Rapid ²H NMR transverse relaxation of perdeuterated lipid acyl chains of membrane with bound viral fusion peptide supports large-amplitude motions of these chains that can catalyze membrane fusion

Abstract

An early step in cellular infection by a membrane-enveloped virus like HIV or influenza is joining (fusion) of the viral and cell membranes. Fusion is catalyzed by a viral protein that typically includes an apolar “fusion peptide” (fp) segment that binds the target membrane prior to fusion. In this study, the effects of non-homologous HIV and influenza fp’s on lipid acyl chain motion are probed with ²H NMR transverse relaxation rates (R₂’s) of perdeuterated DMPC membrane. Measurements were made between 35 and 0 °C which brackets the membrane liquid-crystalline-to-gel phase transitions. Samples were made with either HIV “GPfp” at pH 7 or influenza “HAfp” at pH 5 or 7. GPfp induces vesicle fusion at pH 7 and HAfp induces more fusion at pH 5 vs 7. GPfp bound to DMPC adopts intermolecular antiparallel β sheet structure, whereas HAfp is a monomer helical hairpin. The R₂’s of the no peptide and HAfp, pH 7 samples increase gradually as temperature is lowered. The R₂’s of GPfp and HAfp, pH 5 samples have very different temperature dependence, with ~10× increase in R₂^CD2 when temperature is reduced from 25 to 20 °C, and smaller but still substantial R₂’s at 10 and 0 °C. The large R₂’s with GPfp and HAfp, pH 5 are consistent with large-amplitude motions of lipid acyl chains that can aid fusion catalysis by increasing the population of chains near the aqueous phase, which is the chain location for transition states between membrane fusion intermediates.

Graphical Abstract

Introduction

Many zoonotic pathogens like human immunodeficiency virus (HIV), influenza virus, and coronaviruses are membrane-enveloped viruses. An early step in infection by these viruses is joining (fusion) of the viral and target cell membranes.^1-5 Fusion is catalyzed by a viral protein that is specific to each virus family and which typically has a N-terminal receptor-binding subunit (RbSu) and C-terminal fusion subunit (FsSu). The RbSu binds to specific (to the virus) molecules in a target cell membrane, followed by separation of the RbSu from the FsSu, and then catalysis of membrane fusion by the FsSu. The FsSu is typically a monotopic integral membrane protein of the virus with a substantial N-terminal ectodomain (Ed) outside the virus. HIV, influenza, and coronaviruses all have “class I” fusion proteins for which the Ed’s of three FsSu’s form a complex with three RbSu’s.¹ After receptor-binding and RbSu separation, there is substantial structural change of much of the three FsSu Ed’s to a thermostable trimer-of-hairpins structure. The N-terminal region of the FsSu Ed is not part of the hairpin and a ~25-residue segment in this region is called the “fusion peptide” (fp) and is thought to bind the target membrane during fusion (Fig. 1A, E).⁶ The fp segment is typically a conserved sequence with apolar residues and defined structure in detergent and/or membrane, and exhibits mutations that disrupt FsSu-mediated fusion.^4,5,7-21 For the HIV glycoprotein (GP) 41 kDa and influenza hemagglutinin (HA) subunit 2 FsSu’s, the consensus fp’s are at the N-termini of the respective FsSu’s, with GPfp and HAfp sequences presented in Fig. 2B.

Figure 1.

Fusion mechanism model. Panels A-D display membrane states during fusion. In the absence of fusion subunit protein, calculations suggest that ~25 kcal/mole energy is required to closely-appose two membranes and that there are ~10 kcal/mole barriers to formation of each of the three subsequent states. The membrane apposition energy may be reduced by the thermostable hairpin structure of the soluble ectodomain of the viral fusion protein subunit in conjunction with the fusion peptide bound to the target membrane and the transmembrane domain in the viral membrane (panel E). Both monomer and trimer forms of the hairpin have been reported and the trimer would have three fusion peptides and three transmembrane domains. The barrier for the apposed membranes → stalk step (A →B) may be reduced by perturbation of the acyl chains of the target membrane lipids near the fusion peptide. Panel F shows a schematic model of one type of perturbation that is suggested by molecular dynamics simulations of membrane with monomer HAfp. These simulations show larger motional amplitudes and higher probability of protrusion, i.e. location of the lipid acyl chain closer to the aqueous phase. The HAfp in panel F is represented as the semi-closed helical hairpin structure located in the outer leaflet of the target cell membrane with some hydrophobic sidechains in the membrane interior and some polar sidechains in the aqueous phase.

Figure 2.

Panel A displays the chemical structure of DMPC-d54 lipid with D ≡ ²H. Panel B displays sequences and structural models of GPfp and HAfp based on data for peptide without the rest of the protein. A few residues are identified in the models. One strand of a GPfp intermolecular antiparallel β sheet is displayed. The sheets have distributions of adjacent strand registries that include the residue 1→16/16→1 and 1→17/17→1 registries. HAfp is a monomeric helical hairpin with populations of closed and semi-closed structures. The closed structure is based on PDB 2KXA.

The present study is an investigation of the effects of GPfp and HAfp on lipid acyl chain motion via measurement of ²H NMR transverse relaxation rates (R₂’s) of dimyristoylphosphatidylcholine lipid with perdeuterated acyl chains (DMPC-d54, Fig. 2A). The R₂ has the greatest contribution from “slow” motions which have significant spectral density at frequencies <100 kHz.^22-26 The R₂ depends both on the mean-squared variation of the ²H quadrupolar NMR frequency due to slow motions and the spectral density in the low-frequency regime. The temperature-dependence of R₂ is interesting because lowering temperature has countervailing effects of decreasing amplitudes and increasing the spectral density of slow-frequency motions. The samples of this study contain membrane with only DMPC-d54, in part to take advantage of defined but different predominant structures for GPfp and HAfp in this single lipid (Fig. 2B).²⁷ GPfp is an oligomer and adopts intermolecular antiparallel β sheet structure whereas HAfp is a monomer and adopts monomer α helical hairpin structures.^12,14,18 The number of GPfp in an oligomer is likely fairly small, e.g. <20.²¹ Other structures are adopted by GPfp and HAfp when they are bound to environments other than DMPC. For example, GPfp adopts predominant monomer α helical structure in detergent-rich media whereas in membrane with 30 mole% cholesterol, HAfp adopts predominant oligomeric β sheet structure.^11,28-30 The use of single-component DMPC-d54 membrane in the present study allows comparison of the effects on lipid of the two non-homologous GPfp and HAfp sequences as well as the different intermolecular β and monomer α structures with the goal of discerning fp effects on membrane that are more universal. There is also evidence that the GPfp and HAfp structures are retained for the fp’s in much larger constructs that include the Ed hairpin.^16,17,31-33

Samples with GPfp are prepared at neutral pH which matches the pH of HIV/cell fusion, whereas samples with HAfp are prepared at both pH 5 and pH 7.^34-36 Fusion of influenza virus happens within late endosomes for which pH ≈ 5.¹ The GPfp with oligomeric β structure induces substantial vesicle fusion at pH 7 whereas HAfp with helical hairpin structure induces higher vesicle fusion at pH 5 vs. 7, which also matches the pH dependence of vesicle fusion by large HA2 constructs that include HAfp.^18,20,37,38 Several groups have investigated the basis for why HAfp is more “fusogenic” at pH 5 vs. 7, but there isn’t yet a consensus on the mechanism for the pH difference.^18,39-42 It has been shown that HAfp in membrane adopts two variants of helical hairpin structure, closed and semi-closed (Fig. 2B), and there is larger semi-closed fraction at pH 5 vs. 7.^15,18

Fig. 1 displays one model of the fusion mechanism that illustrates potential contributions of the fp to fusion. In the absence of FsSu, there is ~25 kcal/mole calculated barrier to the membrane apposition step which precedes fusion (Fig. 1A ,E).² This barrier could be reduced by fp binding to the target membrane in conjunction with the thermostable hairpin structure of the C-terminal Ed region of the FsSu.^20,38,43,44 The barrier could be further lowered by membrane dehydration induced by fp.⁴⁵ Apposition is followed by “stalk” and then “hemifusion diaphragm” intermediates which are respectively contiguous with the outer and inner leaflets of the two fusing bodies (Fig. 1B, C).^2,4,34,46 Small pores are formed and at least one pore expands so that the diaphragm is eliminated and there is a single contiguous membrane and compete mixing of contents of the two bodies (Fig. 1D). In the absence of FsSu, there are ~10 kcal/mole calculated barriers between the apposition → stalk, stalk → hemifusion, and hemifusion → fusion pore structures.² The barriers exist in part because lipids have to move through transient configurations with energies higher than those of a typical membrane bilayer, e.g. moving a small group of lipids through water to form a new leaflet in a new bilayer. One way the membrane-bound fp could reduce these barriers is by increasing amplitudes of lipid thermal motions with accompanying larger populations of configurations that are closer to those of the transition-state geometries. There are both data that support increased motional amplitudes with fp and data that support decreased amplitudes, so the fp effect on lipid motions remains unresolved.

