| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Biophys J, September 2002, p. 1501-1510, Vol. 83, No. 3
*Department of Biological Sciences, University of Calgary, Calgary,
Alberta T2N 1N4, Canada; and Department of Bioscience and
Biotechnology, Drexel University, Philadelphia, Pennsylvania 19104 USA
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
The spontaneous formation of the phospholipid bilayer underlies the permeability barrier function of the biological membrane. Tears or defects that expose water to the acyl chains are spontaneously healed by lipid lateral diffusion. However, mechanical barriers, e.g., protein aggregates held in place, could sustain hydrophobic defects. Such defects have been postulated to occur in processes such as membrane fusion. This gives rise to a new question in bilayer structure: What do the lipids do in the absence of lipid lateral diffusion to minimize the free energy of a hydrophobic defect? As a first step to understand this rather fundamental question about bilayer structure, we performed molecular dynamic simulations of up to 10 ns of a planar bilayer from which lipids have been deleted randomly from one monolayer. In one set of simulations, approximately one-half of the lipids in the defect monolayer were restrained to form a mechanical barrier. In the second set, lipids were free to diffuse around. The question was simply whether the defects caused by removing a lipid would aggregate together, forming a large hydrophobic cavity, or whether the membrane would adjust in another way. When there are no mechanical barriers, the lipids in the defect monolayer simply spread out and thin with little effect on the other intact monolayer. In the presence of a mechanical barrier, the behavior of the lipids depends on the size of the defect. When 3 of 64 lipids are removed, the remaining lipids adjust the lower one-half of their chains, but the headgroup structure changes little and the intact monolayer is unaffected. When 6 to 12 lipids are removed, the defect monolayer thins, lipid disorder increases, and lipids from the intact monolayer move toward the defect monolayer. Whereas this is a highly simplified model of a fusion site, this engagement of the intact monolayer into the fusion defect is strikingly consistent with recent results for influenza hemagglutinin mediated fusion.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
The hydrophobic effect is central to the
formation of biological membranes, both lipid bilayer formation and
membrane protein folding. The simplest description of the lipid bilayer
is that all of the molecules are arranged to minimize the contact of
their nonpolar atoms with water. However, the linear arrays of lipids shown in texts, aligned head to head and tail to tail, ignores the
dynamic fluctuations that are being elucidated by molecular dynamics
simulations (Tieleman et al., 1997
; Feller, 2000; Lindahl and Edholm,
2001; Moore et al., 2001). One situation where these fluctuations must
be important is during membrane fusion.
It is believed that the initial bilayer destabilization leading to
fusion is initiated either through a high curvature bending defect
(Chernomordik et al., 1998; Kozlov and Chernomordik, 1998; Siegel,
1999; Melikyan et al., 1999; Lentz and Lee, 2000) or a hydrophobic
defect (Bentz, 2000a,b; Bentz and Mittal, 2000). These proposals agree
on most points, including the appearance of lipid stalks (Chernomordik,
1996; Kozlov and Chernomordik, 1998; Kuzmin et al., 2001; Kozlovsky and
Kozlov, 2002; Markin and Albanesi, 2002) to finally connect the outer
monolayers. The question is how do the stalks form in the first place?
What is the transition state for the formation of the stalk? It is
generally believed that fusion proteins undergo an essential
conformational change to create the initial defect for fusion
(Skehel and Wiley, 1998; Eckert and Kim, 2001). The recipient of
the energy released by this conformational change is the transition
state of interest.
For the fusion mediated by influenza hemagglutinin (HA), an aggregate
of eight or more HAs forms of which only two undergo a slow essential
conformational change, which allows the first fusion pore to form
(Bentz, 2000a; Mittal and Bentz, 2001; Mittal et al., 2002b). Kozlov
and Chernomordik (1998) proposed a model for HA aggregation, wherein
the fusion peptides are embedded in the viral envelope and the nascent
conformational change to the extended coiled coil induces a curvature
in the viral bilayer. An aggregate of HAs can stabilize this curvature
in the form of a lipid dome with a "crown" of HAs. Kozlov and
Chernomordik (1998) proposed that this dome interacts directly with the
target membrane, and outer monolayer destabilization relieves the
curvature stress at the tip of the dome. However, Bentz (2000b) argued
that the fact that only two of the eight or more HAs in the fusogenic
aggregate undergo the essential conformational change seems more
consistent with a hydrophobic defect than a curvature defect. Within
the HA aggregate proposed by Kozlov and Chernomordik (1998), the
conformational change of two HAs to the extended coiled coil would
leave a hydrophobic cavity in the center of the HA aggregate,
stabilized by the "dam" of transmembrane domains, and remaining
fusion peptides (see also Bentz and Mittal, 2000).
