Improvement of HIV fusion inhibitor C34 efficacy by membrane anchoring and enhanced exposure

Abstract

Objectives

The aim of the present work was to evaluate the interaction of two new HIV fusion inhibitors {HIVP3 [C34–polyethylene glycol (PEG)₄–cholesterol] and HIVP4 [(C34–PEG₄)₂–cholesterol]} with membrane model systems and human blood cells in order to clarify where and how the fusion inhibitors locate, allowing us to understand their mechanism of action at the molecular level, and which strategies may be followed to increase efficacy.

Methods

Lipid vesicles with defined compositions were used for peptide partition and localization studies, based on the intrinsic fluorescence of HIVP3 and HIVP4. Lipid monolayers were employed in surface pressure studies. Finally, human erythrocytes and peripheral blood mononuclear cells (PBMCs) isolated from blood samples were used in dipole potential assays.

Results

Membrane partition, dipole potential and surface pressure assays indicate that the new fusion inhibitors interact preferentially with cholesterol-rich liquid-ordered membranes, mimicking biological membrane microdomains known as lipid rafts. HIVP3 and HIVP4 are able to interact with human erythrocytes and PBMCs to a similar degree as a previously described simpler drug with monomeric C34 and lacking the PEG spacer, C34–cholesterol. However, the pocket-binding domain (PBD) of both HIVP3 and HIVP4 is more exposed to the aqueous environment than in C34–cholesterol.

Conclusions

The present data allow us to conclude that more efficient blocking of HIV entry results from the synergism between the membranotropic behaviour and the enhanced exposure of the PBD.

Introduction

Since the discovery of HIV-1 as the causative agent of AIDS,^1,2 the scientific community has been studying different strategies to fight this infection. Despite the intense research since its discovery,³ no cure or vaccine has yet been achieved. Fusion inhibitor peptides have been developed as part of the anti-HIV-1 arsenal.^4,5 Their target is the fusion process between the viral and target cell membranes, necessary for the entry of the viral content into the cell. Viral envelope proteins gp41 and gp120 play crucial and specific functions in HIV-1 entry and are important drug targets due to their conserved structure.⁶ Upon the binding of the surface protein gp120 to CD4 on the host cell, a conformational change occurs and a coreceptor binding site is exposed,⁷ enabling gp120 binding to CCR5 (or CXCR4) receptors, which triggers a major conformational change in the transmembrane protein gp41.⁸ The N-terminal fusion peptide of gp41 is exposed and inserted into the target cell membrane. At this point, a pre-hairpin configuration is adopted in which gp41 is connected to the host cell membrane via its fusion peptide and to the viral membrane via its transmembrane domain. The two helical domains of gp41, N- and C-terminal heptad repeats (NHR and CHR), fold onto each other creating a hairpin structure, the 6-helix bundle (6HB).⁷ This structure brings the two membranes together and promotes the formation of the fusion pore used for the release of the viral content into the host cell.⁹ The fusion inhibitors block this process, preventing infection of the target cell.¹⁰

Enfuvirtide is the most studied HIV-1 fusion inhibitor peptide and one of the two approved drugs in the entry inhibitor class,¹¹ the other approved drug being the CCR5 co-receptor antagonist maraviroc.¹² Enfuvirtide targets the gp41 pre-hairpin state and is less toxic than classical antiretrovirals;¹³ however, it needs to be administered subcutaneously and is very sensitive to degradation while in circulation due to its peptide nature. In order to improve the poor solubility, potency and stability of these fusion inhibitors, several successful approaches have been made, generating new drug candidates.^5,14

In previous reports, we showed that fusion inhibitor peptides such as enfuvirtide, T-1249 and sifuvirtide partition to membranes in a composition-dependent manner,^15–18 including the lipid mixtures of the microdomains in which receptors are preferentially located. Lipophilicity is a key factor for antiviral activity because it determines the local concentration of the fusion inhibitors at the membrane level.

C34, a peptide with 34 amino acid residues that was used to target the trimeric coiled coil structure of gp41,⁹ was one of the most promising early-discovered fusion inhibitors.¹⁹ In contrast to enfuvirtide, C34 lacks a lipid-binding domain and therefore has a weaker interaction with membranes.²⁰ Introduction of a cholesterol moiety to improve the interaction with membranes caused an increase in antiviral potency,²¹ yielding a 50-fold increase in IC₅₀ (HXB2 strain) relative to C34 alone.¹⁹ Importantly, C34–cholesterol was still detected 24 h after injection in a mouse model, while C34 was undetectable after 6 h. Concomitant with its higher antiviral potency against HIV-1, C34–cholesterol has affinity to peripheral blood mononuclear cells (PBMCs) and erythrocyte membranes, while C34 has not.²⁰

An innovative approach to potentiate the activity of C34-related HIV-1 fusion inhibitors without changing the amino acid sequence was adopted by Pessi et al.;²² it relies on the combination of cholesterol tagging with dimerization of the C34 sequence. Furthermore, other multimerization approaches were also conducted, resulting in a remarkable increase in the fusion-inhibitory activity of a three-helical bundle structure of the trimeric form of C34, mimicking gp41 trimers.^23,24

Regarding the approach adopted by Pessi et al.,²² four polyethylene glycol (PEG) residues were used as spacers between the C34 sequence and the cholesterol moiety (Figure 1), aiming at further increasing antiviral potency. Both multimerization and PEG addition are known to increase the potency of some fusion and entry inhibitors.^23–27 As a result of this approach, the dimeric C34 cholesterol-tagged inhibitor HIVP4 proved to be more effective than C34–cholesterol against HIV-1 BaL strain in TZM-bl cells, with IC₅₀ in the picomolar range.²² In the case of the monomeric C34–cholesterol-tagged inhibitor HIVP3, higher antiviral activity was observed in comparison with HIVP4, using HIV-1 IIIB strain and TZM-bl cells.²² Concomitant with the improved antiviral activity, these chemically modified C34 peptides contain some unique features, such as PEG and the maleimide–cysteine pair (MAL), which are safe to use in vivo, making these fusion inhibitors promising candidates for clinical tests.

