AUSTRALIAN FRONTIERS OF SCIENCE, 2008

The Shine Dome, Canberra, 21-22 February

A common fold mediates vertebrate defence and bacterial attack: Structural studies on a MACPF domain containing protein
by Dr James Whisstock

James Whisstock is an NHMRC Principal Research Fellow and Monash University Senior Logan Fellow. His primary fields of research are structural biology and bioinformatics. He leads the NHMRC Program Grant on protease systems biology and is a Chief Investigator on the ARC Centre of Excellence in Structural and Functional Microbial Genomics. He is particularly interested in the function and dysfunction of proteases and their inhibitors. In 2006 he was awarded the Science Ministers Prize for Life Scientist of the year.

I am going to be telling you a story today about how proteins change shape. Nick Dixon mentioned my work on serpins, which looks at how proteins undergo conformational change, and I think there are some interesting parallels with the family I am going to speak about today.

(Click on image for a larger version)

I want to talk about pore-forming toxins. These are molecules which are most famously associated with bacteria, bacterial pathogenesis and tissue destruction. These are very, very unpleasant toxins which, basically, can wreak havoc and destroy tissue.  Such toxins are produced by bacteria such as Staphylococcus aureus, gas gangrene, and other very unpleasant diseases. The bacteria that cause these diseases produce unusual molecules which have to perform an extremely special feat indeed in that they have to switch from a nice, happy water-soluble form into a membrane-bound form.

Most proteins exist in either soluble states or membrane-bound states. These molecules can do both, and the way that they generally do that is to hug their hydrophobic water-hating surfaces to their hydrophobic core, until they are ready to insert into membranes, where they unleash a membrane-spanning segment.

(Click on image for a larger version)

As I said, pore-forming toxins are commonly associated with bacteria and bacterial pathogenesis. They are, of course, found in us as well. Over 100 years ago, the Nobel prize winning immunologist Jules Bordet discovered a factor in blood that he called 'complement', with the capacity to lyse Gram-negative bacteria that is, bacteria which have an outer lipid bilayer. This was very exciting, because it suggested that there was this lytic, lethal factor that had the ability to protect us against infection.

Many, many years later, in 1984, Eckhard Podack, who still works in the  field, discovered a cytolytic protein produced by T cells, immune cells, in our blood that could destroy tumour cells. So it is this protein secreted by these immune cells which has the ability to burst and destroy tumour cells. He called this protein perforin.

(Click on image for a larger version)

A couple of years after this, when genome sequencing and gene sequencing really started in earnest, Jürg Tschopp noticed that both perforin and all of the components of complement C6, C7, C8 and C9 contained a common region that he called the MACPF domain, the membrane attack complex/perforin domain.

This type of picture is the sort of picture chemists put up when they don't know what things look like a blob-a-gram. We know that we have got different types of protein domains here at the left of the slide; the ones in the pink boxes are all the same type; but the one which I am really going to focus on today is the green box in the central section of the slide, because biochemistry and biochemical investigations have suggested strongly that this is the critical region that contains the membrane-piercing, the lytic activity of all of these molecules.

One point I should make is that the two 'classic' members of the MACPF family are C9 and perforin. C9 has the ability to form pores, and perforin has the ability to form pores. Some of the molecules in the green box also have a MACPF domain, but they don't seem to be lytic independently. They seem to act more as scaffolding proteins. But generally many members of this family function to form pores.

(Click on image for a larger version)

So why do we care about the membrane attack complex? It is interesting, in that the MAC is the terminal part of the complement pathway. What happens is that C6, C7 and C5b assemble on the surface of the invading bacteria to recruit C8 all of these contain MACPF domains and then this wonderful pore is formed by C9, which punches a big hole in the membrane of the bacteria, resulting in cell death. This is a very dangerous activity indeed. You do not want unregulated complement. In fact, there is a deep irony, in that complement sometimes seems to cause more damage by its presence than by its absence. You can actually live without the terminal parts of complement, but overactivity of complement can be very dangerous indeed.

So we have a host factor called CD59, a little protein which sits on the surface of all of our cells and, basically, is an effective inhibitor. It binds to C9 and C8, and stops it from bursting our own cells.

