not cheese; it's only brains
January 28, 2019 9:44 AM Subscribe
She doesn’t subject living brains to arrays of electrodes and scanners. She doesn’t divide brains into prosciutto-thin slices and carefully sandwich them between glass slides. She doesn’t seal brains in jars of formaldehyde for long-term storage. Instead, she demolishes them. Each organ she took such great care to protect on her trans-Atlantic journey was destined to be liquefied into a cloudy concoction she affectionately calls “brain soup” — the key to her groundbreaking technique for understanding what is arguably the most complex congregation of matter in the universe. In dismantling the brain, she has remade it.
I suppose we should be thankful birds never learned how to cook. They are rather dextrous, after all.
posted by constantinescharity at 10:33 AM on January 28, 2019
posted by constantinescharity at 10:33 AM on January 28, 2019
I read the cell-counting less as an indicator, and more of a "let's challenge commonly accepted knowledge."
It could be that cell counts or ratios mean nothing, but without proper counts, it's hard to be sure. That the ratio of glia to neurons was commonly accepted as 10x, but turned out to be closer to 1x alone is interesting.
posted by explosion at 11:20 AM on January 28, 2019 [2 favorites]
It could be that cell counts or ratios mean nothing, but without proper counts, it's hard to be sure. That the ratio of glia to neurons was commonly accepted as 10x, but turned out to be closer to 1x alone is interesting.
posted by explosion at 11:20 AM on January 28, 2019 [2 favorites]
I have to wonder how important the specifically cortical structures are outside of a mammalian context, that's all.
Probably there are ways to get to human-level intelligence without respecting the neuroanatomy. The models developed by DeepMind use neural networks, but the model schematics tend not to resemble anything found in prefrontal cortex at all (not saying alphago or alphastar are actually human-level intelligent, but it seems the closest that has been achieved so far). I imagine it's a 'many paths to Buddha' situation - there are different approaches to clever behavior that depend on evolutionary accidents and specific ecological niches.
Considering the importance placed on maximizing calories through cooking as a way of getting around metabolic constraints on neural capacity, I do worry a bit about an advanced pod of octopi hanging out near thermal vents and experimenting with kebabs.
posted by logicpunk at 11:23 AM on January 28, 2019
Probably there are ways to get to human-level intelligence without respecting the neuroanatomy. The models developed by DeepMind use neural networks, but the model schematics tend not to resemble anything found in prefrontal cortex at all (not saying alphago or alphastar are actually human-level intelligent, but it seems the closest that has been achieved so far). I imagine it's a 'many paths to Buddha' situation - there are different approaches to clever behavior that depend on evolutionary accidents and specific ecological niches.
Considering the importance placed on maximizing calories through cooking as a way of getting around metabolic constraints on neural capacity, I do worry a bit about an advanced pod of octopi hanging out near thermal vents and experimenting with kebabs.
posted by logicpunk at 11:23 AM on January 28, 2019
The amazing thing about octopi is that they can do so much with such a decentralized neuroanatomy! It's wild.
I'm not sure I buy the calorie-maximization-via-cooking hypothesis, though, particularly given the wide variety of ecological niches inhabited by other remarkably intelligent species--everything from obligate carnivory (octopi, whales) through to bulk cellulose-heavy grazing and browsing (elephants) to does-it-hold-still-long-enough-to-swallow omnivory (humans). We only have an n of 1 when it comes to known-human-level-capacity, but that doesn't mean that everything humans chose to do on that general path is necessary to getting to this position.
posted by sciatrix at 11:31 AM on January 28, 2019 [2 favorites]
I'm not sure I buy the calorie-maximization-via-cooking hypothesis, though, particularly given the wide variety of ecological niches inhabited by other remarkably intelligent species--everything from obligate carnivory (octopi, whales) through to bulk cellulose-heavy grazing and browsing (elephants) to does-it-hold-still-long-enough-to-swallow omnivory (humans). We only have an n of 1 when it comes to known-human-level-capacity, but that doesn't mean that everything humans chose to do on that general path is necessary to getting to this position.
