You are on page 1of 54

Accepted Manuscript

Effect of general anesthetics on the properties of lipid membranes


of various compositions

György Hantal, Balázs Fábián, Marcello Sega, Balázs Jójárt, Pál


Jedlovszky

PII: S0005-2736(18)30358-4
DOI: https://doi.org/10.1016/j.bbamem.2018.12.008
Reference: BBAMEM 82901
To appear in: BBA - Biomembranes
Received date: 16 October 2018
Revised date: 5 December 2018
Accepted date: 13 December 2018

Please cite this article as: György Hantal, Balázs Fábián, Marcello Sega, Balázs Jójárt, Pál
Jedlovszky , Effect of general anesthetics on the properties of lipid membranes of various
compositions. Bbamem (2018), https://doi.org/10.1016/j.bbamem.2018.12.008

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
ACCEPTED MANUSCRIPT

Effect of General Anesthetics on the Properties of Lipid


Membranes of Various Compositions

György Hantala, Balázs Fábián,b,c Marcello Segaa,1, Balázs Jójárt,d and


Pál Jedlovszkye*

PT
1
Faculty of Physics, University of Vienna, Sensengasse 8/9, A-1090 Vienna,

RI
Austria

SC
2
Department of Inorganic and Analytical Chemistry, Budapest University of
Technology and Economics, Szt. Gellért tér 4, H-1111 Budapest, Hungary
NU
3
Institut UTINAM (CNRS UMR 6213), Université Bourgogne Franche-Comté,
MA

16 route de Gray, F-25030 Besançon, France


D

4
Institute of Food Engineering, University of Szeged, Moszkvai krt 5-7, H-6725
E

Szeged, Hungary
PT

5
Department of Chemistry, Eszterházy Károly University, Leányka utca 6, H-
CE

3300 Eger, Hungary


AC

1
Present address: Helmholtz-Institut Erlangen-Nürnberg, Forschungszentrum Jülich, Fürther
Straße 248, D-90429 Nürnberg, Germany
*
Electronic mail: jedlovszky.pal@uni-eszterhazy.hu (PJ)

1
ACCEPTED MANUSCRIPT

Abstract: Computer simulations of four lipid membranes of different compositions, namely


neat DPPC and PSM, and equimolar DPPC-cholesterol and PSM-cholesterol mixtures, are
performed in the presence and absence of the general anesthetics diethylether and sevoflurane
both at 1 and 600 bar. The results are analyzed in order to identify membrane properties that
are potentially related to the molecular mechanism of anesthesia, namely that change in the
same way in any membrane with any anesthetics, and change oppositely with increasing
pressure. We find that the lateral lipid density satisfies both criteria: it is decreased by
anesthetics and increased by pressure. This anesthetic-induced swelling is attributed to only

PT
those anesthetic molecules that are located close to the boundary of the apolar phase. This
lateral expansion is found to lead to increased lateral mobility of the lipids, an effect often

RI
thought to be related to general anesthesia; to an increased fraction of the free volume around

SC
the outer preferred position of anesthetics; and to the decrease of the lateral pressure in the
nearby range of the ester and amide groups, a region into which anesthetic molecules already
NU
cannot penetrate. All these changes are reverted by the increase of pressure. Another
important finding of this study is that cholesterol has an opposite effect on the membrane
properties than anesthetics, and, correspondingly, these changes are less marked in the
MA

presence of cholesterol. Therefore, changes in the membrane that can lead to general
anesthesia are expected to occur in the membrane domains of low cholesterol content.
D

Keywords: molecular mechanism of anesthesia, general anesthetics, lipid membranes,


E

cholesterol, pressure reversal, computer simulation


PT
CE
AC

2
ACCEPTED MANUSCRIPT

1. Introduction

General anesthesia has been used in the surgical practice for almost two centuries. It
was also clarified at the turn of the 20th century that the site of action of general anesthetics is
located inside the cell membrane, when Meyer [1] and Overton [2] showed, independently
from each other, that the anesthetic efficiency of these molecules depends linearly on their
oil:water partition coefficient. However, in spite of its long and widespread utilization, little is

PT
known so far about the molecular level mechanism of general anesthesia. Understanding this
mechanism is complicated by the fact that general anesthetics show a very large chemical

RI
variety. In addition, anesthesia is known to be reverted at elevated pressures. [3-7] As a

SC
consequence, any possible explanation of the molecular mechanism of anesthesia has to be
valid for all kinds of general anesthetics, and has to account also for the phenomenon of
pressure reversal.
NU
Possible explanations of general anesthesia can be divided into two major groups.
Protein theories assume that the anesthetics act by interacting with an active site at the
MA

transmembrane domain of certain membrane-bound proteins [8-11]. However, since protein-


substrate interactions are usually highly specific, protein theories can hardly account for the
large chemical variety of the general anesthetics. As a consequence, it is rather unlikely that
D

general anesthesia can be explained by a unique mechanism in the frame of the protein
E

theories. To overcome this problem Mihic et al. suggested that anesthetics might change the
PT

dynamics of certain loops of the relevant protein rather than binding to its active site [11].
Nevertheless, neither the protein responsible for the anesthetic action nor its active site has
CE

been unambiguously identified so far.


Lipid theories, on the other hand, assume that anesthetic molecules alter the structure
AC

and/or the dynamics of the cell membrane in a certain way, and these changes alter the
function of some of the membrane-bound proteins. Lipid theories can easily account for the
large chemical variety of the general anesthetics, however, neither the relevant membrane
property that is perturbed by the presence of the anesthetic molecules has been
unambiguously identified, nor the way this altered membrane property changes the function
of the protein has been clarified yet. Concerning the relevant membrane property, several
hypotheses have been proposed in the past six decades. Probably the first of such conjectures
was the critical volume hypothesis of Mullins [12]. According to this hypothesis, anesthetics
increase the volume of the membrane, and anesthesia occurs if the molar volume of the

3
ACCEPTED MANUSCRIPT

membrane exceeds a certain threshold value [12]. One of the advantages of this hypothesis is
that it can easily account for pressure reversal. In interpreting the critical volume hypothesis,
it was generally assumed that anesthetics increase the membrane volume by increasing its
thickness, either by pushing the two leaflets of the bilayer farther away from each other, or by
increasing the order of the lipid tails. However, the molar volume can also increase due to
lateral swelling of the membrane. Membrane fluidity was also assumed to be the relevant
membrane property that is altered by the anesthetics and is responsible for the anesthetic
effect [7,13,14]. Assuming that general anesthetics are ideally dissolving in the disordered ,

PT
but do not dissolve at all in the ordered  phase of the membrane, Heimburg et al. showed

RI
that, in a clear analogy with the well known freezing point depression, anesthetics lower the
- phase transition temperature of the membrane, and the magnitude of the effect depends

SC
only on the water:membrane partition coefficient [15,16]. In a similar spirit, it was proposed
that, besides shifting the temperature of this phase transition (also known as the chain melting
NU
transition), general anesthetics might also lower the critical temperature of the membrane
miscibility transition in multicomponent (cholesterol containing) membranes. In other words,
MA

anesthetics can inhibit lipid heterogeneity by suppressing the formation of lipid rafts. This
way, anesthetics might have an indirect effect on membrane proteins by providing a particular
lipid environment that might promote or suppress protein functionality [17]. Using
D

thermodynamic arguments, Cantor proposed that anesthetic molecules alter the lateral
E

pressure profile in the membrane, and this change shifts the conformational equilibrium of
PT

certain membrane-bound proteins [18,19]. This hypothesis can again easily account for the
pressure reversal of anesthesia, however, its experimental verification or falsification seems to
CE

be extremely difficult, as it would require the accurate measurement of the profile of the
lateral pressure component with a sub-nanometer resolution.
Both experimental and computer simulation results are rather controversial in the
AC

respect of these hypotheses. Thus, Haydon et al. found, in capacitance measurements, a


noticeable anesthetic-induced increase of the membrane thickness, which was attributed to the
ordering of the lipid tails [20]. However, performing X-ray and neutron scattering
measurements on a dimyristoylphosphatidylcholine (DMPC) membrane Franks and Lieb did
not find any evidence of the increased membrane thickness as well as of tail ordering, even if
they used the anesthetics in considerably higher concentration than what is used in surgical
conditions [21]. Further, a decrease of the lipid tail order in the presence of halothane and
methoxyflurane was observed in electron spin resonance (ESR) measurement [22]. In

4
ACCEPTED MANUSCRIPT

performing computer simulation studies of halothane in the dipalmitoylphosphatidylcholine


(DPPC) bilayer Klein et al. did not find any noticeable change of the lipid tail order in the
presence of 6.5 mol% halothane [23], but found a substantial decrease of the lipid tail order
when 50 mol% halothane was used [24]. Computer simulation studies also revealed that while
ethanol decreases the membrane thickness [25-28], xenon leads to an increase of it [29,30].
Hauet et al. found by performing differential scanning calorimetry (DSC) and X-ray
scattering measurements that enflurane is accumulated around the polar lipid headgroups
rather than between the two leaflets of the DPPC membrane [31]. Similarly, computer

PT
simulation studies showed that certain general anesthetics, such as ethanol [26,27,32-36] or
ketamine [37] are located almost exclusively close to the polar headgroups, while halothane

RI
[38,39] and isoflurane [40] stay in this region with the highest probability, although their

SC
density does not drop to zero between the two leaflets. Some other anesthetics, such as
chloroform, diethyl ether, enflurane, and xenon also prefer this position, although a higher
NU
preference corresponds to the middle of the bilayer [41-46]. In the early experiments, it was
generally assumed that anesthetics do not change the lateral density of the membrane. Once
this assumption was no longer used, a clear lateral expansion of the membrane was observed
MA

in the presence of anesthetics [47,48]. These controversies, at least partly, originate from the
fact that different lipids and different anesthetics were used in different studies. However, as it
D

has already been emphasized, any successful explanation of the molecular mechanism of
anesthesia has to be valid for all general anesthetics as well as for different models of the cell
E

membrane.
PT

In our recent studies we simulated the DPPC bilayer in the presence of four different
general anesthetics, namely chloroform, halothane, diethyl ether, and enflurane as well as in
CE

their absence, both at atmospheric pressure and at 1000 bar in the gel [44] as well as in the
liquid crystalline [45] phase. We have tested several membrane properties, including the
AC

thickness and lateral density of the membrane, density profiles of various atoms, self-
association of various anesthetics, overall orientation of the lipid tails and headgroup as well
as lipid tail ordering, using the criteria that any membrane property that is related to the
molecular mechanism of general anesthesia must change (i) in the same way in the presence
of any anesthetic molecule, and (ii) in the opposite way with increasing pressure. In
accordance with results of several other recent simulation studies, we found only one of the
above membrane properties, namely the lateral density of the membrane, to fully meet both of
these criteria. Thus, in systems where anesthesia is supposed to occur (i.e., membranes
containing anesthetics at atmospheric pressure) the average area per lipid always exceeded the

5
ACCEPTED MANUSCRIPT

critical value of 65 Å2, whereas in systems in which anesthesia is not supposed to occur (i.e.,
membranes without anesthetics or at high pressure) it always remained below this value
[44,45]. As a consequence, the molar volume of the membrane also showed similar behavior.
It should be noted that while these results are in accordance with the critical volume
hypothesis of Mullins [12], they explain it in a different way than what was thought for a long
time, i.e., through the anesthetic induced lateral expansion rather than through the broadening
of the membrane. In these systems we calculated also the free volume and lateral pressure
profiles along the membrane normal axis [49]. Our results showed that the inclusion of the

PT
anesthetics in the membrane created additional free volume at 8-16 Å from the middle of the
bilayer, in one of the preferred positions of the anesthetic molecules. This additional free

RI
volume led to a noticeable decrease of the lateral pressure in the nearby region of the ester

SC
groups, at about 13-18 Å from the bilayer center, where the additional space resulted from the
pushing away of the lipid molecules from each other by the anesthetic molecules is still
NU
noticeable, but anesthetic molecules are already not present. The increase of the pressure
evidently reverted both of these effects [49].
These findings are, in general, in line with the results of several recent studies,
MA

involving also other general anesthetics, such as ethanol or xenon [26-30,36-43,46,50].


