Transcript

New Frontiers in Understanding Brain Development

From Allen Institute · Episode page · Watch on YouTube

· 1 h 59 min · 17,644 words · 143 timestamped sections

This is the full, timestamped transcript of "New Frontiers in Understanding Brain Development", an episode of Allen Institute published December 11, 2025 (1 h 59 min, about 17,644 words). It mentions 20 people, works and ideas, including Brain Initiative, Aparna Badori, Allen Institute, Carlo Konuani, Yongsung Kim.

Who speaks in this episode?

What is discussed, and when

  1. – 10:00: Brain Initiative, Nature, Carlo Konuani, Allen Institute
  2. – 20:00: Aparna Badori, Brain Initiative, Nemo Analytics, Nemo
  3. – 30:00: Brain Initiative, UCSC Cell Browser, UCSF Cell Browser
  4. – 40:00: Brain Initiative, Yongsung Kim
  5. – 50:00: Allen Institute, Yongsung Kim, Yuan Ga
  6. – 1:00:00: Allen Institute, Cindiva Belvin
  7. – 1:20:00: Allen Institute, Emily Han, Marilyn Styart, Thomas Noacowski
  8. – 1:30:00: Harris Kaplan, Thomas Noacowski
  9. – 1:40:00: Camille Mannins
  10. – 1:50:00: Lee Wong
  11. – 1:59:10: Brain Initiative, Allen Institute

Notable quotes from "New Frontiers in Understanding Brain Development"

  • I'm Fuisan from the AL Institute. today we'll present to you the work that has been included in a recent publication package in nature. the work was from the brain initiative.
    about Allen Institute · at
  • Several major publications especially from the second phase of the brain initiative effort called brain initiative cell census network or BICN.
    about Brain Initiative · at
  • But as I'll show you these everything I show you today is can be point-and-click online at Nemo Analytics. okay.
    about Carlo Konuani · at
  • On. So the study that we presented as part of the developmental working group package was a meta atlas of the developing human cortex and the goal was to similar to what Carlo…
    about Aparna Badori · at
  • In the stereotactic coordinate frame as well. With this we were able to establish template across seven different stage.
    about Yongsung Kim · at
  • So our data set spans from E11.5 to P28 and we also supplemented this data set with the adult P-56 vocal cortex data from the whole mouse brain project and this schematic highlights.
    about Yuan Ga · at
  • All right. Thank you, Yian. Next, we have Cindiva Belvin, who is also from the Allen Institute, where she is an associate investigator on the informatics and data science team.
    about Cindiva Belvin · at
  • Hopefully a link will be shared to that soon. And I want to thank especially Matt and Tom for their contributions to the study. it was really a joint effort between Matt and I…
    about Thomas Noacowski · at

Who and what is mentioned in this episode?

People

  • Aparna Badori — Person · Main speaker · 2 mentions · first at
    Presents a meta-atlas of developing human cortex and a meta-module method; discusses FEZF2/TSHZ3 regulatory model validated in organoids.
  • Carlo Konuani — Person · Main speaker · 2 mentions · first at
    Presents a data-integration and analysis resource centered on neocortical development, emphasizing Nemo Analytics for point-and-click exploration and cross-dataset signatures.
  • Yongsung Kim — Person · Main speaker · 2 mentions · first at
    Presents the developmental mouse brain common coordinate framework (dev CCF) integrating MRI and light-sheet templates and developmental ontology-based segmentations.
  • Yuan Ga — Person · Main speaker · 1 mention · first at
    Presents a mouse cortex developmental dataset and analysis linking adult cell types to developmental origins, gene modules, and gene regulatory networks across ages.
  • Cindiva Belvin — Person · Main speaker · 1 mention · first at
    Presents deep analysis of transcriptomic and spatial organization of GABAergic neurons in mouse telencephalon, including developmental integration and diversification timelines.
  • Thomas Noacowski — Person · Referenced expert / co-lead · 2 mentions · first at
    Named as a co-lead of the overall publication package; later appears as the lab head for lineage tracing work (spelled similarly as “Tom Nacowski”).
  • Emily Han — Person · Main speaker · 1 mention · first at
    Presents cross-species single-cell analysis of striatal interneuron development, arguing TAC3 interneuron lineage is conserved and homologous to mouse TH interneurons.
    Student in the pollen lab at UCSF and I'm really excited to share our recent study titled conservation and alteration of mamalian stridal interurons.
  • Marilyn Styart — Person · Main speaker · 1 mention · first at
    Presents lineage-resolved atlas of developing human cortex using viral barcoding (“sticker”), highlighting truncated radial glia outputs and late-born deep-layer excitatory neurons.
    All right. so today I'm excited to share a lineage resolved atlas of the developing human cortex that I've been working on alongside a now graduated PhD student Matt Keefe in the lab of Tom Nacowski.
  • Harris Kaplan — Person · Main speaker · 1 mention · first at
    Presents a developmental single-nucleus RNA-seq atlas of the mouse preoptic area (POA), focusing on early diversification and sensory-input dependence (pheromone pathway).
    Next up, we have Harris Kaplan, who is a post-doal fellow at Harvard University in the department of molecular and cellular biology.
  • Camille Mannins — Person · Referenced person (unclear identity) · 1 mention · first at
    Not clearly mentioned in the provided transcript context; may be an extraction artifact or appears outside the included excerpt.
    Next we have Camille Mannins, a post-doal researcher at the Flemish Institute for Biotechnology.
  • Lee Wong — Person · Referenced person (unclear identity) · 1 mention · first at
    Not clearly mentioned in the provided transcript context; may be an extraction artifact or appears outside the included excerpt.
    Thank you. All right. Now we'll move on to our final speaker of the day, Lee Wong from Stanford University, an assistant professor in the biology department. And this is our last presentation of the webinar and we will.

Books & papers

  • Nature — Book · Scientific publication · 3 mentions · first at
    Journal where research findings from the Brain Initiative are published
    I'm Fuisan from the AL Institute. today we'll present to you the work that has been included in a recent publication package in nature. the work was from the brain initiative.

Companies & orgs

  • Brain Initiative — Company · Research consortium · 10 mentions · first at
    Central organization funding and coordinating brain research projects discussed in the podcast
  • Allen Institute — Company · Research organization · 5 mentions · first at
    Key contributor to brain cell type atlases and developmental studies
  • BICN — Company · Research consortium · 1 mention · first at
    Phase of the Brain Initiative focused on cell census network
    This work has culminated in several major publications especially from the second phase of the brain initiative effort called brain initiative cell census network or BICN.
  • Nemo — Company · Research tool · 2 mentions · first at
    Likely shorthand for Nemo Analytics, a data analysis platform
    So that's where you can go online to surf data sets looking one gene at a time. What I'll talk mostly about today is how we can grab focused subsets of these data sets and understand them in a.

Products & tech

  • Nemo Analytics — Product · Research tool · 2 mentions · first at
    Digital platform for analyzing and visualizing brain development data
    Really want to get this in the hands not only of bioinformaticians but as I'll show you these everything I show you today is can be point-and-click online at Nemo Analytics. okay.
  • UCSC Cell Browser — Product · Research tool · 1 mention · first at
    Platform for visualizing cell type data from brain research
    If you're interested in digging into this further, the meta atlas is available at the UCSC cell browser both for development and adult. And the code to make your own meta atlas…
  • UCSF Cell Browser — Product · Research tool · 1 mention · first at
    Platform for visualizing cell type data specific to UCSF research
    If you're interested in digging into this further, the meta atlas is available at the UCSC cell browser both for development and adult. And the code to make your own meta atlas…

Places & events

  • COVID-19 pandemic — Event · Historical context · 2 mentions
    Mentioned as part of the broader context affecting research

Full transcript with timestamps

I'm Fuisan from the AL Institute. Uh today we'll uh present to you uh the work that has been included in a recent publication package in nature. Uh the work was from the uh brain initiative. Um in in the consortium uh we have extended our cell atlasing work in the adult mammanian brain to creating comprehensive cell atlases uh in the developing brain from mouse to human. And this new set of effort we believe uh will create a new frontier in better understanding uh the development of cell types in the mamalian brain and its implications uh in brain function and diseases in particular neurodedevelopmental diseases. So I'm giving you an overview here uh on behalf of my co-leads of the package Baduri from UC and Thomas um Noacowski from UCSF.

The over the past decade uh brain initiative cell senses consortia uh have been using single cell genomics technologies uh as well as other multimodel approaches to create comprehensive and high resolution cell type atlases uh in the mamian brain from mouse to non-human primate to human and this work uh has uh culminated in several major publications especially from the second phase of the brain initiative effort called uh brain initiative cell census network or BICN. And in this publication packages we have generated uh cell uh type taxonomies and atlases starting with individual brain regions uh to the entire brain. In 2021, we published a set of 16 papers in nature and uh describing the consensus cell type taxonomy of primary motor cortex um from mouse to uh non-human primate to human and reveal a series of underlying organizing principles of the brain.

After that uh just two years later uh the consortium also generated um a a large set of uh studies about the human brain and and about 20 papers and and one of the central themes of that uh set of studies is the creation of a draft version of the whole human brain atlas. And immediately following that uh again the consortium uh and with uh multiple teams working together uh 10 publications in nature uh created a multimodel uh single cell transcripttoic epigenomic and spatial transcrytoic atlas of the entire mouse brain with more than 5300 uh cell types or clusters identified.

So following up on this uh comprehensive work now in the third phase of the brain initiative uh called uh brain initiative cell atlas network or by CAN. One of the main goals now for us is to create similarly high resolution and comprehensive cell atlases across the entire pre and postnatal development of the brain. uh again across multiple mamalian species from mouse to uh mama macak and to the human brain itself. The goal of this is really to understand the similarities and differences of cell type development across species.