posted by sciatrix at 11:31 AM on January 28, 2019 [2 favorites]
Sciatrix, do you have any other links that address bird neuroanatomy specifically? I’m interested in learning more about its uniqueness in comparison to mammals.
posted by constantinescharity at 11:48 AM on January 28, 2019
posted by constantinescharity at 11:48 AM on January 28, 2019
Great post and great profile of an important neuroscientist. And the closing lines were wonderfully put:
birds don't have a cortex as we understand it--that's a mammals thing. Birds do have an older analogous structure, but their brains are organized a little differently, and they do not rely on cortical surface area to get things done as mammals do. It's really interesting in that vein that bird brains are so much denser in terms of neurons.
I'd dispute the "older analogous structure" phrasing a bit, particularly the "older" part. I think you've given the generally held opinion in neuroscience, but most folks I know who work in bird brains don't agree with it. The avian pallium is indeed developmentally homologous to the mammalian pallium, which includes the cortex and parts of the amygdala. But I don't think the evidence supports the idea that it's phylogenetically older. Both the mammalian and the avian pallium are heavily derived from the ancestral state, which may have been similar to that in modern amphibians. Non-avian reptiles show a certain amount of derivative specialization in the pallium as well, but birds are really very different even from other reptiles. The pallium of snakes and lizards, for example, seems to be roughly laminar in structure (possibly more similar to the mammalian cortex, in fact!), while birds' are organized into more densely-packed nuclei. As far as I know, it's still an open debate whether the avian or mammalian forebrain more closely resembles the ancestral state, but probably both of them share a similar level of derived characteristics.
To the extent that cortex is special, my opinion is that it is so primarily because of the properties it derives from its position within neural circuitry. In mammals, there is what I would call a canonical forebrain circuit, such that every piece of cortex receives inputs from the thalamus, processes with a local microcircuit (what in vision would be called "columns" but that term seems to be loaded now so let's just say a microcircuit organized according to laminar connections within a local patch of cortex), and projects to the striatum and thalamus; striatum projects to the pallidum; and pallidum projects to thalamus and midbrain. These circuits seem to form a large set of interconnecting "loops" with heavy feedback and crosstalk. Now, what's interesting is the pallial amygdala of mammals also seems to preserve this structure. But if you look at the pallium of birds, the same canonical forebrain loop circuitry is also present! It's harder to see because of the way the tissue is structured, and because the striatum and pallidum actually intermingle during development to form a single tissue, but it's there. And in fact it looks like this canonical loop motif is conserved right across the entire vertebrate phylum. Even teleost fish seem to preserve it. (Side note, I learned somewhat recently that the teleost forebrain undergoes a weird "eversion" during development where it basically flips inside-out. Kind of blew my mind a bit. So even though their forebrain at first glance looks nothing like that of terrestrial vertebrates, it's actually got all of the parts wired up in the same way, just positioned inside-out. So weird. Also, let's ignore the proper placement of hagfish in the vertebrate family tree for now.)
Anyway, all this is to note that Herculano-Houzel's cell counting technique is essentially blind to all of this complexity. Which, as the article says, is both its strength and its weakness. On the one hand, gross changes in numbers and types of cells can still tell you a lot. On the other, I don't think you'd ever be able to understand the true depth of homology among vertebrates without looking deeply at the actual developmental processes at play in each of these tissues, and at the full network of connections between them in the adult.
(Incidentally, sciatrix, I hope it doesn't come across like I'm trying to explain things to you that you already know. I just went on an extended riff based on a minor quibble of phrasing and wanted to give additional context for anyone else reading. I know you follow these issues more closely than most people within neuroscience.)
posted by biogeo at 1:08 PM on January 28, 2019 [9 favorites]
But even if we could count and classify every cell, molecule and atom, we would still lack a satisfying explanation of its remarkable behavior. The brain is more than a thing; it’s a system. So much of intelligence is neither within the brain nor in its environment, but vibrating through the space in between.sciatrix, I agree with your reservations about the limitations of cell-counting, but I do think this is a valuable technique for comparative neurobiology. As an electrophysiologist, I'd love to be able to compare functional activity across species, but the difficulty of recording in even a single species is already so great that comparative work is daunting. Cell counting using Herculano-Houzel's technique is at least fast enough that true comparative work across a large phylogenetic tree is possible.
birds don't have a cortex as we understand it--that's a mammals thing. Birds do have an older analogous structure, but their brains are organized a little differently, and they do not rely on cortical surface area to get things done as mammals do. It's really interesting in that vein that bird brains are so much denser in terms of neurons.