However, contrary to the anesthetic molecules, the role of the membrane composition in this
D

respect has only scarcely been studied. Porasso et al. showed that in the zwitterionic DPPC
the local anesthetics benzocaine stays in the hydrocarbon core, but if the membrane contains
E

also lipid molecules with a negatively charged headgroup, then benzocaine prefers the
PT

position at the boundary between the polar and apolar membrane regions [51]. Reigada
studied chloroform in a neat phospholipid membrane as well as in a mixed membrane of
CE

phospholipid and cholesterol [42,52]. Since cholesterol orders the lipid tails, the neat
membrane was in the disordered (), while the mixed membrane in the ordered () phase.
AC

Chloroform was found to have opposite effects on the two membranes: while it increased the
order in the disordered phase, it induced some disorder in the ordered phase of the bilayer
[42]. Further, it was also shown that in membranes containing rafts of the two phases,
chloroform is clearly accumulated in the disordered phase domains [52].
To better understand which of the anesthetic-induced changes in lipid membranes
might be relevant to the molecular mechanism of general anesthesia, these changes have to be
investigated not only using a large variety of anesthetics, but also in membranes of various
compositions. For this purpose, here we report computer simulations of two general
anesthetics, namely diethyl ether (DE) and sevoflurane (SF) embedded in four lipid bilayers

6
ACCEPTED MANUSCRIPT

of different compositions, together with those of the corresponding anesthetic-free


membranes. The first bilayer is composed solely by DPPC molecules; the second bilayer
contains DPPC and cholesterol in equimolar composition, the third bilayer consists of the
equimolar mixture of N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM), a
sphingomyelin molecule, and cholesterol, while the fourth one is made with only PSM
molecules (even though it does not appear alone in living cells, but uniquely with other
membrane components in lipid rafts), to serve as a reference for our study. Although
phospholipids, such as DPPC, are the main components of eukaryotic cell membranes,

PT
cholesterol is also an abundant component of such membranes. However, cholesterol is not
fully miscible with phospholipid molecules; their miscibility gap covers the thermodynamic

RI
region typical for the membranes of living cells [53-60]. Cholesterol is known to increase the

SC
lateral density [61,62] and decrease the fluidity of the membrane [63,64], and has an ordering
effect on the lipid molecules [65,66]. PSM is present at high concentration in the membranes
NU
surrounding the axons of the nerve cells, primarily in rafts mixed with cholesterol. In the PSM
molecule, one of the lipid tails is connected to the backbone by an amide rather than ester
bond, while the other tail is linked to the backbone via C-C bonds. The schematic structure of
MA

the lipid and anesthetic molecules considered in this study are shown in Figure 1. All
simulations of the anesthetic containing membranes are performed both at atmospheric
D

pressure and at 600 bar, in order to investigate also the effect of pressure reversal. In the
analysis, we focus on the anesthetic and pressure induced changes of those membrane
E

properties (e.g., area and molar volume of the membrane, distribution of the anesthetics along
PT

the membrane normal, dynamics of the lipid molecules, lateral pressure profile) which, in the
light of recent studies [26-30,35-50], seem to be possibly relevant in respect of the molecular
CE

mechanism of general anesthesia.


The paper is organized as follows. In section 2 details of the simulations performed as
AC

well as of the lateral pressure profile calculation are given. The obtained results are presented
and discussed in detail in section 3. Finally, the main conclusions of this study are
summarized in section 4.

7
ACCEPTED MANUSCRIPT

2. Computational details

2.1. Molecular dynamics simulations

Four lipid bilayers, consisting of only DPPC and only PSM molecules, an equimolar mixture
of DPPC and cholesterol, and an equimolar mixture of PSM and cholesterol, respectively,
have been simulated on the (N,p,T) ensemble at 330 K and 1 bar. All bilayers have been built

PT
up by 256 molecules, and have been hydrated by 7680 water molecules, corresponding to full
hydration of the bilayers [67,68]. Besides the neat bilayers themselves, all the four

RI
membranes have also been simulated in the presence of 192 DE, and of 72 SF molecules,

SC
corresponding roughly to the same anesthetic mass density in the membrane. It should be
noted that, in order to amplify any effect exerted by the anesthetic molecules on the properties
NU
of the membrane, and clearly distinguish such effects from the statistical noise, the
concentration of the anesthetic molecules used in the simulations well exceeds that of surgical
conditions, as it was originally suggested by Oh and Klein [39]. Finally, to investigate also the
MA

phenomenon of pressure reversal of anesthesia, the simulations of the anesthetic containing


membranes have been repeated at 600 bar, making the total number of simulations performed
D

thus 20. Again, the pressure of 600 bar used in our simulations is considerably higher than the
range in which pressure reversal was already observed (i.e., 100-200 bar) [3-7]; this choice
E

was again dictated by the requirement of exaggerating the effect of the pressure on the
PT

membrane properties, making the pressure induced changes clearly distinguishable from the
statistical noise.
CE

Lipid and DE molecules have been described by the all atom CHARMM36 force field
[69-71], SF molecules by the CHARMM compatible force field developed by Arcario et al
AC

[72], whereas water molecules have been modeled by the modified TIP3P (mTIP3P) potential
[73], which was designed to be compatible with the CHARMM force field [74]. According to
these models, the total potential energy of the systems has been calculated as the sum of the
intramolecular terms and interaction energies of all molecule pairs, the latter being the sum of
the Lennard-Jones and charge-charge Coulomb contributions of all pairs of their respective
atomic interaction sites. The interaction parameters of the water and anesthetic models used in
the simulations are collected in Table 1. All interactions have been truncated to zero beyond
the group-based center-center cut-off distance of 12 Å; the long range part of the electrostatic
interaction has been accounted for by the smooth Particle Mesh Ewald (sPME) method [75]

8
ACCEPTED MANUSCRIPT

with a mesh spacing of 1.2 Å. The intramolecular term of the potential energy has consisted
of contributions from bond vibration, angle bending and torsional rotations. All bonds
connecting at least one H atom have been kept fixed at their equilibrium values using the
SHAKE algorithm [76].
The simulations have been performed using an in-house modified version of the
GROMACS 5.1 program package [77], which also calculates the virial contribution of each
particle [78]. Equations of motion have been integrated in time steps of 2 fs. The temperature
and pressure of the systems have been controlled by means of the Nosé-Hoover thermostat

PT
[79,80] and the Parrinello-Rahman barostat [81], respectively, with the characteristic time of
0.5 ps and 5 ps, respectively. In the case of the barostat, semi-isotropic coupling has been

RI
employed, i.e., the pressure components parallel and perpendicular to the bilayer plane have

SC
been controlled separately from each other.
The initial structure for the bilayers has been prepared using the CHARMM-GUI web
NU
server [82]. In cases of mixed bilayers, the membrane components have randomly been
distributed by the software. The anesthetic molecules have been placed into the middle of the
membrane. In the first step, systems consisting of 64 lipid molecules only have been prepared,
MA

in order to ensure that energy minimization can successfully be performed. Then, the
following four steps pre-equilibration protocol has been applied for the small systems. First,
D

an energy minimization process with a maximum number of 5000 steps has been performed.
The systems have then been relaxed in a 25 ps long constant volume run, during which the P
E

atoms of the DPPC and PSM, and O atoms of the cholesterol molecules have been restrained
PT

with a 1000 kJ/mol×nm2 force constant. In the next (N,V,T) ensemble run, being again 25 ps
long, these constraints have been removed. Finally, a 800 ps long (N,p,T) ensemble run has
CE

been performed at 1 bar, in order to allow equilibration of the periodic cell. After this pre-
equilibration procedure, the final size systems have been created by replicating the small
AC

systems twice along the Y and Z axes, both of which are laying, according to our convention,
in the bilayer plane. The systems created this way have served as the starting configurations
of the 1 bar simulations, whereas the simulations performed at 600 bar have started from the
final configurations of the corresponding 1 bar runs. Once the starting configurations were
created, the systems have been equilibrated for at least 40 ns, in accordance with the earlier
suggestion of Anézo et al. that reliable simulations of phospholipid membranes require at
least 10-20 ns equilibration [83]. Then, in the production stage, 2000 configurations per
system, sampled from 20 ns equilibrium trajectories, have been saved for the analyses.
Considering the 20 systems simulated, this makes the total simulation time over 1.2 s. The

9
ACCEPTED MANUSCRIPT

lateral pressure profiles have been calculated on-the-fly, over 20000 equilibrium
configurations. Results of all analyses have been averaged not only over the sample
configurations, but also over the two sides of the lipid bilayers. It should be noted that the
symmetrized profiles calculated this way turned out to be almost indistinguishable from the
non-symmetrized ones. Equilibrium snapshots of the four bilayers containing DE, simulated
at 1 bar, are shown in Figure 2.

2.2. Calculation of the lateral pressure profile

PT
In calculating its profile, the lateral component of the pressure is needed to be

RI
decomposed to contributions of well defined positions. While such a decomposition of the

SC
kinetic part of the pressure can trivially be done, that of the potential (or excess) contribution
is not obvious at all. Namely, if the energy is pairwise additive, the elements of the virial
NU
tensor, corresponding to this excess contribution, can be written in a form containing an
integral over an open path:
MA

1 
 ij (r)     F
i
C  (r  s ) ds j
, (1)
2   
 
D

where ij is the ijth element of the virial tensor, F


i
is the ith component of the force acting
E

between atoms  and , C is the open path connecting the two atoms, over which the
PT

integration has to be performed, s and sj are the position vector of the path element and its jth
component, respectively, and  is the Kronecker symbol. Among the possible choices of the
CE

C contour that provide compatible results with each other [84], we have chosen the
Harasima path [85] for two important reasons. First, this way the lateral component of the
AC

virial corresponding to the interaction of atoms  and  is distributed between the positions of
these two atoms, and hence the entire lateral pressure can be decomposed to contributions
located at the atomic positions. As a consequence, lateral pressure contributions can formally
be treated in the simulations as characteristics of the individual atoms, which makes the
calculation of the lateral pressure profile computationally very efficient [86]. Second, the
lateral pressure profile can also be calculated in cases when the potential energy of the system
is not fully pairwise additive [84], such as in the reciprocal space part of the long range
correction, as calculated by the PME method. Thus, the use of the Harasima path allows us to

10
ACCEPTED MANUSCRIPT

distribute even the lateral pressure contribution corresponding to the reciprocal space term of
the PME correction among the atomic sites [87]. This way, the former practice of calculating
the lateral pressure profile by reanalyzing the dumped trajectory by changing the potential
(i.e., using a large cut-off rather than PME correction, as done in the simulation), used in the
vast majority of the earlier pressure profile calculations [27,29,50,88-92], can be avoided.

