Scientist Stories: Arnold Kriegstein, Genomic Insights into Brain Development, Evolution & Disease

Well, it's a delight, a real pleasure and treat for me to be here virtually, and of course quite delayed from my original intention last— I think it was last March. But it's wonderful to be back at Columbia, and I hope to have a chance to meet with some of you over the next day or two. The talk I want to present has to do with work we've done in the last 5 or 6 years focusing largely on cortical development in the human brain and comparing it to other model systems, and also using that as a way of gaining insights into disease. So I probably don't need to remind this audience there's a big difference between the brain and the mouse and the human. These are schematics of the progenitor cells that are found in early developing mouse, and a comparison to human shows a larger variety and a larger number of progenitor types that we'll focus in on in the next 40 minutes or so. So to begin with, during early cortical development, there are two progenitor zones in most mammalian brains. And for example, in the rodent as shown here, there are radial glial cells that line the ventricle in red because they're stained for Sox2, and then their daughter cells, which are intermediate progenitors or basal progenitors, shown expressing Tbr2 in a band above. And these form the ventricular and subventricular zones. Early human cortical development looks very much the same, with radial glia and intermediate progenitors arranged very similarly. But then in the subsequent weeks, the outer subventricular zone forms in the human brain, which is a structure not found in rodents. This is this enormously expanded subventricular zone. And toward the end of neurogenesis, it's the primary focus of neurogenesis, while the ventricular zone becomes much smaller. So we and then others started looking at the progenitors found in this primate expanded zone. And there were several that I want to focus in the course of this talk, but I want to begin with one that looks like a radial glial cell, shown here left, sort of morphologically with a fiber that goes up to the cortical surface, but a cell body that's removed from the adhesion belt at the ventricle. It sits in a zone known as the outer subventricular zone. And if you backfill cells shown here on the right with dye at the surface of the brain, these cell bodies are stained in the outer subventricular zone, but otherwise look very much like radial glia. So we call them outer subventricular radial glia-like cells, or ORG cells, or cells for short. And they've also more recently been sometimes referred to as basal radial glia. We're lucky at UCSF to be able to treat some donated samples the way we do our, our mouse. And so this is an example of a section which has been stained with a viral vector only labeling progenitor cells. And this is a slice culture time-lapse image for almost 2 days showing the radial glial-like cells, these outer radial glial cells, and how dynamic they are, and also how radial they are. And one of the things you may not have noticed is how these cells divide, which is a very characteristic feature of these outer radial glia. So if you look at the cell here in the middle, for example, when it gets ready to go through mitosis, the cell body translocates up the basal fiber prior to actual cytokinesis. And you can see that, for example, on this one cell on the right. When it gets ready to divide, the nucleus will jump along that basal fiber just prior to division— just did so then. So that's a characteristic feature that hadn't been described in cells before. We call it mitotic somal translocation, and it differs from the interkinetic nuclear migration that ventricular radial glia go through. And so we wanted to look at the mechanisms, the motors that actually drive this behavior. And I predicted it would be microtubule motors. So the first experiment we did was to block microtubule with nocotazole, which we predicted would block the jump. And so here on the right is nocotazole-treated outer radial glial cell. And as you may notice, the control jumps on the left, but the treated cell on the right jumped even farther than usual. And then having jumped, it then arrests in mitosis, which you'd expect, because the microtubules are required for spindle formation by an actual cytokinesis. It turns out, to make a long story short, that it was the actomyosin motors that drive the jump, and that's documented here where blebbistatin is shown on the right to block actomyosin, and while control cells jump and divide, the treated cells fail to jump even though they subsequently undergo mitosis. So this allowed us to segregate the jump from the actual cell division and demonstrate that the MST jumps require actomyosin, not microtubule motors. We'll return to this jump and mechanism a couple of times over the course of the talk because I've chosen this as an example of a behavior that we've been able to analyze genetically. So our newest working model for human cortical neurogenesis is that there are ventricular radial glia that give rise to the outer radial glia, that go through multiple asymmetrical self-renewing divisions to produce these transit-amplifying cells that generally make clones of neurons that are of the same type and that ultimately form these cortical columns that are typical of the cortex. At the end of neurogenesis, the radial glia start to transform into astrocytes, and the different subtypes of