 So my first introduction, you can make your way up here. Professor of Microbiology and Immunology, director of the Stanford NHLBI Proteomic Center. He uses flow cytometry and fluorescence-activated cell sorting and technologies developed in his lab. He's been investigating how the immune system becomes dysregulated as disease comes and goes. His lab measures and determines the cellular network states in multiple cell subsets. He holds 17 US patents. He's an outspoken proponent of translating public investment in basic research to serve the public good. And Dr. Nolan's efforts are to enable a deeper understanding not only of normal immune function, but also detailed substructures of leukemias and lymphomas in a manner never before seen, which will enable wholly new approaches to managing disease and clinical outcomes. As it becomes more and more apparent that obesity and insulin resistance are interconnected with the immune system, it's clear that obesity researchers on this campus may find some very rich collaboration opportunities that exist in the future with Dr. Nolan, his laboratory, and the methodologies he's developing. Dr. Nolan. Thank you very much. I think that first part my mother most have written saying all the nice things. So when I got asked to give this presentation at the Obesity Summit, I first thought, OK, somebody's trying to send me a message maybe. Then I need to get back to the gym. But then when I looked into it a little bit more deeply, can we get the lights down in the front? Is that possible, please? And started thinking about the roles that inflammation plays in various disease processes. One of the things that we already know, for instance, is that an area that's near and dear to my heart, which is the study of cancer, we already know that obesity is a risk factor, of course, with cancer. But we also know that inflammatory processes play a wide variety of roles in all kinds of sequela related to obesity. So the domino effect that occurs both with an inflammatory disorder and then perhaps either being obese or having a predilection to become obese has a large number of consequences downstream, both for the body as well as for the inflammatory processes that drive a lot of disease. We, of course, know that inflammation is driven entirely internally, of course, by the immune system, or largely by the immune system. And so that gets right back to the heart of things that I've always been interested in, which is the manners by which immune cells talk to each other and the manners by which immune system cells make decisions about whether they're going to act upon an inflammatory process or a pathogen or suppress an inflammatory response. So I'm going to talk about the single cell, primarily, because at the heart of it, the single cell and how immune system cells talk to each other and how they make those decisions really decide the outcome of disease. So before I even get going, I'm going to talk about information and the value of information. How do we put value on information? Because this really starts and ends how we talk about rewiring cells or how we talk about thinking about a wired system inside of a cell. So how do we put a value on information? How valuable is information? So let's say that we had three elements in a cell or we had three different cells. And each of those cells or elements can have one of two states, either on or off. So the total number of states that the system could exist within is two to the three or eight. So on or off. If we're only interested in one of those states as being relevant to a disease outcome or a clinical outcome, then the difficulty in finding it is one in eight. So we could guess and we'd have a one in eight chance of finding the right thing. But we know that systems don't exist by themselves. Systems actually exist by talking to each other in networks. So for instance, one element could talk to two downstream elements. But we know, though, that there are many other different ways that the system could exist. So the moment that you add relationships between molecules, the number of potential states the system can be in expands greatly. So for instance, I could know Chris or I could know Ed Engelman in this room. That's a yes or no answer. But how I came to know them might be very different. There could be other individuals that I knew first that then introduced me to one or the others of those individuals. So finding the state is harder when there are relationships within them. Now if you talk about just a simple setup like this where you've got three elements and multiple ways that the molecules could talk to each other or the cells or the people could talk to each other, suddenly the number of states that the system can be in increases dramatically from eight to 25. The equation is here that if you were to follow it and start to add additional molecules or cells or additional relationships that could exist within or between the elements in the system, suddenly it becomes an exponential or super exponential relationship where the number of molecules and states that they can be in can suddenly jump up to say 10 to the 17th with just a 10 element system. If only one of those states is of interest to me, it's much harder obviously to find the one in 10 to the 17 than it is to find just the number of molecules and the individual states that they're in. So we have to also understand the relationships between them. So one way to think about this is to say somebody comes into your office for an interview and you sit there and look at them for 30 minutes to determine what they're