Understand how the human brain develops in ways that are similar to the animals and also in ways that are unique to the human itself. And we want to use gene expression uh uh changes uh to correlate with the uh various structural and functional developmental events and from there gain mechanistic understanding of how uh brain development occurs and how it could goes uh could go arai in diseased conditions. And here summarizes the main approaches that we have uh taken to create those uh cell atlases. uh they include single cell RNA seek single nucleus multi ohm which is the combination of RNA seek and attack and spatial transcrytoic uh uh technologies uh and couple that with lineage tracing and from this kind of a large scale comprehensive studies uh we were able to uh create cell type classification schemes and achieve cross species alignment um and reveal the spatial localization and migration of different types of

cells during development. And and from this kind of um atlasing effort, descriptive effort, we also have gained a certain uh insight into the molecular mechanisms of brain development including the uh uh generation of cell um type developmental trajectories or nailages as well as the gene regulatory networks that may be uh uh uh regulating or modulating uh the brain development. Finally uh we think this kind of work cell atlasing work will have a big impact in uh understanding both the mechanisms of brain development as well as in um uh uh discovering ways of uh treat uh better understanding diseases and the treatment of diseases. This include going from uh molecular mechanisms to understanding how uh connectivity and function emerges in the brain. uh and to the understanding of the different genetic variations or mutations uh that could lead to selective vulnerability of different cell types in different kinds of

neurodedevelopmental disorders. uh identifying specific pathways that may be particularly changed diseases. And then for better study of disease models in both animals and in vitro systems and um you know helping with the uh uh functional evaluation of potential drugs or genetic interventions and ultimately may lead to um uh the development of precision medicine. Of course, this is a long way to go, but we think that having um comprehensive higher resolution uh cell atlases would be uh essential, you know, would be serving as essential foundational resources to enable all these things uh to happen. Okay. So with that I'll end my introduction and you'll hear individual talks now about the insights um that the team has gained from a set of really spectacular studies and I hope you will enjoy the talk.

Thank you for your attention. Thank you Hungary for that introduction. Next up we have Carlo Konuani who is adjunct faculty at John's Hopkins and the Institute for Genome Sciences of University Maryland. Thanks. I'm very excited to share our work today in this incredible effort that Hungui just um gave us [clears throat] an overview of. Um our study is quite different than most of the others you'll hear today in that we generated no new data. So our entire mission was to get all of the data we possibly could together in one place not only in to enable our own research but to empower researchers across the globe.

And we focus particularly on neoccortical development. Um, and we really want to get this in the hands not only of bioinformaticians but as I'll show you these everything I show you today um is can be point-and-click online at at Nemo Analytics. Um, okay. So the first piece of what I want to show you today is the data. Um, a little bit about it and then what we can do with it. So this is kind of a summary of what's in Nemo Analytics. We have data from mouse, primates, uh, and human just about as much as we could find um, and bring together. And that's both in vivo development and also in vitro models such as um, IPS cell driven uh, neural differentiation and cerebral organoids.

So that's where you can go online to surf data sets looking one gene at a time. What I'll talk mostly about today is how we can grab focused subsets of these data sets and understand them in a deeper context together. And one thing that we'll highlight is the other thing that you can do online in addition to looking individual genes is to explore complex gene signatures um [snorts] from your own analysis or from across the literature across the same expanse of data that that we've brought together. So here's our first example of uh an analysis. We do a lot of joint decomposition. Please reach out if you're more interested in the methods. I won't talk about them today.

Um so here is a in three rows, three data sets, mouse, monkey, and human single cell RNA sequencing from midjustation focusing on the excitatory neuronal trajectory. So looking at uh the birth of glutamaturgic neurons in the neoortex what the four columns are to the right are four different signals that we got out of the joint decomposition that are found to be common across these systems. So this is sort of a high resolution version of what we know to be true. That's why we call it the neoortex. It's new in evolution. It's in mammals and we know the six layers are conserved and we know some of these states are conserved. So these are four patterns that we got that are shown to be conserved across these three data sets. They correspond to rada IPC's intermediate progenitors that come from those and then new neurons that are being born and finally in the last column maturing neurons.

So here we're looking at the sample space. So the cells in each case in each data set in each species that are expressing these different patterns. But behind this is the gene space that describes which genes across the genome are being uh harnessed in each of these stages. And now we're going to take those signatures. So I did this offline. I can upload those signatures to Nemo analytics and we can explore them in further data to get a much deeper understanding of where those signals came from. So the four columns I'm going to show here are the same four. are the radial signature, IPC signature, new and maturing neurons. And as I go along, I'll show you a new data set just where these signatures are showing up. So this is bulk RNA sequencing from laser microcted human cortex. So in the first column you can see the radle cells are showing up um in these early ages in the ventricular zone right where

we know the radleal cells to be. the next column where the IPCs have delaminated from that ventricular surface that molecular signature is now showing up in the inner subventricular zone and then as the neurons are born they're still in those uh subventricular zones but when you when by the time a neuron is expressing this last signature it is in its final resting place in the uh laminate organization of the cortex in in the cortical plate. So this is a much newer spatial study that can give us even higher resolution understanding of where these patterns come from. This is stereo seek. Um and again um we're looking at each of these same four patterns.

You can see that the radial signature lines the ventricle here. [snorts] The IPCs move out and so on. And really it's just the exquisite data that can be produced today um uh enables this kind of analysis. So and again you can see additional stuff you weren't even thinking about asking perhaps we were looking in the telesphylon for these signatures but the progenitors and also neurons of the entire brain expressed these signatures in different times and places and timing is key here. This is now looking at these same signatures in a study where they birthdated every cell before they sequenced it. So you know the last time this cell divided before you get all its transcriptional information. So this allows us at least in the mouse here in this experiment to look at how long it takes a cell to move through these states.

And this now is the same um two signatures these the nent and maturing neurons showing that they go they turn on the m the nent signature and then turn off the maturing one. So if you kind of keep this in mind we're going to connect this after one slide to the end of the talk. So quickly I'm going to reverse the sort of direction of analysis here and then come back to what we we just looked at. So we took some adult data even though we're interested in development and got a lot of single nucleus data from the adult neoortex from the allen and we uh decomposed it into 20 patterns. These rows are nine of those patterns where we can see across spatial and single cell data that without any instruction or supervision the algorithm can identify very layer specific signals. These are two of those a layer four signature on bottom and a layer three four signature on top. And I just zoomed in here to to show you the

the wonderful data that's out there to for us to use. So you can see in these two different human studies where exactly and there's really sort of fine details in right in layer four or just above it. Um this is in the macac some another wonderful stereo seek data set and when you can look at so much data at um so sort of smoothly and in a directed fashion things start to fall out. For instance, in this mouse data set, we discovered that this layer 4 signature does not occur in the mouse cortex um just because we could see it.

So now we started with those developmental signatures. We now have a collection of adult signatures. This is now starting to bring those together. So this is that layer 4 signature that was primate specific. We're now in the human, so it's showing up. The color scale here shows you age where in a specific layer 4 population this is coming on but it's not just coming on during in uterero or neonatal development. It takes many many years and it looks like about a decade for this particular signature to reach its maximum. Now this is interestingly in stark contrast to canonical transcription factors that are known to set up these patterns. So ROB is a layer 4 master TF but it's peaking right about at birth and then trickles down and I think Aparna Baduri who will speak today will also give you sort of a in-depth view of one gene FEZ F2 that plays in the uh a specific game of this general lesson.

Okay. So last thing I want to show you is again bringing all this together. So this is a data set that has both in uterero and postnatal samples. We're looking at that turning on of the nent signature in neurons and then turning on maturing signatures as that goes off. That's what the up and down is here. And the color is the color of one of those or the strength of any one of those nine layer signatures. So if we look at one row, this is when and where a uh the layer 34 signature comes on. Right about here at 22. So if we extend this all of a sudden we can look at at every age as cells go up and down in this process where do they turn when and where do they turn on the adult signatures and these green arrows can kind of show you that just like the inside out organization of the birth of neurons in the neoortex.

The maturation [snorts] follows the same trajectory with a notable exception of one of these patterns that shows up 10 weeks early which is super interesting and we're we're looking more into that. The last thing I want to show you is the same analysis now, but looking at cerebral organoids, we need to know what organoids model so we know how we can use them and how we can improve them. So, we're going to do the same analysis that we just did in invivo development, but now in organoids. So, the nine rows are again those signatures from the cortex of layer specific neurons and the columns are time points. And really there's only two in this and in many other data sets there's really only two deep layer neuron signatures that show up to mature. Well, now I'll show you one last thing here. One thing that you can push you much farther, this is from Sergu Pasca's lab. If you put that organoid

into a newborn mouse pup's brain, all of a sudden everybody is much more happy to keep maturing with the notable exception of an upper layer signature. And that layer 4 signature that we saw was absent from the mouse does not get induced even though these are human cells. The mouse brain environment can't hit it. So I'd stop by saying please get your own data in here if you have new data and please reach out and use these resources in your own research whether you're coding or clicking. Thank you very much.

Thank you Carlo. Next up we have Aparna Badori a assistant professor of biological chemistry from UCLA and she'll be presenting on behalf of uh a postoc in her lab Patricia Nano. Hi everyone. Hopefully the sharing is working well. I am really delighted to see all these papers out to see the package out and I'm delighted to present today some of the work from our paper in the package which is um led by first author Patricia Nano. Unfortunately, she couldn't give this talk today because she is at a faculty interview. I think she might be literally mid chalk talk.

So sending good vibes her way and excited to share what she's been working on. So the study that we presented as part of the developmental working group package was a meta atlas of the developing human cortex and the goal was to similar to what Carlo was doing but with a different approach identify modules that are driving cell subtype specification in the developing human brain. So um the rationale for doing this is basically that there's a lot of molecular profiles that have been made um at the time of our study. This was primarily from some of the initiatives that Hungi had highlighted as being um previously published and this included some of the previous uh brain initiative packages. However, not all of these studies can span all of human brain development. And this is really a function of some of the rare samples as well as some of the cost considerations and the availability of human tissue

which is going to have many technical differences across different modalities as well as as a function of the fact that it is indeed human tissue. So to this end, Patricia wanted to generate a meta atlas that would put together these data sets and allow us to find broader biological patterns that might be describing some of the trajectories of development. To do so, she leveraged seven single cell RNA sequencing data sets that focused on the peak stages of neurogenesis. This came from 96 individuals and after our additional QC to have uniform standards across the data it was approximately 600,000 cells and to begin for visualization purposes we used a standard RPCA type integration but we wanted to be able to look not only at the cell types and clusters that were emerging from traditional analyses but also to overcome some of the technical variations that are going to vary based on things like sequencing

depth and single cell capture methods as well as to identify subtle differences between developmental cell types. And the overall goal of this study was really to start to begin to understand how we can link gene networks in cell biology. So to do this, Patricia innovated a novel gene co-expression analysis method which we are calling a meta atlasing strategy to make meta modules. The way that this works is it looks at each of the 96 individuals in the meta atlas one at a time. This allows us to use unsupervised clustering approaches for each individual in order to identify the axes of variation that are present in that individual sample. This helps us overcome some of the technical variation. And when we then look at the gene expression signatures within each individual and compare them between individuals, we can start to get these meta modules which are chain which are networks of genes that are changing

across different samples in the data set but also that are important for any given time point or sample. With this approach, Patricia generated 225 meta modules and she annotated these using a combination of prior knowledge, literature gene um ontology enrichment, etc., and was able to see that there was a mix of modules. Some of them seemed to be cell type specific indicating that there were cell types represented that maybe are specialized in something like axon function whereas others seemed to be broader biological processes such as those related to metabolism that we thought might be more continuous across the spectrum. In order to look at this more deeply, we wanted to make sure that this was in fact the case. So we use some positive controls such as looking at vascular modules in vascular cells and microglea and immune modules in the relevant cell types. We also observed some specificity of gal

populations and those relevant modules but these were sparse given this focus of the peak stages of neurogenesis. Now, one of the questions that we had was, should we be doing this from a developmental forward-looking approach or should we be looking backwards from the adult cortex into development? And we wanted to be able to ask the question of which of this these approaches is going to allow us to understand the initiation, refinement, and mechanisms of cell type specification. So to answer this question, Patricia took her meta module approach for a spin and looked at a adult meta atlas that she also generated. This was primarily derived from some of the papers that were published as part of the brain initiative adult atlas um that came out um in the in I believe 2022 or 2023 and from this was able to compile 16 data sets across 274 individuals and 2.6 6 million cells using the same approach.

