Nikolai Axmacher is professor at the Institue for Cognitive Neuroscience at the Ruhr University Bochum where his research focuses on memory, spatial navigation, and neurodegenerative diseases. In this conversation, we talk about how he and his colleagues found that people with a genetic risk factor for Alzheimer's showed reduced grid-cell like activity and path integration ability, despite having no symptoms and still being in their 20s.
BJKS Podcast is a podcast about neuroscience, psychology, and anything vaguely related, hosted by Benjamin James Kuper-Smith. New conversations every other Friday. You can find the podcast on all podcasting platforms (e.g., Spotify, Apple/Google Podcasts, etc.).
Timestamps
0:00:05: The history of Nikolai's 2015 Science paper "Reduced grid-cell-like representations in adults at genetic risk for Alzheimer's disease"
0:15:57: Discussing the paper's main findings
0:38:35: Discussing Bierbrauer...Axmacher (2020), Science Advances
0:49:03: Applying (abstract) cognitive spaces to Nikolai's studies
0:59:10: Could we use grid cells as an early biomarker for Alzheimer's?
Podcast links
Website: https://geni.us/bjks-pod
Twitter: https://geni.us/bjks-pod-twt
Nikolai's links
Website: https://geni.us/axmacher-web
Google Scholar: https://geni.us/axmacher-scholar
Ben's links
Website: https://geni.us/bjks-web
Google Scholar: https://geni.us/bjks-scholar
Twitter: https://geni.us/bjks-twt
References
Bierbrauer, Kunz, Gomes, Luhmann, Deuker, Getzmann, ... & Axmacher (2020). Unmasking selective path integration deficits in Alzheimer’s disease risk carriers. Science advances.
Constantinescu, O’Reilly, & Behrens (2016). Organizing conceptual knowledge in humans with a gridlike code. Science.
Coutrot, Silva, Manley, de Cothi, Sami, Bohbot, ... & Spiers (2018). Global determinants of navigation ability. Current Biology.
Doeller, Barry, & Burgess (2010). Evidence for grid cells in a human memory network. Nature.
Ghebremedhin, Schultz, Braak, & Braak (1998). High frequency of apolipoprotein E ϵ4 allele in young individuals with very mild Alzheimer's disease-related neurofibrillary changes. Experimental neurology.
Hafting, Fyhn, Molden, Moser, & Moser (2005). Microstructure of a spatial map in the entorhinal cortex. Nature.
Hardcastle, Ganguli, & Giocomo (2015). Environmental boundaries as an error correction mechanism for grid cells. Neuron.
Huxley (1959). Brave New World Revisited. Chatto & Windus.
Kunz, Schröder, Lee, Montag, Lachmann, Sariyska, ... & Axmacher (2015). Reduced grid-cell–like representations in adults at genetic risk for Alzheimer’s disease. Science.
Quiroga, Reddy, Kreiman, Koch, & Fried. (2005). Invariant visual representation by single neurons in the human brain. Nature.
Saint-Aubert, Lemoine, Chiotis, Leuzy, Rodriguez-Vieitez, & Nordberg. (2017). Tau PET imaging: present and future directions. Molecular neurodegeneration.
Wills, Cacucci, Burgess, & O'Keefe (2010). Development of the hippocampal cognitive map in preweanling rats. Science.
Wills, Muessig, & Cacucci (2014). The development of spatial behaviour and the hippocampal neural representation of space. Philosophical Transactions of the Royal Society B: Biological Sciences.
-
[This is an automated transcript with many errors]
Benjamin James Kuper-Smith: [00:00:00] Today, I think it's going to be a very clear beginning because, um, as I wrote to you in the email when I contacted you first, the reason I wanted to talk to you is because you are senior author on one of my favorite papers and thank you. Um, so I'm referring to your science paper, which is one of the two papers we're gonna talk about more today.
Um, but I thought it might be nice to kind of say why, uh, what I like about the paper so much as a kind of introduction to the topic. So there's kind of three separate reasons in a way for why I think it's a really cool paper. The first is that it offers. A kind of first glimpse at what a biomarker for Alzheimer's disorder might be before the, before it's too late, basically.
'cause as far as I understand right now, you are, um, diagnosed, you're diagnosed with Alzheimer's disorder, and by that point there's, it's basically late, there's nothing you can do about it. But with this study, it seems like this might, [00:01:00] uh, offer some hope and a suggestion for how this might be achievable.
And I think that's a, I mean, that's a huge, that's a huge step forward. Um, the second thing for me is also, I mean, I work with neuroscience with humans and using FMRI, and I think there's always this, at least for me, this like question of like, does the FMRI signal really measure something useful or not?
And, you know, because it's so indirect and all that kind of stuff. But then. I dunno, every time I have these tools, I think, well it seems like there might be something there with your study where it seems to actually measure something that's real and that, you know, has some real world implications that Yeah.
Are more than just correlating some activity with some task. And the, the third reason is that I like that it's kind of, it has this like main story as in like the title is reduced grid cell performance in patients, well not patients, but in not patients yet. I mean, that's the cool thing. But then there's all these like [00:02:00] lots of, I feel like more than in other papers, lots of additional analysis that really provide a much more nuanced picture about, okay, so you have this reduction in the activity, but the behavior or the, the, the, they don't seem to be worse at the task, but they're compensating for it.
And this is explained by this and then you have like kind of multiple things all in one paper. So I thought I. Yeah, that's why I think it's one of my favorite papers. Thanks. That's why I wanna talk about it.
Nikolai Axmacher: Glad to hear about that. Glad to hear that.
Benjamin James Kuper-Smith: Yeah. So maybe, yeah, maybe as a very basic kind of starting point, uh, how did the project get started?
I mean, and how did, I mean, I think you did, I'm not, I very familiar with the other stuff you've done. It seemed to be more related to memory or something. So how kind of did you go from oscillations and memory to um, this more kind of slightly more applied work? Um, yeah.
Nikolai Axmacher: Yeah. Thanks. So, um, that's a good question.
So, uh, so actually, so at the time when, when we started working at the paper, which was in I guess like [00:03:00] 2000. 12, 2013. And then of course it took a while before it was published. That was a point in time when the new German Center for neurodegenerative diseases started in born. So it's called the Dset and E in German or like the, in English.
It's the German Center for Neurodegenerative Diseases. So it's like a, both the research and uh, um, clinical treatment center related to Alzheimer's and others, other neurodegenerative diseases. And it was just established probably in 2010, 2011. And, um, I was, happened to be in born at that time at, uh, the department of.
Epi ology where mainly did memory research as you mentioned, so shorter memory, long-term memory, memory consolidation. And then I found this idea about this, this, uh, center and the plans, uh, very intriguing. And I actually applied for, to become a PI in a junior research group, uh, at that center about memory [00:04:00] dysfunctions in neurodegenerative diseases.
So in particular Alzheimer's disease. And then I thought about which approach would be most useful to, to understand. Memory and memory changes and memory pathology and Alzheimer's disease. And the first decision sort of was to focus on very early disease stages for several reasons. So, uh, first because as you mentioned, uh, Alzheimer's is, Alzheimer's disease is developing very slowly, so across several decades.
So, uh, where the symptoms only become apparent at a relatively large stage. And so currently there's still no really effective treatment for, for ad despite many drug trials. And as you mentioned, this is probably because the treatment only starts when the patients. Uh, have like advanced or at least symptomatic Alzheimer's disease.
And when you look at the brains of these patients, um, they are very strongly destroyed already to a large degree. And I just thought it's, um, it's, it's quite [00:05:00] unlikely that in the near future, any possible treatment will be successful at that stage. So basically you need to look earlier and this, this slow, uh, progression is go both a good and a bad thing.