radial glia seem to produce different subtypes of astrocytes, and I'll touch on that as we go along. So my lab has been using single-cell RNA sequence techniques to disentangle the cellular composition of developing human brain. Our initial study, quite modest in retrospect, was about 350 cells which we captured on these micro-well plates Fluidigm produced at the time, and we could capture up to about 60 or 70 cells per microchip. Once they were captured and lysed and libraries were prepared, we could look at the individual RNA sequencing of all those captured cells. And so over a course of different ages and time and multiple samples, we had a collection of about 350 cells that were single-cell sequenced, shown here, and clustered bioinformatically into cell types and color-coded for radial glia, intermediate progenitors, excitatory neurons in blue, and inhibitory cells in black. So in retrospect, a group of just 350 cells seems pretty modest, but it was sufficient to actually give us quite a number of insights into brain development. For one thing, we were able to record new or develop new markers to identify these cells. So on the left are new markers that we found for friends, that is the ventricular radial glia that we knew about already, and they're shown here on the left validated with in situ hybridization. On the right, a little bit more interesting to me, are the markers that we discovered for the outer radial glial cells, and these were novel and confirmed here by, you know, by in-situ hybridization and also by labeling cells after we'd shown that they jumped and divided or had other features in situ that resemble these outer radial glial cells. So we confirmed that these are now novel markers for the outer radial glial cells. That led to several insights that I want to summarize in the next few slides. First, a marker like HOPX, shown here in red, stains the cytoplasm of only the outer radial glial cells, and they're shown here in red. But we co-stained this section for markers of the ventricular radial glia in green using a marker called CRYAB. So the CRYAB cells, which are the ventricular radial glia, didn't extend all the way up to the top of the pia as we'd expected them to. The outer radial glia, which don't connect to the ventricle, in fact sent all the red fibers to the cortex, suggesting that we had a disconnected radial glial scaffold. So to confirm that, we used dye eyes, a different approach, and put dye crystals either on the surface or the ventricular surface of the core of these cortical slabs. And in early stages, it looked as we'd expected, that the radial glia at the ventricle ran continuously up to the cortex. They formed— I'm sorry— a continuous radial glial scaffold. But then after a week or two, at the critical stages of gestational week 15 to 18, when the lower cortical layers are formed and now we've switched to upper cortical neurogenesis. At that transition, the dye crystals at the surface only labeled outer radial glial cells that ended in the outer subventricular zone, while dye crystals at the ventricle labeled the ventricular fibers that ended in the outer subventricular zone. Only a few of them made it all the way to the cortical plate, and this confirmed that discontinuous scaffold that our marker expression suggested. And we looked more carefully at the cells at the outer subventricular zone that projected to the cortex, and you can see their cell bodies in magnification here. These are characteristic outer radial glial cell bodies. The fibers from the ventricle ended usually on blood vessels and often made these right-angle turns, and so we defined them as a different subtype of radial glia. So there are ventricular radial glia, outer radial glia, and now these truncated radial glia which start to emerge in the second half of cortical neurogenesis. And these fibers never make it all the way up to the cortex. These cell morphologies have been seen before in, for example, these classic Golgi studies that Pasko Rakic studied in mouse— sorry, in monkey. And so all the forms that he's described here correspond to the forms that I've shown you, but our interpretation is a little different in terms of what's happening over time and how these different subtypes merge one into another during lineage of the cortical progenitors. And then to summarize, putting this all together into a model, this is our current working model of human corticogenesis. It begins with ventricular— with neuroepithelial cells that turn into ventricular radial glia, and they generate those intermediate progenitors that make neurons. But then about halfway through cortical development, when the deeper cortical layers have already formed, there's a wave of outer radial glial generation, they come from the ventricular radial glia that delaminate and jump up into this outer subventricular zone. And then during the rest of neurogenesis, when the upper cortical layers, the supragranular layers, are being formed, they're generated from primarily the outer radial glial cells, and they migrate along outer radial glial fibers. Meanwhile, the ventricular cells transform into these truncated radial glia that I mentioned just a little while ago. So one of the implications of this model is that the upper cortical excitatory neurons have a different lineage than the deeper cortical layers. And this actually fits well with ideas that have been around for over 50 years when Marin Padilla and others noted that in primates, the upper cortical