Louis J Picker l The Antibody Series

Thank you, Dennis. I'd like to thank the organizers, including Dennis, from the bottom of my heart for inviting me here, and mostly to complement them on what a great meeting this is. This combination of cutting-edge science and fun has been rare in my 40 years of doing this, so thank you and congratulations on having such a great meeting. With regard to the science, the technology we just heard about yesterday and the day before is just, as my wife would say, gobsmacking in its sophistication and elegance. But I would remind everybody, including all the young people here, that no battle plan survives first contact with the enemy. What hasn't been emphasized so much here, and I'm going to talk about today, is that all of these ideas, all of these elegant products being made, represent hypotheses that we hope will work in vivo but really have to be stringently tested in order to achieve the therapeutic goals for which we set out. That's what I do, particularly in areas where there are stringent non-human primate models. The non-human primate really allows hypothesis testing as well as iterative development to make products better. The topic I'm going to talk to you about today is one that Gary and Abel mentioned yesterday: the topic of HIV cure. As I'm sure everybody in this room is aware, ART therapy is perhaps certainly one of the top three medical breakthroughs of modern times. The ability to take a deadly disease like HIV AIDS and convert it into a treatable disease where people have essentially normal lifespans. But ART isn't perfect. Even though it prevents deaths from HIV, what happens is an accumulation of people who are infected who have to take pills every day for life. If they stop taking the pills like clockwork, the infection will come back, as well as the risk of developing AIDS. One of the things that has been noticed clinically is that sometimes you take people off antiretroviral therapy and they control the virus. Although this isn't a pure cure in the sense that you've eliminated the virus from these people, this is the state to which we would aspire to get everybody: to be able to achieve an ability to control the virus off therapy so that they don't necessarily have to take therapy for life. Not everything has a good non-human primate model, but for HIV there is, and it's probably one of the best animal models around. SIV, the virus used in monkeys, came through the same cross-species transmission route as HIV. HIV went from pseudomangabey to chimps to humans; SIV went from pseudomangabey directly to Asian-origin rhesus macaques, which don't naturally have SIV. This particular virus, which is a clonal virus, actually a barcoded clonal virus, really effectively recapitulates all the immunopathogenesis of HIV infection in humans, although it's faster and the progression is faster. It's actually a more difficult model in many ways than HIV, but it includes things like efficient immune escape and neutralization resistance. This is in contrast to people who aren't in the business probably don't know this, but combined SHIV models, which are viruses that are combined HIV and SIV, are a bit wimpy when it comes to immune evasion. So this is a much higher bar that better reflects the human disease than using a combined virus. We, in collaboration with others, have developed effective ART for SIV and have developed well-characterized and standardized models of ART therapy interruption and viral rebound. We've done hundreds of animals now, we know what happens, and so we can detect if a therapy changes what happens. This particular virus is barcoded so that it has thousands of isogenic genetic barcodes, which basically means that you can identify by sequence different clones even though they're functionally identical, and it allows you to monitor the clonal composition of an infection or, in this case, a rebound off of ART. This is the model I'm going to talk to you about today. It's a very difficult model to cure because the virus is spread all over the body of the monkey in relatively high levels even if it's being suppressed by antiretroviral therapy. When you take away the therapy, it reactivates in multiple places at once. In other studies right now, we're actually quantifying this on a systemic level, the degree of reactivation, and yes, it basically happens everywhere and simultaneously. So what about the immune control of rebound off of antiretroviral therapy? One thing we noticed right away is that if you look at the rebounding virus in a group of animals coming off ART and compare it to primary infection in animals that are otherwise similar, there's about a one and a half to two log difference. The immunization that occurs during the first infection before ART is administered and develops during ART is not completely protective but provides quite a bit of viral load reduction. In the orange line here is the primary infection for these same animals, showing that they would have been the same, but now when they were put on ART here and released from ART, they do better. This difference is T cell mediated, and you can see that by this experiment in which anti-CD8 beta, which deletes CD8 T cells specifically, is used. Instead of seeing the control looks like this, the CD8-depleted loses that one and a half to two logs and goes back to what it looks like in primary infection. So CD8 T cells clearly have an effect, but that effect only occurs late after peak rebound and mitigates but does not completely control the virus. The hypothesis is, as is the hypothesis in general with HIV and SIV, which are very immune-invasive viruses, that T cell responses come too little and too late to effectively stymie the response. Here's the response developing following viral load, and it controls it a bit, but it's unable to completely control it. The hypothesis we wanted to investigate—let me just point out here, in keeping with the importance of rebound dynamics, the timing and dynamics help determine this. There is a pretty strong correlation between time to rebound and post-ART rebound control, or post-ART set point. This suggests that if the T cells have more time because the animal takes longer to rebound, they tend to do better than if they have a very short time to rebound. The concept we're approaching in terms of immune