The HIV Fusion Peptide and its Cross-Linked Oligomers: Efficient Syntheses, Significance of the Trimer in Fusion Activity, Correlation of β Strand Conformation with Membrane Cholesterol, and Proximity to Lipid Headgroups

Abstract

For enveloped viruses such as HIV, a ~20-residue N-terminal fusion peptide domain in the envelope protein binds to target cell membranes and plays a key role in fusion between the viral and cellular membranes during infection. The chemically synthesized HIV fusion peptide or “HFP” catalyzes fusion between membrane vesicles and is a useful model system to understand some aspects of HIV fusion. Previous studies have shown a common trimeric state for envelope protein from several different viruses including HIV and in this study, practical high-yield syntheses are reported for HFP monomer (HFPmn) and chemically cross-linked HFP dimer (HFPdm), trimer (HFPtr), and tetramer (HFPte). The vesicle fusion rates per strand were ordered HFPmn < HFPdm < HFPtr ≈ HFPte and suggested that HFPtr is the smallest catalytically efficient oligomer. Solid-state NMR measurements of ¹³CO chemical shifts were carried out in constructs labeled at either Ala-6 or Ala-15. For all constructs associated with cholesterol-containing membranes, the chemical shifts of both residues correlated with β strand conformation while association with membranes without cholesterol resulted in a mixture of helical and β strand conformations. The dependence of fusion rate on oligomer size is independent of membrane cholesterol content, so one interpretation of the data is fusion activity of both helical and β strand conformations. Membrane location may be a determinant of fusion activity and for all constructs in both conformations, a large fraction of the Ala-15 ¹³COs were 5–6 Å from the ³¹Ps in the lipid headgroups, while the Ala-6 ¹³COs were more distant.

Infection by the human immunodeficiency virus (HIV) begins with fusion between the viral and host cell membranes and leads to deposition of the viral nucleocapsid in the cytoplasm and viral replication (1, 2). The fusion process is mediated by the HIV gp41 fusion protein which contains a region inside HIV as well as a single-pass transmembrane domain. The ~170-residue ectodomain of gp41 lies outside HIV and is subdivided into a more C-terminal “soluble ectodomain” and a ~20-residue N-terminal fusion peptide (HFP) which is hydrophobic and fairly conserved. Peptides with the HFP sequence catalyze fusion between membrane vesicles and there is good correlation between the mutation/fusion activity relationships of HFP-induced vesicle fusion and gp41-induced membrane fusion (3–8). These data suggest that studies of HFP will provide information about HIV/target cell fusion.

High-resolution structural studies have been carried out on the “soluble ectodomain” of gp41 which begins ~10 residues C-terminal of the HFP. These structures revealed trimeric oligomerization with the three N-termini in close proximity (1, 2, 9–11). Trimeric oligomerization has also been observed in structures of fusion proteins from other “class I” enveloped viruses including influenza virus (1, 2). These results have motivated study of oligomeric viral fusion peptides including a HFP construct formed by cross-linking three HFPs at their C-termini to form a cross-linked HFP trimer (HFPtr) (12–14). The HFPtr construct induced vesicle fusion with a rate that was up to 40 times higher than the rate induced by non-cross-linked HFP monomer (HFPmn) (13). The rates were compared at constant peptide strand:lipid mol ratio, i.e. three times less HFPtr than HFPmn, and suggested that the oligomeric topology enforced by C-terminal cross-linking is functionally important. Outstanding questions include: (1) Does the fusion rate per strand continue to increase for oligomers larger than the trimer? (2) Is there a structural basis for the increased fusion rate of cross-linked oligomers? This paper reports progress in answering both questions.

One experimental challenge in the study of cross-linked HFP oligomers has been production of large quantities of pure material. One synthetic route to HFPtr was cross-linking between HFPmn with one non-native C-terminal cysteine and HFPmn with two non-native C-terminal cysteines but the major product was the HFP dimer (HFPdm) formed from cross-linking between the HFPmn with one cysteine (12). An alternative approach was initial formation of a peptide scaffold with three lysines and amide bonds between the ε-NH₂ of the first lysine and the COOH of the second lysine and between the ε-NH₂ of the second lysine and the COOH of the third lysine (13). HFPtr was then synthesized using standard 9-fluorenylmethoxycarbonyl (Fmoc) chemistry and synchronous growth of the peptide chains from the three main chain α-NH₂. This approach was therefore direct synthesis of a ~90-mer and the consequent yield was at best 5%. In addition, the final product contained significant impurity HFPtr which was non-separable and which contained one additional lysine in one of the chains. This impurity was a consequence of undesired intramolecular removal of the Fmoc protecting group on a Lys α-NH₂ by a free ε-NH₂ (15).

The present paper provides greatly improved approaches to syntheses of HFPmn, HFPdm, and HFPtr and a synthesis of the new HFP tetramer (HFPte). The residues in HFPmn and HFPdm which couple poorly to subsequent residues are identified and conditions are described to improve (yields. A much purer and higher yielding HFPtr synthesis is presented based on cross-linking HFPmn and HFPdm which each contain a single cysteine. The paper also describes efficient chromatographic separation of HFPmn, HFPdm, HFPtr, and HFPte, and quantitative comparison of the rates of vesicle fusion induced by the four constructs.

A separate section of the paper describes comparative solid-state nuclear magnetic resonance (NMR) structural studies of membrane-associated HFPmn, HFPdm, and HFPtr. In particular, the local conformations at the Ala-6 and Ala-15 residues were probed as well as the proximity of these residues to the lipid headgroups. These studies were motivated in part by hypotheses that the conformation and membrane location of HFP are significant structural factors which impact fusion catalysis and which might therefore explain the increased fusion rates of cross-linked HFP oligomers (4, 16).

The HFP structural literature is complex and includes liquid-state NMR studies in detergent micelles. For HFP:detergent mol ratios ≤ 0.01, there is agreement that the residues between Ile-4 and Leu-12 form an uninterrupted α helix, and one study reports that the helix extends to Met-19 (17–21). A variety of biophysical techniques have been used to study the conformation of HFPmn associated with membranes (5, 22). Infrared spectra suggested that for HFPmn:lipid = 0.005 in membranes without cholesterol, a significant HFPmn fraction adopted helical structure while non-helical structure was favored at higher ratios (23). In addition, helical structure was favored with negatively charged lipids while β strand structure was preferred with neutral lipids or with negatively charged lipids with bound Ca²⁺ (4, 16). For membranes which have the approximate lipid headgroup and cholesterol composition of host cells of HIV, solid-state NMR studies have shown that HFPmn adopts uninterrupted β strand conformation from Val-2 to Gly-16 with more disordered structure at Ala-21 (24). Small β sheet aggregates of HFPmn are formed under these conditions with predominant antiparallel alignment of adjacent peptides and adjacent strand crossing near Phe-8 and Leu-9. For this arrangement, most of the apolar 16-residue N-terminal region of HFPmn will form interpeptide hydrogen bonds which is biophysically reasonable because these residues are likely located in the membrane interior where there is little water available for hydrogen bonding. There is much less information about the structure of HFPdm and HFPtr. In membranes without cholesterol, predominant helical conformation was observed in HFPtr between Leu-7 and Phe-11 while in host-cell-like membranes, β strand conformation was observed for HFPdm at Phe-8 and for HFPtr at Leu-7 and Phe-8 (12–14).

