In this episode of PathPulse: Pathology Innovators in Action, we were thrilled to welcome neuropathologist Margaret Flanagan, MD, to talk with us about what is missing from digital pathology, highlighting spatial transcriptomics, multi-omics, and more. Margaret Flanagan, is an Associate Professor and Endowed Chair of the Baptist Health Foundation of San Antonio’s Distinguished Chair in Alzheimer and Neurodegenerative Diseases. She was also just featured on The Pathologist’s 2024 Power List as an Idol of Innovation. Her work focuses on understanding the mechanisms underlying dementia to advance diagnostic and therapeutic solutions.
Currently, Dr. Flanagan serves as Neuropathology Core Co-Leader for the South Texas ADRC, Co-Director of the Biggs Institute Brain Bank, and Director of the Nun Study on Aging and Alzheimer’s Disease. Her multidisciplinary laboratory integrates molecular platforms, biomarker development, biostatistics, and epidemiology to explore dementia mechanisms and develop biomarkers for early neurodegeneration detection. She is also a founding member of the National Digital Pathology Working Group and leads a 24+ month webinar series in collaboration with the National Alzheimer’s Coordinating Center.
Listen here:
Or watch this episode on YouTube.
Transcript:
Jake Brown: All right, welcome to the PathPulse: Pathology Innovators in Action podcast. This is a podcast that showcases pioneers, innovators, and forward thinkers within the digital pathology arena that are making a difference in day-to-day use of the technology. I’m Jake Brown.
I sit on the board for the Digital Diagnostic Summit, and today we are going to be exploring the topic of what’s missing from digital pathology.
With us today is Dr. Margaret Flanagan. She was recently highlighted on the website and magazine The Pathologist in their 2024 power list. Congratulations, Dr. Flanagan, and thanks for joining us today.
Margaret Flanagan: Thank you. Thank you very much.
Jake Brown: Well, would you take a second and maybe introduce yourself to our audience and tell us a little bit about yourself?
Margaret Flanagan: Sure. Yeah. So, Dr. Margaret Flanagan, I go by Maggie. I’m a physician, scientist, and a neuropathologist is my specialty in medicine.
And in addition to my clinical work as a pathologist, I also have a translational neuroscience research laboratory at the University of Texas Health Science Center in San Antonio, Texas. And our work is specifically focused on studying Alzheimer’s disease and other related dementias. And we primarily focus on, of course, brain autopsy material in order to move our research forward.
Jake Brown: Wow.
It’s amazing. That’s a lot to handle, a lot to juggle, but all awesome, amazing things.
Well, thank you for the introduction. And I’m really excited to talk to you today, mostly about your article and the things that you talked about within your article, specifically three things, two of which I had to do some personal research on to learn more about because they’re big words. And I was really excited to understand and ask you what these – how do you see digital pathology playing into it? So the topics that I’ve got for us to talk about are digital spatial transcriptomics, integrating multi-omics data into the diagnostic space, and then cost and accessibility. Those are the three topics that I wanted to highlight. So first of all, would you give us a rundown on digital spatial transcriptomics and how you see that playing into a need for digital pathology?
Digital Spacial Transcriptomics
Margaret Flanagan: Yeah, sure. So spatial transcriptomics is a kind of more recent technology that has become more readily available to researchers. And it’s pretty amazing because it kind of takes the technology that already existed, like single-cell RNA sequencing, but it adds an additional layer of giving you spatial information. So you can actually, you know, get the data that’s collected, you know, similarly, in those older techniques that were used, and then now with, you know, new platforms that are used for the spatial transcriptomics work, you can actually see, you know, which specific cell type or types are impacted and really get at the disease mechanism, I think a little bit more… with a little bit more granularity. So it just adds, I think, a deeper level of understanding to help us understand what is going on in various disease processes.
So it’s not just dementia or other neurodegenerative disease researchers they use this, but a lot of cancer researchers have also used this technology. So I think it’s really a powerful tool that can be used in multiple different disease settings to really help patients moving forward. So it’s really exciting.
I don’t know if you want me to go into kind of more detail about like, so we got this…
Jake Brown: Yeah, my question is, with that understanding, where does digital pathology play into that?
Because I think I understand your description of it. And I’m trying to, my education level is not as sophisticated as yours. And I’m trying to, like, yeah.
Margaret Flanagan: So it’s really cool because I guess one of the steps in the workflow for the spatial transcriptomics for these types of experiments it actually requires digital pathology, more or less. So when you’re kind of picking what we call our collection areas, you actually are able to do this by taking the sample, and we do multiplexed immunofluorescent staining. So taking, you know, three different markers of interest in whatever it – that will vary depending on a lot of things like what disease you’re studying, what your research question specifically is, et cetera.
