It was reported recently that a new aryl methyldiene rhodanine derivative, LJ001, and oxazolidine-2,4-dithione, JL103, act on the viral membrane, inhibiting its fusion with a target cell membrane. The aim of the present study was to investigate the interactions of these two active compounds and an inactive analogue used as a negative control, LJ025, with biological membrane models, in order to clarify the mechanism of action at the molecular level of these new broad-spectrum enveloped virus entry inhibitors. Fluorescence spectroscopy was used to quantify the partition and determine the location of the molecules on membranes. The ability of the compounds to produce reactive oxygen molecules in the membrane was tested using 9,10-dimethylanthracene, which reacts selectively with singlet oxygen (1O2). Changes in the lipid packing and fluidity of membranes were assessed by fluorescence anisotropy and generalized polarization measurements. Finally, the ability to inhibit membrane fusion was evaluated using FRET. Our results indicate that 1O2 production by LJ001 and JL103 is able to induce several changes on membrane properties, specially related to a decrease in its fluidity, concomitant with an increase in the order of the polar headgroup region, resulting in an inhibition of the membrane fusion necessary for cell infection.
The development of new drugs against HIV-1 and other enveloped viruses that cause public health threats has been the focus of intense research for decades . Conventional therapeutic strategies aim at viral proteins responsible for each and every step of viral replication. Although several drugs are now available, targeting these different steps of the virus life cycle, the appearance of resistances and the hope for more efficient drugs have been pushing forward the development of new strategies to treat viral infections. The fusion of the viral and cell membranes, and the consequent entry of the viral content into the host cell, is a critical moment of the life cycle. Efficiently blocking this process prevents all of the subsequent intracellular steps, most importantly the integration of the viral genome, which can stay silent for years. Despite this promising approach, only two HIV-1 entry inhibitors are currently available in the market: maraviroc, an inhibitor of viral envelope protein gp120 (glycoprotein 120) binding to the CCR5 (CC chemokine receptor 5) co-receptor , and enfuvirtide, a fusion inhibitor peptide targeting gp41 (glycoprotein 41) in its pre-fusion conformation . The drawbacks of these approaches include the emergence and selection for drug-resistant viruses . Targeting less variable factors is an attractive concept that is less prone to drug-resistant viruses’ selection . Our increasing understanding of the virus–cell fusion process suggests that physiological differences that exist between viral and cellular membranes may be exploited for the development of antiviral therapeutics. A particularly appealing notion is that viral membrane-targeting agents would necessarily limit the development of resistance, as it is not even conceivable how such resistance may develop. Thus targeting viral membranes represents an exciting new paradigm to explore regarding the development of broad-spectrum antivirals .
Recently, broad-spectrum small-molecule antiviral drugs which prevent enveloped viruses entry at an intermediate step after virus binding but before virus–cell fusion were described [6,7]. These compounds, an aryl methyldiene rhodanine derivative, named LJ001, and an oxazolidine-2,4-dithione, named JL103 (Figure 1), act deleteriously on the virus envelope, but not at the cell membrane level. Both have no significant cytotoxicity and do not affect the cell–cell fusion process. The specificity for inhibiting virus–cell fusion seems to be related to the differences between the biogenic properties of cell plasma membranes and those of static viral membranes. Although previous studies on this subject [6,7] are important milestones for future antiviral strategies, a number of questions remain unanswered.
Molecular structures of the enveloped virus entry inhibitors LJ001 and JL103, together with the inactive analogue LJ025 [6,7]
The aim of the present study was to investigate the interactions of LJ001, JL103 and LJ025 (an inactive analogue [6,7] used as a control) with biomembrane model systems, in order to clarify the mechanism of action of these fusion inhibitors at the molecular level, with a special focus on the change that these molecules are able to induce on membrane properties. With this aim, we studied the partition of these molecules towards selected lipid bilayers, making use of the changes on their intrinsic fluorescence properties when they adsorb on to or penetrate the lipid environment. Using a fluorescence quenching methodology, employing probes located on different regions of the lipid membrane, we evaluated the in-depth location of the three molecules (active and inactive). Additionally, we studied the ability of each compound to produce singlet oxygen (1O2), and the alteration that these drugs are able to cause on membrane fusion and on lipid bilayer parameters that influence it, such as membrane order and fluidity.
