Event: NIH Wednesday Afternoon Lecture
Subject: The Structural Basis of Ebola Viral Pathogenesis
Speaker: Erica Ollmann Saphire, Ph.D., Professor, Department of Immunology and Microbial Science, The Scripps Research Institute
Date: Wednesday, November 06, 2013, 3:00:00 PM
Dr. Collins: Good afternoon, everyone. This is a special day because we are in the first day of the NIH research festival and a special day because we have a remarkable lecturer as part of our regular Wednesday afternoon series who is here to teach us something pretty interesting about viral hemorrhagic fever, specifically Ebola virus.
Erica Ollmann Saphire has an interesting and very productive career bringing her to where she as a professor in immunology and microbial science at The Scripps Research Institute. We did a little digging and found a profile of her in the San Diego Union Tribune where she was called, the virus hunter and various comments were made about her contributions, which are obviously substantial. I won’t comment upon what they called her namely a steel magnolia. I thought that was odd to be putting in a profile of a scientist but you can decide for yourself later on.
She got her undergraduate degree at Rice with a double major in Biochem and Cell Biology and Ecology and Evolutionary Biology and then Ph.D. at The Scripps in the year 2000 and has been there in this remarkable productive enterprise focused on trying to understand how pathogens evade and usurp the innate and adaptive immune responses. She has quite a diversity of projects going on in the lab including Lassa and Marburg and Ebola fever and she is an expert in incorporating different approaches to understanding this, including immunology virology and X-ray crystallography.
So it is a great pleasure to have her here and I want to just point out that at the end of the lecture, we will have time for some questions and there are microphones in the aisle and again welcome to those of you who are watching on the web. We’ll try to be sure that questions are posed from the microphone so you can hear them and then at 4:00, we’re going to adjourn down the hall for continuation of informal conversations with our speaker but also the actual formal unveiling of the new FAES center, which I think you’ll want to come and have a look at because it is really quite beautiful facility and we’ll have a ribbon cutting, we will have a few hopefully short speeches and that will morph into a poster session where the scientific directors are actually themselves standing by their posters talking about their science giving you a chance to hit them up with really hard questions. So it’s going to be quite an afternoon.
But, to get us going here, let me ask you please to give a warm welcome to Erica Ollmann Saphire.
Erica Ollmann Saphire – Professor, Department of Immunology and Microbial Science, The Scripps Research Institute
Thank you, Dr. Collins. It’s a real pleasure to be here. My laboratory works on a lot of different viruses. Today I’m going to show you examples from two of them. The first one is Ebola virus, it’s a long filamentous Filovirus and the second one is Lassa virus – this is smaller rounder particle and it belongs to the arenavirus family.
Now what these viruses in common is a similar disease. They both cause hemorrhagic fever and the symptoms look quite similar especially at first. Whereas Ebola is thankfully quite rare, Lassa is unfortunately extremely common. There are hundreds of thousands of cases of Lassa fever every year in Western Africa and it’s the hemorrhagic most frequently transported to the United States and Europe.
Now what else these viruses have in common is a very small genome. Ebola encodes just seven genes; Lassa only 4. So whereas you have 20,000 – 25,000 genes and you can make 20,000, 25,000 proteins. These viruses make only a few.
So, using this very limited protein toolkit, how does this virus achieve all the different functions of the virus life cycle from attachment to a new host cell, fusion, entry, replication and coding, transcriptions, assembly and exit and some of the more sophisticated functions of immune evasion through lots of different pathways – how do they do that with only a very few proteins at their disposal?
This is the genome of Lassa virus. Those are at 7 genes; this is the genome — that was Ebola and this is Lassa, it’s 4 genes in RNA segments.
So how does a handful of proteins conspire to create such extraordinary pathogenesis in hemorrhagic fever? The answer is that each protein at these viruses do encode is essential. These viruses have no junk. Many of these proteins are multi-functional and some are extremely adaptable. By studying the proteins that these viruses make, we see the vulnerabilities of the virus, the Achilles’ Heel, the place where we can target a drug or vaccine or antibody.
But perhaps more importantly, we can understand something more about proteins themselves. Because evolution has compelled these proteins to be remarkable, to do more with less than other proteins. By studying what these proteins are capable of, we learn something more about the capabilities of proteins and molecular biology in general.
So I’ll show you a few examples. The first one comes from the first step of the virus life cycle. So the first step – the virus has to find and attach to a new host cell. This is achieved by the glycoprotein called GP. Both of these viruses express only one protein on their surface, the glycoprotein called GP and it is solely responsible for attachment and infusion with that cell.
So Ebola virus is filamentous. This is a cartoon of what the virus might look like, it’s got a membrane envelope, that’s green surrounding the nucleocapsid. And studying to the surface of these glycoprotein spikes. Those for Ebola virus form 450 killadalton trimers and they are quite heavily glycosylated.
So the question you might ask is – if this spike is important for attachment and entry, what does it look like and how does it work?