layers are very different than they are in other mammals. They have a larger density of smaller, maybe more diverse excitatory neurons. And so that supragranular layer distinction, which is thought to be primate-specific, may relate to this lineage difference that I'm highlighting here, where the deeper cortical layers mostly come from ventricular radial glia, the upper cortical layers are coming from outer radial glia in primates. Other insights that emerge from that initial study of single-cell sequencing are disease-related, and I want to touch on two of them. One is this disease called lissencephaly. A normal brain is shown here on the left with this highly folded surface cortex, There's a form of lissencephaly shown here in the middle where there's a single gene mutation that produces a smooth rather than folded cortex. And then on the right is a much more severe form of smooth cortex or lissencephaly, also associated with microcephaly—the brain is smaller—and that's the result of truncation of this short arm of chromosome 17 with over a dozen genes that are deleted. We managed to get tissue samples from 3 patients that have a severe form of lissencephaly And Marina Berstein, a postdoc at the time in the lab, generated cell lines from all of these pluripotent stem cell lines, and from those she created organoids and then compared normal to these Miller-Dieker lissencephaly patients. And in particular, we looked at the progenitor cells, and we found the radial glia shown in red, the intermediate progenitors shown here in green, in normal and in Miller-Dieker cases to look very similar. But when we focused on the progenitors over time We first were gratified to see that both control and the sensephaly organoids showed green gene expression in this heat map that we'd identified previously as representing outer radial glial cell genes. And when we mapped the hub genes in these modules of outer radial glial gene expression from fetal cells on the left to our organoid cells on the right, they were really very similar, and many of the hub genes were identical. And so we looked for outer radial glial cells in the organoids. And here's one. This is stained in slices and treated in vivo, imaged under time-lapse microscopy just like our fetal samples. And you can see cells that look exactly like the outer radial glia we found in primary tissue. They behave the same way. This is a looped film, of course, but they jump and divide just as I showed you earlier with the same parameters that we saw in vivo. We then looked from our control into our Miller-Dieker patients and found that the ventricular radial glia were behaving dynamically exactly the same way. They divided in the same time course. They spent the same amount of time in each of the cell cycle phases. But then we looked at the outer radial glia and observed that in the Miller-Dieker samples, unlike the controls, these outer radial glia jumped much farther than normal, and once they jumped, they failed or rested in mitosis. And this is, as you may recall from earlier, the exact phenotype that we saw when we blocked microtubules with nocotrozole. It makes a lot of sense because we know that the LIS1 gene is a microtubule-associated protein, so it's a microtubule defect. But what this study shows us is that the defect was cell-type specific. It affected the behavior of the outer radial glia, not the other progenitors, and it's a finding that couldn't have actually been performed in the mouse, and it highlights how for certain diseases or certain developmental events. If you want to learn what the human brain is like, you're not likely to be able to see all these features in animal models such as the mouse. And another disease that we didn't expect to have insight into is based on the fact that the outer radial glial genes fit into modules of gene expression that are specific for certain functions, and they're highlighted here. Extracellular matrix production, the transition from epithelium to mesenchymal, which occurs we think when the cells arise, and niche formation, namely stem cell self-renewal or maintenance. And so all of these pathway genes have been previously described as being enriched in this brain tumor, glioblastoma multiforme, untreatable, horrible disease, mostly of adults. So we look carefully at samples that were acquired by our neurosurgical colleagues in the operating room. And these are heat maps showing single-cell gene expression of those surgically removed samples. And for the most aggressive glioblastomas, which are highlighted in this top row, we found the pattern of gene expression that we'd come to recognize representing outer radial glial cells. So we were curious if there were oRG-like cells in these tumors. And so in fresh samples, we sectioned them and cultured them as I've shown you before. And here's an example. If we look at the time-lapse image, we find cells that have the morphology of outer radial glia and the behavior, that MST behavior when they divide, they jump and divide. And we found these predominantly in the most aggressive glioblastomas, the form that are known as mesenchymal in type. Now there are many different progenitors that people have hypothesized that could be stem cells for these tumors. What I'm showing above are gene expression patterns in our resected tumor samples, and below are the same genes as expressed in progenitor populations populations, including fibrous astrocytes, protoplasmic astrocytes, OPCs, both adult and developing