Using Stem Cells to Explore the Genetics Underlying Brain Disease

Transcript

• 00:00Grand Rounds It's a pleasure to see
• 00:01so many of you here in the colon and
• 00:04welcome to everyone joining us on Zoom.
• 00:05Also, we'll start today like we start
• 00:08every week with just some reminders
• 00:10about upcoming Grand Round sessions.
• 00:12So next week we'll hear from
• 00:14Carrie Epstein and Brianna Browser.
• 00:16He'll talk about their trauma focused
• 00:18treatment And so exceptionally this
• 00:20presentation will be fully virtual.
• 00:22So please do join us via Zoom next week.
• 00:25And then the following week on February 6th,
• 00:26we'll be back here in the cone
• 00:28for compassionate care rounds.
• 00:30And so do join us for that.
• 00:31And then for any of you joining
• 00:33us today and because of the
• 00:34developmental neuroscience focus,
• 00:35I'll make a plug for our
• 00:37speaker on the 13th of February.
• 00:38Doctor Stacy Bilbo,
• 00:39who's joining us from Duke.
• 00:42And as many of you all know,
• 00:42Stacy has done some really elegant
• 00:44work trying to understand the
• 00:46impact of prenatal stress there
• 00:48is on brain development with a
• 00:50particular emphasis on microglia.
• 00:51Now back to today's speaker.
• 00:53It's a pleasure and a
• 00:54privilege introduced Dr.
• 00:55Kristen Brennan to the Charles City
• 00:57Center community for grand rounds.
• 01:00Now as you'll hear and appreciate
• 01:02from Kristen's presentation,
• 01:03she really has blazed a trail in
• 01:05the use of stem cells to understand
• 01:08complex and psychiatric disorders.
• 01:10Publishing a landmark paper
• 01:11in Nature in 2011,
• 01:13holding a postdoc with Rusty Gage
• 01:15and then in her own independent
• 01:17research group in Mount Sinai in the
• 01:19Pamela Sklar Division of Psychiatric
• 01:21Genetics and now here at Yale.
• 01:24Since 20,
• 01:2421 has really pushed method
• 01:26development and the integration
• 01:28of computational neuroscience with
• 01:30stem cell technologies to understand
• 01:33and the basis the biological basis
• 01:36for complex psychiatric disorders.
• 01:38And and if you speak to Kristen
• 01:40the other thing you'll appreciate
• 01:41from your conversation is that
• 01:43she places a strong emphasis on
• 01:45mentorship and training the next
• 01:47generation of scientists and she
• 01:48Co directed he was the founding Co
• 01:50director of the YSM and Science
• 01:52Fellows program which really tries
• 01:54to provide a structured mentorship
• 01:56program and a pathway to independent
• 01:59faculty positions for those from
• 02:01communities that are traditionally
• 02:03underrepresented in biomedical sciences.
• 02:05And so without any further ado,
• 02:08please join in welcoming Kristen
• 02:09Brennan to the Child Study Center.
• 02:15Thanks, Karen. I'm, oh,
• 02:17I got to keep admitting people.
• 02:19OK. I'm really excited to be here.
• 02:23To those of you in the room
• 02:24and also those of you on Zoom.
• 02:29And I'm going to talk today about
• 02:31something that's probably as far from
• 02:32most of your clinical work as can be,
• 02:34but it's using stem cells to explore
• 02:37the genetics of of brain disorders.
• 02:39And I'm hoping that over
• 02:41the the course of the talk,
• 02:44you'll begin to appreciate maybe
• 02:46how little we've done to change
• 02:47your clinical treatment today.
• 02:49But the potential all these models
• 02:52to hopefully one day change the way
• 02:55that your clinical work is proceeding,
• 02:56is somebody else admitting all these people.
• 02:58Awesome. Then I'm going to stop.
• 03:00OK, you're on. It sounds good.
• 03:02I'm going to hide that then.
• 03:06OK. So and I think to this room,
• 03:09this introduction is obviously overly basic,
• 03:11but I do think it's,
• 03:13it's fundamental to, you know,
• 03:15stop for a moment and talk
• 03:17about the fact that psychiatric
• 03:18disorders are incredibly common.
• 03:20They, in fact they affect about
• 03:22one in five people across the
• 03:24country and around the world.
• 03:26They're, they're they're severe.
• 03:29The disability adjusted impact of
• 03:32psychiatric disorders actually
• 03:34exceeds nerdogenerative disease,
• 03:36which I think a lot of people intuitively
• 03:38always think about as a a bigger problem.
• 03:40In fact, psychiatric disorders
• 03:42come right behind cancer in terms
• 03:44of of worldwide disability.
• 03:45They're diverse.
• 03:46The types of disorders that fall
• 03:48under this umbrella, I think,
• 03:51range from autism and psychosis and
• 03:53bipolar to eating disorders, anxiety,
• 03:55depression and substance abuse.
• 03:57So some of them more rare,
• 03:59some of them more common,
• 04:00but very different clinical presentations.
• 04:03And the two introductory facts
• 04:05that I think really ground my
• 04:07discussion today are these next two.
• 04:09The first one being the average delay of
• 04:12time between symptom onset and treatment,
• 04:15which ranges from 8 to 10 years
• 04:17in this country.
• 04:18Now this is far too long for patients
• 04:20to be experiencing debilitating
• 04:21symptoms impacting the quality of
• 04:23their life without even having a
• 04:26diagnosis to explain what is going on.
• 04:29And the second fact that I think
• 04:31is atrocious is that 60% of adults
• 04:33with psychiatric disorders did not
• 04:35receive treatment last year, right.
• 04:37We have got to do better.
• 04:39We have got to diagnose better
• 04:40and we've got to treat better.
• 04:41And that includes finding new treatments
• 04:43that don't have the side effects
• 04:45of all of our existing approaches.
• 04:47Now, the next thing I want to point out,
• 04:50of course,
• 04:50is that psychiatric disorders are heritable.
• 04:53We know this because there are
• 04:55families that have, you know,
• 04:56a greater percentage of affected
• 04:58individuals than other families.
• 04:59Now, this alone is not definitive, right,
• 05:02because we share DNA with our families.
• 05:06But I want to remind you that we also
• 05:07share environments with our families, right?
• 05:09So if you come from a family of affluence,
• 05:12you're like more likely to have
• 05:14healthier food, better environments.
• 05:15And if you come from a family,
• 05:17from a disadvantaged population,
• 05:19you might live in environments
• 05:21that have increased exposures.
• 05:24OK, OK.
• 05:25I trust you or you know,
• 05:27you know, decreased, you know,
• 05:30quality of foods,
• 05:31decreased exposures through work.
• 05:33And so it's very hard but critical
• 05:36to differentiate these genetic
• 05:38and these environmental effects,
• 05:41impacts that conflate and
• 05:43confound each other.
• 05:44What I can tell you,
• 05:45and I'll start with
• 05:47schizophrenia as an example,
• 05:48is that we know we can estimate from
• 05:51genetic studies that the heritability
• 05:53of schizophrenia is as high as 80%.
• 05:56Now that doesn't mean we can
• 05:58explain all of the heritability.
• 06:00It just says that we you know think
• 06:02from the the best and the largest
• 06:04twin studies that 80% of the cause
• 06:06is in the DNA that we're born with.
• 06:08Now we can explain about 20,
• 06:11like 30% of that heritability through
• 06:13genome wide association studies.
• 06:15We were looking at common
• 06:16variants and underlying it.
• 06:17To expand this across the
• 06:19spectrum of psychiatric disorders,
• 06:20you can see that there are some
• 06:22disorders that are highly heritable,
• 06:23like schizophrenia,
• 06:24Autism spectrum disorder, ADHD,
• 06:26and bipolar are are all over 75% heritable.
• 06:29But reciprocally there are some
• 06:31that are extremely not heritable.
• 06:33And at the bottom end of this plot we
• 06:35have major depressive disorder and PTSD.
• 06:37This should surprise nobody that these,
• 06:39you know,
• 06:39disorders that are in fact you know,
• 06:40often defined by an environmental exposure
• 06:43like PTSD are indeed less heritable.
• 06:46But whether at the top or the bottom,
• 06:48our ability to explain that decreasing
• 06:50heritability is is constant.
• 06:52We can explain about 1/4 to at
• 06:54most 1/3 of genetic heritability
• 06:56for any of these disorders,
• 06:58going back to schizophrenia.
• 07:00To explain how complex this
• 07:02known heritability is,
• 07:04this plot is plotting the frequency
• 07:06in the population on the Y axis
• 07:09and the impact towards diagnosis,
• 07:11the effect size on the X axis.
• 07:14And so to start with,
• 07:15we've got these green dots.
• 07:17These are the oldest,
• 07:18longest known genetic variants
• 07:19linked to schizophrenia.
• 07:21Their copy number variations that are
• 07:23either large deletions or large duplications.
• 07:26They were first identified almost 15
• 07:27years ago now and they're highly penetrant.
• 07:30If for example,
• 07:31you're born with 22 Q 11.2 deletions,
• 07:33you are extremely unlikely to
• 07:35be a neurotypical control,
• 07:37but they're plea atrophic.
• 07:38So you know you you could end up
• 07:41diagnosed with schizophrenia or autism
• 07:44or neurodevelopmental disorders.
• 07:45So that's the green dots,
• 07:46highly penetrant,
• 07:47but not specific to any specific disorder.
• 07:51The red dots are were identified
• 07:53much more recently,
• 07:54in fact in the last year or two.
• 07:55These are protein,
• 07:57protein truncating variations.
