Probing the Properties of Lipopolysaccharide Monolayers and Their

6970

Langmuir 2005, 21, 6970-6978

Probing the Properties of Lipopolysaccharide Monolayers and Their Interaction with the Antimicrobial Peptide Polymyxin B by Atomic Force Microscopy Stefanie Roes,† Ulrich Seydel, and Thomas Gutsmann* Department of Immunochemistry and Biochemical Microbiology, Division of Biophysics, Research Center Borstel, Leibniz-Center for Medicine and Biosciences, Parkallee 10, D-23845 Borstel, Germany Received July 15, 2004. In Final Form: October 29, 2004 In contrast to the majority of all known cell types, Gram-negative bacteria have a second membrane, the outer membrane, which is an asymmetric bilayer composed of a phospholipid inner leaflet and a glycolipid outer leaflet. The glycolipid layer, in most cases being composed of a lipopolysaccharide (LPS), is the first target for antimicrobial agents. To get a basic understanding of the membrane-forming properties of LPS, we reconstituted monolayers of deep rough mutant LPS from Salmonella enterica serova Minnesota (R595 LPS), its lipid A moiety, and of the synthetic tetraacyl compound 406 (resembling the biosynthetic lipid A precursor IVa) at the air-water interface of a film balance. The liquid-expanded (LE) and liquidcondensed (LC) domains in the coexisting region were investigated with epifluorescence and, after transferring the monolayer onto mica, as a Langmuir-Blodgett film, with atomic force microscopy (AFM). The fluorescence and the AFM images showed identical domain structure. The higher resolution of the AFM images, however, contained more topographic details. Different heights and adhesion forces between the LE and LC domains could be observed. Differences in the adhesion forces between the AFM tip and the sample were determined in the repulsive and the attractive dynamic scanning modes, demonstrating the importance of a careful interpretation of height images. We propose that an increase in the lateral pressure causing the LE-LC transition of the monolayers leads to a reorientation of the molecules due to a tilt angle between the alkyl chains and the diglucosamine backbone. LPS monolayers have been utilized as a simplified reconstitution model of the outer membrane to study the interaction with antimicrobial agents. We investigated the action of the polycationic peptide polymyxin B (PMB) and found dramatic influences on the domain structures.

Introduction Biological membranes are composed of lipids, glycolipids, proteins, and glycoproteins, which are known to organize in lateral domains in membranes. The separation and domain formation is thought to play an important role in the functional specialization of specific complexes. The formation of domains in lipid membranes can arise from phase separations of (i) lipids and peptides/ proteins,1-3 (ii) different lipids,4,5 and (iii) one lipid in different phases, e.g., liquid-expanded (LE) and liquidcondensed (LC) phases.6 It is being discussed that proteins of specific signal transduction pathways are combined in domains to form signaling complexes.7 In many cases, these domains are cholesterol-rich and termed rafts.8,9 Furthermore, it has been shown that externally applied * Author to whom correspondence should be addressed. Tel: +49-4537-188291. Fax: +49-4537-188632. E-mail: [email protected]. † Present address: University of Cambridge, Department of Pharmacology, Tennis Court Road, CB2 1PD Cambridge, United Kingdom. (1) Mou, J.; Czajkowsky, D. M.; Shao, Z. Biochemistry 1996, 35, 32223226. (2) Rinia, H. A.; Kik, R. A.; Demel, R. A.; Snel, M. M.; Killian, J. A.; Der Eerden, J. P.; de Kruijff, B. Biochemistry 2000, 39, 5852-5858. (3) Janshoff, A.; Bong, D. T.; Steinem, C.; Johnson, J. E.; Ghadiri, M. R. Biochemistry 1999, 38, 5328-5336. (4) Giocondi, M. C.; Pacheco, L.; Milhiet, P. E.; Le Grimellec, C. Ultramicroscopy 2001, 86, 151-157. (5) Yuan, C.; Johnston, L. J. J. Microsc. 2002, 205, 136-146. (6) Hollars, C. W.; Dunn, R. C. Biophys. J. 1998, 75, 342-353. (7) Hoessli, D. C.; Ilangumaran, S.; Soltermann, A.; Robinson, P. J.; Borisch, B.; Nasir, U. D. Glycoconjugate J. 2000, 17, 191-197. (8) Lawrence, J. C.; Saslowsky, D. E.; Edwardson, J. M.; Henderson, R. M. Biophys. J. 2003, 84, 1827-1832.

