Ali Roghanian, Ph.D. - University of Southampton
Title: The application of humanized mouse models to study LILRB-mediated regulation of myeloid cells
Good afternoon, everybody. Thank you for attending and many thanks to Jackson lab for inviting me today and giving me the opportunity to present our recent data. So as Ralph kindly introduced me on base at the university of South Hampton, UK, and I am part of a larger group called antibody and vaccine group within the center. And then our goal is to develop the next generation of cancer immunotherapies and mass modeling is an integral component of our research here. And this is one of the reasons I was keen to learn this technology a few years back. And therefore I went to MIT in 2015 to work with young zoo Chan. And today I will be introducing a few of those models and giving you more insight into how helpful they've been in helping us to get insight into the mechanism of action of different antibodies and also human receptors.
So my talk will be split into two sub sections. So initially I will introduce some of the older models that we developed while I was at MIT. And then I'll introduce the human LILRB model that was published recently. So this overview kind of introduces the different models that are used to generate humanized mouse models. So as they are, as you appreciate, traditionally, people would have used them human blot. So PBMCs commonly use them to inject into immunocompromised mice, and that would be adult mice, but that would give you a short window because you would often most often get GVHD because of the activation of human T-cells and attacking the mouse tissue. So that's led to that group solution that came around about 10, 15 years ago, and that was injection of human hematopoeitic STEM cells. So again, there are different sources that you could use. Cord blood is a common one human bone marrow, and, , also fetal tissues feature fetal liver, especially it's a very rich source of adult human, fetal STEM cells, which are relatively immature and then great in terms of generating the humanized mice. And generally people tend to inject either young mice or Ram unit.
It takes about three months up to three months. Once you inject the cells to actually get the humanized mice, and you can actually detect these cells as shown in the bottom math facts plot here with the human CD 45 antibodies. And obviously you will have different subsets of the human cells depending on post that you're using. So one of the models that we developed them while I was at the MIT was this humanized mouse model of colorectal cancer. So the way we did that was to generate tumor organized from human biopsies and they were injected into the corner of the mice using colonoscopy guided injections. And you can see over time, these, , organize them, grew in the colon with the mice and the, the, the tumor basically proliferated. And what was interesting is that you do see the infiltration of human CD 45 positive cells as shown on the right here.
And when we looked at the different subsets, the majority of the infiltrating cells into the tumor where T-cells, and they were actually antigen experience or memory CD45ro positive T cells, which is kind of indicative of t-cells being activated and trying to fight off the tumor. And if you take the spleen from the same mass shown on the left here, you will get a kind of a contribution is different than a ratio of RA, which is more naive T cells and are all positive cells. So you have kind of a mixture of the two, whereas in the tumor is mostly memory cells. And that's great because that would give you the opportunity to still be the immune system and try to take advantage of these T cells to target that the tumors, for example, you know, using bi-specific antibodies. So that was one of the models and it was published in nature biotechnology.
Another model that we developed, we have data from the lab that are showing that cancer cells can be and can find refuge in different organs, especially the bone marrow and the liver. And once they reached that kind of organ, they become protective. We can staff therapeutics especially the, the common monoclonal antibodies used to try to treat them cancer cells. So we have data for lymphoma cells. And for example, it was shown that, you know, lymphoma within the lymphoma lymphoid tissues, such as them spleen can be readily depleted using antibodies, whereas in the bone marrow, it was not affected. And therefore you needed to stimulate the myloid itself to, to generate an effective immunotherapy. So we moved on and we generated this human breast cancer model. And then we injected the cells using ultrasound sound guided method intracardiac, and that was to generate this a metastasize feature.
So as you can see, within 10 days, you can detect the tumors within the bone marrow and the liver of the mice. And then we as subjected these mice to the same man therapy. So we did combination of therapy. So combination of a low-dose chemotherapy plus an antibody therapy as i will show in the follow up slides. So if you just treat the humanized mice harboring the tumor cell with clinical grade activity in this case, as he talks it, map it, there there's not much therapy. And as I said, we have, we had seen this with lymphoma cells in the past. If you inject the mice with low-dose cyclophosphamide, there is little therapy, but what is important is that if you do the combination therapy, cyclophosphamide low-dose software, first of all, I can condition human myeloid cells and they become a lot more specific. And therefore that is able to concentrate the proliferation of the tumor cells.
And as you can see in the left-hand graph here, there is much lower tumor burden in the mice that were treated with the chemo and immunotherapeutic regiment. And it also translating into the, in the, the survival. So you can see them, the mice are almost twice longer when you treat them with the combination of therapy. And what is interesting about these humanized mice is that actually you can then go back and enter a gate in the cells and see what, what is cyclophosphamide doing to them, for example, in this case. So that's, that's, that's, you know, the advantage of having access to these models, because, you know, you can't really do this kind of experiment with patients. It's not going to be as feasible. So what we did, we went back and we treat the humanized mice with a low dose of cyclophosphamide.
