6, A7, L8, A9b. The C-terminal data points included ²H quadrupolar splittings for residues A9a, A11, A13, A15, A17, A19, or ¹H/¹⁵N observables for residues A9a, A13, L14, A15, L16 and A17. The combined fits used the ²H quadrupolar splittings together with the indicated ¹H/¹⁵N observables. For these surface helices, the uncertainties in τ₀ and ρ₀ are about 2°. The widths of the distributions around τ₀ and ρ₀ are given by στ and σρ.

^bFor comparison with the Gaussian fit for the ²H data to the C-terminal α-helix, a semi-static GALA analysis gave τ₀ = 86° ρ₀ =33°, S_zz of 0.64 and RMSD of 1.25 kHz.

^cThe results suggest that the ¹⁵N/¹H results could fit either an α-helix or a similarly oriented 3₁₀-helix for either segment (N-terminal or C-terminal), with in all cases a significant discontinuity (change in azimuthal rotation ρ₀) at residue 9. The ²H results, by contrast, show distinct preferences of an α-helix for the C-terminus and a 3₁₀-helix for the N-terminus. The ²H quadrupolar splitting is highly sensitive to the local bond orientation.

^dFor comparison, a semi-static GALA analysis gave τ₀ = 86°, ρ₀ = 3°, S_zz of 0.51 and RMSD of 0.47 kHz.

Figure 7.

GALA quadrupolar wave plots for R¹²GW^4,20ALP23 DOPC bilayers. A. Wave plot for the C-terminal α-helix (τ₀ 86°, ρ₀ 33°), with data points for A9b, A7, A5 and A3 (red) not fitting the curve and omitted in the analysis used to generate the wave plot. B. Wave plot for a 3₁₀-helix for the N-terminal (A3-A11 with A9b).

From the CD spectra (Figure 5), the estimates of the helicity are lower for R¹²GW^4,20ALP23 in DOPC than in thinner membranes, namely 37% in DOPC as opposed to 70% (DMPC) or 71% (DLPC), based on the CONTIN-LL method.^68-69 The K2D method⁷⁰ gave similar predictions of 31% in DOPC, 60% in DMPC and 65% in DLPC. Consistent with these CD predictions, notably, the NMR results (see below) indicate that a portion of the helix converts to a 3₁₀ helix in DOPC.

While uncommon in typical transmembrane domains, the arginine rich voltage sensing domains of channel proteins are sometimes observed to contain 3₁₀-helix motifs;^72-74 see Discussion. Modifying the GALA and Gaussian calculations to fit the N-terminal portion of the peptide to a tighter 3₁₀-helix yields interesting results that provide further insight. The canonical Pauling 3₁₀-helix with exactly 3.0 residues/turn⁷⁵ would not fit the experimental data, since every third residue would occupy the same position on a helical wheel and would therefore generate the same ²H quadrupolar splitting, which is not the case here. In line with our observations, nevertheless, the average helical wheel residue separation in 3₁₀-helices observed in nature is not 120° but rather is roughly 112.5°, which corresponds to about 3.2 residues/turn. We therefore used values of 112°-114° for the radial separation to model the solid-state NMR data for the N-terminal as a 3₁₀-helix.^61,76 When applied to the ²H quadrupolar splittings (Table 2), or in combination with the ¹⁵N data (see below), the best Gaussian and GALA fits predict a surface bound N-terminal 3₁₀-helical segment with 112.5° as the average residue radial separation, oriented with τ₀ of 85°-88°, ρ₀ around 350° and RMSD values between 0.5 and 0.9 kHz (Table 4). These N-terminal average tilt angles above 85° calculated for the tighter 3₁₀-helix are strikingly similar to the predicted surface orientation of the C-terminal α-helix. The discontinuity at residue A9, nevertheless, indicates an unwinding denoted by the 30°-40° difference in azimuthal rotation about the axes of the α- and 3₁₀-helices (Table 4; Figure 7). Notably, the ²H NMR observables can only fit the C-terminal residues to an α-helix and the N-terminal residues to a tighter helix such as a 3₁₀-helix. For the N-terminal ²H data, a modified Gaussian calculation (see ²⁴) was incorporated in order to obtain an estimate of the dynamics if fitted to a 3₁₀-helix. In this case, both σρ and στ would be moderate (table 4). Therefore, a tighter helix, such as a 3₁₀-helix, with likely varying torsion angles,^61,76 is a probable structure for the N-terminal portion of R¹²GW^4,20ALP23 on the DOPC membrane surface.