Experimental support for increased amplitudes includes narrower ²H NMR spectra for DMPC-d54 with vs. without fp.²⁷ Spectra were also obtained over a range of temperatures that spanned the liquid-crystalline and gel phases of the lipid, and relative to without fp, comparable spectral lineshapes and linewidths with fp were obtained at 10-20 °C lower temperature. These effects were observed with HAfp at pH 5 or with GPfp at pH 7 which respectively adopt monomeric helical hairpin or oligomeric β sheet structure (Fig. 2B). By contrast, there were comparable lineshapes and linewidths for no peptide and HAfp at pH 7, where less fusion is induced. The frequency differences between “horns” in the liquid-crystalline spectra are proportional to the order parameters S_n = ⟨(3 cos²θ_n − 1)/2⟩, with 0 ≤ S_n ≤ 1 where n is the index for a particular acyl chain carbon, θ_n is the angle between the C_n-²H bond and the NMR magnetic field, and ⟨…⟩ is the average over “fast” motions whose frequencies are >100 kHz.²⁶ At 35 °C and 1:25 peptide:lipid ratio, the order parameters for GPfp and HAfp, pH 5 are respectively ~15 and ~8% narrower/smaller than no peptide and for HAfp, pH 7, ~4% broader/larger. At 1:50 ratio, the magnitudes of the changes are smaller, i.e. the spectra with peptide become more similar to no peptide. Larger-amplitude lipid motions are also observed in several molecular dynamics simulations for membrane with monomer HAfp, with highest amplitudes for lipids close to a HAfp.^47-49 There is semi-quantitative agreement between the NMR- and simulation- derived reductions in order parameters for lipid with HAfp, pH 5, and also agreement between NMR- and X-ray ray scattering- derived reductions for lipid with GPfp.^27,47-50 For the simulations, the increased motions result in acyl chain protrusion into the aqueous phase and headgroup intrusion into the hydrocarbon interior of the bilayer, i.e. a partial rotation of the lipid molecule that is along the reaction coordinate between topologically-distinct fusion intermediates.^4,47-49,51 In the present study, these simulation and other data are used to interpret large differences in experimental ²H NMR R₂’s of lipid with vs. without fp, with resulting improved understanding of the fp contribution to membrane fusion.

Materials and Methods

Materials.

GPfp has sequence AVGIGALFLGFLGAAGSTMGARSWKKKKKKG and HAfp has sequence GLFGAIAGFIEGGWTGMIDGWYGGGKKKKG, and the underlined residues are respectively the N-terminal residues of the HIV gp41 and the influenza virus HA2 proteins. The GPfp sequence is from the LAV-1 laboratory strain of HIV and the HAfp sequence is a H1 serotype. The C-terminal, non-underlined residues are non-native. The G and K residues increase aqueous solubility and the W in GPfp is a 280 nm absorption chromophore. Synthesis and purification have been previously described.²⁷ DMPC-d54 lipid was purchased from Avanti Polar Lipids (Alabaster, Al). Buffers were 10 mM HEPES and 5 mM MES adjusted to either pH 5.0, 7.0, or 7.4, and contained 0.01% sodium azide as preservative.

NMR sample preparation.

The sample preparation has been previously described and reliably results in HAfp with helical hairpin and GPfp with β sheet structures.²⁷ Briefly, the “no peptide” lipid film was prepared by dissolving dry DMPC-d54 in chloroform:methanol (9:1), followed by solvent removal with nitrogen gas and overnight vacuum. The film was suspended in water, subjected to multiple freeze/thaw cycles, and extruded through a 100 nm diameter polycarbonate filter to form large unilamellar vesicles (LUVs). The LUV suspension was ultracentrifuged and the pellet was harvested, lyophilized, transferred to a 4 mm diameter NMR rotor, and hydrated with ~20 μL of water. Samples with peptide were typically prepared with 1:25 peptide:DMPC-d54 molar ratio. For HAfp samples, DMPC-d54 LUV’s were prepared in buffer at either pH 5.0 or pH 7.0 and subjected to dropwise addition of a HAfp solution while maintaining pH. The HAfp + LUV suspension was then treated similarly to the no peptide except that hydration was done with buffer that maintained the pH of the sample. Preparation of GPfp and DMPC-d54 began with co-dissolution in 2,2,2-trifluoroethanol:chloroform:1,1,1,3,3,3-hexafluoroisopropanol (2:3:2) followed by solvent removal with by nitrogen gas and overnight vacuum. The peptide + lipid film was treated similarly to the no peptide film except that there wasn’t extrusion and buffer at pH 7.4 was used.

NMR spectroscopy and data analysis.

²H NMR spectra were acquired on a 9.4 T Infinity Plus spectrometer using a probe equipped for 4 mm diameter rotors. Nitrogen gas at defined temperature was flowed around the sample for ~1 h prior to data acquisition and spectra were then obtained with T_N2 between 35 and 0 °C, which brackets the ~20 °C and ~10 °C temperatures of the respective L_α→P_β′ transition and P_β′→L_β′ phase transitions of hydrated DMPC-d₅₄, with L_α ≡ liquid-crystalline, P_β′ ≡ rippled gel, and L_β′ ≡ gel phase.^{52 2}H NMR spectra were acquired without spinning and with the solid-echo pulse sequence (π/2)_x − τ_a − (π/2)_y − τ_b – acquire. Typical parameters included 61.5 MHz transmitter frequency, 1.6 μs π/2 pulse (calibrated with ²H₂O), dwell time = 2 μs, recycle delay = 1 s, and acquisition of a few hundred to a few thousand scans. Data processing included removal of the first 11 points ≡ 22 μs so that the initial time is the peak echo, as well as exponential line broadening, Fourier transformation, phasing, and dc offset correction. For each sample at a particular temperature, data were acquired for an array of τ_a,k and τ_b,k values with typical τ_a,k = 30 μs + (k × Δτ/2), τ_b,k = 11 μs + (k × Δτ/2), and k = 0, 1, 2, … k_max. The NMR signal decay includes terms with exp(−R₂ × τ_k) dependence where τ_k = 63 μs + (k × Δτ). The Δτ and k_max values were selected to allow accurate determination of the R₂’s.

