Vaccine and Immunotherapy Technologies

9-11 April 2008, Canberra

Professor Stephen Kent

Head, HIV Vaccine laboratory, Department of Microbiology and Immunology, University of Melbourne

Stephen Kent trained in medicine and infectious diseases at Monash University in Melbourne and the University of Washington in the USA. He has a broad interest in protection and therapy of HIV by vaccines. His research group at the University of Melbourne has studied various live attenuated vaccines, recombinant protein vaccines, virus-like-particle vaccines, DNA vaccines, poxvirus vector vaccines and peptide based vaccines in primate models and some of these approaches have proceeded to human trials. Stephen is also an infectious diseases physician caring for people with HIV infection at the Alfred Hospital and Melbourne Sexual Health Centre.

 

Immunotherapy for HIV

(Click on image for a larger version)

We know that drug therapy is effective for HIV, but there are a lot of side effects, it is daily and lifelong, it is expensive and it leads to drug resistance. I see people with HIV and they all want to get off the drugs – even though if you were living in parts of Africa you would want to get on the drugs. So there is clearly a need for alternative therapies.

(Click on image for a larger version)

It would be no surprise to this audience that immune control of HIV is actually quite strong. There is a 2 log reduction in virus after acute infection. We see people that we used to call ‘long-term non-progressors’ and that are now called ‘Elite Controllers’. (I’m sure there will be another spin on this in a couple of years.) And most of these scenarios are linked to strong T cell immunity. We don’t really see subjects that have tremendously strong and broad neutralising antibodies in the same scenario. But the question is: can this be mimicked by a vaccine or immunotherapy?

(Click on image for a larger version)

T cells recognise short linear peptides presented to CD8 and CD4 T cells. And you all know the incredible diversity of MHC molecules. Most of the time we don’t have the luxury of typing all our patients, so we usually don’t know which peptides are presented by different people.

(Click on image for a larger version)

We have got a conference here on therapeutic vaccination, and everyone has to stand up and say, ‘Well, there really aren’t any great examples. Why would they work? There is plenty of antigen around.’ As Eugene Maraskovsky most eloquently suggested this morning, the presentation of antigen is pretty suboptimal. There is a lot of virus infection going on, there are a lot of defects in antigen-presenting cells, particularly dendritic cells – most of which I don’t really think we have a good handle on in vivo.

I think the totality of the evidence suggests that delivering vaccine antigens by cultured dendritic cells is really showing quite a lot of promise in several diseases, including HIV, where it has been tried, and several cancers as well. But the problem with culturing dendritic cells, as Eric Gowans (painfully, I thought) described yesterday, is that it is really quite impractical. Other than as a proof of principle, it is hard to imagine that many people are going to benefit from that sort of therapy.

(Click on image for a larger version)

As a couple of the other speakers yesterday pointed out, the vaccines have quite a lot of difficulty in safely inducing T cell immunity, although I think speaker after speaker at this conference has shown ways of getting around this. We have to say that most vaccines work by inducing antibodies, but this has so far been unhelpful in the setting of HIV.

What used to be called the ‘new generation’ vaccines, which I guess are now the old kids on the block, are DNA vaccines, pox virus vectors, adenovirus vectors. Some of these are confounded by weaker immune responses or anti-vector immunity; we heard from Wayne Koff yesterday about the disaster of the STEP trial. A big problem with a lot of these vectors is that immunity is often narrow, because it is not easy to safely include many parts of the virus.

I am jumping over a couple of serendipitous steps here, but we took the approach of asking what would happen if you coated fresh cells with virus peptides as a vaccine. Would it act as a vaccine? What peptides would you choose? What cells would you choose?

(Click on image for a larger version)

The peptide antigens that we chose (serendipitously, really) were a set of overlapping peptides. This was really derived from using overlapping peptides in the setting of T cell-based assays in the lab, and Eugene Maraskovsky described this morning how, the more peptides you throw into your mixes, the more T cell responses you tend to see – despite all the important work on defining epitopes and algorithms.

So we just threw in these overlapping peptides that were available from the NIH (National Institutes of Health). What did we throw them onto?