Figure 1.

Characterization of HIVP3, HIVP4 and HIVP5. Schematic representation of the peptide domains of the different fusion inhibitors within HIV-1 gp41 main domains, depicting the relative positions of the fusion peptide (FP), NHR domain, CHR domain and transmembrane (TM) region. Hydrophobic residues (A, F, I, L, M and W) are shown in blue, non-charged polar residues (H, N, Q, S, T and Y) in green and charged polar residues (D, E, K and R) in red. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Given the importance of lipophilicity and selective affinity for different kinds of lipids of peptide fusion inhibitors, the aim of this work was to characterize the interaction with lipid membranes of HIVP3 (monomeric C34–cholesterol-tagged inhibitor), HIVP4 (dimeric C34–cholesterol-tagged inhibitor) and HIVP5 (dimeric C34 without cholesterol; it has low antiviral activity and was used as a control). We used large unilamellar vesicles (LUVs) and lipid monolayers as membrane model systems and human blood cells as a biological model for what may happen in the bloodstream. Our studies focus on understanding possible relations between peptide–membrane interactions and the efficiency of these drugs, offering a rational basis for the improvement of this class of compounds.

Materials and methods

Reagents and sample preparation

The peptide conjugates were derivatives of the peptide C34 (WMEWDREINNYTSLIHSLIEESQNQQEKNEQELL). HIVP3 (C34–GSGC–PEG₄–cholesterol), HIVP4 {[C34–GSGC(MAL–PEG₄)]₂–cholesterol} and HIVP5 {[C34–GSGC(MAL–PEG₁₁)]₂} were prepared as previously described.²² 5-Doxyl-stearic acid (5NS), 16-doxyl-stearic acid (16NS) and cholesterol were from Sigma-Aldrich (Milwaukee, WI, USA). l-Tryptophan, acrylamide, HEPES and NaCl were from Merck (Darmstadt, Germany). 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) and egg sphingomyelin (SM) were purchased from Avanti Polar Lipids (Alabaster, AL, USA). The working buffer used throughout the studies was 10 mM HEPES/150 mM NaCl (pH 7.4). The tryptophan stock solution (500 μM) was prepared in buffer, while HIVP3, HIVP4 and HIVP5 stock solutions (all 500 μM) were prepared in DMSO. LUVs were prepared by extrusion methods, as described elsewhere.^28,29

Fluorescence spectroscopy measurements

HIVP3, HIVP4 and HIVP5 peptides contain tryptophan residues, which make fluorescence techniques suitable tools to probe these molecules. Membrane partition and fluorescence quenching studies using acrylamide were carried out in a Varian Cary Eclipse fluorescence spectrophotometer (Mulgrave, Australia) and time-resolved fluorescence spectroscopy studies in a LifeSpec II Fluorescence Lifetime spectrometer (Edinburgh Instruments, Livingston, UK).

Intrinsic fluorescence measurements of HIVP3, HIVP4, HIVP5 and tryptophan (for the sake of comparison) were performed with an excitation wavelength of 280 nm, except for acrylamide quenching experiments, where the excitation was performed at 290 nm to minimize the relative quencher/fluorophore light absorption ratios. For the quenching experiments, fluorescence emission was collected at a fixed wavelength of 350 nm and for the partition studies integrated spectra from 310 to 450 nm were used. Typical spectral bandwidths were 5 nm for excitation and 10 nm for emission. Excitation and emission spectra were corrected for wavelength-dependent instrumental factors.³⁰ During the quenching and partition experiments, emission was also corrected for successive dilutions, scatter and simultaneous light absorptions by quencher and fluorophore.³¹ All the fluorescence measurements in this study were performed at ∼25°C. Time-resolved intensity decays were obtained by pulse excitation at 275 nm (vertically polarized) and fluorescence was acquired at 350 nm (20 nm bandwidth, at magic angle, 54.7°), using a 20 ns time span and 1024 channels in a multichannel analyser. The fluorescence lifetimes were obtained from intensity decay fits with a sum of exponentials and using a non-linear least-squares method based on the Marquardt algorithm.³² The quality of the fits was evaluated from χ² values, distributions of the residuals and autocorrelation plots.

Partition coefficient determination

Membrane partition studies were performed by successive additions of small aliquots of LUV suspensions with different lipid compositions (pure POPC, pure DPPC, POPC:cholesterol 2 : 1 and 1 : 1, and POPC:cholesterol:SM 1 : 1 : 1) to a 5 μM HIVP3, HIVP4 or HIVP5 solution, with a 10 min incubation time between each addition. The partition coefficients (K_p) were calculated using the equation:³³

(1)

where I_W and I_L are the fluorescence intensities in aqueous solution and in lipid, respectively, γ_L is the molar volume of the lipid^34,35 and [L] its concentration.