How this curved bilayer can initiate the fusion is unknown and quite difficult to probe experimentally. This suggests the value of a molecular simulation to assess the structure of the site. The proposed fusion site with both the proteins and the curved bilayer is too complex to simulate at present. However, we can investigate the simplest element, i.e., what would happen if a persistent hydrophobic defect were formed in a bilayer? How would the hydrophobic effect heal this wound without the help of lipid lateral diffusion according to a molecular dynamics simulation? We used a planar bilayer patch of 8 × 8 equilibrated phosphatidylcholine (PC) molecules in water and deleted 0, 3, 6, 9, or 12 of the PCs from one monolayer while restraining part of the lipids to act like a wall such as might be formed by transmembrane domains and the remaining embedded fusion peptides of the HAs, as described above. This created several persistent hydrophobic defects in the patch, because there are no extra lipids to diffuse into the patch. Our question was simply whether the hydrophobic defects in the patch would percolate together, making a single large hydrophobic cavity, or whether they would increase their acyl chain cross-sectional area and "squat down" to achieve acyl chain contact throughout the defect monolayer. For comparison, we also studied two bilayers in which lipids were removed from one monolayer but no further restraints were applied. Our results suggest that simulations on curved bilayers with persistent hydrophobic defects, when feasible, will predict an even more realistic fusion site.
| |
MATERIALS AND METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Simulations were carried out with Gromacs (Berendsen et al.,
1995; van der Spoel et al., 1999). The united atom lipid parameters are
based on Berger's parameters (Berger et al., 1997) with
additional parameters for the unsaturated carbons taken from GROMOS87.
Water was modeled as simple point charge (Berendsen et al.,
1981). All simulations used a twin-range cutoff for Coulomb of 1.0/1.8
nm and a single cutoff for Lenard-Jones interactions of 1.0 nm.
Temperature and pressure were controlled using the weak coupling
algorithm (Berendsen et al., 1984) with 
T = 0.1 ps,
T = 300 K, p = 1.0 ps, with a
pressure of 1 bar in the z direction. The timestep used was 2 fs with a
neighbor list update every 10 steps. Coordinates were saved every
picosecond. A structure from a previous simulation was used as starting
structure for the 128 lipids palmitoyl-oleoyl-phosphatidylcholine bilayer (available from http://moose.bio.ucalgary.ca) (Tieleman et
al., 1999). Extra water was added, for a total of 4480 water molecules
and 128 lipids (20,096 atoms), resulting in a system of 6.37 nm × 6.59 nm × 7.14 nm. A short equilibration run of 25 ps was used to
equilibrate the added water. The lateral area was kept fixed in all
simulations described, at an area of 0.655 nm2
per lipid, whereas the z dimension of the box was allowed to fluctuate. The area was estimated based on the experimental area of
DPPC of 0.629 ± 0.014 nm2 (Nagle et al.,
1996). However, the area is not critical in the current set of
simulations. An overview of the simulations is given in Table
1. For simulations R0 to R12, 29 lipids
that approximately form the edge of one of the two bilayer monolayers
were restrained with an harmonic position restraint on all atoms of
1000 kJ mol
1 nm1. From
the remaining 64 
29 = 35 lipids that were free to move in
this monolayer, between 0 (R0) and 12 lipids (R12) were removed randomly to create a hydrophobic defect (Fig.
1). The resulting structures were used as
starting structures for simulations R0 to R12. Water atoms rapidly
(within 50 ps) fill the vacuum created by removing lipids as the
pressure coupling in the z direction automatically reduces
the total volume of the system according to the number of lipids that
has been removed. Lipids also readjust somewhat during this fast
initial equilibration, but their motions are much slower than the
diffusion of water. For R12F and R18F, 12 or 18 lipids were removed
from one monolayer, but no lipids were fixed at their positions.