If you lack CD59, if you are deficient in it, it is very bad news and you develop the unpronounceable disease paroxysmal nocturnal hemoglobinuria, which is serious anything which results in blood in the urine generally is serious. So, basically, lack of CD59 results in lysis of your own blood cells by complement.

Complement is more dangerous in inflammatory diseases than in transplant rejection. An overactivity of complement is something which we want to stop. You can live if you are deficient in the terminal parts of complement, but you are more susceptible to infections for example, Neisseria meningitidis or Neisseria gonorrhoeae.

(Click on image for a larger version)

This is a picture of an eye infected by Neisseria gonorrhoeae. It is not just a sexually transmitted disease; it also causes other unpleasant infections.

(Click on image for a larger version)

So what about perforin? Perforin is important for killing virally infected cells and cells which are on their way to becoming cancerous. We think this is part of the tumour surveillance mechanism in humans. It is released by natural killer cells. I look at this as a bit like the Harry Potter 'Dementors': the killer cells wander round the body till they find a cell they want to attack, and they 'kiss' it. When they kiss it, they release a load of perforin that is tightly packaged up inside the cell so it doesn't go and do any inappropriate damage; it also releases a protease called granzyme B. Perforin can lyse the cell and also permit this nuclear warhead to be released into the cell, where it can kill it. So, even if the cell repairs itself, it is too late because it has got a dose of granzyme B.

Basically, if you are deficient in perforin, it is very unfortunate indeed. You develop a commonly fatal immunodeficiency proliferative disease in childhood, called familial hemophagocytic lymphohistiocytosis (FHL) again an unpronounceable disease, which is also reasonably rare. Treatment is by a bone marrow transplant. It is interesting, because plants have perforin-like proteins as well, and plants which are deficient in their perforin also develop a necrotic, overproliferative response which I will talk about a bit later.

(Click on image for a larger version)

The questions we want to answer arise, essentially, because since its discovery people have always wondered what complement looks like and how it works, what perforin looks like and how it works, and nobody has really had any mechanistic insight into what the actual major family of pore-forming toxins in the eukaryotes looks like and how it functions. In order to address that, we use protein crystallography.

(Click on image for a larger version)

These are a peculiarly evil bunch of proteins to produce. (We have worked on this for about nine years.) Most MACPF proteins aggregate quite cheerfully when you produce them, they express very poorly, they don't concentrate very well, they form pores when you don't want them to form pores they're an absolute pain. And so we decided, basically, to use bioinformatics to screen through all of the available genomic data to identify, for starters, all members of the family. We identified around 500 different family members. MACPF proteins are not just present in us; they are present in pretty much all eukaryotes. The only thing I can't find them in is nematodes.

Out of those 500 we picked about 10 which we thought would be good for structural studies. The reason we thought they would be good was that they were smaller and some of them were actually produced by bacteria. We thought that if bacteria could produce it, maybe the strain of bacteria which we use in the lab could also produce it.

We used a high-throughput expression strategy, and I estimate that we tried probably 800 or 900 different constructs, cell lines, variations on scaffolds, different solubility tags and so on.

(Click on image for a larger version)

We were pretty much ready to give up because we really weren't getting anywhere, when Carlos Rosado, in my lab, said, 'No, I just want to try one more thing.' And the one more thing that he tried did work. This was a protein from the wonderful organism Photorhabdus luminescens. I have got to tell you about this thing. This is like the Black Plague of the insect world. It is one of the great relationships. Photorhabdus lives together with a nematode, in a symbiotic relationship. The nematode comes along and sees an insect larva; it proceeds to inject the unfortunate caterpillar with Photorhabdus, and Photorhabdus then spews out about 500 different toxins which pretty much have the ability to destroy most insect life that you can find.

This thing is really quite evil, because not only does it kill the insect larva, or the insect, it also produces antibacterial proteins which stop any bacteria from coming along and growing in it, so the nematode itself has all of the insect larva to feed on without any kind of interference. And when it is finished, it re-ingests Photorhabdus and the whole cycle continues. In the First World War, apparently, if your wounds luminesced that was good news, because that indicated the presence of this bug producing antibacterials and improving your chances of survival.

