Tip:
Highlight text to annotate it
X
Hello, I'm Stephen Harrison from the Harvard Medical
School, Children's Hospital Boston, and the Howard
Hughes Medical Institute. Welcome to Part 2 of this series
on virus structures. This part is about viral membrane
fusion, the process by which enveloped viruses get into
cells. As those of you who watched Part 1 will know,
enveloped viruses, those with lipid bilayer membranes,
acquire their membrane by budding out through the
surface or into an internal compartment of the host cell.
And likewise, they penetrate cells that they are about to
infect by fusion, a reverse of the budding process, by
fusion of viral and cellular membranes. Different viruses
have different triggers or sensors, if you wish, to initiate
the fusion process. Influenza virus, which enters through
endosomes, depends on the low pH of the endosome to
initiate fusion. Viruses such as *** can fuse at the cell
surface, and they depend on the sensing the receptor,
which triggers conformational changes of its own and in,
in the case of ***, a co-receptor as well. What is
membrane fusion? Membrane fusion is, in the simplest
sense, making one bilayer out of two. But it's a relatively
complicated process in practice, although it's
thermodynamically downhill, that is, the fused structure is
ultimately stabler than the two separate structures, but
there's a substantial kinetic barrier, and it's overcoming
that kinetic barrier that is the role of the viral fusion
proteins, or of cellular fusion proteins. So an intermediate
in the fusion process is generally accepted to be a
structure in which the apposed monolayers, the apposed
leaflets, of the two bilayers have merged, but not yet the
distal ones, and that's called a "hemifusion" structure, or
a "hemifusion stalk." And while there's some debates
about the detailed organization of the hemifusion
intermediate, it's clear from a number of studies that that is
an important step on route to fusion. Indeed, the barrier
between two bilayers and the hemifusion structure is one
of the major kinetic barriers in this process of fusion, and
there is probably a kinetic barrier between hemifusion and
the ultimate merging of the distal leaflets that lead to the
formation of a fusion pore. In the case of viral proteins,
there's a sequence of events that's reasonably
stereotypical, it turns out, even though the molecular
machinery driving this series of events may look very
different. That is, the fusion proteins of different viruses,
although from the point of view of their protein
architecture may be very different, the underlying process
that they catalyze (and there's a real sense in which this a
catalysis, since as I said it's thermodynamically downhill,
but with a high kinetic barrier)... the sequence of events
that they catalyze is reasonably stereotypical in all cases.
And so, before events begin, the fusion protein is in some
conformation, and this is a purely schematic
representation, and there are two bilayers: the bilayer of the membrane
in the virus, and the bilayer of the membrane of the cell to
which the virus is attached. Some event, proton-binding
or receptor-binding, induces or fixes a conformational
change in the fusion protein that leads to the formation of
an extended intermediate in which a hydrophobic
element, either an N-terminal peptide or a loop in the
middle of an extended part of the protein structure,
interacts with the target cell membrane. And that
extended intermediate, which is transient, then collapses
into a structure that is ultimately a stable structure for the
fusion protein, and drags the two membranes together. As
I suggested, there are probably kinetic barriers from the
point of view of the lipid bilayer itself, both between the
two bilayers state and the hemifusion state, and between
the hemifusion state and the final formation of a fusion
pore, and it is the role of fusion protein to lower that
kinetic barrier, as suggested by these dashed, red lines.
We'll talk almost entirely about the fusion protein of
influenza virus, the so-called hemagglutinin. It's a member
of a class of viral fusion proteins, all of which have the
following properties, and it's sometimes because they
were the earliest ones characterized in molecular
structural terms, have come to be called "Class I" viral
fusion proteins. These proteins are synthesized as a
precursor, which is cleaved, usually, en route to the cell
surface, by a protease in the late compartments of the
secretory pathway (furin, for example) into an N-terminal
element, which is usually a receptor-binding domain (some
viruses have proteins like this, but have a separate
receptor-binding protein) and a fusion modular, the C-
terminal half in general, which is anchored by a C-terminal
transmembrane segment in the viral membrane. Examples
of this sort of protein (they are all trimeric assemblies of
this sort of organization) are influenza, ***, and the
filoviruses such as Ebola. In the case of influenza, where
the protein hemagglutinin sticks off of the surface of the
virus... along with another protein, which is an enzyme,
called neuraminidase, and we won't talk about that today.
The hemagglutinin is a trimeric structure, as I suggested,
with three functions. It binds the virus to its receptor, the
receptor is sialic acid on glycolipids or glycoproteins on
the surface of the target cell. It has structures on the
outside that can vary without compromising its two other
essential functions, so that the virus can evolve to
escape neutralization by the immune system of its hosts.
