Scientist Stories: Muzlifah Haniffa, Decoding the Immune System

Thank you to the organisers for this opportunity to present at the fantastic European Society of Human Genetics meeting. It's been a fantastic conference and attending it in the virtual setting has allowed many more participants, and I hope this type of hybrid setting will continue, although I must admit that it's really nice to meet people face to face at conferences, and that's been a long time since I experienced one of those. I'm going to spend the rest of the 30 minutes talking about decoding the developing human immune system, and this is really work that has leveraged the enormous progress made within the single-cell genomics and omics fields, where if we were to imagine human tissue as a fruit salad in a bowl, what this has enabled us to do is to measure many parameters at individual cell resolution to deconstruct the tissue. And the work is also very much a contribution to the Human Cell Atlas, which uses single-cell omics technologies to profile all the human cells making our body, 37 trillion cells in total approximately, in order to better understand health and to diagnose, monitor, and treat diseases. This is a global initiative involving 77 countries at the moment, more than 1,000 institutions, and more than 2,000 members, and anyone is able to join. You just need to go to the web link and be, you know, you're most welcome to join the initiative. How do you build a human cell atlas? I mean, essentially it's using the single-cell genomics technologies, not just on cells that have been dissociated from tissues, but also in situ spatial omics methods, which allows you to interrogate a small area of tissue or at the whole organ level. This is really a question of scaling from, you know, a few cells, large numbers of cells, small tissue area to large tissue area, multiple different types of information which you can then integrate to inform and, you know, reveal biological insights. And these datasets will be available for anyone to use, but also, you know, provided in the form where one can actually query the datasets. And I'll show you some of the early examples of this. And as such, the initiative really very much depends on cross-disciplinary team involvement from the biological disciplines, computational scientists, clinical specialties, and also technology development. What can we measure at single-cell resolution? You know, when the first field began, it was very much protein and also RNA, but these days you can combine the different types of parameters from a single cell, and the things that you can measure, it's just increasing exponentially in terms of numbers of individual parameters detected, but also cells that can be analysed. So, I'm going to focus on human development, but also a little bit on human skin. And what I would like to do is to give a background on the developing human immune system. This is quite complex, because it occurs across several organs in gestational time. So, you have you have these extraembryonic tissue, the yolk sac, where progenitors are made which can become differentiated into blood and immune cells. Then you have definitive haematopoietic stem cells which arise in the aortogonad mesonephros in the embryo proper, giving rise to definitive haematopoietic stem cells that then subsequently seed many organs including the liver, which becomes the dominant site of haematopoiesis until 20 post-conception weeks. The bone marrow really develops about 11, 12 post-conception weeks, and bone marrow hematopoiesis is the dominant site after 20 post-conception weeks. Then you have the lymphoid progenitors seeding the thymus and then undergoing T-cell differentiation. Naive B cells made in the bone marrow, for example, differentiating and further differentiating and maturing in the spleen, where germinal centre formation and somatic hypermutation really occurring after birth. And then you have all these blood and immune cells seeding the non-lymphoid tissues such as the skin, kidney, and gut. And here they form the peripheral tissue immune surveillance network. So how do we actually understand the developing human immune system, or how can we study this? You can't just label a cell and track it over time. What you can then do is actually study multiple organs from one embryo or foetus and then integrate that information together. And we've been very lucky to have the Human Developmental Biology Resource based based both in Newcastle and UCL that allows us to— that provides resource for researchers, not just in the UK, but also globally. And this at-scale study from multiple organs has allowed us to understand the emergent properties of the immune system and really beginning to see how the whole is greater than the sum of its parts. And that's what I'm going to do. Really focusing on 3 studies, one on fetal liver hematopoiesis, which also includes analysing yolk sac, skin, and kidney. The second with regards to T-cell differentiation in the thymus, and the third is a manuscript in press on fetal bone marrow hematopoiesis, and to try and put all of that information together. This is very much a collaborative effort, fantastic working collaboratively with many researchers that I've listed here and trying to unravel the biological insights. So, in the Fetal Liver Hematopoiesis manuscript, we studied fetal liver and also the skin and kidney, but for skin and kidney, right up to 12 postconception weeks, because thereafter you have the bone marrow contributing to blood and immune cell development as well. We knew that yolk sac contributed to hematopoiesis before the liver becomes the main site of hematopoiesis, so we profiled some yolk sac like from 4 to 7 post-conception weeks. For bone marrow, basically from when bone formation and bone marrow formation takes place, about 11 post-conception weeks. And then for thymus, from 7 post-conception weeks, which is when the thymus develops. So, we