Analyses were done with Gromacs programs (Berendsen et al., 1995; van
der Spoel et al., 1999). Molecular graphics were made with molscript
(Kraulis, 1991) and raster3D (Merritt and Bacon, 1997).
|
|
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Our primary goal is to determine how lipids adjust to a hydrophobic defect in one monolayer. To do this, we monitored the solvent accessible area of groups of lipids, the depth in the bilayer of several atoms along the lipid chains, and the average distance from the phosphorus atom to the last carbon of the palmitoyl chain during the simulations. We also calculated the deuterium order parameters for the chains and the atomic density profiles across the membrane over the last part of each simulation. Combined with molecular graphics pictures of the starting and the final structures of each simulation these properties allow a reasonable description of the evolution of a hydrophobic defect.
Molecular graphics
Fig. 1 shows molecular graphics views of the lipid bilayer with different defects. R0 to R12 all have 29 lipids restrained in one monolayer of the bilayer (green). Of the remaining 35 lipids, 0 to 12 lipids are randomly removed (red). These lipids will be referred to as "free lipids." The monolayer with the defect will be referred to as the "defect monolayer." The other monolayer is colored yellow, which will be referred to as the "intact monolayer." The z axis runs perpendicular to the lipid-water interface across the bilayer. It should be noted that the snapshots give an impression of the different states and different views of the bilayer, but that significant fluctuations occur. Nonetheless, the general results of the simulations are already visible. In R0 and REF (both not shown), the red and green lipids cover the view from above entirely, and no yellow lipids are visible. In R3, the free lipids redistribute and cover as much of the holes as possible. In R6 this is still reasonably effective. In R9 and R12, the free lipids are unable to cover the yellow lipids effectively. In R12F and R18F, the lipids in the defect monolayer spread out and thin significantly. This general picture is confirmed by the more detailed analyses below. These analyses will also describe what happens to the intact bilayer, which is harder to visualize.
Average lipid length and position
It is useful to consider the positions of different atoms within the lipids as well as the overall length of the lipids to determine how the lipids adjust to the presence of a defect. We have chosen to monitor a few key atoms: P8, the phosphorus atom; C13, the middle carbon atom of the glycerol group; C24, the 9th carbon of the oleoyl chain; C42, the 8th carbon of the palmitoyl chain, and C50, the last carbon (CH3) of the palmitoyl chain, all counting from the headgroup to the end of the tails. Fig. 2 shows the average z coordinate of each of these groups of atoms for the free lipids in the defect monolayer and all the lipids in the intact monolayer. The results for R12F and R18F are relatively straightforward: atoms along the whole lipid are hardly affected in the intact monolayer, but in the defect monolayer all move closer toward the center. Not surprisingly, the effect is largest for atoms furthest away from the center. This is consistent with the molecular graphics in Fig. 1. The z position of atoms in R0 to R12 is more interesting. Not much change is observed in R0 (or in REF, data not shown) and little change in R3. However, R6 to R9 P8 and C13 in the defect monolayer move somewhat closer to the center. C24, C42, and C50 show less change except in R12, where all three clearly move toward the center. Interestingly, the average position of atoms in the intact monolayer also moves closer to the center. Lipids from the intact monolayer move in from the other side to fill in some of the gaps. However, the average z positions of C50 in both monolayers, the last atom of the palmitoyl chain, still differs by 0.4 nm, showing that there is no average interdigitation of the lipids. In an equilibrium bilayer, without defect, the terminal methyl groups of both chains are distributed over the same range of z coordinates, and the distributions for both monolayers overlap. Although the terminal methyl distribution in R12 overlaps more than in an unperturbed bilayer, interdigitation of more than these methyl groups remains small. An interesting question is how the structure of the lipids is changing. A simple criterion to monitor is the "length" of a lipid, taken as the average distance between P8 and C50 in a lipid. Fig. 3 summarizes the lipid length as a function of time. The black lines, the free lipids in the defect monolayer, show that going from R0 to R12 the lipid length decreases substantially from ~1.6 nm to ~1.3 nm for R12. The red lines, the intact monolayers, show that the lipid length in the intact monolayers is mostly unaffected. An exception is R12, where it appears that the lipid length of the free lipids increases somewhat. This correlates with the increase in order parameters for this system (see below). Thus, although the lipids in the intact monolayers show an increasing tendency to move toward the center when removing more and more lipids in a defect in the other monolayer, this has little effect on the average length of the lipids. In R12F/R18F, the lipid length in the intact monolayer increases (blue line for R18F), but the lipid length of the lipids in the defect monolayer rapidly decreases (~0.5 ns) from 1.6 to 1.7 nm to 1.4 to 1.3 nm.