I'd dispute the "older analogous structure" phrasing a bit, particularly the "older" part. I think you've given the generally held opinion in neuroscience, but most folks I know who work in bird brains don't agree with it. The avian pallium is indeed developmentally homologous to the mammalian pallium, which includes the cortex and parts of the amygdala. But I don't think the evidence supports the idea that it's phylogenetically older. Both the mammalian and the avian pallium are heavily derived from the ancestral state, which may have been similar to that in modern amphibians. Non-avian reptiles show a certain amount of derivative specialization in the pallium as well, but birds are really very different even from other reptiles. The pallium of snakes and lizards, for example, seems to be roughly laminar in structure (possibly more similar to the mammalian cortex, in fact!), while birds' are organized into more densely-packed nuclei. As far as I know, it's still an open debate whether the avian or mammalian forebrain more closely resembles the ancestral state, but probably both of them share a similar level of derived characteristics.
To the extent that cortex is special, my opinion is that it is so primarily because of the properties it derives from its position within neural circuitry. In mammals, there is what I would call a canonical forebrain circuit, such that every piece of cortex receives inputs from the thalamus, processes with a local microcircuit (what in vision would be called "columns" but that term seems to be loaded now so let's just say a microcircuit organized according to laminar connections within a local patch of cortex), and projects to the striatum and thalamus; striatum projects to the pallidum; and pallidum projects to thalamus and midbrain. These circuits seem to form a large set of interconnecting "loops" with heavy feedback and crosstalk. Now, what's interesting is the pallial amygdala of mammals also seems to preserve this structure. But if you look at the pallium of birds, the same canonical forebrain loop circuitry is also present! It's harder to see because of the way the tissue is structured, and because the striatum and pallidum actually intermingle during development to form a single tissue, but it's there. And in fact it looks like this canonical loop motif is conserved right across the entire vertebrate phylum. Even teleost fish seem to preserve it. (Side note, I learned somewhat recently that the teleost forebrain undergoes a weird "eversion" during development where it basically flips inside-out. Kind of blew my mind a bit. So even though their forebrain at first glance looks nothing like that of terrestrial vertebrates, it's actually got all of the parts wired up in the same way, just positioned inside-out. So weird. Also, let's ignore the proper placement of hagfish in the vertebrate family tree for now.)
Anyway, all this is to note that Herculano-Houzel's cell counting technique is essentially blind to all of this complexity. Which, as the article says, is both its strength and its weakness. On the one hand, gross changes in numbers and types of cells can still tell you a lot. On the other, I don't think you'd ever be able to understand the true depth of homology among vertebrates without looking deeply at the actual developmental processes at play in each of these tissues, and at the full network of connections between them in the adult.
(Incidentally, sciatrix, I hope it doesn't come across like I'm trying to explain things to you that you already know. I just went on an extended riff based on a minor quibble of phrasing and wanted to give additional context for anyone else reading. I know you follow these issues more closely than most people within neuroscience.)
posted by biogeo at 1:08 PM on January 28, 2019 [9 favorites]
LOL I just got to your appendix--no, I was totally hoping you or others in the field would weigh in! I'm delighted and about to settle down and fully read your comment and chew on it some.
posted by sciatrix at 1:09 PM on January 28, 2019 [2 favorites]
posted by sciatrix at 1:09 PM on January 28, 2019 [2 favorites]
(And for inside-baseball kinds of context, I'm primarily interested in changes of gene regulation in brains and how receptors and hormones communicate in the limbic system, which is why I don't have as much cortical context: I straight up don't care much about the cortex in my day-to-day reading.