There is not yet a consensus for the location of HFPmn in either micelles or membranes. Data from liquid-state NMR experiments have supported either partial micelle insertion or micelle traversal with Ala-15 and Gly-16 at the micelle-solution boundary (17–19). Simulations of membrane-associated HFPmn have also been consistent with either partial membrane insertion or membrane traversal (25, 26). Fluorescence studies on membrane-associated HFP with F8W mutation were consistent with an ~10 Å distance between the tryptophan indole group and the phosphorus longitude (27, 28). Electron spin resonance experiments have indicated that Met-19 is close to the membrane-water interface while Ala-1 is away from this interface (29).

Solid-state NMR has been applied to determine distances between backbone ¹³CO nuclei in HFPmn and ³¹P nuclei in the lipid headgroups (30). Because there was very little information about the membrane location of HFPmn, the initial experiments scanned the HFPmn backbone using samples with ¹³CO labeling at three sequential residues. A ¹³CO-³¹P distance of 5–6 Å was observed for labeling between Ala-14 and Gly-16 while distances >8 Å were observed for labeling between residues Gly-5 and Gly-13. The close proximity of the Ala-14 to Gly-16 region to headgroups in both micelles and membranes is very interesting because the HFPmn conformation is predominantly helical in micelles and β strand in membranes. To date, there have been no studies about the micelle or membrane locations of cross-linked HFP oligomers.

The present solid-state NMR work uses HFPmn, HFPdm, and HFPtr which are ¹³CO labeled at either Ala-6 or Ala-15 and examines samples in which there are both helical and β strand local conformations. Because the ¹³CO chemical shifts of helical and β strand Ala are well-resolved, the data provide information about the dependence of HFP membrane location on conformation as well as residue position. To our knowledge, this is the first study of HFP membrane location for which HFP conformation is explicitly known.

Materials and Methods

Materials

Resins and Fmoc-protected amino acids were purchased from Peptides International (Louisville, KY). ¹³C-carboxy labeled amino acids were obtained from Cambridge Isotope Laboratories (Andover, MA) and were Fmoc-protected using literature methods (31). Lipids were obtained from Avanti (Alabaster, AL). Most other chemicals were obtained from Sigma-Aldrich (Milwaukee, WI). The most commonly used aqueous buffer contained 5 mM N-(2-hydroxy-ethyl)piperazine-N′-2-ethanesulfonic acid (HEPES) at pH 7.0 with 0.01% (w/v) NaN₃ preservative.

Peptide synthesis

Table 1 displays the peptide constructs and Fig. 1 describes the synthetic schemes. Preloaded β-Ala Wang resin was often used because of its low substitution of ~0.2 mmol/g and preloaded Gly Wang resin was also used. The non-native C-terminal lysines were introduced to increase the solubility of the peptides and the non-native tryptophan provided a chromophore for quantification (13). The HFPmn constructs were made by linear synthesis. The HFPdm construct was made by cysteine cross-linking of HFPmn(Cys) and the HFPdm(Cys) construct was made from a dimer scaffold which included a peptide bond formed between the cysteine carboxyl group and a lysine ε-NH₂. The HFPtr construct was made by cysteine cross-linking of HFPdm(Cys) and HFPmn(Cys/Gly) and the HFPte construct was made by cysteine cross-linking of HFPdm(Cys).

Table 1.

Names and sequences of the HIV fusion peptides^a

Name	Sequence b
HFPmn	AVGIGALFLGFLGAAGSTMGARSWKKKKKKAβ
HFPmn(Cys)	AVGIGALFLGFLGAAGSTMGARSWKKKKKCAβ
HFPmn(Cys/Gly)	AVGIGALFLGFLGAAGSTMGARSWKKKKKCG
HFPdm(Cys)
HFPdm
HFPtr
HFPte

^aThe nomenclature of HFP samples labeled at ¹³COs of A6 or A15 in the text are given as HFPmn-A6, HFPmn-A15, etc.

^bA line between K and C denotes a peptide bond between the Cys CO and the Lys ε-NH and a line between two Cs denotes a disulfide bond.

Figure 1.

Synthetic schemes of HFPmn, HFPdm, HFPtr, and HFPte. FP ≡ AVGIGALFLGFLGAAGSTMGARS. A black circle represents a resin bead, lines are drawn to clarify chemical functionalities, an arrow signifies a chemical reaction, and two arrows signify multiple sequential chemical reactions. All reactions were carried out at ambient temperature. The specific reactions are described as follows. Reaction a: Fmoc deprotection in 3 mL of 20% piperidine/DMF (v/v), 15 minutes/cycle, 2 cycles. Reaction b: Peptide synthesis with Fmoc chemistry. For HFPmn syntheses, 2-hour single couplings were used for each amino acid with the following exceptions: 4-hour single couplings for Trp, Ser and Arg residues; 6-hour single couplings for the Leu-12 to Leu-7 residues; double coupling with 3-hours per coupling for the ¹³CO labeled residue. For the HFPdm(Cys) synthesis, 4-hour single couplings were used for each amino acid with the following exceptions: 8-hour single couplings for Trp, Ser, and Arg residues; double couplings with 4-hours per coupling for the ¹³CO labeled residue and for the Leu-12 to Leu-7 residues. Reaction c: Cleavage from the resin using a 4 mL solution containing TFA/thioanisole/ethanedithiol/anisole in 90:5:3:2 volume ratio. After 2.5 hours reaction time, TFA was removed with nitrogen gas and peptide was precipitated with cold methyl t-butyl ether. Reaction d: Coupling using PyAOP and DIPEA (1:2 molar ratio) in 4 mL DMF with 6 hour reaction times for Cys and 2 hour reaction time for Lys. Reaction e: Cross-linking in 5 mM DMAP, pH = 8.4, open to the air. HFPdm or HFPte conditions: 1 µmol HFPmn(Cys) or HFPdm(Cys) in 400 µL solution overnight. HFPtr conditions: 1 µmol HFPmn(Cys) and 1.5 µmol HFPdm(Cys) in 400 µL solution for 2.5 hours. Reaction f: Selective deprotection of Mtt in 3 mL of 1% TFA/DCM (v/v), 6 minutes/cycle, 6 cycles.

Much of the synthesis was done manually in 5 mL polypropylene columns from Pierce (Rockford, IL) and mixing was accomplished with a rotation stage. Resin washing was done with N,N-dimethylformamide (DMF). Deprotection of the Fmoc group was done with two cycles of 20% piperidine in 3 mL and deprotection of the Mtt group was done with six cycles of 1% trifluoroacetic acid (TFA) in 3 mL dichloromethane (DCM). The first step in coupling of an amino acid to the to α-NH₂ terminus of the resin-bound peptide was dissolution in 4 mL DMF of 0.5 mmol protected amino acid, 0.5 mmol O-benzotriazole-N,N,N′,N′-tetramethyl-uronium-hexafluoro-phosphate (HBTU), 0.5 mmol 1-hydroxybenzotriazole (HOBT) and 1.0 mmol N,N-diisopropylethylamine (DIPEA). After ten minutes of activation, coupling was initiated by adding the amino acid solution to the resin. After coupling, acetylation or “capping” of unreacted peptides on the resin was accomplished with a ten minute reaction with a 3 mL solution containing acetic anhydride:pyridine:DMF in a 2:1:3 volume ratio.

For HFPdm(Cys), the dimer scaffold was formed from coupling of the Cys COOH to the ε-NH₂ of the resin-bound Lys and this step began with dissolution in 4 mL DMF of 0.5 mmol Fmoc-Cys(Trt), 0.5 mmol 7-azabenzotriazol-1-yloxy-tris-(pyrrolidino)phosphonium hexafluorophosphate (PyAOP), and 1.0 mmol DIPEA. The amino acid solution was transferred to the resin which had been washed twice with DCM and once with DMF after Mtt deprotection. One key aspect of this procedure was minimization of the time that the resin was at non-acidic pH prior to addition of the amino acid solution. This minimum time suppressed the undesired side reaction of deprotection of the Lys Fmoc group by the Lys ε-NH₂ (13, 15).