Surface pressure

Changes in the surface pressure of lipid monolayers induced by HIVP3, HIVP4 and HIVP5 were measured in a Langmuir–Blodgett trough NIMA ST900 (Coventry, UK) at constant temperature (25 ± 0.5°C). The surface of the buffer contained in a Teflon trough of fixed area was exhaustively cleaned by surface aspiration. Then, a solution of lipids in chloroform was spread on this surface, reaching a surface pressure of 23 ± 1 mN/m. In this range of pressures, the lipid used in this study forms monolayers upon spreading on the air–water interface. Peptide solutions were injected in the subphase and the changes in surface pressure were followed during the necessary time to reach a constant value. The surface pressure of an air–water interface upon injecting the largest concentration of each peptide used throughout the studies was always <15 mN/m (data not shown). For this reason, the lowest initial surface pressure of the lipid monolayers before the addition of the peptides to the subphase was above that value. In this condition, the changes in surface pressure observed upon injection of the peptide can be attributed to an effect of the peptide on the monolayer interfacial tension.

The dissociation constant (K_d) was calculated from the adsorption Langmuir isotherm:

(2)

where ΔΠ are the changes in surface pressure, ΔΠ_max is the maximum change in pressure achieved and [peptide] is the peptide concentration.

The adsorption rate constant (k) was calculated from the equation:³⁶

(3)

Fluorescence quenching

The fluorescence quenching of 5 μM HIVP3, HIVP4 or HIVP5 by acrylamide (0–60 mM)¹⁶ was studied in buffer and in the presence of LUVs containing 3 mM POPC:cholesterol (2 : 1) by successive additions of small volumes of the quencher stock solution. For every addition, a minimum 10 min incubation time was allowed before measurement. Quenching data were analysed by using the Stern–Volmer equation:³³

(4)

where I and I₀ are the fluorescence intensities of the sample in the presence and absence of quencher, respectively, K_SV is the Stern–Volmer constant and [Q] the concentration of quencher.

Fluorescence quenching assays with the lipophilic probes 5NS and 16NS were performed by time-resolved fluorescence spectroscopy. These assays were carried out at the same peptide and lipid concentrations as those used for acrylamide quenching, by successive additions of small amounts of these quenchers (in ethanol) to samples of peptide in POPC and POPC:cholesterol (2 : 1), keeping the ethanol concentration <2% (v/v).³⁷ The effective lipophilic quencher concentration in the membrane was calculated from the partition coefficient of both quenchers to the lipid bilayers.³⁸ For every addition, a minimum 10 min incubation time was allowed before measurement. Quenching data were analysed by using the Stern–Volmer equation (Equation 4), or the Lehrer equation when a negative deviation to the Stern–Volmer relationship was observed:^38,39

(5)

where f_b is the fraction of light arising from the fluorophores accessible to the quencher.

In the case of dynamic quenching, the relationship I₀/I = τ₀/τ is valid; thus, time-resolved quenching data can be analysed by using the same equations (Equations 4 and 5).

Membrane dipole potential assessment using di-8-ANEPPS

Human blood samples were obtained from healthy volunteers, with their previous written informed consent, at the Instituto Português do Sangue (Lisbon, Portugal). This study was approved by the Ethics Committee of the Faculdade de Medicina da Universidade de Lisboa. Isolation of erythrocytes and PBMCs and labelling of these cells with di-8-ANEPPS (Invitrogen, Carlsbad, CA, USA) were performed as described before.^17,40 For erythrocyte isolation, blood samples were centrifuged at 1200 g for 10 min, plasma and buffy coats were removed, and the remaining erythrocytes were washed twice in working buffer. They were incubated at 1% haematocrit in buffer supplemented with 0.05% (m/v) Pluronic F-127 (Sigma) and 10 μM di-8-ANEPPS. PBMCs were isolated with a density gradient using Ficoll–Paque Plus (GE Healthcare, Little Chalfont, UK) and counted in a Neubauer Improved Haemocytometer. They were incubated at 3000 cells/μL in Pluronic-supplemented buffer with 3.3 μM di-8-ANEPPS. Cells were incubated with the fluorescent probe for 1 h with gentle agitation, and unbound probe was washed with Pluronic-free buffer in two centrifugation cycles. HIVP3, HIVP4 or HIVP5 (all in DMSO stock solution) were incubated with erythrocytes at 0.02% haematocrit and with PBMCs at 100 cells/μL for 1 h with gentle agitation before the fluorescence measurements. For lipid vesicle labelling, suspensions with 500 mM of total lipid were incubated overnight with 10 μM di-8-ANEPPS, to ensure maximum incorporation of the probe. The maximum concentration of DMSO in the suspensions was 2.4% (v/v) at 6 μM of peptide or cholesterol. Excitation spectra and the ratio of intensities at the excitation wavelengths of 455 and 525 nm (R = I₄₅₅/I₅₂₅) were obtained with emission set at 670 nm to avoid membrane fluidity-related artefacts.^41,42 Excitation and emission slits for these measurements were set to 5 and 10 nm, respectively. The variation in R with the peptide concentration was analysed with a single binding site model:⁴³

(6)

with the R values normalized for R₀, the value in the absence of peptide. R_min defines the asymptotic minimum value of R.

Data analysis

Fitting of the equations mentioned in this article to the experimental data was done by non-linear regression using GraphPad Prism.