molecules can induce changes in domain structures10 and that proteins preferentially intercalate into the boundary region of LE and LC domains.11 In the past decade, many groups investigated domains in lipid mono- and bilayers using AFM (for reviews, see refs 12-14). The advantages of using atomic force microscopy (AFM) are that the imaging capacity spans about 5 orders of magnitude (1 nm to 100 µm), imaging under near-physiological solution conditions is possible, and the mechanical properties of the sample can be investigated. In this work, we focused on the characterization of monolayers composed of bacterial lipopolysaccharides and their interaction with the antibiotic polymyxin B (PMB). Lipid bilayers are closer to the natural membrane than monolayers; however, monolayers provide some advantages such as the precise measurement of area requirements of moleculesslipids, as well as intercalated antibiotics. Moreover, the access to the fatty acid chains may provide valuable information on their organization. In earlier experiments, investigating the interaction of antimicrobial peptides with lipid membranes, we showed a good agreement between results obtained from film balance and those from bilayer experiments.15 Thus, the monolayer is the most simplified membrane system for (9) Fielding, C. J.; Fielding, P. E. Biochim. Biophys. Acta 2003, 1610, 219-228. (10) Gutsmann, T.; Fix, M.; Larrick, J. W.; Wiese, A. J. Membr. Biol. 2000, 176, 223-236. (11) Ruano, M. L.; Nag, K.; Worthman, L. A.; Casals, C.; Perez-Gil, J.; Keough, K. M. Biophys. J. 1998, 74, 1101-1109. (12) Dufrene, Y. F.; Lee, G. U. Biochim. Biophys. Acta 2000, 1509, 14-41. (13) Rinia, H. A.; de Kruijff, B. FEBS Lett. 2001, 504, 194-199. (14) Janshoff, A.; Steinem, C. Chembiochem. 2001, 2, 798-808.

10.1021/la048218c CCC: $30.25 © 2005 American Chemical Society Published on Web 04/16/2005

AFM of Lipopolysaccharide Monolayers

Langmuir, Vol. 21, No. 15, 2005 6971

in most of the experiments in this paper because it is the membrane-forming moiety of LPS consisting of the hydrophilic disaccharide backbone to which the fatty acids are bound. Furthermore, it represents the “endotoxic principle” being responsible for the induction of the immune response of the host.20 ReLPS is the shortest LPS synthesized by Gram-negative bacteria, and the synthetic compound 406 resembles the biosynthetic lipid A precursor IVa. The formation of LE/LC domains in the coexisting phase was observed in epifluorescence images of the monolayer at the air/water interface, as well as in AFM images of the respective LB films on mica. Interestingly, in AFM images obtained in attractive or in repulsive dynamic (also called AC or tapping) mode, the height differences between LE and LC domains were inverted. The antimicrobial agent PMB led to dramatic changes of the topography, which is consistent with the existing model published earlier.18 Materials and Methods

Figure 1. Chemical Structure of LPS from Salmonella enterica sv. Minnesota strain R595 (R595 LPS), of its lipid A part, and of the synthetic tetraacyl lipid A compound 406. The gray box in the upper structure marks the lipid A portion of the R595 LPS.

