Scientist Stories: Christopher Bartley, Autoantibody Discovery in Neuropsychiatric Syndromes

All right, good morning everyone.

Happy Tuesday. So welcome to our department grand round series. And today it's my pleasure to be able to introduce the second of our trainee research award series presentations. In some ways these two presenters need no introduction, at least on the residency side. They have been superstars in the residency program since they started as PGY-1s. And I am going to actually get out of the way since we have a very limited time schedule and have their PIs present them. We are on a time schedule this morning where we have 25 minutes of presentation followed by 5 minutes of question. And then we will have the second speaker start at 9 o'clock. So I will play timekeeper. Thank you guys for being here and welcome to Grand Rounds. It's my pleasure to introduce Dr. Sam Pleasure.

So I get the easy task of introducing Chris Bartley who joined my and Michael Wilson's lab this last year. I met Chris when he interviewed for the residency program, for the research residency program, and I think we had sort of different thoughts in mind about what he was going to do. And I got very excited that he was interested in the project that Michael and I and Joe DeRisi have been trying to push off the ground for the last few years. Chris came from Yale where he did his MD-PhD with Angelique Bourdieu and also I think went to Yale undergrad where he worked with Pascoe Rakic as an undergrad. And so he's a very well-trained basic neurobiologist, developmental neurobiologist, and he got excited by this new program that we've been starting, which he'll tell you about today. And then, so we're very excited to have Chris in the lab. And no one have I ever experienced joining a lab has done any better at jumping in, like really just jumping in, you know. And so it's really been fantastic.

Thanks, Chris. Yeah, Dr. Kolechia is absolutely right. I changed course mid-residency. I was very interested in pursuing basic science.. Because of my experiences in the residency training program and seeing patients with comorbid autoimmune disorders and psychiatric disorders, I became interested in whether or not there was a way to test for a causal link between autoimmunity and psychiatric disorders. And so today I'm going to talk to you about the beginning of that journey. This is really just an introduction to how we're thinking about it. And I think that this vision is going to take, you know, maybe a decade to really play out, to really affect patients in the clinic. But that's ultimately the goal. And so today my talk is titled Autoantibody Discovery in Neuropsychiatric Syndromes. And I guess at this point I have to talk about disclosures. I'm a very small shareholder in a company called NowRx. It's not really important. They do delivery pharmaceuticals, not related to the talk today. So let's zoom way out and just take a picture of the landscape of mental illness today. So globally, mental illness is ranked second in terms of years lost to disability and disability-adjusted life years. And this is according to the World Health Organization's Survey of Global Burden of Disease. Between 2000 and 2016. And the estimated economic burden of mental illness in the U.S. is estimated to be $16 trillion over 20 years. And despite the global burden and the economic opportunity, currently only 3% of pharmaceutical innovation dollars are spent on mental illness. This category of illnesses is incredibly intractable, which limits investment in this space. And, you know, my personal theory, and I think this is a common theory, is that this is in part because we continue to treat mental illnesses as if they're uniform. If we take a look back 100 years, we can take a look at the precursor of schizophrenia, dementia praecox, which was originally described by Pick and really codified by Kraepelin. And, there were about 9 symptom domains. I'm showing 5 of them here. And, for many of you in the audience, they probably look pretty familiar. If we zoom forward 100 years and we look at the symptom domains that define schizophrenia, we see many of the similar characters that we saw 100 years ago. What's quite interesting is that in order to receive a diagnosis of schizophrenia, you only have to have 2 of these symptom domains, meaning that people with non-overlapping symptoms can be categorized as having the same syndrome. And so, uniform diagnoses, I believe, lead to uniform therapies. And because of this, we have a heterogeneous population defined by a single diagnosis, and we treat them all with dopamine receptor 2 blockers. A treatment that works for one person in this group has to have a reasonable chance of working for someone else in the same group in order order to get approved by the FDA. And so all of our approved medications are D2 blockers, which act as sledgehammers to the nervous system. So, my personal theory is that biological subtyping will help to drive therapeutic innovation and will perhaps allow us to define more homogeneous and smaller subgroups within this