Scientist Stories: Mark Davis, Strategies to Study the Human Immune System

So our next speaker is Mark Davis. So Mark is the director of the Stanford Institute of Immunology, Transplantation, and Infection and a Howard Hughes Medical Institute investigator. He's well known for identification in the 1980s of the elusive T-cell receptor genes, and his group has made many subsequent discovery about T cell receptors and how they function at both biochemical and at the surface of living cells. Of course, like the previous speakers, Mark's contributions to immunology have been recognized by many honors and awards, including a recent election of membership at the Royal Society in London, the National Academy of Sciences, and the National Academy of Medicine. So congratulations for all this, and we look forward to your presentation. For 20 minutes.

Hopefully, um, won't let you down. So, um, so I came at this, uh, from a somewhat different angle, and I just wanted to show that because I think there are much broader implications than vaccines, which is that the immune system is involved in a huge number of different pathologies and different diseases. Uh, you could go on, and I think there are many that we don't don't even know are immunological because the people studying those diseases aren't immunologists, and generally most people are still traumatized by the first immunology lecture they ever heard. They just— you mention the subject, and they get out the garlic and the crucifixes and just try to ward you off from telling you anything. And yet the public wants immunology. There you go. There's probably a whole immune booster aisle of Whole Foods. This is just what you see at the airports a lot. This is totally fake. It's fake because there are no standards. There's, there's no standard for whether or not you should boost an immune system. And in fact, we do boost the immune system of cancer patients quite regularly now with checkpoint inhibitors, and some of them got very severe autoimmune diseases. As a result. So you could boost too much, but unless we actually develop standards for what is actually a healthy immune system— that is, so that's what first got me into this, is say, what are— what are—