(Click on image for a larger version)

We threw them onto fresh blood. Ian Ramshaw actually coined this term: Overlapping Peptide-pulsed Autologous Leukocytes, or OPAL.

(Click on image for a larger version)

So what is really involved? This all started in monkeys. We don’t have a mouse model for this. What happens is that you take blood, you make PBMCs (peripheral blood mononuclear cells) in the first case, you take the PBMC pellet, you throw on some peptide and you infuse it back into the same animal on the same day. The peptide goes on for one hour. So in the grand scheme of things, of culturing dendritic cells and, really, anything else, this was pretty simple.

The startling result was the level of T cell immunity. Most of our work has moved away from ELISPOTs and into intracellular cytokine staining (ICS), where you can phenotype the cells and look at lots of different functions. But we were really seeing several per cent of T cells, both CD4 and CD8 T cells, getting activated by this process – orders of magnitude above what we had seen with any other vaccine concept that we had been involved with.

(Click on image for a larger version)

I have to declare a conflict of interest. The concept was patented. We formed a start-up company to try to move it towards clinical trials, and the university and I have shares in the company. I am not actually making any money and I am not really expecting to, but one day maybe, who knows.

(Click on image for a larger version)

The position we were in about a year and a half or two years ago was, ‘Well, sure, these are lovely immune responses’ – we were phenotyping them, showing that they were broad and they could secrete this cytokine and that cytokine, and that effector molecule – ‘but does it make a difference or not?’ As Debbie Fuller described yesterday, we designed a painfully large study to assess this, asking: does this therapy reduce viral load?

Along the way, just to make our life harder, we asked another question: is broader better? This has confounded the HIV vaccine field, as it has for many other pathogens, I think: is it better to be narrow and strong, or is it better to try and make things much broader? I have to say that our bias at the start was to make things broader.

(Click on image for a larger version)

This is the experimental design. We had 36 monkeys. This is bigger than a lot of phase I trials, except that they all come in on the same day. That is good and bad, depending on whether you design it or you have to actually do the work.

(Click on image for a larger version)

They all got infected with SIV. There was a huge amount of virus surge in the first couple of weeks. Unfortunately, four of the 36 animals died of acute infection, with really severe acute infections.

(Click on image for a larger version)

The animals were then randomised into three groups based on their peak viral load, which in our hands and lots of other hands tends to predict long-term outcome. So they were stratified by peak viral load, weight, gender and an MHC molecule called Mane-A*10, which we knew from previous studies to be a molecule that presents some dominant epitopes that animals tend to do well with.

(Click on image for a larger version)

Then we placed the animals on antiretroviral therapy for seven weeks. Of the 32 remaining animals, 26 controlled virus to below 3 logs, and the primary per protocol endpoint was on monkeys who had controlled virus, because we figured that in a human trial setting people would have controlled virus, so those are the people that you would want to look at.

(Click on image for a larger version)

During the period of antiretroviral therapy they were given four doses of this therapy, where 18ml of blood was taken out, PBMCs were generated, and either just Gag peptides or peptides spanning the entire SIV proteome were pulsed onto them for one hour and reinfused on the same day.

(Click on image for a larger version)

The primary endpoint was the difference in viral load between controls and treated groups over the first 10 weeks.

(Click on image for a larger version)

The secondary endpoint was that extended out to six months and a year after they came off antiretroviral therapy, as well as safety and T cell immunogenicity.

(Click on image for a larger version)

T cell immunogenicity: this was one of those studies big enough that we could say this was what we expected to see but it was a relief to see it, nonetheless – a relief and also interesting. Shown here are CD4 responses, CD8 responses and averages of groups.

The animals that just get Gag have very strong T cell responses, 4 per cent by ICS, which is a big response. The animals that get the ‘All’ group have a Gag response above that of controls but it is blunted as compared with the group that just get Gag. So here is a case where ‘All’ is fine but it is reducing the response as compared with limiting your antigen choice.

There is a similar profile for CD8 responses. They were slightly lower responses. These were 15 with peptides, and they probably stimulated CD4 responses a bit better than CD8 responses.

