Scientist Stories: Tobias Kollmann, The Rules of Immunity

So let's go to the next, the second talk from Tobias Coleman. So Dr. Coleman is investigator at the British Columbia Children's Hospital. And he's currently division head of pediatric infectious disease in Vancouver. His expertise centers around newborn infections, diseases, immune ontogeny, and early life vaccine responses, employing cutting-edge technology like systems biology to extract the most information out of typically small biological samples obtainable in early life. So, as you know, you have 20 minutes.

Thank you. And I also know that I'll be reminded when the 20 minutes are up. All right. And how do I move this thing forward? The green thing in the middle, okay. I don't have to explain this to you in this room that vaccines work. And this is just showing 3 vaccines that we typically give to children, the Haemophilus influenzae, pneumococcal, and rotavirus vaccines. On the left side, you see the numbers of lives of cases prevented. On the right side, in the left column, you see the number of lives saved, just these 3 vaccines. To the right, you see the amount of money that's estimated to be saved. Again, for just 3 vaccines. Vaccines work, there's no question about this, but as Wayne already alluded to, we don't really understand how they work. For most vaccines, we don't have a clear understanding of the mechanisms of how they protect. For some vaccines, for those that we don't have the mechanism, we at least have a correlate, a biomarker that we can measure on how they work. But for very important diseases that Wayne already alluded to, some of those infectious, others cancers, autoimmune allergies, and also emerging diseases, we're not anywhere close to actually designing a vaccine based on insight on how they should be working. And that is a problem. Therefore, the mission, as Wayne already alluded to, for the Human Vaccines Project is to design, to discover and decipher the rules that drive the immune response to vaccines and target them against major global threats. And they're not just all infectious, as you heard in the beginning. So this is the mission of the Human Vaccines Project. And the solution to overcome the problem, you heard part one for that, which are about the parts, the immunome, and the very impressive work from Jim Crowe's group. And what I'm going to talk to you about right now is how the other part, namely the rules, how do these parts function, how do they work together, come together. And Wayne already showed the slide that much of what I'm going to show you right now would not have been possible even 5 years ago. So this is very much dependent on technological innovation and a vision and visionary approach to this that you just saw in Wayne's talk as an example. So that's what I'm going to focus on right now. The way we approach this— and this is nice as a segue to the questions that were just asked— is we were set up to capture as much information in an individual in relation to a specific vaccine as you could potentially do. So this includes blood, because— oopsie— that's most often studied in these assays. But we also include lymph nodes, bone marrow, and more soon-to-come mucosal tissue sampling to have an idea of what's actually going on across the various tissues. And then apply every possible approach to capturing the host response to vaccine in a broad unbiased way using what is typically called omics assistance biology. So RNA for transcriptomics, proteomics, metabolomics, both in plasma as well as in the cells. Epigenomics, the microbiome, and various targeted assays, including very fine-grained detailed analysis using single-cell analysis. And you'll see Richard Sherman talk about this in a little while. And then we couple this with information that we get out of the same individuals using the tools that Jim just showed you, namely the Human Immunome. And then we put the parts and the rules together and identify how do they actually work together. We're going to do this using vaccines initially as probes to to perturb in a defined setup the immune system. With this, we hope to identify not just the rules that happen or the changes that happen in the individual going forward, but also how past history— and Jim already alluded to some of this— imprints itself on the response today. And we're going to do this across populations. And in these key populations, we're targeting not just young healthy adults, but the vulnerable infants and the elderly as well, developing world populations including pregnant women and others as they are relevant for a specific vaccine. And it really is going to be a three-dimensional space where we look at the results using high-intensity information technology across populations, across the entire age spectrum, and across the different vaccines to understand the rules that govern the response to all vaccines, not just to a single vaccine. So the ultimate goal for the Human Vaccines Project is this: one dose of a vaccine given to anyone, anytime, anywhere, providing lifelong protection. And you may say that this is similar to what Jim described, a foolish approach, a foolish goal, and it may still seem completely out of reach. What I'm going to show you today is that we're pretty close to actually reaching some of that, maybe not all of it at once, but it has moved into the realm of feasible, similar to deciphering the entire human immunome. The best example that we set out when we first met about 5 years ago for a current vaccine that with a single shot provides protection for, as far as we know right now, up to 10 years, which is the longest period that we've studied this, is the human papillomavirus vaccine. This is a paper that just came out a few weeks ago and there's a whole special issue coming out comparing single dose to 2-dose to 3-dose trials across across the world using various different versions of the human papillomavirus vaccine. The data that's coming out is that a single dose likely is enough to provide clinically relevant protection for at least a decade, if not life. The reason that this is a good model is that it is also an inactivated vaccine, so it's safe to be given to immunocompromised subjects, the very young, the very old, and pregnancy. The problem with the human papillomavirus vaccine is indicated here by a paper from John Tsang a couple This graph wasn't made specifically for the human papillomavirus vaccine, or HPV. The problem is that it's so good of a vaccine that everybody responds to it and almost everybody is protected. If you have a vaccine that's this good, you're not going to be able to decipher what makes it work, what doesn't make it work, and extrapolate this to other vaccines. So lack of variability or low variability is actually a handicap in this setup, even though it's a gold standard for vaccination. So we looked around and discussed this amongst ourselves and identified the hepatitis B vaccine that's currently licensed from birth to old age in all populations around the world as a very good alternative. Because the hepatitis B vaccine is similar to the human papillomavirus vaccine in the sense that structurally there's some similarities. There's also differences, obviously. But it's importantly very variable in response to the vaccine. About 30% of subjects respond to the with the first dose only. They require up to 3 doses to get close to 100%. So this is actually a pretty decent model that we have a wide spectrum of responses. And because of that, we chose the hepatitis B vaccine as our first model in a demonstration study to identify how do we best approach deciphering the rules that put all the things together that you would like to know about a vaccine. And eventually get back to the gold standard human papillomavirus vaccine. So our demonstration study that I'm going to briefly mention today in outline only was meant to just demonstrate feasibility. First of all, establish a platform and provide data to allow us to