Results

Membrane partition

UV-visible absorption and fluorescence spectra of HIVP5 in buffer nearly overlapped those of tryptophan in aqueous solution (Figure 2a); however, the spectra of HIVP3 and HIVP4 presented a shift to the blue, indicating that the tryptophan residues of these conjugates are in a more hydrophobic environment, as previously observed for C34–cholesterol.²⁰

Figure 2.

Partition of the fusion inhibitors to lipid vesicles. (a) Normalized fluorescence excitation (broken lines) and emission (continuous lines) spectra of 5 μM HIVP3, HIVP4, HIVP5 and tryptophan in HEPES buffer. Evaluation of tryptophan fluorescence variations in 5 μM HIVP3 (b), HIVP4 (c) or HIVP5 (d) upon titration with LUVs, performed by successive additions of POPC, POPC:Chol (2 : 1), POPC:Chol (1 : 1), POPC:Chol:SM (1 : 1 : 1) or DPPC. Continuous lines are fittings of Equation 1 to the experimental data. Trp, tryptophan; Chol, cholesterol. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The partition coefficients between the lipid and aqueous phase, K_p, were determined by fitting Equation 1 to the fluorescence intensity data, in order to quantify the extent of interaction of the peptides with the LUVs (Table 1). As shown in Figure 2(b–d), there was an increase in the fluorescence intensity of HIVP3 and HIVP4 in the presence of LUVs of different compositions. In the case of HIVP5 no significant changes in the fluorescence intensity were observed in the presence of membranes, except for DPPC, indicating an absence of significant peptide–membrane interactions, or an interaction in which the tryptophan residues are not involved.

Table 1.

Partition coefficients

Lipid mixture	Peptide	Kp	IL/IW
POPC	C34–cholesterola	1574 ± 261	1.88 ± 0.05
	HIVP3	854 ± 72	1.93 ± 0.03
	HIVP4	—b	—b
POPC:cholesterol (2 : 1)	C34–cholesterola	3454 ± 718	1.25 ± 0.01
	HIVP3	515 ± 123	1.54 ± 0.07
	HIVP4	387 ± 85	1.41 ± 0.05
POPC:cholesterol (1 : 1)	C34–cholesterola	4612 ± 436	1.32 ± 0.01
	HIVP3	456 ± 77	1.72 ± 0.07
	HIVP4	407 ± 106	1.66 ± 0.10
POPC:cholesterol :SM (1 : 1 : 1)	C34–cholesterola	1810 ± 578	1.18 ± 0.02
	HIVP3	132 ± 68	1.74 ± 0.30
	HIVP4	1519 ± 543	1.088 ± 0.01
	C34–cholesterola	331 ± 44	2.98 ± 0.16
DPPC	HIVP3	121 ± 34	4.45 ± 0.74
	HIVP4	—b	—b

Values are presented as mean ± SD.

All measurements were made at least in triplicate.

Parameters were obtained from the fitting of the fluorescence data of the partition assays of 5 μM HIVP3 and HIVP4 using Equation 1.

^aK_p values for C34–cholesterol were obtained from Hollmann et al.²⁰

^bThe type of partition curve obtained impairs the use of Equation 1.

Similar K_p values were obtained for HIVP3 and HIVP4 in POPC:cholesterol membranes, indicating that the two peptides had comparable affinity for this membrane composition. However both HIVP3 and HIVP4 showed less affinity for those membranes when compared with C34–cholesterol (Table 1). Finally, in the case of POPC:cholesterol:SM (1 : 1 : 1), membranes that mimic liquid ordered membrane microdomains (lipid rafts), HIVP4 had much higher membrane affinity than HIVP3 (Table 1), reaching values comparable to those previously obtained for C34–cholesterol. Taking into account that the only difference between C34–cholesterol and HIVP3 is the addition of a PEG moiety (Figure 1), this indicates that the introduction of a PEG spacer is the main factor responsible for this reduction in affinity.

Lipid monolayer surface pressure perturbation

From the partition results, we observed that HIVP5 seemed not to interact with lipid membranes, while HIVP3 and HIVP4 had lower membrane affinity than C34–cholesterol. However, due to the terminal position of the tryptophan residues in those peptide sequences (Figure 1), one cannot discard the hypothesis that the peptides may be interacting with membranes, leaving the tryptophan residues exposed to the aqueous environment, without changes in the quantum yield (therefore not contributing to the K_p calculation). This may be related to a higher degree of freedom conferred by the PEG linker. To test the hypothesis that membrane interaction occurs but is not sensed by tryptophan, surface pressure measurements on POPC based-monolayers were carried out. As shown in Figure 3(a–c), HIVP5 was not able to induce any significant change in the surface pressure on all tested monolayers, confirming its inability to insert in lipid membranes, in contrast to HIVP3 and HIVP4. Control experiments were carried out using only DMSO (solvent) and no significant changes were observed.

Figure 3.

