Channel: Translational Neuropsychiatric Genomics clear
| # | Channel | Title | Views | Likes | Cmts | Score | Sentiment | Duration | Transcript | Link |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Translational Neuropsychiatric Genomics | 14. Kristen Brennand - iPSC in schizophrenia research | 324 | 11 | 33.6 | 17:18 | And it's my pleasure to introduce Kristen Brinah from the Ikan School of Medicine at Mount Sinai in New York. She's a stem cell biologist and specializing in in-biter work and IPSEs as a model of skits of freeing the floors as yours Kristen. Thank you so much. Thank you. I can't share my screen until Robert stops though. Okay. So it is my my great pleasure. I think no. Yeah, no, you should be possible. There we go. Share and okay. It is my my great pleasure to get to share our work with this community today from the comfort of my own kitchen. And so what I want to talk with you guys about today is the work that we've been doing to apply stem cells to explore the genetics underlying psychiatric disease. And I'll note there's tiny little icons in each slide that there's a green camera feel free to screen grab or tweet it and if there's a red please don't. So why are we using stem cells to teach us about something as complicated and holistic as psychiatry. I think that there are two things that we're really hoping to provide to the field and and one is a new avenue by which to potentially help better diagnose patients. Genetic based diagnosis I think is something that's very obvious to this group. And we really think we have a tool to help functionally validate and link these variants that you guys are uncovering to their function in a cell type specific in donor dependent manner. And I think second and bigger picture is the real goals of course not to diagnose patients but to treat them and to do that either after the onset of symptoms or ideally prior to the onset of symptoms. And so with genetic based diagnosis we can really start thinking about drugs that one might deliver to high risk kids prior to symptom onset. And so what we're talking about here is precision medicine whereby we consider all of the risk factors and the interactions between them and any given patient and how those impact not just clinical outcomes but treatment response. And this slide is is nothing new to any of you but I want you to consider how you would feel the stem cell biologist. Somebody who really didn't know the full complexity of the genetics behind this data and and somebody who wanted to functionally validate a hit. And so this is why I'm so grateful to work with fantastic geneticists like Pell and McClar and now you like Stau and Laura Huckins because the stem cell biologist I'm not sure if you can see my cursor. I was of course drawn to that very very tall one at the top of the graph and they stopped me. They said this was a horrible locus to go after and points us instead at this one here at the gene skewering. And again they really walked me through understanding why of the entire so at the time it was the 108 but it's still true now at 145 loci. At 145 genome wide significant loci what happens at the fear and locus is the only time it happens and that's that you have the single top snip for GWAS. That's the Y-axis is also the single snot top snip for brain EQTL of a coding gene. So at fear and you have that one snip in the top right corner that is most significant for both disease and brain expression. And so that top snip here is a putative causal snip and it was one that we thought we might be able to edit by CRISPR. Now the next best examples looking more like what we see is SNAP 91 or you have 20 or 30 SNPs clustered in the top right hand corner. We don't know what's going on in that top right hand corner. We don't know if there's one put a putative causal snip in there or of each of these 30 SNPs are considering contributing three percent of the risk at the locus. So what I'll talk to you about today is the stem cell and CRISPR-based strategies by which to interrogate risk at locuses that look like or low side that look like fearing or SNAP 91. And so the the the SNAP nine or the fearing work was led by a very talented former postdoc in my lab all forever call her the bravest postdoc in my lab. I showed you on the previous slide the best data for the fearing locus but now what I'm showing you is the other side of the same data. So this is the expression in the common mind consortium brains by genotype at RS4702 the fear and snip length is schizophrenia. And what you can see is yes it is quite significant because there's 600 brains but the air bars there are daunting right and and the reason these air bars are so big is of course each of these brains came from a different individual. Each of these individuals varied at many many locations beyond just the fear and snip. I see some questions here I'll leave those till till the end. So each of these each of these donor brains varied at many genotypes genome wide. They came from donor some of whom had schizophrenia some of whom did not who had different and psychotic treatments at the time of death different drug and alcohol abuse histories. They died at different ages of different causes and had different ring values. And so our hypothesis was that if we could do this as a controlled experiment in the lab whereby nothing was different except for AA to GGG type our hope was that the effects has to be the same but the standard deviations would be much smaller. And so that is what Nadine undertook to test. Now this was a very hard edit to do we started she started this back in like 2014 I think. And the reason it was hard is we couldn't use any of the standard tricks that were developed at the time to increase the