We started this conference with a fantastic presentation by Ian Frazer and the great success of the papillomavirus vaccine, but I am going to take you right back to some of the very fundamentals that might underlie how we think about making vaccines. This is how dendritic cells and T cells come together and talk to each other, and how T cells are told to become protective T cells.
(Click on image for a larger version)
We have heard a lot in this conference about the different sorts of pathogen infections and the success against the poliovirus, and we have also seen that for many of the viral infections that we are confronted with it is actually very challenging to develop effective vaccines that will act not only in the developed world but in the poorer countries.
So the question becomes: what is underlying our failure to generate these very effective vaccines?
(Click on image for a larger version)
As has been pointed out by a number of speakers, often we develop vaccines by simply taking bits of the pathogen, grinding them up and injecting them into animals. We don’t really seem to need to know the basis of positive responses for these vaccines, as represented by the ‘Black Box’ in this slide.
In the Black Box are the things that we study as core in our laboratory. My major interest is in influenza viruses and herpesviruses, and after discovering epitopes to some of these viruses when I was doing my postdoctoral training I have become more and more interested in how dendritic cells interface with the T cells.
A number of companies, such as CSL, who are studying ISCOMATRIX R, have become somewhat interested in some of the basic underlying science that might be driving the sorts of responses we see in vaccines, because regulatory bodies often ask questions that we are able to answer with that information, and also it might uncover different ways in which vaccines might be utilised, targeting innate aspects of the immune system rather than just those that we think should be the outcome.
(Click on image for a larger version)
I don’t need to explain to this audience the lifecycle of a T cell, but really this is the process that we have been interested in: how a naïve T cell becomes a memory T cell. The memory T cells are those things that we are trying to activate and stimulate during a vaccine response.
We all think we have got a fairly good hold on how T cells are generated, that they interact with an antigen-presenting cell of some sort, they expand and make effector cells, and those effector cells are very good at going around killing antigen-presenting cells – or any cell that expresses an antigen.
(Click on image for a larger version)
What has become clear over the last few years, though, is that many of the things that we thought we understood about how naïve and memory T cells are generated are really not so clear at all. And so I am going to address a small number of the rules that we have uncovered over the last few years which change the way in which we think about how we might amplify specific T cells for pathogen infections.
The first thing I want to point out is that originally we thought that any antigen-presenting cell was able to prime a T cell response. We needed a professional antigen-presenting cell to prime a naïve T cell response, but possibly macrophages could do it as well. What has become clear is that there is a very specialised antigen-presenting cell, the dendritic cell, which is necessary to drive those responses.
In addition, promiscuous memory cells, because they have already seen antigen before, were thought to be activated off those cells expressing class I and could load the peptides, but naïve cells needed professional antigen-presenting cells. But, interestingly, when we went into animal models to explore these particular constraints on the different T cell types, we found that in fact both naïve and memory T cells needed to see an antigen-presenting cell, specifically a dendritic cell, in the lymph node draining the mucosal tissues.
(Click on image for a larger version)
This opens up an opportunity for us in thinking about vaccines, in that we now have identified the principal target for amplifying T cells.
The second major and very distinct difference in the way that we have thought about antigen-presenting cells relates to the idea that T cells are killers: they go around trying to find targets – things that express class I or class II antigens – and they want to kill them. They want to get rid of them from the body. That is how T cells always work, simply bombarding antigen-expressing cells.
This turned out not to really be the case. In the case of naïve T cells, as I will show you, they don’t go around killing all the antigen-presenting cells that are driving that initial response, whereas memory T cells sometimes do.
(Click on image for a larger version)
That gives us a temporal separation in the ability to activate different sorts of T cells, and indicates to us that prolonged antigen presentation might really have a role in driving the sorts of vaccine amplifications that we might be able to achieve. So the timing of antigen presentation is really critically important in determining the overall magnitude of the responses that we can achieve by vaccination.
The third story, which I will go into in a bit more detail in a little while, is that we used to think that memory T cells responding to an infection always excluded naïve T cells from the response. In some cases this could be quite good, producing a very fast and robust T cell response; however, it may be that those T cells, as has been indicated before, are not protective. While we are very good at discovering epitopes, we are not so good at determining which of those are really clinically protective. And, as was pointed out in the previous talk, in the gamma-herpesvirus that we work on we have identified 11 epitopes in lytic genes but none of these are actually protective in the long term.
