It is thrilling to be here and see so many people engaged in epitope-based vaccine development. I think we are really pushing the field forward, and that is what I hope to talk to you about today.
I want also to mention that my next endeavour will be the Institute for Immunology and Informatics, at the University of Rhode Island. I am moving my academic appointment from Brown to URI, where under the leadership of the Dean there I will be developing a centre where we will be exploring the application of immunoinformatics for translational medicine.
(Click on image for a larger version)
I am going to talk briefly about why the time is right for genome-derived, epitope-driven vaccines, and I am going to give you a sampler of the work that we are going to be doing. And then I want to introduce to you some of the other things that we have been doing more recently: I have spent the past 18 months looking at regulatory T cell epitopes. We have a new product which we call Tregitope, which may be useful for the treatment of autoimmune disease. This is a direct derivative of the epitope mapping work that we have been doing at EpiVax.
Probably we have said enough about the old way of making vaccines, which I call the ‘shake and bake’ method: basically, you take the bacteria, shake ’em up and grow ’em up, and then you bake ’em and give them to people.
(Click on image for a larger version)
There are some inherent problems with that, which we have talked about briefly but which I want to mention again.
One is the issue of cross-reactive epitopes or heterologous immunity, the potential to skew the immune response if those epitopes have been seen in the context of a different type of inflammatory response – Th2 versus Th1. And the other thing that I think we haven’t been doing as vaccinologists is actually looking at regulatory T cell epitopes, which also are present in the pathogens that we have been trying to make vaccines against. So we have to be really cognisant of those complications. In some of the experiments that we do, when we go from mouse to human, we see contradictory results, and I would predict that some of that may be due to that problem.
(Click on image for a larger version)
I don’t know if you are familiar with the work of Welsh and company – there are a couple of review articles listed on this slide – but I think they have really been pushing forward the concept of heterologous immunity. I just want to bring that up as an issue.
Clearly there are, at least in mice and potentially also in humans, some examples where you may develop an immune response to a pathogen and it affects how you then respond to the next pathogen.
(Click on image for a larger version)
One example of heterologous or cross-reactive immunity is Lyme disease. We know that there is an epitope that is restricted by HLA-DR4 which is cross-reactive with an epitope from the LFA-1alpha molecule, and that has been associated with the development of Lyme arthritis in DR4 patients and also has been associated with a bad outcome for a vaccine that was withdrawn from the market by Merck. There actually is a new vaccine in development which has that particular epitope deleted.
(Click on image for a larger version)
Other examples of heterologous immunity that need to be considered are quite evident in the field of autoimmune disease. For example, CMV, HSV and varicella-zoster virus have been associated with autoreactivity to GAD65. There are some papers that have been published by one group on influenza A exposure predicting the outcome of HCV infection, causing a much more focused T cell response – less of a broad T cell response – and fulminant hepatitis.
And then what is really emerging and I think is very interesting is the concept that worms may affect immune response. You are probably familiar with the idea that people who have exposure to environmental worms – filaria, for example – may have altered immune responses to autoantigens and probably also vaccines. So we have been exploring that concept in collaboration with Rick Maizels at the University of Edinburgh, and he sent us some proteins from Brugia malayi. I just want to show you a glimpse of the work that we are doing with him. (I think we have only scratched the surface here.)
(Click on image for a larger version)
Here we show you that we have mapped in a filaria organism, Brugia malayi, a protein that has probably one of the highest scores we have ever seen in terms of its EpiMatrix score, very high on our immunogenicity scale, which I will talk about a little bit later, and clusters of epitopes that are very, very high in terms of their immunogenic potential. We believe that this filaria protein, which is also highly conserved – 99 per cent conserved with the human protein – may actually be altering the immune response to filaria, most likely inducing a regulatory T cell response. Those experiments are ongoing.
So I take a reductionist perspective. I think that we need to be looking at vaccines in a different way. (You have heard some talks this morning about why that should be true.) And we should be looking at T cell and B cell epitopes, and looking at the function of those epitopes. Are they epitopes that protect against immune response? We should be delivering them in the correct ‘milieu’, with the right delivery vehicle and perhaps those adjuvants that induce the right interaction with TLR ligands. And I believe that we should be really working on this as the next phase of vaccine design. There is even evidence, as you have heard from the other speakers this morning, that those epitope-based vaccines can protect.
