Scientist Stories: Peter Campbell, Somatic mutations in normal tissues

Hi there. My name's Peter Campbell. I'm a group leader at the Wellcome Sanger Institute in the Cancer Aging and Somatic Mutation Program. Recently, we've been doing a lot of work looking at the patterns of somatic mutations in normal tissues, and that's the subject matter that I'm going to cover today. I'm going to cover a little bit about the background, how we identify somatic mutations in normal tissues, and then the patterns of how these mutations change over the lifespan. And then I'm going to drill into a few stories looking at the— how we can use the somatic mutations in normal tissue to uncover patterns of clonal and stem cell dynamics across different organs. Some unpublished data in prostate I'm going to drill into in detail, and then I'm going to skim across a few other organs. So when we sequence non-malignant tissue, so normal tissue, we have to confront the fact that normal tissue is, in fact, polyclonal by and large. And so this image on the left here represents about a square centimeter of skin. And this is a kind of snapshot of the clonal structure of skin with many different clones in really quite a small area of skin. And if we were to take that entire square centimeter, grind it up, and make it into DNA and sequence it, then we would be sequencing molecules of DNA from all of these different clones, and it'll be very, very hard to find the somatic mutations. So what we have to do is drill down into small, much smaller areas of tissue. And the way that we do that is we use laser capture microdissection in this protocol shown on the right. So we would take a region, in this case of liver, take a biopsy and tissue section, and then we use the laser capture microscope to isolate individual regions of tissue, and then sequence those. And we can make sequencing libraries from as few as 100 or 200 cells. And because in many solid tissues, cells don't migrate that far after they divide, that means that we can essentially sequence little clonal patches. So rather than sequencing this entire square centimeter of skin, we're sequencing less than a square millimeter of skin, and that allows us to isolate individual clones. From that data, we can then look at the— reconstruct phylogenetic trees, which is what I'm going to mostly concentrate on today, but also looking at patterns of mutations, mutational signatures, and whether there are mutations that are changing the biology of the cells. So I'm going to show a few phylogenetic trees during the talk today, so it's worth just covering the basic principles of what a somatic mutation phylogenetic tree looks like. So when we think about somatic cells, those cells can trace their— lineage all the way back to the fertilized egg, and through a series of mitoses, somatic cell divisions. And each of those cell divisions can be associated with the accumulation of somatic mutations. And then because of the nature of the genome, each of those somatic mutations will then be passed on to all daughter cells forever after of that original cell that got the mutation. And we can use the patterns of sharing in mutations between different cells to work out what the lineage relationships of those trees are. So if each of these colored dots is a mutation, then we can look at the pattern of sharing. For example, this dark green mutation is shared by these two daughter cells, and that means that those two daughter cells shared a common ancestor. And from that, we can build up these kind of family trees that show the relationships amongst the cells that we've sequenced. So using this approach, we have sequenced now somatic mutations in normal tissues across a range of different organs. And there are some general themes that emerge. And the first of those is that by and large, mutations accumulate linearly with age. So as we go through life, all of our cells and pretty much all of the tissues that we've looked at accumulate mutations at a pretty steady linear rate. So here are data from 3 organ systems, esophagus, non-smoking lung, and endometrium. And you can see that with increasing age of the patients, there's a tendency to increase the mutation burden over time, and that relationship is linear. What that means is that each cell accumulates a certain number of mutations per year, and that rate varies from tissue to tissue. The lowest is in blood and testis. And in blood, we might see a rate of about 15 mutations a year. Highest rate we've seen is in colon. That's more like 30, 40 mutations a year. Mutations a year per cell. But essentially, every cell in our bodies is mutating at least once every week or two, which is really quite striking rates. That, of course, from the point of view of reconstructing clonal dynamics, is quite useful because those mutations are essentially marking each cell every few weeks. And that we can use to really make quite high-resolution inferences about the patterns of cell turnover and over life. To drill in on this in a little bit more detail, this is data from a 59-year-old man, and what we've essentially done is isolate 150 blood stem cells, grown them up in vitro, and sequenced the genomes of those 150 cells and built the tree. And what we found when we did it in this particular man is that there were 2 mutations that completely compartmentalized— split the cells. So about 2/3 of the cell had this mutation