3.1. Effect of Hydrophobic Ions on Model Membranes
To test the effect of sterol inclusion within the membrane, we studied the adsorption of hydrophobic ions added asymmetrically (to one side) onto model lipid membranes of different compositions by measuring differences in the boundary potential, ∆Φ
b. The boundary potential difference was measured using the inner field compensation method. The method allows the measurement of the transmembrane voltage arising from asymmetric adsorption of charged molecules onto the membrane [
26,
27,
28]. However, IFC cannot give information about the position of the molecule in the membrane. Thus, apart from IFC, we used an alternative method for determining the boundary potential of BLM after the addition of hydrophobic ions. This method measures the influence of the tested substances on the energy barrier for the transmembrane transport of the ionophore nonactin [
31]. This barrier is mainly determined by the membrane dipole potential and independent of the surface charge density; thus, the latter technique is sensitive to the penetration of substances into the membrane. Comparison of the IFC and nonactin conductance allowed evaluation of the penetration of hydrophobic ions into the BLM.
Changes in the boundary potential were monitored after the addition of increasing concentrations of hydrophobic ions (
Figure 3). The potential stabilized approximately 5 min after the addition. A positive change in the potential corresponds to the binding of positively charged molecules or dipoles oriented with their positive pole toward the membrane interior. To evaluate the possible penetration of hydrophobic ions into lipid membranes, we measured the effect of increasing concentrations of these compounds on the conductance of lipid membranes induced by nonactin. Each point in the plots of
Figure 3 was obtained by averaging the results of 3–6 experiments. The error bars in the plots represent the standard deviation. If the error bars were smaller than the size of the circle in the plot, the error bars were not shown.
In the case of BAC adsorption on DOPC membranes for concentrations slightly above 1 μM, the membranes became very unstable, which was observed as a substantial decrease in the average lifetime of BLM before its rupture. Most likely, the overall instability of the membranes was responsible for large error bars in the plot (
Figure 3C). Interestingly, the presence of cholesterol in the membrane almost eliminated such effects on ∆Φ
b induced by BAC adsorption (
Figure 3D). This result means that cholesterol increases the membrane ordering, suppressing the disturbance of lipid packing resulting from incorporation of BAC into the lipid membrane. The hypothesized membrane ordering upon cholesterol addition is in accordance with the data obtained from molecular modeling, as described in the next section. Kor105 showed adsorption in a concentration-dependent manner for both membrane compositions used in this study (
Figure 3A,B). The presence of 40 mol.% cholesterol in the membrane significantly enhanced the adsorption of Kor105 at concentrations above 0.4 μM, as demonstrated by higher positive values of the boundary potential difference (
Figure 3B). Most likely, the difference in adsorption of BAC and Kor105 is due to the fact that in Kor105 the aromatic ring is directly linked to nitrogen, which makes the ring less mobile, as compared to BAC (
Figure 1).
To evaluate the penetration of BAC, Kor105, and SDS into and through lipid membranes formed from DOPC and the DOPC/Chol 3:2 mixture, we measured the effect of increasing concentrations of these compounds on the conductance of lipid membranes induced by nonactin. We observed that Kor105, BAC, and SDS did not manifest penetration through the BLMs for both membrane compositions (
Figure 3). For all tested compounds, the boundary potentials determined by IFC and nonactin conductance were almost the same within the experimental error, meaning that Kor105, BAC, and SDS neither pass through the membrane nor crucially disturb the monolayer dipoles. In the case of Kor105, the absence of cholesterol significantly decreased adsorption onto the membrane, in contrast to BAC and SDS, where cholesterol addition had almost no effect on adsorption determined by both IFC and nonactin conductance methods. In the presence of cholesterol, all three tested ions yielded almost the same magnitude of the boundary potential difference within the experimental error.
In
Figure 3, it was essential to demonstrate boundary potential curves obtained by IFC and nonactin conductance methods on the same plot, as their difference can point to possible penetration or strong disturbance of the membrane by the compounds. However, the question of how the presence of sterol changes membrane properties with respect to its interaction with different surfactants is very interesting. Thus, we combined the data of the
Figure 3 to present DOPC vs. DOPC/Chol curves in the same graph (
Figure 4).