studying drug-membrane interactions. Domains observed in lipid monolayers do not necessarily allow the conclusion that they appear in the same way in the cell membrane; however, their characterization may provide a first step in the understanding of the complex natural system. The cell wall of Gram-negative bacteria is composed of the inner cytoplasmic membrane, a peptidoglycan layer, and a second bilayer, the outer membrane. Both leaflets of the inner membrane are composed of phospholipids, and in contrast to this, the outer membrane is extremely asymmetric, consisting of an inner phospholipid leaflet and an outer glycolipid leaflet. For most Gram-negative species, the glycolipid is a lipopolysaccharide (LPS), which is also called endotoxin because of its potency to induce sepsis. The outer membrane is an additional barrier for endogenous and externally applied antimicrobial agents, and in particular, the LPS leaflet is the primary target of membrane-active agents such as defensins,16 cathelicidines,17 and PMB.18 The knowledge of structural properties of LPS membranes is important for an understanding of the interaction between these membranes and the antimicrobial agents and for the development of new drugs. We investigated the topography and, in particular, the phase separation of monolayers of deep rough mutant LPS (ReLPS), of its lipid A moiety, and of the synthetic compound 406 (Figure 1) at the air/water interface of a film balance and supported on mica prepared by the Langmuir-Blodgett (LB) technique.19 Lipid A was used (15) Gutsmann, T.; Hagge, S. O.; Larrick, J. W.; Seydel, U.; Wiese, A. Biophys. J. 2001, 80, 2935-2945. (16) Ganz, T.; Lehrer, R. I. Pharmacol. Ther. 1995, 66, 191-205. (17) Gutsmann, T.; Larrick, J. W.; Seydel, U.; Wiese, A. Biochemistry 1999, 38, 13643-13653. (18) Wiese, A.; Mu¨nstermann, M.; Gutsmann, T.; Lindner, B.; Kawahara, K.; Za¨hringer, U.; Seydel, U. J. Membr. Biol. 1998, 162, 127-138. (19) Blodgett, K. B. J. Am Chem. Soc. 1935, 57, 1007-1022.

Lipids and Other Chemicals. For the formation of monolayers, deep rough mutant LPS from Salmonella enterica sv. Minnesota strain R595 (R595 LPS) was used. The LPS was extracted by the phenol/chloroform/petroleum ether method,21 purified, lyophilized, and transformed into the triethylamine salt form. The amounts of nonstoichiometric substitutions by fatty acids, L-Arap4N, and phosphoethanolamine (P-Etn) were analyzed by MALDI-TOF mass spectrometry. Thus, in the R595 LPS and its lipid A, the L-Arap4N linked to the 4′-phosphate was present at a level of 40% and the alkyl chain distribution was the follwing: 14% tetra-, 11% penta-, 57% hexa-, and 17% heptaacyl. Lipid A was isolated from R595 LPS (in the following, termed as lipid A) by sodium acetate buffer treatment (0.1 M, pH 4.4, 100 °C for 1 h). The synthetic tetraacyl lipid A (compound 406) (Figure 1) was kindly provided by K. Fukase (University of Osaka, Japan) and synthesized as described earlier.22 The fluorescent dye N-(7-nitro-2,1,3-benzoxadiazol-4-yl)-PE was purchased from Molecular Probes (Eugene, OR). PMB was purchased from Sigma (Deisenhofen, Germany). Film Balance Measurements. When amphiphilic or hydrophobic molecules are added to an aqueous phase (subphase), they tend to adsorb to the air/water interface to form a monolayer and, thus, reduce the surface tension of the subphase. Using a film balance, monolayers prepared at the air/water interface can be compressed, and the dependence of the lateral pressure on the surface area can be recorded. From pressure/area curves at a given temperature ((Π-A) isotherms), the area per molecule of the monolayer can be calculated at a given pressure. For the determination of the molecular area of R595 LPS, of its lipid A moiety, and of compound 406, a Langmuir film balance equipped with a Wilhelmy system (Munitech, Mu¨nchen, Germany) was used. The lipids were dissolved in chloroform/methanol (10:1) at a concentration of 1 mM and were carefully dropped on the aqueous subphase (deionized water containing 5 mM HEPES at pH 7.0). The subphase temperature was adjusted to 20 °C to avoid condensation of water. To allow evaporation of the solvent, the monolayers were equilibrated at zero pressure for 5 min. Pressure/area (Π-A) isotherms were then recorded at a compression rate of 3 mm2/s. For epifluorescence microscopic studies,23,24 the fluorescent dye NBD-PE was co-dissolved with the lipid in chloroform/ methanol at a molar ratio [lipid]/[NBD-PE] ) 100:2. Monolayers (20) Rietschel, E. Th.; Kirikae, T.; Schade, F. U.; Mamat, U.; Schmidt, G.; Loppnow, H.; Ulmer, A. J.; Za¨hringer, U.; Seydel, U.; Di Padova, F.; Schreier, M.; Brade, H. FASEB J. 1994, 8, 217-225. (21) Galanos, C.; Lu¨deritz, O.; Westphal, O. Eur. J. Biochem. 1969, 9, 245-249. (22) Oikawa, M.; Wada, A.; Yoshizaki, H.; Fukase, K.; Kusumoto, S. Bull. Chem. Soc. Jpn. 1997, 70, 1435-1440. (23) von Tscharner, V.; McConnell, H. M. Biophys. J. 1981, 36, 409419. (24) Weis, R. M. Chem. Phys. Lipids 1991, 57, 227-239.

