Transcript for:
8.) Lecture on Neuroscience, Hormones, and Behavior

I think it's so considerate that they ask you whether you want to leave the moment it starts recording I think no I do not want to be part of any recordings it's a legal thing you can't record people without their consent or you can in Texas oh yeah so it depends on which state you're in but zoom can't guarantee that you're in a state where that's where two party consent is not the normal Yeah. I am still waiting for Laura. Thank you. Am I missing her? Is anyone else seeing Laura Despin? No, right? I think she's the last speaker. Maybe she thought she only has to show. That can be part of it. And I also think we can start with other speakers otherwise. Kara, you said that we usually wait a little bit longer, right? Yeah, 12.03, 12.04. Okay, I think we can just get started so that we can have as much time listening to our amazing speakers today. I feel very, very honored to moderate this session. I think our speakers today are all absolutely talented scientists with a very cool and unique stories. Thank you very much for Just so you know how it's going to go today, we will have five speakers. Everyone will have 12 to 15 minutes to speak. Afterwards, we will have some time for questions. You can raise your hand if you have a question or just type your question in the chat. And if you type your question in the chat, I will verbalize it for you. And otherwise, the speaker can answer within the chat as well. As you can see, we are being recorded right now. That is because our talks will be posted on YouTube later. So if you do miss something due to an unfortunate event, you can watch all the talks later. With that, I would like to introduce our very first speaker, which is Dr. Simone Sun. Dr. Sun is a scientist, an activist and an artist, and she's sharing her screen. making me not see my notes anymore. I'm sorry. She's a scientist, an activist, and an artist. She's a postdoctoral research fellow at Cold Spring Harbor Laboratory and a senior fellow at the Center for Applied Transgender Studies. She's currently funded by the Simons Foundation and NIMH. Dr. Sun is investigating how gonadal hormones affect neurophysiology through regulation of gene expression. Her overarching research explores how hormones influence sex variable biology. Beyond the bench, Dr. Sun's artwork bridges the rational and emotional aspects of scientific exploration. Without further ado, here's Dr. Sun. Hello, everyone. Are you all able to hear me? Yes. And are you able to see my... Oh, I can't see anybody now. But were the slides visible? I need an auditory conversation. Okay. All right. Great. So, hello, everyone. My name is... Dr. Simone Sun, and I'm here today to talk about some of the work that I have started as a first-year postdoc in Jessica Tolkien's lab at Cold Spring Harbor. And the title of my talk today is Neurophysiological Modulation by Reorganizational Estrogenic Gene Regulation. And so the question that really interests me is how gonadal hormones are able to affect neurophysiological function. to mediate many of the sex variable phenomenon that we see throughout the animal kingdom. And so, granadal hormones are well known. shape sex variability and regulate changes in neural activity and that this is thought to occur through the action of hormones on their canonical hormone receptors and their interactions with various sites on the genome and recently with the advent of many high throughput sequencing technologies the gene sets for that are regulated by these canonical hormone receptors have been recently identified. However, the question remains is how this results in functional changes that are seen with exposure to hormones. And so that's primarily the question that I'm interested in with an eye toward how this type of work and understanding can better inform mental health outcomes of individuals on hormone therapy. And so just to briefly give an overlook of how these gene sites have been identified, recently a graduate student in Jessica's lab, Bruno Gegenhuber, identified through cut-and-run and ATAC-seq, which are sequencing technologies that allow us, or sequencing methods that allow us to identify the sites on the genome that estrogen receptor binds to. as well as the changes in the chromatin accessibility and the open sites of the genome in the presence of estrogen. And so what he found was that in the presence of estrogen, estrogen receptor alpha binds to several thousand different sites on the genome. And that as well as these binding changes, we also see that chromatin accessibility changes, where we see several thousand sites more open and available when estrogen. is present. So we have all of these thousands of genes. What do they do? And how are they important to how neurons function? And so this is sort of where my work starts to come in. And so if we look at these genomic sites, what we see is that a lot of them with regards to the neurons that are sensitive to the presence of these hormones is that in the brain, many of the estrogen receptor binding sites are specific to brain function, as well as we see a small proportion of the genes that are targeted play some role in sort of the other bodily functions that estrogen is involved in. If we look more closely at the specific neurological functions of this large sets of genes, we find that many of them are specifically involved in things such as the organization of the synapse, regulation of the membrane potential, things like postsynaptic organization, and the action potential or the spikes of neurons. And if we look more detailedly at some of these genes that fall within some of these categories, we find things such as voltage-gated ion channels, like calcium channels, which play important roles in the response to input and the processing and regulation of the neurons response to these different kinds of synaptic inputs. We see proteins or genes like can kinase two, which are involved in things like synaptic plasticity, as well as various components of glutamate transmission, various voltage gated potassium channels, and sodium channels. And so by taking a closer look... At the specific functions of the genes, I've begun to divide and categorize these large gene sets into more specific categories that are involved in the postsynaptic responses to inputs to the changes in the intrinsic properties, the excitable properties of these neurons, which can affect things like activity patterns and sensitivity to firing. as well as the conduction of their inputs and signals to their downstream targets through modification of presynaptic transmission. And so just to provide some examples of genes that are Part of these kinds of functions, we have Kemp kinase 2B, which is involved in changing the plasticity threshold for incoming input and changes to those inputs. We see voltage-gated potassium channels like KCNN2, which could affect, based on their expression level, the excitability and the firing frequency of hormone-sensitive neurons, as well as genes like CATNAP4, which can affect the transmission probability for inhibitory synapses. Now, to zoom in on one of these examples, we'll take Kankinase 2B. partially because this is one that I am intimately familiar with in my doctoral work. And so estrogen receptor alpha binds 2.75 times more when estrogen is present, and ChemK2B encodes for a particular isoform of ChemKinase 2. also known as the beta isoform, hence the B. And so what's interesting is that ChemKinase 2, the two predominant forms are the alpha isoform and the beta isoform that are expressed in a majority of neurons. And what's interesting about the beta isoform is that it exhibits a much higher sensitivity to calcium, which is one of the main signals that it uses in order to be activated. And so in my doctoral work, I created a model of postsynaptic components to try and better understand how various aspects of synaptic function like plasticity and responses to changes in activity and attempting to model that to better understand how these different components contribute to these kinds of phenomena. And so one of the things we found. is that if you look at the postsynaptic response by looking at the steady-state calcium level as an indication of how active or how responsive that synapse is at various levels of activity, if you're represented by the variable R, which stands for the rate of input that the synapse is receiving, that without any beta-CAM kinase 2 present in the postsynaptic compartment, the synapse requires a very high level of activity. in order to be put into this more active high calcium state. However, that as we increase the amount of chemokinase 2 that is available in the postsynaptic compartment, this shifts the response curve further to the left, indicating that the threshold of activity that is required to push the synapse into this active state is much lower. And that this threshold gets lower and lower with more and more beta-CAMPYLase II expression. However, this is just a single component of, you know, the many pieces that are involved in the synaptic, post-synaptic compartment. Not to mention that, you know, if we just look back at our gene lists, you know, this is one of thousands of genes that are supposedly changing. And so one of the things that this MyModel is able to allow us to do also is to look at how... These components work holistically together to shape the kinds of responses that the synapse can have. So, for example, another gene that is regulated by estrogen, GRIO1, encodes for a subunit of glutamate receptors that make up the calcium permeable ampere receptors that allow calcium to influx to the postsynaptic compartment in response to glutamate release from the presynaptic neuron. and that can kinase 2, phosphorylates this subunit in order to make it active and available for the synapse. And so this is, again, one of the genes that is regulated by estrogen. Another one that is regulated by estrogen is CACNA1C, which encodes for a voltage-gated calcium channel that is a part of the postsynaptic response that lets in calcium flux when we see presynaptic glutamate transmission. And then... allowing for calcium to flux inside the cell. And so my model also incorporated various other calcium-sensitive aspects and allows us to dynamically explore how alteration of these specific components could holistically affect the system in response to different activity levels. So just to provide a sort of example of this, what you're seeing here on the left is... the response to the model to a broad change in activity. And in this case, what we're doing is reducing activity from a high level to a lower level. And so what we see is that there is some level of plasticity and compensation in the postsynaptic, in the postsynapse. And what we see is the steady state of calcium returning to a sort of almost homeostatically returning to its before this activity level changed in that each of these individual components is activated at slightly different times to elicit this response. And if we change one of these components, what we can see. is that the whole system is affected in a particular way. So for example, if