(Click on image for a larger version)
Our small example is a 14-epitope Francisella tularensis vaccine that protects against live challenge with tularemia. This is a study that we performed with a DNA-prime, peptide-boost in liposomes, and we showed 57 per cent protection against a live aerosol challenge in mice. I will come back to that in a bit.
(Click on image for a larger version)
The point is that you don’t need the whole pathogen to protect against infectious disease. We know that also from vaccinia: there is not a 100 per cent overlap between the vaccinia and the variola genomes, and yet we are able to protect. Some of that is B cell related, some of that is T cell related – the point being that it is not the whole genome that is required but perhaps the immunome.
(Click on image for a larger version)
So what are the kinds of vaccines that we have been working on? I want to just give you an overview.
(Click on image for a larger version)
We have, for example, thanks to George Bush, been able to benefit from the Department of Defense largesse, and we have been working on a smallpox vaccine that is epitope-based. Here we have taken four vaccinia genomes and three variola genomes, and we put them in the top of our epitope-prediction tool. We have identified conserved epitopes that are conserved between vaccinia and variola – the hypothesis being that those are the epitopes responsible for protection – and we selected 110 epitopes for further study.
(Click on image for a larger version)
In studies performed with subjects immunised by the DryVax vaccine, we were able to show that 88 per cent of the peptides that we selected were immunogenic. And to a very high degree our prediction was successful. I have put here for you this line that is the 20 spot cut-off for positive T cell response, and you can see that most of the immune responses were way over and we got broad recognition in this population of 35 patients.
(Click on image for a larger version)
What makes these epitopes so immunogenic is a feature that we have been working on over the past couple of years: the concept that peptides that are very immunogenic in outbred human populations – which are not inbred mouse populations – contain a feature that we call an EpiBar. Here you see the EpiMatrix analysis of the flu epitope sequence, and its predicted affinity for eight class II alleles. You can see in the dark bar in the centre a feature that we find in every promiscuous epitope that we look at. Whether it is tetanus toxin, whether it is PADRE, from Alex Sette, this feature is common to most class II promiscuous epitopes. So now that is what we look for.
(Click on image for a larger version)
In fact, we showed that these are also immunogenic in mice – and I will get back in a minute to the reason why that might be true. We have shown, for example, that we can get very broad immune responses in mice with smallpox peptides administered, again as a DNA-prime, peptide-in-liposome-boost.
(Click on image for a larger version)
So here is our general process, outlined in the context of the smallpox vaccine. We take the proteomes, we analyse and find conserved peptides from the different genomes that we are able to download from GenBank, we then identify selected epitopes to confirm. And out of the 48 that we tested, for example for the class II aspect of this work, we found 94 per cent confirmation using human cells.
The overall reduction is pretty dramatic. You can see here that we went from 700,000 potential peptides to 100,000 in a conserved proteome, and down to about 44 epitopes that are positive. It is a rapid reduction, most of which is performed in silico, which enables you to accelerate the process of vaccine design. In this case it only took us nine months from start to finish.
(Click on image for a larger version)
There are other ways of doing this, and I want to mention them briefly. My former co-student who went to the University of Chicago with me, David Koelle, has been working on expressing fragments of the smallpox or vaccinia genome and then testing in subjects, using these expressed sequences.
(Click on image for a larger version)
David performed an exhaustive study in which, basically, he looked at 30,000 different tests in five patients and found 43 fragments that were in a range of different proteins, some of which were early-late or proteins of unknown function. I don’t actually know how long it took him, from start to finish.
(Click on image for a larger version)
There is also a study by Alex Sette, using pretty much the same approach, mainly a class I mapping approach, in mice.
What we did here was to compare the approaches, and you can see that the expressed sequence approach is quite expensive, about $88,000 per epitope identified; Sette, about $20,000; and our approach, including drawing blood from subjects and the collaborations that we established, about $14,000 per epitope.