what—

how can we define metrics of immunological health? You know, if you, if you go into a doctor's office and say, you know, I have— I worry about my heart, they'll have all kinds of stuff they can sell you which may or may not be actionable, or, uh, it— and never is the whole story. But they've got a whole, whole set of stuff that, uh, is, is available and, and is authorized and validated. If you go into the same doctor's office and ask, how's my immune system, they'll look at you like you're some kind of nut. You know, are you sick? You know, kind of like they have nothing, and they have nothing they can do for you. And, uh, uh, so really, uh, we know actually all medical knowledge is now in a cloud, and And there's the immune system right there, clear as day. And why is it clear as day? Well, because it's just incredible how there's no modern immunology in basic medicine. That these are the two main assays that people do in the clinic involving the immune system, basically white blood cell counts, which is all of 1915, and complete blood cell counts, which is kind of a glorified white blood cell count, 5 whole categories. 1959. In 1959, we didn't even know that lymphocytes made antibodies. Many people thought lymphocytes were some useless cell that just kind of plowed around, probably, you know, had an osmotic effect in terms of the blood, but we didn't really know anything. So, so it's just amazing that here in— since 1960, there have been something like 12 Nobel Prizes awarded in for immune studies, and yet zip is getting into basic medicine. Now, of course, cancer and other areas that more directly involve the immune system have gotten on the stick to some extent, but we are right now in a situation where there are way more experiments, innovative experiments, being done in cancer immunotherapy than we have mouse data to support. So essentially, you just sort of Okay, go through, I have 10 different things, I'll make all possible combinations of those things and see what works. And why is that? Well, it's because the immune system is complicated, and so we had to develop a simplified model, namely the inbred mouse, and that's been great for understanding the basics of the immune system. In fact, I don't think we could have gotten anywhere without the mouse model. But we're now in a situation where we're curing mice right and left. Or as Pedro Lowenstein said, for the mice in the audience, I have wonderful news, you know, that we can cure your autoimmunity, cure your cancer. So I think it's time to move beyond the mouse system. I don't think it's over anytime soon. It's still an incredibly useful and accessible system. We still have lots of mice. There's been this disconnect between clinical data— like, all vaccines work in mice almost, and like, none of them work in humans. That's a slight exaggeration, but in recent days I hear many of those tales. I'm sure Wayne knows more than I do. And so we really have to study the human immune system directly. This is where I'm sort of coming from. This sort of, you know, I avoided medical school, and I think there are a lot of patients that should be grateful for at event or non-event. But, you know, it's, it's certainly interesting. There's certainly lots of important problems. The other thing that occurred to me is that we needed a strategy other than just making a mouse sick and curing it. And that, that brought me to a systems immunology style approach. And we've been working on that for the last 11 years, and, and there are lots of reasons to do that. But one of the things I really thought was important was to to develop a facility that could do, like, what genome centers do. Genome centers sequence— do mass sequencing of genomes. That's how we got the human genome and all the other genomes. And I remember people at that time that Jim knows more than I do, they were saying, "Well, no, we can't do big science. That's going to destroy, you know, research as we know it." No, it was absolutely the right decision to These are data that we want to get in a kind of a pipeline, in an industrial approach. And so for that reason, we developed this Human Immune Monitoring Core at Stanford, now over 11 years ago. And it's just been great in terms of providing immunological data that's done by experts, not by your students and postdocs, and reproducible high technology. We're able to use the latest technology to do that. And then we integrate this into large grant efforts from NIH and now also the Gates Foundation and the Parker Foundation have been funding the groups of investigators. So this is like a human vaccine project in one institution largely where we have expert clinicians, particularly in vaccinology. Corey Decker has been working with us now over 10 years. She recruits patients, bleeds them, gives them a vaccine, bleeds them some more, brings them back year after year if that's part of the thing. And then we— the samples go to this immune monitoring core and we collect all this data. But we always get much more blood than we need for these assays. And this provides material for a whole bunch of science labs at Stanford particularly, but also informatics analysis, microbiome studies, technology development, genomics, et cetera. Just, you know, making human samples available easily to basic scientists really, really enables them to do human work. If— most of these people could not do human work unless someone gave them samples. They just don't have the bandwidth to do— to make the clinical connections properly. This is now systemized and working really well. We have hundreds of papers that have been published. Why did we do vaccine? Well, not because we wanted to develop a better flu vaccine, but because flu vaccine is something you can give to everyone regardless of age. And it's a good stimulant in the immune system. And also in terms of the last point, it has the advantage of also being a crappy vaccine so that there are lots of non-responders. And so just— we didn't have to go to hepatitis B. We just went to flu. It's right there. And really, you know, you don't need to do thousands of people to get a few non-responders. They're right there, especially elderly people. So anyway, without going into that, we've— We have quite a few studies of this that we've done in a systems way. I don't want to dwell on those. Those are— I've got— don't have the time to do that, but just highlight a few things. One is that I'm very proud of— there was just so much literature extolling the connection of some gene polymorphism with some disease, no matter how trivial the connection was. I just thought we really needed to do— something more rigorous, and that was a twin study where we combined the power of systems