we do not allow for any beta chem kinase 2 to be active, the synapses respond much more slowly to this change in input. And not only that, it takes much longer and doesn't... return to the same baseline as before. However, we were to change something like the CAV1 channel, or CACNO1C, by increasing its amount of calcium flux in response to activity, we shift again this response. This response becomes much faster, but does not affect the degree to which that change occurs. And so together this sort of paints a much more holistic view of how we can look at how hormones can change the inputs to hormone sensitive neurons that encode for things like social and sensory cues, how this information is processed by the neuron through changes in its excitability and how this could lead to alterations in the signal transduction. to the downstream targets of these hormone-sensitive neurons. Now, more broadly, how does this type of work and knowledge, how does this shape our understanding of how hormones broadly affect the brain? And so the predominant endocrinological model of how gonadal hormones affect neurons is based on the organizational activation model originally proposed by Phoenix, Skoy, Gerald, and Young back in 1959. And what their model posited was that hormones... affected neurons in the brain differently based on the lifetime of the animal. So for example, the two time points that they were interested in were development and that leading to adulthood. And what they found and posited was that hormones being present at these specific time points in life resulted in different responses. And so in development, hormones served to organize the neural circuitry and the neural cell types in order to later on be activated by those same hormones in adulthood. And an important aspect of what this model suggests is that this developmental organization is permanent and remains permanent such that the activation during adulthood elicits a particular sex variable response. Now, one of the things that Bruno found in his data, which I find particularly interesting, and parallels some of the What I'm interested in, in looking at the physiological consequences of this gene mutation, is that, for one thing, that the epigenomic response in the binding of estrogen receptor alpha to the genome, as well as its alteration of the chromatin, was not sex-specific, and that both sexes were able to respond to estrogen in the same manner. And furthermore, when he looked at how sex variability in the genomic state was elicited in relation to gonadal hormones, what he found was that the wide array of sex variability that is seen when comparing males and females in their epigenetic chromatin organization was lost with gonadectomy. And so what this... specifically suggests is that in adult, and these experiments were done in adults, and what this indicates is that as opposed to a form of genetic activation by these hormones, it rather suggests that the genome is being reorganized depending on the acute hormonal milieu. And so I wanted to incorporate this kind of, this dynamic flexibility to and to inform them of any sort of modeling and experimental work that I will be engaging in, I would like to borrow from a metaphor originally proposed by C.H. Waddington. in which he was attempting to explain the developmental canalization of pluripotent cell types into different specified cell types through development. And how this was illustrated was by a ball rolling down a set of valleys and hills, where as development progressed, a ball was more likely to end up in a particular valley that indicated a particular cell type and that the sort of barrier between these different cell types was established slowly as developments progressed. And specifically, what he sort of presciently proposed was that these hills and valleys were shaped by the underlying genes and what he called chemical tendencies of these genes pulling on the above landscape to shape this development trajectory. Now... To use this metaphor in a more dynamic and broad sense and incorporating what we are finding, how neurons can be affecting both the genetic state of neurons as well as their neurophysiological phenotype, I'd like to frame my work in this adaptive model where we still are using a ball rolling down a set of valleys and hills over a development time point in aging and that the and hills across this landscape represent various gender, sex, variable, phenotypic states that these neurons can exhibit, and that as development progresses, these valleys and hills are changing, and that these changes are mediated by the activation and reorganization, or rather, the organization of the genome pulling on this phenotypic landscape by hormone-responsive elements. And so, for example, here, during the perinatal period where we would be expecting a hormone surge at a particular time, this would shape the probability and trajectory of a particular and bias the phenotype to one over some other phenotypes. And that as the as development progressed. And hormones will do fluctuate, such as in pre-adolescence, where we see less prenatal hormones present endogenously, a change in that phenotypic landscape to a less estrogen or a less hormonally biased phenotype. And that this would continue as the life of the animal were to progress, where we see sort of pubertal surges, again, reshaping this phenotypic landscape well into adulthood, where we see these later, where we see changes later on. And I think that this sort of this depiction of the natural plasticity and variation that we're seeing at these different levels of hormone, of of what hormones do better illustrates the seemingly reorganizational effects of genital hormones on sex variable neurophysiology and can help us better inform the changes that we see in mental health of individuals. who are on hormone therapy. And before I finish for questions, I would like to acknowledge the Tolkien that, especially my postdoctoral mentor, Jessica Tolkien, who has helped me formulate and specify some of these ideas and is guiding me in the neurophysiological explorations that I will be doing, as well as Melody Wu and Bruno Degengruber. who were tour de forces on a lot of the genetic data that I showed today. I would also like to thank my doctoral advisor, Richard Chen, and my collaborator, Dan Levenstein, who helped me create that model, as well as the Center for Applied Transgender Studies and Coran Humphrey Memorial Journal Club for intellectual stimulation and discussion. I would like to acknowledge the native lands that New York University and Cold Spring Harbor, on which Cold Spring Harbor occupy, as well as the generational consequences of a lot of the eugenic work that occurred on Cold Spring Harbor's campus and affects the social and the scientific progress. that my work is a part of. And lastly, I would like to thank my funding sources from the NIMH and Safari. And so with that, I'll take any questions. If there's time, I don't know. That was absolutely fabulous. Let's answer just the first question from Anna. Anna, do you want to just ask it yourself or do you want me to read it out? Yeah, sure. That was a really interesting talk. I noticed that at least two of the genes that you mentioned, CNTNAP4 and CACNA1C are both risk genes for autism, ADHD, and schizophrenia. And I was curious if you've thought about how hormones might be affecting the symptoms and development of those disorders, given that... There's a difference in the risk between males and females for those disorders as well. Right. So what the data is suggesting to us is that what these sort of sex variable biases that we see in these various neuropsychiatric conditions could indeed be because of, or rather interacting with the changes in gene expression that are elicited by the different hormonal states of the individual. And so that is... that's a line of inquiry that Jessica's lab is currently interested in. So, you know, I guess stay tuned. Thank you so much, Simone. There's one more question in the chat. Maybe you can answer that in the chat. Thank you so much. And then our next speaker is Stacy Kiger. Dr. Kiger is a postdoctoral research associate in the Department of Medicine and Psychiatry at the University of Cambridge. She was previously a postdoctoral IRTA fellow at the NIH and IMH in the section for functional neuroanatomy. Stacey is broadly interested in how hormones, stress, and the immune system interact with the brain to influence mental health. I think that is a perfect segue after that last question. So Stacey, take it away. Thank you. One second. Can you see these slides okay? Yes. Sorry. Getting used to Google Slides. Okay. Are you able to see just the presentation? Yeah, okay. Okay, great. I think we would have figured this out by now, but here we are. Okay, so thank you, Micah, for the introduction, and thanks, Simone, for the fascinating talk. Thanks to Garrett McKinley and the Leading Edge organizers for this opportunity. As you can see from my title today, I'll be talking about stress and speaking of it is currently a balmy 37 degrees Celsius, so 99 degrees Fahrenheit here in the UK. So fingers crossed that my outdoor plants don't combust or something as I'm giving this presentation. And I hope that if you're dialing in from the UK that you're not melting. But joking aside. It's a pleasure to be in this group of phenomenal scientists, and today I'll be sharing some of my postdoc work investigating the relationship between stress, the peripheral immune system, and mood disorders. So I want to begin by telling you a little bit about depression. The most recently published data from the Substance Abuse and Mental Health Services Administration indicates that about 1 in 10 Americans over the age of 13 have experienced an episode of depression in 2020 alone. Depression is the leading cause of disability worldwide, according to The Who. It is characterized by prolonged episodes of persistent anhedonia, or a state of lacking interest or pleasure in previously rewarding activities. It's also often comorbid with other mental health issues like anxiety. Importantly, there are a variety of antidepressant drugs on the market, but up to a third of MDD patients show no response to the currently available treatments. And we don't yet understand what accounts for this difference, but in terms of risk factors, exposure to potentially traumatic events or PTEs, specifically early life stress, is a major concern. It has been shown that there is a dose-response relationship between exposure to stressful life events, and that ranges from failing a grade to a major car accident to physical abuse from a caretaker, and adverse mental health outcomes. To put the issue in perspective, here are data from a survey of 10,000 American teens aged 13 to 17. This data was collected between 2000 and 2004. And a strikingly large percentage, 62% of them, reported exposure to at least one PTE in their lifetime. A separate study cohort suggested that approximately a third of these kids would go on to develop a mental health issue, including MDD, with early life stress doubling the risk of developing mental health issues compared to