Liu A.L., Tien H.T. Advances in Planar Lipid Bilayer and Liposomes. V.6

Подождите немного. Документ загружается.

(inserted state or I state) [61–65]. The transition from the S state to the I state has a

sigmoidal peptide-concentration dependence indicating cooperativity in the pep-

tide–membrane interactions. Data gained on this transition of alamethicin measured

in three bilayer conditions were explained by a free energy that took into account

the membrane-thinning effect induced by the peptides. This model was extended

to melittin that was shown to form toroidal pores, instead of barrel-stave pores as in

the case of alamethicin, demonstrating that the membrane thinning effect is a

plausible mechanism for the peptide-induced pore formations [66].

The thinning of the bilayer observed is in the range of 2–3 A

and is an average

value over local dimple deformation [60]. However, an atomic force microscopy

study on the effect of MSI-78, a derivative of magainin, on DMPC bilayers revealed

that the bilayer thickness was not reduced uniformly over the entire bilayer area

[67]. Instead, upon binding of the peptide distinct domains were formed, where the

bilayer thickness was reduced by about 11 A

. Increasing the peptide concentration

resulted in a growth of these domains without affecting the reduction of thickness

and the parallel orientation of this a-helical peptide to the membrane.

X-ray studies also yielded structural information on the channel forming pep-

tide Gramicidin D. This peptide is known to form well-deﬁned dimeric channels in

lipid bilayers [68]. Gramicidin D apparently stretches DLPC bilayers and thins

DMPC bilayers toward a common thickness of 32.470.3 A

owing to hydrophobic

mismatch between the peptide and the hydrophobic core of the individual bilayer

[69,70]. However, further studies with the a-helical synthetic WALP peptides,

which consist of a hydrophobic sequence of leucine and alanine of varying lengths

bordered at both ends by tryptophan as membrane anchors, did not reveal a bilayer

thickness adjustment as observed in the presence of Gramicidin D [71]. Since the

energy cost of hydrophobic mismatch is greater than the energy cost of membrane

Figure 6 Illustration showing the e¡ect of an amphipat ic peptide with a spatial location parallel

to the bilayer su rface. Neighboring phospholipids will be character ized by increased trans-gauche

isomer ization of the hydrocarbon chains to ¢ll the void in the hydrophobic core resulti ng in

membrane thinning.

K. Lohner et al.114

deformation [70], this result is intriguing. The authors suggest that this energy

argument is not applicable for single helix since the way the lipids pack around it

differs greatly from larger proteins like Gramicidin.

2.3. Detergent-Like Effect – Micellar Lipid–Peptide Disk Formation

Another mechanism by which antimicrobial peptides can destroy their target

membrane involves micellization via a detergent-like mode of action [37]. The

character of particularly linear amphipathic a-helical peptides and their interactions

with membranes show analogies to detergent molecules. For example, melittin with

its overall amphipathic character composed of a cluster of cationic amino acids at

the C-terminus and a stretch of hydrophobic amino acids, resembles the features of

many detergents being characterized by a polar/charged head group and a hydro-

phobic moiety. Melittin was found to exhibit pronounced effects on the phase

behavior of DPPC already at very low peptide concentrations (lipid-to-peptide

molar ratio of 1000/1) [72,73]. A similar concentration-dependent behavior was

reported for the detergent cetyltrimethylammonium chloride [74]. However, at

high melittin concentrations (lipid-to-peptide molar ratio of 15/1), disk-shaped

particles were found for the DMPC/melittin mixture [75,76], suggesting a deter-

gent-like solubilization of the membrane under these experimental conditions. The

idea of gradual membrane disintegration is also supported by data gained from

staphylococcal d-lysin/DMPC multilamellar vesicles [77,78]. Modeling of X-ray

data further suggested that the disk-like micelles were surrounded by a ring of

peptide of about 1 nm thickness as sketched in Fig. 7 [78]. There has been some

debate about the peptide arrangement that surrounds the hydrophobic core, i.e., if

the helix axes of the peptide is located normal or parallel to the bilayer plane of the

disk-micelle. Evidence for a location parallel to the disk-plane was shown in case of

discoidal micelles formed by apolipoproteins [79]. However, as the peptide

Figure 7 Drawing of a disk-like lipid micelle with the helical peptide surrounding the disk

preventing exposure of the hydrophobic core of the lipid micelle. Note that the peptide could be

also arranged parallel to the disk surface.

