![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
Department of Chemistry and Biochemistry and Department of Pharmacology, University of California at San Diego, La Jolla, California 92093-0365
 |
Summary |
|---|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
Amphotericin B is a powerful but toxic antifungal antibiotic that is used to treat systemic infections. It forms ionic membrane channels in fungal cells. These antibiotic/sterol channels are responsible for the leakage of ions, which causes cell destruction. The detailed molecular properties and structure of amphotericin B channels are still unknown. In the current study, two molecular dynamic simulations were performed of a particular model of amphotericin B/cholesterol channel. The water and phospholipid environment were included in our simulations, and the results obtained were compared with available experimental data. It was found that it is mainly the hydrogen bonding interactions that keep the channel stable in its open form. Our study also revealed the important role of the intermolecular interactions among the hydroxyl, amino, and carboxyl groups of the channel-forming molecules; in particular, some hydroxyl groups stand out as new "hot spots" that are potentially useful for chemotherapeutic investigations. Our results also help to clarify why certain antibiotic derivatives, with a blocked amino group, are less active. We present a hypothesis for the role of membrane lipids and cholesterol in the channel.
| |
Introduction |
|---|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
AmB is a polyene macrolide antibiotic that is widely used in the treatment of systemic fungal infections (1) (Fig. 1, top). Despite its long clinical history, the molecular antifungal action of AmB is not well understood (2, 3). According to the most widely accepted mechanism, AmB molecules interact with membrane sterols to form channels (2-5). The channels are responsible for the leakage of monovalent ions, particularly K+, and other small molecules from the cell. The resulting irregular ionic distribution eventually causes cell death. The chemotherapeutic use of AmB is based on the higher affinity of this antibiotic for ergosterol (principal sterol in fungal cells) than for cholesterol (sterol in mammalian cells) (6). Because it also has high affinity for cholesterol-containing membranes, AmB is quite toxic, and its use in clinical treatment is limited to rather severe cases.
|
To reduce the toxic features of this powerful drug, its antifungal chemotherapeutic properties should be elucidated more precisely. Many experimental efforts (3, 7) designed to achieve a comprehensive understanding of the membranous AmB/sterol channels have shown that these channels are difficult to study; thus, there is little information available for the molecular level, and the channel structure remains unknown. In addition to experiments, in several recent theoretical studies (8-14), the focus has been on the isolated molecular properties of AmB or sterols (8-12). In some of these studies, the channel structure was also considered (13-15); however, only simple molecular mechanics or electrostatic calculations without proper treatment of the membrane/water environment were used. Ultimately, experimental efforts together with molecular modeling approaches should lead to a more complete understanding of the molecular action of AmB.
According to a proposed model (4), the AmB/sterol channel consists of eight AmB and eight sterol molecules. Experimental data and data from previous molecular modeling studies suggest that the diameter of the channel is 7-10 Å (16). Because the length of the SLC is less than the membrane thickness, it was postulated (4) that two SLCs in a bilayer configuration are needed to form a channel (DLC) spanning the entire membrane. It is now accepted that both SLC and DLC types of channels can exist and function, depending on the membrane environment and thickness and the availability of AmB. The SLC channels are observed to transfer ions, which means that they are able to span the entire membrane. It might be argued that lipids around the AmB channel arrange themselves in such a way the lipid molecules are slightly pressed in to obtain a lipid bilayer thickness approximately equal to the length of the SLC. A similar situation is expected to occur for the DLC; the lipids surrounding the channel are expected to be separated slightly to accommodate the difference between the channel length and membrane thickness. Because AmB is an amphoteric molecule with two charged groups, polar interactions would, to a large extent, forbid AmB molecules from crossing through the cell membrane. Based on this assumption, it has been postulated that an SLC is formed when AmB is present only on one side of the membrane, but a DLC is formed when there are AmB molecules on both sides of the membrane. There is no evidece yet for the channel formation mechanism, but Weakliem et al. (17) suggested that an SLC may form when a critical number of AmB molecules come together at a membrane surface.
In addition to being amphoteric, AmB is an amphiphilic molecule. The
polar and nonpolar electrostatic potential patterns of the AmB molecule
were studied semiquantitatively according to Poisson-Boltzmann methods
(18). On the basis of Baginski and Borowski (18) and other studies (4,
14, 15), it can be assumed that the polar hydroxyl groups of AmB (Fig.
1, top) are placed at the center of the channel and the
chain of conjugated C
C bonds interacts with the
phospholipid/sterol environment. It is known from structure-activity
studies that the polar head of the AmB molecule containing the
amino sugar portion and the carboxyl group are very important for
antibiotic activity; in particular, the presence of the protonatable
amino group has been determined to be essential (19). On the basis of
this observation, a model of interaction between AmB and sterol was
proposed (20); however, the validity of this proposal is unconfirmed.
Because AmB has amino, carboxyl, and many hydroxyl groups, hydrogen bonding is expected to be very important in the interaction of AmB with molecules in its vicinity. For example, inwardly pointed hydroxyl groups can interact with the solvent and with the cations inside the channel (15). It has been observed that the charged amino and carboxyl groups of AmB are involved in hydrogen bonding (14, 20). Similarly, it has been suggested that hydroxyl groups of the macrolide ring can form intermolecular hydrogen bonds (14). Even though they are expected to play an important structural role, current knowledge of hydrogen bonding in the channel complex is very limited. Unfortunately, experimental studies cannot easily reach this level of detail, and past theoretical studies have used only simple representations such as isolated molecules or AmB/AmB or AmB/sterol molecular pair complexes (10, 12, 21).