unexposed peers. So the question that I'm interested in is how is stress encoded in the body? One mechanism involves the HPA axis. As many of you know, a stressful encounter will activate the HPA axis and stimulate the adrenal cortex to produce cortisol. Cort, which is depicted here in this gold color, crosses the plasma membrane of cells where it binds to glucocorticoid receptors. Once bound to cort, GR can enter the nucleus, whereby it promotes transcription. It's been shown in the brains of stressed mice that one of the transcriptional targets following exposure to stress is NF-kappa-B inhibitor alpha, which inhibits inflammation. Which I'll come back to in a moment. A chronically elevated court eventually leads to the downregulation of GR expression via methylation in the GR promoter. As a tragic demonstration of this, postmortem brain tissue from people who died by suicide revealed that lower GR protein levels and higher GR promoter methylation were present in brains of people who had suffered from childhood abuse. This downregulation of GR following chronic stress or chronically elevated court release dampens the brain's ability to turn off the HPA axis, which can cause widespread physiological issues over time. Similarly, white blood cells of the peripheral immune system, including neutrophils and B cells, which I've highlighted here, as I'll be talking about them later, respond to court and can become glucocorticoid resistant, which motivates the investigation of the relationship between stress, immunity, and mood disorders. There is accumulating evidence strongly suggesting a role for the immune system in both the etiology and progression of mood disorders like MDD. For example, patients suffering from autoimmune conditions like psoriasis and rheumatoid arthritis frequently present with comorbid depression. Anti-inflammatory drugs such as TNF-alpha and IL-6 inhibitors improve depression symptoms in these patients, even when there is no improvement in their primary disease. Finally, elevated peripheral inflammation is present in MDD, shown through several meta-analyses, and even more so in patients who are treatment-resistant. This opens two important avenues for further study. One, we need to identify biomarkers that stratify patient populations, which will hopefully result in more effective treatment. And two, we seek to understand the cause-and-effect relationship between stress, depression, and mood. I'll start by talking about my preliminary efforts to identify biomarkers of depression. During or rather over the last year and a half, I've been at Cambridge collaborating on a variety of projects that investigate relationships between immunity and mood. But my primary focus has been on these two studies. One is a phase two clinical trial and the other is a more exploratory clinical study. So I'm going to be walking you through the study design. So ATP, or the antidepressant trial with P2X7 receptor antagonists, which is a mouthful, sadly was a multi-site effort for which I served as the lab coordinator. Patients were recruited to the study from across the UK, and assuming they met the criteria on the left, which I won't get into the details, but if you're interested, feel free to ask me any questions. So assuming they met those criteria, they were entered into the study, enrolled, and then they would come to one of these five sites. where we would do a baseline visit during which time we would take blood, put them in an MRI scanner and do a battery of psychometric testing. So we would ask them questions about things like their depression or suicidality, et cetera. And importantly, for my interests going forward, we'd give them a childhood trauma questionnaire, ask about early life experiences. We used the blood that we took from these participants to generate a large and rich data set of biomarkers, including a glucocorticoid sensitivity assay, which sort of hookens back to what I was telling you about earlier. The participants would then be given the experimental drug and come back for a follow-up visit six weeks later, at which point we would repeat this whole process. Ultimately, there were 15 people that were randomized into the trial, but unfortunately, as of last month, the trial was closed and the study hasn't been unblinded yet, so I don't have any data to show you from ATP. However, the companion study that we had been running at Cambridge meant to sort of flesh out what we were getting from ATP with both healthy controls and non-inflammatory depression has sort of risen in importance as a result of the ATP study closing. So it's basically the exact same protocol. The primary difference between the studies is that in BICBID, it's only at Cambridge, and we're only looking at baseline, so obviously there's no experimental drug here. We are not doing an MRI scan, and we've added in single-cell sequencing to get more deep information about the peripheral blood cells. And additionally, we are doing deep phenotyping of neutrophil activation, which I'll talk a bit more about in the next slide. So here you can see that as expected there is a difference in the depression scores. So PHQ-9 is one of the questionnaires we used to assess depression severity. The controls are obviously much lower and then both depression groups are elevated and there's a trend for the inflamed depression group to be slightly higher than the non-inflamed group. This is preliminary data that we have from the study showing single cell sequencing from four participants. And you can see from this that we're able to capture the expected variety of different blood cell types in blood, with the notable exception of neutrophils. When my collaborator, Dr. Mary Ellen Linnell, was developing this assay, she found that neutrophils did not survive the freeze-thaw process that we needed to be able to multiplex these samples together. So to circumvent the information we're losing from the freeze-thaw, we are doing live cell flow cytometry and looking more deeply at neutrophils in terms of their activation and various states of maturation. And this is, these are bivariate plots that are typical of flow cytometry data, if you're not used to looking at this type of data. Um, If you kind of go clockwise from the top right, you're looking at sort of QC to narrow in on our population of interest, which is neutrophils. And then if you look at the bottom left, I'm showing you an example of self stratification where you can see more mature and more immature neutrophil subtypes. The sort of motivation for really focusing on the neutrophils comes from this study that Mary Ellen. published a few years ago now in the predecessor study to ATP and BICBID, where again, they were looking at depressed patients throughout the UK. And here, what she found was that neutrophil levels in the blood were the strongest predictors of the severity of their depression symptoms. Excuse me. And just to give you a little bit more context about why I'm telling you about mature and immature neutrophils, we were interested in these subclasses because mature neutrophils are thought to be more immunosuppressive, whereas immature neutrophils are thought to be more immune-stimulatory. And so again, this is very preliminary data, but what we can see so far is that it looks like the direction of maturation is as expected where the depressed patients have more of this intermediate potentially pro-inflammatory state. So eventually we will be able to look at relationships between peripheral immune cells and the patient psychometric data which we hope will help us identify different biomarkers to stratify patients and may potentially identify new druggable targets. On the topic of drugable targets and looking to the future, my graduate training in molecular and cellular pharmacology at the University of Wisconsin-Madison is really central to how I intend to continue investigating this cause and effect relationship between stress and mood disorders. Cambridge has given me this incredible opportunity to build relationships with psychiatrists and clinicians, and I will continue to continue building these relationships and continue with these collaborations in the foreseeable future. But I will also be taking the concepts that emerge from the clinical data back into animal models where I can test these relationships in a more causal manner. As an example of this sort of, you know, iteration process. Prior to my UK move, I had already been collaborating with Dr. Mary Ellen Linnell for several years. So I was doing a postdoc at the NIH and Mary Ellen came and was a visiting scholar for a few months. And she and I were both doing. sort of deep immune profiling of tissues in this sort of gold standard preclinical animal model for depression, which is called social defeat stress. So without getting in too much detail, too much detail about the model, again, you can ask me questions if you have any at the end. These animals go through a chronic stress paradigm, and at the end of this stress paradigm, they exhibit anhedonia and anxiety-like behaviors. So as we were doing this profiling, what we found was that there was a really striking decline in the number of B-cell lymphocytes. So B-cells are the antibody producing cells. You've probably heard about them now that we're all in the era of COVID. And we were looking at a variety of tissues, including both peripheral blood and a tissue called meninges, which is a protective layer that surrounds the brain and includes the blood brain barrier. And so what we found was that there was this large drop off in B cells in the stressed mice. And similarly, there was a drop off in B cells in blood, but they recovered to some extent after stopping the stressor. So, to probe further what the function of this decline might be, we used CD19 knockout mice, which are genetically B-cell deficient. What we found was that at baseline, these knockout mice exhibited more anxiety-like behavior, which was similar to that of a wild-type mouse that had undergone the chronic stress paradigm. And there was no further decline with stress, indicating there was some sort of bottoming-out effect. When we did immune phenotyping of these knockout mice at baseline, we found that there was a significant dysregulation of the innate immune system. Specifically, we found that there were increased numbers of neutrophils, which we've already talked a bit about. So you can see that here in this volcano plot, that in the knockout mice, as you go to the right, there are more neutrophils and then shown again here, quantified as a boxing miscarriage plot at the bottom. So, as we come to the end of my talk, there's ongoing work with this collaboration between Cambridge and NIH where we're examining the neutrophils from stress mice. I did a single-cell sequencing experiment of the meninges, and what I hope you can see here from this new map is that in a single cluster of neutrophils, there's a large diversity. There's a lot of heterogeneity just in neutrophils alone. So trying to understand what these different clusters are and what they might be doing is of interest for my future work. We also found that both through single cell and