But I kind of refer to the stains, the multiplexed immunofluorescence staining… so it just basically lets you look at three different, you know, stains on the same slide. I call those masks. And so a lot of the experiments that my lab does using this technology is particularly interested in different types of cells in the brain, such as, you know, neurons, but also we’re very interested in the supporting cells of the brain.
So things such as astrocytes and microglia, which are inflammatory cells within the nervous system specifically. So we kind of select, once we do the, you know, immunofluorescence staining, for example, then we have access to that exact sample as a whole slide image with that, you know, fluorescent staining. So it’s really great because then we also have it for separate digital pathology research studies in the future as well, in addition to using it to select what we call our collection areas. So the brain, the human brain is quite complicated.
We don’t have it as easy as a lot of the cancer researchers necessarily, who can kind of just have a sample of a tumor of some sort.
So we specifically are very interested in neuroanatomy and physiology and connections in the brain. So this technology is fantastic for that because we’re able to
Um, for example, the hippocampus is an area of the brain that’s important for memory. So, in Alzheimer’s disease and other types of dementia research, we really focus on this part of the brain a lot.
And we, uh, are able to kind of make these boxes, uh, to kind of have larger regions that we select, um, which if you’re a neuropathologist such as myself, you know, there are certain things that have been really well established over, you know, decades of research that we know, um, you know, the tau tangles that form inside of neurons and Alzheimer’s disease have been very well described so that we know, um, in the human brain that it spreads in a sequence. So we know, for example, it’s, uh, the tau tangles appear in the neurons in what’s called the entorhinal cortex before they later spread to involve, say, uh, the CA1 region of the hippocampus.
So it’s kind of, we’re able to actually strategically design experiments to make comparisons not just between samples such as control and disease, but also sort of within individual samples, right, like in the earlier and later involved regions. So because of this, the data that’s generated with the digital spatial transcriptomics is truly pretty powerful.
So, you know, you could even do an experiment on one sample sometimes if you have enough collection areas defined and an experiment designed appropriately that you’ll have enough statistical significance and enough cells identified and collected from these regions to actually be able to come to a conclusion perhaps just from one sample. So that’s pretty cool and then when you pick you know the regions with the boxes, then from there you can also design so like I said we like to look at these different cell types. So a great example would be the astrocytes which are a type of glial supporting cell in the brain and different from neurons. So everybody kind of has focused primarily on neurons, but you know the glial field is expanding more and more, which is fantastic, but in addition to neurons getting the neurofibrillary tau tangles in Alzheimer’s disease these astrocytes also accumulate different forms of tau.
So that’s kind of a particular research interest in my lab is kind of understanding the differences between the tau in these different cell types. So you can do that by selecting your masks, and then you know, whichever – and it doesn’t have to be necessarily cells. You can, you know, have different markers if you’re interested in other things such as synapses, you know, like maybe a postsynaptic marker, for example, but there are, of course, limitations and different platforms that can have higher resolution for smaller subcellular targets of interest versus cells themselves and then to kind of piece it all together for the transcriptomic data specifically you also can use these platforms for protein panels.
So, that’s what we do is kind of, you know, take a look at the results from the RNA panel, from the transcriptomic data, and then, you know, we typically also try to run the protein panel to kind of get that additional deeper level of information, which is fantastic because if, you know, you’re looking at things specific to tau with differential expression from the transcriptomic panel, then you can actually look at the protein panel results for using the same type of technology to kind of see, okay, how do the different types of phosphorylated tau differ, you know, in the same samples to really put it all together. So, yeah, hopefully, that makes sense a little bit.
Jake Brown: I think it does. So, if I was to… in layman’s terms, explain what I think you’re explaining. It’s basically that with the digital image, you have these stains or these masks that you can look at with a different image. But then you also have data alongside of it that is helping describe the areas of interest that you’re looking at that you’ve done, like a protein panel.
Margaret Flanagan: I’m almost like, I wish I could just show this to you visually because I have some really nice diagrams. I’m a visual person as a pathologist. So I’m like, I think I need to explain this more clearly. So basically, for both of them, you start with an unstained slide, like a tissue section that’s caught traditionally and placed onto a slide. And so we will try to do, you know, sequential sections right in a row, you know, so that you can have as close as possible to kind of run parallel experiments.