3. Results and discussion

PT
3.1. Density profiles

RI
3.1.1. Overall and lipid density profiles

SC
The mass density profiles of the systems simulated are shown in Figure 3. The inset of
the figure compares the electron density profile of the neat DPPC membrane as resulted from
NU
the present simulation with that of experimental data [67] as well as with our earlier result
[45], obtained using a GROMOS87-based force field [93]. As is seen, apart from the height of
the peaks around 20 Å and depth of the minimum at the middle of the membrane, the present
MA

simulation reproduces the experimental curve very well, and this agreement is certainly
noticeably better than what was previously reached using the GROMOS-based force field.
D

The comparison of the mass density profiles does not show any clear trend as far as
the effects of the anesthetics and pressure are concerned. DE is dissolved, to a small extent, in
E

the aqueous phase, lowering thus the density of this phase, but no such effect is seen for SF.
PT

The general shape of the profiles is characterized by two marked maxima at the positions of
the crowded headgroup regions and a minimum in the middle of the membrane. As is seen,
CE

both the height and the position of the maxima are rather insensitive both to the presence of
anesthetics and to the pressure. Further, in the cases when they slightly change, these changes
AC

show no clear trend. For instance, the presence of DE shifts the two maxima slightly outward
in the case of the DPPC, but slightly inward in the case of the PSM bilayer. Anesthetics either
leave the density unchanged at the position of its minimum in the middle of the membrane, or
slightly increase it; however, the increase of the pressure also leads to the increase of the
density here. The only exception in this respect is SF in the cholesterol-containing
membranes, in the middle of which it induces a large increase of the density, suggesting that
SF prefers to stay in the middle of these membranes. The increase of the pressure, however,
leads to a further increase of the density here even in these membranes.

11
ACCEPTED MANUSCRIPT

On the other hand, cholesterol has a clear and marked effect on the mass density
profiles. Thus, in the presence of cholesterol the two maxima are shifted farther from the
middle of the bilayer, from about 20 to 23 Å, and they exhibit a shoulder at their inner side,
around 20 Å. Further, the minimum in the middle of the membrane becomes clearly deeper in
the presence than in the absence of cholesterol. To understand the origin of these changes, we
have calculated the number density profiles of several lipid atoms in the membranes studied.
The number density profiles of selected lipid atoms in the DPPC- and PSM-containing
membranes are shown in Figures 4.a and 4.b, respectively. The selected atoms include the N,

PT
P, carboxylic O (Ocarb), and chain terminal C (Cterm) atoms of DPPC, the N, P, amide O
(Oamide), and chain terminal C (Cterm) atoms of PSM, as well as the O atom, the C atom that

RI
joins the second and third six-member ring and is closer to the five-member ring (Cr), the C

SC
atom of the methyl group connected to the first atom of the tail (Cm), and the chain terminal C
(Cterm) atoms of cholesterol. The constitutional position of these atoms in the lipid molecules
NU
is indicated in Fig. 1. As is seen, in the presence of cholesterol, the positions of the DPPC and
PSM N and P atoms are shifted farther from the middle of the membrane, indicating that
cholesterol increases the membrane thickness. This finding is in a clear accordance with
MA

earlier experimental [94] as well as simulation results [95-98]. Further, the O atom of the
cholesterol molecules stays at the same depth along the membrane normal than the carboxylic
D

and amide O atoms of the DPPC and PSM molecules, respectively. This finding suggests that
the cholesterols might well form hydrogen bonds with the ester and amide groups of these
E

molecules.
PT

To clarify this point, we have calculated the partial pair correlation functions (g(r)) of
the cholesterol O atom with both O atoms of the ester groups of DPPC as well as with the
CE

amide N and O and sphingosine O atoms of PSM. In addition, the g(r)’s of the corresponding
O-H and N-H atom pairs have also been calculated. The pair correlation functions, obtained in
AC

the anesthetic-free systems, are shown in Figure 5 (similar functions have been obtained in
the systems containing anesthetics). As is seen, the pair correlation functions of the
cholesterol O atoms with both the DPPC and the PSM carboxylic O atoms exhibit a sharp
peak around 2.7 Å, corresponding to the hydrogen bonding distance of two O atoms. A
similar, although much smaller peak is seen also on the cholesterol O – PSM sphingosine O
pair correlation function, which is very likely an artifact caused by the force field [71].
Furthermore, the cholesterol O – PSM N pair correlation function exhibits a marked peak at
slightly larger distances, around 3.4 Å. These findings indicate that the carboxylic O atoms of
both the DPPC ester and PSM amide groups form strong hydrogen bonds with the cholesterol

12
ACCEPTED MANUSCRIPT

O atoms, while a large amount of weaker H-bonds are formed between the cholesterol O and
PSM N atoms. On the other hand, the single bonded (ether) O atoms of the DPPC ester group
do not participate in such hydrogen bonds. These findings are in accordance with the results
of earlier, more detailed studies in the literature [99-101]. The analysis of the H-O g(r)’s fully
confirms these conclusions. It is also evident from these functions that in the PSM N atoms
always participate as the H-donor partner in these hydrogen bonds.
The observed extensive hydrogen bonding of cholesterols with DPPC and PSM
explains the appearance of the shoulder of the mass density peaks around 20 Å in these

PT
systems (see Fig. 3). As is seen from Fig. 4, this is the position of the cholesterol OH as well
as DPPC ester and PSM amide (and, supposedly, also OH) groups, i.e., exactly the region

RI
where this extensive H-bonding occurs, which brings the corresponding atoms closer to each

SC
other, and hence increases the local density. This H-bonding can also explain the marked
density decrease in the middle of the membrane upon adding cholesterol to the bilayer. Thus,
NU
since the head of the cholesterol molecules is kept at a certain position along the membrane
normal by this H-bonding, the shorter of the two molecules simply cannot reach the middle of
the bilayer. This is evidenced by the fact that while the density profiles of the C term atoms of
MA

the DPPC molecules is of Gaussian shape in the middle of the anesthetic-free DPPC-
cholesterol mixed membrane, that of the cholesterol Cterm atoms shows two distinct peaks. An
D

even more marked effect is seen in the middle of the anesthetic-free PSM-cholesterol mixed
membrane, where the cholesterol Cterm atoms have a Gaussian, while the PSM Cterm atoms a
E

clearly bimodal density profile. This indicates that, on average, the cholesterol chains cannot
PT

penetrate as deep into the membrane as the DPPC tails, while the PSM tails cannot reach the
depth of the end of the cholesterol chains. It is also clear that the presence of anesthetics in
CE

these mixed membranes makes even the originally Gaussian shaped Cterm profiles bimodal,
and also enhances the splitting of the originally bimodal Cterm profiles, indicating that the
AC

anesthetic molecules penetrate into the middle of these membranes, pushing the lipid
molecules somewhat outward.

3.1.2. Anesthetic density profiles


The mass density profiles of the anesthetic molecules across the simulated bilayers are
shown in Figure 6. As is seen, unlike SF, a small amount of DE is dissolved also in the
aqueous phase. This behavior is clearly different from what we observed in our previous
simulation, performed with a different, GROMOS-based force field [45]. The profiles
obtained in the cholesterol-free membranes are clearly characterized by three marked peaks,

13
ACCEPTED MANUSCRIPT

corresponding to two preferred positions of the anesthetic molecules along the membrane
normal. The outer one of these two preferred positions, located at about 10 Å from the middle
of the membrane at both sides, is at the outer part of the apolar hydrocarbon region, close to
the region of the carboxylic (in DPPC) and amide (in PSM) O atoms, both of which are
located at about 15 Å from the bilayer center (see Fig. 4). This position is referred to here as
the ‘outer preferred position’ of the anesthetics. The other preferred position of the anesthetics
is right in the middle of the bilayer. This dual preference of the anesthetic molecules is in
accordance with the results of earlier simulations of various anesthetic-containing membranes

PT
[29,30,41-46,102]. These anesthetic density profiles can be very well fitted by the sum of
three Gaussians, two of which are mirror images of each other (see the inset of Fig. 6). At

RI
1 bar, the position of the outer Gaussian is at 11.4 Å and 10.8 Å for DE and SF, respectively,

SC
in the DPPC bilayer, while at 11.3 Å and 11.6 Å, respectively, in the PSM bilayer. The
increase of the pressure shifts this peak somewhat closer to the bilayer center, thus, at 600 bar
NU
it appears in DPPC at 10.5 Å (DE) and 10.2 (SF), while in PSM at 10.9 Å for both
anesthetics. Integration of the fitted Gaussians reveals that at 1 bar 53-57% of the anesthetic
molecules are in the outer position in every case, and the increase of the pressure increases
MA

their amount noticeably, to 57-63%. This observed dual preference can be explained by the
interplay of the electrostatic interaction, which tries to bring the weakly polar anesthetic
D

molecules closer to the polar headgroup region, and the steric interaction, which gives rise to
the preference for the lowest density part (i.e., middle) of the bilayer.
E

The picture is somewhat different in the case of the cholesterol-containing membranes.


PT

Although the above dual preference is still seen in every case, with the exception of SF in the
PSM-cholesterol mixed membrane at 1 bar, the preference for the outer position is
CE

substantially weaker, while that for the inner position is substantially stronger here than in the
cholesterol-free membranes. Due to the large difference in the amplitudes of these peaks,
AC

these density profiles can no longer be well fitted by the sum of three Gaussians. This result is
consistent with our above finding that, due to the headgroup H-bonding and difference in the
chain lengths, the middle of the mixed membranes is characterized by a considerably lower
density than that of the corresponding cholesterol-free membranes, and hence the anesthetic
molecules can easily occupy the additional empty space in the middle of these bilayers. In
other words, the partitioning of the anesthetics between the two preferred positions, caused by
the aforementioned charge-space competition, is substantially shifted towards the middle of
the bilayer by the additional empty space created here by the presence of the cholesterol

14
ACCEPTED MANUSCRIPT

molecules. Similarly, the increase of the pressure decreases the empty space here (see Fig. 3),
and hence shifts this partitioning towards the outer position.

3.2. Thickness, area, and volume of the membrane

To investigate how the presence of the anesthetic molecules and the pressure affects
the size of the membrane both in the lateral direction and along its normal, we have calculated
the thickness (LX), area per lipid (Alip), and volume per lipid (Vlip) of the membranes

PT
simulated. The membrane thickness is estimated here by the distance of the peak positions of
the Gaussians fitted to the DPPC or PSM P atom density profiles at the two sides of the

RI
bilayer, Alip is simply the cross-sectional area of the basic simulation box divided by 128, i.e.,

SC
the number of lipid molecules in one side of the bilayer, while Vlip = LXAlip/2, where the factor
2 in the denominator accounts for the two leaflets of the membrane. The LX, Alip, and Vlip
NU
values obtained for the different membranes simulated are compared in panels a, b, and c,
respectively, of Figure 7, and are also summarized in Table 2. The value of 61.6 ± 0.9 Å,
obtained for the neat, anesthetic-free DPPC membrane agrees excellently with the
MA

experimental values of 62 ± 2 Å [103], and 62.9 ± 0.9 Å [67].