large heterogeneous population. And that may drive more targeted therapies. So, today I'm going to talk to you about my hypothesis, which is that autoantibodies cause mental illness in a subset of individuals. And in part, what I'm going to tell you about the evidence, and the evidence for this theory may not be familiar to everybody in the room. There's a lot of evidence for it and it's sort of focused discussion. I'm going to provide evidence for autoimmunity playing a role in schizophrenia and other psychotic disorders. And then we're going to move forward. We're going to talk about the autoantibody discovery process that we're using to discover novel autoantibodies that may target the brain and may drive psychiatric or neurologic symptoms. So, part 1. So, what is an autoantibody? I've used that word quite a bit. Antibodies are produced by the immune system. These are molecules that have unique patterns on their surface that recognize the patterns of other molecules. And the immune system will produce an antibody to target an infection and will clear the infection in a targeted way. Autoantibodies are produced by rogue immune cells that produce molecular patterns that recognize patterns expressed by the host. These antibodies can cross-react with host tissue, and if that tissue is the brain, it can lead to neuroinflammation and neurological psychiatric illness. So many of you have probably heard about this in some form or another. There was a very popular book called Brain on Fire, which was written by Susannah Callahan. She was a journalist in 2009 who was diagnosed with a recently discovered disease called anti-NMDA NMDA encephalitis. Prior to being diagnosed with this encephalitic syndrome, she was misdiagnosed with bipolar disorder, alcohol withdrawal, and schizoaffective disorder. And this disease is characterized by autoantibodies that mistakenly target the NMDA receptor, which is expressed on the surface of neurons. These antibodies, which are shown as the red Ys in that middle panel, will bind to the NMDA receptor, shown as those green and teal blobs on the end of the dendrites the dendritic spine. And they will increase the rate of internalization or endocytosis of these NMDA receptors. And this leads to psychiatric symptoms in the early phase of the disease. And as you can see there, after a typical viral prodrome, these individuals will experience delusions, hallucinations, mania, agitation, changes in speech, and disorganization. So, NMDA encephalitis is not schizophrenia. These patients go on to develop severe neurologic complications and seizures. But it serves as a model for how antibody that can interact with the host protein expressed by a neuron and drive psychiatric symptoms. So, what evidence is there for a relationship between autoimmunity and psychotic disorders? We're going to go over this slide quickly, but this is some epidemiological data showing studies in genetically distinct populations in Denmark and Taiwan. And they found that a subset of autoimmune disorders shown in green are overrepresented in individuals with schizophrenia.. And those shown in red are not overrepresented. And those shown in blue, rheumatoid arthritis, are underrepresented in people with a diagnosis of schizophrenia or psychotic disorder. This was replicated in a 25-million-person meta-analysis that did not include the original Denmark study. We can take a look at the genetic level. And we find that schizophrenia genetic risk loci are also associated with antibody-producing cells. These are our CD19 and CD20 B cells. And this is a study that was performed by the Psychiatric Genetics Consortium in 2014. Where they identified 108 schizophrenia risk loci. They looked at where those risk loci fell in the genome, and they predominantly fell next to genes that were highly expressed in the brain and genes that were highly expressed in antibody-producing cells. They removed the extended MHC region in the C4 association, and that association remained significant. And if we look at the biological level, we see that psychotic individuals harbor potentially pathogenic antibodies. Using a relatively loose definition, there was a study by Alkomya, Hassan, out of Germany where he found by looking in the cerebrospinal fluid of 124 individuals that upwards of 40% of these individuals have evidence of elevated antibodies in the cerebrospinal fluid. About 14% of them look like they may actually be pathologic. And that's about 3 to 5 times the rate of the healthy population. So we have epidemiological evidence that there's an association between autoimmunity in schizophrenia and psychotic disorders. We have genetic evidence. We have biological evidence. However, we still don't know what most of these autoantibodies target. And in a couple studies that have looked at this, they found that they were able to identify the protein targets of about 15% of the antibodies that were identified in the cerebrospinal fluid of people with schizophrenia and psychotic disorders. So, that's a quick overview of some of the evidence. Evidence that links schizophrenia and