Also, in my field, there's a definite tendency for folks doing neurotranscriptomic work or trying to work out how the brain changes its gene expression according to context to do things like grind up whole brains or massive sections of brains and try to work out the transcriptomic changes across the whole brain. Given how granular a brain is, especially on the level of protein expression, I have always thought this was doable but a little silly, since you'd swamp your signal unless you had a very, very good idea of what you were looking for. Which folks doing whole-brain transcriptomic work generally don't necessarily have; it tends not to be a technique people use when they have a candidate gene or protein in mind to start out with.
So that's a little bit of informing on where my skeptical perspective is coming from; I'm not a comparative neuroanatomist, which is really I think where this work shines, and I'm coming from a background in which people tend to cut a specific corner that involves... taking whole brains... and grinding them up... for further processing. I have no reason to suspect that is relevant at all to the kinds of work that Herculano-Houzel is doing, but it does inform my knee-jerk skepticism... which of course, as I have shown here before, is not always right. ;) )
posted by sciatrix at 1:14 PM on January 28, 2019 [2 favorites]
Also, in my field, there's a definite tendency for folks doing neurotranscriptomic work or trying to work out how the brain changes its gene expression according to context to do things like grind up whole brains or massive sections of brains and try to work out the transcriptomic changes across the whole brain. Given how granular a brain is, especially on the level of protein expression, I have always thought this was doable but a little silly, since you'd swamp your signal unless you had a very, very good idea of what you were looking for. Which folks doing whole-brain transcriptomic work generally don't necessarily have; it tends not to be a technique people use when they have a candidate gene or protein in mind to start out with.
So that's a little bit of informing on where my skeptical perspective is coming from; I'm not a comparative neuroanatomist, which is really I think where this work shines, and I'm coming from a background in which people tend to cut a specific corner that involves... taking whole brains... and grinding them up... for further processing. I have no reason to suspect that is relevant at all to the kinds of work that Herculano-Houzel is doing, but it does inform my knee-jerk skepticism... which of course, as I have shown here before, is not always right. ;) )
posted by sciatrix at 1:14 PM on January 28, 2019 [2 favorites]
Incidentally, there was a fairly provocative paper in Science a few years ago arguing that basal ganglia circuitry (the core of what I call the canonical forebrain circuit) is actually deeply homologous between arthropods and vertebrates, including a pallium-like tissue in arthropods sharing many of the same biochemical markers during development that characterize vertebrate pallium. (Strausfeld & Hirth 2013, academic paywall, Memail me if you need a copy for academic discussion.)
Note that "deep homology" refers to the idea that certain structures may be separately derived from a common ancestor, but some feature of the ancestral trait may have "preadapted" it in both cases. The easiest example being the wings of birds, bats, and pterosaurs. The common ancestor of all three was a stem amniote "reptile," lacking wings. However, certain features of the tetrapod forelimb make it suitable for evolution to shape into a wing given the right selective pressures, and it did so at least three separate times. So the wings of birds, bats, and pterosaurs are homologous as forelimbs, but they are not (shallowly) homologous as wings because the common ancestor lacked wings and there are essential features of their "wing-ness" that cannot be explained by common ancestry; e.g., bird wings are formed by shortening and fusing the bones of the hand and creating a flight surface by extending feathers from the arm, while bat wings are formed by shortening the bones of the arm and lengthening the bones of the hand and joining them with a membrane. But they are deeply homologous as wings, because certain properties of the ancestral trait (strong tetrapod bones, a phylogenetically plastic developmental program, etc.) are essential to explaining how the derived wings came to evolve in each case. I think of deep homology as sort of a combination of homology and convergence.
This is important for thinking about the forebrain because the common ancestor of arthropods and vertebrates (the "ur-bilaterian," also ancestral to molluscs) is typically thought not to have had a brain at all, so it's somewhat surprising to find that arthropods have a structure that developmentally, functionally, and topologically resembles what's thought to be a specialized component of the vertebrate brain. If true, it suggests there's something fundamental about the body plan of bilaterally organized animals that lends itself to brains being organized in a particular way for particular functions.