So the, the bad thing, of course, is that the, the disease starts so early even. But the good thing is that if it is possible to find some, either some biomarker or some cognitive marker, uh, that is able to find these very early disease stages, then um, then you may identify patients or healthy participants still who are still in a, who are already in a, in a treatable, uh, stage.
So this was the first decision not to study, uh, patients, but really. Healthy participants with a risk factor. And then as a risk factor, we focused on a specific genotype, the apo e genotype or apo lipoprotein E And this is this single one, uh, genetic alteration, mutation, polymorphism, however you wanna call it.
Um, that is, um, explaining, [00:06:00] uh, the risk for Alzheimer's disease more than, uh, any other single gene. Um, as a side note, it should be mentioned that there's other kinds of Alzheimer's disease, so the familiar type of Alzheimer's disease where there's also genetic mutations that however, lead with a penetrance of almost 100% to Alzheimer's disease.
So then it's not a risk factor, but really a gene that causes, um, the disease. But this is, this is a very rare case, and I, I found it more rewarding to, to focus on the, like the more common form of Alzheimer's disease.
Benjamin James Kuper-Smith: I have one quick question about the patients here, is that in your paper you mentioned that there's a tenfold increased risk of developing some disorder if you have both of these alleles.
Nikolai Axmacher: Mm-hmm.
Benjamin James Kuper-Smith: Um, but I don't know what the baseline is. Uh, so is it like, and I guess this is already maybe a fairly technical question, but my question, like, what I was wondering when I read that was. Are, how likely, actually, are they absolute in absolute terms to develop Alzheimer's disorders and therefore, how [00:07:00] likely are you actually to have people in your study who, yeah.
Nikolai Axmacher: Yeah. So I, I would need to, to look up the numbers, uh, for, for Alzheimer's disease, but in the general population, the, the, uh, risk of getting Alzheimer's disease changes with age, certain age of about 65 is, is relatively low. It's a few percent. At the age of 75, it's already above 10%. And at 85, it's, I would guess roughly at about 20%.
But I, I would need to look up these numbers.
Benjamin James Kuper-Smith: Okay. So it,
Nikolai Axmacher: so it's like a threefold increase is clinically relevant. And this is also one of the reasons why, uh, when we, uh, recruited the participants. The entire, um, study would be completely double blind. So we would not look up whether a particular person, um, whom we scanned was a risk carrier or not.
So we couldn't tell. We only looked that up after all the data had been, uh, collected and had been fully pseudonymized. And also there was some participants who were asked us whether we could tell them. [00:08:00] And, um, we had a long discussion with the IRB and born at that case, uh, at that time. And, um, the institutional review board told us that we would not be allowed, even if the participant would be interested to look that up, because this is information that can be psychologically very distressing because you know that you have a clinically relevant risk factor for Alzheimer's disease, which is currently not really treatable.
So that's why, why it was fully double blind.
Benjamin James Kuper-Smith: But wait, just, okay. Just a, a quick question here again, because what I assumed what happened was, is that you. I dunno, because somehow I assumed the recruitment of these participants worked something like, I dunno, you had people with Alzheimer's and because it's a genetic thing, you can recruit the children or something, or something like that.
So I assumed they already knew this, but this was just general population. And then you gave them a DNA sampling.
Nikolai Axmacher: These were just students. So basically what we did was to, to go to the, to the big lectures, uh, at Bond University. Um, so there's some lectures with hundreds of [00:09:00] students. And then I asked the, the, the faculty member who was giving the lecture, lecture whether I could have a, like a two minute slot at the end of the lecture and tell about the study and, uh, ask whether there would be some, some volunteers who would, uh, be interested in, in potentially participating.
And then afterwards we were standing outside with some students and colleagues outside of the lecture hall and would genotype as many participants as possibly as possible. And then from this large database or. More than a thousand participants. We, uh, randomly selected risk carriers and non-risk carriers and, and invited them.
But when a given person arrived at the lab, we wouldn't know whether this was, was a risk carrier or not.
Benjamin James Kuper-Smith: Okay. Okay. Yeah.
Nikolai Axmacher: So this, this was like the, the basic setup. So we, we decided to, to look at young participants and we decided, uh, to look at a PE, uh, forest carriers. And, uh, looking at young participants also, uh, had the benefit that we could do [00:10:00] this relatively long and demanding FMRI study.
So this would not be possible with an Alzheimer's patients. It probably also wouldn't be possible with someone suffering from my cognitive impairment and maybe, uh, maybe it would be possible, but it would be very. Demanding and difficult to do, um, even in the average healthy 70 years old because these, for elderly people, it's more difficult to lie still in the scanner without moving a few millimeters for more than an hour and doing the virtual reality paradigm.
I mean, just because of like back, back problems and so. So we, we could do this like relatively complex paradigm in, um, in young participants. And then, then we thought, which, which paradigm should we do and which area should we target? And we relatively early decided to, to focus on the enteral cortex, which is.
One of the most, uh, one of, one of the brain areas that is affected earliest in Alzheimer's disease. There are some neuropathological studies that is, that suggest that it's [00:11:00] even the first area to be affected. Um, and the antinarcotic is an input zone to the hippocampus. And whereas the hippocampus is extremely well characterized, it's maybe the, the both, the best characterized brain region in both rodents and in humans.
This is not necessarily the case for, for the interal cortex in particular with regard to memory. There's many studies on the division of labor between the hippocampus and the neocortex, and also between the hippocampus and in an other adjacent area, the perh cortex, suggesting that the hippocampus is more doing relational memory, associative memory, recollection based memory, and the perineal cortex is doing item based memory or other kinds of memory.
Uh, but the, like, the role of the enteral cortex for memory is not, not really that well, that well known. So if you look in PubMed, you find that the number of studies. Um, that look at memory in humans, specifically in the enteral cortex is, is very low. There's a few studies, but, but it's, it's [00:12:00] relatively few.
So then we thought of alternatives and so other cognitive processes. And at that time the, uh, like one of the most investigated, uh, research areas in with regard to medial temporal lobe were the grid cells that were just discovered in 2004, 2005. And the Moses, uh, lab and the grid cells are a particular type of cells that is probably putatively relevant for spatial navigation.
And, um, that fires in a very specific manner. So like a single grid cell does not fire at one location when you walk around in a room or when the rodent runs around in an environment, but fires in a very repetitive, uh, pattern in, um, across the environment And. Therefore has been, um, suggested to, to be a neural signature or a neural, provide a neural metric for distance estimation and path integration.
I guess [00:13:00] we come back to that a little bit later. And when we were interested in, uh, at the time when we planned the study, there was just one publication from Neil Burgess, uh, group with Christian as a, as a first author that was published in nature in 2010. And that suggested that the grid cells could be measured, could be studied non-invasively using functional MRI.
So a new Christian, uh, from previous, uh, meetings, and, uh, contacted him and asked him whether he would be interested, uh, in sharing his paradigm to apply it in the genotype participants and test whether the genotype participants with a genetic risk for Alzheimer's disease would show any. This function at the level of this network, FMRI pattern of grid cells in adrenal cortex.
And, and he agreed and, uh, send us the paradigm. And, uh, then we, we scanned the risk areas and non-risk areas. So, so that was [00:14:00] basically the, uh, the study developed.
Benjamin James Kuper-Smith: That's really interesting. Somehow. Yeah, it's fine. I, I never realized that there wasn't much research on memory and the hin cortex. I mean, I guess I never, I was never very particularly interested in memory per se.
So I've, you know, I've had like, like I did psychology and cognitive neuroscience in my bachelor of masters. So, you know, I've heard about these things, but it never occurred to Yeah, that's right. The memory research is always about the hippocampus. It's never about, um, internal Tex. And for me, just because I, um, in particular, I took, um, I did my masters, um, at UCL and took Neba justice course with Casper Barry.