For the HFPmn and HFPdm synthetic protocols, coupling at each residue was optimized by ninhydrin monitoring every two hours to detect free α-HN₂ groups (32). This information provided the basis for longer coupling times and double couplings at particular residues. Further explanation of the optimizations is provided in the Results section. Some of the HFPmn, HFPmn(Cys), and HFPmn(Cys/Gly) syntheses were done on an automated peptide synthesizer (ABI 431A, Foster City, CA) using protocols similar to the ones described for manual synthesis. The automated synthesizer was also sometimes used for HFPdm(Cys) peptide elongation (Fig. 1 “b” step).

Crosslinking reactions of HFPdm, HFPtr, and HFPte were done open to the atmosphere at pH 8.4 and were stopped by ~10-fold dilution with water and lyophilization. Further details are provided in the caption of Fig. 1. For the NMR studies, single ¹³C carbonyl (¹³CO) labels were incorporated at either Ala-6 or Ala-15 in HFPmn, HFPdm, and HFPtr and the resultant peptides were denoted HFPmn-A6, HFPmn-A15, etc.

Crude peptides were obtained by cleavage with a mixture of TFA:thioanisole:ethanedithiol:anisole in a 90:5:3:2 volume ratio. The purification of crude peptides was completed by HPLC (Dionex, Sunnyvale, CA) equipped with a semi-preparative C18 column (Vydac, Hesperia, CA). “Buffer A” was water with 0.1% TFA, “buffer B” was 90% acetonitrile, 10% water, and 0.1% TFA, and the gradient was 40% to 80% buffer B over 30 minutes. Peptide masses were measured with MALDI-TOF mass spectroscopy using a Voyager-DE STR biospectrometry workstation (Applied Biosystems, Foster City, CA) and α-cyano-4-hydroxy cinnamic acid matrix. Peptide synthetic yields were quantified using 280 nm absorbance and the extinction coefficients were 5700, 11600, 17300, and 23200 cm⁻¹ M⁻¹ for HFPmn, HFPdm, HFPtr, and HFPte respectively.

Lipid mixing induced by HFPs

Mixing of lipids between membrane vesicles was monitored by a fluorescence assay (33). Two types of large unilamellar vesicles (LUVs) were prepared. The “unlabeled LUVs” contained 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-rac-(1-glycerol)] (POPG) in a 4:1 mol ratio. This composition approximately reflected the ratio of neutral to negatively charged lipids in membranes of host cells of HIV and correlated with the lipid composition used in previous structural studies of viral fusion peptides (34, 35). The “labeled LUVs” contained 77 mol% POPC, 19 mol% POPG, 2 mol% of the fluorescent lipid N-(7-nitro-2,1,3-benzoxadiazol-4-yl)-phosphatiylethanolamine (N-NBD-PE), and 2 mol% of the quenching lipid N-(lissamine Rhodamine B sulfonyl)-phosphatidylethanolamine (N-Rh-PE). HFP-induced fusion was examined in a solution with an unlabeled:labeled vesicle ratio of 1:9 so that a labeled vesicle would likely fuse with an unlabeled vesicle. The resultant lipid mixing would yield a larger average distance between fluorescent and quenching lipids and increased fluorescence.

LUV preparation began with dissolution of lipid in chloroform followed by removal of the chloroform with nitrogen gas and overnight vacuum pumping. The lipid film was suspended in 5 mM HEPES buffer and the lipid dispersion was homogenized with ten freeze-thaw cycles. LUVs were formed by extrusion through a filter with 100 nm diameter pores (Avestin, Ottawa, ON).

Fluorescence was recorded on a stopped-flow fluorimeter (Applied Photophysics SX.18MV-R, Surrey, UK) using excitation and emission wavelengths of 465 and 530 nm, respectively. For a single run, one syringe in the fluorimeter contained HFPmn, HFPdm, HFPtr, or HFPte dissolved at a concentration of 3.00, 1.50, 1.00, or 0.75 µM in HEPES buffer. A second syringe contained labeled and unlabeled LUVs at 300 µM total lipid concentration. At time zero, equal volumes of the two solutions were mixed and fluorescence was recorded every second for 200 s. The HFP concentrations were chosen so that the HFP strand:lipid ratio was always 0.010.

Most reports of fluorescence based lipid mixing have focused on ΔF_fusogen, the net change in fluorescence after the fusogen is added to the vesicles. ΔF_fusogen is typically compared to ΔF_detergent, the change caused by addition of a detergent which completely solubilizes the vesicles. Because of the very large average distance between fluorescent and quenching lipids in the solubilized vesicles, ΔF_detergent is the maximum observable fluorescence change. The “percent lipid mixing” is typically defined as ΔF_fusogen/ΔF_detergent × 100. In order to provide some comparison between our stopped-flow fluorescence data the lipid mixing literature, the raw data at each time point, F_raw(t) were converted to normalized F(t):

$$F(t)=\left\{\raisebox{1ex}{$[F_{\mathit{raw}}(t)-F_{\mathit{initial}}]$}\!\left/ \!\raisebox{-1ex}{$\mathrm{\Delta }{F}_{\mathit{max}}$}\right.\right\}\times 100$$ (1)

F_initial was a typical value of fluorescence at t = 0 and ΔF_max was chosen to provide semi-quantitative comparison between F(t) and earlier studies of percent lipid mixing induced by HFPs (12). A single value of F_initial and a single value of ΔF_max were used for all of the data.

At the end of the 200 s collection time, the fluorescence from HFPmn-induced lipid mixing was still appreciably increasing and it was therefore difficult to fit these data to a buildup function. The fluorescence of the HFPdm, HFPtr, and HFPte constructs had leveled off and these data fitted much better to the sum of two exponential buildup functions than to a single buildup function:

$$F(t)=F_0+F_1(1-e^{-k_{1}t})+F_2(1-e^{-k_{2}t})$$ (2)

where F₀, k₁, F_1, k₂, and F₂ were fitting parameters. The best-fit value of F₀ was close to 0 because of the way F(t) was calculated in Eq. 1. A convention was chosen that k₁ > k₂ so that k₁ and F₁ were respectively the rate constant and overall fluorescence change of the faster lipid mixing process and k₂ and F₂ were the rate constant and overall fluorescence change of the slow process.

Data were collected for each construct at 25, 30, 35 and 40 °C and each HFPdm, HFPtr, and HFPte data set was fitted with Eq. 2. For each construct, the temperature dependence of k₁ was fitted to the Arrhenius Equation ln k₁ = ln A − E_a/RT where R was the ideal gas constant, T was the absolute temperature, and A and E_a were the pre-exponential factor and activation energy, respectively. Three independent runs were done for each construct and temperature.

Preparation of solid-state NMR samples

The membrane lipids were 1,2-di-O-tetradecyl-sn-glycero-3-phosphocholine (DTPC) and 1,2-di-O-tetradecyl-sn-glycero-3-[phospho-rac-(1-glycerol)] (DTPG). For each HFP construct, one sample was made with DTPC:DTPG = 4:1 ≡ “PC:PG” and the other with DTPC:DTPG:cholesterol = 8:2:5 ≡ “PC:PG:CHOL”. The lipid:cholesterol ratio of PC:PG:CHOL correlated with the ratio found in membranes of cells infected by HIV (34, 35). DTPC and DTPG were used because they were ether-linked rather than ester-linked lipids and thus lacked natural abundance ¹³CO NMR signals. Spectral interpretation was therefore simplified because the ¹³CO region of the NMR spectrum was dominated by the labeled ¹³CO nucleus of the HFP. For membrane-associated HFPtr, a previous study showed that the Leu-7 ¹³CO chemical shift and presumably the Leu-7 local conformation was similar in membranes containing either ether-linked or ester-linked lipids (14).

