New Models, Tools, & Targets for the Brain: Breakthroughs in Phosphoproteomics and Neurodegeneration

Welcome to the Axial podcast. Axial is an early-stage investment firm based in San Francisco. We partner with great founders and inventors investing in early-stage life science companies, often when they are no more than an idea. Axial is fanatical about helping the rare inventor who is compelled to build their own enduring business. Hey, Nader, how are you?

Good, good.

Let me make you a moderator because I don't want to be a moderator, you know, tyrant. You know, we should both have green logos by our name. Um, yeah, but thanks for taking the time to talk to me, uh, about your career and your research. Um, and, uh, yeah, it should be a really fun conversation. I think hopefully some buddies of ours show up and, you know, I'll deal with the audience. I think people can raise their hand. So I'll make sure to only let people who are somewhat relevant to kind of engage and ask questions. Sometimes there's trolls on Clubhouse. It is the internet. So I'll deal with the moderation. But everyone, really excited to have Nader, Morshed, postdoc in the Stevens lab at Harvard Medical School, yeah, who then worked in Forrest White's labs at MIT. Nader is kind of an expert in phosphoproteomics and then how then to use that toolkit to study the brain. And he's published some really outstanding research and I recommend you read his papers. There's a paper, I'll tweet it out or something. There's a paper in Nature and there's a paper in AmboPress and you can just follow Nader and he has some incredible research he's been doing. But thank you for taking the time for talking to us and maybe just give a brief introduction on yourself.

Sure. Yeah. So to give some background on myself, went to college at UC Berkeley, and I studied molecular cell bio and computer science there. I worked in a number of different structural biology labs starting from my first year in college going to the end, and that was the time where I really learned that I like doing research. I like answering interesting questions, and I like being able to think about biology as a puzzle to, you know, mull on and, you know, try and come up with some answer to some really deep question in some different field or subfield. So I worked in a number of different computational crystallography labs through all those years. But when I applied to grad school, I knew that I really was interested in studying protein phosphorylation signaling. I'd learned a lot about it across my different classes. And I was really fascinated by this as sort of the, one of the quickest mechanisms that cells use to sense and adapt to their environment. So, you know, we talk about like thinking about transcription factors that might drug a cell, that might drive a cell response, but upstream of that, there's a number of different protein interactions that rely on sort of different post-translational modifications., and the dominant one in that context is protein phosphorylation. So a large negatively charged molecule is added to different amino acids that change the physical property of the protein itself, causing it to interact with new proteins, have changed enzymatic activity, changes localization, and so on. So I looked through a number of different labs, and, uh, I ended up sort of discovering Forrest White, um, who's really a, a pioneer in developing this technology known as phosphoproteomics. Um, and he'd been, uh, applying it for a number of years in the context of, uh, of cancer immunology to try and understand what signaling pathways are telling a cancer cell to divide what are making it resistant to drug treatment, and so on. So that's sort of the early arc of my career. And once I joined the lab, I sort of got to thinking, okay, what do I actually want to do as a project here? But I really appreciated Forrest as a mentor. I think anyone who's worked in his lab can tell you about his booming laugh. And I was really excited about this technology, which I'll really elaborate on in a second.. But, you know, as we're, as we were thinking about these different projects that we could apply the technology to, we're sort of spitballing ideas, and Forrest threw out this idea of studying Alzheimer's disease. And he'd been discussing a collaboration with a professor over in the Picower Institute right across the street whose name is Liwei Cai. She's really a storied expert on neurobiology and has been doing tons of amazing work over the last 30 years. And that was really the point where I think the idea clicked for me. And, you know, I didn't have much of a neuroscience background, but I really got to work at that point to understand, like, you know, what are all the different open questions? How are— what are different subfields? What are techniques that you use to study the brain? And what are interesting questions? And you asked me this in the pre-chat, but, you know, one of the key technologies that we developed was a way to study phosphotyrosine signaling, and that had not really been done at all in the context of Alzheimer's disease or really much in neuroscience at all. So that was really the low-hanging fruit that we decided to start off with going in.

Well, I think maybe to— I'm always fascinated with like people's stories and then how that influences their taste and styles scientifically and beyond. And so at Berkeley, you worked with Susan Marcussey. She's incredible. And then you were a kitchen manager at the Berkeley Co-op. That's a lot of work. Kitchen manager. Geez Louise. That's probably the most stressful thing you did in college. But was there a formidable moment early on in your scientific career, maybe in college, maybe in grad school where you just really kind of fell in love with science research, or maybe even maybe in high school, but what do you think? Was there a certain moment or what kind of got you attracted to doing science and kind of being a professional scientist?

