Scientist Stories: Sarah Teichmann, Mapping the human body one cell at a time

The human body, one cell at a time. That is what the research of the Human Cell Atlas is about. Each of you has different cell types that work together to form the tissues in your body, and each tissue has a different cell composition, as you can see here for the upper airway, the lower airway, the thymus, and the heart, for example. Each of those cells contains exactly the same DNA, pretty much, to a first approximation. And yet they look very different. They have different sizes, shapes, and different functions. And the reason for that is that there's an invisible machinery inside each cell that generates a different set of genes that are active in diverse cell types. So you can see fibroblasts on the left that are structural cells in the body and have one subset of genes, one transcriptome that's active in that subset that generates the proteins that determine its function. And the neuron on the right that has a very different subset of genes that are active that carry out neuronal functions. And so measuring that subset of genes in each single cell in a given tissue is a very powerful technology, and that technology is called single-cell transcriptomics or single-cell genomics. And it's that single-cell transcriptomics at scale that's one of the pillars of the Human Cell Atlas. Project because it allows us to see or characterize each single cell in a more comprehensive way, arguably, than with a microscope. And so the— this resolution revolution in genomics that's happened over the past 10 years or so allows us to take cells in suspension from a tissue, isolate the individual cells, and then sequence the entire transcriptome that's active in that cell or other layers of information like the epigenome or genome. And it's the ability to do that at scale that's catalyzed the Human Cell Atlas consortium, and that was one of the key technologies that was at the root of this work when we started the project over 5 years ago now. So the mission of the Human Cell Atlas is to generate a comprehensive reference map of all the cells in our body, which we believe has implications not only for a basic understanding of the cell biology of human tissues, to better understand what the cells are in our body, but also in applications in terms of diagnosing, monitoring, understanding, and treating disease. And I'll talk a bit about that throughout this presentation. So the history of this project is rooted to some extent in Sanger, in the Cellular Genetics Program. When I joined in 2016, or actually before I joined, I reached out to Aviv, who was at the Broad Institute at the time, and we were both very much aligned on our vision to start this project, and that it was the right time 5 years ago roughly to kick off a larger initiative with the aim of mapping the human body. We organized the first meeting in London in 2016, down at the Wellcome Trust, and now in the meantime the community has grown, you can see, to almost 2,000 members up here, spread across the entire world. And very importantly, I'd like to point out the link down here. So for any scientists that are listening and that are interested in joining this project, there's a link down at the bottom to invite you to join the HCA, which you can do as an individual member or as a project. And, you know, the emphasis on this community is that it's a grassroots scientific community. And for everybody else who's listening from Sanger and from the genome campus, I hope you just enjoy the talk and hearing about what we're up to in the Human Cell Atlas Consortium. So the idea is to create a Google map of the human body. And using these high-resolution technologies that I've mentioned a bit about is we We want to get from the coarse-grained view of the human body as the earth, basically the anatomical view, down to the organ, if you will, and all the way down to the histological view of our cells using high-throughput, high-resolution, comprehensive genomics technologies. A little bit more about the community. I've emphasized that there are many, many members spread all across the world.. And most recently, we've started over the past year or a bit more, there's been an equity working group that's emphasizing equity and diversity in terms of the scientists that are involved, but also the samples that we study. And this is chaired by Alex Scheelek, Parthama Jomdar, and Musun Wang. The other way that our community is organized is into biological networks. So we also have technology working groups focusing on the experimental and computational technologies. And then we also have clusters focused around particular organs or systems, and you can see the different organs and systems that are covered here. It's getting relatively comprehensive, so the project has grown quite significantly over the last year. And what I'll come back to later in the talk is work that was kickstarted by the Lung Biological Network community. You can see the chairs here: Pascal Barbry, Sasha Micharn, Maite Nowén, and Jay Rajagopal in France, the US, Holland, and the US again. And that will be to do with some of