OPCs, and radial glia and intermediate progenitors. And what I want to highlight is the highest correlation between the cells in the tumors and these progenitor populations in primary tissue was between radial glia and the tumors, suggesting that there were radial glia-like cells in these aggressive glioblastomas. And when we looked at the hub genes and the modules, we found compared comparing fetal tissue to these tumor types, that we had many of the same hub genes associated with outer radial glia. And perhaps the most enriched was this gene, PTPRZ1, which is one of the markers for outer radial glia in the fetus and was also a highly expressed marker for the cell type in tumors. That was interesting to us because just recently, a couple of years earlier, it has been found independently that PTPRZ1 is a highly expressed gene in aggressive glioblastomas, and so it was knocked down in this study, and the tumor cells then engrafted into a mouse. And when PTBRZ1 was knocked down, there was far less spread of the tumor in the mouse brain. And this is now being used as a therapeutic, at least in a clinical trial, to see if it can reduce the spread of glioblastoma. So we looked at the function of PTBRZ1, which is highly enriched in outer radial glial cells in our native cells, and found that if we blocked PTBRZ1, we could significantly inhibit the jump, that MST that we talked about, in these outer radial glia from the fetus, and then again in the tumor cells. With that in mind, we wondered if these could be the tumor-initiating cells in glioblastoma. So we took fresh tumor resected samples, divided them into two kinds. We enriched for the outer radial glial cells, and we had depleted and enriched fractions of the tumor, which we then injected into human organoids. Where they grew actually very quickly. But the PTPRZ1 cells injected into the organoids grew much faster and disseminated much more widely within the organoids. And we then were able to extract them and do single-cell RNA sequencing and find that the diversity of cell types found in the adult tumor were recapitulated by the PTPRZ1 enriched fraction. So this gave us encouragement that these oRG cells are contributing to the tumor itself. And that the outer radial glial-like cells in the glioblastoma might be serving as cancer stem cells or tumor stem cells. This may be an opportunity for new targets to treat the disease. Now since that initial sequencing I mentioned, which was very modest, we've now profiled many more cells using the DropSeq approach. So our next data set includes 4,200 cells from 5 different brain regions across time over 48 individual samples. So this is a large dataset of single-cell cells, both neurons and non-neurons during brain development over the period of peak neurogenesis. So these cells can be clustered into cell types. There's a much larger diversity of cell types now than we had with our smaller population. And in addition to just classifying cells based on the transcriptomes and genes that they express, we found that it was very useful to put coherent networks of genes together into modules and use those to help distinguish one cell type from another. And the method we used was weighted gene co-expression network analysis, or WG-CNA for short. And this really identified modules of genes that really made sense in terms of the function of the different cell types. So returning to my favorite cell, the outer radial glial cell shown schematically here, we've kind of decorated it with cartoons of different receptors and intracellular signaling molecules and pathways that our WG-CNA approach identified were enriched in this cell type. And I want to focus on a couple of them right now. One is the LIFR-STAT3 pathway. Some of you who are stem cell biologists know that this is a typical signaling pathway for self-renewal of stem cell populations. What's interesting here is that it was only the outer radial glial cells that express enhanced LIFR STAT3 signaling. That's confirmed in this image where we show the LIFR receptor is expressed uniquely in SOX2 neural stem cells in the outer subventricular zone and is completely absent in the daughters that they produce, which are the eOMES-positive intermediate progenitors. So it's only the radial glial-like outer radial glial cells, not the intermediate progenitors, that express the LIFR receptor. And similarly, they also also express the signaling partner STAT3, which is associated with LIFR activation. So with that in mind, we wondered if you could activate self-renewal of these outer radial glial cells in model systems like, for example, organoids. So in our organoids, as well as organoids in other hands, some of these HOPX-positive outer radial glial cells do emerge, as I showed you earlier, but in much fewer numbers than you'd find in normally developing brain. So we took some of our organoids and treated them or exposed them to LIF, the leukemia inhibitory factor, to see if we can activate this LIFR STAT3 pathway in those cells and stimulate self-renewal. And after 2 weeks in our organoids, shown here on the right, we had a much larger population of HOPX-positive outer radial glial cells. And these are the cells that we'll show you later, the cells that produce the upper cortical layer neurons. So by increasing the number of progenitors in these organoids, we have higher numbers and greater diversity of excitatory neurons, especially in the upper cortical layers later on as these mature. Another pathway that the