• 07:59They're much more rare,
• 08:00they're about tenfold,
• 08:01less common in the population,
• 08:02but they're just as penetrant.
• 08:04So in total,
• 08:05all of the known green and red dots
• 08:07explain less than 5% of schizophrenia.
• 08:09So they're very penetrant when you have them,
• 08:12but most of your cases don't carry
• 08:15these dilutions or duplications.
• 08:16In fact,
• 08:17most of the known genetic risks for
• 08:19schizophrenia is in these blue dots.
• 08:21They're extremely common variants.
• 08:22Each of us as controls were born
• 08:25with over a dozen common risk
• 08:27variants for schizophrenia.
• 08:29Your cases, though,
• 08:30will have over 60 common variants
• 08:32associated with schizophrenia,
• 08:34so all of us have them.
• 08:35But each common variant confers less than
• 08:38a 1% increase risk for schizophrenia.
• 08:40So very small effect sizes that
• 08:42seem meaningful only in aggregate.
• 08:44On the right is the most recent.
• 08:46You know why the association
• 08:47study for schizophrenia.
• 08:48So this is our G wash plot for schizophrenia.
• 08:52Briefly,
• 08:52the Y axis is the P value,
• 08:54the significance of any given loci.
• 08:57And that line you see at 10 to the -8,
• 08:59that's genome wide significance.
• 09:00So every line that goes over every
• 09:03vertical line that goes over that
• 09:05horizontal line is significantly
• 09:06associated with risk for schizophrenia.
• 09:09But they're not actually lines, they're dots.
• 09:11And each dot in those vertical
• 09:13lines is one single DNA base pair,
• 09:15one single nucleotide polymorphism.
• 09:17And so it's actually incredibly
• 09:20challenging at some of these loci to
• 09:22know which DNA variants are actually
• 09:25attributable to schizophrenia risk.
• 09:26And so the first major question
• 09:28that we have outlying is how
• 09:30do we translate risk variants.
• 09:31So again, here you might have thousands
• 09:33of risk variants at chromosome.
• 09:35I think that's chromosome 6,
• 09:37the MHC cluster, and maybe only one
• 09:39of them is like the schizophrenia,
• 09:41whereas like this loci is much more clear,
• 09:44right?
• 09:44There's only six or seven of
• 09:46these snips linked here,
• 09:47so we might have a better.
• 09:48I guess that this top one is it.
• 09:50But top is most significant doesn't
• 09:52actually mean it's causal because
• 09:54we don't scramble our DNA well when
• 09:57when eggs and sperm come together.
• 09:59So major question #1,
• 10:01how to translate risk variants to their
• 10:03target genes and their cellular function?
• 10:06Question #2 How do all of these
• 10:08hundreds of risk variants,
• 10:10And again, there's 250 risk
• 10:12variants here for schizophrenia.
• 10:13For depression,
• 10:14we already have over 350 risk variants.
• 10:16How do all of these risk
• 10:18variants interact and sum?
• 10:20And in fact,
• 10:21do these interactions explain
• 10:23that unknown misinheritability?
• 10:25And I think the really interesting
• 10:27question then is how do these
• 10:29interactions and summings underlie
• 10:31variable penetrance and expressivity?
• 10:33Why can two individuals have
• 10:35the exact same genetic risk,
• 10:37but really different,
• 10:38different,
• 10:38different clinical presentations,
• 10:40either not having any symptoms
• 10:42altogether or changing severity
• 10:44despite sharing the same genetic risk?
• 10:47And so my talk today is really about
• 10:49how do we use genetics and stem
• 10:51cells to answer these questions.
• 10:53The first goal that we have
• 10:55is improving diagnosis.
• 10:56If we could predict who was had
• 11:00schizophrenia from their DNA,
• 11:02right, we would add an extra
• 11:03measurement for all of you,
• 11:04right?
• 11:05Like I think all of you would
• 11:07appreciate blood based biomarkers
• 11:08or DNA genotype that tells you
• 11:10definitively that somebody has
• 11:12schizophrenia versus bipolar or that
• 11:14somebody's high risk or low risk, right?
• 11:16Like this would be informative.
• 11:18You have them in cancer,
• 11:20we have them in many human diseases.
• 11:22We don't have them in psychiatry.
• 11:25And second of all,
• 11:26I'd like to use all this genetic information
• 11:29to improve prevention or treatment.
• 11:31Now again,
• 11:32if we are so good by use,
• 11:34at one day at using DNA
• 11:37to identify diagnosis,
• 11:39our DNA stable across our lifetime.
• 11:41So that means instead of waiting
• 11:42till your patients show up in
• 11:44the clinic with first onset,
• 11:45we could actually,
• 11:46you know,
• 11:46accurately predict who might
• 11:49have these disorders at birth.
• 11:51And if you can define who is high
• 11:53risk and who is not at birth,
• 11:55if you've changed the window
• 11:58for therapeutic intervention.
• 11:59I think it's really intuitive to
• 12:01everybody that if we are going
• 12:02to have Alzheimer's one day,
• 12:04I know for myself,
• 12:05I would much rather treat my Alzheimer's
• 12:08when I'm 50 prior to neuronal death,
• 12:11then when I'm 80, after half of
• 12:13my cortical neurons have died.
• 12:15But it should also be intuitive that
• 12:17the type of treatment might change
• 12:19as we expand that therapeutic window.
• 12:21At age 80, any drug is just focused on
• 12:24keeping my remaining neurons alive.
• 12:26But at age 50, we might actually
• 12:29target microglia activation and
• 12:31try to prevent neurons from dying,
• 12:33try to prevent inflammation.
• 12:34And so if you can expand the therapeutic
• 12:37window by improving diagnosis,
• 12:39you also change the cell types and
• 12:41the pathways that you might have
• 12:43as possible therapeutic targets.
• 12:45And So what I'm really talking about
• 12:46is the idea of precision medicine.
• 12:48How can we take knowledge of all
• 12:50the genetic risk variants carried
• 12:52by any individual as well as the
• 12:54interactions between those variants
• 12:56and use that to best predict the drug
• 12:59that will improve their symptoms?
• 13:02I want to really quickly ground
• 13:04my talk about stem cells in what
• 13:06we know about the human brain.
• 13:08And so I'm going to summarize 50
• 13:11years of human research in in
• 13:14one overly simple slide.
• 13:15But the first thing that I want
• 13:17you all to know is that the brains
• 13:19of patients with schizophrenia,
• 13:21on average,
• 13:22are smaller than the brains
• 13:24of neurotypical controls.
• 13:25We knew this decades ago from
• 13:28autopsy studies where literally
• 13:30the brains were just weighed.
• 13:32And we know this more recently
• 13:33from brain imaging studies.
• 13:34So these are images from Judy
• 13:36Rappaport's group at the NIH
• 13:38almost gosh over 20 years ago now.
• 13:40But the areas in red are brain
• 13:42regions that are smaller,
• 13:43but the brain regions are not
• 13:45smaller because neurons are dying.
• 13:47They seem to be smaller because
• 13:49the neurons themselves are smaller.
• 13:51So this again is postmortem imaging,
• 13:53neuronal reconstruction,
• 13:54showing that cortical neurons are
• 13:56smaller in patients with schizophrenia,
• 13:58and in fact they have fewer synapses.
• 14:00There's fewer connections between neurons
• 14:02in these brain regions that are smaller.
• 14:05And so this is the first
• 14:06truth that I want to hold,
• 14:08that neurons from patients with
• 14:09schizophrenia seem to be less
• 14:11well connected to each other.
• 14:12But there's so much that we still don't know.
• 14:15We don't know if these neurons are the
• 14:16cell type of origin for schizophrenia.
• 14:18Is this where the disorder starts
• 14:20or does something happen Right,
• 14:22Because this is end stage disease.
• 14:23These are patients who are decades
• 14:25and decades old,
• 14:25who probably have had decades of treatment.
• 14:28Where did this start?
• 14:30In their adolescence or in childhood?
• 14:32Which cell type and when and and
• 14:35these are the questions that
• 14:37I'm really interested in.
• 14:39You know it it's it's kind of
• 14:41preposterous that we actually
• 14:42don't know this already right?
• 14:43That we don't know what cell type
• 14:46goes wrong 1st and why and how.
• 14:49And the reason that we don't know
• 14:50this in truth is that there's
• 14:52just not enough live human brain
• 14:54tissue for studies of schizophrenia
• 14:55and for drug discovery.
• 14:57You know,
• 14:58and and in truth across human disease,
• 15:00with the exception of cancer,
• 15:02where patients are, you know,
• 15:04begging and paying their doctors
• 15:05to cut the material out, this is a,
• 15:07you know, a common phenomenon.
• 15:09There are mouse models of
• 15:12psychiatric disorders.
• 15:14But I like to joke that I was
• 15:16not trained as a neuroscientist.
• 15:18And Despite that,
• 15:19I can tell the difference between
• 15:21a mouse brain and a human brain,
• 15:23and I still can tell that
• 15:24difference when we control for size.
• 15:26So there are some fundamental
• 15:28differences here that might
• 15:29not be captured in this model.
• 15:31Now what are the differences
• 15:33that I'm most worried about not
• 15:35being captured in mouse models?
• 15:36So again,
• 15:37mouse models are fantastic at
• 15:39revealing the interactions between genes,
• 15:41circuits and behaviors.
• 15:42But there's no such thing as a
• 15:45perfect model and and so where
• 15:47mouse models actually fall apart
• 15:50is looking at non coding common
• 15:51variants and these are the exact
• 15:53variants I told you are the greatest
• 15:55contributor. Psychiatric risk.