the COVID work that I'll talk about. We also have a very active and exciting development community, a group looking at organoids as models of human tissues, and then groups dedicated to genetic diversity. So I've mentioned single-cell genomics, and you can think about this as a genomics technology that gives us genomic information on the individual pieces of fruit that are inside the fruit smoothie. As contrasted to conventional bulk genomics that basically looks at the ensemble of cells as an average in your smoothie, the single-cell genomics gives you the selection of the chunks of fruit. But then of course what we'd also like to understand is the way those chunks of fruit are arranged in terms of the detailed architecture of our tissues. So how do those different cells come together to form a functioning brain, heart, thymus, lung, etc. And so that's really where the spatial transcriptomics technologies come in, is putting together the pieces of that cake, putting together the pieces of fruit back into that jigsaw puzzle, how they fit together. And so it's the combination of the single-cell genomics and the spatial technologies that will get us to this Google Maps view of the tissues and allow us to zoom in to the three-dimensional architecture of cells and their relationships in tissues. And, you know, having that comprehensive and high-resolution understanding seems incredibly exciting to me in and of itself. To you, you know, you may ask yourself, why does Sarah spend so much of her time unraveling the cells and their relationships? So it's a fun pastime, but what I want to emphasize in this talk is that that knowledge and understanding actually has practical applications in the context of regenerative biology. So in other words, iPS-derived organoid systems as models for our tissues and ultimately applications in regenerative biology, and also for drilling into disease mechanisms so that we understand early stages of disease and what goes awry in disease relative to the healthy reference map of our— the cell atlas of our tissues. So I want to emphasize basically the excitement that people like me get from a basic understanding, but also how that basic understanding can translate into useful applications in biomedical— in the biomedical world. So to give you a snapshot of examples of where we stand in terms of the project and the data today, these are snippets of datasets in the Human Cell Atlas data coordination platform for different organs. So you can see the kidney up here. We've got almost a million cells from different samples, from many different samples and distinct individuals. The skin equally, almost a million cells. The lung, of course, very important in the COVID context, over a million cells, hundreds of samples, over 100 individual people donors. The colon, the large intestine that you can see here, again, many samples, individuals. And then we've got the development community, which obviously encompasses many different organs, and that's why there are over 4 million single-cell transcriptomes that we have information for from human embryonic and fetal development, and then the liver, finally the liver up there. So basically, you know, 5 years into the project, we're still in the early phases of data generation, but you can see it is ramping up and really gaining momentum. And we are starting to map many different tissues at this single-cell resolution and gaining incredible rapidly very new insights about the cells in our body. One of the vignettes that I'd like to drill into that's not on here is unpublished work from Cambridge, a collaboration between us at the Sanger Institute, so Rosa Bento's group, my group, Omar Barakter, and two groups down in Cambridge University, Margareta Turco in the pathology department and Ashley Moffett, where we've been mapping the human uterus. And the reason that I'm so excited about that is that it's a relatively understudied tissue in the human context because it's the tissue architecture is different between the human and the mouse. So rodent models don't actually give us the full story in the human. And of course, the rodent models are the ones that are accessible and have been studied in detail in many tissues. And the other challenge is that we have a dynamic tissue that changes over the monthly cycle, and so understanding that dynamics adds an additional layer of complexity beyond the 3-dimensional challenge that you would have in most tissues. So it's really a 4-dimensional question that we're trying to answer here is how— what is the architecture of this tissue and how does it change over the monthly cycle? And this is where the Cell Atlas technologies really reveal their power, because what we're able to do with these tissue samples from deceased transplant donors, working with Khorasai Parsi and his team in Addenbrooke's Hospital and the Cambridge Clinical School, is that we can get biopsies that cover the entire layer of the uterus. So that means the myometrium, which is the the layer of muscle around the outside of the uterus, and the endometrium, which is the inside layer lining the uterus adjacent to the lumen, the interior. And