And finally, it is, as I've suggested, the protein that
catalyzes the membrane fusion process when suitably
triggered by proton-binding. So as I've said, it's
synthesized as a precursor. This diagram is overly
complicated, but all that matters for today is that, at the N
terminus of the so-called HA2... the precursor is called
HA0, and the two fragments are known as HA1
(hemagglutinin 1) and HA2. At the N terminus of HA2 is a
hydrophobic peptide exposed, if you wish (it's actually not
exposed in the structure, but made N-terminal rather
internal by the cleavage process), that interacts with the
target cell membrane and is known as the fusion peptide.
And then there is a transmembrane segment very near the
C terminus that anchors the protein in the viral membrane.
So, the representation here shows you the overall
structure of the hemagglutinin. This particular
representation is based on x-ray crystallography and does
not show the transmembrane segment or the very short
segment of about 11 residues that extends into the interior
of the virus particle or, before budding, into the cytosol of
the cell. As you see, most of the HA1 part, which would
be, let us say, red (and the HA2 part would be green of
one of the subunits), most of the HA1 part folds into a
globular domain at the top of the molecule. It contains the
site for binding sialic acid. HA2 forms a stalk that projects
it outward from the surface of the virus. The sialic acid-
binding site here (there's one on each of the three
subunits) faces outward; it's the one very conserved
feature of an otherwise antigenically variable surface that
the molecule presents to the outside world. Here's a
slightly more readable representation, both of the
monomer on the left, and of the trimeric, spike-like
hemagglutinin on the right. Let's look at the monomer. As I
said, the HA1 part is largely out at the surface with its
sialic acid-binding site, the HA2 part forms the stalk of the
molecule. The N terminus of HA2, remember that's the
fusion peptide, is here, tucked in along the threefold axis
of the trimer. And so the fusion peptide is hidden and
can't interact with hydrophobic targets in the structure of
the protein as we see it here, but as you'll see, once
exposed to low pH, once protons bind, a major
underfolding occurs that allows this fusion peptide to
emerge and interact with a target membrane. So here's
the low pH-triggered conformational change, and one
way of describing it from the point of view of the
monomer, is that the HA2 part turns itself inside out. That
is, the part of the HA2 (and perhaps it's easier to see in
this representation with colored segments)... the part of
HA2 that's on the outside in the trimer, which is red and
then merging into blue, is on the inside after the
conformational change, and the part that's on the inside
(green and yellow) turns around and comes up the
outside. This structure is most simply described as a trimer
of hairpin conformations. There's a fair amount of twisting
and turning at the turnaround of the hairpin, but
fundamentally, you can think of this as three polypeptide
chains that begin up here (with the purple arrow which
represents the fusion peptide, it's not represented here
since it's based on a crystal structure), comes down, and
then turns around and comes right back up to the
transmembrane segment, which would follow the yellow
arrow. So the hemagglutinin then undergoes two
irreversible changes in the course of its maturation and
exposure to low pH, because indeed the conformational
change I just showed you is irreversible. If you then n
eutralize, you don't go backwards. And that's because of
the first irreversible change, which is the cleavage of a
peptide bond. That now means that the structure we see,
which is very stable if you keep it at pH 7 (soluble flu
hemagglutinin can hang around for months or years stably
in the laboratory), but if you expose it to low pH very, very
rapidly, it rearranges as shown and that rearrangement
doesn't go backwards, and it doesn't go backwards
because there's no way of reknitting that peptide bond,
since this structure is actually not the lowest free energy
state, it's just there's a very high barrier here that's
lowered when protons bind. And it is that second change
and the free energy recovered from that second
conformational change that is coupled to the process of
membrane fusion. And so the fusion mechanism can be
thought of as cleaving the precursor, or priming this fusion
machinery; localizing the virus to the cell by receptor
binding, ultimately by uptake into the endosome; and the
triggering of refolding, in the case of flu, by low pH, in the
case of other viruses, let us say, by a receptor or co-
receptor binding, that leads to this stereotypical sequence
of events: Exposure of the fusion peptide (that's that
extended intermediate), insertion of the fusion peptide into
the target membrane, and a folding back of the protein
that brings together the target and viral membranes. And it
is that folding back that overcomes the first of the kinetic
barriers. There is a substantial kinetic barrier to squeezing
two membranes any closer together than about 10 or 15
angstroms. That is why liposomes, let's say, in solution are
stable, although once fused, they are even more stable.