Fusion inhibitor interactions with lipid monolayers. Changes in surface pressure as a function of addition of HIVP3 (filled circles), HIVP4 (filled squares), HIVP5 (open squares) or DMSO (open circles) to pure POPC (a), POPC:Chol 2 : 1 (b) or POPC:Chol:SM 1 : 1 : 1 (c) monolayers. Continuous lines are fittings of Equation 2 to the experimental data. All assays were carried out at 25°C, using an initial pressure of 23 ± 1 mN/m. Each point is the average of at least triplicates of independent samples. Error bars represent the SEM. (d–f) Kinetic behaviour of interaction of fusion inhibitors with lipid monolayers. Changes in surface pressure as a function of time after addition of HIVP3 (broken lines) or HIVP4 (continuous lines) to achieve a final concentration of 42 nM on POPC (d), POPC:Chol 2 : 1 (e) or POPC:Chol:SM 1 : 1 : 1 (f) monolayers. All assays were carried out at 25°C, using an initial pressure of 23 ± 1 mN/m. Continuous and broken lines are fittings of Equation 3 to the experimental data. Chol, cholesterol. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

K_d and ΔΠ_max values were determined by fitting the experimental data with Equation 2 in order to quantify the interaction of both HIVP3 and HIVP4 with the Langmuir monolayers (Table 2). Although HIVP3 and HIVP4 had induced surface pressure changes (ΔΠ) of the same extent on pure POPC or POPC:cholesterol (2 : 1) monolayers, the K_d values for HIVP3 were ∼7- and 10-fold higher than those for HIVP4 on POPC and POPC:cholesterol (2 : 1) membranes, respectively. These results indicate that HIVP3 has less affinity for membranes when compared with HIVP4, in contrast with the partition results, in which both peptides showed similar affinity for membranes. As the only difference between the two molecules is the dimerization of C34, it seems that dimerization can actually have a role in concentrating the peptide at membrane level (as long as the cholesterol moiety is present).

Table 2.

ΔΠ_max, K_d and k values determined from surface pressure changes

Lipid mixture	Peptide	ΔΠmax (mN/m)	Kd (nM)	k (×10−3 s−1)
POPC	C34–cholesterola	3.5 ± 0.2	59 ± 15	—
	HIVP4	2.68 ± 0.05	19.2 ± 3.4	4.7 ± 0.05
	HIVP3	4.9 ± 0.4	145.2 ± 29.2	1.3 ± 0.01
POPC:cholesterol (2 : 1)	C34–cholesterola	2.51 ± 0.05	5.3 ± 1.4	—
	HIVP4	2.97 ± 0.04	24.15 ± 1.41	7.4 ± 0.16
	HIVP3	3.12 ± 0.43	240.3 ± 64.2	3.7 ± 0.08
POPC:cholesterol:SM (1 : 1 : 1)	C34–cholesterola	2.6 ± 0.2	9.2 ± 1.3	—
	HIVP4	3.57 ± 0.19	68 ± 12	6.9 ± 0.16
	HIVP3	1.9 ± 0.2	61 ± 23	0.8 ± 0.01

Values are presented as mean ± SD.

All measurements were made at least in triplicate.

Parameters were obtained from the fitting of the surface pressure data from Figure 3 using Equations 2 and 3, respectively.

^aΔΠ_max and K_d values for C34–cholesterol were obtained from Hollmann et al.²⁰

Interestingly, the ΔΠ_max and K_d obtained for HIVP4 had the same order of magnitude as that previously obtained for C34–cholesterol,²⁰ in contrast with the partition results, suggesting that in this new inhibitor the C-terminal portion of the peptide, where the tryptophan residues are located, may have a lower interaction with the membrane.

HIVP4 showed faster kinetics of interaction in all lipid monolayers tested, in comparison with HIVP3 (Figure 3d–f). In order to quantify the kinetic behaviours, we calculated k, listed in Table 2. Among the lipid compositions tested, for HIVP4 the fastest k values were obtained for cholesterol-containing monolayers (POPC:cholesterol and POPC:cholesterol:SM). In the case of HIVP3 the fastest adsorption was observed on POPC:cholesterol monolayers.

Depth of insertion in the lipid bilayer

With the aim of confirming the hypothesis that the tryptophan residues in these new inhibitors have a more superficial adsorption on the membrane, we evaluated their depth of insertion in the lipid bilayers by differential fluorescence quenching studies. The first assay was to test the accessibility of the tryptophan residues of the peptides to the aqueous environment using acrylamide as fluorescence quencher. Linear Stern–Volmer plots were obtained and the K_SV values were determined by fitting Equation 4 to the data. Figure 4(a–c) shows the Stern–Volmer plots obtained for HIVP3, HIVP4 and HIVP5 on POPC:cholesterol (2 : 1) LUVs. In the case of HIVP4, the K_SV was similar to that obtained in the absence of the lipid (8.49 ± 0.06 and 8.17 ± 0.05 mM⁻¹ in buffer and lipid, respectively), confirming that, despite HIVP4 being able to insert in the lipid membranes, its tryptophan residues remain exposed enough to the aqueous environment to be quenched by acrylamide. Although HIVP3 and HIVP4 showed similar interaction with POPC:cholesterol (2 : 1) membranes, tryptophan residues of HIVP3 seemed not to be as much exposed to the surrounding aqueous environment (10.29 ± 0.17 and 8.70 ± 0.06 mM⁻¹ in buffer and lipid, respectively) as in the case of HIVP4. Such changes can also be associated with stabilization in the dimer of the peptide domain corresponding to the C-terminal heptad repeat region of gp41 (CHR in Figure 1). For HIVP5, as expected, no changes occurred in the K_SV value (7.23 ± 0.15 mM⁻¹ in buffer and 7.79 ± 0.22 mM⁻¹ in lipid).

Figure 4.