Liposome-Based Biomembrane Mimetic Systems 115

orientation cannot be determined by X-ray, small-angle neutron scattering exper-

iments with deuterated peptide would be necessary to clarify this point.

2.4. Promotion or Prevention of Non-Lamellar Phases

As mentioned above, PE is a major component of bacterial cell membranes. Al-

though PE like PC is zwitterionic, they feature very different molecule geometries.

PC is shaped cylindrically, whereas PE is characterized by a truncated cone shape

owing to its small head group as compared to the cross-section of its hydrocarbon

chains [80,81]. Therefore, PE possesses a large spontaneous curvature making the

lipid prone to the formation of non-lamellar structures [82]. The balance between

lamellar and non-lamellar phase promoting lipids is thought to be essential for many

membrane-associated processes such as function of membrane proteins and mem-

brane fusion [83–85]. Incorporation of small molecules like antimicrobial peptides,

which are able to shift this balance to one direction or the other, is therefore bound

to have a large impact on membrane function and integrity [36]. Peptides that

impose a negative curvature stress on the bilayer can potentiate the intrinsic negative

curvature of these bilayers leading to the formation of non-lamellar lipid structures,

however, peptides having the contrary effect, i.e., stabilizing the PE bilayer by

intercalation into the interfacial region, are also described in the literature [86].

As outlined in Section 1.1, microorganisms such as E. coli or A. laidlawii have

been shown to precisely regulate the amounts of non-lamellar phase preferring

lipids such as PE, DPG or monogalactosyl-diglyceride in a narrow window close to

a lamellar to non-lamellar phase boundary [7,87,88]. AMPs interfere with the

packing constraints arising from the coexistence of lamellar and non-lamellar phase

forming lipids and may lower the lamellar to non-lamellar phase boundary leading

to membrane rupture. In this regard, X-ray diffraction is a highly convenient

technique to detect the presence of non-lamellar phases, with the apparent ad-

vantage over e.g.,

P-NMR that the supramolecular structure of the lipid aggre-

gates can be analyzed in all structural details [89].

The cyclic peptide Gramicidin S from B. brevis, for example, was shown to

disrupt liposomes composed of lipid extracts from E. coli and A. laidlawii by de-

creasing the energetic barriers against the formation of cubic phases by promoting

negative curvature (Fig. 8) [49].

AMPs were also shown to promote the formation of non-lamellar lipid struc-

tures in pure PE model systems. Alamethicin induces a cubic phase, when incor-

porated in small amounts in dielaidoyl PE [90] and an inverse hexagonal phase in

dioleoyl PE liposomes [91]. It was suggested that alamethicin induces such lipid

structures by changing the thickness and/or ﬂexibility of the lipid bilayer. In dip-

hytanoyl PC model membranes alamethicin causes membrane thinning with a

concomitant increase in chain disorder over a large area [62,92]. Therefore, it was

proposed that the decrease in membrane thickness is compensated by an increase of

the hydrophobic cross-sectional area of the lipid acyl chains, which in case of PE

would further enhance the existing mismatch between the cross-sectional areas of

the head group and hydrocarbon chains. This could be a promotive force for the

lipid monolayer to curl.

K. Lohner et al.116

Promotion of an inverse hexagonal phase was observed for nisin in dioleoyl- and

palmitoyloleoyl-PE, respectively [50]. The formation of this non-bilayer structure

can be explained in analogy to hydrophobic molecules such as squalene that also

strongly decrease the phase boundary between ﬂuid lamellar and inverse hexagonal

phase [93]. Thus, the insertion of the large hydrophobic section of nisin (segments

1–19) will lead to an increase of the hydrophobic volume in the bilayer interior and

in turn promote negative curvature. The role of insertion of a large hydrophobic

volume into the bilayer can also be deduced from the synthetic cecropin B analog

CB3, which in contrast to the natural cecropin B was composed of two hydro-

phobic a-helices [94]. Whereas for cecropin B pore formation was proposed, CB3

induced rapid formation of irregularly shaped, non-lamellar structures.