capable of thinking or whether they'll be right for the job. No, you don't. You perturb the system, right? You ask questions of the individual and by asking questions of the individual you can then begin to understand something about how their mind state operates, right? The network within their brain operates. You can see whether or not they're normal. Do they give normal responses or expected responses or you ask them, how are you today? And if they break down crying, you know that neither the question nor the result tells you what the pathology is but it tells you that there is a pathology, right? So perturbing systems allows us to understand the relationship between the elements within the systems and allows us to build models of what's going on inside of the cell. So many of the talks that you'll see later on this morning use this basic perturbation approach to build up a systems level view of how all of the components within the system operate. The idea here and this notion about basal state, what you see when the person walks in the door or when the blood comes out of the patient, being not so informative as asking questions is foundational to everything that I'm gonna show you today. It's always a comparison of the basal state which doesn't tell you very much versus the perturbed state which tells you something about how the cell or the system operates. So perturbation and connections allow us to build out the relationships and as it turns out, if you know how to do the studies it also tells you which of the many states the system could be in are the ones that are correlated to a clinical outcome of interest. Okay, so the technology that we use is all about, as I said before, the single cell. It's building upon work that was originally done by many people but started by Len Hersenberg here at Stanford with the flow cytometer looking at single cell biology and looking at proteins expressed on the surface of cells with the notion being that if you just looked under the microscope with no other measure the cells would look the same except by, if you use and stain those cells with antibodies which have fluorophores on them you can reveal populations that pre-existed within the total heterogeneous mass and you could separate them out by virtue of the surface molecule expressions. This is starting to already lose its power, okay. So two different surface molecules give me two different cell sub-populations. So that was fine in immunology for the last 20 or 30 years but I've always been interested in signaling networks, right? So the notion being that the cells could have all the same surface molecules but they could have network states pre-existing that under one circumstance would cause the network to do one thing, right, to activate certain phosphor protein elements within it and under another network state it would reveal a different set of phosphorylation states that would tell me something different about the biology of the cell. So the same idea, phosphorylation as the markers, a pre-existing mixture, you hit the cells with some kind of perturbation say a cytokine or a drug and that would reveal differences in the cell subsets that there would be something interesting presumably about those two types of cells. Now, the nice thing about doing phosphorylation states which of course are transmitters of information in signaling is that you could look, for instance, at those two cell sub-populations and let's say that there's something about these that I think is relevant to a clinical biology of cancer, for instance. So the green population would be a cell sub-population that, for instance, might have all the elements of apoptotic induction fully available, right, and that the drug would presumably then kill that sub-population but the top one would not have all of those elements across the pathway towards apoptosis and that one or more of the steps in apoptosis would be broken. Now, the nice thing then, of course, about doing phosphorylation measurements or other measurements such as apoptosis and doing this all by flow cytometry is that we can do not just the phosphorylation states that separate the cell populations but we can do phosphorylation states that tell us about the biology of each of these steps. So DNA damage, phosphorylation of ATM or phospho check, here, cross checks and other DNA cell cycle, phosphorylation states that might be relevant and then finally apoptosis, cleavage of PARP and other caspases. So we can then separate the cells and then do biochemistry at the single cell level with this kind of an interrogation approach. So years ago, we had done something like this with AML and we published this in cell circa 2004 where we had showed that with pediatric AML, if you were to stimulate pediatric AML with GMCSF and look at two phospho proteins, phosphostat three and five, you would reveal a sub-population that pre-existed but could only step out if you stimulate the cells with GMCSF. In the paper, which I won't of course go into all the details of, we showed that in fact it's that upper cell population there that actually correlated with the clinical outcome, right? That the presence of that population and the relative size of that population and the biochemistry of what's going on in the signaling proteins within that population actually determined the outcome for the patient. So here, for instance, there's a pre and post the chemotherapy for this patient and you can see that post therapy several months later, the tumor comes back and the only piece of the tumor or mostly the piece of the tumor that comes back is this upper