• 15:56Now why'd the mouse models
• 15:57probably capture this?
• 15:58Well these are actually the variants
• 16:00that are not conserved between
• 16:02humans and mice and so it's very
• 16:04hard to look at the impact of a risk
• 16:06variant that doesn't even exist in mice.
• 16:08Second of all,
• 16:09mouse models are not great at looking at
• 16:12complex interactions between variants.
• 16:14I don't know many.
• 16:15How many of you had the pleasure
• 16:16of working with mouse models?
• 16:18But putting even 2 trans genes
• 16:19together always took me a lot
• 16:21more than the 8 offspring it
• 16:22was supposed to to get a match.
• 16:23And we're talking about looking at
• 16:25the interactions of dozens of risk variants.
• 16:27And I don't think there's a grad
• 16:29student out there willing to
• 16:31look at crosses of 12 or 20 or 50
• 16:33strains trying to get to complex
• 16:35interactions so difficult to engineer
• 16:37and even more difficult to breed.
• 16:40And the final thing is, you know,
• 16:43I don't know how we would ever look
• 16:45at Priya Trophy in a mouse model.
• 16:46What is the difference between
• 16:48schizophrenia and autism in a mouse,
• 16:50right?
• 16:50What is the difference between moderate,
• 16:53severe and mild autism in a mouse?
• 16:56How do we really get AT variable
• 16:59penetrance and expressivity in rodent models?
• 17:01And so these are the types of
• 17:03questions that I most want to
• 17:04ask with our cell based models.
• 17:06What are the complex interactions
• 17:08between common variants to impact
• 17:10cellular phenotypes in vitro?
• 17:12And so when we talk about modeling
• 17:14brain disease and psychiatric disorders
• 17:16with stem cell derived neurons,
• 17:18what I'm really talking about is
• 17:20the fact that since 2006 with
• 17:21the discovery by Shinya Yamanaka,
• 17:23we could reprogram skin cells or
• 17:25blood cells from anybody on the
• 17:27planet and to induce pluripotent stem cells.
• 17:29And these IPS cells as I'll call them,
• 17:32have the capacity to make every
• 17:33cell type in the body.
• 17:35And then in 2011,
• 17:36with the discovery of CRISPR engineering,
• 17:38I can now at scale and relatively
• 17:42easily introduce or remove genetic
• 17:45variants linked to disease.
• 17:46And so I can add risk variants
• 17:49to control cells. I can take away
• 17:52risk variants from patient cells,
• 17:54and then I can make them into all
• 17:56the major cell types of the brain.
• 17:57And I can ask across people and
• 18:00across genetic manipulations,
• 18:01what is the impact?
• 18:04Now before I start telling
• 18:05you about these comparisons,
• 18:06I think it's important to pause
• 18:08and ask how good are the cells
• 18:10that we're making relative to
• 18:11those found in the human brain.
• 18:13So we've done this analysis over
• 18:14and over time and again and we keep
• 18:16getting about the same results.
• 18:18So this is a version of the analysis
• 18:20that we did back in 2017 based
• 18:22on all the available RNA seq that
• 18:24we could find from the postmortem
• 18:26brain and from stem cell models.
• 18:28So I'm coloring here in blue
• 18:31about 800 postmortem brain RNA
• 18:33seek samples that came from Gtex,
• 18:37from the Allen Brainspan Atlas,
• 18:39and from the Common Mind Consortium.
• 18:41They all tend to cluster together
• 18:42in the same, you know,
• 18:44left hand part of the plot you can see
• 18:46over here in red these are blood cells.
• 18:48Over here are skin cells.
• 18:50Over here are those induced
• 18:52pluripotent stem cells.
• 18:53And then I took all the neural progenitor
• 18:55cells and neurons that my lab had ever made,
• 18:57as well as two other labs,
• 18:59Tracy Young Pierce's lab at Harvard and
• 19:01Hong Joon Song's lab at John Hopkins.
• 19:04And all of our stem cell derived
• 19:05neurons and NPCS clustered together.
• 19:07The NPCS are in green and
• 19:09the neurons are in orange.
• 19:10And what I think you can see is
• 19:12that these cells don't pile up
• 19:13right on top of the brain cells,
• 19:15but they do overlap.
• 19:16And the cells that are overlapping
• 19:17are our neurons.
• 19:18And what they're overlapping with are the
• 19:20fetal brain samples here in darker blue.
• 19:23And So what we learn from these
• 19:24analysis is that the cells that we make
• 19:26from stem cells are actually fetal.
• 19:28Like they're immature.
• 19:29This makes sense.
• 19:30We cultured them in the dish
• 19:31for about 3 months,
• 19:32and they look like late 1st trimester,
• 19:35early 2nd trimester cells.
• 19:36It's not that my lab is uniquely bad at this.
• 19:39Every neural stem cell lab has done this.
• 19:41The only way to make cells that
• 19:43look like birth neurons is to
• 19:45culture organoids for over a year.
• 19:47It's not just the neurobiologists
• 19:48who are terrible at this.
• 19:49You see the same thing whether you're
• 19:51making vascular cells or heart cells.
• 19:53We reset age and we make stem cells,
• 19:55and age gets reset back at the same rate
• 19:59that we set it as developing humans.
• 20:02Turns out,
• 20:03Mother Nature's been making people for,
• 20:05you know,
• 20:05hundreds of thousands of years,
• 20:07and she does it as fast as she can
• 20:09and we're not any faster at it.
• 20:12And so I don't like to say we
• 20:13have disease in addition models.
• 20:15I really prefer to say that we have
• 20:17disease risk in a dish models.
• 20:18We're modeling predisposition to disease,
• 20:21not the disease state.
• 20:22But I actually think that's a
• 20:23really informative place to study and it's
• 20:25really different from postmortem approaches.
• 20:27So going back in time,
• 20:29what can I tell you about stem cell neurons?
• 20:32So here I'm showing you representative
• 20:35images from stem cells,
• 20:37neural progenitor cells,
• 20:38and neurons derived from neurotypical
• 20:40controls in cases with schizophrenia.
• 20:43This is really old data,
• 20:44but I like to show this to make one point,
• 20:46which is at this level of magnification,
• 20:48you shouldn't be able to see any differences
• 20:50between neurons from cases and controls.
• 20:51I can't.
• 20:52And that's really important.
• 20:53Patients walk and talk and breathe,
• 20:55just like you and I.
• 20:56If I was up here telling you we
• 20:58couldn't make neurons from patients,
• 20:59that should be concerning.
• 21:01The differences that we
• 21:02see are incredibly subtle.
• 21:05One of the differences that the
• 21:07neurons had fewer branches.
• 21:08This should remind you of the
• 21:09difference that I had previously told
• 21:11you about from the post mortem work.
• 21:13And the neurons have not just fewer
• 21:15connections between each other,
• 21:17but fewer functional connections
• 21:18between each other,
• 21:19so there's less synaptic activity.
• 21:21And neurons derive from cases versus
• 21:24neurons derive from controls.
• 21:25So this is great, but we don't know why,
• 21:28right?
• 21:28So do the molecular causes underlying
• 21:31these in vitro phenotypes represent
• 21:33genetic risk factors for schizophrenia?
• 21:36Or you know,
• 21:37put another way,
• 21:38can we decipher the genetic risk factors of
• 21:40schizophrenia using these stem cell models,
• 21:42right.
• 21:42Are these neurons acting differently
• 21:44in vitro for the very same reasons
• 21:46that patients with schizophrenia act
• 21:48differently in real life? Right.
• 21:50Does this model capture genetic risk?
• 21:53And so that's where you know our our
• 21:55lab pivoted about 10 years ago now.
• 21:57So this is work done by a former postdoc lab,
• 21:59Nadine Shroud where she worked with the
• 22:01the Gwas that existed in circa 2015,
• 22:04the PGC 2.
• 22:05There was about 145 genome wide
• 22:08significant risk loci at this period
• 22:10and she wanted to test one of these
• 22:12common variants to see if we could
• 22:14see a difference in the neurons.
• 22:16Now at the time,
• 22:17nobody had ever tried to manipulate
• 22:19these common variants and I'll
• 22:20remind you that each of them
• 22:21confers like 1% increased risk.
• 22:23So could we see an impact in vitro?
• 22:26Everybody told us that we couldn't do it.
• 22:28And so I'll forever call Nadine the
• 22:30bravest postdoc in the lab because she
• 22:32didn't care what everybody else thought
• 22:34and she just went ahead and tried.
• 22:36Now this going ahead and tried
• 22:37it took her two years.
• 22:39And so,
• 22:40because we knew it was going to be so hard,
• 22:42we really want to pick the
• 22:44best variant in this plot.
• 22:46Now for those of you who are not geneticists,
• 22:49I was really tempted to go after this
• 22:51dot here just because it's so tall.
• 22:53And I was really fortunate to work
• 22:55with some brilliant geneticists
• 22:56who told me to stay clear of this
• 22:58because the tallest plot does not
• 23:00necessarily mean the best candidate.
• 23:02We were looking for the plot
• 23:03where we knew the
• 23:05single snip was most significant.
• 23:07So you're really looking for a plot
• 23:08where you've got better separation
• 23:10between the top snip and the others.
• 23:12And So what we did is we asked 2
• 23:15questions and we asked of all the
• 23:17SNPs that were genome Y significant,
• 23:19which ones were most likely to regulate
• 23:21expression of a nearby target gene.