So I think that, in terms of making vaccines, we can actually reduce the time to identify the epitopes that are critical and thereby accelerate vaccine development.
(Click on image for a larger version)
Another lesson that we learned from this study, when we compared our epitopes with the ones that David Koelle identified, was that there was a little bit of overlap but not extensive overlap. We had 35 subjects, he had five; there were eight epitopes that overlapped. And what I learned from that is that the human immune response is omnivorous.
So how many epitopes would make an effective vaccine? I don’t know. I am getting to the point where I am thinking, in terms of class II, that we need to include at least about 50 epitopes to get broad coverage, and for a class I we can resort to the supertype approach that was mentioned by one of our previous speakers.
The take home message for smallpox is, firstly, that immunoinformatics can be applied efficiently to find the immunome. There are a number of people actually doing this work – I will just mention Alex Sette as one of the most well-known figures.
We can then use the immunome to design vaccines and get 90 per cent coverage with as few as 60 epitopes.
I don’t believe in immunodominance; I apologise. And so I just want to put that right out there. I think that human immune response is omnivorous and multiple epitopes are actually recognised, and it is the collective immune response that actually provides protection.
The heterogeneity is real but you can get coverage if you include enough epitopes.
(Click on image for a larger version)
So we are doing tests of the smallpox vaccine in challenge studies, in collaboration with Mark Buller at St Louis University. Meanwhile, we have also been graciously given funding from the Department of Defense and the NIH (National Institutes of Health) for a vaccine against F. tularensis, another bioterror agent. And in 24 months we took the one genome that was published (the Schu4 genome), mapped class I and class II epitopes, selected 165, confirmed them in humans, cloned them into a vaccine, and actually performed challenge studies.
(Click on image for a larger version)
In those studies again we show broad reactivity of the epitopes. Here you can see that of the class II epitopes, 22 out of 25 were recognised by F. tuli exposed patients from Martha’s Vineyard, and the average response to the individual peptides was very high – way over the 20-spot cut-off – and to the pool, over 1000 gamma-producing cells per million above background.
(Click on image for a larger version)
So I think we have really found something that identifies highly immunogenic peptides. We have now cloned these epitopes into a DNA construct, and we are administering them as a vaccine.
(Click on image for a larger version)
In terms of how you actually approach the design of these vaccines: we have an algorithm called Vaccine-CAD which aligns epitopes and then re-checks them for the presence of pseudo-epitopes at the junctions. In the context of our HIV vaccine program, the tuli vaccine program and the H. pylori vaccine program, we now apply this routinely to design – a string of beads, vaccines that do not have pseudo-epitopes at the junctions.
I should also mention that we blast against human to remove any conservation with human.
(Click on image for a larger version)
The prime-boost tularemia vaccine is shown here. Basically, we have a string of beads DNA construct, and boost with peptide in liposome with CpG, and then challenge by aerosol route. I would also like to emphasise that we give these vaccines IN (intranasally), which is the most effective route.
(Click on image for a larger version)
Here again is that study showing a very protective immune response in 2 x, and 5 x LD50.
(Click on image for a larger version)
Then we followed up in double transgenic mice with a class I and a class II plasmid containing class I and class II epitopes.
(Click on image for a larger version)
We used the same approach as before. This time, because of the differences in our collaborator’s ability to culture the tularemia precisely, the challenge was actually 20 x LD50, so 10 times higher than the first challenge. Again we saw survival of 56 per cent at 10 days, the same as in the previous study. The mice all eventually succumbed. If you can imagine these little mice as what are called ‘war fighters’ at the Department of Defense, this would actually give you the time to get the war fighter into hospital and treat them with an antibiotic. So again it is a pretty effective vaccine – 14 epitopes protecting against a 20 x LD50 challenge.
(Click on image for a larger version)
The tuli project is going quite well. We have also performed binding assays and confirmed these studies in a number of challenge studies – I am just showing you two – and we are moving that forward into development of an actual vaccine. Whether there is going to be some bioterror to protect against remains to be determined, but at least it will be useful for people living on Martha’s Vineyard.