The helix discontinuity was confirmed also with ¹⁵N experiments (Figure 8). The spectra in Figure 8 were drawn from four differently labeled samples of R¹²GW^4,20ALP23 and a control sample of KWALP23^{11, 64} to represent a prototype transmembrane helix. Sample variation led to differing signal intensities and signal-to-noise ratios. One-dimensional spectral slices in the dipolar coupling dimension are shown for each ¹⁵N-labeled residue in Figure S5 of the supporting information. The ¹⁵N NMR spectra indicate and confirm the surface orientation and the discontinuity at residue A9 for the R¹²GW^4,20ALP23 helix. The ²H and ¹⁵N solid-state NMR experiments therefore agree. For example, the spectra obtained from the SAMPI4 experiments (Figure 8) have ¹⁵N chemical shifts between 95 and 70 ppm as well as dipolar couplings between 4 and 6 kHz, which indicate a surface bound orientation,⁷⁷ perpendicular to the membrane normal, in contrast to the spectrum for a control transmembrane helix (Figure 8A). Residue A9 once again shows resonance doubling. Furthermore, the dipolar wave plot depicted in Figure 9 shows a discontinuity in frequency and amplitude of the dipolar wave at residue A9, as the patterns N-terminal and C-terminal to residue A9 are distinctly different, with a much larger amplitude for the wave C-terminal to alanine 9.

Figure 8.

Separated local field ¹⁵N spectra for a transmembrane helix and a distorted surface helix. A. The red peaks arise from resonances for selected ¹⁵N backbone labels on transmembrane KWALP23 in static aligned DLPC bilayers oriented at β = 0°, with the ellipse representing a helix tiled 19° from the bilayer normal. The blue peaks arise from ¹⁵N labels in R¹²GW^4,20ALP23, bound to the surface of DOPC bilayers. B. Expansion and highlights for multiple selectively labeled R¹²GW^4,20ALP23 peptides in static aligned DOPC bilayers (1.3 μmol peptide with 80 μmol lipid) oriented with β = 0°, 1400-1600 scans and temperature of 50 °C over ~22 hrs. The PISA wheels shown are fitted to C-terminal residues (red, α-helix) and N-terminal residues (blue, 3₁₀-helix). See also table 4. The assignments for ¹⁵N backbone labels are shown.

Figure 9.

Dipolar waves depicting the static ¹H-¹⁵N dipolar couplings as a function of residue position for transmembrane GW^5,19ALP23 in DLPC bilayers (top) adapted from ⁶⁰ and surface bound R¹²GW^4,20ALP23 in DOPC bilayers (bottom).

Supporting the ¹⁵N dipolar wave, the expanded region of the two-dimensional SAMPI4 spectrum (Figure 8B) shows separate elliptical patterns for the C-terminal and N-terminal segments. The ¹⁵N polarity index slant angle (“PISA”) pattern in Figure 8B predicts a tilt τ₀ of 89° and a ρ₀ of 42° for the C-terminal residues (Table 4), in agreement with the ²H NMR data. The small difference of 10° in the estimate for ρ₀ (Table 4) is likely due to the lower sensitivity of the SAMPI4 experiment, compared to the ²H experiment, to the helix azimuthal rotation. Once again, the ¹⁵N chemical shift range for the N-terminal residues indicates a surface-bound orientation for this portion of the helix (Figure 8).^78-79

Gaussian analyses reported in table 4 indicate that the ¹⁵N NMR observables “could” fit the N-and C-terminal segments, with different ρ₀ values, to α-helical or 3₁₀-helical segments on the DOPC membrane surface. The ambiguity for the ¹⁵N experiment is expected due to the similar locations of 90° tilted peptide 3₁₀-helix and α-helix PISA wheels.⁷⁶ The wheel positions are furthermore influenced by the extent of dynamic averaging.^{18, 37} Nevertheless, the ²H quadrupolar splittings can fit the C-terminal residues only to an α-helix, and the N-terminal residues only to a 3₁₀-helix (see above). Additionally, only a 3₁₀-helix fits the ¹⁵N data for the N-terminal if the extent of rotational slippage (σρ) is moderate (σρ of 39 - 44°) instead of very high. A hypothetically higher σρ would in turn lead to more signal averaging and “would” make feasible an N-terminal α-helix, with similar PISA wheel size as the blue wheel in Figure 8B. Altogether, nevertheless, the ¹⁵N and ²H NMR observables agree concerning the surface location for R¹²GW^4,20ALP23, the rotational discontinuity of 30°-40° at residue A9, the C-terminal α-helix motif and the N-terminal 3₁₀-helix motif (Table 4).

