Scientist Stories: Xin Jin, In Vivo Perturb-seq & Scalable Investigation of Gene Functions

Hi everyone. I'm so honored and excited to be here as part of the Epoch webinar series on Young Innovators to a Better World. Thank you very much for having me. Our lab is interested in understanding the brain, and particularly we have been developing tools and systems to understand our diverse genetic backgrounds and genetic variants that can contribute to brain function. What we are interested in can be essentially summarized in the schematics here, that we want to know how are these diverse genetic variants and changes in our genome can actually impact our brain function as well as dysfunction. And particularly in the brain, we know that there are heterogeneous cell types and cellular populations. We want to understand for each of these genetic factors, what are their cellular actions among this heterogeneous population of cells? And within these individual cell types, we also want to understand how these genetic alterations will be able to change the cellular states. How are these genetic factors maybe able to alter, for example, gene expression and transcription. And ultimately, a new direction in my own lab is to be able to extend from the cells to the neighboring systems, to be able to understand whether any of these genetic factors can alter cells' ability to be located in certain locations in anatomy, and also understand how these perturbations in one cell may be able to change the cellular neighborhoods and their ability to perform these cell-cell network interactions. So, all these development of tools and systems are enabled by the advancement of human genetic knowledge. And our tools are hugely motivated because of the incredible advancement of human genetic knowledge. Now we know that many of these devastating neuropsychiatric disorders have strong genetic roots. That we know precisely a list of genes and particular genetic loci that is relevant to these devastating diseases. For example, in the case of autism spectrum disorder and developmental delay, one syndrome in case of autism called the Rett syndrome is actually caused by the loss of function of a single gene named MECP2. This is groundbreaking work by many people in the field, including Huda Zomby. Now we know that MECP2, this is a gene that encodes a transcriptional modulator that changes the gene transcription, which is how the DNA transcribed into RNA. But if we switch our attention to a different autism risk gene, For example, in this case, a gene called SYNGAP. And in this case, SYNGAP gene actually encodes a GAP protein, operates on the synapses to regulate receptor trafficking. Essentially, it is a gene that's essential for regulating neuron-neuron communication. And you can see that this is an entirely different biological process compared to transcription. That I showed you earlier. So in the past, there are many of these single genes that we already know the function strongly implicated in neuropsychiatric disorders. And these single gene studies are very important as they are building important and foundational models for us to approach this complex question of neuropsychiatric disorders. But The question is there is also a very long list of these genetic contribution of risk genes to these psychiatric disorders. Estimably, there are about 400 to 500 risk genes contribute to these diseases. So a few years ago when we started, our question is among all these devastating, among all these risk genes that implicated in devastating disorders such as autism and developmental delay, how many different kinds of potential cellular actions there are. Is it possible that each of these hundreds of risk genes actually present an independent question? Or is it possible that they actually converge and diverge into different subtypes? We already know that many of these genes actually express broadly across the nervous system in many different cell types. So the question is, is it possible that some of these risk genes actually impact the same cell type, maybe at the same time of development or even within the same molecular pathway within that cell type? And these will be very important information to know when we're trying to better model these disorders and cluster them together for consideration for future therapeutics. And to start, we really want to focus on this particular risk population of risk gene variants called the de novo variants. These are typically genetic changes that carry large effect sizes, which are really excellent entry point to understand these psychiatric disorders. You can see that 10 years ago, we know about less than a dozen of these genes. But right now, we already have more than 200 risk genes with strong significance that has been shown to contribute to autism and developmental delay. And particularly, 71 of these genes actually passed the genome-wide significance, which is the level of resolution and confidence that we never had before. This is really the true— this is really a true force by the field, including many labs and different institutes through a huge amount of work and collaboration. And now what I'm showing you is a long list of genes that we know they are reproducible and protein-truncating variants, that their functions are essentially tied with these disorders. Right now, the challenge is whether we can draw common mechanisms and potentially being able to cluster them into subtypes for therapeutic considerations. And it is really important to also think about that this genetic breakthrough is not just limited to autism and neurodevelopmental delays. There are many other disorders such as bipolar disorders and schizophrenia that have increasing confidence of gene lists that is coming out. It is important to think about whether we can establish a generalizable approach to scalably and systematically study gene function that is relevant to human disorders and diseases. Motivated by these questions, a few years ago, we proposed this approach that we hope is a generalizable approach to be able to leverage the cutting-edge gene editing technology using CRISPR-Cas9 system. And we want to do so by introducing these genetic perturbations and changes in vivo through viral administration. And then days and weeks later, we will be able to harvest these cells after in vivo perturbation in the living cell in a living organism. And we want to subject analysis of these cells at some level, kind of high-resolution readout that will allow us to capture the cell type-specific effect that is associated with each of these genetic perturbations. And the goal of this approach is that we will be able to test the functional effect across many different perturbations and also being able to test them across all the numerous individual cell types in an organ. And brain is an organ that is famous for its cellular heterogeneity, which of course presents another level of complexity. So I want to highlight that it is important to think about why we need high-resolution readout here. And through this very simple analogy that if we want to consider the tissue we study, as shown here on the slides on the left, a brain made of LEGO. And without high-resolution readouts, if we are performing bulk analysis such as bulk RNA sequencing or bulk proteomics, what we ended up doing is to take different LEGO brains, break them into pieces, and compare this blend of pieces from one to the other. But you can imagine that the differences between a functioning brain versus a brain in the disorder states may be much more subtle. And these bulk analyses are much harder to capture these subtle changes that might be crucial to understand the pathology. And the concept and advancement of single-cell genomics is to not only be able