 Actually, I changed my topic a little bit to keep in line with the recent focus on PD-1 and its mechanism of action, so that's what I'm going to talk about today. I hope you will permit that. And I thank the organizers for inviting me, especially for this immunology session and everybody for staying for the immunology session, which didn't used to always happen, frankly. These are my conflicts of interest. You can read them, et cetera. So here's a patient actually, another patient treated with anti-PD-1. This is a patient of Dr. Hans Homers, who's in the audience. This is a gentleman who, 57 years old, and he had gone through multiple lines of previous therapy with tyrosine kinase inhibitors and mTOR inhibitors and presented to Dr. Homers' clinic where he proceeded to undergo careful molecular profiling, including epigenetic profiling and IHC studies, to determine that he was an ideal candidate for PD-1. Now, actually, that's completely untrue. What happened is he came in because he thought he was going to die and he had given up. This is a true story, actually. And he was just willing to try anything. So Hans enrolled him in the phase one B trial of anti-PD-1. He went on to be treated and really had a nice response. Even as early as six months we could see really dramatic shrinkage in most of his lesions. He went on to be treated for the full two years on study. And he did have a major side effect, actually. That is, he got better and gained weight. As he gained weight, he developed type 2 diabetes, actually, and a ventral hernia. So when he completed his two years of therapy, we performed a CT scan with the idea, as in the last section, if we saw metastatic disease at the time of the surgery for his ventral hernia, we would resect those diseases. We were believing in the surgical resection of metastases. Fortunately for the patient, the PET scan was completely negative, and he has a PET complete response, and it's maintained at least six months now. So the idea going forward, one idea is to try to molecularly profile patients and figure out the patients who would best respond to particular therapy, and that certainly has merit. But another approach is to try to figure out how we can combine immunologic agents like anti-PD-1 with other agents to make more patients look like this fortunate gentleman. In order to do that, though, I think it's important to think a little bit of how anti-PD-1 works immunologically. And I'm going to show you some data suggesting it's maybe a little bit different than you think. First of all, I'll show you a little data, actually, how it works in terms of cell number and cell function, and then we'll talk about its role. Dr. Sharma Pam will speak in the next section about CTLA-4 as well. And this is the story of how a CTLA-4 works. The idea, and I'm going to tell you that this is a little bit of an oversimplification, and this is going to be the point of the talk, is CTLA-4 is thought to work in the priming phase of an immune response. That is when a T cell, that's these blue cells here activated, they get signal 1 through the MHC and the T cell receptor and signal 2 through B7 molecules. In a tumor microenvironment, Dr. Finke will speak about, T cells often upregulate CTLA-4. CTLA-4 hijacks signal 2 and then therefore turns the T cell off. Anti-CTLA-4 allows signal 2 to proceed and the T cells are activated. This is actually not completely true. Recent publications by Jim Allison's group and also by Allen Corman's group at VMS suggested that anti-CTLA-4 depletes regulatory T cells in the tumor and animal models and Dr. Sharma has shown this quite elegantly in both bladder cancer and in prostate cancer. So you think, you know how CTLA-4 works, it might work another way as well. PD-1 was thought to perhaps work more in the effector phase. The idea is that when a T cell comes into the tumor, if the tumor expresses PD-L1 and the T cell expresses PD-1, it can't exert its effector function. And if you block PD-1, T cell can exert its effector function and therefore, PD-1 blockade would work in the effector phase. One thing that's important to realize with PD-1 is that if you see a cell, a T cell, a CD8 cell that's PD-1 high, what does that mean? Is that, and I think that there's a general misperception in the field that that means it's an exhausted cell. So here's a T cell seeing one kind of, so this is a CFSC plot. These are undivided cells. The cells divide from here to here, and these stripes are different divisions of cells. So these are undivided cells. They have no PD-1. As they divide to this stimulus, they rapidly upregulate PD-1. And to this stimulus, they upregulate PD-1 as well. So what you're going to tell me is that maybe these are exhausted cells, correct? Nope, not correct. These are cells dividing in response to a stimulus that results in interferon gamma production. The point being that PD-1 expression alone does not tell you that you have an exhausted T cell. In fact, PD-1 is upregulated in response to cell division, actually. And this is, in fact, a vaccine that drives T cells to divide. It's a vaccinia-based vaccine, like Trovax, which has been used in prostate cancer or Prostvax, which has been used in prostate cancer and Trovax and kidney cancer. So the bottom line is PD-1 is