The one that is protective is encoded in a gene called ORF73, and this particular gene actually causes proteolysis of the terminal part of the actual molecule, and indeed destroys the epitope that is most likely to be fully protective in that infection. Using epitope discovery it has not been possible to find that particular epitope without knocking out one part of the gene to prevent the proteolysis of that gene.
So what we have found is that there are mechanisms that allow both naïve and memory T cells to be amplified during immune responses. We are not exactly sure what situations this is operating in – at least in some mucosal systems it is working, but in the model systems that we have looked at, it doesn’t seem to be working in the spleen, or for a systemic infection.
(Click on image for a larger version)
But this gives us an opportunity to do strategic amplification of different sorts of T cells, if we can get a handle on the molecular mechanisms that are guiding this differential amplification.
(Click on image for a larger version)
As I mentioned, dendritic cells interface with and marshal all the troops in the immune response, and these dendritic cells are required for both the naïve and the memory T cell amplification.
(Click on image for a larger version)
The systems that I work on are systems where we give intranasal infections, infecting the mucosa. That requires the antigen-presenting cells that are found in the peripheral tissues – and there are different sorts of antigen-presenting cells – to migrate to the draining lymph node, where they marshal all the T cells and B cells and really get the immune response to go.
What is interesting about this system is that if you look, in a model influenza infection, at a primary and a recall response, you find that quite strikingly there is very little difference in the ability of the secondary response to actually clear the virus – it does not do so any more quickly than the primary response. There is a one-day difference. So in the primary response you clear the infection by 10 days; in a secondary response you clear it by nine days. Quite simply, there is a rate-limiting step in the migration of these dendritic cells from the peripheral tissues to the lymph node.
That opens up, firstly, avenues for how we might increase the migration of those dendritic cells that are crucial for activating the T cells. This might have been touched on a little bit in the earlier talk about tuberculosis, where monocyte dendritic cells can be recruited into the lung to amplify the immune system. (Warwick Britton spoke about that.)
A second feature that we have found might be plausible is to use different sorts of antigens encoded by the vectors. Our persistent infection is typically a relatively slow-growing infection, as we understand it, yet if we have encoded a soluble antigen into that recombinant virus we have actually accelerated the rate at which we can start to amplify the T cells specific to that infection.
This suggests that there might be different ways of getting antigen to the draining lymph node, to these lymph node resident dendritic cells that are responsible and, actually, critical for activating the naïve T cells.
(Click on image for a larger version)
So the kinetics of antigen presentation became very important to us in analysing what is really going on in these different sorts of responses. I have just listed, at the right-hand side here, the way in which we go about this. It is quite simply to use a reporter T cell hybridoma. This hybridoma is crucial, because it allows us to really enumerate the actual number of antigen-presenting cells that are presenting antigen in a response. This gives us a level of sensitivity in understanding the kinetics of antigen presentation that we are unable to get if we use an ELISA-type approach, and we can really track the kinetics of antigen presentation and then look at whether particular molecules are regulating that antigen presentation.
So we can get a good handle on when antigens are actually expressed, and when antigen-presenting cells are actually there so they can amplify T cells, and it gives us a very good idea of the duration of the antigen-presenting cells and whether that could affect the function of the T cells that are elicited. I think the function or the quality of the T cells that we generate in a protective immune response is essential if we are actually going to generate long-term protection.
(Click on image for a larger version)
Monitoring this antigen presentation has highlighted to us a couple of features of primary and secondary immune responses.
The first one is that antigen presentation during a primary immune response is really prolonged. This opens a doorway to amplify quite a number of the naïve T cell repertoire so that we create in our memory compartment a very large library of potential T cell clones that could respond to secondary infections. So, in effect, we have got a system that allows us to preserve diversity. In this primary immune response, the killer T cells are not in a functional state in those lymph nodes to destroy the antigen-presenting cells, and so we preserve that long-term antigen presentation until the virus in the mucosal site is actually cleared.
What is also interesting about this activation step is that we have rapid modulation of the homing receptors, so downregulation of the surface receptors such as CD62; and almost as quickly as they are downmodulated they are actually upmodulated, allowing them to re-enter the lymph nodes for secondary activation. This allows them to be readily recalled into the immune response when we encounter the antigen again.