We had to make about 140 versions of this GP to get one that would crystallize well and we had to grow about 50,000 crystals to get one with the [frac bone]. Before we have a structure, we typically think of a protein as a schematic like this with an N terminus and C terminus. This GP is cleaved in the producer cell, with GL2 sub units. A GP1 which mediates receptor binding and GP2 which mediates fusion. So the GP1 has receptor binding domains and the GP2 has heptad repeats that undergo a conformational change and collapse the six-helix bundle driving membrane fusion. Also in GP1 is this unusual mucin-like domain, it’s very heavily glycosylated. Each domain is about 75 killadalton, it’s 3 in the trimer, there is a lot of unstructured protein and carbohydrate.
So this is the crystal structure of the Ebola virus GP. The first one we solved — you can see the 3GP1 subunits in blue and green, these mediate receptor binding and they are tied together at the bottom by the GP2 fusion subunits.
Now there is something interesting here. When you think about a fusion peptide for Flu or HIV, it’s a hydrophobic peptide that’s tucked up inside the structure. Well here the fusion loop is tacked under the outside like a [flice water]. The one that belongs to this monomer reaches around the outside of the trimer and binds into the next one over. In order to get this molecule of the crystallized, we had to exize that mucin-like domain.
But we want to understand what the real GP looks like on the viral surface. So it’s got these heavily glycosylated domains attached at the top. Well note GP containing that mucin domain crystallizes, so we had to use a different technique which was small angle X-ray scattering, tiny x-rays and protein molecules tumbling around in solution, you get a low resolution view, maybe 10 angstrom resolution. And in this, that it turns out this is the solution scattering envelope of the complete fully glycosylated Ebola virus GP, with all of its sugars and all of its mucin like domains. So the crystal structure I am showing you in the ribbon, in the center for scale, these are the mucin-like domains attached. So they effectively triple the size of the molecule. When Peter Kwong coined the term glycan shield, this is hell of a glycan shield. They reach about 100 angstrom away from the core of GP and they are quite flexible. So I would actually expect the actual width of this domain to be half that, I think we’re visualizing solution, a lot of flexibility in waving our [camera].
The salient feature of this is that these mucin-like domains are massive and they dominate the structure of the GP.
So this is what is on the virus surface. How does it work? How does it find and get into a new host cell? Well, this I’m showing again the crystal structure, I am coloring the surface white, patches that are colored pink, are areas that mutagenesis tells us are important for infectivity. They are a little bit sequestered inside the bowl shape the trimer makes. The rest of these are most important for a receptor binding are very sequestered. In fact, inside structure underneath this domain. So that is sort of a representation of where the mucin-like domains are. The parts that are important for the receptor binding are these pink ones and they are underneath these domains called the glycan cap, these have a lot of glycan attached to them.
So, does this make sense and how on earth is this a functional receptor binding site underneath this entire canopy of protein and carbohydrates?
The answer is that it is known from biology that GP needs to be cleaved by host capsaicin enzymes for infection to occur. The filovirus is different the requirement for some cleavage but it’s especially important for Ebola virus.
So, why? Well, in solving the crystal structure, we see that all of this structure, the glycan cap and the whole mucin like domain are attached by a single polypeptide tether that connects residue 189 to 213. And that piece of polypeptide is disordered. So something that is disordered in a crystal structure is flexible and it’s moving around. So this looks like a pretty attractive cleavage site.
If proteases were to cleave on that yellow loop, this would be the effect, which is much better exposure of the receptor binding sites. Now we are not making that up. This is actually the crystal structure now cleaved GP and another lap simultaneously shows that yes, cleavage indeed strips off 85% of the mass of GP1 leaving the receptor binding sites exposed.
So this is what the protein looks like on the viral surface. What do we learn from this? Receptor binding probably doesn’t happen at the viral surface. Just by looking at the structure alone, you can see spots needed to bind that receptor are not accessible. They are not well exposed in this kind of protein. Instead, the virus that bears this surface enters cells by macropinocytosis.
Once in the endosome, this is cleaved to strip off all that surface sugar in these mucin-like domains leaving the receptor binding site well exposed and allowing binding by the receptor Niemann-Pick C1, NPC1. And this binding site is right there where the glycan cap used to be.
So what we see here is one polypeptide, that results in two different biologically relevant manifestations. This is the molecule subject to antibody surveillance and this is the molecule that’s functional for receptor binding.
So what does that mean to the immune response? Well, nothing good. Many can be clipped right off. In fact, in a lot of vaccination studies, these sites can be immunodominant. You can see that any antibody that binds to these mucin-like domain epitopes is going to be cut right off in the endosome leaving a perfectly functional receptor binding core that is now antibody free. So those kinds of antibodies don’t neutralize.
The essential conserved sites are not well exposed. So for example, all of these filoviruses share the same receptor, so that’s a conserved binding site, it’s an essential site for the molecule. You’d love to target that with an antibody. It’s completely or partially hidden under the glycan cap and the viral surface, so the antibody might not see it unless you found a way to engineer the antibody.
Because of this conundrum, we are left with a bit of a puzzle that neutralization and protection don’t always correlate for Ebola virus. So neutralization is your ability to inactivate the virus in vitro. Protection is your ability to save the animal in vivo. So for example, antibodies like this, this is the human kz52 from the survivor of the Kikwit, Zaire outbreak neutralizes brilliantly and doesn’t protect the primate. Antibodies like these, including two that bind the mucin like domains, and one that binds up with the glycan cap don’t neutralize but they do protect the primate. So this doesn’t make a lot of sense, leaving you wondering what works here.