through confocal microscopy and in flow cytometry, that the numbers of neutrophils were elevated following social defeat stress, which was consistent with what we had seen in the CD19 knockout B-cell depletion study. So, So we're hoping to combine the rich data that we generated or we are generating from the BigFID study with what we're finding in neutrophils in this preclinical animal model. And with that, I will. I give my acknowledgments, say thank you to the many amazing collaborators I have that have helped me do all this work. I didn't have time to talk about my graduate work today, but I was studying sex differences and risk and resilience. And I'm interested in bringing that in together with these questions about stress. In the middle, I have my collaborators from NIH, including my former advisor, Miles Frickenham. And then finally. Professor Ed Bulmore is my advisor here at Cambridge, and my collaborator, Mary Ellen. Thanks to all my funding sources. Thanks to Leading Edge. Thank you all for listening. And I'll take any questions you have. Thank you so much. That was fascinating. Does anyone have a question? I will ask just a quick question. Is the opposite also true that if you would basically elicit an immune response in either human or mice that there's an elevation of port and is this effect between the mood and the immune system a vicious cycle or that you're trying to disrupt Or is it like a one-way pathway? That is a really interesting question. So as far as I know, there are spikes in court when you have an immune reaction. So you will see elevated levels of court if you have a viral infection, for example. So there is bidirectional communication happening, and that relationship is really complex and interesting. In terms of, you know, I think you might be kind of hinting at like a necessary sufficiency type of question. If you eliminate the quart entirely and just inject cytokines, it's been shown that that's sufficient to drive depressive behavior. And that's actually one of the things that motivated this evolving of this entire field of neuroimmunology. So. It's really interesting. There's a lot more work to be done. Thank you for the great question. That was a great answer. Okay, let's go on to our next speaker. Our next speaker is Dr. Urza May-Guthman. Dr. Guthman is a postdoctoral research fellow at Princeton University where she is funded by an NIH F32 fellowship for her work. She is also a senior fellow at the Center for Applied Transgender Studies. Dr. Guthman is currently She currently studies the role of gonadal hormones in orchestrating the activity of brain-wide hormone-sensitive neural networks and behavior in socially interacting mice. Broadly, she is interested in applying the tools she is developing in the lab to novel translational models for transgender health and to study the basic questions in neuroendocrine neuroendocrinology. I'm glad I'm not studying that because I cannot pronounce that word. But here's Dr. Goodland because she can. Thank you, Micah, for the wonderful introduction. I am just trying to get things set up real quickly to share my screen. And okay, I want to, if I switch this, do you see my notes or do you see? Okay. i have an idea i did this once recently i'm really sorry okay i think this should work now um okay can everyone see my screen fine Awesome. Okay, so thank you, Micah, again for the wonderful introduction. Like she said, my name is Ertha Mae Gutman. I'm a postdoctoral research fellow in the lab of Annegret Faulkner at the Princeton Neuroscience Institute, as well as a senior fellow at the Center for Applied Trans Studies. And I'm going to be telling you about my work today on gonadal hormone regulation of social behavior and neural networks. And I like to think of my work as having two main avenues, one being basic research. where I try to understand the links between hormones, neural activity, and behavior. And the other is applied transgender studies, where I ask questions of how we can improve upon existing therapies to materially benefit the idiosyncratic needs of transgender populations. And one way we can do this is to develop translational models of gender-affirming hormone therapy that are designed specifically to benefit the mental health needs of transgender populations. And this is a major long-term goal of my academic research career. Despite transgender people accessing exogenous genital hormones treatments since hormones were first discovered in the early 20th century, only recently have people begun to attempt to develop. translational models of gender-affirming hormone therapy. In fact, the first- Thanks for your interruption. Yeah. Are you trying to advance your slides? No, I haven't advanced the slides. Okay. Yeah, sorry. I'm about to. In fact, the first model was just published by Teddy Goats five years ago and did not focus on the effects of hormones on the brain or behavior. And so why are these models important? First and foremost, transgender health care or gender-affirming hormone therapy, or GAHT, is under threat across the United States. And lawmakers across the country explicitly appeal to biomedical science when enacting discriminatory laws barring transgender people from accessing health care. And additionally, I need to point out that this map is non-exhaustive. Non-legislative actions against trans healthcare, such as Texas's recent executive orders to block access to gender-affirming healthcare, and Florida's directive to block Medicaid funding for transgender healthcare, are not shown in this map, for example. And so the situation is truly dire. That being said, we have decades of evidence demonstrating that gender-affirming hormone therapy regimes improve mental health outcomes for transgender people. Despite this, and despite the fact that these treatments often last an individual's lifetime, very little is known about how GATT impacts brain functioning. Thus, my long-term goal is to address these issues by developing translational models of GATT with a focus on how these therapies impact neural activity and behavior. As a first step in this project, I'm manipulating the gonadal hormone state of animals to determine how a longitudinal change in adult hormone state impacts both behavior and neural activity. For this work, I'm focusing on what's known as the social behavior network and social behavior. And the social behavior network, or SBN, is a subcortical network of densely interconnected gonadal hormone-sensitive brain regions that are involved in the generation and expression of social behaviors. In general, for my current research, I have two broad questions. First, how does a large longitudinal change An adult hormone state impact the expression of social behavior, and how does such a change influence the activity of the hormone sensitive social behavior network? Additionally, I want to move beyond prior work that is often focused on a limited set of behaviors, and I want to assess mouse social behavior more holistically. And so in order to generate behavioral richness in my dataset, I'm recording CD1 mice dyads across multiple social contexts, partners, and before and after gonadal hormone state manipulation, where I gonadectomize animals, male and female mice. and give them hormone replacements of either longitudinal testosterone, estradiol, or a sesame oil control. And specifically, I record over three non-sequential days before, and then three non-sequential days two weeks after hormone state manipulation. And this allows me to track both behavior and neural activity across time and a change in hormone state. And so I know this is a very ambitious goal to achieve. And in my talk, I'm going to first break it down into behavior and then neural data. And with regards to malsocial behavior, I want to quantify sex differences. across-mouth social behavior holistically, and test the hypothesis that we'll see a change in these sex differences with the change in adult hormone state. Importantly, for understanding these changes in behavior, we have two major goals to try to understand how hormone state influences the behavior. First, I need to be able to quantify individual microbehaviors that make up the totality of the mouse social behavior repertoire. And second, I need to find a computational method to track the totality of these behaviors across time and hormone state. To do this, I'm using SLEEP, which is a machine vision program for multi-animal pose tracking developed by our collaborators at Princeton in the Murthy and Chavis labs. And indeed, demonstrating the applicability of these tools, a recent study from our lab led by the graduate student Lindsay Wilmore pioneered the application of these types of analyses that I'm going to be showing you today to identify specific behaviors associated with resiliency to social stress. So, if we put this all together and what a still frame of a video might look like, I can use sleep to track the position of socially interacting mice, as well as important objects of interest in their environment, such as their nest. From the track posture of these mice, I can then quantify specific time-varying behavior features that evolve as the mice interact with each other. And these include things like proximity to each other, their social heading direction or what kind of angle the mice have between each other, their speed, how close they are to the nest. And then in order to use these time-varying features to identify and quantify behavior, I reduce the dimensionality of this feature set. Using t-SNE into a two-dimensional behavior space. From there, I apply a Gaussian probability density filter to determine the relative amount of time animals spend across the space. And so in this map in particular, you'll see areas of high density represented with warm colors, and these represent areas within the t-SNE space where there are more frames, whereas cooler colors represent areas within the t-SNE space where there are fewer frames. I can then apply watershed clustering to the behavior space to identify specific clusters that represent frames with similar values of behavior features. These clusters in turn represent putative behaviors, and by calculating things like cluster occupancy, I can determine the amount of time animals partake in specific behaviors, as well as the transition rates between these behaviors. Indeed, this method does pretty well at pulling out specific behaviors, including those we may have targeted a priori for hand scoring analyses, such as this dominant-like investigation, where the subject mouse, shown in pink in these videos and subsequent videos, is performing orofacial and anogenital investigation of the partner mouse, shown in blue in these videos and in subsequent videos, and the partner mouse will let the subject mouse engage in this investigation. and will often rear back to provide access to the areas for investigation. Additionally, this method has proven extremely useful in discovering other behaviors that we may not have targeted a priori for analysis, such as this asocial nesting behavior, where the subject mouse is performing its own behavior separately from the partner mouse while