Materials and sample preparation
POPC (1-palmitoyl-2-oleyl-sn-glycero-3-phosphocholine) and DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine) were purchased from Avanti Polar Lipids. NBD-PE [1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-oxa-1,3-benzoxadiazol-4-yl] and rhodamine B-PE (rhodamine B 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine) were from Invitrogen. 5NS (5-doxyl-stearic acid) and 16NS (16-doxyl-stearic acid) were purchased from Sigma. DPH (1,6-diphenyl-1,3,5-hexatriene), TMA-DPH [4′-(trimethylammonio)diphenylhexatriene p-toluenesulfonate], Laurdan (6-dodecanoyl-2-dimethylaminonaphthalene) and DMA (9,10-dimethylanthracene) were from Sigma. The HIV-FP (HIV fusion peptide) (AVGIGALFLGFLGAAGSTMGAA), corresponding to the N-terminus of HIV gp41 (HXB-2 viral clone) was synthesized by JPT. Peptide stock solutions were prepared in DMSO (spectroscopy grade). The working buffer used throughout the studies was 10 mM Hepes (pH 7.4) and 150 mM NaCl. LJ001, LJ025 and JL103 (10 mM) stock solutions were prepared in DMSO. LUVs (large unilamellar vesicles) were prepared by extrusion methods, as described in [8,9].
Fluorescence spectroscopy measurements
Partition and acrylamide-quenching studies were carried out using a Varian Cary Eclipse fluorescence spectrophotometer. Excitation and emission wavelength maxima were 450 and 515 nm respectively for LJ001, 410 and 490 nm respectively for LJ025, and 434 and 516 nm respectively for JL103. Typical spectral bandwidths were 10 nm for excitation and 5 nm for emission. Excitation and emission spectra were corrected for wavelength-dependent instrumental factors . During the quenching and partition experiments, emission was also corrected for successive dilutions, light scattering  and simultaneous light absorption by quencher and fluorophore.
Partition coefficient determination
Membrane partition studies were performed by successive additions of small aliquots of LUV to 50 μM LJ001, LJ025 or JL103 solutions, with a 10 min incubation time in between each addition. Partition coefficients (Kp) were calculated from the fit of the experimental data with eqn (1) [12,13]:
where IW and IL are the fluorescence intensities in aqueous solution and in lipid respectively, γL is the molar volume of the lipid , and [L] is its concentration.
Quenching of 50 μM LJ001, LJ025 or JL103 by acrylamide (0–60 mM)  was studied in buffer and in the presence of 3 mM POPC (LUVs), by successive additions of small volumes of the quencher stock solution. For every addition, a minimal 10 min incubation time was allowed before measurement. Quenching data were analysed by using the Stern–Volmer equation (eqn 2) :
where I and I0 are the fluorescence intensities in the presence and absence of quencher respectively, KSV is the Stern–Volmer constant, [Q] is the concentration of quencher, and fb is the fraction of light emitted by the molecules accessible to the quencher.
Fluorescence-quenching assays with the lipophilic probes 5NS and 16NS were performed by steady-state fluorescence spectroscopy at the same peptide and lipid concentrations, by successive additions of small amounts of these quenchers (in ethanol) to samples of peptide incubated in 3 mM POPC, keeping the ethanol concentration below 2% (v/v) . The effective lipophilic quencher concentration in the membrane was calculated from the Kp of both quenchers to the lipid bilayers . For every addition, a minimal 10 min incubation time was allowed before measurement. Quenching data were analysed by using the same equations as for acrylamide quenching (eqns 2 and 3).