Several groups have tried to explain the origins of the observed higher affinity of AmB for ergosterol by using arguments based on structural differences (20, 22-24). Baginski et al. (8) and Langlet et al. (10) suggested that the more rigid molecular shape of ergosterol, which better complements AmB, is responsible for this higher affinity. Nevertheless, the nature of AmB/sterol interactions and positioning of the sterol in the channel complex remains unknown.
To characterize and further understand the structure and molecular properties of AmB channel, two separate MD simulations were performed of a particular model of AmB/cholesterol channel. In this study, cholesterol rather than ergosterol was included as the sterol; thus, the simulated system corresponds to a mammalian system. This choice was based on the fact that it is the formation of AmB/cholesterol channels that is responsible for the undesirable toxicity of AmB in patients. Further knowledge of the molecular factors that stabilize the structure of AmB/cholesterol channel may help to eliminate or reduce drug toxicity. Comparative studies should also be performed on the AmB/ergosterol channel; this will be the subject of a future project.
The structural information obtained in simulations was compared with the available experimental data. Our analysis emphasizes structural properties that could be important for the functioning and stability of the AmB/sterol channel in its open state and for its formation. The chemotherapeutic implications of our results are also discussed. In particular, we comment on why certain previous modifications of the amino and carboxyl group of AmB might have led to lower antibiotic activity or higher selectivity. Similarly, the roles of some other groups that contribute to the stability of the channel are pointed out. Modifications of such groups might be used in future pharmaceutical investigations.
| |
Materials and Methods |
|---|
|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
Molecular models. Our model for the AmB/cholesterol channel was built based on suggestions from the previous studies (4, 14). The channel consists of eight AmB and eight cholesterol molecules. Other stoichiometries were also considered at the initial stages of the project, but we chose the model with eight pairs of molecules because it gives the most compact structure if the channel diameter is kept within the range of 7-10 Å. The SLC was preferred over the DLC for a couple of reasons. First, it is a simpler and computationally less demanding model to investigate. Second, as discussed, it is biologically more relevant because under physiological conditions, antibiotic AmB molecules enter the cell only from one side (i.e., the external side).
Because the cell membranes in which AmB/sterol channels are formed contain phospholipids, two separate MD simulations using different membrane environments were performed. The MDSI used an SLC solvated in such a way that the inside and both entrances of the pore are occupied by water molecules. To mimic the lipid membrane environment, the second simulation (MDSII) also included one or two layers of phospholipids surrounding the AmB/cholesterol channel. Again, the inside and both sides of the pore contained water molecules, representing an aqueous solution. The lipids in MDSII were modeled with DMPC because DMPC has often been used in experimental membrane studies (25-27), as well as in theoretical studies (28-30), including the gramicidin membrane channel (31). In both simulations, all hydrogen atoms, polar or nonpolar, were explicitly included as interaction sites. The structure of the cholesterol was taken from a previous conformational study, and the conformer with an extended side chain (i.e., B3) was chosen (8). Similarly, the structure of AmB in its zwitterionic form was taken from an earlier work (32). Initial dihedral angle values for the amino sugar position were
(C42/O19/C19/C18) = 
58° and 
(C43/C42/O19/C19) = 126°.
In MDSI, a 1:1 AmB/cholesterol complex was built by optimally aligning
the sterol with AmB. Then, through appropriate replication, a symmetric
channel with eight AmB/cholesterol pair fragments was constructed. The
formed channel was energy minimized for 500 steps of steepest descent
followed by 1000 steps of conjugate gradient algorithms. The resulting
AmB/cholesterol channel (Fig. 1, bottom) was then solvated
with 504 water molecules in such a way that the pore was filled with
water molecules and both sides of the channel had approximately five
solvation layers. This soaking was achieved by immersing the channel in
an equilibrated box of water molecules and then deleting the solvent
molecules that were within 2.6 Å from any of the channel atoms and
deleting the water molecules that were not within a certain distance of
the ends of the channel. The water molecules on the sides of the
channel that would otherwise be occupied by the membrane were also
deleted. Locations of the incorporated water molecules were later
optimized by keeping the channel molecules rigid for 200 and 500 steps
of steepest descent and conjugate gradient, respectively. During the
equilibration MD runs (see the next section), 86 water molecules that
were only loosely bound broke away from the channel complex. These
"evaporated" water molecules also were deleted during the equilibration runs. Thus, there were a total of 2950 atoms in the MDSI
simulation. In the molecular simulations, the AmB/cholesterol channel
and solvating water molecules were treated as an isolated complex
(i.e., no periodic boundary conditions were applied). Although it is a
simple setup, use of the explicit water molecules inside the pore and
as several solvation layers on both ends of the channel, as well as
leaving the lipid region empty, roughly approximates the membranous
cell environment.