remaining on the nest. And so I can use then these social behavior maps to ask if we see sex differences in mouse dyadic behavior. To do this, I initially assign sex to the mice by observation of genitalia and gonads. And if I then separate these behavior maps into male maps shown in blues and purples and female maps shown in yellows and greens, and remembering again, the darker colors, so the darker purples and the darker greens represent areas within the behavior space that have higher density. or the animals spend more time, there are more frames representing those types of behavior features, and we can see that sex differences emerge in the maps. And I can then take a difference of the behavior space maps between males and females and see that this area in the top left of the map is more male-biased, and the areas on the bottom and to the right of the map tend to be more female-biased. Importantly, and as you can see here, these sex variabilities correspond to specific clusters and behaviors and reveal differences both in our expected behaviors, such as this dominant-like investigation behavior shown in the cluster labeled by the orange asterisk, as well as our newly discovered behaviors, such as the asocial nesting behavior shown here with the blue asterisk showing a female black ass. We can then compute how much time animals spend in each behavior cluster per session to quantify sex differences in mouse dyadic behavior, allowing us to peek into the black box, that is the t-SNE, and identify where these differences in behavior occur. In this plot, you can see all 20 of the identified behavior clusters and the relative amount of time animals spend in each of the clusters, with males being shown in purple again and females being shown in green. And we can see that there are a variety of clusters that show differences between the sexes as well as other clusters that don't show any apparent sex differences. And as I mentioned before, we see differences in these behaviors that we may have looked to analyze a priori, like this dominant-like investigation behavior shown by the orange asterisk, as well as differences between behaviors that we may not have looked for initially, and this method has allowed us to discover, such as this asocial nesting behavior that shows this female bias. In addition to comparing across sex and intact animals, I can compare these density maps before and after gonadectomy to quantify how a hormone state change alters behavior expression. And if we look at the raw maps, we can see that after gonadectomy, both for the males and females, there is a migration of the density in the behavior space more towards the bottom of the map. And this is represented by a flattening of the sex differences in the sex difference map. In order to quantify if gonadectomy changes sex differences in behavior, I'm using a metric of distance between probability distributions known as the Jensen-Shannon divergence, or JSD. And the JSD allows me to quantify how different male and female behavior space maps are before and after gonadectomy. And when I look at the same animals after gonadectomy, I find that gonadectomy leads to a significant reduction in the differences between male and female social behavior. Finally, in order to determine the effect of hormone state on neural activity, I'm pairing these behavioral experiments with multi-site fiber photometry that is set up to record from the hormone-sensitive social behavior network using customizable 3D printed implants developed by myself with the help of Jorge Irovedra Garcia and the Faulkner Lab. These implants allow us to record neural activity across the social behavior network, as shown in these example cell-type nonspecific GCaMP recordings. With these data, I'll be able to determine how a change in gonadal hormone state shifts adult neural network activity dynamics to regulate behavior. And in particular, I'm testing the hypothesis that specific gonadal hormone states drive crosstalk between regions in the network, creating specific subnetworks. And indeed, if you peruse these traces, you can see moments of synchrony across the entirety of the SVN, like here. And you can also identify areas where only one or a few regions of the SVN are active at a time. Consistent with the hypothesis that gonadal hormone state promotes specific subnetworks within the social behavior network, my preliminary data in orchiectomized males shows that orchiectomy reduces correlated activity across regions of the social behavior network during dyadic interaction, as you can see in these correlation heat maps with the colors becoming lighter following gonadectomy. If I then quantify this, I see that gonadectomy leads to a reduction in the inter-regional correlations during dyadic social interactions for all recorded regions but one. And additionally, I want to point out that the data from the lateral periaqueductal gray in particular, the PAG is a brain region, the periaqueductal gray is a brain region that we think of as a premotor output for the social behavior network that integrates the signals coming in from the rest of the network to promote specific behavioral outputs. And so when we look at the data from the lateral PAG, we see a large drop in correlations with the rest of the social behavior network following orchiectomy. However, Pearson's correlation coefficients don't tell us much about the temporal relationship of these neural dynamics. And so to assess this, and in particular the directionality of the signals within the social behavior network, I'm using cross correlations to derive what I'm calling a directionality ratio. And here I'm going to just walk you through a bit of a schematic of what that looks like. And so in this example case, where region A and B co-occur in time, we see a cross-correlation that is symmetric across a lag of zero seconds, and this will lead to a directionality ratio of zero. However, if region A leads region B, This leads to a cross correlation that has most of its mass on the negative side of the lag, and this leads to a directionality ratio of greater than zero. Alternatively, if region B leads region A, we see the reverse. We're now on the positive side of the lag. The majority of the cross correlation mass is on that side, and we'll get a directionality ratio that is less than zero. When I apply this analysis to my data, I find that in intact males, I see this neat patterning of signal directionality within the network, where there are regions that sequentially lead each other through the network during social interactions. However, following orchiectomy, this directionality is disrupted, and the neat patterning no longer emerges. Additionally, if we focus back on the lateral periaqueductal prey, we see that this loss of gonadal hormones shifts this region, which we think of as the output of the social behavior network, from following the majority of regions or co-occurring in time with them, to now leading the rest of the social behavior network, suggesting that potentially the input output within this network is now disrupted. And so in conclusion, I've shown you that we're able to generate this computational pipeline to map global social behavior across time and longitudinally across a hormone state change. I think we might be having a fire drill, but I'm going to just try to wrap up really quickly and just say that I also see that gonadotomy reduces sex differences and mousestatic behavior and coordination within this network. And future work is going to focus on specific hormone sensitive populations within the social behavior network, in particular, the estrogen receptor alpha population. And I'm going to employ two color imaging methods to specifically test the hypothesis that this coordination by hormones is specific to these hormone sensitive neurons and not hormone insensitive neurons. And with that, I'd like to thank my lab, my mentor, Anna Greb Faulkner. Jorge Uribeira Garcia, who assisted me in developing these imaging tools, and a really talented undergraduate, Lucy Serres, who is running a lot of the behavior on this project. And I would love to take questions, but I'm going to ask that Micah pass them on to me and I can see if we can get a way to get them back to y'all because I have to leave my building. But thank you very much for the time and to all my funding and the Leading Edge for inviting me to get this off today. Thank you, May. That was absolutely fantastic. Please write your questions for her in the chat and I will pass them along to her later today and make sure that you will get your answers either through the meal or in another fashion. Wow, that's something. Let's go on with our next speaker. Our next speaker is Sarah Issa. Dr. Issa is a postdoctoral research fellow in the computational neuroscience at University of Colorado Boulder. Her current work focuses on how we adapt our decision strategies to match our environment and the corresponding mistakes that we make. In the future, Dr. Issa hopes to study the neurobasis of adaptive decision-making by pairing human behavior and electrophysiology with computational modeling to identify both physiological and pathological mechanisms of cognition. Please take it away, Tacha. All right, thank you so much for having me. So I just want to jump in and start off by really honing in on what we're talking about when we say we're doing adaptive decision making. So often when we're making a choice, we have to weigh information in order to make our decision. But when we talk about adaptive decision making, we're going to take that a step further and say that we also have to consider our contents. So, for example, if you're deciding what clothes to put on on a given day, you're obviously going to consider whether or not it's winter or summer. And also how often you think that that weather is going to change. So, considering our context is critical to really getting at the important information and how much we value and want to weigh those in our choices. So the way that we can approach understanding how the brain can perform these sorts of contextually dependent decision making is using this complementary approach. So as a computational neuroscientist, I want to integrate both data and theory. And there's a few different ways that people often do this. So on the one hand, we can look at neural mechanisms, in which case we can think about data, for example, actual recordings from human brains of epilepsy patients, as I did in my graduate work. And that tells us things like what kind of brain activity do we actually see during a particular behavior? Then on the other side, we can actually think about the theories. How do we understand what that brain behavior means? Then on the flip side of this, what I've been doing in my postdoctoral work comes back to cognitive strategies. And again, we can have data and theory. But in this case, we can use theory to educate these questions by asking questions such as, well, how should I use important information from my environment to optimally make a decision? And then comparing that to actual human behavior and saying, OK, well, how well do as a human actually compare