Cells： from Robert Hooke to Cell Therapy - Professor Bent Jakobsen

So, our next speaker is Dr. Ben Jacobson, who is the Chief Scientific Officer of Adaptive Immune Limited, who's going to talk about fine-tuned T-cell receptors for cancer immunotherapy. Could we please welcome Ben to the podium? Thanks very much, Ben. Pleasure to have you here. Thank you very much. I would like to sincerely thank the organisers for the invitation to speak at this very different and very inspiring meeting, and also to all the other presenters. It's been really interesting. I've really learned so much in such a short space of time. Now, over the last few years, there's been a dramatic surge in the interest of immuno-oncology. This is illustrated by many things, amongst them the cancer immunology journals. That's actually a little bit of a pain, because it's very hard to follow all of them. But there are several reasons for this, and it's, of course, because some of the treatments that have been used recently have shown some great promise. Now, if we look at the different types of treatment that are used in cancer that harness the immune system, particularly T-cells against the tumour cells, vaccines, of course, was a natural place to start. Vaccines have saved far more lives than all other medical science together. So it was not unnatural that vaccines were a natural place to try against cancer. But in many ways, it's actually been vaccines that, for decades almost, gave immuno-oncology a somewhat bad name, because vaccines did not actually work very well. The results have been quite disappointing. Some would say dismally disappointing, but certainly disappointing. And it's only more recently that other types of techniques have sort of shown the potential, we hope, of the immune system to maybe even cure some cancers. One of the ways is to extract T-cells from tumours. It's a very elaborate process. It's also somewhat error-prone. You don't really know what you're getting out of it, and you don't know if there really are T-cells there that actually can mediate a good response to the tumour. But if you isolate them and expand them and give them back to patients, you can sometimes break the tolerance to tumours and get complete, durable responses. But of course, very costly. It's very much a customised type of therapy. More recently, there's been something introduced that is far more general, and this is checkpoint inhibitors. They come mainly in two forms, some that target CTLA-4 and some that target PD-1. The first extends the activation of T-cells, and the other one sort of helps T-cells get activated in the first place. These checkpoint inhibitors have also given complete responses, but in a very small subset of patients where clearly tolerance has been broken. Then there's a class of drugs which are more specific. They are called bispecifics. It can be antibodies, or it can be T-cell receptors that are used for this. They build a bridge between T-cells and the target cell, in this case the cancer cell, and essentially hijacks T-cells that are not specific for cancer cells to kill them. They are showing good results, but it's still very early. It's the early of the T-cell receptors. I wear another hat on that front, but that's not what I'm going to talk about today. Finally, there is a more advanced, one could say, type of therapy where you directly equip the T-cell from the patient with a new receptor that can recognize the cancer. This can, again, come in the form of an antibody. This is called chimeric antigen receptors, and I'm sure you'll hear a lot more about that later from Carl June. Or you can equip them, actually, strangely, with a T-cell receptor, which of course they had in the first place, but with a new T-cell receptor that has a different specificity rather than being whatever it was before. You make a lot of T-cells that have a T-cell receptor specific for cancer antigen. These types of therapies are also showing strong responses against antigen-positive tumors. There's quite a lot of evidence from these types of therapy that the immune system can certainly have an impact on tumors. But, of course, the question is, can it also control it long-term, or can you actually cure cancer as such? There are quite a few lines of evidence that illustrate that maybe the immune system cannot cure the cancer, but it can maybe control it for a very long time. I think one of the most intriguing bits of evidence is shown here, where essentially some patients who had cancer and were, we think, cured of it, certainly they had no signs of the cancer for a very long time. In this case, it was a patient who didn't show any signs of cancer or melanoma for 16 years. Some of these patients, when they become donors to other people who need organs, what you see is that because these patients are then immunosuppressed, the cancer flourishes again and completely overgrows the organ. In this case, this happened with this melanoma patient who had, we thought, been cured for 16 years. I think there are two points to that. One is that maybe it's not possible to actually cure