This now gave us an opportunity to compare modules between the two contexts. And one of the ways that Patricia did this was using a module specificity score. She did this by looking at the specificity of a given module which are shown at the top of this heat map across cell types which are shown on in the rows. So here you can see what the developmental modules look like when applied to development and you can see a lot of that continuity and co-expression that we were discussing and has been well known to exist in development.

Similarly when you look at adult modules on the adult brain you see something very different. Patricia likes to use the word crispy to describe this and I kind of agree with it. It really shows a lot more um specificity to individual cell subtypes or groups of cell types. So, Patricia then asked, "What happens if you apply the developmental modules onto the adult or the adult modules onto development?" And what we observed is regardless of the source of the modules, there was always more continuity in development and more granularity in the adult which suggested to us that if we want to understand how these cell types are made, we need to focus on how that specificity becomes um comes from the broad expression that is seen in development. So focusing on development first and then looking forward into adults might be a good strategy to identify mechanistic modules.

I'll give you a quick vignette on how we did this for one of the modules. Module 20 was a module from the developmental atlas that was expressed in neurons and seemed to increase over time in deep layer populations but not in upper layer populations. This was particularly interesting because when we looked at it in the context of the adult brain, we saw that it actually did have some specificity as was alluded to and in this case it was specific for fuzzf 2 annotated subtypes. Now this was really interesting to us because fezaf 2 is one of the uh terminal transcription factors that has been described in the mamalian brain and a lot of beautiful work has been done in the mouse to describe how fuzzf 2 is capable of specifying specific subtypes of deep layer neurons in the in the brain with um very unique projection patterns. Now what was interesting is that fuzzf 2 was not actually a member of the module 20 but

its expression preceded that of module 20's peak and when we looked at some of the existing mouse data sets that exist um that looked at chipsseek to look at targets of fezf 2 we saw that more than half of module 20 genes were in fact fezf 2 targets suggesting that this might be a regulatory mechanism we also saw that the main transcription factors c PSHC3 was in fact um one of the regulated genes and this was the only transcription factor that existed in module 20. So we wanted to test whether fuzzv 2 and tshd3 are required for module 20 activity and we did this by pro doing a knockdown experiment in cortical organoids. I won't go through all of the details, but basically we leveraged a chyoid system in order to look at multiple lines simultaneously.

And after generating these organoids with knockdown at 8 weeks, we sorted out the knockdown cells from FZF2, TSHDC3, and the double knockdown. What we saw is that there was a decrease in module 20 activity as well as the expected decrease in deep layer populations that seemed to trade off with an increase in radial GAF. Now what was really exciting about this is both FZF2 and TSHD3 seemed to regulate module 20 genes based on the RNA expression. But when we looked at chromatin accessibility from our multi ohm profiling, we observed that TSHD3 not FZF2 was required for chromatin accessibility of FZF 2, TSHD3 and module 20 genes suggesting a new model where TSHC3 is an important intermediate in the specification of deep layer neuron identity where waves of expression of FZF 2, TSHD3 and module 20 are required for this deep layer neuron specification. ation.

If you're interested in digging into this further, the meta atlas is available at the UCSC cell browser both for development and adult. And the code to make your own meta atlas is available on GitHub. With that, I'd like to conclude. I'd like to thank Patricia and all of her um colleagues and uh co-authors on this study. And I will now hand it off to the next speaker. Thank you everyone. And I'm happy to answer questions in the Q&A chat. Great. Thank you, Apara. All right, next up we have Dzang. She is a post-doctoral fellow in the department of biomedical engineering at Yale.

Great. Thank you for the kind introduction. Good morning everyone. My name is D Jang. I'm from department of biomedical engineering at Y University. Today I will talk about how I developed and applied the special matomics technologies to decipher the malian brains asking how cells coordinate make decisions and become disregulated within their native tissue environment. The brain is the most complex organ in the human body orchestrating thoughts, emotions and behaviors through intricate molecular and cellular interactions.

As we know each cell was regulated by proton, transcryto and epigenome including open chromatin hto modifications and other processes. We realized spatial epigenome and transcryto co-p profofiling by using deterministic buckoding in tissue or dbit. Dibbit is a versatile platform. The key idea is that the buckle's information can be written into the tissue section. Beyond that, to investigate the molecular mechanisms across all the layers of the central dogma, we further realized our ambition of doing special triomix mapping on the sim tissue section.

Scientists for brain development research has mainly focused on neurons. However, the non-neuronal cells also require in our attention uh such as understanding when, where and how does malination occur in the brain. Therefore, we mapped the mouse brains from newborn to juvenile stages using special triomix which received clear special um mapaps for both attack and RNA. The mature oligodendrites marker MBP showed RNA attack and protein signals at P10 and P21 in corpus colosum. The gene fox P4 involved in regulating morphogenesis disappeared during development. Perhaps is no longer required when all brain structures has formed.

We further integrated our attack and RNA data leading to a more refined special domains which provide better aerial aerialization when compared to multiplex imlloresence imaging. We dive into the copus colossum and found that the mileing signal start emerging at around P7 to P10 at the lateral parts of the cops colossum reaching full coverage by P21 and extending into the cortex suggesting a lateral to medial progression of malination. To investigate the spatial and temporal variations, we developed a computational framework to model attack and RNA signals in cops colossum.

After modeling, we observed not only concordance between the attack and RNA signals for oligodendro galo genes, but also chromatin priming at P 0 to P5 where RNA expression is observed at later stages. We also applied the same computational framework to the cortical layers. We found alignments of RNA and attack signals for many of the layer specific transcription factors such as COX 2. However, we also observed misalignment such as MEF2C. This might reflect an epigenetic memory of the developmental states. After combined analysis of all the genes for projection neurons and mileing in each special pattern for both cortical layers and the cops colossum, we observed a timed and sequential bidirectional mileination process occuring from the central cops colossum where the OPC cop enter from the postnatal wave to first migrate and differentiate in the lateral part of the cops colossum and then followed by the

second phase of migration and differentiation in the medial direction. We further use a retrograde AA tracing strategy to chase the preference of milealination and found that while corticalatic projection neuron axons in the lateral part of the copscum are already mileinated, the colossal projection neurons in the medial direction are starting to be mileinated. indicating that the axon of neurons might orchestrate the OPC migration to MO minalination process. The different cells are not separate chapters but as movements of the same symphony of time.

Neuroinflammation recycle many of the developmental programs and occurs in almost every brain disorder. In multiple sclerosis, the immune system attacks the mileing shift around the neurons to cause demalination. MS is difficult to treat because the exact cause is unknown. We built a focalod mileination MOS model of MS to study the disease process. We found that the mileing and mature alleged sites were disappeared at the legion sites by 5 days with microfasages and microgia populated there marked by CD1B. By 21 days, the lesion was was much smaller with a clear remalination and reduced microfasages and the microgia.

Here we zoom in on microgenriched regions in the primary lesion and in dist regions. We found that dist microglea are more neurons supportive while microglea at the legion core are pro-inflammatory. Importantly inflammatory signals in the distal region appear later than in the primary lesion suggesting a delayed wave of inflammation. We further mapped the sagin sections. Interestingly, we found distrogal activation spreading in anterior, posterior and ventral directions but returned in the white matter tracks. How about the situation in the homos brain instead of just one tissue section? Using 3D trick disco imaging we saw micro gleo spreading into both hemispheres with stronger signals on the injection side.

These findings suggest that a focal demalanating lesion can trigger distal microgalo activation potentially propagating inflammatory signals along axon tracks. In summary, we generated the spatial triomic atalysis of MOS brain development and demalination. We found that the layer specific projection neurons can coordinate axon genesis and malination during development. In focal dealination model, we uncovered the microgle activation in dist white matter regions potentially transmit the inflammatory signals through neuronal circuits. Using special mitoics, we are not only observing cells, but we are witnessing how life organizes complexity into order and how it be rebuilds itself amid chaos.

I would like to thank all the grants and my amazing collaborators for supporting this work. Thank you for your attention.

Great. Thank you, D. Next up, we have Yongsung Kim, who is a professor in the department of neuroscience and experimental therapeutics from Penn State University.

Hey, hello everybody. Um, today we're going to talk about a developmental framework called developmental mouse brain common coordinate framework in short dev CCF as a part of this consortium effort. Um, let me get started. Um through the Bican effort and many other brain initiative effort and many also individual effort uh there has been uh huge advancement to understand the brain cell type composition and their connectivity and the molecular profiling and all these informations and one of the key aspect of this large scale effort is to uh map all the datas in the coherent spatial framework in adult brain. And for example, I'll let mouse brain common coordinate of version 3 CF v3 serve as a common coordinate systems to map all these informations which has been really critical to integrate these multiscale and multiomic uh multimodel data set in the developing brain. Before we started our this uh the project there

was a few uh two dimension atlases based on hisystological section. So this is all uh uh developing mouse brain atlas based on histologology section and then the segmentations. Although this provide very detailed segmentation in the his based on histologology uh these types of two dimension uh atlas cannot support the 3D atlas a 3D data set. To fill that gap um we started the projects to establish multimodel 3D atlas in the developing brain and we decide to use two different uh imaging modality. The first modality is MRI. MRI provide um inscal data set uh though relatively low resolution. It provide inscal resolution in the intact volume and the brain can be dissected out and then undergo ting process and we use a light sheet for microscope to acquire several resolution informations.