to that optimal behavior? So moving forward, what I will be able to do is actually combine these and ask questions about how those neural mechanisms and that behavior are interconnected. So in order to show you how this will work, I wanted to give a few examples from my current postdoc. So the first question that I want to approach is what happens when we're in an environment to our, a particular environment to our decision strategies, and I'm going to pick one pretty interesting environment, and it's an environment that we call having asymmetric evidence. So just to break down what I mean by asymmetric evidence. I'm going to give this a little hypothetical situation. And so this is actually drawing from my graduate work in epilepsy. So let's imagine that we're a physician and we want to diagnose a patient with epilepsy. We can break down that decision into three main components. On the one point, we're going to start with our expectation. So what is the chance that a random person is going to have epilepsy just to begin with? So before we have any information, this is going to be what we call a prior in math. Then we're going to incorporate with that the evidence. So in this case, that's going to be the symptoms log. Does my patient say that they've had a seizure or a seizure-like event? And tracking that. So we're going to actually gather our information. Then the third piece is how do we combine our evidence and our expectation to make a choice? So this is our decision rule. And in this case, it would be our diagnostic rule. Now, in the case of asymmetric evidence, we're going to focus on just that evidence piece for a second and look at a symptoms log. In this symptoms log, what we see is our patient has said that there's one day that they've had a seizure out of the two weeks. asymmetric evidence would mean that that one seizure in my decision rule is worth more than those 13 non-seizure days. So I'm going to weight it more heavily than one of the other types of evidence. So that's where the asymmetry comes in. So if that's our decision rule in this case, we would say our patient has epilepsy. So the question we then posed was, well, if we have asymmetric evidence, how should we go about weighing this evidence and making a choice? And in order to study this, we built a model and we used a really, really simple task. So in this task, we have two jars. A high jar is going to have more red balls and the low termed jar has fewer red balls. And the goal of the task is that a sample of balls are drawn with replacement. And at the end of the sample, you're asked to decide if you think they came from the high jar or the low jar. Now, the way that we can actually change the symmetry of the evidence is by changing the ratio of red and blue balls. So in the symmetric case, we have that there's as many red balls as there are blue balls in the low jar. And so in other words, we have reciprocals between the two jars. In the asymmetric case, our red balls are going to be rare, making up less than half of the balls in both cases. And what this means is that observing a red ball is going to influence the decision more. So breaking this down into what a Bayesian observer is going to do, again we can write out our three components. So we have our expectation, which now we can say is going to be the probability that we're using either of the two jars. So we're going to say that they're equally likely, so we're going to set it to 0.5. Then, in addition to this, we have to consider our evidence. So now our evidence is going to be each of those ball draws. And what we're going to compute is what's the probability that a particular colored ball came from the high jar compared to the probability that it came from the low jar. Because our observer knows what the ratios are in both of the jars, they can actually compute this themselves. Then we have to have a way to incorporate these together. And this is our decision rule. And we're going to use a rule that's called the log likelihood ratio. And so that means that our belief here is termed ZN is going to update iteratively for every ball I see. I'm going to take the log of that piece of evidence and incorporate it into my belief so that at the end of all of my ball draws, if the belief is positive, it responds in favor of the high jar. And if it's negative, it responds in favor of the low jar. So next thing we did was ask, well, what happens if our observer does this for many, many trials? Now, the expectation says that what would probably happen should be that they respond as often for the high jar and the low jar, right? They're equally likely. But with asymmetric evidence, our optimal observer does something pretty surprising, which is that they responded for the low jar more often than we'd expect. So we asked the question, well, do humans also show this bias or do they try to match that expectation? So we gave this to about 200 online participants. And what we found is that even when we varied the difficulty of the task, our human subjects also showed this bias where they were responding in favor of that low jar more. So in order to identify how similar these two, the humans are to the ideal observer, we can use something called the psychometric function. So in this case, what the psychometric function allows us to do is to compare how often we respond in favor of the high jar. And we're going to compare that to that log likelihood ratio. So what was the belief of the ideal observer? So the trace that you see right now is what the ideal observer's psychometric function would look like. And for subjects, what we can do is take their data points and fit them using logistic regression. Now, what this lets us do is actually look at a few different components of that curve. So, in one hand, we can look at the bias. So, does this curve shift to the right or the left, which tells us something about whether or not our subject is compensating, maybe trying to match that expectation, or if they're accentuating that bias. And then on the other hand, we can look at the shape of the curve itself and see how similar our data points are to the curve, which is variance, and how much noise there is in the system, which tells us by the slope. So looking at the symmetric evidence case, what we found is that all of our subjects pretty much just showed a large amount of variance, which we'd expect. They have some noise. They're not doing things perfectly. But in the asymmetric evidence case, what we found is that there was a lot of subjects that actually showed a favor in favor of the low jar. So they're enhancing that same bias that we saw in the ideal case. So we asked the question as to where this suboptimal favoring was coming from. And in order to do this, what we did was to actually fit each subject's responses to a few different types of models or strategies. So in the one case, we have a nearly ideal set of strategies, which are the most complex, meaning that they're doing things closest to how the ideal observer does it. Then we had mis-tuned Bayesian models, which are trying to emulate what we saw that ideal situation doing, but maybe imperfectly. It's pretty hard to match those same sort of information biases. And then the third group were heuristics. So these are the least complex. They're probably using some simple model, not using all of the information available to them. What we found was that our mistuned Bayesian subjects were the ones that were showing the bias and that increase in asymmetry, while our simple heuristic models were the ones that described most of that variance. And this was really interesting because typically people have discussed the idea of a bias-variance tradeoff. But in our case, we found that the complexity mattered in the opposite direction. So this is one example of how human subjects both use information about their environment, but they do it imperfectly. So we wanted to then ask, well, how do we even learn and represent our environmental distributions? How do we know that it's asymmetric if it's not something as simple as easy jars? So the way that we went about studying this was to look at the limitations of working memory. So working memory is a well-studied cognitive feature. And the nice thing about it is that we understand some of the neural mechanisms that underlie it. And looking at the limitations might help explain some of our optimal or suboptimal cognitive features. So a very typical working memory task that we like to look at is the visual delayed response task. So in this case, what we have is that we are given a cue and working memory works on very short timescales. So we're going to have a delay of just a few seconds and then you're asked to respond for a particular feature of that cue. So in this case, color. Now, something pretty interesting came out recently, which said that even when we were given a set of colors that were uniformly distributed in terms of our targets or our cues, humans showed this bias where they preferred certain focal features. So they had a bias towards preferencing certain colors. Well, we hypothesized that maybe this wasn't an example of suboptimal behavior, but it actually was representing what's real in the environment, which is that certain colors are represented more. So we went about asking this question and asking how the brain could understand this environmental distribution. And so when we think about how the brain performs working memory, we can look at this task again and imagine that we have a bunch of excitatory neurons that have a preference for particular colors. So when we have this cue, we activate a bump of neural activity. It's going to have some fluctuations because the brain has some noise, but the memory gets retained. And wherever that bump of activation is at the end is our response. Now, a bump of activation is really no different than a dot, right? So we can actually make a simple version of this called a particle representation, where we have a dot that we're just going to trace how it moves in time. Now we can break this down and say, well, what does the space it's moving in look like? On the one case, we can say that it's homogeneous, which means that our dot just moves across the line. And so it's only going to be subject to those noisy fluctuations. But on the other hand, we could say that there's a heterogeneous landscape for this, which means that there's these valleys at the same peaks that we expected in the environment. And as we can imagine, this would help our error because it's very hard to climb back up out of a mountain. So we asked what the mechanism could be to actually build this and represent the environment. And the mechanism that we're using is fairly well known within the ideas of memory. And so we're saying that this is experience dependent. So for every task that we see, we're going to have a little activation bump of activity and that's going to actually have recurrent connectivity that's going to modulate so now our connections right at that point are going to be stronger so if we were to do this over and over again we see that the connections are going to change and so if we did this many many times so