cancer. Maybe you can't remove cancer completely. Just maybe. There are always some cells there. We don't know, but this kind of event like this might indicate that that is true. The other point is that despite that, it's also clear that the immune system can control and suppress cancer for decades. That's, of course, very important because that gives us a hope that if we can somehow break the tolerance to the cancer and get the treatments right that harness T-cells, maybe we can control cancer for life. Now, most of the technologies that are advancing now are depending on T-cells being present in the tumour. It's very clear that how hot or cold, as is now referred to in immunology slang, tumours are varies quite dramatically. Some tumours have a lot of T-cells infiltrating them. It's clear that that is very linked to survival. Other tumours do not have very many. It varies between patients in an indication, but it also varies dramatically between different indications. There are some types of tumours, notably melanoma, where there are more T-cells present than there are in some other types of solid tumours. That will probably determine how difficult it will be to actually treat these types of cancers and the individual patients with immunotherapies. If we take the sort of two types of treatments that are maybe dominating the field right now, they actually fall at completely different ends of the spectrum. On the left, you have the checkpoint inhibitors. They blanket activate the immune system. You get polyclonal T-cell responses against the tumour, but you only break tolerance in a small minority. That's probably because it's not possible to give enough of them because, of course, when you blanket activate the immune system, you get very severe toxicity. At the other end of the spectrum, you have targeted immunotherapies, which generally target just one antigen on the tumours. These are things like CAR T-cell therapies and TCR, transduced T-cell therapies. These are monoclonal responses. The toxicity is usually very mild, if you get it right. Toxicity is not generally associated with general activation of the immune system, but more to what they actually target, so if you have on-target activity on normal tissues, or indeed if you have cross-activity through the receptor that's guiding them. The question is, where are we going from here? I'd like to run you through some of the general background about how things look, how they may develop, and a little bit about what we specifically do. Now, if you look at the targeted responses, because of the lower toxicity that you can potentially have with those, they're clearly the most attractive. The problem is finding the right cancer targets, and this is truly an agonising hunt. If you look at cancers and how they develop genetically, they probably start from some founder mutations, but they quickly branch out and become incredibly diverse. Some of that, as illustrated in this paper from Charlie Swanson and co-workers, you can see that the tree on the left is maybe more diverse than the tree on the right. As time goes by, the immune system may be able to cope with some of those mutations and therefore remove some of those cancer cells. You get an editing of the tumour and the mutations that are present in it, but still you end up with something that is incredibly diverse, and there are lots of individual mutations in each tumour. If we look at the types of targets that we can go after, right at the top of the spectrum, what is most specific and will avoid the most toxicity are these types of tumour-specific antigens. It has been shown in a few cases that if you target those with T-cells, you can actually get complete responses. But it's obviously very hard because you've got to identify the mutation and you also have to hope, cross your fingers, that an epitope from that is maybe presented either on a self-surface protein or on an HLA in the form of a peptide, so you can recognise it with a T-cell receptor. And that, of course, is very, very rare. So they're ideal from a toxicity point of view, but they're very, very hard to target and because they're not shared between patients, it would be a completely individual treatment. Then you have something called cancer testis antigens. These are genes that are reactivated in the cancer cells, and they're not always on all the cancer cells, but they're nevertheless good targets because they're usually specific and there's usually no reactivity or very little reactivity to normal cells. Then you have differentiation antigens. That's an interesting type. They have tissue-restricted expression, and they're very good targets because sometimes they are on all the cancer cells. Unfortunately, they will also be on similar types of normal cells, which is sometimes acceptable, but in most types of cancer is not. Then you have overexpressed antigens. Antigens that for some reason are upregulated, but of course there is a danger of normal tissue expression and reactivity. Then you have something called ubiquitous antigens that are expressed in all cells. I'm not