(Click on image for a larger version)
We have also worked on H. pylori. Here is a brief overview. This is again a DNA-prime, peptide-boost. Here we used 50 class II epitopes that were also conserved, in the C57 black strain of mouse, so we were able to predict for those mice as well. And the vaccine was therapeutic post-infection vaccination.
(Click on image for a larger version)
One year later we sacrificed the mice, looked at their immune responses and found broad immune responses, and again I think it is this DNA-prime, peptide-boost that really gives this breadth: 45 of the 50 epitopes were recognised by the mice.
(Click on image for a larger version)
And the really good news was that compared with the standard, which is SS1 H. pylori lysate, we found very good protection: 95 per cent eradication of H. pylori in these mice, which again were immunised intranasally. It turned out that IM (intramuscular) did not work out as well as IN and we only got 56 per cent eradication.
(Click on image for a larger version)
The last sample here is HIV. We remain enthusiastic about the development of a cross-clade HIV vaccine that I have been working on since 1996. I will just summarise the approach.
(Click on image for a larger version)
We have been looking at what we call Achilles’ Heel epitopes. This is a picture showing that if you look for conservation you find very low immunogenicity; if you look for immunogenicity you find very low conservation. We take a balanced approach whereby we analyse the HIV genome of, say, 60,000-plus DNA sequences and we identify the conserved epitopes. Then we run them through EpiMatrix and measure the potential immunogenicity of those epitopes for human class I and class II alleles. And that is the vaccine that we are making and putting into, again, HLA transgenic mouse challenge models.
(Click on image for a larger version)
Shown here is a picture of how we find conserved epitopes. We basically look for epitope strings. This algorithm is called Conservatrix.
(Click on image for a larger version)
We also have an algorithm called EpiAssembler, which takes conserved epitopes and then tries to find where they might overlap in the HIV genome. We call that type of epitope, which is longer, an ‘immunogenic consensus sequence epitope’ – not the same as what Bette Korber is doing, which is a ‘consensus sequence epitope’ and dramatically different in terms of its potential immunogenicity, because we look for immunogenicity.
(Click on image for a larger version)
We also have an algorithm that is called Aggregatrix. Here I am showing you all of our B7 epitopes – this is published – and then their conservation across time. You can see here, in the red, those epitopes that are highly conserved across time, and then this shows you in the individual years how well these epitopes are conserved. We will be focusing on the set of HLA class I B7-restricted epitopes, which look as if they are not only conserved in clades within years but also over time, so they are the Achilles’ Heel of the HIV virus. They are prime candidates for the development of a cross-clade HIV vaccine, and that is what we are going forward with – should we ever receive NIH funding again.
(Click on image for a larger version)
‘GAIA vaccine’ is the name of the vaccine that we are developing. We have identified 286 highly conserved class I and class II epitopes. We have confirmed their immunogenicity in HLA transgenic mice. We have demonstrated that the orientation, based on Vaccine-CAD, is critical to immunogenicity. And, just to get to the bottom of the story: we are now out of funding from NIH, hoping to apply for Gates Foundation funding, but we are still hoping to move forward. We are doing a lot of work on the ground in West Africa, developing a potential clinical trial site with our collaborators there.
(Click on image for a larger version)
So, in terms of the vaccine sampler: I have just shown you a range of vaccine design tools. They are available for collaborations both commercially and academically, and I would encourage you to contact me to use these tools in your work. We have shown you proof of principle, especially in the very extreme model of tularemia challenge.
I want to close with the new discoveries that we have been working on. I think they are very exciting, and they are relevant to vaccine research.
First I want to talk a little bit about autoreactive T cell epitopes, and then about our new discovery, which is what we call Tregitope, an epitope that is very common in the human body and that suppresses immune response to other proteins.
(Click on image for a larger version)
Just a word about in silico mapping: we look for pocket profiles. This is an approach to epitope mapping that was published first by Sturniolo and Hammer in 1999.
(Click on image for a larger version)
The idea is that because you can identify the residues that form the pockets in the HLA alleles, you can actually predict which peptides will bind to them. And even when you don’t have any preliminary data, just by looking for the pocket type you can determine what peptides bind.