• 23:23And so here we're intersecting that
• 23:25G wash the genetic study from over
• 23:28here on the Y axis with a brain
• 23:30postmortem RNA C study on the X axis.
• 23:32And of those 145 genome Y significant loci,
• 23:38only one time did we see a plot this clean,
• 23:41which is to say only one time was there a
• 23:43single dot in the top right hand corner.
• 23:45So this single dot is a putative causal
• 23:48snip for schizophrenia because that
• 23:50single DNA variant is the most significant
• 23:53at that loci for schizophrenia risk.
• 23:55But it's also the most significant DNA
• 23:57variant for regulating expression of a
• 24:00nearby target gene, in this case fear.
• 24:02In the second best example.
• 24:05Oops.
• 24:06Oh, and the and the SNP is called RS47-O2.
• 24:08The second best example happened at
• 24:11this gene of SNAP 91.
• 24:13And what you can see here is that
• 24:15in this case we had about 20 or 30
• 24:17snips in the top right hand corner,
• 24:19all implicated in schizophrenia
• 24:20and regulation of SNAP 91.
• 24:22We don't know which one to edit.
• 24:24Is there one causal SNIP in that cluster
• 24:26or are all 20 or 30 of those causing risk?
• 24:29And so we're going to use different
• 24:31CRISPR edits for these two case examples,
• 24:33our top two ones.
• 24:35So coming back to the Spheron Rs 4/7/02,
• 24:37it took Nadine about two years
• 24:39to achieve a perfect edit.
• 24:41But in a control donor line she
• 24:43was able to take the non risk
• 24:44variant A A and turn it into AGG.
• 24:46And when she did so the first thing that
• 24:48she saw was that fear and expression
• 24:51was in fact down in those GG risk cells,
• 24:53exactly like we had predicted.
• 24:56In the two years that it took
• 24:57her to the edit,
• 24:58it was learned that this SNP which
• 24:59is in the three prime UTR of the
• 25:01of the fear and gene regulating
• 25:02its RNA stability is actually in
• 25:04a microarny binding site,
• 25:05which is probably why this SNIP
• 25:06came out so significant.
• 25:08And if we inhibit that micro RNA
• 25:10we'd actually eliminate the effect.
• 25:11So now we have a context dependent risk
• 25:13variant that only confers risk if this
• 25:16micro RNA is expressed in that cell type.
• 25:18The name was also able to show
• 25:20that neurons with the GG genotype
• 25:21have fewer branches than neurons
• 25:23with the wild type genotype,
• 25:25and that these neurons have reduced
• 25:27activity relative to neurons
• 25:28with the wild type genotype.
• 25:30Until she's really been able to show
• 25:31that we can see differences when
• 25:33we have added just a single common
• 25:35snip on a on a control background.
• 25:38But I'll remind you this was the
• 25:39best example,
• 25:40and I don't know whether the second or
• 25:42third or 250th example would be this clear,
• 25:44but we have a tool where we can
• 25:46test these common variants.
• 25:47So we finished this work in 2019 and
• 25:50then something happened in 2020, right?
• 25:54We shut the lab down when
• 25:56the pandemic started,
• 25:56and we kept hearing reports on the news
• 25:59about how the major difference between
• 26:02the original SARS virus and Cyrus Kobe
• 26:05two was the introduction of a furin
• 26:07cleavage site in in Cyrus Kobe 2.
• 26:10And you know,
• 26:10we kind of laughed about it on Slack.
• 26:11Like, what are the odds our favorite
• 26:13schizophrenia gene is, like,
• 26:15responsible for this pandemic?
• 26:17And then we started thinking like, well,
• 26:19we actually could test if furin was
• 26:22important for this virus and if it was,
• 26:24maybe that would be helpful, right?
• 26:26Like, it's April,
• 26:27May of 2020,
• 26:29and what Christina really decided
• 26:30was that she was sick of being in her
• 26:32really small apartment in Manhattan and
• 26:34that she wanted to come back to work.
• 26:36And so because we had these tools
• 26:38and because we actually had a
• 26:40a really great collaborator,
• 26:41we'd worked with already,
• 26:42Ben Tanover at Mount Sinai,
• 26:43who was actually a virologist
• 26:45in putting SARS COV 2 on cells,
• 26:47we asked well,
• 26:49does Fiorin regulate expression of the
• 26:53receptor for SARS COV 2IN lung cells?
• 26:56And so remotely by zoom in
• 26:58collaboration with Daryl Cotton's lab,
• 27:00Christina learned how to make lung cells.
• 27:02She was able to show that these GG
• 27:05lung cells expressed less furin than
• 27:07their A A counterparts and that they
• 27:10were massively less susceptible
• 27:11to SARS COV TWO infection than
• 27:13their A A counterparts.
• 27:14In fact,
• 27:15so much so you can see it by imaging
• 27:17that these GG lung cells are
• 27:19less susceptible to SARS COV Two.
• 27:21Now it turned out the biggest predictor
• 27:23of who got severe COVID was not genotype,
• 27:26it was antibody repertoire.
• 27:27And so this was not a direction that we
• 27:30actually continue to pursue in the lab,
• 27:32but it was our first Test case of gene
• 27:34by environment interaction, right?
• 27:37There's nothing wrong with GG lung
• 27:38cells until you throw a bunch of stars,
• 27:40COVID 2 in the dish and then they all die,
• 27:42right?
• 27:42So you have a genotype and you
• 27:44have an environmental insult and
• 27:46together you see a phenotype that
• 27:48you didn't see alone.
• 27:49And that's really changed the direction
• 27:51of work in the lab ever since.
• 27:54The question that I'm really
• 27:55interested in asking is,
• 27:56can we modify the impact of genetic risk,
• 27:59right?
• 27:59Are we are genetic fate or are
• 28:02there pop multiple outcomes with
• 28:04the genetic risk we are born with?
• 28:08I think in college we were all
• 28:10taught about genotype,
• 28:10phenotype relationships,
• 28:11but these are not necessarily one to one.
• 28:16I think it's really intuitive,
• 28:17especially to you psychiatrists,
• 28:19that there are environments that
• 28:21can make phenotypes worse.
• 28:22You can be born with really high
• 28:25risk for opioid addiction,
• 28:26but if you never get your hands on any drugs,
• 28:30you won't have this phenotype, right?
• 28:32Abusive childhoods make
• 28:34psychiatric disorders more severe.
• 28:37They can make them onset earlier.
• 28:39Likewise, there are good
• 28:42environments that can minimize
• 28:45phenotypic outcomes of genetic risk.
• 28:47And So what I really want to do
• 28:49is begin to understand what these
• 28:51environments are and if we can
• 28:53understand the mechanisms by which
• 28:55they protect against disease.
• 28:56And and so I'm going to use the
• 28:59word environment throughout the
• 29:00rest of the talk pretty loosely.
• 29:02I'm going to say that the
• 29:04environment includes the other
• 29:06DNA around a genetic variant.
• 29:07So what's the impact of other
• 29:09risk variants on resilience?
• 29:11Cell type, right?
• 29:12How do different variants have different
• 29:14effects and different cell types?
• 29:15And finally, context,
• 29:16which might be putting drugs on a
• 29:19cell or stress hormones on a cell.
• 29:21And so with that,
• 29:22I'm going to dive first into an example
• 29:24of a rare variant where we think we
• 29:26can modify genetic outcomes and then
• 29:28into some common variants and how we
• 29:31think we can modify genetic outcomes there.
• 29:33So these are some of the rare variants
• 29:35linked to schizophrenia and autism.
• 29:37You'll know that there's actually
• 29:39more rare variants linked to autism,
• 29:41but there's one here that's
• 29:42in common on both of these,
• 29:44and that's a a deletion that
• 29:46on chromosome 2 two P 13.3,
• 29:48and it encompasses a single
• 29:49gene nurexanal one.
• 29:50In fact,
• 29:51it's the only copy number variant
• 29:54for schizophrenia or autism that
• 29:56encompasses a single gene.
• 29:57But it's really interesting
• 29:58for a few reasons.
• 30:00A it's pliotrophic,
• 30:01so this deletion can confer risk not
• 30:03just for autism and schizophrenia,
• 30:05but also OCD,
• 30:07epilepsy and several other diagnosis.
• 30:102 is variably penetrant so some people
• 30:14have minimal impact and and some you know
• 30:17are are are nonverbal and have psychosis.
• 30:212 nerexin one is the most highly
• 30:24alternatively spliced gene in
• 30:25the human genome.
• 30:26There are,
• 30:27from mouse studies,
• 30:28hundreds of distinct nerexin 1
• 30:30isoforms and nerexin one's a
• 30:32critical synaptic organizer,
• 30:34with the hypothesis being that
• 30:36each nerexin isoform will have a
• 30:38different post synaptic binding
• 30:40partner and might confer a
• 30:41different type of synaptic function.
• 30:43This is evidence from mouse
• 30:45showing us that we see hundreds
• 30:47of Nurexa 1 splice isoforms with
• 30:49more splice isoform complexity in
• 30:51brain regions like the cortex that
• 30:53have more cellular complexity.
• 30:56Likewise,
• 30:57nurexa 1 isoforms are sufficient to
• 30:59distinguish types of neurons in the brain.
• 31:01Again, this is mouse.
• 31:03So when we started this project back in 2015,
• 31:05we knew there were hundreds
• 31:07of splice isoforms in mice,
• 31:09but we didn't know if this was
• 31:11also true in human.