exposure to our environment over let's say the course of a lifetime we would learn to represent and have stronger connectivity at those more uh common features So we asked if this hypothesis was actually relevant to what human subjects did by looking back at a set of data where humans were actually presented with a heterogeneous environment. And what we found is when we looked at both learning models and static heterogeneous models and a homogeneous model, most subjects showed this learning, suggesting that we really do have this sort of experience-dependent activation of our environments. So just to bring this back to what it is that I want to do, my graduate work again was working just on the neural mechanisms and the beginning of my postdoc was working just on these cognitive strategies. But as you can see in this working memory project, my current and future work is really looking to encompass how we can have the mechanisms for particular behavior. So with that, I would just like to thank my three collaborative mentors here at CU Boulder, University of Houston, and UPenn. I would like to thank our funding sources and Leading Edge for having me. Thank you so much. Thank you so much, Thala. Does anyone have a question? I have a question then. In your example, it seems like the colors that may be more preferenced would be what if I would on that continuum say like, oh, that's the yellow, that's the green, that's the blue instead of yellowish green or yellowish orange. Can it be partially that just the semantic of memorizing during that one second? Oh, I saw a yellow dot or a blue dot and then you have to reassign that color in the response moment that you just find the color that is most similar to the semantics. That's a great question. So there's actually a whole lot of anthropological research that suggests that, first of all, we all have the same general semantics of colors, which is cool. But actually, the hypothesis for the overrepresentation of color comes more from our perception than it does from the semantics. But it's true that it could be like, oh, I just remember it was yellow. It must be somewhere in this yellow category. But. The hypothesis has been more that we just preserve, we interpret those colors more easily than the in-between colors, which maybe those are all the same, right? Semantics and why did we label that as yellow and not yellow green as our main color name? Yeah. Thank you so much. Very fascinating talk. This brings us to our last speaker of the day, which is Dr. Laura Desmond. Dr. Desmond is a postdoctoral research fellow in sensory neurobiology at the University of Oregon. During her PhD work in Paris, Dr. Desmond described critical steps of morphogenesis. morphogenesis of sensory neurons and participated in unraveling the versatility of their sensory functions. Building up on her previous observations, she is interested in exploring how this versatility subverts animal-microbe interactions and in turn shapes behavior. Dr. Desmond is currently addressing these questions in a specific context of olfactory-driven social behavior. And with that, I will give the stage to our final speaker. Thank you, Maike. I will share my screen. It says that my screen sharing is closed. I don't know if I... I'm not seeing anything. I don't see anything on this. Now, can you see my my face? Thank you, Micah, and thank you. This has been a fantastic session, so I'm really glad to have the opportunity to present my work today, all of you all. As Micah mentioned, I am going to talk about my postdoctoral project that aims at investigating the role of microbiota in olfactory-adhering social behaviors. But before heading into that direction, I would like to step back a little bit and tell you more about what I'm more generally interested in, in terms of research, and what are the scientific questions I want to address for my projects in the future. So more generally speaking, I'm interested in chemosensation, through chemosensory processes that are involved in animal-microbe interactions. how in turn they can shape animal behavior. So I'm interested in senses and sensory processes. And so when you think about senses, what might first come to your mind are the five classical senses as defined by Aristotle. So mainly our hearing, touch, vision, taste, and smell. However, as I tried to illustrate on that slide, there is a much greater diversity of sensory modalities that exist in the animal kingdom. some of which being present in only a subset of animal species. This diversity of sensory modalities is very much reflected at the cellular level, where specialized cell types, called sensory cells, are dedicated to the detection of specific sensory cues. So the way it works, the sensory cells express specific receptors that enable them to detect specific sensory cues from the external environment. And these sensory cells are going to transform the sensory information carried by these cues into an atrial chemical signal that can then be read by the central nervous system and interpreted by the central nervous system in order to adjust behavior. And so basically what these cells do is basically enable living organisms to probe their sensory environment in order to address their behavior. So I study sensory cells in zebrafish, and during my PhD I more specifically studied sensory, two sensory cell types in the zebrafish lava. So I studied first lateral line hair cells, which are cells very similar to the hair cells you have located in the inner ear and responsible for auditory function. In the case of lateral line hair cells though, they are located at the body surface of fish, of fish species in general, and they are involved in the detection and processing of hydrodynamic changes in the environment of the zebrafish. of the zebrafish larva in that case. I also studied a specific type of sensorineon located in the ventral spinal cord all along the central canal here. And these sensorineons have been shown to be involved in the detection of tail curvature during locomotion. And so they feed back onto the locomotive circuits in order to modulate locomotion and navigation in the zebrafish larva. During my PhD, what I did was to perform the transcriptome analysis of these two cell types in order to better understand what potential they have in terms of genetic expression. And this analysis really yielded unexpected results. So for the lateral line hair cells, what I found is that in addition to expressing mechanoreceptors that allow them to detect hydrodynamic changes in the environment, they also express a large diversity of chemoreceptors. And so here is the list of no less than 42 chemoreceptors expressed by these lateral line hair cells. And so these chemoreceptors will allow these hair cells to detect a diversity of chemical cues from the environment. And when I investigated the function of these novel chemoreceptors in lateral line hair cells, what I found is that these cells actually integrate both mechanical and chemical cues from the environment in order to shape and modulate locomotion and navigation of these chemical cues from the environment. Very similarly, when I performed the transcriptome analysis of the sensory neurons in the spinal cord, what I found is that in addition to expressing mechanoreceptors, they also express test receptors, so chemoreceptors. And so with these two studies, what I really demonstrated is that As opposed to what is usually thought about sensory cells as being very specialized towards the detection of specific sensory cues, sensory cells would be way more versatile than originally thought and would be able to integrate different types of sensory cues in order to perform their sensory functions. What was also interesting in the case of this sensory nuance in the spinal cord is that when investigating the role of these test receptors that they express, we found that these test receptors would be involved in the detection of chemical cues released by specific microbes that invade the cerebrospinal fluid, so microbial pathogens. And so that is very interesting because what that means is that Not only these sensory cells are more versatile than originally thought, but also this versatility would support the detection of cues from another kingdom, namely the microbiota kingdom, as opposed to only the animal kingdom. And so what that means is that this versatility would support this critical cross-talk between these two kingdoms, between the animal cells and the microbiota cells. So altogether, this will lead me to start asking the question of how hemisensoric processes can really support the cross-tool between animals and microbes and in turn shape animal behavior. So this question is obviously quite broad. I can't really address that question by itself, so I had to kind of reframe it for my postdoctoral project by focusing on specific aspects. So in terms of chemosensory system, I'm going to focus on the auditory system, and in terms of animal behaviour, I'm going to specifically focus on animal social behaviours. So now the question is, how does the olfactory detection of microbes shape animal social behavior? And so what is the link between these three elements, olfaction, microbes, and social behavior? You may not, I mean as human beings, we may not be as aware of it, but a lot of our social interactions actually rely heavily on affection. And we've all seen that in other animals'species. When going to the dog park, for example, we've all seen dogs sniffing each other's. And so what they do really when they perform this behavior is to smell each other to... recognize each other as specific individuals. And so once you have individual recognition that is olfactory driven, you can start having discriminatory behaviors towards the specific individuals. And I have here A set of examples of more complex behaviors that are discriminatory behaviors, either positively or negatively to one other individuals once the more simple task of recognizing individuals based on their body order, so through altercation, is performed. So what about microbes in all of this? Well, very interestingly, almost 50 years ago, a few studies showed that body odors in this body hyena that are responsible for territorialism behavior, so everything that's related to scent marking, is actually, these body odors are actually not only generated by the hyena themselves, but also by the microbes that are located in specific plants called scent glands. Based on these observations, what we know now from these observations and other studies since then, is that body orders are basically generated by the intricate interplay between the secretions of the host, so the animal, and the metabolism of the microbes that are rotating in and on the body of this specific animal and specific individual. And so after these first studies in the spotted hyena, other studies confirmed similar mechanisms in other mammal species, but also in birds and a lot of insects as well. More recently, it has been shown to be also the case in humans. So what we know now is that body odors in human beings are generated by microbes that are located in the atlants and release specific compounds