plan going forward and the next step to then target the gold standard human papillomavirus vaccine or more difficult targets such as the influenza vaccine and other vaccines as they come up. The study setup was rather complicated. I'm not going to go through the details, but it involved enrolling at this point only 15 subjects just for a demonstration pilot, multiple blood samples, large volumes that includes the sequencing of the entire human immunome as you just heard from Jim. It also included the invasive tissue sampling as I mentioned, lymph node, and it will soon also include bone marrow and mucosal tissue sampling including the microbiome as a potential variable in the response assessment. And this is just an example. This on the right side is our first subject on the top. You see her getting blood draws, which we then process. I'll show you this in a minute. Then swabs for the microbiome. And on the left side, you see our setup for the lymph node biopsy, which to our surprise at least, this is to note, this is the first vaccine study that actually took tissue-invasive samples, including the lymph node of the recipients around the vaccine. And this is, to our surprise, the study subjects actually were thrilled to watch themselves being biopsied on the on the ultrasound screen. The same is not quite true for the bone marrow aspiration at this point, but we're working on that. And then we process, in this case I'm just going to show it for the blood. We process the blood in a whole wide range working with experts from each of these domains from around the world conducting not just whole blood transcriptomics but single cell transcriptomics, and this is done at the Craig Venter Institute institute by Richard Trauman and his group, and you'll hear him talk in a little while. We also work with folks at the Institut Pasteur who have a very fancy setup that's broadly implemented around the world through the True Culture system at the Milieu Interieur study. We do next-generation sequencing, as you heard from Jim already, identifying the immunome, working with folks at the Scripps Research Institute. And plasma proteomics and metabolomics is done in part in collaboration with the Wyss Institute and Dave Walt. And there's others. That are part of this team effort. Like, one group alone would not have the top-notch expertise needed to actually do this optimally. And then we put it all together. And I'm not going to show you data right now because, as I said, this was only meant to be a pilot study, but I'll get back to this in a minute. But the point of putting it all together was actually a big hurdle. A, it's a lot of data, but that's not the major hurdle. But to use tools that actually make sense out of this data. Here's an example. If you you just measure a biomarker, pick whatever one you want out of the thousands, close to 100,000 biomarkers that we're capturing, and comparing this to the vaccine response in an individual, you put them together, each one of those by themselves, as expected, wasn't able to differentiate groups. But the moment you put them together, you can actually start seeing groups parse apart. And this integrated multivariate analysis here only shown for 2 parameters. Imagine you do this for 100,000. What you're going to come across is to recognize that you're capturing a picture of a process within the host that actually tells you a biological narrative in fine-grained detail that actually starts making sense. These are not just snapshots from one angle or two. These are snapshots from 1,000 different angles. And with this, the story that emerges from this integration approach is very surprising and it's tremendous.. And the results, I'm not going to go into the detail right now, I'm going to give a preview of what we think this might lead us to, is that not the responses to the vaccine so much as the baseline status at the time of vaccination predict your final vaccine outcome. All the studies, most of the studies that have been conducted using a similar or a more intensive sampling approach and integrative analysis approach, focus on response to the vaccine as finding a signal that predicts vaccine outcome. But actually it's the baseline, immune baseline, that predicts this better than anything else. That's not entirely novel, but it has huge impact if you think about how to go forward in identifying the rules of immunity in response to vaccination. One thing that this already alludes to is that If the pre-vaccine status predicts your final outcome, you should be able to determine the most important parameters of this, miniaturize this in a system that allows this to be done in the field in a rapid manner, and then decide on the spot what vaccine is the best vaccine for the given individual in front of me. Or with existing vaccines, you can decide, does this person need 1 dose, 2 doses, 3 doses? With this, you're not just going to reduce the cost of vaccine implementation for public health systems. But for vaccine evaluation going into the future for new vaccines, you're going to reduce the cost of trials because you're going to get back to what Wayne showed you in the beginning. You'll be able to cut down the number of individuals that needs to be studied from the tens to hundreds of thousands to tens to hundreds. And you also reduce the risk associated with vaccine development because you can target subjects patients that are likely to respond or not respond, and you can therefore reduce the risk for potential failure of the vaccine or adverse events. And most importantly for us, at least at this point, trying to put the rules together of how does immunity work, you can actually begin rational vaccine design based on a database and on a background that is much, much better than what we used to have even 5 years ago. But this is true for most vaccines that we currently have in place. Hepatitis B is one of them. I already told you that the human papillomavirus vaccine is so good that nearly everyone responds. Now, what is it about that vaccine that makes it so good? Not a clue. There's lots of hypotheses, and there's people in this room who know this way better than any of us, and you'll hear from them. But the best way to approach this now is that we can give both vaccines to the same person or different people and figure out what are the differences in response to those vaccines using the exact same tools that I just showed you. So within the same system, we will be able to dissect why, for example, something as good as the papillomavirus vaccine or something not as response-inducing as the hepatitis B vaccine differ from each other. And with that, we will be able to describe comprehensively the vaccine-induced immune response, understand what the best response looks like, and best may differ from vaccine target to vaccine target and/or from population to population. And then aim to replicate this as the one mechanism that is true for all recipients. And with that, really decipher the rules of immunity for vaccine design. And the goal, therefore, of a single dose providing lifelong protection across all populations is certainly within reach. It's not going to be right around the corner, but it is within reach. And I think I'm going to briefly mention, this is obviously, as I mentioned already, a huge team effort. Multiple institutions from Craig Venter to Pasteur, Scripps, and the Wyss Institute, and others in the Human Vaccines Project. And our team at UBC that's been heavily involved in this is pretty large as well. But I want to thank all of the folks from the Human Vaccines Project that are here today that have made this even possible. But we really feel, similar to what Jim described to all of us 20 years ago at the beginning of the verge of the Human Genome Project hitting the road, this is exactly where we're at right now. So we're very much looking forward to sharing the exciting years to come and the exciting data to come with you in the next few years. Thank you.