upregulated when T cells divide. What happens in cancer is that the antigen remains, so there's persistent antigen. So PD-1 stays up in response to that persistent antigen. And these are cases from prostate cancer. These are tumor-infiltrating lipocytes. Sorry, I went the wrong way. And you can see that PD-1 is expressed. And if you look at prostate cancer patients to CD8 cells that infiltrate their prostate, the vast majority are PD-1 positive. And they don't make any interferon gamma. They're exhausted not because they have PD-1. They're exhausted because they have chronic antigen exposure. And that's a very different way to think about that. You'll also notice that the driver that keeps PD-1 upregulated is something systemic as well, because these patients have elevated levels of PD-1 on the CD8 cells in their peripheral blood. And this will soon be studied. This has been studied by Kerry Campbell's group and kidney cancer. And I think that important data in this regard will be coming out in kidney cancer as well. So how does PD-1 blockade work on a very simple basis? To study this, we were fortunate. We had PD-1 knockout T cells. And we can adoptively transfer them into mice where they'll be turned off or they'll be tolerized, either by tumor or by self-antigen. So the experimental system is we adoptively transfer antigen specific T cells. We vaccinate them. And we see a week later whether we have more cells when PD-1 is missing or whether they function more optimally in terms of a killing assay. And the answer is really pretty simple, honestly. So adoptively transfer T cells into self-antigen model. They mostly get deleted. If you use PD-1 knockout cells or block PD-1, you have way more cells. Actually, it's almost infinitely more, right? Because these are mostly deleted over here. If you look at their function, the function is improved in terms of a CTL assay. So again, this is cells that are tolerized. They kill very poorly. If they lack PD-1 or you block PD-1, they can now kill. Same thing for a tumor model. A tumor model, this is a prostate cancer model, adoptively transfer antigen specific T cells. They have absolutely no effector function. They can kill nothing. But if you block PD-1 or PD-1 is genetically absent, the cells can now kill. So pretty simple, right? So more cells and better cells. But again, that doesn't really answer the question whether it happens in the effector phase or in the priming phase. And again, most of the literature or most folks think about PD-1 as operating in the effector phase. But quite frankly, when a T cell is initially activated, PD-1, as I just showed you, is up-regulated. And the question is, does blocking PD-1 have any effect in the priming phase? Now, this is not as complicated as it looks. This is a model, a genetic model, to test whether PD-1 blockade functions in the priming or effector phase. The basis of the model is exactly what I showed you. So you take antigen specific T cells, you stick them in a mouse that has the antigen as a self-antigen or a tumor antigen. You get the same result. And you look about a week later and you see whether they can kill. Are they functionally competent? What I can tell you is if you put these cells into self-antigen or tumor antigen, a week later, they do not kill. So the way to tell if PD-1 blockade operates in the priming phase is to block PD-1 during that priming phase with monoclonal antibodies to either PD-1 or to its ligand, PD-L1 or PD-L2. And I should point out that these molecules were originally discovered by several groups, including Dr. Freeman, who will speak after me and who's really one of the pioneers in this field by far, actually. So this is if we block with antibodies during the priming phase. We had a little trick and we were lucky. So we have mice that lack PD-L1 or L2. So if PD-L1 and PD-1 interaction operates in the effector phase, then you ought to be able to kill these cells really well. Because these cells have no PD-L1 or PD-L2. So this is the way to test whether PD-L1 blockade affects the effector phase. And again, we have positive control. This is mice that have, we treat with the vaccine. So what are the results of this? And actually, this is the readout, just in case I talked about this before. So you basically, you load cells with the target peptide. In this case, we know what it is. It's hemoglutin and we can do this other antigens as well. And if the cells are killed, the peak peptide disappears. So it's a very reliable in vivo assay for killing. This is an in vivo CTL assay. So here are the results. So quite simply, here's your positive control. You vaccinate mice and the targets are beautifully killed, nearly 100% killed, whether the targets have PD-L1 or PD-L2 on them. But check it out. So here's, and here's the negative control. Here's using isotype antibodies and here's cells that are tolerized and they don't kill whether or not the target has PD-L1 or L2 on it. What if you unblock PD-1 during the, so these are the cells that lack PD-L1 in the effector phase and you can see they're also not killed. So the bottom line is that the cells that lack PD-L1 or L2 are really not killed any differently. This really suggests that it might operate more in the priming phase and that's the effect right here. So