And then this live sheet data is also mapping back to the MRI based atlas framework. And we uh uh uh uses combination of different criterias to segment the brain areas uh across entire uh mouse ages and our chosen time point is between uh embry all the way to the p-56 in the in seven different stages. So I'll provide a bit more detail how we uh uh accomplish these goals. So the uh uh uh first uh part is developing template and uh previously uh mouse atlas is based on traditional any one brain uh we want to change that. So this time we pulled about 10ish from the male and female mouse brain at each age to generate morphologically uh averaged and intensity average template across seven brain stages. Here is MRI template and we also have a light sheet based template with a lot higher resolution and because lit imagings require tissue clearings and tissue clearing often include unexpected tissue distortion we align all the ly

data back to the MRI uh space to uh bring the morphology back to the intact space and with this now MRI and the lis are in the synchronized the space and because MRI provide uh uh cut uh 3D coordinate we were able to put the grid to calculate uh to display how this template uh uh uh place in the 3D coordinate systems. Not only this uh we also imported alen CF V3 from adult brain into our adult MRI and live sheet based templates to put the island CF V3 in the stereotactic coordinate frame as well. With this we were able to establish template across seven different stage.

After establishing template we need to establish anatomical segmentations. So the way we uh uh uh draw anatomical segmentations is following. So here is a live sheet autofllororescent uh signals and you can see three different uh plane and here is MRI ADC contrast MRI DWI contrast and FA contrast and T2 weighted contrast. And these different contrast provide a different uh uh uh contrast in different brain region that help us uh to delineate and establish anatomic labels. And I want to empathize emphasize the fact that uh we use a different uh anatomic label compared to the uh adult CCV3 uh labels based on site architectures. In the developing range, site architecture change rapidly. So we have to use a developmentbased atlas uh atlas label scheme. So we borrowed um the developmental ontology based uh scheme of pioneer by Luis Ples.

And here we were able to establish 3D atlas also across all seven ages. And when we establish this uh anatomic labels, we use a various different u evidence including contrast from the different imaging modality as well as uh antibbody label that light up a particular features of the brain and uh also including ner stainings as well. And with this anatomical labels we were able to calculate how the volume of the mouse brain increase across the across the developmental time frame. So what is the use of this to demonstrate the uh the utility of this newly established anatomical frameworks?

We use a transening animal that labels garbage exile type in the development brain. We use a G to create cross with a C-ependent reporter mouse and perform tissue clearing and map their expressions. So here is a raw data. You can see gabage cell types emerge in the different parts of the brain and this is embriionic 11.5 13.5 15.5 when their signal is detected and maps on to the atas framework we can clearly see how uh garbage cell type in different brain areas emerge uh across the time from the different uh areas.

So those are development and time point and our uh development atlas also include adult time point and this adult time point give unique opportunity to compare and contrast our developmentalbased atlas labels to side architecture based CCV3 labels. What you see on the left is our what you see on the left is u uh CC V3 labels and one on the right is our de dep based labels and because all the both atlases are in the same template. We can compare left side how dep label correspond to the right side als uh uh atlas being in the same space we were able to uh uh import uh the spatial transcription data from island institute published in 2023 and then compare how this different cell tonomies is located in different anatomical the segment. The top is uh based on an anatomical segmentation in the CCF V3. Bottom is our dev CCF based arrangement. How these different cell types are arranged.

And to highlight the difference between DECF and the uh and the CCF V3 anatomical segmentation, I want to draw your attention to uh this particular brain area called P3. And the P3 is approximate three in the depth CCF. And and this is combination of zona inserta here. And this is considered as a h part part of hypothalamus in V3. And here is a reticular and part cons area considered as part of a phalamus. But this uh combination of boththalamus and jona insertas are considered as a one structure same structure called promate three based on our developmental ontology based anatomic segmentations. So this provide perspective how anatomical segmentation can differ based on their criteria.

So we establish all these atalas framework. We wanted to be used and shared as broadly as possible. So we uh build a neural glance based uh data sharing uh platform. So if you come to this website our lab visualizations you can see depf uh uh resource and here is download link all these uh files including high resolution datas are viewable and downloadable openly from these links and I want to summarize it. So what I show you today is we establish developmental atlas with a 3D template based on MRI and the life sheet and also anatomical labels based on developmental ontology. And we demonstrate use of it by mapping garbage excel type in the developing brain. And we also shows how developmental common coordinate framework label is compare and contrast to the site architecture based uh CCF v3 labels.

With this I want to acknowledge that Fate Kman was our brilliant former graduate students and Fate led the entire uh project uh with great collaborators. uh and uh Janya from NYU provide MRI support and Luis Ples was pioneer of developmental ontology based segmentations and Lydia and the gym provide new informatic support and uh I also appreciate all the contributors um and the funding source from NIH. Thank you.

Thank you Su. Next up, we have Yuan Ga from the Allen Institute where she is a scientist on the informatics and data science team. Uh hi everyone. Uh I'm Yuen Gal. I'm a scientist at the Allen Institute. So today I'm very excited to share our work uh continuous cell type uh in diversification in mouse cortex development where we explore how these cell types develop and evolve across ages in the mouse cortex. So our data set spans from E11.5 to P28 and we also supplemented this data set with the adult P-56 vocal cortex data from the whole mouse brain project and this schematic highlights the major milestones in the isocortex development across the ages.

We collected the 10x version 3 single cell rick data and also the 10x multimum data. After quality control, we retained more than 500,000 cells from the 10x version 3 [clears throat] and over 200,000 nuclear from the multium data set. The bottom panel here shows the number of cells that pass the QC at each of the age. This is the overview of our analysis workflow from this study. And after collecting data from each of the individual time points as I showed in the previous slides, we uh do the clustering within or between different ages. These steps serve for different purposes including identifying the major developmental patterns and also the cluster level quality control. And then we do the integrations between ages and we systematically propagate the cell type labels from the later stages to the earlier stages. This allows us to build the allows us to connect the adult cell types to their developmental origins and

also to reconstruct the trajectories that trace each of the uh adult cell type back to the developmental counterparts. In the meantime, we use the Murfish data to validate those cell types. So once the trajectory was built, we analyze the gene modules that are co-expressed across different ages and also the cell types. We also examine the gene trains within each of the subclass to understand how those gene functions shift during maturation. In parallel, we also collected the maltium data and we integrate our single cell RNC data and the multium data. Then we transfer the cell type labels from the single cell RNA seek data to the multium data and leverage both the transcripttoic and epigenetic information to build the gene regulatory network. This enables us to identify the key transcription factors that drive the cell type specification and also the maturation.

Our study found very high transcriptional diversity among the early embionic cell populations which may indicate that the selfish specification may occur at a very early stages for okay these three U maps are for the embronic population colored by subclass age and cluster and you can see the U map this is a nonIT population this is it population. We can trace the divergence of the nonIT and IT population in the stage of the intermediate progenitors and even within radlea they show dynamic temporal changes in gene expression. As you can see in this dot plot, in this plot we can see the markers of the nonIT cells for example the FF2 the IT cells like post post3 F2 and also markers for the TNC those emerge in the staggered overlapping pattern which becomes increasingly distinct at the later uh developmental stages.

Our analysis also uncovered the early signals that drives the diversification of the OPC and astro sites. In addition to that, we also find a distinct population which is glyobblast SVZ. This population gives rise to a specific adult astroite uh cell type localized in the subvententricular zone. We also constructed the detailed trajectories for the emergence of each of the subclass with the precise time steps and our findings indicate that almost all the subclasses we observe in the adult stage they are already specified either before or soon after birth around P 0. The only exception is the mature oligodendrites which appear at about P10.

And to study how gene expression changes across development, we have identified about 5,000 developmentally regulated genes and we group them into 36 trajectory patterns. These patterns fall into five broad developmental trends which is up transient up transient down and there there's one more which is constant at different levels which is not shown here. Even within the same category individual patterns can differ in meaningful ways. For example this is trajectory 2 and this is trajectory three. The trajectory 2 increases during the early embronic stages and plateaus at about P 0. But uh trajectory three uh increase mainly after birth and plateaus around the time of eye opening.

Most of the genes show similar patterns um across related subclasses but many also show subclass specific dynamics. For example, as you can see here, this is for TBR1. TBR1 expression broadly increased during neurogenesis, but it diverges across the upper and the upper and the deep layer neurons after the immature neuron stage with the strongest down regulation occurring in the lay in the layer 5 ET cell types and several other key transcription factors. For example, COX 2 and Fox P2, they show similar complex and subclass specific temporal regulation highlighting their tightly controlled roles during the cortical development.

In the post-natal period, we examine the transcrytoic changes across the subclasses. The plot on the left shows the magnitude of the gene expression changes between the consecutive time points where the positive values indicate the upregulation and the negative values indicate the down regulation. This is for the nonIT cells and this is for the IT cells. Um in this plot we observe multiple waves of changes and with over time with I opening around P10 to P13 marks the final bigger wave of ch uh wave of wave of change. Um this is not only observed in the nonIT and IT cells. We actually observed in most of these cell types including the garbagic cells and other nanural populations which is not shown here and we have more details in the manuscript.

activity dependent immediate early genes shows the strongest uh responses. For example NR1 which is triggered by the calcium influx shows dramatic up up regulation and other well-known uh classic uh IEGs like BDNF, NR4 A1, EGR3 they all increase after opening alo with different uh cell type specific timing. Using the gene regulatory network analysis, we identify that the AP1 complex is the central regulator of the activity dependent maturation. AP1 and back two they share many target genes and um our analysis suggests that back two is downregulated after eye opening and it normally repress the late onsite late onsite genes. is down regulation opens those sites allowing the AP1 to such as uh drun and force to activate those target genes and promote the uh neuronal maturation.

We also identify the cell type specific regulator. For example, NR42 as you can see here in the network is a key regulator of the car 3 subclass with its motive enriched in both early and late car 3 subclass specific PI modules and also in the nonIT cells. The post3 F1 is a regulator of the L5ET along with their cell type specific targets. Uh finally I wanted to thank all our authors who have contributed to this work. Especially I wanted to thank Hungu Zindi. Their support, patience and mentorship have been invaluable for me and I'm really grateful for how much they've helped uh me uh grow as a scientist. Thank you.