Examples of peptides inducing positive curvature strain in a membrane thereby

inhibiting the formation of non-lamellar phases include e.g., d-lysin [51], LL-37

[95] and magainin [96].

3. General Aspects of Model Membranes

3.1. Global and Local Membrane Properties

Having outlined the need for studying mimetics of biological membranes in the

introductory section, it appears obvious to look for novel ways to attain a better and

Intensity [a.u.]

0.1 0.2 0.3 0.

s [nm

-1

]

Im3m

Pn3m

Figure 8 Small-angle X-ray di¡ractogram of E. coli membrane total lipid extract (gray line)

showing t he typical scattering of uncorrelated bilayers and in the presence of (4 mol%)

Gramicidin S (black line) recorded at 251C. Position of hkl-re£ections corresponding to a cubic

phase of space group P n3m (dotted) and Im3m (solid), respectively, are indicated in the

di¡ractogram.

Liposome-Based Biomembrane Mimetic Systems 117

characterization of the systems throughout. In doing so, it is important to realize the

different time and length scales involved in the problem. With respect to length

scales, membrane properties may be divided into local and global features, which

are, however, intimately coupled to each other [97,98]. Local membrane properties

would be, for example, different modes of molecular motions and vibrations,

diffusion, chain tilt, but also any disorder in the vicinity of defects (Fig. 9).

In turn, global membrane properties are the overall bilayer structure, bilayer

interactions, elasticity, global curvature, chain packing, to name but a few (Fig. 10).

This general distinction between global and local membrane properties is clearly

artiﬁcial and is due to experimental techniques that address either the one or the

other aspect.

Spectroscopic techniques and all-atom molecular dynamics (MD) simulations

are traditionally applied to study membrane properties on the local scale. Important

distinctions between the different spectroscopic techniques come from the different

timescales involved that range from fast bond vibration modes with correlation

times on the order of 10

14

s to transbilayer diffusion of lipids (‘‘ﬂip-ﬂop’’), which

may take up to several days. Fourier transform infrared (FTIR) and Raman

spectroscopy are the techniques of choice to study the vibrational bands exhibiting

Figure 9 Schematic examples of local bilayer properties including s everal modes of vibrations

and rotations within a single lipid molecule (A), single molecular motions (B) and local

membrane perturbation by the inclusion of a membrane protei n (C).

K. Lohner et al.118

the fastest dynamics, whereas magnetic resonance methods such as electron par-

amagnetic resonance (EPR) and nuclear magnetic resonance (NMR) cover the

‘‘slower’’ motions, such as, molecular wobbling or axial rotations. Here, EPR is

sensitive to faster molecular motions than NMR. On the other hand, EPR requires

the incorporation of nitroxide spin-labels in the membrane, whose concentration

needs to be kept low in order to avoid effects on the bilayer structure. Of recent

interest are also dynamic scattering techniques such as coherent inelastic X-ray

and neutron scattering [99–101], neutron backscattering [102] and X-ray photon

correlation spectroscopy [103].

Global membrane properties, such as the overall structure can be determined

applying elastic X-ray or neutron scattering [104–106], but also H

NMR spectra

can yield information on the hydrocarbon chain length of the lipid bilayer [107].

Bilayer interactions can be measured by the surface force apparatus, atomic force

microscopy, pipette aspiration or the osmotic stress method [108,109]. Also bilayer

elasticity can be studied by a range of techniques including micropipette aspiration

[110], mode-analysis of bending ﬂuctuations [111] and electric-ﬁeld-induced de-

formations on large unilamellar vesicles [112], but also others such as, e.g., optical

Figure 10 Schematic of some global membrane properties like, t he overall structure with the

lamellar repeat distance d, membrane thickness d

, bilayer separation d

, hydrocarbon chain

length d

, head group thickness d

and lateral area per lipid molecule (A). Panel B shows a

bilayer interaction potential and panel C col lective bending £uctuations.