control of post-ART viremia is enhancing the immune response to therapeutic interventions that allow for an earlier intercept of the infection, so that instead of having the normal being up here, the normal will be this post-ART treatment control. That's the hypothesis for testing, and that's the goal. You have to be careful, though. As I show in this experiment, this is a bit of a complicated slide, but basically this N-820 here is an IL-15 agonist, a very strong IL-15 agonist. It also has anti-CD20, which we do to get rid of the follicle so the virus can hide in the follicles. The problem with it is there's very strong T cell stimulation, including CD4 T cells, and when you give it directly before ART release, you can see that it actually enhances infection, creating a bigger barrier for the immune response to overcome. This suggests we need a more subtle T cell stimulation to be effective, one that perhaps also independently suppresses viral spread. We need to figure out a way to enhance immunity without blowing up the situation by activating target cells for the virus. One of the things that has had a lot of attention in the literature these days is the so-called vaccine effect. This is the concept that if you have neutralizing antibodies on board, therapeutic neutralizing antibodies, these antibodies not only would inhibit the virus on their own but actually prime for a better T cell response. Observations in humans have suggested that when you do combination therapy with neutralizing antibodies, you could see an increase in the T cell response. There are examples of clinical trials in which neutralizing antibodies were given to patients. Here you see the antibodies going up and down with therapy, then they go down, and most of the people come right back up when the antibodies get to sub-therapeutic levels. But there are some that continue to control the virus even though the antibodies have waned. This is very rare in the absence of therapy and has been attributed to a vaccine effect. The vaccine effect was first shown in monkeys with a SHIV virus, and again it's the same sort of deal. In this experiment, six of 13 monkeys that were treated were able to control the virus when the antibodies decayed, and this was shown to be CD8 T cell-mediated control. The concept was that the antibody treatment helped the T cell response be more effective. The other reason antibodies might work in this way is the mismatch between ART and antibody half-life. ART drugs basically have a one-day half-life; it takes about three days for the ART to diminish to an ineffective level once discontinued. Antibodies obviously have a longer half-life—this is a bit optimistic for a monkey antibody, which seems to be a little bit shorter—but the idea is that they would stick around longer and provide an obstacle for rebound even when they approach sub-therapeutic levels, allowing the immune system to keep up. Miles Davenport had modeled this and decided that just using neutralizing antibodies would lead to a similar kind of control as we saw with the T cell priming of ART. The studies I'm going to talk to you about today are really testing the hypothesis: just giving antibodies at the time of ART release would allow the immune response to be accelerated and therefore get ahead of the virus, as opposed to the virus being ahead of the immune response in the control situation. That's what we're trying to achieve. Fortunately, although SIV Mac 239 is very difficult to neutralize—you rarely find neutralization in natural infection—very potent neutralizing antibodies have been developed. K11 by Dennis Burton... I'd hate to have to draw this on a blackboard. Anyway, these antibodies are different targets, so they can be used together: glycan hole and V1 V4 loop versus the CD4 binding site. Importantly, they're able to neutralize 100 percent, so there aren't resistant species of this virus in response to these antibodies, and both of them have documented ADCC activity, meaning the FC portions are functional. The development of these antibodies gave us an opportunity to test this hypothesis. In pilot studies, we can show that if you apply these antibodies even to animals that are actively viremic, you suppress viral load, and the virus comes back when the antibodies go down below a titer of about 1:500 to 1:1000. The neutralizing titer isn't enough to keep the active infection at bay, and it pops back up, not because of mutation but just because the antibody levels decay. We've done a number of studies to address this question. The studies are typically organized like this: you take animals, infect them with this barcoded virus so you have lots of barcodes, put them on ART typically at day nine because we've shown that's a full reservoir—you don't gain anything by letting it go longer—and then we put them on ART so that they are aviremic. Ultimately, anywhere from 30 to 70 weeks later, we released them from ART with or without therapy. The first thing we notice is that the animals treated with the neutralizing antibodies, again both antibodies at 20 mg per kg, a relatively high dose, versus a control Ig, significantly delayed rebound from about two weeks to up to 10 weeks. As we would expect, when the animals rebounded, it was because the viral titers had decayed to about a median titer for the day of rebound of about 1:600. That's a typical neutralizing titer that needs to be achieved above if you want to keep viral suppression constant. The other thing is about these barcodes. Animals when they rebound can rebound based on the size of their active reservoir, the timing, and maybe the fertility of the field and innate and adaptive restrictions to rebound. But the bottom line is that the longer they take to rebound, the fewer clones you see coming out. If you look at the controls, the ones that rebound quickly have lots of clones; the ones that rebound slowly have just a few clones. This has to do with the nature of the reservoir in these animals, and this is normal variability. Even though these animals rebounded much later, we saw basically the same thing: the earliest ones had tons of clones, and the latest ones had very few clones. So the reservoir hasn't been affected here; it's just simply everything has been pushed back and delayed. But the real question is whether this had any effect on