But a liposome preparation doesn't spontaneously fuse
because of that kinetic barrier to bringing two bilayers
close together. And it is that process that is at least one of
the crucial ways in which these proteins facilitate
membrane fusion, and they do so by recovering free
energy in this fold-back process because the primed state
is, in one way or another, metastable. So the fusion of
membranes by influenza virus can be through of, then, as
a triggering process (we don't show actually the sialic acid
attachment here)... but a triggering process that leads to
dissociation of the HA1 domains at the top. There
happens to be a disulfide bond down here that keeps
HA1 from actually floating away, but some experiments
done already 10 or 15 years ago (actually, more than that,
nearly 20 years ago, now that I think of it), showed that if
you knit the tops together, then this process can't occur,
so we know that this dissociation of the tops from the
stalk occurs, and that allows the stalk, the HA2 stalk, to
unfold and refold, so to speak. That is, allows the fusion
peptide to flip up, associate with the target bilayer, and
then, along with the rest of the protein, collapse together
to squeeze the two bilayers together, leading to
membrane fusion. I said that the description of the post-
fusion conformation of the flu hemagglutinin of HA2
corresponds to a trimer of hairpin-like structures, and it
turns out that for large numbers of these so-called Class I
viral fusion proteins, that simple analogy is true. Indeed, in
the case of *** and SIV, the hairpin is particularly simple.
It's just a helix coming down, a loop turning around, and a
helix coming up, and so the membrane fusion process is
nicely represented in this animation from Gaël McGill
based on the structure of the post-fusion state of the ***
and SIV conformational proteins, which you can see
going from the extended intermediate at the beginning, to
a fused state at the end. So we can then ask, in the case
of flu hemagglutinin, which makes a post-fusion structure
that's also a trimer of hairpins (although as it happens, as
you saw, the outer layer isn't a simple α-helix, the
structure is a bit more complicated, but it's still
fundamentally just coming one way and going back the
other way), how many trimers are needed to make such a
fusion structure, and indeed, how long does the process
take? And so in some experiments that our laboratory
undertook with the collaboration of Antoine van Oijen,
through the work of a graduate student named Dan
Floyd, we sought to use contemporary techniques in
single-molecule fluorescence microscopy to try to carry
out measurements of fusion, looking at individual virus
particles. Because it was clear that the only way we could
begin to answer the questions I was just raising about
timescale and about numbers of hemagglutinins needed
could only be answered in that way. And the experimental
setup that Dan Floyd devised is shown schematically
here. A lipid bilayer supported on a thin layer of a dextran
polymer is doped with a bit of ganglioside, lipids that have
sialic acid on their head group and therefore are receptors
for flu hemagglutinin. An influenza virus that has been
exposed to two different fluorescent dyes, that has taken
up two different fluorescent dyes, is allowed to bind to this
surface. The green dye is a hydrophobic dye that inserts
into the membrane. The red dye is a more soluble dye that
can be soaked into the virus particle, and then the
excess washed out, and the virus used in the experiment
before any of it leaks back out. And so those two dyes
report, on the one hand, mixture of lipids in the two
membranes, and hence the hemifusion step, and
formation of an aqueous channel between the virus and
the solute in the swollen dextran polymer layer, that shows
the formation of a full fusion pore. And finally, there's a
fluorescein pH sensor to tell us... in the bilayer, fluorescein
is bleached when the pH drops below about pH 6, and
that tells us when, in the experiment you're about to see,
the pH in the region of the virus particle fell below a
critical value. And so, here's the kind of measurement that
is done, and you'll see here the recording both of the
signal from the pH sensor; the signal from the green dye
that's in the bilayer, the hydrophobic dye; and the signal
from the red fluorophore that is inside the virus particle.
And what happens when the pH drops is that, with a
certain time delay, there is suddenly a rise and then a
rapid fall of the fluorescence from the hydrophobic
fluorophore. That's because there's enough of it in the
membrane that the signal is quenched. This represents
de-quenching as the two membranes begin to merge, as
the hemifusion event occurs, and then the fluorophore
diffuses away in the target membrane. Then with a further
time delay, there is mixing of the content of the virus with
the aqueous substrate in the dextran layer underneath the
bilayer, and one sees loss of fluorescence from the dye
that was inside the virus particle as it diffuses away. And
so if one does lots of these measurements, and they can
be done in parallel because in a suitable microscope, as
you see here, there are lots of particles in a field, then one
can get a histogram of the times to hemifusion, that is,
lipid mixing, and the times to pore formation. We can do
this as a function of a variety of parameters, including the
final pH of the buffer that was flowed into the little
chamber in the microscope, and other parameters of the
experiment. Analyzing this kind of experiment, in which...