As is seen, the membrane thickness changes systematically neither with the presence
D

of the anesthetic molecules, nor with the pressure. Thus, for instance, the thickness of the
DPPC membrane is increased upon adding DE, but decreased upon adding SF to it. Opposite
E

changes are seen in the case of the DPPC-cholesterol mixed membrane. Also, the increase of
PT

the pressure to 600 bar is found to increase the thickness of the DPPC membrane in the
presence of SF, but to decrease it in the presence of DE. Pressure, in fact, has a complex
CE

effect on the membrane thickness. Thus, besides making the system denser in all directions,
which leads to a decrease of the membrane thickness, it might also well increase the lipid tail
AC

order and push some water molecules into the headgroup region; both of these effects induce
some increase of the membrane thickness. The observed change of LX with the pressure is
resulted from the interplay of all these effects. Cholesterol, on the other hand, does have a
clear effect on the membrane thickness, as the LX value of the cholesterol-containing
membranes is always 10-20% larger than that of the corresponding cholesterol-free
membranes. This thickening is partly caused by the increased amount of anesthetic molecules
located in the middle of the membrane (see Fig. 6), which pushes the two membrane leaflets
farther away from each other (as evidenced by, e.g., the previously discussed splitting of the
chain terminal C atom density profile of even the longer lipid components in the presence of

15
ACCEPTED MANUSCRIPT

anesthetics, see Fig. 4). However, cholesterol induces a considerable membrane thickening
even in the lack of anesthetics. Hence, this effect is partly caused by the well known ordering
effect of cholesterol on the lipid membranes (seen also in Fig. 2) [65,66,95-97,104-107],
namely that the large rigid ring system of the cholesterol molecules restricts the
conformational flexibility of the nearby lipid chains, and hence increase their order and make
them more elongated. This point is further addressed in a subsequent section.
In a clear contrast to LX, however, the area per lipid value does show a clear,
systematic change both with the presence of anesthetics and with the pressure. Thus, the

PT
addition of anesthetic molecules results in a clear lateral swelling of the membrane (i.e.,
increase of Alip) in every case, with the exception of SF in the PSM-cholesterol mixed

RI
membrane, and this change is always reverted by the increase of the pressure. It is also clear

SC
that cholesterol induces a strong lateral contraction of the membrane in every case; in this
respect the effect of cholesterol is the opposite of that of the anesthetic molecules. This lateral
NU
contraction effect of cholesterol can be explained, together with the above discussed H-
bonding of cholesterol to the other lipids, also by the smaller cross-section area of the
cholesterol ring system than that of the two independent tails of DPPC or PSM, and by the
MA

aforementioned ordering effect of cholesterol on the nearby lipid tails, which, besides making
them longer, also decreases their cross-section area.
D

As it has been previously discussed in the literature [30,102], the anesthetic molecules
staying in the middle of the bilayer push the lipids farther away from each other along the
E

membrane normal, increasing thus the thickness of the membrane (although the membrane
PT

thickness is influenced by the ordering of the tails, as well). However, the swelling of the
membrane is caused by the anesthetic molecules staying in their other preferred position,
CE

closer to the headgroups among the lipid chains, as their presence pushes the lipid tails
laterally farther away from each other. Our results are fully in line with this explanation:
AC

lateral swelling of the membrane is observed in every case when anesthetics are also
occupying their outer preferred position, but no such swelling is caused by SF in the PSM-
cholesterol mixed membrane, in which they exclusively stay in the middle of the membrane
(Fig. 6). This point is further addressed in a subsequent section.
Finally, the molecular volume of the membrane, determined as the product of the
membrane thickness (which does not show any clear trend with anesthetics or the pressure)
and the area per lipid (which does show such a clear trend), shows the same trend as the area
per lipid itself, i.e., it clearly increases upon addition of anesthetics, and decreases upon
increasing the pressure. Further, cholesterol has again an opposite effect on the membrane

16
ACCEPTED MANUSCRIPT

than anesthetics, as the cholesterol-induced thickening is overcompensated by the lateral


contraction of the membrane, resulting in a clear decrease of the membrane volume in every
case. Our present finding is in a clear accordance with the more than sixty years old critical
volume hypothesis of Mullins [12], but explains it in a different way than what was generally
thought in the second half of the past century. Thus, the increase of the membrane volume in
systems where anesthesia is supposed to occur (i.e., in the presence of anesthetics at ambient
pressure) as well as its decrease when the anesthetic effect is supposed to be reverted (i.e., at
high pressures) is governed by the similar changes of the cross-sectional area rather than the

PT
thickness of the membrane. Considering also that the change of the cross-sectional area is
related to the anesthetic molecules that are in their outer preferred position, our results suggest

RI
that the anesthetic molecules staying in this position are probably the ones that are primarily

SC
responsible for the anesthetic effect.
NU
3.3. Lateral mobility of the molecules

Membrane fluidity is often claimed to be a possible membrane property that might be


MA

related to the anesthetic effect [7,13,14,102]. To address this possibility, we have calculated
the lateral diffusion coefficient, Dlat, of the lipid and anesthetic molecules in the various
D

membranes simulated. Dlat has simply been calculated using the Einstein relation [108], i.e.,
Dlat = MSD/4t, where MSD is the mean square lateral displacement of the molecules within
E

the time t. The lipid molecules exited from the ballistic and entered to the diffusive regime
PT

after 1-1.5 ns, thus, the obtained MSD vs. t data could always be very well fitted by a straight
line above 1.5 ns.
CE

The obtained Dlat values corresponding to the lipid molecules are compared in Figure
8 as obtained in the different systems; all Dlat values are also collected in Table 3. As is seen,
AC

the presence of the anesthetic molecules always increases, while the increase of the pressure
always decreases the lateral diffusion coefficient of the lipid molecules. Not surprisingly, the
Dlat value of the cholesterol molecules is very close to that of the DPPC or PSM molecules in
every case. Our finding is thus consistent with the idea that the membrane fluidity, or lateral
mobility of the membrane constituents, is related to the anesthetic effect, as the anesthetic
induced increase of the lateral mobility is reverted at high pressure in every case. This
behavior is likely related to the above discussed trends in the lateral density of the membrane:
smaller lateral density (i.e., higher area per lipid) corresponds to larger space around the lipid
molecules, allowing them to move more freely, while larger lateral density naturally decreases

17
ACCEPTED MANUSCRIPT

the lateral mobility of the molecules. Consistently, the increase of the pressure thus reduces
substantially also the lateral mobility of the anesthetic molecules, see Table 3.

3.4. Order parameters

The thickness of the membrane does not only depend on the presence of anesthetic
molecules in the middle of the bilayer, but also on the orientational order of the lipid tails, i.e.,
on how straight they are elongated along the membrane normal axis. The order of the lipid

PT
chains can be characterized locally, i.e., around the individual C atoms, by the deuterium
order parameter, SCD, defined as [109,110]

RI
SC
1
S CD  3 cos 2  1 , (2)
2

where  is the angle formed by the C-H bond involving the given C atom and the membrane
NU
normal, and the brackets <...> denote ensemble average. It should be noted that SCD is an
experimentally accessible quantity, which can be measured by nuclear magnetic spectroscopy
MA

(NMR) using selectively deuterated samples.


To investigate the effect of anesthetics and the pressure on the local ordering of the
lipid chains, we have calculated the SCD profile along the DPPC and PSM tails in the various
D

systems simulated. In this analysis, we have numbered the C atoms of the hydrocarbon tails,
E

starting from that of the CH2 group bonded to the C=O or C-OH carbon atom, towards the tail
PT

end, as shown in Fig. 1. The order parameter profiles obtained from our simulations are
shown in Figure 9. It should be noted that in the profiles shown in Fig. 9 the SCD values are
CE

averaged over the two tails as well as over all the C-H bonds of the given C atom. However,
apart from the first two C atoms of the chains there is no considerable difference between the
AC

SCD profiles of the two chains, and this difference does not affect the conclusions that are
drawn from this analysis. Further, there is no noticeable difference between the SCD values
corresponding to different C-H bonds of the same C atom along the lipid tails.
To test the performance of the potential model used against experimental data in this
respect, we have also calculated the SCD values corresponding to the C atoms of the choline
headgroup and glycerol backbone of the DPPC molecule in the anesthetic-free DPPC bilayer.
For the  and  C atoms of the choline group (see Fig. 1), the values of SCD resulted in -0.03
and 0.06, respectively, in a good agreement with the corresponding experimental data of -0.05
and 0.03, respectively [111]. In the case of the g1 C atom of the glycerol backbone (Fig. 1),

18
ACCEPTED MANUSCRIPT

the two C-H bonds were found to correspond to markedly different SCD values, namely to 0
and 0.13, in an NMR measurement of a DPPC bilayer [112]. Similarly, in a POPC bilayer
these values were measured to be 0 and 0.14 [113]. Our simulation reproduces this forking
effect quite well, resulting in the SCD values of 0.04 and 0.18. The experimental order
parameters corresponding to the g2 and g3 glycerol C atoms (Fig. 1) of 0.20 and 0.23,
respectively [112], are also reasonably well reproduced by our simulation, the corresponding
values being 0.19 and 0.15-0.20, respectively. Further, the order parameter profile of the
DPPC tails in the neat bilayer agrees excellently with the experimental data, measured at the

PT
same temperature [114], as seen from the inset of Fig. 9. This agreement is certainly much
better than that between the results of our previous simulation with a GROMOS-based force

RI
field [45]. The excellent performance of the CHARMM36 force field in reproducing

SC
deuterium order parameters is in line with the observation of Botan et al. who found this
parameter set to give the best structural description of phosphatidylcholine lipids out of 13
NU
force fields [115].
When comparing the tail order parameter profiles of the different systems, the most
evident finding is that cholesterol has a marked ordering effect on the bilayer: in the
MA

cholesterol-containing systems the SCD profile goes up to 0.4, while in the cholesterol-free
systems it does not exceed 0.3. This ordering effect of cholesterol is known in the literature
D

[65,66,95-97,104-107], and can be explained by the presence of the large, rigid ring system of
the cholesterol molecules, which restricts the conformational flexibility of the nearby lipid
E

tails. This large difference in the order parameter profiles suggests, in accordance with the
PT

snapshots of Fig. 2, that while the cholesterol-free membranes are in the disordered , the
cholesterol-containing membranes are in the more ordered  phase. The effect of the presence
CE

of anesthetics is, however, not as clear as that of cholesterol. Thus, in general, DE is found to
make the lipid tails somewhat less ordered, and this effect is more marked in membranes
AC

containing cholesterol (and hence characterized by higher local tail order). However, this
effect is clearly model dependent, as in our previous study we found an ordering rather than
disordering effect of DE on the DPPC tails [45]. Further, the presence of SF does not change
the lipid tail order parameters noticeably. On the other hand, pressure has a clear ordering
effect on the lipid tails in every case.
The observed disordering effect of the anesthetics (or, at least, of DE) and the
opposite, ordering effect of the pressure on the lipid tails can again be related to the
anesthetics and pressure induced changes of the lateral density. Thus, lower lateral density
(induced by anesthetics) allows larger conformational flexibility of the lipid tails, resulting

19
ACCEPTED MANUSCRIPT

thus in lower order parameter values. Conversely, smaller lateral density, induced either by
cholesterol or the pressure, restricts the conformational flexibility, and thus enhances the
orientational order of the lipid tails.

3.5. Free volume and lateral pressure profiles

The alteration of the lateral pressure profile in the membrane by the presence of
anesthetic molecules is now generally thought to be a strong candidate for the membrane

PT
change that is behind the anesthetic effect [18,19,102]. Indeed, about two decades ago Cantor
provided a thermodynamic model showing how such changes in the lateral pressure profile

RI
might change the conformational equilibrium of certain membrane-bound (e.g., channel-

SC
forming) proteins, which can easily account for the phenomenon of anesthesia [18,19].
However, the lateral pressure profile, determined by the lateral interaction of the various
NU
atoms, is closely related to the packing of these atoms at various depths along the membrane
normal axis, which can be characterized by the profile of the fraction of the empty volume, ,
across the membrane. Thus, the behavior of the free volume profile might help us
MA

understanding the origin of the anesthetic-induced changes in the lateral pressure profile, as
well. For this purpose, we also analyze here the free volume profile across the membranes
D

simulated.
E

3.5.1. Free volume profiles


PT

To determine the fraction of the free volume in a computationally efficient way, we


have randomly placed about 5×105 test points in the sample configurations (the actual number
CE

of the test points has always been ten times the number of atoms in the system), and simply
calculated the ratio of the test points that are not covered by any atom and the total number of
AC

test points along the membrane normal axis. To decide whether a test point is covered by an
atom, the diameters of the atoms have been estimated by their Lennard-Jones diameter, . The
obtained profiles of the fraction of free volume, , across the membranes are shown in Figure
10. To magnify the differences between the different profiles, here we only show the
symmetrized profiles in one side of the bilayer. The profiles are characterized by a maximum
in the middle of the membrane, where the density is of minimum (see Fig. 3), and a minimum
between the middle of the membrane and the aqueous phase. In the cholesterol-free
membranes this minimum is located at about 18 Å from the middle of the bilayer, in the dense