From the plots, it follows that addition of cholesterol increases adsorption of Kor105 onto the membrane, as determined by both IFC and nonactin conductance methods (
Figure 4A,B). On the contrary, cholesterol slightly hampers adsorption of SDS, although this conclusion is supported reliably only by IFC measurements (
Figure 4E). Taking into account the experimental errors, cholesterol does not influence adsorption of BAC (
Figure 4C,D).
In summary, the experiments on BLMs demonstrated that all three tested compounds, Kor105, BAC, and SDS, adsorb onto the membranes of both DOPC and DOPC/Chol 3:2 compositions, yielding substantial boundary potentials. For the compounds, the boundary potentials determined by IFC and nonactin conductance methods were very similar, indicating the ions neither passed through the membrane nor crucially disturbed the monolayer dipoles. Cholesterol substantially increased adsorption of Kor105, slightly hampered adsorption of SDS, and had almost no effect on adsorption of BAC. In the presence of cholesterol, all three tested ions yielded almost the same magnitude of the boundary potential difference within experimental error.
3.2. MD Modeling of Membranes with Incorporated Hydrophobic Ions
To test the preliminary conclusions on the interaction of hydrophobic ions with model BLMs, we attempted molecular modeling. Membranes of 12 different compositions were modeled: (1) 200 DOPC, one Kor105 (lowKorDOPC); (2) 120 DOPC, 80 Chol, one Kor105 (lowKorCholDOPC); (3) 202 DOPC, 86 Kor105 (highKorDOPC); (4) 122 DOPC, 80 Chol, 86 Kor105 (highKorCholDOPC); (5) 200 DOPC, one BAC (lowBacDOPC); (6) 120 DOPC, 80 Chol, one BAC lowBacCholDOPC; (7) 202 DOPC, 86 BAC (highBacDOPC); (8) 122 DOPC, 80 Chol, 86 BAC (highBacCholDOPC); (9) 200 DOPC, one SDS (lowSDSDOPC); (10) 120 DOPC, 80 Chol, one SDS (lowSDSCholDOPC); (11) 202 DOPC, 86 SDS (highSDSDOPC); (12) 120 DOPC, 80 Chol, 86 SDS (highSDSCholDOPC). Each simulation box was filled with at least 10,000 water molecules. The bilayers were analyzed through a 500-ns MD run at constant temperature and pressure.
It appeared that, in lipid membranes, SDS conformation was linear with sulfate located in the region of the lipid phosphate groups. Both BAC and Kor105 adopted a kinked conformation. In this conformation, the N
+ atom was located in the region of the lipid phosphate groups, the alkyl tail was buried in the membrane interior, and the aromatic ring was positioned in the region of lipid ester bonds, i.e., at the membrane polar–hydrophobic interface (
Figure 5).
Let us denote the vector directed from the N
+-atom toward the C-atom in the
para position of the aromatic ring in the hydrophobic ion as
h, the vector directed from the N
+-atom or S-atom toward the terminal C-atom in the alkyl tail in the hydrophobic ion as
t, and the vector of the unit external normal to the membrane plane as
n (
Figure 6).
The orientation of hydrophobic ions was characterized by two angles: (1)
η, the angle between
n and
h; (2)
θ, the angle between
n and
t (
Figure 6). From the MD trajectories, the distributions of the angles
η and
θ were analyzed both for the single molecules of the hydrophobic ions (systems lowKorDOPC, lowKorCholDOPC, lowBacDOPC, lowBacCholDOPC, lowSDSDOPC, lowSDSCholDOPC) and for their ensembles (systems highKorDOPC, highKorCholDOPC, highBacDOPC, highBacCholDOPC, highSDSDOPC, highSDSCholDOPC).