Localization of HIVP3 and HIVP4 in lipid bilayers. Fluorescence quenching by acrylamide of HIVP3 (a), HIVP4 (b) and HIVP5 (c) in the presence (filled circles) and absence (open circles) of POPC:Chol (2 : 1) lipid vesicles (5 μM peptide and 3 mM total lipid). Continuous lines are fittings of the Stern–Volmer equation (Equation 4) to the experimental data. (d and e) Stern–Volmer plots of the quenching of HIVP3 and HIVP4 fluorescence by 5NS or 16NS in POPC:Chol 2 : 1 LUVs, using time-resolved fluorescence measurements. Each point is the average of three independent measures. The continuous lines are fittings of the Lehrer equation (Equation 5) to the experimental data. It should be stressed that [Q]_L is the local concentration of the quencher in the strict bilayer volume.³⁸ (f) Depth of insertion of HIVP3 and HIVP4 tryptophan residues in the membrane, determined using the SIMEXDA method,⁴⁴ yielding an average location of 19.2 and 19.6 Å away from the centre of the bilayer for HIVP3 and HIVP4, respectively. The distributions' half width at half height were 5.14 Å for HIVP3 and 7.5 Å for HIVP4. Chol, cholesterol.

Regarding the in-depth location of tryptophan residues of the peptides inserted in POPC:cholesterol (2 : 1) LUVs, fluorescence quenching measurements were done using two different quenchers: stearic acid molecules derivatized with a doxyl (quencher) group either at carbon 5 (5NS) or carbon 16 (16NS). Generally, 5NS is a more efficient quencher for molecules inserted in the membrane in a shallow position, close to the lipid–water interface, while 16NS is more efficient for molecules buried deeply in the membrane.⁴⁴ Figure 4(d and e) shows the Stern–Volmer plots obtained for HIVP3 and HIVP4, respectively, on POPC:cholesterol (2 : 1), using the effective concentration of 5NS and 16NS in the bilayer matrix.⁴⁴ Fluorescence lifetime quenching data (Figure 4d and e) enabled the application of the simulated experimental data analysis (SIMEXDA) method⁴⁴ to obtain the in-depth distribution of the tryptophan residues of HIVP3 and HIVP4 (Figure 4f). Average locations of 19.6 Å for HIVP4 and 19.2 Å for HIVP3 away from the centre of the bilayer were observed, confirming a mean shallow location, in good agreement with the acrylamide quenching data.

Membrane dipole potential changes

We used the fluorescent probe di-8-ANEPPS, which is sensitive to the membrane dipole potential,⁴¹ as a reporter to detect peptide–lipid interactions. Four different compositions of lipid vesicles were evaluated, namely pure POPC, POPC:cholesterol (2 : 1 and 1 : 1) and POPC:cholesterol:SM (1 : 1 : 1). Both HIVP3 and HIVP4 were able to reduce the dipole potential of all tested membranes, as expected from the previous assays, while no changes were observed with HIVP5 (Figure 5). Regarding the different membrane composition tested, we found that upon increasing the membrane cholesterol content, there was an increase in the extent of the interaction of HIVP3 or HIVP4 (Figure 5), indicating a preference for membranes with high cholesterol content, as is the case for lipid rafts. Furthermore, the highest change on the dipole potential for both compounds was observed for the canonic lipid raft composition (POPC:cholesterol:SM).⁴⁵

Figure 5.

Fusion inhibitor interactions with di-8-ANEPPS-labelled LUVs. Differential spectra of di-8-ANEPPS bound to membranes in the presence of HIVP3 (a) and HIVP4 (b). Spectra were obtained by subtracting the excitation spectrum (normalized to the integrated areas) of labelled LUVs in the presence of peptide from the spectrum in its absence. The shift to the red (decrease in dipole potential) was peptide concentration-dependent. Spectrum traces represent 4 μM fusion inhibitor (broken lines in panels c and d) in the presence of different lipid compositions: POPC (filled circles), POPC:Chol 2 : 1 (filled squares), POPC:Chol 1 : 1 (open circles) and POPC:Chol:SM 1 : 1 : 1 (open squares). Binding profiles of HIVP3 (c) and HIVP4 (d) to LUVs, obtained by plotting the di-8-ANEPPS excitation ratio, R (I₄₅₅/I₅₂₅, normalized to the initial value), as a function of peptide concentration. DMSO, cholesterol and HIVP5 (unconjugated dimer) were also tested, as controls, and no significant changes in dipole potential were observed (data not shown). Chol, cholesterol. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Interaction with blood cell membranes

After the characterization of the interaction of the peptides with membrane model systems, we studied the peptide–membrane interaction in biological settings. We used isolated human erythrocytes and PBMCs, also labelled with the fluorescent probe di-8-ANEPPS because in previous studies it proved to be a good reporter probe for the interaction of HIV fusion inhibitors with the plasma membranes of these cells.^17,20 The membrane dipole potential significantly decreased in the presence of HIVP3 and HIVP4, in contrast to the unconjugated dimer HIVP5, which had no effect either in erythrocytes or PBMCs. This shows that, in agreement with the previous model membrane studies, HIVP5 does not interact with cell membranes, while HIVP3 and HIVP4 do. Apparent K_d values derived from the dipole potential variation curves (Table 3) were determined by fitting the experimental data with Equation 6. Figure 6 shows that HIVP3 and HIVP4 had the same decrease profile for the R value, with similar K_d values for erythrocytes, but significantly different for PBMCs. While HIVP4 and C34–cholesterol showed similar affinity for PBMCs, HIVP3 had less affinity for the membranes of these cells (Table 3), in good agreement with the surface pressure data.