Thanks a lot, Tobias. You've actually been short. So we can have a lot of discussion.

That goes without saying, because we did complete this demonstration trial ahead of time and on budget as well. This is becoming a pattern.

Absolutely, it's becoming a pattern. Not to criticize Jim because he went over time, did you go over budget, Jim, also? Yes, let me just put one comment. Having been in my previous life involved in a lot of discussion about public health vaccination, large-scale vaccination. I think that the idea of having personalised vaccines is something that will be very difficult to implement. It will be very difficult to produce them in the right quality. It will be difficult, you know, the price of it. So you can think about it for if it's a therapeutic vaccine against cancer, potentially yes, you can talk about personalisation. But in terms of Protecting large population against infectious diseases seems difficult. But I would imagine that what you would then try to do when you have these rules of immunity, instead of designing a hepatitis B vaccine that would work for each individual people, try to see how you can make the hepatitis B vaccine like the HPV vaccine and what needs to be changed. In order to get there. So what's the idea about that?

So first of all, what you said in the end is very much true, that the initial goal is going to be trying to reach for the lowest hanging fruit, which is to make all vaccines work in everybody. That's the goal. Looking ahead, we may not want to have all vaccines work exactly the same way in everybody. There is reasons where sometimes you might want to have not 100% zero response because you might want to have a different immune response in different populations. So that's part one. We might not want this for everyone. But what you pointed out was a similar almost ethical hangup on my end to say that, you know what, we can, for $20,000, we can give you your personal footprint from head to toe, genome, immunome, everything. That's not feasible. And is it ethical is a different question. But, and this is the piece where people who are in machine learning, artificial intelligence, and Richard is going to talk about this in a minute, will chime in. If you do this for those populations and/or individuals that can afford it and you keep that data about each individual and you build a database that's becoming more and more rich in its fine granular detail, you'll be able to extract meaning that is going to be useful for even populations for which we cannot do this on an individual level. So personalized medicine, even if there is public health resistance for the very same reasons that you as you just pointed out, I think can still be used as a vehicle to support public health implementation across the financial barriers that are certainly in place. I don't know if that begins to address your question.