I think this is cool in part because of things like what the article mentions about bees, which are capable of surprisingly complex decision-making behaviors even with a tiny number of neurons. It suggests that by studying the small, relatively simple brains of bees, we actually have a good tool for studying the exact same types of circuitry that are important for complex cognition in humans.
posted by biogeo at 1:50 PM on January 28, 2019 [4 favorites]
Note that "deep homology" refers to the idea that certain structures may be separately derived from a common ancestor, but some feature of the ancestral trait may have "preadapted" it in both cases. The easiest example being the wings of birds, bats, and pterosaurs. The common ancestor of all three was a stem amniote "reptile," lacking wings. However, certain features of the tetrapod forelimb make it suitable for evolution to shape into a wing given the right selective pressures, and it did so at least three separate times. So the wings of birds, bats, and pterosaurs are homologous as forelimbs, but they are not (shallowly) homologous as wings because the common ancestor lacked wings and there are essential features of their "wing-ness" that cannot be explained by common ancestry; e.g., bird wings are formed by shortening and fusing the bones of the hand and creating a flight surface by extending feathers from the arm, while bat wings are formed by shortening the bones of the arm and lengthening the bones of the hand and joining them with a membrane. But they are deeply homologous as wings, because certain properties of the ancestral trait (strong tetrapod bones, a phylogenetically plastic developmental program, etc.) are essential to explaining how the derived wings came to evolve in each case. I think of deep homology as sort of a combination of homology and convergence.
This is important for thinking about the forebrain because the common ancestor of arthropods and vertebrates (the "ur-bilaterian," also ancestral to molluscs) is typically thought not to have had a brain at all, so it's somewhat surprising to find that arthropods have a structure that developmentally, functionally, and topologically resembles what's thought to be a specialized component of the vertebrate brain. If true, it suggests there's something fundamental about the body plan of bilaterally organized animals that lends itself to brains being organized in a particular way for particular functions.
I think this is cool in part because of things like what the article mentions about bees, which are capable of surprisingly complex decision-making behaviors even with a tiny number of neurons. It suggests that by studying the small, relatively simple brains of bees, we actually have a good tool for studying the exact same types of circuitry that are important for complex cognition in humans.
posted by biogeo at 1:50 PM on January 28, 2019 [4 favorites]
I have no reason to suspect that is relevant at all to the kinds of work that Herculano-Houzel is doing, but it does inform my knee-jerk skepticism... which of course, as I have shown here before, is not always right.
Well, I certainly think a certain level of skepticism is warranted here! Counting numbers of cortical neurons and giving that as the explanation for human intelligence is definitely simplistic. After all, we know cortex isn't the be-all and end-all of the brain or of complex behavior. For example, an awful lot of the primate cortex is devoted to vision, but in birds and many other animals, visual processing is distributed between the pallium and the optic tectum (equivalent to mammalian superior colliculus). Birds have excellent vision and can do impressive things like categorize paintings according to style or painter, so it's not clear that cortical/pallial neurons are the important thing for complex vision. If we count the neurons in primary visual cortex in humans, are we "overcounting" what's important for complex cognition? Or if we don't include the optic tectum in birds, are we "undercounting"? Counting cells alone has hard limits on what it can tell us.
posted by biogeo at 2:04 PM on January 28, 2019 [3 favorites]
Well, I certainly think a certain level of skepticism is warranted here! Counting numbers of cortical neurons and giving that as the explanation for human intelligence is definitely simplistic. After all, we know cortex isn't the be-all and end-all of the brain or of complex behavior. For example, an awful lot of the primate cortex is devoted to vision, but in birds and many other animals, visual processing is distributed between the pallium and the optic tectum (equivalent to mammalian superior colliculus). Birds have excellent vision and can do impressive things like categorize paintings according to style or painter, so it's not clear that cortical/pallial neurons are the important thing for complex vision. If we count the neurons in primary visual cortex in humans, are we "overcounting" what's important for complex cognition? Or if we don't include the optic tectum in birds, are we "undercounting"? Counting cells alone has hard limits on what it can tell us.
posted by biogeo at 2:04 PM on January 28, 2019 [3 favorites]
Cell counting is interesting, but my research in a previous life involved dissolving fresh embryonic brains (cortices for bulk preparations, dissected hippocampi because they image better - mostly) with papain, then re-plating them on glass (easier to visualize in 2D than 3D) to see how they re-form connections with one another; either in different baseline conditions or after gene replacement with targeted mutants.