psychotic disorders and autoimmunity. That relationship exists for other mental illnesses as well. And now I'm going to talk to you about our autoantibody discovery process that we're using to identify novel protein targets of these potential autoantibodies in the biofluids of people with mental illness. And to start off, I'd like to tell you about the research cohorts that we're looking at. So the first research cohort is the Neuroinflammatory Diseases Cohort. This was put together by my co-mentor Michael Wilson. Who's in the audience here. It's currently comprised of about 1,200 specimens from individuals with suspected neuroinflammatory syndromes. Typically, cerebrospinal fluid and blood samples, sometimes in a longitudinal fashion. And a subset of these individuals have prominent psychiatric symptoms. Cohort 2 is comprised of the early psychosis cohort. This is 162 baseline serum samples from healthy controls, individuals who are clinically high risk for schizophrenia, and people with early onset schizophrenia. This is put together by Rachel Lowing and Mathalon. And we're starting off screening the baseline samples, but we have longitudinal samples as well. Cohort 3 is the University of Freiburg Biobank. We have 100— we're working to get— we've co-submitted an application to the University of Freiburg with Dominik Kendris and Bernhard Fiebet to obtain 180 CSF samples from patients with psychotic disorders. And then Cohort 4 is a growing biobank by Dr. Hassan out of Ludwig Maximilian University. And those last two cohorts don't have control populations. It's incredibly difficult to get control CSF. However, we can use the non-psychiatric patients from Cohort 1 as our controls. Today I'm just going to tell you a little about— a little bit about some data that we've obtained from Cohort 1. And this is just an overview of Cohort 1. Once again, 1,200 subjects. I looked through our database and found about 89 subjects that had prominent psychiatric symptoms. And I screened those for putative unknown autoantibodies. I'll talk to you about how I did that in a moment. And found that about 28 or about a third of those individuals had antibodies against unknown targets. And then on the left what you see is just a histogram showing the age distribution of the individuals in our biobank. You can see that we captured the entire lifespan here and we have about an equal proportion of men and women. And so to really So to get a sense for how this autoantibody discovery works, we have all of this evidence, but how do we actually find a novel autoantibody? We're going to follow the case of a single patient through this autoantibody discovery patient process. So let's meet our patient. Our patient is a 52-year-old male with HIV, well-controlled HIV. He'd been having significant mood swings and was becoming inflexible and stubborn. He later developed confusion, hearing loss, headaches, and had a seizure. And after a week of failed empiric treatment for TB meningitis, he received steroids for possible autoimmune meningoencephalitis and he improved. Now, this is not the sort of most psychiatrically rich patient, but we're going to use this patient as an example of how we can go from a patient to a definitive identification of an autoantibody. And I'll show you a little bit of data from some of the psychiatric, more sort of psychiatrically rich patients that we have. And so the question coming out of the clinic, given this clinical improvement, was really, you know, for the researchers, does he have an autoantibody? Because his clinical testing was negative. And if so, what does it target? And so we consented the patient. We have an active IRB protocol. We obtained his cerebrospinal fluid that was initially obtained in the clinic for clinical testing. And there was surplus cerebrospinal fluid. And we used this for antibody detection. So in order to detect autoantibodies, we perform something called immunofixation. Immunostaining on mouse brain tissue, and I'll explain that in a moment. And as a reference, I'm first going to show you what the immunostaining looks like for a patient who doesn't have an autoimmune disorder. So, what you're seeing here is a mouse brain section. This is a sagittal mouse brain section. On the left, we have the front of the brain and the olfactory bulb. On the right, we have the cerebellum. The dorsum is on the top and the ventrum is on the bottom. And the DAPI stain just gives us a sense of the architecture of the brain and its stained nuclei. And what we're going to be looking I'm looking for as we go through these slides is green. So at the beginning of the talk I mentioned that autoantibodies recognize protein patterns. There's about 85% conservation at the protein level between mouse and human. And so we can use that, we can use mouse brain tissue to pan for human autoantibodies that recognize brain tissue with relatively good sensitivity. We incubate the cerebrospinal fluid or the serum that contain the putative autoantibodies. Antibodies. We allow those autoantibodies to bind to the mouse brain tissue. We wash away whatever