AUSTRALIAN FRONTIERS OF SCIENCE, 2008

The Shine Dome, Canberra, 21-22 February

Session 4: Discussion

Question: I have a question on the methods regarding the X-ray diffraction. Your models of the proteins are extremely beautiful, elaborate and complicated, but they are based on a measurement, the X-ray diffraction photographs. And from what I understand you are inverting those data using basically a three-dimensional Fourier transform, after losing the phase information. So you have got errors that you must propagate through the system. What is the magnitude of those errors, and do they actually impact on the details of the models that you are presenting here?

James Whisstock: It depends on your resolution. Obviously, if you have got a very high resolution structure which diffracts to 1.6 or 1.5 Ångstroms, then you are going to be very confident about the position of the atoms. There is a lot of cross-validation that we use, for example the free R-factor, which is, I guess, an unbiased measure of how well our model agrees with the experimental data.

We are, obviously, not operating at super-high resolution such as small-molecule crystallographers operate at, but when you look at different techniques such as nuclear magnetic resonance – which can be used to solve very large protein structures as well as X-ray crystallography – and X-ray crystallography and cryo-EM (cryo-electron microscopy), all the different methods agree. So I am confident of the accuracy of the data, depending upon the resolution.

Question: James, I am interested in the role, in both your talks, of the CD59 receptor. You mentioned that it bound to the components of the complement, and prevented the lysis et cetera. That is in the human system, where you have got a homogeneous system. But when now you have got a bacterium, what role do you think it plays?

And then, Galina, at the very end you said that perhaps that is the receptor whereby the toxin and the pores are formed. How do you reconcile those two sorts of roles?

James Whisstock: I will kick off by saying, as Galina did, that it is incredibly ironic: the inhibitor of the membrane attack complex is the receptor for Galina's toxin. In other words, it's a pretty evil trick.

From our perspective, CD59 interacts directly with the second transmembrane region, where there is a different binding site for intermedilysin.

Galina Polekhina: Well, it is actually, I think, right underneath but not to it. Domain 4 is for intermedilysin.

James Whisstock: Domain 4, yes. In intermedilysin it is the IG domain, the bit at the bottom, which basically binds CD59, whereas it looks as though the MAC inhibitor interacts with the second transmembrane region. What the evolutionary implications of that are, who knows, but it is a cruel irony.

Galina Polekhina: In the case of intermedilysin, that ensures the human specificity of the toxin. In MAC it ensures that it doesn't attack itself.

Question: In the rings of molecules, that very beautiful structure, you had different colours for each component of the ring. Did they mean anything?

James Whisstock: No.

Question (cont.): So my real question is: how did you persuade a molecule that likes to form rings like that to form a crystal with a long-range order that you could use in the synchrotron? If it is going to form a ring, why would it form a crystal?

James Whisstock: I should point out that the ring models that both of us showed were models. The ring which is reconstructed from single-particle cryo-EM data was published by Helen Saibil, so basically that is a cryo-EM reconstruction based upon liposome imaging – so image your pore in your liposome, repeat 10,000 times and then reassemble all the images. It is incredibly difficult and challenging work. We met with Helen a couple of weeks ago and she said it took about five years of image collection. I think Galina made the point, which is a good one, that a lot of these systems are very difficult to crystallise because they are non-homogeneous, so we have to make do with building models based upon what we see, using EM.

Nick Dixon: I think the most interesting proteins are never crystallised!

Question: Changing shapes of proteins may be important biologically, but it is a nightmare for crystallography. You have solved the point by making 900 different constructs. Looking back, can you tell why the one succeeded and others did not?

James Whisstock: No. This was the sledgehammer approach, because for a lot of them, to obtain them from source you simply couldn't obtain enough. To obtain them from blood, for example, you can purify C9 from plasma, but it is really difficult to work with. It tends to aggregate, it is metal-sensitive, and so on. I guess that if you could answer why some proteins express well and others don't, you would be a billionaire. Invitrogen would be onto you like a flash!

Plu-MACPF in Photorhabdus luminescens just happened to be the one. Ironically, now we have actually expressed this one, we have actually managed to express and purify several others. So why, I don't know.

We use different solubility tags. For example, in the system that was set up by Steve Bottomley we use Nus, GST, MBP and just plain His. Plu-MACPF only expressed at 16° in Rosetta-gami cells with a Nus tag on the N terminus, with not even a sniff of expression elsewhere – apart from unwanted insoluble material. Basically, that was the only condition that worked. We have no idea why.

Question: I am just curious why you think the nematodes didn't have the MACPF.

James Whisstock: That is a really interesting question. I was speaking to Peter Solomon yesterday about a fungal one we have which basically is from a nematode-eating fungus, and I reckon it is probably a toxin in that process. I don't know, but I speculate that it is.

We have looked pretty hard in Caenorhabditis elegans and can't find one. The one thing I would point out – and it is a technical issue – is that the MACPF family is extremely diverged. So you can't identify readily, using sequence-based approaches, the similarity between MACPF and cholesterol dependent cytolysins. It is about 3 per cent. And PSI-BLAST and SAM (sequence alignment and modelling) just really isn't good enough to pick that up.

So it is quite possible that C. elegans does have one, and that we just can't pick it up using a sequence-based approach, and one day somebody is going to solve the structure and go, 'Ooh look, it's a cholesterol dependent cytolysin!'