Table 3.

Cell membrane interaction parameters and antiviral activity of different HIV-1 fusion inhibitors

	Cells/viral strains	C34	C34–cholesterol	HIVP3	HIVP4	HIVP5
Kd (μM)	erythrocytes	—	0.58 ± 0.08a	0.60 ± 0.08	0.49 ± 0.06	—
	PBMCs	—	0.54 ± 0.08a	1.41 ± 0.13	0.66 ± 0.11	—
IC50 (pM)	HIV-1 IIIB	—	—	80 ± 70b	210 ± 370b	—
	HIV-1 BaL	8250 ± 6150c	190 ± 170c	360 ± 330b	10 ± 10b	—

^aK_d values for C34–cholesterol were obtained from Hollmann et al.²⁰

^bIC₅₀ values were obtained from Pessi et al.²²

^cIC₅₀ values were obtained from Harman et al.¹⁹

Figure 6.

Fusion inhibitor interactions with blood cells. Differential spectra of di-8-ANEPPS bound to erythrocyte (a and b) or PBMC (d and e) membranes in the presence of HIVP3, HIVP4, HIVP5 or DMSO. Spectra were obtained by subtracting the excitation spectrum (normalized to the integrated areas) of labelled cells in the presence of the inhibitor from the spectrum in its absence. The shift to the red (decrease in dipole potential) was peptide concentration-dependent. (a, b, d and e) Spectrum traces represent different HIVP3 and HIVP4 concentrations: 0 (DMSO spectrum trace), 0.25, 1 and 5 μM. The represented concentration for HIVP5 is 7 μM. (c and f) Binding profiles of HIVP3 (filled circles), HIVP4 (filled squares), HIVP5 (open squares) and DMSO (open circles) to erythrocyte (c) and PBMC (f) membranes, obtained by plotting the di-8-ANEPPS excitation ratio, R (I₄₅₅/I₅₂₅, normalized to the initial value), as a function of the peptide concentration. DMSO and HIVP5 (unconjugated dimer) were also tested, as controls. (c and f) Curves representing fitting to the single binding site equation (Equation 6). Affinity parameters are presented in Table 3. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Discussion

Recently, we showed that C34–cholesterol, a conjugated form of the classic HIV-1 fusion inhibitor peptide C34, interacts preferentially with membranes that mimic the lipid rafts where HIV-1 fusion is known to occur.²⁰ Moreover, through our previous work we clarified some key points in the mechanisms of action of different HIV-1 fusion inhibitor peptides using membrane model systems and human blood cells. One of the most important findings concerns the peptides' membranotropic properties, which can increase their local concentration at the membrane level, in this way enhancing the efficiency of the drug.^{15–18,20,40}

A combination of cholesterol-tagging and dimerization resulted in improved antiviral potency and extension of the in vivo half-life of the new C34-derived peptides. Additionally, a PEG spacer was inserted between each monomer and the cholesterol core, which has already been reported to be beneficial for other fusion inhibitor peptides, such as VIKI–PEG₄–cholesterol and PIE12 trimer.^26,46

In this context, we decided to evaluate the effect of dimerization and PEG addition on these new C34 derivatives when they interact with different biomembrane model systems and blood cells, comparing these findings with the results already published for the monomeric cholesterol-tagged fusion inhibitor C34–cholesterol.²⁰

Regarding the antiviral activities presented in Table 2, HIVP3 was more effective against the HIV-1 IIIB strain than the Bal strain. On the other hand, HIVP4 was more effective for Bal than for IIIB. Despite the small difference between these IC₅₀ values, the antiviral potencies of HIVP3 and HIVP4 were comparable. This means that a potency plateau has been reached for very high-affinity inhibitors, where an increase in the stability of the 6-helix bundle formed by the peptide does not translate into higher antiviral potency.^22,25,26

As expected, the cholesterol-tagged peptides (HIVP3 and HIVP4) interacted preferentially with cholesterol-rich membranes, as observed previously for C34–cholesterol. In contrast, HIVP5, which lacks the cholesterol moiety, did not interact with lipid membranes, as reported previously for C34,²⁰ due to the lack of the lipid-binding domain (Figure 1).

From the fluorescence spectroscopy data, the K_p values (Table 1) show that HIVP3 and HIVP4 had similar affinity for POPC:cholesterol membranes. However, these molecules showed a lower partition to POPC-based membranes in comparison with C34–cholesterol (Table 1). Regarding the lipid raft-mimicking membrane composition (POPC:cholesterol:SM), HIVP4 showed a higher K_p value than HIVP3, indicating that the cholesterol-tagged dimer has more affinity for this type of membrane composition. It should be stressed that partition constants are assessed based on the change of environment of the tryptophan residues. Analysing the dipole potential changes using di-8-ANEPPS-labelled LUVs (Figure 5), which enables assessment of the interaction of the whole fusion inhibitor with the membrane, it becomes clear that the extension of the interaction of HIVP3 or HIVP4 with membranes increases with the amount of cholesterol present in the membranes. It seems that both fusion inhibitors preferentially bind to liquid-ordered (L_o) membranes (POPC:cholesterol 1 : 1 and POPC:cholesterol:SM 1 : 1 : 1) instead of liquid-disordered (L_d) ones (pure POPC).⁴⁷