6972

Langmuir, Vol. 21, No. 15, 2005

were prepared as described above and compressed to various lateral pressures. Using the fluorescent dye NBD-PE, the LE domains appeared light and the LC domains dark. The images from the epifluorescence microscope (Model 1, Munitech. Mu¨nchen, Germany) were recorded by a camera (C2400-08-C, Hamamatsu Photonics, Japan) and digitized by a video capture card. For intercalation experiments, the monolayers were prepared as described above and compressed to a lateral pressure of 20 mN/m. Twenty microliters of a 0.1 mg/mL (60 µM) PMB solution were injected into the 75 mL subphase below the monolayers, and the PMB-induced increase of the area of the lipid A monolayers was recorded at the same lateral pressure for 60 min. Langmuir-Blodgett Films. Prior to the preparation of the monolayer, a piece of freshly cleaved mica (1.5 cm × 2.5 cm) (Ruby Red Mica Sheets, Electron Microscopy Sciences, Ft. Washington, PA) was placed below the air/water interface. After equilibration at zero pressure, film balance measurements were performed as described above. Time scans of 1 h at certain lateral pressures were performed as controls for the stability of the monolayers. Then, the mica sheet, positioned perpendicular to the air/water interface, was pulled with a stepper motor (v ) 10 µm/s) through the monolayer to transfer the monolayer onto mica. In this way, the hydrophilic heads of the lipid molecules are facing the mica surface. The mica sheets were then glued on a microscope glass slide (OMNILAB, Bremen, Germany) with 5 Minute Epoxy (ITW Devcon, Danviers, MA). Atomic Force Microscopy. AFM imaging was performed in air with a MFP-3D atomic force microscope from Asylum Research (Santa Barbara, CA). The AFM was placed on an inverted optical fluorescence microscope (Olympus IX70, Tokyo, Japan). The setup was located on an air-buffered graphite table. Imaging in the AC mode was performed with AC160TS silicon nitride cantilevers with a spring constant, k, of about 40 N/m and a resonant frequency, fres, of about 300 kHz (OLYMPUS OPTICAL CO., Tokyo, Japan). For DC images, MSCT-AUNM cantilevers (E: k ) 0.1 N/m, fres ) 38 kHz) (Veeco Instruments GmbH, Mannheim, Germany) or NSG-11 (B: k ) 5.5 N/m, fres ) 150 kHz) (NT-MDT, Moscow, Russia) were used, and the forces were adjusted as low as possible. For scratching experiments, AC240TS (k ) 2 N/m, fres ) 70 kHz) (OLYMPUS OPTICAL CO., Tokyo, Japan) was used. Further image processing (flattening, planefitting, and edge detection) was done with the MFP-3D Software under IGOR Pro (Lake Oswego, OR). Height differences were taken from height histograms and also from at least five section profiles, each as an average over 5-10 scan lines. Both methods led to similar results. The percentages of the different domain areas were determined by (i) calculation from the height histograms by fitting these with Gaussian functions and (ii) using edge detection to localize the domains and integrating their areas. The results of the two methods differ less than 5%. Each value for a domain size was averaged over at least 10 different images determined for three different samples.