cell population, right? And this actually predicts outcome for those patients. I actually went on later and started a company that showed that we can actually use that signature to actually predict outcome for patients and we've done now blinded trials to be able to show that it wasn't just an academic flu it actually can be used to the benefit of patients. Again, the notion being that we're using a stimulation to separate cells. Once you've separated the cell, you can then look at the biology of the bad guy cell population and show specifically mechanistically what's going wrong in that cell population. Now, how do we do this? We do it by flow cytometry, which is a single cell analysis tool. We take blood from the patient, right? We stimulate the cells with cytokines or growth factors. At a certain time point post activation because this is a stimulation, right? We will fix and permeabilize the cell. The fixation stops the cell signaling from happening any further. The permeabilization opens the cells in a manner that allows us to get antibodies inside of the cell to stain them, right? So that's the staining with antibodies and surface molecules. We then go to the cell analyzer, the cell sorter, the fluorescence activated cell sorter. We only use the analysis capabilities of the machine. And then we would reveal cell type specificity and cell signal. Now, the flow cytometer, I'm sure many of you are, if you're not immunologists, aren't fully aware of the contribution this machine makes to immunology every day, but suffice to say that this device, and I'm sure Ed Engelman will likely show some plots from using this machine and approach as well, allows us to look at rare cell populations down to as low as 0.1% of the cell population in a heterogeneous mix to determine different cell types. And as we show the different biology of those cells very quickly after stimulation, right? So the growth factor activation within four minutes led to the activation of a cell subset right here that we can then correlate for better or worse to the clinical biology. Now, the problem with flow cytometry has been the fact that fluorophores are relatively limited. The total number of parameters that you can use in a flow cytometer is around effectively 10 to 12, which is interesting because I started as a graduate student with Len Hersenberg years ago in 1983. Four parameters measured at a time was top of the line. The machine that I used at the time actually is literally sitting in the Smithsonian right now. And here we are 20, 30 years later and we've only increased from four to around 12. Meanwhile, we've watched other omics technologies, RNA, DNA, proteomics move from a few proteins measured at a time to hundreds, thousands and hundreds of thousands of elements. So that led, of course, to science envy on my part and I started searching for ways to increase the number of parameters. So I won't take you through the path we went to come to this, but we came upon an approach developed by a guy by the name of Scott Tanner at the University of Toronto where what he was doing is he thought that, well, we can replace fluorophores with isotopes from the bottom of the periodic table, things that are unique and not normally found in cells. And then, of course, rather than going to a flow cytometer, we could go to a special kind of mass spectrometer to do the measurement of these elements sensitively. And as I'll show you, this greatly expands the number of parameters that we can measure to several dozen at a time. And although I can't talk about it today, we have a new technique on the books that will allow us to do hundreds of parameters per cell or thousands of parameters per cell, thousands of cells at a time. That's what the machine looks like. So I'll tell you a little bit about this. So here is a set of fluorophores. These are quantum dots. These are relatively well-behaved for fluorophores. And you can see the problem right up front. They're already bleeding into each other's spectra. It's like trying to have multiple conversations in a restaurant and understand what's going on at any of the individual tables. Those fluorophore bleedovers cause problems statistically and even in the design of the experiment because of having to mix and match colors so as not to overlap with each other. So this is the kind of signal that we get from the isotopic mass spectrometer where we're replacing the fluorophores with isotopes. It's incredibly, has a dynamic range, that's up to 10 to the ninth, which is actually even better than fluorophores. It has a zero background, no auto fluorescence, which is something that is a boogeyman in a lot of flow cytometry experiments, the auto fluorescence. And we have as many as 100 non-biological channels available to us. So how do we do this? We have purified versions of isotopes. So Europium, for instance, comes in two stable flavors. We can get purified versions of this. Most of these metals, the lanthanides, are actually purified in China originally, or at least their mind in China. They're then sent to Russia where they are passed through gas chromatographs to separate them by virtue of weight. And they're purified then into little metal ingots, which we can then prepare as metal salts to attach to our antibodies. So everything that you see in blue there is now available to us as a new tag. So from my point of view, this is the end of fluorescence as an analysis tool for flow cytometry, because now I can have nine cadmium, eight or so palladium, a bunch of mercury, platinum, et cetera. So these are all available to us and attachable. So