In setting up the second simulation (MDSII), phospholipids and
additional water molecules were added by using an equilibrated box of
solvated bilayer membrane containing 72 DMPC molecules (36 in each
layer). The latter coordinates were
taken3 from a 100-psec MD
simulation (28). The use of lipid coordinates from a thermally
equilibrated system enabled us to not repeat portions of the
equilibration simulation runs (33). In Damodaran and Merz (29),
nonpolar hydrogen atoms of the DMPC were modeled as part of the carbons
to which they are bonded (i.e., the united atom model was used for
those hydrogens). These missing hydrogens were added by using the
CHARMM program (34). This solvated box of phospholipids was then
overlapped with the starting configuration of the MDSI simulation. DMPC
molecules whose heavy atoms are within 2.0 Å of any of the channel
heavy atoms as well as the DMPC molecules inside the channel pore were
deleted. A criterion of the short distance of 2.0 Å was preferred to
avoid deletion of too many DMPC molecules because the spatial extent of
the lipid acyl chains can be quite large. Similarly, water molecules
that do not overlap with the AmB, cholesterol, or lipid molecules and
are within approximately the first five solvation layers of the channel
or the lipids were kept and the remainder were removed. The system with
the remaining 1,666 water molecules, 8 AmB molecules, eight cholesterol
molecules, and 34 DMPC molecules (17 in each layer) had a total of
10,706 atoms. The resulting initial structure of the channel complex was subsequently optimized for 500 and 1,000 steps of steepest descent
and conjugate gradient, respectively. During this structural minimization, only the lipid and water molecules were allowed to move.
Computations.
All minimization and MD simulations were
performed using the CHARMM molecular simulation program (34). Only the
bond, angle, dihedral, van der Waals, and electrostatic terms were
included in defining the interaction potentials. The dielectric
constant 
= 2 was used for all electrostatic interactions.
Interaction force field parameters were taken from the CHARMM potential
library version 24
1, except the AmB and cholesterol site charges
were taken from an earlier work. As described previously (18), AmB and
cholesterol site charges were obtained by finding the charge sets that
give the best fits to the electrostatic potentials generated using the
MOPAC'93 program (35). The TIP3P model (36) was chosen to represent
the water molecules, and its parameters were taken from CHARMM force
field version 241.
1 order of magnitude longer.
Unfortunately, it is a large and complicated system, so the cost of
running a simulation of this time scale would be at the order of 1000 CPU hr on 16 nodes of Cray T3D; such a high cost makes it practically
impossible. However, the time scale of our MD simulations is
sufficiently long to allow the study of certain structural properties
of the AmB channel at the microscopic level. In this respect, what is
reported here is the simulation of an already formed open channel that
has a diameter sufficiently large (7-8 Å in this case) to transfer
ions and small molecules (16). This, however, does not necessarily mean
that the channel also accommodates a "closed form" and switches periodically between the open and closed forms, as is observed for many
protein channels. Only long simulations would be able to show that the
channel may have more than one state and switches among them.
The MDSI and MDSII simulations were performed on SGI R4400 and Cray T3D
computers, respectively. In the latter case, 16 nodes were used. The
final MDSII 60-psec simulation took ~60 CPU hr on Cray T3D.
| |
Results |
|---|
|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
We focused our analysis of the MD simulation trajectories on investigating which intermolecular and intramolecular structural and dynamic properties of the AmB channel are important for the channel stability and its functional behavior. The MDSII simulation explicitly includes phospholipids and thus better represents the physical environment; results of this simulation are given. The results of MDSI are very similar to those of MDSII; existing differences are pointed out, and possible reasons for the differences are discussed.
Intermolecular properties. Because the sterols are required in the formation of AmB channels, intermolecular interactions of AmB molecules with other AmB molecules and with sterols are expected to be crucial for the stability of the channel. Visual inspection of the MD trajectories revealed that although the steric restraints are important, it is mainly the hydrogen bonding interactions that keep the channel intact.
Hydrogen bonds. On average, ~75% of the AmB/O8 hydroxyl groups form intermolecular hydrogen bonds with either O9 or O5 hydroxyl groups of the neighboring AmB molecules. The involvement of O8 hydroxyl groups in an intermolecular hydrogen bond chain (Fig. 2) is due to their favorable orientation, which is opposite the other hydroxyl groups in the polyhydroxyl chain (Fig. 1, top). This allows suitable alignment of this hydroxyl group toward the O9 or O5 atom of the adjacent AmB. As discussed below, all hydroxyls of the polyhydroxyl fragment either are involved in forming intramolecular hydrogen bonds or interact with solvent molecules.
|
|
Molecular structure of the channel. Visual analysis of the channel complex revealed that the smallest limiting diameter of the channel pore is defined by the O8 and O3 hydroxyl groups of AmB. However, it should be noted that the channel is not quite cylindrically symmetrical (Fig. 4) and looks slightly elliptical. Judging from the closest distances between O8---O8 and O3---O3 hydroxyls of the oppositely located AmB molecules and from the pore solvent cross section, we estimated the diameter of the channel to be 8.5-10.5 Å. The time dependence of this diameter (defined as the distance between two opposite AmB/O8 hydroxyls) is shown in Fig. 5.
|
|
4 Å) in the direction
perpendicular to the membrane, and their alignment is not necessarily at the same planar level. In addition, the vectorial direction of the
pore axis fluctuates significantly as a function of the simulation
time. The overall shape of the channel complex closely resembles a
barrel formed by twisted and interlocking AmB molecules (Fig.
6). Molecular alignment of AmB molecules
with respect to each other has almost a regular up-and-down pattern.