here, 1/3 of the cells had this mutation here, and every cell had one or other of those mutations, but no cell had both. So essentially what that tells us is that the most recent common ancestor of all of the blood in this person was this cell. And when it divided, one daughter cell got this left-hand mutation, and one daughter— the other daughter cell got this other right-hand mutation, and then all blood cells derive from either one of those two daughter cells. Now, the interesting thing in this gentleman was that when we looked at these two mutations in another tissue, so we had buccal epithelial cells as well, we could see that those same two mutations were present in the buccal epithelial cells from this man at exactly the same proportion. So one-third of buccal cells had this mutation, 2/3 of the buccal cells had this mutation. And what that tells us is that not only was this cell the most recent common ancestor of all blood cells, but it was the most recent common ancestor of all buccal epithelial cells. Now we know that blood derives from mesoderm, whereas buccal epithelium derives from ectoderm. So what that tells us is that the most recent common ancestor of both blood and buccal epithelial cells must have predated gastrulation. And in fact, we think that this is probably the fertilized egg, this most recent common ancestor. And therefore, this cell division here represents the first cell division of the zygote. And these further branches down here must represent further stages of cell division of the embryo. And so really, what this tells us is that somatic mutations begin to accumulate from the moment we become— the moment we're a fertilized egg, essentially. And that then gives us this very high-resolution way of reconstructing lineages and lineage relationships amongst cells, which is what we've exploited in the data that I'm about to show you. So what we wanted to do is to look at patterns of stem cell dynamics in a range of different tissues. And to do this, we chose prostate as an interesting organ to start with. The reasons for doing that were that the prostate is, I mean, it's interesting medically. It's an organ that really imparts a lot of morbidity to the aging male. And much of that morbidity is driven by changes in stem cell dynamics. So cancer and benign prostatic hyperplasia. Wanted to kind of look at what we could say about dynamics of stem cells in the prostate gland. The other reason for looking at prostate is that it's quite small, it's about the size of a walnut. So we can essentially take a slice through an entire prostate and put it on a single microscope slide and then essentially capture the whole section of the prostate. And so what we did, what Sebastian Grossmann, who's a PhD student in my group, did, was to take an entire normal human prostate from a 59-year-old man, and basically section it all the way through the prostate gland. And he wound up with about 800 sections. And each of those sections were put onto a microscope slide, stained, and imaged, and then in this way here, and then he basically from that tracked individual ductal units. So what a prostate looks like microscopically is a series of ducts. They branch from the urethra, so there's about, I think, 24 to 30 individual ductal units, each of which starts at the urethra and kind of invades into the prostate tissue, progressively branching. And what Sebastian did was to follow individual ductal units all the way from the urethra through that pattern of branching out into the periphery of the prostate. Prostate. And then having tracked these individual ductal units, he took a series of microdissections along those ductal units that were then sequenced, and we wound up with about 400 whole genomes from this one prostate, and he concentrated on 2 specific ductal units all the way through their branching pattern. So when we look, prostate, like the other organs, has a linear increase in mutation rate with age, and that means that we can kind of time some of the key events that happened during the development of this prostate. So this is the kind of data that we get. So this represents a heat map of mutations. So what we have is each of the columns represents an individual microdissection from one beta-ductal unit. And then the rows here represent clusters of mutations, and the color here is whether those— the kind of cellular fraction of cells within that microdissection that carry that cluster of mutations. So what you can see here is that there are kind of clusters of mutations like this one here at the top, which are seen in very many microdissections. And here on the left, this heat map here shows the number of mutations that are present in that cluster. And so you can see this cluster here has not very many mutations in it, but those mutations, those few mutations are spread across a very large number of the microdissections in that ductal unit. And then you can see this kind of nested structure underneath. So here's this cluster of mutations, and then there's a subcluster of mutations here, which is seen in a subsample, a subset of the microdissections that the original cluster was seen in. And then further down, there's another cluster of mutations here that's a subset of these ones. And then down here, there's a further couple of subsets. And what that means, so this nested hierarchy basically, it essentially allows us to reconstruct phylogenetic