we can get these samples from donors that are deceased at different stages of the cycle, so menses proliferative and secretory phases, menses follicular and luteal phases. And these are the number of donors from each phase, 4 and 3 and so on. We can establish the phases using pathology analysis with expert pathologists, Louisa Moore, Leah Campos, and Ashley Moffat. The excitement on this slide that you can see here are the tissue sections shown on the top right and on the two left and right on the bottom is that these tissue sections are showing you individual transcriptomic libraries in each voxel here. So on the slide that's about 4 millimeters by 4 millimeters, we have individual libraries that cover a 50-micron voxel. So there are thousands of these in the tissue section. And what we can then do is interrogate that data in terms of what we know about the cell types from our single-cell and single-nuclear RNA sequencing data. From the same tissue samples and define the cell populations and then figure out where the individual cell populations are sitting. And so what we see here are, for instance, glandular epithelial cell markers. This is a gene that marks glandular epithelial cells, and that's high basically sort of throughout the center, this bottom or central layer of the endometrium. You can see the red is high gene expression. And then fibroblasts, those structural cells that I mentioned originally in the beginning of my talk, there's a key population that covers the top layer of the endometrium in the last stage of the menstrual cycle that we also see in the decidua. In other words, the pregnant version of the endometrium after implantation of the placenta takes place, which we showed 2 years ago in a paper on the placenta and the decidua.. In contrast, there's a separate distinct type of fibroblast that sits in the basal layer of the endometrium, which we can see here. And there's a distinct type of epithelial cells that are ciliated that we can see in the top luminal layer lining the inside, that we can see nicely defining the top layer of the endometrium, these wrinkles along the top. And we see an interesting LGR5 population, which we think is a cycling progenitor population of cells, also positioned specifically along the lining of the endometrium, the top in the lumen. So the reason that I'm emphasizing these two different types of cells is the fibroblast cells are thought of as structural cells. What we show is that they secrete key signals to the epithelial cells, the glandular and the ciliated cells, these two different populations that we're showing here, and form the the microenvironment that instructs those cells to differentiate during every monthly cycle. So basically, the fibroblasts that I mentioned, the structural cells, they turn out to be distinct in different layers of the endometrium. We show here the zonated architecture with these distinct layers of fibroblasts that define the zones of the endometrium. And it's those structural cells that we predict are sending signals to the epithelial cells to tell them to go down different cell fates every month. And so what I showed you there are spatial transcriptomics data and spatial transcriptomics tissue sections. We can also analyze tissue sections using imaging, using microscopy and single molecule FISH. And so this is work from Kenny Roberts and Omar Biroctor's group analyzing tissue sections that are taken from different stages of the cycle for different sets of markers. The epithelial cells are marked with EpCAM in blue here, and you can see the secretory epithelial cells marked by PAEP in this green image in the third row. And these green images, and what this shows is that these particular secretory epithelial cells form glands that basically become larger and more distinct during the secretory phase and then shrivel sort of by the late secretory phase. And these beautiful images show how that's a subset of epithelial cells with MMP7-positive cells forming a different subset. You can see here these MMP7-positive cells lining the lumen of the endometrium at the end of the cycle and in a very specific manner. So these beautiful microscopy images give us a different angle and a different resolution of insight into the tissue architecture in parallel with the sequencing data. So this in vivo atlas of the human uterus, you know, it's very beautiful, extremely exciting for us to gain this knowledge. And how does it actually inform us about, you know, diseases like endometriosis, endometrial cancer, and how can we use it to, in terms of diseases, engineering tissues, and so on? Well, one, one application is that we can use the information from the in vivo cell atlas to engineer cells and organoids, sort of organs in a dish, a uterus in a dish essentially. And the way I view all the Human Cell Atlas data is that it's providing us with the recipe for making tissues in vitro. And more specifically, the way I think about it is that the in vivo data tells us what the transcription factors are that are expressed in each cell type. So those are the regulators that switch on modules of genes and activate them. What are the secreted