outer radial glial cells are enriched in is the mTOR signaling pathway. That's shown in a different way in these maps, which are spatial time-related maps that really highlight the expression of these different genes, which are all part of the mTOR signaling pathway in outer radial glial cells. And this was confirmed in tissue sections shown here using Phospho-S6, which is a readout of activated mTOR signaling. And shown here, co-labeling nuclei of these outer radial glial cells, you can see that the activated mTOR signaling is really highlighting the cytoplasm, in this case the fibers of these outer radial glial cells. They're highly enriched in activated mTOR signaling and more so than any other progenitors in the developing cortex. The reason we focused on mTOR signaling—and I'll mention it again during this talk—is because it's been implicated in diseases such as autism, tuberous sclerosis, and macrocephaly. And so this finding suggests that at least in humans, a role that mTOR signaling may have during early brain development in these diseases would be likely a role it has in the function of these outer radial glial cells. So what kind of signaling could the mTOR pathway be activating in specifically outer radial glia? Well, we looked at this by labeling the cells shown here, these outer radial glial cells in green, and then using inhibitors, for example rapamycin, which is an mTOR inhibitor. And the feature that was most dramatic was the loss of radial fibers. As you can see, diminished radial fibers in the rapamycin-treated outer radial glial cells, and this was reversible in a number of different ways by adding other inhibitors to disinhibit that pathway. And if we look more carefully at the morphology of the outer radial glial cells shown here individually in culture, these are normal cells treated with vehicle that have a single primary fiber, a basal fiber. But grown in rapamycin, we find that the cells, in addition to having somewhat abnormal basal fibers, grow multiple other fibers from the cell body that go in other directions. They become multipolar-like cells instead of this very unique uni- or bipolar morphology that they normally have, and that's quantified here, and that was also found to be reversible. So just to make a long story short, the path— the part of the mTOR signaling pathway, which is ubiquitous pathway in many cell types and very important for growth, but is also important for cytoskeletal features, and it turns out in the outer radial glial cells, the mTOR complex is especially related to basal process morphology, and we think has a role in cortical expansion as well as neuronal migration. And we'll return to that at the end of my talk. So I've mentioned organoids a few times, and we're in a very, I think, unique position to be able to compare primary tissue to organoid tissue and see how well those organoids actually reproduce normal brain development, and then again, how well do they reproduce disease pathologies. So just highlighting the primary tissue studies that we've done looking over time and space and how progenitors and also neurons arise during brain development give us a very good benchmark against which we can compare the same processes occurring in organoids. Now there's some things that are very obvious when it comes to organoids. First of all, you coax them to grow from pluripotent cells down a particular lineage, in this case a neuronal lineage, and then dorsalize them and make them ventral forebrain-like progenitors in our case, and let them aggregate and self-organize self-organize to form a cerebral organoid. And those are the protocols that most people use. We've used several of them in our lab, several different protocols. They all give similar results. And this just shows some of them in histological detail. As they aggregate, they form these rosettes, which are essentially little ventricular, subventricular zones, and over time they start producing neurons. So to compare them to the primary tissue, we've stained here with markers for the different classes of progenitors that I've highlighted so far. They're shown in different colors over cortical developmental time in sections of fetal tissue up here in the top row, and then the bottom row, comparable ages in organoids stained with exactly the same markers, color-coded the same way. And I hope you can appreciate the organization, although both of the normal tissue and the organoids have their own characteristic forms of organization. They're very different. These swirls, the circular rosettes that form the progenitor zones in the organoids, are very different than the more linear ventricular and subventricular zones in primary tissue. And then when neurons are formed, they tend at best to form a bilayer, sort of upper and deeper cortical layer, or sometimes of course, mixed cortical layers instead of the exquisite multilayered development in the general— in the usual cortex. So in terms of single-cell sequencing, we looked more carefully at the composition of these organoids over time. And we were able to get, as shown here, cell lines from 5 individuals. I'm sorry, this is the fetal tissue first. We were able to get 5 donated samples ranging from gestational week 6 to 22, which covers the period of peak excitatory neurogenesis, over 7 cortical areas and a total here about 189,000 or almost 200,000 cells from primary tissue. And we look at the markers that highlight some of the key cell types which are labeled here. In the top