(Click on image for a larger version)
That is quite different, and quite unexpectedly we found that in memory T cell responses we have very, very short periods of antigen presentation. In the response to influenza, which is our model infection, primary antigen presentation occurs for as long as 15 days, albeit at a very low level at that stage, whereas in the memory response, antigen presentation is over by three days after infection. This really truncates the period in which we can get T cells into those antigen-presenting cells and activate those T cells.
In this sort of response, the antigen-presenting cells are killed by these rapidly upregulated T cells that express lots of cytolytic molecules and cytokines, and those cells are in direct contact with the antigen-presenting cells while they are expanding.
The second thing is that the T cells that are generated don’t modulate their homing receptors anywhere near as quickly as occurs in a primary immune response. So once again they will downregulate their CD62 ligand expression, and they will retain a downmodulated level of CD62 ligand expression for many months, up to six months. This makes you wonder whether those cells can now get back into the lymph node where they need to be primed, if we are to give a boost to a vaccine preparation.
It raises questions about how frequently we should boost those cells. Do we truncate the number of cells that we can actually amplify in those repeated boostings if we have the period between boostings very short?
A major question that arises out of this sort of information is how do we maintain the diversity of the memory response in the face of what is often quite a dominant and robust response in these secondary immune infections?
(Click on image for a larger version)
That got us wondering how different sorts of dendritic cells actually impact on the types of T cells that are expanded in these responses. We have heard a little bit about dendritic cells at this meeting, but in this slide I am showing you what is really the full library of dendritic cells that exist, as they are defined in mice.
One does have to be a little bit cautious in translating the types of information that we get from dendritic cells in mouse models to human models, because one of the great chasms in this field is understanding what are the dendritic cells in humans that are equivalent to the ones we are looking at in mice. But this system has really allowed us, in a very specific way, to dissect out the functions of different dendritic cells and work out some of the rules that might be guiding the amplification of naïve and memory T cells.
These are fairly spectacular little cells that, when they were first discovered 35 years ago, were thought to be just exciting accessory cells. It was difficult for people to get much of a handle on what these cells did, although it was clear that they were important in driving immune responses – but they represent such a small fraction of lymphoid tissue, fewer than 1 per cent. In addition to that, it is exquisitely difficult to culture them in vivo, and to replicate the sorts of antigen-presenting cells that we are able to isolate from either mice or humans in the in vivo situation.
But in the last several years there has really been a coming together of the different areas of dendritic cell research, in understanding in a real way the in vivo roles of many of these different cells types.
(Click on image for a larger version)
So what we set out to do was to try and understand: amongst that library of dendritic cells, is it all the dendritic cells that are really driving the immune response, or is there a certain population that is specialised to deal with pathogen infection?
In model systems we found, if we looked at a variety of different pathogens, that there was in mice one dendritic cell population that really was doing most of the work, and this was the CD8 + dendritic cell.
What was also interesting was that in an intranasal infection mode there was an additional cell type, the CD11b -CD8 - dendritic cells. These are our trafficking dendritic cells.
(Click on image for a larger version)
Shown here is the model system that we used for the in vitro part. Basically, most people use the spleen when they are looking at dendritic cells, but because we are using a mucosal infection we are actually looking at the mediastinal lymph node, the little guy shown in the centre of the left-hand panel here. We have had to adapt all our assay systems to deal with very small numbers of dendritic cells that we have loaded with pathogens by natural infection in vivo, and to be able to take those dendritic cells out in a way that they would actually be able to work in the in vitro setting without destroying all their function.
So we isolate those cells and phenotype them, sort them into the different types of cells and culture them with reporter T cells that are CFSE-labelled. This is an exquisitely sensitive system where we can use as few as 500 dendritic cells and probe them with 50,000 transgenic T cells, or even endogenously generated memory T cells, and look for responses.
The initial finding that we had some years ago, if we looked at CD8 + dendritic cells and our lung trafficking dendritic cells and compared those with other dendritic cells in the lymph node, was that two populations of dendritic cells were really driving the immune response.