Yeah, I would say that it happened right around year 2. I I joined the lab sort of the spring of my first year in college and was very fortunate to work with a really interesting and thoughtful mentor who really sort of sparked my interest in science at the time. You know, it really came down to the mentors that I had through the years, but that experience really taught me about sort of thinking about Biology is these really physical objects, these physical mechanisms that interact with each other, but at the same time, they were so complex and really hard to really predict in any systematic way. And so this project that I started on was looking at protein alternate conformations in crystal structures. And this is sort of like a hidden feature that most people glossed over when they made when they fit protein structures to their crystallography data. And we could see in these sort of low-level signals that there were these protein dynamics. Proteins would switch in and out of different conformations. They'd have sort of maybe 10 to 30% of their molecules in a crystal structure as having like a flipped residue and, and these changes in the structure could be linked to these really fascinating and really therapeutically relevant, um, uh, features. Like you could find in— this was one way that people were beginning to discover allosteric sites. So you could see that multiple amino acids would flip through different conformations together, and that would mediate how a molecule binding on one side of a protein could change the activity in a catalytic site, you know, nanometers away. And I just thought that was so cool that, you know, that really connected sort of why you would think hard and put all these resources into developing these tools that detect minor features. And but then you, you see how it cracks open entirely new questions. So, you know, I think that was really, for me, one of my, my pivotal parts of my career.

Cool. And to really then build off that, so in college you're getting really good at studying proteins, structural biology, and kind of, I think you worked at Lawrence Berkeley for a while. So I think you probably got to, I had some friends work there. So you got to play with some pretty big machines, big lasers, which is Terrifying. And so you're in Forrest White's lab at MIT. He's the Koch and he's a pretty prominent cancer researcher. And so he suggests that you should do neuro. Why? How did that project emerge where you're in a cancer lab that's an expert in phosphoproteomics? What was the initial premise to say, oh, let's use all these tools we have to measure phosphorylation across the proteome. But let's do it for neuro. Was the premise based on like a genetic observation? Was it based on biochemical observation?

But what was kind of the initial hypothesis of kind of your research project during grad school? I think it encompasses a lot of different reasons, some just like financial or good for society. Alzheimer's disease, which is the disease that we chose to focus in on, is a huge unmet healthcare burden. And there's been almost no progress in developing any therapies for this disease that can actually halt the progression of disease. There's only some drugs that can reduce the symptoms, but inevitably you still go through the same arc of progressive decline. And so, you know, You know, this is a disease where there weren't any good therapies and people didn't really understand the signaling networks that were changing either. And I think, you know, we, we kind of went in with this idea, or at least I did this idea that, you know, these brains are kind of quietly dying, that you have these proteins that build up and then sort of everything just locks down and, and begins to just shut down. And, you know, maybe that's how neurodegeneration is happening in, in these patients, but I think, as I'll talk about with some of the results we found, it turned out to be the opposite. There was, you know, tons of activity. There were many, many different kinases that were active. And I think that sort of made it a really ripe ground to apply these tools of phosphoproteomics. And maybe this would be a good point for me to just really define, what that big long word means. So when we talk about different omics, people use the word omics to talk about different types of technologies that measure many different molecules of some category. So you have like transcriptomics is reading many different transcripts. And in a similar fashion, proteomics means that you're measuring many different proteins at the same time.. And typically this is done using a technology known as mass spectrometry. And, and in this process, you take a collection of proteins, you chop them up at predictable locations, and then use a mass spectrometer to identify the sequence of these short peptide fragments, which are the fragments of the proteins. And the mass spectrometer can at the same time quantify the abundances of those proteins across different samples. And phosphoproteomics is just an extension of that where we're measuring, in this case, thousands to tens of thousands of different protein phosphorylation sites in the same sets of samples. So we end up with this large matrix of different phosphorylation sites quantified in abundance. And if we want, we can also look at the protein levels in the same samples as well. So yeah, that's sort of the basis of the technology.

So on that point where you're explaining phosphoproteomics, from my understanding, the big bottleneck is the enrichment step. But what would you say kind of the biggest issue is, the biggest kind of bottlenecks for scaling, kind of measuring proteomics in general, but then also then measuring phosphorylation across the proteome? Because mass spec machines have gotten a lot better, more sensitive over the years. It seems like the enrichment step is still kind of an issue, but maybe—

