Study of the solubilisation process of bacterial model membranes induced by DDAO

Increasing number of bacteria resistant against conventionally used antibiotics poses a serious health threat for the society. Conventional antibiotics have a specific target within a bacterial cell. Many bacteria have developed resistance to these specific mechanisms of action, rendering the antibiotics no longer effective. Particularly dangerous are the bacteria that are resistant to several conventional antibiotics (Pontali et al., 2013; Xu et al., 2014). In order to overcome multi-drug-resistant bacterial infections, research of substances targeting bacterial cells non-specifically is essential. An example of such substances are long-chain amphiphilic surfactants, such as the one used in the presented study.

N,N-dimethyl-1-dodecanamine-N-oxide (DDAO) is a surfactant with 12 carbon hydrophobic chains and a highly polar N-O group. It is widely commercially used in home cleaning detergents, personal hygiene products and pharmaceutical formulations as well (Singh et al., 2006). Within the research applications, DDAO is used to reconstitute, crystallise and purify membrane proteins, as well as to solubilise membranes, which is the focus of this study. DDAO is an amphiphilic molecule with pK ~ 5, meaning that at physiological pH, DDAO is in a non-ionic form (Búcsi et al., 2014; Herrmann, 1962). DDAO manifests various biological effects, such as antimicrobial (Devínsky et al., 1990) and antiphotosynthetic (Murín et al., 1990) activity or phytotoxic (Šeršeň et al., 1992) and immunomodulatory (Bukovský et al., 1996) effect. In contact with the biological membrane, DDAO interacts mostly with its phospholipids (Kragh-Hansen et al., 1998). DDAO molecules incorporate themselves into the bilayer, influence physical properties of bilayer, and induce formation of defects and pores (Memoli et al., 1999; Ruiz et al., 1988), followed by complete disintegration of the bilayer after a certain concentration of surfactant is reached (Lichtenberg et al., 2013a; Lichtenberg et al., 2013b). For the interaction studies, a binary and ternary mixture of phospholipids was used. Interaction of DDAO with model membranes of various phosphatidylcholines (PCs) or PCs mixed with cholesterol (lipids typical for the mammalian model membranes) was studied in several papers (Belička et al., 2014a; Belička et al., 2014b; Hrubšová et al., 2003; Huláková et al., 2013, 2015; Karlovská et al., 2004a; Karlovská et al., 2004b; Uhríková et al., 2001). The aim of this study was to determine how different is the interaction of DDAO with bacterial model membranes. For most bacteria, phosphatidylethanolamines (PEs) are the most predominant phospholipids. All bacteria have at least 15% of anionic phospholipids such as phosphatidylglycerols (PGs) or cardiolipins (CLs) (Epand & Epand, 2009). PEs are zwitterionic phospholipids with a molecular shape of a truncated cone. Their polar headgroup is smaller in diameter in comparison with its hydrocarbon chain area. Therefore, they tend to form structures with negative curvature, which leads to the formation of non-lamellar phases at higher temperatures. PGs at physiological conditions are negatively charged, and because of their cylindrical molecular shape, they tend to form non-curved bilayers, even at higher temperatures. CLs have a double negative charge at physiological pH and relatively rigid (regarding mobility of the molecule) cylindrical molecular shape (Lopes et al., 2012). As a model organism for our experiments, G⁻ bacteria Escherichia coli (E. coli) were selected. Membrane lipids of E. coli are a mixture of PEs (57%), PGs (15%), CLs (10%) and other lipids (18%) (such mixture of lipids is commercially available as E. coli Total Lipid Extract). Frequently used models are two-component PE and PG mixtures (Lopes et al., 2010). The fatty acids found in the E. coli membrane phospholipids are predominantly 16–18 carbon long, with a substantial portion of them being unsaturated (Meister et al., 2014). For this reason, we chose to use a mixture of palmitoyloleoylphosphatidylethanolamine (POPE) and palmitoyloleoylphosphatidylglycerol (POPG) as our model system, in a molar ratio of POPE/POPG = 0.6/0.4 mol/mol. Gel phase–liquid-crystalline phase transition of this mixture is approximately at 20°C (Lopes et al., 2010; Pozo Navas et al., 2005). The temperature during the experiment was well-maintained at 25°C, ensuring our liposomes are in a liquid-crystalline phase.