And two days later, we sorted the human and the mouse myeloid cells as shown indicating below from the spleen. The spleen, the reason we sort of explained was that as a control, as a positive control, cause we know that the myeloid cells are quite effective and Franco Pacific in the clean, as I said, based on the previous data published by the lab, and we did the bone marrow myloid cell isolation as well, and then we subjected them to RNAC. And as you can see here, both mouse and human cells were differentially affected and you had genes, which where up-regulated and down-regulated the principal component analysis on the left bottom left shows that the macrophages, the myeloid cells that were dried from the spleen had a different profile than the myeloid cells, which were dry from the bone marrow. And that was true for both human and mouse.
So M stands for mouse mouse, and then H for human as, you can see they're quite distinct. And that's interesting. So that tells you that the myeloid cells have had this imprinted signature from the niche that they have in kind of residing in. So that's one interesting fact. And also what we notice is that the, the myeloid cells, which were dried from the spleen had a more activated phenotype ad as we had kind of, you know, seen before in the kind of phenotypic experiment. So they had a more M1 or proelementary signature, as you can see in the bottom, right then gene enrichment analysis. Whereas the more marrow resident macrophages have a more immunosuppressive or M two phenotype. And again, that's kind of supporting the findings that we had because the macrophages on it on their own, they weren't efficient at targeting opsonize cells, but what we found, and again, to cut a long story short was that low dose cyclophosphamide was able to activate the myloid itself.
If you can see gene signatures such as the TNF, alpha signaling molecules and interferon gamma signaling molecules were upregulated in both mouse and human myeloid cells that were treated with low dose cyclophosphamide. And that also affected the expression of active receptors. So the inhibitory FC receptor was down regulated and the activity tripsy receptors, which are basically antibody receptors were upregulated and giving these macrophages an added advantage in terms of digesting well phagocytosing and digesting antibody opsonize that tumor cells. And again, I'm showing only a small portion of the data here, because these are published. Now you're welcome to, to read the whole paper. And that brings me to the second part of my presentation today, which is the human leukocyte IgG like receptor family. So this family is quite large and it's my, you know, they're, they're my favorite receptors in the immune system.
As you can see, there are 11 of them in the human, six of them are activating and five of them are inhibitory. And that depends on the intracellular domain. So the inhibitory molecules contain them multiple I-Team domains, and that means that once they are located, they can confer an inhibitory phenotype via phosphorylation of sheep, for example, and that limits the activation and these, these, you know, the receptors they're important for balancing the immune system, as you can see, you know, I appreciate him involved with, they're always on the negative and positive feedback loops, and they're mostly express on leukocytes and majority of them are expressed them uniquely on myeloid cells and hence my interest because I'm interested in the regulation of myeloid cells, you know, how we put them target these receptors in the context. So the tumor marker environment for cancer immunotherapy. So this is kind of an overview of what is known about these receptors.
So they are generally operated on tolerogenic cell. So for example, tolerogenic dendritic cells have been shown to, to overexpress these and that results in inhibition of T cell. They're quite important in transplantation and also during the, during pregnancy, because they kind of dampen down the, the mother's immune system, so that the matter of feed on maternal interface, they're important, especially the LILRB1 and LILRB2 inhibitory receptors are able to engage with them human lag and that kind of dampens down the immune attack against the newly formed fetus also their expression has been shown to, to enhance the development of modular dryer suppressor cells, which are again prevalent during cancer, but also in other scenarios. And for example, if they were shown in the Tronic, any model two to prolong screen allograft survival, what is interesting is that a number of pathogens have also developed strategies to, to point to these inhibitory LILR's and they can, they can then cause immuno evasion. So basically escape the immune attack. And that includes TMB and salmonella, for example, and from my point of view, they're also interesting because they are express highly within different tumor microenvironment. Now, also they are expressed on a certain tumor. So for example, the myeloid compartment and some of the link point tumors such as chronic lymphocytic leukemia express these proteins, these receptors on their surface.
So mice do not have LILIB's or LILR the LIL family doesn't exist in the ma in the mouse Harbor. You have an ortho local Sam pair of receptors. So they're referred to as paired RG G receptors. So H stands for activated and PRB stands for inhibitory and collectively these receptors represent the human, the whole L 11 human receptors. And it has been shown to have that these receptors have a similar functions relative to their human compartment human counterparts. So for example, PRB knockout mice have been shown to have a more activated myeloid phenotype within the tumor microenvironment. And that means that the mice, which are not TRB have a lower tumor burden, it means that removing the inhibitory receptor on the myloid cells, can, can make them more active against the tumor and that limits the tumor growth.