They taught a course on neuro computation, so that was all about memory and the hippocampus, internal cortex and grids as places like. And it never occurred to me that that's right. The only on a cortex was only really mentioned in relation to grid cells.
Nikolai Axmacher: Yep.
Benjamin James Kuper-Smith: Um, yeah, I didn't realize it was that specific.
Yeah.
Nikolai Axmacher: I think it's still an, an area for, uh, for research. So I think it would be, would still be interesting to, uh, try to [00:15:00] elucidate and investigate the, the function specifically of the enteral cortex for, for memory. It's not, not an area that I'm, that I'm currently working on, but I think it's, it's still a, still a bit of an open question.
Benjamin James Kuper-Smith: Yeah. And, and the, I mean, I talked to Kate Jeffrey on the podcast also a little bit about kind of how this was discovered because she recorded enteral cortex grid cells in like, in the nineties, early, like mid nineties, but didn't realize she was doing that. Mm-hmm. Uh, because the box, the arena wasn't big enough, basically.
It wasn't enough to see the pattern. And I think she also said like, it's, it's kind of, it was the obvious place to look at where play cells get the firing from, just because it's one synapse away from it, but mm-hmm. Like, that's it. It's just basic. You just go like, well, where does it get its signal from?
Let's look at that place. But, huh. Okay. Yeah. I never realized that Interal cortex wasn't as well known as hippocampus, but now that you say it's, yeah. Kind of obvious. Mm-hmm. Yeah. Um, do you maybe want to [00:16:00] outline then, uh, I guess we've been kind of talking about the study without actually saying what you did and found Yeah.
Just a brief summary of that.
Nikolai Axmacher: Yeah. So the, the participants, uh, conducted a, a spa navigation study while they were lying in a scanner. So, uh, in a virtual reality paradigm, which was like adopted from the previous study of, of Christian Dola in 2010. So, um, in this, uh, task, the participant had to learn the locations of different objects in the environment and try to remember the locations and then move to the location.
So specifically in each given trial, they saw one object, so, which were just everyday objects, like a, a clock, or we had an eggplant, a baby bottle, and so on. So all kinds of objects. And, um, there were eight objects in total, and each of the objects was initially shown at one of the locations in the environment.
And the participant was asked to go there, to move there with some button presses in a, on a keypad. So then in each trial they saw one of the objects. So let's say the, uh, eggplant. [00:17:00] And they moved to a location in the, in the arena, in the virtual arena where they believe the, A claim was actually located.
And then they pressed the button. And at the beginning with these eight objects, it's, it is terribly complicated. So the participants essentially perform at chains. So they just place it somewhere, but then after placing it, they see the correct location in each trial. So they can then go to the correct location of, of the eggplant and re-encode or again, learn where the eggplant belongs.
And then in the next trial, there was another object and so on. So across the entire experiment, which took more than an hour, they saw each object more than a hundred times. So, uh, you can nicely see how the performance improves, uh, for the objects, whereas at the beginning, the participants are very bad, they get better and better, um, across the, um, paradigm.
So what this gives you is a measure of. Um, spatial memory, which is the inverse of the drop error. So the further away in a given trial, the participant drop an [00:18:00] object. The larger the drop error and the worse the memory. And this improves across the, uh, paradigm. At the same time, what you can record is how the participants are moving in the arena and in particular in which direction there
Benjamin James Kuper-Smith: moving.
Nikolai Axmacher: And then with a very elegant analysis, you can test whether there's higher activity in the higher bolt responses, uh, in the trinal cortex during movements in one of six different preferred directions, which refer to the main orientations of the grid cells as compared to the directions in between. So basically what you do is identify a hexa directional modulation of the bolt response in the internal cortex as a function of movement direction.
So it's a bit of a complicated analysis, but it has been replicated many times now and, uh, seems to be a quite a robust finding.
Benjamin James Kuper-Smith: Yeah, I, I haven't mentioned it yet, but I always put the references for the papers we discuss in the description of the episode. [00:19:00] So, uh, if you wanna know more about that, I guess the best place is probably just to read the dolo paper from 2010.
Nikolai Axmacher: Absolutely. Yeah. So, um, so then our main hypothesis was that, uh, I mean, first of all, we could replicate the, uh, the, uh, the finding from CELA because the, our 2015 paper was the first to like, try again, um, their analysis algorithm. And then second, uh, we are hypothesized that the grids are like representations were reduced in the internal cortex.
So what we did was to like record our participants. And then after maybe 20, 30 participants, we said, well, we are still not finished with the recordings, but we would like to know. How it's going and whether in the recording so far, we see any grid cell-like representations. And so we applied the analysis that Christiana did and found that there were no grid cell-like representation in enter aquatic in the first 30 participants, which were like a random mixture of genetic risk carriers and non-risk areas.
And we said, oh, maybe it's just too little. So we kept on recordings, [00:20:00] kept on recordings, and of course we had a, like a target number. We wanted to, to, uh, scan 40 risk carriers and 40 non-risk carriers. But I think it's good to do some quality controls, uh, also along the way. So, uh, we recorded more participants and when we were at about 50 or 60, we again ran the analysis and again, didn't, didn't find any significant texta directional modulation and trial aquatics.
Benjamin James Kuper-Smith: Briefly, if, if I remember correctly, the analysis is also. You look for each participant, whether there is this encoding, right? So it's not like you have like one overall analysis that you have for each participant. You didn't find it basically. Right.
Nikolai Axmacher: Exactly. That's
Benjamin James Kuper-Smith: what you're saying. Okay.
Nikolai Axmacher: Exactly. So we thought, oh, so, uh, maybe it, uh, we, we will just find that the, that the wonderful nature paper is not replicable.
So this will be a, like an intense discussion with Christian when we need to tell him that we just cannot, cannot replicate his results. Maybe Chris, we need to
Benjamin James Kuper-Smith: have a word,
Nikolai Axmacher: but maybe if we want to be very optimistic, it could just be that there's, [00:21:00] there are grid like representations in the non-risk carriers and just no bridge track representations in the risk areas.
And therefore when we lump them all together, we cannot find the. The, the pattern, right? Because we hadn't looked at the, the, the d the different genotypes yet. So we just completed the study and, um, after we had, uh, 80 participants and needed to exclude a few, but had more than 35 in each group, we, um, separated the risk carriers and the non-risk carriers.
And then indeed found a bit to our surprise that our hypothesis was absolutely confirmed so that the, the non-risk, the non-risk carriers showed robust and clear grid-like representation. So we could re replicate Christians previous findings. Whereas the risk carriers did not show grid cell like representations.
So, so they were massively reduced to the degree that they could not be recorded, that they could not be found at this network. Uh, level measure of [00:22:00] grid cell like representations. So this was the, the first, um, and maybe the main finding in the paper that in the participants who were, uh, had an average age of 22 years, so were really young university students with no apparent neuropsychological deficits, but if they had a, a risk factor for Alzheimer's disease, so a higher risk of developing Alzheimer's dementia in their sixties, seventies, or eighties, already at an age of 22, they showed this massive reduction of the functionality of grid cells at this network level.
So that was, that was a bit astonishing. And, um, also, um, disturbing maybe that in, in such young participants, you al already find clear evidence for, um, reduction of the functional grid-like representations.
Benjamin James Kuper-Smith: Yeah. Um, I'd like to ask about the, uh, or like get to the other find is in a second before I have one question, which is something that just occurred to me is that.[00:23:00]
Did you ever consider not doing the study because it was a bit too risky? I mean, it kind of seems like the nature paper had never been replicated, as you said. And now you're doing this, I'm assuming you have to sequence the, the, the genome of the participants. So, and then you have to scan them in a long fmri.