We had this result years before and it really cooled everybody’s opinion on antibodies against Ebola virus, maybe it wasn’t even going to be possible to protect animals but then it turns out that you can. These are quite protective even if you wait long enough for hemorrhagic fever to develop. The difference might be that these are given in a cocktail as this was given alone. So does that mean we have to have a cocktail? It’s the replicative capacity or the length of the number of spike of Ebola virus such that we need to have multiple antibodies against multiple sites and if so, which ones do we put together? I mean two-thirds of this cocktail is mucin, does that mean that mucin works? Or is this one the champ that binds the top? We don’t know.
Now in the field we have about 200 different monoclonal antibodies identified against the virus. What do you put together in a cocktail. Well now I’m going divert a little bit from my theme about the proteins of the virus and then tell you how we can use the structure to get at that problem.
This is the website that the viral hemorrhagic fever immunotherapeutic consortium – you will be able to find this link through Scripps very soon. What this is that more than 20PIs in 7 different countries have gotten on the same page. We put almost all the antibodies known against these viruses together in one pool. We blinded them and then compare them side-by-side to see what is more effective. We’re going to try to see if we can correlate efficacy in-vitro and in animals and – to understand how to put together the right cocktail. Right now we have three from the army in a cocktail that neutralized and we have three from Canada in a cocktail that neutralized what if the most effective cocktail is one from japan and one from the army and one from Hamilton. We won’t know until we put them all together. And so it’s nice that everyone is on the same page in the same study.
Okay. So until we make that cocktail, let’s assume that viral infection will proceed. So the next step of viral infection — after the viral membrane is fused with the host endosome membrane, genetic material enters and the virus starts to replicate.
Now, something important happens here. Most people die from Ebola virus infection – 50% to 90%. Some people live. What is the difference? The difference seems to be that those people that survived the Ebola virus infection tend to generate an early and strong immune response against the virus and the viral titer starts to drop by around day 4. Those people that ultimately succumb to the Ebola virus infection are more likely to be characterized by a very poor immune response and their viral titers get quite high — 10 to the 9 to 10 to the 10 per mil at the time of death. So for this decision point to occur by around day 4, that means that the innate immune system is quite important in making this decision of survival or not survival.
So what is the viral factor at play in this innate immune decision point? One of them is a protein called VP35 — viral protein 35 Kilodaltons, that’s how it got its name. It’s a component of the nucleocaps in replication complex. That’s his day job. It also has another job – it’s interferon antagonist. And what it does in that function is to bind double-stranded RNA. Now you typically would only have double-stranded RNA in the context of a viral infection. And so it is a pathogenesis through the molecular pattern. Your innate immune system has sensors that go looking for double-stranded RNA and when they sense they mount to antiviral response, not having to know what the virus is, just the presence of a viral infection.
So how does one this work? This is a crystal structure of VP35 bound to double-stranded RNA. So the double-stranded RNA here is an 18-bp oligo in green. We have got four copies of VP35 bound to it. Now this half is identical to this half in the structure. So you can really only look this half if you want.
This is not the mode of ligand binding that you learned on your mother’s knee as a biochemist. So what you typically think of when you think of a protein binding of ligand is that it has one binding site. This laser pointer is a ligand. My hand is the protein, it binds in the palm and that is its binding site, perfectly shaped for it.
What we have here is the same protein binding in two different ways. Two copies bind the backbone, two copies cap the end. These are the identical proteins. You could pull off the end capping, want to roll it around and attach it by the backbone binding. They use different binding sites to do that. The end capping is a hydrophobic patch, and the backbone binding one uses a hydrophilic patch, this residue being central. So instead of the ligand binding in one site, you have two identical copies of the protein and one binds this way and one binds this way. It turns out that dimerization is essential, point mutations of block at dimer interface attenuate the Ebola virus.
So after you form this asymmetric dimer on the end, it then goes on to spiral around the RNA. But it is interesting that it has repurposed itself from nucleocapsid protein to have this additional function and use different sides of itself in order to make two different binding sites. So polypeptide has got two functions, replication and immune evasion and the same protein has two different binding sites for the same ligand.
Well here is this protein. I am going to show you next – has a different strategy for managing double-stranded RNA. This is the nuclear protein of Lassa virus. So the day job of the nucleoprotein is to bind and play a role in replicating the viral genome. These are RNA viruses that protect RNA genome by having it continually bound by a nucleoprotein.
Lassa genome has four genes again. This protein has another function – it’s also an interferon antagonist. But it was known that it was immunosuppressive but it wasn’t known how it’s immunosuppressive.
So how does this genome binding protein suppress immune signaling? We didn’t know. So we solved the structure. Here is the structure. It’s got strands, helixes and loopy bits and bound zinc. That didn’t tell us anything. The sequence of this protein wasn’t anything in gene bank. It only looks like another arenavirus nucleoprotein. There were no other structures for those. So we haven’t learned anything from the sequence.