Each processed spectrum contained a central region that had contributions from the narrower −C²H₃ powder pattern and the broader −C²H₂ powder patterns, and also outlying regions that only had contributions from the −C²H₂ powder patterns (Fig. S1). For each set of spectra of a sample at one temperature, the integrated −C²H₃ intensity (I^CD3) was typically calculated using the difference between integration over the central spectral region with −C²H₃ and −C²H₂ contributions and integration over nearby ranges that only contained −C²H₂ contributions, i.e. I^CD3 = _−ν1ʃ^ν1I(ν)dν − _ν2ʃ^ν3I(ν)dν + _−ν2ʃ^−ν3I(ν)dν, with ∣ν₁∣ = ∣(ν₃ − ν₂)∣. Integration over the full −C²H₃ spectral width for _−ν1ʃ^ν1I(ν)dν was achieved with ν₁ ≈ Δν_CD3 where Δν_CD3 is the frequency difference between the prominent innermost “horns” of a spectrum that correspond to the most likely θ = 90° orientation of the −C²H₃ group with respect to the NMR field (Figs. 3A and S1).²⁶ The I^CD2,inner was calculated by integration over the (ν₂ → ν₃) + (−ν₂ → −ν₃) regions and I^CD2,outer was calculated by integration over symmetric frequency ranges in the more outlying regions of the spectrum. The specific frequency values for the integration ranges varied with sample and temperature because the spectral lineshapes and linewidths strongly depended on sample and temperature (Fig. S1).

Figure 3.

²H NMR spectra of samples with DMPC-d54 at different temperatures and different relaxation times ≡ τ. The pulse sequence is the solid-echo sequence π/2 − τ_a − π/2 − τ_b - acquisition so that τ = τ_a + τ_b + 22 μs. Each sample contained either DMPC-d54 without peptide (left columns) or DMPC-d54 with peptide and peptide:DMPC-d54 = 1:25 molar ratio (right columns). For each of the A-D panels, the displayed spectra have similar lineshapes for the shortest-time τ = 63 μs spectra. This similarity indicates similar order parameters of the four samples and similar amplitudes of motions with frequencies >100 kHz. For the same column within a panel, each spectrum is the sum of the same number of scans, where 500 is a typical number. For the same column within a panel, each spectrum is processed with the same exponential line broadening, and the typical broadening is 100 Hz for panel A, 500 Hz for panel B, and 1000 Hz for panels C and D. The −C²H₃ “horns” and −C²H₂ horn regions are identified for the 35 °C no peptide spectrum at shortest τ, where the horns are peaks corresponding to the most probable θ = 90° orientation with respect to the NMR field.

Separate relaxation analyses were done for the I^CD2,inner, I^CD2,outer, and I^CD3 integrated intensities from an array of τ_k ≡ τ values. For most datasets, the plot of ln[I(τ)] vs. τ fit well to a straight line and R₂ ≡ best-fit slope. For the I^CD2 datasets of GPfp and HAfp, pH 5 at 20 °C, the ln[I(τ)] plot was non-linear. These data fit well to a sum of two rate processes, i.e. I(τ) = [A_f × exp(−R_2,f × τ)] + [A_s × exp(−R_2,s × τ)] with f ≡ fast, s ≡ slow, R_2,f and R_2,s as rates, and A_f and A_s as the ratio of contributions to the signal intensity when τ = 0 with A_f + A_s = 1.

Results

Fig. 3 displays DMPC-d54 spectra as a function of τ for some temperatures and Fig. S1 displays spectra for the shortest τ for all temperatures. The preparation of a sample and the NMR data acquisition for the sample were typically done within the same month-long time period. Fig. 3A displays spectra of the four samples at 35 °C. The DMPC-d54 is in the liquid-crystalline phase, as evidenced by the maximal spectral width <40 kHz and the sharp symmetric pairs of “horns” that correspond to θ = 90° signals from the −C²H₃ or specific −C_n²H₂ groups along the myristoyl chains, with n = 2, 3, … 13 (headgroup to methyl group). The individual DMPC-d54 molecules are rotating rapidly about the bilayer normal direction and θ approximately corresponds to the angle between the local bilayer normal and the NMR field direction. The most prominent pair of horns with smallest frequency separation Δν are assigned to the −C²H₃ group (n=14) and pairs with increasing Δν are assigned to −C_n²H₂ with monotonically-decreasing n, with the outermost pair assigned to n = 2→6.²⁷ Overall, there are narrower spectra and smaller order parameters for the GPfp and HAfp, pH 5 vs. no peptide and HAfp, pH 7 samples.

The intensities of the Fig. 3A 35 °C spectra of the four samples exhibit similar decays with increasing τ which supports similar R₂ values for all samples. There is also slower decay with increasing τ of the central vs. outer spectral regions that respectively contain −C²H₃ + −C²H₂ vs. only −C²H₂ signals. These visual observations are borne out in the fittings of ln[I(τ)] vs. τ, with similar R₂^CD3’s of the four samples, similar R₂^CD2’s of the four samples, and R₂^CD2/R₂^CD3 ratios in the 2.7 – 3.0 range (Table 1 and Fig. S1). For a particular sample, the R₂^CD2’s are similar for integrations in the outer and inner −C²H₂ horn regions, where the former is dominated by −C²H₂ closer to the headgroup and the latter has significant contributions from −C²H₂ close to the methyl group (Fig. S1). The Discussion section provides a more detailed description of the R₂’s for the different samples and integration ranges. For HAfp, pH 7, there are similar R₂’s determined from NMR data from a different sample with peptide:lipid = 1:50 (Fig. S2). There was higher signal-to-noise for this sample so that R₂^CD2 was also determined from integration in the outermost region of the spectra that is dominated by θ close to 0°, i.e. local bilayer normal parallel to the NMR field. The R₂^CD2(0°)/R₂^CD2(90°) ≈ 1.5 may reflect larger variations in the quadrupolar frequency for 0° vs. 90° for equivalent amplitude acyl chain motion because of the (3 cos²θ − 1) dependence of this frequency.

Table 1.

²H NMR transverse relaxation rates of DMPC-d54 (s⁻¹)^a

	Peptide
T (°C)	No peptide		GPfp		HAfp pH 7		HAfp pH 5
	−C2H2	−C2H3	−C2H2	−C2H3	−C2H2	−C2H3	−C2H2	−C2H3
35	951(6)	313(9)	1147(7)	392(5)	1290(15)	477(13)	1110(11)	407(8)
30	950(9)	334(8)	1096(10)	407(4)	1518(35)	512(34)	1255(12)	413(12)
25	1078(23)	316(9)	1025(20)	418(4)	2116(53)	609(25)	1864(27)	516(8)
20	1878(33)	1009(15)	R2,f = 9480(918) s−1 Af = 0.660(16) R2,s = 1782(59) s−1 As = 0.340(19)	928(32)	1935(40)	1177(29)	R2,f = 14088(374) s−1 Af = 0.965(17) R2,s = 876(163) s−1 As = 0.035(4)	2745(92)
10	2136(22)	1194(20)	7030(127)	2980(80)	2068(15)	1085(26)	4347(131)	2505(106)
0	2994(64)	1548(25)	4078(141)	1736(68)	2772(64)	1710(28)	4286(118)	1409(42)

^aSamples are DMPC-d54 without peptide or with peptide at peptide:DMPC-d54 = 1:25 molar ratio. A relaxation rate is derived from fitting of integrated intensities (I) vs. relaxation time (τ). For −C²H₂, the intensities are sums for symmetric frequency ranges around 0 kHz, with typical values (15 → 25) + (−15 → −25) kHz. For −C²H₃, the intensities are the difference between integration over the central spectral region with −C²H₃ and −C²H₂ contributions and integration over nearby ranges that only contain −C²H₂ contributions, with typical integration frequency ranges of (5 → −5) − [(10 → 15) + (−10 → −15)] kHz. The −C²H₂ rates are typically for the most intense region of the −C²H₂ spectrum that does not overlap with the −C²H₃ signals. Figure S1 lists the frequency ranges for each fitting, fittings for other ranges, and displays spectra at smallest τ. For most data, the R₂ relaxation rate was determined from fitting of ln(I) = ln(I₀) − (R₂ × τ) where ln(I₀) and R₂ are fitting parameters. For the −C²H₂ GPfp and HAfp, pH 5 data at 20 C, the fitting was based I = [A_f × exp(−R_2,f × τ)] + [A_s × exp(−R_2,s × τ)] with f ≡ fast, s ≡ slow, and R_2,f, R_2,s, A_f, and A_s as fitting parameters with A_f + A_s = 1. Each best-fit parameter is followed by the fitting uncertainty in parentheses.