So the reason we, we require humanized mice is obviously because as I explained, mice, mice do not express these receptors. And because these are quite fundamental, the receptors in terms of immune regulation, we would require suitable models. And to date the humanized mouse models are the best platforms to, to study the function of these on, you know, recently people have started to use them to tease out the function of some of these receptors and also to have used them to develop them the next generation of therapeutics. So there are two ways that you can use these receptors for cancer immunotherapy, or for general immunotherapy. So specifically for cancer, you can use direct antibodies that's shown in the top here. So you can have antibodies bind into LILRB's on cancer cells. Out of the mentioned for example acute myeloid leukemia express system, you can have antibody, drug conjugates, and you can also develop car T cells. And there have been groups who are interested in happy developing such as therapeutics, but you can also use them for general immune stimulation or immunosuppression. So for example, when it comes to anti-tumor immunity or anti-infective immunity, you can block the infraction of the logins and the receptor. So you can have blocking receptive to these blocking antibodies to these receptors. And that would kind of remove the immunosuppression, but also if you develop acronystic antibodies, they could induce immunosuppression during let's say, auto immunity or transplantation.
I grew up in the U S generating antibodies against LILRB2. And they, they showed using a humanized mouse models that they can potentiate the activity of myloid cells against tumor cells. So a LILRB2 can, can once, ligated can also confers inhibitory signaling. As I said before, the do treat side team domains, and, you know, that leads to generation of M2 log macrophages. If you generate antibodies and you block these interactions, you can then reprogram the myeloid cells towards the M one phenotype and that results in generation of pro-inflammatory cytokines such as standard alpha as shown here, and that enhances the effect of T-cell function and what these groups did was to co inject human cells with human myloid cells, or actually use human humanized mice and engraft them with human cells. And what we can see in this figure is that the one that the mice, which were treated with the antibody shown in red, they had a, they had the, a more M one skewed phenotype compared to the non-treated mice. And when combined with anti PD one therapy, LILRB2 blockage, gave the best induce the best therapeutic effect in these humanized mice.
And that brings me to human LILRB3, which has been our interest. And the reason we focus on this antibody, and then sorry to, to focus on this receptor was because there, wasn't much known about this receptor a few years back. And so we teamed up with our industry, partner, Bio Invent international in Sweden, and using their phage display library. We generated about 50 antibody clones against them, the human receptor. And in a nutshell, what we found after several rounds of negative and positive selection and characterization, we found that our antibodies could either point with domain 2 or domain 4 of the LILRB domain. And this is important because I didn't mention, but LILRB3, and some of the other family receptors are quite polymorphic. So what this indicates is that the antibodies which bind these two regions domain two and domain four basically recognize the regions and the epitopes, which are homologous and which are not basically polymorphic they're unique to all the LILRB3 receptors.
So yes, we did find these two categories of antibodies. And when we use agonistic LILRB3 antibodies, we find fascinating observation. So one of them for example is shown in this graphical abstract from our publication that we, we treated human myeloid cells with the antibody. And overnight, when you can see the myloid cells become really elongated, and that's an indication of them becoming more immunosuppressed tolerogenic or MQ, whichever way you like to refer to, and that leads to subsequent T cell suppression. So they, they also induce, kind of a supression of T-cell proliferation as shown in the next figure. So we, you seem in the donors and then using PVFC cultures stimulated with them, CD three and CD 28 antibodies plus, or minus LILRB3 antibodies. We, as you can see here shown in red we found antibodies which could potently inhibit proliferation.
And a one was one of the most potent ones and was also specific eight, 13 was called potent, but that was found to be non-specific and it bounced on with the other members. And then we also had other antibodies, for example, eight 16, which was slightly stimulatory. And we had eight 28, which was not doing much compared to the either top control three, the PBMC's all in all, what it suggests is that our antibodies, and this is a small selection of them. You know, we, we have other categories as well, but in this case, what we found was that some of our antibodies can, can agonize the receptor can basically lead to the activation of the Arkema domains. And that was, that was useful because you know, we, we didnt know the function of this receptor, as such as I said, and we wanted to use this activity and this also, this receptor doesn't have very well-defined lichens.