row we have progenitors that are SOX2 positive, the outer radial glia that are expressing HOPX, the intermediate progenitors that express EOMES, and you can see that the population numbers vary as well as, of course, their sequential origin over time. Newborn neurons, excitatory neurons as they mature, and inhibitory neurons were also found in these organoids. They're shown below. You can get some idea of their quantification or representation quantitatively in these organoids based on the number of little blue dots, which represent individual cells. If we do the same now with the organoids— and this I should mention are 4 cell lines that were divided into 3 different protocols and then put together with a total of 109,000 cells over times in development that correspond to the fetal cell stages I showed you earlier. So staining with exactly the same markers shows that these markers are in fact expressed in distinct populations in the organoids. Their representation in terms of numbers are not the same. Generally, the earlier cells are represented better than the later developing cells in terms of numbers. But we look more carefully at the comparison of gene expression, there's some striking differences that I want to highlight here. When we compare primary tissue, we find that first of all, there are about 600 genes that are defining individual cell types enriched in one cell, not in another. If we look at similar defining cell genes in the organoid cells, there are far fewer number. There are about 46 that could be seen as actually defining cell types. If we look at the cells— the genes, rather— that are the same between primary and organoid tissue, it's an even smaller number. There are only 5 or so that are the same. If we look at the heterogeneity of cell types in fetal tissue, it's very rich in both regional and laminar identity in the excitatory neurons, shown in this first column here in orange. But we look at the same diversity in organoids, it's far, far less. That's largely because these cells don't express the same laminar diversity or regional diversity that we find in primary tissue. They're more like generic excitatory upper or deeper cortical layer neurons that lack a lot of those specificities that we see in normal tissue. And that's also true for the radial glia shown in the next column and for almost all the other cell types. So the heterogeneity of normal tissue is reduced. And then perhaps most worrisome, we find a blending of cell identity. By that, let me just use two examples: radial glia and neurons, excitatory neurons. So shown in fetal tissue here in brown are the radial glial gene expression, and this axis here, the neural expression in fetal tissue, and they don't overlap. They're very distinct. The radial glia don't express neural genes, and the neural genes don't— the neurons don't express radial glial genes. But in the organoids shown in blue, We have radial glia that express neuron genes, and we have neurons that express radial glial genes, and so we have a form of blended identity. Not all the genes, but about 20 to 30% are shared in ways that we don't find in primary tissue. So we looked at organoid modules and compared them to primary modules of individual gene types, and that gave us more insights into both similarities as well as differences between organoids and primary tissue. And what I'm showing here are modules of gene expression that characterize these different cell types: radial glia, intermediate progenitors, excitatory neurons, and inhibitory neurons. And by and large, the correlation of these modules from fetal tissue to organoid was reasonably good. The major cell classes were represented reasonably well by gene modules, which, shown here in the bottom, broken down into their hub genes, showed very similar overlapping hub genes in most of these cell types. And so that's the good news: we had the major cell types and they look similar to the cell types we find in fetal tissue. But there were some significant differences that I want to highlight here. The organoids all had enriched glycolysis network genes. So shown here on the right in these violin plots are the first two traces represent these genes, in this case PKG1 and ALDOA gene expression. In primary tissue, and then the subsequent violin plots show the same gene expression levels across different organoids. And you can see the organoids show highly enriched expression of these two genes which are involved in glycolysis, an indication of metabolic stress. That's also true if we look at endoplasmic reticulum stress network genes. So once again, these first two by limb plots are in primary tissue and the rest are in organoids. And you can see that these two genes, which are ER stress genes, along with many others, are highly expressed in organoid tissue cells, but not in primary tissue. So the organoids are generally under a great deal of metabolic stress. And that's not just true in our own hands. If we look at published datasets shown here, single-cell datasets from other labs indicated on the right, we find the same enhancement of these ER stress and metabolic glycolysis network genes that we see in our own hands. And it doesn't change over time. If we look at the organoids from, say, week 3 and compare them all the way through to week 24, we don't see any improvement in correlation with primary tissue. They remain both degraded and showing a significant amount of stress at all ages. So this is a caution when you're interpreting