Furthermore, these pockets are conserved – not only conserved in the human population, so there are a limited number of HLA pockets that perform the binding function in the context of HLA-peptide binding, but also conserved in other species, so we have been able to extrapolate and develop prediction tools for pigs, based on human pocket profiles.
(Click on image for a larger version)
In our class II work we actually look at a limited number of pockets. We have selected eight class II alleles that have representative pockets for the human population, and they cover 95 per cent of human populations worldwide.
(Click on image for a larger version)
This is a picture of ClustiMer, showing what we now call an EpiBar. If you look at those eight class II alleles in a protein sequence, you see there are these regions also known as hot spots – that is what Vladimir Brusic calls them, and we have been calling them clustered regions for a long time. These are the EpiBars that I was talking about.
(Click on image for a larger version)
Overall the prediction accuracy of the EpiMatrix tool, using the eight class II alleles, is 93 per cent.
(Click on image for a larger version)
We also have very high predictive accuracy for A2, and you see here an analysis that we did of a paper published by Vladimir Brusic et al. They looked at a number of different algorithms, and looked for class I A2-restricted epitopes.
(Click on image for a larger version)
Just to summarise: we predicted 100 per cent of the epitopes that they then mentioned – they would use the artificial neural network and hidden Markov model to predict together, in order to get the same level of accuracy.
(Click on image for a larger version)
To go back to the idea of clusters: we have been looking for these.
(Click on image for a larger version)
And we also think that the number of epitopes in a protein is critical.
(Click on image for a larger version)
So we now look for the sum of epitopes in a protein as an indicator of its potential immunogenicity.
(Click on image for a larger version)
We have developed an immunogenicity scale which allows us, if we have the whole genome, to rank which proteins are going to be the most immunogenic for humans.
(Click on image for a larger version)
We have actually applied this to autologous proteins, and have made some surprising discoveries. For example, there are a number of proteins such as human choriogonadotrophin, transferrin, amylase, albumin, that almost seem to have epitopes deleted as if to reduce their immunogenicity, and a number of proteins that have higher epitope content, such as interleukin-2 and interferon-beta, which are actually known to be immunogenic when administered in the context of protein therapy.
(Click on image for a larger version)
We have applied this to monoclonal antibodies. We actually now have a very active business predicting for companies whether their monoclonal antibodies are going to be immunogenic in the clinic. This work is published, and we basically have shown that there is an association between the number of T cell epitopes in a monoclonal antibody and its potential immunogenicity – with an important caveat. That is, in immunoglobulin we find a very strong signal that is a strong T cell epitope that is not an effector epitope; it is a regulatory T cell epitope. And that is our Tregitope that I will speak about in just a minute.
(Click on image for a larger version)
It is important to note that HLA determines immune response, so when you look for the immunogenicity of a protein it obviously depends on the HLA background of the individual. Generally we can predict the immunogenicity of a protein, but an individual will also respond in the context of their HLA allele.
(Click on image for a larger version)
We have also developed an algorithm for that, called iTEM (the individualised T cell epitope measure), which basically takes the sum of the epitopes in a given protein and predicts whether an individual will have an immune response. That work is also published, in collaboration with Amgen.
(Click on image for a larger version)
I like this idea, because basically you can predict which patients in a vaccine trial will develop an immune response. So I think that going forward in vaccine trials we should be thinking about the HLA type of the patients who are receiving our vaccines, and perhaps we might be able to explain the efficacy of the vaccine in the context of their HLA.
(Click on image for a larger version)
To go back to the idea of the relative immunogenicity of autologous proteins: we looked at one protein that was a real outlier in terms of the human protein immunome, C3 (Complement factor 3). It turns out – I wasn’t aware of this, but Paul Knopf, who works with me, pointed it out – that C3 plays a very important role in the immunogenicity of carbohydrate antigens.
(Click on image for a larger version)
So we looked more closely at C3 and found that actually it was a C3d piece of C3 that had the most epitopes, as shown on this slide.
And then again, breaking it down by the region of breakdown of the C3 molecule, we find that the piece that actually binds to antigen and is co-internalised in the B cell during the process of antigen presentation is C3d – here, as you can see by the pink colour, packed full of epitopes.