So the antibodies can act as surrogates for starting the function of the receptor on the human myeloid cells. And then that brings me to the humanized mice. So as I mentioned earlier, humanized mouse models are one of the best models for studying human receptors and especially those which are not present in the mouse. As you can see in the bottom facts plots here, this is just a profile of a typical humanized mice. So they have about 50% of human CD40 circulating cells. But then when we look at the myloid it's, so compartment, the 14, he can see in blue, there is a good expression of LILRB3, but for example, the lymphocytes CD three positive cells do not express this as expected.
And I was basically when I was studying this reset, I was looking for, or a good model to, to, to study the function of the receptor in vivo. At the time we were working on the, the other publication that I just briefly showed you. One thing that we noticed was that if we inject it, alergenic lymphoma cells. So basically coming from a different donor into the humanized mice, they were getting readily rejected. So in this case, while I set out to do, and I'll go through the whole procedure to just to give you an idea of the way we do this. So we isolated human hematopoeitic gem, hematopoietic STEM cells from fetal liver donors. And I said, as I said before, that keeps you quite high number of cells. And, you know, I've had up to 18 million STEM cells and we only use 200,000 per month.
So that can give you, you know, tens of mice, basically in terms of the humanization protocol. So we pre irrigate then the fetal months. And there are about one to two days old. And again, this is one of the best kind of protocols, because if they're young, you get a better differentiation and less GVHD in the future. So yeah, we, we irritate them, irritate them. And then, and ways of there are different ways of introducing the cells ingrafting the cells into the new dates. You can do intracardiac injection or intro hepatic, and we will do both hear in Southampton do the intro hepatic point in the U S the preferred method was intracardiac, but they have the same end point. So within three months, the mice are fully humanized, as I said earlier. And they have, you can see in the bottom plots different human cell compartments, and then coming me back to the LILRB3 stories.
So what I wanted to do was to see whether, if I use our iconicity can keep it a LILRB3 activity. I could induce tolerance in these mice and basically make them make them more kind of accepting for the allergenic lymphoma cells. So this is what I did. So I gave them two shots of our antibody. And then within a week, I injected the humanized mice, which had been preconditioned or not with a 10 million allergenic lymphoma cells. And then our out the cell, the cells, I mean the mice over time. And to be honest, initially, I wasn't expecting much, it was just half of the blue and, you know, like trying to see what we could see, but lo and behold, after a number of weeks, I actually saw the humans grow really, really well in the precondition mind. So if you inject the mice with an control antibody, nothing happens and the cells get rejected, as you can see.
So the cells are re luciferase positive, positive, and you can see that the cell have totally been cleared from these control mice in the top. Whereas the mice that had this pre-conditioning regimen have readily engrafted with the lymphoma cells. And that kind of indicates that, as I said before, the ligation of LILRB3 on the myeloid cells, we're able to efficiently permit engraftment of B cells. And that, that is for the first time that somebody has been able to, to show this you know, in, invivo, that's, that's quite encouraging and you can see that survival again. So the control mice, you know, survive beyond the experiment, whereas the mice, which were preconditioned wilth LILRB3 succumb to their lymphoma in this case. And then we went back and we use their transcript analysis again, to see what's happening after we ligate them receptor or not on these human myeloid cells.
And again, a similar story. So in this case, we see the opposite. So if you ligate LILRB3 antibodies. We can see that there is a skewing towards M two, as you would expect. And then number of interesting molecules were identified from these experiments. So this is from skew for human donors, unrelated donors. And so we, we, so for example, AFI regularly being operated, following the short-term incubation with agnostic LILRB3 antibodies and regularly, and has been shown for example, to be a strong inducer of human regulatory Tcells that could be a potential mechanism for having this effect.
And basically here is just the summary side. As I said, we, we, we have generated antibodies which Once we use the act, the agonistic ones, they can activate the receptor that operates, certain factors include regularly, and that is the subsequently able to limit the proliferation of T-cells and can be a potential therapeutic intervention. So in summary, what I have shown you in the second part is that we've been able to generate specific antibodies to him a LILRB3, which is a higher columnar morphic receptor. And we've shown that ligation LILRB3 on human antigen, presenting cells can suppress T cell proliferation. And importantly, it's an activation by agonistic antibodies can induce tolerance and a prolonged allografts revival in humanized mice. And again, this is a great example of how humanized mice can, can help basic research. And in, in, in summary, we believe that LILRB3 antibodies may have potential therapeutic applications, not only in cancer, also in transplantation. So you could use blocking antibodies in cancer to, to prevent this immune evasion that tumors might be having or inducing, but you can also alternatively use the agonizing activity in transplantation on auto-immune settings. Now they kind of briefly show with our model. So that brings me to the end. And I would like to thank my group collaborators and also the founding body. And thank you for listening again, and I'm happy to answer any questions either now, or you can email me after this meeting. Thank you.