doesn't bind, and then we label those bound autoantibodies with some green fluorescent tags. So let's look at what the mouse brain looked like after we immunostained a section with our patient's cerebral spinal fluid. As you can see, there's bright green that covers the entire brain here, and this green indicates indicates that there is an autoantibody in this patient's cerebral spinal fluid that binds to mouse brain tissue. This is a relatively low-resolution image, but if we zoom in and we take a look at the cortex here, shown in pink, we can get a better sense for the anatomical distribution of the putative autoantibody that's binding to this mouse brain tissue. And so what you're seeing here are potentially two different autoantibodies. So most of our talk we're going to focus on the magenta arrow. And many of you can probably appreciate that there are these bright green hair-like structures that are pointing downward. The dorsum of the brain is on the top. And for the neuroanatomists out there, this probably looks like the axon initial segment, and that's what we said when we saw this as well. But you can see that there are these yellow arrows, and one is pointing to a blue nucleus and the other one is pointing to a teal nucleus. So green and blue may And so we suspect that there's another autoantibody that's expressed in the nuclei that may be different than the autoantigen that's expressed in the snare-like structures. And so we know that our patient has an autoantibody that targets brain tissue and that his autoantibodies appear to target a protein in the axon initial segment and in nuclei. But we don't know what they target. And so we moved this patient We've moved forward in the process to the identification of candidate antigens. Antigens are the molecules that antibodies recognize. And we have answered our first question, does he have an autoantibody? And now we're going to answer the question, what does it target? And so current screening methods for autoantibodies are relatively limited. Clinical assays screen for about 20 known brain autoantibodies. But we know that humans express about 20,000 individual genes. About 80% of those are expressed in the brain. And those genes code for hundreds of thousands or even millions of protein variants. Current research screening methods aren't much better than our clinical assays. And so we turned to a technology that allows us to, in a broad way, pan for novel autoantibodies. And this is called phage display. And this is a somewhat complicated slide. And because we don't have that much time, I'll just move through it quickly. I'd be remiss if I didn't mention that George Hessman and Gregory Winter won the Nobel Prize for their work developing the phage display in 2018. But let's leave them aside. And basically, a phage is a virus that infects a bacterium. We can genetically modify the phage to express human proteins or peptides. And we're going to start in the upper left and move our way down each column. In the upper left, what we've done is we've taken all known proteins and protein variants isoforms that are expressed by humans. We've bioinformatically broken them up into peptides. They're about 49 amino acids long. We've encoded those into DNA oligonucleotides and we've cloned those into individual bacteriophage for a library of about 730,000 individual bacteriophage that express the entire human proteome, all known expressed human proteins and small peptides. We're able to take that phage library and incubate it with blood or cerebral spinal fluid that contains a putative autoantibody. Antibodies bind to their peptide targets expressed on the coats of these bacteriophages. We're able to absorb out the antibodies and the phage to which they bind. That's indicated by the magnetic bead there. We can wash away the phage that don't bind, and then we can sequence the genetic material to identify the peptide that was encoded by the phage for that antigen that was binding. Encoded by your phage. We run some statistics to compare the number of observations of the phage that were pulled down to our negative control, and then we identify a subset that are considered to be significant. So just to remind you where we are, we have a patient who has a known autoantibody that binds to the sterol-like structure that we believe is the axon initial segment, and we're going to try to use phage display to identify the protein target. So I performed phage display on this patient's cerebral spinal fluid. We started off with 731,000 potential antigens. After phage display and a bioinformatics filter that I wrote in R, we whittled that number down to 303 potential protein targets. I then wrote a heuristics filter in R, and that whittled the number from 303 down to 14. And then I did what I call a manual neural filter where I looked at the published protein expression patterns for these 14 predicted targets and compared that to the immunostaining pattern for our patient. And there were 2 faves that came up as significant. And both of those faves mapped to a protein called AnkrynG. They shared an epitope. And