efficiency of CRISPR editing and that's because this was a single non-coding snip and editing a second single non-coding snip downstream in the PAM was just going to confound the analysis. So when she found a way and she got these edits and when she did she was able to show that in isogenic neurons from the same donor cultured in the same plate side by side when nothing more was different except AA to GGG type you could actually see a significant change in expression of the urinal levels. Moreover in the time that it took her to make these edits it was discovered that the snip in the 3-prime UTR appearance also in the binding side of a microRNA near 338. And if you inhibit near 338 in those same neurons now you can completely eliminate this EQTL effect. So it's a cell type context dependent EQTL in neurons. She was further able to show that doing nothing more than changing GGG type of a single non-coding snip was enough to manipulate neuron or right length and neuronal activity. We've been moving forward asking questions now about background. So these first edits were all done in the control background of average polygenic risk and this is now unpublished data led by a new postdoc in the lab Christina Rebick and we're asking to what extent to be snip effects vary if you do the same edits in in controls with particularly low or particularly high polygenic risk risk at syrphonia. So the same edit in both backgrounds. These are in a repository of IPS about 1200 controls from the serum collection in Kevin Egg and the Steve McAerla helped us find these extremists. And this is just our first pass. The very first pass. She's got edits now in low average and high PRS donors. This is that same activity assay and it looks like we're more sensitively resolving AA to GGG and type changes and those high PRS donors. There's a lot more work to be done here but I think it's a really exciting avenue to consider that we can now really tease out cell type and donor specific gene types effects. Now and beyond that there are so many more genes to look at than just fearing it and if you're not able to find a single period of causal to edit you need another strategy. And so here for the the next genes in the top of the list using CRISPR activation or inhibition to manipulate from the endogenous promoter the expression of these top genes and we can do it in the disease relevant direction by taking into account the EQTL direction. So here I'm just showing you by RNA seek that whether we do CRISPR A for two of these top genes, not anyone in T snare. The top gene less changing is this is a target gene, not anyone in T snare but a few you know other genes are also significantly changing and those other downstream genes that are changing seem to be enriched for brain genes, prickly, synaptic genes. We can see changes in synaptic punk die number by up and down regulating snap 91 in T snare and finally reciprocal changes in snap 91 levels up in the top panel and down in the bottom we need to reciprocal changes in a synaptic activity by electrophysiology up in the top and down in the bottom. Now of course common variants don't change in isolation. The really interesting questions what happened and they changed them together. And so here what Nadina and Sokmano did together as they did individual and combinatorial perturbations of these seem for genes in the disease relevant direction. So I'll of course not running one T snare and CLCN3 and down for fear and we have RNA seek of these individual perturbations. You can computationally predict what you think should happen when you change them all together and then you can actually do a combinatorial perturbation side by side on the same plate and ask how well the model predicted the outcomes. And so if the model is exactly right we call that no synergy but sometimes genes were too much up or too much down relative to predictions. So in fact about 82% of the time of genome wide analysis genes changed exactly as they expected the additive model predicted but 7% of the time genes were too much down and 11% of the time they were too much up. What types of genes were too much were down relative to expectations. Well they were actually synaptic genes and all the major no transmitter classes were captured here and those that were too much up while they're enriched for disorder genes particularly the rare and the common variance associated with bipolar and schizophrenia. So what we're really saying here I think is that environment context matters it's important to look at these genes in the combination of the others and here we're just looking at four common variant genes but the next experiment is moving on or 20 genes ideally 100 200 of these genes we really have to understand the interactions between these genes to get a handle on common variant effects. Now of course everything I've talked to you about now are based on the the proximal targets those genes close to this niche but the DNA is not packed into the cell in a straight line we know it's compacted and is compacted in a very deliberate and organized cell type specific fashion. And these loops can bring those schizophrenia risk variants close together to genes that are much more distant in the linear space. So this was work led by Priscilla Mergerion and collaboration with Cheryl McBarrion where he applied high c3d mapping to consider cell type specific interactions in stem cell derived neurons glia and neuroprogeny ourselves and so if there were 224 genes close to the 145 schizophrenia lowest I at the time he added a few hundred more by looking at cell type specific I see he looked at RNA importing interactions in this group and so on rich men's but he also functionally validated that these long range interactions