So it sounds like a lot of effort for something that, you know, is just this, and I think in Ella's paper also, it's one experiment, right? I mean, one with humans, and then they have like a other side thing with rats. But the, so basically like one study found this and now you're putting up a pretty large study.
Yeah, I was just curious like it's. Yeah,
Nikolai Axmacher: yeah, it was, it was definitely a high risk, high gain project. So I mean, it could have been that either we wouldn't be able to replicate the, the initial finding, or it could have been that there would not be any difference between the risk carriers and the non-risk carriers.
It was a risk risky study. But yeah, we were lucky to, uh, to, to, to find some, some interesting results. [00:24:00]
Benjamin James Kuper-Smith: Okay. Okay. So, as you said, big difference in grids. I like representation, but obviously, you know, as we. As we mentioned earlier, people don't develop symptoms until quite a bit longer. So, um, maybe you can link that to the, to the next finding here.
Nikolai Axmacher: Yes. So the next finding was, uh, actually a lack of a finding. Um, so when we looked at the performance or the, the, uh, the drop error and how the change with learning, there was absolutely no evidence for any difference between the risk areas and the non-risk areas. So when we plot the curves of the, um, of the drop error as a function of trials, it drops dramatically at the beginning.
It's very high at the end. It's very low, but it exactly overlaps for the risk areas and the non-risk areas. And we also look for a number of other, uh, markers, like how, how rapidly the participants move, the number of trials that they could do, um, and so on and. No evidence for any, for any difference. So the, um, the most parsimonious [00:25:00] explanation of that is, okay, you can use this complex analysis to identify grid-like representations in the, in the ano cortex of humans, but they have no functional relevance whatsoever.
It's just some, some weird, uh, epiphenomenon or surrogate marker. So next we tried a little bit harder and, uh, conducted a multiple linear regression analysis in which we took multiple different factors into account that may explain together conjoin the inter-individual variance and performance. So in addition to, to the grid, like representations, we considered, um, the, the genotype.
So, uh, whether a participant was an a e four or non a E four carrier, whether the participant would be male or female, whether the participant would be a little bit older or younger. Then we did a few additional control analysis in which we've, uh, included some other factors, but these were the main predictors.
Benjamin James Kuper-Smith: Oh, so predictors of what, what's the Oh,
Nikolai Axmacher: predictors of individual performance. Okay. Performance of these sum drop error across all [00:26:00] trials.
Benjamin James Kuper-Smith: Yep.
Nikolai Axmacher: And in this multiple linear regression analysis, we indeed found, so first of all, the, the, um, genotype by itself did not predict performance. The male participants were substantially better than the female participants.
We hadn't explicitly looked at that, but this was, uh, also finding that was consistent with some previous studies, like showing better spatial navigation performance in a number of, uh, different, um, it also came
Benjamin James Kuper-Smith: out in the Seaquest study. Right? I think I saw something
Nikolai Axmacher: they came in. Yes. It's a, it's a, it's a, like a recurring finding.
We also found that in a number of follow up studies, so it
Benjamin James Kuper-Smith: Yeah,
Nikolai Axmacher: just seems to be that the male participants are doing better than females in many, uh, navigation, uh, experiments. The third finding was that the, those participants who were a little bit younger were doing better than those who were a little bit older.
And interestingly then, the, uh, grid-like representations in a given participants did predict significantly the performance. So those participants were better grid-like representations, more pronounced grid-like representations also had a better performance. [00:27:00] So then we thought, wait a minute, this doesn't really fit together.
So on the one hand we have the genotype effect. So the, the risk carriers have reduced grid-like representations, the reduced grid-like representations, predict performance, but the, the risk areas do not perform worse. How is that possible? Then we thought, wow, maybe it is because the risk carriers use some compensatory other brain system to do the task.
And an obvious candidate is to look at the hippocampus, right? Because, uh, of the, uh, play cells and the known, well-known role of the hippocampus for, for patient navigation. So we, we just took an overall sum measure of hippocampal activity during the task, task related hippocampal activity and tested whether this differed, depending on whether a participant had more pronounced grid-like representations or reduced, like reduced grid-like representations.
And we found indeed that, that there was a negative correlation. Participants who had more pronounced grid-like [00:28:00] representations had relatively lower hippocampal activity as compared to those who had. Reduced gridlock representations. So this suggests that you can do the task with two alternative complementary mechanisms or spatial navigational systems, either a system that relies on grid cells in the internal cortex, or a system that relies on the hippocampus.
So I think this, this was another, uh, fascinating finding because it, it fits to some previous results suggesting that in, uh, very early Alzheimer's disease, there's a hyperactivity, a relative hyperactivity of the hippocampus, not only of the hippocampus, but also of other areas of the default mode network, but including the hippocampus.
And this may actually be, uh, functionally and pathologically relevant because there's evidence from rodents. Which shows that rodents who have a permanent hyperactivity of a default mode [00:29:00] network aggregate more amyloid as the main pathological marker of Alzheimer's disease in this area, which in turn then leads to a stronger progression of the disease.
So it may be that when you at young age have a reduced grid system, this leads to a compensatory hyperactivity in the hippocampus, which helps you to maintain performance in the short term, but could actually be detrimental and causally contributing, at least to the further development of Alzheimer's disease in humans.
This is still still more speculative and it's definitely speculative in these very young participants, but it would, would be consistent with a number of previous, uh, findings. And then the, the final piece of, uh, the final fi finding from this, from the study was, um, that we looked further whether there was any evidence that the recruitment or reliance of either the enteral or the hippocampal navigational system would show up [00:30:00]in a slightly different navigational strategy.
And what we did was to investigate whether the participants moved either in the center of the arena or at the border of the arena, so you can reach each of the objects via different trajectories, of course. Right. And this is independent of the drop error. So the drop error can be very high or can be minimal.
Regardless of your path, um, to this, um, to this object.
Benjamin James Kuper-Smith: The, the arena, there's no like objects in the way or something, right? You can just, the how you get, that's pretty free to the participant, right?
Nikolai Axmacher: Exactly. Yeah. Maybe I should have described how the arena looks like. So this is basically a, a circular area, which with some grass on the floor, which is surrounded by a wall, and in the distance you have some mountains that you can use for, um, directional information.
Um, but at any given point in time, the arena is empty. So you only can, can move around completely, freely and then place the objects where you need, where you think they belong. So you [00:31:00] could either move more along the, along the boundary or straight through the center. And, uh, what we found was that the risk carriers were more likely to move along the border of the trajectory than the non-risk carriers.
So even though they both groups did the task equally well, there was some, some obvious and statistically very significant. Difference in the, in the movement pattern in the arena. And this was, this was something that we've just found out through exploration. But interestingly, this, this finding now has been, has been replicated in a number of other studies also from, from other groups.
Um, so for example, in the C Quest, they also found that a POE four carriers, so genetic risk carriers moved more towards the boundary, which was, which in the, in the C request, uh, consists of the, like the, the, the land part of this, uh, paradigm than, uh, than the non-risk carriers. So it [00:32:00] seems to be, uh, like a common phenotype of APO E four carriers, that they tend to move closer to boundaries.
When we conducted the study, we. We didn't really know why that was. So, uh, we thought, well, if we would've done a rodent experiment, maybe it would've indicated that, um, that the risk carriers show higher anxiety, because rodents obviously don't like to move in open marinas. But, but for humans, it's, it's not that obvious.
I mean, there's, there's some indications. So, for example, if you go to restaurant, typically most people prefer to, to sit at the, uh, at the boundaries or at a, uh, it's middle of ward and then like to expose themselves in the center. But it's, it's obviously not as, not as proud. So we, we just speculated that this may be a reason, but we were not really convinced that it is.