AnkrynG is a well-known protein that's expressed in the axon initial segment. So this is pretty exciting. So we've identified a candidate that looks pretty good, but how do we we actually validated. So the first thing that we quickly did was to buy a commercial antibody that's known to bind to N-crinG and we did co-immunostaining. And so on the left we took a new mouse brain section and we immunostained it with the cerebral spinal fluid and that's shown here on the left. We bought the commercial antibody and we immunostained the exact same section and you can see those red hair-like structures in the center. And when you overlay the two green and red make yellow. And you can see that those hair-like structures have now turned yellow, indicating that the antibody from the patient's CSF and the commercial antibody bind to either the same protein or some molecule that's expressed very nearby Ankryn G. And so to convince ourselves for sure that Ankryn G was the putative— was the actual target of this patient's putative autoantibody, we looked at another anatomic structure in Ankrin G wild-type and knockout mice. And for technical reasons we weren't able to look at the axon initial segment in the Ankrin G knockout mice. And so here you're looking at the node of Ranvier. And the node of Ranvier is characterized by this classical bowtie structure which can be defined by Casper red staining on the outsides of the bowtie and Ankrin G expression or staining on the inside. That stains the nod. And so on the left you see that classic bowtie pattern with anchorin G at the center of the node of Ranvier, and that's what the green arrow is pointing to in the wild-type mouse. And in the knockout mouse you see that the knock anchorin G is missing. What's quite interesting is we see that the nuclear antigen persists, which confirms our theory that there's more than one autoantibody in this individual. So that was a pretty low-resolution image just to convince you a little bit a little bit further. We took higher resolution images and on the upper panels you see NCRNG wild type. Once again you see CSF stains the knot of the node of Ranvier. The bow tie is defined by Casper and the commercial antibody also stains the knot and that is gone in our knockout mice. So that was pretty cool. Our patient has an NCRNG-associated steroid-responsive meningioencephalitis. We were able to answer the questions that came from the clinic. Does this patient have an autoantibody and what does it target? And I have lots of wonderful pictures from psychiatric patients and I can only show you one today. So I'm going to show you another individual from Cohort 1 who we think we may have identified the autoantigen in, but we're not quite sure yet. We're still looking to validate it. And this So, this is from a 30-year-old female who was initially diagnosed with bipolar disorder with psychotic features. 4 years later, she developed an encephalitis. She had a lumbar puncture, and then she developed— she had a stroke and had another lumbar puncture. And, on the left, I'm showing you not a sagittal but a coronal section of a mouse brain with the cortex on top and the ventrum of the brain on the bottom. And, as you can see, there's prominent green staining throughout this coronal section of the mouse brain. It was immunostained with her cerebral spinal fluid prior to the stroke. And this is a higher resolution image of the cortex. And once again you can see prominent green staining. Now this pattern looks a little bit different than our HIV patient. This looks like a membrane-expressed protein. So we performed phage display on her cerebral spinal fluid. I ran it through the same filters that I talked to you about before. And one of the topics that we came up with was this interleukin-15 receptor alpha molecule. So, this is an immune chemokine receptor, but it's also expressed on the surface of neurons. And there's been some great behavioral work that was done in the early 2010s that found that interleukin-15 and its receptor, interleukin-15 receptor alpha, are responsible for regulating baseline activity in mice. As well as modulating depressive phenotypes and depressive acidosis. So we bought the commercial antibody against interleukin-15 receptor alpha, and we did exactly what we did with the previous patient. We co-amino stained mouse brain tissue with her CSF and the commercial antibody. And that CSF is shown on the left, and the commercial antibody is shown in red on the right. The nuclei are missing from this image. When we overlay the image, you see there's a pretty nice co-localization of these two signals. And we're looking to validate this study. So what are the next steps? We're looking to validate the other hits that we found for individuals from Cohort 1 that have prominent psychiatric symptoms. We're then going to apply this antibody— autoantibody discovery process to Cohorts 2 through 4 in order to discover novel, previously undescribed brain-targeting autoantibodies in individuals —ultimately the goal is that— pleasure labs. And there's work going on to clone autoantibodies