Determination of 1O2 production
1O2 production was measured through its effect on a fluores-cent 1O2 chemical trap. DMA reacts selectively with 1O2 to form the non-fluorescent DMAO2 (9,10-dimethylanthracene endoperoxide) . Taking into consideration the fraction of DMA not bound to liposomes, calculated from its binding constant (2.7 ml/mg), and the low fluorescence quantum yield of DMA in aqueous medium, the unbound DMA practically did not contribute to the fluorescence signal . Thus monitoring the disappearance of DMA's fluorescence signal (excitation at 379 nm and emission at 432 nm), we were able to estimate the level of 1O2 in the membrane [20,22]. 1O2-associated fluorescence disappearance data were analysed using the quenching sphere-of-action model (eqn 4) :
where K*SV is the apparent Stern–Volmer constant, indicating the reduction of fluorescence by the conversion of DMA into DMAO2, V is the sphere-of-action volume (i.e. the sphere that surrounds the chromophore within which the ‘quencher’ can be considered to be in contact with the chromophore ) and NA is Avogadro's number.
Fluorescence anisotropy measurements
LUVs of 3 mM POPC or DPMC mixtures were incubated with DPH or TMA-DPH to achieve a final probe concentration of 0.33 mol% (probe/lipid molar). After the addition of the fluorescent probes, all samples were wrapped in aluminium foil to avoid bleaching. Steady-state fluorescence anisotropy, <r>, was calculated using eqn (5) :
where Ivv and Ivh are the parallel and perpendicular polarized fluorescence intensities measured with the vertically polarized ex-citation light and G=Ihv/Ihh is a correction factor accounting for the polarization bias in the detection system. For DPH, excitation and emission wavelengths were 350 nm and 432 nm respectively, whereas for TMA-DPH, excitation was at 355 nm and emission was at 430 nm. Control measurements of DPH anisotropy were carried out in the presence of different amounts of sodium azide, a known 1O2 quencher . All measurements were taken at 37°C. Effective concentrations in the membrane of each compound (LJ001, LJ025 and JL103) were calculated as described in .
Generalized polarization measurements
Generalized polarization measurement was based on the bilayer order-dependent fluorescence spectral shift of Laurdan, which can be attributed to dipolar relaxation phenomena, originating from the sensitivity of the fluorescent probe to the polarity of its environment . Laurdan is located at the hydrophilic/hydrophobic interface of the membrane bilayer, with the lauric acid tail anchored in the phospholipids’ acyl chain region [27,28]. Measurements and temperature controls were made as described above for steady-state emission intensity. The 1:300 Laurdan/lipid molar ratio was chosen in order to increase the sensitivity of the measurements without perturbing the membrane properties [27,29]. Samples were excited at 350 nm and emission intensity was acquired at 435 nm (I435) and 500 nm (I500). Generalized polarization (GPem) was calculated from the emission intensities using eqn (6), adapted from the work of Parasassi et al. :
All measurements were made at 37°C.
Lipid mixing assays
Peptide-induced lipid mixing in vesicles was measured by FRET. This assay is based on the decrease in the efficiency of the resonance energy transfer between two membrane probes, rhodamine B-PE and NBD-PE, when the lipids of the vesicles labelled with both probes are allowed to mix with lipids from unlabelled vesicles. The concentration of each of the fluorescent probes within the pre-fusion liposome membrane was 0.6 mol%. Unlabelled LUVs were pre-incubated for 30 min with 100 μM of each fusion inhibitor, or DMSO for a control. These vesicles were mixed with double-labelled LUVs in a 1:4 proportion (labelled/unlabelled) at a total final lipid concentration of 100 μM, at 37°C, under constant stirring. Fluorescence was measured with excitation at 470 nm and emission recorded between 500 and 650 nm. Excitation and emission slits were set to 10 nm. Phospholipid mixing, resulting from membrane fusion (or hemifusion), was quantified on a percentage basis according to eqn (7) :
where R is the ratio between the fluorescence intensity with emission at 530 nm and 588 nm (corresponding to the fluorescence emission maxima of NBD and rhodamine B respectively) obtained 10 min after HIV-FP addition, R0 is the ratio before peptide addition (constant during the evaluated time range), and R100% was set with vesicles labelled with 0.12 mol% of each of the fluorophores (one-fifth of the mol% of the previous measurements, corresponding therefore to a full lipid mixing between labelled and unlabelled vesicles). Assays were conducted in triplicate. Identical experiments were conducted in the presence of 15 mM sodium azide.