|
|
Density profiles
In Fig. 4 A the normalized distribution of the free lipids in the defect monolayer is plotted. REF (averaged over all 64 lipids in one monolayer, because there are no restrained lipids) and R0 are similar with small differences due to the restraints in R0. From R0 to R3 to R6 to R9 to R12, the distribution becomes increasingly sharper and shifts to the middle of the bilayer, consistent with the analyses already presented. The effect is even stronger for R12F and R18F (both averaged over all lipids in the defect monolayer), both of which narrow and shift toward the middle of the membrane. In the same order, water penetration increases from R0 to R18F, as shown in Fig. 4 B. Note that in Fig. 4 A the lipid distribution is normalized on the number of lipids, whereas the values in Fig. 4 B are absolute in g/cm3. The minimum in the total lipid density profile changes little between the different simulations.
|
Order parameters
A more detailed method to study the change in conformation of lipids is to consider "deuterium" order parameters, calculated from the simulations. In Fig. 5, these are shown for R0 to R12 and R12F/R18F, split in four groups: free lipids in the defect region for the first part of each simulation, free lipids in the defect region for the last part of each simulation, and the same for the lipids in the intact monolayer. This allows an approximate view of the time evolution of the order parameters or the adjustment of the lipids after the sudden creation of a hydrophobic defect. Note that the order parameters are shown for carbons 2 to 15 of the palmitoyl chain of the lipid. Again, let us consider R12F/R18F first. In both of these cases, there is a fast and very large drop in order parameters for the lipids in the defect monolayer, consistent with the observed thinning. The drop is fast because there is little difference between the order parameters averaged over 0 to 2 ns and 2 to 4 ns in both cases.
|
The intact monolayer is affected somewhat with a small increase in order along most of the chain. The last seven carbons in R18F have essentially random order. In R0, there is little difference between the four graphs, as expected. In R3, however, it is clear that a significant drop in order parameters develops in the course of the simulation for the free lipids in the defect area over the last 10 carbons of the tail. This suggests that the bilayer structure near the upper ends of the chains does not change much but that the "floppy" ends of the chains readjust to attempt to fill the extra free volume.
The effects on R6, R9, and R12 become progressively stronger with the
presence of larger defects. The order parameters for the free lipids in
the defect area continue to drop, along the whole chain including the
atoms close to the lipid glycerol moiety. However, in R6 and R9 there
is still little effect on the chain order of the lipids in the intact
monolayer. In R12, there is some increase along the whole chain for the
lipids in the intact monolayer, consistent with the increase in lipid
length discussed above. Comparison of order parameter profiles between
R12F and R6 and R18F and R9 shows that the presence of restrained
lipids significantly limits the effect of removing lipids. R12F has a relative surface concentration of (64 12)/64 = 0.81, comparable with R6 with (35 6)/35 = 0.83. R18F has a
surface concentration of (64 18)/64 = 0.72, comparable
with R9 with (35 9)/35 = 0.74. Thus, the changes in lipid
structure cannot be explained by just a change in surface concentration.
Solvent accessible area
The driving force for the behavior of hydrophobic defects will be
to a large extent the solvent-accessible hydrophobic area. Fig.
6 follows the hydrophilic (Fig. 6
A) and hydrophobic (Fig. 6 B) solvent accessible
area per lipid for the free lipids in the defect area. These graphs
also give an indication of the timescale of the readjustment of the
lipids and the degree of equilibration in our simulations. Like the
order parameters, the solvent accessible areas for R12F and R18F show a
rapid decrease after the creation of the initial defect within a few
hundred picoseconds. The relaxation times for the other systems
increase with the size of the defect and are of the order of 5 ns. This
is quite fast relative to the normal time scales of lipid motion in
bilayers (Lindahl and Edholm, 2001), indicative of the large driving
forces present in the system.
|
The areas appear to converge to a predictable order with the hydrophobic solvent accessible area increasing from R0 to R3 to R6 to R9 to R12. R12F and R18F have a remarkably low solvent accessible area, demonstrating that the presence of restraints on the lipid motions significantly changes the behavior of the remaining free lipids. In Fig. 7, the average solvent accessible areas are shown as a function of defect size, averaged over the last part of each simulation. For both the hydrophilic (Fig. 7 A) and the hydrophobic (Fig. 7 B) solvent accessible area there is a steady increase for all three groups of lipids: Grp 1 consists of the restrained lipids, Grp 2 consists of the free lipids in the defect area, and Grp 3 consists of all lipids of the intact monolayer. The increase is most dramatic for the free lipids in the defect area and occurs for both hydrophobic and hydrophilic atoms. Fig. 7, C and D show that the solvent accessible area, both hydrophobic and hydrophilic, increases significantly for the defect monolayer of R12F and R18F compared with REF, without defect, but the intact monolayer is almost unaffected. This again demonstrates the effectiveness of the free lipids to cover all exposed chains from the intact monolayer by thinning substantially.