Each PC:PG sample contained 16 µmol DTPC and 4 µmol DTPG and each PC:PG:CHOL sample contained 16 µmol DTPC, 4 µmol DTPG, and 10 µmol cholesterol. Sample preparation began with making LUVs in 2 mL HEPES buffer using the protocol described in the previous section. Either HFPmn (0.80 µmol), HFPdm (0.40 µmol), or HFPtr (0.27 µmol) was dissolved in 2 mL HEPES buffer and the HFP and vesicle solutions were then mixed and gently vortexed together with a resulting HFP strand:lipid = 0.040 for all samples. The mixture was refrigerated overnight and ultracentrifuged at ~ 150,000g for 5 hours. Unbound HFPs do not pellet and at least HFPmn binding to membranes has been determined to be approximately quantitative (13, 36). The membrane pellet with associated bound HFP was transferred to a 4 mm diameter magic angle spinning (MAS) NMR rotor.

Solid-state NMR REDOR experiments and data analysis

The evolution of ¹³CO magnetization under the effect of ¹³CO-³¹P dipolar coupling was measured with solid-state NMR rotational-echo double resonance (REDOR) experiments. The REDOR pulse sequence contained in sequence: (1) a 50 kHz ¹H π/2 pulse; (2) 1 ms cross-polarization with 52 kHz ¹H field and 58–69 kHz ramped ¹³C field; (3) a dephasing period τ which contained ~ 50 kHz ¹³C π and in some cases ~ 60 kHz ³¹P π pulses with XY-8 phase cycling on each channel; and (4) ¹³C detection (14, 37). Two-pulse phase modulation (TPPM) ¹H decoupling with 100 kHz Rabi frequency was applied during the dephasing and detection periods. For each sample and each τ, two spectra were acquired. The dephasing period in the “S₁” acquisition contained a ¹³C π pulse at the end of each rotor cycle except for the last cycle and a ³¹P π pulse in the middle of each cycle. There were no ³¹P π pulses in the “S₀” acquisition. MAS averaged the ¹³C evolution due to ¹³C-³¹P dipolar coupling to zero over each rotor period of the S₀ acquisition, while the two π pulses in each rotor cycle of the S₁ acquisition disrupted the averaging and resulted in a net dipolar coupling “d” with consequent reduction of the ¹³C signal. Determination of d was based on the difference of ¹³C intensity between the S₀ and the S₁ spectra. For a single ¹³CO-³¹P spin pair, the ¹³CO-³¹P distance “r” in Å is related to d in Hz by r = 23.05/d^1/3. Further details are provided in the Supporting Information.

Results

Peptide synthesis

Table 1 summarizes the HFPmn, HFPdm, HFPtr, and HFPte constructs synthesized in the current studies and Fig. 1 shows the synthetic schemes. A manual synthesis was carried out on HFPmn and the efficiency after each coupling was monitored with the ninhydrin test. Two-hour single coupling was not sufficient for the Ser, Arg, and Trp residues or for the residues between Leu-12 and Leu-7 and the optimized synthesis used longer coupling times or double coupling for these residues. Fig. 2a displays the HPLC chromatogram of an optimized HFPmn synthesis. The mass spectrum of the dominant fraction had an intense peak with m/z = 3149 Da which was very close to the expected m/z = 3151 Da for HFPmn, cf. Fig. 2b. The HFPmn(Cys) synthesis was carried out under the same conditions, had a very similar chromatogram, and the mass spectrum of the dominant fraction had m/z = 3125 Da which was very close to the expected m/z = 3126 Da of HFPmn(Cys).

Figure 2.

The left column panels are HPLC chromatograms obtained during peptide purification and correspond to the following syntheses: a, HFPmn; c, HFPdm; e, HFPdm(Cys); g, HFPtr; and i, HFPte. The top, middle, and bottom chromatograms in panel g are for syntheses with HFPtr cross-linking times of 0.5, 1.5, and 2.5 hours, respectively. The vertical scales in panels a, c, e, and i are A₂₁₄ and the vertical scale in panel g is A₂₈₀. In panels c and i, the large peaks at 8 minutes are due to DMAP. For a given chromatogram in the left column panel, the HPLC fraction marked with an asterisk was analyzed by MALDI-TOF mass spectroscopy and the corresponding mass spectrum is displayed in the right column panel. The spectral peaks marked with asterisks are discussed in the main text.

HFPdm was made by cross-linking HFPmn(Cys) and the HPLC retention time of HFPdm was very well-separated from that of unreacted HFPmn(Cys), cf. Fig. 2c. The HFPdm fraction had a mass spectral peak with m/z = 6248 Da which was very close to the expected m/z = 6153 Da, cf. Fig. 2d. There was also a peak at m/z = 3124 Da which corresponded to either the HFPmn(Cys) fragment formed from cleavage of the disulfide bond in the mass spectrometer or to doubly charged HFPdm. The insulin standard at 5734 Da was also apparent. Ninhydrin monitoring of the manual synthesis of HFPdm(Cys) showed that the difficult coupling steps were the same as those of the HFPmn synthesis and the optimized synthesis included longer and/or double couplings. Fig. 2e displays the chromatogram of the synthesis and the retention time of HFPdm(Cys) was very close to that of HFPdm. The mass spectrum of the HFPdm(Cys) fraction had an intense peak with m/z = 6190 Da which was very close to the expected m/z = 6188 Da, cf. Fig. 2f. A previously published synthesis showed significant higher molecular weight impurities which were the result of: (1) premature removal of the scaffold lysine Fmoc group by nucleophilic attack of the scaffold lysine ε-NH₂; and (2) subsequent coupling of the next amino acid with both the ε- and the α-NH₂ of the scaffold lysine (13). The mass spectrum of HFPdm(Cys) did not show the impurity corresponding to a peptide with an extra Cys. Minimization of the time between steps f and d in Fig. 1 was critical to eliminating this impurity.

HFPtr was formed from cross-linking HFPmn(Cys) and HFPdm(Cys) in a 1.0:1.5 mol ratio. The non-stoichiometric ratio was based on initial small-scale syntheses which showed that cross-linking of HFPmn(Cys) with itself to form HFPdm was more rapid than cross-linking of HFPmn(Cys) with HFPdm(Cys) to form HFPtr. The top chromatogram in Fig. 2g was obtained after 0.5 hour cross-linking time and the three prominent peaks from left-to-right were HFPmn, HFPdm, and HFPtr, respectively. The middle and bottom chromatograms were obtained with cross-linking times of 1.5 and 2.5 hours, respectively, and showed a relative increase in HFPtr and relative decreases in HFPmn(Cys) and HFPdm(Cys) with longer cross-linking time. The mass spectrum of the HFPtr fraction had a peak with m/z = 9307 Da which was close to the expected HFPtr m/z = 9312 Da, cf. Fig. 2h. The peak at 4653 Da was assigned to doubly charged HFPtr and the peaks at 3112 Da and 6194 Da were assigned to HFPmn(Cys) and HFPdm(Cys) fragments formed from cleavage of the disulfide bond in the mass spectrometer. The insulin standard at 5734 Da was also apparent. HFPte was formed from cross-linking HFPdm(Cys) and the chromatogram showed good separation of the HFPte and unreacted HFPdm(Cys) fractions, cf. Fig. 2i. Relative to the mass spectra of HFPmn, HFPdm, and HFPtr, the mass spectrum of the HFPte fraction had lower signal-to-noise, cf. Fig. 2j. There were broad signals peaked at m/z = 12461 Da and 6235 Da which were respectively comparable to the expected HFPte m/z = 12374 Da and the m/z of the doubly charged species or the HFPdm(Cys) fragment formed from cleavage of the disulfide bond in the mass spectrometer. Relative to the HFPmn, HFPdm, and HFPtr spectra, there were greater uncertainties of the experimental m/z in the HFPte spectra because of both broader signals and lower signal-to-noise.