This crown-type relative positioning of AmB molecules helps to
establish the highly favorable intermolecular hydrogen bond ring formed
by the O8 and O9 or O5 hydroxyl groups of adjacent molecules. As
discussed, this hydrogen bond chain stays intact for almost the entire
simulation and seems to be crucial in keeping the channel in its open
state. Shifted placement of AmB molecules aligns the amino and carboxyl groups of two adjacent molecules in an orientation favoring the hydrogen bond formation between the two groups.
|
90° about
their long axis. The two simulations differed in this rotational motion
of cholesterol molecules; sterol molecules were observed to rotate more
often in the MDSI. This difference is due to the absence of
phospholipids in MDSI. On the basis of the frequency of sterol
rotations (and considering for MDSII the equilibration run lasting 60 psec as well as the data collection run), the estimated time scale of
the cholesterol rotations about their long axes were 25 and ~100 psec
for MDSI and MDSII, respectively. Because the lipid environment was not
incorporated, the MDSI result for this particular property should be
taken with caution. An upper bound for the correlation time 
of the
wobbling motion of cholesterol around its long axis was measured
experimentally using NMR spectroscopy as 
10
9 sec = 1000 psec (37). The agreement
with experiments for the estimated time scale of cholesterol rotational
motion and the observed swinging-type rotations of cholesterols from
+90° to 90° make us believe that regardless of its simplicity,
the membrane model formed by one or two layers of phospholipids was
fairly successful in mimicking the physical membranous environment.
The cholesterol molecules rotate quite freely, and there does not seem
to be a specific way in which they interact with other molecules.
Cholesterol molecules interact simultaneously with both AmB and lipid
molecules, and it appears that their major purpose is to shield the
interaction between the AmB channel and phospholipids. This
interaction-buffering behavior of cholesterol was postulated previously
on the basis of NMR data (27). MD simulations also showed that
cholesterol molecules are rigid up to their side chains and rotate
around their main axis as a rigid plane. Only the tail of the side
chain of cholesterol extending from C22 (see Refs. 8 and 38 for carbon
atom numbering) was observed to have a small bending motion. There is
similar experimental evidence by Dufourc et al. (39), who
reported the rigidity of cholesterol up to its C22---C24 atom.
Solvation of the pore.
The AmB molecules are oriented such
that the hydroxyl groups form the pore surface. There are ~75 water
molecules in the cylindrical part of the channel. If, as above, a
hydrogen bond is defined as having a bond distance of 2.7 Å and a
bond angle of 140°, there are 23 ± 7 hydrogen bonds between
AmB and water molecules. These bonds are present during the entire
MDSII simulation (Fig. 3, bottom). It was observed that
water molecules quite freely enter and leave the channel. For example,
during the 60-psec MDSII run, ~15-pore water molecules were exchanged
with the bulk water. Such high mobility of water molecules is due to
the facts that as reflected in sizable fluctuations, the AmB channel is
not very restrictive, and because the pore diameter is large (~10
Å), the pore water molecules can retain their bulk properties to a
substantial degree.
Role of the lipids. In our simulations, only a thin layer of lipids was used. Therefore, the interaction between AmB and DMPC lipid molecules cannot be studied quantitatively, and the results must be used with caution. However, several interesting features were observed that might be responsible for the channel formation or destruction. Because the distance between the amino and carboxyl groups of AmB and the distance between the phosphate and amino groups of DMPC are almost equal, AmB and DMPC molecules can form different yet energetically favorable configurations. As a result, a DMPC molecule either interacts directly with one AmB molecule, or it can act as an electrostatic bridge in the AmB/DMPC/AmB configuration. Therefore, one might expect that AmB molecules would be anchored to the membrane surface through individual electrostatic interactions with phospholipids, and when the AmB concentration exceeds a certain level, the bridging configuration becomes favorable and the channel starts to form. Such bridging configurations were actually observed in our simulations. By acting like a structural buffer, cholesterol molecules screen the interaction between AmB and DMPC molecules. This may slow down the AmB lateral migration or across the membrane and gives them the opportunity to regroup to form a channel. One may expect the observed interactions between AmB and DMPC molecules to not only help create the channel but also be responsible for destruction of the channel. Depending on the movement and positioning of the cholesterol molecules, either the formation of an active open channel is favored or an operational channel is destroyed.
Intramolecular properties. Because the construction and functioning of the channel depend on the dynamic behavior of constituent molecules, we also investigated the structural intramolecular properties of AmB and cholesterol.
Position of the AmB amino sugar.
The amino sugar of AmB is the
most flexible fragment of the antibiotic due to allowed rotation around
C19---O41 and O41---C42 bonds. The amino sugar is positioned with
regard to the aglycon part of AmB such that typical values of the
dihedral angles and (Fig. 1, top) were 90° and
±180°, respectively. Distribution of the and angles
extracted from the MDSII trajectory for all eight AmB molecules is
presented in Fig. 7, top. The
angle (, ) distribution was very similar in the MDSI simulation.
This spatial positioning of the amino sugar helps to preserve the
intermolecular AmB/AmB hydrogen bonds between the amino group of one
AmB molecule and the carboxyl group of another AmB molecule or between
hydroxyl group O43 of one AmB molecule and hydroxyl group O15 of
another AmB molecule (Fig. 2). It should be mentioned that the position of the amino sugar is slightly different than that found
crystallographically or through molecular simulation of an isolated AmB
molecule and its derivative (11, 39). The current conformation was
predicted in a previous detailed conformational analysis of AmB (32), but data from different sources are not in conflict. As in its crystal
state (39), an isolated AmB molecule tends to form an intramolecular
hydrogen bond between amino and carboxyl group with = 88° and
=142°. In contrast, in the channel complex, an AmB interacts with
its AmB neighbor and forms intermolecular hydrogen bonds using the same
polar groups. In this case, the position of amino sugar is different,
but its conformation in the channel complex is still sterically allowed
(32).
|
Conformational change in the polyhydroxyl part of AmB.