but you're the one doing the experiments. Yeah, I think you actually absolutely hit on the challenge there. Really, the 3 key problems that go into sort of measuring phosphoproteomics data, you know, what separates like measuring 100 phosphorylation sites from 1,000 or 10,000 or 100,000 are these sort of 3 different problems. You have one which is proteins are sticky. They like to stick to everything. They like to stick to every single plastic surface, every tip, every bead in your LC-MS column. And so you can lose your sample to any random uncoated surface along the way. So usually to combat this problem, people just start with tons and tons of sample. But if you're looking at, you know, maybe a conservative chunk of tissue that is hard to get or a small region of the mouse brain like the hippocampus, then you can't really go in with, you know, hundreds of megs of sample unless you really want to be, you know, sacking and pooling many, many animals in these analyses. So that's just the problem of the proteins themselves. You can lose them even before you get to the mass spectrometer due to, these problems. And so this has been a huge barrier for measuring phosphopeptides, which in addition to having this problem of stickiness, their charge group also makes them interact with more charged surfaces along the way. So it's even worse than just proteomics. And then the other two problems are those of the actual sensitivity and ability to analyze complex samples of your mass spectrometer. So the way that these instruments work is that they, you have some incoming stream of peptides that are sort of separated by size and hydrophobicity by some liquid chromatography gradient. And those, as those peptides are being injected into the mass spectrometer, it's measuring the mass-to-charge ratio of each sort of packet of ions that's coming through. And when you want to, when you, when you see some ion that looks like a peptide of interest, then you'll isolate just that peptide and then fragment it into small pieces. And then you take the data from those pieces, the mass-to-charge ratio of all the different fragments that you've generated from each peptide, and then you use that information to reconstruct the sequence of the peptide. So the ability of the mass spectrometer to do that entire operation sensitively and fast, so being able to look at many different molecules simultaneously, it's improved a lot over the years. It's crazy how great it got even through the years of my PhD, but it's still orders of magnitude away from being able to analyze every single phospho site. So I think right now, you know, these are sort of the workhorse instruments of the field for doing phosphoproteomics. But, you know, I try and always pay attention to other technologies that might be able to read out protein sequences more sensitively or specifically. Yeah, but right now there's not really anything that's quite there yet. Protein sequencing is, you know, as I'm sure you're familiar with, is very early stage field at the moment.

Yep, I think it's kind of like sequencing, at least, and sequencing was— DNA sequencing was about 10, 15 years ago. So the kind of proteomics war now, and I'm not sure there's gonna be one standard. It's called probably orthogonal methods. You can look at the various papers and research and you get coverage by using several methods, whether it's mass spec or some other older technology actually. But maybe we can build off, then we can kind of then go to the first kind of your first big signature research paper in EMBO where you kind of use these tools to study mouse models of Alzheimer's and particularly 3. And so maybe you give a quick description about that research. And in particular, whenever I read a paper, I always feel like Cellieri, like Mozart. I'm like, oh man, I could never do that. I read your papers. I'm like, my gosh, I could never do what Nader did. Being honest, I'm reading this stuff. It's like, oh my, it's like reading Mozart and it's just like, I could never do this experiment. And so maybe talk about the paper, but also talk about maybe struggles you've had. Was it all smooth sailing to just measure the phosphoproteins of the three mouse models? What were you struggling with? During this research, and then you could also talk about the discovery you made, and then that can transition towards your Nature paper too.