going to talk about those. I dare not even go there because we have looked at the expression of those and they look unmanageable. Essentially, you have a sort of reverse relationship between the tumour selectivity and the prevalence. The most prevalent are the least tumour selective and vice versa. Therefore, you have this trade-off here that the better efficacy you get, you can usually count on. If you have an antigen that is very prevalent in the cancer, you can usually count on that it will be more expressed in normal tissues as well. It's a real catch-22 situation and it's very difficult to balance. If you look at the two types of targets that are present in cancer cells, one is self-surface proteins, which you can target with antibodies. There are lots of really brilliant antibody technologies to do that. The limitation of it is that it's only some proteins, of course, maybe something like 10% or something like that, that are presented on the self-surface. The other problem is that it's very rare that these are specific to the cancer cells. Some are semi-specific, but it's very difficult to find some that are completely specific to the cancer cells. This is, of course, because the cancer's transformation is much more associated with what is going on in the nuclei. This is what made me interested in T cell receptor 22 years ago. I thought I would work on it for something like five years and give it a really long time, but it's still going because it's a very complex system to work with. But it has one virtue, and that is the antigens that you can target with it. All cells are turned over by the protosome into peptides, and particular peptides are transported into Golgi. Again, particular peptides are captured by HLA molecules and then presented on the cell surface. This is what T cell receptors can work on. This is a reason for working on T cell receptors because with one protein scaffold, you have something that can potentially recognize every single type of cell in the human body, almost. All those that present HLA anyway, and that's most of them. The other nice thing about it is that the HLA peptide engine has a common structure. It doesn't vary. Where the TCR binds, it doesn't vary either. It binds right at the top. The TCRs are very similar in structure as well. So it's a very predictable system when you're trying to engineer it and work with it and predict what cellular outcomes and indeed patient outcomes you might get. Now the way you find these antigens is, at least the way we find them, is through mass spec. That's because we've burnt our fingers many times from incidental evidence that said that one could raise T cells to an epitope and it actually turns out that that epitope doesn't exist on cancer cells. So we use a physical method to do that. We isolate peptides from cancer cells and identify physically which peptides are presented. I'd like to just show you a few examples of some of the antigens that one can address in this way. There are several types. One type is a cancer test antigen. This is one such example. May say 10. You can see it's expressed in quite a few different cancers but at quite restricted frequencies. The most frequent being lung screamer cell cancer and bladder cancer in this particular case. However, its virtue is that if you look in normal cells, and here is an analysis of expression in 72 types of normal cells, there is virtually no expression at all except in testis, which doesn't present HLA proteins. So you should be very safe in terms of targeting an antigen like that. Here's another type of antigen, WT1. This is ranked the most attractive antigen by the NCI of all those that are known in the literature. This is a sort of overexpressed antigen. As you can see, it does have some expression in normal tissues. From our experience, if you go about the level of the red line there, then there is a risk of getting T cell activity if you have really optimised T cells. For that reason, we think it's a slightly risky target. But of course, there's not a direct correlation between this type of expression analysis, which is mRNA, and then the amount of peptide that is actually presented on HLA. So it may not be a risk, but it certainly has expression in normal tissues. When you find targets like this that have this very prevalent presence in cancers, very high percentages of different types of tumours, you invariably also find that they have expression in normal tissues. Here's the last type, lineage-specific antigens. This one is alpha-beta protein, which is present in about 30% of liver cancer. Now, the real virtue of these is that they may, they may just, we don't know for sure, but they may, at least some of them, be present on all of the cancer cells in some patients. With AFP, certainly if one looks at staining, it looks very much like that. The downside is that there's also some expression in normal tissue, in this case in the liver, and we don't know exactly what the significance of that is. Of course, the liver has a great capacity for regeneration, but we don't know exactly which cells in the liver