And then a few years later, there was a, there was a paper coming out from Lisa Giacomo's group at Stanford where she found that risk car, um, not risk areas that grid cells actually need. Encounters with a boundary to stabilize their firing [00:33:00] patterns. So when you look at the stability and the accuracy of the firing of grid cells, depending on whether rodents have been far away from any, any boundary, uh, or probably also other spatial cues, the grids are firing, becomes less stable.
So it's probably due to some error accumulation. There's some, some inaccuracies that develop, whereas like shortly after encounter with the boundary, the grids are firing is very precise. So obviously this, this didn't really work in the risk areas, right, because they, they moved close to the boundary, but, but still their, their grid cell pattern was, was quite inaccurate.
So we, we, our, our current interpretation is that this is actually an attempt to try to stabilize the, the grid pattern, but it's, it's not, it's, it's not, it's not eff effective in the, in the risk areas.
Benjamin James Kuper-Smith: One thing that I think it actually kind of fits with some of the rodent literature on the development of spatial navigation because if I, I correctly there when, [00:34:00] you know, these are studies where they do these kind of study with rats when they're just a few days, or like three weeks or four weeks old, and I think there you start off with these boundary cells and it takes like basically like a day.
14 after birth or something. They have like boundary cells or something like that, and maybe some play cells, but they don't have grid cells yet. That kind of takes more, takes longer to do that. So it seems maybe that, yeah, I mean it's the same thing if, I guess if you don't have this like generalized metric through the grid cells, you kind of have to use whatever's left in this case.
Boundaries and,
Nikolai Axmacher: exactly. Yeah. So, so that's, um, so, so I think on the one hand, one could, one could think that the encounters with the boundary helped, helped, helped to stabilize the, the grid cells be just because you get more visual information, but could also indicate that you fall back onto maybe a on onto genetically older navigation system that is more supported by the hippocampus that is also affected later in Alzheimer's disease.
So that putatively is still, still functional in the, uh, in the young risk carriers, [00:35:00] but that shows up in this specific navigation path so that the participants are orienting themselves by moving close to the boundaries.
Benjamin James Kuper-Smith: Yeah. Yeah. I guess there's are like, yeah, two different interpretations of the same thing.
Right. Either that they're trying to stabilize the grid cells or they're just, they can't use them or use. Right. Exactly. It's not like a on off. Um, one question I think is kind of my final question about this paper, which is, were the participants. The risk carriers, uh, aware that they were using the boundaries more or that they were moving closer to boundaries.
Is this something that people were aware of?
Nikolai Axmacher: Not as far as we know.
Benjamin James Kuper-Smith: I guess you didn't even know who they were during the testing here.
Nikolai Axmacher: Yes, and we also didn't ask them, uh, explicitly. I mean, this is something that we are currently exploring in, in, in follow-up studies where we are looking at boundary effects.
So the propensity of a given person to move close at a boundary, and also there we, we, we are using like questionnaires and, and ask the participants [00:36:00] about their, the subjective navigation strategies and so on. But in that study, it was basically an exploratory analysis that we did after the data collection had been, had been completed.
Just to explain this apparent, just to explain out of this finding that. That the participants seem to rely either on one or the other navigation system, so either the grid cell system or the hippocampus, um, and test whether this was somehow reflected in navigation strategies. But it was very exploratory.
And my, my guess would be that the participants wouldn't notice. But then, I mean, on the other hand, how should they, because they were either always risk areas or non-risk areas, right? So they were not like suddenly in a condition where they moved close to a boundary, so they would need to compare themselves to others and, uh,
Benjamin James Kuper-Smith: or I, or I guess to their previous selves.
Right? Supposed, I guess the idea is that the, the trinal cortex deteriorates throughout their life, right? Um,
Nikolai Axmacher: yes. But the question
Benjamin James Kuper-Smith: is really, or is it that they're born with
Nikolai Axmacher: [00:37:00] that, that, that, that's, that's really good question. I mean, the, the question is when this, uh, pathology start, and there's, um, there's some interesting literature, um, also from, there's some literature from owns, but in particular there's liter literature from neuropathology where the brains of.
Children or adolescents who died from, not from Alzheimer's disease, but from something were dissected. And, um, and then the anti cortic was studied that this is mainly worked by Umra who also also did the famous BRAC staging for Alzheimer's disease, like a famous neuropathologist. And he found that, that in some participants you could even see tau neurodegeneration, uh, which is neurodegeneration that among other diseases you also find in Alzheimer's disease can, can be observed in even in children.
Below 10 years, so, so very young children. And they interestingly also found that the likelihood to see these neuropathological findings is higher in a PE four carriers. So it may [00:38:00] be that, that even before the age of, of 22, we would've found some, um, some effects.
Benjamin James Kuper-Smith: Okay. So what's the paper with the, the risk carriers having even in childhood?
Nikolai Axmacher: Yeah, I believe that is cited in the, in the science paper. Okay. Maybe Giadin and others. So it's, it's all work from, from, so if you, if you cannot find it, just let me know then, then I search it afterwards.
Benjamin James Kuper-Smith: Okay. Yeah. Great. Um, yeah, those are kind of the, the, the questions. I think we've kind of covered the science article pretty already.
Um, so maybe we can just, uh, yeah. Maybe move on to the science advances paper and then kind of have a broader quest, uh, like discussion after that. I think maybe one way to summarize the science advances paper. In like two sentences or whatever is something like in your science paper, you found that people have reduced grid soil like activity, but they can compensate with other mechanisms.
And in the science advanced paper, the question is what do they do [00:39:00] if they can't rely on these mechanisms?
Nikolai Axmacher: Exactly. So, um, the, the, the main, the main idea behind the, uh, science advanced paper was to, um, to develop a paradigm with different conditions. So one condition where, uh, where you conduct a specific patient navigation task that you can only solve, relying on the anal cortex.
So probing the anal cortex with a, with a relatively specific task. So not the spa learning task that we did before that that is known to rely on a variety of different navigational systems. And then on the other hand, use of a control condition, or in that case two control conditions where subtle dysfunction of the enteral cortic can.
Products can putatively compensated for by recruiting other systems. So, so basically the, the, the main, the main idea is that when you navigate in the real world, typically you use a multitude of different mechanisms. So one mechanism is [00:40:00] orienting at allocentric cues. So for example, I mean, space navigation here at war University is a very challenging thing.
I'm not sure whether you've ever been here. This is a huge campus. It's very confusing. And the, the simplest way to, to find your, uh, your path is to, to use a landmark, for example, the big ax, which you can see from many places. And then you just go there. So this is a relatively simple navigation strategy.
You don't. You probably, I would, I would argue that probably you don't need grid cells for that. You just need to keep this spatial queue in, in mind or in, in your, in your site ideally, and go to it. Another strategy then, um, and that is very different, is landmark based navigation. So landmark based navigation is a strategy where you, uh, remember specific salient points on a particular trajectory and just remember whether you need to go left or right when you see this point.
Right. So this is a, an egocentric navigation strategy, obviously. So when you come from [00:41:00] a different direction, you shouldn't go in the same direction.
Benjamin James Kuper-Smith: Yeah. Like the idea being I go to the petrol station and then turn left. But if you're coming from the other direction Yeah.
Nikolai Axmacher: Then you should turn right.
That's exactly, yeah, exactly. Yeah. So this, this is a, uh, a so-called landmark based navigation strategy. Then there's a third system, which is called path integration. And path integration means that you basically estimate the distance that you travel and the direction in which you're traveling, and, um, then use this information to find a path.