All right. Thank you, Yian. Next, we have Cindiva Belvin, who is also from the Allen Institute, where she is an associate investigator on the informatics and data science team. Ah, okay. Now I'm unmuted. Thank you, Rachel, for the introduction. Um, so we took a deep dive into um the transcrytoic and spatial organization.

Hi Cindy, are you with us? Yes, I think I don't know what happened. Okay, we see your slides now. Okay, perfect. Um, okay. So, we took a deep dive into the transcrytoic and spatial organization of um [clears throat] gabaurgic neurons in the mouse tel and sephylon. In 2023, we released um a whole mouse brain cell type. Um this is a hierarchical cell type taxonomy that was created from single cell RNA seek data and a spatial transcripttoic Murphy's data set of covering the entire um adult mouse brain. Um in the accompanying manuscript we describe the landscape of all these cell types and the relationships between these types at the subclass level which is a relatively high level. So for this specific paper, we um kind of isolated the subpelium gabberurgic neighborhood um and took a closer look at the organizing principles um within this specific uh group of talent talentic uh gabberurgic neurons. So this

neighborhood consists of 615,000 cells. Um and we broke that down into seven classes, 51 subasses, 285 super types, and um a little over a thousand different clusters. So this hierarchy that you that you see here is the hierarchy at the superype level. And um below it, you see a breakdown of uh the region that the the cells were dissected from. And what I hope you can see is that some of these classes are very regional specific. So for example, the lateral septum class um contains mostly cells um coming from lateral septum and the same for all factory. But then some of these other classes contain cells coming from different regions.

[clears throat] We use a lot of different visualizations to take a look at um the relationships between types and how they're organized. So um on the left you see a U mapap at the subclass level. So the intermediate level in uh in the hierarchy um and a constellation plot of those subclasses. In this constellation plot which is based on the U map each cent position is the centroid of the subclass in the UA map space and the size of the node tells you something about the size of the number of cells that are included in the cluster. The edges between this the the nodes um show how many of the neighbors um of each of the cells are shared between the two different nodes that are connected. And then we have the spatial transcrytoic data that helps us really put all these transcrytoic types into um a spatial location. So in this case the the murfish sections that are shown are shown showing the class

labels. So very broad high level uh labels. Um but you can see that these classes are very um tightly distributed within the brain. [clears throat] Um this is a similar constellation as shown in the previous slide but now really showing it at this superype level. Um and there is a lot that you can glean from a constellation plot like this as well as from the different U mapaps and um the the spatial transtomic data. Um in this specific constellation plot the um kind of the left to right transition in um in the constellation is related to kind of a dorsal to vententral um positioning within the brain. So uh in this outer ring you have like the types that are located in cortex and all factory. Um and these types tend to be somewhat bigger tend to be more transcrytoically distinct from each other. Then in the middle you have these um types these classes that are populating the cerebral nuclei which

includes the strienum. um the LG derived gabaurgic types tend to be somewhat bigger and tend to be more distinct whereas the MGE derived um CNU types tend to be more transcrytoically similar. And then we have like the more most ventral um classes, the lateral septum um and the anterior hypothalamus where we have a lot of smaller um super types that are highly similar to each other. In the manuscript, we describe um a variety of kind of discrete um and continuous gene expression gradients that um are running through all these different types and some of which are associated to um spatial dimensions as well.

These spatial and molecular gradients, they don't only organize the cell type diversity within the adult mouse brain, but also suggest kind of deeper rules of specification of these types. rules that we can potentially um uncover by looking at the transcription factor profiles. So um in this study we identified a large set of transcription factors um that are specifically marking classes and subclasses and super types um that we have defined in our transcrytoic data set and um several of these are key factors that play a role in uh early developmental processes and there are some examples u below the heat map where um the first images are the are from the ISSH atlas is showing the expression of specific genes um at E13.5 and then the expression of that same gene in our adult Murfish data set. Um where a lot of these patterns are very highly localized in the embryionic period, a lot of these the expression

patterns tend to be more widespread um in the adult brain. So to take a closer look at um where these cells are coming from. So in um in the mouse um there most of the gabberurgic cells are derived from five different um progenitor domains within the embryionic subpelium. From there they um proliferate and migrate um and ultimately differentiate and incorporate into um the developing circuit. Um and so to trace back where these cells are coming from, we generated a new um an additional data set covering um nine different time points from E11.5 to P14.

And we integrated this with external published data sets that are mostly covering um the embryionic period from um E8 to roughly P3. So in total we have um 614,000 cells um coming from uh developing cells that we can integrate with the adult data set. So this is what the integration then looks like. Um these are U mapaps. The first one is colored by age. Um within with dark red is the p-56 and then um it's more like the more you go to the middle the younger the the cells are getting. The middle is the classes that we have assigned to each of these um um to each of these cells and then on the right hand side is the adult uh subclass that um has been assigned in the adult mouse brain taxonomy.

So the labels were transferred in a method similar to what UN just described where we mapped we transferred labels from P-56 to P14 and from P14 to P12 and so on. Um, and as expected, um, a lot of the transcription factors that we've identified, um, previously, we can trace these back to specific progenitor domains, um, early on. Um, so when we take into account the spatial location that we've identified in our adult Murfish data set, um, this really indicates that most of the gabberurgic neurons um, really have migrated over long distances to reach their final destination. But these long um kind of migratory journeys are really just one step in the process. As these cells are dispersing from their progenitor zones, each lineage will undergo further diversification which gives rise to kind of the rich um repertoire of cell types that we observe in the in the adult taxonomy.

This diversification however happens at very different rates. So in the dorsal classes, these cortical um classes coming from CGE and MG, the neurons diversify extensively but mostly later on. Um so after birth [clears throat] um whereas in the more vententral classes the lateral septum and the uh anterior hypothalamus classes um there appears to be kind of a burst of diversification in the embryionic period where most of the types tend to kind of arise all at the same time. Um this space of diversification mir mirrors some of the cluster properties that we've seen in the adult types. Um so the vententral classes show a rapid um kind of embryionic burst of diversification and you yield a lot of small um tightly related types whereas the dorsal classes continue their diversification um postnatally and they give rise to the larger more distinct types.

Um in summary uh we like I just mentioned the dorsal forbrain structures have uh larger more distinct clusters whereas ventral forbrain structures have like smaller more similar clusters we've identified uh various spatial dimensions of discrete and continuous variations um in uh the cell types that we've identified um long distance migration and dispersion happens to kind of nearly all of the classes of the talentic aber veric neurons. Um, and there are distinct developmental timelines for the diversification of these cells. And with that, I would like to thank everyone who was involved in getting all this work done. Um, thank you.

Thank you, Cindy. Next, we'll we'll move on to the next talk by Emily Han, who is a PhD student in the developmental and stem cell biology program at UCSF. Can you see my screen?

Yep, we can see your screen. Okay, great. Um, thank you for that introduction. Um, and hello everyone. I am Emily Corgan. I am a fifth year grad student in the pollen lab at UCSF and I'm really excited to share our recent study titled conservation and alteration of mamalian stridal interurons. So, broadly the theme of our paper is how evolution can modify the brain on a cellular level. And so first, evolution can modify brain structures by simply increasing or decreasing cell numbers. Cells can also move to new locations.

There can be molecular changes and conserved cell types. And then finally, evolution can simply invent a new cell type in one species that others do not have. In fact, a few years ago, a group was performing single cell RNA sequencing on the human and marmicet stratum and they observed a transcriptionally distinct interuron type that had not yet been described and it expressed the neuropeptide TAC 3, thus giving it its name, the TAC 3 interneuron, and it made up about 30% of the stridal interurons in these species.

And then when they surveyed the mouse stridum, they didn't observe this population. And finally when they surveyed the ferret stridum they also didn't observe this population and so due to its absence in mouse and ferret this cell type appeared to be a primate specific population. Luckily its transcriptto gave us some insight into where this population may be coming from during development due to its expression of the transcription factor NKX2.1. So NKX 2.1 along with LHX6 is this signature marker of the medial gangleionic eminence or MGE pictured here in blue in both the developing mouse and developing macacb brain. The MGE is a highly conserved birthplace for the majority of both cortical and sharal inhibitory interneurons. And so following the final cell division within the MGE the initial class of an interneuron is specified. Then these newborn neurons migrate to their final destinations. And it's during this

migration and after that they undergo postmotic fate refinement to form their terminal class. And so the TAC 3 interneuron itself is the terminal class. But our lab is really interested in development. And so for us a question that we had is what is the initial class that gives rise to the TAC 3 interneuron. And we addressed this in some previous work where we examined single cell sequencing of mouse and primate inter neurons during development and then map them to terminal classes. And so what I'm showing here is really just a fraction of that work. But it was shown that the majority of both initial and terminal classes of inhibitory interurons are conserved across mammals with the exception of both the initial and terminal class of the TAC 3 interneuron being primate specific. But there's been some recent work in other systems that has revealed that populations that we thought were primate

specific actually have orthologes and other species. And so seeing work like that really motivated us to examine the phoggenetic extent of this initial class in other mammals. Which brings me to our new study where we now survey approximately 160 million years of evolutionary divergence stretching from primates to marsupials. We perform single cell RNA sequencing on developing strriatum and cortex samples from seven additional species to examine inhibitory neuron initial class diversity across a total of eight species.