here's how we attach them. So we have a chelator cage, which we designed. So that's a chelator up there. We have the metal salts. The metals are chelated into the center of the salt like that, so there's gadolinium chelated. The chelation is about 10 to the minus 16 affinity. We then make a carbon backbone to which we attach multiple versions of those chelators. And then we have multiple of those backbones attached to the antibodies. So you get about 200 or so metal ions per antibody. And then those are used to stain the cells. Okay, so then what we do is we go to this device called the ICPMS, Inductively Coupled Plasma Mass Petrometer. And it's very different than the kind of mass petrometer that you're probably accustomed to thinking about in terms of 100% of the ions that go through this machine or of the material that goes through this machine is ionized, whereas the standard mass petrometers are only about 10,000 fold less sensitive. So this is what lets us get down to being able to measure a few 100 molecules per cell. That's 7,500 degrees Kelvin, 13,000 degrees Fahrenheit. The cells are passed one at a time. They're nebulized at the rate of about 1,000 cells per second, passed into that flame, which is literally the surface of the sun temperature. It's about half an inch or so wide. And the technology was all about getting a microscopic cell, shooting it through loops of flame. And I'd have a keyhole on the far side of campus and I'd be trying to shoot the cell through to get it into that over there. So we're getting good. We're at about 40% efficient. 40% of our cells make it through the hoops of flame all the way to the far side. Okay, so we create this information. If you remember a moment ago, we'll collect the information and I'll show you how we do it, what we do with it. I showed you two dimensional plots, right? Where on one axis there'd be the level of expression of one marker on the other axis, I'd have another marker. So that's a two dimensional plot. And you could see already that there were interesting cell populations that would exist there. The issue is that even for the last 30 years we were dealing with 10 or 11 dimensional data, right? So how do you think in 11 dimensions if you're not Stephen Hawking's, right? And suddenly now we're talking about having 40 and 50 dimensional data and you're thinking of these giant, thinking of these point clouds in n dimensional space. So we needed something better, some novel way to think about this so that we could get 40 dimensional data and represented it in intuitive formats. And I'll show you how we do that. So here for instance is a list of the markers that we've already validated. On the right would be surface molecules, right? Either human or mouse on, sorry, on the left. On the right are intracellular markers of interest to us. We tend to think of these things in panels where we have functional markers, surface molecules that call out the cell subsets that people know and love, T cells, B cells, macrophages, what have you. We have our perturbations that we bring to the system which are to stimulate the cells because different perturbations reveal different things about different cell types, right? No one size fits all perturbation. And then we would have intracellular panels depending upon the biology of what we're interested in, right? So cell cycle panels, apoptosis, signaling, DNA damage, we have metabolism markers, et cetera. So what we're gonna be doing now to study the immune system in the inflammatory process is to look at multiple cell surface populations all at the same time and look at the intracellular biology, the biochemistry of what they are, in the face of multiple perturbations to say, what happens when you have inflammatory cytokine IL-4 or IL-6 or IL-8, right? What happens when you have a suppressive cytokine? What happens when you put the cells into another environment where there are suppressive cells and not just suppressive cytokines, right? So looking then at the individual biology and then reconstructing the system from within. So here are two dimensional plots. That was circa 1988 when I left Stanford as a grad student. When I came back to Stanford as an assistant professor, that was more or less what we were at, eight parameters. Today we're maybe at around 14 and the technology that I've shown you already that we've recently published is already up to 32. And of course, that's the data problem, right? That's the data deluge that we're having to deal with. So how do we deal with this? So the way that we approached it was to think of any experiment as a progression, right? And any immune system analysis that we're doing, let's say, for instance, looking at bone marrow development or development of other subsets of cells from the immune system. You think of it as a developmental progression. At the top would be, say, the stem cell of the immune system and then multiple cell subsets that derive from them. For convenience, biologists have often approached this by placing cells into these compartments. You are a stem cell and then you become a this and then you bifurcate and make a decision to become one or more of these. And we hide our ignorance under this arrow and say that this becomes this and there's an arrow. So we ask the question, can we use the information in the arrow to actually reconstruct the immune system or any developmental progression that we work with? So a little bit of biology background. But it's the same thing in fat metabolism or in a progression towards an inflammatory state. You start with one cell type, it becomes another. We can