(Click on image for a larger version)

This slide is now looking at the other parts that were included in the ‘All’ vaccine: Envelope, Pol and a pool of accessory proteins, Rev, Tat, Nef, Vpu and so on. You can see the Gag responses, and then the responses to the other protein are only stimulated in the group that get all the antigens. So, as you would expect, they get the responses to the antigens that you give.

We saw some really massive responses of up to 50 per cent of all T cells responding to single proteins in this study.

(Click on image for a larger version)

To cut to the chase: this was the viral load. We saw about three-quarters of a log difference out at the primary endpoint, 10 weeks after they came off therapy.

(Click on image for a larger version)

And when you follow this out for longer, you see it is a little bit more statistically significant.

(Click on image for a larger version)

Then, to follow it out for even longer: at six months off treatment they were revaccinated off treatment and we followed them for another six months. And again there is about a log difference in viral load at a year off treatment. Furthermore, as you follow them out you start to lose monkeys, and we lost most of the monkeys in the control group, of borderline significance. So we are seeing durable control of viremia at about a log, and prevention of incipient AIDS in these animals.

So I guess it was a bit like a glass half full, half empty – a log of virus, which is actually quite a lot in this setting, in that we know that both in these monkeys and in a human scenario that would be expected to substantially prolong survival and/or requirement for antiretroviral therapy. But this is no cure; this is partial control, if you like.

(Click on image for a larger version)

I think I have said all the things that are on this slide. The Gag gave strong responses, the ‘All’ gave broad responses but some reduced responses to Gag. There is about a log reduction in both groups. And the therapy was safe; I haven’t gone into that here.

So the question then arose: why was the control of viremia identical between the Gag and the ‘All’ groups? It would make you think, ‘With these fantastically strong broad responses, why didn’t that do better?’

(Click on image for a larger version)

Part of it is that it was difficult to induce both responses at once. The stronger the responses you got to non-Gag antigens, the weaker the responses you got to Gag antigens. And the converse was true: the stronger the responses you got to Gag, the weaker the responses you go to the ‘All’ antigens. That was quite a significant effect. So it was hard to get both responses at once.

(Click on image for a larger version)

As we were doing this study and we were seeing these huge Env responses in the group that got all the proteins, a report came out (Wayne Koff showed a figure from this yesterday) suggesting that Env CTLs were actually bad. We were scratching our heads and wondering why that would be the case, but the authors had looked at 700 or so people and it seemed highly significant. In fact, on the breadth idea, when you looked back in the literature you found that people had been making Gag vaccines – DNA, poxvirus, that sort of stuff – and they were adding in Env and showing better control, but it was a bit murky in that they were SHIV studies and their Envs were nearly identical to what the challenge virus was. Maybe the benefit of the Env was in inducing narrow neutralising antibodies, and it wasn’t really anything to do with broadening out the T cell response.

So Viv Peut, a PhD student in the lab, identified and minimally mapped the epitope in all of the animals where she could find an Env response.

(Click on image for a larger version)

Then we constructed the same sort of viral load graphs, looking at whether they responded to Env or Gag or both. The really startling finding was that if you responded to Gag you did well, if you responded to Env you did no better than controls. Furthermore – admittedly this is a small group of animals now, because they have been honed down – if you responded to both Env and Gag you still did no better than controls. Not only that, but as you would expect from these viral load curves, they matched with survival as well. In fact, the animals in the ‘All’ group that did the best were the ones that just responded to Gag. That’s what this is saying.

So maybe breadth is not everything.

Lastly, making PBMCs was a pain. You had to spin them down and wash them, and it took about an hour, and it was a bit of a pain in the neck. It’s time consuming, and if you actually had to do this in the clinic the Therapeutic Goods Administration would have to come and inspect you, and look up various orifices to make sure you were clean and so on. And so we thought, ‘Well, let’s just use fresh whole blood. That’s what is in PBMCs – it has got a lot of other things in it, but let’s just use fresh whole blood.’

So we studied 20 SIV-infected animals and we randomised them, again stratified by viral load, to receive either Gag peptides pulsed onto 9ml of whole blood or PBMCs made from 9ml of whole blood, still just for one hour ex vivo, or controls.