So then we asked ourselves – well is it structure like anything we’ve seen before though the sequence isn’t. So we did a dolly search for things of similar fold and we found one. So in green, it’s a Lassa virus nucleoprotein. I’m going do overlay some other proteins, they are all exonucleases of the DEDDh super family. This is ISG20, DNA polymerase iii epsilon subunit. You can see the folds are similar. They have the secondary structural elements in the same places. So this is the silent super family of nucleases. It has a Lassa, it has a similar fold even right down to the exonuclease active site. So all of these enzymes are characterized by the DEDDH. So these catalytic residues – at the site. The
Lassa nucleoprotein is colored green. So it’s got the same residues in the same place. And if you went back and looked in the sequence, you could see that those were there and those are conserved across the immune-viruses but the spacing wasn’t anything that you could appreciate that was going to wind up being an exonuclease until we saw the structure.
So it looks like an exonuclease, does it function like an exonuclease? So to answer that, we gave it all kinds of oligos, some DNA, some RNA, single stranded, some double-stranded and it digests some of them. An exonuclease digests nucleic acid, this one does too, the only thing it digests though is double-stranded RNA. So the other exonucleases in the super family can be more catholic in their subtract specificity. That’s a number of things. This one only digests double-stranded RNA. The pathogen associated molecular pattern and we think that enzymatic activity is linked to the immunosuppression that we see in Lassa virus infection, because when you make point mutants in around the active site, or the disruptive structure, so the wild type protein digests double-stranded RNA. The mutants don’t.
If you look at an [IRT] reporter activity, the wild type protein suppresses it, and the mutants don’t. and the ones that disrupt the structure have varying effects. So you knockout the exonuclease activity, you knockout the immunosuppression.
So now here is a structure of the nuclease, complex double-stranded RNA. The yellow strand feeds right into the active site. The paired purple strands and non-substrate strand just above up. And we can look in here and compare this to other exonucleases and see there are only two amino-acids to give the Lassa its unique immunosuppressive substrate specificity.
And so what it is doing is maybe rapidly erasing the thing that the immune system is looking for. A double-stranded RNA is a replication intermediate of a single stranded RNA virus maybe as one domain that the protein binds the genome, the other domain coupled to it comes along and raises it. We are still trying to figure out how that works. But what we do see that this structure and that motif [inaudible] activity seems to be shared among the arenavirus family. This is a family of 50 different viruses that exist in nearly every continent.
So, an enzyme with a number of human pathogens looks like it could be an effective target for broad spectrum antiviral and I’m looking for someone to work with me on that. So what we see here in this example is a polypeptide with multiple activities. Its day job is to bind and replicate the genome. It also erases the key signature that would spark innate immune signaling.
So the virus has entered the cells, suppressed immune signaling and replicated, and its next job in this life cycle is to assemble new variants and bud up. That occurs by a protein called Matrix. For Ebola virus it’s called VP40 – viral protein 40 kilodaltons, it’s this protein here.
So the matrix is the layer right underneath the membrane, so between the membrane and the ribonucleocapsid and that’s what gives the virus its shape. So if you transfect cells of VP40 alone they will assemble and bud out virus like particles that look just like Ebola virus. So all of the information that you need to build and bud are filamentous envelope particle is contained in VP40.
So, how does it do that? What does this protein look like?
The first crystal structure of VP40 was solved 13 years ago now. Here it is. It has an N-terminal domain and a C-terminal domain, was thought to be a monomer. But what is interesting about a matrix protein is not what it looks like as a monomer but how it assembles, how does it build a matrix? So they knew at the time if they tinkered with the VP40, and the tinkering could be cutting off C terminal regions or incubating the protein at 3 mM, they could get it to form rings.
So here is EM of the hexameric ring. Here is a crystal structure of an octomeric ring. So in this crystal structure they express the N-terminal domain by itself without the orange C terminal domain. Eight of them make this ring and unexpectedly, it pulled out RNA from the E coli expression system. There is a little oligo and orange bound to each one of the eight copies of VP40 in this ring.
So for the last decade, that has been our only model for how VP40 could assemble. And there is an awful lot of effort that has gone into designing drugs to inhibit ring formation, to inhibit matrix formation. There is a lot of models of Ebola virus generated by taking this cheerios and making linoleum pattern out of them and wrapping around the filovirus.
But there are a number of problems with this model. The first one is that the rings are not actually found in purified variants. So if they are not in the variant, are they a component of it? They are found in infected cells, however. It’s just not the actual virus.
The second problem is that there is no RNA in the virus matrix layer. So what that RNA bound to that ring is — wasn’t entirely clear. The RNA is bound to the nucleocapsid at the center not in the matrix layer with the membrane.
The third problem is mutations that prevent ring formation gave perfectly normal looking viral particles. So if you abolish the ring you can bud out a normal looking virus. Now the crystallographers that solved this structure didn’t think this is how the matrix was assembled because they did all this work but the field proceeded as even if VP40 made these discrete rings, something about this structure was how the virus assembled.
How does it assemble? We didn’t intend to do any of the work. I’m going to show you next. We were making VP40 for some other reason, we want to pull down the some of the viral protein. And what we noticed is that whenever we purified VP40, it came out as a dimer, not a monomer. So we are using now size exclusion chromatography coupled to multi angle light scattering. So it’s a more sensitive method of determining molecular weight and something that wasn’t widely available when the first structure was determined a decade ago.
So VP40 was always a dimer. Does that matter? We were looking for a different way that VP 40 might [inaudible]. What we had on this protein, we had these robots, so we grew some crystals, it effectively right away solved the structure. Here is the structure. We see the N terminal domain, we see the C terminal domain. So the structure from dimer is colored. Here is the structure from the monomer. No change.