The trends observed for the 35 °C spectra are also generally observed for the 30 and 25 °C spectra, including R₂^CD2/R₂^CD3 ≈ 2.4-3.6 and similar values of R₂^CD2 for −C²H₂ either closer to the headgroup or closer to the methyl group (Table 1 and Fig. S1). At 25 °C, the no peptide and HAfp, pH 7 spectra have a broader background signal that is likely a gel phase, but this signal is subtracted during calculation of I(τ) and the reported R₂’s are for the liquid-crystalline phase in the most intense region (~20 → −20 kHz).

Relative to 25 °C, the no peptide and HAfp, pH 7 spectra are broader at 20 °C and consistent with a gel phase for which the acyl chains are much more ordered than the liquid-crystalline phase (Figs. 3B, C and S1). There is additional broadening when the temperature is reduced to 10 and then 0 °C which correlate with increased ordering (Figs. 3D and S1). At 20 °C, the GPfp and HAfp, pH 5 spectra have some of the horn features of the liquid-crystalline phase and their lineshapes and linewidths are intermediate between those of the 25 and 20 °C spectra of no peptide and HAfp, pH 7 (Fig. 3B). At 10 °C, the GPfp and HAfp, pH 5 spectra are broader and no longer have horn features, and at 0 °C are even broader and resemble the no peptide and HAfp, pH 7 spectra at 10 or 20 °C (Fig. 3C, D).

Fig. 3 displays the effect of increasing τ on spectral intensity, and each panel shows comparison between samples with similar spectral lineshapes and linewidths that reflect similar amplitudes of fast-motion of the acyl chains. Fig. 3A displays the 35 °C data, as discussed above, and Fig. 3B displays the 25 °C data for no peptide and HAfp, pH 7, and 20 °C data for GPfp and HAfp, pH 5. Visual comparison is straightforward between the no peptide and GPfp spectra in Fig. 3B because they were acquired with the same array of τ values. There is much more rapid decay of the intensities of the GPfp vs. no peptide spectra. There is even more striking difference between the very rapid decay for HAfp, pH 5 vs. the much slower decay for HAfp, pH 7 (note that the Δτ is smaller for HAfp, pH 5 vs. 7). Fig. 3C displays 20 °C data for no peptide and HAfp, pH 7, and 10 °C data for GPfp and HAfp, pH 5. Direct comparison can be made between all spectra because they were acquired with the same array of τ values. There are similar slower decays of intensities of no peptide and HAfp, pH 7 vs. more rapid decay of HAfp, pH 5 and even faster decay for GPfp. Similar trends are observed in Fig. 3D which displays comparative 10 °C data for no peptide and HAfp, pH 7, and 0 °C data for GPfp and HAfp, pH 5.

Fig. 4A displays fittings of −C²H₂ intensities measured from the Fig. 3B spectra. For no peptide and HAfp, pH 7 at 25 °C, plots of ln[I^CD2] vs. τ were linear and R₂^CD2 was the best-fit slope. For GPfp and HAfp, pH 5 at 20 °C, plots of ln[I^CD2] vs. τ exhibited more rapid decay and were non-linear, and I^CD2 vs. τ was fitted to a sum of fast and slow exponential decays. Fig. 4B displays fittings of ln[I^CD2] vs. τ based on the Fig. 3C spectra, which show the more rapid decays and larger R₂^CD2’s of GPfp and HAfp, pH 5 vs. no peptide and HAfp, pH 7. Fig. 4C displays fittings of ln[I^CD3] vs. τ based on the Fig. 3B, C spectra. There are more rapid decays and larger R₂^CD3’s of GPfp and HAfp, pH 5 vs. no peptide and HAfp, pH 7. The R₂^CD2 and R₂^CD3 values are presented numerically with uncertainties in Table 1 and displayed visually as bar plots in Fig. 5A, B. The −C²H₂ rates are typically for the most intense region of the −C²H₂ spectrum that does not overlap with the −C²H₃ signals. Fig. S1 displays spectra at shortest τ as well as all integration regions and R₂ values for the four samples. Fig. S2 displays spectra and R₂’s from a second HAfp, pH 7 sample with peptide:lipid = 1:50. For a given temperature and functional group, the R₂’s of the two samples typically agree within 15%.

Figure 4.

Fittings of ²H NMR spectral intensity (I) vs. relaxation time (τ) to determine R₂’s ≡ transverse relaxation rates. Plots and fittings are for: (A) −C₂H₂ data of no peptide and HAfp, pH 7 at 25 °C and of GPfp and HAfp, pH 5 at 20 °C; (B) −C₂H₂ data of no peptide and HAfp, pH 7 at 20 °C and of GPfp and HAfp, pH 5 at 10 °C; and (C) −C₂H₃ data. The data are black squares and are based on integrated intensities from spectra displayed in: (A) Fig. 3B; (B) Fig. 3C; and (C) Fig. 3B and C. The best-fits are displayed in red. The fittings in panels A and B are grouped together because of similar lineshapes and linewidths of the respective Fig. 3B and C spectra. Most data are fitted to ln(I) = ln(I₀) − (R₂ × τ) where ln(I₀) and R₂ ≡ transverse relaxation rate are fitting parameters. For the GPfp and HAfp, pH 5 −C²H₂ data at 20 C, the fitting is based on I = [A_f × exp(−R_2,f × τ)] + [A_s × exp(−R_2,s × τ)] where A_f, R_2,f, A_s, and R_2,s are fitting parameters. Best-fit parameters with uncertainties are also displayed in Table 1, and Fig. S1 has the integration windows that were used to determine the intensities that are fitted.

Figure 5.

Bar plots of temperature-dependent transverse relaxation rates (R₂) and differences in rates (ΔR₂). Panels A and B display R₂’s for −C²H₂ and −C²H₃, respectively, and panels C and D display ΔR₂’s for −C²H₂ and −C²H₃, respectively, where ΔR₂ = R_2,peptide − R_{2,no peptide}. The R₂’s are for DMPC-d54 without peptide and with peptide:DMPC-d54 molar ratio = 1:25. The R₂ values and their uncertainties are numerically presented in Table 1 with additional rate analyses in Figs. S1 and S2. The −C²H₂ R₂’s are typically for the most intense region of the −C²H₂ spectrum that does not overlap with the −C²H₃ signals. For −C²H₂ of GPfp and HAfp, pH 5 at 20 °C, the R₂’s are for the dominant fast components of the bi-exponential decays. At 35, 30, and 25 °C, each ΔR₂ is calculated using R_2,peptide and R_{2,no peptide} at the given temperature, based on predominant liquid-crystalline phase at these temperatures. At 20, 10, and 0 °C, each ΔR₂ for HAfp, pH 7 is also the difference between R₂’s at the given temperature, based on a predominant gel phase and on similar spectral lineshapes and linewidths at each temperature (Figs. 3C, D and S1). At 20, 10, and 0 °C, each ΔR₂ for GPfp and HAfp, pH 5 is calculated using the R_{2,no peptide} at 25, 20, and 10 °C, respectively. This is based on similar spectral lineshapes and linewidths and therefore similar phases and amplitudes of fast-motions of the peptide and no peptide samples (Fig. 3B, C, and D).

Fig. 5C, D displays bar plots of ΔR₂ = R_2,peptide − R_{2,no peptide} for the −CD₂ and −CD₃ groups. The R_2,peptide’s are for peptide:lipid = 1:25. At 35, 30, and 25 °C, each ΔR₂ is calculated using R_2,peptide and R_{2,no peptide} at the given temperature, based on predominant liquid-crystalline phase at these temperatures. At 20, 10, and 0 °C, each ΔR₂ for HAfp, pH 7 is also the difference between R₂’s at the given temperature, based on predominant gel phase and on similar spectral lineshapes and linewidths at each temperature (Figs. 3 and S1). At 20, 10, and 0 °C, each ΔR₂ for GPfp and HAfp, pH 5 is calculated using the R_{2,no peptide} at 25, 20, and 10 °C, respectively. This is based on similar spectral lineshapes and linewidths and therefore similar phases and amplitudes of fast-motions of the peptide and no peptide samples (Fig. 3).