• 31:12And so this was work led by a
• 31:13former PhD student in the lab,
• 31:14Aaron Flaherty,
• 31:15in really close collaboration
• 31:17with my collaborator Gong Fang,
• 31:18and a postdoc, a former postdoc in his team,
• 31:20Shu Zhu.
• 31:21So our first question was,
• 31:22is Narexin one highly determinedly
• 31:24spliced in the human brain as well?
• 31:27And so here,
• 31:28and this was the best way
• 31:30anybody actually in the field
• 31:31is able to figure out
• 31:32how to show this type of data.
• 31:33So this is the Narexin 1 locus at the
• 31:35top each Exxon dot vertical Barb.
• 31:38And what we're showing you here in
• 31:39the very top row is the most abundant
• 31:41Neurexa 1 isoform that we found.
• 31:43At the bottom, the least abundant.
• 31:44This is long range sequencing, so we're
• 31:46reading each isoform start to finish.
• 31:48If an exon's included, it's colored,
• 31:50if it's skipped, it's white.
• 31:51So in this most abundant isoform,
• 31:52exons one and two are included,
• 31:54exons 3 is skipped,
• 31:554-5 and six are included and so on.
• 31:58Now we've colored them either orange
• 32:01or green and that's based on whether
• 32:03the samples came from a human brain
• 32:05tissue or stem cell derived neurons.
• 32:07If an isoform was detected in both
• 32:09the human brain and stem cell Dr.
• 32:11neurons,
• 32:11so we colored it orange and if it was
• 32:13only detected in the postmortem brain,
• 32:15we colored it green.
• 32:16And what I think you can see here
• 32:18is that the most abundant isoforms
• 32:20were actually shared in both.
• 32:21And so we think that we not only
• 32:23but we know that we can see hundreds
• 32:25of spice isoforms in the human
• 32:27brain and that to the sequencing
• 32:28depth that I could afford,
• 32:29that was also true in stem cell
• 32:31derived neurons that the most abundant
• 32:33isoforms are present in both.
• 32:34And so we thought we could model the
• 32:36impact of deletions in nerexin one,
• 32:38asking how it might change the
• 32:40nerexin 1 isoform repertoire.
• 32:42The last really critical thing that
• 32:43you need to know is that these nerexin
• 32:461 phenotypes are human specific,
• 32:47so the people who carry them
• 32:49are heterozygous deletions,
• 32:51but there's actually no phenotype
• 32:53in heterozygous mice.
• 32:54You have to have a full knockout
• 32:55of nerexin one to see behavioral
• 32:57or electrophysiological effects.
• 32:58This was a really clever experiment
• 33:00done by Thomas Sudoff a few years ago
• 33:02where he took skin cells from mice,
• 33:04reprogrammed those to stem cells,
• 33:05made those into neurons,
• 33:07and they did electrophysiology and
• 33:08showed that there was no effect of
• 33:10a heterozygous nerexin 1 deletion.
• 33:12But when he introduced the exact same
• 33:15mutation into humans stem cells and
• 33:17made them into neurons the exact same way,
• 33:19there was a phenotype.
• 33:20And so these are complex spliced genes
• 33:22that only show phenotypes in human neurons,
• 33:25not mice.
• 33:26And we have no idea why.
• 33:28But it's a really good example of
• 33:30a gene that has to be studied in a
• 33:32human context. So here's our human context.
• 33:34These are rare carriers and in
• 33:36collaboration with 2D rapport,
• 33:37we actually got skin samples from 4
• 33:40nerexin 1 heterozygous deletions.
• 33:42The last thing you have to know
• 33:43about nerexin 1 deletions in people
• 33:45is they're non recurrent.
• 33:46They incur in different parts of
• 33:47the gene in every instance.
• 33:49There's no hotspot and so we have
• 33:51a 5 prime deletion.
• 33:53This is a mom and her son who share a
• 33:56deletion in the promoter and 1st 2 exons,
• 33:59both of them are diagnosed with psychosis.
• 34:02And this is a pair of models that got like
• 34:04twins who share a deletion in the second,
• 34:06third and 4th from last exon of this gene.
• 34:09And so we've got stem cells from these
• 34:12four carriers and from 4 match controls.
• 34:14The first thing that I can show you is
• 34:16that from both the five prime patients
• 34:18who I'm coloring in blue and the three
• 34:20prime patients colored colored in red,
• 34:21but we see reduced neural branching.
• 34:24We also see reduced neural activity,
• 34:25again consistent with some
• 34:27of the earlier studies.
• 34:28But the question is why?
• 34:29What's the mechanism for this?
• 34:30To what extent do these phenotypes
• 34:33reflect perturbations in few or many
• 34:35Neurexin 1 isoforms?
• 34:37Put another way,
• 34:38are these neurons from the cases firing
• 34:40less because one really critical
• 34:42neurxin 1 isoform has decreased?
• 34:44And the heterozygous deletions,
• 34:45are they firing less because all 100
• 34:48nerexin 1 isoforms are decreased?
• 34:50Or are they firing less for
• 34:52some other reason altogether?
• 34:54And so we went back and we did
• 34:56sequencing of the nerexin 1 isoforms
• 34:58from the cases and controls.
• 35:00The first thing that I want to
• 35:01show you is that about half of
• 35:03the isoforms were decreased in the
• 35:05cases relative to the controls.
• 35:07Another third weren't even detected,
• 35:08which again I think is a
• 35:09sequencing depth thing.
• 35:10So we've got massive dysregulation
• 35:12of all of the wild type Neurexin
• 35:141 isoforms in these patients.
• 35:16But then something unexpected
• 35:18occurred as well.
• 35:19We've got 31 unique mutant isoforms
• 35:23caused by splicing around this
• 35:25deletion that we never saw in
• 35:28control neurons or in neurons
• 35:29from or in postmortem brain.
• 35:31And so you know,
• 35:33now the question is are the phenotypes
• 35:35from loss of these like 75 plus
• 35:37isoforms or the phenotypes from one
• 35:39or more of these mutant isoforms.
• 35:42And so we began to do exactly that.
• 35:44We tested this,
• 35:45we could take control neurons which are
• 35:48here in in oh in Gray and we can over
• 35:51express 4 different mutant isoforms
• 35:52and all of them were sufficient to
• 35:55decrease activity in control neurons.
• 35:57So control neurons can be
• 35:59impaired by mutant isoforms.
• 36:01In these five prime cases,
• 36:02the ones that don't have mutinisoforms,
• 36:05we can make them better by over
• 36:08expressing just one wild type isoform.
• 36:10But in these three prime cases,
• 36:12the ones expressing mutinisoforms,
• 36:13we cannot make them better by over
• 36:16expressing wild type isoforms
• 36:18and we can't make them any worse
• 36:20by adding more mutinisoforms.
• 36:22So we actually think there's
• 36:23two things going on here.
• 36:25In all cases,
• 36:26phenotypes reflect a loss of nerexin 1 dose,
• 36:29but in a subset of cases,
• 36:31there's an additive effect caused
• 36:33by a mutant isoform activity.
• 36:35Now how big a subset?
• 36:36Right,
• 36:36We were able to get 4 samples and
• 36:40half of them had mutant isoforms.
• 36:41That doesn't mean half of all cases.
• 36:43And so we went back into the genic
• 36:45data sets to ask how often do
• 36:47we think there might be patients
• 36:48who have mutant isoforms?
• 36:49And we think our best guess is about
• 36:52one in five that 20% of Narexan 1
• 36:54deletions might actually be producing
• 36:56mutant isoforms that can be translated.
• 36:59So a lot more work to be done to
• 37:01see how generalizable this is.
• 37:03But I think we weren't so lucky as to find
• 37:05the only example in in all of of humanity.
• 37:08And so this is where Aaron graduated and
• 37:11Michael Fernando picked up the project.
• 37:13Michael's really interested
• 37:14in precision medicine,
• 37:15so he really wanted to understand if
• 37:17we could do anything to help patients.
• 37:20We understand he's done a lot of work
• 37:22that I'm not going to talk about today,
• 37:23that there are different effects of
• 37:25nurexin 1 deletions in glutamatergic
• 37:27neurons and Gabaergic neurons.
• 37:29We understand very well that these
• 37:30wild type isoforms and these mutan
• 37:32isoforms are having different effects,
• 37:34the level of the synapse.
• 37:35But he wanted to know in the case
• 37:37of the patients that only have
• 37:39a loss of function mutation,
• 37:41can we rescue it by turning up
• 37:43expression from the one healthy allele.
• 37:46And so he looked for examples of of
• 37:48drugs that might increase expression.
• 37:51And it turns out that the by chip seek,
• 37:55the estradiol receptor binds
• 37:57just upstream of nurexin.
• 37:58And there's been a couple examples
• 38:00in the literature of estradiol
• 38:02increasing expression of nurexin.
• 38:04So he tested it in our cells.
• 38:06And so if you add estradiol
• 38:09to glutamatergic neurons,
• 38:10the first thing that you see is
• 38:12that yes indeed we can express
• 38:14increase nurexin one expression.
• 38:15And second of all,
• 38:17you can actually rescue activity all
• 38:19of the neurons from that five prime
• 38:21case just by providing beta estradiol
• 38:22and he could quantify it here.
• 38:24So it's a significant and robust rescue
• 38:27of synaptic activity by adding estradiol,
• 38:29that sex hormone to five prime neurons.
• 38:32But that doesn't tell us a lot about those
• 38:34three prime cases with the mutinisoform.