Liposome-Based Biomembrane Mimetic Systems 119

dynamometry [113], optical force measurements [114] or line-shape analysis of

Bragg peaks [115–118] have been applied. Depending on the applied experimental

technique, the reported bending rigidity K

values show some scatter, but are

usually in the range of 10

20

J-10

19

J for phospholipid bilayers.

3.2. Domains-Rafts

Biological membranes are compounds of several lipid components (and proteins,

sugars, etc.). Thus, often binary or ternary mixtures of synthetic lipids are studied in

order to address the complex lipid composition in natural membranes. The mixtures

then have to accommodate the properties of the individual components, which may

lead to a demixing or formation of lipid domains. In turn, a demixing of lipid

components may also be induced by interactions with membrane active com-

pounds, such as, peptides [36,119]. Lipid domains have been observed in model

systems already some time ago [120–123]. However, due their possible implication

in several cellular functions in natural membranes, such as endocytosis, protein

trafﬁcking and signal transduction [124,125], domains or better ‘‘rafts’’ as they are

called these days have recently seen a tremendous increase of research interest. Again

it is important to realize the different time and length scales. Domain formation may

occur as a dynamic process at short-time and small-length scales, respectively. In this

case, the formed domains are small in size and not stable but constantly create and

dissipate and are often referred to as micro-domains (Fig. 11A).

On the other hand, lipid molecules may also segregate and form large areas

composed of single lipids that are stable over longer timescales. This type of do-

mains is often referred to as macroscopic domains. It is, however, important to note

that in the thermodynamic limit only the latter case of domain formation may be

called a phase separation. A lack of this strict distinction has led to severe

controversies in literature with the most prominent example being binary mixtures

of phospholipids with cholesterol. Early phase diagrams [126], based on NMR data

[127], show the existence of a miscibility gap within the ﬂuid phase of

phospholipid/cholesterol mixtures. This ﬂuid–ﬂuid phase separation has been sup-

ported by several further experimental studies such as, [128–131]. Although

McMullen and McElhaney [130] criticized using the terminology of phase diagram

in the context of lipid/cholesterol binary mixtures, their differential scanning

calorimetry (DSC) data clearly showed the co-existence of two phases in a certain

composition range, which can be interpreted according to [126] as the L

(liquid-

disordered) and L

(liquid-ordered) phases, where the latter phase is induced at high

cholesterol concentrations due to its rigid character. In contrast, groups using other

techniques, such as X-ray diffraction [132] or more recently ﬂuorescent microscopy

[133] found no direct evidences for phase-coexistence in such binary mixtures.

Only stable and large domains within the ﬂuid phase have been reported in ternary

mixtures of lipids with cholesterol, using similar techniques [133–135].

While these more recent results will most likely lead to a revision of the original

phase diagrams of binary cholesterol/lipid mixtures, there is another general lesson

to learn from this example. Global techniques, such as, X-ray (or neutron) diffrac-

tion, or ﬂuorescence microscopy average over a large period in time and hence are

K. Lohner et al.120

unable to detect the signal from unstable microscopic domains. On the other hand,

spectroscopic techniques may also observe density ﬂuctuations from the creation

and dissipation of such ‘‘nuclei’’ and DSC seems to be able to detect both. Rather

than thinking in advantages and disadvantages of the individual techniques it is

more important to realize that only through the application of more than one

technique, questions related to domain formation can be addressed adequately, i.e.,

a technique sensitive to a long timescale should be combined with a technique

sensitive to short timescales. In the case of lipid–peptide interactions, the com-

bination of small- and wide-angle X-ray scattering (SWAXS) and DSC has been

proven to be particularly useful as described in the previous section.

4. X-ray Diffraction in Combination with a Global Data

Analysis

Several recent reviews have dealt with the determination of structural para-

meters from lipid bilayer systems using SAXS [98,136,137]. We will therefore

restrict ourselves here only to the essentials and refer for more details to the above-

listed articles.