On the gross level, there are pretty big structural differences even within mammals - you know how the human cortex looks "crinkly?" That's gyrification and mice have very little (but can be induced to develop by exogenous application of Fibroblast Growth Factor 2 during embryonic development.
posted by porpoise at 2:39 PM on January 28, 2019 [2 favorites]
On the gross level, there are pretty big structural differences even within mammals - you know how the human cortex looks "crinkly?" That's gyrification and mice have very little (but can be induced to develop by exogenous application of Fibroblast Growth Factor 2 during embryonic development.
posted by porpoise at 2:39 PM on January 28, 2019 [2 favorites]
Are mice with induced “crinkly” cortex any smarter than regular mice?
posted by monotreme at 4:04 PM on January 28, 2019
posted by monotreme at 4:04 PM on January 28, 2019
I’m so sorry Cortex. We’re not all this way #itgetsbetter
posted by um at 5:31 PM on January 28, 2019
posted by um at 5:31 PM on January 28, 2019
Metafilter: Let's ignore the proper placement of hagfish in the vertebrate family tree for now.
posted by chortly at 6:12 PM on January 28, 2019 [1 favorite]
posted by chortly at 6:12 PM on January 28, 2019 [1 favorite]
monotreme - not generally, but its hard to measure "intelligence" in rodents. Sure, they're smart but they're not smart- smart, or least in the way that the naked apes think of as intelligence.
It also appears that even with "crinkly" the corresponding neural connectivity that arose with human gyrification might not be there or analogous in the induced rodent gyrification.
posted by porpoise at 7:16 PM on January 28, 2019 [2 favorites]
It also appears that even with "crinkly" the corresponding neural connectivity that arose with human gyrification might not be there or analogous in the induced rodent gyrification.
posted by porpoise at 7:16 PM on January 28, 2019 [2 favorites]
But yeah - monotreme that was the exact first question that I asked.
I remember a departmental social where a "mentally disabled" mutant mouse model experiment was discussed to be written up as a paper for submission. During the presentation of (the full dump of) raw data, I noticed that these mutants were feeble in everything except Strength as they aged.
A correlation of that could be that this gene defect creates dumb but super strong mice.
Asking further - it turned out that it was just a developmental milestone in adolescent->juvenile mice's muscle development and inconsistent axes labeling.
posted by porpoise at 10:54 PM on January 28, 2019 [2 favorites]
I remember a departmental social where a "mentally disabled" mutant mouse model experiment was discussed to be written up as a paper for submission. During the presentation of (the full dump of) raw data, I noticed that these mutants were feeble in everything except Strength as they aged.
A correlation of that could be that this gene defect creates dumb but super strong mice.
Asking further - it turned out that it was just a developmental milestone in adolescent->juvenile mice's muscle development and inconsistent axes labeling.
posted by porpoise at 10:54 PM on January 28, 2019 [2 favorites]
How advanced and reliable is scRNA-seq based region calling in the brain? Does nuclear RNA give a decent signal in neurons?
posted by benzenedream at 11:26 PM on January 28, 2019
posted by benzenedream at 11:26 PM on January 28, 2019
Looks like you can get cell type resolution with snRNA: Single-nucleus and single-cell transcriptomes compared in matched cortical cell types
posted by benzenedream at 11:06 AM on January 29, 2019
posted by benzenedream at 11:06 AM on January 29, 2019
« Older Should I curtsey when I meet the Avocado Toast... | The Billy Graham Rule Goes Global Newer »
This thread has been archived and is closed to new comments
I have to wonder how important the specifically cortical structures are outside of a mammalian context, that's all.
posted by sciatrix at 9:46 AM on January 28, 2019 [10 favorites]