Lipid mixing

Fig. 3a shows stopped-flow fluorescence data which track lipid mixing induced by HFPs. Both HFPtr and HFPte induced similar rapid lipid mixing while HFPdm induced slower mixing and HFPmn induced little mixing. For each construct, data were acquired at 25, 30, 35, and 40 °C and the HFPdm, HFPtr, and HFPte data could be fit well to a biexponential buildup function, cf. Eq. 2. The best-fit parameters of the 35 °C data are listed in Table 2 and the parameters for other temperatures are listed in tables in the Supporting Information. For each of the three constructs, k₂ ≈ 0.1 k₁ and F₁ ≈ F₂. In addition, k₁^tr/k₁^dm ≈ 2.5 and k₁^te/k₁^tr ≈1.3. Fig. 3b displays Arrhenius plots for the k₁ rate constants and the best-fit E_as and ln As are listed in Table 2. The values of E_a and ln A for HFPdm and HFPtr are comparable to those reported in a previous study (13). The data show that E_a^dm > E_a^tr > E_a^te and ln A_dm > ln A_tr > ln A_te. An increased number of strands in the oligomer is therefore correlated with decreased ln A and E_a with concomitant opposite effects on k₁. The activation entropies were calculated using the transition state theory equation ΔS^‡ = R × [ln(Ah/k_BT) − 2] where R is the ideal gas constant, h is Planck’s constant, k_B is Boltzmann’s constant, and T is the absolute temperature (13). The resultant ΔS^‡ were all negative with ΔS^‡_dm > ΔS^‡_tr > ΔS^‡_te but we do not understand the sign or trends of the ΔS^‡ values.

Figure 3.

Panel a displays stopped-flow monitored changes in lipid fluorescence induced by addition of different HFP constructs to an aqueous solution containing membrane vesicles. Increased fluorescence is a result of mixing of lipids between different vesicles and this mixing is one consequence of vesicle fusion. The lines are color coded: HFPmn (black); HFPdm (red); HFPtr (blue); and HFPte (green). The total lipid concentration was 150 µM and the HFPmn, HFPdm, HFPtr, and HFPte concentrations were 1.50, 0.75, 0.50, and 0.37 µM, respectively, so that peptide strand:lipid = 0.01. The data were collected at 25 °C, the vesicle composition was 4:1 POPC:POPG, and the initial vesicle diameter was ~100 nm. Additional data (not shown) were obtained for HFPdm, HFPtr, and HFPte at 30, 35, and 40 °C. Each data set for each construct was analyzed as the sum of two exponential buildup functions. Panel b displays Arrhenius plots for the rate constants of the fast buildup function with legend: HFPdm (open square); HFPtr (open circle); and HFPte (open triangle). The best-fit lines are also displayed and result in the respective activation energies 41 ± 3, 26 ± 1, and 20 ± 1 kJ/mol.

Table 2.

Fitting parameters for the lipid mixing kinetics at 35°C ^a,^b

Construct	k1 (10−3 s−1)	k2 (10−3 s−1)	F0 (a.u.)	F1 (a.u.)	F2 (a.u.)	Eac (kJ/mol)	ln A	ΔS‡d (J/mol-K)
HFPdm	54.9(0.3)	4.9(0.2)	0.5(0.1)	19.6(0.3)	19.8(0.1)	40.6(3.4)	13.2(1.2)	−152
HFPtr	139(3)	17.3(0.3)	0.9(0.2)	50.1(0.2)	50.0(0.2)	25.8(1.2)	8.0(0.8)	−195
HFPte	185(7)	20.7(0.3)	0.9(0.3)	49.8(0.3)	50.1(0.2)	20.1(0.7)	6.2(0.5)	−210

^aFitting uncertainties are given in parentheses. The variation of a parameter value from fitting data of different runs was less than the fitting uncertainty of a single run.

^bThe k₁, k₂, F₀, F₁ and F₂ were obtained using Eq. 2 in the main text.

^cE_a and ln A were calculated using ln k₁ = ln A – E_a/RT and k₁ values from temperatures between 25 and 40 °C.

^dΔS^‡ was calculated using ΔS^‡ = R × [ln (Ah/k_BT) – 2].

Solid-state NMR

Fig. 4a displays the ¹³CO S₀ and S₁ REDOR spectra at τ = 24 ms for the PC:PG-associated HFP-A6 samples. The lineshapes of the S₀ spectra did not show a strong dependence on dephasing time and the displayed spectra were therefore representative of the ¹³CO shift distributions of the samples. The ¹³CO S₀ signal had ~3/4 contribution from the labeled Ala-6 and ~1/4 natural abundance contribution from the unlabeled residues. Because the latter contribution was due to ~29 chemically distinct ¹³COs, it would be a broad signal and therefore less apparent than the sharper ¹³CO signals from the single labeled residue. In each spectrum, there are two peaks with chemical shifts of ~180 and ~175 ppm which likely have dominant contributions from separate populations of helical and β strand Ala-6 conformations (38). This correlation of ¹³CO chemical shift with conformation was previously confirmed with backbone ¹³CO-¹⁵N distance measurements on HFPtr (14). For PC:PG:CHOL-associated HFP-A6 samples, the 175 ppm peak is dominant and is consistent with preference for β strand conformation in cholesterol-containing membranes, cf. Fig. 4b. The correlation holds for all three constructs.

Figure 4.

Panel a displays ¹³C-detect/³¹P-dephase REDOR S₀ and S₁ spectra for PC:PG-associated HFPs and panel b displays spectra for PC:PG:CHOL-associated HFPs. The dotted lines are added to provide easier visual comparison of the S₀ and S₁ intensities. The samples for the top, middle, and bottom spectra contained HFPmn-A6, HFPdm-A6, and HFPtr-A6, respectively, and the HFP:lipid mol ratios were 0.040, 0.020, and 0.013. Each spectrum was obtained with 24 ms dephasing time and was processed with 300 Hz Gaussian line broadening and polynomial baseline correction. For panel a, the numbers of scans summed for each S₀ and S₁ spectrum were: HFPmn-A6, 40832; HFPdm-A6, 74756; and HFPtr-A6, 20736. For panel b, the numbers of scans were: HFPmn-A6, 28288; HFPdm-A6, 85136; and HFPtr-A6, 29696. Panels c and d display plots of (ΔS/S₀)^exp vs dephasing time with color coding: HFPmn (black); HFPdm (red); and HFPtr (blue). A 1σ error bar is displayed for each point and lines between points are displayed for visual clarity. Each (ΔS/S₀)^exp value was determined from integrations of 1 ppm regions of the S₀ and S₁ spectra. In the panel c plot, the integration was centered at the ~180 ppm peak for samples containing PC:PG and for the panel d plot, the integration was centered at the ~175 ppm peak for samples containing PC:PG:CHOL. These two peaks are respectively assigned to helical and β strand A6 conformations, respectively.