And so after isolating inhibitory neuron populations in each of these species, we then embedded them in a shared latent space. And here I'm showing that integrated UMAP colored by each individual species contribution. And because we were able to integrate all of these eight species, we were then able to label populations based on conserved and well-characterized markers of initial classes of inhibitory neurons. And we saw that cells across the eight species group by cell type rather than by species. And that all initial classes were conserved and highly correlated across all species sampled. But what really grabbed our attention however were these MGE derived crab P1 expressing classes. We saw surprisingly that again all eight species contributed to both crab P1 expressing clusters including the TAC 3 initial class cluster which was previously thought to be primate specific. And when we

examined the gene expression of the TAC 3 initial class in each species, we were able to conclude that the TAC 3 initial class is conserved across placental mammals and exhibits both marker conservation and marker turnover. And so what I mean by that is we saw this initial class is really defined by a core set of markers including MG lineage based markers like NKX 2.1 and LHX6 as well as other markers like ATV1, CHRNA3, acetylcholine receptors um and ZIC markers. But then when we examined individual species more closely, we were also able to see species and clay specific gene expression. And so for instance we saw that marsupials actually lacked cellular retinoic acid binding protein or crab p1 while macac seemed to be enriched for this bmp antagonist sir one. But what was most striking about this is that when we looked at the mouse we saw the loss of tac 3 in this initial class alongside the gain of th. So this

suggested that the TH interneuron which was previously thought to be an entirely different cell type is actually homologous to the TAC 3 interneuron. And to further examine the possible homology of the TAC 3 and TH interneuron types, we performed integrations of adult mouse and primate stridal inhibitory neuron data sets. And so first we saw the integration of major subtypes between these primate data sets including a large TAC3 cluster here a subset of which expresses LHX8. And then when we looked at the integration of the mouse we found that the TH internons integrated the most with the TAC 3 type in primate. We further validated this homology using SAM map where we observed high homology scores between the mouse TH interurons and primate TAC 3. Um, we performed additional tests including gene module tests and pier piercing correlations which further supported these results but I unfortunately don't have time to

show them today. We also independently examined the gene expression of the mouse triadal inhibitory interurons and within the th interneuron cluster we observed a very small population of inter neurons that express residual TAC 2. So TAC 2 is the orthologue of TAC 3. So same gene just a different name. And then using the Allen Institute's MERS scope data, we found that this TAC 2 expressing TH population was localized primarily within the ventromedial strium. And so taken together, we show that this crab P1 TAC3 initial class is conserved across placental mammals with this conserved initial class producing TAC 3 interneurons in most mammals and TH interneurons in rodents. a small subset of which retains that residual tattoo and it's restricted to the ventromedial straightum.

And broadly, we think that this gives us some more insight into how evolution modifies the brain on the mamalian time scale. We believe that it tends to favor teaching old cells new tricks rather than inventing entirely novel cell types early in development with TAC 3 in interuron evolution actually encompassing all of these mechanisms. I didn't have time to go into it today, but we do see a lower proportion of TH neurons compared to TAC 3 neurons in the primate. We also observe this initial class migrating to the cortex in the pig and ferret, which is not observed in the primate or mouse. And finally, of course, we see modifications of its gene expression throughout evolution.

And with that, I want to thank you for your attention, as well as my lab and all of our amazing collaborators. Thank you, Emily. Next, we have Marilyn Styart, also from UCSF, where she is a PhD student in the biomedical sciences graduate program.

Hi. Uh, can everyone see my screen? Yep, we can see your screen and your cursor.

All right. Um so today I'm excited to share a lineage resolved atlas of the developing human cortex that I've been working on alongside a um now graduated PhD student Matt Keefe in the lab of Tom Nacowski and it's an honor to be here sharing this work. Um, [clears throat] okay. So, it takes no introduction to this crowd to say that there is massive cell type diversity in the adult mamalian brain. And of course, since we're interested in development, we really want to know how does it get here? How does it arise from the relative simplicity that exists during development? and progenitors um give rise to of course more progenitors but also these more distinct cell classes over time sequentially. Um so lineage tracing we think is a really powerful tool that sort of focuses in on the arrows on this map and allows us to look at the um connecting individual progenitors to their cell type output.

We use lineage tracing to generate an atlas of human cortical development sampling across a couple of developmental stages and also um two different germinal zones to uh measure the lineages from two different progenitor types um truncated and outer radial ga. We'll kind of come back to that and start with the developmental stages. But first um how did we do this? We used a tool previously developed in the Nakowski lab called sticker um which has these long complex barcodes um and can be packaged into a lentivirus which we could use to infect the tissue. And we actually infected the tissue with two different pools of this um lineage tracing tool so that we could read out the differences between these two regentor outputs.

and we cultured the cells for 10 to 12 days to allow the lineages to take shape. Afterwards, we sorted out just the lineage trace cells which expressed GFP and could use single cell transcripttoics to read out the cell types that we had traced from and the clonal relations between them. So we collected uh dorsal progenitors and their lineages sort of got the expected cell types plus some interesting ones and um we had we had about eight biological samples and we collected over 8,000 or sorry 6,000 multisellular clones. So to sort of get us started analyzing this, we aggregated all together and tried to get a picture of the global lineage dynamics um in a graph that looks like this. So we could see that um for example there are a lot of radial GA shown here in pink and their clonal relations extend across many cell types whereas some other cell types were more likely to be in shared clones with only themselves.

we saw that there was a shift in um clonal relations across midestation. So in the younger samples we saw lots of these excitatory neurons and in older samples we saw an uptick in generation of oligodendrite precursor cells shown in yellow as well as these locally born inter neurons and their progenitor types um shown in blue at the top. And to dig a little bit further into this cell type because it seemed really interesting to us, um, we validated the identity of these cells and found that they have a distinct identity which could not be found in mouse samples and they specifically express co-express um, secretin along with pack six and we could stain for this in uncultured tissue um, in and find them in the cortical plate especially in the upper layers suggesting that these cells were actually being born and migrating out to their final destination.

So in the beginning I introduced that we used two different pools of this lineage tracer to look at the difference in output from outer radia and truncated rada using their spatial separation. And um we did this because very little is known about the output or the role of truncated rada um since their transcrytoic and morphological description um by work that was done in the creek lab um about 10 years ago now. And so we learned a couple of very interesting things. We learned that first of all truncated radial ga are highly self-renewing. Um this shows that so we looked specifically at clones where we actually collected something we thought was a truncated rada based on its transcrytoic um identity and we found that they were often in shared clones with other rad.

So these could be more truncated radlea or they could also be outer rad. The next most likely thing that they were to be um in a shared clone with was excitatory IPCs. And we also saw some evidence of um them being in the same clone as excitatory neurons. So this suggests that they're actually quite highly neurogenic which was a bit surprising. We really expected them to be more glyogenic if anything. Um so we wanted to dig into this a bit further and validate and see if this was happening also at late stages. And we looked specifically at the late stages. Um this is the UMAP with the teal representing cells that came from the VZ and um coral representing cells that came from the outer subventricular zone. And we could see that both of these cellular compartments even at these late stages were making excitatory neurons.

We validated this um especially in the VZ um by staining for neurode2 which is just a marker of excitatory neurons and we found some cells expressing this marker in the ventricular zone even as late as justational week 24. So really um sort of an unexpected finding and we wanted to know what is the layer identity of these late born excitatory neurons. Um, we know that the cortex is canonically born from the inside out. And so we wanted to see um if this was true and much to our surprise um we found that they were actually deep layer excitatory neurons. So the evidence for this was that um the so these are stains from the outer subventricular zone and the the VZ ISVZ region with a deep and upper layer marker TBR1 and SAP B2 respectively. And you can see at early stages there's just a lot of neurogenesis both um upper and deep layer neurogenesis.

At later stages, we saw a market decrease in SAT B2 positive cells in both of these compartments, which was surprising in and of itself. Um, but we also saw continued expression of TBR1 positive cells um in the ventricular zone compartment. This suggested that deep layer excitatory neurons were still being born in the VZ even at these late stages. And we validated this finding using BRDU birth dating um where we could label just the VZ with BRDU and track the cells that were born from that compartment in late stage tissue samples. And we looked at the subplate and deep layers um of the cortical plate and found Brdu positive cells that express TBR1 but not SAP B2 suggesting that they really are deep layer neurons.

um not like upper layer neurons that could be migrating through or anything. This was really surprising, but it did sort of explain um sort of a long-standing question in the field. We know that the human subplate expands well after subplates were thought to be born um which really are thought to be the earliest born cells of or um uh excitatory neurons of the developing cortex. And so potentially the presence of truncated radial ga and their um contribution of subplate neurons late is how the human subplate keeps expanding for so long.

So to conclude um sort of the the major findings from the study were that dorsally derived gabaurgic neurons possess a unique transcrytoic identity. Um they express secretin and pac 6 um together. The truncated radial ga give rise to glutamaturgic neurons and surprisingly these are late glutamaturgic neurons that take on a deep layer identity. Really the summary of this I think and sort of a bigger lesson that we learned from this study was that there are interesting and unique lineage dynamics that happen especially in the later stages of um midjustation and these were only resolvable because we had the lineage tracing tool at our disposal. Um and um we hope that these data will be useful to the community. They're available on the UCSC style browser. Um hopefully a link will be shared to that soon. And I want to thank especially Matt and Tom for their contributions to the study. Um it was really a joint

effort between Matt and I and Tom's mentorship is amazing. Um and of course I'd like to thank the funding and everyone else that contributed to the study. And thank you for your time. Thank you, Marilyn. Next up, we have Harris Kaplan, who is a post-doal fellow at Harvard University in the department of molecular and cellular biology.

Okay, thanks. Hi everyone. Um, I'm very excited to present our work uh alongside the all these uh wonderful studies today. Um, so I want to start with the uh whole mouse brain transcrytoic atlas, which as Cindy described earlier was published a couple of years ago. um by a large by uh BICCN funded consortium. Um so using single cell sequencing they cataloged over 5,000 transcrytoically defined cell types across the entire mouse brain. And importantly they showed they used spatial transcrytoic approaches such as murfish to show where those cell types are localized um in different brain regions and that's uh summarized in this plot here. Um so each column here is an individual cell type and each row is a kind of major uh brain region. And so you can see which cell types fall into which regions. And one of the most intriguing findings here is that the vast majority of neuronal types are actually subcortical. Um and one way

to think about this is that in cortex there are fewer cell types that together form a kind of general purpose circuit which is the six layered circuit we're all familiar with. Um and this one circuit is capable of uh uh very kind of flexible computations. Um whereas a subcortical region like the hypothalamus has many diverse cell types about 900 or so. Um and these cell types are thought to have a much more dedicated control over specific behaviors. And I think this is interesting uh especially in the context of development because I think we know much more about the development of this small sliver of cortical cell types and relatively little um about how uh the vast majority of neuronal cell types develop.

So for example in the preoptic area which is part of the hypothalamus um at the bottom of the mouse's brain here um this area is involved in different homeostatic functions like thirst, sleep and thermmorreulation as well as different social behaviors like parenting, aggression and mating. And many labs have shown that there are transcriptionally defined cell types within the POA that are dedicated to each of these functions. So just like I described in the previous slide, um this subcortical region, the POA has individual cell types that are dedicated to specific functions but not others.