Discussion

1. Summary of experimental results including distinctive temperature dependence of R₂ of GPfp and HAfp, pH 5.

The present study reports the ²H NMR transverse relaxation rates of the acyl chains of DMPC-d54 membrane as functions of bound fp and temperature. At 35, 30, and 25 °C, the spectral lineshapes of all samples are consistent with membrane with a predominant liquid-crystalline phase. At 20, 10, and 0 °C, the lineshapes of the no peptide and HAfp, pH 7 are consistent with a gel phase whereas at 20 °C, those of GPfp and HAfp, pH 5 are intermediate between the two phases and have some of the “horn” features of liquid-crystalline spectra, and at 10 and 0 °C resemble the lineshapes of no peptide and HAfp, pH 7 at 20 °C. As shown in Fig. 5C, D, ΔR₂ = R_2,peptide − R_{2,no peptide} are separated in three groups: (1) for all peptides in liquid-crystalline lipid (35, 30, and 25 °C), the typical ΔR₂ > 0 and increases with lower temperature with values up to ~10³ Hz for −CD₂ and ~3 × 10² Hz for −CD₃; (2) for HAfp, pH 7 in a gel phase (20, 10, and 0 °C), ΔR₂ ≈ 0; and (3) for GPfp and HAfp, pH 5 at 20, 10, and 0 °C, ΔR₂ > 0, and exhibits a large increase and then decrease as temperature is lowered with maximum ΔR₂ ≈ 10⁴ Hz for −CD₂ and 2 × 10³ Hz for −CD₃.

Earlier studies have shown that acyl chain ²H’s of pure DMPC and other lipids exhibit an increase and decrease in R₂ in a ~5 °C interval near the liquid-crystalline to rippled-gel phase transition.^53,54 This was not observed in our no peptide data, probably because the temperature was changed in larger increments. Earlier studies of deuterated DMPC or other lipid + peptide/protein have typically shown a smaller peak in ²H R₂ near the phase-transition temperature.^55-57 Examples of the peptide/protein include gramicidin, a bacterial ion channel, and lung surfactant proteins that function to reduce the surface tension at the air-water interface in lung alveoli. By contrast, for the present study, GPfp and HAfp, pH 5 exhibit large R₂’s at 20, 10, and 0 °C with highest value typically at 20 or 10 °C (Fig. 5A, B). HAfp, pH 7 shows a small increase in R₂ in the 20 to 0 °C range, similar to no peptide. The GPfp sample was prepared at neutral pH at which GPfp has been shown to induce vesicle fusion, whereas HAfp induces greater fusion at pH 5 vs. 7.

2. ΔR₂ > 0 in the liquid-crystalline phase is consistent with longer correlation time with peptide.

For the liquid-crystalline phase, the larger ²H R₂’s with bound fp correlate with earlier observation at 35 °C of ~2× larger ³¹P R₂ of membrane with bound GPfp.⁵⁸ The ²H R₂ has contributions from the motional spectral densities (J) at 0, ω₀, and 2ω₀ frequencies, where ω₀ is 2π times the ²H NMR Larmor frequency = 61 MHz.²² The J(ω₀) and J(2ω₀) contributions are small based on previously-reported ²H NMR longitudinal relaxation rates (R₁’s) of similar samples at 35 °C.⁵⁸ The R₁’s are ~100× smaller than the respective R₂’s and are a reasonable estimate of the J(ω₀) and J(2ω₀) contributions to R₂.^22,23 A reasonable estimate of the major J(0) contribution to R₂ for a specific type of motion is ΔM₂ × τ_c where ΔM₂ is the mean-squared amplitude of the NMR frequency fluctuations associated with the motion and τ_c ≈ J(0)/2 is the correlation time of the motion, e.g. for angular motion, the typical time to move 1 radian.^23,59 This R₂ estimate is valid for “fast motion” for which ΔM₂ × τ_c² << 1. For deuterated lipid acyl chains, the ²H ΔM₂ ≈ (4 × 10¹⁰s⁻²) × (1 − S²) where S is the order parameter for the motion.²² Smaller S correlates with larger-amplitude motion and therefore larger ΔM₂, i.e. greater fluctuations in NMR frequency experienced by the ²H nuclei. This ΔM₂ estimation is based on large-jump motion and for continuous small jumps, ΔM₂ may be smaller for a given order parameter.⁵⁹ For the temperature range of the present study, the literature values of τ_c for typical lipid chain motions are <10⁻⁶ s so that the fast motion condition is valid.^54,60

For the liquid-crystalline phase, acyl chain fluctuations perpendicular to the bilayer normal have the largest τ_c and are likely the biggest contributor to J(0) and R₂.⁶⁰ For liquid-crystalline DMPC-d54, the typical S ≈ 0.2 for −CD₂ groups so that ΔM₂ ≈ 4 × 10¹⁰s⁻².²⁷ Our typical experimental R₂ ≈ 10³ Hz in this phase, so the calculated τ_c ≈ 2 × 10⁻⁸s⁻¹, which matches literature values of τ_c for chain fluctuations and supports these fluctuations as the dominant motion underlying R₂.⁶⁰ For any fp sample in this phase, the typical ΔR₂ > 0 (Fig. 5C, D). Relative to without peptide, there is a literature report of ~3× larger τ_c with 6 mole% integral membrane peptide that probably reflects peptide-associated disruption of large-amplitude chain motion.⁵⁴ There is probably similar increase in τ_c for fp samples and this increase is likely the basis for the ΔR₂ > 0. The typically larger ΔR₂ for HAfp vs. GPfp may be due to smaller separation between HAfp monomers vs. GPfp oligomers and consequent greater probability of close contact between lipid and peptide. This difference is described more quantitatively by comparison of the typical lipid diffusion distance ≡ d_L vs. average separation between peptides ≡ d_HAfp or d_GPfp. The d_L ≈ 60 Å based on d_L ≈ (4 × D × τ_D)^½ with lateral diffusion constant D ≈ 10⁸ Ų-s⁻¹ and diffusion time τ_D ≈ 10⁻⁵ s, chosen to be short enough for spectral averaging.⁶¹ For a vesicle with ~100 nm diameter, the surface area ≈ 3 × 10⁶ Ų, N_lipid ≈ 6 × 10⁴, and N_fp ≈ 2 × 10³. The d_HAfp ≈ 40 Å and d_GPfp ≈ 120 Å (based on ~10 GPfp-per-oligomer) so that d_L < d_HAfp and d_L > d_GPfp, i.e. more lipids will contact HAfp monomers vs. GPfp oligomers.²¹ For the HAfp, pH 7 sample at 1:50 ratio, d_HAfp ≈ 55 Å so that d_L < d_HAfp still holds. This correlates with R₂’s for the 1:50 sample that are more similar to 1:25 HAfp than 1:25 GPfp (Figs. S1 and S2).

3. ΔR₂ ≈ 0 for HAfp, pH 7 in gel phases is consistent with shorter correlation time with peptide.

