AUSTRALIAN FRONTIERS OF SCIENCE, 2005

Walter and Eliza Hall Institute of Medical Research, Melbourne, 12-13 April

Session 3: Discussion

Question – Jacqui, when you are looking at a four-unit complex like that, which has to assemble and then bind to DNA, theoretically to develop an inhibitor, either peptide or small-molecule, would you be looking for a single hotspot binding, in Jamie’s terminology, or a single-point binding, of low affinity and target that one – perhaps with, of course, a high-affinity binder – or would you be looking for a multipoint binding and high affinity, and therefore high avidity? On theoretical grounds, which one would you choose ab initio as much easier to disrupt, and therefore to allow you to have an effect at a pharmacologically acceptable level of the drug?

Jacqui Matthews – You might expect to inhibit a low-affinity interaction more easily than a high-affinity interaction, as long as you could find a very high-affinity binder that would work. If you want to inhibit something with high affinity, then you may need to administer less compound. So there are two sides to which sort of thing you want to target.

The other thing is also specificity, as opposed to affinity or avidity. So it may be better to actually target a couple of different sites, in order to generate the higher specificity, to see what is going on there. I don’t think there is any one answer for that question.

Question (continued) – But your outline there, which is perfectly logical and sounds very reasonable – experience shows that that is in fact the case?

Jacqui Matthews – I think there are many different methods that would work. One of the things I didn’t show here is that one of the reasons we wanted to target LMO2 protein, particularly, and by using these peptide inhibitors, is that all of the things that I showed you normally take place in the nucleus. But LMO2 doesn’t have to belong in the nucleus; it is not targeted towards the nucleus as many proteins are, in the same way that they are targeted to the mitochondrion. So it usually has to bind to something like that ldb protein in order to get in there and to stay in there. If we can provide it with a peptide that leaves it out in the cytoplasm, then it means it hasn’t got in there in the first place. So that is one of the other rationales for targeting that particular interaction rather than some of the other ones which are nuclear.

Jamie Rossjohn – I would also like to add that the concept of hotspots originated from the work of, I think, De Vos on cytokine receptors, where he solved this huge complex and they wanted to work it out to develop drugs. So they did all this alanine scanning, mutagenesis, and found out that hotspots existed – which was like a sigh of relief for the people who were trying to block those interactions.

Question – Jacqui, you drew that DNA as if it were naked.

Jacqui Matthews – Yes.

Question (continued) – So you didn’t take into account its interaction with histones?

Jacqui Matthews – No.

Question (continued) – Have you thought about the possibility, as seems to be showing up in a number of things, that small RNA molecules are also important in directing the complexing of proteins to histone-DNA complexes?

Jacqui Matthews – I guess the situation would be if we’ve got something that is being aberrantly switched on, then that complex has access to the DNA already. It is going to be cleared of histones. So what we are trying to do is block the interaction. Or it is going to be more or less cleared of histones. If you are going to have proteins binding to the promoter regions of DNA that is going to be expressed, they have got to be able to bind to that DNA.

Question (continued) – Often I think now the interactions involve the histones intimately and you get changing of the composition of the histones. So you get methylation or acetylation changes on the histones.

Jacqui Matthews – I guess my view is that that is where you are making something available to be accessed by transcriptional complexes, and yet there are a lot of more complicated things. A lot of these proteins with their other domains may well be interacting with histone-modifying proteins. It is almost a case that everything seems to interact with everything else, either directly or indirectly, so it is very complex. The more and more we look at it, the more and more complex we discover it is.

I guess that is one of the reasons why we are trying to target specific little regions that we hope will do the job. We don’t know if this will do the job, but we hope it will.

Jamie Rossjohn – I think you also have to apply Occam’s Razor to a lot of these things – the simplest observations, minimal components.

Question – Jacqui, it was enormously disappointing, as we all know, when the gene therapy resulted in inappropriate insertion into the LMO2 gene in SCIDS children. Do you or anyone else know whether there was a particular vector gene combination that resulted in the inappropriate insertion, or is this bad news for gene therapy more generally, in that you found something peculiar in the LMO2 gene?

Jacqui Matthews – It seems that there probably was a combination of the insertion effect and what was being inserted. I think it was an interleukin receptor protein that was being inserted at the same time. One of the reasons it got inserted into the LMO2 region, or that particular region, is that that is a region of the genome that is susceptible to breakpoints. Normally when you get LMO2 leukaemias they are coming from chromosomal translocations, which are coming from the misaligning of the chromosomes.

So I don’t think it is a death knell for gene therapy, but I think there needs to be some more focus on how you direct where things are going to be inserted into genes. And there are quite a few different strategies that are coming across which enable better targeting.

When they were doing this type of experiment, it was going in randomly. It just turns out that this particular cell population where that particular event occurred, and the combination of the two proteins that were produced, meant that you got clonal expansion of one particular cell type – which is how the leukaemia happens. There were many other cells where that didn’t happen. And, obviously, in 80 per cent of the children it wasn’t a problem, as well.