The
hydroxyl groups in the polyhydroxyl part of AmB can form intramolecular
hydrogen bonds that can stabilize the all-trans structure of
the entire carbon atom chain (Fig. 1, top). However, this
uniform polyhydroxyl chain configuration is altered at C6 and C7. Our
simulations showed that the chain can flip around the C6---C7 bond.
Typical dynamic behavior of the dihedral angle defined by four carbon
atoms (C5, C6, C7, and C8) is presented for two different AmB molecules
in Fig. 7, bottom. The conformational change shown in Fig.
7, bottom (solid line), is related to the formation of intramolecular or intermolecular hydrogen bond; when C5/C6/C7/C8 dihedral angle is close to ±180°, the O8 hydroxyl group
can form intermolecular hydrogen bond with either the O9 or O5 hydroxyl
group of the neighboring AmB molecule. In contrast, when this dihedral
angle is 90°, an intramolecular hydrogen bond between O8 and O9
hydroxyl group is formed. For some of the AmB molecules, this torsional
angle changed from trans to gauche during the
equilibration stage of the simulation. Data in Fig. 7,
bottom, show that this change is reversible, and the time
scale of such conformational flip is ~100 psec.
Twisting of the conjugated double-bond system.
The aglycon
ring is the most rigid part of AmB. This rigidity is due mainly to the
presence of seven conjugated CC double bonds in the macrolide ring
of AmB. However, this long conjugated system is not very planar, and it
can twist. It was found in the MDSII simulation that the improper
dihedral angle defined by C20, C25, C28, and C33 fluctuates and differs
from ideal planarity by ~±30°. The flexibility of this conjugated
double-bond system can be important for the interaction of AmB
molecules with sterol molecules.
| |
Discussion |
|---|
|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
Two MD simulations for a particular model of the AmB channel were performed to study the molecular properties of the channel in its open state. These simulations revealed certain structural and dynamic properties of the channel at the molecular level. Our findings are summarized below.
When the diameter of the AmB channel is kept in the range of experimental estimates (16), the channel formed by eight AmB and eight cholesterol molecules is stable, due mainly to strong hydrogen bonding interactions. On the basis of an analysis of the shape of the pore and its nonsymmetric behavior, it can be postulated that the stoichiometry proposed by de Kruijff and Demel (4) is not necessarily the only possibility. Channels formed by fewer than eight molecules can also exist. Even though the channel with seven AmB molecules would be the most likely alternative, because adjacent AmB molecules are shifted along each other, channels with an even number of AmB molecules have the advantage of forming a proper up-down-up crown-type structure. Therefore, with the current level of knowledge, six to eight would be the best answer to the question of how many AmB molecules are involved in the channel formation.
The existence of the intermolecular AmB/AmB hydrogen bonds in the central part of the channel was proposed by Khutorsky (14) and supported by our investigation of the dynamic behavior of these bonds (Fig. 2). It was found that the out-of-plane O8 hydroxyl group is responsible for the formation of the intermolecular bonds between AmB molecules. This intermolecular hydrogen bond is formed when adjacent AmB molecules are mutually shifted along their vertical pore axis. In addition, it was found that the formation of this bond correlates with the conformational change in the C6---C7 bond. Taking into account all the data, it can be stated that formation of intermolecular hydrogen bonds by O8 hydroxyl groups contributes to the stability in the central part of the channel.
The chain of hydrogen bonds between amino and carboxyl groups was also postulated previously (14). The results of the current study further support the finding that this hydrogen bond chain is very stable. These observations of the hydrogen bonding patterns are very important for chemotherapeutic studies because they indicate which amino and/or carboxyl group modifications might have pharmaceutical importance (7). The MD simulations further revealed that it is, again, the strong hydrogen bonding interaction between amino and carboxyl group that stabilizes the channel complex at the pore entrance. This observation helps to explain experimental data showing poor activity of some AmB derivatives with the modified amino group (19). Based on the current simulations, it can be explained that the derivatives of AmB with blocked amino groups cannot form appropriate hydrogen bonds that stabilize the channel structure and therefore a durable channel cannot be formed. Actually, the experimental data (19) suggest that substitution of the carboxyl group and alkylation of the amino group and/or shift of this group improve antifungal versus hemolytic selectivity. The result of our calculations can explain these experimental findings. When the carboxyl group is blocked [e.g., methyl ester or propyl amide derivatives of AmB (19)], the amino group may still form intermolecular hydrogen bonds with other polar groups (e.g., the hydroxyl group O15 of the neighbor AmB molecule). When the amino group is shifted or substituted [e.g., dimethylglycyl or trimethyl derivatives of AmB (19)], the bulky substituents limit the hydrogen bond formation. Such suggestion may be derived from the analysis of mutual orientation of the AmB molecules (Fig. 2) and the trajectory animation of the channel (not shown). For these types of AmB derivatives, hydrogen bonds formed by the altered amino group are not as strong as the bonds formed between free amino and carboxyl groups of the neighboring AmB molecules. These weaker hydrogen bonds probably make the structure of the AmB/cholesterol channel less stable than the AmB/ergosterol channel that results in the improved selectivity; to confirm this, comparable AmB/ergosterol channels must be studied. Nevertheless, as discussed above, results of our simulation readily explain the observed effects of various modifications of AmB undertaken in other laboratories. The goal was to make new AmB derivatives with bulky substituents at the amino group that should still be able to form weak hydrogen bonds. These efforts seem to be very promising. A new, much more selective second generation of derivatives were obtained and tested (40).