|
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Hydrophobic defects in lipid bilayers
The simulations presented in this paper are used to investigate qualitatively how hydrophobic defects that arise in nonequilibrium processes might evolve. We considered two cases. In the most interesting case, the flow of lipids to a defect is restricted by mechanical barriers, here represented by position restraints on a subset of the lipids. For comparison, we also considered defects in a lipid monolayer where lipid flow is unrestricted. When there are no mechanical barriers, the adjustment of lipids to the presence of a defect appears straightforward. The lipids spread over the available surface, decreasing the monolayer thickness but with only a relatively small effect on the final hydrophobic solvent accessible area. The intact monolayer is hardly affected by this process.
When a mechanical barrier is present, the behavior of the remaining lipids in the defect area is more complex. The main effect of randomly removing lipids is still a thinning effect that becomes stronger with the size of the defect. In R3, with three lipids removed, the lower one-half of the chains becomes more disordered, but the headgroup region and the upper part of the chains remains more or less the same. There is little effect on lipid length and on the opposite monolayer. However, in R6, R9, and R12, an additional change occurs. The lipids in the defect area do not thin as much, and there is a coupling with the intact monolayer. The larger the defect, the more lipids from the intact monolayer move toward the defect. In the case of R12, this leads to "stretching" of the lipids from the intact monolayer with increased order parameters and an increased length.
In retrospect, and in an overly simplified way, this is what would be expected if the hydrocarbon cores were treated as a continuous fluid with an initial water filled cavity in the defect monolayer. The area of contact with water for the unconstrained system would be less with thinning than with retaining the cavity, since the surface area of polar headgroups would be the same in both cases. When the defect perimeter lipids were fixed in place, thinning of the defect monolayer is constrained by the exposure of the acyl chains of the fixed lipids. The acyl chains of the intact monolayer were drawn into the defect, because the defect monolayer could not thin as much as in the unconstrained case. If we push the continuous fluid analogy still further and assume we fix the hydrophobic perimeter through both monolayers, as a protein transmembrane domain would be, we would expect that the additional restraint on thinning of the intact monolayer would constrain further the thinning of the defect monolayer. Perhaps, some degree of cavitation in the defect monolayer might occur. This remains to be investigated.
Simulation assumptions and limitations
It is useful to consider some choices made in the design of our simulations and some limitations. There are several possible approaches to considering defects within a restricted area. We chose to restrain a "ring" of lipids in the "outer" monolayer to provide a barrier to the defect zone, on the assumption this roughly resembles the situation where fusion proteins inhibit lipid flow. Alternative approaches could have included perfectly cylindrical walls modeled as some potential or actual proteins or protein models that form a restricted zone. Model potentials would have provided unknown interactions between the walls and the free lipids and would have required considerable investigations into the effect of different types of walls. Protein models would have considerably increased the complexity and the number of atoms of the systems studied. With the relatively long relaxation times on a 5- to 10-ns timescale, this would have made the computation too expensive at present.
One further design simplification that deserves some attention is our choice to restrain only lipids in one monolayer of the membrane, whereas fusion protein transmembrane domains span both monolayers. Lipids are so disordered that it would be problematic to identify a "ring" of lipids in the intact monolayer that have the same lateral location as in the defect monolayer to mimic a cylindrical restraint. In the end, we believe that the trends identified in our simulations are qualitatively accurate, as will be elaborated just below.
The simulations have several limitations. In particular, the length of
the simulations with larger defects (9-12 lipids removed) appears only
a few nanoseconds longer than the necessary minimal length to see the
major adjustments of the system. It is possible that considerably
longer relaxation times are present in some of the systems, although
there are no clear indications for this. A second limitation is the use
of electrostatic cutoffs, which are likely to affect the structure of
the bilayers to some extent. However, due to the qualitative nature of
these simulations we do not believe this is a major concern. For
example, in a recent paper by Marrink et al. (2001), PC lipids
reorganized from random orientations in water into a bilayer even when
long range electrostatic interactions were not taken into account
beyond 1.7 nm.