(Click on image for a larger version)
So Paul Knopf said, ‘Aha! That may explain why C3d is such a good adjuvant’ – as was described by Fearon et al. many years ago – ‘and why it is being used as an adjuvant for vaccines.’ We went and looked at the human T cells, showed that they secrete gamma-interferon in response to stimulation with the C3d epitope and also showed, using tetramers, that the response was actually specific for the C3d peptide.
(Click on image for a larger version)
So we are very happy to say that we are countering dogma: there is autoreactive T cell help. Depicted here is the model showing the carbohydrate antigen being co-internalised with C3d, and that C3d epitope is presented to autoreactive T cells that then generate the second signal, the T cell help that is required for the evolution of antibody response. I hope that those of you who are working on carbohydrate vaccines will take this to heart as a potential mechanism. We have published this. It is in the public domain; it is not patented.
(Click on image for a larger version)
The last thing I want to mention – and I will be brief – is Tregitopes. This is very exciting.
(Click on image for a larger version)
The idea is that clearly we know about effector T helper cells, we know about CTLs, and now as vaccine developers we need to be thinking about regulatory T cells as well. These guys make IL-10, TGFbeta and TNFalpha, and so when we measure immune response to T cell epitopes we might want to consider measuring other cytokines, because they probably do play a role.
(Click on image for a larger version)
Our perspective on T cell epitopes is that they can be either regulatory T cell epitopes or effector T cell epitopes – it totally depends on the milieu. So you can overcome tolerance by administering an epitope that is an autologous epitope in the context of a ‘danger’ signal, as has been described by Paul Matzinger.
(Click on image for a larger version)
What makes epitopes so immunogenic in terms of vaccines is the same thing that makes them really good regulatory T cell epitopes: they have EpiBars. Shown here is an EpiBar in immunoglobulin G.
(Click on image for a larger version)
I had said that immunoglobulin G contains such epitopes. This is one in the Fc domain, which may explain why Fc fusions – which are commonly used in protein therapeutics – are not very immunogenic. It also probably explains why IVIG (intravenous immunoglobulin) can be used in autoimmune disease, because you induce, as has been shown by Ephrem et al. in Blood, that you actually induce T-regs by administering IVIG, possibly due to the presence of the Tregitopes in immunoglobulin.
(Click on image for a larger version)
They not only exist in the Fc region; they also exist in the variable region. Just adjacent to the variable sequences in the framework, you will find Tregitopes that probably act to suppress human anti-human immune responses.
(Click on image for a larger version)
We attended a Breckenridge meeting on Tregitopes and had a kind of ‘aha’ about the strong signals that we found in IgG, and have been working almost exclusively on that for the past 18 months.
(Click on image for a larger version)
And now we are sure that Tregitopes contain these EpiBars; they are in autologous proteins, not just IgG. They are highly conserved. In fact, the same Tregitope will work in mice, with a few amino acid changes.
(Click on image for a larger version)
In the last slide on that, I want to show you that we can suppress immune response in vivo, using mouse Tregitopes. So here we have a protein therapeutic that induces a very strong T cell response – this is IL-4. Shown in blue is the antibody response to the same protein. We are able to reduce that immune response with our Tregitope. And the sham control, obviously, did not have any immunogenicity.
(Click on image for a larger version)
The mechanism of action of the Tregitopes in IgG is, we believe, as follows. In IgG a molecule is taken up by an antigen-presenting cell, presumably a B cell, and at the surface you will get presentation of effector epitopes from the variable region and regulatory T cell epitopes from the Fc region in the framework, and that suppresses the immune response to the antibody itself. That is why humanised antibodies are so much more effective than mouse.
I have taken you on a whirlwind tour of our work over the past couple of years.
We have been working on genome-derived, epitope-based vaccines and, as you have heard from other speakers, they seem to be a very effective approach to making new vaccines. I think that they may be safer and more targeted, and that we will be able to avoid heterologous immunity.
There is clearly evidence in mice – and now in humans, thanks to my co-speakers – that genome-derived vaccines can work.