An extract of polar lipids of E. coli (commercially prepared from total lipid extract by precipitation with acetone and extraction with diethyl ether) contains PEs (67%), PGs (23%) and CLs (10%). The POPE/POPG/CL three-component mixture (0.67:0.23:0.1) shows the highest similarity with the E. coli extracts based on phase transitions (Lopes et al., 2010). Despite CLs having 4 fatty acids in its molecules (in our case, 4 oleic acids), the dominant types of cardiolipin usually contain only one or two kinds of fatty acids. It was found that the fatty acid composition is relatively resistant to dietary manipulations, suggesting a specific composition is needed for its biological functions (Schlame et al., 2005). Tetraoleoylcardiolipin has a gel to liquid-crystalline phase transition at approximately −15°C (Chen, 2012). Therefore, we assume that our ternary mixture was in liquid-crystalline phase during the experiment as well.

In the present paper, the solubilising effect of a surfactant DDAO on two bacterial model membranes was studied using static light scattering (nephelometry). Because of the amphiphilic nature of the DDAO surfactant and relatively high amount of anionic phospholipids in model membranes, all experiments were performed in phosphate-buffered saline (PBS, pH 7.5).

Solubilisation of POPE-POPG and POPE-POPG-TOCL ULLs induced by non-ionic surfactant DDAO was studied at five different lipid concentrations. Decrease of the lipid particle sizes, which is typical for the solubilisation process, was examined using nephelometry. We can see examples of the experiments in Figure 1. The inset of Figure 1 shows raw solubilisation data for POPE-POPG-TOCL membranes at three lipid concentrations. We can see that the method is sensitive not only for the changes in the size of the scattering particles, but for their concentration as well. The main part of Figure 1 shows a solubilisation experiment comparison of the two types of membrane we investigated, at the same lipid concentration 0.1 mmol/dm³. Intensity of the scattered light was normalised (to the highest detected intensity) for better comparison. All dependencies follow the three-stage process described earlier. All dependencies were fitted with bilinear functions (Equation 1) and reverse sigmoid function (Equation 2) to obtain critical concentrations D_t^SAT, D_t^MID, D_t^SOL. An example of the fitting functions can be seen in Figure 2 (critical concentrations are indicated by arrows).

All acquired D_t^SAT, D_t^MID, D_t^SOL values are plotted as a function of concentration of the lipid in Figure 3. We can see their values increase linearly with increasing concentration of lipid for both types of membrane. The only exception is the saturation concentration D_t^SAT for POPE-POPG membranes. The reason is that for POPE-POPG membranes, the intensity of scattered light in the first phase of solubilisation was increasing, as opposed to staying approximately the same, as was the case for POPE-POPG-TOCL membranes (an example can be seen in Figure 1). Surfactant-induced increase in the particle size has been reported in literature (Kragh-Hansen et al., 1998). The mentioned study shows that DDAO induced a fusion of small ULLs to larger vesicles before the transition into micelles. The ULLs interacting with DDAO were composed of sarcoplasmic reticulum lipid extract (lipid composition was not specified), and the buffer in use contained 0.1 mmol/dm³ CaCl₂. Binding of Ca²⁺ ions to the membrane are known as an important factor involved in the membrane fusion (Martens & McMahon, 2008). Our buffer did not contain such ions, and the ULLs lipid composition was different as well. The reason for such increase in intensity needs to be studied further. The increase in the intensity made the fitting of D_t^SAT less reliable, as we can see the error bars (Figure 3) are greater than for any other parameter. Nevertheless, this parameter was not excluded from further analysis, because doing so have not provided significantly different results. All three critical concentrations were fitted with the global function (Equation 3, Figure 3) as described earlier. Instrumental weighing was applied during the fitting, because it uses the square of the reciprocal of the error values, so points with smaller error values have more weight. Results of the fitting are shown in Table 1.