Implications for fusion
Whereas our direct goal here was to investigate the evolution of
persistent hydrophobic defects in a monolayer, we want to return
briefly to the molecular situation during membrane fusion, one case
where we suggest they could arise. If the hydrophobic defect is a
necessary step in influenza hemagglutinin mediated fusion, as proposed
(Bentz, 2000b; Bentz and Mittal, 2000; Mittal and Bentz, 2001), then
the very simplest version had the persistent hydrophobic defects
coalesce together, to provide a clear target for lipids permeating from
the defect monolayer of the target membrane. This, however, did not
occur in the simulation, but what did occur suggests a more realistic
scenario, which is consistent with the most recent and rigorous
findings about HA-mediated fusion.
Two recent studies can shed much light on our results by focusing our
attention on what might be expected at a successful fusion site. First,
Mittal et al. (2001, 2002a) used automated video fluorescence
microscopy to analyze lipid and contents mixing kinetics between
individual red blood cell/erythrocyte-HA expressing cell pairs. For
those pairs showing lipid mixing, one of two phenotypes was observed.
There were those that completed fusion, where aqueous contents and both
lipid monolayers mix, and those that showed only hemifusion, presumably
a mixing of only the outer monolayers. Remarkably, starting from the
onset of lipid mixing, fusion (with contents mixing) was completed more
quickly than hemifusion, suggesting that these two phenotypes arise
from (at least) two distinct defects.
In a second study, the protein basis for these two distinct defects was
suggested. Leikina et al. (2001) found that a soluble proteolytic
fragment formed from the low pH form of HA (FHA2, Kim et al., 1998)
could induce only hemifusion between cells at low pH, i.e., mixing of
outer monlayer lipids only. This fragment is essentially the low pH
equilibrium structure for HA2, i.e., with preformed extended coiled
coil and helix-turn. Generally, the soluble amphipathic peptides seem
only able to destabilize the defect monolayers enough to provoke
hemifusion (for review, see Bentz and Mittal, 2000), i.e., the inner
monolayers are not coupled to the defect formed. Thus, for successful
fusion, the fusion proteins must couple the viral inner monolayer into
the defect.
In one sense it is obvious that fusion, by definition the complete
mixing of lipids and aqueous contents, requires the participation of
the intact monolayers (Bentz and Ellens, 1988). The question is: What
is the molecular meaning of "coupling" the intact monolayer to the
fusion defect? It is known that HA-mediated fusion requires the
essential conformational changes of HA (Qiao et al., 1998) and a full
transmembrane domain (Armstrong et al., 2000; Bentz, 2000b), i.e., HAs
with truncated transmembrane domains provoke only hemifusion. So how do
these factors engage the inner monolayer into the initial fusion defect?
Our simulation showed a coupling of the intact monolayer with the
defect when the defect perimeter lipids were fixed and six or more
lipids were removed. This break point is interesting because two
bilayer-embedded HA fusion peptides would occupy about the same surface
area as six or so lipids so that their removal would create about the
same sized hydrophobic defect. So in the proposed fusogenic HA
aggregate (Bentz, 2000b; Bentz and Mittal, 2000), the removal of two
embedded fusion peptides from the center of the fusogenic aggregate
would create a large enough hydrophobic defect to cause the intact
monolayer acyl chains to enter into the defect area of the defect
monolayer because of the hydrophobic transmembrane domains. Provided
that the "dam" of HAs remains intact, this scenario is reasonable
and provides a first guess as to the molecular definition of
"coupling," which can distinguish between the defects that lead to
fusion from those which lead only to hemifusion. That is, amphipathic
peptides (Leikina et al., 2001; Bentz and Mittal, 2000), low surface
densities of HA (Chernomordik et al., 1998), and some mutant HAs
(Qiao et al., 1998, 1999; Armstrong et al., 2000; Melikyan et al.,
1999) only disturb the outer monolayers leading to hemifusion. Here we
have an example of how a hydrophobic defect in the lipid defect
monolayer, if surrounded by a hydrophobic mechanical barrier, will
engage the inner/intact monolayer. This is summarized in Fig.