Results and Discussion Pressure/Area (Π-A) Isotherms of R595 LPS, Lipid A, and Compound 406. From (Π-A) isotherms, the size of one lipid molecule at a given lateral pressure and phase or structural transitions, e.g., of the alkyl chains, of lipids can be deduced. Both R595 LPS and lipid A show a transition at about 7 mN/m (Figure 2). This transition is temperature independent (data not shown); thus, it is not a phase transition of the alkyl chains. It is remarkable that lipid A undergoes a more distinct structural transformation than R595 LPS, and there is almost no transition in case of the synthetic compound 406. From FTIR experiments, it is known that the tilt angle between the alkyl chains and the diglucosamine backbone is about 47° for lipid A from R595 LPS, 33° for R595 LPS, and 0.2), it is vice versa. Between these two stable regions, an instable region was detected. The height differences in the repulsive mode were about -0.8 nm for all rSP values. In the attractive mode, they were about 0.5 nm. In the near vicinity to the instable region, they were higher but the standard deviation was also significantly higher. As mentioned above, the height values strongly depend on further parameters. This dependence becomes visible from a comparison of the height information in Figures 6 and 7, showing in the attractive mode height differences of 1.2 and 0.5 nm, respectively. Beside the LE domains with diameters up to several micrometers, we found smaller structures with sizes in the range of 100 nm which were not visible in the epifluorescence images. Interestingly, these structures were not circular in shape and did not cause a change in

Figure 8. Three possible quantitative models describing the molecular organization of lipid A molecules on the mica surface: (A) the backbones of the lipid A molecules are oriented parallel to the mica surface, and the tilted fluid alkyl chains are more depressed than the rigid ones; (B) the fluid chains are more flexible and not tilted and are, thus, sticking out of the surface; (C) the backbone of the molecules is tilted, the alkyl chains are perpendicular to the mica surface, and the fluid chains are depressed as in (A). The gray ellipses mimic the diglucosamine backbones and the black lines the fatty acid chains.

the height difference between attractive and repulsive mode. The origin of these structures is not yet clear; however, we speculate that the inhomogeneity of the fatty acid composition (see Materials and Methods section) of the lipid A preparation might be responsible for this observation. Figure 8 shows idealized models for the organization of the lipid A molecules which may serve for a discussion and understanding of the results: The diglucosamine backbone (gray ellipse in Figure 8) of the lipid A molecules is oriented parallel to the mica surface, and the alkyl chains (black lines in Figure 8) are titled by about 47° with respect to the mica surface (Figure 8A and B); following the idea that the molecules are reoriented during the compression, the alkyl chains could be perpendicular to the mica surface and the backbone could be oriented with an angle of 47° to these (Figure 8 C), and of course, the actual orientation may be between these two extremes. Differences in the orientation of the molecules in the LE and LC domains, respectively, might also explain the more pronounced height differences observed in lipid A and DPPC monolayers, respectively. The model presented in Figure 8C can also explain why the height difference in the R595

6976

Langmuir, Vol. 21, No. 15, 2005

Roes et al.