immunology, as we call it, with the ability to look at monozygotic and dizygotic twins in a population and really say, you know, is— on balance, is this trait— is the variation in this trait heritable or not heritable? And it turns out that about 75% of the 200 traits in the immune system that we looked at are not significantly heritable. There might be a small genetic component, but it's basically not an item. Only a few things were really there. And what was— there was a lot of data in this paper, but I just want to highlight this because I think it's particularly interesting, is we had 16 of the 210 twins that we looked at, 16 pairs were monozygotic twins were for cytomegalovirus. And so we asked, well, what's the effect of cytomegalovirus on the sort of disposition of the entire immune system of these people? And this is what we would call an immunome. And it turned out almost 60% of all the 200 variables were changed by CMV in these monozygotic twins. So CMV has an enormous effect on the immune system. And not just any effect. We had a separate study study that we did— we're doing at the same time with a longitudinal cohort that we've been plugging away at for 11 years now, focused on aging. And whereas the older people in the cohort, there wasn't a significant influence of flu antibody response to vaccine between CMV negative and positive, or there was a lot of noise here. But what was interesting was our control cohort of 20-year-olds, perfect in every way, Actually, the CMV-positive young adults had superior flu vaccine responses than the CMV-negative. So CMV is acting seemingly as an adjuvant to a flu response. And in the same paper, we had a colleague, Paul Thomas, had a very striking mouse experiment that go along with this where if you infected mice with cytomegalovirus,, and then 5 weeks later infected with flu, the ones that had been previously infected with cytomegalovirus gave a hugely lower titer of flu, a 4-log lower titer of flu. They were almost entirely protected from the flu. So, and that effect wore off over time, but still it's very interesting, or it supports the idea. I mean, billions and billions of people around the world are infected with CMV, and most of them are just fine. I mean, we think of CMV as a pathogen because it will kill you if somebody just irradiated your immune system. But if you're lucky enough to have avoided that, it's probably a good thing, at least while you're young. When you're old, you know, all bets are off. But it's— we shouldn't just think of it as a pathogen. It's probably doing a lot of good for many more people than it's doing bad for. And then I want to switch here to talk about my almost first love, T cell receptors and how we are always the missing piece in the systems immunology was the repertoire. And James gave a great overview of how important and how vast and how complex that is. This is our own version where we focus on T cells. And particularly, the problem of T cells is getting at their T cell receptors. And so Arnold Han, a clinical fellow in the lab, developed this very simple, high-throughput way of getting single-cell T-cell receptor sequence data and also phenotypic data where you can tell, okay, this was an expanded clone of T-cells here in terms of the same sequences, multiple cells, and then what are the phenotypes of those cells? Do they make IL-2? Do they make FOXP3, et cetera, et cetera? So we can look at the sequence diversity and the phenotype. At the same time. And one of the things that came out of the earliest studies here was how clonal expansion really clues you in as to what's important in a response. So these were colon carcinomas, TIL cells, and these are CD4 cells where we sequenced 600 individual CD4 cells from this colon carcinoma versus 300 from adjacent normal intestinal tissue.. And what you can see with the color wheel here is there's massive clonal expansion. 10% of all the CD4 sequences were one sequence. And it wasn't that these cells with the same sequence all came into the tumor. It's almost certainly that one cell with that sequence came into the tumor and then was very excited by what it saw— saw some antigen— and then clonally expanded. And we followed this up with another great technique our colleague Chris Garcia developed a way in which you could display— sorry, 10 to the 9th, actually, different peptides in the same MHC in a yeast library, and then screen that library with the T cell receptor of interest in a soluble multivalent form. And on a good day, you can find the antigens. And this is an example in this colon carcinoma study that just came out a little while ago. We found two different antigens using T cell receptors that we had sequenced in this approach. And this was a shared one that two different patients basically had T cells. They weren't identical in sequence, but they recognized the same peptide essentially. We also found a neoantigen in the same— one of the patients. So there's a mixture of neoantigen-specific T cells and self-specific unmutated things. And this, this is a proof of principle that this whole system can work. The other problem is that for T cell receptors, you don't have— you just get one sequence out of billions and billions of possible sequences. There are way more T cell receptor sequences than there are antigens that T cells recognize. So a big problem is, how do you find the T cell receptor sequences that that recognize the same antigen. So they'll be different between different people. Even identical twins have only about 1 or 2% overlap in terms of actual sequence identity. So looking at sequence identity in T cell receptors is not the answer. You need to find the rules for recognition. And so what a really brilliant graduate student, Jake Glanville, in the lab did was he took a panel of peptide MHC tetramers, used them to isolate specific T-cell receptors, specific T-cells from different people, and then compared the sequences. And what immediately came out of this was that we could see conserved 3 or 4 amino acid sequences in the CDR3 of the beta chain of these T-cell receptors. And basically, it's an immunogenesis experiment without immunogenesis, which is one of the most painful things you can do in molecular biology. You let the natural variation in T-cell— tell you what's important for this given specificity that you've defined by using this particular peptide MHC label. And it turned out that these conserved sequences were exactly the residues that bound the peptide in the MHC groove in all these cases where we had a structure that we can go to. So this immediately said that For instance, in this case here, you see this conserved sequence, but you also see this other stuff here that's kind of similar