Yeah, I think that all sounds good. Maybe before I dive into the results of the paper, I can talk a little bit more about Alzheimer's and sort of how we came to the different models that we analyzed. So if you talk about— if you talk to any person in the field of Alzheimer's, they're probably going to tell you about at least two molecules. The first is this protein called amyloid beta, which forms these large aggregates that eventually deposit into plaques in patient brains. And then the other protein is called tau protein, which is a protein that's hyperphosphorylated and generates these neurofibrillary tangles inside of neurons as they die. And The disease is really defined by these two, these two key marker proteins. Like this is just how people classify Alzheimer's patients now is if you have amyloid and tau, then you're Alzheimer's. And if you're not, then you're some other form of dementia. And, and that's worked all right to a point. And that's, that's at least been how people define People choose the mouse models of interest. They use either amyloid-driven mouse models or tau-driven mouse models to explore these two pathologies. But what was really interesting to us is that there has been a number of genetic studies, these GWAS studies, in very recent times that have connected a number of genes that are expressed in these glia cells And these genes are specifically expressed by these glia cells. And in certain ways that we still to this day have a poor understanding about is these genes can affect a patient's risk for disease, either being protective or being risk factors, or in some cases, causative factors of dementia in old patients. And so, and so we were really interested in trying to understand what these glia genes were doing, like how are the glia involved in disease? Are they sort of these bystander players that, you know, are just sort of cleaning up the mess as neurons die? Or are they contributing to disease at an early stage. And so we looked across these different mouse models, and we were actually quite fortunate that we were talking to Li Wei because she had an additional mouse model called the CKP25, which is a kinase-driven mouse model. It activates this kinase called CDK5, which produces both amyloid and tau and activates a number of intracellular signaling pathways leading to very early stages of this very strong gliosis phenotype in these mice. And this was an inducible mouse model, so we could sort of age these mice to early adulthood and then start it in a really controlled manner and look before any neurodegeneration is happening, what are the signaling networks that are involved? And so this is sort of the first paper of the first analysis that we did for this paper. We took Forrest's phosphotyrosine method that had been well optimized in the lab, and then we took some CKP25 mice at a very early stage of their neurodegeneration, before there's actual neurodegeneration, I should say, just when there's sort of these early signaling events happening. And we collected their cortex and hippocampus and ground them up and injected them into our instrument.. And what came out was a number of different signaling pathways. We could see that there were lots of kinases that were active. There were, um, lots of different types of signaling pathways. So we could see both tyrosine kinases being active as well as, uh, cyclin-dependent kinase and, uh, MAP kinase, which are sort of fall into a different category of, of kinase classification. There were tons and tons of these events going on and it was really active there. These were very active signaling networks before any sort of neurodegeneration was happening. And so this got us really excited and we wanted to say next, well, what's going on in other mouse models? Are these features just specific to this one model or are these same signaling pathways as a group going up across different amyloid and tau-driven models. So we collected more tissue, we looked in these two other mouse models. One is called the 5XFAD mouse, which has very severe amyloid pathology, as well as the tau P301S mouse, which is very severe tau pathology. And we put them all through the same pipeline again, and, uh, analyzed their phosphoproteomes. And what we found is, you know, sort of maybe, maybe what anyone could tell you is that each mouse model is unique. Each mouse model has different signaling pathways. We could at least map out which ones were occurring in which models. But as we looked across the different mouse models, these two proteins stuck out to us., and we got really excited because these proteins were expressed primarily in microglia. We could see across different sequencing atlases that it was really the microglia that expressed these genes and most of the other cells in the brain did not. And so there were two genes, one of which was a receptor, so we picked that one to really focus in on for the rest of the PhD. So that gene is called Siglec-F, and it's an immunoreceptor that's expressed by microglia. It's specifically expressed by a subset of the inflammatory microglia as opposed to the homeostatic microglia. And I guess I should just define my glia cells actually before I talk too much more at this point about our results. But the brain parenchyma, which is sort of just the, you know, the vast majority of the brain tissue, is composed of really 4 major cell types, you know, without going into all the niche ones. But you have your neurons, which conduct electrical signals and sort of pass around messages through your brain. You have the astrocytes, which are more, they, they serve as supporting cells in the brain. So they make contact with neuronal synapses and sort of buffer their ion release and their neurotransmitter release and make sure that the brain is functioning and healthy, okay, is healthy. And, you know, the neurons are supported in all their very intense activity. There's the oligodendrocytes, which wrap around different neuron axons and myelinate them, which allows the neuron signals to be conducted faster and efficiently across very large stretches of the brain. But the fourth cell type is these really small, really poorly understood cells called microglia. And they're very distinct from those previous three cells. So they come from a different lineage. They sort of colonize the brain. Late in development, and they have this very unique cellular identity which makes them in a way not really related to any other cell type at all in the brain, in the body. So, you know, people commonly refer to them as sort of the macrophages of the brain. There's these immune cells that colonize most of the brain tissue and and eat up different, uh, foreign pathogens that might make their way across the blood-brain barrier. But they look nothing like macrophages, and actually they have these very complex roles in sort of shaping neurodevelopment and protecting or, uh, contributing to neurodegeneration in ways that we really don't understand. And they have this very multifaceted role where they're both trying to act as an immune system against foreign invaders, but also an immune system against targets in the brain. And so they can, for example, gobble up hyperactive synapses that a neuron may not actually need for proper activity. So that's sort of the spark of interest in the microglia, these very unique cells. And so we could find this immune receptor that was activated on these cells, this Siglec receptor., but basically nothing was known about it in the context of the brain at the time. We could see that in the sort of complex relation between mouse and human Siglec receptors that it could have potentially been related to this other gene known as CD33, which had been recently identified at that point as a genetic risk factor for Alzheimer's disease. But because these gene families are really not very well conserved across the different species, we're really just guessing. So we ended up doing a lot of work to express these Siglec genes in an ex vivo environment. So in cells in a dish, which was a cell line model of microglia, and we looked across different Siglecs and in both the human and mouse family to see, do these different receptors at least signal in a similar manner? And we could see that indeed they did. They could all trigger this very active cell death response, which we went on to characterize as what looked like pyroptosis. So it was a pro-inflammatory type of cell death. The cells would die and then release these pro-inflammatory cytokines, which are contribute to more inflammation in, in their environment. And, uh, after, uh, a very long period of sort of painstakingly, uh, combing as many transcriptomic databases as we could find and testing different antibodies on human tissues, we're able to see that there was an additional Siglec known as Siglec-8 that was active on the microglia in narrowed in late-onset Alzheimer's patients. So it's— we, we couldn't exactly figure out what was driving it, but it seemed to be associated with this aging that would happen in patients. So we would see it most in the late-onset Alzheimer's patients and not so much in early onset before the age of 65.