20
ACCEPTED MANUSCRIPT

region of the headgroups. However, in cholesterol-containing membranes this minimum


occurs considerably closer to the middle of the membrane, in the region of the cholesterol
rings (see Fig. 4), and only a shoulder is seen in the headgroup region. This finding is in
accordance with our above explanation that the cholesterol induced lateral contraction of the
membrane can be attributed to the effect of the rigid and extensive cholesterol rings. It is also
clear that the fraction of the free volume in the middle of the bilayer is considerably larger in
the cholesterol-containing than in the cholesterol-free membranes, in accordance with our
above explanation that, due to the headgroup H-bonding, the shorter of the two lipids cannot

PT
reach the middle of the membrane.
Not surprisingly, the increase of the pressure always decreases the fraction of free

RI
volume across the entire membrane. Thus, anesthetic-induced changes that might be relevant

SC
for the explanation of the anesthetic effect should correspond to the increase of the free
volume. Such a free volume increase in the presence of anesthetics is seen at 8-16 Å from the
NU
middle of the bilayer in every case, as it is marked in Fig. 10. This region of free volume
increase coincides with the outer preferred position of the anesthetic molecules. This is
consistent with the previous finding of ours that the anesthetic induced lateral swelling of the
MA

membranes results in additional free volume, and this additional free volume is located
exclusively around the anesthetic molecules [49].
D

The anesthetic molecules located in the middle of the membrane do not induce such an
increase of the free volume, simply because they largely occupy the already existing, rather
E

large empty space, while the outer preferred position of the anesthetics corresponds to a low
PT

fraction of free volume (i.e., to the minimum of the profile in the cholesterol-containing, and
to a marked shoulder of the minimum in the cholesterol-free membranes). It is also seen that
CE

the anesthetic induced increase of the free volume in the outer preferred region of the
anesthetics is noticeably larger in the cholesterol-containing than in the cholesterol-free
AC

membranes (again, with the obvious exception of SF in the PSM-cholesterol mixed


membrane). This effect can be explained in a similar way, namely that the fraction of free
volume here is considerably smaller in the membranes containing cholesterol, i.e., there is
less pre-existing empty volume in these membranes, and hence the effect of the inclusion of
the anesthetics is also more marked here in this respect.

3.5.2. Lateral pressure profiles


The lateral pressure profiles calculated in the different membranes simulated are
shown in Figure 11. As is seen, the obtained profiles are considerably more structured in the

21
ACCEPTED MANUSCRIPT

cholesterol-containing than in the cholesterol-free membranes. The marked structure of the


pressure profiles in the hydrocarbon region of the cholesterol-containing membranes is again
a clear indication that these membranes are in the more ordered  rather than in the  phase.
The profiles in the cholesterol-free membranes are characterized by a marked minimum at
about 15-16 Å from the middle of the bilayer. The position of this minimum roughly
coincides with the location of the ester and amide O atoms of the DPPC and PSM molecules,
respectively (see Fig. 4).

PT
When analyzing the effect of the anesthetics and pressure on these profiles, it is
evident that the increase of the pressure shifts, in general, the entire profiles to higher values.
Therefore, changes of this profile that might be relevant for the molecular mechanism of

RI
general anesthesia should occur in the X ranges where anesthetics lower the lateral pressure.

SC
Such a region along the membrane normal axis occurs around the main minimum, i.e.,
14-18 Å from the middle of the bilayer, in the cholesterol-free membranes, and in the |X|
NU
range of about 20-24 Å in the cholesterol-containing membranes. It should be noted that, in
every case, this range of anesthetic-induced decrease of the lateral pressure is located slightly
farther from the middle of the bilayer than the range of the free volume increase (see Fig. 10),
MA

and roughly coincides with the range where the ester or amide O atoms are located (see Fig.
4). This finding is in a clear accordance with our previous results [49], and suggests that the
D

origin of this decrease of the lateral pressure is that the anesthetic molecules, located in their
E

outer preferred position, push the nearby lipid tails farther away from each other. However,
this effect leads to a marked decrease of the lateral pressure only in the nearby region, into
PT

which the anesthetic molecules already cannot penetrate, and hence, by increasing the local
lateral density, they cannot give rise to the lateral pressure any more. The present results
CE

indicate that if indeed the anesthetic-induced alteration of the lateral pressure profile, and the
corresponding conformational changes of certain membrane-bound proteins [18,19] are
AC

behind the molecular mechanism of general anesthesia, the relevant conformational changes
are expected to occur in the range of the lipid ester (or amide) O atoms along the membrane
normal, i.e., right at the polar side of the boundary between the apolar and polar phases of the
membrane.

4. Summary and conclusions

In this paper, we have analyzed in detail the effect of anesthetics as well as of the
pressure on the properties of lipid membranes of four different compositions by computer

22
ACCEPTED MANUSCRIPT

simulation methods. We have been focusing on such changes in the membrane properties that
(i) occur in the same direction upon adding any anesthetic molecule to any membrane, and (ii)
are reverted by increasing the pressure. In accordance with the results of other, similar studies
[102], we have found several such properties.
Anesthetic molecules are found to have a dual preference for positions along the
membrane normal axis. The first of these preferred positions is located in the middle of the
membrane. The preference for this position can be attributed to the fact that the density of the
membrane has its minimum, while the fraction of the empty volume has its maximum here.

PT
The other preferential position is located in the outer part of the apolar hydrocarbon region,
close to the dense region of the polar headgroups. The preference for this position is related to

RI
the favorable electrostatic interaction between the weakly polar anesthetic molecules and the

SC
polar lipid headgroups. The anesthetic molecules located in this region induce a lateral
swelling of the membrane by pushing the lipid chains farther away from each other, and thus
NU
creating some additional empty space here. This additional space has several consequences on
the membrane properties. Thus, it causes (i) a natural increase of the free volume fraction in
this region, (ii) allows a higher conformational flexibility of the lipid chains, and hence
MA

reduces their orientational order (although this effect is seen much less clearly than the other
ones), (iii) increases the lateral mobility of the lipid molecules, and (iv) decreases the lateral
D

pressure in the nearby region of the amide and ester groups, i.e., at the boundary of the apolar
and polar parts of the membrane, where the anesthetic molecules already cannot penetrate, but
E

their vicinity still pushes, on average, farther away the lipid molecules from each other in this
PT

crowded region. Further, since the lateral swelling of the membrane is clearly reverted by the
increase of the pressure, all these changes are also pressure reversible. It seems that although
CE

the increase of the lateral mobility and decrease of the orientational order of the lipid
molecules are in a clear correlation with the presence of the anesthetic effect, there might be
AC

no cause-effect relation between them. However, such a causality can well exist between the
changes in the lateral pressure profile and the anesthetic effect, as it was discussed in detail
earlier by Cantor [18,19].
An important result of the present study is that cholesterol acts, in many respects,
clearly in the opposite way as the anesthetic molecules. Thus, contrary to anesthetics,
cholesterol increases the lateral density, leading to a lateral contraction of the membrane, and
therefore it decreases the fraction of free volume in the outer preferred region of the
anesthetics, restricts the conformational flexibility of the lipid tails, and hence increases their
orientational order, and decreases also the lateral mobility of the lipid molecules. The

23
ACCEPTED MANUSCRIPT

cholesterol-induced lateral contraction of the membrane, being the cause of all of these
changes, can be explained both by the extensive H-bonding occurring between the
headgroups of cholesterol and the other lipid components (which keeps these molecules close
to each other), and the compact and rigid nature of the large cholesterol ring system. Further,
besides increasing the lateral density of the membrane, there is also another mechanism by
which cholesterol can reverse the effect of the anesthetic molecules. Namely, the
aforementioned extensive headgroup H-bonding and difference in the chain lengths of
cholesterol and the other lipid molecules, cholesterol effectively decreases the density in the

PT
middle of the membrane, altering thus the partitioning equilibrium of the anesthetic molecules
between their two preferred positions. Thus, the fraction of the anesthetic molecules staying

RI
in the outer preferred position, and hence being responsible for the anesthetic effect, is

SC
considerably smaller in the presence than in the absence of cholesterol.
Although cholesterol has a similar effect on the membrane properties in many respects
NU
than pressure, no cholesterol-induced reversal of anesthesia is known, although the membrane
of the living cells contains a considerable amount of cholesterol. However, due to its partial
miscibility with phospholipid molecules [53-60], cholesterol is not distributed uniformly in
MA

the cell membrane, instead, it largely occurs in ordered rafts of high cholesterol content,
which are surrounded by less ordered domains that are poor in, or free from cholesterol.
D

Further, as it was recently elegantly demonstrated by Reigada, anesthetic molecules prefer to


stay in these cholesterol-poor, disordered domains rather than in the cholesterol-rich rafts
E

[52]. Thus, the site of action of general anesthesia seems to be not simply the cell membrane,
PT

but, more specifically, its cholesterol-poor domains, and hence, although cholesterol has an
opposite effect on the membrane properties than general anesthetics, it seems to be irrelevant
CE

from the point of general anesthesia.


AC

Acknowledgements

This work has been supported by the Hungarian NKFIH Foundation under Project No.
119732.

24
ACCEPTED MANUSCRIPT

References

[1] H. Meyer, Zur Theorie der Alkoholnarkose, Naunyn-Schmiedebergers Archiv für


Experimentelle Pathologie und Pharmakologie 42 (1899) 109-118.
[2] E. Overton, Studien über die Narkose zugleich ein Beitrag zur allgemeinen
Pharmakologei, Gustav Fischer Verlag, Jena, 1901.
[3] F. H. Johnson, E. A. Flagler, Hydrostatic pressure reversal of narcosis in tadpoles,
Science 112 (1950) 91-92.

PT
[4] S. M.Johnson, K. W.Miller, Antagonism of pressure and anaesthesia, Nature 228 (1970) 75-
76.

RI
[5] M. J. Lever, K. W. Miller, W. D. M. Paton, E. B. Smith, Pressure reversal of

SC
anaesthesia, Nature 231 (1971) 368-371.
[6] M. J. Halsey, B. Wardley-Smith, Pressure reversal of narcosis produced by
NU
anaesthetics, narcotics and tranquillisers, Nature 257 (1975) 811-813.
[7] J. R. Trudell, D. G. Payan, J. H. Chin, E. N. Cohen, The antagonistic effect of an
inhalation anesthetic and high pressure on the phase diagram of mixed dipalmitoyl-
MA

dimyristoylphosphatidylcholine bilayers, Proc. Natl. Acad. Sci. USA 72 (1975) 210-


213.
D

[8] N. P. Franks, W. R. Lieb, Where do general anaesthetics act? Nature 274 (1978) 339-342.
[9] N. P. Franks, W. R. Lieb, Molecular and cellular mechanisms of general anaesthesia,
E

Nature 367 (1994) 607-614.


PT

[10] D. C. Mitchell, J. T. R. Lawrence, B. J. Litman, Primary alcohols modulate the


activation of the g protein-coupled receptor rhodopsin by a lipid-mediated mechanism,
CE

J. Biol. Chem. 271 (1996) 19033-19036.


[11] S. J. Mihic, Q. Ye, M. J. Wick, V. V. Koltchine, M. D Krasowski, S. E. Finn, M. P.
AC

Mascia, C. F. Valenzuela, K. K. Hanson, E. P. Greenblatt, R. A. Harris, N. L.


Harrison, Sites of alcohol and volatile anaesthetic action on GABAA and glycine
receptors, Nature 389 (1997) 385-389.
[12] L. J. Mullins, Some physical mechanisms in narcosis. Chem. Rev. 54 (1954) 289-323.
[13] B. J. Forrest, D. K. Rodham, An anaesthetic-induced phosphatidylcholine hexagonal
phase, Biochim. Biophys. Acta 814 (1985) 281-288.
[14] A. S. Janoff, K. W. Miller, in D. Chapman (Ed.), Biological Membranes, Academic
Press, London, 1982, pp. 417-476.