So this, this, uh, strategy is, um, can be used. So the advantage of this strategy is that it can be used in the. Absent of most other sensory cues. So for example, my, my main example is always when you, when you're in a hotel room and at night you need to, you want to go to the bathroom and it's completely dark so you don't, don't see anything.
Then you try to remember in which direction to go and, you know, okay, it's about like four meters to go to the, to the door. [00:42:00] So you, you are more or less able to find your way unless there's something standing in your way. Right? So, and this, this is a, a strategy where basically you, you estimate the distance, uh, to the door and the direction and this, and you rely on path integration.
If you can see the door, you don't need that.
Benjamin James Kuper-Smith: Or maybe one example, I dunno whether this is true, uh, but it seems like it might be, um, if you are, uh, in prehistoric or before GPS and that kind of stuff, if you were, uh, at c then you'd probably calculate your speed, uh, and with a compact your direction, and then you kind of calculate where you are Exactly.
Of idea, right? Yeah, yeah.
Nikolai Axmacher: So,
Benjamin James Kuper-Smith: you know, not that conscious.
Nikolai Axmacher: Yeah. Yeah. So that's, that's like for, for a little bit larger distances than hotel rooms. Uh, and, and this is, this is another very good example for, for path integration. Uh, another example is when you are, um, when you're doing a hike in an unknown environment and uh, like go for a while and then think, well now it's, now it's time to go [00:43:00] home.
So basically what you need to do is like, remember how far you went into, into which direction and then, then you try to find back your. To our initial location.
Benjamin James Kuper-Smith: Yeah, that's what I tried to do last weekend. 'cause I moved to like a new place and there's kind of, not, mountains is a big word, it's more like a large hill, but, uh, through the woods.
And I thought, you know, classic cognitive map, I thought I knew shortcut. I, I asked someone whether I was, whether I was going the right direction, I was going the complete opposite direction of where I should have got. Apparently that didn't work In that case,
Nikolai Axmacher: I know the situation and I think it's actually, it's actually relevant because path integration is a pretty poor navigation strategy because, uh, when you just estimate a little bit.
Uh, incorrectly, like short distances, then it all adds up and accumulates, uh, for longer distances. This is called error accumulation. It is of like an intrinsic property of path integration. And this is, this is also one reason why whenever possible people rely and also rodents rely on additional visual cues and use [00:44:00] that to, to correct the path integration error, similar to the correction of the accuracy of grid cell firing when you encounter walls, right?
Yeah. So it's a, I think it's the same mechanism. So, um, uh, so we designed a study in which we, again invited a PE four carriers and non non-carriers. Mainly to do a behavioral study. Then there was also some, some smaller FRI study, but the main part of that paper was a behavioral study where we had our typical path integration paradigm.
So the participants, again, navigated through virtual reality, but now they didn't have to remember the locations of objects in the arena. They just started from one location, from an empty basket in that case, and then went, uh, to a series of trees to find one tree that contained an apple. That's why we called it the Apple game.
It's actually Lucas ksu, uh, is the first author of that study and designed the paradigm who came up with the name. So he should be credited for that. And, uh, after finding the apple, what should you do with that? Bring it back to the basket. [00:45:00] So basically, um, it's an, I would, I would guess like, um, reasonably realistic naturalistic paradigm where then with the apple, the participant should find the location of the initial basket.
And again, you have a drop error, which is the inaccuracy in a given trial between the, so the distance between the location where the people believe the basket is, and the actual location of the basket. So now this is, um, done in. Three different environments. In one environment, this is the most difficult.
The participant only see a grassy plane, so they can only look at the visual flow during movement. Along this arena or space, there's no really no boundary and no no district cues to try to, uh, orient themselves and estimate the distances. And then there's two other conditions, either a condition with a, with a wall that provides some, some additional information or a third condition with a local landmark, so, which you can also use for orientation.
[00:46:00] And what we found was that the a PE four carriers were specifically impaired and only impaired in the task where they had only to rely on the visual cues. Whereas in the two other conditions with the landmark and with the boundary, they were not impaired as compared to the to the controls. So this suggests indeed that when you are in a situation when you can only rely on on par integration, so only on the grid cell system, then the A PE 4K are worse.
But as soon as they can use some compensatory compensatory system, they're at par with the, with the controls. This was a, this was a, a relatively large scale study, so this was, uh, done in the so-called Ethel Game Consortium. So it's a number of, uh, of groups in, um, in Spain, Italy, Germany and Belgium, who each recruited participants, genotype them, and then had them done the, the experiments and was also not only in young students, but also in, in, uh, like towards like, I think [00:47:00] starting from 20 to 75 years, so across the entire age range.
Yeah. Showing that, that the a PE four has a very specific detrimental effect on, on path integration.
Benjamin James Kuper-Smith: Yeah. Yeah, it's really cool. I mean, in, in a, it seems to me in a way it's like a, it's a specification of, uh, of uh, something you believed or not believed have found, but like something of an interpretation of the, um, science paper, right.
Nikolai Axmacher: Exactly. Yes. Yeah.
Benjamin James Kuper-Smith: Does this, I'm wondering does, so we just mentioned earlier that the, the compensation in the science paper through move through, moving closer to boundaries, uh, as we just mentioned, there's at least two possible interpretations. One is they use this to, uh, anchor the grid cell system to help it, and the other is just, they just use a different system altogether.
Does this paper say anything about that?
Nikolai Axmacher: Yeah, so, uh, the, the, the, the finding is a little bit different, but what we found in the landmark condition in particular is. That the risk [00:48:00] carriers move closer to the landmark when they brought the apple back from the tree with apples to the, to the basket. So this suggests that they, again, stick more to a salient visual queue in order to do the path integration task.
So in this per, in this, in this condition, they are not worse than the, than the non-risk carriers, but it again suggests that they, they tend to, uh, rely more on, on additional visual cues.
Benjamin James Kuper-Smith: Uh, so oh, I dunno whether you have anything, uh, like more to say about this study. I kind of,
Nikolai Axmacher: no, I think, I think that's a main finding.
I mean, I, I also think it's interesting because it really shows that like, that there's some very specific cognitive paradigm in, in which you can see. Dysfunction of APO E four carriers, which we didn't find in the previous study. So it, it may pave the way to, to very sensitive neuropsychological tests to find early dysfunction.
Benjamin James Kuper-Smith: Yeah, I mean, I, exactly. So I [00:49:00] mean, I want to slightly broaden out now kind of the discussion. Uh, maybe we can do like the application to clinical research in a second, uh, before that. So one thing I'm really interested in, in this space is the idea of these abstract cognitive maps and that the. Entire system is also used for that.
I mean, I talked quite a bit with Yako about this, um, Yako Mont, and, uh, I, I dunno whether I am per se gonna do research on this, but this is something I'm very interested in and I'm just curious. So, I mean, maybe to summarize that in like a few seconds is the, the idea is that, okay, so this place, cell grid cell, this whole thing we've been talking about, all of that has been initially found or was initially found in special navigation context, but well roughly since your science paper, roughly since that time.
Um, there've been quite a few papers that came out that showed that there seems to be some evidence that these, these grid cells are also involved in processing other dimensions. So we can, you know, if we think of [00:50:00] space as having two spatial dimensions, if you're moving in a plane, then there are studies where there's, so that these two dimensions could just be anything else that.
Conco papers with neck length and leg length, which is really hard to say. Uh, I agree. Um, uh, of, of, of birds in this task. But basically the idea is that maybe this is a much, much more general, uh, phenomena where we just happen to find it in space. So I'm curious just if we apply that kind of idea, or maybe let's just assume for a second that it is the case to your studies.