from patients who enter our research study. There's work going on— excuse me— there's work going on in the pleasure lab to test these putative autoantibodies for pathogenesis. And ultimately we hope that this will allow us to identify novel treatments for these patients or to use existing treatments for these patients with novel inflammatory syndromes. So where am I doing this work? So my mentor is Sam Pleasure, and in his lab we're doing the mouse brain immunostaining. We're also developing in vitro and in vivo assays to test for the pathogenicity of these autoantibodies. Michael Wilson is my co-mentor, and in his lab we're working on recombinant antibody technology. Michael Wilson is the head of the B-cell laboratory for the department. Of neurology. And we're also performing the PHAGE display and doing immune profiling in his lab. Jeff Gelfand is the director of the Neuroimmunology Fellowship and he runs the autoimmune encephalopathy clinic. And he's quite interested in consulting on cases that we see in our clinics that may have immune-mediated psychiatric syndromes. And then Joe Derisi is part of this broader autoantibody discovery group. It was really his lab that developed the PHAGE phage technology that we're applying to encephalitis and neuropsychiatric syndromes. And we're working with them to build an analytics tool that will allow us to filter phage results on genes that are known to be implicated in psychiatric syndromes. And then I'll be continuing my affiliation with the Department of Psychiatry. And so we're going to be joining this broader autoantibody discovery group. So the conclusions are that autoantibodies are associated with schizophrenia and other mental illnesses. We can detect novel autoantibodies using this autoantibody discovery process. We can definitively identify novel, previously undescribed autoantibody targets. And that novel autoantibody discovery in mental illness may open avenues to new treatments. Thank you. Yeah, really simple question. So yeah, I think you're absolutely right that one of the things that's so attractive to me about this work is that there's a functional component, right? And so, genetics is trying to move into functional genetics to try to identify what is the actual consequence of these genetic variants that we're finding. Can we move beyond just the statistical association to actually understand how they're sort of affecting neurobiology? With the autoantibody, after we clone them from humans, we can actually use them as a research tool. We can put them into animal models, and we can actually look at the ways in which they perturb the neurophysiology of the animal. And, there are certain great neurophysiological deficits that we find in schizophrenia, for example, that compare pre-pulse inhibition, that are preserved all the way down to zebrafish. And so that serves as sort of a tractable sort of endophenotype that we can use to test whether or not our autoantibodies are affecting the neural substrate that we expect them to affect. Now, what if we put the autoantibody in and it doesn't do anything? Then presumably the autoantibody isn't itself driving disease. But there's a whole class of autoantibody-associated disorders, these paraneoplastic disorders. They're associated with autoantibodies where it's not clear that the autoantibody itself itself is pathogenic, but it serves as a reliable marker of the disease. And so, in order to determine whether or not these sort of new autoantibodies we find are actually associated, like, clear markers of a unique subset of people that have some different type of mental illness, I think we are going to have to look in the control populations. So, getting control CSF is not— or getting control serum is not a difficult task, but getting control cerebrospinal fluid is an incredibly difficult task. And so, that's one of the challenges that we face. We've looked at a few studies in Europe. And what they're doing is they're actually consenting individuals who undergo minor operations for lumbar punctures. So, someone who has some, you know, some hand problem might undergo an operation to have a lumbar puncture to get controls to this stuff. Anyway, I tried to answer a lot of your question there, but— okay, I'll make it 10 seconds. Yes, Dr. Herschel. Yeah. Yeah, this is a tough question. I think in the neurological world they try to come up with sort of clinical flow diagrams to identify patients who may have autoimmune encephalitis or limbic encephalitis who may not have a known autoantibody. One of the most reliable sort of clinical findings there is impairment in memory. And so, I think if you have patients who have some obvious deficit in short-term memory that's associated with their psychiatric syndrome, that doesn't look like it's related to a dementia disorder, that's a patient that you may want to send to us. Obviously, patients who have known comorbid autoimmune disorders like lupus may be patients that you want to send to us. And then, patients who are atypical and have episodic syndromes, you may want to send to us as well. Thank you.