Interaction with lipid membranes
LJ001, LJ025 and JL103 have low fluorescence quantum yields in aqueous solution, but undergo an approximately 3-fold increase in the quantum yields when inserted into a membrane [6,7]. In a previous study , we showed that these three molecules are able to interact with lipid membranes with different affinities. In the case of LJ001 and LJ025, we showed previously that when lipid membranes were non-limiting (>50-fold molar excess), over 85% of these molecules were protected from acrylamide (a water soluble quencher) and thus buried in the lipid bilayer . Now, we carried out identical experiments with JL103, in order to compare the fraction of compound protected from acrylamide in the membrane, finding that, in this case, less that 50% of the compound is protected (1−fb), indicating an aqueous phase-accessible location (Table 1).
|Kp (×104) ||10.8±3.3||23.8±3.0||0.058±0.059|
|DMA quenching||K*SV (M)||33.5±1.08||~0||88.9±1.78|
|DMA quenching (in methanol)||K*SV (M)||0.321±0.012||~0||1.07±0.205|
|ΔΠ (mN/m) ||3.91±0.25||3.05±0.29||1.90±0.10|
|Fusion inhibition (%)||63.47±2.3||12.39±1.4||84.02±1.5|
|IC50 (nM) ||181.9±244.4||Inactive||12.2±9.7|
|Kp (×104) ||10.8±3.3||23.8±3.0||0.058±0.059|
|DMA quenching||K*SV (M)||33.5±1.08||~0||88.9±1.78|
|DMA quenching (in methanol)||K*SV (M)||0.321±0.012||~0||1.07±0.205|
|ΔΠ (mN/m) ||3.91±0.25||3.05±0.29||1.90±0.10|
|Fusion inhibition (%)||63.47±2.3||12.39±1.4||84.02±1.5|
|IC50 (nM) ||181.9±244.4||Inactive||12.2±9.7|
Differential fluorescence-quenching methodologies were used in order to determine the in-depth location of the compounds inserted in POPC vesicles (Figure 2). Stearic acid molecules derivatized with doxyl (quencher) groups either at C-5 (5NS) or C-16 (16NS) were used. 5NS is a better quencher for molecules located near the lipid/water interface, whereas 16NS is more efficient for molecules buried deeply in the membrane . Stern–Volmer plots were obtained for LJ001, LJ025 and JL103 using the effective concentration  of 5NS and 16NS in the bilayer matrix. In order to obtain the in-depth distribution of molecules, we applied the SIMEXDA (simulated experimental data analysis) method  to these data. As expected from the acrylamide data, a mean location deep inside the membrane was obtained for both LJ025 and LJ001, with LJ025 found to be located significantly deeper inside the bilayer than LJ001 (Figure 2D). However, in the case of JL103, a superficial location was observed, with the maximum of the distribution coinciding with the bilayer surface (Figure 2).
Compound location in 3 mM POPC LUVs
1O2 generation in the lipid membrane
Both active compounds (LJ001 and JL103) are able to produce 1O2 molecules, but not the inactive LJ025, as reported previously . In order to enable a thorough comparison, in the present study, we determined the 1O2 production associated with the three compounds, in 3 mM pure POPC LUVs and methanol (Figure 3). The K*SV constant value for each compound in the presence of POPC was obtained, as well as the sphere-of-action volume (V), i.e. the volume of the sphere surrounding the chromophore within which the ‘quencher’ can be considered to be in contact with the chromophore  (Table 1). As expected, LJ001 and JL103 were able to produce 1O2 both in methanol and in the membrane (Figure 3). Independently of the environment, the extinction of DMA fluorescence was considerably higher for JL103 than for LJ001. On the other hand, LJ025 did not induce any significant changes in the fluorescence intensity, either in methanol or POPC LUVs, indicating an absence of significant 1O2 production, as expected .