At 20 °C, the spectral lineshape of DMPC-d54 without peptide is much broader than at higher temperature and is consistent with transition to a L_β′ rippled-gel phase for which there is higher ordering of the acyl chains and reduced lateral diffusion rate.⁵² The R₂^20C/R₂^25C is ~2 for −C²H₂ and ~3 for −C²H₃. There is greater spectral broadening at 10 °C, and only small increase in R₂’s, with R₂^10C/R₂^20C of ~1.1. There is further broadening at 0 °C, consistent with the transition to the L_β phase for which the chains are fully-ordered in a trans conformation. There is also an increase in R₂’s, with R₂^0C/R₂^10C ≈ 1.4. The increases in R₂ are semi-quantitatively understood by the countervailing changes in ΔM₂ and τ_c. For the gel phases, a typical S^CD2 ≈ 0.7 so that ΔM₂ ≈ 2 × 10¹⁰s⁻², with increases in S and decreases in ΔM₂ as temperature is lowered.²² The effect on R₂ of ~2× smaller ΔM₂ in a gel vs. liquid-crystalline phase is more-than-compensated by increases in τ_c, including ~5× and ~3× increases with transitions to rippled-gel and gel phases, respectively.⁶⁰

At 20, 10, and 0 °C, the spectral lineshapes with HAfp, pH 7 are similar to those of no peptide at the same temperature and consistent with gel phases. Interestingly, for HAfp, pH 7, R₂^20C ≈ R₂^25C and ΔR₂ ≈ 0 at 20, 10, and 0 °C (Fig. 5). These observations are semi-quantitatively explained if HAfp, pH 7 follows an earlier result for a transmembrane peptide of ~2× smaller τ_c with vs. without peptide in gel phases that likely reflects peptide disruption of chain ordering.⁵⁴ This behavior contrasts with the earlier-described ~3× larger τ_c with peptide in the liquid-crystalline phase.

4. Very large ΔR₂ ≈ 10⁴ Hz for GPfp and HAfp, pH 5 at 20 °C is consistent with larger-amplitude motion for lipid closer to vs. further from peptide.

In our view, the most interesting experimental result of our study is the large R₂’s for GPfp and HAfp, pH 5 over the 20 - 0 °C range with maximum ΔR₂^CD2 ≈ 10⁴ Hz at 20 °C and maximum ΔR₂^CD3 ≈ 2 × 10³ Hz at 20 or 10 °C (Fig. 5 C, D). The large increases and then decreases in R₂’s as temperature is lowered through the phase transitions are not explainable with only consideration of monotonic increases in τ_c as temperature is lowered. We provide a more quantitative interpretation of the R₂ behavior based on a model for the large ²H R₂’s near the phase separation temperature of lipid mixtures as well as the results of several molecular dynamics simulations. For the mixtures, lateral lipid diffusion between nm-size domains with different acyl chain order parameters is the proposed motion that results in large R₂’s in these samples.^25,62 The ΔM₂ for this motion is approximately the difference in mean-squared NMR frequency fluctuations between the different domains and the τ_c is the characteristic time for a lipid to diffuse between the domains. The same model is applied to the present study and the “domains” are lipids closer to vs. more distant from the peptide. There is ~6 Å dimension of the peptide-proximal domain which is much smaller than the ~20 – 200 Å dimension of the lipid domains of the earlier study.

Molecular dynamics simulations from multiple groups have shown significantly larger acyl chain motion for lipids close to (<6 Å) to vs. more distant from monomeric HAfp, with −CD₂ S^closer ≈ 0.1 vs. S^further ≈ 0.2 in the liquid-crystalline phase.^47-49 The larger motion results in higher probability of protrusion of acyl chains into the aqueous phase for lipids close to HAfp which may be relevant for reducing barriers between membrane fusion intermediates (Fig. 1F). There is a ΔM₂ ≈ (4 × 10¹⁰ s⁻²) × (S^2,further − S^2,closer) ≈ 10⁹ s⁻² and the diffusion τ_D ≈ τ_c ≈ (6 Å)²/4D with D ≈ 10⁸ Ų/s so that τ_c ≈ 10⁻⁷ s, and the calculated ΔR₂^CD2 ≈ ΔM₂ × τ_c ≈ 10² Hz. This could contribute to the experimental ΔR₂^CD2 of GPfp and HAfp, pH 5 which are typically in the range of 2-8 × 10² Hz.

We propose that this phenomenon also exists at 20 °C and is the basis for the large ΔR₂’s. The lineshapes for GPfp and HAfp, pH 5 are intermediate between those of liquid-crystalline and rippled-gel phases, and we estimate S^closer ≈ 0.2 vs. S^further ≈ 0.5 so that ΔM₂ ≈ 10¹⁰ s⁻². The D ≈ 10⁷ Ų/s so that for 6 Å diffusion length, τ_c ≈ 10⁻⁶ s and ΔR₂^CD2 ≈ 10⁴ Hz, which is in semi-quantitative agreement with the experimental ΔR₂’s.⁶¹ Because 1/τ_c < spectral width, there is spectral averaging between the two lipid locations and S values, which is consistent with the observed single lineshapes at 20 °C.

At 10 °C, the experimental ΔR₂^CD2’s are smaller than at 20 °C but still substantial (up to 5 × 10³ Hz). The lateral diffusion τ_c are larger at 10 °C, so within the present model, the ΔM₂ would have to be smaller at 10 vs. 20 °C. There are similar lineshapes and linewidths of the spectra of the 10 °C peptide and the 20 °C no peptide samples where the latter is in the rippled-gel phase which has larger domains of lipids with more- and less-ordered acyl chains.⁵² If the 10 °C peptide sample also adopts this phase, these larger domains may diminish the distinction between lipids closer to vs. further from the peptide so that ΔM₂ is reduced. At 0 °C, the experimental ΔR₂^CD2 ≈ 10³ Hz which is a continuation of the trend of smaller ΔR₂ with temperature reduction in the gel phases.

5. Similar trends and smaller ΔR₂ for −CD₃ vs. −CD₂ support scaling of ΔM₂ and the diffusion model.

The ΔR₂^CD3’s of the peptide samples follow the same trends as the ΔR₂^CD2’s but with smaller values. The ΔR₂^CD3’s increase as temperature is lowered in the liquid-crystalline phase with largest ΔR₂^CD3 ≈ 3 × 10² Hz at 25 °C. For a peptide sample at a given temperature, the typical ΔR₂^CD3/ΔR₂^CD2 ≈ 1/3 which matches the no peptide R₂^CD3/R₂^CD2 ≈ 1/3 and is a likely consequence of scaling down of the relevant ΔM₂ for acyl chain fluctuations for −CD₃ by rapid rotation about the symmetry axis. For GPfp and HAfp, pH 5, the −CD₃ ΔR₂ also exhibit the increase-then-decrease trend as temperature is further lowered and the lipid adopts gel phases. The maximum ΔR₂^CD3 ≈ 2 × 10³ Hz is ~1/5 the maximum ΔR₂^CD2. For GPfp, the largest ΔR₂ is at 10 vs. 20 °C for −CD₃ vs. - CD₂. The −CD₃ and −CD₂ should experience the same lateral diffusion τ_c, but the −CD₃ has much smaller order parameter so that lower temperature may be needed for appreciable difference in (1 − S²) and ΔM₂ for lipid closer to vs. further from peptide.

6. Much larger ΔR₂ for β sheet GPfp and helical hairpin HAfp, pH 5 correlates with greater hydrophobicity and fusion.

For HAfp, pH 7 at 20, 10, and 0 °C, the ΔR₂ ≈ 0 for −CD₂ and −CD₃ groups, which is very different from GPfp and HAfp, pH 5. HAfp, pH 7 also exhibits spectral lineshapes and order parameters that are comparable to or a little broader/larger than no peptide, whereas GPfp and HAfp, pH 5 exhibit lineshapes and order parameters that are smaller than no peptide and HAfp, pH 7 (Fig. 3). Earlier NMR shows that HAfp is mixture of different helical hairpin structures at pH 5 and 7 (Fig. 2B). At pH 5, NMR data support a greater fraction of more open/”semi-closed” structure that has greater hydrophobic surface area that may be a better solvent for lipid acyl chains and allow greater probability for lipid acyl chain location outside the hydrophobic core of the membrane (Fig. 1F).¹⁸ By contrast, at pH 7, HAfp has smaller hydrophobic surface area and the chain motion may be similar for lipids closer to vs. further from HAfp, which is consistent with pH 7 order parameters similar to pure lipid.²⁷ This similarity would result in ΔM₂ ≈ 0 for diffusion, like pure lipid, so that ΔR₂ ≈ 0.