I also think we need to be looking at autoreactive T cell epitopes, because there is clearly a role. At least in the worm world there seems to be a cross-reactivity, and we believe that those epitopes I showed you on Brugia malayi, which are highly conserved in the human autologous protein, probably induce a regulatory T cell response to the worms that they are able to take advantage of, so that they can live within the human body.
We have only begun to scratch the surface. Now that epitope mapping tools are available, I think we can really get going.
(Click on image for a larger version)
That is why we have founded this new Institute for Immunology and Informatics at URI, also called ‘I-cubed’. I hope to be able to put up a website for neglected tropical diseases like malaria, so that people can map epitopes using our tools for free – hopefully, with Gates funding.
(Click on image for a larger version)
I want to mention my collaborators at EpiVax: Bo Wu, Julie McMurry, Lennie Moise, Joe DesRosiers and Bill Martin. Bill has been my partner for the last 10 years in the company.
(Click on image for a larger version)
I would be remiss if I did not mention my collaborators, who are shown on this slide. I don’t yet have pictures for David Scott, Jeff Bluestone, Kevan Herold and Bob Smith.
(Click on image for a larger version)
Also thanks to my peer reviewers, and to NIH for current vaccine funding.
Discussion
Peter Colman (Chair): Thank you, Annie, for a very provocative talk. It is truly wonderful to see, some 20 years after Don Wiley first saw that peptide lying in the groove of an HLA molecule, just to what ends that information can now be used. The striking correlation that the algorithm is showing with experiment is a great testimony to the breakthrough that was made at that time. There is plenty here, I am sure, for the hard-core immunologists to tear into!
Question: Could I just ask: what is the predictability beyond A2 for your programs? I know A2 works for everybody. We can use that in any predictive program. But beyond HLA-A2, what is the success rate of your programs?
Anne De Groot: Our A2 prediction is better than other algorithms, and I just want to point that out. But we also have 30 class I alleles and we focus on the six for our vaccine work. The class I prediction for HIV is published, so you can take a look at that. It is in Vaccine 2004, I think.
Question (cont.): I am asking in terms of readouts in functional assays. What is [inaudible]?
Anne De Groot: In the B7 paper that was just published in Vaccine, I can’t remember exactly but I think 80 per cent of the epitopes were confirmed in HIV-infected patients.
Question: A great talk. I was wondering: when you were talking about finding these clusters of immunogenic epitopes, is it T cell epitopes and B cell epitopes? Are they all found within the same cluster? And what about T-reg epitopes? Are they in there as well, or do you find a separation?
Anne De Groot: Well, as I said, I don’t think that the algorithm discriminates between effector and regulatory, so we are using the same algorithm to predict T-reg epitopes. The only difference is that they are an autologous protein so presumably in the context of immune development there is a suppression of immune response. So you get regulatory T cells; there is something to do with the affinity of the T cell receptor for the epitope and so on. I am an infectious disease expert learning about tolerance, so I have really been enjoying over the past year learning more about what determines what becomes a regulatory T cell epitope.
Actually, I do think it is the milieu, and you can overcome that just by administering with an adjuvant.
As to your question in terms of T cell and B cell epitopes, I have bugged Bill Martin for a long time about that. Just looking in the literature, he says that about 20 per cent of the time there is co-localisation. I think it is probably higher, and we need to go back to the databases and cull out how they identified the B cell epitopes, how they identified the T cell epitopes, and what the overlap is. Probably the linear B cell epitopes may be more likely to be co-located than we originally thought.
In our work with commercial companies we have shown that that is true, but we just haven’t done a really good analysis of all the literature that is out there now.
Question: To come back to the prediction system: is your program capable of predicting sequences longer than nine amino acids? We are now finding, particularly for B7 and B35 alleles, that there are sequences going up to 12 amino acids long. Could that program accommodate that sort of prediction?
Anne De Groot: We look for nine and 10 right now.
Question (cont.): I see. Would it be possible that the program would do that?
Anne De Groot: Yes, we could certainly do that.
Visit these websites
The Sir Mark Oliphant Conference Series is supported by the Australian Government under the International Science Linkages Program.
© 2008 Sir Mark Oliphant Conferences.
Website design acidgreen