DMA conversion by 1O2 production in methanol (A) and in 3 mM POPC LUVs (B)
Effects on membrane rigidity
In a previous study, we proposed that the 1O2 produced could be able to attack the carbon–carbon double bond of unsaturated phospholipids in the membrane, with a concomitant increase in its rigidity . As the resistance of lipids to peroxidation decreases with the increase in the degree of unsaturation , we compared the effects that LJ001, LJ025 and JL103 are able to induce on membrane rigidity of POPC, a phospholipid with a double bond at C-9 in the oleoyl chain, and DMPC, a completely saturated phospholipid. It should be stressed that, despite the different gel–liquid phase-transition temperatures (Tm) of both lipids (approximately 23°C for DMPC and −5°C for POPC), anisotropy determinations were made at 37°C, in order to guarantee that both lipid systems were in the liquid crystalline phase, with similar DPH anisotropy values (~0.065±0.005) for both phospholipids.
For the three compounds tested, we observed an increase in the anisotropy values in POPC LUVs (Figure 4), both with DPH, which locates in the hydrophobic regions of the bilayer, closer to its centre , and TMA-DPH, which is anchored by its trimethylamine group closer to the transition between the hydrophobic and hydrophilic regions of the bilayer . We found more extensive changes induced by the compounds able to generate 1O2 for both probes (Figure 4). In spite of the higher production of 1O2 by JL103, similar changes in the membrane anisotropy were observed for LJ001 and JL103 (Table 1).
DPH (A) and TMA-DPH (B) fluorescence anisotropy changes induced by the addition of JL103, LJ001 or LJ025 to POPC LUVs, measured at 37°C
Upon replacing POPC with DMPC, we found for both compounds able to generate 1O2 molecules, a marked decrease on their membrane rigidification effect, relative to POPC (Figure 5). However, for LJ025, unable to generate 1O2, no significant changes from the values obtained with POPC were found on the DMPC membrane rigidity (Figure 5). For DMPC, we were also able to follow changes in the Tm, close to 23°C in the absence of additional compounds . For LJ001 and JL103, we observed a Tm shift towards higher temperatures (from 23.0±0.3°C to 25.9±0.3°C and 26.2±0.2°C respectively), but no significant changes were observed for LJ025 (23.4±0.4°C) (Figure 5D).
DPH anisotropy changes induced by the addition of LJ001 (A), JL103 (B) or LJ025 (C) to LUVs of saturated (DMPC) or unsaturated (POPC) phospholipids measured at 37°C, and DPH anisotropy as a function of temperature on DMPC LUVs (D), from which the Tm can be obtained
In order to discard the hypothesis that the smaller changes on anisotropy observed on DMPC using LJ001 or JL103 could be related to a lower production of 1O2, we also evaluated its production, using DMA, in DMPC LUVs. No difference relative to the results obtained for POPC (Figure 3) was observed (results not shown).
To obtain clear evidence of the role for 1O2 on the anisotropy increase, we carried out identical experiments using sodium azide as a 1O2 quencher [35,36]. As it can be seen in Figure 6 for both compounds that are able to produce 1O2 molecules, the addition of sodium azide to POPC LUV suspensions reduces the anisotropy variations, converging to those observed for DMPC. This confirms further that 1O2 production is responsible for the larger increase in anisotropy observed for unsaturated lipids. As expected, no significant changes in anisotropy were found upon the addition of sodium azide to DMPC LUVs (Figure 6). As a control to discard any interference of sodium azide on the membrane interaction of the compounds, we also determined the Kp and quenching data for the different compounds in the presence of sodium azide. We found no significant differences relative to the previous results  without sodium azide, either in the partition or in the final in-depth location of the compounds in the membrane (see Supplementary Figures S1–S3 at http://www.biochemj.org/bj/459/bj4590161add.htm).