I think gene therapy for a while is still definitely going to be, ‘Is it worth the risk?’ but if you are in one of these cases where you are going to be on continual medication for your entire life – very expensive medication, and it may or may not work – then it is the pros and cons.

Question – With the LMO2 and the ldb, the blue and orange, what is the actual function of those two coming together? You are intent on eventually disrupting that interaction?

Jacqui Matthews – Yes.

Question (continued) – So why does LMO2 interact in that fashion with that protein? What does it actually do?

Jacqui Matthews – One of the things we think is that LMO2 needs to bind, to get into the nucleus and stay there. It is quite a tight binder. The ldb protein, the orange one, actually binds to a range of different related proteins, and there is a lot about the competition between binding of those different range of proteins for ldb that allows lots of different things to happen.

So it is partly to stop other things from binding to ldb. But nobody knows the whole story. The ldb protein is turning out to be a very important member of that story as well.

Question – Jamie, you mentioned looking forward to the synchrotron. Is that going to be most useful in giving us many more solutions to the structural interactions than we are able to get now, or is it going to give us finer pictures of the interactions? What do you see?

Jamie Rossjohn – Both. It will transform the research capabilities of many structural biology labs in Australia and, hopefully, it will also expand the structural biology community, meaning that it will make it more accessible. At the moment I have access to a synchrotron facility in Chicago, maybe two or three times a year. I have to wait months to get that beam time. It is a logistical nightmare to get there at times. All our cargo got held up in Customs for six days, seven days – 100 crystals. And on these sorts of series of events it means you are going to be less competitive on an international stage. So it would present structures of more accuracy and there will be many, many more complexes that you will hear about.

Question (continued) – So will we still be limited by the ability to get the interactions in the right state?

Jamie Rossjohn – What is parallelling the technology drive in terms of crystallography is the techniques of protein production, expression and purification, which James could also talk about. Monash and a number of other groups are developing robotics for protein production, for cloning, crystallisation. And because the environment of a protein crystal is such a concentrated environment, we can actually observe very weak interactions – in the 200, 300, 400 micromolar range.

I think James would want to talk about the robotics of expression.

Chair – Protein crystallography per se is really very well suited to technology which actually allows full roboticisation. As an example of the things you can do by using crystallisation robots, we are currently establishing a protein production facility at Monash – more proteins, crystallise more proteins. But there is a critical final step: if you’ve got thousands of crystals and you don’t have a synchrotron, it is the biggest bottleneck of all. And so I think Jamie and Jacqui would agree that the National Synchrotron facility will really allow us to sit at the same table as a lot of the very, very major structural genomics organisations. Certain structural biology groups in the United States have their own beamline, or maybe 50 per cent of access to a beamline. That is a big deal. So it is going to make a huge difference.

Jacqui Matthews – Yes. I would say that to start with you are going to have far more time and far more beamline time than we actually need, but at the moment we have far, far less. So it means we can grow into the synchrotron and use it effectively. At the moment, there has been no particular point in being able to get lots of things ready, if you can’t have a look at them. Or you may have some tiny crystals that you can’t put on your lab equipment and you have to take it to a synchrotron and see if it is going to be good. And lots of crystals, lots of structures couldn’t be solved on the standard home generators and can only be done on a synchrotron. So it certainly increases capability.

Jamie Rossjohn – But it is also the opportunity of big-picture science. That is why I alluded to the structural genomics consortiums. I am sure there are probably mechanisms by which the Australian structural biology community can band together to tackle something really broad and bold. You can’t do that when you have to access an overseas facility.

Question – It seems to me that a lot of your research is, obviously, very experimentally based and there are lots of different computations that you go through in trialling whether there is an interaction between certain binding groups and chemical functionalities. Could either of you comment on the progress of the modelling of those specific interactions between chemical functionalities? I guess that through experimental approaches, use of the synchrotron, we are going to increase the number of experiments that we do. Is there progress on the modelling side of it, in developing skills to predict, to complement the experimental?

Jamie Rossjohn – I think James will talk about modelling, because that is also his forte. I think you were trying to suggest that structural biology, by either technique, can be very labour-intensive. I would also like to point out that the only reason why you would decide to solve the structure is in answer to the very first question, whether it is biologically significant. What are the questions we are addressing? What are the answers that structural biology can give? If there are big ticks to both of those, then that is a green light for me to tackle a structural biology project.

Chair – I think there are huge advances in the modelling field – automated modelling methods and so on and so forth. But again the same thing applies: if you don’t have the appropriate question, you don’t bother building a model.

Jacqui Matthews – On the modelling side of things, I think that the more we know, the better the models are going to be and the better the modelling approaches and algorithms are going to be. At the moment, some of the stuff that we really like is a combination between experimental method and modelling. If you can feed some experimental data into your modelling system, even if most of the components are homology models or whatever and you are trying to see how they interact, if you have a few scraps of information that don’t enable you to solve the structure per se but enable you to make a good experimental model, then they are probably the best ones to have at hand at the moment. So it is the merging of the two that I think is really advancing at the moment.