DISCUSSION

Arginine dominates the GWALP23 peptide helix dynamics and behavior in a manner dependent on its location within the sequence relative to those of the juxta-terminal interfacial Trp residues. In the original GW^5,19ALP23 peptide, the presence of R14 on the opposite face of the helix from that occupied by the two Trp residues (Figure 2) allows the peptide to remain transmembrane in bilayers of DOPC, yet with a 10° increase in tilt and 80° change in helix azimuthal rotation.¹⁴ By contrast, R12, more centrally located and effectively “trapped” within a Trp “cage” defined by W5 and W19, leads to multiple states for the helix, including two competing transmembrane orientations and one at the surface of DOPC bilayers.¹⁴ The multi-state behavior can be “rescued” by moving the tryptophans outward to positions 3 and 21.¹⁶ In the context of W3 and W21, arginine R12 as well as R14 can be accommodated in a suitably tilted transmembrane helix.¹⁶ Toleration of R12 with W3 and W21 as opposed to W5 and W19 has been attributed to the guanidium group occupying a different helix face and no longer needing to compete with the Trp residues for favorable interactions at the interface. Interestingly, R14 also is accommodated with W3 and W21, perhaps because the Trp “cage” is effectively larger than that defined by W5 and W19.

For the present investigation, the structural context for W4 and W20 on the helix framework is entirely different. From this perspective, neither position 12 nor position 14 is located on a helix face containing a Trp residue. Therefore, while the two former cases (W^5,19 and W^3,21) have addressed primarily the possibility that Trp could restrict water access to a central Arg side chain, the W^4,20 sequence removes this side chain competition and introduces instead a new factor, high dynamic motion for the parent helix when no arginine is present. Indeed, the helix of GW^4,20ALP23 undergoes extensive motional averaging in the form of rotational slippage about the helix axis in order to compensate for apparently competing radial locations of the Trp residues and a rotation-dependent hydrophobic mismatch.⁴² The inclusion of Arg at either position 12 or 14 within this sequence severely limits the excessive dynamic averaging and in each case leads to helix structural distortions, previously unobserved in the former GW^5,19ALP23 framework.

Importantly, the Arg side chain carries a positive charge under all conditions.^{15, 80} The finding is verified by numerous experimental and computational results.^{15, 81-84} Indeed, the solution pKa for Arg has been revised upward to a value of 13.8,⁸⁰ and recent experiments showed that R¹⁴GW^5,19ALP23 remains fully charged up to pH 13 within an ether-linked lipid environment.¹⁵ Helix translocation, side chain snorkeling and membrane deformation can also serve to stabilize the positive charge, allowing the Arg side chain to engage in favorable interactions at the membrane interface.^{14, 85-86}

Influence of R14. The unmodified GW^5,19ALP23 helix already displayed a transmembrane orientation with low levels of dynamic averaging.²⁸ Incorporating R14 resulted in a 10° increase in tilt, 80° change in helix rotation and mean helix displacement/membrane thinning (observed via coarse grain simulations)¹⁴ that allow the Arg to snorkel and access the membrane interface. The transmembrane orientations for the helices with W^5,19 and W^4,20 are similar, with low dynamic averaging for each, when R14 is present. The respective helix tilt angles differ by about 9° in DLPC and DMPC (Table 3) but by only about 3° in DOPC. The helix azimuthal rotation differs modestly by 25-36° when the Trp sequence context is changed with R14 present (see Table 3, Figure 4). The arginine residue R14 is therefore the primary determinant of the helix tilt and azimuthal rotation, but the interfacial Trp residues – whether W5 and W19, or W4 and W20 – exert secondary influence for fine tuning of the helix tilt and rotation. Residue R14 also lowers the extent of dynamic averaging, dramatically for the highly dynamic GW^4,20ALP23 helix, and much more modestly for the already low-averaging GW^5,19ALP23 helix.⁶⁴ The rotational slippage in the form of σρ for both helices with R14 is remarkably low in all three lipids (< 25°, Table 3), indeed lower than for GW^5,19ALP23 with arginine absent. Therefore, the single arginine residue governs the overall properties of these transmembrane helices, with the small differences in the tilt and rotation due to the locations of the juxta-terminal Trp residues.