LPS monolayers is lower than that in lipid A monolayers: (i) the smaller angle of 33° (R595 LPS) would lead to smaller height differences between LE and LC domains and (ii) for mixed monolayers of lipid A and compound 406 with a large difference in the angle (47° and >20°, respectively), the height difference is more pronounced. There are two possible reasons for the different height images in attractive and repulsive mode. In models A and C in Figure 8, the height of the molecules in the LE domains is lower than in the LC domains. This assumption easily explains the results from the DC and the repulsive AC modes. To understand the results from the attractive AC mode, stronger attractive forces between the tip and the molecules in the LE domains have to be considered. In the second model (Figure 8B), the height of the molecules is higher in the LE domains. This model easily explains the attractive AC mode images, and the repulsive mode images would result from a deeper indentation of the tip into the LE domains. It may well be accepted that fluid membranes are thinner than rigid membranes; however, the difference cannot explain the height difference of 1.0 nm observed in our experiments. Berger et al. proposed that differences in the attractive forces might be due to a stronger adhesion of the Si3N4 tip to the slightly less hydrophobic CH2 groups of the fatty acids than to the terminal CH3 endgroups.30 In the LC domains, the layer is crystalline and the tip can only interact with the CH3 groups. Thus, we think that both attractive forces and indentation of the tip into the monolayer have to be considered for an understanding of the results. Characterization of Lipid A Domains at Different Lateral Pressures. Monolayers of lipid A were prepared at different lateral pressures and imaged by AFM in the DC mode (Figure 9A-E). The size and the total area of the LE domains decreased with increasing lateral pressure. At lateral pressures < 25 mN/m (Figure 9A-C), the shape of the domains was mainly circular. At higher lateral pressures, more complex shapes appeared. The size of the domains at lateral pressures > 30 mN/m was close to or below the optical resolution of the epifluorescence microscope. From the images, it is evident that the size of the LE domains decreases with increasing lateral pressure; however, the number of molecules per unit area increases. To estimate the average number of molecules per domain, the average size of the domains was calculated from the images by using the iterative or bimodal algorithms in the MFP-3D software, and the area occupied by one lipid A molecule at the respective lateral pressure was calculated from the Π-A isotherms (Figure 2). The calculated area per molecule is the average over the molecules in the LE and LC domains; thus, it provides only an estimate. The calculated numbers of lipid A molecules in LE domains is shown in Figure 9F, showing an exponential decrease of the number of molecules in the fluid LE domains with increasing lateral pressure. The heterogeneous distribution of the numbers of the fatty acids in the lipid A sample (14% tetra-, 11% penta-, 57% hexa-, and 17% heptaacylated) may contribute to the domain formation; however, an assignment of the domains to the certain acylation patterns could not be done. Interaction of Lipid A with Polymyxin B (PMB). For the development of new antimicrobial agents, it is important to understand the basic molecular mechanisms of their interaction with the bacterial membrane. To mimic the in vivo situation, PMB was injected into the subphase beneath lipid A monolayers. At a constant lateral pressure

of 20 mN/m, the injection of PMB induced an increase of the film area of the film balance, which can be taken as evidence for an intercalation of PMB molecules into the monolayer (data not shown). The AFM images of the respective LB films give evidence for dramatic changes of the domain structure of the lipid A monolayers in the presence of PMB. PMB was distributed underneath the monolayer by injecting small amounts at different locations to guarantee an utmost homogeneity of intercalation of PMB. The images shown represent areas in which PMB caused significant changes of the domain structure of the lipid A monolayer. Two effects can be observed (Figure 10), an attachment of circular structures with a diameter of up to 20 µm and a height of up to 50 nm and significant changes of the structure of the LE domains. Obviously, the addition of PMB leads to an increase of the LE fraction of the monolayer but not simply by increasing the size of the LE domains. This can be explained by a fluidization of the alkyl chains by PMB.31 Figure 10C-F reveals PMBinduced domains of irregular shapes; however, from the micrographs it is not obvious whether the domains result

(30) Berger, C. E. H.; van der Werf, K. O.; Kooyman, R. P. H.; de Grooth, B. G.; Greve, J. Langmuir 1995, 11, 4188-4192.

(31) Wiese, A.; Brandenburg, K.; Lindner, B.; Schromm, A. B.; Carroll, S. F.; Rietschel, E. Th.; Seydel, U. Biochemistry 1997, 36, 10301-10310.

