An Emerging Field of Cytokine Pharmacology, K.C. Garcia, 2021 ICIS-BioLegend William E. Paul Award

Hi, this is Chris Garcia. First of all, I'd like to acknowledge the receipt of the Bill Paul Award. I'm extremely honored and flattered, and want to express my gratitude to the board members as well as the members of the Cytokine and Interferon Society. I knew Bill Paul quite well. He was quite instrumental in stimulating a lot of my early interest in cytokines and was also a wonderful person, so this is a really great honor for me to have this award in his name. In the interest of time, I'm just going to dive right in and present my lecture today. So I want to talk about what I call an emerging cytokine pharmacology, which is really our attempts to address pleiotropy, the phenomenon by which cytokines have actions on many different cell types that can counterbalance one another because of shared receptor expression. This has been a barrier to clinical translation as well as to cytokine research, and is something I thought a lot about in terms of mechanism-based approaches to try and narrow cytokine pleiotropy. We've done this in several different ways in my lab over the years, rooted in structural approaches, and that's what I'm going to talk about today. Now, I first want to start with the cytokine receptor dimerization paradigm, which was originally established by people like Harvey Lodish and Jim Wells in the early days. But what we've gone on to definitively establish recently is that cytokines are not preformed dimers; in fact, they bind sequentially to their receptors to dimerize their receptors in a ligand-induced fashion, which causes trans-phosphorylation of the JAKs, phosphorylation of the STATs, and subsequent downstream gene expression. This paradigm offers a great deal of opportunity for cytokine engineering to address the problem of pleiotropy. The concepts I want to talk about today are the ideas of tuning cytokine agonism for therapeutic benefit. I'm going to counter that against GPCR pharmacology, a well-established classical field, which has clearly shown that for GPCRs, ligands can be modified to elicit biased, partial, and inverse agonism. These classical pharmacological concepts have not been translated to cytokine receptors, which work as two proteins brought together by a ligand. There are technical difficulties in applying these concepts to this type of type 1 transmembrane receptor system, but we've done so in a variety of ways and I think we've established what we can start to call a cytokine pharmacology. This is the menu of cytokine receptors which, at a meeting like this, is required to show. We're all interested in one or more members of these receptors, and most of them have some therapeutic relevance. The traditional pharmaceutical path of overexpressing recombinant forms of the cytokines and putting them into people has generally not been very successful for cytokines because of this pleiotropy issue, no matter which family member you're working on. This is a larger problem not only for cytokine research but for clinical translation, and something my lab is very focused on. My work has been rooted in structures of cytokine receptor complexes from the beginning when I started my lab in 2000, because they open doors. We tend to look at them as blueprints for cytokine engineering; they offer very high-resolution blueprints for what areas of the cytokine to modify to try and generate specificity. I'm not going to get into the nitty-gritty of the structures and the different details of the dimerization of these different cytokines, other than to show you that we've generated this large panel of cytokine receptor complexes, which has been the starting point for the projects I'm going to talk about today. The first one involves interferons. To the members of this society, interferons are very near and dear. I've been very interested in interferons since the early days of cytokines. We've done some published work on the type 1 and type 3 interferons that I won't go into, other than to mention that natural type 1 interferons were very stimulating for me as nature's example of cytokine engineering. There are 16 different subtypes of type 1 interferons, and they have been shown to decouple antiviral and anti-proliferative actions depending on the interferon subtype. We provided a structural basis for that in 2011, and it got us thinking that if we could endow natural cytokines with these types of type 1 interferon-like properties, we might be able to create useful therapeutics. So I'm going to talk today about type 2 interferon, gamma interferon, because that was the lingering interferon we had not been able to obtain structural access to until very recently. Gamma is a typical highly pleiotropic cytokine, made by many types of immune cells, with actions on many types of immune cells. Therapeutically, it's