25
ACCEPTED MANUSCRIPT

[15] T. Heimburg, A. D. Jackson, The thermodynamics of general anesthesia, Biophys. J.


92 (2007) 3159-3165.
[16] K.Græsbøll, H. Sasse-Middelhoff, T. Heimburg, The thermodynamics of general and
local anesthesia, Biophys. J. 106 (2014) 2143-2156.
[17] E. Gray, J. Karslake, B. B. Machta, S. L. Veatech, Liquid general anesthetics lower
critical temperatures in plasma membrane vesicles, Biophys. J. 105, 2013, 2751-2759.
[18] R. S. Cantor Lateral pressures in cell membranes: A mechanism for modulation of
protein function, J. Phys. Chem. B 101 (1997) 1723-1725.

PT
[19] R. S. Cantor The lateral pressure profile in membranes: a physical mechanism of
general anesthesia, Biochemistry 36 (1997) 2339-2344.

RI
[20] D. A. Haydon, B. M. Hendry, S. R. Levinson, J. Requena, The molecular mechanisms

SC
of anaesthesia, Nature 268 (1977) 356-358.
[21] N. P. Franks, W. R. Lieb, The structure of lipid bilayers and the effects of general
NU
anaesthetics, J. Mol. Biol. 133 (1979) 469-500.
[22] J. R. Trudell, W. L. Hubbell, E. N. Cohen, The effect of two inhalation anesthetics on
the order of spin-labelled phospholipid vesicles, Biochim. Biophys. Acta 291 (1973)
MA

321-327.
[23] K. Tu, M. Tarek, M. L. Klein, D. Scharf, Effects of anaesthetics on the structure of a
D

phospholipid bilayer: molecular dynamics investigation of halothane in the hydrated


liquid crystal phase of dipalmitoylphosphatidylcholine, Biophys. J. 75 (1998) 2123-
E

2134.
PT

[24] L. Koubi, M. Tarek, M. L. Klein, D. Scharf, Distribution of halothane in a


dipalmitoylphosphatidylcholine bilayer from molecular dynamics calculations,
CE

Biophys. J. 78 (2000) 800-811.


[25] B. W. Lee, R. Faller, A. K. Sum, I. Vattulainen, M. Patra, M Karttunen,. Structural
AC

effects of small molecules on phospholipid bilayers investigated by molecular


simulation, Fluid Phase Equilibria 228-229 (2005) 135-140.
[26] M. Patra, E Salonen,. E. Terama, I. Vattulainen, R Faller,. B. W. Lee, J. Holopainen,
M. Karttunen, Under the influence of alcohol: The effect of ethanol and methanol on
lipid bilayers, Biophys. J. 90 (2006) 1121-1135.
[27] B. Griepernau, R. Böckmann, The influence of 1-alkanols and external pressure on the
lateral pressure profiles of lipid bilayers, Biophys. J. 95 (2008) 5766-5778.

26
ACCEPTED MANUSCRIPT

[28] A. A. Gurtovenko, J. Anwar, Interaction of ethanol with biological membranes: The


formation of non-bilayer structures within the membrane interior and their
significance, J. Phys. Chem. B 113 (2009) 1983-1992.
[29] R. D. Booker, A. K. Sum, Biophysical changes induced by xenon on phospholipid
bilayers, Biochim. Biophys. Acta 1828 (2013) 1347-1356.
[30] Y. Moskovitz, H. Yang, Modelling of noble anaesthetic gases and high hydrostatic
pressure effects in lipid bilayers, Soft Matter 11 (2015) 2125-2138.
[31] N. F. Hauet, F. Artzner, F. Boucher, C. Grabielle-Madelmont, I. Cloutier, G. Keller, P.

PT
Lesieur, D. Durand, M. Paternostre, Interaction between artificial membranes and
enflurane, a general volatile anesthetic: DPPC-enflurane interaction, Biophys. J. 84

RI
(2003) 3123-3137.

SC
[32] S. E. Feller, C. A. Brown, D. T. Nizza, K. Gawrisch, Nuclear Overhauser
enhancement spectroscopy cross-relaxation rates and ethanol distribution across
NU
membranes, Biophys. J. 82 (2002) 1396-1404.
[33] J. Chanda, S. Bandyopadhyay, Distribution of ethanol in a model membrane: A
computer simulation study, Chem. Phys. Letters 392 (2004) 249-254.
MA

[34] J. Chanda, S. Bandyopadhyay, Perturbation of phospholipid bilayer properties by


ethanol at a high concentration, Langmuir 22 (2006) 3775-3781.
D

[35] A. N. Dickey, R. Faller, How alcohol chain-length and concentration modulate


hydrogen bond formation in a lipid bilayer, Biophys. J. 92 (2007) 2366-2376.
E

[36] B. Griepernau, S. Leis, M. F Schneider,. M. Sikor, D. Steppich, R. A. Böckmann, 1-


PT

Alkanols and membranes: A story of attraction, Biochim. Biophys. Acta 1768 (2007)
2899-2913.
CE

[37] H. Jerabek, G. Pabst, M. Rappolt, T. Stockner, Membrane-mediated effect on ion


channels induced by the anesthetic drug ketmine, J. Am. Chem. Soc. 132 (2010) 7990-
AC

7997.
[38] M. Pickholz, L. Saiz, M. L. Klein, Concentration effects of volatile anesthetics on the
properties of model membranes: A coarse-grain approach, Biophys. J. 88 (2005) 1524-
1534.
[39] K. J. Oh, M. L. Klein, Effects of halothane on dimyristoylphosphatidylcholine lipid
bilayer structure: A molecular dynamics simulation study, Bull. Korean Chem. Soc.
30 (2009) 2087-2092.
[40] J. R. Wieteska, P. R. L. Welche, K. M. Tu, M. ElGamacy, G. Csanyi, M. C. Payne, P.
L. Chau, Isoflurane does not aggregate inside POPC bilayers at high pressure:

27
ACCEPTED MANUSCRIPT

Implications for pressure reversal of general anaesthesia, Chem. Phys. Letters 638
(2015) 116-121.
[41] L. M. Stimson, I. Vattulainen, T. Róg, M. Karttunen, Exploring the effect of xenon on
biomembranes, Cell. Mol. Biol. Letters 10 (2005) 563-569.
[42] R. Reigada, Influence of chloroform in liquid-ordered and liquid-disordered phases in
lipid membranes, J. Phys. Chem. B 115 (2011) 2527-2535.
[43] E. Yamamoto, T. Akimoto, H. Shimizu, Y. Hirano, M. Yasui, K. Yasuoka, Diffusive
nature of xenon anesthetic changes properties of a lipid bilayer: Molecular dynamics

PT
simulations, J. Phys. Chem. B 116 (2012) 8989-8995.
[44] M. Darvas, P. N. M. Hoang, S. Picaud, M. Sega, P. Jedlovszky, Anesthetic molecules

RI
embedded in a lipid membrane: A computer simulation study, Phys. Chem. Chem.

SC
Phys. 14 (2012) 12956-12969.
[45] B. Fábián, M. Darvas, S. Picaud, M. Sega, P. Jedlovszky, The effect of anesthetics on
NU
the properties of a lipid membrane in the biologically relevant phase: A computer
simulation study, Phys. Chem. Chem. Phys. 17 (2015) 14750-14760.
[46] J. Chen, L. Chen, Y. Wang, Z. Wang, S. Zeng, Exploring the effects on lipid bilayer
MA

induced by noble gases via molecular dynamics simulations, Sci. Rep. 5 (2015)
17235-1-6.
D

[47] H. V. Ly, D. E. Block, M. L. Longo, Interfacial tension effect of ethanol on lipid


bilayer rigidity, stability, and area/molecule: A micropipet aspiration approach,
E

Langmuir 18 (2002) 8988-8995.


PT

[48] H. V. Ly, M. L. Longo, The influence of short-chain alcohols on interfacial tension,


mechanical properties, area/molecule, and permeability of fluid lipid bilayers,
CE

Biophys. J. 87 (2004) 1013-1033.


[49] B. Fábián, M. Sega, V. P. Voloshin, N. N. Medvedev, P. Jedlovszky, Lateral pressure
AC

profile and free volume properties of phospholipid membranes containing anesthetics,


J. Phys. Chem. B 121 (2017) 2814-2824.
[50] E. Terama, O. H. S. Ollila, E. Salonen, A. C. Rowat, C. Trandum, P. Westh, M. Patra,
M. Karttunen, I. Vattulainen, Influence of ethanol on lipid membranes: From lateral
pressure profiles to dynamics and partitioning, J. Phys. Chem. B 112 (2008) 4131-
4139.
[51] R. D. Porasso, W. F. Drew Bennett, S. D. Oliveira-Costa, J. J. López-Cascales, Study
of the benzocaine transfer from aqueous solution to the interior of a biological
membrane, J. Phys. Chem. B 113 (2009) 9988-9994.

28
ACCEPTED MANUSCRIPT

[52] R. Reigada, Atomistic study of lipid membranes containing chloroform: Looking for a
lipid-mediated mechanism of anesthesia, Plos One 8 (2013) e52631-1-10.
[53] M. R. Vist, J. H. Davis, Phase equilibria of cholesterol/
2
dipalmitoylphosphatidylcholine mixtures: H nuclear magnetic resonance and
differential scanning calorimetry, Biochemistry 29 (1990) 451-464.
[54] E. Sackmann, Biological membranes: Architecture and function, in: R. Lipowsky, E.
Sackmann (Eds.), Structure and dynamics of membranes, Elsevier, Amsterdam, 1995,
pp. 1-64.

PT
[55] T. P. W. McMullen, R. N. McElhaney, New aspects of the interaction of cholesterol
with dipalmitoylphosphatidylcholine bilayers as revealed by high-sensitivity

RI
differential scanning calorimetry, Biochim. Biophys. Acta 1234 (1995) 90-98.

SC
[56] J. P. Slotte, Lateral domain heterogeneity in cholesterol/phosphadidylcholine
monolayers as a function of cholesterol concentration and phosphatidylcholine chain
NU
length, Biochim. Biophys. Acta 1238 (1995) 118-126.
[57] A. Radhakrishnan, H. M. McConnell, Cholesterol-phospholipid complexes in
membranes, J. Am. Chem. Soc. 121 (1999) 486-487.
MA

[58] A. Radhakrishnan, H. M. McConnell, Condensed complexes of cholesterol and


phospholipids, Biophys. J. 77 (1999) 1507-1517.
D

[59] S. L. Keller, H. M. McConnell, Stripe phases in lipid monolayers near a miscibility


critical point, Phys. Rev. Lett. 82 (1999) 1602-1605.
E

[60] F. Richter, L. Finegold, G. Rapp, Sterols sense swelling in lipid bilayers, Phys. Rev. E
PT

59 (1999) 3483-3491.
[61] P. A. Hyslop, B. Morel, R. D. Sauerheber, Organization and interaction of cholesterol
CE

and phosphatidylcholine in model bilayer membranes, Biochemistry 29 (1990) 1025-


1038.
AC

[62] J. M. Smaby, M. M. Momsen, H. L. Brockman, R. E. Brown, Phosphatidylcholine


acyl unsaturation modulates the decrease in interfacial elasticity induced by
cholesterol, Biophys. J. 73 (1997) 1492-1505.
[63] A. Kusumi, M. Tsuda, T. Akino, S. Osnishi, Y. Terayama, Protein-phospholipid-
cholesterol interaction in the photolysis of invertebrate rhodopsin, Biochemistry 22
(1983) 1165-1170.
[64] M. Y. El-Sayed, T. A. Guion, M. D. Fayer, Effect of cholesterol on viscoelastic
properties of dipalmitoylphosphatidylcholine multibilayers as measured by a laser-
induced ultrasonic probe, Biochemistry 25 (1986) 4825-4832.