Effects of sodium azide (NaN3) on the anisotropy value changes induced by the addition of LJ001 (A) or JL103 (B) to POPC or DMPC membranes at 37°C
In order to complement the anisotropy studies, and with the aim of better understanding the effects that these compounds are able to induce on membrane properties, we studied the changes on the order parameter using the fluorescent probe Laurdan, which is located in a more interfacial position than DPH and TMA-DPH . On the basis of the GPem values determined for POPC LUVs, we found the highest variation for JL103, followed by LJ001 (Figure 7 and Table 1). For LJ025, no changes were observed. When we tested the same effects using DMPC LUVs, we found a similar variation trend, but with a lower magnitude of variation relative to POPC LUVs: JL103 caused a GPem increase (ΔGPem) of 0.76±0.03 in POPC and 0.43±0.04 in DMPC, whereas LJ001 induced ΔGPem of 0.27±0.004 and 0.22±0.003 for POPC and DMPC respectively.
Membrane-ordering effects measured by generalized polarization of Laurdan on POPC (A) or DMPC (B) LUVs
Membrane fusion inhibition
Finally, we evaluated whether the ability of LJ001 and JL103 to inhibit the fusion between the virus and cell membranes [6,7] could be also reproduced using lipid vesicles. For this, we tested directly the ability of the compounds to inhibit membrane fusion using POPC LUVs incubated with HIV-FP. In the control LUV system, this peptide induced a fusion efficiency of 27.9±1.7% (Figure 8A), using a HIV-FP/lipid ratio of 1:10, in a good agreement with published data . Then, when we pre-incubated the unlabelled LUV population with each compound for 30 min, before the lipid mixing assay, we found a significant reduction in the fusion efficiency of JL103 and LJ001 (Figure 8B). When we repeated the experiment in the presence of sodium azide, we observed for both compounds able to produce 1O2, a reduction in the fusion inhibition, reaching values comparable with those obtained for the negative control LJ025, in a good agreement with the previous results. As expected, in the case of LJ025, no changes were observed in the presence or absence of sodium azide.
Membrane fusion inhibition analysis
Membranes and their lipid composition play an essential role in viral entry and, in particular, at the enveloped viruses’ fusion steps. Therefore targeting virus and/or cell membranes may be a promising therapeutic strategy. In the present study, we investigated using different biophysical techniques the interaction of LJ001 and JL103 with biomembrane model systems, in order to clarify their mechanism of action at the molecular level. A specific focus was given to the changes induced on membrane properties by those inhibitors of the fusion between the membrane of enveloped viruses (such as HIV) and the target cell membrane, therefore impairing the entrance of the viral content into the cell. For the sake of comparison, LJ025, a variant of LJ001 without antiviral activity, was studied in parallel.
As reported previously by us , LJ001, LJ025 and JL103 are all able to interact with POPC, a lipid with packing density and fluidity properties similar to those of mammalian cell membranes; however, JL103 exhibits a Kp more than 100-fold lower than LJ001 (Table 1).
When the location of the three molecules in the lipid bilayer was studied through the quenching of their intrinsic fluorescence by acrylamide, 5NS or 16NS, we found that LJ025 locates deep inside the acyl region of the bilayer, LJ001 locates in a shallower position, closer to the polar headgroups of the lipids, and JL103 locates in the most superficial water-exposed region. Furthermore, when we quantified the water exposure by acrylamide quenching, we found that LJ001 and LJ025 are almost completely buried in the lipid bilayer, whereas, for JL103, approximately 50% of the molecules are exposed to the aqueous media (Table 1). Several aspects of bilayer structure and lipid physical properties are able to modulate the fusion process, such as bilayer dehydration , or changes in lipid packing [40,41], membrane fluidity [39,42] or its curvature . It was pointed out previously that LJ001 and JL103 are able to produce reactive oxygen species , with a concomitant effect on membrane fluidity. Thus we tried to establish a clear correlation between 1O2 production and the effects on membrane features, and its impact on the ability to inhibit membrane fusion. With this aim, we studied 1O2 production in the membrane and in a homogeneous medium (methanol), using the fluorescent probe DMA, demonstrating further that JL103 and LJ001 are able to produce 1O2 in both media. As expected, JL103 yielded the highest 1O2 production (approximately 3.3- or 2.2-fold higher DMA fluorescence quenching efficiency than LJ001, in methanol or POPC LUVs respectively). In POPC LUVs, the V parameters obtained reveal an action-sphere (the volume where DMA is able to be quenched by the 1O2 produced) higher for LJ001 than for JL103. This difference may relate to the much more interfacial position of JL103, in comparison with the slightly more internal location of LJ001. It should be noted that, using these experimental conditions, only the DMA buried in the membrane contributes significantly for the fluorescence signal [20,21].