In pure DOPC membranes, the alkyl tails of both hydrophobic ions were highly dynamic; they were irregularly bent, and the chain was randomly oriented with respect to the average normal to the membrane surface (
Figure 5A,C,E). In the presence of cholesterol, the tails were more ordered; they were almost straight and oriented nearly perpendicular to the membrane surface (
Figure 5B,D,F). The density of the distribution of angles
η and
θ is shown in
Figure 7 for both single molecules of hydrophobic ions (systems lowKorDOPC, lowKorCholDOPC, lowBacDOPC, lowBacCholDOPC, lowSDSDOPC, lowSDSCholDOPC) and their ensembles (systems highKorDOPC, highKorCholDOPC, highBacDOPC, highBacCholDOPC, highSDSDOPC, highSDSCholDOPC). As SDS lacks the aromatic ring, only the distribution of
θ angle was plotted for this ion (
Figure 7E,F).
In cholesterol-free membranes, the angle η between the vector normal to the membrane surface and the plane of the aromatic ring was approximately 130° in all modeled systems, meaning that the molecule of the hydrophobic ion in such a membrane is strongly bent around its N+ atom (θ ≈ 150°; thus, the angle between h and t vectors is approximately θ − η ≈ 20°), and the aromatic ring is buried deeply into the hydrophobic core of the lipid monolayer. From the plots, especially those built for single molecules, it follows that cholesterol on average decreases η and increases θ, which means that the hydrophobic ion molecule partly unbends in the presence of cholesterol, as the difference (θ − η) increases. The ion alkyl tail aligns with the average direction of the lipid tails, which follows from the increase of θ.
Note that, for a single BAC molecule in both cholesterol-free and cholesterol-containing membranes, the distributions of
η and
θ angles are very wide (
Figure 7C). However, at high BAC concentrations, the distributions become relatively narrow, meaning that the order of the membrane is increased upon incorporation of more BAC molecules (
Figure 7D). For Kor105, the distributions of
η and
θ angles have comparable widths for both the single molecule and the ensemble of the hydrophobic ion (compare
Figure 7A,B). For SDS the distributions of
θ angle are relatively narrow in both DOPC and DOPC/Chol membranes, as well as for a single SDS molecule and the ensemble (
Figure 7E,F).
To analyze the depth of incorporation of hydrophobic ions into the membranes, we considered the distribution of electrical charge density along the direction of the normal vector to the membrane surface,
n, i.e., along the
Oz axis (
Figure 8). In the cholesterol-free membrane, the positive charges of BAC and Kor105 are located predominantly in the region between the phosphorus and nitrogen of DOPC, independent of the ion type (
Figure 8A,C). The addition of cholesterol results in a smaller average charge density of lipids, as cholesterol does not bear any charge and merely dilutes the charges located on nitrogen and phosphorus atoms of DOPC, leading to smaller peaks (compare the amplitudes of the peaks of the green curves in
Figure 8A,C,E, as well as in
Figure 8B,D,F). Besides, cholesterol addition leads to an increase in the membrane thickness, as the nitrogen and phosphorus atoms of DOPC both shift by approximately 0.4 nm in the direction from the membrane center (compare the location of the positive and negative peaks of the green curves in
Figure 8A,C,E, as well as in
Figure 8B,D,F). In particular, the negative peak corresponding to phosphorus atoms of DOPC shifts from
z = ±1.8 nm to
z = ±2.2 nm. Interestingly, the position of the positive peak, corresponding to the N
+ atom of the hydrophobic ion, does not change upon cholesterol addition (compare the position of the positive peaks of the blue curves in
Figure 8A,C, as well as in
Figure 8B,D). Although charged atoms of DOPC do shift, the N
+ atom remains at
z = ±2.5 nm for both Kor105 and BAC, independent of the cholesterol content in the membrane. The effective result is that, in cholesterol-containing membranes, the hydrophobic ion is buried deeper by approximately 0.4 nm into the lipid monolayer compared to the case of membranes formed from pure DOPC. In cholesterol-free membranes, the positive peak is relatively high in the case of high concentrations of hydrophobic ions; nevertheless, the zone of negative electric charge is still detectable at
z = ±1.8 nm (
Figure 8A,C). In cholesterol-containing membranes, the positive peak is lower due to dilution of electric charges by the molecules of electrically neutral sterol, but the negative peaks at
z = ±1.8 nm completely disappear (
Figure 8B,D).