often, for instance, in immune differentiation, follow the change of one cell type into other by the change of a marker expression. So one marker goes down, another marker goes up. So if we think of B cell differentiation, we would start with high CD10 level expression. And as the B cells matured, they would lose CD10 and they would gain CD20. Actually, it doesn't matter where in the immune system you look, you see these kinds of progressions. These are nicely called out just by two markers, but it turns out that often you need many markers. So think back again then to this notion of the immune system where you've got the stem cell and it's turning into all these different cell types. Markers are turning on and turning off across that progression. Now with the ability to measure all of those markers, turning on and off, we're literally going to be able to retrace statistically the things that are turning on and off and recreate the original lineage. We do this by getting rid of the biological information for the time being, capturing cells in a local similar group into what we call a metacell cluster. We create multiple metacell clusters. For all intents and purposes, the cells within that region can be considered the same. They show no statistical difference from each other. Once we've collected them in those little metacell clusters, that's nice, but what we want to do is we want to know where in n-dimensional space they reside. So to do that, we connect them to each other. We basically draw what's called a minimum spanning tree. So think of it again as a coral tree growing out from a base. The cells are changing, new things are turning on and off. You then break the coral tree into multiple little chunks but you maintain some knowledge of where they are. And you maintain that knowledge by connecting them with strings. Now what you can do, we're only doing it in three dimensions, but you can imagine it's 40 or 50 dimensions. You've captured them into these little cages. You flatten it into a circle that represents the cells and where they are. You pull the whole mess out of the n-dimensional or the three-dimensional space and you can flatten it back into two dimensions as you'll see. So here's basically how it happens. Actually, a third grader could do it. We had to do a lot of computation to accomplish this, but a third grader can do it. One marker goes down, orange goes down, blue goes up. That would be thought of as a progression. But in the immune system, again, like I said before, and I'm hammering this point home many times because it's essential for the last bits of the talk, you get all of these markers turning on and off. The problem is that we don't get the arrows. When we pull out the bone marrow, it's a mess. So the problem has always been for the last 30 years reconstructing the original tree, at least starting with that reconstruction. So we do it by asking our third grader to pick any one of the cells and then play the which of these is most like the other game and lining up the cells next to each other according to the markers, statistically the most likely progression that could occur. It's an Occam's razor progression. So we do this and you recreate the tree. And so that plus the first showing, first major showing of the mass spectrometry approach we published in science just earlier this year to show that we could not only measure multiple events, but we literally could from the raw data right out of the bone marrow of normal individuals recreate the hematopoietic trait and find some populations that people didn't even know were there before but also be able to use the approach to analyze not just what we already knew but now the signaling biology of what's going on within the cell populations. So the way that this tree is organized is according to similarity of the markers that were used on the right to create the tree. We of course measured far more than just those markers but that was all that was required to build the tree originally. And then what we do is we size the circles according to the relative number of cells within the subpopulation and then we color the circles according to the level of a marker's expression, CD45 for instance. Here you can see CD34 which is a stem cell marker, you can't quite see it, but it's up here at the tip of the tree where the hematopoietic stem cells are, right? So this allows us to organize the tree in a reproducible manner. Now what's nice about this is I could pick any of our bone marrow and I could recreate this tree and in a registration space that now provides us the ability to do a human immune system reference map, right? Just like people have done the human genome reference map. We now have a reference map that any cells normal or otherwise can be placed onto this tree. What's nice about it is of course because we have the information in this n-dimensional tree structure flattened into something that we can see at a glance, I can see all the cell subsets that exist within the immune system and now I can ask things about phosphoproteins, right? So phosphostat five, the relative expression or inducibility of phosphostat five is here from blue to red being highest. We see that the majority of the immune system in the basal state does not have very much phosphostat five except for those two little subpopulations that we find. Don't have time to talk about them particularly. But what's interesting then is we can come at the cells with an inducer, aisle seven. And what we see from this is that the whole right hand piece of the tree begins to apparently light up, right? These are