The dominance of arginine R14 over the peptide dynamics would also explain the helix unwinding observed at the C-terminal in R¹⁴GW^4,20ALP23. The opposing radial positions of the distal tryptophans, W4 and W20, are responsible for the high dynamic motion exhibited by the host peptide. This arrangement causes the indole side chains to compete with one another for better positions at the lipid/water interface. Furthermore, as the helix cannot solely rely on adjusting its τ₀ and ρ₀ to satisfy hydrophobic mismatch, it additionally exhibits increased oscillations about its average ρ₀ to meet the demands of the membrane interior.⁴² Incorporating R14 into GW^4,20ALP23 introduces an interaction between the Arg side chain and the lipid membrane that is strong enough to drastically limit the rotational averaging about ρ₀, exemplified by the massive drop in σ_ρ from 122° to 10° in DOPC, essentially locking the transmembrane helix into place. In spite of the arginine dominance, the competition between the two Trp residues remains, such that W20 likely causes additional C-terminal residues to unravel from the core helix (Figure 4), in order to obtain a preferential orientation for the W20 indole ring at the lipid/water interface. While residues A3 and A21 often are observed to unwind (see Figure 3),^23-24 now additional fraying of residues W20 and A19 is observed when R14 is present. By contrast, residue A3 near the N-terminal now fits to the central helix of R¹⁴GW^4,20ALP23, indicating a shifting of the midpoint of the core helix. The N-terminus is likely compensating for the unwinding at the other end of the helix (see Figure 10 for a model).

Figure 10.

3D model of R¹⁴GW^4,20ALP23. A selection of possible arginine rotamers from the set in ref. ⁹¹ is illustrated as potentially “snorkeling” toward the membrane interface. The C-terminal unwinding begins at residue A19.

Influence of R12. Placing the Arg residue (R12) at the center of GWALP23 sequences with varying locations for the outer Trp residues has interesting consequences. The peptide R¹²W^5,19ALP23 produces ²H NMR spectra with multiple states for every alanine residue in DOPC bilayers. According to molecular dynamics simulations, it adopts three primary states of which two are transmembrane and one is at the membrane surface.¹⁴ When the tryptophans are moved outward by two residues each, R¹²W^3,21ALP23 has ample room to accommodate R12 between the outer Trp residues.¹⁶ The helix with W3 and W21 is able to remain transmembrane by adopting large tilt angles (24-30°) to accommodate the snorkeling of R12 toward the surface of DOPC, DMPC or DLPC bilayers, without interference from the aromatic side chains.

Introducing a small amount of cholesterol (10 mol %) was enough to drive R¹²GW^5,19ALP23 completely to the surface of DOPC bilayers.¹⁵ Similarly, H12 and K12 both drive GW^5,19ALP23 to the membrane surface, at low pH when H12 or K12 is positively charged.^8-9 Here, R¹²GW^4,20ALP23 adopts a single overall orientation perpendicular to the DOPC bilayer normal, yet unlike the previous peptides, R¹²GW^4,20ALP23 is distorted when bound on the membrane surface. The distortion is apparent from (a) the NMR resonance doubling of A9 and (b) the N-terminal residues preceding A9 not fitting to the same quadrupolar wave plot as the rest of the helix on the C-terminal side of A9 (Figures 6-8). The resonance doubling of A9 is likely observed because the helix actually starts to distort at residue A11, next to the central Arg at position 12. Thus, A9 is found, with about equal probability, within both the N-terminal and C-terminal helical segments. While residue A11 gives only one major ²H NMR signal, additional minor peaks may be evident (Figure 6), although we are unable to assign specific minor peaks. Interestingly, and seemingly by coincidence, the major ²H ∣Δν_q∣ value for A11 fits both the N-terminal 3₁₀-helix and the C-terminal α-helix (Figure 7). While R14 caused unwinding of the R¹⁴GW^4,20ALP23 helix at A19 (see above), the larger and more central distortion with R12 likely also is caused by the competing Trp residues W4 and W20, while the Arg again dominates the peptide dynamics. The C-terminal helix exhibits σρ of 12°-25°, significantly lower than the parent helix when Arg is absent (Table 4). We note that the CD spectra (Figure 4B) are not particularly sensitive to the helix distortion revealed by the ²H and ¹⁵N NMR spectra.