(Click on image for a larger version)

The upshot was that it is basically the same. You see here the whole blood responses, the Gag-specific CD8 responses after three doses. This is in animals who are not on antiretroviral therapy, who see some blunting of the response. It is variable, because some animals have more immunodeficiency than others; the response correlates pretty nicely with whether they are immunodeficient or not.

This tells me that you could probably go ahead and just use blood pulsed with peptides, reinfused.

(Click on image for a larger version)

So what are we doing? I will mention one of the things we want to do. I don’t think there is anything special about the peptides here. I don’t think there is anything special about overlapping peptides, other than that they are broad. You could use any old set of peptides or whatever you make your polytope with, or whatever your favourite antigen or adjuvant is. I think there is something special about putting it onto fresh blood cells. I don’t understand why it doesn’t induce tolerance, but it never does. In a mouse setting, a lot of this stuff would induce tolerance. We don’t have a mouse model; it is all monkeys, and it seems to work.

There is a lot of preparation involved in getting this into clinical trials. GMP (Good Manufacturing Practice) peptides have been made; there is work ongoing with formulation, toxicity studies, and a lot of phase I trial planning. Basically, the upshot will probably be that it will be in healthy HIV-positive people who are on effective antiretroviral therapy; there will be phase I trials to show that it is immunogenic and safe; and in subsequent trials those people will come off their antiretroviral therapy to see if there is an effect.

The concept is that this would be a bedside approach. As shown in the slide, you have this lady who looks as if she is floating on a cloud, having no trouble at all. Her blood gets taken out, into a blood bag. The peptides are already in there. The blood bag gets hung up and the blood all gets flowed back in. It is a closed system so, hopefully, you wouldn’t have to put out bacterial culture plates in this setting. And we would see if it works or not.

I think there are substantial risks that this would fail in clinical trials, as with any vaccine, but I think it is worth a go.

(Click on image for a larger version)

That massive study was led by Rob D Rose, in my lab, with a whole lot of help. We had some statistics help, some auditing help. Antibody studies were done – it doesn’t really induce good antibodies. We had our viral loads all checked independently. We got the peptides from the NIH. We had funding from OPAL Therapeutics, the spin-off company, as well from the NHMRC, the NIH and the Terry Ragon foundation.

 

Discussion

Question: Steve, is one of the interpretations of the presence of Env CTLs confounding the data, that there is something peculiar about those CTLs that are generated? And what would happen if, in the ‘All’ group, you didn’t have the Env peptides but you had all the other peptides in there for the other proteins? I am just wondering if there is something peculiar about the Env CTL, in terms of regulatory cells being induced or some other sort of weird suppressive mechanism.

Stephen Kent: We don’t know. We know what the epitopes are. There were some shared Env epitopes, as you would expect; we know some of the MHC-restricting molecules.

In some cases they were very strong. We saw escape from Env CTL, but we also saw escape from Gag and other CTL responses where we mapped it.

I was interested in the Tregitope concept. We are trying to set up assays where we can ask: are these cells immunosuppressive? (That is CD4 cells, at least, but there is evidence now in the literature about CD8s that have some of these regulatory molecules on them.) I don’t know.

It sort of fits with a bunch of other literature. Bruce Walker has some data where he has cloned out Env CTLs and asked how effectively they kill infected targets in vitro, and he thinks they are less effective at killing. On the other hand, these Env CTLs express CD107 as if it’s going out of style. They look as if they should kill – we don’t do a lot of killing assays these days, but as far as we can tell they should kill. I don’t know.

Question: That was a very nice presentation and you are very brave to go ahead and do this, because there are so many issues that could potentially come up, one of them being clearly: how are you going to address the variability of HIV in your human population? Here you are using mac239 and challenge versus immunisation, the same peptides in the immunisation as in the challenge, and that is not going to be true in the human population. So I applaud your bravery.