Regarding the mechanism by which LJ001 and JL103 produce 1O2 molecules, it was proposed previously that they act as type II photosensitizers . In this type of agent, the process is associated with the transfer of energy (not electrons) to molecular oxygen .
When we studied the effects on the membrane properties induced by the three molecules using DPH and TMA-DPH, we found a rigidification effect (higher fluorescence anisotropy values), especially for JL103 and LJ001. LJ025 was also able to induce an increase on the anisotropy, but less significant than the two active fusion inhibitors tested (Figure 4). Despite no clear differences being found between LJ001 and JL103, when we plotted the anisotropy values as a function of the real concentration of each compound in the strict bilayer volume (Figures 4D and 4E), we were able to correlate the increase in anisotropy values (approximately double that observed for LJ001) with 1O2 production. The larger changes in anisotropy found with DPH, in comparison with TMA-DPH, for LJ001 and JL103 indicate that the main changes on the fluidity occur in the acyl region of the lipids, in good agreement with the double-bond-attack hypothesis previously proposed . On the other hand, in the case of LJ025, no significant differences were found between both probes.
In order to establish the role of 1O2 production and the changes in the membrane properties, we also ran DPH fluorescence anisotropy measurements with DMPC, a saturated phospholipid, which should be less susceptible to reactive oxygen molecule damage . Using this lipid, we observed a drastic reduction in the rigidification effect, both for JL103 and LJ001; however, in the case of LJ025, no significant changes were observed, as expected from the absence of 1O2 production. Finally, by using sodium azide to annihilate the 1O2 produced, we observed for both molecules able to generate 1O2, LJ001 and JL103, a reduction in the rigidification effect in the case of POPC LUVs, absent when DMPC membranes were used. These observations agree fully with the hypothesis that 1O2 production is involved in the rigidification process, by attacking the unsaturated phospholipid acyl chain double bonds. However, as LJ025, which does not produce 1O2, also induces some rigidification, independently of the presence or absence of unsaturated phospholipids, other processes should also be involved. Nevertheless, the fact that LJ025 showed no antiviral activity indicates that oxidative damage is the critical mechanism for the broad-spectrum antiviral activity of JL103 and LJ001.
In order to complement the anisotropy results, we also determined the order parameter GPem, with both POPC and DMPC LUVs, using the fluorescent probe Laurdan. The sensitivity of Laurdan to the polar region of the bilayer can be used to study the structural changes involving modifications in its local water concentration. Laurdan is tightly anchored in the membrane hydrophobic core, by co-operative van der Waals interactions between the lauric acid tail and the lipids hydrocarbon chains, with its fluorescent moiety residing at the level of the phospholipid's glycerol backbone . In this case, the correlation between 1O2 production and the increase in GPem (a lipid-ordering effect) became clear, in good agreement with previous works, which pointed out that Laurdan may be more sensitive than DPH to assess oxidative damage . It is also important to take into account that JL103 locates in a much interfacial location, which is the specific microenvironment where Laurdan senses the changes on polarity, instead of the more hydrophobic locations of LJ001 and LJ025.
When the ability to inhibit membrane fusion (or hemifusion) was tested, we found a good correlation between 1O2 production and fusion inhibition. JL103 was able to decrease HIV-FP-induced membrane fusion in more than 80%, with LJ001 showing a reduction above 60%. Furthermore, when we repeated these fusion experiments, but now eliminating the 1O2 present, we found that the ability of JL103 and LJ001 to inhibit fusion is reduced to values identical with those obtained for LJ025, confirming once more the close relation of 1O2 production with the ability to inhibit membrane fusion and therefore with their antiviral activity.