been very difficult to use because of this. The actions I want to talk about today mainly focus on its ability to enhance PDL1 expression at the same time as class 1 MHC antigen presentation, which is probably one of its best-known actions. These counterbalance each other in the tumor microenvironment and prevent its use as an effective immunotherapeutic. Juan Mendoza in my lab in 2019 determined the structure of gamma interferon, which is a dimer that binds two gamma R1s and two gamma R2s, and it's this beautiful hexameric structure which afforded the opportunity of making structure-based mutants of the gamma interferon dimer that allowed for different numbers of occupancies of receptors on the gamma interferon. One then subsequently made each one of these, for instance 1R1, 1R2, 2R1s, 2R1s, 1R2, and he tested their effects on phospho-STAT1 and found that compared to wild type, he was able to make a series of partial agonists which had lowered EC50s for phospho-STAT1. He went on to look at these partial agonists by looking at their gene expression profiles using RNA-seq, and he found something rather interesting: despite these variants being partial agonists for phospho-STAT1, most of them showed what we call stable gene expression, in the sense that the wild-type level of gene expression was the same as the partial agonists for most of the response of genes to gamma interferon. In contrast, there was a small set of what we call tunable genes, where the wild-type gamma interferon had much higher levels of gene expression than the partial agonist, at least as judged by phospho-STAT1. What we found was that one of the stable genes was MHC class 1; no matter how partial the agonist was, you get very strong upregulation of MHC class 1, whereas that was not true for the checkpoint ligand PDL1. This is now showing that when we exposed the A549 lung cancer line to both wild-type and partial agonist gamma interferons, you can see they all strongly upregulate class 1, but the partials are diminished in PDL1 expression. That was also true across a number of other different cancer cell lines, with different extents of bias: strong in the colorectal line, very weak in the pancreatic line, but in general we saw bias across most of the cancer cell lines. This has opened up the possibility of revisiting gamma interferon's immunotherapeutic potential in the tumor microenvironment, to decouple the concomitant activation of the checkpoint ligand with class 1, and that's something we're exploring. Another cytokine we've explored this tuning concept with is IL-10. This is work from Bobby Saxton published this year. IL-10 is mostly known for its anti-inflammatory actions, but it also has pro-inflammatory actions mediated by STAT1. IL-10 is secreted by macrophages and Tregs and inhibits monocytes and macrophages, but it has some pro-inflammatory actions on T cells. We wanted to see if we could decouple these actions, much as we did with gamma interferon. Bobby solved the cryo-EM structure of the IL-10 receptor complex, which I'm not going to get into in detail, but suffice to say that much like gamma, there are two IL-10R1s and two IL-10R betas. Using the structure, looking at the interface of IL-10 with IL-10R beta, Bobby designed a series of structure-based mutants to impair binding of IL-10R beta to IL-10 in a graded way. These are the mutants shown here. What I'm showing you here are the phospho-STAT3 induction on a monocyte cell line, the THP1, and the Daudi line, which is B cell derived. We found that when we looked at phospho-STAT3, some of the mutants retained actions on the THP1 cells but had greatly diminished activities on the Daudi cells. When we then looked at PBMCs, we found that some of these mutants, for instance the IL-10 DE, retained actions on monocytes but lost actions on T and B cells. That was proportional to the level of IL-10R beta expression; monocytes are very high in IL-10R beta, so through mass action, this partial agonist was able to retain actions on monocytes and lose it on these other types of immune cells. Bobby went on to show that this myeloid-biased IL-10 has reduced capacity to promote pro-inflammatory genes. This is showing here wild-type versus the partial agonist; you can see the partial is greatly diminished in these pro-inflammatory gene expression, and also looking at cytokine secretion, interferon gamma, IL-9, granzyme B, the partial agonist is greatly diminished compared to IL-10. However, when we look at the anti-inflammatory actions, the partial agonists in IL-10 are about the same when, for example, we treat monocytes with LPS and look at the ability of IL-10 to inhibit TNF, IL-6, and IL-8 secretion, they're about equal. So we've lost the pro-inflammatory actions but retained the anti-inflammatory actions. Bobby went on to show in a mouse sepsis model, essentially a cytokine storm model, that his