So yeah, I think I talked for a while, but Did you have any questions up to that point? That's really exciting. So I didn't know before going into that research, you guys already had a kind of an initial hunch that essentially there's been a bunch of GWAS studies implicating microglia genes in Alzheimer's. And so you guys came into the research and said, okay, we have a hunch that if we do this phosphoproteomic screen, we're probably going to find stuff in the microglia, but you wouldn't know that beforehand, but you have a hunch based on the genetics. And so how exciting was it when you discovered Siglec-F's role, at least in early stage onset in these models, and the fact that that gene wasn't found in GWAS, right? TREM2 is, CD33 is, but you discover kind of a new receptor that GWAS didn't catch. I don't know, what was kind of that, what was that moment when you said, oh wow, we found a new, we found a gene that is not implicated in Alzheimer's that has a significant role in microglia.

Like, that's a pretty big discovery. Yeah, it was very start of year 2 that we sort of identified this really at the top of our screen in the CKP25 model. And that was when we really started getting to work in terms of trying to develop some tools to actually understand, like, how it signals. It was really exciting. You know, I think maybe there's this challenge of grad school that you get these really exciting results and then, you know, it takes like 4 or 5 years to really complete the final story of trying to understand what it means. But yeah, it was very early on. And I think, you know, one reason to look at in the phosphoproteome data to understand like what's going on in a tissue is that GWAS isn't going to tell you everything. Even if you sequence every patient on the earth, you're only going to find the genes that are risk factors for Alzheimer's if those mutations exist and they don't somehow cause some other disease well before the age of 65. So I think there's inherent limitations to these genetic studies that as you start to be able to analyze the phosphorylation system, you can find complementary data that you wouldn't ever see in a genetic analysis.

Absolutely. And then cyclic F and then cyclic A, consequentially, they have a very controversial function that's still unknown. And I'm assuming that's follow-up studies in the White lab and beyond. Either they could be inducing pyroptosis in microglia, but they also could be inducing a cytokine response And it may be both, but what's your kind of hunch? You've made this discovery that you've established the role of these receptors in Alzheimer's, at least models in this paper. What's kind of your hunch of what they're actually doing? That's probably a decade's worth of work for somebody else maybe. But they seem to have this dual function that's still kind of, it's still just kind of uncertain. They might be just killing glia cells, or they might be actually then increasing a cytokine response, maybe both.

But what's your kind of sense on things here? Yeah, so we were really thinking hard about this throughout the project, but unfortunately we weren't able to really successfully perturb it in vivo for a number of different challenging reasons. But I Eventually COVID kind of cut the project short and we sort of comment in the paper that we don't really know or we don't have data to really support one way or the other. But I think, you know, there's sort of two ways that we started, that we were thinking about it at the end. One is just as a marker gene. So you can think about these Siglecs as being, if they're being specifically expressed by certain populations, then that means that you have a handle on those cells, and maybe if you're able to, uh, put their ligand together in a, um, some sort of delivery mechanism, either like attaching onto, um, protein, the surface of proteins or viruses or particles, then, you know, maybe you can deliver something to these subset of these inflamed cells that, you know, might be able to modulate their immune response. So I think that was one portion that I think a lot of people who are much smarter at these technologies will hopefully be able to address in more detail. But then the other angle is, well, maybe this response is a good thing. Maybe this is causing these cells not to eat the wrong structures in the brain. Maybe it's not the amyloid plaques or the tau particles that are activating Siglecs, but maybe it's just that when they're inflamed, they eat natural structures that exist in the brain, and you want these receptors to keep them from chewing through every structure that you might want to preserve through infection process. So, it is quite challenging to really address that angle without knowing what the ligands are. So, this is another part of the challenge of Siglecs is there's so many of these genes, there's like 14 or 15 of them in humans and like 8 of them in mice, and they don't map to each other very well. And we barely have any idea what they're binding in the brain. Like we have some idea in some other tissues, but in the brain there's almost no answer to that question. And if you knew that it was not on these amyloid structures and it's on naturally occurring structures, then maybe that would inform that You'd not want to block this response. But I think you could think in the other way though, that maybe if you're able to overactivate these Siglec receptors in sort of an unnatural fashion, then if these microglia are causing other problems, if they're not actually helping with sort of protecting the brain in some disease context, if they're actually contributing to more problems than they're solving, then maybe this would just be a handle which you could use to just pop them. You could specifically take out these inflamed cells and then leave behind a greater fraction of homeostatic cells, or at least delay inflammation in a way that could help other cells pick up the pieces more. So I, and I think like that's sort of a very complex answer, but The problem is that across different early and late stages of different types of neurodegenerative diseases, you may want more of these than the other. So maybe at an early stage, you don't want to activate micro— these Siglecs if they're helping with quelling the amyloid response. But at a late stage, you might want to overactivate them because then you could stop them from eating up synapses, which is more of a late stage disease process.