the T cells and NK cells. So again, the virtue of having these in different subpopulations or having these called out like this in this registration map is that now I can start to do math on them, systems biology, relationships of what's going on. So here for instance is an induction or a full change of background where we can see that only one piece of the tree is lighting up versus others. So since I'm clearly way over already, I do yatter on, don't I? I will jump to the inflammation side of things here. So how do we view the immune system now as a system? How do we use this to measure inflammation, right? So remember I talked about perturbations, right? We perturbed the system to understand the biology. I'm gonna show you how we use people as perturbations to understand relationships within the immune system and how this is actually gonna help us understand inflammation and obesity. So one individual, individual one and individual two, they have relationships. One individual could have a different genetics or a different environmental background, right? That causes one piece of the system to turn on and other system pieces downstream to respond differently. So individual two might be slightly different, right? And so you have a different relationship map. So how do we go after this? So we approach this with rheumatoid arthritis which of course is an inflammatory disease but it could be any inflammatory disease. To ask the question, what happens to the immune system? How do the relationships change as a person progresses through RA? The same idea, healthies and individuals who have disease, multiple stimulations, multiple cell subsets, multiple readouts per cell. The interesting thing that I don't go into all the details of it is that we can measure of course, multiple individuals. Already even amongst normals, these are normal individuals, we see a variation. We see that people with disease in fact, if anything, are suppressed. And now what we can do is look for what we call systems level relationships where a person is high in one and low in another and high in another and low in another, right? We can draw these relationships according to what's called correlation ranks. And by looking across the entire system once what we can see is that an individual responds in a normal, healthy volunteer. For instance, IL-2 activation of a naive cell through phosphostat five gives a negative relationship to other pieces of the immune system. Whereas in RA patients, what we see is that entire negative relationship, that control node is completely ablated, right? So we find this in SLE, we find this in a couple of other inflammatory diseases as well. But what it lets us look at now, and this is the wrap up, is that we can now look across, this is just the entire immune system pulled out into these little circles, much like what you saw before in those circles I showed you previously in the tree structures, which is reorganized in a different way. We're now able to show that changes that happen in T cells are reflected by other changes that happen in other cell types. The yellow that you see here is actually what differs between normal and RA. That entire subset of the network that normally exists within the immune system is completely ablated in these individuals. Their immune system is just not working right any longer. But only what's interesting is that subset of the immune system relationships is missing. The rest of the immune system that in blue is acting quite normally. This now points to hypotheses and approaches that we can take to actually correct RA and as we see already, although I don't have time to show it, we see pieces of these yellow destructive elements of the network coming and going as disease remits and gets better. We now have the ability to look at immune system processes at a level of resolution that we couldn't previously using the CyTOF. We can profile the immune system right down to 0.1% cell types and the responses, the rare responses in those cell subsets in a manner it was never accomplished or accomplishable before. We use a variety of computational approaches and hardware that we've designed to do this as well as online systems that we've actually set up now for work in the clinic. So these are HIPAA compliant systems to give immediate feedback to the clinicians so that they can read out the responses from the materials that they're providing us through a software that was spun out of the laboratory. And the objective is for us to understand not just the inside of the cell biology, the intracellular biology, but the inter-transcellular networks that are existing across the immune system that we can use to understand the clinical outcomes for some of these patients. So the people in the laboratory who did the majority of the work on the mass cytometer are Sean Bendel and Aaron Simmons. Those in blue have contributed to a variety of other approaches, both computational and otherwise, that I obviously didn't have time to talk about today. And one of the things that we're most excited about here now is that not only can we do protein at the single cell level, but we've now enabled the ability to measure RNA at the single cell level down to five copies per cell. So now we'll be able to merge both proteomics and genomics, as well as picking up DNA mutations at the single cell level to merge with the biology that I showed you today, along with hopefully epigenetics. Okay, that's it, thank you. The preceding program is copyrighted by the Board of Trustees of the Leland Stanford Junior University. Please visit us at med.stanford.edu.