tree that tells us the relationship. So what this tells us is that these mutations here at the top are the earliest mutations.. And then they were followed by this next set of mutations, then followed by this next set. And each of those sets of mutations is nested within the other and is present in fewer and fewer microdissections. So that allows us to reconstruct a phylogenetic tree, and this is the tree that we get for the two different ductal units. And there are some really interesting features of the tree here. So what you see here, for those of you who are not familiar with looking at phylogenetic trees, basically the distance along the x-axis represents the number of mutations in that branch.. And where you have a shared branch here, what that means is that these two clones here, each of these open circles represents a kind of final clone that we've identified in the prostate. What this shared branch here tells us is that these two clones here shared this set of a few hundred mutations, and then they branched and and kind of evolved in parallel. So what you can see when you look at this is that the closed circles here represent individual branch points in the phylogenetic tree. And when we're talking about somatic cells, those branch points represent cell divisions of somatic cells. So a daughter cell— a mother cell, sorry, making two daughter cells at that branch point, and then those two daughter cells essentially evolving independently. Over time. So each of these branch points represents a historic somatic cell division. And what you can see is that there's a sort of a group of these branch points with the filled circles very early on in molecular time, less than 50 mutations, certainly 10 to 20 on average. And that, because we know that of the— we know that mutations accumulate linearly with time, we believe that those mutations are occurring during embryogenesis, those first 10, 20 mutations. So this set of branch points in the phylogenetic tree at very early molecular time represent cell divisions occurring during embryogenesis. Then you can see that there's a wave of coalescences here, branch points here and here, occurring at around about 300 to 500 mutations of molecular time, really quite a concentration of them over quite a small period of time. And we, again, we know from our rate estimates of mutation rate that these, the timing of coalescence is around about 300 to 500. Mutations in time must time to puberty, so around about 12 to 17 years of age. And this represents a kind of a burst of cell division activity associated with androgen, the androgen burst in puberty driving maturation of the prostate. Now, because Sebastian had done this microdissection in a very careful way and traced these individual ductal units, we actually know where each of these clones sits, and we can now look at not when are these clones dividing, but we can actually drill into what's happening in the sort of geographical— what's the geographical scope of these clones. And so if we— so essentially what we do is each of these, this is a sort of two-dimensional— of course, these ducts are branching and spreading in three-dimensional space, which doesn't make for very good presentation on a two-dimensional screen. So what we've done is we've projected the three-dimensional branching structure onto a two-dimensional plane here, and that's what you're seeing here. And this is one of those ductal units, and basically everywhere that you see a circle is a region that Sebastian microdissected, and then the lines between the circles represent the ducts that are connecting those regions of microdissection. And so you can sort of see in this particular ductal unit the really very complicated branching trajectory of this branch as it starts at the urethra marked with the asterisk here, and then as it branches into the periphery of the prostate. So each of these circles represents an individual microdissection. I'm going to show a few of these over the next few slides. And then what we've done is we've colored the microdissections by whether they carry a particular cluster of mutations., and the intensity of the color represents what fraction of cells carry that set of mutations. So when we look at the embryonic clones, so these are the ones you'll remember that occur really early on, and here in this particular ductal unit are all of the different embryonic cells. What you can see is that, is that each of these embryonic cells kind of has spread through quite a wide range of the prostate of this particular ductal unit. So this particular embryonic cell that started here and has undergone many, many cell divisions has wound up seeding all of these different microdissections within the same ductal unit. And you can see that essentially we can account for pretty much all of the microdissections from about 5 to 10 of these embryonic cells. So that tells us that every ductal unit in the prostate is seeded by only 5 to 10 embryonic cells. And the scope of each of those embryonic clones is mutually exclusive. In other words, what you tend to see is that a given microdissection, by and large, has contribution only from one embryonic cell. But there are some discontinuities. So you can see that there are patches here of the pink cell, and then quite far-flung other regions that are seeded by the same embryonic cell, with many regions in between that are not seeded by