are expressing AFP. Now, these things are very hard to find. We've been looking for them for years and years and years. We have found about 25 targets where we've been able to identify peptides. With these targets, it's the case that they have high expression levels in at least one cancer and that the expression analysis in normal tissues seems acceptable for targeting the T cells. But it's very complex. The presence of suitable targets vary considerably in different cancers. There are many targets in melanoma, for instance. You can find many, many, many, many targets in melanoma, but it's very hard to find targets in pancreatic cancer. There's also the fact that these cancer test antigens, which seem to be very specific to the cancer, are only present in certain types of cancers. They're certainly present in melanoma. They're also quite frequent in lung, head, and neck, bladder, and so forth and so on. But there are certainly certain types of cancer where they're not present at all. For example, in prostate, leukemia, and a number of other ones. In the linear-specific antigens, they are, of course, only suitable in certain indications. Maybe some leukemias, prostate cancer, breast cancer, and maybe a few others where targeting the normal tissue as well would be acceptable. So, that's sort of the landscape for targets. I'll now turn to talk a little bit about the other side of the equation, and that's the reagent with which we target these antigens. In our case, of course, that is the T-cell receptor. The extraordinary thing about the T-cell receptor is that it's very, very dependent on its affinity to antigens. It's a fantastically fine-tuned system. I think T-cells are the only thing in nature where you have affinities of usually micromolar and you only have something like maybe 30 to 100 targets on a cell, and yet you have this extraordinary activation of T-cells which have very severe consequences because when you activate a T-cell, it will divide and soon you will have a very large number of them hunting down every single cell that presents the target that its T-cell receptor is specific for. This is mediated by an incredibly low affinity interaction between the TCR and the antigen. If we look at this and we plot random TCRs, of which we have isolated many over the years, but if you plot some of them or all of them, you get the same picture. The T-cell receptors against viruses have quite high affinity. That's illustrated by the red ones over on the left, whereas T-cell receptors against cancer antigens have exceptionally low affinities. They are the blue ones on the right. We only ever found one cancer TCR that can mix in between the viral ones. This reflects that most cancer antigens are self-antigens and T-cells undergo a selection mechanism early in life whereby they are removed if they have good recognition of self-antigens. The system doesn't really care that later on some of these antigens, although silenced in all normal tissues, may be reactivated in cancer. Essentially, evolution probably didn't care about cancer. Cancer is a disease of older age and we probably reproduced by that time. Evolution really drove it towards recognising probably viruses mainly, but certainly not cancer very efficiently. What we try to do with the T-cells is engineer T-cell receptors that are cancer-specific to be more like viral ones and then reintroduce them into T-cells. This is a very long process and a very intricate process and it's actually not very predictable. If you look at that scale of different affinities, what we usually do is we generate a whole panel of TCRs that have very tiny increments in affinity. That's illustrated by this barcode-like looking panel of lines there. Each one represents a TCR with a different affinity. What we see is that when you start with a natural TCR, usually they cannot recognise cancer cells that just present natural levels of antigen, or they recognise them very weakly. But when you engineer the affinity and you test all of the TCRs with different affinities, you can find one that has absolutely optimal response to that cancer. In this case, it's this one. This is an ELLIS box. T-cell lysate, you can see there's much more activation from, well, there's no activation from the natural one, but very dramatic activation from the engineered one. The other thing we find is that where the natural selected TCRs are far from optimal, and we can actually identify exactly the point where the potency is optimal. We also find that at some point, if you go too far, you get a lot of cross-reactivity. But not only that, very interestingly actually, the response to the antigen fades off again. It is like many things in nature, a bell-shaped curve. This is illustrated by this data here. You see, we tested a number of TCRs here. This particular TCR is specific for two very related antigens, one called MAGE-A4 and the other called MAGE-B2. And the TCRs go from left to right to higher and higher affinity. You can see the lower affinity TCRs on the left do not really recognize the cancer cells very well, whether it's the one