I'm wondering, so one thing I also asked Yako, and we didn't really have a good answer necessarily for this, is for like, for example, what would a landmark be in this kind of abstract cognitive space? I'm, or maybe I'm, I'm asking maybe one question before I should ask another one. Um. I'm not sure really what the question is.
Um, like I, I guess you are, are you planning to do anything like that in that direction or, um, or you're nodding Yeah, I guess maybe. [00:51:00]
Nikolai Axmacher: Yes. So, uh, I find the, I found the, the constant study very intriguing, and I think it, it filled in a way, like a gap in the theory about the medial temporal lobes or internal cortex and hippocampus for cognitive functions in general, because, so what you, what you see is that in.
In rodents, there's play cells in the hippocampus and good cells and enter aquatics in humans. There's some mixed evidence one has to say about the existence of play cells actually. So, uh, like using microelectrode recordings in epilepsy patients, some studies found play cells, other studies did not find play cells.
There's some empirical reasons why it's. Maybe unlikely to find anything like a place cell, like representation in fmri because adjacent place cells are, do not show any topography. So if you average across a larger voxel, probably you cannot decode any, any specific, uh, locations, although some other [00:52:00] studies suggest that actually it may be possible.
So it's a, it's an ongoing discussion, but there is good evidence now from human single cell recordings that there's the so-called concept cells in the hippocampus. So, which are like the, the famous Jennifer Anderson cell or, uh, flu Skywalker cells or whatever, which code in a perceptually to, to perceptually inva concepts like, like particular people.
So then this suggests that, uh, you have cells and hippocampus, which represent a point. In either a physical space or conceptual space, whereas you have cells in the enter cortex that provide a distance metric, so that code various different locations in physical space. So does this also occur in concepts, uh, space?
And this is, uh, I think exactly what the consent inco paper suggests that actually, uh, you have these two different types of, uh, representations, either, uh, like a metric, uh, representation of distances of both physical distances or [00:53:00] conceptual distances in your internal cortex. And a very specific, um, representation of individual points in either conceptual space or physical space in the hippocampus.
And this also fits the idea to. Like the, the, the hippocampus provides a kind of an of an index or pointer to representations in the neocortex, which can be very flexibly applied depending on task demands. So what drones often do in experiments is running around and also the real world. So, um, it's obvious, or it's quite, may, maybe no surprise that actually what you find in the hippocampus is representations of specific locations because they're running around.
What humans often do is. Reasoning, making comparisons between concepts. So, um, it may be no, no surprise that actually what you find is specific representations of individual concepts. So if the epilepsy patients would run around all the time, and if you would, would be able to record during this period, maybe it [00:54:00]would be easier to, to find also play cells, but the same cell in the hippocampus, um, of an, of an epilepsy patient, which represents a concept.
May in other tasks, in other tasks, situations be used, be recruited to surface an index for specific locations.
Benjamin James Kuper-Smith: I mean, is this an. I guess, is this the secret work you're working on? I can't really say too much about, or is there something concrete planned with Yeah, something like this?
Nikolai Axmacher: Yeah. I mean, this, this is very much ongoing, but, um, I mean the, the, the, the main idea is to investigate this, uh, navigation in conceptual space.
What we are trying is to go a little bit beyond the artificial birds from the Constantine school study, which are, which are wonderful because, because they're so well controlled. But, um, on the other end, of course they're artificial. So, um,
Benjamin James Kuper-Smith: also it's physical still, right? It's still about length distances, which I find is, in some sense it's not as abstract as those, as they say it is.
Nikolai Axmacher: That's right. Yeah. So, so, and, and, [00:55:00] and I think that in any conceptual space where you need to make comparisons between pairs of objects, I, I would expect to find, we haven't really shown that yet, and we don't have any even preliminary results, uh, supporting that. But I would expect and hypothesize to find similar representations of distances between, between objects, between phases, between.
Whatever stimuli you can think of. And then, uh, like one hypothesis, uh, would be that, um, actually path integration and navigation in this conceptual environment would correspond to, um, to comparisons. So for example, when you see a person and, uh, a person's face and compare it similarity across the high dimensional face space, which consists of no links, uh, whether someone has a bear and so on, like, just think of all the features that make make up faces and you compare this, this person's face to another phase, or you[00:56:00] compare two different objects or maybe you even compare, um, in social space.
There's a number of papers about that now, like two people in terms of their. Approachability or, or other, other social dimensions. So whether you like them, whether you don't like them, and so on. So, uh, so I guess this, this navigation and the distance estimation in, uh, in this conceptual space, just, just probably, uh, it's my hypothesis would reflect similarity judgments in any space.
Benjamin James Kuper-Smith: Mm-hmm. Yeah, it's really interesting. I mean, I'm particularly interested in this kind of social aspects because I do kind of social interactions and that kind of stuff. Not, there's nothing like, basically I have like one FMI study left in my PhD and we're not sure whether we're gonna do grid cells with that or not with this social stuff.
Nikolai Axmacher: You should,
Benjamin James Kuper-Smith: yeah, that's, uh, yeah, the problem is kind of that for the stuff I have studied to make it like fit also into the stuff I've done before, no one has done this with. So it, it would, um, the [00:57:00] or rather kind of study one would be to establish whether this is a thing in this way, I'm doing it at all.
And step two would be what I'm actually interested in doing. But that's kind of what I always figure out figuring out till the, in the next few months. Um, basically the, the thing that I also discussed with Yako was this question of how far can you translate the, the physical findings into these abstract conceptual stuff, or to what extent something is specific to space.
But, uh, you'd, you'd think that there is something like. Landmark based navigation in the sense of, in this case, as you mentioned, comparing it to something you already know or,
Nikolai Axmacher: yeah. I mean, I, I agree. It's still, it's still a bit of an open question, what a landmark in conceptual spaces, what boundaries, con, conceptual space are, and so on.
Benjamin James Kuper-Smith: Yeah. Yeah.
Nikolai Axmacher: So I, I guess one possibility for a landmark would be, um, like something like a prototype, right? Like a prototypical representation of let's say a, a particular object, right? So when you, when you [00:58:00] close your eyes and think of a, let's say, let's say a beer, it's getting late. And, uh, then, then something, something comes into your mind.
So this is a prototype of a beer, and then when you're comparing two other beers, you are, you may compare them with regard to your prototype. Right. Yeah. So, um, so when you see some bottle and are asked, uh, to judge how, how typical, for example it is, uh, of a beer bottle, then you compare it to your prototype of a beer.
So this, this may be something like a, like a landmark based navigation and boundaries, I, I guess, would, would, may reflect the boundaries of, uh, like between concepts, which are maybe not, also, not, not very easy to define, but there's, there's probably some, some boundaries where some, some objects where you wonder whether this is a glass or a cup, for example, right.
So, um, there's a prototypical glass, there's a prototypical cup, and there's like some things which are somewhere in between at the boundaries between them.
Benjamin James Kuper-Smith: Like a cup made out of glass or something.
Nikolai Axmacher: [00:59:00] Yeah, exactly.
Benjamin James Kuper-Smith: Yeah. Yeah, I mean, yeah, I guess we're not, we're not gonna get a definitive answer today, but it's, no, I'm sorry.
It's something I find really interesting. Yeah. Um, but okay. I guess we've, we've got maybe like 10 minutes left, so shall we maybe move towards the kind of clinical applications? Um, whilst I was, uh, well, at the same time that I was inviting you, um, onto the podcast, I was, uh, by pure coincidence reading Brave New World Revisited by ti and there was a quote in there that happened to fit, I think perfectly, I guess I'll just read the quote.
The quote is, pure science does not remain pure indefinitely. Soon or later, it is apt to turn into applied science and finally into technology. And it seems to me, I mean, you know, that's, this is also kind of what I found so exciting about your paper that, so we had the basic science with John O'Keefe, just, you know, putting a more or less randomly electrode into hippocampus and letting rats run around.