profile more than 100,000 cells from each of these organs, and the main take-home message is that blood and immune cells development is established very rapidly by 20 post-conception weeks. In the human. And this is in contrast to the mouse, where quite a lot of development still occurs after birth. The first thing to say about the fetal liver haematopoiesis is although you see the myeloid, lymphoid, and erythroid lineages, neutrophils do not develop in the fetal liver. They really come on in the fetal bone marrow, where you can begin to see the granulocytes such as neutrophils, but also eosinophils and basophils. One of the advantages of the kind of like dataset that we've generated, particularly with the fetal bone marrow, and also more recently with all the other organs, is the dataset from CySeq analysis, which allows you to measure protein and RNA simultaneously in one cell. So here we've used the total seq reagents of about 188 surface proteins, and we can then correlate the expression of RNA and protein simultaneously to give us better resolution, particularly for researchers to then subsequently isolate the cell states that we found. Now that we've also got datasets from many organs, we can now begin to put this information together, not just for prenatal development, but using data that's available through the HCA Data Portal, cord blood and adult bone marrow, to begin to understand how things are evolving across the human lifespan. So, one example here shown in the myeloid lineage where you can begin to compare across yolk sac, fetal liver, fetal bone marrow, but also now adult bone marrow. I mentioned earlier on the emergence of granulocytes such as neutrophils in the fetal bone marrow, but also the emergence of greater variety of dendritic cell subsets such as the transitional dendritic cells expressing XL-Siglec-6 and also DC3 subset. One of the things that we asked when we saw granulocytes by RNA was, can we detect them and validate their presence morphologically because they have the characteristic polymorphonuclear appearance. We use the surface markers revealed by single-cell RNA sequencing, design a FACS panel, isolate the cells for SmartSeq2 analysis to confirm their identity based on the primary discovery dataset, and also to look at their morphological appearance. This is the SmartSeq2 overlaid on the 10x dataset, confirming the identity of those cells that we isolated based on surface markers and their morphological appearances very much in keeping with what is known for metamyelocytes and neutrophils of that lineage and also basophils and eosinophils. So really, why do we get granulocytes in the fetal bone marrow? And one of the reasons we think this is, is to do with the relative abundance of transcription factors that drive neutrophil differentiation versus monocyte differentiation. And here, if you look at the GMPs from the bone marrow, some of the cells have already got the monocyte signature and others, neutrophil signature. And we know that CBPA and SPI-1 are the two key drivers for these lineages. And when we compared the GMPs in the fetal bone marrow with the GMPs in the fetal liver, you can see many more cells expressing much higher levels of CBPA and the ratio to SPI-1 being higher. And this probably is one of the reasons why you begin to sort of see neutrophils emerging in the fetal bone marrow. Another lineage that becomes abundant in the fetal bone marrow is the B lineage. And what you can see here is there being no B cells in the yolk sac, very few in the fetal liver, and the proportional representation of cells in the fetal bone marrow and adult bone marrow being very similar for the B lineage. But the types of B lineage cells are different in that in the fetal bone marrow you have a lot of progenitors, the pre-pro-Bs, the pro-B, and the pre-B, whereas in the adult bone marrow the predominant B lineages are of naive transitional memory and plasma cells. And interestingly, childhood leukemias are very much B lymphoblastic leukemia, and the presence of the kind of progenitor substrates may be one reason for why this is the case. What about BCR acquisition? So you can see how heavy chain is first acquired before light chain, and also that, you know, this BCR acquisition in terms of heavy and light chain rearrangements is accompanied by cell cycling. These cells in blue here, the percentage of cells. So there's a group of pre-B cells which are in cycling shown here, and then you have this space where, you know, as B cells differentiate, there's a transition where they cycle and then subsequent selection into naive B cells. What about the lymphoid progenitors? I'm just going to focus a little bit on the T cells to show how this is occurring across atomical space. So T cells differentiate in the thymus, and when we look at the liver progenitors and the thymus progenitors, you can see some of these cells which are annotated as lymphoid progenitors independently in the two datasets. And the ones in the thymus form the trajectory that differentiate into double-negative T cells. And this is the same with the fetal bone marrow. So essentially, you've got these lymphoid progenitors from hematopoietic tissues that enter the thymus, then you have T-cell differentiation. And we showed in our paper on the thymus the different flavours and types of single positives and invariant T-cells, regulatory T-cells that develop. And what we find then in the fetal bone marrow is that these T-cells subsequently egress the thymus and then seed the fetal bone marrow because they have the features, molecular features that correlate with the differentiated single positive cells. And this allows us to sort of like put all of this information together across organs. I'm now going to focus on two insights from our studies, really on the intrinsic properties of haematopoietic stem cells, and secondly, on some of the discoveries of the niche of where blood and immune cells