if you block PD-1 in the priming phase during initial antigen encounter, you can certainly increase killing. And in this model, it turns out that PD-L1 blockade was more effective. In fact, we can, this is actually an amazing role. So if you block PD-L1 during the priming phase, you entirely restore T cell function. Entirely, like the mice had no cancer or no antigen. It's the same as if you just vaccinated the mice. So the bottom line is this experiment suggests that blocking PD-1 might be effective during the priming phase of T cell activation. It's a little tricky though, right? Cause I just showed you that when you do this, you have more cells and on a cell per cell basis, maybe they have more effector function. So how does this work? Is it you have more cells or do you have better cells or both? To test that, we harvested the antigen specific T cells from this model. We performed the same sort of a killing assay, only this time in vitro. So we took the cells out of the mouse and we can say if we blocked PD-1 during the initial antigen encounter, do they look different? Can they kill better? This is the control. These are T cells that come out of a tolerized mouse. They do not kill. The clear circle of cells that come out out of a vaccinated mouse, they kill quite beautifully. And sure enough, if you block PD-1 during the priming phase, you wind up with T cells that can kill better on a cell per cell basis in vitro. If you block PD-1, again in this model, it's a little bit more effective. These cells can kill better. So again, suggesting that this operates during the priming phase of T cell activation. Just not only me, Li-Ping Chen published a paper showing a similar effect. This is an in vivo model of peptide tolerance. So the bottom line is you vaccinate cells with a peptide alone, no adjuvant. They become tolerant. And you can block that with anti-PD-1 or anti-PD-L-1. So this is the initial encounter and then this is the re-encounter. And it turns out if you block either PD-1 or PD-L-1 during the priming phase, at the same time as the initial antigen encounter, you can recover function out here. But if you block PD-1 or PD-L-1 48 hours after antigen encounter, it's not effective. It's actually not effective at all. Again, and there's other data to support this notion. The Merck Serrano group has published similar data in a vaccine model. So again, it means that probably PD-1 operates during at least some degree during the priming phase. So what does that mean in terms of clinical trial design and in terms of going forward with this agent in the clinic? So first of all, what it means in terms of a biomarker is a biomarker for PD-1 activity would be something that shows ongoing antigen encounter. T cells that are seeing antigen dividing, maybe even making little interferon gamma, these would be the patients that we'd expect PD-1 block A to work in. Does PD-L-1 do that? A little bit, right? So when T cells come into the tumor and make interferon gamma, the tumor responds, adapts, by up-regulating PD-L-1. So PD-L-1 is a bit of a surrogate marker for ongoing antigen encounter. But the view is that it's a measure of the particular interaction you're blocking is maybe a little bit of a simplification. Finally, in terms of what Naomi talked about this morning in terms of adjuvant versus new adjuvant trials, if you think about this logically, think about what happens immunotherapy putically in the adjuvant setting. So first of all, when we adoptively transfer these tumor-specific T cells, they go where? They're good ones, they go to the tumor, right? Or they go to the tumor draining lift nodes. So a patient goes to surgery and what happens? The surgeons are good, you just saw a whole bunch of them. They remove the tumor. They remove the tumor draining lift nodes sometimes and they remove the tumor-specific T cells. So there would be less tumor-specific T cells in a pure adjuvant setting. Second of all, you would remove the chronic antigen and so PD-1 is very likely to go down. So you lose not only the T cells but you lose the PD-1 that you are, the very PD-1 you're hoping to block. And not only that, the interaction between PD-1 and PD-1 is gone too because the PD-1 positive tumor is gone. So basically these arguments are against the idea that you could do a pure adjuvant setting, a trial of PD-1 blockade in the pure adjuvant setting and argue for in humans at least considering the idea that maybe a more optimal approach is to block in the neo-adjuvant setting. It also tells you a little bit about other combinations. If you consider that you should block PD-1 during antigen encounter, if you combine with a vaccine like Dr. Storkus will tell us about, you probably need to have the blocking antibody in place as the T cells see their antigen. So you need to block PD-1 and follow it by. And we've actually modeled this with radiation therapy as well. And the result is similar. You need to have PD-1 blockade in place as T cells see their antigen. This is just a quick list of acknowledgments. These are the folks in my lab who performed the studies, JHU phase one PD-1 group. My great collaborator Alan Corman at Metarex who I thank for multiple reagents and my funding. And I think that's all I got. Thank you for your attention.