Another modification worth testing would be to make ion channels composed of amphotericin B dimers formed by covalently bonding the amino group of one monomer with the carboxyl group of the other.4 This might be done using appropriate-sized linkers rather than directly bonding the groups, to allow freer molecular movement. The biophysical and biological studies in membranes with such dimers could show whether the channel forms faster or at lower concentrations. Such an experiment may confirm the mutual orientation of the amino and carboxyl groups in the channel, for which we derived at estimates on the basis of the current simulation.
A second chain of hydrogen bonds between hydroxyl groups O15 and O43 belonging to neighboring AmB molecules was detected at the entrance to the channel. These hydrogen bonds also contribute to the stabilization of the channel. Elimination of these bonds may disturb the structure of the AmB/cholesterol channel and help to reduce AmB toxicity, providing the same modification does not influence stability of AmB/ergosterol channel. This information can be very useful from a chemotherapeutic point of view. The hydroxyl groups O15 and O43 have not been modified previously, and this prediction of the simulations that indicates that they are new "hot spots" can be further tested experimentally. Concerning the interactions between AmB molecules in the channel, we suggest eliminating these hydroxyl groups or substituting related (e.g., methoxy or methyl) groups. However, we are aware that selective modification of the hydroxyl groups involves a very difficult chemical synthesis task. Other AmB hydroxyl groups are very similar; probably only through total synthesis from fragments can AmB derivatives be prepared with modified hydroxyl groups in certain positions.
The MD simulations also give information about the locations of the amino sugars that cause adjacent polar groups to reorient themselves to form as many hydrogen bonds as possible. Torsional angles defining the position of sugar moiety agree well with previous studies (32). Moreover, it was found that these dihedral angles have different values in an isolated AmB molecule and in an AmB molecule that is part of the channel complex.
Analysis of the molecular simulation results shows that the cholesterol molecules do not interact with AmB molecules in any specific way; therefore, it can be stated that cholesterol is a rather nonspecific target for AmB. This property of cholesterol has been measured experimentally through NMR (27). The hypothesis given by Herve et al. (20) that predicts a specific interaction of cholesterol hydroxyl group with AmB amino or carboxyl groups is not supported by the current results. Thus, the explanation of why the affinity of AmB molecules to cholesterol is lower than that to ergosterol requires similar studies for the AmB/ergosterol/DMPC system. In addition to the theoretical simulations, the NMR solid state experiments can be undertaken; with NMR solid state measurements (27), comparative studies can be performed of AmB/cholesterol and AmB/ergosterol channels (in liposomes or lipid bilayers) to determine differences in mobility and orientation of both sterols along the AmB molecules in the channel, which can be simulated later using theoretical approaches. State-of-the-art NMR experiments can also be expected to determine the orientation of the AmB molecules in the channel. The latter results would support or reject the molecular simulations data.
The results of the AmB channel molecular simulations suggest that it is mainly the hydrogen bonding interactions that keep the channel in its open state. Results regarding the rotational motion of cholesterol and the lack of any specific interaction between cholesterol and AmB agree well with available experimental data. Our analysis also pointed out several hydroxyl groups that can have a role in the functioning of AmB, and we presented a hypothesis for the role of membrane phospholipids.
Our study is only a small step toward resolution of this complex system. The current understanding of the AmB/sterol channel and its molecular properties is only at a qualitative or semiquantitative level. Many important questions in this field concerning the structure and properties of the channel are unanswered; future experimental and molecular modeling studies are needed for a more comprehensive understanding.
| |
Acknowledgments |
|---|
We thank Alan J. Robinson for supplying the coordinates of the phospholipids, M. Montal for his help during initial stages of the project, and the anonymous reviewer for suggestions concerning modifications of the AmB channels.
| |
Footnotes |
|---|
Received March 10, 1997; Accepted June 29, 1997
1 Current affiliation: Department of Pharmaceutical Technology and Biochemistry, Technical University of Gdansk, 80-952 Gdansk, Poland.
2 Current affiliation: Department of Physics, Koç University, Istinye, Istanbul 80860, Turkey.
This work was supported by National Science Foundation, National Institutes of Health, and National Science Foundation Supercomputer Centers Metacenter Program Grants to J.A.M. M.B. was supported by a fellowship from the Fulbright Foundation.
3 Alan J. Robinson, personal communication.
4 This suggestion was made by an anonymous reviewer, to whom we are grateful.
Send reprint requests to: J. Andrew McCammon, Ph.D., Department of Chemistry and Biochemistry/Department of Pharmacology, 4238 Urey Hall, University of California, San Diego, La Jolla, CA 92093-0365. E-mail: jmccamo{at}ucsd.edu
| |
Abbreviations |
|---|
AmB, amphotericin B; SLC, single-layer channel; MD, molecular dynamics; MDSI, first molecular dynamics simulation; MDSII, second molecular dynamics simulation; DMPC, 2,3-dimyristoyl-D-glycero-1-phosphorylcholine.
| |
References |
|---|
|
Summary
Introduction
Materials & Methods
Results
Discussion
References
|
|---|
| 1. | Gallis, H. A., R. H. Drew, and W. W. Pickard. Amphotericin B: 30 years of clinical experience. Rev. Infect. Dis. 12:308-329 (1990)[Medline]. |
| 2. |
Brajtburg, J.,
W. G. Powderly,
G. S. Kobayashi, and
G. Medoff.
Amphotericin B: current understanding of mechanisms of action.