8. Clearly, it is possible that there are
other mechanisms to engage the intact monolayer. For example, rapid
removal of the fusion peptide might have not only local effects on the
defect leaflet but on the intact leaflet as well. This could be
investigated by further simulations but is outside the scope of this
study.
|
We are a long way from fusion and from the formation of a hydrophobic
cavity. However, our simulation neglected two elements that seem likely
to favor percolation of the persistent hydrophobic defects into a
single large cavity. First, this simulation assumed a planar bilayer,
because the extension to curved membranes is not yet theoretically
feasible. J. Bentz and A. Mittal (manuscript in preparation) have shown
that the model of HA aggregation forming curved domes proposed by
Kozlov and Chernomordik (1998) is consistent with the kinetic analysis
of fusion data for HA-expressing cells and virions (Bentz, 2000a,
Mittal and Bentz, 2001; Mittal et al., 2002b). Thus, the beginning
point for the fusogenic aggregate should be a lipid dome tightly ringed
by the aggregated HAs. This produces an elastic energy at the peak of
the dome due to its curvature. However, the dome rises less than 3 nm
above the base of HA, i.e., the target membrane is at least 6 to 8 nm
away. One way to minimize the elastic energy of the curved dome would
be to percolate the hydrophobic defects to its peak, thereby creating a
single cavity. Second, there was no hydrophobic barrier around the
inner monolayer required to mimic the other half of the transmembrane domain. With membrane curvature, the defect monolayer would be thinner
than the intact monolayer in the first place; hence the defect
monolayer would not be able to thin as much as in the planar case. On
the other hand, the thicker inner monolayer would be restrained by the
hydrophobic perimeter from further commitment to the defect in the
outer monolayer. This too would promote the percolation of persistent
hydrophobic defects to the peak of the dome because thinning would be
relatively inhibited.
The molecular mechanism of HA-mediated fusion is an important and unsolved problem. Many tools have been brought to bear fruitfully on the problem. Simulating the whole fusion site is not feasible at this time. Here, we have mimicked some of the essential elements of one proposed molecular mechanism and found that a rigorous molecular simulation of that simplification has provided a first guess as to how the fusion site needs to evolve to end up at fusion rather than hemifusion. Further simulations can test other proposed elements of the fusion site, which should lead to the construction of definitive experiments.
| |
ACKNOWLEDGMENTS |
|---|
D.P.T. is a Scholar of the Alberta Heritage Foundation for Medical Research.
| |
FOOTNOTES |
|---|
Address reprint requests to D. Peter Tieleman, Department of Biological Sciences, University of Calgary, Calgary, Alberta T2N 1N4, Canada. Tel.: 403-220-2966; Fax: 403-289-9311; E-mail: tieleman{at}ucalgary.ca or to Joe Bentz, Department of Bioscience and Biotechnology, Drexel University, Philadelphia, PA 19104. Tel.: 215-895-1513; Fax: 215-895-1273; E-mail: bentzj{at}drexel.edu.
Submitted January 29, 2002, and accepted for publication April 15, 2002.
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Biophys J, September 2002, p. 1501-1510, Vol. 83, No. 3
© 2002 by the Biophysical Society 0006-3495/02/09/1501/10 $2.00
This article has been cited by other articles:
 |
|
 |
|
  K. Koshiyama, T. Kodama, T. Yano, and S. Fujikawa Structural Change in Lipid Bilayers and Water Penetration Induced by Shock Waves: Molecular Dynamics Simulations Biophys. J., September 15, 2006; 91(6): 2198 - 2205. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. J. Estes, S. R. Lopez, A. O. Fuller, and M. Mayer Triggering and Visualizing the Aggregation and Fusion of Lipid Membranes in Microfluidic Chambers Biophys. J., July 1, 2006; 91(1): 233 - 243. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Catte, J. C. Patterson, M. K. Jones, W. G. Jerome, D. Bashtovyy, Z. Su, F. Gu, J. Chen, M. P. Aliste, S. C. Harvey, et al. Novel Changes in Discoidal High Density Lipoprotein Morphology: A Molecular Dynamics Study Biophys. J., June 15, 2006; 90(12): 4345 - 4360. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Y. Jiang, Y. Bouret, and J. T. Kindt Molecular Dynamics Simulations of the Lipid Bilayer Edge Biophys. J., July 1, 2004; 87(1): 182 - 192. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Mittal, E. Leikina, L. V. Chernomordik, and J. Bentz Kinetically Differentiating Influenza Hemagglutinin Fusion and Hemifusion Machines Biophys. J., September 1, 2003; 85(3): 1713 - 1724. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