Figure 9. AFM height images (DC mode) of LB monolayers of lipid A prepared at different lateral pressures. The number of fluid domains decreases with increasing lateral pressure: (A) 5, (B) 10, (C) 20, (D) 30, and (E) 40 mN/m. (F) Correlation between the averaged number of lipid molecules per fluid domain and the lateral pressure. The number of molecules per fluid area was calculated using the area per lipid A molecule determined from the respective isotherm. Cantilever: MSCTAUNM E (k ) 0.1 N/m, fres ) 38 kHz).

AFM of Lipopolysaccharide Monolayers

Langmuir, Vol. 21, No. 15, 2005 6977

From earlier investigations describing the PMB-induced permeabilization of planar membranes composed of LPS on the outer leaflet and a phospholipid mixture on the inner leaflet and thus mimicking the outer membrane of Gram-negative bacteria, we concluded that the electrostatic interaction between PMB and negatively charged lipids leads to an increase of the PMB concentration close to the membrane surface finally exceeding the CMC of PMB. The PMB micelles then intercalate into the bilayer. The results presented here are in good agreement with this hypothesis because they show that micelles or aggregates actually bound to lipid A monolayers. It can also not be excluded that PMB micelles/aggregates bound to the negatively charged mica surface in the subphase in the absence of a lipid monolayer. Under these conditions, only minor amounts of PMB were observed, and thus, artifacts can be excluded to be responsible for the effects observed in Figure 10. The formation of the PMB aggregates at the lipid A surface is probably the initial step in lesion formation in membranes. This lesion formation in the outer membrane of Gram-negative bacteria, induced by antimicrobial agents, is probably an important step in the killing of bacteria. When the lesions are large enough to allow further drug molecules to reach the cytoplasmic membrane, the drug can reach its locus of action. Chemical modifications of the LPS structure can influence the interaction with antimicrobial agents and may, thus, lead to resistance of certain bacterial strains. Conclusions

Figure 10. The intercalation of polymyxin B (PMB) leads to fluidization of lipid A monolayers. AFM height images (DC mode) of (A) a LB film of pure lipid A and (B) lipid A in the presence of PMB each prepared at a lateral pressure of 20 mN/ m. AFM height (C and D) and phase (E and F) images in (C and E) repulsive and (D and F) attractive mode. Cantilevers: (A) and (B) MSCT-AUNM E (k ) 0.1 N/m, fres ) 38 kHz); (C-F) AC160TS (k ) 40 N/m, fres ) 300 kHz).

from a separation either of different phases of lipid A or of lipid A and PMB phases. An inversion of the height and phase images in the repulsive and attractive AC mode, respectively, could also be observed in these PMBcontaining lipid A monolayers. Thus, we propose that PMB does not change the general properties of the lipid A monolayers. Binding of the polycationic PMB (five positive charges per molecule) to the negatively charged lipid A (two negative charges per molecule) is mainly driven by an electrostatic interaction.18 The height of the larger circular structures cannot be explained by a simple adsorption of PMB molecules to the lipid A monolayer. It is more likely that small aggregates of PMB molecules bind to the headgroup of the lipid A. In the monolayer experiments described here, the PMB stock solution had a concentration of 60 µM, which is below the critical micelle concentration (CMC) of PMB of about 10 mM. The concentration in the subphase was 16 nM. This value does not exclude micelle formation close to the monolayer.18