29
ACCEPTED MANUSCRIPT

[65] G. Lindblom, L. Johansson, G. Arvidson, Effect of cholesterol in membranes. pulsed


nuclear magnetic resonance measurements of lipid lateral diffusion, Biochemistry 20
(1981) 2204-2207.
[66] B. A. Cornell, M. Keniry, The effect of cholesterol and gramicidin A' on the carbonyl
groups of dimyristoylphosphatidylcholine dispersions, Biochim. Biophys. Acta 732
(1983) 705-710.
[67] J. F. Nagle, R. Zhang, S. Tristram-Nagle, W. Sun, H. I. Petrache, R. Suter, M. X-ray
structure determination of fully hydrated L phase dipalmitoylphosphatidylcholine

PT
bilayers, Biophys. J. 70 (1996) 1419-1431.
[68] J. F. Nagle, S. Tristram-Nagle, Structure of lipid bilayers, Biochim. Biophys. Acta

RI
1469 (2000) 159-195.

SC
[69] J. B. Klauda, R. M. Venable, J. A. Freitas, J. W. O’Connor, D. J. Tobias, C.
Mondragon-Ramirez, I. Vorobyov, A. D. MacKerrel Jr., R. W. Pastor, Update of the
NU
CHARMM all-atom additive force field for lipids: Validation on six lipid types, J.
Phys. Chem. B 114 (2010) 7830-7843.
[70] J. B. Lim, B. Rogaski, J. B. Klauda, Update of the cholesterol force field parameters in
MA

CHARMM, J. Phys. Chem. B 116 (2012) 203-210.


[71] R. M. Venable, A. J. Sodt, B. Rogaski, H. Rui, E. Hatcher, A. D. MacKerrel Jr., R. W.
D

Pastor, J. B. Klauda, CHARMM all-atom additive force field for sphingomyelin:


E

Elucidation of hydrogen bonding and positive curvature. Biophys. J. 107 (2014) 134-
145.
PT

[72] M. J. Arcario, C. G. Mayne, E. Tajkhorshid, Atomistic models of general anesthetics


for use in in silico biological studies, J. Phys. Chem. B 118 (2014) 12075-12086.
CE

[73] E. Neria, S. Fischer, M. Karplus, Simulation of activation free energies in molecular


systems, J. Chem. Phys. 105 (1996) 1902–1921.
AC

[74] A. D. MacKerell, D. Bashford, M. Bellott, R. L. Dunbrack, J. D.Evanseck, M. J. Field,


S. Fischer, , J. Gao H. Guo, S. Ha, All-atom empirical potential for molecular
modeling and dynamics studies of proteins, J. Phys. Chem. B 102 (1998) 3586–3616.
[75] U. Essman, L. Perera, M. L. Berkowitz, T. Darden, H. Lee, L. G. Pedersen, A smooth
particle mesh Ewald method, J. Chem. Phys. 103 (1995) 8577-8594.
[76] J. P. Ryckaert, G. Ciccotti, H. J. C. Berendsen, Numerical integration of the Cartesian
equations of motion of a system with constraints: Molecular dynamics simulation of n-
alkanes, J. Comp. Phys. 23 (1977) 327-341.

30
ACCEPTED MANUSCRIPT

[77] S. Pronk, S. Páll, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, M. R. Shirts, J. C.


Smith, P. M. Kasson, D. van der Spoel, B. Hess, E. Lindahl, GROMACS 4.5: A high-
throughput and highly parallel open source molecular simulation toolkit,
Bioinformatics 29 (2013) 845–854.
[78] The code is freely available from https://github.com/Marcello-
Sega/gromacs/tree/virial/.
[79] S. Nosé, A molecular dynamics method for simulations in the canonical ensemble,
Mol. Phys. 52 (1984) 255-268.

PT
[80] W. G. Hoover, Canonical dynamics: equilibrium phase-space distributions, Phys. Rev.
A 31 (1985) 1695-1697.

RI
[81] M. Parrinello, A. Rahman, Polymorphic transitions in single crystals: A new

SC
molecular dynamics method, J. Appl. Phys. 52 (1981) 7182-7190.
[82] S. Jo, T. Kim, V. G. Iyer, W. Im, CHARM-GUI: A web-based graphical user interface
NU
for CHARMM, J. Comp. Chem. 29 (2008) 1859-1865.
[83] C. Anézo, A. H. de Vries, H. D. Höltje, D. P. Tieleman, S. J. Marrink, Methodological
issues in lipid bilayer simulations, J. Phys. Chem. B 107 (2003) 9424-.9433.
MA

[84] J. Sonne, F. Y. Hansen, G. H. Peters, Methodological problems in pressure profile


calculations for lipid bilayers, J. Chem. Phys. 122 (2005) 124903-1-9.
D

[85] A. Harasima, Molecular theory of surface tension, Adv. Chem. Phys. 1 (1958) 203-
237.
E

[86] M. Sega, B. Fábián, P. Jedlovszky, Layer-by-layer and intrinsic analysis of molecular


PT

and thermodynamic properties across soft interfaces, J. Chem. Phys. 143 (2015)
114709-1-8.
CE

[87] M. Sega, B. Fábián, P. Jedlovszky, Pressure profile calculation with mesh Ewald
methods, J. Chem. Theory Comput. 12 (2016) 4509-4515.
AC

[88] E. Lindahl, O. Edholm, Spatial and energetic-entropic decomposition of surface


tension in lipid bilayers from molecular dynamics simulations, J. Chem. Phys. 113
(2000) 3882-3893.
[89] J. Gullingsrud, K. Schulten, Lipid bilayer pressure profiles and mechanosensitive
channel gating, Biophys. J. 86 (2004) 3496-3509.
[90] M. Carrillo-Trip, S. E. Fellner, Evidence for a mechanism by which -3
polyunsaturated lipids may affect membrane protein function, Biochemistry 44 (2005)
10164-10169.

31
ACCEPTED MANUSCRIPT

[91] O. H. S. Ollila, T. Róg, M. Karttunen, I. Vattulainen, Role of sterol type on lateral


pressure profiles of lipid membranes affecting membrane protein functionality:
Comparison between cholesterol, desmosterol, 7-dehydrocholesterol and ketosterol, J.
Struct. Biol. 159 (2007) 311-323.
[92] S. Ollila, M. T. Hyvönen, I. Vattulainen, Polyunsaturation in lipid membranes:
Dynamic properties and lateral pressure profiles, J. Phys. Chem. B 111 (2007) 3139-
3150.
[93] O. Berger, O. Edholm, F. Jähnig, Molecular dynamics simulations of a fluid bilayer of

PT
dipalmitoylphosphatidylcholine at full hydration, constant pressure, and constant
temperature, Biophys. J. 72 (1997) 2002-2013.

RI
[94] W. C. Hung, M. T. Lee, F. Y. Chen, H. W. Huang, The condensing effect of

SC
cholesterol in lipid bilayers, Biophys. J. 92 (2007) 3960-3967.
[95] A. M. Smondyrev, M. L. Berkowitz, Structure of dipalmitoylphosphatidylcholine/
NU
cholesterol bilayer at low and high cholesterol concentrations: Molecular dynamics
simulation, Biophys. J. 77 (1999) 2075-2089.
[96] E. Falck, M. Patra, M. Karttunen, M. T. Hyvönen, I. Vattulainen, Lessons of slicing
MA

membranes: Interplay of packing, free area, and lateral diffusion in


phospholipid/cholesterol bilayers, Biophys. J. 87 (2004) 1076-1091.
D

[97] S. A. Pandit, S. Vasudevan, S. W. Chiu, R. J. Mashl, E. Jakobsson, H. L. Scott,


Sphingomyelin-cholesterol domains in phospholipid membranes: Atomistic
E

simulation, Biophys. J. 87 (2004) 1092-1100.


PT

[98] M. Doxastakis, A. K. Sum, J. J. de Pablo, J. J. Modulating membrane properties: The


effect of trehalose and cholesterol on a phospholipid bilayer, J. Phys. Chem. B 109
CE

(2005) 24173-24181.
[99] M. Pasenkiewicz-Gierula, T. Róg, K. Kitamura, A. Kusumi, Cholesterol effects on the
AC

phosphatidylcholine bilayer polar region: A molecular simulation study, Biophys. J.


78 (2000) 1376-1389.
[100] T. Róg, M. Pasenkiewicz-Gierula, Cholesterol-sphingomyelin interactions: A
molecular dynamics simulation study, Biophys. J. 91 (2006) 3756-3767.
[101] J. P. Slotte, The importance of hydrogen bonding in sphingomyelin's membrane
interactions with co-lipids, Biochim. Biophys. Acta 1858 (2016) 304-310.
[102] P. Jedlovszky, Simulation of membranes containing general anesthetics, in: M. L.
Berkowitz (Ed.), Biomembrane simulations. Computational studies of biological
membranes, Taylor and Francis, New York, in press, and references therein.

32
ACCEPTED MANUSCRIPT

[103] J. F. Nagle, Area/lipid of bilayers from NMR, Biophys. J. 64 (1993) 1476-1481.


[104] T. Parasassi, A. M. Giusti, M. Raimondi, E. Gratton, Abrupt modifications of
phospholipid bilayer properties at critical cholesterol concentrations, Biophys. J. 68
(1995) 1895-1902.
[105] S. W. Chiu, E. Jakobsson, R. J. Masl, H. L. Scott, Cholesterol-induced modifications
in lipid bilayers: A simulation study, Biophys. J. 83 (2002) 1842-1853.
[106] P. Jedlovszky, M. Mezei, Effect of cholesterol on the properties of phospholipid
membranes. 1. Structural features, J. Phys. Chem. B. 107 (2003) 5311-5321.

PT
[107] C. Hofsäß, E. Lindahl, O. Edholm, Molecular dynamics simulations of phospholipid
bilayers with cholesterol, Biophys. J. 84 (2003) 2192-2206.

RI
[108] M. P. Allen, D. J. Tildesley, Computer simulation of liquids, Clarendon Press, Oxford,

SC
1987.
[109] J. J. López-Cascales, J. García de la Torre, S. J. Marrink, H. J. C. Berendsen,
NU
Molecular dynamics simulations of a charged biological membrane, J. Chem. Phys.
104 (1996) 2713-2720.
[110] M. Mezei, P. Jedlovszky, Statistical thermodynamics via computer simulation to
MA

characterize phospholipid interactions in membranes, in: A. M. Dopico (Ed.), Methods


in molecular biology, vol. 400: Methods in membrane lipid, Humana Press, Totowa,
D

2007; pp. 127-144.


[111] H. U. Gally, W. Niederberger, J. Seelig, Conformation and motion of the choline head
E

group in bilayers of diplmitoyl-3-sn-phosphatidylcholine, Biochemistry 14 (1975)


PT

3647-3652.
[112] J. Seelig, H. U. Gally, R. Wohlgemuth, Orientation and flexibility of the choline head
CE

group in phosphatidylcholine bilayers, Biochim. Biophys. Acta 467 (1977) 109-119.


[113] T. M. Ferreira, F. Coreta-Gomes, O. H. S. Ollila, M. J. Moreno, W. L. C. Vaz, D.
AC

Topgaard, Cholesterol and POPC segmental order parameters in lipid membranes:


Solid state 1H-13C NMR and MD simulation studies, Phys. Chem. Chem. Phys. 15
(2013) 1976-1989.
[114] J. P. Douliez, A. Léonard, E. J. Dufourc, Restatement of Order Parameters in
Biomembranes: Calculation of C-C bond order parameters from C-D quadrupolar
splittings, Biophys. J. 68 (1995) 1727-1739.
[115] A. Botan, F. Favela-Rosales, P. F. J. Fuchs, M. Javanainen, M. Kanduč, W. Kulig, A.
Lamberg, C. Loison, A. Lyubartsev, M. S. Miettinen, L. Monticelli, J. Määttä, O. H.
S. Ollila, M. Retegan, T. Róg, H. Santuz, J. Tynkkynen, Toward atomistic resolution

33
ACCEPTED MANUSCRIPT

structure of phosphatidylcholine headgroup and glycerol bacckbone at different


ambient conditions, J. Phys. Chem. B 119 (2015) 15075-15088.