D_W represents the concentration of surfactant in the water phase at particular stage of the solubilisation process. D_W values were obtained by extrapolation of the global function to zero lipid concentration (see Figure 3, where D_w^SAT is depicted as an example). D_W values (D_w^SAT, D_w^MID, D_w^SOL) were obtained as a y-coordinates of the intersection points of the global function with the y-axis. The critical micellar concentration of DDAO at 27°C in its non-ionic form (at pH >7) is 2.1 mmol/dm³ (Herrmann, 1962). All our calculated values of D_W are smaller than this value, which means that DDAO molecules did not form micelles during the solubilisation experiment. Therefore, we propose that the solubilisation process in this study took place by the transbilayer mechanism (Kragh-Hansen et al., 1998).

The effective molar ratio (R_e) of the amount of DDAO integrated into the bilayer to the amount of lipid is a constant, independent of the concentration of lipid and specific for surfactant–lipid mixture. The R_e (we will be using values for D_t^MID concentration for comparisons) values were 4.6 ± 0.3 for POPE-POPG and 6.3 ± 0.3 for POPE-POPG-TOCL. We have determined the partition coefficient of DDAO in bilayers consisting of POPE-POPG K_p = 5,300 ± 400 and for the POPE-POPG-TOCL K_p = 6,500 ± 500.

Regarding the difference between POPE-POPG and the cardiolipin-containing model membrane, the latter was more stable against the DDAO-induced solubilisation. As we can see in Table 1, the R_e values were higher at all three evaluated critical concentrations. The calculated partition coefficient was higher as well. Assuming it is caused purely by the presence of cardiolipin would be premature. Domain formation in bacteria membranes can also play an important role. These domains, reported to be enriched in cardiolipin, have been suggested to play an important role in regulatory functions of the cell (Epand & Epand, 2009 and references therein). Many studies with bacterial model membranes tend to neglect the importance of the presence of cardiolipin in the model membranes. Further investigation is needed, because it appears cardiolipin is an important part of bacterial membrane, even though its function is not yet completely known. We would like to continue the research of cardiolipin-containing model membranes in the future, for example using fluorescence microscopy with the help of cardiolipin specific dye 10-N-nonyl acridine orange.

The partition coefficients and R_e values calculated in the present study are greater than have been reported for the interaction of DDAO with mammalian model membranes. For example, in our recent article (Želinská et al., 2020), we have researched the interaction of DDAO with dioleoylphosphatidylcholine (DOPC) and cholesterol (CHOL)-enriched DOPC liposomes (33 mol% of CHOL). The R_e values for the D_t^MID concentration were 2.2 ± 0.3 in the case of DOPC ULLs and 1.8 ± 0.3 for DOPC-CHOL type of membrane. Also, the calculated partition coefficients were reported smaller: 2,300 ± 400 (for DOPC) and 2,100 ± 600 (for DOPC-CHOL).

Other studies reported smaller molar partition coefficients for mammalian model membranes as well. For example, Hrubšová et al. (2003) calculated K_p = 500 ± 200 in a system egg yolk phosphatidylcholine (EYPC) ULLs/water. R_e for the D_t^MID concentration was 0.6 ± 0.2. MLLs of EYPC were reported to have a molar partition coefficient of DDAO equal to 1,200 ± 400 (Karlovská et al., 2004a). In these two studies, only a dependence of D_t^MID on the lipid concentration was used for the calculations.

Our data and calculations show us that the partitioning of DDAO between the aqueous phase and the bacterial model membranes is higher than it is in the case of mammalian model membranes, and yet, the bacterial ULLs needed more DDAO to become solubilised. Because the hydrophobic part of the studied phospholipids is very similar, we suggest that the reason for such big differences might be caused by the polar parts of the phospholipids. In comparison with PCs, the polar part of PEs has smaller diameter (the choline group in PCs is bigger than the ammonium group) and binds smaller amount of water molecules. The ammonium group is able to form hydrogen bonds with phosphate and oxygen atoms of adjacent molecules. A molecular dynamics simulation (Murzyn et al., 2005) of POPE-POPG (in the proportion 3:1) with Na⁺ ions (to neutralise the negative charge of POPG) has shown that POPE molecules interact readily with POPG and other POPE molecules with hydrogen bonds and water bridges. These bonds strengthen interlipid contacts in the bilayer. These physico-chemical properties of PCs and PEs–PGs-containing membranes might be the reason for our findings, but further analysis is needed. Therefore, in the future, we would like to widen this research to monitoring the effect of DDAO on fluidity of the membrane and the process of pore formation using fluorescent probes.