to analyze the cell type composition and the molecular changes. Most importantly, we're analyzing these puzzle pieces one at a time so that we will be only comparing the red pieces from this brain to the red pieces on a different LEGO brain. And this way we may be able to have a much higher resolution and in-depth understanding of these brain differences. Such as maybe the color of the red in one brain is slightly darker than the other one, or the number of the yellow pieces in one brain that's different from the other one. So that is the Lego analogy. How does it really work in a real experiment? So to be able to study gene function in the development in a scalable way, we were very fortunate to stand on the shoulders of giants and build our experiments on the already established perturbed system. We can introduce these genetic perturbations in genes that we cared about through introducing guide RNA and injecting this pool of guide RNA in the lateral ventricles in a developing embryo early at embryonic day 12. And this is such an early stage of corticogenesis that you can imagine a subset of these neuronal progenitor cells will be transduced by these lentivirus and therefore express the fluorescent protein as well as the CRISPR loss of function. And in the subsequent days and weeks, all the cellular progeny that came out from this neural progenitor will also carry the blue fluorescent protein as well as the CRISPR loss of function. At our postnatal day 7, a chosen time of analysis, what we can do is to take the tissue and use fluorescent sorting to enrich for these blue perturbed cells, and use single-cell RNA sequencing to capture the transcriptomic profile of each cell. And I want to highlight that importance of single-cell genomic approach, that it not only gave us a high-resolution readout to capture the cell type diversity, but also allow us to multiplex this approach. That is to say that we don't have to introduce one perturbation in one brain at a time, but rather we can mix all the perturbations that we're interested in and inject them as a pool. And only at the data analysis stage, we will be able to take each of these cells and ask, what are their perturbation identity and what are their cellular transcriptomic content? Ultimately, through this kind of approach, our question is, among all these different risk genes that implicated in the disease, what are the cell types that's most vulnerable? And what are the molecular pathways that's most convergently impacted across many of these perturbations? And this approach obviously has many advantages in the way that it is scalable. That we can understand dozens of gene functions at a time in one single experiment. And theoretically, we can always scale up to perform these kind of experiments with hundreds or thousands of genes or even genome-wide. And I want to highlight that precisely because of the efficiency of this approach, these perturb-seq approaches are also more well-controlled because you're always comparing the perturbation effect against a control cell within the same brain that was dissociated in the same experiment, that was sequenced in the same batch. And it really allows us to build a more robust platform, regress against any other possible confounds. Such as sequencing batch or timing effects for different experiments. In total, in this pilot experiment that we published 2 years ago, we have focused on the effect of 35 different autism risk genes and harvested more than 46,000 cells through more than a dozen experimental batches. Long story short, Here, I want to highlight the 20,000-foot view of what we've learned from this body of work. In this analysis, we learned that although we started only with 35 different risk genes, we already start to observe the recurrent and cell type-specific effects by many of these sets of risk gene perturbations. And many of these effects are actually specific to a cell type, but not observed in other cell types. Again, highlighted the importance and the power of the single-cell resolution to capture these phenotypic changes, assign them to the appropriate cell type that could have been masked if you were doing bulk analysis. As a neuroscientist, I should also admit that I was Starting this experiment looking for effects in the neurons. But to our surprise, we found that both neuronal and glial cells are directly impacted by several of these risk gene perturbations. And highlighting the importance of thinking about how to better model these diseases and disorders, especially when in vitro, where the cellular diversity is more limited. We compare our data with developing human brain data and show that many of these gene modules and networks that we extracted from our data are also conserved in the human brain. And ultimately, we hope to convince you that this is also a versatile and scalable screening system and allow us to use systems biology and systems genetic approach to understand the brain function and dysfunction. That is relevant to human diseases. So I want to come back to the cartoon that I showed you earlier to think about the tissue that we are studying as if it's a brain Lego and the importance of using single-cell genomics to extract high-resolution phenotypes when they are subjected to different genetic perturbations. But important things is to register that compared to these single cells and clusters of cells that basically dissociates and lost their spatial coordinate in an experiment, we must respect that the tissue we study, the brain, is inherently a spatially organized and well-patterned organ. And in this case, where the cells located it, how their cellular neighbors are, actually carries meaning. Meanings. And if we truly want to understand the function of a brain and using genomic tools, we're actually losing the most important components of the spatial resolution. And in my new lab, what we are trying to establish is to approach the study using a 3-dimensional organ-wide analysis to capture these spatial resolution as much as possible. So just to summarize, we are very excited to extend the single-cell genomics and building the next generation of Perturb-seq with additional features and additional readouts. We are excited in adapting CRISPR-Cas9 gene editing as well as additional epigenome modulation methods that allow us to not only cut genome and introduce a loss of function, but fine-tune the gene expression and ask what are their effect. We are also interested in analyzing additional tissue types at additional time points in development just to generalize a more complete picture about how the perturbation of a given gene sets can alter the organ-wide action at different time points. And ultimately, we are interested in and continue applying single-cell genomics and multi-omics in conjunction with spatial transcriptomic and high-content imaging to capture the cell-type-specific effect associated with each of the genetic perturbation and trying to grow from the cell-intrinsic effect to their cell-neighborhood effects. And we're very excited about this direction of research in our lab. And hopefully, by using these scalable approach, we can work collectively to expand our understanding into a comprehensive, holistic, dynamic, and unifying view of how our brain work. Last but not least, I'd like to thank my mentors, and collaborators without whom I simply won't be able to tell you the story that was published before. And importantly, I'm also very grateful to have a group of young scientists that joined me in my very young lab at Scripps Research. We started last July and we are actively recruiting. If there's any interest in joining us, If you have any additional questions, please feel free to reach out to me. Thank you very much for your attention.