Lipid monolayers at the air/water interface or on a solid support represent the most simplified reconstitution model of biological membranes. Due to this simplicity, it is possible to describe such properties as domain formation not only of synthetic phospholipids but also of more complex glycolipids such as LPS. A knowledge of the physicochemical characteristics of LPS matrixes is important for an understanding of the antibacterial activity of various agents against Gram-negative bacteria. In many laboratories, phospholipids are used to mimic the bacterial bilayer membrane. However, to understand specific interaction mechanisms and, in particular, mechanisms leading to bacterial resistance, it is important to utilize the lipids composing the outer membrane of Gramnegative bacteria. We showed in earlier publications that the activity of antimicrobial peptides such as cathelicidins,17 amoebapores,32 PMB,18 and porins33 against Gramnegative bacteria can only be understood on the basis of their interaction with LPS. In this work, we focused on monolayers prepared from R595 LPS from S.minnesota and in particular from its isolated lipid A moiety. Both lipids are in the coexistence region of the LE and LC phases when prepared at biologically relevant lateral pressures (10 and 30 mN/ m).34,35 For an interpretation of the results obtained with monolayers representing the outer leaflet of the outer membrane, it must always be considered that, in the complex system of the bacterial bilayer membrane, domain formation may be influenced by the second leaflet and by proteins. The LE and LC domains can be observed by epifluorescence and AFM; however, only AFM images yield information about the height differences between the lipids in the two phases and, moreover, at higher lateral (32) Gutsmann, T.; Riekens, B.; Bruhn, H.; Wiese, A.; Seydel, U.; Leippe, M. Biochemistry 2003, 42, 9804-9812. (33) Hagge, S. O.; De Cock, H.; Gutsmann, T.; Beckers, F.; Seydel, U.; Wiese, A. J. Biol. Chem. 2002, 277, 34247-34253. (34) Marcelja, S. Biochim. Biophys. Acta 1974, 367, 165-176. (35) Blume, A. Biochim. Biophys. Acta 1979, 557, 32-44.

6978

Langmuir, Vol. 21, No. 15, 2005

pressures, when the size of the LE domains decreases below the optical resolution of a light microscope. To interpret the height information of AFM images of lipid monolayers taken in DC, as well as in AC mode, the knowledge of the influence of different forces on the sample is very important. However, the measurements are influenced by a number of effects, e.g., mechanical and chemical properties of the tip, leading to very complex interdependences making a straightforward interpretation difficult. In earlier studies, it has been shown that the imaging contrast can be strongly influenced by the imaging load36 and that a contrast inversion can occur in AC mode.37 The interpretation of the AFM images in dynamic AC mode requires particular care. Due to differences in the adhesion between the probe tip and the LE and LC domains, respectively, height images are inverted in the repulsive mode with respect to the attractive mode. Probably, this behavior is due to different attractive forces between the tip and the CH2 or the terminal CH3 groups, respectively, of the fatty acid chains. This result demonstrates clearly that AFM images are based on forces and that measured differences in height can strongly depend on the forces between tip and sample. The characterization of these properties of the fatty acid chains is only possible using lipid monolayers. Thus, even though the bilayer resembles the natural system more (36) Schneider, J.; Dufrene, Y. F.; Barger, W. R., Jr.; Lee, G. U. Biophys. J. 2000, 79, 1107-1118. (37) Knoll, A.; Magerle, R.; Krausch, G. Macromolecules 2001, 34, 4159-4165.

Roes et al.

closely, the monolayer can still provide some valuable information. An unequivocal interpretation of the height information of AFM images requires further experimental efforts. In contrast to Chen et al., we observed differences in the phase shift between soft (LE domains) and hard (LC domains) material also in the attractive mode.28 In former publications, it was stated that the formation of small lipid domains arises from a dewetting mechanism following film transfer.6 However, for the glycolipids used in this study, we could show that the domains in the monolayer at the air/water interface and in the LB films on mica are closely identical. Interaction of the polycationic antimicrobial peptide PMB with the anionic LPS monolayers led to an intercalation of the peptide and to drastic changes of the domain structure. These results obtained from AFM images of LB films support findings obtained from other techniques, and moreover, they expand the knowledge about the interaction between PMB and LPS. Acknowledgment. We thank Jason P. Cleveland from Asylum Research (Santa Barbara, CA) and Stefan Vinzelberg from Atomic Force F&E GmbH (Mannheim, Germany) for helpful discussions. This work was financially supported by the Deutsche Forschungsgemeinschaft (SFB 470, project B5). LA048218C