on the top, which are A4 and B2 double positive, or it's the A4 ones in the middle only, or it's the B2 positive only ones at the bottom. You can see as you go up in affinity, you get better and better recognition. But then at some point, it fades off again. These are very specific TCRs. They don't have any cross-reactivity at these affinity levels yet. You can see how the response to that antigen actually fades out again. The T cell is clearly sensitive very much to the affinity and the kinetics of the binding to the antigen, but it has to be absolutely optimal for them to function as well as possible. These are killing assays of four different lines with the optimal TCR for the MAGE-A10 antigen that I showed you a little while ago. And you can see that you can get very efficient killing. This is killing over real time. And so when the level plateaus, that means that the cancer culture is not growing anymore, and as far as we can see, all the cancer cells have been killed. Now, the way we use this in the clinic is illustrated in a quite simple diagram here. You screen the patients for the antigen that you want to target, and you then apherese the patient, send the cells off to a central facility where we infect them with a lentiviral vector that encodes the new TCR that is optimized in terms of its affinity for the cancer antigen. And these cells are then expanded, and then they're re-administered to the patient. And it's typically something in the order of 5 to 10 billion T cells. Now, you can get some really spectacular results with this kind of therapy. This is the most spectacular we have had. This is in a sarcoma patient, a synovial sarcoma patient, who had over 200 long metastases and really was terminally ill. And she had a complete response that lasted for something like a year. And I think it's the first kind of response of its kind in synovial sarcoma, at least as far as I know. But, of course, this is, as is many times the case, this is the very best case scenario. It's far from, well, this actually doesn't happen very often. And even when it happens, this patient, unfortunately, eventually also progressed. So, as I mentioned earlier, there is this problem that if you're only targeting one antigen, you may not be able to eradicate the cancer. And that brings us to the future, because despite this good efficacy, breaking tumor tolerance with T cell redirected therapy is really rare. It may be that we can target linear antigens, and we may get complete responses there, but it's not a general solution to the problem. So the question is, can we combine two types of therapy in some way to be synergistic so that we get a breaking of tolerance and avoid excessive toxicity? If you look at a cancer patient, if you look at the sort of general picture, there will be tolerance, obviously, to their own tissues. But if you look in the tumor, there's also tolerance. And if you give checkpoint inhibitors, you very much tip that balance. Certainly systemically, and that's why you get severe toxicity there on the left. And in the tumor, the case is sometimes you actually don't tip the balance, sadly, but occasionally you do. Just occasionally you can tip the balance with the checkpoint inhibitors, which blanket activates the immune system. If you look at a targeted therapy, the picture is slightly different. If you get it right, and it really is targeted, and you've chosen a good target, then the systemic tolerance should be maintained, and you should avoid the toxicity. The picture in the patients is the same. Sometimes you break the tolerance and you get immunity, and you at least get some response against the tumor. But in other patients, you don't. So if we look at how polyclonal the different responses are, clearly checkpoint inhibitors mediate a very polyclonal response. And if you take that as the width of the arrow, whereas the length of it is the power to break through the tumor barrier, the power to do that is not so high for checkpoint inhibitors. For vaccines, unfortunately, it's quite feeble on both accounts. And if you look at TIL expansion therapy, that is probably not as polyclonal, maybe, as checkpoint inhibitors. Maybe it is, but maybe it is not. Most indications seem to be that it's focused on a few clones. If you look at redirector to doxotherapy, that's monoclonal, but very strong. TIL expansion therapy is also quite strong. So is it possible that if we combine a targeted therapy with checkpoint inhibitors, for example, there's certainly the direction I think we need to think. Might we achieve something like this, where we get both a very strong response against the tumor, but also a very polyclonal response? And will we eventually be able to break the barrier? It may not be the checkpoint inhibitors that we have nowadays, and it may not be the targeted therapies that we have right now, but this is certainly where the hope lies for the future. Although, personally, I think it will take quite a long time before we figure that out in many types of cancers. So, thank you very much for your attention. Thank you very much. Enjoy.