and I guess I should go back to explain that, as you see,
hemifusion always involves a rise and then a fall, and
when you have a kinetic event that has a delay of that
kind, and so then if we're looking at the time to
hemifusion, there is a certain delay that has a distribution
from particle to particle that looks like this, then you know
that there are multiple kinetic steps, whereas if there's a
simple, single kinetic step, you would just see a single
exponential decay, as you indeed do if you, on a particle-
by-particle basis, plot the time between hemifusion and
fusion. So to fit the kinetics of the hemifusion event, we
chose a relatively simply kinetic scheme with two
parameters, in which there might be "N" sequential steps,
rate-limiting steps, each with a similar rate constant "k," or
"N" independent parallel steps. And they turn out under
suitable conditions to have essentially the same sort of
functional behavior. And as a result, if we fit these
histograms that I showed you in the previous slide, we
find that the best fit involves an N of 3 and a rate
constant appropriate for the times involved: These
experiments were done at room temperature of about 20
seconds or so, as a kind of mean time to hemifusion
under the conditions of this experiment. Whereas the
pore-forming event was a single kinetic step from the
hemifusion state to pore formation. Now what's the
interpretation of this sort of kinetic analysis? Well, as I
said, there were several possibilities. One might be that
there are N sequential steps, three. Another might be that
there are three parallel steps. By looking at the pH
dependence, as indicated here, we found that N was
essentially independent of the final pH. And it seemed to
us unlikely that one could have N distinct sequential
steps that would vary identically with pH, whereas the
same is much more likely to be true of N parallel steps.
And so we've interpreted N as representing the number of
hemagglutinin molecules, the number of hemagglutinin
trimers, needed to form a successful fusion pore. That
number, of course, might be two in some cases and four
in others, but on average over the large number of events
analyzed, the number comes out to just about three. In
other words, that the free energy recovered from three
hemagglutinin conformational changes appears to be
sufficient to drive the process that I was showing you in
the previous slide. So, which of the various steps in this
sort of scheme are we looking at is the rate-limiting step in
this sort of analysis? From the pH dependence, and I
won't go into the details, we believe that there's an initial
rapid equilibrium between a protonated and unprotonated
state, but that as soon as this extended intermediate
forms, then the process is essentially irreversible. And
indeed, as a result of looking at some variance of the
hemagglutinin, we're pretty convinced that it is this step
that in the measurements I just showed you we're looking
at. So there happens to be a very conserved interaction
just where the fusion peptide tucks into the trimeric stalk.
And mutations here, either in a completely conserved
aspartic acid or a completely conserved glycine residue
that stabilize the tucking in, mutations here accelerate
fusion. And so, we take that as evidence that it is this
step that we're looking at. Now, in practice, on the
surface of the virus, there are very tightly packed
hemagglutinin molecules. There are two of them
superposed on this electron micrograph. And so it is also
plausible that three of these humagglutinins clustered in
one region might well be the minimum needed to catalyze
this fusion process. Also, because of the tight packing, it's
very unlikely that in the surface of the virus the proteins
can move around very much, and so again, the process
is presumably carried out by a local set of interactions at
an attachment point between the viral membrane and the
cell surface. So, these sorts of measurements obviously
are just the beginning at trying to understand the details of
this sort of process, but here, from now just about 50
years ago is an electron micrograph of influenza virus in
what would now be called an endosome, to indicate to
you, to give you a bit of perspective, and to suggest to
you that, from this sort of information, at a stage when one
didn't even know what the molecules on the surface of
the virus might be, we're now at a stage... we're at the
level of dissecting the kinetics of the events and, hence,
trying to understand sensitive points for neutralization by
antibodies, for example. We can actually get at the
molecular details of the process that would lead to the
release of the nucleoproteins from inside the particle (you
can actually see some of them in cross section here
probably) into the cytoplasm through fusion of the lipid
bilayer of the virus with the lipid bilayer of the endosome.
I've mentioned Dan Floyd and Antoine van Oijen, I should
mention also Tijana Ivanovic and John Skehel as
collaborators in the measurements I've been showing you,
illustrating how one can use structure and biophysical
measurements to dissect the fusion mechanism. And I
should add further credit to Gaël McGill, whose animation
of the *** fusion mechanism is particularly helpful in trying
to understand what we believe these fusion proteins are
doing. Thank you very much.