develops. So, looking at the intrinsic properties of haematopoietic stem cells, one, we looked at their molecular characteristics, but also combined that with functional assay whereby we isolate single progenitors, expanded them into colonies, and then perform flow cytometry analysis of the progenies that were formed. And what you can see is that the progenitors— these are all the different types of progenitors that you can find— and the relative abundance of these progenitors down the lymphoid, myeloid, and erythroid lineage changes in the organs. So, for fetal liver, there's a predisposition and enrichment of the abundance for erythroid differentiation, whereas in the fetal bone marrow, you have a greater predisposition of progenitors in the lymphoid and myeloid lineages.. And this, you know, accounts for why you have the differential composition in both organs. But also, the intrinsic potential of the HSCs is different. So, this is the functional assay where we culture single cells and then see what they make, single HSCs. So, if you took HSCs from early gestational stage liver and late gestational stage liver, what you find is that their ability to differentiate down the erythroid lineage declines, but their ability to differentiate down the B lineage and myeloid lineage increases. And If you then compare paired HSCs from fetal liver and bone marrow, you find that the properties have also changed in that the fetal bone marrow produce fewer colonies compared to fetal liver, and also fetal bone marrow HSCs have a greater predisposition to differentiate into myeloid colonies. So, this tells you these molecular changes of these cells can be harnessed to sort of guide differentiation for HSC engineering and clinical therapy. And what about the extrinsic environment? So, this study, our study is the first study to profile bone marrow stroma in humans, and this is the fetal bone marrow stroma, which we then correlated with the known subsets of mouse bone marrow stroma that had been identified, which shows very good correlation across the board. But we found really two new insights, right? One with regards to the endothelial cell subsets. And we saw two states which can be seen by the differential expression of RNA and also protein. And we annotated these cells as TECs and sinusoidal ECs. Really, they have a lot of the markers of what's been known as sinusoidal endothelial cells in the mouse. And when we look at the where these cells are present. So, this is the fetal bone marrow cut cross-section, and this is the metaphysis, the sort of like the end of the bone, and then this is the diaphysis going into the bone shaft. And what you can see is the enrichment of these endothelial cells that have high expression of CD34, the TECs here. And then nearer the diaphysis are the ones that have high expression of VEGFR2, the sinusoidal endothelial endothelial cells. And the relative distribution of these vessels correlate with the Type L and Type H mouse fetal bone marrow vessels that have been described. So, we now show that in humans, there are these two different types of endothelial cells that are lining vessels at different parts of the fetal bone marrow. And here's just to illustrate that point better. So, here are the kind of like more dilated VEGFR2 type sinusoidal endothelial cells. And then here are the kind of like higher CD34 expressing cells that form, you know, looped capillary loops at the metaphyseal end of the bone marrow. What we can now do is to essentially use the reference dataset to map alterations and disease. And I'll show you 3 examples of this. The first relates to Down syndrome caused by trisomy 21, and we were we profiled Down syndrome fetal bone marrow from 12 to 13 post-conception weeks. What we found is that in Down syndrome, you have a higher proportion of cells of the erythroid lineage. This is partly down to the intrinsic potential of HSCs. If you took those HSCs and cultured them on methylcellulose, you will see that they form a lot of erythroid colonies relative to the non-Down syndrome HSCs between 17 to 21. Post-conception weeks. And also, the fetal bone marrow microenvironment in Down syndrome is altered in that if you compare Down syndrome with non-Down syndrome bone marrow endothelial ligands and their interacting receptors expressed on the HSC and MPPs, there is a differential alteration. So, you know, essentially a third copy of genes on chromosome, a third copy of chromosome 21 causes disruption in the intrinsic potential of cells and also the environment of the stem cells that give rise to the blood and immune cells. And if you now look at all of the different cell states in the fetal bone marrow and across chromosomes, you will see that there is indeed higher expression across the board for many genes on chromosome 21, across the lineages, but the effect is also seen on many other genes in other chromosomes and cell states. The second utility of the reference data set is how we can then map infant and childhood leukemias. I mentioned earlier on that B-cell lymphoblastic leukemia is extremely common in childhood, and what we can now do is to see what types of cell states do these leukemic cells best correlate with, and to work out what may be features that we can target for therapy to sort of help these children with leukaemia. The third is the observation that developmental cell programmes also co-opted and re-emerge in inflammatory adult diseases, and in this case it's inflammatory skin disease. So, what we did here is we looked at the developing skin dataset, both immune and non-immune cells, compared it with adult healthy skin and two common inflammatory skin diseases, which are atopic dermatitis and psoriasis. And what we then did was to compare all of the different cell states that we found in the adult skin with developing skin, but also extended the analysis to the other datasets that we had from other developing developing organs to see