• 38:35And so here he's testing antisense
• 38:37oligos which have already been
• 38:39applied as you know as end of 1
• 38:41therapeutics for neurodegeneration
• 38:42and asked can we design Asos that
• 38:44might knock down mutant isoforms.
• 38:46He's been able to design one he
• 38:48can show it has knocked down.
• 38:49But we don't have any data
• 38:51on functional rescue yet.
• 38:52But this is kind of what we envision
• 38:54here moving forward is the type
• 38:55of precision medicine that might
• 38:57be possible once we understand the
• 38:58mechanisms at play for a patient.
• 39:02And so I hope that with what I
• 39:04showed you about estradiol,
• 39:05I've shown you one example of an environment
• 39:08that might make phenotypes less bad.
• 39:11And I think this is actually
• 39:12consistent with what we know about
• 39:14sex bias in brain disorders, right?
• 39:15So we know that males are at a higher
• 39:18risk for autism and schizophrenia,
• 39:20whereas females are at a higher
• 39:22risk for Alzheimer's disease,
• 39:24depression and anxiety.
• 39:26And so maybe sex hormones are one
• 39:28clue about how we can modulate risk,
• 39:30and why two people with the same
• 39:32risk factors might
• 39:33have different outcomes.
• 39:35We had changed gears now and pivot to
• 39:38environments that are less helpful.
• 39:40One thing that we know is that stress
• 39:42increases risk for psychiatric disorders.
• 39:44But that stress alone is not causal.
• 39:47About half of the US population will
• 39:50experience a trauma event sufficient of
• 39:53magnitude to be associated with PTSD.
• 39:56But half of us don't have PTSD, right?
• 39:59And so who are these people who develop PTSD?
• 40:03And do we are there?
• 40:04Ways of predicting or preventing that
• 40:07relationship between trauma and exposure.
• 40:10So one thing that we know is
• 40:12that our environment influences
• 40:14our susceptibility to trauma.
• 40:16People with unstable housing and
• 40:18less social support are more likely
• 40:21to have PTSD when exposed to trauma.
• 40:23But what about their DNA?
• 40:25Does their DNA also predict the likelihood
• 40:28of PTSD when exposed to stress?
• 40:31So the question being,
• 40:32do our genetics determine how
• 40:33we respond to traumatic stress?
• 40:36How do we model that in a dish?
• 40:37Well,
• 40:37we thought about it a lot and we
• 40:40decided to go after cortisol,
• 40:41which is an essential part of the HPA
• 40:43axis and something that we can just
• 40:45add to our cells in a dish, right?
• 40:47We can't put an environmental
• 40:49trauma on cells in an incubator,
• 40:51but we can apply cortisol.
• 40:52And what's nice about this is
• 40:54if we can control the dose,
• 40:55I mean, control the duration.
• 40:57And so this is exactly what we did.
• 40:58We're putting hydrocortisone and
• 40:59synthetic cortisol on neurons
• 41:01and monitoring the response.
• 41:03And we can do this to neurons
• 41:05derived from PTSD cases and controls.
• 41:08We can ask how they're different
• 41:10and how they're the same.
• 41:11So here's the first analysis.
• 41:13We're looking at transcatomic response
• 41:15to Low dose H Corp and high dose H
• 41:19Corp from neurons from 39 donors.
• 41:22I think the first one that you can
• 41:23see is that the more you stress neurons,
• 41:25the more they change gene expression.
• 41:26There's more blue in the high dose case.
• 41:29What types of genes change expression
• 41:30as you stress neurons where they're
• 41:32actually related to neurdevelopment and
• 41:34immune response. So this all makes sense.
• 41:36Now there's 39 donors actually
• 41:38came from a cohort recruited by
• 41:40Rachel Yehuda to the Bronx VA.
• 41:41They were all about combat exposed veterans,
• 41:44half of whom nineteen of whom had PTSD
• 41:46and twenty of whom after really careful
• 41:49clinical characterization did not
• 41:51and suddenly broke down that same data set.
• 41:53Now by diagnosis we actually kind
• 41:55of see the opposite result, right?
• 41:57There's more pink,
• 41:58there's more color at the low
• 41:59dose than at the high dose.
• 42:01What we think this data is telling
• 42:02us is that PTSD cases are hyper
• 42:05responsive to stress, right.
• 42:06So everybody changed at a high dose stress,
• 42:09but PTSD cases changed gene
• 42:10expression at low dose stress before
• 42:13the control neurons due.
• 42:14In fact their their their hyper
• 42:16response to stress was sufficient
• 42:18to separate neurons that came from
• 42:21cases versus controls and the types
• 42:23of genes that change in neurons from
• 42:25PTSD. Cases with stress that didn't change
• 42:28in neurons from controls with stress
• 42:30were enriched for known PTSD risk genes,
• 42:33both the genetic risk variants like the PTSD
• 42:35but also genes that were differentially
• 42:38expressed in PTSD brains postmortem.
• 42:40But that being said,
• 42:41it's not specific to PTSD risk genes.
• 42:44We also have a ton of developmental
• 42:46disorder risk genes, schizophrenia,
• 42:48epilepsy and autism risk genes.
• 42:51And so I actually think that stress
• 42:54broadly dysregulates psychiatric
• 42:55risk gene expression in neurons.
• 42:57Karina, Doug a little bit further
• 42:59trying to understand why neurons
• 43:00were hyper responsive to stress.
• 43:02Here's an example of a transcription factor,
• 43:04Mick, one where its targets were activated
• 43:07higher at increased stress doses and
• 43:09it's at the middle of a hub of genes.
• 43:12And so now we're asking
• 43:13if stress dysregulates,
• 43:14specifically psychiatric risk
• 43:16loci in neurons.
• 43:18So we talked about EQTL at the beginning.
• 43:20We prioritized furin.
• 43:21This is an example of,
• 43:23you know,
• 43:23one location where two nucleotides differ
• 43:26or one nucleotide differs at 2 copies.
• 43:28And you might see for example an increase
• 43:31in expression and that furin example
• 43:33the GG neurons had less remember and the A,
• 43:35A had more.
• 43:36Now with A, a stress interactive EQTL,
• 43:39you see differential stress
• 43:41relationships depending on genotype.
• 43:43The GG neurons show increased
• 43:45expression with stress and the A,
• 43:47A neurons would show decreased
• 43:48expression with stress.
• 43:49And so we asked across our
• 43:5140 donors how often we saw,
• 43:53you know,
• 43:54genotype dependent stress relationships
• 43:56in both glutamatergic neurons and
• 43:58separately in Gabaergic neurons
• 43:59to derive from those 39 donors.
• 44:01And what you can see is that there's
• 44:03actually an enrichment for psychiatric
• 44:05risk disorder loci in both populations.
• 44:08But again,
• 44:08not just psychiatric,
• 44:09We're also getting an enrichment for
• 44:11metabolic and inflammatory risk loci.
• 44:13So we're learning broadly about
• 44:15stress responsiveness and how it
• 44:17underlies a risk for human disease.
• 44:19I think,
• 44:21and I and I think really cool because you
• 44:23could argue that all of this is in vitro.
• 44:24What does this actually tell
• 44:26us about real brain cells?
• 44:27And so in collaboration with Matt Triganti,
• 44:30who has brains from 304 individuals,
• 44:331/3 of whom were controlled,
• 44:34a third of whom had depression
• 44:36and a third of whom had PTSD,
• 44:37we were able to replicate half of
• 44:40our stress interactive genotypes
• 44:42in postmortem brains from people
• 44:44who had trauma exposures, which,
• 44:46you know,
• 44:46I think was a much higher rate of
• 44:49validation than I had expected.
• 44:51And so now I've given you an example of
• 44:53an environment that makes phenotypes
• 44:54less severe and then stress as
• 44:56the example of an environment
• 44:58that makes phenotypes more severe.
• 45:00Now,
• 45:00how specific are these relationships
• 45:02to different stressors?
• 45:04And so now we want to ask if
• 45:06distinct stressors have unique
• 45:07impacts on the regulatory variants.
• 45:11Oops, I think I repeated this slide.
• 45:13OK, so here,
• 45:14how do we pick multiple stressors?
• 45:17Well, we know for example,
• 45:19that during pregnancy, fever,
• 45:21but also trauma and also famine
• 45:24and also infection can all increase
• 45:27risk for psychiatric disorders.
• 45:29Some of these can be mediated by Illinois
• 45:326 activation in the immune system.
• 45:35And likewise,
• 45:35we know in adolescents that stress
• 45:37and drugs of abuse can also increase
• 45:40risk for psychiatric disorders.
• 45:41So we're going to add a few different
• 45:43inflammatory cues to our cells,
• 45:45and we're going to ask at large scale.
• 45:47So this is at the thousands of risk
• 45:49varying simultaneously using an assay
• 45:51called a massively parallel reporter assay.
• 45:52Now, the details of how this assay
• 45:55works aren't really relevant,
• 45:56but what it will do is it will
• 45:58call a DNARNA ratio.
• 45:59So we can ask which risk variants
• 46:01are changing expression in different
• 46:02cell types in different contexts.
• 46:04And so a risk variant that has
• 46:06no effect on regulation will
• 46:08just give the same outcome,
• 46:09the same DNARA ratio for
• 46:11risk and protective variants.
• 46:13But a risk variant that changes
• 46:14expression will all of a different
• 46:16outcome for risk and protective.
• 46:17And we're going to stress the cells in vitro.
• 46:20And so here's an example of
• 46:22risk variants that change their
• 46:23effect in the context of H court,
• 46:25IL 6 and interferon alpha.
• 46:27So A21 stressor H court from earlier and
• 46:29two different inflammatory stressors.