So, the revelation that it was a dimer instead of a monomer hasn’t told us anything about the fold of the protein. But it was the piece of information that we needed to go looking in the crystal packing. Because we knew that it was a dimer in solution. So somehow and how those proteins assembled in the crystals we are going to see the dimer interface.
This is the crystal packing. So the N terminal domains are blue, C terminal domains are orange, so proteins are oriented like this down a filament, so they make this NN-CC-NN-CC filament. Somewhere in this is the dimer that floats around the solution. So it’s the dimer made by the blue-blue N to N interaction or the orange-orange, C to C interaction.
Well, the blue-blue varies a lot more molecular surface but the proof came from a point mutation we made, this one leucine 117]. So the dimer interface is probably the blue one and this is the dimer that floats around the solution, kind of looks like a butterfly.
Incidentally that leucine 117 is on the outside of the ring. It’s not involved in any ring assembling interfaces. So let’s have another look at that filament. Here this belongs to Ebola virus. Looking at it on the side view, roll it around and there is the top view. This is how the crystals assemble. All those filaments line up side-by-side.
Well, we wondered if that was interesting. Is this assembly physiologically relevant or is this just some artifact of crystal packing? Well the odd thing we noticed is that no matter how we tried to crystallize VP40, we always got the same filament. This group C2, base group [inaudible] No matter what species of Ebola virus we worked with or which kind of crystal symmetry arrangement we got, we always got the same filament organized the same way. The NN-CC-NN-CC filament. Here, the rigid and they line up next to each other and the defract well. Here they form four twisted around each other. Here they make a 10-stranded conduit tube and don’t refract too well. But they are always assembled by the same first interfaces.
So, that’s starting to get uncanny. You crystallize the protein four different times and it makes the same assembly every time, maybe it’s something it wants to do. So the protein transcribed as a monomer, forms a dimer in solution, every time you crystallize a full-length protein we get the same filament. Under some other circumstances, you could cut off the C terminal domain and form a ring and there is one crystal structure of that.
So, which one of these assemblies make the viral matrix or do neither one of them? To answer that question we made mutations to each one of these interfaces because this assembly and the ring assembly are built by different surfaces. Amino-acids that make this filament are on the outside of the ring and amino-acids that are important in RNA binding ring formations are not what assembled this filament.
So let me show you those mutations. So this is the dimer interface. You’re looking top down at the butterfly. Blue-blue. So for example leucine 117 and Thr112 are important to the dimer interface. If you mutate them, you get this. So the wild type protein is the dimer. If you mutate the dimer interface, you get monomer and ring. Can you all see that or do we need to dim the light little more?
So if you mutate the dimer interface you get monomer and you get ring. If you transfect cells and now we are staining VP40 green to attract the green. The wild type protein traffic to the cell membrane and buds out these filamentous virus like particles. So you can see some length wise in many cross section. The monomer and ring mutations don’t traffic as well to the membrane and don’t bud anything at all. So that N to N dimer interface is important for matrix assembly and budding. Even if the mutants make ring.
How about the filament that we keep seeing in those crystals made by packing sideways of the dimers? That is made by those C to C interactions. So if that interface is methionine 241, Leu307, so let’s mutate those. Wild type protein is a dimer. This first mutant is also a dimer. That’s we expected, right, we’re mutating a cell that dimers assembles make the filament. So we still have the dimer.
The wild type protein traffic to the membrane and buds out these virus like particles, it also makes these funny little membrane ruffles, we don’t know what they are. But the wild type protein does that. This mutant protein doesn’t traffic quite as well but when it gets there, has this crazy ruffling morphology. It doesn’t successfully bud any viral particles but it makes this membrane have this funny ruffling effect. So we solved the structure of the M241R mutant to find out what was happening. You have the same dimer, here you can see green butterfly, here is the blue butterfly. But instead of being packed side-by-side like every other crystal structure, we mutated the interface and they are twisted relative to each other. So it might be that when they traffic to the membrane, they are making some kind of funny twisted filament that’s making that ruffled morphology but can’t quite get-together and release the virus.
This other mutant is different. It doesn’t make dimer. It only makes rings. And these rings bind RNA. The VP40 dimer doesn’t bind RNA. The VP40 filament doesn’t bind RNA. Only the ring binds RNA. And we looked at these by EM and we’ve got for the same size and shape as the other rings made by deleting the C terminus. The RNA binding rings do not traffic to the membrane. Instead they hug the nucleus and they don’t bud anything at all.
So what we see from those experiments is that disrupting the interaction that built that filament prevents virus assembly and budding even if you’re making only rings.
Now let’s break the ring. This mutation was previously known by Seven Bakers Lab, it prevents RNA binding and ring formation, it’s a dimer. It traffics to the membrane. It buds out virus-like particles and the same kind of membrane ruffles and looks identical to wild type and it assembles and buds just fine, the same morphology, same number of particles. You can’t tell it apart from wild type.
So what we conclude is that something about that dimer and filament is involved in virus assembly, not the ring.