So we thought, ‘Well, memory cells are bigger and better and faster, and they do everything much better than naïve T cells, so surely they are less restricted than naïve T cells in responding to different dendritic cell subsets.’ Quite intriguingly, though, we found exactly the opposite, that these memory T cells simply responded efficiently to CD8 dendritic cells but not to other dendritic cell subsets. This gave us a clue that dendritic cells were actually guiding different sorts of T cells to be generated differentially.
(Click on image for a larger version)
So what is the pattern of presentation amongst these different sorts of dendritic cells? We did this experiment because we wanted to prove that CD8 dendritic cells that are responsible for cross-presentation were actually temporally different, in the timing at which they actually presented antigen. And we would be able to prove that trafficking dendritic cells handed over their antigen to the lymph node resident CD8 dendritic cells, and that would be proof that cross-presentation was essential.
When we did this mapping of the kinetics managing presentation, we actually found the exact opposite of what we expected to find: these double-negative lymph node dendritic cells were presenting antigen for a very long period of time.
But when we combined that with the data that we saw in the previous slide that memory T cells really struggle to be activated by non-CD8 dendritic cells, we thought we could test whether in vivo memory cells were constrained in their ability to be activated by dendritic cells.
(Click on image for a larger version)
We did that by setting up a transgenic system in which we infected our mice with influenza virus, and at different time points we transferred in either naïve or memory T cells and looked for their activation.
If we transfer-in our memory T cells or naïve T cells late in infection – around day 10 – when we think that most of the antigen-presenting capacity of CD8 dendritic cells is lost, we find that only the naïve T cells are actually responding to the dendritic cells that are still presenting antigen in that system. You could argue that our memory T cells are simply not very good and they are not responding, but if we do this same experiment and transfer the cells in at day 3 after infection we find that both naïve and memory T cells respond quite efficiently.
(Click on image for a larger version)
We spent quite a lot of time trying to understand the molecular regulation of this process, because clearly if we could differentially amplify naïve and memory T cells this would be a really good way to guide the sorts of responses that we might want during a vaccine approach.
We identified a whole list of candidate genes or molecules expressed by the trafficking dendritic cells that we thought could be regulating this process. We used the approach of our in vitro assay, and then used monoclonal antibodies that could target the molecules that we had found to try and inhibit the antigen presentation that was going on.
Much to our regret, having screened a large number of different molecules that came up on our arrays and validating that they were completely correct, they had no inhibitory effect on our T cells.
But one molecule which Ralph Steinman had uncovered was differentially expressed on dendritic cells was CD70. The ligand for CD70 is CD27, which is expressed on T cells. And so we did an analysis of in vitro proliferation, looking at whether CD70 could block the proliferation of T cells, whether they be naïve or memory T cells, as proof of principle that differential antigen presentation and regulation through specific molecules on dendritic cells could really be used to regulate the amplification of these different sorts of T cells.
By including CD70 in our sorted dendritic cells that were then exposed to our T cells, we found that there was a considerable amount of inhibition stimulated by blocking the CD70–CD27 ligand interaction. However, if we looked at our lung-derived dendritic cells, when we included CD70 antibody we found that there was virtually no blocking going on at all. So the trafficking dendritic cells are using different mechanisms, potentially, to amplify the T cells that they see during the immune response.
We now realise that dendritic cell subsets, although Ken Shortman has painstakingly over the last 20 years phenotyped dendritic cells into these multiple subsets, really relate to functional differences that occur in these dendritic cells.
The key question for us was: can this really happen in vivo? In the face of preformed memory, is it possible to manipulate the system so that you can amplify naïve T cells, when you have got this robust memory response going on? Clearly, sometimes the memory cells that we recruit into responses are not protective, and are very distracting to the immune response. You want the capacity to actually draw new protective cells into your immune response.
(Click on image for a larger version)
The way we addressed that was to use a series of competitive assays. In these assays we set up an in vivo system where we have, in the green here, our CD8 + dendritic cells, and in the red our trafficking dendritic cells. If we introduced large numbers of naïve or memory T cells, we could block the antigenic sites of particular dendritic cells.
In this situation I have shown the blocking by naïve T cells, and then a small number of responder T cells that are congenically labelled.
(Click on image for a larger version)
If we look at these titrations we find that naïve T cells can block naïve T cells, because both T cells are activated off both the CD8 + and the double-negative dendritic cells; and memory T cells can block naïve T cells if they are in sufficient numbers.