I would also wonder: if you analysed the data based on conservation, if these epitopes are conserved over time in the non-human primates would you also see the same outcome? Gag is notoriously conserved. Env is notoriously variable. And so over the course of the months that you had these non-human primates infected, they certainly could have escaped from the Env peptides that you were trying to immunise with. So one wonders: if you were to retrospectively analyse your data and look at divergence of the epitopes, if you had conserved epitopes like the Gag epitopes those are presumably going to be protective over a long term, rather than the Env peptides.

Stephen Kent: I will address the second question first. We saw evolution which we proved was escape in the Env epitopes; we also saw it in Gag, we saw it in Tat – we saw heaps of escape. Basically, if you had any virus at all above about 2 logs, you saw escape. The groups that didn’t escape had much lower virus. When you escaped you got about a half a log uplift of virus, on average, if you looked across the groups, but that was true whether you escaped in Gag or Env or anywhere else.

So, sure, you see evolution and it is escape, but it didn’t seem to be different across the proteins.

Your first question, about diversity, is a good one. But you and about 5000 other people are working on ways to get around that diversity, so you can toggle the peptide. This is just one set of 100 peptides of Gag. Make another set, or add a few more in, or do whatever you want. In the ‘All’ group they got 800 peptides. You can clearly give as many as you want, give proteins.

So it has got the flexibility, without having to make a new construct and produce it and all that sort of stuff, to do that. And that would definitely be required in a human trial, no question about it, because we are rigging the system in this setting.

Question: A therapy setting is quite different from the prophylactic use of a vaccine, because you have got pre-existing responses. Have you looked to see whether you are boosting pre-existing or whether you are priming naïve, and whether in fact that is very important to the success?

Stephen Kent: Yes, we have, but it is difficult to be certain about what we are doing. In that trial, we saw responses that weren’t there when we started the vaccination but were there at the end. But whether they are just a low response, how can you tell? We have done other studies where we have taken SIV-infected monkeys and immunised them with hepatitis C peptides, and they generate responses to hepatitis C. They are usually weaker, so I think the priming does increase the magnitude of the response. (Whether that is important or not, you could argue.) So you can generate new responses. We see new responses; I can’t be certain that it is doing that, though.

Question: I have a comment and a question. Your data with Env is exactly opposite from what we saw. We saw in our therapeutic trial that Env-specific CTL was associated with improved control, but in that particular study we were also mapping them to three highly conserved Env CTL epitopes – kind of going back to what Andy Morgan was talking about. That might be the essential difference: vaccinating and specifically driving the response toward an Env response that is highly conserved. So in that setting, 100 per cent of the animals that rebounded had no Env CTL against these epitopes, and 100 per cent of the animals that controlled the virus had responses against those epitopes. So I wouldn’t throw out Env as a potential target.

As to my question: in one of your slides it looked as if you came back and revaccinated. Did I see virus load increase during the revaccination?

Stephen Kent: Yes, you did. You see a transient rise in viral load. That has been well described, that if you immunise HIV-infected subjects with any vaccine – flu or whatever – you will see a blip in virus load. We saw that here and I think it tells us, ‘Make sure you are well controlled when you immunise. Don’t immunise in a setting of active virus replication’ – which is, I think, the scenario that you would have in a clinical trial.

Question (cont.): The concept there is just to flush out the virus and expose it to that [inaudible].

Stephen Kent: Yes, right. You can flush it out and send it somewhere else! Send it to a new sanctuary site!

Question: I may have missed it, but when you added the peptides to the blood cells, were you then washing away the unbound peptide beforehand?

Stephen Kent: No. Simplicity was the key here.

Question (cont.): So maybe just free peptide would have done the same thing?

Stephen Kent: Everyone asks me that, and we have looked at that in some pilot studies but you don’t see a lot of response. Dave Jackson and I rail with each other that peptide vaccines get a bad rap, and most of it arises because if you directly inject it in or you infuse it, you don’t really see a lot – presumably because it gets chewed up by peptidase and proteases in the bloodstream.

The regulatory authorities asked me that same question, ‘Are you going to wash out the free stuff?’ And I said, ‘Well, what the heck.’ One of the venture capitalists who has invested in this said, ‘I think the peptides are great, if they’re free. It’s like food!’ I’m not quite sure that that would fly with the regulators, but no, they don’t get washed out.

> Back to program