Scientist Stories: Barbara Treutlein, Tracing and Perturbing Cell Lineages

So next, our next speaker is Barbara Treutlein from ETH Zurich in Switzerland. So Barbara, I'd like to welcome you to share your slides and tell us about tracing and perturbing lineages during human brain organoid development.

Thank you very much, Gary. I hope you see my slides, yes? So yeah, it's my pleasure to present our work on tracing and perturbing lineages during human brain organoid development. We are very excited about organoids. They were mentioned in the previous talks already since they can model human development in the dish. And by doing so, we can actually model the development of any specific individual in the dish, we can perform time course measurements for the same individual. These organoids are genetically tractable, so we can perturb them and we can control the environment they grow in and we can perform high-throughput screening. And so this makes these organoids a very exciting human model system to complement the primary developmental cell atlases and with organoid data. And so I want to focus today fully on brain organoids, cerebral organoids. And in the past, we have used high-throughput single-cell transcriptomics to profile the development of these organoids over time. And what we could show is that from an initial very homogeneous pluripotent cell state, these cells move through different progenitor states and then give rise to different neuronal populations. That reflect neurons in different regions of the human brain. So we have excitatory neurons in the human cortex, we have inhibitory neurons in the ventral telencephalon, the ganglionic eminence, and then we also have neurons developing of the diencephalon, the mid and the hindbrain. And today I want to talk about two projects that try to go— that extend this previous work simply profiling transcriptomes. Where we in the first part now go into gene regulatory mechanisms and we try to infer and perturb sulfate regulomes during brain organoid development. And I want to already here acknowledge two fantastic PhD students that have done this work, Sophie Jansen and Jonas Fleck. And in the second part, I want to present a bit more shorter our new tools that we've developed to directly measure lineage relationships paired with single-cell transcriptomics in cerebral organoids. And here, this has been really great teamwork by 4 people in the lab, Chi Song He, Ashley Maynard, Tobias Gerber, and Akanksha Jain, and I already want to acknowledge them here. So let me start with the first part where now we wanted to go beyond the simple transcriptomic profiling. And we wanted to also measure chromatin accessibility over human brain organoid development in order to get at regulatory networks. And so we generated a new dataset where we have paired single-cell RNA-seq and single-cell ATAC-seq measurements for organoids from 4 different lines over a dense time course of early brain organoid development. And you can see the embedding here. We actually find the matching of RNA and the ATAC. We create metacells that contain both modalities and then integrate also the lines and time points.. And now we have a very dense sampling of these very early states that cells go through as they start to differentiate into the different neuronal cell types. And we have expression accessibility and can also quantify transcription factor binding motif abundances. And so if we look at the accessibility, we can actually see that early on in the pluripotent cells, accessibility is highest. And as the cells move through development and start to restrict their fate and differentiate, accessibility becomes lower and lower, and we can identify very specific regions of the genome that are accessible and are acting as enhancers during the different stages of development, including patterning as well as neuronal— neuroprogenitor cell differentiation and neuronal differentiation. And so since we cover this interesting and exciting time period of fate specification, we wanted to reconstruct the trajectories and really identify the bifurcations that are happening here and the fate decisions. And here's a pipeline that we use to do that. And just looking at this output, which is a graph abstraction of the data, where we can identify that there's an early branching of neuroepithelial cells into non-talencephalic and telencephalic progenitors, and then a later branching of telencephalic progenitors into dorsal and ventral telencephalic cells. And so what we wanted to do now using this highly high-dimensional data from the two different modalities is to build or infer a gene regulatory network that underlies regional fate-specific replication in these organoids. And for that, Jonas actually developed a new framework that he calls Pando, where we— that is based on a generalized linear model. And the different aspects that go into this is the accessibility of chromatin, also the existence of transcription factor binding motifs in these regions. The abundance of transcription factors, the expression level of transcription factors, as well as the abundance of target genes. And so all together, and actually also considering conservation, known cis-regulatory elements, and using transcription factor binding site predictions, we can establish a gene regulatory network from this paired data.. And so here you can see the TRN for this data on human brain organoids. And overall, we see, you know, TFs. So this is just showing the TFs. Each TF is connected to a whole regulome that includes other TFs and downstream genes, non-TF downstream genes. And you can see generally we have pluripotent TFs here, then we go through neuroectoderm and neural progenitor cells, and then we see different TFs that are quite specific to ventral telencephalon, dorsal telencephalon, as well as non-telencephalon and mesenchyme. And so what we wanted to do now next was to actually perturb this network and understand what really the role is of these different transcription factors. And we wanted to use a multiplex approach, a higher throughput approach, not going gene by gene. And what we did was to use the CROP-seq approach where we targeted 20 transcription factors that each has a very specific expression during cortex development, not just in the organoid cortex, but also the primary human cortex. You can see the