Looking into how the 1O2 induces a rigidification in the membrane, we proposed previously  that the increase in anisotropy could be mainly driven by the reaction between 1O2 and acyl chain double bonds. It was reported previously that 1O2 can easily react with cis-double bonds in unsaturated fatty acids through the ene reaction, followed by formation of trans-hydroperoxy groups  (Figure 9). Clustering of these oxidized phospholipids could result in decreased membrane fluidity, which negatively affects the membrane's ability to undergo the extreme membrane curvature transitions occurring during virus–cell membrane fusion [7,48]. All of the data obtained in the present study match and reinforce the idea that the attacks on the double bond present in one of the POPC acyl changes (and absent from the saturated phospholipid DMPC) by the 1O2 molecules is the key factor in the increase in rigidity. Thus photosensitization of viral membranes requires the presence of unsaturated acyl chains and results in a decrease in membrane fluidity, or increased rigidity. It was pointed out previously that changes in membrane fluidity dramatically affect the fusion process [39,42,49]. An increase in membrane rigidity is expected to rise the energy cost of fusion intermediates formed by strongly bent lipid monolayers, thus inhibiting membrane fusion [48,50,51]. All of this may help to explain the mechanism involved in the fusion inhibition caused by the addition of LJ001 or JL103 to membrane.
Hydroperoxy formation by 1O2 molecules 
In contrast with the deeper membrane location of LJ025, the physical locations of LJ001 much closer to the lipid/water interface and JL103 located directly in this interface could eventually difficult or inhibit the transition from positive to negative local curvature necessary for the formation of the viral fusion pore. A study by St Vincent et al.  introduced a class of wedge-shaped RAFIs (rigid amphipathic fusion inhibitors) that seem to block infectivity of unrelated enveloped viruses, apparently through locking the positive curvature of their lipid membranes. It could be reasoned that our compounds could act in a similar way. However, we have demonstrated recently that the mechanism of action of those inhibitors is also mainly driven by 1O2 production (by photoactivation), leading to membrane rigidification, instead of the physical positioning of those compounds in the membrane . Furthermore, the fact that, in the presence of sodium azide, the ability of LJ001 and JL103 to impair fusion is comparable with that of LJ025 allows us to discard the hypothesis that changes in the membrane curvature by the positioning of the compounds in the bilayer are a key factor for membrane fusion inhibition. Even if there is any change on the membrane curvature related with the fusion inhibition, it must also be related to 1O2 molecules production.
It is also noteworthy that the inhibition of membrane fusion by JL103 and LJ001 observed with entire viruses and cells [6,7] was reproduced in the present study using a simple system, including just an unsaturated phospholipid and the HIV-FP. This demonstrates further that the molecular-level target of these broad-spectrum antiviral agents is the membrane lipid themselves, and not another viral or target cell component absent from the simple system tested.
All of the data together allows us to conclude that 1O2 produced by JL103 and LJ001 is the main factor leading to the decrease in acyl changes fluidity, and the concomitant increase in the ordering of the polar headgroups region, resulting in an inhibition of the formation of the fusion pore necessary for cell infection by HIV and other enveloped viruses.
HIV fusion peptide
large unilamellar vesicle
B-PE, rhodamine B 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine
simulated experimental data analysis
Axel Hollmann, Miguel Castanho, Benhur Lee and Nuno Santos conceived and designed the experiments, and wrote the paper. Axel Hollmann performed the experiments. Axel Hollmann and Nuno Santos analysed the data.
We thank Michael E. Jung and Jihye Lee (University of California Los Angeles) for generously providing LJ001, LJ025 and JL103.
This work was funded by Fundação para a Ciência e Tecnologia–Ministério da Educação e Ciência (FCT-MEC, Portugal) [projects PTDC/QUI-BIQ/104787/2008 and VIH/SAU/0047/2011]. A.H. also acknowledges a FCT-MEC fellowship [number SFRH/BPD/72037/2010]. B.L. was supported by the National Institutes of Health [grant numbers U01 AI070495 and U01 AI082100].
Present address: Department of Microbiology, Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, #1124, New York, NY 10029, U.S.A.