There is correlation between the small effects on lipid ²H NMR lineshape and R₂ of HAfp at pH 7 vs. much larger effects at pH 5 and the difference in the appearance of a vesicle suspension after addition of HAfp at pH 7 vs. 5. The vesicles are ~100 nm diameter and the suspension at either pH 7 or 5 is transparent prior to addition. After HAfp addition at pH 7, the suspension remains transparent. By contrast, addition at pH 5 results in a cloudy/turbid suspension that is the likely result of HAfp-induced vesicle fusion and consequent larger vesicles, with increased light scattering when the vesicle dimension is greater than ~300 nm.⁶³ This observation is consistent with pH dependence of vesicle fusion by fluorescence measurement and with large pH dependence of hemolysis of erythrocytes after addition of HAfp.^18,64 Cloudiness/turbidity is also apparent in vesicle suspensions after addition of GPfp and electron micrographs from earlier studies also show larger dimension vesicles after addition of GPfp.^38,65,66 There is correlation between the large ΔR₂’s for GPfp and HAfp, pH 5 and their fusogenicity that we ascribe to increased protrusion of lipid acyl chains into the aqueous phase. This protrusion matches the chain location for transition states between membrane fusion intermediates (Fig. 1F).

The similar spectral lineshapes of the no peptide and HAfp, pH 7 samples raises the possibility that HAfp doesn’t bind to lipid. However, there are also two lines of evidence that support near-quantitative binding of HAfp at pH 7 for our samples. First, among the peptides, HAfp, pH 7 exhibits the largest ΔR₂’s in the liquid-crystalline phase (Fig. 5C, D). Second, A₂₈₀ < 0.01 in the supernatant after centrifugation of the suspension containing lipid and HAfp. This second observation is independently supported by calculated [HAfp_bound]/[HAfp_free] ≈ K_eq × [lipid] ≈ 50, based on K_eq ≈ 1 × 10⁴ M⁻¹, the previously-reported binding constant of HAfp with lipid at pH 7, and [lipid] ≈ 5 × 10⁻³ M during our sample preparation.⁶⁴

There is also the possibility that the different lineshapes and ΔR₂’s for HAfp, pH 7 vs. 5 are a consequence of different HAfp locations in the membrane, e.g. interfacial vs. deeply-inserted or transmembrane. In our view, the membrane location(s) of HAfp remains an unresolved question. There are experimental data that support both interfacial and deep membrane location, and different molecular dynamics simulations also find different “equilibrium” locations that vary between interfacial and transmembrane.^{15,44,47,48,67-69} These differences may be a consequence of different initial conditions for HAfp structure and location. To our knowledge, only small changes in membrane location have been observed at pH 5 vs. 7.^39,40 We therefore think that the large pH dependence of lipid/HAfp interaction is more likely due to a change in HAfp structural distribution and hydrophobicity, as detailed above.

7. Slow-relaxing fraction at 20 °C may be lipids that don’t contact GPfp.

Other than the −CD₂ signals of GPfp and HAfp, pH 5 at 20 °C, the other data are all well-fitted by a single exponential decay and one R₂. This likely means that all analyzed nuclei experience the same motions that dominate R₂. These motions have the largest τ_c and appreciable ΔM₂ with a contribution to R₂ ≈ ΔM₂ × τ_c. As discussed above, for pure lipid, the dominant motion is likely acyl chain fluctuations perpendicular to the bilayer normal. This is also the likely motion for all the peptide samples in liquid-crystalline phase at 35, 30, and 25 °C. There are similar R₂’s for no peptide and HAfp, pH 7 in the gel phases at 20, 10, and 0 °C, which supports similar dominant motion for the two samples (Fig. 5A, B). At a given temperature, there are also similar spectral lineshapes for these two samples.

Fitting with a sum of fast (f) and slow (s) exponential decays was done for −CD₂ analyses of GPfp and HAfp, pH 5 samples at 20 °C. Fitting parameters were populations A_f and A_s and rates R_2,f and R_2,s. These R_2,f’s had the largest ΔR₂ ≈ 10⁴ Hz of all fittings, and as discussed in #4 above, we ascribe the ΔR₂ to lipid lateral diffusion between locations closer to vs. further from the peptide, with different order parameters for the two locations.

We first consider the GPfp analysis for which the fast:slow population ratio A_f/A_s ≈ 2 and the R_2,s is similar to the R₂ of no peptide and HAfp, pH 7 at 20 °C, i.e. ΔR_2,s ≈ 0. This suggests that the slow fraction is lipids that don’t contact GPfp. This hypothesis is supported by considering the model from #2 above with a vesicle with ~100 nm diameter, surface area ≈ 3 × 10⁶ Ų, N_lipid ≈ 6 × 10⁴, and N_GPfp ≈ 2 × 10³. For GPfp oligomers with ~10 peptides and ~16 residues in β sheet structure, the area-per-oligomer is ~(50 Å)² and there is ~120 Å average distance between oligomers.¹⁴ For lateral diffusion constant D ≈ 10⁷ Ų/s and ~10⁻⁵ s diffusion time consistent with spectral averaging, there is ~20 Å diffusion distance by a lipid. There is a difference of ~30 Å between the oligomer separation (~120 Å) and the sum of the oligomer and lipid diffusion dimensions (~90 Å). This difference means there are regions of the vesicle containing lipids that don’t contact GPfp during the diffusion time. Although our ΔR₂ model incorporates molecular dynamics results that show smaller order parameters for lipids closer vs. further from peptide, there are similar experimental lineshapes for the fast- and slow-relaxing lipid populations which is consistent with fast-relaxing lipids spending relatively small fractional time next to a GPfp oligomer.

The above model is also applicable to the biexponential decay fitting of the HAfp, pH 5 data at 20 °C and explains the much smaller slow population for HAfp, pH 5 with A_f/A_s ≈ 28. HAfp is monomeric and the peptides are separated by ~40 Å, which is approximately the lipid diffusion dimension. There are negligible regions of the vesicle with lipids that don’t contact HAfp and have the associated R_2,s.

Unlike −CD₂, the −CD₃ data for GPfp at 20 °C are well-fitted by a single exponential decay and the R₂^CD3 is intermediate between the R₂^CD3’s at 25 and 20 °C for either no peptide or HAfp, pH 7. As discussed above in #5, the −CD₃ and −CD₂ should experience the same diffusion τ_c, but the −CD₃ has much smaller order parameter. At 20 °C, there may a fairly small difference in (1 − S²) for −CD₃ for lipid closer to vs. further from peptide so that ΔM₂^CD3 and ΔR₂^CD3 are smaller which results in similar R₂^CD3’s for lipids that contact and don’t’ contact GPfp. The −CD₃ data for HAfp, pH 5 at 20 °C are also well-fitted by a single exponential decay and the ΔR₂^CD3 ≈ 2 × 10³ Hz. The lack of a second slower decay for −CD₃ is likely due to a combination of small A_s, smaller difference between R_2,f and R_2,s and smaller signals for −CD₃ vs. −CD₂.

At 10 and 0 °C, the GPfp and HAfp, pH 5 −CD₂ and −CD₃ data are well-fitted as single exponential decays, as evidenced by typical (δR₂/R₂) ≈ 0.03 where δR₂ is the fitting uncertainty. The typical ΔR₂ > 0 but is significantly smaller than the ~10⁴ Hz values for −CD₂ fitting of GPfp and HAfp, pH 5 data at 20 °C. We note that even for the −CD₂ GPfp data at 20 °C with R_2,f/R_2,s ≈ 5, the Fig. 4A ln(I^CD2) vs. time plot is reasonably linear, except for the point at shortest time. If there is a second slower decay in the 10 and 0 °C data, the R_2,f/R_2,s ratio would be smaller than at 20 °C and non-linearity likely wouldn’t be apparent in the single decay fitting.