partial agonist, the IL-10 DE, was able to promote survival of mice treated with LPS to the same extent as IL-10. This is very exciting because what we want to do now is revisit this partial agonist in a clinical setting for inflammatory disease, such as inflammatory bowel disease, where we've essentially shimmed out the pro-inflammatory actions and could perhaps give it a whole new life as an anti-inflammatory drug. Finally, I want to show you some recent work on another class of cytokine, the IL-12 family, which are what we call the tall receptor family because they have many more Ig domains than the typical short receptors I just talked about, like gamma interferon and IL-10. IL-12 is an unusual cytokine: it's a four-helix bundle p35 subunit complexed with a cytokine receptor-like p40 subunit, and it's secreted as this heterodimer. We've been interested in this family going back to our early work in the early 2000s on IL-6, but we've never been able to crack the IL-12 structure. Caleb finally managed to do that in the lab through a combination of crystallography and cryo-electron microscopy. What I'm showing you here is the complete quaternary complex of IL-12, with R beta 1 in green, p40 in orange, p35 in blue, and IL-12R beta 2 in purple. It's a very unusual type of geometry. I want to highlight your attention to the IL-12R beta 1 forming a direct interaction with the p40 subunit of IL-12, not the cytokine four-helical bundle portion like most of the cytokines I've previously shown you. This really breaks the mold for how cytokine receptors dimerize. We noticed that this p40-IL-12R beta 1 interaction was of particular interest because one of the main functions of IL-12 is it acts on NK cells and T cells to promote secretion of interferon gamma. In fact, NK cells secrete tremendous amounts of interferon gamma, and that's one of the reasons why IL-12 is so toxic. Clinical trials of recombinant IL-12 were not successful because of its extreme toxicity. Some early studies from 2000 showed that's due to gamma interferon, because gamma interferon knockout mice survive IL-12 administration, and when you deplete NK cells, they also survive. So the gamma interferon is made by NK cells. Caleb asked if he could use the structure to make partial agonists of IL-12 that would retain actions on T cells but not NK cells. This is now showing some residues of p40 engaging the IL-12R beta 1. Caleb tiled these residues with alanine mutants and, looking at phospho-STAT4, he made partial agonists. This blue line here is a 2x IL-12, but he made other alanine mutants which had almost unmeasurable STAT4 yet retained biological activity. Our design strategy was to make partial agonists that retained actions on T cells, which have high IL-12R beta 1 expression, as opposed to NK cells, which have low IL-12R beta 1 expression, through a similar mass action approach that Bobby Saxton used for IL-10 and IL-10R beta. What I'm showing you here is an ex vivo experiment on OT-1 mice where Caleb took T cells and NK cells out and showed that the T cells retained their ability to make interferon gamma by the partial agonist, whereas the NK cells did not. So we seem to have decoupled the T cell and NK cell actions ex vivo. He then went on to show in vivo in an MC38 tumor model that in mice treated with the cytokines, the black line is wild-type IL-12 at 30 micrograms, which is killing the mice. Wild-type IL-12 at 1 microgram is making the mice very sick until the cytokine is stopped. But with the partial agonists, the mice are perfectly healthy; there's no weight loss, and in fact their anti-tumor actions are just as strong as IL-12. So we've retained the anti-tumor properties but shimmed out the toxicity. We're extremely excited about this as a potential immunotherapeutic where we've really solved the toxicity problem, and we're trying to translate this into the clinic. So what I've shown you here is the paradigm that by borrowing concepts from GPCR pharmacology on partial agonism, we've been able to show that different immune cell types are hardwired to respond to cytokines with different response thresholds. By essentially detuning the strength of the cytokine signal, we can shim out certain kinds of cell types and achieve some degree of selectivity for the desired cell types while eliminating toxicity on undesired cell types. We're exploring this concept on many different cytokines beyond what I'm showing you here today. In a different kind of approach, what I've shown you so far is that we've detuned the cytokine signal strength by impairing the binding affinity to one of the two receptor subunits to make a less efficient dimerization. But here again, by comparison to GPCR pharmacology, where medicinal chemistry is able to access the conformational plasticity of GPCRs to elicit different downstream signaling outcomes, we're not really tapping into this type of