Fig. 4c displays a plot of (ΔS/S₀)^exp vs dephasing time for the helical component of the PC:PG-associated HFP-A6 samples and Fig. 4d displays the corresponding plot for the PC:PG:CHOL samples in which the β strand conformation is dominant. In Fig. 4d, the (ΔS/S₀)^exp ≈ 0 for all τ and all constructs and this was also true for the β strand component in the PC:PG samples (not shown). This result is consistent with Ala-6 ¹³CO-³¹P distances greater than 8 Å. For the helical component of HFPtr, (ΔS/S₀)^exp ≈ 0 for all τ while for HFPdm, there was a small increase with increasing τ and for HFPmn, a larger increase which leveled off at 0.2 at large τ. These data support an Ala-6 membrane location which is closest to the lipid ³¹Ps in HFPmn, further away in HFPdm, and furthest away in HFPtr. For the helical conformation, there may therefore be a positive correlation between the number of strands in the oligomer and the depth of membrane insertion. This variation in membrane location may be one reason for the large differences in fusion rates among the different constructs.

Fig. 5 displays the ¹³CO S₀ and S₁ REDOR spectra at τ = 24 ms for the (a) PC:PG-and (b) PC:PG:CHOL-associated HFP-A15 samples. In PC:PG, the peak ¹³CO shift was ~178 ppm and at least for HFPtr, there was also a clear shoulder at ~176 ppm while in PC:PG:CHOL, the peak ¹³CO shift was ~176 ppm. These shifts generally correlate with a mixture of helical and β strand conformations for Ala-15 in PC:PG and predominant β strand conformation in PC:PG:CHOL. This result is consistent with the Ala-6 data. The linewidths of components of individual Ala-15 conformations are significantly broader than the corresponding components of Ala-6 conformations which is consistent with greater structural heterogeneity at Ala-15, cf. Fig. 4b and Fig. 5b.

Figure 5.

Panel a displays ¹³C-detect/³¹P-dephase REDOR S₀ and S₁ spectra for PC:PG-associated HFPs and panel b displays spectra for PC:PG:CHOL-associated HFPs. The samples for the top, middle, and bottom spectra contained HFPmn-A15, HFPdm-A15, and HFPtr-A15, respectively. Panels c and d display plots of (ΔS/S₀)^exp vs dephasing time with color coding: HFPmn (black); HFPdm (red); and HFPtr (blue). Experimental parameters and data presentation are very similar to those described in Figure 4 caption. For panel a, the numbers of scans summed for each S₀ and S₁ spectrum were: HFPmn-A6, 41728; HFPdm-A6, 68096; and HFPtr-A6, 30720. For panel b, the numbers of scans were: HFPmn-A6, 28032; HFPdm-A6, 82176; and HFPtr-A6, 81536. In panel c, integration was centered at the peak chemical shift of the PC:PG samples (~178 ppm) and in panel d, integration was centered at the peak chemical shift of the PC:PG:CHOL samples (~176 ppm).

Fig. 5c displays the plot of (ΔS/S₀)^exp vs τ for the PC:PG-associated HFP-A15 samples and Fig. 5d displays the corresponding plot for the PC:PG:CHOL-associated samples. The (ΔS/S₀)^exp were derived from 1 ppm integration windows at the peak shifts and therefore reflect helical and β strand conformations in the respective samples. For all constructs and all conformations, there was a large increase in (ΔS/S₀)^exp with increasing τ and the (ΔS/S₀)^exp were generally much larger than for the HFP-A6 samples. These data support a membrane location for Ala-15 which is much closer to the lipid ³¹Ps than is Ala-6.

Quantitative analysis of ¹³C-³¹P REDOR data

A more quantitative analysis of the REDOR data was carried out for the HFP-A15 samples. The analysis included approximate separation of the natural abundance contribution from the labeled contribution to (ΔS/S₀)^exp so that the resultant (ΔS/S₀)^lab reflected the Ala-15 ¹³CO-³¹P proximity. The (ΔS/S₀)^lab/(ΔS/S₀)^exp ratio varied with τ with a typical range 0.9 < (ΔS/S₀)^lab/(ΔS/S₀)^exp < 1.2 so a plot of (ΔS/S₀)^lab vs τ was similar to the corresponding plot of (ΔS/S₀)^exp vs τ. As inferred from Fig. 5c, d, there was a plateau value of (ΔS/S₀)^lab in each data set which was between 0.3 and 0.4 and it was therefore considered that only a fraction “f ” of the Ala-15 ¹³COs had detectable d. This plateau effect had also been observed in fitting ¹³CO-³¹P data of a membrane-associated antimicrobial peptide with helical conformation (37). The Supporting Information contains further information and details about the natural abundance correction, calculation of (ΔS/S₀)^sim values as a function of d, and fitting of (ΔS/S₀)^lab to (ΔS/S₀)^sim to determine best-fit d and f.

Table 3 summarizes the best-fit ds and fs for the three constructs in the two different compositions and conformations. The calculation of (ΔS/S₀)^sim as a function of d was based on a single nearby ³¹P nucleus. All of the ds were in the range of 60 – 90 Hz which correspond to ¹³CO-³¹P distances in the 5 – 6 Å range and all of the fs were in the range of 0.3 – 0.4. Because the HFP ¹³CO may be close to more than one lipid headgroup, fittings were also done using simulations with one ¹³CO and two ³¹P spins and resulted in best-fit ¹³CO-³¹P distances of 5 – 6 Å, f = 0.6 – 0.8, and an angle between the two ¹³CO-³¹P vectors close to 180° (14, 39, 40). For this angle, (ΔS/S₀)^sim ≈ 0.5 at long dephasing times. There is insufficient information to choose between the one and two ³¹P spin models.

Table 3.

Best-fit ¹³C-³¹P dipolar coupling (d) and fractional maximum dephasing (f ) parameters for HFP-A15 samples ^a,^b,^c

	PC:PG			PC:PG:CHOL
	d (Hz)	f	χ2	d (Hz)	f	χ2
HFPmn-A15	75 (8)	0.32 (0.02)	8.0	88 (7)	0.38 (0.03)	6.7
HFPdm-A15	62 (4)	0.33 (0.02)	6.8	72 (8)	0.36 (0.03)	3.5
HFPtr-A15	90 (6)	0.36 (0.01)	0.3	90 (3)	0.36 (0.01)	1.9

^aFitting values are presented with uncertainties in parenthesis.

^bThe χ² values were calculated using Eq. S4 in the Supporting Information.

^cThe d values in Hz units are related to distances (r) in Å units using r = 23.05/d^1/3. For r = 5 Å, d = 98 Hz and for r = 6 Å, d = 57 Hz.

Discussion

This manuscript describes syntheses, fusion activities, conformations, and membrane locations of four HFP constructs including the HFPmn monomer and the HFdm, HFPtr, and HFPte cross-linked oligomers that are dimeric, trimeric, and tetrameric, respectively. The putative oligomerization state of HIV gp41 is a trimer and motivations for our studies included understanding the role of oligomerization in the HFP model system and providing insight into the role of oligomerization in intact HIV/host cell fusion.