were manipulating expression of schizophrenia risk genes. So here there was a low side where there was four schizophrenia GWAS SNPs about a hundred kilo basis away from the protein here in alpha low fias use CRISPR to delete those SNPs and was able to show that deleting these wrist SNPs changed expression of two of the three genes of the low stress in the disease relative direction. Finally with the last few minutes that I have a lot of change gears and talk about rare variants. So these are those highly penetrant where mutations in this case deletions that are very likely to result in in disorder. I'll talk to you about some work on directs in one so it's the second most significant schizophrenia where deletion the most significant for autism the only one of the ones on this slide that impacts a single gene but not to make it simple this deletion has a non-recurrent it varies between cases the boundaries and norexinal one is predicted to be spliced in hundreds of different ways. So this is a small cohort because they're rare variants collected in collaboration due to the rap report the study was led by Aaron Flayerti in collaboration with Shuzha Zhu a former postdoc in got my collaborator to go and find slab. So one of the first things Aaron was able to show is that patients from neurons had or neurons from patients had reduced neurite outgrowth. So here I'm showing you four patients the two in blue have five prime deletions where the promoter and first two exons are missing and the two in red have three prime deletions mixing missing the second third and fourth from last exons. The blue ones are in mother sun pair the red ones are mom's erotic twins and again we look at neural activity using the same as I mentioned earlier for the urine. You see reduced activity in patients drive neurons so this is a time course where you're monitoring population wide activity the controls go up with time and the patients really don't. This is a standard differentiation with a mixture of glutamaturgic and gapedrogen neurons but if you do it through a different method that yields entirely glutamaturgic neurons you see the same effect so patient cells don't fire as much. It's the question of being why they fire less because one critical rectal one isoform is missing because all the isoforms are decreased by 50% or because there's new mutant isoforms not seeing any controls. We added long range and short sequencing of the rectal one locust to ask this question we were able to identify 31 unique mutant isoforms from the three prime cases not found in the controls and then purple were indicating abundant so you can see that these mutant isoforms are some of the most abundant in the patient neurons. We found evidence that 49 or 50 50% of the wild type isoforms were significantly decreased in patient neurons and other 28 wild type isoforms were not even detected in the patient neurons but again these are somewhat at least abundant in the controls and so overall there's a mass dysregulation of the rectal one splicing repertoire. I'll add the mutant isoforms were not found in postmortem brain and most of the controls were especially the most abundant ones finally here and some functional studies to look at the impact of overexpressing these isoforms in isolation so if this is control it's neural activity on the left adding a single wild type isoform does not change activity but adding one of any four mutant erectile isoforms dramatically reduces neural activity so control neurons are impaired by a mutant isoform expression. There's five prime patients who are not predicted to have mutant isoforms have reduced activity that is rescued by overexpression of any one of four different wild type isoforms whereas the three prime patients the ones who do have mutant isoforms we can't seem to make them better whether we overexpress wild type isoforms and we can't make them worse if we overexpress mutant ones and so overall we think phenotypes are happening through two independent mechanisms both the loss of nerex and one dose and the additive effect of mutant isoform activity i think this might be more broadly applicable because if you go back into the PGC rare variant data set there are 35 nerex and one cases that have been gene-type and we think about one in five of them might be making mutant isoforms so the future studies then would be to synthesize these predicted mutant isoforms in these other cases and ask what effect they have so overall what we're really envisioning here with stem cell models is bridging the gap from gene-type to function and so i think stem cells and crisper are really good at understanding the link between DNA and RNA levels and the links between RNA levels and snap-dick function the ultimate goal of doing drug screens either the level of gene expression or snap-dick function such that we can predict patient treatment options and with that this is my final and very most important slide these are all the people who would normally be in a lab hard at work getting stem cells instead a trapped in small apartments in Manhattan and very eager to hopefully one day get back in the lab but without them there really would be no data to share with any of you today and i'm so grateful for each of them for their hard work today's talk was led by work done by Nadine Aaron Prashan and Suk in collaboration with Gabriel Gong and Shrong thank you thank you Gerstin some of this was amazing um we've got one question in the q and a for you and it's from a one-wide one um rich region of the gene fearing like x-on-inchon or promoter is the SNP RS4720 located in sorry for not making that more clear it's in the three prime utr which is why the mycoronate binding site seemed more relevant there | ↗ |