So how different are the dimer and the ring? Well they are pretty different. So it’s easy to see how you’re making this filament just by length wise assembly of the dimers. To make the ring, you have to separate the NN-C terminal domains from each other, split the N terminal domains apart, unravel the 70 amino-acids that make the interface and then rotate the two members of the dimer from parallel to antiparallel and backwards. So they are going to reassemble now by the green interface that used to be hidden by the C terminal domain, used to be attached with the L interface. And then four of these antiparallel backwards dimers then make the ring and this ring then has an RNA binding site in the center that wasn’t available for the filament.
So we think this is something to do with the virus assembly. How? There are three questions you might ask yourself. The first one would be what side of this thing interacts with membrane? Well, we knew that the interaction with membrane was electrostatic because you could salt it off. We knew that involved a C terminal domain. So if you look for a basic patch in VP40, there is really only one and they are on the same side of this filament. In that basic patch are five lysines that are conserved across the Ebola viruses. 4 of those 5 are essential for membrane interaction and budding viruses. So probably this surface of the filament is the one that interacts with membrane.
The next question you might ask yourself is, is this it? Is this filament how you build the filamentous virus? Well there are a lot of satisfying things about this model. I mean all of the interfaces we see there we think are essential because any time you mutate them you no longer build and bud a virus.
But, there were two other pieces of information that this model doesn’t address. The first one is that interaction with membrane induces oligomerization of VP40 in Hexamers and by our model, we see 2, 4, 6, 8, 10, with no jump to hexamer.
The second thing is that interaction with membrane seems to do some kind of conformational change between the NN-CC domains and this model didn’t answer that either.
So the third question that we asked ourselves and you might be asking yourself right now, there is something different happened to this structure when it makes it electrostatic interaction with membrane. So we went through a series of attempts to try to satisfy the positive charge with that basic patch. Now it is known that phosphatidylserine is the natural ligand of VP40 in the membrane and that the molecule dextran sulfate will not compete phosphatidylserine. So we soaked a lot of crystals in phosphatidylserine as the head group, we tried to co-crystallize with it. We finally found success with dextran sulphate. If we incubate VP40 with dextran sulphate, grow crystals inside of the structure, what we get is the a VP40 that is now hexameric with N and C separation. We get this.
So let me walk you through this hexameric structure. The N-terminal domains are still blue. C terminal domains are orange. This is one complete monomer with this N-terminal and C terminal domain. Here are two more. It’s molecules 1, 2, 3, 4, 5, 6 in the hexamer. These C terminal domains are still attached. Because if you run the crystal in a jelly, we see intact VP40, they are there but we don’t see them. They are disordered. They have somehow sprung into solvent channels in the crystal where they occupy a lot of positions. So we know they are attached but we don’t see them and they belong to these and we can see which direction they are going from the polypeptide chain that extends.
This hexameric building block forms this long zigzagging filament in these crystals and this is assembled by 3 different interfaces. The same dimer interface we were mutating before with the same Leu117 and Thr112, the same C to C interface before between orange to orange, that had the same methionine isoleucine and something else which we are calling oligomerization interface, that has a Trp95 at the center that’s exposed by release of the C terminal domain. The mutation of Trp95 also prevents matrix assembling. So we know by mutagenesis that every interface of this rearranged zigzag filament is essential for budding.
Well does this filament now fit what we know about the virus?
The Ebola virus looks like this in cross section, cut the fil in half, you have a nucleocapsid at the center and membrane on the outside and then the tomography tells us there are multiple protein layers between the nucleocapsid and the membrane. So matrix has a lot of protein layers in it.
If you look at the radial density of the virus inside to out, you see a big peak for the nucleocapsid and then bum, bum bum, a peek for membrane. So we call this inter-peak, central peak and outer peak in the membrane. Here is our zigzag filament. Turn around on its side and roll it over once more, here are 3 protein layers, inner layer, a central layer and an outer layer. They fit to scale what we know about the width of the variant which is always fixed, the width of the nucleocapsid which is already fixed, the space in between and the dimensions that we know of the C terminal domain and N-terminal core and the reach of these. This also fits the biology. We know from biology that it is the C terminal domain that binds membrane. We also know that it is the C terminal domain that binds nucleocapsid.
How can this happen unless some go this way and some go that way? So it fits what we understand about the virus assembly. It also fits the shape of the virus too. So here protein is white, not protein is black. We are shooting into the side of the Ebola virus instead of N node and the zigzagging filament seems to follow the checkerboard pattern shooting into the size of the virus and scale in repeating distances.
So what we have expressed as a dimer, makes this filamentous intermediate, at the membrane it seems to be some kind of rearrangement that makes this hexameric building block. And this is our current best model for how to make this assemble. Under some other circumstances, you can split the thing apart and rotate around and make this other ring that binds RNA.
What is this ring? Is it real? Remember this mutation by which we should prevent RNA binding and ring formation? It results in perfectly normal looking viruses in perfect shape and perfect number. This is a lethal mutation. You cannot propagate an Ebola virus barring this mutation. Why not? You can still build and bud a virus. Virus can still attach a matrix cell.
Why is this lethal? The RNA binding ring must do something. That’s what we conclude. It must do something essential in the virus life cycle. But what would that be? The recent discovery that VP40 has a second function — in addition to virus assembly and budding, VP40 controls viral transcription inside the infected cells. Well maybe this is what the ring is for. The ring is the only structure of VP40 that binds RNA. We only see it in infected cells, we never see the ring in the virus.