that make up these body odors. This is interesting because that means that microbes that are located in and on your body, namely your microbiota, can generate these body orders and be at the origin. of the individual recognition processes. So this led me altogether to this Voigtman hypothesis that I'm addressing in my postdoctoral project. So this Voigtman hypothesis is that body orders are generated by microbiota, and other ones mediating olfactory driven individual recognition, which then leads to discriminatory social behaviors. So the way it works according to this model in zebrafish is that specific zebrafish individuals are going to have individualized olfactory fingerprints or body odors that are generated by the microbiota. These body odors and their components are going to be detected specifically by olfactory pathways from other individuals. which will lead to individual recognition and therefore discriminatory behaviors during social interactions. And so among these discriminatory behaviors, we can think about social preference, for example, so preference towards specific individuals, which could in turn enable some type of some degree of microbiota transmission, which would feed back onto the model itself. So, for the sake of time, I'm not going to detail all the results I've obtained since the beginning of that project, but I'm going to give you the main results and the main observations that I've collected so far. But obviously, I'm really happy to answer questions if you have any, and I'll write down in the details of these different pieces of evidence. So to start this project, one of the first things I had to do was to really confirm that specific olfactory pathways were capable of detecting cues of microbial origin. So this was something that hadn't been demonstrated before, even though we have evidence showing that body odors are generated by microbes. And so the first thing I did was to take zebrafish larvae at six days, more specifically. and to expose them to different bacterial preparations obtained from zebrafish bacterial isolates, namely bacterial species obtained directly from zebrafish microbiota. And so I used specific staining and specific markers to detect activity in the olfactory pit of this zebrafish larvae. And what I found was that different bacterial strains were capable of robustly activating as pointed out here by the arrows, were capable of obviously activating specific subtypes of olfactory sensor neurons in the olfactory epithelium of the zebrafish larva. And so during these experiments, I was really capable of showing that specific strains can activate specific olfactory sensory neurons, but also that I was capable of describing under which conditions these strains can activate olfactory sensory neurons. This was the first proof of principle on that project. The second thing I had to do was to really demonstrate that zebrafish larvae are capable of individual recognition and individual discrimination. So I established a social assay in which I place pairs of zebrafish, young zebrafish, so from 2 to 6 weeks of age, and I place them in secular arenas where they can freely interact and free screen for 10 minutes during which I record their behavior. And as a first readout, what I did was to measure the average inter-fish distance, so the inter-fish distance averaged across the entire recording of the 10 minutes. And by looking at this inter-fish distance, I can infer as to how this exhibits some kinds of preference towards each other depending on the pair of fish that I put together. And so I basically tested preference to one each other in two contexts. So the context of kinship and the context of kinship. So when he and when he are kin, which means they are siblings, so they are brothers or sisters, or non-kin when they come from different towns. And so what I found is that very interestingly, Zebrafish are capable from very early on, so from two weeks of age, they are capable of discriminating between individuals that are related to, that are related as opposed to individuals that are not related, so kin from non-kin, and they overall display a higher preference toward individuals that are related to them, so by showing an average inter-fish distance that is much shorter and significantly shorter. This was also true at six weeks of age, which means that this preference carries on across development. And so that was interesting because we know that zebrafish are highly social animals, but up to now there was no clear evidence that they are capable of discriminating between individuals and display some kind of discriminatory purpose. And finally, I really spent quite some time in trying to establish an assay in order to measure a microbiota transmission between different individuals. And so to do this, I use a specifically dried bacteria that I manipulate. to zebrafish overnight, and then I place the zebrafish together and let them interact freely and see how they can share their tagged material. So what I found with this assay is that microbiota transmission events do exist between zebrafish individuals, that these microbiota transmission events occur both at the level of the gut and at the level of the skin. At the level of the skin, they seem to arise much earlier during the time of interaction, so starting at one hour of interaction, while in the gut they seem to occur a little bit later, starting at three hours for some type of bacterial species. But these microbiota transmission events show that have increased over time depending on the duration of the interaction between the two zebrafish individuals. And so this is really important for me for my project because it has been quite difficult to assess microbiota transmission. It has been quite a long standing question in the field of how social interactions can lead to social microbiota transmission. And so finally I have the tools here. to really address that question and manipulate some of the social interactions to see how that affects microbiota transmission. So just to summarize about what I've shown you and what I'm at the stage I'm at in that project. Basically, I showed you that specific bacterial isolates are capable of obviously activating olfactory sensory neurons, which means that they are olfactory circuits that are capable of detecting and processing cues of microbial origin. In the future, what I want to really do now is to move on to live functional imaging to characterize this activation of refractory circuits in order to describe which pathways are specifically involved in that detection. I also showed you that the Blackfish individuals are capable of discriminating each other, and more specifically they are capable of discriminating between related versus non-related individuals. What I haven't shown you, but what is also very interesting is that so far I've been measuring social preference or individual preference by looking at inter-fish distance, but what we observed during this recording is that Zebrafish freely interacting as Zebrafish is actually highly complex behaviors. And so one of the goals now is going to be to characterize these behavioral patterns in order to have better insight into the complexity of the social interactions. And obviously to manipulate the olfactory system and the microbiota in order to confirm their roles in these social interactions. And so altogether, this will allow me to establish the causal link between microbiota olfaction and social behavior. Finally, what I showed you is that I have now an assay to measure microbiota transmission in zebrafish. I was capable of showing that these microbiota transmission events do occur, and they occur as early as two weeks of age. Now, what I really want to do in future experiments is to see how social interactions affect these microbiota transmission events. And to do this, I'm going to use specific social DPC models that we have available in the lab to see how that impacts the probability of microbiota transmission in non-Zipper fish. And so I'm going to end my presentation here. I would like to thank you all. And I also would like to thank all the members of my current labs, the Eisen lab and the McKinley lab, as well as all the members from my previous lab, my doctoral lab, the White lab, everybody who has been involved in helping me along that, along this project. as well as my project committee, all the facilities that have also helped me figure out how to establish this organization in order to be able to answer my questions, fundings, and obviously, lineage. And I'm happy to take questions. Thank you. Thank you, Laura. That was very fantastic. Does anyone in the audience have a question? Julie. Yeah, hi, hello, very nice talk. I was wondering, are you beginning to do some cross-fostering experiments to see with your fish if like if it's only like if the individual recognition is based on, you know, some genetic factors that makes that you're attracting some particular, you know, microbial more than some others? Or is that like really, the fact that you've been living in the same space basically and sharing the same, yeah. Yeah, yeah. So this is a fantastic question. So I haven't really shown any of the bigger results. I mean, all the results I've gotten from a particular type of particular part of the project. So I've tested both familiarity and kinship during my experiments. So familiarity, you know, touching a little bit to whether or not it would be because they've been living in the same environment, that would be showing this social preference to one specific individual. And so there is a continuum. So obviously familiarity is going to have a role as well. Whether this familiarity is just visual familiarity or factory familiarity, whether it is based on microbiota or not, I haven't really dived into that question yet. What I really want to do in the future experiments is to start manipulating microbiota. So we have tools to make these zebrafish germ-free and manipulate them with specific microbiota or microbiome consortia. And so the goal, and I've already started doing this a little bit, is to mimic kinship or mimic non-kinship by using the same microbiota or different microbiota. in individuals that are originally genetically kin versus non-kin to really have these different interactions and see how they impact the social preference and individual preference. But I haven't shown any of these results yet. It's actually quite complicated to disentangle all these different elements and different components, but they're all involved in some ways and I have some interesting results already. So I'm really excited about future experiments. Thank you so much and thank you, Julie, for that question. I think this will be the conclusion of our session today. Like I said, all talks were recorded. So if you also got interfered by a fire alarm or outside construction, you can listen everything back on YouTube in a couple of weeks. I would also like to invite everyone to come back on Wednesday, same time, same place for the talks about... I would like to thank all the speakers one more time. I really thought everyone did a fantastic job and I learned so much today.