disease phenotypes, especially neurodegenerative disorders or other diseases where metabolic stress is considered a factor. So we were more interested in how the stress develops and how it might be maybe modulated or attenuated, and so we took our organoids which have these stressed cells, we labeled them and injected them into an in vivo mouse brain We let them grow there for a month or so and then removed them, FACS sorted them, and did single-cell RNA sequencing of those cells having been in a mouse brain environment for a month or so. And to our surprise, we found that these grafted cells eliminated their stress. These are 3 of the stress genes shown in red. The neurons are human— mouse— human neurons in the mouse brain are shown in green. You don't see any red because these gene expressions, which were very high to begin with, are quantified down below. Had essentially normalized over the period of a month. So the cells essentially reduced their stress back to normal, and very interesting, the fidelity of the cells also improved, shown here for ORGs and excitatory neurons. So the gene expression became crisper as well as the metabolic stress going away. We think those are related, and we're doing experiments now to show whether they are. We also did the sort of the flip experiment. We took normal cells, fetal cells that have no stress, label them, and then grafted them into organoids, and then removed them a month later to look at their single-cell gene expression. We found that the organoids were able to activate metabolic stress in these transplanted cells. So the fetal cells began as normal cells, but after sitting for a month in this organoid environment, they upregulated the same metabolic stress genes that the organoids expressed before.. Not only that, but the subtype crispness of each of the cells became degraded, so their gene identity was blunted at the same time as their stress was increased. So this is encouraging in a way because it suggests that the organoid technologies could be improved to actually improve the fidelity of the cells and reduce their metabolic stress. So at this point, it just highlights that the organoid environment is suboptimal and can be— in fact, should be improved. Now more recently, we've been extending our studies to earlier stages, which is a very critical stage of brain development when neuroepithelial cells transform into radial glia. So the neuroepithelia are the true stem cells of the neural system. They form the neural plate and then eventually the neural tube, but when the neural tube closes, they start to transform in a number of different ways into these radial glial cells, and these are the cells that actually start producing neurons. So that early transition hadn't been very well studied, and so we started looking at that in our human samples. And these show some of the markers that we found in the neuroepithelial stages of brain development compared to similar and other markers that emerged during the neuroepithelial to radial glial transition. And looking at single-cell RNA sequencing now of many, many cells at this transition period, we were able to identify 9 progenitor clusters, and these are not all necessarily different cell types—some of them could be different states of cells—but we use it as an indication of how diverse the progenitor population is during this critical transition going from expansive neuroepithelial cells to neurogenic radial glial cells. And taking this data set, Ugoma Eze, an MD-PhD student in the lab, applied RNA velocity to look at—sorry—to look at lineage relationships. And she found evidence for both— sorry— for both direct neurogenesis going from neuroepithelial cells or radial glia directly into neurons, as well as indirect neurogenesis through the intermediate progenitors, which start to be developed even at these early stages. And this is consistent with what people think in the mouse, that direct neurogenesis may occur much more often in early populations than in later ones. And then mapping the expression of individual pathway genes over time shows many genes during the neuroepithelial stages that are enriched and then diminish in the transition to radial glia, and other genes that are lowly expressed in neuroepithelia and then suddenly become highly expressed in that radial glial transition. So this gives us novel markers for that transition that we didn't have before. I just want to show one of them here, validated in tissue sections, which is NOTCH1. It goes up in expression during the neuroepithelial stage and then diminishes as neurofibroblasts transition into radial glia and neurons start to develop. So we're trying to see if we can manipulate some of these genes to shift that transition, the point of transition. The other thing we notice is that some of the genes that are characteristic of these early progenitor types— and I'm showing one of them here, C1orf61— is expressed in human tissue, very highly expressed as you see here. Gratifyingly, it was also expressed at a similar stage in organoids neuroepithelial stage in human organoids as well as in chimp organoids, but we didn't find it at all in mouse. And this is a characteristic of several of these gene expression patterns that we see now—neuroepithelial gene expression that we see in humans but not in mouse. And there are some that we see in mouse and human that we don't see recapitulated in organoids, although I haven't— I'm not showing you any of those right now. The other interesting