potential because these are not changing the receptor dimer shape or topology in the way that small molecules are changing the shape of the GPCR and eliciting biased signaling. So we wanted to develop a system, a platform, where we could change the relative orientation and proximity of cytokine receptor dimers and ask if we could generate biased signaling by a similar concept as GPCR medicinal chemistry. To that end, we wanted to develop a scaffold, an alternative to the cytokine, that would be able to change the dimer orientation and topology in different types of movements in a way that the natural cytokine would be unable to. For that, we collaborated with David Baker on some engineering of a synthetic cytokine receptor dimerizer that we're able to make in different lengths and angles. As you can see, by changing the dimer angle of the EPO receptor, we're able to tune the STAT5 signaling output from maximal to essentially non-signaling. When we then took these different dimer angles and put them into a red blood cell differentiation assay, you can see EPO causes complete red blood cell differentiation from stem cells, whereas these partial agonists arrest the differentiation in proportion to the extremity of the dimer angle of the EPO receptor dimer. So this tuning concept on the outside of the cell translates to being able to influence cell phenotype, and is something we're now looking at in more relevant models, for instance in interleukin-2. Since those synthetic proteins I just showed you don't have drug-like properties, we've gone more towards the idea of using nanobodies linked together to artificially dimerize receptors as a proxy for this kind of tuning of receptor dimer geometry. We've made nanobodies to IL-2R beta and common gamma chain; this is work done by Michelle Yen in my lab. We can link them together in multiple combinations to create these large matrices of synthetic dimerizers that would predict to dimerize the two receptors in different orientations and proximities. We find that they have vastly different STAT5 phosphorylation signal strengths, and even beyond that, in data I'm not going to show you here, they also have biased signaling properties and cell type specificity. So this new area of reorienting the receptor dimer is something that I think is going to have deep pharmacological implications, and there really is a new pharmacological parameter for tuning cytokine receptor signaling that you'll be hearing more about from my lab in the future. Looking forward, I think we're really in the midst of a whole new renaissance about thinking about cytokines as therapeutics as we start to understand how to tune them. I've shown you two different ways that we're thinking about cytokine receptor tuning, but I just want to finish here by showing you what I think is on the horizon: going beyond the cytokine receptor dimer pairings that exist in nature and pairing cytokine receptor dimers that don't normally come together in nature but are expressed on the same cell. By that, I mean we are creating molecules that we call synthekines, where we borrow a receptor subunit from two natural cytokine receptor heterodimers, for instance cytokine 1 and cytokine 2, and we borrow a chain from each to create a new heterodimer on the same cell surface using a biosynthetic bispecific molecule. In this case, unpublished, I'm showing you we borrowed the IL-2R beta chain and the IL-10R beta chain from the IL-2 and IL-10 complexes to create a JAK1-JAK2 heterodimer, and we asked if this thing would be active in STAT5 signaling. Indeed, it is compared to IL-2. We're now looking at phospho-STAT5 in an IL-2R beta-IL-10R beta heterodimer induced by a nanobody-nanobody fusion protein. Not only can we elicit a STAT5 signal, we can tune the amplitude of the STAT5 signal by varying the linker length between the nanobodies. We're very excited by this because I think this gets us out of engineering natural cytokines and takes us towards a platform where we can engineer large numbers of synthetic cytokines with drug-like properties and explore different cytokine receptor pairings. We can tune cytokines as I showed you earlier, but this takes us out of the local evolutionary minima that the structures of natural cytokines keep us in and allows us to explore a much wider spectrum of possibilities. I want to end there and thank the people who did the work: Caleb Klassman led the IL-12 project, Bobby Saxton led the IL-10 project, Kritika Mohan working with the Baker lab led the EPO project, Juan Mendoza was the leader of the gamma interferon project. I do want to disclose that I'm a founder of Synthekine, where several of the molecules I talked about today have been licensed and are being commercialized. Again, I want to thank the Cytokine and Interferon Society for the Bill Paul Award. I'm very honored, and I'll be happy to take any questions.