So you do the morphology markers are the exact same for protein or RNA panels. It’s just you’re either doing like sequencing, right, for the transcriptomic, whereas you are quantifying the protein specifically with a, you know, antibody probe approach for the proteins. But it’s, you know, the same platform, for example, that I’ve used most commonly can do both. It’s just kind of like different panels for the experimental design.
Um, so it’s pretty cool. So you kind of select the areas. So you like, first you do the staining with your markers. So the exact same, you know, you’ll see all the cell types on both. And then it’s just a matter of like, which data is collected, right? If you want the transcriptomics, or if you want the protein, or if you want both, then you just do both. Okay. So, hopefully, that helps a little bit. I’m like, yeah, I wish I can show you my beautiful presentation. I’m like, it’s, you know, diagrams and everything that kind of just feels really nice.
Jake Brown: Maybe we could do that. Maybe after this, we could, if you have a deck or something you want to share as a part of this.
Margaret Flanagan: Absolutely. Yeah, I’d be happy to.
Jake Brown: Okay, cool. Well, that’s really helpful and phenomenal information to hear and to try to learn from you.
Thank you for sharing that.
I would love to maybe transition and talk a little bit more about multi-omics and specifically integrating multi-omics data into the diagnostics. What do you have to say about that?
Multi-omics Data
Margaret Flanagan: Integrating it into like the diagnostic workflow pathology specifically?
Jake Brown: Yeah.
Margaret Flanagan: So I think that that’s exciting and that there’s a lot of potential to do this. But at the same time, I think that there are a lot of things that need to be streamlined and harmonized and maybe developed a little bit further to really feasibly be able to do something like that. Just because right now…
It’s such a new technology that it’s really been used on a research-only basis. As with all technologies in medical research, there’s usually a period of time where the alpha version, for example, and then you kind of have to optimize it and kind of make it adaptable for the specific use clinically. But I do think it has a lot of potential to be extremely informative for patient care, to, you know, really improve diagnoses for, you know, personalized medicine specifically, which is really, it’s really emerged how important personalized medicine is for dementia patients specifically because coexisting brain pathologies are the norm and not the exception in dementia patients.
So for many, many years, people focused a lot on Alzheimer’s disease, the plaques and tangles, which are very important, of course, but there are other things, right?
And it’s very, very common for coexisting brain pathology lesions to exist, even with the plaques and tangles of Alzheimer’s disease. So, for example, if, you know, it’s really exciting, there are a lot of amyloid beta targeting therapies that have been coming out in recent years that specifically target the amyloid beta plaques of Alzheimer’s disease, which is fantastic, but that being said, patients with Alzheimer’s, you know, they also have tau tangles.
So that’s great that we’re able to kind of take care of the plaques, but, or like the amyloid protein, you know, even earlier than when the plaques form, but we don’t really have anything that’s approved and available for patients who would also have, you know, the tau tangles. And in addition to that, the other coexisting proteins that are so common in their brains that we’ve, you know, described in neural pathology research for many years, but are really kind of in more recent years, we’re only really able to evaluate some of these things during life very recently.
So as things develop more and more for biomarkers, I think it’s going to be crucial to, you know, pair clinical medicine with transcriptomics and spatialomics because we’re gonna be able to essentially identify targets for different combinations of these pathology lesions in the brain in dementia patients that then we can take and appropriately give to patients because everybody has kind of different combinations in their brain, right? Like, you don’t treat all cancers the same, I always say, so why would you treat all dementias the same? So I think it’s gonna really have a very powerful translational value for sure.
Jake Brown: That makes a lot of sense when you say it like that. I’m curious to ask this question. In your ideal setting, let’s assume a lot of things. I’m looking over here at your article to make sure that I’m hitting the right thing when I ask this question, but you talk a lot about in your article that it needs rigorous validation, right, in order to be used clinically. So assuming that all those things pass and they are readily available for you as a neuropathologist, what’s your ideal way that it works in an ideal setting for you? Like in this utopian perfect world, what does that look like for you using digital pathology and multi-omics at the same time?
Margaret Flanagan: Sure. So honestly, I think in a perfect world, so I’m going to go extreme here, I’m going to assume that this also means that we have very accurate validated ways to detect, you know, what’s going on in the brain during life, which as of right now is it a hundred percent the case? But you know perfect world here. We’re getting closer and closer. So, assuming that these you know other tools during life for patients from pathologists are also available, like plasma biomarkers for example, which I also am involved with in my role as the medical director for a CLIA-certified laboratory, I know that those are coming out they’re supposed to be approved by the FDA in less than a year for tau specifically. So that’s, you know, close, but not officially, they’re ready for patient use. But assuming that these tools are available during life.