(Click on image for a larger version)
We can now look at whether the preformed memory T cells can completely block the naïve T cells. We had hypothesised that the CD8 + dendritic cells, shown in the green, would amplify the memory T cells, but it is possible that these naïve T cells could sneak in and be amplified off the trafficking dendritic cells.
(Click on image for a larger version)
And when we performed these assays we found that if we use responding naïve T cells but we have lots of memory T cells already in the system, then it is still possible to elicit reasonably good T cell responses.
(Click on image for a larger version)
Why haven’t we seen this before? I think one reason is that we haven’t had a very good handle on what is going on with different dendritic cell subsets, but one of the other reasons, at least in our system, is that if we look at the spleen we find that only CD8 dendritic cells are amplifying T cells in the pathogen infections that we have seen.
So in this situation a preformed population of memory T cells has a very good capacity to actually block most of the antigenic niches that might exist in that particular organ. But where we have got multiple dendritic cells interacting in mucosal tissues, it opens the possibility for those dendritic cell subsets to be performing different functions.
(Click on image for a larger version)
When we looked at an intravenous infection, we found that the naïve T cells were quite easily out-competed by the memory T cells in this system.
(Click on image for a larger version)
Overall we found that there are really quite different rules from what we thought about some years ago, guiding the amplification of naïve and memory T cells.
In primary responses we have long-lived antigen presentation, which gives us the potential to amplify lots of different types of T cells and gives us the opportunity to think about what we try to do with adjuvants and storage adjuvants in promoting that long-term antigen presentation.
However, in memory T cell responses we have slightly different rules that might be guiding the amplification of memory T cells and of naïve T cells, so that we can really maintain a repertoire of T cells that can respond to the sorts of antigens or pathogens that we might encounter in everyday life, not just those that are life-threatening.
(Click on image for a larger version)
So where does it go from here? It is clear that there is a lot of fundamental work to be done in understanding what is really driving these processes that we have observed over the last few years. What are the specific mechanisms, and what are those molecules that are really regulating the amplification of those naïve and memory T cells?
As I mentioned earlier, what are the dendritic cell equivalents in humans, and can we develop systems that allow us to probe really carefully the dendritic cells of humans, possibly in mouse models, so that we get much better translation of the sorts of things we try to do with dendritic cell vaccines? They have not been particularly successful to date.
Ultimately, can we target vaccine strategies to use rational design to actually amplify T cells in a way that gives us protective immunity rather than just lots of T cells that recognise epitopes that are encoded by the particular pathogens?
(Click on image for a larger version)
This work was made possible by great collaborations that we have, both at the Walter and Eliza Hall Institute (WEHI) and outside of WEHI, in doing the dendritic cell studies and the mouse studies, but by also the work with Miles Davenport in modelling all this data to see how realistic it is in uncovering the mechanisms responsible for how T cells are actually amplified during the immune response.
Graham Mitchell (Chair): Thank you, Gabrielle. Beautiful work and beautifully presented. I am going to have to change some of my paradigms. I am still back on the left-hand side, ‘Past’, of the slide you labelled ‘Paradigm shifts’. There have been certain rules of the game with respect to memory T cells and naïve T cells, and you are giving that a bit of a shake so I am going to have to follow through it a bit more closely. Things are changing, that’s for sure.
Question: Correct me if I am wrong, but wasn’t it published last year in Science that when you have got an ongoing response to memory or whatever in a lymph node, it blocks out any entry of naïve T cells by chemokine blocking? How do you account for that in your system?
Gabrielle Belz: That particular system is not persistent infection. We don’t see blocking in our system. We are not seeing massive amplification of naïve T cells; what we are trying to say is that naïve T cells can get back into the system. But persistent infections are really quite different from acutely cleared infections, and are far more complicated. I have not spoken at all about the persistent infection that we work on, but such infections use some very exquisite mechanisms to actually get around the immune system and prevent the immune system from finding the little niches in which they maintain their reservoir.
Graham Mitchell: We are going to have to re-look at original antigenic sin, as well, because that’s a flu phenomenon in a mucosal site.
Gabrielle Belz: But they were all i.v. infections.
Graham Mitchell (cont.): Were they? That’s interesting. So you can account for that phenomenon as well. This is even more robust! Good.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