• 46:31I think all that I really want you to
• 46:33take home is that we have a Venn diagram.
• 46:34We don't have a circle.
• 46:36So it's actually a unique effect
• 46:38of these three different stressors,
• 46:40up regulating but also down
• 46:42regulating risk variants.
• 46:43And so our our take home here
• 46:47across disorders is that different
• 46:49inflammatory stressors have different
• 46:51effects on cross disorder risk.
• 46:54So you know,
• 46:55as as an example here with Interferon Alpha,
• 47:00we see strong up regulation of
• 47:04ASD and schizophrenia,
• 47:06whereas over here with H core,
• 47:11you know,
• 47:11we get a bit more of the bipolar risk,
• 47:13right.
• 47:14So we get different disorders with different
• 47:16stressors and different risk variants,
• 47:18kind of suggesting that there's a
• 47:20specificity to the environmental insult
• 47:22that might make phenotypes worse, right?
• 47:24Some diseases are more
• 47:26worse in some environments.
• 47:27And so Kieran asked me to really
• 47:29focus on clinical take homes here.
• 47:31I don't get to take credit for this slide,
• 47:33an MDPH in the lab, Karina made it,
• 47:35but I think it's fantastic.
• 47:37And so the first one is of course
• 47:39that individual susceptibility is
• 47:41not a personal weakness, right?
• 47:43There's a genetic component to
• 47:45our susceptibility to trauma.
• 47:46Second,
• 47:47it I think seems very likely that
• 47:51heterogeneous outcomes will require
• 47:53heterogeneous interventions.
• 47:55Third,
• 47:56I think it's going to be very
• 47:58important not just for PTSD,
• 48:00but maybe for all brain disorders
• 48:02stretching into Alzheimer's and
• 48:03Parkinson's that we assess for
• 48:05potential trauma in every patient
• 48:07because that's going to inform outcomes.
• 48:09And finally,
• 48:11earlier identification and
• 48:13intervention will expand the net
• 48:16type of therapeutic avenues
• 48:19available to clinicians.
• 48:22I'm checking on time I I also have
• 48:24we've we're also really interested
• 48:25in gene gene interactions now.
• 48:27And so I'll try to just give you the
• 48:29take homes of this part of the talk.
• 48:31We're interested in how risk variants some.
• 48:33So here's an example of two
• 48:35risk genes showing different
• 48:36effects on transchomic effects.
• 48:38But now we're going to look at
• 48:39their effects in combination.
• 48:41So this was led by, again,
• 48:42Nadine a while back now.
• 48:44So we've got 4 schizophrenia
• 48:46common risk genes.
• 48:47We're perturbing them doing RNA seek
• 48:49of these risk gene effects alone,
• 48:51but also in combination she
• 48:53created from those individual
• 48:55perturbations an additive model.
• 48:56If each of these things happened
• 48:58with these four perturbations,
• 48:59here's what I think would happen if we
• 49:00did all of them at once and then in
• 49:02parallel we did all four at the same
• 49:03time and we asked how good the model was.
• 49:06And so genome wide,
• 49:07the model is pretty good.
• 49:08About 82% of the time,
• 49:10downstream genes changed exactly as you
• 49:12would predict from an additive model.
• 49:14But those times the model failed
• 49:16were particularly interesting.
• 49:18The 7% of the time that genes
• 49:20were more down than expected,
• 49:21we were enriched for
• 49:23neurotransmitter risk genes,
• 49:24and the 11% of time that genes
• 49:25are more up than expected,
• 49:27we are enriched for the rare
• 49:28and common variants linked
• 49:29to schizophrenia and bipolar.
• 49:31This really suggests there's an
• 49:33emergent biology that comes from
• 49:34manipulating genes in combination
• 49:36that you could never predict from
• 49:38one at a time perturbations,
• 49:39and that ultimately we're going to
• 49:41have to test these risk variants in
• 49:44combination to understand their full impacts.
• 49:46The caveat to this particular
• 49:47study was that we picked these
• 49:49four risk genes on the strength of
• 49:50genetic evidence without paying
• 49:51attention to their biological roles,
• 49:53and it turns out that three of
• 49:54the four of them were synaptic.
• 49:55And so we've gone back now and
• 49:57repeated this approach,
• 49:58but on a larger list of genes to
• 50:00ask if what we saw was only because
• 50:02those 3 genes had related function.
• 50:05So going from the PGC 2 to the PGC 3,
• 50:08using the same prioritization approach,
• 50:10instead of having just four genes
• 50:11that she was from, we had 25.
• 50:13And from those 25 prioritized genes,
• 50:15we picked 5 synaptic genes,
• 50:175 regulatory genes,
• 50:18and five multifunction genes
• 50:20that were neither synaptic nor
• 50:22regulatory and they had no apparent
• 50:24relationship to each other.
• 50:25And so for all three of these examples,
• 50:27we perturb,
• 50:27We perturb risk genes alone and
• 50:29in combination and asked how
• 50:31good the additive model was.
• 50:32So in the synaptic model
• 50:33with the five synaptic genes,
• 50:35we saw the exact same thing
• 50:36we'd seen previously,
• 50:37but 15% of the genome wide effects
• 50:40did not follow the additive model.
• 50:43Likewise in the regulatory
• 50:44genes that we again saw,
• 50:45about 17 or 18% of genes did
• 50:48not follow the additive model.
• 50:50But unexpectedly in the
• 50:52multifunction genes that
• 50:54all the effects came
• 50:55through the additive model.
• 50:56So what's going on here?
• 50:58How does gene function predict
• 51:00the combinatorial effects?
• 51:01I don't have the full answer to you here,
• 51:04but I can say amongst the 17 I guess
• 51:06percent of non additive changing genes,
• 51:09we had a huge investment for genes with that
• 51:11were convergent that had shared effects.
• 51:13And I'll I'll explain more carefully
• 51:14what I mean on the next slide.
• 51:16Likewise in the regulatory perturbation,
• 51:20those genes with non additive effects were
• 51:22again in which for the convergent targets,
• 51:24but in the multifunction group we had very,
• 51:27very few convergent targets and
• 51:29that might be explaining why we
• 51:31didn't see any non additive effects.
• 51:33So what do I mean by convergent?
• 51:36Well, first of all,
• 51:37I guess I should know that these
• 51:38convergent risk genes are enriched
• 51:40for schizophrenia risk variants
• 51:41and they're different across the
• 51:43synaptic and regulatory effects.
• 51:45So they're different convergent
• 51:47downstream networks in these
• 51:49different biological functions.
• 51:51And in here I'm going to explain
• 51:52what I mean by convergence.
• 51:53So let's say we have 10 risk genes and each
• 51:55of them have a different individual effect.
• 51:57The expected additive
• 51:59model is the full rainbow,
• 52:00the full add additive value of
• 52:03each of these individual effects.
• 52:06But if I manipulate them in
• 52:08combination experimentally,
• 52:08what we see is something just
• 52:10less than the prediction.
• 52:11Things don't change as expected.
• 52:13Now convergent effects are those shared
• 52:15things that are in all of these rainbows.
• 52:18So what are the genes that show up in each of
• 52:21these individual single gene perturbations?
• 52:24And here I'm coloring them black.
• 52:26So these are downstream genes of all
• 52:28of the of the 10 individual genes.
• 52:32And those downstream genes we think
• 52:34are the ones explaining the the the
• 52:37non additive interactions between
• 52:38risk variants and schizophrenia.
• 52:41With the last minute,
• 52:42I want to talk about drugs and therapies,
• 52:45'cause I think that's where
• 52:47everybody wants to see this go.
• 52:49It's been very hard in neuroscience
• 52:51to discover new drugs, right?
• 52:53We haven't had new indications
• 52:56in schizophrenia for 50 years,
• 52:58and I like to point out this
• 53:00figure from a review of Nature Drug
• 53:02Discovery written for biotech people.
• 53:04If you're investing in biotech
• 53:06and you want to make money,
• 53:09this chart tells you where to put your money,
• 53:11right?
• 53:12Companies that have more
• 53:13publications per R&D dollar,
• 53:15I'm more likely to make money.
• 53:16The in fact only negative predictor
• 53:18of whether a company will make
• 53:20money is if they're looking
• 53:22at a neuroscience indication.
• 53:23And I think this really points to how
• 53:25hard the challenge has been and why
• 53:28it's so important that we keep trying
• 53:30in academics to move the needle here.
• 53:33So how can we use stem cells
• 53:36as a drug screening platform?
• 53:39Is clinical drug responsiveness predictable?
• 53:43So can we show that a drug that we know
• 53:46patients respond to in the clinic,
• 53:49can we show that their neurons
• 53:50respond to that drug too?
• 53:52So if we were to do this experiment properly,
• 53:54you would want to have a drug where
• 53:57patients show a lifetime stable response.
• 54:00Now as psychiatrists, I think you know
• 54:02how small a list of drugs we have where
• 54:05patients show show clear lifetime,
• 54:07stable responsiveness.
• 54:07Ideally, you would also use a drug where
• 54:11patient responsiveness is predicted
• 54:13by the responsiveness of their family
• 54:15members who might have the same disorder.
• 54:18So we're looking for a heritable
• 54:20lifetime stable drug response.
• 54:22I don't know how many are
• 54:23coming to mind for you,
• 54:23but the best example I've seen
• 54:25of this gold standard example is
• 54:28lithium responsive bipolar disorder.
• 54:30So not true for every patient,
• 54:32but there is, you know,
• 54:33a very heritable and lifetime
• 54:35stable proportion of patients here.