Yes, let me just come back to you. I had a first question down there. Yeah. So I think in your study design, it looks like you're looking at peak titers after 3 doses.—

that's not right?

It looks like you— I saw 3 doses and then looking at pre-titer. So you'd really be missing the key differences between the response to HPV and HPV vaccine in terms of longevity of the response and also the strong response after a single dose.

So I wonder if you'd address that issue. Yeah, so this is the— the graph may have been misleading. We look at it after every single dose. So the data that I showed you that predicts the response pre-vaccine baseline was actually not only in relation to a month post the third dose, which is the gold standard, and that's why we did that. But also to the titers in the few individuals that respond with just one dose of hepatitis B. So, we do what you just suggested at every single dose for hepatitis B. [INAUDIBLE] Absolutely. And this is where this becomes an issue right now. So, this is where doing this in the same person over over time is going to be coming up in the future. We obviously can't do this in 6 months because we only had 9 months really to complete the whole study. That long-term question is in the planning stage right now. But you're right, that is the goal. Single dose and then follow them over time.

Stanley. Well, obviously age is going to be important because as we all know, infants are not responding the way older people are and don't have the lymph nodes older people do. But the point I would also make is that where I could see what you're doing as being extremely important is comparing populations, different populations, because as you know, there are real differences between vaccination in developing world population and developed And I don't understand what the differences are. That would be very important. Even rotavirus, which people are talking about the microbiome, et cetera. But the microbiome may be a reflection of other things. And so I look forward to your data and other populations.

Thank you for making that point, Stan. But we are looking forward to that data. But that's precisely where we're heading.

Go ahead. Kind of building on that and on Paul's question, in a public health context where we want to know what are the rules of the road for an infant, but by necessity ethically your volunteers are all adults. No? Are you trying this on 3, 2-year-olds?

That's correct. Yes, so we just completed, or we're in the process of conducting the study in newborns and over the the first few days of life. And it's important, this is an HPSI NIAID-funded study. That does not include yet lymph node, bone marrow, and all those biopsies. But it includes blood, and that's an important piece to point out there. The studies that I've just shown you, and this was a major criticism in the beginning, how can we translate this to very small, even premature infants, where you can get let's say half a mil of blood or a mil of blood or 2 mils of blood at most. We've miniaturized the entire platform, that I just showed you to work with less than 500 microliters of blood. We can't do the whole immunome, which is obviously because that's the number of frequency of B cells that you need to do in there, but everything else you've seen in there works with 500 microliters of blood, and we've proven that to work. So go ahead. [INAUDIBLE] Yes, there is a— as Stampp already pointed out, there's differences not just across populations but even within a population.. And those differences that have major impact on your early life trajectory, C-section being one of them because really birth is a major inflammatory process by nature, that's just normal, will have major impact on whether you have labor or not before you get the sample. Fully understood, fully agreed. We've been doing this for over 10 years right now in these early life populations. We at least know some of the variables that we will have to capture and C-section, breastfeeding and the likes Those are all major, major variables that we're trying to capture.

Okay. We take Nathalie as the last one.

[INAUDIBLE] If I may, without— I'm not sure that— I'm not going to steal your thunder, Richard. Precisely the point. The data that we are getting out of this demonstration project, and that literally it was only meant to be a feasibility study, but we were surprised to see that integrating it across, similar to Wayne's slide, with 10,000 data points for 10 subjects, we're already getting major signals out of there. And the signal that we're getting suggests that if you would give an innate immune stimulus before you give the vaccination, you would standardize the response response to the vaccine. So adjuvant and antigen don't always have to be in the same dose. And obviously this is at this point a hypothesis that we're in the process of testing. But the data that's emerging out of this little demonstration pilot already strongly suggests that that would be the case. And that would get along with what Stan already pointed out. Differences in the population, they're not just genetic. They may also be due to exposures to different microbes and infections. It's kind of like an innate immune set point change. So if we can control the set point, change the baseline, we can do— have everybody respond to a given vaccine after we adjust the baseline status. That's exactly the probable final rule that's going to come out of this. [INAUDIBLE] Acceptable to agencies, but that's—

yeah. Okay, thanks a lot. Thanks again, Thomas.