Antimicrob. Agents Chemother.
34:183-188 (1990) |
| 3. | Hartsel, S. C., C. Hatch, and W. Ayenew. How does amphotericin B work? Studies on model membrane systems. J. Liposome Res. 3:377-408 (1993). |
| 4. | De Kruijff, B. and R. A. Demel. Polyene antibiotic-sterol interactions in membranes of Acholeplasma laidlawii cells and lecithin liposomes. III. Molecular structure of the polyene antibiotic-cholesterol complexs. Biochim. Biophys. Acta 339:57-70 (1974)[Medline]. |
| 5. | Bolard, J. How do the polyene macrolide antibiotics affect the cellular membrane properties? Biochim. Biophys. Acta 864:257-304 (1986)[Medline]. |
| 6. | Teerlink, T., B. De Kruijff, and R. A. Demel. The action of pimaricin, etruscomycin and amphotericin B on liposomes with varying sterol content. Biochim. Biophys. Acta 599:484-492 (1980)[Medline]. |
| 7. | Schaffner, C. P. Amphotericin B derivatives, in Recent Trends in the Discovery, Development and Evaluation of Antifungal Agents (R. A. Fromtling, ed.). L. R. Prous Science Publishers, Barcelona, 595-632 (1987). |
| 8. | Baginski, M., A. Tempczyk, and E. Borowski. Comparative conformational analysis of cholesterol and ergosterol by molecular mechanics. Eur. Biophys. J. 17:159-166 (1989)[Medline]. |
| 9. | Baginski, M., P. Bruni, and E. Borowski. Comparative analysis of the distribution of the molecular electrostatic potential for cholesterol and ergosterol. J. Mol. Struct. 311:285-296 (1994). |
| 10. | Langlet, J., J. Berges, J. Caillet, and J. P. Demaret. Theoretical study of the complexation of amphotericin B with sterols. Biochim. Biophys. Acta 1191:79-93 (1994)[Medline]. |
| 11. | Mazerski, J. and E. Borowski. Molecular dynamics of amphotericin B. I. Single molecule in vacuum and water. Biophys. Chem. 54:49-60 (1995). |
| 12. | Anachi, R. B., M. Bansal, K. R. K. Easwaran, K. Namboodri, and B. P. Gaber. Molecular modeling studies on amphotericin B and its complex with phospholipid. J. Biomol. Struct. Dyn. 12:957-970 (1995)[Medline]. |
| 13. | Bonilla-Marin, M., M. Moreno-Bello, and I. Ortega-Blake. A microscopic electrostatic model for the amphotericin B channel. Biochim. Biophys. Acta 1061:65-77 (1991)[Medline]. |
| 14. | Khutorsky, V. E. Structures of amphotericin B-cholesterol complex. Biochim. Biophys. Acta 1108:123-127 (1992)[Medline]. |
| 15. |
Khutorsky, V. E.
Ion coordination in the amphotericin B channel.
Biophys. J.
71:2984-2995 (1996) |
| 16. | De Kruijff, B., W. J. Gerritsen, A. Oerlemans, R. A. Demel, and L. L. M. VanDeenen. Polyene antibiotic-sterol interactions in membranes of Acholeplasma laidlawii cells and lecithin liposomes. I. Specificity of the membrane permeability changes induced by the polyene antibiotics. Biochim. Biophys. Acta 339:30-43 (1974)[Medline]. |
| 17. | Weakliem, C. L., G. Fujii, J.-E. Chang, A. Ben-Shaul, and W. M. Gelbart. Effect of tension on pore formation in drug-containing vesicles. J. Am. Chem. Soc. 99:7694-7697 (1995). |
| 18. | Baginski, M. and E. Borowski. Distribution of electrostatic potential around amphotericin B and its membrane targets. J. Mol. Struct. 389:139-146 (1997). |
| 19. | Cheron, M., B. Cybulska, J. Mazerski, J. Grzybowska, A. Czerwinski, and E. Borowski. Quantitative structure-activity relationships in amphotericin B derivatives. Biochem. Pharmacol. 37:827-836 (1988)[Medline]. |
| 20. | Herve, M., J. C. Debouzy, E. Borowski, B. Cybulska, and C. M. Gary-Bobo. The role of the carboxyl and amino groups of polyene macrolides in their interactions with sterols and their selective toxicity: a 31P-NMR study. Biochim. Biophys. Acta 980:261-272 (1989). |
| 21. | Mazerski, J. and E. Borowski. Molecular dynamics of amphotericin B. II. Dimer in water. Biophys. Chem. 57:205-217 (1996)[Medline]. |
| 22. | Nakamura, T., M. Nishikawa, K. Inoue, S. Nojima, T. Akiyama, and U. Sankawa. Phosphatidylcholine liposomes containing cholesterol analogues with side chains of various lengths. Chem. Phys. Lipids 26:101-110 (1980)[Medline]. |
| 23. |
Clejan, S. and
R. Bittman.
Rates of amphotericin B and filipin association with sterols: a study of changes in sterol structure and phospholipid composition of vesicles.