ligands that are sending signals to cells in their microenvironment? And how— and really those are the kinds of key regulators that we need to understand to replicate. The tissue in a dish. And so the endometrium-in-a-dish system has been established by Marguerite Turco and the University of Cambridge several years ago now. And using the signals that we predict to be secreted from the fibroblasts that are adjacent to the glands and the fibroblasts that are adjacent to the ciliated cells, we could modify an engineered culture condition that made those two different subtypes of endometrial cells. So we're able to kind of precision engineer the organoid and in that way also confirm the predictions from the in vivo cell atlas about how the cells are talking to each other and how the cells are influencing each other to form the tissue architecture of the uterus. So in that way, you know, the human— we've shown, and there are many other examples from other people, how the Human Cell Atlas can— is directly having a direct impact on the field of model systems in a dish and regenerative biology, making human tissues in a dish. And what I want to go on to next is give you an example of how the Human Cell Atlas is providing insights into disease mechanisms. So the top right here. So going back a year or so, we published a paper on contrasting data from healthy reference bronchoscopy samples from the bronchi of healthy donors as compared to asthmatic donors, asthmatic patients, as well as alveoli from diseased transplant donor tissue. We published this last year in Nature Medicine, and that gave us insight into the structure of the bronchi, which forms this epithelial monolayer with immune cells and stromal cells peppered in that you can see here, and the alveoli, which contain these large alveolar type 2 cells and alveolar type 1 cells among the the vessels and immune cells. So asthma, as I'm sure you all know, is a condition where there's inflammation that leads to thickening of the airway walls and then the asthmatic attack. So at a cellular molecular level, what's the difference between bronchoscopy samples from healthy versus deceased donors? We sought out to answer this question using the precision technology of single-cell genomics. And what we discovered throughout that work was changes in different cell compartments, both the epithelial cells, the functional cells in the lung, as well as immune cell compartments. And during this work, we— I should say that we also studied the upper airways, in other words, the nasal epithelium, the bronchi, and the parenchyma. From top to bottom of the airways. And one of the things that we saw in the nasal cavity was a subset of epithelial cells, of goblet cells, so these functional cells inside the nose that have a particular phenotype where they have a particular immune phenotype where they are secreting signals and talking to other cells of the immune systems, T cells and critic cells and so on. As compared to a more sort of quiescent, distinct population of goblets and ciliated cells that are shown here in the nasal epithelium. And what we saw later was that those cells in the nose express ACE2, which of course, you know, starting at the beginning of this year, is a gene that's become very prominent, let's say, not for its enzymatic activity but for its role, its sort of happenstance role, as the receptor for SARS-CoV-2. And so I want to use this to illustrate one of the ways that the healthy reference data can inform us about disease, and that is basically to map where the receptor and the protease, so ACE2 and TMPRSS2, for the coronavirus, for the SARS-CoV-2 are expressed around the tissues of our body. And this was work that was basically became a consortium effort, started off with the Lung Biological Network collaborators. So that particular community, because the airways of course are so important in the pathology of this disease, but then it really became a broader effort of the the broader Human Cell Atlas community. And what was incredibly, basically amazing about this effort was how fast people contributed data towards understanding the SARS-CoV-2 receptor expression. And so this included unpublished data from the eye, unpublished data from the heart, the liver, and these are really groups from around the world that contributed this. In this very open and very rapid way. And so we published this work earlier this year as a community paper. And I'll walk you through a few of the key barrier tissues that we think are particularly interesting in this context because of course the Human Cell Atlas data, what we're looking at here is mostly healthy reference data. It's not the infected tissue, which I'll come on to later.. And so looking at the healthy reference data, I think where this is particularly informative is providing predictions about where the virus might be entering the healthy cells at the very early phase of infection. And so I've mentioned the nose earlier and these goblet cells that are particularly high in the receptor, that's why they're dark blue. If we look at the lungs in the bronchi, the lower airway, then they are both club and