Concerning the surface pressure measurements (Figure 3a–c and Table 2), HIVP3 showed more affinity for lipid raft-mimicking membranes, while HIVP4 had a preference for POPC and POPC:cholesterol 2 : 1 membranes. For HIVP3, the K_d values obtained for all lipid monolayers tested were significantly higher than those for C34–cholesterol (Table 2), meaning that this new molecule has a lower affinity for membranes. However, HIVP4 exhibited a lower K_d in POPC monolayers, but higher values on cholesterol-containing monolayers. This result does not indicate that the peptide concentration at the membrane level is lower, because in HIVP4 there are two C34 peptide chains per cholesterol moiety, while C34–cholesterol has one peptide chain and a cholesterol moiety. Concerning the kinetics of the interaction of HIVP3 and HIVP4 with the POPC-based monolayers (Figure 3d–f), we found that HIVP4 had much faster binding kinetics in all tested monolayers when compared with HIVP3 or C34–cholesterol.²⁰ This result can be explained by the dimerization, with two C34 per cholesterol moiety, which doubles the electric charge per molecule, this being is important in the first approach to the membrane.

HIVP5 did not induce any change in the membrane partition, surface pressure or dipole potential measurements, confirming that cholesterol tagging can in fact boost the interaction of cholesterol-tagged peptides with membranes, as previously observed for C34–cholesterol.²⁰

Regarding the depth of HIVP3 and HIVP4 in the lipid bilayer, our quenching data show that the tryptophan residues of both peptides were located at a shallow position, close to the membrane surface and highly exposed to the aqueous environment (Figure 4). The interfacial location indicated by the fluorescence quenching methodologies confirms that the HIVP3 and HIVP4 pocket-binding domains (PBDs) (where the tryptophan residues are located) are less involved in the interaction with the membrane, in good agreement with the partition data. Furthermore, comparing these results with those previously published for C34–cholesterol, it is possible to infer that the PEG spacer may be responsible for this behaviour. The addition of a PEG spacer between the peptide region and the cholesterol moiety may also contribute to this orientation, besides the presence of the cholesterol moiety itself. These two factors may increase the exposure of the peptide chains, close to the membrane surface. As a result of this higher exposure, it should be easier for the peptide chains to acquire a proper orientation and align in an anti-parallel fashion with the NHR domain of gp41, resulting in higher antiviral activity (Figure 7). In the case of the dimer, HIVP4, this effect seems to be even more significant.

Figure 7.

Putative mode of action of HIV-1 fusion inhibitors HIVP3, HIVP4 and HIVP5. It was demonstrated that both active peptides partition to cell membranes, especially to those rich in cholesterol and SM, while HIVP5 does not. HIVP3 and HIVP4 anchor to the membrane via their cholesterol moiety and also, as expected, with weaker binding via their tryptophan-rich N-terminal domain. In the context of HIV-1 gp41 engagement with the target cell, a confined space exists between the two membranes. Concentrated in the lipid raft environment, the drug may reach its target (gp41) more efficiently than through simple diffusion in aqueous solution. Moreover, the anchoring promoted by the cholesterol at the C-terminus and the higher exposure of the PBD to aqueous media due to the PEG spacer bring the peptide into contact in the correct orientation to compete with the NHR binding site. Thus, gp41-mediated fusion may be inhibited, blocking viral content entry into the cell.

In order to obtain a biological perspective of what may happen in the bloodstream, we also studied the interaction of the peptides with human erythrocytes and PBMCs. These two types of blood cell are the main target of the virus for distribution and replication all over the body. On this basis we decided to evaluate the changes in membrane dipole potential induced by the antiviral agents, labelling the cells with di-8-ANEPPS. Our data show that HIVP3 and HIVP4 decreased the membrane dipole potential of the two cell types (Figure 6), indicating that both peptides interact with erythrocytes and mononuclear lymphocytes. As expected from the partition and surface pressure data, the untagged dimer HIVP5 did not interact with these cells. The same concentration of cholesterol alone was not enough to induce a detectable membrane potential change,⁴⁸ showing that the peptide region of the conjugate (with the PEG spacer) is inducing the membrane potential change. The K_d values (Table 3) show that HIVP3, HIVP4 and C34–cholesterol had comparable affinities for erythrocytes, but with HIVP3 presenting a considerably lower affinity for PBMCs.

Overall, our data confirm that the more efficient blocking of HIV entry associated with the PEG spacer and the dimerization of the peptide region of the inhibitors derives from enhanced exposure of the PBD in the context of the membranotropic behaviour promoted by the cholesterol moiety, as depicted in Figure 7. These findings offer a rational basis for the design of improved fusion inhibitors, since they suggest that maximizing antiviral activity requires finding the proper balance of membrane affinity and exposure of the peptide moiety, through variations in the lipid-binding domain, the PEG spacer region and the number of peptide moieties in the construct. Moreover, we offer an experimental strategy to guide the development of the structure–activity relationship towards this goal.

Funding

This work was partially supported by Fundação para a Ciência e a Tecnologia—Ministério do Ensino e da Ciência (FCT-MEC, Portugal) projects PTDC/QUI-BIQ/104787/2008 and DELIN-HIVERA/0002-2013, and by National Institutes of Health (NIH, USA) project #R21NS076385 (to M. P.). A. H. also acknowledges FCT-MEC fellowship SFRH/BPD/72037/2010.

Transparency declarations

A. P. has an ownership interest in JV Bio SRL, which has intellectual property rights on some of the compounds used in the present study. The other authors have none to declare.