Anyway, we found a MACPF protein in this, a two-domain protein with an N-terminal MACPF domain and a C-terminal domain of unknown structure. Carlos expressed and crystallised it. We can't find any lytic activity, but what I want to convince you about and what we used this for was to give us structural information on the entire MACPF superfamily.

(Click on image for a larger version)

This is one of the few times when, as a biochemist, I get to show a nice big tool. This is the synchrotron which we used, the Advanced Photon Source, in Chicago. (We now have our own synchrotron in Australia, which is fantastic.) This is a wonderful instrument and was important for determining the structure. I'm not going to talk about methodology, other than to say that we solved it using MIRAS (multiple isomorphous replacement and anomalous scattering) and we have a very nice, well-refined structure.

The first question is: what does the MACPF domain actually look like? And is it actually a domain, for example?

(Click on image for a larger version)

You see here the X-ray crystal structure. I know there are a lot of non-structural biologists in the audience, so I am going to try and keep the structural biology to a minimum, other than to say that the pink curly things here are a-helices and the big things with arrows are beta-strands. Lots of b-strands come together to form a b-sheet.

The first thing that we can see is that we were right in our prediction, in that this is a two-domain protein. We have the N-terminal domain here in pink and blue, and we can see the MACPF domain. The C-terminal domain is a b-prism domain and I am not going to be speaking about this today at all. I am going to focus entirely on the MACPF domain.

Shape is really important. It can tell you a lot of things. This is like a little machine, and so looking at its shape can give us insight into how it works. The first thing we noticed was a really weird at least it was to us bent b-sheet. There is a 90° bend in it. There are four strands and a huge bend in the centre. That started ringing some bells for us that is, we thought we might have seen that before somewhere. In fact, I was reading a review of Helen Saibil's and we were looking at it and things started to fall into place.

So we started performing comparisons with other toxins.

(Click on image for a larger version)

The first one that we really looked at was the cholesterol dependent cytolysin (CDC) from Clostridium perfringens. This was actually solved in Michael Parker's group, and our next speaker, Galina Polekhina, works on this family of proteins. I think one of the great things today is that we are bringing this family together for the first time, so I am not going to spend a lot of time on CDCs, other than to invite you to look at the bent sheet shown here it looks the same.

This is a really famous family of toxins, because it is produced by bacteria that we fear the most: things like Bacillus anthracis, which causes anthrax, Clostridium perfringens, which causes gas gangrene. They spew out this type of toxin and it is a really unpleasant and highly efficient lytic toxin.

(Click on image for a larger version)

The next thing we did was to look, really quite carefully, to see if the two structures were genuinely homologous, or genuinely related to each other. And we can see that they clearly are. I really want to draw your attention to the middle bit at the top left of this slide: the four-stranded b-sheet with two clusters of a-helices, and at the lower left we have exactly the same: a four-stranded b-sheet with two clusters of helices. So a complex core fold, central to both molecules, is conserved. This suggests that many hundreds of millions of years ago these two proteins shared a common ancestor, and that they split and have been evolving ever since.

That gave us confidence that the MACPF is actually a cholesterol dependent cytolysin. At least it belongs to the same family. But does it work in the same way?

With structural biology and bioinformatics I am often reminded of the following quote from James Whitcomb Riley: 'When I see a bird that walks like a duck and swims like a duck and quacks like a duck, I call that bird a duck.' Well, we have got two families of toxins. We have one in us, basically the MACPF family, which has the ability to form pores and lyse cells. We have another family in bacteria which has the ability to form pores and lyse cells. We look at the structures, and they look like the same type of structure. So do they work in the same way?

(Click on image for a larger version)

We were really fortunate, because of work from Michael Parker's group, Rod Tweten's group and Helen Saibil's group, in that we have a very good idea about how cholesterol dependent cytolysins work. They operate by taking the two helical regions at the left on this slide, allowing them to unwind, and then inserting into membranes to form what we call amphipathic b-strands. Imagine this happening when you have got 20 of these molecules oligomerised into a big ring: you've got a giant hole-punch.