Notably, the MD simulations suggested very similar behavior of both BAC and Kor105 in cholesterol-containing membranes (
Figure 7 and
Figure 8). However, the data presented in
Figure 3 and
Figure 4 clearly indicate the difference between the effects of cholesterol on BAC and Kor105 interaction with membranes (
Figure 4, A vs. C and B vs. D). Most likely, this difference is due to the fact that, in Kor105, the aromatic ring is directly linked to nitrogen, which makes the ring less mobile, as compared to BAC. This means that Kor105, being relatively more rigid, disturbs the lipid monolayer more strongly than BAC. This disturbance is more pronounced in the membranes rich in cholesterol, as cholesterol orders the Kor105 hydrophobic tail, thus imposing a stricter orientation of the aromatic ring.
Quite surprisingly, the negatively charged sulfate group of SDS is located in the region of phosphorus atoms of lipids (
Figure 8E,F). Such localization appears to be independent of cholesterol content. Addition of cholesterol shifts the negative peak of phosphorus atoms of the lipids from the center of the membrane (compare locations of negative peaks of green curves in
Figure 8E,F). The negative peak of SDS sulfate group shifts in the same direction by the same distance, from
z = ±1.8 nm to
z = ±2.2 nm (compare locations of negative peaks of blue curves in
Figure 8E,F). This behavior is the opposite to that demonstrated by BAC and Kor105; the location of their nitrogen atoms with respect to membrane center did not change upon addition of cholesterol, although phosphorus atoms of lipids shifted. The behavior of SDS can be explained by it generally preferring to bury its sulfate in the region of positively charged choline groups of the lipids. However, this would place SDS hydrophobic tails near to the polar region of the lipid phosphate groups, leading to a high energy penalty. Burying the SDS deeper into the lipid monolayer would lead to contact of lipid hydrophobic tails with electric charges of the SDS sulfate group. It appears that, in such a situation, it is energetically favorable to minimize the contact between polar and hydrophobic groups, despite the electrostatic repulsion between the neighboring sulfate of SDS and phosphate groups of the lipids.
For all modeled systems, the molecules were distributed evenly between the monolayers, and laterally randomly along each monolayer. However, in the course of the MD runs, we observed that cholesterol formed small clusters. We analyzed the size distribution of cholesterol clusters using the last 100 ns of the trajectories for the analysis; the number of clusters was calculated at each integration step. The cholesterol molecules were considered as clustered if the distance between their O atoms was smaller than 0.7 nm. The size distribution of cholesterol clusters is presented in
Figure 9. It was found that the size distribution depended on the concentration of the ion; higher concentrations yielded smaller clusters. This effect can be explained by favorable interaction of saturated alkyl chains of the ions with cholesterol, leading to disruption of cholesterol clusters and heterodimerization of the ion and cholesterol molecules. However, simple dilution of cholesterol by the ions cannot be excluded. Each modeled system contained approximately the same number of cholesterol molecules. Thus, for a high concentration of hydrophobic ions, the molar fraction of cholesterol was lower compared to the membranes with a low concentration of hydrophobic ions. Such a dilution might be the main reason responsible for the smaller size of the clusters at high concentrations of hydrophobic ions.
Generally, no redistribution of ions (BAC, Kor105, or SDS) between the lipid monolayers was observed during 500 ns of each simulation trajectory. The flip-flop of the cholesterol was also absent with some exceptions. For the system highBacCholDOPC composed of 122 DOPC, 80 Chol, and 86 BAC molecules, cholesterol molecules twice plunged into the bilayer, laid at the monolayer interface for 4 ns and 16 ns (
Figure 10), and returned to the same monolayer. For the system highKorCholDOPC composed of 122 DOPC, 80 Chol, and 86 Kor105 molecules, cholesterol molecules several times plunged into the bilayer, but did not reach the monolayer interface, and immediately returned to the same monolayer.