this particular embryonic cell. And occasional regions, such as this one and this one here, show contributions from more than one embryonic cell. And those regions that are seeded by, or show contribution from, two or more embryonic clones tend to be somewhat closer to the urethra than the ones that are seeded by only one embryonic clone, which tend to be right in the periphery of the ductal tree. So when we now look at those pubertal clones, we can see that puberty is associated similarly with kind of entire— so we know that from morphology studies that puberty drives formation of new side and terminal branches. So you get a rudimentary ductal structure laid down in embryogenesis, and then during puberty, puberty, that gets increasingly complex. There are new ducts formed from that existing structure. And what this shows, what our data show, are that each of those new terminal branches, those new side branches, are seeded by single clones present locally at adolescence. So you can see here's a set of coalescences or branch points that occur during puberty. And what you see is that the scope of those— the geographic scope of those clones is different. So this entire side branch here is seeded by one of those cells created in adolescence. And this other side branch here that's right next door to it is seeded by another one of the daughter cells created in adolescence. So that allows us to say that these ducts are formed by single cells present locally at adolescence, and then kicked into action by the burst of androgens in puberty. Now when we look at adulthood, we find that actually the scope of clones acquired during adulthood is really very limited. So here are a series of adult clones, and they're really not spread across very many microdissections at all, suggests that during adulthood tissue may maintenance is pretty quiescent. There's not a lot of cell division going on. What clones there are don't tend to spread very, very far, and cell division, cell turnover, must be quite limited in adult prostate under normal conditions. There was one exception to that, and this was a clone where we did see a series of branch points in adulthood, where we could absolutely time these to adulthood, and that clone had expanded over quite a wide range of the prostate.. And what was striking about this particular branch was that it was the only, the only out of 400 whole genomes, was the only one to contain what we would recognize as a classic driver mutation that you might see in prostate cancer, a hotspot mutation in FOXA1. And so it is tempting to suggest that this kind of wider geographic scope and greater expansion of an adult clone is initiated by this driver mutation. But overall, in normal prostate, driver mutations are relatively rare. Interestingly, although there was— this clone had a driver mutation, the morphology of these ducts was relatively normal. Certainly, there was no sign of this being actually a prostate cancer or even an intraepithelial neoplasia. So to summarize the prostate data, what I've shown is that each of the 24 to 30 ductal units in prostate are laid down by 5 to 10 embryonic cells. That rudimentary ductal structure laid down in embryogenesis is then quiescent until puberty when there's a kick of androgens that drives the formation of new side and terminal branches. And that new duct formation is generated monoclonally by local stem cells that are already scattered throughout that ductal tree. And so the androgens act locally to induce that new duct formation. In adulthood, tissue maintenance is undertaken by local progenitor cells, leading to limited geographic scope of clones, and driver mutations are rare, but when they do occur, would seemingly be able to trigger more extensive clonal expansion that respects the ductal tree. So I'm just going to quickly skip through this sort of— this kind of approach in a range of other organs, and just to give you a sense of the kind of different patterns that we can observe. So when we look at colon, for example, when we look at normal colon, —this is this process of microdissecting individual colon crypts—what we find is that in healthy colon, the phylogenetic tree is very lacking in branch points. And what this suggests is that essentially once each colon crypt is laid down, it pretty much acts in isolation from all its neighboring crypts. Any kind of spillover or division of crypts into multiple crypts is pretty rare in healthy colon. In contrast, when we look at diseased colon, in particular inflammatory bowel disease— and this is a study that was just published earlier this year— what we find is that in the presence of inflammatory disease, there's a lot more structure in the phylogenetic tree, a lot more branching in the phylogenetic tree, suggesting that when the colon is is inflamed and the crypt structure is damaged and then repaired and regenerates, that's when you begin to see extensive branching. So what that tells us is that stem cells in the colon have this massive latent potential that basically isn't tapped during healthy normal states. But in the presence of disease, that recovery can be can be really quite extensive, and you can get quite large stretches of colon all deriving from a single ancestral stem cell. There's a similar sort of pattern that we see in liver. So when we look at healthy normal liver, actually the kind of— the liver is— the parenchyma is essentially