I think how this would be paired for, you know, multi-level omics technologies and clinical care would be to leverage this technology for the spatial omics to inform us on how best to treat these patients. So, for example, patients are different. So like, you know, there may be some patients who only have amyloid beta, for example, right? And so we have kind of that, some medications are readily available to at least kind of offer to patients who are confirmed to have the amyloid beta in their brain. So that’s wonderful, but, and there, you know, there’s even a subset of patients that would have amyloid beta, but not tau, for example, like in preclinical disease or vice versa.
And then we have things such as TDP-43, which is an abnormal protein accumulation that was, you know, initially described in
ALS, or motor neuron disease – that is actually very commonly found in dementia patients. And the same with alpha-synuclein, which is the abnormal protein that accumulates in Parkinson’s disease, but is actually extremely common in Alzheimer’s disease. And I believe the UPenn group has described that it’s actually present in over half of patients with Alzheimer’s disease. So that’s pretty huge, right? So I think that by, you know, designing and performing very carefully stratified experiments with this, you know, cutting edge new technology to kind of have the different profiles and compare the different order that the proteins appear in the brain. So it’s going to take a lot of data collection and building on that. And, you know, putting it all together. But once we have all of that finished, I think it’s going to be huge because we’re gonna be able to identify, validate, you know, the findings. So, typically, how my lab does it is once we find what we call a hit with this technology, we then validate it using more affordable approaches.
So, you know, multiplexed amino fluorescence and whole slide images, for example, because each slide costs a very large amount of money to do with this amazing technology, although I wish I could do it on every single sample.
So once the validation is done with what we’ve discovered, then we can really take those findings to, you know, rapidly accelerate drug development, have, you know, targets that we think are very likely or most likely to develop pharmacologic interventions to target that we think will have the highest chance of being effective in certain patient groups. But, like I said, the combinations are complex. So we really kind of need to do a lot of experimental design data collection to really figure out which things are best for which subgroups of patients. And then, kind of simultaneously going back to the clinical side, in theory, these biomarkers during life would be readily available.
And I’ve actually done some research in collaboration with a company that’s based out of Canada now called ReadySPEC with a collaborator, Dr. Swati More, at the University of Minnesota, where I was able to work with them to validate this pretty cool eye scanner. So, you know, something like that. So she basically was able to develop this technology and she’s in a very strong engineering background. So this is my version of explaining how it works, but I’m not the engineer; I’m the neuropathologist.
So essentially, it’s like taking a picture of the back of your eye with a camera, and the light like bounces off the retina and any proteins that accumulate there because it seems that, you know, the amyloid beta plaques and the tau tangles actually are present in the retina. And in mechanistic model systems, it’s suggestive that maybe even these pathologies accumulate in the retina, which people often say is an extension of the brain because it looks very similar. So people think that, yeah, these proteins could accumulate here before the brain, even decades earlier than when memory problems start. And we’ve also been finding that there are other proteins such as TDP-43 that have also been confirmed to be present in the human retina. So best case scenario is everybody comes in for their, when they need reading glasses, right?
Like 40s, so before dementia, and boom, it’s just like a picture back of your eye. There’s different unique signatures based on Raleigh scattering for soluble and insoluble amyloid beta.
But then validating this technology for all of the possible combinations of these proteins, then the patient would have this. And it’s pretty low cost, like about 20, I think, $20 to get the test done. You don’t even need to have air blown into your eye like that other test. So, hopefully it would be not an unpleasant experience for patients.
And then, years before any memory problems develop, you’d have that information on the patient.
And then you’d know exactly what combinations of things are happening in their brain and building up so that you could start giving these really tailored therapies to patients, depending on what the combination is that they have. And ideally, also, this could allow it to be given to patients as early as possible. Right? Like the key is before… irreversible damage happens because the brain is different than a lot of organs like the skin is it can heal easily if you cut yourself, it’ll heal. The brain has some, you know, regenerative repair capabilities, but it’s not really like that, so the earlier you can stop potentially irreversible damage the better for preserving memory function and quality of life for individuals so to really be able to To come up with the, you know, best therapies for these different subgroups of patients with different combinations that’s kind of where the omics stuff comes in the spacial omics to really identify delineate validate disease pathways to find the targets to, you know, drive drug development forward
I know it’s a lot, but you said in a perfect world.