but different. It turns out this doesn't matter because this is structural. This is holding— this is the sides of the CDR loop. What's really important is this contact region, and that's why it's conserved. So if you design, as Jake did, an algorithm, which he calls GLIF, that can take thousands of T-cell receptor sequences and look for ones that share just a motif, just the 3 amino acids or so, plus some other— there are some other rules that go into this, you might have a system by which you could crunch thousands of T cell receptors and look for shared specificities. Even the sequences are not identical, but they— that doesn't matter. As long as they share motifs, they will share specificity. And I have to hurry along. Cure. So we did this for CD4 T cells from TB-infected people, lately infected people, and it was really nice that we could— we took almost 6,000 sequences to start with. We crunched them, we got 100 different groups, specificity groups we call them, and these are the ones that were most shared between 3 or more people. And we could also see with the HLA type that if if people shared a particular TCR motif, they also shared an MHC allele, which was fantastic because this is a South African cohort and they expressed 69 different class II alleles. It would have been a nightmare to look at all those individually, but since we had this strong indication of what the MHC was, we could go right to reporter constructs and right to a collection of peptides that Alex Settey had curated for CD4 responses in TB. And we could find the peptide MHC target for every one of those 5 groups. And so this says the algorithm works. And that means you can use the algorithm just by itself to cruise through sequences and understand the diversity of a particular T cell response, either in a group or in a collection. And so what we're doing now for flu and other diseases like TB, we're assembling disease-specific TCR ohms so that we could have a very large collection of specificity groups in a given disease for a given MHC. So that if somebody comes in with that MHC and infected or vaccinated, you could then gauge the diversity of their response just by looking at this index this database that we've been developing. Right now we have something like 30,000 unique flu-specific sequences that we're using. And just in general, what we're finding is that this clonal expansion that I first showed you in the colon carcinoma is absolutely characteristic of T-cell responses, particularly CD8 clonal expansion. We see them in Lyme-infected people. We see them in MS. We see this in chronic fatigue patients versus normal people have very few in the blood, very few clonally expanded CD8 cells. They're really rampant in just about every disease we've looked at or in a vaccine response. So very lastly, I want to tell you about this— how an artificial lymph node has long been a goal in vaccinology. For the obvious reason that it takes so long, it costs so much to get a single vaccine into a single person— years and years and millions and millions of dollars. If you could do this in vitro initially, and you can do on a much broader scale, you can look at hundreds of vaccine candidates, you can look at thousands of adjuvants, and that's exactly what we want to do. But artificial lymph nodes have not happened yet. So we went another route to say, let's just take a real lymph node, namely a tonsil. That is, something like half a million tonsils are whacked out of innocent little children every year, and we can catch some of those if they're usually just thrown away. So if you're standing there with an ice bucket, you can get a chunk of tonsil. And Lisa Wager really made this work where you can basically simulate the tonsil with the live attenuated flu flu vaccine, for example, you get a real response. You get germinal centers. These are clusters of B cells. And then there's a T cell zone here, just like in a lymph node typically. But more importantly, the proof of the pudding is we get flu-specific antibody responses. If in terms of these ELISA assays, we also get plasma blasts. Lots of sequences and somatic mutation, just like you'd hope. If you looked at affinity of these antibodies, we get quite decent affinity flu-specific T cells. Some are mostly HA head, but some stem as well. So there's, there's hope there. And we also have done this with the RSV fusion protein where we can also get decent antibody responses. We can manipulate the system. So one of the common problems that people have with human work is the inability to do mechanistic experiments in many cases. I don't think that's absolute, but I think that's the— that's what most people think. This is mechanistic experiments in spades. You can manipulate the genes, you can manipulate the molecules, manipulate the cells, and we've done some of that. If you remove the dendritic cells from the culture, nothing happens. Although if you remove memory cells, it's, it's actually okay, nothing bad happens. Anyway, we're picking that apart. So all these stories, uh, uh, are, uh, I think really helpful in, uh, both in understanding human immunology and in, uh, working towards, uh, we hope will be an immunometer, uh, metrics of immunological health in general. We go along with the study we just heard about and others I know of where the pre-immune status seemed to be the biggest correlate with a good vaccine response versus a bad vaccine response. I think this tonsil organoid system will be a really valuable system for vaccine development, adjuvant development, mechanistic studies of one of the most basic things that happens in immune response, adaptive immune response, and probably there's lots of innate stuff going on too. And I think I will go to the last slide and thank the people. Peter Brodin really led the flu, the twin work. Jacob and Huang Huang did the TB work. Harnell Han and others did the single-cell T-cell sequencing, and Lisa Wager and others have really powered through the tonsil organoid system.

So I will stop there. Thanks a lot. So you were first one, and I'll take only one question.

Go ahead.

Probably recall responses because of flu and RSV.

Can you speak up? Basically— You generated a primary response in anything? We can get a bit of a primary response with some HIV antigens, but they're IgM. They're not very high affinity. But what we can do, because we can freeze down lots of aliquots of the tonsils and we can work with just those frozen aliquots, What is on the list right now is, can we jump from one aliquot to another? Can we start a response, primary response, and then boost it by adding more cells from the same individual, from the same tonsil? And I think that should be possible.

Thank you very much.