ciliated cells that express the two. Key proteins. If we look deep down in the lung, in the parenchyma, the alveoli, then we see these alveolar type 2 cells, which were described already 16 years ago as expressing ACE2, which is the coronavirus receptor, using immunohistochemistry. Now in the eye, and this was data from Linda Laycombe in Manchester that she shared very, very quickly, what amazed us was that we see expression in cells of the conjunctiva and the cornea. And so this implicates the nasolacrimal duct, so the connection between the eye and the nose, as potentially involved in early infection and transmission. In the gut, we see in the small intestine enterocytes that express the receptor. And of course, there are gastrointestinal features of the disease. And the placenta, so this harks back to the endometrium that I talked about earlier. This is the maternal-fetal interface and the decidua and the placenta, the uterus and the placenta. We see expression in trophoblast cells of the placenta as well as decidua cells on the maternal side. So both on the fetal and the maternal side. And this suggests that there can be potential maternal-fetal transition of the virus But of course, we know clinically that if it happens, it's a very rare event indeed. So, in a mammoth kind of effort, collating data from around the world, Wardhan Somnack, a postdoctoral fellow in the lab, and Nhi Hoang, a computational scientist in the lab, worked fantastically with the consortium and with brilliant support from the CellGenIT team, kind of really playing to the strengths of the technical teams, the support teams in Sanger who are just amazing. And just let me say thank you to everybody on the campus who works in a role like that. You know, many of these teams are absolutely amazing in their level, their quality, and their dedication. And what we're able to do was to put together this COVID-19 Cell Atlas for the community. Now, of course, all of that data that I've referred to so far is healthy reference human cell atlas data. And what that healthy reference data provides is a hypothesis about where the virus could be entering. In order to understand more about what happens during the disease, what we really want is patient samples, patient data. And looking at interior tissues is mostly only feasible using samples from deceased patients who've died of COVID-19. And we're very lucky in the UK to have a mortuary at Westminster that allows pathology specimen collection. And so this is from Michael Osborne, Brian Hanley at Imperial, together with Michaela Maceda. And also an effort in Newcastle that's available where tissue samples are collected from deceased COVID-19 patients and then Again, Kenny and Omar have looked at, have analyzed this tissue using their multiplex FISH approach and also the new NanoString GeoMx instrument. And so what you're seeing here is a lung sample and a section where there's SARS-CoV-2 probes that are lighting up in green and AT2 cells, which are those alveolar type 2 cells that I mentioned earlier, and why you can see them peppered around. And what this histogram shows is that there are very variable loads in virus in different donors. So there's a lot of donor-to-donor variability. If we zoom into different regions from the same donor here in white and in yellow, there are some regions that have a low viral load and a lot of AT2 cells, and then other regions that have a much higher viral load. The virus again is in green, and many fewer AT2 cells. The AT2 cells are white, so presumably because they're dying of infection and disappearing. And this is where the damage really— where you can see the damage at the molecular and cellular level happening. A much broader— Omar and Michaela together, their groups collaborating with all of us, are doing a much broader survey of different tissues. And of course, the kidney, the heart, and the vasculature have been implicated in the pathology of COVID-19. And you can see here steps at interrogating what's going on in these tissues at a lot of detail. So CDH5 is a marker of endothelial cells of blood vessels. And you can see that the changes in, or what basically structures across different tissues— nose, pharynx, heart, kidney, and so on. And this is very early work, I should say. This is an ongoing effort, and there are similar pathology collections available in the US and in Italy that we're aware of, and we're coordinating across this community using the infrastructure and the Human Cell Atlas community and also the Allied Lifetime Consortium for Study of Diseased Tissues in Europe. And so with those two stories on regenerative biology and the uterus in a dish and the disease mechanistic work on COVID-19 as well as asthma, I hope I've told you— I've given you some insights into what we're up to in the Human Cell Atlas community and how it can actually be useful. And the tissues I've talked about are, again, just to remind you, the uterus, the maternal-fetal interface, and detail on the airways to some extent. Thank you everybody for listening, both on the Wellcome Genome Campus and anywhere else that you may be, and I'm happy to take questions.