feature is that many of these genes that are enriched in the progenitors, those 9 populations I mentioned earlier, are associated with neurodevelopmental disorders, and some of them are shown here, and they are enriched in one or another of those different populations of early progenitors I started to show you. And this is ongoing work, so we haven't sorted all this out yet. And finally, I wanted to end on an evolutionary note. We are also interested in comparative studies between humans and non-human primates as well as mouse. And to understand the human-specific features of the brain, we really have to compare them to our nearest living relatives, which are the chimpanzees. They diverged approximately 6 million years ago from humans, and they are our closest living ancestors. But we really are not able to access the developing fetal brain of chimpanzees. So Alex Pollan, an evolutionary biologist who joined my lab a number of years ago, focus on this problem. And he developed protocols for growing not only human but great ape organoids, including gorilla, chimpanzee, and others. And then having developed that protocol, we started looking at single-cell comparisons between human and non-human primates. So this shows primary samples we were able to look at from macaque fetal as well as human fetal brain. And those we were able to cluster into the major cell types. And start looking at comparisons between human and non-human primate. But perhaps more interestingly, we grew organoids from human and chimpanzee, compared them to each other, and also compared them to primary tissue from a cat. So that allowed us to start looking at how well the organoids are mimicking what happens in human development. And then especially when it came to chimpanzee organoids, start to look at how— first of all, how good they are, and then secondly, can we use them to look at human-specific gene expression? And this has been published, so I just want to get the highlight, which is yes, we can do all this and look at gene expression in different species in specific cell types. So what I'm showing in these violin plots on the right are genes that are enriched in human outer radial glia. So ETNK1 is enriched in primary human tissue and human organoids, but not in macaque or chimpanzee organoids. So this is a gene that is uniquely expressed in human radial glia, not in any other primate. And there's another gene with the same pattern of expression. It's seen in human but not in macaque or in chimpanzee. And there are other genes in radial glia that are the opposite. They are found in macaque and in chimpanzee, but have been lost in evolution of human brain. And now I'd like to return in my final slides to the mTOR story I mentioned earlier. If we look at the mTOR genes, which are selectively shown here, they're enriched in primary human tissue, which is what I told you earlier. It's gratifying to see it repeated here. We also find those same pathway genes are enriched in organoids as they are in primary tissue, so there's a very good congruence of the organoid to primary tissue. But they're not enriched in non-human primates, neither in the chimpanzee or in the macaque. So what this tells us that we didn't know before is that the mTOR signaling enrichment in aberratively that I highlighted earlier is a human-specific feature. And that's documented here where we went to macaque tissue and showed that the upregulated pathway in atheroma in human is not upregulated in macaque. And we also looked more carefully and found that there were two receptors for the mTOR signaling pathway that are highly expressed in atheroma in human but not in non-human primates. And if we knock them down in a human, we get levels of expression that resemble more what we see in primate. So what that tells us is that the mTOR signaling pathway that is uniquely expressed in outer radial glial cells is a human-specific feature. And so to the extent that that is a characteristic that's disease-relevant for diseases like autism or tuberous sclerosis, it suggests that this is a disease that may not be modeled— not only not modeled in mouse, but possibly not even modeled in non-human primate.. It may require human or human organoid-type platforms to understand. So just to conclude, there's a greater diversity of neural stem cell subtypes in the human developing brain than in the mouse. The validated features of organoid models can reveal developmental features and disease mechanisms in ways that you may not be able to in non-human models. But they demonstrate activated stress pathways and they have degraded cell type sub-identity that has to be considered if you're modeling either normal development or planning to model normal development or look at disease phenotypes. And then finally, I want to thank the people in my lab who did the work, including both Tom Nowakowski and Alex Pollen, who initiated single-cell work in our lab and have now established their own labs. And I also want to highlight the work that I mentioned from these three investigators. Aparna and Madeline are both terrifically skilled postdocs, and Ogoma is an MD-PhD student. They all did great work that I talked about later, especially in the single-cell sequencing of early stem cell populations and organoids. And the work is supported by our granting organizations, and I want to specifically thank the BRAIN Initiative that got us started on our single-cell sequencing work. And now I'm happy to take any questions. Thank you.