So, that's sort of my non-committal answer. But that's a great point. I think the next step is glycomics, and it's after proteomics, right? The sequel, the trilogy. But that's a great way to kind of transition towards your Nature Aging paper where you probably created more questions than answers, to be honest. But you also, you address kind of this question, you built a map of phosphorylation changes between early and late stages in Alzheimer's. And so, you know, the Nature Aging paper builds off the EMBO paper where EMBO's focused on 3 mouse models. Now you focus on humans for the Nature Aging paper. You know, maybe you can discuss kind of that initial hypothesis where, you know, where these glia events you were seeing in mice in Alzheimer's, are they in humans? How did that, you know, how did you think about getting tissue samples? And then, how exciting was it to actually work on human samples now after kind of—

and what did you discover in the Nature Aging paper? Yeah, so I think you stated our question right there, that we wanted to know just what of these changes are connected in mice to humans. Are these all just phantasms of these mouse models, or are they real changes? Maybe that could help you narrow narrow down the list of, you know, thousands of dysregulated phosphocytes to maybe a few that are important for disease. Um, and we are quite fortunate that as we are characterizing these different, uh, signaling networks in cell line models, we met Diego Mastroni at ASU in the Banner Sun Health Institute who, you know, had been collecting these primary microglia cells from patients and trying to understand after recovery, if they, since they're behaving differently when you're culturing these primary Alzheimer's microglia in a dish, what are the signaling networks that are connected to these changes? So we helped him out with a pilot experiment early on, and he was so excited about the phosphotyrosine data that we were generating, like seeing all these very well-known transcription factors and kinases in our list of hits that he offered to collect some brains from their, from their biobank and send them to really see, like, in an actual chunk of human tissue what signaling networks were changing. And so we're very fortunate. You know, I was very thankful for every brain that I received throughout this project because that really helped accelerate us to get through the early stages and collect the data that we needed. And he ended up sending along 10 initially and then another 20. And then I pushed for another 20 eventually to make this, you know, large data set of different proteomics and phosphoproteomic changes in these 50 different patient samples. And as we're collecting and looking through this data, we started to see that there was just tons of heterogeneity. I mean, you could pick out these proteins like tau, which were very strongly increased in AD versus control. But then as we started to look across different Alzheimer's patients, we'd see that, you know, in some patients we'd see like 4 or 5-fold activation of different kinase phosphorylation sites or different protein abundances, and they'd be back low in another Alzheimer's patient. And we, It took us a very long time to understand, like, what does that heterogeneity really mean and how can we understand it? What are the statistical tools that can parse through this data in sort of a less biased manner? And I think probably one of the most interesting analyses of this heterogeneity to me came out of these two different approaches. The first was just to cluster the peptides. So we'd take our matrix, which was about 50 columns for each patient by about 88,000 rows for all the different peptides that we measured. And then we clustered it down to about 100 different clusters, which we then went through with some tools that could pick out sort of which clusters had interesting features. Were they enriched for tau? Did they have an overabundance of different glia-specific proteins and so on? And not allowed to at least say, okay, here are these different features of co-regulated, uh, here's these different co-regulated features that either go up or go down across different patients together. So, you know, 10 different proteins that follow the same trends, um, across the different patients. And that gave us sort of a set of disease features that then we could start thinking about as like, okay, well, this patient is positive for amyloid and tau. And we also see markers of neurodegeneration go down. And we also see these markers of either astrocyte or microglia changes going up. And so we then start to try and piece together which, which of these different disease features were associated with different types of stages of disease, say, or which ones were related to one another. But then the other insight came when we just used a dimensionality reduction technique called principal component analysis, or PCA, to find the major axes of variation in our data. And what we found is that when we looked across these patient samples, that we could sort of segment them into along two major axes. And one of these axes was very closely aligned with tau pathology. So as you went up in that major variable, you had a greater degree of tau phosphorylation events in those same sets of samples. But then the set, the first axis or the other axis, the one that explained an even larger fraction of the data's variance was really aligned with these two markers, one of which looked sort of like a neurodegeneration cluster, included a lot of different synaptic proteins, which had been very well characterized as markers of neurodegeneration. And in an sort of anti-correlated fashion, what went up in these same exact sets of patients was these oligodendrocyte proteins, as well as these very specific neurofilament proteins, which are expressed by neurons and have been really well characterized as these cerebrospinal biomarkers that are much more closely associated with neurodegeneration. So here were all these markers that were sort of discussed in the literature as these neurodegeneration features that had been validated across many different research groups. And we were seeing that they were all correlated with this feature that seemed to explain most of the data variance. And so we sort of termed this as the late-stage variable because it's seem to be more associated with, you know, actual synapse loss and neurodegeneration. And thus, across our different samples, we could begin to say, well, these are ones which have quote unquote an early stage because they have just tau pathology without neurodegeneration, or I should say, sorry, tau and amyloid pathology.. And then these are the ones that have late stage because they have these two other things as well as these neuroprotein, neuro, these neuroproteins going down. So once we had those classifications of these tissues based on their actual molecular profiles and not just what someone had written down qualitatively about a different chunk of tissue, then we could start fitting in each and every protein phosphorylation site that we're measuring into these early and late-stage disease processes. So that's sort of, as you said, this work raises more questions than we answer because now we have, at that point, the most detailed view to date of what different signaling factors are associated with early or late-stage responses. And so we could, for example, pick out a kinase like CaMKII and see that its phosphorylation site was closely associated with tau. And so therefore, maybe it's more of an early-stage kinase activation. We don't know if it's causing disease or if it's just a downstream effect, but at least we can group those things together at that point. And so at that point, basically the rest of the analysis just kind of fell out and we said these are interesting sites and interesting changes.