Cost and Accessibility
Jake Brown: So, yeah, there we go. Well, hey, we’re, we’re running short on time, but I wanted to make sure you also had a chance to hit it. The last thing that you mentioned in your article is cost and accessibility. Um, so as we wrap this conversation up, uh, please take a minute to describe, you know, I think the way it was kind of, uh, displayed was like, these are the biggest hurdles to getting to where we want to go. Um, and so what thoughts do you have on that and any final things you’d like to say?
Margaret Flanagan: Yeah. So I mean, because it’s kind of more of an, in an experimental, you know, drug discovery development phase right now, I do still think that accessibility for hospitals and researchers everywhere so that they can work together to build, you know, these data sets to really figure out what’s going on like that’s going to be crucial, I think, is for access and for everyone to have access because not everybody is the same right like there’s sex-specific differences between male and female brains.
There’s also a lot of genetic differences depending on you know, your ethnicity. So, like two cohorts that I work with, a Honolulu Asian aging study was based out of Oahu, and Hawaii was all Japanese American men, a lot of them in the military. Whereas the nun study which I work with this is all women, largely Caucasian of European descent. So they’re very different, right?
So we found, you know, significant differences between those two groups.
So I think that’s just I want to highlight how important it is that we all work together to put together all of the data so that we can really have the information to know, you know, to apply to be most helpful for everyone, not just like one group of people, for example, and to collect all of this data. We’ll also, you know, really clarify this for us, not just, okay, well, maybe in the Haas, this happened, but like, maybe not so much in the nun study, and they have different, you know, more Lewy body disease, for example, the vascular disease was more common in the Haas and kind of piece it all together to really optimize um, the drug development and discovery. And in order to do that, I think, you know, collaboration is extremely important.
And, um, in addition to that standardization, um, and keeping things simple so that you can have a streamlined workflow that, you know, your data collection, everything is as harmonized as possible so that everybody’s kind of speaking the same language.
Um, and you’re following, you know, established guidelines for sample preparation, um, the data acquisition, uh, et cetera, so that you can reduce the variability as much as possible, um, and collect the best, you know, strongest data possible. And then, of course, from there, um, some other ideas that I have are because these platforms are so expensive that they can be really difficult for, um, you know, medical researchers to access even at this point in time. So I know in different countries, um…
For example, I did my medical school in Ireland at Trinity College Dublin, and in Ireland, in the HSE for their health care system, they have what’s called centers of excellence. So, you know, all the patients who have breast cancer will go and be treated at the center of excellence for breast cancer. So I wonder if a model such as that might be helpful so that, you know, maybe not every single Alzheimer’s disease research center in the country. Maybe not everybody needs the entire platform installed, right? Like maybe there are centers of excellence that you could collaborate and work together across the network to, you know, send your samples to, and then they’d, you know, be much more efficient running samples in higher throughput, which could also help to reduce overall cost in addition to variability, right, because it’s going to be collected at, you know, by the same person.
So I think that those things are crucial and hopefully will be adopted.
I think there’s also a lot more movement over to cloud-based data storage and access, which is going to, I think, really be helpful for driving everything forward and addressing cost and access in general. So, you know, centers of excellence, everybody needs to put everything together to really get to the bottom of these questions.
Similar to kind of a digital pathology research grant that I’m an investigator on, we have multiple sites nationally, and a couple internationally as well, that it’s cloud-based, and then we can all agree and follow, make sure we’re compliant for de-identification, but then all of that data will become readily available to researchers everywhere. So I guess kind of building off of some of the work that’s been done in digital pathology and kind of having the OMEX technology be an additional layer to kind of the infrastructure that’s been established there already. So yeah, those are, I guess, my… my thoughts.
Jake Brown: Yeah, I love it. I think those are awesome, and very, I mean to me, they’re logical.
Margaret Flanagan: Well, thank you, I think so too. But, you know.
Jake Brown: Okay, well, this has been an awesome discussion; thank you so much for joining us today. As we as we wrap this up is there anything else that you want to shout out or say?
Margaret Flanagan: I just encourage people to consider neuropathology as a career choice; whether it’s research or clinical or education-focused, it’s a subspecialty right now that it appears there’s a shortage of us. So, just advocating, consider neuropathology. It’s fantastic. Give it a shot.
Jake Brown: I love it. Okay. Well, yes, we encourage neuropathology. Let’s do it. Let’s get more people there. Lots of awesome stuff going on. All right. Thank you again.
And hopefully, we’ll have you back soon, and we can talk about some more cool stuff.
Margaret Flanagan: Awesome. Well, thank you so much.
Jake Brown: All right. Thanks, Dr. Flanagan.