8. Lipid ordering with fp with EPR and fluorescence may reflect fp-lipid label interaction.

The interpretation of the large ΔR₂ for GPfp and HAfp, pH 5 in terms of large-amplitude acyl chain motions is consistent with smaller average lipid order parameters relative to no peptide and HAfp, pH 7 from ²H NMR spectra, X-ray scattering, and molecular dynamics simulations but not with larger order parameters from EPR and fluorescence spectra.^{19,27,41,47-50,70,71} These latter experiments share common features including membrane with a small fraction of lipid with an organic ring spin- or fluorescence- label at a particular site, and measurement of the EPR lineshape or fluorescence anisotropy of the label. Increased anisotropy for the label for membrane with fp vs. without fp is considered representative of increased order for all membrane lipids. There may be a specific interaction of fusion peptide with the EPR and fluorescence label which results in ordering of the label. This hypothesis is consistent with the observation that when [fusion peptide] ≈ [spin-labeled lipid], the ordering of the label is already close to its maximal value. By contrast, the simulation, NMR, and X-ray results were based on data that probe all of the lipids in the sample. For most of the above studies, the lipid was in the liquid-crystalline phase and phosphatidylcholine was the major or only lipid headgroup, so the difference in disordering vs. ordering between techniques is likely not due to lipid differences. In addition, the X-ray studies detected disordering with GPfp for a variety of lipid compositions that included a single phosphatidylcholine lipid with or without cholesterol as well a mixture that approximately matches the lipid composition of the plasma membrane of host cells of HIV. In most studies, lipid disordering or ordering was also correlated with fusion by comparison between different peptide:lipid ratios, and/or comparison between wild-type and mutant peptide or between conditions that affect peptide-induced vesicle fusion, e.g. pH 5 vs. 7 for HAfp.

We also note conflicting results between analyses of X-ray vs. neutron scattering data about other effects on membrane by GPfp. For both approaches, the GPfp had significant fraction oligomeric β sheet structure. Analysis of the low-angle X-ray scattering showed ~3 Å thinner membrane width in the presence of GPfp whereas analysis of small-angle neutron scattering showed ~3 Å thicker width.^50,72 Analysis of the X-ray data showed typical 5-10 fold reduction in the bending modulus of membrane when GPfp was bound whereas analysis of neutron spin-echo spectra showed typical 1.6 fold increase in the bending modulus.^72,73 The X-ray vs. neutron data are respectively consistent with membrane with larger vs. smaller amplitudes of lipid motion. The reasons for these conflicting results are not understood but we note there was some difference in the GPfp sequence used for the X-ray vs. neutron experiments.

9. Detailed fitting gives precise R₂^CD3, R₂^CD2,outer, and R₂^CD2,inner’s, and R₂^outer > R₂^inner is consistent with τ_c^outer > τ_c^inner for chain fluctuations.

For the present study, separate fittings are required for the I^CD2 and I^CD3 intensities because: (1) I^CD2 is dominant at shorter times, based on 24 −C²H₂ vs. 6 −C²H₃; and (2) I^CD3 is dominant at longer times, based on R₂^CD3 < R₂^CD2 (Fig. 3). The spectral integrations include I^central from the central region that has both −C²H₂ and −C²H₃ contributions, and I^outer and I^inner from outlying regions that are only −C²H₂ (Fig. S1). Typically, I^outer is weighted towards −C²H₂’s closer to the headgroup with larger S, and I^inner is weighted towards −C²H₂’s closer to the chain termini with smaller S. Each R₂^CD3 is well-fitted using I^central − I^inner, with typical δR₂^CD3/R₂^CD3 ≈ 0.02 (Table 1 and Fig. 4C).

The −CD₂ R₂^outer and R₂^inner are determined from separate fittings of I^outer and I^inner (Figs. S1 and S3). Between 35 and 10 °C, no peptide and HAfp, pH 7 exhibit typical R₂^outer:R₂^inner ≈ 1.1 and 1.2, respectively. As discussed above in #2 the R₂ ≈ ΔM₂ × τ_c for these samples and the R₂^outer > R₂^inner is reasonably explained by τ_c^outer > τ_c^inner which is consistent with lower mobility of −C²H₂ closer to the headgroup vs. chain termini. The larger R₂^outer:R₂^inner for HAfp, pH 7 vs. no peptide could be due to HAfp location near the headgroups. At 0 °C, there is significant broadening of the −C²H₃ contribution so that the inner and outer −C²H₂ ranges are moved to regions corresponding to −C²H bond orientations respectively closer to perpendicular and parallel to the external magnetic field direction. This may be related to R₂^inner > R₂^outer at 0 °C, with ratio of ~1.1. Interestingly, the GPfp and HAfp, pH 5 typically have similar R₂^outer and R₂^inner at a given temperature but don’t exhibit clear trends like no peptide and HAfp, pH 7 (Fig. S3). For −C²H₂’s nearer the headgroup vs. chain termini, there may be different temperature dependences of S^closer and S^further which results in different temperature dependences of ΔR₂^outer and ΔR₂^inner.

Summary

In this study, ²H NMR R₂’s were determined for DMPC-d54 with bound GPfp and HAfp fusion peptides that adopt very different structures, intermolecular antiparallel β sheet and monomer helical hairpin, respectively. For the liquid-crystalline phase between 35 and 25 °C, the typical R_2,peptide > R_{2,no peptide}, which is consistent with larger correlation time τ_c for acyl chain fluctuations with peptide. For no peptide in the gel phase at 20 °C, the R₂’s are ~2× times larger than in liquid-crystalline phase, and there is ~1.5× further increase when temperature is reduced to 0 °C, which correlates with increases in τ_c in the gel phases. At 20 °C, there are similar R₂’s and spectral lineshapes for no peptide and HAfp, pH 7, and also similar R₂’s at 10 and 0 °C, which correlates with earlier observations of similar τ_c’s without vs. with peptide in gel phases. GPfp and HAfp, pH 5 exhibit very different R₂’s and lineshapes at 20, 10, and 0 °C. The ΔR₂^CD2 ≈ 10⁴ Hz at 20 °C, with smaller but still substantial ΔR₂’s at 10 and 0 °C. At a given temperature, the lineshapes are also narrower than those of no peptide and HAfp, pH 7. The large ΔR₂’s for GPfp and HAfp, pH 5 are interpreted to be due to lipid diffusion between locations closer to vs. further from peptide, with respective smaller vs. larger order parameters. This difference in order parameters was previously detected in molecular dynamics simulations of membrane with HAfp. Larger-amplitude chain motion, including protrusion into the aqueous phase, is likely characteristic of transition states between membrane fusion intermediates, and such motion is supported by the present study as a contribution of fusion peptides to fusion catalysis. This contribution exists for both helical and β sheet fusion peptide structures.

Abbreviations

d

distance

D

diffusion constant

DMPC-d54

dimyristoylphosphatidylcholine with perdeuterated acyl chains

DOPC

dioleoylphosphatidylcholine

DPPC-d62

dipalmitoylphosphatidylcholine with perdeuterated acyl chains

Ed

ectodomain

f

fast

fp

fusion peptide

FsSu

fusion subunit

GP

HIV glycoprotein

HA

influenza hemagglutinin

I

intensity

LUV

large unilamellar vesicle

M₂

spectral second moment

n

carbon number in acyl chain

R₁

NMR longitudinal relaxation rate

R₂

NMR transverse relaxation rate

RbSu

receptor-binding subunit

s

slow

S

order parameter

SARS-CoV-2

severe acute respiratory syndrome coronavirus 2

δ

uncertainty

ΔM₂

mean-squared amplitude of the NMR frequency fluctuations

θ

angle with respect to NMR field direction

ν

frequency

τ

NMR time delay

τ_c

correlation time

τ_D

diffusion time