And then 40 years later, we have the beginning of clinical [01:00:00] applications, right? So. I guess maybe as a very broad question, so how far are we away from this being in any way clinically usable?
Nikolai Axmacher: So, very good question. Of course. So, um,
Benjamin James Kuper-Smith: predicting the future is of course, always easy.
Nikolai Axmacher: So, um, so one caveat of course is that it's, I think, quite unlikely that you can like, or what you cannot do is just take the, the paradigm, the the spatial navigation paradigm, also the Apple game and test 60 or 70 years old in a scanner in order to see whether they have some alterations of the grid pattern.
For practical reasons, this is not really feasible. And, uh, what is still an open question is, uh, I mean, what we've, what we've seen is that, that there are differences in grids, cell like representations and also navigational patterns and typicall activation between risk carriers, apo e risk areas and [01:01:00] non-risk areas.
But then on the other hand, like the, the easier, easier approach to find the difference between risk areas and non-risk caries is to genotype, then not, not to an FM I study. So in other words, it's currently not clear to which degree these, um, these findings of reduced grid select representations or auto representational strategies provide any additional.
Clinical clinically useful information to, uh, predict, um, Alzheimer's development or Alzheimer's pathology in a, in a, um, in a given participant. So what I could imagine, um, what is, what is much more feasible, of course than FM RI study in participants at risk for Alzheimer's disease or with like very early stages of Alzheimer's disease, is to, to do a, a cognitive paradigm and, um, to try to develop, um, a cognitive paradigm, maybe something like the c quest to find subtle alterations, um, in, in early disease stages.
Then of course, many validations, [01:02:00] um, steps are necessary. So one, of course is a longitudinal study. So test whether in a 60-year-old with subtle dysfunction or even only a change in navigation strategies in these paradigms, whether this participant is more likely to develop Alzheimer's disease, um, than another person who doesn't show this, these deficits or these, these alterations and strategies and whether this, these early changes predict the development of Alzheimer's disease.
More than established clinical markers, including aqui four. So this is just an empirical study, uh, an empirical question. Obviously, longitudinal studies take a time and are like, uh, time consuming and, and expensive to conduct. I would love to do a similar study in the, in the future. I think it would be a necessary next step.
I know that there's ongoing studies also with the c request paradigm on, on this question. So it's, um, I think in the end it's also [01:03:00] an open question, which paradigm is best suited, is most sensitive, most robust, but, uh, I think there will be development on that. Another, uh, question of course then is whether the changes in navigational strategy, uh, are actually reflecting Alzheimer's pathology in order to, uh, to address that one would need to, uh, relate them to other biomarkers of Alzheimer's disease.
So, CSF, uh, would be one obvious, uh, biomarker, which you cannot really, uh, obtain from healthy participants. Another one is blood markers of amyloid, uh, which are, have been established very recently. And then of course, what is very exciting is the new, uh, possibility to do amyloid and tau pet imaging, right?
Where you do not only, um, measure the, uh, overall amount of amyloid or, or tau in the brain, but even their regional distribution. So you can distinguish between, let's say hippocampal or extra [01:04:00] hippocampal, maybe even trinal tau, uh, pathology. Of course the, the special resolution of PET is somewhat limited, but, um, there's now the second generation of, of Tau PET markers, uh, out and, and there definitely will be further developments.
Benjamin James Kuper-Smith: What I find really exciting about this is kind of not the, the pet based, uh, scanning or whatever, because you know, who has that scanners, right? That's very, very few people can do that kind of scanning per year, very few patients. Whereas what's much more exciting, as you mentioned, I mean, for example, the C here, quest is a game.
I mean, for those who don't know, it's like a smart phone based game. You could just download it in the app store or whatever. And I mean, as you hear, there's of course some like ethical problems with just anyone who's, who uses boundaries a lot to just tell them, Hey, you're using boundaries a lot. You might wanna see a doctor.
That's probably not how it's gonna work, but I really like the idea of, yeah, I mean, you're at the doctors and they say, Hey, just play this smartphone game for like 10 minutes while you're waiting. And that maybe that [01:05:00] already tells you something, right?
Nikolai Axmacher: Yeah, exactly. So I, I think this, this is a, uh, a likely development and, uh, I mean, for me, what is so, uh, fascinating about this, this research direction is that you can draw a line between, on the one hand, very applicable methods like apps or behavioral paradigms on a laptop even, which are relatively easy to conduct in, in larger populations.
So in the Apple game study, we had more than 300 participants who conducted the paradigm on the one hand, and then the possibility to, to do the same, uh, experiment in the scanner and look at the bolt responses. Then what we also did, what we didn't talk about, uh, is the same run, the same paradigm in Intre lead G in epilepsy patients, so that you could look at the electrophysiological pattern, um, in, in enteral cortex and, and hippocampus.
Um, you could even look at the single cell pattern in, in some epilepsy patients who have micro wires implanted. And then you can see whether any changes in a, [01:06:00] in a given person are related to, um, to top pathology or, uh, amyloid pathology. So this, this obviously will not be done in a single study, so you will not have like an epilepsy patients who afterwards doesn't top it.
But, um, that's a
Benjamin James Kuper-Smith: big study. Yeah.
Nikolai Axmacher: But, but, but basically you can combine these different angles and these different pieces of evidence and then, uh, try to develop paradigms that are very sensitive and very specific and easy to run, which may allow you to increase the prediction, the predictability of Alzheimer's disease in a given person.
And at the same time, you can validate the same paradigms by looking at red cell-like representations in FRI, single cell activity and intracranial eg. Oscillations in epilepsy patients and amyloid and tpe and other biomarkers in, um. Various individuals.
Benjamin James Kuper-Smith: Mm-hmm. Um, final question, uh, so if I understand it correctly, you got a consolidated ground from the [01:07:00] ELC to work on this, right?
On this kind of research area. Can you say kind of what's, that was fairly recently, right? Um, so yeah, kind of what's, what's the plan there? What, yeah, what's the rough direction you want to go down, uh, with this research?
Nikolai Axmacher: Yeah, I mean, uh, in, in just one sentence, exactly what I just suggested. So, okay. So, uh, like trying to understand how behavioral alterations.
Are related across paradigms, how they relate to fm i grid cell, like representations to electrophysiological signatures of grid cell, like representations and grid cells in, in humans and to Alzheimer's pathology.
Benjamin James Kuper-Smith: And that's in physical space or also abstract cognitive space
Nikolai Axmacher: that's in, in different spaces, including conception space.
Okay.
Benjamin James Kuper-Smith: Okay, cool. Um, so I dunno whether there's anything else, any final words you wanna say otherwise I'll stop recording. I don't have like a formal ending. Um,
Nikolai Axmacher: yeah. Well, I, I don't know. I mean, a few years ago there [01:08:00] was a. So-called grid cell meeting in, uh, London. Um, I'm not sure whether you participated in that as well.
Uh, whether the different lines of direction of, of grid cell research came together. So in addition to what we talked about, of course, all the, the, uh, rodent uh, research, the really fascinating research in beds from Lansky's lab and also the modeling. And this was one of the most exciting meetings, um, that year.
And, and, and for, for multiple years, I think. So what you could really see is that there's a, a very strongly divergent, but at the same time. Very, um, interdisciplinary and methodologically diverse field, um, which is progressing, progressing very rapidly. And, uh, I think it's a, it's a very active and, and fascinating research direction.
And, uh, yeah, for anyone who too isn't working on grid cell grid salts yet, or grids arelike representations, I just recommend, uh, to go in that direction. [01:09:00]
Benjamin James Kuper-Smith: Okay.