PT
RI
SC
NU
MA
E D
PT
CE
AC

34
ACCEPTED MANUSCRIPT

Tables

Table 1.
Interaction Parameters of the Anesthetic and Water Models Used.a

Molecule Atom /Å /kJ mol-1 q/e

PT
O 2.9400 0.4184 -0.340
C(CH2) 3.5814 0.2343 -0.010

RI
DE C(CH3) 3.6527 0.3264 -0.270
H(CH2) 2.3876 0.1464 0.090

SC
H(CH3) 2.3876 0.1004 0.090

O 2.9400 0.4184 -0.086


NU
C(CH) 3.5635 0.1139 0.003
C(CH2) 3.3854 0.2510 0.001
MA

C(CF3) 4.0981 0.0837 0.002


SF F(CH2F) 2.9043 0.5648 -0.202
F(CF3) 2.8509 0.4058 -0.048
D

H(CH) 2.3876 0.1882 0.300


E

H(CH2) 2.3520 0.1171 0.134


PT

O 3.1506 0.6364 -0.834


water
H 0.4000 0.1925 0.417
CE
AC

a
 and  are the Lennard-Jones distance and energy parameters, respectively, while q is the
fractional charge carried by the corresponding atomic site.

35
ACCEPTED MANUSCRIPT

Table 2.
Thickness, Area Per Lipid, and Volume Per Lipid of the Membranes Simulated

with DE with SF
quantity membrane without
anesthetics 1 bar 600 bar 1 bar 600 bar

DPPC 39.97±0.03 40.97±0.01 39.43±0.02 39.05±0.02 40.79±0.05

46.12±0.01 45.73±0.01 47.66±0.01 46.89±0.01 47.60±0.01

PT
DPPC+chol.
Lx/Å
PSM 40.51±0.01 39.66±0.01 40.84±0.02 41.38±0.01 41.94±0.02

RI
PSM+chol. 45.51±0.01 48.28±0.01 47.87±0.01 49.14±0.01 47.60±0.01

SC
DPPC 61.6 ± 0.9 70.5 ± 1.4
NU 66.6 ± 1.5 65.7 ± 0.9 58.3 ± 0.9

41.0 ± 0.3 46.5 ± 0.7 42.9 ± 0.5 43.7 ± 0.5 41.5 ± 0.3
DPPC+chol.
2
Alip/Å
PSM 55.8 ± 0.7 64.1 ± 0.9 60.0 ± 0.7 60.0 ± 0.7 55.8 ± 0.5
MA

PSM+chol. 40.8 ± 0.3 44.0 ± 1.0 41.3 ± 0.3 41.0 ± 0.3 40.4 ± 0.3
D

DPPC 1231 ± 17 1444 ± 28 1313 ± 29 1282 ± 17 1196 ± 19


E

946 ± 7 1062 ± 16 1022 ± 11 1024 ± 13 988 ± 8


DPPC+chol.
PT

Vlip/Å3
PSM 1130 ± 14 1271 ± 18 1224 ± 14 1241 ± 15 1162 ± 10
CE

PSM+chol. 929 ± 7 1063 ± 24 990 ± 7 1007 ± 7 962 ± 6


AC

36
ACCEPTED MANUSCRIPT

Table 3.
Lateral Diffusion Coefficients of the Molecules (in 10-3Å2/ps Units) in the Membranes Simulated
DPPC cholesterol PSM DE SF
without 1.87±0.02
anesthetics
with DE 4.21±0.01 95.16±0.05
1 bar
DPPC with DE 3.43±0.01 63.31±0.06
600 bar

PT
with SF 1.941±0.002 27.21±0.03
1 bar
with SF 0.501±0.001 9.20±0.02

RI
600 bar

SC
without 0.49±0.03 0.502±0.003
anesthetics
with DE 3.40±0.01 3.54±0.01 149.2±0.1
1 bar
NU
DPPC + with DE 0.61±0.01 0.735±0.001 114.4±0.1
cholesterol 600 bar
with SF 1.48±0.003 1.572±0.003 48.84±0.05
MA

1 bar
with SF 0.15±0.001 0.191±0.001 13.79±0.05
600 bar
D

without 0.237±0.001
anesthetics
E

with DE 1.507±0.002 76.18±0.02


PT

1 bar
PSM with DE 0.524±0.002 46.43±0.05
600 bar
CE

with SF 0.631±0.003 20.70±0.01


1 bar
with SF 0.196±0.001 6.15±0.01
AC

600 bar

without 0.187±0.001 0.134±0.001


anesthetics
with DE 1.661±0.007 1.418±0.006 163.2±0.1
1 bar
PSM +
with DE 0.470±0.001 0.535±0.001 112.8±0.1
cholesterol 600 bar
with SF 0.898±0.004 0.933±0.004 47.90±0.04
1 bar
with SF 0.148±0.001 0.155±0.001 17.20±0.07
600 bar

37
ACCEPTED MANUSCRIPT

Figure legends

Fig. 1. Schematic structure of DPPC, PSM, cholesterol, and the anesthetic molecules DE and SF.
The numbering scheme of the atoms used throughout this paper is also indicated.

Fig. 2. Equilibrium snapshot of the four membranes, simulated at 1 bar, that contain DE. The N, P,
O, and C atoms of the lipid molecules are shown by dark blue, yellow, red, and light blue colors,

PT
respectively. Water molecules are shown by red thin sticks, while the anesthetic DE molecules are
marked by green color, and are shown enlarged for better visibility. For clarity, H atoms are omitted

RI
from the snapshots.

SC
Fig. 3. Mass density profiles of the membranes simulated. Top left: DPPC bilayer, bottom left:
DPPC-cholesterol mixed bilayers, top right: PSM bilayers, bottom right: PSM-cholesterol mixed
NU
bilayers. Black solid curves: membranes without anesthetics (at 1 bar), blue circles: membranes with
DE, red squares: membranes with SF. Full and open symbols corresponds to membranes at 1 bar and
MA

600 bar, respectively. The inset shows the comparison of the electron density profile of the pure
DPPC bilayer at 1 bar, as obtained from our present simulation (full circles) with that of our earlier
simulation, performed at 310 K using the GROMOS force field [45] and with experimental data,
D

obtained at 323 K [66]. All profiles shown are symmetrized over the two leaflets of the bilayer.
E
PT

Fig. 4. Number density profiles of selected lipid atoms in the membranes containing (a) DPPC, and
(b) PSM, simulated at 1 bar. The notation of the atoms considered is shown in Fig. 1. All profiles
CE

shown are symmetrized over the two leaflets of the bilayer.

Fig. 5. Partial pair correlation functions of the cholesterol O atoms with the DPPC ester O, PSM
AC

amide O and N, and PSM sphingosine atoms (bottom panel), and between the corresponding O-H
and N-H atom pairs (top panel), as obtained in the anesthetic-free systems. Functions involving the
DPPC =O, DPPC –O-, PSM N, PSM =O, and PSM –O(H) groups are shown by black, red, green,
blue, and orange colors, respectively. Functions involving the cholesterol O and a H atom are shown
by filled circles, all other functions are shown by lines.

38
ACCEPTED MANUSCRIPT

Fig. 6. Mass density profiles of the anesthetic molecules across the membranes simulated. Top left:
DPPC bilayer, bottom left: DPPC-cholesterol mixed bilayers, top right: PSM bilayers, bottom right:
PSM-cholesterol mixed bilayers. Blue circles: membranes with DE, red squares: membranes with
SF. Full and open symbols corresponds to membranes at 1 bar and 600 bar, respectively. The data
corresponding to SF in the cholesterol-free and cholesterol-containing membranes are shifted up by
0.05 g/cm3 and 0.1 g/cm3, respectively, for better visibility. The inset shows the Gaussian
decomposition of the SF profile in the PSM bilayer at 1 bar (see the text). Open circles: simulated

PT
data, green lines: individual Gaussians, red line: sum of the individual Gaussians. All profiles shown
are symmetrized over the two leaflets of the bilayer.

RI
SC
Fig. 7. (a) Average thickness, (b) average area per lipid, and (c) average volume per lipid values of
the membranes simulated. Asterisks: membranes with DE, red squares: membranes with SF. Full
NU
and open symbols corresponds to membranes at 1 bar and 600 bar, respectively. Error bars are only
shown when larger than the symbols.
MA

Fig. 8. Lateral diffusion coefficients of the DPPC, PSM and cholesterol molecules in the membranes
simulated. Asterisks: membranes with DE, red squares: membranes with SF. Full and open symbols
D

corresponds to membranes at 1 bar and 600 bar, respectively. Symbols of darker and lighter shades
correspond to the DPPC or PSM and to the cholesterol molecules, respectively. Error bars are
E

always smaller than the symbols.


PT

Fig. 9. Deuterium order parameters along the hydrocarbon chains of the DPPC and PSM molecules
CE

in the membranes simulated. Top left: DPPC bilayer, bottom left: DPPC-cholesterol mixed bilayers,
top right: PSM bilayers, bottom right: PSM-cholesterol mixed bilayers. All values are averaged over
AC

the two chains of the molecules as well as over all the H atoms bonded to the given C atom. Error
bars are always smaller than the symbols. The numbering scheme of the C atoms is shown in Figure
1. The inset compares the data obtained from our simulation of the neat DPPC bilayer at 1 bar
(asterisks) with experimental data, measured also at 330 K (open circles) [114].

39
ACCEPTED MANUSCRIPT

Fig. 10. Profile of the fraction of empty volume across the membranes simulated. Top left: DPPC
bilayer, bottom left: DPPC-cholesterol mixed bilayers, top right: PSM bilayers, bottom right: PSM-
cholesterol mixed bilayers. Black solid curves: membranes without anesthetics (at 1 bar), blue
circles: membranes with DE, red squares: membranes with SF. Full and open symbols corresponds
to membranes at 1 bar and 600 bar, respectively. In order to magnify the details, the profiles, being
symmetrized over the two leaflets, are only show in one side of the bilayer. Encircled is the region in
which the change of the profile seems to be relevant from the point of anesthesia.

PT
Fig. 11. Profile of the lateral pressure component across the membranes simulated. Top left: DPPC

RI
bilayer, bottom left: DPPC-cholesterol mixed bilayers, top right: PSM bilayers, bottom right: PSM-

SC
cholesterol mixed bilayers. Black solid curves: membranes without anesthetics (at 1 bar), blue
circles: membranes with DE, red squares: membranes with SF. Full and open symbols corresponds
NU
to membranes at 1 bar and 600 bar, respectively. In order to magnify the details, the profiles, being
symmetrized over the two leaflets, are only show in one side of the bilayer. Encircled is the region in
which the change of the profile seems to be relevant from the point of anesthesia.
MA
E D
PT
CE
AC

40
ACCEPTED MANUSCRIPT

Conflict of interest

Besides the institutions of the authors, this work was only financed by the Hungarian NKFIH
(formerly OTKA) foundation, as stated in the Acknowledgement. We do not declare any
conflict of interest.

PT
RI
SC
NU
MA
E D
PT
CE
AC

41
ACCEPTED MANUSCRIPT

Highlights

• Lipid membranes are simulated with and without general anesthetics at 1 and 600 bar

• Properties that can be relevant in the molecular mechanism of anesthesia are identified

PT
• All the possible changes are originated from the lateral swelling of the membrane

RI
• Alteration of the lateral pressure profile is found in the range of the ester groups

SC
NU
• Effect of cholesterol in the respect of anesthesia is also addressed
MA
E D
PT
CE
AC

42
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11

You might also like