MD modeling showed that the presence of cholesterol increases the ordering of the lipid membrane, leading to a greater thickness of the lipid bilayer. The increased bilayer thickness does not change the position of the N+ atom of BAC or Kor105 with respect to the monolayer interface. Thus, cholesterol addition results in effectively deeper penetration of BAC and Kor105 into the lipid monolayer as related to the positions of the lipid headgroups. On the contrary, for SDS, it was shown that, upon addition of cholesterol, the location of the SDS sulfate group follows the position of phosphates of the lipids; as the lipid monolayer becomes thicker, the sulfate group also shifts from the bilayer center, always being co-localized with the phosphates. The MD simulations showed a principal difference between the interactions of SDS and the cationic surfactants with the membranes. For all considered ions, the positions of the charged groups are close to the negatively charged phosphate groups of DOPC. This means that, due to electrostatic interactions, the cationic surfactants are significantly more membranophilic than anionic SDS. Thus, although the tested model membranes are electrically neutral, their interaction with hydrophobic ions depends on the sign of the charge of the ion. Another principal difference comes from the fact that, while the SDS molecule is essentially linear, the BAC and Kor105 molecules are not; their aromatic heads are wider that the tails. For this reason, the membrane disturbance caused by the cationic surfactants is sensitive to cholesterol; this neutral lipid dampens the surfactant-induced misalignment of the hydrophobic tails of DOPC. On the contrary, within the range of tested concentrations, SDS does not alter the packaging of the hydrophobic tails, which explains why cholesterol presence does not significantly affect SDS–membrane interaction, as shown by the experiments on BLM.
3.3. Experiments on Yeast Cells with Deletion of PM Sterol Transporters
Our results suggested that the sterol content of the PM strongly influences the effects of benzalkonium and Kor105 on cell physiology. To test this hypothesis, we used yeast cells harboring mutations in
Lam genes, which encode PM sterol transporters. Lam1–4 proteins were shown to transport ergosterol, the main yeast sterol, from the PM to the ER (reviewed in References [
22,
38]). Their deletions are believed to affect either the ergosterol content of the PM, its distribution within the PM, or both [
20,
21]. There are several lines of evidence pointing to the
Lam mutations leading to an increase in PM ergosterol. Firstly, Lams are passive sterol transporters and the PM is the richest cellular compartment in terms of sterol concentration. Secondly, delta-Lams are sensitive to amphotericin B [
20]. Amphotericin B is a drug which forms ion channels in ergosterol-containing but not in ergosterol-free plasma membranes [
39]. Thus, increased amphotericin B sensitivity can be attributed to an increased ergosterol content. Thirdly, it was demonstrated that the deletion of mammalian
Lam homologs leads to a decreased flow of cholesterol from the PM inside the cells [
23].
We constructed three
Lam deletion mutants:
△Lam1△Lam3,
△Lam2△Lam4, and quadruple
△Lam1△Lam2△Lam3△Lam4 mutants. Firstly, we found that deletions do not strongly affect the growth rate of the cells, at least when the cells are grown in a standard rich medium (
Figure 11A).
Next, we compared the resistances of the control strain and the deletion mutants to SDS, benzalkonium, and Kor105. The cells were grown in rich liquid medium in the presence of the indicated concentrations of the surfactants. The drug resistance was estimated by measuring the growth rates (
Figure 11). The double deletion
△Lam1△Lam3 had no significant effect on the resistances (
Figure 11B–D). Furthermore, the
△Lam2△Lam4 and
△Lam1△Lam2△Lam3△Lam4 deletion strains were more sensitive to SDS than the control strain (
Figure 11B). In contrast, these two mutant strains were less sensitive to the cationic surfactants than the control (
Figure 11C,D).
Together, the data presented in
Figure 11 indicate that even relatively minor alterations (most likely an increase) in the PM sterol can cause a major change in the resistance to the surfactants. Also, our data show that the same alterations in the PM sterol content can change the resistance to SDS and the cationic surfactants in opposite directions. Our finding that
△Lam mutants are more sensitive to SDS than the wild type is not surprising; deletion mutations tend to make the cells generally weaker and, thus, more sensitive to a variety of poisonous substances. For the same reason, an increase in the resistance caused by even the quadruple mutation requires a more specific explanation. We reason that an increase in ergosterol levels in the PMs of
△Lam mutants can explain their increased resistance to BAC and Kor105. Indeed, the MD showed that cholesterol reduces the disturbance to the lipid packaging caused by these surfactants (but not that caused by SDS, which seems to be not influenced by sterols at all).