Yeah, I think the key— I need to— this— I need to— I've been I need to read this paper a few more times. There's just a lot of great work there. But the key thing is measuring phosphorylation changes across early to late stages in humans. And one discovery is really interesting is phosphorylation at serines and threonines is correlated with at least tau pathology in these patients. And then you also discover a variety of tyrosine kinases and moreover. Could you talk more about like, this is a lot of work and I'm still trying to appreciate the paper myself, but out of all this kind of all this research, kind of what do you think the kind of the most significant like observation you've made in terms of phosphorylation changes in these patients? Is there a particular target that's attractive? Is there a particular pathway? What's your sense on, out of all this work, what was the most, maybe what's a very meaningful lead to pursue?

And maybe the White lab, or maybe you are pursuing it. Yeah. So I think I, I don't really want to say that any one like newly discovered or previously known kinase is the best or worst target in this data, just because it's really difficult to answer that without a follow-up model with, um, you know, drugs that you can test. Um, but I think that one thing that stood out to me as like this, um, interesting phenomena that seemed to separate the patients that were early stage from the patients that were late stage was a number of these proteins that are expressed by the glia cells. And so we could see that there were these clusters of astrocyte and microglia proteins that were going up sort of slightly later than tau, so to speak. I'm putting later in quotes there, but later than tau but earlier than neurodegeneration. And within these groups of cells, we could see that a number of the proteins that were being upregulated were complement proteins, which you may know are these factors that are expressed in the brain that drive synapse elimination. So it's sort of a complex chain of event where synapses are first marked with this protein called C1q, and then that protein forms a— that protein is in a complex that goes on to activate another protein by catalytic proteolytic cleavage, and then that protein activates another protein, and then it eventually triggers the microglia to come in and eat up synapses. So this is one way that humans regulate synapse development and sort of the refinement of neural circuits early on in development. And here it was again showing up in these very old patients at very high levels and really closely correlated with this mid-stage of disease. So that sort of really got me thinking a lot about complement and glia signaling in terms of sort of mediating this progression from an early stage to a late stage. And, you know, this sort of leads me into where I'm doing my postdoc now in Beth Stevens' lab, who's really an expert in glia biology and complement signaling.

Absolutely. So complement, I mean, complement's the workhorse of innate immunity. And so kind of a So maybe we can shift gears and talk about your research in the Stevens lab, and he's one of the experts in microglia. And so you do your postdoc in the lab. Maybe talk about your research there. And is it a— are you building off your work from grad school?