Figure 11 Schematic of microscopic (A) and macroscopic domains ( B), where each circle

represents a single lipid molecule. Microscopic domai ns involve only a few lipid molecules and

are highly unstable. Macroscopic domains in cont rast exhibit sizes upto microns and are stable

over an extended time period.

Liposome-Based Biomembrane Mimetic Systems 121

X-rays are known to interact with the electron cloud of atoms and scatter in the

elastic case without loosing or taking up energy from the atoms. The contrast in

X-ray scattering experiments comes from the number of electrons per volume

( ¼ electron density) and its modulation across the sample. Thus, the real space

information that is obtained from scattering experiments is obtained by calculating

the electron density from the scattering data. In the case of lamellar lipid phases, the

electron density varies only across the lipid bilayer. The corresponding electron

density proﬁle can be modeled to sufﬁcient accuracy by the summation of

three Gaussian distributions [138–140], two centered at the position of the elec-

tron-dense lipid head groups and one of negative amplitude in the middle of the

bilayer, where the hydrocarbon chains meet (Fig. 12A).

In order to compare with experimental data, one needs to Fourier transform the

electron density proﬁle, which can be calculated analytically. The thereby derived

function is called the form factor F (q), which varies as a function of the modulus of

the scattering vector q, which is related to the X-ray photon wavelength l and the

scattering angle y by q ¼ 4p sin(q)/l. Figure 12B shows the absolute square of

the form factor, which corresponds to the intensity recorded in a diffraction

experiment.

Frequently, lipid bilayers self-assemble into a stack of several bilayers. In this case,

positional correlations between the bilayers leading to the observation of Bragg

peaks in scattering patterns need to be considered. The shape of the peaks is affected

due to the signiﬁcant disorder present in the lipid multibilayers. This is accounted

for by two theoretical descriptions. The paracrystalline theory [141,142] takes into

account a Gaussian distribution of distances between idealized solid layers, which

additionally display some thermal ﬂuctuations (Fig. 13B). This leads in comparison

to thermal ﬂuctuations of the layers around regularly distributed distances

(Fig. 13A) to an increase of the Bragg peak width with diffraction order. The

decrease of peak intensity is also observed in the case of simple thermal ﬂuctuations.

Figure 12 Electron density pro¢le of lipid bilayers across the membrane (panel A) and the

absolute square of its Fourier transform given by the form factor ( panel B). z

is the position of

the lipid headgroup and s

its width. Panel A further shows de¢nitions for the membrane

thickness d

and the hydrocarbon chain thickness d

K. Lohner et al.122

The second theory, named after Caille

[143,144], considers the ﬂexibility of the

bilayers yielding bending ﬂuctuations. This translates into cusp-like diffraction peaks

(Fig. 13C), which also broaden with increasing diffraction order. It has been shown

experimentally that the Caille

theory accounts for the shape of Bragg peaks of

multilamellar vesicles in the ﬂuid L

phase [145]. In turn, bilayer elasticity decreases

by about one order of magnitude upon transforming lipid bilayers into the gel phase.

Thus, in the lamellar gel phase the bilayer can be considered as solid, and the

paracrystalline theory should apply. Nevertheless, there seem to be some inconsist-

encies in ﬁtting higher order Bragg peaks [98], which may be due to the simplifying

assumption of a Gaussian distribution of bilayer separations. Nevertheless, we have

been successful in applying this theory in several cases as will be detailed below.

In X-ray scattering experiments on a liposomal dispersion, one records the

scattered intensity as a function of wave vector q. For lipid–peptide interaction it is

preferable to work under physiological relevant conditions, i.e., at full hydration

and adjusting pH and ionic strength of the buffer solution. In this case, SAXS

Figure 13 Disorder in lamella r stacking and its consequences on the Bragg peaks. (A) Thermal

disorder, (B) paracrystalline theory and (C) Caille

theory.

Liposome-Based Biomembrane Mimetic Systems 123