post-ART rebound. Sadly, in this experiment, it did not. Here are the viral loads adjusted for time of rebound so they can be looked at the same, and you can see that this looks pretty much like this. In fact, if you measure peak plasma viral load or set point or viral area under the curve, there's no difference. If you look at the CD8 T cell responses, using bronchoalveolar lavage as a way to follow an effector site, you can see that with rebound there's a boost, but there's no difference between the two groups. So it didn't seem to work in this experiment. We always like to repeat things to make sure we're correct, and we did a separate study almost identical to the first one except we waited 72 weeks before releasing from ART. Again, we saw a very good delay of rebound with these neutralizing antibodies, but even though there were some animals that controlled in these two groups, they were in both groups, and again there is no difference between post-rebound peak viral load and post-rebound set point or viral burden. That was a bit disappointing. We then thought, one of the things we've noticed in monkeys is that in many monkeys, the immune response is so feeble that it's hard to boost. So we said, let's test this hypothesis further by trying to tilt the equation a little bit towards the immune system. One way to do that is to use monkeys which have known protective MHC 1a alleles. These animals are more likely to be elite controllers, although by no means do they all become elite controllers, but they have more effective responses. The other possibility we thought about is to delay ART a little bit to give more time for the immune system to be exposed to virus to maybe make a more robust response. We decided to do a third cohort where we used monkeys only with these alleles and delayed ART start to day 35 rather than day nine. Protective MHC alleles don't affect time to rebound in the normal situation but do affect set point viral load by about a log or a little over a log. So clearly these animals have more effective immunity. This is study two, same sort of format except we started ART later and used all animals that had protective alleles, balanced with respect to males and females. Interestingly, we saw the same delay; in fact, the delay was even longer with these animals, coming out at maybe day 98. But when you look at viral load adjusted for rebound, you can see that basically the viral load profiles, even though they're improved from what we were looking at before, are still no difference between the control and the neutralizing antibody. Here are the viral loads adjusted to rebound, and again no difference. T cell responses at rebound were identical. So again, we're beginning to show that at least with these reagents, this hypothesis does not seem to be passing the test of experimental rigor. One thing that was interesting, though, is that if you compare the day nine ART guys to the day 35 ART guys—the difference is really mostly protective alleles—there's really no difference in the post-ART rebound survival curves of the controls; they all come back at exactly the same time. But the post-ART rebound survival curve of the ones with protective alleles was significantly delayed from the ones that didn't have protective alleles in that first experiment. We're not quite sure why this is, but it suggests that the prolongation of the pre-rebound period by antiretroviral therapy did allow the protective alleles to do a little bit more in preventing rebound. So it's a little hint of activity, but as I said before, when it came to the end game—what happens after rebound—we did not see a difference. Finally, we asked whether neutralizing antibody therapy would enhance anti-PD1 therapy. We have previously shown that in the absence of neutralizing antibodies, if you just use anti-PD1 in the post-ART period, you could increase control of the virus at least as long as the PD-1 was being administered. We asked the question: would delaying rebound and having anti-PD1 on board show a more pronounced PD1 effect? Sadly, again, the answer is no. There's an effect with PD1, but it's certainly no different, maybe not even as good as we saw without the neutralizing antibody. In summary, neutralizing antibodies at the time of ART significantly delay time to rebound. That's great because it shows the neutralizing antibodies are working as advertised. However, the application of neutralizing antibodies alone did not provide enhanced post-ART viral control. We could find no real evidence of a functional vaccine effect, except for that very small difference I showed you: the delay in rebound we saw in the animals with protective alleles. The rebound dynamics associated with the decay of neutralizing antibody titers indicated that fully suppressive levels of neutralizing antibodies effectively limit viral spread, meaning they're basically acting like antibody-mediated antiretroviral therapy, which could be useful for people who need drug holidays and so forth. But it doesn't seem to be offering an immunologic advantage, even when we add checkpoint blockade and tried to see if the neutralizing antibodies would help that. It did not. It's possible that longer suppression periods or neutralizing antibodies with engineered functional enhanced FC would reveal more therapeutic benefit, but at least with a natural antibody, we did not see the desired effect. We don't give up on this. As most people know, I'm a T-cell guy. This is a T-cell vaccine we're giving in the same exact model, in which T-cell boosts were given during ART and prior to ART release. Interestingly, this T-cell vaccine, which is targeted to Gag only, delayed rebound pretty significantly and has resulted so far—this experiment is ongoing—in a significant reduction in post-ART viral loads. So maybe if we're trying to do T cells, we need to be perhaps a little more aggressive with a T-cell vaccine and not count on immune complexes to act as a vaccine to raise T-cell responses. That's all I have to say. It takes a huge group to do these kinds of non-human primate studies. These are my senior partners, particularly a shout out to Jeff Lifson, who I've worked with for 23 years, who developed our viral studies; Brandon Keele, who made the barcoded virus; and all of our other collaborators shown here. With that, I thank you.