And so we know a lot about the POA in terms of uh the functions of individual cell types, but all of this work comes from studies in adults. Um at the beginning of this study, uh these cell types had really not been described at any earlier age. And so my goal has been to address this question, how do POA cell types develop? And so I established a single nucleus RNA sequencing atlas uh of the mouse preoptic area um at the eight ages shown here um from the end of neurogenesis at E16 all the way to adult at P65. So for each of these eight ages I dissect out the POA extract nuclei and perform sequencing um and so overall I profiled over 200,000 nuclei across the eight different ages. This allowed me to define about 150 different neuronal cell types shown by the different colors here. um 64 excitatory and 83 inhibitory subtypes. Um importantly, these cell types uh were identified by mapping them

to previously published uh murfish atlas of the POA. And because that atlas has spatial and functional information, I can kind of inherit that information into my data set. Um and so I'll tell you about two key questions we asked using this data set. Uh the first is how does the POA generate gen generate this really staggering diversity of about 150 neuronal cell types. And so here you see the same UMAP colored by cell type or by age. And what we see is that each cell type consists of cells uh from all eight ages um in this beautiful gradient pattern which is kind of repeated throughout this plot.

Um, and so what really surprised us when we looked at these data was that even at E16, these cells are already grouping with just a single adult uh cell type in UMAP space, which suggests that they're already diversified uh really early in life. And so to quantify this, we turn to label transfer. Um, and what we found is that indeed each E16 cell type maps in essentially a onetoone fashion um onto the adult cell types. And so this tells us that the full complement of about 150 cell types is already present at this really early age of E16.

Um and so this prompted us to ask when this uh diversity first emerges. Um so we decided to sequence at E14. Uh and we chose E14 because this is roughly the midpoint of neurogenesis in the POA. So at this time about half of the cells are still dividing progenitors and the other half are newborn neurons which were born in the previous one or two days. And so we could ask how diversity looks like in each of these populations. And so first for the neurons, we find that again they're largely diversified already at E14. Um we can map them onto the adult cell types in a onetoone fashion. Um whereas the progenitor clusters largely don't map.

Um and so this tells us that the E14 progenitors really lack any mature regional or cell type identity, which means that these neurons diversify really rapidly um upon neurogenesis in just one or two days. Um and so to summarize this first part, we find that these POA neurons diversify rapidly and very early in life. And one idea um is that this makes sense because the POA is important for functions that might be essential really early in life like thirst or sleep um as opposed to maybe cortex where cell types emerge more gradually um and maybe don't need to be diversified until cognitive functions emerge uh later in life.

Um and so next we asked whether POA development is sensitive to environmental inputs. Decades of research has shown that sensory circuits um in for sensory circuits there are critical periods in development uh where for example visual experience affects the development of visual cortex. And we know that deprivation of early life social experience can lead to long-lasting deficits um in social behavior. For example uh from decades old uh social isolation rearing studies in primates. And yet we don't know how social inputs affect the development of circuits involved in social behavior.

And so to address this uh we decided not to use social isolation but instead took a more targeted approach um by sequencing in different mutant mouse lines that are impaired in different sensory modalities like olfaction pheromone sensation touch uh or thermos sensation. And so for each of these mutants we sequenced the POA in mutants and litter mate controls focusing on three key ages. what we thought when we thought social input might be especially important. Um and so for each mutant we can ask how each POA cell type is affected. Um and to do so we used a distance metric in PCA space to simply ask if the mutant and control cells kind of cluster separately.

Um and what we find is that for the mutants impaired in touch sensation no cell type is significantly affected across uh these replicates. Um for mutants impaired in thermosensation a couple of cell types show small uh effects but nothing uh really major stands out. Um similar for the cell type the mutants involved in olfaction. Um whereas for mutants involved in pheromone sensation uh we found that many cell types are affected um and that the effects are actually really uh uh significant really large. Um and so what is this mutant trip 2? Um so trips C2 is expressed in the varasal organ or the VNO which is an accessory alactory organ that's present in most mammals and which is specialized for feromone detection.

So for detecting these chemical signals for conspecific sex, age or state um and feromone information from sensory neurons in the VNO reaches the POA down here in just three synapses. And so trip C2 is a gene that is expressed almost exclusively in the VNO sensory neurons and it's required for sensory transduction in those neurons. So trip 2 mutants. So in trips C2 mutants phmonal information does not reach the POA. And so what we find is that in these mutants many cell types are affected um but only actually at specific ages. So not at birth um or in adults but only at these kind of intermediate ages of P10 and P18.

And so to understand this better we quantified the gene expression maturity of these neurons using a pseudotype analysis. And so you can see the control cell types show increasing maturity uh in gene expression across age. Um whereas the trip 2 knockout cell types are less mature especially at these intermediate ages of P10 uh and P18. And so from these data we conclude that trip 2 mutants show a developmental delay in the maturation of POA cell types. And so what we think is happening is that at these intermediate ages of around P10, ferommonal information does not reach the POA and this leads to a delay in POA cell type maturation. Um and due to previous work and other analyses, we think that this might lead to long-lasting effects on POA cell type function and adult social behavior.

And so in conclusion, we find that POA cell types show a very striking pattern of early diversification as well as a sensitivity to early life environmental inputs, especially feronal input. And so with that, I want to thank all the uh co-authors um the dual lab and uh collaborators um and especially the banan program. Um thanks so much for listening.

Thank you Harris for that presentation. Next we have Camille Mannins, a post-doal researcher at the Flemish Institute for Biotechnology.

Uh all right, thanks Rachel. Um yes so um this is a project that I conducted during my PhD in the standerson lab uh where we looked at chromatin accessibility during the first trimester of human neuro development. So um an ongoing project in the lab over the past 10-15 years has been the aim to create a comprehensive atlas of the human brain uh in collaboration also with the Alen brain institute. Um which has resulted eventually into landmark papers u characterizing the cell types both for the adult brain as well as the first trimester in development. uh but while working on uh the the developmental part we started to realize that for some of the kind of uh key questions that we wanted to ask about sulfate transitions uh we needed to look also beyond single cell RNA seek which we had worked for worked with primarily before uh and look into gene regulation primarily in the form of chromatin accessibility. Uh so this is

exactly uh what we did. We collaborated with the clinic at the Kolinska hospital as well as the human developmental biological uh resource and Roger Barker at the University of Cambridge. Uh and we collected 26 embryos staged between uh five and 14 postconceptional weeks for which we conducted either single cell aex sequencing or single cell multihomem sequencing. Uh the multihomem we strategically placed along the timeline to try and capture all the different time points. so that we could impute then uh gene expression across the rest of the data set. Um and all these uh samples were dissected into the different posic regions. So we could compare between regions and identify specific cell types for each of them.

Um so this resulted in a a dead set of about half a million cells um where we could compare gene expression with accessibility of the gene regulatory elements driving the expression and we could identify for instance genes that were primed but not yet expressed and vice versa. Um but perhaps uh more interesting for the rest of the talk is that we also looked extensively at gene regulatory elements and what defines uh their specificity in relation to cell types. uh in most cases this means that we're looking at transcription factor binding motifs um and kind of nicely uh illustrated also uh before by Cindy and Emily um uh in the gangleionic eminences where many of the inter neurons are born um there's an early differentiation between the two regions of the medial and the lateral ganglionic eminence and to kind of illustrate why these multiomic approaches can be quite useful um we identified that the key regulators

for those two regions s were both homeod domains home domain transcription factors which you cannot find the difference just by the binding motif but with the RNA data on top of this we could identify LHX6 and DLX2 as the two kind of central players in that field so just to kind of illustrate the approach here um next I'll show a few vignettes from the paper interesting things that we did uh one of the things that we've been interested in for a long time is the the switch from radlea to glyobblast and the timing there. Um so when you look at the different parts of the brain this switch occurs at different time points. So red ga uh in the tennis for instance exists throughout a very long time in development whereas in the hind brain uh they start to switch to a glyobblast state quite early which means that they stop producing neurons and only produce uh gal cells like OPC's and and eststerytes.

uh and we wanted to look at uh some of the the regulators that caused this uh this switch. So we did u differential abundance testing uh related to the time points uh and identified NFI binding factors as uh important here [clears throat] for that switch. Um but it was difficult to see if this is a timing effect or if this was primarily due to the switch uh from reda to glyobblast. So we decided to also look at the other cell types that we had in the data set. um um and compare the earlier to the later versions there. So for instance, we would take uh the neurons and the neuroblast uh split them between the earlier cells so the ones from before 10 postconceptional weeks and those after and then identify uh the transcription factors there that were different and we found that the NFI transcriptious factors in fact were uh enriched in most of these later cell types. So it seems to be an important uh timing uh

regulator rather than necessarily just for for glyobblast transitions. Uh we also found on top of that that many of the later cell types tended to have more accessible regions than the earlier cell types kind of indicating uh a more uh uh specific regulatory programs being active there. Um then we wanted to dive in uh a little bit more in trying to characterize the generatory elements u and what kind of networks of co accessibility that they form. So we did a topic modeling on all of these regions based on their activity in different cell types and I identified about 175 different topics. So um here on the left u the Disney these are genomic elements not cells. Um and we found uh that quite a lot of them had very distinct programs that they could be labeled as. Many of these programs they're related primarily to cell types but some of them uh also showed broader functions. So for instance we could

identify that many of the TSS sites form uh a single cluster as well. Uh but uh most interesting to us was that we found that there were two topics for CTCF binding sites. Uh so these of course are important in the formation of cohesion loops and transactivating domains. Uh and that these topics were specific either to the gal state or uh to the neuronal state. Um so that means that from the attack data we could see uh the formation of this new layout of the of the chromatin in the sense of the the transactivating domains uh which was quite cool. Um then finally um we also trained a convolutional neural network so sequence to function model uh with the aim of taking as inputs uh different gene regulatory elements and trying to predict what kind of cell type they would be active in. Um and then you can use these models to also back propagate through them to try and understand what parts of the input sequence were

important for the prediction. Um so we did this using uh PyTorch. If you're interested in trying anything like this, there's also now some packages that do a lot of the heavy lifting for you. Um, but we use this model to then identify the sets of transcription factor binding motifs that uh the model picks up as being the most important ones for identifying specific uh neuronal cell types. Um uh for that uh then to conclude um this larger data set allowed us to investigate some gene regulatory process processes in more detail. Uh we could use the data also to do enhancer syntax learning.

um and not shown here but if you're interested um there's also in the paper some uh more figures on the temporal delays that we see in enhancer activation and how we can explain those based on the presence of specific transcription factors um and integration with GW was data to try and prioritize vulnerable cell types in psychiatric disorders. Uh with that I want to thank everybody in the lab all the collaborators as well the funding agencies and uh the donors who make this work possible. Thank you.

Thank you. All right. Now we'll move on to our final speaker of the day, Lee Wong from Stanford University, an assistant professor in the biology department. And this is our last presentation of the webinar and we will have some time for some brief question and answer at the end. So please continue to enter your questions in the Q&A box.