Well now we have new tools to bring to bear in situations, we found these point mutations that make only ring and point mutations that never make rings. So as a first path to put this into a minigenome assay. This we can do this at VS2, we have an Ebola start, Ebola stop, luceine placed in the middle and the wild type VP40 exhibits that control function. The VP40 that we’ve locked into that ring controls a little better. So if you anchor VP40 into the ring you get the same function. If you prevent VP40 from forming the ring, you get a lot less, it’s not a total knock out. Maybe it’s making partial structure. So that ring does seem to have some kind of function inside the infected cell of transcriptional control.
So the wild type unmodified end product here makes a dimer. The dimer is critical for trafficking to the membrane. It makes a filament to build and bud a virus and makes an RNA binding ring to control transcription inside infected cells. So VP40 is both a structural and a non-structural protein.
We are doing this all in bacteria so we don’t need a post-translational modification to do it. Maybe there is one in infection. But where there is no mutation there is no splash variation, so same polypeptide making different structures for different functions at different times.
What do you call a protein that does that? Well, we toyed around with some different names, bistructural,ambee form, finally we decided the Transformer was the right analogy. So transformers are these toys that refold a robot into a vehicle – truck, cars, it’s never a cup of coffee or grain, but they refold from one to the other. And what I like about this analogy is that you see the same secondary structural elements achieving different roles in the different manifestations. So for example, the tires are the seat of his pants and his ankles and the tires are of course tires on the truck. If you did not know that this truck existed, I’m going to block it out. All you knew is this robot existed. And you knew that sometimes this protein could walk and talk and shoot and sometimes this robot could carry a lot of cargo and drive very fast, and you wanted to find out why. They mutated the tires. It’s easy to see how the tires would flatten the cargo carrying capacity but if you didn’t know the structure existed and you were looking at this, you would conclude that this robot had rocket powered pants because the seat of his pants and the hand of his pants are essential for carrying a lot of cargo, driving very fast.
The head here is the hydrophobic core about which the truck is folded. So if you mutate the head, you knockout everything. So you’d say, the head is the thinking center. We were not allowed to use that analogy because the toy company wouldn’t let us and so instead, we called it molecular origami. And so, what you can think about in this situation is the protein as a blank sheet of paper that folds into different structures according to different needs in the virus life cycle.
And so we’ve just showed you how we think the dimer makes RNA binding ring and the same dimer rearranges to make the hexamer that builds and buds the virus. Viruses are compelled by evolution to be small especially RNA viruses. They don’t have a proof reading machinery.They have to keep below that error threshold.
How do viruses keep their genomes lean and mean? How do they do more with less? They can hijack host proteins for central functions. They can encode proteins in overlapping reading frames. It’s the same piece of nucleic acid that makes different proteins. They can have moonlighting 8proteins, so the same protein does different functions so the nucleocapsid also suppresses interferon signaling. This is a fourth flavor, here is the VP40 crystal structure, here is VP40 crystal and here is the VP40 crystal structure. The polypeptide that the gene encodes rearranges into different structures for different functions at different times, to get more function from less gene.
So this is what we bring to go to the virus’s territory. This is what it brings. It travels lightly. Because this actually, the few proteins that it does make, achieve a multitude of functions.
This is my lab at Scripps. We collaborate with a number of wonderful labs and very supportive. [inaudible] lab did all of the VP40 and these groups are working with us to develop antibodies against the Ebola virus and I’d like to thank the NIH very much for funding also Burroughs Wellcome Fund and Skaggs Institute For Chemical Biology. And I’d like to thank you for your attention.
Dr. Collins: What great stories. Please, if people have questions, there are microphones in the aisles. Feel free to come forward and ask what is on your mind. While people are thinking, I have to come back to what something you said early in the talk about the difference in who survives Ebola and who doesn’t in terms of who is able to mount some kind of an immune response in those four days. Do you have any idea what that is about? What determines whether you’re in the survival category or not?
Erica Ollmann Saphire: We do not. There are some obvious answers that you can rule out — health care status, nutritional status. The German scientists infected with Marburg had a better prognosis than people that were hungry in remote villages. But if you ruled that out, I think we still need to do the work to understand. I think that we know that the viral factors are at play but we don’t the human genetic factor that control resistance. We know these for Lassa virus. Ebola is quite new. Lassa is quite old. For example, Lassa virus has been in Nigeria for thousands of years, one at the Sierra Leone 150 years ago. You can see that people have evolved mutations in their receptor, sort of like a sickle cell anemia kind of thing. So they are less susceptible to the virus.
What’s happening in Ebola I think we don’t know. I think we need the people on the ground that know that tell us and look at the human genetics of the survivors.
Female Audience: Very interesting talk. And in a different version of question that Dr. Collins asked, my question is, do we know — you mentioned you don’t know but you know the thought is, are you impacting different composition of cells perhaps the mucin-like cells that can actually neutralize? Or is there anything known about the profile of immunoglobulin synthesis and/or the affected cells that could either neutralize and produce immune responses, or they are not affected immediately?
Erica Ollmann Saphire: I don’t think we really know the answers to that question yet, because we haven’t done as much in-depth analysis of the human survivors as we need to have done. So a lot of the studies are ongoing. It does infect most cell types, monocytes and macrophages to begin. So it’s quite a prolific virus but I think that we haven’t really looked enough at the survivors to find out what the difference is in the cell types.