So this thing assembles on the surface of the unfortunate cell which is targeted, and then those two clusters of helices on the left there punch their way in and, basically, cause a big hole.

(Click on image for a larger version)

When we look at the MACPF protein and compare it with the cholesterol dependent cytolysin, we can see that the key functional unit is indeed conserved. It is not the greatest superposition in the world remember these are very, very distantly related molecules but it is conserved and they do have the same functional bits.

(Click on image for a larger version)

So what else can we tell? Remember I was talking about CD59, which is the host factor which controls our own MACPF proteins and stops them from getting out of control. Jim Sodetz's group has done some lovely work where he has actually mapped the peptide binding sequence on C8a and C9, and we were able to map that sequence onto our structure. We predict that the CD59 binding site in C9 and C8a localises to the second transmembrane region. This is nice, because it suggests a mechanism for inhibition of the MAC by the host factor, CD59.

(Click on image for a larger version)

We can also start to explain and map dysfunctional disease-linked mutations in perforin, and these match regions which we know from cholesterol dependent cytolysins are important for how the molecule changes shape; for example, the big cluster of glycine residues, which includes G305 and G306. In CDCs we know this region is important for how the molecule opens up and during insertion into membranes.

We can also start looking at residues like this alanine 91 in perforin, at the top of this slide, which is carried by up to 17 per cent of the population and may result in a mild perforin deficiency. There is a lot of interest in this, because it is suggested that this may predispose to diseases like leukaemia. So there is a growing literature in that field.

(Click on images for larger versions)

To sum up my last few slides: we think that our structure suggests the MACPF proteins in fact belong to the CDC family, or vice versa; and we can have a model for MACPF membrane insertion, where basically the MACPF protein associates the membrane and inserts those helices as amphipathic b-strands.

(Click on image for a larger version)

The final point I would like to make about these molecules is that in our bioinformatics study we started finding a lot of proteins that really didn't fit into the mould as defence proteins. I was particularly interested in Peter Dodds's talk yesterday in terms of plants. Plants have a protein called CAD1, which is involved in immune defence. If you knock out CAD1, you get a necrotic phenotype caused by a cytokine storm which is freakishly similar to what happens if you are deficient in perforin and develop FHL.

If you ever get stung by a sea anemone, you can partially blame MACPF, because some sea anemone toxins are MACPF proteins. Malaria uses a MACPF. Two beautiful PNAS papers explain that malaria uses a MACPF protein to invade the liver cell lining, and also to invade the mosquito mid-gut. And Chlamydia, which is such bad news for our koala population as well as our human population, also produces MACPF proteins that we predict are probably toxins.

So these are all reasonably routine roles for MACPF proteins: defence, attack, pain and misery, and so on. However, an interesting point is that sea urchins have MACPF proteins which are involved in eggshell development, and don't seem to perform a defence function but instead have a developmental role. In insects, torso-like protein is fundamental for anterior/posterior development that is, development of a head and a bum. Again, nobody knows how it works, but it has a MACPF domain in there. And in mammals, astrotactin 1 and 2 are involved in neural guidance: astrotactin is important for nerve cells growing and getting to their destination.

(Click on image for a larger version)

So I hope I have convinced you that MACPF proteins are CDCs and this fearsome family of toxins is actually present in us, and that we use these for our defence activities but potentially for development activities as well.

(Click on image for a larger version)

I guess I look at this as a nuclear war, that over the last few hundred million years we have been in a nuclear arms race with bacteria they develop weapons, and we steal their weapon and throw it back at them. But, ironically, we also use this 'weapon' in developmental processes.

(Click on image for a larger version)

I would like to finish up by acknowledging Michelle Dunstone, who co-led the project, Carlos Rosaldo, who was critical for actually producing the molecule which we managed to crystallise, Ash Buckle, who built a lot of the structure, Ruby Law, who produced the heavy atom derivatives, and Bec Butcher, who really worked for a long time on a lot of the proteins which we were not able to produce. A lot of the project was really started by Bec.

I would also like to thank Phil Bird, Ilia Voskoboinik and Joe Trapani for their assistance with the lytic assay, and all of my other collaborators.