Hi everyone. Uh this is a work done in Dr. Arnold Christine's lab and at UCSF and I was a post-doctoral fellow and today I wanted to share with you uh this paper we just published on and defining the molecular and cellular dynamics of the developing human cortex. The neoortex in human is larger expanded compared with the other species and the new cortex contains about 16 billion neurons in the human and it is important for planning, reasoning, language and many other behaviors that humans excel at. Therefore and we really need to understand the neoortex you know in order to understand uniqueness of the human brain and many of the other speakers have already mentioned the developmental state the how the neocortis developing human uh there are certainly many different cell types including cells like radleal cells uh neurons and inhibitory and excitatory neurons as well as cleo cells such as OPC and

astroytes. This is a um really complex uh structure therefore and also be regulated by both intrinsic factors such as gene regulatory networks as well as extrinsic factors such as cell interactions. Therefore to truly understand how the whole process being regulated we need to understand it single cell resolution across modalities. To do that we have collected uh human necortical brain samples from the first trimester and all the way to adolescent stages. And we for the earliest stage samples we focused on the whole neo cortex but for later stages we focused on two major regions. One is prefrontal cortex the other one is a primary visual cortex to highlight regional differences.

We collected the samples and performed nuclear and preparation 10xium sequencing to get the epiggenome and transcriptto information from the same nuclei. In the resulting data sets, we identified in total 33 major cell types in the developing human neuro cortex including different subtypes of the neuro stem cell, radle cells, excited for neurons, inhibitary neurons and glyos cells including astroytes, OPC as well as alotenocytes. We also went further to understand the spatial context of these individual cell types. So we perform analysis by designing a 300 gene panel based on the single nucleus multium data. uh here I'm showing you six of those 300 genes and these are known to be layer specific markers. They're already demonstrating layer specific patterns in this second trimester human neuro cortex. And based on these 300 gene markers, we can further define the cell types including different excitatory neuron subtypes in

the cortical plate as well as neuro stem cell subtypes in the germinal zone. And then based on the local composition of cell types, we can define domains including cortical layers as well as terminal terminal zone layers. Perform such um spatial transform analysis across three different major developmental stages from second trimester to infancy and also across two major locations from PFC to V1. Wanted to mention that these cell types are defined based on the uh markers we derived from the single modium. Therefore we have onetoone correspondence of these cell types to the single nucleus multium data allowing a very uh streamlined integration. We identified in total 10 different domains within the human cortex again across all the cortical layers as well as the uh subcortical regions including the terminal and white matter.

So based on these rich data set we have gained multiple biological insights into human neocortex development at both molecular and cellular levels. One of the interesting things uh we decided to um to pursue is a gene regulatory network that regulate fate decisions of excitatory neuron neurogenesis. To do that we have um identified these irregularones which consist of the transcription factor. It's targeting cancers and targeting gene activities across individual cell types. You can see many of them are displaying cell type specific activities collectively.

We also integrate that information with a trajectory analysis and and identify bifocation point where each of the neuron have to decide which trajectory to follow to become a terminally differentiate neuron or subtypes. At each of each each of these bifocation point you can identify key transition factors that are differentially um have differential differential activity across the two emerging branches and suggesting that they are key regulators of the fate decision of those neurons. We also integrate the single cell single nucleus multium data with our spatial transom data to identify cell cell interactions and in this case we're employing this in in this uh noentric expression model analysis excuse me noentric expression model analysis by understanding how a particular cell type as a sender cell will affect the transcriptto and another cell nearby as a receiver cell. You can see that many

excited neurons are indeed affecting the transmut nearby excited to neurons. But we also find very interestingly that inhibitory neurons are able to regulate the nearby excited trainum even at this very early stage. So we think that probably besides synaptic connections there may be something that is something such as secreted factors that are regulating the nearby excited neurons by inhibitory neurons. And we find that one particular excite inhibitory neuron subtype smosing positive inhibitory neurons indeed are predicted to be able to affect uh regulate or impact uh the nearby excitatory neurons through the somatic signaling and further um further validation experiments demonstrated that activating smitoing signaling actually delays the maturation of excitatory neurons across different neuron types.

We also uh try to understand the lineage potential of individual progenator cells in the human neo cortex by perform prospective uh linear tracing. To do that we performed tissue dissection um uh fact sorting based on cell cell type specific markers for different neuropentor cells. this developing human cortex between gestation week 20 to 24 to understand which potential cells are actually important for the switch from neurogenesis to glyogenesis and we focus our analysis on three of them the truncated radlea and alpha radlea and this so-called IPC ga population based on their self type surface cell surface markers um by tracing their differentiation in the single cell RNA sequencing we find that this IPC cleo population actually give rise to three different lineages astrite OPC as well inter neurons but not excited to neurons suggesting that they're tripotent cells and therefore we have named this

population as tripot potential intermediate progentor cell or tripc on the other hand you can see that the tribc population can actually be produced by both the transfaded radle and outer radlea in 2D culture and these cells will further give rise to astroite obescri neurons later during the culture we also demonstrate the tripotency in vivo um uh using transplantation assays and interestingly this population seems to have some transcripttoic similarity to uh cancer cells in glyobblastoma suggesting some of the developmental mechanisms may be hijacking tumor as well.

Finally, we also wanted to understand the dis if there's any um enrichment of particular neodal or psychiatric disorders uh in particular cell types at specific developmental stages using our uh data set. So in this case we uh combined we integrate the GWAS data with the epiggenome data to understand how the common variants are enriched are associated with different disorders are enriched in the open grounding regions of different neuronal subtypes. Uh here I'm showing you two examples. One is autism spectrum disorders the other one is bipolar disorder. As you can see, they indeed displayed uh different enrichment patterns across all uh neuronal subtypes with autism spectrum disorders are primarily uh risk of primarily enriched in immature excited neurons you find in the second trimester. But for bipolar disorders, those are mostly in the later stage neurons as those inhibitory neurons. We

can also group these risk scores based on developmental stages from the first trimester all the way to adolescent stages. And consistent with previous analysis using um using rare variants in autism certain disorder, we find that the highest risk of disorder is has been concentrated in the second trimester uh from the common variants. And for bipolar disorders, you can see that high risk come actually from infancy stage suggesting disease specific uh risk factors that affect the cortical human cortical development.

All of these data are available from this interactive browser and hopefully you can also take on data to obtain new bilateral insights beyond what we have demonstrated. With that like to thank Dr. Arnold Quickstein for his um fantastic mentorship as well as collaborators Dr. Jing Lee and Dr. Chen who actually contributed substantially for this project and many pro many collaborators at UCSF and founding sources from the brain initiative as well as other sources. Happy to answer questions as well. Thank you for your attention.

All right, we have just a few minutes left for questions. Um, so please if you have any questions, submit those in the chat box in the Q&A function. And one question that we can start with is what are some ways um that others can use some of these larger data sets to influence their own work or how can they use some of these data sets in their own labs?

I'm happy to take this one if no one else is speaking up. Um I know that there's a lot of resources and that there's a lot through the brain initiative as well as through um the Allen Institute and these include places to download data. You can get them from Nemo um which is also repository and there's many interactive pieces. Um I also want to highlight that others have also engaged with the UC Santa Cruz cell browser um which is also an interactive tool that requires no informatic expertise to engage with it but then also if you want to download the underlying objects. So hopefully each of the papers and the brain initiative and by websites have links to many of these tools that allow for both interaction as well as download and engagement with the um underlying data.

The links to Nemo are in the first or second question in the question box too, but I think Arena is right. They're in the the resources as well, the papers.

And this webinar is being recorded. It'll be up on the uh Byan YouTube channel afterwards. So, if you've missed some of the URLs that were linked in the slides, don't worry. This will be posted online later so you can review and and pause at those screens and get some of those URLs. I'd like to make a a maybe perhaps final comment. I think there are a lot of questions about the resources you know with all these different great data sets uh how how they can be uh accessible. Right now the the many of the data are um being made available in the individual labs or an institute you know that can be found in the individual papers but I think you know in banan collectively we should have a a you know some kind of a central page to list all the availabilities of the different resources uh data sets and things like that. you know this is something that we should work on going forward already in

the plan but but uh it will be realized in the coming year. All right I just saw in the chat that we had a last question was just answered by Leewan which disease association analysis did you use and the response was we apply a method called um scavenge for which details can be found in the link there. All right. So, thank you all of our speakers for a great webinar today and all these wonderful talks. And again, this will be up on the YouTube channel afterwards if you'd like to rewatch or revisit some of these slides.

And thank you everyone for attending today's webinar. Thanks everyone.

Questions about this episode

Who speaks in "New Frontiers in Understanding Brain Development"?

Aparna Badori (Main speaker), Carlo Konuani (Main speaker), Yongsung Kim (Main speaker), Yuan Ga (Main speaker), Cindiva Belvin (Main speaker), Emily Han (Main speaker), Marilyn Styart (Main speaker), Harris Kaplan (Main speaker).

What is discussed in "New Frontiers in Understanding Brain Development"?

In order: 0:00 Brain Initiative, Nature, Carlo Konuani, Allen Institute; 10:00 Aparna Badori, Brain Initiative, Nemo Analytics, Nemo; 20:00 Brain Initiative, UCSC Cell Browser, UCSF Cell Browser; 30:00 Brain Initiative, Yongsung Kim; 40:00 Allen Institute, Yongsung Kim, Yuan Ga; 50:00 Allen Institute, Cindiva Belvin; 1:10:00 Allen Institute, Emily Han, Marilyn Styart, Thomas Noacowski; 1:20:00 Harris Kaplan, Thomas Noacowski; 1:30:00 Camille Mannins; 1:40:00 Lee Wong; 1:50:00 Brain Initiative, Allen Institute.

What is mentioned in this Allen Institute episode?

20 people, works and ideas are mentioned, including Brain Initiative, Aparna Badori, Allen Institute, Carlo Konuani, Yongsung Kim, Yuan Ga. Each one on this page links to its own page with every episode that mentions it.

How long is "New Frontiers in Understanding Brain Development"?

The episode runs 1 h 59 min; the transcript is about 17,644 words in 143 timestamped sections.

More from Allen Institute

About this transcript: the episode's spoken words as captioned on YouTube, split into timestamped sections. Reclip.Me locates every person, work and idea in the text, keeps the time it is first named, and quotes the words around it verbatim. Last updated January 24, 2026.