One result of this work is an improved synthesis of HFPtr and the synthesis of HFPte. The previous synthetic strategy for HFPtr used a trimeric scaffold and simultaneous coupling of amino acids onto each chain of the scaffold (13). This approach often led to synthetic failure because of the requirement for ~180 successful deprotection and coupling reactions. In addition, use of a 4-hour-coupling time for all residues did not consider inefficient coupling for residues that contained large side chains or side chain protecting groups. The present study increased the HFPtr yield and purity using the following modifications: (1) HFPtr was formed from a cysteine cross-linking reaction between HFPmn(Cys) and HFPdm(Cys). Because HFPdm(Cys) was synthesized using a dimeric scaffold, a successful synthesis required 1/3 fewer reactions than the earlier HFPtr synthesis. In addition, the purification of the cross-linking reaction was fairly straightforward because of the separation of the HPLC peaks corresponding to HFPmn(Cys), HFPdm and HFPdm(Cys), HFPtr, and HFPte, cf. Fig. 2g. (2) Ninhydrin monitoring of coupling reactions in the manual syntheses of HFPmn and HFPdm(Cys) showed that longer coupling times were required for the Trp, Ser, Arg, and Leu-12 to Leu-7 residues. The new synthetic protocol used longer coupling times or double coupling at these residues. (3) The HFPdm(Cys) synthetic protocol was modified to minimize the time between the cleavage of the Mtt group of the Lys ε-NH₂ and the subsequent coupling to Cys. This modification reduced undesired deprotection of the Fmoc group of the Lys α-NH₂ by the ε-NH₂ (15).

The lipid mixing at long-time was ordered HFPmn < HFPdm < HFPtr ≈ HFPte, cf. Fig. 3a. The k₁^HFPtr/k₁^HFPdm ≈ 2.5 while k₁^HFPte/k₁^HFPtr ≈ 1.3, cf. Table 2, where k₁ was the rate constant of the fast component of lipid mixing. In addition, E_a^HFPdm − E_a^HFPtr ≈ 15 kJ/mol whereas E_a^HFPtr − E_a^HFPte ≈ 6 kJ/mol. These data indicate: (1) cross-linking increases the rate and extent of HFP-induced lipid mixing and decreases activation energy; and (2) the increase in lipid-mixing-per-strand and decrease in activation energy with cross-linking levels off at HFPtr. It might be expected that oligomer folding would be more difficult with an increasing number of monomer units so the putative trimeric oligomerization state of gp41 and other class I viral fusion proteins may be the optimal balance between higher catalytic efficiency and more difficult folding. We do not know why lipid mixing appears to be a combination of a fast (k₁) and slow (k₂) processes.

The ¹³CO chemical shift distributions in the S₀ spectra of Fig. 4 and Fig. 5 provided information about local conformation with the analysis based on the literature conformational dependences of the ¹³CO shifts (38). For a single membrane composition, qualitatively similar NMR spectra and therefore conformational distributions were observed for the HFPmn, HFPdm, and HFPtr constructs. It was therefore unlikely that the large variation in lipid mixing rates among the constructs were due to conformational differences.

The S₀ spectra also provided specific information about the populations of helical and β strand conformation at the Ala-6 and Ala-15 residues. There is a significant literature on the membrane-associated conformation of different HFP constructs with reports of both helical and β strand conformation (2–5, 16, 22, 23). Higher HFP:lipid ratio is one factor which favors β strand conformation and correlates with formation of β sheet aggregates at higher peptide concentration in the membrane (23). All of the NMR samples in the present study had peptide strand:lipid mol ratio = 0.04 but differed in the residue position of the labeled ¹³CO and in membrane cholesterol content. Consideration of membrane cholesterol may be important because cholesterol is ~30 mol% of host cell membranes and ~45 mol% of HIV membranes (34, 35).

The effect of membrane cholesterol on conformation is most clearly understood with the HFP-A6 S₀ spectra, cf. Fig. 4a, b. In PC:PG membranes, the spectra showed a mixture of two signals with peak shifts of ~180 and ~175 ppm which were assigned to populations of helical and β strand conformations, respectively (38). In PC:PG:CHOL membranes, there were single peaks with shift of ~175 ppm, i.e. predominant β strand conformation. Further evidence for correlation of the β strand conformation with the 175 ppm shift is an earlier complete ¹³C chemical shift assignment of residues 1 to 16 of HFPmn associated with cholesterol-containing membranes (24). The ¹³C shifts of these residues, including Ala-6, were consistent with β strand conformation.

For the HFP-A15 S₀ spectra, the effects of membrane cholesterol were more subtle, cf. Fig. 5a, b. In PC:PG membranes, there was a dominant peak at ~178 ppm for all three constructs and at least for HFPtr, there was also a clear shoulder at ~176 ppm. In PC:PG:CHOL membranes, there were single peaks at ~176 ppm. The 178 ppm signal was assigned to helical conformation based in part on the 181.2 ppm and 179.1 ppm ¹³CO shifts of Ala-6 and Ala-15, respectively in detergent-associated HFPmn (17). These residues were helical in the detergent structure and the 2 ppm difference in shifts correlated with the 2 ppm difference between the putative helical Ala-6 and Ala-15 shifts in PC:PG membranes. The 176 ppm shift in PC:PG:CHOL membranes was assigned to β strand conformation based on: (1) literature chemical shifts; (2) the previously described work on uniformly labeled HFPmn; and (3) earlier work showing specific antiparallel β strand registries for a large fraction of HFPmn in PC:PG:CHOL (24, 38). In particular, this latter study demonstrated that an antiparallel β sheet was formed for residues Ala-1 to Gly-16 with adjacent strand crossing near Phe-8 and Leu-9. For these registries, Ala-6 was in the middle and Ala-15 was near the edge of the β sheets. The Ala-6 and Ala-15 residues would likely be in more and less ordered environments, respectively, which generally correlated with narrower ~3 ppm and broader 4–5 ppm linewidths observed for these residues in PC:PG:CHOL membranes. Our interpretation of the HFP-A15 spectra was thus similar to that of the HFP-A6 spectra with helical and β strand populations in PC:PG and predominant β strand conformation in PC:PG:CHOL.

One overall conclusion of the chemical shift analysis was that membrane cholesterol was associated with β strand conformation of the HFP. Although this result had been previously suggested by work from our group, the present study provides much clearer evidence for this observation because the only difference between the PC:PG and PC:PG:CHOL membranes was membrane cholesterol content. A similar correlation between membrane cholesterol and β strand conformation has been observed for the influenza virus fusion peptide so the correlation may be a general property of fusion peptides (41, 42). Although the reasons for the structural effect of membrane cholesterol are poorly understood, it is useful to consider the increased lateral molecular packing density in cholesterol-containing membranes and the difference in sizes of a putatively small monomeric HFP helix and a larger HFP β sheet aggregate (43). Relative to the aggregate, the small helix might experience a more positive increase in free energy of membrane insertion with higher packing density.

Spectra from other PC:PG samples with peptide strand:lipid ratio ≈ 0.04 (not shown) had populations of helical and β strand chemical shifts that were similar to those displayed in Fig. 4a. The general observations based on multiple samples were: (1) the population of helical conformation was lower in HFPmn samples relative to HFPdm and HFPtr samples; and (2) for all constructs, there was some β strand population. The latter observation was consistent with earlier studies on HFPmn samples near this peptide strand:lipid ratio (23, 42). For HFPmn, HFPdm, and HFPtr samples with PC:PG:CHOL membranes, there was always predominant β strand conformation, cf. Fig. 4b. An earlier study showed that the β strand conformation was also formed in samples made by the different protocol of initial cosolubilization of HFPmn, lipids, and cholesterol in organic solvent followed by removal of the solvent and then hydration (44). Therefore, the β strand conformation in PC:PG:CHOL membranes is probably a thermodynamic equilibrium rather than a kinetically trapped structure.