As noted, it is likely that the R¹²GW^4,20ALP23 helix on the DOPC membrane surface contains a 3₁₀-helical segment as well as an α-helical segment. The ²H quadrupolar splittings of the C-terminal alanines fit only to an α-helix. Fitting the ²H quadrupolar splittings of the N-terminal alanines to an α-helix is not possible, nevertheless, even in combination with the ¹⁵N data (table 4). On the other hand, a tighter 3₁₀-helix predicts a surface bound N-terminal segment with only a 40° difference in azimuthal rotation from the C-terminal helix. The rotational difference and the tighter N-terminal 3₁₀-helix then would allow the W20 side chain to reside at the membrane interface, oriented toward the lipids (see Figure 11), instead of projecting out of the membrane as would have been dictated by an extension of the C-terminal α-helix. Indeed, the ¹⁵N chemical shifts also discount the possibility of a transmembrane orientation and are instead characteristic of an orientation perpendicular to the bilayer normal.^29,76 In a 3₁₀-helix, the carbonyl oxygens are more exposed, such that the folding of such a transmembrane structure is unfavorable due to the low dialectic of the bilayer interior.⁶⁶ On the membrane surface, nevertheless, a 3₁₀-helix becomes a reasonable motif for adjusting the relative radial locations of the side chains of W4 and W20, such that both of the indole rings can face the membrane interface (Figure 11B,C). We note as well that both the C-terminal α-helix and N-terminal 3₁₀-helix orientations are similar to the interfacial state among the multiple orientations for the R¹²GW^5,19ALP23 helix (Figure 11 A). Interestingly, the addition of 5% cholesterol to the DOPC membrane drives essentially all of the R¹²GW^5,19ALP23 helix population to the membrane surface.¹⁵ Moving the tryptophans outward to positions 3 and 21, nevertheless, retains a tilted transmembrane helix (Figure 11 D.)¹⁶

Figure 11.

Models to illustrate the varying orientations, in DOPC bilayers, of 23-residue peptide helices with a central arginine residue, R12, and varying juxta-terminal tryptophans. A. Multiple states observed, with about equal probabilities,¹⁴ for the helix of R¹²GW^5,19ALP23. The three states consist a surface helix and two transmembrane helices in which the R12 side chain snorkels either up or down.¹⁴ B. Distortion of the surface-bound helix of R¹²GW^4,20ALP23 to yield a 3₁₀-helical segment (blue) for the N-terminal, based on the NMR evidence in figures 7-8. C. A model with lipids and the hybrid alpha(left)-3₁₀(right) helix of R¹²GW^4,20ALP23, rotated 180° from the view in panel B. D. Illustration of the stable transmembrane helix of R¹²GW^3,21ALP23,¹⁶ wherein the interfacial tryptophans are moved outward and away from the central R12 residue.

The ¹⁵N separated local field experiments are in full support of a surface-bound helix, yet essentially show little sensitivity to the geometric differences between an α-helix or 3₁₀-helix at this orientation. The insensitivity is largely because the plane perpendicular to the magnetic field/bilayer normal is additionally a reflection plane for the resonance frequencies^50-87 resulting in overlapping PISA wheel arcs over a small range of ¹H-¹⁵N dipolar couplings and ¹⁵N chemical shifts, which are both further reduced by motional averaging. The ²H experiments span a much wider frequency range (0-50 kHz)^{15, 88} for surface bound helices and are therefore more sensitive in distinguishing the particular type of helix. The distinction would be easier for a transmembrane oriented peptide, as α-helices and 3₁₀-helices aligned on a plane parallel to the magnetic field each produce a distinctive PISA wheel pattern ⁷⁶ Indeed there are precedents for an Arg-rich motif containing an α-helix that kinks into a 3₁₀-helix, for example in the voltage sensing domains of membrane channel proteins.^89-92 A model for such a helix transition may be manifest here.

CONCLUSIONS

Incorporating one single central Arg residue into the highly dynamic GW^4,20ALP23 helix framework has led to unique consequences. Placing Arg at position 14 arrests the dynamics, reorients the helix and causes the C-terminal residues around W20 to unwind from the helix, probably to optimize the interfacial interactions of residue W20. By contrast, an Arg residue at position 12 brings the entire helix to the surface of DOPC bilayer membranes and distorts the helix so that residues 3-11 form a 3₁₀-helix while residues 9-19 remain α-helical, with deuterated Ala-9 itself giving two distinct ²H NMR spectral signals that represent both of the helix motifs.