Is something related or just something orthogonal? Yeah, it's somewhat of a new direction. So I think the The 3 ways that I'd sort of describe my work now is working on the models, working on the tools, and working on the targets. And this is all building on a lot of work that's being done by many of my colleagues in Beth's lab. But, you know, one thing that really stood out to us across all these different mouse analyses is these mouse models are not very good. They don't— capture a lot of the same genes. And the way that they develop these pathologies is not, is seemingly not that similar to what's the, the ways that these pathologies develop in humans. So we're still trying to go back to the base models and see, well, how can we generate tau species in human cells that actually looks something like the tau protein aggregates that are that exist in human tissue, which is very different from the tau protein that we're generating in these overexpression mouse models. And so we're looking across the different Alzheimer's genetic mutation landscape to see which factors might be best able to recapitulate these tau aggregation events, sorry, in patient stem cell-derived models. And so that's sort of one direction that we're working on now. And also looking at, you know, how this might be connected to different pathological glia cell activities. And then we're also trying to make some new proteomics tools that can more specifically interrogate sort of these, what we would term as pathogenic hubs in these patient tissues. So maybe the signaling network in a particular glia cell type as opposed to just looking in a brain cell homogenate, that would go a long way to sort of enhancing your specificity and clearing out a lot of the background signal while still giving you all the benefits of the in vivo signaling environment. So that's, I don't want to describe exactly what those tools are for the time being, but being able to hone in on the compartments and the cells would really go a lot to advancing our ability to understand these diseases and really getting more high-quality data going into these, these, sorry, these downstream analyses. And then finally, sort of connecting the signaling factors that we are able to identify with more functional knowledge of how different signaling— how different kinase perturbations affect the entire signaling network in these cells. So bringing in more CRISPR and more functional genomics to, um, you know, have more, uh, more of an understanding of, of how perturbations affect these signaling networks. And so I, you know, I really hope that long-term people are able to look at these studies and look at these data sets and better understand what it means for the system to decline in Alzheimer's, and especially having a much better idea of how different therapies that we're giving to patients affect these signaling networks as a whole. Like, for example, is amyloid therapy just able to affect a couple of tau phosphorylation sites but not the other glia signaling events or the other types of neuronal signaling events that are changing. Um, you know, we, we really want to answer that at a whole network scale rather than just relying on a couple biomarker targets as we've been doing for the past 20 years, because, you know, that's, that's really failed the field so far.

Absolutely. I think your work, it's already proven out that if you can use kind of proteomics methods, you uncover new mechanisms and new ways to measure disease progression that has been missed out by genetic methods. And so yeah, I think better models are incredibly needed. There's a guy, the Blanton Jones lab at UC Irvine. They do some cool research I like around new models. And then tools, totally agree. If you can measure the phosphoproteome and probably other modifications too, right? Sugar modifications are really underexplored in every disease, in particular the brain. And then the last part that will lead to new targets and new ways to understand how neurodegeneration progresses, but also then maybe potentially new drug targets that go beyond beta amyloid and tau, which are incredibly needed. That's the biggest— what makes your research so meaningful and impactful is identifying at least new leads for diseases like Alzheimer's, which has so much unmet patient need. It's one of the biggest problems to work on over the next few decades. But this is really, really helpful and really awesome. But long-term, you're doing your postdoc in Steven's lab, long-term, what are your goals? Do you want to start a lab? Do you want to get involved in drug development?

Maybe everything, but what would you say your long-term goals are? Yeah, I would say Long term, I'm still hoping to, uh, start a lab and be able to support other people, be able to sort of mentor people and inspire the next generation of scientists who will work on these problems, and hope that they're equipped with good technical and quantitative skills for trying to interrogate these, uh, really pressing questions. And, um, Yeah, I think beyond that, I'm still generally interested in sort of staying at this more. It's still a translational angle, but more on the basic side of just understanding how things work. Because I think that's, you know, a problem that can sometimes be missed in more industrial biotech environments where, you know, they want to pursue a drug target or, you know, have a fairly small program based around some type of technology. But, you know, I'm still interested in trying to adapt different types of tools together in sort of a way that may integrate different types of omics datasets and, you know, continue to understand or improve our understanding of how things are working when they are working or when they're working as they should in the brain. And then also better understanding what differentiates a pathogenic signaling pathway from one that's helpful for patients.

So still thinking on a basic translational research angle, I think, is sort of how I'd explain it. Cool. Yeah, I mean, I think your research is incredibly fascinating and incredibly valuable. So I'm sure the best is yet to come. And personally, I'm really excited to just continue to read the work you put out. And so, yeah, thanks for taking the time to talk to me. This has been a lot of fun. And I think a lot of people are gonna get a lot of value from this. But yeah, I think we're at the hour. I know you're pretty busy. So let me know if you're ever in Berkeley.

We can go to Saul's Deli, we can go to Cheese Board.

I love all these places. So, you know, come back to Berkeley. We go to Fire Trails. What I like to do in Berkeley. Berkeley. I like, uh, what's a cool thing? I like going to the Rose Garden. That's the thing I like doing. And so, uh, uh, you know, come back to Berkeley or come back to SF. Boston's too cold. Um, but yeah, thanks for taking the time, and I had a lot of fun.

And, uh, yeah, we'll talk soon.

Yeah, it was, uh, great to do this. So talk to you later. Talk to you later, Nader.

But have a great day.

Thank you, everyone.

Bye.