For a particular construct, the lipid mixing rate was approximately independent of the absence or presence of membrane cholesterol (13). One interpretation of these data is that both the helical and the β strand HFP conformations induce vesicle fusion while an alternate interpretation is that fusion is induced by unstructured HFP (45, 46). This transient HFP state would not be apparent in the NMR samples which reflect the long-time end-state HFP structure. Experimental support for the first interpretation is a HFPmn study which showed that the rates of membrane binding and secondary structure formation were faster than the rate of lipid mixing (47).

As there were not large conformational variations between the HFPmn, HFPdm, and HFPtr constructs in a single membrane composition, HFP location in membranes was considered as a factor to explain the significant differences in fusion activities. In addition, detection of distinct signals for helical and β strand ¹³COs provided an opportunity to examine the dependence of membrane location on conformation. An earlier ¹³CO-³¹P REDOR study was carried out on HFPmn constructs containing ¹³CO labels at three sequential residues (30). For Ala-14 to Gly-16 labeled HFPmn in either PC:PG or PC:PG:CHOL membranes, (ΔS/S₀)^exp ≈ 0.3 – 0.4 at 24 ms dephasing time while (ΔS/S₀)^exp ≤ 0.15 for Gly-5 to Leu-7, Phe-8 to Gly-10, or Phe-11 to Gly-13 labeled HFPmn. The overall conclusion was that the Ala-14 to Gly-16 residues were closer to the lipid phosphate headgroups than were the Gly-5 to Gly-13 residues. The singly ¹³CO labeled samples of the present study were consistent with these results and provided the following additional insights: (1) Ala-15 is closer to the phosphate headgroups than is Ala-6; (2) this proximity difference is observed for both helical and β strand conformations; and (3) this difference is observed for the HFPmn, HFPdm, and HFPtr constructs. These insights are supported by the smaller (ΔS/S₀)^exp for all HFP-A6 constructs for both helical and β strand conformations and the larger (ΔS/S₀)^exp for all HFP-A15 constructs for both helical and β strand conformations, cf. Fig. 4 and Fig. 5. The proximity of the Ala-15 residue to the phosphate groups may be an intrinsic property of the HFP sequence. This is perhaps explained by Ala-15 being at the junction of the more apolar N-terminal and the more polar C-terminal regions of the sequence. These regions likely have negative and positive free energies of membrane insertion, respectively (48).For signal-to-noise reasons, the experiments described in this paper were done at a nominal temperature of −50 °C in which the PC:PG membranes were in the gel phase and the PC:PG:CHOL membranes were in a glass of the liquid-ordered phase (49). Although we do not have data in the more biologically relevant higher temperature liquid-crystalline and liquid-ordered phases, similar ¹³CO-³¹P REDOR studies have been carried out for the influenza virus fusion peptide in the lower temperature gel and higher temperature liquid-crystalline phases and similar (ΔS/S₀)^exp values were obtained in both phases (50).

There is general consistency between our results and earlier experimental work on HFPmn location in membranes and micelles. Liquid-state NMR studies on helical HFPmn in detergent micelles showed that the Ala-15 residue was close to the detergent headgroups. In addition, the Ala-6 residue was in the micellar interior and was further away from these headgroups (17, 18). The membrane location of HFPmn with a F8W mutation has been examined by fluorescence and it was concluded that the Trp sidechain was in the bilayer interior and was 10 ~ 11 Å away from the phosphate group longitude. Very similar locations were found in samples with HFPmn-F8W:lipid = 0.0025 and 0.02 and for membranes without or with cholesterol (27, 28). These data generally correlate with the small (ΔS/S₀)^exp of HFPmn-A6 in PC:PG and PC:PG:CHOL.

Computer simulations have yielded two distinct models of the membrane location of helical HFPmn (25, 26). For the “deep insertion” model, there was traversal of both membrane leaflets while in the “shallow insertion” model, HFPmn was restricted to the outer leaflet. For the deep insertion model, the Ala-6 and Ala-15 residues were respectively ~15 Å and ~5 Å away from the phosphorus longitude while in the shallow insertion model, they were ~5 Å and ~3 Å from this longitude. Both models are consistent with the experimentally observed proximity of Ala-15 to the phosphate groups. However, the non-zero (ΔS/S₀)^exp for the helical Ala-6 peak in HFPmn and HFPdm samples were not consistent with 100% deep insertion. There may be discrete populations of deeply and shallowly inserted helical HFPs (21).

Table 3 shows that the best-fit ds for all HFP-A15 samples were consistent with 5–6 Å ¹³CO-³¹P distances. Qualitative consideration of these distances and the van der Waals radii of HFP and the phosphate group suggests close contact of the Ala-15 residue with the phosphate group. The fittings also resulted in best-fit fs of 0.3 – 0.4 which were consistent with two populations of HFPs that differed in having Ala-15 close and far from the phosphate groups. For β sheet structure, these populations could be strands at the edges and the middle of the sheet, respectively (24).

Although the data to-date generally supported similar membrane locations of HFPmn, HFPdm, and HFPtr, there was some indication of differences in (ΔS/S₀)^exp for the helical peak in the HFP-A6 samples. In particular, Ala-6 appeared to be closest to the phosphate groups in HFPmn followed by HFPdm and HFPtr which would correlate with shallowest membrane insertion for HFPmn and deeper insertion for HFPdm and HFPtr. This suggests the intriguing hypothesis that the membrane insertion depth correlates with fusogenicity.

Supplementary Material

1_si_001. Supporting Information Available.

Detailed description of NMR experiments and analysis and listing of values of (ΔS/S₀)^exp, (ΔS/S₀)^lab, and (ΔS/S₀)^na for the HFP-A15 samples and detailed listing of the best-fit parameters for the lipid mixing data. This material is available free of charge via the Internet at http://pubs.acs.org.

ABBREVIATIONS

a.u.

arbitrary units

d

dipolar coupling

DCM

dichloromethane

DIPEA

N,N-diisopropylethylamine

DMAP

4-dimethylaminopyridine

DMF

N,N-dimethylformamide

DMPC

1,2-dimyristoyl-sn-glycero-3-phosphocholine

DPPC-¹³C

1,2-dipalmitoyl[1-¹³C]-sn-glycero-3-phosphocholine

DTPC

1,2-di-O-tetradecyl-sn-glycero-3-phosphocholine

DTPG

1,2-di-O-tetradecyl-sn-glycero-3-[phospho-rac-(1-glycerol)]

Fmoc

9-fluorenylmethoxycarbonyl

HBTU

O-benzotriazole-N,N,N′,N′-tetramethyl-uronium-hexafluoro-phosphate

HEPES

N-(2-hydroxy-ethyl)piperazine-N′-2-ethanesulfonic acid

HFP

HIV fusion peptide

HFPdm

HFP dimer

HFPmn

HFP monomer

HFPte

HFP tetramer

HFPtr

HFP trimer

HIV

human immunodeficiency virus

HMB

N-2-hydroxy-4-methoxybenzyl

HOBT

1-hydroxybenzotriazole

HPLC

high-performance liquid chromatography

LUVs

large unilamellar vesicles

MALDI-TOF

matrix-assisted laser desorption/ionization-time of-flight

MAS

Magic Angle Spinning

NMR

nuclear magnetic resonance

N-NBD-PE

N-(7-nitro-2,1,3-benzoxadiazol-4-yl)-phosphatiylethanolamine

N-Rh-PE

N-(lissamine Rhodamine B sulfonyl)-phosphatidylethanolamine

POPC

1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine

POPG

1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-rac-(1-glycerol)]

PyAOP

7-azabenzotriazol-1-yloxy-tris-(pyrrolidino)phosphonium hexafluorophosphate

REDOR

rotational-echo double resonance

TFA

trifluoroacetic acid

TPPM

two-pulse phase modulation