transcription factors here. And then we infect inducible Cas9 iPS cells with lentiviruses that contain 3 guides per gene. We make mosaic brain organoids, and then we do single-cell RNA-seq and amplicon sequencing to measure the transcriptomes, but also the abundance of the guides and the detect the guides in the cells. Then we move on to do analysis of both fate changes— what is the fates that emerge, are there changes in the abundance of fates— and also state changes. Are there deviations from the normal state that are manifested by differential expression? And here I go very fast over this, but we see a lot of guides in all these cells, and we can look at composition changes. And just highlighting one gene here, Glee3, that very specifically leads to enrichment of the dorsal lineage and depletion of dorsal lineage and enrichment of ventral lineage. And we followed up on this by generating isogenic knockout lines for Glee3 and control lines, grow organoids and perform single-cell transcriptomics at 2 months.. And what we found was that indeed, knockout of Gli3 leads to a full lack of the cortex. So we can confirm the finding in this multiplex experiment. And moreover, what we wanted to do is now really look into the regulome that defines the dorsoventral patterning of the telencephalon, because we see that this is perturbed when Gli3 is knocked out. And what we did here was to choose the time point where really this fate decision is made, grow organoids from control and Gli3 knockout lines, and then do single-cell multi-omics. So for the same cell, measure RNA and ATAC from these organoids. And what we can see is that we see a lot of regulatory regions that are differentially accessible nearby peaks that are differentially expressed. And when we overlay our perturbation measurements, the differential expression we measure, onto the subnetwork of the Glee3 transcription factor from our DRN, we can see that this, the inferred direction from the DRN, is highly consistent with the measurements of DE that we have. And so this shows that our DRN is really meaningful, and it also now reveals the gene regulatory network or the regulome that that is downstream of Glee3 and that controls dorsal ventral fate specification in the telencephalon. And I'm running out of time here, I will go very fast over the second part. I should mention this first part, it's a preprint that actually came out just today, so you can check it out more in that preprint. The second part, also a preprint that was published last year, where we developed a new tool to directly measure lineage relationships in cerebral organoids. And this tool is called iTRACEr. What we do is on one hand have barcoded GFP in the iPS cells, where from day 0, we record a lineage. And then in an inducible manner, we can introduce a second lineage mark by targeting Cas9 to GFP and introducing SCARs.. And so then we have a two-level or two-channel recorder here where we can record lineage at two time points. How does this look? So here you can see single-cell RNA-seq data, a transcriptome embedding where you see all the different fates that emerge. Again, the major branches, telencephalon, diencephalon, mesencephalon, and rhombencephalon. And we can now in addition infer these whole organoid lineage maps where We can measure barcode families that mark all the cells that derive from one cell, one original cell in the embryoid body. And then we have scar family that mark a family of cells that derive from one cell at the time point of scarring. And we use this tool now to analyse more the lineage relationship between the cells during organoid development. And what we could see is that there is an enrichment of barcode families in different brain regions. So for example, here a family enriched in the telencephalon and another one really highly enriched in the hindbrain regional identity. And we could actually confirm that this links to spatial enrichment of lineages when we did spatial transcriptomics on these eye-tracer organoids. And so this really confirms that there is a clonality of the brain regions that develop in these brain organoids. And we could then finally confirm this when we performed light sheet measurements on these organoids, tracking them over time, so kind of a 4D measurement, and directly measuring the lineage relationship between cells. And when we quantified this data, we could see indeed there is clonality of these luminal regions as the brain organoid develops. And so with this, I'm at the end. I'm happy to take questions if there's time, and you can read more about both projects in these preprints. Thank you. And I should mention that both projects have been performed in collaboration with Gray Camp at ITB in Basel.

Thank you. Thank you very much, Barbara. That was great. So we will wait for questions. Just remind everybody that The video feed might be delayed up to a few minutes depending on where you are in the world. So, and also the speakers, if there are questions that come up after you've left, you can answer them using the comment feature in the platform. But while we're waiting, I just wanted to ask one question. You know, I'm wondering how variable these processes are. We know that mutations occur during normal development. And I guess we don't really know a lot about what, you know, how those affect development, but, you know, do you— it seems like if you repeat your experiments on different organoids, you might see, you might get some information about that.

I'm wondering if we know about that. Yes, so of course in these mosaic experiments also, you know, every mutation is in a way different, and so So this is definitely something that we could address with these types of experiments. I think in order to really understand the heterogeneity, we would maybe need more data on different or on a lot of different batches. But yeah, I think this is interesting to look at, at the variability. We try to extract whatever is a homogeneous response by all the mutations that we induced, right? But it would be interesting to also go into the analysis of the variability for each individual gene.

Okay, great, thank you. So yeah, just a reminder to speakers, if you see questions coming in after, you can answer them in the online chat or the the question system, so I don't see any other questions, so thank you again, Barbara, and[No speech]