a patchwork of clones, all very small, no more than 100 to 1,000 cells or so, with very little branching in the phylogenetic tree, as you can see here. But when we look in the presence of liver disease, and in particular cirrhotic liver, what you see is much more branching in the phylogenetic tree. And then these clones that extend here, these kind of colored oblongs or rectangles here, are essentially patches of clones, of a single clone that's spread over quite a large distance. So in the presence of liver disease, again, we can get large patches of liver that are derived from a single ancestral cell. So the presence of disease unmasks this huge latent regenerative potential in liver stem and progenitor cells. In esophagus, what we see is a really quite remarkable finding. So this is work from colleagues of mine, Inigo Martin-Corina and Phil Jones, showing that that with age we see increase in— this is an image of what esophagus looks like in 9 individuals as they increase in age, and each of these circles represents a clone. And you can see that with age, the size of clones increases dramatically. So the clones are much bigger in older people than in younger people in the esophagus. And the presence of colored circles basically represents clones that have driver mutations. And so you can see with age that not only is there an increase in the number of clones and the size of those clones, but also a massive increase in the number of fraction of cells that carry driver mutations, and many cells carrying more than one driver mutation. So in this case, that sort of clonal expansion is associated with selective advantage for clones carrying particular mutations. In genes that are known to affect squamous epithelium differentiation and division. Finally, in lung, what we see is that there seem to be two populations of cells. So particularly when we look at in the setting of smoking, and people who've stopped smoking, what we find is that there are— populations of cells that have elevated mutation burden in ex-smokers, and then a population of cells that have normal mutation burden in ex-smokers. And this population of normal cells is polyclonal, as you can sort of see here. This is the population of cells with a normal mutation burden in an ex-smoker. Here are the cells with elevated mutation burden. And you can see this population of cells with normal mutation burden is polyclonal. And what that suggests is that there's a stem cell population in bronchial epithelium that essentially lives in a protected niche and escapes the mutational consequences of tobacco smoking. And when patients stop— when people stop smoking, that population can expand and gradually repopulate the bronchial epithelium with cells that essentially have escaped mutational consequences of somatic mutation and are essentially normal. So when, I guess, to draw some of those themes together, I think the key findings from this data are that spontaneous somatic mutations enable us to perform lineage tracing, and that makes, enables us to make inferences about the patterns of clonal dynamics and stem cell dynamics across a range of tissues. Clearly, the stem cell dynamics and numbers are different in solid tissues than they are in blood, where much of our inferences previously have come from. And those stem cell dynamics in solid tissues tend to be shaped and constrained by the physical structure. And I've shown data today from squamous epithelium, like the esophagus, versus glandular epithelium, like the prostate, versus colon crypts. And I didn't show it, but endometrial crypts. Glands as well. And so, and these, and the particular kind of physical structure that those clones are working in do shape that, the patterns of clonal expansion and spread that we see. We see really different patterns across these different types of tissue. Furthermore, disease such as the inflammatory bowel disease or cirrhosis toxicity, such as the cigarette smoking and alcohol in the context of liver, driver mutations like we see in endometrium and in esophagus. All of these things can unmask a huge latent regenerative potential in many tissues. And that potential basically isn't accessed— seemingly isn't accessed under healthy conditions. But when disease, toxicity, and driver mutations occur, then that can trigger a regenerative potential or a clonal expansion potential that is otherwise untapped. What's interesting and is, of course, a fundamentally important question for stem cell biologists is what fraction of cells in these solid tissues carry this regenerative potential? How many cells are there in the liver that can respond in that regenerative way to create entire clones or serotic nodules. We just don't know what that fraction of cells really is. So that's where I want to finish. I won't rehash all of these points, but I will thank a number of people who've been involved in this work, particularly Sebastian Grossmann and Yvette Hooks, who worked tirelessly on the prostate data, along with our collaborator in Newcastle, Rakesh Heer. The data in the other organs was generated by the people listed here. Kanichi did the lung work. Henry did the blood and the colon work. Simon did the liver work. And Kate helped Kanichi with the lung work. Siggy did the inflammatory bowel disease work. And my senior colleagues listed here, who really have been fantastic collaborators and mentors for me. Thank you.