• 54:38So this was back when I was still
• 54:40a postdoc with Rusty but work that
• 54:42was you know really nicely finished
• 54:44by Jerel Mertens and Jun Yao.
• 54:46We were able to recruit 6 patients
• 54:48with bipolar disorder,
• 54:50three of them lithium responsive,
• 54:51three of them non lithium responsive
• 54:53and three controls.
• 54:54The neurons from the bipolar cases
• 54:57were universally hyper excitable,
• 54:58but only those neurons from lithium
• 55:01responsive cases showed reduced
• 55:02activity when treated with lithium.
• 55:04The non responsive cases showed no
• 55:07reduced activity in lithium treatment
• 55:09and Rusty then subsequently went and
• 55:11got a second cohort lithium responsive
• 55:12cases and was there able to show that
• 55:15by doing activity recording from just
• 55:16five neurons with or without lithium,
• 55:19he could predict which patients had
• 55:21lithium responsive or non responsive
• 55:24bipolar with an accuracy over 90%.
• 55:26So I think this is the gold standard example,
• 55:29but it's the one that we hold out for
• 55:31that we might be able to to follow up.
• 55:33And so we've gone back to this convergence
• 55:35idea and asked if we can use this to
• 55:38predict drug responsiveness or sorry,
• 55:39if we can use convergence to predict
• 55:41novel drugs that we could then
• 55:43test responsiveness to in cells.
• 55:45And so Kayla's looking at drugs that
• 55:47reverse our convergent signatures.
• 55:49I think these are convergent targets
• 55:51of schizophrenia risk genes.
• 55:53But one of the top drugs that she
• 55:55predicted is actually glucocorticoid
• 55:57receptor stress hormones, right?
• 55:59So everything kind of keeps
• 56:02coming back full circle.
• 56:03This was not the first drug that we
• 56:05tested on her convergent signatures.
• 56:06Here we're looking at a toxamir.
• 56:08This data is not being
• 56:10presented the most clear way,
• 56:12but the green is the genome wide
• 56:15effect of our risk variants on gene
• 56:17expression and the yellow dots I
• 56:20think you can see are flattened.
• 56:22So the the genome wide effect of risk is
• 56:25being minimized by treatment with a TOXAMIR.
• 56:27And so this is kind of the way
• 56:29that we're trying to screen for
• 56:31impact on convergent effects.
• 56:33I don't have any more data
• 56:34on drug NIST at this point,
• 56:36but it's kind of the model
• 56:37that we're looking at.
• 56:37Can we identify new not new drugs
• 56:39that might perturb not those risk
• 56:41genes but their shared downstream
• 56:43effects and ultimately will this
• 56:45inform patient outcomes.
• 56:46And so the idea that I want to leave
• 56:48you with is that genetics alone does
• 56:50not determine clinical outcomes.
• 56:51But as I think all of you know,
• 56:53psychiatric disorders are much
• 56:55more complex in etiology than
• 56:57I think anybody anticipated.
• 57:00With advancing genetic studies,
• 57:02we have now identified hundreds of risk
• 57:05variants across psychiatric disorders.
• 57:08And I think one of the most obvious outcomes
• 57:10will be that we'll have genetic risk scores.
• 57:13Now they're not accurate yet,
• 57:15but I think this is coming.
• 57:18That being said,
• 57:19these genetic risk scores will never be
• 57:21perfect if we don't take into account the
• 57:23environment that our patients live in.
• 57:25So it's important to think about things
• 57:28like social supports and housing security,
• 57:31but it's also important to think
• 57:33about drug exposures,
• 57:34nutrition and pollutants in the environment.
• 57:39All of these things together are
• 57:41going to inform the outcomes in
• 57:43our patients and it's going to be
• 57:46important to advocate for equity,
• 57:47right in some of the context
• 57:49in which people live.
• 57:51And the final,
• 57:51more optimistic thought that I
• 57:53want to leave you with is that
• 57:55if we can begin to really dissect
• 57:57the mechanisms of these gene,
• 57:59gene and gene environment interactions
• 58:01that I'm hoping it will inform what
• 58:04I'm calling genetic resilience.
• 58:05If we understand why when you put
• 58:07some genes with some environment,
• 58:09the effects are less than we would expect,
• 58:11are those pathways that we can
• 58:13target to decrease the impact
• 58:15of genetic risk in our cases?
• 58:17And this is really what I'm hoping that
• 58:19the lab will ultimately be able to do.
• 58:20How do we allow everyone to
• 58:22achieve their genetic maximum,
• 58:24the best possible outcome given the
• 58:26genetic risk that they were born with?
• 58:28Can we inform clinical treatments by gene,
• 58:31gene and gene environment interactions?
• 58:33And so with that,
• 58:34this is the most important slide in the deck.
• 58:36These are all the people who contributed
• 58:38to the work that I've shown you.
• 58:40A lot of the gene environment stuff
• 58:42was by Kayla Townsley and Karina.
• 58:43So Michael led the autism studies.
• 58:47It's been, you know,
• 58:48a remarkable effort over the last 10
• 58:50years to really pull this story together.
• 58:52And with that,
• 58:53I know I I'm kind of at time,
• 58:55but I'm happy to take questions too.
• 59:02Well, let's stop
• 59:04playing. Yeah, I believe the
• 59:06audience might should be unmuted.
• 59:08Anyone has question for example
• 59:10and I know we are at times.
• 59:12So if anyone wants to leave
• 59:13and anyone wants to stay,
• 59:14we continue the conversation.
• 59:22Thanks for the next
• 59:23talk. I'd like to the time frame
• 59:26was the stress exposure to the
• 59:28courses or how long was it?
• 59:30How do you reconcile your findings
• 59:34with findings that hydrocortisone
• 59:37might be preventive for PTSD?
• 59:39Yeah, the the hydrocortisone
• 59:41field is really confusing.
• 59:42I'm going to be honest.
• 59:44One of the things that we don't
• 59:46even really know is whether
• 59:47what's important is exposure to
• 59:49H court or recovery from H court.
• 59:51And given that we only had one time point,
• 59:53I can't even resolve that
• 59:55simple question, right?
• 59:56Is it that the patient just wanted faster or
• 59:59came down slower on our 24 hour treatment?
• 01:00:01There's a ton of work to be done there.
• 01:00:04We picked a 24 hour treatment
• 01:00:06honestly because the field had
• 01:00:08been using that in blood.
• 01:00:09I think stress can be both acute and chronic.
• 01:00:13This is not all of it and even more
• 01:00:15so remember that I told you that
• 01:00:16neurons that we're making are young,
• 01:00:18they're fetal like.
• 01:00:19So I'm not even sure we're modeling
• 01:00:21the type of environment that
• 01:00:22would cause PTSDI think we're more
• 01:00:24accurately modeling the type of fetal
• 01:00:26environment that might cause outcomes,
• 01:00:28right.
• 01:00:28But then again that's why it's
• 01:00:30so amazing that we were able to
• 01:00:32reproduce our stress dynamic EQTLS
• 01:00:34in a postmortem base that where
• 01:00:36people had adult exposures, right.
• 01:00:38Like I was shocked at the
• 01:00:40percentage and that half of our our,
• 01:00:42our variants replicated here.
• 01:00:44We've got adult human brains with
• 01:00:47adult trauma exposures decades
• 01:00:49ago and fetal like human cells
• 01:00:52with acute exposures a day ago and
• 01:00:54and half of it's replicating.
• 01:00:56And so it might mean that neurons
• 01:01:00respond to stress pretty similarly
• 01:01:02regardless of their maturation.
• 01:01:04That's possible,
• 01:01:04but there's a lot of specificity
• 01:01:07here that has to get untangled.
• 01:01:11Thanks for the great talk, Kristen.
• 01:01:12I was wondering with any of the
• 01:01:15patient derived stem cell samples
• 01:01:17schizophrenia that you've worked with,
• 01:01:20has there ever been a correlation between
• 01:01:23what you've observed in these cells,
• 01:01:25any sort of phenotype with symptoms,
• 01:01:27severity, onset of illness or
• 01:01:29any other behaviorally observable
• 01:01:31characteristic of the patient?
• 01:01:33It's a great question.
• 01:01:36I think this kind of thing depends
• 01:01:38on the cohort that you get right.
• 01:01:40So the patients that a lot of my work
• 01:01:43initially represented were with Judy
• 01:01:45Rapport on childhood onset schizophrenia,
• 01:01:48all of whom are particularly severe.
• 01:01:50I thought at the time this was 15 years ago,
• 01:01:53this would be a great cohort because
• 01:01:55they were the youngest on set
• 01:01:56and I knew my cells were young.
• 01:01:57But in hindsight it was a terrible choice
• 01:02:00because they were all very severely impacted,
• 01:02:02but through very different rare variants.
• 01:02:05I don't think any of my cohorts we,
• 01:02:07I guess what I can say in the PTSD,
• 01:02:09we checked for correlations,
• 01:02:11didn't really see it.
• 01:02:12Honestly, Florabacarina has one
• 01:02:14of the best demonstrations of
• 01:02:15this in the context of autism.
• 01:02:17I think it was looking at language delays,
• 01:02:21but you really have to have the right
• 01:02:23cohort to be able to quantify symptoms,
• 01:02:25severity and have a good enough
• 01:02:27power to do it.
• 01:02:28So I haven't been able to, a few others have.
• 01:02:35Well, that was through the force.
• 01:02:36Thank you so much for an
• 01:02:37excellent talk and thank you.