whether this was unique to the developing skin or whether more generalised to other developing organs. First thing to say about the vascular endothelium that we saw was the transcriptional enrichment of a rare endothelial cell subset in the adult skin, which we named V3, which seemed to correlate with, you know, vascular endothelial cells from many developing organs. And the same for this macrophage 2 subset, which is most enriched with all the developing macrophages. What was really interesting about these MAC2 and VE3 were that they were proportionally increased in both atopic dermatitis and psoriasis lesional skin, which led us to look at these cells in greater detail. So, if you look at VE3, you know, microanatomically, this looks like the cells that form dilated capillary loops. And indeed, the description of these cells, both molecular and sort of anatomical level, resemble very much the high endothelial venules in the lymph node, which is where how the immune cells kind of like enter the actual lymph node. So we wondered whether what these cells were doing was to basically facilitate the recruitment of immune cells into the skin. And we asked what were the gene expression programmes that were conserved in developing vascular endothelial cells and macrophage 2, and also diseased endothelial V3 and MAC2. And essentially, this all came down to gene programs that were involved in leukocyte recruitment and angiogenesis. And because we had these two cell types, we wondered whether they were, you know, in fact interacting, which led us to look at receptor ligands that the cells may be expressing to see if there was any prediction of statistically significant interaction between those two cells. And we came down to this relationship between— or interaction between the receptor ACKR1 and its ligand CXCL8, which you see here most highly expressed by MAC2 and BE3, particularly in disease states, but also observed in the developing macrophages and vascular endothelial cells. In disease, both atopic dermatitis and psoriasis, these cells were interacting close together with in situ suggesting that they may be mediating those effects in situ. So, what we propose is that CXCL8 produced by MAG2 interacts with its receptor ECKR1 and VE3, and this drives leukocyte recruitment and also angiogenesis during development. This is to seed lymphocytes or leukocytes in the tissue, but during disease, this recruits immune cells, which then causes inflammation.. What is still not clear to us is whether this program is just unique to atopic dermatitis or psoriasis and psoriasis, or whether it's more broadly applicable to any immune-mediated inflammatory disorders, and how it may be regulated differentially during development and in disease, which will give us, you know, new therapeutic targets potentially. So what I've done is to show how the Human Cell Atlas profiling has given us a lot of information that can be used for tissue engineering, cell therapy, to better understand childhood disorders, mapping leukemias and chromosomal alterations, how it can now begin to inform us about lifespan and ageing, and also the relevance of developmental pathways in adult pathology, which may be underappreciated. All of our datasets is available through a web portal. This is the link to it: developmentcellatlas.ncl.ac.uk, and you will see From there, you can go to the dataset for the fetal liver, skin, and so on and so forth. What you can do there is to browse the expression of genes of the different cell states. Here is the expression of MS4A1 or CD20 on B cells. And you can also look at the expression of, you know, several genes on the cell states that are of interest to you, visualised as a spot plot or a heat map. I'm going to very briefly segue into what happened during the pandemic. And, you know, we had a lot of these kind of— we had acquired expertise in single-cell multi-omics, working collaboratively in a team science framework, establishing all of this infrastructure. So we leveraged these abilities to then better understand the immune response in COVID-19 patients. And this was very much a large collaborative effort across many centres. Involving many researchers. And this really then involved many junior researchers working together collaboratively. And I think, you know, this is one example of how I think, you know, science is no longer about one person, one project, but actually a group of individuals coming together. And the progress we make could not have been achieved if we just had one or two individuals. And I think it's also important these individuals are recognised for their contribution and why we had all of them as co-first authors. And I'm trying to promote this idea that, you know, we can do science at a much faster scale and speed, but we also need to think about how we make sure that, you know, everyone involved is acknowledged. So with regards to COVID-19, I'm not going to dwell too much on what we found, but the fact that the data and the web portal is available and you can browse that through COVID-19 cell atlas at nihplus.org. And very briefly, essentially what we did was to try and understand how the changes that we saw across all of the different cell types in the peripheral blood informed us of the response to the primary site infection in the lung and the response of the bone marrow. And there were many, you know, interactions and loops that were involved. The whole thing is one intricate interplay between all of these cells, which you can read in the manuscript that was published. So, I'm going to stop now to thank my lab for the fantastic journey over the years and to acknowledge all of my many collaborators. Much of the work is very much in close collaboration with the Sanger Cellular Genetics Program, particularly with Sarah Tyceman, and our funders. And also, for anyone interested in joining our team, either at Sanger or Newcastle, please do contact me. Thank you very much.