Male Audience: The ring and the filament, they coexist. So, basically something is determining the folding pattern. Do you know what that is?
Erica Ollmann Saphire: No, and it’s killing me. What is the trigger for this? What makes it do one thing and – so this is specific game number one. So I don’t know. We can speculate. So maybe at some stage of the virus life cycle there is a lot of viral RNA and what I have drawn in that model is RNA binding and drawing a wedge between the domains and kicking off the C terminal and opening up. If it’s not RNA, maybe it’s an RNA protein complex like the polymerase or something like that.
Is there a chaperon? It could be. What we are trying to do now is use these point mutations that we have that lock it into only ring or never ring or either and see what they pull down and based on what it pulls down does it tell us if it means a host factor to drive these or is it purely a viral factor? Is it RNA? And if it’s RNA, which RNA. The RNA that picked out from the E coli expression system in that ring crystal structure was UGA, UGA, UGA. So does it recognize stop codons? Is that what it is trying to find in the virus life cycle, does it look for the end of the gene? We don’t know. And those are exactly the kind of questions we are trying to ask. Also what is the thermodynamics of this? Are these all equally stable or do you need the input of energy or chaperon in order to get from one to the other?
Male Audience: There is a kinetics difference in the synthesis of this?
Erica Ollmann Saphire: We don’t know. We know that – we have trafficked where VP40 is at different stages of the virus life cycle. Early it hugs the nucleus, and maybe it’s in the ring formation that’s making copies. Later in the virus life cycle, it catches a ride on microtubules up to the surface and then it makes filaments and buds out. So the location traffics may be with its function. What causes it to make the different structures we still need to figure that out.
But what is interesting about that is it is a different perspective on the protein folding problem, right? Instead of an infinitely unknown, unfolded thing to one single folded structure, we have a folded structure, another folded structure and into a convergent and are these thermodynamically equal or different or what can we learn about the protein folding problem getting from one to the other? Can we learn something about information and coding that we have multiple functions encoded in this one piece of code.
Male Audience: So, the exonuclease you talked about, one would envision that it might actually do harm to the replication of the virus too if it get activated in an opportune time. So is there a timing in terms of when this viral encoded exonucleus gets activated during the life cycle of the virus so that it allows it to evade immune detection but doesn’t hurt its own production?
Erica Ollmann Saphire: That’s a fantastic question. That’s similar to what I wanted to ask Dr. Kraugg [ph] earlier today. We don’t know. We know that one domain binds genome and the other chooses double-stranded RNA and for the protein to function they must be genetically linked. You cannot supply those separately. So they have to be tethered together. So you need the one function attached to the other. The NP doesn’t exist as a monomer, it’s an oligomer that makes the ribonucleoprotein particle. So somehow this interacts with its friends. So, we also know that the linker is quite protease sensitive in that in actual infection a lot of C terminal domains go free. So is it the C terminal domain having gone free that is scrubbing out double-stranded RNA or is that just an accident of protease contamination? Or is this exonuclease function physically tethered to the replication site or to erase some intermediate as it is being made or does it have some entirely other function that we never thought about?
Male Audience: Or does it do any genetic editing to the viral genome itself?
Erica Ollmann Saphire: Oh that is another question. We wanted to keep cycles of replication going to see if it had a proof reading function. We don’t know. We love to do those kind of studies.
Male Audience: So I have a second question about this RNA binding towards by the protein at the end and also on the side. So, I guess you mentioned that chemical nature is totally different, one is hydrophobic and another is electrostatic. Which one actually dominants – I mean if you analyze the binding affinity – obviously the cap end binding has a disadvantage because it needs each strand to bind two. If it binds on the side you can bind multiple copies.
Erica Ollmann Saphire: So we haven’t — what we should have done and haven’t done is see how well it binds to a circular double-stranded RNA that has no end. We see what happens is it makes the cap first and then this backbone binder does the same interaction all the way down the rest. So if you keep polymerizing once it has this cap on there. Marburg virus doesn’t need the cap and it is happy to just polymerize. And so, another group has been doing similar structures and they see the same thing. Every time you crystallize an Ebola virus VP35 you have an end cap and every time you crystallize the Marburg VP5, you have a spiral. So based on that, I would say it is the backbone binder that might be dominant but we haven’t been able to really tease them apart yet.
You do lose binding when you have an overhang that projects into the space that needs to be occupied by the end cap. So, I don’t know what that says about relative infinity but it seems you need the end cap in order to get it going. The affinity is not high. Maybe 10 or maybe micromolar. So maybe you need some kind of affinity effect.
Male Audience: In the crystal structures of VP40, you showed a hexameric ring I didn’t see later. Is that involved in the cell during infection at any point?
Erica Ollmann Saphire: Hexamer versus octomer? We don’t know and we don’t even know if infected cells is even a complete ring or what it is — it’s the same interface 7but split open and spiraling around the nucleocapsid. We don’t know. We just know that you can get it to form both kinds of rings and one crystalized and the other didn’t.
Dr. Collins: We will now invite everybody to walk up the hall to the ribbon cutting. Before you do so, please join me in thanking Erica for a really fascinating seminar.