J. Biol. Chem.
260:2884-2889 (1985) |
| 24. | Haynes, M. P., P. L.-G. Chong, H. R. Buckley, and R. A. Pieringer. Fluorescence studies on the molecular action of amphotericin B on susceptible and resistant fungal cells. Biochemistry 35:7983-7992 (1996)[Medline]. |
| 25. | Bolard, J. Interaction of polyene antibiotics with membrane lipids: physicochemical studies of the molecular basis of selectivity. Drugs Exp. Clin. Res. 12:613-618 (1986)[Medline]. |
| 26. | Balakrishnan, A. R. and K. R. K. Easwaran. Lipid-amphotericin B-complex structure in solution: a possible first step in the aggregation process in cell membranes. Biochemistry 32:4139-4144 (1993)[Medline]. |
| 27. | Dufourc, E. J., J. C. P. Smith, and H. D. Jarrell. Amphotericin and model membranes: the effect of amphotericin B on cholesterol-containing systems as viewed by 2H-NMR. Biochim. Biophys. Acta 776:317-329 (1984)[Medline]. |
| 28. |
Damodaran, K. V. and
K. M. Merz.
A comparison of DMPC- and DLPE-based lipid bilayers.
Biophys. J.
66:1076-1087 (1994) |
| 29. |
Robinson, A. J.,
W. G. Richards,
P. J. Thomas, and
M. M. Hann.
Head group and chain behavior in biological membranes: a molecular dynamics computer simulation.
Biophys. J.
67:2345-2354 (1994) |
| 30. |
Chiu, S.-W.,
M. Clark,
V. Balaji,
S. Subramaniam,
H. L. Scott, and
E. Jakobsson.
Incorporation of surface tension into molecular dynamics simulation of an interface: a fluid phase lipid bilayer membrane.
Biophys. J.
69:1230-1245 (1995) |
| 31. | Woolf, T. B. and B. Roux. Structure, energetics, and dynamics of lipid-protein interactions: a molecular dynamics study of the gramicidin A channel in DMPC bilayer. Proteins Struct. Funct. Genet. 24:92-114 (1996). [Medline] |
| 32. | Baginski, M., P. Gariboldi, P. Bruni, and E. Borowski. Conformational analysis of amphotericin B. Biophys. Chem., in press. |
| 33. | Pastor, R. W. and R. M. Venable. Molecular and stochastic dynamics simulation of lipid membranes, in Computer Simulation of Biomolecular Systems. Theoretical and Experimental Applications: Vol. 2 (W. F. Van Gunstern, P. K. Weiner and A. J. Wilkinson, eds.). ESCOM, Leiden, 443-463 (1993). |
| 34. | Brooks, B. R., R. E. Bruccoleri, B. D. Olafson, D. J. States, S. Swaminathan, and M. Karplus. CHARMM: a program for macromolecular energy, minimization, and dynamics calculations. J. Comput. Chem. 4:187-217 (1983). |
| 35. | Stewart, J. J. P. MOPAC 93, QCPE program 455. Quantum Chemistry Program Exchange, Bloomington, IN (1993). |
| 36. | Jorgensen, W. L., J. Chandrasekhar, J. D. Madura, R. W. Impey, and M. L. Klein. Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79:926-935 (1983). |
| 37. | Dufourc, E. J. and I. C. P. Smith. A detailed analysis of the motion of cholesterol in biological membranes by 2H NMR relaxation. Chem. Phys. Lipids 41:123-135 (1986)[Medline]. |
| 38. | Dufourc, E. J., E. J. Parish, S. Chitrakorn, and I. C. P. Smith. Structural and dynamical details of cholesterol-lipid interaction as revealed by deuterium NMR. Biochemistry 23:6062-6071 (1984). |
| 39. | Ganis, P., G. Avitabile, W. Mechlinski, and C. P. Schaffner. Polyene macrolide antibiotic amphotericin B: crystal structure of the N-iodoacetyl derivative. J. Am. Chem. Soc. 93:4560-4564 (1971)[Medline]. |
| 40. | Borowski, E., J. Grzybowska, J. Gumieniak, P. Sowinski, B. Cybulska, T. Zieniawa, F. Mignini, F. Santacroce, and J. Bolard. Second Generation of Amphotericin B Derivatives of Low Toxicity: Design and Properties. American Society for Microbiology, 251 (1995). |
This article has been cited by other articles:
|
|
 |
|
  A. Nedal, H. Sletta, T. Brautaset, S. E. F. Borgos, O. N. Sekurova, T. E. Ellingsen, and S. B. Zotchev Analysis of the Mycosamine Biosynthesis and Attachment Genes in the Nystatin Biosynthetic Gene Cluster of Streptomyces noursei ATCC 11455 Appl. Envir. Microbiol., November 15, 2007; 73(22): 7400 - 7407. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Czub and M. Baginski Comparative Molecular Dynamics Study of Lipid Membranes Containing Cholesterol and Ergosterol Biophys. J., April 1, 2006; 90(7): 2368 - 2382. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Huang, Z. Zhang, X. Han, J. Tang, J. Wang, S. Dong, and E. Wang Ion Channel Behavior of Amphotericin B in Sterol-Free and Cholesterol- or Ergosterol-Containing Supported Phosphatidylcholine Bilayer Model Membranes Investigated by Electrochemistry and Spectroscopy Biophys. J., December 1, 2002; 83(6): 3245 - 3255. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
|
 |
 |
 |
